Intestinal stem cell overproliferation resulting from inactivation of the APC tumor suppressor requires the transcription cofactors Earthbound and Erect wing

Wnt/β-catenin signal transduction directs intestinal stem cell (ISC) proliferation during homeostasis. Hyperactivation of Wnt signaling initiates colorectal cancer, which most frequently results from truncation of the tumor suppressor Adenomatous polyposis coli (APC). The β-catenin-TCF transcription complex activates both the physiological expression of Wnt target genes in the normal intestinal epithelium and their aberrantly increased expression in colorectal tumors. Whether mechanistic differences in the Wnt transcription machinery drive these distinct levels of target gene activation in physiological versus pathological states remains uncertain, but is relevant for the design of new therapeutic strategies. Here, using a Drosophila model, we demonstrate that two evolutionarily conserved transcription cofactors, Earthbound (Ebd) and Erect wing (Ewg), are essential for all major consequences of Apc1 inactivation in the intestine: the hyperactivation of Wnt target gene expression, excess number of ISCs, and hyperplasia of the epithelium. In contrast, only Ebd, but not Ewg, mediates the Wnt-dependent regulation of ISC proliferation during homeostasis. Therefore, in the adult intestine, Ebd acts independently of Ewg in physiological Wnt signaling, but cooperates with Ewg to induce the hyperactivation of Wnt target gene expression following Apc1 loss. These findings have relevance for human tumorigenesis, as Jerky (JRK/JH8), the human Ebd homolog, promotes Wnt pathway hyperactivation and is overexpressed in colorectal, breast, and ovarian cancers. Together, our findings reveal distinct requirements for Ebd and Ewg in physiological Wnt pathway activation versus oncogenic Wnt pathway hyperactivation following Apc1 loss. Such differentially utilized transcription cofactors may offer new opportunities for the selective targeting of Wnt-driven cancers.


Introduction
The evolutionarily conserved Wnt/β-catenin signal transduction pathway directs fundamental cellular processes across metazoans, whereas deregulation of this pathway is associated with numerous human congenital disorders and cancers [1,2]. In the absence of Wnt exposure, βcatenin, a key transcription coactivator, is phosphorylated and targeted for proteasomal degradation by a "destruction complex" comprised of the scaffold protein Axin, the tumor suppressor Adenomatous polyposis coli (APC), and two kinases: glycogen synthase kinase 3 (GSK3) and casein kinase 1α (CK1α). Wnt stimulation inactivates the destruction complex and thereby stabilizes β-catenin, which subsequently translocates to the nucleus and interacts with the DNA-binding transcription factor T-cell factor (TCF) to regulate Wnt target genes [3][4][5].
Through a forward genetic modifier screen for suppressors of Apc1 in the Drosophila retina, we identified Earthbound1 (Ebd1) and Erect wing (Ewg) as context-specific transcription cofactors in the Wingless pathway [53,54]. Ebd1, a member of a protein family containing Centromere Binding Protein B (CENPB) DNA binding domains, physically associates with and bridges β-catenin/Armadillo (Arm) and TCF, thereby promoting the formation and stability of the β-catenin-TCF complex and the recruitment of β-catenin to chromatin [53]. Ewg is a DNA binding transcriptional activator that shares DNA binding specificity with its human homolog, Nuclear Respiratory Factor-1 (NRF-1) [54][55][56][57]. We found that Ewg is a physical and functional partner of Ebd1 that promotes the recruitment of Ebd1 to specific chromatin sites [54]. We postulated that recruitment of Ebd1 to chromatin by Ewg enhances the transcriptional activity of the β-catenin-TCF complex, thus promoting Wingless signaling.
Herein, we report that these two Wnt pathway transcription cofactors have distinct functions in the Wnt-directed regulation of the adult Drosophila intestine. Similar to the mammalian intestine, the adult Drosophila midgut undergoes rapid turnover and is replenished by intestinal stem cells (ISCs) [58,59]. We find that both Ebd and Ewg are required for all major consequences of Apc1 inactivation in the adult midgut: the hyperactivation of Wingless target genes, excess number of ISCs, and hyperplasia of the epithelium. By contrast, during intestinal homeostasis, Ebd is essential for the Wingless-dependent control of ISC proliferation, whereas Ewg is dispensable. These studies identify transcriptional cofactors that are differentially required for Wnt signaling in physiological conditions versus the pathological states resulting from hyperactivation of the pathway, providing potential selectivity for therapeutic strategies that target Wnt-driven cancers.

Apc1 is essential for both the development and homeostasis of the adult midgut
As the severe intestinal defects present in Apc1 mutants were readily detected as early as two days of adulthood, we hypothesized that these phenotypes arise during formation of the adult midgut. To test this hypothesis, we examined the Apc1 mutant gut epithelium shortly after eclosion. Strikingly, compared with the age-matched controls, the midguts of recently eclosed Thus, an excess number of progenitor cells are present prior to eclosion, whereas the disruption of both EC structure and epithelial architecture arise after eclosion. To trace the initial requirement for Apc1, we examined midguts earlier in their development. The adult intestine is derived from the larval gut, but undergoes major histolysis and reformation during pupation [80][81][82]. Therefore, we examined the gut epithelium of Apc1 mutant third instar wandering larvae, the developmental stage that immediately precedes formation of the adult gut. Notably, supernumerary adult midgut progenitors (AMPs) [80] were not detected in the Apc1 mutant larval guts (S4A-S4B" Fig) and the epithelial structure was normal (S4C-S4D" Fig). Thus, loss of Apc1 initiates intestinal defects during formation of the adult gut during pupation, and these defects increase in severity during adulthood.
To further test the temporal and cell-specific requirements for Apc1 in the midgut, we utilized the temperature-sensitive progenitor cell driver (esg ts : esg-Gal4 tub-Gal80 ts UAS-GFP) for RNAi-mediated Apc1 knockdown in ISCs and EBs either during formation of the adult gut or during adult gut homeostasis. Of note, dramatic increases in the progenitor cell number were knockdown resulted in increased ISC proliferation. Together, these results indicate that Apc1 is required in stem/progenitor cells during both adult gut development and homeostasis, supporting our observations in Apc1 null mutants (Fig 1 and S1-S4 Figs) and previous reports [72,73].
Apc1 prevents the constitutive activation of Wingless target genes in the adult midgut Both the initiation and sustained growth of human colon cancers harboring APC mutations rely on the aberrant activation of Wnt target genes [10,15,16,18,[20][21][22][23][24][25][26][27][28][29][30][31]. To examine whether loss of Drosophila Apc1 also induces the aberrant activation of Wingless target genes in the midgut, we tested three transcriptional reporters for direct target genes of Wingless signaling: frizzled3 (fz3), notum, and naked cuticle (nkd) [84][85][86][87][88]. The Drosophila midgut, like its mammalian counterpart, is subdivided into compartments with distinct histology, gene expression, and physiological functions (Fig 2A) [89][90][91]. The activation of Wingless signaling is graded along the length of each intestinal compartment; Wingless target genes are activated at high levels at intestinal compartment boundaries and at lower levels within compartments as a function of distance from the boundary [89,92]. We found that inactivation of Apc1 resulted in strong ectopic expression of notum-lacZ [93,94] [99]. Even at compartment boundaries, where Wingless pathway activity is normally at its peak, a further fivefold increase in fz3-RFP levels was present in Apc1 null mutant clones as compared to the surrounding control tissue (Fig 2J-2L; quantification: Fig 2M). These findings indicate that Apc1 is required to prevent the constitutive activation of Wingless target genes at both compartment boundaries and within compartments of the midgut.

Apc1 inactivation results in extensive deregulation of gene expression
We sought to determine the extent to which Apc1 loss deregulates gene expression. Using Affymetrix microarrays, we found that the expression of approximately 1000 genes, which can be grouped in broad categories, was either up-or down-regulated by more than twofold in Apc1 mutant midguts as compared to wild-type controls (GEO database: GSE99071; S9 Fig). The changes in expression of selected up-or down-regulated genes were validated by real-time quantitative PCR (Fig 3 and S10 Fig). Consistent with the Wingless target gene reporter analysis described above (Fig 2 and S7 and S8 Figs), fz3, notum and nkd transcription were activated significantly in Apc1 mutants (Fig 3A and S10A Fig). Furthermore, a previous study identified Janus kinase/signal transducer and activator of transcription (Jak/Stat) and Epidermal growth factor receptor (Egfr) signaling as key mediators of ISC hyperproliferation in Apc1 mutants [73]. Indeed, the expression of both unpaired 3 (upd3), a ligand of the Jak/Stat pathway, and Socs36e, a downstream target gene of this pathway, were induced upon loss of Apc1 ( Fig 3A  and S10A Fig). Furthermore, the expression of vein (vn) and spitz (spi), two ligands of the Egfr pathway, was also increased (Fig 3A and S10A Fig). Together, these results suggest that loss of Apc1 results in an extensive deregulation of gene expression.   Transcription cofactors Earthbound and Erect wing mediate intestinal defects due to APC inactivation Both Ebd and Ewg are required for the excess progenitor cells and intestinal epithelial hyperplasia that result from Apc1 inactivation In a forward genetic screen in the Drosophila retina, we previously identified both Ebd1 and Ewg [53,54] as novel suppressors of Apc1. Furthermore, we found that these two proteins function as transcriptional cofactors that physically interact both with each other and with Arm/ Tcf to promote the context-specific activation of Wingless signaling during pupal muscle development [53,54]. However, the limited genetic tools available to analyze Wnt signaling in pupal muscle restricted our ability to identify the roles of Ebd and Ewg in physiological versus pathological Wnt signaling. Herein, we sought to overcome this obstacle by utilizing powerful in vivo assays in the Drosophila intestine.
First, we sought to determine whether Ebd and/or Ewg are required for the phenotypic consequences of Apc1 inactivation in the intestine by examining ebd Apc1 and ewg Apc1 double mutants. Strikingly, the three major defects in the midguts of Apc1 mutants were largely suppressed upon inactivation of either ebd or ewg. First, the numbers of progenitor cells in ebd Apc1 or ewg Apc1 double mutants were similar to those in controls ( Both Ebd and Ewg are required for the aberrant activation of Wingless target genes in the midgut resulting from Apc1 inactivation To determine whether Ebd or Ewg are required for the aberrantly high expression of Wingless target genes that results from Apc1 inactivation, we examined the expression of notum-lacZ, fz3-RFP and nkd(UpE2)-lacZ in ebd Apc1 and ewg Apc1 double mutants. Strikingly, upon loss of either ebd or ewg, the hyperactivation of all three Wingless pathway reporters in Apc1 mutant midguts was reduced nearly to control levels ( Fig 2B-2E, Fig 2F-2I, S7 and S8 Figs). Thus, not only Arm/β-catenin and TCF, but also Ebd and Ewg are required for the aberrantly increased activation of Wingless target genes in Apc1 mutant midguts.
Both Ebd and Ewg are required for the extensive deregulation of transcription resulting from Apc1 inactivation in the midgut We sought to test whether Ebd and Ewg are required for the hyperactivation of only a subset of direct Wingless target genes or also have broader effects in the extensive deregulation of gene expression that occurs in Apc1 mutants. Therefore, we analyzed the expression of genes that are selectively up-or down-regulated genes by Apc1 inactivation in either ebd Apc1 or ewg Apc1 double mutants by real-time quantitative PCR (Fig 3 and S10 Fig). Of note, the transcriptional deregulation resulting from loss of Apc1 was rescued in ebd1 Apc1 or ewg Apc1 double mutants for all genes analyzed (Fig 3 and S10 Fig). These findings provide further evidence that the aberrant transcriptional response in Apc1 mutant midguts requires both Ebd and Ewg.
Ewg is a known sequence-specific DNA binding protein [54][55][56][57]. Therefore, we sought to determine whether consensus Ewg DNA binding sites are present in the enhancers of genes deregulated by Apc1 loss. As the Wingless target gene reporters notum-lacZ, nkd(UpE2)-lacZ, and fz3-RFP are each hyperactivated in an Ewg-dependent manner following Apc1 loss, and the enhancers within these reporters are well-characterized, we searched for potential Ewg and TCF binding sites in these enhancers. The transcriptional enhancers that drive expression of both the notum-lacZ and nkd (UpE2)-lacZ reporters, which are 2.2 kb [93,94] and 0.6 kb [97], respectively, are directly bound and regulated by TCF through distinct pairs of core consensus sites (SSTTTGWWSWW) and Helper sites (GCCGCCR) [5,[96][97][98]100] (S11A-S11C Fig). We identified similar TCF core consensus binding sites and Helper sites in the 2.3 kb enhancer of the fz3-RFP transgene [95] (S11D Fig). In addition, we found that the fz3 enhancer contains an Ewg consensus binding site (GCGCABGY) [54][55][56][57] (S11A and S11D Fig), and that this site is conserved among sequenced Drosophila species (S12 Fig) [101]. In contrast, neither the notum nor the nkd enhancer contains an Ewg consensus binding site (S11A-S11C Fig). Therefore, these findings suggest that the hyperactivation of at least some Wingless target genes in Apc1 mutants may not require direct binding of Ewg to DNA, or alternatively, that Ewg may bind non-consensus DNA sites upon Apc1 inactivation.
Ebd and Ewg are required in progenitor cells to mediate the defects resulting from Apc1 loss RNAi-mediated knockdown of Apc1 specifically in progenitor cells phenocopies the supernumerary progenitors observed in Apc1 null mutants (S5 Fig). Therefore, we hypothesized that Ebd and Ewg act in progenitor cells to mediate the phenotypic consequences of Apc1 loss. To test this hypothesis, we used RNAi-mediated knockdown to concomitantly reduce both Apc1 and ebd or ewg in progenitors. The aberrant increase in progenitor cell number resulting from Apc1 knockdown was largely suppressed upon simultaneous knockdown of either ebd or ewg ( Fig 4A-4H'; quantification: Fig 4I). Based on these findings, we conclude that Ebd and Ewg are required in progenitors to mediate the gut defects resulting from Apc1 loss.

Ebd1 promotes Wingless target gene activation in the adult midgut under physiological conditions, whereas Ewg is dispensable
Our findings indicate that both Ebd1 and Ewg are required for the aberrantly increased expression of the Wingless target genes resulting from Apc1 loss in the adult midgut (Fig 2 and Fig 3; S7, S8 and S10 Figs). To determine whether Ebd1 and Ewg also promote Wingless target gene expression in the adult midgut under physiological conditions, we analyzed the expression of the Wingless target gene reporter notum-lacZ. Under basal conditions, notum-lacZ expression peaks at both the foregut/midgut ( Fig 2B) and the midgut/hindgut boundaries ( Fig 5). This expression of notum-lacZ is completely dependent on Wingless pathway activation, as revealed by its loss in fz Dfz2 double null mutant [102] or dsh null mutant clones [103] (Fig 5A-5D"). We found that in many, but not all ebd1 null mutant clones, notum-lacZ expression was eliminated (Fig 5E-5F"), providing evidence that Ebd1 promotes the activation of Wingless target genes not only in hyperactivated states, but also during homeostasis in the adult midgut. In contrast, ewg null mutant clones resulted in no detectable reduction in the expression of notum-lacZ ( Fig 5G-5H"). In addition, ewg null mutant clones also did not affect the expression of the other two Wingless target gene reporters, fz3-RFP (S13A-S13B" Fig) or nkd-lacZ (S13C-S13D" Fig), suggesting that Ewg is dispensable for Wingless target gene activation in the adult midgut under physiological conditions. Thus, these findings indicate although both Ebd and Ewg are essential for the hyperactivation of Wingless signaling upon Apc1 inactivation, only Ebd is required for Wnt pathway activation during intestinal homeostasis, whereas Ewg is dispensable.

Ebd is required for adult intestinal homeostasis, whereas Ewg is dispensable
Wingless pathway activation is crucial for the maintenance of adult midgut homeostasis [89,92]. We thus sought to determine whether Ebd and Ewg are required for this process. To test whether Ewg has a role during adult midgut homeostasis, we first analyzed ewg P1 , a hypomorphic allele containing an ewg missense mutation ( [54]; note that complete inactivation of ewg results in embryonic lethality). In comparison with controls, ewg P1 midguts contained comparable numbers of progenitor cells (marked by esg>GFP) and displayed normal epithelial architecture ( Fig 6A-6B'). Thus, although this allele revealed that Ewg is crucial for the hyperactivated Wingless signaling and intestinal hyperplasia that results from Apc1 inactivation in the adult midgut (Figs 1-3 and S1, S2, S7, S8 and S10 Figs), it exhibits no detectable defects under physiological conditions. To further reduce the level of ewg activity, we examined the midguts of flies transheterozygous for the null allele ewg 2 and the hypomorphic allele ewg 1 , which is the most severe viable combination of ewg alleles (all mutant flies exhibit "erect wing" defects) [54,55]. Notably, no excess progenitors were observed in ewg 2 /ewg 1 transheterozygotes (Fig 6C and 6D; quantification: Fig 6E). These results suggested that consistent with our observation that Ewg is dispensable for physiological Wingless target gene activation, Ewg does not have a role in Wingless-dependent adult intestinal homeostasis. Gut cell type specific RNA-seq under homeostatic condition revealed previously [104] that expression of Ewg is very low during intestinal homeostasis, while Ebd1 is expressed in all gut cell types ( Fig 6F). Thus, the possibility arose that loss of Apc1 induces overexpression of ewg and this might explain why Ewg is essential for hyperactivated Wingless signaling but dispensable for physiological signaling. However, RT-qPCR of ewg expression revealed a less than 2-fold increase in Apc1 mutants (Fig 6G), suggesting that an increase in Ewg levels is unlikely the mechanism by which Ewg mediates the consequences of Apc1 loss in the adult midgut.
We next analyzed whether Ebd is required during adult intestinal homeostasis. In Drosophila, two Ebd proteins, Ebd1 and Ebd2, possess partially redundant functions [53]. To elucidate the function of Ebd1 in the midgut, we compared the intestinal epithelium of control (ebd1/+) to ebd1 240 null mutants [53]. We found that the number of progenitor cells (marked by esg>GFP or combination of Arm/Prospero staining) was significantly increased in ebd1 mutants (S14A-S14F Fig; quantification: S14M). In addition, in contrast to the control midguts, very few of which (8%) displayed chains of progenitors and none of which exhibited clusters of progenitor cells, the majority of ebd1 mutant midguts (60%) contained chains of progenitor cells and 15% exhibited clusters (S14A-S14F Fig, quantification: S14N Fig). To further determine whether this requirement for Ebd1 in the regulation of ISC proliferation occurs during adulthood, we generated marked ebd1 null mutant clones in adults. We found that ebd1 mutant clones were markedly larger than control clones: 44% of ebd1 mutant clones contained more than 4 cells, as compared to 14% of the control clones (S15 Fig). Together, these findings indicate that in contrast to Ewg, Ebd1 is required for intestinal homeostasis during adulthood, resembling other Wingless pathway components [89,92].
We further sought to determine whether the combined inactivation of ebd1 and ebd2 would result in a more severe phenotype than inactivation of ebd1 singly. Staining for esg>GFP revealed that by comparison with ebd1 single mutants, a larger proportion of ebd1ebd2/ebd1 mutant midguts displayed clusters of progenitor cells (42%), and this proportion increased further in midguts homozygous mutant for both ebd1 and ebd2 (58%) (S14G-S14L Transcription cofactors Earthbound and Erect wing mediate intestinal defects due to APC inactivation mutant phenotype. To further differentiate the subtypes of progenitor cells that are deregulated by loss of the Ebd1 and Ebd2 proteins, we examined the ISC-specific marker Delta (Dl) [105] and the EB-specific marker GBE-Su(H)-lacZ in ebd1 ebd2/ebd1 mutants, and detected a significant increase in the number of both cell types as compared to controls (Fig 7A-7L; quantification: Fig 7M and 7N). Furthermore, 33% of ISCs (esg+, GBE-Su(H)-) were associated Transcription cofactors Earthbound and Erect wing mediate intestinal defects due to APC inactivation with an EB (esg+, GBE-Su(H)+) in controls, but this number increased to 78% in ebd1 ebd2/ ebd1 mutants (Fig 7O). Together, both Ebd1 and Ebd2 are required for homeostasis of the Drosophila midgut as their inactivation leads to an aberrant increase in both ISCs and EBs.

Ebd1 non-autonomously prevents aberrant ISC overproliferation
To determine the cell types in which Ebd1 is expressed in the midgut, we immunostained intestines with Ebd1 antibody [53] (S16A-S16A" Fig). Using ebd1 null mutant clones to test the specificity of the Ebd1 antibody, we found that Ebd1 is expressed in enterocytes (S16B-S16C" ' Fig). We also detected Ebd1 in progenitors and EEs, but based on mutant clonal analysis, it was not clear that this staining was above background (S16B-S16C" ' Fig). Therefore we conclude that there is strong Ebd1 expression in ECs, and any Ebd1 expression in progenitors cells or EEs is lower than the detection limit of the Ebd1 antibody. We also tested ebd1-Gal4 lines [53] to drive reporter expression, which revealed that Ebd1 is strongly expressed in ECs, but also detectable at lower levels in progenitors and EEs (S17 Fig).
The activation of Wingless signaling in ECs inhibits the proliferation of ISCs non-autonomously to regulate adult intestinal homeostasis [89,92]. Similarly, we found that an abnormally large number of progenitor cells were clustered around ebd1 240 or ebd1 5 null mutant clones (Fig 8A-8B'; quantification: Fig 8C). The excess progenitor cells present near ebd1 clones resulted from aberrantly increased proliferation, as revealed by the number of pH3 positive cells (Fig 8D). Since Wingless signaling is required specifically in ECs to regulate the proliferation of surrounding ISCs [92], we tested whether Ebd1 functions similarly by reducing ebd1 expression in ECs using RNAi-mediated knockdown with the EC-specific driver Myo1A-Gal4 [83]. As compared with controls, knockdown of ebd1 in ECs resulted in increased numbers of progenitor cells (marked by esg-lacZ or a combination of Arm and Prospero staining) and ISCs (marked by Dl) that were present in chains or grouped in clusters (Fig 8E-8K and S18A-S18F Fig). Furthermore, the number of pH3-positive cells increased upon ebd1 knockdown in ECs, confirming the overproliferation of ISCs (Fig 8L). Moreover, as reported previously for inactivation of other Wingless pathway components [92], increased ISC proliferation was observed when ebd1 expression was disrupted during adulthood, but not during development (S18G-S18I Fig). These results were obtained with three independently derived transgenic ebd1 RNAi lines to rule out the possibility of off-target effects. Together, our findings suggest that loss of ebd1, like that of Wingless pathway components, non-autonomously promotes the proliferation of neighboring ISCs.
As Wingless signaling controls the proliferation of ISCs through the Jak/Stat pathway [92], we examined whether Ebd1 analogously controls Jak/Stat signaling. RNAi-mediated knockdown of ebd1 expression in ECs led to significant increases in the expression of upd2 and upd3, ligands for the Jak/Stat pathway (Fig 9A). In contrast, little increase was detected in the expression of decapentaplegic (dpp) or keren (krn), EC-expressed ligands for the TGF-β and EGF pathway, respectively [106][107][108][109] (Fig 9A). Similarly, expression of puckered (puc) and keap1, target genes for the two major stress response signaling pathways, JNK (c-Jun N-terminal kinase) and Nrf2 (Nuclear factor 2) respectively [110][111][112], was not affected (Fig 9B). Thus, Ebd1 specifically regulates the expression of Jak/Stat pathway ligands in ECs, and could thereby control the activation of Jak/Stat signaling in adjacent ISCs. In support of this idea, we found that RNAi-mediated knockdown of ebd1 in ECs induced expression of Socs36e, a direct target gene of the Jak/Stat pathway [113] (Fig 9B). To further test whether Jak/Stat signaling is activated in ISCs upon loss of ebd1 in ECs, we analyzed the expression of the Jak/Stat pathway reporter, stat-GFP [114]. We found that stat-GFP expression increased markedly in ISCs near ebd1 mutant clones (Fig 9C-9C"; quantification: Fig 9D), indicating that the Jak/Stat pathway was activated non-autonomously upon ebd1 inactivation. To determine whether the ectopic activation of Jak/Stat signaling mediates the overproliferation of ISCs resulting from loss of ebd1, we concomitantly knocked down both upd and ebd1 in ECs using RNAi. Dual knockdown of ebd1, and either upd2 or upd3, reduced ISC proliferation in posterior midguts, as indicated by Dl and pH3 staining (Fig 9E and 9F). Therefore, Ebd1 activity in ECs, like that of Wingless pathway components, prevents the non-autonomous activation of JAK/STAT signaling in neighboring ISCs, and thereby inhibits their proliferation.
The observation that ebd1 inactivation results in ISC overproliferation in physiological conditions, but prevents ISC overproliferation in Apc1 mutants presented us with a paradox. Analysis of the spatial and temporal requirement for Ebd1 provided an explanation for these unanticipated results. The Apc1 mutant phenotype emerges during formation of the adult gut during pupation (S3-S5 Figs and S19A, S19C and S19E Fig), a stage in which ebd1 knockdown has no effect (S18G-S18I Fig). Indeed, the midguts of newly eclosed ebd1 mutants exhibited a similar number of EBs by comparison with the age-matched controls (S19A and S19B Fig;  quantification: S19E Fig). Furthermore, Ebd1 is non-autonomously required in ECs to prevent ISC overproliferation during adult homeostasis (Figs 8 and 9 and S18 Fig), in contrast to its autonomous requirement in progenitor cells for the gut defects resulting from Apc1 loss ( Fig  4). Together, these findings indicate that Ebd1 plays qualitatively different roles in transducing physiological and pathological Wingless signaling, which are temporally and spatially distinct.
In summary, our analysis of two transcription cofactors, Ebd and Ewg, in the Drosophila midgut revealed that both Ebd and Ewg are required for all major consequences of Apc1 inactivation: the hyperactivation of Wingless target genes, excess number of progenitor cells, and epithelial hyperplasia (Fig 10A). By contrast, during intestinal homeostasis, only Ebd, but not Ewg, is essential for the Wingless-dependent control of ISC proliferation (Fig 10B). Together, these findings provide evidence that some context-specific transcription cofactors are differentially required for physiological Wnt pathway activation during homeostasis versus the oncogenic hyperactivation of the Wnt pathway resulting from loss of Apc1, and thus may present opportunities for the therapeutic targeting of Wnt-driven diseases.

Discussion
Our findings indicate that both Ebd and Ewg are necessary for the aberrantly high-level Wnt target gene activation that mediates the consequences of Apc1 loss. These results provide in vivo evidence that the core β-catenin-TCF transcriptional machinery is insufficient for the transformation of intestinal epithelial cells in Apc1 mutants; cooperation of β-catenin-TCF with Ebd and Ewg is also necessary. As Ebd and Ewg are known to physically interact with each other and with β-catenin, we postulate that the Ebd-Ewg complex acts with β-catenin-TCF to activate the high-level transcription of Wnt target genes in ISCs in Apc1 mutants. Moreover, Ebd, but not Ewg, is required for the Wnt-dependent control of ISC proliferation during homeostasis. Together, these studies reveal that transcription cofactors with contextspecific roles in Wnt target gene activation under physiological conditions can be co-opted to function with β-catenin-TCF to promote the global hyperactivation of Wnt target genes following APC loss.
Ebd and Ewg are essential mediators of the hyperactivated Wnt target gene expression and intestinal epithelial defects that result from Apc1 inactivation.
In both mammals and Drosophila, two APC paralogs have partially redundant roles that are dependent on cell context. However, as in humans and mice, inactivation of a single Transcription cofactors Earthbound and Erect wing mediate intestinal defects due to APC inactivation Drosophila APC homolog alone is sufficient to induce ISC overproliferation, as well as defects in intestinal epithelial cell adhesion, cell polarity, and intestinal architecture that recapitulate many aspects of human colorectal cancer. Furthermore, similar to inactivation of human and mouse APC, loss of Drosophila Apc1 results in aberrantly high levels of Wnt target gene expression in the intestine. Our analysis of one Wnt target gene reporter reveals a fivefold increase in its expression in Apc1 mutant cells compared to wild-type cells even at intestinal compartment boundaries, which are the sites with the highest levels of Wingless protein and the highest activation of physiological Wingless signaling in the adult gut. Overall, the expression of approximately 1000 genes is significantly deregulated in Apc1 mutant guts. These results provide evidence that inactivation of Drosophila Apc1 singly results in intestinal hyperplasia and Wingless target gene hyperactivation, in a manner analogous to the pathological consequences that result from loss of mammalian APC.
Our findings also reveal that Ebd and Ewg mediate the intestinal epithelial defects and oncogenic levels of Wnt target gene expression that result from loss of Apc1. In addition, we find that although Ewg is a known sequence-specific DNA-binding protein and is required following Apc1 loss for the high level expression of the Wnt target genes fz3, nkd, and notum through their well-characterized enhancers, Ewg consensus DNA binding sites are present in only one of these three enhancers. Therefore, the direct association of Ewg with DNA might not be required for Ewg's role in the hyperactivation of Wnt signaling, or Ewg might also interact with non-consensus binding sites. Thus, in these contexts, Ebd might access DNA through its own CENPB-type DNA binding domains [53,115]. Alternatively, these findings raise the possibility that Ewg and Ebd access chromatin via protein-protein interactions instead of direct association with DNA. A precedent for this type of mechanism was documented previously for Fushi tarazu, which activates transcription even when its DNA-binding homeodomain is deleted, through interaction with the DNA-binding transcription factor Paired [116,117].

Ebd, but not Ewg, is essential for Wingless-dependent control of ISC proliferation during homeostasis
In the Drosophila intestine, activation of Wingless signaling in ECs non-autonomously restricts the proliferation of surrounding ISCs during homeostasis. Our findings herein suggest that this process requires Ebd. We further find that Ebd is also required for the autonomous hyperactivation of Wingless signaling in ISCs that results in their overproliferation following Apc1 loss. This novel finding reveals that Ebd is required for Wnt signaling during both normal homeostasis of the intestine and its aberrant hyperplasia, in addition to Ebd's previously documented roles in muscles and neurons [53,54]. Similar to that in other tissues, the role of Ebd in the intestine is context-specific, as not all Wnt-mediated processes are dependent on Ebd; Ebd promotes the Wnt-mediated regulation of ISC proliferation during homeostasis, but is dispensable for the Wnt-dependent specification of cell fate near intestinal Transcription cofactors Earthbound and Erect wing mediate intestinal defects due to APC inactivation compartment boundaries [92]. Conversely, Ewg has no observed role in either of these Wntdependent processes. Thus, Ebd functions in an Ewg-independent manner in the adult gut under physiological conditions. These results suggest that Ebd and Ewg do not always function in a complex, and that recruitment of Ebd to chromatin by Ewg [54] is context-specific.
Context-specific transcription cofactors in Wnt pathway hyperactivation: Implications for new therapeutic strategies in Wnt-driven cancer Based on our findings, we propose that mechanistic differences in the Wnt transcriptional machinery underlie target gene activation in physiological versus pathological states. These novel distinctions likely underlie the markedly increased expression of Wingless target genes in the hyperactivated state that results from Apc1 inactivation. Analogously, the mammalian transcription cofactors Pygo and BCL9 also form a complex that enhances target gene Transcription cofactors Earthbound and Erect wing mediate intestinal defects due to APC inactivation activation by β-catenin-TCF in the Wnt hyperactivated state. Neither mammalian Pygo nor BCL9 is required for Wnt-mediated ISC proliferation or maintenance during homeostasis, but both promote Wnt target gene expression in colorectal cancer. Most targeted therapies under investigation disrupt Wnt signaling not only in tumors, but also in normal tissues. Thus, the discovery that transcription cofactor complexes, such as Ebd-Ewg in Drosophila or Pygo-BCL9 in mammals, are essential for supraphysiological signaling but dispensable for most Wnt-dependent physiological processes may distinguish tumors from normal tissues and provide selectivity for therapeutic strategies that target Wnt-driven diseases.
Our findings suggest that the human homologs of Ebd and Ewg might provide novel therapeutic targets for the treatment of Wnt-driven cancers. Jerky (also known as JRK or JH8; [118][119][120][121][122][123]), the human homolog of Ebd, rescues ebd mutant phenotypes when expressed in Drosophila [53] and promotes the aberrant increase of both cell proliferation and β-catenin-TCF mediated transcription in colon cancer cell lines [53,124,125]. Moreover, aberrantly high levels of Jerky are present in several carcinomas, including colon, breast, and ovarian serous cystadenocarcinoma. Elevated Jerky expression is associated with increased β-catenin nuclear localization and the aberrantly increased expression of Wnt target genes in human colorectal tumors [125]. A possible role for Nuclear Respiratory Factor 1 (NRF1), the human homolog of Ewg, in Wnt signaling awaits future investigation. Together, these findings suggest that inhibition of Jerky, NRF1, or their physical interaction may provide promising therapeutic strategies for colorectal cancer.
Adult fly intestines were dissected in PBS and fixed in 4% paraformaldehyde in PBS for 45 minutes at room temperature. For Delta staining, intestines were fixed in 8% paraformaldehyde, 200mM Na cacodylate, 100mM sucrose, 40 mM KOAc, 10 mM NaOAc, and 10mM EGTA for 20 minutes at room temperature [130]. Tissues were then washed with PBS, 0.1% Triton X-100, followed by incubation in PBS, 0.1% Tween-20 and 10% BSA for 1 hour at room temperature. The samples were then incubated with primary antibodies at 4˚C overnight in PBS, 0.5% Triton X-100. Samples were stained with secondary antibodies for 2 hours at room temperature. Specimens were stained with DAPI (2μg/ml) and mounted in Prolong Gold (Invitrogen). To assess the gut layers, specific mounting set-ups were performed according to a protocol from the Micchelli lab [74]. Larval guts were immunostained in the same way except that wandering third instar larvae were fixed in 4% paraformaldehyde in PBS for only 20 minutes and were incubated with primary antibodies in PBS, 0.1% Triton X-100. Fluorescent images were obtained on a Nikon A1RSi confocal microscope except those in (S6 Fig), which were captured on a Zeiss LSM 780 confocal microscope. Images were processed using Adobe Photoshop software.

Clonal analysis
Mitotic clones were generated using the MARCM system [99]. Developmental clones were induced in third instar larvae by a single 2-3 hour heat shock at 37˚C and examined 1 to 2 days after eclosion. To generate clones in the adult gut, flies were heat shocked for 30 minutes in a 37˚C water bath four days post-eclosion. After heat shock, flies were maintained at 25˚C for five days before analysis. For quantification of clone size, flies were maintained at 25˚C for 14 days post heat shock and only clones in the posterior midguts were included in the analysis.

Transgene expression using temperature-sensitive Gal4 and flip out systems
To induce temporal knockdown in ISCs, control or specific RNAi lines were crossed to the esg ts (esg-Gal4 tubGal80 ts UAS-GFP/CyO) driver. For knock down during development, crosses were set up at 18˚C and shifted to 29˚C 6 days later (during the second instar larval stage). Progeny of desired genotypes were dissected 2-3 days after eclosion. For knock down during adulthood, crosses were maintained at 18˚C until eclosion, and progeny of desired genotypes were then shifted to 29˚C for 14 days before analysis.
To induce temporal knockdown in ECs, RNAi experiments were performed using Myo1A-Gal4 with the temperature-sensitive Gal4 repressor, Gal80 ts . Crosses were maintained at 22˚C and on the day of eclosion, progeny of desired genotypes were shifted to the restrictive temperature (29˚C) for 7 days.
To induce temporal knockdown using the "escargot flip out" system (esg ts F/O: esg-Gal4 tub-Gal80 ts UAS-GFP; UAS-flp Act>CD2>Gal4), crosses were maintained at 18˚C and 3-5-day-old progeny of the desired genotypes were shifted to 29˚C. The marked esg + cell lineages were analyzed 14 days later.

Quantification and statistics
For quantification of ISCs, flies were stained with anti-Delta (Dl) and anti-Prospero antibodies. Images of the midgut R5a region [89] were obtained with a 60x lens and the total number of Dl-positive cells in a field of 0.051mm 2 was counted. Similarly, progenitor cells inside a defined field were quantified by counting esg>GFP, esg-lacZ, or small cells with strong Arm staining and absence of Prospero staining. EBs inside a defined field were quantified by counting GBE-Su(H)-lacZ positive cells. For quantification of pH3-positive cells, the total number of pH3-positive cells in the posterior midgut of the indicated genotypes was counted. For quantification of pH3-positive cells near MARCM clones, the number of pH3-positive cells in a field of 4000 µm 2 around the clone was counted.
For quantification of intestinal structure, 40x Z-stack confocal images of posterior midguts of desired genotypes were acquired. The maximum number of epithelial layers and maximum epithelial height were measured using NIS-elements software (Nikon).
Quantification of GFP intensity in "esg ts flip out" guts was performed by measuring overall GFP intensity within two areas per posterior midgut and normalizing that value by the total number of cells in the field.
Quantification of fz3-RFP intensity was performed with ImageJ (NIH). For each clone examined (total of 57 clones derived from more than 20 guts), intensities of three separate areas within the clone and areas of identical size outside the clone were measured. The average intensities of the three areas were compared to the average intensities of their control counterparts.
Quantification of stat-GFP intensity was performed using Imaris software (Bitplane). Stat-GFP-positive cells within a field (40μm × 40μm) surrounding an ebd1 mutant clone, or in an equal field at least 50μm away from the ebd1 mutant clone, were identified and their intensity was measured.
All statistical tests were performed using Prism (GraphPad Software, USA).

Microarray
Whole midguts from Canton S (control) or Apc1 Q8 7-day-old females were dissected in nuclease-free PBS and processed for transcriptomic analysis. Total RNA from 30 adult midguts per sample was extracted using Trizol following manufacturer's instructions. Triplicate samples of each of the genotypes were prepared. RNA was sent to the Microarray facility at the University of Manchester where it was used to hybridize Drosophila Affymetrix 2.0 chips. The CEL files were subject to RMA normalization and log2 transformation followed by differential gene expression analysis by the Beatson Institute Bioinformatics department. Microarray data were deposited in the GEO database (GSE99071). GO term analysis was performed via "PANTHER GO-slim" [131].

RT-qPCR validation
Whole midguts from 15-20 flies of desired genotypes were dissected in PBS and total RNA was extracted using the RNA miniprep kit (Zymo research). The RNA was subsequently treated with RQ1 DNase (Promega). 1 μg of RNA was reverse transcribed using pdT15 primers and M-MLV reverse transcriptase (Invitrogen). Expression level of candidate genes was quantified using the StepOne Real-time PCR system (Life Technologies) with SYBR green (Life Technologies/Biorad). RNA extraction of three biologically independent samples was performed. Two independent repeats are presented in Fig 3 and S10 Fig, respectively, as mean fold change relative to the internal control (rpl32), with standard deviation. The primers used are listed in Table 1.