MtSerB2 a structural testbed for PPI-directed discovery: structure-based virtual screening against the ACT domain interface of an essential Mycobacterium tuberculosis phosphoserine phosphatase

Promoter

Prof. Johan WOUTERS, UNamur, Department of chemistry, Unit of theoretical and structural physico-chemistry (UCPTS), Laboratory of structural biological chemistry (CBS) 

Jury

• Prof. Steve LANNERS (UNamur), Président
• Prof. Johan WOUTERS (UNamur), Secrétaire
• Prof. Pierre FRANCOTTE (ULiège)
• Dr Marie HAUFROID (UCB)
• Prof. Lionel POCHET (UNamur)

Summary

Currently, Mycobacterium tuberculosis ranks as the second most lethal infectious agent, causing 1.6 million deaths in 2021. The lengthy and costly treatment (6 months and 4,000 euros), along with the emergence of antibiotic-resistant strains, underscores the urgent need for new therapeutic molecules. This work focuses on phosphoserine phosphatase of Mycobacterium tuberculosis (MtSerB2), a protein essential for serine synthesis and crucial for the pathogen's survival. Additionally, this protein plays a significant role in host invasion (by interacting with NF-κB and the cell cytoskeleton), making MtSerB2 an ideal target for the development of new and potent drugs against Mycobacterium tuberculosis. A promising strategy explored in this thesis is protein disruption, aiming to create a novel molecule that disrupts the protein’s structure, leading to a loss of protein activity.

To this end, the structural bottleneck was addressed first. Following the screening of several hundred conditions, a crystal of MtSerB2 was obtained and its structure solved at 2.3 Å at the SOLEIL synchrotron. This structure, the first reported for this enzyme, confirms a domain-swapped homodimer mediated by the ACT domains, which are absent from the human enzyme. As the electron density remained discontinuous for the 206–226 loop and for several solvent-exposed side chains, a hybrid protocol was implemented: the experimental backbone served as the reference, while the AlphaFold 3 model, superimposable onto the crystal structure (RMSD 0.92 Å), restored the side chains required for screening.

Three sites were then interrogated successively by virtual screening of a ~2,500-compound library of approved drugs (DrugBank), with prioritized compounds evaluated in a malachite green phosphatase assay. The catalytic site yielded a series of Mg²⁺-chelating dithiocarbamates (IC₅₀ ≈ 450 µM) that showed no selectivity over hPSP (IC₅₀ ≈ 440 µM) and proved chemically unstable. The interface between the PSP domains yielded a more potent dihydropyridine series (IC₅₀ 73–118 µM), but selectivity remained limited (1.6- to 2.4-fold). Finally, the pathogen’s protein-specific ACT domain interface was screened using a two-stage funnel combining rigid docking (Glide) with an AlphaFold 3 co-folding conformational filter, automated by an in-house tool. Four of the ten prioritized candidates inhibit MtSerB2; dequalinium (IC₅₀ 37.7 µM) and indacaterol (94.0 µM) display modest but reproducible selectivity, 5- to 6-fold in favour of the bacterial enzyme. Both compounds also inhibit the growth of M. tuberculosis H37Rv (MIC 12.5–25 µM and 25–50 µM), demonstrating that they cross the mycobacterial envelope. They represent unoptimized starting points whose binding mode remains to be established experimentally, notably by co-crystallization.