Molecular replacement with pseudosymmetry and model dissimilarity: a case study.

Crystals of human T-cell leukemia virus protease (HTLV-1 PR) have been very difficult to prepare and only native data extending to 2.6 angstroms resolution could be collected. Initial attempts to solve the structure with a variety of low-sequence-identity models utilizing proteases from other retroviruses and using a number of molecular-replacement programs were unsuccessful. The structure was finally solved using Phaser, revealing extensive pseudosymmetry and significant deviations from the starting models, features that were likely to be responsible for the initial failures. The steps taken to solve this structure and some of its intriguing crystallographic aspects are discussed.