© 2006 Nature Publishing Group Identification of pathways regulating cell size andcell-cycle progression by RNAiMikael Bjo¨rklund1,3*, Minna Taipale1,3*, Markku Varjosalo1,3, Juha Saharinen3,4, Juhani Lahdenpera¨2& Jussi Taipale1,2,3Many high-throughput loss-of-function analyses of the eukaryoticcell cycle have relied on the unicellular yeast species Saccharo-myces cerev isiae and Schizosaccharomyces pombe. In multicellularorganisms, however, additional control mechanisms regulate thecell cycle to specify the size of the organism and its constituentorgans1. To identify such genes, here we analysed the effect of theloss of function of 70% of Drosophila genes (including 90% ofgenes conserved in human) on cell-cycle progression of S2 cellsusing flow cytometry. To address redundancy, we also targetedgenes involved in protein phosphorylation simultaneously withtheir homologues. We identify genes that control cell size, cyto-kinesis, cell death and/or apoptosis, and the G1 and G2/M phasesof the cell cycle. Classification of the genes into pathways byunsuper vised hierarchical clustering on the basis of these pheno-types shows that, in addition to classical regulatory mechanismssuch as Myc/Max, Cyclin/Cdk and E2F, cell-cycle progression in S2cells is controlled by vesicular and nuclear transport proteins,COP9 signalosome activity and four extracellular-signal-regulatedpathways (Wnt, p38bMAPK, FRAP/TOR and JAK/STAT). Inaddition, by simultaneously analysing several phenotypes, weidentify a translational regulator, eIF-3p66, that specifically affectsthe Cyclin/Cdk pathway activity.The cell cycle can be divided into distinct phases including asynthesis (S) phase, where DNA is replicated, and a mitosis (M)phase, where cell division occurs. In animal cells, growth and thesynthesis of components required for these phases are regulated byextracellular growth factors and occur mainly in two gap phases, G1(between M and S) and G2 (between S and M)2–4. The orderedprogression of the cell-cycle phases is orchestrated by cyclin-depen-dent kinases (Cdks), whose activity is controlled by phosphorylationand by association with specific regulatory subunits, the cyclins2–4.The cell cycle is a robust system in which compensatory mecha-nisms control its overall length5. We therefore analysed the effect ofRNA-mediated interference (RNAi)-induced loss of function ofDrosophila genes in S2 cell-cycle distribution using flow cytometry,which allows direct determination of the fraction of cells in differentphases of the cell cycle (Supplementary Fig. S1).Consistent with the high efficiency of RNAi in Drosophila6–9, clearphenotypes including arrest in G1, G2/M and S, cytokinesis andDNA replication defects, and apoptosis and/or cell death wereobserved when targeting known regulators of these processes(Fig. 1a and Supplementary Fig. S2). We next carried out a pilotscreen targeting Drosophila kinases and phosphatases individually(Fig. 1b and Supplementary Fig. S3), and in pools of homologues toaddress redundancy (Fig. 1c and Supplementary Fig. S4). Unexpect-edly, the analysis of redundancy did not reveal any additional cell-cycle regulators (Fig. 1c and Supplementary Table S3). These resultssuggest that a considerably lower fraction of Drosophila kinasesregulate the cell cycle (19%) than has been reported previously(35%, ref. 10; for explanation and comparison, see SupplementaryTable S4 and Fig. S5). Despite this, our screen identified also kinasesfor which a role in the S2 cell cycle has not been appreciated (forexample, AKT1, CG7177 and MEKK1/MEKK4).We reasoned that, by screening most Drosophila genes, we shouldbe able to identify the pathways regulating the S2 cell cycle. Screeningof 11,971 double-stranded RNAs (dsRNAs) generated from theDrosophila Gene Collection (DGC) releases 1 and 2 (see Supplemen-tary Methods and refs 7, 8) showed that loss of 270 and 169 genesresulted in significant changes in G1 and G2 populations, respectively(Supplementary Fig. S6). A strong correlation between cell size at theG1 and G2 phases was observed in individual samples, suggestingthat Drosophila cells do not have a ‘strong’ cell-size checkpoint11that forces cells to a particular size at a defined cell-cycle phase(Supplementary Fig. S7). Components acting on the same directionin a particular process, such as ribosomal proteins or Dp, E2f andCyclin E, appeared in discrete areas of a plot describing G1 cell size asa function of fraction of cells in G1 (Fig. 2a). The strongestphenotypes, characterized by specifically decreased cell size, wereobserved with dsRNAs targeting Rbf, a negative regulator of E2fpathway, and cropped (crp), the Drosophila AP4 transcription factor(Supplementary Tables S3 and S5).To identify genes involved in cell death and cytokinesis, weanalysed changes in the populations of cells with less than 2N DNAor more than 4N DNA, respectively (Fig. 2 and Supplementary Fig. S8).The increase in the ,2N population in all cases consisted of apoptoticor dead cells; the genes whose loss caused this phenotype includedpreviously unknown effectors, the known inhibitor of apoptosisThread and several mitosis regulators, consistent with induction ofapoptosis by mitotic catastrophe12(Fig. 2b, and Supplementary Figs S8and S9 and Table S6). The increase in the .4N population wasdue to cells that had undergone two rounds of DNA replicationwithout cytokinesis (8N DNA); dsRNAs causing this phenotypetargeted several previously undescribed genes and known cytokinesisregulators9(Fig. 2c, and Supplementary Fig. S8 and Table S6).There have been several large-scale RNAi screens, and in manythere is very little overlap between the gene sets identified even whenthe same phenotypes are analysed. This is apparently due to a verylarge number of false positives (see Supplementary Tables S7 and S8,and Fig. S10). To reduce the number of false positives, we analysed allphenotypes by using criteria that excluded all 564 control samples(.99.99994% confidence). Roughly 4% of genes (488; SupplementaryTable S3) had a cell-size, cell-cycle or cell-death phenotype. From ananalysis of 19 different pathways and/or protein complexes, weestimate that our screen identified ,80% of strong non-redundantLETTERS1Molecular and Cancer Biology Program, and2High Throughput Center, Biomedicum Helsinki, PO Box 63 (Haartmaninkatu 8), FI-00014 University of Helsinki, Finland.3Department of Molecular Medicine, National Public Health Institute, and4Biomedicum