Mg or water?

From: bie gao
Date: 14 December 2011 22:45

Hi every,

I'm working with 2 crystal forms of a protein from 2 different crystallization conditions. Condition 1 has 100mM MgCl2. Condition 2 doesn't. Both are ~2.9 angstrom.  The 2 structures are virtually identical except in condition1, there is a clear positive density surrounded by a Glu side chain carboxyl and a couple of main carboxyl groups. (Again, condition 2 doesn't have this density).

My initial thought is that a Mg atom is incorporated and it fits well. But the problem is we can not role out the possibility of a water molecule. Refining with Mg gives a b-factor of 42 (about average for the whole protein). The b-factor is 21 when refining with a water. Both cases there is no positive/negative density at contour=2.0.

Based on the current data, is there any other role we can apply to see how likely it is a Mg or water. Or  anomalous scattering is the only way? Thanks for your suggestions.

Best,
Gao

----------
From: Phil Evans


I doubt that you can tell the difference between Mg and water just from the height of the density, but Mg2+ is always octahedrally coordinated with Mg-O bond lengths ~2.0A At 2.9A resolution you may not be able to distinguish

Phil

----------
From: Eleanor Dodson

At 2A I think you will see the difference between the 2A bond Mg-GLU and the likely 2.8A HOH-GLU.

Look at your difference map - put a peak there and check distances to surrounding atoms.

Marjorie Hardings metal protein server gives a guide to other putative Mg binding.

Eleanor

----------
From: bie gao


Thank you all for the help. These are the key factors I collected so far:
1. Distance, Mg--O is shorter (2.0 -- 2.4A)
2. Coordination, Mg is octahedrally coordinated.
3. B factors, local B factors (i.e. the residues that coordinate with the ion) should be similar.
4. Use Mn++ to replace Mg.
I will look into these more.

Best,
Gao

----------
From: John R Helliwell


Dear Bie Gao,
You can obtain an estimate of the standard deviation on your putative
Mg to ligand distance using the diffraction precision index (DPI)
approach of Cruickshank and Blow (1999 and 2002 Acta Cryst D). ie:-
D M Blow Acta Cryst. (2002). D58, 792-797
Synopsis: The formulae for the diffraction-component precision index
introduced by Cruickshank [(1999), Acta Cryst. D55, 583-601] are
simplifed using two approximations. A rearranged formula for the
precision index is presented which can readily be calculated from
experimental data.


Using this you can add a quantitative test of statistical significance
to the, of course, very senible earlier inputs to your questions. Do
note though that the DPI is calculated as a number for the position
error, sigma (x),  on an atom with an average B factor. You have to
adjust your DPI value for the atoms in question via the square root of
their B value ratioed to the average B. For example a lower than
average B gives a more precise sigma (x) than the average. Finally the
sigma on the bond distance itself is calculated from the sigma (x)
pair of values for each atom via the quadrature formula. The usual
statistical test of significance is if the ligand distance (L) is then
>3sigma (L).

Best wishes,
Prof John R Helliwell DSc.

--
Professor John R Helliwell DSc