In Vitro Formation of Disulfide-bonded Fibronectin Multimers*

Fibronectin purified from a plasma protein side frac- tion in the absence of denaturant contained 1.5 to 1.9 cryptic free sulfhydryl groups per 200- to 250-kDa sub- unit. Exposure of sulfhydryl groups in physiologic salt solutions required at least 1 M guanidine, and 3 M gua- nidine was required for optimal exposure. “he sufiy-dry1 groups were not exposed by collagen, a fibronec- tin-binding collagen fragment, fibrinogen, heparin, hyaluronic acid, calcium ion, EDTA, deoxycholate, or methylamine. One- and two-dimensional gel electrophoresis indicated that a molecule of in and fibro- nectin. In addition, traces of disulfide-bonded multimers were present in preparations of purified fibronec- tin. The proportion of fibronectin in disulfide-bonded multimers increased in guanidine-containing solutions. Compared to dimeric fibronectin, these multimers had solubility in physiologic buffers, could be read- ily cross-linked by Factor XI&, and exhibited altered tryptic susceptibility. In free sulfhydryl groups were blocked by prior alkylation with iodoacetamide, fibronectin did not form disulfide-bonded multimers in guanidine-containing solutions. The patterns of altered tryptic susceptibility and cyanide cleavage suggested that multimer formation both of

tryptic susceptibility. In free sulfhydryl groups were blocked by prior alkylation with N-ethylmaleimide or iodoacetamide, fibronectin did not form disulfidebonded multimers in guanidine-containing solutions. The patterns of altered tryptic susceptibility and cyanide cleavage suggested that multimer formation is mediated by both sulfhydryls of fibronectin. The transition from dimeric to multimeric fibronectin can serve as a model for the formation of disulfide-bonded fibronectin multimers in the extracellular matrix.
Fibronectin is a glycoprotein which exists in soluble and insoluble forms. Plasma fibronectin and fibronectin synthesized and secreted by cultured cells are largely 400-to 500-kDa disulfide-bonded dimers of similar 200to 250-kDa subunits (1-4). The interchain disulfide bridges responsible for linking the subunits together are at the carboxyl terminus ( 5 ) . There are also a number of intrachain disulfides which create small homologous loops (5). Finally, there are probably two free sulfhydryl groups per subunit (6,7). These free sulfhydryl groups do not react with sulfhydryl reagents in physiologic saline (7). Insoluble fibronectin in cell culture exists as disulfide-bonded multimers as well as dimers (8-14). Multimer formation presumably occurs by oxidation of free sulfhydryl groups because fibronectin can be released from the matrix of cultured cells by treatment with sulfhydryl-containing re-* This work was supported by Grants HL 21644 and 24885 from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertzsement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

+ Recipient of an Established Investigatorship from the American
Heart Association and its Wisconsin Affiliate. agents (15,16) and modification of fibronectin with N-ethylmaleimide at alkaline pH prevents binding to the cell layer (6). Multimeric fibronectin extracted from the matrices of cultured cells is more active than plasma fibronectin in causing cultured transformed cells to assume a normal shape (17), in the agglutination of trypsinized fixed red blood cells (17), in the binding of hyaluronic acid (la), and in the promotion of growth by primary fibroblasts in culture (19).
The observations described above suggest that the conversion of fibronectin from a disulfide-bonded dimer to multimer is an important event. As described in the present paper, we have examined the sulfhydryl groups and subunit composition of human plasma fibronectin in some detail in order to gain insight into factors which may control the transition from dimer to disulfide-bonded multimer and to learn how to make disulfide-bonded multimers in vitro.

EXPERIMENTAL PROCEDURES
Materials-Human Factor XI11 and fibrinogen were prepared as described previously (20). Rabbit anti-human fibronectin sera were pooled from rabbits immunized with 100 pg of 200-kDa fibronectin subunit (prepared by preparative polyacrylamide slab gel electrophoresis (21)) in complete Freund's adjuvant injected subcutaneously in multiple dorsal sites. The rabbits were boosted a month later with an identical injection as above, then monthly with two boosts of 100 pg of 200-kDa fibronectin in Tris-buffered saline and finally two boosts of 100 pg of fibronectin in Tris-buffered saline. Human thrombin was a gift from Dr. John Fenton I1 (New York State Department of Health, Albany, NY), type I11 collagen and cyanogen bromide fragment 7 of the al(1) chain of type I collagen (al(I)-CB7) were gifts from Dr. Hynda Kleinman (National Institute of Dental Research, Bethesda, MD), and monoclonal antibody to the 31-kDa fragment of early tryptic digests of fibronectin (7, 22) was a gift from Drs. Dennis Smith and Leo Furcht (University of Minnesota Medical School, Minneapolis, MN). The following were purchased heparin, iodoacetamide, and hyaluronic acid, Sigma; DEAE-cellulose (DE-52), Whatman; fluorescein isothiocyanate-labeled goat anti-rabbit IgG and rabbit anti-mouse IgG, Cappel; guanidine hydrochloride, Schwarz/Mann; 1-tosylamido-2-phenylethyl chloromethyl ketone-treated trypsin, Worthington; 5,5"dithiobis-(2-nitrobenzoic acid) and N-ethylmaleimide, Pierce; and 2,a"dipyridyl disulfide, Aldrich.
Purification of Fibronectin-Fibronectin was purified from fibronectin-and fibrinogen-rich protein side fractions of human Factor VI11 production as follows (Table I). Approximately 80 g of frozen paste (provided by Dr. Mike Hrinda, Revlon Health Care Group, Tuckahoe, NY) were dissolved for 120 min at 22 "C with constant stirring in 600 ml of 0.01 M Tris, 0.4 M sodium chloride, pH 7.4. Undissolved material was removed by centrifugation. Fibrinogen was precipitated by heating at 56 "C for 3 min and removed by centrifugation. The supernatant was dialyzed against 4 liters of 0.01 M Tris, 0.01 M sodium chloride, pH 7.4, and applied to a column (6.5 X 13.5 cm) of DEAE-cellulose which had been equilibrated with 0.01 M 'rris, 0.07 M sodium chloride, pH 7.4. After the passthrough peak was eluted, a 2-liter gradient of 0.07 to 0.30 M sodium chloride in 0.01 M Tris, pH 7.4, was applied to remove bound fibronectin (Fig. 1). The fibronectin was precipitated by dialysis against 20% (w/v) ammonium sulfate and dissolved in and dialyzed against Tris-buffered saline. Portions were snap-frozen at concentrations of 15-20 mg/ml and stored at -70 "C.
In an alternate purification, the supernatant after heat precipita-6596

Disulfide-bonded Fibronectin Multimers
tion was passed over a column of gelatin agarose, and bound fibronectin was eluted with 1 M sodium bromide at pH 5.0 (20).
Polyacrylamide Slab Gel Electrophoresis-Electrophoresis in one dimension was performed using the discontinuous slab system of Ames (23). In some experiments, a gradient separating gel was formed by rapidly pouring gels whose bottom third was 12%. middle thud was 8%. and top third was 4% acrylamide. Nonreduced-reduced twodimensional electrophoresis was performed as described elsewhere (24). Unless indicated otherwise, slab gels were stained with Coomassie brilliant blue to visualize the protein bands.
Electroblotting (25) was done in an apparatus purchased from Bio-Rad Laboratories. After the protein was transferred to nitrocellulose paper by electroblotting, replicate sections were analyzed by protein  Fig. 1 on which the samples were analyzed.
Calculated from the concentrations of fibronectin (determined by immunoassay) and total protein (determined by assuming that contaminating proteins have the same extinction coefficient as fibronectin).
' Estimated by densitometry of polyacrylamide gels. staining and immunostaining. One section of paper was stained with 0.1% naphthol blue black in 45% methanol and 10% acetic acid and destained in 45% methanol and 10% acetic acid. A replicate section of paper was soaked in Tris-buffered saline containing 3% bovine albumin for 1 h at 37 "C, rinsed in Tris-buffered saline, and soaked overnight in Tris-buffered saline containing 3'i; albumin, lo$ fetal calf serum, and either 2% rabbit anti-human fibronectin antiserum or mouse monoclonal anti-fibronectin, 50 pg/ml. The section was washed in Tris-buffered saline and soaked for 1 h in Tris-buffered saline containing 3% albumin, 10% fetal calf serum, and I? fluorescein isothiocyanate-conjugated goat anti-rabbit IgG or rabbit anti-mouse IgG. The section was washed with Tris-buffered saline, and fluorescent bands were photographed.
Silver staining of polyacrylamide slab gels were done by the procedure of Oakley (26) with modifications described by Giulian et al. (27).
The molecular size markers included: fibronectin dimer, 400 kDa; fibronectin monomer, 200 kDa; phosphorylase, 93 kDa; bovine serum albumin, 68 kDa; ovalbumin, 43 kDa; chymotrypsinogen, 24.5 kDa: and hemoglobin, 16.5 kDa. The sizes of fibronectin are based on previous experiments in which the apparent size of reduced plasma fibronectin was determined by comparison to the heavy chain of myosin and the subunit of Factor VIII-related antigen/von Willebrand factor (28). In this analysis, the sizes assigned to myosin and Factor VIII-related antigen/von Willebrand factor were literature values for the subunit sizes of these proteins as determined by sedimentation equilibrium in 6 M guanidine.
Other Methods-Sulfhydryl groups were quantified with 5,5'-dithiobis-(2-nitrobenzoic acid) or 2.2"dipyridyl disulfide as described elsewhere (7) except that titrating reagent was added prior to addition of guanidine and the molar extinction coefficient for 2-thiopyridone was assumed to be 7060 M" cm" (29). Fibronectin with blocked sulfhydryl groups was prepared by incubating the protein, 5 mg/ml. with 10 mM N-ethylmaleimide or iodoacetamide and 3 M guanidine for 2.5 h a t 22 "C. Excess alkylating agent and guanidine were removed by dialysis. For cleavage of fibronectin at free sulfhydryl groups (30) fibronectin was incubated with 0.25 mM 5,5'-dithiobis-(2-nitrobenzoic acid) and 3 M guanidine in Tris-buffered saline, pH 9.0, for 15 min at 22 "C, and then treated with 20 m~ potassium cyanide. The reaction was stopped after 30 min by adding 1 M acetic acid, and the samples were dialyzed into 0.1 M Tris, 0.15 M sodium chloride, pH 6.8.

-
Fibronectin antigen was measured by the immunoturbidometric method (31) with a kit supplied by Boehringer-Mannheim Biochemicals (Indianapolis, IN).

RESULTS
Large quantities of human plasma fibronectin were purified from a fibronectin-rich side fraction of commercial Factor VI11 production (Table I and Fig. 1). Aside from the step in which fibrinogen was precipitated by heating at 56 "C for 3 min, fibronectin was kept a t near physiologic conditions throughout the purification. Preparations with the same qualities were obtained when chromatography on gelatin-agarose with elution with 1 M sodium bromide, pH 5.0 (20), was done in place of chromatography on DEAE-cellulose.
When purified fibronectin was analyzed by polyacrylamide gel electrophoresis after reduction, 99% of the protein was in a band of 200 kDa (Table I and Fig. 1). The 200-kDa band could be resolved into 2 closely spaced of equal intensity if only 1-2 pg of protein were analyzed (32, 33). When purified fibronectin was analyzed by electrophoresis without reduction, minor bands of 180, 200, 240, 460, approximately 800, and greater than loo0 kDa were observed in addition to the major band of 400 kDa (Figs. 1 and 2). The heterogeneity of the unreduced sample was best appreciated by indirect immunofluorescent staining of electroblotted protein (Fig. 2). When whole human plasma was analyzed by the same technique, fluorescent bands of 180, 200, 240, 400, and 460 kDa were detected whereas the bands of approximately 800 and greater than lo00 kDa were not (Fig. 2).
Gel electrophoresis of purified fibronectin without reduction followed by gel electrophoresis after reduction (Fig. 3) indicated that the 460-, 400-, and 240-kDa bands in the nonreduced dimension were composed of 200-kDa subunits. The 200-and 180-kDa bands, in contrast, fell on the diagonal. On the basis of this analysis, we hypothesize that a molecule of 40-60 kDa is disulfide-linked to a small proportion of fibronectin molecules (see under "Discussion").
Fibronectin in Tris-buffered saline had no reactive free sulfhydryl groups whereas fibronectin in 3 M guanidine had 1.5 sulfhydryl groups per 200-kDa subunit or 1.9 sulfhydryl groups if a subunit size of 250 kDa is assumed (7). Greater than 1 M guanidine was required to expose sulfhydryl groups, and additional reactive sulfhydryl groups were not exposed at guanidine concentrations greater than 3 M (Fig. 4). Greater than 2 M urea also caused exposure of sulfhydryl groups, and 4 M urea caused optimal exposure (data not shown). A variety of substances which are known to interact with fibronectin did not cause exposure of sulfhydryl groups (Table 11). In addition, incubation with calcium ion, EDTA, deoxycholate, or methylamine did not cause exposure of sulfhydryl groups (Table 11).
When fibronectin, 7 mg/ml, was incubated in 3 M guanidine, disulfide-bonded multimers were formed in a time-and temperature-dependent manner (Fig. 5). Upon removal of guanidine by dialysis against Tris-buffered saline, a granular precipitate formed on the sides of the dialysis bag. With time, the granules coalesced to form a thick stringy precipitate (Fig. 6).
After dialysis a t 22 "C, the precipitate contained approximately 30% of the fibronectin when the protein concentration in the bag was greater than 1.5 m g / d (Table 111) and was enriched in the larger multimers (Fig. 5). After dialysis at 4 "C, the precipitate contained approximately 75% of the fibronectin (Table IV) fibronectin (right) were analyzed by electrophoresis in 2' ; sodium dodecyl sulfate on discontinuous 4 to 12% polyacrylamide slab gel. Separated proteins were transferred to nitrocellulose paper. For each pair, the strip to the left was analyzed by protein staining, and the strip to the right was analyzed by immunofluorescent staining with rabbit antifibronectin. Molecular size markers are indicated on the left. The line on the right indicates the top of the separating gel. Two-dimensional polyacrylamide gel electrophoresis in sodium dodecyl sulfate of purified fibronectin. Fibronectin, purified from a commercial concentrate on a gelatin-Sepharose column as described under "Experimental Procedures," was analyzed by electrophoresis in sodium dodecyl sulfate without reduction (.VR) in 4-mm cylindrical gels to provide separation in the first dimension (from left to right). This gel was reduced ( R ) and analyzed on the slab gel for the second dimension (from top to bottom). The u r r o w indicate protein spots of slightly larger size in the nonreduced dimension than fibronectin dimer or monomer in the nonreduced dimension. Size markers are shown on the left. CAPS-buffered saline, pH 11, and fibronectin dialyzed directly from guanidine into CAPS-buffered saline did not precipitate. Multimeric fibronectin in CAPS-buffered saline precipitated when diluted in or dialyzed against Tris-buffered saline, pH 7.4. In contrast, dimeric fibronectin in CAPS-buffered saline did not precipitate when diluted in or dialyzed against Trisbuffered saline.
Fibronectin with blocked free sulfhydryl groups did not form a precipitate after incubation in 3 M guanidine and dialysis against Tris-buffered saline (Table IV) and did not form disulfide-bonded multimers in 3 M guanidine (data not shown).
Multimeric fibronectin in Tris-buffered saline was readily cross-linked by Factor XIII, under conditions in which little cross-linking of dimeric fibronectin occurred (Fig. 7).

TABLE I1
Failure of potential modifiers to expose the free sulfiydryl groups of fibronectin in physiologic saline Samples were preincubated with the potential modifier for 60 min at 22 "C and then assayed with 5,5'-dithiobis-(2-nitrobenzoic acid), 0.1 mg/ml, a t 412 nm. The incubation with methylamine was done at 37 "C. Samples contained fibronectin, 1 mg/ml, as ascertained by the absorbance at 280 nm prior to assay. The molar extinction coefficient for p-nitrothiophenol anion and fibronectin subunit was assumed to be 13,600 M I cm I and 256,000 M cm ', respectively. Multimers were tested for ability to bind to gelatin-agarose after dilution in Tris-buffered saline, pH 7.4, to a final concentration (500 pg/ml) a t which precipitation did not occur. Under standardized conditions, 79% bound to gelatin-agarose as compared to 77% for dimeric fibronectin which had been previously dialyzed into CAPS-buffered saline, pH 11.

Free "SH
Multimers and dimers were digested differently by trypsin (Figs. 8 and 9). The differences were of fragments recognized by a monoclonal antibody to an epitope close to the more carboxyl-terminal of the two free sulfhydryls ( 7 ) . Thus, the intensities of previously described 180-, 135, 71-, and 31-kDa fragments ( 7 , 2 2 ) were much decreased both by protein staining and by immunofluorescent staining (Figs. 8 and 9). In contrast, fragments of 165, 150, and 39 kDa, which were not recognized by the monoclonal antibody, were present in digests of both multimers and dimers (Fig. 8). Fragments of 80 kDa, which have previously been shown to contain a free sulfhydryl group (7), were present in digests of both dimers and multimers (Fig. 8). We were not able to identify disulfidelinked pairs in comparisons of protein-stained gels run without and with reduction, i.e. we could not identify fragments which I 2 3 4 5 6 7 8 9 101112 400 -2 0 0 - FIG. 5. Formation of disulfide-bonded fibronectin multimers after exposure to guanidine. Fibronectin, 7 mg/ml, was incubated in Tris-buffered saline containing 3 M guanidine, pH 9.4, for 0, 5. 30, 60, 120. and 240 min a t 22 "C (lanes 1-61. for 120 min at 37 "C (lane 7). and for 120 min a t 0 "C (lane 8). A control sample which did not contain guanidine was incubated for 240 min at 22 "C (lane 9). At the end of the incubation, the sample was diluted with 2 parts of Trisbuffered saline and shortly thereafter was dialyzed against Trisbuffered saline a t 22 "C. Hence, the sample in lane I was incubated in Tris-buffered saline containing 1 M guanidine for about 15 min at 22 "C prior to dialysis. Dialysis of the 120-and 240-min samples from the incubation at 22 OC and the 120-min sample from the incubation at 37 "C produced precipitates (lanes IO, 11, and 12). The precipitates were removed with a stirring rod, rinsed with water, and suspended in the original volume of Tris-buffered saline. Samples were analyzed by electrophoresis on 6% polyacrylamide gels after dilution in an equal volume of 2% sodium dodecyl sulfate. Reduced samples analyzed by electrophoresis all had a single band of 200 kDa (not shown). In other experiments, incubation of fibronectin in Tris-buffered saline containing 1 M guanidine for 120 min at 22 "C did not result in multimer formation. The lines on the right indicate the tops of the stacking (upper) and separating (louver) gels.

6599
split into smaller fragments with reduction. This finding was confirmed by nonreduced-reduced two-dimensional gel electrophoresis, in which all of the spots of e200 k D a fell on the diagonal (data not shown). Several new fragments were recognized in digests of multimers: fragments with apparent sizes nonreduced/reduced of 54/57 and 36/37 kDa, detected b y protein staining (Fig. 8) and nonreduced of 70 and 32 kDa, detected by immunofluorescence (Fig. 9). In addition, there were increased amounts of small antigenic fragments detected by immunofluorescence at the front (~2 0 kDa) of the gel of nonreduced digests (Fig. 9).
Cyanide cleavage of fibronectin dimer and multimers resulted in patterns in sodium dodecyl sulfate-polyacrylamide gels similar to those described b y Wagner and Hynes (30). However, the yield of the principal 156-kDa cleavage component (determined by densitometry) was 5.44 in the preparation of multimers as compared to 12.7% in the preparation of dimer. Cleavage products were not detected i n the 20-to 50k D a region of the gels, even after silver staining.

TABLE 111
Concentration dependence ofprecipitation of fibronectin multimers when dialyzed againsf physiological saline Fibronectin, at the concentrations indicated, was treated with 3 M guanidine for 2 h a t 22 "C and then dialyzed extensively against Trisbuffered saline a t 22 "C. The amount of precipitate in each bag was noted, and the amount of soluble fibronectin was determined by A?,,.

TABLE IV
Dependence ofprecipitation of fibronectin multimers when dialyzed against physiological saline on the presence of free sulfhydryl groups Fibronectin, fibronectin treated with iodoacetamide (CAM-fibronectin), and fibronectin treated with N-ethylmaleimide (NEM-fibronectin) were incubated in 3 M guanidine for 2 h a t 22 "C and then dialyzed extensively against Tris-buffered saline a t 4 "C. The amount of precipitate in each bag was noted, and the amount of soluble protein was determined by A~w .
The fibronectin derivatives contain <O.l free sulfhydryl moup per 200-kDa subunit as ascertained bv assay with 5,5"dithiobis-(2-n~itrobenzoic acid) in 3 M guanidine. Titration of free sulfhydryl groups of fibronecfin dimer a n d multimer Samples contained either fibronectin dimer or multimer, 1 mg/ml, in Tris-buffered saline with or without 4 M guanidine-HCl (GnHCI) and were assayed as described in Table 11.

A.tt/Aw!rul
Free "SH  I-3). or multimeric fibronectin, 0.4 mg/ml (lanes4-6). in Tris-buffered saline were treated with nothing (lanes I and 4 ) or with thrombin, 1 unit/ml, and Factor XIII, 20 pg/ml, in the presence of 10 mM EDTA (lanes 2 and 5 ) or mM calcium ion (lanes 3 and 6). Prior to the experiments, both proteins had been dialyzed against CAPS-buffered saline, pH 11. Therefore, the proteins were diluted 8-fold in Tris-buffered saline prior to the experiment. The samples were incubated at 37 "C for 2.5 h and then treated with electrophoresis buffer containing 25 2-mercaptoethanol and analyzed on 10% polyacrylamide slab gels. Multimeric or dimeric fibronectin in CAPS-buffered saline, pH 11, was diluted 8-fold in Tris-buffered saline, pH 7.4. The final concentration of each was 1 mg/ml. The proteins were digested at 37 "C with trypsin, 1 pg/ml. At 0, 1, 5, 10, 30,60, and 120 min, samples were removed, denatured in 2% sodium dodecyl sulfate without ( N R ) or with (R) 2% 2-mercaptoethanol, and analyzed by electrophoresis on a 10% polyacrylamide slab gel (lane I , 0 min; lane 2. 1 min; etc.). Proteins were stained with Coomassie brillant blue. In other gels, not shown, we found that dimeric fibronectin that was not cycled through CAPS-buffered saline had the same bands as dimeric fibronectin that was cycled through CAPS-buffered saline. The arrous point to fragments which were found in one preparation and not in the other. . Sequential trypsinization of multimeric fibronectin as studied by immunoblotting. Multimeric or dimer fibronectin in CAPS-buffered saline, pH 11, were digested and analyzed on slab gels as described in Fig. 8. Separated proteins were transferred to nitrocellulose and detected with naphthol blue-black and by immunofluorescent staining as described under "Experimental Procedures." Immunofluorescent staining was with a mouse monoclonal antibody which has been shown previously to react with the 31-kDa fragment of early tryptic digests of fibronectin (7). Shown are the results of immunofluorescent staining. Molecular sizes are not marked, but marker standards were the same as in Fig. 8. Arrows indicate the immunoreactive fragments which were found in one preparation and not in the other. As in Fig. 8, the fragments are designated by their size nonreduced/reduced. Protein staining (not shown) indicated that all of the bands in Fig. 8 were transferred to the nitrocellulose paper.

DISCUSSION
Both purified fibronectin and fibronectin in freshly prepared plasma exhibited considerable heterogeneity when examined by polyacrylamide gel electrophoresis in sodium dodecyl sulfate without reduction (Figs. 2 and 3). Molecules of 180,200,240,400, and 460 kDa were detected. It is likely that the heterogeneity results partially from proteolysis (32). Cleavage close to the COOH-terminal interchain disulfides (5) would create fragments of approximately 200 kDa. Additional cleavage at the NH2 terminus would create fragments of 180 kDa (see Refs. 1 and 2). In order to explain the 240-and 460-kDa bands, however, it is necessary to hypothesize that something of approximately 40-60 kDa is disulfide-bonded to the 200-kDa fragments or to the intact 400-kDa dimer. We are currently trying to isolate and identify the 40-to 60-kDa molecule. Nothing in the 40-to 60-kDa region reacted with polyclonal anti-fibronectin antiserum upon immunoblotting of sodium dodecyl sulfate-polyacrylamide gels of purified reduced fibronectin, and, therefore, it is unlikely that the 40-to 60-kDa piece is a fragment of fibronectin. On the basis of negative immunoprecipitation and immunoblotting studies with an antiserum generously provided by Professor Anders Grubb (University of Lund, Lund, Sweden), we do not think that the 40-to 60-kDa molecule is HC protein (HC protein has a size of 30-kDa and is found disulfide-linked to many plasma proteins, including IgA and albumin (34)). The fact that fibronectin fragments apparently contain more of the 40-to 60-kDa molecule ( i e . the ratio of 240 to 200 kDa molecules is greater than the ratio of 460 to 400 kDa molecules) suggests that the circulating fibronectin which contains the 40-to 60-kDa molecule originates from tissues with active proteolysis. Fibronectins of approximately 800 and greater than 1000 kDa, which were present in purified preparations but not in plasma (Fig.  2), probably represent protein which was denatured during purification.
Plasma fibronectin (35), as well as fibronectin synthesized in cell culture (13), forms disulfide-bonded multimers upon incorporation into the extracellular matrix. A variety of substances known to bind to fibronectin did not expose free sulfhydryl groups (Table 11). Thus, we can offer little insight into if or how free sulfhydryl groups of fibronectin are exposed in the cell layer and participate in disulfide-bonded multimer formation. Multimer formation is a slow process (12, 13). If a small fraction of free sulfhydryl groups are exposed at any one time and the fibronectin molecules in the cell layer are aligned optimally, multimer formation could occur without a conformational change to further expose free sylfhydryl groups. Exposure could be facilitated by binding of fibronectin to two substances simultaneously (e.g. collagen and a glycosaminoglycan) or to a substance not tested in our experiments. The free sulfhydryl groups of fibronectin synthesized by cells may be exposed in a manner different from plasma fibronectin. For instance, the difference in antigenicity of the sulfhydryl-containing COOH-terminal region of hamster cellular fibronectin (36) may be accompanied by a conformational change which makes the sulfhydryl group more accessible. Alternatively, modification at a potential site of N-glycosylation which has been identified adjacent to a free sulfhydryl group in bovine plasma fibronectin (37) might make the free sulfhydryl group more accessible. Finally, disulfide multimer formation in the cell layer may not involve free sulfhydryl groups at all, but disulfide exchange of existing disulfides.
The failure of methylamine to expose free sulfhydryl groups suggests that fibronectin does not contain cysteinyl residues in thiolester linkages (as is the case with a2-macroglobulin and C3 (38, 39)).
Disulfide-bonded fibronectin multimers could be produced in solution by exposing concentrated fibronectin to 3 M guanidine at 22 or 37 "C (Fig. 5). These multimers had limited solubility in physiologic buffers, bound to gelatin agarose, and exhibited altered tryptic susceptibility in the 31-kDa COOHterminal region. Formation of such multimers may explain the limited solubility that is sometimes observed with fibronectin prepared by elution of gelatin-agarose with 4 M urea.
Previous studies have demonstrated that a 31-kDa early tryptic fragment and an 80-kDa late tryptic fragment of human plasma fibronectin contain free sulfhydryl groups (7). The 31-kDa fragment and precursors of the 31-kDa fragment were largely missing from trypsinates of multimers whereas the 80-kDa fragment and precursors of the 80-kDa fragment were present and apparently not disulfide-bonded into a larger aggregate (Figs. 8 and 9). The decreased yield of the 156-kDa fragment upon cyanide cleavage, however, suggests that there was also some oxidation of the sulfhydryl group in the 80-kDa region. Disulfide bond formation potentially could involve free sulfhydryl groups in both the 31-and 80-kDa regions, and disulfide bonds could form between the two halves of the dimers as well as between two dimers. Thus, six different disulfide arrangements are possible.' Only half of the sulfhy-dry1 groups were blocked in the multimers (Table V), and the trypsinization study suggests that the majority of the arrangements involve the sulfhydryl group in the 31-kDa fragment. The altered tryptic susceptibility suggests that exposure and oxidation of the cysteinyl residue in the 31-kDa region ( 7 ) causes a local conformation change.
The multimers, like fibronectin in the cell layer (40), were readily cross-linked by Factor XIII,. As described in the introduction, multimeric cell surface fibronectin has several biologic properties which are not shared by dimeric plasma fibronectin (17-19). We have found that multimeric plasma fibronectin is more potent than dimeric fibronectin in the promotion of growth by primary guinea pig glomerular cells in defined medium (41). Multimeric plasma fibronectin, but not dimeric fibronectin, caused aggregation of platelets in platelet-rich plasma.3 Experiments are currently underway to study whether multimeric plasma fibronectin, like cell surface fibronectin, causes transformed cells in culture to assume a more normal shape, agglutinates fixed red blood cells, and binds hyaluronic acid. It will also be interesting to learn whether multimeric fibronectin can bypass the initial binding pool (35) and become incorporated directly into the detergentinsoluble extracellular matrix when added to cultures of nontransformed fibroblasts. Thus, we believe that the multimers will be a valuable derivative for studies of the polymerization and insolubilization of fibronectin and in structure-function studies of the interactions of fibronectin with cells.