Hippo signaling promotes Ets21c-dependent apical cell extrusion in the Drosophila wing disc

The Hippo (Hpo) pathway is conserved from Drosophila to mammals (Dong et al., 2007; Pan, 2010). Upstream key components of this pathway, including Hpo and Warts (Wts), form a core kinase cascade to regulate the transcriptional co-regulator Yorkie (Yki) (Huang et al., 2005; Kim et al., 2015; Praskova et al., 2008; Udan et al., 2003; Wu et al., 2003). When the Hpo pathway is inactive, dephosphorylated Yki is translocated into the nucleus and interacts with the DNA-binding transcription factor Scalloped (Sd) (Zhang et al., 2008). The Yki-Sd complex induces the expression of a wide range of target genes, including the Cyclin E (CycE) and Myc cell-cycle regulators, bantam (ban) microRNA and Drosophila inhibitor of apoptosis protein 1 (Diap1) (Pan, 2010). Numerous studies have indicated that the Hpo pathway functions as a tumor suppressor. YAP1 (yes-associated protein 1, a Yki ortholog) is hyperactivated in some human cancers, including lung cancer (Wang et al., 2010) and uveal melanoma (Feng et al., 2014; Yu et al., 2014). However, YAP is also downregulated in individuals with multiple myeloma and in those with acute myeloid leukemia (Cottini et al., 2014).

The Hpo pathway is essential for epithelial tissue development and homeostasis by regulating cell survival (Tapon et al., 2002), cell proliferation (Halder and Johnson, 2011), organ size (such as that of the Drosophila eye and thorax and the mouse liver) (Dong et al., 2007) and cell-cell adhesion (Schroeder and Halder, 2012). It has been reported that yki mutant clones grow poorly in the third instar larval wing disc (Qing et al., 2014). In addition to roles in apoptosis and cell proliferation, the Hpo pathway has recently been implicated in cell extrusion and invasion, which may help to explain the poor survival of yki mutant clones. Expression of ykiRNAi along the anterior-posterior (AP) boundary of Drosophila wing discs induces cell migration across the AP boundary and basal cell extrusion (BCE) (Ma et al., 2017). However, except for apoptosis-induced BCE, whether cell extrusion causes the abnormal Hpo-Yki signaling-induced poor recovery rate is unclear.

In both invertebrate and vertebrate models, cell extrusion plays an important role in epithelial homeostasis and development as well as in cancer cell metastasis (Eisenhoffer et al., 2012; Gu and Rosenblatt, 2012; Gudipaty and Rosenblatt, 2017; Marinari et al., 2012; Ohsawa et al., 2018). Moreover, extruded tumor cells have invasive abilities (Dunn et al., 2018; Pagliarini and Xu, 2003; Shen et al., 2014; Vaughen and Igaki, 2016). Here, we tested the hypothesis that the Hpo pathway maintains tissue homeostasis by suppressing cell extrusion. We found that Hpo pathway activation, by repressing yki or expressing hpo or wts, induced both apical cell extrusion (ACE) and BCE. Furthermore, ban expression could suppress BCE but not ACE. We further demonstrated that JNK signaling was necessary and sufficient to induce ACE downstream of Yki. Mechanistically, we present genetic evidence that ACE induced by the Hpo-JNK pathway is mediated by Ets21c, a member of the ETS-domain transcription factor family.

This study was designed to investigate whether poor yki mutant clone growth is due to the effect of cell extrusion, in addition to cell death and proliferation constriction. We used mosaic analysis with a repressible cell marker (MARCM) (Lee and Luo, 2001) to generate GFP-marked yki mutant clones and monitored their cellular behavior. Compared with the control clones (Fig. S1A,F), wing pouch yki mutant clones were rare and small (Fig. S1B,F). Next, we generated UAS-ykiRNAi and UAS-GFP co-expressing clones using flp-out technology (Harrison and Perrimon, 1993). After 72 h of growth, ykiRNAi clonal cells showed poor survival (Fig. S1C,D,G). Given that Yki is a cell death suppressor (Huang et al., 2005), the loss of clones may be caused by apoptosis. To address this possibility, p35, encoding a baculovirus pan-caspase inhibitor (Hay et al., 1995), was co-expressed with ykiRNAi in the clonal cells to suppress potential apoptosis. However, even when apoptosis was suppressed by p35 expression, ykiRNAi clones were still very small and rare, especially in the wing pouch (Fig. S1E,G). Therefore, the low recovery rate of clones lacking Yki activity was not due to apoptosis. These results support the possibility of cell extrusion in yki mutant clones.

To check whether cells with downregulated yki are extruded from the wing disc, vertical wing disc sections were examined by y-z scanning on a confocal microscope. The apical side of disc proper (DP) cells faces the lumen of the wing sac, and their basal side is the external surface, which contacts the hemolymph. We used Discs large 1 (Dlg1) (Knust and Bossinger, 2002) as an apical maker of the epithelia during y-z scanning. Compared with wild-type clones (Fig. 1A), yki mutant clones were rare and invaded into the lumen (Fig. 1B, arrow). ykiRNAi clones were rare and were extruded either basally or apically (Fig. 1C,D). To overcome the low recovery rates during clonal analysis, UAS-ykiRNAi was expressed under the control of a pan-disc-specific driver C765-Gal4 (Fig. S2A,A′). There were two stripe regions close to the wing hinge folds with greatly reduced Dlg1 staining in the apical wing disc confocal sections when we decreased yki expression (two white circles in Fig. S2A′). We inspected the wing disc vertical section and found cells undergoing ACE when yki was silenced, and most of the apically extruded cells were located in the two regions with reduced Dlg1 staining (Fig. 1E,F, arrow). At the same time, BCE was also evident (Fig. 1E,F, arrowhead). ACE and BCE were observed in over 90% of the wing discs (Fig. S3A,C). We confirmed that the same ACE and BCE phenotypes were observed using another, independent, UAS-ykiRNAi line (Fig. S2B).

Hpo and Wts suppress Yki in the Drosophila wing disc (Huang et al., 2005). To confirm that activated Hpo signaling regulates cell extrusion, we inspected the vertical sections of wing discs expressing hpo and wts. Consistently, both ACE and BCE were observed in more than 95% of UAS-hpo-expressing wing discs (Fig. 1G; Fig. S3). Both ACE and BCE were observed in the UAS-wts-expressing wing disc pouch at high frequency (76.5% and 100% for ACE and BCE, respectively; n=34) (Fig. 1H; Fig. S3). These results suggest that the activated Hpo pathway induces both ACE and BCE in Drosophila wing discs. The absence of yki expression induces apoptosis (Huang et al., 2005), and this was confirmed by anti-cleaved-caspase-3 staining (Fig. S4A). Vertical sections showed that most of the dying cells were located either in the lumen or at the basal side of the epithelia (Fig. S4A). To examine whether ACE and BCE are induced as a side effect of apoptosis, we co-expressed p35 and ykiRNAi. Both ACE and BCE still occurred (Fig. S4B,C) at high frequency in clones co-expressing p35 and ykiRNAi (74.2%, n=31) (Fig. S3A,C). These data indicate that cell extrusion induced by Hpo pathway activation is independent of apoptosis.

We next tested whether canonical Hpo-Yki target genes mediate the observed cell extrusion. We checked four targets downstream of Yki: Cyclin E (CycE), a cell cycle regulator (Tapon et al., 2002); Diap1, an apoptosis inhibitor (Wu et al., 2003); ban, a microRNA that promotes growth and inhibits cell death (Nolo et al., 2006; Thompson and Cohen, 2006); and Myc, a crucial cellular growth effector (Neto-Silva et al., 2010). Co-expressing ban and ykiRNAi still induced ACE (Fig. 2A,B) at a high frequency (70%, n=30) (Fig. S3A,C), but the frequency of BCE in wing discs was reduced (16.7%, n=30). Co-expressing Diap1 (Fig. 2C,D), Myc (Fig. 2E,F) or CycE (Fig. 2G,H) with ykiRNAi had little effect on ACE and BCE frequencies (Fig. S3A,C). These results suggest that BCE induced by silencing yki is largely dependent on ban, but ACE induced by silencing yki is not dependent on the canonical Hpo-Yki targets examined.

To further investigate the mechanism of ACE induced by ykiRNAi, we performed RNA-seq analysis in ykiRNAi-expressing wing discs. In C765>ykiRNAi versus C765> samples, we identified 833 transcripts with more than a 1.2-fold change in expression [false discovery rate (FDR) less than 0.05], including 500 upregulated genes and 333 downregulated genes (Fig. 3A,B; Table S1). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of the 833 transcripts revealed the enrichment of many pathways, including Hpo (Table S2) and JNK pathways (Fig. 3C). In the Drosophila wing disc, ACE is associated with tumor growth and invasion (Dunn et al., 2018; Tamori et al., 2016; Vaughen and Igaki, 2016), indicating that genes involved in these processes are potentially related to the ACE observed. By comparing our RNA-seq data with that produced through analyzing invasive tumors in Drosophila wing discs (Atkins et al., 2016), we found 90 genes that overlapped (Fig. 3E; Table S1). Gene ontology (GO) analysis suggested that these 90 genes were enriched in carbohydrate metabolic processes, salivary gland cell autophagic cell death, cellular response to gamma radiation, epithelial cell development and other developmental processes (Fig. 3F). Enrichment of carbohydrate metabolic processes indicates the need for macromolecule biosynthesis to support tumor growth (Külshammer et al., 2015). Enrichment in salivary gland cell autophagic cell death and cellular response to gamma radiation terms reflects increased apoptosis. To further verify our RNA-seq results, yki, Ets21c and another four genes [Kmn1, puckered (puc), rpr and hth] were selected and their mRNA expression levels tested using RT-qPCR (reverse transcription quantitative real-time PCR). Consistent with the transcriptomic data, RT-qPCR results indicate that in the ykiRNAi wing discs hth, Kmn1 and yki were downregulated, whereas puc, rpr and Ets21c were upregulated (Fig. 3D; Table S3). We noticed that the expression levels of several JNK pathway components were altered in the ykiRNAi wing discs (Fig. 3C). These results indicate that JNK signaling may be a key mediator for yki-dependent ACE and BCE.

Our transcriptome data indicated increased expression of the JNK targets puc and Matrix metalloproteinase 1 (Mmp1) (Fig. 3C; Martín-Blanco et al., 1998; Xue et al., 2007). To confirm activation of the JNK pathway by yki suppression, we examined the expression levels of a puc-LacZ reporter and Mmp1 protein levels. Indeed, both puc and Mmp1 were upregulated in ykiRNAi wing discs (Fig. S5). To test whether JNK signaling is required for ykiRNAi-induced ACE and BCE, we inhibited JNK signaling by co-expressing a dominant-negative form of JNK (bsk^DN) (Weber et al., 2000). Both ACE and BCE were largely blocked by bsk^DN in ykiRNAi wing discs (Fig. 4A,B). The percentage of wing discs with ACE and BCE was reduced to 10% (n=30; Fig. S3A,C). Activation of the JNK pathway by expressing hep^CA (a constitutively active form of hep), which encodes a JNK kinase (Glise et al., 1995), induced both ACE and BCE (Fig. 4C; Fig. S3B,D). To rule out the effect of apoptosis, we suppressed cell death by co-expressing p35 with hep^CA. ACE and BCE were still observed (Fig. 4D) with high frequencies (55.6 and 100% for ACE and BCE, respectively; Fig. S3B,D). Taken together, these results suggest that activated JNK signaling is required and sufficient to induce ACE and BCE in the Drosophila wing epithelia.

To identify genes downstream of Yki-JNK that mediate ACE, we performed a targeted genetic screen by either overexpressing or knocking down the expression of candidate genes identified in the transcriptome analysis. We found that the Ets21c ETS-domain transcription factor, encoded by an ortholog of the proto-oncogenes FLI1 and ERG, was a regulator of ACE. Similar to the results in ykiRNAi wing discs, expression of Ets21c was highly upregulated in wing discs expressing hep^CA (Fig. S6). UAS-Ets21c^HA (HA-tagged Ets21C) clones were eliminated from the wing disc pouch (Fig. 5A; Fig. S1G), phenocopying the ykiRNAi clones (Fig. S1D,G). RT-qPCR showed that Ets21c transcription was upregulated in ykiRNAi wing discs (Fig. 3D), which is consistent with our RNA-seq data (Fig. 3B). Ets21c expression was also examined using a GFP-trap line (Khan et al., 2017). Consistently, compared with the control (Fig. 5B,B′), ykiRNAi expression increased Ets21c-GFP levels (Fig. 5C,C′). These results suggest that Ets21c acts downstream of yki. The y-z views showed that the Ets21c-GFP level was relatively higher in cells undergoing apical extrusion (Fig. 5C,C′, yellow arrow) than that in non-extruded cells (Fig. 5C,C′). We also observed higher levels of Ets21c-GFP in apically extruding cells induced by hep^CA (Fig. S6B, yellow arrow). These results suggest that Ets21c might act downstream of yki and JNK to regulate ACE. To further confirm that Ets21c functions in Yki-JNK-mediated cell extrusion, Ets21cRNAi was co-expressed with ykiRNAi. Remarkably, suppressing Ets21c efficiently rescued the ACE but not the BCE induced by ykiRNAi (Fig. 5D,E, arrowheads). Co-expression of Ets21cRNAi with ykiRNAi resulted in 18.2% and 60.6% of wing discs with ACE and BCE, respectively (Fig. S3A,C; n=33). These results suggest that Ets21c upregulation mediates ACE induced by silenced yki. We then expressed Ets21c^HA to examine whether ACE can be induced by Ets21c. Notably, expression of UAS-Ets21c^HA was sufficient to induce ACE in 51.5% of wing discs (n=33) (Fig. 5F; Fig. S3B). We inspected anti-cleaved-caspase-3 staining in wing discs expressing activated Ets21c and did not find apparent cell death at the apical side of the epithelia (Fig. S7), which implies that there was no correlation between Ets21c^HA-induced ACE and cell apoptosis. These results further confirm that Ets21c mediates Hpo-Yki-JNK-dependent ACE.

Cell extrusion plays an important role in epithelial homeostasis and development as well as in cancer cell metastasis (Eisenhoffer et al., 2012; Gu and Rosenblatt, 2012; Gudipaty and Rosenblatt, 2017; Marinari et al., 2012; Ohsawa et al., 2018). In Drosophila epithelia, BCE occurs during dorsal closure and epithelial-mesenchymal transition (EMT) as well as in apoptosis (Andrade and Rosenblatt, 2011; Riesgo-Escovar et al., 1996; Riesgo-Escovar and Hafen, 1997; Pocha and Montell, 2014), whereas ACE occurs in tumor invasion and extrusion of apoptotic enterocytes in the Drosophila adult midgut (Dunn et al., 2018; Tamori et al., 2016; Vaughen and Igaki, 2016; Martin et al., 2018). However, the molecular mechanisms underlying BCE and ACE in Drosophila epithelia are not well understood. Our results demonstrate that inappropriate Hpo-Yki-JNK signaling induces ACE and BCE in Drosophila wing disc epithelia (Fig. 6). We also show that in the wing disc epithelia, ban acts downstream of Yki to regulate BCE (Fig. 6B) and Ets21c acts downstream of JNK to regulate ACE (Fig. 6C).

The Hpo pathway regulates tissue growth in Drosophila (Halder and Johnson, 2011). It has been reported that yki^B5 mutant clones grow poorly in the wing and eye discs (Huang et al., 2005; Koontz et al., 2013; Qing et al., 2014). Consistent with these reports, our results showed small ykiRNAi and yki^B5 mutant clones (Fig. 1B,D; Fig. S1B,D). Cells with depleted yki expression are extruded either apically or basally from the epithelia independently of apoptosis (Fig. 1F, Fig. 2D; Fig. S4C), indicating that cell extrusion is one explanation for the low recovery rate of Yki-depleted clones. In the Drosophila wing disc, overexpression of hpo by MS1096-Gal4 and nub-Gal4 dramatically decreases adult wing size (Koontz et al., 2013). Meanwhile, overexpression of wts by nub-Gal4 also reduces the wing size (Deng et al., 2015). When we activated hpo and wts expression, using C765-Gal4, cells were intensively extruded to the lumen and the basal side of the epithelia (Fig. 1G,H). Therefore, in addition to the proliferation defect, cell extrusion is one reason for the reduced tissue size induced by Hpo pathway activation. Diap1 levels are decreased in the small yki mutant clones (Qing et al., 2014), and co-expression of Diap1 and ykiRNAi could not block ACE or BCE (Fig. 2D). These results indicate that Diap1 does not regulate cell extrusion downstream of Yki. Hpo, wts mutant and yki overexpression in clones confers on cells supercompetitive properties that can lead to elimination of surrounding wild-type cells (Tyler et al., 2007; Ziosi et al., 2010). This suggests that cell competition could promote elimination of Yki-depleted clones. In our results, however, elimination of Yki-depleted cells could be triggered autonomously, even when Yki was depleted in the whole wing pouch. Cells expressing low levels of Myc are extruded basally through cell competition (Lampaya and Basler, 2002). Expressing Myc alone is not sufficient to prevent the elimination of yki mutant cells (Ziosi et al., 2010). Consistently, overexpression of Myc could not block BCE induced by silenced yki (Fig. 2F), indicating that other factors regulate BCE downstream of Yki. We found that ban could inhibit ykiRNAi-mediated BCE but not ACE (Fig. 2B). It is known that activated Hpo plays a role in cell migration (Ma et al., 2017). Cells with depleted yki expression migrated across the AP boundary and were extruded basally, and this cell migration was suppressed by ban. Our results show that ban can suppress ykiRNAi-induced BCE in the Drosophila wing disc but does not regulate ykiRNAi-induced ACE.

In vertebrate epithelia, cells dying through apoptosis or crowding stress are extruded apically into the lumen (Eisenhoffer et al., 2012; Rosenblatt et al., 2001; Slattum and Rosenblatt, 2014). The S1P-S1P2 pathway regulates both apoptosis-induced and apoptosis-independent ACE (Eisenhoffer et al., 2012; Gu et al., 2011; Kuipers et al., 2014). The oncogenic KRAS^V12G mutation in MDCK (Madin-Darby canine kidney) epithelial cell monolayers can downregulate both S1P (sphingosine 1-phosphate) and its receptor S1P2 (also known as S1PR2) to promote basal extrusion (Slattum et al., 2014). In Drosophila epithelia, the direction of apoptotic cell extrusion is reversed with most apoptotic cells undergoing BCE (Casas-Tintó et al., 2015; Dekanty et al., 2012). Apoptosis-induced BCE is regulated by JNK signaling (Andrade and Rosenblatt, 2011). One exception is in Drosophila adult midgut, where enterocytes are lost through apical extrusion (Martin et al., 2018). However, we know little about the mechanism of ACE in Drosophila epithelia.

In Drosophila epithelia, apical extrusion of scrib mutant cells is mediated by the Slit-Robo2-Ena complex, reduced E-cadherin and elevated Sqh levels. In normal cells, slit, robo2 and ena overexpression only results in BCE when cell death is blocked (Vaughen and Igaki, 2016). More importantly, in our RNA-seq results, expression of slit, robo2 and ena were not changed in the Yki-depleted Drosophila wing disc, which means Slit-Robo2-Ena does not associate with the Hpo pathway to regulate ACE. scrib mutant cells activate Jak-Stat signaling and undergo ACE in the ‘tumor hotspot’ located in the dorsal hinge region of the Drosophila wing disc (Tamori et al., 2016). Moreover, ACE can precede M6-deficient Ras^V12 tumor invasion following elevation of Cno-RhoA-MyoII (Dunn et al., 2018). Our RNA-seq results revealed that the expression of Jak-Stat pathway genes and RhoA (Rho1) were not altered, indicating that ACE can be regulated by novel signaling pathways.

In Drosophila, the JNK signaling pathway is essential for regulating cell extrusion in phenomena including wound healing, cell competition, apoptosis and dorsal closure (Ohsawa et al., 2018). JNK signaling mediates the role of Dpp and its downstream targets in cell survival regulation in the Drosophila wing (Adachi-Yamada et al., 1999). Cell extrusion and retraction toward the basal side of the wing epithelia induced by the lack of Dpp activity is independent of JNK (Shen and Dahmann, 2005; Shen et al., 2010). In one case of ectopic fold formation at the AP boundary of the Drosophila wing, loss of Omb activates both Yki and JNK signaling. In this case, JNK signaling induces the AP fold by cell shortening, and Yki signaling suppresses JNK-dependent apoptosis in the folded cells (Liu et al., 2016). During cell competition induced by Myc manipulation, JNK-dependent apoptosis mediates the death of ‘loser’ cells and their extrusion to the basal side of the epithelia. Apoptosis-induced BCE can be blocked by Diap1, which suppresses JNK-dependent apoptosis (Lampaya and Basler, 2002). Taken together, these results show that JNK signaling mediates or interacts with Yki signaling in a cellular context-dependent manner during the regulation of wing epithelial morphogenesis and apoptosis.

JNK is required for the migration of Csk mutant cells across the AP boundary and for their extrusion to the basal side of the epithelia (Vidal et al., 2006). puc encodes a JNK-specific phosphatase that provides feedback inhibition to specifically repress JNK activity. Expression of puc can prevent ptc>CskRNAi cells from spreading at the AP boundary (Vidal et al., 2006). JNK activity is also needed for ykiRNAi cells to invade across the wing disc AP boundary, and co-expression of bsk^DN and ykiRNAi blocks this invasion (Ma et al., 2017). Consistent with the role of JNK in BCE regulation, blocking JNK signaling by bsk^DN expression prevented ykiRNAi cells from being extruded to the basal side of the wing epithelia (Fig. 4B). More importantly, we found that JNK activation by hep^CA was sufficient to induce BCE (Fig. 4C), independently of apoptosis (Fig. 4D). Furthermore, few JNK targets have been shown to regulate cell migration and BCE. An exception to this are caspases that function downstream of JNK, which can promote cell migration when activated at a mild level (Fan et al., 2020; Rudrapatna et al., 2013).

In Drosophila eye imaginal discs, elevated JNK signaling in scrib mutant cells regulates both ACE and BCE. JNK and Robo2-Ena constitute a positive-feedback loop that promotes the apical and basal extrusion of scrib mutant cells through E-cadherin reduction. Meanwhile, in normal cells, p35 upregulation when Robo2 and Ena are overexpressed only induces BCE (Vaughen and Igaki, 2016). Our results showed that blocking JNK signaling could suppress ACE induced by silenced yki (Fig. 4B). Meanwhile, activation of JNK by hep^CA was sufficient to induce the extrusion of cells into the lumen (Fig. 4C). Cell debris may be trapped in the disc lumen when overexpressing hep^CA (Herrera et al., 2013). We suppressed apoptosis by co-expressing p35, to confirm that the ACE observed was independent of cell death (Fig. 4D). Taken together, these results indicate that there are additional regulators downstream of JNK to mediate ACE in normal cells.

E-twenty-six (ETS) family transcription factors have conserved functions in metazoans (Hollenhorst et al., 2011; Sharrocks et al., 1998). These include apoptosis regulation (Sevilla et al., 1999), cell differentiation promotion (Taylor et al., 1997), cell fate regulation (Miley et al., 2004) and cellular senescence (Ohtani et al., 2001). Ets21c encodes a member of the ETS-domain transcription factor family and is the ortholog of the human proto-oncogenes FLI1 and ERG (Mundorf et al., 2019). In Drosophila eye imaginal discs, 30-fold increased Ets21c expression is induced by Ras^V12 and eiger, an activator of JNK (Toggweiler et al., 2016). In the Drosophila adult midgut, Ets21c expression is increased when JNK is activated by the JNK kinase hep (Mundorf et al., 2019). Ets21c can also promote tumor growth downstream of the JNK pathway (Külshammer et al., 2015; Toggweiler et al., 2016). Our results have confirmed that Ets21c functions downstream of JNK. Indeed, we showed that Ets21c-GFP level was elevated following JNK activation (Fig. S6A). Expression of Ets21c^HA was sufficient to induce ACE (Fig. 5F) and silencing of Ets21c was sufficient to rescue ykiRNAi-induced ACE in the wing discs (Fig. 5E). However, the mechanism through which yki regulates JNK-Ets21c remains to be determined.

In Drosophila imaginal discs, ACE promotes polarity-impaired cells to grow into tumors (Tamori et al., 2016; Vaughen and Igaki, 2016). Therefore, it is possible that Ets21c can promote Hpo-Yki-JNK-related tumorigenesis by facilitating ACE in Drosophila. It is difficult to infer a putative pro-tumoral function of Et21c in mammals through its effect on ACE. ACE is rather associated with the elimination of tumor cells in mammals (Kajita and Fujita, 2015; Kon et al., 2017), whereas BCE is traditionally associated with higher invasive capacity. Yki/YAP gain-of-function promotes cancer cell invasion in non-small-cell lung cancer (Wang et al., 2010), neoplastic transformation (Yang et al., 2013), uveal melanoma (Yu et al., 2014) and pancreatic cancer (Yang et al., 2015). Additionally, Yki/YAP loss of function helps tumor cells to escape from apoptosis in hematologic malignancies, including multiple myeloma, lymphoma and leukemia (Cottini et al., 2014). Consistent with the latter role, Yki suppressed cell extrusion from the Drosophila wing epithelia by suppressing Ets21c (Fig. 5E). Therefore, the role of Ets21c in Hpo-Yki-related tumor models should be further examined.

The following transgenes were used in this study: C765-Gal4 (Nellen et al., 1996), UAS-bsk^DN (Weber et al., 2000), UAS-ban (Ying and Padgett, 2012), puc-lacZ (Martín-Blanco et al., 1998), FRT42D-yki^B5 and FRT42D (gifts from Z. H. Li, Capital Normal University, Beijing, China), UAS-Diap1 (gift from A. Bergmann, University of Massachusetts Medical School, Worcester, MA, USA), UAS-hippo and UAS-wts (gifts from L. Xue, Tongji University, Shanghai, China), UAS-GFP [Bloomington Drosophila Stock Center (BL), #4775], UAS-CycE (BL, #30725), UAS-Myc (BL, #9674), UAS-p35 (BL, #5073), UAS-hep^CA (BL, #38639), Ets21c-GFP (BL, #38639), UAS-Ets21cRNAi (Vienna Drosophila Resource Center, #106153), UAS-Ets21c^HA (FlyORF, #F000624), UAS-ykiRNAi [Tsinghua University (THU), #0579] and UAS-ykiRNAi (THU, #3074).

Larvae were raised at 25°C. For efficient expression of RNAi and UAS transgenes driven by C765-Gal4, larvae were raised at 29°C. To generate UAS-ykiRNAi and UAS-Ets21c clones, larvae of genotype y w hs-Flp/+; act5c>CD2>Gal4, UAS-GFP/+; UAS-ykiRNAi/+ and y w hs-Flp/+; act5c>CD2>Gal4, UAS-GFP/+; UAS-Ets21c/+ were subjected to heat shock at 35°C for 30 min. To generate yki^B5 MARCM clones, larvae of genotype y w hs-flp, tub-Gal4, UAS-GFP/+; FRT42D, Tub-Gal80/FRT42D yki^B5 were subjected to heat shock for 1 h at 37.5°C. Larval genotypes for the data shown in every figure are listed in Table S5.

Drosophila wing imaginal discs were dissected from middle to late third instar larvae, and antibody staining was performed according to standard procedures (Sui et al., 2012). The primary antibodies used were: mouse anti-Dlg1 (1:10; 4F3; Developmental Studies Hybridoma Bank, DSHB); rabbit anti-cleaved caspase-3 (1:200; Asp175; Cell Signaling); mouse anti-Mmp1 (1:10; 5H7B11; DSHB); and mouse anti-β-galactosidase, (1:2000; Z3783; Promega). Secondary antibodies (diluted 1:200) used were anti-mouse Cy3, anti-rabbit Cy3 and anti-rat Cy3 (Jackson Immuno Research). The cell membrane was stained using rhodamine-phalloidin (1:1000; PHDR1; Sigma). Cell nuclei were stained with DAPI (1:200,000; 32670; Sigma).

Images were collected using Leica SP8 and Zeiss LSM 800 confocal microscopes. The figures were assembled in Adobe Photoshop CC with minor image adjustments (brightness and/or contrast). To compare the Ets21c-GFP fluorescence intensity in Fig. 5 and Fig. S6, staining and microscopy conditions were identical, and images were edited in parallel using Adobe Photoshop CC.

More than ten Drosophila wing pouches were y-z scanned to statistically assay BCE and ACE. In controls, DP cell nuclei were observed in the middle/center of the cell layer, but never close to the basal membrane or within the overlying wing disc lumen. The BCE and ACE were defined as the presence of a cell nucleus close to the basal membrane and within the overlying wing disc lumen, respectively, especially when the cell nucleus was located outside of the basal membrane or in the lumen (Fig. 6).

C765>ykiRNAi wing discs were dissected as test samples, and C765> wing discs were dissected as controls. For each sample, 100 wing discs were dissected in PBS on ice. Total RNA was prepared using Trizol reagent (Invitrogen, code no. 12183555) following the recommendations of the manufacturer. RNA quality was verified by agarose gel electrophoresis and using a Nanodrop 2000. RNA-seq libraries were sequenced on an Illumina Hiseq platform. Raw reads were processed to clean data using in-house Perl scripts. Clean data were mapped to the Drosophila genome using TopHat v2.0.12 (Kim et al., 2013). Analysis of differential expression in test and control samples was performed using the DESeq2 R package (Anders and Huber, 2010). Genes with DESeq2 adjusted P-values<0.01 were considered to be differentially expressed. KOBAS software (http://kobas.cbi.pku.edu.cn/kobas3/?t=1) was used to test the statistical enrichment of differentially expressed genes in KEGG pathways (http://www.genome.jp/kegg/). We also used DAVID to test the GO functional enrichment of differentially expressed genes (https://david.ncifcrf.gov).

cDNA was reversed transcribed from RNA using the TaKaRa PrimeScript RT reagent kit with gDNA Eraser (code no. RR047A). RT-qPCR was performed using the PerfectStart Green qPCR SuperMix (+Dye II) (TransGen Biotech) in the QuantStudio 6 Flex platform (Applied Biosystems), and each reaction was repeated in triplicate. Transcription levels were normalized to those of act. Data were calculated using Microsoft Excel and presented as mean+s.d. P values were calculated using two-tailed Student's t-test: *P<0.05; **P<0.01; ***P<0.001. The gene-specific primer sequences used are shown in Table S4.

For quantifying the number of clones in the pouch, between eight and ten third instar transgenic larvae were selected at random from three groups, images were taken at 0.5 μm increments along the z-axis using a Leica SP8 confocal microscope. Image stacks were used to quantify the number of separate clones in the pouch. Clone numbers were analyzed using GraphPad Prism 7. For quantification of the percentage of ACE and BCE, transgenic flies were raised as previously described in ‘Transgene expression and clonal induction’, three groups (each group having more than 25 larvae) of third instar wing discs were dissected and scored for the percentage of ACE and BCE. The experimental data were calculated using Microsoft Excel and presented as mean+s.d. P values were calculated using two-tailed Student's t-test. *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001; n.s., no significant difference.

We thank Dr L. Xue, Dr Z. H. Li, Dr A. Bergmann, Bloomington Stock Center, Vienna Drosophila RNAi Center and Tsinghua University Stock Center for fly stocks, and Dr N. Jiang and Dr L. Qi for the confocal facilities.

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.190124.reviewer-comments.pdf