Tissue-autonomous immune response regulates stress signaling during hypertrophy

Postmitotic tissues are incapable of replacing damaged cells through proliferation, but need to rely on buffering mechanisms to prevent tissue disintegration. By constitutively activating the Ras/MAPK-pathway via RasV12-overexpression in the postmitotic salivary glands (SGs) of Drosophila larvae, we overrode the glands adaptability to growth signals and induced hypertrophy. The accompanied loss of tissue integrity, recognition by cellular immunity, and cell death are all buffered by blocking stress signaling through a genuine tissue-autonomous immune response. This novel, spatio-temporally tightly regulated mechanism relies on the inhibition of a feedback-loop in the JNK-pathway by the immune effector and antimicrobial peptide Drosomycin. While this interaction might allow growing SGs to cope with temporary stress, continuous Drosomycin expression in RasV12-glands favors unrestricted hypertrophy. These findings indicate the necessity to refine therapeutic approaches that stimulate immune responses by acknowledging their possible, detrimental effects in damaged or stressed tissues.


Introduction
Immune and stress responses have evolved to protect the organism from both exogenous and endogenous stimuli (Eming, 2014;Adamo, 2017;Rankin and Artis, 2018). By sensing deviations from homeostasis and inducing compensatory mechanisms, immune and stress responses keep physiological parameters within tolerable limits (Hoffmann and Parsons, 1991;Vermeulen and Loeschcke, 2007).
Heat shock, radiation, starvation, toxic metabolites, hypoxia, or hyperoxia are all well-characterized stressors. They activate stress responses which can ultimately lead to the induction of programmed cell death (Lowe et al., 2004;Loboda et al., 2016). In Drosophila, the Keap1-Nrf2-, JNK-, and p38-signaling pathways are crucial for mounting these anti-stress-responses (Stronach and Perrimon, 1999;Fuse and Kobayashi, 2017). Immune responses on the other hand, like the Drosophila Toll and imd pathways, are typically activated by molecular structures exposed on the surfaces of pathogens (Medzhitov and Janeway, 2002;Kurata, 2004). These pathways coordinate the humoral and cellular immune system to eliminate intruding pathogens (Lemaitre and Hoffmann, 2007;Buchon et al., 2014).
Humoral immune responses in Drosophila are characterized by the production and secretion of large sets of effector molecules, most notably antimicrobial peptides (AMPs) like Drosomycin (Drs) (Imler and Bulet, 2005). AMPs not only target extrinsic threats in the form of intruding pathogens, but also react to intrinsic stimuli such as tumorigenic transformation with the possibility to induce apoptosis (Araki et al., 2019;Parvy et al., 2019). However, it remains poorly understood whether AMPs also have functions beyond promoting apoptosis when sensing and reacting to accumulating stress such as during wound healing and tumor formation.
Apart from their described individual roles, immune and stress pathways are proposed to be either induced successively or concomitantly dependent on the level of deviation from homeostasis (Chovatiya and Medzhitov, 2014;Ammeux et al., 2016). However, detailed characterization of wound healing and tumor models in Drosophila revealed a more complex picture (Park et al., 2004;Buchon et al., 2009;Meyer et al., 2014;Wu et al., 2015;Liu et al., 2015). Accordingly, immune and stress responses often neither occur separately, nor do they follow a simple linear cascade, but rather regulate each other via context-dependent mutual crosstalk (Wu et al., 2015;Liu et al., 2015;Fogarty et al., 2016;Pérez et al., 2017). One recurring motif throughout most of these models is the central role of the stress-responsive JNK-pathway and its frequent interaction with the Toll and imd immune pathways (Park et al., 2004;Rämet et al., 2002;Galko and Krasnow, 2004;Uhlirova et al., 2005;Igaki et al., 2006;Enomoto et al., 2015;Andersen et al., 2015). However, while JNK-signaling can function either in a tumor-promoting, anti-apoptotic or in a tumor-suppressive, pro-apoptotic manner depending on the context, Toll-and imd-signaling have only been shown to display a tumor-suppressing, pro-apoptotic role in Drosophila (Uhlirova et al., 2005;Igaki et al., 2006;Enomoto et al., 2015;Uhlirova and Bohmann, 2006;Cordero et al., 2010;Vidal, 2010).
These tumor-suppressive, pro-apoptotic functions of immune responses have been well characterized and attributed to the secretion of humoral factors or the recruitment of immune cells through the systemic immune system (Fogarty et al., 2016;Pérez et al., 2017;Babcock et al., 2008;Pastor-Pareja et al., 2008;Parisi et al., 2014;Hauling et al., 2014). In addition, during clonal cell competition in imaginal discs, Toll-and imd-signaling were implicated in the elimination of less fit cell clones by inducing apoptosis (Meyer et al., 2014). Importantly, the selective growth disadvantage of these less fit cells is thought to be a response to systemic infection (Germani et al., 2018) and it remains an open question whether genuine tissue-autonomous immune responses can contribute to adaptation of growth during wound healing and tumor formation.
The larval salivary gland (SG) of Drosophila is a powerful system to study adaptive growth control, because growth is not completely predetermined, but modulated by the nutritional status eLife digest Tissues and organs work hard to maintain balance in everything from taking up nutrients to controlling their growth. Ageing, wounding, sickness, and changes in the genetic code can all alter this balance, and cause the tissue or organ to lose some of its cells. Many tissues restore this loss by dividing their remaining cells to fill in the gaps. But some -like the salivary glands of fruit fly larvae -have lost this ability.
Tissues like these rely on being able to sense and counteract problems as they arise so as to not lose their balance in the first place. The immune system and stress responses are crucial for this process. They trigger steps to correct the problem and interact with each other to find a common decision about the fate of the affected tissue. To better understand how the immune system and stress response work together, Krautz, Khalili and Theopold genetically manipulated cells in the salivary gland of fruit fly larvae. These modifications switched on signals that stimulate cells to keep growing, causing the salivary gland's tissue to slowly lose its balance and trigger the stress and immune response.
The experiments showed that while the stress response instructed the cells in the gland to die, a peptide released by the immune system called Drosomycin blocked this response and prevented the tissue from collapsing. The cells in the part of the gland not producing this immune peptide were consequently killed by the stress response. When all the cells in the salivary gland were forced to produce Drosomycin, none of the cells died and the whole tissue survived. But it also allowed the cells in the gland to grow uncontrollably, like a tumor, threatening the health of the entire organism.
Mapping the interactions between immune and stress pathways could help to fine-tune treatments that can prevent tissue damage. Fruit flies share many genetic features and molecular pathways with humans. So, the next step towards these kinds of treatments would be to screen for similar mechanisms that block stress activation in damaged human tissues. But this research carries a warning: careless activation of the immune system to protect stressed tissues could lead to uncontrolled tissue growth, and might cause more harm than good. (Smith and Orr-Weaver, 1991;Britton and Edgar, 1998). In contrast to mitotically active tissues, growth in post-mitotic tissues like the larval SG is based on endoreplication and hypertrophy rather than on cell division (Edgar et al., 2014;Orr-Weaver, 2015). Modifying the underlying, tight growth regulation, for instance by continuous growth signaling via constitutively activated Ras/MAPK-signaling can easily lead to the accumulation of oxidative stress and DNA damage (Mason et al., 2004;Bartkova et al., 2005;Bartkova et al., 2006;Di Micco et al., 2006;Shim et al., 2013;Shim, 2015). However, the parameters defining the natural limit of growth adaptation and the buffering mechanisms in place to cope with prolonged or continuous stress remain poorly understood.
Here, we uncover a genuine tissue-autonomous immune response which directly regulates hypertrophic growth and adaptation to accumulating stress in larval SGs. By overexpressing a dominantactive form of the small GTPase Ras, Ras V12 , we induced hypertrophic growth. This activates a tissue-autonomous immune response which allows the hypertrophic gland to cope with the resulting stress through spatio-temporally regulated inhibition of the JNK-mediated stress response. We present evidence that tissue-autonomous expression of the AMP Drosomycin (Drs) is at the core of this inhibition: Drs directly interferes with JNK-signaling and inhibits JNK-dependent programmed cell death. This prevents recognition of the stressed tissue by the cellular immune response and allows survival and unrestricted growth of hypertrophic SGs.

Local immune reaction accompanies Ras V12 -dependent hypertrophy
In order to identify buffering mechanisms that compensate for continuous stress, we made use of our previously published hypertrophy model in the SGs of Drosophila larvae (Hauling et al., 2014). We expressed a dominant-active form of Ras, Ras V12 , across the entire secretory epithelium of SGs throughout larval development by using the Bx MS1096 enhancer trap (Figure 1-figure supplement 1A,C for 96/120 hr after egg deposition, AED). To further enhance Ras V12 -dependent hypertrophy, we combined Ras V12 -expression with RNAi-mediated knockdown of the cell polarity gene lethal (2) giant larvae (lgl; Figure 1-figure supplement 1C-H; Jacob et al., 1987;Strand et al., 1994). Their individual and cooperative role in tumor formation in mitotic tissues has been well characterized (Bilder et al., 2000;Pagliarini and Xu, 2003;Herranz et al., 2016). Cell-and tissue-morphology was assessed using Phalloidin staining ( Figure 1A At 96 hr AED, Ras V12 -expressing SG cells retained most of their normal morphology compared to w 1118 -control glands. However, their integrity and polarity were severely disrupted at 120 hr AED ( Figure 1A; Figure 1-figure supplement 1C). Nuclei of Ras V12 -glands showed a continuous increase in volume at 96 hr and 120 hr AED (1.33 fold compared to w 1118 controls at 96 hr AED; 5.66 fold at 120 hr AED; Figure 1C) and signs of nuclear disintegration at 120 hr AED implying the induction of programmed cell death (PCD; Figure 1B). Both loss of cell integrity and nuclear disintegration coincided temporarily ( Figure 1A-B) and were exacerbated upon coexpression of l(2)gl RNAi (Figure 1-figure supplement 1C-F). These results confirm our previous findings and demonstrate that continuous growth signaling in larval SGs leads to increased organ size accompanied with additional endocycles at 96 hr AED, both hallmarks of compensatory hypertrophy (Tamori and Deng, 2014). Furthermore, the additional, Ras V12 -induced endoreplications without obvious effect on tissue integrity imply an adaptability to excess growth signaling, whereas the subsequent collapse of nuclear integrity and cellular polarity at 120 hr AED delineate its limitations.
To characterize the mechanisms involved in the early phase of SG growth adaptation at 96 hr AED, we performed total RNA sequencing of complete Ras V12 -expressing and w 1118 -control SGs dissected at 96 hr AED prior to cellular and nuclear disintegration. The most significantly upregulated gene in Ras V12 -glands compared to their w 1118 -counterpart was Ras85D itself (q-value = 6.51Á10 À282 ), which validates the experimental set-up ( Figure 1D). The most differentially expressed gene in turn was the AMP Drs, which indicates the activation of a local immune response in Ras V12glands. To evaluate this further, we employed a GFP reporter for Drs and observed strong induction in Ras V12 -glands, but not in any other larval tissue ( Figure 1E; Figure 1-figure supplement 1B,G,I; Ferrandon et al., 1998). At 96 hr AED, the entire secretory epithelium of the SG expressed Drs with a strong tendency for increased induction in the proximal part (PP) closest to the duct (Figure 1-D Figure 1 continued on next page figure supplement 1B,G). At 120 hr AED Drs was almost exclusively expressed in the PP, an expression pattern that persisted until Ras V12 -expressing larvae belatedly puparate ( Figure 1E; Figure 1figure supplement 1B,G,I). To rule out artifacts derived from driver related unequal expression across the SG epithelium, we measured Bx MS1096 -driven RFP-expression across the longitudinal gland axis at 96 hr and 120 hr AED and normalized the acquired signal with driver-independent Phalloidin-fluorescence (Figure 1-figure supplement 1A; see 'Materials and methods'). The average Drs-reporter signal already peaks at 96 hr AED in the middle of the gland epithelium, while Bx MS1096 -driven expression increases further until 120 hr AED. Hence, while Bx MS1096 -driven expression of Ras V12 is crucial for activating Drs-expression, both are not qualitatively correlated (Figure 1figure supplement 1A-B). In order to assess whether hemocytes were recruited as part of a parallel, cellular immune response, we stained glands with an antibody against the pan-hemocyte-marker Hemese. While Ras V12 -glands were completely devoid of hemocytes at 96 hr AED, at 120 hr AED they were recruited to the gland surface. However, hemocyte attachment was restricted to the distal, non-Drs expressing part (DP), rendering Drs expression and hemocyte attachment across the gland epithelium mutually exclusive ( Figure 1E-F). Coexpression of l(2)gl RNAi elevated the level of recruited hemocytes at 120 hr AED and pre-empted this recruitment to the DP already at 96 hr AED ( Figure 1-figure supplement 1G-H). We next investigated whether the change in nuclear volume as a marker for growth adaptation follows a similar proximal-distal-divide as Drs-expression and hemocyte attachment (Figure 1-figure supplement 1E). Nuclei in the DP of the SG at 96 hr AED showed a moderate volume increase upon Ras V12 -expression compared to distal w 1118 -control nuclei. However, after 120 hr AED distal nuclei had undergone a drastic increase in nuclear volume (6.28 fold compared to distal w 1118 -control nuclei) while nuclei in the PP of the SG displayed only a moderate increase in size compared to w 1118 -control nuclei, that did not increase over time. This indicates that nuclei in the DP of Ras V12glands undergo more rounds of endoreplication than their proximal counterparts coinciding with the decline of Drs-expression and an increase in hemocyte attachment in this part. Moreover, this difference also explains the marginal reduction in Bx MS1096 -driven expression from DP to PP (Figure 1figure supplement 1A; PP shows 86% of DP-expression level).

Dorsal-dependent Drs expression is part of a genuine tissueautonomous immune response
As barrier epithelia, the lumen of the SG forms part of a continuum with the exterior, exposing them to extrinsic stimuli including nutritional cues and pathogens (Andrew et al., 2000). Since systemic infection can modulate tissue growth, we sought to clarify whether the observed local immune response in the gland epithelium fulfills the criteria of a genuine tissue-autonomous immune response or was rather embedded in a wider systemic immune response (Germani et al., 2018). Therefore, we eradicated the majority of putative systemic infections, food-derived signals and pathogens or bacterial contamination by raising larvae with Ras V12 -glands under sterile conditions, on minimal medium or by bleaching embryos (Figure 2-figure supplement 1A-A''; Shaukat et al., 2015;Kenmoku et al., 2017;Asri et al., 2019). None of these changes diminished the Drs expression, strongly indicating that Drs is indeed induced in a bona fide tissue-autonomous manner as a response to Ras V12 -dependent hypertrophic growth (Figure 2-figure supplement 1A-A''; Colombani et al., 2005;Mirth et al., 2014).
We further sought to identify the upstream factors controlling Drs expression. The homeobox-like transcription factor Caudal is necessary for Drs expression in the female reproductive organs and the adult SG (Ferrandon et al., 1998;Ryu et al., 2004;Han et al., 2004). However, RNAi-lines directed against Caudal or its canonical interaction partner, Drifter (dfr / vvl), did not reduce Drs expression   Figure 2 continued on next page in Ras V12 -glands indicating that the regulation of Drs expression as part of the tissue-autonomous immune response is clearly different from its counterpart in the adult SG ( Figure 2-figure supplement 2A; Junell et al., 2010).
In order to evaluate whether either Toll-or -as is the case for local infections -imd-signaling plays a role in Drs expression, we used the reproducible fluorescence pattern of the Drs-GFP reporter at 96 hr AED to assay RNAi-lines directed against canonical components of both pathways in Ras V12 -glands ( Figure 2A; see 'Materials and methods' for scored phenotypes; Ferrandon et al., 1998;Tzou et al., 2000;Takehana et al., 2004;Wagner et al., 2009). Of the 11 tested RNAi-lines, most of which were previously published to cause phenotypes, only one targeting the NFkB-transcription factor Dorsal (dl) significantly reduced the fluorescence signal of the Drs-reporter ( Figure 2A; Supplementary file 3). However, Drs expression is completely independent of the upstream modules of the two classical Drosophila immune pathways, Toll and imd.
The dl-Drs-relationship in Ras V12 -glands was further confirmed by the strong correlation of nuclear signals between dl-immunofluorescence and Drs-reporter intensities across PP and DP at 96 hr AED (Figure 2-figure supplement 2B; see 'Materials and methods'). Moreover, at 96 hr AED dl was present in the entire secretory epithelium of both Ras V12 -and w 1118 -control glands. In contrast, at 120 hr AED its expression was solely confined to the PP ( Figure 2B). This overlapped with the Drs-mRNA expression as determined by in situ hybridization (ISH; Figure 2-figure supplement 2C). Furthermore, both SG-specific knock-down or whole organism homozygous knock-out (dl 15 ) of dl abolished the majority of Drs-expression in the DP of the gland at 96 hr AED and reduced it in the PP at both time points ( To assess the extent to which Drs relies on dl for its expression, we measured the effect size on Drs expression of the employed dl RNAi -construct in Ras V12 -glands (Figure 2-figure supplement 3A). A reduction of Drs-expression to 6% in the DP of 96-hr-old dl RNAi ;Ras V12 -compared to Ras V12glands indicates not only a strong reliance for Drs on dl in this part at 96 hr, but also a high efficiency of the dl RNAi -construct. However, a less pronounced decrease of Drs-expression in the PP at 96 hr and 120 hr AED to 27% and 52% upon dl RNAi -coexpression implies the presence of additional Drsregulating factors. Putative candidates for this Drs-regulating role were identified by screening transcription factor binding sites amongst the up-and downregulated genes in Ras V12 -glands and all other acquired transcriptomes ( Together, our results indicate that the spatio-temporal dynamics of Drs expression are strongly correlated to the decrease in endogenous dl expression, independent of canonical Toll-and imd-signaling. While dl expression is Ras V12 -independent, Drs is only expressed in the presence of dl during Ras V12 -induced hypertrophy. Alongside dl, Mef2 regulates Drs-activation emphasising the notion of non-canonical activation of this immune effector even further. -log 10 (p-value) ribosome biogenesis salivary gland development salivary gland morphogenesis salivary gland histolysis EGFR sig. pathway regulation of EGFR sig. pathway positive regulation of EGFR sig. pathway negative regulation of EGFR sig. pathway growth cell growth regulation of cell growth developmental growth regulation of developmental growth cell death regulation of cell death programmed cell death regulation of programmed cell death positive regulation of programmed cell death defense response immune response regulation of immune response immune response process regulation of immune response process innate immune response humoral immune response organ or tissue specific immune response  Significantly differentially expressed genes (log2(beta) !1; q-value 0.05) and genes belonging to GO-term 'immune response' (GO:0006955) highlighted in yellow and blue. Right: Gene expression in Ras V12 , lgl RNAi ;Ras V12 or lgl RNAi ;Ras V12 -PP compared to w 1118 -glands for immune genes significantly upregulated in the PP. Blue arrows indicate strongest expression in the PP for the indicated genes between all three groups. Missing bars indicate absence of expression values in the RNAseq data. (D) GO term enrichment among significantly upregulated genes in Ras V12 -glands including terms related to activation of JNK (green) and immune responses (red). Numbers in bars indicate amount of upregulated genes belonging to associated GO term. (E) qPCR results for canonical JNK target genes (log2-transformed, fold-change over Rpl32) at 96 hr after egg deposition (AED). Lower/upper hinges of boxplots indicate 1st/3rd quartiles, whisker lengths equal 1.5*IQR and bar represents median. Significance evaluated by Student's t-tests (***p<0.001). (F) TREGFP1b reporter (green) signal in Ras V12 -and control-glands at 96 hr and 120 hr AED. Scalebar: 100 mm. Figure 3 continued on next page

Hypertrophy in SGs induces parallel immune and stress responses
Between 96 hr and 120 hr AED, the decrease in Drs-expression ( Figure 1E; Figure 2B-C) is correlated with deterioration of tissue integrity ( Figure 1A-B) in the DP of hypertrophic Ras V12 -glands. Thus, we hypothesized that the tissue-autonomous immune response revealed by Drs-expression aids in preventing the collapse of nuclear as well as cellular integrity until 96 hr AED in the DP and due to its prolonged expression in the PP until 120 hr AED and beyond ( Figure 1-figure supplement 1I).
To identify possible targets and effector mechanisms of the immune response that buffer the detrimental effects of hypertrophic growth, we analyzed the transcriptome data acquired for the w 1118control and Ras V12 -glands at 96 hr AED in further detail. In order to distinguish whether the differences between Ras V12 -and lgl RNAi ;Ras V12 -glands ( The sets of significant and differentially upregulated genes (i.e. log2(beta) !1; q-value 0.05) for all three conditions (i.e. Ras V12 / lgl RNAi ;Ras V12 / lgl RNAi ;Ras V12 -PP each normalized to w 1118 -controls) were intersected to determine common and specific genesets ( Figure 3A). Notably, the two biggest genesets are differentially upregulated genes shared between all three conditions and genes specifically upregulated in the PP ( Figure 3A). Furthermore, while all 6 Ras V12 -and lgl RNAi ;Ras V12 -replicates are in close proximity along the first two principle components in the PCA, the sets of replicates for w 1118 and especially the PP are distant from the rest along PC2 emphasizing the distinctiveness of the PP compared to the rest of the gland ( Figure 3B).
By screening for enriched gene ontology (GO) terms amongst the significantly upregulated genes in the PP of lgl RNAi ;Ras V12 -glands in comparison to the entire w 1118 -glands, we identified genes belonging to the GO-term 'immune response' (i.e. GO:0006955) as significantly enriched ( Figure 3C; p-value=1.45Á10 À4 ). Detailed examination of the fold-enrichment of these genes across all three experimental groups (i.e. Ras V12 , lgl RNAi ;Ras V12 , lgl RNAi ;Ras V12 -PP) showed a preferential expression in the PP, too (17 of 30 genes highest expressed in PP; blue arrowheads in Figure 3C). In addition, 5 of the top 20 upregulated genes in the PP belong to this GO-term as well. Thus, the Drsexpression we observed using reporter lines serves as a proxy for a more complex, tissue-autonomous immune response in hypertrophic glands, especially in the PP. Nonetheless, Drs itself remains one of the top significantly, differentially upregulated genes across all three conditions ( Figure 1D; Figure 3C; not shown for lgl RNAi ;Ras V12 ). In fact, Drs-expression in the PP compared to the two conditions for entire glands is even further increased, confirming the strong tendency for proximal over distal Drs-GFP reporter activation ( Figure 1E; Figure 2B; Figure 3C). Moreover, AttD and Drs are the only two AMPs that showed considerable expression in the two groups involving whole gland samples (i.e. Ras V12 , lgl RNAi ;Ras V12 ). Strong expression of particular genes in the PP might occur as expression throughout the whole gland samples. Therefore, we tested AttD expression separately in PP and DP, which showed almost exclusive AttD-expression in the PP at 96 hr and 120 hr AED (Figure 3-figure supplement 1D). Similarly, CecA1 as the 4 th highest expressed AMP in the PP-RNAseq was also confined exclusively to the PP as evaluated by ISH (Figure 3-figure supplement 1C, E). Thus, based on the RNAseq and additional evaluations, Drs appears to be the only AMP to be significantly expressed in the DP until 96 hr AED, while in parallel the PP shows a much broader immunocompetence.
A similar GO-term analysis amongst significantly, upregulated genes in entire Ras V12 -or lgl RNAi ; Ras V12 -glands revealed the enrichment of genes associated with GO-terms related to 'growth', 'salivary gland development' and 'EGFR signaling' consistent with the studied tissue and the induced Ras V12 -overexpression. Amongst the genes enriched in both Ras V12 -and lgl RNAi ;Ras V12 -glands were also sets significantly overlapping with GO-terms indicating activated JAK-STAT-signaling (data not shown). By using a construct reporting Stat92E-expression, we were able to confirm this geneset enrichment and show restriction of Stat92E-expression to the PP throughout development of Ras V12glands (Figure 3-figure supplement 1F; Bach et al., 2007). This may implicate JAK-STAT-signaling in the maintenance of immune competence in the PP and underlines the separation into PP and DP along the gland epithelium further. Moreover, the lack of unique, significantly enriched GO terms and thus distinct gene expression signatures for either Ras V12 -or lgl RNAi ;Ras V12 -glands at 96 hr AED excludes qualitative differences between these two genotypes as an explanation for their phenotypic differences (Figure 3-figure supplement 1B).
Importantly, we detected signatures of an activated JNK-cascade as well as cell death in Ras V12and lgl RNAi ;Ras V12 -transcriptomes indicating the presence of an activated stress response and confirming the stimulation of PCD as implied by nuclear disintegration at 120 hr AED ( Figure 1B; Figure 3D; Figure 3-figure supplement 1B; GO:0006955; GO:0008219). To validate these signatures further, we performed qPCR for canonical JNK-targets at 96 hr AED and found that the expression of all tested genes was significantly increased compared to w 1118 -control glands ( Figure 3E; Figure 3-figure supplement 2A). We used the TRE-GFP1b-reporter construct, which recapitulates JNK-activation by expressing GFP under the control of binding sites for JNK-specific AP-1 transcription factors ( Figure 3F; Chatterjee and Bohmann, 2012). This not only confirmed JNK-signaling in Ras V12 -glands, but also uncovered its prevalence in the DP of these glands at 96 hr and 120 hr AED consistent with the transcriptome data ( Figure 3F; Figure 3-figure supplement 2B). Moreover, the elevated fluorescence signal in the DP at 120 hr AED compared to 96 hr AED implies an increase in activation over time correlating with the collapse of nuclear and cellular integrity during this period.
In summary, the transcriptome analysis confirms our findings that Ras V12 -overexpression induces a strong tissue-autonomous immune response in the PP of the SG, beyond sole Drs expression. In contrast, the DP shows a striking increase in JNK-signaling which correlates with decreasing Drs expression and cellular and nuclear disintegration at 120 hr AED, consistent with the described role of JNK target genes in PCD.

Drs overexpression and JNK inhibition prevent Ras V12 -induced tissue disintegration
The increase in JNK-signaling in the DP of the SG as revealed by our transcriptome analysis coincided in space and time with the downregulation of Drs suggesting an interaction between the tissue-autonomous immune response and the stress response.
To test this assumption, we first overexpressed either Drs or a dominant negative form of the Drosophila Jun kinase jnk (basket) individually with Ras V12 throughout the entire secretory epithelium of the SG ( Figure 4A). Either Drs overexpression or JNK inhibition had a profound effect on the gland size, which was significantly increased at 120 hr AED, compared to Ras V12 or UAS-dilution control mCD8::RFP;;Ras V12 only ( Figure  Importantly, despite this size increase, glands of both genotypes (i.e. Drs,Ras V12 / jnk DN ;;Ras V12 ) showed no morphological abnormalities and resembled control w 1118 -much more than Ras V12glands ( Figure 4A-B). Strikingly, jnk DN ;;Ras V12 -and Drs,Ras V12 -glands were completely devoid of attached hemocytes ( Figure 4B,D). Given the basement membrane's (BM) role in directly regulating organ morphology, the rescue of the gland shape upon coexpression of either Drs or jnk DN with Ras V12 pointed towards changes in the integrity of the BM (Ramos-Lewis and Page-McCaw, 2019). In addition, previous reports suggested that hemocytes are only recruited to tissue surfaces upon tissue disintegration and when the integrity of the BM is lost (Kim and Choe, 2014). To trace the BM we used an endogenous GFP-trap in the viking gene, which encodes one subunit of CollagenIV ( Figure 4B). Both, in the presence of Drs or by inhibiting jnk, the BM remained a continuous sheet surrounding the entire gland, whereas the BM on the surface of Ras V12 -glands was clearly disrupted.
Matrix metalloproteinases (Mmps) are likely candidates for executing the disruption of the BM and known target genes of the JNK-pathway (Uhlirova and Bohmann, 2006;Hauling et al., 2014;Stevens and Page-McCaw, 2012). In order to see whether hypertrophic glands express Mmps, we performed qPCRs for Mmp1 and Mmp2 at 96 hr and 120 hr AED on Ras V12 -and control w 1118glands. At 96 hr AED, Ras V12 -glands already exhibit increased Mmp1 expression ( Figure 4E). However, Mmp2 only reaches a significant level of expression at 120 hr AED coinciding with the appearance of hemocyte attachment ( Figure 4D-E). SG-wide, Ras V12 -independent overexpression of Mmp2, but not of Mmp1, caused opening of the BM and hemocyte attachment to the surface (Jia et al., 2015). In contrast, knock-out of Mmp2 reduces hemocyte recruitment significantly compared to Ras V12 -glands ( Figure 4F-G; 33% in Mmp2 -/+ ;Ras V12 ; 34% in Mmp2 -/-;Ras V12 ). In turn, residual hemocytes still attached to the gland surface of Mmp2 -/-;Ras V12 -glands were not activated (i.e. no filo-or lamellipodia, no spreading; Figure 4-figure supplement 1C). This indicates the necessity for a JNK-dependent expression of Mmp2 in the hypertrophic Ras V12 -glands prior to the recruitment of hemocytes to the tissue surface. However, the presence of residual hemocytes indicates additional cues such as reactive oxygen species (ROS) as recently suggested for neoplastic tumors (Diwanji and Bergmann, 2019). In fact, while Ras V12 -glands show accumulation of ROS in the DP, this accumulation is reverted upon Drs-overexpression or JNK-inhibition (Figure 4-figure supplement 1D). A parallel block of ROS-accumulation and Mmp2-dependent BM-degradation in Drs,Ras V12 and jnk DN ;;Ras V12 could explain the absence of any attached hemocytes on the surface of these glands ( Figure 4B,D).
Taken together, overexpression of Drs alone is sufficient to mimic the inhibition of the JNK-pathway in Ras V12 -glands: both lead to excess hypertrophic growth compared to Ras V12 -glands, but simultaneously prevent tissue disintegration and PCD. Prohibiting the disruption of the basal membrane reduces hemocyte recruitment and activation, thus preventing the cellular immune response from sensing hypertrophic growth. This in turn suggests that the endogenous Drs expression in Ras V12 -glands is seminal for maintaining nuclear and tissue integrity and thus part of the buffer mechanism to adapt to continuous growth signaling.

Drs inhibits JNK-signaling
The strong correlation between loss of Drs and the increase in JNK-signaling in the DP of Ras V12glands between 96 hr and 120 hr AED indicated an active interaction between Drs and the JNKpathway, which prompted us to resolve their hierarchy by epistatic analysis.
Coexpression of jnk DN with Ras V12 did not deplete Drs in the PP of 120-hr-old glands, since neither the fluorescence signal of the Drs-GFP reporter nor staining for endogenous Drs-mRNA via ISH showed any effects ( Figure 4A; Figure 5-figure supplement 1A). This excludes a direct regulation of Drs by JNK-signaling and also negates indirect reduction of Drs-expression in the DP due to JNKinduced PCD at 120 hr AED (see Figure 6 for more details). qPCR for Drs in Ras V12 -and jnk DN ; Ras V12 -glands confirmed these results further ( Figure 5-figure supplement 1B). In contrast, qPCR in hypertrophic Ras V12 -glands showed a significant reduction in expression of JNK target genes upon coexpression of Drs at 96 hr and 120 hr AED ( Figure 5B; Figure Figure 5B,D,F). Thus, the overexpression of the AMP Drs actively and tissue-autonomously inhibits the JNK-dependent stress response in hypertrophic Ras V12 -glands beyond 96 hr AED.
In addition, we used an efficient RNAi-line against Drs, which in combination with Ras V12 reduced Drs-expression significantly compared to its expression in Ras V12 -glands as shown by qPCR and ISH (  Figure 1E). This result demonstrates that until 96 hr AED the endogenously expressed Drs has the same inhibitory effect on the JNK-pathway as the Drs-overexpression has at 120 hr AED.

Drs prevents cell death in Ras V12 -glands
The capacity to induce apoptosis is a well-established function of the JNK-pathway in Drosophila (Uhlirova et al., 2005;Igaki et al., 2006;Enomoto et al., 2015;Uhlirova and Bohmann, 2006;Cordero et al., 2010). Combined with the identification of 'cell death' as a significantly enriched term in the GO-analysis, we hypothesized that the observed nuclear disintegration in Ras V12 -glands after 96 hr AED was a consequence of the JNK-dependent induction of PCD (Figure 1B-C; Figure 1D-F; Figure 3D; Figure 3-figure supplement 1B).
In mitotic tissues, the apoptotic inducer head involution defective (hid) is inhibited by Ras/MAPKsignaling (Bergmann et al., 1998;Kurada and White, 1998). However, in Ras V12 -glands hid expression gradually increased from 96 hr to 120 hr AED coinciding with the increase in nuclear disintegration ( Figure 6A). Coexpression of Drs with Ras V12 significantly decreased hid expression at both time points, while Drs-knock-down in Ras V12 -glands increased hid expression even further already at 96 hr AED. Together with the reduction in hid expression upon JNK-inhibition in jnk DN ;;Ras V12glands, Drs emerges as a negative regulator of the apoptotic inducer hid by inhibiting JNK-signaling ( Figure 5-figure supplement 1B).   To evaluate the induction of PCD in Ras V12 -glands, we either inhibited JNK-signaling at the level of jnk activation or coexpressed the caspase-inhibitor p35 with Ras V12 and examined nuclear volume and integrity ( Figure 6B-D). p35 inhibits Drice and thus blocks PCD at the level of effector caspase activation (Hay et al., 1994;Meier et al., 2000;Hawkins et al., 2000). While both interventions successfully blocked nuclear disintegration, the additional rounds of endoreplication as observed in Ras V12 -glands were not significantly suppressed ( Figure 6C-D). Crucially, Drs-coexpression in Ras V12 -glands phenocopies the inhibition of JNK-signaling and effector caspases both in terms of restoring nuclear integrity as well as the persistence of excess endoreplications, in spite of a trend of overall nuclei size reduction similar to p35;Ras V12 -glands.
Last, to validate the inhibition of PCD by Drs, we monitored Dronc activity using the cleaved caspase 3 (CC3) antibody. In fact, the strongest Dronc activity occurred in cells that also displayed heavy disintegration of nuclei, confirming the relation between caspase activation and nuclear disintegration as part of PCD ( Figure 6-figure supplement 1B). However, in Drs,Ras V12 -glands Dronc was significantly less activated compared to Ras V12 -glands, a phenotype which we showed to be independent of UAS-dilution effects ( Figure 6E-F; Figure 6-figure supplement 1A-D; Fan and Bergmann, 2010). Given the strong reduction of Dronc activity upon knock-down of hid in Ras V12glands, the inhibitory effect of Drs on PCD is most likely to a large extent dependent on hid-inhibition ( Figure 6-figure supplement 1E-F).
Taken together, we show that JNK-dependent induction of PCD is a consequence of the collapse in adaptation to Ras V12 -induced hypertrophy. Importantly, overexpression of Drs in Ras V12 -glands prevents PCD by blocking full JNK-activation. This is in contrast to recent observations where AMPs, including Drs, act pro-apoptically aiding in the elimination of tumor cells and limiting tumor size (Araki et al., 2019;Parvy et al., 2019).

Drs inhibits the JNK-feedback loop
Various Drosophila models for tissue transformation have shed light on the functional separation of the JNK-pathway into an upstream kinase cascade leading to jnk-activation and a downstream feedback-loop converging again on jnk activation ( Figure 7A; Fogarty et al., 2016;Pérez et al., 2017;Shlevkov and Morata, 2012;Muzzopappa et al., 2017).
Our results revealed a negative regulatory impact of Drs on JNK-signaling in Ras V12 -glands, but did not determine which part of the pathway is targeted by Drs. To answer this question, we uncoupled the feedback loop from the upstream kinase cascade by solely overexpressing hid ( Figure 7A). Irrespective of the differing models for the signal propagation downstream of Dronc, this initiator caspase remains seminal for establishing the actual feedback with jnk (Fogarty et al., 2016;Shlevkov and Morata, 2012). Thus, we stained glands for activated Dronc and jnk after a pulse of hid-overexpression during the larval wandering stage ( Figure 7A). As expected, hidexpressing glands showed highly elevated levels of activated Dronc, but also jnk, which indicates the presence of feedback activation. Moreover, both phenotypes were reversed upon coexpression of Drs during the hid-expression pulse ( Figure 7B-C,E).
To clarify that Drs operates in a similar fashion during Ras V12 -induced hypertrophic growth, we stained Ras V12 -glands ( Figure 6E Figure 6 continued on next page downstream of hid and upstream of Dronc and jnk, emphasizing an inhibition of the JNK-feedback loop rather than the initial kinase cascade.

Discussion
Tissue-autonomous vs. cellular immune response mediated via JNKsignaling By overexpressing Ras V12 in larval SGs, we made use of and overrode the gland's ability to adapt to growth signals. Activating constitutive Ras/MAPK-signaling allowed us to identify local immune and stress responses as part of a buffering mechanism to compensate the accumulation of stress and decipher natural limits of growth adaptation (Hauling et al., 2014). Remarkably, in this context, the local immune response inhibits the parallel stress response, an effect that to our knowledge has not been described before. Central to this inhibition is the AMP Drs, which directly impinges on the JNK-pathway and thereby subsequently also on inducing PCD, a function completely opposite to previous observations in other tissues (Araki et al., 2019;Parvy et al., 2019). These results also indicate an unprecedented role for an AMP as a signal transducer enabling tissue-autonomous crosstalk between immune and stress pathways. Dependent on the extent of JNK-inhibition by the local, Drsdependent immune response and thus the integrated decision of both on the state of the tissue's homeostasis, the gland epithelium attracts hemocytes as part of a wider, cellular immune response. By virtue of inhibiting JNK-signaling, tissue-autonomous and cellular immune responses antagonize each other, a balance that determines continuous hypertrophic growth or its restriction (Figure 8). The latter has far reaching implications for therapeutic approaches that need to consider the adverse effects that stimulating immune responses might have on tissue growth after damage and under stress.

Drs is expressed as part of a genuine tissue-autonomous immune response
The SGs of Drosophila larvae are an integral part of its gastrointestinal system and the lumen of the mature glands forms a continuum with the exterior. As such the glands are constantly exposed to microbial and pathogenic influences, which predestines them to act as a dedicated immunological barrier epithelium. However, raising larvae with Ras V12 -glands under strictly sterile conditions corroborated the authenticity of the immune response as tissue-autonomous and with high likelihood independent of exogenous or systemically distributed pathogenic stimuli. This is further emphasized by the dependency of Drs expression on the tissue-specific overexpression on Ras V12 and tissue-autonomous manipulations (i.e. Drs or jnk DN overexpression) leading to the inhibition of JNK-activation and hemocyte recruitment. The activation of immune effectors in the absence of exogenous or endogenous pathogens is also consistent with the definition of sterile inflammation and aligns with finding in other models Kenmoku et al., 2017;Asri et al., 2019). However, Drs is embedded into a wider spatio-temporally regulated activation of immune effectors and immune mechanisms providing immunocompetence to the PP until 120 hr AED and thus constituting a genuine immune response ( Figure 3C-D; Figure 3-figure supplement 1B-C). Follow-up studies will determine the mode of their activation under physiological and sterile conditions and thus clarify their nature in the context of inflammation and immunity.
Apart from dl and Mef2 neither the homeobox transcription factor Caudal as in the adult glands nor any other of the canonical members belonging to the Toll-and imd-pathway are involved in Drs expression (Hauling et al., 2014;Ferrandon et al., 1998;Ryu et al., 2004). Under wild-type conditions, dl is expressed throughout the entire gland epithelium until 96 hr AED, but only remains  Figure 2B). This spatio-temporal expression pattern in turn determines to a large extent the expression of Drs in Ras V12 -glands essentially separating the gland at 120 hr AED into an immunocompetent duct-proximal and a stress-responsive duct-distal compartment ( Figure 2B; Figure

Drosomycin impinges on JNK feedback activation
In depth analysis of wound regeneration and tumor formation has shed light on the intricate architecture of the JNK-pathway and its signal propagation (Santabárbara-Ruiz et al., 2015;Pinal et al., 2018). This has led to the discovery of feedback-or self-sustenance-loops as part of JNK-signaling (Fogarty et al., 2016;Pérez et al., 2017;Shlevkov and Morata, 2012;Muzzopappa et al., 2017).
Similarly, several lines of evidence validated the existence of a complex quantitatively and qualitatively regulated activation of the JNK-pathway in Ras V12 -glands. In fact, the strong reduction of jnk activation upon knock-down of the initiator caspase Dronc in the Ras V12 -background at 120 hr AED indicates that a feedforward-loop is also part of the propagation of JNK-activation in hypertrophic glands ( Figure 7D,F; Figure 7-figure supplement 1A-B). Consistently, even sole hid overexpression in SGs led to Dronc as well as jnk activation emphasizing the presence of this feedback-regulation as part of the JNK-pathway further ( Figure 7B-C,E). In general, feedback loops serve as a predestined platform to integrate additional signals via crosstalk with other pathways to dynamically modulate the originally transmitted signal and thus either amplify or weaken the response (Kholodenko, 2006;Antebi et al., 2017). Here, we describe a novel mode of signal attenuation by the tissue-autonomous immune response that rather than eliminating an exogenous stimulus directly interferes with the signal propagation in the JNK-feedback loop. Fundamentally, this interaction is part of an emerging picture of crosstalk between immune and stress responses involved in organ growth and maintenance in Drosophila (Wu et al., 2015;Liu et al., 2015;Parisi et al., 2014;Hauling et al., 2014).
Drs regulates the signal propagation of the intracellular module of the JNK-pathway and throughout our experiments only directly Drs-expressing cells prevented Ras V12 -induced cell disintegration and PCD. However, it remains an outstanding question whether Drs functions exclusively cell-autonomously and intracellularly or whether secreted Drs operates in an autocrine manner too. Clonal analysis and rescue experiments will serve this purpose in the future. Further work on the effector mechanism of Drs will also elucidate further details about the components of the JNK-feedback loop in-or directly regulated by Drs. This will also contribute to mapping the manifold modes of immunestress-crosstalk in Drosophila and find general patterns among them beyond a sole dependency on the specific context.

Drs promotes hypertrophic growth and inhibits PCD
As our Drs,Ras V12 -experiments indicated, under conditions of continuous growth and therefore chronic stress induction, the ability to suppress the JNK-pathway in a Drs-dependent manner supports continuous, hypertrophic growth of Ras V12 -glands and the survival of the tissue. Moreover, a prolongation of Drs-expression beyond its endogenous decrease inhibits the induction of PCD and the recognition of the cellular immune response and thus renders the hypertrophic gland unchallenged.
In fact, while Ras/MAPK-signaling in mitotic tissues crucially suppresses apoptosis by downregulating hid expression, this effect is revoked in hypertrophic Ras V12 -glands (Bergmann et al., 1998;Kurada and White, 1998). Although the reason for this difference remains to be elucidated, it is in fact Drs which operates as the inhibitor of apoptotic inducers in hypertrophic glands. This function is central to the suppression of PCD in Ras V12 -glands ( Figure 6A).
This differs fundamentally from the pro-apoptotic function of AMPs, which was recently described for two tumor models. In both, disc (discs large, dlg) (Parvy et al., 2019) and leukemic tumors (mxc mbn1 ) (Araki et al., 2019), AMPs were shown to target tumor cells and limit tumor size by inducing apoptosis. In addition to the differences between the tumorous tissues (i.e. proliferative discs . Tissue-autonomous antagonizes cellular immune response via JNK-inhibition in Ras V12 -glands . dl-mediated Drs expression inhibits JNK activation in the entire Ras V12 -gland until 96 hr after egg deposition (AED). At 120 hr AED, dl and thus Drs expression is reduced to the PP, derepressing full JNK-activation in the DP. Consequently, JNK-dependent expression of hid and Mmp2 participate in stimulating nuclear and tissue disintegration, which eventually leads to the interference with tissue growth and integrity by the cellular immune response. and lymph glands), the fact that Drs is induced locally in SGs and acts tissue-autonomously may explain its different activities.
Hypertrophic SGs are a remarkable system to discover buffer mechanisms Being incapable of cell proliferation, damaged postmitotic tissues cannot rely on regenerative cell plasticity like imaginal discs or stem cell-derived tissue regeneration as in the Drosophila adult midgut (Ohlstein and Spradling, 2006;Herrera et al., 2013;Herrera and Morata, 2014;Schuster and Smith-Bolton, 2015;Ahmed-de-Prado and Baonza, 2018). Instead, they need to cope with endogenous and exogenous influences via elaborate mechanisms to prevent or buffer detrimental consequences.
Here, we show that continued Ras V12 -expression in the larval SG overrides the dependency on nutritional cues and stimulates excess endoreplications that eventually have damaging effects on the tissue integrity by inducing elevated levels of stress. Uninterrupted induction of endoreplications has its natural limits in every system, even in SGs that are already polyploid. Eventually, continuous induction of more endocycles is challenged by nutritional restrictions in synthesizing more DNA, replication stress, spatial limitations in the gland nuclei and continuously more unsynchronized metabolic turnover. Hence, in spite of the anti-apoptotic function Ras V12 conveys in mitotic tissues, unrestricted stimulation of excess endoreplications ultimately leads to cell death. Since this characterizes the final collapse of tissue homeostasis, it also allows to study the extent of buffering capacity conveyed by immune and stress-responsive signaling. While JNK-signaling has been shown to be seminal during development (i.e. tissue morphogenesis) (Agnès et al., 1999;Zeitlinger and Bohmann, 1999), its activation in Ras V12 -glands appears however rather likely to be due to its function in relaying stress signals, emphasizing the presence of stress as part of hypertrophic growth further.
Thus, hypertrophic Ras V12 -glands constitute an outstanding system to study the involvement of tissue-autonomous immune and stress-induced responses to buffer deviation from homeostasis.

Immune surveillance theory
According to the immune surveillance theory, the immune system has evolved to reduce the risk of somatic cells accumulating cancerous mutations (Burnet, 1957;Burnet, 1970). In order to reduce the danger of cell-transformation, cells express or expose molecules upon recognition of stress or damage during transformation. These markers are sensed by the immune system, which in turn eradicates the potentially harmful cells (Jung et al., 2012;Vantourout et al., 2014;Schmiedel and Mandelboim, 2018). Given the absence of exogenous or endogenous pathogenic agents, the activation of immune effectors (i.e. Drs) in the here presented system is most likely a consequence of sensing Danger or Damage Associated Molecular Patterns during hypertrophic gland overgrowth (Matzinger, 1994;Seong and Matzinger, 2004). However, the immune surveillance theory remains controversial, since tumor-associated inflammation was also shown to promote rather than suppress tumor growth (Balkwill and Mantovani, 2001;Mantovani et al., 2008). Our model bridges the gap between these two opposing views, since the effects of the tissue-autonomous and cellular immune responses appear to be antagonistic regarding the regulation of JNK-activation and thus ultimately PCD. In fact, only the integration of the various stress and immune mechanisms in hypertrophic Ras V12 -glands allows a concerted decision to eradicate a putatively dangerous cell via inducing PCD or not. Given the evolutionary conservation from insects to mammals of signaling pathways that govern growth control (Edgar, 2006), it is likely that mechanisms to detect and counteract a loss in regulation of these pathways, such as stress and immune pathways, are similarly conserved between both phyla.

Materials and methods
Key resources

RNASeq library preparation and analysis
To avoid variability and thus confounding influences among the various RNASeq sample groups, we controlled rigorously for age of female parents, larval density to avoid larval crowding, age differences as well as developmental age itself and bacterial influences by using axenic culture conditions. For each genotype, three biological replicates were dissected with 30 pairs of SGs for the three groups including whole glands and 60 pairs of PPs for the corresponding group to acquire 5 mg of total RNA. No power or sample size calculations were performed. Replicate numbers were determined by the available resources and funds. Poly(A)-containing mRNA molecules from totalRNA-samples were purified with oligo(dT)-magnetic beads, subsequently fragmented and cDNA synthesized with random primers using the TruSeq RNA Sample Preparation Kit v2. Adapter ligation and PCR-amplification precede cluster formation with a cBot cluster generation system. All samples were sequenced on a HiSeq 2500 Illumina Genome Sequencer as PE50. Reads were pseudoaligned with kallisto (v0.44.0) to a transcriptome index derived from all Drosophila transcript sequences of the dmel release r6.19. Subsequent analysis of transcript abundances was performed in R with sleuth (v0.30.0) including principal component analysis for dimensionality reduction, statistical and differential expression analysis based on the beta statistic derived from the wald test. Enriched gene ontology terms were identified by calculating hypergeometric p-values via the GOstats (v2.48.0) R package. Gene IDs were converted via the AnnotationDbi (v1.44.0) package and the reference provided with the org.Dm.eg.db (v3.7.0) database and geneset intersections visualized as UpSetR plot (v1.3.3). Please, see the accompanied R markdown (i.e. 'RNAseq_sg_analysis.Rmd') deposited on GitHub for details (https://github.com/robertkrautz/sg_analysis). The data is available under the accession number GSE138936 on the NCBI Gene Expression Omnibus.

Detection of enriched motifs and Mef2 visualization
Significantly, differentially up-(i.e. q < 0.001 and b > 1) or downregulated (i.e. q < 0.001 and b < À1) genesets were determined for all three sample populations by subsetting q-and b-values as determined by sleuth with the latter indicating fold-change over expression in BxGal4>w 1118 -control crosses. The RcisTarget R package (v1.6.0) together with the accompanied file including all dm6 motif rankings (i.e. 'dm6-5kb-upstream-full-tx-11species.mc8nr.feather') were used for motif enrichment analysis in the various genesets and adding annotations. Enriched motifs were manually screened for functionally related motif groups (e.g. 'NFkB' with dl-, Dif-, Rel-relaetd motifs) and motifs plotted according to their affiliation to these groups.
The sequence of the 'dm6' genome assembly was acquired as fasta-file from the Ensembl database and nucleotide sequence frequencies across the genome calculated via customized functions including all major chromosomes. In parallel, promoter sequences for all 'dm6' genes were collected via the biomaRt package (v2.42.1) and converted into a DNAStringSet with the Biostrings package (v2.54.0). A curated list of all 'dm6' transcription factor motifs as position probability matrices was obtained from the Cis-BP database and converted to position weight matrices (PWMs) with the help of the PWMEnrich R package (v4.22.0) by using the above cacluated nucleotide sequence frequencies. Based on these PWMs and all promoter sequences, background scores for all 'dm6' transcription factor motifs across all promoters of the 'dm6' genome were quantified. Sequences for the wider Drs-locus and the shared Dif-/dl-loci environments were obtained via biomaRt and the enrichment of the Mef2-motif 'M08214_2.00_Mef2' compared to all background scores was calculated via PWMEnrich. Mef2 motif enrichment scores were plotted as a function of the location of the detected motif in the screened loci (i.e. Dif/dl; Drs) with the help of the novel, devised function geom_bar_rastr(), which servers as an extension to the ggrastr R package. For details please follow the entire computational workflow as provided in 'RNAseq_motif_enrichment.Rmd' deposited on GitHub (https://github.com/robertkrautz/sg_analysis).

ROS-staining
H2DCFDA (D399, Thermo Fisher Scientific) was reconstituted in DMSO just prior to the experiment. This stock solution was further diluted to 10 mM in PBS. Freshly dissected glands were incubated in the 10 mM-working solution for 30 min in the dark. Three 5-min wash steps in PBS ensured removal of excess H2DCFDA-dye. Glands were mounted with PBS and confocal pictures obtained with the Zeiss LSM780 microscope system.

Hemocyte-/ pJNK-/ CC3-quantification and size measurement
Pictures of stained glands were taken with a Zeiss Axioplan two microscope equipped with an ACH-ROPLAN Â4 lens and a Hamamatsu ORCA-ER camera (C4742-95). Images were extracted with Axio-Vision software (v40V 4.8.2.0) and analyzed with ImageJ. Cumulated area of hemocyte attachment, pJNK or CC3 fluorescence per SG or per SG part was filtered in the Red-channel and gland size determined by outlining glands, PPs or DPs with the 'Polygon selection'-tool. The distribution of hemocytes in Drosophila can be approximated by a natural logarithm, which required transformation of hemocyte attachment-and SG-areas before calculating ratios (Sorrentino, 2010). Normality across all samples of a particular genotype and where necessary separated by gland part (i.e. proximal or distal) or time (i.e. 96 hr and 120 hr AED) was evaluated with the Shapiro-Wilk-test and by bootstrapping via the fitdistrplus R package (v1.0.11). Significant differences between experimental groups were determined for pairwise comparisons via the Student's t-test after validating un-/equal variance via the Bartlett's test. Data and statistical analysis was performed in R. No power or sample size calculations were performed. All experiments were performed at least twice, while preparing and acquiring meaurements for all specimens and replicates of the respective experimental set-up. The amount of replicates was determined as a trade-off between reproducibility (maximum number) and feasibility of the experimental procedure to ensure same treatment and timing for all specimens. No measurements were categorized as outliers or excluded. Effect sizes were calculated by normalizing the effect and the reference samples by the mean of the reference samples, which anchors the reference mean at '1' and reports all deviations from it in procent.

Nuclei volume
Z-stacks of entire SGs were captured with a Zeiss LSM780 confocal microscope for DAPI signal. Obtained stacks were further processed in ImageJ via the '3D Objects Counter'-plugin (v2.0.1.). Proximal and distal compartments were defined with the 'Polygon selection'-tool. Transfer of the region of interest to all z-stack slices, signal thresholding, object identification and volume determination were integrated in a macro workflow. Data were plotted as average nucleus volume for all nuclei of individual glands or gland parts. Representative sections of individual z-stack slices showing the transition between proximal and distal compartments were cropped and added for illustration. Statistics were performed similar to the analysis of Hemocyte-, pJNK-and CC3quantifications.

Drs-dl-correlation
BxGal4;;DrsGFP>Ras V12 -glands were stained for dl (mouse anti-dl-Ab; 1:50, DSHB, 7A4) and genomic DNA (DAPI), whereafter z-stacks were obtained on a Zeiss LSM780 confocal microscope. Proximal and distal gland compartments were delineated as regions of interest (ROI) in ImageJ with the 'Polygon selection'-tool and nuclei therein defined via their DAPI-signal and the '3D Objects counter'-ImageJ plugin. Transfer of the 3D-nulei outlines across the dl-and Drs-channels via the ImageJ 'RoiManager 3D' (v3.96) allowed quantifying an accumulated intensity of the corresponding signals per nucleus. Data analysis was conducted in R including calculation of linear regression and Pearson coefficients for correlations between nuclei-wide Drs-and dl-signal for individual gland compartments and visualization of all nuclei data points after log 10 -normalization for all five screened samples.

Driver expression strength
DrsGFP-reporter or Phalloidin-stained, GFP-glands were captured with the 'Zeiss Axioplan two microscope / ACHROPLAN 4x lens / Hamamatsu ORCA-ER camera (C4742-95) / AxioVision software (v40V 4.8.2.0)'-system. ROIs were added manually to the imaged glands to outline their position in ImageJ with the 'Polygon selections'-tool. Feret diameter and angle were measured to delineate the longitudinal axis of each gland. For each position along the longitudinal axis, the corresponding rectangular line region intersecting with the glands ROI is determined and the encapsulated average fluorescence intensity measured. This workflow is implemented as ImageJ-macro (i.e. 'scanner.ijm') for screening of entire picture batches and available on GitHub (https://github.com/robertkrautz/sg_ analysis). The resulting data was further analyzed in R to bin and align measurements along the longitudinal axis. Measurements were further averaged and normalized per bin across all screened glands. Given the statistically limited amount of observations (i.e. 200-350 bins), a loess function was used for smoothing the mean during visualization as implemented in geom_smooth() (i.e. span = 0.05).
SGs were dissected in PBS, fixed in paraformaldehyde (4% PFA in PBSTw: 0.1% Tween-20 in PBS) for 40 min at room temperature (RT), washed four times in PBSTw for 5 min and transferred to Methanol. Permeabilization was performed for 1 hr in Ethanol, followed by washing once in Methanol for 5 min and twice in PBSTw for each 5 min. The samples were treated with Proteinase K (30 mg/ml; ThermoFisher Scientific; EO0491) for 6 min, washed twice with glycine (2 mg/ml) for 1 min, fixed for 20 min in 4% PFA and washed four times in PBSTw for each 5 min.
. Supplementary file 3. Overview of all RNAi-lines used in the DrsGFP-reporter assay, including references indicating prior use and evidence for active inhibition of the outlined target of the respective RNAi-line.
. Transparent reporting form

Data availability
All sequencing data has been deposited at NCBI GEO under the record GSE138936.
The following dataset was generated: