Dear Napoleão,
I will try to summarize our experience and give some suggestions.
Few reasons why MR with coiled coils can be very tricky, such as
their asymmetric shape and their ability to overlap onto themselves
upon a shift and rotation (for a heptad-based coiled coil, this would be a
shift by 7 residues) have been already mentioned.
And I can add two more.
First, we very often see that coiled coils get bent due to
crystal contacts. This means that the conformation may be
quite different compared to your search model.
Second, many coiled coil sequences were seen to
assemble into structures different to what
they were supposed to be. Examples I know include
a trimer when a dimer was expected, an antiparallel
coiled coil when a parallel one was expected,
a monomer when a a dimer was expected,
chains offset with respect to each other, etc etc
We were able to phase several short (40-60 residues)
coiled coils by MR in the past. The paper
Strelkov SV et al (2002) EMBO Journal describes
two such cases (please look at the Methods section
which provides a fairly detailed explanation of what
we did). Both were done with Molrep.
I do have to say that we also failed on MR
miserably in many other cases, whatever we tried.
One recent example of a coiled coil fragment
that assembled to something entirely different
than we expected it to, is in Nicolet S et al (2010)
J. Struct. Biol. There, we really had to use heavy atoms.
I collected
various data sets (home source, Brookhaven and Diamond), including some
at the resolution of 1.65 A, for which the space group appears to be
C222 or C2221.
You have collected many data sets and you are still not sure
about the space group - ? Have you looked at the systematic absences
carefully? If you are still missing the (00l) axis then you
probably could try collecting it again while considering the
orientation of your crystal.
Knowing the correct space group is a valuable
information. Yes there are cases when you can not
distinguish between two space groups from the diffraction pattern
(e.g. P61 and P65) and then you /have/ to run your MR search
twice in both groups. Yes modern programs will do
it almost by default for you. But if you do know your correct space
group (for instance C2221 and not C222) then when you
do the MR search in the two space groups you /may/ (with some
luck) see better results with the first than with the second.
This /may/ be a hint that your C2221 solution is correct.
I have tried A LOT of Molecular Replacement using Phaser and Phenix
AutoMR.
As mentioned already, is really advisable to try other MR programs
(MolRep, epmr etc),
as in fact they are all based on different algorithms.
I'm using a 80% identity coiled coil helix as search model.
Regarding the search model, I already tried trimming some or all
side chains and removing 2, 3 or 5 residues on each/both sides.
From your description I am guessing that you use a monomeric helix
currently - ?
Indeed you really should try different models, anything you can
get your hands on. In fact this is a number one factor to
get MR to work.
On one hand, you should indeed try shortening your
model (even systematically trimming one residue at a time).
Do not hesitate to use much shorter models (less than
half of your full helix). You can trim the wrong side chains
but I would not advise chopping off all of them.
With a short helix, once you have the first candidate
solution, try searching for a second copy with the packing penalty
switched off. If you get another helix overlapping with your first
solution (with some offset) this may be a sign of success.
(see the EMBO J paper)
On the other hand, our experience shows that in many cases
you can only get a solution if you use a full coiled coil
(dimer, trimer,...) and not a monomeric helix.
If your real structure is a non-crystallographic dimer etc,
then you should definitely search with a dimer etc.
If your oligomer is due to a crystallographic axis, then
you may try this as well, after switching off the
packing penalty. But beware that the oligomer
will almost certainly land on the axis, even for a wrong solution.
Hope this will help, and best of luck with your MR
searches -- which are fun, since they still require
some thinking!
Sergei
--
Prof. Sergei V. Strelkov
Laboratory for Biocrystallography
Department of Pharmaceutical Sciences, Katholieke Universiteit Leuven
O&N2, Campus Gasthuisberg, Herestraat 49 bus 822, 3000 Leuven, Belgium
Work phone: +32 16 330845 Fax: +32 16 323469 OR +32 16 323460
Mobile: +32 486 294132
Lab pages: http://pharm.kuleuven.be/anafar
