Macromolecular Powder Diffraction:

1
Macromolecular Powder Diffraction
Structure Solution via Molecular Replacement

• Jennifer Doebbler
• and
• Robert Von Dreele

2
Acknowledgements

• Use of APS sector 1-BM
• Thank you for coming
• Tenniel for the wood cuts
• Use of the Advanced Photon Source was supported
  by the U. S. Department of Energy, Office of
  Science, Office of Basic Energy Sciences, under
  Contract No. DE-AC02-06CH11357.
• Further information on the data can be obtained
  fromJ. of Appl. Crystallography (2007) 40,
  133-143By Robert Von Dreele

3
Synopsis

• Introduction to powder diffraction
• Single crystals vs.. powders
• Down the Rabbit Hole
• Or, powder refinement of proteins
• Finding the real false minima
• Tutorial on multi-pattern molecular replacement
• Looking to the future

But first- why powder diffraction of proteins?
4
Wherefore Powder Diffraction?

• Single crystal data provides
• Iobs for each observed diffraction spot
• Nice method for subtracting background
• Tried and true methods for structure
  determination
• Though not always easy!

5
What happens when the dormouse misbehaves?
Or you cant get a nice single/non-twinned/ low
mosaicity/ highly diffracting crystal!
6
Welcome to Wonderland.
Integrate wedge

• Powder Data
• Image plate (shown)
• 3D 2D
• Overlap problem
• All Freidel pairs exactly overlapped
• Identical d-spacings
• Higher 2?
• Larger structures proteins!

Single crystal data 2D location of points, 3D
when you add angle Precession. Powder data is
only 2D
7
Reduce impact of overlaps

• Multi-pattern strategy
• Lattice parameters sensitive to pH, salt,
  radiation
• Leads to offset Bragg peak positions
• Reduces impact of some overlaps
• Extract better intensities

8
If you are going to win at cards.

• Mimic single crystal diffraction techniques
• End goal - solve unknown structures
• Rietveld refinement
• Previous work
• SH3 domain of Ponsin
• Margiolaki, et al. J. Am. Chem. Soc., 129
  (38), 11865 -11871, 2007.
• 67 residues about 1/2 the size of lysozyme
• Our goals
• Find the wall for protein size and resolution
• Image plate data
• Better signal to noise (generally)
• Slightly better resolution (generally)
• Broader peaks- more overlaps

9
Appropriate Starting Model

• Step 1 find a suitable starting model
• Human lysozyme
• 1.5 Å structure pdb code 1LZ1
• Single crystal data
• P212121 spacegroup
• 130 residues
• Hen egg white lysozyme
• 1.9 Å data, 5 data sets
• P43212 spacegroup
• 129 residues

gap
10
Pawley Refinement
Powder data statistics Fitted
-Bknd Average Bank
Ndata Sum(wd2) wRp Rp wRp Rp
DWd Integral Hstgm 1 PXC 1 1847 2478.7
0.0394 0.0529 0.0481 0.0554 0.147 0.924
Hstgm 2 PXC 1 1847 2548.4 0.0414 0.0536
0.0506 0.0559 0.144 0.924 Hstgm 3 PXC 1
1847 1661.4 0.0336 0.0447 0.0415 0.0473
0.150 0.924 Hstgm 4 PXC 1 1847 2282.2
0.0427 0.0558 0.0545 0.0593 0.145 0.924
Hstgm 5 PXC 1 1847 5141.5 0.0495 0.0725
0.0607 0.0772 0.133 0.927 Powder totals
9235 14112. 0.0423 0.0574 0.0525 0.0608
0.141 No serial correlation in fit at 90
confidence for 2.072 lt DWd lt 1.928 Cycle 800
There were 9235 observations. Total
before-cycle CHI2 (offset/sig) 1.4112E04 (
4.1588E01) Reduced CHI2 1.633 for 591
variables
Reflection data statistics Histogram 1 Type
PXC Nobs 627 R(F2) 0.0190 Histogram 2
Type PXC Nobs 627 R(F2) 0.0207
Histogram 3 Type PXC Nobs 627 R(F2)
0.0173 Histogram 4 Type PXC Nobs 628
R(F2) 0.0172 Histogram 5 Type PXC Nobs
626 R(F2) 0.0236

• Final values for multi-pattern Pawley refinement
• ?21.5 for 0-7o range
• 55-5 Å

11
The Caucus Race

• Determine number of molecules
• Matthews coefficient calculation (ccp4i)
• 1 mol in assym unit (mw 14331.2 Da)
• 39.73 solvent
• Coefficient 2.04
• Extract intensities from data
• Use overlp function in GSAS to minimize overlap
  impact
• No model
• partitioning of intensities will be off
• Determine a reasonable search space
• Underdetermined, use whatever info we have
• Choice of origin can vary
• x -1/8 - 1/8, y 0 - 1/2, z 1/4 - 3/4

Backward, forward, outward, inwardBottom to the
topNever a beginningThere can never be a stop
12
Do You Play Croquet?

• Do the spacegroup conversion from P212121 to
  P43212
• PSSP by Dr. Peter Stephens
• randomly assign Euler angles and center of mass
  coordinates
• Use simulated annealing to minimize obs-calc
  integrated intensities (1000 steps)
• Lather, rinse, repeat (10x)
• Accounts for overlaps when calculating cost
  function
• 6 Å data, 370 reflections

13
Almost Twins

• C-? trace of oriented human lysozyme
• red
• Superposition HEWL structure (194L)
• black
• Previously solved
• Serves as control to verify molecular
  replacement solution

14
Summary/Future Goals

• Molecular Replacement of powders works
• Reproducible
• Finalize refinement
• Working on algorithm to produce better maps
• Current maps mostly reproduce the model
• Tackle larger proteins
• Good technique to study drug or inhibitor
  interactions

Thank You!
15
Profile Function Values
Gu0, Gv3.6, Gw1.4, Lx1, Ly0
16
Change in Lattice Size

• Change in a not significant

• C shows a increase over exposure time

17
The hard part is over

• Globally the structure is in the right spot
• Mol rep solution with modified model
• Use this as a basis for Rietveld refinement
• Extend refinement region to 2-12o 2?
• Low angle shifts- solvent model
• High angle - overlaps
• Rigorous method for structure optimization
• Rigid body and restrained refinement (ccp4i-
  Refmac)
• Only works up to 3 Å (12o)
• Overlaps hurt

18
Omit Map

• Manually refine backbone and mutate side chains
  to match map extent
• Conversion from human to HEWL
• Map generated using CNS based DCO-X total
  composite annealed omit map routine.
• SA shown to be best at highlighting incorrect
  regions
• Hodel 1992 Acta Cryst
• Learn to restrain yourself