The conditional medaka mutation eyeless uncouples patterning and morphogenesis of the eye

Vertebrate eye formation starts during gastrulation with the determination of an eye field in the anterior ectoderm. Subsequently well-coordinated inductive events lead to the formation of a single retinal primordium in the anterior neuroectoderm and the presumptive lens placodes in the abutting head ectoderm (for review see Grainger, 1996; Jean et al., 1998). At late gastrula/early neurula stages, this single retinal field is split into two bilateral symmetrical retinal primordia by cell movements under the influence of signals emanating from the ventral midline (Gritsman et al., 1999; Li et al., 1997; Varga et al., 1999). Morphogenesis during neurula stages results in the evagination of the optic vesicles from the lateral wall of the forebrain, a process that requires changes in cell proliferation and programmed cell death (Schmitt and Dowling, 1994). Concomitantly, lens specification occurs when the optic vesicle contacts the overlying ectoderm of the lens placode (reviewed in Grainger, 1996). The optic vesicles invaginate under the influence of the developing lens to form the optic cups, in which differentiation of the neuroretina and the pigmented retinal epithelium (PRE) starts. The optic cup remains attached to the ventral diencephalon by the optic stalk, which eventually serves as a leading track for the projection of retinal axons to target regions in the midbrain.

Insect eye development has been analysed by a combination of genetic and molecular techniques. Surprisingly, key players, acting high up in the genetic pathway controlling Drosophila eye development, were found to be evolutionary conserved (Halder et al., 1995; Hammond et al., 1998; Oliver et al., 1995; Quiring et al., 1994; Xu et al., 1997) and to exert a comparable function during vertebrate eye development (Hill et al., 1991; Loosli et al., 1999; Oliver et al., 1996; Ton et al., 1991). In homozygous Small eye (Sey) mouse and rat, mutations in the transcription factor Pax6 lead to the complete absence of the eyes, including lens, and the nose (Hill et al., 1991; Hogan et al., 1986; Matsuo et al., 1993). In these embryos, the optic vesicles initially evaginate, but degenerate during subsequent stages of development (Hogan et al., 1986). Recently, it was shown that overexpression of Pax6 mRNA in Xenopus embryos can lead to the formation of an ectopic eye (Chow et al., 1999). Six3, a homeobox gene, is expressed early in the anterior neuroectoderm and subsequently in the optic vesicles and prosencephalon (Bovolenta et al., 1998; Kobayashi et al., 1998; Loosli et al., 1998; Oliver et al., 1995; Seo et al., 1998). Overexpression of Six3 mRNA in medaka initiates the formation of ectopic retina tissue in the midbrain and anterior hindbrain involving a regulatory interaction of Six3 and Pax6 (Loosli et al., 1999). This indicates a role for Six3 and Pax6 in early steps of vertebrate retina formation.

Targeted inactivation of the transcription factor Rx in mouse results in the absence of eye structures and in severe defects of the anterior forebrain, demonstrating the necessity of this gene for eye development (Mathers et al., 1997). Mice lacking sonic hedgehog gene function exhibit cyclopia phenotypes, demonstrating an involvement of sonic hedgehog in the early proximodistal patterning of the retinal primordia (Chiang et al., 1996), as also suggested by overexpression experiments in fish embryos (Ekker et al., 1995; Macdonald et al., 1995).

Mutations affecting eye development have been isolated in mutagenesis screens in zebrafish (Heisenberg et al., 1996; Malicki et al., 1996; Schier et al., 1996). Early steps of eye development are affected in the zebrafish mutation masterblind (mbl), which results in forebrain patterning defects and reduced or missing eyes (Heisenberg et al., 1996; Masai et al., 1997).

Here we present the spontaneous, recessive medaka mutation eyeless (el). el mutant (el⁻) embryos do not develop morphologically visible eye structures and die at early larval stages. el is a temperature-sensitive mutation and its activity is required between the late gastrula and the early neurula stage prior to the evagination of the optic vesicle. Our analysis of the el mutation reveals that patterning of the eye primordia and optic vesicle morphogenesis are uncoupled by the el mutation.

The medaka mutant eyeless was originally identified by Hideo Tomita. The el mutation was crossed into and kept in two different genetic backgrounds of wild-type medaka (O. latipes), which are kept in closed stocks at the EMBL, as described (Köster et al., 1997). Embryos were staged according to Iwamatsu (1994). To identify el carrier fish, pairwise matings were set up and the embryos were kept at 28°C or 18°C, respectively, and screened for the el phenotype.

A 550 bp medaka Emx1 fragment was amplified by RT-PCR on stage19-21 total RNA, using degenerate PCR primers specific for Emx1 (up, 5′ACIATCGAGWSIYTIGTIGGIAARGA; low, 5′CYGTTYTGRAACCAAACYTTIACYTG).

A 520 bp Vsx2 fragment was amplified by RT-PCR stage 19-21 total RNA, using degenerate PCR primers specific for Vsx2 (up, 5′GGITTYGGIATHCARGARATHYTIGG; low, 5′CCAYTTRGC-TCKICKRTTYTGRAACC).

PCR conditions were 5 cycles at 94°C/1 minute, 48°C/2 minutes and 72°C/4 minutes, followed by 35 cycles with annealing at 53°C.

A 560 bp Vax1 fragment was amplified by RT-PCR on stage 19-21 total RNA, using degenerate PCR primers specific for Vax1 (up, 5′AGRATCCTIGTSAGRGAYGCIAARGG; low, 5′TIGTYCTIC-TGTAIGGCTCRAAIGC). PCR conditions were 5 cycles at 94°C/1 minute, 51°C/2 minutes and 72°C/4 minutes, followed by 35 cycles with annealing at 56°C.

A 300 bp Vsx1 fragment was amplified by RT-PCR from stage 36 total RNA using degenerate PCR primers specific for Vsx1 (up, 5′GARGARYTIGARAARGCYTTYMAYGA; low, 5′YTTYTTR-TGCATICCSARSARCCA). PCR conditions were 5 cycles at 94°C/1 minute, 52°C/2 minutes and 72°C/4 minutes, followed by 35 cycles with annealing at 57°C.

A 1.3 kb Vsx1 fragment was then isolated by PCR from a somitogenesis stage cDNA library (stage 23) using a Vsx1-specific primer (5′GCAAAGTGGCGAAAGCGTGAG) and a T7 primer (GTGAATTGTAATACGACTCACTATAG). Cycling conditions were 35 cycles at 94°C/1 minute, 62°C/1 minute and 72°C/4 minutes.

Resulting PCR products were cloned into the TopoTA vector (Invitrogen) and sequenced.

A 560 bp Rx2 fragment (Loosli et al., 1999) and a 520 bp Vsx2 fragment were used to screen a 23-stage λZAP cDNA library at high stringency. Positive clones were plaque purified, converted to Bluescript SK plasmids and sequenced.

EMBL database accession numbers: Vax1, AJ250401; Emx1, AJ250402; Vsx1, AJ250403; Vsx2, AJ250404; Rx2, AJ250405.

Whole-mount in situ hybridization was performed as described (Loosli et al., 1998). The entire cloned cDNAs of α-Crystallin, Emx1, Fgf8, Pax6, Rx2, Six3, Vax1, Vsx1 and Vsx2 were transcribed for RNA riboprobes. In double-labeling experiments, the fast red detection (red) was performed first, followed by the NBT/BCIP (blue) staining. Vibratome sections of whole-mount stained embryos were done as described (Loosli et al., 1998).

Embryos were fixed in 4% paraformaldehyde in PBS, 0.5% Tween-20 and stored in 100% methanol at −20°C. Embryos were rehydrated and incubated in 2 mg/ml DAPI/PBS, washed (PBS) and mounted in 80% glycerol, 20% PBS/N-propyl gallate. 6 μm longitudinal sections of the whole embryo were scanned using confocal microscopy and nuclei in the optic stalk region of comparable longitudinal planes were counted. Two different wild-type embryos were analyzed in four focal planes each (average number of cells: 52 in the optic stalk region; 161 in both optic vesicles). Three different el⁻ embryos were analyzed in 6 comparable focal planes each at the presumptive optic stalk region (average = 73 cells).

Embryos fixed in 4% PF1 and stored in 100% methanol were rehydrated stepwise in PBST/methanol. Apoptotic cells were visualized using the ‘in situ cell death detection kit’ (Roche Molecular Biochemicals) in combination with the chromogens NBT/BCIP according to the manufacturers protocol.

Cells transplantation was done as described (Ho and Kane, 1990; Hong et al., 1998) with the following modifications. Embryos were dechorionated enzymatically using proteinase K (10 mg/ml, 40 minutes at room temperature) followed by incubation in medaka hatching enzyme as described (Wakamatsu et al., 1993). Wild-type donor cells were taken at the blastula stage from a transgenic fish line stably expressing GFP under the control of the ubiquitous cytoskeletal actin promoter (T. H. and J. W., unpublished results). el⁻ donor cells were derived from crosses of el carriers stably expressing GFP. Alternatively cells were labeled by injection of Rhodamine dextran (Sigma, 5% in 1× Yamamoto) into one blastomere at the 1-cell stage. At late blastula stage, 30-50 labeled cells from the animal pole region of donor embryos were transplanted into the same position of host embryos using a dissecting microscope equipped with UV illumination (Leica). Transplanted embryos were raised at the restrictive temperature of 18°C and analyzed from late neurula stage (stage 18) onwards.

The medaka mutant eyeless (el) was identified as a spontaneous, recessive mutation that affects eye development. The expressivity of the el phenotype is temperature sensitive. When raised at the permissive temperature of 28°C, homozygous el⁻ embryos show a complete penetrance but variable expressivity of the phenotype at somitogenesis stages. The variability ranges from the complete absence of optic cups (Fig. 1, 48% of homozygous el⁻ embryos) to the formation of one laterally located optic cup (data not shown) or of optic cups smaller than in the wild type (Fig. 2B, 52% of the homozygous el⁻ embryos with one or two small optic cups). When raised at the restrictive temperature of 18°C, no morphologically visible optic cups form in homozygous el⁻ embryos. The results below refer to embryos raised under restrictive conditions.

The el phenotype becomes morphologically apparent at mid to late neurula stage (stage 17/18, Iwamatsu, 1994), when optic vesicle evagination from the lateral wall of the prosencephalon starts in wild-type embryos (Fig. 1A). This evagination is not observed in el⁻ embryos (Fig. 1B,D). In a few cases, minute lateral protrusions of the forebrain form transiently between late neurula and early somitogenesis stages (Figs 1B, 3H). The formation of other sensory organs, such as nose or ears, is not affected by the el mutation (Fig. 1F; data not shown). Homozygous el fish die at day 1 to 5 after hatching. Escapers (1.5%) can be raised to adulthood; males are sterile, but females are fully fertile and mate successfully. Adult el fish show no morphologically visible eye structures (Fig. 1H).

The offspring of el carrier crosses were shifted at different stages of development from the restrictive temperature (18°C) to the permissive temperature (28°C), or vice versa. Embryos were analysed at later stages (10-somite stage, stage 22/23) when, in wild-type embryos, optic cups have formed (Fig. 2D). A temperature shift from 18°C to 28°C prior to the late gastrula stage does not affect the variable expressivity of el phenotype (Fig. 2B). However, when shifted at the early neurula stage (stage 17) or later, the variable expressivity is lost and all el⁻ embryos exhibit a complete lack of morphologically visible optic cups (Fig. 2C). Conversely, a temperature shift from 28°C to 18°C prior to the late gastrula stage (stage 16) leads to a complete lack of optic cups in el⁻ embryos, while a similar shift at the early neurula stage (stage 17) or later does not affect the variable expressivity of the el phenotype (data not shown). Thus, the phase between the late gastrula (stage 16) and the early neurula (stage 17) stage is the time window during which el activity is required.

In el⁻ embryos, optic vesicles do not evaginate from the prosencephalon at the neurula stage. To analyze whether this is due to the lack of tissue to be evaginated, we performed DAPI staining and used confocal microscopy to determine the cell distribution in this region of the brain. Optic vesicle evagination in wild-type embryos results in reduced cell numbers at the site of evagination (Fig. 3A, arrow), while el⁻ embryos show a uniform distribution of cells along the lateral wall of the forebrain (Fig. 3A,B). Cell numbers in the total eye region of the prosencephalon at stage 18 of el⁻ embryos were 70% less than the wild-type cell number (dashed yellow line; Fig. 3A,B). In contrast, counting cells in the lateral forebrain walls only (Fig. 3A,B, area surrounded by red dashed line), showed that el⁻ embryos have 40% more cells than wild type, in good agreement with the more uniform distribution of cells in the lateral wall of the prosencephalon in the mutant (see above). This indicates that, in the mutant, cells instead of evaginating remain in the lateral wall of the prosencephalon.

To investigate whether the lower total cell number in the forebrain of el⁻ embryos is caused by apoptosis, we used TUNEL labeling. The offspring of el carrier crosses were analyzed at different stages of development. At early stages (late gastrula to early neurula) no obvious differences in the number of apoptotic cells were detected in wild-type and el⁻ embryos (data not shown). At stage 18 (mid neurula stage), however, 9 hours (18°C) after el activity is required, enhanced apoptosis is detected in the anterior prosencephalon of el⁻ embryos during a short time period (2 hours, 18°C; Fig. 3C,D).

A second apoptotic wave is observed in el⁻ embryos at the 12-somite stage (see below). Thus, the increase in cell numbers within the lateral walls of the prosencephalon of el⁻ embryos due to the lack of evagination is in part counteracted by ectopic apoptosis affecting this region.

The transcription factors Pax6 and Six3 are key players for early eye and forebrain development and are expressed in the presumptive eye and forebrain region during late gastrula stages (Chow et al., 1999; Hill et al., 1991; Hogan et al., 1986; Kobayashi et al., 1998; Loosli et al., 1999; Oliver et al., 1995). To examine if the expression of Pax6 and Six3 is affected in el⁻ embryos, we performed whole-mount in situ hybridization. The expression patterns of Pax6 and Six3 were indistinguishable among sister siblings of an el carrier cross at late gastrula to early neurula stages (data not shown). Also, at late neurula stages, the expression patterns of Pax6 and Six3 in the forebrain are indistinguishable between wild-type and el⁻ siblings (data not shown). Strikingly, the expression of Pax6 and Six3 in the presumptive lens placode of el⁻ embryos is unaffected (Fig. 3E-H).

Altered expression of Pax6, however, was observed at the 12-somite stage. The dorsal diencephalic expression domain, which resembles an inverted V in wild type, is compressed in el⁻ embryos (Fig. 4A,B; arrow). Ectopic Pax6 expression is detected in the ventral diencephalon at the site of the optic stalk in wild-type embryos (arrowheads in Fig. 4A,B). Since this may reflect the presence of non-evaginated retinal precursor cells in the lateral wall of the diencephalon, we analyzed the expression of Rx2, a gene that is exclusively expressed in the retina of wild-type embryos from early somitogenesis (stage 19) onwards (Loosli et al., 1999). In el⁻ embryos, Rx2-positive tissue is detected in two lateral domains located in the ventral forebrain (Fig. 4D) at the 6-somite stage. In some cases, the Rx2-positive domains exhibit a concave shape resembling an invaginating retina (arrowhead in Fig. 4F). This indicates the formation of retinal precursor cells in el⁻ embryos, which are located in two ventral domains contiguous with the lateral wall of the diencephalon (Fig. 4E,F), at the site where the wild-type optic stalk connects.

Thus, except for the lack of evagination, morphogenetic movements of the developing diencephalon are normal. In mutant embryos, retina precursor cells initially born dorsally move with the rotating diencephalon to a ventral position during subsequent development and eventually end up at the attachment site of the optic stalk in the diencephalon. It is noteworthy that the split of the initial retinal field into two primordia is not affected by the el mutation, as two distinct domains of Rx2-positive retinal tissue form in the ventral diencephalon.

To investigate whether early patterning that generates proximal and distal identities in the optic vesicle occurs in homozygous el⁻ embryos, we checked for the presence of optic stalk tissue.

In wild-type medaka embryos, the transcription factor Vax1 is strongly expressed in the optic stalk and proximal retina and at lower levels in the ventral diencephalon and the olfactory placodes at the 12-somite stage (Fig. 5A), similar to Vax1 expression in mouse and Xenopus (Hallonet et al., 1998). In el⁻ embryos, high levels of Vax1 expression are found at a position where, in the wild-type embryos, the optic stalk is located (stage 23; Fig. 5B). The region expressing Vax1 at lower levels is expanded anteriorly (Fig. 5A,B). Vax1 expression in the olfactory placode is not affected in el⁻ embryos (data not shown).

We also examined the expression of Fgf8, which is expressed in the optic stalk at somitogenesis stages (stage 27, Figs 5C, 7B) (Carl and Wittbrodt, 1999; Reifers et al., 1998). In el⁻ embryos, Fgf8 transcripts are detected in a structure reminiscent of the optic stalk at stage 27. As in the wild-type eye, where the Fgf8-expressing optic stalk connects to the retina (Figs 5C,7C), the Fgf8 expression domain in el⁻ embryos is located adjacent to the Rx2-positive retinal tissue in the wall of the ventral diencephalon (Figs 5D,7E). Thus, in el⁻ embryos, the optic primordia are patterned and receive proximal and distal identities, which are maintained even though morphogenesis of the optic vesicles does not occur.

Our data show that, in el⁻ embryos, eye structures derived from the neuroectoderm, namely retina and optic stalk, form as parts of the lateral wall of the diencephalon. In amphibian embryos, lens formation in the head ectoderm is induced by the anterior neuroectoderm (Saha et al., 1989). To test whether the anterior neuroectoderm and the remnants of retinal tissue in el⁻ embryos have the potential to induce lens formation and differentiation, we examined lens formation by morphological criteria, the expression of Pax6 and Six3 in the lens placode, and α-crystallin expression as a molecular marker for lens fiber differentiation (Loosli et al., 1998). At the 16-somite stage, lenses are detected in 80% of the el⁻ embryos (Fig. 5G,H). The size of the mutant lenses is variable and smaller than in wild type (Fig. 5G,H). Remarkably these lenses, which express Pax6 at this stage (data not shown), do not express α-crystallin, although, in wild-type embryos, α-crystallin expression is detected by the 10-somite stage (Fig. 2B). In el⁻ embryos, α-crystallin expression is detected only in a few cases and only from the 18-somite stage onwards (Fig. 5F). In summary, lens induction as shown by wild-type Pax6 and Six3 expression (Fig. 3F,H) occurs, but, lens differentiation is delayed and, in most cases, incomplete, indicating that for proper differentiation the presence of an optic cup is required.

In the developing wild-type retina, Rx2 is initially expressed in proliferating cells of the entire optic vesicle and is subsequently restricted to the neuroretina as opposed to pigmented retina epithelium (PRE) (Loosli et al., 1999). To investigate whether, in el⁻ embryos, the non-evaginated Rx2-positive retinal tissue has the potential to differentiate further, we examined later stages of eye development for the presence of PRE. In wild-type embryos, PRE is first detected at the 30-somite stage (stage 28). In el⁻ embryos, however, PRE is never detected at any stage of development (Figs 1F,H, 5F).

To investigate whether the Rx2-positive retinal tissue differentiates further into neuroretina cells, we performed whole-mount in situ hybridization with genes specific for the differentiating retina. In fish, two different homologues of murine Chx10, Vsx1 and Vsx2 (Alx), have been identified (Barabino et al., 1997; Levine et al., 1994, 1997; Passini et al., 1998). Similar to the situation in other fish species, medaka Vsx2 is expressed prior to the onset of retinal cell differentiation in the temporal part of the ventral retina, starting at the 12-somite stage (stage 23) (Fig. 5G). As differentiation proceeds, Vsx2 expression is confined to the temporally located marginal zone of the neuroretina. In el⁻ embryos, Vsx2 is not detectable in the Rx2-positive area in the ventral diencephalon (Fig. 5H). Medaka Vsx1 is a marker for differentiated neurons of the retina. In the wild-type embryo, Vsx1 is expressed in postmitotic cells in the inner nuclear layer of the neuroretina, where it is confined to the bipolar cells (Fig. 5I; Belecky-Adams et al., 1997). In el⁻ embryos, Vsx1 is not expressed in the domain defined by Rx2 expression in the lateral wall of the ventral diencephalon (Fig. 5J). Taken together, even though retinal precursor tissue forms in el⁻ embryos, this tissue remains undifferentiated and forms neither PRE nor neuroretina.

As indicated by the distorted Pax6 expression domain in the dorsal diencephalon, it is likely that the retinal tissue in the ventral forebrain of el⁻ embryos affects the morphogenetic movements of the diencephalon. To examine further the forebrain of el⁻ embryos, we analyzed the expression of Emx1 and Emx2 as markers for specific regions of the developing forebrain (Boncinelli et al., 1993; Morita et al., 1995; Pannese et al., 1998; Loosli et al., 1999). The pattern of Emx2 expression in the telencephalon and the ventral forebrain is unaffected in el⁻ embryos (data not shown). Emx1 is expressed dorsally in the telencephalon of wild-type and el⁻ embryos. This indicates that the el mutation does not affect the formation of the telencephalon (Fig. 6A,B). However, the shape and location of the Emx1 domain are altered in el⁻ embryos. While the wild-type domain bends ventrally at the anterior tip of the brain during somitogenesis stages (stage 22), this occurs only partially in el⁻ embryos (Fig. 6A,B), indicating a retained movement of the telencephalon. In contrast, more caudal structures such as the hypothalamus are not affected as demonstrated by the unaltered morphology and expression of Emx2 (compare Fig. 6A and B; data not shown).

Analysis of apoptosis by TUNEL labeling revealed an increased number of apoptotic cells in the ventral forebrain and the lateral head ectoderm of el⁻ embryos at the 12-somite stage (Fig. 6C,D, stage 23). Thus, in el⁻ embryos, a second wave of ectopic apoptosis occurs in tissues that at later stages are affected in differentiation namely the retina and the lens.

To investigate whether el activity is required for the evagination of the optic vesicles in a cell-autonomous manner, we performed cell transplantation experiments. About 30-50 cells were transplanted at the late blastula stage from wild-type donors to isochronic offspring of an el carrier cross. To allow unambiguous identification of transplanted cells, a transgenic line stably expressing GFP in all cells was used as wild-type donor (T. H. and J. W., unpublished results). After transplantation, host embryos were raised at the restrictive temperature (18°C). The embryos were analyzed for 6 days from late neurula stage onwards.

Wild-type cells transplanted into el⁻ embryos are able to rescue the el phenotype and lead to the evagination of an optic vesicle and eventually to the formation of an eye (Fig. 6E,F). In all cases where optic vesicle formation was observed (n=3), evagination was unilateral and the tip of the outbudding optic vesicle exclusively comprised labeled wild-type cells, whereas the proximal part of the evaginating vesicle contained unlabeled mutant cells (2-somite stage, Fig. 6E). The contralateral side, however, showed complete failure of evagination. While in control transplantations, donor and host cells intermingle (Fig. 6H), wild-type cells transplanted into el⁻ embryos do not mix with mutant cells. The transplanted wild-type cells appear to initiate lateral outbudding of mutant cells. (Fig. 6E). However, at later stages, el⁻ cells do not contribute to the forming optic cup but are confined to more proximal regions (Fig. 6F).

To analyze the developmental potential of el⁻ cells in a wild-type background, 30-50 el⁻ cells stably expressing GFP were transplanted into wild-type embryos at the late blastula stage. Donor and corresponding host embryos were raised together (18°C) and analyzed for 6 days from the late neurula stage onwards. In all cases (n=8) where transplanted el⁻ cells were found in the head region of wild-type hosts, they were excluded from the optic vesicles, but located in the forebrain and midbrain as well as the lens. Thus, the neighboring wild-type cells cannot rescue the mutant cells. Also, at later stages, no el⁻ cells were found in the optic cups (Fig. 6G). In summary, the cell transplantations show that el functions in a cell-autonomous manner.

In this study, we show that optic vesicle morphogenesis is defective in the temperature-sensitive medaka eyeless mutant. Retinal tissue forms in the forebrain but does not evaginate (Fig. 7C,D), leading to increased cell numbers in the prosencephalon. The extra forebrain tissue affects later steps of brain morphogenesis such that telencephalic structures are retained in a more dorsal location. Proximal and distal identities as characteristic for wild-type optic vesicles are established in the absence of optic vesicle evagination (Fig. 7). Thus, early patterning occurs in the absence of optic vesicle formation, indicating that eye patterning and early eye morphogenesis are uncoupled by the el mutation.

The early expression of both Pax6 and Six3 during mid and late gastrula stages are unaltered. Thus, the determination of a single retina anlage and the early steps of lens induction are unaffected in el⁻ embryos. Temperature-shift experiments indicate a sensitive phase during the late gastrula and early neurula stage. Thus el activity is required prior to optic vesicle evagination, which in wild-type embryos starts 9 hours later (18°C).

Transplantation of wild-type cells into an el mutant background and the reciprocal experiment indicate a cell-autonomous requirement for el activity. Wild-type cells evaginate to form optic vesicles in el mutant embryos. Evagination on the contralateral side, when devoid of wild-type cells, is not observed. Mutant cells are found in the proximal region of optic vesicles rescued by transplanted wild-type cells but are not present in the resulting optic cups. This is consistent with a cell-autonomous nature of the el mutation where mutant cells are passively moved by the neighboring evaginating wild-type cells during initial stages of evagination. The reciprocal experiment shows that el mutant cells do not contribute to the optic cup confirming the cell-autonomous nature of the el mutation. Thus, neighboring wild-type cells cannot rescue the phenotype of el mutant cells.

Our data show that retinal primordia in el⁻ embryos do not form at an ectopic position, but rather fail to evaginate during development. These retinal primordia, as visualized by Rx2 expression, are ventrolaterally located, adjacent to the position where in the wild-type embryo the optic stalk emerges from the ventral diencephalon (Fig. 7). Furthermore, optic-stalk-like tissue as defined by the expression of Fgf8 and Vax1 is directly abutting the Rx2 expressing tissue. Thus, in the absence of optic vesicle morphogenesis, retinal precursors with proximal and distal identities form within the lateral wall of the diencephalon. The presence of two retinal primordia and adjacent optic stalk like tissue consistently indicate that the split of the single eye anlage and subsequent patterning are unaffected. However, differentiation of the retinal primordia into pigmented retinal epithelium and neuroretina is lost.

At a stage when, in wild-type embryos, optic vesicles have already evaginated, a short phase of apoptosis is detectable in the prosencephalon of el⁻ embryos in the region where the optic vesicles fail to evaginate. Since apoptosis in this region in el⁻ embryos occurs later than optic vesicle evagination in wild-type embryos, ectopic cell death is the consequence rather than the cause for the lack of optic vesicle morphogenesis.

In el⁻ embryos, a second apoptotic wave occurs at the stage when, in wild-type embryos, differentiation of the neuroretina is initiated. It results in ectopic cell death in the lens placode, the anterior part of the diencephalon and in the Rx2-positive retina precursor tissue of the mutant embryos; in wild-type embryos, no increase in the apoptotic index is detectable in these regions. This second wave of apoptosis suggests that the retinal precursor cells are either unable to respond to differentiation signals at this stage or do not receive the differentiation signals. This may in part be due to the lack of inductive signals between the developing lens and retina causing the observed apoptosis in both tissues.

In el⁻ embryos, the lenses that form are reduced in size and often do not differentiate properly as indicated by a lack of α-crystrallin expression. The expression of Pax6 and Six3, key players in the early steps of lens formation (Loosli et al., 1999; Oliver et al., 1996; Quinn et al., 1996), is unaltered in el⁻ embryos indicating that lens determination occurs properly.

Mutations affecting early steps of eye development have been described in other vertebrate species. All of these mutants exhibit a phenotype that differs significantly from that of the el mutation in medaka.

The zebrafish mutation masterblind (mbl) affects eye development to a variable degree. Severely affected embryos lack eyes and show an expansion of the dorsal diencephalon at the expense of the telencephalon (Heisenberg et al., 1996; Masai et al., 1997). Head sensory placode development is also affected in mbl mutant embryos, reflecting the proposed role of mbl in early forebrain patterning prior to the induction of the respective placodal structures. On the contrary, in el⁻ embryos, formation of the telencephalon and sensory placodes is normal and early forebrain patterning, as revealed by the expression of Pax6, Six3, Emx1 and Emx2, respectively, is not affected.

Absence of eye structures, mispatterning and, most severely, loss of entire forebrain was observed upon targeted inactivation of the mouse gene Rx (Mathers et al., 1997). Based on this phenotype, it has been proposed that Rx is required for eye formation from its initial stages (Mathers et al., 1997). The el mutation on the contrary affects optic vesicle evagination only. Marker gene analysis does not indicate a loss of forebrain structures in el mutant embryos and the initial steps of eye development are unaffected.

Mutations of the murine and rat Pax6 gene, Small eye (Sey), lead to the absence of eyes and nose as well to severe mispatterning of the forebrain. Initially, however, optic vesicles do evaginate, but eventually degenerate later during development (Hill et al., 1991; Hogan et al., 1986; Matsuo et al., 1993). In el⁻ embryos, in contrast, optic vesicle evagination does not occur, Pax6 is expressed normally in the nasal and lens placode, and nose and lenses eventually form. Our data show that the el mutation interferes with optic vesicle morphogenesis rather than the determination of the retinal, lens and nose primordia. The gross morphology of the forebrain is only marginally affected in el⁻ embryos, while Pax6 mutant mice exhibit severe mispatterning of the forebrain.

In summary, analysis of the eyeless mutation showed that initial steps of retina formation and patterning as well as optic vesicle morphogenesis are two genetically separable processes. el function is required for proper evagination of the optic vesicle and for differentiation of retinal cells in the optic cup. Future experiments, using the temperature sensitivity of the medaka el mutation, will address the link between patterning and morphogenesis.

We would like to dedicate this work to the memory of Professor H. Tomita, who initially discovered and named the eyeless mutation. We thank H. Jäckle for his support, M. Carl and D. Arendt for continuous discussion and helpful suggestions, S. Cohen, P. Rørth and T. Bouwmeester for critical reading of the manuscript, former laboratory members R. Kühnlein and R. Köster for their contribution, A.-M. Michon for help with confocal analysis, and A. Krone and E. Grzebisz for technical assistance. We are indebted to P. Gruss for financial support (S. W.). This work was supported by a grant of the SFB271 and an EC Biotechnology Grant (J. W.).