In this study, we took a forward genetic approach to identify genes involved in zebrafish visually controlled behaviors. In order to capture a large number of mutants, we screened almost 2,000 F2 families and cast a wide, dense net by screening with three complementary behavioral assays. We report here on the initial characterization of 53 specific mutations in 41 genes, only two of which had previously been described.
OKR versus OMR versus VBA as Screening Assays
Choice of a suitable assay is paramount to the success of any genetic screen. We found that each of the three assays employed had its specific strengths and limitations. The OKR assay requires each fish to be mounted individually, dorsal side up, in methylcellulose and is therefore much more time-consuming than the OMR assay, for which each group of fish can just be poured into an elongated tank. The OKR assay therefore dictated the pace of the screen, and we were thus unable to test as many fish as with the OMR assay (and may therefore have missed some mutants). However, since the OKR assay records fish individually, whereas the OMR assay records a population, the OKR has the potential to find less-penetrant phenotypes than the OMR. In the primary screen, OMR and OKR assays each discovered a largely nonoverlapping set of visual mutants, which, upon retesting, showed defects in either assay. Thus, the high throughput of the OMR assay complemented the specificity of the OKR assay. This tradeoff also applies to genetic linkage mapping, which we have so far completed for 25 of the 41 loci. We found that the OMR is most useful for presorting of mutants, while the OKR is most suitable for the subsequent “weeding-out” of false positives. The VBA response, on the other hand, is extremely effective in sorting mutants for linkage mapping, but is less suited as a primary screening assay, because it is prone to missing important mutant classes. Screening with all three assays increased the likelihood of finding all mutants and often provided independent confirmation of a behavioral phenotype.
How Many and What Kinds of Genes Control Visual Behavior?
We found that at least one-quarter, and probably more than half, of the behavioral mutations discovered here affect photoreception. Their phenotypes include defects in photoreceptor formation or maintenance, phototransduction, and adaptation to sudden light changes (whose likely cellular and molecular substrate is located in the outer retina). Another sizable fraction (at least a quarter) of mutations affect RGCs and their projections to the brain. As far as we can conclude so far from our ongoing analysis, mutations affecting the development of higher visual centers (beyond the retinofugal projections) are largely absent from our collection. This could mean that the genes involved in the formation of circuits in higher brain regions are either essential for embryonic development (i.e., their loss of function would lead to early lethality), or they are redundant, which would prevent their discovery by classical mutagenesis screens.
The number of genomes screened should have been sufficient to uncover at least one mutation in each gene of interest, based on the mutation rate measured in the F0 founder males. However, the empirical allele frequency clearly contradicts this optimistic scenario. Of the 41 loci in our collection, 35 are represented by a single allele and four by two alleles. The other two genes for which we found five alleles each, mti and wud, appear to be outliers. Excluding these two loci, and assuming that the probability of finding a mutation follows a Poisson distribution, the number of genes with no hits is estimated at about 150. This back-of-the-envelope calculation shows that our screen was not saturating, and that many more genes may be discovered using our approach. Potential obstacles to future screens include the intrinsic difficulty of detecting mutants in behavior, as opposed to, say, pigmentation (which was used to measure the mutation rate), and the low mutability of some loci, as has been observed in other large-scale zebrafish screens [31,32].
Satisfyingly, we discovered new alleles of several previously identified genes. These include mutants falling within the limits of our screening criteria, such as bel and nof (Table 1), as well as others with more severe phenotypes, such as chk , bru [21,22], ome, and nok  (unpublished data). It is possible that some of our mutations have generated weak (or maternally rescued) alleles of housekeeping or other essential genes, although the molecular identification of the first set of genes shows that this is not generally the case. For a precise estimate of the number of genes whose mutations lead to specific, nonlethal visual system phenotypes, a much larger screen will have to be carried out.
Genes Involved in Photopic Vision and Photoresponse Dynamics
Zebrafish fill an important niche for the genetic study of photoreception. Human pattern vision, like that of zebrafish, is largely cone-driven. Because most genetic work has been done on the rod-dominated retinas of rodents, less is known about phototransduction in cones. Here we have already discovered two mutant alleles of zatoichi (zats125 and zats376), the gene for cone-specific guanylyl cyclase (Gc3), as well as a new allele of nof, which encodes the alpha subunit of cone transducin . It is likely that there are additional mutants in phototransduction in our collection, and it will be interesting to study their genetic interactions. Zebrafish are appealing for this work, because all their cone opsin genes have been identified , and their photoreceptors are amenable for biochemical  and psychophysical studies .
The visual system operates over a wide range of luminance intensities by adjusting its sensitivity to ambient light levels. At least two adaptation mechanisms are operational in the vertebrate retina, one acting on the phototransduction cascade itself [36–38] and the other on synaptic strengths within the network of neurons . We have discovered five mutants that exhibit delayed recovery of the OKR following a sudden transition from dark to light. These mutants are otherwise normal and adult viable. We speculate that these mutants have defects in light adaptation, although further analyses, such as electroretinogram recordings, will be needed to define and localize the underlying defect. The mutations identified here should provide novel entry points into a molecular dissection of light adaptation.
Zebrafish Mutants as Human Eye Disease Models
We identified five genes whose mutations result in loss of photoreceptors. Several processes can lead to retinitis pigmentosa or macular degeneration in mammals, including structural defects of outer segments, excessive light illumination, and genetic disruption of the phototransduction cascade, but the molecular mechanisms of cell death induction are largely unknown . Photoreceptors are lost quickly in our zebrafish mutants (over days), in contrast to rodent models of retinal degeneration, in which the same process takes months . This is advantageous for the screening of therapeutic drugs that block photoreceptor degeneration. Tests of pharmacological rescue could be carried out in conjunction with our high-throughput behavioral assays. Our collection of zebrafish mutants with rapidly degenerating cones provides us with novel tools to examine the molecular mechanisms of macular degeneration in a model system that is not only genetically tractable, but amenable to small-molecule screens .
Axon Targeting and Functional Neuroanatomy
Our screen successfully identified a small assortment of specific axon-guidance mutants. These mutants will serve as starting points for the discovery of proteins involved in axon targeting and synaptic specificity in the visual pathway. But their phenotypes are also significant for assigning function to certain pathways in the zebrafish visual system . While most RGCs project to the midbrain tectum, nine smaller areas, or AFs, also receive direct retinal input . Different AFs are innervated by molecularly and spatially distinct subpopulations of RGCs  and probably mediate different visual behaviors. Laser ablations have shown that the tectum is required for localization of prey , but is dispensable for OMR, OKR, and VBA . An intact AF-7 is also not necessary for OMR or OKR . Some of the new mutants now help us narrow down the optomotor pathway further by providing “lesions” that are impossible to obtain using surgical, pharmacological, or optical ablation techniques. For instance, in the OMR-deficient misss522 mutant, AF-4 and AF-9 are reduced. This suggests, but does not prove, that one of these underdeveloped AFs is necessary for the OMR. Conversely, darls327 mutants lack AF-2, AF-3, and AF-6, but have an intact OMR, indicating that these three AFs are dispensable for this behavior. Based on these phenotypes, we predict that either AF-4 or AF-9 (or both) are required for the OMR.
Systematic forward genetic approaches have been applied with great success to many areas of biology in a variety of model species. Mutants are not only starting points for gene discovery; their phenotypes often elucidate underlying biological mechanisms even before molecular identification of the mutated genes (e.g., ). Our behavioral screen focusing on the zebrafish visual system has achieved three major goals. First, the mutant phenotypes found here have revealed novel genes, or new functions for known genes, which can be identified by positional cloning. Second, these mutations provide novel tools to study central nervous system development and behavior, to localize functions in the brain and to explore the ways in which neuronal circuits reorganize in response to genetic perturbations. Third, our unbiased screen is yielding fundamental insight into the genetic architecture of brain functions and their pathologies. A mutational approach to circuit formation and function, while being an essential first step, should be complemented in the future by targeted manipulations of cells and synapses. Zebrafish are slated to become an excellent system for an integrated genetic approach to unravel cellular and molecular mechanisms of behavior.