sine oculis, a member of the six family of transcription factors, directs eye formation

16
Genomes & Developmental Control Sine oculis, a member of the SIX family of transcription factors, directs eye formation Brandon Weasner, Claire Salzer, Justin P. Kumar Department of Biology, Indiana University, 1001 E. 3rd Street, Bloomington, IN 47405, USA Received for publication 7 September 2006; revised 18 October 2006; accepted 19 October 2006 Available online 1 November 2006 Abstract The initiation of eye formation in all seeing animals is controlled by a group of selector genes that together forms the retinal determination cascade. In Drosophila, mice and humans, loss-of-function mutations lead to defects in eye and/or head development. While ectopic expression of these genes is sufficient to direct non-retinal tissues towards an eye fate, the ability of each gene to initiate eye formation is neither unlimited nor equal. A particularly enigmatic observation has been that one member of the cascade, sine oculis (so), which is a member of the SIX family of homeobox transcription factors, is unable to initiate eye development in non-retinal tissues. It is in contrast to every other retinal determination gene including optix, another Six family member, which can induce eye formation when expressed on its own. Here we demonstrate that, in contrast to published reports, expression of so on its own is sufficient to induce eye development within non-retinal tissues. We have extended results from prior reports on binding partner selectivity and DNA binding sites by conducting a structure/function analysis of the SO and OPTIX proteins. Here we demonstrate that the SIX domains and C-terminal portions of the SO and OPTIX proteins are required for functional specificity of SIX class transcription factors while the homeodomain of these proteins are interchangeable. Taken together, these results shed new light on the role that so plays in eye specification. © 2006 Elsevier Inc. All rights reserved. Keywords: Drosophila; Eye; Retina; Sine ocuilis; Optix Introduction The ability to specify the fate of specialized tissues and organs is a fundamental requirement of all metazoans and involves the use of specialized networks of selector genes. A well-studied example is the developing compound eye of the fruit fly Drosophila melanogaster, which is controlled by the concerted activity of eight genes that comprise the eye specification or retinal determination cascade. These genes include eyeless (ey), twin of eyeless (toy), eyegone (eyg), twin of eyegone (toe), sine oculis (so), optix, eyes absent (eya) and dachshund (dac)(Bonini et al., 1993; Cheyette et al., 1994; Czerny et al., 1999; Jang et al., 2003; Mardon et al., 1994; Quiring et al., 1994; Seimiya and Gehring, 2000; Serikaku and O'Tousa, 1994). These genes have a special role during eye development in flies as witnessed by the complete absence of eye tissue in loss-of-function mutants and the redirection of non- retinal tissues towards an eye fate in forced expression experiments (Bonini et al., 1997; Czerny et al., 1999; Halder et al., 1995; Seimiya and Gehring, 2000; Shen and Mardon, 1997; Yao and Sun, 2005). It should be noted that loss-of- functions do not exist for all eight genes (optix and toe) and their place in the retinal determination cascade is based mainly on their expression pattern and ability to induce ectopic eye development. Two additional genes that play key roles in eye specification are teashirt (tsh) and homothorax (hth). Loss of either gene leads to defects in retinal specification while ectopic expression of tsh can, in limited circumstances, induce eye formation (Bessa and Casares, 2005; Bessa et al., 2002; Pai et al., 1998; Pan and Rubin, 1998; Pichaud and Casares, 2000; Singh et al., 2002). These genes are thought to have multiple roles in retinal development. Early on, both genes are thought to function with ey to promote eye specification followed by a switch in activity in which eye development is repressed at later stages (Bessa and Casares, 2005; Bessa et al., 2002; Singh et al., 2002). Developmental Biology 303 (2007) 756 771 www.elsevier.com/locate/ydbio Corresponding author. Fax: +1 812 856 1566. E-mail address: [email protected] (J.P. Kumar). 0012-1606/$ - see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.ydbio.2006.10.040

Upload: brandon-weasner

Post on 30-Oct-2016

213 views

Category:

Documents


0 download

TRANSCRIPT

Developmental Biology 303 (2007) 756–771www.elsevier.com/locate/ydbio

Genomes & Developmental Control

Sine oculis, a member of the SIX family of transcription factors,directs eye formation

Brandon Weasner, Claire Salzer, Justin P. Kumar ⁎

Department of Biology, Indiana University, 1001 E. 3rd Street, Bloomington, IN 47405, USA

Received for publication 7 September 2006; revised 18 October 2006; accepted 19 October 2006Available online 1 November 2006

Abstract

The initiation of eye formation in all seeing animals is controlled by a group of selector genes that together forms the retinal determinationcascade. In Drosophila, mice and humans, loss-of-function mutations lead to defects in eye and/or head development. While ectopic expression ofthese genes is sufficient to direct non-retinal tissues towards an eye fate, the ability of each gene to initiate eye formation is neither unlimited norequal. A particularly enigmatic observation has been that one member of the cascade, sine oculis (so), which is a member of the SIX family ofhomeobox transcription factors, is unable to initiate eye development in non-retinal tissues. It is in contrast to every other retinal determinationgene including optix, another Six family member, which can induce eye formation when expressed on its own. Here we demonstrate that, incontrast to published reports, expression of so on its own is sufficient to induce eye development within non-retinal tissues. We have extendedresults from prior reports on binding partner selectivity and DNA binding sites by conducting a structure/function analysis of the SO and OPTIXproteins. Here we demonstrate that the SIX domains and C-terminal portions of the SO and OPTIX proteins are required for functional specificityof SIX class transcription factors while the homeodomain of these proteins are interchangeable. Taken together, these results shed new light on therole that so plays in eye specification.© 2006 Elsevier Inc. All rights reserved.

Keywords: Drosophila; Eye; Retina; Sine ocuilis; Optix

Introduction

The ability to specify the fate of specialized tissues andorgans is a fundamental requirement of all metazoans andinvolves the use of specialized networks of selector genes. Awell-studied example is the developing compound eye of thefruit fly Drosophila melanogaster, which is controlled by theconcerted activity of eight genes that comprise the eyespecification or retinal determination cascade. These genesinclude eyeless (ey), twin of eyeless (toy), eyegone (eyg), twin ofeyegone (toe), sine oculis (so), optix, eyes absent (eya) anddachshund (dac) (Bonini et al., 1993; Cheyette et al., 1994;Czerny et al., 1999; Jang et al., 2003; Mardon et al., 1994;Quiring et al., 1994; Seimiya and Gehring, 2000; Serikaku andO'Tousa, 1994). These genes have a special role during eyedevelopment in flies as witnessed by the complete absence of

⁎ Corresponding author. Fax: +1 812 856 1566.E-mail address: [email protected] (J.P. Kumar).

0012-1606/$ - see front matter © 2006 Elsevier Inc. All rights reserved.doi:10.1016/j.ydbio.2006.10.040

eye tissue in loss-of-function mutants and the redirection of non-retinal tissues towards an eye fate in forced expressionexperiments (Bonini et al., 1997; Czerny et al., 1999; Halderet al., 1995; Seimiya and Gehring, 2000; Shen and Mardon,1997; Yao and Sun, 2005). It should be noted that loss-of-functions do not exist for all eight genes (optix and toe) and theirplace in the retinal determination cascade is based mainly ontheir expression pattern and ability to induce ectopic eyedevelopment. Two additional genes that play key roles in eyespecification are teashirt (tsh) and homothorax (hth). Loss ofeither gene leads to defects in retinal specification while ectopicexpression of tsh can, in limited circumstances, induce eyeformation (Bessa and Casares, 2005; Bessa et al., 2002; Pai et al.,1998; Pan and Rubin, 1998; Pichaud and Casares, 2000; Singh etal., 2002). These genes are thought to have multiple roles inretinal development. Early on, both genes are thought to functionwith ey to promote eye specification followed by a switch inactivity in which eye development is repressed at later stages(Bessa and Casares, 2005; Bessa et al., 2002; Singh et al., 2002).

757B. Weasner et al. / Developmental Biology 303 (2007) 756–771

Within the mammalian eye, homologous genes are expressedand loss-of-function mutations often lead to head and retinaldefects in mouse model systems and human patients (Azuma etal., 2000; Gallardo et al., 1999; Glaser et al., 1994; Hill et al.,1991; Klesert et al., 2000; Pasquier et al., 2000; Sarkar et al.,2000; Wallis et al., 1999; Winchester et al., 1999; Xu et al.,1997). As seen in flies, forced expression invertebrates can leadto the formation of ectopic eyes (Altmann et al., 1997; Chow etal., 1999; Hanson, 2001; Loosli et al., 1999; Onuma et al., 2002;Traboulsi, 1998; Zuber et al., 2003). Excitingly, the introductionof mouse Pax6 and Eya2 into files is sufficient to induce ectopicretinal development in a variety of imaginal discs and to restoreeye development to flies harboring loss-of-function mutations(Bonini et al., 1997; Halder et al., 1995). These experimentshave also worked in the reciprocal manner in which fly eyelessmRNA has been introduced into Xenopus and ectopic eyes havebeen induced (Onuma et al., 2002).

While the retinal determination genes may govern theearliest decisions in eye formation, the ability of each gene todirect the initiation of eye development is neither equal norunlimited. For example, ey is the most potent initiator of retinaldevelopment; ectopic expression of ey can transform the fatesof most post-embryonic tissues including the antenna, legs,wings and halteres (Halder et al., 1995). At the other extremelies so, which reportedly cannot initiate eye development on itsown in any tissue (Chen et al., 1997; Pignoni et al., 1997). Theabilities of the remaining retinal determination genes residesomewhere between these two extremes and are limited tovarying degrees. For instance, eya, on its own, can direct eyedevelopment at a relatively low frequency when compared to eyjust within the antenna and leg imaginal discs (Bonini et al.,1997; Yao and Sun, 2005). Interestingly, EYA forms abiochemical complex with SO in vitro and forced co-expressionof these two genes results in a synergistic increase in thefrequency and size of ectopic eyes (Pignoni et al., 1997;Seimiya and Gehring, 2000). The current model is that SO canbind to DNA through its homeodomain and EYA, not havingDNA binding properties of its own, binds to SO and serves as atranscriptional co-activator for downstream target genes(Pignoni et al., 1997). The recent demonstration that EYA is aprotein tyrosine phosphatase (Li et al., 2003; Rayapureddi et al.,2003; Tootle et al., 2003) provides an attractive mechanism inwhich EYA regulates SO phosphorylation states during eyedevelopment and this may be mechanistically important for eyespecification. Equally feasible is for EYA to work on additionaltarget proteins. The identification of EYA substrates will beimportant in distinguishing among these possibilities.

The inability of so to initiate eye development on its own isintriguing, in part, because this deficiency appears to be specificto this retinal determination gene. In Drosophila, there are threemembers of the SIX family of homeobox transcription factors:so, optix and DSix4 (Cheyette et al., 1994; Kawakami et al.,2000; Seimiya and Gehring, 2000; Seo et al., 1999; Serikakuand O'Tousa, 1994). Although so lacks the ability to induce eyeformation, expression of optix is sufficient to initiate retinaldevelopment in the antennal disc (Seimiya and Gehring, 2000).The apparent differences in initiating eye development in non-

retinal tissues suggest that these genes may also play verydistinct roles in normal eye development, a view which issupported, in part, by differences in the expression patterns ofso and optix in the developing eye imaginal disc; optixexpression is restricted to cells ahead of the morphogeneticfurrow while so is expressed throughout the eye field (Cheyetteet al., 1994; Seimiya and Gehring, 2000; Serikaku and O'Tousa,1994). Dsix4 is not expressed in the retina and is therefore notthought to play a role in eye formation but instead functions inmuscle and gonad development (Kirby et al., 2001).

As with all SIX class transcription factors, SO and OPTIXeach contain a homeodomain (HD) for DNA binding.Recently, a consensus binding site for SO has been identifiedand a genome-wide search has uncovered an auto-regulatoryloop in which SO protein binds to an enhancer element withinthe so gene itself (Pauli et al., 2005). Additionally, amonggenes that are already known to function in eye development,SO binding sites are present within the eyeless (ey), lozenge(lz) and hedgehog (hh) genes (Pauli et al., 2005; Yan et al.,2003). These reports serve to link so to genes that promote eyespecification ahead of the morphogenetic furrow (so, ey) andto those that participate in cell fate decisions behind the furrow(Pauli et al., 2005; Yan et al., 2003). Additionally, so is alsolinked to regulating signaling pathways that communicateinstructions across the furrow (Pauli et al., 2005). Studies ofthe mammalian Six1, Six2, Six4 and Six5 genes, whichrepresent two of the three Six subfamilies, have revealedsimilar binding properties for each of the encoded transcriptionfactors (Harris et al., 2000; Kawakami et al., 1996a,b; Suzuki-Yagawa et al., 1992). While binding sites for optix and themammalian Six3 and Six6 homologs are yet to be identified, itis possible that they will also be similar to that of the otherfamily members. This would suggest that the specificity inactivity lies elsewhere in the protein, most likely within theprotein interaction domain.

SO and OPTIX are also characterized as having a SIX-domain (SD) for protein–protein interactions. Recent identifi-cations of binding partners have provided an attractivebiological basis that may explain differences in the functionsof SIX family members. For example, the transcriptional co-repressor GROUCHO (GRO) binds both SO and OPTIX(Kenyon et al., 2005a; Silver et al., 2003). On the other hand,EYA and the novel protein SBP (SO binding protein) interactsstrongly with SO (EYA binds very weakly to OPTIX) while thezinc finger containing protein OBP (OPTIX Binding Protein)binds preferentially to OPTIX (Kenyon et al., 2005a; Pignoni etal., 1997; Seimiya and Gehring, 2000). Understanding thebiological consequences of the differences in binding partnerselection will be a key step towards elucidating how SIX classtranscription factors differ functionally.

In this report, we have set out to answer two questions. First,what is the biological basis for the limitations in inducingectopic eye formation by SO and OPTIX? In other words, whatrestricts the initiation of ectopic eye development by OPTIX tothe antenna and why is SO completely incapable of inducingeye formation? Using a set of GAL4 driver lines, wedemonstrate that the expression of so, on its own, is sufficient

Table 1Induction of ectopic eyes by SO and OPTIX

Phenotype Sine oculis OPTIX

No. of G4 lines No. of G4 lines

No effect 139 78Embryonic lethal 53 83Larval lethal 0 26Pupal lethal 11 16Eye defects 4 8Wing defects 7 3Ectopic eyes 4 4Leg defects 3 1

GAL4 expression patterns that yield ectopic eyes when so is expresseddpp-GAL4 (antenna)cb41-GAL4 (antenna)cb49-GAL4 (antenna, ventral head)sd-GAL4 (ventral head)

758 B. Weasner et al. / Developmental Biology 303 (2007) 756–771

to initiate eye development in non-retinal tissues. We furthershow that expression of so induces the expression of two otherretinal determination genes eya and dac. And finally, we willpresent evidence that only a subset of cells expressing so, eyaand dac can be coaxed into adopting a retinal fate. Our secondquestion focused on which regions of the SIX proteins mediatefunctional differences in the roles that SO and OPTIX playduring normal eye development? We followed up on publishedreports of binding partner specificity and DNA binding targetsequences with functional tests for the support of normal eyedevelopment and the induction of ectopic eyes in non-retinaltissues. Using a set of SO deletion and SO/OPTIX chimericproteins, we demonstrate that the SIX and C-terminal regions ofthe proteins confer specificity while the homeodomains arelikely to bind to common target genes and are to some degreeinterchangeable. We believe that these findings shed new lighton the role that so plays in eye specification.

GAL4 expression patterns that yield ectopic eyes when optix is expresseddpp-GAL4 (antenna; Shen and Mardon, 1997)rn-GAL4 (wing, haltere)cb49-GAL4 (antenna, ventral head)c309-GAL4 (antenna, ventral head)c253-GAL4 (antenna)

Experimental procedures

Fly stocks

The following fly stocks were obtained for the experiments performed in thisreport: ey-GAL4 (gift of Walter Gehring), dpp-GAL4 (gift of Janice Fischer),GMR-GAL4 (gift of Lucy Cherbas), Bloomington Stock Center GAL4collection v05.06.29, UAS-GFP, sev-GAL4, optix[P]/CyO (BloomingtonStock Center).

Microscopy

The following reagents were used in this report: mouse anti-DACHSHUND(1:5, DSHB), mouse anti-EYES ABSENT (1:5, DSHB); mouse anti-ELAV(DSHB); goat anti-mouse TRITC (1:100, Jackson Laboratories), goat anti-mouse FITC (1:100, Jackson Laboratories); goat anti-rat FITC (1:100, JacksonLaboratories); goat anti-rat TRITC (1:100, Jackson Laboratories); andphalloidin-TRITC (1:100, Molecular Probes). Third instar imaginal discs wereprepared for confocal microscopy by dissection in phosphate buffer, fixation in4% formaldehyde and immunohistochemistry with the above listed primary andsecondary antibodies. Adult compound eyes were prepared for scanningelectron microscopy by dehydration through an ethanol gradient series followedby an ethanol–HMDS gradient series. Whole adult flies were viewed andphotographed with a Zeiss Discovery light microscope.

Induction of ectopic eyes

UAS-so and UAS-optix responder lines were crossed to 219 GAL4 lines(25°C) that comprise the Bloomington Stock Center GAL4 Collectionv05.06.29. Each line drives expression of a target gene in a unique spatial andtemporal pattern. The adult F1 progeny were scored for the presence of ectopiceye development using a light microscope. The results are presented in Table 1and Figs. 1 and 2. UAS-so deletion and UAS-so/optix chimeric responder lineswere crossed to the subset of GAL4 lines that generated ectopic eyes with thefull-length constructs. These results are presented in Figs. 3 and 4. A minimumof 5 deletion and chimeric responder lines were obtained and tested for eachconstruct.

Generation of UAS lines

We used PCR and the Gateway Cloning System from Invitrogen to clonefull-length so and optix coding sequences into the pUAST expression vector.Similar methods were used to generate molecules that encoded SO deletionvariants and SO-OPTIX chimeric proteins. The limits of each domain wereobtained from published reports (Pignoni et al., 1997; Seimiya and Gehring,

2000). The SO protein deletions are as follows: the SO ΔNT protein containsamino acids 98–417, the SO ΔSD protein contains amino acids 1–97 fused to218–277, the SO ΔHD protein contains amino acids 1–217 fused to 278–417and the SOΔCTcontains amino acids 1–277. The SO-OPTIX chimeric proteinsare as follows: the SO-OPTIX NTchimera contains amino acids 1–36 of OPTIXfused to amino acids 98–217 of SO, the SO-OPTIX SD chimera was generatedby replacing the SD of SO with amino acids 37–153 from OPTIX, the SO-OPTIX HD chimera was generated by replacing the HD of SO with amino acids154–214 of OPTIX, the SO-OPTIX CT chimera contains amino acids 1–277 ofSO fused to amino acids 215–488 of OPTIX. All wild-type, deletion variantsand chimeric proteins are depicted in Fig. 3. Detailed steps of the cloningprocedures and all primer sequences are available upon request. Germlinetransformants were generated and genetically mapped using standard methods(Ashburner et al., 2005).

Results

SO initiates eye formation within the developing antenna

A key step in the initiation of eye development is thought tobe the formation of the SO-EYA heterodimer. The SO-EYAcomplex influences eye specification by binding to DNA targetsequences through the HD of so and activating transcriptionthrough the EYA1 domain of eyes absent. The formation of theheterodimer is supported by in vitro binding assays and in vivoexpression studies in which so cannot induce eye formation onits own but will synergize with eya to generate ectopic eyes innon-retinal tissues. Interestingly, eya is capable of supportingeye development on its own at low frequencies in the antennaldisc. How eya initiates eye development without its obligatebinding partner remains enigmatic. One possible scenario is thatboth SO and EYA independently interact with additionalnuclear factors to induce eye formation. Recently, a yeasttwo-hybrid screen identified Sine oculis binding protein (Sbp)as a putative binding partner (Kenyon et al., 2005a). It containsa proline-rich region, which has been implicated in

Fig. 1. SO induces ectopic eyes in the antennal disc. (A–C) Scanning electron micrographs of adult normal and ectopic eyes. (D–S) Confocal images of 3rd instar eye-antennal imaginal discs. (A) Wild type. (B) cb41-GAL4/UAS-so, the yellow arrow marks the ectopic eye. (C) cb49-GAL4/UAS-optix, the yellow arrow marks theectopic eye on the ventral portion of the head. (D) cb41-GAL4/UAS-GFP, the GAL4 line drives expression within the developing eye, ocelli and a large portion of theantenna. (E) cb41-GAL4/UAS-so, ELAVexpression marks the ectopic eye within the antennal disc. (F) cb41-GAL4/UAS-so, ELAVand EYA are distributed in subsetsof cells within the cb41 expression pattern. (G) cb41-GAL4/UAS-so, DAC is also distributed in a smaller subdomain of the cb41 expression pattern. (H) cb41-GAL4/UAS-ey, expression of ey has a minimal effect on EYA distribution and no ectopic eyes are specified. (I) cb41-GAL4/UAS-ey, dac expression is not initiated by theexpression of ey in this expression domain. (J–K) cb41-GAL4/UAS-toy, expression of toy is insufficient to induce either eya, dac or elav expression. (L) dpp-GAL4/UAS-GFP, GFP is distributed along the posterior-lateral margins of the eye disc and in a sector of the ventral antennal disc. (M) dpp-GAL4/UAS-ey, expression of ey issufficient to induce ectopic eyes. (N–O) dpp-GAL4/UAS-so, expression of so induces the expression of eya, dac and induces eye formation in the antenna. (P) cb49-GAL4/UAS-GFP, expression of the GAL4 line is restricted to the distal most regions of the antennal disc. (Q–S) cb49-GAL4/UAS-so, note the ectopic photoreceptorsare located at a distance from the cells that express GAL4. Genotypes are listed at the top right of each panel. Visualized molecules are in the top right of each panel.Anterior is to the right.

759B. Weasner et al. / Developmental Biology 303 (2007) 756–771

transcriptional activation thus raising the possibility that so canactivate transcription independently of eya.

A prediction of this scenario is that so may, in fact, becapable of inducing eye development on its own (without theco-expression of eya). We set out to test this prediction by usingthe UAS/GAL4 system to forcibly express so in different spatialand temporal patterns within the developing fly. A UAS-soresponder line was crossed to 219 unique GAL4 driver lines. Incontrast to prior published results, we identified 4 instances in

which the expression of so is sufficient to initiate eyedevelopment (Table 1). In each instance, ectopic eyes wereobserved on either the adult antenna or the ventral portion of thehead (Figs. 1A–C, E, M, Q). An examination of the developingeye-antennal discs demonstrated that the GAL4 drivers inquestion are expressed, as expected, within the regions of theantennal disc that give rise to the head cuticle and the antennaproper (Figs. 1D, L, P; c253-GAL4 data not shown). It shouldbe noted here that the ectopic eyes are located in regions of the

Fig. 2. SO and OPTIX induce ectopic eyes in different cell types. (A–(D) Confocal images of 3rd instar eye-antennal imaginal discs. (E–G) Light microscope imagesof adult flies. (A) cb41-GAL4/UAS-GFP, note that the GAL4 line is expressed widely within the antennal disc. (B) cb41-GAL4/UAS-so, note the ELAV-positive cellslie within the broader GAL4 expression pattern. (C) cb309-GAL4/UAS-GFP, note the GAL4 expression pattern is in the medial–proximal segments of the antennaldisc. (D) cb309-GAL4/UAS-optix, note that the ELAV-positive cells are located in regions of the antennal disc that are distinct from those that are transformed by so.(E) Wild type. (F) rn-GAL4/UAS-so, note that expression of so within the wing causes the wing to shrink in size and has patterning defects. (G) rn-GAL4/UAS-optix,note the replacement of wing tissue with retinal tissue. Genotypes are listed at the bottom right of each panel. Visualized molecules are in the top right of each panel.Anterior is to the right.

760 B. Weasner et al. / Developmental Biology 303 (2007) 756–771

antennal disc that normally do not express eya suggesting thatthe induction of retinal tissue is not the result of simply handingSO protein its binding partner (data not shown). This suggeststhat the expression of so is sufficient to initiate the eyespecification cascade and redirect non-retinal tissues towards aneye fate. However, the ability of so to induce eye formation islimited. We were only able to induce ectopic eyes within theantenna and head cuticle. Furthermore, even within the antennaldisc, we observed several instances in which only small subsetsof cells that express so are actually transformed intophotoreceptors (Figs. 1D and E; compare GFP and ELAVdistribution profiles).

It should be noted that expression of so via the dpp-GAL4driver is also able to induce ectopic eyes. This is interestingbecause prior use of this driver by other groups with UAS-sofailed to show ectopic eye formation. We observe ectopicphotoreceptors and eye specification gene markers in theantennal disc but not ectopic ommatidia in the adult. Theommatida seen in the antennal disc are small and may beeliminated during later stages of development. It is possible thatearlier efforts by other groups focused on adult tissues, a timepoint that may be too late to see the ectopic eyes.

If eye development can be induced in antennal tissue inresponse to the individual expression of so, then what roledoes eya play in this process? In order to address thisquestion, we examined the expression pattern of eya thatresults in response to the forced expression of so. EYA proteinwas detected in the ectopic retinal tissue (Figs. 1F, N, R)suggesting that so is able to activate its expression. It is notclear if this regulation occurs through a direct interaction orthrough one of the many feedback loops that have beendocumented for the retinal determination pathway. Wecompared the expression of eya to the size of the ectopiceye and observed that EYA protein is distributed in a broaderpattern than ELAV (Figs. 1F, N). In summary, the distribution

of the three proteins can be written as SO>EYA>ELAV. Acomparison of the expression of dac, a downstream target ofboth so and eya reveals a similar pattern. DAC protein ispresent in more cells than ELAV but is still in only a smallersubset of cells that contain SO protein (Figs. 1G, O, S). Twopossible scenarios can account for these results. In onescenario, so requires factors that are necessary to activate eyaand dac and these factors may be only expressed in a subsetof cells that ectopically express so. Additional factors may bethen required to further commit cells towards a retinal fate(assayed by ELAV distribution) and these factors areexpressed in an even more restricted pattern; thus, only asmall number of cells become bona fide retinal neurons. Asecond scenario envisions a set of negative factors that blockso from inducing ectopic eyes. An expectation is that thesefactors would be expressed throughout the antennal discexcept in the few cells that can be transformed into retinaltissue. More sophisticated technologies such as laser capturemicroscopy coupled to DNA microarrays may provideopportunities to discriminate between the two models. It isalso possible to envision a scenario in which a combination ofboth models may have to taken into account.

RD genes are not equal in their ability to induce ectopic eyes

Since the retinal determination genes are not equal in theirability to induce eye formation, we were interested indetermining the ability of the other members to induce ectopiceyes when expressed with the same GAL4 drivers that gaveectopic eyes with so. We focused our efforts on the two Pax6homologs ey and toy whose encoded proteins (1) sit atop theeye specification network; (2) are the most potent inducers ofectopic eye formation; and (3) are known to bind regulatoryelements within the so gene and induce its expression.Surprisingly, in three of four instances in which expression of

Fig. 3. Schematic diagrams of SO deletion and SO/OPTIX chimeric proteins along with summary of rescue and overexpression assays. Individual domains of SO andOPTIX proteins were determined from published reports. All deletions and chimeric proteins were used in ectopic eye experiments (Fig. 4), so mutant rescueexperiments (Figs. 5 and 6) and overexpression experiments (Fig. 7). Results from these figures are shown in tabular form.

761B. Weasner et al. / Developmental Biology 303 (2007) 756–771

so produces ectopic eyes in the antenna, expression of either ofthese two genes was insufficient to induce retinal fates. Welooked at how cells expressing ey or toy regulate otherdownstream targets within these GAL4 expression patterns.For example, in response to the expression of the Pax6homologs, cells within the cb41-GAL4 expression patternturn on eya and dac weakly or not at all (Figs. 1H–K compareto Figs. 1D–G). Similar results are obtained for the cb49-GAL4and sd-GAL4 (data not shown). The absence (or extremely lowlevels) of expression of these two genes is likely, in part, to bethe underlying reason for the absence of ectopic eyes. Since sobut not upstream regulators such as ey and toy can induce

ectopic eyes in regions of the antenna (i.e., cb41-GAL4expression pattern in Fig. 1D), we postulate that there are yetto be identified factor(s) that lie between the Pax6 homologs eyand toy and their downstream target so (Fig. 8A). A predictionof this model is that within the antenna there are at least threegroups of cells (each corresponding to a location that can beconverted into retinal cells in response to so but not ey and toy)in which these putative factors are absent. Such factors may bepresent in other tissues as well since we have made similarobservations in other tissues (data not shown; C. Salzer and J.Kumar unpublished). Any such factor(s) would also bepredicted to be present in cells in which all three genes (ey,

Fig. 4. Induction of ectopic eyes by SO/OPTIX chimeric proteins. (A–C) Lightmicroscope images of whole adult flies. (A) cb41-GAL4/UAS-so/optix NTchimera, the arrow marks the location of the ectopic eye. The ectopic eye haspartially merged with the normal compound eye. (B) cb41-GAL4/UAS-so/optixCTchimera, arrowmarks the location of the ectopic eye. Note that the location isslightly different than that seen in panel A. (C) cb41-GAL4/UAS-so/optix NTCT chimera, the arrow marks the location of the ectopic eye. Note that in thiscase the ectopic eyes are very small in size compared with those depicted inpanels A and B. Expressed proteins are listed at the top left of each panel.Anterior is to the right.

762 B. Weasner et al. / Developmental Biology 303 (2007) 756–771

toy and so) can induce ectopic eyes. While we have notidentified these potential new genes, we have identified cellswithin the antennal dpp-GAL4 expression pattern that can betransformed into photoreceptor cells by ey, toy and so (Figs.1L–O; dpp-GAL4/UAS-toy not shown). Interestingly, only asubset of cells within the dpp-GAL4 expression pattern is EYAand DAC positive and even a smaller subset of cells aretransformed into photoreceptors (Figs. 1M–O).

SO and OPTIX induce ectopic eyes in different regions of thefly

This ability of so to induce eye formation suggests that itmight function similarly to optix (the other Six family memberto function in eye development), which can induce retinaldevelopment within antennal cells of the dpp-GAL4 expressionpattern (Seimiya and Gehring, 2000). In order to determine howsimilar these genes are in regards to redirecting tissue fates, weused the 219 GAL4 drivers to independently express optix andinduce ectopic eyes (Table 1). In addition to the four GAL4 linesthat direct ectopic eye development through expression of so,we have identified four new GAL4 patterns in which optix caninduce ectopic eyes (Table 1). A comparison of these resultsindicates that of this combined set of GAL4 expression patternsonly two (cb49-GAL4 and dpp-GAL4) contain cells that can beconverted into retinal tissue by both so and optix. Each of theother expression patterns contain cells that can be transformedinto photoreceptor neurons in response to either so or optix butnot both (Table 1). We also noted that the most commonlocations of the ectopic eyes are in the antenna and ventral head;both adult structures are derived from the antennal disc. Withinthe antennal disc, so and optix induce ectopic eyes in differentlocations of the antenna; two representative examples areprovided in Figs. 2A–D. Outside of the antenna, ectopic eyescan be generated in the wings and halteres by optix (Table 1;Figs. 2E, G). In contrast, the ability of so to induce ectopic eyesappears restricted to the antennal disc and its adult derivatives(Table 1; Figs. 1A–C and 2E–F). What accounts for thedifferences between these Six family members? The obviouspossibilities are that SO and OPTIX interact with different co-activator/repressor proteins, bind and activate different targetgenes or both. The former model is supported by yeast two-hybrid assays that have identified proteins that are boundpreferentially by either SO or OPTIX (Kenyon et al., 2005a,b).The latter model has not been tested as the consensus bindingsites for OPTIX have not been identified (Pauli et al., 2005).

Induction of ectopic eyes with SO deletion and SO/OPTIXchimeric proteins

We were interested in extending our understanding of thefunctional differences that distinguish the activities of these twoSIX family members during eye development. To this end, wehave generated a series of SO deletion and SO/OPTIX chimericproteins; each SO deletion protein is lacking individual ormultiple protein domains and each SO/OPTIX chimeric proteincontains single or multiple domain substitutions in which

domains of SO are replaced by those fromOPTIX (Fig. 3). Thesedeletion and chimeric proteins were first used to induce ectopiceyes and then later to rescue the no-eye phenotype of so loss-of-function mutants (Figs. 4–7). The logic behind these experi-ments is simple: induction of ectopic eyes and/or rescue ofretinal phenotypes by related deletion and chimeric proteins (i.e.,same domain in both constructs) is predicted to indicate whichdomains are dispensable for SO activity. In contrast, a failure toinduce eye development and/or rescue normal eye developmentby both protein types would identify essential domains. Andfinally, the induction and/or rescue of eye development by anSO/OPTIX chimeric protein but not its related SO deletionprotein would indicate a domain that is not only essential for SOfunction but also one whose activity has been conserved inOPTIX. As we are now expressing deletion and chimericproteins for the purpose of inducing ectopic eyes, we tested aminimum of 5 independent transformant lines in order to rule outthe possibility that expression levels of any one given constructwould effect the induction of ectopic eye development.

Our first attempt was to initiate eye development within thedeveloping antenna with each of the SO deletion and SO/OPTIXchimeric proteins described in Fig. 3 with the intent ofrecapitulating the results of full-length SO and OPTIX. Theseproteins were expressed in the antenna using the cb41-GAL4,cb49-GAL4 and c309-GAL4 lines. We recovered ectopic eyeswhen we expressed the SO/OPTIX NT, CT and NT+CTchimeras within the antenna suggesting that these domains mayhave a conserved function in eye development (Figs. 3 and4A–C). However, the SO ΔNT, SO ΔCT and SO ΔNT+CTdeletion proteins were incapable of initiating eye formation(Fig. 3). These domains share very limited sequence homo-logies; therefore, one possible explanation is that these

Fig. 5. Rescue of so mutants by SO deletion and SO/OPTIX chimeric proteins. (A–O) Scanning electron micrograph of adult compound eyes and head. (A) Wild type.(B) so1, note the complete absence of the compound eyes. (C) so1; ey-GAL4/UAS-so, the compound eye is restored to near wild-type levels. (D) so1; ey-GAL4/UAS-optix, expression of optix is insufficient to rescue the so1 mutant. (E) so1; ey-GAL4/UAS-so ΔNT, rescue is similar to wild-type so. (F) so1; ey-GAL4/UAS-so/optixNT chimera, rescue is similar to wild type. (G) so1; ey-GAL4/UAS-so ΔSD, no rescue. (H) so1; ey-GAL4/UAS-so/optix SD chimera, no rescue. (I) so1; ey-GAL4/UAS-so ΔHD, no rescue. (J) so1; ey-GAL4/UAS-so/optix HD chimera, compound eye development is partially rescued. (K) so1; ey-GAL4/UAS-so ΔCT, rescue issimilar to wild type. (L) so1; ey-GAL4/UAS-so/optix CT chimera, no rescue. (M) so1; ey-GAL4/UAS-so ΔNT CT, rescue is similar to wild type. (N) so1; ey-GAL4/UAS-so/optix NT CT chimera, no rescue. (O) so1; ey-GAL4/UAS-so/optix SD HD chimera, no rescue. Anterior is to the right.

763B. Weasner et al. / Developmental Biology 303 (2007) 756–771

domains are required to merely stabilize SO. We think that thisis unlikely since both domains appear to be completely dis-pensable during normal eye development (see below). Thus,any role(s) played by the N- and C-terminal domains appearsto be restricted to ectopic eye development. This result is

consistent with anecdotal evidence that suggests that somedifferences exist in the mechanisms underlying normal andectopic eye development.

We also failed to recover ectopic eyes when we expressedproteins that (1) lacked either the SD and/or HD and (2)

Fig. 6. Rescue of soD by SO deletion and SO/OPTIX chimeric proteins. (A–F) Scanning electron micrographs of adult compound eyes and heads. (A) soD/+, note thatheterozygotes completely lack the compound eyes. (B) soD/+; ey-GAL4/UAS-so, note that expression of wild type so partially restores eye development. Multiplecopies of the UAS-so insertion increase the number of ommatidia (data not shown). (C) soD/+; ey-GAL4/UAS-optix, no rescue. (D) soD/+; ey-GAL4/UAS-so ΔNT,note that the eye is partially restored. (E) soD/+; ey-GAL4/UAS-so/optix NT chimera, note that the compound eye is partially restored. (F) soD/+; ey-GAL4/UAS-so/optix HD chimera, note that the compound eye is partially restored. The remaining deletion and chimeric proteins listed in Fig. 3 did not rescue the no-eye phenotype ofsoD. All genotypes are listed at the bottom left of each panel. Anterior is to the right.

764 B. Weasner et al. / Developmental Biology 303 (2007) 756–771

substituted these domains from SO with those from OPTIX(Figs. 3 and 4). The results from the deletion constructs areunderstandable because we removed either the DNA bindingdomain or the main protein–protein interaction domain. Theseproteins would not be expected to function like wild-type SOand support eye development. The results from the expressionof SO/OPTIX SD, HD and SD+HD chimera are interestingbecause it suggests that the binding partner specificity and DNAtarget recognition properties of SO and OPTIX have divergedsignificantly. The identification of different in vitro bindingpartner specificities by Kenyon and co-workers are consistentwith these functional results for the SIX domain. The resultscentered on the HD are even more interesting in light of the factthat the SO/OPTIX HD substitution can rescue the so1 loss-of-function mutant (see below). This result further suggests thatthere may be real difference between the genetic contexts ofnormal and ectopic eye development. Identifying the biologicalbasis that underlies these differences will certainly be importantfor understanding how SIX family members (along with othermembers of the eye specification cascade) are able to direct non-retinal tissues towards and eye fate.

Functional conservation between SO and OPTIX

We have used the SO deletion and SO/OPTIX chimericproteins as a lever into further examining the functionalrelationship between the so and optix genes in normal eye

development. We expressed each deletion and chimeric proteindescribed in Fig. 3 within the ey-GAL4 expression domain(ahead of the morphogenetic furrow) of developing so1 mutanteyes in an attempt to rescue the structural defects seen in adulteyes (Figs. 5A, B). Expression of full-length SO protein (SOFL) restores the structure of the adult eye to near wild-typewhile full-length OPTIX cannot substitute and rescue the retinaldefect of so1 (Figs. 3 and 5C, D). This result indicates that thefunction of SO and OPTIX have diverged significantly since theoriginal duplication event and these proteins do not playredundant roles in normal eye specification.

We then set out to test the requirement and degree offunctional conservation for each domain. Expression of proteinsin which the N-terminal domain of SO was either deleted (SOΔNT) or replaced by the N-terminal of OPTIX (SO/OPTIX NT)were sufficient to support eye development at the levels of wild-type SO suggesting that the N-terminal regions of the proteinare not required for SO to initiate normal eye development(Figs. 3 and 5E, F). A similar conclusion can be reachedregarding the C-terminal tail of SO when we observed nearcomplete structural rescue of the eye in so1 flies that wereexpressing proteins in which either the C-terminal individually(SO ΔCT; Figs. 3 and 5K) or the N- and C-terminal regionstogether were deleted (SOΔNT+CT;Figs. 3 and 5M). In effect,this suggests that an SO protein containing just the core SIX andhomeobox domains is fully functional. However, it should benoted that while the C-terminal tail is not required for normal

Fig. 7. The C-terminal tail of OPTIX inhibits eye development. (A–F) Scanning electron micrographs of adult compound eyes. (G) Light microscope image of an adultretinal section. (A) GMR-GAL4/UAS-optix, note that the eye is devoid of ommatidia. The eye field has a glassy appearance and is replete with small bristles. (B)GMR-GAL4/UAS-so/optix NT chimera, note that the eye is slightly rough. The posterior edge of the eye has the severest roughening. (C) GMR-GAL4/UAS-so/optixSD chimera, note that the eye is wild type in its appearance. (D) GMR-GAL4/UAS-so/optix HD chimera, note that the eye is wild type in appearance. (E) GMR-GAL4/UAS-so/optix CT chimera, note that the eye has a similar phenotype to that seen when wild-type optix is expressed (A). (F) GMR-GAL4/UAS-so/optix NT CTchimera, note that the eye has a similar phenotype to that seen when wild-type optix is expressed (A). Please note that the 3 instances in which the eye is severelyaltered (panels A, E, F) are the result of the expression of molecules that contain the C-terminal regions of the OPTIX protein. (G) GMR-GAL4/UAS-optix, note thatthere are no photoreceptor cells within this retinal section. Instead the eye is filled with pigment and bristle cells. Retinal sections of the eyes shown in panels E and Flook identical to that shown here (data not shown).

765B. Weasner et al. / Developmental Biology 303 (2007) 756–771

SO function, replacement with the C-terminal tail of OPTIX(SO/OPTIX CT and SO/OPTIX NT+CT) appears to have aninhibitory effect and prevents the rescue of so1 (Figs. 3 and 5L,N). An alignment of the SO and OPTIX C-terminal tailsindicates that there is less than 10% amino acid similaritybetween these two regions. In contrast, an alignment of the C-terminal regions of OPTIX and the mammalian homologs SIX3and SIX6 has revealed two regions of conservation that mightrepresent new functional motifs (Fig. 8D). These results are insharp contrast to the N-terminal regions of SO and OPTIXwhich have little homolog to their mammalian counterpartsequences and appear to be completely dispensable andinterchangeable. One possible explanation is that the C-terminal

tail is a crucial element in individualizing the activities of theseevolutionarily related proteins; possibly through the recruitmentof additional binding partners (Fig. 8C).

We then focused our examination at the protein–proteininteraction and DNA binding domains. First, deletion orreplacement of the SD regions (SO ΔSD and SO/OPTIX SD)rendered the modified SO proteins incapable of restoring eyedevelopment (Figs. 3 and 5G, H). Since the SD regions of SIXproteins are involved in protein–protein interactions, theseresults suggest that the SD regions of SO and OPTIX bind tounique binding partners and this specificity in partner selectionis crucial to the role that each protein plays in eye development.This assertion is supported by several lines of evidence that

Fig. 8. Models for SO and OPTIX activity and regulation during eye development in Drosophila. (A) A portion of the eye specification cascade is shown. Our resultssuggest that in certain areas of the antenna expression of so but not toy or ey is sufficient to induce ectopic eyes (despite TOYand EYproteins binding to the promoterof so). One possible model is that there are additional (and yet to be identified) players that reside genetically between the two Pax6 genes and so. (B) We observed (inlimited circumstances) that expression of so could non-autonomously induce ectopic eyes. A similar effect has been documented in the eye disc when dpp isoverexpressed ahead of the morphogenetic furrow. Dpp is expressed in the antennal disc. One potential model is that SO interacts with or regulates Dpp, which in turncan induce ectopic eyes non-autonomously. (C) Our results suggest that the C-terminal tail plays a role in eye development and our sequence analysis has indicated thatthere are regions of amino acid conservation. In one model these regions of OPTIX might be bound to additional co-factors and that this helps to modulate OPTIXactivity. (D) A schematic diagram showing an alignment of the C-terminal regions of OPTIX, SIX3 and SIX6. The red blocks are regions of the highest conservationthat might serve as potential protein–protein interaction domains.

766 B. Weasner et al. / Developmental Biology 303 (2007) 756–771

includes genetic and in vitro biochemical experiments demon-strating an exclusive interaction between SO and the transcrip-tional co-activator EYA (Pignoni et al., 1997; Seimiya andGehring, 2000, Kenyon et al., 2005a). Additionally, yeast twohybrid assays using the SD domains of the Six proteins as baitshave identified differing sets of putative binding partners for SOand OPTIX (Kenyon et al., 2005a).

The homeodomains of SO and OPTIX are partiallyinterchangeable

We then manipulated the homeodomain and, as expected, aprotein lacking the HD of SO (SO ΔHD failed to rescue so1;Figs. 3 and 5I). Unexpectedly, expression of a protein in whichthe HD domain of SO was replaced with the HD of OPTIX (SO/OPTIX HD) partially rescued so1 (Figs. 3 and 5J). We hadpreviously isogenized our so1 line and the no-eye phenotype is

100% penetrant; thus, the presence of ommatidia in the so1; ey-GAL4/UAS-so/optix HD flies likely reflects a partial conserva-tion in the DNA binding specificity of the SO and OPTIXhomeodomains. We then expressed a protein in which both theSD and HD domains of SO were replaced with the correspond-ing domains of OPTIX but did not observe any restoration of eyedevelopment (Figs. 3 and 5O). We conclude from this result thatalthough SO and OPTIX are capable of binding to at least somecommon transcriptional targets, binding partner selection(which occurs through the SD domain) also plays a major rolein distinguishing SO from OPTIX.

Inhibition of eye development in soD mutants

A key feature that distinguishes the activities of SIXtranscription factors appears to be the selection of proteinpartners that are bound to the SIX domain (Kenyon et al.,

767B. Weasner et al. / Developmental Biology 303 (2007) 756–771

2005a). A dominant-negative so allele (soD) has been shown tocontain a valine to aspartic acid substitution (V98D) within theSD (Roederer et al., 2005; Kenyon et al., 2005b). Flies harboringa single copy of this mutation lack compound eyes whilehomozygous mutants die during embryogenesis (Roederer et al.,2005; Kenyon et al., 2005b). The underlying biological basis ofthis phenotype is thought to be different than that of traditionalloss-of-function mutants; in one scenario, the V98D amino acidsubstitution might alter the activity of SO by recruiting proteinsthat normally physically interact with OPTIX thus making theSO-D protein function more like OPTIX than SO (Kenyon et al.,2005b; Roederer et al., 2005). This scenario is supported byexperiments in which expression of the soV98D protein in anotherwise wild-type background deletes the compound eye in amanner that exactly phenocopies the extant soDmutant (Kenyonet al., 2005b) while expression of wild-type SO protein has noeffect (Roederer et al., 2005). We sought to test this model byexpressing the SO/OPTIX SD chimera within the developingeye with both ey-GAL4 and GMR-GAL4 drivers. Since the HDsappear to be interchangeable (Figs. 3 and 5J), the SO/OPTIX SDchimera is predicted to function just like soD (due to thereplacement of the SO SIX domain with that of OPTIX).Interestingly, in both cases we did not observe any significantalteration in eye structure (data not shown). If the current modelwas correct, then we should have been able to phenocopy thesoD mutant phenotype by expressing a protein that has the entireSIX domain of OPTIX substituted into SO. Since we did notobtain this result, we conclude that the dominant-negativephenotype of soD is not due to a switch in binding partnerselection but is rather due to another yet to be determinedmechanism. One plausible alternate mechanism might be thatthe V98D substitution results in a higher affinity of SO-D forEYA and/or other binding partners (when compared to thebinding of SO to these factors), which in turn may hyperactivatethe transcription of target genes. Such hyperactivation of targetgenes can result in eye loss, as the eye appears to be sensitive tothe dosage of eye specification proteins. Overexpression ofmany eye specification genes such as optix, eyes absent, dach-shund, eyegone and twin of eyegone within the developing eyeresults in moderate to severe retinal loss (J. Kumar unpublisheddata).

We extended our examination of the mechanism underlyingSO-D activity by expressing each SO deletion and SO/OPTIXchimeric proteins throughout developing soD mutant retinas andthen assaying the ability of each protein to rescue the structuraldefects. Expression of the wild-type SO protein partially rescuedsoD (150–200 ommatidia) while expression of OPTIX appearedto have no visible effect (Figs. 3 and 6A–C), indicating that theeffect of the SO-D protein can be titrated by an increase in thelevels of wild-type SO and further supports an inhibitory role forOPTIX in normal eye development. We then expressed each ofthe deletion and chimeric proteins throughout the retinas of soD

flies and observed a partial restoration of eye development (30–200 ommatidia) in the cases in which the SO ΔNT, SO/OPTIXNT and SO/OPTIX HD proteins were expressed (30–200ommatidia; Figs. 3 and 6D–F). Each of the four proteins thatrescue the retinal phenotype of soD contained both the SD and

C-terminal domains of SO. In contrast, deletions or chimeras thataffected the SD or C-terminal regions were unable to restorenormal eye development to soD retinas. The results takentogether with the overexpression of the SO/OPTIX SD chimera(above) further implicate both the SIX domain and theC-terminal region in differentiating functional activities of SOand OPTIX. Furthermore, these results suggest that themechanism by which SO-D blocks eye development may bemore than simply functioning like its evolutionary cousinOPTIX.

The C-terminal of OPTIX inhibits eye development

We were particularly intrigued by the apparent incompat-ibility among the C-terminal tails of SO and OPTIX. Our resultsto date indicate that although the C-terminal is dispensable fornormal SO protein function, the C-terminal of OPTIX cannotserve as a substitute. This led us to speculate that the C-terminaltail of OPTIX functions to inhibit eye development. In order totest this hypothesis, we expressed full-length SO, OPTIX andeach chimeric protein described in Fig. 3 ahead and behind themorphogenetic furrow using ey-GAL4 (data not shown) andGMR-GAL4 drivers, respectively (Fig. 7). Expression of SO inall cells posterior to the morphogenetic furrow (using GMR-GAL4) causes only a very slight roughening of the eye (data notshown). Likewise, distribution of the SO/OPTIX NT, SO-OPTIX SD and SO/OPTIX HD proteins in the same expressionpattern appeared to have only a minor effect on the developmentand structure of the eye (Figs. 7B–D). A common featurebetween full-length SO and these three chimeric proteins is thateach contains the C-terminal from SO (Fig. 3).

In contrast, expression of full-length OPTIX in all cellsposterior to the morphogenetic furrow (using GMR-GAL4)results in a severely altered external surface and a complete lossof photoreceptor neurons (Figs. 7A, H; compare to Figs. 1A and7G). Interestingly, distribution of the SO/OPTIX CT and SO/OPTIX NT+CT chimeras inhibited eye development to thesame degree as full-length OPTIX (Figs. 7E, F). A criticalfeature that is found in common to these three proteins is thepresence of the C-terminal tail region of OPTIX (Fig. 3). Weobtained similar patterns when we expressed all full-length,deletion and chimeric proteins ahead of the furrow using an ey-GAL4 (data not shown). We have interpreted these results asfurther evidence that the OPTIX C-terminal tail does inhibit eyedevelopment. A comparison of the amino acid sequence ofOPTIX and its mammalian homologs has revealed two regionsof high conservation (Fig. 8D). While it is not yet clear if theseregions play a role in the inhibition, one can imagine a scenario(one of several) in which these domains interact with atranscriptional co-repressor to inhibit the activation of targetgenes. The identification of any such factors will be animportant next step in furthering our understanding of howOPTIX functions in eye development. Furthermore, it will beinteresting to determine if the other SIX family members recruitco-factors via their C-terminal tails also.

Since our analysis of optix overexpression suggested that itplays a role in normal eye development, we made retinal mosaic

768 B. Weasner et al. / Developmental Biology 303 (2007) 756–771

clones of a hypomorphic mutation that results from the insertionof a transposable element in the second intron of optix. Basedon sequence analysis, there are no ORFs predicted to lie withinthis intron thus the P-element is likely to have disrupted aregulatory element of optix. An examination of retinal mosaicclones indicates a slight roughening in the structure of the eyeand a slight disruption in eye specification (data not shown),furthering a potential role for optix in eye specification. Thesedata suggest that this allele is a weak hypomorph and efforts tomake a null allele are in progress.

The results that we have presented here indicate that SO, likeeach of the other retinal determination genes, is capable ofinitiating eye development. We have also demonstrated that thedifferences in the activities of SO and OPTIX are not dueexclusively to the DNA binding domain. In fact, our findingssuggest that there might be minimal differences in target geneactivation and that the SD and C-terminal domains are crucial inestablishing distinct functions for the members of the SIXfamily of homeobox transcription factors.

Discussion

The specification of the compound eye of Drosophila (andby extension the vertebrate eye) is a multi-step process in whicha primordial tissue is directed towards an eye fate through aseries of stepwise events. Prior work on eye specification hasheld that a critical step in this process involves the retinaldetermination proteins SO and EYA forming a physicaltranscriptional complex that goes on to promote retinaldevelopment by regulating the expression of downstream targetgenes (Pignoni et al., 1997). Another key hypothesis is that SOand its evolutionary cousin OPTIX play distinct roles in eyedevelopment (Seimiya and Gehring, 2000). This model wasbuilt upon several lines of evidence. First, unlike all retinaldetermination genes, so was seemingly incapable of inducingectopic eyes when expressed individually. Second, co-expres-sion of so and eya results in a synergistic increase in ectopic eyeformation over levels seen with either so or eya alone. Andfinally, in vitro binding assays demonstrated a biochemicalinteraction between the SO and EYA (Pignoni et al., 1997).Additional data indicate that optix and so are expressed indifferent patterns within the developing eye: optix expression isrestricted to cells ahead of the morphogenetic furrow while so istranscribed both ahead and behind the furrow. Finally,expression of optix, on its own, is sufficient to induce eyeformation and this effect does not require an interaction witheya (Seimiya and Gehring, 2000). Together, these data suggestthat the formation of the SO-EYA complex is a key step in eyedevelopment. They also suggest that SO and OPTIX regulatedifferent aspects of eye specification. In this paper, we set out to(1) determine if so is indeed incapable of promoting eyeformation on its own and to (2) conduct a structure/functionanalysis of the SO and OPTIX proteins in an effort to identifythe molecular and biochemical mechanisms that are responsiblefor the differences in activity.

In order to address our first question, we used 219 GAL4lines to express SO and OPTIX individually throughout

developing tissues and we have been able to demonstrate that,in contrast to previous reports, the expression of so on its own issufficient to initiate eye development in the antennal disc (Table1, Figs. 1 and 2). The induction of ectopic eyes does not requirethe co-expression of eya as previously held since EYA protein isnot normally distributed in the areas in which the ectopic eyesare generated (data not shown). However, it should be noted thattranscription of eya (and the downstream retinal determinationgene dac) is activated in response to so although it is not clear ifthis interaction is direct (Fig. 1). One possible interpretation ofour results is that the first step in eye development is theactivation of so transcription by EY and TOY proteins. This is,in turn, followed by the activation of eya expression by SO.Subsequently, the SO-EYA complex is assembled and functionsto promote eye specification by activating target genetranscription. It should be noted, however, that the expressionof so is not always sufficient to induce eya transcription nor isthe co-expression of so and eya sufficient in all circumstancesto promote eye development. As we have shown, the expressionof so induces eya expression in only a subset of antennal cellsand only an even smaller subset of these cells are converted intoretinal photoreceptor neurons (Fig. 1). One explanation is thatthere are yet to be identified factors that mediate (1) theactivation of eya by so and (2) the specification of retinal fatesby the SO-EYA complex. These factors may be expressedwithin normal eye tissue and only in the small regions of non-retinal tissues in which ectopic eyes can be generated. Theidentification of such factors, if they exist, will be a key steptowards gaining a greater appreciation of the how each step ineye specification is regulated.

Our survey has also shown that the induction of eyedevelopment by optix is not restricted to the antenna, aspreviously suggested, but can also occur within the wing (Table1, Fig. 2). This result, along with those for so, clearly suggeststhat the SIX-type homeodomain proteins have a wider potentialfor inducing retinal development than has been previouslydescribed. A similar conclusion can be drawn for each of theremaining eye specification genes (C. Salzer and J. Kumar;manuscript in preparation). This is important since severalreports have attempted to elucidate biological mechanisms forthis restriction. For instance, it has been reported that the TGFβsignaling pathway functions within the eye specificationpathway and might serve as a pre-requisite for the inductionof ectopic eyes (Chen et al., 1999). However, we observe thatboth so and optix can induce ectopic eyes in locations withinthe antenna that are not under the influence of this pathway.Thus, it is likely that the requirements for the induction of eyedevelopment are more complicated. A complete map of whereand when eye formation can be induced within the fly will beuseful in identifying the positive and inhibitory influences thatregulate retinal specification.

A potentially important observation is that several regions ofthe antennal disc that are converted by so into retinal tissue arerefractory to the activity of other retinal determination genes.For example, expression of either ey or toy via cb41-GAL4 isinsufficient to induce ectopic eyes despite the fact that so caninduce ectopic eyes in this expression zone and that EY and

769B. Weasner et al. / Developmental Biology 303 (2007) 756–771

TOYproteins bind to regulatory regions within so and activatesits transcription. Similar discrepancies were seen for each retinaldetermination gene in several imaginal discs (C. Salzer and J.Kumar; manuscript in preparation). We predict that additionalfactors will be required for each step in the eye specificationpathway. One possible approach to identifying such factorswould be to use laser capture microscopy (LCM) to isolatesmall groups of cells from which a molecular fingerprint can bedetermined through the use of DNA microarrays. It is likely thatthese types of approaches will reveal an increased level ofcomplexity within the cascade.

In order to address our second interest, we created a series ofSO deletion and SO/OPTIX chimeric proteins that along withfull-length proteins (described in Fig. 3) were used to initiateeye formation in non-retinal tissues as well as to restore retinaldevelopment in so1 loss-of-function and soD dominant-negativemutants (Figs. 4–7). This structure/function approach allowedus to define potential differences between how two SIX-typehomeodomain proteins function in eye development. Prior tothis report, it was traditionally thought that the differencesbetween these proteins could be solely attributed to eitherbinding partner selection through the Six domain (SD) and/ortranscriptional target selection through the homeodomain (HD).Our results confirmed that the SD, through exclusive bindingpartner selection, is crucial to distinguishing between theactivities of SIX family members (Figs. 5 and 6). To oursurprise, we have been able to also demonstrate that the C-terminal region of SIX proteins is important in conferringfunctional specificity (Figs. 6 and 7). These regions do not sharesignificant homology amongst different family members but dohave regions of conservation within members of the samesubfamily (Fig. 8C). A potential model might include themodulation of OPTIX activity by the interaction of additionalbinding proteins at these amino acids. Since prior reportsidentifying OPTIX binding partners made exclusive use of theSix domain, good candidates for potential binding partners arenot yet available. Their identification will be an important steptowards a better understanding of the functional differencesbetween the Six-type transcription factors.

Also surprising was the apparent functional conservationbetween the HDs of SO and OPTIX. A chimeric protein inwhich the HD of SO was replaced with the correspondingregion of OPTIX (SO/OPTIX HD) was able to partially rescuethe so1 loss-of-function mutant. One potential explanation isthat both SO and OPTIX proteins bind to a similar set of targetgenes. This hypothesis is supported by the fact that several ofthe mammalian SIX homologs have very similar binding sitesare known to regulate several common targets. While the DNAbinding sequence of OPTIX is not known, it is predicted to bevery similar to the consensus sites that have been described forSO and the mammalian homologs. If SO and OPTIX proteinsbind to common targets, then what confers functional specificityduring eye development? The results presented here and thosedescribed by Kenyon and colleagues suggests that the answerlikely resides within the Six and C-terminal domains.

In total, our results are consistent with a model in which bothSIX-type homeodomain proteins SO and OPTIX are placed

sufficiently high within the retinal determination cascade toinitiate eye specification although they may promote eyedevelopment in distinct ways. Interestingly, the vertebrateSIX1 and SIX2 genes, orthologs of so, do not appear to functionduring early eye formation but instead play significant roles inmuscle specification. This is in contrast to the SIX3 and SIX6genes, homologs of optix, which have demonstrated roles in theearly development and disease of the mammalian eye (Boucheret al., 2000; Kawakami et al., 2000; Ruf et al., 2004). While theresults presented here represent a step towards understandingthe functional relationship between two SIX family members,further dissections of both insect and mammalian SIX proteinswill be required to completely elucidate the biological basis thatunderlies the functional divergence of members of this family oftranscription factors.

Acknowledgments

We would like to thank Kathy Matthews, Peter Cherbas andTomDonahue for comments and suggestions on the manuscript.We would also like to thank Seymour Benzer, Lucy and PeterCherbas, Graeme Mardon, Janice Fischer, Walter Gehring, theDevelopmental Studies Hybridoma Bank and the BloomingtonDrosophila Stock Center for fly strains and molecular reagents.This work is supported by a grant to J.P.K. from the NationalEye Institute R01 EY014863.

References

Altmann, C.R., Chow, R.L., Lang, R.A., Hemmati-Brivanlou, A., 1997. Lensinduction by Pax-6 in Xenopus laevis. Dev. Biol. 185, 119–123.

Ashburner, M., Golic, K.G., Hawley, R.S., 2005. “Drosophila: A LaboratoryHandbook”. Cold Spring Laboratory Press, Cold Spring Harbor.

Azuma, N., Hirakiyama, A., Inoue, T., Asaka, A., Yamada, M., 2000. Mutationsof a human homologue of the Drosophila eyes absent gene (EYA1) detectedin patients with congenital cataracts and ocular anterior segment anomalies.Hum. Mol. Genet. 9, 363–366.

Bessa, J., Casares, F., 2005. Restricted teashirt expression confers eye-specificresponsiveness to Dpp and Wg signals during eye specification in Droso-phila. Development 132, 5011–5020.

Bessa, J., Gebelein, B., Pichaud, F., Casares, F., Mann, R.S., 2002.Combinatorial control of Drosophila eye development by eyeless,homothorax, and teashirt. Genes Dev. 16, 2415–2427.

Bonini, N.M., Leiserson, W.M., Benzer, S., 1993. The eyes absent gene: geneticcontrol of cell survival and differentiation in the developing Drosophila eye.Cell 72, 379–395.

Bonini, N.M., Bui, Q.L., Gray-Board, G.L., Warrick, J.M., 1997. The Droso-phila eyes absent gene directs ectopic eye formation in a pathway conservedbetween flies and vertebrates. Development 124, 4819–4826.

Boucher, C.A., Winchester, C.L., Hamilton, G.M., Winter, A.D., Johnson,K.J., Bailey, M.E., 2000. Structure, mapping and expression of thehuman gene encoding the homeodomain protein, SIX2. Gene 247,145–151.

Chen, R., Amoui, M., Zhang, Z., Mardon, G., 1997. Dachshund and EyesAbsent proteins form a complex and function synergistically to induceectopic eye development in Drosophila. Cell 91, 893–903.

Chen, R., Halder, G., Zhang, Z., Mardon, G., 1999. Signaling by the TGF-bhomolog decapentaplegic functions reiteratively within the network of genescontrolling retinal cell fate determination in Drosophila. Development 126,935–943.

Cheyette, B.N., Green, P.J., Martin, K., Garren, H., Hartenstein, V., Zipursky,S.L., 1994. The Drosophila sine oculis locus encodes a homeodomain-

770 B. Weasner et al. / Developmental Biology 303 (2007) 756–771

containing protein required for the development of the entire visual system.Neuron 12, 977–996.

Chow, R.L., Altmann, C.R., Lang, R.A., Hemmati-Brivanlou, A., 1999. Pax6induces ectopic eyes in a vertebrate. Development 126, 4213–4222.

Czerny, T., Halder, G., Kloter, U., Souabni, A., Gehring, W.J., Busslinger, M.,1999. twin of eyeless, a second Pax-6 gene of Drosophila, acts upstream ofeyeless in the control of eye development. Mol. Cell 3, 297–307.

Gallardo, M.E., Lopez-Rios, J., Fernaud-Espinosa, I., Granadino, B., Sanz, R.,Ramos, C., Ayuso, C., Seller, M.J., Brunner, H.G., Bovolenta, P., Rodriguezde Cordoba, S., 1999. Genomic cloning and characterization of the humanhomeobox gene SIX6 reveals a cluster of SIX genes in chromosome 14 andassociates SIX6 hemizygosity with bilateral anophthalmia and pituitaryanomalies. Genomics 61, 82–91.

Glaser, T., Jepeal, L., Edwards, J.G., Young, S.R., Favor, J., Maas, R.L.,1994. PAX6 gene dosage effect in a family with congenital cataracts,aniridia, anophthalmia and central nervous system defects. Nat. Genet. 7,463–471.

Halder, G., Callaerts, P., Gehring, W.J., 1995. Induction of ectopic eyes by targetexpression of the eyeless gene in Drosophila. Science 267, 1788–1792.

Hanson, I.M., 2001. Mammalian homologues of the Drosophila eye specifica-tion genes. Semin. Cell Dev. Biol. 12, 475–484.

Harris, S.E., Winchester, C.L., Johnson, K.J., 2000. Functional analysis of thehomeodomain protein SIX5. Nucleic Acids Res. 28, 1871–1878.

Hill, R.E., Favor, J., Hogan, B.L., Ton, C.C., Saunders, G.F., Hanson, I.M.,Prosser, J., Jordan, T., Hastie, N.D., van Heyningen, V., 1991. Mouse smalleye results from mutations in a paired-like homeobox-containing gene.Nature 354, 522–525.

Jang, C.C., Chao, J.L., Jones, N., Yao, L.C., Bessarab, D.A., Kuo, Y.M., Jun, S.,Desplan, C., Beckendorf, S.K., Sun, Y.H., 2003. Two Pax genes, eye goneand eyeless, act cooperatively in promoting Drosophila eye development.Development 130, 2939–2951.

Kawakami, K., Ohto, H., Ikeda, K., Roeder, R.G., 1996a. Structure, functionand expression of a murine homeobox protein AREC3, a homologue ofDrosophila sine oculis gene product, and implication in development.Nucleic Acids Res. 24, 303–310.

Kawakami, K., Ohto, H., Takizawa, T., Saito, T., 1996b. Identification andexpression of six family genes in mouse retina. FEBS Lett. 393, 259–263.

Kawakami, K., Sato, S., Ozaki, H., Ikeda, K., 2000. Six family genes-structureand function as transcription factors and their roles in development.BioEssays 22, 616–626.

Kenyon, K.L., Li, D.J., Clouser, C., Tran, S., Pignoni, F., 2005a. Fly SIX-typehomeodomain proteins Sine oculis and Optix partner with different cofactorsduring eye development. Dev. Dyn. 234, 497–504.

Kenyon, K.L., Yang-Zhou, D., Cai, C.Q., Tran, S., Clouser, C., Decene, G.,Ranade, S., Pignoni, F., 2005b. Partner specificity is essential for properfunction of the SIX-type homeodomain proteins Sine oculis and Optixduring fly eye development. Dev. Biol. 286, 158–168.

Kirby, R.J., Hamilton, G.M., Finnegan, D.J., Johnson, K.J., Jarman, A.P., 2001.Drosophila homolog of the myotonic dystrophy-associated gene, SIX5, isrequired for muscle and gonad development. Curr. Biol. 11, 1044–1049.

Klesert, T.R., Cho, D.H., Clark, J.I., Maylie, J., Adelman, J., Snider, L., Yuen,E.C., Soriano, P., Tapscott, S.J., 2000. Mice deficient in Six5 developcataracts: implications for myotonic dystrophy. Nat. Genet. 25, 105–109.

Li, X., Oghi, K.A., Zhang, J., Krones, A., Bush, K.T., Glass, C.K., Nigam, S.K.,Aggarwal, A.K., Maas, R., Rose, D.W., Rosenfeld, M.G., 2003. Eya proteinphosphatase activity regulates Six1-Dach-Eya transcriptional effects inmammalian organogenesis. Nature 426, 247–254.

Loosli, F., Winkler, S., Wittbrodt, J., 1999. Six3 overexpression initiates theformation of ectopic retina. Genes Dev. 13, 649–654.

Mardon, G., Solomon, N.M., Rubin, G.M., 1994. dachshund encodes a nuclearprotein required for normal eye and leg development in Drosophila.Development 120, 3473–3486.

Onuma, Y., Takahashi, S., Asashima, M., Kurata, S., Gehring, W.J., 2002.Conservation of Pax 6 function and upstream activation by Notch signalingin eye development of frogs and flies. Proc. Natl. Acad. Sci. U. S. A. 99,2020–2025.

Pai, C.Y., Kuo, T.S., Jaw, T.J., Kurant, E., Chen, C.T., Bessarab, D.A., Salzberg,A., Sun, Y.H., 1998. The Homothorax homeoprotein activates the nuclear

localization of another homeoprotein, extradenticle, and suppresses eyedevelopment in Drosophila. Genes Dev. 12, 435–446.

Pan, D., Rubin, G.M., 1998. Targeted expression of teashirt induces ectopic eyesin Drosophila. Proc. Natl. Acad. Sci. U. S. A. 95, 15508–15512.

Pasquier, L., Dubourg, C., Blayau, M., Lazaro, L., Le Marec, B., David, V.,Odent, S., 2000. A new mutation in the six-domain of SIX3 gene causesholoprosencephaly. Eur. J. Hum. Genet. 8, 797–800.

Pauli, T., Seimiya, M., Blanco, J., Gehring, W.J., 2005. Identification offunctional sine oculis motifs in the autoregulatory element of its own gene,in the eyeless enhancer and in the signalling gene hedgehog. Development132, 2771–2782.

Pichaud, F., Casares, F., 2000. homothorax and iroquois-C genes are required forthe establishment of territories within the developing eye disc. Mech. Dev.96, 15–25.

Pignoni, F., Hu, B., Kenton, H.Z., Xiao, J., Garrity, P.A., Zipursky, S.L., 1997.The eye-specification proteins So and Eya form a complex and regulatemultiple steps in Drosophila eye development. Cell 91, 881–891.

Quiring, R., Walldorf, U., Kloter, U., Gehring, W.J., 1994. Homology of theeyeless gene of Drosophila to the Small eye gene in mice and Aniridia inhumans [see comments]. Science 265, 785–789.

Rayapureddi, J.P., Kattamuri, C., Steinmetz, B.D., Frankfort, B.J., Ostrin, E.J.,Mardon, G., Hegde, R.S., 2003. Eyes absent represents a class of proteintyrosine phosphatases. Nature 426, 295–298.

Roederer, K., Cozy, L., Anderson, J., Kumar, J.P., 2005. Novel dominant-negative mutation with in the six domain of the conserved eye specificationgene sine oculis inhibits eye development in Drosophila. Dev. Dyn. 232,753–766.

Ruf, R.G., Xu, P.X., Silvius, D., Otto, E.A., Beekmann, F., Muerb, U.T., Kumar,S., Neuhaus, T.J., Kemper, M.J., Raymond Jr., R.M., Brophy, P.D.,Berkman, J., Gattas, M., Hyland, V., Ruf, E.M., Schwartz, C., Chang, E.H., Smith, R.J., Stratakis, C.A., Weil, D., Petit, C., Hildebrandt, F., 2004.SIX1 mutations cause branchio-oto-renal syndrome by disruption of EYA1-SIX1-DNA complexes. Proc. Natl. Acad. Sci. U. S. A. 101, 8090–8095.

Sarkar, P.S., Appukuttan, B., Han, J., Ito, Y., Ai, C., Tsai, W., Chai, Y., Stout, J.T.,Reddy, S., 2000. Heterozygous loss of Six5 in mice is sufficient to causeocular cataracts. Nat. Genet. 25, 110–114.

Seimiya, M., Gehring, W.J., 2000. The Drosophila homeobox gene optix iscapable of inducing ectopic eyes by an eyeless-independent mechanism.Development 127, 1879–1886.

Seo, H.C., Curtiss, J., Mlodzik, M., Fjose, A., 1999. Six class homeobox genesin Drosophila belong to three distinct families and are involved in headdevelopment. Mech. Dev. 83, 127–139.

Serikaku, M.A., O'Tousa, J.E., 1994. Sine oculis is a homeobox gene requiredfor Drosophila visual system development. Genetics 138, 1137–1150.

Shen, W., Mardon, G., 1997. Ectopic eye development in Drosophila inducedby directed dachshund expression. Development 124, 45–52.

Silver, S.J., Davies, E.L., Doyon, L., Rebay, I., 2003. Functional dissection ofeyes absent reveals new modes of regulation within the retinal determinationgene network. Mol. Cell. Biol. 23, 5989–5999.

Singh, A., Kango-Singh, M., Sun, Y.H., 2002. Eye suppression, a novel functionof teashirt, requires Wingless signaling. Development 129, 4271–4280.

Suzuki-Yagawa, Y., Kawakami, K., Nagano, K., 1992. Housekeeping Na,K-ATPase alpha 1 subunit gene promoter is composed of multiple cis elementsto which common and cell type-specific factors bind. Mol. Cell. Biol. 12,4046–4055.

Tootle, T.L., Silver, S.J., Davies, E.L., Newman, V., Latek, R.R., Mills, I.A.,Selengut, J.D., Parlikar, B.E., Rebay, I., 2003. The transcription factor Eyesabsent is a protein tyrosine phosphatase. Nature 426, 299–302.

Traboulsi, E.I., 1998. Ocular malformations and developmental genes. J. AAPOS2, 317–323.

Wallis, D.E., Roessler, E., Hehr, U., Nanni, L., Wiltshire, T., Richieri-Costa, A.,Gillessen-Kaesbach, G., Zackai, E.H., Rommens, J., Muenke, M., 1999.Mutations in the homeodomain of the human SIX3 gene causeholoprosencephaly. Nat. Genet. 22, 196–198.

Winchester, C.L., Ferrier, R.K., Sermoni, A., Clark, B.J., Johnson, K.J., 1999.Characterization of the expression of DMPK and SIX5 in the human eye andimplications for pathogenesis in myotonic dystrophy. Hum. Mol. Genet. 8,481–492.

771B. Weasner et al. / Developmental Biology 303 (2007) 756–771

Xu, P.X., Woo, I., Her, H., Beier, D.R., Maas, R.L., 1997. Mouse Eyahomologues of the Drosophila eyes absent gene require Pax6 for expressionin lens and nasal placode. Development 124, 219–231.

Yan, H., Canon, J., Banerjee, U., 2003. A transcriptional chain linking eyespecification to terminal determination of cone cells in the Drosophila eye.Dev. Biol. 263, 323–329.

Yao, J.G., Sun, Y.H., 2005. Eyg and Ey Pax proteins act by distincttranscriptional mechanisms in Drosophila development. EMBO J. 24,2602–2612.

Zuber, M.E., Gestri, G., Viczian, A.S., Barsacchi, G., Harris, W.A., 2003.Specification of the vertebrate eye by a network of eye field transcriptionfactors. Development 130, 5155–5167.