PNNL

Abstract

Refinement of inhibitor libraries used to discover new tuberculosis (TB) drugs usually requires high resolution structures of the active site pockets of Mycobacterium tuberculosis (Mtb) proteins. We selected 179 potential Mtb drug targets for structure determination, but obtained structures for only 19, a modest gene-to-structure success rate (11%). Therefore, we added 1675 targets from nine closely related non-tuberculosis Mycobacterium (NTM) species based on sequence similarity to the Mtb target. By using this approach, we obtained structures from orthologues of another 51 Mtb targets, thereby increasing the structural coverage by 3.7-fold, to 39%. In order to determine whether inhibitors designed against these non-tuberculosis structures will likely be effective against TB, we carried out an enzyme pocket comparison of all 106 Mtb-NTM pairs of structures with known Mtb active sites in the Protein Data Bank. This analysis showed that most pairs with >50% overall sequence identity have very similar active site pocket shape and chemistry, suggesting that the NTM protein structures will be useful as surrogates in TB inhibitor design. This hypothesis can now be tested using the purified proteins and structures generated in this study, and we highlight the promising test case of cytidylate kinase. These results provide strong support for a strategy to increase the availability of high-quality structures for the design of molecular libraries used to screen for new drugs and thereby accelerate TB drug discovery.