Structure determination and refinement
The structure of M. tuberculosis Rv2623, which contains two USP domains in tandem, and whose first domain shares 25% sequence identity with the USP M. tuberculosis Rv1636, was solved using the molecular replacement method and a CHAINSAW-generated search model consisting of the Rv1636 dimer (PDB ID 1TQ8 chains A, B), using a 2.9 Å, C2221 dataset (a = 173, b = 241.5, c = 241.7). A starting polyalanine model (R/Rfree = .56/.57) of four dimers was subject to four refinement cycles, each consisting of multi-domain rigid-body refinement in Molrep, a single cycle of restrained MLF refinement in Refmac5 (to obtain input FOMs for DM), 20 cycles of phase extension in DM (as above), and manual rebuilding of the polyalanine backbone in Coot. As R-factors converged (R/Rfree = .40/.42), ∼80% of the side chains were positioned and the Rv2623 dimers were further rebuilt and refined (R/Rfree = .31/.33) in CNS using high NCS restraint weights (400 kcal/mol) with rigid-body, energy minimization, grouped isotropic B factor, and simulated annealing refinement protocols. ATP and Mg2+ were built within composite omit density (calculated in CNS) during the final rebuilding/refinement cycles conducted with relaxed NCS restraints in Arp-waters and Refmac5, yielding final R/Rfree = 24.5/26.5. SigmaA-weighted difference maps calculated with the refined model reveal weak, fragmented density for a pseudotranslated copy of the Rv2623 dimer whose corresponding NCS translational vector (uvw = .500, .012, .494) appears in the native patterson at 7.7% of the origin peak height. Data collection and refinement statistics are summarized in Protocol S1. The coordinates of M. tuberculosis Rv2623 have been submitted to the protein databank (PBDID 3CIS).