Integration of Orthogonal Signaling by the Notch and Dpp Pathways in Drosophila

Abstract

The transcription factor Suppressor of Hairless and its coactivator, the Notch intracellular domain, are polyglutamine (pQ)-rich factors that target enhancer elements and interact with other locally bound pQ-rich factors. To understand the functional repertoire of such enhancers, we identify conserved regulatory belts with binding sites for the pQ-rich effectors of both Notch and BMP/Dpp signaling, and the pQ-deficient tissue selectors Apterous (Ap), Scalloped (Sd), and Vestigial (Vg). We find that the densest such binding site cluster in the genome is located in the BMP-inducible nab locus, a homolog of the vertebrate transcriptional cofactors NAB1/NAB2. We report three major findings. First, we find that this nab regulatory belt is a novel enhancer driving dorsal wing margin expression in regions of peak phosphorylated Mad in wing imaginal discs. Second, we show that Ap is developmentally required to license the nab dorsal wing margin enhancer (DWME) to read out Notch and Dpp signaling in the dorsal compartment. Third, we find that the nab DWME is embedded in a complex of intronic enhancers, including a wing quadrant enhancer, a proximal wing disc enhancer, and a larval brain enhancer. This enhancer complex coordinates global nab expression via both tissue-specific activation and interenhancer silencing. We suggest that DWME integration of BMP signaling maintains nab expression in proliferating margin descendants that have divided away from Notch–Delta boundary signaling. As such, uniform expression of genes like nab and vestigial in proliferating compartments would typically require both boundary and nonboundary lineage-specific enhancers.

NOTCH signaling is a metazoan mechanism that promotes different fates in adjacent cells (Artavanis-Tsakonas 1999; Barad et al. 2011; Guruharsha et al. 2012). Cell–cell signaling between membrane-bound Notch receptor and its membrane-bound ligands, Delta and Serrate/Jagged, leads to cleavage and nuclear import of the Notch intracellular domain (NICD) (Schroeter et al. 1998). In the nucleus, NICD binds the transcription factor (TF) Suppressor of Hairless, Su(H), to activate target genes via Su(H)-bound transcriptional enhancers (Fortini and Artavanis-Tsakonas 1994). This role of Su(H) is further complexified because it can recruit a Hairless repressor complex in the absence of NICD (Bang et al. 1995; Barolo et al. 2002; Maier et al. 2011; Ozdemir et al. 2014). This operation is central to diverse developmental contexts including tissue compartment boundaries, where such signaling defines adjacent epithelial domains. For this reason, Notch signaling to an enhancer is frequently integrated with tissue-specific, developmental signaling cues (Voas and Rebay 2004; Ward et al. 2006; Liu and Posakony 2012; Housden et al. 2014).

Notch-target enhancers can be characterized as either Notch instructive or Notch permissive (Bray and Furriols 2001), although other types are also evident (Janody and Treisman 2011). In the Drosophila embryo, the E(spl)m8 enhancer is a neurogenic target of an instructive Notch signal (Furukawa et al. 1995; Lecourtois and Schweisguth 1995; Schweisguth 1995). Ectopic expression of NICD in this context drives E(spl)m8 expression throughout the dorsoventral (D–V) axis, except in the mesoderm where its enhancer is inhibited by the Snail zinc finger repressor (Cowden and Levine 2002). In contrast, the Notch-target sim mesoectodermal enhancer and rhomboid (rho) neurogenic ectoderm enhancer (NEE) are driven ectopically in a way that is limited by the morphogenic gradient of the Rel-homology domain containing factor Dorsal, which cotargets these enhancers (Cowden and Levine 2002; Markstein et al. 2004; Crocker et al. 2010). Thus, in the context of the sim and rho enhancers, the Notch signal is only permissive because it is not sufficient for expression.

Wing margin enhancers, which define the border separating the dorsal and ventral compartments of wing imaginal discs, can also receive instructive or permissive Notch signals (Jack et al. 1991; Williams et al. 1994; Lecourtois and Schweisguth 1995; Neumann and Cohen 1996b). Wing margin enhancers from E(spl)m8 and cut use Notch signaling instructively, whereas enhancers from vestigial (vg) and wingless (wg) use the signal permissively (Janody and Treisman 2011). Tellingly, these enhancers respond differently to mutations of genes encoding the Med12 and Med13 subunits of the Mediator coactivator complex, which are required for Notch signaling (Janody and Treisman 2011). In clones with Med12 or Med13 deletions, Notch-instructive margin enhancers from E(spl)m8 and cut fail to drive any expression, while Notch-permissive margin enhancers from vestigial (vg) and wingless (wg) drive expression that is limited to cells close to the margin. Thus, diverse developmental enhancers encode contextual information specifying whether the Notch signal is sufficient and therefore instructive or only permissive of activation by other signaling effectors.

How an enhancer integrates Notch signaling with other signaling pathways in a developmental context is an important question. Some insight comes from studies of the nonhomologous, Notch-permissive NEEs at rho, vn, brk, vnd, and sog (Erives and Levine 2004; Crocker et al. 2008, 2010; Crocker and Erives 2013; Brittain et al. 2014). These enhancers are driven by the Dorsal morphogenic gradient patterning system of Drosophila. Activation is mediated by a pair of linked binding sites for Dorsal and Twist:Daughterless (Twi:Da) heterodimers, as well as by a separate site for the MADF BESS domain containing factor Dip3, which is important for SUMOylation of Dorsal and for Dorsal/Twist cooperativity (Bhaskar et al. 2002; Erives and Levine 2004; Ratnaparkhi et al. 2008). Notch input is mediated by a Su(H) binding site as shown by overexpression of constitutively active NICD and mutation of the Su(H) site (Markstein et al. 2004; Crocker et al. 2010). There are also conserved binding sites for the pioneer factor Zelda (Brittain et al. 2014), which primes embryonic enhancers (Harrison et al. 2011; Nien et al. 2011).

Activator sites for Dorsal, Twi:Da, Dip3, and Su(H) exhibit a constrained organization in each NEE (Erives and Levine 2004). Furthermore, Dorsal gradient readouts by NEEs are sensitive to the length of a spacer element that separates the Dorsal and Twi:Da binding sites (Crocker et al. 2008, 2010; Crocker and Erives 2013), and which is exploited in the evolutionary tuning of gradient responses (Crocker et al. 2008, 2010; Brittain et al. 2014). This functional spacer sensitivity of NEEs may involve the polyglutamine (pQ)-enriched trans-activation domains of NEE activators: Dorsal, Twi, Da, Su(H), and Zelda (see Figure 1A). Interestingly, Dorsal:Twi coactivation involves the SUMOylation system (Bhaskar et al. 2002; Ratnaparkhi et al. 2008), which can attenuate pQ-mediated aggregation (Mukherjee et al. 2009). The carboxamide side chains of glutamine (Q) and asparagine (N) allow additional hydrogen bonding, which is a key feature of pQ/pN-mediated protein aggregation. Although proteins with pathological expansions of pQ tracts ≥40 residues can self-assemble into cross-β-sheet amyloid fibers (Perutz et al. 2002a,b), proteins with shorter pQ tracts can aggregate into complexes when imported into the nucleus and brought together by a DNA scaffold (Perez et al. 1998). As such, the lengths of functional cis-spacers may modulate the degree of pQ aggregation and/or β-strand interdigitation (Rice et al. 2015).

Figure 1.

Consideration of the polyglutamine (pQ) and polyasparagine (pN) content of factors in computational genomic searches for novel Notch-signal integrating enhancers. (A) Histogram of pQ tracts for the Dpp effectors, pMad (violet) and Medea (Med, blue); the Notch effectors, Su(H) (yellow) and its coactivators, the Notch intracellular domain (NICD, green) and Mastermind (Mam, red); and the temporal patterning factor Zelda (Zld, orange). Binding sites for these factors are enriched in the nab DWME. Also shown is the pQ content for activating TFs targeting the analogous NEEs: dorsal, Twist, Daughterless (Da), Zld, and Su(H). Snail (Sna) is a NEE-targeting transcriptional repressor and is devoid of pQ content. Each pQ tract, defined as a contiguous sequence of Qs ≥ 3, is represented by a single box in a bin corresponding to its length. In contrast to the patterning factors, the selector DNA binding factors (Ap, Dll, Sd, and Hth) and cofactors (Vg) are devoid of pQ tracts, likely indicating distinct modes of regulation separate from the signaling pathway effectors (see text). Similar trends are also seen with pN (B) or with mixed polycarboxamide (X) tracts (C).

In Drosophila melanogaster, the NICD coactivator contains a nearly uninterrupted pQ tract (≥31 residues long) that is conformationally variable and functionally important (Wharton et al. 1985; Kelly et al. 2007; Rice et al. 2015). Many such pQ repeats have been identified in transcriptional activators/coactivators and some have been found to be required for synergistic activation (Wharton et al. 1985; Courey and Tjian 1988; Courey et al. 1989). Su(H)-bound NICD recruits the pQ-rich Mastermind coactivator (Yedvobnick et al. 1988; Newfeld et al. 1993; Helms et al. 1999; Schuldt and Brand 1999; Kovall 2007), as well as pQ-rich Mediator components (Tóth-Petróczy et al. 2008; Janody and Treisman 2011). We hypothesize that Notch-permissive enhancers might function through pQ-aggregated complexes that accumulate in the nucleus. Such enhancer-specific complexes might also continue forming for the duration that these pQ-rich signals are being received such that they constitute a type of trans-epigenetic memory. To understand the functional repertoire of Notch-permissive enhancers, we thus seek to identify and study model Notch target enhancers that integrate instructive morphogenic signals via pQ-rich effectors. One candidate pQ/pN-rich effector complex is Mad:Medea, which mediates Decapentaplegic (Dpp)/BMP morphogenic signaling in Drosophila (Raftery et al. 1995; Hudson et al. 1998; Wisotzkey et al. 1998). Furthermore, in several of these Dpp/BMP signaling contexts Notch signaling is also active (de Celis 1997; Steneberg et al. 1999; Crocker and Erives 2013).

Here we identify and analyze a novel, Su(H)-dependent, Notch/BMP-integrating enhancer in the BMP-inducible gene nab, which encodes a conserved transcriptional cofactor (Clements et al. 2003; Terriente Félix et al. 2007; Ziv et al. 2009; Hadar et al. 2012). The nab enhancer drives expression in two domains abutting the dorsal wing margin and flanking the stripe of Dpp expression in Drosophila. We find that this dorsal wing margin enhancer (DWME) is licensed by the selector Apterous (Ap) to read out Notch and BMP/Dpp signaling in the dorsal compartment of wing imaginal discs. Ap is a homeodomain-containing factor specifying the dorsal compartment of wing imaginal discs (Cohen et al. 1992; Blair et al. 1994). We show that the nab DWME, which has multiple Ap-binding sites, is affected by mutations of ap. We further find that Nab is required in the dorsal compartment for morphogenetic patterning of both the thorax and wings. We show that activity of the nab DWME is driven by Notch and Dpp signaling through two unpaired Su(H) binding sites and one to three Mad:Medea binding sites and propose that Dpp signaling helps to maintain enhancer activity in off-margin lineal descendants of NICD-positive margin cells. Last, we show that global nab expression is driven by the combined activity of the nab DWME, a wing pouch quadrant enhancer (QE), a proximal wing enhancer (PWE), and a larval brain enhancer (BrE), all of which function as dual enhancers and silencers. Importantly, we find that some of their Su(H) sites function in interenhancer silencing, revealing another aspect of Su(H)-targeted enhancers.

Materials and Methods

Derivation of binding motifs

Binding motifs follow International Union of Pure and Applied Chemistry (IUPAC) DNA consensus coding and were derived from position-weighted matrices from various datasets as illustrated in Supplemental Material, Figure S1 (also see Table 1). The Mad:Medea binding motif 5′-CNBYGDCGYSNV is a consensus IUPAC motif that we derived from embryonic ChIP-chip data (Bergman et al. 2005) available on JASPAR (Sandelin et al. 2004; Mathelier et al. 2015). The Zelda binding motifs, 5′-CAGGYAR, 5′-CAGGTAV, and 5′-YAGGTAR, were based on previously published characterization of Zelda ChIP-chip data (Harrison et al. 2011; Nien et al. 2011). The Apterous binding motif 5′-RYTAATKA is a consensus IUPAC motif, which we derived from bacterial one-hybrid (B1H) data from the FlyFactorSurvey project (Zhu et al. 2011). We also derived the Pangolin (dTcf) binding motif 5′-TTTGWWS from B1H data available from the FlyFactorSurvey. For the Scalloped dimer binding motif, which is necessary for recruitment of Vestigial, we realigned the known binding sites from the cut, sal, and kni enhancers (Halder and Carroll 2001) and found we could derive the tighter consensus 5′-RVATTNNNNRVATH by using the reverse complement sequences of some sites in a consensus alignment (see Figure S1). Sequences containing the above sites were identified from the 1344 conserved regulatory belts (i.e., clusters of nonprotein-coding conservation peaks) containing sites matching the Su(H) motif 5′-YGTGRGAAH. Motifs were searched using regular expression pattern matching, perl, and grep commands in a UNIX shell environment.

Table 1. Regulatory belts with sites for pQ-rich effectors of Notch/BMP signaling and pQ-deficient wing disc selectors.

Filtering step	Parameter	Output
Unique D. mel. belts with Su(H) site(s)a	5′-YGTGRGAAH	1344 beltsb
+ChIP-chip pMad binding motifc	5′-CNBYGDCGYSNV	221 belts
+ChIP-seq Zelda binding motifd	5′-CAGGYAR, 5′-CAGGTAV, or 5′-YAGGTAR	98 belts
+B1H Apterous binding motife	5′-RYTAATKA	31 belts
+Sd:Sd:Vg in vivo binding consensusf	5′-RVATTNNNNRVATH	13 belts
−Tcf/Pangolin in vivo consensusg	Missing: 5′-TTTGWWS	2 belts (nab intronic, cv-c intergenic)
+homotypic site clusteringh	≥2 Su(H) sites, or	1 belt (nab intronic)
	≥3 Mad sites, or	1 belt (nab intronic)
	≥2 Zld sites, or	1 belt (nab intronic)
	≥2 Ap sites	1 belt (nab intronic)

^aMany D. melanogaster belts map to two or more discontinuous D. virilis blocks, which were either spaced far apart in the D. virilis assembly or else were present in separate contigs.

^bThese are composed of 1157 belts with a single Su(H) site, 158 belts with two sites, and 29 belts with 3–5 clustered sites.

^cIUPAC developed from reported ChIP-chip profile (Bergman et al. 2005).

^dIUPAC developed from reported embryonic ChIP-chip profile (Harrison et al. 2011).

^eBacterial one-hybrid assay (Noyes et al. 2008).

^fConsensus matches known Sd:Sd:Vg regulatory binding sites at cut, sal, and kni enhancers (Halder and Carroll 2001).

^gTcf is a pQ-deficient transcriptional effector of Wg/Wnt signaling. Motif is based on ChIP-chip data (van de Wetering et al. 1997) and in vivo consensus (Archbold et al. 2014).

^hA comparison of the regulatory belts at nab and cv-c.

Molecular cloning

DNA fragments were amplified from either genomic DNA extracted from w¹¹¹⁸ flies (DNA → nab-C, and DNA → nab-A → nab-Ax), or DNA prepared from the Berkeley Drosophila Genome Project (BDGP) BAC clone BACR48M07 and its subcloned derivatives (BAC → nab-CDAB → nab-AB → nab-B, nab-CDAB → nab-DAB → nab-DA, and BAC → nab-CDA). All enhancer fragments were sequenced in both directions to confirm identity of clones and the absence of unwanted mutations, although the larger fragments were not necessarily sequenced through their entire lengths (see File S1 and File S2). Mutations of individual Su(H) sites were created using two-step PCR-mediated stitch mutagenesis to introduce changes as indicated in the text and sequenced to confirm these mutations. Triple mutation of the Mad:Medea binding sequences in the nab-A fragment was synthesized using gBlock synthesis (Integrated DNA Technologies) of the cloned nab-A fragment sequence in which the prominent 5′-CG dinucleotide of each site was replaced with a 5′-TT dinucleotide (see Figure S1). Outside of the introduced mutations, the sequences obtained for mutagenized clones are otherwise identical to their parent clones (nab-A, nab-AB, or nab-CDA). To sequence inserts cloned into the pH-Stinger vector, we used Stinger-fwd: (5′-ATA CCA TTT AGC CGA TCA ATT GTG C) and Stinger-rev: (5′-CTG AAC TTG TGG CCG TTT ACG).

Amplified nab intronic fragments were cloned into the XbaI site of the pH-Stinger vector (TATA-box containing hsp70 core promoter driving nuclear eGFP) (Barolo et al. 2004). The nab-A fragment was also cloned into the EcoRI site of the “−42 eve-lacZ” pCaSpeR vector. The nab intronic fragments were amplified using a high-fidelity Taq polymerase (NEB Platinum Taq mix) and the following oligonucleotide primer pairs designed from the reference iso-1 assembly: nab-A: A-fwd (5′-TGGACGCAACTGGTCTGATA) and A-rev (5′-GACCAAGGATGCGATACGAT); nab-B: B-fwd (5′-TTTCAGAAGGGGTTGAACC) and B-rev (5′-CGTATGCATAAGAAACTGGC); nab-C: C-fwd (5′-ACAAGTACAATGGACATGG) and C-rev (5′-GAAAAGATACATATGAGTAATGC); nab-Ax: A-fwd and Ax-rev (5′-CCAGCAAGGATTGCCAGG); nab-AB: A-fwd and B-rev; nab-DA: D-fwd (5′-CTCATATGTATCTTTTC, which spans reverse complement of C-rev) and A-rev; nab-DAB: D-fwd and B-rev; nab-CDA: C-fwd and A-rev; and nab-CDAB: C-fwd and B-rev. All primers included flanking restriction sites for XbaI (5′-TCTCAGA) or BsaI/EcoRI (5′-GGTCTCGAATTC).

Dissection and antibody staining

Wandering third instar larvae were dissected and fixed with 11.1% formaldehyde in PBS for 30 min. Tissue was washed for 30 min in PBT and then blocked with 1% BSA in PBT for 1 hr. Tissue was incubated with primary antibodies overnight and with secondary antibodies for 1.5 hr. After each antibody incubation, a series of washes was done for 30 min. Nuclear stained tissue was incubated with 17.5 μM DAPI in PBT for 5 min and then washed for 30 min. Subsequently the imaginal discs or larval brains were dissected from the remaining tissue and cuticle. This dissection was done on a slide in 80% glycerol. The slide was then covered with a supported coverslip and imaged with a confocal microscope.

The following primary antibodies were used: chicken anti-GFP (1:250) (Abcam: ab13070), rabbit anti-Gal4 (1:250) (Santa Cruz Biotechnology: sc-577) mouse anti-Wg (1:25), mouse anti-Ptc (1:50), mouse anti-En (1:50), and mouse anti-β-galactosidase/40-1a (1:12). The last four monoclonal antibodies were developed by S. M. Cohen, I. Guerrero, C. Goodman, and J. R. Sanes, respectively, and were obtained as concentrates/supernatants from the Developmental Studies Hybridoma Bank, created by the National Institute of Child Health and Human Development (NICHD) of the NIH and maintained by the University of Iowa. Primary antibodies were detected with Cy2-conjugated goat antichicken (1:1000) (Abcam: ab6960), Cy3-conjugated goat antirabbit (1:1000) (Abcam: ab6939), or Cy5-conjugated goat antimouse (1:1000) (Invitrogen: A10525) secondary antibodies.

Drosophila stocks and crosses

Functional analysis of nab and the nab DMWE were performed using stocks obtained from the Bloomington and Kyoto stock centers and the Vienna Drosophila RNAi Center (VDRC). The following stocks from the Bloomington Drosophila Stock Center were used: N¹ (6873), ap^GAL4 (3041), en^GAL4 UAS:Dcr (25753), UAS:GFP (6874), and the 3-kb dpp-lacZ reporter line (8404). The nab-RNAi (1607) stock was obtained from the VDRC. The following nab enhancer trap stocks were obtained from the Kyoto Stock Center: nab-NP1316 (112622) and nab-NP3537 (104533).

For the mutant Notch experiments, virgin N¹/FM7c flies were crossed with X-linked nab-A-Hsp70-GFP (DWME pH-Stinger reporter) males. An overnight collection of F₁ embryos from this cross was then sorted into two populations based on the presence or absence of GFP expression, which would indicate the presence of the FM7c balancer, which carries twist:GAL4 UAS:GFP. Again at 12 and 24 hr postcollection, a few newly GFP-positive embryos were removed from the initially GFP-negative population. These two populations of sorted embryos (±GFP) were then grown to the wandering third instar stage and dissected and stained as described.

For the mutant Apterous experiments, virgin ap^GAL4/CyO females were crossed with X-linked nab-A-GFP males. Virgin F₁ females without CyO were then crossed with ap^GAL4/CyO males. F₂ larvae with background salivary gland GFP expression, indicating presence of the pH-Stinger vector (Zhu and Halfon 2007), were then dissected and stained.

Data availability

Transgenic lines carrying all nab enhancer reporters described in this study are available upon request.

File S1 is a text file containing FASTA sequences for the entire nab-CDAB enhancer complex (reference iso-1 genome) as well as the cloned nab-A sequence. Annotated sequences for the cloned nab-A fragment and the nab-CDAB enhancer complex have also been deposited at GenBank under the accession numbers KU375573 and KU375574, respectively. File S2 is a pdf file showing an annotated alignment of all cloned sequences relative to the reference genome. Figure S1 shows our derivation of IUPAC transcription factor motifs that we used in this study. Figure S2 shows images of nab DWME-driven GFP expression in live larvae and pupae and in live dissected imaginal discs. Figure S3 shows overnight beta-gal staining of wing discs dissected from all seven independent P-element lines that we established for the nab-A (DWME) lacZ reporters. Figure S4 shows how nab DWME binding sites relate to (default) fixed nucleosome positions inferred from embryonic data from Langley et al. (2014).

Results

Identification of conserved regulatory belts with binding sites for Notch/BMP effectors and wing disc selectors

The Su(H) binding motif (5′-YGTGRGAAH) is remarkably constant across bilaterians and unique to Su(H) (Tun et al. 1994). To identify novel enhancers integrating Notch signaling with pQ-rich morphogenic effectors akin to the NEEs, we first developed a computational pipeline that identified 1344 evolutionarily conserved regulatory belts containing one or more adjacent Su(H)-binding sites (pipeline and other identified enhancers will be described in a separate article). By “regulatory belt” we mean a characteristic cluster of nonprotein coding conservation peaks consistent with a regulatory module (e.g., enhancer, insulator, and core promoter). We find that most regulatory belts span a range of 15–25 peaks of genus-wide conservation. Each such peak typically corresponds to one to three overlapping binding sites for TFs. This is consistent with enhancers requiring multiple cis-elements for both nuanced activity patterns and restricted tissue specificity. Here, we use these candidate Su(H)-targeted genetic elements to find a model enhancer that is well suited for testing hypotheses about functional enhancer grammar.

To find enhancers that are integrating both Notch and Dpp/BMP signaling, we searched for the subset of 1344 D. melanogaster regulatory belts that also contain a binding site for the pQ/pN-rich Mad:Medea complex, which mediates activation of Dpp/BMP targets (Sekelsky et al. 1995; Newfeld et al. 1996, 1997; Wiersdorff et al. 1996; Hudson et al. 1998; Wisotzkey et al. 1998; Campbell and Tomlinson 1999; O’Connor et al. 2006; Weiss et al. 2010). Specifically, we developed and searched for the subset of belts containing the IUPAC consensus motif that we derived for phosphorylated-Mad (pMad) ChIP-seq peaks (see Materials and Methods and Table 1). We also searched for sequences targeted by Zelda, which is a pQ-rich pioneer factor for embryonic enhancers that is also expressed in wing imaginal discs (Staudt et al. 2006; Harrison et al. 2011). Altogether, we found 98 unique regulatory belts containing binding sites for this pQ-rich set of activators Su(H), pMad:Medea, and Zelda (Table 1 and Figure 1).

To find enhancers that are specific to the wing imaginal disc, we refined the subset of 98 down to those that also had binding sites for three wing imaginal disc selectors: (i) Apterous (Ap), which is expressed throughout the dorsal wing compartment (Cohen et al. 1992), and (ii) Scalloped (Sd) and Vestigial (Vg), which are expressed at the dorsal/ventral compartment boundary where there is active Notch–Delta signaling (Halder et al. 1998; Halder and Carroll 2001; Koelzer and Klein 2006). By definition, selectors are TFs that are responsible for a stable binary cell fate decision (García-Bellido 1975; Lawrence et al. 1979; Akam 1998). For Ap, we derived the IUPAC consensus motif 5′-VYTAATKA from its DNA binding profile and found 31/98 regulatory belts with sequences matching this motif (Table 1). For Sd and Vg, we found that 13 of these 31 regulatory belts contain the canonical Sd dimer binding site that recruits the Sd:Sd:Vg complex (Table 1) (Halder and Carroll 2001). All three selectors (Ap, Sd, and Vg) are deficient in pQ/pN tracts unlike the graded signal-dependent effectors (e.g., see Figure 1), Mediator coactivators subunits, and TATA binding protein (TBP).

Finally, to focus on pQ-rich graded signal integration of just the Notch and BMP pathways, we set aside 11 of the 13 regulatory belts that had potential binding sites for Tcf/Pangolin. Tcf/Pangolin mediates transcriptional activation in the Wnt and β-catenin/Armadillo signaling pathway (van de Wetering et al. 1997; Archbold et al. 2014). This last filtering step left us with two regulatory belts at the nab and crossveinless c (cv-c) loci, which are located on the left and right arms of chromosome three, respectively. The gene nab encodes a BMP-induced transcriptional cofactor involved in embryonic and imaginal disc development (Clements et al. 2003; Terriente Félix et al. 2007; Ziv et al. 2009; Hadar et al. 2012), while the gene cv-c encodes a BMP-induced Rho-type GTPase activating protein (RhoGAP) involved in cross-vein morphogenesis (Matsuda et al. 2013). We later describe how RNAi knockdown of nab phenocopies crossveinless-type wings, suggesting both nab and cv-c function in a common developmental genetic pathway. In this study, we focus on the nab regulatory belt and its interactions with adjacent regulatory elements because it has more sites for each factor than the cv-c regulatory belt (Table 1).

The nab DWME drives D–V wing margin expression modulated along the anterior–posterior axis

The dense site cluster overlying the nab regulatory belt is located in the long first intron of nab in Drosophila (thick bar underline in Figure 2A). This same belt is part of a smaller 2.7-kb intronic fragment containing several regulatory belts conserved in both D. melanogaster and D. virilis (see boxed region in Figure 2A). We cloned the predicted 764-bp nab-A fragment and found it corresponds to a DWME with augmented expression in regions corresponding to peak pMad levels (Figure 2, B–L). (Note: this fragment is 764 bp long in the reference iso-1 genome, but is 762 bp long in the cloned fragment due to polymorphic indels in two separate poly-A tracts.) The nab-A fragment was tested using two different core promoters with different reporter transgenes: (i) a 270-bp minimal hsp70 core promoter–eGFP–nls fusion, and (ii) a 205-bp minimal eve core promoter–lacZ fusion (Figure 2C). These constructs also tested the nab-A enhancer fragment in both orientations relative to these core promoters (Figure 2, B and C). We generated seven and five independent P-element integrations of the lacZ and eGFP–nls reporter constructs, respectively, and found all 12 to drive the same expression pattern.

Figure 2.

A highly conserved regulatory block in the Drosophila nab locus is a robust dorsal wing margin enhancer (DWME). (A) Diagram of the nab locus from D. melanogaster showing sites matching binding motifs (color-coded boxes on separate tracks in large intronic box) within a 2.7-kb intronic block that contains several regulatory belts conserved across the genus. We cloned a predicted 764-bp region, the nab-A fragment (thick bold underline), for having one of the highest concentrations of sites for Su(H), Mad, Zelda, Apterous (Ap), and the Scalloped (Sd) and Vestigial (Vg) complex (Sd:Sd:Vg) without having any Tcf sites. (B and C) We cloned the nab-A fragment in front of two core promoter reporter genes: an hsp70 core promoter fused to an eGFP-nls reporter gene in a gypsy insulated construct (B), and an eve core promoter fused to the lacZ reporter gene (C). Independent P-element transgenic lines with each reporter cassette gave identical expression patterns despite differences in integration sites, core promoters, and enhancer orientation relative to the core promoter (arrows in enhancer box, and + or − signs in parentheses). The cloned 762-bp fragment is 2 bp shorter than the reference sequence due to single base pair contraction polymorphisms in two separate poly-A runs. (D–F) Expression from the nab-A (DWME) EGFP reporter along the dorsal wing margin. Different color channels indicate DAPI (D, cyan); GFP (E, green); and Wingless (Wg) (F, magenta), which mark the D–V compartment margin, a ring around the wing pouch, and a broad stripe across the proximal part of the wing disc. (G) Merged images of D–F. (H–J) Additional discs double-labeled with antibodies to Wg (H), and Ptc (I), and En (J), which mark the A–P margin and posterior compartment, respectively. The anterior DWME expression pattern is characteristically longer than the posterior compartment side, which in turn stretches deeper into the dorsal compartment than the anterior compartment expression pattern. (K–M) DWME-driven lacZ reporter activity in late third instar discs (K and L) or in germ-band extended embryos (M). DWME-driven expression was detected with overnight X-gal staining (K) or with a digU-labeled antisense lacZ RNA probe (L and M). Both wing (K and L) and haltere (L) discs show the characteristic twin spots of DWME-driven activity, while the embryonic expression is detected in a subset of lateral neuroblasts.

We find that the nab-A fragment works equally well across these different reporter constructs in live dissected and undissected discs (Figure S2), in fixed and double-stained imaginal discs (Figure 2, D–J), in fixed discs stained for β-gal expression (Figure 2K and Figure S3), and in fixed discs incubated with an antisense lacZ RNA probe (Figure 2L). In situ hybridization with the same antisense lacZ RNA probe also shows that the nab-A fragment drives expression in stage 9/10 long germ-band-extended embryos in what may be a subset of embryonic neuroblasts, a known site of nab expression (Clements et al. 2003).

In the wing imaginal disc, the nab DWME drives expression from the D–V border region and into the dorsal compartment by several cells but only in two spots roughly centered in the anterior and posterior compartments. These spots occur in regions with peak pMad levels as previously shown (Tanimoto et al. 2000; Restrepo et al. 2014). The nab DWME also drives expression in a highly stereotyped pattern that is unique to each compartment. In addition to the thin interrupted row of expression along the D–V margin, expression occurs as an elongated anterior spot and a broader mitten-shaped posterior spot abutting the margin (e.g., compare anterior and posterior spots on the left and right sides of each disc in Figure 2K). This robust DWME-driven expression pattern is consistent with our desired goal of finding a wing disc compartment-specific enhancer integrating two orthogonal developmental signals: (i) a D–V Notch margin signal and (ii) a graded anterior–posterior (A–P) BMP morphogenic signal. Because endogenous nab is expressed in a broader wing pouch pattern than what we observe for the DWME-driven reporters, below we address how the DWME relates to other possible enhancers at nab. We also address how some of the predicted sites contribute to DWME activity and how the DWME contributes to Nab function in wing development.

The first intron of nab harbors four separate enhancers

To understand the role of the nab DWME activity in the overall nab expression pattern, we made a series of constructs containing various intronic fragments that we cloned from the first intron of the nab locus (Figure 3A). These were tested by anti-GFP antibody staining of dissected larval tissues from independent P-element integrated lines as well as pools of these lines (see Materials and Methods). Consistent with the presence of insulators in the pH-Stinger vector, independent lines of the same construct recapitulated each other’s expression patterns.

Figure 3.

The Drosophila nab harbors several regulatory modules functioning as both enhancers and mutual silencers/attenuators. (A) We cloned and tested the indicated series of nab intronic fragments to understand the regulatory logic of nab’s expression in wing imaginal discs. The colored boxes follow the key in Figure 2 except Zelda sites are not shown for clarity. To refer to the different enhancers we identified, we defined four different intronic regions as the nab-C, nab-D, nab-A (DWME), and nab-B fragments. We also labeled the four best matches to Su(H) binding motifs (sites S1–S3) and mutated these sites in the indicated constructs (S1, S2, or S3 MUT and Xs in construct). These regions can be understood as having four major enhancer activities: the dorsal wing margin enhancer (DWME); a larval brain enhancer (BrE), which drives dense expression in neuronal lineages; a wing imaginal disc quadrant enhancer (QE), which complements the DWME activity to match the endogenous expression pattern; and a proximal wing enhancer (PWE). ++ indicates strong expression, + indicates weak expression, while maintaining the indicated pattern. * means there is a more nuanced description of the expression in the main text. Both the sizes and the direction of the cloned insert relative to the core promoter are shown for each construct. (B) Each cloned enhancer was found to drive a distinct expression activity associated with the endogenous nab locus, a distinct ectopic activity, or expression level that was not associated with the endogenous nab locus, and a distinct silencing or attenuation activity acting on ectopic expression patterns and levels (red repression symbols connecting one silencer/attenuation activity in one enhancer to the ectopic activity/level in another intronic enhancer).

After comparing a PGawB nab GAL4 enhancer trap line (nab^NP1316, insertion site 48 bp upstream) (Hayashi et al. 2002) with the various constructed transgenic reporter lines, we discovered that we identified separate enhancer activities for all known tissues in which nab is expressed (Figure 4A vs. Figure 4, B–I). The expression pattern of the endogenous nab locus in late third instar wing imaginal discs (Figure 4A) is driven by a combination of (i) the DWME activity from the 764-bp nab-A fragment (Figure 4B), (ii) QE activity from a 1.4-kb nab-C fragment (Figure 4C), and (iii) PWE activity from the combined 2.4-kb nab-CDA fragment (triple arrows pointing to dispersed expression pattern in Figure 4D). The PWE activity, which we also find in both the endogenous enhancer trap line (see triple arrows in Figure 4A) and a larger 2.7-kb nab-CDAB fragment (Figure 4E), likely requires elements present in both nab-C and nab-D fragments because neither the nab-C (Figure 4C) nor the nab-DA fragment (Figure 4H) recapitulates this pattern on its own.

Figure 4.

Activities of dissected regulatory modules from the nab intronic enhancer/silencer complex. (A–I) GFP reporter expression driven by nab-GAL4>UAS-eGFP (A) or by different nab enhancer modules driving eGFP-nls (B–I). The first row depicts third instar wing imaginal discs; the middle row (single prime) depicts additional third instar discs labeled for halteres (h), leg (l), and wing (w) discs; and the third row (double prime) depicts expression patterns in third instar larval brains, if any. (A–I) The endogenous nab expression pattern in wing imaginal discs (A) appears to be the result of discrete activities from: the nab-A fragment, which contains the dorsal wing margin enhancer (B); the nab-C fragment, which contains a wing pouch/quadrant enhancer (C); a nondescript proximal wing disc enhancer seen with most intronic fragments containing the nab-D region (triple arrows in D and E); and the nab-B fragment, which contains a hinge enhancer that functions in the absence of the nab-C fragment (arrowheads in F–I). Note that the nab-C fragment is missing expression at the margin and has more uniform levels of expression throughout the pouch relative to either the nab-GAL4 enhancer trap (A) or fragments containing both the nab-C and nab-A regions (empty arrow in C). (A′–I′) Fragments with activities from both the nab-A and nab-C fragments recapitulate in endogenous nab expression in the haltere disc. (A′′–I′′) Most of the endogenous expression of nab in the larval brain and ventral nerve cord (A′′) is recapitulated by nab intronic fragments carrying the nab-B fragment (E′′–H′′) but it is noticeably weaker by itself (F′′). This pattern corresponds to dense expression in neuronal lineages and four cells in the posterior tip of the ventral nerve cord. The nab-C fragment has ectopic brain activity that is completely silenced by activities present in the nab-DA fragment (see entire brain in D′′ and optic lobes and posterior ventral nerve cord regions in E′′). The nab-AB fragment also drives a strong ectopic leg disc ring pattern (see G′ and H′) that is repressed in the presence of the nab-C fragment (see E′). Similarly, the dorsal wing margin activity of nab-A (B) is silenced in the presence of nab-B (see empty arrows pointing to margin in nab-AB disc in G) except when nab-D is also present (see nab-DAB disc in H). The S3 Su(H) binding site in nab-B appears to mediate repression of the wing hinge activity inherent to nab-B and the dorsal wing margin activity inherent to nab-A because there is augmentation of both of these patterns in the nab-AB S3 mutated construct (I, arrows point to twin spots along the dorsal margin of the compartment boundary, and arrowheads point to expanded hinge pattern). This same site is not absolutely required for the nab-B brain activity (I′′). Neither the nab-DA nor the nab-A are able to drive any brain expression (representative blank nab-DA disc is shown in B′′), strongly suggesting that the nab-B fragment is necessary and sufficient for the overall gross expression pattern of nab in the brain. (B′′′ and C′′′) Shown are blown-up images of the nab-A and nab-C activities in the wing disc (green) double labeled for Wg (magenta).

The robust expression of the endogenous enhancer trap reporter in the dorsal margin of the wing pouch (single horizontal arrow in Figure 4A) is recapitulated in all reporter lines carrying the nab-A fragment (Figure 4B and single horizontal arrows in Figure 4, D and E) with one exception summarized in the next section. Furthermore, this dorsal margin activity is absent in the nab-C fragment, which only drives a quadrant pattern in wing imaginal discs (see hollow arrow pointing to gap of expression at margin in Figure 4C). The combined nab QE + DWME expression pattern, which we see in the nab-CDA and nab-CDAB reporters and the nab^NP1316 P{GawB} enhancer trap (Figure 4, A, D, and E), is also recapitulated by a second P{GawB} enhancer trap (nab^NP3537, data not shown), which has an insertion site 1 bp upstream of nab but in the opposite orientation as nab^NP1316, and a third, previously characterized P{GabB} nab enhancer trap line, S149, which has a reported insertion site 23 bp upstream (Gerlitz et al. 2002; Ziv et al. 2009; Hadar et al. 2012).

We also observed that all nab enhancer fragments driving wing pouch expression patterns (DWME and QE) also did so in the haltere imaginal discs (“h” discs in single prime lettered panels of Figure 4). This suggests to us that nab may play a role in both wing and haltere development. This could indicate that this function predates evolution of haltere balancing organs from wing discs in dipteran ancestors (Lewis 1978; Weatherbee et al. 1999).

The endogenous enhancer trap reporter is expressed in entire neuronal lineages of the larval brain and in four distinct neurons/cells at the posterior tip of the ventral nerve cord (Figure 4A′′). We find that a 292-bp nab-B fragment, located immediately downstream of nab-A (Figure 3A), drives the same pattern but at lower levels than the endogenous locus (Figure 4F′′). These lower levels of expression are rescued by larger fragments containing the nab-B region: nab-AB (Figure 4G′′), nab-DAB (Figure 4H′′), or nab-CDAB (Figure 4E′′) fragments. This suggests that the minimalized larval BrE present in the nab-B fragment normally functions with additional elements that span into the adjacent nab-A fragment. Nonetheless, a 693-bp truncated nab-A fragment (see nab-Ax fragment in Figure 3A), which is missing a conserved sequence that is present on the side flanking the nab-B fragment, does not drive any detectable activity in either imaginal discs or larval brain (data not shown). This might signify that the conserved block present in nab-A and separating it from the nab-B fragment is required for both the DWME and BrE activities. Alternatively, BrE boosting elements may be located elsewhere in the nab-A fragment.

The nab enhancers also function as mutual silencers

For each of the four distinct and separable enhancer activities in the first intron of nab, we also find distinct silencer activities (Figure 3B). For example, while the nab-C fragment does not recapitulate the larval BrE activity, it does drive ectopic expression in a more superficial set of larval brain cells, possibly neuroepithelial cells or glia, and some distinct neurons in the ventral nerve cord (Figure 4C′′). This ectopic activity associated with nab-C is not seen from the endogenous locus (Figure 4A′′) or from any fragment that also contains the nab-DA region (Figure 4, D′′ and E′′). This suggests that the ectopic larval brain activity endowed by nab-C is silenced by elements present in nab-DA (Figure 3B).

In a second example, we find that the nab-AB enhancer is a potent ring enhancer in leg imaginal discs (“l” disc in Figure 4G′), which is an activity that is not seen for the nab enhancer trap reporter (l disc in Figure 4A′). This activity is attenuated (i.e., made less robust) in the nab-DAB fragment (l disc in Figure 4H′) and is completely silenced in the larger nab-CDAB fragment. This suggests that the ectopic leg disc activity endowed by nab-AB is silenced by elements spanning nab-C and nab-D (Figure 3B).

In yet a third example, which we previously mentioned as an exception to the presence of DWME activity in all fragments containing nab-A, we find that the nab-AB fragment obliterates dorsal wing margin expression (arrows in Figure 4G). However, this same silencing activity is only able to attenuate the DWME in the context of the larger nab-DAB fragment (Figure 4, H and “w” disc in H′). These results suggest that the nab-B fragment is simultaneously a larval brain enhancer and a dorsal wing margin silencer/attenuator. In a related fourth example, the nab-D fragment also appears to modulate and attenuate nab-A’s DWME activity in the context of the nab-DA fragment (Figure 4I′).

A fifth example of dual enhancer/silencer activity is intertwined with the above third example of the BrE’s wing disc silencing activity. The nab-AB fragment drives ectopic expression in the wing imaginal disc hinge region near the stripe of peak Dpp (arrowheads in Figure 4G). This activity is attenuated in the nab-DAB fragment (see arrowheads in Figure 4, H and “w” disc in H′) and completely silenced in the nab-CDAB fragment (Figure 4E).

Thus, we find that all of the distinct nab enhancer fragments possess both endogenous (i.e., nab-related) enhancer activities and ectopic enhancer activities. However, these same enhancers also function as interenhancer silencers of ectoptic activities and attenuators that modulate their nonectopic activities. These activities are summarized in Figure 3B. Later, we describe finding that at least one of the intronic Su(H) binding sites involved in inducible activation is also involved in interenhancer silencing and attenuation.

Activity of nab DWME requires Notch signaling and Su(H) sites

Single nucleotide transversions to any of the core positions in the invariant Su(H) binding sequence, 5′-YGTGRGAA (core positions underlined), are sufficient to eliminate binding in vitro (Tun et al. 1994). Therefore, to verify that the nab DWME is targeted by Notch signaling via its two separate (unpaired), canonical Su(H) binding sites, we compared DWME activity from nab-A fragments containing either wild type Su(H) sites, a mutated S1 site, or a mutated S2 site (Figure 5, A–E).

Figure 5.

The nab DWME is induced by Notch signaling via its Su(H) sites. (A) The DWME contains two Su(H) sites, S1 and S2, and drives a unique stereotypical pattern. To test the role of these sites, we mutated each individually in the minimalized nab-A enhancer fragment. (B) The S1 mutated nab-A fragment exhibits no expression in wing imaginal discs. (C) The S2 mutated nab-A fragment results in discs with either no expression or weak dorsal margin expression as shown in these representative discs. (D) The nab-CDA fragment with an S1 mutated Su(H) exhibits reduced expression at the D–V compartment margin (arrow). (E) Quantitative comparison of wild-type, S1 mutated, and S2 mutated discs according to levels of expression. (F) Quantitative comparison of nab-A DWME activity in wild-type or mutant Notch backgrounds. Also shown is representative DWME reporter activity from discs in wild-type (G) or mutant Notch (H) backgrounds.

The central S1 site is the only sequence within the nab-A fragment that matches the motif 5′-YGTGRGAAH, which we used in our computational screen. We found that the reporter lines with the mutated S1 site (5′-TGTGAGAAT → 5′-TcatAGAAT) lack expression in both wing and haltere discs (Figure 5, B and E). When we mutated the S1 site in the context of the nab-CDA fragment (Figure 3A), we found diminished expression at the margin similar to the nab-C fragment (compare Figure 5D to Figure 4C). Thus, the Su(H) S1 site is absolutely necessary for nab DWME activity in both the cells along the dorsal margin and the anterior–posterior compartment cells located farther into the wing pouch.

The flanking Su(H) S2 site matches the consensus motif 5′-YGTGDGAAH, which is slightly less stringent than the genome screen motif 5′-YGTGRGAAH (Figure 3A). Nonetheless, this S2 site is still conserved across Drosophila. We found that reporter lines with the mutated S2 site (5′-CGTGTGAAA → 5′-CcatTGAAA) drive diminished expression that is collapsed to two small attenuated spots along the margin near the predicted peak pMad regions (Figure 5, C and E). Thus, the flanking S2 Su(H) site is also important for robust activation along the dorsal wing margin as well as in the off-margin spots.

To confirm that the Su(H) sites work canonically in Notch-dependent inducible activation, we also compared DWME activity from the nab-A-driven reporter in both wild-type and mutant Notch backgrounds (Figure 5, F–H). These experiments show that whereas wild-type Notch allows robust DWME activity in nearly 100% of wing imaginal discs, a Notch mutant background (N¹) results in ∼60% of the discs not having any expression at all and ∼40% of the discs having very weak expression close to the margin (Figure 5, F and H). Thus, the nab DWME requires both wild-type Notch signaling and both of its Su(H) binding sites for robust expression in the wing pouch. This signaling is also required for the off-margin expression in the anterior–posterior compartment “spots,” suggesting that BMP input into the DWME is not independent of Notch signaling.

Su(H) binding sites are involved in interenhancer silencing

In the absence of NICD, Su(H) recruits Hairless and establishes a repression complex at many targeted enhancers (Bang et al. 1995; Barolo et al. 2002; Maier et al. 2011). Moreover, there is evidence for long-range looping of multiple Notch/Su(H)-targeted promoters from the E(spl) gene complex into an interaction domain associated with cohesin and the PRC1 polycomb repression complex (Schaaf et al. 2013). We thus hypothesize that some of the predicted Su(H) binding sites involved in Notch-dependent inducible activation might also be involved in interenhancer silencing or attenuation.

To test the hypothesis that the S1 Su(H) site within the DWME is responsible for silencing the ectopic larval brain activity associated with nab-C, we mutated this site in the nab-CDA fragment (Figure 3A). We find that mutation of this site does not result in ectopic expression in larval brains (data not shown). Nonetheless, it is still possible that the S2 site and/or additional lower-affinity sites, which we did not mutate, might participate singly or collectively in an intronic interenhancer repression complex featuring Su(H).

To test the hypothesis that the strong S3 Su(H) site present within the BrE is responsible for the silencing of the DWME, we mutated this site (5′-TGTGAGAAC → 5′-TcatAGAAC) in the nab-AB fragment (Figure 3). Normally, the DWME activity is completely silenced in the wild-type nab-AB fragment, and partially attenuated in the nab-DAB fragment. We find that mutation of this S3 Su(H) site in nab-AB rescues the twin spots of DWME activity in about half of the discs (16/30 discs), although at levels much diminished relative to the wild-type nab-A fragment (arrows in Figure 4I). Moreover, we also find that the hinge activity is much expanded along both the A–P axis and along the proximal–distal axis of the imaginal discs (Figure 4I). This suggests that the S3 Su(H) site is involved in both intraenhancer repression and interenhancer silencing/attenuation.

Apterous licenses the nab DWME to receive Notch/BMP signals in the dorsal wing compartment

We propose that selectors are not likely to act via recruitment of pQ-rich Mediator subunits as do the pQ-rich signaling effector activators. This is another reason to eschew the use of “activation” to describe the role of selectors and simply refer to their action as “licensing” (see Discussion). Motivated by a hypothesis of selector licensing of signal effector integration by enhancers, we used the binding motifs for the pQ-deficient selectors Ap, Sd, and Vg to identify the nab DWME as described above. In addition to the cluster of Ap binding sites, and the single Sd:Sd:Vg site, we also find a binding site for the homeotic selector Distalless (Dll) (5′-ATAATYAT), which has a similar expression pattern to Sd and Vg at the wing margin (Campbell and Tomlinson 1998). We thus suspect that the DWME might be licensed collectively by Sd, Vg, Dll, and Ap. Furthermore, like Ap, Sd, and Vg, Dll is also deficient in pQ and pN content.

While it will take future studies to show how multiple selectors might act collectively to regulate the nab DWME, we did seek to test the role of Ap, which has many candidate binding sites throughout the DWME. Expression of the selector Ap is restricted to the dorsal compartment of the wing pouch and the proximal parts of the wing imaginal disc (Cohen et al. 1992; Blair et al. 1994) (Figure 6A). In the absence of Ap, the DWME might be refractive to Notch and Dpp signaling, which are active in many contexts that do not feature nab DWME activity. To test the hypothesis that the nab DWME is licensed in part by Ap to respond to Notch and BMP developmental signals only in the dorsal wing compartment, we crossed our nab DWME reporter line into backgrounds differing in the dosage of normal Ap. As the levels of normal Ap is reduced, we observe that DWME-driven GFP reporter expression collapses to a smaller and smaller cluster of cells along the D–V border of each A–P compartment (Figure 6, B–D). This result is consistent with Ap being required for both robust expression along the margin and maintenance of DWME activity in off-margin descendants of margin cells.

Figure 6.

Apterous licenses the nab DWME to be receptive to Notch and Dpp signaling in the dorsal compartments of wing imaginal discs. (A) Cartoon of Ap expression (blue) in the dorsal compartment of a third instar wing imaginal disc. Wingless (Wg) expression (magenta) is found at the D–V compartment boundary in the wing pouch. (B–D) Wing imaginal discs stained for Wg (magenta), Gal4 (cyan), and GFP (green) in a line carrying the nab-A (DWME) eGFP reporter in a wild-type Ap background (B), in a heterozygous ap-GAL4 hypomorph caused by a GAL4 element integration (C), and a homozygous ap-GAL4 hypomorph (D). Note as the intensity of the Gal4 signal increases, the nab DWME-driven GFP signal decreases, which is consistent with its many Ap binding sites. (E and E′) The same ap-GAL4 hypomorph was used to drive nab RNAi to knock down expression in the dorsal compartment via both RNA and enhancer interference. Defects include an overly creased wing at wing vein II (arrows in E), a delta-like patterning defect where wing veins I and II intersect (arrow in E′), and occasionally missing or diminished anterior cross-veins (*acv). (F) Wild-type wing for comparison to E. (G) Wild-type notum (n) and scutellum (sc) with macrochaetes circled. (H) One balanced copy of the ap-GAL4 does not affect macrochaete patterning nor development of the scuttelum. (I–K) When ap-GAL4 drives nab-RNAi, adult flies develop grossly misshapen scutellums with severe macrochaete patterning defects. Three representative thoraxes are shown.

Dorsal-compartment expression of nab is required for normal patterning of wing and thorax

Expression of nab in the dorsal compartment of the developing wing pouch is the result of cells using DWME and/or QE activity. To understand better the role of the DWME vs. the QE, we used the ap-GAL4 hypomorph (ap^GAL4) to drive nab RNAi. This likely brings down nab mRNA levels in the dorsal compartment while also specifically disrupting the Ap-dependent activity of the endogenous nab DWME. Thus, in this experiment, the most severe disruption of nab expression is expected to be in the DWME-active cells.

Using one copy of the ap^GAL4 allele to drive expression of UAS:nab-RNAi, we observe the following set of phenotypes. We observe an exaggerated creased wing along vein II (Figure 6E), which may be consistent with differences in Nab signaling between dorsal and ventral wing epithelia. This is often accompanied by a malformed delta-like junction between the first and second wing veins at the anterior margin (Figure 6E′). Sometimes this is accompanied by a disrupted bristle pattern at the juncture of the margin (Figure 6E′ vs. Figure 6F, wild type). In addition, the scutellum of these flies is malformed and diminished in size with typically one or more anterior and posterior scutellar bristles missing (compare Figure 6, G and H to Figure 6, I–K). In these flies, the anterior and posterior dorsocentrals in the notum are also frequently absent (Figure 6, I–K). Thus, there are missing or mis-specified macrochaetes in both the notum and scutellum. Wings from ap^GAL4:nab-RNAi adults show other defects including occasional ectopic wing veins, gaps in the wing vein patterning, and diminished or absent anterior cross-veins (data not shown). Therefore, we propose that normal patterns and/or levels of nab are required in the dorsal margin cells to ensure proper wing vein specification and patterning.

The results of dorsal compartment-specific knockdown of nab suggests that expression levels are critical to morphogenetic pattering of the notum and wing veins (Figure 6). To explore further the importance of nab expression in wing imaginal discs, we expressed nab-RNAi in two additional ways. First, we used an endogenous GAL4-nab enhancer trap line (nab-NP3537) to drive the same nab-RNAi transgene. The prediction is that in this experiment, both endogenous Nab and Gal4 proteins are being synthesized simultaneously as the result of the different regulatory enhancers at nab inducing transcription of both nab and GAL4. Thus, this experiment might be expected to result in only a partial knockdown of Nab protein because of the delayed transcriptional/translational cycle of Gal4-driven nab-RNAi. In this experiment, we observe the mildest effect of all the nab-RNAi experiments in that we only observed one phenotype: occasional loss of the posterior cross-veins but only in adult females (see arrows in Figure 7, B and C).

Figure 7.

Distinct levels of nab expression are required for normal developmental patterning of the notum, scutellum, and wing veins. Wing vein patterning in female (A–D) and male (E–H) fly wings carrying a UAS-nab-RNAi transgene without a GAL4 driver (A, E, and the red wing outline in G), a nab locus GAL4 enhancer trap driver (B, C, F, and blue wing outline in G), or an en locus GAL4 enhancer trap driver augmented with additional UAS-dicer expression (D and H). Posterior cross-veins are often lost or incomplete in female wings with nab-RNAi knockdown arrows (B–D). The fourth wing vein is often lost in male wings (H). Thus, normal patterns of nab expression in both the dorsal and ventral compartments are crucial to wing vein patterning.

To test the role of nab expression in other wing disc compartments, we used an engrailed (en) GAL4 enhancer trap line to drive nab-RNAi in the posterior compartment, which spans both D–V compartments (see Material and Methods). En is a selector protein expressed early in the posterior compartments of wing imaginal discs, where it determines posterior-compartment identity (growth/size, shape, and wing vein patterning) (Lawrence and Morata 1976; Kornberg 1981; Blair 1992; Brower 1986). Importantly, in the en-GAL4 experiment, nab-RNAi would be induced and maintained prior to transcription of nab in early third instar wing imaginal discs (Terriente Félix et al. 2007). To increase the severity of any phenotypes, we used the en-GAL4 transgene to drive both UAS-nab-RNAi and UAS-dicer. This experiment results in a more severe wing vein patterning defect in both females and males (Figure 7, D and H, respectively). This phenotype, which includes incomplete and missing longitudinal veins as well as cross-veins, is restricted to the posterior compartment. We do not see any morphological defects in scutellar development, suggesting that ap hypomorphs sensitize scutellar development to loss of Nab function.

In summary, nab-RNAi knockdown experiments demonstrate that nab expression is important in all compartments of wing imaginal discs, including both proximal and distal parts of discs, for the normal developmental patterning of the scutellum, thoracic macrochaetes, and wing veins. This is consistent with nab’s intronic complex of three wing imaginal disc enhancers (QE, PWE, and DWME).

The nab DWME is a lineage-specific enhancer maintaining expression in off-margin clonal descendants of margin cells

To better understand the individual roles of the nab-A binding sites mediating Notch and Dpp signaling, we examined the developmental onset of expression in wing imaginal discs. We find that DWME-driven expression begins around late second/early third instar discs in a few cells at the margin (Figure 8, A and B). This is consistent with the earliest reported detection of Nab protein in early third instar discs (Terriente Félix et al. 2007). As the discs develop, there are many examples of dividing daughter cells in which one of the daughter cells is dividing away from the D–V boundary while still maintaining equivalent amounts of expression (see circled pairs of cells in Figure 8, A′–D′ and I′). This is not simply due to stable GFP inheritance in margin daughter cells because these dividing pairs of cells can also be seen at much farther distances from the margin in later stages (Figure 8I′). We also observe decreasing DWME-driven reporter activity in early-to-mid third instar cells along the margin when peak pMad is still centralized as a single (i.e., nonbimodal) peak (see series of arrows of decreasing size in Figure 8I′).

Figure 8.

The nab DWME is stably maintained in off-margin clonal descendants. (A–L) GFP reporter activity driven by the nab DWME (nab-A fragment) starting during late second instar to late third instar wing imaginal discs are shown arranged according to the number of expressing cells based on nuclear GFP intensity. (A′–D′, H′, and I′) Zoomed-in images. Inspection of recently divided daughter cells (e.g., ovals in A′–D′ and I′) demonstrate that GFP activity is equally maintained in both cells even when the direction of cytokinesis places one of the daughter cells farther away from the dorsal compartment boundary. Cells far removed from the D–V boundary but still near the A–P compartment boundary display robust GFP expression (large arrows in H′). However, cells farther away from the A–P compartment boundary display decreasing levels of GFP expression even when they are on the dorsal margin itself (smaller arrows in H′ and I′). During early–late third instar (J and K) and late–late third instar/prepupal (L) discs, the DWME activity is maintained in clonal descendants of dorsal margin cells that are now located deep into the wing pouch. We thus propose that the DWME functions as a lineage-specific margin enhancer that maintains expression in dorsal off-margin clonal descendants. The disc in J is doubled stained for β-gal (magenta) driven by a 3-kb dpp enhancer (see Material and Methods). (M) The nab QE (nab-B fragment) is noticeably weaker or not active in dorsal compartment off-margin cells (arrowheads) from late third instar/prepupal discs. (N) DWME fragment in which all three Mad:Medea binding sequences have been mutated (5′-CG to 5′-TT mutations at each site) drives diminished expression in both margin and off-margin cells. Two representative discs are shown.

This temporal profile of nab DWME activity suggested to us that the DWME marks a clonal patch of cells descended from a margin cell, which experienced early Notch–Delta signaling (see Figure 9B and Discussion). This would also explain a role for the Dpp input in stabilizing activity of the DWME in clonal descendants removed from the border of active Notch–Delta signaling. Like clonal fate mapping in wing imaginal discs (Dahmann and Basler 1999), cells expressing DWME-driven GFP are patch-like clusters of dividing cells that respect both D–V and A–P compartment boundaries. Consistent with this idea, DWME-driven patches of GFP expression resemble the D–V-oriented cell proliferation patterns made by marked clones in wing imaginal discs (González-Gaitán et al. 1994). To explore this idea further, we also looked at the nab QE (nab-C) reporter in late third instar/prepupal wing imaginal discs specifically to see if the deficiency in dorsal margin expression becomes more noticeable over time. We find that this is possibly the case as the margin gap appears more pronounced than earlier stages (compare arrowheads in Figure 8M to Figure 4C or Figure 4C′′′).

Figure 9.

An enhancer model featuring selector licensing for pQ-mediated signal integration. (A) Shown are the trascription factor binding site motif distributions for the nab DWME (left) and the cut WME (right), both of which are induced by Notch signaling and Su(H) binding sites. As much as possible, matches to the indicated motifs are shown on separate tracks for ease of visualization. Tick marks represent 100-bp intervals. Boxes represent the minimalized enhancers, but in both cases, augmented expression is seen with larger fragments. (B and C) Below each enhancer are models of how each enhancer works in the context of both homeotic licensing and graded pQ signal integration or lack thereof. In both examples, selectors are envisioned as allowing certain cells (cell without Xs) to be receptive to transcriptional effectors of signaling pathways (Notch, Dpp, and Wg). The pQ/pN-rich Mad:Medea complexes are envisioned as stabilizing the active DWME in daughter cells of margin cells (B). In contrast, no such stabilizing effect is envisioned for the cut WME, which lacks both Mad:Medea sites (purple boxes) and instead has only Tcf (green boxes) and Su(H) (red boxes) binding sites (C). Thus, this enhancer would become inactive in margin daughter cells that divide away from the margin border, where there is active Notch–Delta signaling.

The DWME temporal profile suggests that the nab DWME and QE are complementary lineage-specific enhancers that have evolved to deliver Nab expression in proliferating wing pouch cells more uniformly than what the QE alone could drive. To test the hypothesis that the identified Mad:Medea binding sequences in the DWME are critical for maintaining expression in the margin gap evident in QE reporter activity in late third instar, we mutated all three Mad:Medea sites in the nab-A (DWME) fragment by changing the motif-prominent 5′-CG dinucleotide in each binding sequence to 5′-TT (see Figure S1). We find that these mutations result in substantially weakened expression in both the margin cells and in the off-margin descendants of margin cells (Figure 8N). This suggests that the Dpp/BMP signaling input is required for both synergistic activation with Notch signaling at the margin and for stabilization of DWME activity in off-margin descendants. Furthermore, it also suggests that the DWME functions to rescue a QE “signaling hole.”

Discussion

We used a comparative genomics approach to identify and characterize a novel enhancer mediating tissue-compartment specific activation based on the integration of two signaling pathways: Notch and BMP/Dpp. This approach led us to a dense intronic cluster of binding sites for the desired activators and selectors at the Drosophila gene nab. “Nab” was named for “NGFI-A binding” and is present in vertebrates as Nab1 and Nab2, where they work as corepressors of the C₂H₂ zinc finger factors Egr1 (NGFI-A/Krox24) and Egr2 (Krox20) to control proliferation, patterning, and differentiation (Russo et al. 1995; Svaren et al. 1996). Similarly in the nematode Caenorhabditis elegans, this gene encodes MAB-10/NAB, which is a corepressor of the LIN-29/EGR C₂H₂ zinc finger factor and is involved in cellular differentiation (Harris and Horvitz 2011). In Drosophila, LIN-29 functions appear to be maintained across three paralogous genes encoding C₂H₂ zinc finger factors: rotund (rn), squeeze, and lin19 (Vilella et al. 2009). Correspondingly, Drosophila Nab functions as a corepressor of Rn in wing imaginal discs and as a coactivator of Squeeze in a subset of embryonic neuroblasts (Terriente Félix et al. 2007), but a Nab cofactor function for the putative fly Egr1/2 ortholog encoded by the stripe gene has yet to be identified. Nonetheless, nab is conserved across bilaterian animals and encodes a cofactor of specific C₂H₂ zinc finger transcription factors involved in developmental patterning and differentiation.

The dense site cluster that we identified in the first intron of nab drives a novel expression pattern that is consistent with integration of a D–V Notch signal and an orthogonal A–P BMP signal. The nab locus is already known to be induced by Dpp/BMP signaling (Ziv et al. 2009; Hadar et al. 2012). This is consistent with our finding several canonical Mad:Medea binding sequences in the nab DWME (Figure 3A) and several low-affinity Mad:Medea binding sequences in the nab QE (not shown). Furthermore, both the DWME and QE drive A–P-modulated expression in the wing pouch (Figure 4, B, C, B′′′, and C′′′). We find that the nab DWME has at least two canonical Su(H) binding sites that are critical to its activity at the D–V margin. This activity is also impacted by a mutant Notch background. Similarly, we find that one to three SMAD binding sites are essential for the Notch/BMP signal integration. Thus, this may be the first example of a wing disc enhancer that is downstream of both Notch and BMP signaling. By virtue of its obligate Notch signaling input, the DWME’s transcriptional readout of the BMP morphogen gradient is an idealized two-dimensional graph of pMad levels (y-axis) at different A–P positions (x-axis).

To deconstruct the developmental role of nab expression in different wing compartments, we used the ap-GAL4 enhancer trap to drive nab-RNAi in the dorsal compartment. This RNAi experiment would also knockdown activity of the Ap-dependent dorsal wing margin enhancer in the dorsal wing compartment. With this manipulation, we observed morphogenetic defects of both the thorax and the wing, suggesting an important role for Nab in developmental patterning of the dorsal compartment. More specifically, the patterning defects affect the scutellum and thoracic macrochaetes, consistent with both nab expression and its PWE. This experiment also resulted in an increased rate of wing vein patterning defects. We also observed similar wing vein patterning defects, including loss of the posterior cross-vein in males and females and loss of longitudinal veins IV and V in males, when we drove UAS:nab-RNAi in the posterior compartment using an en-GAL4 enhancer trap line. A milder knockdown using a nab-GAL4 enhancer trap line resulted in an increased loss of the posterior cross-vein in adult females. Thus, of all the knockdown phenotypes, developmental specification of the posterior cross-vein was the most sensitive to diminished Nab function. This is an interesting finding given that crossveinless c was the only other locus in the genome having a similar cluster of binding sites as the nab DWME (Table 1).

Lineage-specific aspects of wing margin and wing quadrant enhancers

Inspection of the site composition of the nab DWME shows it is similar to the WME at cut (Figure 9A). Both wing margin enhancers are induced by Notch signaling and have functional sites for Su(H) (Jack et al. 1991; Neumann and Cohen 1996a). We find both enhancers also have sequences matching the binding preferences for Zelda and Scalloped:Scalloped dimers, although the cut WME has many more of the latter. However, the nab and cut wing margin enhancers each have multiple binding sites either for Mad:Medea or for Tcf, but not for both (compare purple SMAD sites in the nab DWME vs. green Tcf sites in the cut WME in Figure 9A). Unlike the Mad:Medea complex, Drosophila Tcf/Pangolin and its β-catenin coactivator Armadillo, which together mediate Wg/WNT signaling, are not pQ/pN-rich. We thus propose the following model for pQ-mediated Notch/BMP signal integration at the nab DWME (Figure 9B) that distinguishes it from the cut WME (Figure 9C).

We hypothesize that pQ/pN-mediated aggregation of Mad:Medea and Su(H):NICD complexes produced by the nab DWME stabilizes and maintains an active DWME in off-margin clonal descendants of D–V margin cells. Thus, the nab DWME would remain active only in daughter cells that are also simultaneously experiencing high levels of nuclear pMad (see arrows in Figure 9B). In contrast, the cut WME would become inactive in these same off-margin daughter cells because they are no longer experiencing Notch or Wg signaling (see arrows in Figure 9C). In this way, the nab DWME functions as a lineage-specific enhancer in regions with active phosphorylated-Mad (pMad). In contrast, the cut WME is not lineage-specific because its activity is extinguished in cell descendants as they divide away from the margin. Consistent with this model of Notch/BMP integration at the nab DWME, the presumed pMad input is not sufficient for driving DWME expression in cells that although having peak pMad levels are not descended from margin cells.

The cut locus is not expressed throughout the wing pouch like the nab and vg genes, both of which have separate margin/boundary and quadrant enhancers. This also highlights the lineage specificity of some wing pouch enhancers as follows. The vg locus contains a Notch-target boundary enhancer (Williams et al. 1994) as well as a separate BMP-target quadrant enhancer (Kim et al. 1997). The vg boundary enhancer drives expression broadly throughout the presumptive wing border region while the nab DWME drives expression in dorsal margin cells and their off-margin descendants. While the vg locus has separate “single-channel” enhancers responsive to either Notch or BMP signaling pathways, the nab locus has enhancers responsive to combined Notch/BMP signals and (presumably) BMP-only signals. Nonetheless, separate margin/boundary and quadrant enhancers would be required by loci expressed uniformly in proliferating cells of the developing wing pouch. Potentially, this “rule” might generalize across diverse developmental systems in which a gene is expressed broadly in a proliferating tissue patterned by a smaller number of compartment organizer cells. One enhancer would respond to signals experienced specifically by organizer cells, and a second enhancer would respond to signals experienced more broadly across the developmental field.

Our study of the nab intronic enhancer complex also demonstrates a remarkable aspect of transcriptional enhancers. We found that all of the nab intronic enhancers are also silencers. This is reminiscent of studies in Drosophila and mammals identifying arrays of dual enhancer–silencer modules acting in concert to drive complex patterns of gene expression (Sasaki and Hogan 1996; Bessis et al. 1997; Holohan et al. 2006; Perry et al. 2011). Examples of mutual enhancer inhibition also have been found to involve insulators and core promoters (Lin et al. 2007).

Importantly, we discovered the many nab dual enhancer–silencer activities because we were monitoring expression across several tissues of the embryo, larva, and pupa. These results and others suggest that we should be cautious about the effects of enhancer (“over”)-minimalization. The “cryptic” silencer aspect of enhancers may predispose tissue-specific enhancer screens to high false-positive rates of detection if short regulatory DNA fragments are associated with ectopic activities. In any case, these results suggest a new role for Su(H) in mediating interenhancer silencing and attenuation. This functional role is worth testing in greater depth across several additional examples.

Selectors: licensing factors via nucleosomal positioning?

Homeotic selectors were defined initially as TFs that decide and maintain a choice of alternate cell fate potentials for a clonal population of cells and their descendants (García-Bellido 1975; Lawrence et al. 1979). One early clue suggesting that they correspond to a specific regulatory mechanism was that (homeotic) selectors in animals were predominantly found in the form of homeodomain-containing factors (McGinnis et al. 1984). While selectors were initially discussed in terms of cell fate determination and cell fate potential, a consideration of evolutionary processes suggested an emphasis on the role of transcriptional enhancers as important components of selector-based regulation (Akam 1998). Thus, we hypothesize that the phenomenology of selectors acting on cell fate decisions lies in their roles as factors licensing enhancer DNAs to receive graded developmental signals through any number of pathways (e.g., Notch, BMP/Dpp, WNT/Wg, and Hh). For reasons explained below, we further hypothesize that selectors could work simply through nucleosome repositioning, while signal pathway effectors could work through pQ-mediated aggregation and Mediator recruitment.

We have referred to selector activity as licensing and the pathway effector activity as (true) activation for two additional reasons. First, licensing via the fixation of alternate fixed nucleosomal positions could function phenomenologically both in transcriptional activation and in repression, depending on whether effector sites are being revealed or occluded with the positional shifting of a fixed nucleosome. Second, signaling pathway-based effectors tend to be pQ-rich like the Mediator coactivator complex (Tóth-Petróczy et al. 2008) and TBP (Koide et al. 1999), which are recruited by pathway-effector TFs. Thus, signaling pathway integration via pQ-mediated aggregation would be inherent to eukaryotic activation. It would also be functionally rich because it reserves an important role for cis-element organization to strengthen an otherwise nonspecific protein–protein interaction.

The nab DWME, nab QE, cut WME, and vg WME (i.e., the boundary enhancer, BE) have binding sites for homeotic selectors in addition to those for the pathway-mediating effectors. Selectors can act in a negative or positive manner as demonstrated by known wing disc enhancers. The vg WME uniquely has matching sites for the posterior Hox factor Abd-B homeotic selector (5′-YYTTTATGK) that are not found in either the nab or cut WMEs (data not shown). Abd-B plays a negative selector role by prohibiting expression of vg in posterior segments (Carroll et al. 1995). Similarly, the nab QE possesses binding sites for Homothorax (Hth) (5′-ARYDATSRC), which is known to ChIP to this enhancer (Slattery et al. 2013). Hth is expressed throughout the disc except in the wing pouch and is thus likely to prohibit expression outside of the wing pouch. This would be consistent with the clean border of nab QE reporter expression at the wing pouch/hinge border. In contrast, Hth sites are absent in the vg WME, which drives D–V margin expression beyond the pouch.

The wing margin enhancers at nab and cut are likely to use positively acting selectors. Consistent with the site composition of the nab and cut WMEs (Figure 9A), the former would be positively licensed primarily by the selector Ap for the dorsal wing compartment (see cells without Xs in Figure 9B), while the latter would be positively licensed by the selector complex of Sd and Vg (see cells without Xs in Figure 9C). Thus, a mutant Ap background causes DWME-driven expression to collapse to a few cells along the wing margin. Because of the dorsal compartment-specific expression of Ap, and the known localization of activity of Notch signaling to the margin, it is the Notch pathway, secondarily augmented by the BMP input, that drives its expression pattern. Ap only marks a larger domain in which DWME activity can occur. Thus, we interpret this as Ap “licensing” the Notch-dependent DWME for expression exclusively in the dorsal compartment. Additional positively acting selectors are likely to work at the nab DWME based on conservation of known binding sequences, but further work will be required to elucidate these more definitively.

Certain sequence features in the nab DWME hint at the role of selectors in making the enhancer sequences either receptive or refractory to effector DNA binding via fixed or remodeled nucleosomal positions. The core sequence of the DWME overlaps fixed nucleosomal positions in early embryonic stages prior to DWME activity (Figure S4) (Mavrich et al. 2008; Langley et al. 2012). At specific sites flanking these fixed nucleosomes, there are also polyT, homopolymeric runs (Figure S4), which can act as nucleosome repulsing sequences (Anderson and Widom 2001; Segal and Widom 2009). In addition, we also see evidence for competing TT/AA/TA dinucleotide phasing embedded at these fixed nucleosomal windows (Lowary and Widom 1998; Thåström et al. 1999; Segal et al. 2006). Based on the known Su(H) binding structures (Arnett et al. 2010), these fixed positions would make the central S1 Su(H) site unavailable for binding (Figure S4). We thus suggest that some selectors may act by binding and stabilizing alternate nucleosomal positions that reveal effector binding sites. Intriguingly, we also find that Apterous binding sites coincide with the nucleosomal DNA entry regions that are also associated with the repressor linker histone H1 (Syed et al. 2010). Consistent with the inability of Su(H) to bind the DWME sites in the absence of Ap, we find that Su(H) ChIP signals at the nab DWME have not been detected in embryonic stages (Nègre et al. 2011).

We thus speculate that the above homeodomain-containing selectors (Ap, Dll, Hth) and other selectors (e.g., Sd and Vg) work via a mechanism that is different from the graded spatial (Notch/BMP) and temporal (Zld) patterning effectors. The above set of selectors are deficient in pQ (Figure 1A), pN (Figure 1B), and mixed p(Q/N) carboxamide (Figure 1C) tracts unlike the spatiotemporal effectors and the Mediator coactivator complex, which they recruit in their role as activators. We suggest that the phenomenon of graded signal integration by effectors also occurs via generic polycarboxamide aggregation as dictated by the composition of binding sites present in different DNA enhancer scaffolds. In stark contrast, pQ/pN-deficient selectors could bind and fix (i.e., stabilize) default or alternate nucleosomal positions without cross-interacting with pQ/pN-rich effectors. Furthermore, the relationship between selector and pathway effector binding sites would be informed by a helically phased “regulatory reading frame” based on nucleosomal positioning.