NMR Conformation of (−)-β-d-Aristeromycin and Its 2‘-Deoxy and 3‘-Deoxy Counterparts in Aqueous Solution

Abstract

The solution conformations of aristeromycin (1), 2‘-deoxyaristeromycin (2), and 3‘-deoxyaristeromycin (3) have been determined from an integrated analysis of X-ray (for 1 only), NMR data (i.e., 3JHH coupling constants), and ab initio calculations. One-dimensional NOE difference experiments showed that the adenin-9-yl ring in 1 and 2 is involved in a ∼50% syn ⇄ ∼50% anti equilibrium around the C-glycosyl torsion angle, whereas an anti orientation (χ = −113°) is found in the X-ray crystal structure of 1. The preferred conformation around the γ torsion angle is γt both in solution for 1−3 and in the solid state (for 1 only). The plots of energy as a function of the phase angle of pseudorotation (Figure 2) for the structures optimized by ab initio calculations (HF/3-21G*) show that there are two major wells of low energy conformers for 1−3, supporting the two-state North-type ⇄ South/West-type equilibrium of the constituent cyclopentane rings in 1−3. The ab initio calculations suggested that the South/West-type conformers are more stable than the North-type forms for 1 [ΔE (120° < P < 240°) − (330° < P < 30°) ∼10 kcal/mol], for 2 [ΔE (210° < P < 240°) − (330° < P < 60°) ∼ 4 kcal/mol] and for 3 [ΔE (P = 240°) − (330° < P < 0°) ∼ 6 kcal/mol]. Newly developed A and B sets of parameters correlating the H−C−C−H torsions to the endocyclic torsions based on the ab initio optimised structures of 1−3 have been subsequently used to interpret the time-averaged 3JHH couplings using the program PSEUROT. The discrepancy found between the X-ray crystal structure (P = 89°, Ψm = 41°) of aristeromycin (1) and its structure calculated by NMR-PSEUROT conformational analysis (35° < P [ T − T] < 65°, 35° < Ψm < 45°) ⇄ (128° < P [1E] < 131°, 34° < Ψm < 36°) based on observed 3JHH couplings in aqueous solution, as well as the relatively high error in the NMR-PSEUROT analyses for 1−3 [ΔJmax ≤ 1.6 Hz (i.e., maximal difference between experimental and PSEUROT-calculated 3JHH) and root mean square (rms) error ≤ 0.7 Hz] prompted us to reparametrize the Karplus equation implemented in the PSEUROT program by using torsion angles derived from solid-state geometries of conformationally constrained nucleosides and their corresponding experimental 3JHH. The precision of our reparametrized Karplus-type equation (rms error = 0.40 Hz) became comparable to that expected for the standard Haasnoot−Altona Karplus (0.48 Hz) equation. The results of the PSEUROT analyses performed with the standard Haasnoot−Altona Karplus equation are also very comparable in terms of geometry with those based on our reparametrized equation (eq 4). Both series of PSEUROT analyses suggest that the predominant conformation of the cyclopentane ring in 1−3 is defined by 128° < P < 140° for 1, 105° < P < 116° for 2, and 118° < P < 127° for 3, with the puckering amplitude in the range from 34° to 40° for 1−3. However, PSEUROT analyses based on our Karplus equation produced a smaller rms error by ≤0.14 Hz and ΔJmax error by ≤0.5 Hz than those performed with the standard Haasnoot−Altona equation. This work therefore highlights two important points: (i) the solution- and the solid-state structures of aristeromycin (1) are indeed different, and (ii) the close similarity of geometries derived from Haasnoot−Altona's equation or from our Karplus equation suggest that the solution structures for 1−3 are correctly defined.