Further Locallization of the Gelatin-bindimg Determinants within Fibronectin ACTIVE FRAGMENTS DEVOID OF TYPE I1 HOMOLOGOUS REPEAT MODULES*

Digestion of a 42-kDa gelatin-binding fragment (GBF) of fibronectin with pepsin followed by affinity chromatography on gelatin-Sepharose produces three fractions, a drop-through non-binding fraction, a retarded fraction that is dominated by a 13-kDa fragment whose NH2 terminus is identical to that of 42-kDa GBF, and a binding fraction that contains a homogeneous fragment of apparent mass 21 kDa with an NH2 terminus corresponding to Arg484. This 21-kDa GBF binds repeatedly to gelatin-Sepharose, eluting near 2.6 M in a urea gradient. It also binds in the fluid phase to a fluorescent-labeled collagen peptide with Kd = 10 microM and inhibits the binding of 42-kDa GBF to the same peptide with KI = 7.3 microM. Thus, major gelatin-binding determinants of fibronectin are located within a 21-kDa region that contains two type I homologous "finger" modules and is devoid of the type II "kringle-like" modules that were previously thought to be essential for this activity.

Digestion of a 42-kDa gelatin-binding fragment (GBF) of fibronectin with pepsin followed by affinity chromatography on gelatin-Sepharose produces three fractions, a drop-through non-binding fraction, a retarded fraction that .is dominated by a 13-kDa fragment whose NH2 ter~minus is identical to that of 42-kDa GBF, and a binding fraction that contains a homogeneous fragment (of apparent mass 21 kDa with an NH2 terminus corres:ponding to This 21-kDa GBF binds repeatedly to gelatin-Sepharose, eluting near 2.6 M in a urea gradient. It also binds in the fluid phase to a fluorescent-labeled collagen peptide with K d = 10 p M and inhibits the binding of 42-kDa GBF to the same peptide with Klr = 7.3 pM. Thus, major gelatinbinding determinants of fibronectin are located within a 21-kDa region that contains two type I homologous "finger" modules and is devoid of the type I1 "kringlelike" modules that welre previously thought to be essential for this activity. Fibronectin is a large glycoprotein that mediates the adhesion of numerous types of cells to various surfaces (1-3). It has two nearly identical polypeptide chains, each comprised of multiple domains that recognize other macromolecules such as collagen, heparin, and. fibrin. The collagen binding domain is located near the NH:2 terminus within an approximately 40-kDa region that contains four type I and two type I1 homologous repeat units or modules (Fig. 1) (4, 5). The type I "finger" modules are also found in other domains of fibronectin, whereas the type I1 "kringle-like" modules are unique to the gelatin binding dlomain. It is thus logical to suspect that the type I1 modules are involved in the interaction with gelatin. Further evidence for this idea was provided by Owens and Baralle (6) based on experiments with fusion proteins expressed in Escherichia. coli. This work also concluded that a short stretch of 14 amino acids linking the second type I1 with the adjacent type I i.s critical for gelatin binding activity. * Supported in part by Grant HL21791 from the National Heart, Lung, and Blood Institute. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solelly to indicate this fact.
$ To whom correspondence should be addressed Biochemistry Laboratory, American Red Cross Biomedical Research & Development, 15601 Crabbs Branch Way, Rockville, MD 20855. We have worked extensively with a 42-kDa gelatin-binding fragment (GBF)' derived from chymotryptic and/or thermolytic digests of fibronectin (7-10). During the course of large scale preparation of this fragment, small quantities of an approximately 21-kDa gelatin-binding fragment were obtained as a by-product. Although there is evidence in the literature for the existence of gelatin-binding fragments smaller than 40 kDa, the binding properties and exact position of these fragments within the fibronectin molecule were not determined (11,12). It thus appeared worthwhile to further purify and characterize the 21-kDa GBF and to attempt to generate it under controlled conditions. It is shown here that the 21-kDa fragment is comprised of two type I finger modules located in the COOH-terminal region of 42-kDa GBF, a region distinct from that which was concluded by Owens and Baralle (6) to be critical.

MATERIALS AND METHODS
Pepsin and thermolysin were obtained from Sigma. Fibronectin was purified by affinity chromatography on gelatin-Sepharose according to method B of Miekka et al. (13), starting with a side-fraction generated during large scale preparations of antihemophilic factor from human plasma. Fragments were generated by digestion of 1.5 g of purified fibronectin with thermolysin according to Zardi et al. (14). The 42-and 21-kDa gelatin-binding fragments were recovered by affinity chromatography on gelatin-Sepharose and further purified by exclusion chromatography on Bio-Gel P-60 (Bio-Rad). Small amounts of larger GBFs were removed from 42-kDa GBF by adsorption on heparin-Sepharose (10). A 21-kDa GBF was also obtained by digestion of 42-kDa GBF with pepsin as described in the text. The binding of these smaller fragments to gelatin is weaker than that of 42-kDa GBF, and the best yields are obtained with freshly prepared gelatin-Sepharose. Concentrations of 42-kDa GBF are based on an extinction coefficient of 7.3 X lo' M" cm" (10). The corresponding value for 21-kDa GBF is 2.7 X lo' M" cm", calculated by the method of Edelhoch (15) assuming 3 Trp, 7 Tyr, and 4 disulfides. Unless otherwise noted, all experiments were performed in 0.02 M Tris, 0.15 M NaCl, pH 7.4 (TBS).
Analytical affinity chromatography was performed as described by Isaacs et al. (10) using gelatin-Sepharose prepared according to Miekka et al. (13). SDS-PAGE was performed in the Pharmacia Phast system using 8-25% gradient gels followed by Coomassie Blue staining. NH2-terminal sequences were determined on a Beckman model 890M automated sequenator, using a Waters high performance liquid chromatography system equipped with a Zorbax ODS C-18 column to analyze the phenylthiohydantoin-derivatives.
Fluorescence measurements were made at 25 "C with an SLM-8000C fluorometer as described previously (9). The collagen peptide used in the titration experiments was obtained by digestion of type I calf-skin collagen (Sigma) with CNBr (1% w/v, 0.1 M HCl, 4 h, 30 "C). After dialysis against 0.1 M NaHC03, pH 8, the mixture was labeled with fluorescein isothiocyanate as described previously (9) and applied to a column of immobilized 42-kDa GBF at 40 'C. A fluorescent peptide having an apparent mass of 70 kDa on SDS-PAGE was preferentially adsorbed under these conditions. It was eluted with 6 M urea, dialyzed into TBS, and stored at -70 "C for later titrations. Titration of dilute (lo-* M) solutions of this peptide with fibronectin or gelatin-binding fragments produced a concentration-dependent increase in anisotropy, AA, which was fit to the following equation: where [GBF] = total concentration of fragment and Kd is the disso-The abbreviations used are: GBF, gelatin-binding fragment; TBS, Tris-buffered saline; SDS-PAGE, sodium dodecyl sulfate-polyacrylamide gel electrophoresis.
Gelatin-binding Determinants within Fibronectin ciation constant (9). The quantity Urnax was treated as a fitting parameter because the amount of fragment added is seldom sufficient to give complete saturation.
Titration with 42-kDa GBF in the presence of fixed amount of 21-kDa GBF produced apparent inhibition from which an inhibition constant, K,, was calculated as follows: where K; is the apparent dissociation constant for 42-kDa GBF in the presence of 21-kDa GBF at concentration [I], and K d is the dissociation constant in the absence of 21-kDa GBF. This equation is analogous to that used in enzyme kinetics to describe the effect of a competitive inhibitor on the K, of an enzyme for its substrate. In the present case, K, provides an estimate of the dissociation constant for the binding of 21-kDa GBF to the fluorescent gelatin peptide. Fig. 2A illustrates the time course of digestion of 42-kDa GBF with pepsin at pH 2.5 and 35 "C as analyzed by SDS-PAGE. Within 5 min, the parent fragment is converted into smaller ones, two of which appear to resist further proteolysis. The largest of these has an apparent size of 24 kDa prior to reduction and 21 kDa after reduction. The smaller has an apparent mass of 13 kDa before and after reduction. The appearance of these fragments in SDS-PAGE and their failure to decompose after reduction of disulfides suggests that a major pepsin cleavage site is located between disulfide-bonded modules (Fig. 1).

RESULTS
Application of a 10-min pepsin digest of 42-kDa GBF to a column of gelatin-Sepharose results in the production of three fractions, as shown in Fig. 3A. The first is a non-binding fraction containing numerous fragments of varying size (Fig.  2B, lanes 5). The second is a retarded fraction containing primarily the 13-kDa fragment (Fig. 2B, lanes 6). The third is a binding fraction that is essentially homogeneous, being comprised of the 21-kDa fragment, hereafter referred to as 21-kDa GBF (Fig. 2B, lanes 7). NH2-terminal sequence analysis of this material produced a single sequence beginning at Arg4%, between the second and third type I modules of the Single-letter codes are given within the circles, except for cysteines which are filled in to accentuate disulfide bonds. Shaded circles represent deletions in the alignment of type I modules. Arrows indicate cleavage sites for pepsin. Carbohydrates are represented by filled squares. The dashed line reflects uncertainty in the exact length of this fragment, which terminates a t Trp5w in the chymotrypsin and plasmin-generated versions (5). The brackets in the center of the molecule designate the region that was deemed critical for gelatin binding by Owens and Baralle (6). In contrast to the published cDNAderived sequence of human fibronectin (24), we consistently find an arginine a t position 380, in agreement with the bovine (5) and rat (25) species.  (Table I, Fig. 1, arrow 1 ). The appearance of a single sequence for the nonreduced fragment indicates that the apparent decrease in mass from 24 to 21 kDa upon reduction is not due to the presence of a small disulfideassociated peptide. The 13-kDa fragment has an NH2-terminal sequence identical to that of the parent 42-kDa GBF (Table I). Also shown in Fig. 2B (lanes 8) is the 21-kDa GBF obtained as a byproduct while preparing 42-kDa GBF from a thermolytic digest of whole fibronectin. It has an NH2 terminus beginning a t Leu483 (Table I). An approximately 16-kDa fragment purified in small quantities from pool I by reverse-phase chromatography had an NH2-terminal sequence beginning a t Val377, between the 2 type I1 modules (Fig. 1, arrow 2). However, a major portion of this fragment was also cleaved within a disulfide loop between Met462 and Met463 (Fig. 1, arrow 3), perhaps contributing to its lack of affinity for gelatin. Fig. 3B illustrates the results of analytical scale affinity chromatography of the three pools from Fig. 3A. The 21-kDa GBF re-binds quantitatively to gelatin-Sepharose, eluting near 2.6 M in a urea gradient. The nonbinding fraction, Pool 1, passed through the analytical column unretarded. The elution of Pool 2, consistingprimarily of the 13-kDa fragment, was again retarded relative to the nonbinding fraction. Thus, its failure to bind more strongly cannot be attributed to insufficient capacity of the preparative column. It would appear that the 13-kDa fragment has a weak affinity for gelatin.
The results of titration experiments designed to quantitate the affinity of 21-kDa GBF for gelatin in the fluid phase are shown in Fig. 4. These experiments utilized a fluoresceinlabeled CNBr fragment of calf-skin collagen whose fluorescence anisotropy increases when complexed with fibronectin or gelatin-binding fragments. Titration of this material with 42-kDa GBF (upper curve) produced a dose-dependent increase in anisotropy which was consistent with a dissociation Heterogeneous, two sequences differing by 1 amino acid, present in 2:l ratio. constant of 0.8 p~, about twice that reported previously for binding of 42-kDa GBF to a fluorescein-labeled a1 chain of rat type I collagen under similar conditions (9). Titration with 21-kDa GBF (lower curue) produced a smaller increase, consistent with its smaller size, with Kd = 10.3 pM. The data shown in Fig. 4 were obtained with the thermolytic 21-kDa GBF. Similar titrations using the pepsin-generated 21-kDa GBF produced Kd values of 3.1, 7.4, and 33 ~L M (average 14.5 p~) .
Addition of 42-kDa GBF to a solution which had already been titrated with 21-kDa GBF produced a further increase in anisotropy that was also hyperbolic (Fig. 4, middle curue). The apparent Kd in this case was 7.2 pM, almost 10-fold greater than obtained for 42-kDa GBF in the absence of any competitor. From these data one calculates an inhibition constant, KI = 7.3 p~ (see "Materials and Methods"). Thus, 21-kDa GBF competes directly with 42-kDa GBF for binding to the collagen peptide, presumably at the same site (or sites), with an inhibition constant similar to the dissociation constant obtained by direct titration.

DISCUSSION
The 42-kDa gelatin-binding fragment of fibronectin is relatively resistant to proteolysis, and efforts to produce smaller gelatin-binding fragments have previously been unsuccessful. Calorimetric investigations to be published elsewhere have shown that 42-kDa GBF exhibits a reversible thermal transition whose midpoint is sensitive to pH.' The temperature used for the pepsin digestion is close to the midpoint of the thermal transition at that pH. Under these conditions, 42-

Gelatin-binding Determinants within Fibronectin
kDa GBF is rapidly cleaved by pepsin to produce two fragments, one of which retains the ability to bind tightly to gelatin-Sepharose. Titration experiments showed that the affinity of this 21-kDa fragment for a fluorescent collagen peptide was about 10-fold lower than that of the parent 42-kDa GBF, which itself binds such peptides 10-20-fold weaker than whole fibronectin (9). It is possible that the recognition site has a different conformation in the smaller fragment or that residues not present in 21-kDa GBF contribute to the stronger binding of 42-kDa GBF. A hint of the latter possibility was provided by the observation that a 13-kDa fragment derived from the opposite end of 42-kDa GBF appeared to interact weakly with the gelatin-Sepharose column. Using the expression AG = -RT In K , one can show that 80% of the free energy associated with the binding of 42-kDa GBF to gelatin in solution can be accounted for by the 21-kDa fragment. An additional site which by itself bound gelatin with a K d of only 30 mM would be sufficient to restore the difference. This additional binding would presumably involve a separate site in close proximity to that occupied by the 21-kDa fragment. The possibility of multiple contact points between a collagen chain and the larger gelatin-binding fragments of fibronectin would be consistent with the extensive internal sequence homology in both reactants. There is ample evidence in the literature for multiple sites of varying affinity on collagen chains (Ref. 9, and references therein).
The 21-kDa GBF is comprised of two type I finger modules, the first of which contains two carbohydrate moieties (5). This is consistent with the observation that increased glycosylation of placental fibronectin weakens the affinity for gelatin (16). Such carbohydrate may also account for the aberrant behavior of 21-kDa GBF on SDS-PAGE. The apparent mass is about 25% higher than the value of 16.5 kDa estimated from the amino acid composition plus two carbohydrate moieties of 2 kDa each (17). Furthermore, reduction of this fragment causes an unusual decrease in apparent mass of about 3 kDa, despite the lack of evidence for a disulfidelinked peptide, i.e. only one NH2-terminal sequence was obtained for both the thermolytic and the pepsin-generated 21-kDa fragments. This decrease, also observable in the 42-kDa precursor, is not a general feature of type I finger containing fragments since the NH2-terminal 29-kDa heparin/fibrinbinding fragment displays a substantial increase in apparent mass upon reduction (14). It is possible that disruption of disulfide bonds in 21-kDa GBF alters the interaction between the carbohydrate groups and the protein in such a way as to alter the amount of bound detergent during SDS-PAGE. The most surprising result of our study was the finding that the type I1 modules are not essential for gelatin binding activity. These modules are unique to the gelatin binding region of fibronectin and are commonly thought to be responsible for the interaction with collagen. This notion is reinforced by the observation of similar type I1 modules in two other proteins that are devoid of type I fingers and can be purified by affinity chromatography on gelatin-Sepharose. These are type IV collagenase (18,19) and bovine seminal plasma protein (20, 21), whose type I1 modules respectively exhibit approximately 50 and 33% identity with those of fibronectin. Type I1 modules are also found in coagulation factor XI1 (22) and mannose-6-P04 insulin-like growth factor I1 receptor (23), for which gelatin binding properties, if any, have not been documented. In our hands, factor XI1 had no activity when tested by assays of the kind in Fig. 4 (not shown). Nor do we have a problem with factor XI1 contamination of fibronectin prepared from plasma by affinity chromatography on gelatin-Sepharose. Thus, the mere presence of type I1 modules does not guarantee gelatin-binding activity, as first concluded by Owens and Baralle based on the expression in E. coli of fusion proteins containing fibronectin type I1 modules (6). However, our 21-kDa GBF is one finger removed from the region reported by these workers to be critical (see Fig. 1). We have no obvious explanation for this discrepancy. The conclusions of Owens and Baralle (6) were based on the ability of fusion proteins to bind to gelatin-Sepharose. The yield of active protein in these experiments was very low and problems of proteolytic degradation and improper folding could produce false-negative results. Furthermore, constructs coding for the region identified here as being important were not part of their series. Future work utilizing site-directed mutagenesis to delineate specific residues involved in gelatin-binding should not ignore the 21-kDa region. It is possible that several sites spread throughout the 42-kDa region, and perhaps beyond, are required to confer maximum affinity for gelatin.