Highly accurate protein structure prediction-based virtual docking pipeline accelerating the identification of anti-schistosomal compounds

TL;DR

This study uses AI to predict protein structures in parasitic worms and identifies new drug candidates to treat schistosomiasis, a tropical disease.

Contribution

A novel drug discovery framework combining AI-based protein structure prediction with virtual screening and molecular validation for schistosomiasis.

Findings

Virtual screening identified 268 candidate compounds, seven of which showed strong anti-schistosomal activity.

Carubicin hydrochloride (CH) was the most potent compound, targeting the splicing factor U2AF65.

CH treatment and gene knockdown of U2AF65 disrupted splicing events and parasite viability.

Abstract

Schistosomiasis is a major neglected tropical disease that lacks an effective vaccine and faces increasing challenges from praziquantel resistance, underscoring the urgent need for novel therapeutics. Target-based drug discovery (TBDD) is a powerful strategy for drug development. In this study, we utilized AlphaFold to predict the structures of target proteins from Schistosoma mansoni and S. japonicum, followed by virtual molecular screening to identify potential inhibitors. Among 202 potential therapeutic targets, we identified 37 proteins with high-accuracy structural predictions suitable for molecular docking with 14,600 compounds. This screening yielded 268 candidate compounds, which were further evaluated ex vivo for activity against both adult and juvenile S. mansoni and S. japonicum. Seven compounds exhibited strong anti-schistosomal activity, with HY-B2171A (Carubicin hydrochloride, CH) emerging as the most potent. CH was predicted to target the splicing factor U2AF65, and knockdown of its coding gene Smp_019690 resulted in a phenotype similar to CH treatment. RNA sequencing revealed that both CH treatment and Smp_019690 RNA interference (RNAi) disrupted splicing events in the parasites. Further studies demonstrated that CH impairs parasite viability by inhibiting U2AF65 function in mRNA splicing regulation. By integrating RNAi-based target identification with structure-based virtual screening, alongside ex vivo phenotypic and molecular analyses of compound-treated schistosomes, our study provides a comprehensive framework for anti-schistosomal drug discovery and identifies promising candidates for further preclinical development.

Schistosomiasis is a debilitating disease caused by parasitic worms and affects over 230 million people worldwide. For decades, control of the disease has relied on a single drug, praziquantel, which is less effective against juvenile parasites and increasingly threatened by the risk of drug resistance. To accelerate the discovery of new treatments, we integrated advanced computational tools with molecular validation. Using AlphaFold-predicted structures of 37 key schistosome proteins, we performed large-scale virtual screening to identify promising drug candidates. Over 200 compounds were prioritized for phenotypic screening on two major human-infective schistosome species, leading to the discovery of seven with potent anti-parasitic activity. Among them, Carubicin hydrochloride showed strong efficacy by targeting a critical splicing factor, U2AF65. Inhibiting this protein—either chemically or through gene silencing—disrupted parasite movement, suppressed cell proliferation, and altered splicing events. Our work highlights the power of integrating AI-based protein modeling with phenotypic and mechanistic validation to accelerate drug discovery for neglected tropical diseases like schistosomiasis.

Figures

Workflow for the screening of putative anti-schistosomal compounds.This representation summarizes the inputs, outputs and filter criteria used in this study, leading to the selection of 268 compounds for ex vivo screening (see methodology for full details).

Phenotypes of adult S. mansoni and S. japonicum treated with high-efficiency compounds.The figure depicts the phenotypic changes in S. mansoni (A) and S. japonicum (B) after a 7-day co-incubation with high-efficiency compounds at a concentration of 10 μM. Negative control group: DMSO (0.1% DMSO, at the same DMSO concentration as in the experimental group.); Positive control group: Praziquantel (1 µM in 0.1% DMSO). Pink box denotes species-specific high-activity. Representative images from 3 biological replicates. Scale bars, A, 1000 μm; B, 2000 μm.

Phenotypes of juvenile S. mansoni and S. japonicum treated with high-efficiency compounds.Bright-field images showcasing the phenotypes of juvenile worms after 7-day co-incubation with potent compounds (10 μM). (A) S. mansoni (14 compounds); (B) S. japonicum (16 compounds). 0.1% DMSO as negative control; 1 µM Praziquantel in 0.1% DMSO as positive control. Pink box denotes species-specific high-activity. Representative images from 3 biological replicates. Scale bars, 500 μm.

Molecular Docking of effective compounds with their targets and effects on schistosomes.(A) Molecular docking diagrams of active compounds with the target. The C backbone of the target protein is depicted in green, nitrogen atoms (N) in blue, oxygen atoms (O) in red, and hydrogen atoms (H) in white. The compounds are shown as sticks with the following color codes: dark blue for HY-12462 and HY-16322, magenta for HY-112852, light green for HY-15004, and light blue for HY-15196, HY-12382, and HY-B2171A. π-π interactions are indicated by green dashed lines, and hydrogen bonds are shown as purple dashed lines, with longer hydrogen bond lengths reflecting weaker interactions. Light blue dashed box: molecular docking of compound CH with binding site of SjU2AF65 (EWB00_008591) and SmU2AF65 protein (Smp_019690). (B–E) Bright-field images showing representative compound-induced phenotypes following 7-day co-incubation with 1 μM compounds. (B) Adult S. mansoni; (C) Adult S. japonicum; (D) Juvenile S. mansoni; (E) Juvenile S. japonicum. Scale bars: B, 1000 μm; C, 2000 μm; D–E, 500 μm. Representative images from 3 biological replicates.

Stem cell maintenance and transcriptomic changes in parasites following Smp_019690 knockdown or CH treatment.(A) Phenotype of adult S. mansoni following Smp_019690 or control RNAi (left, n = 15-18), and EdU labeling of proliferative cells in male parasites (right). Representative bright-field images from 3 biological replicates. Scale bars, 1000 μm. For EdU labeling, nuclei were stained with DAPI (blue) and proliferative cells were shown in magenta. Testes are outlined with dashed yellow circles. n = 15 worms from 3 biological replicates, Scale bars, 100 μm. (B) Relative mRNA expression levels of Smp_019690 in parasites treated with Smp_019690 dsRNA or control dsRNA (mean ± SD). P < 0.0001. (C) Phenotype of adult S. mansoni following 0.618 μM CH or DMSO treatment (left, n = 15-18), and EdU labeling of proliferative cells in male parasites (right). Representative bright-field images from 3 biological replicates. Scale bars, 1000 μm. For EdU labeling, nuclei were stained with DAPI (blue) and proliferative cells were shown in magenta. Testes are outlined with dashed yellow circles. n = 15 worms from 3 biological replicates, Scale bars, 100 μm. (D) Schematic diagram of sample collection for RNA-seq following RNAi and CH treatment. (E–F) Volcano plots of differentially expressed genes following Smp_019690 RNAi (E) or CH treatment (F). Differentially expressed genes were defined as those with Padj < 0.05 and |Log₂FoldChange| ≥ 1. (G) Venn diagram of the differentially expressed genes between RNAi and CH treatment group.

CH inhibits formation of schistosome U2AF65–RNA complex.(A) The model of U2AF-dependent splicing in model organisms [25]. NLS, nuclear localization signal; KH, K Homology Domain; UHM, U2AF Homology Motif; ULM, UHM-ligand motif; RRM, RNA Recognition Motifs; PPT, the poly-pyrimidine tract. (B-C) Differential alternative splicing events after RNAi (B) or CH treatment (C). AS, alternative splicing; SE, skipped exon; A5SS, alternative 5' splice sites; A3SS, alternative 3' splice sites; MXE, mutually exclusive exons; RI, retained intron. (D) SDS-PAGE analysis of purified recombinant Smp_019690 (SmU2AF65). Lane M: protein molecular weight marker; remaining lanes: purified Smp_019690 protein. (E) Assessment of the binding affinity between SmU2AF65 and FAM-labeled RNA. (F) Assessment of the effect of varying CH concentrations on the binding interaction between SmU2AF65 and RNA. DDS-11, an unrelated compound, served as a control. (G) Quantitative analysis of SmU2AF65–RNA shift band intensity at different CH concentrations. The quantitative analysis data are derived from S11 Fig. The shift bands were analyzed for grayscale intensity using the ImageJ software. *P < 0.05, P < 0.01.

Full text

Introduction

Schistosomiasis, a zoonotic parasitic disease caused by infection with Schistosoma spp., is a significant public health concern. Globally, schistosomiasis has been reported in 78 countries, with 52 classified as endemic [1]. It is estimated that approximately 230 million individuals are infected worldwide, while 700 million people live at risk of infection [2]. As of 2021, at least 251.4 million people required preventive treatment [3]. The World Health Organization ranks schistosomiasis as the second most neglected tropical disease, surpassed only by malaria. Notably, the disease can infect both humans and a wide range of domestic animals, exacerbating its impact on human health and hindering economic development.

Schistosomiasis is closely linked to inadequate living conditions and poor sanitation. A comprehensive approach is emphasized by effective control strategies, including the improvement of sanitation facilities, promotion of health education, access to safe water supplies, and provision of adequate medical care [4]. In addition, the development of vaccines is considered a critical strategy for combating schistosomiasis. However, no effective vaccine has been developed to date. Another alternative approach focuses on the elimination of intermediate host snails. Historically, molluscicides were widely used for snail control prior to the 1970s, but their significant ecological impacts have limited their long-term application [5]. Despite these efforts, each of the aforementioned methods has inherent limitations, necessitating the exploration of alternative strategies. Increasing attention has been directed toward the discovery of anti-schistosomal drugs designed to interrupt the parasite's life cycle and thereby reduce transmission. Currently, praziquantel (PZQ) is the only drug used for large-scale treatment and control of schistosomiasis. However, PZQ exhibits efficacy solely against adult worms and does not prevent reinfection. Moreover, its prolonged and repeated use has been associated with decreased sensitivity of parasites to the drug, leading to reduced therapeutic efficacy and the emergence of drug-resistance [6]. These challenges underscore the urgent need for the development of novel anti-schistosomal drugs to accelerate the eradication of this disease.

Current drug discovery approaches include phenotypic drug discovery (PDD) and target-based drug discovery (TBDD). Both have contributed to the screening of anti-schistosomal compounds. For PDD, it searches for compounds that produce a desired effect on a cellular or organismal level, regardless of the exact molecular target. Abdulla et al. developed a partially automated screen workflow that interfaces schistosomula with microtiter plate-formatted compound libraries [7]. This workflow enabled high-throughput, phenotypic ex vivo screening of 2,160 structurally diverse natural and synthetic compounds. The implementation of this method markedly improved the efficiency and throughput of drug discovery for schistosomiasis. For the TBDD strategy, it focuses on identifying compounds that bind to a specific known molecular target. Following the sequencing and annotation of the major human schistosome genomes, several putative drug targets have been identified and subjected to investigation [8]. Notable examples of these targets include S. mansoni thioredoxin glutathione reductase (SmTGR) [9], polo-like kinase 1 (SmPLK1) and tyrosine kinase [10], histone deacetylases (HDACs) and sirtuins [11,12], cysteine protease cathepsin B1 (SmCB1) [13], serotonin (5-HT) GPCR [14], and S. japonicum 3-oxoacyl-ACP reductase (OAR) [15]. Research efforts targeting these molecules have yielded several promising compounds with the potential to inhibit schistosomal activity effectively. Recently, we performed a large-scale RNA interference (RNAi) screening and identified around 200 potential druggable targets in S. mansoni [16], which provides an essential resource for the TBDD against schistosomes.

The in silico TBDD workflow begins with the identification and validation of drug targets, followed by simulating the binding interactions between candidate compounds and the target. A critical prerequisite for these processes is the determination of the target protein's three-dimensional (3D) structure. However, elucidating protein structures remains one of the most formidable challenges in biology. Despite advances in techniques such as nuclear magnetic resonance (NMR) spectroscopy [17], X-ray crystallography [18], and cryogenic electron microscopy (Cryo-EM) [19], only around 200,000 protein structures have been resolved to date, representing less than 0.1% of all known proteins (https://www.rcsb.org/) [20]. Unlike human proteins, for which thousands of crystal or cryo-EM structures have been resolved, very few three-dimensional structures of Schistosoma proteins are available in public databases. This lack of structural data presents a major bottleneck for in silico TBDD, which relies heavily on accurate protein models to simulate molecular interactions. To overcome this limitation, computational structure prediction has become a valuable alternative. In December 2020, AlphaFold2 (AF2), a machine learning-based protein structure prediction model developed by DeepMind, achieved a groundbreaking milestone by winning the 14th Critical Assessment of Structure Prediction competition. In 2021, DeepMind, in collaboration with the European Molecular Biology Laboratory, launched the AlphaFold Protein Structure Database, providing high-precision 3D protein structure predictions. This publicly accessible database now contains over 200 million predicted protein structures, encompassing nearly all known protein sequences, including the human proteome and the proteomes of many other organisms [20]. Among these predictions, approximately 35% exhibit experimental-level accuracy, while an additional 45% are considered highly reliable [20]. The database has significantly advanced fundamental research, aiding in the understanding of protein functions and disease mechanisms. Furthermore, it has become an indispensable resource for drug discovery, enzyme engineering, and biotechnology, offering unprecedented opportunities for innovation in these fields. Importantly, this includes predicted structural models for proteins in S. mansoni and S. japonicum, offering new possibilities for rational drug design against previously inaccessible targets.

In this study, we selected 202 potential therapeutic targets, primarily based on our previous large-scale RNAi screening on S. mansoni [16], along with seven newly identified genes from recent RNAi experiments on S. mansoni and S. japonicum. Using the highly accurate predicted structures of 37 target protein among these based on AlphaFold Protein Structure Database, we performed virtual docking against MCE active compound library and identified top 200 compound targeting each protein. We then prioritized 268 compounds and evaluated their inhibitory effect on both adult and juvenile worms of S. mansoni and S. japonicum. In total, we identified 7 compounds that showed anti-schistosomal activity at 1 µM. Among them, HY-B2171A (Carubicin hydrochloride, CH), as a compound binding to the target protein splicing factor U2AF65 (Smp_019690) exhibited the most potent efficacy. We further demonstrated that this compound affects the worm activity by inhibiting the function of the splicing factor U2AF65 in regulating the mRNA splicing. This study integrates high-accuracy predicted structures of potential targets with molecular docking, as well as ex vivo phenotypic and molecular assessments of compound-treated schistosomes, thereby providing a robust approach that could significantly accelerate the drug discovery process for schistosomiasis.

Results

Virtual docking for potential druggable targets in schistosomes

To identify potential molecular inhibitors against schistosomes, we selected 202 protein targets essential for parasite viability, as determined by RNAi studies on S. mansoni and S. japonicum. This included 195 genes from our previous study [16] and 7 newly identified genes (4 genes from S. mansoni and 3 genes from S. japonicum) from recent screenings (S1 Fig). As shown in Fig 1, protein structure predictions for these targets were conducted using AlphaFold. After excluding low-quality models with an average per-residue model confidence score (pLDDT) < 70, 37 high-quality protein models were retained for virtual screening (S1 Table). These included 20 enzymes and 6 signal recognition particles. Hydrogen atoms were added to the protein models using Schrödinger Maestro 11.4, followed by energy optimization with the OPLS2005 force field. The Sitemap tool was used to predict the binding sites for each model, with the center of the docking box (20 Å × 20 Å × 20 Å) defined as the best binding site (Site1).

Workflow for the screening of putative anti-schistosomal compounds.This representation summarizes the inputs, outputs and filter criteria used in this study, leading to the selection of 268 compounds for ex vivo screening (see methodology for full details).

To identify potential active compounds targeting these proteins, we selected the MedChemExpress (MCE) Bioactive Compound Library Plus, which contains 14,600 compounds, for virtual docking (Fig 1). After adding hydrogen atoms and performing energy optimization on the 2D structures of the compounds (using the same procedure as for the protein models), 3D structures were generated. High-Throughput Virtual Screening (HTVS) was performed to identify primary compound candidates from the MCE library, with the top 10% subjected to Standard Precision (SP) docking. Subsequently, the top 10% of compounds from the SP results underwent a final round of docking using Extra Precision (XP) mode. The top 200 compounds with the highest docking scores for each target protein were selected (S2 Table). To prioritize the compounds, we excluded those with a molecular weight over 700 or endogenous compounds, and selected the top 10–15 compounds with the highest docking scores for each protein target. In total, 268 compounds targeting 37 potential druggable proteins in S. mansoni and S. japonicum were identified from the MCE Bioactive Compound Library Plus (S3 Table).

Ex vivo screening of effective compounds against adult S. mansoni and S. japonicum

During the virtual screening process, we observed a high degree of similarity between the target protein sequence of S. mansoni and its homolog in S. japonicum (S4 Table). Furthermore, protein structure predictions revealed a remarkable structural resemblance of the target protein in both species (S4 Table and S2 Fig). These findings suggest that the selected compounds may target the protein in both species, potentially inhibiting worm activity in S. mansoni and S. japonicum. Based on this, we conducted an extensive screening of compounds targeting both species.

To assess the effects of compounds on adult schistosomes, the viability of female and male worms was evaluated over seven days of culture in medium treated with 10 µM of the compounds. Key indicators of worm abnormalities included reduced attachment ability, swelling, shrinkage, darkened coloration, and abnormal motility. The activity of the cultured parasites was recorded daily, and the compound that caused 100% abnormality in worms on the final day of cultivation was considered effective, with PZQ used as a control.

For S. mansoni adult worms, a total of 26 effective compounds were identified (S5 Table), of which 13 exhibited pronounced effects, leading to worm mortality or a moribund state (Fig 2A). These compounds targeted seven distinct protein targets, with phosphatidylserine decarboxylase proenzyme 1 (Smp_021830.3) being the most frequently targeted, followed by chitobiosyldiphosphodolichol alpha-mannosyltransferase (Smp_005010). Phenotypic analysis revealed significant activity for eight compounds (HY-100493, HY-15196, HY-112852, HY-12462, HY-12382, HY-B2171A, HY-12844, and HY-18018) after a single treatment, as observed on day 2 (D2) (S6 Table). Worms treated with these compounds exhibited key phenotypic changes, including dissociation of male-female pairs, failure to adhere to the bottom of the culture plates, reduced motility, and, in some cases, death (S6 Table). By D7, all highly effective compounds caused complete loss of adherence to the culture plate bottom, severely impaired viability, and states of death or near-death, characterized by intermittent, minimal movement restricted to the head or tail regions (S6 Table and Fig 2A). Enlarged phenotypic images of worms treated with lethal compounds highlighted significant differences between the treatment and control groups (S3A Fig). In the control group, worms exhibited well-extended bodies and strong adherence to the substrate. In contrast, worms treated with either the effective compounds or PZQ displayed darkened coloration, shortened bodies, and, in some instances, curling into a circular posture. Additionally, worms in some treatment groups showed varying degrees of epidermal erosion, with blurred surface structures indicative of dissolution-like damage.

Phenotypes of adult S. mansoni and S. japonicum treated with high-efficiency compounds.The figure depicts the phenotypic changes in S. mansoni (A) and S. japonicum (B) after a 7-day co-incubation with high-efficiency compounds at a concentration of 10 μM. Negative control group: DMSO (0.1% DMSO, at the same DMSO concentration as in the experimental group.); Positive control group: Praziquantel (1 µM in 0.1% DMSO). Pink box denotes species-specific high-activity. Representative images from 3 biological replicates. Scale bars, A, 1000 μm; B, 2000 μm.

To further evaluate their efficacy, these active compounds were subjected to secondary screening at a lower concentration (1 µM), resulting in the identification of two compounds with significant effects (Tables 1 and S7).

Table 1: Compounds exhibiting potent inhibitory activity against S. mansoni and S. japonicum at 1 µM.

Similarly, an initial screening at 10 µM was conducted on S. japonicum adult worms, resulting in the identification of 23 effective compounds, 11 of which demonstrated high efficacy (S5 Table and Fig 2B). The most frequently targeted proteins were consistent with those identified in S. mansoni, namely phosphatidylserine decarboxylase proenzyme 1 (Smp_021830.3) and chitobiosyldiphosphodolichol alpha-mannosyltransferase (Smp_005010). The detailed results of phenotypic recordings were presented in S8 Table. High-resolution images revealed the damages on the parasites (S3B Fig). Unlike the control group, worms in the PZQ group were dead, displaying darker coloration and a curled, ring-like posture. Additionally, worms in the treatment groups exhibited near-death characteristics, including darkened epidermal coloration, varying degrees of erosion and damage, and swelling. At a lower concentration of 1 µM, one compound (HY-B2171A) exhibited a significant effect (Tables 1 and S7), with its predicted target identified as splicing factor U2AF subunit (EWB00_008591).

Ex vivo screening of effective compounds against juvenile S. mansoni and S. japonicum

Next, ex vivo screening of compounds against juvenile worms was also performed. Compounds were deemed effective when all worms exhibited loss of attachment and reduced mobility. Those causing 100% mortality or moribundity were considered strongly active. Phenotypes for effective compounds were recorded in S9 and S10 Tables.

For juvenile S. mansoni, 29 compounds were identified with activity, 14 of which exhibited pronounced potency (S5 Table), inducing 100% mortality or moribund states (Fig 3A). These most active compounds targeted seven distinct protein targets (S5 Table). Subsequent screening at a reduced concentration of 1 µM revealed five compounds with significant effects (Tables 1 and S7). These five compounds targeted four distinct proteins, among these, two compounds—HY-12382, and HY-B2171A—were identical to those effective against adult worms (Table 1).

Phenotypes of juvenile S. mansoni and S. japonicum treated with high-efficiency compounds.Bright-field images showcasing the phenotypes of juvenile worms after 7-day co-incubation with potent compounds (10 μM). (A) S. mansoni (14 compounds); (B) S. japonicum (16 compounds). 0.1% DMSO as negative control; 1 µM Praziquantel in 0.1% DMSO as positive control. Pink box denotes species-specific high-activity. Representative images from 3 biological replicates. Scale bars, 500 μm.

For S. japonicum juvenile, 29 effective compounds were identified, of which 16 demonstrated high efficacy (S5 Table and Fig 3B). These strong inhibitors targeted seven distinct proteins, with one target differing from those identified in S. mansoni. This unique target (Smp_157160), corresponding to compound HY-16322, exhibited higher efficacy in S. japonicum than in S. mansoni. When the screening concentration was reduced to 1 µM, only six compounds remained effective (Tables 1 and S7). The corresponding targets were Smp_021830.3, Smp_005010, Smp_157160, and EWB00_008591. Among these, Smp_157160 was unique to S. japonicum, while the other three targets were consistent with those identified in S. mansoni.

Compounds demonstrating strong efficacy against both adult and juvenile schistosomes in two species

As previously described, at a concentration of 1 µM, seven compounds showed high anti-schistosomal activity, whose docking efficiencies with their respective targets were displayed in Fig 4A. In detail, two compounds demonstrated activity against S. mansoni adult worms (Fig 4B), whereas only one compound exhibited efficacy against S. japonicum adult worms (Fig 4C). For juvenile worms, four compounds were effective against both species; however, compound HY-12382 showed activity exclusively against S. mansoni juvenile (Fig 4D), while HY-112852 and HY-16322 were specifically effective against S. japonicum juvenile (Fig 4E). Notably, compound HY-B2171A (CH) exhibited broad-spectrum efficacy across all four types of schistosomes, demonstrating the most significant activity.

Molecular Docking of effective compounds with their targets and effects on schistosomes.(A) Molecular docking diagrams of active compounds with the target. The C backbone of the target protein is depicted in green, nitrogen atoms (N) in blue, oxygen atoms (O) in red, and hydrogen atoms (H) in white. The compounds are shown as sticks with the following color codes: dark blue for HY-12462 and HY-16322, magenta for HY-112852, light green for HY-15004, and light blue for HY-15196, HY-12382, and HY-B2171A. π-π interactions are indicated by green dashed lines, and hydrogen bonds are shown as purple dashed lines, with longer hydrogen bond lengths reflecting weaker interactions. Light blue dashed box: molecular docking of compound CH with binding site of SjU2AF65 (EWB00_008591) and SmU2AF65 protein (Smp_019690). (B–E) Bright-field images showing representative compound-induced phenotypes following 7-day co-incubation with 1 μM compounds. (B) Adult S. mansoni; (C) Adult S. japonicum; (D) Juvenile S. mansoni; (E) Juvenile S. japonicum. Scale bars: B, 1000 μm; C, 2000 μm; D–E, 500 μm. Representative images from 3 biological replicates.

CH is a potent anti-schistosomal compound

Based on previous results, nine compounds demonstrating significant efficacy against both adult and juvenile schistosomes were identified for S. mansoni and S. japonicum (Table 2). In addition, two compounds showed species-specific activity—one effective against S. mansoni, and the other against S. japonicum (Table 2). To assess their relative anti-schistosomal potency, IC50 values were calculated based on gradient concentrations and corresponding phenotype scores (Table 2 and S4 and S5 Figs). Among these, HY-B2171A (CH), predicted to target the splicing factor U2AF through virtual docking, exhibited the highest efficacy in both assays for S. mansoni (0.1422 μM) and S. japonicum (0.566 μM). Notably, IC50 analysis revealed that most compounds exhibited comparable activity against the two Schistosoma species. However, a few demonstrated pronounced species-specific effects. For example, HY-100493 showed activity exclusively against S. mansoni, whereas HY-16322 exhibited efficacy restricted to S. japonicum.

Table 2: Cytotoxicity (CC50) in HepG2 cells and IC50 values against S. mansoni and S. japonicum for high-efficacy compounds.

To evaluate compound safety, cytotoxicity of 11 potent hits was measured in HepG2 cells using the CCK-8 assay. Dose-response curves fitted well (R² > 0.8), enabling CC50 calculation (Table 2 and S6 Fig). Most compounds showed moderate to low toxicity, while HY-16322 maintained > 50% cell viability at 50 μM, indicating minimal toxicity. Selectivity indices (SI), defined as the ratio of CC50 to IC50 against S. mansoni and S. japonicum, were calculated and summarized in Table 2. Except for HY-12462, all compounds had CC50 values exceeding their respective IC50s. Most compounds exhibited SI values greater than 3 for both species, demonstrating favorable selectivity toward schistosomes over mammalian cells. These results support further optimization of these hits for anti-schistosomal drug development with improved safety profiles.

Characterization of the coding gene for CH's potential target

CH was predicted to target the protein encoded by EWB00_008591 in S. japonicum, annotated as the splicing factor U2AF subunit (U2 small nuclear RNA auxiliary factor) (Fig 4A). Notably, this protein shares high sequence and structural similarity with Smp_019690 in S. mansoni, which is also annotated as a U2AF subunit (S4 Table and S2 Fig). Interestingly, CH also exhibited potent activity against S. mansoni (Tables 1 and 2). Given the high degree of similarity between the SjU2AF and SmU2AF, we further performed molecular docking between CH with Smp_019690, which resulted in similarly high docking scores (Boxed in Fig 4A). These findings suggest that CH may exert its anti-schistosomal activity by targeting the conserved U2AF subunit in both species. Based on this hypothesis, we selected Smp_019690 as a candidate target for further functional and mechanistic studies in S. mansoni.

U2AF is a critical auxiliary component in RNA splicing, functioning as a heterodimer consisting of two subunits: U2AF65 and U2AF35 [21]. To classify the S. mansoni splicing factor U2AF subunit, we conducted a phylogenetic analysis, identifying it as part of the U2AF65 family (S7 Fig). Protein sequence alignments with homologs from other schistosome species and model organisms, including humans, mice, Danio rerio, and Caenorhabditis elegans, revealed conserved regions across these species (S8 Fig). These findings highlight evolutionary similarities and suggest the presence of functionally conserved domains in U2AF65.

As indicated by the single-cell atlas of S. mansoni (https://www.collinslab.org/schistocyte/), Smp_019690 is predicted to be expressed across all cell types in both sexes (S9A and 9B Fig). Whole-mount in situ hybridization and fluorescent in situ hybridization further confirmed this, revealing higher expression levels in the sexual organs of the parasites, such as the testes and ovaries (S9C and 9D Fig).

CH disrupts exon splicing in schistosomes by targeting U2AF65

To test whether CH inhibits schistosomal activity by targeting SmU2AF65 as predicted (Fig 4A), we first evaluated the performance of parasites subjected to Smp_019690 RNAi or CH treatment. As expected, silencing the expression of Smp_019690 via ex vivo cultivation with dsRNA resulted in all worms detaching from the plate bottom and exhibiting significantly reduced motility by day 10 (Fig 5A). qPCR analysis showed that the expression level of Smp_019690 was reduced by 66% (Fig 5B). When worms were treated with a low concentration (0.618 μM) of CH, all exhibited signs of morbidity within one day (Fig 5C). Given that U2AF65 has been previously reported to play a role in cell proliferation [22,23], we further assessed the effect of RNAi or CH treatment on parasite cell proliferation using EdU labeling. Both treatments resulted in a significant decrease in proliferative cells in the soma and testes of male parasites (Fig 5A and 5C). To better understand the molecular mechanisms underlying these phenotypic changes, RNA-seq was performed to profile transcriptomic changes in both Smp_019690-knockdown adult male worms and CH-treated adult male worms (Fig 5D). Dramatic transcriptional changes occurred after both treatments (S10 Fig). Using selection criteria of Padj < 0.05 and |Log_2_FoldChange| ≥ 1, we identified 556 downregulated and 414 upregulated genes in the Smp_019690 RNAi group compared to the control group (Fig 5E and S11 Table). In the CH-treated group compared to the DMSO group, 268 downregulated and 223 upregulated genes were identified (Fig 5F and S12 Table). A comparison of differentially expressed genes (DEGs) between the two groups revealed 145 shared genes (Fig 5G and S13 Table).

*Stem cell maintenance and transcriptomic changes in parasites following Smp_019690 knockdown or CH treatment.(A) Phenotype of adult S. mansoni following Smp_019690 or control RNAi (left, n = 15-18), and EdU labeling of proliferative cells in male parasites (right). Representative bright-field images from 3 biological replicates. Scale bars, 1000 μm. For EdU labeling, nuclei were stained with DAPI (blue) and proliferative cells were shown in magenta. Testes are outlined with dashed yellow circles. n = 15 worms from 3 biological replicates, Scale bars, 100 μm. (B) Relative mRNA expression levels of Smp_019690 in parasites treated with Smp_019690 dsRNA or control dsRNA (mean ± SD). ***P < 0.0001. (C) Phenotype of adult S. mansoni following 0.618 μM CH or DMSO treatment (left, n = 15-18), and EdU labeling of proliferative cells in male parasites (right). Representative bright-field images from 3 biological replicates. Scale bars, 1000 μm. For EdU labeling, nuclei were stained with DAPI (blue) and proliferative cells were shown in magenta. Testes are outlined with dashed yellow circles. n = 15 worms from 3 biological replicates, Scale bars, 100 μm. (D) Schematic diagram of sample collection for RNA-seq following RNAi and CH treatment. (E–F) Volcano plots of differentially expressed genes following Smp_019690 RNAi (E) or CH treatment (F). Differentially expressed genes were defined as those with Padj < 0.05 and |Log₂FoldChange| ≥ 1. (G) Venn diagram of the differentially expressed genes between RNAi and CH treatment group.

In humans and plants, U2AF plays a key role in mediating alternative splicing [24–26]. A schematic representation of the working model for U2AF, based on these studies, is shown in Fig 6A. Specifically, U2AF35 recognizes and binds to the AG dinucleotide at the 3' splice site, while U2AF65, which contains two RNA recognition motifs (RRM1/RRM2), interacts with the poly-pyrimidine tract (PPT) at the 3' splice site [25]. Additionally, the C-terminal UHM (U2AF homology motif) of U2AF65 associates with the ULM (UHM-ligand motif) domain of the auxiliary protein SF1 [24], stabilizing the early spliceosome complex. Together, U2AF35 and U2AF65 facilitate early spliceosome assembly, regulate exon inclusion or skipping, and influence the dynamic selection of splice sites [26]. Given the high sequence similarity between SmU2AF65 and its human homologs, we hypothesized that SmU2AF65 may also be involved in mRNA splicing in S. mansoni. To test this, we analyzed alternative splicing events using RNA-seq data from Smp_019690 RNAi and CH treatment experiments (Fig 6B and 6C). Five types of alternative splicing events were assessed: skipped exon (SE), mutually exclusive exons (MXE), alternative 3' splice sites (A3SS), retained intron (RI), and alternative 5' splice sites (A5SS). Knocking down Smp_019690 resulted in changes across all splicing event types, with SE events showing the greatest number of changes: 2,419 reduced and 201 increased events (Fig 6B). These findings suggest that SmU2AF65 promotes SE events in S. mansoni, consistent with U2AF65's role in other organisms (Fig 6A) [27]. Similarly, changes in splicing events were observed after treating worms with CH, a predicted binding compound for SmU2AF65 (Fig 6C). Although the total number of altered splicing events was lower than in the RNAi group, SE events predominated. Notably, CH treatment resulted in an equal number of reduced and increased SE events, contrasting with the RNAi results where SE events mainly decreased.

*CH inhibits formation of schistosome U2AF65–RNA complex.(A) The model of U2AF-dependent splicing in model organisms [25]. NLS, nuclear localization signal; KH, K Homology Domain; UHM, U2AF Homology Motif; ULM, UHM-ligand motif; RRM, RNA Recognition Motifs; PPT, the poly-pyrimidine tract. (B-C) Differential alternative splicing events after RNAi (B) or CH treatment (C). AS, alternative splicing; SE, skipped exon; A5SS, alternative 5' splice sites; A3SS, alternative 3' splice sites; MXE, mutually exclusive exons; RI, retained intron. (D) SDS-PAGE analysis of purified recombinant Smp_019690 (SmU2AF65). Lane M: protein molecular weight marker; remaining lanes: purified Smp_019690 protein. (E) Assessment of the binding affinity between SmU2AF65 and FAM-labeled RNA. (F) Assessment of the effect of varying CH concentrations on the binding interaction between SmU2AF65 and RNA. DDS-11, an unrelated compound, served as a control. (G) Quantitative analysis of SmU2AF65–RNA shift band intensity at different CH concentrations. The quantitative analysis data are derived from S11 Fig. The shift bands were analyzed for grayscale intensity using the ImageJ software. *P < 0.05, *P < 0.01.

To explore the function of SmU2AF65 in the parasite and its interaction with compound CH, we cloned and expressed the full-length SmU2AF65 protein in E. coli (Fig 6D). The purified protein was then used for further analysis. As described previously [21,26], U2AF65 plays a critical role in pre-mRNA splicing by recognizing and binding to the polypyrimidine tract near the 3' splice site of introns, facilitates the assembly of the early spliceosome and regulates exon inclusion or skipping. Using rMATS, we identified the introns of the top 10 differentially skipped exons from our RNA-seq data of CH treatment experiments. After extracting and aligning these sequences with MAFFT, we identified a consensus 50 bp sequence (5'-UACAUAAAUUAAUAAGUAUUUUUACUUUUUUUUAUUUUCAAUUUCAACAG-3') in the latter region, assumed to be the RNA fragment recognized by SmU2AF65. This 50 bp RNA fragment was shown to bind to the purified SmU2AF65, forming a shift band on native polyacrylamide gel, as detected by EMSA (Fig 6E). To determine whether CH disrupts the binding of U2AF65 to its target RNA, we conducted binding assays with SmU2AF65, the RNA fragment, and varying concentrations of CH. The shift bands weakened as the CH concentration increased (Figs 6F, 6G, and S11), whereas the unrelated compound DDS-11 had no effect. These findings suggest that CH targets U2AF65, disrupting the U2AF65–RNA complex and affecting RNA splicing in schistosomes.

To further confirm that CH acts through specific residues within SmU2AF65, we performed mutagenesis analysis guided by our docking model. The predicted CH-binding interface (residues 271–329) includes four key residues—LEU255, ARG270, GLN284, and VAL289—that form hydrogen bonds with CH (S12A Fig). Two mutants were generated, expressed, and purified: Mutant 1 (L255A/R270A/Q284A/V289A) and Mutant 2 (L255A/R270E/Q284R/V289A) (S12A and S12B Fig). EMSA assays revealed that Mutant 1 retained RNA-binding activity, whereas Mutant 2 completely lost this activity (S12C Fig). Notably, CH inhibited the RNA-binding activity of wild-type SmU2AF65 at 10 μM (Fig 6G), whereas inhibition of Mutant 1 was only evident at 30 μM (S12D and S12E Fig). These findings demonstrate that the identified residues are critical for CH binding and that disruption of this interface attenuates CH-mediated inhibition of SmU2AF65–RNA interactions. Collectively, these results provide strong biochemical evidence identifying SmU2AF65 as the molecular target of CH in schistosomes.

In conclusion, we established a workflow combining highly accurate protein structure modeling and virtual docking to identify compounds binding to target proteins in schistosomes. Through ex vivo screening, we prioritized compounds with high efficacy against adult and juvenile worms of S. mansoni and S. japonicum. We demonstrated that U2AF65 is a functional target, with CH identified as the most potent compound. CH treatment or u2af65 knockdown resulted in significant changes in exon-skipping events, with CH inhibiting the binding of U2AF65 to its target RNA. This study provides a robust workflow for high-throughput screening of anti-schistosomal compounds and highlights a novel protein-compound interaction, accelerating the development of anti-schistosomal drugs.

Discussion

Research on anti-schistosomal drugs dates back to 1918 with the discovery of potassium antimonyl tartrate [28]. Since then, continuous efforts have been made to develop effective treatments. A major breakthrough in schistosomiasis therapy was achieved in the mid-to-late 1970s with the discovery of praziquantel [29]. Despite ongoing research to develop new therapeutics, praziquantel remains the sole drug available for the treatment of this disease. In this study, we selected 202 target proteins essential for the viability of S. mansoni and S. japonicum from our previous RNAi screening. Among these, we prioritized 37 proteins with highly accurate predicted structures for virtual docking, leading to the identification of 268 compounds with high binding potential. Through ex vivo cultivation and evaluation of parasite activity, we further identified 7 distinct compounds that exhibit significant inhibitory effects at 1 μM on S. mansoni and S. japonicum. Among them, HY-B2171A, which targets schistosome U2AF65, demonstrated the most robust efficacy against both species. This workflow provides a promising approach to accelerate the development of alternative drugs to praziquantel.

During the ex vivo evaluation of the anti-parasite efficacy, we observed distinct susceptibilities between adult and juvenile worms to the same compound. Specifically, compounds HY-12844, HY-112647 and HY-N0794 demonstrated higher efficacy against S. mansoni adults than juveniles, whereas compounds HY-138696, HY-15002, HY-B0119 and HY-11101 were more effective against juveniles than adults (S5 Table). A similar pattern was observed in S. japonicum. This differential susceptibility between developmental stages was further exemplified by the anti-schistosomal effects of PZQ, where stage-specific efficacy variations were particularly evident. Previous studies have reported that juvenile S. mansoni (3–4 weeks post-infection) and S. japonicum (2 weeks post-infection) are not susceptible to PZQ in vivo [30–32], preventing the treatment of recent infections [33]. However, the mechanisms underlying this reduced susceptibility remain unclear. Investigations into oxamniquine derivatives further highlighted that two novel compounds exhibited weaker activity against juveniles compared to adults [34]. Similarly, compounds such as artemether and curcumin have demonstrated preferential efficacy against juveniles [35]. Collectively, these findings indicate that differential compound susceptibility arises from physiological and structural divergences between juvenile and adult worms. This stage-specific susceptibility is not restricted to individual compounds but represents a critical factor that must be systematically considered in anti-schistosomal drug discovery.

Furthermore, our work revealed marked interspecies variations in compound efficacy against different Schistosoma species. Notably, HY-100493 displayed exclusive activity against S. mansoni, while HY-16322 showed species-specific efficacy restricted to S. japonicum. This pharmacological divergence aligns with previous findings. A comprehensive review by Abdul-Ghani et al. highlighted oxamniquine's exclusive activity against S. mansoni [36]. Similarly, prior studies have demonstrated that CDDD-0149830 exhibits greater efficacy against S. haematobium, CIDD-150303 shows superior activity against S. mansoni, and CIDD-066790 is more effective against S. japonicum [34]. The observed differences in compound efficacy across Schistosoma species may stem from variations in parasite biology. Further investigation is warranted to elucidate the precise mechanisms underlying these interspecies pharmacological discrepancies.

Structure-based drug discovery plays a crucial role in identifying anti-schistosomal compounds by leveraging the 3D structures of target proteins to facilitate precise compound screening, improve hit rates, and reduce the workload and cost of experimental screening. In this study, we employed AlphaFold to predict the structures of target proteins, followed by molecular docking to assess interactions between the target proteins and small-molecule compounds, aiming to identify potential anti-schistosomal candidates. A key strength of this approach is that AlphaFold generates high-confidence 3D structure predictions for proteins lacking experimentally resolved structures, thereby expanding the range of schistosomal targets available for structure-based drug screening. However, this method has inherent limitations. First, AlphaFold predictions rely on sequence homology and physicochemical properties, which may not fully capture dynamic conformational changes, particularly the plasticity of ligand-binding pockets, potentially affecting virtual screening accuracy. Second, when a target protein is highly conserved between the host and schistosomes, selected compounds may pose a risk of host toxicity. To mitigate this, an additional filtering step that prioritizes targets with parasite-specific structural or sequence features should be incorporated. Third, because highly conserved proteins tend to yield higher AlphaFold prediction scores, obtaining experimentally resolved structures (e.g., via X-ray crystallography or cryo-EM) for schistosome-specific targets, in combination with our pipeline, would further strengthen the drug discovery process. Nonetheless, protein expression and structural determination remain significant challenges. Thus, in practical applications, SBDD based on AlphaFold-predicted structures should be integrated with target conservation analysis, molecular dynamics simulations, and experimental validation to enhance screening efficiency and improve the success rate of identifying selective anti-schistosomal drugs.

In this study, we found compound CH displayed a strongly inhibitory effect on schistosome viability ex vivo. Virtual docking analysis predicted its potential target as U2AF65, a key protein involved in pre-mRNA splicing [26]. Transcriptomic analysis revealed significant differences between samples treated with CH and those subjected to RNAi of the target gene (Fig 5E and 5F). We hypothesize that this discrepancy arises from the fundamental differences in their mechanisms of action: CH directly targets the protein, whereas RNAi affects gene expression. Specifically, RNAi reduces U2AF65 protein levels by degrading its mRNA, gradually impairing RNA splicing. In contrast, CH directly targets the U2AF65 protein to disrupt its function. Alternative splicing analysis further confirmed that CH affects splicing events in a manner similar to, but not entirely identical to, *u2af65 *knockdown, as both treatments significantly influenced exon skipping (Fig 6B and 6C). To determine whether these splicing changes are pathway-specific or reflect general effects of worm death, we compared CH-treated worms with PZQ-treated controls. Interestingly, PZQ also induced abnormal SE events (S13 Fig), indicating that certain transcriptomic alterations may result from general drug effects, worm death, or direct perturbation of the splicing machinery. This comparison highlights the complexity of interpreting RNA-seq data in the context of drug target validation. Previous studies have shown that U2AF65 plays a crucial role in pre-mRNA splicing by recognizing and binding to the polypyrimidine tract near the 3' splice site of introns, facilitating early spliceosome assembly and regulating exon inclusion or skipping [21,26]. Our findings support this established mechanism, as ex vivo evaluation confirmed that U2AF65 binds to a polypyrimidine RNA probe.

CH, a member of the Bioactive Compound Library, is a microbially derived compound previously reported to target the Golgi complex, inducing apoptosis in clear cell renal carcinoma cells [37,38]. It also inhibits the template activity of DNA in the DNA-dependent RNA polymerase system by complexing with DNA [39] and acts as a potent inhibitor of VHL-deficient (VHL ⁻ /⁻) CCRCC cell proliferation [38]. In this study, we found that both CH treatment and u2af65 knockdown significantly suppressed parasite cell proliferation (Fig 5A and 5C), consistent with previous findings [22,23,38]. Our results demonstrate that CH interferes with RNA splicing in schistosomes by inhibiting U2AF65's RNA-binding ability, leading to stem cell death and ultimately resulting in parasite death. While U2AF65 inhibition alone is sufficient to induce parasite lethality, CH may also interact with additional targets in schistosomes. Notably, CH inhibited U2AF65–RNA binding in EMSA only at tens of micromolar concentrations, yet showed low-nanomolar IC50 values in whole-worm assays. This difference likely reflects the simplified nature of EMSA, which does not fully replicate the intracellular environment, and the cumulative downstream effects of splicing disruption in vivo. Possible engagement of additional parasite targets may further enhance potency. Despite CH's promising anti-schistosomal activity, the high sequence similarity between schistosome U2AF65 and its human homolog raises concerns about potential host toxicity. Therefore, rational structural modifications of CH are recommended to enhance its binding selectivity toward the parasite protein over the human counterpart, thereby improving its safety and therapeutic index.

Although CH ranked 19th in docking to schistosome U2AF65 (S2 Table), it exhibited the strongest anti-schistosomal activity (Table 2). In contrast, several higher-ranked compounds with superior docking scores, such as HY-N0768 and HY-13624A, failed to show comparable ex vivo anti-parasitic effects or to disrupt U2AF65–RNA binding (S14 Fig). This discrepancy highlights that docking scores alone are not reliable predictors of bioactivity or target engagement. Contributing factors may include protein conformational flexibility, solvation effects, and parasite-specific differences in compound uptake or permeability. Accordingly, experimental validation remains indispensable for accurately identifying functional inhibitors in virtual screening pipelines.

Given the potent ex vivo activity of CH against two schistosome species, we conducted preliminary in vivo evaluation in a murine model of S. mansoni infection to assess its therapeutic potential. Based on LD50 data from the PubChem database (Carubicin Hydrochloride | C26H28ClNO10 | CID 11570269 - PubChem), CH was administered orally at 720 µg/kg (1/10 LD50). Under this regimen, no significant anti-schistosomal effect was observed, although no signs of host toxicity were detected (S15 Fig). The disparity between ex vivo potency and in vivo efficacy is likely attributable to poor pharmacokinetics, limited parasite exposure, or suboptimal bioavailability. These findings suggest that substantial chemical optimization will be required to advance CH as a therapeutic candidate.

In summary, the integrated strategy employed in this study—combining AlphaFold-based structural prediction, virtual screening, and functional validation—holds significant promise for application to other parasitic diseases, such as malaria or leishmaniasis, where high-resolution structural data for protein targets remain scarce. This approach may accelerate target prioritization and compound discovery across a broader range of neglected tropical diseases.

Materials and methods

Ethics statement

All experiments involving vertebrate animals were conducted in compliance with protocols approved by the Institutional Animal Care and Use Committee (IACUC) of Fudan University (approval APN: 2021JS0078).

Parasites

Adult S. mansoni (6–7 weeks post-infection) and juvenile stage worms (4 weeks post-infection) were harvested from infected female Kunming (KM) mice by hepatic portal vein perfusion using ice-chilled 0.9% (w/v) NaCl containing 0.005% (w/v) heparin. For S. japonicum, adult worms (4 weeks post-infection) and juvenile stage worms (2 weeks post-infection) were used. Parasites were rinsed in DMEM containing 10% horse serum, then cultured in DMEM (high glucose) supplemented with 10% fetal bovine serum and 1 × antibiotic-antimycotic. Juvenile worms were cultured in Basch's medium 169 [40]. The host mice utilized in the experiment were 4-week-old female Kunming mice (KM), weighing between 18 and 22 g. The mice were maintained under standard laboratory conditions, with the temperature controlled at 22 ± 2°C, relative humidity maintained at 50–60%, and a 12-hour light/dark cycle.

Virtual screening workflow for protein-ligand docking

The 3D structures of proteins were retrieved from the AlphaFold database (https://www.alphafold.ebi.ac.uk/). After filtering out low-quality models with an average per-residue model confidence score (pLDDT) < 70, high-quality protein models were retained for further analysis. Using Schrödinger Maestro 11.4, hydrogen atoms were added to the protein models, followed by energy optimization using the OPLS2005 force field (RMSD = 0.30 Å). The binding sites of the protein models were predicted using the Sitemap tool, and the best site (Site1) was selected as the center of a docking box (20 Å × 20 Å × 20 Å).

To identify potential active compounds targeting these proteins, the MCE Bioactive Compound Library Plus (14,600 compounds) was used as the source for candidate compounds. The 2D structures of the compounds were processed by adding hydrogen atoms and performing energy optimization, following the same procedure as for the protein models, to generate 3D structures for docking.

Ligand docking was performed using the compound library against each protein model in three stages: first, High-Throughput Virtual Screening (HTVS) was conducted to identify primary compound candidates; next, the top 10% of compounds from the HTVS results, based on docking scores, were subjected to Standard Precision (SP) docking; finally, the top 10% of compounds from the SP results underwent a final round of docking using Extra Precision (XP) mode. After these steps, the top 200 compounds with the highest docking scores were selected as candidates for further analysis.

Sequence and structural comparison of target proteins

The protein sequence of S. mansoni was obtained from the Sm.V7 genome assembly, while the sequence of S. japonicum was retrieved from the PRJNA520774–HuSjv2 dataset. Sequence alignment of target proteins from S. mansoni and S. japonicum was conducted using the WormBase ParaSite database to determine sequence similarity. For structural comparison, predicted protein structures were sourced from the AlphaFold Protein Structure Database. Structural alignments of homologous proteins from S. mansoni and S. japonicum were performed using PyMOL, and root mean square deviation (RMSD) values were calculated to quantitatively assess structural similarity.

Evaluation of compound effects on parasites

Ex vivo evaluation of the selected compounds (at a single concentration of 10 μM in 0.1% DMSO) was conducted on adult worms (4–6 pairs in 12-well plates with 3 mL medium, or 2 pairs in 24-well plates with 1 mL medium) and juvenile worms (10–16 individuals in 48-well plates with 500 μL medium). Worms were cultured for 7 days at 37°C in a humidified atmosphere with 5% CO_2_, with the medium replaced every 2 days. Phenotypes were recorded daily under a microscope. Control groups included worms treated with 0.1% DMSO (negative control) and 1 μM PZQ (positive control) in DMSO. At day 7, microscopy images were captured using a digital camera system (OLYMPUS, SZX2-ILLTS).

To quantify adult worm attachment to the substrate following compound treatment, freshly perfused adult worms were placed in 24-well plates with 1 mL media and cultured overnight. On the following day (D1), unattached worms were removed, and compounds were added to the media. The media and compounds were replaced on D1, D3, and D5. Dose-response titration (30 μM–0.010 μM) was performed for effective compounds to assess their anti-schistosomal potency. Eight titration points with two-fold or three-fold stepwise dilutions from the top concentration were prepared, and each titration point was performed in triplicate (4 male worms per replicate). Short videos (5 seconds) were recorded under a microscope at day 7 to facilitate scoring. Compound effects were evaluated using 3 indicators [41]: 1) attachment: 1 = attached, 0 = unattached; 2) motility: 0 = static, 1 = slight drifting, 2 = intermittent swinging, 3 = energetic swinging; 3) morphological changes: 0 = tegumental disruption or dark worms, 1 = normal appearance. Worm viability was scored based on the sum of these three indicators, with a scale ranging from 0 to 5. Data on phenotypes and viability scores from the dose-response titration were analyzed using GraphPad Prism to generate approximate IC50 values for each effective compound.

Evaluation of CH against S. mansoni in a mouse model

Female BALB/c mice (6 weeks old, weighing 18–22 g) were percutaneously infected at the tail with 100 S. mansoni cercariae per mouse. Drug treatment was initiated at 4 and 6 weeks post-infection, with daily oral administration for five consecutive days. CH was dissolved in a vehicle consisting of 10% DMSO, 40% PEG300, 5% Tween-80, and 45% saline, and administered orally at a dose of 720 μg/kg. Control mice received an equivalent volume of vehicle. At 7 weeks post-infection, mice were euthanized, and worm burdens were quantified. Gross pathological assessment of liver and spleen were subsequently performed.

Cytotoxicity assay

HepG2 human liver carcinoma cells were maintained in DMEM supplemented with 10% fetal bovine serum and 1 × antibiotics at 37°C in a humidified atmosphere containing 5% CO_2_. Cells were seeded into 96-well plates at a density of 5 × 10^3^ cells per well and incubated overnight to allow adherence. Test compounds were serially diluted in culture medium with a final DMSO concentration not exceeding 0.5%, and applied to cells at various concentrations in triplicate or quadruplicate wells. Blank controls (medium only, no cells) and negative controls (cells with medium and 0.5% DMSO) were included. After 24 hours of compound exposure, cell viability was measured using the Cell Counting Kit-8 (CCK-8) assay according to the manufacturer's protocol. Absorbance was recorded at 450 nm using a microplate reader. The percentage of viable cells was calculated relative to the negative control. CC50 values, representing the compound concentration causing 50% cytotoxicity, were calculated by nonlinear regression analysis using GraphPad Prism software.

Parasite staining and imaging

Whole-mount in situ hybridization (WISH), fluorescent in situ hybridization (FISH) [42], DAPI staining, and EdU labeling [43] were performed as previously described. Worms stained with WISH were imaged using light microscopy (Axio Zoom.V16, ZEISS), while fluorescent images were captured with confocal scanning microscopy (A1, Nikon). Riboprobe synthesis for WISH/FISH: Primers were designed using the Primer-BLAST tool (https://www.ncbi.nlm.nih.gov/tools/primer-blast/index.cgi?LINK_LOC=BlastHome) based on the coding sequence (CDS) of the target gene to amplify fragments ranging from 400 to 700 bp, as listed in S14 Table. Sequence specificity was carefully analyzed to ensure accuracy. The PCR products were inserted into the pJC53.2 vector using TA cloning. The resulting plasmids were purified from E. coli, validated by sequencing, and then used as templates for riboprobe generation.

dsRNA synthesis

dsRNA production was conducted as previously described [44,45]. Briefly, the recombinant pJC53.2 plasmid, containing the target gene fragment as previously described, was utilized as a template for dsRNA synthesis. The target dsRNA was synthesized ex vivo using a T7 High Yield RNA Synthesis Kit (Yeasen BioTechnologies, China) for RNAi applications. The oligonucleotide sequences used for dsRNA template generation are provided in S14 Table.

Sample collection for RNA-seq

To assess gene expression changes following the knockdown of Smp_019690, 6 adult male worms were placed in each well of a 12-well plate with 3 mL of Basch's medium 169, supplemented with 30 µg/mL dsRNA. The medium was replaced every two days, and dsRNA was administered at day 0, 2, and 6. Worms were collected at day 10 for sequencing. As controls, worms cultured in parallel were treated with dsRNA derived from a bacterial gene [44]. To evaluate gene expression changes following treatment with the compound CH, 6 adult male worms were placed in each well of a 12-well plate with 3 mL of Basch's medium 169. Parasites were treated with 0.618 μM CH for 24 hours or 0.3 μM PZQ for 12 hours, with worms treated with DMSO as control. After Smp_019690 RNAi, CH or PZQ treatment, worms were harvested and washed with PBS, and then flash frozen in liquid nitrogen for RNA extraction and Illumina sequencing (Novogene, China). For both experiments, each biological replicate consisted of worms from two wells, with 3 biological replicates for each condition.

RNA-seq data processing and splicing analysis

Schistosome total RNA was isolated using Qiagen miRNeasy Mini Kit and qualified using an Agilent 5400 system. RNA sequencing was performed according to the manufacturer's protocol on Illumina platforms with PE150 strategy in Novogene Bioinformatics Technology Co., Ltd. (China). For differential expression analysis, raw sequencing reads were aligned to the reference genome using Hisat2 [46], a fast and sensitive splice-aware aligner optimized for RNA-seq data. The resulting alignment files were subsequently processed with StringTie [47] to assemble transcripts and quantify gene-level expression. The generated count matrix was then used as input for DESeq2 [48], a widely utilized R package that employs a negative binomial distribution model to identify differentially expressed genes (DEGs). DESeq2 performed data normalization, dispersion estimation, and statistical testing to detect significant expression changes between experimental conditions. Genes with an adjusted P-value < 0.05 and |Log_2_FoldChange| ≥ 1 were considered significantly differentially expressed.

For alternative splicing analysis, we employed rMATS (replicate Multivariate Analysis of Transcript Splicing) [49], a robust computational tool for detecting differential alternative splicing events from RNA-seq data. BAM files generated by StringTie were used as input for rMATS, and the analysis was conducted with default parameters. This approach enabled the identification of five major types of alternative splicing events: skipped exons (SE), retained introns (RI), mutually exclusive exons (MXE), and alternative 5' and 3' splice sites (A5SS and A3SS). rMATS implemented a statistical framework to compare splicing patterns between experimental groups and estimated the inclusion level (ψ values) for each splicing event. Significance was assessed using false discovery rate (FDR)-corrected P-values with a threshold of 0.05.

Expression and purification of recombinant Smp_019690 and its mutants

Two mutants of Smp_019690 (SmU2AF65) were generated based on the CH-binding interface predicted by molecular docking. Mutant 1 (L255A/R270A/Q284A/V289A) and Mutant 2 (L255A/R270E/Q284R/V289A) genes were synthesized by a commercial provider (Tsingke Biotechnology,China) and verified by Sanger sequencing.

The splicing factor U2AF subunit (Smp_019690) of S. mansoni and its mutants were expressed in E. coli strain ArcticExpress (DE3) as a His+SUMO fusion protein in the pET28a vector. The recombinant proteins were purified using Ni^2+^ affinity chromatography on a HisTrap column (Cytiva), followed by size-exclusion chromatography on a Superdex-75 prep-grade column (GE Healthcare). The oligonucleotide sequences used to generate the full-length Smp_019690 template were provided in S14 Table.

Each recombinant protein was initially bound to a HisTrap column equilibrated with a buffer containing 20 mM Tris-HCl, 300 mM NaCl, and 20 mM imidazole at pH 8.0. Elution was performed using a buffer containing 20 mM Tris-HCl, 300 mM NaCl, and 500 mM imidazole at pH 8.0. Protein purification was carried out using the AKTA protein purification system, and elution was monitored via UV detection at 280 nm to track the protein in the eluted fractions. Purity was confirmed by SDS-PAGE gel analysis, which was stained with Coomassie Brilliant Blue.

To cleave the His+SUMO tag, the recombinant proteins were treated with ULP1 protease during dialysis against a buffer containing 20 mM Tris·HCl, 300 mM NaCl, pH 8.0, 10% (v/v) glycerol, and 0.1 mM EDTA. The cleaved His+SUMO tag was separated from U2AF65 using Ni² ⁺ affinity chromatography, followed by size-exclusion chromatography on a previously equilibrated Superdex-75 prep-grade column with 20 mM Tris·HCl, 300 mM NaCl, pH 8.0, and 0.1 mM EDTA. The purified proteins were concentrated using an Ultra Centrifugal Filter (Millipore) with a 50 kDa MWCO, and protein concentration was estimated by its extinction coefficient (38,655 M ⁻ ¹cm ⁻ ¹) and absorbance at 280 nm.

Electrophoretic mobility shift assay (EMSA) for protein/RNA binding

FAM labeled and unlabeled RNA fragments were synthesized (Genscript, China) and diluted to a final concentration of 100 μM in TE buffer (10 mM Tris-HCl, 0.1 mM EDTA, pH 8.0). The RNA fragments were denatured by incubating at 70°C for 5 minutes, followed by immediate cooling on ice. Half pmol of FAM-labeled RNA was incubated with 1 μg of recombinant wild-type U2AF65 protein or 2 μg of mutant proteins in reaction buffer (Beyotime, China) for 20 minutes at room temperature, in a final volume of 10 μL. For the competitive control group, 200 pmol of unlabeled RNA was added. In the compound-binding groups, the compounds (10–30 μM) were pre-incubated with the protein for 30 minutes at room temperature before adding the RNA fragment. The reactions were analyzed by electrophoresis on 6% non-denaturing polyacrylamide gels, prepared in 1 × TBE buffer (89 mM Tris, 89 mM boric acid, 2 mM EDTA, pH 8.3). A pre-run was performed at 200 V for 10 minutes, followed by electrophoresis at 200 V for approximately 25 minutes. The gel was visualized using a FLA Typhoon 9000 (GE), and signals from the entire lanes were quantified using ImageJ software.

Statistical analysis

Statistical analysis of data was performed using GraphPad Prism software, and results are presented as mean ± SD. Statistical significance was determined using an unpaired two-tailed parametric t-test, where P-values < 0.05 were considered significant. The significance levels are indicated as follows: NS (not significant); *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Supporting information

S1 FigRNAi-induced phenotypes of target genes in S. mansoni and S. japonicum.(A) Bright-field images showing phenotypes of adult S. mansoni after RNAi treatment. (B) Bright-field images showing phenotypes of adult S. japonicum after RNAi treatment. Representative images from 3 biological replicates, Scale bars, 1000 μm.(TIF)

S2 FigStructural alignment of homologs between S. mansoni and S. japonicum for 37 target proteins.Each alignment demonstrates the structural similarity between homologous proteins from the two schistosome species. Protein structures were superimposed using PyMOL, and RMSD values were calculated to quantify structural similarity. The protein structure of S. mansoni is shown in cyan, while that of S. japonicum is depicted in salmon. The small inset in the lower left corner represents the alignment of the full-length protein structures, whereas the enlarged section highlights the alignment of conserved regions or specific protein domains.(TIF)

S3 FigPhenotypic details of adult worms treated with lethal compounds.(A) Phenotypes of S. mansoni treated with different compounds at concentration of 10 μM for 7 days. (B) Phenotypes of S. japonicum treated with different compounds at concentration of 10 μM for 7 days. DMSO as negative control; PZQ as positive control. Scale bars, 500 μm. Red arrows denote tegumental erosion.(TIF)

S4 FigDose–response analyses of highly effective compounds in *S. mansoni.*IC50 values were determined for ten high-efficacy compounds targeting adult S. mansoni. The fitted IC50 curves are shown in the figure, with praziquantel (PZQ) used as a control. Dose-response curves were generated using GraphPad Prism based on IC50 values calculated from experimental data. The IC50 value was evaluated on Day 7 (D7) post-treatment.(TIF)

S5 FigDose–response analyses of highly effective compounds in *S. japonicum.*IC50 values were determined for ten high-efficiency compounds targeting adult S. japonicum., The fitted IC50 curves are shown in the figure, with praziquantel (PZQ) used as a control. Dose-response curves were generated using GraphPad Prism based on IC50 values calculated from experimental data. The IC50 value was evaluated on Day 7 (D7) post-treatment.(TIF)

S6 FigDose–response curves of cytotoxicity for highly active anti-schistosomal compounds in HepG2 cells.Cytotoxicity of 11 potent anti-schistosomal compounds was assessed in HepG2 cells using the CCK-8 assay. Dose–response curves were fitted using a four-parameter logistic (4PL) regression model in GraphPad Prism software.(TIF)

S7 FigPhylogenetic analysis of the protein product of *Smp_019690.*The green leaf represents U2AF35, the red leaf represents U2AF65. The tree was constructed using the Maximum Likelihood method in MEGA 11, based on amino acid sequences from 7 species. Bootstrap values from 1000 replicates are indicated at the nodes. Smp_019690 belongs to the U2AF65 family.(TIF)

S8 FigMultiple sequence alignment of U2AF65 protein.Multiple sequence alignment of U2AF65 proteins across various species, including S. mansoni (Smp_019690.1), S. haematobium (MS3_00003258.1), S. japonicum (EWB00_008591), Caenorhabditis elegans (NP_001022967.1), Danio rerio (NP_991252.1), Mus musculus (NP_001192160.1), and Homo sapiens (NP_009210.1). The percentage of conserved residues was calculated for each column based on physicochemical properties. Columns with a similarity score above 0.7 are considered highly conserved, with residues highlighted in red and framed in blue. Strictly identical residues appear in white on a red background. The blue-font-labeled line above the alignment represents the predicted secondary structure of human U2AF65, while the blue-font-labeled line below the alignment corresponds to that of S. mansoni U2AF65.(TIF)

S9 FigLocalization of Smp_019690 expression in S. mansoni.(A, B) UMAP plots showing the expression pattern of Smp_019690 in adult female (A) and adult male (B) S. mansoni, obtained from the S. mansoni single-cell atlas (https://www.collinslab.org/schistocyte/). Color intensity represents the expression levels, with darker shades indicating higher expression. (C) WISH showing expression of Smp_019690 in adult female and male worms. Representative images from n > 5 worms. Scale bars, 100 μm. (D) FISH showing expression of Smp_019690 in somatic tissues and reproductive organs of adult female and male worms. Representative images from n > 5 worms. Nuclei are labeled by DAPI (grays), Smp_019690 mRNA shown in magenta. Scale bars, 50 μm.(TIF)

S10 FigTranscriptomic profiles of male schistosomes under Smp_019690 RNAi and CH treatment.(A) Principal Component Analysis (PCA) on the RNA-seq data under Smp_019690 RNAi. (B) Heatmap of the Euclidean distance matrix with hierarchical clustering of transcriptomes from Smp_019690 RNAi and control RNAi samples. (C) Heatmap showing the cluster of top 2000 highly variable genes in the parasites from Smp_019690 RNAi and control group. (D) PCA on the RNA-seq data under CH treatment. (E) Heatmap of the Euclidean distance matrix with hierarchical clustering of transcriptomes from CH and DMSO treatment samples. (F) Heatmap showing the cluster of top 2000 highly variable genes in the parasites from CH and DMSO treatment group.(TIF)

S11 FigEMSA analysis of SmU2AF65 binding activity at varying CH concentrations.(A, B) Effects of varying concentrations of CH on the binding interaction between SmU2AF65 and RNA. Grayscale intensity analysis of the shift bands was performed using ImageJ software for quantitative evaluation.(TIF)

S12 FigMutagenesis analysis of SmU2AF65 residues essential for CH binding.(A, upper panel) Predicted 2D interaction diagram of SmU2AF65 with CH, showing four key residues (LEU255, ARG270, GLN284, and VAL289) at the binding interface. (A, lower panel) Schematic representations of wild-type (WT) SmU2AF65 and two mutants: Mutant 1 (L255A/R270A/Q284A/V289A) and Mutant 2 (L255A/R270E/Q284R/V289A). (B) SDS-PAGE analysis of purified mutant SmU2AF65 proteins. Lane M: protein molecular weight marker; Lane 1: purified mutant protein prior to His and SUMO tag cleavage; Lane 2: purified protein after His and SUMO tag removal. (C) EMSA results showing RNA-binding activity of Mutant 1 and Mutant 2. (D) EMSA assessing the effects of increasing CH concentrations on the interaction between Mutant 1 and RNA. (E) Quantitative analysis of Mutant 1–RNA shift band intensity at different CH concentrations, based on grayscale measurements from panel D using ImageJ. ***P *< 0.01, ns, P > 0.05.(TIF)

S13 FigAlternative splicing changes after PZQ treatment.Differential alternative splicing events after 0.3 μM PZQ treatment for 12 hours. AS, alternative splicing; SE, skipped exon; A5SS, alternative 5' splice sites; A3SS, alternative 3' splice sites; MXE, mutually exclusive exons; RI, retained intron.(TIF)

S14 FigEMSA assessing the impact of HY-N0768 and HY-13624A on SmU2AF65–RNA interaction.(A) Representative EMSA gel depicting the binding of recombinant SmU2AF65 protein to RNA in the presence of 30 μM HY-N0768 or HY-13624A. No appreciable disruption of the SmU2AF65–RNA complex was observed relative to the DMSO vehicle control. (B) Quantitative analysis of band intensities from 3 independent replicates. ns, P > 0.05.(TIF)

S15 FigIn vivo evaluation of CH in a murine S. mansoni infection model.(A) Images of liver and spleen tissues from control and CH-treated mice following oral administration of CH (720 µg/kg) starting from 4 or 6 weeks post-infection. (B) Worm burden analysis in control and CH-treated groups under the same dosing regimen (mean ± SD, n ≥ 5). No significant difference was observed compared with control (ns, P > 0.05). cntl, control. Scale bars, 2 cm.(TIF)

S1 TableTarget proteins for virtual docking.(XLSX)

S2 TablePotential binding compounds for each target protein from MCE library.(XLSX)

S3 TableSelected 268 compounds against 37 target proteins.(XLSX)

S4 TableComparison of target protein sequences and structures between S. mansoni and S. japonicum.(XLSX)

S5 TableInformation of effective compounds against S. mansoni and S. japonicum at 10 μM.(XLSX)

S6 TablePhenotypes of adult S. mansoni under the treatment of effective compounds.(XLSX)

S7 TablePhenotypes of effective compounds against S. mansoni and S. japonicum at 1 μM.(XLSX)

S8 TablePhenotypes of adult S. japonicum under the treatment of effective compounds.(XLSX)

S9 TablePhenotypes of juvenile S. mansoni under the treatment of effective compounds.(XLSX)

S10 TablePhenotypes of juvenile S. japonicum under the treatment of effective compounds.(XLSX)

S11 TableList of DEGs identified in the Smp_019690 RNAi group vs. control group.(XLSX)

S12 TableList of DEGs identified in the CH-treated group vs. DMSO group.(XLSX)

S13 TableList of overlapping DEGs between RNAi and CH treatment groups.(XLSX)

S14 TableOligonucleotides used for RNAi, protein expression, in situ hybridization and qPCR.(XLSX)