24.3 n&v 443 MH

begun to provide a comprehensive picture of the hereditary information required for various forms of life. But what do all the genes so identified do to warrant their existence? In the unicellular yeast Saccharomyces cerevisiae, for example, all predicted genes have been deleted,and the resulting strains subjected to a battery of analyses. These genome-wide studies have shown that only about 19% of yeast genes are involved in essential processes. In multicellular organisms, deciphering gene function on a genome-wide scale is a much tougher prospect. Here, one of the main laboratory subjects is the nematode worm, Caenorhabditis elegans (Fig. 1). On page 462 of this issue Sönnichsen et al. report their testing of nearly every C. elegans gene for functional requirements during embryo formation and larval development. They report that 1,668 genes display a visible phenotype, or defect, when reduced in function. A surprisingly small number (661) are essential for distinct cell processes, and embryonic viability, during the first two embryonic cell divisions. In 1998, the release of the C. elegans genome sequence provided the first complete catalogue of the predicted gene content of a multicellular organism. These data have allowed bioinformatic comparisons with other genomes. For example, sequence analyses reveal that about 50% of human genes have homologues in yeast, fruitflies or worms. Also, so-called microarrays — glass slides with thousands of spots, each containing many copies of DNA specific for one gene — have allowed the rapid and systematic profiling of gene expression in numerous environmental conditions, at different developmental stages, and following genetic perturbations. Next came methods to understand gene function in relation to the phenotypes produced when genes are eliminated. Most recently, the remarkable phenomenon of RNA interference (RNAi) — the ability of double-stranded RNA (dsRNA) to interfere with the production of the corresponding gene product — has allowed worm researchers to ‘knock down’ the expression of any given gene. RNAi was first applied by microinjecting dsRNA into worms. More recently, it was shown that feeding worms dsRNA produced in bacteria (C. elegans’ favourite meal) also reduces gene function, leading to the first genome-wide tests for gene function in C. elegans. These feeding RNAi studies have revealed phenotypes for about 10% of the worm’s genes, with half of these (929) resulting in death of the embryo when they are reduced in function. But the phenotypes described in this first genome-wide study were divided into very broad categories (for example ‘embryonic lethal’, ‘sterile’, ‘larval arrest’ and so on). Although this information, and the reusable feeding RNAi library it generated, are extremely useful, more specific phenotypic descriptors were not provided. For example, embryonic lethality can result from any number of defects in just the first embryonic cell division. Sönnichsen et al. have addressed this limitation. They performed the herculean task of microinjecting dsRNAs corresponding to 98% of C. elegans genes into adult worms, and then scoring the progeny of microinjected animals for phenotypes (‘progeny tests’). The injected worms’ progeny were examined for embryonic lethality, larval arrest, morphological and neurological defects, and sterility. The authors thereby discovered the functions of 9% of worm genes. Compared to the feeding RNAi screens, microinjection identified a greater number of genes involved in embryonic development (1,266 genes in the injection study versus 929 in the feeding study). But fewer genes yielding larval and adult phenotypes were found. Thus, different RNAi methods may target different processes more or less effectively; other studies support this conclusion. The major advance provided by Sönnichsen et al. is a much more detailed analysis of all the early embryonic-lethal phenotypes reported. The embryos of C. elegans offer exceptional accessibility and clarity for observing early cell divisions,allowing analysis, at the cellular level, of the effects of gene knockouts. Thus, if an RNAi progeny test yielded an embryonic-lethal phenotype, Sönnichsen et al. proceeded to record the first two cell divisions using live cell imaging with standard light microscopy methods. The authors made over 40,000 time-lapse movies of developing embryos, following events in the first two cell divisions in each movie. These movies were then annotated for 45 distinct cellular phenotypes, including defects in such processes as passage through meiosis, centrosome attachment, mitotic spindle assembly and sister chromatid separation. Obviously, these phenotypes go well beyond the designation of ‘embryonic lethality’ and point to specific cellular requirements. These data are available for searching and viewing on the authors’ website. Furthermore, the 661 genes important for early embryogenesis were divided into groups with related functions. Thus, gene pathways were shown to function in specific cell processes. For example, subunits of the proteasome, a complex involved in protein disposal, are highly enriched in the genes required for passage through meiosis, confirming that protein degradation is required during the meiotic cell cycle. Sönnichsen and colleagues’ data offer remarkable insights into the cell biological functions of a large but manageable set of genes. This data set represents only a jumping-off point for even more specific functional studies, on a gene-by-gene basis. For example, determining the subcellular localization of the proteins required for a distinct process will provide a straightforward first step in determining whether such proteins news and views