Identification of Transcriptional Targets of the Dual Function Transcription Factor/Phosphatase Eyes Absent

Abstract

Drosophila eye specification and development relies on a collection of transcription factors termed the retinal determination gene network (RDGN). Two members of this network, Eyes absent (EYA) and Sine oculis (SO), form a transcriptional complex in which EYA provides the transactivation function while SO provides the DNA binding activity. EYA also functions as a protein tyrosine phosphatase, raising the question of whether transcriptional output is dependent or independent of phosphatase activity. To explore this, we used microarrays together with binding site analysis, quantitative real-time PCR, chromatin immunoprecipitation, genetics and in vivo expression analysis to identify new EYA-SO targets. In parallel, we examined the expression profiles of tissue expressing phosphatase mutant eya and found that reducing phosphatase activity did not globally impair transcriptional output. Among the targets identified by our analysis was the cell cycle regulatory gene, string (stg), suggesting that EYA and SO may influence cell proliferation through transcriptional regulation of stg. Future investigation into the regulation of stg and other EYA-SO targets identified in this study will help elucidate the transcriptional circuitries whereby output from the RDGN integrates with other signaling inputs to coordinate retinal development.

Introduction

Regulation of gene expression is a primary means by which signaling networks control cell fate specification. Studies of the compound eye of Drosophila melanogaster have provided numerous insights into how multiple signaling pathways are integrated at the level of transcription to control proliferation and cellular differentiation during development. The Drosophila eye is composed of approximately 800 units, called ommatidia, which each contain eight photoreceptor neurons and 12 accessory cells. The adult eye develops from a structure called the eye imaginal disc, which consists of cells set aside in the embryo that subsequently proliferate and differentiate during larval and pupal development (Wolff, 1993). In the third instar larval eye imaginal disc a wave of differentiation, termed the morphogenetic furrow (MF), initiates at the posterior of the disc and moves anteriorly across the field, marking the transition from an asynchronously proliferating population of cells to G1 arrested cells (Wolff, 1993). After specification of the initial five photoreceptors just posterior to the MF, the remaining undifferentiated cells undergo a final mitotic division, called the second mitotic wave, and subsequently differentiate to give rise to the additional photoreceptors and accessory cells (reviewed by (Wolff, 1993).

Vital to the process of eye development is a network of transcription factors, known as the retinal determination gene network (RDGN), which are required for early eye specification. The genes comprising the RDGN include eyeless (ey), twin of eyeless (toy), sine oculis (so), eyes absent (eya), optix and dachshund (dac). Loss of function of these genes results in the complete absence of eye tissue, while overexpression in non-retinal tissues often induces ectopic eye formation (Bonini et al., 1997; Bonini et al., 1993; Chen et al., 1997; Cheyette et al., 1994; Czerny et al., 1999; Halder et al., 1995; Mardon et al., 1994; Pignoni et al., 1997; Quiring et al., 1994; Shen and Mardon, 1997; Weasner et al., 2006). Within the RDGN, genetic epistasis and expression analyses have positioned EY and TOY upstream of EYA, SO and DAC (Czerny et al., 1999; Halder et al., 1998; Niimi et al., 1999; Ostrin et al., 2006; Shen and Mardon, 1997; Zimmerman et al., 2000). Supporting the genetic data, EY has been shown to directly regulate transcription of eya and so, with regulation of the latter involving cooperation with TOY (Niimi et al., 1999; Ostrin et al., 2006; Punzo et al., 2002). A number of positive feedback loops, some of which operate at the level of direct transcriptional control, further reinforce expression of network components to drive eye development (Bonini et al., 1997; Bui et al., 2000b; Chen et al., 1997; Pauli et al., 2005; Pignoni et al., 1997; Shen and Mardon, 1997).

While the regulatory relationships within the RDGN have been worked out, much less is known about how RDGN members modulate patterns of gene expression to yield specific developmental outcomes. As a downstream component of the RDGN, EYA provides a logical place to begin examining how transcriptional output from the RDGN leads to retinal specification. EYA family proteins are conserved from worms to humans and are defined by a conserved C-terminal domain, termed the EYA domain (ED). The ED is required for interaction with SO and DAC, and also contains a phosphatase catalytic motif (Chen et al., 1997; Li et al., 2003; Pignoni et al., 1997; Rayapureddi et al., 2003; Tootle et al., 2003; Zimmerman et al., 1997). EYA mediates transactivation function through its more divergent N-terminal half, which contains a second moderately conserved domain, the EYA domain 2 (ED2), embedded in a proline-serine-threonine (P/S/T)-rich stretch of amino acids (Zimmerman et al., 1997). The P/S/T-rich region is required for transactivation, while the role of the ED2 domain remains unclear (Silver et al., 2003; Xu et al., 1997). Given that EYA does not have DNA binding activity, it must bind a cofactor to be recruited to target DNA. While EYA binds both SO and DAC through the ED (Chen et al., 1997; Pignoni et al., 1997), only SO has been demonstrated to recruit EYA to target DNA (Ohto et al., 1999; Silver et al., 2003). Previous studies of EYA-SO transcriptional targets have focused on identifying SO binding sites in target genes and showing the ability of EYA to coregulate expression. These studies have lead to the identification of five EYA-SO targets in Drosophila: lozenge (lz), sine oculis (so), eyeless (ey), hedgehog (hh) and atonal (ato) (Pauli et al., 2005; Yan et al., 2003; Zhang et al., 2006), all of which are required for proper eye development.

As mentioned above, EYA not only functions as a transcription factor, but also as a phosphatase. This unique juxtaposition of functions is intriguing and begs the question of whether phosphatase activity is required for transcriptional regulation of EYA-SO target genes or whether the two functions are independent. Experiments using transcriptional reporter assays in transfected cultured cells have yielded conflicting answers to this question (Li et al., 2003; Tootle et al., 2003), emphasizing the necessity of additional analysis. Although the relationship between these two EYA functions remains to be determined, it should be emphasized that both are required for eye development (Rayapureddi et al., 2003; Silver et al., 2003; Tootle et al., 2003).

In order to begin exploring the relationship between EYA phosphatase and transcription functions, we needed first to improve our understanding of EYA's role as a transcription factor by expanding the list of EYA transcriptional targets. Toward this goal, we used a microarray-based approach to identify genes whose expression was altered upon overexpression of a wild type eya (eya^WT) transgene. In parallel, we examined the ability of eya phosphatase mutant (eya^MUT) transgenes to regulate these potential target genes. From our microarray analysis, we identified 577 genes that underwent at least a two-fold change in expression upon overexpression of wild type or phosphatase mutant eya. Comparison of the expression profiles for tissue overexpressing wild type or mutant eya reveals that the majority of genes have similar expression changes across all samples. Further comparison of the ability of EYA^WT and EYA^MUT to upregulate putative targets by quantitative PCR analysis, confirms that reduced phosphatase activity may decrease EYA's transactivation potential in some contexts, but does not eliminate its transcriptional ability. Focusing on those genes that were upregulated in all of our experimental samples, we asked which might be directly regulated by EYA and SO by first identifying potential SO binding sites in silico and then experimentally confirming their functionality using chromatin immunoprecipitation (ChIP). Among the five validated targets that resulted from this analysis, we have focused on string (stg), a Cdc25a phosphatase that regulates the G2 to M phase transition of the cell cycle by dephosphorylating Cdk1. Our results suggest that EYA and SO might influence cell proliferation by transcriptionally regulating stg.

Materials and Methods

Western Blots

Adult flies of the genotypes 1) w¹¹¹⁸, 2) w¹¹¹⁸; hsp>eya^WT, 3) w¹¹¹⁸; hsp>eya^D493N, and 4) w¹¹¹⁸; hsp>eya^E728Q were heat shocked at 37°C for 1 hour, followed by a 3 hour incubation at 25°C. Fly heads were prepared by collecting adult males and females into microcentrifuge tubes, freezing in liquid nitrogen, vortexing, and repeating three times. Fly heads were collected into a new tube and 50μl of 2x reducing sample buffer (100 mM Tris-Cl (pH 6.8), 4% SDS, 20% glycerol, 200mM DTT) was added. Tissue was homogenized, boiled and run on an 8% SDS-PAGE gel. Protein levels were detected with gpαEYA (1:10,000) and mαtubulin E7 (1:5000; Developmental Studies Hybridoma Bank) using the Li-COR imaging system and quantitated using ImageJ software (Abramoff, 2004).

Microarray sample preparation and hybridization

Flies of the genotypes described above were used for microarray analysis. Adult males and females were collected into 15ml falcon tubes, frozen in liquid nitrogen, and vortexed, repeating three times. Fly parts were poured onto an RNaseZap (Ambion) treated glass plate and 50 heads were collected into an eppendorf tube using a paintbrush. After homogenization in 50μl of Trizol reagent (Invitrogen), an additional 200μl of Trizol was added. Four samples were pooled for a total of 200 heads and RNA was isolated in Trizol according to the manufacturer's instructions and purified using the RNeasy Minikit (Qiagen). Total RNA concentration and 260/280 ratio was determined using a NanoDrop ND-1000 UV-Vis Spectrophotomer (NanoDrop Technologies). cDNA synthesis was carried out according to the Expression Analysis Technical Manual (Affymetrix, standard sample) using 2μg of total RNA for each sample. cRNA reactions were carried out using one cycle cDNA synthesis according to the GeneChip Expression Analysis Technical Manual (Affymetrix). Ten micrograms of labeled cRNA were fragmented and hybridized to Affymetrix Drosophila Genome Array version 2.0 following the manufacturer's instructions.

Microarray data analysis and binding site identification

Raw data were exported to Excel from the Affymetrix Microarray Suite, normalized and expression ratios and t-test values calculated. For genes whose expression was considered to be absent, the normalized filtered value was initially set to one, and changes in expression were confirmed by repeated analysis in which the normalized filtered value was set to 50. Only the 577 genes which showed a two fold change in expression levels (based on a log₂ value of 1) and had a t-test value of <0.05 in at least one of the experimental samples were used for further analysis.

For binding site analysis the publicly available program Genome Enhancer (http://genomeenhancer.org/fly, (Markstein et al., 2002)) was used. Each chromosome arm was examined individually to allow for identification of the maximal number of clusters. Binding sites for SO targets were identified using the following conditions: four YGATAY (Y=T/C) sites within a 50bp window; a single copy of the longer more degenerate sequence described by Pauli et al (Pauli et al., 2005), GTAANYNGANAYS (Y=T/C, S=C/G, N=any nucleotide) in the forward and reverse direction; one to four copies of the short sequence in conjunction with binding sites for the transcription factors ETS (GGA(T/A)), TCF/LEF ((T/A)(T/A)CAAAG), STAT (TTCCCGGAA), MAD (GCCGNCG), Su(H) ((T/C)GTG(A/G)GAA) and GLI (TGGG(T/A)GGTC). See Supplementary Materials and Methods for further details regarding search parameters. Conservation of binding sites was assessed using the USCS Genome Browser (http://genome.ucsc.edu/) and examining sequence alignments across twelve Drosophila species.

Quantitative real-time RT-PCR

Adult flies of the genotypes w¹¹¹⁸, w¹¹¹⁸; hsp>UAS-eya^WT and w¹¹¹⁸; hsp>UAS-eya^E728Q were heat shocked, 25 heads collected, and RNA isolated in Trizol and resuspended in 16μl nuclease-free H₂0 as described for microarray sample preparation. RNA was DNase (Invitrogen) treated for 1 hour in a total volume of 20μl at 37°C, followed by addition of 1μl of 25mM ETDA and inactivation at 70°C for 10 minutes. cDNA was generated from 1μg of DNase-treated total RNA using the Reverse Transcription System (Promega) according to the manufacturer's instructions. Quantitative PCR was carried out in triplicate on the Stratagene Mx4000 machine using 1.0 μl of the undiluted RT reaction, 0.5μl each of 100ng/μl sense and antisense primer, and 12.5μl of Quantitect SYBR Green (Qiagen) in a total volume of 25μl. For analysis in eye-antennal imaginal discs, 40 pairs of eye-antennal discs from third instar wandering larvae of the genotypes GMR-Gal4 and GMR-Gal4>UAS-eya, which had been shifted to 29°C two days after egg laying, were dissected in Schneider's S2 cell medium and processed as described above. Quantitative real-time PCR was carried out as described above using 0.5μl cDNA on the ABI Prism 7700. Relative quantification of real-time PCR product was performed using the comparative C_T method (as described in the ABI Prism 7700 Sequence Detection System User Bulletin no. 2) and normalized by subtracting the C_T value of the control gene ribosomal protein 17 (RpS17). Primers for this analysis are in Supplementary Table S1.

Chromatin immunoprecipitation

Eye-antennal discs were dissected from third instar larvae in Drosophila S2 media. One hundred pairs of eye-antennal discs were fixed for 15 minutes at room temperature in 1ml of S2 media with 40.5μl of 37% formaldehyde. Glycine was added to 125mM and incubated for 5 minutes on ice. Discs were washed 3 times in 1xPBS and resuspended in ChIP lysis buffer (50mM KHepes pH 7.8, 140mM NaCl, 1mM EGTA, 1mM EDTA, 1% Triton X-100, 0.1% Nadeoxycholate) with protease inhibitors. Samples were kept on ice and pooled for a total of 200 pairs of discs for each condition. Discs were homogenized with a pestle in an eppendorf tube 3 times and syringe passed 10 times with a 25 gauge needle and 10 times with a 27 gauge needle, followed by incubation in lysis buffer 20 minutes at 4°C with rocking. Discs were sonicated on ice using a Branson digital sonifier for 15 rounds with two minutes rest in between at the following settings: 15% amplitude, 15 sec (.9 seconds on/.1-.2 seconds off). Following sonication, samples were spun at 13.2K for 10 minutes at 4°C. The supernatant from 400 total disc pairs was pooled in a new tube, and 10μl was removed for the input and frozen. The remainder of the supernatant was divided into equal volumes for the control and the experimental sample (∼600μl each). For the control sample, pre-immune serum was added at a concentration of 1:500. For the experimental sample gpαSO antibody was added at a concentration of 1:500 (Mutsuddi et al., 2005). Samples were incubated overnight at 4°C. On the second day, Gammabind G-Protein-Coupled Sepharose beads (Amersham Biosciences) beads were washed 2 times in 1xPBS and 20μl of a 1:1 slurry of the beads and lysis buffer were added to each sample and incubated for 3 hrs. Samples were spun down for 2 min at 2K and the supernatant removed. Beads were washed 3×5min with ChIP lysis buffer, 1×5 min with high salt ChIP lysis buffer (same as above with 500mM NaCl), 1×5 min in TE. Beads were resuspended in 150μl TE/SDS (10mM Tris pH 8, 1mm EDTA, 1%SDS) and 140μl TE/SDS was added to the input sample. Samples were incubated at 65°C for 10 min, with vortexing every couple of minutes. Beads were pelleted at 14K for 1 min and the supernatant was removed to a fresh tube. Tubes were sealed with parafilm and incubated at 65°C overnight to reverse crosslinks. Qiagen PCR purification kit was used to extract DNA. Sample was eluted in 30ul elution buffer and PCR analysis was performed with appropriate primers and 1-3μl of eluted DNA. (See Supplementary Table S2 for a list of ChIP primers.) Chromatin enrichment in the IP sample over the mock treated sample was determined using ImageJ (Abramoff, 2004).

Gel shift assays

For labeling of probes, 25pmol of annealed oligos (see Supplementary Table S3 for oligo sequences) were labeled with T4 polynucleotide kinase (New England BioLabs) and 75μCi γ–³²P-dATP in a total volume of 25μl for 30 min at 37°C. Kinase was inactivated by ∼20 min incubation at 65°C. Recombinant GST-full length SO was purified from bacteria as previously described (Sullivan, 2000). Binding reactions were carried out in 15μl binding buffer (10mM HEPES pH 7.9, 35mM KCl, 10mM NaCl, 4mM MgCl₂, 20% glycerol, 1mM DTT, 1mM EDTA, 0.1% BSA), 1μg poly dI-dC, 1μl labeled probe, 2μl protein and H₂O in a total volume of 20μl. Samples were incubated for 20 min and analyzed on 6% non-denaturing polyacrylamide gels followed by autoradioagraphy. For competition experiments, samples were pre-incubated with 100X excess of unlabeled mutant or wild type probe for 10 min at room temperature before adding labeled probe.

In situ hybridization

stg fragments were generated by PCR amplification from genomic DNA using the following oligos: STG-Sense 5′- GATAATTTAGGTGACACTATAGAAGAGATGACGGAGAGCAACACCAACAGC-3′ and STG-Antisense 5′-GAATTAATACGACTCACTATAGGGAGAGAAGCGGGACATTTTCGGTC-3′. The PCR product was gel purified using the QIAquick Gel Extraction kit (Qiagen) according to manufacturer's instructions. One microgram of DNA was used in a total volume of 15μl and was heated to 95°C and cooled on ice. Two microliters of Dig DNA labeling mix (Roche), 2μl hexanucleotide mix (Roche) and 1μl of Klenow (New England BioLabs) were added and the labeling reaction was incubated overnight at room temperature. The labeled probe was precipitated and resuspended in 50μl nuclease-free H₂0. Eye-antennal or wing imaginal discs were dissected from third instar larvae in Schneider's S2 cell media. Discs were fixed in 4% paraformaldehyde for 20 min on ice, followed by incubation in 4% paraformaldehyde w/ 0.6% Triton X-100 for 15 min at room temperature. Discs were washed 3x5 min in 1xPBS w/ 0.3%Triton and 1x5 min in PBT (PBS + 0.1% Tween-20), post-fixed in 4% paraformaldehyde for 20 min at rooom temperature, washed 3× 5min in PBT, 1×5 min PBT/hybridization buffer (50% formamide, 5xSSC, 100μg/ml salmon sperm DNA, 100μg/ml tRNA, 50μg/ml heparin, 0.1% Tween-20) and 1×5 min hybridization buffer. Discs were prehybridized in hybridization buffer for at least 1 hr at 48°C. Probe was heated to 95°C for 3min, cooled and added at a concentration of 1:10 for overnight hybridization at 48°C. Discs were washed in a PBT/hybridization buffer series, followed by 5× 20 min washes in PBT and incubation with preabsorbed alkaline phosphatase-conjugated anti-dioxygenin antibody (1:2000) (Roche) overnight at 4°C. Development was carried out in AP Buffer (100mM Tris pH 9.5, 100mM NaCl, 50mM MgCl2, 0.1% Tween-20, 1mM Levamisol) with 3.375μl NBT/3.5μl BCIP (Roche).

Immunostaining

Wing imaginal discs from wandering third instar larvae were dissected in Drosophila S2 media, fixed for 10 min in 4% paraformaldehdye with 0.1% Triton X-100, washed 3x in PT(1x PBS, 0.1% Triton), and blocked for 30 min in PNT (1x PBS, 0.1% Triton, 1% normal goat serum). Samples were incubated with mouse anti-dac (1:10; Developmental Studies Hybridoma Bank) in PNT at 4°C overnight. Samples were washed in PT, incubated with goat anti-mouse Cy3 (1:2000l Jackson ImmunoResearch, West Grove, PA) for 2 hr at room temperature, washed in PT and mounted.

Genetics

Stocks used for microarray analysis have been previously described: HSP70-Gal4 (Bloomington), UAS-eya^WT, UAS-eya^D493N and UAS-eya^E728Q (Tootle et al., 2003). GMR-Gal4 (S. L. Zipursky) was used to drive eya expression in the eye disc for qRT-PCR analysis. For genetic interactions and in vivo expression analysis the following stocks were used in addition to those above: DPP-Gal4 (40C6) (Staehling-Hampton et al., 1994), UAS-eya^WT/Cyo actin-GFP; stg⁰²¹³⁵/TM3Ser actin-GFP, UAS-ey, eya²/ Cyo actin-GFP; DPP-Gal4/TM6B, eya²/Cyo actin-GFP; UAS-ey/TM6B, UAS-eya^WT, UAS-so^WT/Cyo actin-GFP, and UAS-eya^E728Q, UAS-so^WT/Cyo Dfd-YFP.

Results

Understanding the role of EYA activity in Drosophila eye development and the extent to which EYA's phosphatase activity influences transcriptional output requires knowledge of its downstream transcriptional targets. At the time this project was initiated, only a single EYA target gene, lozenge(lz), had been identified; therefore, we used a microarray-based approach to isolate additional genes with altered expression levels upon eya overexpression. In parallel, we profiled tissue in which an eya transgene mutant for phosphatase activity was overexpressed, reasoning that if phosphatase activity is required for transcriptional activity, then the patterns of altered gene expression would be quite different. Despite the inherent limitations of an overexpression based analysis, we selected this approach first because the only eya mutant alleles predicted to have reduced phosphatase activity behave as embryonic lethal nulls and therefore are not suitable for array analysis in the eye (Bui et al., 2000a; Mutsuddi et al., 2005), and second because the requirement for EYA phosphatase activity for normal eye development precludes us from setting up a genetic rescue context in which eya^MUT transgenes are expressed in a background lacking endogenous eya expression (Rayapureddi et al., 2003; Tootle et al., 2003).

Specifically, we compared the expression profile of wild type control tissue to that of tissue overexpressing an eya^WT transgene or one of two eya phosphatase mutant transgenes, eya^D493N and eya^E728Q, collectively referred to as eya^MUT (Li et al., 2003; Rayapureddi et al., 2003; Tootle et al., 2003), under the control of the HSP70-Gal4 driver in the adult head. Adult heads were selected both for the ease in isolating significant quantities of RNA, and because so, which encodes the DNA binding moiety of the EYA-SO transcription factor, is endogenously expressed (data not shown). Eya transgenes with similar expression levels were chosen based on quantification of protein levels on Western blots (Figure 1). Tissue was harvested three hours post-induction in order to maximize chances that resulting changes in gene expression would reflect direct transcriptional control, rather than downstream secondary effects. RNA was extracted from adult heads in duplicate for each genotype, processed and hybridized to Drosophila Affymetrix chips (Materials and Methods). Linear regression analysis demonstrated that the replicates of each genotype were highly similar, with control, eya^WT and eya^E728Q samples having R² values >0.99 and eya^D493N samples an R² value of 0.96.

Figure 1.

Eya transgenes are expressed at equivalent levels. Western blot of head lysates from flies overexpressing eya^WT, eya^D493N, and eya^E728Q.

Comparison of the eya^WT, eya^D493N and eya^E728Q expression profiles to the control sample resulted in a list of 577 genes that demonstrated a two-fold or greater change in expression with a t-test value of <0.05 in at least one of the experimental conditions. In order to determine if reducing phosphatase activity altered transactivation, we compared the expression level changes of the 577 genes in the eya^WT and eya^E728Q samples. For this comparative analysis we relaxed the p value to 0.10 and the fold change to ∼1.87 in order to survey a slightly larger data set. Using these parameters, out of the 577 genes with altered expression in at least one of our array conditions, 452 genes had statistically significant changes in gene expression in both the eya^WT and eya^E728Q. Of these genes, 80% were regulated similarly between the eya^WT and eya^E728Q samples, suggesting that reducing EYA phosphatase activity does not globally alter transcriptional output. These genes were also similarly regulated in eya^D493N samples, but due to the lower correlation coefficient, often had higher p values. Of the remaining 89 genes that were not similarly regulated among all three array conditions, in only 9 cases were the expression changes more similar between the two mutant samples than between either mutant to wild type. Because this latter pattern would be the expected signature of a gene whose transcriptional regulation depended on EYA phosphatase activity, our microarray-based survey suggests that loss of phosphatase activity does not broadly alter EYA's transcriptional output.

In silico identification of SO binding sites in putative EYA target genes

Given that EYA functions as a transcriptional activator (Ohto et al., 1999; Silver et al., 2003), we focused our analysis of potential targets on the 226 genes out of the initial set of 577 whose expression was upregulated in both eya^WT and eya^MUT samples. To identify potential direct targets of the EYA-SO complex, two consensus sequences were used to detect the presence of potential SO binding sites in genomic regions surrounding candidate targets: a short sequence, (T/C)GATA(T/C), identified through in vitro binding studies (Hazbun et al., 1997); and a longer consensus sequence, GTAAN(T/C)NGANA(T/C)(C/G) (N=any nucleotide), generated by comparison of SO binding sites in the Drosophila transcriptional targets lz (Yan et al., 2003) and so with sequences from targets of the mammalian Six family genes (Pauli et al., 2005).

Three independent computational strategies, all using the Genome Enhancer tool (Markstein et al., 2002), were implemented. We first searched for the presence of the short consensus site (Hazbun et al., 1997)(exact search parameters described in Materials and Methods). Due to the short and somewhat palindromic nature of this sequence, we searched for the presence of four binding sites within a 50 base pair window, but found only two candidate targets with predicted binding sites nearby. In a separate analysis we searched for one copy of the larger consensus sequence in regions proximal to genes upregulated in our microarray analysis, yielding an additional nine targets. Finally, considering the limitations of searching for such a short binding site and the fact that the longer consensus sequence was based only on a handful of Drosophila SO targets and is likely to evolve further, we developed a third approach based on the knowledge that binding sites for different transcription factors often cluster in cis-regulatory elements (Stanojevic et al., 1991). Specifically we searched for the short SO binding site (Hazbun et al., 1997) in combination with identified consensus binding sites for transcription factors downstream of canonical signaling pathways. Our search included the binding sites for the Wingless pathway effector TCF/LEF/PAN (van de Wetering et al., 1997), the Hedgehog effector GLI (Kinzler and Vogelstein, 1990), the Epidermal Growth Factor Receptor pathway ETS family effectors (Sharrocks et al., 1997), the JAK/STAT pathway effector STAT (Yan et al., 1996), the Notch pathway effector Suppressor of Hairless (Su(H)) (Rebeiz et al., 2002), and the Decapentapalegic (DPP) pathway effector MAD (Kim et al., 1997). Searching for these binding sites in combination with the SO binding site allowed us to search for fewer copies of the SO binding site and to cover a larger genomic window, thereby isolating an additional 32 potential candidate target genes. The combined results for all three searches produced a list of 43 potential direct EYA-SO targets that showed similar upregulation in the eya^WT and eya^MUT array samples (Supplementary Table S4). For many of our potential targets, more than one SO binding region was predicted proximal to the gene of interest, suggesting the possibility of complex or tissue-specific regulation.

qRT-PCR validation of EYA targets confirms impaired phosphatase activity has minimal impact on transcriptional output

We next implemented a number of secondary tests to distinguish relevant candidates from the inevitable background. First, in order to validate our microarray results independently, we performed quantitative real-time PCR (qRT-PCR) on cDNA derived from adult head tissue, prepared in the same manner as that used for the microarray analysis. Genes were chosen for analysis if they had an ∼three-fold change in expression and/or a potentially relevant molecular function. Using qRT-PCR we observed the upregulation of fourteen of seventeen tested target genes upon overexpression of eya^WT (Figure 2A and data not shown). Although the fold changes in expression were not identical between the microarray and qRT-PCR data, the trends in upregulation were consistent, confirming our microarray results.

Figure 2.

Overexpression of eya induces expression of putative target genes. (A) Quantitative RT-PCR analysis of target expression in adult heads overexpressing eya^WT and eya^MUT under control of the HSP-Gal4 driver compared to adult heads of flies carrying the driver alone. (B-D) Dac antibody staining in wing imaginal discs. (B) Wild type. (C) DPP-Gal4 driving coexpression of eya^WT and so transgenes. (D) Wing imaginal disc overexpressing eya^MUT and so under the control of the DPP-Gal4 driving coexpression of the phosphatase dead eya^EQ and so transgenes. (C-D) Dac expression is induced strongly by eya^WT and eya^MUT when overexpressed with so. (E) Quantitative RT-PCR of target expression in third instar larval eye-antennal imaginal discs overexpressing eya^WT under the control of the GMR-Gal4 driver compared to discs only carrying the driver alone.

As discussed above, our microarray data suggests that eya^MUT transgenes retain transcriptional activity, at levels similar to eya^WT. To further validate this result, we compared the ability of an eya^WT and an eya^MUT transgene to upregulate expression of the genes confirmed above by quantitative real-time PCR (Figure 2A). All ten targets tested were significantly upregulated in response to eya^MUT, consistent with the microarray data, and six showed statistically comparable fold activation upon expression of either eya^WT or eya^MUT transgenes. Two, mal and CG6560 showed a modest, less than two-fold reduction in induction by eya^MUT versus eya^WT transgenes, while stg and CG8211, exhibited a more striking ∼three-fold reduction in induction by eya^MUT relative to eya^WT transgenes. However, even in the latter two cases, upregulation of target gene expression by eya^MUT was significantly (three to four-fold) greater than the control (Figure 2A). These results suggest that the ability of an EYA phosphatase mutant to activate expression of target genes is overall quite similar to that of wild type EYA, but that there may be some context dependency, consistent with transcription assay data from cultured cells in which EYA^MUT activated some reporters as robustly as EYA^WT, but others to a lesser degree (Li et al., 2003; Tootle et al., 2003).

In addition, we compared the ability of eya^WT and eya^MUT transgenes co-overexpressed with so to induce ectopic expression of dac, a gene predicted to be a direct transcriptional target of EYA-SO, based on previous studies (Anderson et al., 2006; Chen et al., 1997; Kenyon et al., 2005; Pignoni et al., 1997). Both wild type EYA and the phosphatase mutant induced strong ectopic expression of dac when overexpressed with so (Figure 2B-D), demonstrating that EYA phosphatase activity is not required for the expression of this downstream EYA-SO target. Thus, loss of EYA phosphatase activity does not appear to globally or drastically alter direct downstream transcriptional output.

As one of the goals of this study was to identify transcriptional targets of EYA relevant to eye development, we asked if these targets could be similarly upregulated in response to overexpressing eya in the third instar eye-antennal imaginal disc. Using GMR-Gal4 to drive eya expression in all cells in and behind the morphogenetic furrow, we observed upregulation of twelve out of fourteen target genes by qRT-PCR (Figure 2E, Table 1 and data not shown). These data suggest our adult head microarray analysis was successful in identifying target genes that could also be regulated by EYA in the developing eye.

Table 1.

Predicted direct target of EYA and SO

Gene	Full name and predicted function	Fold change* WT/MUT
stg	string; phosphatase	2.72/2.24
mal	maroon-like; molybdopterin sulfurase	2.83/2.63
CG12030	Galactose metabolism	5.45/6.88
CG15879	Hydrolase fold	6.11/3.12
CG8449	Related to human RabGAP related protein	3.85/3.77
CG8211	novel	4.56/4.09
CG6560	GTPase activity	2.95/2.73
CG15203	novel	2.70/2.89
osbp	Oxysterol binding protein; cholesterol biosynthesis	2.94/3.01
CG17002	novel	2.54/2.62
Eip63E	Ecdysdone-inducible protein	14.83/15.19

^*Fold change is indicated for eya^WT and eya^E728Q samples.

In vivo and in vitro confirmation of predicted SO binding sites in EYA target genes

To distinguish direct EYA transcriptional targets from genes upregulated several steps downstream, we looked for the ability of SO, which contributes DNA binding activity to the EYA-SO transcription factor, to bind predicted target sequences both in vivo and in vitro. First, we investigated whether endogenous SO protein in the eye imaginal disc associated with chromatin regions encompassing predicted binding sites using chromatin immunoprecipitation (ChIP). Out of the ten candidate target genes examined, five, plus the postitive control, lz, showed enrichment of SO at one or more computationally predicted binding sites (Figure 3). CG8449 and CG15879 each had enrichment at a single predicted binding region, stg showed enrichment at each of its two predicted binding regions, while maroon-like (mal) and CG12030 showed significant occupancy at each of three predicted regulatory elements (Figure 3). We observed SO enrichment at binding sites predicted by all three computational strategies described above, demonstrating the efficacy of this approach in identifying potential targets. The association of endogenous SO with these genomic regions during the course of normal eye development suggests we have identified five new direct targets of EYA-SO transcriptional regulation, while the remaining five negatives could still be relevant direct targets in other developmental contexts.

Figure 3.

Chromatin immunoprecipitation analysis of SO association with genomic regions containing predicted SO binding sites. PCR amplified product obtained from input chromatin is in the first lane, from mock treated sample incubated with pre-immune serum is in the second lane, and from sample immunoprecipitated using the SO antibody is in the third lane. Chromatin enrichment in the IP sample over the mock treated sample is given below each panel.

Because our computational methods for identifying SO binding sites involved searching for clustered sequence motifs, most of the ChIP confirmed regions contained multiple potential SO binding sequences. In order to confirm SO binding to the predicted elements within these genomic regions and to determine how these sequences compare with the previously generated SO consensus binding motif, we performed electrophoretic mobility shift assays (EMSAs) using the two stg genomic regions bound by SO as a test case. In these experiments, purified recombinant full-length GST-SO fusion protein was tested for its relative ability to bind radio-labeled wild type oligonucleotide probes covering the predicted binding sites versus mutant probes that no longer matched the consensus sequence. Two predicted SO binding sites within the stg-A region and three probes covering five potential SO binding sites within the stg-B region were tested. While none of the mutant probes formed a complex with SO, all wild type probes demonstrated association with SO as evidenced by a strong mobility shift that was competed away by excess unlabeled wild type competitor DNA, but not by excess mutant competitor (Figure 4). However, SO did not bind to all probes with equal strength. Site stg-A2 showed strongest association with SO within the stg-A region, while stg-B3 was the most effective in complexing with SO in the stg-B region (Figure 4A). Assuming comparable affinities in vivo, these data predict that stg-A2 and stg-B3 should be the primary sites of SO recruitment to the stg genomic locus during eye development.

Figure 4.

SO binding analysis. (A) Electrophoretic mobility shift assays with recombinant GST-full-length SO on double-stranded oligonucleotides containing predicted wild type or mutant SO binding sites. (B) Alignment of predicted SO binding sites highlights the core consensus sequence.

To determine if the strength of SO binding could be correlated with a specific sequence, we aligned the probe sequences and looked for conserved motifs. From this analysis, only the short SO binding site (T/C)GATA(T/C) (Hazbun et al., 1997) emerged (Figure 4B). Direct comparison of this sequence and other binding sites identified in this study to the longer sequence GTAAN(T/C)NGANA(T/C)(C/G)(N=any nucleotide) (Pauli et al., 2005) revealed no obvious conservation of flanking sequences, suggesting the core sequence requirement for SO binding is (T/C/G)GA(A/T/G)A(T/C). Thus, how flanking sequences contribute to regulatory specificity in vivo remains an open question.

Stg is an in vivo target of EYA-SO

In order to determine if EYA and SO regulation of stg expression is required during development, we first tested the ability of eya and stg to interact genetically. Specifically, we took advantage of the fact that misexpression of EYA can induce ectopic eye formation (Bonini et al., 1997; Pignoni et al., 1997). In this assay, modulating the expression level of a gene predicted to cooperate with eya in eye specification can be tested for its ability to dominantly modify the frequency of ectopic eye induction (Hsiao et al., 2001). Therefore, we examined the frequency of EYA ectopic eye induction when one copy of stg was removed. Heterozygosity for stg resulted in a 40% decrease in the frequency of ectopic eye induction by EYA (Figure 5), suggesting that eya and stg function cooperatively during the specification of eye tissue.

Figure 5.

Heterozyosity for stg reduces the frequency of ectopic eye induction upon eya misexpression under the DPP-Gal4 driver.

We next examined the ability of EYA and SO to regulate stg expression in vivo. First, using a series of stg-lacZ enhancer trap lines, two of which, pstg β-E4.9 and pstgβ-E6.7, contained our identified SO binding sites (Lehman et al., 1999), we tested the ability of DPP-Gal4 driven EYA and SO to induce lacZ expression in the wing imaginal disc. No induction in lacZ expression in the region where EYA and SO are overexpressed was observed (data not shown). Examination of eight enhancer trap lines spanning the 5'UTR of stg (Lehman et al., 1999) also failed to reveal an expression pattern consistent with that of stg in the eye imaginal disc (Alphey et al., 1992)( data not shown). Together these results suggest that stg expression requires input from multiple cis-regulatory regions and multiple transcription factors. This is consistent with other studies that have shown stg expression is directly regulated by the epidermal growth factor transcriptional regulators, Pointed and Tramtrack69 (Baonza et al., 2002), and directly or indirectly by Notch and Wingless signaling (Deng et al., 2001; Johnston and Edgar, 1998).

Therefore, as an alternative to in vivo reporter analysis, we overexpressed eya and so singly or in combination using a DPP-Gal4 driver in third instar wing and antennal imaginal discs, both tissues that normally lack expression of these genes, and followed stg expression by in situ hybridization. While overexpression of either eya or so alone did not induce stg (data not shown), co-overexpression of both genes resulted in increased stg along the dorsal/ventral margin of the wing and in the ventral part of the antennal disc (Figure 6A-D).

Figure 6.

EYA and SO regulation of stg expression. (A-H) In situ hybridization for stg. Wild type eyeantennal (A) or wing imaginal disc (B) show strong stg. Eye-antennal (C) or wing (D) imaginal discs overexpressing eya^WT and so under the control of the DPP-Gal4 driver. In both tissues stg expression is induced in the region where eya and so are being overexpressed. Eye-antennal (E) and wing (F) imaginal discs in which eya^MUT and so are overexpressed under the control of the DPP-Gal4 driver demonstrate ectopic stg expression. Arrows in C-F indicate regions of increased stg expression. Overexpression of ey using a DPP-Gal4 driver induces stg expression in eye-antennal (G) and wing imaginal discs (H). Discs overexpressing ey, but homozygous for the eya² allele no longer exhibit induction of stg expression (I, J). In eya²/eya²; DPP-Gal4/ UAS-ey, wing discs significant folding is observed, which results in the darker stg staining in the region of ey overexpression but does not reflect an overall increase in stg expression.

Using the same assay, we next examined the ability of the eya^MUT to activate stg expression when coexpressed with so. Similar to eya^WT, eya^MUT also induced ectopic stg expression in both the wing and antennal discs (Figure 6E-F). The results support our earlier microarray and qPCR data, suggesting that reducing EYA phosphatase activity does not severely compromise its transcriptional ability. However, the overproliferation and resulting tissue distortion observed upon coexpression of so and eya^MUT (Figure 6C-D) was much less pronounced than that observed with so and eya^WT (Figure 6E-F), suggesting that although the immediate transcriptional output might not be significantly compromised, reduced phosphatase activity can alter the downstream developmental program.

Although the above results are consistent with EYA and SO directly regulating stg expression, overexpression of ey, itself a transcriptional target of EYA-SO (Pauli et al., 2005; Pignoni et al., 1997), has also been shown to result in upregulation of stg (Michaut et al., 2003; Ostrin et al., 2006). Thus it was formally possible that EYA-SO mediated regulation of stg might occur indirectly through EY. To rule out this possibility, we compared the ability of EY to induce stg expression in wild type wing and antennal imaginal discs versus discs homozygous for the eye-specific allele eya². This allele deletes 0.3 kb of an eye enhancer element through which EY regulates eya expression (Zimmerman et al., 2000). As expected based on the results of Michaut and colleagues (Michaut et al., 2003), overexpression of ey using a DPP-Gal4 driver resulted in increased stg expression in wing and antennal discs (Figure 6G,H). Antibody staining revealed induction of eya and so expression in these tissues (data not shown), suggesting that stg upregulation could be directed by EYA and SO. In contrast, in discs overexpressing ey, but lacking eya, we no longer observed stg induction in wing or antennal discs, indicating that EYA and SO function downstream of EY to directly regulate stg transcription (Figure 6G-J).

The ability of EYA and SO to induce ectopic stg expression demonstrates that EYA and SO are sufficient for stg expression, and the fact that EY cannot induce stg in the absence of eya suggests eya is also necessary. Further supporting the argument that eya is necessary for stg expression, loss of eya also compromises the normal pattern of stg expression at the morphogenetic furrow (Figure 6A, I), demonstrating that eya is required for stg expression during eye development. However, we still observe tissue overgrowth in wing discs overexpressing ey and mutant for eya (Figure 6I, J), indicating that EY may act through additional parallel pathways to influence cell proliferation and organ size. This may also be the case for EYA and SO, as heterozygosity for stg was unable to rescue tissue overgrowth in wing discs overexpressing eya and so (data not shown). These observations are consistent with previous data demonstrating that overexpression of stg alone does not increase cell proliferation (Neufeld et al., 1998). Thus, the ability of EYA and SO to induce stg outside of its normal expression region, coupled with our demonstration that eya is necessary for stg expression in the eye imaginal disc, suggest a direct mechanism through which EYA and SO might regulate cell cycle progression during eye development.

Discussion

We have used microarray analysis combined with computational and experimental validation to identify potential EYA-SO transcriptional targets. Two general conclusions have resulted from this work: first, the similarity in expression profiles between tissue overexpressing wild type and phosphatase-dead eya transgenes and analysis of gene expression by quantitative PCR suggest that EYA's phosphatase is not generally required for EYA transcriptional activity, although it may be required for maximal transactivation of some target genes; and second, the short sequence (T/C/G)GA(A/T/G)A(T/C) appears to be the only recognizable motif shared among all SO binding sites. As exemplified by our in vivo validation of EYA-SO-mediated regulation of the cell cycle regulator stg, further analysis of the target genes identified in this study will likely shed new light into the mechanisms underlying EYA-SO function during development.

Identification of New EYA-SO Targets

The main goal of this study was to identify new targets of EYA transcriptional activity. Although we used adult head tissue for our overexpression experiments, our 86% success rate in confirming changes in expression of our potential targets in developing Drosophila eye-antennal imaginal discs overexpressing eya supports the ability of this system to identify similar data sets in different developmental stages. Out of the ten genes upregulated by eya overexpression in both adult head tissue and eye-antennal imaginal discs, five demonstrated enrichment of endogenous SO at one or more predicted binding sites. These predicted binding regions were conserved across a minimum of two and up to nine Drosophila species, emphasizing their likely biological relevance (data not shown; see methods for details). Two binding sites that did not demonstrate SO enrichment were not conserved across other Drosophila species, while binding sites in the remaining three genes were conserved across multiple species and could be EYA-SO targets in other tissues.

The core sequence shared by all of these targets is (T/C/G)GA(A/T/G)A(T/C), a pared down version of the previously proposed GTAAN(T/C)NGANA(T/C)(C/G) SO binding sequence (Pauli et al., 2005). In Drosophila EYA-SO targets, the sequence flanking the core (T/C/G)GA(A/T/G)A(T/C) has only been shown to be important in the case of the target so (Pauli et al., 2005), and is absent in one of the two SO binding sites identified in the lz locus (Yan et al., 2003) and the binding sites in stg we confirmed by gel shifts. While specific flanking sequences may further stabilize SO-DNA interactions, characterization of such a flanking sequence consensus awaits further analysis. Confirmation of additional targets predicted by our microarray and binding site analysis should provide for further characterization of the SO binding sequence. Out of the remaining 31 potential targets, all except one have binding sites conserved across multiple Drosophila species, suggesting that additional EYA-SO targets will be confirmed within this data set.

While none of the previously identified EYA-SO targets were included in our final list, two targets, so and lz were upregulated upon eya overexpression, although less than our two-fold cutoff. The expression of the previously identified targets hh, ato and ey was either absent or changes were not statistically significant. One explanation for this observation is that other signaling pathways required for the expression of these genes may not be activated, or, conversely, inhibitory signaling pathways could be activated in adult head tissue.

In addition to examining expression levels of previously identified EYA-SO targets, we also compared our list of upregulated EYA-SO target candidates to genes that were upregulated by ey overexpression in microarray analyses (Halder et al., 1998; Niimi et al., 1999; Ostrin et al., 2006; Pauli et al., 2005; Punzo et al., 2002). Because ey both induces eya and so expression and is itself transcriptionally regulated by EYA-SO, one would expect to see a number of genes similarly regulated by overexpression of either ey or eya; however, as detailed below, pairwise comparisons between our data set to lists of candidate ey targets derived from two independent array studies (Michaut et al., 2003; Ostrin et al., 2006), reveals a surprisingly limited overlap. Michaut et al. identified 371 genes with at least 1.5-fold upregulation across two array experiments, only 55 of which were similarly upregulated in both arrays (Michaut et al., 2003). Comparison of this data set to a more recent report by Ostrin et al. of 300 candidate genes upregulated in response to ey overexpression revealed only 24 common targets. Comparison to our list of potential eya-so targets yielded 10 shared with the Michaut et al. data set and 3 common to the Ostrin et al. results. Encouragingly, despite this limited overlap, stg, a gene we have shown here to be transcriptionally regulated by eya and so, was one of the two targets consistently upregulated in all three studies (Michaut et al., 2003; Ostrin et al., 2006).

As we continue to confirm additional targets, it is important to note that EYA may also associate with transcription cofactors other than SO to regulate gene expression. Although EYA can associate with DAC (Chen et al., 1997), and X-ray crystallographic analysis suggests DAC can bind DNA (Kim et al., 2002), targets of an EYA-DAC complex or a consensus DAC binding site have not been identified. In addition, EYA also contains an engrailed homology 1 (eh1) domain (Goldstein et al., 2005), suggesting it may be able to bind to the transcriptional repressor Groucho (GRO). However, as current in vivo data only supports a role for EYA as a transactivator complexed to SO, identification of additional EYA cofactors in vivo will be necessary to explore the potential of SO-independent EYA transcriptional functions further.

Interplay between EYA transcription and phosphatase activities

Previous studies using transcriptional reporter assays in transfected cultured cells have yielded conflicting results as to whether phosphatase activity is important for proper transcriptional output from EYA-SO (Li et al., 2003; Tootle et al., 2003). However, the emerging consensus from these experiments is that missense mutations in the ED that reduce phosphatase activity also compromise the ability of EYA to transactivate targets (Li et al., 2003; Mutsuddi et al., 2005). Importantly, these mutations do not affect SO binding (Mutsuddi et al., 2005; Tootle et al., 2003), suggesting it is actually loss of EYA phosphatase activity that alters transcriptional output, presumably as a consequence of improper phosphoregulation of critical substrates, although altered protein-protein interactions cannot be ruled out. How faithfully these cultured cell based reporter assays mirror in vivo regulation remains an open question.

In order to examine further the relationship between EYA's dual functions as transcriptional coactivator and protein tyrosine phosphatase in vivo, we compared the expression profiles of tissue overexpressing wild type versus phosphatase mutant eya transgenes. We expected that if phosphatase activity is required for transactivation, then direct transcriptional targets showing altered expression in eya^WT samples would exhibit reduced changes in eya^MUT samples. Our microarray results demonstrated that ∼80% of a set of 577 candidate targets were similarly regulated among eya^WT and eya^MUT samples, supporting the alternate hypothesis that phosphatase activity is not broadly required for EYA's transcriptional activity. Only 2%, or 9 genes, exhibited a consistent change in expression in response to the two eya^MUT transgenes that differed from the response to eya^WT. Our follow-up analysis of a subset of these genes by quantitative PCR confirmed that an EYA phosphatase mutant has transcriptional activity quite similar to wild type EYA. In the cases in which phosphatase mutant EYA induced expression at a significantly lower level that wild type EYA, one hypothesis would be that those genes might represent indirect targets; arguing against this, stg, one of two genes whose pattern of induction fell into this category, was confirmed as a direct EYA-SO target. In addition, ectopic expression of dac, a likely direct downstream EYA-SO target, was induced by the eya^MUT as efficiently as eya^WT. In conclusion, while mutating EYA phosphatase activity has a minimal impact on the direct downstream transcriptional profile overall, at certain target genes it may reduce, but not eliminate transcriptional output. Given these results, future exploration of in vivo roles for EYA phosphatase activity independent of its transcriptional functions may be warranted.

Disease and developmental implications of EYA-SO targets

Many of the genes identified as direct EYA-SO transcriptional targets are largely uncharacterized “CGs” whose expression patterns in the eye will have to be studied in detail to gain further insight to EYA-SO-mediated regulation, but a few have predicted or well-studied functions that may provide insight into how EYA-SO functions during normal development and how misregulation can result in disease. Most notable on this list is stg. Given that overexpression of eya and so results in overproliferation, while their loss leads to tissue reduction (Bonini et al., 1997; Pignoni et al., 1997), EYA-SO control of stg expression provides a mechanism for how EYA-SO regulation of the cell cycle may in turn affect cell proliferation. An interesting question for future investigation is how the relatively broad expression of EYO and SO throughout the developing retina activates stg expression only in a relatively narrow stripe of cells just anterior to the morphogenetic furrow. Given the apparent complexity of stg cis-regulatory elements revealed by our and previous studies (Baonza et al., 2002)(Deng et al., 2001; Johnston and Edgar, 1998), a likely explanation is that EYA-SO act combinatorially with transcriptional effectors of other signaling pathways to effect this developmental precision.

Consistent with eya and so overexpression leading to increased tissue overgrowth in Drosophila (Bonini et al., 1997; Pignoni et al., 1997), elevated levels of Eya and Six family members have been observed in a variety of cancers (Coletta et al., 2004; Ford et al., 1998; Yu et al., 2006; Zhang et al., 2005). Studies of the transcriptional targets of mammalian Eya and SO/Six proteins have identified the cell cycle regulatory genes, cyclin D1 and cyclin A1 (Coletta et al., 2004; Yu et al., 2006), the proto-oncogene c-Myc (Li et al., 2003) and ezrin (Yu et al., 2004), a regulator of the cytoskeleton and contributor to metastasis, suggesting intermediates through which Eya and Six family genes regulate proliferation and contribute to cancer. Identification of stg as a transcriptional target of EYA and SO in Drosophila provides not only the first direct cell cycle target in Drosophila, but also suggests another target through which EYA and SO might regulate proliferation in other organisms.

Before we can draw parallels to how EYA-SO targets important for Drosophila retinal development might be relevant to development and disease in other organisms, it will be necessary to examine the conservation of the transcriptional regulatory circuits. However, given the predicted functions of the gene products encoded by our candidate EYA-SO targets, together with knowledge of Eya-Six function in mammalian systems, it is tempting to speculate. For example, CG12030, the Drosophila homolog of the human Gale, encodes a sugar epimerase required for galactose metabolism (predicted by Interpro (Apweiler et al., 2000)). As metabolic abnormalities have been demonstrated to play a part in cataract formation, and mutations in eya have been observed in patients with congenital cataracts (Azuma et al., 2000), the identification of CG12030 as an EYA-SO target suggests intermediates through with eya might function to maintain homeostasis in the eye. Mal, which encodes a molybdenum cofactor sulfurase important for ommochrome biosynthesis, is expressed in Drosophila pigment cells in the eye (Amrani et al., 2000) and would seem a logical target of the RDGN. Mutations of the human homolog of mal, HMCS, can result in renal failure and myositis (Ichida et al., 2001), both intriguing phenotypes given the importance of Eya-Six in vertebrate kidney and muscle development (Abdelhak et al., 1997; Grifone et al., 2005; Heanue et al., 1999; Laclef et al., 2003; Li et al., 2003; Ozaki et al., 2001; Xu et al., 1999; Xu et al., 2003). CG15879 encodes the Drosophila homolog of human SERHL2, a member of a serine hydrolase-like family predicted to regulate muscle growth, a developmental context in which Eya and Six family genes function in Drosophila and vertebrates (Clark et al., 2006; Grifone et al., 2005; Heanue et al., 1999; Laclef et al., 2003; Li et al., 2003). Lastly, CG8449 has a predicted RabGAP/TBC domain. While RabGAPs function in a variety of developmental contexts, RabGAP-like proteins have been predicted to function in phototransduction and synaptic transmission in Drosophila (Xu et al., 1998) and mutations in RabGAP genes have been isolated in cases of Warburg Micro syndrome (Aligianis et al., 2005), a severe autosomal recessive disorder characterized by abnormalities in the eye, as well as the central nervous system and genitals, all contexts in which Drosophila eya and so are expressed (Bonini et al., 1998; Boyle et al., 1997; Cheyette et al., 1994; Fabrizio et al., 2003). Identification of additional EYA-SO targets and the examination of the conservation of EYA-SO transcriptional regulation across homologous genes in different species will be necessary to determine how EYA and contribute to development and disease.

Given the importance of achieving appropriate levels of gene expression during the course of development, it is not surprising that multiple signaling pathways converge to regulate common target genes at the level of transcription. For example, hh and lz are coordinately regulated by receptor tyrosine kinase (RTK) downstream effectors of the ETS family (Behan et al., 2002; Rogers et al., 2005) in conjunction with EYA and SO. Here, we have identified stg as an EYA-SO target, and work by us and others suggests stg transcription is also regulated Notch and Wingless (Wg) signaling (Deng et al., 2001; Johnston and Edgar, 1998) , and by RTK signaling (Baonza et al., 2002). Thus, our results suggest a mechanism by which members of the RDGN are integrated with Notch and Wg signaling to coordinate cell proliferation. Identification of additional EYA-SO targets is likely to reveal new nodes for integration of the RDGN with other signaling pathways, explaining how signaling pathways cooperate to yield specific developmental outcomes.