Shared and distinct mechanisms of atonal regulation in Drosophila ocelli and compound eyes

Abstract

The adult fly has two types of external visual organs, a pair of compound eyes and a group of three ocelli. At the time of neurogenesis, the proneural transcription factor Atonal mediates the transition from progenitor cells to differentiating photoreceptor neurons in both organs. In the developing compound eye, atonal (ato) expression is directly induced by transcriptional regulators that confer retinal identity, the Retinal Determination (RD) factors. Little is known, however, about control of ato transcription in the ocelli. Here we show that a 2 kb genomic DNA fragment contains distinct and common regulatory elements necessary for ato induction in compound eyes and ocelli. The three binding sites that mediate direct regulation by the RD factors Sine oculis and Eyeless in the compound eye are also required in the ocelli. However, in the latter organs, these sites mediate control by Sine oculis and the other Pax6 factor of Drosophila, Twin of eyeless, which can bind the Pax6 sites in vitro. Moreover, the three sites are differentially utilized in the ocelli: all three are similarly essential for atonal induction in the posterior ocelli, but show considerable redundancy in the anterior ocellus. Strikingly, this difference parallels the distinct control of ato transcription in the posterior and anterior progenitors of the developing compound eyes. From a comparative perspective, our findings suggest that the ocelli of arthropods may have originated through spatial partitioning from the dorsal edge of an ancestral compound eye.

Introduction

Two compound eyes and three ocelli comprise the external photosensitive organs of the fruit fly, Drosophila melanogaster. The large compound eyes mediate vision of the surrounding world (including motion, color, and polarized light perception), whereas the much smaller ocelli contribute mainly to flight stabilization through perception of relative light intensity. The three ocelli are located on the dorsal surface of the fly head and are arranged into a posterior pair and a single anterior ocellus (Fig. 1A). Each consists of nearly a hundred photoreceptor neurons surrounded by corneagenous (lens-producing) and pigment cells (Stark et al., 1989; Brockmann et al., 2011). Like the compound eyes, the ocelli develop during the larval and pupal stages from the eye-antennal imaginal disc, the progenitor epithelium that gives rise to most of the fly head. The two posterior, or lateral, ocelli (OC^P) form separately, one from each eye disc; whereas the anterior, or medial, ocellus (OC^A) forms from the fusion of two ocellar ‘halves,’ one from each disc (Fig. 1B).

Figure 1. Ato in the development of eyes and ocelli.

Where needed, orientation is shown by coordinates (A=anterior, P=posterior, D=dorsal, V=ventral). (A) Micrograph of dorsal fly head showing the ocellar complex. (B) Location of developing ocellar primordia (OCP) in the late L3 eye disc. Schematic of the larval eye-antennal imaginal disc with eye and ocellar fields shaded in gray. (C) Evolutionary relationship of eyes and ocelli as derived from a precursor organ devoted to light sensation in a common ancestor of arthropods (Mya= million years ago). (D) Diagram of ato expression pattern during ocellar and retinal development. In both, founder neuronal precursors are selected from the larger pool of Ato-positive proneural cells. Ato expression in the proneural domain is Ato-independent, but primary precursors maintain Ato-expression through feedback regulation. Neurogenesis in surrounding cells is induced through an ato-independent process in the eye. As suggested by the lineage tracing experiment (Fig. S1), in the ocelli, ato-dependent precursors may recruit additional photoreceptor neurons in Ato-independent fashion, similarly to the eye. However, an Ato-dependent recruitment step of the type described for the femoral chordotonal organ cannot be entirely excluded (zur Lage et al., 2004). An accurate understanding of ocellar development awaits a detailed analysis of the neurogenetic process in this organ. (E) Summary diagram of the ato gene and its regulatory regions based on published reports. The single exon of the gene is shown as a thick black rectangle marked ‘ato’ in white, black arrow above shows direction of transcription. The 1.1 kb genomic segment, shown as a thick line, includes the promoter region and extends into the exon to include the short ato 5’UTR. This segment does not drive expression in eyes or ocelli but was used as ‘endogenous promoter’ fragment in all constructs analyzed. Fragments implicated in the regulation of ato expression in eyes and ocelli are shown, numbers to the right refer to relevant references (1=Sun et al. 1998; 2=Zhang et al., 2006; 3=Tanaka-Matakatsu and Du., 2008; 4=Zhou et al., 2014). (F-F’) Confocal image of a late L3 eye disc to highlight relative position of the large eye field and the ocellar primordial (F). Enlarged view of ocellar region is shown in F’. Disc is stained for Eya (blue), which marks both the eye field and the ocellar primordia, and for Sens (red), a direct target of Ato in the eye that is also expressed in the ocellar region very late in L3. Here and in the following figures, the ocellar primordia are marked by dashed lines, orange for the OCP^P and yellow for the OCP^A.

In support of their common evolutionary origin from a precursor visual organ (Fig. 1C), the compound eyes and ocelli share a large number of developmental regulators, including the proneural transcription factor Atonal (Ato) (Friedrich, 2006). During development, this bHLH protein plays an essential role in the transition from progenitor cells to differentiating photoreceptor neurons (Jarman et al., 1994; Jarman et al., 1995). In both compound eye and ocelli, ato is expressed transiently at the onset of neurogenesis. At first, its broad expression defines a domain of neuro-competent cells (proneural function) (Fig. 1D). Soon after, it becomes restricted to a small number of primary neuronal precursors. In the eye disc, these founder neurons drive the recruitment of additional photoreceptor cells that do not express ato (Fig. 1D) (Jarman et al., 1994). In the ocelli, the process of neurogenesis is less well characterized. Hence, the role of the primary neurons in generating the full complement of photoreceptor cells remains to be determined. Nonetheless, in the developing ocelli as in the compound eye, the initial and broad phase of ato expression depends on upstream factors, while the subsequent restricted phase reflects maintenance of ato transcription through an autoregulatory loop (Jarman et al., 1995; Sun et al., 1998). In addition, a lineage tracing experiment using the ato^5’EYE-GAL4 driver (which controls the restricted phase of ato expression; Yu et al., 2015) suggests that, as in the developing retina, some but not all ocellar photoreceptor cells derive from ato-expressing neuronal precursors (Fig. S1, see legend). These similarities strongly suggest related neurogenetic processes in the two types of visual organs.

The transcriptional control of ato expression, in both compound eye and ocelli, has been mapped to two large genomic DNA regions, a 3’ fragment (2 kb) controlling onset of expression and a 5’ fragment (3 kb) mediating direct or indirect autoregulation (Sun et al., 1998). Further analysis has focused on the dissection of the 3’ region in the developing compound eye (Zhang et al., 2006; Tanaka-Matakatsu and Du, 2008; Zhou et al., 2014). Here, the onset phase is dependent on transcription factors of the retinal determination gene network (RDGN) through two enhancers, 3’ENH^I and 3’ENH^II (Fig. 1E). Two RD proteins, the Pax6 factor Eyeless (Ey) and the Six1/2-type protein Sine oculis (So), directly bind at sites within the 3’ enhancers and together promote ato induction in the retina progenitor cells (Fig. 1E) (Zhang et al., 2006; Tanaka-Matakatsu and Du, 2008). Interestingly, the requirement for these sites was found to differ in the posterior versus anterior regions of the developing eye field, thereby defining two types of retinal progenitors (Zhou et al., 2014).

As mentioned above, ocellar control of ato transcription has also been broadly mapped to the same genomic regions active in the eye (Sun et al, 1998). Hence, the regulatory elements for proneural ato induction in the ocelli also lie in the same 3’ genomic fragment that contains the RDGN-responsive enhancers 3’ENH^I and 3’ENH^II (Fig. 1E). Although this may suggest shared mechanisms of ato regulation, it is clear that the genetic networks that specify the ocular and ocellar primordia are not identical, comprising several organ-specific components (Treisman, 2013; Domínguez-Cejudo and Casares, 2015). For instance, while So and its partner Eyes absent (Eya) promote the formation of both organs, the Pax6 input is mediated through different Pax6 gene family members, ey (eyeless) in the compound eye and twin of eyeless (toy) in the ocelli (Czerny et al., 1999; Blanco et al, 2010; Brockman et al., 2011). Further, the transcription factor Orthodenticle (Otd) and the Wingless (Wg) signaling pathway promote ocellar specification but antagonize compound eye development; whereas, the RD factor Teashirt (Tsh) is expressed in, and required for, formation of the compound eyes but not the ocelli (Royet and Finkelstein, 1996; Royet and Finkelstein, 1997; Baonza and Freeman, 2002; Bessa et al., 2002; Singh et al., 2004; Blanco et al., 2009 and 2010). These observations raise questions about how distinct or shared are the molecular mechanisms that mediate the transcription of ato in the two visual organ types.

Here, we report on the transcriptional regulation of ato in the ocelli, revealing the presence of distinct regulatory regions and of cis-elements shared with the compound eye. We demonstrate that the 3’ENH^I and 3’ENH^II retinal enhancers are also active in the ocelli and that their So and Pax6 binding sites are specifically required for ocellar ato expression. Strikingly, we find differential utilization of these sites in the anterior and posterior ocellar primordia in a way that parallels their use in anterior and posterior progenitors of the developing eye field. We discuss our findings in the context of the evolution of compound eyes and ocelli from a precursor photosensitive organ.

Materials and methods

Fly lines and transgenics

Drosophila lines were obtained from the Bloomington Drosophila Stock Center (BDSC) or the Vienna Drosophila RNAi Center (VDRC): UAS-Optix-RNAi (VDRC-KK110813), UAS-Dcr-2 (VDRC-60008), tsh[md621]-GAL4 (BDSC-3040), dpp[blk1]-GAL4 (BDSC-1553), UAS-Dcr-2; ey-FLP (BDSC-5580), UAS-GFP-nls (BDSC-4776), ato^5’EYE-GAL4 (Yu et al., 2015), the G-TRACE line UAS-FLP Ubi-p63E(FRT.STOP)Stinger (BDSC-28282; Evans et al., 2009) and Actin5C>y+>GAL4 (BDSC-3953); the latter two are referred to as UAS-FLP Ubi>IC>GFP and Act>IC>GAL4, respectively, in the text and figures (IC=interruption cassette). For reporter constructs, genomic DNA fragments were cloned into the vectors pCasper-β-gal (encoding cytoplasmic β-Galactosidase) or pStinger (encoding nuclear eGFP) (Pirrotta, 1988; Barolo et al., 2000). Both vectors were first modified to contain a 1.1 kb fragment spanning the ato promoter region and 5’UTR (Sun et al., 1998). All constructs were confirmed by sequencing (data are available upon request) and injected into w CantonS for P-element transformation (Spradling and Rubin, 1982). All crosses for Fig. 4 were carried out at 30°C to enhance Optix gene silencing.

Figure 4. Summary of ato regulation and proposed evolution of eye and ocelli from a compound-eye-like ancestral organ.

(A) Diagram of regulatory region for the onset of ato expression in eyes and ocelli. Position relative to transcription unit is shown above and contribution of DNA segments and binding sites to reporter gene expression is summarized below. Orientation of arrows reflects predicted effect on gene expression based on report constructs analyses, up- or down-regulation. The three binding sites function cooperatively in posterior retinal progenitors and OCP^P, but largely additively in anterior retinal progenitors and OCP^A. (B) Evolution of the compound eye and the ocelli from an already complex ancestral visual organ, a compound eye with genetically distinguishable posterior and anterior ommatidia. (C) Proposed evolution of ocelli by dorsal partitioning from of an ancestral compound eye, with posterior (green) and anterior (magenta) ommatidia.

Immunohistology and expression analysis

Antibodies (Ab) were obtained from the Developmental Studies Hybridoma Bank (DSHB) or other source, as indicated: mouse anti-Eya (1:200; DSHB); guinea pig anti-Sens (1:1000; gift of Dr. H. Bellen) (Nolo et al., 2000); rabbit anti-Optix (1:500) (Kenyon et al., 2005); rat anti-Elav (1:100; DSHB); rabbit anti-GFP (1:1000; Invitrogen); rabbit anti-β-gal (1:1000; Cappell). Anti-mouse, anti-rat, anti-guinea pig, and anti-rabbit Cy2-, Cy3- or Cy5-conjugated secondary Ab (Jackson ImmunoResearch Laboratories) were used at a dilution of 1:250. Samples were imaged with a Leica DM5500 Q confocal microscope and images were processed for publication with Adobe Photoshop. Multiple independent transgenic lines were analyzed in all experiments. Number of independent lines analyzed per construct is shown in Supplementary Fig. S2. Changes in expression were apparent and consistent across different reporter constructs based on multiple independent lines (see Fig. S2); comparisons were restricted to lines based on the same vector against separate pCasper-β-gal or pStinger full-length M”-fragment-containing control reporters. For quantification, images of 3 discs from 2–3 independent lines of each construct were analyzed in Photoshop CS4. The Adjusted Integrated Density was calculated by comparison of reporter signals from each developing ocellus (as defined by Eya expression) and an Eya-negative control region close to the ocellar domains and of equivalent size. Bar graphs report wild type (M”) expression readings set as 100 % and relative expression levels of other constructs. Statistical significance was analyzed by Student’s t-test by comparison of reporters to M” controls and is reported in figure legends.

Electrophoretic mobility shift assay (EMSA)

Toy protein was produced using a reticulocyte lysate in vitro transcription-translation system (Promega). DNA fragments used as probes were generated by PCR. Unlabeled competitor DNAs, containing the Pax6^EI site, with wild type (wt) sequence or mutated, were generated by annealing of synthesized fragments and used in 100× excess. DNA labeling and binding reactions were carried out as per the DIG Gel Shift Kit-2nd generation (Roche), followed by electrophoresis in 5% non-denaturing polyacrylamide gels and signal detection by phosphorimager.

Results and Discussion

To better understand the control of ato in the developing ocelli, we studied the expression of various ato reporter constructs (Supplementary Fig. S2). In these analyses, we identified the ocellar primordia, OCP^P and OCP^A, by the expression of the Eya protein in both progenitor and differentiating ocellar cells (Fig. 1F-F’). Since ocellar morphogenesis begins late in the third and last larval stage (L3), we used the onset of Senseless (Sens) expression to identify discs suitable for reporter analysis (Fig. 1F-F’). The gene sens is directly activated by the Ato transcription factor in the R8 founder neuron, which derives from the primary neuronal precursor (Nolo et al., 2000; Pepple et al., 2008). Although sens has not been investigated in the ocelli, Sens expression is observed within the ocellar primordia in the very late L3 disc providing a suitable stage marker.

Similarities and differences in the cis-regulatory control of ato in eyes and ocelli

In previous studies, we identified two evolutionarily conserved enhancers for retinal expression, 3’ENH^I (183 bp) and 3’ENH^II (300 bp), within a 2 kb-long 3’ fragment (M”) that mediates onset of ato in both eyes and ocelli (Fig. 2A) (Zhang et al, 2006). To investigate whether 3’ENH^I and 3’ENH^II are essential for ato expression in the developing ocelli, we assessed for ocellar expression of reporters deleted for 3’ENH^I (M”-Δ183) or 3’ENH^II (M”-Δ300) (Fig. 2A). Both deletion reporters failed to be expressed in OCP^P and were expressed in only a few cells of the OCP^A (Fig. 2A, C–D’). These results show that, as in the retina, reporter expression in the ocelli is controlled through the 3’ENH^I and 3’ENH^II enhancers, raising the possibility that the previously identified Six1/2 and Pax6 binding sites mediate transcriptional control in both sensory organs.

Figure 2. The 3’ENH^I and 3’ENH^II retinal enhancers are required but not sufficient for expression in the ocelli.

All constructs compared carry the β-Galactodidase (cytoplasmic) reporter. (A-A’) Diagram of constructs analyzed with summary results for relative expression level. Standard deviation is shown. Student’s t-test p-values in relation to M”-lacZ were p≤0.01 for M’-D183, M”-D300 and EI+EII, and <0.05 for 1.2. (B–F’) Confocal images of representative samples with immunostaining for Eya (blue) to identify the ocellar primordia, Sens (red) to confirm late L3 stage, and β-Galactosidase (green) to assess reporter constructs expression. In all panels, orientation is as in Fig.1F-F’ and posterior and anterior ocellar primordia are marked by orange and yellow dashed lines, respectively.

In contrast, reporter constructs covering different subregions of the M” fragment provided evidence for organ-specific regulatory inputs. First, multiple independent lines of the 5’-truncated reporter 1.2 (lacking 879 bp) consistently showed lower expression in the ocelli than M” reporter lines (Fig. 2A and E-E’). Notably, the opposite effect is observed in the retina where the expression of 1.2 reporter lines is consistently higher than for M” lines (Zhang et al., 2006). Thus, the region of M” DNA deleted in the 1.2 promotes expression in the ocelli but downregulates expression in the eye. Second, an EI+EII reporter (consisting of only 3’ENH^I and 3’ENH^II DNA) was not expressed in the ocellar region (Fig. 2A and F-F’), although robustly expressed in the retina (Zhou et al., 2014). Hence, the DNA region between EI and EII must contain one or more cis-elements that are essential for reporter activation in the ocelli, but that are dispensable in the compound eye.

Taken together, these results confirm that the regulatory regions for ato expression in eyes and ocelli overlap and that the enhancer regions 3’ENH^I and 3’ENH^II are required in both organs. However, two lines of evidence suggested that ocelli-specific control elements must also exist: 1) the 3’ENH^I and 3’ENH^II are sufficient for retinal but not ocellar expression and 2) the 879 bp region contributes differently to ato expression in the developing ocelli and compound eyes. Whether the latter regulation involves distinct or shared (though differentially utilized) cis-regulatory sites remains to be determined. It is conceivable, for instance, that Toy and/or Otd (which are required in the development of the ocelli but not the retina) and Ey and/or Tsh (which are required in retina progenitors but not in the ocelli) contribute to the organ-specific regulation of ato.

Differential use of shared So and Pax6 binding sites in anterior vs. posterior ocellar progenitors

The requirement for the 3’ENH^I and 3’ENH^II control regions in both compound eyes and ocelli suggested that some of the mechanisms for ato induction might be identical or very similar in the two organs. To test this hypothesis, we focused on well-characterized binding sites known to drive ato expression in retinal progenitor cells, including one Pax6 site in 3’ENH^I and two sites, one for So and one for Pax6, in 3’ENH^II. Assessing the expression of a triple mutant M” reporter lacking all three sites (M”-ΔPax6^EISo^EIIPax6^EII), we observed a complete absence of expression in OCP^P and a drastic reduction in OCP^A (Fig. 3A and I-I’). This result indicated that at least some of these sites function in the ocelli. We therefore proceeded to analyze the regulation of ato expression using single and double mutant reporters.

Figure 3. Pax6 and So sites are required and differentially utilized in posterior versus anterior ocelli.

All constructs compared carry the eGFP (nuclear) reporter. (A-A’) Symbolic representation of M”-reporter constructs (wt and altered) with summary results of relative expression levels. Filled shapes indicate normal binding sites, unfilled shapes indicate deleted binding sites. Letters on the left correspond to following panels B-I’. Standard deviation is shown. Student’s t-test p-values in relation to M”-GFP were p≤0.01 for all modified reporters except for M”-ΔPax6^EII and M”-ΔSo^EII in the OC^A, which showed no change; p-value for M”-ΔPax6^EIPax6^EII or M”-ΔPax6^EISo^EII compared to M”-ΔPax6^EI is p≤0.01. (B–I’) Confocal images of representative samples with immunostaining for Eya (blue) and Sens (red), as explained in legend of Fig. 2, and GFP (green) to assess reporter constructs expression. In all panels, orientation is as in Fig.1F-F’ and posterior and anterior ocellar primordia are marked by orange and yellow dashed lines, respectively. (J) EMSA for Toy protein binding at the Pax6^EI site. From left to right: a shift of the 120 bp fragment spanning part of 3’ENH^I is observed when Toy protein is present (compare lane 1 to lane 2). The shift is suppressed by cold DNA competitor with the Pax6^EI binding site (wt), but not by DNA competitor with a mutated Pax6^EI sequence (mut). Arrow marks position of Toy-bound DNA probe.

In the OCP^P, deletion of any one site (M”-ΔPax6^EI, M”-ΔPax6^EII or M”-ΔSo^EII) resulted in severe reduction of reporter gene expression (Fig. 3A–E’). Interestingly, the effect of deleting both 3’ENH^II sites (M”-ΔSo^EIIPax6^EII) was indistinguishable from deletion of either site alone (Fig. 3A and F-F’). In contrast, double-deletions of the 3’-ENH^I Pax6^EI site with either of the sites in 3’-ENH^II (M”-ΔPax6^EIPax6^EII or M”-ΔPax6^EISo^EII) resulted in the complete absence of expression (Fig. 3A and G–H’). These results suggest that the two adjacent binding site in 3’ENH^II together facilitate one trans-regulatory input, whereas through their respective Pax6 and So binding sites, the 3’ENH^I and 3’ENH^II function synergistically to induce gene expression in the posterior ocelli.

In the OCP^A, by contrast, deletion of any one site (M”-ΔPax6^EI, M”-ΔPax6^EII or M”-ΔSo^EII) resulted in either a moderate (~50%) decrease in reporter expression (M”-ΔPax6^EI) or no significant change (M”-ΔPax6^EII or M”-ΔSo^EII) (Fig. 3A–E’). Further analyses of Pax6^EII or So^EII demonstrated that these two sites, similarly to the Pax6^EI site, contribute partially to the expression of ato in the OCP^A. First, the double-deletion reporter M”-ΔSo^EIIPax6^EII showed only ~1/3 of the expression of the M” reporter (Fig. 3A, F-F’). Second, loss of either the Pax6^EII or the So^EII site in a Pax6^EI-deleted background caused a further reduction in expression compared to loss of Pax6^EI alone (compare M”-ΔPax6^EIPax6^EII or M”-ΔPax6^EISo^EII to M”-ΔPax6^EI) (Fig. 3A, C-C’ and G–H’). These results show that all three sites regulate expression in the OCP^A and, once again, uncover an interaction between 3’ENH^II and 3’ENH^I. As mentioned above—and in agreement with partially redundant roles for the three sites in the OCP^A—deletion of all three left only minimal expression (Fig. 3A and I-I’).

Altogether, these findings reveal that eyes and ocelli rely on the same cis-binding sites for control of ato expression by specification factors, but that the regulatory mechanisms at work in the posterior and anterior ocelli are not identical.

Interestingly, evidence of distinct regulation of ato has also been observed along the anterior-posterior axis of the compound eye (Zhou et al., 2014). In the developing eye, the Pax6^EI, Pax6^EII and So^EII sites are differentially required in two groups of retinal progenitor cells: the posterior progenitors and the anterior progenitors (Zhou et al., 2014). Remarkably, the OCP^P displays a regulatory logic of ato induction similar to that observed in posterior retina progenitors, with each of the three binding sites playing profoundly essential roles. Induction of ato transcription in the OCP^A, instead, displays a regulatory logic similar to that observed in anterior retinal progenitors, with all three sites showing significant partial redundancy. Thus, the developing compound eye and the ocelli not only rely on the same cis-bindings sites, but also display parallel differences in their usage between posterior and anterior progenitors in both organs.

In this regard, it is interesting to note that the requirement for the transcription factor Optix also differs similarly along the anterior-posterior axis of both organs. In the ocelli, Optix is expressed and required in the developing OC^A, but not OC^P (Domínguez-Cejudo and Casare, 2015). Similarly, in the developing compound eye, Optix is not expressed in the posterior-most region of the eye field, even prior to the start of neurogenesis (Supplementary Fig. S3A), and it is required for the formation of anterior ommatidia but not of several rows of posterior ones (Li et al., 2013). In agreement with these reports, we observed loss of anterior ommatidia and developmental defects when Optix was silenced preferentially in the anterior progenitors of the compound eye field (tsh-GAL4 UAS-Optix^RNAi) (Supplementary Fig. S3B-B”), but not when preferentially inactivated in the posterior ones (dpp-GAL4 UAS-Optix^RNAi) (Supplementary Fig. S3C-C”). Accordingly, a reduced, Bar-like, eye (Supplementary Fig. S3D–E’) and the loss of the anterior ocellus (Supplementary Fig. S3F–G’) were observed when Optix was inactivated throughout the eye disc (ey-FLP Act>IC>GAL4 UAS-Optix^RNAi). Altogether, these parallels in differential ato regulation and Optix function may reflect a shared distinction of anterior-type versus posterior-type progenitor cells in both retina and ocelli.

Toy and So directly regulate ato expression through the Pax6^EI, Pax6^EII, and So^EII sites

Previous studies have shown that the Pax6 homolog Toy, but not Ey, is expressed during and required for ocellar development (Jacobsson et al., 2009; Blanco et al., 2010), suggesting that Toy regulates ato expression through the sites Pax6^EI and Pax6^EII in the ocelli. Consistent with this conclusion, previous work reported that not only Ey but also Toy can bind the Pax6^EII element in vitro (Zhang et al., 2006). To determine whether Toy can also bind the Pax6^EI site, we carried out an electrophoretic mobility shift assay. As shown in Fig. 3J, Toy can bind and shift a DIG-labeled, 120-bp DNA fragment containing the Pax6^EI site. This binding was sequence-dependent because inclusion of unlabeled DNA fragments suppressed the gel shift, whereas DNA fragments mutated at the Pax6^EI site could not (Fig. 3J). Thus, we conclude that Toy directly regulates the M” reporters through the Pax6^EI and Pax6^EII sites of the 3’ENH^I and 3’ENH^II enhancers.

Our findings also support a model whereby So regulates reporter expression in the ocelli by direct binding at the So^EII site. Previous studies have shown that So is expressed in the ocelli at the relevant developmental stage (Cheyette et al., 1994), that so function is absolutely required for ocelli formation (Heitzler et al., 1993), and that the So protein binds the So^EII sequence both in vitro and in eye field cells (Zhang et al., 2006; Zhou et al., 2014). Here we show that the So^EII site is required for reporter gene expression in the ocelli (Fig. 3). Hence, we propose that So directly regulates M” reporters’ expression in these organs, at least in part, through the So^EII site of 3’ENH^II.

In summary, the available data support a model whereby the RD factors Toy and So are direct regulators of the ato gene in the ocelli through the same cis-binding sites (Pax6^EI, Pax6^EII and So^EII) that control ato expression in retinal progenitors.

Conclusions and perspectives

Our molecular genetic dissection of ato activation in the developing ocelli confirms its dependence on the 3’ region of the ato gene (Sun et al., 1998), and further reveals that, along with OC-specific element(s), select input mechanisms are shared with the compound eyes (see Fig. 4A for summary diagram). The partial overlap of regulatory mechanisms in the two visual organs is well in line with their shared descent from an ancestral photosensitive organ present >500 million years ago (Fig. 1C), reflecting the occurrence of conservation and organ-specific modification (Friedrich et al., 2006). In the context of this phylogenetic framework, the differential requirement for the Pax6 factors Ey and Toy in compound eyes versus ocelli can be explained as the outcome of spatial subfunctionalization following gene duplication (Force et al., 1999). Further, our finding that the ato regulatory DNA also contains organ-specific, or differentially utilized, control elements is consistent with the structural and functional diversification of eyes and ocelli (Friedrich, 2006). Most likely, the latter process occurred in concert with the modification of the gene specification networks for the two types of visual organs, resulting in the distinct roles of several regulators, such as Otd, Wg, Tsh, Hedgehog and Engrailed, in the organization of distinct organ primordia (see introduction; review by Treisman, 2013; Royet and Finkelstein, 1996; Royet and Finkelstein, 1997; Baonza and Freeman, 2002; Bessa et al., 2002; Singh et al., 2004; Blanco et al., 2009 and 2010; Domínguez-Cejudo and Casares, 2015).

Arguably, the most surprising finding from our study is that of unexpected parallels in the differential regulation of ato in posterior versus anterior primordia of both the compound eyes and ocelli. This is evidenced by the differential activation mechanism via the So, Ey, and Toy inputs in the anterior ocellus and anterior compound eye versus the posterior ocelli and posterior compound eye (Fig. 3 and Zhou et al., 2014). Intriguingly, Optix also displays differences along the anterior-posterior axes of both organs; specifically, it is required in the anterior ocellus and the anterior region of the compound eye, but not for posterior ommatidia and posterior ocelli formation (Fig. S3; Li et al., 2013; Domínguez-Cejudo and Casare, 2015). These parallel differences cast interesting light on the origin of both organs from their shared precursor. The most parsimonious explanation for the use of two types of progenitors that are similarly defined in both organs is that the underlying genetic control mechanisms predate the emergence of eyes and ocelli. Hence, we propose that these shared aspects of ocellar and compound eye development reflect regulatory constraints inherited from their shared precursor organ. More precisely, it is tempting to speculate that the ocelli emerged as a partition from the dorsal edge of an ancestral compound eye (Fig. 4B–C) that was already characterized by the differential regulation still observed in present day insects like Drosophila. Our findings thus speak to the ongoing discussion on the organization of the ancestral visual organ of the early arthropod (Oakley, 2003; Mayer, 2006; Strausfeld et al., 2016) and deliver first candidate mechanisms that can be tested for conservation in the developing visual systems of a broader range of arthropod species.

Supplementary Material

1. Supplementary Figure S1: lineage tracing of ocellar neurons derived from primary precursors.

The ato^5’EYE-GAL4 driver contains the 5’ regulatory region for Ato-dependent expression of the ato gene in the developing eye and ocelli. It is therefore expressed in the primary neuronal precursors of eye and ocellar photoreceptors. We crossed ato^5’EYE-GAL4 with the G-TRACE line UAS-FLP Ubi>IC>GFP in order to irreversibly mark with GFP the photoreceptor neurons derived from the primary precursors. Panels show developing OCP^P (A-A’) and OCP^A (B) from ~1-day-old pupal disc, at a time when disc fusion is already in progress (note side-by-side anterior ocellar fields in the process of fusing to form one anterior ocellus; B). Discs were dissected and immunostained for Elav (red), to mark all developing neurons, and GFP (green), to mark direct descendents of primary neuronal precursors. Note that many developing photoreceptors (red) do not express GFP (green). This suggests that, as in the developing eye, additional photoreceptors are recruited from surrounding, non-ato-expressing cells.

2. Supplementary Figure S2: ato reporter constructs.

List (A) and diagram (B) of all transgenes analyzed. Comparisons were carried out only between constructs based on the same vector.

3. Supplementary Figure S3: Optix is required for the formation of anterior, but not posterior, ommatidia and ocelli.

(A) L2 and L3 wt eye-antennal discs immunostained for Eya (red) to mark the eye field and Optix (green) to assess its expression. Optix is expressed in most but not all Eya-positive cells (eye primordium) in the L2 stage, prior to the start of neurogenesis. During L3, Optix expression is maintained in retinal progenitors, but is then downregulated as progressive rows of progenitor cells initiate neurogenesis and then form the ommaditial units. (B–E’) Optix is absolutely required for ommatidia formation in the anterior but not posterior of the eye, although the development and organization of the posterior ommatidia is not entirely normal. (B–C”) Effect of RNAi-mediated silencing of the Optix gene on eye formation. When UAS-Optix^RNAi is expressed specifically in anterior retinal progenitors (tsh-GAL4) (B), but not when expressed mainly in posterior progenitors (dpp-GAL4) (C), the fly eye is severely reduced in size (compare eye in B’ to eye in C’). Diagrams in B and C show expression of GAL4 lines at L2/early-L3 stage. The development of the ommatidial array, as assessed by immunostaining for Eya (red) and the pan-neural marker Elav (green), is impaired (compare eye disc in B” to eye disc in C”; insets in B” and C” show Eya alone). (D) Control wt eye. (E-E’) Optix silencing throughout the eye disc epithelium (ey-FLP Act>IC>GAL4 UAS-Dcr-2 UAS-Optix^RNAi) results in a severely reduced adult eye (E) and abnormal development in the L3 disc (E’). Disc in E’ is stained for Eya (red) to visualize eye development and Optix (green) to confirm effective RNAi-mediated gene silencing. (F–G’) Effect of Optix silencing on the ocelli: specific loss of the OC^A (compare panel G to F) is observed in ey-FLP Act>IC>GAL4 UAS-Dcr-2 UAS-Optix^RNAi flies, and correlates with decreased Eya expression in OCP^A (compare panel G’ to F’). Inset in F’ shows double staining for Eya (red) and Optix (green) in the OCP^A marked by the dashed line. Yellow arrows point to missing OC^A in G and missing OCP^A in G’.

Highlights.

• 3’ ato regulatory DNA for eyes and ocelli contains shared and unique cis-elements.

• So directly regulates ato expression in ocelli through sites used by So in retinal progenitors.

• Toy directly regulates ato expression in ocelli through sites used by Ey in retinal progenitors.

• So and Pax6 sites are differentially used in OC^A and OC^P, as also observed in eye.

• Shared mechanisms of ato regulation suggest descent of ocelli from precursor compound eye.