Interaction of fibronectin and its gelatin-binding domains with fluorescent-labeled chains of type I collagen.

Fluorescent probes have been used to obtain dissociation constants for the fluid-phase interaction of human plasma fibronectin and several of its gelatin-binding fragments with purified alpha chains of type I rat tail collagen, as well as with a cyanogen bromide fragment (CB7) of the alpha 1 chain in 0.02 M Tris buffer containing 0.15 M NaCl at pH 7.4. Addition of fibronectin to fluorescein-labeled collagen chains caused a dose-dependent increase in the fluorescence anisotropy which continued over several logs of titrant concentration. Scatchard-type plots of the anisotropy response were biphasic indicating the presence of one or more weak sites (Kd greater than microM) along the collagen chain in addition to a strong site characterized by Kd = 1.3 X 10(-8) M at 25 degrees C. Gelatin-binding fragments with Mr = 42,000, 60,000, and 72,000 also produced a biphasic response with Kd values for the high affinity site being 10- to 20-fold greater than for intact fibronectin. Binding of fibronectin and its fragments to fluorescent-labeled CB7 was essentially the same as to the whole alpha 1 chain. In all cases, the anisotropy response could be reversed or prevented by addition of excess unlabeled gelatin or CB7, but not by synthetic peptides spanning the collagenase cleavage site of alpha 1 (I). Studies of the temperature dependence of Kd for binding of fibronectin to the high affinity site on alpha 1 produced a value of +16 kcal/mol for the enthalpy of dissociation below 30 degrees C. Above this temperature, fibronectin appeared to undergo a subtle conformational transition characterization by a reduced affinity for collagen. This transition occurred in whole fibronectin but not in the gelatin-binding fragments and may involve disruption of intramolecular interactions between different domains.

Interaction of Fibronectin and Its Gelatin-binding Domains with Fluorescent-labeled Chains of Type I Collagen* (Received for publication, October 26, 1987) Kenneth C. InghamS, Shelesa A. Brew, and Benjamin S . Isaacs

From the Biochemistry Laboratory, American Red Cross Biomedical Research and Development, Rockuille, Maryland 20855
Fluorescent probes have been used to obtain dissociation constants for the fluid-phase interaction of human plasma fibronectin and several of its gelatin-binding fragments with purified a chains of type I rat tail collagen, as well as with a cyanogen bromide fragment (CB7) of the a1 chain in 0.02 M Tris buffer containing 0.15 M NaCl at pH 7.4. Addition of fibronectin to fluorescein-labeled collagen chains caused a dose-dependent increase in the fluorescence anisotropy which continued over several logs of titrant concentration. Scatchard-type plots of the anisotropy response were biphasic indicating the presence of one or more weak sites (Kd > 1 pM) along the collagen chain in addition to a strong site characterized by Kd = 1.3 X lo-' M at 25 "C. Gelatin-binding fragments with M, = 42,000, 60,000, and 72,000 also produced a biphasic response with Kd values for the high affinity site being 10-to 20-fold greater than for intact fibronectin. Binding of fibronectin and its fragments to fluorescent-labeled CB7 was essentially the same as to the whole al chain. In all cases, the anisotropy response could be reversed or prevented by addition of excess unlabeled gelatin or CB7, but not by synthetic peptides spanning the collagenase cleavage site of al(I). Studies of the temperature dependence of Kd for binding of fibronectin to the high affinity site on al produced a value of +16 kcal/mol for the enthalpy of dissociation below 30 "C. Above this temperature, fibronectin appeared to undergo a subtle conformational transition characterized by a reduced affinity for collagen. This transition occurred in whole fibronectin but not in the gelatin-binding fragments and may involve disruption of intramolecular interactions between different domains.

~~ ~ ~ ~~
Fibronectin is a 500-kDa adhesive glycoprotein found in plasma and other body fluids, in the extracellular matrix, and on the surface of various cells (for a review, see Refs. 1-4). Its two nearly identical polypeptide chains are each comprised of several distinct domains which can be isolated from proteolytic digests with retention of specific macromolecular recognition properties. Fibronectin binds to numerous macromolecules and mediates the attachment of many types of cells to surfaces containing those macromolecules. A thorough understanding of the numerous functions of this complex protein will require quantitative knowledge of its affinity for various substances. Here, we focus on the interaction with collagen, * This work was supported by Grant HL21791 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 "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ T o whom correspondence should be addressed American Red Cross, 15601 Crabbs Branch Way, Rockville, MD 20855. a ubiquitous component of the extracellular matrix. Fibronectin is known to bind to several types of collagen and the binding site on fibronectin has been localized to a 30-40 kDa domain, near the N terminus of each polypeptide chain, where the unique type I1 homologous repeat units are located. The importance of the type I1 units for collagen binding was recently demonstrated by means of fusion proteins expressed in Escherichia coli ( 5 ) . An intriguing aspect of this interaction is the fact that, with most types of collagen, the denatured forms exhibit a much stronger binding than the native forms, suggesting that the binding sites are at least partially masked in the native triple helix (6-9). A major site on type I collagen has been assigned to CNBr fragment 7 of the a1 chain (10,11).
While much valuable qualitative information has accumulated about the interaction of fibronectin and its fragments with the collagens and their fragments, the quantitative picture is less clear. Most studies have focused on elucidating the regions of the various molecules that are involved in the recognition process with, at best, a comparison of the concentrations required for 50% inhibition in cell attachment or immunoprecipitation assays. The few dissociation constants available in the literature have been determined by solidphase methods, with unfractionated gelatin as the substrate. No dissociation constants are available for purified a chains. We have developed the use of fluorescent probes for the purpose of quantitating these interactions in the fluid phase under true equilibrium conditions (12,13). Here, we present a systematic study of the binding of fibronectin and its fragments to purified chains of type I collagen and to the CB7 fragment of al(I). In addition, we provide evidence that fibronectin undergoes a temperature-dependent conformational change near 33 "C which reduces its affinity for the cy1 chain.

DISCUSSION
The results show that binding of fibronectin to isolated chains of type I collagen and to the CB7 fragment of the a1 chain is a complex process; the anisotropy response extends over several logs of titrant concentration (Fig. 2) and Scatchard-type plots of the data are biphasic (Fig. 3). This was true for whole fibronectin as well as its fragments. Such heterogeneity can be interpreted in several ways. One possibility is that each collagen chain contains two (or more) high affinity sites which interact in such a way that occupation of the first somehow diminishes the affinity of the others, perhaps through steric interference. A second possibility is the  formation of ternary or higher order complexes mediated by interactions between fibronectin molecules at high fibronectin concentration. The fact that the anisotropy response to fragments of fibronectin was qualitatively similar to that of the whole protein argues against this interpretation since selfassociation, to the extent that it has been characterized, appears to involve heterologous interactions between different domains widely separated in the polypeptide chains (14-17). The third and simplest explanation for the observed heterogeneity in binding is that each chain contains two or more independent sites that differ widely in their intrinsic affinity for fibronectin. In most cases, the initial linearity of the Scatchard-type plots extended almost to the abscissa (see This same linearity also suggests that if multiple high affinity sites exist, they must be virtually identical. Studies by Kleinman and co-workers (10,18) suggest that binding of fibronectin to the a1 chain of type I collagen is dominated by a single site about three-fourths of the distance between the N and C termini, near the cleavage site for mammalian collagenase. This site is contained within the CB7 fragment and our results confirm that binding of fibronectin to the al chain can be explained entirely in terms of the behavior of the CB7 fragment which appears to contain not only the strong site but one or more weak sites. However, synthetic peptides which span the collagenase cleavage site showed no binding or inhibitory activity in our system, indicating that further studies are required to define the precise location of the high affinity site.
Indirect evidence that each chain contains only a single high affinity site comes from a comparison of the apparent K d for whole fibronectin with those of its fragments. Since whole fibronectin is bivalent with respect to gelatin binding, the presence of more than one strong site on a given collagen chain would confer a much higher affinity for fibronectin relative to the isolated domains, since the flexible fibronectin molecule (19)(20)(21) could conceivably attach in two places to the flexible a chain. In fact, if Kd for the isolated domains is taken as 30 X lo-' M at 25 "C (Table I) gests that all three contain an intact collagen-binding domain which is located entirely within the 42,000 portion and has not been damaged by removal from its parental environment. How then to account for the observed 10-to 20-fold lower affinity relative to whole fibronectin? Here also the bivalency of intact fibronectin must be considered. Given the presence of a strong site and one or more weak sites on the same collagen chain or CB7 fragment, a model in which one chain of fibronectin binds to a strong site while the second chain binds to a weak site seems plausible, especially in view of the flexible nature of both reactants. However, additivity of free energies would still predict a much greater difference between the bivalent and monovalent species than is observed. It is possible that intramolecular associations between domains in fibronectin could affect the environment or conformation of the gelatin-binding domain in such a way as to increase its affinity for the strong site. Another possibility is that more distant domains of fibronectin have an affinity for collagen which is too weak to be detected independently but which contributes to the primary interaction when present as part of the same molecule.
Another important difference between whole fibronectin and its gelatin-binding fragments is the dependence of the binding on temperature. In the range of 10-30 "C, van't Hoff plots had similar slopes for fibronectin and 42,000 reflecting similar enthalpies of dissociation of about 16 kcal/mol (Fig.  5). This indicates that the weaker binding of the fragment in this temperature range arises from a larger entropy of dissociation. Above 30 "C, fibronectin, but not the fragment, underwent a transition to a form characterized by a 10-fold weaker affinity for the a1 chain such that the difference in affinity between fibronectin and 42,000 was diminished above this temperature. Triple-helical collagen structures are known to unfold near this temperature and the existence of such structures could explain a discontinuity in fibronectin binding. Although isolated a chains and even CNBr fragments of type I collagen are capable of reforming triple helical structures, the required concentrations and incubation times are orders of magnitude greater than those involved here (22)(23)(24)(25). Furthermore, the effect seems to be in the wrong direction since it is well-established that denatured collagen has much higher affinity for fibronectin than does native triple-helical collagen (6)(7)(8)(9). Finally, as shown in Fig. 1, the anisotropy of FITC-a, varied smoothly with temperature over the entire range investigated with no evidence for unfolding of the kind easily detected by the same method when FITC-labeled native collagen was heated (9). We therefore conclude that the observed discontinuity in the van't Hoff plot reflects a change in fibronectin and not in the a chains. If the CY chains were responsible, one would expect to see the effect with the 42,000 fragment as well as with fibronectin.
Studies of the thermal stability of fibronectin and its fragments have focused primarily on the gross denaturation which occurs above 50 "C (26, 27, and Refs. therein). However, several observations in the literature support the possibility of a more subtle thermal transition near physiological temperature. Lai and Tooney (28) examined the dependence of the ESR spectrum of a spin-labeled derivative of fibronectin on temperature; their published spectra show new features appearing above 30 "C. Brown et al. (29), interpreted differences in monoclonal antibody binding, surface binding, and trypsin sensitivity as evidence for a conformational change in fibronectin between 4 and 37 "C. Williams et al. (30) observed a break near 25 "C in the Perrin plot of dansylcadaverinelabeled fibronectin, similar to the one we have seen near 33 "C with 1-anilino-8-naphthalenesulfonate-labeled fibronectin (Fig. 6). In their case, the probes were attached close to the N termini whereas in ours, the probes were attached to sulfhydryl groups confined to the C-terminal third of the fibronectin chains. The occurrence of a break in the Perrin plot of both derivatives suggests that the increased flexibility at higher temperature may reflect a disruption of intramolecular interactions between domains which are widely separated in the primary structure. Such interactions have been proposed to explain the unusual sensitivity of the hydrodynamic properties of fibronectin to changes in ionic strength and pH (30-32). The involvement of similar interactions in the low temperature transition discussed above could explain its absence in the isolated gelatin-binding domain.