Concerted Action of Tenascin-C Domains in Cell Adhesion, Anti-Adhesion and Promotion of Neurite Outgrowth

Journal of Cell Science 110, 1513-1522 (1997) 1513 Printed in Great Britain © The Company of Biologists Limited 1997 JCS4412

Concerted action of tenascin-C domains in cell adhesion, anti-adhesion and promotion of neurite outgrowth

Doris Fischer1, Marianne Brown-Lüdi1, Therese Schulthess2 and Ruth Chiquet-Ehrismann1,* 1Friedrich Miescher Institute, PO Box 2543, CH-4002 Basel, Switzerland 2Biocenter, University of Basel, CH-4056 Basel, Switzerland *Author for correspondence (e-mail: [email protected])

SUMMARY

We used a new approach to identify domains of chicken When tenascin-C was added to the medium of fibroblasts tenascin-C required for interaction with cells. Instead of plated on fibronectin-coated wells, cell adhesion was expressing the parts of interest, we deleted them from an blocked by intact tenascin-C, but not by mutants missing otherwise intact tenascin-C molecule and scored for the con- the fibrinogen globe. In neurite outgrowth assays using comitant change in activity. As a starting point for all dorsal root ganglia, processes formed on all substrates mutant constructs we expressed the smallest naturally except on the mutant missing only the fibrinogen globe, occurring tenascin-C splice variant in vertebrate cells. The where the ganglia failed to adhere. The mutants missing the tenascin-C mutants had either deletions of all EGF-like fibronectin type III repeats allowed more rapid neurite repeats, all fibronectin type III repeats or of the fibrinogen outgrowth than all other tenascin-C variants and the globe. In double mutants the fibronectin type III repeats mutant consisting essentially of oligomerized EGF-like were deleted together with either the EGF-like repeats or repeats was as active a substrate for neurite outgrowth as the fibrinogen globe, respectively. All tenascin-C variants laminin. From the combined data, it is concluded that the assembled correctly to hexameric molecules of the expected activities of intact tenascin-C cannot be mimicked by inves- molecular characteristics. Intact tenascin-C and the mutant tigating domain by domain, but the concerted action of missing the fibrinogen globe did not promote adhesion of several domains leads to the diverse cellular responses. chick embryo fibroblasts, whereas both, the hexamers containing solely the fibrinogen globe or the EGF-like repeats Key words: Tenascin-C, Extracellular matrix, Cell adhesion, Neurite were adhesive substrates and even supported cell spreading. outgrowth, Recombinant protein

INTRODUCTION system, whereas tenascin-C, the longest known member of the family, seems to be the most versatile protein being transiently Extracellular matrix proteins have an important function in expressed during many processes in embryogenesis accompa- providing structural integrity to tissues, as well as in present- nied by cell migration and cell differentiation (for reviews see ing proper environmental clues for cell migration, growth and Erickson and Bourdon, 1989; Chiquet-Ehrismann et al., 1995). differentiation during development (for review see Adams and In the adult, tenascin-C is often expressed at sites of tissue Watt, 1993). Prominent members of non-collagenous extracel- remodeling (Mackie et al., 1988; Gatchalian et al., 1989; lular matrix proteins are laminins (Timpl and Brown, 1995), Chiquet-Ehrismann, 1993). It is therefore of great interest to fibronectin (Hynes, 1990) and tenascins. The tenascins form a investigate the function of tenascin-C in influencing the family of four members, namely tenascin-C, tenascin-R, behavior of different cell types that encounter this protein in tenascin-X, and tenascin-Y (for review see Erickson, 1993; vivo. Chiquet-Ehrismann, 1995). The tenascins have in common Because of its large size and multimeric nature it is not with fibronectin that they all contain a large number of possible to express the entire tenascin-C molecule in bacteria. fibronectin type III (FN III) repeats. In the case of fibronectin Therefore parts of the protein chains have been produced in several of these repeats have been shown to interact with several laboratories and have been used for structural as well integrins, the cellular extracellular matrix receptors (Hynes, as functional studies. The FN III repeats of tenascin-C are the 1992). In the case of the tenascins this is less clear. Very little most amenable parts for bacterial expression and such recom- is known yet about the function of tenascin-X and -Y, except binant tenascin-C fragments enabled the determination of the that they may be important for muscle integrity, since they are 3-dimensional structure of a single FN III domain (Leahy et highly expressed in the extracellular matrix of the connective al., 1992). Certain FN III repeats of tenascin-C have been tissue surrounding muscle fibers (Matsumoto et al., 1994; shown to promote or inhibit cell adhesion (Joshi et al., 1993; Hagios et al., 1996). Tenascin-R is restricted to the nervous Prieto et al., 1993; Götz et al., 1996; Dörries et al., 1996) and 1514 D. Fischer and others others to bind to heparin (Aukhil et al., 1993; Weber et al., overlap extension’ (Horton et al., 1989). We deleted the EGF-like 1995). It has, however, been demonstrated that although a repeats (nucleotides 795-2,004 of the EMBL data base entry single FN III repeat was able to promote cell adhesion, in the M23121), the fibronectin-type-III-repeats (nucleotides 2,005-4,875 of context of its adjacent repeat this activity was blocked (Leahy the EMBL data base entry M23121) and the fibrinogen domain et al., 1992; Joshi et al., 1993). Such discrepancies may not be (nucleotides 4,999-5,658 of the EMBL data base entry M23121), resulting in the plasmids pCTN EGF−, pCTN FN− and pCTN FB−. surprising considering the structural differences between a On the basis of these deletion mutants we constructed in the same way single FN III repeat and an assembly of four of them as deter- double deletion mutants, namely tenascin-C variants lacking two mined by X-ray crystallography (Leahy et al., 1996). In domain types. We deleted the fibronectin type III repeats from the fibronectin, the loop containing the famous cell binding peptide plasmid pCTN EGF− (nucleotides 2,005-4,875 of the EMBL data base Arg-Gly-Asp was flexible when the corresponding FN III entry M23121) as well as from the plasmid pCTN FB−, respectively, repeat was expressed alone (Main et al., 1992); however, this resulting in the plasmids pCTN EFN− and pCTN FF−. All constructs loop was shown to acquire a fixed position in a protein were analyzed by sequencing the PCR-modified regions and fragment assembled from four FN III repeats (Leahy et al., subcloned in the eukaryotic expression vector pCDNAI/NEO (Invit- 1996). The FN III repeats, although being independently rogen, San Diego, USA) as described (Fischer et al., 1995). The domain structure of the corresponding recombinant proteins is repre- folding protein units, interact with each other within one sented in Fig. 1. protein chain revealing distinct and fixed relative positions To express the recombinant tenascin-Cs, HT1080 human fibrosar- (Leahy et al., 1996). Clearly such structural constraints will coma cells (American Tissue Culture Collection, Maryland, USA) affect functional sites and, by expressing single domains or were stably transfected with the appropriate constructs and the even fragments thereof in bacteria, the danger of either loosing proteins were isolated from the conditioned medium by affinity chro- active sites or uncovering cryptic activities is imminent. matography as described (Fischer et al., 1995). The purification Besides the FN III repeats, tenascin-C contains cysteine-rich procedure included the removal of fibronectin by gelatin-agarose domains, such as the N-terminal part involved in disulfide- chromatography as well as the removal of endogenous human linking two tenascin-C trimers into a hexameric molecule, the tenascin-C by passing the conditioned medium over an anti-human tenascin-C coupled Sepharose column before isolating the recombi- heptad repeats flanked by cysteines responsible for the trimer- nant chicken tenascin-C by affinity chromatography to chicken- ization, and the many EGF-like repeats as well as the C- specific anti-tenascin-C monoclonal antibody columns. In contrast to terminal globe homologous to the globular parts of γ- and β- the isolation of the majority of recombinant tenascin-Cs, using a fibrinogens. Therefore, bacterial expression of properly folded Sepharose 4B column (Pharmacia, Uppsala, Sweden) coupled with tenascin-C fragments is often difficult. We therefore decided anti-chick tenascin-C anti-TnM1 (Chiquet and Fambrough, 1984), we to use tenascin-C proteins expressed in eukaryotic cells for our used in the case of TN-EGF− a Sepharose-4B column coupled with anti-chick tenascin-C anti-Tn68 (Chiquet-Ehrismann et al., 1988) and functional studies to determine the structural requirements for − cell adhesion, anti-adhesion and neurite outgrowth promoting in the case of TN-EFN a Sepharose 4B column coupled with anti- activity. A similar approach has been used previously by chick tenascin-C anti-Tn60 (Pearson et al., 1988). The purified recombinant proteins were analyzed by SDS-PAGE under reducing con- Chung et al. (1996) to prove the focal adhesion disassembling ditions and by solid phase ELISA (Chiquet-Ehrismann et al., 1988) activity of the alternatively spliced fibronectin type III repeats using anti-chick tenascin-C anti-Tn60, TnM1, Tn68 and Tn4 (Fischer of tenascin-C. et al., 1995) and peroxidase-labeled goat anti-mouse IgG (CAPPEL In our experiments we decided to delete entire assemblies of Organon Teknika, Turnhout, Belgium). For transmission electron given domain types from an intact hexameric chicken tenascin- microscopy proteins were processed as described (Chiquet-Ehrismann C protein in order to score for the loss (or gain) of function et al., 1988). accompanied by the respective deletion. All our proteins were Short term cell adhesion assays and immunological isolated from the conditioned media of stably transfected cells, procedures and they all revealed proper assembly into hexameric Fibronectin used for the cell adhesion assays was isolated from bovine molecules as judged by electron microscopy. For the first time serum (Gibco-BRL) by affinity chromatography using a gelatin- we could on the one hand reveal cell adhesion and neurite agarose column (Sigma, Buchs, Switzerland). After washing the outgrowth activity associated with the EGF-like repeats, and column with phosphate-buffered saline (PBS), bound fibronectin was on the other hand show that deletion of the fibrinogen globe eluted with 4 M urea in PBS. Fibronectin-containing fractions were caused the complete loss of inhibition of cell adhesion to dialyzed overnight against PBS and stored frozen at −70¡C. For fibronectin, which is a prominent activity of tenascin-C coating, fibronectin as well as recombinant tenascin-Cs were diluted (Chiquet-Ehrismann et al., 1988). to 50 nM in PBS containing 0.01% Tween-20 (Fluka, Buchs, Switzer- land). Concentrations of tenascin-Cs were calculated for hexameric molecules (e.g. in the case of TN-190, 50 nM = 43 µg ml−1). The proteins were allowed to adsorb overnight at 4¡C either to 60-well MATERIALS AND METHODS tissue culture plates (NUNC, Roskilde, Denmark) for counting of cells or to glass coverslips for immunofluorescence staining. Equal coating Construction, expression and isolation of recombinant efficiency between tenascin-C variants was verified by immunoblot tenascin-C variants analysis. Coated proteins were removed by adding sample buffer for All recombinant tenascin-C molecules represent artificial deletion SDS-PAGE and aliquots were analyzed by immunoblotting using anti- variants of the smallest naturally occurring splice variant of chick Tn60, equally recognizing all tenascin-C variants. Before cells were tenascin-C, tenascin-190 (TN-190). The appropriate tenascin-C plated the coated substrates were blocked with 1 mg ml−1 heat construct, pCTN 190, is encoded by nucleotides 1-3,360 and 4,180- denatured bovine serum albumin (BSA) in PBS for 1 hour at room 5,658 of the EMBL data base entry M23121 (cf. Fischer et al., 1995). temperature and finally washed with PBS. For antibody inhibition To construct tenascin-Cs lacking precisely one domain type, pCTN experiments 60-well plates were either coated with TN FF− or 190 was deleted using a PCR-based method called ‘splicing by fibronectin and blocked as described. Wells were either incubated with Active domains in recombinant hexameric tenascin-C 1515 a polyclonal anti-fibronectin antiserum (Ehrismann et al., 1981) diluted 1:50 in PBS or the monoclonal anti-tenascin-C anti-M1 (40 A 4 µg ml−1) followed by a polyclonal goat anti-mouse IgG (CAPPEL Organon Teknika, Turnhout, Belgium) diluted 1:50 in PBS, respec- 68 tively, before the final PBS wash. All antibody incubations were done for 1 hour at room temperature. Chick embryo fibroblasts prepared from the skin of 11-day-old chick embryos were used for cell adhesion assays 2-3 days after subculture. Cells were trypsinized and washed once in DMEM (Gibco-BRL) containing 10% fetal calf serum (Gibco-BRL) and once in serum-free medium and finally resuspended × 5 −1 µ in DMEM at a concentration of 2 10 cells ml . Then 10 l samples M1 were added to each well of the 60-well plate or 30 µl samples were added to each coverslip, respectively. After incubation at 37¡C for 30 minutes cells were fixed with 3.7% formaldehyde and either stained with Crystal Violet (0.1% in H2O; Sigma, Buchs, Switzerland) or with TRITC (Texas Red isothiocyanate)-phalloidin (diluted 1:50 in PBS; Sigma, Buchs, Switzerland). Crystal Violet stained cells were photographed and counted. Phalloidin stained samples were examined 60 under epifluorescence using a Zeiss Axiophot microscope. To examine the influence of tenascin-C on cell adhesion to fibronectin 60-well plates were coated with 50 nM fibronectin and blocked as described above. Tenascin samples at a concentration of 100 nM were premixed in a 1:1 ratio with a cell solution of 4×105 cells ml−1 in DMEM and added to the wells. Cell attachment and staining was carried out as described above for the cell attachment assay. Neurite outgrowth assays Glass coverslips were washed in 70% ethanol, wiped with cotton gauze and air dried. Coverslips were incubated with 70 µl samples of mouse laminin-1 isolated from mouse Engelbreth-Holm-Swarm tumor as described (20 µg ml−1; Paulsson et al., 1987) or recombinant tenascin-C diluted to 50 nM in PBS containing 0.01% Tween-20 (Fluka, Buchs, Switzerland) overnight at 4¡C in a closed humidified chamber. The coverslips were washed twice in PBS and placed in wells of a 24-well culture dish (Falcon, Becton-Dickinson, Franklin- Lake, USA) containing L15 medium (Gibco-BRL). Sensory ganglia from 6-day-old chick embryos were dissected and cultured as described (Wehrle-Haller and Chiquet, 1993). The ganglia were cultured for 24 hours and photographed under phase contrast at different time points. For antibody inhibition experiments tenascin-C coated glass coverslips were blocked as described and incubated with the monoclonal anti-chick tenascin-C anti-M1 (20 mg ml−1) followed by a polyclonal goat anti-mouse IgG (1:50 diluted in PBS) before ganglia were plated.

Fig. 1. Domain structure and characterization of the chick tenascin-C RESULTS variants by SDS-PAGE and ELISA. (A) Models of one subunit of each hexameric recombinant tenascin-C variant. Each subunit is Production of the tenascin-C deletion variants depicted as a linear array from the N- (bottom) to the C- (top) terminus showing the central domain (segment of circle), heptad All our recombinant tenascin-C variants were produced in repeats (wavy line), EGF-like repeats (diamonds), FN-III repeats stably transfected HT1080 cells and were purified from the (rectangles) and the fibrinogen globe (circle), and the designation of conditioned medium by affinity chromatography to immobi- each protein variant is indicated. The location of the epitopes of lized anti-tenascin-C monoclonal antibodies as described in monoclonal anti-tenascin-C antibodies anti-Tn4, anti-Tn68, anti- Materials and Methods. Models of the different tenascin-C TnM1 and anti-Tn60 are indicated by numbered triangles. (B) Below proteins including the positions of the epitopes of the mono- each model the corresponding purified recombinant tenascin-C clonal antibodies used for this study are shown in Fig. 1. We protein was run on a 6.75% SDS-polyacrylamide gel after reduction expressed the smallest naturally occurring intact recombinant and stained with Coomassie Blue. To the right of the gel, positions of tenascin-C containing the central oligomerization domain, the the molecular mass markers are indicated in kDa. (C) A solid phase ELISA with antibodies directed against the different domain types of 13 EGF-like repeats, eight FN III repeats and the fibrinogen tenascin-C confirmed correct assembly of all variants. Below each globe. We termed this variant TN 190, owing to its subunit − lane of the SDS-gel the reaction pattern of the corresponding variant molecular mass of 190 kDa. In the tenascin-C variant TN EGF with the monoclonal antibodies indicated (numbered triangles) is − , the EGF-like repeats have been deleted, in TN FN , the FN shown. All recombinant tenascin-C variants show the expected sizes III repeats have been removed and in TN FB− the fibrinogen upon SDS-PAGE and are lacking the desired domain types as globe is missing. In two further constructs two types of visualized by the loss of the respective epitope(s). 1516 D. Fischer and others

Fig. 2. Electron microscopy of recombinant tenascin-C proteins and their corresponding domain models. All recombinant deletion variants assemble into an authentic hexameric tenascin-C molecule with expected structural features and dimensions (cf. structural models in the right panel). Bar, 50 nm. domains have been removed, namely TN FF− is missing both with mostly round cells was seen on TN EGF− and TN FN−. the FN III repeats as well as the fibrinogen globe, whereas TN In contrast, on both TN FF− and TN EFN, cells not only EFN− is devoid of EGF-like repeats as well as the FN III adhered but were remarkably well spread. In comparison to the repeats. Below each model a sample of the purified proteins is fibroblasts plated on fibronectin, stress fibers were not as pro- shown after SDS-PAGE (Fig. 1B). Since all constructs contain nounced on the tenascin-C substrates, and many of the spread the central oligomerization domain, they all carry the epitope cells on TN FF− and TN EFN− showed actin-rich protrusions of anti-Tn60. This is evident from the ELISA assay shown in instead of stress fibers. Quantitative analysis of cell adhesion Fig. 1C. It shows that all proteins contain the epitope of the to these substrates is shown in Fig. 4. It was surprising to see central globe recognized by anti-Tn60, whereas each variant cell adhesion and spreading on TN FF−, since earlier experi- missing domains is indeed devoid of the corresponding ments using bacterial expression proteins showed an anti- epitope(s) and thus is not recognized by the respective adhesive behavior of tenascin-C fragments containing EGF- antibody. All tenascin-C variants have been analyzed by like repeats (Spring et al., 1989; Scholze et al., 1996; Dörries electron microscopy after rotary shadowing. As can be seen in et al., 1996). In order to confirm specific adhesion to the EGF- Fig. 2, they all show the expected dimensions and structural like repeats in TN FF−, we compared cell adhesion to features and occur as hexameric structures. fibronectin with that to TN FF− (Fig. 5). Whereas adhesion to TN FF− was completely blocked by the monoclonal antibody Fibroblast adhesion to the tenascin-C variants recognizing the EGF-like repeats of tenascin-C but not by anti- We tested chick embryo fibroblast adhesion to all tenascin-C fibronectin, the opposite was the case for adhesion to variants and compared their cell shape and actin cytoskeleton fibronectin. We then investigated whether the expression of the to that of cells plated on fibronectin or on uncoated glass cov- C-terminally located EGF-like repeats in TN FF− resulted in erslips. As can be seen from the phalloidin-stained cells shown the exposure of cryptic adhesive sites normally not present in in Fig. 3, no adhesion above background could be detected to intact tenascin-C, or whether the deletion of the rest of the intact recombinant TN 190 or TN FB−. Moderate adhesion tenascin-C molecule in TN FF− resulted in the loss of the anti- Active domains in recombinant hexameric tenascin-C 1517

attachment assays (Chiquet-Ehrismann et al., 1988). In order to identify the structural requirements for this activity of tenascin-C we tested our deletion mutants for their anti- adhesive activity. As shown in Fig. 5, chick embryo fibroblast adhesion to fibronectin was inhibited by the addition of full length recombinant TN 190. Whereas TN EGF− was still inhibitory, deletion of the fibrinogen globe in TN FB− completely abolished the anti-adhesive activity of soluble tenascin- C. The same loss of anti-adhesive activity by added TN FB− was also observed when the cells were plated on TN FF− (Fig. 5). Quantitative analysis of anti-adhesive activity by all tenascin mutants for cells plated on fibronectin is shown in Fig. 6. The addition of TN 190 reduced cell adhesion to fibronectin by over 80%. Deletion of the EGF-like repeats in TN EGF− reduced the anti-adhesive activity to some extent, while deletion of the FN III repeats in TN FN− lead to an even greater reduction. Both proteins, however, still reduced cell adhesion by 30-55%. In contrast, deletion of the fibrinogen globe in TN FB− completely abolished the anti-adhesive activity of tenascin-C. Also the second mutant missing the fibrinogen globe, namely TN FF−, did not considerably inhibit cell adhesion to fibronectin. Although removal of the fibrinogen globe from an otherwise intact tenascin-C molecule completely abolished its inhibitory activity, this function could not be recovered in the small tenascin-C mutant consisting essentially of a hexamer of fibrinogen globes (TN EFN−). Neurite outgrowth from DRGs is amplified on TN FF− Tenascin-C has previously been shown to promote neurite outgrowth from dorsal root ganglia (Wehrle and Chiquet, 1990). We therefore decided to map the active site(s) mediating neurite formation using our tenascin-C variants. In Fig. 7 two time points showing the emergence of processes from DRGs 8 hours and 24 hours after plating are shown. Outgrowth on TN 190 is not yet visible after 8 hours in culture. It is delayed in comparison to laminin and after 24 hours the neurites are much more Fig. 3. Comparison of cell adhesion and morphology in response to recombinant tenascin-C substrates and fibronectin. Chick embryo fasciculated on TN 190. This is in agreement with the observa- fibroblasts were plated on glass coverslips which were uncoated (−), tions reported using tenascin-C purified from chick embryo coated with fibronectin (Fn) or coated with the recombinant tenascin- fibroblast conditioned medium (Wehrle-Haller and Chiquet, C variants indicated. Cells are shown after staining with TRITC- 1993). Deletion of the EGF-like repeats caused a reduction of labeled phalloidin to label the actin cytoskeleton. The majority of the neurite outgrowth and a pronounced fasciculation, as evident cells on glass, TN 190, TN EGF−, TN FN− and TN FB− are round, from DRGs plated on TN EGF−. In contrast, deletion of the FN whereas the cells on fibronectin, TN FF− and TN EFN− have spread. III repeats in TN FN− resulted in an increased number of In contrast to the cells on fibronectin which show prominent actin- neurites emanating from the DRGs already 8 hours after plating. rich stress fibers, the cells on TN FF− and EFN− tend to have actin- − µ A similar behavior was observed on TN EFN missing both the rich protrusions and fewer stress fibers. Bar, 100 m. FN III and the EGF-like repeats, and after 24 hours in cultures the neurites tended to be even longer than on TN 190. adhesive activity normally overriding the effect of the EGF- Absolutely no neurite outgrowth was observed on TN FB−, like repeats. We plated cells on wells coated with TN FF− in since none of the DRGs were able to attach to this substrate. the presence of other tenascin-C variants in the medium. As Surprisingly, however, the most neurite outgrowth promoting can be seen in Fig. 5, the addition of TN 190 or TN EGF− to variant was the mutant containing the EGF-like repeats but the medium resulted in a loss of cell adhesion to TN FF−. It missing the FN III domains as well as the fibrinogen globe. The therefore seems likely that intact tenascin-C counteracts the neurites were highly branched and grew out very rapidly on TN adhesion-promoting activity of the EGF-like repeats both in an FF−. No difference was visible between this tenascin-C intramolecular (cis-acting) as well as intermolecular (trans- substrate and laminin (Fig. 7). To prove the specificity of this acting) manner. effect, we confirmed that the neurite-promoting activity of TN FF− was inhibited by the monoclonal anti-TnM1 against its The fibrinogen globe is necessary to inhibit cell EGF-like repeats (Fig. 7). From the combined data, it seems adhesion to fibronectin that both EGF-like repeats and the fibrinogen globe of tenascin- It has been found earlier that the addition of tenascin-C to the C have independent neurite-promoting activity, which is atten- medium inhibits cell adhesion to fibronectin in short term cell uated by the neighboring FN III domains in the same molecule. 1518 D. Fischer and others

Fig. 4. Analysis of cell attachment and 700 round cells 500 A round cells B spreading on recombinant tenascin-C spread cells spread cells substrates. (A) Chicken embryo 600 400 ﬁbroblasts were plated in culture wells coated with BSA blocking solution 500 (BSA) or a solution containing 50 nM of 300 400 field

the different recombinant tenascin-C / variants indicated and incubated for 30 300 200 cells minutes. The cells were fixed, stained, cells / field photographed and counted, 200 discriminating between round (white 100 bars) and spread cells (grey bars). TN- 100 190 and the deletion mutant lacking the − 0 0 1 2345 12345 12345 fibrinogen domain (TN-FB ) did not BSA 190 EGF − FN − FB − EFN − FF − FN − EFN − FF − show cell-adhesive activity above substrates substrates background (BSA) and very few cells adhered to TN EGF−. Cells did adhere to TN FN− but remained round, while on TN EFN− and TN FF− the majority of the attached cells had spread. (B) The same experiment as described in A was performed using a series of 3-fold dilutions of the coating solution of the tenascin-C variants indicated starting at 50 nM (1) and ending at 0.6 nM (5). These cell adhesion experiments have been repeated eight times always confirming about 80-90% of input cells to adhere to TN FF−, the majority of which showed cell spreading. Also cells on TN EFN− were spread, while the cells on Tn FN− remained round in each case. Cells were scored as spread when they showed clearly visible protrusions and process formation.

DISCUSSION

In this manuscript we present data obtained from a novel approach taken to analyze the mechanism of action of tenascin- C in cell adhesion and neurite outgrowth. Instead of using antibody perturbation of intact tenascin-C or the bacterial expression of small fragments of tenascin-C, we tested intact hexameric tenascin-C and its deletion mutants produced by a eukaryotic expression system. This method has proven to be particularly useful for the study of extracellular matrix proteins, since they are difficult to purify from tissues without destroying their structure through highly denaturing extraction procedures. Furthermore, they are large multidomain proteins which are difficult to produce in bacteria without compromis- ing a loss of proper folding of the protein. Similar methods have been applied successfully for the production of mouse (Fox et al., 1991) and human (Mayer et al., 1995) nidogen, fibulin-1 (Sasaki et al., 1995), agrin (Gesemann et al., 1995), parts of perlecan (Schulze et al., 1995; Chakravarti et al., 1995) and the α-chain short arm of laminin-1 (Colognato-Pyke et al., 1995). The investigation of fibronectin-fibronectin interaction was only possible through the use of eukaryotic expression of intact molecules and deletion mutants, since this interaction required the presence of N- as well as C-terminal parts of the molecule in addition to the natural dimeric structure (Schwarzbauer, 1991). In the case of tenascin-C, we have previously shown that contradictory results were obtained concerning heparin-binding activity when bacterially expressed

Fig. 5. Inhibition of cell adhesion to TN FF− and fibronectin by antibodies and tenascin-C variants, respectively. To prove the specificity of cell adhesion to the respective substrates, wells coated with TN FF− (FF−) or fibronectin (Fn) were incubated with anti- TnM1 against the EGF-like repeats of tenascin-C (+αTN) or by anti- fibronectin (+αFn) before plating the cells. Whereas αTN inhibited cell adhesion to TN FF− and αFn interfered with cell attachment to fibronectin, the addition of TN 190 (+190) or TN EGF− (+EGF−) inhibited cell adhesion to both substrates while the addition of TN FB- (+FB−) had no effect. Bar, 100 µm. Active domains in recombinant hexameric tenascin-C 1519

is necessary for the initial interaction of tenascin-C with cells, thereby enabling other domains of the molecule to act on the 80 cells in an anti-adhesive manner. However, when this globe is (%) missing the cells are not able to recognize the added tenascin- Fn

to C. Concerning neurite outgrowth, there are some tenascin-C 60 mutants that allow a faster response than intact tenascin-C. These are the three mutants missing the FN III repeats. Thus, the presence of the FN III repeats seems to suppress neurite

adhesion 40 outgrowth. In contrast, the EGF-like repeats and the ﬁbrinogen cell globe have a high intrinsic neurite promoting activity which is of

20 attenuated in the presence of adjacent FN III repeats. Again the tenascin-C missing the fibrinogen globe stands out as the only completely inactive substrate. This is due to the inability of the inhibition 0 − − − − − dorsal root ganglia to adhere to this mutant at all, and as a con- TN 190 EGF FN FB EFN FF sequence neurite outgrowth is precluded. media additions Many attempts have been made previously to dissect the Fig. 6. Activity of soluble recombinant tenascin-C variants in functions of tenascin-C. In earlier work, proteolytic fragments inhibiting cell adhesion to fibronectin. Cells were premixed with the and monoclonal antibodies against mapped epitopes were used recombinant tenascin-C mutants indicated (50 nM) and plated on to inhibit certain activities of tenascin-C. It was found by frag- fibronectin-coated wells. After an incubation of 30 minutes cells mentation of tenascin-C (cytotactin) into pools of peptides that were fixed, stained and counted. The number of cells attaching to cells adhered to the C-terminal fragments and not to the N- fibronectin in the absence of any tenascin-C was set to 100% cell attachment (or 0% inhibition, respectively). Cell adhesion in the terminal ones (Friedlander et al., 1988). Therefore a cell- presence of the tenascin-C variants is presented as percentage binding site was postulated in the C-terminal part of the inhibition of cell adhesion to fibronectin. This experiment was molecule (Friedlander et al., 1988). This site could coincide repeated five times, confirming the qualitative differences between with the cell binding site on the fibrinogen globe mapped again the activities of the tenascin-C variants. in the present study. In another study it was shown that tenascin-C interfered with cell adhesion to fibronectin, and that this activity could be blocked by monoclonal antibody anti- protein fragments are compared with recombinant tenascin-C Tn68 which seemed to bind to the terminal globe of the produced in vertebrate cell cultures. Bacterially expressed FN tenascin-C arms or nearby (Chiquet-Ehrismann et al., 1988). III repeats showed binding to heparin (Aukhil et al., 1993; This antibody was postulated to mask a site for interaction with Weber et al., 1995), but the same region in the context of a the cells on the terminal globe (Chiquet-Ehrismann et al., protein expressed in eukaryotic cells did not show any heparin- 1988). In the meantime it is known that the epitope of anti- binding activity anymore (Fischer et al., 1995). Tn68 is located on FN III repeat 7 and not on the fibrinogen In our present study we focused on cell adhesion, anti- globe (Spring et al., 1989), but nevertheless the bound antibody adhesion and neurite outgrowth activity of tenascin-C. We cor- obviously comes to lie side by side with the fibrinogen globe related the loss or gain of function of tenascin-C mutant as was observed in the electron micrographs and thus may ster- proteins with the absence of certain domain types. An overview ically block its function. Therefore, our present finding that summarizing our results is shown in Fig. 8. This table shows removal of the fibrinogen globe from an otherwise intact the functional characteristics of each tenascin-C variant con- tenascin-C molecule prevents cells from reacting to the added cerning cell adhesion, cell spreading, inhibition of adhesion to tenascin-C is supported by this early work. Further agreement fibronectin, and neurite outgrowth promotion in a qualitative with the present results is seen in experiments with proteolytic way. When presented as substrates, poor cell adhesion is tenascin-C fragments (Chiquet et al., 1991). Pepsin-resistant observed on intact recombinant tenascin-C and on the mutant fragments from the N-terminal part of tenascin-C were shown missing the fibrinogen globe. Deletion of either the EGF-like to be inactive in inhibiting cell adhesion to fibronectin, whereas repeats or the FN III repeats render tenascin-C more adhesive. partial activity could be recovered in fragments encompassing In contrast, mutants with only EGF-like repeats or with the fib- the fibrinogen globe plus several FN III repeats, again pointing rinogen globe alone are very adhesive substrates for cells and towards the importance of the fibrinogen globe for interaction even promote cell spreading. We conclude from this that cells with cells (Chiquet et al., 1991). are able to interact with both, the EGF-like repeats as well as Antibody inhibition experiments have been used to localize with the fibrinogen globe, but that the response of the cells is regions within the tenascin-C molecule important for neurite governed by the context in which these domains are presented. outgrowth promotion. It was found that anti-TnM1 as well as For the different variants, there seems to be an inverse rela- anti-Tn68 partially inhibited neurite outgrowth of peripheral tionship between their cell adhesion activity and the degree by neurons, implying a role for both the EGF-like repeats as well which they inhibit adhesion to fibronectin. The variants making as the C-terminal part of tenascin-C (Chiquet and Wehrle- up poor adhesion substrates are the ones that strongly inhibit Haller, 1994). While these results fit well with our present cell adhesion to fibronectin when added to cells in solution and observations, there are other studies which seem less in vice versa. The tenascin-C variant missing the fibrinogen globe agreement possibly due to the use of very different types of appears to be an exception. It is a poor substrate for cell neuronal cells and assay conditions. Using neurons of the adhesion, but nevertheless is not capable of inhibiting cell central nervous system, it was found that a monoclonal adhesion to fibronectin. We postulate that the fibrinogen globe antibody reacting in the region of the sixth constant FN III 1520 D. Fischer and others

Fig. 7. Neurite outgrowth on recombinant tenascin-C variants. Dorsal root ganglia from 6-day chick embryo fibroblasts were cultured for 24 hours on coverslips coated with laminin (LN), TN 190 (190), and the different recombinant tenascin-C mutants indicated. Ganglia were photographed after 8 hours (8 h) and 24 hours (24 h) in culture. To prove that the neurite outgrowth on TN-FF− was directly dependent on the EGF- like repeats present in this variant, coverslips coated with TN-FF− were further incubated with anti-TnM1 before plating the ganglia (FF−/+M1). This resulted in a dramatic reduction in neurite outgrowth. All tenascin-C variants promoted neurite outgrowth except for TN FB− which did not allow adhesion of the explants. The variants containing FN III repeats plus the fibrinogen globe showed a delayed outgrowth (TN 190, TN EGF−) in comparison to the variants containing EGF-like repeats and/or the fibrinogen globe but no FN III repeats (TN FN−, TN EFN−, TN FF− ). TN FF− was the most active substrate, indistinguishable from laminin. This experiment was repeated eight times and a total of about 50 ganglia per type of substrate was recorded. The differences reported for the examples shown were consistently observed for the majority of the ganglia. Bar, 200 µm. repeat or the extra repeat D was able to inhibit neurite Prieto et al. (1992) also found anti-adhesive activity for bac- outgrowth (Lochter et al., 1991; Husmann et al., 1992). By terially expressed EGF-like repeats, as well as adhesive activity other antibody inhibition studies, a cell binding site for PC12 for the FN III repeats 2-6, 3 and the fibrinogen globe. Another cells was inferred in the EGF-like repeats 3-5 and the FN III report confirms the adhesive activity of the fibrinogen globe repeats 3-4, while neuroblastoma cells were inhibited to adhere (Joshi et al., 1993) and many different investigations support to tenascin-C by anti-Tn68 against FN III repeat 7 (Husmann Cell Inhibition Neurite et al., 1995). Tenascin-C variants adhesion of adhesion outgrowth Many of the more recent publications have made use of to Fn 8 h 24 h recombinant tenascin-C fragments expressed in bacteria. In the TN 190 first such study Spring et al. (1989) found that a cell-adhesion TN EGF site exists in fusion proteins encompassing repeat 7, while a TN FN fusion protein containing the EGF-like repeats was anti- TN FB adhesive. In a later report (Prieto et al., 1992) showed that the cell-adhesive activity of repeat 7 was only recognized by cells TN FF that were extensively trypsinized in the presence of EDTA. TN EFN This fits well with our own finding that the ‘active site’ within repeat 7 could be tracked down to a heptapeptide of the Fig. 8. Summary of the functional characteristics of recombinant tenascin-C variants. Each variant is scored for promotion of cell sequence GLVVMNI(T), a hydrophobic sequence encompass- − ᭺ ing a potential carbohydrate attachment site (Spring, 1991). adhesion by , + or ++ and the cell shape is indicated by if the majority of the cells were round and by ᭝ if the majority of the Thus this originally postulated adhesion site could have been attached cells were spread. Activities in inhibiting adhesion to caused by a twofold artifact created by hydrophobic interac- fibronectin (Fn) or the promotion of neurite outgrowth at two time tion of denuded cell membranes with a lipophilic peptide, points, namely 8 hours (8 h) or 24 hours (24 h) after plating are which in native tenascin-C might not be accessible because it indicated by −, +, ++, or +++. The scores are based on the results is glycosylated or not exposed on the surface of the molecule. presented in Figs 4, 6 and 7. Active domains in recombinant hexameric tenascin-C 1521 the existence of a cell binding site in isolated FN III repeat 3 cellular responses are governed by the combination of domains (Prieto et al., 1993; Yokosaki et al., 1994; Schnapp et al., 1995; recognized. Scholze et al., 1996; Joshi et al., 1993). However, concerning the importance of the RGD sequence present in this repeat and We thank Drs M. Chiquet, J. Engel, B. Rubin and S. Schenk for whether adhesion is inhibitable by RGD-peptides or anti-β1 critical reading of the manuscript and helpful suggestions. We are integrin antibodies, no consensus has been obtained. Further- grateful to I. Obergfoell for the help with the photography. more, it was shown by Joshi et al. (1993) that coexpression of the FN III repeat 3 with an adjacent one reduces its adhesive activity. Neuron adhesion or neurite outgrowth was shown to REFERENCES be promoted by bacterial expression proteins of FN III repeats Adams, J. C. and Watt, F. M. (1993). Regulation of development and 3 and 6 (Phillips et al., 1995), 1-3 (Scholze et al., 1996), 1-3 differentiation by the extracellular matrix. Development 117, 1183-1198. and the extra repeats (Götz et al., 1996), 1-2, 6-8, the extra Aukhil, I., Joshi, P., Yan, Y. and Erickson, H. P. (1993). Cell- and heparin- repeats and the EGF-like repeats (Dörries et al., 1996). Many binding domains of the hexabrachion arm identified by tenascin expression of the expression proteins have been shown to be anti-adhesive proteins. J. Biol. Chem. 268, 2542-2553. Chakravarti, S., Horchar, T., Jefferson, B., Laurie, G. W. and Hassell, J. R. for neuronal cells. These were the EGF-like repeats, FN III 4 (1995). Recombinant domain III of perlecan promotes cell attachment and 5 as well as the extra repeats (Scholze et al., 1996), the through its RGDS sequence. J. Biol. Chem. 270, 404-409. extra repeats and the EGF-like repeats (Kiernan et al., 1996), Chiquet, M. and Fambrough, D. M. (1984). Chick myotendinous antigen. I. the EGF-like repeats, FN III 2,4 and the extra repeats A1,2,3 A monoclonal antibody as a marker for tendon and muscle morphogenesis. J. (Götz et al., 1996) and finally again the EGF-like repeats Cell Biol. 98, 1926-1936. Chiquet, M., Vrucinic Filipi, N., Schenk, S., Beck, K. and Chiquet- (Dörries et al., 1996). It has to be kept in mind that most exper- Ehrismann, R. (1991). Isolation of chick tenascin variants and fragments. A iments made use of different cell types as well as of different C-terminal heparin-binding fragment produced by cleavage of the extra assay conditions. It is therefore very difficult to find a common domain from the largest subunit splicing variant. Eur. J. Biochem. 199, 379- consensus between the studies using bacterially expressed 388. Chiquet, M. and Wehrle-Haller, B. (1994). Tenascin-C in peripheral nerve tenascin-C fragments. morphogenesis. Persp. Dev. Neurobiol. 2, 67-74. What seems to be confirmed is an anti-adhesive effect by the Chiquet-Ehrismann, R., Kalla, P., Pearson, C. A., Beck, K. and Chiquet, M. bacterially expressed EGF-like repeats. This is in contrast to (1988). Tenascin interferes with fibronectin action. Cell 53, 383-390. the present study, where we find the best activity for cell Chiquet-Ehrismann, R. (1993). Tenascin and other adhesion-modulating proteins in cancer. Semin. Cancer Biol. 4, 301-310. adhesion and neurite outgrowth in our tenascin-C variant con- Chiquet-Ehrismann, R. (1995). Tenascins, a growing family of extracellular sisting of EGF-like repeats. This might be due to the method matrix proteins. Experientia 51, 853-862. of expression of the protein. The bacterially expressed EGF- Chiquet-Ehrismann, R., Hagios, C. and Schenk, S. (1995). The complexity like repeats may be hampered in folding up properly and the in regulating the expression of tenascins. BioEssays 17, 873-878. over 40 disulfide bridges might not link up appropriately. In Chung, C. Y., Murphy-Ullrich, J. E. and Erickson, H. P. (1996). Mitogenesis, cell migration, and loss of focal adhesions induced by tenascin- contrast, in the protein expressed in eukaryotic cell cultures we C interacting with its cell surface receptor, annexin II. Mol. Biol. Cell 7, 883- can assume that the structure including the postranslational 892 modifications is correct and therefore the activity better Colognato-Pyke, H., O’Rear, J. J., Yamada, Y., Carbonetto, S., Cheng, Y.- preserved. It will be interesting to use this hexameric tenascin- S. and Yurchenco, P. D. (1995). Mapping of network-forming, heparin- binding, and α1β1 integrin-recognition sites within the α-chain short arm of C variant consisting of EGF-like repeats to identify the inter- laminin-1. J. Biol. Chem. 270, 9398-9406. acting cell surface component(s), since no receptor for this Dörries, U., Taylor, J., Xiao, Z., Lochter, A., Montag, D. and Schachner, M. region of the tenascin-C molecule has been proposed yet. Fur- (1996). Distinct effects of recombinant tenascin-C domains on neuronal cell thermore, it will be important to investigate whether in intact adhesion, growth cone guidance, and neuronal polarity. J. Neurosci. Res. 43, tenascin-C the EGF-like repeats are indeed recognized by a 420-438. Ehrismann, R., Chiquet, M. and Turner, D. C. (1981). Mode of action of cellular receptor, or whether this interaction is caused by the fibronectin in promoting chicken myoblast attachment. J. Biol. Chem. 256, artificial C-terminal exposure of these repeats in the deletion 4056-4062. variant and in vivo access to this part of native tenascin-C is Erickson, H. P. and Bourdon, M. A. (1989). Tenascin: an extracellular matrix sterically hindered. Even if this was the case, the recognition protein prominent in specialized embryonic tissues and tumors. Annu. Rev. Cell Biol. 5, 71-92. of C-terminally exposed EGF-like repeats could serve a phys- Erickson, H. P. (1993). Tenascin-C, tenascin-R, and tenascin-X Ð a family of iological function during processes of pathogenesis and regen- talented proteins in search of functions. Curr. Opin. Cell Biol. 5, 869-876. eration accompanied by high proteolytic extracellular matrix Fischer, D., Chiquet-Ehrismann, R., Bernasconi, C. and Chiquet, M. remodeling, since the EGF-like repeats are the region of the (1995). A single heparin binding region within the fibrinogen-like domain is tenascin-C molecule most resistant to proteolysis (Chiquet et functional in chick tenascin-C. J. Biol. Chem. 270, 3378-3384. Fox, J. W., Mayer, U., Nischt, R., Aumailly, M., Reinhardt, D., Wiedemann, al., 1991). H., Mann, K., Timpl, R., Krieg, T., Engel, J. and Chu, M. L. (1991). In conclusion, our experiments have shown that the fibrino- Recombinant nidogen consists of three globular domains and mediates gen globe holds an important function in promoting the binding binding of laminin to collagen type IV. EMBO J. 10, 3137-3146. of tenascin-C to cells. Without this initial binding the activities Friedlander, D. R., Hoffman, S. and Edelman, G. M. (1988). Functional mapping of cytotactin: proteolytic fragments active in cell-substrate of the other domains are precluded. The EGF-like repeats have adhesion. J. Cell Biol. 107, 2329-2340. cell adhesion and neurite outgrowth stimulating activities, Gatchalian, C. L., Schachner, M. and Sanes, J. R. (1989). Fibroblasts that which in the context of the FN III repeats are counteracted. proliferate near denervated synaptic sites in skeletal muscle synthesize the Thus, the reaction of cells to intact tenascin-C is a complex adhesive molecules tenascin(J1), N-CAM, fibronectin, and a heparan sulfate process, involving the interplay of multiple domains. It is not proteoglycan. J. Cell Biol. 108, 1873-1890. Gesemann, M., Denzer, A. J. and Ruegg, M. A. (1995). Acetylcholine possible to infer the function of intact tenascin-C from the receptor-aggregating activity of agrin isoforms and mapping of the active addition of effects evoked by isolated domains, since the site. J. Cell Biol. 128, 625-636. 1522 D. Fischer and others

Götz, B., Scholze, A., Clement, A., Joester, A., Schütte, K., Wigger, F., Pearson, C. A., Pearson, D., Shibahara, S., Hofsteenge, J. and Chiquet- Frank, R., Spiess, E., Ekblom, P. and Faissner, A. (1996). Tenascin-C Ehrismann, R. (1988). Tenascin: cDNA cloning and induction by TGF-β. contains distinct adhesive, anti-adhesive, and neurite outgrowth promoting EMBO J. 7, 2977-2982. sites for neurons. J. Cell Biol. 132, 681-699. Phillips, G. R., Edelman, G. M. and Crossin, K. L. (1995). Separate cell Hagios, C., Koch, M., Spring, J., Chiquet, M. and Chiquet-Ehrismann, R. binding sites within cytotactin/tenascin differentially promote neurite (1996). Tenascin-Y: a protein of novel domain structure is secreted by outgrowth. Cell Adhes. Commun. 3, 257-271. differentiated fibroblasts of muscle connective tissue. J. Cell Biol. 134, 1499- Prieto, A. L., Andersson-Fisone, C. and Crossin, K. L. (1992). 1512. Characterization of multiple adhesive and counteradhesive domains in the Horton, R. M., Hunt, H. D., Ho, S. N., Pullen, J. K. and Pease, L. R. (1989). extracellular matrix protein cytotactin. J. Cell Biol. 119, 663-678. Engeneering hybrid genes without the use of restriction enzymes: gene Prieto, A. L., Edelman, G. M. and Crossin, K. L. (1993). Multiple integrins splicing by overlap extension. Gene 77, 61-68. mediate cell attachment to cytotactin/tenascin. Proc. Nat. Acad. Sci. USA 90, Husmann, K., Faissner, A. and Schachner, M. (1992). Tenascin promotes 10154-10158. cerebellar granule cell migration and neurite outgrowth by different domains Sasaki, T., Kostka, G., Göhring, W., Wiedemann, H., Mann, K., Chu, M.-L. in the fibronectin type III repeats. J. Cell Biol. 116, 1475-1486. and Timpl, R. (1995). Structural characterization of two variants of fibulin-1 Husmann, K., Carbonetto, S. and Schachner, M. (1995). Distinct sites on that differ in nidogen affinity. J. Mol. Biol. 245, 241-250. tenascin-C mediate repellent or adhesive interactions with different neuronal Schnapp, L. M., Hatch, N., Ramos, D. M., Klimanskaya, I. V., Sheppard, D. cell types. Cell Adhes. Commun. 3, 293-310. and Pytela, R. (1995). The human integrin alpha-8-beta-1 functions as a Hynes, R. O. (1990). Fibronectins. Springer Verlag, New York. receptor for tenascin, fibronectin, and vitronectin. J. Biol. Chem. 270, 23196- Hynes, R. O. (1992). Integrins: versatility, modulation, and signaling in cell 23202. adhesion. Cell 69, 11-25. Scholze, A., Goetz, B. and Faissner, A. (1996). Glial cell interactions with Joshi, P., Chung, C. Y., Aukhil, I. and Erickson, H. P. (1993). Endothelial tenascin-C: Adhesion and repulsion to different tenascin-C domains is cell cells adhere to the RGD domain and the fibrinogen-like terminal knob of type related. Int. J. Dev. Neurosci. 14, 315-329. tenascin. J. Cell Sci. 106, 389. Schulze, B., Mann, K., Battistutta, R., Wiedemann, H. and Timpl, R. Kiernan, B. W., Goetz, B., Faissner, A. and Ffrench-Constant, C. (1996). (1995). Structural properties of recombinant domain III-3 of perlecan Tenascin-C inhibits oligodendrocyte precursor cell migration by both adhesion-dependent and adhesion-independent mechanisms. Mol. Cell. containing a globular domain inserted into an epidermal-growth-factor-like Neurosci. 7, 322-335. motif. Eur. J. Biochem. 231, 551-556. Leahy, D. J., Hendrickson, W. A., Aukhil, I. and Erickson, H. P. (1992). Schwarzbauer, J. E. (1991). Identification of the fibronectin sequences Structure of a fibronectin type III domain from tenascin phased by MAD required for assembly of a fibrillar matrix. J. Cell Biol. 113, 1463-1473. analysis of the selenomethionyl protein. Science 258, 987-991. Spring, J., Beck, K. and Chiquet-Ehrismann, R. (1989). Two contrary Leahy, D. J., Aukhil, I. and Erickson, H. P. (1996). 2.0 Å crystal structure of a functions of tenascin: dissection of the active sites by recombinant tenascin four-domain segment of human fibronectin encompassing the RGD loop and fragments. Cell 59, 325-334. synergy region. Cell 84, 155-164. Spring, J. (1991). The structure and the function of tenascin. Dissertation, Lochter, A., Vaughan, L., Kaplony, A., Prochiantz, A., Schachner, M. and University of Basel, Basel. Faissner, A. (1991). J1/tenascin in substrate-bound and soluble form Timpl, R. and Brown, J. C. (1995). The laminins. Matrix Biol. 14, 275-281. displays contrary effects on neurite outgrowth. J. Cell Biol. 113, 1159-1171. Weber, P., Zimmermann, D. R., Winterhalter, K. H. and Vaughan, L. Mackie, E. J., Halfter, W. and Liverani, D. (1988). Induction of tenascin in (1995). Tenascin-C binds heparin by its fibronectin type III domain five. J. healing wounds. J. Cell Biol. 107, 2757-2767. Biol. Chem. 270, 4619-4623. Main, A. L., Harvey, T. S., Baron, M., Boyd, J. and Campbell, I. D. (1992). Wehrle, B. and Chiquet, M. (1990). Tenascin is accumulated along peripheral The three-dimensional structure of the tenth type III module of fibronectin: nerves and allows neurite outgrowth in vitro. Development 110, 401-415. an insight into RGD-mediated interactions. Cell 71, 671-678. Wehrle-Haller, B. and Chiquet, M. (1993). Dual function of tenascin: Matsumoto, K.-I., Saga, Y., Ikemura, T., Sakakura, T. and Chiquet- simultaneous promotion of neurite growth and inhibition of glial migration. Ehrismann, R. (1994). The distribution of tenascin-X is distinct and often J. Cell Sci. 106, 597-610. reciprocal to that of tenascin-C. J. Cell Biol. 125, 483-493. Yokosaki, Y., Palmer, E. L., Prieto, A. L., Crossin, A. L., Bourdon, M. A. Mayer, U., Zimmermann, K., Mann, K., Reinhardt, D., Timpl, R. and Pytela, R. and Sheppard, D. (1994). The integrin α9β1 mediates cell Nischt, R. (1995). Binding properties and protease stability of recombinant attachment to a non-RGD site in the third fibronectin type III repeat of human nidogen. Eur. J. Biochem. 227, 681-686. tenascin. J. Biol. Chem. 269, 26691-26696. Paulsson, M., Aumailley, M., Deutzmann, R., Timpl, R., Beck, K. and Engel, J. (1987). Laminin-nidogen complex: extraction with chelating agents and structural characterization. Eur. J. Biochem. 166, 11-19. (Received 21 January 1997 Ð Accepted 23 April 1997)