Table of Contents

Editor biographies xv

Section authors xvi

1 Introduction and overview 1-1

1.1 Visualizing biological molecules to understand life's principles 1-1

1.1.1 A brief historical perspective on scattering-based structural biology methods 1-1

1.1.2 Unique capabilities of cryo-EM: polymers and viruses 1-3

1.1.3 Unique capabilities of cryo-EM: integral membrane proteins 1-4

1.1.4 Unique capabilities of cryo-EM: large assemblies 1-5

1.1.5 Unique capabilities of cryo-EM: scarce samples 1-5

1.1.6 Unique capabilities of cryo-EM: compositionally heterogeneous samples 1-6

1.1.7 Unique capabilities of cryo-EM: conformationally complex samples 1-7

1.1.8 Current limits of cryo-EM and things yet to come 1-8

1.2 Recovery of 3D structures from images of weak-phase objects 1-9

1.2.1 The signal that we care about is attributed to elastic scattering of electrons 1-10

1.2.2 The electron accumulates information as it passes through a specimen 1-10

1.2.3 The image wave function, and thus the image intensity, suffers from imperfections in the microscope optics 1-12

1.2.4 Intermediate summary: the image intensity is linear in the projected Coulomb potential of the object 1-14

1.2.5 Structure-factor phases, as well as amplitudes, are retained in the computed Fourier transforms of image intensities 1-15

1.2.6 The projection theorem: the Fourier transform of an image corresponds to a 2D 'central' section within the 3D Fourier transform of the object 1-17

1.2.7 The 3D object can be reconstructed from multiple projections 1-18

1.2.8 Similarities and differences between sub-tomogram averaging and single-particle cryo-EM 1-19

References 1-20

2 Sample preparation 2-1

2.1 Overview 2-1

2.2 Initial screening of samples in negative stain 2-2

2.2.1 Introduction 2-2

2.2.2 Negative staining for TEM 2-2

2.2.3 Purpose of negative staining when starting a project 2-3

2.2.4 Techniques for the preparation of negatively stained samples 2-4

2.2.5 Use of data processing to provide feedback to optimize samples for cryo-EM 2-10

2.3 Standard method of making grids for cryo-EM 2-13

2.3.1 Grids and support films 2-14

2.3.2 Plasma cleaning or 'glow discharging' grids 2-15

2.3.3 Types of apparatus used for plunge freezing 2-17

2.3.4 Blotting and plunging the grid using plunge freezers 2-18

2.3.5 Common issues faced in making grids for cryo-EM imaging 2-19

2.4 Requirement to make very thin specimens for cryo-EM 2-19

2.4.1 Inelastic electron scattering causes the image quality to deteriorate with increasing sample thickness values 2-20

2.4.2 The projection approximation may fail if the sample is too thick 2-21

2.4.3 Areas of a grid where the sample is obviously too thick can, and should be, avoided during data collection 2-21

2.4.4 Areas where the sample is much too thin, perhaps even air-dried, can sometimes be avoided just on the basis of their subjective appearance 2-23

2.5 Current strategies for optimizing preparation of cryo-grids 2-25

2.5.1 Behavior of particles in the thin film environment 2-25

2.5.2 Approaches to alter particle behavior in the thin film 2-28

2.5.3 New technologies for sample preparation 2-32

References 2-37

3 Data collection 3-1

3.1 Overview 3-1

3.2 Radiation damage in cryo-EM 3-2

3.2.1 Introduction 3-2

3.2.2 Interaction cross sections, elastic, and inelastic interactions 3-2

3.2.3 Cryoprotection and primary, secondary, and tertiary radiation damage 3-3

3.2.4 Radiation damage dependence on electron energy 3-4

3.2.5 Practical implications of radiation damage: image averaging in cryo-EM 3-5

3.2.6 Resolution dependence and exposure weighting 3-6

3.2.7 Radiation damage versus beam-induced motion and charging 3-9

3.3 Low-dose protocols for recording images 3-10

3.3.1 Automated low-dose imaging 3-10

3.3.2 Improving throughput 3-13

3.3.3 Electron exposure levels used during high-resolution data collection 3-15

3.4 Practical considerations: defocus. stigmation, coma-free illumination, and phase plates 3-15

3.4.1 Why do we need to defocus the microscope? 3-15

3.4.2 Effects of defocus on the image and its information content 3-16

3.4.3 Defocus variation is necessary to obtain uniform information coverage in reciprocal space 3-17

3.4.4 Optical correction of astigmatism and coma aberrations 3-19

3.4.5 Use of phase plates to improve image contrast and the expected benefits 3-20

3.5 Practical considerations: movie-mode data acquisition 3-23

3.5.1 Magnification and resolution 3-24

3.5.2 Dose rate 3-25

3.5.3 Strategies for motion correction 3-26

3.5.4 Total dose or exposure time 3-29

3.5.5 File size of movie datasets 3-29

3.5.6 Summary 3-29

References 3-30

4 Data processing 4-1

4.1 Overview 4-1

4.2 Automated extraction of particles 4-2

4.2.1 From micrographs to particles 4-2

4.2.2 Manual selection 4-3

4.2.3 Unbiased automated approaches 4-4

4.2.4 Particle extraction 4-8

4.2.5 Cleaning up the results through classification 4-9

4.3 CTF estimation and image correction (restoration) 4-10

4.3.1 CTF estimation 4-13

4.3.2 Image correction 4-16

4.3.3 Magnification distortion 4-18

4.3.4 Concluding remarks 4-20

4.4 Merging data from structurally homogeneous subsets 4-20

4.4.1 How many particle images are needed for a 3D reconstruction? 4-21

4.4.2 Obtaining a 3D reconstruction 4-28

Acknowledgments 4-37

4.5 3D classification of structurally heterogeneous particles 4-37

4.5.1 Introduction 4-37

4.5.2 Global 3D classification 4-38

4.5.3 Masked 3D classification 4-40

4.5.4 3D classification of particles with pseudo-symmetry 4-43

4.5.5 Dealing with continuous motions 4-44

4.5.6 Conclusion 4-45

4.6 Preferred orientation: how to recognize and deal with adverse effects 4-45

4.6.1 Protein interaction with the air-water interface 4-46

4.6.2 Preferred orientation and its effects in cryo-EM 4-47

4.6.3 Quantifying preferred orientation and its effects on cryo-EM reconstructions 4-50

4.6.4 Overcoming the effects of preferred orientation 4-53

4.6.5 Areas of research 4-57

Acknowledgments 4-59

4.7 B factors and map sharpening 4-59

4.7.1 An ideal 3D reconstruction has a predictable radial amplitude spectrum 4-59

4.7.2 Actual 3D reconstructions feature dampened amplitudes at high frequencies 4-59

4.7.3 Several factors contribute to signal decay at high frequencies 4-62

4.7.4 Gaussian falloff, parametrized by a B factor, is a useful model of signal loss 4-62

4.7.5 Estimating B factors 4-63

4.7.6 Sharpening a map 4-64

4.7.7 A single inverse Gaussian filter using a global B factor does not always lead to the optimal map 4-65

4.8 Optical aberrations and Ewald sphere curvature 4-67

4.8.1 Further considerations on the aberration function γ(s) 4-68

4.8.2 Common types of aberrations 4-70

4.8.3 Practical considerations for aberration correction 4-72

4.8.4 Thick objects and the Ewald sphere 4-76

4.8.5 Ewald sphere correction 4-77

References 4-80

5 Map validation 5-1

5.1 Overview 5-1

5.2 Measures of resolution: FSC and local resolution 5-2

5.2.1 The 'gold-standard' FSC 5-5

5.2.2 Resolution thresholds 5-5

5.2.3 FSC artifacts due to masking, filtration, and CTF 5-6

5.2.4 Local resolution 5-8

5.2.5 Resolution anisotropy 5-10

5.3 Recognizing the effect of bias and over-fitting 5-10

5.3.1 Introduction and nature of the problem arising from iterative refinement 5-10

5.3.2 Assessing the consistency of maps with projection data 5-12

5.3.3 Detecting over-fitting at high resolution in maps and effect on the FSC 5-15

5.3.4 Local over-fitting 5-18

5.3.5 Conclusion 5-21

5.4 Estimates of alignment accuracy 5-21

5.4.1 Correlation and the signal-to-noise ratio (SNR) 5-22

5.4.2 Analysis of alignment accuracy with synthetic data 5-23

5.4.3 The relationship between alignment accuracy and resolution 5-25

5.4.4 Estimating alignment accuracy from tilt pairs 5-27

5.4.5 Estimating alignment accuracy from the reconstructed map 5-27

5.4.6 Estimating alignment accuracy from projection-matching results 5-28

5.5 Discussion 5-28

Acknowledgements 5-29

References 5-29

6 Model building and validation 6-1

6.1 Overview 6-1

6.2 Using known components or homologs: model building 6-2

6.2.1 Identifying known/modeled structures of individual subunits 6-3

6.2.2 Rigid-body fitting 6-4

6.2.3 Flexible fitting 6-4

6.3 Building atomistic models in cryo-EM density maps 6-6

6.3.1 Introduction 6-6

6.3.2 Building models into cryo-EM density maps 6-10

6.3.3 Model refinement 6-15

6.3.4 Model validation 6-17

6.3.5 Model uncertainty 6-20

6.3.6 Model deposition 6-21

6.3.7 Revisiting the cryo-EM model challenge 6-22

6.3.8 Toward the future 6-22

6.3.9 Conclusions 6-25

Acknowledgments 6-25

6.4 Quality evaluation of cryo-EM map-derived models 6-26

6.4.1 Introduction 6-26

6.4.2 Map-model metrics 6-27

6.4.3 Model-only metrics 6-37

6.4.4 Summary and conclusions 6-39

Acknowledgment 6-40

6.5 How algorithms from crystallography are helping electron cryo-microscopy 6-40

6.5.1 Introduction 6-40

6.5.2 Map improvement 6-42

6.5.3 Map interpretation and model building 6-45

6.5.4 Model optimization 6-48

6.5.5 Validation 6-50

6.5.6 Validation-guided corrections 6-52

6.5.7 Conclusions 6-54

Acknowledgments 6-54

6.6 Archiving structures and data 6-55

6.6.1 Introduction 6-55

6.6.2 Single-particle cryo-EM structure deposition 6-56

6.6.3 Preparing files for deposition 6-57

6.6.4 Data validation 6-59

6.6.5 Sample sequence and ligands 6-61

6.6.6 Deposition using OneDep 6-61

6.6.7 Post-deposition: what happens next? 6-64

6.6.8 Accessing cryo-EM structure data 6-65

6.6.9 Conclusions 6-66

Acknowledgment 6-66

References 6-66