REVIEW doi:10.

1038/nature19948

Unravelling biological macromolecules
with cryo-electron microscopy
Rafael Fernandez-Leiro1* & Sjors H. W. Scheres1*

Knowledge of the three-dimensional structures of proteins and other biological macromolecules often aids understand-
ing of how they perform complicated tasks in the cell. Because many such tasks involve the cleavage or formation of
chemical bonds, structural characterization at the atomic level is most useful. Developments in the electron microscopy
of frozen hydrated samples (cryo-electron microscopy) are providing unprecedented opportunities for the structural
characterization of biological macromolecules. This is resulting in a wave of information about processes in the cell that
were impossible to characterize with existing techniques in structural biology.

B
iological macromolecules adopt complicated three-dimen- study the structures of proteins and how rapid progress in structure
sional (3D) structures that are crucial to their function. Some determination through electron microscopy in the past few years
macromolecules, such as enzymes, act alone to provide chemi- has been heralded as the start of a revolution in structural biology3.
cal environments that favour the catalysis of specific chemical reac- We also highlight how the unique characteristics of this technique
tions. Other macromolecules form larger complexes with protein have already changed structural biology and identify opportuni-
partners, nucleic acids, lipids or sugar molecules. Many such com- ties through which electron microscopy will continue to transform
plexes perform their functions through the relative movements of understanding of how macromolecules perform intricate tasks in
individual parts, in a way that resembles how man-made machines the cell.
work1. A fundamental goal of modern biology is to understand how
these complicated structures perform their tasks. Cryo-electron microscopy
Macromolecular complexes are too small to be seen with visible Because electrons are scattered by molecules in the air, electron
light. Photons with wavelengths that are short enough to visualize microscopes must be operated in a high vacuum. This poses a prob-
details at the atomic level are found in the X-ray region of the elec- lem when studying biological samples, most of which occur naturally
tromagnetic spectrum. X-rays interact weakly with biological matter, in an aqueous environment. Biological structures are also sensitive
which makes it difficult to use them to study individual protein com- to radiation damage. For each electron that contributes to the forma-
plexes. But when many copies of the same protein are arranged into a tion of an image, there will be three electrons that deposit energy in
3D crystal, information about the atomic structure of the protein can the sample. This energy causes the cleavage of chemical bonds and
be obtained through X-ray diffraction experiments. This technique, ultimately destroys the structures of interest. By keeping samples at
known as X-ray crystallography, has been the most important tool cryogenic temperatures, cryo-electron microscopy (cryo-EM) ena-
in structural biology for more than two decades (Fig. 1). Another bles their preservation in a high vacuum and provides them with
technique that is used to characterize the structures of proteins is some protection against the effects of radiation damage4.
nuclear magnetic resonance (NMR) spectroscopy, which measures The placement of a cryo-EM sample inside an electron microscope
distance-dependent interactions between atoms. NMR can be used is shown in Box 1. To prepare the sample, a few microlitres of a
to infer the structure of relatively small proteins, and it provides purified protein solution is applied to a metal (usually copper) grid,
unique information about the dynamics of proteins and their inter- on top of which lies a thin film of amorphous carbon that contains
actions with other molecules. holes. After any excess liquid is blotted with filter paper, the grid
Electrons can also be used to look at protein structures. Proteins is plunged into liquid ethane5. Ideally, this results in the formation
scatter electrons about ten-thousand times more strongly than they of a thin layer of non-crystalline or vitreous ice, in which copies of
do X-rays, and electrons can be accelerated by electric fields of sev- the protein are deposited in a range of orientations. Images that are
eral hundreds of thousands of volts to wavelengths that are much captured through the holes in the carbon film contain two-dimen-
shorter than the distances between the atoms in protein structures. sional (2D) projections of individual protein complexes, which are
Moreover, the electric charge of electrons makes it relatively easy called particles. Projections of particles in various orientations pro-
to focus them with electromagnetic lenses. Microscopes can there- vide complementary information about the underlying 3D object.
fore be built that use electrons to make images with atomic-level Numerous 2D projections can therefore be combined into a sin-
detail. The contributions of electron microscopy to structural biol- gle 3D reconstruction, provided that their relative orientations are
ogy have been modest in comparison with X-ray crystallography known. Unfortunately, this information is lost in the experiment
and NMR, but present trends indicate that this is changing (Fig. 1). because the individual particles tumble randomly in solution before
In 2016, the one-thousandth atomic structure derived from elec- the sample is vitrified.
tron microscopy images2 was entered into the Protein Data Bank The relative orientations of individual particles can still be deter-
(PDB), the main repository for protein structures. In this Review, mined a posteriori by processing the 2D projection images using
we describe how images from an electron microscope can be used to a computer6, a process known as single-particle analysis. Images
1
MRC Laboratory of Molecular Biology, Cambridge Biomedical Campus, Cambridge CB2 0QH, UK. *These authors contributed equally to this work.

1 5 S E P T E M B E R 2 0 1 6 | VO L 5 3 7 | NAT U R E | 3 3 9
©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.

these methods are more robust to noise than collected with single-particle cryo-EM are of low contrast because those that were already in common use27.000 capture. which improved the efficiency of detection to around half 4. And in acquisition procedures that were developed originally for CCDs the past 5 years.000 200 results in images that are too noisy to determine reliably the orienta- tion of particles. mixture problem was provided first by 3D maximum-likelihood classification algorithms25. experimental pro. numerous structures previously blurred cryo-EM structures. Complexes which predicted that structure determination should be possible with molecular weights of less than 200 kDa have become feasible to atomic resolution for protein complexes with molecular weights targets for structure determination33. from which the best par- scope of cryo-EM-based structure determination. By 2010. These successes built because crucial parameters no longer needed to be tuned. Many of the newly determined structures represent proteins tion18. Because they incorporate a statistical description of the data. By 2008.3% 15. the develop- ment of cameras containing a digital charge-coupled device (CCD)17 Opportunities for structure determination opened the path to higher throughput and automated data acquisi. separate particles from distinct structural states was also improved. The percentage of membrane protein structures discovered in 2015 using structures therefore need to be separated. that are naturally embedded in membranes (Fig. The molecular machines adopt a range of conformations in solution. Images were originally surpass 2 Å in favourable cases34. Whereas the presence of could be used to determine small differences in image intensities. INSIGHT REVIEW a 10. This means that samples prepared for cryo-EM often between 1975 and 2015. much like the burst mode of contemporary photographic cameras. rate of less than one-fifth of the incoming electrons19. and the achievable limits of as low as 100 kDa (ref. . which blurs the resulting images. Ribosome structures could also be ian approach made image processing more accessible to non-experts calculated to a resolution of 6 Å (refs 10–12). The EM field.23. valu- However. 16).000 X-ray crystallography NMR Electron microscopy cryo-EM-based structure determination exactly at its bottleneck: 250 the low number of electrons that is used to limit radiation damage 8. Alternative approaches to the proteins scatter electrons only about 30% more than does the classification soon followed. The Structures that heralded this era include the coenzyme F420-reduc- presence of internal symmetry in a complex also helps to improve ing hydrogenase29. the modernized electronics that surround these direct electron detectors facilitated fast image 2. The innovative chips were suffi- 50 ciently resistant to radiation damage to enable the direct detection of 0 electrons. to reduce the effects of radiation damage.  a. in 2013. The difference could be explained in resolution now extend to below 3 Å (refs 2. Moreover. Fast 0 image capture also addressed a problem that is associated with frozen 1980 1985 1990 1995 2000 2005 2010 2015 hydrated samples: energy released by the incoming electrons causes Year movement inside the ice layer. CCDs are even less efficient. which becomes the main bottleneck. 30) icosahedral virus capsids have long been at the forefront of the cryo. the number of able insight into protein dynamics could now be obtained from a electrons used must be carefully limited. recorded on photographic film.000 digital electron detectors by 2012. the automated data- cedures and image processing in the preceding 40 years13–15. resulting in extremely noisy single experiment28. estimated from the data instead32. NMR (red) and electron microscopy (green) dom perfect. 3 4 0 | NAT U R E | 5 3 7 | 1 5 S E P T E M B E R 2 0 1 6 © 2 0 1 6 M a c m i l l a n P u b l i s h e r s L i m i t e d . images. When such mixtures are subjected electron microscopy between 1995 and 2015 is also shown (inset). which to single-particle analysis. tures to atomic resolution for a wide range of samples. otides give rise to images with higher signal-to-noise ratios. Photographic film detects only about one-third of the incom. a number of developments have further changed the facilitated the recording of large datasets. This affects Detergents also make crystallization notoriously difficult to achieve. As a result. but in the early 2000s.24. number of structures recorded in the PDB. The ability to Larger protein complexes and complexes that contain oligonucle.5% 2. A general solution to the each of the three techniques. A l l r i g h t s r e s e r v e d .34. the mammalian ion channel transient receptor the resolution. Such pro- ing electrons. b. Moreover. and the large subunit of the mitochondrial ribosome in yeast31. Number of structures 150 Three companies had produced prototypes of a new generation of 100 6. the 2D projections of a number of 3D highlights the recent growth in structure determination using this method. as determined by the techniques and the purification or formation of these protein complexes is sel- of X-ray crystallography (blue). because an extra teins are difficult to purify in solution because their hydrophobic conversion from electrons to photons leads to an overall detection membrane-spanning domains must be stabilized with detergents. p a r t o f S p r i n g e r N a t u r e . ticles could then be selected using image classification. b 3. This would not be a problem if many electrons the mixture problem into an opportunity. Such noise impedes the accurate assignment of orientation When digital electron detectors became available commercially to the particles. which Membrane protein facilitated structure determination to an unprecedented resolution structures in 2015 from far fewer data than before21. Meanwhile. the structures of several viruses had been solved extension of maximum-likelihood methods into an empirical Bayes- to near-atomic resolution7–9. The unprec- edented image quality that arose from these detectors enabled Progress in cryo-EM orientations to be assigned with greater accuracy.8% Movies recorded during exposure of the sample to electrons could be processed to effectively remove the blurring effects22. which resulted in methods that turned surrounding ice. This explains why the structures of ribosomes and potential cation channel subfamily V member 1 (TRPV1) (ref. Many macro- Figure 1 | Growth in structural biology over the past 40 years. the quality of available cryo-EM protein structures did in the past 3 years many research groups have solved cryo-EM struc- not support theoretical considerations about radiation damage. 2). Counting individual electron events on the fastest camera available led to an even better efficiency of electron detection21. however. 35–37) and might even part by the inefficient detection of electrons.26. the cryo-EM field was poised for a revolution. another main impediment to high-resolution cryo- X-ray crystallography NMR Electron microscopy EM-based structure determination had been solved. they were on important developments in instrumentation. The number of structures that were determined by contain a variety of structures.000 1995 2000 2005 2010 2015 of the incoming electrons20.

. the structure of responsible for the burning sensation that chilli peppers impart and the human γ-secretase complex was solved despite it containing at it is an important drug target for pain. these structures have yielded a wealth of crystallization. The grid is stored in of a structure determined through cryo-EM. 1). 30). the active component of chilli peppers. This protein is type proteins can be used for cryo-EM. Instead. the structures of many other medically relevant ion biology (Fig. fully glycosylated wild- the new cryo-EM technology was TRPV1 (ref. The first membrane-protein structure to be solved using protein-engineering approaches46. human γ-secretase is the smallest cryo-EM struc- determined41.1 (ref. This membrane- in an empty state that had been solubilized in amphipols. After the initial TRPV1 structure recognizes receptors on the surface of immune cells and mediates the BOX 1 Cryo-EM structure determination Protein structure determination through cryo-EM involves several liquid nitrogen until it is transferred to the transmission electron stages: sample grid preparation. the heavily glycosylated structure of the human immunodeficiency zation of lipid substrates and provided more detailed mechanistic virus type 1 (HIV-1) envelope glycoprotein trimer. scaffold of an amphipathic protein belt39 or the saposin–lipoprotein flexible loops or sugars are typically removed using complicated system40. For example.5 Å. it is possible to image membrane proteins information on how cells regulate ion transport across membranes. The the same protein into a 3D reconstruction of the protein’s structure. A l l r i g h t s r e s e r v e d .). structure determination through cryo-EM does not require glycine receptor45. including the voltage-dependent calcium of all known small-molecule drugs bind to these proteins. A complex embedded protease generates amyloid-β peptides that aggregate in of TRPV1 with the spider toxin DkTx and a small molecule that is the brains of people with Alzheimer’s disease. With an ordered mass similar to capsaicin. which is unfortunate because approximately half channels were published. p a r t o f S p r i n g e r N a t u r e . which flash-freezes the sample The enzyme glutamate dehydrogenase34 is presented as an example and embeds the particles in vitreous ice. 43). Its structure was solved first least 11 sugar chains and a long disordered loop33. An improved TRPV1 structure that was determined ture to be determined at a resolution below 3. channel CaV1. How. membrane proteins are a blind spot for structural was determined. REVIEW INSIGHT Consequently. data collection and data processing microscope. Transmission electron microscope Filter Sample preparation paper Tweezers 90º Ethane Grid Sample application Blotting Plunge freezing 3 mm 80 µm Incident electron beam Ice-embedded proteins 20 nm 2 µm Grid Carbon film Carbon film Data processing and 3D reconstruction Detector Glutamate dehydrogenase 1 5 S E P T E M B E R 2 0 1 6 | VO L 5 3 7 | NAT U R E | 3 4 1 © 2 0 1 6 M a c m i l l a n P u b l i s h e r s L i m i t e d . was also of about 130 kDa. This protein insight into TRPV1 function42. Together. However. a sodium–potassium channel44 and the ever. directly that have been solubilized in detergents or amphipols38. or Another advantage of cryo-EM is that flexible regions of proteins stabilized in a lipid environment using nanodiscs formed within the do not impede structure determination. Two-dimensional images of proteins in various followed by 3D reconstruction (Box Fig. The data are processed to combine images of perforated carbon film and the excess liquid is blotted away. Another example is in the more natural environment of a nanodisc enabled the visuali. To facilitate crystallization. A few microlitres of orientations within the grid’s holes are then captured by the a purified protein solution are applied to a grid composed of a microscope’s detector. grid is plunged into liquid ethane.

shock protein 90 (ref. In almost all of these studies. that reveals how this mobile DNA element catalyses self-splicing in more than 50 high-resolution ribosome structures have been deter. in the development of HIV-1 vaccines47.62. flexible regions and glycosylation. In the past 2 years. TRPV1 in a complex with the spider toxin DxTx42. INSIGHT REVIEW a Extraction of membrane proteins b Examples of membrane proteins Glycosylation Long loop Cell membrane Photosystem II–LCHII supercomplex TRPV1–spider toxin Extraction Flexible region Human Cell N C γ-secretase Membrane protein Protein Detergent engineering • Removal of glycosylation. because wild-type proteins can be used and only micrograms of For example.73. 50) interactions with each other. 4). macromolecular machines make mTOR82. tion of the final map. 43). inside the capsid of double- receptors55–58.100. However. The crystallization of membrane proteins for X-ray diffraction. The structures of embedded in biological membranes must be extracted from the membrane glycosylated membrane proteins and very large membrane complexes can be using detergent solubilization techniques before their structures can be determined through cryo-EM. . photosystem II–light stranded RNA viruses93. the structures of vari. several glutamate even been studied in situ by cryo-EM. These all change conformation. including the ryanodine how DNA is transcribed into RNA88–92. of the particles in the initial dataset is selected for use in the calcula- the exosome72. tional freedom of such complexes in a single experiment (Fig.97. A l l r i g h t s r e s e r v e d . Most particles in cryo-EM datasets are unsuit- teasome75. mammalian complex I (ref.71. the dynein–dynactin complex79. and a ous large membrane-protein complexes that had resisted crystal. tion with the mitochondrial genome has affected their structures. A 3 4 2 | NAT U R E | 5 3 7 | 1 5 S E P T E M B E R 2 0 1 6 © 2 0 1 6 M a c m i l l a n P u b l i s h e r s L i m i t e d . and structures mined. For many structures.78. rearrange their subunits use of movements of parts relative to each other in their function. For example.  a. which has yielded a wealth of information on the control of that have aided our understanding of the mechanisms of type II and protein biosynthesis. it has proven challenging provide clues about immunity to Ebola in humans and how the virus to make or purify many large nuclear complexes.1 (ref.66. nuclear complexes include: a structure that provides a fresh under- thases61. the SNARE (soluble NSF attachment protein receptor) able in some way for high-resolution structure determination and complex77.94.80. number of structures of RNA polymerases have led to insight into lographic studies have been determined.76. vari- Complexes that have been purified from native sources may also ous cryo-EM-determined structures are starting to shed light on the be suitable for cryo-EM-based structure determination.48.83. Similarly. N C long loops and flexible N regions C Crystallization • Detergent optimized (Detergent exchange) Glycine X-ray diffraction Cryo-EM CaV1. the inositol triphosphate receptor54. entry of HIV-1. as exemplified by the structures that have emerged structure determination. Cryo-EM enables the study of these often short. a structure of the bacterial group II intron tion of soluble macromolecular machines (Fig. Examples of membrane protein structures determined. studies of both mammalian63. This is molecular details of some of the most fundamental processes of life. structures that show how the The possibility of determining the structures of large complexes retroviral recombination machinery engages with the host nucleo- from native sources has also had a huge impact on the characteriza. some to insert its DNA96. Other examples of cryo-EM studies of harvesting supercomplex (PSII–LHCII) (ref. potential of cryo-EM for studying highly dynamic complexes. image classification played a crucial Other macromolecular machines have also now become amenable to part in selecting structurally homogeneous subsets of particles for structural studies. human to remove long loops. the signallosome70. 3). 59) are shown as examples. and bind various partners and substrates during their assembly and Whereas dynamic complexes would need to be trapped in a single working cycles. the new image classification algo- lived conformations and interactions. 60) and ATP syn. only a small fraction of the inflammasome65. that were determined through cryo-EM: the photosystem II–light harvesting II based structure determination is difficult and often requires protein engineering supercomplex60. In a showcase of the fuses its membrane with that of the host. 59). the structures Both the proteins and the oligonucleotide substrates in these com- of Ebola virus glycoprotein GP1 bound to neutralizing antibodies49 plexes tend to be highly dynamic molecules that engage in transient and to the transporter protein Niemann–Pick C1 (NPC1) (ref. the anaphase-promoting complex73. 81) and the serine/threonine protein kinase much like man-made machines. p a r t o f S p r i n g e r N a t u r e . Moreover. b. the spliceosome67–69. state to facilitate crystallization. Proteins that are detergents or more natural environments such as nano-discs. conjunction with a small intron-encoded protein98. Consequently. standing of genetic recombination95. (CRISPR) systems99. DNA replication machinery structures have unveiled purified protein are required. the voltage-gated calcium channel CaV1. Consequently. Knowledge of the structure of the fully glycosylated Nuclear complexes that act on DNA and RNA form another group native protein has provided important information that might help of molecular machines that is difficult to analyse conventionally.64 type III clustered regularly interspaced short palindromic repeat and yeast31 mitochondrial ribosomes have revealed how coevolu. some of the molecular details of how genomes are copied84–87. cryo-EM γ-secretase33. the 26S pro. the glycine structure determination enables membrane proteins to be imaged directly in receptor45 and mammalian complex I (ref. RNA polymerases have receptor51–53.1 receptor Mammalian complex I Figure 2 | Membrane protein structural biology. the chaperone heat image classification enables only the best to be selected. which are difficult to isolate rithms offer the unique opportunity to visualize the full conforma- or stabilize biochemically.

there is an urgent need to develop afford. The future of cryo-EM The refinement of sample preparation methods provides another Several challenges must be overcome for cryo-EM-based structure important opportunity for enhancing cryo-EM-based structure determination to continue its transformative growth in structural determination. alternative solutions such as cloud com. microlitres of a purified sample of protein. generates contrast that is improved by up to an order of magnitude which acts as a molecular turbine. Such detectors could even be improved further. there is ample scope for improving the performance of microscope hardware. some of the highest resolution and smallest structures that are available were reconstructed from energy-filtered images. or by using films made of graphene or graphene oxide114.33. 1 5 S E P T E M B E R 2 0 1 6 | VO L 5 3 7 | NAT U R E | 3 4 3 © 2 0 1 6 M a c m i l l a n P u b l i s h e r s L i m i t e d . p a r t o f S p r i n g e r N a t u r e .1% of the sample of daltons. such as faster electronics and Hsp90–Cdc37–Cdk4 Transcription pre-initiation kinase complex complex improved chip design. Automated data acquisition on a high-end microscope can tigations that have pursued this direction include spraying picolitre- yield several terabytes of images every day. Purification of the sample on the grid itself. Noise in images can be decreased further through the use of energy filters. Of the three commercially available detectors. . which requires fast read-outs21. through the use of all in-phase. But the cheaper microscopes that are available at present. To avoid the need to buy and maintain costly high-performance U4/U6.0 μmol per litre.1–5. volume will remain on the grid after blotting. Similar methods can reach hundreds of thousands of computing core hours per dataset. As well as reducing costs and increasing the accessibility of high- end cryo-EM instruments. specifically adhered affinity tags116.103.  Examples of cryo-EM Optical devices known as phase plates are another promising structures are shown for the U4/U6. cryo-EM aim to reduce or stop this motion. are unable to generate electrons that are grid.107. in particular at higher resolutions. are needed to prepare a single cryo-EM often operate at 100–120 kV. could enable the production of detectors with efficiencies of up to 90%. to correct for beam-induced motion in samples. for example by replac- end microscopes. which tion of 0. requires access to an electron microscope on a daily or weekly basis. They produce a difference yeast67. but their cost is still on during this early period112. The use of more modern technology. they lack the coherence that is required to visualize par. However. at present only one is fast enough to enable counting. also be modified to boost cryo-EM outcomes. which con. These optical devices remove electrons that have lost part of their energy in the sample and can no longer contribute con- structively to the image. in which more electrons lose energy while passing through the sample. and only a fraction able screening microscopes with more coherent electron sources. and processing times can sized droplets of samples on cryo-EM grids117. For example. rectors might be helpful when higher resolutions of around 2 Å must be achieved110. can the machine from a single cryo-EM dataset61. and the human transcription pre-initiation complex91. counteract the effects of imperfections in the optical system of the sist of multiple ribosomes bound to a single molecule of mRNA. progress has been achieved by increasing the efficiency of electron detection from about 30% when film is used to 50% for direct electron detectors. of the grid’s surface is typically used for data acquisition. several expensive high-voltage machines are not needed for sample screen. Consequently. An immediate concern is the elevated cost of cryo-EM. This Cdk4 kinase complex81. Although followed by extensive image classification led to the production of not needed at present to reach resolutions of 3 Å or lower.111. This places a considerable constraint on and are accompanied by expensive maintenance contracts. in images at low resolution — of particular interest when imaging tion to synthesize molecules of ATP. which cost in excess of US$5 million fast to be corrected21. And because less than 0. At present. yield better images.34.29. How- ever. complexes that are too small to yield enough contrast with existing Image classification revealed the presence of three rotated states of optics109. To facilitate broad access to high. REVIEW INSIGHT striking example of this is the membrane-embedded ATP synthase. A l l r i g h t s r e s e r v e d .36. Inves- problem. These devices the imaging of actively translating human polysomes. Developments in sample preparation for the order of millions of dollars. spliceosome complex helicase puting105 and the implementation of image-processing algorithms on cheaper graphics processing units (GPUs) are being explored106. typically at a concentra- ing. large movements High-resolution structure determination is best performed using of particles during the early stages of electron exposure are often too 300 kV electron microscopes. the determination of structures at high resolution because most of scopes that operate at 200 kV are cheaper and can also be used to the high-resolution information is destroyed by radiation damage produce atomic-resolution structures102. microscope through complicated combinations of lenses.108. savings in The cost of storing and processing large volumes of data is also a sample volumes of multiple orders of magnitude are possible. which indicates that the removal of these electrons is also beneficial for thinner samples2. the energy currency of the cell. a further type of optical device. Although movie processing has progressed enough biology. Requirements that concern the quantity and purity of samples can which is not practical when using centralized facilities.U5 tri-snRNP spliceosomal complex in development for microscope hardware. Micro. Aberration correctors.U5 tri-snRNP CMG computing infrastructure. converting a proton flux into rota. another needs to lower the intensity of the electron beam to avoid flooding the chip with too many events. the human Hsp90–Cdc37– in the phase of the scattered and the unscattered electron waves. In principle. the optimization of samples gold113. Energy filters were originally considered to be most useful for thick samples such as whole cells. the eukaryotic replicative CMG helicase86. such cor- structural snapshots of the entire translation cycle101. can relax demands on sample ticles with molecular weights below several hundreds of thousands concentration and purity. Figure 3 | Soluble macromolecular machines. An important aspect of efficient electron detection is the ability to count individual elec- trons as they hit the detector.115. regional or national cryo-EM facilities are quickly ing both the copper grid and the amorphous holey carbon film with gaining in popularity 104. In another example.

R.120. et al. W. Rev. 156–163 (2014). 1130–1134 (2008). R. Henderson.. micrographs of 40S ribosomal subunits. 415–419 (2008). Backbone structure of the infectious ε15 virus capsid revealed by electron cryomicroscopy. 2. 755–765 (2009). for atomic resolution microscopy of unstained biological molecules. hydrated protein ponents are mixed together in a precisely timed manner118. Science 326.3 Å resolution cryo-EM structure of human p97 and mechanism of allosteric inhibition. at present. is now on the horizon. Freeman. Applications of slow-scan CCD cameras in around 100 kDa will widen the applicability of the technique to many transmission electron microscopy. A. J. Sci. Together with improved image-processing algorithms. et al. 3... Received 11 April. 2. . Nature 451. 2. With the potential to achieve resolutions well beyond 3 Å (ref. W. are A comparison of the three commercially available direct electron detectors. be necessary to calculate structures of higher resolution. Natl Acad. Frank.. The detailed structural information about individual complexes that can be obtained through single-par- ticle analysis needs to be understood in the context of the complexes’ cellular environment. in Advances in Imaging and Electron Physics Vol. based drug design will become routine for targets that. extremely difficult to characterize using alternative techniques. Frozen aqueous suspensions. A. ■ be imaged directly using cryo-EM techniques. 34) 20. Ultramicroscopy receptors should become amenable to structure determination by cryo. Generalized single-particle cryo-EM—a historical perspective. Hawkes. 871–875 (2016). Kühlbrandt. Seidelt. possibly in combination with sample preparation on the nanolitre scale. E. complexity of highly dynamic molecular machines67. Alternative procedures that 10. A. R. cryo-electron tomography can be used to build 3D reconstructions of frozen cells by taking multiple images of the same sample at different tilt angles inside the electron microscope124. EM. Proc.. USA 105. Electron diffraction of frozen. G-protein-coupled receptors are 18. 17. These techniques could bridge the gap between biophysics dynamics. generation of molecular biologists. more drug targets.. the structural characterization at the molecular (or even the atomic) level of extremely large and flexible machines such as the nuclear pore complex122. USA 106. W. Verschoor. In complex with electron micrographs a day. Jin. Sci.-J. & Mooney. 1867–1872 (2008). In particular. H. electrons and X-rays and of more complicated mixtures than is possible at present. B. 1412–1415 (2009). Potter. 1443–1444 (2014). & Henderson. R. Excitingly. 3 4 4 | NAT U R E | 5 3 7 | 1 5 S E P T E M B E R 2 0 1 6 © 2 0 1 6 M a c m i l l a n P u b l i s h e r s L i m i t e d . images will yield higher resolution structures of smaller complexes 16. 1036–1037 (1974). will also expand the applicabil- ity of cryo-EM to complexes with low abundance in the cell. Faruqi. issues and future perspective.-C. be used in time-resolved studies in which multiple protein com. providing unprecedented insight into the function of large macromolecular machines. For example. Schuette. EMBO J. & Boublik. blue and green objects) can at a molecular level. Natl Acad. Lepault. 113–137 (Elsevier. Henderson. B. Comparison of optimal for complexes that are bound to drug candidates121. When crystals. G. INSIGHT REVIEW Macromolecular complex Cryo-EM The ability to obtain structures from small subsets of the data in solution (V-ATPase) imaging through image processing. cryo-EM structure. Biophys. J. Villa. even transient complexes and intermedi- ate conformational states can be studied. e13 (2016). E. J. and will lead to the collection of images with higher signal-to-noise Microscopy 65. R. Leginon: a system for fully automated acquisition of 1000 sought-after targets for the pharmaceutical industry. this could enable reactions 6. L. 1126–1143 (2009). Single particle electron cryomicroscopy: trends.. multicomponent mixtures and to identify less common 9. 1063–1068 (2009). Banerjee. The tomographic approaches will benefit from the same developments in hardware and software that are driving forwards cryo-EM-based single-particle Figure 4 | Image classification enables the study of macromolecular analysis. 171–193 (1995). O. K. combined with modern microfluidic systems and the ability to tackle 5. The potential and limitations of neutrons. tial to enlarge the scope of cryo-EM-based single-particle analysis 14. accepted 15 July 2016. K. Ultramicroscopy 77. et al. Rev. Cheng. Science 214. superior image-formation models might 7. imaging with electrons could also provide unique opportunities in this task. GTPase activation of elongation factor EF-Tu by the ribosome during decoding. 3. Agard. M. domains still represents a challenge. Ribosome-induced changes in elongation factor Tu conformation describe the ensemble of structures in a dataset (instead of dividing control GTP hydrolysis. X. R. 28. et al. such 15. G. S. et al. Science 343. 4. Glaeser. McMullan. Improved 8.. & Henderson. P. Q. S. Krivanek. Ultramicroscopy 10.88 Å structure of cytoplasmic polyhedrosis virus by image-processing algorithms are also needed to deal with com. the data into subsets) have the potential to reveal the conformational 11. ratios. C.. 185 (ed. Dubochet. p a r t o f S p r i n g e r N a t u r e .) Ch. et al. Ultramicroscopy 49. the presence of numerous continuously flexing single-particle reconstruction.. 153–161 (1999). Ultramicroscopy 147. P. Q. The resolution revolution. as well as the characterization of complexes 3D classification involved in the organization of chromatin123.. Such classification therefore enables the functional cycles of dynamic molecular machines such as the eukaryotic V-ATPase61 to be characterized 1. S. R. Clare. 3–8 (2016). M. McMullan. 95–108 (1993). M. J. Yu. Zhang. 28. 1353–1355 (1981). & Glaeser. For example. Prospects 13. A l l r i g h t s r e s e r v e d . Chang. S. A. 49. Frank. The cell is densely packed State 1 State 2 State 3 with many molecules that interact to form highly intricate networks that are finely tuned to keep it alive. Three-dimensional image classification is then used to obtain structures for each state of the complex. Yet macromolecules do not act in isolation. cryo-electron microscopy. Detective quantum conformation-specific antibodies or heterotrimeric G proteins. Science 186. & Faruqi. Structural insight into nascent polypeptide chain-mediated translational stalling. 291–294 (1998). Z. Y. Chen. & Zhou. 109. A. The cell as a collection of protein machines: preparing the next from a single experiment. For example. performance at 300keV of three direct electron detectors for use in low dose electron microscopy. L. Cell 92. Proc. Alberts. Nature 453. D. Near-atomic resolution using electron cryomicroscopy and states. W. 2014). & Subramaniam. In software development. these efficiency of electron area detectors in electron microscopy. J. X. R. et al. plicated. all were shown to perform better than photographic film. R. Computer averaging of electron to be followed on a millisecond timescale119. 12. but the resolution of the substructures that are repeated in the cell can be increased by applying averaging methods125. D. 55–62 (1982). Each of the developments described in this Review has the poten. mixtures through image classification. R. Radiation damage is still a major limita- tion. Jiang. Science 351. Taylor. 19.  A mixture containing a macromolecular complex in distinct and cell biology and provide a road map for understanding of the cell conformations or compositions (shown as red. & McDowall. The ability to visualize proteins complexes with molecular weights of Biophys. Vinothkumar. Using time-resolved methods. J.

Cheng. Nature K+ channel. Structure of the human 26S proteasome at a resolution of using cryo-electron microscopy. Science 343. Near-atomic resolution visualization of human transcription (2015). J. Cryo-EM elucidated by electron cryo-microscopy. eLife 2. Nature 534. A. 63–68 (2016). Potter.. 328–334 (2014). H. 80. et al. Mills.. M. Greber. 1043–1048 (2016). et al. 1542–1547 (2016). & Grigorieff.8 Å phosphorylation. 303–308 (2015). E. G. I.. T. A. 63. 94. D. 404–409 (2015). et al. J. Ohashi. 328–339 (1998). Horizontal membrane-intrinsic α-helices in the stator classification of cryo-EM images using FREALIGN. Althoff. 336–341 (2015). J.1 complex. 39. et al.. J. Agafonov.. A. et al. E. 30. Science 350. Nature 522. Brown. activation and the cullin–RING ubiquitin ligase deneddylation cycle. T. 198–203 (2015). 1485–1489 (2014). Architecture and Proc. coli replicative DNA polymerase reveal its dynamic 46. & Cheng. F. cholesterol transfer and Ebola infection. et al. Struct. 89. 1698–1707 (2016). Y. Wan.. aad3873 (2016). Bernecky. et al. Allegretti. The complete structure of the 55S mammalian mitochondrial 28. et al. Scheres.8 Å) of a single-particle reconstruction. Molecular architecture of the human U4/U6. Science 348. G. et al. 353–358 (2016).. et al. p a r t o f S p r i n g e r N a t u r e . 43. S. Lyumkis. at (2015). et al. D. 1147–1151 (2015). eLife 5. 1823–1828 (2012). J. Yan. Nature 517. 70. Natl Acad. Kühlbrandt. Science 350. Julius. S. 11016 (2016). H. 23. Structural and kinetic analysis of the COP9–signalosome 32. Lyumkis. Mosadeghi. 584–590 glutamate receptors. Leitner. C. Science 349. 359–365 (2016). e01963 (2014). H.8 Å structure of the U4/U6. et al. Structure and organization of heteromeric AMPA-type near-atomic-resolution single-particle cryo-EM. S. Structure of the yeast mitochondrial large ribosomal subunit. Structure 20. 237–240 (2015). et al. X. Allegretti. et al... 298–302 (2016). 87. G. R. Y. Lü. F. 1 5 S E P T E M B E R 2 0 1 6 | VO L 5 3 7 | NAT U R E | 3 4 5 © 2 0 1 6 M a c m i l l a n P u b l i s h e r s L i m i t e d . Zhu. & Kühlbrandt. Liao. Fernandez. Electron cryomicroscopy observation of e00461 (2013).. 53. cryo-EM using a 2. W. Amunts. Z. N. Mol.-C. J. M. H. D. 64. D. A. cleaved HIV-1 envelope trimer. et al. G. Arrangement of electron Res. Structure of spinach photosystem II–LHCII supercomplex at 3. Wei. Cell 149. et al. et al. McMullan. 73. Activation of NMDA receptors and the mechanism of inhibition 93. et al. Electron counting and beam-induced motion correction enable 57. structure of the 26S proteasome lid reveals the mechanism of deubiquitinase 37. eLife 4. 25. R. & Lander. The 3. A. 27–29 (2007). Structure of a mammalian ryanodine receptor. J. 90. Commun. 69–74 (2016). M. Struct. D. Q. Nature 526. G. et al.. Nature 504. Cheng. 69. M. Nature 518. L.. J. Structure of the rabbit ryanodine receptor RyR1 at near-atomic polymerase III. Yuan. Verba.. Baumeister. (2015). 224–229 (2015). et al. S. M. Atomic structure of Hsp90-Cdc37-Cdk4 reveals that Hsp90 42. E. structures of the E. S. Meyerson. Campbell. Proc. W. Nature 531. Breaking cryo-EM resolution barriers to facilitate drug discovery. Glycine receptor mechanism 84. complex. S. 13543–13548 (2015). Y. Berndt. Misasi. film. Membrane protein assembly into nanodiscs. 74. 260–264 (2016). Ketcham. Nature 528. G. S. Nature 517. et al.. Structural and molecular basis for Ebola virus neutralization by (2016). et al. M. Nature 527. 83. & Williams. Transcription initiation complex structures elucidate DNA cryomicroscopy.8 Å. A. binding and -folding mechanics of the GroEL chaperonin. Abid Ali. Merk. 81. 122. C. T.. 59. e11134 Nature Methods 7. resolution. 29. J. J. conformations reveal activation mechanisms. Cryo-EM structure of SNAP-SNARE assembly in 20S particle. et al. . Zhang. G. Yang. Nature 534. P. 212–217 72. X. 450–454 The highest reported resolution (1. envelope trimer. e06980 (2015). S. Mechanism of NMDA receptor inhibition and activation. Zhang. USA 113. 60. S. Z. & Cramer. Nature Methods 10. T. 186–191 (2015). resolution reconstruction of the Thermoplasma acidophilum 20S proteasome 75. 399–404 (2015). Gong. R. et al. a pumpjack motion for translocation. The 27. & Carragher. C. Nature 515. An atomic structure of human γ-secretase. et al. present. D. N. X. 4652–4664 (2011). L. Sci. Architecture of human mTOR complex 1. Scheres. Gating machinery of InsP3R channels revealed by electron 92. Efremov. exonuclease and τ. J. Z. et al. 56. FEBS 79.6 Å reconstruction of rotavirus VP6. The structure of the dynactin complex and its interaction 41. CryoEM structure of yeast cytoplasmic exosome complex. 345–351 (2016). 183. H. M. 48. E. et al. Plaschka. eLife 3. Science 351.. aad2395 (2015). Murakami. Cryo-EM shows the polymerase structures and a nonspooled 704–714 (2016). J. eLife 5. 598–603 (2016). Beam-induced motion of vitrified specimen on holey carbon and desensitization. Brilot. Mol. Res.. Nature 524.U5 tri-snRNP at Refs 29–31 each present one of the first three high-resolution structures that 3. Herzog. D. D. Nature 521. Structure of 39–43 (2015). Struct. 80–84 (2014). Zhu. Mol. Bartesaghi. interactions with the DNA sliding clamp. Nature Methods 13. J. Cryo-EM structure of the yeast U4/U6. Amunts. J. et al. Science 351. E. A. determined by electron cryo-microscopy. L. Y. A. S. Bai. Fan. Tajima. Li. ribosome. e12102 (2016). 2. Structural mechanism of glutamate receptor activation 22. Du.. 7816–7821 (2016). S. & Sligar..-J. Natl Acad. 377–388 a-subunit of an F-type ATP synthase. 347–351 (2016). Likelihood-based 62. A. Baker. H. Vinothkumar. H. & Barford. proteins. 231–236 (2015). X. 26. e13027 (2016). Nature 517. R. 61–67 (2015). Cullin–RING ubiquitin E3 ligase regulation by the COP9 of cryo-EM structures that reduces the need for expert supervision. S. & Vonck. Nature 525. K. J. A. Clare. D..2 Å resolution cryo-EM structure of β-galactosidase in 76. Maximum-likelihood approach to single-particle image structure of the human mitochondrial ribosome. Structural and biochemical basis for induced self-propagation of of the F420-reducing [NiFe] hydrogenase by electron cryo-microscopy using a NLRC4. states of dynamic molecular machines from a single experiment. Atomic complex with a cell-permeant inhibitor. Structure of the eukaryotic MCM complex at 3. Scheres. signalosome. 415. H. L. & Cheng. W. & Julius. to near-atomic resolution from thirty thousand cryo-EM particles. polymerase complex inside a dsRNA virus. 10708 (2016). Nguyen. & Gouaux. Science 342.7 Å resolution. Ribosome structures resolution. et al. 551–554 (2016). J. Popot. 630–637 (2012). Mills. Sci. B.-C. A.9 Å. & Lamers. Structure of an RNA polymerase II preinitiation complex. e06380 (2015).. Nature 521. Toots. Atomic model 66. He. 1441–1446 (2015). 2. Hu.U5 tri-snRNP: insights into 31. F. & Ramakrishnan. Liao. Structural insights into the Niemann–Pick C1 (NPC1)-mediated to a translocation substrate. H. L. spliceosome assembly and catalysis.. et al. & Raunser. 1484–1490 (2013). Making membrane proteins for structures: a trillion tiny tweaks. Structure of the TRPV1 ion channel Science 351. 1347–1350 (2015). Biol. Zalk. 7. et al. 48–52 (2016). Cryo-electron microscopy structure of the Slo2. X. Science 348. Dambacher. mechanisms of ligand and lipid action. Nature 504.. TRPV1 structures in distinct with dynein. J. Martin. Hite. Science 351.. Scheres. Movies of ice-embedded particles enhance resolution in complex I. et al. Schroer. Cryo-EM structure of the activated NAIP2-NLRC4 inflammasome (2012).. A. Science 348. A. 44–49 91. Zhao. eLife 4. conformational switch mechanism of the ryanodine receptor. Cell (2015). USA 112. 95–98 (2015). Ozorowski. W. S. 113–123 65. et al.. direct electron detector. protective human antibodies. Urnavicius. Cavadini. & Ward. A l l r i g h t s r e s e r v e d . Hoffmann. et al. 77. M. 88. H. 584.. S. Z. electron cryo-microscopy. Zhou. 1416–1420 (2016). McMullan. A. REVIEW INSIGHT 21.. Biol. C. Campbell. J. et al. Chowdhury. R. reveals nucleated polymerization. C. Cryo-EM structure of a fully glycosylated soluble cleaved HIV-1 85. R. McLaughlin. Tor forms a 44. Biol. & Grigorieff. Y. R. Worden. J. M. Chang. et al. B. K. W.. J. et al. Zhang. Science 347. B. ATP-triggered conformational changes delineate substrate. G. Biol. J.. Wu. In situ structures of the segmented genome and RNA by ifenprodil. Aylett. 68. D. Nature 534. 177. Veesler. TRPV1 structures in nanodiscs reveal traps and stabilizes an unfolded kinase. L. B. & Scheres. N. Johnson. 1343–1346 (2016). Aebersold. S. Herguedas. Nature Methods 4.. Nature 527. Nature 514.. Liu. Benlekbir.. J. Bayburt. et al. K. Nature Struct. opening.. Nature 527. Biol. Science 350. Biol. transport chain components in bovine mitochondrial supercomplex I1III2IV1. Nature 530. 33. 61.. S. Cell 38. S. Structural Lett. W. G. C. D. D. H. Wu. & Hirst. X. 47. eLife 4. 58. the APC/C and its mechanism of protein ubiquitination. Cao. Baretić. M. et al. 345–347 (2015). A. 45. D. fully 86. Structure of the eukaryotic replicative CMG helicase suggests glycosylated. Nature 533.. Fernández-Leiro. & Rubinstein. Molecular structures of unbound and transcribing RNA 51. M. H. L. et al. F. & Lander. C. S.2 Å 24. 55. Architecture of mammalian respiratory 23. A statistical framework for the classification and high-resolution refinement 71. Julius.. 1467–1478 (2016). et al. Y. organization of the dynein–dynactin complex bound to microtubules. E. et al. promoter opening. Zhang. A Bayesian view on cryo-EM structure determination. R.. Sigworth. A. J. 50–55 (2015). Nature 529. 82. & Cheng. Cao. D. Conrad. genome within a dsRNA virus. 25. W. C. 429–434 (2010). Science 352. 466–475 (2016). V. Molecular mechanism of APC/C activation by mitotic 35. Cryo-EM structures of the eukaryotic replicative helicase bound 50. L. Gao. 1721–1727 (2010). J. Atomic structure of Cell 165. Liu. Frauenfeld. et al.. 107–112 (2013). Structure of the voltage-gated calcium channel CaV1. Disentangling conformational states of macromolecules An example of how image classification can reveal numerous functional in 3D-EM through likelihood optimization. (2013). (2013). Lee.. W. 22. 34. M. A saposin-lipoprotein nanoparticle system for membrane Struct.. Grant. 36. et al. 406–418 (2012). 822–837 (2016). 7.2 Na+-activated dimer through an N-terminal helical solenoid with a complex topology. Nature 533. Y. Cell 165. Brilot. et al. J. Cheng. F. 26. A. Nature Commun. 531–534 (2015). Nature 40. Schweitzer. 52. T. Li. G.. Herzik.U5 tri-snRNP. rotational states in a eukaryotic V-ATPase. J. 54. 217–224 49. et al. refinement. K. R.. et al.. 3. 67. et al. N. Zhao. Theobald. 113–118 (2013). A. Science 351. Measuring the optimal exposure for single particle inhibition. A.. R. 30. heralded the resolution revolution in cryo-EM-based structure determination. N. 241–245 (2015). Science 352. 551–560 (2015). Cryo-EM structure of a native. Bai. transcribing mammalian RNA polymerase II. Plitzko. M. Cell 165. Mechanistic insights into the recycling machine of the SNARE EMBO J. Cao. K. 78. Nature 533.

Song. & Passmore. T. Acetylcholine receptor channel imaged in the open state. E. Scheres. 170.nature. Biol. (2014). Structural snapshots of actively translating human 37–43 (1995). Nature 523. Kimanius. Chen. G. Cryo-EM single particle analysis with the Volta phase plate. J.W. S. coli ribosome–EF-Tu complex at <3 Å comments on an early version of this manuscript. Maskell. Trajectories of the ribosome as a Brownian nanomachine. structure determination with parallelisation using GPUs in RELION-2. 179. A. Science 351. 659–671 (2009). L. Cell 161. W. 109. 101. Liang. Biol. Schmeisser. Cheng. Structure. et al. Fischer. 1377–1380 (2014). Ultrastable gold substrates for electron (scheres@mrc-lmb. 119. .and function-based design of Plasmodium-selective electron microscopy. preparing EM specimens. Crum. Barford and R. Structural dynamics of ribosome subunit association studied by 102. Saibil. 679–684 (2002). Struct. Nature Struct. electron cryomicroscopy. U.H. Structure 23. eLife 4.doi. Scheres. distributed and GPU computing technologies in twisted by tetranucleosomal units. 16 (suppl.. sub-tomogram averaging can bridge the gap between the structural biology of 1411–1416 (2009).1101/059717 (2016). B. eLife 3. Molecular architecture of the inner ring scaffold of the human of single particle cryo-electron microscopy data in the cloud. 1138–1152 (2015). (2015). Parallel. & Passmore.. Sheehan. 124. 549–557 (2016).. C.org/10. Henderson for 110. D 71. J. Medical Research Council (MC_UP_A025_1013). erratum 163. electron microscope operated at 200kV. 68–75 (2012). USA 111. Danev. Cryo-EM reveals a novel octameric integrase structure oxide: a substrate for optimizing preparations of frozen-hydrated samples. F. & Plitzko. A. 358–361 (2016). 123. 3 4 6 | NAT U R E | 5 3 7 | 1 5 S E P T E M B E R 2 0 1 6 © 2 0 1 6 M a c m i l l a n P u b l i s h e r s L i m i t e d . B. D. D.. Behrmann. A. 581–585 (2015). Struct.. J. M. A national facility for biological cryo. et al. 366–369 (2015).S is funded by the UK resolution by Cs-corrected cryo-EM. G. Acknowledgements We thank Y. et al. Baumeister. E. B. Enhanced imaging in This work presents an exciting vision of how cryo-electron tomography and low dose electron microscopy using electron counting. M. Near-atomic resolution reconstructions using a mid-range mixing-spraying time-resolved cryogenic electron microscopy. 118. & Leschziner. D. K. Science 344. cryo-EM using low-energy hydrogen plasmas. New applications for affinity grids in transcriptase. Campbell. Proc. Opin. Microanal. Ballandras-Colas. D. Qu. D. Beam-induced motion correction for sub-megadalton cryo-EM particles. et al.. Science 352. D. The authors declare no competing financial interests. et al. Taylor. Li.H. & Boettcher. Pantelic.. Unwin. D 65. INSIGHT REVIEW 95. N. Dashti. Author Information Reprints and permissions information is available at 112. I. et al. Structure of a group II intron in complex with its reverse 116. F. Ultramicroscopy 63. Mahamid. G. periphery. 867–871 (2016). Cianfrocco. Nature 520. Structures of the CRISPR–Cmr complex reveal mode of RNA Biol. W. 127–135 (2015). J. Struct. J. Acta Crystallogr. Russo. 376–380 (2014). A l l r i g h t s r e s e r v e d . 121. ribosomes. 105. F.uk). Cell 163. B. 152–156 (2010). S. 113.. e13046 (2016). C. & Lindahl. Accelerated cryo-EM organization of the cytoplasm. Nature 530. 567–570 (2015). R. Preprint 125. Structures of a CRISPR–Cas9 R-loop complex primed for DNA 117. A. et al. Nature Methods 11. H. W. A. go. Jain. Electron radiation damage to protein www. A. Biol. Controlling protein adsorption on graphene for (2015). 99. Electron tomography: towards visualizing the molecular 107. H. M. Biol. A. et al. T. 115. 183–187 (2014). J. & Potter. et al. Science 346. Readers are welcome to comment on the online version of this paper at 75–79 (1996). et al. Carragher. P. proteasome inhibitors.com/2bqncxm. J. Ru. 363–365 (2016). et al. E. Ultramicroscopy 109. 100. Russo. R. 1097–1105 (2015). R. 840–841 (2010). N. Structure of the L protein of vesicular stomatitis virus from 120. et al. & Baumeister.. S. Microsc. Turchetta. et al. Acta Crystallogr. Grünewald. Clark. Baumeister. McMullan. S. Kelly. et al. for an integrated inkjet dispense and vitrification system for cryo-TEM. 23. 17492–17497 (2014). J. et al. single-particle electron microscopy. 104.W. C.. Nature 373. S. 12. F. Graphene 97.com/reprints. D. isolated complexes and cell biology. Jiang. 111. Correspondence should be addressed to S. Struct. Natl Acad. & Stuart. C. for betaretroviral intasome function. Stark. R. Molecular mechanism of V(D)J recombination from synaptic RAG1– cryomicroscopy. Kosinski. Sci. Structure of the E. p a r t o f S p r i n g e r N a t u r e . Cryo-EM study of the chromatin fiber reveals a double helix 106. 1807 114. Nature 530. Cell 162. 188.. Structural basis for retroviral integration into nucleosomes. W. Mol. Curr. e06664 nuclear pore complex. Kaiser. Visualizing the molecular sociology at the HeLa cell nuclear at http://dx. Forsberg. Spotiton: a prototype cleavage. T. et al. high performance processing 122. 969–972 (2016). crystals of bacteriorhodopsin at different temperatures.ac. L. Dukovski.cam. S2). Meyer. 649–652 96. J. H. 103. 845–857 (2015). Science 351. 108. e03665 (2014). 98. M.nature. eLife 5. RAG2 complex structures. H. Science 348. O. A. 314–327 (2015). R. C. 233–236 (2016). P.. J.S. Zemlin. Low cost. target positioning.. et al. & Faruqi. H. & Walz.