Research Article: ARCIMBOLDO on coiled coils

Date Published: March 01, 2018

Publisher: International Union of Crystallography

Author(s): Iracema Caballero, Massimo Sammito, Claudia Millán, Andrey Lebedev, Nicolas Soler, Isabel Usón.


The ARCIMBOLDO method of phasing through the location of small fragments combined with density modification and autotracing is particularly suited to helical structures, but coiled coils remain challenging. Features designed for solving coiled coils at resolutions of up to 3 Å were tested on a pool of 150 structures.

Partial Text

The phase problem is central to crystallography, and in the case of macromolecular crystals it is often not trivial to solve (Hendrickson, 2013 ▸). Starting phases for the structure factors that are missing from the results of a diffraction experiment are initially approximated by experimental phasing through heavy-atom derivatives or anomalous scattering at particular wavelengths (Hendrickson, 1991 ▸) or using previous structural knowledge from a similar structure in the method of molecular replacement (Rossmann, 1972 ▸; Navaza, 1994 ▸; Read, 2001 ▸). In chemical crystallography, molecules with less than 200 atoms that diffract to atomic resolution are routinely solved ab initio from the native diffraction intensities alone by direct methods (Karle & Hauptman, 1956 ▸; Woolfson, 1987 ▸). Small proteins of up to 1000 atoms that diffract to atomic resolution can also be phased by direct methods using the Shake-and-Bake algorithm (Miller et al., 1993 ▸; Sheldrick et al., 2012 ▸). Restrictions on data quality and structure size can be relaxed by means of various techniques. These include sophisticated use of the Patterson function (Caliandro et al., 2008 ▸), the use of expected values of structure amplitudes outside the actual resolution limit of the experimental data (Caliandro et al., 2005 ▸; Usón et al., 2007 ▸) and high-resolution density-modification algorithms such as low-density elimination (Shiono & Woolfson, 1992 ▸; Refaat & Woolfson, 1993 ▸), the sphere of influence (Sheldrick, 2002 ▸) and VLD (Burla et al., 2012 ▸). Small but highly accurate substructures can provide starting phases leading to successful phasing through density modification, as has been shown with ACORN (Foadi, 2003 ▸). As little as 10% of the main-chain atoms may suffice to solve a structure at 2 Å resolution (Millán et al., 2015 ▸). Thus, the atomicity constraints that are essential to direct methods can be substituted by enforcing secondary- or tertiary-structure stereochemistry. A related proof of principle was established using α-helices (Glykos & Kokkinidis, 2003 ▸) or nucleotides (Robertson & Scott, 2008 ▸; Robertson et al., 2010 ▸) as fragments to seed phasing. ARCIMBOLDO (Rodríguez et al., 2009 ▸, 2012 ▸) solves structures by combining the search for small polyalanine-model fragments with Phaser (McCoy et al., 2007 ▸) with expansion to a fairly complete structure through density modification and autotracing with SHELXE (Thorn & Sheldrick, 2013 ▸). Depending on the complexity of the case, a single-multicore workstation may suffice or a grid of computers may be needed (Sammito et al., 2015 ▸). Extremely successful approaches based on more complete models of lower accuracy (Rigden et al., 2008 ▸) have been developed based on the improvement of models derived from remote homologues or de novo model generation using ROSETTA (Qian et al., 2007 ▸) or QUARK (Xu & Zhang, 2012 ▸) combined with molecular replacement with Phaser (Read & McCoy, 2016 ▸) or MOLREP (Vagin & Teplyakov, 1997 ▸; Vagin & Teplyakov, 2010 ▸). This design underlies methods such as MR-Rosetta (DiMaio et al., 2011 ▸), AMPLE (Bibby et al., 2012 ▸, 2013 ▸; Keegan et al., 2015 ▸) and other implementations (Shrestha et al., 2011 ▸; Shrestha & Zhang, 2015 ▸).

ARCIMBOLDO_LITE succeeds in solving 140 out of a pool of 150 test coiled-coil structures with sizes ranging from 15 to 635 residues and resolutions between 0.9 and 3.0 Å on a single workstation. The fragments placed are 1–12 straight polyalanine helices made up of 6–50 amino acids. Run times for ARCIMBOLDO_LITE jobs typically take a couple of hours to one day on a single machine with eight physical cores. The successfully solved cases cover the full range of resolution data in the set, from a highest resolution structure at 0.9 Å (PDB entry 1byz) to a lowest resolution structure at 3.0 Å (PDB entry 4qkv). In terms of length and complexity a wide range is covered as well, from a smallest structure with just a single chain in the asymmetric unit comprising 15 residues (PDB entry 1kyc) to a largest structure with four chains in the asymmetric unit totalling 618 residues (PDB entry 2efr).




Leave a Reply

Your email address will not be published.