Delilah, prospero, and D-Pax2 constitute a gene regulatory network essential for the development of functional proprioceptors.

Coordinated animal locomotion depends on the development of functional proprioceptors. While early cell-fate determination processes are well characterized, little is known about the terminal differentiation of cells within the proprioceptive lineage and the genetic networks that control them. In this work we describe a gene regulatory network consisting of three transcription factors-Prospero (Pros), D-Pax2, and Delilah (Dei)-that dictates two alternative differentiation programs within the proprioceptive lineage in Drosophila. We show that D-Pax2 and Pros control the differentiation of cap versus scolopale cells in the chordotonal organ lineage by, respectively, activating and repressing the transcription of dei. Normally, D-Pax2 activates the expression of dei in the cap cell but is unable to do so in the scolopale cell where Pros is co-expressed. We further show that D-Pax2 and Pros exert their effects on dei transcription via a 262 bp chordotonal-specific enhancer in which two D-Pax2- and three Pros-binding sites were identified experimentally. When this enhancer was removed from the fly genome, the cap- and ligament-specific expression of dei was lost, resulting in loss of chordotonal organ functionality and defective larval locomotion. Thus, coordinated larval locomotion depends on the activity of a dei enhancer that integrates both activating and repressive inputs for the generation of a functional proprioceptive organ.

Introduction

A central question in developmental biology is how different cells that originate in the same lineage and develop within the same organ, acquire unique identities, properties and specialized morphologies. One of the common mechanisms involved in cell fate diversification within a cell lineage is asymmetric cell division in which cytoplasmic determinants of the mother cell differentially segregate into one of the two daughter cells. This asymmetry is then translated into differential gene expression and the activation of cell-type-specific gene regulatory networks (GRN) that dictate the differentiation programs of cells with unique properties. The transition from a primary cell fate to the characteristic phenotype of a fully differentiated cell involves complex GRNs in which numerous genes regulate each other’s expression. Despite this complexity, genetic analyses in well-characterized developmental systems can often reveal elementary interactions in small GRNs which dictate a specific cell fate, or a specific feature of the differentiating cell.

Many of the core components and the central processes underlying asymmetric cell divisions and primary cell fate decisions have been uncovered in studies performed on the central and peripheral nervous system (PNS) of Drosophila (e.g. Knoblich, 2008; Schweisguth, 2015). The PNS of Drosophila contains two classes of multicellular sensory organs, external sensory organs and chordotonal organs (ChOs), whose lineages share a similar pattern of asymmetric cell divisions (Lai and Orgogozo, 2004). In both types of organs, the neuron and support cells, which collectively comprise the sensory organ, arise from a single sensory organ precursor cell (SOP) through a sequence of precisely choreographed asymmetric cell divisions. Antagonistic interactions involving Notch and Numb are key regulators of the asymmetry generated between each two sibling cells within these lineages (Gönczy, 2008; Rebeiz et al., 2011; Reeves and Posakony, 2005). Unlike the primary cell-fate specification, which has been extensively investigated, the process of terminal differentiation of the post-mitotic progeny is poorly understood. We are using the larval lateral pentascolopidial ChO (LCh5) as a model system to study cell fate diversification within a sensory lineage.

The LCh5 organ is composed of five mechano-sensory units (scolopidia) that are attached to the cuticle via specialized epidermal attachment cells. Each of the five scolopidia originates in a single precursor cell that divides asymmetrically to generate five of the six cell types that construct the mature organ: the neuron, scolopale, ligament, cap, and cap-attachment cell (Brewster and Bodmer, 1995; Figure 1A–C). Three of the five cap-attachment cells are rapidly removed by apoptosis, leaving two cap-attachment cells that anchor the five cap cells to the epidermis (Avetisyan and Salzberg, 2019). Later in development, following the migration of the LCh5 organ from the dorsal to the lateral region of the segment, a single ligament-attachment cell is recruited from the epidermis to anchor the five ligament cells to the cuticle (Inbal et al., 2004). The mature LCh5 organ responds to mechanical stimuli generated by muscle contractions that lead to relative displacement of the attachment cells and the consequent shortening of the organ (Hassan et al., 2019).

Figure 1.

The LCh5 organ and the dei gene.

(A) Four abdominal segments of a stage 16 embryo carrying the sv/Pax2 reporter (green) which labels the cap and scolopale cells, stained with anti-Pros (red) which labels the nuclei of scolopale cells, and anti-Futsch (blue), which labels the neurons. One LCh5 organ is circled and the cap cells (C), scolopale cells (S) and neurons (N) are indicated. The cap-attachment, ligament-attachment and ligament cells are not stained. Scale bar = 20 μm. (B–C) the ChO lineage (B) and schematic illustration of an LCh5 organ (C). The neurons are depicted in blue, the scolopale cells in orange, the ligament cells in cerulean, the cap cells in green, the cap-attachment cells in red and the ligament-attachment cell in purple. (D) Schematic representation of the dei locus showing the two exons (black boxes) and the cis regulatory modules (CRM) that drive gene expression within the ChO lineage: the dei CRM which drives expression in the attachment cells (red box) and the dei CRM which drives expression in the cap and ligament cells (green box). The dei CRM was originally mapped to a 1353 bp fragment (dei) and was then delimited to a smaller 389 bp fragment located immediately upstream to the second exon (dei). As part of this work, the ChO-specific CRM was further delimited to a 262 bp fragment (dei). (E–H) The embryonic expression patterns driven by the dei (E), dei (F), dei (G), and dei (H) enhancers are shown. C, cap cell; L, ligament cell; CA, cap-attachment cell, LA, ligament-attachment cell.

(A) Schematic representation of the screening system that used the dei enhancer as a bait and the Aurobasidin A selection system to screen an embryonic cDNA library for preys that bind to the bait (Hybergenics Services). (B) A DomSight graph showing the top candidates identified in the screen in very high confidence in the interaction (red) or moderate confidence (green). This score represents the probability of an interaction to be non-specific based on the comparison between the number of independent prey fragments found for an interaction and the chance of finding them at random (background noise). For each interaction, a Predicted Biological Score (Global PBS) is computed to assess the interaction reliability and the results are ranked in four categories from A (the highest confidence rank) to D (Formstecher et al., 2005). The orange bars mark the SID fragment (selected interaction domain), which is the amino acid sequences shared by all prey.

Very little is known about the unique cell-type-specific differentiation programs that characterize each of the ChO cells, whose morphologies and mechanical properties differ dramatically from each other. To address this issue, we focus on the transcription factor Taxi wings/Delilah (Dei), an important regulator of cell adhesion (Egoz-Matia et al., 2011), which is expressed in the four accessory cell types (cap, ligament, cap-attachment and ligament-attachment) but is excluded from the neuron and the scolopale cell. Even though dei is expressed in all four accessory cells, its expression in these cells is differentially regulated. The transcription of dei in the ChO is controlled by two cis-regulatory modules (CRMs): The dei enhancer, located ~2.5 Kb upstream of the dei transcription start site, drives expression in the cap-attachment and ligament-attachment cells (as well as tendon cells), whereas the dei enhancer, an intronic 1353 bp DNA fragment, drives dei expression specifically in the cap and ligament cells (Nachman et al., 2015; Figure 1D–F). The dei enhancer was shown to be activated by the transcription factor Stripe, which is considered a key regulator of tendon cell development (Becker et al., 1997) and a known determinant of attachment cell identity (Klein et al., 2010). Here, we provide a high-resolution dissection of the dei enhancer and show that it integrates both activating and repressive cues to drive dei expression in cap and ligament cells while suppressing it in scolopale cells. We find that D-Pax2/Shaven (Sv), which is expressed in both branches of the cell lineage, is a positive regulator of dei and that Prospero (Pros) inhibits dei expression specifically in the scolopale cell. This small GRN is required for the realization of differentiation programs characterizing cap versus scolopale cell fates and is therefore essential for ChO functionality and coordinated larval locomotion.

Results

Identifying D-Pax2/Sv as a putative direct activator of dei expression

To reveal the gene network that regulates dei expression in the ChO cells, we used the previously identified ChO-specific dei enhancer or the shorter version of it - dei both described in Nachman et al., 2015; Figure 1D and F–G, as an entry point. As part of this work, we have further narrowed down the critical regulatory region to an evolutionarily conserved 262 bp fragment, dei, which drives ChO-specific expression in a pattern indistinguishable from the expression pattern driven by the larger fragments dei and dei (Figure 1D, F and H). These regulatory fragments were used in two types of screens aimed at identifying genes that regulate the expression of dei in the ChO lineage. The first screen was an RNAi-based phenotypic screen conducted in larvae (described in Hassan et al., 2018). It capitalized on a reporter fly strain in which the cap and ligament cells expressed green fluorescent protein (GFP) under the regulation of the dei enhancer, while the attachment cells expressed red fluorescent protein (RFP) under the regulation of the dei enhancer (Halachmi et al., 2016). One of the 31 genes identified in that screen as being required for normal morphogenesis of the larval LCh5 organ was shaven (sv), which encodes for the Drosophila Pax2 homologue D-Pax2. The knockdown of sv within the ChO lineage led to loss of expression of the dei reporter from the cap cells, suggesting that D-Pax2 is a positive regulator of dei transcription in this cell type (Hassan et al., 2018).

Here, we describe the second screen, a yeast one hybrid (Y1H) screen, aimed at identifying proteins that bind directly to the dei enhancer. The screen was performed by Hybergenics Services on a Drosophila whole embryo library (ULTImate Y1H screen vs Drosophila Whole Embryo RP2 0–12 + 12-24 hr), using the dei sequence as a bait. Screening was performed using the Aureobasidin A selection system (described in detail in the Materials and methods section). Out of 124 million clones screened, 146 clones that were found to grow on the selective medium containing 400 ng/ml of the yeast antibiotic agent Aureobasidin-A were sequenced (Supplementary file 1). Two proteins were identified to bind the bait with very high confidence in the interaction: D-Pax2/Sv (eight independent clones) and LamC (21 independent clones). Three additional candidates were identified as binders with moderate confidence in the interaction: Fax (one clone), Lola (two independent clones), and Toy (three independent clones) (Figure 1—figure supplement 1 and Supplementary file 1). Together, the results of the two independent screens identified D-Pax2/Sv as a putative direct transcriptional activator of dei and an important player in ChO development.

Figure 1—figure supplement 1.

Identifying potential direct regulators of dei in the ChO using a Y1H screen.

(A) Schematic representation of the screening system that used the dei enhancer as a bait and the Aurobasidin A selection system to screen an embryonic cDNA library for preys that bind to the bait (Hybergenics Services). (B) A DomSight graph showing the top candidates identified in the screen in very high confidence in the interaction (red) or moderate confidence (green). This score represents the probability of an interaction to be non-specific based on the comparison between the number of independent prey fragments found for an interaction and the chance of finding them at random (background noise). For each interaction, a Predicted Biological Score (Global PBS) is computed to assess the interaction reliability and the results are ranked in four categories from A (the highest confidence rank) to D (Formstecher et al., 2005). The orange bars mark the SID fragment (selected interaction domain), which is the amino acid sequences shared by all prey.

D-Pax2/Sv activates dei expression in the cap cell

To validate the sv RNAi-induced phenotype, we characterized the LCh5 organs of sv mutant embryos (a null allele - sv). In accordance with the knockdown phenotypes, the expression of dei in the cap cell was abolished in sv-deficient embryos (Figure 2A–B). The expression of additional genes associated with proper differentiation of the cap cell, such as αTub85E, was reduced as well (Figure 2C–D). The loss of sv did not eliminate the expression of either scolopale-specific proteins (Crumbs, Pros) or neuronal markers (Futsch, Elav, Nrg), suggesting that it did not affect primary cell fate decisions, however, its loss prevented normal morphogenesis of the sensory unit (Figure 2E–H).

Figure 2.

Sv/D-Pax2 activates dei expression in the cap cell.

(A–B) Representative abdominal segments of stage 16 embryos stained for Dei (cyan) and the neuronal marker Futcsh (red). The anti-Dei staining is shown separately in the lower panel. (A) A sv heterozygous embryo demonstrating the expression of Dei in the cap cell nuclei (circled). (B) The expression of Dei is lost in homozygous sv embryos. (C–D) Heterozygous (C) and homozygous (D) sv embryos stained for αTub85E (red) and Futsch (cyan). The arrows mark the cap cells. The anti-αTub85E staining is shown separately in the lower panel. (E–F) Heterozygous (E) and homozygous (F) sv embryos stained for Elav (cyan) and Pros (red). The arrows mark the nuclei of scolopale cells. The anti-Pros staining is shown separately in the lower panel. (G–H) Heterozygous (G) and homozygous (G) sv embryos stained for Nrg (cyan) and Crumbs (red). The arrows mark the nuclei of scolopale (Sco) cells. The anti-Crumbs staining is shown separately in the lower panel. (I–L) Representative abdominal segments of stage 16 embryos that express GFP (I, K) or GFP and sv (J, L) under the regulation of en-Gal4. The embryos carry the dei marker (anti bGal staining is shown in red) and are stained with anti-Dei (blue in I-J) or anti-αTub85E (blue in K-L). The staining patterns of dei-lacZ, Dei and αTub85E are shown separately on the right. Note the ectopic expression of Dei, dei-lacZ and αTub85E in epidermal cells within the en domain (arrows in J, L).

To further test the ability of D-Pax2/Sv to activate dei expression in vivo, we ectopically expressed sv under the regulation of en-Gal4 and examined the resulting changes in gene expression patterns. The ectopic expression of sv led to ectopic expression of both the endogenous dei gene and the dei transcriptional reporter, as well as the cap cell marker αTub85E (Figure 2I–L). The ectopic expression of D-Pax2 in the en domain had detrimental effects on the pattern of ChO migration and possibly on other morphogenetic processes in the embryo. Thus, the identification of cells based on their position was not feasible in these embryos. However, this experiment clearly showed that ectopic expression of D-Pax2 could activate expression of the dei gene and the dei reporter in epidermal cells within the en domain. Thus, D-Pax2/Sv ability to activate dei is not restricted to the ChO lineage, or the peripheral nervous system in general. Altogether, these observations corroborate the notion that D-Pax2/Sv is an activator of dei that plays a critical role in ChO morphogenesis.

Pros represses dei in the scolopale cell

The sv gene is expressed broadly within the ChO lineage during early stages of organ development and is then gradually restricted to the cap and scolopale cells (Czerny et al., 1997; Fu et al., 1998; Fu and Noll, 1997), where its expression level remains high during late embryogenesis and larval stages. Yet, the expression of its downstream target gene dei is normally activated in the cap cell but is excluded from the scolopale cell. This discrepancy in the expression pattern of D-Pax2/Sv and dei may point to the presence of a scolopale-specific repressor that prevents the transcription of dei in this D-Pax2/Sv-expressing cell. The results of the abovementioned RNAi screen identified the transcription factor Pros as a good candidate for being that repressor. pros and sv RNAi had opposite effects on the expression of the dei reporter. Although the knockdown of sv led to a loss of the reporter expression from the cap cell, the knockdown of pros led to expansion of its expression into the scolopale cell suggesting that, normally, Pros represses dei in this cell (Hassan et al., 2018).

To validate the pros knockdown phenotypes and to further test the idea that Pros acts as a repressor of dei in the scolopale cell, we characterized the ChOs of pros embryos. The loss of pros led to ectopic expression of both the endogenous dei gene and the dei reporter, as well as the cap cell marker αTub85E in the scolopale cells of the LCh5 organs (Figure 3A–B). Similar effect was evident in the scolopale cells of the LCh1 and VChA/B organs (Figure 3—figure supplement 1). The pros-deficient scolopale cells developed into cap-like cells rather than ligament-like or attachment cell-like cells. This was indicated by the upregulation of dei which was not accompanied by upregulation of the transcription factor Sr that is normally co-expressed with Dei in the ligament and attachment cells but is excluded from the cap cells (Figure 3C–D). The observed alterations in gene expression pattern do not reflect a full scolopale-to-cap cell fate transformation, as the Pros-deficient scolopale cells still maintain some of their scolopale-specific characteristics, such as Eyes Shut and α—Catenin expression (Figure 3E–H). As the number of neurons, ligament, cap, and cap-attachment cells remained normal in pros mutant embryos (Figure 3—figure supplement 2), we conclude that Pros does not influence primary cell-fate decisions within the LCh5 lineage.

Figure 3.

Pros represses dei in the scolopale cell.

(A–B) Representative abdominal segments of stage 16 pros heterozygous (A) and homozygous (B) embryos carrying the dei marker (shown in red) and stained for Dei (blue) and αTub85E (green). Each of the three channels is shown separately below. Note the expansion of Dei and the dei marker into the scolopale cells in pros mutant embryos. The arrows point to the scolopale cells. (C–D) pros heterozygous (C) and homozygous (D) embryos stained for Dei (blue) and Sr (red). Note that the ectopic expression of Dei in the scolopale cells of pros mutant embryo is not accompanied by ectopic expression of Sr (the arrow labeled with asterisks in D). (E–F) pros heterozygous (E) and homozygous (F) embryos carrying an α-Catenin-GFP reporter and stained for Dei (blue, shown separately in the inset) and αTub85E (red). (G–H) pros heterozygous (G) and homozygous (H) embryos carrying the dei reporter (shown in green) and stained with the scolopale marker anti-Eyes Shut (MAb21A6, red, shown separately in the inset). Note that the expression of both Eyes Shut/21A6 and α—Catenin is maintained in the pros-deficient scolopale cells. (I–J) dei (I) and dei pros (J) embryos stained for Dei (cyan, shown separately in I’-J’) and αTub85E (red, shown separately in I’’-J’’). The cap and ligament cells are circled. Note the loss of Dei and αTub85E expression upon Pros expression. (K–L) dei (K) and dei pros (L) embryos carrying the dei reporter (shown in green) stained for Sal (blue) and Repo (red). The cap, scolopale (Sco) and ligament (Lig) cells are indicated; the asterisks mark oenocyte cell nuclei. (M–N) dei (M) and dei pros (N) embryos carrying an α—Catenin-GFP reporter (green, shown separately in M’-N’) and stained for Crumbs (red, shown separately in M’’-N’’) and αTub85E (blue). Note the duplication of scolopale-specific structures in the cap cell expressing Pros ectopically (circled).

(A–B) Representative segments of stage 16 heterozygous (A) and homozygous (B) pros mutant embryos that carry the dei reporter (red) and stained for Dei (blue) and αTub85E (green). The staining of Dei is shown separately in A’ and B’. Note the ectopic expression of Dei and the dei reporter in the scolopale cells of pros-deficient LCh5, LCh1, and VChB organs (arrows in B and B’).

(A–B) Representative segments of stage 16 wt (A) and pros mutant (B) embryos stained for Dei (blue), Repo (red), and Sr (green). Repo staining shows the normal number of ligament cells in the mutant. (C–D) Representative LCh5 organs of wt (C) and pros mutant (D) embryos stained for CPO (cyan) and 22C10 (anti Futsch, red). CPO is expressed in all of the LCh5 cells. (E–F) Representative LCh5 organs of wt (E) and pros mutant (F) embryos stained for the neuronal markers Elav (cyan) and 22C10 (red). The data shown in C-F demonstrate that the number of neurons remains unaltered in pros mutant embryos. (G) An LCh5 of pros mutant embryo stained for the neuronal marker 22C10 (anti-Futsch), Dei and Sr. The arrowhead points to the typical axonal pathfinding defect of pros mutant embryos.

Figure 3—figure supplement 1.

The loss of pros affects similarly the various types of larval ChOs.

(A–B) Representative segments of stage 16 heterozygous (A) and homozygous (B) pros mutant embryos that carry the dei reporter (red) and stained for Dei (blue) and αTub85E (green). The staining of Dei is shown separately in A’ and B’. Note the ectopic expression of Dei and the dei reporter in the scolopale cells of pros-deficient LCh5, LCh1, and VChB organs (arrows in B and B’).

Figure 3—figure supplement 2.

The number of neurons and ligament cells remain normal in pros mutant embryos.

(A–B) Representative segments of stage 16 wt (A) and pros mutant (B) embryos stained for Dei (blue), Repo (red), and Sr (green). Repo staining shows the normal number of ligament cells in the mutant. (C–D) Representative LCh5 organs of wt (C) and pros mutant (D) embryos stained for CPO (cyan) and 22C10 (anti Futsch, red). CPO is expressed in all of the LCh5 cells. (E–F) Representative LCh5 organs of wt (E) and pros mutant (F) embryos stained for the neuronal markers Elav (cyan) and 22C10 (red). The data shown in C-F demonstrate that the number of neurons remains unaltered in pros mutant embryos. (G) An LCh5 of pros mutant embryo stained for the neuronal marker 22C10 (anti-Futsch), Dei and Sr. The arrowhead points to the typical axonal pathfinding defect of pros mutant embryos.

The ability of Pros to repress dei was not restricted to the scolopale cell. Ectopic expression of pros in the LCh5 lineage, using a dei driver, abolished the expression of both the endogenous dei gene and a dei reporter in the cap and ligament cells as well (Figure 3I–L). In parallel to the repression of dei, ectopic expression of pros was sufficient for upregulating scolopale-specific genes and, moreover, for driving the formation of ectopic scolopale-specific structures within the affected cap cells. Most notably, the Pros-expressing cap cells manifested scolopale rods and ectopically expressed Crumbs and α-Catenin in a scolopale-characteristic pattern (Figure 3M–N).

D-Pax2/Sv and Pros regulate the transcription of dei via dei

The fact that the expression of the dei transcriptional reporters was affected similarly to the endogenous dei gene by both sv and pros loss- and gain-of-function, suggested that both D-Pax2/Sv and Pros regulate dei’s transcription in the ChOs through this regulatory module. To test this hypothesis, we deleted the dei region from the fly genome by CRISPR/Cas9-mediated genome editing, resulting in a new regulatory allele of dei (dei) (Figure 4—figure supplement 1). To verify that the deletion of this intronic enhancer does not affect splicing of the transcript, cDNA was synthesized from homozygous dei and control flies. A 416 bp fragment was amplified by PCR from the cDNA samples using primers located on both sides of the intron (in the 1st and 2nd exons) (Figure 4—figure supplement 1B and Materials and methods). Sequencing of the PCR products verified the presence of normally structured dei transcript in dei mutants (Figure 4—figure supplement 1C).

Figure 4—figure supplement 1.

Deletion of the dei enhancer does not affect splicing.

(A) Structure of the dei (tx) locus and the two expected transcripts. The gene contains one non-coding and one coding exon that are separated by a ~ 5.3 kb intron. The parentheses mark the sequence deleted in the dei allele. (B) A partial sequence of the dei locus showing the first exon (blue), the intron (black) and part of the coding exon (red). The sequence deleted in the dei allele is highlighted in green; the initiation codon is highlighted in orange, and the primers used for the amplification of a 460 bp cDNA fragment for sequencing are highlighted in yellow. (C) Electropherograms of sequences read in the reverse orientation on cDNA prepared from flies homozygous for the dei allele and control flies from the Cas9 strain to which injection was done. The splicing site is marked with arrows and the initiation codon is highlighted in orange in both sequences. The arrow at the top marks the direction of transcription.

As expected, in homozygous dei embryos the expression of dei was lost from the cap and ligament cells but remained intact in the rest of the dei-expressing cells: CA, LA and tendon cells (Figure 4A–B). This observation strongly suggests that the dei enhancer constitutes the sole regulatory module driving dei expression in the cap and ligament cells.

Figure 4.

D-Pax2/Sv and Pros regulate the transcription of dei via the dei regulatory module.

(A–B) Stage 16 wt (A) and dei (B) embryos stained for Dei (cyan) and αTub85E (red). Note that in (B) Dei is still evident in the cap-attachment and ligament-attachment cells (CA and LA, arrows in A, circled in A’’), as well as in tendon cells (T; arrowheads), but is lost from the cap and ligament cells. The double edge arrows in A and B demarcate the length of the cap cells. (C) An embryo in which sv was expressed under the regulation of en-Gal4 in a dei background. Sv was unable to induce Dei expression in the cap and ligament cells in the absence of the dei enhancer. Ectopic expression of αTub85E is evident within the en domain (arrowheads in C’). (D–E) wt (D) and A pros homozygous embryo (E) stained for Dei (cyan) and anti-Futsch (red). The pros embryo (E) presents a dei-like and not pros-like expression pattern of the Dei protein, indicating that the dei deletion is epistatic to pros loss-of-function. The neurons present the typical pros axonal pathfinding defects.

(A) Structure of the dei (tx) locus and the two expected transcripts. The gene contains one non-coding and one coding exon that are separated by a ~ 5.3 kb intron. The parentheses mark the sequence deleted in the dei allele. (B) A partial sequence of the dei locus showing the first exon (blue), the intron (black) and part of the coding exon (red). The sequence deleted in the dei allele is highlighted in green; the initiation codon is highlighted in orange, and the primers used for the amplification of a 460 bp cDNA fragment for sequencing are highlighted in yellow. (C) Electropherograms of sequences read in the reverse orientation on cDNA prepared from flies homozygous for the dei allele and control flies from the Cas9 strain to which injection was done. The splicing site is marked with arrows and the initiation codon is highlighted in orange in both sequences. The arrow at the top marks the direction of transcription.

To further test the notion that D-Pax2/Sv regulates dei expression via the dei enhancer we examined the ability of D-Pax2/Sv to activate dei expression in the dei background. We found that ectopic expression of D-Pax2/Sv failed to induce ectopic dei expression in the dei background (Compare Figures 4C and 2L), indicating that the dei enhancer is indispensable for the ability of D-Pax2/Sv to activate dei transcription. In a complementary experiment, we tested the effect of pros loss-of-function on the expression of dei in the dei background. We found that the dei regulatory mutation was epistatic to pros loss-of-function, so that no ectopic expression of dei was observed in the pros-deficient scolopale cells in embryos homozygous for the dei mutation (Compare Figure 4D–E and Figure 3B and D). Based on these results, we concluded that the expression of dei in ChOs depends on the presence of the dei regulatory region and that this enhancer integrates the positive and negative inputs of D-Pax2/Sv and Pros, respectively, to drive a lineage-specific dei expression.

D-Pax2/Sv and Pros are direct transcriptional regulators of dei

Our genetic analyses revealed that D-Pax2/Sv and Pros regulate the expression of dei in opposing manners through the function of the dei enhancer. In addition, the Y1H screen identified a direct interaction between D-Pax2/Sv and dei (which includes the dei module). We therefore hypothesized that the dei enhancer contains binding sites for D-Pax2/Sv and Pros. Motif search analysis predicted that the dei sequence encodes one canonical binding site for D-Pax2/Sv (Figure 5B, D-Pax2/Sv Site 2). On the other hand, we were unable to predict binding sites for Pros in the dei sequence.

Figure 5.

dei contains two binding sites for the activator D-Pax2/Sv and three binding sites for the repressor Pros.

(A) Schematic representation of the dei enhancer. Green and orange boxes represent the location of the D-Pax2/Sv and Pros binding sites identified by systematic EMSAs using oligos corresponding to the regions represented by Boxes 1–7. (B) Sequence alignment between six Drosophila species for the region of the dei enhancer containing the three D-Pax2/Sv and Pros sites (labeled and highlighted in green and orange, respectively). Dashes indicate gaps in the aligned sequence. Mutations of the D-Pax2/Sv and Pros sites used for the in vivo assays are shown at the bottom. (C) D-Pax2/Sv binds to two sites in dei, one low affinity site in fragment 3 (D-Pax2 site 1) and one high affinity site in fragment 7 (D-Pax2 site 2), as demonstrated with EMSA. The full screen is shown in Figure 5—figure supplement 1. (D) Pros binds to three sites in dei in fragment 3, Pros site 2 in an overlapping sequence in fragments 4 and 5, Pros site 3 in fragment 6, as demonstrated with EMSA. The full screen is shown in Figure 5—figure supplement 2. (E–L) Expression of wild-type (E) and mutated (F–L) dei reporter constructs in abdominal segments A2-A6 of representative stage 16 embryos. The name of the construct is indicated in the bottom of each panel. The green arrows in E point to the cap cells of the various ChO of one abdominal segment: LCh1-lateral ChO1, LCh5-pentascolopidial organ, VChB and VChA are two ventral ChOs. The green arrows in J and L point to cap, scolopale (Sco) and ligament (Lig) cells of LCh5 organs where elevated level of reporter expression is evident. The arrowheads in I and K point to ligament cells of the ventral ChOs. in (M–N) Quantification of reporter activity in nuclei of cap (M) and scolopale cells (N) from LCh5 of segment A2 in embryos carrying the indicated constructs (n = 10 embryos for each genotype). In violin plots, each point represents an individual nucleus, median is shown as dark gray dashed line. Asterisks denote significant difference from wild-type activities, (*) - p < 0.05, (****) – p < 0.0001, n.s. – not significant (Dunnett’s multiple comparison test).

(A) A schematic representation of the regions tested for their ability to bind D-Pax2/Sv in EMSAs. In gray are fragments that show specific binding. (B) EMSA scan for D-Pax2/Sv binding sites in dei using the oligos marked in (A). Two sub-regions of dei (fragments 3 and 7) bind to D-Pax2/Sv. (C–E) Dissection of the D-Pax2/Sv binding sites through competition assays. EMSAs were performed in the presence of increasing concentrations of unlabeled oligos (competitors). (C) Sub-fragments 3b and 7b unlabeled probes completely compete the binding of fragments 3 and 7, respectively, to D-Pax2/Sv. (D–E) On the left, alignment of the mutated competitors used in the competition EMSAs shown on the right. The nucleotides that comprise the D-Pax2/Sv binding sites are marked in gray.

(A) A schematic representation of the regions tested for their ability to bind Pros in EMSAs. In gray are fragments that show specific binding (stable upon increase of non-specific and specific competitors). (B) EMSA scan for Pros binding sites in dei using the oligos marked in (A). Four sub-regions of dei (fragments 3, 4, 5, and 6) bind specifically to Pros. (C–F) Dissection of the Pros binding sites through competition assays. EMSAs were performed in the presence of increasing concentrations of unlabeled oligos (competitors). (C) Sub-fragments 3b, 4b, 5a, and 6a unlabeled probes completely compete the binding of fragments 3, 4, 5, and 6, respectively, to Pros. (D–F) On the left, alignment of the mutated competitors used in the competition EMSAs shown on the right. The nucleotides that comprise the Pros binding sites are marked in gray.

Sequence alignment of the wild-type dei enhancer and the mutated constructs used in Figure 5. Verified D-Pax2/Sv and Pros binding sites are marked with green and orange boxes, respectively. Red letters denote mutated sites.

Consensus positional weight metrics (PWM) motifs for D-Pax2/Sv (A) and Pros (B) aligned to binding sites discovered in this study.

b-galactosidase expression pattern driven by the wild-type and mutated dei enhancers in pupal femoral ChOs. Pupae were stained with X-gal 40 hr after pupa formation. (A–C) An overview of pupae carrying the dei (A), the dei (B) and the dei (C) enhancers. A representative pupa of each genotype is circled and shown in higher magnification in (D–F). (G–I) A closeup view of a femur (Fe) and tibia (Ti) of legs dissected from the pupae shown in (D–F). The arrows point to femoral ChOs and arrowhead pont to the tibial ChOs (in genotypes where staining is evident).

Figure 5—figure supplement 1.

Identification of the regions in the dei enhancer that bind D-Pax2/Sv in vitro.

(A) A schematic representation of the regions tested for their ability to bind D-Pax2/Sv in EMSAs. In gray are fragments that show specific binding. (B) EMSA scan for D-Pax2/Sv binding sites in dei using the oligos marked in (A). Two sub-regions of dei (fragments 3 and 7) bind to D-Pax2/Sv. (C–E) Dissection of the D-Pax2/Sv binding sites through competition assays. EMSAs were performed in the presence of increasing concentrations of unlabeled oligos (competitors). (C) Sub-fragments 3b and 7b unlabeled probes completely compete the binding of fragments 3 and 7, respectively, to D-Pax2/Sv. (D–E) On the left, alignment of the mutated competitors used in the competition EMSAs shown on the right. The nucleotides that comprise the D-Pax2/Sv binding sites are marked in gray.

Figure 5—figure supplement 2.

Identification of the regions in the dei enhancer that bind Pros in vitro.

(A) A schematic representation of the regions tested for their ability to bind Pros in EMSAs. In gray are fragments that show specific binding (stable upon increase of non-specific and specific competitors). (B) EMSA scan for Pros binding sites in dei using the oligos marked in (A). Four sub-regions of dei (fragments 3, 4, 5, and 6) bind specifically to Pros. (C–F) Dissection of the Pros binding sites through competition assays. EMSAs were performed in the presence of increasing concentrations of unlabeled oligos (competitors). (C) Sub-fragments 3b, 4b, 5a, and 6a unlabeled probes completely compete the binding of fragments 3, 4, 5, and 6, respectively, to Pros. (D–F) On the left, alignment of the mutated competitors used in the competition EMSAs shown on the right. The nucleotides that comprise the Pros binding sites are marked in gray.

To test whether D-Pax2/Sv actually binds to the predicted site, and to search for additional D-Pax2/Sv binding sites and for Pros binding sites, we systematically screened the dei sequence with electrophoretic mobility shift assays (EMSAs, Figure 5 and Figure 5—figure supplements 1 and 2). As shown in Figure 5C, purified D-Pax2/Sv DNA-binding domain bound strongly to the region containing the predicted binding site (probe 7). In addition, we identified another region (in probe 3), that bound D-Pax-2/Sv at a lower affinity (Figure 5C and Figure 5—figure supplement 1). Using purified Pros-S DNA-binding domain we found that four fragments–probes 3, 4, 5, and 6–bound Pros in vitro (Figure 5D). Comprehensive mutagenesis and competition assays revealed three Pros conserved binding sites within these fragments (Figure 5—figure supplement 2), one of them partially overlapping the low-affinity binding site of D-Pax2/Sv (Figure 5B and Figure 5—figure supplement 3). Interestingly, none of these identified binding sites resemble the known binding sites for Pros identified by target detection assay (TDA) (Hassan et al., 1997), by functional studies (Cook et al., 2003), or by single-cell omics analyses (Bravo González-Blas et al., 2020; Figure 5—figure supplement 4).

Figure 5—figure supplement 3.

Sequence of the dei enhancer and tested mutations.

Sequence alignment of the wild-type dei enhancer and the mutated constructs used in Figure 5. Verified D-Pax2/Sv and Pros binding sites are marked with green and orange boxes, respectively. Red letters denote mutated sites.

Figure 5—figure supplement 4.

The dei enhancer contains non-canonical binding sites for Pros and D-Pax2.

Consensus positional weight metrics (PWM) motifs for D-Pax2/Sv (A) and Pros (B) aligned to binding sites discovered in this study.

To test the in vivo role of the D-Pax2/Sv sites identified in vitro, we mutated them either individually, or in combinations, in the context of a dei reporter transgene encoding for nuclear β-galactosidase (the mutations are shown in Figure 5—figure supplement 3). Mutation of the canonical D-Pax2/Sv site 2 reduced the expression level driven by dei in cap cells (Figure 5G and M). Mutation of D-Pax2/Sv site 1 had no significant effect on its own (Figure 5F and M) but led to a complete suppression of dei function when combined with a mutation in D-Pax2/Sv site 2 (Figure 5H and M). This effect was evident in all types of larval ChOs (Figure 5H). These results suggest that D-Pax2/Sv regulates the expression of dei in cap cells by binding to two D-Pax2/Sv binding sites within the dei enhancer.

We next tested whether the Pros sites identified by EMSA function in vivo to suppress dei activity in scolopale cells. Mutating the Pros site 2 resulted in ectopic expression of the reporter gene in the scolopale cells of LCh5 (Figure 5J and N). In contrast, disruption of Pros site 1 or site 3 had very small or no detectable effects, respectively, on the expression of dei in the scolopale cells of LCh5 (Figure 5I, K and N) but induced ectopic expression in scolopale cells of the LCh1 and VChA/B organs (Figure 5I and K, arrowheads). In addition, the disruption of Pros site 3 led to elevation in the reporter’s level in the LCh5 ligament cells (Figure 5K). Simultaneous mutation of all three Pros binding sites in dei resulted in intensified ectopic expression in the scolopale cells of all types of larval ChOs (LCh5, LCh1, VChA/B; Figure 5L and N). Based on these results we conclude that Pros represses dei expression in the scolopale cell by binding to three low-affinity binding sites in dei that function additively. It is important to note that the mutations introduced in Pros site 1 also disrupted the low-affinity site for D-Pax2/Sv (site 1, Figure 5B) and affected D-Pax2/Sv binding in vitro (Figure 5—figure supplement 1D). While the D-Pax2/Sv site 1 is dispensable for dei expression in cap cells at the presence of the high-affinity site, it is possible that mutations in Pros site 1 would cause a greater effect on the reporter expression in scolopale cells if the D-Pax2/Sv site would not be disrupted simultaneously.

To test whether the identified regulatory interactions apply to additional types of ChOs, we examined the expression pattern driven by the wild-type and mutated dei enhancers in developing femoral ChOs (Figure 5—figure supplement 5). This analysis suggests that the Dei/Pros/D-Pax2 GRN plays a similar role in larval and adult ChOs and that the dei enhancer encodes pleiotropic transcription factor binding sites (Preger-Ben Noon et al., 2018) that integrate the activating and repressing inputs of D-Pax2/Sv and Pros, respectively.

Figure 5—figure supplement 5.

The D-Pax2 and Pros binding sites regulate the activity of the dei enhancer in femoral ChOs.

b-galactosidase expression pattern driven by the wild-type and mutated dei enhancers in pupal femoral ChOs. Pupae were stained with X-gal 40 hr after pupa formation. (A–C) An overview of pupae carrying the dei (A), the dei (B) and the dei (C) enhancers. A representative pupa of each genotype is circled and shown in higher magnification in (D–F). (G–I) A closeup view of a femur (Fe) and tibia (Ti) of legs dissected from the pupae shown in (D–F). The arrows point to femoral ChOs and arrowhead pont to the tibial ChOs (in genotypes where staining is evident).

The dei enhancer is essential for normal ChO function and larval locomotion

We have previously shown that the cap cell plays a crucial role in propagating muscle-generated mechanicals signals to the sensory neuron (Hassan et al., 2019). To test whether dei expression in the cap and ligament cells, mediated solely by the dei enhancer, is essential for the proprioceptive function of the ChO, we analyzed the pattern of locomotion of freely moving dei larvae and compared it to wild-type larvae and larvae homozygous for the dei null allele (Hassan et al., 2018). As shown in Figure 6, wild-type larvae crawled persistently 94.5% ± 12.4% (n = 27) of the time with a very few changes of direction or head swipes (Figure 6A, E and H, and Videos 1–2). In contrast, the dei larvae exhibited frequent changes of moving direction and longer pauses, walking on average only 38.9% ± 19.4% (n = 25) of the time. While pausing, the dei larvae swiped their heads extensively (Figure 6B, F and H, and Videos 3–4). The dei larvae exhibited locomotion phenotypes that were similar to, though slightly less severe and more variable than, those of the dei larvae (Figure 6C, G and H, Videos 5–6). On average, the dei larvae crawled 51.9% ± 24.9% (n = 24) of the time and swiped their heads often; in 15.8% of the time their body angle was higher than 250° or lower than 110°, compared to 9.9% in dei and 1.2% of the time in wt larvae (Figure 6H). These results demonstrate that the dei regulatory element is crucial for proper function of the ChO as its deletion resulted in sensory dysfunction and uncoordinated movement.

Figure 6.

The dei enhancer is essential for normal larval locomotion.

(A–C) Crawling trajectories of 25 wild-type (A) dei (B) and dei (C) larvae. Each trajectory is shown in a square that represents 40 × 40 mm area. (D) Schematic representation of the body angle, γ, defined as the angle between the head and the body axis. (E–G) Representative time evolutions of the body angle of a wild-type (E), dei (F) and dei (G) larvae. The wild-type larva walks persistently, and the body angle stays 180 degrees most of the time (a 40 s interval is shown, after which the larva exited the filmed arena). The dei and dei mutant larvae display frequent changes in the direction of motion and long pauses accompanied by extensive head swiping (120 s intervals are shown). (H) A graph showing the average fraction of the time the larvae were crawling (GoPhase) and the fraction of time in which the body was bended more than 70 degrees (the measured angle γ was higher than 250 degrees or lower than 110 degrees). n = 24–27 for all genotypes; error bars represent the standard deviation. *** p < 0.0001, * p < 0.05, ns = non-significant. p Values were calculated using the unpaired two-tailed Mann-Whitney test.

Video 1.

A video showing the locomotion of a wt (Canton-S) larva.

Video 2.

A video showing the locomotion of a wt (Canton-S) larva.

Video 3.

A video showing the locomotion of a dei larva.

Video 4.

A video showing the locomotion of a dei larva.

Video 5.

A video showing the locomotion of a dei larva.

Video 6.

A video showing the locomotion of a deiΔChO larva.

Discussion

Opposing activities of D-Pax2/Sv and Pros dictate cap versus scolopale differentiation programs by regulating the dei gene

In this work, we identify a small GRN that governs the alternative differentiation programs of two cousins once removed cells within the ChO lineage - the cap cell and the scolopale cell. We show that Pros and D-Pax2/Sv are direct regulators of dei that together dictate its expression in the cap cell and its repression in the scolopale cell. Both D-Pax2/Sv and Pros exert their effects on dei transcription via a 262 bp chordotonal-specific enhancer (dei) in which two D-Pax2/Sv and three Pros binding sites were identified.

Following primary cell fate decisions within the ChO lineage, Pros expression becomes restricted to the scolopale cell (Doe et al., 1991; Vaessin et al., 1991), whereas D-Pax2/Sv expression becomes restricted to the scolopale and cap cells (Figure 1A), similar to its behavior in the external sensory lineages (Johnson et al., 2011; Kavaler et al., 1999). D-Pax2/Sv activates the expression of dei in the cap cell but is unable to do so in the scolopale cell where Pros is co-expressed. If D-Pax2/Sv activity is compromised, the cap cell fails to express dei and loses some of its differentiation markers, such as the expression of αTub85E. In contrast, if Pros activity is lost, dei is ectopically expressed in the scolopale cell that, as a consequence, acquires some cap cell features including the expression of αTub85E (Figure 7). The observed D-Pax2/Sv- and Pros-associated phenotypes do not reflect genuine cell fate transformations, suggesting that D-Pax2/Sv and Pros do not affect primary cell fate decisions within the ChO lineage. Rather, the observed phenotypes reflect a failure of the cap and scolopale cells to follow the cell type-specific differentiation programs responsible for bringing about their characteristic cellular phenotypes. The D-Pax2/Sv-deficient cap cells fail to express unique differentiation markers (such as αTub85E) and are therefore hardly detectable. It is also possible that the Sv/Pax2-deficient cap cells fail to survive. Thus, we cannot exclude the possibility that some of the findings reflect more upstream roles of Sv/D-Pax2 in the specification of cap-cell identity.

Figure 7.

Summary of the relations between Sv, Pros and Dei in the ChO lineage and their effect on ChO development.

(A) A wt ChO lineage. The cap (C) and ligament (L) cells are depicted in green, the scolopale cell (S) is depicted in orange and the neuron (N) is depicted in blue. Cells that expressed Dei are circled in light blue. (B) The loss of pros leads to upregulation of Dei in the scolopale cell and to failure of scolopale cell differentiation. In contrast, the loss of Sv leads to loss of Dei expression from the cap cell and failure of cap cell differentiation. The Sv-deficient scolopale cells are also abnormal. The CA cells which depend on the cap cell for their development/maintenence also appear abnormal is sv mutants. Misexpression of Pros leads to repression of Dei in the cap and ligament cells, preventing their normal differentiation. The Pros-expressing cap cells adopt some scolopale-specific features. In contrast, over-expression of Sv leads to ectopic expression of Dei. Due to the presence of Pros, the level of expression of dei in the scolopale cell is restricted. (C) A schematic summary showing the relations between Pros, Sv and Dei and their relations to cell-type-specific differentiation programs. In the CA cells, dei is activated by Sr via the dei enhancer. Both Sr and Dei are required there for the activation of CA-specific genes. In the cap cell dei is activated by Sv via the dei enhancer. Sv is required for activating cap-specific genes in both Dei-dependent and independent ways. In the scolopale cells, dei is repressed by Pros via the dei enhancer. Pros is required in addition for activating scolopale-specific genes. Dei is also expressed in the ligament cells and is required for their correct differentiation. The regulators of dei in the ligament cell are yet to be identified.

The switch between the differentiation programs of cap and scolopale identities cannot be simply explained by the nature of asymmetric cell divisions within the ChO lineage. The effects on the ChO lineage of major regulators of asymmetric cell division, such as Notch and Numb, and the expression pattern of cell differentiation determinants such as Pros and D-Pax2/Sv, were mainly postulated based on knowledge gained by analyzing external sensory lineages (Kavaler et al., 1999; Lai and Orgogozo, 2004; Manning and Doe, 1999; Reddy and Rodrigues, 1999). According to the similarity between the lineages, the cap cell parallels the Notch-non-responsive hair (trichogen) cell, whereas the scolopale parallels the Notch- responder sheath (thecogen) cell (based on Rebeiz et al., 2011). Thus, D-Pax2/Sv is expressed in one Notch responder and one non-responder cells in the lineage. The presence of Pros in the Notch-responder cell represses the cap-promoting activity of D-Pax2/Sv. Somewhat similar cousin-cousin cell transformation was found in external sensory organs in the adult where mutations in hamlet transform the sheath cell into a hair cell (parallel to scolopale-to-cap transformation) (Moore et al., 2004). Ectopic expression of hamlet induced pros expression and repressed the hair shaft-promoting activity of D-Pax2/Sv.

In the adult external sensory lineage, Pros was shown to be important for the specification of the pIIb precursor, which gives rise to the neuron and sheath cell (the scolopale counterpart). However, the absence of Pros from the pIIa precursor, which gives rise to the hair and socket cells (the cap and cap-attachment cells counterparts) was even more critical for proper development of this branch of the lineage (Manning and Doe, 1999). This phenomenon is somewhat conserved in the larval ChO. While Pros is required for proper differentiation of the scolopale cell, its absence from the cap cell is critical for adopting the correct differentiation programs within the lineage.

Opposing effects of D-Pax2/Sv and Pros activities on cell differentiation have been also identified in the regulation of neuronal versus non-neuronal cell fate decisions in the developing eye, where they play a role in modulating the Notch and Ras signaling pathway (Charlton-Perkins et al., 2011). Interestingly, in the R7 equivalence group Pros and D-Pax2/Sv can only alter the cell-type-specific differentiation program of cells that already express the other gene (Charlton-Perkins et al., 2011). Similarly, in the ChO lineage, ectopic expression of Pros in the cap and ligament cells transforms the D-Pax2/Sv-positive cap cell toward a scolopale cell identity but does not affect the D-Pax2/Sv-negative ligament cell in a similar fashion, even though the ectopic expression of Pros does repress the transcription of dei in both cell types. Additionally, a loss of Pros activity in the scolopale cell can transform the identity of this cell toward a cap cell identity only in the presence of D-Pax2/Sv.

We have shown that the opposing influences of Pros and D-Pax2/Sv on dei expression is integrated by the dei enhancer in both larval and adult ChO lineages. To the best of our knowledge, this is the first example of an enhancer that responds to these opposing signals to dictate cell-specific differentiation programs in a sensory lineage. While the identified enhancer is ChO-specific, it is plausible that other enhancers of sensory organ lineage-specific genes encode coupled Pros and D-Pax2/Sv binding sites. The expression of the dei gene in other (non-ChO) organs is regulated via different enhancers (as described in Nachman et al., 2015). Some of these enhancers are responsible for regulating dei expression in tissues where Pros and Pax2 play opposing roles, such as the eye and wing margin ES organs (the dei enhancer; Nachman et al., 2015). It is beyond the scope of this work, but in the future, it will be interesting to decipher whether these enhancers also serve as molecular platforms for integrating opposing effects of Pax2 and Pros.

The dei enhancer encodes non-canonical binding sites for Pros and D-Pax2/Sv

We have identified two D-Pax2/Sv and three Pros binding sites in the dei enhancer. Apart from D-Pax2/Sv site 2, none of these sites match the published binding motifs for D-Pax2/Sv or Pros. These results agree with recent studies that showed that many transcription factors function in vivo through low-affinity (Crocker et al., 2016; Crocker et al., 2015) or suboptimal (Farley et al., 2015) binding sites that differ from their predicted binding motifs. It was suggested that low-affinity binding sites provide specificity for individual transcription factors belonging to large paralogous families, such as the homeodomain family of transcription factors, that share similar DNA-binding preferences (Crocker et al., 2016; Crocker et al., 2015; Kribelbauer et al., 2019). To compensate for their weak binding capabilities, low-affinity binding sites are often organized in homotypic clusters that can increase the cumulative binding affinity of an enhancer (Crocker et al., 2016; Crocker et al., 2015; Kribelbauer et al., 2019). Our findings, that the homeodomain transcription factor Pros functions through a cluster of low-affinity binding sites in dei, may represent another example for the suggested tradeoff between transcription factor binding affinity and specificity (Crocker et al., 2016; Crocker et al., 2015).

We do not know how Pros opposes the effect of D-Pax2/Sv in the context of dei to inhibit the expression of dei in scolopale cells. Our results suggest that the inhibitory effect of Pros is not mediated through binding competition with D-Pax2/Sv at the D-Pax2/Sv high-affinity site (site 2), since this site does not overlap with a Pros-binding site (Figure 5B). The D-Pax2/Sv low-affinity site does overlap with a Pros binding site and mutations in the Pros binding site affect D-Pax2/Sv binding in vitro (see Figure 5—figure supplement 1D), however, while being important for robust dei expression, this site is dispensable in the presence of the high-affinity site. It is possible that binding of Pros to the dei enhancer targets this sequence to a repressed heterochromatin domain as was recently shown for other Pros target genes in differentiating neurons (Liu et al., 2020).

How is dei regulated in other ChO cell types? dei is expressed in four out of six cell types comprising the ChO: the cap-attachment and ligament-attachment cells, in which dei transcription is activated by Sr via the dei regulatory module (Nachman et al., 2015), and the cap and ligament cells in which the expression of dei is regulated via the dei enhancer. We now show that D-Pax2/Sv activates dei transcription in the cap cell, and that Pros inhibits its expression in the scolopale cell. The identity of the positive regulator/s of dei in the ligament cell, whose cell-fate is determined by the glial identity genes gcm and repo (Campbell et al., 1994; Halter et al., 1995; Jones et al., 1995), and the identity of the negative regulator/s of dei in the neuron remains unknown. Interestingly, the expression of dei was found to be altered in response to ectopic expression of gcm in the embryonic nervous system; its expression was upregulated at embryonic stage 11, but was repressed in embryonic stages 15–16 (Egger et al., 2002). This observation points to GCM as a potential regulator of dei expression in the ligament cells. Another interesting candidate for repressing dei in the sensory neuron is the transcriptional repressor Lola. Lola has been identified as a putative direct regulator of dei in the Y1H screen and was shown to be required in post-mitotic neurons in the brain for preserving a fully differentiated state of the cells (Southall et al., 2014). The possible involvement of Gcm and Lola in the regulation of dei awaits further studies. The observed upregulation of the dei reporter in the ligament cells of embryos with mutated Pros-binding sites may reflect an early role of Pros in the pIIb precursor before its restriction to the scolopale cell, which prevents dei expression in the ligament cell.

The dei-driven dei expression is critical for organ functionality

Although the loss of dei in the genetic/cellular milieu of the ligament cell (unlike the cap cell), even when accompanied by ectopic expression of Pros, is not sufficient for transforming ligament cell properties towards those of scolopale cells, we know that the expression of dei in the ligament cell is critical for its proper development. Ligament-specific knockdown of dei leads to failure of the ligament cells to acquire the right mechanical properties and leads to their dramatic over-elongation (Hassan et al., 2018). By analysing the locomotion phenotypes of larvae homozygous for a dei null allele and the newly generated cap and ligament-specific dei allele, we could show that the expression of dei in the cap and ligament cells is crucial for normal locomotion. Thus, we conclude that the correct expression of dei within the ChO is critical for organ functionality. Surprisingly, the gross morphology of LCh5 of dei larvae appears normal (Data not shown). Yet, in a way that remains to be elucidated, the Dei-deficient cap and ligament cells fail to correctly transmit the cuticle deformations to the sensory neuron, most likely due to changes in their mechanical properties.

Materials and methods

Fly strains

The following mutant and reporter alleles of sv and pros were used: sv (Kavaler et al., 1999), Dpax2 (Johnson et al., 2011), pros (BDSC:5458). The following Gal4 drivers and UAS strains were used: dei (Halachmi et al., 2016), dei (Hassan et al., 2018), en-Gal4 (Brand and Perrimon, 1993), P{UAS-3xFLAG-pros.S}14 c, y1 w*; Pin (BDSC:32245). UAS-sv-RNAi (VDRC:107343), UAS-sv (Kavaler et al., 1999), UAS-D-α-Catenin-GFP (Oda and Tsukita, 1999).

The dei driver was constructed by cloning the dei regulatory module described in Nachman et al., 2015 into the pChs-gal4 vector which was then used for the generation of transgenic fly strains (insertions are available on the X, 2nd and 3rd chromosome). The deiChO–262-lacZ strain was generated as previously described in Nachman et al., 2015. For mutational analysis of the putative binding sites, wild-type and mutated dei fragments were synthesized by GenScript (USA) and cloned into the reporter constructs placZattB. Plasmids were integrated into the attP2 landing site by BestGene Inc (Chino Hills CA, USA). (Supplementary file 2 lists all the transgenes used in this study, vectors used, landing sites and the sequence of the inserted mutations). The dei allele was generated by GenetiVision (Houston TX, USA) via multiplex targeting with two sgRNAs: 5’GCACTTGTTTGCGTTTACATTAC3’ and 5’GGCGAGAAGTATTCCCTGCG3’; creating a defined deletion of 307 bps spanning the dei fragment. The presence of the desired deletion was verified by sequencing. To verify that this deletion does not affect splicing or other structural features of the transcript, cDNA was synthesized and sequenced from the homozygous dei flies and control flies (the M{nos-Cas9.P}ZH-2A strain to which the injection was done). Total RNA was isolated from 10 adult flies using TRI Reagent (Sigma #T9424) according to the protocol described by Green et al. (Cold Spring Harb Protoc; doi:10.1101/pdb.prot101675). One mg of total RNA was used to generate cDNA using qScript cDNA Synthesis kit, according to the manufacturer protocol (Quanta BIOSCIENCE). PCR amplification was performed on 100 ng of cDNA using PrimeSTAR Max DNA polymerase (TAKARA #R045A), a forward primer from exon 1: TGCCAAATTTATGCATGAGC and reverse primer from exon 2: GCTTCTGTCGCAGGGAATAC.

Embryo staining and image analysis

Immunostaining of whole-mount embryos was performed using standard techniques. The following primary antibodies used in this study were: Rabbit anti Sv/D-Pax2 (1:10,000) (Johnson et al., 2011), Rabbit anti α85E-Tubulin (1:200) (Klein et al., 2010), Mouse anti α85E-Tubulin (1:20) (Nachman et al., 2015), Rabbit anti Dei (1:50) (Egoz-Matia et al., 2011), Rabbit anti Spalt (1:500) (Halachmi et al., 2007), Rat anti NRG (1:1000) (Banerjee et al., 2006), anti-Cpo (1:5000) (Bellen et al., 1992), and Mouse anti-βGal (1:1000, Promega). The following antibodies were obtained from the Developmental Studies Hybridoma Bank: Mouse anti Pros (MR1A, 1:20), Mouse anti Futsch (22C10, 1:20), Rat anti-ELAV (7E8A10, 1:50), Mouse anti-Eys (21A6, 1:20), Mouse anti-Repo (8D12, 1:10), Mouse anti Crb (Cq4, 1:10). Chicken anti-Sr (1:20) was made against amino acids 707–1180 of the Sr protein fused to GST in the pGEX-KG expression vector. The ~80 kDa fusion protein was purified on Glutathione-agarose beads followed by elution with reduced glutathione. Antibodies against the GST-Sr fusion protein were produced in Chickens by Dr. Enav Bar-Shira (Department of Animal Sciences, Robert H. Smith Faculty of Agriculture Food and Environment, The Hebrew University of Jerusalem, Rehovot, Israel). The IgY antibody fraction was isolated from the egg yolk and cleaned on Glutathione-agarose beads to reduce background of anti-GST antibodies. Secondary antibodies for fluorescent staining were Cy3, Cy2, Cy5 or Alexa Fluor-647-conjugated anti-Mouse/Rabbit/Rat/Chicken/Guinea pig (1:100, Jackson Laboratories, Bar-Harbor, Maine, USA). Stained embryos were mounted in DAKO fluorescence mounting medium (Agilent Technologies, Santa Clara, CA, USA) and viewed using confocal microscopy (Axioskop and LSM 510, Zeiss).

For the analysis of reporter gene expression, images were analyzed using ImageJ software (http://rsb.info.nih.gov/ij/) as previously described (Preger-Ben Noon et al., 2018). Briefly, maximum projections of confocal stacks were assembled, and background was subtracted using a 50-pixel rolling-ball radius. Then, we manually segmented visible nuclei of cap and scolopale cells from LCh5 of abdominal segments A2 and measured the fluorescence mean intensities of each nucleus. Statistical analyses and graphing were performed using GraphPad Prism version 8, GraphPad Software, La Jolla California USA, https://www.graphpad.com/.

X-Gal staining

Staining was done on staged pupae collected 40 hr after pupal formation. Pupae were removed from the pupal case and fixed at room temprature for 15 min in 4% formaldehyde in PBS. Following two washes in PBT (PBS + 0.1% Triton X-100), the pupae were incubated for five minutes in X-Gal staining buffer (without X-Gal) (5 mM K4[Fe+2(CN)6], 5 mM K3[Fe+2(CN)6]) in PBT and then incubated in staining buffer containing 1 mg/ml X-Gal for 4 hr at 37°C.

Yeast one-hybrid analysis

Yeast one-hybrid screening was performed by Hybrigenics Services, S.A.S., Evry, France (http://www.hybrigenics-services.com). The sequence of dei389 was PCR-amplified and cloned into the integrative vector pB301 (pAbAi, Clontech Laboratories, Inc). The construct was checked by sequencing the entire insert and transformed into the YM955 yeast strain to integrate the DNA bait into the yeast genome. Screening was performed against a random-primed Drosophila Whole Embryo cDNA library constructed into pP6 that derives from the original pGADGH (Bartel, 1993) plasmid. 124 million clones (12-fold the complexity of the library) were screened using a mating approach with YHGX13 (Y187 ade2-101::loxP-kanMX-loxP, matα) and the dei-containing yeast (mata) strain as previously described (Fromont-Racine et al., 1997). 146 His + colonies were selected on a medium lacking uracil, leucine and supplemented with 400 ng/ml Aureobasidin A. The prey fragments of the positive clones were amplified by PCR and sequenced at their 5’ and 3’ junctions. The resulting sequences were used to identify the corresponding interacting proteins in the GenBank database (NCBI) using a fully automated procedure. A confidence score (PBS, for Predicted Biological Score) was attributed to each interaction as previously described (Formstecher et al., 2005; Rain et al., 2001; Wojcik et al., 2002).

Motif search analysis

The D-Pax2/Sv site 2 was predicted by JASPAR (Mathelier et al., 2016). The PWM for Pros was generated using the sequences selected in TDA by Hassan et al., 1997 and the MEME suite (Bailey et al., 2015). The MEME suite was used to search for the Pros motif in dei and to compare the experimentally identified binding sites to the D-Pax2/Sv and Pros PWMs.

Protein purification

D-Pax2-HD and Pros-S-HD expression plasmids were a kind gift from Brian Gebelein and Tiffany Cook (Cook et al., 2003; Li-Kroeger et al., 2012). Proteins were purified from E. coli (BL21) as described previously (Uhl et al., 2010) with the following modifications. Protein expression was induced at 37 °C using 0.1 mM IPTG for 4 hr (Pros-S/L-HD) or 0.4 mM IPTG for 4 hr (D-Pax2-HD).

Pros-S-HD: After induction, bacterial pellet was resuspended in PBS supplemented with complete protease inhibitor mix (Roche) and lysed on ice using sonication (10 cycles of 30 s on/off). GST-tagged Pros-S-HD in soluble fraction was purified using Glutathione-Agarose beads (Sigma). The bound proteins were eluted in elution buffer (50 mM Tris, pH 8, 10 mM reduced Glutathione).

D-Pax2-HD: The bacterial pellet was resuspended in 20 mM Tris, pH 7.5 supplemented with protease inhibitor. His-tagged proteins in soluble fraction were purified using cOmplete His-Tag Purification Columns (Roche). The columns were washed with 10 column volumes of wash buffer 1 (20 mM Tris, pH7.5, 300 mM NaCl, 50 mM NaH2PO4 pH 8.0), 2 column volumes of wash buffer 2 (20 mM Tris, pH7.5, 300 mM NaCl, 5 mM DTT, 10 mM Imidazole), and bound proteins were eluted in the same buffer supplemented with 250 mM Imidazole.

All samples were dialyzed against 500 ml of dialysis buffer (20 mM HEPES, pH 7.9, 200 mM NaCl, 10% Glycerol, 2 mM MgCl) for 18 hr at 4 °C. Protein concentrations were measured with NanoDrop and confirmed by SDS-PAGE and Coomassie blue analysis.

Electromobility shift assay (EMSA)

DNA probes were generated by annealing 5’ IRDye700 labeled forward oligonucleotides with unlabeled reverse oligonucleotides (Integrated DNA Technologies) to a final concentration of 5 µM in PNK buffer (New England Biolabs). One hundred femtomoles of labeled IRDye700 probes were used in a 20 µl binding reaction containing 10 mM Tris, pH 7.5; 50 mM NaCl; 1 mM MgCl2; 4% glycerol; 0.5 mM DTT; 0.5 mM EDTA; 50 µg/ml poly(dI–dC); 200 µg/ml of BSA and purified proteins (see Supplementary file 3 for amount of each protein used). The binding reactions were incubated at room temperature for 30 min, and run on a native 4% polyacrylamide gel for 1.5 hr at 180 V. For competition assays, the appropriate amount of cold competitor was added with the IRDye700-labeled probe prior to the incubation. The polyacrylamide gel cassettes were imaged using an Odyssey Infrared Imaging System and image analysis was performed using ImageQuant 5.1 software. All experiments were performed at least three times.

Locomotion assays

Larvae used in the locomotion assays were collected from 8 to 12 hr egg collections that aged at 24°C until reaching the wandering 3rd instar stage (115–140 hr). 25–30 larvae of each genotype were individually transferred to a fresh 2% agar 10 cm plate, prewarmed to 24 °C. Larvae were let to adjust for 30 s prior to 2-min recording at a rate of 30 frames per second. The wild-type larvae often exited the filmed arena before the completion of the full 2 min recording time. Larval locomotion was recorded using a Dino-Lite digital microscope placed above the plate. We used VideoPad software to convert Dino-Lite files into Tiff files. ImageJ and FIMTrack (Risse et al., 2017) tracking software were used for following larval (center of mass) movements and body angle.

The study provides compelling evidence for a gene regulatory network involved in generating different sensory cell types from a common progenitor. The careful work shows how an enhancer can integrate the antagonistic relationship between two transcription factors for Drosophila sensory system development.

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "delilah, prospero and D-Pax2 constitute a gene regulatory network essential for the development of functional proprioceptors" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Utpal Banerjee as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Sonia Sen (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

The three reviewers found the work to be of quality but they have identified a few serious issues on three points.

The first one, shared by reviewers #1 and #3 is the significance of the findings that would gain by being more general. Of course, we do not want you to start new experiments on other systems, but you should discuss in more depth how the GRN and the regulations can apply to other sensory organs or to other systems.

There is a point that needs to be addressed and was raised by reviewer #3 about the nature of the enhancer deletion, to make sure that this does not affect splicing or other structural features of the transcript.

The binding sites for Pros and Pax2 appear to mediate regulation. However, you should show that you can really separate the binding of these proteins to the enhancer and that mutations in one site do not affect the other.

There is a list of other significant issues raised by the reviewers and you should try your best to address these suggestions.

Reviewer #1 (Recommendations for the authors):

In general, the paper expands on previous work by this group demonstrating that Pros and Pax2 play opposite roles on the dei ChO enhancer in "cousin" cell types. While well-executed, the overall impact of the study is rather limited, expecially given that a similar relationship has been observed in other development contexts. Providing additional mechanistic insight into Pros-Pax2 antagonism and discussion of these factors in broader contexts besides Drosophila sensory systems may increase impact.

Reviewer #3 (Recommendations for the authors):

(1) The yeast 1-hybrid screen is not well described. It is unclear what the authors mean by "two proteins were identified to bind the bait with high confidence" and what defines the other 3 candidates as being "moderate confidence"? The methods only describe that the "Aureobasidin A selection system" was used – with no description of what define high/moderate/low confidence. For those not familiar with this system the authors should clearly define in either the Results section or Methods how genes were defined as high/moderate/low confidence, how the results were validated, and how many independent clones of each gene were isolated – especially of D-Pax2/sv.

(2) In Figure 1E-1H, it is unclear what transgenes were used to study these enhancers, what location they are inserted into, and what stain is shown in the Figure. The Figure legend simply states embryonic expression patterns driven by each enhancer and the methods imply that the placZattb was used with a reference to Table S4. But I could only find one supplemental Table (the yeast 1-hybrid results). The authors need to (A) clearly state what transgenes were made in this study versus others, (B) state the explicit attP2 landing location used to insert each transgene, (C) state if each transgene was always inserted into the same or different landing locations for the comparative analysis, (D) state if similar results obtained using two different landing sites, and (E) state if all transgenic flies were studied as homozygotes (i.e. two copies of the transgene) or as heterozygotes, etc.

(3) In Figure 2J and 2L, the authors state that ectopic Pax2 induces extra Dei and deicho-lacZ expression – but it is unclear in how many cells? Given that Dei is expressed in other cell types – it is hard to tell how consistently Pax2 induces extra Dei expression. It would also be informative to know what specific cell types Pax2 is sufficient to induce Dei. Is Dei only induced by ectopic Pax2 within other cells of the ch organ lineage or does it also induce Dei in cells within other sensory organ lineages or even outside of the PNS? Addressing such questions would help to determine the sufficiency of DPax2 to induce this enhancer element.

(4) In Figure 4 – the authors need to provide more information and more controls when using and interpreting the data of the fly line with the 307bp enhancer deletion generated by genome editing. Since this deletion is located within an intron of the gene – and very close to the intron/exon boundary between intron 1 and exon2 – the authors need to both clearly highlight (A) where the intron/exon boundaries are within dei; (B) the exact location of the 307bp deletion; and (C) the authors should experimentally demonstrate that the dei transcript in the enhancer mutant fly line is not disrupted in any way that impacts the making of the Dei protein. This data is essential, as the authors conclude that failure to express the Dei protein in the cap/attachment cells is what leads to defective larval movement. However, if the intronic deletion also alters mRNA splicing and results in an abnormal transcript followed by an abnormal protein – then interpreting the phenotypes generated by these mutants is difficult given that Dei is not only expressed in the cap/attachment cells but also within tendon cells that could also impact Drosophila movement.

(5) In the EMSA methods, the authors site a Table SX with all the protein concentrations used in the in vitro DNA binding studies – but I could not find any table with such information.

(6) In Figure 5 – The authors highlight that the Pros site 1 and dPax2 site 1 overlap each other and the authors subsequently test a mutation for Pros site 1 and Pax2 site 1 in the reporter assays. However, it is unclear which specific mutations were tested in the reporter assays and importantly IF the Pros site 1 mutation was specific to Pros and does not impact Pax2 binding and IF the Pax2 site 1 mutation was specific to Pax2 and does not impact Pros binding. The authors need to explicitly state if each in vivo mutation was designed to be specific to either Pax2 or Pros binding or if the mutations impact the binding of both factors. Thus, it would be very helpful, if the authors showed the mutation sequence tested for each site listed under the sequence confirmation in panel B and use a name for each mutation that matches the mutations tested in the Supplemental Figures.

(7) With the exception of transgenic reporter expression in Figures 5M and 5N and the larvae movement assays, there is very little quantification of phenotypes throughout the paper. In point #4 above, I already mentioned one such example where Pax2’s impact on Dei and Dei enhancer activity could be quantified. While it may not be possible to quantify every type of expression changes studied throughout the manuscript, the authors should consider if it is possible to quantify the impact of Pros on Dei expression/Dei enhancer activity in a similar manner.

(8) To better understand the variability in DNA binding sites for Pros and Pax2, it would be helpful to align each of the Pax2 and Pros sites identified within the dei enhancer with each other and under the published consensus PWM motifs for Pax2 and Pros respectively. Such information would be very helpful given that the authors stated the identified sites, especially for Pros do not match the consensus motifs well.

(9) In Figure 2, the authors use two different sv alleles but do not describe either one. Are these alleles protein null alleles (no Pax2 protein detected) or hypomorphic? Moreover, why show the impact on Dei using the sv6 allele in panels A/B, but then show the impact on aTub85E in sv7 alleles in C/D? Were similar results observed in both? Would be better to show the data for one allele in Figure 2 and then include a supplemental figure with the other and explicitly state whether each give identical or different severities of phenotypes – ideally with quantitation and statistical analysis.

(10) The presentation of the results, especially the first few sections, made it difficult to assess what data was based on previous papers and what data was new in this study. For example, the first paragraph of the results describes the mapping of the minimal core dei enhancer (shown in Figure 1D through 1H) and the second paragraph describes an RNAi screen identifying shaven (DPax2) as a key regulator of this enhancer (not shown). However, both findings were previously published by the Salzberg lab – and thus while I found Figure 1 to be helpful to understand the paper, the authors should consider presenting these findings in the Introduction or only briefly summarized at the beginning of the results in a single statement or two.

(11) It would be more convincing if the authors quantify the number of ectopic Dei positive cells within an EnGal4 stripe vs between the EnGal4 stripes in the control (UAS-GFP vs UAS-sv/UAS-GFP) animals – which would determine autonomous vs non-autonomous impacts on Dei expression. In addition, the authors could use additional marker genes of specific cell fates to determine if Pax2 induces Dei in only ch organ lineages, in other sensory cell lineages, and/or in cells outside the PNS.

(12) The authors should use RT-PCR followed by sequencing to make sure the transcripts of the dei mRNA are properly spliced and not disrupted in flies homozygous for the deletion.

Essential revisions:

The three reviewers found the work to be of quality but they have identified a few serious issues on three points.

The first one, shared by reviewers #1 and #3 is the significance of the findings that would gain by being more general. Of course, we do not want you to start new experiments on other systems, but you should discuss in more depth how the GRN and the regulations can apply to other sensory organs or to other systems.

The role of opposing Pros and Pax2 activities in dictating cell specific differentiation programs in sensory organs has been previously documented (as thoroughly discussed in pages 14-16). However, to the best of our knowledge, our work identifies the first enhancer which serves as a molecular platform for integrating these opposing signals. While the identified dei enhancer is ChO-specific, it is reasonable to assume that other enhancers that regulate gene expression in other types of sensory organs encode coupled Pros and D-Pax2 binding sites. For example, the expression of dei itself in other (non-ChO) organs is regulated via different enhancers (as described in Nachman et al., 2015). Some of these enhancers are responsible for regulating dei expression in tissues where Pros and Pax2 play opposing roles, such as the eye and wing margin ES organs (the dei enhancer; Nachman et al., 2015). It is beyond the scope of this work, but in the future it will be interesting to decipher whether these enhancers also integrate opposing effects of Pax2 and Pros. These points are now included in the Discussion. To address reviewer’s #3 comment regarding the generality of the dei function in other ChOs, we tested the function of the wild type and mutated dei enhancer in the adult leg ChOs and found that the regulatory interactions within this GRN are conserved in the adult ChO lineages. This new data is presented in Figure 5—figure supplement 5. In addition, we highlight better the similar trends evident in other type of larval ChOs (Lch1, VChA/B) as pointed in Figures 2, 3, 5 and Figure 3—figure supplement 1.

There is a point that needs to be addressed and was raised by reviewer #3 about the nature of the enhancer deletion, to make sure that this does not affect splicing or other structural features of the transcript.

This point has been addressed and is discussed in detail in our response to comment #4 of reviewer #3.

The binding sites for Pros and Pax2 appear to mediate regulation. However, you should show that you can really separate the binding of these proteins to the enhancer and that mutations in one site do not affect the other.

In the revised manuscript we elaborate on the issue of overlap between Pros and D-pax2 binding sites. The introduced changes are described in detail in our response to comment #6 of reviewer #3.

Reviewer #3 (Recommendations for the authors):

1) The yeast 1-hybrid screen is not well described. It is unclear what the authors mean by "two proteins were identified to bind the bait with high confidence" and what defines the other 3 candidates as being "moderate confidence"? The methods only describe that the "Aureobasidin A selection system" was used – with no description of what define high/moderate/low confidence. For those not familiar with this system the authors should clearly define in either the Results section or Methods how genes were defined as high/moderate/low confidence, how the results were validated, and how many independent clones of each gene were isolated – especially of D-Pax2/sv.

Following the reviewer’s comment we have significantly expanded the description of the Y1H screen in the Results section and, mainly, in the Methods section. We now describe in greater details the constructs used, the use of Aureobasidin-A as a selection system, the number of clones screened, and number of clones sequenced (which is also shown in the Figure 1—figure supplement 1 and Supplementary File 1). We also describe what defines high/moderate/low confidence (see pages 6 and 21-22 and the legend to Figure 1—figure supplement 1). The relevant references were added to the manuscript.

2) In Figure 1E-1H, it is unclear what transgenes were used to study these enhancers, what location they are inserted into, and what stain is shown in the Figure. The Figure legend simply states embryonic expression patterns driven by each enhancer and the methods imply that the placZattb was used with a reference to Table S4. But I could only find one supplemental Table (the yeast 1-hybrid results). The authors need to A) clearly state what transgenes were made in this study versus others, B) state the explicit attP2 landing location used to insert each transgene, C) state if each transgene was always inserted into the same or different landing locations for the comparative analysis, D) state if similar results obtained using two different landing sites, and E) state if all transgenic flies were studied as homozygotes (i.e. two copies of the transgene) or as heterozygotes, etc.

The reporter constructs shown in Figure 1E-H were cloned into the pG-Pelican vector and transgenic strains were generated using P-element transformation. For each construct, independent transgenic strains with insertions on the X, 2nd and 3rd chromosomes were isolated and tested (described in Nachman et al., 2015). The dei reporter construct was later generated again using the pLacZ-attB vector and injected into the attP2 site (attP2 is the landing site located at 68A4), into which all the mutated reporter constructs were inserted as well. All the transgenic reporters have been studied as homozygotes. Supplementary File 2 (and the Star methods table) summarizes all the transgenes, vectors used and landing sites when known. We apologize for the loss of this table from the previous version of the manuscript.

3) In Figure 2J and 2L, the authors state that ectopic Pax2 induces extra Dei and deicho-lacZ expression – but it is unclear in how many cells? Given that Dei is expressed in other cell types – it is hard to tell how consistently Pax2 induces extra Dei expression. It would also be informative to know what specific cell types Pax2 is sufficient to induce Dei. Is Dei only induced by ectopic Pax2 within other cells of the ch organ lineage or does it also induce Dei in cells within other sensory organ lineages or even outside of the PNS? Addressing such questions would help to determine the sufficiency of DPax2 to induce this enhancer element.

It is very difficult to determine unambiguously how many cells express dei or the dei reporter ectopically upon Pax2 expression. The main reason is that ectopic expression of Pax2 has detrimental effects on the pattern of ChO migration and on other morphogenetic processes in the embryo. Thus, it is very difficult to identify cells based on their location and to some extent on the expression of other markers (for example, aTub85E is normally restricted to ChO cells, but upon ectopic Pax2 expression it is expressed in epidermal cells). However, clearly, cells outside the PNS express the dei gene and the dei reporter upon Pax2 ectopic expression. As seen in Figure 2J and L, the expression of dei is induced in the majority of epidermal cells within the en domain (labeled with GFP) that do not express it under normal conditions. The expression of Dei itself and the ChO marker aTub85E is also evident in many of these cells. In the revised manuscript we refer more clearly to this point (page 7 and the legend to Figure 2).

4) In Figure 4 – the authors need to provide more information and more controls when using and interpreting the data of the fly line with the 307bp enhancer deletion generated by genome editing. Since this deletion is located within an intron of the gene – and very close to the intron/exon boundary between intron 1 and exon2 – the authors need to both clearly highlight (A) where the intron/exon boundaries are within dei; (B) the exact location of the 307bp deletion; and (C) the authors should experimentally demonstrate that the dei transcript in the enhancer mutant fly line is not disrupted in any way that impacts the making of the Dei protein. This data is essential, as the authors conclude that failure to express the Dei protein in the cap/attachment cells is what leads to defective larval movement. However, if the intronic deletion also alters mRNA splicing and results in an abnormal transcript followed by an abnormal protein – then interpreting the phenotypes generated by these mutants is difficult given that Dei is not only expressed in the cap/attachment cells but also within tendon cells that could also impact Drosophila movement.

We fully agree with this comment and followed the reviewer’s suggestion. To verify that deleting the intronic enhancer does not affect splicing or other structural features of the transcript, cDNA was synthesized from homozygous dei and control flies. A 416 bp fragment was amplified by PCR from the cDNA samples using primers located on both sides of the intron (in the 1st and 2nd exons). Sequencing of the PCR products verified the presence of normally structured dei transcript in the dei mutant. We have now added this new experimental data to the manuscript together with more detailed presentation of the gene structure (result section – page 9; Materials and methods, and Figure 3—figure supplement 2).

5) In the EMSA methods, the authors site a Table SX with all the protein concentrations used in the in vitro DNA binding studies – but I could not find any table with such information.

We thank the reviewer for drawing our attention to this missing information. We have now added the missing table (Supplementary File 3). The table includes the sequence of the oligos used in the different EMSAs along with the concentrations of the proteins and the associated Figures. We also included a table that summarizes all the transgenic lines used in this study (Supplementary File 2).

6) In Figure 5 – The authors highlight that the Pros site 1 and dPax2 site 1 overlap each other and the authors subsequently test a mutation for Pros site 1 and Pax2 site 1 in the reporter assays. However, it is unclear which specific mutations were tested in the reporter assays and importantly IF the Pros site 1 mutation was specific to Pros and does not impact Pax2 binding and IF the Pax2 site 1 mutation was specific to Pax2 and does not impact Pros binding. The authors need to explicitly state if each in vivo mutation was designed to be specific to either Pax2 or Pros binding or if the mutations impact the binding of both factors. Thus, it would be very helpful, if the authors showed the mutation sequence tested for each site listed under the sequence confirmation in panel B and use a name for each mutation that matches the mutations tested in the Supplemental Figures.

We thank the reviewer for drawing our attention to this missing information. We have now revised Figure 5 to include the mutated nucleotides that were tested in transgenic flies. In addition, we added a Supplementary Figure (Figure 5—figure supplement 3) and Table (Supplementary File 2) that presents the full sequence of all the mutants used for the in vivo assays.

Regarding the overlap between Pros site1 and D-Pax2 site 1: The mutation in Pax2 site 1 was specific to the Pax2 site and did not affect Pros binding to the Pros site 1 in EMSA. As shown in Figure 5—figure supplement 1D, mutH in probe 3b, which is equivalent to the in vivo mutation in Pax2 site 1, was able to compete the WT probe for the binding to Pros. The mutation in Pros site 1, on the other hand, did affect the binding of Pax2 to Pax2 site 1 in vitro. As shown in Figure 5—figure supplement 1D, mutE in probe 3b, which is equivalent to the in vivo mutation in Pros site 1, failed to compete the WT probe for the binding Pax2. As mutations in Pax2 site 1 did not affect the expression of the reporter gene in cap cells on its own (Figure 5M), we do not anticipate that the mutation in Pros site 1 will have a major effect on the function of dei. Saying that, it is possible that if Pax2 binding was not affected by the mutation in Pros site 1, we would observe even higher elevation in the expression levels of the reporter in scolopale cells of this line (Figure 5N). We have revised the manuscript to address this concern (page 12).

7) With the exception of transgenic reporter expression in Figures 5M and 5N and the larvae movement assays, there is very little quantification of phenotypes throughout the paper. In point #4 above, I already mentioned one such example where Pax2’s impact on Dei and Dei enhancer activity could be quantified. While it may not be possible to quantify every type of expression changes studied throughout the manuscript, the authors should consider if it is possible to quantify the impact of Pros on Dei expression/Dei enhancer activity in a similar manner.

While such quantifications could be desirable, we feel that they are not necessary because the effects of Pros and Pax2 on dei expression shown in Figures 2-4 are discussed in terms of spatial distribution rather than level of expression. At this point, quantifying the level of expression in the different genotypes is impossible since only in the experiments shown in Figure 5 all the samples were imaged using the exact same imaging parameters thus allowing for quantitative comparisons. It is important to note that the representative pictures showed in the manuscript represent a large number of examined segments and that the observed phenotypes were fully penetrant. For example, we have examined 116 segments of pros mutant embryos, all of them showed the described ectopic expression of dei.

8) To better understand the variability in DNA binding sites for Pros and Pax2, it would be helpful to align each of the Pax2 and Pros sites identified within the dei enhancer with each other and under the published consensus PWM motifs for Pax2 and Pros respectively. Such information would be very helpful given that the authors stated the identified sites, especially for Pros do not match the consensus motifs well.

We thank the reviewer for this good advice. We have added Figure 5—figure supplement 4 that shows the suggested alignments, which are also discussed in the text (in the results -page 11 and the Discussion – page 16).

9) In Figure 2, the authors use two different sv alleles but do not describe either one. Are these alleles protein null alleles (no Pax2 protein detected) or hypomorphic? Moreover, why show the impact on Dei using the sv6 allele in panels A/B, but then show the impact on aTub85E in sv7 alleles in C/D? Were similar results observed in both? Would be better to show the data for one allele in Figure 2 and then include a supplemental figure with the other and explicitly state whether each give identical or different severities of phenotypes – ideally with quantitation and statistical analysis.

In the revised version of the manuscript, we present in Figure 2 a single sv allele (sv) which is a protein null allele.

10) The presentation of the results, especially the first few sections, made it difficult to assess what data was based on previous papers and what data was new in this study. For example, the first paragraph of the results describes the mapping of the minimal core dei enhancer (shown in Figure 1D through 1H) and the second paragraph describes an RNAi screen identifying shaven (DPax2) as a key regulator of this enhancer (not shown). However, both findings were previously published by the Salzberg lab – and thus while I found Figure 1 to be helpful to understand the paper, the authors should consider presenting these findings in the Introduction or only briefly summarized at the beginning of the results in a single statement or two.

We fully agree with this comment and revised the first paragraph of the Results section (pages 5-6) accordingly (without moving Figure 1). We hope that the distinction between old and new data is much clearer now.

11) It would be more convincing if the authors quantify the number of ectopic Dei positive cells within an EnGal4 stripe vs between the EnGal4 stripes in the control (UAS-GFP vs UAS-sv/UAS-GFP) animals – which would determine autonomous vs non-autonomous impacts on Dei expression. In addition, the authors could use additional marker genes of specific cell fates to determine if Pax2 induces Dei in only ch organ lineages, in other sensory cell lineages, and/or in cells outside the PNS.

It is not clear to us why the question of autonomous versus non-autonomous effect of Pax2 on Dei expression is raised as none of the experimental results support a non-autonomous effect. Specifically for the induction of Pax2 expression under the regulation of en-gal4, such analysis is tricky because the ChO cells are born within the en domain and then migrate out to a more anterior compartment. The pattern of cell migration is disrupted when Pax2 is expressed, making the analysis of autonomous vs non-autonomous effects quite impossible. The data in Figure 2J, L demonstrates unambiguously that Pax2 can induce the ectopic expression of dei outside the PNS.

12) The authors should use RT-PCR followed by sequencing to make sure the transcripts of the dei mRNA are properly spliced and not disrupted in flies homozygous for the deletion.

We have followed the reviewer’s recommendation and used RT-PCR followed by sequencing to verify the correct structure of the dei transcript in the enhancer deletion mutant. The results are described in the text and in Figure 4—figure supplement 1.

Key resources table

Reagent type (species) or resource	Designation	Source or reference	Identifiers	Additional information
Gene (Drosophila melanogaster)	dei/tx	FlyBase	CG5441, FBgn0263118
Gene (Drosophila melanogaster)	pros	FlyBase	CG17228, FBgn0004595
Gene (D. melanogaster)	sv/D-Pax2	FlyBase	CG11049, FBgn0005561
Genetic reagent (D. melanogaster)	sv 6 /act-GFP	Kavaler et al., 1999	N/A
Genetic reagent (D. melanogaster)	Dpax2D1-GFP	Johnson et al., 2011	N/A
Genetic reagent (D. melanogaster)	pros17/TM6B, Tb1	Bloomington Drosophila Stock Center	BDSC:5458
Genetic reagent (D. melanogaster)	dei ChO-1353 -GFP,dei attachment -RFP;en-Gal4	Halachmi et al., 2016	N/A
Genetic reagent (D. melanogaster)	dei KO-mCherry	Hassan et al., 2018	N/A
Genetic reagent (D. melanogaster)	en-Gal4	Brand and Perrimon, 1993
Genetic reagent (D. melanogaster)	P{UAS-3xFLAG-pros.S}14 c, y1 w*; Pin1/CyO	Bloomington Drosophila Stock Center	BDSC:32245
Genetic reagent (D. melanogaster)	UAS-sv-RNAi	Vienna Drosophila Resource Center	VDRC:107343
Genetic reagent (D. melanogaster)	UAS-sv	Kavaler et al., 1999	N/A
Genetic reagent (D. melanogaster)	UAS-D-α-Catenin-GFP	Oda and Tsukita, 1999	N/A
Genetic reagent (D. melanogaster)	deiattachment-lacZ	Nachman et al., 2015	N/A
Genetic reagent (D. melanogaster)	deiChO-1353-lacZ	Nachman et al., 2015	N/A
Genetic reagent (D. melanogaster)	deiChO-1353 -GFP	Halachmi et al., 2016	N/A
Genetic reagent (D. melanogaster)	deiChO-389 -lacZ	Nachman et al., 2015	N/A
Genetic reagent (D. melanogaster)	deiChO-Gal4	This study	N/A
Genetic reagent (D. melanogaster)	deiChO-262- lacZ(in pH-Pelican)	This study	N/A	P element transgenesis. Available on 1st, 2nd, and 3rd chromosomes
Genetic reagent (D. melanogaster)	deiChO-262-placZ-wildtype(in placZattB)	This study	N/A	Inserted at attP2
Genetic reagent (D. melanogaster)	deiChO-262-placZ-Pros-site 1(in placZattB)	This study	N/A	Inserted at attP2
Genetic reagent (D. melanogaster)	deiChO-262-placZ-Pros-site 2(in placZattB)	This study	N/A	Inserted at attP2
Genetic reagent (D. melanogaster)	deiChO-262-placZ-Pros-site 3(in placZattB)	This study	N/A	Inserted at attP2
Genetic reagent (D. melanogaster)	deiChO-262-placZ-Pros-site 1 + 2 + 3(in placZattB)	This study	N/A	Inserted at attP2
Genetic reagent (D. melanogaster)	deiChO-262-placZ-Pax2-site 1(in placZattB)	This study	N/A	Inserted at attP2
Genetic reagent (D. melanogaster)	deiChO-262-placZattB Pax2-site 2	This study	N/A	Inserted at attP2
Genetic reagent (D. melanogaster)	deiChO-262-placZ-Pax2-site 1 + 2(in placZattB)	This study	N/A	Inserted at attP2
Genetic reagent (D. melanogaster)	deiΔChO/(TM6)	This study	N/A
Genetic reagent (D. melanogaster)	M{nos-Cas9.P}ZH-2A	Bloomington Drosophila Stock Center	BDSC:54591
Antibody	Anti Sv/D-Pax2 (rabbit polyclonal)	Johnson et al., 2011	N/A	(1:10,000)
Antibody	Anti-α85E-Tubulin (rabbit polyclonal)	Klein et al., 2010	N/A	(1:200)
Antibody	Anti α85E-Tubulin (mouse monoclonal)	Nachman et al., 2015	N/A	(1:20)
Antibody	Anti-Dei (rabbit polyclonal)	Egoz-Matia et al., 2011	N/A	(1:50)
Antibody	Anti-Spalt (rabbit polyclonal)	Halachmi et al., 2007	N/A	(1:500)
Antibody	Anti-NRG (rat polyclonal)	Banerjee et al., 2006	N/A	(1:1000)
Antibody	Anti-βGal (mouse monoclonal)	Promega	Z3781	(1:1000)
Antibody	Anti-Pros (mouse monoclonal)	The Developmental Studies Hybridoma Bank	MR1A	(1:20)
Antibody	Anti Futsch (mouse monoclonal)	The Developmental Studies Hybridoma Bank	22C10	(1:20)
Antibody	Anti-ELAV (rat monoclonal)	The Developmental Studies Hybridoma Bank	7E8A10	(1:50)
Antibody	Anti-Eys (mouse monoclonal)	The Developmental Studies Hybridoma Bank	21A6	(1:20)
Antibody	Anti-Repo (mouse monoclonal)	The Developmental Studies Hybridoma Bank	8D12	(1:10)
Antibody	Anti Crb (mouse monoclonal)	The Developmental Studies Hybridoma Bank	Cq4	(1:10)
Antibody	Anti-Cpo (rabbit polyclonal)	Bellen et al., 1992	N/A	(1:5000)
Antibody	Anti-Sr (Chicken polyclonal)	This study	N/A	(1:20)
Antibody	Cy3-conjugated goat anti-mouse	Jackson Laboratories, Bar-Harbor, Maine, USA	115-165-166	(1:100)
Antibody	Cy3-conjugated goat anti-rabbit	Jackson Laboratories, Bar-Harbor, Maine, USA	115-165-144	(1:100)
Antibody	Cy2-conjugated goat anti-rabbit	Jackson Laboratories, Bar-Harbor, Maine, USA	111-225-144	(1:100)
Antibody	Alexa Fluor-647-conjugated goat anti-mouse	Jackson Laboratories, Bar-Harbor, Maine, USA	115-605-166	(1:100)
Antibody	Alexa Fluor-647-conjugated goat anti-rabbit	Jackson Laboratories, Bar-Harbor, Maine, USA	115-605-144	(1:100)
Antibody	Cy3-conjugated donkey anti-chicken IgY	Jackson Laboratories, Bar-Harbor, Maine, USA	703-165-155	(1:100)
Antibody	Alexa Fluor-647-conjugated Donkey anti-Rat	Jackson Laboratories, Bar-Harbor, Maine, USA	712-605-153	(1:100)
Commercial assay or kit	qScript cDNA Synthesis kit	Quanta BIOSCIENCES	95047–100
Commercial assay or kit	PrimeSTAR Max DNA polymerase	TAKARA	R045A
Chemical compound, drug	DAKO fluorescence mounting medium	Agilent Technologies, Santa Clara, CA, USA	S3023
Chemical compound, drug	TRI Reagent	Sigma	T9424
Software, algorithm	ImajeJ	National Institute of Health	SCR_003070
Software, algorithm	FIMTrack tracking software	Risse et al., 2017	https://www.uni-muenster.de/FRIA/en/FIM/
Software, algorithm	MATLAB	Mathworks	SCR_001622
Software, algorithm	Imaris	Bitplane	SCR_007370