Freeze Frame

How to troubleshoot sample preparation for cryo-electron microscopy, an up-and-coming structural biology technique.

Feb 1, 2009
Jeffrey M. Perkel

Cryo-electron microscopy may be the new kid on the structural biology block, but it is a technique on the rise. Although X-ray crystallography remains the dominant technique for solving structures because of its fine atomic resolution, not every protein (especially large complexes) will crystallize, and those that can are sometimes not sufficiently abundant to work with. That's where cryoEM comes in.

CryoEM is a form of electron microscopy that produces sub-nanometer-resolution 3D structures from 2D images of flash-frozen samples. It takes four different forms, of which single particle reconstruction and cryo-electron tomography are the most common variants. In the former, thousands of individual complexes are imaged and computationally averaged to produce a 3D rendering; in the latter, a sample is imaged from a variety of tilt angles to reproduce a 3D volumetric structure, without averaging.

Typically, a sample is spotted onto a copper grid covered with a thin film of holey carbon. (The sample is blotted with filter paper, which spreads the sample into a layer 50-100 nm thick, quickly plunged into -170°C liquid ethane, and imaged.)

The technique's success rests on sample preparation. Flash freezing, which creates ice that is vitreous, like glass, rather than crystalline, preserves the sample as it was in solution, avoiding the distortion that accompanies crystallization. But it's a tricky process, influenced by variables such as humidity, concentration, blotting time and force, and carbon thickness. "CryoEM is a technique which is much younger than x-ray crystallography and therefore less mature and less well documented," says Neil Ranson, a cryoEM specialist at the University of Leeds, UK. "Certainly, specimen preparation is still a bit of a black art."

The Scientist asked five researchers how they approach the sample-preparation problem. Here's what they said:

Mutational mimicry

<figcaption>A raw cryoEM image (left), and a slightly filtered cryoEM image from which the structure of GroEL-ATP was determined. Credit: Courtesy of Neil Ranson</figcaption>
A raw cryoEM image (left), and a slightly filtered cryoEM image from which the structure of GroEL-ATP was determined. Credit: Courtesy of Neil Ranson

Researcher: Neil Ranson, University Research Fellow, Astbury Center for Structural Molecular Biology, University of Leeds, UK

Project: Solving the structure of the molecular chaperone protein GroEL complexed with ATP

Problem: The ATP and ADP-bound forms of GroEL represent different functional states, and presumably have different structures. But, how to freeze the complex rapidly enough to prevent ATP hydrolysis?

Solution: Ranson first tried non-hydrolyzable ATP analogs, but that didn't work; they don't bind tightly enough to mimic the desired functional state. So he tried a high-tech solution: spraying ATP at the grid as it plunged towards the ethane. This reduced the amount of time GroEL was exposed to ATP before freezing from 5 seconds to about 10 milliseconds, he says, "But, as you can imagine, it becomes far more technically challenging." For instance, how can you tell where on the grid the ATP was deposited? (solution: spike the ATP with colloidal gold).

That created another problem: blotting the grids without damaging them. Thinner carbon films tended to rip, but thicker ones meant thicker ice, which "changes the views of the molecule that you get on the grid when you image it," he explains. "I spent more than a year of my life trying to get this to work with a singular lack of success."

Finally, based on previously solved structures of GroEL, he and his collaborators created a mutant with ATPase activity about 2% that of wild-type-a subtle structural change that mimics the wild-type protein but works slowly enough to capture the complex via standard plunge freezing methods (Cell, 107:869-79, 2001). "The message is that informed biochemical knowledge is a much quicker way of solving these problems than exotic freezing techniques," he concludes.

Caution: fragile particles

<figcaption>Surface representation of a virus capsid reconstruction calculated from cryo-electron micrographs (colors indicate the positions of capsomers and decoration molecules) This virus, bacteriophage SPO1, is the first non-herpesvirus to exhibit the same arrangement of capsomers previously diagnostic of herpesvirus capsids. Such conserved structural features point to a common ancestral virus that infected cells before the split into separate domains. (Current Biol, 16:R11-13, 2001) Credit: Courtesy of James Conway</figcaption>
Surface representation of a virus capsid reconstruction calculated from cryo-electron micrographs (colors indicate the positions of capsomers and decoration molecules) This virus, bacteriophage SPO1, is the first non-herpesvirus to exhibit the same arrangement of capsomers previously diagnostic of herpesvirus capsids. Such conserved structural features point to a common ancestral virus that infected cells before the split into separate domains. (Current Biol, 16:R11-13, 2001) Credit: Courtesy of James Conway

Researcher: James Conway, Associate Professor of Structural Biology, University of Pittsburgh School of Medicine

Project: Single particle reconstruction of virus capsids

Problem: Viruses can be fragile; broken capsids litter the EM grid with nucleic acid and protein, cluttering the image.

Solution: A cluttered grid is an indecipherable grid. Thus, the key to good single particle reconstruction is sample quality, says Conway.

In the case of virus particles, that means samples that are fresh, intact, and at high concentration-enough to get hundreds of particles per micrograph. To maximize his chances of success, Conway never stores samples in the fridge for long. "As soon as I get the sample I make frozen grids, because if I leave them in the fridge for a couple of days, just at 4°C, they will fall apart," he says. "Ideally, you look at them at the same time to check that you've made good ice with good sample coverage in it."

He might make three grids with slightly different blotting times, just to hedge his bets. He also takes a quick "negative-stain" transmission electron micrograph, to assess sample quality and concentration, which allows him to know in minutes if the sample should be diluted (cryoEM "is pretty much an all-day experiment," he explains).

"If you want to do yourself a favor," he concludes, "you would treat a sample fresh, and you'll go through the exercise of working out how good it is in negative stain and what kind of things you need to do to get a good cryo grid, and you'll do that quickly, before it has a chance to do anything untoward."

Vesicular wallpaper

<figcaption>An EM image of liposomes tethered to a streptavidin crystal, frozen in a thin layer of vitreous ice. Left: raw micrograph; Right: the same after the crystal periodicity is removed computationally; Inset: a model image of the large liposome, which contains three copies of a membrane protein, the BK potassium channel. Credit: Courtesy of Liguo Wang and Fred Sigworth</figcaption>
An EM image of liposomes tethered to a streptavidin crystal, frozen in a thin layer of vitreous ice. Left: raw micrograph; Right: the same after the crystal periodicity is removed computationally; Inset: a model image of the large liposome, which contains three copies of a membrane protein, the BK potassium channel. Credit: Courtesy of Liguo Wang and Fred Sigworth

Researcher: Fred Sigworth, Professor of Cellular and Molecular Physiology, Yale School of Medicine

Project: Determining the structure of the BK ("big potassium") ion channel protein by single particle reconstruction

Problem: Traditionally, membrane-bound proteins are removed from their membranes with detergent, but how well detergentsolubilized membrane proteins resemble their in-membrane counterparts is unclear. Sigworth wanted another approach.

Solution: Sigworth went for membrane reconstitution. He purifies the ion channel proteins with detergent, and then replaces that detergent with lipids to create vesicles, each of which contains one or two protein molecules. "So the idea now is, let's take pictures of these things and let's treat the protein particles as single particles and we'll do the same single particle reconstruction business, but this time it will be a protein in a membrane."

The approach allows Sigworth "to do a couple of cool things," including ion transport and membrane potential assays. But Sigworth also wanted a way to quality-assess his images. So, he "dopes" his liposomes with biotinylated lipid, which he uses to attach the micelles to the EM grid via a periodic 2D crystalline streptavidin "wallpaper" that was laid down on the grid previously.

"A crystallographer takes his or her crystal to the x-ray machine, and they immediately know whether they got good data or not," he explains. But with cryoEM, "you don't know how good your data are, because the individual particles are so hard to see, technically, you would say the signal-tonoise ratio is below unity."

"So," he concludes, "what we wanted was just a way to know whether or not we had a good image, and the idea is to have a crystal present that gives us that information." And, because the streptavidin wallpaper is periodic, it can be removed from the dataset computationally, leaving only the vesicle images for analysis.

Freezing for thickness

<figcaption>A 40 nm section of a cryo-electron tomogram of polyoma viruses within intact c127 host cells (red frame). Cell membranes can be seen best in an orthogonal orientation showing up as flexible lines. The filamentous structures to the left and at the bottom of the image are actin filaments. Insets show an average of virus particles that were extracted from a tomographic 3-D reconstruction. The smaller inset shows an inside view of the capsid. Credit: Courtesy of Andreas Hoenger and Mary Morphew</figcaption>
A 40 nm section of a cryo-electron tomogram of polyoma viruses within intact c127 host cells (red frame). Cell membranes can be seen best in an orthogonal orientation showing up as flexible lines. The filamentous structures to the left and at the bottom of the image are actin filaments. Insets show an average of virus particles that were extracted from a tomographic 3-D reconstruction. The smaller inset shows an inside view of the capsid. Credit: Courtesy of Andreas Hoenger and Mary Morphew

Researcher: Andreas Hoenger, Associate Professor of Molecular, Cellular, and Developmental Biology, University of Colorado at Boulder

Project: Studying microtubule dynamics and associated proteins in structures such as mitotic spindles using 3D cryo-electron tomography

Problem: With thicknesses on the order of microns, cells are too thick for direct cryoEM imaging (which caps out at 300 nm or so). How then to visualize proteins in situ?

Solution: Standard freezing methods are too slow for thick samples, so Hoenger uses high-pressure freezing. Basically, the sample is subjected to "a couple hundred bars" of pressure in the presence of liquid nitrogen (because the sample tends to heat up when the pressure increases). Then the pressure is rapidly released. "That [release] gives you a flash freezing effect, which goes through the entire tissue," Hoenger says. "Not only from the outside but actually within the tissue itself."

Once the sample is frozen in this manner, it is then sectioned (vitrified sectioning), just as in a standard pathology lab, and imaged from a variety of tilt angles to produce a 3D tomogram.

Cells present unique EM challenges: Unlike in single-particle work, cells are chock-full of biomolecules that are irrelevant to the experiment. Microtubules are easily recognized in a micrograph; how can you tell what proteins are associated with it? "This is a big issue," Hoenger says. He and his colleagues are developing methods to modify proteins so that they will chelate heavy metals-something electron- dense. It's the equivalent of GFP -tagging in fluorescence microscopy, he says. "We try to get a high-density signal so that we have a dense blob there, which we can say, okay, that must be this protein."

Stretch & stick

<figcaption>3D reconstruction of calcium release channel, top view (above) and side view (below). Credit: Courtesy of Wah Chiu, Irina Serysheva, Steven Ludtke, and Susan Hamilton</figcaption>
3D reconstruction of calcium release channel, top view (above) and side view (below). Credit: Courtesy of Wah Chiu, Irina Serysheva, Steven Ludtke, and Susan Hamilton

Researcher: Irina Serysheva, Associate Professor, Department of Biochemistry and Molecular Biology, The University of Texas Medical School, Houston

Project: Single particle reconstruction of the ryanodine receptor (RyR), a calcium release channel in muscle cells

Problem: Serysheva had two basic problems: getting her detergent-solubized proteins to stick to the grid, and doing so in such a way that the particles would not have a preferred orientation, but would orient randomly. RyR has four-fold symmetry, and to solve its structure, both side and top-down views of the protein must be present.

Solution: Servsheva chose to tweak her carbon support film's surface chemistry. Carbon supports can be continuous or "holey," like Swiss cheese. In earlier studies she used a continuous carbon support for adhering protein particles to the EM grid, but RyR particles tended to deposit with preferred orientations on that surface. When vitrified over a holey carbon grid, protein particles orient randomly in a thin film of aqueous solution that spans the holes in the film.

But she also needed to tweak the surface's hydrophobicity, to get the membrane proteins to stick at all (because they are solubilized in detergent, which exposes a hydrophilic face to the grid). One approach is called "glow discharge," in which Serysheva applies an electrical current to the grid. Another strategy: wash the grids in organic solvents before using them.

Serysheva says long experience provides her the expertise to look at the ice she gets from a freezing run and then know how to proceed to optimize vitrification. Given a new membrane protein, she says she would opt for holey film, glow discharge it ("because usually commercial grids are very hydrophobic"), and then freeze. "If it doesn't work, I will put continuous carbon and see what's going to happen," she says.

Tools of the Trade

1. The Vitrobot

Though plunge freezing can be done manually, many researchers have migrated to an automated freezing instrument called a Vitrobot (vitrification robot), which controls for factors that make manual plunge freezing so difficult, including humidity and the force and duration of blotting.

"Whoever starts working with me, I encourage everyone, move on and just use Vitrobot," says Serysheva, "because the learning curve for manual freezing usually is very steep." Serysheva has two Vitrobots from FEI, and is preparing to get a third. Each costs about $70,000.

2. High-Pressure Freezing

Daniela Nicastro, Assistant Professor of Biology at Brandeis University, uses high-pressure freezing for her research into cellular structures in situ. She recently purchased a BalTech HPM-100 (now sold by Leica), which costs about $210,000, she says (Leica's alternative, the EM PACT2, costs about the same, she says). Both systems feature a "rapid transfer unit" that can quickly shunt a sample from a light microscope to the freezing apparatus, which allows the researcher to correlate the light and electron microscopic information of the same cells.

3. Freeze Substitution

Between freezing and sectioning, says Nicastro, many researchers perform "freeze substitution," in which water in the sample is slowly replaced with organic solvents like acetone at around -80°C. The addition of fixatives and contrasting agents (e.g., osmium tetroxide and uranyl acetate), as well as the resin embedding are also performed at low temperatures.

"By doing this at really low temperatures everything moves much slower and you don't get the same structural distortions and extractions of your specimen," she says. "So it's a much softer way of preparing your specimen for EM." Leica's EM AFS2 automated freeze substitution device costs about $35,000.

4. Cryoultramicrotome

To cut sections, Nicastro uses a cryoultramicrotome, essentially a souped-up deli slicer. The specimen "block" sits in a box above a pool of liquid nitrogen, whose vapor cools the sample and diamond knife to temperatures below -130°C. Nicastro then cuts "really thin slivers of somewhere between 50 to 300 nanometers thickness from the block face," she says.

The technique is difficult and prone to artifacts, Nicastro cautions, "because the specimen is very brittle and there's no real fluid at these low temperatures that will have the same properties that water has (e.g., surface tension) helping with the floating of the sections and working against the compression that you have through the sectioning."

Since the cryo-diamond knife is dry, the manufacturer specially treats the surface to make the sections slide better. Nicastro uses a little probe, like a toothpick with a thin hair attached, to catch the section as it emerges from the cryoblock face. "People are actually using things like eyelashes and Dalmatian hair tips, it's really sophisticated, like voodoo sometimes," she says. "It really is an art." The Leica cryo-ultramicrotome costs about $110,000.