TR from Trypanosoma brucei (TbTR) was expressed and purified as previously reported (Fiorillo et al., 2022; Exertier et al., 2024). In details, the gene of TbTR was subcloned in a pET15b vector. After transformation in E. coli BL21, a 10 mL starter culture was grown overnight to inoculate the next day 1 L of LB medium supplemented with antibiotics. The culture was incubated at 37°C, 180 rpm until the optical density reached 0.6. TbTR overexpression was induced upon addition of 1 mM IPTG. The cells were incubated at 37°C, 180 rpm for four more hours prior to harvest by centrifugation.

Cell pellets were resuspended in lysis buffer (50 mM Tris pH 8.0, 300 mM NaCl, 5 mM imidazole, 10 mM MgCl₂, 0.1 mM PMSF and one tablet of Pierce antiprotease cocktail). Cells were lysed by sonication, cell debris were discarded by centrifugation and the soluble phase was loaded onto a HiTrap column for immobilized-metal affinity chromatography (IMAC). Elution of TbTR occurred upon application of an imidazole gradient. The His-tag was cleaved from TbTR construct using 1U of Thrombin per 0.1 mg of protein. Tag-free TbTR was further purified from the protease using a benzamidine resin and from the uncleaved protein by reverse-IMAC. Finally, TbTR buffer was exchanged into 20 mM HEPES pH 7.4 by dialysis.

Crystallization plates were partly set up at the XCHEM facility of Diamond Light Source using a Mosquito crystal robot from SPT Labtech and partly set up in our home laboratory using an Oryx4 robot from Douglas Instruments. In both cases, TbTR crystallizes at 17 mg/mL in 22% 2-Methyl-2,4-pentanediol (MPD), 14% PEG3350, imidazole 40 mM pH 8 and 50 mM NaBr. Surprisingly, the crystallization conditions were similar to those used in the previous screening, yet yielded crystals with a different symmetry (monoclinic vs. orthorhombic). This behavior, which we could not rationalize or control, is attributed to variations between sample batches. An ECHO acoustic liquid dispenser was used to dispense 503 fragments from second half of DSI Poised and the entire EubOPEN libraries onto the crystallization drops, resulting in a maximum concentration of 50 mM fragment and 10% DMSO (Collins et al., 2017). Soaked crystals were then incubated for 1 h at 20°C. Crystals were rapidly mounted on cryoloops thanks to the OLT shifter (Wright et al., 2021) and directly flash frozen in liquid nitrogen prior the diffraction experiments, which were carried out both at the I03 and I04-1 beamline. Data processing was carried out using the XCHEM Explorer (XCE) pipeline (Krojer et al., 2017). Hits were identified using Pan-Dataset Density Analysis (PANDDA) (Pearce et al., 2017), a first refinement was performed by Buster (Smart et al., 2012), implemented on XCE. The final refinement of the occupancy and geometry was performed using Phenix-1.21.2-5419 (Liebschner et al., 2019). Supplementary Table S1 reports the data collection and refinement statistics.

Surface Plasmon Resonance (SPR) experiments were carried out to assess the thermodynamic parameters of the interaction between TbTR (ligand) and analytes (fragments) in a Sartorius Octet SF3 apparatus. The ligand was immobilized on an Octet CDL Sensor Chip, which consists of a carboxymethyl dextran 3D matrix which allows a highly stable covalent bond with the ligand. The ligand was immobilized with amine coupling chemistry. The amount of immobilized TbTR was detected by mass concentration-dependent changes in the refractive index on the sensorchip surface, and corresponded to about 3,800 resonance units (RU). Analytes were dissolved in 100% DMSO at a concentration of 50 mM, and subsequently diluted in sterile HEPES 20 mM, pH 7.4, NaCl 150 mM, 0.005% surfactant Tween20 to yield 2% DMSO final concentration (HSP-2%D buffer) and final analyte concentration: 1 mM. Further dilutions and all the experiments were carried out at 25°C in degassed HSP-2%D buffer. After ligand immobilization, analytes were injected at different concentrations, i.e., 33 μM, 100 μM, and 300 μM, into a moving stream of buffer (Valentini et al., 2022; Masciarelli et al., 2014). At higher concentrations, the signal traces became increasingly difficult to interpret, likely due to interference from non-specific binding and contributions from solubilized components in the solution. In all experiments, the increase in RU relative to baseline indicates complex formation between the immobilized TbTR ligand and the fragments (analytes). The plateau region, when present, represents the steady-state phase of the interaction. The decrease in RU after 200 s indicates analyte dissociation from the immobilized ligand after HSP-2%D buffer injection. As a negative control, sensor chips were treated as described above in the absence of immobilized TbTR. Values of the plateau signal at steady-state (Req) and full fittings with 1 and 2 sites were calculated from overall kinetic evaluation of the sensorgrams using the Octet SPR Analysis software with a one-site kinetic model.

Fragments were purchased from Enamine and were solubilized in DMSO to reach a stock concentration of 50 mM. Typically, TbTR enzymatic reaction at 25°C was followed in the presence of 50 nM TbTR, 100 μM fragments (2% DMSO final concentration), 150 μM trypanothione, 100 uM NADPH in 50 mM HEPES pH 7.4, 40 mM NaCl at 340 nm using a spectrophotometer (JASCO V650) equipped with a Peltier (JASCO EHC 716). Experiments were carried out only for the fragments identified in catalytic cavities (trypanothione and NADPH cavities). The reaction initiated upon NADPH addition was followed at 340 nm. Initial velocities were used to calculate the residual activity of TbTR compared to the activity in the absence of fragments. Each data point is the average of triplicate experiments. Data analysis was performed with Prism 9, and graphs were made with Labplot2.

Fragment screening has been performed on TbTR monoclinic crystals grown in 13% PEG3350, 24% MPD, 40 mM imidazole pH 8, 50 mM NaBr. It is worth mentioning that although the crystallization conditions are rather identical, the previously reported fragment screening experiment was performed on orthorhombic crystals of TbTR (Fiorillo et al., 2022). The monoclinic packing did not affect the quality of crystals, as these TbTR P2₁ crystals diffracted at high resolution (1.62–2.11Å) compared to the orthorombic ones (1.65–1.97Å, Fiorillo et al., 2022). Molecular replacement revealed the presence of two TbTR homodimers in the asymmetric unit instead of the single homodimer observed for the orthorhombic crystals, and the different packing arrangement did not significantly alter the solvent-accessible surface area (SASA) of the TbTR dimer (Supplementary Figures S1, S2).

The 1-h soaking of monoclinic TbTR crystals with 50 mM fragments from the second fraction of DSIpoised and the entire EubOPEN libraries, designed for favoring a follow-up chemical expansion (Cox et al., 2016) was performed at the XCHEM facility while X-ray diffraction was carried out at both Diamond Light Source I03 and I04-1 beamlines.

Eight clear hits have been identified using the Pan-Dataset Density Analysis (PanDDA) implemented on the semi-automatic XCHEM-Explorer pipeline (Supplementary Table S2). These fragments were observed in five different sites with 0.6–1.0 occupancy (Figure 1; Table 1): (i) the most populated site is the trypanothione cavity, where compounds target several sub-pockets; (ii) the NADPH cavity, in the close vicinity of the doorstop pocket; (iii) three other sites accounting for pockets with no known enzymatic function. Notably fragment 79, a rather promiscuous compound, is found in three distinct cavities.

Over the years, the trypanothione cavity has been the preferred target for the design of TR inhibitors (Exertier et al., 2024; Baiocco et al., 2013; Battista et al., 2020; Ilari et al., 2018; Battista et al., 2022; Turcano et al., 2018; Colotti et al., 2020; Patterson et al., 2011). This is primarily due to the larger size of the cavity and its distinct surface charge distribution compared to that of the glutathione cavity found in the closest human homolog, glutathione reductase (GR), which provide a significant advantage in the development of selective inhibitors through structure-activity relationship (SAR) studies. We counted four fragments accommodating in the trypanothione cavity. The “trypanothione cavity” inset of Figure 1 clearly shows that fragments may be divided in two groups: those that interact with the so-called mepacrine binding site (MBS), i.e., a hydrophobic patch localized at the entrance of the trypanothione cavity; and those binding to the z-site, a more profound hydrophobic alcove.

In the z-site, event maps arising from the PanDDA analysis clearly indicated the presence of fragments 79 and 350 (Figures 2A,B; Supplementary Table S2). Importantly, 79 contains a piperazine ring linked to a phenyl group by an amide. The piperazine is stabilized due to interactions with E467; this interaction has also been observed previously with piperazine-containing fragments binding to TbTR (Supplementary Figure S3) (Fiorillo et al., 2022), although abolished upon merging and elongation of these fragments (Exertier et al., 2024).

Other two fragments, namely 28 and 133, bind to the mepacrine binding site but in rather separate and mutually exclusive locations (Figures 2C,D). Interestingly, fragment 133 occupies a sub-cavity generated by the displacement of M113. A similar mode of binding has been observed for dihydroquinazoline compounds (DHQ), that account for some of the most potent TR inhibitors (Figure 2C; Patterson et al., 2011). Indeed, fragment 133 enters deeper into this sub-cavity compared to DHQs, although the moiety exposed towards the cavity is less bulky.

We identified only one hit in the NADPH binding region, close to the so-called doorstop pocket (Ardini et al., 2024). F198 plays the role of a gatekeeper: in the absence of NADPH, F198 is locked in a perpendicular position with respect to the FAD. When NADPH enters the cavity however, F198 swings out from its position to allow the nicotinamide moiety to interact with the FAD isoallaxazine ring (Battista et al., 2020). Ligand binding within the doorstop pocket can restrict the movement of F198, consequently hindering NADPH binding and TR activity.

Fragment 108 is stabilized in this position mainly through hydrophobic interactions. In fact, it is surrounded by numerous apolar side chains, amongst them F198, F230, L332, L334 and V362. Nonetheless, the position of this fragment is also supported by H-bonds with water molecules (Figure 3).

Fiorillo and colleagues previously reported the identification of several other fragments accommodating in the doorstop pocket, with the hypothesis that they could act as allosteric inhibitors. Amongst them, Fiorillo’s 68 (hereafter referred to as F-68) also contains two amide groups connected to a nitrogen heterocycle (piperidine instead of piperazine), and its position partially overlaps to that of fragment 108 (Supplementary Figure S4; Fiorillo et al., 2022).

Four fragments were found in other binding sites with no known enzymatic activity. One of these sites, containing the three fragments 35, 42 and 79 (Figure 4), was already detected in the previous fragment screening performed by Fiorillo and coworkers.

This site is rather exposed to solvent and close to the dimerization interface, involving both TR monomers through the protrusion of W61’ in the bottom of the cavity lined by M70, R74, V88 and K211. It must be noted that the binding of fragment 79 in this site is likely an artifact due to crystal packing. Indeed, although the binding pose of 79 is stabilized by a meaningful polar interaction between the amide oxygen and the side chain of K211, its benzene ring extends out of the cleft apparently towards the solvent but actually toward adjacent protein subunit. In contrast, 35 and 42 fully adhere to the cleft and share a common benzene ring whose position is conserved. Notably, other ligands have been previously detected in the same site, namely the fragment F-60 (Fiorillo et al., 2022) and a drug-like inhibitor binding this cleft as a secondary site (Turcano et al., 2020).

The two other non-enzymatic binding sites contain either fragment 79 or 115 (Supplementary Figure S5). These two sites are mostly superficial grooves that however show a certain affinity, although probably low, for these two fragments. While Fiorillo and colleagues already observed fragments binding to 115 groove, 79 binds to a shallow site that has never been identified before. In these last three binding sites, the occupancy of each fragment is relatively moderate (0.6–0.7) and interactions with the peptidic chains are rather scarce.

The eight fragments identified here plus seven selected from the previous screen were tested by SPR to assess their affinity for TbTR. As previously stated, fragments are characterized by low mass (150–300 Da) and low affinity (μM-mM), making the characterization of their binding through SPR a challenging task that requires extensive experimental effort for optimization (Tiwari et al., 2021). However, in our study, we were more interested in obtaining a rough estimation of binding affinity rather than an exact measurement of the dissociation constant (K_D) to rank the fragments for their binding properties. Thus, by testing three concentrations (33, 100, 300 μM), we could detect binding for 12 out of 15 fragments with TbTR, albeit with K_D values in the submillimolar-millimolar range (Table 2; Supplementary Figure S6). The compounds endowed with the highest affinities were 28 and 350 in the trypanothione cavity, 108 and F-117 in the NADPH binding site, and 42 and F-60 in the non-enzymatic cavity. Although none of these ligands emerges as a particularly strong binder, their detectable binding indicates that some may potentially exert inhibitory effects on TbTR.

The fragments found in the substrate cavities, namely the trypanothione and NADPH cavities, during the fragment screening reported in this work and in Fiorillo et al., 2022 were tested against TbTR to evaluate their inhibition activity (Table 2). Most of these fragments do not significantly alter the activity of TbTR even at a concentration of 100 μM. This is not unexpected as fragments are usually small low-affinity ligands and, in this context, their binding may not be sufficient to impede the entry and positioning of the substrates. Consequently, these fragments do not substantially hinder TbTR’s ability to perform its full and effective reduction of trypanothione.

Unlike hypothesized above, 108 does not affect TbTR activity, suggesting that the fragment does not sterically hinder the swinging movement of the gatekeeper F198.

However, it is worth mentioning that fragment F-117 at least inhibits TbTR reduction of trypanothione of ∼11% (Residual TbTR activity = 89.4%). F-117 has been identified in the doorstop pocket and appears to be among the fragments that protrude the most towards the FAD gatekeeper F198 (Supplementary Figure S4). This suggests that F-117 is capable to obstruct the movement of F198 side chain, thereby preventing NADPH nicotinamide moiety from positioning above the FAD isoalloxazine ring, disturbing the transfer of electrons from NADPH to the FAD. Thus, F-117 emerges as a good candidate for elongation and merging and a promising starting point for the design of a potent inhibitors.

Notably, a similar a strategy has been successfully applied for the design of noncovalent inhibitors targeting the doorstop pocket of the thioredoxin glutathione reductase from Schistosoma mansoni (Petukhova et al., 2023).