Posted in: information

Cell review:new drug research and development in the era of freeze electron microscopy

Structure based drug discovery (SBDD) is a necessary method to design and optimize innovative drugs. This review will deeply explore the rapid rise of cryo em in the field of SBDD and its main role, and explain how it provides rich new structural information for high-value pharmacological targets. Compared with X-ray crystallography, the main advantage of freeze electron microscopy is that it can skip the cumbersome crystallization steps and directly image the vitrified biological macromolecules; Cryoelectron microscopy can also provide more dimensional information, including heterogeneity and dynamics. In addition, this review will also discuss the recent and future development of cryoelectron microscopy, and explore the wide impact of this technology in the pipeline of SBDD.

SBDD in the era of freeze electron microscopy

SBDD is a research and development method of rational drug design for the target based on the basic atomic structure information of the target. In the 1980s, with the approval of enzyme targeted drugs such as captopril, captopril and Dozolamide, SBDD method first appeared. This batch of drugs approved by FDA combines the two emerging technologies of crystal structure modeling and computer-aided molecular modeling, and successfully solves the problems of high-throughput screening method (HTS) in traditional wet laboratory, such as high cost, time-consuming and low return. Since then, with the continuous innovation of computing technology, the crystal structure of a large number of drug targets has been analyzed, and SBDD method has entered a stage of rapid development. From 1999 to 2013, 78 of the 113 first in class drugs approved were found based on the SBDD method. Although the development of SBDD is fast enough, the expectations of academia and pharmaceutical industry are obviously higher. SBDD method can often find another way to verify the targets that were not considered as drugs in the past, and further develop new drugs. Such as the K-ras (G12C) target, which uses the crystallographic structure to identify a previously unknown binding pocket to avoid competition with the GDP/GTP of picomolar affinity. Because target validation is one of the main difficulties in discovery and development, first in class drug molecules can provide new insights into the effectiveness of targets and disease applications, such as bromodomain bromine domain inhibitors (+) – JQ-1 and i-bet762. These compounds have been successfully used to characterize and verify the importance of bromine domain in various diseases, and gave birth to a large number of clinical candidate drugs. Even for known drug targets approved by FDA, further SBDD is often needed in clinic. For example, some drugs need better selective optimization design. After targeted design and modification, erdafitinib showed a higher choice for fibroblast growth factor receptor than the original drug. In addition, some drugs may need to optimize the efficacy or efficacy, or provide the selectivity of specific receptor subtypes. For example, improving the selectivity of sphingosine-1-phosphate (S1P) inhibitor siponimod to S1P1 rather than s1p3 is the key to improve its efficacy and safety over non selective S1P inhibitors. In addition, the structural improvement of anti-cancer drugs based on S1P1 and s1p3 is facing the problem of continuous drug resistance to many anti-cancer drugs. The bottleneck of SBDD is to obtain high-resolution biological target structure information. Although some small and ordered biomolecules meet the research scope of X-ray crystallography, the proteins in most known targets, such as transmembrane receptors or dynamic complexes, are difficult to crystallize, resulting in the inability of these target proteins to use crystal means for high-resolution structural analysis. In addition, X-ray crystallography often modifies the target protein, such as truncated body design, introducing thermal stability mutation or inserting an exogenous domain, which affects the subsequent SBDD structure information analysis. A key factor to be considered is that a large number of target proteins cannot meet the requirements of crystallization conditions. However, these difficulties are being overcome one by one by cryoelectron microscopy technology. The resolution of cryoelectron microscopy is high enough, and a large amount of data generated by cryoelectron microscopy can also be used in computational aided drug design (CADD), which is also the core topic of this review. Different from X-ray crystallography, cryoelectron microscopy does not need to crystallize the target:the purified target biological macromolecules will be instantly frozen in a thin layer of amorphous glass ice, and then imaged by transmission electron microscopy to record hundreds of thousands to millions of cryoelectron microscopic particle data, which can be used to reconstruct the three-dimensional electrostatic potential diagram and accurately model the macromolecules. Therefore, this technique is very suitable for the structural determination of protein complexes, proteins with low thermal stability and high dynamic motion, and transmembrane proteins in lipid micelles. With the continuous improvement of resolution, cryoelectron microscopy has become a powerful tool for drug design.

Cryoelectron microscopy and drug discovery

Before 2014, the freeze electron microscope could hardly analyze the structure with a resolution better than 4.0 Å, which directly led to its inability to provide effective data support for SBDD work. However, in the past few years, the explosive breakthrough of freeze electron microscopy has produced a large number of high-resolution structural data, which could not be achieved before. This qualitative leap is due to many technological innovations, such as direct electronic detectors for recording images, improved computing methods and hardware clusters for processing large data sets. These technological leaps are reviewed in detail in other literature. In addition, as a direct visualization technology, frozen electron microscopy can quickly judge the aggregation and stability of samples, so as to quickly improve the quality of samples by genetic and biochemical means, stabilizing proteins with interaction factors, or extracting membrane proteins from cell membrane environment by optimizing detergents. Based on the above, the number of freeze electron microscope structures with a resolution of 4.0 Å or higher in PDB has increased from a total of 16 before 2014 to 1753 new structures submitted in 2020 alone (Fig. 1, a). In the newly uploaded structure, the proportion of resolutions higher than 4.0 and 3.5 Å increased from 36%and 12%in 2015 to 75%and 50%in 2020, respectively. What’s more exciting is that by 2020, the proportion of freeze electron microscope structure with resolution higher than 3.0 Å and 2.5 Å has reached 18%and 3%respectively, realizing the future breakthrough before freeze electron microscope structure analysis (Fig. 1, b). In order to systematically evaluate the impact of cryoelectron microscopy on the field of SBDD, we (the author) investigated the target related structural data of 200 most commonly used prescription drugs in the United States in 2018. 72%of the targets contained structural information in the PDB database. In terms of subdivision, these structural information are determined by X-ray crystallography (42%), freeze electron microscopy (15%) or a combination of both (15%) (Fig. 1, c). The targets analyzed by freeze electron microscopy include many transmembrane proteins, such as ion channels (GABAA, CAV, NAV and KATP), activated G protein coupled receptors (GPCRs) and transporter proteins (serotonin transporter, NaCl transporter).

Figure 1 The improvement of the resolution of freeze electron microscopy and its contribution to the structural characterization of protein drugs. (A) The absolute number of freeze electron microscope structures uploaded in PDB below a specific resolution; (B) Percentage of frozen electron microscope structures uploaded in PDB below a specific resolution. (C) Target map of 200 popular prescription drugs in 2018, classified according to the structural characteristics of the target; (D) Target maps of 44 popular GPCRs prescription drugs, classified by structural characteristics; (E) The target map of the 200 drugs with the highest sales volume in 2018 (as the representative of new drugs), classified according to the structural characteristics of the target. The data in 2020 are the protein targets of 200 most popular prescription drugs and 200 drugs with the highest sales volume in 2018 (if applicable) publicized by njardardson laboratory, and then the relevant structures are determined in PDB for manual screening.

Of the more than 200 most common prescription drugs, GPCRs account for 44. These drugs include agonists, antagonists and reverse agonists targeting GPCRs (Fig. 1, D; note that antagonists and reverse agonists are pharmacologically different, but here we (the author) classify them as antagonists). 32 (73%) of these GPCRs have been subjected to some form of structural analysis, including crystal structure combined with antagonist (44%) or agonist (7%), freeze electron microscope structure combined with agonist (9%), or structural analysis by X-ray crystallography and freeze electron microscope (20%). It is worth noting that the highly dynamic structure of GPCR makes it difficult to obtain high-quality crystals. Therefore, most GPCR crystal structures can be analyzed only after they are combined with antagonists. In conclusion, cryoelectron microscopy technology has a profound impact on prescription drugs that have existed in the market for many years. In order to further understand the role of freeze electron microscopy in future drug discovery, we (the author) also investigated 200 drugs that achieved the highest profit in 2018 to represent those newly discovered drugs on the market (Fig. 1, e), which are referred to as new drugs for short. There are significant differences between these new drugs and the most commonly used drugs mentioned earlier. A considerable number of new drugs have been characterized by crystallography, which reflects the importance of structural data in today’s drug research and development:even if they are not driven by structure, there are few cases that do not pursue structure, because structural information can provide key data for the optimization and further discovery of lead compounds. In addition, considering the long time of drug development, cryoelectron microscopy, a new technology rising in recent years, accounts for a small proportion in this list, but its contribution is still considerable. These drugs and targets include biopharmaceuticals, ion channels and GPCRs, as well as other highly active macromolecules that are not suitable for crystallization.

Contribution of freeze electron microscope to SBDD and analysis of new structure

Although many FDA approved drug target structures can be analyzed by X-ray crystallography, cryoelectron microscopy is opening the door to more and more difficult or even non crystallizable targets, such as proteins and protein complexes with larger molecular weight and more dynamic. Cryoelectron microscopy also significantly reduces the difficulty of studying intracellular complexes, such as ribosomes, chromatin modification complexes and transcription machines of pathogens. For example, the structure of a first in class inhibitor related to mitochondrial RNA polymerase complex was recently analyzed by freeze electron microscopy. It is worth noting that in the field of membrane proteins, the contribution of freeze electron microscopy is unparalleled. Whether traditional drugs or new prescription drugs, many drugs target GPCRs, ion channels and transporter proteins. However, it is very difficult to analyze the structure of membrane proteins by X-ray crystallography. Although lipid cubic crystallization has made some progress in the field of GPCR, GPCR protein usually needs thermal stability mutation or fusion with other proteins to promote the formation of crystals. Moreover, in order to obtain a stable conformation after transformation, a large number of cumbersome and complex screening of clone construction, experimental methods and conditions are needed. In contrast, the structure of freeze electron microscope can be directly used to analyze the biochemically stable membrane proteins treated with previous stain or nano disk, and obtain the structure of proteins in or close to physiological state. The ability of cryoelectron microscopy is unstoppable in analyzing the structure of complex membrane proteins, and a large number of high-resolution structures have been successfully analyzed. Membrane proteins have long been popular targets for approved drugs, and their structures have only recently been revealed by cryoelectron microscopy (Fig. 2).

Figure 2 G protein coupled receptors, transporters (upper row) and ion channels (lower row)排),每个受体有相应的FDA批准的配体分子(蓝框)。

利用冷冻电镜解析膜蛋白结构的突出进展,部分原因受益于新试剂的设计和使用。这些试剂可以在体外纯化过程中维持跨膜蛋白的结构,在冷冻制样过程中保护蛋白,并为高分辨率的结构解析提供均质样品。去垢剂如正十二烷基β-D-麦芽糖苷(DDM)和月桂基麦芽糖新戊二醇(LMNG),可以有效地从细胞膜上溶解跨膜蛋白,并维持蛋白质的生理状态构象。去垢剂的使用也会产生一些问题,如去垢剂形成的空胶束和与包裹蛋白质的去垢剂同时存在存在会引起样品的不均一,对后期的数据处理处理产生影响;也可能会导致冷冻样品制备时的气液界面收到破坏,产生一些不好的结果。脂质纳米盘是去垢剂的一种替代品,原则上可以为结构和生物物理研究提供接近胜利状态的脂质双分子层。脂质纳米盘在膜蛋白药物靶点上的应用已经非常关键和广泛。举例而言,将纳米盘与冷冻电镜技术相结合,成功阐明了TRPV1和TRPV5离子通道(在TRPV1的情况下,脂质对抑制剂的结合至关重要)、GABAA配体门控离子通道、人类P-糖蛋白以及GPCR-β-arrestin复合物的高分辨率结构和机制。关于纳米盘的进一步介绍可查阅。冷冻电镜还可以用来解析嵌入脂质体中的蛋白质的结构,允许在更接近生理状态的的电化学梯度中对离子通道以及孔蛋白进行可视化研究。在过去的几年中,冷冻电镜也在生物制药领域产生了巨大影响。在较新的药物中,生物制药的占比正越来越高。如果仅将目光聚焦于药物靶点识别这一领域,生物制药的结晶技术确实称得上有所改善。然而,冷冻电镜已经为一些关键的生药物研发提供了基于全长蛋白的结构信细节息胰岛素受体一种二聚化的酪氨酸激酶受体蛋白,在调节人体的葡萄糖平衡方面起着关键作用。胰岛素受体信号通路的失调会引起一些疾病,如II型糖尿病,全球约有9.3%(4.63亿人)受到影响两个独立的研究小组利用冷冻电镜在胰岛素受体结构解析方面取得了突破进展;第一个小组以4.3Å和7.2Å的分辨率分别解析了与一个或两个胰岛素分子结合的胰岛素受体胞外结构域结构,第二个小组以3.1Å的分辨率获得了与四个胰岛素分子结合的胰岛素受体胞外结构域结构(图3, A)。这些结构解释了胰岛素受体结合胰岛素的不同结合位点,以及激活这一关键药物靶点所进行的构象变化。

类似的例子比比皆是:从HER2-trastuzamab-pertuzumab复合物到SARS-CoV-2和中和抗体的结构解析,冷冻电镜为生物治疗的新老靶点提供了新的视点,为进一步发现和开发仿制药和first-in-class药物铺平了道路。另一个值得注意的例子是B淋巴细胞抗原CD20,它是治疗白血病和自身免疫性疾病的一个重要的治疗靶点,尽管其功能作用仍不清楚。尽管CD20的分子量较小,只要35kDa左右,但分别与单克隆抗体利妥昔单抗(rituximab)、奥法图单抗(ofatumumab)和奥比努单抗(obinutuzumab)的Fab结合形成复合物后,都解析获得分辨率较高的CD20复合物结构(图3, B)。负染结果显示,利妥昔单抗与CD20结合后,可诱导形成高度有序的高级结构,这一发现对激活先天免疫的补体系统提供了全新见解。由于复合物中的高度动态和跨膜结构域的存在,利用结晶手段结构解析几乎不可能实现,冷冻电镜技术的应用实现了这一可能。

图3.冷冻电镜(cryo-EM)在小分子和生物制药发现方面的效用。(A)与胰岛素结合的胰岛素受体(PDB ID 6PXV)和(B)CD20与利妥昔单抗复合物(PDB ID 6VJA)冷冻电镜密度图。(C)使用GemSpot(PDB ID 6CVM)将小分子PETG精确地建模到β-半乳糖苷酶的冷冻电镜图像中。(D)基于片段的PKM2的发现,冷冻电镜密度允许正确识别和放置发现片段(PDB ID:6TTF)



冷冻电镜单颗粒技术利用数百万个颗粒的可视化投影来重建静电势图,这通常涉及数十万亿字节的原始数据。因此,该方法从计算方法的快速发展中获益匪浅,这些计算方法同时满足了对更高的分辨率的需求并加深了对粒子动力学的理解。然而,与X射线晶体学相比,冷冻电镜在获取配体-靶点复合物的高可信度模型时仍然面临着一些难题。其中一个难题是冷冻电镜难以解析得到高于2.5Å的蛋白结构,而这通常是建模人员能够精确放置配体并解析出结合位点处水分子的最低分辨率。此外,冷冻电镜的结构建模流程与晶体学完全不同:在晶体学中,模型和密度图之间有一套严格而完善的统计测量方法,该方法能够提供和模型精度相关的关键信息。而在冷冻电镜方法中,基于密度图的建模是一个完全独立的过程,仅适用收集的电镜投影来进行密度图重构,然后基于密度图进行结构建模和实空间下的微调。该过程的独立性使得模型的精度被降低了。这一问题在最近已得到改善。此外,两种方法之间还存在一些物理上的差异,如晶体学依赖电子密度图,而冷冻电镜依赖静电势图。这些差异加在一起,使得晶体学的模型验证工具无法应用于冷冻电镜模型。因此,我们可能需要为精确性开发一些新的指标。一种解决方案是使用强大的计算技术和精确的分子力场对大分子及其配体在冷冻电镜结构中的相互作用进行模拟。比如PHENIX软件包结合实空间和傅里叶空间微调和OPLS3e力场的分子动力学模型,从而生成生物分子和小分子的几何统计精修模型。OPLS3e微调工具已经被整合进到我们(作者)的自研软件GemSpot,它将各种计算方法整合为一个工作流程,从而提高冷冻电镜密度图中配体位置的准确性(图3C)。新的计算工具也推动冷冻电镜在基于片段的药物发现(Fragment-based drug discovery)中发挥作用,其中高溶解度的小片段化合物被浸泡在由多个不同结构的化合物组成的生物分子靶点中。解析复合体的结构可以解释配体与结合口袋之间关键位点的相互作用,然后可以将其组合成一个先导化合物。然而,这种方法要求配体密度质量高、分辨率高,才能正确区分配体的姿态和原子类型,目前对于冷冻电镜来说还是一个难题。最近,Saur等人在高度棘手的β-半乳糖苷酶和颇具治疗意义和挑战性的激酶PKM2的场景中成功地将冷冻电镜用于FBDD。尽管他们为了将配体置放于密度图中,而不得不将干法和湿法实验结合,但他们成功地建立了一个与β-半乳糖苷酶结合的大约150kDa的精准片段模型。更令人印象深刻的是,他们能够从四种化合物的鸡尾酒中确定哪些片段与PKM2结合(图3, D)。因此,不断发展的计算方法为冷冻电镜密度图的构建提供了一个强大的平台,可以在高分辨率下对大分子复合物进行建模。


冷冻电镜是极为精密且昂贵的仪器,需要大量的费用和人力成本来搭建、维护与操作。这一特性在很大程度上限制了冷冻电镜的发展,并将冷冻电镜的机时资源集中在了那些受政府资金扶持的大型机构上。因此,在科研界中,冷冻电镜资源的获取门槛极高。然而,这一门槛正在被逐渐降低:许多国家级设施都启动了冷冻电镜人才培养计划,以降低冷冻电镜运维的人力成本。一些大型制药公司也开始进行内部投资,设立最先进的冷冻电镜设施。此外,冷冻电镜设施的可复制性远超晶体学极其昂贵的同步加速器和线性加速器,使得该技术更有发展前景。随着100kV电子束技术的发展,未来可能会出现性价比极高的冷冻电镜,增加其在药物发现领域中的应用场景。鉴于2018年FDA批准的药物中有49%来源于中小型公司,降低冷冻电镜的成本将使冷冻电镜技术得到更广泛的应用。最近对SARS-CoV-2相关蛋白的结构表征证明了冷冻电镜的无限潜力。在病毒爆发后的几个月内,科学家们利用冷冻电镜,以极快的速度解析了新冠病毒刺突蛋白的几种构象,以及它与人源血管紧张素转换酶或许多中和人源抗体片段的复合物的结构。最近获得FDA批准的用于治疗COVID-19的再利用药物瑞德西韦(Remdesivir)与SARS-CoV-2 RNA聚合酶结合的结构也已被冷冻电镜解析。鉴于X射线晶体学一直是病毒RNA聚合酶结构测定的传统方法,对新冠病毒的冷冻电镜结构解析是一个颠覆性的创新,凸显了冷冻电镜的高时效性特点在快速反应研究中的应用。此外,冷冻电镜的分辨率仍在大幅提高,最近的一份报告指出,作为冷冻电镜的代表性复合物结构,去铁蛋白apoferritin的分辨率达到了1.25Å,该分辨率足以对单个原子进行精准定位,在某些情况下甚至可以解析氢原子和质子化态。毋庸置疑,在样品制备良好的情况下,冷冻电镜的不断改进将持续打破结构解析的分辨率记录。冷冻电镜在药物发现和开发方面的应用将进一步受益于该技术的全面自动化。在载网准备方面,一些自动化工具正在出现,以解决不可重复性和样品浪费的难题。这些技术的改进不仅会提高自动化的程度和可及性,还可能解决冷冻电镜载网制备中的其他难题,如减少颗粒在空气及水中的暴露程度。此外,机器学习方法和深度神经网络也是提高颗粒筛选速度和准确性的关键。这些自动化方法甚至有望在未来成为冷冻电镜的核心技术,从而推动冷冻电镜在药物发现领域的发展。主流硬件和软件的改进也有望提高冷冻电镜在SBDD领域的可及性。例如,更高效的检测设备能显著提高冷冻电镜的产能。在一个标准的数据收集过程中,老式的检测器相机可以每次收集1个影像,每小时产生50个影像,而较新的检测器可以每次收集9-16个影像,每小时可以产生超过200个影像,进而转化为每24小时收集的数百万颗粒投影数据。此外,虽然今天许多最高分辨率的结构是用300kV冷冻电镜获得的,但这些机器非常庞大,且前期和维护成本昂贵。在许多情况下,对于单颗粒分析中使用的薄样品,200kV的显微镜可能就足够了,甚至100kV的显微镜也可以用来获得分辨率高达3.4 Å的结构。



图4.单一的冷冻电镜数据集,投影的三维分类显示了两种不同的构象,代表了两种不同的G蛋白偶联受体-G蛋白相互作用的状态,代表了两种热力学上可比较的构象。在典型状态下(左边,PDB ID 6OS9),受体以典型的方式与G蛋白结合,其中核苷酸结合口袋为GTP结合做准备。在非经典状态下(右图,PDB ID 6OSA),G蛋白异源三聚体与经典状态相比旋转了45°,代表了沿G蛋白偶联途径的中间配体结合受体状态。缩写:α-N=G蛋白的N端α螺旋;cryo-EM=冷冻电镜;TM=跨膜螺旋。





1. 激动剂


2. 拮抗剂(也称中性拮抗剂)


3. 生物制药


4. 电子密度图


5. 静电势图(即库仑势图)


6. 基于片段的药物发现FBDD


7. 反向激动剂


8. 脂质立方相结晶


9. 脂质纳米盘


10. 冷冻电镜负染


11. 时间分辨的冷冻电镜方法