Computational design and experimental study of tighter binding peptides to an inactivated mutant of HIV-1 protease

Abstract

Drug resistance in HIV-1 protease, a barrier to effective treatment, is generally caused by mutations in the enzyme that disrupt inhibitor binding but still allow for substrate processing. Structural studies with mutant, inactive enzyme, have provided detailed information regarding how the substrates bind to the protease yet avoid resistance mutations; insights obtained inform the development of next generation therapeutics. Although structures have been obtained of complexes between substrate peptide and inactivated (D25N) protease, thermodynamic studies of peptide binding have been challenging due to low affinity. Peptides that bind tighter to the inactivated protease than the natural substrates would be valuable for thermodynamic studies as well as to explore whether the structural envelope observed for substrate peptides is a function of weak binding. Here, two computational methods-namely, charge optimization and protein design-were applied to identify peptide sequences predicted to have higher binding affinity to the inactivated protease, starting from an RT-RH derived substrate peptide. Of the candidate designed peptides, three were tested for binding with isothermal titration calorimetry, with one, containing a single threonine to valine substitution, measured to have more than a 10-fold improvement over the tightest binding natural substrate. Crystal structures were also obtained for the same three designed peptide complexes; they show good agreement with computational prediction. Thermodynamic studies show that binding is entropically driven, more so for designed affinity enhanced variants than for the starting substrate. Structural studies show strong similarities between natural and tighter-binding designed peptide complexes, which may have implications in understanding the molecular mechanisms of drug resistance in HIV-1 protease.

Figure 1

Molecular environments of the Glu-P3 (A) and Thr-P2 (B) peptide residues found to be suboptimal for electrostatic binding in the RT–RH crystal complex. The position of the P3 glutamate (A, atom colors, center) is pointed away from Arg8B and makes contact with Phe53A. Calculations suggest an alternative conformation (yellow) that makes better electrostatic interactions with Arg8B at the expense of packing. The wild-type P2 threonine residue (B, center) is situated in a pocket composed of four hydrophobic residues and makes no polar interactions. The crystal structure of a designed mutant peptide with a single threonine-to-valine substitution at the P2 position exhibits a structural rearrangement of the Glu-P3 residue (C, atom colors) as compared to the starting sequence (A). This rearrangement is well supported by calculation (A, C, yellow). This figure was prepared with VMD and Raster3D .

Figure 2

Changes in binding free energy contributions computed for mutants derived from protein design calculations. The worst-case changes across two van der Waals (vdW) (A) and two electrostatics/solvation (Elec/Solv) (B) parameter sets were computed for all single mutations (except proline) at each peptide position relative to the appropriate starting RT–RH sequence. Changes upon mutation in excess of +3 kcal/mol, or mutations ranked worse than +5 kcal/mol from the original sequence in stability of the complex, were dropped from consideration in the protein design calculation and are represented as the darkest red. Sequences and relative energetics for several of the best electrostatically ranking single, double, and triple mutations are also presented (C), broken down by energy term and parameter set. Mutations to the sequence are underlined, relative energies are in kcal/mol, and negative numbers indicate computed improvements to binding.

Figure 3

Isothermal titration calorimetry data for the binding of RT–RH and designed peptides to the inactivated protease. In (A), the sequences of peptides tested (mutations underlined), their design origin, and their determined thermodynamic parameters of binding are shown. Units of the dissociation constant K_d are μM, and energies are in kcal/mol. Errors on thermodynamic parameters are derived from the fitting error after repeating the experiment at least three times. A comparison of the ITC traces for the RT–RH peptide (B) and Peptide 2 (C) shows that a sharp transition is only present for the tighter binding designed peptide.

Figure 4

Superposition of the crystal structures for the RT–RT peptide (green), Peptide 1 (cyan), Peptide 2 (purple), and Peptide 3 (yellow) exhibits structural similarity. This figure was prepared with Midas Plus .

Figure 5

Double difference plot between Cα atoms in the crystal structures of the RT-RH peptide complex and the tightest-binding designed complex (Peptide 2). The largest deviations correspond to the surface residues 17, 51, and 67 in monomer B.

Figure 6

Comparison between predicted and experimentally determined structures. For reference, comparisons are made between the designed (atom colors) and crystal structure (yellow) for the starting RT–RH peptide sequence (A). Crystal rotamers were selected by design at all but two residues. Designed mutation to valine at P2 in Peptide 2 (atom colors) has good structural agreement with both the crystal structure of the mutant (yellow) and the wild-type threonine (green) (B). The glutamate to glutamine structural prediction at P3 (atom colors) for Peptide 3 also agrees well with its experimental structure (yellow) (C). Mutations at P2 to isoleucine (D) and P2′ to leucine (E) in Peptide 1 occupy similar volumes with some structural differences. This figure was prepared with VMD and Raster3D .