(Received for publication, September 17, 1996, and in revised form, October 28, 1996)
From the American Red Cross Holland Laboratory, Rockville, Maryland 20855
The first type III module of fibronectin (Fn) contains a cryptic site that binds Fn and its N-terminal 29 kDa fragment and is thought to be important for fibril formation (Morla, A., Zhang, Z., and Ruoslahti, E. (1994) Nature 367, 193-196; Hocking, D. C., Sottile, J., and McKeown-Longo, P. J. (1994) J. Biol. Chem. 269, 19183-19191). A synthetic 31-mer peptide (NAPQ ... TIPG) derived from the middle of domain III1 was also shown to bind Fn, but the site of its interaction was not determined (Morla, A., and Ruoslahti, E. (1992) J. Cell Biol. 118, 421-429). By affinity chromatography on peptide-agarose, we tested a set of fragments representing the entire light chain of plasma Fn. Only 40-kDa Hep-2 (III12-15) failed to bind. The concentration of urea required for peak elution of Fn and the other fragments decreased in the order Fn > 42-kDa GBF (I6II1-2I7-9) > 19-kDa Fib-2 (I10-12) > 110-kDa CBF(III2-10) > 29-kDa Fib-1 (I1-I5). Neither Fn nor any of the fragments bound immobilized intact III1, confirming the cryptic nature of this activity. 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 III1 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 III7 and III15. Fn itself did not bind to the denatured fragments. Thus, domain III1 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 III1 (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 III1, 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 III1, when adsorbed to plastic above its denaturation temperature, was able to bind Fn and its N-terminal 70-kDa fragment; other fragments C-terminal to this were not tested, nor were a sufficient number of denatured type III domains tested to determine the specificity of the observed interaction. More recently, the same group reported that native III1 was able to bind heat-denatured III10 and hypothesized that integrin receptors might induce a conformational change in III10 that would expose a binding site for III1 (16).
Denaturation of any protein or independently folded domain will expose hydrophobic groups that are normally buried in the native structure. These sticky residues, at least some of which are also present in the synthetic III1 peptide, could interact with other proteins or fragments in a nonspecific manner, in the same way for example that a number of Fn fragments have been shown to interact with alkyl- and aryl-Sepharose columns (17-19). In the present study, we used analytical affinity chromatography to test a complete set of fragments representing almost the entire light chain of Fn for their ability to bind native or denatured III1. In addition, we tested each of the native fragments against each of the others to identify additional interactions. We found that denatured III1, as represented by the above-mentioned synthetic peptide, bound numerous Fn fragments. The only interaction detected among the native fragments was that of fluid phase III1 with immobilized 40-kDa Hep-2 (III12-15) and 110-kDa cell-binding fragment (III2-10). These interactions were dramatically increased after deliberate denaturation of the immobilized fragments. Additional experiments with subfragments indicate that module III7 in 110 kDa and III15 in 40 kDa contain the cryptic sites for binding of native III1 to the respective fragments.
Human plasma fibronectin was prepared as described previously
(20). The major fragments of Fn were generated with thermolysin (21).
The methods of purification and documentation of their purity have been
published previously. The heparin-binding fragments (40-kDa
III12-15 and 30-kDa III12-14) and 8-kDa
III15 were purified according to Ingham et al.
(22). The gelatin-binding fragments (42, 30, and 21 kDa) were purified
according to Ingham et al. (23). The 56-kDa gelatin-binding
fragment was separated from the others by adsorption of the
gelatin-binding fraction on heparin-Sepharose; it was homogeneous by
SDS-PAGE. Two versions of III1, designated III-1a and
III-1b by Litvinovich et al. (24), were used. The first
begins with ITETP and was used in all of the experiments in which
III1 was chromatographed on affinity columns. It is herein
referred to as 9-kDa III1. The other is slightly shorter,
beginning with IQWN, and is the one coupled to Sepharose. The 110-kDa
III2-10, 54/60-kDa III2-5/6, 39-kDa
III7-8, 26-kDa III8-9, and 11-kDa
III10 from the cell-binding region were purified as
described (25).2 The 29-kDa Fib-1/Hep-1
(referred to herein as simply Fib-1) and 19-kDa Fib-2 fragments were
purified as described (26). Recombinant maltose-binding protein fused
to Fn module III7 (MBP-III7) was prepared in
Escherichia coli by methods similar to those described (27);
it was homogeneous by SDS-PAGE. The synthetic 31-mer peptide from the
middle of III1 (NAPQ ... TIPG) was prepared with
C-terminal amide on a Milligen model 9050 peptide synthesizer using
Fmoc chemistry as described (28). The locations of all fragments and
the synthetic peptide within Fn are summarized in Fig.
1.
Affinity chromatography matrices were prepared by coupling Fn and its
fragments to CNBr-activated Sepharose 4B following the directions of
the manufacturer (Pharmacia Biotech Inc.). Generally, proteins were
coupled in 0.1 M NaHCO3, 0.5 M
NaCl, pH 8.0, overnight at 4 °C except for Fn, which was coupled at
room temperature. Unreacted sites were blocked with
tris(hydroxymethyl)-aminomethane (Tris) buffer. The degree of
substitution varied between 3 and 6 mg of protein per ml of packed gel.
Matrices that were eluted with urea had been washed with urea
previously. Those eluted with guanidinium chloride (GdmCl) were washed
only with alternating acid, base, and 1 M NaCl solutions
before equilibration with TBS (0.02 M Tris, 0.15 M NaCl, pH 7.4) to prevent possible denaturation of the
coupled protein before the first chromatography run. Heat-denatured affinity matrices were prepared by incubating a slurry of the resin in
TBS for 10 min at 100 °C. Chromatography was performed at room
temperature on a Pharmacia FPLC system. One hundred-µl volumes of
protein solutions with A280 0.5 were injected onto affinity matrices in TBS at a flow rate of 1.0 ml/min. After a 5-ml
wash with TBS, a 10-ml linear gradient from 0 to 6 M urea in TBS or from 0 to 6 M GdmCl in 0.02 M Tris, 1 M NaCl, pH 7.4, was applied. Elution was monitored by
absorbance at 280 nm with the Pharmacia detector or by fluorescence at
340 nm with excitation at 280 nm using a Shimadzu RF-535 detector.
Elution profiles were corrected for baseline shifts by subtracting the
profile obtained when only TBS was injected.
Blocking of sulfhydryl groups was accomplished by incubating immobilized 8-kDa III15 that had been previously heat-denatured with a 100-fold molar excess of HgCl2 for 15 min at room temperature, followed by washing with TBS. Alternatively, 40-kDa Hep-2 was exposed to 20 mM iodoacetamide in the presence of 6 M urea followed by dialysis into 0.1 M NaHCO3, 0.5 M NaCl, pH 8.0, for coupling to Sepharose as above.
SDS-polyacrylamide gel electrophoresis was performed on eluted fractions of a thermolysin digest (1:200 w/w for 4 h at room temperature) of 1.0 mg of Fn that had been applied to a 2.5-ml column of III1 peptide-Sepharose and eluted with a gradient from 0 to 6 M urea. The fractions were concentrated and electrophoresed on 8-25% polyacrylamide gels using a Phastgel system (Pharmacia). Protein bands were stained with Coomassie Brilliant Blue. Urea was from Life Technologies, Inc., and GdmCl was from U. S. Biochemical Corp. All other chemicals were reagent grade or better.
The synthetic III1 peptide (14)
was coupled to Sepharose, and a thermolysin digest of Fn was applied to
the column in TBS. Bound proteins were eluted with a gradient of urea.
Pass-through and bound fractions were collected and analyzed by
SDS-PAGE (Fig. 2). The nonbinding fraction contained 30- and 40-kDa Hep-2 fragments and two low molecular weight fragments,
9-kDa III1 and 8-kDa III15. Bound fragments
included 19-kDa Fib-2, 29-kDa Fib-1, 42-kDa GBF, and 110-kDa CBF. Fn
itself bound quantitatively and was eluted at 4 M urea
(data not shown).
Fn and a variety of its purified thermolytic fragments and subfragments
were also applied individually to III1 peptide-Sepharose in
TBS and eluted with a urea gradient (Fig.
3A). All of the major fragments except 40-kDa
Hep-2 and 9-kDa III1 (ITETP... . ) bound to the column
and were eluted with urea. The pattern is identical to that observed
when the thermolytic digest was applied (Fig. 2), indicating that the
fragments that bound in that experiment were interacting directly with
the immobilized peptide and not indirectly via other bound fragments.
None of the major fragments bound as tightly as Fn itself, as judged by
the concentration of urea required for their elution. In additional
experiments not shown, the 30- and 21-kDa subfragments of 42-kDa GBF
also bound but eluted earlier than the parent fragment, whereas the 56-kDa GBF, which contains type III1, eluted at a similar
position as 42-kDa GBF, which does not. Several subfragments of the
110-kDa CBF (25), including 11-kDa III10, 39-kDa
III7-8, and 54/60-kDa III2-5/6 either failed
to bind or were merely retarded (data not shown).
The synthetic III1 peptide is extremely basic with a net positive charge between five and six. The Fn fragments that fail to bind peptide-Sepharose are also basic (30- and 40-kDa Hep-2 and 9-kDa III1), whereas most of those that do bind are acidic. This suggested that electrostatic forces might be responsible for the the observed interactions. To test this possibility, an attempt was made to elute Fn from the peptide column by increasing the ionic strength. No elution was detected with a NaCl gradient up to 2 M. Therefore, the binding must involve other types of forces that are insensitive to ionic strength.
Fibronectin Fragments Do Not Bind Native III1-SepharoseFragment 8-kDa III1 (fragment III-1b of Litvinovich et al. (24) with N-terminal sequence IQW ... ) was coupled to Sepharose. Fn and each of its fragments were applied separately in TBS, eluting with a urea gradient (Fig. 3B). No binding was detected with any of the applied proteins. Thus, the binding sites that are present in the synthetic III1 peptide are truly cryptic and not accessible in the native domain from which its sequence was derived. Attempts to render it accessible by exposure of the immobilized domain with 6 M GdmCl (during elution) or by heat (100 °C for 10 min) were unsuccessful, perhaps because of the highly efficient refolding of this domain (24).
Denaturation of Type III-containing Fragments Exposes Binding Sites for Native III1Fn and all fragments and subfragments
were tested on affinity columns containing each of the major fragments,
including 29-kDa Fib-1, 42-kDa GBF, 8-kDa III1, 110-kDa
CBF, 40-kDa Hep-2, and 19-kDa Fib-2, coupled to Sepharose (Fig.
4). Elution was accomplished with a gradient of 0 to 6 M GdmCl. The only evidence for an interaction was in the
case of 9-kDa III1 binding to immobilized 110-kDa CBF and
40-kDa Hep-2. However, the percentage of the applied III1 which bound was low, less than 15%, although the columns contained at
least a 5-fold molar excess of the immobilized ligand. All of the
columns had been washed with GdmCl prior to application of
III1, raising the possibility that the observed binding
might have been due to partial denaturation of the immobilized
proteins. Therefore, fresh columns of 110-kDa CBF and 40-kDa Hep-2 were prepared, and III1 was applied to determine if it would
bind to the native fragments. As shown in Fig. 5,
A and B, the amount of protein that was eluted
with GdmCl was trivial in comparison to the nonbound fraction.
After washing the columns with TBS, a second batch of III1 was applied to determine if exposure to 6 M GdmCl during the previous elution would increase the amount of binding. As shown in Fig. 5, C and D, that was indeed the case; approximately 40% of the applied material bound to each column. In the case of the 110-kDa CBF column, the elution occurred in two peaks. Because there was no evidence for heterogeneity of the III1 preparation, this biphasic elution pattern was attributed to heterogeneity of binding sites on the partially denatured immobilized CBF. The percentage of material bound in these experiments is probably underestimated by the fluorescence detector because the native unbound protein has a higher fluorescence efficiency than the denatured bound one, which would be unfolded by the GdmCl used for elution (24).
The affinity resins were then removed from the column, incubated at 100 °C for 10 min, washed with 6 M GdmCl, and again tested in the column. After this treatment, the binding was essentially quantitative in the case of 110-kDa CBF and nearly so with 40-kDa Hep-2 (Fig. 5, E and F). Although the overall recovery of fluorescent material was diminished because of the quenching effect of the GdmCl, the peak representing the unbound unquenched native III1 passing through the column was much lower than in Fig. 5, A and B. The conclusion that both 110-kDa CBF and 40-kDa Hep-2 contain cryptic binding sites for native III1 is inescapable.
Specificity of Binding of Native III1 to Heat-denatured 110-kDa CBFFn and its fragments and subfragments were applied to
heat-denatured immobilized 110-kDa CBF in TBS followed by elution with a gradient of GdmCl as before. As shown in Fig. 6, the
only binding detected was that of native 9-kDa III1 and, to
a lesser extent, 56-kDa GBF, which contains III1. A similar
pattern was seen with heat-denatured 40-kDa Hep-2 (data not shown).
Thus, the binding is quite specific in terms of the number of sites in
Fn that recognize heat-denatured type III domains. This is in contrast
to the situation with the III1 peptide, assumed to
represent denatured III1. Note that the III1
peptide did not bind significantly to the 110-kDa column, whether
native or denatured, although the 110-kDa fragment did bind to the
immobilized peptide. Note also that Fn itself did not bind to denatured
110-kDa, although it contains III1, indicating that the
active site in native III1 is cryptic in whole Fn.
Localization of Binding Sites for Native III1 within Denatured Subfragments of 110-kDa CBF and 40-kDa Hep-2
Subfragments of 110-kDa CBF, including 11-kDa
III10, 26-kDa III8-9, 39-kDa
III7-8, MBP-III7, and a mixture of 54-kDa
III2-5 and 60-kDa III2-6, were coupled to
Sepharose and tested for binding of 9-kDa III1 (Fig.
7, A-E, respectively). The only significant
binding was seen with 39-kDa III7-8 and
MBP-III7, suggesting that III7 is the active
module. The amount of binding to these immobilized fragments increased
significantly after they had been purposely denatured. The increase was
from about 15 to >90% with 39-kDa III7-8 and from about
25 to 50% with MBP-III7. It is clear, however, that some
binding occurred, even to the native III7-8 and
MBP-III7 columns, suggesting that the active site in
III7 is partially exposed in these fragments simply by
separating them from other domains. In the case of
MBP-III7, it is not certain that all of the recombinant
domain was properly folded, even before deliberate denaturation. The
other subfragments of 110 kDa as well as MBP alone failed to bind
detectable amounts of III1, even after exposure to GdmCl
and heat. The failure of 26-kDa III8-9 to show significant
binding reinforces the conclusion that III7 is the active
module in 39-kDa III7-8. Similar experiments were
conducted with 8-kDa III15 and 30-kDa Hep-2
(III12-14), both of which are subfragments of 40-kDa Hep-2
(III12-15). As shown in the lower part of Fig. 7, binding
occurred only to 8-kDa III15 (Fig. 7F), and then
only after it was denatured. Thus, binding of III1 to
denatured 40-kDa Hep-2 appears to occur via module
III15.
Modules III7 and III15 are unique among type III modules in that each has a free buried sulfhydryl that becomes exposed in the presence of denaturants (29). To determine whether free sulfhydryls might be involved in binding of III1, denatured 8-kDa III15-Sepharose was exposed to 44 mM HgCl2 in TBS for 15 min, a treatment that was shown in separate experiments to block the incorporation of sulfhydryl-specific probes into the protein (data not shown). Such treatment had no effect on the binding of 9-kDa III1. A second experiment in which the sulfhydryl in 40-kDa Hep-2 was blocked with iodoacetic acid in 6 M urea prior to immobilization led to the same conclusion.
Because III7 and III15 are acidic and III1 is basic, we tested whether III1 could be eluted from immobilized heat-denatured 39-kDa III7-8 and 8-kDa III15 with salt alone. No elution was detected up to a NaCl concentration of 1 M.
Based on experience with other polymerizing systems such as fibrinogen, it should be possible to localize sites of interaction of Fn by identifying fragments that either self-associate, interact with other fragments, or compete with protomers for binding to complementary polymerization sites and thereby inhibit matrix assembly. Although N-terminal fragments have long been known to inhibit matrix assembly, the sites of their interaction have not been established. The fact that they bind reversibly to cultured cell layers prompted a search for a putative "matrix assembly receptor" (3, 9, 10). Failure to identify a convincing candidate has strengthened the idea that the long sought receptor may be Fn itself, perhaps conformationally altered by its interaction with cells (5, 11, 12). Indeed, experiments in which 125I-labeled 29-kDa Fib-1 was covalently cross-linked to the matrix identified Fn as one of the targets (9, 30).
In the present study, we systematically checked for self-interactions between native fragments of Fn by testing all thermolysin fragments in pairwise fashion by affinity chromatography. Finding no significant interactions between the native fragments, we then followed the lead of Hocking et al. (11, 16) and began to explore the possibility that some of the sites might be cryptic, becoming exposed upon denaturation. The only interaction thus detected was that of native 9-kDa III1 (or 56-kDa GBF, which contains III1) with immobilized denatured type III-containing fragments 110-kDa CBF and 40-kDa Hep-2. These sites were indeed cryptic because: (a) they were much better expressed after denaturation; and (b) immobilized native III1 failed to bind native 110-kDa or 30/40-kDa. The observed binding was rather specific in that III1 was the only native thermolysin fragment that was able to recognize cryptic sites in other heat-denatured type III domains. There was also specificity in terms of the number of type III domains that were recognized by native III1. Testing of a range of subfragments indicated that the ones being recognized upon denaturation were III7 and III15 and that this recognition was independent of the free sulfhydryls contained therein. Subfragment 39-kDa III7-8 and MBP-III7 appeared to bind significant amounts of III1, even before denaturation, as if this site might be at least partially exposed in the smaller subfragment without the need to unfold.
Hocking et al. (16) reported that recombinant III1 (presumably native) as well as whole Fn were able to bind recombinant III10 but only when the latter was adsorbed to plastic wells at 80 °C, just below the midpoint for thermal denaturation of natural 11-kDa III10 (25). However, a slightly shorter version of recombinant III10, truncated by several kilodaltons from the C terminus, was active even in the fluid phase, as evidenced by its ability to partially inhibit the binding of Fn to heated full-length, plastic-adsorbed recombinant III10 (16). This suggested that a binding site for native III1 was exposed in III10 by proteolytic truncation (or by unfolding, if truncation destabilized the compact structure). In our hands, native 9-kDa III1 failed to bind Sepharose-conjugated 11-kDa III10, even after exposure of the latter to harsh denaturing conditions. Unfolding of 11-kDa III10 is highly reversible in solution (25).3 Although it seems unlikely, we cannot exclude the possibility that refolding of immobilized III10 after boiling or exposure to 6 M GdmCl was sufficiently complete to preclude detectable binding of III1 (Fig. 7, top panel). Note also that domain III11 is not represented by any of the fragments studied and thus cannot be excluded as a possible target of binding by III1 or other fragments.
A role for module III1 in self-interactions of Fn was
evident from earlier reports that Fn matrix assembly could be inhibited by a monoclonal antibody that mapped to this region (13). Measurements of specific melting enthalpy suggest that III1 differs from
other type III domains in having a smaller compact core flanked by
unstructured regions that in other type III domains are part of the
overall fold (24). Thus, in a sense, module III1, at least
when isolated from neighboring domains, is already partly unfolded
(Scheme I), and its unstructured regions and/or its
exposed core could be available for interaction with complementary
regions on other type III domains through what might be termed
"subdomain swapping" (31). The fact that chymotrypsin cleaves Fn in
III1 after Trp599 (see Fig. 1) (14), which is
highly conserved and deeply buried in other type III domains (32-36),
indicates that this region of III1 is accessible in whole
Fn. Yet, whole Fn fails to bind denatured immobilized 110-kDa CBF (Fig.
6, top) or 40-kDa Hep-2 (data not shown), suggesting that
the binding site(s) in III1 are not accessible in the whole
protein. The mechanism by which these and other cryptic sites in type
III domains would become exposed in native Fn remains unknown but may
depend on mechanical forces exerted by cytoskeletal filaments that
align with Fn fibrils across the membrane (1, 5, 12, 37, 38).
Cytochalasin and other agents that disrupt actin-containing stress
fibers also block or reverse Fn matrix assembly and abolish binding
sites for N-terminal fragments (5, 39), consistent with the idea that
tensile forces expose cryptic sites. Erickson (40) estimated that
forces of readily achievable by contracting cells would be sufficient
to completely unfold a type III domain with a G of unfolding of
~10 kcal/mol (24). Partial unfolding would require even less
energy.
The site or sites in native III1 that recognize denatured III7 and III15 appear to be different from that represented by the 31-residue synthetic peptide of Morla and Ruoslahti (14). This peptide, derived from the core of III1, did not bind to denatured 110k-Sepharose under conditions where intact 9-kDa III1 did. However, we did confirm their finding that the 31-residue peptide, when coupled to Sepharose, is able to bind Fn. This is not surprising in view of the fact that the same peptide also binds most thermolysin fragments, including 29-kDa Fib-1, 42- and 56-kDa GBF, 110-kDa CBF, and 19-kDa Fib-2. The site represented by the peptide is truly cryptic because immobilized intact III1 failed to bind Fn or any of its fragments. It is possible that the structure of III1 in whole Fn is altered by interactions that play a role in maintaining the molecule in a soluble conformation. Additional work is required to test this possibility and to determine which of the numerous interactions of III1 are important for matrix assembly, a process which based on electron microscopic studies (41, 42) can be assumed to be orderly and to involve specific recognition sites between molecules in the fibrils.
Protomeric Fn in solution has long been known to undergo reversible changes in its hydrodynamic properties in response to changes in pH and ionic strength (43). This is thought to arise from long range intramolecular electrostatic interactions between oppositely charged "domains" (44-46). One imagines that an interaction within or between chains of protomeric Fn could involve some of the same recognition elements that govern fibril formation and that the role of cells is to somehow open the molecule and facilitate a switch from intra- to intermolecular self-association. The work presented here as well as additional unpublished work in our laboratory reveals a lack of self- or hetero-interactions between native purified fragments, which collectively cover almost the entire chain. It is possible that multiple interactions involving modules from separate thermolytic fragments, perhaps on separate chains of the covalent dimer, are required to stabilize the solution conformation (47). Additional studies are required to elucidate the structure of Fn in solution, that of its fibrils in the extracellular matrix, and the mechanism by which the former is transformed to the latter. The present work provides additional support for the role of cryptic sites in the self-interactions of this complex protein. Some of these sites may be simply masked by domain-domain interactions within the solution conformation, but others appear to be hidden within the folded domains themselves.