MOLECULAR AND CELLULAR BIOLOGY, Sept. 2003, p. 5989–5999
Copyright © 2003, American Society for Microbiology. All Rights Reserved.
Vol. 23, No. 17
Functional Dissection of Eyes absent Reveals New Modes of
Regulation within the Retinal Determination Gene Network
Serena J. Silver,1,2Erin L. Davies,1† Laura Doyon,1and Ilaria Rebay1,2*
Whitehead Institute for Biomedical Research,1and MIT Department of Biology,2
Cambridge, Massachusetts 02142
Received 26 December 2002/Returned for modification 18 February 2003/Accepted 19 May 2003
The retinal determination (RD) gene network encodes a group of transcription factors and cofactors
necessary for eye development. Transcriptional and posttranslational regulation of RD family members is
achieved through interactions within the network and with extracellular signaling pathways, including epi-
dermal growth factor receptor/RAS/mitogen-activated protein kinase (MAPK), transforming growth factor
?/DPP, Wingless, Hedgehog, and Notch. Here we present the results of structure-function analyses that reveal
novel aspects of Eyes absent (EYA) function and regulation. We find that the conserved C-terminal EYA
domain negatively regulates EYA transactivation potential, and that GROUCHO-SINE OCULIS (SO) inter-
actions provide another mechanism for negative regulation of EYA-SO target genes. We have mapped the
transactivation potential of EYA to an internal proline-, serine-, and threonine-rich region that includes the
EYA domain 2 (ED2) and two MAPK phosphorylation consensus sites and demonstrate that activation of the
RAS/MAPK pathway potentiates transcriptional output of EYA and the EYA-SO complex in certain contexts.
Drosophila S2 cell two-hybrid assays were used to describe a novel homotypic interaction that is mediated by
EYA’s N terminus. Our data suggest that EYA requires homo- and heterotypic interactions and RAS/MAPK
signaling responsiveness to ensure context-appropriate RD gene network activity.
Proper development of an organism requires a formidable
amount of cell-cell communication, wherein successful infor-
mation transfer is effected by signaling cascades that ultimately
alter the gene expression profile of a cell. Additionally, cells
must integrate signals from multiple pathways to coordinate
morphogenesis and differentiation in a spatially and temporally
appropriate manner. Examples of this combinatorial control
paradigm have been elucidated by using the Drosophila eye as
a model system and have demonstrated that unique combina-
tions of general and tissue-specific transcription factors are
required to specify and maintain distinct cell fates (15, 18, 43).
Components of the retinal determination (RD) gene net-
work, which includes twin-of-eyeless (toy), eyeless (ey), eyes ab-
sent (eya), sine oculis (so), and dachshund (dac), are essential
for eye fate specification in metazoans. The RD gene network
collectively encodes a cohort of nuclear transcription factors
and/or cofactors whose expression is regulated by a conserved
hierarchy of transcriptional regulation, such that TOY acti-
vates ey expression, EY induces so and eya, and EYA turns on
dac (9). In addition to assuming pivotal roles during visual
system development, the RD genes function in a variety of
other contexts, including gonadogenesis (2), myogenesis (19),
limb formation (16, 44), neurogenesis (30), and the cell cycle
(29). Consequently, null mutations are lethal and exhibit com-
plex phenotypes that are reflective of the pleiotropic roles
assumed by RD network proteins during development (6, 11,
27, 32). Additionally, the expression patterns of the RD genes
are not wholly coincident (4), suggesting that reiterative de-
ployment of the entire RD network module is not obligatory
for the function of specific RD network proteins.
Much analysis of RD gene function and regulation has fo-
cused on the visual system, particularly in Drosophila, where
gene activity can be manipulated without compromising viabil-
ity or fertility of the animal. RD genes are best known for the
“eyeless” phenotype associated with eye-specific hypomorphic
mutations and for the ability to induce formation of ectopic eye
tissue upon overexpression (5, 9, 13, 17, 37, 41). Use of these
two assays has revealed a complex system of positive feedback
loops superimposed on the defined linear hierarchy of tran-
scriptional regulation, which most likely amplifies and stabi-
lizes expression of the RD gene products.
eya is the founding member of a novel gene family charac-
terized by a highly conserved C-terminal motif that contains
both SO (31) and DAC (9) binding sites, termed the EYA
domain (ED; amino acids [aa] 486 to 760) (Fig. 1A). Verte-
brate homologs, such as murine EYA1-4, are strikingly similar
in their ED, yet their N termini, with the exception of a small
tyrosine-rich region, the ED2 (Fig. 1A), are largely divergent
(45, 47). so and vertebrate Six family genes encode nuclear
proteins with a homeobox-type DNA binding domain and the
conserved SIX domain (23). The latter is an important medi-
ator of EYA-SO interactions (37). DAC, a novel nuclear pro-
tein, and its vertebrate DACH counterparts contain two re-
gions of high homology: DACH-box-N and DACH-box-C (32).
Like SO, DAC has been shown to physically interact with the
ED, via DACH-box-C (19).
Analysis of loss-of-function eya mutants in Drosophila and in
human patients suffering from branchio-oto-renal (BOR) syn-
drome underscores the importance of EYA-SO and/or EYA-
DAC interactions in vivo. BOR syndrome, a disease charac-
terized by craniofacial, ear, and kidney defects, arises from
* Corresponding author. Mailing address: Whitehead Institute for
Biomedical Research, Nine Cambridge Center, Cambridge, MA 02142.
Phone: (617) 258-6399. Fax: (617) 258-0376. E-mail: firstname.lastname@example.org.
† Present address: Department of Developmental Biology, Stanford
University School of Medicine, Stanford, CA 94305.
mutations in the human Eya1 gene (1, 8). In Drosophila, point
mutations in the conserved ED appear to cause loss-of-func-
tion phenotypes by impairing EYA’s ability to interact with SO
and DAC (8). Similarly, a recent study has shown that human
BOR alleles that map to the ED also have impaired interac-
tions with SIX and DACH family members, emphasizing the
evolutionarily conserved importance of interactions between
these three RD family members in vivo (35).
The combined physical interaction, colocalization, and ge-
netic data have led to a model whereby EYA and SO together
constitute a functional transcription factor, with EYA provid-
ing the activation domain and SO contributing the DNA bind-
ing moiety. Consistent with this, in mammalian cell culture,
SIX2, SIX4, and SIX5 are able to synergize with EYA to drive
expression from a reporter construct (33). The functional con-
sequences of EYA-DAC interactions remain less well under-
stood. Although initial characterization of Drosophila dac and
its vertebrate counterparts, Dach1 and Dach2, failed to identify
a DNA binding domain, it has been postulated that DACH-
box-N may encode a novel DNA binding motif similar to the
winged helix/forkhead subgroup (24). Thus, like SO, DAC may
play a role in recruiting EYA to the promoters of target genes
In addition to the complex interactions observed within
the RD gene network, numerous signaling pathways, includ-
ing epidermal growth factor receptor (EGFR)/RAS/mito-
gen-activated protein kinase (MAPK) (20), transforming
growth factor-? (TGF?)/Dpp (10, 38), Wingless (3), Hedge-
hog (12), and Notch (2, 28), have been shown to interact
genetically with members of the RD network. However, with
the exception of the EGFR pathway (20), little is known
about how extracellular signaling pathways regulate RD
gene expression and/or activity.
In order to better understand RD gene network regulation,
we have performed an extensive structure-function analysis of
the EYA protein. Using a series of Drosophila S2 cell-based
transcriptional activation assays, we have defined a proline-,
serine-, and threonine-rich (P/S/T-rich) region of EYA, en-
compassing both the MAPK phosphorylation consensus sites
and the tyrosine-rich ED2, that is necessary for EYA transac-
tivation potential in cell culture and for ectopic eye induction
ability in vivo. We demonstrate that RAS/MAPK signaling can
positively regulate EYA transactivation and that GROUCHO
is a potent repressor of EYA-SO-mediated transcription
through its interactions with SO. Finally we show that EYA is
able to self-associate, adding yet another layer of functional
FIG. 1. The N-terminus of EYA is a potent transactivator. (A) Drosophila EYA contains two conserved regions, the ED2 and ED. The
P/S/T-rich region includes both the ED2 and the two MAPK phosphoacceptor sites. (B) Gal4DBD-EYA fusions were used to assay the ability of
EYA to activate transcription from a UAS-luciferase reporter gene. Transactivation potentials were calculated by taking the luciferase/?-
galactosidase activity ratio for each construct and were plotted relative to the activity of the Gal4DBD vector alone. As shown in panel B, for
construct 2, full-length EYA can activate transcription 3.5-fold above Gal4DBD alone. The N terminus of EYA (construct 3), a construct that lacks
the ED, is a potent transactivator, activating transcription over 250-fold (note scale change on axis). Gal4DBD-EYA 1-353 (construct 4), a
truncation that contains ED2 but removes part of the P/S/T-rich region, reduces transactivation potential to only fivefold. ED2 plus part of the
P/S/T-rich region is able to activate transcription at low levels (construct 5), indicating that the entire N terminus is necessary for full transactivation
potential. Deletion of the conserved ED2 within the N terminus of EYA (construct 6) sharply reduces transactivation to 41-fold above background,
one-fifth the activity of the intact N terminus. Strikingly, deletion of the entire P/S/T-rich region (construct 7) results in complete loss of
transactivation potential. (C) Deletion of the EYA domain does not affect protein expression levels. WT, wild type.
5990 SILVER ET AL.MOL. CELL. BIOL.
complexity to the elaborate hierarchy of interactions that exists
among the RD gene products.
MATERIALS AND METHODS
Construction of transactivation assay expression plasmids. pUAST-luciferase
was constructed by amplifying the luciferase cDNA from pGL3luc (Promega)
with primers LUC1 (5?-TTGGAATTCCAACATGGAAGACGCCAAAAAC-
3?) and LUC2 (5?-TTGGGTACCTTACACGGCGATCTTTCCGC-3?), diges-
tion with EcoRI and KpnI, and insertion into the EcoRI-KpnI sites of pUAST.
All eya constructs were made with the eya1 cDNA. pRmHa3-Gal4DBD-eyawt,
S-A, and S-D/E full-length fusion constructs have a similar design. The MAPK
consensus sequence is defined as P-X-S/T-P, and Drosophila eya contains two
adjacent phosphorylation sites at aa S402 and S407. The construct referred to as
pRmHa3Gal4DBD-eyaS-Acontains two S-A point mutations in place of the
phosphoacceptor residues, and conversely, the pRmHa3-Gal4DBD-eyaS-D/E
construct contains S-D and S-E point mutations at these sites. A three-piece
ligation was performed to insert a 660-bp, N-terminal eya PCR-amplified frag-
ment and a 1.6-kb, C-terminal eya restriction fragment containing either wild-
type or mutated MAPK sites into pRmHa3-Gal4DBD, cut with KpnI and SalI.
The N-terminal PCR product was generated with primers EYA1 (5?-TGGGTA
CCTTGTATAATGTGCCGTGCTATC-3?) and EYA2 (5?-CGAAGAGTTGA
CCGCCACTG-3?) and was digested with KpnI and BamHI. BamHI-SalI
restriction fragments from pRmHa3-eya, pRmHa3-eyaS-A, and pRmHa3-eyaD/E
(constructs described previously in reference 20) were then combined with the
digested eya PCR product in a ligation with pRmHa3-Gal4DBD, cut with KpnI
Similarly, three truncated pRmHa3-Gal4DBD-eya constructs that lack the
conserved EYA domain (aa 486 to 760) were made, which contain each of the
MAPK site variants described above. Primers EYA1 and EYA3 (5?-TTGGTC
GACTTACACACTGCTGCCTCCGCTC-3?) were used to amplify a 1.3-kb
product from a pRmHa3-eya, pRmHa3-eyaS-A, or pRmHa3-eyaD/Etemplate.
PCR products were digested with KpnI and SalI and ligated into the pRmHa3-
Gal4DBD vector. A construct encoding the first 353 aa of EYA, including the
N-terminal portion of the P/S/T-rich region, was also generated. This con-
struct, pRmHa3-Gal4DBD-eya 1-353, was generated with primers EYA1 and
EYA1545A (5?-TTGGTCGACGTAGTTGGCCGGACTGTA-3?). The 920-bp
PCR product and pRmHa3-Gal4DBD were digested with KpnI and SalI and
ligated directionally. An internal deletion construct that lacks the entire P/S/T-
rich region, pRmHa3-Gal4DBD-eya ?223-438, was made by inserting an an-
nealed, double-stranded linker with 5? BamHI and 3? KpnI sticky ends into
pRmHa3-myc-eya, cut with the aforementioned enzymes. Primers EYAD1 (5?-
GATCCATTTTGTACGGTACC-3?) and EYAD2 (5?-CGTACAAAATG-3?)
were annealed and used in the directional ligation described above to generate
pRmHa3-myc-eya ?223-438. pRmHa3-Gal4DBD-eya ?223-438 resulted from a
two-piece ligation between pRmHa3-Gal4DBD-eya, cut with BamHI and SalI,
and a complementary restriction fragment from pRmHa3-myc-eya ?223-438. A
construct encoding a truncated EYA in which ED2 (aa 318 to 353) is also
deleted, pRmHa3-Gal4DBD-eya ?ED2 ?ED, was made by the Stratagene Quik-
change site-directed mutagenesis protocol. Primers dEya 1351/1455 (S) (5?-CA
GCTGTACAGCAGTCCGTCACCGTATGCGGTCAGC3?) and dEya1351/
were used to generate pBSSK-eya ?ED2. A 600-bp BamHI-SacII eya ?ED2
fragment was then shuttled into the truncated pRmHa3-Gal4DBD-eya construct
in a directional ligation. pRmHa3-Gal4DBD-eya 318-436 was made with primers
KpnI-EYA D2 (S) (5?-TTGGGTACCTACGCCGGCTACAACAACTTC-3?)
and EYA3 to amplify a 350-bp product. The PCR product and pRmHa3-
Gal4DBD were digested with KpnI and SalI and used in two-piece ligations.
Development of S2-2H assay. The S2 cell two-hybrid (S2-2H) assay was de-
veloped as follows. A vector containing the DNA binding domain of yeast Gal4
(aa 1 to 147), pRMHa3-Gal4DBD, was constructed as follows. Gal4DBD was
amplified from pCasperUbGal4 with the primers 5?-TGGAATTCCAACATGA
AGCTACTGTCTTCTATCG-3? and 5?-TGGGTACCCGATACAGTCAACTG
TCTTTG-3?, which contain 5? and 3? EcoRI and KpnI sites, respectively. The
digested PCR product was ligated into pRmHa3, a pUC9-derived vector con-
taining a metallothionein-responsive promoter upstream of the multiple cloning
pRmHa3-Gal4AD was made by amplification of the Gal4 activation domain
(aa 768 to 874) from pCasperUbGal4 via PCR with primers Gal4AD S2302
(5?-TGGAATTCCAACATGGCCAATTTTAATCAAAGTG-3?) and Gal4AD
A2673 (5?-TTGGTACCGTATCTTCATCATCGAATAGA-3?), cut with EcoRI
and KpnI, and inserted into pRmHa3. To ensure nuclear expression of these
constructs, a nuclear localization signal (NLS) was added with these two oligo-
nucleotides, Gal4AD NLS S-50 (5?-TGTGACCCCCCCCAAGAAGAAGCGC
AAGGTGGAGGACGATGGTAC-3?) and Gal4 AD NLS A-54 (5?-CATCGT
nucleotides were annealed and then ligated into the KpnI site of pRmHa3-
Gal4AD to make pRMHa3-Gal4AD-NLS, which we refer to as pRMHa3-Gal4AD.
pRmHa3-Gal4AD-eya resulted from a three-piece ligation using KpnI-BamHI
eya and BamHI-SalI eya restriction fragments and with pRmHa3-Gal4AD cut
with KpnI and SalI. Primers SoS4 (5?-TTGGTACCTTACAGCATCCCGCCAC
AG-3?) and So1234 (5?-TTGGTCGACTCATAAGTGCTGGTACTC-3?) were
used to amplify full-length so cDNA from pBSSK so (gift of G. Mardon), with
unique 5? KpnI- and 3? SalI-cut sites. The digested PCR product was inserted
into pRmHa3-Gal4DBD in a directional, two-piece ligation to make pRmHa3-
Gal4DBD-so. To make pRmHa3-Gal4AD-so, so was cut out of pRmHa3-
Gal4DBD-so with KpnI and SalI inserted into pRmHa3-Gal4AD cut with the
same enzymes. To make dac constructs, DacS622 (5?-TTGGTACCGATTCTG
TGACAAGTGAAC-3?) and DacA1370 (5?-AAG TGCTTCAGGAAGAGCTC
G-3?) primers were used to amplify the 5? region of dac from pBSSK-dac (gift of
G. Mardon), adding a 5? KpnI site. The PCR product was cut with KpnI and StuI,
which is found internally in the amplified dac fragment. A three-piece ligation
was performed with KpnI-StuI 5?dac and a StuI-SalI restriction fragment from
pBSSK-dac, into the KpnI-SalI sites in pRmHa3-Gal4DBD. For the AD con-
struct, full length dac was cut out of pRmHa3-Gal4DBD-dac with KpnI and SalI
and ligated into those sites in pRmHa3-Gal4AD to create pRmHa3-Gal4AD-
dac. For pRmHa3-Gal4DBD-eya ?D2, the N-terminal eya ?D2 construct de-
scribed above was shuttled into full-length pRmHa3-Gal4DBD-eya by using
BamHI and SacII. For the pRmHa3-Gal4DBD-eya domain, the eya domain was
PCR amplified with primers EYADS (5?-TTGGGTACCGAACGGGTGTT
CGTCTGG-3?) and EYA DA (5?-TTGGGATCCTCATAAGAAGCCCATG
TC-3?), which contain KpnI and BamHI sites used to clone into the pRmHa3-
Construction of transcription assay expression plasmids. To construct ARE-
luciferase, luciferase cDNA was isolated as an XhoI-XbaI fragment (?1.6 kb)
from Promega pGL3-Luciferase and inserted into XhoI-XbaI sites of pBluescript
SK?. To add the hsp70 TATA box, oligonucleotides EBS link 1 (5?-CCATAT
GATCTGCAGAGGGTATATAATGC-3?) and EBS link 2 (5?-TCGAGCATT
ATATACCCTCTGCAGATCATATGGGTAC-3?) were annealed and inserted
into the KpnI-XhoI sites of pBSSK-luciferase. KpnI and XhoI sites were retained,
and NdeI and PstI sites and a TATA box were inserted. ARE-luciferase was made
by multimerizing the AREC3 (SIX4) binding site, as defined in reference 22, us-
ing oligonucleotides ARES (5?-TCGAGGGTGTCAGGTTGCG-3?) and AREA
(5?-TCGACGCAACCTGACACCC-3?). Oligonucleotides were annealed and
ligated and then cut with XhoI and SalI. A resultant 7-mer was cloned into the
MCS of pBSSK and then shuttled into pBSSK-TATA-luciferase. To make
pRmHa3-flag-eya, full-length eya1 cDNA was PCR amplified with primers
EYAI 5397 (5?-TTGTATAATGTGCCGTCGTATC-3?) and EYA STOP (5?-
TTTCATAAGAAGCCCATGTCGAGG-3?) and then digested with SmaI,
which cuts the eya cDNA internally. The 0.7-kb blunt/SmaI fragment was
inserted into SmaI-cut pBSSK ? Flag vector (gift from R. Fehon) to produce an
in-frame fusion of the Flag epitope with the second amino acid of EYA. A
three-piece ligation was done to join the 5? end of flag-eya (obtained as a
SacI-SmaI fragment from pBSSK-flag-eya) with the 3? end of eya (obtained from
pRMHa3-eya) as a SmaI-SalI fragment. These were inserted into SacI-SalI-cut
pRMHa3. Similarly, pBSSK-myc-eya was generated by PCR amplification with
primers EYAI and EYA STOP and blunt ligated into the StuI site of pBSSKmyc.
A three-piece ligation between the EcoRI-BamHI fragment from pBSSK-myc-
eya and BamHI and SalI from pRmHa3-eya into the EcoRI and SalI sites of
pRMHa3 resulted in pRMHa3-myc-eya. In order to construct pRMHa3-dac, first
a ClaI fragment was cut out of pBSSK-dac (gift from G. Mardon) and cloned into
pBSSK to remove most of the large 5? untranslated region (UTR) (Cla-dac). A
3? fragment was cloned into pBSSK by using BamHI and HindIII sites (HI/HIII-
dac). A full-length construct was made by four-piece ligation with EcoRI-StuI
from Cla-dac plus a StuI-BamHI fragment from pBSSK-dac plus the BamHI-
XhoI fragment from HI/HIII-dac into the EcoRI-XhoI sites of pUAST. The
entire full-length dac cDNA, including approximately 45 bp of the 5? UTR and
245 bp of the 3? UTR, was then excised with EcoRI and XhoI and ligated into the
EcoRI-SalI sites of pRMHa3. pRMHa3-myc-so was generated by PCR amplifi-
cation of full-length so cDNA by using primers So S805 (5?-TTACAGCATCC
CGCCACAGAT-3?) and So A2268 (5?-AACTAGAATCATAAGTGCTGG-3?)
and blunt end ligated into the StuI site of pBSSK-myc, resulting in an in-frame
fusion of the MYC epitope to the second amino acid of SO. This was moved into
pUAST by using EcoRI and XbaI. To move the full-length tagged so into
pRMHa3, myc-so was cut out of pUAST-myc-so with XbaI, blunted with Klenow
fragment, and then digested with EcoRI and subcloned into EcoRI-SmaI sites of
VOL. 23, 2003 STRUCTURE-FUNCTION ANALYSIS OF Eyes absent5991
pRMHa3. A full-length groucho expression construct, pMT-GROUCHO, was
generated by PCR amplification of the N terminus of GROUCHO, using the
groucho cDNA clone LD33829 as a template and primers Groucho-start (5?-A
TGAATTCAACAACATGTATCCCTCACCGG-3?) and Groucho-A879 (5?-T
GTGCGATACTTCTCACGATCGG-3?), digestion with EcoRI and XbaI, and
ligation with an XbaI-XhoI fragment from LD33829 into EcoRI-SalI-cut pRMHa-3.
All regions generated by PCR were verified by sequencing, and all constructs
were tested for expression and localization in S2 cells by immunohistochemistry.
Further subcloning details are available upon request.
Co-IP and Western blots. Transfected cells were harvested and then lysed by
rocking at 4°C for 20 min in 1 ml of lysis buffer (100 mM NaCl, 50 mM Tris [pH
7.5], 2 mM EDTA, 2 mM EGTA, 1% NP-40, one Roche Complete, Mini
protease inhibitor cocktail tablet per 10 ml). Clarified lysates were subjected to
immunoprecipitation (IP) with anti-Flag-conjugated agarose beads (Sigma) for
1.5 h at 4°C. Beads were washed three times with lysis buffer. The immunopre-
cipitates were boiled in 40 ?l of 2? sodium dodecyl sulfate (SDS) buffer, and
Western blotting was carried out as previously described (34) with mouse anti-
MYC (1:300), mouse anti-GRO (1:50), and Rb anti-Flag (1:5,000) antibodies.
For Western blots to analyze protein levels, half of the transfected cells were
lysed as described below for ?-galactosidase assays, and the other half were
pelleted, vortexed briefly, and resuspended in 40 ?l of 2? SDS buffer. Trans-
fection efficiency was measured by ?-galactosidase assays, and appropriate
amounts of crude lysate were run on the gel. Efficiency was confirmed by West-
ern blotting for ?-galactosidase (Rb anti-?-galactosidase, 1:20,000). Protein lev-
els were examined with GP anti-EYA (1:10,000), mouse anti-MYC (1:300), and
Rb anti-Flag (1:5,000) antibodies.
Generation of transgenic lines. pRmHa3-flag-eya and pRmHa3-myc-eya ?D2
were cut with SmaI and SalI, and a directional ligation was performed to create
pRmHa3-flag-eya ?D2. Likewise, a SmaI-SalI double digest was performed on
pRmHa3-myc-eya ?223-438 to generate pRmHa3-flag-eya ?223-438. EcoRI-SalI
double digests were used to excise the flag-eya, flag-eya ?D2, and flag-eya ?223-
438 cDNAs from the pRmHa3 vector for insertion into pUAST, cut with EcoRI
and XhoI. pUAST-flag-eya, pUAST-flag-eya ?D2, and pUAST-flag-eya ?223-438
were subsequently used to generate transgenic lines as previously described (39).
S2 cell transactivation and transcription assays. Drosophila S2 cells were
transiently transfected with calcium phosphate as described previously (36). Each
construct (2.5 ?g) indicated was transfected for 6 h along with 2.5 ?g of pUAST-
luciferase or 10 ?g of ARE-luciferase as the reporter gene and with 1 ?g of pActin
5.1-V5His-lacZ (Invitrogen) to normalize for transfection efficiency. Cells were
allowed to recover for 17 h, whereupon expression was induced by addition of 0.1
M CuSO4. After 24 h, cells were harvested by spinning at 200 ? g for 1 min and
lysed by rocking at 4°C for 20 min in 250 ?l of lysis buffer (Tropix/Applied
Biosystems). Samples were subsequently microcentrifuged at 20,800 ? g for 1
min at 4°C, and supernatants were transferred to fresh tubes. Luciferase and
?-galactosidase activities were quantified with a Tropix/Applied Biosystems lu-
ciferase assay kit or Galacto-Star assay kit. Assays were performed in triplicate
on whole-cell extracts according to the manufacturer’s instructions (TROPIX/
Applied Biosystems). A minimum of four independent transfections were per-
formed for each condition. The average luciferase or ?-galactosidase signal for
Gal4DBD/pUAST-luciferase or ARE-luciferase alone was set to 1, and the exper-
imental averages were normalized relative to this value. Data were analyzed and
graphed with Microsoft Excel. Error bars denote one standard deviation above
and below the mean for each construct. In Fig. 3, 5, and 6, SO is always tagged
with the Flag epitope. In Fig. 3, EYA constructs are tagged with the Flag epitope,
and in Fig. 5 and 6, EYA constructs are tagged with the MYC epitope.
EYA functions as a transactivator. To address the question
of whether EYA can function as a transcriptional coactivator,
we took advantage of the well-characterized yeast transcription
factor Gal4 and its target sequence, UAS (14), to design an
assay for transactivation in Drosophila S2 cells. The DNA bind-
ing domain of Gal4 (Gal4DBD) was fused in frame to the eya
coding region (Gal4DBD-EYA) (Fig. 1B) and subcloned into
a vector containing a metallothionein promoter, which allows
inducible expression in S2 cells. Gal4DBD-EYA fusion pro-
teins were tested for their ability to activate expression of a
UAS-luciferase reporter gene; cotransfection of a constitutively
expressed actin-LacZ plasmid enabled us to normalize activity
levels based on transfection efficiency. Immunohistochemistry
with anti-Gal4DBD antibodies confirmed the expression and
nuclear localization of all Gal4DBD-EYA fusion proteins in
transfected S2 cells (data not shown).
First, we tested the full-length eya coding region fused in-
frame to the Gal4DBD (Fig. 1B). As shown in Fig. 1B (con-
structs 1 and 2), Gal4DBD-EYA exhibits 3.5-fold-greater ac-
tivity than the Gal4DBD alone. A series of deletion and
truncation constructs were designed to define a minimal and
sufficient domain for this activity. Strikingly, a fusion protein
expressing the N-terminal 485 aa of EYA but lacking the
conserved ED, Gal4DBD-EYA ?ED, displays an approxi-
mately 70-fold increase in transactivation potential relative to
the full-length Gal4DBD-EYA construct (Fig. 1B, construct 3
versus construct 2 [note scale]). The converse fusion protein
expressing only the C-terminal ED, Gal4DBD-EYA Domain
(aa 486 to 760), does not transactivate (data not shown). This
difference is likely not due to changes in protein stability, as
EYA and EYA ?ED are expressed at similar levels (Fig. 1C).
The discrepancy in activity levels of the full-length versus
Gal4DBD-EYA ?ED chimeras suggests that the ED may
function as an autoregulatory inhibitor in this context.
A P/S/T-rich region is critical for EYA transactivation. EYA
?ED was dissected further to determine the regions critical for
transactivation. EYA ?ED contains the conserved ED2 (aa
318 to 353), a tyrosine-rich region that has not been function-
ally characterized (Fig. 1A). ED2 lies within a larger P/S/T-rich
region (aa 223 to 438) that includes two consensus MAPK
phosphorylation sites previously shown to be important for
EYA regulation in vivo (20).
We found that Gal4DBD-EYA 2-353, an N-terminal con-
struct that contains the ED2 but truncates the P/S/T-rich re-
gion, exhibits very low transactivation activity, suggesting a
critical requirement for the latter domain (Fig. 1B, construct
4). Consistent with these results, the fusion protein that con-
tains the last two-thirds of the P/S/T-rich region, Gal4DBD-
EYA 318-436, exhibits a 22-fold increase in transactivation
potential relative to the Gal4DBD-alone control (Fig. 1B, con-
struct 5 versus construct 1). Therefore the P/S/T-rich region of
EYA is critical but not entirely sufficient for transcriptional
coactivation, and upstream regions of the protein (aa 2 to 223)
are required to achieve maximal transactivation levels.
Interestingly, deletion of the conserved ED2 (Gal4-DBD-
EYA ?ED2, ?ED; aa 2 to 317 and 353 to 485; Fig. 1B, con-
struct 6) results in a 6-fold reduction in transactivation poten-
tial relative to that of EYA ?ED (Fig. 1B, construct 3), yet
shows a 41-fold increase relative to the Gal4DBD control (Fig.
1B, construct 1). Although ED2 is not essential for EYA trans-
activation per se, our results suggest that it is needed to achieve
maximal levels and ascribes a function to this previously un-
Although the full-length EYA is a weaker transactivator in
this assay than any of the “active” deletion constructs that lack
the C-terminal EYA domain, we wanted to determine whether
the P/S/T-rich region is necessary for the transactivation po-
tential of full-length EYA. To address this question, the entire
P/S/T-rich region, including ED2, was deleted to generate
Gal4DBD-EYA ?223-438. We found that this internal dele-
tion abrogates transcriptional activation in full-length EYA
(Fig. 1B, construct 7 versus construct 3). An identical result,
5992 SILVER ET AL.MOL. CELL. BIOL.
namely a complete lack of transactivation, was obtained when
the C-terminal EYA domain was also deleted from construct 7
(data not shown).
RAS/MAPK signaling increases EYA transactivation poten-
tial. Because the P/S/T region important for EYA transactiva-
tion potential also contains the MAPK phosphorylation sites
(Ser402 and Ser407) shown to affect EYA activity in vivo, we
asked whether RAS signaling modulates EYA’s transactivation
potential. To examine this possibility, we cotransfected acti-
vated ras (rasV12) with the Gal4DBD-EYA fusion constructs.
We did not observe an increase in the transactivation potential
of Gal4DBD-EYA (data not shown), but observed a variable
but significant increase in Gal4DBD-EYA ?ED transactiva-
tion potential in the presence of RASV12(Fig. 2). In repeated
trials, this increase in transactivation ranged from 10 to 66%,
the latter of which is depicted in Fig. 2.
Because we had previously shown that site-specific muta-
tions in the MAPK consensus phosphoacceptor residues affect
EYA function in vivo (20), we wanted to test whether these mu-
tations might also influence transactivation in the Gal4DBD
assay system. Mutation of the two MAPK consensus sites S402
and S407 to alanine (EYAS-A) significantly reduces EYA
activity in ectopic eye induction assays. Conversely, mutation
of the phosphoacceptor residues to aspartic or glutamic acid
(EYAS-D/E) produces a hyperactive protein, presumably by
mimicking its phosphorylated state (20). Surprisingly, we found
that both EYAS-Aand EYAS-D/Epoint mutations increased
transactivation activity of Gal4DBD-EYA ?ED to slightly
higher levels than were seen upon addition of RASV12(Fig. 2).
Because EYA transactivation potential is increased upon
cotransfection of rasV12or mutation of the MAPK phosphoac-
ceptor residues, it seems likely that these residues are impor-
tant for proper regulation of EYA transactivation. Our finding
that in this system both EYAS-Aand EYAS-Dmutations result
in increased transactivation relative to wild-type EYA may re-
flect the complexity of the consequences of RAS/MAPK sig-
naling. However, it is important to note that the chimeric
Gal4DBD-EYA fusion proteins and/or the truncations we en-
gineered may have an altered conformation relative to native
EYA, such that MAPK phosphorylation events have different
consequences in this context.
ARE-luciferase is responsive to the EYA-SO transcription
factor. The synthetic assay system described above enabled us
to map the regions crucial for EYA transactivation activity to
the ED2 and the surrounding P/S/T-rich region. RAS/MAPK-
dependent effects on EYA transactivation levels indicate that
RAS signaling can activate EYA transactivation but that such
effects may not afford themselves to straightforward interpre-
tation. In order to examine this question in a more native
system, we designed a transcription assay system to measure
the transactivation potential of EYA-SO complexes.
Currently, the genomic DNA target sequences bound by
Drosophila SO are not known, but several studies in mam-
malian cells and tissues have identified SIX family response
elements. In particular, the Na, K-ATPase ?1 subunit gene
(ATP1a1) regulatory element (ARE) has been shown to re-
spond to SIX family members SIX2, SIX4 and SIX5 in vivo
and in vitro (22). Because SO and SIX2 belong to the same
SIX family subgroup (SIX1/2) (23) and their homeodomains
are 93% identical (40), we reasoned that they are likely to bind
similar sequences. We therefore multimerized the core ARE
binding site (see Materials and Methods) and placed this en-
hancer in front of a minimal promoter followed by luciferase
cDNA, which we will refer to as ARE-luciferase.
We found that ARE-luciferase is responsive to cotransfection
of eya and so, and together they activate ARE-luciferase 27-fold
over the reporter alone (Fig. 3A). ARE-luciferase is not
appreciably activated alone nor upon transfection of eya or
so individually (Fig. 3A). Therefore, ARE-luciferase activa-
tion provides a measure of EYA’s efficacy as a transcriptional
coactivator when bound to SO. We also asked whether the
addition of DAC affects ARE-luciferase transcription and
found that DAC did not affect the transactivation potential of
the EYA-SO transcription factor (data not shown).
The P/S/T-rich region of EYA is necessary for transactiva-
tion but not for EYA-SO interactions. We used the ARE-
luciferase reporter and full-length EYA and SO to ask whether
the ED2 and the P/S/T-rich region determined to be critical in
the synthetic Gal4DBD assay system are essential in a more
physiologically relevant context. In this assay, we used other-
wise full-length EYA constructs that lack either the ED2 or the
entire P/S/T-rich region, EYA ?ED2 and EYA ?223-438. As
FIG. 2. RAS/MAPK signaling activates EYA transactivation.
Transactivation via Gal4DBD-EYA ?ED is increased significantly but
variably upon the addition of RASV12. Mutations of the MAPK phos-
phoacceptors to alanine (EYAS-A) to prevent phosphorylation, or to
aspartic or glutamic acid (EYAS-D/E) to mimic phosphorylation, both
increase Gal4DBD-EYA ?ED transactivation activity, suggesting that
these sites are important for regulation of EYA transactivation poten-
tial. Taken with the increase seen upon addition of RASV12, we con-
clude that RAS signaling can increase EYA transactivation potential.
VOL. 23, 2003STRUCTURE-FUNCTION ANALYSIS OF Eyes absent 5993
shown in Fig. 3A, deletion of the ED2 reduces transactivation
to only 14-fold relative to that in controls, while deletion of the
entire P/S/T-rich region, EYA ?223-438, virtually eliminates
One possible explanation for this lack of activity might be a
decrease in protein expression levels in the EYA deletion
constructs. In order to test this, we performed quantitative
Westerns blots to assay tagged EYA protein levels. Transient
transfections were normalized for efficiency by using ?-galac-
tosidase assays, and appropriate amounts of lysate were
loaded. We found that while the ?ED2 deletion did not alter
protein expression levels (Fig. 3B, lane 2), EYA ?223-438 was
less abundantly expressed than full-length EYA (Fig. 3B, lane
3). In order to correct for this difference, we transfected suf-
ficient EYA ?223-438 plasmid to produce protein levels above
that of full-length EYA (Fig. 3B, lanes 4 and 5). Using these
larger amounts of EYA ?223-438 plasmid, we still observed
only low levels of transactivation (Fig. 3A), indicating that
deletion of this region of EYA compromises activity.
To confirm that the loss of transactivation potential seen in
these deletions was not due to a loss of EYA-SO binding, we
developed a system to test for direct protein-protein interac-
tions in S2 cells, which we term S2-2H assays (see Materials
and Methods for details). Use of Drosophila cultured cells,
rather than the more common yeast or mammalian cell-based
systems (14, 15), allows interactions between Drosophila pro-
teins to be assayed in a more physiologically native environ-
ment, thereby increasing the probability that necessary cofac-
tors and/or protein modifications are present.
Using our S2-2H assay, Gal4DBD-EYA, EYA ?ED2, and
EYA ?223-438 fusions were each tested for interaction with a
Gal4AD-SO fusion protein. As shown in Fig. 3C, Gal4AD-SO
is able to interact with all three EYA proteins and demon-
strates even stronger interactions with Gal4DBD-EYA ?223-
438. This, combined with our Western analysis, indicates that
the reduced transactivation observed upon deletion of the ED2
and the P/S/T-rich region (Fig. 3A) reflects a change in activity
levels of the EYA-SO transcription factor rather than simply
the loss of EYA protein or loss of the ability to form an
We confirmed that the P/S/T-rich region and ED2 are nec-
essary for EYA activity in vivo by assaying the ability of these
deletions to induce ectopic eyes when overexpressed with the
UAS/Gal4 system (7). We found that when driven by the
57A1dpp-Gal4 driver, EYA ?ED2 and EYA ?223-438 exhib-
ited drastically reduced activity relative to wild-type EYA (Fig.
3D), although all constructs were expressed at comparable
levels as assayed by Western blots of Ub-Gal4-driven expres-
sion of UAS-EYA constructs in embryos (data not shown).
Thus, ED2 and the P/S/T-rich region are critical for EYA
function in vivo.
DAC does not interact with EYA or SO in S2-2H assays.
Because DAC did not affect EYA-SO-mediated transcription
of ARE-luciferase, we asked whether DAC was able to interact
with EYA or SO according to our S2-2H system. Gal4DBD
and Gal4AD fusions were made with full-length EYA and
DAC, and tests were performed in both directions. Surpris-
ingly, no interaction was observed in either direction (data
shown in one direction, Fig. 3C). We also asked whether SO
might be required to nucleate an EYA-DAC complex, but did
FIG. 3. ARE-luciferase is responsive to the EYA-SO transcription
factor. (A) EYA and SO alone do not affect transcription of the ARE-
luciferase reporter gene, but together can activate transcription 27-fold.
Full-length EYA ?ED2 is unable to fully activate transcription of this
gene, and neither is a construct missing the entire P/S/T-rich region,
EYA ?223-438. This construct was not expressed at the same level
as wild-type EYA (EYAWT), so we transfected two (??) and three
(???) times the amount of plasmid to raise protein levels to and
above the levels of EYA. These levels still do not activate ARE-
luciferase. (B) Quantitative Western blotting shows that EYA (lane 1)
and EYA ?ED2 (lane 2) are expressed at similar levels, while the same
amount of EYA ?223-438 plasmid (lane 3) is not. However, transfec-
tion of two and three times more EYA ?223-438 plasmid results in
robust expression, as shown in lanes 4 and 5. (C) EYA and SO show a
strong interaction in an S2-2H assay. The deletions in EYA do not
affect interactions with SO, and in fact, EYA ?223-438 appears to have
a stronger interaction with SO. DAC does not interact with EYA in
S2-2H assays. (D) Overexpression of EYA using the 57A1dpp-Gal4
driver causes ectopic eye induction in over 98% of animals (n ? 413).
Overexpression of EYA ?ED2 results in ectopic eyes in only 35% of
animals (n ? 506). Deletion of the entire P/S/T-rich region, EYA
?223-438, results in a protein that can only rarely induce ectopic eyes,
seen in only 1.5% of animals examined (n ? 204).
5994 SILVER ET AL.MOL. CELL. BIOL.
not observe an interaction (data not shown). To determine
whether RAS/MAPK signaling might be required for forma-
tion of an EYA-DAC complex, we cotransfected RasV12or
used the EyaS-Aand EyaS-Dconstructs described above, but still
did not observe an EYA-DAC interaction (data not shown).
SO and DAC also fail to interact in the S2-2H assay (data not
EYA-EYA interactions are mediated via the N terminus.
Because the RD gene network is known to function in reiter-
ative feedback loops, we wondered if EYA might interact with
itself to potentiate or restrict function. Using our S2-2H assay,
we found that full-length EYA shows a significant interaction
with itself (using Gal4DBD-EYA and Gal4AD-EYA; Fig. 4),
at sevenfold above background. We then used our series of
deletion and truncation constructs fused to the Gal4DBD do-
main coexpressed with the Gal4AD-EYA fusion to ask which
domain mediates this EYA-EYA interaction. In order to dis-
tinguish between the activity of the fusion alone and the S2-2H
interaction, we transfected each alone and with AD-EYA
As shown in Fig. 4, EYA 2-317 (construct 3) shows slightly
greater interaction with AD-EYA than full-length EYA, while
a construct including ED2, EYA 2-353 (construct 4), shows a
striking increase in interaction relative to full-length EYA.
EYA 223-353 (construct 5) shows interaction levels compara-
ble to those of full-length EYA; however, deletion of ED2 in
the context of full-length EYA also results in a striking in-
crease in EYA-EYA interaction (Fig. 4, construct 6). The ED,
which mediates EYA-SO interactions, does not interact above
background level with full-length EYA (Fig. 4, construct 7).
This result, coupled with the striking increase observed be-
tween EYA 2-353 and EYA, leads us to conclude that EYA-
EYA interactions are likely mediated by amino acids in the N
terminus and that ED2 (aa 318 to 353, Fig. 1A) and the regions
immediately following it, including the MAPK phosphoryla-
tion sites, while not necessary for this interaction, may strongly
potentiate it. We could not test other N-terminal EYA con-
structs in this assay, such as EYA ?ED and EYA ?ED2?ED,
because when expressed alone they strongly activate the re-
porter (Fig. 1B). Because this interaction maps to regions we
describe above as crucial for full EYA transactivation potential
(Fig. 1B) and for EYA function in vivo (Fig. 3D), it seems
likely that the EYA-EYA interaction is functionally significant
and could contribute to regulated transcriptional activity.
Phosphorylation increases EYA transactivation potential in
the context of the EYA-SO transcription factor. Having estab-
lished an assay in which full-length EYA acts as a transcrip-
tional coactivator when complexed to SO, we returned to the
question of whether EYA activity is regulated by RAS/MAPK
signaling. Although activated RAS did not affect EYA-SO
mediated transcriptional regulation, nor did mutation of the
EYA phosphoacceptor sites to alanine (Fig. 5A), we found
that the EYAS-D/E-SO complex consistently exhibited a 50%
increase in transactivation levels relative to the EYA-SO com-
plex (Fig. 5A). This result is consistent with previously report-
ed transgenic analyses, wherein overexpression of EYAS-D/E
led to stronger and more penetrant phenotypes than did EYA
(20). Analysis of protein levels in flies expressing EYAS-Aor
EYAS-D/Etransgenes indicated no differences in protein sta-
bility (20), and we repeated these results in cell culture with
FIG. 4. EYA-EYA interactions are mediated by the EYA N terminus. Using the S2-2H system, Gal4DBD-EYA fusions shown on the left were
coexpressed with (?) and without (?) full-length Gal4AD-EYA. Full-length EYA interacts with itself sevenfold above the background level. This
interaction is mediated by aa 223 to 317, because all constructs that contain this minimal region can interact with Gal4AD-EYA. Strikingly,
Gal4DBD-EYA 2-353 and Gal4DBD-EYA ?ED2 (constructs 4 and 6) show a more than threefold-stronger interaction than that of full-length
EYA. The Gal4DBD-ED fusion (construct 7) does not interact with Gal4AD-EYA, consistent with our finding that the EYA N terminus mediates
VOL. 23, 2003 STRUCTURE-FUNCTION ANALYSIS OF Eyes absent5995
quantitative Western blots as shown in Fig. 5B, where wild-
type and mutant EYA proteins are expressed at the same
levels. The S2-2H system was used to rule out the possibility
that the transactivation increase seen with the EYAS-D/E-SO
complex results from an increase in EYA-SO interactions
(data not shown). Therefore, the increase in transactivation
exhibited by the EYAS-D/E-SO complex suggests that phos-
phorylation directly increases EYA’s transactivation potential.
Because we have seen this result only upon mutation of the
phosphoacceptor site, but not in response to RAS/MAPK ac-
tivation, it remains possible that a different signaling pathway
mediates this phosphorylation event. However, previously re-
ported genetic and biochemical evidence (20) indicates that
RAS/MAPK signaling is responsible for phosphorylation of
this site in vivo. Thus we favor the interpretation that EYA
transactivation is potentiated by MAPK phosphorylation, as
evidenced by the increased activity of the EYAS-D/E-SO com-
plex, and that the lack of response to RAS stimulation likely
reflects a more complex role for RAS within the RD gene
GROUCHO is a repressor of the EYA-SO transcription
factor. Recent studies of zebra fish (26), mice (46), and
medaka fish (31) have revealed a functional role for SIX3
interactions with GROUCHO (GRO), a transcriptional core-
pressor. One of these studies also demonstrates weak interac-
tions between SO and a murine GROUCHO homolog, GRG5
(46). Because SO belongs to a different class of SIX homologs
and the SIX3/SIX6 families are distinct from other families in
that they do not interact with EYA or EYA homologs, we
wanted to ask if this SO-GRG5 interaction indicated a func-
tional role for GRO in regulation of the EYA-SO transcription
factor. We found that coexpression of GRO strongly reduces
but does not eliminate activation via the EYA-SO transcrip-
tion factor. As shown in Fig. 6A, EYA-SO activates transcrip-
tion more than 33-fold, while coexpression of GRO abrogates
activation to only 20-fold.
This corepressor function of GRO may be mediated by
interactions with SO through the previously characterized en-
grailed homology 1 (eh1) domains within the SIX domain (26)
or may be mediated through an eh1 domain found within the
ED (Z. Paroush, personal communication). In order to address
this question, we performed coimmunoprecipitation (co-IP) to
look at direct interactions between GRO and EYA or SO.
Strikingly, we found that GRO can co-IP with SO alone, but
FIG. 5. The EYA-SO transcription factor is regulated by phosphor-
ylation. (A) As shown in Fig. 3, the EYA-SO transcription factor can
activate expression of ARE-luciferase. This expression is not affected by
addition of RASV12nor upon mutation of the MAPK phosphoacceptor
sites to alanine (EYAS-A). A striking increase in activation is seen
when the EYAS-D/Ephosphomimetic mutant is used, showing that
phosphorylation acts to increase EYA transactivation potential. That
RASV12itself does not produce the same increase on wild-type EYA
(WT) in this assay suggests that RAS signaling may have multiple
effects on the RD gene network and in particular may negatively
regulate SO. (B) EYA phosphoacceptor mutations do not affect pro-
tein expression levels.
FIG. 6. The EYA-SO transcription factor is negatively regulated by
interactions with GROUCHO. (A) When coexpressed with EYA and
SO, GRO is a repressor of the EYA-SO transcription factor. (B) Lanes
1 and 2 show that MYC-EYA and GRO are not pulled down by
anti-Flag beads. In lane 3, IP of SO can co-IP EYA. Lane 4 shows that
IP of SO can co-IP EYA but not GRO; however, in lane 5, we see that
without EYA, GRO can associate with SO. Thus, co-IP of GRO with
SO is disrupted by cotransfection of EYA. This result does not seem to
be due to direct competition between EYA and SO for GRO, because
IP of EYA cannot co-IP GRO. All proteins were expressed at similar
levels in crude cell lysates (data not shown).
5996 SILVER ET AL.MOL. CELL. BIOL.
not in the presence of EYA (Fig. 6B, lanes 5 and 4). This is not
due to competition with EYA for GRO binding, because IP
of EYA cannot co-IP GRO (Fig. 6B, lane 6). We therefore
propose an additional negative regulatory mechanism for
EYA-SO targets, whereby in the absence of high levels of
EYA, GRO interactions with SO lead to repression and down-
regulation of target genes. Thus, SO may function both as a
transcriptional activator and repressor dependent upon the
context-specific expression levels of particular cofactors.
The RD gene network encodes proteins that operate in
multiple contexts to effect differentiation of various cell types.
Input from extracellular signaling pathways, such as the RAS/
MAPK cascade, may provide instructive, context-dependent
cues that regulate the expression and/or activity of RD gene
family members. We have shown that Drosophila EYA is a
potent transactivator, either on a heterologous promoter or in
conjunction with SO, and that this activity maps to an internal
P/S/T-rich region encompassing the ED2 and MAPK consen-
sus sites. This activity is negatively regulated by the ED and
positively modulated by phosphorylation, likely through RAS/
MAPK signaling. We also provide evidence for direct EYA-
EYA interactions and demonstrate that the ED2 may be crit-
ical in this context. Together our results suggest a complex
cooperation and interplay among the distinct structural motifs
of EYA that reflects the importance of proper regulation of
the RD gene network.
Our transcription assay results correlate well with those ob-
tained in mammalian cell culture studies of murine EYA ho-
mologs, mEYA1 to -4, which showed that their N termini can
function as transactivators on a heterologous promoter (44).
Our functional dissection of Drosophila EYA enables us to
propose a role for a second, and previously uncharacterized,
conserved domain in EYA, ED2, in mediating EYA transac-
tivation potential. The P/S/T-rich region surrounding ED2,
which includes two MAPK phosphorylation consensus sites, is
absolutely necessary for EYA transactivation, and both ED2
and the P/S/T-rich region are essential for EYA function in
Ectopic expression of EYA is associated with a wide range of
deleterious phenotypes (20), suggesting that EYA activity must
be precisely regulated to ensure appropriate growth and de-
velopment. Our observation that the conserved EYA domain
functions as an autoregulatory inhibitor of EYA transactiva-
tion potential suggests that regulation of and by this domain is
critical for proper EYA function. Relief of this inhibition in
vivo may require cofactor binding or protein modification.
Alternatively, the inhibition mediated by the ED might be
dependent on interactions with an unidentified negative regu-
lator of EYA. In this context, it would be interesting to ask
whether any of the BOR alleles that map to the ED (35) affect
the transactivation potential of human EYA1.
Another negative regulatory component of the EYA-SO
transcription factor arises from our finding that striking repres-
sion is achieved in the presence of GROUCHO. Furthermore,
we provide evidence that SO-GRO interactions are disrupted
by EYA, providing an intriguing model for SO target gene
regulation. Because EYA and SO are not entirely coexpressed
and GRO is widely expressed (25), a SO-GRO complex may
provide tight regulation of EYA-SO targets, functioning as an
“off” switch in the absence of EYA. This may explain the lack
of ectopic eye induction seen upon overexpression of SO alone
(37) compared to that with EYA alone (5) and may be a
mechanism for the cooperativitity observed when EYA and SO
are coexpressed (37), since EYA is necessary to overcome
Our finding that DAC does not interact with EYA or SO in
S2-2H assays is surprising, but it does not preclude their inter-
action in vivo. It is possible that the EYA-DAC interaction
requires cofactors or modifications not made in Drosophila S2
cell culture or that some factor present in S2 cells inhibits the
interaction. As well, the use of Gal4DBD and Gal4AD fusions
may in some way disrupt an EYA-DAC interaction. It will be
interesting to ask whether cotransfection of one known mam-
malian cofactor, CREB binding protein (21), can nucleate an
EYA-DAC interaction in S2 cells. Alternatively, EYA may
indirectly associate with DAC in the context of an as-yet-
uncharacterized macromolecular complex that may vary ac-
cording to the particular transcriptional target being regulated.
The observation that EYA interacts with itself reveals yet
another potential mechanism for complex and reiterative in-
teractions within the RD gene network. This interaction ap-
pears to be mediated by the N-terminal half of EYA, a region
that we have found necessary for full transactivation potential.
Furthermore, we find that the ED2 potentiates EYA-EYA
interactions. Because the EYA-EYA interaction maps to re-
gions of EYA necessary for transactivation, we propose that
homotypic interactions may contribute to EYA function as a
transcription factor in vivo.
Determination of cell fate is dependent on both the pres-
ence of a particular complement of transcription factors and
the appropriate activation state of these factors. Here we
provide evidence that while RAS/MAPK activation is not nec-
essary for EYA transactivation potential, it can potentiate
EYA-mediated transactivation. This leads to an intriguing
mechanism for modulation of the EYA-SO transcription fac-
tor, whereby in the absence of RAS/MAPK signaling, it is
competent to activate some transcription, but in the presence
of signal, this function is potentiated such that target genes
may be activated to higher levels. An alternative but not mu-
tually exclusive role for RAS/MAPK activation of EYA may be
to allow the higher activation potential of EYA to overcome
negative regulation of specific target genes.
Activation of EYA by RAS/MAPK signaling provides a
direct point of cross talk between a signal transduction mod-
ule and the RD gene network. Our results suggest that RAS/
MAPK signaling may regulate multiple aspects of RD network
function. In the context of Gal4DBD-EYA fusions acting on
UAS-luciferase, RAS signaling clearly increases transactivation
activity. Surprisingly, we found that mutation of the MAPK
phosphoacceptor sites in this fusion protein to either alanine,
to prevent phosphorylation, or a negatively charged residue, to
mimic phosphorylation, results in higher transactivation levels.
We believe that these results may be due the nature of this
assay, which uses chimeric and truncated Gal4DBD-EYA
?ED fusions, since we did not observe the same effect in the
context of full-length EYA working with SO to promote tran-
scription. Rather, in the context of an EYA-SO complex acting
VOL. 23, 2003STRUCTURE-FUNCTION ANALYSIS OF Eyes absent 5997
on ARE-luciferase, RASV12does not affect transcription, yet
there is a consistently strong increase in activation when the
phosphomimetic EYAS-D/Eis used. One possible explanation
for the lack of RAS responsiveness is that RAS signaling could
simultaneously upregulate EYA but downregulate SO. This
would provide a mechanism for fine-tuned transcriptional reg-
ulation, whereby RAS signaling can activate the EYA-SO tran-
scription factor through phosphorylation of EYA but nega-
tively regulates SO to prevent sustained high levels of
activation. Such dual and conflicting inputs by the RAS/MAPK
pathway are consistent with previous genetic observations.
Specifically, our previous work has demonstrated a positive
role for the pathway with respect to RD network function using
EYA as a point of cross talk, whereas work by others has
implicated the RAS pathway as antagonizing RD gene func-
tion, although in this case, the molecular mechanisms under-
lying the inhibitory regulation are unknown (28). Thus, it re-
mains to be seen how RAS/MAPK signaling regulates SO or
other members of the RD gene network or whether this direct
interaction is unique to EYA.
In addition to RAS/MAPK signaling, Notch, Hedgehog,
Wingless, and TGF?/DPP signaling all play important roles in
eye development (42). The integration of multiple signaling
pathway inputs with our existing knowledge of RD gene net-
work transcriptional regulatory loops suggests a mechanism for
unique specification of multiple cell fates in the eye. Our work
provides evidence that EYA is a crucial modulator of its own
activity, through autoinhibition and homotypic interactions.
We also show evidence that the RAS/MAPK pathway directly
enhances the transactivation potential of EYA and that the
corepressor GROUCHO inhibits EYA-SO-mediated tran-
scription. It will be important to discover whether other sig-
naling pathways interact directly with RD gene network pro-
teins and how such inputs effect the expression of distinct
cadres of target genes, thereby establishing and/or reinforcing
unique cell fates.
We thank S. Artavanis-Tsakonas for GRO monoclonal antibody, Z.
Paroush for gro cDNA, G. Mardon for so and dac plasmids, and R.
Fehon and T. Orr-Weaver for helpful comments on the manuscript.
We thank all members of the Rebay laboratory for their valuable
advice on this project and the manuscript.
S.J.S. is a Howard Hughes Medical Institute Predoctoral Fellow.
I.R. is a recipient of a Burroughs Wellcome Fund Career Award in the
Biomedical Sciences and is a Rita Allen Foundation Scholar. This
work was supported in part by National Institutes of Health grant RO1
1. Abdelhak, S., V. Kalatzis, R. Heilig, S. Compain, D. Samson, C. Vincent, D.
Weil, C. Cruaud, I. Sahly, M. Leibovici, M. Bitner-Glindzicz, M. Francis, D.
Lacombe, J. Vigneron, R. Charachon, K. Boven, P. Bedbeder, N. Van Rege-
morter, J. Weissenbach, and C. Petit. 1997. A human homologue of the
Drosophila eyes absent gene underlies branchio-oto-renal (BOR) syndrome
and identifies a novel gene family. Nat. Genet. 15:157–164.
2. Bai, J., and D. Montell. 2002. Eyes Absent, a key repressor of polar cell fate
during Drosophila oogenesis. Development 129:5377–5388.
3. Baonza, A., and M. Freeman. 2002. Control of Drosophila eye specification
by Wingless signalling. Development 129:5313–5322.
4. Bessa, J., B. Gebelein, F. Pichaud, F. Casares, and R. S. Mann. 2002.
Combinatorial control of Drosophila eye development by Eyeless, Homotho-
rax, and Teashirt. Genes Dev. 16:2415–2427.
5. Bonini, N. M., Q. T. Bui, G. L. Gray-Board, and J. M. Warrick. 1997. The
Drosophila eyes absent gene directs ectopic eye formation in a pathway
conserved between flies and vertebrates. Development 124:4819–4826.
6. Bonini, N. M., W. M. Leiserson, and S. Benzer. 1998. Multiple roles of the
eyes absent gene in Drosophila. Dev. Biol. 196:42–57.
7. Brand, A., and N. Perrimon. 1993. Targeted gene expression as a means of
altering cell fates and generating dominant phenotypes. Development 118:
8. Bui, Q. T., J. E. Zimmerman, H. Liu, and N. M. Bonini. 2000. Molecular
analysis of Drosophila eyes absent mutants reveals features of the conserved
Eya domain. Genetics 155:709–720.
9. Chen, R., M. Amoui, Z. Zhang, and G. Mardon. 1997. Dachshund and eyes
absent proteins form a complex and function synergistically to induce ectopic
eye development in Drosophila. Cell 91:893–903.
10. Chen, R., G. Halder, Z. Zhang, and G. Mardon. 1999. Signaling by the
TGF-beta homolog decapentaplegic functions reiteratively within the net-
work of genes controlling retinal cell fate determination in Drosophila. De-
11. Cheyette, B. N., P. J. Green, K. Martin, H. Garren, V. Hartenstein, and S. L.
Zipursky. 1994. The Drosophila sine oculis locus encodes a homeodomain-
containing protein required for the development of the entire visual system.
12. Curtiss, J., and M. Mlodzik. 2000. Morphogenetic furrow initiation and
progression during eye development in Drosophila: the roles of decapen-
taplegic, hedgehog and eyes absent. Development 127:1325–1336.
13. Czerny, T., G. Halder, U. Kloter, A. Souabni, W. J. Gehring, and M. Bus-
slinger. 1999. twin of eyeless, a second Pax-6 gene of Drosophila, acts up-
stream of eyeless in the control of eye development. Mol. Cell 3:297–307.
14. Fields, S., and O. Song. 1989. A novel genetic system to detect protein-
protein interactions. Nature 340:245–246.
15. Flores, G. V., H. Duan, H. Yan, R. Nagaraj, W. Fu, Y. Zou, M. Noll, and U.
Banerjee. 2000. Combinatorial signaling in the specification of unique cell
fates. Cell 103:75–85.
16. Galindo, M. I., S. A. Bishop, S. Greig, and J. P. Couso. 2002. Leg patterning
driven by proximal-distal interactions and EGFR signaling. Science 297:256–
17. Halder, G., P. Callaerts, S. Flister, U. Walldorf, U. Kloter, and W. J. Gehr-
ing. 1998. Eyeless initiates the expression of both sine oculis and eyes absent
during Drosophila compound eye development. Development 125:2181–
18. Halfon, M. S., A. Carmena, S. Gisselbrecht, C. M. Sackerson, F. Jimenez,
M. K. Baylies, and A. M. Michelson. 2000. Ras pathway specificity is deter-
mined by the integration of multiple signal-activated and tissue-restricted
transcription factors. Cell 103:63–74.
19. Heanue, T. A., R. Reshef, R. J. Davis, G. Mardon, G. Oliver, S. Tomarev,
A. B. Lassar, and C. J. Tabin. 1999. Synergistic regulation of vertebrate
muscle development by Dach2, Eya2, and Six1, homologs of genes required
for Drosophila eye formation. Genes Dev. 13:3231–3243.
20. Hsiao, F. C., A. Williams, E. L. Davies, and I. Rebay. 2001. Eyes absent
mediates cross-talk between retinal determination genes and the receptor
tyrosine kinase signaling pathway. Dev. Cell 1:51–61.
21. Ikeda, K., Y. Watanabe, H. Ohto, and K. Kawakami. 2002. Molecular inter-
action and synergistic activation of a promoter by Six, Eya, and Dach pro-
teins mediated through CREB binding protein. Mol. Cell. Biol. 22:6759–
22. Kawakami, K., H. Ohto, K. Ikeda, and R. G. Roeder. 1996. Structure, func-
tion and expression of a murine homeobox protein AREC3, a homologue of
Drosophila sine oculis gene product, and implication in development. Nu-
cleic Acids Res. 24:303–310.
23. Kawakami, K., S. Sato, H. Ozaki, and K. Ikeda. 2000. Six family genes—
structure and function as transcription factors and their roles in develop-
ment. BioEssays 22:616–626.
24. Kim, S. S., R. Zhang, S. E. Braunstein, A. Joachimiak, A. Cvekl, and R. S.
Hegde. 2002. Structure of the retinal determination protein dachshund re-
veals a DNA binding motif. Structure (Cambridge) 10:787–795.
25. Knust, E., K. A. Bremer, H. Vassin, A. Ziemer, U. Tepass, and J. A. Campos-
Ortega. 1987. The enhancer of split locus and neurogenesis in Drosophila
melanogaster. Dev. Biol. 122:262–273.
26. Kobayashi, M., K. Nishikawa, T. Suzuki, and M. Yamamoto. 2001. The
homeobox protein Six3 interacts with the Groucho corepressor and acts as a
transcriptional repressor in eye and forebrain formation. Dev. Biol. 232:315–
27. Kronhamn, J., E. Frei, M. Daube, R. Jiao, Y. Shi, M. Noll, and A. Rasmuson-
Lestander. 2002. Headless flies produced by mutations in the paralogous
Pax6 genes eyeless and twin of eyeless. Development 129:1015–1026.
28. Kumar, P., and K. Moses. 2001. EGF receptor and Notch signaling act
upstream of Eyeless/Pax6 to control eye specification. Cell 104:687–697.
29. Kurusu, M., T. Nagao, U. Walldorf, S. Flister, W. J. Gehring, and K.
Furukubo-Tokunaga. 2000. Genetic control of development of the mush-
room bodies, the associative learning centers in the Drosophila brain, by the
Eyeless, Twin of eyeless, and Dachshund genes. Proc. Natl. Acad. Sci. USA
30. Li, X., V. Perissi, F. Liu, D. W. Rose, and M. G. Rosenfeld. 2002. Tissue-
specific regulation of retinal and pituitary precursor cell proliferation. Sci-
5998SILVER ET AL.MOL. CELL. BIOL.
31. Lopez-Rios, J., K. Tessmar, F. Loosli, J. Wittbrodt, and P. Bovolenta. 2003.
Six3 and Six6 activity is modulated by members of the groucho family.
32. Mardon, G., N. M. Solomon, and G. M. Rubin. 1994. dachshund encodes a
nuclear protein required for normal eye and leg development in Drosophila.
33. Ohto, H., S. Kamada, K. Tago, S.-I. Tominaga, H. Ozaki, S. Sato, and K.
Kawakami. 1999. Cooperation of Six and Eya in activation of their target
genes through nuclear translocation of Eya. Mol. Cell. Biol. 19:6815–6824.
34. O’Neill, E. M., I. Rebay, R. Tjian, and G. M. Rubin. 1994. The activities of
two Ets-related transcription factors required for Drosophila eye develop-
ment are modulated by the Ras/MAPK pathway. Cell 78:137–147.
35. Ozaki, H., Y. Watanabe, K. Ikeda, and K. Kawakami. 2002. Impaired inter-
actions between mouse Eyal harboring mutations found in patients with
branchio-oto-renal syndrome and Six, Dach, and G proteins. J. Hum. Genet.
36. Pascal, E., and R. Tjian. 1991. Different activation domains of Sp1 govern
formation of multimers and mediate transcriptional synergism. Genes Dev.
37. Pignoni, F., B. Hu, K. H. Zavitz, J. Xiao, P. A. Garrity, and S. L. Zipursky.
1997. The eye-specification proteins So and Eya form a complex and regulate
multiple steps in Drosophila eye development. Cell 91:881–891.
38. Pignoni, F., and S. L. Zipursky. 1997. Induction of Drosophila eye develop-
ment by decapentaplegic. Development 124:271–278.
39. Rebay, I., R. G. Fehon, and S. Artavanis-Tsakonas. 1993. Specific trunca-
tions of Drosophila Notch define dominant activated and dominant negative
forms of the receptor. Cell 74:319–329.
40. Seo, H. C., J. Curtiss, M. Mlodzik, and A. Fjose. 1999. Six class homeobox
genes in Drosophila belong to three distinct families and are involved in head
development. Mech. Dev. 83:127–139.
41. Shen, W., and G. Mardon. 1997. Ectopic eye development in Drosophila
induced by directed dachshund expression. Development 124:45–52.
42. Treisman, J. E., and U. Heberlein. 1998. Eye development in Drosophila:
formation of the eye field and control of differentiation. Curr. Top. Dev.
43. Xu, C., R. C. Kauffmann, J. Zhang, S. Kladny, and R. W. Carthew. 2000.
Overlapping activators and repressors delimit transcriptional response to
receptor tyrosine kinase signals in the Drosophila eye. Cell 103:87–97.
44. Xu, P. X., J. Cheng, J. A. Epstein, and R. L. Maas. 1997. Mouse Eya genes
are expressed during limb tendon development and encode a transcriptional
activation function. Proc. Natl. Acad. Sci. USA 94:11974–11979.
45. Xu, P. X., I. Woo, H. Her, D. R. Beier, and R. L. Maas. 1997. Mouse Eya
homologues of the Drosophila eyes absent gene require Pax6 for expression
in lens and nasal placode. Development 124:219–231.
46. Zhu, C. C., M. A. Dyer, M. Uchikawa, H. Kondoh, O. V. Lagutin, and G.
Oliver. 2002. Six3-mediated auto repression and eye development requires
its interaction with members of the Groucho-related family of co-repressors.
47. Zimmerman, J. E., Q. T. Bui, E. Steingrimsson, D. L. Nagle, W. Fu, A.
Genin, N. B. Spinner, N. G. Copeland, N. A. Jenkins, M. Bucan, and N. M.
Bonini. 1997. Cloning and characterization of two vertebrate homologs of
the Drosophila eyes absent gene. Genome Res. 7:128–141.
VOL. 23, 2003STRUCTURE-FUNCTION ANALYSIS OF Eyes absent 5999