All Six Modules of the Gelatin-binding Domain of Fibronectin Are Required for Full Affinity*

The gelatin-binding sites of fibronectin are confined to a 42-kDa region having four type I and two type II modules in the following order: I6-II1-II2-I7-I8-I9. To determine the relative importance of each module for recognition of gelatin, recombinant green fluorescent fusion proteins were prepared in which individual modules or groups of modules were deleted, and the resulting proteins were tested for binding to gelatin by analytical affinity chromatography. Deletion of both type II modules did not eliminate binding, confirming that at least some of the type I modules in this region are able to bind gelatin. It was found that deletion of type I module 6 tends to increase the affinity, whereas deletion of any other module decreases it. Deletion of module I9 had a large effect but only if module II2 was also present, suggesting an interaction between these two noncontiguous modules. Analysis of more than 20 recombinant fusion products led to the conclusion that all modules contribute to the interaction either directly by contacting the ligand or indirectly through module-module interactions.

The gelatin-binding sites of fibronectin are confined to a 42-kDa region having four type I and two type II modules in the following order: I 6 -II 1 -II 2 -I 7 -I 8 -I 9 . To determine the relative importance of each module for recognition of gelatin, recombinant green fluorescent fusion proteins were prepared in which individual modules or groups of modules were deleted, and the resulting proteins were tested for binding to gelatin by analytical affinity chromatography. Deletion of both type II modules did not eliminate binding, confirming that at least some of the type I modules in this region are able to bind gelatin. It was found that deletion of type I module 6 tends to increase the affinity, whereas deletion of any other module decreases it. Deletion of module I 9 had a large effect but only if module II 2 was also present, suggesting an interaction between these two noncontiguous modules. Analysis of more than 20 recombinant fusion products led to the conclusion that all modules contribute to the interaction either directly by contacting the ligand or indirectly through module-module interactions.
Fibronectin is a large modular glycoprotein found in the extracellular matrix and body fluids of higher organisms. It interacts with a variety of other macromolecules including integrin receptors, heparan sulfate proteoglycans, fibrin, tenascin, and several types of denatured collagens (gelatins). Here we focus on the interaction with gelatin. The gelatin-binding sites of fibronectin are confined to a region having four type I and two type II modules in the following order: I 6 -II 1 -II 2 -I 7 -I 8 -I 9 (Fig. 1). This region can be isolated as a 42-kDa fragment (42-kDa GBF) 1 that binds to gelatin with affinity only slightly lower than that of the parent protein (1). Fragments outside this region including an amino-terminal 29-kDa fragment (type I modules 1 through 5) and a carboxyl-terminal 19-kDa fragment (type I modules 10 through 12) do not bind gelatin. The 42-kDa fragment contains the only type II modules in fibronectin, modules that are thought to play a role in binding to gelatin not only in fibronectin but in other gelatin-binding proteins (especially some matrix metalloproteinases where type II mod-ules are also found) (2). However, the 42-kDa GBF can be further cleaved into two non-overlapping subfragments, I 6 -II 1 -II 2 -I 7 and I 8 -I 9 , both of which also bind gelatin, albeit with ϳ10-fold lower affinity than the parent fragment (3). This suggests that type II modules are not essential for binding of fibronectin to gelatin, although they may contribute, and that some type I modules can bind. It was recently reported that a recombinant two-module fragment, II 1 -II 2 , bound weakly to gelatin but that the affinity was about 5-fold higher in a threemodule fragment, I 6 -II 1 -II 2 (4). The latter fragment was shown in the same study to have a hairpin-like structure in which module I 6 interacts strongly with II 2 despite the intervening II 1 . This highlights the fact that affinity for gelatin may be affected not only by the presence of a given module but by the interaction of that module with neighboring modules.
In the present study we have prepared a large number of recombinant fragments in which one or more modules have been deleted from 42-kDa GBF to further understand the importance of all the modules for binding to gelatin. We have shown that modules II 2 and I 9 are most critical for binding but that all six modules contribute directly or indirectly to the binding site.

EXPERIMENTAL PROCEDURES
Materials-Polyclonal anti-green fluorescent protein (GFP) antibody was purchased from Clontech. Spodoptera frugiperda (Sf9) cells adapted to grow in serum-free medium Sf900-II were obtained from Invitrogen. Swine skin gelatin (type I) was from Sigma. The 42-, 30-, and 21-kDa gelatin-binding fragments were purified from a thermolysin digest of human plasma fibronectin as described previously (5).
Plasmid Constructs-A vector (pENTR11-MHG) encoding a secretion signal, His 6 tag, and GFP tag at the amino terminus was described previously (6). cDNAs encoding the gelatin-binding domain of rat fibronectin and its variants were amplified by Pfu DNA polymerase (Stratagene) followed by BamHI-HindIII digestion and then subcloned into BglII-HindIII sites of pENTR11-MHG. The primers used for amplification are given in Table I where amino acids are numbered from the amino terminus (pyroglutamic acid) of the processed fibronectin. The inserts were transferred to pDEST8 (Invitrogen) using the Gateway cloning system (Invitrogen).
Preparation of Recombinant Fragments-Recombinant fragments were expressed in Sf9 insect cells using the Bac-to-Bac expression system (Invitrogen). A bacterial strain (DH10Bac) was transformed with recombinant plasmids, and colonies containing recombinant bacmids were grown for isolation of the resulting viral DNA according to the manufacturer's instructions. Supernatants of transfected Sf9 cells were used to infect more Sf9 cells through two or three stages of amplification. Sf9 cells (2 ϫ 10 8 ) were then infected with recombinant viruses, and supernatants were harvested 48 or 72 h after infection. Recombinant GFP-fibronectin fragments were purified on Talon metal affinity resin (Clontech), eluting with 150 mM imidazole, and then dialyzed against phosphate-buffered saline (PBS, pH 7.4). The identity of each fragment was verified by detection of the appropriate viral DNA in the infected cells with PCR and by the apparent molecular weight of the purified protein on SDS-PAGE. Yields varied between 0.5 and 1.2 mg/100 ml of culture supernatant.
Gelatin Affinity Chromatography-Analytical affinity chromatography was performed as described previously using an FPLC system * 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.
(Amersham Biosciences) (7). Briefly, recombinant protein was injected onto a gelatin-Sepharose column equilibrated with PBS and pumped at a flow rate of 1 ml/min. After washing, a linear gradient of urea (0 -6 M) was applied between 10 and 30 min, and the bound GFP fusion proteins were detected with a fluorescence monitor (Shimadzu Model RF535). In some experiments, unbound and bound fractions were collected and subjected to immunoblot analysis using anti-GFP antibody (8). The variable amount of unbound green fluorescence in the analytical runs is due to varying degrees of proteolytic cleavage at the flexible linker between GFP and the GBFs during handling and/or storage of the samples.
Solid-phase Binding Assay-Ninety-six-well microtiter plates were coated with 10 g/ml gelatin in PBS for 16 h at 4°C. After washing and blocking with 1% bovine serum albumin in PBS, the plates were incubated with various concentrations of recombinant proteins for 2 h at room temperature. Plates were washed three times with 0.2% bovine serum albumin in PBS and then incubated with polyclonal anti-GFP antibody for 1 h at room temperature followed by incubation with horseradish peroxidase-conjugated donkey anti-rabbit IgG antibody for 40 min at room temperature. After three washes with PBS, plates were developed with peroxidase substrate. Absorbance was measured with a Dynatec plate reader, and C 50% values were obtained by fitting the data to a simple binding isotherm using Sigmaplot.

RESULTS
Since each of the six modules in the 42-kDa gelatin-binding fragment of fibronectin has a unique number, 6, 1, 2, 7, 8, and 9, those numbers are used as a shorthand notation to specify the modular composition of the various mutants, and dashes (-) are used to designate deletions. Various combinations of these modules were expressed as fusion proteins with enhanced GFP using a baculovirus expression system. Initially we used gelatin-Sepharose in conjunction with SDS-PAGE and immunoblotting to assess binding. Typical results are shown in Fig. 2 where it can be seen that the pair of type II modules alone does not bind to gelatin-Sepharose, whereas a fragment containing type I modules 7, 8, and 9 does bind. These data suggest that type II modules alone are neither sufficient nor essential for binding to gelatin and that at least some type I modules are able to bind without the help of type II modules.
Analytical affinity chromatography on gelatin-Sepharose was used to rank the relative affinities of the various fusion proteins using green fluorescence to monitor elution in a urea gradient. The results for all products are presented in Fig. 3, where panel A refers to the wild type fusion protein, designated 612789. Its peak elution occurs at 3.3 M urea, very close to that observed with natural 42-kDa GBF as reported previously (6). This elution profile is reproduced as a dotted line in all other panels for easy comparison. Except for those panels where no peak is evident, the profiles have been normalized to give the same amplitude at their peak elution. The urea concentrations at peak elution are summarized in Table II. The next six panels, B through G, in Fig. 3 show results for all single module deletions from amino to carboxyl terminus, respectively. Although none of these deletions abolishes binding, all of them have a measurable effect on the elution position. Interestingly removal of module 6 actually increases the concentration of urea required for elution to about 3.7 M (panel B). The other deletions shift the elution to lower urea in varying degrees with loss of module 7 having the smallest effect and module 9 having the largest. These results alone lead to the conclusion that all six modules play a role, directly or indirectly, in binding to gelatin.
Panels G, H, and I of Fig. 3 show the effects of successive deletions from the carboxyl-terminal end. Once module 9 has been removed, further deletion of module 8 has very little effect as if its only role is to tether 9 in the appropriate position. Further deletion of module 7 yields a three-module fragment, 612---, which fails to bind to gelatin-Sepharose, although slight retardation is evident. This lack of binding of 612---was confirmed in a separate experiment using an immunoblotting approach like that in Fig. 2 (data not shown).
Panels J, K, and L of Fig. 3 represent successive deletions of modules from the amino-terminal end. As noted above, deletion of module 6 increases the affinity (panel J is identical to panel B). Further deletion of module 1 seems to reverse the effects of deleting module 6 so that the elution of the four-module fragment --2789 is indistinguishable from that of the full-length fragment. The next deletion produces the three-module fragment ---789, which despite its lack of type II modules is able to bind (as in Fig. 2), albeit with lower affinity than --2789. This proves again that type II modules are not essential for binding FIG. 1. Schematic illustration of the modular composition of human plasma fibronectin and its gelatin-binding region. Type I modules are shown as ovals, type II modules are shown as circles, and type III modules are shown as squares; the small diamonds designate glycosylation sites. The gelatin-binding domain has modular composition I 6 -II 1 -II 2 -I 7 -I 8 -I 9 , referred to herein as 612789. Also shown is a rendition of the structure of subfragment I 6 -II 1 -II 2 illustrating the folding back of module II 2 onto I 6 (4). of fibronectin to gelatin, although the second one has some effect. Fig. 3, panel M, shows the results for a four-module fragment lacking both terminal type I modules 6 and 9. It was shown above (panel G) that deletion of module 9 causes the largest decrease in affinity of any single module deletion, while deletion of module 6 (panel B) caused an increase in affinity. Here it is seen that the enhancing effect of deleting module 6 also occurs upon its removal from 61278-, shifting the peak elution from 2.15 to 2.45 M urea and partially compensating for the effect of removing module 9.
Deletion of either terminal module 1 or 8 from -1278-had very little further effect (Fig. 3, panels N and O), but deletion of both to produce the bimodular fragment --27--(panel P) caused a noticeable weakening with substantial amounts of protein bleeding from the column during the wash phase prior to application of the urea gradient. Module 7 binds slightly better when partnered with module 8 (panel R) than with module 2 (panel P). Panel Q shows that a lone pair of type II modules (-12---) does not bind tightly enough to survive the wash phase, consistent with its absence in the sixth lane of Fig. 2. However, deletion of this type II pair from the full-length fragment to yield 6--789 significantly lowers the affinity (panel S).
Fragment --2789 is the only four-module fragment that binds with full affinity. Removal of either of its terminal modules substantially lowers that affinity by a similar amount (Fig. 3,  panels K, L, and N). However, deletion of the interior modules 7 and 8 from --2789 generates a bimodular fragment containing only 2 and 9, which failed to bind with significant affinity although retardation was evident (panel T). If modules 7 and 8 in either --2789 or in the full-length fragment are replaced by homologous modules 4 and 5 from outside the gelatin-binding region, the binding is essentially abolished (panels U and V). The effect of deleting modules 7 and 8 from the full-length fragment (panel W) is much less severe than replacing them with modules 4 and 5. That modules 7 and 8 serve more than a spacer function is evident from the fact that they can bind independently as a pair (panel R). From the foregoing results it is apparent that all six modules in the gelatin-binding domain of fibronectin play a role in ligand recognition, either directly through contact with gelatin or indirectly by influencing neighboring modules.
ELISA analysis was used as an alternative measure of the affinity of some of the fusion proteins for gelatin. Plastic microtiter wells were coated with gelatin and incubated with increasing concentrations of the fusion protein, and binding was detected immunologically with antibody to GFP. Representative data are shown in Fig. 4. The concentrations required for half-saturation (C 50% ) in this and other experiments not shown varied between 0.2 and ϳ5 M as summarized in Table  II. The range of values of C 50% obtained for the full-length fragment is in good agreement with K d values between 0.6 and 1.2 M determined previously in the fluid phase (Ref. 3 and references therein). In Fig. 5, the values of C 50% are plotted against the urea concentration required for peak elution. Despite considerable scatter in the data, there is a reasonable

612--9
Val correlation. Most of the scatter can be attributed to uncertainties in the C 50% values determined by ELISA, which is not the optimum method for weak interactions. The affinity chromatography results, although only qualitative, are quite reproducible and thus provide a more reliable determination of the rank order of affinities. The difference in urea concentration required for elution of any mutant relative to the full-length fragment, calculated from the data in column 3 of Table II, or for any pair of mutants differing by a single module is summarized visually in Fig. 6 where results are grouped according to the module that was deleted. Note how module 6 stands out from the others in that its deletion consistently enhances binding, whereas most of the others, especially 1 and 9, have a weakening effect when deleted. DISCUSSION The results presented here indicate that recognition of gelatin is a robust property of fibronectin. Any one of the six modules within the 42-kDa gelatin-binding domain (612789) can be deleted without abrogation of binding. At the same time, all six modules appear to play some role since each of their deletions has a measurable effect on affinity. Thus the situation is more complicated than is often found with domain deletion experiments of this type. For example, the fact that fragment --2789 binds with the same affinity as the full-length fragment would suggest that modules 6 and 1 are not important. But this conclusion is rendered invalid by the observation that removal of 6 alone actually increases the affinity, while removal of 2 alone decreases it by a similar degree. When both are removed the effects appear to compensate.
There are two ways in which a given module might contribute to the interaction with gelatin. The first is direct, i.e. the module engages the ligand, and the other is indirect, i.e. the module interacts with and affects the conformation or disposition of neighboring modules that actually make contact. Some modules could operate at both of these levels. The three-dimensional structure of the trimodular fragment I 6 II 1 II 2 (Fig. 1) shows that modules need not be contiguous to engage in such interactions; I 6 forms an extensive interface with II 2 despite the intervening II 1 (4). The carboxyl terminus of II 2 in that structure is located such that the following module, I 7 , could also interact with I 6 as was inferred from earlier observations that the thermal stability of I 6 was greater in fragments where I 7 was present (5). Thus, there is reason to suspect that the gelatin-binding domain of fibronectin is folded back upon itself providing numerous opportunities for interactions between noncontiguous modules such that the deletion of any one of them could perturb the overall arrangement of the others.

Summary of binding parameters for green GBFs
The letters in the first column refer to the panels in Fig. 3 from which the data in column 3 are derived. Column 4 presents the range of C 50% values obtained in at least two independent ELISA experiments. Trace (Fig. 3 Additional evidence for long range interaction between modules comes from examination of the effect of deleting module 9. As shown in Fig. 6, deletion of this module from 612789, -12789, or --2789 consistently reduces the affinity for gelatin by a similar amount; the ⌬[urea] values vary between Ϫ1.1 and Ϫ1.25 M (Fig. 6). The fact that this effect is lost in fragment ---789 could suggest some type of interaction between modules 2 and 9 that is important for binding. It should also be mentioned that deletion of modules could allow new interactions between other modules. For example, a weak interaction between modules 1 and 2 that was observed in the isolated pair (9) was not present in a larger fragment that also contained module 6 (4).
Replacement of type I modules 7 and 8 with type I modules 4 and 5 from the amino-terminal region abolishes binding both in the full-length fragment and in --2789. This is consistent with modules 7 and 8 serving as more than mere spacers as further evidenced by the fact that the pair alone is able to bind even better than the pair of type II modules. Modules 4 and 5 constitute a rigid pair with an extensive interface (10).
Whether a similar interaction exists between modules 7 and 8 remains to be determined. There is no such interaction between type I modules 1 and 2 in the fib-1 region of fibronectin (11). If the connection between modules 7 and 8 is flexible, then replacing them with modules 4 and 5 may introduce constraints that prevent other interactions that are important for the overall structure and for binding. This is consistent with the fact that simply deleting modules 7 and 8 had less effect than replacing them with 4 and 5.
Module 6 stands out as the only one whose deletion actually enhances binding (Fig. 6). Note, however, that the enhancing effect is greatly diminished in fragments that lack module 2. This suggests that the above-mentioned interaction between modules 2 and 6, which is well documented in the known structure of 612 (Fig. 1), imposes a constraint on the overall structure that slightly impedes the interaction with gelatin. This is opposite to what was reported by the group who determined that structure. Pickford et al. (4) reported that both 612---and -12---bound to gelatin-Sepharose and provided evidence from surface plasmon resonance measurements to suggest that removal of module 6 from 612---increased the affinity toward immobilized ␣(1) chain of human type I collagen ϳ10fold. The discrepancy could relate to the different species of collagen and fibronectin used in the two studies. 2 The K d val- 2 We routinely use swine skin gelatin for affinity chromatography, and the recombinant fragments used in the current study were encoded by cDNA from rat fibronectin. The full-length fragment behaves identically to natural 42-kDa GBF derived from human plasma fibronectin. The species of gelatin used by Pickford et al. (4) was not specified. We also find that the affinity of the 42-kDa GBF for ␣ chains and CNBr fragments of bovine type I collagen (Ref. 12 and Y. Katagiri, S. A. Brew, and K. C. Ingham, unpublished results) is significantly lower than for those from rat tail collagen (1).  Table II. Abs, absorbance.  6. Summary of the effects of each module on the affinity for gelatin. The effects of deleting a given module (indicated at the right) from various parent fragments (indicated at the left) on the peak elution position in the urea gradient ( Fig. 3 and Table II) are grouped according to the module that was deleted. ues that were estimated from limited SPR data by Pickford et al. (4) are designated by ϩ symbols in our Fig. 5. Their position on the urea axis would be consistent with the failure of our corresponding fragments, if they had similar affinities, to bind to our gelatin-Sepharose.
A recent report from this laboratory highlighted the fact that each of the ␣ chains in type I collagen contains multiple sites with similar affinity for 42-kDa GBF (12). The nature of those sites remains unclear. Since it is denatured collagen, i.e. gelatin, that is recognized by fibronectin, it is assumed that what is being recognized is some linear sequence as opposed to a well defined tertiary structure. As of yet the relevant sequence has not been defined, i.e. there is no synthetic collagen-like peptide that has been shown to bind fibronectin with significant affinity. It is not even certain whether the various active subfragments of 42-kDa GBF, such as those described here, are all recognizing the same sequence in gelatin. It is conceivable that neighboring sequences of the collagen chains interact with different parts of 42-kDa GBF. This would provide additional opportunities for module-module interactions to play a role in defining the binding site. The present analysis of more than 20 recombinant fusion products clearly illustrates that all six modules in 42-kDa GBF contribute to its interaction with gelatin either directly by contacting the ligand or indirectly through module-module interactions. Determining which is which will probably require a high resolution structure of a complex between GBF and an appropriate segment of gelatin.