Serial Femtosecond Protein Crystallography SFX. Henry Chapman Center for Free-Electron Laser Science DESY and University of Hamburg

Size: px

Start display at page:

Download "Serial Femtosecond Protein Crystallography SFX. Henry Chapman Center for Free-Electron Laser Science DESY and University of Hamburg"

Nigel Washington
5 years ago
Views:

1 Serial Femtosecond Protein Crystallography SFX Henry Chapman Center for Free-Electron Laser Science DESY and University of Hamburg UXSS, SLAC, June 2012

$diffraction$ experiment was

2 The first X-ray diffraction experiment was carried out 100 years ago o qlb 'O' B I ' O O, t B o Q o,p W. Friedrich, P. Knipping & M. Laue (1912) Sitzungsberichte (München), pp

3 The number of protein structures solved is now increasing linearly Cumulative number of structures in the PDB ribosome 80,000 70,000 virus myosin 60,000 Structures 50,000 40,000 30,000 20,000 transfer RNA antibody nucleosome 10,000 0 myoglobin 1972 hemoglobin 1978 actin Year X-ray NMR Electron

4 Space-grown Insulin crystals NASA

5 The scattered field is a sum of paths over point scatterers f(q) k2 4 (1 n 2 (x)) exp(iq x) dx f(q) = r e (x) exp(iq x) dx scattering potential a density of point scatterers k out q = k out k in f 1 k =2 / k in x 1 1 = x 1 k out x 1 k in = x 1 q O f = f 1 exp i 1 5

6 The scattered field is a sum of paths over point scatterers f(q) k2 4 (1 n 2 (x)) exp(iq x) dx f(q) = r e (x) exp(iq x) dx scattering potential a density of point scatterers k out q = k out k k f in in i f = x i f(q) = i i f i exp i i f i exp(iq x i ) O 6

The scattered wave is the Fourier transform of the scattering potential f(q) = r e (x) exp(iq x) dx r e (x) = 1 2 3D#Fourier#transform f(q) exp( i q x) dq Inverse#3D#Fourier#transform Consider light

7 The scattered wave is the Fourier transform of the scattering potential f(q) = r e (x) exp(iq x) dx r e (x) = 1 2 3D#Fourier#transform f(q) exp( i q x) dq Inverse#3D#Fourier#transform Consider light that is scattered to one particular f(q) = (q 2 G) r e (x) = 1 (q 2 G) exp( i q x) dq 2 = exp( 2 ig x) G =1/d This%is%a%Volume'gra+ng'with% period%1/ G %in%the%direc0on%of%g q =2 G The%diffrac0on%condi0on%is%simply% q =2 G q =2 G =2 /d q = 4 sin d 2 7 =2d sin Bragg s Law

8 Only specific components of a structure diffract an incident monochromatic wave k out q 1 =2d sin d = 1 2 sin 2 kin = kout - kin Ewald&sphere Diffraction direction: k out k in = q q = 2 d k = 2π λ 8

$A single diffraction pattern only records 2D data on$ direction of kin (rotate the sample) kin Vary

9 A single diffraction pattern only records 2D data on the Ewald sphere Ways to collect full 3D: Vary direction of kin (rotate the sample) kin Vary magnitude of kin (change wavelength) 9

10 The weak X-ray scattering cross section requires amplification from the crystal signal is proportional to the number of unit cells

11 The weak X-ray scattering cross section requires amplification from the crystal signal is proportional to the number of unit cells

12 The truncated lattice F (u) =M(x) {L(u) S(u)} l(x) = S(u) f(x) =m(x) {l(x) s(x)} n= (x na) L(u) = N a sinc N au = N = Na h h = sin( Nau) sin( au) L(u) =1/a 1 a h (u sinc Nau (u sinc N(au h) h h a ) h a ) a Na (u h/a) N 2 1 Na s(x) = rect ( x Na ) x S(u) =Nasinc Nau 1/a N wiggles including the two Bragg peaks

13 Crystallography can produce atomic resolution 3D images X-ray diffraction pattern of crystallized 3Clpro, a SARS protease Jeff Dahl, wikipedia.org 12

$Coherent diffractive imaging (including crystallography) is lensless$

14 Coherent diffractive imaging (including crystallography) is lensless λ θ Prior knowledge about object λ θ Algorithm Resolution: δ = λ /sinθ

15 Images are synthesized from the Fourier amplitudes M(q) m(x) Fourier transformation q 2 / q M(q) q

16 Images are synthesized from the Fourier amplitudes M(q) Fourier transformation m(x) M(q) q 290 nm

17 Bragg sampling excludes information about the molecular transform M(q) M(q) L(q) m(x) l(x) 290 nm

18 Bragg sampling excludes information about the molecular transform M(q) L(q) m(x) l(x) 290 nm

High radiation dose causes changes in molecular structure Tolerable dose in cryogenicallycooled crystals is 30 MGy 1 Gy = 1 J/kg Elspeth Garman, U.

19 High radiation dose causes changes in molecular structure Tolerable dose in cryogenicallycooled crystals is 30 MGy 1 Gy = 1 J/kg Elspeth Garman, U. Oxford micrograph of crystal after exposing to x-rays and warming up 30 MGy 0.3 ev / Da 0.02 ev / atom (about one ionization per 20 amino-acid residues) ph/µm 2

20 X-ray free-electron lasers may enable atomicresolution imaging of biological macromolecules 2 fs 5 fs 10 fs 20 fs 50 fs R. Neutze, R. Wouts, D. van der Spoel, E. Weckert, J. Hajdu, Nature 406 (2000)

21 We first demonstrated diffraction before destruction at FLASH 1 micron 1st shot at full power 19 Chapman et al. Nature Physics (2006) 1 micron SEM of structure etched into silicon nitride membrane

22 First EUV-FEL experiments show that pulses are indeed destructive ϑ Nomarski micrograph of crater Si/C multilayer 40 micron After the pulse Plasma forms, layers ablate Electron temperature reaches 28 ev (300,000 K)

23 First EUV-FEL experiments show that structural information can be obtained before destruction ϑ Si/C multilayer Reflectance (%) % 16 % 30 fs pulse reflectivity at 32 nm wavelength, W/cm 2 increasing fluence Low-fluence Reflectivity unchanged Multilayer d spacing not changed by more than 0.3 nm Angle of incidence ϑ (degrees) 50 S. Hau-Riege et al. PRL 98, (2007)

24 We used the same strategy as at FLASH to monitor sample destruction during the pulse ϑ 0 fs 20 fs Si/C multilayer Stefan Hau-Riege, LLNL FLASH: Wavelength 100 Å Structures: 100 Å to microns LCLS: Wavelength 6.8 Å Structures: 6 Å to microns

25 a = b = 288 Å c = 167 Å Petra Fromme, ASU

26 Nanocrystallography is carried out in a flowing water microjet 80 mm 200 µm Liquid jet Liquid jet s e s l u p y a r SX LCL Interaction point Interaction point Front detector (564 mm) (68 mm) Front pnccd (z = 68 mm) CAMP Chamber: L. Struder et al µm crystals 4 µm jet Rear pnccd Rear (z = detector 564 mm) 3 µm X-ray beam

27 Samples are delivered to the beam in a liquid jet 20 µm Dan Deponte, CFEL

28 Optical emission is observed for dose rates above about 20 MGy/fs

29 a = b = 288 Å c = 167 Å Petra Fromme, ASU

30 Single pulse diffraction from Photosystem 1 nanocrystals at LCLS Upper front CCD beam center Single shot at LCLS E = 1.8 kev 80 fs pulse 2 mj pulse energy Lower front CCD Resolution at corner = 8.6Å

31 Femtosecond-pulse nanocrystal data appears better than diffraction from large crystals Petra Fromme, Mark Hunter, ASU

$diffraction from large$

32 Femtosecond-pulse nanocrystal data appears better than diffraction from large crystals Petra Fromme, Mark Hunter, ASU

33 The crystal shape can be used to obtain additional information about the molecular transform

34 Each pattern is indexed Tom White (CFEL) Rick Kirian (ASU) b* c* a*

35 We can sum patterns to create a virtual powder pattern Lysozyme nanocrystals Ilme Schlichting

$We have merged indexed patterns into a 3D diffraction pattern CrystFEL$

36 We have merged indexed patterns into a 3D diffraction pattern CrystFEL software now available: Tom White (CFEL)

We have a new DESY system for processing and storage LCLS Data Dec 2009:

5 weeks 60 Hz >120 TB data Jan-Feb 2011: 2 weeks 120 Hz > 250 TB data SGI

Data Direct Networks SFA10000 60-bay HDD / 4U unit ~1 PB/rack (formatted)

37 We have a new DESY system for processing and storage LCLS Data Dec 2009: 1 week 30 Hz >20 TB data >6,000,000 patterns May-June 2010: 4 experiments 5 weeks 60 Hz >120 TB data Jan-Feb 2011: 2 weeks 120 Hz > 250 TB data SGI Altix 72 physical cores 360GB RAM Shared memory Direct connected storage Data Direct Networks SFA bay HDD / 4U unit ~1 PB/rack (formatted) (600 x 2 TB HDDs) Anton Barty, Tom White Can process 30 patterns / second

38 We need 10,000 to 100,000 indexed patterns R split = 1 p 2 P Ieven I odd P (Ieven + I odd )/2 R_split (%) Number of indexed patterns (x10 3 )

39 Molecular replacement reconstructs the 8.5Å structure 3000 MGy / pulse Nature (2011) Axel Brunger (Stanford) using DEN

40 Needle-shaped crystals can be grown in vivo by infection of cells by a modified baculovirus! 15 µm 0.25 µm Rudolph Koopman, Karolina Cupelli, Michael Duszenko U Tübingen Lars Redecke, Dirk Rhedes, U Lübeck & Hamburg Christian Betzel U Hamburg

41 Bragg peaks are observed even with 300 fs pulses 0 femtoseconds 20 femtoseconds 70 femtoseconds Resolution at corner = 8.6Å with 300 fs pulse Stefan Hau-Riege, LLNL

42 A crystal only gives Bragg diffraction when it is a crystal! '0'fs 25'fs 50'fs Accumulated# Bragg#intensity Low%resolu,on%Bragg%peak High%resolu,on%Bragg%peak Time#during#pulse#(fs)

$A crystal only gives Bragg diffraction$ '0'fs 25'fs 50'fs 0.30 0.25 0.20 0.15 0.

43 A crystal only gives Bragg diffraction when it is a crystal! '0'fs 25'fs 50'fs Accumulated# Bragg#intensity Low%resolu,on%Bragg%peak High%resolu,on%Bragg%peak Diffuse%sca7ering Time#during#pulse#(fs)

44 We see a degradation of the sample at longer pulse durations! 3! 2! 7!&8, $!!&8, $#!&8, 1!!&8, 1#!&8, 2!!&8, Diffraction counts 4.5+.,65) 1! $! 100 fs 150 fs 40 fs 300 fs!!"!!"# $"! $"# %&'()(*+,-./0 q = 2 sin Barty et al, Nature Physics 6, 35 (2012)

45 Only the first 30 fs contributes to the diffraction! Diffraction intensity relative to 40 fs 12344&5+36&7+(3)&'2389:0 $"# $"!!"# fs 100 fs 150 fs 300 fs ;!&<, $!!&<, $#!&<, =!!&<, =#!&<, >!!&<, Turn-off times 30 fs 40 fs 30 fs 80 fs 35 fs 150 fs 35 fs 200 fs 40 fs 250 fs 45 fs 300 fs!"!!"!!"# $"! $"# %&'()(*+,-./0 q = 2 sin

46 Time (fs) Population dynamics of Fe charge states during an XFEL pulse Ionized states have higher binding energies of 8 kev, photons/µm2, 10 fs FWHM Fe0+ 1s22s22p63s23p63d64s2 Fe10+ 1s22s22p63s23p4 Fe10+ 1s12s22p63s23p5 Fe20+ 1s22s22p2 Fe20+ 1s12s22p3 10 f f Photon energy (kev) 9 10 Dispersion corrections of atomic form factors of Fe and its ions S.-K. Son, H.N.C., R. Santra, PRL 107, (2011). > >

47 Calculations show that anomalous signals are enhanced by high X-ray intensity Effective scattering factors for Fe with 2 mj pulse Average ionization by end of pulse is +14 for highest fluence b(λ) f 0 +f c(λ) f photon energy (kev) photon energy (kev) Undamaged µm W/cm 2 45 µm W/cm 2 12 µm W/cm W/cm MGy/fs 5 MGy/fs 20 MGy/fs 200 MGy/fs S.-K. Son, H.N.C., R. Santra, PRL 107, (2011).

48 Calculations show that anomalous signals are enhanced by high X-ray intensity Harker section (difference Patterson map) at 2 Å resolution calculated from scattering factors at 6.5 kev and 7.4 kev at W/cm 2 Protein: Reaction center with Fe atoms Intensity difference at synchrotron: 9.95% Intensity difference at W/cm 2 : 15.9% Lorenzo Galli, CFEL

49 Anomalous dispersion breaks Friedel symmetry f(q) = r e (x) exp(iq x) dx x f( q) = r e Z = r e applez (x)exp( iq x) dx (x)exp(iq x) dx Randy Read

50 The crystal shape can be used to obtain additional information about the molecular transform

$We first demonstrated diffraction$ micron 1st shot at full power 51

51 We first demonstrated diffraction before destruction at FLASH 1 micron 1st shot at full power 51 Chapman et al. Nature Physics (2006) 1 micron SEM of structure etched into silicon nitride membrane

52 We perform ab initio image reconstruction with our Shrinkwrap algorithm q x (1/µm) θ x (deg) S. Marchesini et al. Phys Rev B (2003)

$measured FT -1 diffraction amplitude, Keep phase Impose$ $diffraction amplitudes Set of all objects that obey$

53 Phase retrieval can be accomplished with iterative transform algorithms Object function, g FT Impose measured FT -1 diffraction amplitude, Keep phase Impose support constraint; Impose positivity New object function, g Set of all objects that have measured diffraction amplitudes Set of all objects that obey support constraint Initial g 53

$resolution Coherent X-ray diffraction data λ =1.$ 6 nm, from a sample of 50- nm gold spheres arranged on a pyramid on a synchrotron

$Resolution = 10 nm Coherent X-ray diffraction data, rotating the sample -70 to +70$

54 We have reconstructed a 3D X-ray image of a noncrystalline object at 10 nm resolution Coherent X-ray diffraction data λ =1.6 nm, from a sample of 50- nm gold spheres arranged on a pyramid on a synchrotron Complete image reconstruction achieved, without any prior knowledge, using our shrinkwrap algorithm, parallelized for 3D on 32-CPU cluster. Resolution = 10 nm Coherent X-ray diffraction data, rotating the sample -70 to +70 degrees ( data points) 54 H. Chapman, et al., JOSA A (2006) 1 micron

55 Bragg sampling excludes information about the molecular transform M(q) M(q) L(q) m(x) m(x) l(x) 290 nm

56 Bragg sampling excludes information about the molecular transform M(q) M(q) L(q) m(x) m(x) l(x) 290 nm

$Diffraction pattern of a rectangular aperture f(x, y) = rect(x/a) rect(y/b) F (u, v)$

57 Diffraction pattern of a rectangular aperture f(x, y) = rect(x/a) rect(y/b) F (u, v) =ab sinc (au) sinc (bv) I(u, v) = F (u, v) 2 =(ab) 2 sinc 2 (au) sinc 2 (bv) b 1 a 1 b a

58 The truncated lattice F (u) =M(x) {L(u) S(u)} l(x) = S(u) f(x) =m(x) {l(x) s(x)} n= (x na) L(u) = N a sinc N au = N = Na h h = sin( Nau) sin( au) L(u) =1/a 1 a h (u sinc Nau (u sinc N(au h) h h a ) h a ) a Na (u h/a) 1 Na s(x) = rect ( x Na ) x S(u) =Nasinc Nau 1/a N wiggles including the two Bragg peaks

59 The finite crystal shape gives access to finer sampling of the molecular transform M(u) L(u) S(u) John Spence et al. Opt. Express (2011)

60 The finite crystal shape gives access to finer sampling of the molecular transform M(u) L(u) S(u) John Spence et al. Opt. Express (2011)

The finite crystal shape gives access to finer sampling of the molecular transform M(u) Method 1: Compute average shape transform by summing over all peaks, then divide this out to give M(u) at all

61 The finite crystal shape gives access to finer sampling of the molecular transform M(u) Method 1: Compute average shape transform by summing over all peaks, then divide this out to give M(u) at all samples Method 2: Determine asymmetry of each summed peak, and use this (plus intensities at peaks) to solve the 3D structure X L(u) S Ni (u) i = L(u) X i S Ni (u) There are no zeroes except where the molecular transform itself is zero John Spence et al. Opt. Express (2011)

$Atomic-resolution diffraction from single$ 28 nm 10 14 ph/µm 2 60 GGy 6000 MGy/fs 10 fs 3 Å

62 Atomic-resolution diffraction from single particles should be possible with ph/μm 2 28 nm ph/µm 2 60 GGy 6000 MGy/fs 10 fs 3 Å resolution RMS displacement: 0.5Å half electrons ionized

63 FLASH Experiments: LLNL: A. Barty, M. J. Bogan, M. Frank, S. P. Hau-Riege, S. Marchesini, B. W. Woods, S. Bajt, W. H. Benner, R. London, R. W. Lee, E. Spiller, A. Szoke U. Uppsala: J. Hajdu, S. Boutet, M. Bergh, C. Caleman, G. Huldt, M. M. Seibert, F. R. N. C. Maia, N. Timneanu, D. van der Spoel, M. Svenda, I. Andersson, J. Andreasson, D. Westphal, B. Iwan DESY: E. Plonjes, M. Kuhlmann, R. Treusch, S. Dusterer, T. Tschentscher, J. R. Schneider TU Berlin: T. Moller, C. Bostedt, M. Hoener LCLS Experiments: DESY: A. Barty, T. White, A. Aquila, J. Schulz, D. P. DePonte, A. Martin, K. Nass, F. Stellato, M. Liang, M. Barthelmess, C. Caleman, F. Wang, S. Bajt, L. Gumprecht, S. Stern, L. Galli, K. Beyerlein, G. Potdevin, H. Graafsma Arizona State University: J. C. H. Spence, P. Fromme, R. Fromme, M. S. Hunter, R. A. Kirian, U. Weierstall, R. B. Doak, K. E. Schmidt, X. Wang, I. Grotjohann U. Uppsala: F. R. N. C. Maia, J. Hajdu, N. Timneanu, M. M. Seibert, J. Andreasson, A. Rocker, B. Iwan, D. Westphal, O. Jonsson, M. Svenda, I. Andersson Max Planck Society: I. Schlichting, L. Lomb, R. L. Shoeman, S. Epp, R. Hartmann, D. Rolles, A. Rudenko, L. Foucar, N. Kimmel, G. Weidenspointner, P. Holl, B. Rudek, B. Erk, C. Schmidt, A. Homke, C. Reich, D. Pietschner, L. Struder, G. Hauser, H. Gorke, J. Ullrich, S. Herrmann, G. Schaller, F. Schopper, H. Soltau, K.-U. Kuhnel, R. Andritschke, C. Schroter, F. Krasniqi, M. Bott, T. R. M. Barends, H. Hirsemann SLAC: S. Boutet, M. Bogan, J. Krzywinski, C. Bostedt, M. Messerschmidt, J. Bozek, C. Hampton, R. Sierra, D. Starodub, G. J. Williams LLNL; S. Hau-Riege, M. Frank LBNL: J. M. Holton, S. Marchesini European XFEL: N. Coppola, J. Schulz, A. Aquila Gotheburg: R. Neutze TU Berlin: S. Schorb, D. Rupp, M. Adolph, T. Gorkhover U. Hamburg C. Betzel, L. Redecke U. Tübingen: M. Duszenko, R.Koopman, K. Cupelli

64 Quiz questions What#are#the#factors#that#dictate#the#smallest#crystal#size#that#can#be#used#in# an#experiment? What#determines#how#many#diffracEon#paFerns#are#required? How#could#the# Monte#Carlo #method#be#improved?#(how#can#we#get# greater#accuracy#with#fewer#paferns?) How#can#we#esEmate# hit#rates #and#the#amount#of#protein#needed? What#experimental#improvements#could#be#made#to#reduce#protein# consumpeon? What#advantages#are#there#to#using#femtosecond#pulses#compared#with# conveneonal#measurements? How#could#EmeMresolved#measurements#be#made?

Structure determination of a photosynthetic reaction centre by serial femtosecond crystallography

Structure determination of a photosynthetic reaction centre by serial femtosecond crystallography Gergely Katona Department of Chemistry University of Gothenburg SFX collaboration Gothenburg R. Neutze,