A whole genome RNAi screen of Drosophila S 2

Cells can adopt a wide range of morphologies, from round lymphocytes to highly branched Purkinjee neurons. Major challenges of modern cell biology are to identify the proteins that are involved in cell shape determination and understand how these proteins are regulated by external and internal signals. Many of the best-studied proteins involved in cell shape determination are components or direct regulators of the actin cytoskeleton (Faix and Rottner, 2006), although many other proteins appear to play roles as well (Randazzo et al., 2007). Not surprisingly, different sets of proteins contribute to the determination of morphology in different cell types (Liu et al., 2009). RNAi screens provide powerful approaches for identifying proteins involved in cell biological activities. However, only a limited number of screens have been performed on cell shape. Initial screens by Kiger et al. (2003) and Rogers et al. (2003) examined the effect of knocking down a limited set of genes (994 and 96 genes, respectively) on cell shape in Drosophila melanogaster tissue culture cells. At that time, the acquisition of image data was performed manually on the microscope, and the image data were analyzed by laborious visual inspection. Since that time, technological advances in robotic microscopy have allowed fully automated image acquisition, making it possible to obtain images from a whole genome RNAi screen in only a few weeks. As a result, the image analysis has become the rate-limiting step, and in most cases, the amount of image data for a whole genome screen exceeds what can be reasonably and reliably analyzed through visual inspection. Complex visual data, such as cell shape, are also best compared quantitatively rather than by qualitative assessment. Thus, developing new approaches for automated quantitative image analysis has now become a greater challenge than collecting the primary image data. An important step in automated, computational analysis was taken by Bakal et al. (2007), who used computational methods and neural networks to classify phenotypes in a screen with 249 genes predicted to play a role in cell morphology. More recently, Sepp et al. (2008) performed an automated whole genome screen of Drosophila primary neuron cells for genes that regulate neurite outgrowth using algorithms designed to analyze Recent technological advances in microscopy have enabled cell-based whole genome screens, but the analysis of the vast amount of image data generated by such screens usually proves to be rate limiting. In this study, we performed a whole genome RNA interference (RNAi) screen to uncover genes that affect spreading of Drosophila melanogaster S2 cells using several computational methods for analyzing the image data in an automated manner. Expected genes in the Scar-Arp2/3 actin nucleation pathway were identified as well as casein kinase I, which had a similar morphological RNAi signature. A distinct nonspreading morphological phenotype was identified for genes involved in membrane secretion or synthesis. In this group, we identified a new secretory peptide and investigated the functions of two poorly characterized endoplasmic reticulum proteins that have roles in secretion. Thus, this genome-wide screen succeeded in identifying known and unexpected proteins that are important for cell spreading, and the computational tools developed in this study should prove useful for other types of automated whole genome screens. A whole genome RNAi screen of Drosophila S2 cell spreading performed using automated computational image analysis