Onset of atonal expression in Drosophila retinal progenitors involves redundant and synergistic contributions of Ey/Pax6 and So bindings sites within two distant enhancers

Abstract

Proneural transcription factors drive the generation of specialized neurons during nervous system development, and their dynamic expression pattern is critical to their function. The activation of the proneural gene atonal (ato) in the Drosophila eye disc epithelium represents a critical step in the transition from retinal progenitor cell to developing photoreceptor neuron. We show here that the onset of ato transcription depends on two distant enhancers that function differently in subsets of retinal progenitor cells. A detailed analysis of the crosstalk between these enhancers identifies a critical role for three binding sites for the Retinal Determination factors Eyeless (Ey) and Sine oculis (So). We show how these sites interact to induce ato expression in distinct regions of the eye field and confirm them to be occupied by endogenous Ey and So proteins in vivo. Our study suggests that Ey and So operate differently through the same 3′ cis-regulatory sites in distinct populations of retinal progenitors.

Introduction

Nervous system development involves the generation of specific neuronal types in complex and precise spatial and temporal patterns. The proneural factors, transcriptional regulators of the basic helix–loop–helix (bHLH) type, play a critical role in this process by conferring neural competence to groups of cells and thereby imposing both spatial and temporal developmental guidelines. Different proneural factors are not only necessary to initiate neurogenesis, but are also often sufficient to induce formation of specific neuronal types (Brunet and Ghysen, 1999; Fode et al., 2000; Inoue et al., 2001; Jarman et al., 1993; Lee et al., 1995; Ma et al., 1996; Ma et al., 1998; Takebayashi et al., 1997; Vetter and Brown, 2001). Proneural factor expression is dynamic and transient. As the first step in the complex process of neurogenesis, the carefully orchestrated induction of proneural gene transcription reflects the synthesis of inductive inputs by signaling pathways and instructive inputs by specification factors.

The role of proneural genes in neurogenesis has been studied extensively in the sensory organs of both vertebrates and invertebrates. In the developing fly eye, the proneural gene atonal (ato) is expressed transiently with exquisite temporal and spatial precision to initiate eye morphogenesis. At first, ato is expressed broadly and confers neural competence to retinal progenitors. Then, its expression becomes restricted in a process that ensures the correct cellular composition and spacing of some 750-800 single eye units (ommatidia). Early in the third and last larval period (L3) of Drosophila, hundreds of ommatidia composed of 8 neurons (R-cells) plus accessory cells begin to develop within the eye-antennal imaginal disc (the fly head-forming epithelium) and continue to form over a ∼3 day period (Wolff and Ready, 1993). This process begins when eye progenitor cells stop dividing and initiate neurogenesis, starting with the activation of ato transcription. However, instead of triggering ato expression synchronously across the entire eye field, groups of retinal progenitor do so in a sequential pattern. Progenitor cells at the very posterior margin of the disc activate ato transcription first, generating a dorsoventral stripe of expression (Fig. S1; Jarman et al., 1994; Jarman et al., 1995). Soon after, this stripe changes into a row of evenly spaced cell clusters, and then into a line of single, ato-positive cells (the future R8 photoreceptors) (Fig. S1). As this change occurs, the eye progenitor cells located immediately anterior activate ato transcription; once again in a broader stripe that morphs then into the more restricted pattern linked to R8 formation (Fig. S1). This dynamic pattern of ato expression is repeated across the disc epithelium such that a ‘stripe of gene activation’ appears to move across the eye-field from posterior to anterior (Figs.1A; S1) leaving in its wake emerging neuronal clusters with ato-expressing R8 neurons. The initial broad activation of ato transcription in a stripe is the critical first step in the onset of eye neurogenesis and marks the transition from retina progenitors to developing ommatidial clusters.

Fig. 1. Identification of 3′ENH^I enhancer region.

In this and all other figures: discs orientation is as per A<->P, D<->V coordinates; arrowheads indicate the position of the morphogenetic furrow (MF), an indentation in the disc epithelium that marks the transition zone from ato expressing progenitor cells to developing neurons; size bars = 25 μm; (A) In situ detection of ato mRNA in posterior progenitors (green) and anterior progenitors (magenta) and the corresponding genomic regions (∼3.6 kb and 338 bp) associated with expression. Note that the disc continues to proliferate becoming larger as progressively more anterior cells transiently express ato. Central diagram shows the disc epithelium divided in anterior-progenitors region (green) and posterior-progenitors region (magenta) based on previous reports (Tanaka-Matakatsu and Du, 2008; Zhang et al., 2006; see Fig. S2). Bottom diagram shows the 3′ regulatory region downstream of the ato transcription unit (t.u.) in relation to the expression shown above and the genomic DNA fragments shown in B and C. (B) A 288 bp DNA region is shared among reporters expressed in posterior progenitors and thus identifies the potential site for 3′ENH^I. Diagram of reported expression patterns are shown to the right: the 3.6BP drives expression in only posterior cells; the M” (∼2 kb) and the 1.2 fragments drive expression in both anterior and posterior regions in essentially indistinguishable patterns. Hollow bars mark the location of two evolutionarily conserved DNA sequences (IC1 and IC2) identified in Zhang et al. (2006). (C) Deletion analysis of the 288 bp interval maps the 3′ENH^I enhancer to a smaller, 183 bp region. First 3 panels show p-galactosidase staining, right-most panel shows in situ hybridization to reporter mRNA (eGFP). Though not expressed in posterior cells, as shown here, the 1.2-A281 and M”-A183 are normally expressed in anterior progenitors (see Fig. S3A). (D) Summary of the analysis showing location of the two enhancers 3′ENH^I and 3′ENH^II downstream of the ato t.u.

How the expression of ato, and its vertebrate ortholog Ath5 (also known as Atoh7), is first induced has been subject of studies in fly, frog, chick and mouse, leading to the identification of several highly conserved cis regulatory elements (Hufnagel et al., 2007; Hutcheson et al., 2005; Matter-Sadzinski et al, 2001; Matter-Sadzinski et al, 2005; Skowronska-Krawczyk et al., 2004; Sun et al., 1998; Sun et al., 2003; Riesenberg et al., 2009; Tanaka-Matakatsu and Du, 2008; Willardsen et al., 2009; Zhang et al., 2006). Relatively unexplored, however, are the specific mechanisms of enhancer control, i.e. how the readout from some of these sequences is integrated. In the fly, genetic evidence points to multiple regulatory inputs including cell-cell signaling cascades (such as the Hh and Dpp/BMP4 pathways) and several transcription factors (particularly the Retinal Determination (RD) proteins -a cascade of transcription factors that together confer eye identity and establish an eye organ primordium in the L2 disc) (Bonini et al., 1997; Borod and Heberlein, 1998; Greenwood and Struhl, 1999; Pignoni et al., 1997; Shen and Mardon, 1997). Evidence of direct regulation, however, is only available for some of the RD factors and in only a subset of the progenitor cells (Tanaka-Matakatsu and Du, 2008; Zhang et al., 2006).

Based on these studies, a distal 3′ enhancer element appears to control expression specifically in progenitor cells located in the mid-to-anterior portion of the eye disc (called here ‘anterior progenitors’) (Fig. 1A). This element (named here 3′ENH^II) is located ∼4.2 kb downstream of the ato transcription unit and contains consensus binding sites for the RD factors Eyeless/Pax6 (Ey) and Sine oculis (So) (Zhang et al., 2006). Zhang and colleagues (2006) were able to show that both sites are required in vivo and confirmed binding by Ey and So in vitro (Fig. 1A). Less is known about the control of ato induction in retinal progenitors located along or near the posterior margin of the disc (called here ‘posterior progenitors’). One study suggests that ato expression in these cells depends on putative regulatory element (s) also located in the 3′ flanking genomic DNA, but within 3.6 kb of the ato transcription unit and thus upstream of the 3′ENH^II enhancer (Fig. 1A). However, lack of information about the latter regulatory element and about how these enhancers work separately and together severely limits our current understanding of ato regulation.

Here we report the identification of the putative posterior progenitors enhancer (3′ENH^I) and investigate its function vis-à-vis the more distal 3′ enhancer element (3′ENH^II) (Fig. 1A). We find that neither regulatory element is actually specific to posterior or anterior retinal progenitor cells, but that the two enhancers work together in all cells of the eye field to generate the ato expression pattern. The 3′ENH^I enhancer lies ∼600 bp upstream of 3′ENH^II and also contains a critical RD-factor binding site, specifically for Ey. We show that the two Pax6 sites and the one So site present in these two enhancers are required for gene expression in all eye progenitor cells, display different types of synergism in posterior and anterior progenitor cell populations, and are occupied by the Ey and So proteins in vivo. This work suggests a more complex regulation of ato expression by both Ey and So in retinal progenitors, and leads to a novel view of ato induction in the developing eye field.

Materials and Methods

Genetics

We utilized a combination of traditional P-element transgenesis as in the earlier studies (Tanaka-Matakatsu and Du, 2008; Zhang et al., 2006) and the more recent site-specific-insertion method (Bischof et al., 2007) to generate transgenic lines in our lab. For P-element transformation, DNA fragments were cloned into the modified pCasper-β-gal or pStinger (eGFP) vectors (Barolo et al., 2000; Spradling and Rubin, 1982) modified to contain the 1.1 kb region from the atonal locus (Sun et al., 1998; Zhang et al., 2006). In all cases of P-element-mediated insertion, multiple independent lines were analyzed and results are based on consistent data from all or a majority of the lines generated (see Fig. S2 for summary of constructs analyzed). For site-specific transformation, a SphI restriction site was first introduced into the pattB vector. Then, SphI-SpeI DNA fragments containing the enhancer regions to be tested, the 1.1 kb, and the eGFP-NLS-SV40 region from pStinger were cloned into the modified pattB vector cut with SphI/XbaI. Sequences of all primers are reported in Table S1. A complete list of constructs and transgenic lines is provided in Supplemental Fig. S2. All constructs were confirmed by sequencing. Fly lines used: attP (y¹ M{vas-int.Dm}ZH-2A w*; M{3xP3-RFP.attP'}ZH-51C (Bischof et al., 2007; BDSC) fly for injection of site-specific constructs; w Canton S (BDSC) for injection of P-element constructs; UAS-ey (Pignoni et al., 1997); 30A-Gal4 (Brand and Perrimon, 1993); dpp-Gal4 (Staehling-Hampton et al., 1994); UAS-wg ^RNAi (VDRC13352); UAS-dsh ^RNAi (TRiP. JF01253); UAS-arm^RNAi (TRiP. JF01251); and UAS-tsh (Gallet et al., 1998). Strong activity of the RNAi lines was confirmed using constitutive expression throughout the eye-antennal disc (as in Zhang et al., 2011).

Immunohistology and β-galactosidase stanining

Standard Ab and β-gal staining protocols were used (Sullivan et al., 2000). Ab used: rat anti-Elav (1:100; DHB), mouse anti-Eya (1:200; DHB), guinea pig anti-Sens (1:1000) (Nolo et al., 2000), rabbit anti-β-gal (1:1000; Cappell), rabbit anti-GFP (1:1000; Upstate Biotech). Anti-mouse, anti-rabbit and anti-guinea pig Cy2-, Cy3- or Cy5-conjugated (Jackson Immuno Research Laboratories) secondary Ab were used at 1:200 dilution. Discs of were analyzed and imaged on a Nikon E600 microscope (DIC & fluorescence) and a Leica DM5500 Q confocal microscope.

In situ hybridization and transcripts level quantification

In situ hybridization experiments were carried out per standard protocol (Sullivan et al., 2000) using DIG-labeled probes (DIG RNA Labeling Kit; Roche). Samples were processed in parallel, under identical conditions through, the staining reaction. Discs were imaged on a Leica DM5500 Q microscope using DIC and identical settings (8-bit images; 256 gray values). For each type of construct, images of 4-7 discs from 2-3 independent lines were analyzed using the gray value measurement function in Photoshop (version CS4). Mean gray values were computed for the same central area of the eye disc, where the MF crosses the disc midline. Relative Optical Density (ROD) was calculated using the formula: ROD = log (256/mean gray value).

Electrophoretic mobility shift assay (EMSA)

Ey protein was produced using a reticulocyte lysate in vitro transcription-translation system (Promega). DNA fragments used as probes were generated by PCR. Competitor fragments including the full length 120 bp fragment III and short double stranded DNA fragement (∼40 bp) containing the Pax6 site, unchanged or mutated (see Figs. 4A & S4), were generated by annealing of synthesized oligonucleotides. Labeling of DNA to generate probes and gel shift reactions were carried out as per the DIG Gel Shift Kit -2^nd generation (Roche) using 5% non-denaturing polyacrylamide gel. For competition experiments, 100x excess unlabeled DNA was used. Signals were detected by a phosphorimager.

Fig. 4. Identification of the Pax6^EI binding site and interactions between Pax6 and So sites in posterior progenitors.

(A) EMSA for the binding of Ey to 3′ENH^I. DNA fragment III contains a Pax6 consensus site (asterisk) and shows an Ey-protein-dependent shift in vitro (arrow). The shift is suppressed by unlabelled competitor DNA containing a wt Pax6 site (+), but not with a mutated site (+^m). (B) Diagram of binding site deletions used in C-F. (C-D) Panels show reporter protein expression (green) in posterior progenitors. Discs were also stained for the proteins Eya (blue) to mark all retina cells and the pan-neural marker Elav (red) to identify discs in the early stages of neurogenesis (i.e. with few rows of Elav-positive developing photoreceptor neurons). Discs are essentially wild type thanks to the normal expression and function of the endogenous ato gene. Stainings are shown for early discs just before formation of the first row of neurons (C) and after a few rows have emerged (D). (C) First panel: M” shows strong expression in cells along the entire posterior margin of the disc. Second panel: Deletion of the Pax6^EI site results in a drastic loss of reporter protein expression in posterior margin cells similarly to deletions of So^EII or Pax6^EII (compare to single deletions in Fig. 3). Third and fourth panels: reporter protein expression is completely lost when Pax6^EI is deleted along with Pax6^EII (M”-ΔPax6^EIPax6^EII) or So^EII (M”-ΔPax6^EISo ^EII). (D) Slightly older discs show consistent results than in C in posterior progenitors away from the margin. (E-F) In situ hybridization to reporter gene mRNA confirms regulation at the level of transcription for the double-deletion constructs M”-ΔPax6^EIPax6^EII and M”-ΔPax6^EISo ^EII in both posterior margin cells (E) and cells some distance away from the margin (F).

Chromatin immunoprecipitation (ChIP)

ChIP for So binding was performed as previously described (Jemc and Rebay, 2007). PCR primers used were ato-S (5′-AGTATTCCGCATTTGGCAACAC-3′) and ato-A (5′-GCACTCTGGACGCATTTTTCAC-3′). ChIP for Ey binding was performed as follows. Four hundred pairs of eye-antennal discs from 3^rd instar larvae were dissected and crosslinked with 1% formaldehyde at room temperature for 15 minutes. After three PBS washes, the discs were lysed and sonicated with Branson Digital Sonifier. The sonicated chromatin was pre-cleared with protein A sepharose beads (#17-5280-01, GE healthcare) for 1 hour at 4°C, followed by incubation with an anti-Ey antibody (kindly provided by Dr. Uwe Walldorf) at 4°C overnight with rotation. The Ey bound chromatin was pulled down by protein A sepharose beads, washed and eluted with TE/SDS (1x TE, 1% SDS). After incubation at 65°C overnight, the reverse-crosslinked samples were treated with Proteinase K for 1 hour at 45°C and purified with QIAquick PCR purification kit (#28106, Qiagen). PCR primers used were as follows - Pax6^I: ato-B-FW (5′-AAATAATTCGCACGGCCAAC-3′) and ato-B-RV (5′-TGCAGCG-TGAGATACTGAGA-3′); Pax6^II: ato-A-FW (5′-ATTTGTCCAGGTCTGCGTCT-3′) and ato-A-RV (5′-AGTCTCCGAAGATTCCATGC-3′); negative control region: control-FW (5′-GAACTGACCGCATTTGTTGA-3′) and control-RV (5′-TTTTGTTGTGCCTGAGATGG-3′).

Results

The 3′ENH^I enhancer maps to a 183 bp region ∼3.4 kb downstream of the ato transcription unit

We began our search for the putative posterior-progenitor enhancer, 3′ENH^I, by comparing genomic fragments contained in three reporters. These included two constructs from our previous work called M” and 1.2 (Zhang et al., 2006), and the construct 3.6BP from Dr. Du's lab (Tanaka-Matakatsu and Du, 2008). As summarized in Fig. 1A and B, these three reporters used fragments of DNA from the ato genomic region just 3′ downstream of the transcription unit (Fig. 1A), and expressed in both posterior and anterior progenitors (M” and 1.2) or exclusively in posterior progenitors (3.6BP) (Fig. 1B). This comparison identifies a 288 bp sequence as the only region present in all three constructs (Figs. 1B; S2B). We therefore considered that this region was likely to harbor all or part of the 3′ENH^I enhancer.

Interestingly, the 288 bp sequence contained two DNA sequences highly conserved among Drosophilids, one in its entirety, IC1, and the other partially, IC2 (Zhang et al., 2006) (Figs. 1B; S2B). We investigated, therefore, whether these conserved sequences harbored sites required for expression in posterior progenitors. We first tested a reporter with a deletion spanning both highly conserved regions (1.2-Δ281) and found it to lack expression in posterior progenitors (Fig. 1C), confirming that the conserved sequences likely contained 3′ENH^I. We then tested two additional reporters with smaller deletions of the 281 bp interval: 1.2-Δ78 and M”-Δ183 (Figs. 1C; S2B). The 1.2-Δ78 reporter was normally expressed, whereas M”-Δ183 was not (Fig. 1C). Importantly, lack of expression from the 1.2-Δ281 and M”- Δ183 transgenes did not reflect the loss of a universally required element, insertion into generally inactive chromatin, or damage to the reporter protein coding sequence, because multiple transgenic lines carrying these constructs (independent insertions) consistently lacked expression in the posterior progenitors but showed robust expression elsewhere, e.g. in anterior progenitors within the eye disc as well as in other tissues (Fig. S3A; not shown).

Thus, these findings narrow down the region containing the putative posterior-progenitor's 3′ENH^I enhancer to a 183 bp DNA segment located ∼3.2 kb downstream of the ato transcription unit and ∼0.6 kb upstream of 3′ENH^II (Fig. 1D).

3′ENH^I requires 3′ENH^II to drive expression in posterior progenitors

Since Tanaka-Matakatsu and Du (2008) constructed all their reporters (including 3.6BP) by cloning ato genomic fragments directly upstream of the basal hsp70 promoter (followed by the reporter-protein encoding sequence), we next sought to confirm the ability of this enhancer to independently drive expression through the ato 5′ regulatory region. Therefore, we analyzed, expression of reporter constructs in which DNA fragments with the 3′ENH^I enhancer controlled expression through an ato genomic fragment containing the ato promoter, 5′ UTR and some of the 5′-upstream DNA. This fragment called ‘1.1 kb’ (Fig. S1) was originally identified by Sun and colleagues as sufficient to mediate start of transcription and to respond to various modular enhancers present upstream and downstream of the ato transcription unit, but insufficient to drive expression in the eye epithelium or any other imaginal disc on its own (Sun et al., 1998; Zhang et al., 2006).

Thus, we analyzed three constructs that retained 3′ENH^I but lacked 3′ENH^II (Fig.2A). Two were based on our M” fragment: 1) M”-Δ300, consisting of an M” fragment deleted in and around 3′ENH^II and 2) M”-Δ63, consisting of an M” fragment lacking 63 bp critical for 3′ENH^II function (Figs. 2A; S2A) (Zhang et al., 2006). The third construct reproduced the 3.6BP reporter of Tanaka-Matakatsu and Du (2008) except for utilizing the 1.1kb (3.6BP ^atoP), as in the other two M”-based constructs, instead of the heterologous hsp70 promoter (Figs. 2A; S2A).

Fig. 2. The 3′ENH^I enhancer cannot drive expression in posterior progenitors without 3′ENH^II.

(A) Diagram of genomic fragment analyzed; all three modified constructs contain the 3′ENH^I enhancer, but lack a functional 3′ENH^II. (B) Left-most panel: the M” construct drives strong expression in posterior progenitors, right along the disc margin and some distance away. Central two panels: M”-Δ300 and M”-Δ63 cannot drive expression in posterior progenitors. Right-most panel: the 3.6BP fragment linked to the ato promoter (3.6BP ^atoP) also cannot drive expression in posterior progenitors. In addition, none of the modified constructs express in anterior progenitors, though all express in other tissues (Fig. S3B-C; and not shown). (C) Lack of expression is also observed in posterior cells away from the margin. Arrowhead marks MF region where reporter gene expression is expected. The M”-Δ300 and M”-Δ63 reporters show a weak signal at the posterior central margin of the discs. This ectopic expression begins well after shut down of ato transcription in cells that are differentiating. Images for M”-Δ300 and 3.6BP ^atoP show results of β-galactosidase stainings, M” and M”-Δ63 show in situ hybridizations to reporters' mRNA (eGFP for M” and lacZ for M”-Δ63).

Based on the previous result with 3.6BP (Tanaka-Matakatsu and Du, 2008), we expected all three reporters to be expressed in posterior progenitors thanks to the presence of 3′ENH^I, but not in anterior progenitors due to the absence of 3′ENH^II. Surprisingly, we found that none of these constructs, not even 3.6BP ^atoP, could drive reporter gene expression anywhere in the eye disc, including the posterior region (Figs. 2B; S3B). As in the previous experiments, lack of expression was observed in multiple independent transgenic lines that nonetheless expressed robustly in other tissues, e.g. eight independent 3.6BP ^atop lines confirmed lack of expression in the eye disc but expression was easily detected in the antenna and other issues (Fig. S3C, and not shown). Thus, the 3′ENH^I enhancer can induce transcription from the hsp70 basal promoter, but is insufficient to initiate transcription from endogenous ato DNA.

These results (Fig. 2B) together with the previous data (Fig. 1C), establish that 3′ENH^I is required but not sufficient for expression in posterior progenitors. Moreover, since the M” fragment lacking only 63 bp of the 3′ENH^II could not drive expression of the reporter, our findings also implicate the latter enhancer in the onset of ato transcription. In short, 3′ENH^II is also required but not sufficient for expression in cells near the posterior margin of the eye disc. We conclude, therefore, that the 3′ENH^II and 3′ENH^I enhancers work together to induce ato expression in posterior progenitors.

Adjacent So-Pax6 binding sites in 3′ENH^II synergize to induce expression in posterior progenitors

As reported previously (Zhang et al., 2006), two adjacent So and Pax6 sites within the 3′ENH^II (So^EII and Pax6^EII) play a critical role in the onset of ato expression in anterior progenitors. Since 3′ENH^II is also required in posterior progenitors (Fig. 2B), we sought to establish whether these binding sites also functioned in the posterior region of the eye disc.

For this reason, we made specific, small deletions of the Pax6^EII and the So^EII sites singly or in combination within the M” DNA fragment (Fig. 3A) and assessed reporter expression in posterior progenitors. In very early L3 discs, just prior to the emergence of neurons, endogenous ato gene expression can be seen right along the posterior margin, whereas, in slightly older discs, as the first rows of developing neuronal clusters form, ato mRNA is found in a stripe just ahead of the emerging clusters (Fig. 1A, top left panels). In these and later experiments, we relied on the expression of the Senseless (Sens) or Elav proteins in emerging neuronal clusters to identify discs at these early stages (see M” in Figs. 3B-C, 4C). The gene sens is a direct, early transcriptional target of Ato in the developing R8 neuron, whereas the gene elav is expressed in all emerging neurons. Both genes are induced early during eye neurogenesis and persist thereafter (Frankfort et al., 2001; Nolo et al., 2000; Robinow and White, 1991; Zheng et al., 1995). As our analysis of all reporters is carried out in a wild type genetic background, the expression of these factors is normal and serves here and below to stage discs.

Fig. 3. So^EII and Pax6^EII binding sites are required in posterior progenitors.

(A) Diagram shows So^EII and/or Pax6^EII binding site deletions used in B and C. (B-C) Panels show reporter protein expression (green) in posterior progenitors. Discs were also stained for the proteins Eya (blue), to mark all retina cells, and Sens (red), to identify discs in the early stages of neurogenesis (i.e. with few rows of Sens-positive R8 photoreceptors/precursors). Discs are essentially wild type thanks to the normal expression and function of the endogenous ato gene. (B) Left-most panel: M” shows strong expression in cells along the entire posterior margin of the disc. Central two panels: single-site deletions for Pax6^EII or So^EII show a dramatic reduction in reporter protein. Right-most panel: deletion of both binding sites is indistinguishable from deletion of either site alone. (C) Slightly older discs than in B show consistent results in posterior progenitors away from the margin. Top row shows staining for reporter protein (green) as well as Eya (blue) and Sens (red). Bottom row shows Eya (blue) and Sens (red) only.

The deletion of either site (M”-ΔPax6^II and M”-ΔSo^EII constructs) resulted in a severe reduction in reporter protein expression along the posterior margin; weak expression could be detected only near the disc midline (Fig. 3B). As the first few rows of neuronal clusters formed, weak expression of the reporters was also observed in cells just ahead of the forming neuronal clusters (Fig. 3C). Residual expression was indistinguishable in the two types of constructs, whether deleting one or the other site.

Deletion of both sites (M”-ΔSo^EIIPax6^EII) did not appear to result in more severe effects than deletion of either site alone, suggesting that once one site is lost, the other site is essentially non-functional (Fig. 3C). As shown in Fig. 2B-C, deletion of the 3′ENH^II (as in M”-Δ300) or 63 bp centered around the binding sites (M”- Δ63) resulted in complete loss of expression, indicating that the residual expression seen in M”-ΔSo^EIIPax6^EII still reflects some activity of the 3′ENH^II enhancer. The deletion in M”-Δ63 extends only 21 bp to the left of So^EII site and 10 bp to the right of the Pax6^EII site, and within these short stretches of DNA lie sequences related to homeodomain binding sites (M”-Δ21+10; Fig. S2). However, the activity of any of these sites appeared to be mainly supportive because reporters in which these sequences were specifically deleted (leaving the Ey and So sites unaffected) show normal expression (Fig. S2; not shown). In short, loss of either RD-factor binding site is equivalent to loss of both sites, but the 3′ENH^II still appears to function at a very low level through one or more site in the immediately surrounding DNA.

Based on these findings, we conclude that the So^EII and Pax6^EII sites play a major role in the induction of ato transcription within posterior retinal progenitors. Most importantly, since transgenes with both sites drive expression at far higher than additive levels (i.e. much higher than the sum of the constructs bearing either site), the So^EII and Pax6^EII sites function synergistically in posterior progenitors.

3′ENH^I contains a Pax6 site critical for expression in posterior progenitors

Since both enhancers control expression in posterior progenitors and 3′ENH^II does so, at least in part, through RD factors binding sites, we decided to investigate whether Ey and/or So might also act through the 3′ENH^I enhancer.

A search for potential binding sites within a 604 bp region containing the 3′ENH^I revealed the presence of two consensus Pax6 (paired-box) binding sites but no So (homeobox) site. To identify Pax6 sites that could mediate strong binding, we relied on electrophoretic mobility shift assays (EMSA) using in vitro transcribed/translated Ey protein. EMSA analysis of six overlapping DNA fragments spanning the 604 bp genomic region showed robust binding of Ey to fragment III (Fig. S4). Interestingly, this 120 bp fragment mapped largely within the 183 bp interval containing 3′ENH^I and partially overlapped both conserved sequences, IC1 and IC2 (Figs. 4A; S2B; S4). Binding was indeed specific since competition by unlabeled DNA depended on presence of the wt Pax6 binding sequence (Fig. 4A, legend). We therefore named this site Pax6^EI and proceeded to investigate its role in the regulation of ato expression using single and double mutant constructs (Fig. 4B).

Deletion of the Pax6^EI sequence in M”-ΔPax6^EI led to a severe reduction in reporter gene expression in early L3 discs (Fig. 4C). In slightly older discs, expression could be detected, albeit severely reduced, in progenitor cells just anterior to the first few rows of developing neuronal clusters (Fig. 4D). Thus, loss of the Pax6^EI site appeared to have a similar effect as loss of the Pax6^EII site, affecting expression in all posterior progenitors and particularly strongly in progenitors along the margin.

Interestingly, loss of Pax6^EI along with either Pax6^EII or So^EII (M”-ΔPax6^EIPax6^EII and M”-ΔPax6^EISo^EII) resulted in the complete loss of reporter expression (Fig. 4C-F). This finding is in contrast to the presence of residual expression along the posterior margin in M”-ΔSo^EIIPax6^EII double mutant constructs (which is dependent on nearby sequences as mentioned above; Fig. 3C), but is consistent with the disruption of a necessary interaction between the two enhancers as suggested by our earlier findings (Figs. 1C; 2B).

Notably, the presence of all three sites in the wt control construct contributes a far higher level of reporter expression than predicted from an additive contribution of each site alone. This is consistent with the synergistic effect seen with 3′ENH^I and 3′ENH^II enhancers, and thus suggests a role for the Pax6^EI, Pax6^EII and So^EII binding sites in mediating enhancer interactions in posterior progenitors

The Pax6^EI site from 3′ENH^I is active in anterior progenitors

Given the requirement for both enhancers in posterior progenitors, we decided to investigate whether both enhancers could also be active in anterior progenitors. We had previously shown an absolute requirement for 3′ENH^II, and its Pax6^EII and So^EII binding sites, in anterior expression (M”- Δ300 and M”– Δ63 in Fig. S3B) (Zhang et al., 2006). On the contrary, substantial expression was detected in anterior cells in absence of the 3′ENH^I region (e.g. in 1.2-Δ281 or M”– Δ183; Fig. S3A). These previous studies, however, did not investigate the possibility of a partial or redundant contribution of the 3′ENH^I enhancer. To address this possibility, we decided to carry out a more detailed, quantitative analysis of reporters lacking the Pax6^EI, So^EII or Pax6^EII using both P-element-based and site-directed transgenes in order to detect partial effects.

As expected, deletion of either the So^EII or the Pax6^EII site (M”-ΔSo^EII or M”-ΔPax6^EII) reduced reporter gene expression substantially in anterior progenitors (Fig. 5A, A'). Surprisingly, loss of the Pax6^EII (M”-ΔPax6^EII) led to de-repression of gene expression in the posterior region of the late L3 discs, where differentiation of ommatidial cluster is well along the way (Fig. 5A). This observation suggests an additional and converse role for this site in repression at the later stage. More importantly, deletion of the 3′ENH^I Pax6^EI site (M”-ΔPax6^EI) also led to a significant loss of reporter gene expression (Fig. 5A-A').

Fig. 5. Interactions between Pax6 and So sites in anterior progenitors.

(A-B') Effect of single-site deletions on reporter gene expression assessed by in situ hybridizations to reporter mRNA (eGFP) in L3 eye discs (2hr staining reaction). (A-A') Representative discs (A) and graph of relative expression levels (A') for P-element insertion lines. (B-B') Representative discs (B) and graph of relative expression levels (B') for site-specific insertion lines. In both A and B, deletion of any one site (Pax6^EI, Pax6^EII or So^EII) reduces reporter gene expression significantly (p<0.01; see Fig. S5); deletion of the So^EII site consistently shows a more severe effect than deletion of either Pax6 site (p<0.05; see Fig. S5). (C-D) Effect of double-site deletions on reporter gene expression. (C-C') Expression level in P-element insertion lines was assessed by in situ hybridizations to reporter mRNA (eGFP) in L3 discs (2 hr staining reaction). When no expression was detected within 2 hr, the staining reaction was carried for 24 hr (shown in insets and panels as marked) in order to detect weak expression. Panels show representative discs (C) and graph of relative expression levels (C'). Deletion of the Pax6^EI and So^EII resulted in complete loss of expression, whereas loss of both Pax6 sites had the weakest effect (p<0.01; see Fig. S5). (D) Expression in site-specific insertion lines was assessed at the protein level (eGFP) in L3 eye discs. Discs were also stained for the pan-neural marker Elav (red) to show normal neurogenesis (due to the normal expression and function of the endogenous ato gene). As in C, deletion of Pax6^EI and So^EII sites leads to complete loss of expression and deletion of the two Pax6 sites has a weaker effect than So^EII and Pax6^EII deletions. Note that constructs with the deletion of the Pax6^EII site (including M”-ΔPax6^EII in A-B, and M”-ΔPax6^EIPax6^EII and M”-ΔPax6^EIISo^EII in C-D) show de-repression of reporter expression posterior of the MF, in a region of the disc where neurons are differentiating and Ey is not expressed suggesting that an unknown negative regulator may function through the Pax6^EII site in this region of the disc.

Transcript abundance was lower in M”-ΔSo^EII, than in M”-ΔPax6^EI or in M”-ΔPax6^EII discs (Fig. 5A' and B'; see Fig. S5A-B for p values from Student's t-test; see Materials and Methods for measurements of Relative Optical Density). Smaller, but significant differences in expression levels (e.g. between M”-ΔPax6^EI and in M”-ΔPax6^EII) were detected more readily with the site-specific transgenes than with the P-element mediated insertions (Fig. S5A-B). Nonetheless, the relative order of expression levels (M”-ΔPax6^EII > M”-ΔPax6^EI > M”-ΔSo^EII) was the same in both site-directed (PhiC31) and random (P-element) insertion lines (Fig. 5A'-B', S5A-B). Loss of the So binding site consistently had a stronger effect than deletion of either Pax6 site, suggesting a predominant role for the RD factor So. Regardless, the Pax6^EI site made a substantial contribution to gene expression in anterior progenitors.

In conclusion, our findings show that the activity of the 3′ENH^I and its Pax6^EI site is not restricted to posterior cells but contributes to anterior-progenitor expression. In short, both enhancers, 3′ENH^I and 3′ENH^II, regulate onset of ato expression in anterior cells as well.

Cooperation of Pax6^EI, Pax6^EII and So^EII sites in anterior progenitors

We then further investigated the contribution of all three sites to anterior expression. Our analysis of double deletion constructs revealed complex interactions among these sites. Deletion of any two sites (M”-ΔPax6^EIPax6^EII; M”-ΔSo^EIIPax6^EII; M”-ΔPax6^EISo^EII) led to a dramatic decrease in gene expression regardless of trasgenesis or detection methods (in both P-element-based and site-specific reporters and as assessed by in situ hybridization or immunostaining). In this analysis, we benefited from testing constructs of both types as explained below.

The P-element based reporters gave consistent results across lines carrying the same construct. However, whereas expression was generally stronger than for site-directed lines, independent insertions of the same construct showed greater variability in expression level (Figs. 5C; S5C). Weak expression was detected in discs from the M”-ΔPax6^EIPax6^EII transgenic lines after 2 hr of staining (Fig 5C), but only a faint signal was visible in similarly stained discs from the M”-ΔSo^EIIPax6^EII lines (Fig 5C). After 24 hr incubation of the discs with the chromogenic substrate, both double mutant constructs clearly showed staining in a stripe, confirming that the weak signals seen at 2 hr did reflect very low expression (Fig 5C insets) (as well as ectopic expression in posterior cells undergoing differentiation; an observation consistent with the findings described above for M”-ΔPax6^EII; Fig. 5A). In contrast, no expression at all was detected in M”-ΔPax6^EISo^EII discs after 2 hr or 24 hr of chromogenic reaction (Figs. 5C). As shown in Fig. S7, the latter construct showed expression in the antenna (as well as other tissues; not shown), confirming that the construct is not fundamentally defective. This relative order of expression (M”-ΔPax6^EIPax6^EII > M”-ΔSo^EIIPax6^EII > ΔPax6^EISo^EII) was also reflected at the level of the reporter protein (Fig. S6).

In site-specific lines, the analysis of double mutant constructs was complicated by the very low level of reporter gene expression (expression was lower for all site-directed constructs as compared to P-element lines, including the wt M” construct). Thus, a quantitative analysis at the transcript levels was not possible. Nonetheless, reporter gene expression could be detected at the protein level and showed the same overall pattern (relative level and ectopic activation) as the P-element inserted lines (Figs. 5D; S6). Expression in M”-ΔPax6^EIPax6^EII was higher than in M”-ΔSo^EIIPax6^EII; whereas, no expression at all could be detected in M”-ΔPax6^EISo^EII discs (Fig. 5D).

In conclusion, our analyses show that, in anterior progenitors, any two sites are sufficient to mediate gene activation at a reduced but still robust level, whereas each of the sites alone induces little to no expression. The more severe effect of deletions of the So^EII site in combination with Pax6^EII or Pax6^EI, is consistent with the more critical role to the So uncovered in single mutant constructs and with an interaction between the 3′ENH^I and 3′ENH^II enhancers that is critical for expression in anterior progenitors.

The 3′ENH^I and 3′ENH^II regions are sufficient for ato-like expression and are bound by the RD factors Ey and So in vivo

We then asked whether the 3′ENH^I and 3′ENH^II regulatory regions were in fact sufficient to recapitulate ato expression in all retinal progenitor cells. To this end, we generated a construct containing the 183 bp and 338 bp (Zhang et al., 2006) DNA fragments linked to the ato promoter (E^I+E^II construct). The E^I+E^II transgene can drive expression in both posterior and anterior progenitors very robustly (Fig. 6A), in a pattern reminiscent of constructs containing larger fragments (e.g. M” and 1.2; Figs. 1 & 2) (Sun et al., 1998; Zhang et al., 2006). Thus, 3′ENH^I and 3′ENH^II together are sufficient to reproduce the transient expression of ato throughout the eye field and are likely the main enhancers controlling the onset of ato transcription in eye progenitors at the start of neurogenesis.

Fig. 6. 3′ENH^I and 3′ENH^II are sufficient to drive expression in the ato pattern and the Ey and So binding sites are occupied in vivo.

(A) The 3′ENH^I and 3′ENH^II enhancers together (E^I+E^II) can drive expression in both posterior and anterior progenitors recapitulating the ato-independent phase of ato transcription throughout the eye field. (B) In 30A-Gal4 UAS-ey wing discs, ectopic eye formation is in progress where strong expression of the RD factor Eya (marking retinal progenitors and developing cells) and the pan-neural marker Elav (marking sites of neurogenesis) is present (arrowheads). Expression of the M” reporter (top panels), but not the M”-ΔPax6^EISo^EIIPax6^EII reporter (bottom panels) is detected at Eya-/Elav-positive sites. (C) ChIP experiments using anti-Ey and anti-So antibodies to detect binding of endogenous proteins. All three sites are occupied by the corresponding transcription factors in the L3 eye disc.

The regulation of ato through these enhancers by the RD factors Ey and So is also confirmed by ectopic induction experiments. Ectopic eye formation is induced at two locations within the wing epithelium by targeted expression of Ey (in 30A-Gal4 + UAS-ey wing discs) (Fig. 6B). These locations show strong ectopic induction of multiple RD factors including So in response to the exogenously provided Ey protein (Chen et al., 1999). In 30A-Gal4 UAS-ey discs, expression of the M” reporter is induced at these ectopic sites, whereas expression of M”-ΔPax6^EISo^EIIPax6^EII is not (Fig. 6B). Thus, not only normal eye disc expression but also ectopic expression is dependent on the Pax6 and So sites present in 3′ENH^I and 3′ENH^II.

Lastly, we sought to confirm that Ey and So bind in vivo to these sites within the 3′ regulatory region of the endogenous ato gene. To establish whether this was indeed the case, we set out to carry out chromatin immunoprecipitation (ChIP) experiments using the anti-Ey and anti-So antibodies and PCR primers specific to each site. Using these reagents, we were able to show that all three sites are occupied in wild type L3 discs at a time when eye progenitors are undergoing neurogenesis (Fig. 6C).

Thus, we believe that the requirement for Pax6^EI, So^EII and Pax6^EII uncovered through this extensive analysis reflects the activity of the Ey and So RD factors in regulating ato expression at these sites during eye neurogenesis.

Discussion

The activation of ato expression in retina progenitor cells represents a critical step in the transition from progenitor cell to developing photoreceptor neuron. We show here 1) that onset of ato transcription depends on two distant enhancer regions that interact differently in subset of retinal progenitor cells; 2) that these interactions are mediated largely by three specific Pax6/Ey and So binding sites; 3) that ato expression is induced by So not only in anterior progenitors but posterior ones as well; and, 4) that all three sites are in fact occupied in vivo. This analysis leads to a more complex model for ato regulation at the start of eye morphogenesis and provides a first example of how multiple enhancer elements synergize in the regulation of ato transcription.

In 1998, Sun et al. showed that the cis regulatory elements controlling onset of ato transcription in eye progenitor cells were located within a ∼6 kb fragment of 3′-flanking genomic DNA. Surprisingly, further analysis led to the identification of two distinct regions that functioned independently in different subsets of eye progenitor cells (Tanaka-Matakatsu and Du, 2008; Zhang et al., 2006). Thus, the eye field appeared to contain two types of progenitors (posterior and anterior) with independent mechanisms for the induction of ato expression (Fig. 1A).

In this work, we have shown that the transgenic vectors used in these analyses influenced the results and led to this early model. In particular, the 3′ENH^I was originally found to drive express in posterior but not anterior progenitors when using the hsp70 promoter (Tanaka-Matakatsu and Du, 2008). However, when linked to the endogenous ato control region, the 3′ENH^I enhancer alone fails to drive expression in any eye progenitor cells, but it does play an essential role in posterior progenitors when coupled with the 3′ENH^II enhancer. Differential regulation of distinct promoter regions by enhancers and their transacting factors has been documented in multiple cases and shown to depend on the combination and configuration of core promoter elements (Butler and Kadonaga, 2001; Juven-Gershon et al., 2008; Kapoun and Kaufman, 1995; Kwon et al., 2009; Ohler and Wassarman, 2010). A detailed analysis of the region around the ato transcriptional start site may shed some light on this effect. Although use of the gene promoter may more faithfully reproduce endogenous regulation, use of heterologous promoters can still be informative. As in the case of the 3′ENH^I of ato, a versatile promoter such as hsp70 can help unmask the activity of some regulatory sites and thus facilitate their detection (Tanaka-Matakatsu and Du, 2008).

In addition to driving expression in posterior progenitors (together with the 3′ENH^II enhancer), the 3′ENH^I also contributes to 3′ENH^II-induced expression in anterior progenitors. Thus, neither regulatory region is specific to a subset of retinal cells. On the contrary, the two work together to drive expression in all progenitors. However, the activity of and interaction between these enhancers is ‘progenitor population specific.’ These effects are mediated by two Pax6 and one So binding site. The Pax^EI site within the 3′ENH^I region is located ∼600 bp away from the adjacent Pax^EII and So^EII sites of the 3′ENH^II enhancer (Figs. 1; 4B). In both progenitor populations, the combined activity of the binding sites contributes to gene expression, but their contribution and specific mode of interaction differ.

Differential regulation of ato transcription in posterior and anterior progenitors: the cis-acting elements

In posterior retinal progenitors, DNA fragments lacking either enhancer cannot drive reporter gene expression (Figs. 1; 2), whereas the combination of the 183 bp and 338 bp DNA fragments, containing 3′ENH^I and 3′ENH^II respectively, can do so robustly (Fig. 6A). Therefore, both enhancers are required but neither is sufficient for expression in the posterior region of the disc. Interestingly, the three RD factors binding sites present in these sequences show a similar pattern, with all three sites being critically important; loss of any one site results in a nearly complete loss of reporter gene expression (Figs. 3; 4). Deletion of the Pax^EII site or the So^EII site of 3′ENH^II results in very low residual expression in the posterior-central portion of the disc. Surprisingly, this expression persists even when both sites are deleted. This result suggests that, in posterior progenitors, loss of either site is equivalent to loss of both. Two simple models that would account for this effect include 1) the cooperative binding of Ey and So to the closely juxtaposed Pax^EII So^EII sites (Fig. 3), and 2) the requirement for the transcription factors together to promote the association of a third factor; but other models are also possible. Interestingly, this residual expression, is completely abolished when the Pax^EI site of 3′ENH^I is deleted along with either one of the 3′ENH^II sites, Pax^EII or So^EII. We interpret this to indicate that the damage inflicted upon 3′ENH^I and 3′ENH^II by these particular double deletions result in a profound disruption of the crosstalk between the two enhancers. On the contrary, in Pax^EII So^EII double mutants very weak expression is induced by the presence of an intact 3′ENH^I enhancer together with some residual activity of the 3′ENH^II (due to sequences in the DNA flanking the So^EII-Pax^EII sites, as discussed in the results; see M”-Δ63 in Fig. 2 and M”-Δ21+10; Fig. S2).

In anterior retinal progenitors, the 3′ENH^II enhancer is both required and sufficient for reporter gene expression, whereas the 3′ENH^I contributes to full expression but is not sufficient on its own. Once again, the contribution of the three RD factors binding sites reflects this modified relationship between 3′ENH^I and 3′ENH^II. Here, all three sites are still important for normal expression, but loss of any one site has a relatively modest effect. For instance, the loss of Pax6^EI, or even of the entire 3′ENH^I region, does not lead to a lack of expression, but to a relatively modest reduction (down to ∼50% of M” control; Figs. 1C; S5A). Similarly, the lowest expression (observed in site-directed constructs bearing both Pax6 sites but no So^EII site) still exceeds 30% of the M” control (Figs. 1C; S5A). This also implies that, unlike in posterior progenitors, the So^EII and Pax^EII sites are less dependent on one another for their activity in anterior cells. This effect suggests some redundancy within the 3′ENH^II element that preserves enhancer function in absence of one or the other site. Nonetheless, each site makes a significant, cooperative contribution to ato transcriptional activation in anterior progenitors because the expression level of the M” control or reporter constructs bearing 2 normal sites (single-site deletions) far exceed the predicted additive contributions from each single site (Fig. S5).

In summary, our findings strongly suggest that, in posterior progenitors, strong synergistic interactions between the two enhancers trigger the onset of ato expression, and that the Pax6/So binding sites are significant mediators of this crosstalk (Fig. 7B). In anterior progenitors, instead, collaborative interactions between pairs of sites achieve substantial expression, although all three sites are required for full expression levels (Fig. 7C). Since the Ey and So proteins have been shown to directly interact in vitro (Zhang et al., 2006), some of these effects may involve direct protein-protein contacts between these RD factors while bound to the DNA.

Fig. 7. Diagram of different modes of ato gene activation by 3′ENH^I and 3′ENH^II and their RD factors binding sites in different regions of the developing eye disc epithelium.

(A) Drawing of the disc with anterior (green) and posterior (magenta) progenitor domains. Diagrams of the 1.2 reporter fragment [with 3′ENH^I (light gray) and 3′ENH^II (dark gray)] as well as related modified constructs (3′ENH^II alone, 3′ENH^I alone and 3′ENH^II+3′ENH^I) are shown on each side of the disc; reporter expression results are reported (YES or NO) in the relevant progenitor domain. Pax6 and So binding sites are marked by triangles (Pax6) and circles (So). (B, C) The solid black arrows represent the synergistic effects of the binding sites on onset of transcription from the ato promoter region. The dashed bidirectional arrows indicate the interactions between enhancers and/or pairs of binding sites that are necessary and sufficient for robust expression. (B) In posterior progenitors, the 3′ENH^I and 3′ENH^II together promote transcription from the ato promoter region. None of the sites alone can induce significant expression through the ato promoter, nor can any combination of two sites. Thus, the presence of both Pax6^EII and So^EII is required for 3′ENH^II function, and cooperation between the two enhancers 3′ENH^I and 3′ENH^II (dashed double-arrow) is necessary to initiate ato expression in posterior progenitors. (C) In anterior progenitors, interactions between any two sites (dashed double-arrows) are sufficient to induce robust expression in reporter constructs, although all three sites are required for highest expression. Hence, consistent with previous enhancer mapping experiments (Zhang et al., 2006), the 3′ENH^II alone, bearing two of the sites, can drive expression in anterior progenitors. On the contrary, the 3′ENH^I alone, bearing only the Pax6 ^EI site, cannot drive expression anywhere in the eye disc through the ato promoter region, but is able to do so when at least one of the two sites in the 3′ENH^II enhancer is present (this work). This effect suggests some redundancy within the 3′ENH^II element that may render the mutation of either Pax6^EII or possibly So^EII DNA tolerable for endogenous gene function in anterior progenitors.

Differential regulation of ato transcription in posterior and anterior progenitors: the trans-acting factors

As discussed above, our findings support a model whereby Ey and So bind to specific DNA sequences in the 3′ENH^I and 3′ENH^II regulatory elements and interact (directly or through other factors) to synergistically promote ato activation in all retinal progenitor cells. However, the nature of these interactions differs in posterior versus anterior progenitors. Differences in the genetic control of eye morphogenesis between cells along the disc margin and away from it have been previously reported (Curtiss and Mlodzik, 2000; Domínguez, 1999; Firth and Baker, 2009; Fu and Baker, 2003; Greenwood and Struhl, 1999; Hazelett et al., 1998; Pappu et al., 2005) and may reflect local differences in signaling systems or transcription factors expression.

Genetic evidence shows that several other regulators promote Ato expression, including the RD factors Eyes absent (Eya), Teashirt (Tsh) and Dachsund (Dac), the transcription factor Lilliputian (Lilli), the chromatin modifiers Trithorax and Kismet, and the Mediator complex subunits Kohtalo and Skuld. Unlike Ey and So, however, none of these factors has been shown to bind ato regulatory DNA. The chromatin modifiers are thought to affect ato transcription indirectly through their negative regulation of the retinal antagonist homothorax and their positive regulation of the RD genes ey, tsh and eya (Janody et al., 2004; Lim et al., 2007; Melicharek et al., 2008). Lilli may promote ato expression in the late L3 disc through cis elements within the 3′ ato enhancers (although direct binding has not been shown) (Distefano et al., 2012). However, its ubiquitous expression does not support a role for Lilli in determining the difference between posterior versus anterior progenitors (Distefano et al., 2012). Among the RD factors, Eya is expressed and partners with So (Pignoni et al., 1997); thus, the So-Eya complex most likely functions to regulate ato in all retinal progenitors (Zhang et al., 2006). On the contrary, Tsh and Dac are of particular interest because the activity of these two factors differs in posterior versus anterior eye field cells. Dac is absolutely required for neurogenesis in the posterior but not in the anterior region of the disc, as shown by clonal loss-of-function analysis (Mardon et al., 1994). Tsh promotes eye formation in the anterior region but not in posterior progenitors, because it is not expression in cells at or near the posterior margin of the disc epithelium (Pan and Rubin, 1998). Between the two, it is the Zn-finger retina-promoting factor Tsh that best fits our findings. The presence of Tsh in the anterior region could account for the less stringent requirements for ato expression in these cells.

Alternatively, the influence of a retinal antagonist on posterior but not anterior progenitors could account for the effects described in this work. Negative regulation of ato expression across the differentiating ommaditial array is mediated by the bHLH factor Daughterless (Da) and the homeobox proteins BarH1 and BarH2. However, in both cases, loss of function leads to ectopic Ato expression posterior to the MF in late L3 discs (Melicharek et al., 2008; Lim and Choi, 2003). Thus, negative regulation of ato by the Da or Bar proteins is not restricted to posterior progenitors but extends throughout the epithelium. The one candidate that fits our profile (negative regulation near the posterior disc margin at the time when neurogenesis begins in this region) is the WNT pathway because 1) the WNT-type ligand Wingless (Wg) is indeed expressed along the margins of the eye disc and 2) the Wg pathway can and does antagonize eye formation in the disc (Cho et al., 2000; Hazelett et al., 1998).

We therefore tested possible roles for either Tsh or the Wg pathway in modifying ato regulation in different retinal progenitor types. We did so by manipulating the expression/activity of these two candidates in vivo and then assessing ato reporter gene expression (see legend of Fig. S8 for details). In these experiments, the expression of our reporter (388) did not change significantly (Fig. S8), suggesting that neither Tsh nor the Wg pathway mediate the distinction between ato regulation in posterior versus anterior progenitors. Hence, additional unidentified factors must be involved in this process. Several putative binding sites (based on sequence conservation) are present in the 3′ENH^I and 3′ENH^II enhancers and are essential for expression (Zhou and Pignoni, unpublished). Further analysis of these sequences and the identification of factors that bind to them may provide useful insights on this issue.

Regulation of other target genes by Ey and/or So in the Drosophila retina

Several other direct targets of Ey and/or So in the eye have been identified in recent years. Ey is thought to directly regulate the expression of eya, so, optix, and dac; So controls the expression of ey, dac, hh and lozenge; and both genes regulate their own expression (Niimi et al., 1999; Ostrin et al., 2006; Punzo et al, 2002; Pauli et al., 2005; Yan et al., 2003). In all cases, 2 or more binding sites for Ey and/or So have been identified. However, only in the case of the target genes dac and so, any evidence of interactions between regulatory sites has been reported. In the case of dac, synergy was observed between a pair of So binding sites that lie just 13 bp apart within the 3′ regulatory region of the gene (Pappu et al., 2005). In the case of so, synergy was observed between several Pax6 binding sites located in an intron and contained within a ∼320 bp interval (Punzo et al., 2002). In all other cases, the authors did not investigate the sites' contribution to gene expression other than establishing that they are required (Hauck et al., 1999; Ostrin et al., 2006; Pauli et al., 2005; Yan et al., 2003). Thus, the regulation of ato described here remains the first example of interactions between multiple, distant Ey and So sites (∼600 bp apart) and of their effect on a promoter (in the 1.1 kb ato fragment) that is normally located more than 5 kb upstream. Nonetheless, the occurrence of multiple binding sites in all these genes suggests that interactions among RD proteins bound to several sites may be a common mechanism in the regulation of RD gene targets.

Pax6 and Six factors in the regulation of vertebrate proneural genes

Over the past 15 years, considerable evidence has been uncovered for the genetic control of proneural factors by Pax6 in vertebrates. For instance, the mouse bHLH genes neurogenin1, neurogenin2, Mash1, Mash2, Math1, and Math5, and the Xath5 locus of Xenopus, all require Pax6 for normal expression in a number of different tissues including retina, spinal cord, and/or cerebral cortex (Blader et al., 2004; Brown et al. 1998; Helms et al.; 2000; Hutcheson et al.; 2005; Hufnagel et al., 2007; Landsberg et al.; 2005; Marquardt et al.; 2001; Nakada et al., 2004; Riesenberg et al., 2009; Scardigli et al., 2001; Scardigli et al., 2003; Toresson et al., 2000; Verma-Kurvari et al., 1998; Willardsen et al., 2009). Moreover, in vitro and in vivo evidence has shown that this regulation is direct for some of these targets.

In the early vertebrate retina, expression of the ato homologues Xath5 and Math5 (required for ganglion cells formation; Brown et al. 1998; Kanekar et al., 1997) depends on binding of Pax6 to a distant 5′ cis-site present in both the Xenopus and mouse genes. In these organisms, as in the fly, Pax6 binding induces Ath5 expression in a bHLH-independent fashion. This is followed in Xenopus, but not in mouse, by a bHLH-dependent expression phase (similar to the Ato-dependent phase of ato gene transcription; see Fig. S1) (Hufnagel et al., 2007; Hutcheson et al., 2005; Riesenberg et al., 2009; Willardsen et al., 2009). In addition, the expression of Mash1 in the developing retina and of neurogenin2 in both retina and CNS also require direct binding of Pax6 to specific enhancer elements (Marquardt et al., 2001; Scardigli et al., 2003; Verma-Kurvari et al., 1998). Thus, a direct regulatory relationship at the transcriptional level is present between Pax6 and several proneural genes in vertebrates.

What about proneural gene regulation by transcription factors of the Six family? Transcriptional regulation of Math5 by the So homologues Six1 and Six2 in the early retina is unlikely, because expression of Six1/2 factors is only detected late in the developing murine eye (Kawakami et al., 1996). Two other Six family members, Six3 and Six6, are expressed and required in the early stages of eye development (Jean et al., 1999; Liu et al., 2010; López-Ríos J, 1999; Oliver et al., 1995; Toy et al., 1998; Zuber et al., 1999). Six3 antagonizes the anti-retina signal Wnt8b and promotes proliferation of retinal progenitor cells, whereas Six6 controls eye size and is later associated with amacrine cells development (Liu et al., 2010; Zuber et al., 1999; reviewed by Sinn and Wittbrodt, 2013). However, no evidence of direct transcriptional regulation of Ath5 genes in the retina has been uncovered for either factor.

Nonetheless, a link between Six1 and other mammalian proneural genes is apparent in another sensory organ. Expression of the bHLH genes Atoh1 and neurogenin1 in the mouse inner ear requires the function of Six1 and its partner Eya1 (Ahmed et al., 2012a; Ahmed et al., 2012b). Moreover, ectopic expression of Six1/Eya1 is sufficient to induce activation of the both gene targets. Atoh1, in particular, is directly regulated by Six1; whether regulation of neurogenin1 is direct or indirect, remains to be determined (Ahmed et al., 2012a; Ahmed et al., 2012b).

In summary, whereas a direct regulatory relationship between Pax6 and proneural genes is well established in both fly and vertebrates (and was likely present in some form in their last common ancestor), evidence of direct regulation of proneural genes by Six factors is currently very limited but also largely unexplored. Considering that Pax, Six and bHLH proteins are present (and often related by co-expression and/or genetic regulation) in several basal metazoans, including sponges (Richards et al., 2008; Rivera et al., 2013), the direct regulatory relationships described here may reflect the long and complex evolution of an ancient gene network.

Highlights.

• Onset of ato depends on two distant enhancers in all retinal progenitors.

• Enhancers interact differently in subsets of progenitor cell populations.

• Two Pax6 and one So binding sites mediate enhancer cross-talk.

• All binding sites are occupied by endogenous Ey and So proteins in vivo.

• We present a novel view of ato induction in the developing eye field.