Deconstruction of Aspartic Protease Inhibitors Enables Fragment‐Based Discovery of Plasmepsin V Inhibitors

Abstract

Malaria, caused by the Plasmodium parasites, remains a major global health burden, and resistance to current antimalarials drives the need for drugs with new mechanisms of action. Plasmepsin V (PMV), an essential aspartic protease required for PEXEL processing and protein export, is structurally divergent from human aspartic proteases, offering a path to selective inhibition. Here, we applied a structure‐informed, deconstruction‐based approach to identify non‐peptidomimetic PMV inhibitors by mining catalytic dyad‐binding motifs from experimentally solved aspartic protease–inhibitor complexes and assembling a focused fragment‐like library. Fragment‐inspired model compounds based on pyrrolidine, piperidine, and piperazine cores showed measurable PMV inhibition in a Forster resonance energy transfer (FRET)‐based assay, and a trans‐3,4‐disubstituted pyrrolidine hit (7a; IC₅₀ 70  µM) was selected for optimization. Guided by PMV structural data and a renin cocrystal structure of a related pyrrolidine inhibitor, we explored SAR around substituents intended to engage the S1 and S2 regions. Optimization identified N‐sulfonamide analogs bearing two aromatic substituents as a preferred chemotype, with meta‐substitution on the N‐aryl group improving potency to the low micromolar range. The most potent compounds obtained displayed ~10 µM potency, establishing a promising non‐peptidomimetic scaffold for further development of selective PMV inhibitors.

A structure‐informed, fragment‐based approach was used to identify non‐peptidomimetic inhibitors of plasmepsin V. Pyrrolidine, piperidine, and piperazine containing fragments showed activity in a FRET assay. The optimization of the pyrrolidine scaffold resulted in N‐sulfonamide analogs with potency up to ~10 µM, establishing a promising scaffold for further inhibitor development.

1. Introduction

Malaria is a deadly parasitic infection that poses an enormous threat to global health [1]. It is caused by Plasmodium parasites, with P. falciparum being the most lethal human pathogen. While drugs are available to treat the disease, widespread resistance to practically all antimalarials in clinical use is motivating scientists to seek drugs with novel mechanisms of action [1, 4]. The life cycle of P. falciparum involves three stages: the mosquito, liver (pre‐erythrocytic), and blood (erythrocytic) stages, of which the blood and liver stages occur within the human body [5, 6].

The P. falciparum genome encodes 10 aspartic proteases, named plasmepsins (PMI to X). The PMs have diverse roles across the parasite’s lifecycle, and several essential PMs are considered attractive antimalarial drug targets. These enzymes can be fundamentally categorized into distinct subclasses based on their spatial localization, stage‐specific expression, and biological functionality. PMs I, II, IV and histo‐aspartyl protease (HAP) are localized to the digestive vacuole of the asexual parasite and degrade hemoglobin, providing sustenance for parasite development [6, 7]. For a long time, these proteins were considered essential for parasite survival and thus attractive antimalarial drug targets; however, despite their clear physiological role, targeted gene‐disruption studies have revealed a high degree of functional redundancy among these enzymes, meaning that single‐knockout variants remain viable [8]. Plasmepsins VI–VIII are expressed in the mosquito stages, and their roles remain undefined. Evidence suggests PMs VI and VIII are essential for P. berghei development in the mosquito, while PMVII appears to be redundant and nonessential [9, 10, 11]. PMs IX and X have high structural homology and are both indispensable for asexual parasite development. PMIX is essential for erythrocyte invasion, acting on rhoptry secretory organelle biogenesis, whereas PMX is essential for both egress and invasion, controlling maturation of the subtilisin‐like serine protease SUB1 in exoneme secretory vesicles [12, 13, 14, 15, 16]. Since PMIX and PMX share high sequence homology with digestive PMI–PMIV, these enzymes are considered the main targets of the digestive vacuole plasmepsin inhibitors that have shown antimalarial activity in vivo [14, 17, 18, 19]. Plasmepsin V (PMV) is an essential protease that plays a critical role in the export of proteins from the parasite cytosol to the host erythrocyte. Exported proteins play crucial roles in parasite survival, such as providing sustenance for the parasite to replicate, evasion of the host immune system, and nutrient/waste efflux from the host erythrocyte. Many proteins that are exported have a five‐amino acid N‐terminal motif, with the consensus sequence RxLxQ/E/D, known as the Plasmodium export element (PEXEL) [20, 21]. Due to the high PEXEL motif and PMV conservation across all Plasmodium parasites, and the fact that PMV is quite divergent from the other plasmepsins, it is a promising essential antimalarial drug target [22, 23, 24]. Most digestive vacuole PM inhibitors are not potent against PMV.

An important issue when developing pathogen enzyme inhibitors is selectivity over human aspartic proteases. PMs share high sequence similarity with several human aspartic proteases (beta‐secretases (BACEs), cathepsins, renin, pepsin), with the most similar being the lysosomal enzyme cathepsin D (catD); therefore, it is commonly used as a marker for cross‐inhibition [25, 26, 27]. Due to the high protein structural similarity, the design of potent plasmepsin inhibitors that do not inhibit human aspartic proteases remains a challenging task [28]. PMV is the most structurally divergent of the plasmepsins and shows low similarity to human aspartic proteases; thus, it could be one of the most promising targets for the design of selective antimalarials. Moreover, higher selectivity is typically achieved by non‐peptidomimetic inhibitors, since PMs can bind such inhibitors partly under the flap loop (a single long β‐hairpin structure that lies perpendicularly over the aspartic dyad), where an additional hydrophobic pocket is formed [29]. Up until now, however, there is no structural information that confirms the existence of such a hydrophobic pocket for PMV. Nevertheless, encouraged by the structural uniqueness of PMV, we set out to identify non‐peptidomimetic inhibitors that should enable the design of selective PMV inhibitors.

2. Results and Discussion

Decades ago, the first PMII inhibitors were designed by repurposing scaffolds from known aspartic protease inhibitors, primarily those targeting renin, BACE, and HIV proteases [29]. To date, more than 200 aspartic protease and inhibitor cocrystal structures have been reported in the literature, approximately 80 of which are with non‐peptidomimetic inhibitors. Here, we sought to identify non‐peptidomimetic PMV inhibitors by deconstructing known aspartic protease inhibitors with experimentally resolved binding modes. Based on available cocrystal structures, predominantly of BACE and renin, we constructed a tailored library of scaffolds that match the general structures 1 and 2 (Figure 1B). Since these scaffolds mimic the transition state of the peptide bond cleavage (Figure 1A), it is expected that these compounds will interact with a wide variety of aspartic proteases, including PMV. In these compounds, the hydrogen‐bond donor (D) mimics the catalytic water molecule, while the hydrogen‐bond acceptor (A) mimics the peptide bond carbonyl group. In all cocrystallised non‐peptidomimetic inhibitors, the hydrogen‐bond donor is an amino group: a primary amine in compounds corresponding to general structure 1, and a secondary or tertiary amine in compounds corresponding to general structure 2. Notably, compounds matching general structure 2 lack an explicit hydrogen‐bond acceptor.

FIGURE 1.

(A) The proposed mechanism of the aspartic protease‐catalyzed reaction [30]. (B) General structures of non‐peptidomimetic aspartic protease inhibitors (A indicates hydrogen bond donor, D—hydrogen bond donor). (C) Aspartic dyad‐binding fragments extracted from reported aspartic protease non‐peptidomimetic inhibitor complexes. The transition state mimicking hydrogen bond donor is highlighted in red. For each scaffold, the target protease and a representative complex structure ID are shown. (D) Model compounds designed by functionalizing the extracted aspartic dyad–binding fragments (from panel C) with synthetically accessible hydrophobic substituents. Compound potency against PMV is reported as the fraction of protein inhibited at an inhibitor concentration of 250 μM. The compound potency determination here and further was performed in duplicate, and measurement standard deviations (SD) are available in Table S1.

The scaffolds of potential aspartic dyad‐binding fragments (Figure 1C) were extracted from resolved aspartic protease structures, focusing on compounds that match the general structure 2. These were predominantly pyrrolidine, piperazine, and piperidine‐containing fragments, in which a secondary amine is positioned to mimic the catalytic water molecule and engage in hydrogen bonding with the catalytic dyad. To evaluate the potential of these scaffolds as potential PMV inhibitors, model compounds (Figure 1D) were designed and synthesized. Selected scaffolds were functionalized with isopropyl and phenyl group‐containing substituents intended to occupy the S1 and S2′ subpockets, as indicated by the available PMV‐inhibitor cocrystal structures (PDB IDs: 4ZL4, 6C4G, 8TYF, 8TYG) [23, 31, 32]. Linker lengths were systematically varied to enable appropriate positioning of these substituents within the respective binding sites.

Given the fragment‐like nature of these compounds, low inhibitory potency was anticipated. Accordingly, initial PMV inhibition was evaluated at a single concentration of 250 μM, and IC₅₀ values were determined only for the most potent compounds. Compounds derived from all investigated scaffolds exhibited partial inhibition of PMV in a Forster resonance energy transfer (FRET)‐based assay (see Methods), supporting the validity of the proposed strategy to identify potent hits. Among these, compound 7a emerged as the most potent, inhibiting PMV by 78% at 250 μM, with an IC₅₀ value of 70 μM. Based on its favorable activity and synthetic tractability, the pyrrolidine scaffold 7 was selected for further optimization.

The availability of the binding mode of chlorinated 7a in complex with renin (PDB ID: 4GJ6) [33], and subsequently optimized trans‐3,4‐disubstituted pyrrolidines (PDB IDs: 4RYC, 4RYG, 4RZ1, 4GJ7, 4GJA) [34, 35, 36], enabled structure‐based drug design even without experimental 7a‐PMV binding data (Figure 2). The computational model was created by merging the chlorinated 7a‐renin complex 4GJ6 with the available PMV structure 4ZL4. The minimized model served as the basis for guiding the hit optimization campaign. Notably, prior optimization efforts in the renin series [35, 36] demonstrated that modifications targeting the S1 subpocket resulted in substantially greater potency gains than those directed toward the S2′ subpocket. Consistent with these observations, subsequent 7 optimization efforts in the present study were focused primarily on improving interactions within the S1 subpocket.

FIGURE 2.

The binding mode of peptidomimetic PMV inhibitor WEHI‐842 (salmon; PDB ID: 4ZL4) and compound 7a (teal). Inhibitors and aspartic dyad residue sidechains (gray) are shown as sticks. Yellow dashed lines indicate hydrogen bonds. Hydrogen atoms are omitted for clarity. Superimposed is the aspartic protease binding subpocket numbering scheme.

Since both the tertiary and secondary amine‐containing trans‐3,4‐disubstituted pyrrolidine hits (7a and 7b, respectively) showed high potency, the initial efforts focused on understanding whether the tertiary amine at this position is required or whether a secondary amine is preferable. Various hydrophobic acyclic and cyclic substituents were installed at this position to explore the structure–activity relationship (SAR) of mono‐substituted S1 binding substituents (Figure 3A). The highest PMV inhibition potency was observed for the phenyl (7b) and diaryl substituent‐containing compounds (7f and 7g), reaching 74% inhibition at 250 μM; however, no substantial activity gain over 7a was observed. Compounds bearing non‐aromatic substituents (7m and 7n) showed low potency (>40% inhibition at 250 μM), whereas analogs containing a decorated benzene ring provided moderately active compounds (60%–40% inhibition at 250 μM).

FIGURE 3.

SAR of a secondary (A) and tertiary (B) amine containing trans‐3,4‐disubstituted pyrrolidines. The fraction of PMV inhibited at 250 μM inhibitor concentration is shown under the molecular structures. Compounds are arranged by activity. The IC₅₀ value was determined for the most potent compounds.

Since none of the secondary amines showed substantially improved PMV potency over the tertiary amine containing 7a, the SAR of the tertiary amines was explored next. Within this series, one of the substituents, benzene, was kept constant, while the second substituent was varied (Figure 3B). Substituents installed were benzene rings with various linkers, isopropyl, and methyl groups. Apart from carbonyl group‐containing 7s, all compounds inhibited PMV by at least 50% at 250 μM. The highest inhibition at this concentration was observed for the original hit 7a and sulfonamide 7o. While both compounds showed the same fraction of PMV inhibited at a 250 μM concentration, the determined IC₅₀ values showed that 7o was slightly more potent than 7a (60 and 70 μM, respectively). The discrepancy between the determined IC₅₀ value and the fraction inhibited at a single point on the dose–response curve could result from differences in curve slope, partial efficacy, and/or experimental noise. Unlike trans‐3,4‐disubstituted pyrrolidines as renin inhibitors, the isopropyl‐containing inhibitor (7q) showed lower potency than the benzene‐containing analogs [33].

The initial optimization efforts indicated that the N‐substituted sulfonamides bearing two aromatic substituents (e.g., 7o) exhibit superior PMV inhibitory activity. Consequently, subsequent optimization efforts focused on systematic modification of these substituents. The two aromatic groups are presumed to occupy the S1 and S2 subpockets; however, at this stage, molecular modeling could not unambiguously determine the orientation of the sulfonamide moiety or assign each substituent to a specific pocket. Therefore, both substituents were independently optimized to effectively engage the more enclosed S1 subpocket, while the remaining substituent was kept relatively small, reflecting the greater solvent exposure of the S2 subpocket.

Initially, the SAR of the N‐substituent was investigated within the sulfonamide subset 9, while the phenyl sulfonamide group was kept constant (Figure 4). Small hydrophobic meta‐substituents on the phenyl ring were well tolerated, with compounds 9a and 9b showing the highest potency (IC₅₀ of 21 and 44 μM, respectively). In contrast, introduction of a linear ethynyl group 9c or bulkier substituents at the meta‐position (9e,g,h) led to reduced activity, suggesting that the corresponding binding site cannot accommodate substantially larger substituents. The para‐substituted analogs (9l–n) were considerably less potent than their respective meta‐substituted analogs (9g,h,a). Moreover, replacement of the aromatic N‐substituent with an aliphatic group (9o–s) led to a significant loss of potency (<42% PMV inhibition at 250 μM), highlighting the importance of an aromatic moiety at this position. Installation of flexible benzyl carbamate substituents (9i–k) did not improve the compound potency; nevertheless, compounds showed slightly higher potency than the para‐substituted analogs. The observed SAR trends, together with molecular modeling studies, suggest that the N‐substituent occupies the S1 subpocket, where the benzene ring engages in π–π interactions with Phe137, while the meta‐substituent is directed into a small adjacent cavity. According to molecular modeling, the poor potency of the para‐substituted analogs could be a result of steric clashes with Ile78.

FIGURE 4.

Parent compound 9 and SAR of the N‐substituent. Compound potency (IC₅₀ or fraction of PMV inhibited at either 100 or 250 μM inhibitor concentration) is shown under the molecular structures. Compounds are arranged by activity. The reference sulfonamide 7o is shown in teal. Docked pose of compound 9a in complex with PvPMV. Compound 9a (teal) and key PMV residues (gray) are shown as sticks; cocrystallised PMV inhibitor WM36 (PDB ID: 8TYF) is shown as salmon lines. The flap loop is shown in blue. Yellow and blue dashed lines indicate hydrogen bonds and aromatic stacking, respectively. Nonpolar hydrogen atoms are omitted for clarity.

In the subsequent optimization series, efforts were directed toward the optimization of the sulfonamide substituents while retaining a phenyl group as N‐substituent (Figure 5). Within this series, biaryl systems (10a,d,e,g,h,j,k) and bicyclic analogs (10b,f) were generally well tolerated, with several compounds exhibiting IC₅₀ values in the 20–30 μM range. Comparable activity was observed for the dichloro‐substituted thiophene 10c and its phenyl bioisostere 10i (IC₅₀ 24 and 31 μM, respectively), indicating that the expected π–π stacking interactions introduced by the terminal ring of the biaryl system do provide a significant gain in potency. The terminal aromatic ring of the biaryl analogs is expected to bind in a rather shallow S2 pocket, where it is π–π stacking with His320. Substitution pattern analysis revealed that meta‐ and para‐substituted biaryl analogs consistently displayed superior potency compared to ortho‐substituted counterparts (e.g., 10d and 10k vs. 10q), highlighting the importance of substituent orientation for optimal binding in the S2 subpocket and minimizing the steric clashes with the sulfonamide group. In addition to the substitution pattern, linker geometry proved to be important for compound potency. Biaryl systems connected through an sp²‐hybridized linker (10e,h) exhibited greater potency than compounds with an sp³‐hybridized or ether linker (10n,o,r), suggesting a preference for planar and conformationally restricted motifs. No potency increase with respect to 7o was observed upon heteroatom or polar substituent (phenols, methoxy) introduction on the phenyl ring, or replacement of the aromatic moiety with a methyl group.

FIGURE 5.

Parent compound 10 and SAR of the sulfonamide substituent. Compound potency (IC₅₀ or fraction of PMV inhibited at either 100 or 250 μM inhibitor concentration) is shown under the molecular structures. Compounds are arranged by activity. The reference sulfonamide 7o is shown in teal. Docked pose of compound 10a in complex with PvPMV. Compound 10a (teal) and key PMV residues (gray) are shown as sticks; cocrystallised PMV inhibitor WEHI‐842 (PDB ID: 4ZL4) is shown as salmon lines. The flap loop is shown in blue. Yellow and blue dashed lines indicate hydrogen bonds and aromatic stacking, respectively. Non‐polar hydrogen atoms are omitted for clarity.

Finally, the preferred N‐substituent identified, an m‐methylphenyl group (9a, IC₅₀ 21 μM), was combined with the best‐performing sulfonamide substituents 10a–c (IC₅₀ ~ 20 μM), yielding compounds 11a–c with approximately twofold improved potency (IC₅₀ ~ 10 μM, Figure 6). The improvement in potency indicates that the contributions of the N‐aryl and sulfonamide moieties are complementary and can be simultaneously accommodated within the binding pocket. The selectivity of these compounds was evaluated by testing the inhibition of the human catD, commonly used as an aspartic protease cross‐inhibition marker. Compounds 11a and 11b were equally potent inhibitors of PMV and catD, with poor selectivity, whereas compound 11c showed some selectivity, inhibiting PMV fivefold more than catD.

FIGURE 6.

Trans‐3,4‐disubstituted pyrrolidines designed by combining the most potent N‐substituents and sulfonamide substituents. Compound potency (IC₅₀) is indicated under the molecular structures.

3. Conclusion

Here, we have presented a method for identifying non‐peptidomimetic PMV inhibitors by mining catalytic dyad‐binding motifs from experimentally solved aspartic protease–inhibitor complexes. We exploited a structure‐informed, deconstruction‐based approach to assemble a focused fragment‐like library that contained pyrrolidine, piperidine, and piperazine‐based compounds. The fragment‐like model compounds showed measurable PMV inhibition, with tertiary amine‐containing pyrrolidine 7a being the most potent one (IC₅₀ 70 μM). Structure‐guided optimization of 7a led to the development of sulfonamide‐based trans‐3,4‐disubstituted pyrrolidines as a new class of PMV inhibitors. Conversion to the sulfonamide scaffold afforded an initial potency improvement and introduced a motif, enabling exploration of S1 and S2 subpockets binding substituents. These studies revealed clear preferences for hydrophobic aromatic substituents in both regions, with optimal activity achieved for inhibitors bearing a m‐methylphenyl group and biaryl or dichlorothiophene sulfonamide substituents. A combination of these features delivered compounds 11a–c with IC₅₀ values of ~10 μM, representing a substantial improvement over the initial hit. The most potent compounds showed poor selectivity over human catD. Taken together, these findings establish key SAR for this scaffold and identify sulfonamide‐containing trans‐3,4‐disubstituted pyrrolidines as a promising starting point for further optimization of PMV inhibitors.

Funding

This work was supported by NextGeneratioEU (5.2.1.1.i.0/2/24/I/CFLA/001).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

The Supporting Information contains synthetic, protein expression, enzymatic testing, and computational methods used.

Supplementary Material

Data Availability Statement

The data that supports the findings of this study are available in the Supporting Information of this article.