Date Published: December 01, 2019
Publisher: International Union of Crystallography
Author(s): Madhumati Sevvana, Michael Ruf, Isabel Usón, George M. Sheldrick, Regine Herbst-Irmer.
Examples are presented of successful data-processing, phasing and refinement strategies for non-merohedral twins, covering the range from minerals to proteins.
Twins are defined as regular aggregates consisting of individual crystals of the same species joined together in some definite mutual orientation (Giacovazzo, 2002 ▸). Therefore, twins may be defined by a symmetry operator that transforms one orientation into another, the so-called twin law, and by the fractional contribution ki of each component. In reciprocal space, the twin law describes the symmetry operator that transforms the h1k1l1 indices of one domain into the indices h2k2l2 of a second domain.
The methods described above have successfully been used for twinned small molecules for many years, and SHELXD and SHELXE have also been used to assist in the SAD phasing of merohedrally twinned macromolecules (Dauter, 2003 ▸; Rudolph et al., 2003 ▸). To show that the procedures described in this paper are also valid for macromolecular structures, we grew non-merohedrally twinned crystals of two benchmark protein structures: cubic insulin and glucose isomerase (Sevvana, 2006 ▸; Fig. 11, right).
The same procedures may be used for the treatment of non-merohedral twins in minerals, organometallic structures and proteins when the data are processed using the programs CELL_NOW, SAINT and TWINABS. CELL_NOW and SAINT are also incorporated into the Bruker APEX3 system. The resulting HKLF 4- and HKLF 5-format files can be used for structure solution and refinement with the SHELX and several other program systems. The detwinned HKLF 4 data are more widely applicable, but refinement against the composite reflections without detwinning using the HKLF 5 format may be slightly more accurate. If all domains are of similar quality and all of them are well centred in the beam, refinement against the HKLF 5 data should lead to the best results because the multiplicity is the highest. Quite often data from one domain might be of superior quality to those from other domains. In this case, only reflections with a contribution from that domain should be used for model refinement. However, in order to use Rfree the HKLF 4 format may be required.