The transcription factor Spalt and human homologue SALL4 induce cell invasion via the dMyc-JNK pathway in Drosophila

Spalt-like (Sall) gene family proteins are zinc finger transcription factors evolutionarily conserved in many organisms from Caenorhabditis elegans to human beings. These proteins can act as both transcriptional repressors and activators in different contexts (de Celis and Barrio, 2009; Sánchez et al., 2011). They play instrumental roles in stem cell development, cell specification and morphogenesis, cancer progression and inherited disorders (Sweetman and Münsterberg, 2006; de Celis and Barrio, 2009). Understanding the regulation of Sall genes is vital to decipher their biological functions.

The first member of the Sall gene family, spalt (sal), was identified as a homeotic gene during Drosophila embryonic development (Frei et al., 1988; Kühnlein et al., 1994). There are two Drosophila spalt homologues, spalt major (salm) and spalt-related (salr), which have complementary functions (Barrio et al., 1996, 1999). Numerous studies have been devoted to the role of sal in patterning and growth control of the Drosophila wing imaginal disc, an epithelial tissue that proliferates during larval development. In the wing disc, the expression of sal is activated by Decapentaplegic (Dpp) signaling in specific regions and leads to tissue growth (de Celis et al., 1996; Barrio and de Celis, 2004; Doumpas et al., 2013; Akiyama and Gibson, 2015). Loss of sal shows abnormal vein formation and reduction in wing size (de Celis et al., 1996; Grieder et al., 2009; Wang et al., 2017). At the cellular level, mitotic cells are strongly reduced in sal mutant wing discs (Organista and De Celis, 2013). Cell death pathways and the JNK signaling are activated in sal knockdown cells, but these two processes only have a minor role in generating the sal mutant phenotypes (Organista and De Celis, 2013; Organista et al., 2015). Conversely, ectopic sal expression promotes cell proliferation (Skottheim Honn et al., 2016; Wang et al., 2017) via positive regulation of the microRNA bantam (Wang et al., 2017). These results suggest that sal is vital in organ size control by accelerating cell proliferation, but the relation of Drosophila sal to tumorigenesis is not yet known.

In vertebrates, there are four Sall paralogues, named Sall1 to Sall4. All four vertebrate Sall members are involved in embryonic development and their mutations lead to severe genetic disorders (Sweetman and Münsterberg, 2006; de Celis and Barrio, 2009). Particularly, SALL4, a mutation that causes Okihiro syndrome (Al-Baradie et al., 2002; Kohlhase et al., 2002), is highly expressed during embryonic development and plays a crucial role in maintaining pluripotency and self-renewal of embryonic stem cells (Wu et al., 2006; Zhang et al., 2006; Yang et al., 2008a). As tissues and organs mature, the expression of SALL4 is gradually decreased. By contrast, there is substantial evidence that SALL4 is highly upregulated in numerous human cancers and regulates multiple cellular processes responsible for cancer progression (Zhang et al., 2015). First, SALL4 regulates the self-renewal of cancer stem cells by targeting a variety of genes, such as upregulation of Bmi-1, Wnt/β-catenin and HoxA9 and repression of PTEN, a tumor suppressor gene (Ma et al., 2006; Lu et al., 2009; Li et al., 2013; Zhang et al., 2014). Second, SALL4 regulates cell proliferation and apoptosis. Overexpressing SALL4 in liver cancer cell lines enhances cell proliferation through Cyclin D expression (Oikawa et al., 2013). In addition, SALL4 negatively regulates the transcription of apoptotic genes (Yang et al., 2008b; Li et al., 2015) through activating the oncogene Bmi-1 (Yang et al., 2007; Lu et al., 2011). Correspondingly, silencing of SALL4 results in less proliferation and differentiation (Elling et al., 2006; Sakaki-Yumoto et al., 2006; Zhang et al., 2006), which is significantly correlated with cell cycle arrest (Böhm et al., 2007; Lu et al., 2011; Oikawa et al., 2013; Zhang et al., 2017) and/or increased apoptosis (Li et al., 2015; Zhang et al., 2017). Third, SALL4 regulates cell migration and invasion. SALL4 improves epithelial-mesenchymal transition (EMT), as indicated by increasing Twist1 and N-cad expression and decreasing expression of E-cad (Zhang et al., 2014; Li et al., 2015; Liu et al., 2015). The EMT activator ZEB1 (Itou et al., 2013) and oncogene cMyc (Yang et al., 2008a; Li et al., 2015; Liu et al., 2015) are positively regulated by SALL4, therefore leads to EMT. Transplantation of SALL4-expressing cells into immunodeficient mice gives rise to subcutaneous tumor growth and tumefaction of many organs (Ma et al., 2006; Oikawa et al., 2013). Lastly, SALL4 is associated with drug resistance, which, in turn, hampers treatment of tumor cell growth (Oikawa et al., 2013; Liu et al., 2015). Thus, SALL4 plays an essential role in regulating tumorigenesis, tumor growth and tumor progression. Yet, how SALL4 regulates invasive cell movement at the molecular level needs to be elucidated.

In this article, we make use of a Drosophila genetic model for epithelial tumor invasion to explore the molecular mechanism of SALL4 in cancer cell invasion and metastasis. Overexpressing the Drosophila sal or human SALL4 generated migrating cells with invasive behavior in the Drosophila larval tissues. The additional cellular and genetic data revealed that sal/SALL4-induced cell invasion depended on dMyc-JNK signaling and was independent of the apoptosis pathway. These results provide new insights into the molecular mechanisms of sal/SALL4-induced cancer invasion and metastasis.

Given the expression level of SALL4 is increased in many types of tumors, to uncover whether SALL4 is capable of inducing cell migration and invasion in vivo, we increased Sal levels in a central region within the spalt expression domain by expressing salm, salr or human SALL4. In the wing disc, when GFP was expressed in the dpp-Gal4 domain in the wild-type background, the boundary (indicated by dotted lines in Fig. 1A) was relatively linear and no GFP-positive cells could be found in the P compartment. In contrast, a significant number of GFP-labeled cells were present both in anterior and posterior regions far away from the dpp-Gal4 domain when sal/SALL4 was overexpressed (Fig. 1B–D). These cells were largely two types. One was grouped cells extruding into the posterior region, which had connections to the major dpp expression region (Fig. 1B–D, yellow arrowheads) and may be either proliferated (Wang et al., 2017) or migrated from the main part. The other was single cells, which were separated from the dpp expression region (Fig. 1B–D, red arrowheads) and probably migrated from the main part. Because the dpp region is anterior cell fate, if these anterior GFP cells emerge in the posterior region, it means they could go across the compartment boundary and invade into the posterior region (Fig. S1). Hence, we considered the GFP signals in the P compartment of the pouch region as invasive cells. To verify that the GFP-tagged cells represent the sal/SALL4-overexpressing cells, Sal and SALL4 were labeled with anti-Sal and anti-HA tag antibodies, respectively. Cell migration occurred exactly in the Sal/HA positive regions (Fig. 1C″,D″). These data demonstrate that Drosophila salm, salr and human SALL4 are highly conserved. For convenient genetic manipulation, we used human SALL4 and one of the Drosophila homologues (either salm or salr) for the following experiments.

Next, clones were performed to further confirm that sal/SALL4 regulates cell movement. In control clones, cells descending from one progenitor tended to remain clustered and the rugged clone outlines (GFP positive cells) showed similar adhesive properties with their unmarked neighbors (GFP negative cells) (Fig. 1E). When sal/SALL4 was overexpressed, some clone cells were dispersed to the single cell level (Fig. 1F,G, red arrowheads), which is similar to expressing another Dpp target gene optomotor-blind (Shen et al., 2014), indicating increased mobility of sal/SALL4-expressing cells. Tumor-like proliferating cell clusters were seen in the hinge region (Fig. 1F,G, yellow arrowheads), a tumor hotspot where tumors often originate (Tamori et al., 2016). Co-expression of the membrane marker CD8-GFP with sal showed that the migrating cells had filopodia-like structures (Fig. 1I), which is a property of migratory and invasive cells (Shen et al., 2014). Taken together, our results demonstrate that the Drosophila salm, salr and human SALL4 are highly conserved in stimulating cell proliferation and cell motility in the wing disc.

To examine whether sal/SALL4 is able to modulate cell movement in other tissues, we turned to the salivary gland, where sal was endogenously expressed at a moderate level (Fig. S2A). Overexpressing sal/SALL4 by AB1-Gal4 triggered cell invasion throughout the body (Fig. S2C,D). After dissecting the body wall of third-instar larvae, invading cells (GFP positive) were detected and completely co-localized with the HA antibody staining (Fig. S2E′), confirming that the GFP-labeled invading cells showed high sal/SALL4 expression. Collectively, our data suggest that ectopic sal/SALL4 expression is sufficient to trigger cell invasion into other tissues.

The invasive behavior of transformed cells is commonly associated with EMT, whose characteristics include increased cell motility, destabilization of adhesion junctions and loss of cell polarity. In order to better visualize the property of sal/SALL4-overexpressing cells, we performed cryosectioning in the wing discs. At the late third-instar stage, the basal membrane of wing disc epithelia was marked by α-integrin (Fig. 2A). In contrast, the salr-overexpressing cells, which were extruded toward the basal side of epithelia, were deficient in α-integrin expression and substantially lost contact with the epithelia (Fig. 2B, arrowheads). These observations suggest that the salr-hyperactive cells were penetrating the extracellular matrix (ECM) during invasive migration. The apical DE-cadherin (DE-cad) protein level did not change significantly, but its localization in cytoplasm and basal distribution were increased (Fig. 2C–E). Cytoplasmic distribution of soluble E-cad, which is generated from extracellular cleavage by matrix metalloprotease (Mmp), is known to promote epithelial cell extrusion (Grieve and Rabouille, 2014). Interestingly, hyperactivation of salr/SALL4 resulted in upregulation of the mesenchymal fate marker DN-cadherin (DN-cad) (Fig. 2G,H), indicating that sal/SALL4 overexpression induces some consequence related to EMT.

As the large size of salivary gland cells makes it easier to observe the cell morphology and cellular protein localization, we used this tissue to further observe the changes of cell polarity. The apical markers DE-cad and β-catenin/Armadillo (Arm), which were expressed on the cell membrane (Fig. S3A,C), were both mis-localized cytoplasmically in sal-expressing cells (Fig. S3B,D). We further marked the apical membrane by antibody against Discs large (Dlg). Dlg was apparently disorganized in sal-expressing cells (Fig. S3F). A severe disruption of the actin and microtubule cytoskeleton may have contributed to the disruption of apical polarity due to the morphological changes of sal-expressing cells (Fig. S3H) (Tang et al., 2016). The above data suggest that sal activation promotes cell invasion by disruption of the apico-basal polarity.

Because the JNK pathway is an essential pathway driving tumor growth and invasion, we investigated whether the JNK pathway mediates sal/SALL4 overexpression-induced cell invasion. Degradation of the ECM components and basement membrane requires the activity of Mmp1, a transcriptional target of JNK signaling (Uhlirova and Bohmann, 2006). We first examined the Mmp1 level. salr/SALL4 overexpression by dpp-Gal4 or in clone cells within the wing discs led to a strong increase in Mmp1 protein level (Fig. 3B–D). The deposition of Mmp1 was also found in the salivary gland (Fig. S4B, dotted lines). Then, the JNK signaling level was probed by a specific antibody against the activated JNK isoform pJNK. The pJNK level was elevated when salr was overexpressed (Fig. 3F). The JNK pathway target puckered (puc) was transcriptionally upregulated (Fig. 3H). Besides in the sal/SALL4-expressing regions, the location of Mmp1, pJNK and puc usually occurred at or close to the edge of salr/SALL4-overexpressing domains (arrowheads in Fig. 3). The non-autonomous activation of JNK pathway in neighboring wild-type cells may also contribute to invasive cell migration, such as in mutant clones for the tumor-suppressor scrib (Ohsawa et al., 2011).

To examine whether JNK is required for sal/SALL4-induced cell invasion, we blocked JNK signaling by expressing several JNK pathway inhibitors. As puc is a JNK-specific inhibitor (Martin-Blanco et al., 1998), increasing puc expression is thought to inhibit the JNK activity. As a result, the invasive migration in sal/SALL4-overexpressing wing discs was repressed by expressing puc (Fig. 3I,J). The Mmp1 level, both in sal/SALL4-expressing regions and adjacent wild-type cells, was rescued (Fig. 3I,J), indicating that the non-autonomous activation of JNK pathway depends on JNK signals from the sal/SALL4-expressing cells. A dominant-negative form of the Drosophila JNK homologue basket (bsk^DN) also greatly repressed salr/SALL4-induced cell invasion (Fig. 3L,M). Consistently, downregulation of Mmp1 by expressing tissue inhibitor of matrix metalloprotease (Timp) (Visse and Nagase, 2003) compromised salr-induced cell invasion (Fig. 3O). In cryosectioning discs, the restoration of basal membrane integrity by Timp was apparent (as indicated by anti-α-integrin staining, Fig. 3P). Statistically, the GFP area in the P compartment was significantly reduced when JNK signaling was repressed. The area of invading cells was reduced more than 60% compared with that of salm, salr, or SALL4 (Fig. 3Q). The above data suggest that inhibition of the JNK pathway largely reduces sal/SALL4-induced cell invasion and epithelial disruption.

As the activation of JNK signaling is often accompanied by the appearance of apoptosis and apoptosis can cause delamination and/or migration of epithelial cells (Rudrapatna et al., 2013; Gorelick-Ashkenazi et al., 2018), we assessed the function of apoptosis in sal/SALL4-overexpressing cells. Caspase-3 (Cas3) was activated in and close to the salr/SALL4-overexpressing domain (Fig. S5B,C, yellow arrowheads), as well as non-autonomously activated elsewhere (Fig. S5B,C, red arrowheads). Further TUNEL assay showed that the migrating cells were not dead cells (Fig. S5D,E). When apoptosis was inhibited by overexpression of p35, an inhibitor of the caspase drICE, salr/SALL4-expressing cells still maintained the ability of horizontal invasion (Fig. S5G,H). To avoid the fact that expressing p35 induces ‘undead’ cells to produce migration signals (Martin et al., 2009), we used Diap1 (Fan and Bergmann, 2008) to suppress caspase Dronc-mediated cell death. Co-expression of Diap1 and salr/SALL4 still induced a large number of invading cells (Fig. S5J,K). Thus, co-expression of p35/Diap1 and salr/SALL4 cannot rescue sal/SALL4-induced cell invasion. Apoptosis does not play a major role in this process.

The human MYC is an oncogene that contributes to tumorigenesis and metastasis. So does the single Drosophila homologue dMyc (Dang, 2012). Previous reports also showed that loss of dMyc promotes cell migration by activating JNK signaling (Ma et al., 2017a; Tavares et al., 2017). Here, overexpression of salr/SALL4 led to a downregulation of the dMyc level in the dpp-Gal4 domain (Fig. 4B′,C′, arrowheads). To confirm the regulation by sal/SALL4, we produced salr/SALL4-overexpressing clones in which the dMyc level was consistently downregulated (Fig. 4E,F, arrowheads). Higher-resolution images illustrated that dMyc was reduced in clone cells (Fig. 4E″,F″, arrowheads). Consistently, dMyc was reduced in the salivary gland (Fig. S4D, dotted lines). Therefore, dMyc was cell-autonomously repressed by sal/SALL4.

Although overexpression of dMyc showed weak cell migration in the wing disc (Fig. 5A), we attempted to rescue sal/SALL4-induced cell invasion by expressing dMyc. Co-expression of dMyc and salr/SALL4 significantly reduced the cell invasion rates (Fig. 5B,C). Statistical results indicate that more than 70% of the GFP cells in the P compartment was lost (Fig. 5I). At the same time, the JNK signal activated by salr/SALL4 ectopic expression was repressed by dMyc expression as indicated by the Mmp1 staining (Fig. 5D,E). In turn, knock-down of dMyc by dMyc-RNAi showed obvious single cell movement (arrowheads in Fig. 5F′). Reducing dMyc also induced activation of the JNK pathway, which was more obviously seen in the x-z view (Fig. 5G). Thus, we deduce that concurrently expressing dMyc-RNAi and sal/SALL4 will enhance sal/SALL4-induced cell invasion and the results were as expected (Fig. 5H,I). These findings demonstrate that dMyc inhibits the JNK signaling and the Drosophila epithelial cell invasion induced by sal/SALL4 depends on dMyc-JNK signaling.

Human SALL4 has been reported to be significantly elevated in metastatic cancer cells. Here, we provide genetic evidence for a model in which sal/SALL4 regulates cell invasiveness by dMyc-JNK signaling. The JNK pathway is an important cellular signaling pathway that regulates a variety of cellular activities relevant to tumorigenesis, such as cell migration, apoptosis and proliferation. JNK promotes the expression of Mmp1, which acts as an enzyme to degrade basement membrane and ECM components to promote tumor cell motility (Uhlirova and Bohmann, 2006). Manipulation of expression of many genes can lead to cell death, cell extrusion and invasive cell migration through activation of JNK signaling (Petzoldt et al., 2013; Rudrapatna et al., 2014; Ma et al., 2017a,b; Sun et al., 2019). sal/SALL4 overexpression activates Mmp1 and reducing JNK can suppress cell invasion and Mmp1 level (Fig. 3; Fig. S4). In addition to Mmp1, some other markers in the JNK pathway such as pJNK (activated bsk) and puc showed a significant increase in expression (Fig. 3). Promotion of cell invasion by sal/SALL4 induction was accompanied by activation of the apoptotic pathway, but it was not dependent on apoptosis because caspase inhibition did not prevent cell invasion upon sal/SALL4 expression (Fig. S5). Therefore, the JNK pathway probably mediates the role of sal/SALL4 overexpression to regulate cell invasion through an apoptosis-independent mechanism.

The MYC gene is one of the most highly amplified oncogenes among many human cancers (Dang, 2012). For instance, in some certain cancer cells, Myc is upregulated through directly transcriptional activation by SALL4 (Yang et al., 2008a; Li et al., 2015; Liu et al., 2015). Besides promoting cancer progression and metastasis, MYC has a bivalent role in regulating tumorigenesis and cell invasion. MYC restrains breast cancer cell motility and invasion through transcriptional silencing of integrin subunits (Liu et al., 2012). In Drosophila, dMyc inhibits JNK signaling in retinal progenitors to block non-autonomous glia over-migration (Tavares et al., 2017). The Drosophila puc gene, encoding the sole JNK-specific MAPK phosphatase and inhibitor (Martin-Blanco et al., 1998), and its mammalian homologue Dusp10 are directly bound by Myc as shown in ChIP-sequencing data (Yang et al., 2013; Sabò et al., 2014). In Drosophila tissues, direct evidence illustrates that dMyc and cMyc activate puc transcription through binding to the Myc binding-motif EB3, and consequently inhibit JNK signaling to suppress cell invasion (Ma et al., 2017a). We found that dMyc is repressed in sal/SALL4-expressing regions and introducing dMyc partially rescues cell invasion (Figs 4 and 5), indicating a repressive role of dMyc in tumor cell migration. As Sal is a transcriptional repressor in both Drosophila and human cells (Sánchez et al., 2011), it is possible that Sal/SALL4 binds to Myc and suppresses its expression because the cMyc promoter has putative binding sites that are available to Zinc finger binding (Wu et al., 2015). Sall2, another emerging cancer player in the Sall family, binds to the cMyc promoter region and represses cMyc expression (Sung et al., 2012; Wu et al., 2015). Thereby, sal/SALL4 may activate JNK signaling through the repression of puc, which is activated by dMyc in Drosophila.

Cell competition occurs when Myc is unevenly distributed between cells. Clones expressing high levels of Myc expand and eliminate the surrounding cells by apoptosis. On the contrary, downregulation of Myc in clones leads to their elimination (de la Cova et al., 2004; Moreno and Basler, 2004). Given sal/SALL4-expressing cells are relatively lower Myc expression, it is possible that the surrounding cells with higher Myc expression become competitors and eliminate those lower Myc expression cells. Intriguingly, sal/SALL4-induced migrating cells are not dead and inhibiting cell death cannot repress sal/SALL4-induced cell invasion (Fig. S5), so the mechanism may not be apoptosis-driven cell elimination (Levayer and Moreno, 2013; Levayer et al., 2015). Previous studies found that JNK activation in surrounding wild-type cells promotes elimination of their neighboring scrib mutants by activating the PVR-ELMO/Mbc-mediated engulfment pathway, and the surrounding JNK is independent of JNK activation in mutant clones (Ohsawa et al., 2011; Nagata and Igaki, 2018). Distinct from this, sal/SALL4-activated non-autonomous activation of JNK is dependent on JNK activation in sal/SALL4-expressing cells (Fig. 3J,K). Whether JNK-dependent engulfment plays a major role in sal/SALL4-mediated extrusion needs to be addressed in the future.

Fly lines were cultured at 25°C on standard fly food unless otherwise noted. The transgenes used were as follows: UAS-salr (de Celis et al., 1996), UAS-salm (from the Bloomington Drosophila Stock Center #29716, short for BL#29716), UAS-SALL4-HA (BL#65835), UAS-Timp (BL#58708), UAS-bsk^DN (Weber et al., 2000), UAS-p35 (BL#5073), UAS-Diap1 (BL#6657), UAS-GFP (nuclear expression, BL#4775), UAS-CD8-GFP (membrane expression) (Lee and Luo, 1999), UAS-dMyc (BL#9674), dMyc-RNAi (BL#36123), puc-lacZ (Martin-Blanco et al., 1998), UAS-puc (Dobens et al., 2001), dpp-Gal4 (Shen and Mardon, 1997), actin5c>CD2>Gal4 (Pignoni and Zipursky, 1997), and AB1-Gal4 (BL#1824). To promote the GFP phenotype in a larval body, salm, salr, or SALL4-overexpressing larvae were raised at 29°C after egg laying. Clones in the larval wing imaginal discs were generated with the genotypes y w¹¹¹⁸ hs-Flp; actin5c>CD2>Gal4 UAS-GFP/CyO; UAS-salr/UAS-SALL4-HA by heat shock at 35.5°C for 30 min. Then, late third-instar larvae were dissected after a recovery period of 3 days at 25°C.

Dissected imaginal discs from third-instar larvae were fixed and immunostained using standard procedures for confocal microscopy. Appropriate primary antibodies and staining reagents include rhodamine-phalloidin (1:50, Invitrogen A12380, Waltham, USA), DAPI (1:500, Sigma-Aldrich 32670, Shanghai, China), rabbit anti-HA [1:500, Cell Signaling Technology (CST) #3724S, Danvers, USA], rat anti-Ci [1:200, Developmental Studies Hybridoma Bank (DSHB) 2A1, IA, USA], mouse anti-α-integrin (1:20, DSHB DK.1A4), rat anti-DE-cadherin (1:100, DSHB DCAD2), mouse anti-DN-cadherin (1:10, DSHB DN-EX #8), mouse anti-Dlg (1:10, DSHB 4F3), mouse anti-Arm (1:100, DSHB N2 7A1), mouse anti-Mmp1 (1:20, DSHB 5H7B11), rabbit anti-pJNK (1:200, CST #4668), rabbit anti-dMyc (1:400, Santa Cruz Biotechnology sc-28207, CA, USA), rabbit anti-β-galactosidase (1:2000, Promega Z378B, Madison, USA), rabbit anti-cleaved caspase-3 (1:200, CST #9661), and rabbit anti-p35 (1:500, Novus Biologicals NB100-56153, Centennial, USA). Rabbit anti-Sal antibody (1:500) was a gift from Professor Rosa Barrio at CIC bioGUNE, Spain. Secondary antibodies (1:200, Jackson ImmunoResearch, West Grove, USA) were anti-mouse Cy2 (115-225-146), Cy3 (115-165-146) and Cy5 (115-175-146); anti-rabbit Cy2 (111-225-144), Cy3 (111-165-144), and Cy5 (111-175-144); and anti-rat Cy3 (112-165-143). The samples were mounted in 50% glycerin before imaging.

After secondary antibody staining, discs were re-fixed in freshly made 4% paraformaldehyde for 30 min and washed three times with 1× PBS, then stored in 30% sucrose solution at 4°C overnight. Wing discs were oriented in Tissue-Tek (Sakura Finetek, Japan), frozen and cut into 20 μm sections on a cryostat (YD-1900, YIDI, China). All samples were mounted in 50% glycerin before imaging.

Imaging of prepared samples was collected by a Leica SP8 confocal microscope. Adult wing images were collected using an inverted microscope (AMG EVOS, USA). To recognize the P compartment boundary before statistical analysis of the invasive cell area, Ci was stained as the A compartment marker (Fig. S1). The invasive cell area in the P compartment of wing discs was calculated by the ImageJ program (National Institutes of Health). Statistical figures were generated by the GraphPad Prism 5 project.

The wing discs were dissected from wandering third-instar larvae in PBS. The discs were fixed in 4% paraformaldehyde for 20 min and washed with PBST (0.2% Triton100) three times for 45 min at room temperature. TUNEL (TdT-mediated dUTP Nick-End Labeling) staining was performed using the in situ Cell Death Detection Kit (TMR red) produced by Sigma-Aldrich (Cat No. 12156792910).

We thank Dr Rosa Barrio for the anti-Sal antibody, Dr Gert O. Pflugfelder for critical reading the manuscript, the Bloomington Drosophila Stock Center for fly stocks, and Dr Na Jiang and Dr Linlu Qi for the confocal facility.