Engrailed acts with Nejire to control decapentaplegic expression in the Drosophila ovarian stem cell niche

In adult organisms, stem cells undergo asymmetric divisions to self-renew and produce differentiated cells to maintain tissue homeostasis. The decision between self-renewal and differentiation of stem cells is strongly influenced by niche-derived signals (Hsu and Fuchs, 2012; Morrison and Spradling, 2008; Scadden, 2014).

The Drosophila ovary provides an ideal system to study how niche-associated signals regulate germline stem cell (GSC) self-renewal versus differentiation (Fuller and Spradling, 2007; Harris and Ashe, 2011; Lin, 2002; Losick et al., 2011; Xie, 2013). At the anterior tip of the ovary, the GSC niche is formed by several types of stromal cells, including terminal filament (TF) cells, cap cells and escort cells (ECs, previously known as inner germarial sheath cells). Besides providing a physical location to house GSCs, the niche also produces a range of signaling molecules, including Decapentaplegic (Dpp; the fly ortholog of BMP2/4), Hedgehog (Hh), Wnt factors and Unpaired (Upd), which act in concert to control GSC activity (Losick et al., 2011). Among these signaling molecules, Dpp is the primary factor that promotes GSC self-renewal by repressing the transcription of differentiation-promoting factor bag of marbles (bam) (Chen and McKearin, 2003a; McKearin and Spradling, 1990; Song et al., 2004). dpp is transcribed mainly in cap cells and, to a lesser extent, in ECs (Liu et al., 2015; Xie and Spradling, 2000). However, it remains largely unknown how its expression in these niche cells is regulated.

Engrailed (En) is a multifaceted homeodomain-containing transcription factor that plays an essential role in the development of Drosophila appendages and segments, and is also involved in the development of the nervous system (Morgan, 2006). En binds directly to specific DNA sequence and can function as a transcriptional activator or a transcriptional repressor depending on its associated co-factors (Alexandre and Vincent, 2003). In the germarium, En is expressed in both TF and cap cells and is used as a cap cell marker (Forbes et al., 1996; Ward et al., 2006). Removing En function from TFs leads to disorganization of TF stacks during ovary development, indicating a role of En in the proper organization of the TFs (Bolívar et al., 2006). Interestingly, one previous publication showed that germaria bearing cap cells mutant for En exhibited a GSC loss phenotype (Rojas-Rios et al., 2012), indicating a role for cap cell-expressed En in maintaining GSCs non-cell-autonomously. Although this report also suggested that En maintains GSCs indirectly by promoting dpp expression in ECs via Hh signaling (Rojas-Rios et al., 2012), several recent studies showed that Hh signaling in ECs instead promotes GSC differentiation (Li et al., 2015a; Liu et al., 2015; Lu et al., 2015). It was also reported that ectopic expression of En, but not Hh, in ECs results in an expansion of GSC-like cells with expanded Dpp signaling activation outside the niche (Eliazer et al., 2014), suggesting that En can promote Dpp signaling independently of Hh. Thus, the mechanism underlying En-mediated GSC maintenance remains elusive.

Nejire (Nej) is the Drosophila ortholog of vertebrate CREB-binding protein CBP/p300 (Chan and La Thangue, 2001). The CBP/p300 proteins contain multiple protein-protein interaction domains, including three cysteine-histidine (CH)-rich domains, one kinase-inducible domain interacting (KIX) domain, one bromodomain (Br), one SRC1-interacting domain (SID), and one histone acetyltransferase (HAT) domain. They function as versatile transcriptional co-activators with the ability to interact with a myriad of partners, including transcription factors and signaling molecules (Vo and Goodman, 2001).

In this study, we show that En controls dpp expression in cap cells by binding to a 5′ enhancer region, which drives heterologous reporter expression in the niche in a manner similar to endogenous dpp. Compromising En activity in cap cells leads to downregulation of dpp expression in the niche, while ectopic En expression in ECs results in ectopic dpp transcription. We further provide evidence showing that En forms complexes with Nej, directs Nej to this enhancer region, and that the En activity required to turn on dpp expression in cap cells is dependent on Nej function.

Previous studies showed that dpp transcripts are expressed in niche cells, with high expression levels detected in cap cells and low expression levels in ECs (Liu et al., 2010, 2015; Song et al., 2004; Wang et al., 2008; Xie and Spradling, 2000). To investigate how dpp expression is regulated in these niche cells, we first examined the available dpp transcriptional reporters, including BS1.0-lacZ that recapitulates the dpp expression pattern in larval imaginal discs (Blackman et al., 1991). None of these lines, under the control of enhancer elements located in the 3′ cis-regulatory region, exhibited reporter activity in the ovarian niche (not shown).

We then focused on the 5′ cis-regulatory region and generated nine transgenic reporter lines (D1-D9) covering a span of ∼36 kb 5′ to the dpp coding region (Fig. S1A, Table S1). lacZ reporter activity was detected using an antibody against β-galactosidase (β-gal). Two lines exhibited reporter activity in niche cells, with D3-lacZ showing β-gal expression mainly in TF cells and D6-lacZ exhibiting reporter activity in cap cells and one to two adjacent TF cells (Fig. S1B,C). Further dissection of the D6 region identified a 4 kb fragment (D6.3) that drove strong reporter activity in all cap cells and several ECs (Fig. S1A,D-F), indicating that this region contains a cap cell-associated enhancer. Other regions (D6.1 and D6.2) exhibited weak enhancer activity in cap cells. Fine mapping (together with ChIP, see below, Fig. 2C) led to the identification of a 2.0 kb fragment located at the 3′ end of this region that drove strong β-gal expression in virtually all cap cells (Fig. 1A,B); we named this reporter line dpp2.0-lacZ. About 15.2% of the germaria examined (n=730) also showed weak β-gal expression in some ECs (Fig. 1C), a pattern reminiscent of endogenous dpp expression as detected by FISH (Liu et al., 2015).

In line with the notion that Notch signaling is required for the maintenance of cap cells in adult ovary (Song et al., 2007), β-gal expression was strongly reduced in cap cells with compromised Notch signaling activity (Fig. 1D, Fig. S1G). We further examined whether this reporter also reflects dpp transcriptional activation in ectopic cap cells induced by ectopic Notch signaling during niche formation (Song et al., 2007; Ward et al., 2006), and found β-gal expression in these cap cells located outside the normal niche position (Fig. 1E). Furthermore, dpp2.0-lacZ activity was upregulated in ECs with compromised Hh signaling (Fig. S1H), consistent with our previous study showing that Hh signaling in ECs dampens dpp expression (Liu et al., 2015). Together, these data show that dpp2.0-lacZ faithfully recapitulates dpp expression in niche cells.

We next sought to investigate the mechanism regulating dpp2.0-lacZ in the niche. Through bioinformatics analysis using PROMO (alggen.lsi.upc.es/cgi-bin/promo_v3/promo/promoinit.cgi?dirDB=TF_8.3), we found that this region contains consensus binding sequences for several well-characterized transcription factors (Table S2), including Engrailed (En), a homeodomain-containing factor that is generally used as a cap cell marker (Fig. 2A). To test whether En regulates dpp2.0-lacZ activity we used the bab1-gal4 driver, which is expressed in TF and cap cells, in combination with an RNAi construct against en, and found that germaria with compromised En function (referred to as enⁱ germaria) failed to express this reporter (Fig. 2B).

Since En is a DNA-binding protein, we examined whether En binds directly to the 2.0 kb enhancer region. We ectopically expressed En with an epitope tag in S2 cells, which do not express En, and conducted ChIP experiments followed by real-time PCR. En was found to be highly enriched in this 2.0 kb region, as compared with other genomic regions (Fig. 2C).

To further define the region responsible for the cap cell-associated enhancer activity, we generated a series of small deletions (∼100 bp) across the 2.0 kb region and identified a 300 bp region essential for driving β-gal expression in cap cells (Table S3). Interestingly, four putative En binding sites (Kassis et al., 1989; Sanicola et al., 1995; Serrano and Maschat, 1998) were identified in this 300 bp region (Fig. S2A). To test the functional importance of these potential En binding sites, we generated transgenic reporter lines bearing single or double deletion of these sites and examined their enhancer activity in the niche. A small deletion that removes two adjacent putative binding sites [dpp2.0-lacZ (Δ2)] abolished enhancer activity in cap cells (Fig. S2A-D). We further investigated whether these two putative sites are involved in En binding and conducted far-western blot analysis. Whereas bacterially expressed GST-tagged En bound strongly to the 2.0 kb fragment, its binding to the dpp2.0-lacZ (Δ2) derivative was clearly compromised (Fig. 2D). In vivo, whereas dpp2.0-lacZ reporter activity was ectopically activated in ECs upon En overexpression, dpp2.0-lacZ (Δ2) did not respond to ectopic En expression (see below, Fig. 4B, Fig. S3K). Taken together, these data suggest that En acts through these sites to regulate enhancer activity in cap cells.

In addition to loss of β-gal reporter expression, enⁱ germaria also exhibited a GSC loss phenotype, which was observed using two independent RNAi constructs (Fig. S3A-C). In wild-type (WT) germarium, each niche houses two or three GSCs that are in contact with cap cells and undergo asymmetric divisions to produce a self-renewing GSC daughter that remains within the niche and a differentiating cystoblast (CB) daughter that moves away from the niche. GSCs possess an anteriorly positioned spectrosome, which is a spherical intracellular structure enriched in cytoskeletal proteins such as α-Spectrin, whereas CBs contain a randomly positioned spectrosome and differentiating cysts possess a branched fusome interconnecting individual cystocytes. Each control germarium contained 2.3±0.05 (n=124) GSCs, whereas each enⁱ germarium harbored only 0.9±0.08 (n=126) GSCs (P<0.001). To exclude potential off-target effects of RNAi, we generated cap cells mutant for En using a null allele [en^E (Gustavson et al., 1996)]. As reported previously (Rojas-Rios et al., 2012), these germaria exhibited a GSC loss phenotype (Fig. S3D,E), demonstrating a role of En in maintaining GSCs non-cell-autonomously.

Since En is expressed in TF and cap cells during their formation and plays a role in TF organization (Bolívar et al., 2006), disrupting En expression could potentially affect niche organization or function. We performed temperature-shift experiments for en knockdown to bypass its requirement during development and to check its function at the adult stage. These enⁱ germaria gradually exhibited a GSC loss phenotype (Fig. S3F), although their niches contained a similar number of cap cells as control counterparts (Fig. S3G), excluding the possibility that GSC loss is a result of defective niche formation.

We then investigated how En maintains GSCs in the niche. Our data indicate that the GSC loss observed in enⁱ germaria is not a result of germ cell death (Fig. S3H,I) but a consequence of precocious differentiation, as those spectrosome-containing cells within the niche expressed Pbam-gfp, a transcriptional reporter of bam (Chen and McKearin, 2003b). In control germaria, Pbam-gfp was activated in CBs and germline cysts outside the niche (Fig. 3A). However, its activity was detected in the spectrosome-containing cells in the niche of enⁱ germaria (Fig. 3B).

Given earlier findings showing that Dpp signaling activation in GSCs directly represses the transcription of bam (Chen and McKearin, 2003a; Song et al., 2004), we examined phosphorylated Mad (pMad), a real-time reporter for Dpp signaling activation. Compared with the high levels present in control GSCs, pMad levels were significantly reduced in GSCs of enⁱ germaria (Fig. 3C,D), indicating defective signaling activation. Considering our results above showing that En binds the cis-regulatory region of dpp and regulates dpp2.0-lacZ reporter activity in cap cells, we hypothesized that En might directly regulate dpp expression in cap cells and therefore performed FISH experiments to detect dpp transcripts. As reported previously, dpp transcripts were mainly detected in cap cells of control germaria; however, dpp expression was strongly reduced in cap cells of enⁱ germaria (Fig. 3E,F). Together, these data show that En regulates dpp expression in cap cells and thereby maintains GSCs non-cell-autonomously.

To test whether ectopic En expression is sufficient to turn on dpp expression, we forced En expression in ECs, which are believed to share a common precursor pool with cap cells (Song et al., 2007). We employed c587-gal4 in combination with a temperature-sensitive version of Gal80 (Gal80^ts) to drive En expression in ECs and prefollicular cells. As reported previously (Eliazer et al., 2014), these germaria contain ectopic spectrosome-containing cells outside the niche (Fig. S3J). Many of these ectopic spectrosome-containing cells also exhibited pMad (Fig. 4A), indicating ectopic Dpp signaling activation. In line with this, dpp2.0-lacZ reporter activity was also ectopically activated in these En-expressing ECs (Fig. 4B). Furthermore, elevated dpp transcript levels were detected in the En-expressing ECs by FISH (Fig. 4C compared with Fig. 3E). We also noted that dpp2.0-lacZ (Δ2) (Fig. S2A), which exhibits defective En binding activity (Fig. 2D), was not activated in these En-expressing ECs (Fig. S3K), supporting the notion that these two putative En binding sites are important for dpp activation in vivo. It is also worth noting that these ectopic En-expressing ECs did not express cap cell markers such as Hh (Fig. S3L), indicating that ectopic En does not transform ECs into cap cells. To examine the causal relationship between ectopic expression of En and the dpp transcripts observed in ECs, we tested genetic interaction by knocking down dpp in these En-expressing ECs. Our results show that formation of the ectopic spectrosome-containing cells, as induced by ectopic En expression, was strongly suppressed when dpp was knocked down in those cells (Fig. 4D, Fig. S3M-P).

Rojas-Rios et al. (2012) showed previously that cap cell-expressed En maintains GSCs indirectly by regulating hh expression in cap cells, which in turn promotes dpp expression in ECs (Rojas-Rios et al., 2012). To address whether En regulates cap cell-expressed dpp via hh expression, we examined hh expression in cap cells with compromised En function. Using a different source of anti-Hh antibody to that of Rojas-Rios et al. (2012) (see Materials and Methods), we found that Hh expression was still detectable (although at a lower level) in 55% (n=49) of enⁱ germaria, including those exhibiting GSC loss (Fig. S4A,B). Furthermore, in cap cells mutant for en^E, Hh was also detected at a reduced level (Fig. S4C). We next conducted FISH experiments to detect hh transcripts in these backgrounds. In control germaria, hh transcripts were detected in cap cells as well as in some ECs (Fig. S4D). Although the expression level was reduced, hh mRNA was still detected in 67% (n=21) of enⁱ germaria (Fig. S4E). In line with this, recent studies have shown that knocking down hh in the niche by bab1-gal4 does not result in GSC loss, in contrast to the GSC loss phenotype observed in enⁱ germaria (Li et al., 2015a; Liu et al., 2015; Lu et al., 2015). We further conducted genetic rescue experiments by ectopically expressing hh in cap cells of the enⁱ germarium. When ectopically expressed in a WT background, these Hh-expressing germaria contained fewer spectrosome-containing cells (Fig. S4F,H). Similarly, ectopic Hh expression in cap cells of the enⁱ germaria did not rescue the GSC loss phenotype (Fig. S4G,H). Taken together with earlier results showing direct binding of En to the enhancer region of the dpp locus, these data support the notion that En directly regulates dpp expression in cap cells.

As a sequence-specific DNA-binding protein, the ability of En to activate target gene expression depends on the co-activator with which it associates. To determine how En activates dpp transcription in cap cells we searched for its transcriptional co-activator. We tested Nejire (Nej), the fly ortholog of CBP/p300, a well-established transcriptional co-activator that can link sequence-specific transcription factors with the transcriptional machinery by serving as a protein bridge or by nucleating a multicomponent transcriptional regulator complex (Goodman and Smolik, 2000).

We first tested whether nej interacts genetically with en in the ovary. Whereas germaria singly heterozygous for either en^E or nej^S342, an EMS-induced hypomorphic allele (Florence and McGinnis, 1998), contained GSCs, germaria double heterozygous for en^E and nej^S342 exhibited a GSC loss phenotype (Fig. 5A,B), suggesting that En and Nej function in the same genetic pathway for GSC maintenance. Since Nej was detected in all cells in the germarium (Fig. S5A), we tested whether the synergistic relation between Nej and En in maintaining GSCs reflects Nej function in the niche cells. We knocked down nej using bab1-gal4 and found that these germaria exhibited premature GSC loss and reduced pMad levels (Fig. 5C,F). Furthermore, these germaria exhibited a strong reduction in dpp2.0-lacZ activity in cap cells (Fig. 5D), similar to that observed in enⁱ germaria. These data support a role of Nej in regulating dpp expression in the niche. Consistently, further knocking down Nej activity in cap cells of enⁱ germaria enhanced the GSC loss phenotype (Fig. 5E,F). Lastly, compromising Nej activity in En-expressing ECs strongly suppressed the formation of ectopic spectrosome-containing cells, prevented ectopic pMad presence outside the niche and, importantly, suppressed ectopic dpp transcription in the En-expressing ECs (Fig. S5B-F). These data suggest that Nej is required for En to activate dpp expression in cap cells.

To test whether En forms a complex with Nej we performed co-IP in S2 cells. Our results showed that the N-terminal portion of Nej including the KIX domain can interact with En, whereas other parts of Nej, including the Br, HAT and SID domains, are not essential for En binding (Fig. 5G,H). Additional results showed that this N-terminal region of Nej forms a complex with the C-terminal portion of En. Interestingly, this N-terminal portion also interacted with En[Act], a truncated variant of En that lacks the repressive domains and functions as a transcriptional activator in vivo (Fig. 5G,I) (Alexandre and Vincent, 2003). Together, these results support the notion that Nej is a transcriptional co-activator for En.

Finally, we investigated whether, like En, Nej occupies the 2.0 kb dpp enhancer region. However, we were not able to detect occupancy for endogenous Nej at this region in S2 cells (Fig. 5J). We reasoned that this could be due to lack of En expression in S2 cells and performed ChIP experiments in S2 cells with ectopic En expression. Indeed, in these conditions Nej was found to be enriched at the 2.0 kb enhancer region (Fig. 5J). These data support a role for En (as a sequence-specific transcription factor) as a link to bring Nej to this enhancer region.

The niche plays a vital role in maintaining stem cell function throughout the lifespan of an organism. The Drosophila ovary was one of the first in vivo systems shown to demonstrate the niche concept at the structural and molecular level, with Dpp being the primary niche-associated signal promoting GSC self-renewal (Xie and Spradling, 2000). In the GSC niche, dpp mRNA is mainly detected in cap cells and is weakly expressed in ECs. Several players, including epigenetic regulators such as the histone demethylase Lsd1 [Su(var)3-3], the histone H3K9 methyltransferase Eggless, H3K4 methyltransferase Set1, Piwi, and the Polycomb complex subunits, repress dpp expression in ECs (Eliazer et al., 2014; Jin et al., 2013; Li et al., 2016; Ma et al., 2014; Wang et al., 2011; Xuan et al., 2013). Knocking down the function of these genes in ECs activates ectopic Dpp signaling outside the niche, although the detailed mechanisms await further investigation. By contrast, the regulation of dpp in cap cells is less understood. JAK/STAT signaling plays a role in promoting dpp expression in cap cells via a mechanism that has yet to be identified (Lopez-Onieva et al., 2008; Wang et al., 2008). In this study, we show that the cap cell-expressed transcription factor En binds to a 2.0 kb fragment in the 5′ cis-regulatory region of dpp and controls its expression in cap cells. Our data further suggest that Nej is recruited by En to this cis-regulatory region and serves as the transcriptional co-activator for En to regulate dpp expression (Fig. S5G). Thus, this study unveils the underlying mechanism controlling dpp expression in cap cells.

Dpp, a member of the transforming growth factor β (TGFβ) superfamily, is widely employed during fly development, including in dorsal/ventral patterning of the embryo, morphogenesis of the larval gut, and anterior/posterior patterning of wing imaginal discs, as well as in stem cell regulation in the adult (Morata and Lawrence, 1975; O'Connor et al., 2006). This suggests a complex regulation of dpp expression during development and, indeed, multiple enhancers have been identified to direct dpp expression in different developmental contexts (Blackman et al., 1991; St Johnston et al., 1990). Adding to this complexity, our study shows that one previously uncharacterized 2.0 kb fragment in the 5′ cis-regulatory region directs dpp expression in cap cells. This cis-regulatory region is distinct from the well-characterized ʻdisc enhancer' located at the 3′ UTR (Blackman et al., 1991; Masucci et al., 1990). Consistent with their cellular context-dependent enhancer activities, this 2.0 kb enhancer does not drive β-gal expression in a pattern similar to endogenous dpp expression in imaginal discs (not shown) and, likewise, the disc enhancer does not exhibit activity in niche cells. We also note that dpp2.0-lacZ reporter activity does not respond to ectopic Upd expression in ECs (not shown), a condition shown to activate ectopic dpp expression in ECs (Decotto and Spradling, 2005; Lopez-Onieva et al., 2008; Wang et al., 2008), thus indicating an intricate and complex regulation of dpp expression in the niche.

We further identify several En consensus binding sequences within this 2.0 kb fragment and our results show that removing two of these sites (region 10, Table S3) abolishes its enhancer activity in cap cells. We also note that deleting two adjacent regions (regions 9 and 11, Table S3) also abolishes reporter activity in the niche, suggesting that the proper organization of this genomic region is important for reporter activity. Furthermore, compromising En activity in cap cells abolishes the enhancer activity of the 2.0 kb fragment, diminishes dpp expression in cap cells and consequently leads to GSC loss. Conversely, ectopic expression of En in ECs results in ectopic dpp expression, ectopic activation of Dpp signaling outside the niche and the formation of ectopic pMad-positive GSC-like cells, which is suppressed by knocking down dpp in these cells. These data support a role of En in directly regulating dpp expression in the GSC niche via this 2.0 kb enhancer.

Our study also unveils distinct roles of En in controlling dpp expression in different developmental contexts. While early studies showed that, in imaginal discs, En directly associates with and acts through the disc enhancer in the 3′ UTR to suppress dpp expression in the En-expressing posterior compartment (Sanicola et al., 1995), our results show that En acts as a transcriptional activator to turn on dpp expression in the niche. It has been shown that En, similar to other homeobox genes, can act as a transcriptional activator or repressor depending on the co-factors that it associates with (McGinnis and Krumlauf, 1992). Thus, the distinct activity of En towards dpp expression in the imaginal disc versus the GSC niche is likely to be due to the different co-factor associated with En. Although the co-factor required for En-mediated suppression of dpp in imaginal discs remains elusive, our data support a role for Nej as the co-activator for En in the niche. Knocking down Nej activity in the niche compromises 2.0 kb enhancer activity in cap cells and leads to GSC loss as a consequence of reduced Dpp signaling. Furthermore, Nej also binds to the 2.0 kb fragment in an En-dependent manner in S2 cells. Consistent with Nej being the co-activator, the formation of ectopic spectrosome-containing cells induced by ectopic En expression in ECs is strongly suppressed by compromising Nej function in these cells. Nej contains one HAT domain that can acetylate histone H3 lysine 27 (H3K27ac) to antagonize Polycomb complex-mediated transcriptional suppression via histone H3 lysine 27 trimethylation (H3K27me3) (Goodman and Smolik, 2000). A recent paper showed that Polycomb group genes function in ECs to suppress dpp expression and promote germ cell development (Li et al., 2016). Thus, it would be interesting to investigate whether the role of Nej as a co-activator for En-mediated dpp expression in cap cells encompasses its role in epigenetic regulation by antagonizing Polycomb group activity.

The following lines were used in this study. y¹w¹¹¹⁸ (as WT control), c587-gal4 (or c587; from T. Kai, Temasek Life Sciences Laboratory, Singapore), Pbam-gfp (from D. Chen and D. McKearin, UT Southwestern Medical Center, USA), UAS-hh (from J. Jiang, UT Southwestern Medical Center, USA), en^E (from K. Basler, University of Zurich, Switzerland), Dad-lacZ (from T. Tabata, University of Tokyo, Japan) and bab1-gal4.UAS-flp (from A. Gonzalez-Reyes, Centro Andaluz de Biologia del Desarrollo-CABD, Spain). Other stocks were obtained from Bloomington Drosophila Stock Center (BDSC), Kyoto Stock Center, NIG-Fly or Vienna Drosophila RNAi Center (VDRC): dpp^dsRNA (BDSC #25782), en^dsRNA (VDRC #35697 [#1], VDRC #105678 [#2]), nej^dsRNA (BDSC #27724), bab1-gal4 (BDSC #6802). For overexpression or knockdown experiments, crosses were raised at 18°C and progeny of correct genotype were collected upon eclosure and fattened with freshly prepared wet yeast paste at 31°C before dissection. For crosses using bab1-gal4, progeny were fattened for 3-7 days. For crosses using c587-gal4, progeny were fattened for 9 days.

For enhancer/promoter dissection experiments, fragments of the dpp genomic region were amplified using the primers listed in Table S1 and inserted into the pattBLacZ vector (from K. Basler). The injection of these constructs and transgene generation were carried out by BestGene.

Immunostaining and fluorescent in situ hybridization were performed as described previously (Li et al., 2014, 2015b). The following primers were used to generate FISH probes: dpp, 5′-AGGACGATCTGGATCTAGATCGGT-3′ and 5′-ACTTTGGTCGTTGAGATAGAGCAT-3′; hh, 5′-ATTCGTCGATCAGTTCCCACGTGC-3′ and 5′-GATGGAATCCTGGAAGAGCGATCC-3′. Primary antibodies were: rabbit anti-α-Spectrin (1:3000; generated in-house), rabbit anti-pMad (1:500; Cell Signaling, 9516S), rabbit anti-Hh (1:3000; from P. Beachy, Stanford University School of Medicine, USA), guinea pig anti-Vasa (1:6000; from T. Kai), guinea pig anti-Nej (1:1000; from F. Yu, Temasek Life Sciences Laboratory, Singapore), mouse anti-α-Spectrin [3A9, 1:100; Developmental Studies Hybridoma Bank (DSHB)], mouse anti-Bam (1:5; DSHB) and chicken anti-GFP (1:5000; Abcam, AB13970). Fluorescein (FITC)-, CyTM3- and Alexa Fluor 647-conjugated goat secondary antibodies against rabbit, mouse, chicken and guinea pig primary antibodies were purchased from Jackson ImmunoResearch Laboratories. The DNA dyes used were TO-PRO-3 (1:5000; Invitrogen) or Hoechst 33258 (1:5000; Invitrogen). Samples were analyzed using a Zeiss LSM510 Meta upright or Leica SP8 inverted confocal microscope and compiled by LSM image browser or LAS X. Images were processed in Adobe Photoshop CS6 and Illustrator CS6.

Drosophila S2 cell line was obtained from the Drosophila Genomics Resource Center and cultured in Shields and Sang M3 Drosophila Insect Medium (Sigma-Aldrich) at 25°C without CO₂.

Co-IP and ChIP were performed according to previous protocols (Luo et al., 2015). For each experiment, 1×10⁷ to 2×10⁷ cells were used. The following antibodies were used: mouse anti-Flag (Sigma, F3165-1MG; 1.5 µl undiluted), rat anti-HA (Roche, 11867423001; 1.5 µl undiluted) and guinea pig anti-Nej (from F. Yu; 1.5 µl undiluted). Primers for real-time PCR are listed in Table S4. Rp49 (RpL32) was used as control for normalization.

GST and GST-En were expressed in BL21(DE3) bacterial cells, which were then lysed with lysis buffer (PBS, 50 mM Tris pH 8.0, 0.1% Triton X-100, 0.5 mM MgCl₂, 1 mg/ml lysozyme) and placed on ice for 15 min. Lysed cell suspension was then mixed with 6× loading dye at a ratio of 1:1 and incubated at 97°C for 10 min. Lysate containing equal amounts of GST or GST-En protein (quantified by Coomassie Brilliant Blue R250 staining on a gel before running the actual experiment) were then loaded onto three separate polyacrylamide gels. Protein lysate was separated by SDS-PAGE and two gels were transferred onto PVDF membrane, while the third gel was stained with Coomassie Blue as a loading control. Membranes with proteins were incubated in 8 M urea at room temperature for 15 min on an orbital shaker to denature the proteins. Proteins were then renatured using ten different concentrations of urea, which were prepared by serially diluting 8 M urea in TBS at a volume ratio of 2:3, at 4°C for 15 min each. Membrane was then rinsed in TBS and blocked at 4°C for 2 h in blocking buffer (25 mM NaCl, 10 mM MgCl₂, 10 mM Hepes pH 8.0, 0.1 mM EDTA, 1 mM DTT, 5% skimmed milk). 1 μg/ml biotin-labeled dpp2.0 or dpp2.0 (Δ2) DNA probe was added to the hybridization buffer (25 mM NaCl, 10 mM MgCl₂, 10 mM Hepes pH 8.0, 0.1 mM EDTA, 1 mM DTT, 2.5% skimmed milk) and hybridized with membrane at 4°C overnight. Membranes were then washed three times with hybridization buffer at 4°C for 15 min. Biotin-labeled DNA probes were then detected using Chemiluminescent Nucleic Acid Detection Module (Thermo Scientific) and membranes were exposed and imaged using the ChemiDoc Imaging System (Bio-Rad). The following 5′-biotin-conjugated primers were used to amplify the DNA probe from dpp2.0-lacZ or dpp2.0-lacZ (Δ2) template for far-western blotting: 5′-biotin-ATATTTCGCTACTCTTAATAGACCT-3′ and 5′-biotin-CAATCCGTTGACTGCACATCAAA-3′.

Spectrosomes were counted using a Nikon Eclipse 80i microscope with HBO mercury lamp. Results of real-time PCR were exported from ProteinAssist (Applied Biosystems). All statistical data were plotted using GraphPad Prism 7.0. P-values were calculated by unpaired t-tests in GraphPad Prism. P<0.05 was considered statistically significant. Error bars represent s.e.m.

We thank P. C. Chai for critical comments on the manuscript; Drs K. Basler, P. Beachy, D. Chen, A. Gonzalez-Reyes, J. Jiang, T. Kai, D. McKearin, T. Tabata, F. Yu, DSHB, BDSC, VDRC, NIG-Fly (Japan) and TRiP at Harvard Medical School (NIH/NIG R01-GN084947) for reagents; the Temasek Life Sciences Laboratory confocal facility and sequencing facility for support; and Dr G. H. Baeg and the members of the Y.C. laboratory for discussion.