Identification of the Wallenda JNKKK as an Alk suppressor reveals increased competitiveness of Alk-expressing cells

Anaplastic lymphoma kinase (Alk) is a receptor tyrosine kinase of the insulin receptor super-family that functions as oncogenic driver in a range of human cancers such as neuroblastoma. In order to investigate mechanisms underlying Alk oncogenic signaling, we conducted a genetic suppressor screen in Drosophila melanogaster. Our screen identified multiple loci important for Alk signaling, including members of Ras/Raf/ERK-, Pi3K-, and STAT-pathways as well as tailless (tll) and foxo whose orthologues NR2E1/TLX and FOXO3 are transcription factors implicated in human neuroblastoma. Many of the identified suppressors were also able to modulate signaling output from activated oncogenic variants of human ALK, suggesting that our screen identified targets likely relevant in a wide range of contexts. Interestingly, two misexpression alleles of wallenda (wnd, encoding a leucine zipper bearing kinase similar to human DLK and LZK) were among the strongest suppressors. We show that Alk expression leads to a growth advantage and induces cell death in surrounding cells. Our results suggest that Alk activity conveys a competitive advantage to cells, which can be reversed by over-expression of the JNK kinase kinase Wnd.

The receptor tyrosine kinase anaplastic lymphoma kinase (ALK) was initially described as oncogenic driver in anaplastic large-cell lymphoma and is implicated in a wide range of cancer types including non-small cell lung cancer and neuroblastoma 1 . During development, the vertebrate ALK receptor is activated by ALKAL (also known as FAM150 or Augmentor) family proteins while LDL-domain protein ligands activate Alk in invertebrates [2][3][4][5][6][7] . In contrast, cytosolic ALK fusion proteins, such as NPM1-ALK and EML4-ALK, are constitutively active and common drivers of tumor progression in anaplastic large-cell lymphoma and non-small cell lung cancer 1,[8][9][10] . These ALK fusion proteins are ligand-independent and often ectopically expressed in tissues that would normally not express ALK 1 . In addition to the activation of ALK by fusion events, the full length receptor can be activated by point mutations in its kinase domain which has been described in familial and sporadic neuroblastoma 1,[11][12][13][14][15][16] . Clinical studies suggest that ALK activation drives tumor growth and indeed ALK-driven cancers respond well to initial therapy with ALK tyrosine kinase inhibitors 17 . However, ALK-driven tumors frequently become resistant to ALK inhibitor treatment, underscoring a need to better understand signaling events that occur during aberrant ALK activity.
To shed light on these signaling events, we and others have used various model organisms, such as Drosophila melanogaster, to complement in vitro cell culture studies. In Drosophila, a single Alk orthologue is required to specify the muscle founder cells of the larval gut musculature 2,3,18 . In addition, Alk signaling plays a complex role in many aspects of neuronal function [19][20][21][22][23][24][25] . Interestingly, ectopic expression of Alk in the developing fly eye results in a characteristic phenotype reflecting the activation state of the receptor 26,27 . Therefore, the fly eye has become a valuable model system to analyze putative activating human ALK mutations identified in neuroblastoma patients [28][29][30][31] . The sensitivity of the fly eye model to both Drosophila and human ALK activity suggests that Scientific RepoRtS | (2020) 10:14954 | https://doi.org/10.1038/s41598-020-70890-6 www.nature.com/scientificreports/ a genetic modifier screen in the fly holds the potential to reveal important downstream factors of Alk signaling in vivo. We crossed flies ectopically expressing Alk under control of the sevEP-Gal4 driver to flies carrying genetically mapped deficiencies from the DrosDel collection 32 and scored progeny for suppression or modification of the Alk-induced eye phenotype. Among the strongest suppressors, we identified gain of function alleles of the LZK, MAP3K13 (leucine zipper bearing kinase, mitogen-activated protein kinase kinase kinase 13) and DLK, MAP3K12 (dual leucine zipper kinase, mitogen-activated protein kinase kinase kinase 12) homolog Wallenda (wnd). Wnd has previously been shown to regulate JNK-mediated cell death in Drosophila 33 . JNK signaling in Drosophila tumorigenesis models can promote either cell-invasion or apoptosis dependent on context 34,35 . Investigating cell death in the context of Alk expression revealed increased cell death in Alk-negative neighboring cells, suggesting that ectopic Alk signaling leads to a competitive advantage that is further strengthened by the finding that Alk-expressing clones have a slight, but significant, growth advantage. This Alk-dependent increased competitiveness can be abrogated by elevated levels of JNK activation, through expression of Wnd or the JNKK Hemipterous (hep) in Alk expressing cells, suggesting that Wnd restrains Alk anti-apoptotic signaling output via modulation of JNK.

Results
Ectopic expression of Alk interferes with Drosophila eye development. We previously showed that the signaling activity of Drosophila and human ALK receptors can be assessed in the fly eye 4,[26][27][28][29][30][31]36 . When Drosophila Alk was ectopically expressed in the developing eye under control of the sevEP-Gal4 driver, we observed severe loss of ommatidia and patches of necrotic tissue in the anterior part of the adult fly eye as well as a "rough-eye-morphology" in the posterior part ( Fig. 1A-C). This phenotype was suppressed when co-expressing an RNAi construct (UAS-Alk JF02688 ) targeting Alk (Fig. 1D). In order to characterize the sevEP-Gal4 driver further, we performed lineage analysis using the G-TRACE system 37 , which revealed driver activity posterior to the morphogenetic furrow in third instar larval eye discs ( Fig. 1E-E''). The presence of sevEP-Gal4 > UAS-RFPexpressing and EGFP-positive, lineage derived cells suggested that the sevEP reporter resembles the transient expression pattern of sevenless in early photoreceptors and remains active only in a subset of retinal cells [38][39][40] . This observation was more apparent at later pupal stages when G-TRACE analysis indicated continued RFP expression in the R7 photoreceptor and the cone cells that are recruited later to the ommatidial cluster while other cells of the ommatidium solely expressed EGFP (Fig. 1F). To further analyze the effects of ectopic Alk expression on eye development, we dissected eye discs 50 h after pupa formation (apf) (Fig. 1G-J'). Activity of the sevEP-Gal4 driver revealed by UAS-lacZ (Fig. 1G,G''') resembled the previously described pattern of sevenless in R1, R6, R7 photoreceptors and cone cells [38][39][40] . In control eye discs, DECad staining indicated a regular arrangement of ommatidia upon β-Galactosidase expression ( Fig. 1G'' ,G'''). Cut (CT), a lineage marker for cone cells, was expressed in four cells of each ommatidium (Fig. 1G' ,G'''; quantified in Fig. S1A,B). The neuronal marker Elav was expressed in all photoreceptor cells (Fig. 1I') and Pros was expressed only in R7 cells (Fig. 1I,I'; quantified in Fig. S1C,D). In sevEP-Gal4 > UAS-Alk/ + eye discs, the expression pattern of these markers was highly irregular ( A genetic modifier screen identifies targets of Alk-signaling in Drosophila. We conducted a genetic modifier screen ( Fig. 2A) based on the rationale that dose-reduction of factors acting downstream of Alk would modify the observed eye phenotype. We first tested a number of potential suppressors in a validation screen (Fig. S2). These included loss-of-function (lof) alleles of jelly belly (jeb) encoding the Alk-activating ligand in Drosophila 2,3,18,41 , and RTK-associated signaling molecules such as C3G 42,43 , Son of sevenless (Sos) 44 , 14-3-3ε 45 , Ras oncogene at 85D (Ras85D) 46 , kinase suppressor of Ras (ksr) 47 , rolled (rl) encoding for Drosophila ERK1,2 48 , and pointed (pnt) 49 . When crossed into the sevEP-Gal4 > UAS-Alk/ + background, single copies of these alleles suppressed the eye phenotype to varying degrees ( Fig. S2A-H), indicating that genome-wide screening of deficiency strains should allow us to identify genomic loci implicated in Alk signaling.
We screened 388 lines from the DrosDel collection of deficiencies 50 covering about 60% of the annotated Drosophila genome and scored 58 modifying genomic regions covered by 133 deficiencies (Fig. 2B, green; Table S1). Ninety-eight deficiencies suppressed the Alk-induced eye phenotype while ten deficiencies led to other modifications such as enhanced necrosis in the eye (Fig. 2B, red; Table S1). We did not obtain adult progeny with 25 deficiencies, of which most were located on the X-chromosome. Further genomic mapping revealed 38 individual loci within these deficiencies that modified the Alk-induced eye phenotype (Fig. 3B, Table S1, see also discussion section). Based on this information, we further tested related factors leading to the identification of 12 additional loci that were not covered by the originally screened DrosDel deficiencies (Table S1). For several modifying deficiencies we were not able to map down individual loci that modified the Alk-induced eye phenotype (Table S1). PANTHER14.0-driven gene ontology (GO) enrichment analysis (https ://geneo ntolo gy.org/, Dec 2019) 51 of all Alk modifiers (Fig. 3) revealed a high abundance of factors associated with optic placode development and signal transduction as well as cell dedifferentiation and negative regulation of apoptosis and autophagy (Fig. 3A, Table S1). Among these modifiers, we identified Ras/Raf/ERK network components such as MESR3, cnk and aveugle (ave, also referred to as hyphen) (Fig. 3B). Other identified signaling molecules www.nature.com/scientificreports/ included PI3K21B, STAT92E, the protein serine/threonine phosphatase alphabet (alph), split ends (spen), Star and moleskin (msk) (Fig. 3B). A number of transcription factors also modified the Alk-induced eye phenotype including the forkhead box, sub-group O transcription factor foxo, tailless (tll), eyeless (eye) and eyes absent (eya). We also identified lilliputian (lilli) and factors involved in regulation of RNA polymerase II-induced transcription (TFIIA-S, TFIIA-L, TFIIEα, TFIIEβ and the TFIID-associated factors Taf4, Taf6, TFIIS, Taf12L) (Fig. 3B). In some cases, more than one gene within a suppressing deficiency was found to suppress Alk-mediated signaling. For example, while mapping down the suppressor contained within deficiency ED6346 we noticed that mutations in both the tailless (tll) transcription factor 52 and the warts/lats serine/threonine kinase 53,54 suppressed Alk. We also noticed that certain balancer chromosomes modified the Alk-induced phenotype. In case of the CyO chromosome, we identified mutations in the Duox (Curly) gene as moderate suppressors (Table S1). www.nature.com/scientificreports/ The activity of human oncogenic ALK is modulated by the identified modifiers. Aberrant ALK activation occurs in a range of cancers 55 . We therefore tested whether the modifiers found in our screen were also able to modulate signaling output from human ALK. To do this we activated human ALK ( 1C). Next, we tested the ability of loss of function alleles of 48 candidate genes to modify the human ALK-induced eye phenotype. Many alleles identified in the initial modifier screen also affected the activity of all human ALK variants tested (Fig. 4F), while dose-reduction of ave, cnk, exex, foxo, jeb, Mask, MESR3, Nhe2, Pi3K21B, Ret, spen, Taf12L, and Thor had no effect. Interestingly, P{XP} d00622 and Df(3L)Exel6135 suppressed the phenotype induced by both hALK + ALKAL1 and hALK-F1174L and since neither P{XP} d00622 nor Df(3L)Exel6135 had previously been connected to the function of a particular gene, we decided to analyze these genomic aberrations in more detail.
UAS-sequences are also present on the Df(3L)Exel6135 chromosome that harbors a hybrid P{XP-U}Exel6135element immediately upstream of the wnd transcription start site 57 . We therefore reasoned that the sevEP-Gal4 ). However, fused ommatidia (Fig. 5I, circle) and ommatidia with a reduced number of photoreceptor cells (Fig. 5I, arrow) could still be observed in the eye discs of these animals (Fig. 5I). Notably, UAS-wnd transgene expression was also able to rescue the rough-eye phenotype observed in all three activating conditions of human ALK, likely reflecting higher levels of expression of the UAS-wnd transgene compared to P{XP} d00622 and Df(3L)Exel6135 (Table S1). Wnd has been described as a cell death regulator in the JNK pathway, and over-expression of UAS-wnd with GMR-Gal4 results in cell death and a characteristic "small-eye" phenotype 33 . In line with these findings, we observed a "small-eye" phenotype when we combined GMR-Gal4 with either P{XP} d00622 or P{XP-U}Exel6135 (Fig. 5J,N, controls shown in Fig. S3). This "small-eye" phenotype was suppressed by wnd-specific RNAi supporting that both P{XP} d00622 or P{XP-U}Exel6135 result in ectopic expression of Wnd when combined with a Gal4-driver (Fig. 5K,O). In agreement with the proposed role of Wnd in the JNK-pathway, co-expression of either a dominant-negative variant of the Jun kinase Basket (UAS-bsk DN ) or the JNK-antagonist protein phosphatase Puckered (UAS-puc) suppressed the "small-eye" phenotype (   Fig. S4), as did expression of puckered (UASpuc) (Fig. 6H), a negative regulator of JNK.

Alk expression induces non-cell autonomous apoptosis. Our previous experiments indicated that
accumulation of Alk expressing cells in the anterior part of the fly eye correlated with the loss of ommatidia organization (Fig. 1). Therefore, it could be possible that Alk-negative neighboring cells play an important role in restricting the expansion of Alk-positive cells, eventually suppressing the phenotype in the posterior region of the eye. Supporting this assumption, broad expression of the cell death inhibitors Diap1 and p35 under direct control of the GMR promoter but not under sevEP-Gal4 control, strongly suppressed the Alk-induced eye phenotype ( Fig. 7A-E, controls in Fig. S5). Moreover, reducing the dose of hid, grim and rpr using the deficiency Df(3L)H99 resulted in strong suppression of the Alk-induced eye phenotype (Fig. 7F), further suggesting that protecting Alk-negative neighbors from cell death enhances their ability to resist the Alk expressing cells.
To analyze this further, we used the TRE-JNK activity reporter system 59 in sevEP-Gal4 > UAS-Alk/ + eye discs. No TRE-JNK activity was observed in control 50 h hpf eye discs expressing UAS-YFP under the control of the sevEP-Gal4 driver (Fig. 7G-G''). In Alk-expressing discs however, the TRE-DsRed reporter was activated in both YFP-negative and YFP-positive cells (Fig. 7H-H''). Co-expression of Wnd in sevEP-Gal4 > UAS-Alk, UAS-wnd/ + eye discs resulted in strong TRE-DsRed reporter activity in Alk-expressing cells (Fig. 7I-I''). One explanation for these observations could be that sevEP-Gal4 > UAS-Alk expressing cells acquire a competitive advantage that is abrogated by Wnd-induced JNK activity. The role of JNK in cell competition involves both pro-and anti-tumorigenic responses and depends on the duration of JNK activation as well as on the genetic context 60,61 . For instance, JNK activation has been observed in invasive and migrating cells in Drosophila tumor models 62,63 , but also efficiently triggers death of cells with reduced fitness during cell competition [64][65][66] . To discriminate between these effects we analyzed cell death in sevEP-Gal4 > UAS-Alk expressing eye discs with antibodies against cleaved Dcp-1, an effector caspase downstream of Dronc 67 . Increased Dcp-1 staining could already be observed in third instar larval eye discs, when compared to the controls (Fig. 8A-B'). G-TRACE analysis further suggested that many Dcp-1 positive cells were in close vicinity to, but not derived from, the sevEP-Gal4 lineage www.nature.com/scientificreports/ (Fig. 8B,D). Increased Dcp-1 staining could still be observed in sevEP-Gal4 > UAS-Alk (Fig. 8D,D') but not in control ( Fig. 8C-C') eye discs at pupal stages (50 h apf; quantified in Fig. 8E), suggesting that Alk-expressing cells www.nature.com/scientificreports/ induce cell death in neighboring cells from early stages of ectopic Alk expression. To further test the hypothesis that ectopic Alk-expression provides a competitive advantage at the expense of cell death in neighboring cells we used the hsFLP/Ay-Gal4 system 68 to induce Alk-expressing clones in developing wing discs. Wing discs with Alk-expressing clones displayed significantly increased numbers of Dcp-1-labelled cells when compared to control discs with only GFP-expressing clones (Fig. 8F-G''' , quantified in 8H). In addition, the number of Dcp-1-positive cells within a range of 8-10 cells around Alk-expressing clones was also significantly increased compared to GFP-clone controls (Fig. 8I). Moreover, Alk-expressing clones were slightly increased in size relative to GFP controls, which further supports a competitive advantage of Alk-expressing cells (Fig. 8J). Taken together, these results show that expression of Alk results in cells with enhanced fitness and suggest a role for Alk in non-cell autonomous induction of cell death in surrounding cells.

Discussion
Malignant transformation by ALK involves several mechanisms including hyper-activation of Ras/MAPK-, PI3K-and STAT3 signaling pathways as well as increased expression of transcriptional regulators and stem-cell factors, which shifts the affected cell to a more undifferentiated state [69][70][71][72] . In this study, we conducted a genomewide deficiency screen in Drosophila to identify modifiers of Alk signaling, resulting in the identification of many known and anticipated signaling molecules, such as Ras/Raf/ERK network components (MESR3, Cnk, Aveugle), PI3K21B, STAT92E, and the protein serine/threonine phosphatase Alphabet which is in agreement with previous screening efforts 73 . Amongst the signaling molecules we also found split ends (spen) 74 and Star 75,76 , consistent with their known roles in RTK signaling 73 . Another gene identified as a suppressor of the Alk-induced eye phenotype was moleskin, which is a homologue of the vertebrate Importin-7, and has been proposed to be important for the nuclear import of activated ERK 77 . Among the identified transcription factors, Foxo and the orphan nuclear receptor transcription factor Tailless (Tll) were modifiers of Alk signaling activity. Nuclear localization of Foxo is regulated by Alk-driven AKT/PI3K signaling in insulin-producing cells in the Drosophila brain 23 and, similarly, oncogenic ALK variants affect FOXO3a localization in human anaplastic large-cell lymphoma (ALCL) and neuroblastoma cell culture models 71,78 . In addition to its function in terminal patterning in the Drosophila embryo 52,79 , tll is expressed in the optic lobe primordium 80 , where it controls cell-fate decisions during Bolwig's organ and optic lobe precursor development 81 . Interestingly, the mammalian Tll homolog TLX/ NR2E1 acts as self-renewal factor in neural stem cells and neuroblastoma tumor spheres, and elevated expression of TLX/NR2E1 is correlated with an unfavorable prognosis in neuroblastoma patients 82,83 . We also identified tissue-specific transcription factors, like the PAX6 homolog Eyeless and Eyes absent (EYA) which have been linked to human neuroblastoma 84,85 . Moreover, eight factors involved in initiation of transcription from RNA polymerase II regulated promoters (TFIIA-S, TFIIA-L, TFIIEα, TFIIEβ and the TFIID-associated factors Taf4 (Taf II 110), Taf6 (Taf II 60), TFIIS (TCEA1), and Taf12L (Taf II 30α-2)), were revealed as downstream effectors of Alk signaling. However, mutations in core TFII-complex members could lead to a general reduction of transcriptional activation that would also affect Alk transgene expression. Similarly, mutations of the AF4/FMR2 family (AFF) homolog lilliputian (lilli) were strong suppressors of the Alk-induced eye phenotype 86,87 . As for the identified TFII-complex members, suppression by lilli alleles should be interpreted with caution as they affect hsp70-and sE-promotor-dependent transgene expression and are therefore frequently found as suppressors in genetic overexpression screens 87 . Most strikingly, inducible gain of function alleles of the MAP3K Wallenda (Wnd) that were serendipitously identified during the candidate mapping process were amongst the strongest suppressors. Wnd is one of several JNKKKs in Drosophila and has previously been shown to induce JNK-mediated cell death 33,34,88 . Testing other components of the JNK signaling pathway, we noted that only activators of the pathway, such as the Wnd JNKKK and an activated version of the Hep JNKK, were able to suppress the Alk-induced eye phenotype, while transgenes that inhibit JNK signaling had no effect, suggesting that activation of the JNK pathway leads to suppression of the phenotype. Interestingly, Wnd was more effective than Tak1 in suppressing the Alk-induced eye phenotype, which may reflect earlier observations that the JNKKKs provide context-dependent output, in contrast to Hep, which is part of the core JNK signaling module 33,89 . These observations led us to hypothesize that Wnd overexpression in our sevEP-Gal4 > UAS-Alk model might have shifted the cells into a state where competitive advantages gained by ectopic Alk signaling are abrogated upon high levels of JNK pathway activity. A possible mechanism how Alk-expressing cells acquire this advantage could be the repression of pro-apoptotic factors like Hid upon increased Ras/ERK and PI3K/Akt signaling which not only involves direct phosphorylation of Hid by MAPK/ERK but also transcriptional regulation by Foxo and E2F1 which were also found as modifiers in our screen 90 . Alk-expressing cells activate multiple signaling pathways that result in promotion of the protumorigenic response, and we know from previous work that Alk overexpression leads to activation of MAPK signaling in both eye discs and clones of imaginal discs 26 .
Cell death is a central mechanism in the process of cell competition and the JNK signaling pathway is thought to play an important role 64,91 . Work from many investigators has identified numerous context-dependent effects of JNK signaling in this process, including scenarios where cells with higher fitness ("winners") eliminate their lesscompetitive neighbors by activating the JNK-pathway in these "loser-cells" 34,35,65,66,91 . Conversely, JNK activation can lead to increased cellular fitness if the apoptotic pathway downstream of JNK is blocked 60,92 . A number of genes with oncogenic potential have been reported to confer a competitive advantage on Drosophila cells. One of the best studied examples is Myc that transforms cells into super-competitors at the expense of surrounding cells that are induced to undergo apoptosis 66,93 , which has been reported to be mediated by both JNK-dependent 66 and JNK-independent mechanisms 93 . A number of observations support a role for Alk in conveying a competitive advantage. Firstly, inhibiting cell death in all cells of the eye disc substantially rescued eye morphology, while blocking cell death in Alk-expressing cells alone had no effect. Secondly, expression of Alk led to an increased www.nature.com/scientificreports/ TRE-DsRed signal in eye discs. Thirdly, Alk-negative neighbors appear to be sensitized to undergo cell death as reflected in the increased Dcp-1 staining of wing discs containing Alk-expressing clones. Intriguingly, apoptosis is increased in the Drosophila eye when EGFR-triggered cell differentiation is blocked 94,95 , suggesting that undifferentiated cells involved in cell competition may be more sensitive to JNK-pathway activation, and we noted more Dcp-1 staining in the third instar larval eye disc in the region that is closest to the morphogenetic furrow which may reflect this. Finally, clones expressing Alk were slightly, but significantly increased in size compared with controls. We also noted TRE-activity in Alk-expressing cells of the sevEP lineage as well as YFP-negative cells which could resemble different contexts of JNK activation in the two cell populations. In that case, it would be plausible to speculate that Alk signaling could protect cells by activating the Ras/Raf/ERK pathway which in turn could inhibit pro-apoptotic effector genes. Repression of apoptosis by EGFR signaling has indeed been observed in the developing eye disc and revealed the pro-apoptotic gene hid as direct target of the Ras/MAPK pathway [96][97][98] . In addition, suppression of apoptosis by activated ALK in a ALK-F1174L/MYCN-driven zebrafish neuroblastoma model 99 , and pro-apoptotic responses in human cell cultures after ALK inhibitor treatment 100,101 have been reported further strengthening a connection between Alk-signaling and cell-intrinsic inhibition of apoptosis. These findings shed light on a previously unappreciated potential of Alk to induce cell death in the surrounding environment that may ultimately promote growth of Alk-expressing cells. Whether this is also true in human ALK-driven cancer remains to be elucidated, although our initial analysis with ALK-F1174L and the NPM1::ALK fusion oncogenes that are known oncogenic variants in human cancer would suggest that this may be the case. In mammalian cells, ALK signaling leads to activation of multiple signaling pathways, including anti-apoptotic responses 55 that may act in concert to convey a competitive advantage. Moreover, as mentioned above, Drosophila Myc transforms cells into super-competitors 66 , which may be relevant in human cancers such as neuroblastoma where ALK is known to cooperate with MYCN to promote tumorigenesis 11,55 . In conclusion, while our screening efforts identified many genes that will require further analysis to define their role in Alk signaling, it also allowed us to identify a role for Alk in cell fitness and competition that may be of relevance in human ALK-driven cancer.

Methods
Drosophila melanogaster stocks and genetics. Standard Drosophila husbandry procedures were followed 102 . Stocks were maintained on a potato-mash-based diet at room temperature and crosses were performed in constant climate chambers at controlled 25 °C and 60% humidity levels, 12:12 h day and night cycle. Generation of jeb™ .3xOLLAS flies. We used CRISPR/Cas9 genome editing to induce homology-directed repair (HDR) in the Drosophila jeb locus 108,109 . The pBluescriptII KS [-] HDR donor construct (GenScript) contained homology arms (corresponding to 357 bp immediately upstream and 406 bp downstream of the jeb stop codon) that flanked a central segment encoding for a six amino acid linker (Gly-Ser-Gly-Gly-Ser-Ala), the transmembrane domain of Drosophila Neurotactin (Ala325-His347), a triple OLLAS-tag and a TAA stop codon. Two CRISPR target sites (AGC GGA TGC GTT ACT ACG CTGGG and TGT GTG TGC TGT GTA TTA GACGG) in close proximity to the endogenous jeb stop codon were identified using the flyCRISPR optimal target finder tool 109 . Targeting gRNAs were cloned into pU6-BbsI-chiRNA (Addgene) as described 109 . Donor and U6-chiRNA plasmids were injected into y 1 M{vas-Cas9}ZH-2A embryos (BestGene Inc.). HDR-events were identified by single fly PCR screening using a primer combination specific to the knocked-in sequence (5′-GTT CCT CTT GCA GAG TGT CCT-3′ and 5′-CTA ATT GTT TGT CA-ATG TGA GTT CGC T-3′). Candidates were further analyzed by Sanger sequencing (GATC Services-Eurofins Genomics) and tested for non-complementation with the jeb weli null allele 18,36 . Construction of the UAS-hNPM1::hALK transgene. We employed the deduced amino acid sequence of the human nucleophosmin 1 (hNPM1)-human anaplastic lymphoma kinase (hALK) chimera reported by Morris et al. ( 8 , GenBank: AAA58698.1) as template to generate the UAS-hNPM1::hALK transgene. gBlock fragments (Integrated DNA Technologies) of the hNPM1 n-terminus and the hALK kinase domain were assembled into EcoRI/XbaI-cut pUASTattB vector using the NEBuilder HiFi assembly cloning kit (NEB). The construct was verified by Sanger sequencing (GATC Services-Eurofins Genomics) and injected in y 1 w 67c23 ; P{y +t7.7 = CaryP} attP1 embryos (BestGene Inc.).
Genetic screening. Transgenic flies ectopically expressing the Alk receptor in a subset of cells of the developing ommatidia (sevEP-Gal4, UAS-Alk/CyO) exhibit a temperature-sensitive phenotype. Stocks reared at 29 °C exhibited pupal lethality while a characteristic eye phenotype (Fig. 1C) is observed at 25 °C. Screening was performed under controlled conditions (constant 25 °C and 60% humidity levels, 12:12 h day and night cycle). sevEP-Gal4, UAS-Alk/CyO flies were crossed to stocks of the DrosDel deficiency collection kit 32 which offers the advantage of an isogenic background and high-percentage coverage of the Drosophila genome. More than 20 F1 female flies per cross that did not exhibit balancer chromosome-associated markers were scored for modifications of the eye phenotype. Suppression was defined as reduction or absence of the area of "undifferentiated cells" in the anterior part of the eye. Expansion of this area, as well as dark spots and excessive necrotic tissue www.nature.com/scientificreports/ were scored as "other modifications" (detailed in Table S1). All scored flies had to exhibit a similar effect on the sevEP-Gal4, UAS-Alk phenotype although slight variations in the phenotypic strength of the observed effect were tolerated. Crosses that did not yield viable offspring of the desired genotype were performed in triplicate before assigning them to the "lethal/no progeny" group. Non-curly winged female offspring of sevEP-Gal4, UAS-Alk/ CyO and the DrosDel project starter stock (#5905) were used as controls. Deficiencies that modified the eye phenotype in this initial screen were tested again to ensure that the results were reproducible. In order to map down the various modifier regions, we scored additional stocks carrying either smaller chromosomal deficiencies, or available mutations in loci located within the suppressing region, or Gal4/UAS-based transgenic constructs of the associated genes. Alleles used to reveal the modifying effect of the identified candidates are listed in Table S1. More than 20 non-balanced F1 female flies per cross were scored, crosses with each allele were performed at least twice. We did not score for partial suppression/modification but slight variations in phenotypic strength were tolerated. For the human ALK modifier experiment, more than 15 non-balanced F1 female flies per cross were scored for modifications of the induced "rough-eye" phenotype. Only complete penetrance was scored, and crosses were performed twice.

Preparation of adult Drosophila eyes for thin sectioning and microscopic analysis.
To visualize ommatidia patterns, heads from adult flies were dissected and fixed in a mix of 2% glutaraldehyde, and 4% formaldehyde in 50 mM sodium cacodylate buffer (pH 7.4) and post-fixed in 4% Osmium and 2% KFeCN in 200 mM sodium cacodylate buffer (pH 7.4) followed by stepwise dehydration in 30%, 50%, 70%, 85%, 95%, 100% ethanol. Samples were embedded in Hard Plus Resin 812 (Electron Microscopy Sciences) for 48 h at 60 °C. Eye Section (1 µm) were obtained using a Reichert-Jung Ultracut E ultramicrotome with Ultra 45 0 diamond knife (Electron Microscopy Sciences) on glass slides and stained with a 1% toluidine-blue solution. Images were obtained using a ZEISS Axio Imager.Z2 microscope equipped with a ZEISS Axiocam 503 color camera and the ZEN Blue edition software.
Imaging of adult Drosophila eyes. Image stacks of adult fly eyes were acquired with a ZEISS AxioZoom.
V16 stereo zoom microscope equipped with LED ring light and an Axiocam 503 color camera. The enhanced depth of focus (EDF) module of the ZEN Blue edition software was used for further image processing.  Table S1) were analyzed by PANTHER14.0-driven gene ontology (GO) enrichment analysis (https ://geneo ntolo gy.org/) 51 ). Interaction analysis of Alk modifying loci (Fig. 3B, listed in Table S1) was analyzed with STRING version 11 (https ://strin g-db.org) 111 employing a medium confidence score of 0.4 for interactions. GraphPad Prism 7 software was used for all statistical analyses. www.nature.com/scientificreports/