BTB-Zinc Finger Oncogenes Are Required for Ras and Notch-Driven Tumorigenesis in Drosophila

During tumorigenesis, pathways that promote the epithelial-to-mesenchymal transition (EMT) can both facilitate metastasis and endow tumor cells with cancer stem cell properties. To gain a greater understanding of how these properties are interlinked in cancers we used Drosophila epithelial tumor models, which are driven by orthologues of human oncogenes (activated alleles of Ras and Notch) in cooperation with the loss of the cell polarity regulator, scribbled (scrib). Within these tumors, both invasive, mesenchymal-like cell morphology and continual tumor overgrowth, are dependent upon Jun N-terminal kinase (JNK) activity. To identify JNK-dependent changes within the tumors we used a comparative microarray analysis to define a JNK gene signature common to both Ras and Notch-driven tumors. Amongst the JNK-dependent changes was a significant enrichment for BTB-Zinc Finger (ZF) domain genes, including chronologically inappropriate morphogenesis (chinmo). chinmo was upregulated by JNK within the tumors, and overexpression of chinmo with either RasV12 or Nintra was sufficient to promote JNK-independent epithelial tumor formation in the eye/antennal disc, and, in cooperation with RasV12, promote tumor formation in the adult midgut epithelium. Chinmo primes cells for oncogene-mediated transformation through blocking differentiation in the eye disc, and promoting an escargot-expressing stem or enteroblast cell state in the adult midgut. BTB-ZF genes are also required for Ras and Notch-driven overgrowth of scrib mutant tissue, since, although loss of chinmo alone did not significantly impede tumor development, when loss of chinmo was combined with loss of a functionally related BTB-ZF gene, abrupt, tumor overgrowth was significantly reduced. abrupt is not a JNK-induced gene, however, Abrupt is present in JNK-positive tumor cells, consistent with a JNK-associated oncogenic role. As some mammalian BTB-ZF proteins are also highly oncogenic, our work suggests that EMT-promoting signals in human cancers could similarly utilize networks of these proteins to promote cancer stem cell states.


Introduction
The Epithelial-to-mesenchymal transition (EMT), a developmental process involved in morphogenesis, organogenesis and wound healing (reviewed in [1]), can be coopted by epithelial cancer cells to gain metastatic potential (reviewed in [2]). Over recent years, it has also become apparent that the activation of an EMT can endow cancer cells with stem cell-like properties essential for tumor maintenance (reviewed in [3]). Triggers driving the induction of EMTs during tumorigenesis are beginning to be elucidated, and can include heterotypic interactions between tumor and associated stromal cells as a result of localized inflammation [4][5][6]. Cytokines such as IL-6 can promote an EMT and endow tumor cells with cancer stem cell properties [7], and TGFβ, which has well-established roles in the induction of EMT, can cooperate with TNF to induce EMT, stemness and tumorigenicity [8]. Well-characterized downstream regulators of the EMT programme include transcription factors of the ZEB, Snail, Twist and NF-κB families [9][10][11], many of which converge upon the repression of E-cadherin expression. How this is linked to self-renewal programmes however, remains poorly understood. The acquisition of cancer stem cell properties induced by inflammation is associated with NF-κB and STAT-dependent pathways [5], however, the down regulation of E-cadherin could also help drive self-renewal through the release of β-catenin, and activation of Wnt signaling. Indeed, the loss of polarized epithelial constraints may promote self-renewal through deregulation of the Scrib cell polarity module, and subsequent activation of the Hippo negative tissue growth pathway effector, TAZ [12]. Deciphering the complex interrelationship that exists between the EMT and self-renewal pathways in cancer cells is a major challenge and will require the use of powerful model systems to tease out the separate and interconnected aspects of these two key developmental properties.

Mosaic analysis
Clonal analysis utilized MARCM (mosaic analysis with repressible cell marker) [47] with FRT82B and ey-FLP1 to induce clones and mCD8-GFP expression to mark mutant tissue. All fly crosses were carried out at 25°C and grown on standard fly media unless otherwise stated.

Analysis of adult midguts
esg-GAL4,tub-GAL80 ts flies were crossed to the UAS-transgene of interest at 18°C. Progeny flies carrying esg-GAL4,tub-GAL80 ts with the transgene of interest were collected over 5 days and stored at 18°C until shifting to 29°C, on standard food, for the specified number of days. Midguts were then harvested for immunohistochemical analysis.

Microscopy and image processing
All samples were analysed by confocal microscopy on an Olympus FV1000 or Leica TCS SP5 microscope. Single optical sections were selected in FluoView software before being processed in Adobe Photoshop CS2 and assembled into figures in Adobe Illustrator CS2.

Expression array and bioinformatic analysis
Eye/antennal discs were dissected from~5 day old larvae bearing either FRT82B control clones, scrib 1 + Ras ACT clones, scrib 1 + Ras ACT + bsk DN clones, scrib 1 + N ACT clones, or scrib 1 + N ACT + bsk DN clones. For the expression array, 50 pairs of discs per genotype were used to prepare RNA. Samples were prepared in triplicate, and the RNA isolated using TRIZOL, before being column purified (Qiagen). Probes were hybridized to Affymetrix GeneChip Drosophila Genome 2.0 Array.
The raw data was analysed using the R [50] statistical software and Bioconductor [51]. The data was loaded, background adjusted and normalized using the Affy [52] and gcRMA [53] packages and the arrays were assessed for quality using the affyPLM [54] package. Following quality assessment a linear model was fitted to the data using the LIMMA [55] package. For each contrast of interest moderated t-statistics were computed using empirical Bayes moderation of the standard errors towards a common value. The resultant p-values were adjusted for false discovery using the Benjamini & Hochberg [56] method and those with a 2 fold change and an adjusted p-value less than 0.05 were considered significant.
Expression arrays were deposited in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) repository as CEL files, under the accession number GSE42938.
We acknowledge our use of the gene set enrichment analysis, GSEA software, and Molecular Signature Database (MSigDB) [57]. GO term enrichment was performed using AmiGO [58].

Results
A comparative microarray analysis identifies JNK-dependent gene expression changes in Ras and Notch-driven tumors Previous analysis of Ras ACT and N ACT -driven tumors in Drosophila had established that continual tumor overgrowth, migration and invasion, throughout an extended larval stage, were dependent upon JNK signaling. Blocking JNK within the tumors, through the ectopic expression of a dominant negative JNK transgene (bsk DN ), restored pupariation, thus curtailing tumor overgrowth during larval development, and blocked invasive cell morphology [16,20,21]. The diverse effects of JNK signaling within the tumor (invasion and overgrowth) are reminiscent of how EMT-promoting signals can activate cancer stem cell properties in human tumor cells. To further understand how overgrowth and mesenchymal cell behaviour could be interlinked in Drosophila tumor cells, we exploited a comparative microarray approach to identify, in a relatively unbiased manner, JNK-induced transcriptional changes within both Ras and Notch-driven tumors.
The expression profile for mosaic eye-antennal discs of equivalent age were determined for scrib -+ Ras ACT and scrib -+ N ACT samples, as well as for the same genotypes expressing bsk DN , and control eye-antennal discs carrying wild type clones (see Materials and Methods). Using a log base 2 fold change>1 and p<0.05 as cut-off values for significantly deregulated genes, we first compared the four tumor samples to the control discs. This revealed that 1203 probe sets were deregulated in scrib -+ Ras ACT tumors, and 761 probe sets in scrib -+ N ACT tumors ( Fig  1A). Of these, 517 probe sets (43% of the Ras tumors, and 68% of Notch-driven tumors) were shared between the two tumor types, indicating considerable genetic similarity. Upon expressing bsk DN within the tumors, and comparing once again to control discs, 629 probe sets were deregulated in the scrib -+ Ras ACT + bsk DN sample (with only 315, or 50%, shared with scrib -+ Ras ACT tumors), and 1086 probe sets were deregulated in the scrib -+ N ACT + bsk DN sample (with only 430, or 40%, shared with scrib -+ N ACT tumors) ( Fig 1B). Thus, blocking JNK exerted a major impact upon the profile of transcriptional deregulation within the tumors.
To specifically focus upon the JNK-dependent changes within the Ras and Notch-driven tumors, we next compared the expression profile of each tumor sample back to their respective bsk DN -expressing samples (Fig 1C and S1 File). This showed that 828 probes were deregulated in scrib -+ Ras ACT tumors compared to scrib -+ Ras ACT + bsk DN , and 1034 probes were significantly deregulated in scrib -+ N ACT tumors compared to scrib -+ N ACT + bsk DN . Gene Set Enrichment Analysis (GSEA) indicated that amongst the JNK-dependent changes in both tumor types there was enrichment for functional roles consistent with the proposed roles of JNK signaling in tumor development. These included; 1) mesoderm and muscle-related gene sets (eg Contractile fibre/Myofibril/Myosin complex), consistent with cell migration and a mesenchymal-like phenotype; 2) Toll signaling/Inflammation (eg Regulation of Toll signaling showing the number of differentially expressed probes (log base 2 fold change>1 and p<0.05) in scrib 1 + Ras ACT and scrib -+ N ACT mosaic eye-antennal discs compared to control FRT82B eye-antennal discs. (B) Venn diagram showing the number of differentially expressed probes (log base 2 fold change>1 and p<0.05) in scrib 1 + Ras ACT + bsk DN and scrib 1 + N ACT + bsk DN mosaic eye-antennal discs compared to control FRT82B eye-antennal discs. (C) Venn diagram showing the number of differentially expressed probes (log base 2 fold change>1 and p<0.05) in scrib 1 + Ras ACT and scrib 1 + N ACT mosaic eyeantennal discs compared to scrib 1 + Ras ACT + bsk DN and scrib 1 + N ACT + bsk DN mosaic eye-antennal discs, respectively. (D) Examples of gene sets identified by GSEA from genes differentially expressed in scrib 1 + Ras ACT versus scrib 1 + Ras ACT + bsk DN mosaic eye antennal discs (top 3 samples), and scrib 1 + N ACT versus scrib 1 + N ACT + bsk DN mosaic eye-antennal discs (bottom 3 samples). (E) Venn diagram showing the number of differentially expressed probes (log base 2 fold change>1 and p<0.05) in scrib 1 + Ras ACT , scrib 1 + N ACT , scrib 1 + Ras ACT + bsk DN and scrib 1 + N ACT + bsk DN mosaic eye-antennal discs compared to control FRT82B eye-antennal discs. (F) Selected genes identified from the 168 probes (corresponding to 103 genes) differentially expressed in scrib 1 + Ras ACT and scrib 1 + N ACT mosaic eye-antennal discs (compared to FRT82B), but not in scrib 1 + Ras ACT + bsk DN and scrib 1 + N ACT + bsk DN mosaic eye-antennal discs (compared to FRT82B). Red genes are upregulated, green genes downregulated, by JNK in the tumors. pathway/Defense response/Immune response), consistent with an inflammation response and the attraction of hemocytes; and 3) Hormone signaling (eg Hormone activity), consistent with the failure of the tumor-bearing larvae to undergo an ecdysone-induced pupariation response ( Fig 1D and S1 File). Thus, the JNK-dependent changes in both Ras and Notch-driven tumors exhibited significant functional overlap, and indeed, 399 probes, or close to a half of the JNKdependent changes (48% of Ras, 39% of Notch) were shared between the two tumor types.
To validate the expression array data, we determined if previously characterized JNKinduced target genes, such as the negative regulators of the pathway, puckered (puc) [59] and scarface (scaf), [60], would be amongst the JNK-dependent changes common to both Ras and Notch-driven tumors. Indeed, both genes were upregulated by JNK within the tumors, thus confirming the arrays' ability to identify bona fide JNK targets (S1 Table). In addition, the GSEA identification of hormone signaling reflected the repression of ecdysone response genes within the tumors, and previous studies have indicated that this is due to the JNK-induced expression of Ilp8 (CG14059), which prevents the release of ecdysone [30,31]. Consistent with this, Ilp8 was upregulated by JNK in both tumor types (S1 Table). The GSEA also indicated significant enrichments for genes associated with motile activity and mesodermal cell fate, and indeed, four previously characterized JNK targets within Drosophila tumors, the Matrix metalloproteinase 1 (Mmp1) [21], Paxillin (Pax) [16], cheerio (cher) [23] and PDGF-and VEGFreceptor related (Pvr) [61], were also amongst the significant JNK-dependent changes identified in the array (S1 Table), as were the PVR ligands, Pvf1 and Pvf2, which are similarly known to be induced by JNK [62]. Finally, JNK signaling within tumors induces the expression of cytokines capable of activating JAK/STAT signaling (Upd1/2/3), which is known to be required for scrib -+ Ras ACT tumor overgrowth [25], and these genes were also induced by JNK within the tumors (S1 Table), further confirming the reliability of the expression array data.
Interestingly, however, although cytokines capable of activating JAK/STAT signaling were induced by JNK within both Ras and Notch-driven tumors, other proliferative pathways known to be associated with JNK activation were not implicated by the array data. Thus, although Drosophila JNK can induce the expression of growth-promoting morphogens during compensatory proliferation, including the Drosophila Wnt and TGFβ homologs, Wingless (Wg) and Decapentaplegic (Dpp), and JNK can also promote Yki-dependent proliferation [28], neither dpp and wg, nor Hippo pathway components (expanded (ex), fat (ft), four-jointed (fj), Merlin (Mer), warts (wts), salvador (sav), yorkie (yki) and thread (th)), were generally perturbed in a JNK-dependent manner (S1 Table). Furthermore, known regulators of cell cycle progression and cell growth (including the Retinoblastoma homologues, Rbf and Rbf2, cycE, cycD, cycA, Myc/diminuitive (dm), E2f1, E2f2) were also not significantly deregulated by JNK signaling within the tumors (S1 Table).
Also notable by their absence from the JNK-induced expression changes within the tumors were known mediators of the EMT. Although JNK targets from the array included multiple regulators of cell migration, Drosophila homologues of Twist (twi), Snail (sna, esg and worniu), E-cadherin (shg), and ZEB1/2 (zfh1, zfh2), were generally not identified as JNK-regulated genes, nor as genes that were significantly deregulated within the tumors compared to wild type control discs, although both dorsal (dl) and Mef2, which act with Twi and Sna to coordinate mesoderm formation in the embryo [63], were induced by JNK within the tumors (S1 Table). Other pathways known to promote EMT in Drosophila involve activation of the GATA factor, Serpent (srp), which downregulates crumbs (crb) expression [64], and inhibition of the Deleted in Colorectal Cancer (DCC) gene homologue, Frazzled (fra) [65,66], a receptor for Netrins (NetA and NetB in Drosophila). Whilst components of these pathways did not show consistent deregulation in both tumors, srp levels were increased by JNK signaling in Notchdriven tumors, and fra was downregulated by JNK in Ras-driven tumors (S1 Table). It is therefore possible that multiple developmental programs, including yet to be identified ones, may cooperate to promote mesenchymal behaviour in response to JNK signaling.
The effects of JNK signaling upon eye-antennal disc differentiation in Ras and Notch-driven tumors We next examined whether JNK signaling influenced the differentiation state of the tumors, since it was possible that the JNK-induced proliferation and mesenchymal behaviour of the tumor cells was also associated with a progenitor-like, or cancer stem cell-like, state. Indeed, although Drosophila eye antennal discs are not known to contain stem cells, they do contain progenitor cells that can overproliferate in response to STAT activity [35]. Previously, we showed that although scrib -+ Ras ACT eye disc tumors failed to express the photoreceptor differentiation marker, Elav (Embryonic lethal abnormal vision), and that blocking JNK (by expressing bsk DN ) was associated with a restoration to Elav expression [16], surprisingly, blocking JNK in scrib -+ N ACT tumors failed to restore Elav expression [16]. This indicated that curtailing tumor overgrowth was not necessarily associated with restoring Elav. The expression array further validated these observations by showing Elav expression was downregulated in both tumor samples, but only restored by blocking JNK signaling in the Ras-driven tumors (Table 1). However, Elav is expressed relatively late with respect to cell fate commitment, and only in cells committed to a photoreceptor fate. It was therefore possible that the continual overgrowth of Ras and Notch-driven tumors was characterized by the failure to upregulate earlier-acting cell fate commitment regulators, and/or the continued expression of progenitor cell markers, and that the expression of these would be normalized by blocking JNK signaling. To determine if this was the case, we used the array data to examine the expression of other markers of cell fate commitment and progenitor cell states in the eye-antennal disc. Differentiation factors included the proneural factor Atonal (Ato), which is expressed just before Elav, and Sine oculis (So), Dachshund (Dac), Eyes absent (Eya), Distal antenna (Dan) and Distal antenna-related (Danr), the expression of which all precede Ato (reviewed in [67]). Dan and Danr are also expressed during antennal disc differentiation, together with the homeodomain protein Distal-less (Dll). Markers of progenitor cells in the eye disc included Homomthorax (Hth), which is downregulated as cells upregulate Dac and Eya, Teashirt (Tsh), Eyeless (Ey), Twin of eyeless (Toy) and Optix.
Interestingly, all six markers of eye-antennal cell fate commitment (ato, dac, dan, danr, Dll, eya and so) were downregulated within both Ras and Notch-driven tumors (Table 1). This indicated that tumor overgrowth was indeed associated with a failure to differentiate. However, blocking JNK within scrib -+ Ras ACT and scrib -+ N ACT tumors, by co-expressing bsk DN , failed to increase ato, dac, dan and so expression in either Ras or Notch-dependent tumors; and although danr, eya and Dll levels were marginally increased in Ras-driven tumors expressing bsk DN , their levels remained downregulated in scrib -+ N ACT + bsk DN discs (Table 1). This suggested that blocking JNK within Ras and Notch-driven tumors was not acting through a common pathway to restore differentiation to the tumor cells. Furthermore, although we hypothesized that progenitor state markers might be increased within the tumors if JNK signaling was promoting their expression to maintain tumor overgrowth, the expression array data indicated that hth, tsh, ey, toy and optix expression were either not significantly altered, or were downregulated, in both Ras and Notch-driven tumors compared to control eye-antennal discs. In fact, blocking JNK signaling in Notch tumors was associated with an upregulation of both hth and toy expression compared to control mosaic discs. Therefore, it did not appear likely that JNK was promoting Ras and Notch-driven tumor overgrowth by maintaining a progenitor cell state characterized by the expression of hth and other known progenitor cell-expressing factors.
To confirm these data, and to specifically observe whether the tumor cells failed to express these proteins, we examined mosaic eye-antennal discs using immunohistochemical analysis with available antibodies directed against the products of three of the differentiation genes, Eya, Dac and Ato, and the progenitor state marker, Hth. This analysis again validated the expression array data, since all three differentiation markers were repressed in scrib -+ Ras ACT (S1 Fig Thus, we conclude that, although both Ras and Notch-driven tumors fail to transition to Dac/Eya expression, the JNK-dependent maintenance of an Hth-dependent progenitor state, was not likely to be key to their continual overgrowth.

The expression of BTB-ZF genes are deregulated by JNK in Ras and Notch-driven tumors
The analysis of cell fate markers in the Drosophila eye-antennal disc indicated that tumor overgrowth was associated with a block to differentiation, but failed to identify specific JNK- effectors, common to both Ras and Notch-driven tumors, that could be involved with maintaining a progenitor-like cell fate. To determine, in a less biased manner, if any other progenitor state factors could be induced by JNK signaling within the tumors, we further mined the array data by narrowing down the list of candidate genes. We did this by not only assuming that key JNK effectors would be common to both Ras and Notch-driven tumors, but also, that upon blocking JNK activity, the expression of these candidates would be normalized to approximately wild type levels. In other words, that by comparing the expression profiles of all four tumorigenic and non-tumorigenic samples back to wild type control discs, these genes would only be significantly deregulated (log base 2 fold change>1, p<0.05) in Ras and Notch-driven tumors, but not in the non-tumorigenic samples expressing bsk DN (Fig 1E). This four-way comparison yielded groupings of probes specifically deregulated by Ras ACT , but not N ACT , expression (eg. EGFR, sty), or by N ACT , but not Ras ACT , expression (eg. HLHm3, Ser, neur), as well as generating a focussed list of 168 probe sets, corresponding to 103 genes, deregulated by JNK and common to both Ras and Notch-dependent tumors (32 genes downregulated; 71 genes upregulated by JNK; Fig 1E and S2 File). Significantly, this JNK signature included Mmp1, Pax, cher and Upd3, as well as other genes known to be induced by JNK activation in other developmental contexts (eg scaf [60]), and genes potentially involved in the invasive phenotype (eg the ARP family member, Actin-related protein 2/3 complex subunit 3B (Arpc3B), and myosin light chain 2 (Mlc2) [68]). Of note, from this list of 103 genes, there was a high enrichment for genes belonging to the BTB-ZF family (4 identified, out of a genome encoding 15 family members): broad (br) and tramtrack (ttk) were repressed in response to JNK signaling, whereas fruitless (fru) and chronologically inappropriate morphogenesis (chinmo) were upregulated in a JNK-dependent manner (S2 Table). BTB-ZF proteins are increasingly implicated in the aetiology of human cancers, as both tumor suppressors and oncogenes, and are also highly oncogenic in Drosophila when ectopically over-expressed. Our own previous screening for Drosophila oncogenes identified the BTB-ZF gene abrupt, as a potent inducer of tumorigenesis when ectopically overexpressed in scrib mutant cells [69]. Although ab was not significantly deregulated by JNK in the expression arrays (S2 Table), the BTB-ZF protein Chinmo bears many striking similarities to the functional activity of Ab. This includes both being targets of let-7 mediated repression, and also regulating the temporal differentiation of neural cells in the brain [38,[70][71][72]. Furthermore, previous analysis had indicated that chinmo expression is associated with a stem cell state; its ectopic overexpression can promote stem cell proliferation, and, in response to JAK/STAT signaling, chinmo is expressed within the progenitor domain of the eye disc and required for eye disc growth and/or proliferation [26,73]. Together, these data suggested that Chinmo could be an important STAT effector of progenitor cell maintenance in the eye-antennal disc tumors, downstream of JNK signaling.
To confirm that the expression of chinmo was upregulated by JNK within the tumors, we examined the activity of a previously characterized enhancer trap reporter for chinmo expression, chinmo-lacZ [73]. As the reporter was inserted on the same chromosome as the Ras ACT and N ACT transgenes, we facilitated this analysis by examining Raf gof -driven tumors (+/bsk DN ), since the Raf gof transgene was on a different chromosome to the chinmo-lacZ reporter, and we had previously shown that Raf gof mimics the effects of Ras ACT in driving scribtumor overgrowth [14]. In wild type discs, chinmo-lacZ was expressed in the centre of the antennal disc, and also in the posterior cells of the eye disc (Fig 2A). However, in scrib -+ Raf gof tumors, chinmo-lacZ was ectopically expressed within the tumor cells ( Fig 2C). Furthermore, the reporter was also expressed in basally located cells that had dropped out of the epithelium, and in cells that appeared to be migrating between the brain lobes (Fig 2E), previous analysis of which had indicated are JNK positive [16]. However, upon expressing bsk DN within the scrib -+ Raf gof tumors, the expression of chinmo-lacZ was normalized, consistent with it's expression being JNK-dependent ( Fig 2D).
Overexpression of chinmo is sufficient to cooperate with Ras ACT or N ACT and drive JNK-independent tumor overgrowth in the eye-antennal disc If Chinmo acts as an important oncogenic mediator of JNK signaling, its ectopic expression might be sufficient to drive tumor overgrowth in cooperation with Ras ACT or N ACT . To determine if this was the case, we used a transgene to ectopically express a full-length version of chinmo, both alone and in combination with Ras ACT or N ACT in eye disc clones. Strikingly, whereas the overexpression of chinmo alone did not prevent organismal pupariation and clones did not overgrow, larvae overexpressing chinmo with either Ras ACT or N ACT often failed to enter pupariation, and massive tumor overgrowth ensued throughout an extended larval stage The ectopic expression of chinmo-lacZ in scrib 1 + Raf gof tumor cells is dependent upon JNK signaling. Mosaic eye-antennal discs, anterior to the right. Clones are generated with ey-FLP, and are positively marked by GFP (green, or magenta when overlaid with white). chinmo-lacZ expression is shown by antibody detection of β-Galactosidase (white, or magenta when overlaid with GFP in the merges). Brain lobes (BL) remain attached to the eye discs in E, and in E and E' tissue morphology is shown with Phalloidin staining to highlight F-actin (white). Yellow scale bar corresponds to 40μm. (A) In control FRT82B eyeantennal discs, chinmo-lacZ is expressed in the centre of the antennal disc and in the posterior half of the eye disc. (B) In scrib 1 mosaic discs, chinmo-lacZ is ectopically expressed in some mutant antennal disc clones (arrowhead). (C) In scrib 1 + Raf gof mosaic discs, chinmo-lacZ is ectopically expressed in mutant clones in the antennal and eye disc (arrowheads). (D) Expressing UAS-bsk DN in scrib 1 + Raf gof clones abrogates the ectopic expression of chinmo-lacZ in the mutant clones of tissue (arrowhead). (E) In scrib 1 + Raf gof tumors, chinmo-lacZ is ectopically expressed in the tumor cells that appear to be migrating between the brain lobes (arrowhead, E' and E''').
doi:10.1371/journal.pone.0132987.g002 (Fig 3). The brain lobes were also markedly enlarged in Ras-driven tumors, consisting of masses of clonal tissue, suggestive of excessive neuroepithelial proliferation (S4 Fig). Thus chinmo is sufficient to drive tumorigenesis in cooperation with ectopic Ras or Notch signaling.
The capacity of chinmo overexpression to cooperate with Ras or Notch in tumorigenesis is consistent with chinmo being an important tumorigenic effector downstream of JNK. We therefore hypothesized that the overgrowth of chinmo-driven tumors throughout an extended larval stage could be independent of JNK activity. Indeed, the overgrown eye-antennal discs of chinmo + Ras ACT or chinmo + N ACT tumors appeared benign, since they remained as separate entities and did not fuse to the brain lobes, consistent with a failure to activate a JNK-dependent invasion pathway. Furthermore, blocking JNK signaling within chinmo + Ras ACT tumors by coexpressing bsk DN in the mutant clones failed to restore pupariation to the tumor-bearing larvae, and the tumors continued to grow throughout an extended larval stage (Fig 3G). Thus, unlike scrib -+ Ras ACT /N ACT tumors, JNK signaling is not essential for chinmo-driven tumorigenesis.
chinmo overexpression is sufficient to block epithelial differentiation in the larval eye disc and promote stem cell/enteroblast overproliferation and cooperation with Ras ACT in the adult midgut Previous studies have revealed that chinmo expression is associated with progenitor-like states in some Drosophila tissues: it is involved in stem cell maintenance in the cyst cells of the Drosophila testis, as well as functioning within a heterochronic pathway controlling the timing of neural differentiation in the brain [38,70,73]. To determine if its oncogenic ability in the eye-antennal disc was also likely to be associated with roles in maintaining a progenitor-like state, we examined the expression of cell fate markers in chinmo-expressing clones. This revealed that the overexpression of chinmo alone was sufficient to block the expression of Dac, Eya and Elav in the eye disc (Fig 4). Similarly, chinmo + Ras ACT tumors were also characterized by the failure to express the differentiation factors, Dac, Eya and Elav (Fig 4). The data are therefore consistent with Chinmo functioning to prime cells towards transformation by blocking differentiation.
Since the eye-antennal disc is not a good model for investigating stem cell properties, we turned to an epithelial tissue that is maintained by stem cell divisions, the midgut of the adult fly. Using esg-GAL4, which is expressed in the epithelial stem cells and their progeny, the enteroblast (prior to the enteroblast differentiating into either an enterocyte and enteroendocrine cell) [74,75], we overexpressed chinmo FL for 5-7 days in adult flies. Strikingly, this significantly increased the number of GFP positive stem cells/enteroblasts within the epithelium, suggesting that ectopic expression of chinmo was able to promote their proliferation. To determine if this also predisposed cells to transformation, we next coexpressed chinmo FL with Ras ACT . Ras ACT does not produce tumors in the adult midgut [76], however, combined expression of chinmo FL + Ras ACT produced massive overgrowth of esg>GFP expressing cells that filled the lumen of the midgut (Fig 5 and S5 Fig), eventually causing host lethality. Thus, the ectopic expression of chinmo can maintain an epithelial stem cell or enteroblast state, which primes cells for transformation by Ras ACT .   Chinmo, and the functionally-related BTB-ZF protein, Abrupt are required for scrib -+ Ras ACT /N ACT tumor overgrowth Our analysis of chinmo overexpression had confirmed its sufficiency to promote tumorigensis in cooperation with Ras ACT or N ACT in different epithelial tissues, and the data were consistent with the possibility of chinmo functioning downstream of JNK in scrib -+ Ras ACT /N ACT tumors to promote overgrowth. To determine if chinmo does play a role in the development of scrib -+ Ras ACT /N ACT tumors, we used an RNAi transgene to knockdown chinmo levels. Immuno-histochemical analysis of Chinmo protein in the wing confirmed that the overexpression of chinmo RNAi significantly reduced Chinmo levels ( S6 Fig). However, the ectopic expression of chinmo RNAi in either scrib -+ Ras ACT or scrib -+ N ACT tumors exerted little effect upon the size or invasive properties of tumors at day 9 (Fig 6). This suggested that chinmo was not significantly rate-limiting for invasive, tumor overgrowth.
It was possible, however, that a functional role for chinmo in tumor formation might be being masked by redundancy with another BTB-ZF protein expressed within the tumors. We therefore turned to the functionally related BTB-ZF protein, Abrupt. Abrupt is also expressed within the eye disc progenitor cells [70,71], and our own previous work had indicated that Abrupt is highly oncogenic when overexpressed in the Drosophila eye-antennal disc [69]. Furthermore, although the transcriptional array did not indicate that ab expression was significantly altered by JNK activity within Ras and Notch-driven tumors (S2 Table), Ab protein was present in basal and migrating cells of Ras-driven tumors (Fig 7), and, using the enhancer trap msn-lacZ as a reporter for JNK activity, Ab was strongly expressed in msn-lacZ positive cells migrating between the brain lobes (Fig 7A and 7B). Immuno-staining to detect chinmo-lacZ expression in scrib -+ Raf gof tumors also showed that Ab protein was present in chinmo-lacZ positive cells tumor cells (Fig 7C). However, consistent with the array data, Ab did not appear to be a JNK-induced gene, since Ab levels were not increased in all msn-lacZ positive tumor cells, and ectopically activating JNK signaling within eye disc clones by expressing an activated allele of JNKK, hep ACT , did not lead to increased levels of Ab (S7 Fig). Thus, although Ab is not likely to be a direct target of JNK signaling, it is co-expressed with chinmo in JNK-positive cells, and could therefore play a functional role in JNK-driven tumor development.
To analyse the role of ab in tumor formation, we used RNAi transgenes to knockdown ab in eye disc clones. This showed a strong reduction in Ab protein levels, thereby validating the RNAi lines (S8 Fig). Strikingly, expression of ab RNAi in scrib -+ Ras ACT or scrib -+ N ACT tumors significantly reduced tumor overgrowth at day 9, thus indicating that Ab was required for tumor development (Fig 6). To next determine if reducing ab activity would expose a functional requirement for chinmo in tumorigenesis, we coexpressed ab RNAi and chinmo RNAi in scrib -+ Ras ACT /N ACT tumors. Indeed, this produced a significantly greater reduction to tumor development at day 9, than ab RNAi alone, and nearly eliminated tumor overgrowth (Fig 6). Thus, when the activity of Ab, a BTB-ZF protein functionally related to Chinmo, is reduced, a key role for Chinmo in tumor development is exposed. fru overexpression, but not br or ttk knockdown, also promotes oncogene-mediated transformation Our analysis of BTB-ZF proteins expressed in the tumors focussed upon Chinmo and Ab, since evidence suggested that they could be important promoters of a progenitor-like cell state. However, other BTB-ZF genes were also identified in the array as being deregulated by JNK signaling in the tumors. fru expression was upregulated by JNK, raising the possibility that fru, like chinmo, could have oncogenic activity; and br and ttk were repressed by JNK in the tumors, Note that the quantification of GFP is based upon the amount of GFP in two-dimensional sections, as opposed to volumetric calculations of the entire tumor mass in three dimensions. It thus underestimates the true extent of tumor size reduction. n for each genotype: FRT82B = 6; scrib -+ Ras ACT = 7, + chinmo RNAi-17156R-1 = 6, + ab RNAi-104582 = 9, + chinmo RNAi-17156R-1 + ab RNAi-104582 = 6, + ab RNAi-4807R-2 = 7, + chinmo RNAi-17156R-1 + ab RNAi-4807R-2 = 12; scrib -+ N ACT = 10, + chinmo RNAi-17156R-1 = 6, + ab RNAi-104582 = 9, + chinmo RNAi-17156R-1 + ab RNAi-104582 = 4, + ab RNAi-4807R-2 = 5, + chinmo RNAi-17156R-1 + ab RNAi-4807R-2 = 2. Error bars are 95% Confidence Intervals (CI). **** p < 0.0001; *** p = 0.0001 to 0.001; ** p = 0.001 to 0.01; * p = 0.01 to 0.05; ns = not significant.
raising the possibility that they could normally represss tumorigenesis, in a similar way to which some mammalian BTB-ZF proteins are known to function as tumor suppressors.
To determine if fru could act as an oncogene in Drosophila, we overexpressed fru in eyeantennal disc clones. The transcriptional regulation of fru is complex, involving multiple isoforms of differentially expressed products, however, the overexpression in clones of a fru isoform known to be normally expressed in the eye disc [45], did not result in massive clonal overgrowth (Fig 8A), although pupariation was often delayed. In contrast, when ectopic fru expression was combined with either Ras ACT or N ACT , massive, but non-invasive, tumor overgrowth ensued during an extended larval stage (Fig 8B and 8C). Similar to chinmo-driven tumors, the overgrowth was at least partly JNK-independent, since pupariation was not restored by co-expressing bsk DN in fru + Ras ACT or fru + N ACT tumors, and tumor overgrowth continued throughout an extended larval stage (Fig 8E and 8F). Thus fru over-expression is sufficient to drive cooperative tumorigenesis in the eye-antennal disc, with a similar potency to chinmo over-expression. In contrast to fru, br and ttk were identified in the expression arrays as genes that were repressed by JNK signaling in the tumors. Indeed, immuno-staining of scrib -+ Ras ACT and scrib -+ N ACT eye-antennal disc tumors confirmed that Br levels were significantly reduced in the tumors, both within the main tumor mass, as well as in tumor cells migrating between the brain lobes (S9 Fig). To test if the downregulation of either br or ttk would be sufficient to elicit cooperative tumor overgrowth in combination with oncogenic signals, we knocked down either br or ttk in clones with RNAi, and co-expressed either Raf gof or N ACT in the knock-down clones. Interestingly, however, although Br and Ttk protein levels were reduced in clones ectopically expressing br RNAi or ttk RNAi , respectively (S10 Fig), neither Raf gof nor N ACT was sufficient to elicit br RNAi or ttk RNAi clonal overgrowth throughout an extended larval stage of development (S10 Fig). Thus, in summary, whilst the overexpression of either fru or chinmo is sufficient to cooperate with Ras or Notch in Drosophila tumorigenesis, the downregulation of BTB-ZF genes does not sensitize cells to Raf or Notch-induced transformation.

Discussion
In this report we have defined the transcriptional changes induced by JNK signaling within both scrib -+ Ras ACT and scrib -+ N ACT tumors by carrying out comparative microarray expression arrays. This showed that JNK exerts a profound effect upon the transcriptional profile of both Ras and Notch-driven tumor types. The expression of nearly 1000 genes was altered by the expression of bsk DN in either Ras or Notch-driven tumors, and less than half of these changes were shared between the two tumor types, indicating that JNK signaling elicits unique tumorigenic expression profiles depending upon the cooperating oncogenic signal. Nevertheless, of the 399 JNK-regulated probe sets shared between Ras and Notch-driven tumors, we hypothesized that these had the potential to provide key insights into JNK's oncogenic activity, and to prioritize these targets, we considered that the expression of the critical oncogenic regulators would not just be altered by bsk DN , but would be normalized to close to wild type levels. This subset of the 399 probe set was identified by comparing the expression profile of each genotype back to control tissue, thereby producing a more focussed JNK signature of 103 genes. Notably, this included previously characterized targets of JNK in the tumors, such as Mmp1, cher and Pax, thereby providing validation of the approach [16,21,23]. Also amongst these candidates were 4 BTB-ZF genes; two of which were upregulated by JNK in the tumors (chinmo and fru), and two downregulated (br and ttk). Focussing upon chinmo, we showed that chinmo overexpression is sufficient to prime epithelial cells for cooperation with Ras ACT in both the eye antennal disc and in the adult midgut epithelium, and that chinmo is required for cooperative Ras ACT or N ACT -driven tumor overgrowth, although it's function was only exposed when it's knockdown was combined with knockdown of a functionally similar BTB-ZF transcription factor, abrupt. This family of proteins is highly oncogenic in Drosophila, since previous work has shown that ab overexpression can cooperate with loss of scrib to promote neoplastic overgrowth [69], and in these studies, we also show that overexpression of a fru isoform normally expressed in the eye disc is capable of promoting cooperation with Ras ACT and N ACT in the eye-antennal disc, in a similar manner to chinmo overexpression. Thus, whether fru also plays a role in driving Ras or Notch-driven tumorigenesis warrants further investigation. Indeed, a deeper understanding of the oncogenic activity of these genes is likely to be highly relevant to human tumors, since of the over 40 human BTB-ZF family members, many are implicated in both haematopoietic and epithelial cancers, functioning as either oncogenes (eg. Bcl6, BTB7) or tumor suppressors (eg. PLZF, HIC1) [77]. Furthermore, over-expression of BTB7, can also cooperate with activated Ras in transforming primary cells [78], and its loss makes MEFs refractory to transformation by various key oncogenes such as Myc, H-ras V12 and T-Ag [79], suggesting that cooperating mechanisms between BTB-ZF proteins and additional oncogenic stimuli might be conserved.

The relationship between JNK-induced overgrowth and stemness in Drosophila tumors
JNK signaling in Drosophila tumors is known to promote tumor overgrowth through both the STAT and Hippo pathways [19, [24][25][26]29]. Deregulation of the STAT pathway was evident in the arrays through the upregulation of Upd ligands by JNK in both Ras and Notch-driven tumors. In contrast, although cher was identified in the arrays as being upregulated in both tumor types and previous studies have shown that cher is partly required for the deregulation of the Hippo pathway in scrib -+ Ras ACT tumors [23], more direct evidence for Hippo pathway deregulation amongst the JNK signature genes was lacking. In part, this could be due to JNK regulating the pathway through post-transcriptional mechanisms involving direct phosphorylation of pathway components [80]. However, the failure to identify known Hippo pathway target genes, and proliferation response genes in general, may simply highlight limitations in the sensitivity of the array assay and the cut-offs used for determining significance, despite it's obvious success in correctly identifying many known JNK targets.
Whether tumor overgrowth through STAT and Yki activity is somehow associated with a stem cell or progenitor-like state remains uncertain. Although imaginal discs exhibit developmental plasticity and regeneration potential, and JNK signaling is required for both of these stem-like properties (reviewed in [81]), there is no positive evidence for the existence of a population of asymmetrically dividing stem cells within imaginal discs [82]. Instead, symmetrical divisions of progenitor cells may be the means by which imaginal discs can rapidly generate enough cells to form the differentiated structures of the adult fly. To date, progenitor cells have only been characterized in the eye disc neuroepithelium. These cells have a pseudostratified columnar epithelial morphology and express the MEIS family transcription factor, Hth, which is downregulated as cells initiate differentiation and begin expressing Dac and Eya. Interestingly, they also require Yki for their proliferation [83], and can be induced to overproliferate in response to increased STAT activity [35]. However, analysis of cell fate markers indicated that tumor overgrowth was not likley to be solely due to the overproliferation of these undifferentiated progenitor cells. Although scrib -+ Ras ACT /N ACT tumors, were characterized by the failure to transition to Dac/Eya expression in the eye disc, blocking JNK in scrib -+ Ras ACT /N ACT tumors did not restore tumor cell differentiation, despite overgrowth being curtailed, and Hth expression was not maintained in the tumors in a JNK-dependent manner. Nevertheless, a JNK-induced gene such as chinmo is likely to be associated with promoting a progenitor-like state, since it is a potential STAT target gene required for adult eye development that is expressed in eye disc progenitor cells in response to increased Upd activity [73,84] and its overexpression alone is sufficient to block Dac/Eya expression. Furthermore, chinmo is also required for cyst stem cell maintenance in the Drosophila testis [73], and our own work has shown that chinmo overexpression promotes increased numbers of esg>GFP expressing stem cells or enteroblasts in the adult midgut. As the BTB-ZF protein Ab is also highly oncogenic and expressed in the eye disc progenitor cells, we hypothesize that the JNK-induced expression of chinmo in scrib -+ Ras ACT /N ACT tumors could cooperate with Ab to maintain a progenitorlike cell state in the eye disc, and that this is required for scrib -+ Ras ACT /N ACT tumor overgrowth. However, although Ab was expressed in chinmo-expressing, JNK positive tumor cells, Ab does not appear to be a JNK-induced gene. What JNK-independent mechanisms control ab expression will therefore require further analysis. Interestingly, we have previously observed that ab overexpression in eye disc clones upregulates chinmo expression [69] and although the effect of chinmo expression upon ab is yet to be described, the data at least suggest that the control of their expression is interlinked in a yet to be defined manner.
Consistent with Chinmo being important for scrib -+ Ras ACT /N ACT tumor overgrowth, chinmo overexpression itself is also highly oncogenic. Over-expression of chinmo with Ras ACT or N ACT drives tumorigenesis in the eye-antennal disc, and also resulted in enlarged brain lobes, presumably due to the generation of overexpressing clones within the neuroepithelium of the optic lobes. In the adult midgut, the overexpression of chinmo with Ras ACT in the stem cell and its immediate progeny, the enteroblast, promoted massive tumor overgrowth, resulting in esg<GFP expressing cells completely filling the lumen of the gut, and eventual host lethality. The luminal filling of esg<GFP cells is reminiscent of the effects of Ras ACT expression in larval adult midgut progenitor cells [74]. Together with the data linking Chinmo function to stem or progenitor cells, these data reinforce the idea that epithelial tumorigenesis can be primed by signals, such as chinmo over-expression, that promote a stem or progenitor cell state.
The function of some Drosophila BTB-ZF proteins including Chinmo and Ab, has also been linked to heterochronic roles involving the conserved let-7 miRNA pathway and hormone signals, to regulate the timing of differentiation [38,[70][71][72]85]. Indeed, Ab can directly bind to the steroid hormone receptor co-activator Taiman (Tai or AIB1/SRC3 in humans), to represses the transcriptional response to ecdysone signaling [85]. Thus, the capacity of BTB-ZF proteins to influence the timing of developmental transitions, particularly if they impede developmental transitions within stem or progenitor cells, could help account for their potent oncogenic activity. Indeed, ecdysone-response genes were repressed by JNK in the tumorigenic state, consistent with the failure of the larvae to pupate and a delay in developmental timing. Whether repressing the ecdysone response cell autonomously might contribute to tumor overgrowth and/or invasion will be an interesting area of future investigation, given the complex role of hormone signaling in mammalian stem cell biology and cancers.

The relationship between JNK-induced invasion and progenitor states
Previous studies have suggested that JNK-dependent tumor cell invasion is developmentally similar to the JNK-induced EMT-like events occurring during imaginal disc eversion [32]. Thus the capacity of JNK to also promote tumor overgrowth is reminiscent of how EMT inducers such as Twist (Twi) and Snail (Sna) are associated with the acquisition of cancer stem cell properties [86]. In Drosophila, however, twi and sna were not induced by JNK in the tumors, although transcription factors involved in mesoderm specification, including the NF-κB homologue, dl (a member of the 103 JNK signature), and Mef2 (a member of the 399 JNK signature), were amongst the up-regulated JNK targets. Mesoderm specification is not necessarily associated with a mesenchymal-like cell morphology, however, dl is involved in the induction of EMT during embryonic development, and both dl and Mef2 act with Twi and Sna to coordinate mesoderm formation [63]. Interestingly, we recently identified dl in an overexpression screen for genes capable of cooperating with scribin Drosophila tumorigenesis [69], and Mef2 has been identified as a cooperating oncogene in Drosophila, and possibly also in humans, where a correlation exists between the expression of Notch and Mef2 paralogues in human breast tumor samples [87]. It is therefore possible that dl and Mef2 either act in combination with Twi or Sna, or independently of them but in a similar oncogenic capacity, to promote a mesodermal cell fate in scrib -+ Ras ACT /N ACT tumors. The potential relevance of this to the mesenchymal cell morphology associated with tumor cell invasion, as well as the acquisition of progenitor states is worthy of further investigation.
In mef2-driven tumors both overgrowth and invasion depend upon activation of JNK signaling [87], suggesting that Mef2 is not capable of promoting invasive capabilities independent of JNK. In contrast, chinmo + Ras ACT /N ACT tumors appeared non-invasive and retained epithelial morphology despite the massive overgrowth, although closer examination of cell polarity markers will be required to confirm this. Furthermore, the overgrowth of chinmo + Ras ACT / N ACT tumors was not dependent upon JNK signaling, suggesting that the maintenance of a progenitor-like state could be uncoupled from JNK-induced EMT-effectors associated with invasion. Whether clear divisions between mesenchymal behaviour and progenitor states in tumors can be clearly separated in this manner is not yet clear, however, overall, it is likely that multiple JNK-regulated genes will participate in both promoting tumor overgrowth as well as migration/invasion. Although we used the 103 JNK signature as a means to focus upon potential key candidates, an analysis of the 399 JNK-regulated probe sets common to both Ras and Notch-driven tumours has the potential to provide deeper insights into the multiple effectors of JNK signaling during tumorigenesis. Whilst the individual role of these genes can be probed with knockdowns, the complexity of the response, potentially with multiple redundancies and cross-talk, will ultimately need a network level of understanding to more fully expose key nodes participating in overgrowth and invasion. This approach has considerable potential to further expose core principles and mechanisms that drive human tumorigenesis, since it is clear that many fundamental commonalities underlie the development of tumors in Drosophila and mammals. Mosaic eye-antennal discs, anterior to the right. Brain lobes (BL) are also shown in (F). Clones are generated with ey-FLP, and are positively marked by GFP (green, or magenta when overlaid with white). Hth or Elav is white, or magenta when overlaid with GFP in the merges. Yellow scale bar corresponds to 40μM. (A-D) In control FRT82B eye-antennal mosaic discs, Hth is expressed in the antennal disc, the progenitor domain of the eye disc, and in the posterior of the eye disc (A, arrowheads). In scrib 1 + Ras ACT tumors, Hth levels are reduced in all three regions (B). In contrast, scrib 1 + N ACT tumors maintain Hth expression throughout the eye disc and show mild ectopic expression (C, arrowheads), this ectopic expression is maintained in scrib 1 + N ACT + bsk DN clones (D, arrowhead). (E-F) Expressing UAS-N ACT in scrib 1 hth P2 double mutant clones results in large clones (E) and does not abrogate tumor development throughout an extended larval stage (compare F to control FRT82B mosaic eye-antennal discs attached to brain lobes in Fig 3A). Mosaic eye-antennal discs, anterior to the right. Clones are generated with ey-FLP, and are positively marked by GFP (green, or magenta when overlaid with white). Ab is detected by immunohistochemical staining (white, or magenta when overlaid with GFP in the merges). (A) The expression of an activated allele of JNKK (UAS-hemipterous(hep) ACT ) in eye-antennal disc clones produces very small clones due to cell death, but the co-expression of the caspase inhibitor UAS-P35 permits the analysis of larger clones of tissue. Ab is normally expressed in the anterior progenitor domain of the eye disc and in the antennal disc, and its levels are not increased in hep ACT + P35 clones. Mosaic eye-antennal discs, anterior to the right. Clones are generated with ey-FLP, and are positively marked by GFP (green, or magenta when overlaid with white). Ab is detected by immunohistochemical staining (white, or magenta when overlaid with GFP in the merges). Yellow scale bar corresponds to 40μM. (A-C) Control FRT82B mosaic discs show the endogenous expression of Ab in the eye progenitor domain (A, arrowhead) and antennal disc. UAS-ab RNAi#104582 expressing clones show decreased Ab protein levels (B, arrowheads). UAS-ab RNAi#4807R-2 expressing clones also show decreased Ab protein levels (C, arrowheads). (TIF) S9 Fig. Br is repressed within Ras and Notch-driven tumors, and within tumor cells migrating between the brain lobes. Mosaic eye-antennal discs, anterior to the right. Brain lobes (BL) are also shown in (D). Clones are generated with ey-FLP, and are positively marked by GFP (green, or magenta when overlaid with white). Br is detected by immunohistochemical staining (white, or magenta when overlaid with GFP in the merges). Tissue morphology is shown with phalloidin staining F-actin in (D, red). Yellow scale bar corresponds to 40μM. (A-D) In control FRT82B mosaic discs, Broad is expressed in both the eye and antennal disc (A). In scrib 1 + N ACT (B, arrowhead) and scrib 1 + Ras ACT tumors (C, arrowhead), Br levels are reduced. scrib 1 + Ras ACT tumor cells, which are known to be active for JNK-pathway activity [16], migrate between the brain lobes and do not express Br (D). (TIF) S10 Fig. Knockdown of br or ttk in eye-antennal disc clones is not sufficient to promote cooperation with Raf gof or N ACT in tumorigenesis. Mosaic eye-antennal discs, anterior to the right. Clones are generated with ey-FLP, and are positively marked by GFP (green, or magenta when overlaid with white). Br, Ttk and Elav are detected by immunohistochemical staining (white, or magenta when overlaid with GFP in the merges). Yellow scale bar corresponds to 40μM. (A-C) UAS-br RNAi#104648 expressing clones show decreased Br protein levels (A, arrowheads). Co-expressing UAS-br RNAi#104648 in UAS-Raf GOF (B) and UAS-N ACT (C) clones does not result in tumorigenesis. (D-F) Ttk levels are reduced in UAS-ttk RNAi#101980 expressing clones (D, arrowheads), and co-expression of UAS-ttk RNAi#101980 in UAS-Raf GOF (E) and UAS--N ACT (F) clones does not result in tumorigenesis. (TIF) S1 File. List of differentially expressed probes (log base 2 fold change>1 and p<0.05) in scrib 1 + Ras ACT mosaic eye-antennal discs compared to scrib 1 + Ras ACT + bsk DN discs, and scrib 1 + N ACT mosaic eye-antennal discs compared to scrib 1 + N ACT + bsk DN discs (see Fig  1C). Of the 1463 probes deregulated in both comparisons, 429 probes (pattern 2) are unique to scrib 1 + Ras ACT versus scrib 1 + Ras ACT + bsk DN discs, 635 probes (pattern 1) are unique to scrib 1 + N ACT versus scrib 1 + N ACT + bsk DN discs, and 399 probes (pattern 3) are common to both scrib 1 + Ras ACT versus scrib 1 + Ras ACT + bsk DN and scrib 1 + N ACT versus scrib 1 + N ACT + bsk DN discs. 17489 probes were not significantly deregulated in either comparison. (CSV) S2 File. List of differentially expressed probes (log base 2 fold change>1 and p<0.05) in scrib 1 + Ras ACT , scrib 1 + N ACT , scrib 1 + Ras ACT + bsk DN and scrib 1 + N ACT + bsk DN mosaic eye-antennal discs compared to control FRT82B mosaic eye-antennal discs (see Fig 1E). Of the 2117 deregulated probes, 168 probes (pattern 12) were deregulated in both scrib 1 + Ras ACT and scrib 1 + N ACT mosaic eye-antennal discs (compared to the FRT82B control), but not in scrib 1 + Ras ACT + bsk DN and scrib 1 + N ACT + bsk DN mosaic eye-antennal discs (compared to the FRT82B control), and are considered to represent a JNK-signature. (CSV) S1 Table. Expression of candidate genes in scrib -+ Ras ACT and scrib -+ N ACT tumors (+/bsk DN ) compared to control FRT82B eye-antennal discs, and in scrib -+ Ras ACT and scrib -+ N ACT tumors compared to their respective genotypes expressing bsk DN . (DOC) S2 Table. Expression of BTB-ZF genes in scrib -+ Ras ACT and scrib -+ N ACT tumors (+/bsk DN ) compared to control FRT82B eye-antennal discs.

(DOC)
Drosophila RNAi Centre (VDRC), the National Institute of Genetics (NIG) Fly Stock Centre, the Developmental Studies Hybridoma Bank (DSHB) for contributing fly stocks and/or reagents, and Flybase.