A targeted genetic modifier screen in Drosophila uncovers vulnerabilities in a genetically complex model of colon cancer

Abstract Received on 16 January 2023; accepted on 21 February 2023Kinases are key regulators of cellular signal transduction pathways. Many diseases, including cancer, are associated with global alterations in protein phosphorylation networks. As a result, kinases are frequent targets of drug discovery efforts. However, target identification and assessment, a critical step in targeted drug discovery that involves identifying essential genetic mediators of disease phenotypes, can be challenging in complex, heterogeneous diseases like cancer, where multiple concurrent genomic alterations are common. Drosophila is a particularly useful genetic model system to identify novel regulators of biological processes through unbiased genetic screens. Here, we report 2 classic genetic modifier screens focusing on the Drosophila kinome to identify kinase regulators in 2 different backgrounds: KRAS TP53 PTEN APC, a multigenic cancer model that targets 4 genes recurrently mutated in human colon tumors and KRAS alone, a simpler model that targets one of the most frequently altered pathways in cancer. These screens identified hits unique to each model and one shared by both, emphasizing the importance of capturing the genetic complexity of human tumor genome landscapes in experimental models. Our follow-up analysis of 2 hits from the KRAS-only screen suggests that classical genetic modifier screens in heterozygous mutant backgrounds that result in a modest, nonlethal reduction in candidate gene activity in the context of a whole animal—a key goal of systemic drug treatment—may be a particularly useful approach to identify the most rate-limiting genetic vulnerabilities in disease models as ideal candidate drug targets.


Introduction
Cancer is a genetically complex and heterogeneous disease. For most solid tumors, tumorigenesis and progression into metastatic disease typically require the accumulation of multiple genomic alterations (Bailey et al. 2018;Sanchez-Vega et al. 2018). For instance, broad alterations in multiple signaling networks and global changes in transcriptomic, proteomic, phospho-proteomic, and epigenomic profiles have consistently emerged from recent tumor profiling studies as hallmarks of most advanced tumors (Das et al. 2020;Weber et al. 2020;Zhang et al. 2022). Individual genes and pathways recurrently altered in human tumors have been well characterized in multiple experimental systems, and multigenic cancer models have started to shed light on emergent interactions between concurrent cancer-driving genetic alterations (Kersten et al. 2017;Guerin et al. 2020;Sajjad et al. 2021).
However, the complex, multigenic nature of tumor genome landscapes makes it particularly challenging to identify genes required to establish and maintain tumor phenotypes and prioritize candidate targets for drug discovery approaches (Haley and Roudnicky 2020).
With its sophisticated genetic tools and practical advantages, Drosophila has a strong track record as a disease model and has also emerged as a useful drug discovery platform (Cheng et al. 2018;Mohr and Perrimon 2019;Munnik et al. 2022). Disease models generated either by genetically manipulating Drosophila orthologs of human disease genes or directly introducing disease-associated variants of human genes into Drosophila have provided important insights into molecular mechanisms underlying human diseases, including cancer (Verheyen 2022). Furthermore, multiple studies over the past decade have demonstrated a high degree of conservation of compound activity in Drosophila (Su 2019;Munnik et al. 2022); several compounds identified through these studies have been effective in vertebrate experimental models and human patients (Dar et al. 2012;Bangi et al. 2016Bangi et al. , 2019Bangi et al. , 2021Das et al. 2018;Sonoshita et al. 2018).
We have previously leveraged the genetic simplicity and power of Drosophila as a model system to functionally explore genome landscapes of human colon tumors (Bangi et al. 2016(Bangi et al. , 2019. These multigenic, tumor genome-based models capture many hallmarks of human tumors and demonstrate that intrinsic drug resistance is an emergent feature of genetic complexity. In this study, we take advantage of a classic genetic modifier screen approach to identify bona fide regulators of a 4-hit model targeting Drosophila orthologs of KRAS, TP53, PTEN, and APC, 4 genes recurrently mutated in colon tumors, as well as KRAS alone, which represents one of the most commonly observed pathway alterations in human tumors (Cancer Genome Atlas Network 2012). As kinases are critical regulators of signaling networks and frequent targets of cancer drug discovery efforts (Attwood et al. 2021), we focused our genetic screen on the Drosophila kinome, which consists of 377 kinases representing all branches of the human kinome tree (Manning et al. 2002;Gramates et al. 2022). Our parallel screens of these 2 models identified partially overlapping hits, indicating that as the genetic complexity of the model increases, its dependence on the activity of some kinases decreases and new genetic vulnerabilities that were absent in simpler models emerge. By comparing the sensitivity of our models to a genetic copy loss vs. strong, tissuespecific knockdown of kinase activity, we found that classic modifier screens in heterozygous mutant backgrounds, which result in a modest, nonlethal reduction of kinase activity in the whole animal may be a valuable approach for identifying most rate-limiting genetic vulnerabilities that could be exploited in therapy, while tissuespecific gene knockdown or knockout approaches can be used to uncover genetic dependencies.
Fly lines that combine all transgenic elements required for targeted expression of KRAS TP53 PTEN APC or KRAS to generate the lethal screening phenotypes into a single genetic background (i.e. screening stocks) were generated using standard Drosophila genetic crosses: w 1118 UAS-dcr2/Y, hs-hid; UAS-ras G12V UAS-p53 RNAi UAS-pten RNAi UAS-apc RNAi ;byn-gal4 UAS-GFP tub-gal80 ts /S-T, Cy, tub-gal80, Hu, Tb and w 1118 UAS-dcr2/Y, hs-hid;UAS-ras G12V ;byn-gal4 UAS-GFP tub-gal80 ts /S-T, Cy, tub-gal80, Hu, Tb. UAS-GFP and UAS-dcr2 transgenes are included to fluorescently label targeted hindgut epithelial cells and to facilitate RNAi-mediated knockdown, respectively. A tub-gal80 ts transgene on the third chromosome (byn-gal4 UAS-GFP tub-gal80 ts ), which drives the ubiquitous expression of a temperature-sensitive (ts) allele of the Gal4 inhibitor Gal80 (McGuire et al. 2004), is used to temporally regulate transgene induction. A tub-gal80 transgene (Lee and Luo 1999) introduced into the balancer chromosome (S-T, Cy, tub-gal80, Hu, Tb) is used to inhibit gal4 activity in screening stocks, preventing transgene expression and lethality. The Y chromosome hs-hid transgene, which results in ubiquitous activation of apoptosis when induced (Starz-Gaiano et al. 2001), is used to kill all male progeny and facilitate mass virgin female collection required for a large number of crosses of the genetic modifier screens. Kinase alleles were balanced using balancer chromosomes carrying a Tb marker, using CyO, Cy, Tb, RFP (CTR) and FM7c, Tb, RFP (FTR) balancers for the second and X chromosomes, respectively (Pina and Pignoni 2012), and the TM6b, Hu, Tb balancer for the third chromosome.

Genetic screen
Mutant kinase alleles were introduced into the KRAS TP53 PTEN APC and KRAS backgrounds with a simple F1 cross (see Supplementary  Fig. 1 for examples). For the 2nd, 3rd, and 4th chromosome screens, males from kinase stocks were crossed to virgin females from the KRAS TP53 PTEN APC and KRAS-only screening stocks. For kinases on the X chromosome, kinase − /FM7c, Tb, RFP virgin females were crossed to males from the KRAS TP53 PTEN APC and KRAS-only screening stocks. KRAS TP53 PTEN APC and KRAS virgins crossed to w 1118 /Y males (no kinase mutations) were used as baseline controls for our screening read-outs. Virgins required for these crosses were generated en masse by 2 independent 1-hour heat shocks of screening lines using a 37°C water bath during development using the following schedule: 3-day egg lay at room temperature in bottles with standard Drosophila medium followed by removal of parents, first heat-shock of the progeny on day 4, and second heat shock on either day 5 or day 7.
Crosses for the screen were set up on Bloomington semidefined medium (Bangi et al. 2019) at 29°C using 18-20 virgin females and 8-12 males for each cross. Parents were removed after 2 days of egg laying. Progeny was scored 12-14 days after crosses were set up. Experimental pupae were identified by the absence of the Tb marker. Survival to the pupal stage was calculated by counting the experimental and control pupae in each vial ( Supplementary Fig. 1). The fraction of empty experimental pupal cases (indicating an adult fly has emerged) was used to calculate survival to the adult stage. Each kinase allele was tested in triplicate. Kinase alleles that resulted in a statistically significant increase in survival compared to KRAS TP53 PTEN APC or KRAS alone (Fig. 1, c and d) were considered candidate hits (multiple t-tests with Holm-Sidak correction for multiple hypotheses, PRISM software). Candidate hits were then retested in an independent set of experiments, and kinases that showed statistically significant rescue in the retest experiments were considered confirmed hits. Screening crosses that were inconclusive due to low n were retested similarly. Each hit was also crossed to w 1118 UAS-dcr2/Y, hs-hid; + ; byn-gal4 UAS-GFP tub-gal80 ts /S-T, Cy, tub-gal80, Hu, Tb to confirm that mutant alleles did not affect organismal lethality or the size of the hindgut imaginal ring area in an otherwise wildtype background. The hits obtained from the screen were also tested using different alleles to control for background mutations that may contribute to the phenotypic rescue.

Dissections, imaging, and scoring
Crosses to analyze the overall size of the hindgut imaginal ring area were set up similarly on Bloomington semi-defined medium using 18-20 virgin females and 8-10 males for each cross, but at 18°C to prevent early larval lethality. Parents were removed after 3 days of egg laying, and progeny were kept at 18°C for an additional 3 days before transgenes were induced by a temperature shift to 29°C. Experimental larvae were identified based on the absence of the Tb phenotype, and hindguts from larvae at the late third instar stage were dissected 3 days after induction. Dissections were performed in phosphate-buffered saline (PBS), and hindguts were fixed in ice-cold 4% paraformaldehyde in PBS for 15 min at room temperature, followed by 3 rinses and a 15-min wash in PBS. Hindguts were mounted the next day on Vectashield with DAPI and imaged at 10× magnification at 1.5 Zoom using Leica SPE DM6 confocal microscope using 488 and 405 nm lasers for visualizing the GFP labeled hindgut epithelium and nuclei (DAPI), respectively. About 10-12 hindguts of each genotype were imaged, and the area of the imaginal ring region of each hindgut was quantified using Fiji Image J Software in pixels (Schindelin et al. 2012). KRAS TP53 PTEN APC and KRAS screening stocks crossed to w 1118 /Y was used as positive controls and GFP-only expressing hindguts generated by crossing w 1118 UAS-dcr2/Y, hs-hid; + ; byn-gal4 UAS-GFP tub-gal80 ts /S-T, Cy, tub-gal80, Hu, and Tb virgins to w 1118 / Y males served as negative controls. Experimental and control groups were compared using a one-way ANOVA (PRISM software).

Western blot analysis
Crosses to generate experimental larvae for dissections were set up as described in the previous section. Hindguts for protein extraction were dissected from late third instar larvae lacking the Tb marker after 3 days of induction. Ten hindguts for KRAS TP53 PTEN APC and 15 hindguts for KRAS alone, along with matching GFP-only control larvae, were used for making protein lysates (3 biological replicates/genotype). Larval hindguts were homogenized with a motorized pestle (at 10-second pulses) in icecold lysis RIPA buffer (37.5 µl, Sigma-Aldrich #R0278) fortified with protease and phosphatase inhibitors (Sigma-Aldrich #4693132001 and Millipore Sigma #524627). Lysates were then centrifuged at 4°C for 10 min at 13,000 rpm. Thirty-five microliters of supernatants were transferred to a fresh tube, followed by the addition of 12.5 µl of Sample buffer (Bio-Rad #1610792) and 2.5 µl of 20× Reducing Agent (Bio-Rad #1610792). The lysates were boiled for 10 min in a 100°C heat block, centrifuged at 4°C for 5 min at 13,000 rpm, and 50 μl of supernatant were transferred to new tubes. Lysates were stored at −80°C until use. For western blot analysis, proteins were separated using 4-12% Criterion XT Bis-Tris Protein Gels (Bio-Rad #3450125). Primary antibodies used were Rabbit anti-phospho-AKT (1:1,000, Cell Signaling  (KRAS alone, b) to the developing hindgut using the hindgut specific byn-gal4 along with a UAS-GFP transgene. c and d) Screening strategy using rescue of lethality as a read-out. A mutant allele for each kinase is introduced into KRAS TP53 PTEN APC (c) and KRAS (d) backgrounds to evaluate their effect on lethality. Kinases with important, rate-limiting roles in these backgrounds may improve survival.

Screen strategy
Rescue from lethality has been shown to serve as a useful readout in high throughput chemical and genetic screens to identify novel regulators of disease-relevant phenotypes (Rudrapatna et al. 2014;Levinson and Cagan 2016;Bangi et al. 2019). To generate a robust, lethal phenotype suitable for large-scale screening, we used the byn-Gal4 driver expressed in the hindgut epithelium (Takashima et al. 2008) to target our 4-hit model KRAS TP53 PTEN APC to the developing hindgut. This model comprises an oncogenic UAS-ras G12V transgene and 3 UAS-RNAi transgenes to knockdown Drosophila orthologs of tumor suppressors APC, TP53, and PTEN (Bangi et al. 2016). We found that targeting KRAS TP53 PTEN APC multigenic combination to the developing hindgut epithelium resulted in a highly penetrant and consistent early larval lethal phenotype (Fig. 1a), while targeting KRAS alone (UAS-ras G12V ) to the same tissue resulted in a similarly robust lethal phenotype, albeit at a later stage during late pupal development (Fig. 1b). We then used these lethal phenotypes as surrogate read-outs in parallel genetic screens to evaluate vulnerabilities of genetically complex and simple models and identify genes with essential roles in intestinal transformation.
Kinases are essential in governing cancer-relevant pathways (Gross et al. 2015;Cicenas et al. 2018); therefore, we focused our screens on the Drosophila kinome. Like in humans, most Drosophila loci are recessive; they show no phenotype or lethality when a single genomic copy is removed. However, losing a copy of a gene that promotes oncogenic transformation and organismal lethality in the context of a multigenic model may prove ratelimiting and improve survival. Therefore, we decided to adopt a classic genetic modifier screening approach by reducing the gene dosage of each kinase by introducing a mutant allele into our models using publicly available fly lines. Building on prior studies taking a similar approach (Levinson and Cagan 2016;Sonoshita et al. 2018), we reasoned that reducing the level of a kinase that may be critical for tumor progression would rescue the lethality we observe in our models (Fig. 1, c and d).

Screen design
The ability to perform complex genetic manipulations and carry out large-scale in vivo screens are 2 key strengths of Drosophila as a model system. However, as our multigenic models are already genetically complex-for instance, experimental animals in our KRAS TP53 PTEN APC model carry 8 different transgenes-performing additional genetic manipulations in these backgrounds, especially in the context of a large-scale genetic screen, can be challenging. To address this problem, we established screening lines for KRAS TP53 PTEN APC and KRAS alone by consolidating all transgenic elements required to generate each lethal screening read-out into a single genetic background (Supplementary Fig. 1). These screening lines also include: (1) a ubiquitously expressed gal80 transgene (Lee and Luo 1999) to prevent gal4 activity and lethality within screening stocks and (2) a hs-hid transgene on the Y chromosome (Starz-Gaiano et al. 2001) that leads to ubiquitous activation of apoptosis and lethality in male progeny when induced, a strategy that allows en masse virgin generation necessary for large-scale screens. This design allowed us to introduce mutant kinase alleles into KRAS TP53 PTEN APC and KRAS alone backgrounds with simple F1 genetic crosses (Supplementary Fig. 1).

Screen results
The Drosophila kinome comprises 377 kinases (Gramates et al. 2022). We were able to obtain mutant lines for 206 kinases, covering 54.6% of the entire fly kinome (Supplementary Table 1). Our parallel screens of the same set of kinases identified 3 confirmed hits from the KRAS TP53 PTEN APC screen (aPKC, par-1, cdk2), while the KRAS screen resulted in 5 hits (par-1, cdk2, hppy, dsor1, akt) (Fig. 2, a and b, Supplementary Table 2). Confirmed hits did not affect organismal lethality in GFP-only control animals ( Supplementary Fig. 2a). Retests of confirmed hits using a different mutant allele for each kinase provided results consistent with the original screen, except par-1, where the new allele resulted in the statistically significant rescue of KRAS TP53 PTEN APC-induced lethality but not that of KRAS alone ( Supplementary Fig. S2, b and c). Given this discrepancy, we did not pursue par-1 further as a hit for the KRAS-only screen in this study (Fig. 2c). However, these results do not rule out a role for par-1 in KRAS-mediated intestinal transformation; additional experiments using other alleles and transgenic constructs will be necessary to investigate its function further.
The only shared hit between the 2 screens was cdk2, a highly conserved gene with cancer-relevant functions (Fig. 2c). cdk2 regulates the G1/S transition of the cell cycle in association with CycE and plays an essential role in modulating cell division and differentiation (Sauer et al. 1995;Roesley et al. 2018). The human cdk2 ortholog, CDK2, is altered in various cancers, including colorectal, and its downregulation can have antitumor effects depending on the genetic context (Tadesse et al. 2020).
aPKC, 1 of the 2 hits unique to KRAS TP53 PTEN APC (Fig. 2, a and c, Supplementary Fig. 2b), is an atypical protein kinase with many downstream targets and a key regulator of cell polarity (Vorhagen and Niessen 2014;Hong 2018). aPKC overexpression is also linked to the progression and development of different cancers (Eder et al. 2005;Regala et al. 2005;Nayak et al. 2019). par1, the other hit unique to KRAS TP53 PTEN APC, is the single Drosophila ortholog of the PAR-1/MARK family of serine/threonine kinases in mammals (Wu and Griffin 2017) and a critical regulator of cytoskeleton structuring, cell polarity, and Hippo signaling (Huang et al. 2013;Doerflinger et al. 2022). PAR-1 has been implicated in cancer progression and is upregulated in various tumors like pancreatic, breast, and lung cancer (Boire et al. 2005;Cisowski et al. 2011;Tekin et al. 2018). Hits like aPKC and par-1, which are effective against genetically complex cancer models, are promising candidate drug targets as they may represent broad genetic sensitivities shared among cancer genome landscapes.
The 3 genes identified as unique hits from the KRAS-only screen were hppy, akt, and dsor (Fig. 2, b and c, Supplementary  Fig. 2c), which are critical components of Hippo, PI3K, and MAPK signaling cascades, respectively (Zheng et al. 2015;Yang et al. 2019;Guo et al. 2020). Hppy, a serine-threonine kinase, activates Warts alongside Hippo and regulates JNK signaling, inflammation, cellular migration, and invasion (  cell proliferation and downregulation of antiproliferative genes (Chambard et al. 2007;Guo et al. 2020). The third hit, akt, lies downstream of PI3K/AKT signaling cascade, another pathway that regulates cell growth, survival, proliferation, and metabolism in Drosophila and mammals (Saxton and Sabatini 2017) and is dysregulated in various cancers (Yang et al. 2019). Our observation is that KRAS TP53 PTEN APC is not sensitive to a reduction in the gene dosage of these kinases is interesting, particularly in the case of dsor and akt, given the strong activation of both MAPK and AKT signaling pathways in RAS-PI3K co-activated tumors (Wee et al. 2009). Overall, our findings suggest that as the genetic complexity of the model increases, it may become less dependent on kinases that have essential roles in simpler genetic backgrounds.

Characterizing hits using cancer-relevant secondary assays
While rescue from lethality is a useful surrogate screening readout, it is critical to establish more disease-relevant secondary assays to test whether hits from such screens can also modify disease phenotypes and explore their mechanisms of action (Rudrapatna et al. 2014;Levinson and Cagan 2016;Bangi et al. 2019). We have previously demonstrated that multigenic cancer models in Drosophila capture multiple key hallmarks of human tumors and using multiple phenotypic read-outs is particularly important for exploring mechanisms of action of hits identified from genetic and drug screens (Bangi et al. 2012(Bangi et al. , 2016(Bangi et al. , 2019. To illustrate this principle, we use one such assay to test our hits from the genetic screens as a proof of concept. Targeting KRAS TP53 PTEN APC and KRAS to the larval hindgut epithelium results in the expansion of the hindgut imaginal ring zone (Fig. 2d), the anterior region of the hindgut where proliferating progenitor cells reside (Murakami and Shiotsuki 2001). We next tested whether hits from our screens could reduce KRAS TP53 PTEN APC or KRAS-induced expansion of this region as a secondary read-out. We found that a reduction in cdk2 resulted in a significant decrease in the size of the imaginal ring area in the KRAS TP53 PTEN APC background but not in KRAS alone (Fig. 2, e and f). We also found that reduction in akt or dsor levels resulted in a significant reduction in the imaginal ring zone size in KRAS alone (Fig. 2f), further confirming the essential roles these 2 kinases play as downstream mediators of the RAS pathway. None of the hits affected the size of the anterior hindgut of GFP-only control animals ( Supplementary Fig. 2d).
Combined, these results show that not all hits that rescued KRAS TP53 PTEN APC or KRAS-induced lethality were able to reduce the size of the transformed hindgut area, demonstrating that the expansion of the imaginal ring area cannot fully account for the lethal phenotype we observe in our models. As the size of this area is likely to be determined by a combination of proliferation, apoptosis, senescence, and cell growth, subsequent assays that quantitatively analyze these markers will be necessary to investigate the mechanisms of action of these hits. Furthermore, given the complex nature of intestinal transformation, hits that had no effect in this particular assay are likely to contribute to intestinal transformation by affecting other hallmarks of cancer. Future studies using additional phenotypic read-outs of intestinal transformation, such as epithelial-mesenchymal transition and cell polarity, will be necessary to explore the mechanisms of action of these hits further. Overall, these results demonstrate that a single secondary assay is unlikely to be sufficient to confirm hits from lethality-based screens and emphasize the importance of using multiple assays that capture different hallmarks of tumors to fully explore the mechanisms of action of hits generated from genetic screens.

Tissue-specific knockdowns uncover genetic dependencies on dsor and akt
Ras/MAPK and PI3K/AKT signaling cascades are 2 of the most common pathways dysregulated in human cancers and multiple experimental models, including our previous work (Ericson et al. 2010;Bangi et al. 2016Bangi et al. , 2019Reggiani Bonetti et al. 2018;Jiang et al. 2020;Chen et al. 2021), making them relevant drug targets in cancer therapy. Still, we found that reducing dsor or akt gene dosage using the loss of function alleles is insufficient to rescue KRAS TP53 PTEN APC-induced lethality in our screen, suggesting that this model was not very sensitive to changes in akt or dsor levels. To investigate the genetic dependency of our multigenic model on these 2 genes, we tested the effect of strongly knocking down each gene in the hindgut on KRAS TP53 PTEN APC-induced lethality. RNAi-mediated knockdown of dsor or akt in the hindgut epithelium using 2 independent RNAi lines for each significantly improved survival to the pupal stage in KRAS TP53 PTEN APC background (Fig. 3, a and c, Supplementary Fig. 3b) and reduced the expansion of the imaginal ring area (Fig. 3, e and g, Supplementary Fig. 3c). Furthermore, an RNAi line targeting luciferase had no effect on KRAS TP53 PTEN APC-induced lethality or the expansion of the anterior hindgut area ( Supplementary  Fig. 3, b and c), indicating that the rescue observed upon dsor and akt knockdown is not an indirect effect of introducing another UAS-transgene into this already complex genetic background.
Consistent with our KRAS-only screen results, RNAi knockdown of akt and dsor also rescued KRAS-induced lethality and expansion of the imaginal ring area (Fig. 3, b, d, f, and h). These results demonstrate that even though both akt and dsor are required for intestinal transformation in KRAS TP53 PTEN APC background, this model is more resistant to moderate changes in dsor or akt levels; a more substantial reduction in their activity is necessary to lead to the rescue. This contrasts with the KRAS-only background, where reducing the akt or dsor gene dosage is sufficient to mount a rescue.
Identifying genetic dependencies of tumors is a crucial step in target identification and prioritization in drug discovery approaches. Our findings demonstrate that genetically complex models may not be sensitive to moderate reductions in the activity of all genes, on which they are dependent. This has important implications for target prioritization for drug discovery, as it may not always be feasible to achieve a strong enough reduction in target activity comparable to what we observe in our tissue-specific knockdowns by systemic drug treatment without significant side effects. As genetic modifier screens performed in heterozygous genetic backgrounds select for hits that suppress disease phenotypes without causing any organismal lethality by design, they may represent a valuable approach to identifying the most ratelimiting genetic vulnerabilities that may be more suitable candidate drug targets.

Molecular correlates of genetic dependency and sensitivity to akt and dsor
To further investigate how reducing the gene dosage of akt or dsor and their RNAi knockdowns alter Ras/MAPK and AKT pathway activity in our models, we directly analyzed the levels of phoshpo-Akt (pAKT) and diphospho-Erk (dpERK), the main readouts of the Ras/MAPK and AKT pathways, respectively, in KRAS TP53 PTEN APC and KRAS-only hindguts (Fig. 4). Heterozygous loss of akt or dsor resulted in a modest but statistically significant downregulation of dpERK in KRAS TP53 PTEN APC hindguts but had no significant effect on pAKT levels ( Fig. 4, a, c, and e), suggesting that increased AKT activity may be more critical than ERK activity in the context of KRAS TP53 PTEN APC. dsor knockdown in KRAS TP53 PTEN APC hindguts was not sufficient to alter pAKT levels but resulted in a much stronger reduction in pERK levels, indicating that KRAS TP53 PTEN APC is still dependent on ERK activity and can be rescued if a strong enough reduction in its activity is achieved (Fig. 3, c and g). akt knockdown, on the other hand, resulted in a significant decrease in both pAKT and pERK levels, albeit a much stronger reduction in pAKT compared to dpERK (Fig. 4, c and e). These results demonstrate that in KRAS TP53 PTEN APC, a strong decrease in either AKT or MAPK pathway activity is sufficient to result in rescue; however, this level of reduction in activity could not be achieved in heterozygous mutant backgrounds for either akt or dsor.
In comparison, in KRAS-only hindguts, we observed a more significant decrease in MAPK and AKT pathway activities in dsor and akt heterozygous mutant backgrounds, respectively (Fig. 4, b, d, and f). We also noted a significant increase in pAKT levels in dsor heterozygous mutant background, most likely due to the disruption of a previously established negative feedback regulation of AKT activity by MEK (Turke et al. 2012). Our finding that heterozygous dsor loss is sufficient to rescue both KRAS-induced lethality and imaginal ring area expansion despite an increase in pAKT levels suggests that in a KRAS-only background, transformed tissue is more dependent on MAPK activity than AKT signaling. In contrast, KRAS TP53 PTEN APC transformed tissue is more dependent on the AKT pathway. Combined, these results show that strong downregulation of either MAPK or AKT signaling pathways results in the rescue of lethality and a significant decrease in the size of the transformed tissue (Fig. 3), demonstrating that both pathways are essential for KRAS TP53 PTEN APC and KRAS alone. However, it is easier to achieve a level of reduction in MAPK or AKT pathway activity sufficient to rescue lethality or overgrowth in KRAS alone compared to KRAS TP53 PTEN APC.

Concluding remarks
AKT and MEK have been attractive targets for cancer drug discovery given their central roles in Ras/MAPK and PI3K/AKT pathways and functional studies demonstrating a requirement for these 2 genes in tumor progression and development (Yap et al. 2011;Caunt et al. 2015;Saura et al. 2017;Bose and Kalinsky 2021). Despite extensive efforts, single-agent MEK and AKT inhibitors have not been very successful in clinical trials for most solid tumor types in part due to severe toxicities Serra et al. 2011;Yap et al. 2011;Faião-Flores et al. 2019;Zhan et al. 2019). In addition, moderate inhibition of the MAPK pathway by MEK inhibitors can also lead to the development of resistance mainly via upregulation of the PI3K/AKT pathway (Balmanno et al. 2009;Turke et al. 2012;Irvine et al. 2018;Tsubaki et al. 2019). Importantly, AKT and MEK inhibitor combinations designed to inhibit both pathways simultaneously and prevent the activation of feedback loops that promote resistance have also been unsuccessful due to adverse effects that were more severe than single-agent trials and observed at lower doses (Tolcher et al. 2015(Tolcher et al. , 2020. Overall, our findings are consistent with clinical studies suggesting that, despite the importance of these genes in cancer, pharmacological inhibition of their activity to a level that could be detrimental to genetically complex tumors may not be possible in most cases due to significant side effects and toxicities ( (a and b) or dsor (c and d) in the hindgut epithelium. Error bars represent SEM. **P ≤ 0.01, ***P ≤ 0.001 (t-tests, PRISM software). e-h) Quantification of hindgut imaginal ring area in hindguts with indicated genotypes. Error bars represent SEM. **P ≤ 0.01, ****P ≤ 0.0001 (one-way ANOVA, PRISM Software).
As most drugs are administered systemically, identifying doses that provide a clinical benefit without any adverse effects and toxicities can be challenging. (Muller and Milton 2012). In fact, a significant fraction of drugs that enter clinical trials fail due to unmanageable toxicities or lack of clinical efficacy at maximum tolerable doses, even though they target critical cancer genetic dependencies (Sun et al. 2022). While tissue-specific gene knockdown and knockout approaches are helpful in identifying genes essential for mediating disease phenotypes, the dependence of a disease on a gene does not necessarily reflect its potential as a drug target. Our results suggest that classic genetic modifier screens performed in heterozygous mutant backgrounds that result in a modest, nonlethal reduction in gene activity in the context of a whole animal may be a particularly useful complementary strategy to identify promising genetic vulnerabilities that can be successfully exploited by therapy. Fig. 4. Analysis of MAPK and AKT pathway activity in response to tissue specific knockdown vs. heterozygous loss. a and b) Western blot analysis of pAKT and pERK levels in protein extracts from hindguts with indicated genotypes (3 biological replicates/genotype). Syntaxin: loading control. c-f) Quantification of western blot data presented in (a) and (b). For quantification of pERK levels, all genotypes are normalized to GFP-only negative controls (c and d). As there were no detectable pAKT bands in GFP-only negative controls (top blots in a and b), all genotypes were normalized to KRAS TP53 PTEN APC (e) or KRAS (f). Error bars represent SEM. *P ≤ 0.05, **P ≤ 0.01, ****P ≤ 0.0001 (one-way ANOVA, PRISM Software).

Data availability statement
Fly stocks generated through this work are available upon request. The BDSC IDs and full genotypes of the kinome stocks used in this study are available in Supplementary Table 1 and online at https://bdsc.indiana.edu/index.html. More information on the kinase alleles used in this study can be found online at https:// flybase.org/.
Supplemental material is available at G3 online.