Investigations of Protein Folding by Solid State NMR

在我们小组以前的工作中(Havlin和Tycko，PNAS 2005)，我们进行了蛋白质折叠的第一个固体核磁共振研究。这些研究集中在35个残基的蛋白质结构域HP35上，已知HP35含有三个处于折叠状态的α-螺旋片段，并且热稳定到约70℃。之所以选择HP35，是因为它可以很容易地通过标准的固相肽合成方法合成，而且因为它已经通过实验技术和计算机模拟进行了大量研究。在这项工作中，我们研究了选择性同位素标记的HP35的固态核磁共振信号与冷冻甘油/水溶液(玻璃化转变温度约为-70℃)中化学变性剂(GdnHCl)浓度的关系。在低变性剂浓度下，如预期的那样，观察到了螺旋蛋白质特有的13C核磁共振信号。在较高的变性剂浓度下，出现了两个相当意想不到的观察结果：(1)在“完全展开”状态(7M GdnHCl)，三个螺旋片段的核磁共振信号明显不同，表明第三个螺旋片段中存在高度无序，第二个螺旋片段中存在剩余螺旋含量和无序的混合，第一个螺旋片段中的无序程度明显较低，但没有螺旋含量。这与未折叠状态为均匀随机卷曲的简单假设相矛盾；(2)在展开中点附近(4.5M GdnHCl)，三个螺旋片段似乎沿着各自的展开路径进入了不同的阶段，即HP35的变性不能用只有全折叠状态和完全展开状态共存的简单两态模型来描述。这项工作表明，固体核磁共振测量确实可以提供关于蛋白质折叠的新信息。 特别令人感兴趣的是观察到HP35的第一个螺旋片段在展开状态下显然是高度有序的(但不是螺旋的)。这就提出了一个问题，即这一部分实际上是什么构象处于展开状态。我们已经在2007财年解决了这个问题，方法是进行固态核磁共振测量，直接探测第一个螺旋段中特定位置(Valine-50)的主干phi和psi扭转角。这些测量采用了我们小组以前开发的三种技术，缩写为CT-DQFD、DQCSA和2DEXMAS。在不深入这些技术的细节的情况下，只要说每种技术对主干phi和psi扭转角提供独立的约束就足够了。多个独立的约束允许我们将数据适合于phi和psi群体分布的简单模型(即，结构无序的简单表示)。当这些技术应用于处于折叠状态的HP35时，数据最适合单一构象，非常接近由结晶学确定的折叠HP35的螺旋phi和psi角。在未折叠状态下(7M GdnHCl)，组合数据不能用单一构象来拟合。相反，数据由两个phi，psi对附近的重要种群很好地描述，即-75,155度和-115，75度。其中第一个(约占总数的33%)对应于多聚脯氨酸II的构象，其他组已经认为它是未折叠蛋白质中的主要构象。第二个(约占总人口的67%)位于α-螺旋构象和β-链构象之间的“过渡区”，这是意想不到的。有趣的是，13C化学位移的从头计算表明，这两种构象具有相似的化学位移，这解释了固体核磁共振谱中观察到相对尖锐的线条，这表明结构有序度很高。 上段描述的工作代表了定量固体核磁共振方法的第一次应用，以确定展开状态下特定部位的结构参数。一份描述这部作品的手稿正在准备中。 上述实验探测到了一种“热力学展开”的状态。我们目前正在尝试研究“动力学展开”状态，即由于溶剂条件的快速变化而陷入不平衡的结构状态，随后是快速冻结猝灭。HP35是一个具有挑战性的案例，因为主要的折叠转变被认为发生在大约10微秒内，太快了，不能被冷冻淬灭。尽管如此，我们已经建造了一种设备，可以在大约100微秒内冷冻淬火(通过将最初加热到展开温度以上的蛋白质溶液喷入冷异戊烷中，使用50微米直径的喷嘴和高压泵来产生非常精细的高速溶液喷射)。利用这台设备，我们目前正在研究HP35的折叠是否可能分为两个阶段，即在10微秒的时间尺度上初步形成二级结构，然后是较慢的三级接触形成阶段和侧链填充优化阶段。对于用来表征HP35折叠动力学的光学技术来说，这样一个缓慢的阶段可能是看不见的，但在固体核磁共振中可能是可见的。我们还计划对折叠速度较慢的蛋白质进行动力学折叠研究，包括67个残基的阿尔法光影蛋白SH3结构域，我们最近成功地用固相方法直接合成了该结构域，因此可以选择性地进行同位素标记。

In previous work in our group (Havlin and Tycko, PNAS 2005), we carried out the first solid state NMR studies of protein folding. These studies focussed on the 35-residue protein domain HP35, which is known to contain three alpha-helical segments in its folded state and to be thermally stable up to approximately 70 C. HP35 was chosen because it can be readily synthesized by standard solid-phase peptide synthesis methods and because it has been the subject of numerous previous studies by experimental techniques and by computer modelling. In this work, we examined the dependence of solid state NMR signals from selectively isotopically-labeled HP35 on chemical denaturant (GdnHCl) concentration in frozen glycerol/water solutions (glass transition temperature of roughly -70 C). At low denaturant concentrations, 13C NMR signals characteristic of a helical protein were observed, as expected. At higher denaturant concentrations, two quite unexpected observations were made: (1) In the "fully unfolded" state (7 M GdnHCl), NMR signals from the three helical segments were markedly different, indicating a high level of disorder in the third helical segment, a mixture of residual helix content and disorder in the second helical segment, and an apparently low degree of disorder but no helix content in the first helical segment. This contradicts the simple assumption that the unfolded state is a uniform random coil; (2) Near the unfolding midpoint (4.5 M GdnHCl), the three helical segments appeared to have progressed to different stages along their respective unfolding paths, i.e., denaturation of HP35 cab not be described by a simple two-state model in which only the fully-folded and fully-unfolded states coexist. This work demonstrated that solid state NMR measurements can indeed provide new information about protein folding. Of particular interest was the observation that the first helical segment of HP35 was apparently highly ordered (but nonhelical) in the unfolded state. This raises the question of what the conformation of this segment actually is in the unfolded state. We have addressed this question in FY2007 by carrying out solid state NMR measurements that directly probe backbone phi and psi torsion angles for a particular site (Valine-50) in the first helical segment. These measurements employ three techniques developed previously in our group, with abbreviated names CT-DQFD, DQCSA, and 2DEXMAS. Without going into details of these techniques, suffice it to say that each technique provides independent constraints on backbone phi and psi torsion angles. Multiple independent constraints allow us to fit the data to simple models for the distribution of phi and psi populations (i.e., simple representations of the structural disorder). When these techniques are applied to HP35 in its folded state, the data are fit best by a single conformation, very close to the helical phi and psi angles determined by crystallography for folded HP35. In the unfolded state (7 M GdnHCl), the combined data can not be fit by a single conformation. Instead, the data are well described by significant populations near two phi,psi pairs, namely -75,155 degrees and -115,75 degrees. The first of these (with approximately 33% of the population) corresponds to the polyproline II conformation that has been suggested by other groups to be a dominant conformation in unfolded proteins. The second (with approximately 67% of the population) is in the "transition region" between alpha-helical and beta-strand conformations, and was not anticipated. Interestingly, ab initio calculations of 13C chemical shifts indicate that these two conformations have similar chemical shifts, accounting for the observation of relatively sharp lines in solid state NMR spectra, which had suggested a high degree of structural order. The work described in the above paragraph represents the first application of quantitative solid state NMR methods for determination of site-specific structural parameters in an unfolded state. A manuscript describing this work is in preparation. Experiments described above probe a "thermodynamically unfolded" state. We are currently attempting to study "kinetically unfolded" states, i.e., structural states that are trapped out of equilibrium by rapid change of solvent conditions, followed by rapid freeze-quenching. HP35 is a challenging case, because the main folding transition is believed to occur in approximately 10 microseconds, too fast to be freeze-quenched. Nonetheless, we have constructed an apparatus that permits freeze-quenching in approximately 100 microseconds (by spraying a protein solution, initially heated above the unfolding temperature, into cold isopentane, using a 50-micron-diameter spray nozzle and high pressure pumps to create a very fine, high-velocity jet of solution). With this apparatus, we are currently investigating whether folding of HP35 may occur in two stages, consisting of an initial formation of secondary structure on the 10 microsecond time scale, followed by a slower stage in which tertiary contacts form and sidechain packing is optimized. Such a slow stage might be invisible to optical techniques that have been used to characterize HP35 folding kinetics, but may be visible in solid state NMR. We also plan to carry out kinetic folding studies of slower-folding proteins, including the 67-residue alpha-spectrin SH3 domain, which we have recently succeeded in synthesizing directly by solid-phase methods and which is therefore amenable to selective isotopic labeling.