Dear Crystallographers,

I have solved the structure of a complex, which is the average of its component proteins. The problem is as follows:

The complex is composed of two proteins, which share 67%/84% sequence identity/similarity in the resolvable region. The proteins are stoichiometric (SDS-PAGE of the crystals), and form a dodecamer composed of two hexameric rings, stacked on top of one another.

The crystals are H3. Analysis with Xtriage, Pointless, Aimless, Scala do not find other space groups likely, and do not find any apparent data pathologies. Data processing in H3 (to 1.9 A) is straight forward (Rfactor/Rfree: 22.3%/26.4% using one of the proteins as MR model). Attempts to solve the structure in other space groups (except P1) fail, or produce meaninglessly poor statistics.

I have good anomalous data from isomorphous SeMet crystals, and the anomalous difference density peaks for the Se are approximately additive: anomalous peaks for positions where SeMet is unique to one of the two proteins are about 1/2 the size of peaks common to both proteins. This composite effect is apparent when the structure is processed in P1 also.

Crystal contacts are not specific to either of the two proteins, and so it appears that the crystal doesn't distinguish between up- and down-orientations of the dodecameric rings. Therefore, each subunit in the resulting structure is a composite of the two proteins.

Despite the occurrence of heteromeric rings in protein structures, I have not been able to find a description of this problem in the literature, nor the archives of the CCP4BB.

I am keen to hear from anyone who has observed similar things in structures of their complexes, and how best to approach this problem.

Many thanks,

Graham Robinson

Postdoc, University of Geneva

Graham Robinson, Ph.D.

Fitzpatrick Group

Département de Botanique et Biologie Végétale

Université de Genève

30 Quai Ernest-Ansermet

CH-1211, Genève 4

T: +41 (0) 22/379.30.12