PROTEINS Structure, Function, and Genetics 26353-357 (1996) CRYSTALLIZATION ANNOUNCEMENTS Crystallization of the Receptor Binding Domain of Vascular Endothelial Growth Factor Hans W.Christinger,' Yves A. Muller,' Lea T. Berleau? Bruce A. Keyt? Brian C. Cunningham,' Napoleone Ferrara? and Abraham M.de Vos' Departments of 'Protein Engineering and 2Cardwvascular Research, Genentech, Inc., South Sun Francisco, California 94080 ABSTRACT Vascular endothelial growth factor (VEGF) is a potent angiogenic factor with a unique specificity for vascular endothelial cells. In addition to its role in vasculogenesis and embryonic angiogenesis, VEGF is implicated in pathologic neovascularization associated with tumors and diabetic retinopathy. Four different constructs of a short variant of VEGF sufficient for receptor binding were overexpressed in Escherichia coli, refolded, purified, and crystallized in five different space groups. In order to facilitate the production of heavy atom derivatives, single cysteine mutants were designed based on the crystal structure of platelet-derived growth factor. A construct consisting of residues 8 to 109 was crystallized in space group P2,,with cell parameters a = 55.6 A, b = 60.4 A, c = 77.7 A, p = 90.0", and four monomers in the asymmetric unit. Native and derivative data were collected for two of the cysteine mutants as well as for wild-type VEGF. 0 1996 Wiley-Liss, Inc. Key words: VEGF, angiogenesis, tumor vascularization, inclusion bodies, cysteine mutants, X-ray crystallography, crystals INTRODUCTION Angiogenesis (the development of new blood vessels) is required for many physiological processes such as embryogenesis, wound healing, and corpus luteum formation. Vascular endothelial growth factor (VEGF), also known as vascular permeability factor (VPF), is a highly endothelial cell-specific angiogenic factor and has been implicated in the neovascularization of Although several other potential mediators of tumor vascularization, including basic and acidic fibroblast growth factors, tumor necrosis factor-a, and transforming growth factors-a and -p, have been identified in situ, these factors lack specificity for endothelial cells. The ap0 1996 WILEY-LISS, INC. parent involvement of VEGF in diabetic retinopathy and other retinal neovascular disorder^,^ as well as the inhibition of tumor growth in vivo by antibodies directed against VEGF,' make it a prime candidate as the mediator of normal and pathological angiogenesis. Therefore, VEGF antagonists will potentially be useful therapeutic agents to regulate excessive vascularization. VEGF is a homodimeric glycoprotein which naturally occurs in four different isoforms, each monomer having 121, 165, 189, or 206 amino acids (VEGF-121, VEGF-165, VEGF-189, and VEGF206). The different isoforms result from alternative C-terminal splicing and whereas VEGF-121 lacks affinity for heparin, all other isoforms are highly basic proteins tightly bound to extracellular heparin-containing prote~glycans.~ The absence of the heparin binding domain does not reduce the binding affinity of VEGF to its receptors KDR and FLT, but does decrease the activity of VEGF in a cell-based assay, suggesting that the role of the heparin binding domain is to sequester soluble VEGF to the cell surface.* A shorter isoform of VEGF retaining full receptor binding affinity can be obtained in vivo upon plasmin cleavage after residue 110 (VEGF110).8 Sequence homology suggests that the receptorbinding domain of VEGF (residues 1 to 110) is a member of the cystine-knot fa mil^.^*'^ This family is characterized by three disulfide bridges forming a knot-motif. Other members of this family with known three-dimensional structure are nerve growth factor," transforming growth f a ~ t o r - p , ' ~ , ~ ~ platelet-derived growth factor (PDGF),14and chorionic g~nadotropin.'~,~' These molecules all form dimers, but even though the monomer units share a conserved fold, the dimer interfaces are completely Received March 29,1996; revision accepted April 5, 1996. Address reprint requests to Abraham M. de Vos, Department of Protein Engineering, Genentech, Inc., 460 Point San Bruno Boulevard, South San Francisco, CA 94080. 354 H.W. CHRISTINGER ET AL. different? With the exception of the cysteine residues, VEGF has very little sequence homology to the other members of the family; the connectivity of intermolecular disulfide bridges suggests that VEGF is most closely related to PDGF.' Due to the central role of VEGF in angiogenesis and in order to better understand its biological activity on a molecular level, we decided to determine the crystal structure of the receptor binding domain of VEGF. Here we report the purification, refolding, and crystallization of four different VEGF variants (VEGF-8~109,VEGF-110, VEGF-121, VEGF-1651, undertaken because of our initial failure to obtain suitable crystals for some of the variants. In addition, when production of heavy atom derivatives using conventional soaking methods proved unsuccessful, we designed mutants each with a single residue replaced by cysteine. The success of these approaches demonstrates the power of the application of molecular biology methods in support of structural studies. MATERIALS AND METHODS Two isoforms of VEGF (VEGF-121 and VEGF165), an artificial construct consisting of residues 8 to 109 (VEGF-8:109), and two single cysteine mutants (P53C and Q98C) of the latter were expressed as insoluble proteins in E. coli in amounts of 60 to 80 mg per 100 g of cell paste. Inclusion bodies were isolated by passing homogenized cells in 20 mM TRIS pH 7.5, 5 mM EDTA through a French pressure cell, and centrifuging the homogenate for 15 minutes at 4,OOOg. The inclusion bodies were then resuspended and centrifugation repeated. The pellet consisting predominately of VEGF was dissolved in 7.5 M urea, 20 mM TRIS pH 7.5, and stirred for 1 hour in 20 mM dithiothreitol (DTT) in order to achieve complete reduction of incorrectly formed disulfide bridges. VEGF-121, VEGF-165, and VEGF8:109 were refolded by diluting the above solution to a final concentration of about 0.75 mg/ml and then dialyzing for 24 hours against 20 mM TRIS-HCl (pH 8.4), 0.4 M sodium chloride, 1 mM cysteine. Identical results were obtained with 7 pM CuC1, in the refolding buffer. The single cysteine mutants were refolded in the absence of free cysteine and CUCl,. The refolded proteins VEGF-121 and VEGF-8:109 were purified to homogeneity in consecutive steps of ion exchange chromatography (Q-Sepharose; Pharmacia, Uppsala, Sweden), hydrophobic interaction chromatography (Alkyl-Sepharose; Pharmacia) and size exclusion chromatography (S-100; Pharmacia). In order to protect the free cysteines in the cysteine point mutations of the VEGF-8:109 variant, the first two purification steps were performed in the presence of 2 mM D'I"I'. The highly basic protein VEGF165 was purified by ion exchange chromatography (S-Sepharose, Pharmacia) followed by a copper chelating column and a second ion exchange chromatography column (SB-Toyo Pearl, Toso Haas, Philadelphia, PA). Portions of the protein were then subjected to plasmin cleavage in order to generate VEGF-110.' VEGF-110 was purified by negative absorption to a heparin affinity column followed by ion exchange chromatography (Q-Sepharose; Pharmacia). The purity and homogeneity of all proteins were monitored by SDS-PAGE and mass spectrometry. In addition, receptor-binding activity to the extracellular domain of the KDR receptor was tested with IgG-KDR fusion proteins in a radio receptor assay. Initial screening for crystallization conditions was done according to the sparse matrix method17 using commercially available buffers (Hampton Research, Riverside, CA), suspending droplets obtained after mixing 2 p1 of protein solution [lo-15 mg/ml protein in crystallization buffer (20 mM TRIS-HCl, pH 7.5, 0.4 M sodium chloride)] with 2 pl of reservoir solution over 900 pl of reservoir solution in cell culture plates from Linbro (Flow Laboratories, McLean, VA). After optimization of the initial crystallization conditions, large single crystals were grown in sitting drops consisting of 40 pL protein mixed with 40 pL reservoir solution in nine-well sitting-drop trays from Corning (Corning, NY). All crystallographic data were recorded in-house using graphite-monochromated CuKa radiation generated from a Rigaku RU200 rotating anode (Rigaku-USA, Danvers, MA) operating at 45V, 12OmA, and a MAR Research imaging plate (MAR Research, Hamburg, Germany) with a 300 mm diameter. Usually data were recorded in 1" rotation steps with an exposure time of 15 minutes and indexed using the automatic indexing routines of programs XDS1' and DENZO." RESULTS AND DISCUSSION Design of Cysteine Mutants When conventional attempts at producing derivatives were unsuccessful, single cysteine mutations were introduced into VEGF. The design of the mutations was achieved by displaying the sequence of VEGF onto the backbone of the existing PDGF structure14 and looking for partially exposed side chains in well-ordered portions of the model. In order to prevent misfolding of the cystine-knot disulfide bridges, sites chosen were distant in space from existing cysteine residues. No restrictions were imposed on the amino acid type a t the residues considered for mutation. We selected a final set of four potential mutation sites (Pro53, Thr71, Leu97, Gln98) distributed evenly over the molecule (Fig. 1); two of the mutants were crystallized, but in practice, a single mutant (P53C) was sufficient to determine the structure. CRYSTALLIZATION OF VEGF 355 Fig. 1. Backbone (in gray) and disulfide bonds (in green) of the PDGF dimer, showing VEGF side chains for residues selected for substitution with cysteine in yellow. Residues on one monomer are labeled using VEGF numbering. The Cys53 and Cys98 point mutants were expressed, refolded and purified, and crystallized isomorphously with wild-type VEGF. Purification and Refolding Mass spectrometry confirmed the correct mass and homogeneity of the purified variants. All constructs showed similar receptor binding activity (Bing Li, personal communication). The final yield of purified wild-type VEGF was about 30 to 40 mg per 100 g of cell paste (approximately 50%); some of the losses represented VEGF running anomalously on the hydrophobic interaction chromatography column, and we assume this was misfolded protein. Initial purification and refolding attempts for the single cysteine mutants using the refolding protocol developed for the wild-type variants failed, as in all cases the determined mass differed from the expected mass. The mass difference could be accounted for by cysteine molecules from the refolding buffer bound to the free cysteines of the mutants. In other attempts, observed mass differences suggested partial oxidation by molecular oxygen of the free sulfhydryl groups of the mutants. Only after using DTT in the refolding buffer and maintaining reducing conditions during the first two purification steps, could the correct masses be observed for the single cysteine substitutions. After re-optimization of the refolding and purification protocols, the yield of the mutants increased from 10 to 20 mg per 100 g of paste to the levels observed for wild-type VEGF. Dithionitrobenzoate assays showed two free sulfhydry1 groups per VEGF dimer. Crystallizationof VEGF-165, VEGF-121, and VEGF-110 Crystallization attempts with the VEGF-165 variant resulted in the formation of poorly diffracting crystals which could not be improved. X-ray quality crystals were first obtained using VEGF-110 generated by plasmin cleavage of VEGF-165; the same space groups were later observed for VEGF-121. Hexagonal rods grew aRer mixing 40 p1 of protein in crystallization buffer with 40 p1 of reservoir solution (0.2 M ammonium sulfate, 2.0 M sodium formate in 0.1 M sodium acetate buffer pH 4.6) and equilibrating against reservoir solution (Fig. 2a). The space group was determined as P61(5j (a=96.9 A, c = 127.2 A), and considerations of solvent content suggested two VEGF dimers in the asymmetric unit (VM= 3.4 b3/Da). However, the crystals diffracted only to about 4 A and were sensitive to X-ray exposure. In one case, a spontaneous transition from P6, 5 ) (a=96.9 A, c=127.2 A) to P6,(,,22 (a=96.9 c = 63.6 A1 was observed upon radiation exposure. A second crystal form consisting of orthorhombic h, 356 H.W. CHRISTINGER ET AL. Fig. 2. Five different crystal forms obtained with different constructs of the receptor binding domain of VEGF. a: Hexagonal rods (space group p6,(,,; b: orthorhombic blocks (R,2,2,) obtained with VEGF-110 and VEGF-121; c: monoclinic d: tri- clinic (Pl) crystals obtained with VEGF-8:109; e: monoclinic crystals (F2,) of VEGF-8:109 used to determine the structure by multiple isomorphous replacement. blocks originally grew using a reservoir solution consisting of 0.2 M ammonium acetate, 0.1 M sodium acetate, 30% polyethylene glycol (PEG) 3350, pH 4.6 (Fig. 2b). These crystals were determined to be of space group P2,2,2, (a=48.2 b, b=61.5 A, c= 186.2 A), diffracted to 3.3 b resolution and contained two dimers per asymmetric unit (V, = 2.7 b31 Da). After failing to determine the structure by molecular replacement using the PDGF structure as a search model, an extensive search for heavy atom derivatives was undertaken. However, soaking with heavy atom solutions resulted in large changes in the cell axes (>5%)and no stabilizing mother liquid could be found. 40% 2-methyl-2,4-pentanediol (MPD) in 0.1M MES buffer pH 6.0 (Fig. 2c). The crystals grew within 2 weeks and diffracted to better than 3.3 A resolution. Assuming a V , of 2.0 b3/Da, the crystals contained four dimers in the asymmetric unit. Crystals in space group P1 (a=45.9 A, b=68.7 b, c=85.2 A, a = 104.9", f3 =90.0", y = 101.7') were obtained from droplets equilibrating against a reservoir of 14% PEG3350, 10% isopropanol, 0.2 M ammonium acetate pH 5.6 (Fig. 2d). Assuming a V, of 2.7 A3/Da, the crystals contained four dimers in the asymmetric unit. The crystals diffracted to better than 3.0 b resolution. A third crystal form was obtained by adding calcium chloride to a condition that had previously resulted in precipitate. Crystals in space group P2, (a= 55.6 A, b = 60.4 A, c = 77.7 A, f3 = 90') grew from 30% PEG3350, 20% isopropanol, 0.2M ammonium acetate, 0.2M calcium chloride (pH 5.6), reaching a size of 0.6 x 0.3 x 0.3 mm3 within 2 days (Fig. 2e). Decreasing the PEG concentration slowed crystal growth and improved crystal morphology, and conditions were further improved by adding sodium chloride. The final conditions were 14% PEG3350, 20% isopropanol, 0.2M ammonium acetate, 0.2M calcium chloride, 0.4M sodium chloride (pH 5.6). Stabilizing artificial mother liquor was found by reducing the isopropanol concentration to 5%, adding 5% MPD, and increasing the PEG concentration to (a); Crystallization of VEGF-8109 Preliminary experiments using nuclear magnetic resonance spectroscopy to determine the three-dimensional structures of the receptor-binding domain of VEGF suggested that the first seven residues, as well as the last residue of VEGF-110, are disordered in solution (W.J. Fairbrother, personal communication). Therefore, a shortened variant, VEGF-8:109, was constructed and used in crystallization screenings, resulting in a total of three different crystal forms suitable for crystallography. Crystals in space group C2 (a=235.5 A, b=44.4 b, c = 75.2 A, p = 101.6") were grown from sitting drops equilibrating against reservoir containing CRYSTALLIZATION OF VEGF 30%. Reducing the isopropanol concentration from 20 to 5%greatly simplified the handling of the crystals as the evaporation of isopropanol caused severe convection in the drops as a result of concentration and temperature gradients. The crystals contained two dimers in the asymmetric unit (V, = 2.7 A3iDa) and diffracted anisotropically to better than 2.5 A in the (a*,c*) plane but only to 2.7 A in the b* direction. The search for heavy atom derivatives was again hindered by nonisomorphism reflected in large changes in the b cell axis upon soaking the crystals with heavy atom solutions. However, we could readily obtain isomorphous crystals of the single cysteine mutants in this space group, enabling us to produce a heavy atom derivative by soaking the mutant crystal with lead acetate and using an unsoaked mutant crystal for a native dataset. “he solvent-flattened single isomorphous replacement density clearly allows for the determination of the relationships between monomers in the asymmetric unit, and structure determination making extensive use of non-crystallographic symmetry is in progress. ACKNOWLEDGMENTS We thank Jim Wells and Tony Kossiakoff for support and comments on the manuscript; Wayne Fairbrother for suggesting the 8-109 construct; Dot Reilly and Sylvia Wong for fermentation runs; Henry Heinsohn, Rex Hayes, Marge Winkler, Mark Ultsch, and Mike Randal for suggestions on refolding and purification; Beth Gillece-Castro and Jim Bourell for mass spectrometry data; Angie Namenuk and Alan Padua for N-terminal sequencing; and Bing Li for binding assays. REFERENCES 1. Ferrara, N. The role of vascular endothelial growth factor in phatological angiogenesis. Breast Cancer Res. Treat. 36:127-137, 1995. 2. Folkman, J . Angiogenesis in cancer, vascular, rheumatoid and other disease. Nature Med. 1:27-31, 1995. 3. Dvorak, H.F., Brown, L.F., Detmar, M., Dvorak, A.M. Vascular permeability factorhascular endothelial growth factor, microvascular hyperpermeability and angiogenesis. Am. J. Pathol. 146:1029-1039, 1995. 357 4. Neufeld, G., Tessler, S., Gitay-Goren, H., Cohen, T., Levi, B.Z. Vascular endothelial growth factor and its receptors. Prog. Growth Factor Res. 5:89-97, 1994. 5. Aiello, L.P., Avery, R.L., Arrigg, P.G., Keyt, B.A., Jampel, H.D., Shah, S.T., Pasquale, L.R., Thieme, H., Iwamoto, M.A., Park, J.E., Nguyen, H., Aiello, L.M., Ferrara, N., King, G.L. Vascular endothelial growth in occular fluid of patients with diabetic retinopathy and other retinal disorders. N. Engl. J. Med. 331:1480-1487, 1994. 6. Kim, K.J., Li, B., Winer, J., Armanini, M., Gillet, N., Phillips, H.S., Ferrara, N. Inhibition of vascular endothelial growth factor-induced angiogenesis suppresses tumour growth in vivo. Nature 362:841-844, 1993. 7. Park, J.E., Keller, G.A., Ferrara, N. The vascular endothelial growth factor (VEGF) isoforms: Differential deposition into the subepithelial extracellular matrix and bioactivity of extracellular matrix-bound VEGF. Mol. Biol. Cell. 4:1317-1326, 1993. 8. Keyt, B.A., Berleau, L.T., Nguyen, H.V., Chen, Heinsohn, H., Vandlen, R., Ferrara, N. The carboxyl-terminal domain of vascular endothelial growth factor is critical for its mitogenic potency. J. Biol. Chem. 271:7788-7795, 1996. 9. Sun, P.D., Davies, D.R. The cystine-knot growth-factor superfamily. Annu. Rev. Biophys. Biomol. Struct. 24269291,1995. 10. Isaacs, N.W. Cystine knots. Curr. Opin. Struct. Biol. 5:391-395, 1995. 11. McDonald, N.Q., Lapatto, R., Murray-Rust, J., Gunning, J., Wlodawer, A., Blundell, T.L. New protein fold revealed by a 2.3 A resolution crystal structure of nerve growth factor. Nature 354:411-414, 1991. 12. Schlunegger, M.P., Grutter, M.G. Refined crystal structure of human transforming growth factor-p2 at 1.95 A resolution. J. Mol. Biol. 231:445-458, 1993. 13. Daopin, S., Li, M., Davies, D.R. Crystal structure of TGF-p2 refined at 1.8 A resolution. Proteins 17:176-192, 1993. 14. Oefner, C., D’Arcy, A., Winkler, F.K., Eggimann, B., Hosang, M. Crystal structure of human platelet-derived growth factor BB. EMBO J. 11:3921-3926, 1992. 15. Lapthorn, A.J., Harris, D.C., Littlejohn, A., Lustbader, J.W., Canfield, R.E., Machin, K.J. Morgan, F.J., Isaacs, N.W. Crystal structure of human chorionic gonadotropin. Nature 369:455-461, 1994. 16. Wu,H., Lustbader, J.W., Liu, Y., Canfield, R.E., Hendrickson, W.A. Structure of human chorionic gonadotropin at 2.6 A resolution from MAD analysis of the selenomethionyl protein. Structure 2:545-558, 1994. 17. Jancarik, J., Kim, S.H. Sparse matrix sampling: A screening method for crystallization of proteins. J . Appl. Crystallogr. 24:409-411, 1991. 18. Kabsch, W. Evaluation of single crystal X-ray diffraction data from a position sensitive detector. J. Appl. Crystallogr. 21:916-924, 1988. 19. Otwinowski, 2. Oscillation data reduction program. In: Proceedings of the CCP4 Study Weekend “Data Collection and Processing.” Sawyer, L., Isaacs, N., Bailey, S. (eds.). UK: SERC Daresbury Laboratory, Warrington, 1993:5662.