Binding Investigation of Integrin α v β 3 With Its Inhibitors by SPR Technology and Molecular Docking Simulation

Abstract

Integrins play critical roles in the process of angiogenesis and are attractive targets for anticancer therapies. It is desirable to develop new types of small-molecule inhibitors of integrin. Herein, the binding features of several inhibitors to integrin α_vβ₃ have been studied by surface plasmon resonance (SPR) biosensor technology and molecular docking analyses. The SPR results indicated that the equilibrium dissociation constant (K_D) values are evaluated for the inhibitors and showed that the K_D value of cyclopeptide c-Lys is much lower than the reference molecule. In addition, the 3D structural model of integrin α_vβ₃ was generated according to the crystal structure of the integrin α_vβ₃ complex, and the molecular docking simulation analyses revealed that the predicted binding sites for the most active cyclopeptide c-Lys were consistent with the reported structure. These results thus implied that cyclopeptide c-Lys could be developed as a novel inhibitor for integrin α_vβ₃. The current work has potential for application in structure-based integrin α_vβ₃ inhibitor discovery.

Keywords

integrin inhibitors RGD motif SPR molecular docking simulation

Introduction

Integrins are heterodimeric cell surface receptors that consist of 2 noncovalently associated alpha and beta subunits and function as receptors for extracellular matrix (ECM) protein ligands such as fibronectin, laminins, collagens, and vitronectin.¹ Integrins play a key role in the process of angiogenesis and are expressed in a variety of malignancies, including melanoma, breast cancer, prostate cancer, colon cancer, and gliomas.^2,3 The integrins interact with the ECM proteins through the arginine-glycine-aspartic acid sequence (RGD).⁴ During the past decade, the universal recognition site (RGD) became the basis for the design of a variety of RGD-containing peptides and RGD peptidomimetics^5-7 and has been used as a lead structure in the development of α_vβ₃ inhibitors.⁸ Few studies have presented the direct binding data of the inhibitors to integrin α_vβ₃ and details of the binding modes of their interactions, which is a major constituent of structure-based integrin α_vβ₃ inhibitor discovery.

Recently, the surface plasmon resonance (SPR) biosensor has been recognized as a valuable tool in monitoring biological interactions with advantages of being label-free, real-time, noninvasive measurements; having low sample consumption; and being widely used in hit determination.⁹ In our study, the SPR technology was successfully applied to determine the binding affinities of several cyclopeptides to integrin α_vβ₃. The equilibrium dissociation constant (K_D) values evaluated by the Biacore X100 for the inhibitors demonstrated that the K_D value of cyclopeptide c-Lys is much lower than that of the reference inhibitor.

To further investigate the inhibitor/integrin α_vβ₃ binding mode at the atomic level, the LigandFit method of molecular docking was employed to dock the newly designed molecules in the active site of integrin α_vβ₃. Molecular docking has become a widely used technology in drug design to identify and optimize lead compounds.^10,11 During the past few years, some molecular docking algorithms based on different search methods have been developed and implemented.^10,12 Molecular docking involves the prediction of a ligand (small-molecule) conformation and orientation within the active site of the molecular target. The molecular docking program LigandFit,¹³ a shape-based method for docking ligands into protein active sites, was used to perform the virtual screening. The aim was to build models able to (1) distinguish active from inactive compounds for integrin receptor α_vβ₃, (2) predict binding affinity for active compounds, and (3) find selective compounds. In the current work, molecular docking was performed with the 3D structural model of integrin α_vβ₃ generated on the basis of the crystal structure of the integrin α_vβ₃-ligand complex (PDB entry code 1L5G).¹⁴ The predicted binding sites for the most active cyclopeptide c-Lys were consistent with the reference structure, but the inactive cyclopeptide c-Leu was not, which suggests that cyclopeptide c-Lys could be developed as a potential inhibitor for integrin α_vβ₃.

In our previous work, a new family of compounds (17 cyclopeptides) was prepared according to the published procedure,¹⁵ as shown in Figure 1 . In the current work, the binding features of several inhibitors to integrin α_vβ₃ have been studied by SPR technology and molecular docking simulation.

Fig. 1.

Structure of cyclo[-Arg-Gly-Asp-ψ(triazole)-Gly-Xaa-].

Materials and Methods

Materials

Protein integrin α_vβ₃ was purchased from Millipore (Billerica, MA). N-octyl-β-D-glucoside was purchased from Dojindo Laboratories (Kunamoto, Japan). The cyclopeptides cyclo[-Arg-Gly-Asp-ψ(triazole)-Gly-Xaa-] (c-Xaa) ligands for integrin α_vβ₃ receptor as previously described¹⁵ were synthesized for surface plasmon resonance experiments. N-hydroxysuccinimide (NHS) and 1-ethyl-3-(3-dimethyl aminopropyl) carbodiimide hydrochloride (EDC) were obtained from JK Chemical (Beijing, China). All other chemicals were of reagent grade or ultrapure quality and purchased from Shanghai Sangon Biological Engineering & Technology and Service Co. Ltd. (Shanghai, China).

Methods

SPR technology-based binding assay

To inspect the possible active compounds with binding affinities to integrin α_vβ₃, SPR technology-based Biacore X100 biosensor was used (GE Healthcare, Piscataway, NJ). The SPR experiment was performed according to the manufacturer’s instruction. In brief, integrin α_vβ₃ was immobilized on a CM5 chip (Biacore, Piscataway, NJ) by the amine-coupling method. The dextran surface was activated by injection of NHS/EDC mixture (0.1M NHS and 0.4M EDC, 1:1 mixture) for 7 min. Integrin (in 10 mM acetate, pH 4.0) was injected over activated surface. Finally, unreacted activated groups were blocked by a 35-µL injection of ethanolamine. All the binding experiments were performed at 25°C at a continuous flow rate with 10 mM HEPES buffered saline (10 mM HEPES, 150 mM NaCl, 1 mM MgCl₂, 1 mM MnCl₂, and 5 mM n-octyl-β-D-glucoside, pH 7.4). A few sets of experiments, using cyclopeptides as analytes, were run on the chip surface at the concentration of 0, 3.2, 6.3, 12.5, 25, 50, and 100 µM. The cyclopeptide cRGDfK, as a positive control, which displayed a high binding affinity to integrin α_vβ₃, was reported previously.¹⁶ The sensor surfaces were regenerated with a short pulse of 10 mM glycine, pH 2.2.

All the equilibrium dissociation constants (K_Ds) used for evaluating the ligand-integrin α_vβ₃ binding affinity were determined with Biacore evaluation Version 4.1 software (Biacore). During the assays, the signal was corrected against the control surface response to eliminate refractive index changes due to buffer change.

SPR data analysis

During the assays of integrin α_vβ₃ and its inhibitors with Biacore X100, the binding response signals in RUs were continuously recorded graphically as a function of time. The data were fit using a homogeneous single-site interaction model, A + B = AB, where A represents the injected analyte, B is the immobilized ligand, and AB is the complex formed during the reaction. The Biacore results indicated that, in reaching equilibrium, both the association with and dissociation from the immobilized integrin α_vβ₃ of the analytes involved slow phases, as shown in Figure 2 (using c-Lys as a typical example). Therefore, the data were analyzed by global fitting to a 1:1 binding model (Langmuir) of both the association and dissociation phases for several concentrations simultaneously, using the BIAevaluation 4.1 software.¹⁷ The curve-fitting efficiency was evaluated by chi-square. The chi-square value denotes the fitting degree between the estimative and the experimental curves.

Fig. 2.

Kinetic analysis of the cyclopeptide c-Lys binding to integrin α_vβ₃ by surface plasmon resonance (SPR) technology. Representative sensorgram obtained from injection of cyclopeptide c-Lys at concentrations of 0, 3.2, 6.3, 12.5, 25, 50, and 100 µM.

Molecular modeling systems

The X-ray structure of integrin α_vβ₃ was obtained from the PDB (1L5G, resolution 3.20 Å).¹⁴ The structure of α_vβ₃, a heterodimeric protein (183 kDa) of 1649 amino acids, consists of noncovalently linked alpha V subunit polypeptide (CD51) and beta 3 subunit polypeptide (CD61). The 3D-optimized structures of cyclopeptides were generated by the program ChemSketch/3D Viewer (Version 11.0, from Advanced Chemistry Developments, Toronto, Canada; http://www.acdlabs.com). LigandFit was used for molecular modeling on a Linux workstation. The resultant structures were then analyzed using the LigPlot 4.22¹⁸ program to identify specific contacts between ligands and the protein. The visualization of complexes was employed by the Pymol 0.99 program¹⁹ on a Windows XP workstation.

LigandFit docking procedure

LigandFit, a docking tool, was used to dock ligands in the protein binding site, calculate scores, and rank them accordingly (http://www.accelrys.com/). The docking site was defined using the LigandFit site search utility. Two possibilities for binding site definition were offered by LigandFit: cavities (i.e., areas of unoccupied grid points) can be detected based on the shape of the protein or from a bound ligand.²⁰ The former strategy has to be applied when only the apo (ligand-free) structure of the protein is known, which involves more uncertainty of the obtained results, but the latter strategy is more reliable and the technique of choice for holo (ligand-bound) protein structures. In our case, the binding site was generated from the X-ray pose of the bound ligand. The grid resolution was set to 0.5 Å, and the ligand-accessible grid was defined such that the minimum distance between a grid point and the protein was 2.0 Å for hydrogen and 2.5 Å for heavy atoms. Such a general binding site was used to calculate the nonbonded interactions between all atoms of each ligand and the protein residues. If necessary, the initial binding site cavity was expanded with the program expansion tool to a distance of 3 Å in all directions and then edited manually to remove points protruding far outside the molecular surface. After characterization of the binding pockets, a Monte Carlo method was employed for the conformational search of the ligand. Ligands were docked within the identified site with 100 possible conformations, with each ranked according to the dock score provided in LigandFit. Dock scoring was performed with LigScore1 scoring function.¹³

Results and Discussion

SPR-based binding assays between cyclopeptides and the integrin

The interactions of the cyclopeptides were investigated with the immobilized integrin α_vβ₃ on the CM5 sensor chip at a level of 11,000 response units (RU) with the standard amine coupling procedure. Subsequently, the chip was used to measure binding parameters for the interaction with different cyclopeptides (analytes) at concentrations ranging from 3.2 to 100 µM. As shown in Figure 2 , the RU values evaluating the cyclopeptide c-Lys binding to the integrin α_vβ₃ revealed an obvious concentration-dependent manner, which demonstrated that the determined equilibrium dissociation constant (K_D) of 1.3 × 10⁻⁹ M is much lower than the K_D value of 1.3 × 10⁻⁶ M for the reference molecule cRGDfK.¹⁶ The binding kinetic constant was fitted using the 1:1 binding fitting model (Langmuir) supplied in the Biacore software. Using this model, the chi-squared value of c-Lys was calculated to be 0.79, indicating a good fit.

Shown in Table 1 are the K_D values of all the tested inhibitors binding to integrin α_vβ₃ from the SPR assay. In addition to the c-Lys mentioned above, for c-Gly, it is nearly within the same quantitative grade (2.4 × 10⁻⁶ M) compared with the reference molecule, and for other cyclopeptides, their K_D values (>10⁻⁶ M) are higher. The chi-squared value was used to judge the quality of fit between experimental data and that derived from binding models. The smaller the chi-square, the smaller their variation. Typically, good fittings derive chi-square values less than 10.²¹ All the results are summarized in Table 1 . To sum up, we compared the K_D values of all compounds obtained from the SPR assay, and the K_D value of c-Lys was much lower than the reference molecule.

Table 1.

Equilibrium Dissociation Constant (K_D) Values of Cyclopeptides Determined by the SPR Assay

Entry	Compound	MW, Da	K_D, M	χ²
1	c-Lys	594.63	1.3 × 10⁻⁹	0.79
2	c-Gly	523.51	2.4 × 10⁻⁶	2.43
3	cRGDfK^a	608.72	1.3 × 10⁻⁶	1.37
4	c-Pro	563.57	3.5 × 10⁻⁵	1.45
5	c-Gly-Gly^b	580.56	1.7 × 10⁻⁴	2.32
6	c-Glu	595.57	2.5 × 10⁻⁴	1.68
7	c-Asn	580.56	3.3 × 10⁻⁴	5.49
8	c-Trp	652.67	4.5 × 10⁻⁴	6.13
9	c-Cys	569.59	6.7 × 10⁻⁴	2.32
10	c-Phe	613.63	8.4 × 10⁻⁴	3.24
11	c-Met	597.65	2.4 × 10⁻³	1.77
12	c-Ile	579.61	3.1 × 10⁻³	3.25
13	c-Asp	581.54	4.7 × 10⁻³	2.31
14	c-0^c	466.46	5.5 × 10⁻³	3.52
15	c-His	603.6	6.1 × 10⁻³	1.87
16	c-Ala	537.53	6.5 × 10⁻³	3.27
17	c-Arg	622.64	7.3 × 10⁻³	6.83
18	c-Leu	565.59	8.2 × 10⁻³	4.39

cRGDfK was used as reference in the receptor binding analysis.

Cyclo[-Arg-Gly-Asp-ψ(triazole)-Gly-Gly-Gly-].

Cyclo[-Arg-Gly-Asp-ψ(triazole)-Gly-].

Model evaluation of LigandFit docking

To explore the binding characteristics of the inhibitors to integrin α_vβ₃ at the molecular level, molecular docking was applied to construct the inhibitor/integrin α_vβ₃ binding model. The LigandFit program was employed to predict the potential binding of cyclopeptides to integrin α_vβ₃ based on the cavity detection algorithm. In our case, the binding site was generated from the X-ray pose of the bound ligand. Previous crystallographic and modeling analysis showed that cRGDfNMeV interacts with integrin α_vβ₃ mainly using the R-G-D moiety.¹⁴ Evaluation of the site definition was achieved by superimposition of the docked and the X-ray conformations and orientations of cRGDfNMeV, as shown in Figure 3 . The root mean square deviation (RMSD) of the best docked structure from the X-ray form was 0.234 Å. This promising result confirmed the applicability of our binding site model for further studies.

Fig. 3.

Superimposition of the docked conformation (red) with the X-ray conformation (green) of the ligand cRGDf NMeV (PDB entry 1L5G).

Analysis of docking complexes

The docked conformations were scored using different scoring functions (LigScore1, LigScore2, PLP1, PLP2, JAIN, and PMF). Searching the lists of scoring functions for the cyclopeptides showed that the actives obtained highest positions, when ranked according to their LigScore1 values. Hence, it can be stated that the good performance of LigScore1 justifies the use of this scoring function to rank the results of a LigandFit docking process for this system. Three descriptors are used to calculate the LigScore (a program-internal scoring function): a softened Lennard-Jones potential, 1 count of the buried polar surface area between a protein and ligand that involves attractive protein-ligand interactions, and the sum of squared attractive and repulsive protein-ligand contact terms. The 3-descriptor system that could be found is equation (1):

\begin{matrix} F (LigScore 1) = C - β_{1} ⟨ E_{vdW} ⟩ + β_{2} C pos_tot \\ - β_{3} (SolvPlty_{lig}^{2} + SolvPlty_{prot}^{2}) \end{matrix}

where E_vdW indicates the van der Waals (vdW) energy of protein-ligand interactions, which is computed via a Lennard-Jones potential¹³ using van der Waals radii and energy parameters of either the consistent force field (CFF) or the DREIDING force field. Cpos_tot is the total surface area of the ligand involved in attractive polar interactions with the protein. The SolvPlty_lig² term represents the overall polar ligand surface that is buried inside the protein binding site, and the SolvPlty_prot² term indirectly accounts for the polar protein surface in contact with the bound ligand molecule. Therefore, in the case of cyclopeptide/integrin α_vβ₃ complexes, the contributions of 3 key descriptors (vdW and 2 surface descriptors) may be identified effectively.

Take the example of c-Lys, the most active compound in SPR assay. Molecular docking results thus describe the general binding mode of the inhibitor c-Lys to integrin α_vβ₃. The hydrophobic and hydrogen bond interactions between integrin α_vβ₃ and the various cyclopeptide ligands were visualized by employing the program LigPlot. The schematic binding mode analysis of c-Lys/integrin α_vβ₃ complex shows the residues involved in the receptor site shown in Figure 4 . On the whole, c-Lys has possible strong hydrogen bonds with 9 amino acid residues—Asp150(A), Tyr178(A), Asp218(A), Asp219(A), Ser123(B), Asn 215(B), Ala218(B), Glu220(B), and Arg241(B)—of the α_vβ₃ receptor, as judged by the LigPlot program. Four residues of them—Asp150(A), Tyr178(A), Asp218(A), and Asn215(B)—are in good agreement with those usually observed for other complexes in the literature.¹⁴ Thus, new hydrogen bonds have been found in the complex. The presence of additional hydrogen bonds can stabilize the complex structure but could contribute to the R-G-D moiety conformation. Moreover, analysis with the LigPlot program suggests that c-Lys can form 3 hydrophobic interactions with Tyr178(A), Ala218(B), and Tyr122(B) of the α_vβ₃ receptor. The docking conformation of c-Lys around the receptor binding domain of integrin α_vβ₃ is depicted in Figure 5 . The Asp residue adopted a U-shape with the 2 Arg residues in the docking conformation of c-Lys. However, there is a strong salt bridge between the Asp carboxylate oxygens of c-Lys and the Mn4001(B) ion, which is also involved in hydrogen bond interactions with the hydroxyl group of Ser123(B). Overall, the above results indicate that c-Lys has strong interactions and thus binding with integrin α_vβ₃.

Fig. 4.

A schematic illustration of the interactions of c-Lys with the residues around the receptor site of integrin α_vβ₃. Integrin α_vβ₃ residues and corresponding atoms of c-Lys involved in hydrogen bond and hydrophobic contacts are drawn using the LigPlot program. Dashed lines represent hydrogen bonds, and spiked residues form hydrophobic contacts with the c-Lys inhibitors.

Fig. 5.

The docking conformation of c-Lys into the active site of the integrin α_vβ₃ receptor. The ligand, c-Lys, is shown in stick and the important residues around the receptor site surface are shown in line. The α_v and β₃ residues are labeled red and black, respectively. Hydrogen bond interactions are represented by dashed lines. The c-Lys is accommodated within the binding pocket.

Similar to the case of c-Lys/integrin α_vβ₃ complex studies, the binding mode of the c-Leu/integrin α_vβ₃ complex was obtained, which is the most inactive compound in the SPR assay. As shown in Figure 6 , in comparison with the c-Lys/integrin α_vβ₃ complex, the hydrophobic and hydrogen bond interactions of the c-Leu/integrin α_vβ₃ complex were less prominent, showing a weak hydrogen bond with only 3 residues—Asp218(A), Arg214(B), and Ser123(B)—of integrin α_vβ₃. Figure 7 shows that the conformation of c-Leu failed to dock into the integrin α_vβ₃ receptor site, in which the Asp residue of the RGD motif in c-Leu is pushed further away from the receptor site than it does in c-Lys. Thus, the RGD motif of c-Leu may have a preferred inactive conformation, depending on the effect of its leucyl side chain. As has been mentioned, the molecular docking results have indicated that the c-Lys/integrin α_vβ₃ complex has a strong multiple hydrogen bond and hydrophobic interactions, whereas the c-Leu/integrin α_vβ₃ complex only shows weak interactions.

Fig. 6.

A schematic illustration of the interactions of c-Leu with the residues around the receptor site of integrin α_vβ₃.

Fig. 7.

The docking conformation of c-Leu to the active site of integrin α_vβ₃ receptor. The ligand, c-Leu, failed to bind to the binding pocket domain of the integrin α_vβ₃, in which the Asp residue of the RGD motif flips out of the binding grooves as a whole. The c-Leu is protruded outside the binding pocket.

Conclusions

In summary, the current work employed SPR technology and molecular docking simulation analyses to study the binding features of several inhibitors to integrin α_vβ₃. The equilibrium dissociation constant (K_D) values evaluated by Biacore X100 for the inhibitor binding to integrin α_vβ₃ showed that the K_D value of cyclopeptide c-Lys is much lower than the reference molecule. In addition, molecular docking technology was performed with the 3D structural model of integrin α_vβ₃ generated on the basis of the crystal structure of the integrin α_vβ₃ complex, which has helped us to investigate the inhibitor/integrin α_vβ₃ binding at the atomic level. The predicted binding sites for the most active cyclopeptide c-Lys were consistent with the reported structure. These results suggested that cyclopeptide c-Lys could be developed as a potential inhibitor for integrin α_vβ₃. It is hoped that our work will find application in the platform construction for structure-based integrin α_vβ₃ inhibitor discovery.

Footnotes

Acknowledgements

We are grateful to Prof. Yongzhou Hu for providing the LigandFit package and to Dr. Wenhai Huang and Xiaowu Dong for providing help during the LigandFit docking procedure.

References

Ruegg

Postigo

Sikorski

Butcher

Pytela

Erle

: Role of integrin α₄β₇/α₄β_P in lymphocyte adherence to fibronectin and VCAM-1 and in homotypic cell clustering. J Cell Biol 1992;117:179-189.

Max

Gerritsen

Nooigen

Goodman

Sutter

Keilholz

: Immunohistochemical analysis of integrin α_vβ₃ expression on tumor-associated vessels of human carcinomas. Int J Cancer 1998;71:320-324.

Abdollahi

Griggs

Zieher

Roth

Lipson

Saffrich

: Inhibition of α_vβ₃ integrin survival signaling enhances antiangiogenic and antitumor effects of radiotherapy. Clin Cancer Res 2005;11:6270-6279.

Smith

Cheresh

: The Arg-Gly-Asp binding domain of the vitronectin receptor: photoaffinity cross-linking implicates amino acid residues 61-203 of the beta subunit. J Biol Chem 1988;263:18726-18731.

Haubner

Gratias

Diefenbach

Goodman

Jonczyk

Kessler

: Structural and functional aspects of RGD-containing cyclic pentapeptides as highly potent and selective integrin αvβ₃v3 antagonists. J Am Chem Soc 1996;118:7461-7472.

Wermuth

Goodman

Jonczyk

Kessler

: Stereoisomerism and biological activity of the selective and superactive α_vβ₃ integrin inhibitor cyclo--RGDfV-- and its retro-inverso peptide. J Am Chem Soc 1997;119:1328-1335.

Lohof

Planker

Mang

Burkhart

Dechantsreiter

Haubner

: Carbohydrate derivatives for use in drug design: cyclic α_v-selective RGD peptides. Angew Chem Int Ed Engl 2000;39:2761-2764.

Haubner

Finsinger

Kessler

: Stereoisomeric peptide libraries and peptidomimetics for designing selective inhibitors of the α_vβ₃ integrin for a new cancer therapy. Angew Chem Int Ed Engl 1997;36:1374-1389.

Eglen

Pribilla

: High-throughput screening using label-free technologies. J Biomol Screen 2004;9:465-466.

10.

Shoichet

McGovern

Wei

Irwin

: Lead discovery using molecular docking. Curr Opin Chem Biol 2002;6:439-446.

11.

Jorgensen

: The many roles of computation in drug discovery. Science 2004;303:1813-1818.

12.

Leach

Shoichet

Peishoff

: Prediction of protein-ligand interactions: docking and scoring: successes and gaps. J Med Chem 2006;49:5851-5855.

13.

Venkatachalam

Jiang

Oldfield

Waldman

: LigandFit: a novel method for the shape-directed rapid docking of ligands to protein active sites. J Mol Graph Model 2003;21:289-307.

14.

Xiong

Stehle

Zhang

Joachimiak

Frech

Goodman

: Crystal structure of the extracellular segment of integrin α_vβ₃ in complex with an Arg-Gly-Asp ligand. Science 2002;296:151-155.

15.

Liu

Zhang

Wan

Pan

: Design and synthesis of cyclo[-Arg-Gly-Asp-ψ-triazole--Gly-Xaa-] peptide analogues by click chemistry. Tetrahedron 2008;64:10728-10734.

16.

Pfaff

Tangemann

Müller

Gurrath

Müller

Kessler

: Selective recognition of cyclic RGD peptides of NMR defined conformation by alpha IIb beta 3, alpha V beta 3, and alpha 5 beta 1 integrins. J Biol Chem 1994;32:20233-20238.

17.

Rich

Day

Morton

Myszka

: High-resolution and high-throughput protocols for measuring drug/human serum albumin interactions using BIACORE. Anal Biochem 2001;2:197-207.

18.

Wallace

Laskowski

Thornton

: LIGPLOT: a program to generate schematic diagrams of protein-ligand interactions. Protein Eng 1995;8:127-134.

19.

DeLano

: The PyMOL molecular graphics system. DeLano Scientific 2002. Retrieved from http://www.pymol.org

20.

Steindl

Langer

: Docking versus pharmacophore model generation: a comparison of high throughput virtual screening strategies for the search of human rhinovirus coat protein inhibitors. QSAR Comb Sci 2005;24:470-479.

21.

Baglia

Badellino

Dasari

Walsh

: A binding site for the kringle II domain of prothrombin in the apple 1 domain of factor XI. J Biol Chem 2000;41:31954-31962.