Oncogenic cooperation between Yorkie and the conserved microRNA miR-8 in the wing disc of Drosophila

How organ growth is controlled is a crucial issue in developmental biology. However, we still have little understanding about the mechanisms regulating tissue size during animal development. The wing imaginal disc of Drosophila has been used extensively to characterize the signaling pathways and molecular mechanisms that modulate tissue growth (Hariharan, 2015).

MicroRNAs (miRNAs) are endogenous small non-coding RNAs that have emerged as important regulators of gene activity. miRNAs can target multiple target genes and thereby regulate gene expression networks. The identification of miRNA target genes is crucial to determine the gene networks in which each miRNA participates. The miR-200 family is highly conserved. It has been shown to control cell proliferation and cell viability when expressed in several human cell lines, and miR-200 microRNAs are deregulated in certain human cancers (Bracken et al., 2015; Hyun et al., 2009).

miR-8 is the only member of the miR-200 family in Drosophila. miR-8 controls growth in two opposite ways. On the one hand, it promotes systemic growth by regulating the insulin pathway in the fat body. As a consequence of that, miR-8 mutant flies are smaller than wild-type flies (Hyun et al., 2009; Jin et al., 2012). On the other hand, miR-8 inhibits growth in proliferating tissues. miR-8 inhibits neuroepithelial expansion in the central nervous system by repressing the EGFR ligand spitz (Morante et al., 2013). Likewise, miR-8 represses growth in the proliferating wing epithelium, where clones of cells mutant for miR-8 grow bigger than wild-type cells, and clonal overexpression of miR-8 reduces cell survival (Bolin et al., 2016). Previous studies have identified miR-8 target genes involved in growth control (Bolin et al., 2016; Hyun et al., 2009; Jin et al., 2012; Kennell et al., 2008; Lee et al., 2015; Morante et al., 2013; Vallejo et al., 2011). However, the mechanisms by which miR-8 inhibits tissue growth in the wing epithelium are not completely understood.

The Hippo signaling pathway regulates normal and oncogenic growth in different animal species, including humans. The central components of this pathway comprise a regulatory kinase cascade and a transcriptional module. In the regulatory cascade, Hippo acts with Salvador to phosphorylate Warts, which then acts with its co-factor Mats to phosphorylate and inactivate Yorkie (Yki) (Pan, 2010). Yki is a key element in the transcriptional module. Mutation of Yki inhibits growth and cell survival, whereas Yki overexpression induces tissue overgrowth (Huang et al., 2005).

Here, we show that upregulation of miR-8 represses Yki. Interestingly, co-expression of Yki and miR-8 results in the formation of neoplastic tumors. We find that upregulation of miR-8 leads to a reduction in the growth repressor brinker (brk), and our results suggest that this regulation is involved in the formation of those tumors. Taken together, our studies show that miR-8 upregulation results in the repression of a positive growth regulator, Yki, and a negative growth regulator, brk. Deregulation in this regulatory network can result in the formation of tumors.

We used the wing imaginal disc of Drosophila as a model system to study how miR-8 regulates growth. The wing primordium of Drosophila is an epithelial monolayer that proliferates extensively during the larval stages and provides an excellent model for studying the role of miRNAs during development (Herranz et al., 2010; Waldron and Newbury, 2012). The wing disc can be subdivided along the proximal-distal axis into three main regions: the notum (proximal), the hinge (medial) and the wing pouch (distal). During metamorphosis, the notum will develop into the thorax of the adult fly, the hinge will form the region where the wing blade attaches to the thorax and the wing pouch will give rise to the adult wing blade.

We first used the Gal4 driver nubbin-Gal4 (nub-Gal4), which is specifically expressed in the wing pouch and allowed to manipulate gene expression during wing development (Fig. S1). Expression of UAS-miR-8 under the control of nub-Gal4 caused a reduction in adult wing size (Fig. 1A,B). These wings also showed defects in the wing margin, presumably owing to the role of miR-8 in regulating the Notch pathway (Vallejo et al., 2011). miRNA sponges encoded by UAS-transgenes permit spatially controlled depletion of miRNAs, when expressed under Gal4 control (Loya et al., 2009). Reciprocal to the effect observed in wings expressing UAS-miR-8, localized depletion of miR-8 by expression of two different UAS-miR-8 sponges (Fulga et al., 2015; Loya et al., 2009) resulted in bigger wings (Fig. 1A,B; Fig. S2).

To study the consequences of manipulating miR-8 expression during the development of the wing disc, we made use of the dorsal-specific Gal4 driver apterous-Gal4 (ap-Gal4). By using ap-Gal4, we induced direct transgene expression in the dorsal compartment of the wing disc epithelium (Fig. 1C). This allowed the consequences of gene manipulation to be compared with the adjacent wild-type cells of the ventral compartment as an internal control. In agreement with the results obtained in the adult wing, expression of a UAS-miR-8 transgene reduced the size of the dorsal compartment of the wing disc (Fig. 1D). This was associated with elevated levels of the activated form of Caspase 3 (Cas3-act) in dorsal cells, compared with the adjacent ventral tissue (Fig. 1E). Caspase activation reflects cells undergoing apoptosis. This suggests that miR-8 may regulate elements involved in apoptosis and growth control. This is consistent with previous results reporting that miR-8 induces apoptosis in the wing primordia (Bolin et al., 2016). Reciprocally, expression of UAS-miR-8 sponge under the control of ap-Gal4 caused an increase in wing disc size (Fig. S2).

The Hippo pathway inhibits cell proliferation and induces apoptosis. Activation of the pathway results in phosphorylation and inactivation of the Hippo nuclear effector Yki (Dong et al., 2007; Oh and Irvine, 2008). Yki controls apoptosis and cell proliferation through targets such as the cell cycle regulator Cyclin E, the inhibitor of apoptosis DIAP1 and the miRNA bantam (Huang et al., 2005; Nolo et al., 2006; Thompson and Cohen, 2006). The fact that miR-8 affected growth and apoptosis in the wing epithelium suggested that it might regulate the activity of the Hippo pathway.

To test whether miR-8 regulates Hippo pathway activity, we analyzed the expression of Yki target genes in discs upregulating miR-8. Overexpression of miR-8 in the dorsal compartment of the wing disc resulted in robust downregulation of the Yki target genes Cyclin E and Diap1, when compared with the cells in the otherwise normal ventral compartment (Fig. 2A-D). This is consistent with previous observations made by Buttittas' group showing that miR-8 upregulation causes a reduction in DIAP1 (Bolin et al., 2016). To visualize the expression of the miRNA bantam, we made use of a GFP sensor transgene (Brennecke et al., 2003). bantam activity reduces sensor expression in such a way that high GFP levels reflect low bantam, and vice versa. Cells overexpressing miR-8 showed an increase in GFP levels, when compared with the ventral compartment, indicating that bantam levels were reduced (Fig. 2E,F). These results suggest that miR-8 overexpression causes a reduction in Yki activity.

Computational target prediction identified Yki as a potential miR-8 target gene (Fig. 3A). During the preparation of this manuscript, a paper was published showing that miR-8 can regulate Yki mRNA in Drosophila S2-DRSC cells through that binding site (Umegawachi et al., 2017). We then asked whether miR-8 regulates Yki in the wing primordium. Yki is homogeneously expressed in normal wing discs (Fig. 3C). Expression of UAS-miR-8 under the control of ap-Gal4 led to a robust reduction in Yki protein levels (Fig. 3D). Yki lacks a DNA-binding domain and functions in concert with DNA-binding transcription factors, such as Scalloped (Sd), to induce tissue growth (Goulev et al., 2008; Peng et al., 2009; Ruiz-Romero et al., 2015; Wu et al., 2008; Zhang et al., 2008). Remarkably, computational target prediction also identified the Yki transcriptional partner Sd as a potential miR-8 target gene (Fig. 3B). Sd is expressed throughout the wing disc, showing higher protein levels in the cells of the wing pouch (Fig. 3G). As observed with Yki, overexpression of miR-8 in the dorsal compartment reduced Sd levels (Fig. 3H). We have shown that miR-8 overexpression leads to apoptosis (Fig. 1). To rule out the possibility that the reduction in Yki and Sd protein levels observed was due to the presence of dying cells, we co-expressed miR-8 with the inhibitor of apoptosis p35. Yki and Sd were still downregulated in the context of apoptosis inhibition (Fig. 3E,I). Interestingly, those discs grew bigger in size. Suppression of apoptosis can lead to the generation of undead cells, which can drive tissue overgrowth by expression of the mitogens wingless (wg) and dpp (Huh et al., 2004; Perez-Garijo et al., 2004; Ryoo et al., 2004). In agreement with that, discs co-expressing UAS-miR-8 and UAS-p35 led to ectopic wg and dpp (Fig. S3). Therefore, upregulation of wg and dpp might account for the increase in size observed in discs co-expressing UAS-miR-8 and UAS-p35.

We next asked whether endogenous miR-8 was able to limit Yki and Sd levels in the wing disc. We did not detect any obvious change in Yki protein levels when we expressed the UAS-miR-8-sponge in the dorsal compartment of the wing disc (Fig. S4). During the development of the wing disc, changes in the growth rate of a given compartment are compensated by the neighboring compartment to generate a well-proportioned and functional organ. The Hippo pathway is involved in this compensatory mechanism (Cho et al., 2006; Mesquita et al., 2010; Udan et al., 2003). As this might mask the regulatory effect of miR-8 on Yki, we decided to use a Gal4 driver that was not expressed in a whole compartment but in a more restricted domain. patched-Gal4 (ptc-Gal4) is expressed in a stripe close to the anterior-posterior compartment boundary of the wing disc. Expression of the UAS-miR-8-sponge under the control of ptc-Gal4 allows the comparison of Yki protein levels with the surrounding wild-type cells. Expression of UAS-miR-8-sponge under the control of ptc-Gal4 caused a modest increase in Yki protein levels in most of the discs analyzed (Fig. 3F). However, this phenotype was not totally penetrant and, in some cases, we did not observe any change in Yki levels (more examples are provided in Fig. S4). We did not detect any obvious change in Sd protein levels in discs expressing the UAS-miR-8-sponge transgene (Fig. 3J and Fig. S4).

Together, this suggests that, even though upregulation of miR-8 is efficient in repressing Yki, depletion of miR-8 has a mild effect. Thus, other miR-8 target genes, including members of the wingless pathway, Notch signaling, insulin signaling or cytoskeletal regulators (Bolin et al., 2016; Hyun et al., 2009; Jin et al., 2012; Kennell et al., 2008; Vallejo et al., 2011), should contribute to the increase in size observed upon miR-8 depletion.

Yki mRNA stability has been shown to be controlled by a stem loop (SL2) in its 3′UTR. Interestingly, that regulation depends on a miR-8 seed sequence present in SL2 (Umegawachi et al., 2017). This observation is consistent with our results and supports the role of miR-8 regulating Yki. In animals, most known miRNA targeting occurs within the 3′UTR of mRNAs. However, miRNAs can also pair and target the protein-coding region and 5′UTR (Schnall-Levin et al., 2010). To assess whether miR-8 modulated Yki in vivo by targeting other regions different from the SL2-miR-8 binding site, we cloned the complete Yki cDNA downstream of a tubulin promotor-EGFP reporter plasmid (Yki-sensor, Fig. 3K). Consistent with our previous results, overexpression of miR-8 caused a robust reduction of the Yki-sensor in wing disc cells (Fig. 3L,M). Next, we generated a Yki-sensor carrying a mutant version of the SL2-miR-8 binding site (Yki-miR-8^SL2-mut-sensor, Fig. 3K). To our surprise, the Yki-miR-8^SL2-mut-sensor was still sensitive to miR-8, and the rescue in the regulation of the Yki-miR-8^SL2-mut-sensor by miR-8 was only partial (Fig. 3N,O). This suggests that miR-8 might regulate Yki through other regions, in addition to the binding site present in the SL2, as previously reported by Umegawachi and colleagues.

To assess whether the miR-8-predicted binding site identified in the Sd-3′UTR (Fig. 3B) was functional, we cloned the Sd-3′UTR downstream of a tubulin promotor-EGFP reporter plasmid (Sd-sensor, Fig. 3P). Overexpression of miR-8 caused a reduction in the Sd-sensor levels in the wing disc (Fig. 3Q,R). Mutation of the miR-8-predicted binding site restored the regulation by miR-8 (Fig. 3S,T), indicating that the miR-8-predicted binding site in the Sd-3′UTR was functional in vivo.

The fact that miR-8 plays an important role during wing development led us to study the regulation of miR-8 expression in the wing imaginal disc. We used a miR-8 GFP sensor transgene to visualize the expression of miR-8 (Kennell et al., 2008). This sensor is expressed under the control of the α-tubulin promotor and contains two miR-8-binding sites in its 3′UTR. The presence of miR-8 leads to a decrease in the GFP levels. Thus, cells with high GFP express low levels of miR-8 and vice versa. The miR-8 sensor has a complex expression pattern (Bolin et al., 2016; Karres et al., 2007). It shows high GFP levels in the hinge region of the wing disc, whereas GFP levels are lower in the wing pouch and notum (Fig. 4A). This indicates that miR-8 is downregulated in the hinge. To study whether the downregulation of miR-8 in the hinge was required for normal wing development, we analyzed the consequences of expressing miR-8 in the hinge. The Gal4 driver 18h03-Gal4 is expressed in the hinge (Fig. 4B). Expression of UAS-miR-8 under the control of 18h03-Gal4 reduced the expression of the miR-8 sensor to levels that were comparable to those observed in the thorax and in the wing pouch (Fig. 4C, compare to Fig. 4A). The gene homothorax (hth) is expressed in the hinge (Fig. 4A,B), and is involved in the specification of the hinge during wing development (Azpiazu and Morata, 2000; Casares and Mann, 2000). hth expression overlapped in the hinge with the expression of the miR-8 sensor and with that of the 18h03-Gal4 driver (Fig. 4A,B). Interestingly, Hth protein levels were reduced in cells overexpressing miR-8 (Fig. 4D, compare with Fig. 4A,B). An important player in wing hinge development is the morphogen Wingless (Wg). wg is expressed as two concentric rings in the wing hinge, the so-called Wg inner ring (IR) and the Wg outer ring (OR) (Fig. 4E; Baker, 1988; Couso et al., 1993), and is required for the proliferation of the hinge cells (Neumann and Cohen, 1996; Zirin and Mann, 2007). Hth is required to maintain the expression of the Wg OR (Perea et al., 2009). Consistent with the defects observed in hth, expression of miR-8 in the hinge resulted in a reduction in the Wg OR (Fig. 4E,F). The wing hinge is a complex structure formed by many folds and sclerites (Bryant, 1975). In agreement with the reduction in the expression of genes required for the development of the hinge, expression of miR-8 under the control of the 18h03-Gal4 driver led to wings with defects in the hinge: the wing hinge was reduced in size, and some parts of the most proximal hinge, such as the tegula, were not present (Fig. 4G and H). These observations indicate that the downregulation of miR-8 in the wing hinge is required for proper wing hinge development.

We were intrigued by the expression pattern of miR-8 and decided to study the signals regulating its expression the wing disc. The genes hth and teashirt (tsh) are expressed in the hinge region of the wing disc (Fig. 4A,B,J), and control hinge development (Azpiazu and Morata, 2000; Casares and Mann, 2000). This, together with the fact that hth and tsh are expressed in cells where miR-8 levels are reduced, made hth and tsh two promising candidates as potential regulators of miR-8. To test whether hth regulates miR-8, we analyzed the miR-8 sensor in discs expressing a UAS-hth-RNAi transgene. Expression of UAS-hth-RNAi under the control of the ap-Gal4 driver was efficient in reducing Hth protein levels in the dorsal compartment of the wing disc (Fig. 4I, compare with 4A). Interestingly, the miR-8 GFP sensor was downregulated in the cells expressing the UAS-hth-RNAi transgene (Fig. 4I), indicating that miR-8 levels were increased upon hth depletion. In contrast to that, tsh depletion did not affect miR-8 expression (Fig. 4J,K). This indicates that hth is required to repress miR-8 and is therefore involved in shaping the expression pattern of miR-8.

Yki interacts with Sd to regulate gene expression and induce tissue growth (Goulev et al., 2008; Wu et al., 2008; Zhang et al., 2008). Repression of Yki and Sd by miR-8 upregulation may contribute to the miR-8 growth-repressing role in the wing disc. To test that hypothesis, we restored Yki in discs expressing UAS-miR-8. In the context of regulation of tissue size, miR-8 and Yki have opposing roles. While miR-8 overexpression leads to smaller wing discs, overexpression of Yki induces tissue overgrowth. To our surprise, co-expression of UAS-Yki and UAS-miR-8 did not balance their opposing growth outcomes but led to the development of big tumors (Fig. 5A-C,E; Fig. S5). These larvae did not pupate and died as giant larvae, a phenotype that is characteristic of larvae with neoplastic tumors (Bilder, 2004).

A closer analysis revealed that Yki+miR-8 tumors, far from being homogeneous lumps of cells, were organized in a complex way. While some parts of the tumor maintained normal epithelial organization (as revealed by apical accumulation of F-actin), other regions showed loss of polarized expression of F-actin, indicating defects in epithelial polarity (Fig. 5F). The latter resembled tissue undergoing neoplastic transformation that can result from loss of apical-basal polarity (Bilder, 2004). Interestingly, the parts of the tumor with polarity defects comprised cells with different nuclear sizes, when compared with the surrounding regions with normal epithelial polarity and homogeneous nuclear sizes (Fig. 5F). Altogether, this suggests that the tumor parts undergoing neoplastic transformation were composed of cells with different DNA content. DNA content analysis by fluorescence-associated cell sorting (FACS) revealed a higher number of aneuploid cells in tumors expressing UAS-Yki and UAS-miR-8, when compared with discs expressing UAS-Yki on its own (Fig. 5H).

We recently showed that the gene peanut (pnut) is a miR-8 target gene (Eichenlaub et al., 2016). Pnut belongs to the septin family of proteins and is required for cytokinesis (Neufeld and Rubin, 1994). Cytokinesis failure (CF) can lead to the formation of tetraploid cells. Subsequent divisions can result in aneuploidy, which can be oncogenic in animal models (Schvartzman et al., 2010). Yki can drive tumorigenesis in cells with cytokinesis defects (Fig. 5D,E) (Gerlach et al., 2018). Cytokinesis failure due to pnut repression by miR-8 might contribute to the aneuploid phenotype observed in the tumors expressing UAS-Yki and UAS-miR-8.

Aneuploidy drives tumorigenesis in Drosophila epithelia by activating the c-Jun N-terminal kinase pathway (JNK) (Dekanty et al., 2012). Consistent with that, the regions of the tumor with nuclei of different sizes expressed the JNK target gene Matrix metalloproteinase 1 (Mmp1) (Fig. 5G). Malignant fly tumors express Mmp1 to degrade the basement membrane of the imaginal discs, allowing tumor cell migration and invasion (Beaucher et al., 2007; Uhlirova and Bohmann, 2006). These tumor parts also expressed TRE-RFP, a transgenic reporter of JNK activity consisting of binding sites for AP1 (Chatterjee and Bohmann, 2012) (Fig. S6). Interestingly, JNK was induced only in the regions containing bigger nuclei and remained inactive in those parts of the tumor with nuclei of homogeneous sizes and normal epithelial polarity.

Next, we studied the invasive potential of cells co-expressing UAS-miR-8 and UAS-Yki. The Drosophila wing disc is subdivided into compartments (Mann and Morata, 2000). In normal discs, cells of different compartments do not mix and this can be used as an assay of invasion (Vidal et al., 2006). We used the posterior-specific driver hedgehog-Gal4 (hh-Gal4) to study whether tumor cells migrate through the compartment boundary. Cells expressing UAS-Yki did not mix with anterior cells and remained in the posterior compartment. Interestingly, posterior cells co-expressing UAS-Yki and UAS-miR-8 were observed in the anterior compartment (Fig. S7). This suggests that these tumor cells are invasive.

The observations that the Yki+miR-8 tumors led to the formation of giant larvae, showed defects in polarity, contained cells with different nuclear sizes, expressed an invasive marker such as Mmp1 and invaded the adjacent compartment suggest that these tumors have acquired malignant features. However, only some cells in those tumors had the potential to undergo neoplastic transformation. To determine the origin of those cells, we compared the effects of expressing Yki and miR-8 in the wing pouch (by using the wing pouch Gal4 driver nub-Gal4) or in the hinge region (by using the hinge Gal4 driver 18h03-Gal4). Co-expression of Yki and miR-8 in the wing pouch did not enhance the Yki-overexpressing phenotype, and those discs did not lose epithelial polarity and did not express Mmp1 (Fig. S8). Remarkably, discs overexpressing Yki and miR-8 in the hinge developed big tumors that were positive for Mmp1 and showed defects in epithelial polarity (Fig. S8). This suggests that the epithelial cells in the hinge can undergo neoplastic transformation in the presence of Yki and miR-8 overexpression. In contrast to this, the cells in the wing pouch are not transformed upon Yki and miR-8 overexpression.

miR-8 expression causes apoptosis in the wing imaginal disc. Apoptosis is a well-known tumor-suppressor mechanism, and suppression of apoptosis can lead to compensatory cell proliferation and tumor formation in Drosophila (Huh et al., 2004; Perez-Garijo et al., 2004; Ryoo et al., 2004). Yki inhibits cell death by regulating anti-apoptotic genes such as Diap1 (Huang et al., 2005). To determine whether the cooperative interaction of Yki+miR-8 is explained by the suppression of apoptosis, we co-expressed miR-8 with Diap1. Diap1 expression in that context did not drive tumorigenesis. Co-expression of the apoptosis inhibitor p35 produced similar results (Fig. 6). Hence, the cooperative effect of Yki and miR-8 cannot be explained by the anti-apoptotic role of Yki.

We have recently shown that Yki can drive tumorigenesis in cells with cytokinesis failure as a result of pnut depletion (Gerlach et al., 2018). Interestingly, tumors expressing UAS-Yki and UAS-pnut-RNAi were smaller in size than the tumors expressing UAS-Yki and UAS-miR-8 (Fig. 5A-E). This suggests that other genes controlled by miR-8 might contribute to the formation of the Yki+miR-8 tumors.

Computational target prediction programs identify the transcription factor brk as a potential miR-8 target gene (www.targetscan.org; http://34.236.212.39/microrna/home.do). Brk is a growth repressor (Martín et al., 2004), making it a promising candidate for further analysis. brk is repressed by the Dpp pathway (Campbell and Tomlinson, 1999; Jaźwińska et al., 1999; Minami et al., 1999). Dpp, a member of the TGFβ family of secreted proteins, signals via the SMADs Mad and Medea to control wing growth (Burke and Basler, 1996; Capdevila and Guerrero, 1994; Haerry et al., 1998; Martín-Castellanos and Edgar, 2002; Zecca et al., 1995). It is expressed in a stripe close to the anterior-posterior compartment boundary. Dpp protein diffuses to form a concentration gradient that can be visualized using an antibody that recognizes a form of Mad that is phosphorylated (P-Mad) by the activated Dpp receptor (Tanimoto et al., 2000) (Fig. 7A). The Dpp gradient is converted into an inverse gradient of Brk (Müller et al., 2003). Cells with high P-Mad show low Brk levels, and vice versa (Fig. 7A). Brk levels are very similar in the anterior and posterior compartments of the wing disc (Fig. 7A). We used the posterior-specific Gal4 driver hh-Gal4 to study whether miR-8 regulates brk. hh-Gal4 allowed us to express miR-8 in the posterior compartment and use the adjacent anterior compartment as an internal control. We observed that expression of miR-8 in the posterior compartment of the wing disc resulted in reduced Brk levels (Fig. 7B, compare with 7A). Interestingly, Brk downregulation was not accompanied by an increase of P-Mad, suggesting that miR-8 might regulate brk independently of the Dpp pathway (Fig. 7B, Fig. S9). miRNAs are post-transcriptional regulators of gene activity. We reasoned that, if brk is a miR-8 target gene, miR-8 overexpression should reduce Brk protein levels, yet it should not affect the expression of brk. To monitor the expression of brk, we made use of the brk.M12-lacZ reporter that reproduces the expression of brk in the wing disc (Jaźwińska et al., 1999). We found that, although miR-8 overexpression resulted in a reduction of Brk protein levels, it did not affect the expression of the brk.M12-lacZ transgene (Fig. 7D,E). These observations indicate that miR-8 regulates Brk post-transcriptionally. To study whether this regulation was direct, we cloned the brk-3′UTR downstream of a tubulin promotor-EGFP reporter plasmid (brk-sensor). The effect of miR-8 overexpression on the brk-sensor was mild, when compared with the one observed on Brk protein (Fig. S10). Regions in the brk transcript other than its 3′UTR might contribute to this regulation. Alternatively, repression of brk by miR-8 might be, at least in part, indirect. We have shown that miR-8 induces apoptosis (Fig. 1), and apoptotic cells express dpp for compensatory proliferation (Huh et al., 2004; Perez-Garijo et al., 2004; Ryoo et al., 2004). Ectopic dpp in the context of miR-8 overexpression (Fig. S3) might contribute to brk repression in cells overexpressing miR-8. Depletion of miR-8 did not have an obvious effect in Brk protein levels, suggesting that miR-8 does not regulate brk during normal development (Fig. S11).

To determine whether the synergistic effect of miR-8 with Yki was due to downregulation of brk, we used a UAS-RNAi transgene to deplete brk in Yki-overexpressing discs. This combination produced massive tissue overgrowth, far in excess of the effects of Yki overexpression on its own, and comparable to the Yki+miR-8 tumors (Fig. 8A-C). We obtained a similar result when co-expressing UAS-Yki together with UAS-dpp, as an independent way of depleting brk (Fig. 8D). If downregulation of brk by miR-8 contributes to the formation of tumors, we reasoned that restoring brk expression should suppress the formation of tumorous discs. Co-expression of a UAS-brk transgene with UAS-Yki and UAS-miR-8 completely suppressed the formation of tumors (Fig. 8E,F).

We next analyzed in more detail the tumors expressing Yki and depleting brk. These tumors resembled the Yki+miR-8 tumors. In both cases, giant larvae were formed. As observed in the Yki+miR-8 tumors, the tumors expressing UAS-Yki and UAS-brk-RNAi were organized in a complex way: whereas some regions of the tumor showed normal epithelial polarity, other regions of the disc were more disorganized and epithelial polarity was disrupted (Fig. 8G,H). Some parts of the tumors expressing UAS-Yki and UAS-brk-RNAi were positive for the pro-invasive marker Mmp1. As observed in the Yki+miR-8 tumors, there was a good correlation between the presence of epithelial polarity defects and the expression of Mmp1 (Fig. 8G,H). These observations suggest that downregulation of the Dpp target gene brk is required for the formation of tumors in a context of UAS-miR-8 and UAS-Yki co-expression.

The control of organ size requires the coordination of cell growth, cell proliferation and apoptosis, and deregulation in these processes can lead to cancer. Understanding how miRNAs regulate tissue and tumor growth depends on the identification of the specific mRNA targets that mediate these outputs.

We show here that the conserved miRNA miR-8 controls organ growth and tumor formation. miR-8 overexpression causes a reduction in tissue size, induces cell death and represses the Hippo transcriptional complex formed by Yki and its transcriptional partner Sd. To our surprise, co-expression of Yki and miR-8 results in loss of growth control and the formation of malignant tumors. This suggests that repression of Yki by miR-8 might serve as a homeostatic mechanism to prevent the formation of tumors. Interestingly, those tumors have their origin in the hinge. In normal development, miR-8 is active in the thorax and wing pouch; however, it is downregulated in the hinge region of the wing disc (Bolin et al., 2016; Karres et al., 2007). Given that miR-8 and Yki are oncogenic partners in the hinge, downregulation of miR-8 in those cells might also provide a mechanism to prevent the formation of neoplastic tumors in a context of Yki upregulation. Hinge cells in the wing disc are protected against apoptosis (a central tumor suppressor mechanism) and can survive external apoptotic inputs, including irradiation, drug-induced apoptosis and cell competition (Tamori and Deng, 2017; Tamori et al., 2016; Verghese and Su, 2016, 2017). Consistently, recent studies have shown that the hinge is a hotspot in the formation of epithelial tumors (Khan et al., 2013; Tamori and Deng, 2017; Tamori et al., 2016). The results shown here provide a new example supporting the notion that the hinge is an oncogenic hotspot (Tamori and Deng, 2017).

We found that miR-8 upregulation leads to a reduction in the negative growth regulator Brk. Our findings show that depletion of brk in a context of Yki upregulation causes the Yki-expressing tissue to lose polarity and express an invasive marker such as Mmp1. These characteristics are associated with the development of malignant tumors (Beaucher et al., 2007; Bilder, 2004).

Although miR-8 upregulation causes a robust reduction in Yki, depletion of miR-8 causes a rather modest increase in Yki protein levels. Interestingly, this effect is not totally penetrant. The molecular processes that regulate cellular functions are noisy, and intrinsic and extrinsic factors can affect levels of gene expression. Incomplete penetrance can be due to stochastic variation in gene expression (Raj et al., 2010). As in the case of miR-8, most miRNA targets undergo small changes at the protein level upon miRNA depletion (Baek et al., 2008; Selbach et al., 2008); this causes phenotypes with incomplete penetrance (Li et al., 2006; Medeiros et al., 2011). This has led to the idea that miRNAs have an important role in suppressing the fluctuations in target gene expression to ensure that noisy gene expression does not have a detrimental impact on animal development or homeostasis.

miR-8 belongs to the miR-200 family, members of which have been reported to act as tumor promotors and tumor suppressor in a context-dependent manner. Although a large number of studies have shown a suppressive role of the miR-200 family on cancer initiation and progression (Humphries and Yang, 2015), some cancer types, such as ovarian cancer, display an upregulation of the miR-200 family genes (Mateescu et al., 2011). Similarly, miR-8 can promote and suppress tumor growth in a context-dependent manner. Although miR-8 inhibits the formation of Notch-driven tumors (Vallejo et al., 2011), miR-8 cooperates with the oncogene Egfr in the formation of metastatic tumors (Eichenlaub et al., 2016). This emphasizes the importance of understanding the context-dependent role that different genes play during the process of tumorigenesis.

The stocks used are described in the following references: ap-Gal4, hh-Gal4 and nub-Gal4 (Calleja et al., 1996); ptc-Gal4 (Wilder and Perrimon, 1995); UAS-miR-8-RFP (Karres et al., 2007); UAS-miR-8-sponge-GFP (Loya et al., 2009); UAS-miR-8-sponge-mCherry (Fulga et al., 2015); UAS-Yki (Huang et al., 2005); UAS-brk (Jaźwińska et al., 1999); miR-8-sensor (Kennell et al., 2008); bantam-sensor (Brennecke et al., 2003); TRE-RFP (Chatterjee and Bohmann, 2012); UAS-tsh-RNAi (BDSC ID: 35030); UAS-hth-RNAi (BDSC ID: 34037); brk-M12-lacZ (Jaźwińska et al., 1999); UAS-pnut-RNAi (VDRC ID:11791); and UAS-brk-RNAi (VDRC ID: 101887). 18h03-Gal4, UAS-p35, UAS-GFP, UAS-lacZ, UAS-Diap1, UAS-LUC-miR-8, UAS-dpp, dpp-lacZ, ci-Gal4, Cyclin E-lacZ and tub-Gal80ts are described in FlyBase and were obtained from the Bloomington Drosophila Stock Center. Further details of the strains shown in each figure can be found in the supplementary Materials and methods.

The Gal4/Gal80 system was used to allow conditional transgene activation in order to bypass early lethality due to transgene expression. Embryos were collected from crosses of the indicated genotypes for 48 h at 18°C and allowed to develop for 5 days at 18°C to maintain the Gal80-dependent repression of Gal4 until the larvae reached early third instar. Larvae were then transferred to 29°C to induce Gal4 activity and raised for 2-10 days at 29°C before being processed for immunostaining.

Primary antibodies were as follows: rabbit anti-Caspase-3act (1:100 dilution, Cell Signaling Technology, #9661S); mouse anti-β-Gal (1:20 dilution, Developmental Studies Hybridoma Bank), mouse anti-Wg (1:50 dilution, Developmental Studies Hybridoma Bank), anti-DIAP1 (a gift from Bruce Hay, California Institute of Technology, Pasadena, CA, USA, dilution 1:100), rabbit anti-Gal4 (1:100 dilution, Santa Cruz Biotechnology, sc-577), rabbit anti-Hth (a gift from Natalia Azpiazu, Centro de Biología Molecular Severo Ochoa, Madrid, Spain, 1:1000 dilution), rabbit anti-Tsh (dilution 1:1000; Ng et al., 1996), rabbit anti-pMAD (a gift from Ed Laufer, Columbia University, NY, USA, dilution 1:100), guinea pig anti-Brk (a gift from Ginés Morata, Centro de Biología Molecular Severo Ochoa, Madrid, Spain, 1:500 dilution), rabbit anti-Yki (a gift from Duojia Pan, Howard Hughes Medical Institute, MD, USA, 1:1000 dilution), rabbit anti-Sd (a gift from Kirsten Guss, Dickinson College, Carlisle, PA, USA, 1:1000 dilution; Guss et al., 2013), Alexa Fluor 635 or 568 phalloidin was used to label F-actin (Life Technologies).

Third-instar larvae were dissected in PBS and fixed in 4% formaldehyde/PBS for 20 min at room temperature, then washed three times for 10 min in 0.1%Triton/PBS (PBT) and blocked for 30 min in 3% BSA in PBT with 5 mM NaCl (BBT). Samples were incubated at 4°C overnight with primary antibody diluted in BBT, washed three times for 15 min in BBT at room temperature, and incubated with fluorescent secondary antibody and DAPI for 1 h at room temperature. After four 15 min washes with PBT, wing imaginal discs were mounted in 90% glycerol and PBS containing 0.05% N-Propyl Gallate.

Wing discs of larvae carrying the different UAS transgenes (including UAS-GFP) were dissected in ice-cold PBS, dissociated with trypsin-EDTA and fixed with formaldehyde 4% for 20 min and ethanol 70% (vol/vol) for 2 h. Fixed cells were stained with DAPI. The fluorescence of the DAPI and the GFP signal was determined by flow cytometry using a BD FACSAria Fusion. Doublets were discriminated using an integral/peak dot plot of DAPI fluorescence.

Images were taken using a Leica SP8 confocal microscope and analyzed using Fiji software (Schindelin et al., 2012) and Adobe Photoshop. In the images presented in the figures, the orientation and/or position of the wing discs were adjusted in the field of view. No relevant information has been affected. The original images are available on request.

The size of the third instar wing imaginal discs and adult wings were measured using the Fiji Software. The tissue sizes of the different genotypes were normalized to the controls. Mean and standard deviations were calculated using two-tailed unpaired Student's t-test. When standard deviations were unequal, a two-tailed unpaired t-test with Welch's correction was carried out. Multiple groups were compared using an ANOVA test and Tukey's post-hoc analysis. Calculations and bar graphs were made using the Graph pad Prism 7 software.

The sizes of the dorsal compartment of the wing discs were measured using Fiji Software. Wing discs were dissected at day 3 after induction for ap>GFP, and at day 6 after induction for ap>Yki, ap>Yki+miR-8 and ap>Yki+pnut-RNAi to allow direct comparison. The specified cell population (dorsal compartment) is labeled using GFP. The tissue sizes of the different genotypes were normalized to the control. The mean and standard deviations were calculated using an ANOVA test and Tukey's post-hoc analysis. Calculations and bar graphs were made using the Graph pad Prism 7 software.

The Yki cDNA was amplified from plasmid DNA (DGRC: LD21311) with the following primers: 5′-GCTCTgcggccgcAACACCTTAATGTTATAGTT (Yki-fw) and 5′-GGGGCCGCctcgagATGTAAAGAAATACTATAAA (Yki-rv); Sd 3′UTR from plasmid DNA (DGRC: IP16090) with 5′-ATAGAgcggccgcCAACAAGCAACCACAACCAT and 5′-AGCTActcgagAATAAGCCTCAACCTTGCTT; and Brk 3′UTR from plasmid DNA (DGRC: 18244) with 5′-TATATgcggccgcACCCCAAACCCTAGGTATAA and 5′-CCCGAGctcgagGGTGTGTATTTGTTTCGTTT. They were cloned into the control sensor plasmid as NotI-XhoI fragments.

To prepare the mutant versions with the miR-8 seed sequences mutated, the following primer pairs were used: Yki-mutant-rv, 5′-GCACCTCGAGATGTAAAGACCTGCTATAAAATTTGAAAAATATTTAATTTAG; Sd-mutant-fw, 5′-CTAAATCTCAGCAGGACGGGCGAAAAAGAAGCGGCATTGG; and Sd-mutant-rv, 5′-TCTTTTTCGCCCGTCCTGCTGAGATTTAGAGCAGCCGCGG.

The miR-8 binding site in Yki was mutated by a single PCR step using the Primer Pair Yki-fw and Yki-mutant-rv. The miR-8 binding site in Sd was generated by two simultaneous PCR reactions, producing overlapping DNA fragments (mutated in the miR-8 see sequence) that are annealed in a second consecutive PCR step.

We thank Stephen M Cohen for his help in the design of the project. We thank Melissa A. Visser for technical support. We thank Patricia Jarabo and Stephan U Gerlach for their comments during the preparation of this manuscript. We thank Bruce Hay, Carlos Estella, Marco Milán, Natalia Azpiazu, Ginés Morata, Ed Laufer, Duojia Pan and Kirsten Guss for reagents, and the Developmental Studies Hybridoma Bank, Vienna Drosophila RNAi Collection (VDRC) and Bloomington Stock Center for antibodies and fly strains.