The Chemistry Nobel committee seems to have taken everyone by surprise today with their award for cryo-electron microscopy (cryo-EM). That’s not because it isn’t Nobel-worthy, though – it certainly is. But they tend to take their time before recognizing discoveries (ask 95-year-old John Goodenough, a key inventor of the lithium battery who was on many short lists this year, as he has been for many years to date).
I’ve written several times on the blog about cryo-EM, most recently just back in July, and it’s a very hot topic in analytical chemistry and structural biology. Here’s why: chemistry’s entire rise as a science depended on an understanding of molecular structure, what atoms are bonded to what atoms in a compound and their three-dimensional arrangement. The properties and reactivities of every substance depend on these details: the oxygen in our blood, the semiconductor layers in our computer chips, the receptor proteins in your brain that are changing shape as you read these lines, the catalysts that make all the gasoline in the world, and every drug for every disease.
But getting good three-dimensional structure information isn’t easy. The breakthrough (worth a well-deserved Nobel of its own!) was the discovery of X-ray crystallography. If you have a crystal (we’ll come back to that phrase) you can pass X-rays through it and see the arrangement of spots that its crystal lattice has scattered out the other side. (Figuring out the basic laws behind that process was worth another Nobel the very next year). Taking a whole set of such readings, from different angles, and collecting the information on just where those spots emerge allows you to work backwards and reconstruct what the crystal must have been that produced them. The more data you can collect, the more sure you can be about how accurately you’ve narrowed down the possibilities. Finding new ways to refine that data and converge on the real structure has been a Nobel-worthy discovery all its own.
But in general, the more atoms in each repeating molecule of your crystal, the more complex the patterns that emerge, and the more data you have to collect. The earliest X-ray crystal work was on things like table salt, with very simple patterns. But since the calculations based on all those X-ray reflection spots had to be done by hand, even that was a solid amount of work, and moving on to more complex substances quickly strained the ability of the current technology to process the data. And that was when you could get enough data to start with – the strength and quality of the X-ray sources meant that you didn’t get as many spots as you wanted in the first place, and the film that was used to record them wasn’t as sensitive as it needed to be, either.
So X-ray crystal work has advanced to a degree that would stun those pioneers, due to improvements in all those things: brighter, tighter X-ray beams, more sensitive detectors (both of which give you huge amounts more data to work with per shot), faster computers (and how), and far better algorithms for them to use while working up the numbers. This allowed the extension of the technique to molecules having dozens, hundreds, or thousands of separate atoms arranged in 3-D space, giant repeating units of the sort found in protein crystals and those of other biomolecules (just the discovery that proteins could actually be crystallized got a share of the 1946 Chemistry Nobel). J. D. Bernal and Dorothy Hodgkin (later a solo Nobelist for her crystallography work) showed in 1934 that a protein crystal would indeed diffract X-rays, a major advance, but working out the structure from that data was a problem just too big to be solved. The first protein crystal structure was determined in 1958, and that was yet another Nobel (and rightly so). Determining the structure (and thus working out many of the functional details) of key biomolecules led to Nobels or shares of them in 1988, 1997, 2006, 2009, and 2012.
So it’s safe to say that the committee is attuned to the impact that structural studies have had on the field! But X-rays are not the whole story. Electrons can be diffracted by crystals in much the same way as X-rays (Nobel in 1937). In 1982, though, a second electron-diffraction Nobel was awarded, this time with a key difference: the structural information was obtained by electron microscope work without forming a crystal of the protein complex under investigation. (In one of many odd Nobel moments, a Nobel for the actual invention of said electron microscope didn’t get awarded until 1986).
That was a remarkable advance. You’ll have noticed that everything mentioned up until now has depended on having a crystalline sample. The problem is, growing sufficiently high-quality crystals of many important molecules – or frankly, any crystals of them whatsoever – can be extremely difficult. Now that we have synchrotron sources and X-ray lasers, massive amounts of computing power and advanced computational techniques, everything we need to go after huge and difficult crystal data sets. . .the rate-limiting step is often just getting the damn crystals in the first place. If you walk into a protein crystallography lab, the problem will become immediately, painfully apparent. You will see stacks after stacks of sample plates, each well of each plate filled with a bit of protein sample and a slightly different brew of solution to try to induce it to please, please form a crystal. It is brutally empirical voodoo, wildly frustrating and time-consuming, and there’s a long list of important things that have never been studied by X-ray because they just can’t or won’t form crystals to let you do it.
“Can’t?”, you wonder? Well, consider a living cell. It’s full of various membranes, which have a huge variety of proteins imbedded in them to some degree or another. And all of these are in motion; their motions are the key to their function as they change shape subtly to accommodate ligands and binding partners. How is a person to rip these things out of the membranes in which they live, along with all their various associated other proteins that help determine their activities, and get a meaningful crystal? In recent years there actually have been X-ray structures of a few such things, but they strain the abilities of the technology and have to be done in carefully constructed artificial systems to even work at all.
But electron diffraction offers a way out. I’m not going to go into all the details here, but electrons can be focused in ways that X-rays can’t, and the resulting data contains information (particularly phase information) that X-ray diffraction data doesn’t. This simplifies structure determination of complex molecules tremendously, but it’s been difficult to work out how to do these experiments without destroying the sample under a powerful electron beam. In 1975, one of today’s laureates, Richard Henderson, was already getting some broad structural information about membrane proteins, and in 1990 he and his team worked out the first complete atomic-resolution protein structure using electron microscopy. This was done on a protein crystal, and was at the limit of the technology of the best electron microscopes available at the time. Henderson and his co-workers were truly pioneers in this area; working on this problem not only delivered a structure, but also illustrated what needed to be done (in both hardware and software) to actually make this technique into something dependable.
Meanwhile, another of today’s awardees, Jacques Dubochet, worked out another key advance, one that at first sounds bizarre. He found a way to keep crystals from forming in electron microscopy samples – specifically, water crystals. Proteins live in water, and samples for either X-ray or electron diffraction are invariably cooled down as much as possible to slow down the motions of their atoms. But when you do that to an aqueous protein sample, you get masses of ice crystals that damage the protein and complicates the data workup horribly. Ice crystals diffract electrons on their own, and that can just overwhelm the signals from what you’re trying to study. Dubochet worked out how to freeze thin films of such solutions quickly in such a way that the water doesn’t have time to line up in crystalline form – it forms a vitreous (glassy) state instead, preserving the protein structure as it does so. This idea had been proposed in a general way back in the 1940s, and many groups had tried variations of it, but getting it to work reproducibly under the sample conditions needed for electron microscopy is something else again.
And Joachim Frank, the third laureate today, contributed on the computational end, which is crucial for any attempt to deal with such data. If you have a thin film of glassy frozen protein molecules, they’re going to be lined up any old way. You can see each individual molecule with your electron microscope, but each one you encounter is tumbled into a different position. Since the electron beam can be focused, you can zero in on them one at a time, but the resulting data is a huge, messy pile. Frank, though, worked out the ways to reconstruct these into a single structural picture. X-ray crystallography machines feature high-precision mounts that precisely turn and twist the single crystals mounted on them so that the X-ray beam can be fired through them for all sorts of angles, and Frank found a way to take the various random poses of frozen electron microscopy samples and work backwards to the same sort of arrangement. This had first been done in 1968 in the lab of Aaron Klug, that 1982 Nobel mentioned above, but it was brutally hard work and difficult to extend to more general cases (for one thing, it worked best, by far, for particles that were very symmetrical to start with, and a lot of proteins are anything but). Frank’s methods are ingenious and generally applicable, and he absolutely laid the foundations for the software that’s being used today.
Cryo-EM is still advancing rapidly (as those blog posts referenced above will illustrate). The idea of getting protein structures without having to crystallize the protein is, as you’d imagine, wildly appealing to structural biologists. As is the idea of getting such structures in something much closer to a protein’s native environment (after all, most proteins don’t spend their time as beautiful faceted crystals). The resolution of such structures has gone from “tantalizing fuzzballs” to “Holy cow” over the last 20 years, and the challenge now is to generate the “Holy cow” data sets more easily, more generally, and more reliably. Every research program around the world with an interest in structural biology is paying close attention to the field – the Nobel for it is certainly deserved.
As an addendum, the Nobel committee’s scientific background article on today’s award is, as usual, an excellent detailed summary, going into many things that I don’t have space to cover. If you want more, that’s the place to go!