Fibronectin mRNA splice variant in articular cartilage lacks bases encoding the V, III-15, and I-10 protein segments.

Fibronectin is an extracellular matrix glycoprotein encoded by a single gene. Alternative RNA splicing has been reported at three sites, ED (extra type III domain)-A, ED-B, and the variable or V region. Articular cartilage fibronectin monomers are rarely (ED-A)+, but approximately 25% are (ED-B)+. RNA gel electrophoresis and Northern blot analysis identified two (ED-B)+ and two (ED-B)− fibronectin transcripts in cartilage, each pair differing by ~750 bases. This difference results from a previously unreported RNA splicing pattern that eliminates not only the V region but also nucleotides encoding protein segments III-15 and I-10. This new splice variant, which we designate (V+C)−, represents the majority of fibronectin transcripts in equine, canine, and rabbit articular cartilage but is absent in the liver. Reverse transcriptase-polymerase chain reaction analyses of 11 additional equine tissues failed to detect the (V+C)− splice variant, except for very low levels in lymph node, bone, aorta, and skin. Furthermore, chondrocytes grown in monolayer culture maintain high levels of fibronectin expression but stop expressing (V+C)− transcripts over time. The tissue-specific expression pattern of this novel fibronectin isoform suggests that it may have an important function in the matrix organization of cartilage.

Fibronectin is an extracellular matrix glycoprotein encoded by a single gene. Alternative RNA splicing has been reported at three sites, ED (extra type III domain)-A, ED-B, and the variable or V region. Articular cartilage fibronectin monomers are rarely (ED-A) ؉ , but approximately 25% are (ED-B) ؉ . RNA gel electrophoresis and Northern blot analysis identified two (ED-B) ؉ and two (ED-B) ؊ fibronectin transcripts in cartilage, each pair differing by ϳ750 bases. This difference results from a previously unreported RNA splicing pattern that eliminates not only the V region but also nucleotides encoding protein segments III-15 and I-10. This new splice variant, which we designate (V؉C) ؊ , represents the majority of fibronectin transcripts in equine, canine, and rabbit articular cartilage but is absent in the liver. Reverse transcriptase-polymerase chain reaction analyses of 11 additional equine tissues failed to detect the (V؉C) ؊ splice variant, except for very low levels in lymph node, bone, aorta, and skin. Furthermore, chondrocytes grown in monolayer culture maintain high levels of fibronectin expression but stop expressing (V؉C) ؊ transcripts over time. The tissuespecific expression pattern of this novel fibronectin isoform suggests that it may have an important function in the matrix organization of cartilage.
Fibronectin (FN) 1 is an extracellular matrix glycoprotein present in body tissues and fluids. Functionally, it is important in such diverse activities as cell adhesion, cell migration, cellular differentiation, blood clotting, opsonization, wound healing, and neoplastic transformation (Hynes, 1990). Fibronectin protein structure consists predominantly of three types of homologous repeating units (designated I, II, and III). It is encoded by a single gene, but significant protein heterogeneity results from alternative splicing of the pre-mRNA at three sites, termed extra type III domain A (ED-A), extra type III domain B (ED-B), and the variable (V) region (Schwarzbauer, 1991). The V region is sometimes also referred to as the connecting segment between the 14th and 15th type III homologous repeats (IIICS). Exons encoding ED-A and ED-B are spliced in or out in their entirety. In the V region of rat FN transcripts, however, a single 5Ј-splice donor site combines with one of three different 3Ј-splice acceptor sites (exon subdivision). In the human, an additional internal 5Ј-splice donor site is present and results in two more V region splice variants (Vibe-Pedersen et al., 1984;Kornblihtt et al., 1985;Odermatt et al., 1985;Schwarzbauer et al., 1987).
Fibronectin is an important matrix constituent in cartilage, and its content is markedly elevated in articular cartilage lesions within osteoarthritic joints (Wurster and Lust, 1982;Burton-Wurster et al., 1986). Although the precise functional role of FN in normal and diseased cartilage is unknown, several unique structural features have been described. Relatively high levels of the (ED-B) ϩ isoform have been found in both canine and human cartilage FN Zhang et al., 1995a;Zhang et al., 1995b;Rencic et al., 1995). A small subset of cartilage FN appears to be post-translationally modified with the addition of a chondroitin or dermatan sulfate glycosaminoglycan (Burton- Wurster and Lust, 1993). We have also observed that canine cartilage FN has a subunit with an apparent molecular mass ϳ15 kDa less than the smallest subunit of plasma FN. This subunit fails to react with two monoclonal antibodies that recognize epitopes in the III-15 segment, although at least some of these smaller subunits were still found within FN dimers (Burton- . These protein data are consistent with those of a subunit that retains the two sulfhydryl groups in the carboxyl terminus necessary for dimerization but has an internal deletion of the III-15 segment. In this study, we have examined the alternative splicing patterns of FN mRNA in articular cartilage. The results demonstrate a previously unreported splice variant that extends beyond any of the known 3Ј-acceptor sites in the V region and deletes nucleotides that would normally encode the 15th type III homology repeat (III-15) and the 10th type I homology repeat (I-10). This new splicing pattern is present in a majority of FN transcripts within articular cartilage and accounts for the small subunit of FN protein previously described.

EXPERIMENTAL PROCEDURES
Experimental Samples-Equine tissues were collected from a 1.5year-old intact male horse immediately after euthanasia under a protocol approved by the Institutional Animal Care and Use Committee of Cornell University. The horse was euthanized because of neurological deficits caused by a congenital malformation of a cervical vertebra but had no other evidence of illness. Articular cartilage was collected from this same horse, a 3-week-old male foal, an adult male Labrador Retriever dog, and an adult male Flemish Giant/Chinchilla cross-bred rabbit. Articular cartilage samples from within each animal were pooled from multiple joints, including shoulder, elbow, hip, and stifle. All tissue samples were snap frozen in liquid nitrogen and stored at Ϫ70°C until the time of RNA isolation.
Chondrocytes in Culture-Chondrocytes were isolated from equine articular cartilage (3-week-old foal) as described by Nixon et al. (1992). In brief, 1 g of diced cartilage fragments was incubated at 37°C with 10 ml of 0.075% (w/v) collagenase type CLS1 from Clostridium histolyticum (Worthington) until cartilage pieces were no longer visible. The cell/enzyme mixture was filtered to remove debris, and the chondrocytes were seeded in a monolayer at a density of 3 ϫ 10 4 cells/cm 2 into Dulbecco's modified Eagle's medium (Life Technologies, Inc., catalog no. 11965) supplemented with ␣-ketoglutaric acid (30 g/ml), penicillin G (20 units/ml), streptomycin (20 g/ml), and 10% fetal bovine serum. Chondrocytes were cultured for up to 46 days and were subcultured at weekly intervals.
RNA Isolation-Total RNA was isolated from cultured chondrocytes and all tissue samples except the articular cartilage by acid guanidinium thiocyanate/phenol/chloroform extraction followed by differential alcohol and salt precipitations (Chomczynski and Sacchi, 1987). Cartilage tissue, however, presents several technical problems for RNA isolation. The tissue is hypocellular and largely composed of extracellular matrix proteins. This results in low RNA yields/g of cartilage when compared with normal parenchymal tissues. In addition, chondrocytes are contained within the matrix and protected from routine extraction procedures. Finally, the direct application of RNA isolation methods that rely on differential alcohol and salt precipitations will co-precipitate large amounts of matrix proteoglycans with the RNA. To overcome these difficulties, we have combined modifications of a method originally published by Adams et al. (1992) with a commercial protocol that utilizes silica gel-based spin columns (RNeasy, QIAGEN Inc., Chatsworth, CA). Specifically, articular cartilage samples were pulverized to fine granules in liquid nitrogen using a Spex Freezer Mill (Spex Industries, Metuchen, NJ). While still frozen, 0.5 g of the cartilage powder was transferred to a 50-ml disposable polypropylene tube containing 4 ml of 4 M guanidinium thiocyanate, 0.1 M Tris-HCl, 25 mM EDTA, and 2.5 l of 2-mercaptoethanol, final pH 7.5. The sample was then homogenized at room temperature for 30 s and centrifuged at 1,500 ϫ g for 10 min. The resulting supernatant was transferred to a new 50 ml disposable polypropylene tube. Homogenization and centrifugation steps were repeated on the matrix pellet, and the supernatants were pooled. 650 l of 25% Triton X-100 was added to the pooled supernatant, mixed with a Vortex mixer, and chilled on ice for 15 min. 8.0 ml of 3 M tihydrate sodium acetate, pH 6.0, was then added, mixed with a vortex mixer, and chilled on ice for an additional 15 min. The majority of solubilized matrix proteins were then removed by sequential (usually 3) extractions (1:1, v/v) with phenol:chloroform:isoamyl alcohol (24:24:1) until a white precipitate at the organic/aqueous interface was no longer evident. Total RNA was precipitated from the aqueous layer by the addition of 0.8 volume of isopropyl alcohol, chilled on ice for 5 min, and centrifuged at 15,000 ϫ g in a SW 28 rotor (Beckman Instruments). Final purification of articular cartilage RNA was achieved by using the commercial RNeasy spin columns and following the manufacturer's protocol, starting with the solubilization of the pellet from the SW 28 rotor centrifugation step in 900 l of lysis buffer RLT. Experiments with liver (data not shown) and monolayer chondrocyte cultures ( Fig. 7) determined that this modified technique for directly isolating RNA from cartilage does not artifactually alter the primary structure of fibronectin transcripts.
RNA Gel Analysis-For Northern analyses, RNA was electrophoretically separated in 6.5% formaldehyde, 1.2% agarose gels and blotted to nylon membranes (Magna Charge, MSI, Westboro, MA) using standard protocols (Sambrook et al., 1989). The fibronectin transcripts were resolved on 20-cm-long gels for 30 h at a constant 32 V. Fibronectin cDNA fragments specific for ED-B (Zhang et al., 1995b), ED-A (amplified from equine genomic DNA using ACATTGATCGCCCTAAAG and GTGGACTGGGTTCCAA as primers), III-15 (SfcI ϫ HincII fragment isolated from the amplified RT-PCR product of primers 6s and 6a, shown in Table I, starting with liver RNA), I-10 (HincII ϫ BanI fragment isolated from the amplified RT-PCR product of primers 6s and 6a, shown in Table I, starting with liver RNA), and the carboxyl terminus (amplified RT-PCR product of primers 7s and 7a, shown in Table I) were individually cloned into pGEM-3Zf(ϩ) (Promega, Madison, WI). A type II procollagen cDNA fragment extending from exon 1 to 7 was generated by RT-PCR of equine cartilage RNA using CTGGTGCTGCTGCT-GACGCTGCTCGT and GCACCTTTTTCACCTTTGTCAC as primers. The elongation factor Tu cDNA was a gift from Dr. Roy Levine (Levine et al., 1993). Radiolabeled probes were prepared from gel-purified cDNA insert preps using [ 32 P]dCTP and random hexanucleotide primers (Prime-a-Gene, Promega) and purified with Sephadex G-50 spin columns (Boehringer Mannheim). Prehybridization, hybridization, and wash conditions followed protocols recommended by the manufacturer of the nylon hybridization membrane.
RT-PCR, Cloning, and Sequence Analysis of Equine and Canine FN cDNA-Fibronectin-specific oligonucleotides were generated from conserved regions in the human and rat FN sequence (Table I). First strand cDNA was synthesized from 4 g of total RNA with RNase H Ϫ reverse transcriptase (Superscript, Life Technologies, Inc.) using random hexamers, oligo(dT), or FN-specific antisense oligonucleotides as primers. The targeted region of FN cDNA was then amplified for 35 PCR cycles using pairs of specific primers (Table I) and Taq polymerase. Reaction products were resolved in agarose gels and visualized by staining with ethidium bromide. For cloning, amplified FN cDNA fragments were purified by agarose gel electrophoresis, isolated by glass bead extraction (Qiaex, QIAGEN Inc.), blunt-ended with Klenow, and ligated into SmaI-digested pGEM-3Zf(ϩ) (Promega). Recombinant plasmid was then electroporated into JM-109 E. coli, grown in LB medium containing ampicillin (100 g/ml), and purified using anion-exchange chromatography (QIAGEN Inc.). DNA sequence was determined with an automated PCR-based system using fluorescent labels (Applied Biosystems, Foster City, CA). Sequence data reported in Fig. 4 were determined from four dog and three horse independent clones, each analyzed in the sense and antisense orientations.

RESULTS
Cartilage Splicing Patterns of FN mRNA-Total RNA was isolated from equine articular cartilage, liver, second passage monolayer chondrocyte cultures, and peripheral blood lymphocytes. These samples were initially analyzed by Northern blot hybridization using three FN cDNA probes: an 841-base pair fragment flanking the stop codon, full-length ED-B, and fulllength ED-A (Fig. 1). The first probe, which is outside of all known alternatively spliced regions in FN, detected four distinct mRNA bands in cartilage with sizes estimated at 7.3, 7.6, 8.1, and 8.4 kb. In contrast, only two FN-specific bands were detected in the liver and cultured chondrocyte samples. The 7.3-kb band was the major FN transcript in cartilage but appeared to be completely absent in both the liver and chondrocyte RNA samples. Hybridization with ED-B cDNA detected only the 7.6-and 8.4-kb bands in cartilage and a single band in chondrocyte RNA. There was no evidence of ED-B positive hybridization in the liver. The ED-A probe recognized a single band in the monolayer chondrocytes but did not demonstrate any specific hybridization in the liver or cartilage RNA sam- ples. Lymphocyte RNA was negative to all three FN probes.
The 7.3-kb band from articular cartilage was substantially smaller than the expected sizes of FN mRNAs based on existing sequence data and the published patterns of alternative splicing. To determine its origin and to compare more specifically FN mRNA structure between articular cartilage, cultured chondrocytes, and liver, we scanned the entire FN coding region by RT-PCR (Fig. 2). A series of FN-specific oligonucleotide primers were synthesized (Table I) and used to amplify each region independently. Primer pairs 3, 5, and 6 were purposefully designed to flank ED-B, ED-A, and the V region, respectively. Primer pairs 1, 2, 4, and 7 were outside of all known regions of alternative splicing in FN and detected only a single band of the predicted size in each RNA sample. In the ED-B region, two bands were generated using RNA obtained from both cartilage and cultured chondrocytes. The size of the larger band was consistent with retention of the ED-B exon. The presence of (ED-B) ϩ mRNA in cartilage and cultured chondrocytes, but not in liver (Fig. 2), was in agreement with the Northern blot results (Fig. 1). RT-PCR with primers flanking the ED-A region amplified two bands using RNA obtained from the cultured chondrocytes. These two bands differed by 270 base pairs, which is consistent with retention of the ED-A exon in a portion of the chondrocyte FN transcripts. There was no evidence of (ED-A) ϩ transcripts in cartilage or liver. A novel result was obtained in the V region. RT-PCR of RNA isolated directly from articular cartilage produced a major band approx-imately 400 base pairs smaller than the lower band amplified from RNA obtained from liver and cultured chondrocytes. Although the two major bands amplified from liver and cultured chondrocytes were of the size expected for V ϩ and V Ϫ mRNA, the band amplified from cartilage suggested a deletion of additional nucleotides. The presence of this smaller FN transcript in articular cartilage was also confirmed in samples collected from two additional species, canine and rabbit (Fig. 3).
To determine the exact nucleotide sequence identity of the major (but smaller) cartilage band and to make comparisons with the two liver bands amplified using primer pair 6, these three cDNA fragments were independently cloned and sequenced. As predicted by their size, the two bands in liver result from inclusion, V ϩ , and exclusion, V Ϫ , of the 360-base V region. The major band in cartilage, however, results from a previously unreported RNA splicing pattern that extends from the normal 5Ј-splice donor site of the V region to the 3Ј acceptor site upstream of the exon encoding the type I homologous segment 11 (I-11). This splicing pattern maintains the same translational codon reading frame but completely eliminates from the mRNA nucleotides that normally encode segments V, III-15, and I-10 in FN protein. The complete nucleotide sequence for the full-length cDNA fragment amplified with primer pair 6 from canine chondrocytes and equine liver is given in Fig. 4A. Base differences with FN sequences published for rat and human are indicated. The amino acid sequence FIG. 3. Species comparison of articular cartilage and liver fibronectin RNA splice variants. Total RNA samples prepared from equine, canine, and rabbit articular cartilage and liver were analyzed by RT-PCR using primer pair 6 ( Table I). Amplified fragments were electrophoretically resolved in an agarose gel and visualized by ethidium bromide staining. V ϩ and V Ϫ symbolize retention and deletion of the V region, respectively. (VϩC) Ϫ symbolizes a new RNA splice variant in which nucleotides encoding the III-15 and I-10 domains are deleted together with the complete V region. bp, base pairs.

TABLE I Fibronectin oligonucleotide primers
The primer names correspond to the diagram in Fig. 2 depicting the RT-PCR scanning analysis. The position of each oligonucleotide was derived from the human FN cDNA sequence published by Kornblihtt et al. (1985) and modified to include all of exon 1, ED-B, and the complete V region.  2. Analysis of fibronectin mRNA structure by RT-PCR scanning. Total RNA samples prepared from equine articular cartilage (C), cultured chondrocytes (Ch), and liver (L) were analyzed by RT-PCR using fibronectin-specific oligonucleotide primers (Table I) (Fig. 2) and canine (Fig. 3) articular cartilage and the PCR-amplified 1567-base pair cDNA fragments from equine liver and canine chondrocytes were purified, cloned, and se-quenced as described under "Experimental Procedures." These data are available through GenBank TM accession numbers U52105, U52106, U52107, and U52108. A, the 1567 canine sequence is given in its entirety. Divergence from the equine sequence and from the published rat (Schwarzbauer et al., 1987) and human (Kornblihtt et al., 1985) fibronectin sequences are indicated. The locations of the borders between protein segments are also indicated. Alternative splice sites within the V region are starred. Canine sequence data were determined from four independent clones analyzed in both orientations. Equine sequence data were determined from three independent clones analyzed in both orientations. Numbers 5941 and 7507 at the beginning and end of the sequence, respectively, correspond to the published human fibronectin cDNA sequence (Kornblihtt et al., 1985) modified to include all of exon 1, ED-B, and the complete V region. B, the novel splice junction that results in the juxtaposition of protein segments III-14 and I-11 in the (VϩC) Ϫ splice variant is indicated (arrow), as well as the predicted amino acid change at the splice junction. predicted by the new mRNA splice junction in articular cartilage is shown in Fig. 4B. We designate this new splicing pattern for FN transcripts (VϩC) Ϫ .
To confirm independently of a PCR-based assay that the 7.3and 7.6-kb FN transcripts in articular cartilage lack nucleotides encoding protein segments III-15 and I-10, additional Northern blot analyses were conducted (Fig. 5). Separate cDNA probes specific for the III-15 and I-10 segments both hybridized to the 8.1-and 8.4-kb bands but failed to recognize the smaller cartilage transcripts as distinct bands. In contrast, the banding patterns in RNA isolated from liver and 21-day (second passage) chondrocyte cultures were not altered by using the III-15 and I-10 probes.
Tissue Specificity of (VϩC) Ϫ FN mRNA Splicing Pattern-To determine the distribution of the new (VϩC) Ϫ splice variant of FN, we analyzed 11 additional tissues by RT-PCR using primer pair 6 (Fig. 6). In all tissues with the exception of articular cartilage, the majority of FN transcripts were spliced within the V region in a pattern similar to that of liver. Although the relative intensity of V Ϫ and V ϩ bands changed between tissues, V Ϫ ,C ϩ and V ϩ ,C ϩ mRNA were still the two major splice variants. We did find RT-PCR evidence for very low levels of (VϩC) Ϫ transcripts in several other tissues, however, high level expression of the (VϩC) Ϫ splice variant was restricted to cartilage. This finding is supported by examining the progressive loss of differentiated phenotype in monolayer chondrocyte cultures over time. RNA isolated from chondrocytes after two passages and 21 days in culture demonstrated an almost complete absence of the (VϩC) Ϫ FN splice variant (Figs. 1, 2, and 5). Even after 35 PCR cycles, a band corresponding to (VϩC) Ϫ transcripts is only barely detectable. The kinetics of this change was examined by comparing chondrocytes after 1, 3, 7, 10, 14, 21, and 46 days in culture. The RT-PCR band profile generated with primer pair 6 appeared to be identical between RNA isolated directly from articular cartilage and day 1 chon-drocyte cultures. Over time, however, expression of the (VϩC) Ϫ splice variant declined, while V Ϫ ,C ϩ and V ϩ ,C ϩ isoforms increased (Fig. 7). This change in the pattern of FN splicing was broadly parallel to the loss of type II procollagen expression.

DISCUSSION
In this report, we demonstrate a new FN splicing pattern in articular cartilage that extends beyond the normal V region. This new FN splice variant is not present in the liver and is lost over time when chondrocytes are removed from their extracellular matrix and placed in monolayer cultures. We have termed this splice variant (VϩC) Ϫ . "C" denotes the cartilage-sensitive region that includes 411 nucleotides that would normally be translated into the III-15 and I-10 segments in FN protein. The absence of nucleotides encoding III-15 in the (VϩC) Ϫ splice variant readily explains our previous observation that two monoclonal antibodies specific for epitopes within this segment FIG. 7. Loss of (V؉C) ؊ fibronectin transcripts in monolayer chondrocyte cultures. Total RNA was purified from equine articular cartilage, liver, and chondrocytes maintained as monolayer cultures for the times indicated. The RNA was analyzed by RT-PCR using primer pair 6 (Table I) and by Northern analyses for type II procollagen and elongation factor Tu (EFTu) transcripts. V ϩ and V Ϫ symbolize retention and deletion of the V region, respectively. (VϩC) Ϫ symbolizes a new RNA splice variant in which nucleotides encoding the III-15 and I-10 segments are deleted together with the complete V region. bp, base pairs.

FIG. 5. Northern blot analysis of fibronectin transcripts for bases encoding the III-15 and I-10 protein segments.
Total RNA was purified from equine articular cartilage (3-week-old foal), liver (adult), and cultured chondrocytes (monolayer, passage 2). The RNA was resolved electrophoretically in triplicate, transferred to nylon membranes, and hybridized individually with 32 P-labeled fibronectin cDNA probes. The first probe recognized all splice variants of fibronectin (A11 FN). The second probe was specific for transcripts containing III-15. The third probe was specific for transcripts containing I-10. fail to recognize the small cartilage FN protein subunits in Western blots (Burton- . Chondrocytes cultured in monolayer for up to 46 days demonstrate a progressive decrease in steady-state levels of (VϩC) Ϫ FN broadly parallel to the loss of type II procollagen expression. Since an RT-PCR assay was used to follow the changes in FN splicing, however, these data do not determine exact quantitative relationships. Alternative splicing of ED-A has previously been associated with chondrocyte de-differentiation. (ED-A) ϩ FN is not normally expressed in cartilage but appears when chondrocytes are cultured and is further increased in passaged monolayer cells (Burton-Wurster et al., 1988;Bennett et al., 1991). The expression of (ED-A) ϩ FN can be modulated by the addition of dibutyryl cAMP (Leipold et al., 1992) or transforming growth factor ␤1 (Zhang et al., 1995a). In contrast, (ED-B) ϩ FN mRNA, which is expressed at high levels by chondrocytes within articular cartilage (15-35%), remains relatively high in primary chondrocyte cultures (18%), is insensitive to the addition of transforming growth factor ␤1, but is decreased by the addition of dibutyryl cAMP to the culture medium (Zhang et al., 1995a). Loss of the cartilage-specific (VϩC) Ϫ splicing pattern of FN along with the appearance of the (ED-A) ϩ isoform may prove to be an early and even more sensitive marker of chondrocyte de-differentiation in culture than is the loss of type II procollagen and aggrecan core protein expression.
Since the liver is the primary source of plasma FN, the two splice variants generated by RT/PCR from both equine and canine liver RNA are consistent with a distribution in which approximately half of the protein subunits lack the V region and half include all or a part of the V region. This is similar to what has been reported for mouse, human, rat, and cow but unlike that for the chicken in which no FN transcripts have the V region totally excluded (Schwarzbauer et al., 1985;Kornblihtt et al., 1985;Norton and Hynes, 1987;Schwarzbauer, 1991). For human and rat FN, the V region and the first half of the III-15 segment are encoded by one exon (Tamkun et al., 1984;Vibe-Pedersen et al., 1986). There is, however, an internal acceptor site immediately preceding the bases encoding III-15 that permits the entire V region to be spliced out. This region is conserved in the dog and horse, as is the alternative splice acceptor site that results in a deletion of the first 25 amino acids of the V region. In humans but not in the rat, an alternative splice donor site (GTGAG, beginning at base 881, Fig. 4A) permits deletion of the final 31 amino acids in the V region and accounts for the total of five different FN splice variants in the V region of humans versus only three in the rat. The comparable sequence in the dog and horse is ATGAG, which is identical to that in the rat. Loss of the invariant GT dinucleotide suggests that the dog and horse, like the rat, will have only three splice variants within the V region and not five.
In addition to the smaller and cartilage-specific (VϩC) Ϫ isoform, we also detected by Northern blot hybridization and RT-PCR, FN transcripts in cartilage that have a size consistent with V region splicing patterns comparable to that observed in the liver. Burton-  previously estimated that cartilage FN was 80% V ϩ . This was based on the generation of a high percentage of a 30-kDa heparin binding fragment after thermolysin digestion, reflecting the presumed insertion of the thermolysin-sensitive V region between the type III-14 and III-15 segments (Pande et al., 1987). The accuracy of this assay, however, is invalidated by the deletion of the III-15 domain in the (VϩC) Ϫ isoform. Therefore, based on the transcriptional results reported here, the earlier quantitative estimate of V ϩ appears to be much too high. Rencic et al. (1995) reported that FN splicing in seven human articular cartilage samples exhibited a pattern in which the V region was deleted. They would not, however, have been able to identify the (VϩC) Ϫ splice variant since the antisense primer they chose for PCR amplification was within the I-10 region.
From the Northern analysis data in Fig. 1, it is clear that the cartilage-specific (VϩC) Ϫ splice variant may include or exclude the ED-B segment. Direct quantitation of 32 P-labeled decay events on the hybridization membrane by phosphoimager analysis (Fuji BioImaging, Stamford, CT) suggests the following approximate distribution of FN transcripts in adult equine articular cartilage: 63% ((B Ϫ ,A Ϫ ,(VϩC) Ϫ )), 21% (B ϩ ,A Ϫ ,(VϩC) Ϫ ), 11% (B Ϫ ,A Ϫ ,V ϩ/Ϫ ,C ϩ ), and 5% (B ϩ ,A Ϫ ,V ϩ/Ϫ ,C ϩ ). In total, the 26:74 ratio of B ϩ :B Ϫ transcripts is entirely consistent with published observations (Zhang et al., 1995a, Rencic et al., 1995. It is also interesting to note that the percentage of (ED-B) ϩ transcripts is 25% in (VϩC) Ϫ splice variants, which is roughly comparable to the 31% in the V ϩ/Ϫ ,C ϩ splice variants. Based on the observed patterns of alternative RNA splicing, the predicted major protein isoforms of FN expressed in adult equine articular cartilage are summarized in Fig. 8. PCR data with primer pair 6 also support, however, the presence of other minor V region splice variants in cartilage (Figs. 2, 3, and 7). These minor variants were not independently resolved by our Northern blot analyses. The possibility of subdivisions within FIG. 8. Schematic representation of fibronectin protein isoforms expressed in articular cartilage. The principal fibronectin isoforms predicted by RNA splicing patterns in adult equine articular cartilage are indicated. Approximately 25% of fibronectin transcripts retain ED-B (Zhang et al., 1995a). In contrast, ED-A is normally deleted. Data presented in this report demonstrate that approximately 80% of fibronectin transcripts have the (VϩC) region deleted, the remainder being either V ϩ or V Ϫ . Expression of the (VϩC) Ϫ splice variant, however, appears to be regulated independently of retention or deletion of ED-B. The diagrammatic representation of fibronectin protein in this figure was modified from Hynes (1990). the C region (nucleotides encoding III-15 and I-10) must also be considered and is not resolved by these data.
Functions for some of the splice variants in the V region have been investigated and include the identification of sites that are important for cell adhesion and FN dimer secretion and the ability to affect covalent cross-linking in fibrin clots (Schwarzbauer, 1991). Functions for the ED-B and ED-A regions remain speculative but may be related to the assembly of FN dimers into pre-existing matrices. Consistent with that idea, Zhang et al. (1995b) have shown that newly synthesized (ED-B) ϩ FN is preferentially retained in the cartilage matrix compared with that of (ED-B) Ϫ FN.
The tissue-specific pattern of (VϩC) Ϫ expression and its loss by chondrocytes in monolayer culture suggest that this FN isoform may play an important role in articular cartilage matrix organization. It is not yet known if the (VϩC) Ϫ splice variant is also a predominant isoform in other cartilaginous tissues such as meniscal, nasal, costal, and tracheal cartilage. Tracheal cartilage in particular will be of interest since Zhang et al. (1995a) found that it is the only cartilaginous tissue that does not express the (ED-B) ϩ isoform of FN. At present, we have no idea what properties the loss of the III-15 and I-10 segments may confer on the mature FN protein. The III-15 domain contains a sulfhydryl group that may be involved with intermolecular disulfide bonding, although there is some evidence to the contrary (Morla et al., 1994). Ichihara-Tanaka et al. (1995) recently reported that III-15 and the I-10 through I-12 segments are actively involved in matrix assembly and that deletion of only one of the three type I segments will markedly impair matrix assembly activity. A disulfide crosslinked "super-fibronectin" with enhanced adhesive properties (Morla et al. 1994) may be close to the natural matrix form of FN but may be inappropriate for cartilage. In addition, the juxtaposition of the III-14 and I-11 segments in the (VϩC) Ϫ isoform of FN, which replaces the I-10 and I-11 junction, predicts amino acid substitutions of an acidic glutamate for a basic lysine residue in the equine and canine sequences and a glycine residue for the basic arginine residue in humans (Fig. 4B). When combined, these changes may result in alterations of the FN tertiary structure that convey additional new properties to this isoform. FN fragments but not intact FN have been reported to induce the synthesis of proteolytic enzymes in cartilage explants (Homandberg et al., 1992). If the (VϩC) Ϫ splice variant mimics a FN fragment in this regard, it may have an important regulatory role to play in matrix metalloproteinase activity within articular cartilage. Although these and other possibilities need to be independently investigated, the tissuespecific pattern of (VϩC) Ϫ FN expression suggests that this isoform has an important function in articular cartilage.