Assessment of protein solution versus crystal structure determination using spin-diffusion-suppressed NOE and heteronuclear relaxation data.

A spin-diffusion-suppressed NOE buildup series has been measured for E. coli thioredoxin. The extensive 13C and 15N relaxation data previously reported for this protein allow for direct interpretation of dynamical contributions to the 1H-1H cross-relaxation rates for a large proportion of the NOE cross peaks. Estimates of the average accuracy for these derived NOE distances are bounded by 4% and 10%, based on a comparison to the corresponding X-ray distances. An independent fluctuation model is proposed for prediction of the dynamical corrections to 1H-1H cross-relaxation rates, based solely on experimental structural and heteronuclear relaxation data. This analysis is aided by the demonstration that heteronuclear order parameters greater than 0.6 depend only on the variance of the H-X bond orientation, independent of the motional model in either one- or two-dimensional diffusion (i.e., 1-S2 = 3/4 sin2 2 theta sigma). The combination of spin-diffusion-suppressed NOE data and analysis of dynamical corrections to 1H-1H cross-relaxation rates based on heteronuclear relaxation data has allowed for a detailed interpretation of various discrepancies between the reported solution and crystal structures.