Yes, this sort of thing happens a lot more often than one might think,
but people who have crystals with such "high-copy asymmetric units" tend
to not solve them. Hence, they don't end up in the PDB. In cases where
the structure is eventually solved, it is usually done by finding an
alternative crystal form. At least, that is what I usually see when
people bring these things to the beamline. Yes, the high NCS sounds
like it would be really cool, but the reality of high-copy ASUs is that
they are particularly prone to radiation damage problems. Not because
of any special chemistry, but because the big ASU means that a complete
data set from a single crystal requires collecting a lot of photons (see
Holton & Frankel, 2010, and http://bl831.als.lbl.gov/xtalsize.html), and
that high-copy ASUs tend to also have non-isomorphism "issues" (making
it difficult to merge data from many crystals). Yes, the NCS advantages
"cancel" this hindrance, but only at a much later stage (after you have
found all the sites). That, and I think there is some psychological
barrier to building and refining 24 copies of the same thing!
Why do high-copy ASUs happen? Sometimes they occur naturally, like
certain invertebrate hemocyanins where one molecule contain dozens to
hundreds of copies of a single domain. Bart Hazes can tell you all
about these!
However, it is also not that uncommon for one or more crystal symmetry
operators to "collapse" upon cryo-cooling (or other forms of crystal
abuse). I have seen this a lot! There are many examples of
nearly-crystallographic NCS in the PDB, many of which I suspect are
cryo-cooling artifacts. Doesn't change the structure all that much, but
should you choose to "go with it" you do have to be VERY careful with
NCS like this! It is very easy to invalidate the Rfree. An extreme
example is taking a crystal that is actually P2, but instead processing
it as P1, picking a random "free" set, and refining with a twofold NCS
operator. You will find that Rfree will drop like a rock and become
essentially equal to Rcryst. This will be the case even if the
structure you are refining is totally wrong! This is because every
"free" reflection actually has an "NCS symmetry mate" in the working
set. There are an alarming number of cases like this in the PDB, but I
will not name names here.
What is the "result" that makes you think your crystal is not tetragonal?
-James Holton
MAD Scientist
On 9/30/2010 4:54 AM, Mario Milani wrote:
Dear all,
i have a 30 kDa protein that crystallize so far in three different conditions
but with the same space group. It initially looks like tetragonal (I4, a=141,
b=141, c=208) and then results triclinic (P1, a=141, b=141 c=144, alpha=119,
beta=119, gamma=90), hosting about 24 mol. in the unit cell. Other data: self
rotation shows the presence of 4 peaks with chi=180; molecular replacement
shows the presence of a pseudo-translation peak; DLS made at protein
concentration close to crystal growth conditions shows a Rh compatible with
something like a tetramer with low polydispersity (about 15%). Do you have any
experience with similar ‘asymmetric’ associations? Do you have any suggestions,
beside the addition of ligands to the crystal growth conditions, in order to
get a ‘simpler’ crystallographic assembly? I have some models (with sequence
identity less than 25%) in order to try MR but all trials so far did not solve
the structure (using balbes, molrep, phaser and epmr). Any suggestion is
welcome.
Thank you,
Mario Milani
