Functional requirements of protein kinases and phosphatases in the development of the Drosophila melanogaster wing

Abstract Protein kinases and phosphatases constitute a large family of conserved enzymes that control a variety of biological processes by regulating the phosphorylation state of target proteins. They play fundamental regulatory roles during cell cycle progression and signaling, among other key aspects of multicellular development. The complement of protein kinases and phosphatases includes approximately 326 members in Drosophila, and they have been the subject of several functional screens searching for novel components of signaling pathways and regulators of cell division and survival. These approaches have been carried out mostly in cell cultures using RNA interference to evaluate the contribution of each protein in different functional assays and have contributed significantly to assign specific roles to the corresponding genes. In this work, we describe the results of an evaluation of the Drosophila complement of kinases and phosphatases using the wing as a system to identify their functional requirements in vivo. We also describe the results of several modifying screens aiming to identify among the set of protein kinases and phosphatases additional components or regulators of the activities of the epidermal growth factor and insulin receptors signaling pathways.


Introduction
Reversible protein phosphorylation was first described in the 1950s (Krebs and Fischer 1955) and since then many studies have emphasized that phosphorylation is one of the main regulatory mechanisms modifying protein activity and consequently a variety of cellular behaviors including cell cycle progression, cell death, metabolism, tissue homeostasis, cell motility, and cell differentiation (Cohen 2001). The phosphorylation state of a protein is a determinant of its biochemical activity and defines protein stability and subcellular location. Protein phosphorylation also allows transitions between active and inactive conformations and influences the repertoire of interactions with other proteins. Not surprisingly, several diseases such as obesity, cancer, and inflammation are related with aberrant phosphorylation, emphasizing its essential role in the regulation of cellular biology (reviewed in Shchemelinin et al. 2006;Tonks 2006;Hendriks et al. 2013).
The phosphorylation/dephosphorylation of proteins is mediated by protein kinases and protein phosphatases, enzymes that catalyze the transfer of phosphate groups to or from its targets, respectively (Hunter 1995;Shchemelinin et al. 2006;Hendriks et al. 2013). Kinases represent one of the largest protein families encoded in eukaryotic genomes, accounting for around 500 genes in humans and 328 genes in Drosophila melanogaster (Morrison et al. 2000). Phosphatases constitute a smaller group, including about 200 and 192 genes in humans and fly, respectively (Morrison et al. 2000). There are no Drosophila-specific families of kinases or phosphatases, and each subfamily presents small complexity and low redundancy (Manning et al. 2002). These characteristics, and the facility of genetic manipulation in this organism, make Drosophila a suitable model for the functional study of these gene families in developing tissues and cell cultures (Mattila et al. 2008;Read et al. 2013;Swarup et al. 2015). One organ that is particularly well suited for such functional approaches is the wing, a flat structure of epidermal origin that has been systematically used as a model system to dissect the molecular components and cell biology underlying epithelial development (Molnar et al. 2011;Hariharan 2015).
The Drosophila wing is a cuticular structure resulting from the differentiation of an epidermal tissue named wing imaginal disc. All features decorating the wing such as sensory organs, pigmentation, and veins are the results of the differentiation, during pupal development, of epidermal cells that were genetically specified during the growth of the wing imaginal disc (Ostalé et al. 2018). In this manner, wing patterning, as well as its size and shape, is determined during the development of the wing disc. There are multiple cellular processes impinging on wing development that are regulated by the opposing actions of kinases and phosphatases on their targets. These processes include cell growth and division, the acquisition and maintenance of apical-basal and planar polarities and vein differentiation among others (Bettencourt-Dias et al. 2004;Chen et al. 2007;Read et al. 2013;Parsons et al. 2017). In addition, protein phosphorylation pervades as a regulatory mechanism in multiple signal transduction pathways regulating pattern formation and cell differentiation.
One significant advantage of the wing for genetic analysis is the variety and specificity of phenotypic responses to genetic perturbations. For example, altering the activity of signaling pathways results in precise and pathway-specific phenotypes affecting the size and shape of the wing, the formation and polarity of the trichomes differentiated by each epithelial cell, and the position and differentiation of veins (Molnar et al. 2011;Ostalé et al. 2018). These phenotypes allow the grouping of novel mutations or knockdown conditions and can be used as a first approximation to assign gene functions by phenotypic comparison. An additional advantage of the wing for genetic analysis is the possibility of carrying out "modifier" screenings using sensitized backgrounds in which the activity of a given signaling pathway is altered. It is expected that sensitized genetic backgrounds help to identify additional components of these pathways or other molecular elements affecting their activities. For example, modifying screens have been instrumental in identifying components of the EGFR and Wnt pathways during imaginal development (Friedman and Perrimon 2006;McElwain et al. 2011;Swarup et al. 2015).
In this work, we describe the adult wing phenotypes resulting from the individual knockdown of most annotated Drosophila kinases and phosphatases, with particular emphasis in protein kinases and phosphatases. We find that 53% of protein kinases and 40% of protein phosphatases result in mutant wing phenotypes affecting the size, pattern, and differentiation of this organ. This percentage is higher compared to the percentage found for Carbohydrate, Lipid, and Nucleoside kinases (101 genes; 29% knockdowns with a phenotype). In addition, we have constructed and used sensitized genetic backgrounds in which the activities of the epidermal growth factor receptor (EGFR) and insulin receptor (InR) pathways are altered to screen the same collection of protein kinases and phosphatases for genetic interactions.

Drosophila stocks and genetics
We used the Gal4 lines sal EPv -Gal4 and nub-Gal4. The expression of sal EPv -Gal4 is restricted to the wing blade territory located between the vein L2 and the intervein L4/L5 (Cruz et al. 2009). The expression of nub-Gal4 is generalized in the entire wing pouch and hinge. For the modifier screens, we used the UAS lines UAS-GFP, UAS-dicer2, UAS-InR DN (PfUAS-InR.K1409Ag; BSCD8252), UAS-InR Act (PfUAS-InR.R418Pg; BSCD8250), UAS-ERK sem (Brunner et al. 1994), UAS-ERK-RNAi (VDCR 109108), UAS-EGFR ktop (BDSC59843), and UAS-EGFR-RNAi (VDCR 107130). These lines were combined or recombined with sal EPv -Gal4. Virgin females of sal EPv -Gal4 UAS-GFP/CyO, sal EPv -Gal4/CyO; UAS-InR DN /TM6b, sal EPv -Gal4 UAS-InR Act / CyO, sal EPv -Gal4/CyO; UAS-ERK sem /TM6b, sal EPv -Gal4 UAS-ERK-RNAi/ CyO, UAS-EGFR ktop ; sal EPv -Gal4/CyO, and sal EPv -Gal4/CyO; UAS-EGFR-RNAi/TM6b were crossed with males from the collection of UAS-RNAi of the complement of protein kinases and phosphatases. The UAS-RNAi lines used for kinases and phosphatases are listed in Supplementary Table S1. Most UAS-RNAi strains were obtained from the Vienna Drosophila RNAi Center (VDCR; 478 strains), and some from the Bloomington Stock Center (BDSC; 7 strains), and the National Institute of Genetics (NIG-FLY; 6 strains). The knockdown phenotypes of these genes were determined in UAS-dicer2/þ; nub-Gal4/UAS-RNAi and UAS-dicer2/ þ; sal EPv -Gal4/UAS-RNAi combinations. We aimed to describe each mutant wing using a simplified nomenclature summarizing the main components of its phenotype. Many combinations displayed late larval (LL) or pupal lethality (PL). In many cases, dead pupae observed in the puparium showed necrotic patches in the position normally occupied by the wings (nec). Flies showing a total failure in the formation of the wings were named "nW" (no-wing). Wings showing wing size changes were defined as "S" (wing size smaller than normal) and "S(L)" (wing size larger than normal). When changes in size were accompanied by changes in the pattern of veins, the phenotype was named "S-P." Changes affecting primarily the wing veins were defined as VÀ (loss of veins) and Vþ (excess of veins). All defects related to the wing margin consisting in the loss of wing margin stretches were defined as "WM." Defects in the apposition of the dorsal and ventral wing surfaces, observed in the form of blisters, were considered as failures in dorsoventral adhesion, and were named "WA." Similarly, defects in the global shape of the wing were defined as wing shape ("WS"), and they include lanceolate wings (lan) and dumpy wings (dp). In some cases, the wing cuticle appeared with an abnormal general appearance, brighter than normal, not entirely unfolded or with necrotic patches. These wings were classified as wing differentiation defects ("WD"). In other cases, wing cuticle was darker than normal, and these cases were named "WP" (wing pigmentation defects). Changes in the number of trichomes formed by each cell, which normally differentiate only one trichome, as well as alterations in trichome polarity and spacing, were defined as alterations in cell differentiation ("CD"). A very frequent phenotype observed in combinations between nub-Gal4 and UAS-RNAi strains of the KK VDCR collection result in the formation of adults with the wings totally folded ("WF"). This phenotype is a consequence of a UAS insertion affecting the gene tiptop (Green et al. 2014;Vissers et al. 2016). As discussed elsewhere (Ló pez-Varea et al. 2021), the same KK UAS-RNAi lines in combination with the driver sal EPv -Gal4 result in the formation of normal wings, and consequently, all WF wings where we could not observe any other phenotype were considered as wild type for all quantifications. Finally, we included the bins "strong" (s) and weak (w) in the phenotypic description, to give an indication of relative phenotypic strength. Unless otherwise stated, crosses were done at 25 C.
We did not measure the efficiency of mRNA knockdown in these genetic combinations. It was estimated in a collection 64 UAS-RNAi/act-Gal4 viable combinations that the reduction in mRNA levels varies from 95% to 10%, and that an estimated 15-40% of UAS-RNAi insertions are inactive (Dietzl et al. 2007;Perkins et al. 2015). For these reasons, a fraction of combinations without a mutant phenotype could be due to insufficient knockdown efficiency. In addition, we generally used only one UAS-RNAi strain per gene. However, from our data (Ló pez-Varea et al. 2021, G3 submitted), we know that lines targeting the same gene result in similar qualitative phenotypes (202 out of 281 cases analyzed; see Ló pez-Varea et al. 2021) and that in the remaining cases (82% of 79 genes), the more frequent situation is that one nub-Gal4/UAS-RNAi combination results in a mutant phenotype and the other in wild-type flies, again pointing to different knockdown efficiencies between independent strains. In agreement, when we compared our results with a previous RNAi screen of Drosophila protein kinases and phosphatases that used multiple UAS-RNAi lines to target each gene (Swarup et al. 2015), we found a coincidence for genes showing a wing phenotype in 82% of the genes we identified. The remaining 18% of genes correspond to cases described in Swarup et al. (2015) as "mutant wing" where we could not detect a mutant phenotype. These genes are indicated in red lettering in Supplementary Table S1.

Wing and disc measurements
Wing pictures were made with a Spot digital camera coupled to a Zeiss Axioplan microscope, using the 5X and 40X objectives for wings and for wing regions, respectively. Cell size was estimated from the number of trichomes in a dorsal region located between the L2 and L3 longitudinal veins. The number of cells was calculated using cell density and wing size values.

Immunohistochemistry
We used the rabbit antibodies anti-phospho-Histone3 and anticleaved Cas3 (Cell Signaling Technology). Alexa Fluor secondary antibodies (used at 1:200 dilution) were from Invitrogen. To stain the nuclei we used TO-PRO-3 (Invitrogen). Imaginal wing discs were dissected, fixed, and stained as described in de Celis (1997). Confocal images were taken in an LSM510 confocal microscope (Zeiss). All images were processed with the program ImageJ 1.45 s (NIH, USA) and Adobe Photoshop CS3.

Statistical analysis
All numerical data including wing size and cell size were collected and processed in Microsoft Excel (Microsoft Inc.). The data and ratios between number of cells were expressed as means þ standard error of the mean (SEM) and were compared using a Ttest. P-values were adjusted by false discovery rate method using R-studio platform. We consider a significant P-value lower than 0.05 (*), 0.01 (**) and 0.001 (***).

Gene expression
We used RNA-Seq reads from run SRR3478156, corresponding to control larvae expressing Gal4/GFP data obtained from dissected wing imaginal discs (Flegel et al. 2016) and GeneChip TM Drosophila Genome 2.0 Affymetrix array data (Organista et al. 2015) to determine expression or not expression in the wing disc for all genes encoding kinases and phosphatases.
The authors affirm that all data necessary for confirming the conclusions of the article are present within the article, figures, tables, and Supplementary information.

Results and discussion
Phenotypic screen of kinases in the wing Kinases catalyze the transfer of a phosphate group from ATP to a substrate molecule. To compile a list of kinases (and phosphatases, see below), we used the classification provided in the FlyBase gene group list (http://flybase.org/lists/FBgg/) and the annotation of protein kinases provided by Morrison et al. (2000). We included in our analysis carbohydrate, lipid, nucleoside, and protein kinases, resulting in a group of 328 genes (Figures 1 and 2A). As a general procedure for the screen, we used only one UAS-RNAi line per gene. We first crossed UAS-RNAi males (Supplementary Table S1) with UAS-dicer2; nub-Gal4/CyO females. In all cases, where the progeny UAS-dicer2/þ; nub-Gal4/UAS-RNAi was lethal or resulted in flies with rudimentary wings (42 out of 310 crosses performed), we crossed the corresponding UAS-RNAi lines with UAS-dicer2; sal EPv -Gal4/CyO females. The UAS-dicer2/þ sal EPv -Gal4/UAS-RNAi combinations were always viable and were used to classify phenotypically the corresponding RNAi lines.
Carbohydrate, Lipid, and Nucleoside kinases include 101 proteins mostly involved in metabolic pathways (71%; Supplementary Table S1). The corresponding genes are generally expressed in the wing disc (84%; Figures 1 and 2A) and their knockdowns result in lethality or a wing phenotype in a low percentage of cases (29%; Figures 1 and 2B). The phenotypes most frequently observed after knockdown of nonprotein kinases consisted in a reduction of the size of the wing (S, 31%; Figures 1 and 2B) and defects in wing cuticle differentiation (WD, 13%; Figure 2B).
Protein kinases comprise a single protein superfamily having a common catalytic structure (Morrison et al. 2000). These enzymes are further subdivided into distinct groups based on their structural and functional properties (Hanks and Hunter 1995). Most of the 227 protein kinases genes are expressed in the wing disc (83%; Figures 1 and 2A) and in 53% of them we identified lethality or a mutant wing phenotype in UAS-dicer2/þ; nub-Gal4/UAS-RNAi or UAS-dicer2/þ; sal EPv -Gal4/UAS-RNAi combinations (Figures 1  and 2, B and C and Table 1). The most frequent alterations observed were changes in the size of the wing (S), in many cases accompanied by changes in the position (size and pattern; S-P) or the differentiation (size and vein formation; S/V) of the veins (Table 1; Figure 2, B and C). Other changes in wing morphology consist in blisters, caused by a failure in the adhesion between the dorsal and ventral wing surfaces (wing adhesion; WA), or failures in the formation of the wing margin (WM; Figure 2, B and C). In general, protein kinases with a known function have a higher frequency of knockdown phenotypes than other kinases with less well-characterized functions (67% vs 40%, respectively). The phenotypes of gene knockdowns for kinases that have been previously characterized generally fits with the expectation. For example, knockdown of kinases regulating the phosphorylation and inactivation of Yorkie in the Hippo pathway result in wings larger than normal (Supplementary Figure S1B). Similarly, knockdown in components of the MAPK signaling pathway cause loss of veins and wing size-reduction phenotypes (Supplementary Figure Figure S1I), and for genes which activity is required for cell growth, division, adhesion, and survival (Supplementary Figure S1, D and J-I, respectively). These results suggest that the phenotypes of not previously characterized kinases in the wing disc would be informative as to their functional requirements.
We were able to identify a phenotype for 40% of protein kinases not previously characterized in the Drosophila wing. These phenotypes could now be used as an entry point to perform a more detailed functional characterization of the corresponding genes and proteins. Despite the high fraction of genes that knockdown results in wings with altered morphogenesis, there are still many cases of genes expressed in the wing disc and for which we could not detect a mutant phenotype upon expression of the corresponding RNAi (208 genes). The reason for this result could be either a genuine lack of requirement of the gene during wing development, gene redundancy in those cases where multiple kinases affect a similar set of targets, or insufficient reduction in the level of mRNA following the RNAi knockdown. Focusing on those cases in which the expression of RNAi results in wings with altered size and/or vein patterns, we did not find a particular phenotypic enrichment for a given family of protein kinases ( Figure 2C). Many of the phenotypes we found are reminiscent of those caused by alterations of specific signaling pathways in the wing. For example, knockdown of genghis khan (gek), the fly orthologous to human CDC42 binding protein kinase alpha, results in wings larger than normal ( Figure 2F), similar to increased Yorki activity. The Gek protein is a putative effector for Drosophila Cdc42, which promotes Actin polymerization during Drosophila oogenesis (Luo et al. 1997), and the Actin cytoskeleton is a key mediator of the regulation of Hippo signaling (Seo and Kim 2018). In contrast, loss of Ret reduces wing size and causes a wing blisters ( Figure 2E), which is compatible with the requirement of the gene in extracellular matrix adhesion during dendrite development (Soba et al. 2015). Loss of cdk12, encoding a transcription elongation-associated CTD kinase (Bartkowiak et al. 2010), results in ectopic vein formation and loss of wing margin structures reminiscent of loss of Notch signaling ( Figure 2G). Strong effects in wing size and pattern were observed upon knockdown of several kinases such as Cdk9 (Supplementary Figure S2), which is involved in RNA polymerase II elongation control (Peng et al. 1998), CKIalpha (Fig. 2H), which is involved in multiple signaling pathways (see, e.g., Apionishev et al. 2005) and nonC (Supplementary Figure S3), related to the nonsensemediated mRNA decay pathway (Rehwinkel et al. 2005). Other protein kinases affecting the veins may do so by altering the early secretory pathway (CG10177 in Supplementary Figure

Phenotypic screen of phosphatases in the wing
Phosphatases catalyze the hydrolysis of a phosphate group from a given substrate. We included in our analysis 79 nonprotein phosphatases, 99 protein phosphatases, and 14 unclassified phosphatases (Figures 1 and 3A). These genes are expressed in the wing disc with percentages varying from 64% for unclassified phosphatases to 73% for protein phosphatases ( Figure 3A). Nonprotein phosphatases include proteins with broad substrate specificity (acid and alkaline phosphatases), lipid phosphate Figure 1 Global parameters of kinases and phosphatases expression and knockdown phenotypes. Summary of the number of genes (TOTAL), genes analyzed (DONE), genes expressed in the wing disc (EXP), genes with a knockdown wing phenotype (PHE), gene knockdowns causing altered wing size (S), and gene knockdowns causing loss of wing or strong defects in wing size and pattern phenotype (S-P/nW). phosphatases (LPP), which are integral membrane proteins that catalyze the dephosphorylation of a variety of lipid phosphates, phosphatidylinositol lipid phosphatases, sugar phosphatases, and HAD family nonprotein phosphatases. The genes CG9115, CG3632, CG3530, and CG5026, which have Phosphoinositide 3 phosphatase activity, also have Dual-Specificity Phosphatases (DSP) activity, and they were classified in this last group. A large fraction of these genes (88%) is related to metabolism (Supplementary Table S1). The frequency of lethality or wing mutant phenotype for this group of genes is low (31%; Figures 1 and 3B), and is only above average for phosphatidylinositol lipid phosphatase enzymes (45%; Table 1). These proteins remove phosphate groups from positions 3, 4, or 5 of inositol molecules, participating in the metabolism of phosphoinositides. Although these lipids bind a variety of target proteins mediating cell membrane functions including vesicular trafficking, signaling, and cytoskeletal function (Balakrishnan et al. 2015) phosphatidylinositol lipid phosphatases were classified mostly in the metabolism class.
The frequency of nub-Gal4/UAS-RNAi combinations with a lethal or altered wing phenotype for protein phosphatase genes was 40% ( Figure 3B), reaching higher values for cytoplasmic tyrosine phosphatases (60%; Figure 1) and DSP (52%; Figure 1). For proteins with a known function the phenotype was as expected.         Figure 3F). Inositol and Lipid phosphatases, such as 5PtaseI and laza ( Figure 3G), display a similar extra-vein phenotype, suggestive of increased EGFR signaling. Both of them also have adhesion defects between dorsal and ventral surfaces of the wing (WA phenotype). This is a common feature of the knockdown of other phosphoinositide phosphate phosphatases such as CG9784, CG11477, and CG17029 (Supplementary Figure S6). Particularly strong phenotypes were observed in the case of genes encoding different subunits of the protein phosphatase type 2A complex (PP2A), which modulates the insulin , Hedgehog (Su et al. 2011), and Wingless (Luo et al. 2007) signaling pathways. For example, knockdown of Pp2A-29B, encoding the structural A subunit of PP2A phosphatase enzyme ) prevents wing development ( Figure  3H). A similar phenotype is observed in Pp1a-96A knockdown flies (Supplementary Figure S6). This protein also has multiple functions including the regulation of the Hedgehog and Wingless signaling pathways (Su et al. 2011). The knockdown of several PPP Serine/Threonine phosphatases results in lethality (nub-Gal4) and defects in wing size and pattern (sal EPv -Gal4) with a phenotype similar to Pp2A-29B knockdown ( Figure 3H). Some examples are mts, Pp1-87B, Pp1alpha-96A, Pp4-19C, PPP4R2r, a component of the protein phosphatase 4 complex that may coordinate centrosome maturation and cell migration ), Pp2A-29B and PpV, encoding the catalytic subunit of PP6 [Supplementary Figure  S6, PPP family and see Ma et al. (2017)]. A similar strong phenotype, in which all the central domain is differentiated as vein tissue, is also observed for Pp2C1 ( Figure  3I). In contrast, knockdown of the protein tyrosine phosphatases Ptp69D and Ptp4E, which might mediate negative regulation of the receptors EGFR, Breathless, and Pvr (Jeon et al. 2012), results only in defects in wing size (Supplementary Figure S7). The DUSP family offers a wide range of wing phenotypes including extra veins (CG10089), lack of veins (twe), size defects (CG13197, Mtmr6), and severe size and pattern defects (stg, mRNAcap, Mkp4 and Puc). The complete collection of phenotypes for protein and inositide phosphatases is shown in Supplementary Figure S6 and S7.
Developmental bases for "wing size" and "wing size and pattern" defects The most common phenotypes observed in UAS-RNAi/nub-Gal4 and UAS-RNAi/sal EPv -Gal4 combinations are those in which the size of the wing is altered, most frequently reduced (see, e.g., Figures 2E and 3, E, G, and I). This phenotype could be caused by a reduction in the number of wing cells (due to cell death or reduced cell division in the imaginal disc), by a reduction in the size of the cells, or by a combination of these two effects. We analyzed cell division (mitotic index) and death in the wing imaginal disc and cell size in the adult wing for four genetic combinations with different degrees of wing size reduction (Figure 4). In wildtype imaginal discs, cell division (mitosis) occurs throughout the presumptive wing blade and cell death is only testimonial and scattered in the disc (Figure 4, A and B). In the combinations analyzed the mitotic index in the wing pouch region was reduced, from 47% (nub-Gal4/UAS-fab1-RNAi; Figure 4C) to 24% (nub-Gal4/ UAS-CG14297-RNAi; Figure 4E). Cell size in the adult wing was also generally reduced, from 29% (nub-Gal4/UAS-Cdc7-RNAi; Figure 4D) to 14% (nub-Gal4/UAS-fab1-RNAi; Figure 4C). The occurrence of cell death in wing discs corresponding to smaller adult wings was generally low (Figure 4, C-F). These observations suggest that reduced wing size is mostly due to a lower rate of mitosis accompanied by different degrees of cell size reduction.
The second most frequent class of mutant phenotypes includes strong changes in the size of the wing accompanied by alterations in the pattern of veins. For many of these cases, the expression of RNAi in the entire wing (nub-Gal4) resulted in PL, and the effects in the wing could only be analyzed in combinations with the weaker driver sal EPv -Gal4 (Table 1). We analyzed cell death and mitosis in three sal EPv -Gal4/UAS-RNAi combinations leading to the formation of small wings with aberrant venation patterns and found that some but not all of them are accompanied by massive cell death in the wing disc ( Figure 5). This result indicates that the corresponding genes are required for cell viability and suggest that many genetic combinations in which the size and pattern of the wing are severely affected are a consequence of continuous and massive cell death in the imaginal disc epithelium.
Quantitative changes in the activity of the EGFR signaling pathway are translated into phenotypic series affecting wing vein formation and wing size The EGFR signaling pathway contributes to the regulation of imaginal cell division, growth, viability, and differentiation (Shilo 2003). The pathway includes a Tyrosine kinase transmembrane protein as receptor (EGFR) and several protein kinases and phosphatases that participate as core components of the receptor intracellular signal transduction cascade (Shilo 2003). In order to search for additional protein kinases and phosphatases that could impinge on the EGFR signaling cascade, we used genotypes in which the activity of the pathway is modified at the level of the receptor or at the level of the MAP kinase ERK (rolled). For both EGFR and ERK, we aimed to modify the phenotype resulting from higher than normal activation (EGFR ktop and rolled sem , respectively) or by lower than normal activation (EGFR-RNAi and rolled-RNAi, respectively) by the coexpression of RNAi's targeting all protein kinases and phosphatases. As a preliminary experiment, we generated genotypes with different degrees of EGFR and ERK variants overexpression. To do this, we changed the number of doses of the Gal4 insertions used and also the temperature at which the flies were raised. We were able to establish for each case a clear phenotypic series of effects, suggesting a linear translation between EGFR signaling output and wing phenotype ( Figure 6). For example, in the cases of EGFR pathway insufficiency caused by the expression of RNAi directed against EGFR or ERK the wing becomes progressively smaller as the level of RNAi expression increases ( Figure  6A, EGFR-i and rolled-i columns). Simultaneously, the number of veins is also progressively reduced, from small gaps in the L4 vein (low expression of RNAi, upper panels in Figure 6) to the absence of all the veins included in the domain of sal EPv -Gal4 expression (L2, L3, and L4; high expression of RNAi; lower panels in Figure 6A). Conversely, expression of activated forms of EGFR (EGFR-ktop) or ERK (Rolled Sem ) results in the differentiation of ectopic veins and wing size reduction, and these phenotypes are stronger in genotypes with maximal overexpression (Figure 6, second and fourth columns). We expect that changes on the level of EGFR or ERK activity, caused by knockdown of other genes, will modify the background phenotype of each individual combination along similar phenotypic series.

Modifier screen of kinases and phosphatases in EGFR mutant backgrounds
We crossed a collection of UAS-RNAi targeting protein and inositide kinases (211 genes; Supplementary Table S2) and phosphatases (88 genes; Supplementary Table S2) into four different genetic backgrounds with higher (UAS-EGFR-kTop/þ; sal EPv Gal4/þ and sal EPv -Gal47þ; UAS-rl sem /þ; Figure 7) or lower (sal EPv -Gal4/þ; UAS-EGFR-RNAi/þ and sal EPv -Gal4 UAS-rl-RNAi/þ Figure 7) than normal EGFR signaling pathway activity. From the resulting phenotypes, we identified those which consistently increased the background wing size and vein differentiation phenotypes (enhancers) and those which reduced these phenotypes (suppressors). In most cases, the expression of UAS-RNAi lines resulted in additive phenotypes (89% for kinases and 91% for phosphatases in average; see Supplementary Table S2). We found modifiers in cases of genes which knockdown have a phenotype by itself (26 genes; Supplementary Table S2) and also for genes which knockdown does not affect wing development (22 genes). In general, the modifiers affected one (11 genes) or more than one background phenotype (24 genes), with cases in which two (6 genes), three (9 cases), or the four (9 cases) backgrounds we used were modified by the knockdown (Supplementary Table S2). Consistently, genes acting as enhancers of EGFR gain of activity conditions usually behave as suppressors of EGFR knockdown conditions and vice versa (Figure 7, A and B). Not unexpectedly, the genes with more hits correspond to core members of the EGFR signaling pathway (Dsor, phl, and rl; Figure 7, B and H-L). Other genes identified as positive regulators because of the opposite effects of their knockdown on the EGFR-kTop and EGFR-RNAi phenotypes, are members of other signaling pathways (babo, Akt1, PI3K92E, and mts), phosphatidylinositol 3-kinases (nonC), cytoplasmic tyrosine kinases (Src42A), and a regulatory subunit of the protein phosphatase 2A (tws; Figure 7B). Similarly, genes identified as negative regulators of EGFR signaling are either components of other signaling pathways (hop, Ptn, csk, wts, Tao, alph, and sgg; Figure 7, B and M-K for the case of sgg), and also include a regulator of clathrin dynamics (aux; Hagedorn et al. 2006), Casein kinase II b subunit (an enhancer of position effect variegation, see McCracken and Locke 2014) and the phosphatases protein phosphatase 4 regulatory subunit 2-related (PPP4R2r) and Ptp61F ( Figure 7B).

The components of the InR pathway modify consistently the phenotypes of loss and gain of InR activity
InR signaling is required for wing imaginal cells growth and cell division (Edgar 2006). Consistently, expression of dominant negative or constitutively activated forms of the InR in the wing disc (sal EPv -Gal4/UAS-GFP; UAS-InR DN /þ and sal EPv -Gal4 UAS-InR Act /UAS-GFP) results in the formation of smaller and larger wings, respectively (Figure 8, A-C). These wings are formed by less and smaller cells (InR DN ) or by more and larger cells (InR*; Figure 8D). We used these two genotypes as backgrounds to search for kinases Figure 6 Phenotypic series of increased and reduced EGFR signaling in the adult wing. Wings from females grown at 17 C, 25 C, and 29 C (indicated in the left column) of genotypes containing one (salG4) or two [(salG4)x2]) copies of the sal EPv -Gal4 driver in combination with UAS-EGFR-RNAi (EGFR-i column), UAS-EGFR ktop (EGFR-ktop column), UAS-rl-RNAi (rolled-i column), and UAS-rl Sem (rolled-Sem column). Note how the severity of each mutant wing increases (top to bottom) with the level of Gal4/UAS expression. and phosphatases that in knockdown conditions can modify the wing size phenotypes resulting from altered InR signaling. As a preliminary experiment, we tested whether known components of the InR pathway can modify the characteristic InR DN or InR Act wing phenotypes (Figure 8, E-H). We found that loss of Akt, Pdk1, InR, Tor, and PI3K consistently enhance the wing size and cell size defects caused by InR DN expression ( Figure 8G). The same knockdowns also significantly correct the larger than normal wing and cell size caused by expression of activated InR ( Figure 8H). The examples of Akt-RNAi and Pdk-RNAi are shown in Figure 8, I-K and M-O, respectively. We also measured wing size for a collection of UAS-RNAi lines corresponding to genes that were identified under the dissecting microscope as "neutral" regarding InR DN or InR Act effects on wing size. In all cases, we could not find quantitative differences in the size of the corresponding combinations (Figure 8, E and F).

Modifier screen of kinases and phosphatases in InR mutant backgrounds
We combined the collection of UAS-RNAi lines directed against protein kinases and phosphatases to generate sal EPv -Gal4 UAS-In Act /UAS-RNAi and sal EPv -Gal4 UAS-InR DN /UAS-RNAi flies, and selected those with wing sizes distinct to the corresponding sal EPv -Gal4 UAS-InR Act /UAS-GFP and sal EPv -Gal4 UAS-InR DN /UAS-GFP background phenotypes. We only found one enhancer of the InR Act phenotype (Tao) and two suppressors of the InR DN phenotype (Csk and Pten). In contrast, we found 30 suppressors of the InR Act phenotype and 34 enhancers of the InR DN phenotype ( Figure 9A). Interestingly, 24 of these genes modify the InR Act and InR DN phenotypes in opposite manners, indicating that our screen has the potential to identify genes with a direct connection with Insulin signaling. In fact, we identified as "positive regulators" of InR signaling several known components of the pathway (InR, Tor, Pdk1, Akt1, and PI3K92E; Figures 9B and 10) and Cadherin 96Ca (Cad96Ca), encoding a receptor tyrosine kinase that cooperates with the InR during wing growth (O'Farrell et al. 2013). Other members of signaling pathways related to growth control identified in the screen were Src42A, ksr, EGFR, rl, and phl (EGFR signaling), the Hippo pathway member Activated Cdc42 kinase (Ack; Hu et al. 2016), and the TGFb pathway components punt, babo, and sax ( Figure 9B). We also identified as "positive regulators" of InR signaling several Cyclin-dependent kinases ( Figures 9B and 10), including Cdk2, regulating G1, and S phases of the cell cycle, Cdk7, a component of the Cdk activating kinase complex with a function in promoting tissue growth through Yorki stabilization (Cho et al. 2020), Cdk9, involved in RNA polymerase II elongation control (Eissenberg et al. 2007), and Cdk8, a component of the Mediator complex (Loncle et al. 2007) that also participates in lipid homeostasis (Zhao et al. 2012). Other genes related to lipid metabolism were Salt-inducible kinase 2 (Sik2), encoding a serine/ threonine kinase that regulates lipid storage and energy homeostasis (Hirabayashi and Cagan 2015), and the regulatory (CkIIb) and catalytic (CkIIa) subunits of the CKII ( Figure 8B). Casein kinase II is a broad specificity Ser-Thr kinase involved in a variety of processes including cell signaling, neuronal physiology, transcription factor activity, and lipid and polyamine metabolism (Stark et al. 2011;Bandyopadhyay et al. 2016;McMillan et al. 2018). Gcn2, related to the regulation of amino acid metabolism (Kang et al. 2017) and translation initiation (Olsen et al. 1998) was identified as suppressor of the InR Act large size phenotype ( Figure 8B). Other genes identified in the screen as positive regulators of InR signaling encode proteins involved in vesicular trafficking such as fab1 kinase (fab1), encoding a phosphatidylinositol-3phosphate 5-kinase promoting endo some-to lysosome trafficking (Rusten et al. 2006), gilgamesh (gish), encoding a plasma membrane-associated kinase regulating Rab11-mediated vesicle trafficking (Gault et al. 2012) and auxilin (aux), encoding a cofactor for the ATPase Hsc70 that regulates Clathrin dynamics (Kandachar et al. 2008). Finally, we also identified several genes regulating actin or tubulin dynamics, including microtubule star (mts), encoding the catalytic subunit of protein phosphatase 2A, Protein Kinase D (PKD), and the Phosphatidylinositol 4-Phosphate-5 kinase skittles (Gervais et al. 2008). Other kinases acting as positive regulators of InR signaling were CG8485 (fly ortholog of human SNF-related kinase), CG8878 (fly ortholog of VRK serine/threonine kinase 3; Figure 9, I-K), CG3277 (fly ortholog of human Colony-stimulating factor 1 receptor), Darkener of apricot (Doa), and minibrain (mnb).

Concluding remarks
We used the Drosophila wing to identify the in vivo requirements of the Drosophila complement of kinases and phosphatases. Only a low percentage of Carbohydrate, Lipid, and Nucleoside kinases and phosphatases (29%) are required for the correct development of the wing. In contrast a higher percentage of protein kinases, phosphatidylinositol lipid phosphatases, cytoplasmic tyrosine phosphatases, and DSP are required for wing development (45-60% of genes). One caveat of our screen is that we used only one UAS-RNAi line per gene, and this can lead to a wrong estimation of phenotypic frequencies. However, the high coincidence of genes showing a wing phenotype (82%) identified in our screen and in a similar screen in which several independent lines were used suggests that the numbers of false positives and negatives are low. The most frequent phenotypes we observed for these genes were lethality and changes in the size of the wing, associated or not to changes in the position of the veins. These phenotypes are caused by changes in cell division, cell size, and cell viability. We also carried out several modifying screens aiming to identify protein kinases and phosphatases acting as regulators of the EGFR and InR signaling pathways. The correct activation of these pathways is a requisite for the growth and differentiation of the imaginal epithelium, and alterations on the level of their activities led to characteristic adult wing phenotypes that were used as sensitized backgrounds for these screens. We identified modifiers affecting one (11 genes) or more than one (24 genes) EGFR genetic background phenotypes, with genes acting as enhancers of EGFR gain of activity conditions usually behaving as suppressors of EGFR knockdown conditions and vice versa. We also identified a significant group of genes acting as enhancers of InR DN and/or suppressors of InR Act expression. These genes include kinases and phosphatases regulating lipid and amino acid metabolism, cytoskeleton dynamics and vesicle trafficking, other signaling pathways regulating wing growth and several Cyclin-dependent kinases such as Cdk2, Cdk7, Cdk8, and Cdk9 with a variety of functions in cell cycle regulation, tissue growth, RNA polymerase II elongation, and transcription.

Data availability
The data underlying this article are available in the article and in its online supplementary material.
Supplementary material is available at G3 online.