The crystal structure of M. tuberculosis Rv2623: Dimer assembly and ATP-binding capacity 
To examine the biochemical mechanisms responsible for Rv2623 function, we determined the crystal structure of wild-type Rv2623 at a resolution of 2.9 A.
The structure reveals a compact, 2-fold symmetric dimer.
Each monomer is composed of tandem USP domains [residues 6-154 (domain 1), 155-294 (domain2)] that share 26% sequence identity and significant structural homology (residues 6-154 and 155-294 comprise domains 1 and 2, respectively; interdomain rms=2.04 A for 140 equivalent Calpha's).
Individual domains, which consist of a twisted, five-stranded, parallel beta sheet flanked by four alpha helices, unite through an antiparallel, cross-strand (beta5-beta10) interaction that produces a central dyad axis between beta5/beta10 and a continuous, ten-stranded, mixed beta sheet in the complete monomer.
Each domain possesses a pair of conserved betaalphabeta motifs (domain 1: beta1-L1-alpha1- beta2, beta4-L2-alpha4-beta5; domain 2: beta6-L3-alpha5-beta7, beta9-L4-alpha8-beta10) that encompass four loops (designated L1-L4) responsible for ATP recognition (Figure 6A and C).
A "U-shaped" ATP molecule that lies within a cleft near the monomer surface is stabilized by 1) a cluster of hydrophobic residues (I14, V41, H42, V116/132/261/277/281, L136, A175) that forge the adenine/ribose-binding scaffold, 2) a pair of conserved L1/L3 aspartates (D15-L1/D167-L3), and 3) small phosphoryl/ribosyl-binding residues within the G-2X-G-9X-G (S/T) motifs that comprise L2/L4 (G120/265/267/268 and S131/276) (Figures 6A,C and 7A).
Dimerization of Rv2623 occurs along a 2-fold axis orthogonal to the intramonomer dyad and juxtaposes ATP binding pockets from opposing monomers (Figure 6B).
Phylogenomic analysis places Rv2623 in a Uniprot/TrEMBL family (Q5YVE7) of 370 tandem-domain USPs, and a 113-member subfamily (N631) that consists almost exclusively of actinobacterial representatives (Text S1).
Structure-based sequence alignments of both Rv2623 domains with the N631 consensus suggest that domain 2, which exhibits significantly higher conservation than domain 1 across global and ATP-binding subfamily consensus sequences, represents the ancestral domain among ATP-binding USPs with tandem-type architectures.
Interestingly, the domain fold and interdomain organization observed for Rv2623 is broadly conserved: these features are shared among single domain USP structures, both monomeric and dimeric, that are presently represented within the PDB.
As this manuscript was under preparation, a second, lower resolution (3.2 A) crystal form of Rv2623 (PDB ID 2JAX) was released for public access.
This structure is nearly identical to the present model as demonstrated by superposition over the ATP ligands and the monomeric and dimeric forms (rmsds are 0.57 and 0.81 for 258 and 517 matched CA's, respectively).
The differences localize primarily to flexible loop regions (residues 44-58, 150-159) that, while disordered in 2JAX, are partially stabilized in the present structure by local crystal contacts.
To gain insight into the ATP-binding mode(s) exhibited by Rv2623, the structural features of the ATP-binding pocket of domains 1/2 were compared to the monomer fold of the representative ATP-binding USP, MJ0577 (PDBID 1MJH) [26].
Overlay of these structures reveals very considerable similarity for the residues that form the binding pockets and the associated ATP molecules, for which the triphosphoryl moieties assume virtually indistinguishable conformations.
Relatively subtle structural and phylogenetic differences that exist between the ATP-binding pockets might nevertheless confer divergent binding and/or regulatory properties to the tandem domains.
