Protein structure determination using molecular fragment replacement and NMR dipolar couplings
Protein structure determination using molecular fragment replacement and NMR dipolar couplings
复制标题
DOI：
10.1021/ja993603n
复制
发表时间：
2000-03-08
影响因子：
15
通讯作者：
Bax, A
中科院分区：
文献类型：
--
作者：
Delaglio, F;Kontaxis, G;Bax, A
根据 NMR 数据确定溶液中蛋白质的三维 (3D) 结构主要依赖于大量质子间距离 (NOE) 的测量，并辅以源自 J 耦合和化学位移的扭转角约束。 1 最近开发的偶极耦合测量方法2 提供了额外的结构信息，可用于提高 NMR 衍生的蛋白质结构的准确性。 3 在这里，我们描述了一种仅通过偶极耦合确定蛋白质主链结构的新方法。我们方法的第一阶段，我们称之为分子片段替换（MFR），类似于 Kraulis 和 Jones 描述的用于从 NOE 模式确定局部片段结构的方法，也类似于用于拟合蛋白质 X 射线结构的主链电子密度的常用数据库方法。 5 它还与最近描述的通过搜索数据库从偶极耦合中识别蛋白质折叠的方法有一些相似之处，6 但后一种方法需要数据库中存在非常相似的结构，因此不是从头开始的方法。 MFR 方法适用于蛋白质泛素，可获得 1.8 Å X 射线晶体结构 7，并且已通过 NMR 进行了广泛研究。 8, 9 其 NMR 结构的有序部分（残基 2-72）与 X 射线结构非常一致，均方根偏差 (rmsd) 为 0.35 Å。 10 之前已对泛素中的大多数残基测量了四种主链偶联（NH；C'-N；C'-HN；CR-HR），但对 Pro、Pro 之前的残基以及具有加宽或缺失酰胺共振的残基测量较少。 11 此外，这些耦合是在两种不同的液晶相中测量的，从而产生了相对于两个不同轴系的核间矢量方向的信息。 12在目前的实施中，MFR 方法使用 7 个残基的片段大小。对于每个片段，其测量的偶极耦合组与 Brookhaven 蛋白质数据库 (PDB) 中发现的每个 7 残基片段之间的最佳拟合是通过使用线性最小二乘法确定的。 13 为了加快搜索速度，我们创建了 PDB 的简化版本，仅包含 1560 个蛋白质，其中三分之二的分辨率为 2.2 Å 或更高。从 350 000 个 PDB 片段的整个集合中，根据测量的偶极耦合和最佳拟合的偶极耦合之间的最低 2，以及实验化学位移和使用结构/化学位移数据库预测的每个 PDB 片段之间的 2（在更弱的程度上）选择 20 个。 14 通过每次将 7 个残基片段移动一个残基来重复此过程；也就是说，对于 N 残基蛋白质，搜索进行 N-6 次。支持信息中显示了偶极数据通常与最合适的 7 残基 PDB 片段的匹配程度如何的示例。忽略每个数据库片段的第一个和最后一个残基的扭转角，7×20个最佳拟合片段的重叠集合在查询蛋白质的每个残基处提供了5×20对的φ和ψ角（蛋白质的第一个和最后五个残基较少）。在有利的情况下，通常在 R 螺旋中心附近发现，给定残基簇的所有“命中”都位于 Ramachandran 图的同一区域。然而，经常会出现异常值，因为偶极耦合可能无法唯一地定义每个单独的 7-残基延伸段的构象（见下文），因此数据库中不止一种类型的 7-残基肽构象与实验偶极耦合相匹配。根据经验，我们...
Determination of the three-dimensional (3D) structure of a protein in solution from NMR data has relied primarily on the measurement of a large number of interproton distances (NOEs), supplemented by torsion angle restraints derived from J couplings and chemical shifts. 1 Recently developed methods for measurement of dipolar couplings2 provide additional structural information which can be used to improve the accuracy of the NMR-derived protein structure. 3 Here, we describe a novel approach for determining the backbone structure of a protein solely from dipolar couplings. The first stage of our method, which we refer to as molecular fragment replacement (MFR), is analogous to a method described by Kraulis and Jones for determining local fragment structures from NOE patterns4 and also is similar to the commonly used database approach for fitting the main chain electron density of protein X-ray structures. 5 It also bears some similarity to a recently described approach for identifying the fold of a protein from its dipolar couplings by searching a database, 6 but this latter method requires a very similar structure to be present in the database, and therefore is not a de novo method. The MFR method is demonstrated for the protein ubiquitin, for which a 1.8 Å X-ray crystal structure is available, 7 and which has been studied extensively by NMR. 8, 9 The ordered part of its NMR structure (residues 2-72) is in excellent agreement with the X-ray structure, with a root-mean-square deviation (rmsd) of 0.35 Å. 10 Four backbone couplings (NH; C′-N; C′-HN; CR-HR) have previously been measured for most residues in ubiquitin, but less for Pro, residues preceding Pro, and residues with broadened or missing amide resonances. 11 Also, these couplings were measured in two different liquid crystalline phases, yielding information on the internuclear vector orientations relative to two different axis systems. 12In its present implementation, the MFR method uses a fragment size of 7 residues. For each fragment, the best fit between its set of measured dipolar couplings and each 7-residue fragment found in the Brookhaven Protein Data Bank (PDB) is determined by using a linear least-squares method. 13 To expedite this search, a reduced version of the PDB was created, containing only 1560 proteins, of which two-thirds are of a resolution of 2.2 Å or better. From this entire ensemble of 350 000 PDB fragments, 20 are selected on the basis of the lowest 2 between measured and bestfitted dipolar couplings and, to a much weaker degree, the 2 between experimental chemical shifts and those predicted for each PDB fragment using a structure/chemical shift database. 14 This procedure is repeated by shifting the 7-residue fragment by one residue at a time; that is, for an N-residue protein the search is carried out N-6 times. An example of how well the dipolar data typically match those of the best fitting 7-residue PDB fragment is shown in the Supporting Information. Ignoring the torsion angles of the first and last residue of each database fragment, the overlapping collection of 7× 20 bestfitting fragments provides 5× 20 pairs of φ and ψ angles at each residue of the query protein (less for the first and last five residues of the protein). In favorable cases, as typically found near the center of R-helices, all “hits” for a given residue cluster in the same region of the Ramachandran map. Frequently, however, there will be outliers, as the dipolar couplings may not define uniquely the conformation of each individual 7-residue stretch (see below) so that more than one type of 7-residue peptide conformation in the database matches the experimental dipolar couplings. Empirically, we …