We've updated our Privacy Policy to make it clearer how we use your personal data.

We use cookies to provide you with a better experience. You can read our Cookie Policy here.

Advertisement
Why Automation Has a Major Role to Play in the Future of Cryo-EM
Article

Why Automation Has a Major Role to Play in the Future of Cryo-EM

Why Automation Has a Major Role to Play in the Future of Cryo-EM
Article

Why Automation Has a Major Role to Play in the Future of Cryo-EM

The Simons Electron Microscopy Center runs the FEI Titan Krios series of Cryo Transmission Electron Microscopes. Credit: Thermo Fisher Scientific
Read time:
 

Want a FREE PDF version of This Article?

Complete the form below and we will email you a PDF version of "Why Automation Has a Major Role to Play in the Future of Cryo-EM "

First Name*
Last Name*
Email Address*
Country*
Company Type*
Job Function*
Would you like to receive further email communication from Technology Networks?

Technology Networks Ltd. needs the contact information you provide to us to contact you about our products and services. You may unsubscribe from these communications at any time. For information on how to unsubscribe, as well as our privacy practices and commitment to protecting your privacy, check out our Privacy Policy

Cryogenic electron microscopy (cryo-EM) is an indispensable technique for structural biologists probing down to the basic materials that make up our cells. But this wasn’t always the case; only through recent advances in cryo-EM technology has its true potential been realized.

Bridget Carragher, Adjunct Professor of Biochemistry and Molecular Biophysics at Columbia University and Co-Director of the Simons Electron Microscopy Center at the New York Structural Biology Center, knows a thing or two about advancing cryo-EM techniques. Her work in the automation of cryo-EM and other electron microscopy techniques has helped shift cryo-EM from an imprecise and painstakingly slow method to one that can go toe-to-toe with other protein analysis techniques.

We spoke to Bridget to discover how automation has enhanced cryo-EM, where there are still improvements to be made, and what a fully automated future for cryo-EM might mean for researchers. 

Ruairi Mackenzie (RM): Why is cryo-EM so useful to structural biologists?

Bridget Carragher (BC): Structural biology in itself is an incredibly useful biological method.  With structural biology we are looking to understand the structure of molecular machines; our body’s fundamental machinery, and there’s hundreds of thousands of these machines that get all the work done in a cell.  So structural biology by itself is essential to understand how these machines work, what their function is and how to fix them when they go wrong.  For years and years x-ray crystallography and to some extent NMR dominated as the structural biology techniques used to solve the structure of proteins. 

Cryo-EM became popular about five or six years ago, not that it hadn’t been there for decades, but it had been plagued by being a low-resolution technique. In other words, you could see these machines but just as a blurry “blobs” and you couldn’t see where every atom in the machine was and therefore you couldn’t understand how exactly it worked. Cryo-EM went through a “resolution revolution” about five or six years ago, and that was partly finalized by the emergence of a brand-new camera that could detect electrons directly instead of photons. 

Every other camera out there is a photon detector, such as the one in your phone, but transmission electron microscopes (TEMs) produce electrons, so now we had this camera that could detect electrons directly which improves resolution. Even better than that, it takes movies instead of single frames.  Just like your iPhone can do now. So, these cameras, to some extent, gave us these advantages, and so cryo-EM went from being a “blobbology” method to a near atomic resolution method. The advantage of cryo-EM against crystallography is that you don’t need to crystallize your sample; sometimes things can take a very long time to crystallize and sometimes they never crystallize.

There are many, many structures being solved by cryo-EM right now, every day there’s many more. These are targets that crystallographers had worked on for years! They had beautiful clean fabulous protein stuck in their fridge because they would not crystallize and now, they’re going over to cryo-EM and solving these structures very rapidly.


A busy day in the control room at the Simons Electron Microscopy Center. All the center's microscopes are contained in rooms behind closed doors. They are operated from computers and screens in this centralized control room. Credit: Bridget Carragher/SEMC


RM: What are the challenges that automation can overcome in cryo-EM?

BC: One of the challenges is how long it takes to produce these structures. In some ways it seems fantastically fast in that what we need to do is collect thousands of individual images, or movies actually, from the microscopes, then we pull our structures out of those images and then we process them with a whole lot of computer algorithms. At the step of acquiring the images we until recently had about 1000 images a day coming off these microscopes. When you need many thousands of these images, that means you spend days and days on a single project at the microscope.  That’s ridiculously long in some ways.  At synchrotrons (x-ray tomography machines) these days, you can get data sets in seconds or minutes, so you can analyze hundreds of structures in a day.

Cryo-EM is a long way behind that, but we hope it will catch up. We expect with better and faster cameras and new approaches to acquire images faster, we might see an improvement by a factor of about ten of these instruments’ throughput. We’d like to automate and improve technology to get many more images out of the microscope and then we’d like to automate the processing of those images to get to the 3D structure. While you’re sitting at the microscope with your precious sample mounted inside, you should be seeing a 3D map of that sample and know when you have enough data.  Then, I can stop collecting and go onto my next very important precious sample.  Right now, we don’t have that kind of real time feedback and a lot of automation is necessary to do that.  People do still sit at the microscope and make some decisions; It’s only usually for a few hours in the morning and then they let the thing run for the rest of the day.

But I think that soon we will need to automatically change the samples so that many can be streamed through in one day; we’ll have learning algorithms that know how to set the sample up for collection and then move onto the next one when enough data has been acquired. At the moment, that’s not such a bottleneck because, as I say, we collect for a day or two days, and so the few hours I’ve spent  setting up isn’t a bottleneck, but if we were only collecting on each sample for an hour, now you’re changing samples every hour, and so you have automate that step too. 

In developing an automated pipeline, I always think of it as just dealing with the next bottleneck. As you soon as you solve one at one point of the pipeline, some other point of the pipeline immediately becomes the bottleneck. As we go faster and faster, we have to address different points along this pipeline.

RM: Which step of the workflow has proved most difficult to automate? 

BC: Most of the analysis pipeline is quite automated.  What isn’t so automated is making the specimen in the first place Most everything about cryo-EM since it was developed in the early 1980s has improved but the one thing that hasn’t changed much at all is making the sample in the first place. To do cryo-EM you have to take your sample, basically a molecular machine tumbling around in bulk solution and reduce that to a very thin film because otherwise the electrons can’t get through it. Once it is reduced to a thin film you flash freeze it; turn it into a vitrified (glassy), solid layer that is very like water except it’s frozen. That process turns out to be a lot more difficult than we thought about a decade ago. The sample when it’s reduced to this thin film comes into contact with the air water interface at the surfaces. Proteins are sometimes very unhappy at this interface and a lot of them can misbehave under those circumstances. Many groups are now starting to think very hard about what to do about this.   

In our lab we’ve developing methods where we spray the sample onto a grid using a piezoelectric dispenser, something similar to what you’d have in your inkjet printer. We use very tiny amounts of samples; that’s an advantage as sample is sometimes hard to produce, and we spray it onto a so-called self-wicking grid that will spread the sample into a thin film quickly. We can get that thin film into its frozen state much more rapidly than we can using the methods that were developed earlier, and so we can escape some of the deleterious effects of the air water interface. We have been developing this technology for quite a few years now and it’s now being commercialized so that other people can also use it. The advantages of this approach are that sometimes you see less aggregation, or less preferred orientation, or perhaps see your protein in a better-preserved state. I’m sure there’s going to be many other methods coming along in the future because everybody knows now that this is a real problem that is a real hindrance for some samples. Obviously, we are solving a lot of structures at the moment, so it’s not a problem for everything, but for some things it is a total showstopper.

Secondly, not everybody is great at making grids the old-fashioned way and new people to the field can find it quite difficult.  It can take them months and months to get good at it, so we need automation for this process too. This machine that spots these samples onto the self-wicking grids is very highly automated as well and has the potential to completely automate grid making.

RM: What other advances in cryo-EM technology would you highlight?

BC: There are two big branches of cryo-EM. One is single particles, where we look at many similar particles spread out onto a field.  The other is tomography, and there, the holy grail is to look at molecular machines inside the cell i.e. solve structures in situ. Cells are much too thick to go into a transmission electron microscope, so one method we use to create thin regions  is to use a focused ion beam inside a scanning electron microscope, to literally cut a thin layer out of a frozen cell, a thin lamella; I think Wolfgang Baumeister and his group coined the term “cutting windows into the cell”. You use this focused ion beam knife to cut a thin window and then you take that window across to the TEM. There, in a process called tomography, you take tilted views of the thin region and then you can reconstruct a three-dimensional view of this area.

From these 3D tomograms you then must find your molecules of interest that are now in a very cluttered and complex cellular environment, extract them as individual particles and then use similar averaging techniques as we do for single particles. That sounds very complicated, it is, and it’s hard. Automating this process is just starting to happen, but again I think we need a very high level of automation if that method is going to go out to a large community and be widely used.

RM: Once we overcome those challenges, how would a researcher approach cryo-EM differently?

BC: Then it will be more like the process of crystallography. I think at one of our meetings somebody once said the future belongs to the biochemists. I would say the future belongs to the biochemists and the cell biologists, maybe the neuroscientists in there too, because the microscopy will become more and more automated and will simply become a tool much like x-ray crystallography is now. That technique is highly automated, and you don’t have to become an expert to use it. You do have to make really good protein and good crystals but then you just ship them in batches to the synchrotrons. Some of them are good, some of them bad, you don’t have to worry too much as it’s all automated, you find the good ones and get data quickly and you’re done. So, I think cryo-EM will move towards that and it will become a technique that is available to anybody, any biochemist, any cell biologist, neuroscientist who wants to use it. I think we’re a little bit far away from that being the case for cell biology and neuroscience, but not so far for structural biology. I think that cryo-EM is going to get much more highly automated in the next five years.

RM: How has funding helped make this all possible?

BC: None of this happens without funding. Funding technology development is quite hard to do. There are very few outlets for doing that, but the NIH through the P41 program has been very generous in providing funding for developing new technology. Our entire center is also generously funded by the Simons Foundation and we’re very grateful for that because without that support, none of this would be happening and we would all be a long way away from where we are now. I should also mention that the NIH common fund has also now funded three national centers to provide access and training for cryo-EM across the nation and anyone can apply to use these at no cost. 

You can find out more about these national centers to provide access and training for cryo-EM using the links below:

Meet The Author
Ruairi J Mackenzie
Ruairi J Mackenzie
Senior Science Writer
Advertisement