Single Particle EM José-Jesús Fernández, José-María Valpuesta,

Introductory article

Centro Nacional de Biotecnología – CSIC, Madrid, Spain Centro Nacional de Biotecnología – CSIC, Madrid, Spain

Article Contents • Introduction • Specimen Preparation • Image Acquisition in the Electron Microscope • Structure Determination

Based in part on the previous version of this eLS article ‘Single Particle EM’ (2009) by José-Jesús Fernández and José-María Valpuesta.

• Structure Validation • Structure Interpretation • Advances towards Atomic Resolution • Illustrative Examples • Acknowledgements

Online posting date: 16th March 2015

Single particle electron microscopy (EM) plays an important role in structural biology because it allows derivation of biologically relevant information about proteins and macromolecular complexes. A large amount of randomly oriented images of the specimen under study (so-called particles) are collected from micrographs taken with an electron microscope. These particles are then computationally aligned and combined to yield the 3D structure, which is subsequently subjected to visualisation and interpretation. In most cases, the resolution attained with this technique precludes tracing of the polypeptide chain or the clear visualisation of the secondary structure elements. Nevertheless, the integrative combination of the information provided by the different structural techniques (X-ray crystallography, EM, etc.) at different resolution levels has allowed a comprehensive interpretation of the structure. The recent advancements in instrumentation and computational procedures are now making it possible to obtain maps at sub-nanometre, and even near-atomic, resolution.

Introduction Knowledge of the structure of biological specimens at all levels of detail is essential in life sciences to understand their functions. Electron microscopy (EM) is playing an increasingly important role in structural biology for the analysis of eLS subject area: Cell Biology How to cite: Fernández, José-Jesús and Valpuesta, José-María (March 2015) Single Particle EM. In: eLS. John Wiley & Sons, Ltd: Chichester. DOI: 10.1002/9780470015902.a0021846.pub2

proteins and macromolecular complexes. As EM has matured, its complementarity with other traditional high-resolution structural techniques such as X-ray crystallography or nuclear magnetic resonance (NMR) has become clear. The combination of EM with these other techniques allows the integration of the multiresolution structural information gathered at multiple levels of the biological complexity with the ultimate goal of a comprehensive interpretation. In addition, the recent significant advances in EM instrumentation and computational procedures have opened up the possibility of achieving near-atomic resolution with this technique in a rapidly increasing number of cases. See also: Macromolecular Structure Determination by X-ray Crystallography; Macromolecular Structure Determination: Comparison of X-ray Crystallography and NMR Spectroscopy; Proteins: Fundamental Chemical Properties; Two-dimensional Electron Crystallography; EM Analysis of Protein Structure; History of the Electron Microscope in Cell Biology Single particle EM allows structure determination of biological samples from EM images of individual molecules or complexes (hence single particles). The underlying principle relies on the collection of a large amount of images of the specimen under study in different orientations, which are then computationally aligned and combined to obtain a three-dimensional (3D) density map. Afterwards, this 3D map is subjected to visualisation and interpretation, mainly in terms of secondary structure elements and, more recently, even at the level of atomic model building. See also: Electron Cryomicroscopy and Three-dimensional Computer Reconstruction of Biological Molecules; Proteins: Fundamental Chemical Properties Single particle EM has a number of advantages compared to the traditional structural techniques: (1) structural information of isolated specimens can be derived from relatively small amounts of purified samples. (2) There is no need to grow protein crystals. (3) The specimens may be captured in their native conformation in near-physiological buffers. (4) This technique allows, in principle, to deal with heterogeneity (e.g. populations from different conformational states that are mixed in the sample or impurities in the preparation) as long as the different populations in the heterogeneous sample may be visually or computationally separated. (5) There is no upper limit to the size of molecules, though a

eLS © 2015, John Wiley & Sons, Ltd. www.els.net

1

Single Particle EM

minimum size is required (approximately 150 kDa for cryomicroscopy; approximately 50 kDa for negative staining; see section titled ‘Specimen Preparation’). (6) Finally, EM is suitable for the structural analysis of large, flexible macromolecular assemblies, purified only in modest quantities and difficult to be studied by conventional methods (X-ray crystallography and NMR). Single particle EM currently allows structure determination of proteins and macromolecular assemblies from about 30 to 3Å resolution. At about 25 Å resolution, the overall shape can be identified. Resolutions better than 15 Å permit the identification of functional domains and subunits within the whole structure. Thanks to the technical and computational advances, in the last few years an increasing number of structural studies of large-to-medium complexes by single particle EM have attained sub-nanometre resolution and, recently, up to near-atomic level. This level of resolution allows the identification of secondary structure elements (α helices and β sheets) and even the tracing of the protein backbone. There are exciting prospects for routinely reaching resolutions that allow the assignment of a sequence of amino acid side chains to particular density features in the maps.

Specimen Preparation The target protein or complex must be purified to obtain a sample as homogeneous as possible, as single particle EM relies on the assumption that every particle derives from the same specimen. The sample is then transferred to an EM grid to be vizualized under the electron microscope. However, the sample must be treated prior to the electron exposure since the vacuum conditions in the microscope and electron radiation are deleterious for the biological structure. The goal of specimen preparation is to prevent dehydration of the sample in the vacuum and to reduce the sensitivity to radiation damage, while keeping the biological sample as close as possible to its native state. There are two main preparation techniques: negative staining and electron cryomicroscopy. Each technique has its own advantages and limitations (Figure 1). See also: Electron Microscopy; Transmission Electron Microscopy: Preparation of Specimens Negative staining consists in embedding the sample in a medium with a contrast agent (a heavy metal salt) (Brenner and Horne, 1959). The primary source of contrast is therefore the stain cast rather than the biological material itself. This technique is fast and easy to use and yields images with relatively high contrast, which provide structural information of the specimen and internal cavities. However, there are several important disadvantages. First, the maximum resolution attainable is limited by the size of the salt grain used for staining, usually around 15–20Å. Second, the images may present artefacts derived from the dehydration, the distortion or collapse of the specimen due to the staining procedure or the unevenness or the penetrating ability of the stain (Figure 1). For those reasons, negative staining is not used for high-resolution studies, although it is very useful for the earlier stages of research to get a first glimpse of the structure. Negative staining is also indispensable for analysis of small proteins (approximately 50–150 kDa) that, otherwise, could not be visualised with cryomicroscopy due to the lack of contrast. 2

For cryomicroscopy, the sample in its native buffer solution is rapidly frozen by being plunged into liquid ethane and kept frozen in a thin layer (50–80 nm) of vitreous (amorphous) ice at cryogenic temperatures, which ensures preservation in near-physiological conditions during electron exposure (Dubochet et al., 1988). Images of vitrified unstained specimens intrinsically present low contrast due to the fact that the biological material mainly consists of light atoms (C, H, O, N, etc.), and the density is very close to that of the embedding vitreous ice. This technique does not imply any direct limitation on the achievable resolution. Therefore, cryomicroscopy of vitrified specimens (cryo-EM for short) is the method of choice for high-resolution structural studies in a near-native environment. See also: Electron Cryomicroscopy Cryonegative staining is an intermediate technique that combines negative staining and cryomicroscopy (Adrian et al., 1998). Here, the biological material is placed in a concentrated contrast agent solution just before vitrification. As a result, the sample is kept in a thin layer of vitrified salt solution. This technique keeps the specimen frozen, similar to cryomicroscopy, while having the much higher contrast of negative stained preparations. It is however uncertain to what extent the high concentration of salt in cryonegative staining may have an effect on the molecule structure.

Image Acquisition in the Electron Microscope The optical principles of the transmission electron microscope (TEM) are similar to the widely known light microscope (Figure 2). A source of illumination emits electrons at the top of a cylindrical column about 2 m high. The electrons are accelerated and form a beam that travels down the column in vacuum and is focused towards the specimen by magnetic lenses. The incident electron beam then passes through the specimen, and electrons are either unscattered or scattered by the atoms of the sample. This scattering occurs either elastically (i.e. with no loss of energy) or inelastically (i.e. with energy deposition resulting in radiation damage). Scattered and unscattered electrons emerging from the specimen are then collected by the objective lenses and focused to form an interference pattern that produces the image (Baker and Henderson, 2012). Under the assumption of single-scattering events, images taken with TEMs can be considered as 2D projections of the specimen, that is, as formed by the integration of the 3D information of the specimen along the beam direction. In other words, the images look like ‘radiographs’ of the sample. The images provided by the TEM (so-called micrographs) are recorded on photographic films, which will eventually be scanned to produce digital information for subsequent computer processing. Alternatively, and nowadays more frequently used, a camera based on a charge-coupled device (CCD) or a recently introduced direct detection device (DDD) can be placed inside the TEM, which directly produces digital images. Simplistically speaking, the image formation procedure can be viewed as follows. The passing of the electron beam along its path through the specimen is hampered by the

eLS © 2015, John Wiley & Sons, Ltd. www.els.net

Single Particle EM

(a)

(b) (f) (c)

(d)

(g)

(e)

Figure 1 Sample preparation. Sketches of the preparation of the biological material by negative staining and cryomicroscopy are shown in (a–d) and (e), respectively. Middle panel (a–e, right) shows sketches of the structure that would be projected in the acquired images from the preparations shown in (a–e, left). In negative staining, the sample is embedded in a stain medium (a) that produces images of the specimen with good contrast (a, right). However, this technique is prone to artefacts (b–d, right) coming from partial (b) or uneven (c) staining, or due to the specimen distortion (d). In cryomicroscopy, the sample is embedded in a layer of vitreous ice (e), ensuring preservation of the structure in near-physiological conditions while imaged (e, right). Micrographs of a sample of bacteriophage T4 prepared with negative staining and cryomicroscopy are shown in (f) and (g), respectively. Note the better and inverted (hence negative) contrast in (f) with respect to (g). Also note that the viruses in (f) present artefacts and deformations compared to the good preservation in cryomicroscopy (g).

greater number atoms of the protein or complex and thus fewer electrons get into the detector, either film or camera, resulting in a projection image that reflects a superposition of the structural features at different layers in the 3D structure. See also: Electron Microscopy The TEM, as any other physical instrument, is not perfect and thus leaves an imprint in the acquired images. This imprint is modelled by the transfer function of the TEM (so-called CTF – contrast transfer function), which arises from the aberrations of the lenses and from the defocus used in imaging. The artefacts of the CTF consist of the attenuation of structural details and, even worse, in contrast reversals at certain spatial resolution ranges. The contrast reversal is especially harmful as the structural details at some resolution ranges are presented as white density over a black background, whereas at other resolution ranges they appear as black over white. Estimation of the CTF and correction for its effects are therefore essential for any image to faithfully represent a projection of the specimen, which is of paramount importance for high-resolution structural studies (Baker and Henderson, 2012). The theoretical resolving power in a TEM is imposed by the electron wavelength, which depends on the accelerating voltage, and is typically around 0.02 Å. In practice, due to lens aberrations and other limitations, real instruments resolve to about 2Å. However, the current effective resolution limit ranks around 3–10Å because of the specimen preparation techniques, sample thickness, low contrast, radiation damage, specimen motion and so on.

Biological material is very sensitive to radiation. Electron doses must therefore be kept very low to minimise beam-induced radiation damage to the specimen and to preserve as much resolution as possible. For most specimens, images are typically recorded with doses around 5–20 e− Å−2 . The combination of low dosage and weak contrast makes cryomicroscopy images extremely noisy, with a signal-to-noise ratio (SNR) around 0.1 or lower. This poor SNR is substantially increased by combining very large sets of images of the same specimen (currently, tens to hundreds of thousands of particles) in high-resolution structural studies.

Structure Determination The general problem in structural determination by EM is the 3D reconstruction of a biological specimen from a finite set of 2D images (single particles) taken from the specimen at different orientations (Baker and Henderson, 2012). It is mathematically proven that this problem can be solved provided that a sufficient number of 2D images, enough to sample the 3D object in the Fourier space, are available. In practice, because the EM images are very noisy, a much higher number of images than the theoretical images are needed to increase the SNR. From an image processing point of view, there are several problems when addressing such structural analyses: the extremely low SNR of the images; alignment and classification of the images; accurate determination of the relative orientations of the particles;

eLS © 2015, John Wiley & Sons, Ltd. www.els.net

3

Single Particle EM

To high voltage power supply Filament Anode Microscope column Condenser lens

Illuminating system

Objective lens Specimen stage

Intermediate and projector lenses

Phosphorescent screen

Specimen plane

Imaging system

To vaccum pumps

Photographic plate or digital camera

Figure 2

Electron microscope. A photograph and a schematic representation (courtesy of Dr. C. San Martín) are shown.

the 3D reconstruction problem itself; and finally, once the structure is obtained, possible post-processing and interpretation of it. Figure 3 shows a sketch of the major image processing steps involved in structure determination by single particle EM, which will be described in the following. See also: Electron Cryomicroscopy and Three-dimensional Computer Reconstruction of Biological Molecules Micrographs obtained from the TEM show EM fields where individual particles of the specimen under study are widely dispersed and randomly oriented. The very first step is to detect and box out the single particles from the micrographs. This process is normally carried out by visual inspection of the digital micrographs, manual selection and extraction of the particles. Automatic particle selection from micrographs is increasingly used and becomes particularly important as the required number of collected particles increases (tens or hundreds of thousands) with the resolution needs. The selected particles are then subjected to a number of steps that end up with the particles with the density levels normalised and corrected for the contrast reversals due to the transfer function of the TEM. The next major steps are alignment and classification of the particles. Particle alignment intends to reposition each particle by finding out the appropriate rotation and translation so that all particles can be placed in register, that is, in the same relative 4

position. The simplest approach for alignment is based on comparison of each particle against a reference, forcing the particle into an arrangement similar to the reference. On the other hand, particle classification tries to identify similar views of the same object, or different objects, out of the aligned particles. The principles of classification are based on the identification of common features in the data. Classification thus allows identification and separation of different views of the 3D object from the set of particles, and allows increasing the SNR by averaging particles that belong to the same class or object view. As alignment and classification mutually influence each other, they are usually combined in an iterative way. A more straightforward approach that intrinsically combines alignment and classification in a single, iterative process is multireference refinement. Here, the set of particles is aligned with respect to a predefined number of reference images, which are assumed to represent the different views of the object and the potential structural diversity among the data. As a result of these steps of alignment and classification, averages that represent the different 2D views of the specimen under study are obtained, which present much better SNR than the original particles contributing to those averages. The relative 3D orientations (the angular parameters, the Euler angles) of the average views obtained as a result of alignment and classification must be found out to compute the 3D structure.

eLS © 2015, John Wiley & Sons, Ltd. www.els.net

Single Particle EM

Particle selection

Micrograph

Aligned and classified particles

Selected particles Alignment and classification

Iterative refinement

3D structure Angular assignment and 3D reconstruction Average views

Figure 3 Structure determination. Particles of the specimen under study are randomly dispersed and oriented in the micrographs taken with the microscope (top left). First, particles are selected and extracted (top right). After alignment and classification, averages that represent the different 2D views of the specimen under study are obtained (bottom left). The relative 3D orientations of the average views are determined by angular assignment, after which the tomographic reconstruction can be carried out to yield the 3D structure (bottom right). This structure can be refined afterwards by iteratively combining angular assignment and 3D reconstruction and using the original particles instead of average views. (See main text for a more detailed explanation of the whole process.) (Material for this figure is by courtesy of Dr. E. Arias-Palomo.)

The angular determination is normally carried out by comparing the average views with computer-simulated projections of a model resembling the specimen under study (so-called projection matching procedure). The orientation of the best-matching model projection is then assigned to the average view. An alternative approach is based on the fact that any two projections of a volume must share a common line in Fourier space, which allows angular determination without a reference volume. Regardless of the angular assignment method, determination of the 3D structure is only possible as long as the angular parameters of the average views are known. There exist different 3D reconstruction methods that share the goal to minimise the error between the average views and the corresponding projections obtained from the reconstructed volume. The 3D structure can be improved afterwards by iterative refinement, that is, it is refined by iteratively combining angular assignment and 3D reconstruction using the original particles instead of average views. Single particle EM is the most broadly applicable EM structure determination modality in the sense that it is suitable for non-crystalline and asymmetric specimens. For specimens with symmetry, single particle EM can take advantage of the symmetry relationships to accurately align the particles and to decrease the number of images required to determine the 3D structure. Imposing an n-fold symmetry during the single particle EM processing

is essentially equivalent to an n-fold increase in the actual number of particles collected and processed to yield the structure. This approach has been especially powerful when applied to icosahedral particles because of their very high symmetry (60 symmetry elements). See also: Bacteriophages: Structure

Structure Validation The calculated 3D structure needs to be validated to confirm that it truly represents the biological structure under study in full detail. This validation must be done at different levels. First, the general correctness of the structure must be proved. Next, it is necessary to assess the resolution attained or, in other words, the level of detail that is undoubtedly present and reliable in the structure. This is essential since all unreliable details should be discarded from the map for a proper interpretation of the structure. The tilt-pair validation (Rosenthal and Henderson, 2003; Henderson et al., 2011) allows overall check of the 3D structure based on the consistency of the orientations of the particles. Pairs of images of the same field of single particles are recorded at two different tilt angles (typically, 0 and 10; ±5, ±10) around a tilt axis. The orientation of all particles is then estimated against the

eLS © 2015, John Wiley & Sons, Ltd. www.els.net

5

Single Particle EM

3D structure. The differences in the orientation of each pair of corresponding particles should then cluster around the experimental tilt axis and tilt angle. The structure is reliable if a significant amount of particles (typically 60%) are within the cluster. The tilt-pair analysis also allows determination of the absolute hand of the structure (Rosenthal and Henderson, 2003). If the tilt angle found by the tilt-pair analysis is the negative of the experimental one, the handedness of the 3D map must be inverted. The tilt-pair validation ensures validity of the 3D structure at low resolution, up to around 20 Å. Tools to assess the information at higher resolution are thus required. This is of paramount importance in cryo-EM because the low SNR of the images affects the orientation estimation of the particles, and this ultimately may lead to artefactual details. The widely accepted method to prevent these potential artefacts is referred to as gold-standard Fourier shell correlation (FSC) (Scheres and Chen, 2012; Chen et al., 2013). It consists in randomly splitting the collected particles into two half data sets, from which two absolutely independent 3D reconstructions are computed as described in the previous section. The two independent maps are then compared in Fourier space, in resolution shells (hence FSC), so as to find the maximum spatial resolution up to which both are mutually consistent. This resolution is used to filter a map calculated from the whole set of particles, which yields the final 3D structure. Intrinsic properties of the specimen (e.g. flexibility, structural heterogeneity of the sample), and technical and computational factors (e.g. uneven orientation distribution of the particles, errors in the orientation estimation) may translate into a 3D structure with locally varying level of detail. For instance, flexible protrusions in a 3D map are usually resolved at lower resolution than very well-ordered features. To properly interpret these maps, there now exist tools to measure the resolution at a local scale and to filter the maps accordingly with locally adaptive approaches (Cardone et al., 2013; Kucukelbir et al., 2014).

Structure Interpretation The ultimate aim of structure determination is the extraction of biologically relevant information that allows drawing conclusions and provides new knowledge about the biological system under consideration. In general, the resolution achieved in structural studies by single particle EM still precludes tracing of the polypeptide chain, thereby providing limited information about the biological specimen. Global shapes (at about 25 Å resolution) or domains (at about 10–15 Å) are identified in most cases. As sub-nanometre resolution is becoming approachable almost routinely, secondary structure elements are discernable in a vast amount of cases. Only in a few particular cases, the protein backbone and side chains have been possible to be traced (at 3–4 Å) (Liao et al., 2013; Amunts et al., 2014). In this regard, there exist several approaches to bridge the gap between atomic resolution and the resolution level of the structure solved by single particle EM (Villa and Lasker, 2014). These approaches are based on either linking with atomic information available from previous X-ray crystallography studies or computationally prediction of the secondary structure elements (Figure 4). See also: 6

Computational Methods for Interpretation of EM Maps at Subnanometer Resolution Rigid-body docking, which has been extensively used in the field, allows fitting atomic models into the density map with the aim of better defining protein domains and creating pseudo-atomic models of the structure under study. Here, the atomic model is considered as a rigid body that is placed into the density map in terms of the best visual fit or according to correlation-based computational methods. This approach has been particularly useful to define domains into large complexes made up of several domains or proteins. Biological systems undergo large functional rearrangements that involve collective motions between small regions of the proteins, which cannot be modelled by rigid-body fitting. Flexible fitting emerges as an approach capable of reorganising the atomic models while being docked into the density map to rationalise the conformation observed in the structure solved by single particle EM. The reorganisation of the atomic model is normally carried out by exploring only the natural collective deformational motions that lead to physically realistic conformational changes, according to normal mode analysis or molecular dynamics. See also: Molecular Dynamics; Normal Mode Analysis Techniques in Structural Biology In density maps solved at sub-nanometre resolution, secondary structure elements become discernable. In those cases, instead of fitting known atomic models, the actual prediction of secondary structure elements can be accomplished. A current approach that is used in the field predicts the location and orientation of α helices and β sheets based on density skeletonisation, template-based search (looking for rod or plane-like features) and local geometry analysis within the EM density map (Baker et al., 2007). As a result, the propensity for sheet and helix-like features is obtained for representative points in the map, which allows the construction of the secondary structure topology. At near-atomic resolution, construction of an atomic model based on de novo modelling becomes possible (Figure 4). Here, the structural features in the EM map, including the backbone and side-chain traces at high resolution and secondary structure elements at lower resolution, are compared with those obtained from sequence-based secondary structure prediction. This mapping allows building of a topological model from which individual atoms can be placed using spatial constraints derived from the map (Baker et al., 2010). After validation and interpretation, the structures solved by single particle EM are normally deposited within structural databases to make them accessible to the scientific community. The Electron Microscopy Databank (EMDatabank) is a joint initiative among the Protein Databank in Europe (PDBe) at the European Bioinformatics Institute, the Research Collaboratory for Structural Bioinformatics (RCSB) at Rutgers and the National Center for Macromolecular Imaging (NCMI) at Baylor College of Medicine. EMDatabank (http://www.ebi.ac.uk/pdbe/emdb; http://www.emdatabank.org) acts as a host to store and access the structures solved by any EM modality in much a similar way as Protein Data Bank (PDB) does for atomic coordinates mainly obtained by X-ray crystallography or NMR (Lawson et al., 2011). See also: Structural Databases of Biological Macromolecules

eLS © 2015, John Wiley & Sons, Ltd. www.els.net

Single Particle EM

η

α

η

δ

ε

δ

θ

ζ

θ

γ

β

α ε ζ

γ

β

ATP

(a)

(b)

βP

βS

α3

α1

α2

(c)

30 Å

(d)

Figure 4 Structure interpretation. (a) Rigid-body fitting allowed interpretation of the mechanism of the eukaryotic chaperonin containing T-complex polypeptide 1 (CCT) to assist the folding of actin by fitting its atomic structure into specific regions of the map (Llorca et al., 2001). (b) Flexible fitting allowed comparison of the single particle EM map of an archaeal prefoldin (a complex that stabilises and delivers unfolded proteins to a chaperonin for facilitated folding) with the atomic structure of a homologue (Martín-Benito et al., 2007). (c, left) Rigid-body fitting of the atomic structure of a trimer of infectious bursal disease virus T = 1 subviral particle into a map solved by single particle EM allowed assessment of the latter (Luque et al., 2007). At this resolution (close to 7Å), secondary structure prediction yielded good results (c, right) in the detection of α helices (green rods) and β sheets (blue planes), as compared with the atomic structure of the monomer (Fernandez et al., 2008). Material for this panel is by courtesy of Dr. D. Luque. (d, left) High-resolution EM structure of the TRPV1 ion channel at 3.4 Å (Liao et al., 2013) allowed identification of many side-chain densities and (d, right) enabled de novo atomic model building for most residues. (d) Reproduced from (Liao et al., 2013) © Nature Publishing Group.

Advances towards Atomic Resolution A significant advance in instrumentation has revolutionised the field of single particle cryo-EM in the last few years: the advent of digital cameras that are based on a DDD. The traditional ways to produce the digital images in this field rely on either CCD cameras installed in the TEM or digitised photographic films. The former are fast but do not perform well at high resolution. The latter are well suited to high-resolution studies, but they are not compatible with high throughput since the films have to be scanned to produce the images. The modern CMOS-based DDD sensors detect electrons directly and produce digital images with

improved SNR at high resolution at an unprecedented speed, thus combining the advantages of CCD and films (Faruqi and Henderson, 2007). The fast readout rate of DDD cameras has an additional, considerable advantage. It allows detection of the small movements that the particles undergo when the electron beam strikes the sample in the TEM (Campbell et al., 2012; Brilot et al., 2012). These movements translate into an image blurring or, in other words, a steep degradation of the high-resolution information. This problem has long been considered apparently insurmountable. However, the new cameras now allow recording images as a set of frames of a movie. Thus, the beam-induced movements can be identified, calculated and compensated for by means of computational procedures.

eLS © 2015, John Wiley & Sons, Ltd. www.els.net

7

Single Particle EM

(a)

(b)

Figure 5 Identification and compensation for the beam-induced specimen movement. (a) The fast readout rate of modern DDD cameras allows recording of images as frames of a movie. Estimation of the translational motion in individual frames, or rather averages of the frames, can be carried out with computational techniques based, for instance, on cross-correlation. For this panel, 60 frames from the same field of double-layered rotavirus particles were taken. By comparing consecutive 10-frame averages, the shifts undergone by the particles were estimated. The first 10-frame average and three other representative ones are shown here. The white lines represent the shifts with regard to the previous 10-frame average, scaled by a factor of 70× for illustrative purposes. (a) Reproduced from Brilot et al. (2012) © Elsevier. (b) Motion correction. On the left, the average of the 60 frames was computed in a straightforward way, which results in a substantially blurred image. On the right, the 60 frames were first aligned to compensate for the shifts describing the particle movements, and their average shows substantial reduction of the blurring and improved contrast. Adapted from Grigorieff, 2013 © Creative Commons Licence.

Several approaches have been introduced for computational motion correction. The motion can be estimated and corrected for using the whole image (Li et al., 2013), or individually for each particle (Bai et al., 2013) or for patches, where sets of neighbour particles are treated as affected by the same movement (Scheres, 2014). Motion compensation thus allows generation of corrected images, or corrected single particles directly, with significant sharpness and contrast (Figure 5). Currently, there are research groups that are also studying other mechanisms to reduce beam-induced movement, for instance using alternative sample supports (Russo and Passmore, 2014). The effect of motion correction on DDD images has turned out to be dramatic in terms of resolution improvement. It has opened the door to high-resolution structural studies, in the range 3–4 Å, of low-symmetry macromolecular complexes using just a few thousands of particles (Amunts et al., 2014, Liao et al., 2013; see next section), even for small proteins (Lu et al., 2014). There are exciting prospects for atomic-resolution structure determination with single particle cryo-EM, as theory anticipated 20 years ago (Henderson, 1995).

Illustrative Examples An illustration of the variety of specimens, sizes and resolutions encompassed by single particle EM is shown in Figures 6 and 7. A brief description of these specimens and details about the 8

elucidation of their structure follows. Most of these structures have been obtained from the EMDatabank (the accession code is indicated in the following in the form emd-xxxx). Figure 6(a–c) shows representative examples of structures solved at near-atomic resolution (around 4Å resolution) before the advent of DDD cameras. This level of resolution was possible, thanks to the technical progress of cryo-EM and the combination of a significant number of particles along with extensive symmetry operations. GroEL is a prokaryotic molecular chaperone (molecular mass of 800 kDa) that promotes protein folding in bacteria. Its structure (emd-5001) was determined by combining 20 000 particles and using sevenfold symmetry relationships (Ludtke et al., 2008). The structure of the viral protein VP6 (41 kDa) in the icosahedral rotavirus inner capsid particle (emd-1461) was solved from 8400 viral particles combined with extensive use of symmetry (icosahedral 60-fold followed by further averaging of the 13 equivalent monomers in the capsid that were not icosahedrally related) (Zhang et al., 2008). Finally, the structure of the cytoplasmic polyhedrosis virus (47 MDa, emd-1508) was elucidated from 13 000 particles with the imposition of the icosahedral symmetry (60-fold) (Yu et al., 2008). In these studies, it was possible to clearly unravel secondary structure elements, loops, polypeptide backbones and even bulky side chains. Since late 1990s, the advances in single particle cryo-EM have enabled sub-nanometre resolution structure determination almost

eLS © 2015, John Wiley & Sons, Ltd. www.els.net

Single Particle EM

GroEL chaperone 800 kDa, 4.2 Å

Bovine rotavirus VP6 protein 41 kDa, 3.8 Å (b)

(a)

Bacteriophage T 7 connector 700 kDa, 8 Å (d)

Cytoplasmic polyhedrosis virus 47 MDa, 3.88 Å (c)

Human parvovirus B19 5.4 MDa, 7.5 Å (e)

Human DNA-dependent protein kinase catalytic subunit 470 kDa, 13 Å (g) (h)

Eukaryotic 80s ribosome 4 MDa, 7.4 Å (f)

Human geminin 90 kDa, 18 Å

Archaeal prefoldin 87 kDa, 19 Å (i)

Figure 6 Illustrative structures determined by single particle EM. Structures solved at near-atomic resolution (a–c), at sub-nanometre resolution (d–f) and at medium resolution (g) by single particle cryo-EM are included as representative examples. Two representative structures of small proteins where the use of negative staining is of paramount importance are also shown (h–i). For all these structures, the actual resolution achieved is included. As an indication of their sizes, the molecular mass is indicated too. The structures in (a–h) are accessible through the Electron Microscopy Data Bank under the following accession codes: emd-5001, emd-1461, emd-1508, emd-1231, emd-1466, emd-1217, emd-1102 and emd-1190, respectively. The structure in (i) is from a previous work of ours. Bar, 5nm.

routinely not only for specimens with different levels of symmetry but also for asymmetric specimens. Figure 6(d–f) shows some representative structures in this category that were solved at around 8Å resolution. At this level, α helices can be visualised or at least predicted by computational techniques. The structure of the bacteriophage T7 connector (700 kDa, emd-1231) was determined from 26 000 particles and imposition of 12-fold symmetry (Agirrezabala et al., 2005). Structural studies of icosahedral viruses normally yield sub-nanometre resolution because of the 60-fold symmetry, as happened with the infectious human

parvovirus B19 (5.4 MDa, emd-1466) using 8800 particles (Kaufmann et al., 2008). Finally, the ribosome is a representative of structures where the lack of symmetry has to be compensated with a significant increase in the number of particles if sub-nanometre resolution is to be achieved. Several studies with the eukaryotic 80S ribosome (4 MDa) have attained this level of resolution (e.g. emd-1217) by using around 100 000 particles (Halic et al., 2006). Otherwise, structural studies of asymmetric complexes using a relatively low number of particles would yield structures in the range 10–20Å resolution, as that obtained

eLS © 2015, John Wiley & Sons, Ltd. www.els.net

9

Single Particle EM

Archaeal 20s proteasome 700 kDa, 3.3 Å (a)

TRPV1 ion channel 300 kDa, 3.4 Å (b)

Yeast mitoribosomal large subunit 1.9 MDa, 3.2 Å (c)

Human Y-secretase 170 kDa, 4.5 Å (d)

Figure 7 Recent major breakthroughs by high-resolution single particle cryo-EM. The use of direct detection device cameras and compensation for beam-induced specimen motion has been of paramount importance. As in Figure 6, the actual resolution achieved and the molecular mass are indicated for all structures. Maps in (a–d) are accessible through the Electron Microscopy Data Bank under the following accession codes: emd-5623, emd-5778, emd-2566 and emd-2677, respectively. Bar, 5nm.

for the human deoxyribonucleic acid (DNA)-dependent protein kinase catalytic subunit (470 kDa, emd-1102), which has a central role in DNA double-strand break repair, computed from 7000 particles (Rivera-Calzada et al., 2005). Figure 6(h–i) shows examples of small proteins where negative staining is indispensable to derive the structure by single particle EM. As described earlier, the resolution in these studies is limited to around 15–20 Å. Human geminin, a protein that ensures one round of DNA replication during each cell cycle, is one of the smallest asymmetrical proteins (90 kDa) whose structure (emd-1190) has been determined by this technique by collecting and processing 3300 particles (Okorokov et al., 2004). Archaeal prefoldin, a small complex (87 kDa) that stabilises and delivers unfolded proteins to the chaperonin that assists its folding, was solved at 19 Å resolution by single particle EM using about 3800 particles and twofold symmetry (Martín-Benito et al., 2007). Finally, Figure 7 focuses on very recent major breakthroughs determined at near-atomic resolution from relatively few thousands of particles with DDD cameras and motion correction. Archaeal 20S proteasome (emd-5623), a protein complex of 700 kDa that degrades unnecessary or degraded proteins, was one of the first examples showing the benefits from these technical and computational advancements (Li et al., 2013). It was solved at 3.3 Å from 126 729 particles using 14-fold symmetry. The structure of the mammalian TRPV1, the ion channel responsible for sensing heat, determined at 3.4 Å from just 35 645 particles and fourfold symmetry (emd-5778) was a remarkable achievement (Liao et al., 2013). The large subunit of the yeast mitochondrial ribosome was solved at 3.2 Å from just 47 124 particles (emd-2566), which was a striking application of the technique to an asymmetric complex (Amunts et al., 2014). These three specimens might be thought of well suited to high-resolution cryo-EM (e.g. internal symmetry or molecular mass in the MDa range). However, the structure of the human γ-secretase (emd-2677), a protease that cleaves other proteins inside the membrane and has 10

an important role in Alzheimer’s disease, has recently been solved at 4.5 Å from 144 545 particles (Lu et al., 2014). This is an asymmetric membrane protein, previously thought to be very small (170 kDa) for cryo-EM, and its structure turns out to be a landmark success that indicates that high-resolution cryo-EM may be applicable to a large variety of proteins. See also: Electron Cryomicroscopy; Electron Cryomicroscopy and Three-dimensional Computer Reconstruction of Biological Molecules; Molecular Dynamics; Optical Mapping; Proteins: Fundamental Chemical Properties

Acknowledgements The authors thank the support of the grants MINECO-TIN201237483-C03-02 and MINECO-BFU2013-44202-P.

References Adrian M, Dubochet J, Fuller SD and Harris JR (1998) Cryonegative staining. Micron 29: 145–160. Agirrezabala X, Martín-Benito J, Valle M, et al. (2005) Structure of the connector of bacteriophage T7 at 8Å resolution: structural homologies of a basic component of a DNA translocating machinery. Journal of Molecular Biology 347: 895–902. Amunts A, Brown A, Bai XC, et al. (2014) Structure of the yeast mitochondrial large ribosomal subunit. Science 343 (6178): 1485–1489. Baker ML, Ju T and Chiu W (2007) Identification of secondary structure elements in intermediate-resolution density maps. Structure 15 (1): 7–19. Baker ML, Zhang J, Ludtke SJ, et al. (2010) Cryo-EM of macromolecular assemblies at near-atomic resolution. Nature Protocols 5: 1697–1708.

eLS © 2015, John Wiley & Sons, Ltd. www.els.net

Single Particle EM

Baker TS and Henderson R (2012) Electron cryomicroscopy of biological macromolecules. International Tables for Crystallography F Chapter 19.6: 593–614. Bai XC, Fernandez IS, McMullan G, et al. (2013) Ribosome structures to near-atomic resolution from thirty thousand cryo-EM particles. eLife 2: e00461. Brenner S and Horne RW (1959) A negative staining method for high resolution electron microscopy of viruses. Biochimica et Biophysica Acta 34: 103–110. Brilot AF, Chen JZ, Cheng A, et al. (2012) Beam-induced motion of vitrified specimen on holey carbon film. Journal of Structural Biology 177 (3): 630–637. Campbell MG, Cheng A, Brilot AF, et al. (2012) Movies of ice-embedded particles enhance resolution in electron cryo-microscopy. Structure 20 (11): 1823–1828. Cardone G, Heymann JB and Steven AC (2013) One number does not fit all: mapping local variations in resolution in cryo-EM reconstructions. Journal of Structural Biology 184 (2): 226–236. Chen S, McMullan G, Faruqi AR, et al. (2013) High-resolution noise substitution to measure overfitting and validate resolution in 3D structure determination by single particle electron cryomicroscopy. Ultramicroscopy 135: 24–35. Dubochet J, Adrian M, Chang JJ, et al. (1988) Cryo-electron microscopy of vitrified specimens. Quarterly Reviews of Biophysics 21: 129–228. Faruqi AR and Henderson R (2007) Electronic detectors for electron microscopy. Current Opinion in Structural Biology 17 (5): 549–555. Fernandez JJ, Luque D, Caston JR, et al. (2008) Sharpening high resolution information in single particle electron cryomicroscopy. Journal of Structural Biology 164: 170–175. Grigorieff N (2013) Direct detection pays off for electron cryomicroscopy. eLife 2: e00461. Halic M, Gartmann M, Schlenker O, et al. (2006) Signal recognition particle receptor exposes the ribosomal translocon binding site. Science 312: 745–747. Henderson R, Chen S, Chen JZ, et al. (2011) Tilt-pair analysis of images from a range of different specimens in single-particle electron cryomicroscopy. Journal of Molecular Biology 413 (5): 1028–1046. Henderson R (1995) The potential and limitations of neutrons, electrons and X-rays for atomic resolution microscopy of unstained biological molecules. Quarterly Reviews of Biophysics 28 (2): 171–193. Kaufmann B, Chipman PR, Kostyuchenko VA, et al. (2008) Visualization of the externalized VP2 N termini of infectious human parvovirus B19. Journal of Virology 82: 7306–7312. Kucukelbir A, Sigworth FJ and Tagare HD (2014) Quantifying the local resolution of cryo-EM density maps. Nature Methods 11 (1): 63–65. Lawson CL, Baker ML, Best C, et al. (2011) EMDataBank.org: unified data resource for CryoEM. Nuclear Acids Research 39 (suppl 1): D456–D464. Li X, Mooney P, Zheng S, et al. (2013) Electron counting and beam-induced motion correction enable near-atomic-resolution single-particle cryo-EM. Nature Methods 10 (6): 584–590. Liao M, Cao E, Julius D, et al. (2013) Structure of the TRPV1 ion channel determined by electron cryo-microscopy. Nature 504 (7478): 107–112.

Ludtke SJ, Baker ML, Chen DH, et al. (2008) De novo backbone trace of GroEL from single particle electron cryomicroscopy. Structure 16: 441–448. Llorca O, Martín-Benito J, Grantham J, et al. (2001) The ‘sequential allosteric ring’ mechanism in the eukaryotic chaperonin-assisted folding of actin and tubulin. EMBO Journal 20: 4065–4075. Lu P, Bai XC, Ma D, et al. (2014) Three-dimensional structure of human 𝛾-secretase. Nature 512 (7513): 166–170. Luque D, Saugar I, Rodriguez JF, et al. (2007) Infectious bursal disease virus capsid assembly and maturation by structural rearrangements of a transient molecular switch. Journal of Virology 81: 6869–6878. Martín-Benito J, Gómez-Reino J, Stirling PC, et al. (2007) Divergent substrate-binding mechanisms reveal an evolutionary specialization of eukaryotic prefoldin compared to its archaeal counterpart. Structure 15: 101–110. Okorokov AL, Orlova EV, Kingsbury SR, et al. (2004) Molecular structure of human geminin. Nature Structural and Molecular Biology 11: 021–1022. Rosenthal PB and Henderson R (2003) Optimal determination of particle orientation, absolute hand, and contrast loss in single-particle electron cryomicroscopy. Journal of Molecular Biology 333 (4): 721–745. Rivera-Calzada A, Maman JD, Spagnolo L, et al. (2005) Threedimensional structure and regulation of the DNA-dependent protein kinase catalytic subunit (DNA-PKcs). Structure 13: 243–255. Russo CJ and Passmore LA (2014) Ultrastable gold substrates for electron cryomicroscopy.. Science 346: 1377–1380. Scheres SH (2014) Beam-induced motion correction for submegadalton cryo-EM particles. eLife 3: e03665. Scheres SH and Chen S (2012) Prevention of overfitting in cryo-EM structure determination. Nature Methods 9 (9): 853–854. Villa E and Lasker K (2014) Finding the right fit: chiseling structures out of cryo-electron microscopy maps. Current Opinion in Structural Biology 25: 118–125. Yu X, Jin L and Zhou ZH (2008) 3.88 Å structure of cytoplasmic polyhedrosis virus by cryo-electron microscopy. Nature 453: 415–419. Zhang X, Settembre E, Xu C, et al. (2008) Near-atomic resolution using electron cryomicroscopy and single-particle reconstruction. Proceedings of the National Academy of Sciences of USA 105: 1867–1872.

Further Reading Fernandez JJ, Sorzano COS, Marabini R and Carazo JM (2006) Image processing and 3-D reconstruction in electron microscopy. IEEE Signal Processing Magazine 23 (3): 84–94. Frank J (2006) Three-Dimensional Electron Microscopy of Macromolecular Assemblies: Visualization of Biological Molecules in Their Native State. New York, NY: Oxford University Press. Frank J (2009) Single-particle reconstruction of biological macromolecules in electron microscopy – 30 years. Quarterly Reviews of Biophysics 42 (3): 139–158. van Heel M, Gowen B, Matadeen R, et al. (2000) Single-particle electron cryo-microscopy: towards atomic resolution. Quarterly Reviews of Biophysics 33: 307–369. Henderson R (2013) Avoiding the pitfalls of single particle cryo-electron microscopy: Einstein from noise. Proceedings of the National Academy of Sciences of USA 110 (45): 18037–18041.

eLS © 2015, John Wiley & Sons, Ltd. www.els.net

11

Single Particle EM

Henderson R, Sali A, Baker ML, et al. (2012) Outcome of the first electron microscopy validation task force meeting. Structure 20 (2): 205–214. Jensen GJ (2010a) Cryo-EM, part A: sample preparation and data collection. Methods in Enzymology 481: 2–410. Jensen GJ (2010b) Cryo-EM, part B: 3-D reconstruction. Methods in Enzymology 482: 2–410.

12

Jensen GJ (2010c) Cryo-EM, part C: analyses, interpretation, and case studies. Methods in Enzymology 483: 2–360. Orlova EV and Saibil HR (2011) Structural analysis of macromolecular assemblies by electron microscopy. Chemical Reviews 111 (12): 7710–7748.

eLS © 2015, John Wiley & Sons, Ltd. www.els.net