So, you are a macromolecular
crystallographer and have crystals of an interesting protein.
You know you
can get a set of native diffraction data, but phasing is a
problem. You can't do molecular replacement because this is a
brand new protein, and you haven't been able to find any usable
isomorphous heavy atom derivatives. However, you can make a non-isomorphous
derivative, or you know you can get cloned protein incorporating
selenomethionine - looks like a case for MAD (multiple
wavelength anomalous diffraction). |
| |
In order for MAD phasing to be successful, the
signal from the anomalous scatterers must be sufficiently large
relative to the total scattering from anomalous plus
non-anomalous scatterers. The anomalous signal from each heavy
atom depends on what that atom is, and you must also consider
the available x-ray flux at the wavelength of the appropriate
absorption edge. A
database of absorption edges is
available, courtesy of Ethan Merritt.
|
| |
At the
CHESS F2 station, with the standard Si (111) monochromator
crystals, the range of possible energies is 7.9 - 14 keV (1.57 -
0.89 A). This allows use of the K edges of elements 28 - 36
(including
Cu, Se, Br, etc.) or the L-III edges of elements 67 - 84
(including Hg, Pt, Au, some lanthanides, etc.).
The total signal depends on the number of (well-ordered)
anomalous scatterers as well as on their
nature. As an example, consider a 100 residue protein containing
2 Se atoms. The total number of
electrons is approximately 5200. Of this, the two selenium atoms
contribute 68. The single-wavelength anomalous (Bijvoet) signal
depends on the imaginary component of the scattering factor,
f'', while the multiple-wavelength (dispersive) signal depends
on the difference in scattering factors at the two wavelengths,
f'(1)-f'(2). These factors vary with the environment of the Se
atoms, but typical values are at most 3.7 and 6.6 electrons,
respectively [1], assuming suitable selection of wavelengths.
The resulting ratio of anomalous signal to total structure
factor may then be calculated according to the method of
Hendrickson [2], and is at most 4.0% for the Bijvoet and 3.5%
for the dispersive case.
Table 1 gives the maximum number of residues per heavy atom for
a reasonable signal (i.e. 3%
Bijvoet differences and 2% dispersive differences) from some
representative anomalous scatterers,
using f' and f'' values determined in protein crystals. Your
crystals may have a larger f'', and you may
be able to get useful information from a smaller signal, if you
are very careful in collecting data, but
this table is a good place to start in deciding whether a MAD
experiment is likely to succeed.
|
| |
Table 1 |
Maximum
Amino Acid Residues per Heavy Atom for Good MAD Signal |
| Heavy
atom |
Absorption edge |
Maximum f" |
Maximum delta (f') |
Max.
no residues
per heavy atom |
| Cu |
1.381 ? (K) |
4.2 |
5.0 |
90 |
| Se |
0.980 ? (K) |
3.7 |
6.6 |
90 |
| Br |
0.920 ? (K) |
3.4 |
5.5 |
75 |
| Hg |
1.009 ? (L-III) |
10.2 |
13.3 |
640 |
| Ho |
1.536 ? (L-III) |
25.7 |
19.9 |
1430 |
| |
|
|
|
|
|
The last column of this table was
calculated using an inversion of the equation given by
Hendrickson in Reference 2 for calculating anomalous signal
ratios. The f" and delta (f') values are taken from References 3
(Cu), 1 (Se), 2 (Br), 4 (Hg), and 5 (Ho). |
| |
| |
Because the MAD
signal is so small, it is especially important to minimize
errors in intensity measurements. This means: use your best
crystals, freeze them, and be careful in selecting data
collection parameters. It is worth spending considerable time
optimizing crystallization and freezing conditions (before
coming to CHESS, of course). During data collection, select each
crystal carefully and optimize exposure time, oscillation range,
etc. for it - see information on
"Efficient collection of oscillation data"
for hints.
|
| |
Return to Contents |
| |
Having decided to attempt a MAD
experiment at CHESS, follow the
usual procedures to request
beam time. As with any experiment involving heavy atoms, you
must consider the total amount and
form of heavy atom compounds to be brought to CHESS. If heavy
atoms are only present in
prederivatized crystals, the experiment is non-hazardous - just
check the appropriate place on the
proposal form. If you are bringing unfrozen crystals in a soak
solution containing heavy atoms, you
will need to include a "Heavy Atom Compounds Declaration"
describing the type and amount of
heavy atom compounds that you will be bringing. The small amount
of heavy atoms present in the
crystals themselves plus a few milliliters of millimolar soak
solutions poses no hazard and no
administrative complications beyond supplying the Declaration.
If you expect to bring concentrated
stock solutions of toxic compounds (not recommended but
occasionally necessary), you will also
need to complete a "Hazardous Materials Declaration". See the
heavy atom guidelines for more
detail.
|
| |
You will need a sample containing
your heavy atom for taking reference energy scans (see below).
CHESS has reference samples for a number of commonly used heavy
atoms on hand; check with
the staff to see if they have (or can make) one for your
experiment. If necessary, make up a standard
and bring it along. A simple and satisfactory method is to
sandwich a pinch of powdered material
(e.g. mercuric acetate for a mercury sample) between two pieces
of x-ray transparent sticky tape
(Kapton tape is good), and mount the sandwich in a slide mount.
The area covered by the powder
should be at least a centimeter or so square, so that it is easy
to be sure the x-ray beam is passing
through it. If you are bringing a reference sample containing
toxic material, such as a mercury
compound, be sure to include it on the Heavy Atom Compounds and
Hazardous Materials
Declarations. If you think you might need to make up another
sample while at CHESS and are
bringing a bottle of material to do so, include that also. When
you are at CHESS, handle any
reference samples you bring with care - label them, try not to
spill them, know how to clean them up
safely if you do spill them, and don't bring any more than
necessary in the first place.
|
| |
Return to Contents |
| |
Energy scans are essential to
select the energies (wavelengths) at which to collect data and
to check
for drifting of the monochromator during the experiment. These
are taken by scanning the
monochromator through the appropriate energy range and
monitoring the x-ray fluorescence from the
sample. CHESS staff will set the station monochromator at the
absorption edge of the element you
are using; scans may then be easily performed in the vicinity of
this energy. Reference samples
containing high concentrations of commonly used elements are
provided (see above for how to
make your own if you are using something unusual). The reference
sample is mounted towards the
front of the hutch; the "refscan" command moves it into the
beam, scans the energy over a range of
about 40 eV (specified using the "madsetup" command), lists and
plots the resulting fluorescence
spectrum, moves the sample out of the beam, and resets the
energy to its initial value. Reference
scans are run periodically during the experiment to check, and
if necessary recalibrate, the
monochromator. In most cases, you will also want to take an
energy scan on your crystal; this can be
a little more tricky.
|
| |
Ideally, you take an energy scan
on the crystal you will be using for data collection. However,
if this
crystal is too small to give a usable fluorescence signal,
another (bigger) crystal, or multiple crystals,
may be used. In order to get the best signal-to-noise ratio, the
fluorescence detector should be
placed at 90 degrees to the beam and in the plane of the ring,
i.e. pointing along the spindle axis. As
normally configured, the cold stream nozzle comes in at an angle
and does not interfere with the
fluorescence detector, so you may do a crystal scan on a frozen
crystal. If you prefer to have the cold
stream coming in along the spindle axis, you will have to
temporarily remove the cooling nozzle and
use unfrozen crystal(s) for the crystal scan. Unfrozen crystals
are perfectly satisfactory for this
purpose, as the environments of the anomalously scattering atoms
change very little on freezing.
Here is a good place to use those junky crystals that are not
suitable for diffraction - pack a bunch of
them into a capillary and do a crystal scan with them. If
possible, tilt the capillary relative to the spindle axis, so
that the fluorescent x-rays don't have to go through too much
glass. Once you have a sample in the beam, the "xtal_scan"
command runs the scan and outputs the results. Panels a - c of
the figure below show scans across the Se edge. In panel a is a
reference scan taken on a foil. The scan in panel b was taken
using a medium-sized crystal with the detector poorly
positioned; the panel c scan used a large crystal sample with
the detector correctly placed. Panel d of the figure shows a
reference scan for Hg. It is typical of this element that the
L-III edge is rather broad, with no "white line".
|
| |
Experimenters have also found
empirically that there is very little variation in the edge from
one
sample to another, so this is a case where you may get away
without a crystal scan at all, if you are
sure there is mercury in the crystal and are pressed for time.
|
| |
|
 |
| |
Select your desired
wavelengths from the crystal scan. For a 4-energy MAD
experiment, use (1) a
point slightly below the edge, (2) the inflection point of the
edge, (3) the top of the "white line" peak,
and (4) a point about 200 eV above the edge. For a typical
3-energy experiment, omit energy (1). For
some elements, such as Hg, it may be better to omit energy (3)
instead. In case of severe time
limitations, useful data may be obtained using just two
energies, e.g. (2) and (3) or (1) and (4). An
alternative to a full MAD experiment is the single wavelength
anomalous diffraction (SAD) technique;
this method uses only the Bijvoet differences collected at a
single wavelength. For SAD, the preferred
energy is (3), at which f" is largest. Look at the f' and f''
curves for your element (see Ref. 4 for how to
calculate these curves from an x-ray absorption spectrum) to
decide what selection of energies will
give you the best signal in case you can't collect all the data
you would like to. It is generally
preferable to have complete data at fewer wavelengths rather
than partial data at more wavelengths.
|
| |
Note that a
reasonable anomalous signal can be obtained over a range of
energies above the
absorption edge, and this range can be quite large for some
elements. Hence SAD data may be
collected at A1 for Se-containing samples (station tuned 40 eV
above the Se edge) and at F1 for
Br-containing samples (station tuned 40 eV above the Br edge).
At F2, a SAD experiment may be
possible even if the station energy cannot be tuned as low as
the desired absorption edge. Consult
CHESS staff about this possibility if you are working with an
element such as I or Xe.
|
| |
Return to Contents
|
| |
Geometry |
Once you have
selected the wavelengths for data collection, get a good crystal
mounted and determine the appropriate oscillation range and
exposure time.
Typical values are 1 degree and 1 minute, but these depend
heavily on your
particular crystal. Decide whether you will be using "mirror,
"inverse beam", or no
special geometry. In "mirror" geometry, anomalous pairs appear
on each image,
and in the "inverse beam" case anomalous mates appear on pairs
of images
collected with spindle angles differing by 180 degrees. Mirror
geometry requires
the crystal to be oriented with a mirror plane perpendicular to
the spindle axis. At
CHESS, the only means for getting this orientation is the use of
the arcs on the
goniometer head, so that it may not be possible to get a mirror
where you would
like it. There can also be a problem in scaling images from a
precisely oriented
crystal together, particularly in the case of a monoclinic
crystal of low mosaicity.
Nonetheless, mirror geometry is clearly preferable in terms of
reducing the number
of exposures and insuring good scaling between anomalous pairs,
and should
generally be used when possible, especially when another data
set is available for
scaling purposes. For inverse beam collection, software settings
allow collection of
wedges of data separated by 180 degrees. It is also possible to
collect data in a
single sweep, from a randomly oriented crystal, as for a
monochromatic
experiment; this can be perfectly satisfactory for a robust
crystal of relatively high
symmetry.
|
| |
CCD detector |
MAD
users at CHESS usually use a CCD detector. Considerations in
your planning due to the detector include: |
| |
|
 |
You will need to
take a background image every time you change the exposure time.
This doesn't take long, but is important. |
| |
|
|
The readout time, including
overhead, is about 3 seconds for the ADSC Q-210
detector (at F2 and A1) and about 11 seconds for the ADSC Q-4
detector (at
F1); you must include this time when calculating how many
exposures you can
get per hour. |
| |
|
|
The detector area is limited. To
get high resolution data, it may be necessary
to offset the detector, so that the direct beam position is no
longer in the center
of the detector. For mirror geometry, a vertical offset, i.e.
perpendicular to the
spindle, will still allow recording of anomalous pairs on each
image; a
horizontal offset will eliminate some pairs. In the inverse beam
case,
corresponding reflections appear in opposite quadrants on the
paired images,
and any offset will eliminate some of them.
|
| |
Timing |
Data
collection proper consists of a series of energy changes,
oscillation exposures, and crystal rotations. A sequence for a
3-wavelength experiment using inverse beam geometry, for
example, might be: |
| |
|
|
go to wavelength 1,
take 5 degrees of data, rotate crystal -5 degrees, |
| |
|
|
go to wavelength 2,
take 5 degrees of data, rotate crystal -5 degrees, |
| |
|
|
go to wavelength 3,
take 5 degrees of data, rotate crystal (180 - 5 = 175) degrees, |
| |
|
|
go to wavelength 1,
take 5 degrees of data, rotate crystal -5 degrees, |
| |
|
|
go to wavelength 2,
take 5 degrees of data, rotate crystal -5 degrees, |
| |
|
|
go to wavelength 3,
take 5 degrees of data, rotate crystal -180 degrees.
|
| |
|
Repeat
until data set is complete, crystal dies, or fill ends.
|
| |
|
The major contributor
to data collection time is the actual exposure time (at least at
F2), which can vary from a few seconds to a few minutes,
depending on the crystal
quality. Use a series of snapshots to determine the optimal
exposure time. Knowing
the detector readout time, the time to change wavelengths (about
3 seconds), and
the time to rotate the spindle 180 degrees (about 2 seconds) ,
it is then straightforward to calculate how much data you can
collect per hour. Adding in the
time to refill CESR (about 10 minutes every 2 - 4 hours), will
you have time to get
your data? If not, consider reducing the number of wavelengths
or the resolution, or
plan to come back another time.
|
| |
|
Return to Contents
|
| |
You need to monitor the data
collection - sorry, it's not safe to set up an 8-hour data
collection and
take the day off. It is particularly important to keep an eye on
the images being produced to be sure
that the crystal is still diffracting. It is also highly
desirable to pause data collection whenever the ring
is being refilled. The data collection software will halt
collection if collimator counts fall too low, but
even if this does not happen the beam motion that occurs during
filling makes images collected at
that time of dubious quality. If collection is halted
automatically, you will need to restart it once beam
returns. Reference scans should be taken at least once per fill,
to check for any monochromator
drifting. Once some data have been collected, start processing
them to be sure that the quality is
sufficient to be worth continuing with the current crystal.
Processing as you go will also let you know
when enough data have been collected, and may even let you solve
a structure at the beamline.
Don't forget to keep up with data backup, so as not to be left
with several hours worth of file transfer to
do at the end of your run.
|
| |
Return to Contents |
| |
 |
A suitable molecular
system with enough expected signal for phasing. |
| |
|
Good freezable
crystals, and (preferably) a suitable sample for a crystal scan. |
| |
|
Beam time at CHESS. |
| |
|
Well-thought out
strategy for data collection, with contingency plans. |
| |
|
Careful experimental
technique, with due consideration for safety. |
| |
|
Good bookkeeping
skills. |
| |
|
A sense of humor to
get you through the inevitable equipment failures. |
| |
1 |
Yang, W.,
Hendrickson, W. A., Crouch, R. J., and Satow, Y. "Structure of
RNase H Phased at 2A Resolution by MAD Analysis of the
Selenomethionyl Protein", Science, 249, 1398-1405
(1990)
|
| |
2 |
Hendrickson, W. A.
"Determination of Macromolecular Structures from Anomalous
Diffraction of Synchrotron Radiation", Science 254,
51-58 (1991)
|
| |
3 |
Walter, R. L., Ealick, S. E.,
Friedman, A. M., Blake, R.C., II, Proctor, P., and Shoham, M.
"Multiple Wavelength Anomalous Diffraction (MAD) Crystal
Structure of Rusticyanin: a Highly Oxidizing Cupredoxin with
Extreme Acid Stability", J. Mol. Biol. 263, 730-51
(1996).
|
| |
4 |
Tesmer, J. J. G.,
Stemmler, T. L., Penner-Hahn, J. E., Davisson, V. J., and Smith,
J. L. "Preliminary X-ray Analysis of Escherichia coli GMP
Synthetase: Determination of Anomalous Scattering Factors for a
Cysteinyl Mercury Derivative", Proteins: Structure, Function,
Genetics 18, 394-403 (1994)
|
| |
5 |
Weis, W. I., Kahn, R., Drickamer,
K., and Hendrickson, W. A. "Structure of the Calcium-Dependent
Lectin Domain from a Rat Mannose-Binding Protein Determined by
MAD Phasing", Science 254, 1608-15 (1991).
|
| |
|
Return to Contents |
|
