Department of Chemistry, Purdue University, 560 Oval Drive, West Lafayette, Indiana 47907-2084, USA.
J Am Chem Soc. 2011 Sep 21;133(37):14614-28. doi: 10.1021/ja201605c. Epub 2011 Aug 24.
Molecular dynamics (MD) simulations have been widely used to analyze dynamic conformational equilibria of folded proteins, especially in relation to NMR observables. However, this approach found little use in the studies of disordered proteins, where the sampling of vast conformational space presents a serious problem. In this paper, we demonstrate that the latest advances in computation technology make it possible to overcome this limitation. The experimentally validated (calibrated) MD models allow for new insights into structure/dynamics of disordered proteins. As a test system, we have chosen denatured ubiquitin in solution with 8 M urea at pH 2. High-temperature MD simulations in implicit solvent have been carried out for the wild-type ubiquitin as well as MTSL-tagged Q2C, D32C, and R74C mutants. To recalibrate the MD data (500 K) in relation to the experimental conditions (278 K, 8 M urea), the time axes of the MD trajectories were rescaled. The scaling factor was adjusted such as to maximize the agreement between the simulated and experimental (15)N relaxation rates. The resulting effective length of the trajectories, 311 μs, ensures good convergence properties of the MD model. The constructed MD model was validated against the array of experimental data, including additional (15)N relaxation parameters, multiple sets of paramagnetic relaxation enhancements (PREs), and the radius of gyration. In each case, a near-quantitative agreement has been obtained, suggesting that the model is successful. Of note, the MD-based approach rigorously predicts the quantities that are inherently dynamic, i.e., dependent on the motional correlation times. This cannot be accomplished, other than in empirical fashion, on the basis of static structural models (conformational ensembles). The MD model was further used to investigate the relative translational motion of the MTSL label and the individual H(N) atoms. The derived segmental diffusion coefficients proved to be nearly uniform along the peptide chain, averaging to D = 0.49-0.55 × 10(-6) cm(2)/s. This result was verified by direct analysis of the experimental PRE data using the recently proposed Ullman-Podkorytov model. In this model, MTSL and H(N) moieties are treated as two tethered spheres undergoing mutual diffusion in a harmonic potential. The fitting of the experimental data involving D as a single adjustable parameter leads to D = 0.45 × 10(-6) cm(2)/s, in good agreement with the MD-based analyses. This result can be compared with the range of estimates obtained from the resonance energy transfer experiments, D = 0.2-6.0 × 10(-6) cm(2)/s.
分子动力学 (MD) 模拟已广泛用于分析折叠蛋白质的动态构象平衡,尤其是与 NMR 观察结果相关的构象平衡。然而,这种方法在研究无规蛋白质时很少使用,因为在无规蛋白质中,对巨大构象空间的采样是一个严重的问题。在本文中,我们证明了计算技术的最新进展使得克服这一限制成为可能。经过实验验证(校准)的 MD 模型可对无规蛋白质的结构/动力学提供新的见解。作为测试系统,我们选择了在 pH 2 和 8 M 脲存在下变性的泛素。我们对野生型泛素以及 MTSL 标记的 Q2C、D32C 和 R74C 突变体进行了高温 MD 模拟。为了使 MD 数据(500 K)与实验条件(278 K,8 M 脲)相校准,我们对 MD 轨迹的时间轴进行了缩放。调整缩放因子以使模拟和实验(15)N 弛豫率之间的一致性最大化。所得轨迹的有效长度为 311 μs,确保了 MD 模型的良好收敛性。所构建的 MD 模型通过一系列实验数据进行了验证,包括附加的(15)N 弛豫参数、多组顺磁弛豫增强 (PRE) 和回转半径。在每种情况下,都获得了近乎定量的一致性,表明该模型是成功的。值得注意的是,基于 MD 的方法严格预测了固有动态的量,即依赖于运动相关时间的量。除了基于静态结构模型(构象集合)的经验方法外,无法完成其他方法。进一步使用 MD 模型研究了 MTSL 标记和各个 H(N) 原子的相对平移运动。所得的分段扩散系数沿肽链几乎均匀,平均为 D = 0.49-0.55 × 10(-6) cm(2)/s。通过使用最近提出的 Ullman-Podkorytov 模型直接分析实验 PRE 数据,验证了这一结果。在该模型中,MTSL 和 H(N) 部分被视为两个相互扩散的连接球体,在调和势中运动。涉及 D 作为单个可调参数的实验数据拟合导致 D = 0.45 × 10(-6) cm(2)/s,与基于 MD 的分析结果非常吻合。该结果可以与从共振能量转移实验获得的范围进行比较,D = 0.2-6.0 × 10(-6) cm(2)/s。