Understanding protein folding, evolution and function via molecular simulation

该项目在过去一年中解决了以下领域的问题： 1. 形成完全无序的蛋白质复合物。我们与 Ben Schuler（苏黎世大学）和 Birthe Kragelund（哥本哈根大学）合作，研究了组蛋白 H1 和蛋白质原胸腺肽 α 之间复合物的形成，据信该复合物充当 H1 的伴侣。实验发现，该复合物具有极高（皮摩尔）亲和力，但没有证据表明结合后形成结构。使用简约的粗粒度模型，我们能够准确地重现实验 FRET 数据并解释 NMR 数据。我们的模拟显示了静电在确定结合复合物的结构整体中的关键作用。该复合物本质上是完全无序的，接触形成的倾向与序列中的局部电荷密度相关。这种仍然完全无序的高亲和力复合物的发现代表了蛋白质-蛋白质相互作用的新范例（手稿目前正在《自然》杂志上修订）。 R·贝斯特 2.阿尔茨海默病Abeta单体的表征。肽 Abeta40 和 Abeta42 构成了阿尔茨海默病患者大脑中发现的大部分斑块。因此，了解这些由 Abeta 肽原纤维形成的斑块的组装机制是一个主要的研究目标。组装的起点当然是单体；发现用于原纤维形成的非常小的初生核（通常n=2），这表明将单体转化为“有聚集能力”的形式可能是关键步骤。许多研究（主要来自模拟）提出了一种紧凑的、异质的构象系综，其中形成了稳定的二级结构，尽管对结构的性质存在一些分歧。我们与 Hoi-Sung Chung 团队（LCP、NIDDK）合作研究单体的结构形成。 Chung 小组的 FRET 实验显示出单一状态，没有证据表明存在长期稳定的物种，并且构型弛豫时间为 40 ns。我们小组的全原子模拟也显示出在相似时间尺度上的快速构型弛豫，并且没有证据表明存在聚集的塌缩或结构化状态。模拟结果与测量的 FRET 效率以及之前测量的流体动力学半径和 NMR 可观测值一致。总而言之，数据表明 Abeta40 和 Abeta42 单体中几乎完全不存在结构化物质（手稿正在准备中）。 M.贝莱什、R.贝斯特 3. 膜蛋白的折叠（与牛津大学 Mark Sansom 合作）。我们的重点是使用粗粒度和原子模拟进行膜蛋白折叠，由于脂质膜的高粘度，这非常具有挑战性。我们发现，使用双螺旋二聚体作为测试平台，虽然当前的粗粒度模型产生合理的解离常数值，但它们经常预测折叠状态的错误结构（参考文献 1）。我们已经在所有原子模拟中解决了同样的问题；在这种情况下，力场正确预测了天然二聚体的结构，但稳定性太低。由于需要非常广泛的采样（正在准备的手稿），这一点以前并未得到重视。 J. 多曼斯基、R. 贝斯特 4. 蛋白质序列进化与设计。最近的工作表明，序列比对参数化模型在捕获折叠蛋白残基之间的相关性方面具有潜力。这种模型本质上能够为任何序列分配统计能量，因为它能够折叠成与序列比对相对​​应的结构。我们将此类序列衍生模型与统计力学采样方法结合使用，以评估折叠为给定结构的序列总数（参考文献 2）。从逻辑的极端来看，这种类型的模型可用于设计新的蛋白质序列，我们也尝试过这一点，设计了折叠到葡萄球菌蛋白 G 的每个 GA 和 GB 结构域以及 SH3 结构域的序列。我们现在已经通过这种方法获得了折叠到每个结构域的新序列的良好折叠示例，并由 John Louis（LCP，NIDDK）进行了实验测试，我们的目标是确定这些蛋白质的代表性示例的 NMR 结构。在该项目的最后一个线程中，我们试图使用统计模型设计折叠成不同结构的序列，从而形式化由菲利普·布莱恩（马里兰大学）凭经验开创的一个想法。 P. 田, R. 贝斯特 5.肌联蛋白的共翻译蛋白折叠。我们正在使用粗粒度模拟来解释 Jane Clarke（剑桥大学）小组关于核糖体上肌联蛋白共翻译折叠的实验。实验使用“捕获肽”通过卡在核糖体出口通道中来停止翻译。然而，逃脱逮捕的自发率可能会因蛋白质施加的任何力而增加。在给定的孵育时间后逃离核糖体的蛋白质的产量被用作所施加的“力”的量度。利用模拟中测量的力以及早期实验中逃逸率与力相关的实验数据，我们开发了一个动力学模型，使我们能够直接与实验中获得的蛋白质产量进行比较。最大产量是折叠状态（翻译早期）施加的最大力与折叠状态的总体（翻译后期最高）之间的权衡。 P. 田, R. 贝斯特 6. 折叠 psi 值的解释。 psi 值已被提议作为 phi 值的替代方案，其中一对组氨酸残基被工程化到蛋白质中，以便为锌等金属离子创建二价结合位点。然而，一些实验学家对 psi 值的解释提出了疑问。另一个问题是离子结合/解离相对于折叠的时间尺度，因为离子结合的时间尺度与蛋白质折叠转变路径的时间尺度相当。我们正在使用粗粒度模拟来研究在什么实验条件下 psi 值可以很容易地用折叠机制来解释。李伟 7.淀粉样原纤维的形成。我们正在与Tuomas Knowles（剑桥大学和Sara Linse）（隆德大学）合作，以描述淀粉样纤维的形成，特别是二次成核的分子机制。我们采取两种方法。第一种是基于已知的纤维结构建立一个简单的粗粒度模型，我们仍在研究这个模型。第二种是表征肽与现有纤维表面的亲和力、关联率和结合模式，并将针对该模型进行验证我们还获得了来自 Linse 实验室的突变体数据，并在匹兹堡的 Anton 超级计算机上进行了原纤维伸长的全原子模拟。 8. 开发蛋白质相分离的粗粒度模型（与 Lehigh 大学 Jeetain Mittal 合作）。最近的研究表明，由于应激而在细胞中形成的所谓应激颗粒是富含蛋白质的相，可以在体外从选定的纯化成分中重建。我们开发了一种粗粒度模型，可以捕获相行为的序列依赖性以及稳定高密度相的分子间相互作用。结合特殊的模拟方法，我们可以从模型中确定相图（参考文献 3 和正在审阅的手稿）W. Cheng, R. Best 每个项目涉及的小组成员列于最后

The project has addressed the following areas in the past year: 1. Formation of a completely disordered protein complex. In collaboration with Ben Schuler (University of Zurich) and Birthe Kragelund (University of Copenhagen), we have been studying the formation of a complex between histone H1 and the protein prothymosin alpha, believed to act as a chaperone for H1. Experimentally, it was found that the complex has extremely high (picomolar) affinity, but with no evidence for formation of structure on binding. Using a minimalist coarse-grained model, we were able to reproduce the experimental FRET data accurately and to explain the NMR data. Our simulations show the key role of electrostatics in determining the structural ensemble of the bound complex. The complex is essentially completely disordered, with the propensities for contact formation correlating with local charge density in the sequence. The finding of such a high affinity complex which remains completely disordered represents a new paradigm in protein-protein interactions (Manuscript currently under revision in Nature). R. Best 2. Characterization of Alzheimer's Abeta monomer. The peptides Abeta40 and Abeta42 constitute the majority of the plaques found in the brains of Alzheimers patients. Understanding the mechanism of assembly of these plaques formed by fibrils of the Abeta peptides is therefore a major research objective. The starting point of assembly is of course the monomer; the finding of a very small primary nucleus for fibril formation (usually n=2) has suggested to some that conversion of the monomer to an "aggregation-competent" form may be a key step. Many studies, mostly from simulation, have suggested a compact, heterogeneous conformational ensemble in which stable secondary structures are formed, although there is some disagreement on the nature of the structures. We have collaborated with the group of Hoi-Sung Chung (LCP, NIDDK) to study structure formation in the monomer. FRET experiments in Chung's group showed a single state with no evidence for long lived stable species and a configurational relaxation time of 40 ns. All-atom simulations in our group also showed rapid configurational relaxation on a similar time scale and no evidence for populated collapsed or structured states. The simulations are in agreement with the measured FRET efficiencies as well as with hydrodynamic radii and NMR observables measured previously. Taken together, the data indicates the near complete absence of structured species in Abeta40 and Abeta42 monomers (manuscript in preparation). M. Bellaiche, R. Best 3. Folding of membrane proteins (Collaboration with Mark Sansom, University of Oxford). We are focussing our efforts on membrane protein folding using both coarse-grained and atomistic simulations, which is very challenging due to the high viscosity of lipid membranes. We find, using two-helix dimers as a testbed, that while current coarse-grained models yield reasonable values for dissociation constants, they often predict incorrect structures for the folded state (Ref 1). We have addressed the same problem with all atom simulations; in that case, the force field correctly predicts the structure of the native dimer, but with much too low stability. This was not previously appreciated due to the very extensive sampling needed for this (manuscript in preparation). J. Domanski, R. Best 4. Protein sequence evolution and design. Recent work has shown the potential of models parameterized on sequence alignments to capture correlations between residues in folded proteins. Such models intrinsically are able to assign a statistical energy to any sequence for its ability to fold into the structure corresponding to the sequence alignment. We have used such sequence-derived models in conjunction with statistical mechanics sampling methods to evaluate the total number of sequences which fold to a given structure (Ref 2). Taken to its logical extreme, this type of model could be used to design novel protein sequences, and we have tried this also, designing sequences which fold to each of the GA and GB domains of staphyloccocal protein G, as well as to an SH3 domain. We have now obtained well-folded examples of novel sequences folding to each of these domains by this approach, tested experimentally by John Louis (LCP, NIDDK), and we are aiming to determine NMR structures for representative examples of these proteins. In the final thread of the project, we are trying to design sequences which fold into different structures using the statistical models, thus formalizing an idea which has been pioneered empirically by Philip Bryan (University of Maryland). P. Tian, R. Best 5. Co-translational protein folding of titin. We are using coarse-grained simulations to interpret experiments in the group of Jane Clarke (Cambridge University) on the co-translational folding of titin on the ribosome. The experiments use an "arrest peptide" to stop translation by becoming stuck in the ribosome exit tunnel. However, there is a spontaneous rate of escape from the arrest, which may be increased by any force exerted by the protein. The yield of protein escaping the ribosome after a given incubation time is used as a measure of the "force" exerted. Using the forces measured in simulations, and experimental data on the force-dependence of the escape rate from earlier experiments, we have developed a kinetic model that allows us to compare directly with the yield of protein obtained in the experiment. The maximum yield is a trade-off between the maximum force exerted by the folded state (early in translation) and the population of the folded state (highest late in translation). P. Tian, R. Best 6. Interpretation folding psi-values. Psi-values have been proposed as an alternative to phi-values, in which a pair of histidine residues is engineered into a protein in order to create a divalent binding site for metal ions such as Zinc. However, several experimentalists have raised questions over the interpretation of psi-values. An additional concern is the time scale of ion association/dissociation relative to folding, since the time scale for ion binding is comparable to that of protein folding transition paths. We are using coarse-grained simulations to investigate under what experimental conditions psi values might be easily interpreted in terms of folding mechanism. W. Li 7. Formation of amyloid fibrils. We are working in collaboration with Tuomas Knowles (Cambridge and Sara Linse (Lund University) in order to describe the formation of amyloid fibers, and in particular the molecular mechanism of secondary nucleation. We are taking two-approaches. The first is to build a simple coarse-grained model based on the known structure of the fiber, which we are still working on. The second is to characterise the affinity, association rate and binding mode of the peptides with the surface of an existing fiber, which will be validated against mutant data from the Linse Lab. We have also obtained time on the Anton supercomputer in Pittsburgh to perform all-atom simulations of fibril elongation. W. Li, M. Bellaiche, R. Best 8. Development of coarse-grained models for protein phase separation (collaboration with Jeetain Mittal, Lehigh University). Recent work has shown that so-called stress granules formed in cells as a result of stress are a protein-rich phase which can be reconstituted in vitro from selected purified components. We have developed a coarse-grained model which can capture the sequence-dependence of the phase behaviour and the intermolecular interactions stabilizing the high density phase. Together with special simulation methods, we can determine phase diagrams from the model (Ref 3 & manuscript under review) W. Zheng, R. Best Group members involved in each project are listed at the end