Prp8 regulates oncogene-induced hyperplastic growth in Drosophila

Intensive research in the recent past that has combined molecular profiling approaches with in vivo and in vitro functional studies has resulted in the identification of genes and pathways that drive tumour formation (Hanahan and Weinberg, 2011). In this regard, the use of the fruit fly Drosophila melanogaster as a model organism has been particularly powerful (Gonzalez, 2013; Sonoshita and Cagan, 2017; Tipping and Perrimon, 2014). Indeed, seminal studies using Drosophila have led to the identification of multiple genes and signalling pathways, including the Notch (N) and Ras/MAPK pathways that, when mutated, not only cause severe developmental defects but are also involved in tumourigenesis (Gonzalez, 2013). Indeed, different aspects of tumourigenesis have been studied in Drosophila and the vast majority of cancer hallmarks are conserved in flies (Hanahan and Weinberg, 2011; Tipping and Perrimon, 2014).

Signalling pathways underpin cellular behaviour and, when disrupted, lead to developmental defects and/or cellular transformation. Virtually all signalling pathways are controlled by post-translational protein modifications, with phosphorylation being the most frequently associated with signalling events (Hynes et al., 2013). However, it is clear that additional post-translational modifications are vital for tightly controlling developmental events. Ubiquitylation, a multi-step cascade that results in the covalent attachment of the small protein ubiquitin onto a substrate, has emerged as a crucial process in signalling that regulates virtually all functions within a cell (Heride et al., 2014). Despite being historically linked with regulation of protein levels and protein degradation, ubiquitylation can also have non-proteolytic effects, leading to changes in protein-protein interactions, protein function and subcellular localisation (Rape, 2017). In a manner akin to phosphorylation, ubiquitylation is reversible, and the removal of ubiquitin moieties from target proteins is controlled by deubiquitylating enzymes (DUBs) (Heride et al., 2014; Rape, 2017). However, the in vivo role of DUBs remains poorly explored. This is especially true in the context of developmental and oncogenic growth, despite the fact that many DUBs have recently been linked with tumourigenesis (Fraile et al., 2012).

We performed a Drosophila in vivo screening approach to study the role of genes containing domains that are involved in the removal of ubiquitin and ubiquitin-like proteins in the regulation of tumourigenesis. Our top hit was the spliceosome component Prp8, which we identified as a crucial regulator of developmental and hyperplastic growth in several Drosophila models of cancer. Prp8 is a core protein of the spliceosome complex and its protein structure includes an MPN/JAB domain typical of the JAMM family of DUBs (Grainger and Beggs, 2005; Komander et al., 2009). Based on sequence and structural analysis, Prp8 is thought to be an inactive DUB, as conserved residues of the JAMM ubiquitin hydrolase domain are absent (Clague et al., 2013; Pena et al., 2007). Nevertheless, the MPN/JAB domain is essential for Prp8 function and can bind ubiquitin with an affinity comparable with that of other ubiquitin-binding domains (Bellare et al., 2006). Our data suggest that Prp8 regulates hyperplasia in a context-dependent manner, which is consistent with previous observations that identified prp8 as a regulator of organ growth in vivo, in a genetic modifier screen that used overexpression of a kinase-dead phosphoinositide 3-kinase (Coelho et al., 2005).

Together with recently published data, our work identifies the spliceosome as a potential target in cancers and suggests that tumours display different sensitivity to disruption of Prp8 function depending on the driver oncogene (Hsu et al., 2015). Thus, our results imply that future therapies that target the spliceosome in cancer may require the identification of the exact context-dependent condition of individual tumours to maximise their efficacy.

To elucidate the role of DUBs in the regulation of developmental and pathological growth, we performed in vivo RNAi screens using lines targeting all Drosophila genes that carry a ubiquitin hydrolase domain (Broemer et al., 2010). To explore the role of ubiquitin-related modifications, we also included Drosophila orthologues of SUMO and NEDD8 hydrolases in our library of 123 RNAi lines targeting 54 genes (designated herein as DUB^RNAi for simplicity) (Table S1). To avoid potential early lethality phenotypes, we regulated RNAi expression spatially and temporally using an act-Gal4/Gal80^ts module and a FLP/FRT STOP cassette (FLPout) (Fig. S1A). We expressed the FLPase enzyme under the control of the eye-specific eyeless promoter (ey-FLP), such that DUB^RNAi expression was limited to the developing eye and was induced by shifting larvae from 18°C to 29°C 120 h after egg laying (AEL) to inhibit Gal80^ts function.

We initially assessed the role of DUBs in the normal growth of the developing Drosophila eye, and identified three genes which, when depleted, caused eye disc hypoplasia: prp8 (Fig. 1C,G), usp10 (Fig. S1B) and npl4 (Fig. S1C). We selected Prp8 for further study as the hypoplasia phenotype was fully penetrant, and was observed in several RNAi lines that target prp8, which were predicted to not have off-target effects. We next tested whether DUBs could influence tumour growth. To this end, we co-expressed the DUB^RNAi library with an oncogenic form of Ras (Ras^V12), thereby mimicking a well-established Drosophila tumour model in which expression of Ras^V12 causes hyperplasia (Lee et al., 1996; Pagliarini and Xu, 2003) (Fig. S1D,L and Table S2). This Ras^V12 model has been used to identify new regulators of growth and metastasis and, for example, previous research has uncovered the fact that combining Ras^V12 expression with loss-of-function mutations for polarity genes causes metastasis in larvae (Chabu et al., 2017; Ohsawa et al., 2012; Pagliarini and Xu, 2003). To validate our genetic model, we co-expressed Ras^V12 with RNAi lines that target the polarity genes scribbled (scrib^RNAi), lethal giant larvae (lgl^RNAi) and bazooka (baz^RNAi). Consistent with previous reports, a combination of Ras^V12 with RNAi against polarity genes resulted in enhanced overgrowth phenotypes in eye discs and, in some cases, in the appearance of distant metastases (compare Fig. S1D with Fig. S1E-G). Therefore, our model mimicked previously used systems to study oncogene-mediated growth and metastasis, and is an appropriate setting to test the role of DUBs in these processes.

Analysis of Ras^V12-expressing eye discs with simultaneous depletion of prp8 (Fig. 1D,H) revealed a dramatic reduction of eye disc overgrowth and a partial rescue of disc morphology when compared with Ras^V12 expression alone (Fig. 1B,F). Indeed, the major effect of prp8 depletion in Ras^V12 tumours was a decrease in size, such that the GFP-positive area in developing eye discs appeared to be similar to that of the controls (compare Fig. 1E with H, quantified in Fig. S4K). When analysed collectively, prp8^RNAi lines led to decreased hyperplastic growth in ∼80% of cases, with some of these eye discs displaying a near rescue of eye disc morphology (∼25% of cases) (Fig. 1I and Fig. S1L). In ∼20% of eye discs we still observed tumours and, in rare cases (8.5%), prp8 depletion in Ras^V12 tumours caused metastases to appear in developing larvae (Fig. S1L). For the majority of our experiments, we used a prp8^RNAi line that resulted in a higher frequency of complete rescue of Ras^V12-induced hyperplasia (prp8^{RNAi 18567GD}). Importantly, although we detected a significant percentage of tumours in whole larvae when this RNAi line was combined with Ras^V12, upon eye disc dissection in subsequent experiments, the vast majority of Ras^V12; prp8^RNAi tissues were significantly smaller than Ras^V12-expressing tissues, suggesting that our analysis of intact larvae in fact overestimates the existing tumour growth. We confirmed the effect of prp8 using the MARCM system to combine loss-of-function of prp8 (prp8^KG03188) with expression of Ras^V12. Our experiments revealed that, as in the prp8^RNAi, prp8 loss-of-function clones are smaller than control clones (compare Fig. 1J with 1L). Moreover, combining Ras^V12 with prp8 loss-of-function resulted in a phenotype similar to prp8 loss alone (compare Fig. 1M with 1L). Together, our results suggest that Prp8 influences Ras^V12-mediated hyperplastic growth and that, to a large extent, depletion of prp8 impairs Ras^V12-mediated hyperplasia.

To elucidate how Prp8 regulates tissue growth, we tested whether the eye disc hypoplasia phenotype obtained with depletion of Prp8 was because of cell proliferation defects. For this, we assessed the levels of phosphohistone-H3 (PH3), a marker of cells that are undergoing mitosis. When compared with controls (Fig. 2A), prp8^RNAi discs displayed reduced cell proliferation (Fig. 2C, quantified in Fig. 2I). In contrast, the number of PH3-positive cells seen in discs that expressed Ras^V12 (Fig. 2B) or the Ras^V12; prp8^RNAi combination (Fig. 2D) was similar to controls (Fig. 2A, quantified in Fig. 2I). We also analysed the G2/M cyclin, Cyclin B (CycB) (Fig. S2A-D) and found that, in both Ras^V12 and prp8^RNAi samples, CycB distribution was altered. CycB levels in the presumptive second mitotic wave were reduced in Ras^V12-expressing cells (Fig. S2B), whereas in prp8^RNAi the sharp boundary of CycB expression was lost and its expression was more uniform throughout the disc (Fig. S2C). In both cases, the morphogenetic furrow is absent. Interestingly, depleting prp8 in Ras^V12-expressing cells leads to a partial rescue of the CycB phenotype and the appearance of a rudimentary morphogenetic furrow (Fig. S2D). We also analysed the number of cells entering S phase by assessing 5-bromo-2′-deoxyuridine (BrdU) incorporation (Fig. S2E) and found that, as expected, Ras^V12 increased the number of BrdU-positive cells, which was suppressed when combined with prp8^RNAi. These results suggest that the eye phenotypes that are associated with prp8 depletion may be due to cell proliferation defects, which is consistent with a previous report that stated prp8 depletion causes a G2/M arrest (Andersen and Tapon, 2008). Depleting prp8 from Ras^V12-expressing tissues reduced entry into S phase but not progression through mitosis, as there was no significant difference in the number of PH3-positive cells between Ras^V12 and Ras^V12; prp8^RNAi tissues. Therefore, the effect of Prp8 on cell proliferation appears to be insufficient to explain why the loss of Prp8 blocks Ras^V12-induced hyperplasia.

These observations raise the possibility that Prp8 may affect cell survival. We assessed this using an antibody that recognises cleaved caspases (anti-Dcp1 antibody) and, therefore, reflects the overall level of cell death. Control eye discs have a relatively low level of cell death, which is mostly restricted to the area that juxtapose the morphogenetic furrow (Fig. 2E, quantified in Fig. 2J) (Rusconi et al., 2000). Expression of Ras^V12 alone did not significantly alter cell death levels (Fig. 2F,J). In contrast, prp8^RNAi led to a dramatic increase in caspase staining (Fig. 2G,J). Combining Ras^V12 with prp8 depletion resulted in a significant decrease in the levels of active caspase compared with prp8^RNAi alone, which suggests that Ras^V12 can rescue the cell autonomous defects that lead to cell death when prp8 is depleted (Fig. 2H,J). Nevertheless, depletion of prp8 in Ras^V12 tumours resulted in reduced hyperplastic growth (Fig. 2H). Together, these results suggest that loss of Prp8 function leads to defects in eye disc morphology due to a combination of decreased proliferation and increased cell death. However, when oncogenic Ras is present, Prp8 blocks hyperplasia, despite only having a modest effect on cell proliferation and cell death levels relative to Ras^V12 alone. To confirm that cell death alone cannot explain the phenotypes associated with loss of Prp8, we used the caspase inhibitor P35. Expression of P35 in controls did not result in any overt changes in tissue size or number of GFP-positive cells in developing eye discs (Fig. S2F). Consistent with our hypothesis, co-expression of prp8^RNAi and P35 was insufficient to fully rescue the hypoplasia phenotype that was seen when prp8 was lost, despite reducing the levels of activated caspases (compare Fig. S2G with Fig. 2G, quantified in Fig. S2H). Therefore, we conclude that although cell death contributes to the prp8^RNAi phenotype, there appear to be other processes that are regulated simultaneously (including cell proliferation) that contribute to regulation of developmental and Ras^V12-induced hyperplastic growth.

We next assessed whether Prp8 could regulate other processes that influence eye disc development. We first tested cell differentiation, as Prp8 has been associated with differentiation defects (Keightley et al., 2013; Wu et al., 2016) and, in Drosophila, eye disc development involves close coupling of cell proliferation, death and differentiation (Cagan, 2009). Moreover, crucial eye disc determinants are thought to be regulated via alternative splicing events (Fic et al., 2007; Roignant and Treisman, 2010), a process for which Prp8 function is crucial (Grainger and Beggs, 2005). To assess whether Prp8 regulates cell differentiation in eye discs, we stained for the photoreceptor differentiation marker Embryonic lethal abnormal vision (Elav, an RNA-binding protein that acts as a neuronal marker), the transcription factor Reversed polarity (Repo, which is restricted to glial cells) and the transcriptional co-activator Eyes absent (Eya, expressed in progenitors before differentiation) (Bonini et al., 1993; Lee and Jones, 2005; Soller and White, 2004). In controls, Elav and Repo were detected primarily in the region posterior to the morphogenetic furrow, which determines the ‘front’ of the cell differentiation wave (Fig. 3A,E). In prp8^RNAi discs, Elav staining was completely lost (Fig. 3C). Repo staining was still detectable in prp8-depleted discs, but the localisation and morphology of Repo-positive cells was dramatically changed (Fig. 3G). In contrast, whereas some Ras^V12-expressing cells maintained Elav and Repo expression, the majority were negative for these differentiation markers and, therefore, are presumably undifferentiated (Fig. 3B,F). Interestingly, when Ras^V12 was combined with prp8^RNAi, both Elav and Repo were expressed in the overgrown tissue and, when compared with Ras^V12 alone, these tissues appeared to have a higher percentage of Elav-positive and Repo-positive cells and resembled the control situation (Fig. 3D,H, quantified in 3M). With regard to Eya, its expression pattern was disrupted in Ras^V12- (Fig. 3J) and prp8^RNAi-expressing tissues (Fig. 3K), compared with controls (Fig. 3I) but, contrary to Elav and Repo, this was not rescued in the Ras^V12; prp8^RNAi combination (Fig. 3L), indicating that the effect of prp8 may be limited to specific differentiation markers.

We also assessed whether prp8 depletion affected signalling downstream of Ras^V12, as it has been previously reported that alterations in spliceosome genes cause dramatic changes in the splicing pattern of MAPK (Ashton-Beaucage et al., 2014). We analysed total and phospho-MAPK levels in vivo (Fig. S3A-J) and splicing changes in S2 cells (Fig. S3K and L) and found minor changes, under the conditions tested. Accordingly, expression of MAPK (rolled; rl) in prp8^RNAi-expressing tissues was unable to rescue prp8^RNAi-mediated hypoplasia, despite the fact that the eye disc morphology was partially rescued (Fig. S3M-P). Therefore, it is unlikely that the effect of prp8 in Ras^V12 hyperplasia is due to an effect on MAPK regulation.

Our data suggest that Prp8 can regulate cell differentiation in developing eye discs and that, in the presence of oncogenic Ras, absence of Prp8 function prevents uncontrolled growth, not by affecting MAPK signalling downstream of Ras but, at least in part, by causing premature or enhanced differentiation of cells. This would render the cells postmitotic, thereby limiting the growth of Ras^V12 hyperplastic tissues.

Our data suggest that prp8^RNAi can partially rescue the overall morphology of Ras^V12 hyperplastic eye discs (Fig. 1). Therefore, we hypothesised that Prp8 regulates cell processes and the components that are crucial for establishing and/or maintaining tissue morphology, such as polarity and the actin cytoskeleton (Pickup et al., 2002). To address this, we performed immunofluorescence staining for F-actin and the polarity protein Discs large (Dlg) in developing eye discs (Fig. 4). F-actin has a stereotypical organisation, with a prominent accumulation in the morphogenetic furrow and at the periphery of the posterior region of the disc (Fig. 4A). In Ras^V12 discs, the loss of overall tissue organisation and structure is reflected in the localisation of F-actin. F-actin is abnormally accumulated in large patches, which are adjacent to regions in which total F-actin appears to be significantly downregulated (Fig. 4B). Surprisingly, prp8^RNAi-associated eye disc hypoplasia was not accompanied by prominent changes in F-actin organisation beside an interruption of the F-actin signal in lateral membranes (Fig. 4C). In prp8-depleted eye discs, F-actin still accumulates at the periphery of the posterior region of the disc (Fig. 4C). When Ras^V12 and prp8^RNAi were combined (Fig. 4D), we observed that, despite a significant rescue of overall tissue morphology and tissue size, the F-actin pattern was still disorganised and did not resemble either wild-type (Fig. 4A) or prp8^RNAi phenotypes (Fig. 4C). Our results suggest that Prp8-mediated regulation of hyperplasia is mostly independent of a potential minor role of Prp8 in the modulation of the actin cytoskeleton. This hypothesis is consistent with our observation that F-actin structure is still disorganised when prp8 is depleted from Ras^V12-expressing hyperplastic tissue.

We next examined whether modulation of cell polarity could explain the rescue of tissue morphology that was seen when prp8^RNAi was combined with Ras^V12. Dlg is a basolateral polarity determinant that localises to septate junctions and is visualised in the lateral side of cells (Fig. 4E). Expression of Ras^V12 results in Dlg mislocalisation, which appears to spread throughout the cell membrane (Fig. 4F). Loss of prp8 leads to a dramatic increase in Dlg levels, loss of apico-basal polarity and epithelial organisation (Fig. 4G). When combined with Ras^V12, prp8^RNAi partially rescued Dlg localisation, which is more lateral than in Ras^V12 tumours alone, recapitulating the situation seen in the controls (compare Fig. 4H with Fig. 4E and Dlg localisation in cross-section images). Collectively, these results indicate that Prp8 regulates levels and/or localisation of Dlg and this modulation of cell polarity function may be, at least in part, responsible for the rescue of tissue morphology seen during hyperplasia.

Next, we determined whether the effect of Prp8 in the regulation of oncogene-induced hyperplasia was a tissue-specific function or whether it was limited to the regulation of Ras^V12-induced hyperplasia. For this, we generated alternative hyperplasia models in the developing eye using activated versions of the EGF receptor (Egfr^λtop) and Notch (N^ΔECD) (Pallavi et al., 2012; Queenan et al., 1997) (Fig. S4). Similar to Ras^V12, expression of EGFR (Fig. S4A) or N (Fig. S4C) led to overgrowth phenotypes. prp8 depletion significantly reduced the hyperplasia caused by both genes and, in some cases, produced a dramatic rescue of tissue organisation (Fig. S4B,D). Significantly, this suggests that prp8 regulates hyperplastic growth that is induced by different oncogenes. Moreover, we tested whether the effect of prp8 on Ras^V12-induced hyperplastic growth was a general role for the spliceosome by depleting the expression of alternative spliceosome components, such as Mfap1 (part of the spliceosome complex B; Fig. S4E,F), Prp38 (part of complex B; Fig. S4G,H) and Bx42 (part of complexes B, C and P; Fig. S4I,J). We found that all the spliceosome components efficiently suppressed Ras^V12-induced hyperplasia in the eye, suggesting that, at least in this tissue, the effect of prp8 is likely to be mediated by its role in the spliceosome (Fig. S4E-J, quantified in Fig. S4K).

We also assessed whether Prp8 regulated Ras^V12-induced hyperplasia in other tissues. To test this, we selected the adult gut as a model, as expression of Ras^V12 in intestinal stem cells (ISC) is known to cause tissue hyperplasia (Ragab et al., 2011; Jiang et al., 2011). The adult gut is maintained by ISCs, which can be identified by the expression of the N ligand Delta (Dl) and small nuclear size. ISCs give rise to enteroblasts (EB) that can differentiate into either enteroendocrine cells (ee) or absorptive enterocytes (EC) (Fig. 5A). Both ISCs and EBs express the transcription factor gene escargot (esg) and, in our experiments, we combined esg-Gal4 with a temperature-sensitive version of Gal80 to control gene expression (Fig. 5A) (Micchelli and Perrimon, 2006; Ohlstein and Spradling, 2006). When compared with controls (Fig. 5B), esg-Gal4-mediated expression of Ras^V12 in the adult gut led to an increase in the relative area of GFP-positive cells (ISCs and EBs) in the posterior midgut, 7 days after induction (Fig. 5C, quantified in 5G). The total cell number (Fig. 5F) and, in relative terms, the number of GFP-positive cells were similar to controls (Fig. S5I), which suggests that most of the GFP-positive area is because of large GFP-positive cells, which are likely to be ECs in which GFP expression now persists or progenitor cells that become enlarged. Accordingly, the number of small GFP-positive cells (ISCs and/or progenitors) is similar to control (Fig. S5J). Ras^V12-mediated hyperplasia influenced tissue architecture and, in some cases, caused GFP-positive cells to invade the gut lumen (Fig. 5C). In contrast, prp8 depletion resulted in a significant reduction in GFP-positive area (Fig. 5D,G), accompanied by a significant reduction in the relative number of GFP-positive cells (Fig. S5I). Similar to what we observed in the developing eye, co-expression of Ras^V12 and prp8^RNAi resulted in a significant reduction in the levels of Ras^V12-induced hyperplasia (Fig. 5E,G and Fig. S5I), which suggests that Prp8 influences Ras^V12-mediated growth in different tissues.

Next, we assessed whether prp8 depletion also affected cell proliferation, differentiation and death in the adult gut. Therefore, we stained adult guts with the ISC marker Dl (Fig. 5H and Fig. S5A-D), the ee marker Prospero (Pros) (Fig. 5I and Fig. S5E-H) and an antibody against active caspases (Fig. 6 and Fig. S6). Expression of Ras^V12 did not significantly change the ratio of Dl-positive cells in the gut, which suggests that although Ras^V12 promotes the proliferation of ISCs or uncommitted progenitors, these are likely still differentiating appropriately (Fig. 5H and Fig. S5B). In contrast, prp8^RNAi, alone or in combination with Ras^V12, led to a significant decrease in the ratio of Dl-positive cells relative to Ras^V12 alone, suggesting that it blocks Ras^V12 hyperplasia by blocking ISC and/or progenitor proliferation, promoting their differentiation or inducing apoptosis (Fig. 5H, Fig. S5C,D). With regard to ee cells, Ras^V12 expression had no effect on the proportion of ee cells in the adult gut, which suggests that the increased proliferation of ISCs and/or progenitors does not result in increased levels of differentiation to the ee lineage (Fig. 5I and Fig. S5F). In contrast, the proportion of ee cells was markedly reduced in prp8^RNAi-expressing guts (Fig. 5I, Fig. S5G and H). As the proportion of ee cells decreased when Prp8 was absent, these results suggest that it is unlikely that Prp8 loss blocks Ras^V12 tumours by promoting the differentiation of ee cells.

Finally, we analysed the levels of activated caspases in the posterior midgut at different times after gene induction to determine whether modulation of caspase-dependent cell death could explain the phenotypes that are seen with prp8^RNAi alone or in combination with Ras^V12. In controls (Fig. S6A), the levels of apoptosis 4 days after induction were low and the majority of caspase-positive cells were GFP-negative (therefore not esg-Gal4 positive), which is in agreement with the fact that most of the remodelling in the gut happens at the level of ECs and ee cells (Amcheslavsky et al., 2009; Jiang et al., 2009). In guts that expressed Ras^V12, there was an increase in the number of caspase-positive cells, which were found both in areas that contained GFP-positive cells and in adjacent GFP-negative areas (Fig. S6B). This suggests that Ras^V12 increases cell proliferation but also cell death, both in ISCs/progenitors and GFP-negative cells. This would explain why the total number of cells in the posterior midgut was similar between controls and Ras^V12 (Fig. 5F). When we depleted prp8 using the esg-Gal4 driver, we observed very low levels of caspase staining 4 days after induction (Fig. S6C). This was also the case when prp8 was depleted in combination with Ras^V12 expression (Fig. S6D). At later stages (7 days after transgene induction), the number of caspase-positive cells was low in all genotypes (data not shown). Therefore, we assessed an earlier time point, 2 days after induction, which yielded similar results (Fig. 6A-D). However, in this situation we observed caspase activity in GFP-positive cells of Ras^V12; prp8^RNAi animals, which indicated a potential loss of progenitor cells (Fig. 6D). In addition, we tested whether blocking apoptosis using P35 could rescue the reduction in the number of GFP-positive cells that was seen when prp8 was depleted. Expression of P35 alone reduced active caspase levels but had no overt effect on the adult gut 2 days after induction (Fig. 6E). Flies that expressed prp8^RNAi and P35 displayed phenotypes that were similar to prp8^RNAi alone (compare Fig. 6C with F), which indicates that blocking cell death is insufficient to rescue the prp8^RNAi phenotype. These results suggest that either Prp8 does not affect cell death in these conditions or that cell death occurs at an earlier time point and/or is caspase-independent. Depleting prp8 in the context of Ras^V12 hyperplasia was associated with cell death, but it was not overtly different from the caspase activation that was seen in Ras^V12-expressing tissues. Together, our data suggest that Prp8 reduces Ras^V12-mediated hyperplastic growth in different tissues, potentially using context-dependent tissue-specific mechanisms (i.e. modulation of proliferation and differentiation in the eye versus modulation of proliferation in the gut).

Next, we tested whether Prp8 regulates oncogene-induced hyperplastic growth and invasion in the developing wing disc. For this, we combined prp8^RNAi with expression of an oncogenic version of the receptor tyrosine kinase RET (RET^MEN2B; M955T point mutation) in the anterior-posterior boundary of the developing wing disc using patched-Gal4 (ptc-Gal4), as this system has previously been used to study invasion and metastasis and the receptor tyrosine kinase RET (Das et al., 2013; Read et al., 2005; Rudrapatna et al., 2012). Activating mutations in RET lead to the cancer syndrome multiple endocrine neoplasia type 2, which is associated with the occurrence of multiple tumours, including the highly metastatic medullary thyroid carcinoma. RET was used in this instance as ptc-Gal4-mediated expression of Ras^V12 led to early lethality, even when crosses were performed at 18°C (data not shown). As previously shown, when compared with controls (Fig. 7A), expression of oncogenic RET increased the ptc-Gal4-expressing area (Fig. 7B, quantified in Fig. 7E) (Das et al., 2013). In contrast, prp8 depletion resulted in disruption of the wing morphology and the appearance of GFP-positive cells outside of the ptc domain, indicative of potential cell invasion, which is in sharp contrast to what we observed in the eye disc and adult gut in the presence of oncogenic Ras (Fig. 7C,E). This phenotype was significantly enhanced when prp8 depletion was combined with RET expression, with a clear increase in the number of GFP-positive cells outside of the ptc-Gal4 domain (Fig. 7D,E). These results suggest that, in this tissue, prp8 may act as a tumour suppressor gene. Alternatively, the distinct outcome that is seen when removing prp8 function in both settings could be due to the activation of different signalling modules by Ras^V12 or oncogenic RET, that is, the function of Prp8 would be context-dependent. We also assessed the levels of caspase activation in this setting and found that both control (Fig. S7A) and RET-expressing wing discs exhibited low levels of caspase activation (Fig. S7B). In agreement with our eye disc data, depletion of prp8 led to moderate caspase activation in the ptc-expressing domain, which is consistent with previous reports (Fig. S7C) (Claudius et al., 2014). Surprisingly, when prp8^RNAi was combined with RET expression, the levels of caspase activation were dramatically increased, particularly in areas that were adjacent to wild-type cells (Fig. S7D). This suggests that the increased invasion of prp8^RNAi; RET^MEN2B cells may be driven by an increase in caspase activity. This hypothesis is in line with previous studies that implicate caspases in cell invasion via the activation of JNK signalling and subsequent activation of matrix metalloproteases that remodel the extracellular matrix and destroy the basement membrane, allowing cells to invade (Rudrapatna et al., 2013). We also assessed whether the effect of Prp8 in RET-induced hyperplasia could be a general feature of affecting spliceosome function. We depleted mfap1 (Fig. S7E,F), prp38 (Fig. S7G,H) or bx42 (Fig. S7I and J) in the ptc-Gal4 domain and found that, in marked contrast to prp8 depletion, removing the function of these spliceosome components significantly suppressed the RET^MEN2B phenotype (compare Fig. S7F,H,J with Fig. S7B). Collectively, our results identify Prp8 as a crucial regulator of hyperplastic growth. The precise function of Prp8 in tumours still requires further studies but appears to depend on the driving oncogene and may involve tissue-specific mechanisms, which may be dependent or independent of the function of Prp8 in the spliceosome.

In this report, using an in vivo RNAi screening approach, we identify Prp8 as a crucial regulator of oncogene-induced hyperplastic growth in Drosophila. Depletion of prp8 in the developing eye caused significant hypoplasia, suggesting that Prp8 is required for eye disc development. Depleting prp8 in Ras^V12-expressing tissues suppressed the Ras^V12 overgrowth phenotype and, in some cases, resulted in a rescue of the global tissue structure. Despite the fact that prp8^RNAi tissues display increased apoptosis and reduced proliferation, this alone is unlikely to account for all of the phenotypes that were observed when prp8 was depleted alone or in combination with Ras^V12. Indeed, simply blocking apoptosis in prp8-depleted tissues was insufficient to rescue the eye disc hypoplasia phenotype. This suggests that Prp8 has a complex pleiotropic effect during development and in the context of deregulated growth, in line with its role as a crucial spliceosome component (Grainger and Beggs, 2005; Shi, 2017). Accordingly, in tissues that express both Ras^V12 and prp8^RNAi, hyperplasia was decreased, despite the fact that apoptosis levels were significantly reduced compared with prp8^RNAi alone. Moreover, the levels of proliferation were not significantly affected. This raises an important question: how does prp8 depletion suppress hyperplasia in the absence of changes in proliferation and caspase-dependent cell death? One possible explanation is the apparent increase in cell differentiation that is seen in Ras^V12; prp8^RNAi tissues, which would preclude further proliferation, as cells would enter a postmitotic state (Silies et al., 2010). This is a potentially conserved strategy, as there are examples of differentiation acting as a tumour-suppressive mechanism in mammals, most notably in relation to p53 function and Notch signalling in skin cancer (Bieging et al., 2014; Guinea-Viniegra et al., 2012; Restivo et al., 2011). In several Drosophila tissues, Ras^V12 expression blocks differentiation, at least in part, by co-opting JNK activity (Brumby et al., 2011; Pastor-Pareja and Xu, 2013; Zeng et al., 2010). Loss of prp8 function in this context may affect the ability of Ras to block differentiation programs. Indeed, depletion of prp8 alone led to changes in cell differentiation in the eye disc and a specific loss of neuronal markers. Interestingly, a zygotic Prp8 mutant in zebrafish displayed extensive neuronal cell loss, as well as defects in the differentiation of myeloid cells (Keightley et al., 2013). This observation, combined with the suggestion that genes that are essential for eye disc development undergo alternative splicing, suggests that Prp8 has a conserved function in the regulation of developing neurons and is involved in cell fate decisions (Fic et al., 2007; Keightley et al., 2013; Roignant and Treisman, 2010). Alternatively, depletion of prp8 in Ras^V12-expressing tissues could influence their ability to outcompete wild-type cells or cause a cell cycle delay that promotes cell differentiation, as it has been previously shown that prp8 regulates G2/M transition (Andersen and Tapon, 2008).

Interestingly, we also observed a role for prp8 in the regulation of Ras^V12-induced hyperplasia in the adult gut. Depletion of prp8 in Ras^V12-expressing tissues significantly reduced hyperplasia and the cells no longer invade the gut lumen. Despite this, we failed to detect increased levels of apoptosis in Ras^V12; prp8^RNAi cells compared with cells expressing Ras^V12 alone, which is consistent with our observations in the eye disc. However, we did not detect cell differentiation changes in the adult gut, at least at the level of the ee population. Our results do not rule out an effect of prp8 in the regulation of EC differentiation, which could potentially be masked by the fact that the majority of the cell renewal in the gut involves ECs (Guo et al., 2016). However, we would favour the hypothesis that Prp8 affects mostly proliferation or stem cell dynamics, rather than directly affecting differentiation, as the number of GFP-positive cells and ISCs was reduced when prp8 was depleted. Our data indicate that the precise mechanism by which prp8 regulates oncogene-induced hyperplasia may differ according to the affected tissue. Consistent with this, despite presumably regulating splicing ubiquitously, mutations in the human orthologue of prp8, PRPF8, are generally only associated with diseases in specific organs, such as retinitis pigmentosa which affects the eye (Grainger and Beggs, 2005).

Our results also raise issues regarding the role of the spliceosome in the regulation of tumour growth. We found that depletion of spliceosome components resulted in suppression of Ras^V12-induced hyperplasia. This would suggest that prp8 regulates Ras^V12-mediated hyperplasia through its function within the spliceosome. This would be in agreement with the observation that mutation in other spliceosome components besides PRPF8 can cause retinitis pigmentosa in humans (Krausova and Stanek, 2017). However, in the context of oncogenic RET, we found that prp8 depletion enhances the RET phenotype, whereas depleting other spliceosome components suppresses it. This would suggest that Prp8 regulates RET^MEN2B-mediated hyperplasia and invasion in a spliceosome-independent manner, which requires further investigation. Notably, it has been previously shown that other spliceosome components have pleiotropic roles and additional non-splicing-related functions, as exemplified by Prp19 (Chanarat and Strasser, 2013). It will also be important to determine the molecular requirements for the function of prp8 in the regulation of hyperplasia, that is, which are the protein domains involved. Prp8 is considered to be an inactive DUB because of specific amino acid alterations in the MPN/JAB domain that serves as the catalytic domain in JAMM family DUBs (Clague et al., 2013; Grainger and Beggs, 2005; Komander et al., 2009; Pena et al., 2007). However, as the Prp8 MPN/JAB domain is essential for its function and can bind ubiquitin, it is possible that the role of Prp8 requires this domain and that it may involve the interaction of Prp8 with ubiquitylated proteins (Bellare et al., 2006). Detailed in vivo structure-function analysis will be required to fully elucidate this point but, interestingly, previous studies have suggested that ubiquitylation is important for the modulation of spliceosome protein-protein interactions (Das et al., 2017; Song et al., 2010). As for the downstream mechanisms involved, our data do not support a major role for modulation of the MAPK signalling cascade. Despite the fact that previous studies have shown that alterations in spliceosome genes cause dramatic changes in the splicing pattern of MAPK and affect Ras downstream signalling, we failed to uncover a major effect of Prp8 in this process and, accordingly, expressing MAPK in prp8^RNAi tissues was insufficient to rescue their hypoplasia (Ashton-Beaucage et al., 2014).

Prp8 appears to be required for Ras-driven hyperplasia and this role appears to be conserved with other oncogenes, such as activated Egfr and N. Therefore, our data extends recent observations in the context of human cancers, in which the spliceosome has been identified as a potential therapeutic target in Myc-driven tumours (Hsu et al., 2015). Our results suggest that Prp8 could be a specific target in tumours driven not only by Ras, Notch and EGFR but, potentially, by other receptor tyrosine kinases that signal through Ras and downstream pathways. However, as prp8 depletion enhanced the phenotype of RET^MEN2B in the developing wing, it will be important to define the mechanisms that control the function of Prp8 (and of the spliceosome) in the regulation of hyperplasia and tumour growth. In conclusion, clearly more work is needed to determine in which conditions inhibiting the spliceosome will be beneficial for cancer treatment.

MARCM experiments were performed using the y, w, hsFLP, UAS-GFP-nls; tub-Gal4, FRT42D tub-Gal80 MARCM maker stock. Below are the respective genotypes for the MARCM experiment: y, w, hsFLP, UAS-GFP-nls; tub-Gal4, FRT42D tub-Gal80/FRT42D blank (Fig. 1J); y, w, hsFLP, UAS-GFP-nls; tub-Gal4, FRT42D tub-Gal80/FRT42D blank; +/UAS-Ras^V12 (Fig. 1K); y, w, hsFLP, UAS-GFP-nls; tub-Gal4, FRT42D tub-Gal80/FRT42D prp8^KG03188 (Fig. 1L); and y, w, hsFLP, UAS-GFP-nls; tub-Gal4, FRT42D tub-Gal80/FRT42D prp8^KG03188; +/UAS-Ras^V12 (Fig. 1M).

For additional details of fly strains used, see supplementary Materials and Methods.

Third instar larval imaginal discs and adult guts were dissected in PBS and fixed for 20-30 min at room temperature (RT) in PBS containing 4% formaldehyde. After washing with 0.1% Triton X-100 (TX)/PBS, tissues were permeabilised with 0.1% or 0.3% TX/PBS for 30 min and, following additional washing steps with 0.1% TX/PBS (five times for 5 min), blocked for 30 min in blocking buffer [10% normal goat serum (NGS), 0.1% TX/PBS] and incubated overnight at 4°C with primary antibody diluted in blocking buffer. After washing with 0.1% TX/PBS and a blocking step, tissues were incubated for 1-4 h at RT with secondary antibodies. Samples were mounted in Vectashield (Vector Laboratories) after additional washing steps. For F-actin and DNA staining, tissues were incubated with TRITC-conjugated phalloidin (1:500; Sigma-Aldrich) or Hoechst 33342 (1:1000; Thermo Fisher Scientific), respectively, for 15 min during one of the final washing steps.

The following antibodies were used: rabbit anti-GFP (A-11122; Thermo Fisher Scientific; 1:1000), mouse anti-GFP (A-11120; Thermo Fisher Scientific; 1:1000) or chicken anti-GFP (ab13970; Abcam; 1:1000); rabbit anti-cleaved Caspase 3 (DCP-1) (9578; Cell Signaling Technology, 1:250); rabbit anti-phospho Histone H3 Ser10 (06-570; Merck; 1:1000); mouse anti-Dlg [4F3; Developmental Studies Hybridoma Bank (DSHB); 1:250]; rat anti-Elav (7E8A10; DSHB; 1:10); mouse anti-Repo (8D12; DSHB; 1:10); mouse anti-Delta (C594.9B; DSHB; 1:100); mouse anti-Prospero (MR1A; DSHB; 1:20); mouse anti-BrdU (G3G4; DSHB; 1:10); mouse anti-Cyclin B (F2F4; DSHB; 1:5); mouse anti-Eya (eya10H6; DSHB; 1:100); rabbit anti-MAPK (M5670; Sigma-Aldrich; 1:500) rabbit anti-phospho-MAPK (p44/42 MAPK ERK1/2–137F5; Cell Signaling Technology; 1:500). Secondary antibodies used were coupled to FITC (1:1000), Alexa Fluor 488, Alexa Fluor 568, Alexa Fluor 633, Alexa Fluor 647 (1:2000), Cy3 or Cy5 (1:200 or 1:500, depending on the primary antibody used) (Molecular Probes).

BrdU analysis was performed as previously reported (Chioda et al., 2010). Briefly, wandering third instar larvae were collected and eye imaginal discs were dissected in PBS. Discs were incubated in 1× PBS containing 20 μM BrdU (Sigma-Aldrich) for 30 min. Discs were fixed in PBS with 4% formaldehyde for 30 min at RT. DNA was denatured with 3 M HCl for 30 min. Samples were washed 3× with PBS containing 0.3% TX, followed by incubation in blocking buffer (10% NGS, 0.1% TX/PBS) for 1 h. Samples were incubated with mouse anti-BrdU antibody (G3G4; DSHB; 1:10) overnight at 4°C. Subsequent steps were performed as described in the immunostaining section.

Nuclei were segmented in 3D using a user-defined fluorescence threshold and watershed to separate nuclei in contact. Segmented nuclei were then classified into big or small cells based on their volume (the threshold was ∼90 μm³ but it was optimised for each image) and within each category they were classified as part of a clone or not based on their mean green intensity value, which was also optimised for each image. To avoid changes in fluorescence owing to the depth of the tissue, only nuclei in the half of the gut closest to the coverslip were considered. The code was implemented in Matlab and is available at github.com/juliafs93/CellCounter.

Drosophila S2 cells were grown in Drosophila Schneider's medium (Thermo Fisher Scientific) supplemented with 10% (v/v) foetal bovine serum, 50 μg/ml penicillin and 50 μg/ml streptomycin. Expression plasmids were transfected using Effectene transfection reagent (Qiagen) according to the manufacturer's instructions. Expression plasmids were generated using Gateway technology (Thermo Fisher Scientific). All vectors were verified by sequencing. S2 cells and the Ras^V12 cDNA were a kind gift from Nic Tapon (The Francis Crick Institute, London, UK).

dsRNAs were synthesised using the Megascript T7 kit (Thermo Fisher Scientific) according to the manufacturer's instructions. DNA templates for dsRNA synthesis were PCR amplified from genomic DNA or from plasmids that encoded the respective genes using primers that contained the 5′ T7 RNA polymerase-binding site sequence. dsRNA primers were designed using the DKFZ RNAi design tool (www.dkfz.de/signaling/e-rnai3). The following primers were used: lacZ (forward, TTGCCGGGAAGCTAGAGTAA; reverse, GCCTTCCTGTTTTTGCTCAC) and prp8 (forward, CGAGTCTGGCTGTTCTTTATGC; reverse, ATGTACGGACCGTCCTTTAAGTAG). After seeding, S2 cells were incubated with 15-20 μg dsRNA for 1 h in serum-free medium, before complete medium was added. Cells were lysed and processed for further analysis 72 h after dsRNA treatment.

Total RNAi was extracted from S2 cells using the QIAshredder and RNeasy kits (Qiagen) according to the manufacturer's protocols. RNA purity and concentration were assessed using a Nanodrop One UV-Vis spectrophotometer (Thermo Fisher Scientific). cDNA was synthesised using the QuantiTect Reverse Transcription kit (Qiagen) following the manufacturer's instructions. RT-PCR analysis was performed using 1 μl of cDNA per PCR reaction and the following primers: MAPK full-length (forward, CGCCGTCGATTTTGATAAATCATATTTACGC; reverse, AGGCGCATTGTCTGGTTGTCGT) (Ashton-Beaucage et al., 2014). RT-PCR products were run in 2% UltraPure agarose (Thermo Fisher Scientific) gels and imaged in an Amersham Imager 600 (GE Healthcare).

For in vivo RNAi screen studies, whole larva images were acquired using a Zeiss SteREO Lumar V12 stereomicroscope. Confocal images were acquired at ×20 or ×40 magnification using a Zeiss LSM710 confocal microscope or an Olympus FV1000 confocal microscope equipped with 20×/0.85 oil and 40×/1.35 oil iris objectives. All images were taken as z-stacks of 1 µm sections in eye and wing imaginal discs and in the posterior midgut region immediately anterior to the hindgut (R4-R5 region). For cross-sections and 3D reconstructions, images were acquired as z-stacks of optimal sections. Image processing, analysis and 3D reconstruction were performed with ImageJ and Imaris XT8.0.

3D reconstruction images were quantified using Imaris 8.4 and ImageJ and quantifications were performed throughout the volume of the reconstruction. GFP area was calculated in 3D volume using Imaris 8.4 or ImageJ. Quantification of cell numbers, Delta-positive and Prospero-positive cells was performed manually in ImageJ. Statistical analyses were performed in Microsoft Excel or GraphPad Prism. Significance (P) values were determined using one-way ANOVA analysis (with Tukey's post test for multiple comparisons or Kruskal–Wallis and Dunn post tests for non-parametric t-tests). Unless otherwise stated, data is represented as mean±s.e.m.

We thank the Transgenic RNAi Project (TRiP) at Harvard Medical School [National Institutes of Health (NIH)/National Institute of General Medical Sciences, R01-GM084947], the Kyoto Stock Center (DGRC) and the Vienna Drosophila Resource Center for providing transgenic RNAi fly stocks used in this study. Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) were used in this study. We thank Ross Cagan, Sarah Bray, Nic Tapon, Barry Thompson and Pascal Meier for fly stocks. We thank Pascal Meier for compiling the first iteration of the Drosophila in vivo DUB RNAi library and for assistance in assembling the current library. The antibodies 4F3, 8D12, 7E8A10, C594.9B, MR1A, G3G4, F2F4 and eya10H6, developed by C. Goodman (4F3 and 8D12), G. Rubin, S. Artavanis-Tsakonas, C. Doe, S. J. Kaufman, P. H. O'Farrell, and S. Benzer and N. M. Bonini (eya10H6), were obtained from the Developmental Studies Hybridoma Bank, which was created by the National Institute of Child Health and Human Development of the NIH and maintained at The University of Iowa, Department of Biology. We thank Linda Hammond for assistance with microscopy. We thank Nick Brown for support and access to microscopy facilities. We thank members of the Ribeiro lab for helpful discussions and comments on the manuscript. We thank Sarah Martin, Susana Godinho, Tyson Sharp, Prabhakar Rajan, Pascal Meier and Nic Tapon for critical reading of the manuscript.