The Westerfield Lab Institute of Neuroscience 1254 University of Oregon Eugene, OR 97403-1254 USA

Zebrafish models of human Usher syndrome:

Human Usher syndrome, the most frequent cause of deaf blindness, affects about 1 in 17,000 Americans and is characterized by retinal degeneration due to progressive loss of photoreceptors and congenital deafness. Although at least 12 different loci have been linked to Usher syndrome and we know the identity of nine genes, etiology of the disease and underlying mechanisms are still poorly understood. As a first step toward development of strategies for treatment, we need to learn how the Usher proteins function and how mutations result in sensory defects.

We used TBLASTN searches to locate zebrafish orthologues of all known human USH genes, and analyzed sequence similarity to establish homology within the functional domains of the genes. We amplified transcripts of ush1c, ush2a and ush2c from adult zebrafish retinas and used cRNA probes to characterize their expression patterns in larval tissues. We injected antisense morpholino oligonucleotides designed to knock down expression of these genes in developing embryos and larvae. TUNEL labeling and vital staining of apoptotic cells were used to assay retinal cell death, and antibodies to double cones and to synaptic markers were used to characterize the structural integrity of the retina. We assayed visual function in live larvae with the optokinetic response test.

All three genes are expressed in the neural retina and sensory hair cells of the ear and lateral line from neurogenesis stages through early larval stages. Transcripts are also present in the adult retina. Knock-down of zebrafish Ush1c, Ush2a and Ush2c by morpholino injection results in swimming and balance defects in young larvae similar to those seen in ush1b (myo7a) mutant fish. We observe reduced visual function and increased cell death in the retina within the first week of development. Cone photoreceptor integrity is slightly disrupted in the Ush2a morpholino-injected animals. Pre-and post-synaptic proteins in morpholino-injected animals at the photoreceptor synapses are abnormally localized, appearing patchy and poorly aligned.

Zebrafish USH orthologues are expressed in tissues consistent with the human disease pathology and at time points that suggest roles in development as well as maintenance of sensory cells. Depletion of functional Ush1c (Harmonin), Ush2a (Usherin) or Ush2c (VLGR1) proteins results in poor visual function and retinal cell death, although these defects are not coupled with a pronounced change in photoreceptor morphology. Photoreceptor synapses are perturbed in morpholino injected animals, suggesting that some Usher proteins may be important for establishing or maintaining synaptic integrity.

Visual System Basic Research:

During development, vertebrate embryos normally form two, bilateral eyes. The retinas are derived from the anterior neural plate. Under the influence of various genetic (Cohen, 1989; Hatta et al., 1991; Muenke et al., 1994; Schier et al., 1996; Solnica-Krezel et al., 1996; Ming and Muenke, 1998) and environmental factors (reviewed in Adelmann, 1936a; Roach et al., 1975; Blader and Strþhle, 1998), however, a single, median, cyclopic eye can form. Such observations have led to two hypotheses to explain the developmental origin of bilateral eyes. The first proposes that two fields of retinal precursor cells in the neural plate, separated by medial diencephalic precursors, independently produce the two eyes (Meckel, 1826; Spemann, 1904; 1912; reviewed in Adelmann, 1936a,b). This hypothesis predicts that during abnormal development, the two fields of retinal precursors fuse to form a cyclopic eye. The second hypothesis postulates a single field of retinal precursor cells that separates into left and right eyes (Huschke, 1832; Stockard, 1913; LePlat, 1919; Adelmann, 1929a,b,c; reviewed in Adelmann, 1936a,b), leaving diencephalic precursors medially. Failed separation of the single primordial neural plate eye field would result in cyclopia.

Fate map analyses of the gastrula and anterior neural plate in several species have indicated the presence of a single median field of retinal precursors (Ballard, 1973; Jacobson and Hirose, 1978; Hirose and Jacobson, 1979; Woo and Fraser, 1995). Cells labeled in the medial part of the anterior neural plate can contribute progeny to either or both eyes, suggesting that the field of retinal precursor cells extends across the midline. Movement of axial mesendodermal cells and the subsequent elongation of the neural plate midline have been proposed as playing a role in separation of the eyes (Woo and Fraser, 1995; Heisenberg and NŸsslein-Volhard, 1997; Marlow et al., 1998). In contrast, other fate map studies indicated that cells in the anterior midline of the neural plate do not contribute to the retina, but to diencephalic brain regions between the eyes (Couly and Le Douarin, 1988; Eagleson and Harris, 1990; Eagleson et al., 1995; Li et al., 1997) consistent with the hypothesis that two fields of neural plate cells give rise to the retinas. Based on this second set of observations, an alternate interpretation has been proposed. Under the influence of the underlying prechordal plate, median neural plate cells down regulate expression of genes characteristic of retinal precursors and instead contribute to the ventral diencephalon (Pera and Kessel, 1997; Li et al., 1997). This model predicts that in the absence of prechordal plate signaling, median neural plate cells would continue to develop as retinal precursors resulting in a single retina fused across the midline. It is unclear, however, how this model can account for the previous fate map studies that indicated that median cells contribute to the retinas (Ballard, 1973; Jacobson and Hirose, 1978; Hirose and Jacobson, 1979; Woo and Fraser, 1995).

To distinguish between these apparently contradictory hypotheses and to understand the morphogenesis of the diencephalon and the eyes, we fate mapped the anterior neural plate of zebrafish embryos at high resolution, measuring the movements of labeled cells relative to morphological landmarks and patterns of gene expression. We found a single field of retinal precursor cells that express the odd paired-like gene (opl, Grinblat et al., 1998) at the end of gastrulation. This field extends across the midline and all cells within the field contribute to the eyes. Diencephalic precursor cells express the forkhead gene, mariposa (mar, Moens et al., 1996; also called forkhead 3, Odenthal and NŸsslein-Volhard, 1998), and are located posterior to the opl expressing eye field. Median mar expressing cells move anteriorly along the midline, separating the retinal precursors into left and right eyes and forming the ventral diencephalon. In embryos mutant for Cyclops (Hatta et al., 1991), a nodal-related member of the Transforming Growth Factor-§ superfamily of signaling molecules (Rebagliati et al., 1998; Sampath et al., 1998), these posteriorly located cells express opl instead of mar and fail to move anteriorly. Thus, the eye field fails to separate and a single cyclopic eye forms. Ablation of mar expressing median ventral diencephalic precursors and underlying prechordal plate from wild-type embryos also prevents separation of the eye field and produces cyclopia. Our results suggest that movement of diencephalic precursors anteriorly along the midline is required to separate the primordial eye field into left and right eyes and that this morphogenetic movement requires cyclops gene function.

Two major questions in developmental biology are how do cells become different from each other and how are different cell types organized into functional tissues or organs. Analysis of mutant mice has identified many genes that control the organization of cells into layers in the brain (for review, see Rice and Curran, 1999). Generally, these mutations do not affect neurogenesis, suggesting that cell positioning is largely independent from cell-fate determination. As yet, no targeted mutations in mammals have been reported to cause prominent lamination defects in the retina, suggesting that the neural retina, given its unique relationship with adjacent tissues, the lens and retinal pigmented epithelium (RPE), uses fundamentally different processes to generate cell layers than those used for brain lamination. Recently, several mutations in zebrafish have been isolated that cause retinal lamination defects (Malicki et al., 1996; Malicki and Driever, 1999) and all of these mutations have accompanying RPE defects, suggesting that the RPE may be required for normal retinal lamination. Alternatively, the genes that regulate the positioning of retinal cells may also function in the development of the RPE. Evidence from in vitro and in vivo experiments suggests that the RPE is required for normal retinal organization (Vollmer et al., 1984; Rothermael et al., 1997; Raymond and Jackson, 1995). Vollmer and colleagues (1984) demonstrated that the inclusion of RPE cells in retinal reaggregate cultures could induce a highly organized arrangement of cells that was similar to that observed in vivo. Rothermel and colleagues (1997) further demonstrated that RPE conditioned media could have a similar affect on retinal reaggregates, suggesting that the RPE factor(s) is diffusible. In vivo experiments by Raymond and Jackson (1995), in which they genetically ablated RPE cells using diphtheria toxin, demonstrated that the development of layers in the retina was severely perturbed in the absence of RPE. Although these experiments suggest that the RPE has an important role in retinal organization, the mechanisms by which the RPE signals to the retina and the genes that regulate this process are not yet known.

We have isolated two mutations of the mosaic eyes (moe) gene in which there are RPE defects, a loss of lamination in the retina, a loss of localization of dividing retinal cells to the RPE surface, but no apparent disorganization of cells in the brain. Although the retina is severely disorganized, retinal cells differentiate in moe mutant retinas. The localization of dividing moe mutant retinal cells transplanted into wild-type hosts is normal, suggesting that the moe mutation does not act cell-autonomously in retinal cells. We show that moe function is required in the RPE; transplanted wild-type RPE cells can rescue the mislocalization of dividing cells in moe mutant retinas. Our analysis provides the first genetic evidence that signaling from the RPE to the retina is required for the induction of proper retinal organization. We suggest that this signaling depends upon moe gene function.