Technical Challenges Nikolaus Grigorieff Brandeis University - PowerPoint PPT Presentation

Technical Challenges Nikolaus Grigorieff Brandeis University Larson, The Far Side

What Technical Challenges? 3.6 Å resolution Wolf et al. 2010

An Old Prophecy How many images must be averaged to reach near-atomic resolution? Theoretical prediction (Henderson, Glaeser): a few thousand Papillomavirus L1 subunits averaged to reach 3.6 Å: 1.5 million!

Contrast and Noise 8.3 Å SNR 1/10 ≈ Bacteriorhodopsin

Aligning Small Particles 100 kDa Bacteriorhodopsin

B-Factor Analysis 0.5 native bR ln(image/diffraction amplitudes) native bR 0 (2.8 Å) DOC bR -0.5 B = 140 Å 2 -1 (200 kV, FEG, l. helium, -1.5 10 e - /Å 2 ) B = 180 Å 2 -2 (120 kV, tungsten -2.5 10 lipids filament, l. helium, -3 10 e - /Å 2 ) per bR -3.5 0 0.02 0.04 0.06 0.08 0.1 DOC bR Resolution [Å -2 ] 2.6 Å) B damage = 40 Å 2 (liquid helium) B detector = 70 Å 2 (film) → B total = 110 Å 2 6 lipids per bR Grigorieff et al. 1995

B-Factor Analysis 3.6 Å resolution 3,977 particles B damage = 60 Å 2 (liquid nitrogen) 60-fold icos. sym. 6-fold non-icos. sym. B detector = 70 Å 2 (film) B motion = 160 Å 2 (Campbell et al. 2012) Wolf et al. 2010 B alignment = 90 Å 2 (σ shift = 0.2 Å, σ rot = 0.2º, 12 σ defocus = 200 Å) Papillomavirus without 10 non-icosahedral averaging → B total = 380 Å 2 ln(amplitudes) (resolution = 4.4 Å) 8 B = 510 Å 2 6 B observed = 510 Å 2 B = 260 Å 2 4 → B unexplained = 130 Å 2 2 0 0.01 0.02 0.03 0.04 0.05 Resolution [Å -2 ]

B-Factor Analysis 3.6 Å resolution 3,977 particles 60-fold icos. sym. 6-fold non-icos. sym. → 24,000 “60-fold” particles Wolf et al. 2010 12 Papillomavirus without 10 non-icosahedral averaging ln(amplitudes) (resolution = 4.4 Å) 8 B = 510 Å 2 6 B = 260 Å 2 4 2 0 0.01 0.02 0.03 0.04 0.05 Resolution [Å -2 ] Rosenthal & Henderson 2003

The Challenges • Beam-induced motion & charging • Detector DQE • Beam damage • Alignment errors Larson, The Far Side

Better Sample Support Thin carbon support shows crinkling due to shrinking of the copper grid (0.3%) and paraffin crystals (1% – 2%) at liquid nitrogen temperature. Thick carbon (350 Å) reduces or eliminated movement. → Flatness and mechanical strength of the support film are important. Molybdenum grids may help. 60º tilt Glaeser et al. 2011

Movies Recorded with direct electron detector DE-12 (Direct Electron) Frame rate = 40 fps Dose/frame = 0.5 e - /Å 2 Duration = 1.5 s No. of frames = 60 Total dose = 30 e - /Å 2 1 movie = 720 MB (1 byte/pixel) → Data Tsunami! Brilot et al. 2012

Frame Alignment 60-frame average 60-frame average (no alignment) (translational alignment) Brilot et al. 2012

Paraxial Charge Compensation C2 Berriman & Rosenthal 2012

The Challenges • Beam-induced motion & charging Contrast • Detector DQE • Beam coherence • Alignment errors

Improving Contrast • Better detectors • Low voltage • Phase plate • Inelastic scattering • Astigmatism Larson, The Far Side

Direct Electron Detectors 300 kV Chris Booth (Gatan) Benjamin Bammes (Direct Electron) Panel discussion (David Agard) McMullan & Henderson, 2009

Perfect Detector Δ = 22% McMullan & Henderson, 2009 Gatan webpage, 2012

Low Voltage Trimers of HIV gp140 in ice M = 420 kDa 80 kV 20 e - /Å 2 DQE of film and scintillator-based cameras improved at lower voltage 500 Å Harris et al 2011

Phase Plate GroEL in ice 20 nm 20 nm Defocus contrast Zernike phase plate Wah Chiu, Bob Glaeser Danev & Nagayama 2008

Phase Plate with Lasers Elastic Compton scattering Spherical resonant cavity 40 W laser with λ = 2 μm Müller et al. 2010

Inelastic Scattering Unfiltered 0 eV 25 eV 300 kV, 6 μm underfocus, 15 eV energy window Assuming 700 Å sample thickness: Electrons scattered elastically: 9% scattered inelastically: 18% → C c correctors will increase image contrast. Chen Xu (unpublished)

Astigmatic CTF Hemocyanine 3.8 MDa D 2 symmetry Martin et al. 2007 DF 1 = 1000 Ǻ, DF 2 = 14000 Ǻ Grant & van Heel (unpublished)

Potential Improvements 1 1 1 B=520 Å 2 B=520 Å 2 B=520 Å 2 0.8 0.8 0.8 B=360 Å 2 B=360 Å 2 B damage = 60 Å 2 B=290 Å 2 Envelope Envelope Envelope 0.6 0.6 0.6 B detector = 70 Å 2 0.4 0.4 0.4 B motion = 160 Å 2 B alignment = 90 Å 2 0.2 0.2 0.2 0 0 0 0 0 0 0.05 0.05 0.05 0.1 0.1 0.1 0.15 0.15 0.15 0.2 0.2 0.2 0.25 0.25 0.25 0.3 0.3 0.3 Factor @ 3.5 Å: Resolution[Å -1 ] Resolution[Å -1 ] Resolution[Å -1 ] 110 12,000 x signal

The Challenges • Beam-induced motion & charging • Detector DQE • Beam damage • Alignment errors Larson, The Far Side

Optimal Dose 2.5 N e = 12 e - /Å 2 2 (50 Å resolution) SNR (arb. units) 1.5 N e = 6 e - /Å 2 (7 Å resolution) 1 19x 5x 0.5 B = 60 Å 2 N e = 1.2 e - /Å 2 B = 30 Å 2 (3 Å resolution) 0 0 5 10 15 20 25 30 Dose (electrons/Å 2 ) John Rubinstein Unwin & Henderson 1975; Hayward & Glaeser 1979; Stark et al. 1996, Baker et al. 2010

High-Dose Imaging 104 electrons/Å 2 24 electrons/Å 2 Voltage = 200 kV Defocus = 1.5 μm Flu hemagglutinin (2FK0, Stevens et al. 2006) Trimer = 180 kDa Melody Campbell Peter Lee (unpublished)

Thon Ring Patterns 6 Å No frame alignment With frame alignment Melody Campbell Peter Lee (unpublished)

Potential Improvements 1 1 1 1 B=520 Å 2 B=520 Å 2 B=520 Å 2 B=520 Å 2 0.8 0.8 0.8 0.8 B=360 Å 2 B=360 Å 2 B=360 Å 2 B damage = 30 Å 2 B=290 Å 2 B=290 Å 2 Envelope Envelope Envelope Envelope 0.6 0.6 0.6 0.6 B detector = 70 Å 2 B=260 Å 2 0.4 0.4 0.4 0.4 B motion = 160 Å 2 B alignment = 90 Å 2 0.2 0.2 0.2 0.2 0 0 0 0 0 0 0 0 0.05 0.05 0.05 0.05 0.1 0.1 0.1 0.1 0.15 0.15 0.15 0.15 0.2 0.2 0.2 0.2 0.25 0.25 0.25 0.25 0.3 0.3 0.3 0.3 Factor @ 3.5 Å: Resolution[Å -1 ] Resolution[Å -1 ] Resolution[Å -1 ] Resolution[Å -1 ] 200 40,000 x signal

The Challenges • Beam-induced motion & charging • Beam damage • Detector DQE • Alignment errors Larson, The Far Side

Dealing With Noise 100 kDa Bacteriorhodopsin

Maximum Likelihood Correlation alignment N = 4000 SNR = 1/200 ML estimation Sjors Scheres Sigworth 1998

ML Classification SNR = 1/50 Difference N = 2000 map Sjors Scheres Pawel Penczek Correlation alignment

All Things Considered • Accurate modeling of all parameters – CTF, envelope, variability … • Statistical models – noise models, parameter distributions, weighting … • Score dependent on all known facts/data – cross-linking, homology, total mass … • Reproducibility tests – multiple starts, consistency checks … • Large data sets – automation

Helical Processing with Frealix Emphasis on flexible filaments Rotational alignment of a single crossover (amyloid fibrils) Correlation in neighboring areas Full-filament processing helps find correct alignment (no segment boxing) Other filament types (TMV, microtubules) Constraints during image processing low score (persistence length...) high score Alexis Rohou, unpublished

Computer Games (Doom et al.) 10 6 7 10 5 10 6 10 10 4 Voxel MIPS 5 10 1000 4 10 100 10 1000 1990 1995 2000 2005 2010 Year

Sample Limitations • Sample heterogeneity/stability – biochemistry – new algorithms • Transient complexes – affinity grids, streptavidin crystals • Detergent and lipid – amphipol, GLC/GDN Larson, The Far Side – amphiphilic β-strand peptides • Low molecular weight Holger Stark Debbie Kelly

Scaffolds Yifan Cheng HA 6-helix bundle adapter VP7 Ni-NTA nanogold Rotavirus DLP His-tagged flu hemagglutinin Junhua Pan, Yuhang Liu, unpublished unpublished

Thank You!