Cooperation between oncogenic Ras and wild-type p53 stimulates STAT non-cell autonomously to promote tumor radioresistance

Oncogenic RAS mutations are associated with tumor resistance to radiation therapy. Cell-cell interactions in the tumor microenvironment (TME) profoundly influence therapy outcomes. However, the nature of these interactions and their role in Ras tumor radioresistance remain unclear. Here we use Drosophila oncogenic Ras tissues and human Ras cancer cell radiation models to address these questions. We discover that cellular response to genotoxic stress cooperates with oncogenic Ras to activate JAK/STAT non-cell autonomously in the TME. Specifically, p53 is heterogeneously activated in Ras tumor tissues in response to irradiation. This mosaicism allows high p53-expressing Ras clones to stimulate JAK/STAT cytokines, which activate JAK/STAT in the nearby low p53-expressing surviving Ras clones, leading to robust tumor re-establishment. Blocking any part of this cell-cell communication loop re-sensitizes Ras tumor cells to irradiation. These findings suggest that coupling STAT inhibitors to radiotherapy might improve clinical outcomes for Ras cancer patients. Dong, Vadla et al. find that oncogenic Ras and p53 cooperatively stimulate JAK/STAT cytokines to promote non-autonomous tissue overgrowth in Drosophila. Further, blocking STAT signalling prevents the growth-inducing effect of p53 overexpression and ionising irradiation in human cancer lines in culture and in mouse xenografts, suggesting a therapeutic route.

O ncogenic Ras mutations activate a complex network of interacting signals to cause aggressive cancers 1,2 . Gold standard treatment options include radiation therapy and conventional chemotherapies that cause irreversible genomic damage and trigger apoptosis 3 . However, oncogenic Ras mutations enable cancer cells to resist these genotoxic agents, ultimately leading to cancer recurrence [4][5][6][7][8][9][10][11][12][13] . We define tumor radioresistance as incomplete sensitivity and/or the capacity of tumors to rapidly re-form following radiation therapy.
Various mechanisms have been proposed to explain the resistance of Ras cancers to treatments, including the presence of cancer stem cells in the tumor microenvironment 5,[14][15][16] . Another view is that therapy-resistant cancer cells possess robust DNA repair mechanisms that curtail the proapoptotic effect of the treatment. In many cancers, including lung and colorectal cancers where oncogenic Ras mutations are common, an association between polymorphisms in DNA damage response genes and an improved clinical response to genotoxic agents has been observed [17][18][19][20][21][22][23] . However, cellular responses to DNA damage are complex and include activation of cell-cell interactions that we do not fully understand 24 . How these nonautonomous effects influence the response of Ras-driven cancers to genotoxic therapies is an underexplored area of research. Animal tumor models provide the advantage of interrogating tumor resistance mechanisms at the tissue level, enabling the identification of novel and broadly applicable mechanisms.
In genetic screens for suppressors of oncogenic Ras (Ras V12 )mediated tissue overgrowth in Drosophila 25,26 , we isolated genotoxic mutations, including null alleles of the Pax2 transactivation domain-interacting protein coding gene (ptip −/− ). Interestingly, ptip −/− inhibits the growth of Ras V12 cells but also triggers the overgrowth of the surrounding tissues. PTIP is essential for maintaining genomic stability under normal conditions and after DNA damage [27][28][29] . Disruption of the PTIP DNA repair complex causes genomic instability and triggers a DNA damage response that culminates in the activation of p53 (dp53 in Drosophila), which orchestrates DNA repair or triggers apoptosis of damaged cells 29,30 .
It is becoming evident that p53 biology is far more complex than initially thought and involves nonautonomous functions that are not well-understood [31][32][33][34] . We found that ptip −/− causes genomic instability and consequently upregulates dp53 in Ras V12 cells. This upregulation of wild-type dp53 cooperates with oncogenic Ras signaling to stimulate the secretion of JAK/STAT (Janus kinases/signal transducers and activators of transcription) ligands (interleukin 6-related cytokines known as unpaired in Drosophila). These ligands activate JAK/STAT in the surrounding cells, leading to tissue overgrowth. Ionizing radiation (IR) of Drosophila Ras V12 tissues or of human cancer cells harboring oncogenic Ras mutations triggers similar nonautonomous effects. Blockade of any part of this p53/Ras V12 -STAT signaling relay inhibits the nonautonomous growth effect and resensitizes Ras V12 tissues to IR treatment.
In addition to highlighting the complexity of p53 biology, our work defines a treatment-induced cell-cell interaction dynamic that promotes the recurrence of oncogenic Ras mutant tumors after genotoxic therapies. Our data also provide a possible explanation for why some Ras mutant cancers resist genotoxic therapies despite the lack of p53 mutations.

Results
Ptip −/− promotes nonautonomous tissue overgrowth. In Drosophila, the MARCM (mosaic analysis with a repressible cell marker) technique permits the expression of oncogenic Ras (Ras V12 ) in clones of cells within the developing eye epithelium 26 .
These clones coexpress a fluorescent protein, making them distinguishable from the surrounding wild-type cells. Ras V12 -mediated tissue overgrowth is readily detectable by the appearance of large and hyperplastic fluorescent clones (tumors) that ultimately kill the animal [35][36][37] . Ras V12 suppressor mutations are isolated through the identification of mutations that significantly reduce the clone size and rescue animal viability when introduced in Ras V12 -expressing cells 35 .
We isolated several Ras V12 suppressors, including mutation #3804, using this approach. Mutation 3804 potently suppresses Ras V12 -mediated tumor overgrowth and yields viable adult animals ( Fig. 1a versus 1b; 1c versus 1d; 1g-l). To determine whether this mutation synthetically suppresses oncogenic Ras or is cell deleterious on its own, we generated wild-type or 3804 mutant clones in developing and adult eye tissues to determine whether this mutation synthetically suppresses oncogenic Ras or is deleterious to the cell itself. Adult eye clones are marked by the absence of the red pigment. As expected, wild-type cells contributed~50% to the eye field. In contrast, 3804 mutant cells were barely detectable in tissues, suggesting that the affected gene is essential for cell viability (Fig. 1m-o).
Interestingly, although 3804 suppressed Ras V12 tumor overgrowth in the developing eye tissue and correspondingly yielded adult animals, the eyes of the rescued animals were overgrown to varying extents, as evidenced by the appearance of tissue folds (Fig. 1e, f versus 1g-l, w). The overgrown tissues exclusively consisted of wild-type cells (GFP-negative) ( Fig.1p-s).
We used deficiency mapping and allele sequencing approaches and determined that 3804 represents a null mutation in the PAX transcriptional activation domain-interacting protein (PTIP). Homozygosity of the 3804 mutation is animal lethal. Two independent deficiency alleles (ED4529 and ED4536) that overlap at the PTIP gene fail to complement 3804 animal lethality (Fig. 1t). Direct sequencing of the 3804 allele revealed a G>A mutation leading to a premature stop codon in the protein (Fig.1u, v). Thus, we hereafter refer to 3804 as ptip −/− and conclude that while ptip −/− suppresses growth in a cell-intrinsic manner, it cooperates with oncogenic Ras to promote the growth of the surrounding tissue.
Ptip −/− promotes nonautonomous tissue overgrowth via p53. We sought to delineate the underlying mechanism. PTIP was originally identified in a yeast two-hybrid screen 38 . In mammals, PTIP interacts with histone methyltransferase complexes to control developmental transcription programs 38,39 . Additionally, PTIP is essential for maintaining genomic stability 29,40,41 . Genomic instability triggers the DNA damage response signals, which culminate in the sequential activation of p53 and p21 (Dacapo or dap in Drosophila) to drive cell cycle arrest or apoptosis [42][43][44] . We investigated whether ptip −/− causes DNA damage in Drosophila Ras V12 epithelial cells and found that it does. DNA damage triggers the phosphorylation of a histone 2A variant (γH2Av), which is readily detected in immunostaining experiments using phospho-specific antibodies 45,46 . We examined γH2Av in ptip −/− cells or in Ras V12 cells with or without the ptip −/− mutation. We observed an increase in the number of γH2Av nuclear foci in ptip −/− and Ras V12 ptip −/− double mutant cells, but not in cells expressing Ras V12 alone ( Fig. 2a-c′). In addition, quantitative polymerase chain reaction (qPCR) assays revealed transcriptional upregulation of dp53 and dap/p21 in ptip −/− or Ras V12 ptip −/− double mutant tissues compared to wild-type controls (Fig. 2j, k). Consistent with DNA damage, ptip −/− caused an increase in dp53 and dap/p21 protein levels in Ras V12 cells (Fig. 2d-g′).
We posited that the cellular response to genomic stress likely underlies the nonautonomous growth effect of ptip −/− on the Ras V12 clones. Janus N-terminal kinase (JNK, also known as Bsk in Drosophila) and p53 play central roles in cellular response to DNA damage [47][48][49] . In addition, we and other researchers have shown that JNK promotes nonautonomous tissue growth in Drosophila 35,50,51 , suggesting that ptip −/− acts via JNK to drive nonautonomous growth in Ras V12 mosaic tissues. Consistent with this hypothesis, JNK was activated in Ras V12 ptip −/− mutant cells compared to Ras V12 control cells (Fig. 2h-i′). We inhibited JNK by expressing a potent dominant-negative JNK transgene (Bsk DN ) in Ras V12 ptip −/− cells (Ras V12 ptip −/− Bsk DN triple defective cells) and asked whether this genetic manipulation suppresses the nonautonomous tissue overgrowth phenotype to directly test this hypothesis. Bsk DN failed to suppress Ras V12 ptip −/− nonautonomous tissue overgrowth (Figs. 2l-n and 2l'-n'), making it unlikely that JNK plays a significant role in this phenomenon.
We explored alternative mechanisms. Nonautonomous growthinducing clones (Ras V12 ptip −/− mutant cells) showed higher levels of the wild-type p53 protein (p53 wt ) than Ras V12 cells, which did not cause nonautonomous growth (Fig. 2d-e′). Because the ptip −/− mutation occurs very early and is permanent, the resulting high p53 wt protein levels (Fig. 2d, d′, j, and k) likely persist throughout the life of Ras V12 cells. Normally, p53 wt has a high turnover rate 52 . We wondered whether the elevated p53 wt protein levels observed in the Ras V12 ptip −/− mutant cells play an active role in the nonautonomous tissue growth effect. Indeed, RNAi knockdown of dp53 in Ras V12 ptip −/− cells remarkably suppressed nonautonomous tissue overgrowth (Fig. 2m, m′, o, o′, and s). Similarly, blocking dp53 transcriptional activity in Ras V12 ptip −/− cells by expressing a DNA binding-defective dp53 mutant version (p53 R155H ) 53 also suppressed the overgrowth of the surrounding wild-type cells (Fig. 2m, m′, p, p′, and s). In addition, direct overexpression of p53 wt (p53 OE ) in clones of Ras V12 cells was sufficient to trigger the overgrowth of the surrounding wild-type tissue, mimicking Ras V12 ptip −/− clones (Fig. 2l, l′, r, and r′). The ability of p53 wt to drive nonautonomous tissue overgrowth requires oncogenic Ras. Overexpression of dp53 wt alone failed to generate a similar effect (Fig. 2l, l′, q, and q′).
We used the MARCM technique to juxtapose Ras V12 p53 OE clones (RFP-labeled) with Ras V12 clones (GFP-labeled) and assessed whether this Ras V12 /p53 cooperation similarly accelerates the growth of adjacent Ras V12 cells (see Methods). This alteration caused massive overgrowth of Ras V12 clones compared to controls (abutting Ras V12 clones without dp53 OE ) (Fig. 2t-u″ and s). Ras V12 ptip −/− clones exerted a similar nonautonomous effect on Ras clones ( Supplementary Fig. 1). Taken together, these findings indicate that oncogenic Ras cooperates with elevated levels of the wild-type dp53 protein to drive tumor overgrowth via a novel nonautonomous mechanism.
Oncogenic Ras and p53 cooperatively stimulate JAK/STAT cytokines to promote nonautonomous tissue overgrowth. We set out to delineate the underlying mechanism. The Drosophila JAK/STAT ligands Unpaired1-3 (upd, upd2, and upd3) mediate nonautonomous tissue growth 35,50,51 . We asked to what extent oncogenic Ras and dp53 OE act via the JAK/STAT pathway. Immunostaining experiments using a upd reporter line, upd-lacZ 54 , to monitor upd transcriptional activity revealed that dp53 OE causes Ras V12 cells to upregulate upd ( Fig. 3a-b′). In a complementary qPCR approach, we found that dp53 OE causes Ras V12 cells to upregulate all of the upd ligands (upd1-3) in tissues (Fig. 3d). The ptip −/− mutation exerted similar effects on Ras V12 tissues, including activation of JAK/STAT in cells surrounding the mutant (Ras V12 ptip −/− ) clones (Fig. 3a, a′, c, c′ and Supplementary Fig. 2a-d and 2g-j). We inhibited p53 via RNAi knockdown or by expressing dominant-negative dp53 (p53 R155H ) in Ras V12 ptip −/− cells to further establish that Ras V12 ptip −/− tissues rely on dp53 for the stimulation of upd and found that each of these manipulations blocked the upregulation of upd ligands (Fig. 3e). These findings suggest that Ras V12 p53 OE clones induce the growth of surrounding cells via the secretion of JAK/ STAT cytokines. We blocked the secretion of JAK/STAT cytokines into the tissue surrounding Ras V12 p53 OE clones and asked whether this blockade suppresses the nonautonomous growth effect to functionally test this hypothesis. We simultaneously knocked down upd and upd2 in Ras V12 p53 OE clones by combining upd-RNAi expression with upd2 deletion mutants. This manipulation dramatically reduced the tissue size ( Fig. 3f-h). Knockdown of Upd in Ras V12 ptip −/− clones similarly suppressed nonautonomous tissue overgrowth ( Supplementary Fig. 2e, f, k, and l).
We evaluated our findings in human breast and lung cancer cells using supernatant transfer experiments. MCF-10A breast epithelial cells were cultured in media conditioned with MCF-10A cells (controls) or MCF-10A cells overexpressing wild-type p53 alone or coexpressing oncogenic HRAS. STAT signaling status was assessed using western blot experiments with antibodies that specifically detect activated STAT (anti-phosphorylated STAT). Growth was determined by scoring cell numbers. In this and subsequent experiments, the superscripts "P53OE" or "HRasG12V" denote overexpression of wild-type p53 or oncogenic Ras, respectively. As expected, MCF10A HRasG12V, P53OE cells showed elevated levels of the p53 protein and Ras signaling (determined by phospho-ERK) levels compared to untransfected controls ( Fig. 4a, b). MCF10A HRasG12V, P53OE -conditioned media stimulated STAT signaling and correspondingly induced the growth of MCF-10A cells (Fig. 4c, f, and supplementary Fig. 6). This growth-promoting effect was significantly reduced when the conditioning cells lacked oncogenic Ras (MCF-10A P53OE ) ( Fig. 4f and supplementary Fig. 6).
Moreover, we tested our findings in vivo by performing mouse xenograft experiments. We inoculated one million A549 cells into each of the flanks of nude mice. Left flank inoculants were unmodified, while right flank inoculants consisted of a 50-50 mixture of untreated and irradiated cells. Eight weeks after treatment, tumor xenografts from the mixed population (right flanks) grew markedly larger than xenografts arising from the homogenous inoculants (left flanks) in the same animal ( Fig. 4j and k; 83%; N = 6 animals). We treated animals with the validated pharmacological STAT blocker Ruxolitinib to test whether STAT plays a role in the observed tumor overgrowth, as suggested by our tissue culture and fly data 57 . Animals were treated orally with Ruxolitinib at 10 mg/kg, a dose that is well tolerated in nude mice 57 . Ruxolitinib suppressed the overgrowth of tumors arising from the mixed cells (Fig. 4p, q). Endpoint western blot analyses of tumor xenografts harvested from animals that were treated with or without Ruxolitinib confirmed the inhibition of STAT signaling in the treated and slower growing mixed tumors (Fig. 4r).
JAK/STAT signaling supports tissue growth by promoting cell survival or cell proliferation 58,59 . We examined cell death and cell proliferation in human cells using flow cytometry approaches to distinguish between these two mechanisms. Notably, p53 OE and IR-induced nonautonomous JAK/STAT signaling mainly stimulated cell proliferation ( Supplementary Figs. 3, 4).
Taken together, the above data indicate that stimulation of wild-type p53 cooperates with oncogenic Ras to induce JAK/ STAT signaling in the surrounding cells, resulting in nonautonomous growth.
We used Drosophila Ras V12 tumor tissues exposed to IR to test whether IR-stimulated dp53 generates similar growth-promoting effects in tissues. Specifically, we asked whether dp53 cooperates with oncogenic Ras to establish tumor recurrence via the nonautonomous STAT signaling relay described above. (a-i′) Representative images of dissected eye imaginal discs containing ptip −/− or Ras V12 or Ras V12 ptip −/− double mutant clones (GFP) stained with DAPI to detect DNA or anti-phosphorylated H2AV antibodies to detect DNA damage (a-c') or anti-p53 (d-e′) or anti-dacapo (dap/p21) (f-g′) or anti-phosphorylated JNK (h-i′) antibodies to detect cellular response to DNA damage. Scale bars are 20 µm. (j, k) Quantitative Polymerase Chain Reaction (qPCR) data showing expression of p53 or dap/p21 in wild type versus ptip −/− eye imaginal discs (j) or the expression of p53 in Ras V12 or ptip −/− or Ras V12 ptip −/− eye imaginal discs (k). Expression was normalized to the transcript abundance of the housekeeping gene rp49. Error bars denote standard deviation (SD) values. P values are derived from Student's t test analyses. (l-r) Matched light and fluorescence images of adult eyes containing GFP-labeled clones. The respective clone genotypes are indicated at the top of each panel. The corresponding fluorescent images are shown below in (l′-r′). GPF-negative tissues represent wild-type tissues. Scale bars are 150 µm. (s) Quantification of the nonautonomous growth phenotype of adult eyes containing clones of the indicated genotypes: Ras V12 ptip −/− , Ras V12 ptip −/− Bsk DN , Ras V12 ptip −/− p53 R155H , or Ras V12 ptip −/− p53 RNAi . (t-u″) Genetic juxtaposition of GFP-labeled Ras V12 clones with RFP-labeled Ras V12 clones (t-t″, controls) or with RFP-labeled clones of cells coexpressing Ras V12 and wild-type p53 (Ras V12 , p53 OE ) (u-u″). GFP-positive Ras V12 clones are surrounded by RFP/GFP double-positive Ras V12 clones (t-t″) or by RFP/GFP double-positive Ras V12 , p53 OE clones (u-u″). Brain cephalic complex images showing the growth of Ras V12 clones when juxtaposed to Ras V12 or to Ras V12 , p53 OE clones are shown in t and u, respectively. Dotted white lines (t′, t″, u′, u″) represent tissue boundaries. Scale bars are 100 µm. (v) Quantification of eye tissue sizes from (t-u″). Sample size N = 10 tissues per genotype. Error bars denote standard error of the mean (SEM) values. P values are derived from Student's t test analyses. Effect size (Cohen's d values) for (j), (k), and (v) is greater than 0.8.
We first performed a study to determine a dose that generates cellular effects without grossly impeding animal development. Second-instar larvae harboring GFP-labeled oncogenic Ras clones in eye imaginal discs were treated with 600 R, 1000 R, or 2000 R. Each dose was administered three times in 6 hours intervals to mimic clinical settings where total radiation treatments are administered in fractions 60,61 . Larvae treated with the 3 × 600 R dose developed normally into adults without any detectable abnormalities, and those treated with 3 × 2000 R died during the pupal stage. Larvae that received 3 × 1000 R yielded adult flies with mild rough eyes, making it an ideal dosing regimen for our study.
We next determined the extent to which IR recapitulates fundamental aspects of the Ras V12 /p53 OE -STAT signaling relay, namely, stimulation of both dp53 and JAK/STAT cytokine production. Compared to nontreated Ras V12 control tissues, IR increased levels of the dp53 protein in the immunostaining experiments. Notably, dp53 stimulation was nonuniform, and dp53 was undetectable in portions of wild-type (GFP-negative) and Ras V12 cells (GFP-positive) (Fig. 5a-b′). This mosaicism supports our finding that Ras V12 cells expressing high levels of dp53 protein stimulate the growth of the surrounding Ras V12 cells with lower dp53 levels (Fig. 2o-p″). Our qPCR data showed that IR transcriptionally stimulates all unpaired cytokines (upd1-3) (Fig. 5e). Similar to ptip −/− -induced upd stimulation, IRtriggered upregulation of upd cytokines in Ras V12 cells was lost when we introduced dominant-negative dp53 (p53 R155H ) (Fig. 5f). Collectively, our data indicate that irradiation transcriptionally induces the production of upd cytokines downstream of dp53.
We sought to directly test the functional relevance of IRinduced STAT signaling in Ras V12 tumor radioresistance in Ras tissues. IR reduced the size of wild-type clones and the overall tissue size, as expected (Fig. 5h, h′, and m). In sharp contrast, IR increased the Ras V12 clone size and failed to reduce the overall tissue size (Fig. 5i, i′, m, n, and n′). We extended our analysis to other tumor signaling contexts, tissues containing clones of cells carrying homozygous null mutations in tumor suppressors (Salvador/Sav or tuberous sclerosis/Tsc) or tissues containing clones of cells overexpressing the oncogene dMYC, to determine whether this resistance is unique to oncogenic Ras signaling. In all of these tissues, IR effectively reduced clone and overall tissue sizes ( Fig. 5i-l′, and m). Thus, similar to humans, Drosophila oncogenic Ras-driven tumors are uniquely radioresistant.
Next, we asked to what extent the depletion of dp53 or JAK/ STAT cytokines sensitizes Ras V12 tissues to IR. We induced Ras V12 clones in wild-type or dp53 −/− null mutant discs, assessed whether the dp53 −/− mutation sensitizes Ras V12 tumor tissues to IR and found that it does (Fig. 5n-p, and u).
Next, we simultaneously depleted Upd1 and Upd2 cytokines from Ras V12 cells and asked whether this manipulation also sensitizes Ras V12 tumor tissues to IR. Two independent approaches were used. First, we introduced an upd2 null mutation into Ras V12 cells coexpressing Upd1-RNAi (Ras V12 ,upd −/− ,Upd1-RNAi clones). Second, we generated clones of cells coexpressing Ras V12 , Upd-RNAi and Upd2-RNAi. Both approaches abolished the capacity of Ras V12 tissues to grow following IR treatment, supporting a sensitization effect (Fig. 5n, n′, q-r′, and u). Similarly, specific inhibition of the JAK-STAT receptor domeless in Ras V12 cells via the expression of a potent dominant-negative protein version (domeless-DN) 62 sensitized Ras V12 tissues to IR (Fig. 5n, n′, s-t′, and u).
Thus, in Drosophila and human cancer cells, the p53 response to genomic instability cooperates with oncogenic Ras to induce JAK/STAT activation in surrounding cells. This nonautonomous effect stimulates tumor growth and promotes the rapid recurrence of oncogenic Ras tumors.

Discussion
Oncogenic Ras mutations are associated with resistance to genotoxic therapies. The underlying resistance mechanism remains poorly understood. The molecular responses of tumor cells to genotoxic stress have been investigated mainly in isolated cells, which limits our ability to capture the broader tissue-level tumor biology.
Using Drosophila eye tissue in a clonal genetic screen for Ras V12 suppressors, we unexpectedly isolated the genotoxic mutation ptip −/− . Interestingly, in addition to blocking Ras tumor growth, ptip −/− stimulates the surrounding Ras V12 tissue to overgrow, mirroring the resistance of Ras V12 tumors to genotoxic therapies. This nonautonomous effect stems from cooperation between oncogenic Ras signaling and ptip −/− , as oncogenic Ras or ptip −/− mutant cells alone do not cause nonautonomous growth. Our mechanistic studies reveal that this cooperation is centered on p53.
P53 is broadly known as a tumor suppressor gene. The majority of p53 mutations in human cancers are missense mutations that stabilize the p53 protein, leading to elevated levels of the mutant p53 protein in cancers. These gain-of-function mutations interfere with the canonical tumor suppressor role of p53 while causing it to function as an oncogene [63][64][65][66] . The accumulation of mutant p53 is associated with aggressive cancers [63][64][65][66][67] . Interestingly, overexpression of wild-type p53 is also observed in many cancers lacking p53 mutations, including in lung cancers where oncogenic Ras mutations are common [68][69][70][71][72] . How wild-type p53 overexpression relates to oncogenic Ras cancers and their resistance to genotoxic therapies remained unclear. Here, we show that genotoxic stress-activated p53 acts non-cell autonomously to promote the radioresistance of Ras mutant tumor tissues. The ptip −/− mutation causes genomic instability in Ras V12 cells, resulting in the upregulation of the dp53 protein. This stimulation of p53 is essential for the nonautonomous tissue overgrowth effect of Ras V12 ptip −/− tumor clones. RNAi depletion of dp53 in Ras V12 ptip −/− clones abrogates the nonautonomous tissue overgrowth effect. In addition, direct overexpression of dp53 in Ras V12 clones is sufficient to trigger overgrowth of the surrounding tissues, mimicking the ptip −/− mutation.
The nonautonomous tissue overgrowth effect is mediated by the dp53 transcriptional program. Expression of a DNA bindingdefective dp53 mutant (p53 R155H ) 53 in Ras V12 ptip −/− cells blocks the nonautonomous tissue overgrowth effect. Consistent with this finding, the transcriptional program of wild-type p53 is modified in cancer-associated fibroblasts to promote cancer progression 73 . Additionally, under ectopic wild-type p53 conditions, oncogenic Ras modifies the p53 transcriptional program, leading to the senescence-associated secretory phenotype (SASP) 74 depletion of Upd cytokines in Ras V12 ptip −/− or Ras V12 p53 OE clones suppresses nonautonomous tissue overgrowth. Increases in wild-type p53 levels either via IR, the ptip −/− mutation alone, or direct dp53 overexpression are sufficient to stimulate upd cytokine production ( Fig. 5e and Supplementary Fig. 5c-e), but these effects do not cause nonautonomous growth, likely because these cells are quickly eliminated in the absence of oncogenic Ras.
The cooperative effect of oncogenic Ras and p53 on paracrine STAT signaling may also reflect an ability of Ras to rewire the dp53 transcriptional program and/or to increase the exocytosis of STAT cytokines above a required threshold. Consistent with these possibilities, oncogenic Ras stimulates exocyst in flies and mammals, and we identified and validated dp53-binding sites near upd genes. Interestingly, the deletion of these sites significantly suppressed upd expression but did not restore upd expression to basal levels in reporter assays ( Supplementary  Fig. 5), possibly because of cryptic dp53-binding sites located near upd genes. Notably, p53 binds to noncanonical DNA sites to expand its transcriptional network 77 . In the future, it would be desirable to map these noncanonical p53 sites on upd in order to better understand p53 function.
The nonautonomous Ras/p53-STAT signaling relay allows Ras mutant clones to resist the damaging effects of IR treatment in Drosophila, as dp53 or STAT depletion sensitizes Ras mutant tumor tissues to IR. Wild-type dp53 is stimulated nonuniformly in irradiated Ras mutant tissues. This heterogeneity might be due to stochastic variation or reflect different cell-inherent capacities to successfully withstand genotoxic stress. Treatment-induced p53 heterogeneity within Ras mutant tumor tissues would allow cells with extensive genomic insults (high dp53 levels) to directly induce the upregulation of JAK/STAT ligands, which instructs nearby less-damaged (low dp53) Ras V12 cells to overproliferate and reestablish the tumor following treatment.
In vitro supernatant transfer and mouse xenograft experiments revealed a similar mechanism in human Ras cancer cells. Indeed, activation of STAT signaling is associated with resistance to genotoxic agents in human cancer cells [78][79][80] . Compared to controls, media conditioned with irradiated or p53-overexpressing Ras cancer cells elevate STAT signaling and stimulate cell proliferation across genetically diverse cancer cells. We propose that the Ras-p53 nonautonomous STAT signaling relay likely represents a tissue-level adaptive mechanism for selecting and expanding therapy-resistant tumor clones in the tumor microenvironment. This mechanism is reminiscent of the paracrine activation of TGFα/amphiregulin signaling by oncogenic Ras to establish resistance to EGFR blockade in colorectal cancers 81 .
In addition to highlighting an emerging role for p53 in cell-cell interactions, our findings provide a possible explanation for the paradoxical resistance of Ras cancers to genotoxic therapies, despite functional p53 82,83 . Thus, our data suggest that combining STAT inhibition with radiation therapy may improve clinical outcomes.
Developmental and regenerative signaling contexts may functionalize p53 in a similar manner to maintain tissue homeostasis. Neighboring cells with different levels of wild-type p53 influence the growth of other cells in Drosophila and mammalian tissues 84 .
(e, f) Quantification by qPCR of upd, upd2, and upd3 expression in eye-antennal discs containing wild-type or Ras V12 clones after 36 h of first fraction of IR treatment (IR+) or without IR treatment (IR−). Column bars represent the mean of fold changes for the expression level of indicated genes (e). Relative expression of upd2 and upd3 in irradiated eye-antennal discs containing wild type, Ras V12 and Ras V12 p53 R155H clones (f). Three independent experiments were carried out. Error bars denote SD. P values are derived from Student's t test analyses. (g) Diagram of setting Drosophila irradiation models. Larvae after egg laying (48 h) were irradiated with three fractions of 10 Gy and allowed to recover to late third-instar larval stage. All eye-antennal discs were dissected at the late third-instar larval stage to evaluate the irradiation results by measuring the relative size between GFP-labeled clones and whole eyeantennal discs. (h-l′) GFP-labeled clones homozygous for Ras V12 (i, i′), sav 3 (j, j′), Tsc1 Q600X (k, k′), or expressing dMyc (l, l′) as well as wild-type controls (h, h′) were induced in the eye-antennal discs of larvae, irradiated at 48 h, and then collected discs on day 5. (h-l) the eye-antennal discs without irradiation treatment (IR−). (h′-l′) show irradiated discs (IR+). (m) Quantification of relative eye disc size (blue) and GFP-clone size (green) treated with IR (IR+) or without IR (IR−). For each genotype, eye-antennal disc and GFP-clone were normalized to age-matched eye discs without IR. Column bars represent the mean size of samples (N = 5-10). Scale bar is 50 µm. (n-t′) GFP-labeled Ras V12 , p53 −/− , Ras V12 p53 −/− , upd RNAi upd2 Δ , Ras V12 Upd RNAi upd2 Δ , Dome DN , and Ras V12 Dome DN clones were induced in the eye-antennal discs and half were then irradiated at the second-instar larval stage. After 3 days of recovery, all eye discs at the late third-instar larval stage were dissected to evaluate the differences in response to irradiation. (n-t) Eye-antennal discs without irradiation treatment (IR−) and (n′-t′) eye discs treated with irradiation (IR+). Scale bar is 50 µm. (u) Quantification of clones and eye discs treated with or without irradiation. Eye-antennal disc and GFP-clone areas were measured by ImageJ and normalized to the eye-antennal discs with the same genotype at the same age without IR. Column bars represent the mean size of samples (N = 5-9). Blue columns represent the mean size of the entire eye-antennal tissue for the indicated genotypes; green columns represent the size of GFP-labeled tumors. Error bars denote SEM. P values are derived from Student's t test analyses. Effect size (d) values for e, f, m, and u are greater than 0.8.
anti-p21 (1:200, DSHB) and mouse anti-H2Av monoclonal antibody (1:200, DSHB #UNC93-5.2.1). Secondary antibodies were purchased from Life Technologies. Images were acquired on a Leica SP8 confocal microscope. Measurements of tumor clone size within imaginal discs were performed from confocal pictures using Fiji imageJ software. Adult eyes were imaged with a Leica DFC 300FX camera in a Leica MZ FLIII fluorescence stereomicroscope.
Ganetespib cytotoxicity. MTT assays on cancer cells treated with increasing concentrations of Ganetespib (0.1-100 nM) determined the half maximal inhibitory concentration (IC50) at 25 nM. Cells were seeded in 96-well plates at a density of 1c10 5 cells per well and treated with Ganetespib (Biosciences, #A11402,) for 48 h. A 100 μl solution of MTT (100 μg/ml) was added to each well. Cell viability was measured with a spectrophotometer at 570 nm.
Human cancer cells irradiation. For x-irradiation conditioned media experiments, cells were cultured in fresh media prior to irradiation (8 Gy, 280 cGy/min exposure) using an X-RAD 320 Biological Irradiator. Media were replaced immediately following irradiation, and the irradiated cells were cultured for 24 h to generate conditioned media. Supernatant from the irradiated cells was collected after centrifugation at 500 × g for 2 min to remove cellular debris.
Flow cytometry analysis. GFP cell sorting was performed on the MoFloXDP (Beckman Coulter), and the regular flow cytometer analysis was performed using a CyAn ADP (Beckman Coulter). Flow cytometry cell proliferation assays were performed using the 5-ethynyl-2'deoxyuridine (EdU) assay kit (Invitrogen #C10634). Cells were incubated with EdU for 24 h and labeled. The Click-iT reaction was performed using the Click-iT EdU assay kit with Alexa Fluor 647 fluorophore, according to the manufacturer's instructions. The propidium iodide (PI) solution was at 10 µg/ml in PBS containing 1% BSA. Gating was set at 488/636 nm (excitation/emission) or at 633/660 nm to detect and score PI or EdU_Alexa647-positive cells, respectively.
Mouse xenograft experiments. The mouse study was performed with the approval and oversight of the Institutional Animal Care and Use Committee at the University of Missouri (Protocol#9501). 13-weeks old Fox n1<Nu> homozygous male and female mice were used. Thirteen-week-old athymic nude mice (homozygous Foxn1 nu ) were purchased from the Jackson Laboratory for xenografts experiments. Experiments were conducted in compliance with the National Institute of Health's guide for the care and use of animals. All animals were housed under pathogen-free conditions on a 12/12 h light-dark cycle. The A549 cells were grown in Dulbecco's Modified Eagle Medium with L-glutamine and high glucose supplemented with 10% FBS. Cells were collected in pharmaceutical grade PBS, counted, and resuspended in pharmaceutical grade PBS at 1 × 10 6 /100 µL. All the left flanks were inoculated with 1 × 10 6 A549 cells. Right flanks were inoculated with a 50/50 mixture of A549 and irradiated A549 (0.5 × 10 6 + 0.5 × 10 6 cells) in a total of 100 µL. Tumor measurements were taken weekly using a digital caliper. The STAT inhibitor Ruxolitinib (Sigma Millipore #ADV390218177) was administered (10 mg/kg body weight) through oral gavage once a day for a period of three weeks.
Statistics and reproducibility. All experiments were performed in three biological replicates for reproducibility. Standard deviations represent at least three biological replicates. Student's t test was used to determine the statistical significance of differences between groups. Effect size was determined by calculating Cohen's d value [d = |mean (group1) -mean (group2) |/Pooled standard deviation].

Data availability
All source data are provided in Supplementary Data 1. All other data are available from the corresponding author on reasonable request.