Domain structure of fibronectin and its relation to function. Disulfides and sulfhydryl groups.

Hamster cell fibronectin is a glycoprotein consisting of two 230,000-dalton subunits in a disulfide-bonded dimer. The molecule is composed of domains which can be separated by partial proteolytic cleavage. The carbohydrates, disulfide bonds, and a single free sulfhydryl group per chain are distributed nonuniformly among these regions. All the interchain disulfides are within 10,000 daltons of the end of the molecule and are removed by mild proteolysis which also generates 200,000- and 25,000-dalton fragments which do not contain interchain disulfides. The 200,000-dalton fragment contains all or most of the carbohydrate side chains, and the free sulfhydryl group, but is relatively poor in cystine. The 25,000-dalton fragment is carbohydrate-free and cystine-rich but has no free sulfhydryl groups. There is heterogeneity in carbohydrate content among the monomeric chains of intact fibronectin and the 200,000-dalton fragments. The gelatin binding site of fibronectin is in the 200,000 fragment. Intact disulfide bonds are required for binding of fibronectin to cells and to gelatin and blockage of the free sulfhydryl groups prevents binding of fibronectin to cells, suggesting that intermolecular disulfide bonding may be important.

and exists in a fibrillar network between cells and substrata (6), in regions of cell-cell contact (6,7), and in dense cultures as an elaborate network covering the cells (6,8,9). It appears to be involved in adhesion of cells to substrata and its presence affects a variety of phenotypic properties of cells including adhesion, spreading, morphology, overlapping, and alignment of cells (10,11). Fibronectin has a high affinity for collagen, a property which could be involved in some of its biological functions (12)(13)(14)(15). Addition of fibronectin to transformed cells causes reappearance of microfiiament bundles within the cells (11,16) and promotes cell migration (17). In normal cells the distribution of fibronectin correlates with that of actin (18

30).
To study the fragments formed by tryptic digestion, iodinated dishes of NIL8 cells were treated with trypsin at varying concentrations and times. Reaction was stopped by soybean trypsin inhibitor. Samples of the liquid phase and of the cell lysates were run on SDS-polyacrylamide gels with or without reduction (Fig. 1). In both cases strong bands corresponding to polypeptides of molecular weight around 200,000 (co-migrating with myosin) appeared on the cells and were released into the digests. Precipitation with antibody to fibronectin showed that these bands have antigenic determinants of fibronectin (not shown). This means that cell surface fibronectin has an asymmetric distribution of interchain disulfide bridges. A stretch of 200 kd in each half of the disulfidebonded dimer possesses no disulfide bridges binding it covalently to the other polypeptide chain. This 200-kd fragment therefore behaves in the same way if run reduced or nonreduced.
To test whether the resistance of the 200-kd fragment to proteolysis was intrinsic or caused by its associations on the cell surface, we studied fibronectin in solution. Fibronectin purified from chick embryo fibroblasts or from NIL cells was also rapidly cleaved by trypsin (1 pg/ml) to give 200-kd fragments which were then relatively resistant to further cleavages. The 200-kd fragments were again released with or without subsequent reduction (see below). This 200-kd polypeptide was also obtained whether or not the protein was previously reduced with dithiothreitol (not shown). At short digestion times, fragments of intermediate sizes 215,000 and 220,000 daltons could be observed (Fig. 2). Both these fragments migrated at the same positions reduced or nonreduced.
An additional fragment of 240,000 appeared only on the nonreduced half of the gel. These results indicated several tryptic cleavage sites in the susceptible 30,000-dalton region and indicate that the interchain disulfide bonds are within 10 kd of the end of the intact 230-kd molecule. Procedures"). Results are shown in Fig. 3.
Several conclusions can be drawn from these results. First, the carbohydrate/amino acid ratio (3H/35S or 3H/14C) is consistently higher at the slower migrating portion of the fibronectin band (Fig. 3, A and C). In high resolution gels we are sometimes able to resolve two bands of fibronectin after reduction. ' The results in Fig. 3 suggest that they may differ * Unpublished data.

B.
in their carbohydrate content. After cleavage with trypsin, the asymmetry of carbohydrate content persists (Fig. 3 If one assumes that 100% of the carbohydrate is in the 200-kd fragment, the observed recovery of ["5S]cystine in this fragment would predict that 50% of the half-cystines are in the 200-kd fragment. Combining the two sets of data, it is clear that only 34 to 50% of the halfcystines are present in the 200-kd fragment. This leaves 50 to 66% of the half-cystines in the 30-kd region removed by proteolysis. Amino acid analyses of fibronectin, carried out according to several different protocols with particular attention to halfcystine residues, indicate 40 to 48 half-cystine residues per chain (Table I)   detectable glucosamine (Fig. 4B). It was labeled after cleavage of fibronectin from iodinated cells (data not shown) indicating that it, like the 200-kd fragment, is exposed outside the cell. The absence of carbohydrate and enrichment of half-cystine in the 25-kd fragment is consistent with the conclusions drawn above concerning distribution of these residues. Since there are 40 to 48 half-cystines per monomer of fibronectin and 50 to 66% of these (20 to 32) are in the 30-kd region removed by trypsin, the released fragment, most of which is contained in the 25-kd fragment, could be around 10% half-cystine.

Presence of Free Sulfhydryl
Groups--In a preliminary attempt to detect sulfhydryl groups, fibronectin was reacted with ["Hliodoacetic acid (see "Experimental Procedures"). Nomeduced fibronectin in 8 M urea incorporated some radioactivity. After reduction, incorporation of iodoacetate increased 20-fold. This indicates a maximum level of 2 cysteines per chain in nomeduced fibronectin.
Insufficient radioactivity was available for analysis of labeled residues to assure specificity of labeling. However, the efficiency of labeling did not vary significantly over the pH range 7.4 to 11. In any case, it is clear that the large majority of the half-cystines in fibronectin are in disulfide bonds. it much more likely that the 19-kd band represents anomalous migration of an intrachain disulfide-bonded 25-kd monomer.
Scanning of autoradiograms showed that the 25kd frag ment was rich in half-cystines (it consistently labeled as well as or better than the 200-kd band) and it contained no To verify the presence of free sulfhydryl groups on fibronectin we have used an activated thiol-Sepharose 4B column which covalently binds proteins containing free sulfhydryl groups. Fibronectin binds to this column and can be eluted with 10 mM dithiothreitol (Fig. 5). Pretreatment of the column with NBz& did not prevent binding of fibronectin while pretreatment of fibronectin with NBz& blocked its binding to the column (Fig. 5). When the partial tryptic digest fragments were loaded on this activated thiol-Sepharose column, the 200-kd fragment bound while the 25-kd fragment was in the flow through (Fig. 6).   To determine the number of free sulfhydryl groups per fibronectin molecule, we reacted hamster fibronectin with 1 mM NBzS2 in 8 M urea during purification on gelatin-sepharose (see "Experimental Procedures"). Excess NBz& was then dialyzed away and the protein concentration and the increase in absorption at 412 nm after reduction of the NBz&-reacted fibronectin were measured. Fig. 7 shows the results of a typical experiment.
Calculation showed the presence of one free sulfhydryl group per 230,000 daltons of protein. Lower values (0.35 to 0.51) were obtained after reaction with NBz& in 4 M urea, suggesting incomplete reaction of the sulfhydryl groups at this lower urea concentration.
In summary, the activated thiol-Sepharose quantitatively binds both intact fibronectin and the 200-kd fragment indicating the presence of free sulfhydryl groups in each of these molecules. Quantitation shows one (and not more than two) sulfhydryl groups per polypeptide chain of fibronectin. This free sulfhydryl is located in the 200-kd region. Biological Relevance of Sulfhydryl Groups and Disulfides-It was previously shown that intact disulfide bridges in fibronectin are necessary for its biological activity when added Left half is from nonreduced medium, right half is from reduced. S indicates starting sample. All samples were reduced for electrophoresis. Note that in the nonreduced sample, fibronectin is all retained and elutes with urea (!fruck 22) whereas, in the reduced sample, it all comes in the flow-through and none is retained (Track 24). to transformed cells, and for its retention at the surfaces of normal cells (19). One interaction possibly involved in biological activities of fibronectin is its interaction with collagen (12)(13)(14). As fibronectin binds to gelatin (denatured collagen), we have used this property to investigate the requirement for intact disulfide bonds for gelatin binding. Aliquots of culture medium containing [%]cystine-labeled fibronectin were run on gelatin-Sepharose columns with or without prior reduction with 50 mM dithiothreitol.
Bound fibronectin was eluted with 4 M urea. In the nomeduced sample, all the fibronectin was bound; after reduction, practically none bound (Fig. 8). Hence the disulfide bonds are required for efficient binding of fibronectin to gelatin and, presumably, to collagen. The gelatin binding site is situated in the 200-kd fragment which can be purified on a gelatin-Sepharose column (not shown). NIL cells will bind fibronectin added to the culture medium (10,11,23). We have purified [35S]methionine-labeled fibronectin, blocked one part of the sample with N-ethylmaleimide and added it to NIL8 cells. Samples from media and cell lysate were then counted. There was a lo-fold decrease in the binding of fibronectin treated with MalNEt in comparison with un- [""SlMethionine-labeled fibronectin, treated or untreated with N-ethyl maleimide, was added in 2 ml of medium to 3-cm dishes of NIL8 cells. Samples from the medium and from the cells (lysed in 220 ~1) were counted.
A treated samples (Fig. 9). A similar result was obtained after alkylation with iodoacetic acid. These results indicate that blocking the free sulfhydryl groups on fibronectin interferes with the binding of fibronectin to the cell surface. DISCUSSION The result,s described above are incorporated into models for the structure of hamster cell fibronectin which are depicted in Fig. 10. Three distinct regions within each chain of the dimer are identified and can be separated from each other by partial proteolysis. These are: 1) a 200-kd region containing most or all the carbohydrate and less than half the cystine residues but all the cysteine. 2) a 25-kd region which is carbohydrate-free and cystine-rich but which contains no sulfhydryl groups, and 3) a short region (5 10 kd) which contains all the interchain disulfide bond(s). Various reports in the literature are consistent with this general picture. Fibronectins, both the cellular form (20,21,31) and the plasma form, cold-insoluble globulin (32,33), are disulfide-bonded dimers. Proteolytic digestion of plasma fibronectin releases a 200-kd fragment (34,35) and a 27-to 29-kd fragment (35,36) neither of which contains the interchain disulfides. These results are very similar to some of those reported here for cellular fibronectin.
We are not able to decide conclusively the order of these three regions but presently favor that shown in Fig. 10A  This result would favor the model shown in Fig. 10B. Further work will be required to resolve this discrepancy.
Our results bear more directly on the question of distribution of constituents between the different domains. We have presented evidence that at least 66 to 80% of the carbohydrate is located on the 200-kd fragment, but some of our results suggest that all of the carbohydrate may be in this part of the molecule. There is none detectable on the 25-kd fragment. All the carbohydrate appears to be of the complex asparaginelinked type since none is sensitive to endoglycosidase H (38)" or to mild alkaline digestion (39). Carter et al. (38) report a single class of glycopeptide of 2,000 daltons after pronase digestion which, on the basis of 5 to 7% carbohydrate, suggests five to seven side chains. In any event, the carbohydrate is concentrated, perhaps exclusively, in the 200-kd region and may be responsible for its relative resistance to proteolysis. Consistent with this supposition, Olden et al. (39) report that fibronectin which is synthesized in the presence of tunicamytin and lacks carbohydrates is more readily degraded in culture.
The asymmetry of the glucosamine/amino acid ratio across the fibronectin band (Fig. 3) indicates that some chains contain more carbohydrate than others. This has also been suggested for amniotic fluid fibronectin (40). There are reports of double bands for fibronectin, particularly for the plasma form (32,34,36,41,42) but also for the cellular form (41,43). These bands could differ in their carbohydrate content. This brings us to a comparison of the present results on cellular fibronectin with those on plasma fibronectin. We have reported elsewhere that fibronectins from these two sources show differences in specific activity for attachment of transformed cells (23) and recently Yamada and Kennedy (42) have also reported differences in biological activity. In comparisons of hamster cellular and plasma fibronectin, we consistently observe that the plasma form generates two well separated bands after reduction; whereas the cellular form runs as a single band or a very closely spaced doublet. Our results (Fig. 3) suggest that the two members of this doublet from cellular fibronectin may differ in carbohydrate content. The same might be true for the two bands of the plasma form. During tryptic digestion, the difference between cellular and plasma forms persists in the 200-kd fragments (42).4 Whether this difference in structure of cellular and plasma fibronectins represents a difference in primary sequence or in post-translational modifications, such as glycosylation, remains unclear. It is clear that the overall structures are rather similar.
Whereas the carbohydrate appears concentrated in the 200kd fragment, the cystine residues of cellular fibronectin are enriched in the other domains of the molecule. The 200-kd fragment which comprises 87% of the molecule contains only 34 to 50% of the half-cystine. This leaves 50 to 66% of the halfcystines (20 to 32 residues) concentrated in 30,000 daltons of each chain. Thus, while fibronectin itself is not particularly enriched in half-cystine (2.3 mole %) this small region is highly enriched (around 10 mole %). This concentration of disulfides in a restricted region is reminiscent of the disulfide knot region of fibrinogen (44). The interchain disulfides appear to be even more localized, being within 10,000 daltons of the end of the molecule. The 25-kd piece probably contains most of the intrachain disulfides. Its behavior on gels (Fig. 4) suggests that it contains significant amounts of intrachain disulfide bonding, and it labels relatively heavily with cystine.
There appears to be a single sulfhydryl group per chain, since all the fibronectin and 200-kd fragment are retained on an activated thiol-Sepharose column and estimates using Ellman's reagent lead to values of one sulfhydryl per chain. Mosesson et al. (32) failed to detect a sulthydryl group in human plasma fibronectin using NBz&. We have been able to detect one both by NBzSz and by the thiol-Sepharose column.5 The single sulfhydryl of cellular fibronectin is located on the 200-kd fragment (Fig. 6).
The presence of sulfhydryl groups on fibronectin is of considerable interest. Cell surface fibronectin is present both as dimers and in high molecular weight aggregates. Both dissociate to release monomeric libronectin on reduction and several other iodinatable polypeptides also dissociate from these complexes (20). On the basis of these earlier results it is not possible to decide whether these complexes represent a noncovalent association of disulfide-bonded fibronectin with other molecules or intermolecular disulfide bonding. The presence of free sulfhydryl groups on dimeric fibronectin makes it possible, even likely, that tibronectin forms intermolecular disulfide bonds, either with itself or with other cell surface proteins. Consistent with this idea is the observation that blocking of the sulfhydryl groups with MalNEt or iodoacetic acid interferes with binding of fibronectin to cells, although this result could reflect either steric hindrance or interference with intermolecular disulfide bonding. Hence it appears from these results and earlier ones (19) that disulfide bonding of fibronectin is important for attachment and function of this protein at the cell surface.
The structural analysis reported here has identified a domain structure for fibronectin with different regions being enriched for carbohydrate and disulfide bonds and has allowed us to locate the gelatin binding site and the free sulfhydryl group in the 200-kd fragment. Further analysis along these lines should allow location of the several different biologically important binding sites on tibronectin. These include binding sites for cells, collagen, fibrin (33,36,(45)(46)(47), and sulfated glycosaminoglycans or proteoglycans (48,49).

D D Wagner and R O Hynes
sulfhydryl groups.