Mutations affecting development of the zebrafish retina

Neurons of the vertebrate central nervous system, including the retina, originate from an initially uniform sheet of pseudostratified neuroepithelium (Jacobson, 1991). In the neural retina seven major cell types, including six types of neurons and the Muller glia, combine in three cellular layers to form a powerful and versatile sensory structure (Dowling, 1987). Several problems of general significance are associated with the development of the vertebrate retina. How is the unique identity of the retina specified within the neuroepithelial sheet of the neural tube? How are the identities of the six retinal neurons and one glial cell type acquired during the process of development? What cues guide individual cell types to occupy specific positions within the stratified structure of the retina? These questions are not unique to the retina and are equally relevant to many regions of the vertebrate brain.

For several reasons the zebrafish retina is particularly suited for studies of neuronal specification and patterning in the vertebrate central nervous system. In teleost embryos the eye is relatively large and easily accessible. The cellular architecture of the retina is relatively simple and well characterized (Dowling, 1987). Subtypes of retinal neurons are easily recognizable by their position and morphology. Molecular markers specific to many retinal neurons are available (Larison and Bremiller, 1990; Trevarrow et al., 1990; Raymond et al.,1993; Sandell et al., 1994). Importantly, in the central region of the zebrafish retina, the vast majority of neurons are born and organized in distinct laminae by 60 hours postfertilization (hpf). Taken together, these characteristics make the development of the zebrafish retina particularly amenable to genetic analysis.

Early development of the zebrafish eye has been previously described in detail (Schmitt and Dowling, 1994). The optic lobe can be first distinguished at the 6 somite stage. The pigmented epithelium and the neural retina become distinct starting at the 11-12 somite stage. Invagination of the optic lobe is initiated at the 14 somite stage and is accompanied by the formation of the lens rudiment. By the end of somitogenesis (30 somites; 24 hpf) the lens is spherical and has detached from the epidermis. The optic cup consists of two distinct layers: a thick layer of columnar pseudostratified neuroepithelium from which the neural retina will form, and a thin layer of flat pigmented epithelial cells. The first pigment granules appear in the pigmented epithelium at approximately 24 hpf.

As in other vertebrates, ganglion cells are the first neurons to be born in the zebrafish retina (Nawrocki, 1985; reviewed by Altshuler et al., 1991). Birth-dating studies indicate that the first postmitotic neurons appear between 29 and 34 hpf (Nawrocki, 1985). At 36 hpf none of the neuronal cell layers are clearly distinguishable (Fig. 1A). By 60 hpf the vast majority of neurons in the central retina have already been born (Nawrocki, 1985) and are organized into three nuclear layers separated by two plexiform layers (Fig. 1B). The stratification of the retina becomes progressively more distinct at later stages of development (Fig. 1C,D).

The photoreceptor cell layer of the zebrafish retina contains five photoreceptor types: rods, short single cones, long single cones, and long and short members of the double cone pair (Branchek and Bremiller, 1984). Initially, different photoreceptor types are not distinguishable by morphological criteria. Short single cones can be first distinguished from other photoreceptor cells at about 4 days postfertilization (dpf) (Fig. 1D). By 12 dpf all photoreceptor types can be distinguished on the basis of morphological criteria (Branchek and Bremiller, 1984).

Genetic analysis has been exceptionally successful in dissecting the development of invertebrate sensory organs. In the Drosophila eye, genes involved in the specification of the eye imaginal discs (Halder et al., 1995), progression of the morphogenetic furrow (Haberlein et al., 1993; Ma et al., 1993; Brown et al., 1995) and specification of photoreceptor cell fates (Tomlinson et al., 1988; Baker et al., 1990; reviewed in Zipursky and Rubin, 1994) have been identified. It is not clear however, to what extent the genetic circuitry utilized in Drosophila is conserved in the development of the vertebrate eye. Several genes known to play important roles in Drosophila eye development have vertebrate homologs with expression patterns in developing eyes (Della et al., 1993; Fjose et al., 1993; Quiring et al., 1994). The function of most of these genes has not yet been determined. Thus far, the strongest point of similarity between insects and vertebrates is specification of the eye primordium. The vertebrate gene Pax-6 (Small eye) and its fruit fly homologue eyeless have similar loss of function phenotypes causing reduction or absence of eyes in mice and fruit flies, respectively (Hill et al., 1991; Quiring et al., 1994). Amazingly, both eyeless and Pax-6 are sufficient to induce ectopic specification of eyes in the developing fly (Halder et al., 1995).

In a large scale mutagenesis screen in zebrafish (Driever et al., 1996), we attempted to identify recessive, zygotic loci involved in several aspects of vertebrate eye development. Our search led to isolation of 49 mutations falling into the following six phenotypic categories: neuronal patterning defect; cyclopia; defect of the outer retina; growth retardation; nonspecific retinal degeneration; and retinal degeneration associated with a pigmentation defect. In this work we report the initial genetic and phenotypic characterization of the isolated mutants.

Mutations were induced using N-ethyl-N-nitrosourea (ENU) as described previously (Solnica-Krezel et al., 1994). All mutants were initially identified in the progeny of crosses between individuals of the F2 generation bred from mutagenized G₀ males (Driever et al., 1996). All crosses which led to the identification of mutant phenotypes were repeated. For further analysis, heterozygous carriers of mutant alleles were outcrossed to wild-type AB (Chakrabarti et al., 1983), Tubingen (Mullins et al., 1994) or India (Knapik et al., 1996) strains. Mutant heterozygotes were identified in the progeny of outcrosses by sibling mating.

Complementation tests were performed between heterozygous carriers of mutations falling into the same phenotypic category. Both mutants in the ‘miscellaneous’ category were tested with the neuronal patterning group. At least 30 embryos, often from two independent crosses, were scored for the mutant phenotype. For mutations with no assigned locus name (Table 1), complementation testing was not performed.

Embryos were maintained at 28°C in egg water (Westerfield, 1994) with addition of methylene blue (Sigma, Inc.) at the concentration of 1 mg/l. Staging was performed as described previously (Kimmel et al., 1995). For observations, embryos were anesthetized with a 0.02% solution of 3-aminobenzoic acid methyl ester (Sigma, Inc.) and embedded in 3% methylcellulose in egg water. Living embryos were initially observed and photographed under a dissecting microscope. Detailed observations were conducted using an Axiophot microscope and Nomarski optics (Zeiss, Inc.).

For histological analysis, embryos were fixed in PFA/GA/Acrolein fix (Kuwada et al., 1990), embedded in JB-4 resin (Polysciences, Inc.) and sectioned at 3 μm. For each mutant strain 3-6 embryos were sectioned in a single block. Sections were collected on slides, stained with methylene blue-azure II (Humphrey and Pittman, 1974) for 10 seconds, rinsed in distilled water for 10 minutes, dried and mounted with Permount (Fisher Scientific, Inc.). Depending upon the onset of phenotype, mutant retinae were sectioned at either 3 or 5 dpf. In most cases a wild-type sibling of the mutant individuals was embedded for sectioning in the same block.

As part of a large scale mutagenesis screen we searched for zygotic recessive mutations affecting development of the zebrafish eye (Driever et al., 1996). We paid particular attention to several aspects of eye development: morphogenesis of the optic cup; appearance of the pigmented epithelium; the size and shape of the eye. Relevant aspects of development were evaluated at 1, 2, 3 and 5 dpf.

Not all mutations affecting development of the eye were subjected to a detailed analysis. For example, at 5 dpf, a small eye phenotype was frequently associated with an overall reduction of brain size and a delay in the development of branchial arches and pectoral fins. Similarly, at earlier stages, mutant embryos frequently displayed a severe reduction of eye and brain size. These phenotypes appeared to reflect a general delay of development or non-region-specific neuronal degeneration. Mutations leading to such phenotypes were not analyzed further.

For all mutations, retinae were sectioned to assess the fate of individual neural cell types. Based on observation of living embryos and analysis of histological sections we grouped the isolated mutations into six phenotypic categories: neuronal patterning defect; cyclopia; defect of the outer retina; growth retardation; nonspecific retinal degeneration; retinal degeneration associated with a pigmentation defect (Table 1). Complementation tests were performed for mutations classified in the same groups. Below we discuss the phenotypic characteristics of mutants in each of the identified categories.

In wild-type zebrafish pigmentation appears at 24 hpf both in the pigmented epithelium and melanophores. The mutants oko meduzy (ome)^m98, glass onion (glo)^m117 and nagie oko (nok)^m227exhibit an eye specific pigmentation defect (Table 1). The eye phenotype of these mutations becomes visible shortly after the onset of pigmentation. Wild-type eye pigmentation is uniform (Fig. 2A,F). In contrast, eye pigmentation of the mutant animals is patchy or absent (Fig. 2B,C,E,G). Fig. 2 shows the mutant phenotypes in this category at 36 hpf. Melanophores have a normal appearance in the mutant embryos.

We inspected retinae of the mutant embryos at 3 dpf on histological sections. At this stage, wild-type retinal neurons form an easily recognizable pattern of nuclear and plexiform layers (Fig. 1C). In all three mutants the laminar arrangement of retinal neurons is disorganized (Fig. 3A-C). Instead of forming layers, the plexiform matter is distributed in patches. Although photoreceptor cells have a distinct, elongated shape, rudiments of the photoreceptor cell layer were never observed in any mutant retina. We evaluated whether the amount of cell death is abnormal in ome^m98 embryos. Cell corpses are easily recognizable on histological sections by their condensed, round appearance. In the ome^m98 retina at 48 hpf the amount of cell death does not exceed the wild-type level. By 3 dpf the number of cell corpses in the mutant retinae appears to be approximately 4 times higher than in the wild type (data not shown). These observations suggest that the patterning defect in ome^m98 is not a consequence of an extensive degeneration of retinal neurons. All mutations in this group produce fully penetrant phenotypes in the retina.

Mutations in this category also lead to an abnormal brain shape (Schier et al., 1996), curved body axis and frequently reduced or absent circulation. The phenotypes caused by ome^m98 and nok^m227 are very similar to each other. The glo^m117 phenotype is more pleiotropic, causing a much stronger brain defect and abnormal tail development (Fig. 2B). In ome^m98 and nok^m227, overall brain patterning appears to be normal on histological sections, whereas the brain pattern of glo^m117 is disorganized (data not shown).

Mutations in four loci lead to a single-eye phenotype (cyclopia) (Table 1). The cyclops (cyc) locus (Hatta et al., 1991) has been previously shown to be involved in patterning of the ventral brain (Hatta, 1992). The three other loci producing cyclopic phenotypes are bozozok (boz)^m168, one eyed pinhead (oep)^m134 and uncle freddy (unf)^m768. The degree of eye fusion is variable in these mutants. In the most severe cases of the boz^m168 phenotype, the cyclopic eye is very small or even absent. In less affected individuals two eyes form closely next to each other (Fig. 3G). In order to assess whether fusion of the eye anlage affects patterning of retinal neurons, we prepared histological sections through the eyes of cyc^b16 (Hatta et al., 1991), oep^m134 and boz^m168 mutants at 4 dpf. Although retinal development is somewhat delayed in the cyclopic mutants, neuronal lamination forms at least to some degree (Fig. 3E-G). Patterning abnormalities are present in some mutant individuals and are particularly evident in the mutant boz^m168. On histological sections, abnormal patterning of the retina is accompanied by extensive cell death.

Mutations at seven loci predominantly affect development of the outer retinal layers (Table 1). With the exception of elipsa (eli)^m649, mutants in this category exhibit reduced eye size at 5 dpf in comparison to their wild-type siblings (Fig. 4). In the lateral view the eye diameter of the mutant larvae is reduced by 20% to 30% at this stage. The eye of eli^m649 has an abnormal, oval shape at this stage (Fig. 4D). Histological sections through the mutant retinae reveal three types of photoreceptor deficit in this group.

The first type is present in mikre oko (mok)^m632 and niezerka (nie)^m743. In these two mutants the photoreceptor cell layer is absent at 5 dpf while the ganglion and the inner plexiform layers appear to be relatively normal (Fig. 5B,G). Some deficits may also be present in the inner nuclear layer. At 3 dpf some photoreceptor cells can be found in the central retina in both mutants. Fig. 5A shows the appearance of the mok^m632 retina at 3 dpf. Many cells in the photoreceptor layer do not have the elongated appearance characteristic of wild-type photoreceptors. The brain shape of both mutants appears to be unaffected (Fig. 4B,C). Numerous cell corpses, predominantly in the photoreceptor cell layer, are present in the retinae of both mutants at 3 dpf.

In 5 dpf retinae of elipsa (eli)^m649 and brudas (bru)^m148, most photoreceptors are missing. In contrast to mok^m632 and nie^m743, however, some morphologically normal cells are present in the periphery of the photoreceptor layer (arrowheads in Fig. 5D,H). The onset of the eli^m649 phenotype is late relative to other mutations in this group. At 3 dpf the photoreceptor cell layer of eli^m649 is relatively normal (Fig. 5C). Excessive cell death is noticeable mainly in the center of the retina (arrowhead in Fig. 5C). The retinal phenotype of bru^m148 is much more evident at this stage (Fig. 5H). Most photoreceptors in the bru^m148 retina are missing at 3 dpf. Similar to eli^m649 at 5 dpf, some photoreceptors persist in the periphery of the bru^m148 retina (arrowhead in Fig. 5H). For both loci the retinal abnormalities are associated with other defects. Starting during day 3 of development, bru^m148 mutants are characterized by a reduced touch response and a darker pigmentation. eli^m649 mutants have curled body axis and pronephric cysts (Iain Drummond, personal communication).

In krenty (krt)^m699, sinusoida (sid)^m604 and discontinuous (dis)^m704, the photoreceptor cell layer is characterized by discontinuities (arrowhead in Fig. 5E,F) which are already distinct at 3 dpf. The pattern of photoreceptor deficiencies persists unchanged till 5 dpf (Fig. 5F). In krt^m699, gaps in the photoreceptor cell layer seem to be associated with abnormalities of the inner nuclear layer. On histological sections of wild-type retinae, a subset of cells in the inner nuclear layer (presumptive amacrine cells) is less intensely stained (Fig. 1C). In krt^m699 these weakly staining cells are more numerous in the vicinity of the photoreceptor-deficient areas (arrowhead in Fig. 5E). The krt^m699 phenotype is variable. In the least affected individuals the photoreceptor cell layer is normal; in the most severe cases the krt^m699 phenotype is associated with extensive cell death. The phenotypes of all three mutants involve a slight decrease in brain size, which in krt^m699 and sid^m604 is already present during day 4 of development. In dis^m704 the brain defect is not obvious until day 5.

The nine mutations in this category are characterized by reduced eye size (Table 1). Histological sections reveal that all neuronal laminae are present in these mutants at all inspected stages (Fig. 6C, and data not shown), and cell death in excess of wild-type levels is not observed. The most severe phenotype in this group is present in out of sight (out)^m233mutants (Fig. 6A,B). Three alleles were found at this locus. The out^m233 phenotype is distinguishable during day 2 of development. The other two alleles, out^m306 and out^m390 are weaker and do not produce any obvious phenotype until 5 dpf. Pigmentation appears to be darker in out^m233 than in its wild-type siblings at 5 dpf. From a dorsal view the brain appears narrower in some out^m233 individuals. Other organs have no obvious mutant phenotype.

Mutants in this class (Table 1) are characterized by reduced eye size and brain shape defects (Fig. 6D,E,G,H). Histological sections reveal extensive cell death in their retinae. Cell corpses appear as small, round, intensely staining particles (Fig. 6F). Cell death is not localized to any particular cell layer in the retinae of these mutants. The shape and localization of plexiform layers is frequently abnormal and variable (Fig. 6F,I). These patterning defects are most likely due to cell death. In ziemniok (zem)^m709 the photoreceptor cells appear to be more affected by cell death than other neurons (not shown). All mutations in this category lead to slightly reduced brain size starting at 2 to 3 dpf. The brain abnormalities are most evident in the forebrain and midbrain.

Reduction of eye size and increased amount of cell death in the retina are associated with a general loss of pigmentation in this group of four mutations belonging to three complementation groups (Fig. 7). The pigmentation defect becomes distinct during day 2 of development and the eye size defect during day 4. In all mutants the majority of melanocytes are condensed and round at 5 dpf (Fig. 7B,E,H). Histological sections through the mutant retinae reveal excessive cell death (Fig. 7). The highest concentration of cell corpses is present near the marginal zone of the piegus (pgu)^m286 retina (arrow in Fig. 7K). In contrast, cell corpses in the mizerny (miz)^m293 retina are evenly distributed along the neuronal laminae (Fig. 7L). In the most severely affected mutant individuals, gaps appear in the photoreceptor cell layer of punktata (pkt)^m288 and miz^m293. In pkt^m288 at 5 dpf photoreceptors have abnormal appearance and are frequently reduced in number (Fig. 7F). Growth of pkt^m288 and miz^m293 is slightly retarded at 5 dpf and both mutants have somewhat abnormal brain shape. Ear otoliths are smaller in all mutants belonging to this group (Malicki et al., 1996).

The mutation pandora (pan)^m313 produces a very pleiotropic phenotype. At 24 hpf pan^m313 embryos are characterized by abnormal tail, brain and otic vesicle (not shown). Later in development pan^m313 mutant embryos are smaller than their wildtype siblings, do not develop body pigmentation, and display heart and circulation defects. At 1 dpf, eyes of mutant embryos appear to have abnormal morphology in the area of the optic stalk. Pigmentation develops only in the dorsal eye (not shown). Sections through the mutant retinae reveal a variable degree of cell death. At 3 dpf the retinal neurons frequently do not form a wild-type, laminar pattern. In the least affected embryos the retinal stratification develops normally (Fig. 3H). In wild-type embryos, the exit point of the optic nerve is located somewhat ventral to the mediolateral axis of the retina (Fig. 1). Interestingly, at 3 dpf in pan^m313 the portion of the retina ventral to the optic nerve (arrowhead in Fig. 3H) is missing.

The second locus in the miscellaneous category is heart and soul (has) (Fig. 2D). Similar to mutants in the neuronal patterning class, has^m129 has an eye-specific pigmentation defect, abnormal brain shape and a defective heart. has^m129 embryos become retarded during day 3 of development. Plexiform layers do not develop in the has^m129 retina and we observed numerous cell corpses on sections prepared from 60 hpf embryos (Fig. 3D). Due to the delay in development, it is difficult to assess whether retinal neurons in has^m129 are capable of forming normal lamination. Both mutations in this category complement Group I loci.

In a large scale mutagenesis screen we analyzed progeny of over 10,000 crosses between siblings from 1,808 F2 generation families bred from ENU mutagenized males (Driever et al., 1996). We recovered 49 recessive mutations with defects in the retina of the zebrafish eye. Several characteristics of zebrafish made this approach possible. High fecundity allows the efficient search for recessive phenotypes in a single clutch of embryos. Extrauterine development enables easy visual inspection of the developing embryos. Small size and relative hardiness allow for maintenance of a large number of lines at a relatively low cost. These characteristics will also prove helpful in further genetic and phenotypic characterization of the identified loci.

The screening procedure used in this study was aimed at the isolation of relatively obvious morphological abnormalities in the developing eye. Minor deficiencies of eye size or shape were probably not recorded. During day 5 of the screening protocol we were searching for abnormal size or shape of the eye. This effort resulted in the isolation of mutations which produce rather drastic deficiencies, mainly in the outer cell layers of the retina. Mutations resulting in minor changes in cell count or survival rates of retinal neurons would not have been isolated in this screen.

Genetic analysis is subject to two general constraints. These are redundancy and pleiotropy of function at some genetic loci. The second constraint is particularly relevant in the analysis of eye development, which takes place relatively late in ontogenesis. Late functions of pleiotropic loci are frequently obscured by an early function. Numerous examples of this are known from invertebrates (Simon et al., 1991; Cheyette et al., 1994; Pan and Rubin, 1995). Clearly, for this reason, some pleiotropic loci important to retinal development may have been discarded as not sufficiently specific, or included in phenotypic categories not obviously related to eye development (Abdelilah et al., 1996; Schier et al., 1996; Solnica-Krezel et al., 1996). In invertebrates, late functions of such loci are routinely studied in mosaic animals (Simon et al., 1991; Xu and Rubin, 1993; Pan and Rubin, 1995). Mosaic analysis is well established for zebrafish (Ho and Kane, 1990) and will undoubtedly prove useful in future studies of selected mutants. The degree of saturation achieved in a genetic screen can be estimated from the average number of alleles recovered at each locus. In the group of mutants affecting eye development we isolated an average of 1.4 alleles per locus. Approximately three quarters of the loci included in complementation testing are represented by a single allele (Table 1). This indicates a low degree of saturation and implies that many additional loci with function in eye development could be recovered by future morphological screening.

The mutants presented here are relevant to several aspects of central nervous system development. One of the major questions concerning formation of the retinal pattern is how particular cell types are able to find their final, proper positions within the neuronal layers. Three zebrafish mutants, oko meduzy^m98,nagie oko^m227 and glass onion^m117, cause loss of neuronal stratification in the retina. Presence of well developed plexiform patches and lack of extensive cell death in ome^m98 suggests that the retinal neurons in this mutant differentiate but are unable to form neuronal laminae. The phenotypes of ome^m98 and nok^m227 are very similar and largely confined to the retina, suggesting that these loci may belong to a single genetic pathway. On the other hand, glo^m117, produces an earlier phenotype and affects the entire neural tube. The product of the glo locus is thus more likely to play a general role in the development of neuroepithelium.

Many structures of the vertebrate brain consist of several neural cell types organized into distinct laminae (Jacobson, 1991). Some interesting insights into the formation of neuronal lamination came from studies of mouse mutants (Caviness and Rakic, 1978). Studies of two loci have been particularly informative. The reeler (rl) mutation causes extensive disorganization of neuronal laminae in both cerebral and cerebellar cortex of the mouse. The reeler locus has been recently shown to encode an extracellular matrix protein with a homology to F-spondin (D’Arcangelo et al., 1995). Another mutation, weaver (wv), produces degeneration of granule cells prior to their migration into the appropriate cell layer in mouse cerebellum (Caviness and Rakic, 1978). This phenotype appears to be associated with a defect in a homotypic signaling event among precursors of the granule cells (Gao et al., 1992; Gao and Hatten, 1993). Further study of the neuronal patterning loci ome, nok and glo may lead to identification of novel patterning mechanisms relevant to the neuronal lamination of the retina and, possibly, other regions of the vertebrate brain.

Mutations producing defects in the outer retina are relevant to the issue of cell fate acquisition in the central nervous system. Only preliminary information is available concerning factors involved in neuronal specification and differentiation in the vertebrate retina (Watanabe and Raff, 1990; Pittack et al., 1991; Altshuler et al., 1993; Kelley et al., 1994) . The pheno-types of the mutants mikre oko^m632, niezerka^m743, krenty^m699, sinusoida^m604 and some others may be associated with defective specification of the photoreceptor cells. Alternatively, they may involve deficiencies of factors necessary for photoreceptor survival. It seems likely that at least some mutants in this category also affect cells of the inner nuclear layer.

Interestingly, defects in the outer retina involve three different patterns of photoreceptor loss. In mok^m632 and nie^m743 the photoreceptor cells are missing in both the central and the peripheral retina. A different pattern is seen in krt^m699, sid^m604 and dis^m704. In these mutants, photoreceptor cells differentiate in patches along the ventricular margin of the retina. The mutations elipsa^m649 and brudas^m148 display the third pattern in which the defect is localized in the center of the retina, while some peripheral photoreceptors remain, at least temporarily, unaffected.

Mutations in genes of the phototransduction cascade have been shown to cause loss of photoreceptor cells (Bowes et al., 1990; Dryja et al., 1990). In mice, mutations of this type lead to late onset phenotypes (Chang et al., 1993). In the marginal zone of the teleost retina neurons continue to be generated throughout the lifetime (Raymond, 1991). One would expect that mutations in the phototransduction cascade loci should primarily affect differentiated photoreceptor cells of the central retina while having little impact on newly born, peripheral cells. Two mutants, eli^m649 and bru^m148, indeed display such a pattern. Unlike the phototransduction cascade mutants known in the mouse, both of them also produce strong phenotypes outside the visual system (Table 1).

A common human inherited retinopathy, retinitis pigmentosa, involves photoreceptor degeneration. The genetic causes of this disorder are very diverse. It is estimated that mutations in 50 to 100 loci are capable of causing this disorder (Dryja and Berson, 1995). Only a small fraction of them has been identified to date (Dryja et al., 1990; McLaughlin et al., 1993; Kajiwara et al., 1994). Further studies of the genetic defects included in group III may shed light on other causes of this disease.

Growth of the retina appears to be a precisely regulated process. The factors that play a role in this regulation are currently poorly understood. The mutation out of sight (out)^m233 as well as others in its group may provide an entry point to the study of this issue. eyeless-1,-2, (ey-1, ey-2), ocular retardation (or), fidget (fi), Small eye (Sey) and a few other mouse mutants produce microphthalmia or anophthalmia (Green, 1990). Their phenotypes are different from out^m233 and related mutants. For example, Sey is also involved in development of the nasal pits (Hogan et al., 1986). The optic nerve of or^J does not exit the optic cup (Silver and Robb, 1979). In addition to the eye defect, fidget mice do not form semicircu lar canals in the otic vesicle (Truslove, 1956). Such defects are not observed in out^m233 or related mutants. The fi retina is properly stratified. The reduction of eye size in this mutant has been shown to involve prolongation of the cell cycle in the retinal anlage (Konyukhov and Sazhina, 1976).

Mutations included in Group VI are reminiscent of the microphthalmia (mi) phenotype characterized in the mouse (Hertwig, 1942). Similarly to the zebrafish mutants piegus^m286, punktata^m288 and mizerny^m293, mouse mutations in the mi locus affect eye size, ocular pigmentation and differentiation of melanocytes. Molecular characterization of the mouse locus revealed that it encodes a putative transcription factor expressed in the pigmented epithelium (Hodgkinson et al., 1993). Given the phenotypic similarity, it is possible that one of the mutations included in this group affects the zebrafish homolog of the microphthalmia gene. The human auditorypigmentary syndrome, Waardenburg syndrome type 2, is caused by mutations in the MITF gene, the homologue of microphthalmia (Tassabehji et al., 1994). Further studies of the zebrafish mutants that combine eye and pigmentation defects may prove valuable to understanding of this type of human disorders. The defect of the photoreceptor cell layer in punktata^m288 may have its primary causes in the pigmented epithelium, which appears to be abnormal in this mutant. The development of the photoreceptor outer segments is known to depend on interaction with the pigmented epithelium, (Hollyfield and Witkovsky, 1974). pkt^m288 and the related mutants may thus reveal novel aspects of interaction between photoreceptor cells and the pigmented epithelium.

The mutation pandora^m313 appears to be relevant to yet another aspect of eye development. The dorsoventral polarity of the vertebrate eye is evident by both morphological and molecular criteria (Constantine-Paton et al., 1986; McCaffery et al., 1991; Schmitt and Dowling, 1994). pan^m313 causes absence of the ventral retina at 3 dpf. This phenotype suggests that the pandora locus may play a role in the specification of the dorsoventral axis in the zebrafish eye. Retinoic acid has recently been shown to influence the ventral cell fates in the zebrafish optic cup (Marsh-Armstrong et al., 1994). It is note-worthy that the eye defect produced by citral, an inhibitor of retinoic acid synthesis, is strikingly similar to the pan^m313 phenotype (Marsh-Armstrong et al., 1994).

Alternative screening methods for loci participating in the zebrafish eye development are conceivable. A screening procedure for retinal defects based on a functional test is available in zebrafish (Clark, 1981; Brockerhoff et al., 1995). It takes advantage of the optokinetic response, which in zebrafish appears early (4 dpf) in development (Clark, 1981). The assumption of this approach is that fish larvae with retinal defects will not display a normal optokinetic response. A screen based on the optokinetic response can potentially detect mutants with minor developmental abnormalities, which do not result in obvious morphological changes of the eye. Similarly, mutations producing abnormal physiological characteristics of the retina can be recovered in such a screen (Brockerhoff et al., 1995). The disadvantage of screening procedures based on the optokinetic response is that they are substantially more laborious than screens based on morphological criteria.

Our studies demonstrate that zebrafish provide a powerful genetic system to study various aspects of vertebrate eye development, including: specification of the eye primordium, neuronal lamination, specification/differentiation of distinct neuronal classes and possibly the specification of axes. A detailed analysis of the identified mutants will enhance our understanding of vertebrate central nervous system development.

We thank Colleen Boggs, Jane Belak, Ioannis Batjakas, Heather Goldsboro, Lisa Anderson, Snorri Gunnarson and Pamela Cohen for technical help during the various stages of the screen. We are also grateful to Eliza Mountcastle-Shah for critical reading of the manuscript and Lisa Anderson for help with naming mutants. This work was supported in part by NIH RO1-HD29761 and a sponsored research agreement with Bristol Myers-Squibb (to W. D.). Further support in the form of fellowships came from HFSP and the Fullbright Program (to Z. R.), EMBO and Swiss National Fond (to A. S.), Helen Hay Whitney Foundation (to D. L. S. and D. Y. S.), and the Damon Runyon-Walter Winchell Cancer Research Fund (to J. M.).