HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Pentikäinen, O. Articles by Johnson, M. S. Search for Related Content PubMed PubMed Citation Articles by Pentikäinen, O. Articles by Johnson, M. S. Social Bookmarking                      What's this?

J Biol Chem, Vol. 274, Issue 44, 31493-31505, October 29, 1999

"RKKH" Peptides from the Snake Venom Metalloproteinase of Bothrops jararaca Bind Near the Metal Ion-dependent Adhesion Site of the Human Integrin ₂ I-domain^*

Olli Pentikäinen^§, Anna-Marja Hoffrén^¶, Johanna Ivaska^**, Jarmo Käpylä, Tommi Nyrönen^§§, Jyrki Heino^**, and Mark S. Johnson^§¶¶

From the  Department of Biochemistry & Pharmacy, Åbo Akademi University, Tykistökatu 6 A, FIN-20520 Turku, ^§ Turku Centre for Biotechnology, University of Turku and Åbo Akademi University, Tykistökatu 6 B, FIN-20520 Turku, ^** MediCity Research Laboratory, Department of Medical Biochemistry, University of Turku, Tykistökatu 6 A, FIN-20520 Turku,  Department of Biological and Environmental Science, University of Jyväskylä, FIN-40351 Jyväskylä, and ^§§ Center for Scientific Computing, FIN-02120 Espoo, Finland

	ABSTRACT

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Integrin ₁₁ and ₂₁ are the major cellular receptors for collagen, and collagens bind to these integrins at the inserted I-domain in their  subunit. We have previously shown that a cyclic peptide derived from the metalloproteinase domain of the snake venom protein jararhagin blocks the collagen-binding function of the ₂ I-domain. Here, we have optimized the structure of the peptide and identified the site where the peptide binds to the ₂ I-domain. The peptide sequence Arg-Lys-Lys-His is critical for recognition by the I-domain, and five negatively charged residues surrounding the "metal ion-dependent adhesion site" (MIDAS) of the I-domain, when mutated, show significantly impaired binding of the peptide. Removal of helix C, located along one side of the MIDAS and suggested to be involved in collagen-binding in these I-domains, does not affect peptide binding. This study supports the notion that the metalloproteinase initially binds to the ₂ I-domain at a location distant from the active site of the protease, thus blocking collagen binding to the adhesion molecule in the vicinity of the MIDAS, while at the same time leaving the active site free to degrade nearby proteins, the closest being the ₁ subunit of the ₂₁ cell-surface integrin itself.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Integrins are cell-surface heterodimeric glycoproteins that anchor cells to their surroundings through cell-cell and cell-matrix interactions and are thus able to transduce bi-directional signals between the cytoplasm and the extracellular matrix or other cells. Integrins are composed of one  subunit (17 varieties) and one  subunit (8 varieties) and associate non-covalently to form over 20 different complexes. Both  and  subunits consist of a large extracellular domain, a membrane spanning sequence, and a cytoplasmic tail.

A well characterized function of the integrins is to recognize and bind extracellular matrix proteins containing the Arg-Gly-Asp (RGD)¹ sequence (the fibronectin cell adhesion motif), which is also found on the surfaces of a wide range of other extracellular matrix glycoproteins in all metazoans from marine sponges to mammals (1). One ₁, ₃, ₅, or ₆ subunit, in combination with one  subunit, can form an RGD-dependent receptor. In the ₁ subunit, the recognition site for the RGD motif is located near the amino terminus. Most integrins that contain the ₁ subunit must have additional binding sites since they recognize their ligands in an RGD-independent manner and bind to other protein motifs. For example, ₄₁ binds the sequence Leu-Asp-Val in fibronectin and Ile-Asp-Ser-Pro in VCAM-1. Integrins ₆₁ and ₇₁ are RGD-independent laminin receptors, and both ₁₁ and ₂₁ integrins are RGD-independent collagen receptors. Their binding motifs are unknown.

The ₁ and ₂ integrin subunits, and the recently identified ₁₀ I-domain (2), differ in their structure from other ₁-associated  subunits and have an additional ~200 amino acids long "inserted" I-domain, which is homologous to the von Willebrand factor A domain (3). It is this I-domain in the ₁₁ and ₂₁ integrins that is responsible for the recognition of native collagen. The ₂ I-domain also binds two other ligands, laminin-1 and echovirus-1, but the echovirus-1-binding site is distinct from the matrix protein site on the I-domain. Integrin ₂₁ is the major collagen receptor of platelets and many cell types, such as epithelial cells and fibroblasts. It has been associated with various cellular functions, including migration on collagen, invasion through collagenous matrix, and reorganization of paracellular collagen fibers. In cancer biology, ₂₁ integrin may be essential for metastasis by certain tumor cells.

Several snake venoms contain disintegrin-like proteins, which block integrin function. Many of these disintegrins contain the RGD motif or a closely related motif and inhibit the function of _IIb3 or _V3 integrins. Another toxin, jararhagin, from Bothrops jararaca, seems to inhibit adhesion to collagen (platelet aggregation) in an ₂₁-dependent manner (4-5) but uses a different mechanism. Jararhagin is a polyprotein containing a propeptide, a metalloproteinase domain, a disintegrin-like domain, and a cysteine-rich domain. The disintegrin-like domain does not contain the RGD motif, whereas it is replaced by Glu-Cys-Asp. The present hypothesis about jararhagin action suggests that it first binds to the ₂ I-domain and then degrades the ₁ subunit (5). The interaction between jararhagin and ₂ I-domain has never been shown directly, and the domain in jararhagin interacting with ₂₁ integrin is not known. There is some evidence that the disintegrin domain of jararhagin, called jaracetin, as a dimer could alone inhibit platelet adhesion to collagen, but it seems to be a weaker inhibitor than jararhagin. Our attempts to show that the ₂ I-domain binds to jaracetin or to disintegrin domain-derived peptides failed (6), and therefore the putative interaction between the metalloproteinase domain and ₂ I-domain became more interesting. Furthermore, in the snake venom, a part of jararhagin is normally degraded, and the metalloproteinase domain is dissociated from the disintegrin-like domain, suggesting that the metalloproteinase domain might have some independent functions.

Based upon the prediction of surface loop regions along the metalloproteinase sequence, peptides were synthesized and evaluated for inhibition of collagen binding to recombinant ₂ I-domain (6). Each of these peptides included a negatively charged residue, since it was suggested that an acidic residue would play the part of the sixth missing ligand to the bound metal of the MIDAS in the I-domains (7-8). A nine-residue aspartate-containing peptide was obtained from the metalloproteinase that, when cyclized through the addition of terminal cysteines and oxidized to form a disulfide bond, was found to inhibit collagen binding to the ₂ I-domain. Surprisingly, alanine replacement of individual residues of the peptide showed that the aspartate was not critical for binding, instead the sequence Arg-Lys-Lys-His was essential. Based on these results, a model of the jararhagin metalloproteinase was made and used as the basis for the design of a smaller peptide capable of stronger blocking of collagen binding to the I-domain. The stable conformation of this loop allowed us to predict the complementary characteristics on the surface of a model of the structure of the ₂ I-domain (as the three-dimensional structure had not yet been reported) and to identify 5 negatively charged residues as potential sites of interaction with the "RKKH" peptides. Binding studies made with mutant ₂ I-domains demonstrated that these five residues on the surface of the ₂ I-domain surrounding the MIDAS are critical for ligand binding, whereas the removal of a short helix, suggested to be important for collagen binding and located along one side of the MIDAS, exerted no effects on peptide ligand binding.

	EXPERIMENTAL PROCEDURES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Generation of ₂ I-domain and Mutant I-domains-- Recombinant ₂ I-domain was produced as described earlier (6). Site-specific mutations were introduced into the ₂ I-domain using the Stratagene QuickChange mutagenesis kit, essentially following the manufacturer's instructions as follows: PCR primers having desired point mutations, indicated by the name of the mutant, were designed. PCR was then performed using the Pfu polymerase (Stratagene), which at 68 °C makes exactly one copy of the whole GEX-2T vector (Amersham Pharmacia Biotech) containing the ₂ I-domain sequence. The PCR reaction was digested with DpnI, which only cuts methylated DNA (i.e. only the template is digested). In the mutant C₂I, amino acids 284-288 (GYLNR) were deleted by using PCR with specifically designed primers. In each case, the resulting GEX-2T having mutated ₂ I-domain was transformed into Escherichia coli strain DH5, and the construct was verified by sequencing the entire ₂ I-domain. For protein production, each construct was transformed into E. coli strain BL21. Protein concentrations were determined using the Bradford method (9). Protein purity and folding were checked in both native and SDS-polyacrylamide gel electrophoresis.

Binding Assay Using Biotinylated 229ox or 248ox-- The peptides were synthesized on an automated peptide synthesizer (Applied Biosystems 431A) using Fmoc (9-fluorenylmethylcarbonyl) chemistry. After synthesis, peptides were oxidized to form disulfide bridges. The peptides (229ox, "CTRKKHDNAQC"; 248ox, "CTRKKHDC"), described earlier (6), were solubilized at 1 mg/ml concentration with 0.1 M ammonium carbonate buffer and incubated for 16-24 h at 4 °C. The oxidation was checked by reverse-phase high pressure liquid chromatography, and the oxidized peptides were lyophilized. Biotinylation of the peptides was carried out as follows: lyophilized peptide was solubilized in PBS and 1/5 volume of 0.1 M NaHCO₃, 0.5 M NaCl (pH 8.0) was added to elevate the pH for biotinylation. Sulfo-NHS-biotin (Calbiochem) was added 1:2 (w/w) peptide:biotin and incubated for 2 h at room temperature. The biotinylation reaction was stopped by adding 1/10 volume 0.5 M Tris-HCl (pH 8.0).

The binding assays using biotinylated peptide were performed as described earlier (6). Briefly, 96-well amine binding plates (Costar) were coated with recombinant ₂ I-domain or mutant variants (1 mg/ml) according to the manufacturer's instructions. Residual protein absorption sites on all wells were blocked with 2% bovine serum albumin in PBS for 1 h at 37 °C. 100 µM biotinylated 229ox or 248ox in PBS, 2 mM MgCl₂, and 1 mg/ml bovine serum albumin were added at the indicated concentrations to the coated wells and incubated for 3 h at 37 °C. For the studies involving anti-₂ antibodies, 0.1 µg/ml antibodies (5E8, a kind gift from Dr. Bankert, Roswell Park Cancer Institute; 12F1 (10); G_i9 (11)) were added to the wells together with the biotinylated 229ox peptide at a concentration of 100 ng/ml. After incubation, wells were washed 6 times with PBS, 2 mM MgCl₂, and 1 mg/ml bovine serum albumin for 30 min at room temperature. Wells were again washed 6 times. Finally, 0.1 ml of Delfia enhancement solution (Wallac) was added to each well, and Europium signal was measured by fluorometry (model 1232 Delfia, Wallac).

Structural Modeling-- Three-dimensional structures were obtained from the Brookhaven Protein Data Bank (12). Sequence alignments were made with the programs MALIGN and MALFORM (13, 14). Based on the alignments, initial models (for review see Ref. 15) of both the metalloproteinase and the ₂ I-domain were constructed using COMPOSER (16-18) (SYBYL 6.5, Tripos Associates, St. Louis). Final models were made with MODELLER 4.0 (a kind gift from Andrej ali, Rockefeller University) (19), which allowed us to incorporate additional constraints as follows: disulfide bonds, in the case of the metalloproteinase, and the published secondary structure, in the case of the ₂ I-domain. Models built with COMPOSER were energy-minimized using the TRIPOS force field and the steepest descent method. Initially, each model's backbone was kept rigid, and only the side chains were allowed to move. Subsequently, all atoms were allowed to move, and energy minimization was performed until all short contacts and inconsistencies in geometry were rectified. The main purpose of this minimization was to remove steric hindrances and bad geometry.

Conformation and Flexibility of RKKH Peptides-- The conformational flexibility of the original cyclic peptide in the cyclic form was assessed using molecular dynamics simulations. The starting conformation of the peptide was taken from the metalloproteinase model structure; cysteines were added to each end, and a disulfide bond was created between them. Peptides derived from metalloproteinase were first minimized to remove atom-atom clashes and then further refined by molecular dynamics simulations. Simulations were performed in vacuum at 300 K and consisted of 20-ps equilibration followed by a 200-ps production run. The SHAKE algorithm from the SYBYL package was applied to constrain the lengths of all bonds between heavy atoms and hydrogen atoms. With the SHAKE algorithm we are able to use a longer 1-fs time step, instead of a 0.5-fs time step, thus achieving a longer production run for the same computational time. Electrostatics were excluded, because small peptides tend to form intramolecular hydrogen bonds to make the structure globular, especially when many charged residues are present within a peptide. All calculations were made using SYBYL and the TRIPOS force field on a Silicon Graphics Onyx II workstation.

Peptide-I-domain Docking Studies-- The essential chemical interactions in the ₂ I-domain binding site were mapped using the program GRID version 16 (20). GRID calculates energies of interaction between a probe and the receptor. In the calculation, the probe that mimics a chemical group is placed at different positions throughout the binding site, and the receptor side chains are allowed to move to minimize the interaction energy (using the side chain flexibility option in GRID). The GRID maps were visualized using the program CERIUS 2 (Molecular Simulations Inc., San Diego) and GRASP (21).

The program Autodock 2.4 (22-23) was used to both flexibly (allowing rotation of up to 25 torsion angles) and rigidly dock the 248ox peptide to the crystal structure of ₂ I-domain. Autodock combines Monte Carlo-simulated annealing for conformational searching with a rapid, atomic resolution, grid-based method of energy evaluation utilizing the AMBER force field (24-25). Standard AMBER parameters for Mg²⁺ were added to the default Autodock parameter set. A distance-dependent dielectric constant was used to account for the solvent screening effects. The interaction of a probe group (corresponding to each type of atom in the ligand) with receptor model was calculated at grid positions 0.3 Å apart in a 30-ų box centered at the binding site using the program Autogrid (23).

Fifty separate docking simulations were performed; for each simulation there were 100 constant temperature cycles with 15,000 steps accepted or rejected. The initial value of RT (the gas constant times the absolute temperature) was set equal to 300 cal/mol and was reduced by a factor of 0.95 in each cycle. The maximal torsional rotation and molecular center translation steps were 15° and 0.2 Å, respectively, and they were reduced by a factor of 0.99 in each cycle. In this way, over 70 million different conformations were studied for the "rigid" and the "flexible" peptide. Representative alternative modes of binding were identified using cluster analysis of the 50 docked structures from the flexible and rigid docking simulations. A 2.0-Å cut-off value of the root mean square deviation calculated over all atoms in the 248ox peptide was used to define a new cluster.

The computer program GOLD (26) was used as a second method to produce alternate binding modes for the 248ox-₂ I-domain complex. GOLD allows full ligand flexibility and partial protein flexibility. The energy functions in GOLD are based partly on experimental information about hydrogen bond geometries taken from the IsoStar data base (27). The program exploits the observed distributions of torsional angles seen in the Cambridge Structural Data Base to search the conformational space available to the ligand. GOLD also considers the fundamental requirement that the ligand must displace loosely bound water on binding. A genetic algorithm is used to search the conformational space. Twenty separate dockings were made within a 17-Å radius sphere centered at the Mg²⁺ at the MIDAS using the default genetic algorithm parameters.

The docked conformations produced with Autodock and GOLD were visually inspected to identify similarities in the predicted binding modes produced by each method. The similar binding modes were then superimposed onto the GRID maps. The 248ox conformations produced with both programs that best fitted the affinity surfaces calculated with GRID were chosen as the representative ones. Autodock and GOLD were run on computers at the Center for Scientific Computing, Espoo, Finland. The results were visualized using the programs CERIUS 2, GRASP, SYBYL and gOpenMol (28-29).

	RESULTS

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

With the knowledge that we had identified a peptide from the metalloproteinase of jararhagin capable of blocking collagen binding to the ₂ I-domain (6), we sought here to characterize the interactions between the peptide and the I-domain. In this respect we 1) modeled the metalloproteinase domain of jararhagin; 2) identified the location and conformation of the peptide 229ox on the metalloproteinase model structure; 3) used computer simulations to suggest alterations to the original cyclic 9-residue RKKH peptide that would lead to a tighter binding peptide; 4) modeled the 2 I-domain structure to help identify the likely site of interaction between the metalloproteinase and derived peptides with the I-domain; 5) used function blocking antibodies to provide general support for the predicted sites of interaction proposed by the modeling results; 6) used site-directed mutagenesis of key residues to demonstrate in binding studies that 5 negatively charged residues are important for binding the positively charged peptides from the jararhagin metalloproteinase; 7) and, as described under "Discussion," we used computational chemistry to dock the ₂ I-domain with the RKKH peptide and the metalloproteinase of jararhagin.

A Model Structure of the Metalloproteinase Domain of Jararhagin-- A three-dimensional model of the structure of the zinc metalloproteinase domain of jararhagin was constructed based on homology of the jararhagin metalloproteinase sequence (SwissProt code: DISJ_BOTJA, (30)) with the 2.0-Å resolution structure of another snake toxin zinc metalloproteinase, the collagenase adamalysin II from the Eastern diamondback rattlesnake Crotalus adamanteus (Brookhaven Protein Data Bank code 1IAG (31)). The adamalysin II structure consists of a single twisted sheet of five -strands, surrounded by five -helices, for 201 of the total 202 amino acids in the processed proteinase; residue 1, a pyrrolidone carboxylic acid, is not present in the structure. The catalytic glutamic acid, Glu¹⁴³ (numbered according to the structure coordinate file; Fig. 1), is located within a cluster of residues that include the three histidine ligands (His¹⁴², His¹⁴⁶, and His¹⁵²) of the essential zinc metal bound at the active site. In addition, adamalysin II has 4 cysteine residues that form two disulfide bonds between Cys¹¹⁷ and Cys¹⁹⁷, and Cys¹⁵⁷ and Cys¹⁶⁴.

View larger version (26K): [in this window] [in a new window]   Fig. 1.   Sequence alignment of the snake venom metalloproteinases from B. jararaca and C. adamanteus. The secondary structure of the C. adamanteus metalloproteinase structure (30) is also shown (-helices, cylinders; -strands, arrows). Identically aligned residues are boxed. Key residues in the C. adamanteus structure that match residues in B. jararaca are shown in italics: the catalytic glutamate, the histidine ligands of the essential bound zinc, and the cysteine residues involved in disulfide bonds; Cys312 and Cys334 in the jararhagin metalloproteinase sequence are near each other, exposed to solvent, and potentially could form an additional disulfide bond; the remaining Cys339 appears to be buried on the inner face of the carboxyl-terminal helix. The 9-residue peptide from the metalloproteinase domain of jararhagin, which inhibits collagen binding to the 2 I-domain when cyclized through a terminal disulfide bond, is shown in bold. The figure was produced with ALSCRIPT (59).

The sequence alignment of the two snake venom metalloproteinase domains is shown in Fig. 1 along with the secondary structure assignments made from the x-ray structure of adamalysin II; these proteins share approximately 50% sequence identity, and there are no insertions or deletions present in the alignment. The 201-residue-long model begins with residue Tyr¹⁵⁵ (residue numbering begins with the first residue of the pro-proteinase sequence), the fifth residue in the putative processed protein, and continues until the carboxyl-terminal position Pro³⁵⁵. The key catalytic glutamate (Glu²⁹⁶ in the jararhagin metalloproteinase) and the three ligands of the bound zinc (His²⁹⁵, His²⁹⁹, and His³⁰⁵), common to the snake venom metalloproteinases, are completely conserved in the alignment with jararhagin (Fig. 2). In addition, the jararhagin metalloproteinase sequence contains 7 cysteine residues lying within the carboxyl-terminal half of the metalloproteinase; Cys²⁷⁰ and Cys³⁵⁰, and Cys³¹⁰ and Cys³¹⁷ align with the disulfide-bonded pairs seen in the adamalysin II metalloproteinase structure, but there are no cysteines in adamalysin II corresponding to the additional Cys³¹², Cys³³⁴, and Cys³³⁹ seen in the jararhagin metalloproteinase sequence (Fig. 1). Cys³¹² and Cys³³⁴ are close enough to each other in space in our preliminary models that they could easily form a disulfide bond, and this is included in the final model. The remaining Cys³³⁹ is not located on the surface of the metalloproteinase model but lies on the inner buried surface of carboxyl-terminal helix E and is unlikely to participate in forming a disulfide bond. Given the very high level of sequence identity and the additional constraints provided by the three disulfide cross-links, the model of the metalloproteinase domain should be quite close to the authentic three-dimensional structure of the jararhagin metalloproteinase.

View larger version (69K): [in this window] [in a new window]   Fig. 2.   Modeled structure of metalloproteinase domain of jararhagin. Ribbon drawing (made using MOLSCRIPT (60) and RASTER3D (45)) (A) showing the side chains of key residues involved in catalysis, coordination of the essential zinc metal, Ca2+ binding, and cysteines forming predicted disulfide bonds. Molecular surface (B) showing the surface charge distribution about the RKKH peptide calculated using the program GRASP (21) and rendered using RASTER3D (45): positive charge (blue), negative charge (red), and neutral (white). Critical residues that are essential for binding to the 2 I-domain are labeled: Arg242-Lys243-Lys244. His245 is buried and located directly below the central portion of the loop where a hydrogen bond links the imidizole NH group of His245 to the main chain carbonyl of Lys243.

The model of the jararhagin metalloproteinase pinpointed the location of the sequence TRKKHDNAQ contained within the peptide 229ox that was shown to block collagen binding to the ₂ I-domain (6). Peptide 229ox has, in addition to this sequence of 9 residues from jararhagin, a cysteine at each end, and oxidation to form the cyclized peptide was shown to be required to block collagen binding to the ₂ I-domain; the reduced linear peptide did not block collagen binding (6).

The Arg²⁴²-Lys²⁴³-Lys²⁴⁴-His²⁴⁵ motif of the peptide is a distinct feature on the surface of the metalloproteinase next to the Ca²⁺-binding site, is distant from the active site, and forms a solvent-exposed loop connecting -helix C to the following -strand 3 (Figs. 1 and 2). Interestingly, the aspartate used as one criteria for selection of the peptide, given reports that an acidic residue binding to the MIDAS might be generally important (7-8), was in fact a ligand of the bound Ca²⁺ in the rattlesnake metalloproteinase structure (31) and is shielded from the solvent (Fig. 2A). Furthermore, alanine replacement of the aspartate (in the peptide derived from the jararhagin metalloproteinase) led to no change in the ability of the cyclic peptide to block collagen binding in comparison with the original peptide (6). Alanine replacement of all other residues within the peptide did demonstrate, however, that the three surface-exposed positively charged residues (Fig. 2B) of the Arg-Lys-Lys-His motif are essential for binding to the ₂ I-domain (6), and alanine replacement of the histidine showed some effect on binding too.

In the fibrinogen-cleaving ruberlysin of the red diamondback rattler Crotalus ruber ruber (HRT2_CRORU (32)) a similar sequence of three consecutive positively charged residues is found, Arg-Lys-Arg-His. However, this loop region differs in sequence from the other snake venom sequences. In adamalysin II, the sequence in this region was determined from electron density maps, and reports vary from Trp-Lys-Arg-His in the deposited crystal coordinate file (Protein Data Bank code 1IAG) to Arg-Lys-Arg-His in the crystallographic publications (31, 33-34) to Lys-Lys-Lys-Lys in the current Swiss-Prot entry (ADAM_CROAD (31)). The corresponding sequence is Arg-Lys-Ser-His in atrolysins C, D, and E of the Western diamondback rattler Crotalus atrox (HRTD_CROAT and HRTE_CROAT (35-36)) that cleaves type IV collagen and gelatin and prevents platelet aggregation and whose crystal structure is also known (37). Two other hemorrhagic metalloproteinases have sequences Arg-Thr-Ser-His (Indian green tree viper Trimerisurus gramineus, DISA_TRIGA (38-39)) and Arg-Ile-Ser-His (Bushmaster snake Lachesis muta muta (HRL2_LACMU (40)). Thus, this motif is not strictly conserved among the snake venom zinc metalloproteinases, yet the following aspartate involved in calcium binding is found in each of these sequences, as well as in several mammalian metalloproteinase sequences too (33).

Conformational Flexibility and Predicted Binding of RKK-containing Peptides-- As shown in Fig. 2, the side chains of the Arg-Lys-Lys sequence are exposed to solvent, and the three positive charges form a planar feature on the surface of the metalloproteinase. The histidine lies internally and forms a side chain hydrogen bond to the main chain carbonyl of Lys²⁴³ within the loop and hydrogen-bonds to the side chain of a buried arginine Arg²³⁵. The loop conformation seen in the adamalysin II structure used to model the jararhagin metalloproteinase matches very closely the conformation seen in the metalloproteinase structure from the venom of the Western diamondback rattlesnake too (37). Experimental evidence has shown that the loop conformation is important for blocking collagen binding to the ₂ I-domain, since the linear peptide was non-functional (6). Consequently, we explored the conformational flexibility of peptide 229ox and other peptides derived from this region of the jararhagin metalloproteinase in order to optimize the binding properties of the RKKH-containing peptide. We sought to reduce the size of the cyclic peptide and to increase its rigidity in such a way that the peptide would maintain, as best as possible, the backbone conformation present in the model structure.

The conformational flexibility of the original cyclic peptide in the cyclic form was assessed using molecular dynamics simulations. The starting conformation of the peptide was taken from the metalloproteinase model structure, and cysteines were added to each end, and a disulfide bond was created between them. Table I lists the sequences and lengths of peptides studied by molecular dynamics simulations and their experimentally determined IC₅₀ values.

The backbone and side chain conformations of the minimized cyclized peptide 229ox were very similar to the original loop region from metalloproteinase model, which was used as a basis for the peptide models. After simulation, the peptide corresponding to 229ox (9 residues from the loop region of the metalloproteinase plus two added cysteines), as well as peptides having 8 and 7 residues, had distorted twisted backbone structures when compared with the minimized 229ox (Fig. 3A). Interestingly, the backbone of each of these peptides was distorted from the histidine residue onward to the carboxyl terminus, suggesting that this region of the peptides is more flexible than the amino-terminal region where the RKKH sequence is located. Although the 9-residue peptide did block collagen binding with an IC₅₀ = 52 ± 20 µM, the 8-residue peptide was a poor blocking agent with an IC₅₀ >10 mM, and the 7-residue peptide failed to block collagen binding (Table I (6)).

View larger version (21K): [in this window] [in a new window]   Fig. 3.   Molecular dynamics simulations of cyclic RKKH peptides and comparison with the loop conformation in the jararhagin metalloproteinase model. Superimposed C atom traces of the metalloproteinase loop conformation (green, residues Thr241-Gln249) with the 9-residue-long 229ox peptide (yellow, Thr241-Gln249 plus terminal half-cystines), the 8-residue peptide (red, Thr241-Ala248 plus terminal half-cystines), and the 7-residue peptide (cyan, Thr241-Asn247 plus terminal half-cystines) (A); the 5-residue (red, Thr241-His245 plus terminal half-cystines) and 4-residue (cyan, Arg242-His245 plus terminal half-cystines) peptides (the conformation of the histidine side chain is shown for each peptide) (B); and the 6-residue 248ox peptide (Thr241-Asp246 plus terminal half-cystines; the side chains of the RKKH motif are shown) (C). For those side chains shown as ball and stick figures, carbon atoms are colored gray and nitrogen atoms blue. In these stereo figures, the amino terminus of each structure is on the right and the carboxyl terminus is on the left; the disulfide bond between terminal cysteines is not shown. Figure was made using MOLSCRIPT (60).

Simulations of the 5- and 4-residue peptides suggest that they are too short and constrained: when the disulfide bond is formed between the terminal cysteines the histidine side chain is forced to flip onto the other side of the backbone (Fig. 3B). Consistent with these computer-based observations, the 5-residue peptide plus terminal half-cysteines, CTRKKHC, were synthesized but did not cyclize when oxidized.

The six-residue peptide plus terminal half-cysteines, CTRKKHDC (248ox), is more rigid than the longer peptides, does not distort under prolonged simulations (Fig. 3C), and has less ring strain than is seen for the 4- and 5-residue peptides when they are forced into a cyclic conformation. The 6-residue peptide most closely approximated the metalloproteinase loop conformation, both backbone and side chains. We proposed that only this peptide would function better than the original 9-residue peptide. When tested, the 6-residue peptide had an IC₅₀ = 1.3 ± 0.2 µM (Table I), consistent with this prediction.

Modeling of the Human ₂ I-domain-- In order to help identify possible sites of binding of the metalloproteinase and proteinase-derived peptides on the ₂ I-domain, a model of the I-domain three-dimensional structure was made. A three-dimensional structure of the human ₂ I-domain has since been reported (41), and the coordinates are now available. The majority of this work, including our proposal that the metalloproteinase-derived peptides bind to negatively charged residues surrounding the MIDAS, was carried out in advance of that structural report.

The three-dimensional model of the structure of the ₂ I-domain was constructed on the basis of homology with two available x-ray structures of integrin I domains in the Brookhaven Protein Data Bank (12), the 2.0-Å resolution structure of human "A-domain" of the _M2 leukocyte complement receptor type 3 CD3 (CD11b/CD18; Protein Data Bank code 1JLM (7)) and the 1.8-Å resolution I-domain structure of the _L2 (CD11a/CD18; Protein Data Bank code 1LFA (42)), which is a leukocyte receptor for intercellular adhesion molecules ICAM-1, ICAM-2, and ICAM-3. The I-domains represented by _M and _L are classic "Rossmann" folds having five parallel -strands plus one -strand antiparallel to the others, surrounded by six -helices. A metal ion, Mg²⁺ or Mn²⁺, is bound at the carboxyl-terminal end of the five -strands and thus near the surface of the I-domain. The metal ion is coordinated by residues of the "MIDAS" motif that includes the "Asp-Xaa-Ser-Xaa-Ser" sequence (residues 140-144 in _M; Xaa refers to a more sequence variable position) and Thr²⁰⁹ and Asp²⁴² in _M; a sixth ligand to the bound metal is thought to be provided by the side chain of an acidic amino acid from the molecules recognized by the I-domain (7-8).

The alignment of the ₂ I-domain sequence with the sequences of both _M and _L is shown in Fig. 4, where the alignment between _M and _L is based on their structural superposition over their C backbone prior to alignment with the ₂ I-domain sequence (43). The sequence identity between the ₂ I-domain and _M and _L is approximately 26 and 25%, respectively (_M and _L are approximately 35% identical). The matching of the ₂ I-domain sequence to the sequences of the known structures is quite reliable; there are only 5 insertions and 1 deletion compared with _L and 3 insertions and 2 deletions compared with _M, and the key residues forming the MIDAS-binding site are conserved (Fig. 4). Indeed, on the basis of the published details of the ₂ I-domain structure, the alignment of Fig. 4 was not changed.

View larger version (43K): [in this window] [in a new window]   Fig. 4.   Sequence alignment of the human integrin I-domains. Secondary structure assignments are shown for the known three-dimensional structures: the L (42), M (7), 1 (44)2, 2 (41) I-domains and the von Willebrand factor (46-47); the structure of the 10 I-domain has not yet been determined. Residues conserved throughout all six sequences are boxed. The five negatively charged amino acids in the 2 I-domain and matched residues in the 1 and 10 I-domains are shown in bold type. In the 2 I-domain, these residues are candidates for binding the metalloproteinase of jararhagin and derived RKKH peptides, and these residues are located in the vicinity of the MIDAS: Asp219, Glu256, Asp259, Asp292, and Glu299 in the 2 I-domain structure (numbering according to the full-length 2 subunit sequence). The figure was produced with the program ALSCRIPT (59).

The original homology model of the ₂ I-domain (Fig. 5A) was built with COMPOSER (15-18) and rebuilt (when the report by Emsley et al. (41) was published, we incorporated the C helix into the model) using MODELLER 4.0 (19) on the basis of the alignment in Fig. 4. The proposed ₂ I-domain binding site of the cyclic peptide derived from the metalloproteinase of jararhagin and identification of the key Asp²¹⁹ on the I-domain critical for binding the peptide were initially located using our original model (Fig. 5A; see below).

View larger version (83K): [in this window] [in a new window]   Fig. 5.   Modeled structure of the 2 and 10 I-domains and the x-ray structures of the 2 and 1 I-domains. Surface views of the human integrin 2 I-domain at the proposed binding site of the RKKH-containing peptides: original model (A) and x-ray determined structure (41) (B). Similar views are shown for the x-ray structure of human integrin 1 I-domain (44)2 (C); a model structure of the 10 I-domain made on the basis of the 1 and 2 I-domain structures (D). The surface charge distributions were calculated using the program GRASP (21) and rendered using RASTER3D (45): positive charge (blue), negative charge (red), and neutral (white). For each display, the charged residues surrounding the central MIDAS metal-binding site are labeled, as are the locations of the bound metal (Mg) and helix C.

The ₂ I-domain structure (Fig. 5B) described by Emsley et al. (41) pinpointed one major difference with the _L and _M structures: the ₂ I-domain has an additional helix, residues Gly²⁸⁴-Arg²⁸⁸, also present in the structure of the ₁ I-domain (44)² and predicted on the basis of the sequence alignment (Fig. 4) to be present in the other collagen-binding I-domain (the recently reported ₁₀ I-domain (2)). This helix is located at the surface of the I-domain in the vicinity of the metal-binding site and, in the alignment of the I-domains (Fig. 4), corresponds to a deletion in _L (42), _M (7), and the von Willebrand factor A-domain (46-47) structures (Fig. 4). This helix leads to a deeper pocket above the MIDAS metal-binding site and may be responsible for binding specificity since this region is suggested to be the location of the collagen-binding site (41); as we shall show below, our experimental results support this proposition. The surface of the ₂ I-domain x-ray structure (Fig. 5, A and B) about the MIDAS is clearly different in terms of charge in comparison with the recently solved structure (44)² of the ₁ I-domain (Protein Data Bank code 1QCY; Fig. 5C) and our model structure (based on the alignment in Fig. 4 and using the structure of the ₂ and ₁ I-domains) of the ₁₀ I-domain (Fig. 5D).

Binding of the RKK Peptide to the ₂ I-domain-- The proposed ₂ I-domain binding site of the cyclic peptide derived from the metalloproteinase of jararhagin was initially located by visual inspection of our original I-domain model (Fig. 5A). Since the peptide that prevents collagen binding to the ₂ I-domain presents a triangular array of three positively charged side chains in both the model of the metalloproteinase structure and the constrained cyclized peptides, we examined the surface of the I-domain model structure for a complementary triangular array of negatively charged residues. Only one region of the surface of the model met these requirements, and it is located at the carboxyl-terminal ends of five of the six strands, just above and partly surrounding the metal-binding MIDAS. Indeed, there are five acidic residues that could provide complementary ionic interactions with the three positively charged residues of the metalloproteinase loop: Asp²¹⁹, Glu²⁵⁶, Asp²⁵⁹, Asp²⁹², and Glu²⁹⁹ in the ₂ I-domain (Figs. 4 and 5A).

This proposal that RKKH peptides bind to the ₂ I-domain in the vicinity of the MIDAS was generally supported by the effects of function-blocking antibodies on 229ox peptide binding to the ₂ I-domain (Fig. 6). In this study, we tested a set of ₂ I-domain-binding antibodies that are known to have different functions in order to learn more about the binding site of the RKKH peptides on the surface of the ₂ I-domain with respect to the binding sites for collagen and echovirus-1. Of the antibodies tested, 12F1 inhibits echovirus-1 binding (10-11, 48), Gi9 inhibits collagen binding (11), and 5E8 inhibits both echovirus-1 and collagen binding (11). As seen in Fig. 6, both IgG (used as the nonspecific control) and 5E8 have no effect on the binding of biotinylated 229ox peptide to ₂ I-domain bound to the solid phase in the assay. Both antibodies 12F1 and Gi9, however, did inhibit the interaction significantly. The binding site of the antibody 5E8 has been mapped to Tyr²¹⁶ on the ₂ subunit of the I-domain (49), whereas 12F1 and Gi9 are both known to bind either to residues 173-199 or to residues 217-259 (49-50). Tyr²¹⁶ is nearby but peripheral to the MIDAS; residues 173-199 form two consecutive -strands and map to the side of the I-domain well away from the MIDAS, whereas residues within the segment 217-259 include 3 of the 5 negatively charged residues that surround the MIDAS and were suggested by us to be involved in peptide binding.

View larger version (20K): [in this window] [in a new window]   Fig. 6.   The effect of 2 I-domain blocking antibodies on RKKH peptide binding. Recombinant 2 I-domain (1 mg/well) was bound to 96-well amine-binding plates, and biotinylated 229ox (100 mM) was added in the presence (IgG as a control or 2 I-domain-specific antibodies: 12F1, 5E8, or Gi9) or absence (No Ab) of antibody (100 ng/ml) and allowed to bind for 3 h at 37 °C. The wells were washed six times, and Europium-labeled streptavidin at a concentration of 500 ng/ml was added for 30 min at room temperature. Wells were washed six times, and Europium signal was measured. The data shown are the mean values (±S.D.) of a representative experiment done in triplicate.

In order to test our proposal for peptide-₂ I-domain interaction, a series of mutant I-domains were produced. The ₂ I-domain variants were immobilized on the bottom of a microtiter well, and their ability to interact with biotinylated RKKH peptide (Bio-248ox) was tested. As seen in Fig. 7A, deletion of the C helix (residues 284-288 in the ₂ I-domain; see Fig. 4), a unique characteristic of the collagen-binding ₁, ₂, and ₁₀ I-domains, has no effect on the recognition of the peptide since the mutant C₂I binds to the peptide as efficiently as the wild type I-domain, but the deletion does affect collagen binding to the ₂ I-domain.³ Aspartate 219 was the first residue we suspected on the basis of modeling studies of having involvement in peptide binding, and an initial binding study with a D219N mutant (6) showed a clear 5-fold reduction in the binding of the RKKH peptide to the ₂ I-domain (Fig. 7A). When this residue was replaced by alanine (methyl group side chain), no difference from the effects with the wild type were seen (Fig. 7A) but, surprisingly, when replaced by arginine (a large positively charged amino acid) the peptide bound significantly better to the mutant ₂ I-domain (Fig. 7A). Arginine is found at the position equivalent to Asp²¹⁹ in both the ₁ and ₁₀ I-domains (Figs. 4 and 5, C and D), and the RKKH peptides 229ox, 248ox, and a T241A mutant of the 248ox peptide have also been shown by us to bind to isolated ₁ I-domain and to block collagen-I binding to ₁ I-domain.³ The arginine side chain has a large hydrophobic part (three methylene groups) in addition to the positively charged guanido group at the end of the side chain. The peptide itself has a large hydrophobic surface, and the D219R mutation may function to increase the hydrophobic surface interactions between the two. This is supported by the D219N mutation (negative charge  neutral, but where hydrogen bonds can still be formed), which suggests that a charge pair interaction is not critical for binding in the case of Asp²¹⁹, and by the D219A mutation (small hydrophobic side chain) where no effects are seen on ligand binding.

View larger version (13K): [in this window] [in a new window]   Fig. 7.   Mutagenesis of residues surrounding the MIDAS in the 2 I-domain and RKKH peptide recognition. A and B, wild type (WT) or mutant recombinant 2 I-domain (1 mg/well) were bound to 96-well amine-binding plates, and biotinylated 248ox was added at the indicated concentrations and allowed to bind for 3 h at 37 °C. The wells were washed six times, and Europium-labeled streptavidin was added at a concentration of 500 ng/ml for 30 min at room temperature. Wells were washed six times, and Europium signal was measured. The data shown are the mean values (±S.D.) of a representative experiment done in triplicate.

Four other acidic residues were mutated, and each of these mutations showed an overwhelming loss of binding for 248ox to the mutant ₂ I-domain (Fig. 7B). These mutations include D292N, E256Q, D259N, and E299Q. Each of these residues lies at the surface of the ₂ I-domain in the vicinity of the MIDAS (Fig. 5, A and B).

	DISCUSSION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Integrin ₂₁ is a major receptor for collagens with collagen binding being mediated through the I-domain of the ₂ subunit (51-52). Human ₂₁ is also a receptor for the human pathogen echovirus-1, and like with collagen, the binding site for the virus has been localized to the ₂ I-domain (53) but to a site distinct from the collagen-binding site (48). The newly identified ₂ I-domain interacting RKKH peptides derived from the snake venom protein jararhagin block collagen binding but also modify the echovirus-1 interactions with the ₂ I-domain. While potently inhibiting the interaction between ₂ I-domain and collagen, the peptide also increases the binding of the virus to ₂ I-domain 10-fold (6). This suggests that the binding of the peptide to the ₂ I-domain blocks the collagen-binding site on the surface of the receptor either directly or through allosteric structural changes, while at the same time helping to expose or alter the virus-binding site. The exact locations of these binding sites have not been established. Here we provide evidence from experimental and computer-based approaches to show that the RKKH peptides bind in the vicinity of the MIDAS and that negatively charged residues make key contributions toward the binding interactions.

The Loop Conformation of the Peptide Is Essential for Binding-- With both the "Arg-Lys-Lys-His"-containing peptides and the metalloproteinase model structure, the side chains of the "Arg-Lys-Lys" sequence are completely accessible to solvent. The preceding threonine and the following histidine and aspartate are not. It seemed clear to us that if this region was responsible for blocking collagen binding to the ₂ I-domain, as shown experimentally for the peptide (6), then the major interactions between the peptide and the I-domain would be via the three positive charges that form a distinctive feature on the solvent-exposed surface of the metalloproteinase (Fig. 2), although these side chains provide substantial hydrophobic surfaces that can interact with the I-domain too. The loop conformation is important for binding, and linear peptides failed to block collagen binding to the I-domain (6). Molecular dynamics simulations suggested that a 6-residue peptide, TRKKHD, when cyclized via terminal cysteines and a disulfide bond, would optimally maintain the loop conformation observed in the metalloproteinase model. Consistent with this prediction, the peptide did block collagen binding with the highest affinity (Table I). Longer peptides demonstrated excessive flexibility and either functioned poorly or not at all (6). Peptides of 4 and 5 intervening residues were excessively strained in their cyclic forms in computer simulations, and experimentally the 5-residue peptide could not be cyclized by oxidation of the terminal cysteines. In each of the simulations of peptides with 6 or more residues, the histidine remained on the "inner" face of the loop as it does in the metalloproteinase crystal structures and our model of the jararhagin metalloproteinase domain. This histidine is conserved throughout most of the metalloproteinases. In the jararhagin metalloproteinase model structure, the histidine stabilizes the loop containing the "RKK" sequence by forming a hydrogen bond between the imidizole ring NH of histidine and the main chain carbonyl of the central Lys²⁴³ in the sequence (Fig. 2A). This role appears to be important for the peptide since alanine replacement of the histidine resulted in reduced binding of ₂ I-domain to the mutated peptide, although it exerted much less effect on the prevention of ₂ I-domain binding to type-I collagen (6).

Acidic Residues Surrounding MIDAS in the ₂ I-domain Are Key to the Mechanism of Action of the RKKH-containing Peptide-- The models of the metalloproteinase, the peptides derived from it, and the ₂ I-domain suggested that Asp²¹⁹, Glu²⁵⁶, Asp²⁵⁹, Asp²⁹², and Glu²⁹⁹ would be among the key residues likely to be involved in binding the RKKH peptides and thus the metalloproteinase itself. In the metalloproteinase-derived peptides, the main feature we focused on was the triangular array of three positively charged residues; as suggested by the metalloproteinase model, the histidine probably does not play a direct role in binding but serves instead to stabilize the conformation of the peptide or loop of the metalloproteinase. Five acidic residues were located on the surface of the ₂ I-domain model in the vicinity of the MIDAS, and the peptide, if bound to them, would overlap the surface of the metal-binding site. Metal was previously shown to be required for RKKH peptide binding (6).

A number of monoclonal antibodies shown to interact with the ₂ I-domain have been identified. Function blocking antibodies are considered to be allosteric regulators of ligand binding to the receptors, which means that their binding site does not necessarily overlap with that of the ligand. Nevertheless, they are considered to be valuable tools in the characterization of receptor-ligand interactions. For example, recent data from studies made with antibodies against _M I-domain showed for the first time that the I-domain itself undergoes conformational changes leading to an activated I-domain form (54). In the case of the ₂ I-domain, we found that two antibodies that interact with a region mapped to the surroundings of the MIDAS do affect peptide binding, supporting the idea that the binding site for the RKKH peptides and the metalloproteinase domain of jararhagin is located in that region of the structure. Given the evidence from our modeling studies, Asp²¹⁹, Glu²⁵⁶, Asp²⁵⁹, Asp²⁹², and Glu²⁹⁹ were subsequently mutated. In each case, the binding of the peptide 248ox was dramatically affected, strongly supporting our assertion that the peptide-binding site is located in the vicinity of the MIDAS and that these residues have key roles in recognition of the peptide and the snake venom metalloproteinase itself.

The MIDAS Touch Does Not Grasp an Asp in the Collagen-blocking Peptide-- It has been suggested (7-8) that the missing sixth ligand of the bound metal at the MIDAS of integrins would be an acidic residue. Within the crystal structure of the _M I-domain, the sixth position is indeed occupied by the side chain of glutamate of a second I-domain within the crystal structure (7). The metal dependence of ligand binding to I-domains is also well known, although other ligands can bind in a metal-independent fashion and presumably to a different site on the I-domain. For example, in ₂₁ integrins, echovirus-1 exhibits metal-independent binding to the ₂ I-domain and at a site distinct from the peptide-binding site. In the case of the jararhagin-derived peptides, metal was required for binding (6).

Originally, the presumption made by us (6) was that an acidic residue would also be required for binding in the case of a peptide derived from the metalloproteinase domain. Since evidence pointed to the metalloproteinase of jararhagin being involved in blocking collagen binding to the I-domain, peptides were identified that were likely to be surface loops containing either an aspartic acid or a glutamic acid. Indeed, peptide 229ox did have an aspartate; it is conserved throughout the snake venom metalloproteinases, but surprisingly, it was shown not to be required for binding. In our model, it is very likely that this highly conserved aspartate functions to bind a calcium ion as it does in the structures of the adamalysin II (30) and atrolysin C (36) metalloproteinases and is thus unavailable to bind to the metal and is normally shielded from the solvent. In the 229ox peptide, alanine replacement clearly showed that the aspartate was not necessary for binding (6).

How, then, can we explain the requirement for metal for the blocking of collagen binding to the I-domain? One possibility is that bound metal changes the conformation at the MIDAS significantly. Lee et al. (55) observed significant differences between the structures of Mn²⁺- and Mg²⁺-bound forms in the I-domain of _M, where the carboxyl-terminal helix moved by as much as 10 Å, and they suggested that the Mn²⁺ form represented the inactive receptor conformation, whereas the Mg²⁺ structure represented the active one. Qu and Leahy (56), however, found that the structure of the I-domain of _L, with and without Mn²⁺, was not much different from the Mg²⁺-bound form except in the region of the carboxyl-terminal helix. More recently, Baldwin et al. (57) found no significant changes in different _M metal complexes, but Li et al. (58) have trapped different conformation states by mutations to _M. In the case of the ₂ I-domain, there is no direct evidence that changes do occur in the presence of bound metal, although addition of peptide 229ox resulted in a 10-fold increase in the affinity of echovirus-1 for the I-domain at a site distinct from the MIDAS, suggesting that alterations in the I-domain structure do occur (6).

A second possibility is that the peptide does fulfill the sixth ligand position of the MIDAS, not through the direct participation of any side chain but through the interaction of main chain carbonyl oxygens, possibly via a bound water molecule lying between the peptide and the metal of the MIDAS. In the Mn²⁺-bound structure of _M, Lee et al. (55) did observe that a molecule of water acted as the sixth ligand to the metal, instead of a glutamate from a second I-domain as seen in the Mg²⁺-bound crystal structure (7). The third possibility is that metal both alters the conformation of the binding site and interacts with the peptide with the only suitable ligand group: main chain carbonyl oxygens of the peptide via an intervening water molecule. A structure of the complex between the ₂ I-domain and the peptide will in the future clarify these issues.

In the absence of a crystal structure of the complex, we have analyzed the surface features near the MIDAS (Fig. 8) and used this information in combination with docking simulations and the experimental mutagenesis results to dock the peptide and the metalloproteinase model structure (Fig. 9). The GRID map in Fig. 8 displays the locations on the surface of the ₂ I-domain where different types of atoms or atomic groups would ideally interact surface-exposed groups. Note the large hydrophobic interaction surface (white surface over the MIDAS: sp³ hydrocarbon interactions) whose outline matches well the surface features of the peptide in the model of the metalloproteinase (Fig. 2B). Furthermore, at the ends of these hydrophobic regions are located "blue" interaction surfaces that represent good positions for positively charged nitrogens (Fig. 8). Although we made an extensive conformational search of the 248ox peptide, using both rigid and flexible representations of the peptide and I-domain, the docking of the peptide structure proved to be very difficult, since the peptide side chains are very flexible (over 20 rotatable torsion angles). One of many different docking proposals about the MIDAS is shown in Fig. 9A. In Fig. 9B, we show the docking result made manually using the result from Fig. 9A and incorporating the details on ideal interactions from Fig. 8 and other experimental details. One possible docking interaction between the ₂ I-domain structure and the metalloproteinase model is shown in Fig. 9C.

View larger version (92K): [in this window]