An intrinsic tumour eviction mechanism in Drosophila mediated by steroid hormone signalling

Polycomb group proteins are epigenetic regulators maintaining transcriptional memory during cellular proliferation. In Drosophila larvae, malfunction of Polyhomeotic (Ph), a member of the PRC1 silencing complex, results in neoplastic growth. Here, we report an intrinsic tumour suppression mechanism mediated by the steroid hormone ecdysone during metamorphosis. Ecdysone alters neoplastic growth into a nontumorigenic state of the mutant ph cells which then become eliminated during adult stage. We demonstrate that ecdysone exerts this function by inducing a heterochronic network encompassing the activation of the microRNA lethal-7, which suppresses its target gene chronologically inappropriate morphogenesis. This pathway can also promote remission of brain tumours formed in brain tumour mutants, revealing a restraining of neoplastic growth in different tumour types. Given the conserved role of let-7, the identification and molecular characterization of this innate tumour eviction mechanism in flies might provide important clues towards the exploitation of related pathways for human tumour therapy.

P olycomb group (PcG) proteins are evolutionarily conserved chromatin regulators modulating histone modifications and suppressing target gene expression, required for the maintenance of cellular memory 1,2 . PcG proteins can bind to particular genomic regions, where they mediate specific histone modifications and chromatin compaction, therefore suppressing the expression of the target genes in these loci. Dysregulation of PcG genes is associated with various human cancers, but the mechanisms are incompletely understood 3 .
PcG proteins form two major Polycomb repressive complexes, PRC1 and PRC2, to silence the expression of target genes. Previous studies have shown that the PRC1 components can act as tumour suppressors in Drosophila 4,5 . In the developing eyeantennal imaginal discs, for instance, cells homozygous mutant for ph overgrow and give rise to neoplastic tumours 4,5 . These tumours can be transplanted and continue to grow in wild-type adult flies 5 .
Here, we carry out studies to investigate the mechanisms underlying tumour formation and growth in ph mutants. Unexpectedly, we observe that the tumorigenic ph mutant cells are transformed into nontumorigenic cells after metamorphosis, and eventually evicted in adult flies. We show that ecdysone signalling is responsible for the transformation of tumorigenicity. By performing transcriptome analyses we identify miRNA let-7 as a key target of the ecdysone response in this process. We further demonstrate that mis-expression of chronologically inappropriate morphogenesis (chinmo), a direct target of both Ph and let-7, is required for tumour growth. Furthermore, we show that the let-7 cascade could also suppress the overgrowth of brain tumours in brain tumour (brat) mutant flies. Our analyses reveal an intrinsic mechanism that is able to reprogram tumorigenic cells and suppress their malignant growth in adult Drosophila.

Results
Conversion of tumorigenic ph 505 cells during metamorphosis. The Drosophila genome encodes two ph genes, ph proximal (php) and ph distal (ph-d) 6 . ph 505 is a loss of function allele of both genes 7 . Homozygous ph 505 clones, generated genetically by MARCM (mosaic analysis with a repressible cell marker) 8 and marked by GFP, overgrow and give rise to large tumours in the larval eye-antennal discs at the wandering third instar (Fig. 1a). The morphology of these clones is in sharp contrast to wild-type GFP-expressing clones (Fig. 1b). After transplanting ph 505 eye disc tumours into wild-type adult hosts (Fig. 1c, arrow), ph 505 cells continued to proliferate, resulting in the formation of neoplastic tumours (Fig. 1c, d). This indicates that larval ph 505 cells are tumorigenic and is also consistent with previously reported results 4,5 . These tumours can recapitulate proliferation after serial retransplantation into new hosts, but they did not give rise to metastatic tumours in other parts of the body (Fig. 1d). In newly eclosed adult flies, GFP-marked ph 505 cells can be observed all over the body, including the head, legs, thorax, and abdomen (Fig. 1e). However, this was caused by the expression of the ey-flp in all leg discs and the genital disc, producing GFP-marked clones in these tissues as well (see Methods).
In the head, these marked ph 505 cells formed grape-like, singlelayered epithelial structures ( Fig. 1f; Supplementary Fig. 1a). Surprisingly and in contrast to transplanted tumour tissue, the ph 505 tumour cells disappeared gradually during fly adulthood ( Fig. 1g; Supplementary Fig. 1b). Immunostainings showed that the single layer of ph 505 cells in these spherical structures did not proliferate and did not differentiate into neurons (Fig. 1f). Moreover, after transplantation of these ph 505 structures into wild-type hosts (Fig. 1h, arrow), these cells did not grow and also disappeared within a few days (Fig. 1i). A subset of the ph 505 cells found in adult flies expressed Drosophila cleaved death caspase-1 (cDCP-1) (Fig. 1j), an apoptosis cell marker 9 . In addition, the autophagy marker UAS-mCherry:Atg8 (Ch:Atg8) 10 was also expressed in a number of cells (Fig. 1k). These results suggest that there is a conversion of tumorigenic larval ph 505 cells into nontumorigenic adult ph 505 cells at metamorphosis (henceforth named metamorphed cells), which are then eliminated by either apoptotic and/or autophagic cell death in the adult.
Ecdysone controls the transformation of ph 505 tumour cells. 20-hydroxyecdysone (ecdysone) is the key molting steroid hormone controlling metamorphosis of flies 11 . Ecdysone is produced as a series of brief low-level pulses during embryonic and early larval stages. Near the end of third larval instar, a mid-level pulse of ecdysone triggers pupariation 11,12 . The expression of ecdysone increases and reaches the peak level around 48 h after pupa formation. Afterward, ecdysone expression gradually decreases to a low basal level, which is then maintained during adulthood 11,12 . To assess the role of ecdysone in altering the oncogenic potential of ph 505 cells, we ectopically expressed a dominant-negative form of the ecdysone receptor (EcR) 13 in ph 505 mutant cells (ph 505 ; UAS-EcR DN ). In contrast to metamorphed ph 505 cells (Fig. 1e, g), ph 505 ; UAS-EcR DN cells continued to grow in the adults and resulted in the accumulation of large tumours throughout the body (Fig. 2a). In the head, the tumour mass (Fig. 2b, arrow) could reach a similar size as the adult brain (Fig. 2c, arrowhead). As a result, these flies showed a significantly reduced lifespan (Fig. 2d) and ph 505 ; UAS-EcR DN cells were able to give rise to neoplastic tumours after transplantation (Fig. 2e). A knockdown of EcR co-receptor Ultraspiracle (Usp) using transgenic RNAi (ph 505 ; UAS-RNAi-usp) showed similar results substantiating ecdysone involvement ( Supplementary Fig. 2). Hence, the manifestation of the ecdysone pulse at metamorphosis appears directly responsible for suppressing the tumorigenic character of larval ph 505 cells. Conversely, a cell-autonomous block of ecdysone signalling retains the tumorigenicity of ph 505 cells in the intact adult flies.
To better understand how ecdysone exerts its tumoursuppressor function in ph 505 cells, we performed transcriptome analyses by mRNA sequencing. RNA from metamorphed ph 505 cells in adults, 4, 8, 14 weeks old tumours in adult hosts after weekly re-transplantations, and wild-type transplanted discs as control were extracted and sequenced. In total, we identified 2015 significantly differentially expressed genes. Applying kmeans clustering on the gene expression profiles across wild-type, tumour and metamorphed samples, groups of genes with similar profiles can be found ( Fig. 2f; Supplementary Fig. 3). Gene ontology (GO) term analysis of the resulting five groups revealed that transcription regulators and genes required for metabolic processes were upregulated in the tumours, but genes involved in neuronal differentiation were downregulated ( Fig. 2f; Supplementary Fig. 3). Cell-cycle-related genes were downregulated in metamorphed ph 505 cells in addition to the differentiation-related genes ( Fig. 2f; Supplementary Fig. 3), when compared to a wildtype transplanted disc. Lower levels of neuronal marker Elav can corroborate this (Fig. 1f). We next focused on genes that are known to respond to ecdysone ( Supplementary Fig. 4). Expression levels of a group of genes was upregulated in the metamorphed ph 505 cells but decreased in the transplanted tumours (Fig. 2g). Among these is let-7-C, which is a polycistronic locus encoding three miRNAs including let-7 14 .
As the mammalian homologues of let-7 are underexpressed in various cancer types 15 , we decided to further investigate the potential role of let-7 in altering ph 505 tumorigenicity.
Ecdysone-induced let-7 suppresses the growth of ph 505 tumour. In Drosophila, let-7 expression is induced by ecdysone at the start of pupariation, and the expression of let-7 correlates with the dynamic changes of the ecdysone level during metamorphosis 16 . A first step was to confirm mature let-7 levels in the various samples using the Taqman quantitative PCR assay. As previously reported 16 , the expression of mature let-7 was low at the wandering third larval instar but increased significantly to a high level at 48 h after pupa formation (Fig. 3a). Indeed, let-7 expression was also elevated in the metamorphed ph 505 cells, but not in the ph 505 transplanted tumours (Fig. 3a). Moreover, let-7 level in ph 505 ; UAS-EcR DN cells collected from adults remained low (Fig. 3a), indicating let-7 was not induced when ecdysone signalling was blocked.
To further characterize the role of let-7, we reduced the activity of endogenous let-7 in ph 505 cells using a transgenic miRNA sponge 17 (ph 505 ; UAS-let-7-SP) (Fig. 3a). We observed that ph 505 ; UAS-let-7-SP cells maintained the tumorigenic growth in the adults (Fig. 3b). After dissection and transplantation, ph 505 ; UASlet-7-SP cells collected from adults retained tumour formation in the hosts (Fig. 3c), indicating that let-7 is necessary to mediate the suppression of tumorigenic growth of ph 505 cells. Next, we overexpressed let-7 in ph 505 cells. Overexpression of either the entire let-7-C (ph 505 ; UAS-let-7-C) or let-7 alone (ph 505 ; UAS-let-7) could strongly suppress tumour growth in the eye-antennal discs ( Supplementary Fig. 5a, b). The average tumour volume was reduced to 21% in ph 505 ; UAS-let-7 larvae (Fig. 3d). The effect of let-7 overexpression appeared to be independent of apoptosis, as overexpression of let-7 did not lead to elevated cell death in the eye-antennal discs ( Supplementary Fig. 5c, d). Moreover, ph 505 ; UAS-let-7 cells did not give rise to tumours after transplantation into adult hosts (n > 50). let-7-C also encodes another two miRNAs, miR-100 and miR-125. We next tested if these two miRNAs might also play a role in transforming the tumorigenic ph 505 cells. First, unlike let-7-SP, ph 505 ; UAS-miR-100-SP cells stopped growing and disappeared in the adult flies ( Supplementary Fig. 6a). Immunostaining showed that the apoptosis cell marker cDCP-1 was expressed in the ph 505 ; UAS-miR-100-SP cells ( Supplementary Fig. 6b), suggesting that the ph 505 ; UAS-miR-100-SP cells can still be transformed into nontumorigenic cells and undergo apoptosis. For miR-125 we observed that the ph 505 ; UAS-miR-125-SP cells stopped growing in the adult flies, but were still visible in these flies 2 weeks after eclosion ( Supplementary Fig. 6c), unlike the metamorphed ph 505 cells that eventually are eliminated. Immunostaining showed that ph 505 ; UAS-miR-125-SP cells did not express the apoptosis cell marker cDCP-1 ( Supplementary Fig. 6d), nor did they express the mitosis marker PH3 ( Supplementary Fig. 6e). Furthermore, when we transplanted adult ph 505 ; UAS-miR-125-SP cells into host flies ( Supplementary Fig. 6f), they did not give rise to neoplastic tumours, indicating the ph 505 ; UAS-miR-125-SP cells are not tumorigenic anymore. However, even at 4 weeks after transplantation, the ph 505 ; UAS-miR-125-SP cells were still present ( Supplementary Fig. 6g). These results show that ph 505 ; UAS-miR-125-SP cells neither proliferate nor undergo apoptotic cell death. Because miRNA sponges inhibit miRNA/mRNA interaction by sequestration, the function of targeted miRNAs might not be disrupted completely. Therefore, we tested if the miRNA sponges used in our experiments were functional. In flies, the zinc finger transcription factor chinmo is a known target of let-7 and miR-125 18 . Using quantitative PCR, we determined the expression of chinmo, and some computationally predicted targets of miR-125 and miR-100, as there are no experimentally validated miRNA/ mRNA interactions for these two miRNAs. The expression of chinmo indeed increased in ph 505 ; UAS-let-7-SP cells (Supplementary Fig. 6h), but did not in ph 505 ; UAS-miR-125-SP cells ( Supplementary Fig. 6i). However, the expression of the two computationally predicted targets, Zasp52 and RecQ4, increased in ph 505 ; UAS-miR-125-SP cells (Supplementary Fig. 6i). On the other hand, the expression of predicted targets of miR-100 did not increase in ph 505 ; UAS-miR-100-SP cells. These results indicate that the let-7 sponge and miR-125 sponge appear to be functional in ph 505 cells, whereas the miR-100 sponge may not.
To further evaluate the role of miR-125 and miR-100, we carried out overexpression experiments. First, overexpression of miR-125 (ph 505 ; UAS-miR-125) could partially reduce the tumour volume in the eye-antennal discs ( Supplementary Fig. 5a). However, when we transplanted the ph 505 ; UAS-miR-125 cells into wild-type hosts, we could still observe the formation of tumours in 52% of the hosts (n = 31), indicating ph 505 ; UAS-miR-125 cells are still tumorigenic. Similarly, overexpression of miR-100 (ph 505 ; UAS-miR-100) only slightly reduced the tumour volume in the eye-antennal discs ( Supplementary Fig. 5a) and did not alter the tumorigenicity.
Taken together, these analyses show that let-7 is the key regulator in the let-7-C. The miRNA appears indispensable and needed for suppressing the tumorigenic character of ph 505 cells. While our tests do not identify a significant involvement of miR-100, miR-125 does appear to have a certain role in the elimination of the metamorphed cells. Previous work has revealed that there are cross-regulatory relationships among the three miRNAs 19 , which may explain the complicated phenotype of the ph 505 ; UAS-miR-125-SP cells described above.
Ph and let-7 target chinmo is required for ph 505 tumour growth. To further elucidate the mechanism of how let-7 controls the tumorigenesis of ph 505 cells, we searched for genes that are bound by Ph in their promoter region 20 and are let-7 targets. Among others, we identified the gene chinmo 21 (Fig. 3e). The expression of chinmo was significantly increased in the ph 505 tumours but decreased in the metamorphed ph 505 cells (Fig. 2f, black curve). Immunostaining showed that Chinmo was ubiquitously expressed in the transplanted tumours ( Fig. 3f) but was not detectable in metamorphed ph 505 cells (Fig. 3g). In the eye discs, Chinmo was expressed in the ph 505 mutant clones, as well as the neighbouring heterozygous cells (Supplementary Fig. 7a). However, at 48 h after pupa formation, when ecdysone level was at the peak and let-7 was highly expressed (Fig. 3a), the expression of Chinmo became undetectable ( Supplementary Fig. 7b). Another known let-7 target is the abrupt gene 22 , which has been shown to promote the development of some types of tumours in the eyeantennal discs 23 . However, by immunostaining we did not detect the expression of Abrupt in the transplanted ph 505 tumours, in the eye-antennal discs, as well as in the metamorphed ph 505 cells. This indicates that abrupt does not play a significant role in regulating the tumorigenesis of ph 505 cells.
To test the function of Chinmo, we overexpressed the protein in the tumour cells (ph 505 ; UAS-chinmo) and observed that these cells continued to grow in the adult flies (Fig. 3h) and gave rise to tumours after transplantation (Fig. 3i). Furthermore, when chinmo was knocked down (ph 505 ; UAS-RNAi-chinmo), the average volume of the ph 505 ; UAS-RNAi-chinmo clones in the eye-antennal discs was reduced to 60% (Fig. 3d, Supplementary  Fig. 5e). After transplantation of ph 505 ; UAS-RNAi-chinmo eye discs, tumours formed in only 16% of the hosts (n = 67), significantly less frequent than after transplantation of larval ph 505 eye discs (79%, n = 74). These results show that chinmo is an important downstream effector of let-7 in regulating the proliferation of ph 505 -driven tumours.
Ecdysone-induced let-7 suppresses the growth of brain tumours. Because let-7 expression is induced by ecdysone in a variety of tissues during metamorphosis 16 , we next tested if let-7dependent tumour suppression also acts in a different tissue. In Drosophila, brat gene mutations lead to the formation of malignant brain tumours 24 . To generate such tumours, we used wor-Gal4 ase-Gal80 to knockdown brat (UAS-RNAi-brat) in the type II neuroblasts 25 . Compared to the wild-type adult brain (Fig. 4a), brat RNAi resulted in the growth of large brain tumours in all  (Fig. 4b). However, when let-7 was overexpressed in brat-negative cells (UAS-RNAi-brat; UAS-let-7), a strong suppression of tumour growth in all adult brains was observed (Fig. 4c). Chinmo has recently been shown to play a key role in sustaining the tumour growth in brat mutants, and the overgrowth is markedly inhibited in brat; chinmo double mutant cells 26 . We suspected that let-7 inhibits the brat tumour growth also by suppressing chinmo. However, since the brat mutant cells continue to grow in adult flies, we wondered why these tumours cannot be suppressed by the ecdysone pulse during metamorphosis as seen in ph 505 tumours. A recent study showed that the expression of the neuronal-specific ecdysone receptor isoform EcRB1 is temporally regulated in the neuroblasts; it is expressed from the mid third instar, when Chinmo is no longer expressed 27 . We therefore speculated that in brat tumours, ecdysone receptors are not expressed because of the sustained expression of Chinmo. Indeed, immunostaining with EcRB1-specific antibodies showed that the receptor was only weakly expressed in wild-type type II neuroblasts (Fig. 4d, arrows) and undetectable in brat tumours (Fig. 4e). Consequently, we tested whether ectopic expression of the EcRB1 in the tumour cells would restore the eviction mechanism. In both cases, larvae of the genotype brat-RNAi as well as larvae coexpressing EcRB1 (UAS-RNAi-brat; UAS-EcRB1) show large brain tumours at the third larval instar (Fig. 4f, g). However, in the case of EcRB1 coexpression, the size of the brain tumours was significantly reduced in all adult brains (Fig. 4h), demonstrating that also brat tumour cells can be metamorphed. Indeed, the pupal ecdysone/let-7 pulse induces a substantial remission of the tumorous tissue restoring almost wild-type brain morphology (Fig. 4h).

Discussion
Deregulation of PcG gene expression has been associated with various types of human cancer 3,28 . For example, loss of expression of the human ph homologue has been linked to the formation of osteosarcomas 29,30 . As the molecular mechanisms of PcG proteins in human cancers are largely unknown, understanding the tumour-suppressor function of PcG genes in Drosophila therefore could provide insights in human cancer biology. During showing numerous tumour cells (green) that express Deadpan (Dpn) (red). Scale bar is 50 μm. g Confocal image of a larval brain hemisphere of UAS-RNAi-brat; UAS-EcRB1; note that a large tumour still forms. Scale bar is 50 μm. h Confocal image of an adult brain of UAS-RNAi-brat; UAS-EcRB1; note that the number of green brain tumour cells is significantly reduced. Scale bar is 50 μm. i Model for the intrinsic tumour eviction mechanism mediated by ecdysone-induced miRNA let-7 in ph 505 tumours. In the drawing, ecdysone refers to 20-HE titres during prepupal and pupal stages 11[,12 ; chinmo and let-7 refer to the mRNA expression in the eye-antenna discs. Genotypes: a, d w UAS-dicer2; wor-Gal4 ase-Gal80; UAS-mCD8::GFP; b, e, f w UAS-dicer2; wor-Gal4 ase-Gal80/UAS-RNAi-brat; UAS-mCD8::GFP/+; c w UAS-dicer2; wor-Gal4 ase-Gal80/UAS-RNAi-brat; UAS-mCD8::GFP/UAS-let-7; g, h w UAS-dicer2; wor-Gal4 ase-Gal80/UAS-RNAi-brat; UAS-mCD8::GFP/UAS-EcRB1 the past decades imaginal discs have been used as a powerful paradigm to investigate mechanisms underlying the formation and progression of several types of tumour, including Ras-, PcG-, or Hippo pathway-induced tumours 31 . In addition, it is worth noting that the let-7 consensus sequence is identical from Caenorhabditis elegans to humans, suggesting that let-7 may control functionally conserved targets in regulating proliferation and differentiation during development 15,32,33 . In various types of human cancer, downregulation of one or more let-7 members has been observed [33][34][35][36][37] . Moreover, induced expression of let-7 in cancer cell lines can suppress cell proliferation and tumour growth 33,38 . In the human genome, the let-7 family consists of more than ten members. However, the transcriptional regulation, spatial and temporal expression, and their tissue-specific and/or redundant functions of the let-7 family in human are far more complicated and still remain elusive 15,39 .
Here, we identified an intrinsic mechanism reprogramming tumorigenic to nontumorigenic cells of at least two different tumour types, by marking the cells for destruction in adult Drosophila (Fig. 4i). We found that the steroid hormone-induced miRNA let-7 is a key mediator of this mechanism. Interestingly, let-7 and its target genes, including chinmo have been shown to act as heterochronic genes that regulate developmental transitions 40 . Other findings from our lab indicate that ph 505 tumour cells are reprogrammed from the original larval imaginal disc identity to an early embryonic state 41 . By artificially overexpressing a differentiation factor, these cancer cells can be induced to lose their neoplastic state, however. It appears as if these tumour cells are trapped in an immature condition, unable to differentiate. Pulses of ecdysone are a major timer of developmental transitions in flies and their target, let-7, is for example required for enforcing the terminal cell cycle arrest in pupal stage wing discs 22,28 . Our findings suggest that flies have evolved tumour suppressive mechanisms by inducing let-7-controlled heterochronic gene networks to enforce cellular differentiation in epigenetically derailed tumours (Fig. 4i). Indeed, differentiation therapy is considered a promising approach for curing human cancers 42 . However, the strategy has been applied only in limited cases. As such, our identification of an innate tumour eviction mechanism in flies based on these principles may provide new ideas how such cancer treatments could be further improved in human patients.

Methods
Fly genetics. Fly strains were maintained on standard medium. All genetic experiments were performed at 25°C, except that the brat-RNAi experiments were completed at 29°C to increase the knockdown efficiency. To test the roles of candidate genes (let-7, chinmo, etc.) in the tumorigenic growth of ph 505 mutant cells, we attempted to generate fly strains carrying double mutations including the candidate genes in combination with ph 505 . However, we were unable to establish double mutant strains of ph 505 and another gene mutant. Therefore, in this study, we used transgenic fly strains expressing either RNAi or miRNA sponge to reduce the activities of the target genes.
To generate homozygous ph 505 MARCM clones, virgin females of ph 505 FRT19A/FM7 act-GFP were crossed with tub-Gal80 FRT19A; ey-flp act>STOP>Gal4 UAS-GFP males. For the control MARCM clones, virgin females of y w FRT19A were crossed with tub-Gal80 FRT19A; ey-flp act>STOP>Gal4 UAS-GFP males. Because the ey-flp is expressed not only in the eye-antennal discs, but also in all leg discs and the genital disc, clones were observed in these tissues as well. As a result, GFP-labelled ph 505 cells can be seen in the heads, legs, thorax, and abdomen in the adult flies. Other MARCM clones using different transgenic strains in combination with ph 505 or FRT19A were generated in the same way by crossing corresponding virgin females with tub-Gal80 FRT19A; ey-flp act>STOP>Gal4 UAS-GFP males. Immunohistochemistry and antibodies. All the tissues (larval discs, larval brains, metamorphed ph 505 cells, adult brains, and transplanted tumours) were dissected in cold PBS on ice, fixed in 2% paraformaldehyde (in 1× PBS) for 25 min at room temperature, and washed several times in PBST (1× PBS with 0.5% Triton X-100). Tissues (larval discs, larval brains, or metamorphed ph 505 cells) were incubated overnight with primary antibodies at 4°C, followed by several washes at room temperature, incubated with secondary antibodies at 4°C overnight. Adult brains and transplanted tumours were incubated with primary antibodies for 48 h at 4°C, followed by several washes at room temperature, incubated with secondary antibodies for 48 h at 4°C. After several washes, all tissues were incubated with DAPI (1:200 in PBST) at room temperature for 20 min. Tissues were then mounted in Vectashield and stored at −20°C before analysis.
Microscopy. Immunofluorescent images were recorded on a confocal microscope (TCS SP5; Leica). Adult fly pictures were taken on a Leica MZ16 or a Nikon SMZ1270 microscope. Images were processed using ImageJ, Imaris, Photoshop, and Adobe Illustrator. To measure the volume of tumour clones, confocal images of eye-antennal discs from wandering third instar larvae of the corresponding genotypes were collected and processed in Imaris.

Transplantation.
Transplantation experiments were carried out as previously described 43 . In brief, 4-6-day-old adult w 1118 females were used as hosts (n > 20 for each transplantation). The host flies were immobilized on an ice-cold metal plate and stuck on a piece of double-sided sticky tape, with their ventral sides up. The dissected eye discs or other tumour tissue from adult flies were cut into small pieces and each piece was transplanted into the abdomen of one host using a custommade glass needle. All transplantation was made under a GFP microscope to ensure labelled cells were injected into the hosts. After transplantation, host flies were allowed to recover at room temperature for 1-2 h in fresh standard Drosophila medium before transferred to and maintained at 25°C.
Survivorship measurement. To measure the survivorship of ph 505 and ph 505 ; UAS-EcR DN adult flies, newly eclosed adults were collected and separated into three groups (for ph 505 flies, n = 20 in each group; for ph 505 ; UAS-EcR DN flies, n = 25 in two groups and n = 20 for the third group). Flies were maintained at 25°C, flipped into fresh food vials every 2 days, and the number of living flies were counted on each Monday and Friday during the course of experiment. The numbers of flies were imported into Excel and the survivorship at different times was calculated. The survival curve was produced in Excel.
Quantitative PCR. As previously reported 18 , the expression of let-7 is low at wandering third larval instar but high during pupal stage in response to ecdysone, we dissected eye-antennal discs together with brains from larvae (wt L3) or pupae (wt pupa) as controls. We collected metamorphed ph 505 cells from adults, transplanted tumours from hosts, ph 505 ; UAS-EcR DN cells from adults, and ph 505 ; UASlet-7-SP cells from adults. RNA was extracted from all these samples with mirVana miRNA Isolation Kit (Invitrogen), and reverse transcribed using TaqMan Micro-RNA Reverse Transcription kit (Applied Biosystems) with let-7 RT primer and 2S rRNA reference RT primer. qPCR was performed with TaqMan Small RNA Assays (Applied Biosystems) using each let-7 and 2S rRNA TaqMan Small RNA Assay primer mix (Thermo Fisher Scientific).
Transcriptome analyses. Transcriptome analyses were performed by mRNA sequencing. RNAs were isolated from the metamorphed ph 505 cells, as well as RNAs from the ph 505 tumours transplanted in adult hosts after 4 times, 8 times, and 14 times weekly retransplantation. As a control, we used the wild-type discs that were transplanted into the abdomen of adult hosts and extracted RNAs 1 day after their transplantation. This step was necessary as earlier evidence showed that the transplantation process itself introduced differential expression of a number of genes 44 . RNA was extracted using an Arcturus PicoPure RNA Isolation kit (Applied Biosystems), library prepared with SMARTSeq2 NexteraXT and sequenced on an Illumina NextSeq2500 for control and tumour samples and NextSeq500 for metamorphed samples. Bioinformatics analyses. Short reads were aligned to BDGP dm6 genome assembly using TopHat 2.0.12 45 with parameters "--very-sensitive" for Bowtie 2.2.3. From the aligned reads, differential expression was called using R 3.3.1 with maSigPro 1.46.0 46 . Genes were stratified into expression profile clusters with kmeans clustering. Subsequent biological process gene ontologies for each cluster were found with topGO 47 . Respective p values were calculated with Fisher's exact test. For the hierarchical clustering (UPGMA) underlying the heat map, we used library normalized log2 counts per million reads determined by edgeR 3.16.5 48 on genes involved in the ecdysone response as listed by Flybase with scaling per sample/column. Experimental design. All fly genetic experiments were repeated at least three times. Each transplantation experiment was performed at least twice independently. RNA samples for the transcriptome analysis were collected two or three times independently. RNA samples for qPCR were extracted from the mixture of 5−10 independent animals and qPCR was performed biologically twice.
Sample size was not predetermined before the experiments. Sample size in each experiment was randomized based on the number of viable flies and the amount of other materials. No data were excluded. The investigators were not blinded to group allocation during data collection and analysis.
Data availability. All deep-sequencing data pertaining to this study can be found on GEO GSE101455. Materials and all relevant data from this study are available from the corresponding author upon reasonable request.