Activated Conformations of the ras-gene-Encoded p21 Protein. 1. An Energy-Refined Structure for the Normal p21 Protein Complexed with GDP

Abstract

A complete three-dimensional structure for the ras-gene-encoded p21 protein with Gly 12 and Gin 61, bound to GDP, has been constructed in four stages using the available α-carbon coordinates as deposited in the Brookhaven National Laboratories Protein Data Bank. No all-atom structure has been made available despite the fact that the first crystallographic structure for the p21 protein was reported almost four years ago. In the p21 protein, if amino acid substitutions are made at any one of a number of different positions in the amino acid sequence, the protein becomes permanently activated and causes malignant transformation of normal cells or, in some cell lines, differentiation and maturation. For example, all amino acids except Gly and Pro at position 12 result in an oncogenic protein; all amino acids except Gin, Glu and Pro at position 61 likewise cause malignant transformation of cells. We have constructed our all-atom structure of the non-oncogenic protein from the x-ray structure in order to determine how oncogenic amino acid substitutions affect the three-dimensional structure of this protein. In Stage 1 we generated a poly-alanine backbone (except at Gly and Pro residues) through the α-carbon structure, requiring the individual Ala, Pro or Gly residues to conform to standard amino acid geometry and to form trans-planar peptide bonds. Since no a-carbon coordinates for residues 60–65 have been determined these residues were modeled by generating them in the extended conformation and then subjecting them to molecular dynamics using the computer application DISCOVER and energy minimization using DISCOVER and the ECEPP (Empirical Conformational Energies for Peptides Program). In Stage 2, the positions of residues that are homologous to corresponding residues of bacterial elongation factor Tu (EF-Tu) to which p21 bears an overall 40% sequence homology, were determined from their corresponding positions in a high-resolution structure of EF-Tu. Non-homologous loops were taken from the structure generated in Stage 1 and were placed between the appropriate homologous segments so as to connect them. In Stage 3, all bad contacts that occurred in this resulting structure were removed, and the coordinates of the α-carbon atoms were forced to superimpose as closely as possible on the corresponding atoms of the reference (x-ray) structure. Then the side chain positions of residues of the nonhomologous loop regions were modeled using a combination of molecular dynamics and energy minimization using DISCOVER and ECEPP respectively. All of the residues of the structure were then allowed to move under restrained energy minimization where the restraints were gradually removed. In Stage 4, the nucleotide GOP was added to the model and further energy minimization was carried out. The energy of the protein-GOP complex was minimized by allowing the atoms of GOP to move with the protein held fixed and then by allowing both the nucleotide and the residues of the protein to move together. The reconstructed model agrees with the published features of the p21 protein-GOP complex including the hydrogen bonding scheme, the distribution of backbone dihedral angles, the residues contacting the nucleotide, and the orientation of loops with respect to one another in the protein. The structure also agrees with one that was predicted previously (Chen, J.M. et al., J. Biomol. Struct. Dynamics 6, 850–875 (1989)). In our molecular dynamics-energy minimization procedures, we also have been able to place all residues except Ala 66, which occurs in a poorly-defined region crystallographically, in local single residue minima, including residues reported to be in high energy regions in the x-ray structure. The constructed model can explain observed physical phenomena such as autophosphorylation by GTP on Thr 59 in proteins containing Thr in place of Ala 59.