【作者】 李永乐；

【导师】 谢代前； 张增辉；

【作者基本信息】 南京大学 ， 物理化学， 2011， 博士

【摘要】 世界上几乎近三分之一的蛋白质都含有金属离子,金属离子对于蛋白质的结构和功能甚至大多数生命过程都是极其关键的。金属蛋白在细胞中发挥着重要作用,如酶催化,蛋白的传输和储存,信号传递等。例如,钙结合蛋白与钙离子结合是细胞信号传递的重要步骤。铜蛋白与体内的氧化还原反应密切相关。含铁蛋白如肌红蛋白和血红蛋白具有携氧和储氧的能力。锌指蛋白与多种生物功能相关,如DNA识别、RNA聚合、转录,激酶活性调控等等。目前对于蛋白质等生物大分子的研究主要局限在经典力学的范畴。常用的分子力场,比如AMBER, CHARMM, GROMOS, OPLS等,在蛋白、核酸、糖类以及脂质等方面的研究取得了巨大的成功。然而目前对于金属蛋白的理论研究尚处于初级阶段。现在较成熟的处理金属离子的方法是键连模型,即金属离子显式地同与之配位的原子相连。但是在具有催化功能的金属蛋白的活性区域,金属离子价态会随所处状态变化,这就需要应用非键连模型。但是目前流行的分子力场中的非键连模型处理配位数可变体系均不够精确。因此,开展对金属离子以及金属蛋白的定量化的理论计算研究对于我们深入了解相关生物物理性质,预测和设计新的蛋白及功能等有着非常重要的意义。为精确描述蛋白质内金属离子的行为,本文基于最近开发的一种新型电荷,蛋白质专一的极化电荷(PPC),着力发展了可应用于金属蛋白的电荷方案。我们利用PPC的思想,用全量子力学的方法计算金属离子的配位区域,利用包含极化效应的电子密度拟合出金属以及其第一配位层的电荷。本文中,我们首先详细研究了拟使用的量化计算方法的有效性,和分子动力学模拟中的抽样方法,最后基于量化计算和动力学模拟中获得的经验,将金属体系的PPC电荷应用于双锌蛋白DFsc以及DF1这两个体系。受以上思路启发,第三章中,我们考察了一个具有催化氧化活性的人工设计的蛋白质DFsc。DFsc是一个人工设计的小型蛋白,包含两个金属离子。这两个金属离子可以是锌、铁、锰、铬等,但是只有铁和锌配位的DFsc蛋白才具有催化氧化活性。在该蛋白中,两个相邻的锌离子位于溶剂可及通道底部,不仅具有催化活性,而且配位数可变。这种情况下只能应用非键模型来对其模拟。在该章中,我们应用了可极化电荷PPC模拟DFsc,并使电荷随时间更新以模拟电荷的变化。然后我们用四种AMBER力场(AMBER94, AMBER99, AMBER99SB, AMBER03)进行模拟作为对比。MD模拟的结果表明,首先对RMSD的分析发现,DFsc在我们设计的模拟过程中结构稳定。和核磁共振(NMR)耦合数据的比对发现,我们模拟得到的结构同NMR信号具有很高的一致性。这说明我们的方案对此蛋白质的骨架的模拟是成功的。其次,对金属配位集团的结构分析(配位键的键长平均值,以及它们随时间的变化)表明,两个锌离子在模拟过程中的配位数发生了变化,平均之后发现一个为四配位,另一个为五配位。这说明两个锌离子的不对称性得到了保持。而且,我们的模拟正确反映出,两个锌离子即使在蛋白质骨架达到平衡之后,仍具有很大的涨落。根据前人的结果,我们推测该涨落是由于锌离子具有反应活性引起的。但是在对照的AMBER力场的结果中,锌离子极其稳定。最后,我们通过对锌离子间距和附近水分子运动的关系的交叉相关分析,发现水分子对锌离子的扰动是造成该体系中锌离子位置涨落的原因。但是,DFsc中的锌离子的位置的涨落是否是APPC电荷本身带来的效果,而不是体系内在的特征,尚不完全明确。于是在第六章中,我们选择DFsc的一个类似物,DF1,进行了进一步的模拟。DFl与DFsc的不同之处在于,在DFsc中存在的溶剂可及通道在DFl中被变异的残基阻隔,使得锌离子被包裹在蛋白质内部,不能跟水接触,这样就没有催化氧化活性。我们经过模拟发现,我们模拟得到的锌离子的涨落,在DF1中并不存在。DF1中的两个锌离子之间的距离稳定在3.77A左右(实验值为3.92A)。这进一步证明了我们的电荷方案可以正确反映金属离子在蛋白质中的动力学行为。而AMBER力场高估了锌离子在蛋白质的催化部位的稳定性。目前密度泛函方法是最流行的量子化学计算方法。但是对于含金属体系的性质的计算,比如光谱,并没有完全普适的泛函与基组的组合方案。要决定采用何种泛函与基组的组合,需要预先在小体系中进行测试。第五章中,我们详细地研究了目前流行的密度泛函对含金属的体系二茂铁(ferrocene)的几何结构和激发光谱的影响。计算结果表明,流行的泛函和基组的组合几乎均可给出正确的几何结构。但是不同的组合对紫外可见光谱的计算值影响则非常大。在该章中我们发现PBEO/6-31++G**配合可极化导体模型的组合可以给出非常精确地紫外-可见吸收谱,并成功应用于二茂铁的类似化合物二苯铬,以及几个实验合成的化合物(acetylferrocene(AcFc), hydroxyethylferrocene(HyFc), vinylferrocene(ViFc), and ethynylferrocene(EtFc))的光谱模拟。抽样是分子动力学模拟的一个重要环节,正确而高效的抽样方案可以得到非常精确的自由能。这对蛋白质-配体相互作用能的计算非常重要。为了测试在已有力场下应用高精度抽样技术,是否可以得到足够精确的自由能,第六章中,我们使用了目前高精度的热力学积分(TI)方法,在传统的AMBER力场下,研究了蛋白质Avidin与带电配体biotin形成的复合物Avidin/biotin的配位作用能,和该蛋白质与另一个带电配体形成的复合物Avidin/BTN2的配位作用能的差值,△△G。我们发现,即使模拟的轨迹非常稳定,模拟时间足够长,计算得到的△△G(约为lkcal/mol)仍然比实验值(6kcal/mol)小很多,反而类似于此前另一研究组(B. Kuhn and P. A. Kullman)使用AMBER力场,用另外的方法(MM-PB/SA)计算得到的数值。在模拟过程中,体系中的四个氢键发生了变化。其中两个是配体作为氢键受体,而另外两个是配体作为氢键给体。这使得主要由氢键的变化导致的自由能变正负抵销,最终使A△G过小。这证明了传统不包含极化效果的力场处理带电体系会有根本性的偏差。更多还原

【Abstract】 There are nearly one third proteins contain metal ions in the world. The metal ions play cardinal role in both the structure and function of proteins in most of life processes. Metalloproteins serve crucial functions in cell, such as enzyme catalysis, transport and storage of proteins, signal transmissions. For detail example, the calcium binding process of calcium contained protein named calbindin is the central process in signal transmission in cell, copper-contained protein, cuprein relates intimately to the redox reactions in living body, and iron-contained protein, ferritin, such as myoglobin and hemoglobin can transmute and store oxygen. As a member of zinc-contained protein, zinc finger protein, zinc finger, involves to various biology functions, such as DNA recognition, resembling of RNA, transcription, and regulation of kinase. In current research, the study of macromolecules in biology, such as proteins, is mainly limited into area of classi-cal mechanics. The general molecular force fields, such as AMBER, CHARMM, GROMOS, OPLS, etc. obtained triumph in the simulations of proteins, nucleic acids, and saccharides. While till today the theoretical research of metalloprotein is still in the initial state. The more developed model is bonded model, in which the metal ions explicitly bonded to ligating atoms. But in the active site of cat-alytic metalloproteins, the coordinate number of metal ion may change according to environment. In this case, the non-bonded model is demanding. Alas, in the popular molecular force fields, the use of non-bonded model is limited. So, the study of molecular dynamics (MD) modeling metalloprotein is of profound mean-ing for help understand the biophysical properties of metalloproteins, predict and design novel metalloproteins. For precisely describe the metal ions in metalloproteins, this thesis aims to novel charge scheme for metalloprotein, based upon a kind of partial charge devised from our group, named protein-specific polarized charge (PPC). We utilize and advance the thought of PPC, calculate the metal-binding site using quantum chemistry, then fit the partial charges of the metal ions and its first ligating shell, with the calculated polarized charge density. The organization of this thesis is like this:first we study the viability of existed functionals in modeling metal system, and then we study how the enhanced sampling technique helps to refine the results of MD simulations. At last, with the experience from modeling, we devise the PPC charges for metalloprotein and apply this MD simulation scheme into two metalloproteins, DFsc and DF1.Based on that thought, in chapter three, we study an active de novel de-signed protein, DFsc. It contains two metal ions. The ions can be iron, zinc, manganese, chromium, etc. While only the iron or zinc coordinated DFsc is with activity to catalyze oxidation reaction. In this protein, there are two neighboring zinc ions posited in the bottom of a solvent-accessible channel. The geometry of the zinc-binding group is asymmetric. Furthermore, the coordinate numbers of the two zinc ions are also changeable. We study this system using MD simulation with PPC charge, and update the charges of zinc binding group on-the-fly (adap-tive PPC, APPC). Meanwhile, four AMBER force fields (AMBER94, AMBER99, AMBER99SB, and AMBER03) are used as control group. Firstly, from RMSD analysis we find the backbone of DFsc is stabilized during MD simulation under APPC. And the NMR violations data taken from MD trajectory indicates the highly accordance to experimental value. Both of the results validate the simu-lation of backbone is correct. Then, from the analysis of both the time-evolution plots of bond length and their averaged values indicate one of the zinc ions is four-fold coordinated and the other is five-fold coordinated, so the asymmetry of the zinc binding group has been maintained. And the trajectories reveal there can be great fluctuation of the zinc ions even after the backbone has reached its equilibrium. According previous works in other group, we assign this fluctuation to the activity of this zinc binding group. At last, after the analysis of the cross correlation between water molecule around the zinc ions and the zinc-zinc dis-tance, we have found the cause of the fluctuation is the perturbation of water to zinc ions. On the contrast, in all the trajectories from AMBER force fields, the zinc ions are over stable and the coordinate numbers are all incorrect.While, whether the fluctuation of zinc ions initiates from the character of the DFsc or from our MD simulation strategy is not clear in the single case of DFsc. So in the chapter four, we have simulated an inactive analogue of DFsc, DF1. In DF1, the solvent accessible channel is blocked by mutant residues. So the zinc ions are buried in the protein and can never interact to solvent. The simulated results of DF1 indicate the zinc-zinc distance is much stable during the MD run, and its value is stabilized around 3.74A. So this evidence approves the charge scheme used here can reflect the character of metal ions in proteins correctly. And the AMBER force field always overestimates the stability of zinc ions.Currently, density functional theory (DFT) is the most popular calculation method in quantum chemistry. But for metal-contained system, there is no elixir which can be used in every system. For specific system, one must test the viability of the combination of functional and basis. In chapter five, we have thoroughly studied the influence from different functional to the geometry and UV-vis spec-tra of ferrocene. The results indicate the calculated geometries of these com-pounds are nearly the same under different functional, although their UV-vis spectra change much. We found the combination PBEO/6-311++G** with po-larizable continuum solvation model (PCM) can give the most accurate UV-vis spectra. And we also successfully apply this combination to the UV-vis spectra calculations of other metal-contained systems, including bis-(benzene) chromium and four of derivatives of ferrocene, namely, acetylferrocene (AcFc), hydroxyethyl-ferroene (HyFc), vinylferrocene (ViFc), and ethynylferrocene (EtFc).Sampling is an important step of MD simulation. Via correct and highly efficient sampling, one can obtain much accurate free energy. This is the central task of the studying of protein-ligand binding. To test whether the application of enhanced sampling technique under conventional force field can obtain sufficiently accurate free energy, in chapter six, we have studied relative binding free energy,△△G, between the binding of protein Avidin and charged ligand biotin and other complex, BTN2/avidin, using the high accuracy sampling technique, thermody-namic integration (TI), under classical AMBER force field. We find although the trajectories are stable, and the simulation time is long enough, the calculated△△G is still under estimated comparing with the experimental value (calculated value is Ca. lkcal/mol, while the experimental value is 6kcal/mol). This re-sult is much similar to the previous value produced by P. A. Kollman et al. in which the△△G was calculated by MM-PBSA method under AMBER force field. During the simulation, there are four of hydrogen bonds had changed. In two of them, ligand serves as donor, and in the other two, ligand serves as acceptor. This may cause the cancellation of the△△G, which is generated mainly from the hydrogen-bond change between the two complexes. This result also proves that there may be large deviation when doing simulations of charged system under the non-polarizable classical force field.更多还原