Oncogenic Ras-driven Dorsal/NF-κB signaling contributes to tumorigenesis in a Drosophila carcinoma model

Cancer-driving mutations synergize with inflammatory stress signaling pathways during carcinogenesis. Drosophila melanogaster tumour models are increasingly recognized as models to inform conserved molecular mechanisms of tumorigenesis with both local and systemic effects of cancer. Although initial discoveries of the Toll-NFκB signaling pathway in development and immunity was pioneered in Drosophila, limited information is available for its role in cancer progression. Using a well-studied cooperative RasV12 -driven epithelial-derived tumour model, we here describe functions of Toll-NF-κB signaling in malignant RasV12, scrib- tumors. The extracellular Toll pathway components ModSP and PGRP-SA and intracellular signaling Kinase, Pelle/IRAK, are rate-limiting for tumor growth. The Toll pathway NFκB protein Dorsal, as well as cactus/IκB show elevated expression in tumors with highest expression in invasive cell populations. Oncogenic RasV12, and not loss of scribble, confers increased expression and heterogenous distribution of two Dorsal isoforms, DorsalA and DorsalB in different tumour cell populations. Mechanistic analyses demonstrates that Dorsal drives growth and malignancy by suppressing differentiation, counteracting apoptosis and promoting invasion of RasV12, scrib- tumors genetically dependent on twist and snail.


Introduction:
Inflammation, defined as the body's response to harmful external inputs, is a hallmark of cancer [1].This response is mediated by four major inflammatory pathways: the MAPK, PI3K-AKT, JAK-STAT and NF-kB (canonical and non-canonical) signaling pathways which integrate intra-and extracellular alarms and trigger cellular responses [2].NF-kB activation has been observed in many cancers [3,4] and correlates with a poor prognosis in breast and non-small cell lung cancer [5,6].NF-kB functions appear cancer-relevant, as a significant body of in vitro studies reports its involvement in cellular processes such as invasion [7][8][9], proliferation [10,11], stemness [12], survival [10][11][12] and chemoresistance [12][13][14][15].Similarly, a limited number of human cancer xenograft studies demonstrated a positive effect of NF-kB on cancer proliferation [11], survival [11] and metastatic behavior [8,9].Despite several attempts, pharmaceutical targeting of NF-kB in cancer has proven ineffective due to toxicity resulting from its essential role in regulating cellular and innate immunity at the organismal level [16,17].Therefore, understanding both the upstream and downstream mechanisms governing NF-kB activation and functions in cancer, as well as deciphering whether such mechanisms are specific to cancer cells, is crucial.This understanding may facilitate the design of NF-kB -targeting cancer therapies that spare healthy cells.In vivo studies would be informative for achieving this goal.
Drosophila melanogaster is increasingly being used to study both cancer cell-autonomous mechanisms as well as the biology of complex local and systemic paraneoplastic effects instigated by tumor growth [18].Among the Drosophila melanogaster cancer models, the Ras V12 , scrib -/-carcinoma model, arising from the concomitant expression of the constitute activated oncogenic form of the KRAS/NRAS orthologue, Ras85D (here referred to as Ras V12 ), and the loss of cell polarity through scribble loss-of-function, has been most extensively studied and displays a remarkable array of hallmarks similar to human cancer [19] [20].In Ras V12 , scrib -/-tumors, generated in the larval eye-antennal disc epithelium (EAD), the evolutionarily conserved Ras-MAPK, PI3K-AKT and JAK-STAT inflammatory pathways all constitute major drivers [19,20].Surprisingly, at the organismal level, Ras V12 , scrib -/-tumour-bearing flies also display both microenvironmental and systemic alterations that mimic human cancer, such as wasting of adipose and muscle tissues, gut atrophy, microenvironmental extracellular matrix remodeling and immune cell infiltration [21][22][23][24].Although the discovery and functions of the Toll-NF-kB and IMD-NF-kB pathways were pioneered in flies and found to control developmental patterning and infection response of the innate immune system [25][26][27], less information is available for its possible roles in carcinogenesis.
Here, we describe the aberrant expression and function of a NF-kB transcription factors in the Ras V12 ; scrib -tumour model.We demonstrate that in Ras V12 ; scrib IR tumours, the NF-kB transcription factor Dorsal (Dl) plays a pro-tumorigenic role by opposing differentiation and promoting survival and invasion.In this context, Dorsal can amplify JNK signaling activation, a known driver of invasion in this model.We also found that the known Dorsal targets snail and twist contribute to tumour growth.Interestingly, we report the expression of two different splicing variants of Dl, DorsalA (DlA) and DorsalB (DlB), in non-overlapping tumour cell populations.

The Toll pathway promotes tumour growth in an autonomous manner
To identify potential tumour-driving mechanisms, we surveyed upregulated genes in Ras V12 ; scrib -/-malignant tumours by RNA Sequencing (Fig 1A).We identified 364 genes with elevated expression levels.These include 14 previously identified genes with defined functions in driving Drosophila tumor growth or paraneoplastic effects, like organ wasting and diuretic dysfunction (Table 1).Notably, among the upregulated genes -PGRP-SA, GNBP2, Easter, Spz3, and Spz5 -encode extracellular components of the Toll pathway (Fig 1A).The Toll pathway, one of the two established NF-kB signaling pathways in Drosophila, is differentially activated during development and in response to fungal or bacterial infection.PGRP-SA (Peptidoglycan Recognition Protein SA) is a secreted protein that binds to bacteria-derived peptidoglycans upon bacterial infection, forming a trimeric complex with a member of the GNBP family (Gram-Negative Binding Protein) and the serine protease ModSP (Modular Serine Protease).This association initiates a multi-step proteolytic cascade, eventually cleaving the ligand Spz (Spaetzle).

Subsequently Spz binds to the Toll receptor, activating the Toll pathway intracellularly (Fig 1B).
To date, six Spz ligands and nine Toll receptors have been identified and the functions and associations of the different ligands and receptors is not yet fully understood [27].Intracellularly, Toll receptor activation triggers the activation and formation of a complex composed of the adaptor proteins Myd88 and Tube as well as the serine-threonine protein kinase Pelle.Pelle subsequently phosphorylates the inhibitor Cactus, the homolog of mammalian IkappaB, which is targeted to proteasome degradation.As Cactus sequesters the NF-kB Dorsal or Dif into the cytoplasm, its degradation triggers the translocation of the NF-kBs into the nucleus and the transcription of their target genes (Fig 1B).
We decided to first investigate whether PGRP-SA is involved in Ras V12 ; scrib -/- tumorigenesis.To do so, we genetically induced the formation of Ras These results collectively suggest that Ras V12 ; scrib -/-tumours produce extracellular Toll-activating components that contribute to a Toll pathway pro-tumorigenic signal activity in an autonomous manner.

Dorsal is ectopically expressed in oncogenic Ras V12 -driven tumours
Toll pathway activation leads to the release and nuclear translocation of the NF-kB transcription factors Dl and/or Dif, leading to transcription of target genes.While most studies of Dl and Dif in developmental patterning and infection have focused on their main splicing variants (DlA and DifA), both NF-kB transcription factors have been reported to also have alternatively spliced B isoforms (DlB and DifB) [28,29].For instance, both DlA (678 amino acids) and DlB (994 amino acids) contain a Relish-homology domain, composed of a DNA binding domain and a dimerization domain (for homo/heterodimerization and Cactus binding), as well as a transactivation domain (for transcription activation) [28].However, the transactivation domain of DlA is much shorter than Dorsal-B.Noteworthy, only DlA possesses an NLS sequence.Therefore, DlB cannot translocate to the nucleus by itself but is thought to be able to bind to DlA for translocating to the nucleus of fat body cells upon infection [28].Contrary to DlA, DlB is not expressed and involved in embryonic dorso-ventral patterning.It is rather specifically expressed in the subsynaptic reticulum of neuromuscular junctions, where it stabilizes Cactus.Similarly, Dif-B is found in the mushroom body of the CNS, where it also stabilizes Cactus [29].Although these observations support the idea that B isoforms play a specific role in nervous system tissues, their precise functions remain to be elucidated [30]. .We identified a DlA high cell population that had four-fold higher DlA protein levels compared to the DlA low remaining cell population.Importantly, the DlA mean intensity in the DlA low cell population was consistently higher than in the surrounding wt epithelial cells of the EAD (Fig 2C).These observations suggest that although we observed a peak of DlA expression at the posterior part of the tumour, there is a general elevation of the basal level of DlA throughout the entire tumour.In contrast, we did not observe a general elevation of DlB throughout the tumour when compared to wt cells of the TME, but we did observe a local increase in the posterior part of the tumour of a similar magnitude to DlA (Fig 2B and 2D).Tumour cells located in the posterior area of the tumour are closest to the CNS, which is the first organ these tumours invade [31][32][33].Intriguingly, tumour cells that have initiated invasion into the optic lobes (OL) of the CNS displayed high levels of DlA or DlB (S1B Finally, a strong enrichment of the repressor Cactus, the ortholog of mammalian IkappaB (IKKB), was observed in DlA high tumour cells but not in DlB high tumour cells, as assessed with an endogenous GFP fusion protein reporter, Cactus::GFP (S1C Fig).This is particularly interesting as cactus is a positive transcriptional target of the Toll pathway [34].Cactus upregulation constitutes a negative feedback loop which could cap Toll pathway activation, a mechanism that has also been described in mammalian NF-kB signalling pathways [35,36].In conclusion, simultaneous elevation of DlA, DlB and Cactus expression suggests that in the posterior part of Ras V12 ; scrib IR tumours, Toll pathway activation is significantly higher than in the rest of the tumour.

Dorsal aberrant accumulation is genetically and spatiotemporally-regulated
We investigated what could be the upstream mechanism leading to aberrant Dl accumulation.Initially, we examined Dl isoforms levels in various genetic backgrounds: control wild type clones, scrib IR tumours, Ras V12 tumours and Ras V12 ; scrib IR tumours.We observed aberrant expression of DlA as well as DlB exclusively in Ras V12 and Ras V12 ; scrib IR tumours (Fig 2A), suggesting that Ras V12 instructs tumour cells competent to produce these isoforms at abnormal levels.Furthermore, we followed DlA levels over time in Ras V12 ; scrib IR tumours, using an endogenous Dorsal genetic reporter, that acts as a Dorsal-A specific reporter in this context Together, these observations demonstrate that the aberrant expression NF-kB/Dl in Ras V12 ; scrib IR tumours is both genetically and locally controlled.General elevated levels of DlA in the whole tumor population suggests a transcriptional response downstream of oncogenic Ras V12 signaling.On top of this Ras V12 -dependent genetic control, a temporal and or local control triggers elevated levels of DlA, DlB and Cactus specifically in the posterior part of the tumour in the prospective eye region, as well as in invading tumour cells to the brain.

Dorsal represses Ras V12 scrib IR tumour differentiation and death
To investigate the primary functions of Dl in Ras V12 ; scrib IR tumours, we conducted knockdown experiments using two independent RNAis that efficiently target both Dorsal isoforms (S3A and S3B tumour growth results from a combination of cell growth, cell proliferation, inhibition of the differentiation of epithelial tumour cells into Elav-positive photoneurons and survival [31,37].To delineate the potential functions of Dl in Ras V12 ; scrib IR tumours, we assessed alterations in proliferation, differentiation and apoptosis.Quantification of mitotic cells suggests that Dorsal IR Ras V12 ; scrib IR tumours display similar proliferation rates to control tumours, but both their differentiation and apoptosis rates were significantly higher than in control tumours (Fig4B, 4C and 4D).Although, a few neurons form in Ras V12 ; scrib IR tumours, they appeared spaced apart and did not form the full complement of eight neurons per ommatidia.However, upon Dorsal knockdown, Elav-positive cells increased in number and density, suggesting a partial rescue of differentiation (Fig4C).The fact that we frequently observed cell death coinciding with areas where Elav-positive neurons were present (Fig 3D ) suggests that cell death might be a consequence of aberrant tumour cell differentiation.Importantly, Pelle IR Ras V12 ; scrib -/-tumours also exhibited increased apoptosis and differentiation rates (S3C Fig).This emphasizes the need for the Toll signalling pathway, and not an alternative way to activate Dorsal/NFkB activity, to promote tumour growth.Thus, the Toll-Dl signaling pathway promotes tumour growth primarily by preventing differentiation and promoting survival.

Conclusion/Discussion
Here, we establish that autonomous activation of the inflammatory Toll pathway plays a pivotal role in supporting the growth of Ras V12 ; scrib IR tumours.The NF-kB transcription factors Dorsal, and key effectors of the Toll pathway, exhibit aberrant expression in this model.In this context, Dorsal activity facilitates tumour growth by preventing differentiation and promoting survival without measurable impact on proliferation.Additionally, Dorsal promotes invasion genetically dependent on the EMT factors Snail and Twist.
An exciting feature in this tumor model is tumor heterogeneity in both NF-kB protein composition (DlA and DlB) and Toll pathway activity (Dl high cactus high vs Dl low , Cactus low ), as judged by classical target gene expression (cactus and dorsal) and stimulation of invasion.Dl is unevenly accumulated within the population of transformed cells in a spatial and regional manner and on at least two levels: there is a low-level DlA accumulation within the entire region of Ras V12 expression and a heightened level in the posterior region situated in the differentiating prospective eye region that is physically attached to the brain through the stalk.In this region a notable accumulation of both splicing variants of Dl, DlA and DlB, is seen within distinct tumour cell populations.
While NF-kB protein heterogeneity within the same tumour has not been studied yet in solid mammal tumours, the NF-kB fingerprinting and their subsequent homo-/heterodimerization within diffuse large B-cell lymphoma hold predictive values for the tumour cell response to microenvironmental activating cues [47].With our study, this finding, suggest that NF-kB composition heterogeneity might also occur in solid mammal cancers and modulate their response to microenvironmental inflammation.Moreover, heterogeneity in the level/activity of a single NF-kB mirrors observations in mammalian cancers.In pancreatic cancer, where oncogenic KRAS is the main driver, tumor heterogeneity is a central hallmark and frequently associated with variable NF-kB signaling.A central question is thus, what causes some cells to respond with inflammation whereas others do not?Possibilities include non-autonomous external input from nearby cells of the microenvironment or circulation versus intrinsic mechanisms.In vitro studies of KRAS G12V -transfected immortalized human pancreatic epithelial derived cells (HPDE) have demonstrated that variable levels of oncogenic KRAS can produce different outcomes.High KRAS G12V expression can direct NF-kB signaling and EMT phenotypes driven in part by feed forward loop through IL6 signaling, whereas lower level of KRAS activity retains epithelial character with lower level of NF-kB signaling dependent on IL1 [48].
In the Ras V12 , scrib -/-model, a "flip out" cassette turns on Ras V12 expression under control of the actin promoter in all cells with the same genetic dose, and by first principle should signal evenly.Yet, Ras V12 -driven NF-kB expression is temporally and spatially controlled in a stereotypic pattern.One possibiity is that this region is prepatterned to respond to increased levels of Ras V12 with a higher response.Indeed, in this region, we observe a higher level of pMAPK and where tumour region neurogenesis is incomplete as we do not observe fully formed ommatidia emphasizing the inhibitory effect of Dorsal on differentiation.In the prospective eye region, behind the morphogenetic frurrow, founder R8 neurons recruit further neurons through Receptor Tyrosine Kinase-MAPK signaling that potentially could add up with Ras V12 -MAPK signaling to reach a higher signaling levels [49].Notably, previous studies have identified PGRP-SA, Cactus, and Dorsal as positive targets of the Toll pathway [34].Furthermore, it has been shown that the CNS serves as a significant source of Spz ligands that regulates Ras V12 ; lgl -/-tumour migration tropism through Toll6 [33].Hence, it is tempting to think that upon the onset of neurogenesis, tumour cells suddenly gain access to elevated levels of Spz concentration from the brain, significantly enhancing local Toll activation and subsequent PGRP-SA and Dorsal expression.This hypothesis may explain the spatial restriction of aberrant Dorsal expression to the posterior part of the tumor.

Extrinsic regulation of
We have yet to assess whether DlA and DlB exert different functions within Ras V12 ; scrib IR tumours, as our knockdowns and overexpression transgenes targeted both isoforms simultaneously.Consistent with previous reports, DlB expression in the tumor context is restricted to cell types from the nervous system.Colocalization of DlB with pJNK in neurons that seem prone to die, allows us to speculate if both isoforms have distinct functions, that DlB may promote migration through JNK, while DlA may favor stemness and survival.Additionally, we cannot yet rule out the involvement of other NF-kBs, such as Dif and Relish in Ras V12 ; scrib IR tumours.
Using splicing variants may be a mechanism to diversify the range of NF-kB actions in various contexts in Drosophila, as it possesses only three NF-kBs (Dorsal, Dif, and Relish).In In conclusion, the findings herein extends the understanding of how innate immune signaling is engaged by and cooperates with oncogenic Ras, and opens up for more detailed mechanistic studies in a genetically tractable preclincial tumor model with clear paralells to human cancer

Drosophila lines
The EyMARCM system is known to generate flip-out clones in the adjacent optic lobes of the CNS in addition to the eye discs [50].For assessing tumour cell migration from true carcinoma cells arising from the eye disc epithelium, we therefore used another genetic system designed in our lab, called EyaHost, that generates flip-out tumours exclusively in the EAD but not in the CNS (to be described in detail elsewhere).
Unless otherwise specified, crosses were conducted at 25°C by default; however, for enhanced knockdown efficiency, most knockdown experiments were carried out at 29°C, as detailed in the figures and legends.

Imaging
Samples were imaged on Zeiss LSM880 and Zeiss LSM980 confocal microscopes as well as on Nikon HCI SoRa spinning disc microscope.They were further processed on FIJI.

Tumour Volume Analysis
Z-stacks were acquired with a 3µm step for early-stage tumors (Day 6) and a 4µm step for latestage tumors (after Day 6).Tumor volume was estimated using a FIJI Macro [52], which automatically performs threshold-based 3D reconstruction of the GFP+ signal, followed by volume measurements (3D Object counter plugin).Before 3D reconstruction, the FIJI Macro applies two consecutive Gaussian blur steps (sigma=2) to reduce noise.

Image processing
• Fluorescence intensity measurements and heatmaps Fluorescence intensity measurements were performed using a FIJI Macro, which subtracts background intensity to the signal of interest inside or outside the tumour, using a tumour mask generated from intensity thresholding.When comparing values across different conditions, they were normalized to the mean value of the control condition, ensuring the control mean is set at 1.In Fig 2B and Fig 2C, the tumours were split in two compartments "high" and "low", according to arbitrarily set intensity thresholds (250 for DlA, 10000 for DlB).Normalized mean intensities were then automatically quantified within the different compartments: tumourHigh, tumourLow and TME.DlA and DlB intensity heatmaps were generated after background subtraction, using the Fire LUT in FIJI.

• Colocalization analysis
Colocalization analyses were performed using a FIJI Macro, which automatically computed the Pearson's Correlation Coefficient (PCC) between two channels of a confocal image using the "coloc2" plugin.The PCC, ranging from -1 to 1, denotes the degree of colocalization between signals.A higher absolute value indicates stronger colocalization, with the sign indicating correlation direction ("-" for anti-correlation, "+" for correlation).

• Proliferation, differentiation and apoptosis ratio
The proliferation, differentiation and apoptosis ratio are defined as the area of the meaningful marker within the tumour (pH3, Elav and cDcp respectively) divided by the area of the tumour.These ratios were measured with a FIJI Macro which measures both the tumour area and the tumour marker area using an automatic-thresholding.For

Statistical analyses
At least, two biological replicates were performed for each experiment.According to the normality and size of the datasets, appropriate statistical tests were applied using GraphPad  (A) A transcriptomics analysis comparing the profile of Ras V12 ; scrib -/-tumour cells to GMR>GFP control cells in the eye-antennal disc (EAD) identified 364 genes that are upregulated in the tumours.Among them, PGRP-SA, GNBP2, easter, spz3 and spz5 are genes taking part in the extracellular machinery leading to Toll pathway activation.(B) Simplified version of Toll pathway activation upon bacterial infection.PGRP-SA (Peptidoglycan recognition protein SA) is a carboxypeptidase which recognizes and binds to peptidoglycans upon bacterial infection.Follows the formation of a trimeric complex together with a member of the GNBP family (Gram Negative Binding Protein) and the serine protease (ModSP).This complex initiates a multi-step proteolytic cascade that ultimately leads to the cleavage and activation of the ligand Spz.After the binding of Spz to the Toll receptor, the pathway is activated intracellularly through the activation and formation of a complex composed of the adaptor proteins Myd88 and Tube as well as the serinethreonine protein kinase Pelle.Pelle subsequently phosphorylates the inhibitor cactus, the homolog of mammalian IkappaB, which is targeted to proteasome degradation.As Cactus sequesters the NF-kB Dorsal into the cytoplasm, its degradation triggers the translocation of Dorsal into the nucleus and the transcription of its target genes.(C) Cartoon of the genetic setting used in this study for manipulating Ras V12 ; scrib -/-tumours.GFP-labelled tumours are generated from randomly-selected single epithelial cells from the EAD.They grow and fuse to form large tumours that start invading the central nervous system (CNS) through the Optic Lobe (OL) first and the Ventral Nerve Cord (VNC) later.Tumour cells also sometimes reach the leg discs (LD).(D) Representative confocal pictures and quantifications of the mean tumour volumes of GFPlabelled control tumours (n=21, m=3,71x10 7 µm 3 , SD=±0,47x10 7 µm 3 ) and PGRP-SA Ras V12 ; scrib -/-tumours (n=17, m=2,11x10 7 µm 3 , SD=±0,19x10 7 µm 3 ) at day8 after egg laying, statistical significance was determined with an unpaired T-test with Welch's correction.(E) Representative confocal pictures and quantifications of the mean tumour volumes of GFP-labelled control (n=20, m=8,46x10 7 µm 3 , SD=±1,77,53x10 7 µm 3 ) and ModSP IR Ras V12 ; scrib -/-Dcr2 tumours (n=30, m=5,07x10 7 µm 3 , SD=±1,00x10 7 µm 3 ) at Day8 (29°C), statistical significance was determined with an unpaired T-test with Welch's correction.(F) Representative confocal pictures and quantifications of the mean tumour volumes of GFP-labelled control tumours (n=10, m=2,64x10 7 µm 3 , SD=±0,76x10 7 µm 3 ) and pelle IR Ras V12 ; scrib -/-tumours (n=10, m=0,78x10 7 µm 3 , SD=±0,22x10 7 µm 3 ) at Day6 (29°C), statistical significance was determined with a Mann-Whitney test.Scale bars=50µm.(A) Immunostainings against Dorsal-A (DlA in green) and Dorsal-B (DlB in magenta) in BFPlabelled wt clones (n=20 EADs), scrib IR tumours (n=23 EADs), Ras V12 tumours (n=20 EADs) and Ras V12 ; scrib IR tumours (n=24 EADs) at day 6.DlA and DlB accumulation is observed only in Ras V12 -driven tumours and appears mainly mutually exclusive.(B) Immunostaining against DlA (green) and DlB (magenta) in BFP-labelled Ras V12 ; scrib IR tumours (white) (n=24 EADs) and close-ups showing that DlA and DlB are not expressed in the same subset of cells.Colocalization analysis assessed with the Pearson's correlation coefficient reveals a poor colocalization between DlA and DlB in the tumours (n=24, m=0,11, SD=±0,08).(C) Heat map of DlA intensity within a representative BFP-labelled Ras V12 ; scrib IR tumour-bearing EAD at day 6.The tumour is outlined with the straight white line.An arbitrary intensity threshold set at 250 (gray line on the graph) allows the clear segregation of a DlAhigh ("tumour High") and a DlAlow ("tumour Low") cell population within the tumour.Quantifications of the mean DlA intensity of the "tumour High" (n=23, m=458,7, SD=±28,4), "tumour Low" (n=23, m=135,2, SD=±7,0) and wt Tumour microenvironment (TME) (n=23, m=82,2, SD=±6,7) demonstrates a general elevation of DlA levels within the tumour compared to wt cells.Statistical significance was determined with a Tukey's multiple comparison test.(D) Heat map of DlB intensity within a representative RasV12; scribIR tumour-bearing EAD at day 6.The tumour is outlined with the straight white line.An arbitrary intensity threshold set at 10000 (gray line on the graph) allows the clear segregation of a DlBhigh ("tumour High") and a DlBlow ("tumour Low") cell population within the tumour.Quantifications of the mean DlB intensity of the "tumour High" (n=24, m=20331,9, SD=±3108,7), "tumour Low" (n=24, m=3955,8, SD=±584,5) and wt Tumour microenvironment (TME) (n=23, m=4660,9, SD=±1187,3).Normalization was done by subtracting the background intensity.Statistical significance was determined with a Dunn's multiple comparison test.Scale bars=20µm.

Fig 7. Working Model
In this working model, we propose that the protein level of the NF-kB Dorsal regulate different aspect of tumorigenesis.Invasion: In the most posterior part of the tumour, high levels of Dorsal promote invasion of the tumour cells in the adjacent organs through elevation of JNK signaling and the EMT transcription factor Snail and Twist that would promote cytoskeleton remodelling.Tumour growth: In the rest of the tumour where more intermediate levels of Dorsal are observed, Dorsal would stimulate tumour growth by repressing differentiation and apoptosis.Regulation: At least two factors seem two trigger aberrant level of Dorsal within the tumours.Genetically, Ras V12 make tumour cells competent to accumulate higher levels of Dorsal throughout the entire tumour.Locally, a spatiotemporal control is exerted at the posterior part of the tumour where neurons are formed, to reach very high levels of Dorsal.A possibility is an increase local Toll pathway activation triggered by high Spz concentration from the brain which will promote Dorsal expression.There, we observe alternative splicing of DlA versus DlB which segregate in different tumour subpopulations.We hypothesize that the heterogeneity in Dorsal isoforms and/or levels promote a balance between tumour growth and invasion favoring tumorigenesis.(A) Representative confocal pictures and quantification of the mean DlA intensity of GFP-labelled cherry IR Ras V12 ; scrib IR (n=6, m=1,00, SD=±0,14) and dorsal IR-1 Ras V12 ; scrib IR tumours (n=6, m=0,54, SD=±0,03) as well as Cherry IR Ras V12 ; scrib IR Dcr2 control tumours (n=5, m=1,00, SD=±0,14) and dorsal IR-2 Ras V12 ; scrib IR Dcr2 tumours (n=6, m=0,32, SD=±0,05) at day 6 (29°C).Intensity normalization was done by subtracting the background intensity on each picture and finally dividing each data point by the mean intensity value for the control tumours for each replicate.Statistical significance was determined with a Mann-Whitney test.Scale bars=20µm.(B) Representative confocal pictures and quantification of the mean DlB intensity of GFP-labelled cherry IR Ras V12 ; scrib IR (n=7, m=1,00, SD=±0.19) and dorsal IR-1 Ras V12 ; scrib IR tumours (n=9, m=0.40,SD=±0.12) as well as Cherry IR Ras V12 ; scrib IR Dcr2 control tumours (n=5, m=1.00,SD=±0.31) and dorsal IR-2 Ras V12 ; scrib IR Dcr2 tumours (n=6, m=0.54,SD=±0.15) at day 6 (29°C).Intensity normalization was done by subtracting the background intensity on each picture and finally dividing each data point by the mean intensity value for the control tumours for each replicate.Statistical significance was determined with a Mann-Whitney test.Scale bars=20µm.(C) Z-projection of representative confocal pictures of GFP-labelled RFP IR Ras V12 ; scrib IR control tumours and pelle IR Ras V12 ; scrib IR tumours at day 6 (29°C), stained for Elav (green) and cDcp1 (magenta).The tumour is outlined with a straight white line.Quantification of the apoptosis ratio of RFP IR Ras V12 ; scrib IR control tumours (n=16, m=4,97%, SD=±2,23%) and pelle IR Ras V12 ; scrib IR tumours (n=15, m=17,84%, SD=±4,16%) at day 6 (29°C).Here, the apoptosis ratio is defined as the ratio between the cDcp1 area within the tumour and the area of the tumour itself.Statistical significance was determined with an unpaired T-test with Welch's correction.Quantification of the differentiation ratio of RFP IR Ras V12 ; scrib IR control tumours (n=5, m=18,04%, SD=±5,99%) and pelle IR Ras V12 ; scrib IR tumours (n=8, m=44,14%, SD=±8,83%) at day 6 (29°C).Here, the differentiation ratio is defined as the ratio between the Elav area within the tumour and the area of the tumour itself.Statistical significance was determined with a Mann-Whitney test.Scale bars=50µm.

S4 Fig. Dorsal is sufficient to induce dysplasia and delamination but not overgrowth and invasion.
(A) Representative confocal pictures and quantification of the mean clone volume of GFP-labelled wt clones (n=20, m=5,59x10 5 µm 3 , SD=±0,69x10 5 µm 3 ) and dorsal OE clones (n=30, m=4,87x10 5 µm 3 , SD=±0,66x10 5 µm 3 ) at day 6, stained for DlA (magenta) and Hoechst (blue).Displasic clones are observed in 100% of the dorsal OE clones-bearing EADs.GFP-labelled cells are never found in the VNC as indicated with the VNC score values.Scale bars=50µm.(B) Reconstruction of an orthogonal view of wt clones and dorsal OE clones representing the EAD tissue (magenta) and the GFP-labelled clones (green).GFP-labelled cells extend from the apical to the basal side of the pseudostratified epithelium of the EAD of wt clones whereas some Dorsal OE clones spread basally (yellow arrow).Select RNASeq read counts and fold changes between Ras V12 , scrib -/-GFP+ tumor cells and control GmrGal4,UAS-GFP, GFP+ cells.Classic: Previously identified genes known to have a role in Drosophila tumour growth, invasion, or Tumour-Host interactions.Upregulated Toll pathway genes are indicated in grey.
V12 ; scrib -/-tumours from randomly induced single cells in the Eye-Antennal Disc (EAD) of the Drosophila larva.This way, we can knockdown or overexpress any gene of interest within the tumour specifically without affecting expression in the rest of the body (Fig 1C).To improve RNA interference (IR) expression levels and processing respectively, knockdown experiments were sometimes performed at 29°C with the additional expression of Dicer-2, as indicated in the legends.Knock-down of PGRP-SA (PGRP-SA IR ) significantly reduced Ras V12 ; scrib -/-tumour growth compared to control tumours (Fig 1D).Similarly, knock-down of ModSP (ModSP IR ) significantly decreased tumour size (Fig 1E), suggesting that the extracellular components of the Toll pathway surprisingly play a pro-tumorigenic role in this context.To address whether Toll signaling is required within tumour cells or in another microenvironmental compartment, we performed knockdown of the intracellular kinase Pelle/IRAK (pelle IR ), a hub in Toll signal transduction.pelle IR Ras V12 ; scrib -/-tumours exhibited a drastic size reduction compared to control tumours (Fig 1F).
V12 ; scrib IR tumours through immunostainings against DlA and DlB (Fig 2A, 2B, 2C, 2D and S1A, S1B Fig).Both Dorsal isoforms were highly expressed in tumors in the posterior part of the disc that attaches to the brain through the prospective optic stalk (Fig 2A and 2B), while neither DlA or DlB proteins were observed in control clones (Fig 2A).DlA and DlB were almost entirely mutually exclusive, in different cell populations (Fig2A, 2B and S1B Fig).Within Ras V12 ; scrib IR tumours, cells can be easily separated according to their levels of DlA mean intensity (Fig 2C)

Fig), suggestive
Fig), suggestive of a potential role of Toll-NFkB signaling in tumor cell migration.

(
Fig 3A and S2A and S2B Fig).The onset of DlA expression correlated with the appearance of Elav-positive neurons in the prospective eye region of the tumours around Day5 (Fig 3A).Interestingly, DlA accumulate in cells surrounding the neurons but was rarely detected in the neurons themselves (Fig 3A yellow arrows and Fig 3B).Concurrently, DlB expression was exclusive to the tumour neurons (Fig 3C).Within the tumours, apoptosis was predominantly observed in the differentiating areas where we observed dying neurons (Fig 3D, yellow arrows).Therefore, we hypothesized that the local aberrant Dl expression might be triggered by cell deathderived signals.However, efficient blocking of apoptosis in Ras V12 ; scrib IR tumours through Diap1 overexpression (Diap1 OE ) did not affect increased DlA and DlB levels (Fig 3E and 3F).
Fig).Dorsal IR , Ras V12 ; scrib IR tumours exhibited a significant reduction in size compared to Cherry IR Ras V12 ; scrib IR control tumours (Fig 4A), mimicking the phenotypes observed upon PGRP-SA, ModSP and pelle knockdowns (Fig1C, 1D and 1E).Ras V12 ; scrib IR conducted colocalization analyses, assessed by the Pearson's Correlation Coefficient (PCC), between the Dorsal isoforms and JNK activity in the tumours, assessed either through activated JNK (phosphor-JNK, pJNK) or a JNK activity reporter, TRE-eGFP.Overall, we found a low colocalization between DlA and TRE-eGFP (Fig 5A) while we observed a good PCC between DlB and pJNK within the tumours (Fig 5B).These observations align with the finding that tumor cells invading the CNS often have high levels of either DlA or DlB.To test whether Dorsal may function upstream of JNK signaling in Ras V12 ; scrib IR tumours, we assessed the levels of the JNK target MMP1 following Dorsal OE and ModSP OE overexpression which enhances Toll signaling and subsequently Dorsal activity.MMP1 mean intensity was significantly higher in tumors with elevated Dorsal or ModSP expression.compared to control tumours.Therefore, Toll-Dl signaling can augment expression of the JNK transcriptional target, mmp1 in Ras V12 ; scrib IR tumours (Fig 5C).As both DlA, DlB and Cactus are highly expressed in brain lobe proximal and invading cells and Dl amplifies JNK signalling in the tumours, we investigated the effect of Dorsal knock-down on tumour cell migration.We scored for three parameters reflecting the mobility of tumour cells at late stages of tumour development (Day 12): 1/ the VNC invasion score, which assigns a higher score to tumors with greater migration toward the tip of the ventral nerve cord (VNC) of the CNS, 2/the frequency at which we observe tumour cells within the organs adjacent to the VNC called leg discs and 3/ the fusion of tumours generated in the two distinct EAD, indicating the ease with which cells move from one EAD to the other and EADs and come into contact with each other (Fig 6A).To mitigate the influence of tumour cell number on migration assessment, we compared control and Dorsal IR tumours of similar size.Upon Dorsal knock-down, at late stages, the three parameters showed a drastic decrease: the VNC score decreased threefold compared to control, reaching a low value similar to early-stage control tumours (Fig6B).The frequency of invasion into leg discs decreased sevenfold and tumors from neighboring EADs often failed to merge (Fig 6A).Similarly, Dorsal overexpression (Dorsal OE ) increased tumour cell mobility even at early stages (Day 6), while tumour size is signicantly smaller (Fig 6B), suggesting that 1/ the increase in invasion is not simply a secondary consequence of higher cell number and 2/ that the balance between invasion and tumour growth tilted toward invasion when tumour cells have high levels of Dorsal.Given the functions of Dorsal in the tumour context, we wondered whether Dorsal OE is sufficient to induce tumorigenesis.Although, Dorsal OE control clones were not bigger than wt clones, we systematically observed the formation of dysplasic GFP-labeled Dorsal OE clones within the EADs that appeared as large-rounded cell masses often located basally (S4A and S4B Fig, yellow arrow).We hypothetize that they arise from the inhibition of differentiation.Moreover, in this context, we never observed Dorsal OE cells in the VNC (VNC score in S4A Fig).Thus, Dorsal at high levels is sufficient to trigger dysplasia but not invasive hyperplastic tumours.The Epithelial-to-Mesenchymal (EMT) transcription factor Snail has been implicated in promoting invasion of Ras V12 ; lgl -/-tumours (a similar tumour model to Ras V12 ; scrib -/-tumours) through stimulation of JNK signaling and EMT-like cytoskeleton remodeling [42].Snail, along with Twist, another EMT transcription factor, are well-known targets of Dorsal that govern mesoderm invagination during developmental patterning in cells with the highest Dorsal activity [43-46].Interestingly, knockdowns of both Snail and Twist significantly decreased Ras V12 ; scrib IR tumour size (S4A and S4B Fig), mimicking Dorsal knockdown.At late stages, twistIR Ras V12 ; scrib IR tumour also displayed a lower VNC score compared to control tumours, suggesting impaired tumour cell migration (S4B Fig).We propose a model where the Dl high tumour cells activate Twist and Snail expression, provoking migration, in part through JNK signalling enhancement (Fig 7A).Taken together, these experiments suggest that Dl promotes tumor growth through repressing cell death and differentiation as well as promoting EMT like delamination and invasion dependent on Twist and Snail activity.
Toll signaling does have a role in the Ras V12 , scrib -/-tumor model for growth.The Toll-activating components implicated in tumour growth, PGRP-SA and ModSP, are part of a bacterial recognition alert system.The observed overexpression of PGRP-SA in Ras V12 ; scrib IR tumours may serve to locally increase the concentration of active cleaved Spz ligands, thereby stimulating Toll activation and promoting tumor growth.The source of activation governed by PGRP-Sa and ModSP remains unknown.As the posterior part of the disc experience higher levels of apoptosis, we considered whether apoptotic cell death and debris could active PGRP-SA/ModSP -driven Toll sigaling in this context through death-induced danger signals and bacterial component mimicry, but inhibition of death through DIAP1 expression did not alter NFkB levels.
contrast, mammalian NF-kBs are not reported to have splicing variants, although they possess five different NF-kBs.To our knowledge, the concept of intratumoral NF-kB heterogeneity remains relatively unexplored in mammalian cancer biology and the fly model offer an experimentally accessible model to investigate this question in vivo.Future research efforts will be crucial in elucidating the importance and mechanisms underlying intratumoral NF-kB heterogeneity during malignant transformation.The Ras V12 , scrib tumor model engages five known inflammatory pathways during transformation.Ras V12 directly drives MEK-ERK and PI3K-AKT signaling while engaging Toll-NFkB, whereas loss of cell polarity through scribbled, initiates TNFR-JNK MAPK stress signaling and as a result, an Upd/IL6-JAK-STAT autocrine signaling loop altogether favoring malignant transformation.
Fig 3A, 3B and 3C, measurements were conducted on single confocal pictures whereas, for S1 Fig, the apoptosis ratio was measured on the Z-projection of confocal Z-stacks.• Migration score Migration of the VNC and leg discs was manually scored on Z-projection (Max Intensity mode) of confocal Z-stacks.
Prism9 and are indicated in the figure legends.Graphs were also generated on GraphPad Prism9 and are presented as boxplots (10-90 percentile) or dot plots indicating the mean of the datasets.Statistical significance was categorized as follows: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, and n.s when not significant.

Fig 2 .
Fig 2. Two splicing variants of the NF-kB Dorsal are aberrantly expressed in Ras V12 -driven tumours.

S1Fig.
Localization of DlA, DlB and Cactus.(A)Representative confocal pictures of GFP-labelled Ras V12 ; scrib -/-tumours (n=10 EADs) at day 6, generated with the conventional EyMARCM genetic system and stained for DlA (magenta) and Hoechst (blue).Consistent with the data generated with our new EyaHost genetic system in Fig 2,DlA levels are aberrant in the tumours.The tumour is outlined with a straight white line.Scale bars=50µm.(B) Representative confocal pictures of GFP-labelled Ras V12 ; scrib IR tumours at day 6, generated with our novel EyaHost genetic system and stained for DlA (green), DlB (magenta) and Hoechst (blue).The tumour cells that already migrated inside the optic lobe (OL) of the CNS express high levels of DlA or DlB, suggesting their potential involvement in tumour cell migration.Scale bars=50µm.(C) Representative confocal pictures of BFP-labelled genomic-cactus::GFP, Ras V12 ; scrib IR tumours stained for DlB (magenta) at day 6 (upper line) (n=6 EADs) or stained for DlA (magenta) at day 8 (lower line) (n=6 EADs).The tumour is outlined with a straight white line.Scale bars=20µm.S2Fig.The endogenous Dl reporter used in this study is a DlA specific reporter.(A) Representative confocal pictures of BFP-labelled genomic-dorsal::GFP Ras V12 ; scrib IR tumours at day 6, stained for DlA (magenta) and DlB (blue).The tumour is outlined with the straight white line.Scale bars=50µm.(B) Colocalization analyses assessed with the Pearson's correlation coefficient reveal a good colocalization between DlA and Dorsal::GFP (n=24, m=0,44, SD=±0,06) and no colocalization between DlB and Dorsal::GFP in the tumours (n=24, m=0,09, SD=±0,07) suggesting that this reporter is a DlA specific fusion protein.Statistical significance was determined with an unpaired T-test with Welch's correction.S3 Fig. Dorsal knockdown efficiencies and pelle knockdown mimicks the function of Dorsal on differentiation and survival.

Table 1 .
Select list of genes previously reported to have a role in Drosophila tumor models alongside Toll pathway components