In an effort to detect interactions between other Fn domains, all fragments were coupled to Sepharose, and each fragment was tested on each affinity matrix before and after denaturation. The only interaction detected was that of fluid phase III 1 with immobilized denatured 110-kDa CBF and 40-kDa Hep-2, both of which contain type III domains. Analysis of subfragments revealed this activity to be dominated by domains III 7 and III 15 . Fn itself did not bind to the denatured fragments. Thus, domain III 1 contains two cryptic "self-association sites," one that is buried in the core of the fold but recognizes many Fn fragments when presented as a peptide and another that is concealed in Fn but exposed in the native isolated domain and recognizes cryptic sites in two other type III domains. These interactions between type III domains could play an important role in assembly of Fn multimers in the extracellular matrix.
Fibronectin (Fn)1 circulates in plasma as a 550-kDa 2-chain monomer that can be transformed by cultured fibroblasts into an insoluble fibrillar structure during the cell-driven process of matrix assembly (1). Several regions of Fn have been implicated in this process. N-terminal fragments containing the first five type I modules bind to cell layers and inhibit matrix assembly but are not themselves incorporated into the insoluble matrix unless bivalent, and then most efficiently in the presence of intact Fn (2-8). For a long time, it was thought that N-terminal fragments were interacting with a cell-surface "matrix assembly receptor" that was distinct from integrins that interact with Fn primarily via the Arg-Gly-Asp in the 10th type III domain. At least two unidentified molecules have been proposed as candidates for the matrix assembly receptor by virtue of their interaction with N-terminal fragments of Fn (9, 10). However, none has been further characterized or shown to play a role in matrix assembly. More recently, it is beginning to appear that the matrix assembly receptor, i.e. the molecule responsible for binding N-terminal fragments, might be Fn itself, perhaps conformationally altered by incorporation into the matrix (5, 11, 12).The fact that Fn matrix assembly could be inhibited with a monoclonal antibody directed to module III 1 (13) prompted efforts to examine the interaction of this module with Fn. Morla and Ruoslahti (14) showed that a synthetic 31-mer peptide with N-terminal sequence NAPQ . . . , derived from the middle of III 1 , was able to bind Fn, but the site of its interaction was not determined. The same group later reported that a longer recombinant peptide with the same N terminus, when incubated with whole Fn, was able to induce the formation of polymers that were stable in SDS in the absence of reducing agents and exhibited superior adhesive properties toward fibroblasts (15). At the same time, it was shown by Hocking et al.(11) that module III 1 , when adsorbed to plastic above its denaturation temperature, was able to bind Fn and its N-terminal 70-kDa fragment; other fragments...