INTRODUCTION

The tenascins are a family of extracellular matrix glycoproteins with a typical multidomain structure. These proteins are widely encountered in the animal kingdom (1, 2). In invertebrates, tenascin-like molecules were demonstrated in leech (3) and in Porifera, the most primitive phylum of multicellular organism (4). In Drosophila, two molecules related to tenascins, coded by the genes ten^a and ten^m, were characterized (5, 6). Four members of the tenascin family have been identified in vertebrates so far: tenascin-C (also called cytotactin), tenascin-R (for restrictin), tenascin-X, and the recently characterized tenascin-Y. Tenascins consist of an amino terminus, which is involved in polymerization into oligomers, a series of epidermal growth factor (EGF)¹-like repeats, a variable number of fibronectin type III (FNIII)-like repeats, and a carboxyl-terminal globular fibrinogen-like domain (1, 2). In contrast to other known tenascins, tenascin-Y harbors a domain containing repeated serine-proline-Xaa motifs, which interrupts the series of FNIII domains (7). The numbers of FNIII and EGF repeats differ among the tenascins and among species. Different isoforms of tenascin-C, tenascin-R, and tenascin-Y arise from various splicing events in the region coding for FNIII repeats (7-9). An isoform lacking part of the NH₂-terminal cysteine-rich region was reported in chicken tenascin-R (10).

Tenascin-X was initially reported as a partial sequence coded by gene X, found on the opposite strand of the human 21-hydroxylase gene (P450c21B) and located in the human major histocompatibility complex class III region (11-13). By extending genomic sequences, Bristow et al. (14) obtained enough data to predict the modular structure of the tenascin-X protein, which comprises an amino-terminal cysteine-rich region with four heptad repeats, 18.5 EGF-like repeats, at least 29 FNIII repeats, and a fibrinogen-like domain. The carboxyl termini of mouse (15), rat (16), and pig tenascin-X (17) have been characterized by cDNA cloning. Northern blot analysis showed that the major species of tenascin-X mRNA was about 13 kb (15, 17, 18). Other, smaller bands were detected, suggesting the possibility of alternative splicing (15).

Tenascin-X is expressed much more widely than tenascin-C or tenascin-R (2). Most tissues express detectable levels of tenascin-X mRNA, but some fetal tissues (skeletal and heart muscles, dermis, testes, nerves, and digestive tract) express high levels (14-17), which are maintained in the adult tissues (except testes) and are also found in tendons, ligaments, and peripheral nerves (17). Several cell lines of various origins, including normal and transformed fibroblasts, carcinoma, and glioma cells, also express tenascin-X (15, 18). Biochemical analysis indicated that the molecular mass of the tenascin-X monomer is 450-500 kDa (15, 18). A smaller variant of 220 kDa has been described in cultured fibroblasts (18).

Little is known about the function of tenascin-X, but indirect evidence suggests that it is vital. Numerous mutations in the gene coding for steroid 21-hydroxylase, which cause adrenal hyperplasia, are observed in humans. The fact that none of these deletions extends to the region coding for tenascin-X suggests that this protein has an essential function (11). The pattern of tenascin-X expression in heart, skeletal muscle, and limbs during development is consistent with this hypothesis (16).

In a previous study, we characterized flexilin, a bovine extracellular matrix glycoprotein (19). This high molecular mass molecule, which appears to be a flexible monomer, is located on collagen fibrils. To characterize flexilin further, specific antibodies were used to screen a fetal calf skin expression library. We report here the complete cDNA structure of flexilin. The deduced amino acid sequence indicates that the overall domain organization of flexilin is similar to that of tenascins and that it corresponds to the bovine orthologue of human tenascin-X. Thirty FNIII repeats were found in the bovine cDNA, some of which were unusually long (up to 134 amino acids), compared with typical FNIII repeats identified previously (90-100 amino acids). These results, biochemical, cell attachment, rotary shadowing data, and immunofluorescence localizations of flexilin are discussed in relation to the available data on tenascin-X.

EXPERIMENTAL PROCEDURES

Mouse monoclonal antibodies were produced according to previously described procedures (20). After immunization with a complex antigenic mixture (containing flexilin and types XII and XIV collagens) and immunofluorescence screening, nine monoclonal antibodies specific for flexilin were selected. They were characterized by comparing their properties with those of the 4E7 clone, which was described previously (19), by tests that included enzyme-linked immunosorbent assay (ELISA), Western blotting against purified flexilin, and immunofluorescence assays in various organs. The 8F2 clone was chosen for this study. Polyclonal antisera were obtained by immunizing mice with three injections (10 µg each) of immunopurified flexilin.

Skin from fetal calf (18 weeks old) was dissected, frozen, and powdered in liquid nitrogen. Total RNA was prepared with RNA-B^TM (Bioprobe Systems), and poly(A)-rich RNAs were selected by chromatography on oligo(dT)-cellulose. A cDNA library was prepared from 5 µg of poly(A)-rich RNA, using random hexamer primers and reagents from a cDNA synthesis kit and a MOSElox cloning kit (Amersham Corp.), according to the manufacturer's protocol. This expression cDNA library was screened initially with monoclonal antibodies 8F2 and 4E7. About 200,000 recombinant MOSElox clones were plated with Escherichia coli strain BL21 DE3 and grown for 6 h at 37 °C. Nitrocellulose filters (Hybond C, Amersham), which had previously been treated with 10 mM isopropyl-1-thio--D-galactopyranoside, were placed on the plates and incubated at 37 °C for 4 h. The filters were washed in 25 mM Tris-HCl, pH 8.0, 140 mM NaCl, and 2.7 mM KCl (Tris-buffered saline) and submitted to the immunodetection procedure using alkaline phosphatase conjugate, as described previously for Western blotting (19).

Three antibody-positive cDNA clones (Flex-1, Flex-2, and Flex-3; Fig. 1) were selected and screened again once or twice with the same monoclonal antibodies. Purified clones were further tested with a polyclonal antibody (see above). Insert cDNAs of the three clones were therefore used to screen the same cDNA library under highly stringent conditions, as described previously (21). From 150,000 plaque-forming units, 25 positive clones were purified; four of them (Flex-11-Flex-13) are presented in Fig. 1.

To avoid the purification and characterization of numerous cDNAs, an oligonucleotide strategy was preferred. For this purpose, 5- and 3-specific oligonucleotides (Fig. 1) were used to screen the skin cDNA library, using ³²P-end-labeled oligonucleotides by standard procedures (22). Briefly, duplicate filters were probed at 50 °C for 2 h in hybridization buffer containing 0.05% sodium pyrophosphate, 6 × SSC (1 × SSC is 15 mM sodium citrate, 150 mM NaCl), 0.1% sodium dodecyl sulfate (SDS), 1 × Denhardt's solution (0.02% each of polyvinylpyrrolidone, Ficoll, and bovine serum albumin), and 100 µg/ml denatured salmon sperm DNA. The filters were washed twice in 3 × SSC and 0.05% sodium pyrophosphate at room temperature for 15 min and then with 6 × SSC and 0.05% sodium pyrophosphate at 50 °C for 15 min. Three successive oligonucleotide screenings were performed, and five cDNA clones (Flex-21-Flex-23, Flex-31, and Flex-41; see Fig. 1) were obtained. None of the cDNA clones extended upstream of Flex-21. With the previously isolated clones, 9253 base pairs of sequence, including the 3-untranslated region, were obtained.

The available data predicted that flexilin is the bovine orthologue of human tenascin-X. To extend the sequence in the 5 direction, human genomic data were used to generate two genomic polymerase chain reaction (PCR) fragments. The human genomic regions that were amplified corresponded to nucleotides 122-452 and 3353-3574 of the human tenascin-X gene (numbering from Bristow et al., Ref. 14), which code for the domain responsible for the oligomerization and the furthest amino-terminal FNIII repeat of human tenascin-X, respectively. The amplified fragments were cloned, sequenced, and used as a probe to screen 400,000 plaque-forming units under moderately stringent conditions, as described previously (21). Two positive clones, TX-1 and TX-2, were isolated (Fig. 1).

The last two gaps between clones TX-1-TX-2 and TX-2-Flex-21 were filled up by the reverse transcriptase (RT)-PCR strategy. First-strand cDNA of poly(A)-rich RNA was synthesized using Expand^TM RT (Boehringer Mannheim) and random hexamer primers, according to the manufacturer's instructions. First-strand cDNAs were used for PCR with Long Expand^TM Taq polymerase (Boehringer Mannheim). For the first gap, between clones TX-1 and TX-2, two primers were used to generate the PCR-1 clone (Fig. 1): PCR-1, forward (5-GGACTGCGGTACGCGTGCCTGCCCTGGCGA-3), corresponding to nucleotides 1652-1681; and PCR-1, reverse (5-GCAGTCCTCACCCTCGTAGCCGTCGTC-3), corresponding to nucleotides 1910-1937.

For the second gap, between clones TX-2 and Flex-21, two primer pairs with a common 5-oligonucleotide were used to generate clones PCR-2 and PCR-3 (Fig. 1). The primers are: PCR-2 and -3, forward (5-CCAGCCTGACACCTTTACCCACTTCCAGC-3), corresponding to nucleotides 3131-3159; PCR-2, reverse (5-TTGCGGCCGACGTCCAGGTTGGAGATGAGG-3), corresponding to nucleotides 3524-3553; and PCR-3, reverse (5-AGAGCCCTGGGCCACCGTCC-3), corresponding to nucleotides 4021-4040.

PCR was performed in the reaction mixture described in the Boehringer Mannheim protocol. Initial denaturation at 94 °C for 2 min was followed by 30 PCR cycles at 94 °C for 10 s, 61 °C for 30 s, 68 °C for 45 s and an additional elongation time of 20 s/cycle for the last 20 cycles. After a final extension at 68 °C for 7 min, the PCR products were purified from agarose gels, cloned into pBluescript II SK(+) (Stratagene), and sequenced.

To obtain information about human sequences coding for the FNIII motifs VII to IX (numbering from Bristow et al., Ref. 14), RT-PCR experiment was carried out as above using poly(A)-rich RNAs purified from MG63 cells. The RT-PCR product coded for amino acids 1284-1350 of the human tenascin-X (14). Using the poly(A)-rich RNAs, a MG63 cDNA library was prepared in the MOSElox cloning kit (see above) and screened with the RT-PCR product. This allowed the purification of one cDNA clone (HFX1). For genomic analyses, PCR experiment was performed using human genomic DNA (Promega). The amplified human genomic region corresponds to nucleotides 6486-6797 of the human tenascin-X gene (numbering from Bristow et al., Ref. 14).

Sequencing was performed with the T7 sequencing kit (Pharmacia Biotech Inc.) using the dideoxynucleotide chain termination procedure on double-stranded DNA. Both DNA strands were sequenced using universal primers and synthetic oligonucleotides (Isoprim, Toulouse, France) corresponding to the appropriate cDNA sequences. For G- and C-rich regions, the sequencing reactions were carried out with dITP, an analogue of dGTP. The DNA sequences were analyzed by the DNAid computer program (23). Multiple alignments and identity studies were performed via the IBCP site server² with the Antheprot (24) and Multalin (25) programs.

Northern blotting was performed as described elsewhere (22). Briefly, 10 µg of total RNA were electrophoresed on a 0.8% agarose, 0.66 M formaldehyde gel. The gel was soaked in 50 mM NaOH and 10 mM NaCl for 20 min and neutralized twice in 0.1 M Tris-HCl, pH 7.4, for 20 min. The RNAs were transferred overnight onto nylon membranes (Hybond N+, Amersham). The filters were hybridized with a ³²P-labeled cDNA probe (5 PstI fragment of clone Flex-12) overnight at 65 °C in 6 × SSC, 5 × Denhardt's solution, 0.1% SDS, and 100 µg/ml denatured salmon sperm DNA. The final washing was done in 0.1 × SSC and 0.1% SDS at 65 °C. The 0.28-6.58-kb RNA ladder from Promega was used as a size marker.

Flexilin was extracted from bovine fetal skin with 0.5 M NaCl in the presence of proteases inhibitors and immunopurified on a column prepared with the 8F2 antibody. The purification procedures, SDS-polyacrylamide gel electrophoresis (PAGE), immunoblotting, rotary shadowing, and immunofluorescence were performed as described previously (19).

The osteosarcoma cell line MG63 was obtained from American Type Culture Collection (Rockville, MD) and was maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum and 50 µg/ml gentamicin. Adhesion substrates were prepared by adding 50 µl of immunopurified bovine tenascin-X (at 5 µg/ml) to 96-well ELISA plates (Nunc) overnight at 4 °C. The plates were blocked with 1% bovine serum albumin for 2 h. Cell suspension was obtained by a brief treatment of the cell layer with 0.25% trypsin and 0.05% EDTA, centrifugation with medium containing 1 mg/ml soybean trypsin inhibitor (Sigma), and resuspension in DMEM without serum. To perform the adhesion, 100 µl of cell suspension was added to each well and left for 50 min at 37 °C. For inhibition studies, cells were suspended with DMEM containing synthetic peptides, GRGDS as inhibitor, or SDGRG as control (Sigma). Incubation with peptides was performed for 10 min at room temperature before adding cell suspension to the wells. Non-adherent cells were removed and adherent cells submitted to fixative treatment using the buoyancy method (26). Wells were rinsed with phosphate-buffered saline, and adherent cells were quantitated by coloration with 0.1% crystal violet, lysis with 0.2% Triton, and reading absorbance at 570 nm on a MR5000 spectrophotometer (Dynatech). Three wells were averaged for each point tested, and the standard deviation of the three points was above 10% of the mean. For morphological studies of adherent cells, eight chamber culture slides (Falcon) were coated with tenascin-X or fibronectin as described above. After blocking, cells were added, left for 50 min at 37 °C, and gently washed with phosphate-buffered saline. After fixation with 2% glutaraldehyde, cells were examined by phase contrast microscopy and photographed. Inhibition studies were performed as described above.

RESULTS

Poly(A)⁺ RNAs extracted from early embryonic calf skin, a flexilin-rich tissue (19), were used to construct an expression cDNA library. Screening of this library with monoclonal antibodies specific for flexilin resulted in the detection of three cDNA clones (Flex-1, Flex-2, and Flex-3; Fig. 1), which were also immunoreactive with polyclonal anti-flexilin serum. The 5 ends of these clones were sequenced according to the expression vector. The sequence NLYGFHDR, previously obtained by amino acid sequencing of flexilin tryptic peptides (19), was identified in two of these cDNA clones (Flex-1 and Flex-2; Fig. 2). Moreover, complete characterization of the three immunoreactive clones shows that none of them overlap. Analysis of their sequences reveals that they potentially code for FNIII repeats, the highest score being shared with the central FNIII repeats of human tenascin-X (14). Owing to large gaps between the bovine and human sequences, the percentage of identity could not be determined precisely.

Multiple screenings of the cDNA library with specific cDNA and oligonucleotide probes (see "Experimental Procedures"), allowed us to purify eight new clones (Flex-11-Flex-13, Flex-21-Flex-23, Flex-31, and Flex-41; Fig. 1). These overlapping cDNA clones cover 9253 bp, with an open reading frame of 3045 codons ended by a stop codon TAG, and 113 bp of 3-untranslated region. The open reading frame could be divided into two parts (Fig. 1), comprising a succession of 27 FNIII repeats and a fibrinogen-like carboxyl-terminal motif. The last FNIII repeats (numbered 26-29, Fig. 1) and the fibrinogen-like motif share more than 80% of identity with comparable domains of human, mouse, and pig tenascin-X (14, 15, 17). Moreover, the 3 end of flexilin cDNA is 100% identical to the 3 end of the bovine cytochrome P450c21 gene (data not shown; Ref. 28). The same overlap was identified in human, rodents, and pig genes (11, 15-17). Taken together, these findings indicate that bovine flexilin and human tenascin-X are orthologous proteins.

Since even extensive screening was insufficient to obtain the full cDNA sequence, PCR products coding for either the amino terminus or the furthest amino-terminal FNIII repeats of human tenascin-X were generated and used to screen our cDNA library again. With this procedure, we obtained two clones, TX1 and TX2 (Fig. 1). The two last gaps, of 69 and 135 bp, were filled by sequence analysis of RT-PCR products. The complete cDNA sequence is 12,706 bp long and contains an open reading frame of 4135 codons. The deduced amino acid sequence is presented in Fig. 2. The 5-untranslated region (185 bp long) is comparable to the counterpart sequence of the major mRNA of human tenascin-X (29). Northern blot analysis of total RNAs from fetal skin showed that bovine tenascin-X transcript is about 13 kb (Fig. 1).

Bovine tenascin-X harbors a typical tenascin structure, which is presented in Figs. 1 and 2. The first 22 amino acids may correspond to a typical signal peptide (30). Thus, the predicted molecular size of mature bovine tenascin-X monomer was calculated to be 445 kDa. The signal peptide is followed by a cysteine-rich region (amino acids 23-152). Except for the furthest amino-terminal sequence, this domain is comparable to the region responsible for the trimerization of tenascin-C and tenascin-R monomers and is composed of four heptad repeats flanked by seven cysteine residues. Next is found the carboxyl-terminal half of an EGF-like domain, separated from 18 complete EGF-like domains by a proline- and serine-rich spacer of 15 amino acids (amino acids 153-744). The cysteine residues in these EGF-like repeats are organized in the same way as in other tenascins, i.e. X₄CX₃CX₅CX₄CX₁CX₈C (26). As in human tenascin-X, the series of 18 EGF-like modules is interrupted by a four-amino acid insertion between repeats 16 and 17. The sequence following the EGF-like motifs is composed of 30 FNIII repeats (amino acids 745-3910). The size of each FNIII domain varies from 88 to 134 residues. Finally, a fibrinogen-like domain is found at the carboxyl terminus (amino acids 3911-4135). The predicted amino acid sequence also reveals five N-linked oligosaccharide acceptor sites in the amino and carboxyl termini of the molecule (Figs. 1 and 2). A single RGD sequence, suggesting a cell adhesive site, is found in the b10 FNIII motif (Figs. 1 and 2).

For this analysis, the human and bovine tenascin-X sequences were separated into three parts. Two of them, the amino-terminal region (comprising the probable oligomerization region and the EGF-like domains) and the fibrinogen-like domain, could be aligned perfectly in the two species, with an identity greater than 80%. The third region, which includes the FNIII repeats, is more difficult to analyze since the human tenascin-X genomic sequence has not been fully characterized, especially in the 5 region.

The 30 bovine (b0-b29) and the 28 human (h1-h29) FNIII repeats were also compared; the main results are presented in Fig. 3. The furthest amino-terminal bovine FNIII repeat (b0, Fig. 3A) has not yet been characterized in human tenascin-X, but the third bovine FNIII repeat (b2) may represent the human FNIII-2 repeat (h2), which has been identified only by Southern blotting (13). The b0 and b2 repeats showed strongest identity with b1 and b28 (31%) and with b26 (28.4%), respectively. As shown in Fig. 3 (A and B), three regions of bovine and human tenascin-X appear to be colinear. The first contains FNIII repeats 1-5 in both species (Fig. 3A). The second corresponds to FNIII motifs b9-b13 and h7-h11 (Fig. 3A). The third encompasses FNIII repeats 22-29 in both species (Fig. 3B).

The remaining FNIII repeats are essentially located in the central part of both bovine and human tenascin-X. Three classes of repeats could be identified (Fig. 3C). The first, X1, comprises highly homologous FNIII repeats (greater than 80% identity), and the prototypes of this class are FNIII motifs b7 and h6. The major characteristic of these FNIII motifs is the presence of a variable number of nonade (nine amino acids in length) motifs at the amino terminus. These nonade motifs, which have the consensus sequence T(E/A)XEETPSP, are generally followed by the PEEPPEP sequence (Fig. 2). Two of the human X1 FNIII domains (h6 and h15) lack the nonade motif. In one of them, h15, the choice of an upstream AG splicing acceptor site from the published data (14) leads to the addition of 12 amino acids at the amino terminus, including a nonade motif. The second class of FNIII repeats, X2, also represents strongly identical repeats with prototype domains, b13 and h11. The third class of FNIII repeats, observed only in bovine, is called type X1-2, as the repeats are related to both of the other FNIII types (70-80% identity). The bovine FNIII repeats b6 and b14, which are included in the X1-2 class, harbor nonade motifs at their amino termini, like type X1. As shown in Fig. 4, the three FNIII classes are arranged differently in bovine and human tenascin-X; it is therefore difficult to assume any colinearity for this region in the two species.

SDS-PAGE analysis under reducing conditions and Coomassie Blue staining show freshly isolated, immunopurified tenascin-X as three high molecular mass bands (Fig. 5A, lane 1). The faster migrating bands (380-400 kDa) correspond to those described previously (19). The molecular mass of the more slowly migrating band (420 kDa) was estimated by comparison with the migration of unreduced bovine plasma fibronectin. This value is in correct agreement with the mass calculated from the deduced amino acid sequence. The latter band disappears rapidly during storage, whatever the conditions.

To verify that some of these bands do not correspond to proteins that interact with tenascin-X, we performed immunoblotting using the 8F2 antibody on purified protein. All three bands were clearly revealed (Fig. 5A, lane 3), demonstrating that monomers of different sizes were present in our purified preparation. Only two bands were observed after SDS-PAGE under nonreducing conditions, the band with intermediate migration being stronger than under reducing conditions (Fig. 5A, lane 2). Moreover, very faint bands are found on the top of the gel under nonreducing conditions. When examined by rotary shadowing, the immunopurified preparations were found to be highly homogeneous, consisting of flexible units of 150-180 nm ending with a globule (Fig. 5B).

The expression of tenascin-X has been tested in various bovine organs. It has been shown to be present in skin, tendon, and kidney glomeruli and under endothelia (19). In this study, we extended our observations to muscle and intestine. Connective tissue associated with skeletal muscle (epimysium and perimysium) was intensely labeled with anti-flexilin antibodies (Fig. 6A), and similar results were obtained within the striated muscle of the hypodermis (Fig. 6B). In cardiac tissue, typical extracellular matrix staining is observed along muscle cells (Fig. 6C). In gut sections, thin fibers of the lamina propria are stained (Fig. 6D).

In a previous study (14), no RGD site was characterized in human tenascin-X. From Fig. 4, the domain comparable to bovine FNIII b10, which harbors the RGD sequence, is the human h8 motif. To detect the presence of an RGD site in the uncharacterized amino-terminal part of the h8 motif, we performed RT-PCR experiments covering part of the sequence coding for the h7 and h8 domains (Fig. 7). As shown in Fig. 7B, the amino-terminal side of the FNIII motif h8 contains an RGD sequence. This result was confirmed by analysis of a human cDNA clone, although some discrepancies with the published human sequence were noted (Fig. 7). Several sequence conflicts are clustered just after a gap of 500 bp, including the 5 part of the exon coding for h8 repeat. With our data, an improved identity score was obtained between the FNIII motifs b10 and h8 (87.5% instead of 72.9%). Moreover, we found that the sequence of the amino-terminal part of the human FNIII motif h9 also differs from the published data. At the nucleotide level, the divergence is located upstream of a PstI site. Analysis of a genomic PCR product reveals that a 327-bp PstI fragment was previously omitted (Fig. 7B and Ref. 14). From this PstI fragment, it appears that the human FNIII motifs h8 and h9 are coded by two exons but not by an exon fusion as described previously (14). After these corrections, the identity between the h9 and b11 repeat sequences increases from 80.2 to 84.6%.

The function of the RGD site in bovine tenascin-X was tested by cell adhesion studies using the osteosarcoma cell line MG63. These cells were shown to adhere on tenascin-X-coated wells (Fig. 8). Moreover, they exhibited a round morphology and did not spread on tenascin-X (Fig. 8B), in contrast to the results obtained on immobilized fibronectin (Fig. 8A). Inhibition studies conducted with a synthetic peptide containing the RGD sequence clearly demonstrate that this adhesion is RGD-dependent (Fig. 8C), and that inhibition increases with peptide concentration (Fig. 8D).

DISCUSSION

In this study, we have characterized the primary structure of bovine flexilin. Sequence analysis indicates that this large extracellular matrix glycoprotein is the bovine orthologue of human tenascin-X. Moreover, the immunofluorescence localizations are in good agreement with those observed in other species (15, 16). Our results therefore represent the first complete sequencing of the nucleotides and deduced amino acid for the full length of the cDNA of a vertebrate tenascin-X.

Electron micrographs of purified bovine tenascin-X (Fig. 5B and Ref. 19) show the presence of monomeric molecules appearing as 150-180-nm flexible structures with a knob at one end, which is probably the fibrinogen-like module. The tenascin-X molecule looks like the tenascin-C, -R, and -Y monomers, except for its size and its flexible aspect. The length of the molecule is due to the large number of FNIII repeats (30, whereas tenascin-C has only 8-16); its flexible aspect may reflect the unusual size (up to 134 amino acids) of some FNIII repeats in bovine tenascin-X, in comparison with 90-100 amino acids in these repeats in other proteins (31). The additional amino acids are located at the amino terminus and are mostly multiple nonade motifs with the canonical sequence T(E/A)XEETPSP. It is unknown whether these conserved regions are an integral part of the FNIII repeats or act as a link between them. The amino-terminal extensions observed in bovine tenascin-X FNIII repeats, which are rich in glutamic acid and proline, are also present in chicken tenascin-Y but in a shorter form. Other similarities between tenascin-Y and tenascin-X were described by Hagios et al. (7), who indicated that the FNIII domains YB, YC, YD, and YE of tenascin-Y are similar to the 23 central FNIII repeats of human tenascin-X. Our comparisons between bovine tenascin-X and chicken tenascin-Y led us to the same conclusion.

As suggested by Bristow et al. (14), the four heptad repeats located at the amino terminus of human tenascin-X could provide the structural basis for the assembly of three monomers, which might form a triple-stranded, coiled helix analogous to TN-C and TN-R trimers. These heptads are flanked by seven cysteine residues that may stabilize the trimer by forming disulfide bonds. Human and bovine tenascin-X lack the amino-terminal cysteine found in tenascin-C and -R, which is involved in the polymerization process in tenascin-C, by which two trimers form an hexamer. The production of monomeric tenascin-X molecules from fetal bovine skin may be due to proteolysis in a sensitive site at the amino terminus. In some reports, tenascin-C and -R were also shown to be extracted from tissues as monomers (10, 32). This hypothesis is likely since the 420-kDa band tends to disappear during storage. We postulate that other bands observed in SDS-PAGE (Fig. 5A) arise from either another protease-sensitive site or from alternatively spliced isoforms.

Bovine tenascin-X contains five putative sites of N-glycosylation (NXS/T), which are observed in same locations as in the human tenascin-X molecule. Another interesting sequence, a putative RGD cell attachment site, is located in the FNIII repeat b10 (Fig. 2). By RT-PCR and cDNA analysis, we have shown that the corresponding human FNIII domain (h8) contains this RGD site. It is noteworthy that the tenascin-X RGD site is located at the amino terminus of the FNIII b10 and h8 modules (Figs. 2 and 7A), whereas in human and chicken tenascin-C and in fibronectin the RGD site is located at the carboxyl terminus of FNIII repeats (25, 33, 34). It is tempting to speculate that its location (between two folded FNIII repeats) might promote its interactions. Cell adhesion experiments using purified bovine tenascin-X and inhibition with RGD peptides are consistent with this hypothesis. Other adhesive sites may also be present in the tenascin-X molecule, such as the fibrinogen-like motif, which was demonstrated to be active in tenascin-C; adhesion to this domain was inhibited by RGD peptides, even though it has no RGD sequence (35). It is likely that both RGD site and fibrinogen-like domain are involved in cell adhesion to tenascin-X.

An overall comparison of bovine and human tenascin-X sequences shows good correlation between the two termini of these proteins. Greater than 80% identity is observed for the cysteine-rich potential polymerization region including the four heptad repeats, the 18.5 EGF-like domains, and the fibrinogen-like domain. Within the FNIII region, this score is reached for the first five FNIII repeats, bovine b7 and human h6 FNIII domains, bovine b9-b12 and human h7-h10 FNIII modules, and the eight carboxyl-terminal FNIII repeats (Figs. 3 and 4). Large uncharacterized sequences between the exons coding for the human FNIII repeats h5 and h6 and for h6 and h7, of 1.8 and 6.5 kb (14), respectively, may contain genetic information coding for repeats corresponding to bovine FNIII b6 and b8. Following the same argument, bovine repeats b0 and b2 may have their counterparts in human tenascin-X.

The central parts of human and bovine tenascin-X seem to be organized differently (Figs. 3 and 4). They are made of three highly homologous types of FNIII repeats, which we have named X1, X2, and X1-2. The last group, characterized only in bovine, is related similarly to types X1 and X2 and may represent a common ancestor. The absence of colinearity in the central part of bovine (deduced from cDNA data) and human (deduced from genomic analysis) tenascin-X arises from a distinct number and a different organization of the three types of FNIII repeats. The first hypothesis to explain this organization is that the genomic region coding for the central part of the tenascin-X has evolved differently in the two species. Similarly, the available data indicate that human tenascin-C harbors two more FNIII repeats than mouse and chicken proteins. In the second hypothesis, bovine and human tenascin-X genes have the same genetic information, and the discrepancy observed between the two species arises from alternative splicing events. Several lines of evidence are consistent with the latter hypothesis. In the region coding for the human FNIII repeats h11-h21, there are five gaps of 0.5, 1, 3.2, 1, and 0.8 kb (Fig. 4). Although the 11 human FNIII repeats may reflect the entire genetic information of this region, the possibility that one or more additional FNIII repeats are contained within these gaps cannot be excluded. In bovine tenascin-X, only nine FNIII repeats (b13-b21) are present in this region, and one of them, FNIII b14 (type X1-2), has not been identified in the human sequence. It might be suggested that the exons coding for the central FNIII repeats of tenascin-X are involved in alternative splicing events. It is noteworthy that these repeats have the strongest identity with the FNIII repeats of tenascin-Y, -C, and -R, which undergo alternative splicing. This possibility does not contradict the results of Speek et al. (29), who showed that human tenascin-X transcripts always contain the sequence coding for the last FNIII repeats (h23-h29). An alternate explanation for the differences between bovine and human tenascin-X could be a combination of the two hypotheses.

The present study has elucidated the primary structure of bovine tenascin-X by cDNA cloning and has correlated these results with biochemical data. Morphology of cells in contact with tenascin-X suggests a function similar to that of tenascin-C, i.e. the modulation of cell-matrix interactions, and the RGD sequence found in tenascin-X seems of particular importance for its biological properties.