Stuxnet fine-tunes Notch dose during development using a functional Polycomb response element Free

Notch signaling is evolutionarily conserved, and it controls many important biological processes in metazoan development and adult homeostasis (Henrique and Schweisguth, 2019; Sprinzak and Blacklow, 2021). Dysfunction of Notch signaling often leads to birth defects and cancer in humans (Aster et al., 2017; Mašek and Andersson, 2017; McIntyre et al., 2020). Activation of Notch signaling depends on the proteolytic cleavage and direct translocation of the intracellular domain of the Notch receptor (N^icd) into the nucleus, where N^icd binds to transcription factor Suppressor of Hairless [Su(H)] and activates the expression of downstream target genes. Notch signaling activation is dose sensitive to Notch receptors due to direct signaling from the plasma membrane to the nucleus. In Drosophila, heterozygotes carrying one copy of the loss-of-function Notch mutation show a stereotypical notched wing phenotype, whereas females harboring three copies of the Notch gene display the Confluens wing vein phenotype (Artavanis-Tsakonas and Muskavitch, 2010). Similarly, Notch receptor haploinsufficiency and increased Notch gene copy number in mammals are associated with multiple forms of developmental anomalies and cancer (Aster et al., 2017; Mašek and Andersson, 2017). Therefore, the dose of Notch receptors must be tightly controlled to ensure normal development and adult tissue homeostasis.

Modulation of Notch receptor dose occurs at multiple levels. At the transcriptional level, a few transcription factors and co-factors have been shown to bind directly to the Notch loci and control their transcription (Lambertini et al., 2010; Lefort et al., 2007; Koyama et al., 2014; Taranova et al., 2006; Wang et al., 2019; Wu et al., 2005). At the protein level, several E3 ligases (Hori et al., 2004; Jehn et al., 2002; Mukherjee et al., 2005; Qiu et al., 2000; Sakata et al., 2004; Wilkin et al., 2004) and γ-secretases (Cras-Méneur et al., 2009; De Strooper et al., 1999; Struhl and Greenwald, 1999; Wu et al., 2001) have been shown to mediate the ubiquitylation and processing of Notch receptors, thereby regulating their stability or signaling activity. Ubiquitylation-mediated degradation of Notch receptors can be further regulated by other forms of post-translational modifications, including phosphorylation, acetylation and methylation (Foltz et al., 2002; Fryer et al., 2004; Guarani et al., 2011; Hein et al., 2015; Li et al., 2014; Mo et al., 2007; Sjöqvist et al., 2014). Furthermore, O-fucosyltransferase 1 (O-fut1) interacts with the extracellular domain of Notch to facilitate its endocytosis and turnover (Sasamura et al., 2007).

In addition to transcriptional and post-translational regulation, there is growing evidence that epigenetic repression mediated by Polycomb group (PcG) proteins is an integral part of the Notch signaling regulatory network (Acharyya et al., 2010; Felician et al., 2014; Feng et al., 2011; Jin et al., 2017; Loubiere et al., 2016; Martinez et al., 2009; Schwanbeck, 2015). For Notch receptors, PcG protein recruitment has been detected at the Notch loci, highlighting the possibility that the Notch receptor dose may be additionally regulated at the epigenetic level. Specifically, members of Polycomb repressive complex 2 (PRC2) Su(Z)12 and EZH2 bind to the promoter region of the Notch1 and Notch3 loci in cultured mammalian cells, but the effect of the PRC2 recruitment on Notch expression has not been determined (Acharyya et al., 2010; Jin et al., 2017). In the Drosophila eye, Notch expression is elevated when the polyhomeotic (ph-p and ph-d) gene, which encodes a core component of the PRC1, is mutated. This is closely related to the observed deposition of Ph and Pc (another member of the PRC1) in the promoter region of the Notch locus (Loubiere et al., 2016; Martinez et al., 2009). However, PRC1 proteins are known to bind to loci other than Notch, including key members of the JNK, JAK-STAT and Wingless (Wg) signaling cascades (Classen et al., 2009; Loubiere et al., 2016). Given the extensive crosstalk between Notch and the developmental signaling pathways described above (Beira et al., 2018; Feng et al., 2011; Torres et al., 2018), it is uncertain whether the observed effect of PRC1 on Notch expression during fly eye development is direct.

Epigenetic phenomena were first reported in Drosophila, and experimentally validated epigenetic targets all contain cis-regulatory elements called Polycomb response elements (PREs) that recruit PcG proteins to their respective genomic loci (Grossniklaus and Paro, 2014; Kassis et al., 2017; Schuettengruber et al., 2017). As no functional PREs have been detected at the Notch locus, it is necessary to identify physiologically relevant PREs to establish a direct role of PcG in controlling the Notch receptor dose in Drosophila. In this study, we found that Stuxnet (Stx), an upstream epigenetic regulator that controls PRC1 stability and assembly (Du et al., 2016), positively regulates Notch mRNA expression, thus providing us with an excellent opportunity to evaluate the direct contribution of epigenetic regulation of the Notch receptor dose. We showed that Stx is a positive regulator of Notch signaling in multiple developmental processes. It epigenetically controls Notch receptor gene transcription by removing Pc from a previously uncharacterized PRE at the Notch receptor gene locus. Unlike those PcG-binding sites mapped to the Notch locus by chromatin immunoprecipitation sequencing (ChIP-seq) (Ahmad and Spens, 2019; Loubiere et al., 2016; Martinez et al., 2009), this new PRE, named Notch PRE, is responsive to altered Stx and PRC1 activities, providing in vivo evidence that Notch is a bona fide PcG target. Although PcG deposition has been found at many genomic loci through ‘omics’ studies, in vivo evidence supporting the physiological relevance of epigenetic regulation to specific targets is lacking, in part because, in most cases, epigenetic regulation can only fine-tune development, which is often counteracted by the intrinsic genetic and cellular plasticity of the embryo to ensure developmental robustness (De et al., 2016; Mihaly et al., 1997; Ogiyama et al., 2018; Sipos et al., 2007; Xiao et al., 2022). We show that the new Notch PRE identified in our study is functional in vivo. Deleting this PRE in situ from the Notch locus increases Notch mRNA expression, resulting in loss of macrochaetes in the notum and excessive crystal cell differentiation in the lymph gland: two stereotypical phenotypes associated with hyperactivation of Notch signaling. As cell fate specification in macrochaetes and in lymph gland development is highly sensitive to changes in Notch activity, in situ PRE deletion in such systems, which requires precise and robust control, may help to definitively determine the physiological role of epigenetic regulation in development.

To examine whether the upstream epigenetic regulator Stx regulates Notch signaling, we manipulated stx expression in the posterior compartment of the Drosophila wing (Fig. 1A-F″) and found that knockdown of stx by RNAi resulted in loss of marginal tissue in adult wing blades (compare Fig. 1C with 1A), resembling the stereotypical phenotype associated with reduced Notch signaling (Xu et al., 1990). To eliminate off-target effects of RNAi, we used two additional RNAi lines that target different regions in the stx-coding sequence (Fig. S1A). When overexpressed, both RNAi lines led to downregulated Notch signaling in adult wings (Fig. S1B-C′). Furthermore, the wing-notching phenotype observed in stx RNAi flies could be largely rescued by stx overexpression (Fig. S1D,E), using either a GS line that inserts the UAS element upstream of the stx-coding sequence (Toba et al., 1999) or a UAS-stx transgenic line (Du et al., 2016).

To confirm that Notch signaling is indeed regulated by stx, we examined the expression of Notch signaling targets cut and wg, as well as NRE-gfp, a Notch signaling reporter indicating that cells exhibit Su(H)-dependent activation of Notch signaling (Saj et al., 2010), in third instar larval wing imaginal discs. As expected, when stx expression was knocked down by RNAi in the posterior compartment of the wing disc, we observed decreased expression of Cut and Wg (Fig. 1D-D″), and reduced activity of the NRE-gfp reporter (Fig. S1G,G′). Furthermore, a similar wing-notching phenotype and reduced expression of Notch signaling target genes were observed when the YFP protein trap line CPTI-004181, with a yfp cassette inserted in frame at the stx locus and a reported trapping efficiency of 68% (Du et al., 2016), was used for destabilizing endogenous Stx-YFP fusion proteins using the anti-GFP nanobody (Fig. S1I-J′; Caussinus et al., 2012). In contrast to the effects associated with reduced stx activity, overexpression of stx resulted in loss of vein tissue (Fig. 1E), a phenotype associated with increased Notch signaling (Xu et al., 1990), and corresponding increases in Cut and Wg expression, and in NRE-gfp activity (Fig. 1F-F″; Fig. S1H,H′).

To further determine the involvement of stx in Notch signaling, we examined the genetic interactions between stx and various Notch pathway components. We found that the wing-notching phenotype caused by stx RNAi could be further enhanced by loss-of-function alleles of positive regulators of Notch signaling, such as Notch¹, Notch^PL24, Ser¹ and mam⁸ (Fig. S1K-O). Consistently, this stx RNAi phenotype was suppressed by the loss-of-function allele of Suppressor of deltex [Su(dx)], a negative regulator of Notch signaling (Fig. S1P; Wilkin et al., 2004). These data suggest that stx plays a positive role in Notch signaling in wing development.

As the regulation and outcome of Notch signaling often depends on the cellular context (Bray, 2016), we investigated whether stx promotes Notch signaling in tissues other than the developing wing. The specification of Drosophila sensory organ precursors (SOPs) in the developing adult notum and crystal cells in the lymph gland is highly sensitive to fluctuating Notch signaling (Brennan et al., 1999; Jung et al., 2005; Kopan, 1999; Lan et al., 2020; Lebestky et al., 2003). Impaired Notch activity leads to supernumerary SOPs and reduced crystal cell specification, whereas overactivation of Notch signaling results in SOP loss and crystal cell overproduction. The adult scutellum contains four bristles derived from two SOPs, aSC and pSC, in the notum of the larval wing disc (Fig. 1G,H; Hartenstein and Posakony, 1989). We examined the requirement of stx in scutellar bristle development by manipulating stx expression using a patched (ptc)-Gal4 driver whose expressing domain in notum spans aSC and pSC (Fig. 1H; Brennan et al., 1999). Reduced stx expression by RNAi increased the number of scutellar bristles and the number of neuralized (neur)-lacZ-labeled SOPs (Fig. 1I,J,M; Huang et al., 1991). Conversely, increased stx activity prevented scutellar bristle development and SOP specification (Fig. 1K-M). We used lozenge (lz)-Gal4 to regulate stx expression in crystal cells and found that knocking down stx significantly reduced the number of crystal cells labeled by lz>GFP and Notch (Fig. 1O,O′,Q; Lebestky et al., 2000), whereas elevated stx expression enhanced crystal cell specification (Fig. 1P,Q). To rule out the possibility that the decrease in the number of crystal cells caused by stx knockdown is a consequence of induced apoptosis, we examined the protein levels of cleaved Dcp-1 caspase, an indicator of apoptosis, in the lymph glands and found that knocking down stx expression did not induce the cleavage of Dcp-1 (Fig. S1Q-Q″). Based on the above observations, we conclude that stx is a general positive regulator of Notch signaling in Drosophila.

Our previous study showed that Stx promotes proteasomal degradation of the Pc protein, a core component of the epigenetic repressive PRC1 complex, thereby disrupting the assembly and activity of PRC1 (Du et al., 2016). In addition to Pc, Drosophila PRC1 is made up of three other proteins, including Ph, Posterior sex combs (Psc) and Sex combs extra (Sce). As increased expression of the Notch receptor and its ligand Serrate (Ser) but not the other ligand Delta (Dl) was observed in the ph and Psc mutant eye discs (Loubiere et al., 2016; Martinez et al., 2009), we investigated whether Stx positively regulates Notch signaling in the developing wing by regulating the expression of Dl, Ser or Notch. We found that when stx was knocked down in the wing disc, the expression levels of both Notch ligands, Dl and Ser, were not affected (Fig. S2A-D′). However, the expression pattern of Dl changed from two stripes adjacent to the D-V boundary to one stripe overlapping the D-V boundary (Fig. S2E-E″). This phenotype has been observed in wing discs carrying temperature-sensitive Notch (Notch^ts) mutations (de Celis and Bray, 1997), thus suggesting that Notch receptor expression may be affected. We consistently found a significant reduction in the amount of Notch protein in stx knockdown cells (Fig. 2B,B′, compare with 2A). Conversely, when stx was overexpressed, Notch expression increased significantly (Fig. 2C,C′; compare with 2A).

To further determine whether regulation of Notch by Stx occurs at the transcriptional or post-transcriptional level, we performed fluorescence in situ hybridization in the wing disc and found that Notch mRNA was greatly reduced when stx was knocked down (Fig. 2E, compare with 2D). Conversely, overexpression of stx resulted in elevated Notch expression (Fig. 2F, compare with 2D). Regulation of Notch receptor mRNA expression by stx was also quantified by quantitative real-time PCR (qPCR). When stx expression in the wing disc was knocked down by T80-Gal4, Notch mRNA levels were reduced by 80%, whereas increased stx activity produced by MS1096-Gal4 resulted in a fivefold upregulation of Notch mRNA expression (Fig. 2G). Furthermore, we showed that the wing-notching phenotype caused by stx knockdown was largely rescued by overexpression of Notch^FL (Fig. 2H-J). Therefore, we conclude that Stx positively regulates Notch signaling by promoting Notch mRNA expression.

As Stx is known to promote Pc protein degradation (Du et al., 2016), we investigated the possibility of Stx regulating Notch mRNA expression through Pc. As expected, removing a single copy of the Pc was sufficient to rescue the adult wing margin defects caused by reduced stx expression (Fig. 3B, compare with 3A). Consistent with this observation, Notch expression and activation of the Notch signaling targets cut and wg were also largely restored (Fig. 3E-F′, compare with. 3C-D′). Conversely, Notch upregulation induced by stx overexpression was abolished when Pc was co-expressed (Fig. 3H-H″, compare with 3G-G″). These results suggest that Pc is epistatic to stx and that the regulation of Notch mRNA expression by Stx may depend on Pc activity. This inference is further supported by the cell-autonomous enhancement of Notch protein production observed in the loss-of-function Pc^XT109 clones in the wing disc (Fig. 3I-I″).

Increased Notch expression is observed when the ph or Psc function of the Drosophila eye disc is impaired (Loubiere et al., 2016; Martinez et al., 2009). Therefore, we asked whether the regulation of Notch by Stx also depends on other components of the PRC1 complex. As expected, removal of one copy of ph or Psc largely rescued the stx RNAi-induced wing-notching phenotype (Fig. 3J,K, compare with 3A). In contrast, this phenotype was unaffected when 50% of Enhancer of zeste [E(z)],which encodes the enzymatic subunit of the PRC2 complex responsible for depositing the H3K27me3 mark, was removed (Fig. 3L, compare with 3A). This result is consistent with a previous study showing that H3K27me3 modification may be dispensable for epigenetic repression of Notch (Loubiere et al., 2016). Taken together, the above results demonstrate that Stx promotes Notch mRNA expression by reducing PRC1 activity.

In Drosophila, canonical PcG target genes often contain PRE at their respective genomic loci, which is a specific cis-regulatory sequence required for PRC recruitment (Grossniklaus and Paro, 2014). To identify functional PREs at the Notch locus, we performed DNA adenine methyltransferase identification by sequencing (DamID-seq) in wing imaginal discs and analyzed these data together with the two previously published ChIP-seq datasets (Ahmad and Spens, 2019; Loubiere et al., 2016). Similar to the two published ChIP-seq datasets, DamID-seq showed significant Pc recruitment to canonical PcG target genes, including Ultrabithorax (Ubx), bithoraxoid (bxd), abdominal A (abd-A), iab-8 and Abdominal B (Abd-B). Consistent with the role of Stx in reducing Pc activity, Pc recruitment was lost at these loci when stx was overexpressed (Fig. S3A; Du et al., 2016). However, Pc recruitment to the Notch locus revealed by the DamID-seq data in this study did not fully conform to what was found in a previously published ChIP-seq data (Ahmad and Spens, 2019; Loubiere et al., 2016; Fig. S3B). Therefore, we treated all Pc-bound regions in the Notch locus revealed by three different datasets as putative PREs, which we named E1-E7 (Fig. 4A; Fig. S3B). Among them, E1 was identified by all three datasets, while E4 and E6 were identified by ChIP-seq data from Ahmad and Spens (2019) and by DamID-seq data from this study. Furthermore, E2 and E7 were only identified by ChIP-seq data from Ahmad and Spens (2019), while E3 and E5 were identified only by DamID-seq data from this study (Fig. S3B). Similar to the canonical PcG targets, the binding of Pc to E1, E3, E4, E5 and E6 was greatly reduced when stx was overexpressed (Fig. S3B), further supporting the observation that Stx promotes Notch receptor mRNA expression by preventing PRC1 recruitment to the Notch locus.

To determine whether these Pc-binding regions are functional PREs in vivo, we constructed a set of PRE-GFP reporter plasmids in which nuclear GFP expression is jointly controlled by a quadrant enhancer of the vestigial gene (vg^QE) and one of seven putative PREs flanked by two FRT sites (Sengupta et al., 2004). The presence of functional PRE is expected to repress vg^QE-mediated GFP expression, whereas excision of the PRE should restore GFP expression (Fig. 4B). These plasmids were integrated into the same attP2 landing site in the Drosophila genome by φC31 integrase, generating seven transgenic fly lines for in vivo analysis of putative PRE function. We found that GFP expression regulated by E1 or E5, but not by E2, E3, E4, E6 or E7, showed a significant reduction in wing imaginal discs compared with vg^QE alone (Fig. 4C-E,L; Fig. S4B-I′,L). Notably, E5 inhibited vg^QE-mediated GFP expression more strongly than E1 (Fig. 4D,E,L, compare with 4C′; Fig. S4C′,G′,L compare with S4B′). When E1 or E5 was excised by en-Gal4-driven flippase expression in the posterior compartment of the wing disc, vg^QE-GFP expression was restored (Fig. 4F-G′,M), implying that E1 and E5 are responsible for the decrease in GFP expression. However, when Pc was knocked down or stx was overexpressed in the posterior compartment of the wing disc, only the inhibitory effect of E5 was attenuated, whereas E1 was not (Fig. 4H-K′,M), suggesting that only the function of E5 depends on endogenous PcG activity, and E1 cooporates with factors other than PcG proteins to repress Notch expression. This conclusion is further supported by the observation that loss of ph (Fig. S4J,J′,M compare with Fig. S4G′) or knockdown of Psc (Fig. S4K,K′,N, compare with S4G′) significantly increased GFP expression of the E5 reporter. E5 spans the ∼1 kb region (ChrX: 3,153,915-3,155,043; r6.49) located in the second intron of the Notch locus (Fig. S3B). It does not overlap with any known Notch regulatory sequence, but contains 32 putative binding sites for Pho, GAF, or DSP1 (Fig. S3C), all of which are transcription factors reported to be involved in PcG recruitment (Déjardin et al., 2005; Grossniklaus and Paro, 2014; Schuettengruber et al., 2017). We believe that a detailed analysis of the binding of these transcription factors to E5 will provide mechanistic insights into epigentic control of Notch receptor gene transcription.

Notch has previously been classified as a neo-PRC1 target because only PRC1 recruitment, but no H3K27me3 modification marks, was detected around the Notch locus (Loubiere et al., 2016). Therefore, we performed quantitative chromatin immunoprecipitation (qChIP) and DamID-qPCR in the wing imaginal disc to examine whether E5 is sufficient to recruit PRC1 in the absence of H3K27me3 modification. As shown by the qChIP and DamID-qPCR assays, significant recruitment of Pc, Ph and Sce was detected at E5, but without the H3K27me3 modification mark (Fig. 4N-R). Moreover, these PRC1 components bound E5 more strongly than E1 (Fig. 4O-R), echoing the conclusion that E5 may be more important than E1 in PRC1 recruitment. Combined with the above PRE reporter analysis, we identified E5 as a functional PRE at the Notch locus, regulated by PcG. Therefore, we name E5 as the ‘Notch PRE’. In conclusion, we provide evidence that Notch is a bona fide neo-PRC1 target and that epigenetic regulation of Notch dose is mainly mediated through our newly discovered Notch PRE. Given that Notch PRE is the first experimentally verified PRE of neo-PRC1 targets, this PRE will serve as a starting point for understanding the mechanisms underlying the dynamic regulation of Notch, as well as the growing class of PRC1 target genes (Loubiere et al., 2016) that acquire little or no H3K27me3, resulting in transient silencing.

To investigate the physiological role of this Notch PRE, we generated a Notch^ΔPRE mutant allele in which the Notch PRE was removed in situ from the Notch locus by CRISPR-Cas9-mediated homologous recombination (Fig. S5A). The removal of Notch PRE, confirmed by PCR and Sanger sequencing (Fig. S5B,C), resulted in an ∼30% increase in Notch mRNA expression in Notch^ΔPRE mutant larvae (Fig. 5A,B). Based on the above PRE reporter analysis and genetic interactions, we believe that after removing the PRE in situ from the Notch locus, Notch expression is no longer regulated by Stx. Consistent with this speculation, the notched adult wing phenotype and downregulation of Notch, Wg and Cut expression caused by stx RNAi were largely restored in the context of Notch^ΔPRE mutation (Fig. S5G-I′, compare with S5D-F′). Genetic interactions further support this view, as the adult wing morphology of Notch^ΔPRE could not be altered against the background of Pc³ heterozygotes (Fig. S5L, compare with S5K,J). Of note, the posterior wing margin curvature phenotype shown by the Pc³ heterozygotic mutation (Fig. S5K,L) is a homeotic transformation phenotype unrelated to Notch hyperactivation (Bi et al., 2022).

Given the importance of Notch receptor dose in Notch signaling homeostasis, the loss of Notch PRE-mediated PRC1 repression may impair developmental processes that are sensitive to changes in Notch activity. As expected, we found that Notch^ΔPRE flies exhibited loss of notal and scutellar macrochaetae, and excessive crystal cell differentiation in the lymph gland, two stereotypical phenotypes associated with hyperactivation of Notch signaling (Brennan et al., 1999; Jung et al., 2005; Kopan, 1999; Lan et al., 2020; Lebestky et al., 2003). The number of dorsocentral bristles (DCs) and scutellar bristles (SCs) in adult Notch^ΔPRE flies was significantly decreased (Fig. 5D,E, compare with 5C), which was consistent with the decrease in the number of neur-lacZ-labeled SOPs in the wing imaginal disc of Notch^ΔPRE larvae (Fig. 5G, compare with 5F). Furthermore, Lz-GFP (piggyBac insertion allele of lz) and Notch antibody labeled crystal cells increased by approximately 50% in the primary lobes of the lymph glands of the third instar Notch^ΔPRE larvae (Fig. 5I-J, compare with 5H′). These results highlight the crucial role of Notch PRE in cell fate decisions that are highly sensitive to changes in Notch activity.

Although the loss of Notch PRE did not result in apparent defects in the developing wings, probably due to the need for further refinement and correction of the wing patterning process during the pupal stage, the genetic interactions between Notch^ΔPRE and classical Notch alleles still support the role of this PRE in Notch receptor dose control in wing development. Specifically, one copy of Notch^ΔPRE completely rescued the phenotypes caused by loss-of-function Notch^55e11/+, including overproduction of scutellar bristles and notched wings (Fig. 5K-N; de Celis et al., 1993, 1991). Furthermore, in trans-heterozygotes of Notch^ΔPRE and Notch^Ax-E2, a gain-of-function allele that does not show any defects in scutellar bristle development under heterozygosity (Fig. 5T; Xu et al., 1990), scutellar bristles were more extensively lost than in heterozygous Notch^ΔPRE alone (Fig. 5U,V, compare with 5S). Taken together, the above results suggest that Notch PRE is an integral part of the Notch receptor dose-controlling developmental process in the Notch signaling network.

Taking advantage of the power of Drosophila genetics, we identified Stx as a new epigenetic regulator that positively controls Notch receptor dose. This mode of regulation is mainly mediated by a bona fide PRE in the Notch locus, repressing Notch mRNA expression independently of H3K27 trimethylation. Importantly, this PRE plays a physiologically crucial role in Notch receptor dose control in multiple developmental contexts.

PRE is an important cis-regulatory element for PcG-mediated epigenetic repression (Grossniklaus and Paro, 2014; Steffen and Ringrose, 2014). It was originally discovered in Drosophila as DNA sequences that recruit PcG proteins and maintain transcriptional silencing of reporter genes (Chan et al., 1994; Kassis, 1994; Simon et al., 1993). To date, extensive genome-wide ChIP-seq data have revealed clustered PcG protein-binding sites in the Drosophila genome, of which only a few have been confirmed as functional PREs by reporter gene analysis (Bauer et al., 2016). This partly explains why most PcG-binding genes are not derepressed when the PcG complex is dysfunctional (Cohen et al., 2018; Gargiulo et al., 2013; Loubiere et al., 2016; Morey et al., 2015; Pherson et al., 2017). In cultured Drosophila BG3 cells with reduced ph expression, only about 5% of the genes occupied by PcG proteins are significantly upregulated (Pherson et al., 2017), whereas in the eye discs of PcG mutants, this number does not exceed 30% (Loubiere et al., 2016). A similar phenomenon was observed in mice. When Bmi1, Ring1a or Ring1b (Rnf2) expression is reduced, only about 4.8-15% of PcG target genes are derepressed (Cohen et al., 2018; Gargiulo et al., 2013; Morey et al., 2015). These results confirm why many cis-regulatory elements that recruit PcG proteins have little effect in in vivo reporter assays (Cuddapah et al., 2012; Cunningham et al., 2010). By integrating our DamID-seq and published ChIP-seq data (Ahmad and Spens, 2019; Loubiere et al., 2016), we identified seven PRC1-binding regions at the Notch locus with varying degrees of binding strength. Our study suggests that binding strength is not necessarily a good predicator of the physiological requirement for these binding regions. Only in combination with in vivo reporter assays did we identify two PRC1-binding regions sufficient to mediate Notch gene repression, only one of which is shown to respond to epigenetic regulation by Pc and Stx in vivo. Notably, this experimentally validated PRE does not belong to the PRC1-binding sites with the highest binding strength identified in previously published ChIP-seq datasets. Therefore, our study demonstrates that, in PRE prediction, it is not only necessary to determine the binding of PcG proteins to genomic loci but also, more importantly, to directly test the role of PcG-binding sites by reporter assays, as well as in situ deletion of these PcG-binding sites in the corresponding genomic loci. This conclusion echoes concerns that high-throughput sequencing analysis alone only increases the number of associations and should be further validated to reveal causality in biological processes (Stern, 2022).

Although PRE is considered crucial for epigenetic silencing of target genes, there is little in-depth study of their physiological function. To date, only a few PREs in the Drosophila genome have been deleted to directly test their in vivo requirement, and the phenotypic consequences of PRE deletion vary. In one case, deletion of Ubx PRE leads to ectopic Ubx expression and homeotic transformation (Sipos et al., 2007), supporting the crucial role of PRE in mediating epigenetic repression. This contrasts with the effect of removing a characterized PRE of iab-7, which does not result in a visible phenotype (Mihaly et al., 1997). This may be due to the presence of other cis-regulatory PRE-like elements required for epigenetic gene repression. Alternatively, in the absence of a canonical PRE, compensatory mechanisms for developmental robustness can be activated, as in the case of epigenetic control of the invected-engrailed (inv-en) gene complex. After in situ removal of inv-en PRE, the remaining weak PcG-binding interactions between PcG domains can maintain epigenetically silenced 3D chromatin structures (De et al., 2016). A similar result was observed in PRE deletion of the vg gene (Ahmad and Spens, 2019). The exact mechanisms behind these paradoxical phenomena require further investigation.

Metazoans use a variety of genetic, epigenetic and cellular mechanisms to ensure developmental robustness to buffer against genetic perturbations or environmental fluctuations (Félix and Wagner, 2008; Spatz et al., 2021). In our study, the increased expression of Notch caused by PRE deletion alters transient and rapid decision-dependent cell fate specification in the notum and lymph gland, but does not result in significant defects in wing margin formation, a gradual developmental process that requires further refinement and correction in later developmental stages. These seemingly contradictory consequences of PRE deletion in different tissues may reflect the different plasticity of the developmental process conferred by genetic and cellular compensation, thereby adapting to changes in developmentally regulatory gene expression and heterogeneous cellular responses (Xiao et al., 2022). Although our Notch PRE deletion mutation does not readily exhibit the classical gain-of-function Confluens phenotype in the developing wing, it is sufficient to compensate for the loss-of-function wing-notching phenotype in the sensitized background. Similarly, in vivo deletion of PRE in the dachshund (dac) gene results in only a very mild tarsal transformation of the first leg, a phenotype that can be further enhanced when flies are reared at higher temperatures (Ogiyama et al., 2018). Consistent with these observations, ectopic Notch or dac activity is known to lead to broader defects (Anderson et al., 2006; Baonza and Freeman, 2005; Córdoba and Estella, 2014; Dong et al., 2001; Ku and Sun, 2017; Shen and Mardon, 1997), as opposed to the more limited phenotypes observed when the corresponding PRE is deleted in situ. Thus, context-dependent sensitivity to PRE-mediated epigenetic control highlights the integral role of epigenetic adaptation to genetic and environmental fluctuations during development.

All fly crosses were maintained at 25°C, except those listed in Table S1. The following fly stocks were obtained from the Bloomington Drosophila Stock Center: ap (apterous)-Gal4 (#3041), ap-lacZ (#5374), dpp (decapentaplegic)-Gal4 (#1553), en (engrailed)-Gal4 (#30564), hh (hedgehog)-Gal4, nub-Gal4 (#25754), Lz-GFP (#43954), mam⁸ (#1596), MS1096-Gal4 (#8860), Notch¹ (#6873), Notch^55e11 (#28813), neur-lacZ (#4369), NRE-gfp (#30727), pPc-Pc-GFP (#9593), Pc³ (#1730), ph⁴¹⁰ (#5813), Psc^e24 (#24155), ptc-Gal4 (#2017), rotund (rn)-Gal4 (#7405), Ser¹ (#89), spalt major (salm)-Gal4 (#5818), Su(dx)² (#293), T80-Gal4 (#1878), UAS-flp (#4539), UAS-Notch^FL (#26820), UAS-Nslmb-vhhgfp4 (#38422), UAS-Pc RNAi (#33622) and w¹¹¹⁸ (#3605). Three UAS-stx RNAi lines (GD27036, GD23946 and KK109495) and a UAS-Psc RNAi line (GD30587) were obtained from the Vienna Drosophila RNAi Center (VDRC). The stx-yfp protein trap line CPTI-004181 and the GS line for stx (GS200070) were obtained from the Drosophila Genetic Resource Center (DGRC). UAS-stx-Flag and UAS-Flag-Pc have been described previously (Du et al., 2016).

E(z)⁶³ was a gift from Peter Harte (Case Western Reserve University, Cleveland, OH, USA). hs-flp;; ubi-gfp, FRT2A was a gift from Jocelyn McDonald (Kansas State University, Manhattan, KS, USA). lz-Gal4, UAS-gfp (BL#6313) was a gift from Ying Su (Ocean University of China, Qingdao, China). Notch^Ax-E2 was a gift from Michelle Longworth (Cleveland Clinic, OH, USA). Notch^PL24 was a gift from Alain Vincent (Université Toulouse III, France) (Bourbon et al., 2002). Pc^XT109 was a gift from Jürg Müller (Max-Planck Institute of Biochemistry, Munich, Germany) (Müller et al., 1995). ph^del was a gift from Jian Wang (University of Maryland, College Park, USA) (Feng et al., 2011). UAS-Dam was a gift from Andrea Brand (University of Cambridge, UK) (Southall et al., 2014). UAS-pPc-ph-GFP was a gift from Donna Arndt-Jovin (Max-Planck Institute for Biophysical Chemistry, Göttingen, Germany) (Ficz et al., 2005). UAS-Sce-Flag was a gift from Judith Kassis (National Institutes of Health, USA) (Langlais et al., 2012).

Transgenic flies expressing UAS-Pc-Dam were generated by P-element mediated germline transformation. UAS-stx-HA and GFP reporters for putative Notch PREs were integrated into attP2 site using φC31 integrase-mediated recombination to generate transgenic flies. Notch^ΔPRE mutant was generated by CRISPR/Cas9-mediated homologous recombination (Port et al., 2014). Details of related plasmids and primers are provided in Table S2.

Standard procedures were used for wing disc immunofluorescence staining and in situ hybridization (Su et al., 2011). Lymph glands were immunofluorescently stained using the described protocol (Evans et al., 2014). The following primary antibodies were used for immunofluorescence staining: rabbit anti-β-galactosidase (1:4000; 55976; Cappel), mouse anti-β-galactosidase [1:200; 40-1A; Developmental Studies Hybridoma Bank (DSHB)], mouse anti-Cut (1:100; 2B10; DSHB), rabbit anti-cleaved Dcp-1 (1:400; 9578; Cell Signaling Technology), mouse anti-Dl (1:200; C594.9B; DSHB), rabbit anti-GFP (1:4000; A11122; Invitrogen), mouse anti-N^icd (1:200; C17.9C6; DSHB), rabbit anti-Pc (1:500; this study), rat anti-Ser (1:2000; a gift from Kenneth Irvine, Rutgers University, NJ, USA; Papayannopoulos et al., 1998), mouse anti-Wg (1:100; 2A1; DSHB) and rabbit anti-Wg (1:1000; this study). Primers used for RNA probes synthesis are listed in Table S2.

Fluorescence images were acquired with a Leica SP8 confocal microscope or a Zeiss Axio Imager Z1 microscope equipped with an ApoTome. Heatmap images were generated in ImageJ using the heatmap plugin. Figures were assembled in Adobe Photoshop CC. Minor image adjustments (brightness and/or contrast) were made in Adobe Photoshop CC.

Adult wings and nota were dissected and mounted as described previously (Zhu et al., 2003). The images of these adult tissues were acquired with a Leica DMIL inverted microscope (wings) or a Keyence VHX-7000 digital microscope (nota).

Rabbit polyclonal antibodies were raised against amino acids 191-354 of the Pc protein (Schuettengruber et al., 2009) and amino acids 3-468 of the Wg protein (Brook and Cohen, 1996) (Abclonal Biotech), and affinity purified for immunofluorescence staining. The specificity of rabbit anti-Pc antibody was validated by Pc RNAi in the wing disc (Fig. S6A'). Staining pattern of rabbit anti-Wg antibody in the wing pouch (Fig. S6B) is similar to that of mouse anti-Wg antibody (4D4; DSHB) (Fig. S6C).

Total RNA of third instar larval wing discs was extracted using Eastep Super Total RNA Extraction Kit (LS1040; Promega). Reverse transcription was performed with Eastep RT Master Mix Kit (LS2050; Promega). cDNA levels were quantified by real-time PCR in a 7500 real time PCR system (Applied Biosystems) using PowerUp SYBR Green Master Mix (A25741; Thermo Fisher Scientific). Relative fold changes of Notch mRNA levels were calculated using comparative CT method. Samples from three independent experiments were prepared and run in triplicate. Primers used for qPCR are listed in Table S2.

qChIP analyses of the wing disc were performed using the previously described protocol (Loubiere et al., 2017). For each qChIP experiments, ∼900 pairs of wing discs were collected and three biological replicates were performed. w¹¹¹⁸ was used for H3K27me3 qChIP experiments. pPc-Pc-GFP transgenic flies, in which GFP-tagged Pc is expressed under the control of an endogenous Pc promoter (Dietzel et al., 1999), were used for qChIP experiments on Pc. For qChIP experiments with Ph or Sce, wing discs expressing UAS-pPc-ph-GFP (Ficz et al., 2005) or UAS-Sce-Flag (Langlais et al., 2012) under nub-Gal4 control were used. Rabbit anti-GFP (6 μg per IP; A11122; Invitrogen) for Pc-GFP and Ph-GFP, rabbit anti-H3K27me3 (3 μg per IP; ABE44-S; Millipore) and mouse anti-Flag (5 μg per IP; AT0022; Engibody) for Sce-Flag were used. Primers used for qPCR are listed in Table S2. The results were normalized using the PGRP-LE gene, which served as a negative control.

The Pc cDNA was cloned into the pUAST-attB-LT3-Dam vector (a gift from Andrea Brand) (Southall et al., 2014), and transgenic flies were generated by integrating the vector at the attP2 site. UAS-Dam transgenic flies were used as negative controls for nonspecific Dam activity. The genotypes used for the experiments were: nub-Gal4/+; UAS-Dam/+, nub-Gal4/+; UAS-Dam-Pc/+ and nub-Gal4/+; UAS-Dam-Pc/UAS-stx-HA. Crosses were raised at 25°C. For DamID-qPCR, 100 pairs of wing discs from each of three independent experiments were dissected from third instar larvae in ice-cold Schneider's Drosophila medium. For DamID-seq, 300 pairs of wing discs were dissected. The DamID was performed according to a previously described protocol (Vogel et al., 2007). The PCR products were purified by QIAquick PCR purification kit (Qiagen).

For DamID-qPCR, purified PCR products were used for real-time quantitative PCR analyses. Primers for qPCR are listed in Table S2. For DamID-seq, purified PCR products were used to generate the next-generation sequencing (NGS) libraries that were further sequenced via HiSeq 2500 and multiplexed to yield ∼50 million mapped reads per sample (Novogene). Using the damidseq_pipeline (http://owenjm.github.io/damidseq_pipeline/) (Marshall and Brand, 2015), NGS reads were aligned to the Drosophila melanogaster reference genome version r6.41 by bowtie2 (Langmead and Salzberg, 2012), and a final log2 ratio file in bedgraph format was generated. These files were visualized and analyzed by Integrative genomics viewer (IGV) (Robinson et al., 2011).

To quantify the hair-loss region in adult wing margin, mounted adult wings were imaged and the length of hair-loss regions was measured by QCapture Pro software (QImaging). For quantification of the intensity of GFP fluorescence in the wing disc, images were taken with the same confocal settings. The GFP fluorescence intensity of wing pouch region, posterior or anterior compartment of the wing pouch was measured using NIH ImageJ software.

Statistical analysis was performed using Graphpad Prism 8. For comparison between sample pairs, a two-tailed Student's t-test was used (Fig. 2G; Fig. 5A,B,E,J; Fig. S4M,N); for comparisons between three or more conditions, one-way ANOVA followed by Dunnett's multiple comparison tests were used (Fig. 1M,Q; Fig. 4L-R; Fig. 5N,R,V; Fig. S4L).

We thank Donna Arndt-Jovin, Andrea Brand, Peter Harte, Judith Kassis, Kenneth Irvine, Michelle Longworth, Jocelyn McDonald, Jürg Müller, Ying Su, Alain Vincent, Jian Wang, the Bloomington, DGRC, THFC and VDRC Stock Centers, and DSHB for fly stocks, antibodies and plasmids. We thank Dr Guilan Li at the National Center for Protein Science at Peking University for sample preparation, and Xinhan Meng and Di Yang in Zhu laboratory for assistance with ChIP and DamID analyses.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201297.reviewer-comments.pdf.