What is cryo-EM?
Conventional transmission electron microscopy poses two problems for biological samples. The column runs under high vacuum, where water evaporates, and the electron beam breaks chemical bonds. Staining and drying get around the first problem, but they do not preserve a protein as it exists in solution.
Cryo electron microscopy takes another route. The sample stays hydrated, frozen so quickly that the water turns into a glass, and it is held at around -196 °C (77 K), the temperature of liquid nitrogen. Cold reduces beam damage and the frozen water survives the vacuum. The price is a very faint image, which is why the method depends so heavily on averaging and computation.
In practice, "cryo-EM" covers a family of techniques. The most widely used is single-particle analysis (SPA), which is what most people mean by the term. Cryo-electron tomography (cryo-ET) is the second main branch.
Vitrification: why the ice must not crystallise
Ordinary freezing would ruin the sample, because ice crystals damage biomolecules and interfere with the electron beam. Vitrification cools water so fast that its molecules have no time to arrange into crystals, leaving an amorphous solid that acts as a uniform background.
Jacques Dubochet's group found that liquid nitrogen alone did not work. Success came when they used ethane cooled by liquid nitrogen, and the same principle survives in today's plunge freezers. A small drop of sample goes on a grid, blotting leaves a thin film, and the grid is plunged into liquid ethane or an ethane-propane mixture.
What makes a good grid
- Ice thickness: thin enough for good contrast, thick enough to hold the particles.
- Particle distribution: an even spread with random orientations, so that every view of the molecule is captured.
- Sample quality: a homogeneous sample. Mixtures and flexible complexes complicate the analysis.
- Concentration: much higher than for negative stain, because blotting removes many molecules.
The air-water interface is a notorious trouble spot. One review estimates that about 90% of particles on a typical grid adsorb to it, which can cause preferred orientation, where the same views appear over and over. Compared with the fixation and drying steps in SEM sample preparation, the chemistry here is minimal, but the process still needs careful, repeated optimisation.
Cryo-EM workflow: from purified protein to 3D map
A single-particle project moves through a fairly standard pipeline. Software such as RELION and cryoSPARC handles the computational stages.
- Prepare the sample. Purify and stabilise the target, and reduce compositional and conformational variety where possible.
- Screen by negative stain. A quick check shows whether the particles are intact and homogeneous.
- Vitrify. Blot and plunge-freeze grids, adjusting conditions to improve ice and particle distribution.
- Screen under cryo conditions. Find out which grids and areas are worth collecting from.
- Collect data. Record movies rather than single frames, spreading the dose across many frames.
- Correct motion and weight the dose. Align movie frames to undo motion, and weight them to account for accumulating damage.
- Estimate the CTF. Measure the defocus and astigmatism of each image so that the contrast transfer function can be corrected.
- Pick particles and classify in 2D. Locate the particles, then sort them into classes and discard junk.
- Reconstruct in 3D. Determine the orientation of each particle and combine the images into a map, separating different states if necessary.
- Validate and build a model. Assess resolution, then fit an atomic model into the density.
Why the images are so noisy
Dose has to stay low because the beam destroys what it images. For single exposures, the dose has traditionally been kept below about 20 electrons per square ångström, while movies spread a larger total dose across many frames. Vitrified specimens also have very low contrast, so images are taken slightly out of focus to make particles visible.
The result is that a single particle is barely recognisable. Averaging thousands of images of identical particles lifts the signal out of the noise, which is the idea at the heart of Joachim Frank's work.
Resolution values should be read with care. The usual measure is the Fourier shell correlation between two independently refined half-maps, and the authors of a widely cited primer advise treating a reported figure as a broad guideline rather than a firm number. Inspect the map itself as well.
The resolution revolution and the Nobel Prize
The 2017 Nobel Prize in Chemistry went to Jacques Dubochet, Joachim Frank and Richard Henderson for developing cryo-EM for high-resolution structure determination of biomolecules in solution. Each laureate solved a different part of the problem:
- Richard Henderson showed what electrons could do. With Nigel Unwin he produced a rough 3D model of bacteriorhodopsin in 1975 at 7 Å, and in 1990 he presented its structure at atomic resolution.
- Joachim Frank developed the image-processing methods that align and average noisy images of many identical molecules. His algorithms were complete in 1981.
- Jacques Dubochet vitrified water in the early 1980s, and in 1984 published the first images of viruses in vitrified water.
For years most maps showed little more than shapeless blobs. The change came in the early 2010s, when direct electron detectors and better algorithms arrived. This "resolution revolution" pushed structures towards the range where amino acid positions can be resolved. As of 2020, most single-particle structures were at 3-4 Å, and the best had reached 1.22 Å. The speed of the method also became clear when researchers used it to visualise the Zika virus within a few months.
Cryo-EM vs X-ray crystallography
X-ray crystallography remains the larger source of protein structures, but the balance is shifting. One 2026 analysis counted roughly 7,200 cryo-EM depositions in the Protein Data Bank in 2025, against about 10,000 for X-ray methods. Another expects the annual totals to level out at the end of 2026 or during 2027. Cryo-EM has already moved ahead for membrane proteins.
| Feature | Cryo-EM (single-particle) | X-ray crystallography |
|---|---|---|
| Crystals needed | No | Yes, and growing them can take months or years |
| Typical sample | Under 10-100 µL at 0.5-5 mg/mL | 100-200 µL at 5-40 mg/mL |
| Sample state | Near-native, frozen from solution | Packed in a crystal lattice |
| Typical resolution | Often around 3-4 Å, better for favourable samples | Median near 2 Å, depends on crystal quality |
| Size range | Best for large complexes; difficult below about 50 kDa | Not limited in the same way at the small end |
| Flexible or mixed samples | Different states can be sorted computationally | Crystallisation favours a single packed form |
The two methods are complementary. Small proteins remain hard for cryo-EM because their signal is too weak, although haemoglobin at 64 kDa has been resolved and scaffolds such as nanobodies or antibody fragments can increase the effective size of a target. Instrument access is the other practical limit: high-end microscopes are expensive, so many groups work through shared facilities.
Cryo-EM vs cryo-ET
Single-particle analysis looks at purified copies of one molecule. Cryo-electron tomography images one specific object, such as a slice of a cell, from many tilt angles and computes a 3D volume. Cryo-ET trades resolution for context, which makes it useful for seeing molecules where they actually work.
| Question | Single-particle cryo-EM | Cryo-ET |
|---|---|---|
| What is imaged? | Many copies of a purified molecule | A unique object, such as a cell region |
| How is 3D obtained? | Combining particles in random orientations | Tilting the same area through many angles |
| Typical resolution | Near-atomic | Roughly 1-4 nm, better with subtomogram averaging |
| Main strength | High-resolution molecular structure | Structure in its cellular context |
Cells are usually too thick for electrons to pass through, so they are thinned into lamellae by cryo-focused ion beam milling. Fluorescence can guide the milling to the right spot, and lamella thickness strongly affects the detail that subtomogram averaging can reach.
Where cryo-EM fits in the microscopy toolbox
Choose single-particle cryo-EM when you have a purified complex that resists crystallisation, or when you want to compare several functional states. Choose cryo-ET when location and context matter more than atomic detail. For small, well-behaved proteins that crystallise readily, X-ray methods may still be the quicker path. Success in every case depends on sample quality, so time spent on purification and grid screening is rarely wasted.