Large-scale RNAi screening uncovers new therapeutic targets in the human parasite Schistosoma mansoni

Schistosomes kill 250,000 people every year and are responsible for serious morbidity in 240 million of the world’s poorest people. Despite their profound global impact, only a single drug (praziquantel) is available to treat schistosomiasis, highlighting the need to better understand schistosome biology to drive the development of a new generation of therapeutics. A major barrier to this goal is the paucity of large-scale datasets exploring schistosome gene function. Here, we describe the first large-scale RNA interference screen in adult Schistosoma mansoni examining the function of over 2000 genes representing approximately 20 percent of the protein coding genome. More than 250 genes were found to have phenotypes affecting neuromuscular function, tissue integrity, stem cell maintenance, and parasite survival. Leveraging these data, we bioinformatically prioritized several compounds with in vitro activity against parasites and validated p97, a component of the ubiquitin proteasome system, as a drug target in the worm. We further reveal a potentially druggable protein kinase-signaling module involving the TAO and STK25 kinases that are essential for maintaining the transcription of muscle-specific mRNAs. Importantly, loss of either of these kinases results in paralysis and death of schistosomes following surgical transplantation into a mammalian host. We anticipate this work will invigorate studies into the biology of these poorly studied organisms and expedite the development of new therapeutics to treat an important neglected tropical disease.

Genome sequences are available for the major species of medically-relevant schistosomes 1-50 3 ; nevertheless, studies of gene function have been limited to relatively small numbers of genes 4,5 . 51 To address this issue, we developed a platform for large-scale RNAi screening on adult 52 schistosomes (Fig. 1a). To establish the efficacy of this platform to detect phenotypes in adult S. 53 mansoni, we prioritized a list of 2,320 of the worm's ~10,000 protein coding genes, including 54 those encoding enzymes, cell-surface receptors, ion channels, and hypothetical proteins of 55 unknown function (Supplementary Table 1). After filtering for genes expressed in adult 56 schistosome somatic tissues using existing expression datasets 6 , we performed Polymerase Chain 57 Reactions (PCR) from schistosome cDNA, generated dsRNAs, and performed RNAi by treating 58 adult pairs of male and female worms with five dsRNA treatments over the course of a 30-day 59 experiment (Fig. 1b). After filtering genes that either did not amplify during PCR steps, or failed 60 to generate sufficient concentrations of dsRNA, a total of 2,216 genes were screened 61 (Supplementary Table 1). 62 These parasites live in the veins surrounding the host intestines, and attachment to the 63 vascular endothelium is essential in vivo for parasites to be kept from being swept away in the 64 blood and trapped in host organs. Since detachment from an in vitro tissue culture substrate has 65 been shown to precede more deleterious phenotypes 7 , and since under our in vitro culture 66 conditions, healthy parasites firmly attached to the substrate using a combination of their oral and 67 ventral suckers (Supplementary Video 1), we reasoned that substrate attachment would be a 68 useful quantitative metric to define RNAi treatments that affect parasite vitality and predict in vivo 69 survival. Therefore, during our 30-day experiments we monitored parasites every 48 hours for 70 substrate attachment and any other visible defects. Schistosomes possess adult somatic stem cells, 71 called neoblasts, that rejuvenate key parasite tissues, including the intestine and tegument (skin) 6,8 , 72 and are likely to be essential for long-term parasite survival in the blood. The parasites also contain 73 large numbers of proliferative germline stem cells (GSCs) in their reproductive organs 8 which are 74 essential for producing eggs that represent the central driver of parasite-induced pathology in vivo 9 . 75 Therefore, we also monitored the maintenance of neoblasts and GSCs by labeling with the 76 thymidine analog EdU prior to the conclusion of the experiment (Fig. 1b). Due to the variable rate 77 at which the reproductive organs of female worms degenerate during in vitro culture 10 , stem cell 78 proliferation was only monitored in male worms. At the conclusion of this initial screen we 79 performed two major quality control steps for RNAi treatments resulting in attachment-or stem 80 cell-related phenotypes. First, we confirmed the identity of every gene producing a phenotype by 81 DNA sequencing. Second, where possible, we examined the specificity of our RNAi knockdown 82 by designing new oligonucleotides targeting a non-overlapping region of genes that produced 83 phenotypes (Fig. 1a, Extended Data Fig. 1). To be considered a "hit" a gene must have shown a 84 fully penetrant phenotype in three independent experiments. These studies identified 195 genes 85 that were essential for parasite attachment, and thus potentially essential for worm survival in vivo. 86 In addition to facilitating parasite substrate attachment, we also observed that 121 of these 195 87 genes were associated with other visible phenotypes including tissue and intestinal edema (36), 88 head (26) and/or tegument (78) degeneration, muscular hypercontraction (6), and complete 89 cessation of movement (death) (36) (Fig. 2a, Supplementary Table 2). In addition to these genes, 90 we found that RNAi of an additional 66 genes resulted in stem cell maintenance defects but caused 91 no other visible phenotypes (e.g., substrate attachment) suggesting a selective role in stem cell 92 maintenance (Supplementary Table 3, Extended Data Fig. 2). 93 Of the 66 genes essential for stem cell survival over 90% (60/66) led to defects in the 94 maintenance of both neoblasts and proliferative cells in the male testes (Extended Data Fig. 2). 95 However, in a minority of cases some genes appeared to play more significant roles in maintaining 96 proliferative cells in either the male germ line (e.g., a RAD51 homolog) or the neoblasts (e.g., 97 fgfrA, a previously-described FGF receptor homolog 8 ) (Fig. 2b). In addition to genes necessary 98 for cell cycle progression (e.g., polo-like kinase), Gene Ontology enrichment analysis highlighted 99 genes important for protein translation, including gene products involved in ribosomal structure, 100 tRNA aminoacylation, and rRNA processing as important regulators of proliferative cell 101 maintenance (Fig. 2c, Extended Data Fig. 3). Although this could reflect an enhanced sensitivity 102 of actively proliferating cells to alterations in protein translation, recent work has highlighted "non-103 housekeeping" roles for translational regulators in controlling stem cell function 11 . Thus, it is worth 104 exploring whether specific roles for translational control exist for regulating schistosome stem cell 105

function. 106
Similar to previous whole organism large-scale RNAi studies in other metazoa 12,13 , we 107 found that a large fraction of the 195 genes essential for parasite vitality (attachment) share 108 sequence similarity (BLAST E-value < 1e-5) with genes in other organisms including C. elegans 109 (91%), Drosophila (93%), the planarian Schmidtea mediterranea (97%), and humans (93%) 110 Table 4). Some of these 195 schistosome genes with detachment phenotypes 111 have C. elegans/D. melanogaster orthologs that lack any phenotypes (Supplementary Table 5); 112 such genes could regulate novel schistosome-specific biology or represent opportunities for studies 113 of S. mansoni to shed light on the function of poorly characterized animal gene families. Further 114 examination of genes with attachment phenotypes by Gene Ontology analyses revealed that 115 although this dataset was enriched for genes encoding regulators of protein transport and mRNA 116 transcription (Fig. 2c, Extended Data Fig. 3), the dominant group of enriched genes were those 117 encoding components necessary for protein turnover via the ubiquitin-proteasome system (UPS) 118 ( Fig. 2c, Extended Data Fig. 3). RNAi and pharmacological studies have implicated proteolysis 119 by the proteasome as important for larval, and, more recently adult viability in vitro 14,15 . However, 120 our data points to a much broader requirement for UPS components in these worms. Indeed, 121 inspection of our RNAi dataset found that key components from virtually every arm of the UPS 122 were required for adult parasite vitality during in vitro culture including: E1/E2 ubiquitin ligases 123

(Supplementary
and Deubiquitinating Enzymes (DUBs), the AAA-ATPase p97 that delivers proteins to the 124 proteasome 16 , and nearly all regulatory and catalytic subunits of the proteasome complex 17 (Fig.  125   2d). Indeed, RNAi of nearly all of UPS components resulted in extensive tissue degeneration and 126 in some cases (e.g., p97(RNAi)) adult parasite death (Extended Data Fig. 4). Taken together, these 127 data suggest that disruption not just of proteasome function, but any critical UPS components, 128 results in reduced schistosome vitality in vitro. 129 To determine if any genes associated with attachment phenotypes encoded proteins 130 targeted by existing pharmacological agents, we performed a combination of manual searches of 131 the literature and bioinformatic comparisons with the ChEMBL database 18 (Supplementary 132 Table 6). This analysis uncovered 205 compounds potentially targeting 49 S. mansoni proteins. 133 To gauge the utility of this approach to prioritize compounds with activity on adult parasites, we 134 selected 14 of these compounds (Supplementary Table 7), including: FDA-approved drugs (e.g., 135 Ixazomib, Panobinostat), drugs currently or previously explored in clinical trials (e.g., CB-5083, 136 HSP990), or experimental compounds with activity in rodent models of disease (e.g., 137 Thapsigargin, NMS-873). We then examined their activities on worms cultured in vitro using an 138 automated worm movement tracking platform 19 and by measuring the effects on parasite 139 attachment following drug treatment. This analysis found that more than half of the compounds 140 tested (8/14) on worms at 10 µM reduced parasite movement >75% and half of the compounds 141 (7/14) caused fully penetrant substrate attachment defects by D7 post-treatment ( Fig. 3a-b,  142 Supplementary Video 2). Among the compounds that emerged from these studies was 143 simvastatin, an HMG-CoA reductase inhibitor, that was previously shown to have effects on 144 parasites both in vitro and in vivo 20 . We also evaluated these compounds on post-infective larvae 145 (schistosomula), observing that 7 had profound effects on parasite movement (Supplementary 146 Table 8), suggesting the potential of these compounds to target multiple schistosome life-cycle 147 stages. Consistent with our observation that UPS function is critical for schistosome vitality (Fig.  148 2d-f), we found that the proteasome inhibitor ixazomib caused profound effects on both 149 schistosome movement (Fig. 3a) and attachment (Fig. 3b), mirroring a recent report using the suggesting that these compounds have similar pharmacological effects on the parasite (Fig. 3c). 159 Given the prominent role for the UPS in schistosomes (Fig. 2c-d), we assessed if NMS-873 and 160 CB-5083 affected UPS function by measuring the accumulation of ubiquitinated proteins using an 161 antibody that recognizes K48 polyubiquitinated proteins marked for proteasome-mediated 162 destruction 23 . Not only did we observe the accumulation of polyubiquitinated proteins following 163 RNAi of p97, treatment of schistosomes with either CB-5083 or NMS-873 enhanced anti-K48 164 polyubiquitin labeling (Fig. 3d). We observed similar accumulation of polyubiquitinated proteins 165 following either RNAi of proteasome subunit beta type-2 or treatment with ixazomib (Extended 166 Data Fig. 5). These effects on the degradation of ubiquitinated proteins appeared to be specific to 167 inhibition of UPS function, rather than a non-specific effect due to reduced worm vitality, as 168 treatment with the sarco/endoplasmic reticulum Ca 2+ -ATPase inhibitor thapsigargin, which also 169 caused profound effects on worms (Fig 3a, 3b), did not alter the accumulation of polyubiquitinated 170 proteins (Extended Data Fig. 5). 171 To determine if UPS function is broadly required for adult schistosomes in vivo, we 172 depleted UPS components using RNAi and surgically transplanted these worms into the 173 mesenteric veins of recipient mice (Extended Data Fig. 6) to measure parasite egg deposition in 174 host tissues and parasite survival 7 . Following hepatic portal perfusion, we recovered about 55% 175 of control RNAi-treated worms originally transplanted (Fig. 2e, Extended Data Fig. 6) and these 176 parasites established patent infections depositing large number of eggs into the livers of recipient 177 mice (Fig. 2f, Extended Data Fig. 6). In contrast, we failed to recover parasites following hepatic 178 portal perfusion from mice receiving p97 (Fig. 2e) or proteasome subunit beta type-2 (Extended 179 Data Fig. 6) RNAi-treated worms. Additionally, the livers in these mice were devoid of eggs, as 180 a consequence, we observed no signs of egg-induced granulomas (Fig. 2f, Extended Data Fig. 6). 181 We did, however, observe RNAi-treated parasites at various stages of being infiltrated by host 182 immune cells in the livers of recipient mice (Fig. 2g, Extended Data Fig. 6), suggesting these 183 parasites are unable to remain in the portal vasculature and are cleared via the immune system in 184 the liver. Thus, several components of the UPS are essential for schistosome survival in vivo. 185 Recent studies from a variety of human parasites have highlighted the potential for therapeutically 186 targeting UPS function by inhibition of the proteasome 14,24,25 . Our data suggest that targeting 187 another critical (and druggable 21,22 ) mediator of UPS function (i.e., p97) may have therapeutic 188 potential, not just against schistosomes, but against a variety of important human parasites. 189 Another prominent group of potentially druggable targets to emerge from our RNAi screen 190 were protein kinases, 19 of which led to defects in either parasite attachment or stem cell 191 maintenance. The most striking protein kinase-related phenotypes resulted from RNAi of two 192 STE20 serine-threonine kinases: tao and stk25, which are homologs of the human TAO1/2/3 and 193 STK25/YSK1 protein kinases, respectively. RNAi of either of these kinases led to rapid 194 detachment from the substrate (Extended Data Fig. 7) and a concomitant posterior paralysis and 195 hypercontraction of the body, such that the parasites were shorter than controls and took on a 196 distinctive banana-shaped morphology (Fig 4a-b, Supplementary Video 4). Aside from RNAi 197 of stk25 and tao, this banana-shaped phenotype was unique, only observed in our screening 198 following RNAi of a CCM3/PDCD10 homolog (Smp_031950), a known heterodimerization 199 partner with the mammalian STK25 kinase 26 . We failed to observe death of either stk25-or tao-200 depleted parasites during in vitro culture; however, following surgical transplantation we noted a 201 significant reduction in the recovery of tao or stk25 RNAi-treated parasites from recipient mice 202 and these recipient mice displayed little signs of egg-induced granuloma formation (Extended 203 Data Fig. 7). Thus, both tao and stk25 appear to be essential for schistosome survival in vivo. 204 Given the unique and specific nature of the stk25 and tao associated "banana" phenotype 205 we reasoned that these kinases may be acting in concert to mediate similar signaling processes in 206 the worm. Recent data suggests that the Drosophila STK25 ortholog (GCK3) is a substrate of 207 TAO and that these proteins function in a signaling cascade essential for tracheal development 27 . 208 Consistent with these studies, we too observed that recombinant S. mansoni STK25 (SmSTK25) 209 could serve as a substrate for the S. mansoni TAO (SmTAO) in an in vitro kinase assay (Extended Given their phenotypic similarities and our biochemical observations, we reasoned that the 225 schistosome TAO and STK25 might be acting in a signaling module to mediate critical processes 226 in the parasite. To define these processes, we performed transcriptional profiling on RNAi-treated 227 parasites just prior to the timepoint in which we observed detachment and hypercontraction (Day 228 6 and Day 9 for tao and stk25 RNAi treatments, respectively) (Extended Data Fig. 9). We 229 reasoned that transcriptional changes common to both stk25 and tao RNAi data sets would provide 230 details about any processes regulated by these proteins. Consistent with the model that these 231 kinases cooperate in the parasite, we found that expression of differentially regulated genes 232 following RNAi of either tao or stk25 were highly correlated (Fig. 4d) and that more than half of 233 these differentially regulated genes were common in both datasets (Extended Data Fig. 9, 234 Supplementary Table 9). Importantly, RNAi of either tao or stk25 was specific, not affecting 235 expression of the other kinase gene of this pair (Fig. 4c, d). To better understand the genesis of 236 the phenotype associated with loss of tao or stk25, we examined the tissue-specific expression of 237 differentially-regulated genes on an adult schistosome single cell expression atlas using cells from 238 schistosome somatic tissues 29 . Strikingly, we found that roughly 40% (51/129) of the most down-239 regulated genes following tao and stk25 RNAi (Log2 Fold Change < -0.5, adjusted p-value < 240 0.00001) were specific markers of parasite muscle cells (Extended Data Fig. 10, Supplementary 241 Table 9). Indeed, nearly half of all mRNAs specifically-enriched in muscle cells (60/135) from 242 this single cell atlas, including key muscle contractile proteins (e.g, Troponin subunits Actins, 243 Myosin light/heavy chains, and Tropomyosin), were significantly down-regulated following RNAi 244 of both tao and stk25 (Fig. 4e, Extended Data Fig. 10, Supplementary Table 10). Importantly, 245 these transcriptional effects appeared to be largely specific to parasite muscles, since 246 comparatively few markers specific to other major somatic organ systems (neurons, gut, 247 parenchyma) were affected by RNAi of these kinases (Fig. 4e, Extended Data Fig. 10, 248 Supplementary Table 10). In principle, loss of muscle-specific mRNAs could be due to either 249 loss of muscle cells or down-regulation of muscle-specific mRNAs. To distinguish between these 250 possibilities, we performed labeling with phalloidin to mark F-actin in schistosome muscle fibers 251 and in situ hybridization to detect muscle-specific mRNAs. Within a few days of RNAi-treated 252 parasites adopting their banana-shaped phenotype, we noted a dramatic reduction in the expression 253 of mRNAs encoding the contractile proteins Tropomyosin 1 and a Myosin Light Chain by in situ 254 hybridization (Fig. 4f), but observed no major qualitative defects in phalloidin labeling in the 255 muscle fibers within anterior or posterior body wall muscles (Fig 4g, Extended Data Fig. 8). 256 Thus, it appears that these kinases are required to maintain the transcription of a large number of 257 muscle-specific mRNAs in intact muscle cells. Interestingly, we noted that the heads of tao and 258 stk25 RNAi parasites, which retained their capacity for movement (Supplementary Video 4), 259 partially maintained the expression of muscle-specific mRNAs (Fig. 4f). Thus, there appears to 260 be a relationship between the maintenance of muscle-specific mRNA expression and locomotion. 261 Taken in their entirety, our data are consistent with the model that STK25 and TAO kinases 262 cooperate (perhaps with TAO directly activating STK25) in the schistosome to mediate a signaling 263 cascade essential for sustaining transcription of muscle-specific mRNAs. As a consequence, loss 264 of either SmSTK25 or SmTAO activity results in muscular function defects and this compromises 265 parasite survival in vivo. Although the essentiality of the three mammalian TAO homologs is 266 unclear, whole body knockouts of mouse STK25 are homozygous viable displaying no obvious 267 deleterious phenotypes 30 . Thus, SmSTK25 function appears to be a schistosome-specific liability 268 for survival when compared to mammals. Given this, and the druggable nature of kinases, we 269 suggest that SmSTK25 represents a high-value target for therapeutic intervention. RNAi studies, together with bioinformatics, have allowed us to effectively prioritize targets 275 essential in vivo (e.g., STK25, TAO, and p97) and potential specific inhibitors with in vitro 276 activities on worms (Fig 3a-b). Thus, future efforts should not only explore compounds our 277 bioinformatic approaches have already uncovered (Supplementary Table 6), but also larger 278 libraries of compounds with known molecular targets (e.g., the REFRAME collection 34 ). Such 279 studies are likely to be an efficient means to identify existing drugs for potential repurposing 280 against schistosomes. Not only does this study enhance our understanding of schistosome biology, 281 and serve as a template for conducting further genome-scale studies of schistosome gene function, 282 it provides a new lens to prioritize genes of interest in other medically-and agriculturally-283 important parasitic flatworms (e.g., tapeworms and flukes). Collectively, we anticipate this study 284 will expedite the discovery of new anthelmintics.

435
Parasite labelling and imaging

436
Whole-mount in situ hybridization 6 , EdU detection 8 , and phalloidin staining 37 were performed as previously described.

437
For in situ hybridization, riboprobes were generated from cDNA fragments amplified using primers for tropomyosin-

471
The blot was washed in TBST 3x before being imaged on a Li-Cor Odyssey Infrared Imager.

474
To manually search for existing drugs targeting "detachment" hits from our RNAi screen, we performed protein-475 protein BLAST against the Homo sapiens proteome to find the closest human homolog to our RNAi hits. We then 476 manually searched a variety of databases (e.g., genecards, google, DrugBank, Therapeutic Targets Database) and 477 chemical vendors (e.g., seleckchem) for inhibitors against these human proteins. In each instance, we consulted the 478 published literature to give preference to compounds likely to be selective for a given target. If several such drugs 479 were available, preference was given to those that were also orally bioavailable and/or FDA approved/in clinical trials.

480
For larger-scale discovery of compounds, the S. mansoni protein sequences of genes with 'detachment' phenotypes 481 were used to search the ChEMBL database 18 , to identify compounds predicted to interact with them. To do this, we 482 followed the protocol previously described 38 with the following differences. First, for each S. mansoni gene, we

519
To examine gene expression changes following loss of tao or stk25, 10 adult worm pairs were placed into 6-well plates 520 and cultured in 3 mL Basch 169 supplemented with 30 µg/mL dsRNA for 3 days. Media and dsRNA were replaced 521 daily. On D3, dsRNA-containing media was removed and worms were maintained in 6 mL Basch 169 media that was 522 replaced every other day. On day 6 (tao(RNAi)) or D9 (stk25(RNAi)) worms were anesthetized with 0.25% tricaine 523 and separated by sex. As controls, worms cultured in parallel were treated with an irrelevant dsRNA 8 . For RNA 524 extraction, male worms were collected, excess media removed, and 100 µL of TRIZOL was added. Parasites were 525 then flash frozen in liquid N2, homogenized with a micro pestle, the volume of TRIZOL was brought to 600 µL before 526 RNA was purified using a Zymo Direct-zol RNA miniprep kit and processed for Illumina sequencing. RNAseq data 527 was mapped to the S. mansoni genome (v7) using STAR and differential expression was analyzed by DESeq2 as