The level of oncogenic Ras controls the malignant transformation of Lkb1 mutant tissue in vivo

The genetic and metabolic heterogeneity of RAS-driven cancers has confounded therapeutic strategies in the clinic. To address this, rapid and genetically tractable animal models are needed that recapitulate the heterogeneity of RAS-driven cancers in vivo. Here, we generate a Drosophila melanogaster model of Ras/Lkb1mutant carcinoma. We show that low-level expression of oncogenic Ras (RasLo) promotes the survival of Lkb1 mutant tissue, but results in autonomous cell cycle arrest and non-autonomous overgrowth of wild-type tissue. In contrast, high-level expression of oncogenic Ras (RasHi) transforms Lkb1 mutant tissue resulting in lethal malignant tumors. Using simultaneous multiview light-sheet microcopy, we have characterized invasion phenotypes of Ras/Lkb1 tumors in living larvae. Our molecular analysis reveals sustained activation of the AMPK pathway in malignant Ras/Lkb1 tumors, and demonstrate the genetic and pharmacologic dependence of these tumors on CaMK-activated Ampk. We further show that LKB1 mutant human lung adenocarcinoma patients with high levels of oncogenic KRAS exhibit worse overall survival and increased AMPK activation. Our results suggest that high levels of oncogenic KRAS is a driving event in the malignant transformation of LKB1 mutant tissue, and uncover a novel vulnerability that may be used to target this aggressive genetic subset of RAS-driven tumors. One Sentence Summary A multivariable Ras-driven Drosophila model reveals a novel LKB1 mutant lung adenocarcinoma patient subpopulation and targetable effector pathway.


Introduction
KRAS is the most commonly mutated oncogene in human cancer, and is frequently mutated in cancer types associated with high mortality such as non-small cell lung cancer (NSCLC). Efforts to directly target the KRAS protein have been challenging, although renewed efforts are currently in clinical trials 1 . Large-scale sequencing of lung adenocarcinoma has uncovered heterogeneity in mutant KRAS tumors due to concomitantly mutated tumor suppressor genes such as TP53 and LKB1, genetic subtypes that are largely mutually exclusive and which harbor distinct biologies and therapeutic susceptibilities 2 . An added layer of complexity arises due to the extensive metabolic rewiring observed in RAS-driven tumors 3 , which can arise due to Kras-mutant dosage and alterations in signaling pathways downstream of mutated tumor suppressor genes 4 . Increasingly, metabolic rewiring is known to be dependent on tissue-level dynamics within the tumor and the tumor microenvironment.
Therefore, there is a need to develop rapid and powerful models of RAS-driven cancers that mimic the complex landscape of these tumors in vivo.
Liver Kinase B1 (LKB1) is a master serine/threonine kinase that phosphorylates 13 downstream kinases of the AMP-activated protein kinase family (AMPK) family to control cell growth and cell polarity 5 . LKB1 activity is lost in a wide spectrum of human cancers and the gene that encodes LKB1 (STK11) is the third most frequently mutated tumor suppressor in human lung adenocarcinoma. Loss of LKB1 frequently occurs in KRAS-driven lung adenocarcinoma, and has been shown to promote metastasis, shorten overall survival, and confer resistance to targeted therapies and checkpoint inhibitors [6][7][8][9][10] . Altogether, these differences in survival and treatment outcomes highlight the importance of in vivo models that recapitulate the complexity and heterogeneity of these tumors when developing and implementing cancer treatments.
Drosophila melanogaster is a powerful model system for studying cancer biology due to the high conservation of human oncogene and tumor suppressor pathways 11,12 . Elegant genetic mosaic techniques in Drosophila allow tissue-specific overexpression of oncogenes and knockdown of tumor suppressors within distinct subpopulations of cells, which bestows the ability to build complex tumor landscapes in vivo. Seminal work using these methods has identified cooperating mutations that promote the metastasis of benign Kras-4 mutant tumors in vivo, and has identified such cooperating models as amenable to pharmacologic approaches 13,14,15,16 . However, despite evidence from mouse models that loss of Lkb1 is sufficient to promote tumor progression and metastasis in Kras-mutant lung tumors 17 , there has been no report of malignant synergy between Ras and Lkb1 using the rapid and genetically tractable Drosophila model.
Here, using a novel Drosophila model of Ras/Lkb1-driven malignant progression, we found that the relative levels of oncogenic Ras determine clonal growth dynamics in Lkb1 mutant tissue. Low levels of oncogenic Ras promote non-autonomous growth of surrounding wild-type tissue, while high-levels promote malignant progression and organismal lethality. To further characterize the metastatic capability of Ras/Lkb1 malignant cells we used simultaneous multiview light sheet microscopy to image live tumor-bearing larvae for up to 48hrs, and show that Ras/Lkb1 cells actively degrade basement membrane, and ultimately invade distant tissues. To further define the mechanism driving the progressive synergy between high oncogenic Ras and loss of Lkb1 we investigated signaling networks in mosaic tissue. We show that malignant Ras/Lkb1 tumors activate AMPK and are dependent on the activation of the Drosophila ortholog of CAMKK2. We validate the translational potential of our work by showing high level KRAS with concurrent mutation in LKB1 represents a unique subset of patients with worse overall survival and increased AMPK activation. Our work uncovers a novel mechanism that may include oncogenic KRAS copy number gains or amplification as a novel synergistic mechanism that drives the aggressive nature of LKB1 mutant tumors. In addition, our work proves Drosophila as a powerful model for the rational design of targeted therapies for genetic subsets of RAS-driven cancers, and suggests that the LKB1 subset of KRAS-driven cancers may benefit from targeting of the CAMKK/AMPK circuit.

Clonal loss of Lkb1 in vivo results in autonomous cell death
Recent work has highlighted effects of the dosage of oncogenic Ras on the progression of Ras-dependent cancers 18,19 . Previous work in Drosophila has identified myriad pathways that collaborate with mutant Ras to promote tumor progression and metastasis 20 , but how the dosage of Ras affects tumor progression in these multiple hit models is unknown. To address this question, we identified oncogenic Ras transgenes with differing expression levels. One expresses oncogenic Ras at levels similar to endogenous Ras (Ras Lo ). The other expresses Ras at levels several fold higher (Ras Hi ) (Fig. 1b). To mimic the genetic landscape of human KRASdriven cancers we chose to co-mutate the tumor suppressor LKB1 in Ras Lo and Ras Hi tissue. Most tumor specific LKB1 mutations are homozygous deletions or loss-of-heterozygosity with somatic mutation [21][22][23] . Among the latter, nonsense or frameshift mutations leading to protein truncation are the most common 24 . To identify the Drosophila Lkb1 loss-of-function allele with the strongest reduction in Lkb1 protein levels we first generated an antibody to Drosophila Lkb1. We then assayed for Lkb1 protein in transheterozygous larvae using three previously published Lkb1 loss-of-function alleles (X5 25 ), 4B1-11 and 4A4-2 26 ) over a large deletion that removes the Lkb1 gene. The Lkb1 X5 and Lkb1 4B1-11 loss-of-function alleles reduced Lkb1 protein expression by 60% compared to control. However, the Lkb1 4A4-2 allele reduced protein expression by 80% (Fig. 1a), which agrees with prior published genetic data suggesting Lkb1 4B1-11 as having residual protein activity 27 . The Lkb1 4A4-2 allele was chosen for further study and will be referred to as Lkb1 -/-.
We used the GFP-labeled eye expression system 14 to express Ras Lo in discreet patches or 'clones' of developing eye epithelial tissue. Expression of Ras Lo resulted in ablation of eye tissue and benign outgrowths of eye cuticle similar to what has been reported in prior reports using a UAS-Ras V12 transgene 14,28 (Fig. 1c). We then used the GFP-labeled eye expression system to inactivate the Lkb1 tumor suppressor (Lkb1 -/-) in clones of cells in the developing eye. Inactivation of Lkb1 in clones resulted in adult flies with small, rough eyes ( Fig.   1c), suggesting high levels of apoptosis. To test this, we assayed for cleaved death caspase 1 (DCP-1) in mutant clones using immunofluorescence in wandering 3 rd instar eye imaginal discs. As expected, loss of Lkb1 (marked 6 by GFP+ tissue) resulted in a large increase in autonomous cleaved DCP-1 expression as compared to discs carrying control FRT82B clones (Fig. 1d). These data suggest that homozygous loss of Lkb1 within an otherwise wild-type epithelium can result in a high level of apoptosis in vivo.

Low-level Ras and loss of Lkb1 synergize to promote non-autonomous benign overgrowth
Data from genetically engineered mouse models (GEMMs) suggests loss of Lkb1 is sufficient to promote the progression and metastasis of nascent Kras mutant lung adenocarcinoma 17 . Due to the redundancy of the vertebrate genome and paucity of rapid genetic mosaic analyses in GEMMs, we sought to use the GFP-labeled Drosophila eye expression system to build a Ras/Lkb1 model of cooperative tumorigenesis. We simultaneously expressed Ras Lo and depleted Lkb1 (Ras Lo /Lkb1 -/-) in clones of developing eye epithelial tissue, and found that autonomous DCP-1 levels returned to those observed in control eye imaginal discs (Fig. 1d). These data suggest that low levels of oncogenic Ras promote the survival of Lkb1 -/mutant tissue in vivo. In addition, eye imaginal disc complexes carrying Ras Lo /Lkb1 -/clones were larger than mosaic control discs but contained only a small amount of mutant GFP+ tissue compared to the expression of Ras Lo alone. In agreement with these results, analysis of adult Ras Lo /Lkb1 -/mosaic eyes revealed a large, overgrown eye phenotype composed of mostly GFP-wild-type cells (Fig. 1c,e). To confirm the overgrown eye phenotype was due to synergy between Ras and Lkb1 and not to simply preventing cell death in Lkb1 mutant cells we expressed the baculoviral caspase inhibitor p35 in Lkb1 mutant clones. Expressing p35 in Lkb1 mutant clones resulted in a majority of flies with eyes that are phenotypically similar to expression of p35 alone (normal size eye), with 20% of flies exhibiting a more severe smaller malformed eye ( Supplementary Fig. 1).
To investigate the mechanism that results in an increase in organ size in Ras Lo /Lkb1 -/flies, we analyzed BrdU incorporation in mosaic Ras Lo /Lkb1 -/eye imaginal disc tissue. Eye disc tissue carrying Ras Lo /Lkb1 -/clones exhibits BrdU incorporation in GFP-wild-type cells surrounding mutant clones (Fig. 2a). In addition, we analyzed mosaic Ras Lo /Lkb1 -/eye imaginal disc tissue by fluorescence activated cell sorting (FACS). This analysis revealed an increase in the percentage of GFP+ mutant cells in G1 when compared to GFP+ cells from 7 control FRT82B discs (Fig. 2b,c). Altogether, these data suggest that although Ras Lo /Lkb1 -/mutant cells survive, they undergo G1 arrest while promoting the increased hyperplastic proliferation of surrounding wildtype tissue.

High level oncogenic Kras promotes the neoplastic transformation of Lkb1 mutant tissue
Previous studies have implicated the dose of mutant Kras in tumor progression, cell motility, and metabolic reprogramming 7,18,19,29 , therefore we used the GFP-labeled eye expression system to clonally express Ras Hi and mutate Lkb1 in developing eye epithelia (Ras Hi /Lkb1 -/-). When combined with Lkb1 loss-of-function, expression of Ras Hi resulted in severely overgrown and disorganized 3 rd instar larval eye-imaginal disc tumors composed of mostly GFP+ mutant tissue (Fig. 3a). FACS analysis of mutant tissue revealed a shift in cell cycle phasing that favored G2/M, suggesting that mutant cells were precociously completing G1 (Fig. 2b,c). The majority of larvae carrying Ras Hi /Lkb1 -/mosaic discs did not pupate but continued to grow into 'giant larvae' while expression of Ras Hi alone resulted in late pupal lethality (Fig. 3b). The giant larval phenotype is shared by lossof-function mutations in the Drosophila neoplastic tumor suppressor genes 30 and suggests that Ras Hi /Lkb1 -/tumors are malignant. To test this, we performed an allograft assay by implanting control, Ras Lo /Lkb1 -/-, and Ras Hi /Lkb1 -/-GFP+ tumor tissue in the abdomens of wild-type hosts. Transplanted control and Ras Lo /Lkb1 -/tissue failed to grow in host abdomens (Fig. 3e). Surprisingly, the lifespan of hosts with transplanted Ras Lo /Lkb1 -/tissue was shortened which suggests that residual GFP-'wild-type' tissue from the transplant could be partially transformed. In contrast, only transplanted Ras Hi /Lkb1 -/tissue was able to grow into visible secondary tumors that significantly shortened host survival (Fig. 3e,f) thus confirming the malignancy of

High-level Ras promotes the invasion and metastasis of Lkb1 mutant tissue
Mutations in cell polarity proteins cooperate with oncogenic Ras to drive tumor cell invasion and metastasis 20 .
Previous studies have shown that Lkb1 regulates cell polarity and epithelial integrity across species 31,32 , 8 therefore, we hypothesized that malignant Ras Hi /Lkb1 -/tumors would have invasive properties. To test this, we first examined whether Ras/Lkb1 mutant cells compromised basement membrane structure by examining the expression of GFP-tagged Collagen IV (Viking (Vkg)-GFP) using conventional fixation and confocal microscopy. Compared to control and Ras Lo /Lkb1 -/tissue which shows contiguous Vkg-GFP expression in epithelia, Ras Hi /Lkb1 -/tissue exhibits breaks in Vkg-GFP expression (Fig. 3c). Expressing Ras Hi on its own is lethal (albeit at the pharate adult stage), so we investigated Vkg-GFP in this genotype and once again found no breaks in the structure of Vkg-GFP. We next assayed matrix metalloproteinase (MMP) expression, as MMPs degrade basement membrane. Compared to control, Ras Lo /Lkb1 -/-,and Ras Hi clones, Ras Hi /Lkb1 -/mutant tissues express high-levels of autonomous MMPs (Fig. 3d). Lastly, we measured the extent to which Ras Hi /Lkb1 -/cells invade local tissues by dissecting cephalic complexes and assaying extent of migration over the ventral nerve cord (VNC). Compared to Ras Hi control tissue which exhibits benign overgrowths confined to the eye-antennal discs, Ras Hi /Lkb1 -/cells completely invade contiguous organs like the brain hemispheres and VNC (Fig. 3g).
These data suggest Ras Hi /Lkb1 -/tumor cells escape the basement membrane using an active proteolytic process and invade local tissues.
Invasion and metastasis are difficult processes to visualize in living organisms . Thus far, Drosophila tumor-bearing larvae have been precluded from fast, high resolution long-term intravital imaging techniques due to their size, degree of movement, and the significant amount of light scattering throughout the body due to the larval cuticle. To address this, we prepared live tumor-bearing larvae for long-term intravital imaging and used simultaneous multiview (SiMView) light-sheet microscopy (see Methods and 33  Video 1). We defined two independent regions of interest in each larva that encompassed a tumor-adjacent tracheal branch and calculated Vkg-GFP pixel intensity every 2 hours over a 14hr imaging window. Using the wing disc of each animal as an internal control, we observed a statistically significant difference in the change in levels of Vkg-GFP over the imaging window in the tracheal branches (Fig. 4b). Volumetric rendering and surface reconstruction of the tracheal branches revealed tumor cells in contact with trachea at several hundred µm away from the primary tumors ( Fig. 4c-e)) and on rare occasion were found on the 'interior' surface of Vkg-GFP. These data suggest Ras Hi /Lkb1 -/mutant cells actively invade tracheal vascular cells to potentially spread to distant organs.  Fig. 5d to 1c). A recent study from the Guo group found that autophagy may sustain AMPK activity upon Lkb1 loss to support tumor growth 39 . In support of this, we detected increased lipidated ATG8a in Ras Hi /Lkb1 -/tumors, indicative of an increase in autophagic flux (Fig. 5a). Altogher, these data support the conclusion that activation of Ampk is maintained in Ras Hi /Lkb1 -/tumors and is required autonomously to promote malignant progression of Kras/Lkb1 tumors in vivo.

Ras Hi /Lkb1 -/malignant tumors depend on CaMK/Ampk signaling in vivo
The Ca2+/calmodulin-dependent protein kinase kinase (CaMKK2) is a nucleotide-independent activator of AMPK 37 , therefore we assayed activation of the Drosophila ortholog CamkIIB (48% identical/63% similar to CaMKK2) in our panel of mutant tissue. We found that activation of CamkIIB was elevated in Ras Hi /Lkb1 -/tumors ( Fig. 5a), suggesting a conserved role for this kinase in activating Ampk in the presence of oncogenic Ras tumors lacking Lkb1. To test whether Ras Hi /Lkb1 -/tumors are dependent on CamkIIB activity we used pharmacologic inhibition of the CaMK cascade by feeding developing Ras Hi /Lkb1 -/larvae with the inhibitor KN-93 40 , which in our model inhibited activation of the Drosophila CamkIIB by 47% (Supplemental Fig. 2b).
Treatment of Ras Hi /Lkb1 -/larvae resulted in a significant rescue of whole-organismal lethality, with an increase in the number of flies surviving to the pupal and adult stage (6.5% adult survival for KN-93 vs. 0% adult survival for vehicle control) (Fig. 5f). Taken together, these data suggest that in the context of loss of Lkb1, high levels of oncogenic Ras result in activation of Ampk by the alternative sole Drosophila CAMKK2 ortholog. Moreover, our pharmacologic results suggest that targeting the upstream AMPK/CAMKK complex may offer therapeutic benefit to KRAS/LKB1 mutant lung adenocarcinoma patients.
High levels of oncogenic KRAS and loss of LKB1 result in decreased patient survival and AMPK signaling circuit activation in the TCGA lung adenocarcinoma cohort.
To test the translational relevance of our findings in Drosophila we analyzed human lung adenocarcinoma genomic and clinical data using cBioPortal 41,42 to study how differences in levels of oncogenic KRAS affect tumor progression in LKB1 mutant patients. We used the TCGA Lung Adenocarcinoma PanCancer Atlas and TCGA Provisional Lung Adenocarcinoma datasets to select the proportion of patients with KRAS mutations in codon 12 (G12C, G12D, or G12V) for further study. We then used available RNA sequencing data to stratify patients as either RAS Lo or RAS Hi . We next investigated overall patient survival by comparing cohorts of RAS Lo or RAS Hi alone, to those that contained mono, bi-allelic loss and/or loss-of-function mutations in LKB1.
We found no difference in overall survival in RAS Lo vs. RAS Lo /LKB1 Mut patients, but strikingly RAS Hi/ LKB1 Mut patients exhibited significantly worse overall survival when compared with RAS Hi patients (Fig. 6a,b). We then tested whether KRAS copy number changes could account for the change in overall survival. Similar results were obtained when patients were stratified into either oncogenic Ras Diploid or Ras Gain/Amp (Fig. 6c,d).
Interestingly, the ability of high level vs. low level KRAS to drive survival differences did not extend to patients with TP53 mutations (Supplemental Fig. 3).
A recent study has reported that AMPK has a pro-tumorigenic role in lung cancer GEMMs with Kras and p53 mutations 43 . Moreover, data from our Drosophila Lkb1 mutant tumor model indicate that halving the genetic dose of ampk is sufficient to partially reverse whole-organism lethality. To test whether AMPK signaling may be involved in human KRAS/LKB1 mutant lung adenocarcinoma we performed a correlation analysis between pAMPK and oncogenic codon 12 KRAS mRNA for LKB1 loss-of-function and LKB1 wildtype patients using TCGA data. We detected a positive correlation trend between pAMPK and oncogenic KRAS levels, but only in LKB1 mutant patients (Spearman's correlation coefficient = 0.3, p = 0.068 for LKB1 loss-offunction vs coefficient = -0.076, p = 0.683 for LKB1 wild-type patients) (Fig. 6e,f). To further test our hypothesis, we used canonical circuit activity analysis (CCAA) 44 which recodes gene expression data into measurements of changes in the activity of signaling circuits, ultimately providing high-throughput estimations of cell function. We performed CCAA to estimate activity of the AMPK pathway in RAS Hi /LKB1 Mut lung adenocarcinoma patients compared to RAS Hi patients. The activity of three effector circuits are significantly (FDR<0.05) upregulated in KRAS Hi /LKB1 Mut patients, one ending in the node that contains PPARGC1A (encodes PGC1alpha), the second one ending in the node with the MLYCD gene, and the third ending in the node containing EIF4EBP1 (Fig. 6g). These three genes control the cellular processes of circadian control of mitochondrial biogenesis, fatty acid metabolism, and translation regulation, and are known to be upregulated in various cancers 45,46,47 . These data confirm the translational relevance of our Drosophila model, and suggest that high oncogenic KRAS levels, perhaps through copy number gains, activate specific sub-circuits of the AMPK signaling pathway to drive the malignant progression of LKB1 mutant tumors.

Discussion
Co-occurring genomic alterations in oncogene-driven lung adenocarcinoma are emerging as critical determinants of tumor-autonomous and non-autonomous phenotypes 2 . Here, we have generated the first It has been proposed that RAS-induced senescence functions as a tumor suppressive mechanism 48 . More recent data have built upon these studies to show that high levels of Hras are required to activate tumor suppressor pathways in vivo 18 , and that doubling the levels of oncogenic Kras is sufficient to cause metabolic rewiring leading to differences in therapeutic susceptibilities 19  Using CCAA, we discovered PPARGC1A, which encodes the protein PGC1α, as significantly upregulated in KRAS Hi /LKB1 Mut lung adenocarcinoma patients. Interestingly, studies in human prostate cancer have discovered metabolic adaptations through PGC1α-mediated mitochondrial biogenesis in response to CAMKKβ/AMPK signaling 51,52 . Future studies should focus on whether similar adaptations drive tumor growth and survival in KRAS/LKB1 mutant lung adenocarcinoma. In addition, work is needed to elucidate the mechanism used by high-level Ras signaling to engage the CaMK pathway. Lastly, our work is the first to 14 show that Ampk can have a pro-tumorigenic role in Lkb1 mutant cancer in vivo, and suggests that KRAS/LKB1 mutant lung adenocarcinoma patients may benefit from CAMKK inhibitors.

Materials and Methods
Drosophila stocks and maintenance. Flies were grown on a molasses-based food at 25°C. Western blotting. Twenty 3 rd instar larvae were dissected in 1X PBS and eye-imaginal discs were transferred to a 1.5ml microcentrifuge tube containing 1ml of fresh 1X PBS. Discs were spun down at 4°C for 1 min at 9,600g and supernatant was removed. 2X Laemmli Sample Buffer was added and discs were boiled for 10 minutes at 100°C, and spun down. Approximately 10µg of protein was loaded into a 12% polyacrylamide gel.
Alternatively, 3 rd instar larvae were dissected and 20µg of crude extract was loaded into a 10% polyacrylamide gel. Samples were run at 100V and separated by SDS-PAGE before transferring to a polyvinylidene difluoride (PVDF) membrane overnight at 0.07amps at 4°C. Membranes were blocked for 1 hour with 10% skim milk in 1X tris-buffered saline plus Tween 20 (TBST) and placed in primary antibody overnight in 1X TBST with 5% skim milk or BSA at 4°C. The following day, membranes were washed three times for 10 minutes each in 1X TBST and placed in secondary antibody in 1X TBST with 5% skim milk or BSA for 1 hour at RT. After three 16 additional 10-minute washes in 1X TBST, ECL-reagent (Amersham, RPN2232) and X-ray film were used to detect signals. When necessary, membranes were stripped using GM Biosciences OneMinute Plus Western Blot Stripping Buffer (GM6011). Primary antibodies and dilution: affinity purified guinea pig anti-Drosophila Lkb1

SiMView Light Sheet Microscopy
Prior to mounting, live wandering 3 rd instar Drosophila larvae and giant larvae (13 days AEL) were selected for stage and proper expression then cooled in a petri dish placed on top of an ice bucket. After sufficiently cooled to minimize movement, the samples were attached posterior side up to a 3mm diameter stainless steel post using gel-control super glue (Ultra Gel Control, Loctite). When mounting, the sample's mouthparts were adhered in an extended state in order to improve image quality (i.e. reduce object depth) of the tumors. After allowing the adhesive to dry, the sample and post was loaded into an adapter that is magnetically attached to a multi-stage stack with degrees of freedom in the X-Y-Z and rotational directions. The sample chamber is sealed using custom-made rubber gaskets and filled with Schneider's Medium. The instrument is constructed as previously published with slight modification 55,56 . All data was collected using a Nikon 16x/0.8 NA LWD Plan Fluorite water-dipping objective and Hamamatsu Orca Flash 4.0 v2 sCMOS cameras. Exposure time for all experiments was 15ms per frame. We collected data using a single camera view and two illumination arms, exciting with each arm in sequence for each color and timepoint. In our SIMView implementation for one-photon excitation, multiview image stacks are acquired by quickly moving the specimen over the desired z range and alternating light-sheet activation in the two illumination arms for each volume. This bidirectional illumination and detection capture recordings from two complementary views of each z plane in two illumination steps. Notably, no mechanical rotation of the specimen is required. The switching of laser shutters in the two illumination subsystems is performed within a few milliseconds. GFP and RFP fluorophores were excited using 488nm and 561nm Omicron Sole lasers, respectively.

Analysis of SiMView Data
Following data acquisition, images were processed prior to analysis. All data had 90 counts subtracted to account for dark counts of the sCMOS cameras. Images from each illumination arm corresponding to the same Z slice were merged and corrected for intensity variation. Details on these algorithms are previously published 57 . Vkg-GFP pixel intensity over time was measured by using maximum intensity projections of 3D volumes from 8 different time points between 0 and 14 hours. The pixel intensity for the tracheal region of interest was measured for each time point using FIJI/Image J. 3D volumetric time-lapse data were visualized using Bitplane Imaris 9 (Fig. 4 c-e). Subsets of the entire 2000-3000 timepoint series (~3 -5 TBs in size) were selected for 3D inspection and visualization from maximum intensity projection (MIP) images. 3D regions of interest (3D-ROI) were created using Imaris' intensity-based Surfaces function.
Pharmacology. Molasses-based food was melted and 10ml of food was aliquoted to vials. While warm, 10µl of H 2 0 or 10µl of 5mM KN-93 (Millipore Sigma, 422711) were added to vials, respectively. Food vials were cooled and allowed to solidify before use. Vials not immediately used were placed at 4°C. Adult y,w, eyFLP1; Act >y+> Gal4, UAS-GFP; FRT82B, Tub-Gal80 virgin female flies were crossed to FRT82B or UAS-RasV12 Hi /Lkb1 4A4-2 males, respectively. Flies were moved to embryo "egg-laying cups" and allowed to egg-lay onto grape juice agar plates at 25°C. Flies were moved onto fresh agar plates every 24 hours. After each 24hr period, embryos were collected using forceps and placed onto a fresh vial of food. Embryos were placed at 25°C and allowed to hatch. Once of age, 2 nd -instar larvae were collected and placed onto drug containing media at 25°C. Survival was quantified as the percentage of total embryos placed that survived to pupation and adulthood.
Survival analysis of patient data. cBioPortal was used to obtain survival, copy number, mRNA expression, and RPPA expression data available through the Cancer Genome Atlas (TCGA). For survival analysis, specific studies used included: TCGA Pan-Lung Cancer study 58  Statistical Analysis. GraphPad Prism 7 and 8 were used to generate P values using the two-tailed unpaired Student's t-test to analyze statistical significance between two conditions in an experiment, ordinary one-way ANOVA with a Tukey's multiple comparisons test for experiments with three or more comparisons, and Logrank (Mantel-Cox) test for analysis of survival data. Significance was assigned to P values less than 0.05 unless otherwise indicated. For Figure 6e and f, statistical analysis was conducted using RStudio. Data was divided into two groups, LKB1 loss of function (n = 40) and LKB1 wild type (wt) (n= 31). A single outlier sample in the LKB1 mutation category was excluded and calculated z-score for pAMPK and KRAS expression data was used. The correlation between the AMPK and KRAS was conducted and a spearman's correlation test. Due to the relatively small sample size, a p-value of <= .1 or 10% was considered significant.

Supplementary Materials
Materials and Methods  Images are representative of n=10 independent flies per genotype. Scale bar, 100µm.