1. Determining protein structures
Andrew Torda, wintersemester 2006 / 2007
X-ray
numerically most important
NMR
more detail
What is our goal ?
a set of x, y, z coordinates
short detour to coordinate files …
Lecture plan
X-ray first, then NMR

2. Coordinate files and the PDB
PDB = protein data bank
only good repository of protein structures
usually required for publications
format
from old fortran based programs (columns / punch cards)

3. Coordinate information
general headers and information
HEADER PROTEINASE INHIBITOR (TRYPSIN) FEB BPI BPI 2 COMPND BOVINE PANCREATIC TRYPSIN INHIBITOR (BPTI) (CRYSTAL FORM II) 1BPI 3 EXPDTA X-RAY DIFFRACTION BPI 5
….. ATOM N ARG BPI 137 ATOM CA ARG BPI 138 ATOM C ARG BPI 139 ATOM O ARG BPI 140 ATOM CB ARG BPI 141 ATOM CG ARG BPI 142 ATOM CD ARG BPI 143 ATOM NE ARG BPI 144 ATOM CZ ARG BPI 145 ATOM NH1 ARG BPI 146 ATOM NH2 ARG BPI 147 ATOM N PRO BPI 148 ATOM CA PRO BPI 149 ATOM C PRO BPI 150 ATOM O PRO BPI 151 ATOM CB PRO BPI 152 ATOM CG PRO BPI 153 ATOM CD PRO BPI 154 ATOM N ASP BPI 155 …
in case you drop them on the floor
temperature factor
x, y, z coordinates

4. X-ray sociology / geography
History
1896 X-rays from Wilhelm von Röntgen
1913 Bragg first small molecule
1950's or early 60's first proteins (Mb)
X-ray sociology / geography
biggest, meanest X-ray source ?
DESY (down the street)

5. Proteins and crystals Proteins can form crystals
like table salt or sugar
just much more difficult
a, b, c define the unit cell
may not be perpendicular
may have more than 1 molecule a b

6. Proteins and X-rays
light wavelength 4 – 700 x 10-9 m (about 4000 bonds !)
x-rays have wavelengths near 1 Å (10-10 m)
no such thing as X-ray lens
they will diffract
cute explanation
x-ray frequency about 2 x 1018s-1
electrons move at about 2 x 106 ms-1
effectively standing still

7. proteins as a diffraction grid
remember high school diffraction
depends on wavelength
x-rays bounce off electron clouds
will eventually give information about electron density (ρ(x, y, z))
like light in a diffraction grid
intuitively
shine light on grid and try to work out separation

8. Diffraction extra path length ABC if it is a full wavelength2θ A B C
Diffraction
extra path length ABC
if it is a full wavelength
x-rays come out in phase
we see a spot
formalise this in Bragg’s law
we have lots of d’s
the bigger the d’s, the closer spaced the diffraction spots
we know λ, but have the information from all the d’s at once
can this be separated ?

9. Collecting data rotate sample
detector
x-ray source
rotate sample
fuzzy looking spots, indexed by position and angle
diagram from www-structmed.cimr.cam.ac.uk/Course/Overview/Overview.html

10. electron density expression
spots are intensity
lots of electrons scatter more
intensity is square of structure factors F(hkl)
example in one dimension
where ρx is the density at our one dimensional coordinate x
α is a phase, h frequency
|F| is a structure factor, absolute value
a comes from unit cell
everything except α defined
can be done in three dimensions

11. where does this come from ?
the structure factors will be periodic
property of spacing within the crystal and between atoms
should be the same on different days, copies of crystal FT
real
real FT
real FT
colours represent phase
every dot has its own phase
from Kevin Cowtan's

12. Phasing direct methods not practical for many points other methods
replace atoms (MIR)
guesses based on model (MR)

13. Multiple isomorphous replacement (MIR)
if we only have a few points, phases are easy
proteins have many points
make them act like a few points
bind some heavy atoms (lots of electrons)
they will then dominate scattering
modern method..
engineer in sites for selenomethionine
phase directly
heavy atoms

14. Molecular replacement (MR)
make a model and use it to get rough phases
I have unknown protein "A"
similar to protein "B" whose structure I do know
phases from "B" should be similar to those from "A"
idea
from "B"
calculate density
back transform
get phases
apply to data from "A"
in pictures…

15. Molecular replacement (MR)
Background - if I have a protein I can calculate the spots
protein coordinates
Put protein on a grid
calculate ρx at each grid point, use
but apply backwards to get Fh (back transform)
with phases
if I know the structure, I can calculate the expected α 's
density on a grid

16. Molecular replacement (MR)
requires a starting model for structure
much luck – often works if model is 30 to 40 % identical to correct answer
very important procedure
many proteins of interest are similar to known ones
requires no chemistry (unlike MIR)

17. Overall procedure Making crystals
make crystals
collect data
phase
fit to initial map
refine
Making crystals
do you normally see protein crystals ?
concentrated protein + salts + robotic trials
lots of trial and error

18. Data collection can be done in the lab
more powerful X-rays from a synchrotron
may damage crystal
often done in the cold
takes 10 minutes (synchrotron) to few days

19. Fitting to a map the Fourier transform gives you electron density
not nuclei
not protons (H atoms)
may look like a protein
atoms have to be placed within skeleton

20. Fitting atoms errors can be made backwards wrong sequence
cannot tell O from N

21. Refinement removing noise fixing phases adjusting coordinates
from a model, calculate structure factors
move atoms to as to match predicted vs measured

22. Quality Quality of data some proteins diffract better than others2θ A B C
Quality of data
some proteins diffract better than others
some crystals are better than others
completeness
Resolution
physical meaning
most scattered x-rays from smallest "d"
best resolution, smallest θ

23. Disorder / mobility what if proteins do not pack perfectly ?
data will be smeared / blurry
what if there are differences between molecules ?
proteins are not perfectly static
what is real resolution
typical 1.5 to 3.0 Å
best < 0.8 Å (small friendly proteins)
worst > 5 Å (large membrane bound)

24. Local disorder
protein may crystallise even when parts are not well ordered
typical of loops
and termini
can we quantify this ?
with a model and approximations

25. Atoms - how well determined ?
model of Gaussian density
believable ?
does reflect refinement
contribution from overall disorder
atoms not really Gaussians
certainly reflects relative mobility
can we see this in coordinates ?
individual B-factors
where U is mean square displacement (Å)
σ is width
μ is centre

26. Coordinate information
HEADER PROTEINASE INHIBITOR (TRYPSIN) FEB BPI BPI 2 COMPND BOVINE PANCREATIC TRYPSIN INHIBITOR (BPTI) (CRYSTAL FORM II) 1BPI 3 EXPDTA X-RAY DIFFRACTION BPI 5
….. ATOM N ARG BPI 137 ATOM CA ARG BPI 138 ATOM C ARG BPI 139 ATOM O ARG BPI 140 ATOM CB ARG BPI 141 ATOM CG ARG BPI 142 ATOM CD ARG BPI 143 ATOM NE ARG BPI 144 ATOM CZ ARG BPI 145 ATOM NH1 ARG BPI 146 ATOM NH2 ARG BPI 147 ATOM N PRO BPI 148 ATOM CA PRO BPI 149 ATOM C PRO BPI 150 ATOM O PRO BPI 151 ATOM CB PRO BPI 152 ATOM CG PRO BPI 153 ATOM CD PRO BPI 154 ATOM N ASP BPI 155 …
temperature factor
x, y, z coordinates

27. Lastly, what can we do Time to solve a structure days to years
crystallisation, phasing
biggest structures
macromolecular complexes, ribosome, photosynthetic centre, …
most difficult and important
membrane bound proteins
account for many drug targets
more applications
solve structures with ligands / complexes
inhibitors
DNA