Decision letter: Shared enhancer gene regulatory networks between wound and oncogenic programs

Abstract

Full text Figures and data Side by side Abstract Editor's evaluation Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Wound response programs are often activated during neoplastic growth in tumors. In both wound repair and tumor growth, cells respond to acute stress and balance the activation of multiple programs, including apoptosis, proliferation, and cell migration. Central to those responses are the activation of the JNK/MAPK and JAK/STAT signaling pathways. Yet, to what extent these signaling cascades interact at the cis-regulatory level and how they orchestrate different regulatory and phenotypic responses is still unclear. Here, we aim to characterize the regulatory states that emerge and cooperate in the wound response, using the Drosophila melanogaster wing disc as a model system, and compare these with cancer cell states induced by rasV12scrib-/- in the eye disc. We used single-cell multiome profiling to derive enhancer gene regulatory networks (eGRNs) by integrating chromatin accessibility and gene expression signals. We identify a ‘proliferative’ eGRN, active in the majority of wounded cells and controlled by AP-1 and STAT. In a smaller, but distinct population of wound cells, a ‘senescent’ eGRN is activated and driven by C/EBP-like transcription factors (Irbp18, Xrp1, Slow border, and Vrille) and Scalloped. These two eGRN signatures are found to be active in tumor cells at both gene expression and chromatin accessibility levels. Our single-cell multiome and eGRNs resource offers an in-depth characterization of the senescence markers, together with a new perspective on the shared gene regulatory programs acting during wound response and oncogenesis. Editor's evaluation This study is an important progression in our understanding of wounding response and its relationship to malignancy. Although this topic has been previously addressed in genetic studies, the use of a systems biology approach here provides compelling support for the dual use of regulatory sequences to achieve context dependence for two linked but non-redundant tasks. Investigators in the fields of gene regulation, developmental biology as well as basic cancer research will find this manuscript to be both important and useful. https://doi.org/10.7554/eLife.81173.sa0 Decision letter eLife's review process Introduction The Drosophila wing imaginal disc (WID) is a classical model system to study developmental patterning and cell differentiation. This larval primordium is composed of two epithelial cell layers, the peripodial epithelium and the disc proper, along with muscle precursors (Figure 1a). Despite its rather complex structure, the wing disc is extensively studied for its regeneration capacities. Damaged wing discs can trigger a set of wound-response mechanisms allowing for disc repair and the formation of normal wings (Smith-Bolton, 2016; Tripathi and Irvine, 2022). The molecular pathways identified as key drivers for this regenerative process include the regulation of cell apoptosis (JNK, JAK/STAT; La Fortezza et al., 2016), cell proliferation (EGFR, Wnt, Wingless, Scalloped; Blanco et al., 2010; Herrera et al., 2013; Irvine and Harvey, 2015; Smith-Bolton et al., 2009; Yu et al., 2015), re-epithelialization (ERK, Grainy head; Mace et al., 2005), and developmental timing (insulin-like peptide 8; Katsuyama et al., 2015). Interestingly, the same regulatory pathways may lead to uncontrolled cell apoptosis or neoplastic growth when unrestricted (La Marca and Richardson, 2020; Pérez-Garijo et al., 2013; Pinal et al., 2019). Such tumor-like outcomes are observed in rasV12scrib-/- transformed cells, where the loss of cell polarity triggers a cellular stress that cells cannot escape via apoptosis (Atkins et al., 2016; Brumby and Richardson, 2003; Cosolo et al., 2019; Davie et al., 2015; La Fortezza et al., 2016; Igaki et al., 2006; Külshammer et al., 2015; Pagliarini and Xu, 2003; Pinal et al., 2019; Uhlirova and Bohmann, 2006). In light of these outcomes, it is still unclear how such antagonistic mechanisms (apoptosis and proliferation) interact in the vicinity of a wound to orchestrate tissue repair while controlling for overproliferation. To study this process at the gene regulatory level, we generated a multi-omic dataset, jointly measuring chromatin accessibility and gene expression changes at single-cell resolution, in wild-type and genetically ablated wing imaginal discs from third-instar larvae (Figure 1a). Figure 1 with 2 supplements see all Download asset Open asset Gene expression patterns in wild-type and wounded wing discs. (a) Top: schematic of the wing imaginal disc subdomains from a Drosophila third-instar larvae. The disc proper (DP) is composed of three domains and is shoehorned between the peripodial epithelium (PE) and the adult muscle precursors (AMP, or myoblasts). bottom: Design of the wound experiment, eiger expression is induced in the pouch for 24 hr, resulting in a localized apoptosis. (b) Representation of the wing disc scRNA Atlas tSNE (left) and the scRNA multiome data UMAP (right), where wound populations α and β can be detected. (c) Gene expression of marker genes across cell types (left) and wound subclusters in contrast to pouch (right) clusters. (d) Gene expression pattern of wound markers in multiome UMAP (top) and immunostaining (bottom). Ilp8, rn and upd3 are found to colocalize with JNK reporter (TRE-RFP), present at the wound site. (e) Relative expression (log2FC) of markers of wound response (first row, wounded/wild-type, ▲) and/or developmental timing (second row, 96 hr/120 hr after egg laying, ●). Shared markers are marked by an asterisk (*). Leveraging the multidimensionality of our dataset, we constructed enhancer gene regulatory networks (eGRN), centered around transcription factors (TFs) and comprising both gene and enhancer signatures (Janssens et al., 2022). We detect two classes of wound populations, respectively expressing markers of proliferation and cellular senescence. The senescent cells are driven by eGRNs belonging to the C/EBP bZip family (Irbp18, Slow border, Vrille; Blanco et al., 2020), which we also found to be present in the tumor cells from the rasV12scrib-/- model. Results Single-cell multi-omics of the normal and wounded wing imaginal disc To study the gene regulatory program of a wound response at single-cell resolution, we used a published genetic model that induces a sterile wound (La Fortezza et al., 2016; Smith-Bolton et al., 2009; see ‘Materials and methods’). In this model, the expression of the TNF ligand eiger (egr) is induced in the wing pouch, where rotund (rn) is expressed, at specific times through a temperature shift regime. The expression of egr causes a wound by inducing extensive local cell death (Figure 1a). Wing discs were dissected 24 hr after the induction of rn-Gal4, the time point when most rn-expressing cells have undergone apoptosis, and markers of wound response are observed (Herrera et al., 2013; Pérez-Garijo et al., 2013). After disc dissociation, we performed multiome experiments using 10X Genomics (combined scRNA & scATAC from the same cell), as well as on wild-type control discs. This yielded 17,402 high-quality cells, with a median of 1124 detected genes per cell, and a median of 3060 unique ATAC fragments per cell (Figure 1—figure supplement 1). To increase the power to robustly detect cell types and simultaneously validate our single-cell data, we integrated our scRNA dataset with several previously published scRNA-seq datasets of wild-type wing imaginal discs, generating an integrated atlas. This wing atlas is available through SCope, along with clustering and marker gene information (Figure 1b, Figure 1—figure supplement 2a, https://scope.aertslab.org/#/WingAtlas/*/welcome). The integrated atlas contains 70,230 cells from 10X Genomics, across 10 replicates (four from this work, four from Everetts et al., 2021, one from Bageritz et al., 2019, and one from Deng et al., 2019). To annotate the cell types of the wild-type disc, we compared markers from literature with significantly upregulated gene sets obtained from each cluster (see ‘Materials and methods’), resulting in the identification of the previously known wing disc cell types and patterning domains (Figure 1c, Supplementary file 1). We also confirmed our annotation by comparison with the annotations from the integrated public datasets (Figure 1—figure supplement 2b). The annotated clusters form a continuum of epithelial cells, from the pouch over the hinge to the notum (globally marked by the epithelial marker grainyhead); and separate clusters of myoblasts (marked by myogenic genes such as twist, holes in muscle and zn finger homeodomain 1), hemocytes, and tracheal cells (Figure 1b, Figure 1—figure supplement 2c). One cluster was noticeably enriched for cells from the wounded disc samples, with markers linked to stress response pathways (e.g. kayak, insulin-like peptide 8, unpaired3, p-adj <10e–3, log2FC >1.7; Figure 1c, Figure 1—figure supplement 2d) and localized at the wound site in the pouch domain (Figure 1d), suggesting that this cluster represents a wound-response cell state. This cluster can furthermore be subdivided at higher resolution into two distinct cell populations (population α of 1211 cells and population β of 94 cells, Figure 1b and c). Next, we analyzed the scATAC-seq part of our multiome dataset separately using cisTopic (Bravo González-Blas et al., 2019; Figure 1—figure supplement 2e). Since multiome data delivers same-cell measures for RNA and ATAC, we could label the scATAC-seq based on the previously derived scRNA-seq annotations. The detected ATAC clusters, based solely on chromatin accessibility, also identify the hinge, pouch, notum, myoblast, peripodial epithelium, and wound clusters (Figure 1—figure supplement 2f). Thus, both chromatin accessibility and gene expression independently identify normal cell types and a separate wound cell state. The wounded disc is delayed in its developmental timing The proper regeneration of a damaged wing disc is tied to the introduction of a developmental delay via a reduced ecdysone signaling. This delay provides the necessary time for tissue repair before pupation (Jaszczak and Halme, 2016; Katsuyama et al., 2015; Sanchez et al., 2019). To assess whether we can detect this regulatory response in our integrated dataset, we combined cells from multiple conditions (wounded/wild-type) but also normal discs dissected at different developmental time points (96 and 120 hr after egg laying, from Everetts et al., 2021). By comparing the up and downregulated genes with respect to the developmental time, we found that marker genes of late time points (e.g. ecdysone-inducible gene E2) are globally downregulated in the wounded disc samples dissected at the same time point. We furthermore found a significant overlap of downregulated markers for both wound response and developmental timing (21%, p-adj <10e–3, Fisher’s exact test, Figure 1e). We confirmed this result when restricting the analysis outside of the wound site, in the notum domain (34% shared downregulated genes, p-adj <10e–3, Fisher’s exact test, Figure 1—figure supplement 2g). This resemblance of wounded disc cells with those from earlier stage wild-type larvae confirms that wounding triggers a global reaction across the whole disc which delays development to give more time for the wound to repair completely before metamorphosis. This delay is likely driven by insulin-like peptide 8 (ilp8), a critical long-range regulator of ecdysone signaling, highly expressed in the wound cluster (Figure 1c and d). We additionally found genes significantly upregulated in the entire wing disc following wounding, with no strong change between developmental time points (p-adj <10e–3, log2FCwound > 1.7, log2FCtiming < 0.5, Figure 1e). Among these wound markers, suppressor of cytokine signaling at 36E (socs36E) is a known target of the JAK/STAT and EGFR pathways (Berez et al., 2020; Kang et al., 2018). We also noticed an upregulation of pickled eggs (pigs), a potential Notch regulator (Pines et al., 2010), in the hinge and peripodial cells of wounded discs. Multi-ome gene regulatory network reconstruction An important advantage of single-cell multiomics data is the power to detect regulatory interactions by synchronous changes in expression and/or accessibility across cells (Fiers et al., 2018). Here, we set out to infer enhancer-GRNs (eGRN) following a similar strategy as we recently applied to the fly brain (Janssens et al., 2022; Figure 2a, see ‘Materials and methods’). First, we defined differentially accessible regions (DARs) for each cell type, including the wound cluster, using cisTopic (Bravo González-Blas et al., 2019). Next, these DARs were tested for enrichment in TF binding motifs using cisTarget (Herrmann et al., 2012), which resulted in a list of TF-to-regions edges (i.e. regulatory links) with a significant motif hit. Next we determined co-variability between gene expression and accessibility of neighboring enhancer regions to generate a list of region-to-gene edges. We then completed the edges loop by using the TF-to-gene adjacency scores from pySCENIC (Van de Sande et al., 2020). Lastly, we filtered for high-quality TF-to-region-to-gene interactions by keeping the leading edges of a gene set enrichment analysis (GSEA), taking the TF-to-gene scores as ranking values. Figure 2 with 1 supplement see all Download asset Open asset Construction of enhancer-mediated gene regulatory networks (eGRNs). (a) eGRN construction is based on TF motif enrichment (1), co-variability of gene expression and chromatin accessibility (2–3), and followed by functional edge selection (4). (b) Aggregated wild-type chromatin accessibility signals around a target region (black rectangle, gold shaded box) of the blistered eGRN. The target region comprises a blistered motif hit, it is significantly correlated with expression of defective proventriculus (dve, orange arcs), and it overlaps a flylight reporter (green rectangle) expressed in the pouch domain. The activities of the blistered TF and its associated eGRN signatures are all similarly localized in the pouch (UMAP and violin plots). (c) Dotplot of the average gene-based (left) and region-based (right) activity of selected eGRNs with highest cell type specificity in wild-type conditions. (d) Aggregated chromatin accessibility signals similar to panel (b) for the target regions of two other eGRNs, active in the myoblast (twist, left) and the hinge domains (Sox15, right). This eGRN inference approach has the advantage of including distal enhancers, found up to 50 kb away from a putative target gene. This procedure resulted in 147 high-quality eGRNs (85 activating, 62 repressing), spanning 98 TFs, with on average 54 target genes and 58 target regions (Supplementary file 2). An example of an inferred eGRN is the TF blistered (bs), which is predicted to target the repressor defective proventriculus (dve) via multiple intronic enhancers (Figure 2b). Using the target genes and target regions sets as proxies, we scored the activity of a TF and its subsequent domain specificity via its entire eGRN by AUCell (Figure 2c, see ‘Materials and methods’). In the case of bs, we found the eGRN to be specifically active in the pouch, both from the gene expression and the region accessibility perspectives (Figure 2b and c). We can further divide our final list of eGRNs into repressor and activator categories, based on linear correlation between TF expression and accessibility. For example, we identified mirror (mirr) as a repressor TF in control wing discs because its expression negatively correlated with accessibility of both its target genes and regions (Figure 2—figure supplement 1a). The repressive action of Iroquois TFs like mirr has already been demonstrated in Drosophila (Andreu et al., 2012; Bilioni et al., 2005). We used the TF activity scores derived from gene expression and chromatin accessibility to compute the regulon specificity score (RSS) of each eGRN across the disc domains (Suo et al., 2018). A high RSS score indicates an enrichment for the TF signature (gene or region targets) among the top markers of a given domain. One of the strongest regulatory programs was observed in the hinge with the Sox box protein 15 (Sox15) (Dichtel-Danjoy et al., 2009) eGRN (22 genes, 36 regions, Figure 2c and d) where the top predicted target genes included zn finger homeodomain 2, frizzled, dachsous, and homothorax. Several other well-known wing development TFs were identified, including tailup and odd-paired in the notum, ultrabithorax and C15 in the peripodial epithelium and twist in the myoblasts (Figure 2c and d, Figure 2—figure supplement 1a–c). Another interesting example is nubbin (nub), a POU/homeodomain transcription factor targeted by vestigial (Rodríguez D del Á et al., 2002), which is found specifically expressed in the pouch domain (Figure 2—figure supplement 1a). An extended list of TFs and their target genes can be queried via SCope (https://scope.aertslab.org/#/WingAtlas/*/welcome), and is provided as a Supplementary file 2. The wound site shows a strong JNK and JAK/STAT eGRN activity Both α and β wound clusters are unique to the wounded disc and display markers of wound response at both gene expression and chromatin accessibility levels. These two clusters are associated to 3980 enhancers (DARs, see ‘Materials and methods,’ Supplementary file 3), among which 24% have been previously described as damage-responsive regulatory elements (Harris et al., 2020; Vizcaya-Molina et al., 2018), including the BRV118 locus (Gracia-Latorre et al., 2022; Figure 3—figure supplement 1). Their signatures at the gene expression level also share enrichment for GO terms related to wound response and paracrine signaling (p-adj <10e–3). The two wound clusters express high levels of stress-response genes, including ilp8 (Katsuyama et al., 2015), matrix metalloproteinase 1 (Mmp1; Harris et al., 2020), moladietz (mol; Khan et al., 2017), PDGF- and VEGF-related factor 1 (pvf1; Wu et al., 2009), and jun-related antigen/kayak (jra/kay, homologs of JUN and FOS, forming the AP-1 complex, involved in the JNK cascade); (Cosolo et al., 2019; Figure 1c and d), (p-adj <10e–3, log2FC > 1). The activity of the JNK pathway is further confirmed by a high and specific activity of the Jra/Kay eGRNs in the wound populations, at both gene expression and chromatin accessibility levels (Figure 3a). We also confirm the strong involvement of JAK/STAT as the eGRN of Signal-transducer and activator of transcription protein at 92E (Stat92E) is specifically active in the wound response clusters (Figure 3a), in agreement with the enrichment for the Stat92E binding motif in the wound-specific accessible regions (NES = 3.6) and the co-expression of the unpaired ligands (upd1/2/3, Figure 1c). Figure 3 with 5 supplements see all Download asset Open asset bZIP TF activity in senescent and proliferative wound populations. (a) Dotplot of the average gene-based (left) and region-based (right) activity of selected enhancer gene regulatory networks (eGRNs) with highest α and/or β specificity in wounded conditions. (b) Feature maps of six types of TF motifs on regions specifically accessible in the wound populations α and/or β. AP1 bindings are homogeneously present, while Stat92E and C/EBP motifs are specifically enriched in wound population α and β, respectively. One region targeted by both AP1 and C/EBP GRNs regulates two genes with antagonistic α/β expression patterns (mys and Gclc). (c) Scatterplot of the relative changes in target gene expression and chromatin accessibility for all enhancers targeted by the C/EBP eGRNS (vri, Irbp18, Xrp1, slbo). We note the presence of the CEBP-TEAD dimer motif in regions strongly upregulated in β. (d) Wing disc immunostaining (left) and normalized average expression of three wound marker genes and a proliferation marker (CycE, right). Both α and β marker genes are expressed and localized at the wound site. (e) Expression change of Xrp1 eGRN target genes following Xrp1 overexpression, we note the strong upregulation of the Unpaired ligands. Other transcription factors with induced eGRN activity in both wound clusters include the Cyclic-AMP response element binding protein A (CrebA, CREB3L2 ortholog, involved in resistance to infection), Cap'n'collar (Cnc, Nrf1/2 ortholog, involved in oxidative stress response), and Cryptocephal (Crc, ATF4 ortholog, involved in unfolded protein response) (Figure 3a), three basic-leucine zipper (bZip) TFs known to be involved in stress response (Brock et al., 2017; Brown et al., 2021; Ragheb et al., 2017; Sorge et al., 2020; Sykiotis and Bohmann, 2008; Troha et al., 2018). Interestingly, we note an additional upregulation of both Stat92E and AP-1 eGRN activity in the peripodial epithelium (PE) of the wounded disc (Figure 3a). Although the PE was not directly targeted by the genetic ablation, this domain is in close proximity with the pouch territory in wild-type wing discs (Figure 1a). We therefore hypothesized that stress signaling is capable of local diffusion across epithelial layers. We further identified a JAK/STAT repressor apontic (apt), to be specifically expressed in the peripodial epithelium in wounded conditions (p-adj < 10e–3, log2FC > 1). This finding supports a protective role of Apontic to block a response to Stat92E proliferative signaling in the vicinity of wounded tissues (Harris et al., 2020). Taken together, our results highlight JNK and JAK/STAT as the most prominent markers of the global wound response program, shared by the two cell populations α and β. AP1 is the largest inferred wound eGRN (470 target genes, 689 target regions) and its binding motifs (CRE : TRE variants, Fonseca et al., 2019) are strongly enriched in DARs from both wound populations α an β (Figure 3b). Proliferative and senescence signals separate the α and β populations In the wound samples, we expect the rn-expressing pouch cells to be almost entirely lost upon apoptosis from the rn-Gal4 induction. However, we find persistent rn-positive cells in the wound cluster α (p-adj < 10e–3, log2FC > 2) that are excluded from cluster β (p-adjα/β < 10e–3, log2FCα/β > 2). These cells are distinct from the normal rn-positive pouch cells though as they also show stress response markers. We further observe an upregulation of markers of tissue patterning and proliferation in cluster α relative to cluster β, like wingless (wg), Wnt oncogene analog 6 (Wnt6) (Figure 1c, p-adjα/β < 10e–3, log2FCα/β > 1) and CyclinE (CycE) (Figure 3d, log2FCα/β > 0.8). Consistent with the higher enhancer activity of the Stat92E eGRN in wound α compared to β (Figure 3a and b), we find a significant overlap between the α marker genes and a study that identified JAK/STAT as coordinating cell proliferation during wing disc regeneration (GSEA, NES = 1.292) (Katsuyama et al., 2015). We also find evidence for an upregulation of the pro-regenerative marker Ets at 21C (Ets21C) (Worley et al., 2022), with a combined enrichment of its gene expression (p-adj < 10e–3, log2FC > 8) and an enrichment of the Ets21C binding motif in cluster α (cistarget, NES = 3.35). In line with Ets21C upregulation, we integrated our data with the study from Worley et al., 2022 that focused on disc regeneration and found strong similarities between their reported regenerative blastemas and our wound population α (Figure 3—figure supplement 2). These results demonstrate that pro-proliferative and pro-regenerative characteristics are specific to the α population in the wound. In contrast, cells from cluster β do not show clear proliferative nor regeneration markers. Genes found upregulated in β compared to α include genes associated with innate immunity (Dorsal-related immunity factor), glutathione metabolism (Glutamate-cysteine ligase catalytic subunit, Glutathione S transferase D9)(p-adjβ/α<10e–3, log2FCβ/α > 2), cell migration signaling (e.g rho1, slow border and pvf1, p-adj <10e–3), response to irradiation (e.g. inverted repeat-binding protein, p-adj <10e–3) and negative regulation of cell cycle (e.g. growth arrest and DNA damage-inducible 45, tribbles). Among these markers, tribbles (trbl) and slow border cells (slbo) encode a known repressor and a target of the JAK/STAT pathway in border cell migration, respectively (Berez et al., 2020; Dobens et al., 2021; Harris et al., 2020; Rørth et al., 2000; Starz-Gaiano et al., 2008). The cells expressing these markers are located at the center of the wounded pouch, despite their lack of pouch-specification markers (Figure 3d). These results suggest that cluster β contains cells derived from the wild-type, rn-expressing domain that have de-differentiated and hence no longer express wing disc markers. In addition to their loss of wing fate, the cells from cluster β show interesting similarities with cellular senescence. Indeed, a recent study in wounded wing imaginal discs associate the emergence of senescent-like cells with the presence of markers of stress-response (jra, kay), DNA-damage response (gadd45), paracrine signaling (pvf1), and cell cycle stalling (trbl), as observed in our population β (Cosolo et al., 2019; Jaiswal et al., 2022). Precisely, one marker of our cluster β population, tribbles (trbl), was found to be partly responsible for cell cycle stalling in wounded cells. Nevertheless, it still remains unclear whether these wounded cells will later die or whether they will revert to the pouch disc fate after wounding (Jaiswal et al., 2022). Hence, we refer here to cellular senescence and associated secretory phenotype in terms of gene expression programs rather than evidence of a terminal cell cycle arrest. The slbo expression in population β further corroborates the proximity with senescence-associated secretory phenotype (SASP) as slbo is homologous to the human C/EBPB homodimer, a major mediator of oncogene-induced senescence (Kuilman et al., 2008; Lee et al., 2010a; Lee et al., 2010b; Reactome ID R-DME-2559582). Together with the lack of pouch markers, our results suggest that wound-response population β corresponds to cells that have lost their initial pouch fate, arrested cycling and express classical senescence markers. The senescent population is characterized by C/EBP eGRN activity By examining eGRN activities, we find several TFs with a strong specificity score for the senescent population (β), namely the heterodimer partners inverted repeat binding protein 18 kDa (irbp18) and xrp1, slbo, vrille (vri), and sox box protein 14 (sox14) (top 6% RSS, Figure 3a). Among them, the two bZip TFs irbp18 and vri are significantly upregulated in the senescent population compared to the proliferative population (α) (p-adj <2 * 10e–3, log2FCβ/α>2.4) and share similarities in their TF binding motifs with the mammalian, senescence-associated C/EBP proteins (Figure 3b; Blanco et al., 2020). The Irbp18 eGRN (extended category, see method) is composed of 149 predicted target regions (80 wound-specific and 11 senescent-specific regions) and 116 target genes (52 markers of senescent population, p-adj <5 * 10e–3, log2FC > 1.5) including slbo, vri, ets at 98B (ets98B), socs36E, head involution defective (hid), growth arrest and DNA damage-inducible 45 (gadd45) and p53. The presence of these TFs in the senescent cluster (β) is further corroborated by the significant enrichment for C/EBPB-like binding motifs in senescent-specific peaks (NES = 5.25, Figure 3b and c) generated using MACS2 bdgdiff (Zhang et al., 2008a). Our eGRN approach has an important limitation, namely that TFs can be identified only if their expression co-varies with chromatin accessibility and target gene expression. A key TF that remains undetected in the eGRN approach is Scalloped (Sd). This homolog of mammalian TEAD factors is the effector TF of the Hippo signaling pathway (Zhang et al., 2008b). Despite the absence of Sd mRNA upregulation, nuclear Sd protein may increase in the senescent population since we find a significant enrichment for the Sd motif in the senescent-specific accessible regions (NES = 4.89). In fact, the top enriched motif in these senescent-specific regions is a C/EBP-TEAD dimer motif (NES = 7.48, Figure 3c), notably found in the enhancer of CG13024, a strong marker of both senescent and rasV12 scrib-/- cells (Figure 3—figure supplement 3, see next section). Our results suggest that the eGRN activity of C/EBP orthologs, such as the Irbp18-Xrp1 heterodimer, is a strong marker of cellular senescence. To confirm the activity of C/EBP eGRNs in our wound populations, we used bulk RNA sequencing to compare gene expression changes induced by a tem poral overexpression of the short Xrp1 isoform (Xrp1-S) in the developing wing imaginal discs using rnts >driver (rnts >Xrp1S) in comparison to control discs (rnts>crossed to w1118; see ‘Materials and methods’). The 200 significantly upregulated genes (DEseq2, logFC > 1.5, p-adj < 5.10e–3) in response to Xrp1 overexpression were strongly enriched for markers of cellular senescence (Reactome Pathway Database, Gillespie et al., 2022) and α and β wound populations. As an additional validation of our eGRN construction, we also find the inferred target gene sets of C/EBP eGRNs significantly enriched among the upregulated genes (Figure 3e, Figure 3—figure supplement 4). Given that these eGRNs target multiple markers of cell migration, cell cycle stalling, and DNA damage response, we hypothesize that C/EBP signaling may be responsible for establishing a state of cellular senesc