对话约阿希姆·弗兰克（二）冷冻电镜：只要我冻得足够快，分子就以为自己在家里_风闻

导读

　　冷冻电镜技术是如何研发出来的？它又有什么作用？欢迎大家持续关注我与约阿希姆·弗兰克的对谈视频。

　　视频链接：

　　西瓜视频：

　　https://www.ixigua.com/6906033244232221191

　　本视频发布于2020年12月14日，播放量已超四百万

　　美国科学家约阿希姆·弗兰克、瑞士科学家雅克·杜博歇以及英国科学家理查德·亨德森，因为研发冷冻电子显微镜，获得了2017年的诺贝尔化学奖。冷冻电镜技术是如何研发出来的？它又有什么作用？欢迎大家持续关注我与约阿希姆·弗兰克的对谈视频。

　　【约阿希姆·弗兰克与袁岚峰对话】

　　袁岚峰：

　　You said that there is such a technique that lets the molecules feel at home. So it seems to me that the technique of the super-fast cooling right? The temperature decreases by more than 10,000 degrees in one second. Is that the technique that made the name cryo-EM right?

　　你说有这样一种技术，“让分子有在家的感觉”。所以在我看来这就是超快冷却技术对吗？温度以一秒钟内一万多度的速度下降。是这种技术让它被叫做冷冻电镜的吗？

　　约阿希姆·弗兰克：

　　Yes, the molecules are embedded in water. So they’re embedded by very quickly freezing, and when the freezing is fast enough, then the water doesn’t have a chance to expand and form crystals. 

　　是的，分子嵌入水中。所以它们是通过很快的冻结来嵌入的，当冻结足够快时，水就没有机会膨胀并形成晶体。

　　And so the molecule is again like in a native environment because the water stays in the same structure like glass, you know, whereas if you cool very slowly, then the water forms crystals, which expand and then they crush the molecule.

　　因此这个分子感觉像是在它的原生环境中，因为水保持在相同的结构上，就像玻璃一样。然而如果你冷却得很慢，水就会形成晶体，晶体会膨胀，然后压碎分子。

　　袁岚峰：

　　Yeah, because my background is a theoretical and computational chemistry, so I can understand that under such a super-fast freezing the water becomes amorphous ice, not crystal ice, because the hydrogen bonds in the crystal ice have some ice rule, so the density of crystal ice is lower than that of liquid water. And so the volume of a crystal ice is larger than liquid water, so they develop to squeeze the cells and destroy the molecule, is my understanding correct?

　　是的，因为我的背景是理论与计算化学，所以我能理解在如此快速的冷冻速度下，水变成了无定形冰，而不是晶体冰。因为晶体冰中的氢键有一定的冰规则（每个水分子都跟周围的四个水分子形成氢键），晶体冰的密度比液态水低，所以晶体冰的体积比液态水大，因此它们会挤压细胞，破坏分子，我的理解正确吗？

　　约阿希姆·弗兰克：

　　Yes, exactly.

　　是的，非常正确。

　　袁岚峰：

　　So, if you are cooling them in a such a super-fast speed, then you will get amorphous ice, but my question is this, is the density of amorphous ice the same as the liquid water?

　　所以，如果你以超高速冷却它们，你会得到无定形冰，但我的问题是，非定形冰的密度和液态水的密度一样吗？

　　约阿希姆·弗兰克：

　　I think it’s very close. 

　　是的，非常接近。

　　袁岚峰：

　　Okay, I see. So you can explain this phenomenon in terms of density. Okay, this sounds very interesting. So I read some introduction to the contributions of you three, you and Professor Henderson and Professor Dubochet. In my understanding, your contributions are like this, So Professor Henderson found a way to find a structure that if the molecules are aligning in the crystalline way and in this way, it can accept very little radiation, right? so they can tolerate the radiation.

　　好吧，我明白了。所以你可以用密度来解释这个现象。好吧，这听起来很有趣。我读了一些介绍你们三个，你，亨德森教授和杜波谢教授的贡献。在我的理解中，你的贡献是这样的：亨德森教授找到了一种方法，来确定分子的结构，如果分子以晶体的方式排列，它可以受到更少的辐射，对吧？因此他们能承受辐射。

　　约阿希姆·弗兰克：

　　That’s again the same idea to distribute the radiation load among many copies of the molecule. 

　　And in one case, these copies are aligned with one another in an order arrangement and the other case, they are single. But the idea of distributing radiation load is the same, and then the later important step, is averaging, only by averaging, do you get the signal up again.

　　这和把辐射负载分配到分子的多个拷贝中的想法是一样的。

　　在一种情况下，这些副本按照顺序排列，而另一种情况下，它们是单一的。但是分配辐射负载的想法是一样的。然后接下来重要的一步，就是求平均，只有通过求平均，你才能再次得到信号。

　　袁岚峰：

　　Yes, so after Professor Henderson, you said that actually the molecules are not needed to be crystallized, they can just be random and you can still get the signal. 

　　And it seems to me that if we say it can be random, we mean it is the best thing that they are totally random. If they are somewhat crystallized and somewhat of a random, it will be a disaster to us, right?

　　那么在亨德森教授之后，你表示实际上分子不需要结晶，它们可以是随机的，你仍然可以得到信号。

　　在我看来，如果我们说它可以是随机的，我们的意思是最好它们是完全随机的，但如果他们有点结晶又有点随机，那对我们来说就是一场灾难，对吧？

　　约阿希姆·弗兰克：

　　Yes, yes. Totally random and totally, you know, it’s not good if they are only partially random that the orientations are still very much align, then you don’t get information from a certain angle, that’s not good. 

　　So cryo-molecule is really, completely randomly Orient.

　　是的，是的。完全随机。你知道，如果它们只是部分随机的，方向仍然在很大程度上对齐，那么你就不能从某些角度获得信息，这就不太好。

　　所以在这里分子是真正完全取向的。

　　袁岚峰：

　　So it’s very interesting. So there are two extremes. It supposes that you took the extremes to us. The first one is totally ordered, the second one is totally random, but in between, that’s no good, right? 

　　这很有趣。所以这里有两个极端。第一个是完全有序的，第二个是完全随机的，但在这两者之间，就是不好的，对吗？

　　And in some reviews I see that the flexibility of the molecules is not good to us, right? So it’s bond length and bond angles can change. You can change. That is not good. So we expect a very rigid molecule.

　　在一些综述中，我看到分子（结构）的可变形性对我们不好，对吧？如果键长和键角是可以改变的，这就对我们不好了。所以我们期待一个非常刚性的分子。

　　约阿希姆·弗兰克：

　　Well, I wouldn’t say it’s not good. It is good because that’s what’s supports life, the ability of the molecule to change its shape, all the properties that give rise to life have something to do with molecules interacting with one another and be able to change their conformations. 

　　我不会说这是不好的，其实这是好的，因为这正是支撑生命的东西，分子改变形状的能力，以及所有产生生命的其他属性都与分子间能够相互作用有关，以及和能够改变它们的构型相关。

　　And I am always saying that X-ray crystallography does not give us a lot of information.

　　It gives us the information of the molecule in just one state, and that state might not be a state that the molecule assumes when it actually is doing its interactions. 

　　我总是说X射线结晶学不能给我们提供很多信息。

　　它只给我们分子在一种状态下的信息，而这个状态可能不是真实情况下分子进行相互作用时的状态。

　　So Cryo-electron microscopy gives us all the additional information of molecules changing their shape.

　　所以低温电子显微镜给我们提供了分子改变形状方面的附加信息。

　　So now, but you’re saying it’s bad. Yes, it’s bad if you cannot differentiate between the different shapes, but today’s computer programs are able to actually sort the molecules out into a different one.

　　所以如果你说这很糟糕，那的确是的。如果你不能区分出不同的形状那的确是糟糕的，但是今天的计算机程序实际上以及能够将分子分类成不同的形状了。

　　So if one molecule is doing this and another one is doing that, then you can by looking through hundreds of thousands of molecules. You can pick all the ones that are like this and pick all the ones that are then form averages. 

　　所以如果一个分子某个状态而另一个分子是另一个状态，那么你可以通过观察成千上万的分子来做实验，你可以选择所有的像这样的分子，或者所有的像那样的分子，然后平均。

　　袁岚峰：

　　So you are partitioning molecules into subgroups.

　　我明白了，您是在把分子分成亚组。

　　约阿希姆·弗兰克：

　　Yes, and we call it classification. So there are very sophisticated programs now that perform this kind of classification, and you have to realize that the raw data are very faint, single projections of molecule. 

　　是的，我们把它称为分类。现在有非常复杂的程序来进行这种分类，但你必须意识到原始数据是非常微弱的，是分子的单一投影。

　　And still these algorithms are able to sort them into populations of three -dimensional structures. So it’s a really fantastic ability now that we have. 

　　而且这些算法仍然能够将它们分为三维结构的种群。所以现在我们有了这种能力真是太棒了。

　　And I called it, story in a sample. Story in a sample, which means that, you know, you can pull all that information, all the different structures out of the same sample. 

　　我称之为样本中的故事。样本故事的意思是，你可以从同一个样本中提取所有的信息以及所有不同的结构。

　　You know you don’t have to switch to another sample, but it’s all of this comes out from the same sample, and then you can see two or three-dimensional constructions from the different populations and then see how they are related to each other.

　　你不必切换到另一个样本，所有这些都来自同一个样本，然后你可以看到来自不同群体的二维或三维结构，然后看它们是如何相互关联的。

　　袁岚峰：

　　This sounds like the basic idea of Statistical Mechanics. So it’s the basic concepts of Statistical mechanics is ensemble, it’s a whole lot of replicas, and you can just analyze one example and get the all the information. 

　　这听起来像是统计力学的基本概念。统计力学的基本概念是系综，它包含大量的复本，你只需分析一个例子，就可以得到所有的信息。

　　约阿希姆·弗兰克：

　　Yeah, speaking of that, when you have a very large number of molecules, hundreds of thousand millions of them, then even states that are very lowly populated come to you. And then you can actually, you can see a continuum of states. 

　　是的，说到这，当你有很大量的分子，比如数以亿计的时候，那么即使是布居数很低的状态，你也能看到。然后你实际上就会看到一系列连续变化的状态。

　　And when you see a continuum of states, you can use cryo-electron microscopy in order to enumerate all these different states, and they are you know, the relationship to each other is like, multi-dimensional, and then you have to sort of, you know, do some kind of dimension reduction technique to sort them, to order them in some way. 

　　当你看到一系列连续变化的状态，你可以用冷冻电镜来列举所有这些不同的分子态，它们之间的关系是类似于多维的，因此你必须做一些降维技术来把它们分类，以某种方式排列它们。

　　But when you have done this, then you have each element of this imagined array, you have a number of counts of a state. 

　　Now, if you have that, then you have the energy landscape. 

　　当你做了这些，你就得到了这个想象中的阵列的每个元素，你有一个分子态的计数。

　　如果你有这个，那么你就有了能量全景图。

　　袁岚峰：

　　Yes, because they have the Boltzmann distribution. So we can tell the energy difference.

　　是的，因为它们符合玻尔兹曼分布。所以我们可以分辨出能量的差别。

　　约阿希姆·弗兰克：

　　Exactly, so we are now able to reconstruct the energy landscape of molecule. Yeah, so there’s just been an article that we published in Nature Communications.

　　没错，所以我们现在可以重建分子的能量全景图了。【袁：这真是太美妙了！】这正我们刚刚在《自然通讯》上发表的一篇文章的内容。

　　袁岚峰：

　　That’s great. So is that the benefit of the cryo-EM is the better the XRD right? XRD cannot do this? 

　　太棒了。那么这是冷冻电镜比较于X射线衍射的一个优点吗？X射线衍射不能做这个，对吧？

　　约阿希姆·弗兰克：

　　No, it cannot, yeah.

　　是的，它不能。