A potential gene target in HIV-1: rationale, selection of a conserved sequence, and determination of NMR distance and torsion angle constraints.

Recently, the capability for determining the high-resolution, sequence-dependent structure of oligonucleotides in solution via careful analysis of multidimensional NMR spectra and structure refinement procedures has been developed. Consequently, the rationale for selection of a genome sequence as a target for drug design based on the detailed three-dimensional structure of the target is presented. The concept is illustrated by the successful search for a highly conserved region of the HIV-1 genome's long terminal repeat which could serve as a molecular target. A compound which could selectively bind the target sequence could inhibit both RNA transcription from the integrated provirus and the reverse transcription process. Of 148 unique HIV-1 sequences examined, 147 exhibit a 21-base conserved sequence (nucleotides 70-90 in HIVHXB2R) in the R region of the long terminal repeat. The only exception, a minor constituent for one individual, has a change in the penultimate base. A 13 base pair duplex sequence, [d(AGCTTGCCTTGAG).d(CTCAAGGCAAGCT)], from this conserved region was selected and synthesized for NMR structure studies. Phase-sensitive proton two-dimensional nuclear Overhauser enhancement (2D NOE) and double-quantum-filtered correlation (2QF-COSY) spectra were obtained at 500 MHz for the DNA duplex. Exchangeable and nonexchangeable proton resonances were assigned. Quantitative assessments of the 2D NOE cross-peak intensities for different mixing times were carried out using conventional Fourier transform NMR and the maximum likelihood method (MLM). Distance constraints, along with upper and lower bounds, were obtained from the 2D NOE intensities using the iterative complete relaxation matrix algorithm MARDIGRAS. Distances entailing both exchangeable and nonexchangeable protons were determined: 7-11 experimental distance constraints per residue including interresidue and interstrand distances. Simulations of the scalar coupling effects manifest in 2QF-COSY cross-peaks by means of the program SPHINX/LINSHA were compared with experimental data to yield torsion angle constraints for the sugar rings. A single conformer was inadequate to describe any of the sugar puckers, but a rapid two-state equilibrium with one conformer strongly dominant (75-95%) provided a good fit of the 2QF-COSY cross-peaks. The sugar pucker of the major conformer exhibited significant variability for the various nucleotides but was roughly 2'-endo. Though derived independently and subject to different time-averaging effects, the 2QF-COSY and 2D NOE results are in accord.