Fragment Merging, Growing, and Linking Identify New Trypanothione Reductase Inhibitors for Leishmaniasis

Trypanothione reductase (TR) is a suitable target for drug discovery approaches against leishmaniasis, although the identification of potent inhibitors is still challenging. Herein, we harnessed a fragment-based drug discovery (FBDD) strategy to develop new TR inhibitors. Previous crystallographic screening identified fragments 1–3, which provided ideal starting points for a medicinal chemistry campaign. In silico investigations revealed critical hotspots in the TR binding site, guiding our structure- and ligand-based structure-actvity relationship (SAR) exploration that yielded fragment-derived compounds 4–14. A trend of improvement in Leishmania infantum TR inhibition was detected along the optimization and confirmed by the crystal structures of 9, 10, and 14 in complex with Trypanosoma brucei TR. Compound 10 showed the best TR inhibitory profile (Ki = 0.2 μM), whereas 9 was the best one in terms of in vitro and ex vivo activity. Although further fine-tuning is needed to improve selectivity, we demonstrated the potentiality of FBDD on a classic but difficult target for leishmaniasis.


Docking calculations of fragments 1-3
Docking studies of fragment-derived 6, 9, 10 and 14   Docking calculations of fragments 1-3.In Figure S1, we report the best results in terms of docking score for each fragment together with the respective crystallographic binding mode as a reference.
All the docking calculations share the same size and location of the box, as shown in Figure S2.
Interestingly, 3 was the only fragment matching the crystallographic binding mode, in terms of the interactions with the Z-Site (π-π stacking with P369, and the water-mediated hydrogen bond between the amidic group and T463, E467, and M400).In particular, the predicted docking pose adopted a kinked conformation, where the p-fluoro-phenyl ring was accommodated below the catalytic tetrad, and it was surrounded by lipophilic residues (L399, M400, P462).
Conversely, the best pose predicted for fragment 1 diverged significantly with respect to its crystallographic reference.In particular, the salt bridge between the positively charged piperazine ring and E467 was replaced by an equivalent interaction with nearby E466, and the entire molecule appeared to be shifted toward the solvent-exposed portion of the binding site.Notably, the departure from the Z-site is justified by the establishment of additional hydrophobic interactions of the propylbenzene tail with a solvent-exposed cluster of lipophilic residues (V362, P336, I339, and I458, among the others).
A remarkable difference between the predicted and experimental binding modes was also observed for fragment 2. Here, the best docking pose turned out to be highly influenced by the direct hydrogen bond established between the ureidic moiety and L399 and the water-mediated hydrogen bond of the same group with E467 and T463.Furthermore, the positively charged nitrogen of the piperazine ring interacted with E467 and F369 through a salt bridge and cation-π interaction, respectively.
Intriguingly, the orientation of the ethyl-p-fluorophenyl moiety is overall consistent with that displayed by the same group in fragment 3, which was located in the hydrophobic cavity, rather than towards the solvent.
We decided to take advantage of the information derived from the crystal structures and docking calculations to merge the fragments, exploiting the common functional group like the amidic core or the piperazine ring as pivotal features for binding the Z-Site.

S5
Docking studies of fragment-derived compounds 6, 9, 10, and 14.Docking studies were performed for compounds representative of each group of modifications to confirm that the compounds designed by knowledge-guided growing strategy were able to address the desired hotspots while preserving the network of crucial interactions.
Compounds 6-8 were designed to probe modifications of the aromatic region interacting with the MBS in fragment-derived compound 5, and the binding mode of compound 6 was predicted by docking calculations (Figure S2).In the best ranked pose, the tricyclic scaffold (PTZ) of 6 favourably interacts with the MBS, particularly with H461, T110 and G49.Similarly, the orientation of the furan portion directed to the Z-site is preserved compared to the hit fragment 5, but the compound loses the interaction with the γ-Glu site.
Computational studies were performed on compounds 9 and 10 bearing N-alkylation on the piperazine ring (Figure S2).From docking calculations, the binding mode of 9 seemed preserved compared to 5 upon the introduction of a permanent positive charge by N-methylation of piperazine ring.Indeed, the quaternary nitrogen guides the electrostatic recognition with E466.Similarly, we preserved the cationic charge of quaternary nitrogen whilst conferring to the molecule a more hydrophobic surface contact by substitution of the methyl group with a diClBn moiety (10).
Consistently with the behaviour of 9, the interactions of 10 resulted preserved in the region of the ethyl-p-fluorophenyl and furan ring bound to the Z-site wall, and also regarding the salt bridge with E466.The prediction of the diClBn group contacting the MBS was explained by a series of weak Van der Waals interactions with T110.
As suggested from the analysis of docking pose of hit fragment 5 and crystal structure of 2, compounds 11-14 were designed by modification on region C.Among those, the docking pose of 14 (Figure S2) confirmed the possibility of removing the hydrophobic p-fluorophenyl ring exposed to the solvent without affecting the overall interactions.The conformation adopted by the derivative 14 is indeed similar to that of 10 (analogue, bearing the p-fluorophenyl group).
Overall docking studies confirmed that the desired hotspots were addressed by the different strategies of fragments optimization that led to the design of compounds 6-14.Moreover, in all cases, the interactions with the small cavity targeted by furan-moiety are consistent notwithstanding modifications of the other substituents directed to other hotspots.

Figure S1 .
Figure S1.Left: comparison between the binding modes of fragments 1-3 as predicted by the docking calculation (thin sticks) and the experimental binding mode (thick white sticks).Right.2D representation of the main interactions established by the best docking pose for each fragment with TR residues.

Figure S6 . 2 S12Figure S7 .S13Figure S8 .S14Figure S9 .S15Figure S10 .
Figure S6.Omit maps.Binding site of TR in complex with compounds 9 (A), 10 (B) and 14 (C).In the figure are represented the 2Fo-Fc maps (blue) contoured at 1σ of TR and Fo-Fc omit maps (green and red) contoured at 3σ of the complexes (in the absence of the compounds).2Fo-Fc (blue) andPolder Omit Fo-Fc (green and red) maps are contoured respectively at 1 and 3σ.Polder maps make weak densities, that can be obscured by bulk solvent, become visible, which is particularly recommended for ligands.2

Table S2 . Data collections, refinement, statistics, and validation.
The statistics for the highestresolution shell are shown in parentheses.R free is based on 5% of the data randomly selected that were not used during refinement.