Misexpression of the Eyes Absent Family Triggers the Apoptotic Program*

Genetic studies in Drosophila and mice have shown that eyes absent ( eya ) is an important and conserved transcriptional regulator of development. Along with eyeless / Pax6 , sine oculis , and dachshund, eya genes function as master regulators in eye development and can induce ectopic eye formation. Furthermore, the loss-of-function mutants of these genes in the fly causes partial or complete loss of the compound eye, and this is associated with inappropriate apoptosis. Conversely, ectopic eyeless expression in the context of eyes absent or sine oculis mutations results in apoptosis, suggesting that the proper ratio of these factors regulates apoptosis. Here we report that enforced expression of fly eya or of one of its mammalian homologs, Eya2 , triggers rapid apoptosis in interleukin-3-dependent 32D.3 murine myeloid cells, which express Eya family members but not Pax6. Eya-induced cell death overrides survival factors and has many features typical of apoptosis, including plasma and mitochondrial membrane changes and caspase activation. Eya-induced apoptosis is blocked by Bcl-2 overexpression but not by the broad-spectrum caspase inhibitor z-VAD.fmk, suggesting that mitochondria are a major target in Eya-induced apoptosis. These results support the concept that inappropriate changes in the steady state levels of Eya proteins may trigger programmed cell deaths during development. The development of the compound eye in the Drosophila eye-antennal imaginal disc

The development of the compound eye in the Drosophila eye-antennal imaginal disc involves a complex sequence of events that include cell proliferation, differentiation, migration, and death (1,2). During larval stages, a small set of progenitor cells proliferate to form the eye disc. This is followed by a wave of differentiation that occurs when a morphogenetic furrow sweeps across the eye field in a posterior to anterior fashion. In the wake of this furrow, cells differentiate into cell clusters that eventually form the ommatidia of the eye. By contrast, cells continue to proliferate ahead of the furrow, and as the furrow moves anteriorly, their cell cycle becomes synchronized (3).
Mutagenesis studies have established that key regulators in Drosophila eye development include the transcription factors eyeless, sine oculis, dachshund, and eyes absent (eya). 1 All are highly conserved throughout evolution and play essential roles in eye formation (1, 4 -7). For example, ectopic expression of eyeless or its murine homolog Pax6 is sufficient to promote ectopic eye development (8). Similarly, ectopic expression of eya along with either sine oculis or dachshund is capable of inducing ectopic eyes (9 -11). Furthermore, there appears to be a hierarchical regulation, such that eyeless regulates the expression of eya, sine oculis and dachsund (9,12,13). In addition, feedback loops exist such that once eyeless induces the expression of eya, sine, oculis, and dachsund, these factors are then capable of regulating the expression of one another, ensuring the proper implementation of the eye development program (12). Once the program is initiated, other downstream eye targets are activated including those involved in photoreceptor development, which involves the induction of dectapentaplegic and the repression of wingless (14). Finally, this network is even more complex in vertebrates, which have multiple orthologs of each fly gene. For example, four murine homologs of eya have been identified as Eya1, Eya2, Eya3, and Eya4 (15)(16)(17)(18). A common denominator of the eyeless, sine oculis, dachshund, and eya mutations is either partial or complete loss of the compound eye (1, 4 -7), and this is associated with the deaths of progenitor cells located anterior to the morphogenetic furrow (1, 4 -7). In addition, when eyeless is ectopically overexpressed in the wing disc of eya or sine oculis mutants, it promotes ectopic cell death rather than eye development (12). Similarly, dachshund overexpression in either the antennal, leg, or wing disc also leads to inappropriate cell death (13).
The precise role these developmental regulators perform in controlling programmed cell deaths is unresolved. It has been suggested that eya may inhibit cell death either directly or indirectly as a default pathway that occurs following inappropriate proliferation or differentiation (1). Immortal interleukin-3 (IL-3)-dependent 32D.3 murine myeloid cells provide an excellent system to evaluate apoptotic regulators, similar to primary hematopoietic progenitors, as they continuously require hemopoietins for their survival, and in vivo, this fail-safe mechanism strictly regulates hematopoietic cell numbers (19 -23). Expression analyses demonstrated that the Eya family genes Eya1-3 were expressed in these cells, but that they failed to express Pax6, suggesting that this network of factors may play a role in regulating hematopoietic cell survival. Surprisingly, the overexpression of murine Eya2 or human EYA2 induced rapid apoptosis that overrides the survival functions provided by serum and IL-3. Similarly, inducible expression of the fly eya gene also triggered rapid cell death, demonstrating that this apoptotic pathway is conserved from arthropods to vertebrates. Moreover, biochemical and genetic data indicate that the major target of the Eya-induced cell death pathway is mitochondria. A model for how this apoptotic pathway may function during normal development is proposed.
Cloning of Human EYA2-To isolate a human homolog of Drosophila eya, a partial human clone, HPBEJ79, from an EST hemangioma cDNA library was obtained from Human Genome Sciences. This EST contained an insert of 1445 nucleotides with most of the 5Ј-718 bases being 65% identical to Drosophila eya. To acquire a complete cDNA, 5Ј-RACE (Invitrogen) was employed using mRNA from the human neuroblastoma cell lines LA-N-5 and CHP-134 (both obtained from ATCC, Manassas, VA), which had abundant levels of EYA2 transcripts as determined by Northern blotting. 2 Three sequential 5Ј-RACE reactions were required to retrieve the entire EYA2 cDNA. Three partial cDNA clones, one from each set of 5Ј-RACE extensions, were ligated by standard cloning procedures to generate a complete EYA2 cDNA, which is predicted to encode a protein of 538 amino acid residues.
Cloning of Murine Eya2-Based on areas of high homology between human EYA2 and eya, EYA2 primers were used for the isolation of murine Eya2, sense primer 5Ј-GGGACATTTGCATCCAGATAC-3Ј and antisense primer 5Ј-AGCTTCCTCATCCAGTCCAC-3Ј. RT-PCR was performed as described above, and a 300-base pair fragment was isolated from a murine T-cell leukemia cell line . To extend the 3Ј-end of the murine cDNA, 3Ј-RACE was performed as described by manufacturer protocol (Invitrogen). To extend the 5Ј-end of the murine cDNA, 5Ј-RACE was performed (Invitrogen). The first strand cDNA synthesis was performed using a gene-specific antisense oligonucleotide (2.5 pmol), sample mRNA (500 ng), and SuperScript II reverse transcriptase. The mRNA template was then degraded with RNase H, and the purified cDNA was tailed with dCTP and terminal deoxynucleotidyltransferase (producing an oligo(dC) n -3Ј tail). The oligo(dC)-tailed cDNA was then amplified by PCR using an anchor primer to oligo(dC) tail and a gene-specific nested primer. In addition to 5Ј-RACE, we searched the EST data base at NCBI and identified an EST that was identical to the 5Ј-end of our cDNA. Based on the sequence of this EST and PCR, we isolated this EST and with 5Ј-RACE extended it until the Eya2 cDNA contained the most 5Ј-ATG and several upstream in-frame stop codons.
Establishment and Maintenance of Inducibly Expressed Eya 32D.3 Cells-A sequence encoding the FLAG epitope MDYKDDDDK (Sigma) was attached to the 5Ј-end of fly eya, murine Eya2, and human EYA2 cDNAs (the eya cDNA was kindly provided by Dr. Seymour Benzer, California Institute of Technology, Pasadena, CA). This cloning was accomplished with primers designed to include seven base pairs (CAAGGCA) 5Ј of the Eya2 ATG and the FLAG sequence in frame with Eya2. High fidelity PCR was then performed using Expand Taq polymerase (Roche Molecular Biochemicals). DNA sequencing was performed to ensure no mistakes were generated during PCR. The tagged Eya2s and fly eya were then cloned into the XhoI (blunted) site of the dexamethasone (dex)-inducible vector pMAM-Neo (CLONTECH, Palo Alto, CA).
32D.3 cells were maintained in RPMI 1640 medium, 10% fetal calf serum, 1% L-glutamine medium supplemented with 20 units of IL-3 as described previously (24). Parental 32D.3 cells were electroporated by washing the cells (90% viable) with RPMI 1640 medium (without additives), adding 20 g of the linearized FLAG-tagged pMAM-Neo-Eya expression plasmids into the electroporation chamber and mixing with 1 ml of cells (at 5 ϫ 10 6 cells/ml) (for review see Refs. 23 and 24). The chamber was then placed into a cell electroporator and electroporated with 1180 farads, 375 volts, and low resistance. Transfected cells were selected in medium containing gentamycin 418 and individual clones were isolated as described previously (23,24). Control clones containing the vector pMAM-Neo were isolated in parallel.
Prior to the start of all experiments, cells were set twice at 0.5 ϫ 10 6 cells/ml on consecutive days to ensure that they were in equivalent exponential phase of their growth. To induce expression of the Eya proteins, clones were treated with 1.5 M dex (Sigma).
Eya Localization by Confocal Imaging-Cytospins were made of Eya2/eya clones with or without dex (100 l of cells at 0.5 ϫ 10 6 cells/ml). The slides were then fixed in 10% buffered formalin (Fisher) for 10 -20 min at room temperature, and the cells were permeabilized with cold acetone for 2 min. The slides were then washed with 1ϫ PBS for 5 min and blocked in 10% bovine serum albumin/1ϫ PBS for 30 min at room temperature. Primary FLAG antibody (M2, Sigma) was diluted 1:1000 in 10% bovine serum albumin/1ϫ PBS and placed on the slides for 2 h at room temperature. The slides were then washed three times with 1ϫ PBS (5 min each). The slides were incubated for 2 h with fluoroscein-conjugated secondary antibody diluted 25-50-fold in 10% bovine serum albumin/1ϫ PBS. After three washes with 1ϫ PBS, the slides were mounted with 0.1% p-phenylenediamine in 1:9 PBS:glycerol before microscopic examination.
Confocal imaging was performed using a Leica DM-IRBE microscope together with Leica TCS-NT software and equipped with a Nikon ϫ 100, 1.4 numerical aperture objective, and an argon/krypton laser (ϭ 488/565) to excite fluoroscein isothiocyanate (FITC) and Texas Red fluorescence.
Apoptosis Assays-The viability of vector-only and Eya2/eya cells cultured in IL-3 growth medium was assessed following treatment with or without dex using a hemocytometer and trypan blue dye exclusion as an indicator of cell viability. Cell morphology was assessed following cytospins and staining with Wright-Geimsa.
Mitochondrial membrane potential was assessed by loading dextreated and untreated cells with either 3,3-dihexyloxacarbocyanine iodide (DiOC 6 , 250 nM) or tetramethylrhodamine ethyl ester (TMRM, 500 nM) (Molecular Probes, Eugene, OR) for 30 min. The cells were then washed twice in growth medium and then resuspended in growth medium containing 100 nM of the respective dye. The cells were incubated for 2 h, washed in PBS, and then analyzed for fluorescence changes by flow cytometry (for DiOC 6 ) or confocal microscopy (for TMRM) (25,26).
Early plasma membrane changes were studied by using an Annexin-V-FITC kit (Roche Molecular Biochemicals). Cells were washed in PBS, resuspended in binding buffer (10 mM HEPES-NaOH, pH. 7.4, 140 mM NaCl, 2.5 mM CaCl 2 ), and cell density was adjusted to 5.0 ϫ 10 5 /ml. To 190 l of cells, 10 l of Annexin-V-FITC was added and incubated for 10 min in the dark. The cells were then washed in binding buffer, 10 l of 20 g/ml propidium iodide (PI) was added, and the sample was analyzed by fluorescence activated cell sorter (FACS) analysis.
The analysis of genomic DNA fragmentation was carried out as described previously (23). 1.0 ϫ 10 6 cells were washed with PBS, pelleted, resuspended in 20 l of buffer (10 mM EDTA, pH 8.0, 50 mM Tris-HCl, pH 8.0, 0.5% sodium laurylsarcosine, 0.5 mg/ml proteinase K), and incubated for 1 h at 50°C. DNA loading buffer was added, the samples incubated for 2 min at 70°C, loaded into the dry wells of a 2% agarose gel (0.1 g/ml ethidium bromide), run at 20 volts overnight in Tris-based boric acid EDTA, and the gel was photographed. Electron micrographs were performed by the St. Jude Children's Research Hospital electron microscope facility using standard protocols. z-VAD.fmk (Sigma) was dissolved in Me 2 SO and used at a final concentration of 400 M, which effectively delays death of 32D.3 cells that are deprived of IL-3 (25). As a control, an equal volume of the vehicle Me 2 SO was added to cells.
Cell Cycle Analyses-Cells were collected by centrifugation and re-2 B. E. Fee et al., manuscript in preparation. suspended in 0.1% sodium citrate with 50 g/ml propidium iodide and analyzed as described previously (23,24).
Western Blot Analysis-Cells were harvested, and 75-100 g of proteins were loaded/lane and resolved by SDS-12% polyacrylamide gel electrophoresis. Proteins were transferred to nitrocellulose using a semi-dry transfer cell (20 volts for 1 h, Trans-Blot SD, Bio-Rad). After transfer, blots were incubated in a fast green FCF staining solution (0.1% fast green FCF (Sigma F7258), 20% methanol, 5% acetic acid) for 2 min and destained for 5 min (20% methanol, 5% acetic acid). The blots were then washed in TBS-Tween 20 three times for 10 min and blocked for 1 h (5% nonfat dry milk, TBS-Tween 20). The primary antibody was then diluted in blocking solution and placed on the blots for 1h. Three washes followed in TBS-Tween 20 for 10 min each. Horseradish peroxidase-conjugated secondary antibody was then added in blocking buffer for 30 min to each blot and then washed in TBS-Tween as stated above before luminescence detection (SuperSignal, Pierce). The primary antibodies used were as follows: caspase-9 (Calbiochem), caspase-3 (PharM-

RESULTS
The Eya Family Is Expressed in Hematopoietic Cells-By screening cDNA libraries from Human Genome Sciences, we identified a human homolog of Drosophila eya, EYA2, from a hemangioma library (clone HPBEJ79). Using this EYA2 cDNA, we probed multiple tissue blots of fetal murine tissue. These results revealed that in addition to its expression in the eye, Eya2 is also expressed in E15.5 fetal brain, lung, and liver (data not shown). The fetal liver is the principal site of definitive hematopoiesis at this stage of murine development (26). These results suggested that the collection of eye developmental regulators may also play a role in the development and/or survival of hematopoietic progenitors. Therefore, we assessed expression of the Eya family and Pax6 in IL-3-dependent 32D.3 myeloid cells, which share many properties of primary hematopoietic progenitors. In particular, 32D.3 cells continuously require IL-3 for their survival (23). Surprisingly three of the four murine Eya members (Eya1, Eya2, and Eya3) were expressed in exponentially growing 32D.3 cells, and their expression was not dependent upon IL-3 ( Fig. 1). By contrast, Pax6 expression was not detected in 32D.3 cells, although it was detected in RNA isolated from the developing murine eye (Fig. 1).
Eya2 and eya Trigger Apoptosis of Murine 32D.3 Myeloid Cells-To determine whether Eya proteins could influence the survival of 32D.3 cells, we performed 5Ј-RACE of the partial human EYA2 cDNA from a Human Genome Science clone (HPBEJ79) to obtain a full-length cDNA (GenBank TM accession number Y10261) and then cloned the full-length murine Eya2 from RNA prepared from an Eya2-expressing murine T-cell leukemia cell line (RL-12) by RT-PCR and 5Ј-and 3Ј-RACE (GenBank TM accession number BC003755). We then FLAG-tagged fly eya, murine Eya2, and human EYA2. The FLAG-tagged Eyas were then cloned into a pMAM-Neo vector, and stable clones were generated in 32D.3 cells following electroporation and growth in G418 medium and cloning by limiting dilution. The pMAM-Neo vector is dex-inducible, and after induction with dex, we detected proteins for eya, murine, and human Eya2 at 95, 64, and 70 kDa, respectively ( Fig. 2A). The predicted molecular masses for these proteins are 80 (eya), 58 (murine Eya2), and 59 kDa (human EYA2) (15,27). This finding suggests that all Eya proteins may be posttranslationally modified.
Eya proteins have been variably reported to be nuclear or cytosolic (1,27,28), and Eya localization has been suggested to be regulated by interactions with the sine oculis/Six homeobox family of proteins (15). Therefore, we assessed murine Eya2 localization by confocal microscopy following induction with dex. Eya2 was localized to both the cytoplasm and nucleus yet was excluded from the nucleolus (Fig. 2B). Similar results were also evident in clones that inducibly express fly eya and in cells that overexpress human EYA2 (data not shown). The nuclear localization of a significant fraction of Eya2 is consistent with its generally accepted function as a regulator of transcription (11,29).
To assess the role of fly eya and murine and human Eya2 in regulating 32D.3 cell survival, vector-only and Eya-expressing clones were treated with dex and then deprived of IL-3. As a control, cells were also cultured in medium containing both IL-3 and serum and treated with dex. Surprisingly, human and murine Eya2-expressing clones rapidly died following the addition of dex, even when these cells were cultured in replete IL-3 growth medium containing the full complement of survival factors (Fig. 3). Similarly 32D.3 cells engineered to express fly eya also rapidly died following the addition of dex, whereas (as expected) dex had no effect on the survival of parental 32D.3 cells or on vector-only expressing cells (Fig. 3). Thus, eya and Eya2 overexpression induces cell death, and this pro-apoptotic function is conserved between fly, mouse, and human proteins.
Eya loss-of-function mutants in the fly exhibit excessive proliferation, suggesting that eya may regulate the cell cycle (11). Thus, we also measured whether the induction of Eya expression led to changes in the cell cycle distribution in exponentially growing cultures or in cells deprived of IL-3, a condition in which 32D.3 cells arrest in G 1 (23). No obvious changes were evident in Eya-overexpressing clones (data not shown), suggesting that Eya-induced cell death was not a consequence of overt influence on cell cycle traverse.
Eya-induced Cell Death Is Apoptotic-To assess whether eyainduced death was apoptotic in nature, we analyzed both the DNA integrity and the ultrastructure of eya and Eya2-expressing cells before and after treatment with dex. A hallmark of apoptosis is the condensation of chromatin into "micronuclei" and the cleavage of genomic DNA into oligonucleosomal units, resulting in a "DNA ladder" (30). After dex treatment, there was obvious internucleosomal DNA cleavage occurring in Eya2 clones (Fig. 4A), and Wright-Geimsa staining of cytospins of dex-treated Eya-expressing clones revealed typical condensation of chromatin (Fig. 4B). DNA condensation and condensation of the cytosol were also apparent by electron microscopy (Fig. 4C).
Another hallmark of apoptosis is the flipping of phosphatidylserine residues from the inner to the outer leaflet of the plasma membrane (31). Annexin-V binds to exposed phosphatidylserine residues, and FACS analyses can identify early apoptotic cells as those that stain with Annexin-V but exclude propidium iodide (PI), because the cell membrane is intact. FACS analysis demonstrated that dex treatment of Eya2 expressing clones, but not vector-only clones, led to a substantial increase in Annexin-V-positive/PI-negative early apoptotic cells (Fig. 4D). Thus, we conclude that eya-induced death of 32D.3 cells displays many of the hallmarks of apoptosis.
Caspase-3 Is Activated during but Is Not Essential for Eya2and eya-induced Death-Oligonucleosomal DNA fragmentation is mediated by DNA fragmentation factor 40, which is activated following caspase-3-mediated cleavage of its inhibitor DNA fragmentation factor 45 (32)(33)(34). The activation of caspases involves a cascade that first cleaves and activates initiator caspases, such as caspase-9, which then cleave and activate downstream effector caspases including caspases-3, -6, and -7 (35). Once activated, caspase-3 cleaves key targets required for cell integrity including DNA fragmentation factor 45 and PARP. We examined eya clones for caspase activation following the addition of dex using antibodies specific for caspase-9, caspase-3, and PARP. Immunoblot analyses demonstrated cleavage of caspase-9, caspase-3, and PARP following either eya or Eya2 induction (Fig. 5A). Therefore, eya-induced apoptosis involves caspase activation.
To confirm that caspase activation was required for eyainduced apoptosis, we assessed whether the broad-spectrum caspase inhibitor, z-VAD.fmk (36), could block the death of Eya clones treated with dex. In parallel, we evaluated the effects of z-VAD.fmk on the death of parental 32D.3 cells deprived of IL-3. As expected (25), z-VAD.fmk effectively delayed 32D.3 cell death associated with IL-3 withdrawal (Fig. 5B). However, z-VAD.fmk failed to protect 32D.3 cells from Eya2-induced death and actually slightly accelerated their death (Fig. 5B), despite the fact that it abolished DNA cleavage in these cells (Fig. 5C). Thus, Eya-induced apoptosis involves both caspasedependent and caspase-independent pathways.
Eya-induced Apoptosis Targets Mitochondria-Mitochondrial changes are associated with many forms of apoptosis, including caspase-independent pathways (37). A common mitochondrial alteration are reductions in membrane potential (⌬) (38,39), therefore, we assessed whether Eya2 expression influenced mitochondrial function by altering its ⌬. Initially we used the fluorophore DiOC 6 , which is taken up by mitochondria and other organelles of viable cells, but after changes in ⌬ diffuses out of the mitochondria and cell (40). Eya2 clones cultured in medium with and without dex were loaded with DiOC 6 and analyzed 14 h later by flow cytometry (Fig. 6A). 32D.3 cells grown in medium with and without IL-3 were used as a negative and positive control, respectively, for alterations in ⌬ (Fig. 6A). After treatment with dex, 40% of Eya2 clones underwent rapid decreases in membrane potential, whereas with or without dex was isolated and resolved on a 12% polyacrylamide gels, and Western blot analysis was used to determine Eya2, EYA2, and eya expression. Because Eya2, EYA2, and eya are FLAG-tagged, the primary antibody is anti-FLAG. B, To determine Eya2 localization within 32D.3 cells, Eya2-expressing cells treated with dex to induce Eya2 expression were fixed for immunohistochemistry and analyzed by confocal microscopy using the anti-flag antibody (green stain). Propidium iodide (PI) (red stain) was used to stain DNA in the nucleus.

FIG. 3. Overexpression of Eya2 and eya in 32D.3 cells overrides the survival factors IL-3 and serum to trigger cell death.
Cultures of 32D.3 clones growing in IL-3 and serum and engineered to inducibly express Eya2, eya, or the vector alone (Neo, electroporated at the same time as Eya2, EYA2, and eya) were treated with dex for the indicated intervals. Cell viability was determined at the indicated intervals following the addition of dex by trypan blue dye exclusion (n ϭ 3). When deprived of IL-3 and treated with dex, Eya-expressing clones displayed a marked acceleration in their rates of death (data not shown).
only 5% of untreated cells exhibited overt changes in their membrane transition state. To address whether this reduction in membrane potential was specific to mitochondria, we used the fluorophore TMRM, a potentiometric fluorescent dye that incorporates into mitochondria of viable cells in a ⌬-dependent manner (41). Eya2 clones and control cells in growth medium were cultured with and without dex, loaded with TMRM, and then observed by confocal microscopy (Fig. 6B). Again, dex-treated Eya2 clones underwent a marked decrease in their ⌬ visualized by a decrease in the fluorescence of their mitochondria, whereas these changes were not evident in control cells or in Eya2 clones not treated with dex (Fig. 6B). Therefore,

FIG. 4. Eya2 and eya overexpression induces apoptosis in 32D.3 myeloid cells.
A, 1 ϫ 10 6 Eya2-expressing or Neo vector cells growing in IL-3 and serum were treated with the vehicle alone (Ϫ) or were treated with dex for 24 h (ϩ). After 24 h, genomic DNA was prepared and loaded on a 2% agarose gel. As a positive control for internucleosomal DNA cleavage, genomic DNA was isolated from 32D.3 cells overexpressing c-Myc and deprived of IL-3 (Myc.2-IL-3), which displays an accelerated course of apoptosis (23). B, chromatin condensation was examined by making cytospins of Eya2 and Neo clones stained with Wright-Geimsa 24 h after dex treatment. C, 32D.3 clones expressing either fly eya or the Neo vector were treated with dex for 24 h, isolated, and fixed for electron microscopy. The bar at the bottom of each picture denotes a size of two microns. Note the condensed micronuclei in cells expressing fly eya. Similar results were obtained in cells expressing Eya2 (data not shown). D, to examine phosphatidylserine flipping, a clone that inducibly overexpressed Eya2, 32D.3/ Eya2.5, was treated with or without dex for 10 h, and the cells washed with PBS and Annexin-V were added. FITC-conjugated anti-Annexin-V was then added to the cells along with PI, and the cells were analyzed by flow cytometry. Annexin-V ϩ and PI Ϫ are early apoptotic cells (bottom right rectangle) that were 8-fold higher in Eya2 cells treated with dex. Similar results were obtained in clones engineered to express fly eya, and dex treatment did not induce Annexin-V-positive cells in vectoronly control cells (data not shown).

FIG. 5. Caspase-3 is activated by but is not essential for Eya2and eya-induced death.
A, protein (100 g) was isolated from the indicated cells treated with or without dex (14 h) and run on a 12% polyacrylamide gel. Western blot analysis was performed with FLAG (for Eya2 and eya expression), caspase-9, caspase-3, PARP, and tubulin antibodies. Both the proenzyme form (Proform) and the cleaved product (Cleaved) are shown for caspase-9 and PARP, but only the cleaved form of caspase-3 is shown. Tubulin served as a loading control. B, the broad caspase inhibitor z-VAD.fmk fails to protect 32D.3 cells from Eyainduced apoptosis. An Eya2-expressing clone (Eya2.5) was treated with dex and the caspase inhibitor z-VAD.fmk (400 M) for up to 24 h. As a control, parental 32D.3 cells cultured in the presence or absence of IL-3 were incubated with z-VAD.fmk. As shown, z-VAD.fmk protected 32D.3 cells from growth factor withdrawal-induced apoptosis but was unable to inhibit Eya-induced death. Viability was determined as indicated in Fig. 3 (n ϭ 3). C, to confirm that the caspase inhibitor was still effective in Eya2-expressing cells, we assessed its ability to inhibit DNA cleavage. Eya2.5 cells were treated with dex for 20 h and with or without z-VAD.fmk. DNA was isolated and run on a 2% agarose gel. L, DNA size markers. As shown, z-VAD.fmk was able to inhibit DNA cleavage in Eya2-expressing cells. Nonetheless, these cells still underwent cell death (B).

Eya-induced apoptosis is associated with marked alterations in ⌬.
Bcl-2 family members are known to regulate apoptosis in part through their effects on mitochondrial function (42). Cytokines suppress apoptosis of myeloid progenitors at least in part through their ability to selectively regulate the expression of the anti-apoptotic protein Bcl-X L (43), and c-Myc-induced apoptosis in 32D.3 cells is associated with the selective repression of Bcl-2 expression (44). We therefore assessed whether Eya-induced apoptosis was associated with alterations in the levels of anti-apoptotic or pro-apoptotic Bcl-2 family members by immunoblot analyses. However, the induction of Eya by dex did not result in the reduction of any of the anti-apoptotic proteins studied and also failed to induce the pro-apoptotic members Bax, Bak, or Bad (Fig. 7).
Caspase-dependent and caspase-independent cell death is blocked by Bcl-2 overexpression (45), therefore, we tested whether Bcl-2 overexpression could antagonize eya-induced death. The clones of 32D.3 cells engineered to overexpress human Bcl-2, which have a marked survival advantage when deprived of IL-3 (44), were electroporated with the murine Eya2 expression construct, and pools overexpressing both Bcl-2 and Eya2 were identified by immunoblot analyses. These cells were then analyzed for their viability after the induction of Eya2 with dex. Bcl-2 overexpression had no effect on the induction of Eya2 protein by dex (data not shown) but effectively blocked Eya2-induced apoptosis (Fig. 8). Overall, these data suggest that Bcl-2 overexpression prevents Eya-induced apoptosis by protecting mitochondrial functions. DISCUSSION The role of the conserved class of transcription factors eyeless, eya, sine oculis, and dachsund in development is well documented (2,3,45). Loss-of-function mutants and ectopic overexpression studies in the fly have linked these proteins to the regulation of programmed cell death, yet their direct role as activators of the apoptotic program has heretofore been lacking. Here we have shown that at least when overexpressed eya can directly activate the apoptotic program in 32D.3 myeloid cells, and that this function is conserved between the fly, mouse, and human Eya proteins. 32D.3 cells express Eya1-3 and are a useful model to evaluate apoptotic pathways, because like normal hematopoietic progenitors, they continuously require IL-3 to inhibit the endogenous apoptotic program (23). The model originally proposed by Benzer and colleagues (1) suggested that eya may function as an anti-apoptotic protein, therefore, an expectation of our studies was that Eya overexpression would inhibit the death of 32D.3 cells when they were deprived of IL-3. Surprisingly, we found the opposite to be true, and that eya acted as a pro-apoptotic signal in this cell context. In part, this is perhaps not entirely unexpected as an imbalance of this regulatory network in the fly does lead to inappropriate apoptosis (1, 4 -7, 12, 13, 46). Overall, our data and those of others suggest that when one of the genes of this network is overexpressed, a default apoptotic program is triggered. This death program could simply be the result of inappropriate FIG. 6. Eya-induced apoptosis targets the mitochondria. A, to examine overall cell membrane integrity, DiOC 6 was loaded into 32D/ Eya2.5 cells treated with or without dex for 14 h and washed to determine membrane leakage. Cells undergoing membrane changes will leak the dye, and their fluorescence intensity will decrease. The values indicate the number of cells that have a decrease (a shift to the left) in their fluorescence intensity and thus decreases in their ⌬. As a control, 32D.3 cells were grown in medium containing IL-3 or were deprived of IL-3, which induces their apoptosis (23). B, to examine whether the changes in ⌬ were specifically associated with mitochondrial changes, we examined mitochondrial ⌬ changes with the dye TMRM, which is mitochondria-specific, in Eya2.5 cells treated (14 h) with or without dex. Cells were loaded with TMRM (500 nM), washed in growth medium, placed in growth medium with 100 nM TMRM, incubated for 2 h, washed in PBS, and prepared for confocal microscopy. Again, whether mitochondrial ⌬ decreased, the dye would not be retained by the mitochondria, and a decrease in fluorescence would occur. The hatched circle indicates the outline of the Eya2-expressing cells. FIG. 7. Eya2-and eya-induced apoptosis is not associated with significant changes in the levels of Bcl-2 family proteins. Protein was isolated from exponentially growing cultures of clones expressing Eya2 or eya treated with or without dex for 14 h. Western blot analysis was performed with antibodies to the following Bcl-2 family members, Bad, Bak, Bax, Bcl-2, Bcl-X L , and Mcl-1. Tubulin was used as a loading control, and 100 g of protein was loaded/lane. A slight decrease in protein-specific signal that was seen in most lanes treated with dex was found to be because of the release of histones from DNA following cleavage (data not shown). stoichiometry of these factors or could be attributed to a complete misregulation of eya target genes.
To define how eya provoked the cell death, we examined biochemical changes. Eya-induced apoptosis was associated with many of the hallmarks of typical apoptosis including condensation of the cytosol and nucleus, plasma cell and mitochondrial membrane changes, the activation of caspases, and the degradation of genomic DNA into a typical oligonucleosomal ladder. However, eya-induced death was not inhibited by the broad caspase inhibitor z-VAD.fmk. It has been suggested that other proteases such as calpain can be activated directly during apoptosis (45) or following the release of pro-apoptotic factors from mitochondria such as apoptosis-inducing factor (38), and this may explain the failure of z-VAD.fmk to rescue some forms of apoptosis (45). Consistent with what others (55) have observed in caspase-independent pathways, we did see extensive cell membrane blebbing in Eya-expressing cells in the presence of z-VAD.fmk (data not shown), even though internucleosomal cleavage of genomic DNA was blocked. Furthermore, eya-induced death was not associated with overt changes in the expression of any Bcl-2 family members. However, Eya-induced apoptosis was blocked by the overexpression of Bcl-2, which is consistent with the ability of Bcl-2 to block both caspase-dependent and independent death (42,45). Thus, eya appears to induce apoptosis by triggering both caspase-dependent and independent pathways. One unresolved issue is whether Eya somehow results in changes in the localization of Bcl-2 family proteins. For example, Bax translocates from the cytoplasm to the outer mitochondrial membrane following the receipt of apoptotic signals (42), and this could be occurring during eyainduced apoptosis. Furthermore, we suspect that since most intrinsic cell death signals appear to require the combined functions of Bax and Bak (47), Eya-induced death would have similar constraints.
It has been suggested that the cell deaths that occur in Drosophila eya mutants is not because of their inability to control death but rather their failure to properly control proliferation (11). This concept is supported by excessive proliferation in the eye disc in these mutants, and the fact that cell death is only observed following traverse of the morphogenetic furrow (11). By contrast, in 32D.3 cells eya and Eya2, overexpression triggers apoptosis without overt effects on cell growth and cell cycle traverse. Of course, the simplest interpretation is that Eya functions may be cell context-specific, but another interpretation is that the loss-of-function fly eya mutants display excessive proliferation as a consequence of inappropriate survival of progenitor cells because of the lack of Eya-induced apoptosis.
The Drosophila eye regulator eya and murine Eya2 are both capable of inducing apoptosis when ectopically expressed in eyeless mutants in vivo (9). Thus, eya and other eye regulators may perform dual roles in development by both promoting eye development and activating cell death. We propose that proper development versus cell death is dependent on the appropriate levels of the eye regulators. Specifically, when eya genes (or others such as sine oculis or dachshund) are overexpressed or when other members of the regulatory network are not present in the proper stoichiometry, a default pathway is activated that triggers apoptosis (Fig. 9). Again, this could be attributed to inappropriate regulation of the target genes of this network. Such a pathway would ensure that misguided development rarely occurs. Several lines of evidence support this concept. Firstly, when overexpressed in a wild-type background, eya, eyeless, sine oculis, and dachshund all result in a reduction of the eye (46). These deficits are phenotypically similar to lossof-function mutants in each of these genes and are again associated with massive cell deaths in the eye field (1,5,7,8). Secondly, when eyeless is ectopically expressed in the wing disc of either eyes absent or sine oculis mutant flies, an eye is not formed but rather abnormal structures caused by massive cell death (12). Similar observations have been reported when dachshund is overexpressed in either the leg or wing of wild type flies (13). Finally, we have shown that the overexpression of eya or Eya2 in 32D.3 cells, which do not express Pax6 (Fig.  1), induces apoptosis, and we have also shown that sine oculis overexpression also triggers the death of these cells. 3 Thus, the overexpression of these eye regulators seems in most cases sufficient to trigger apoptosis.
Recent studies have also implicated eyeless/Pax6 in trigger-3 W. C., and J. L. C., unpublished results.
FIG. 9. A proposed model for how overexpression of Eya family members leads to apoptosis rather than eye development. A, when the eye regulators are expressed at their proper levels, normal development occurs. In this scenario, target genes are correctly regulated. B, if the levels of one of the eye regulators are either deficient or in excess, a default apoptotic pathway is triggered, which may be because of simple changes in stoichiometry of these factors and/or alterations in the regulation of critical target genes. Arrows indicate feedback loops that exist within the network. eyeless (ey) is required for the expression of eyes absent (eya), dachshund (dac), and sine oculis (so), yet these factors can also regulate the expression of eyeless ( (1) and (2)) (for review, see Refs. 9 and 13) and each other.
ing vertebrate cell deaths. For example, during Xenopus development, Pax6 expression in the neural -fold stage strikingly overlaps with TUNEL-positive cells (48). Furthermore, highcopy-number Pax6 transgenic mice have a severe microphthalmia phenotype (49). Finally, it has been shown that PARP, the target of caspase-3, is itself a regulator of Pax6 gene expression in the neuroretina (50). Together with findings demonstrating that the eya genes require Pax6 for expression (15) suggests that a feedback loop exists in which Pax6, Eya2, and PARP regulate cell death in the developing neuroretina.
There is also ample evidence that reduction in expression of the eye master regulatory genes also triggers apoptosis. In addition to the wealth of data in the fly where mutations in eya, eyeless, sine oculis, or dachshund correlate with excessive apoptosis (1,5,7,8), deletion of Eya1 in mice leads to defective otic vesicle formation, and this is associated with excessive apoptosis (51). Furthermore, the knockout of the ski gene (a proposed dachshund family member) results in embryos with excessive apoptosis during neurulation (52,53). Lastly, an analysis of mouse mutants defective in Pax6 expression indicates abnormal apoptosis (54). Therefore, overexpression or a failure to express eye master regulatory genes appears sufficient to trigger the cell death program. This finding suggests that the relative amounts of these proteins must be tightly controlled for proper development. We propose that the apoptotic pathway is activated only when the stoichiometry of these regulators is distorted because of changes in tissue-specific expression or after mutations that inactivate or silence the expression of these genes or alter their function or regulation of their target genes.