Remember the Walkman, Betamax, and floppy disk? In their prime, those devices were the height of technology. But each has been replaced by something better—something that corrected flaws in the original tool.
Just like the electronics industries, structural biology is a field where innovation is relentless. Researchers are constantly aiming for better, faster, and easier ways to do things. For example, HHMI Investigator Axel Brunger is using an x-ray laser he hopes will help x-ray crystallographers solve protein structures that are currently elusive. HHMI Investigator David Agard, on the other hand, helped built an electron microscope camera that is pushing the limits of microscopy to the molecular level.
Beyond the Synchrotron
The first determination of a macromolecule’s atomic structure by x-ray crystallography, in the late 1950s, was a Nobel Prize–winning achievement. By exposing a protein crystal to an intense beam of x-rays, researchers could deduce the protein’s structure from the resulting x-ray diffraction pattern. The application of synchrotrons in the 1970s as a source of x-rays for crystallography pushed the technique to the next level—the intense x-ray beams produced by these large circular electron accelerators allowed scientists to determine structures that could not be solved otherwise and to dramatically speed up the structure determination process.
Still, modern x-ray crystallography is not without its problems. “Membrane proteins and large complexes often don’t produce very good crystals,” explains Brunger, a structural biologist at Stanford University. “They often diffract x-rays weakly or don’t diffract at all. They also decay very rapidly when you start exposing them to x-rays.” Because of such difficulties, these structures often go unsolved.
Brunger and collaborators are testing the idea that a new source of x-rays may circumvent these problems altogether. With the support of an HHMI Collaborative Innovation Award (HCIA), he heads a team of scientists that is developing methods of sample delivery, data collection, and analysis for x-ray crystallography using the world’s first x-ray laser. “With the HCIA funding, what we’re trying to show is that we can do things we couldn’t do at synchrotrons,” Brunger says.
The Linac Coherent Light Source (LCLS) in Menlo Park, California, is what’s known as an x-ray free-electron laser (XFEL). Built and operated with funding from the U.S. Department of Energy and opened in 2009, it is part of the world’s longest linear accelerator—it extends across 2 miles of real estate at the SLAC National Accelerator Laboratory. The laser produces intense short pulses of x-rays, much like flashes from a strobe light. The pulses are more than a billion times brighter than most existing x-ray sources, including the powerful synchrotrons many crystallographers use today. Until a few years ago, LCLS was the only XFEL around. Today, Japan has one and others are being built in Germany, Korea, and Switzerland.
With a synchrotron, scientists must use crystals that are at least 5 µm long, about the size of a small grain of sand. With a free-electron laser, crystals as tiny as 0.2 µm—the size of some bacteria—are fair game. That’s because the laser beam can focus tightly on a very small region. This technical advance means that scientists can study proteins that were once off limits because they didn’t produce large crystals.
There is a down side, however, to all this intensity. The lasers are so powerful that they destroy crystals within nanoseconds. “After the pulse, the crystal is gone,” Brunger says. “It literally evaporates. The atoms have ionized and the crystal disappears.” But the intensity of the x-rays also means that scientists can theoretically “outrun” the damage, collecting all the data they need from a crystal before it is obliterated. This process is sometimes referred to as “diffraction before destruction.”
So far, only four protein structures have been examined using XFELs. Although these structures, or structures similar to them, had been determined before with conventional x-ray sources, Brunger maintains that their determination still represents a huge accomplishment. It demonstrates that the x-ray lasers can provide useful data.
What hasn’t been shown is that XFELs will offer a significant advantage over synchrotrons when applied to challenging biological systems, like membrane-embedded proteins. “Our goal is to develop both the experimental methods and the computational tools and then to actually test them on a number of challenging systems so we know if we can get better data using x-ray free-electron lasers,” says Brunger.
The team is working on several ways to integrate XFELs into x-ray crystallography. For example, to determine a molecule’s structure using XFELs, scientists will need to combine data from hundreds of crystals for a complete three-dimensional picture. So Brunger is working with biochemist and structural biologist James Berger (formerly of the University of California, Berkeley, now at Johns Hopkins University) to create ways to deliver crystals to the x-ray pulses. One device on their drawing board is a microfluidic chip that deposits crystals in a two-dimensional array.
In addition to the logistics of data collection, computational problems need to be tackled. Once the scientists get their data snapshots from a set of crystals, they need to merge the information for analysis. Since the computational methods for XFEL crystallography differ slightly from regular x-ray crystallography, Brunger’s team is working on new algorithms.
These methods will be developed and tested by the team’s other members—William Weis of Stanford University and HHMI Investigators David Eisenberg at the University of California, Los Angeles, and Douglas Rees at the California Institute of Technology as well as collaborators at the SLAC Linear Accelerator Laboratory/Stanford Synchrotron Radiation Lightsource and the Lawrence Berkeley National Laboratory. Brunger has high hopes. “There’s no intrinsic showstopper,” he says. “It’s really just figuring out the technology and engineering.” And once they confirm its usefulness, he believes the method will catch on quickly.
If the strategy pans out, researchers envision future studies on molecules in solution rather than crystals and, farther down the road, perhaps even single molecules. Another possibility is watching how a molecule changes over time, like the breaking and making of chemical bonds. HHMI Investigator Eva Nogales, a structural biologist at the University of California, Berkeley, is intrigued about the technology’s potential. “It’s a completely new tool and where it will take us, we don’t know, but it could open up new possibilities.”
Page 3 of 4
Reaching the Summit
Like Brunger, Agard is trying to improve on technology. He would like to see the electron microscope become a common tool for solving protein structures. It is great for looking at cells, organelles, and even the outlines of large molecular complexes. Atomic structures, however, have generally been just beyond its reach.
Very much like the x-ray beam in crystallography, the electron beam in electron microscopy is extremely harmful to samples. “The radiation damage is so bad that it is comparable to being close to the center of a nuclear bomb going off,” explains Nikolaus Grigorieff, a biophysicist and lab head at HHMI’s Janelia Farm Research Campus. “There’s really not much left after that.”
Unfortunately, reducing the dose levels also limits the amount of high-resolution data that can be collected. This produces a conundrum for researchers who are trying to use electron microscopes to look at the atomic structures of molecules. “At the dose levels that will keep the sample from turning into charcoal, you can’t get enough information to actually see the structure at high resolution,” says Agard, a structural biologist at the University of California, San Francisco.
So, microscopists have come up with a solution—they blast hundreds of thousands of molecules briefly with a high-dose electron beam and then combine the resulting images. Because each image came from a molecule in a random orientation, it is necessary first to determine the precise rotation and translation that relates each image. This is much easier to do for large, highly symmetric assemblies, such as viruses, than for your average molecule. As a consequence, until this year, the only near-atomic-resolution electron microscope structures have been from viruses.
There are two other obstacles to getting atomic-level data from the microscopes for more typical samples. The first comes from limitations in camera technology. Most electron microscope cameras convert electrons to light and then transform the light into an electronic image. At every step in the process, noise is introduced and information is lost.
The second obstacle comes from something called beam-induced motion. Over the past two years, Grigorieff and his colleagues at Brandeis University, the Scripps Research Institute, and Harvard University have used a specialized camera called a direct electron detector to record movies of a sample in an electron beam. This procedure revealed that when an electron beam hits the sample, it bows out like a drum that’s been hit. “The sample is actually swimming around in the electron beam,” explains Agard. “All the time we’ve been recording data, we’ve been getting blurring of the particle because it’s moving during the exposure. That’s one of the main reasons we haven’t been able to get high-resolution information.” Molecules move, resolution plummets. For large viruses, Grigorieff showed that each frame of the movie could be treated as a single image that when aligned would circumvent the blurring. However, smaller objects are essentially invisible within each frame, so this approach won’t work. Enter Agard.
Page 4 of 4
In 2009, Agard joined forces with Gatan, Inc., and Peter Denes, a physicist at the Lawrence Berkeley National Laboratory. Their goal: create a camera that can conquer these obstacles. Key to the design was a direct detection camera that could count individual electrons, essentially eliminating all noise in the imaging process.
With $300,000 in seed funding from HHMI, Agard, Denes, and the Gatan team began developing a next-generation direct electron detector. Like the camera that Grigorieff used, the new camera would detect electrons directly rather than having to convert them to light to create an image. Thus, no information was lost. The big challenge, however, was to figure out a rapid way to observe and count the individual electrons that hit the camera. Denes had spent years building charged-particle detectors for use in high-energy physics studies and was able to tailor the technology for an electron microscope.
Because of the prototype’s strong performance, the team received $2 million from the National Science Foundation to bankroll production of the camera’s production. Building a camera that could process 16 million pixels 400 times a second was an amazing feat of engineering for the Gatan team. By 2012, Gatan produced a full-scale prototype. Agard and his colleague Yifan Cheng showed how the new camera could be used to solve much smaller structures to very high resolution. By combining the camera with a new algorithm to align complete frames, they were able to correct beam-induced motion for much smaller complexes. In June 2013, Cheng and Agard published a near-atomic resolution view of the 20S proteasome, a large complex of proteins involved in breaking down cellular proteins, in Nature Methods.
The commercial version of the camera was christened the K2 Summit—a nod to the famous mountain peak in Asia and the fact that the camera is a second-generation direct electron detector. And it has been a huge success. Beyond the University of California, San Francisco, the next eight produced went to HHMI labs, and another 15 have been delivered to other labs around the world.
It takes several months to solve a structure using the camera. But Agard envisions a much quicker turnaround down the road. “This is just the beginning,” he says. “I can imagine that five years from now you’ll put a sample in the microscope and a day or two later you’ll get its structure at 3 Å or better resolution. It’s clearly headed in that direction, but there’s still a lot of software development and optimization that needs to be done.”
Agard and the camera team continue to tinker. “This camera changes what has been the accepted procedure for collecting cryoelectron microscope data over the last 20 or 30 years,” Agard says. “So now we can start fresh. We can figure out all the new and wonderful things we can do.”