Here we identify multiple novel regulators of cell morphology, cytoskeletal organisation and migration, many of which are implicated in human diseases and signal transduction processes. Sixteen genes were chosen based on their identification in a genome-wide Drosophila cell morphology screen followed by bioinformatic analysis of their domain structure, interacting partners and potential functions in other species. Of the 26 human genes identified as homologues of these Drosophila genes, nearly all affected cell shape and/or migration in PC3 prostate carcinoma cells, implying that these genes have conserved functions between Drosophila and human cells. Based on morphological analysis, these genes could be divided into six distinct groups, and this grouping was reflected in their effects on migration. This could in the future allow prediction of migratory behaviour based on cytoskeletal and shape analysis.
The PMM genes have not previously been identified in RNAi-based migration screens; for example, they were not in the gene set analysed by Simpson et al.  because they are not kinases, phosphatases or previously known regulators of cell migration. They were not identified in another migration screen of 5,234 genes , but the complete gene list is not provided for this screen.
Of the seven PMM genes chosen for in-depth characterization, the closely related FAM40A and FAM40B genes had clearly distinct phenotypes: FAM40A-depleted cells had an increase in F-actin around the periphery and a circular shape, whereas FAM40B-depleted cells were highly elongated. However, knockdown of each of these genes only led to a mild decrease in migration in scratch wounds. Indeed, other genes that induced cell elongation in PC3 cells did not affect migration, indicating that under the conditions we have analysed elongation is not predictive of a cell migration defect. Very little is known about FAM40 genes but interestingly they are conserved from fungi and amoebae to humans, although we have been unable to find a homologue in plants. The yeast homologue is FAR11 (pheromone arrest 11), which in S. cerevisiae interacts with other FAR proteins and causes a Far3-like pheromone arrest phenotype when disrupted . Protein-protein interaction databases report that FAR11 interacts with Rom2, a GEF for the Rho GTPases Rho1 and Rho2 (BioGRID; ), as well as Rho4 and Tpd3 . In human cells, FAM40A and FAM40B have recently been isolated as part of a striatin-interacting phosphatase and kinase (STRIPAK) complex, that also contains PP2A phosphatases, striatins, MOB1/3 and members of the Ste20 kinase family . Interestingly, S. cerevisiae Far8 shares sequence similarity with the striatins, and Tpd3, a PP2A-A orthologue, has been found as an interacting partner with Far11 in yeast two-hybrid assay . Altogether these data suggest that FAM40A and FAM40B could be part of a protein complex conserved from yeast to humans. Functions of the STRIPAK complex are still to be determined but it is possible it plays a role in intracellular trafficking, and/or in cytoskeletal changes, since yeast MOB1 is involved in cytoskeletal reorganization during exit from mitosis and cytokinesis .
FMNL1, 2 and 3 are part of the Diaphanous-related formin family of proteins, several of which have been shown to be regulated by Rho GTPases [23, 24]. Depletion of each of the three FMNL proteins induced an increase in peripheral F-actin rich lamellipodial protrusions in PC3 cells, but only FMNL3 knockdown significantly inhibited migration in scratch wounds. FMNL1/FRLα has been shown to stimulate actin polymerization, bind to Rac1, and regulate macrophage motility [25, 26]. FMNL2 and FMNL3 have recently been reported to bind to RhoC and regulate migration and invasion [27, 28]. Our observation that FMNL3 severely impairs migration and localizes in part to lamellipodia is in accordance with these observations.
Formin family proteins have multiple binding partners, known as formin-binding proteins (FBPs or FNBPs). It is interesting that we identified that depletion of a FBP, FNBP3, strongly inhibits migration, similar to FMNL3 knockdown. Its close relative HYPC did not have a marked effect on migration or morphology of PC3 cells, which might reflect low-level expression compared to FNBP3. FNBP3 (also known as FBP11/PRPF40A) was identified in a screen for formin-1-binding proteins, which identified multiple proteins with WW domains [29, 30]. However, of these WW domain-containing FBPs, only FBP17 has been characterized in detail . FNBP3 and HYPC are orthologues of the S. cerevisiae splicing factor PRP40, but whether they have a role in splicing in mammals is not known. FNBP3 has been reported to interact with N-WASP and inhibit its translocation to the cytoplasm  and both FNBP3 and HYPC interact with the N-terminus of Huntingtin (HTT) , but little is known of the functional relevance of these interactions. Our results, showing that FNBP3 depletion strongly inhibits cell migration in PC3 cells and cytokinesis in HeLa cells, imply that FNBP3 is likely to act through a formin family protein, which is known to affect cell protrusion and cytokinesis [24, 33]. It would be interesting to know whether FNBP3 and FMNL3 act together to regulate cell migration, although only FNBP3 has an effect on cytokinesis so would be expected to act through a different formin for this process.
LIMD1 and WTIP are members of the LIM domain-containing the Ajuba/Zyxin family. Several members of the family, such as Zyxin, are involved in the regulation of cell adhesion and also translocate to the nucleus where they regulate gene expression . LIMD1 and WTIP also appear to be multifunctional proteins. LIMD1 is implicated in osteoclast development through effects on transcription factor activity , but has also been reported to localise to focal adhesions . WTIP interacts with and affects the transcriptional activity of Wilm's tumour protein, yet also localizes to cell-cell junctions and podocytes in kidney cells , and interacts with the receptor tyrosine kinase ROR2 . LIMD1 and WTIP depletion induce distinct changes in actin organization: LIMD1 knockdown leads to an increase in peripheral F-actin bundles and stress fibres (in HeLa cells), whereas WTIP knockdown induces the formation of small F-actin-rich protrusions. These effects could be mediated by nuclear or cytoplasmic functions of LIMD1 and WTIP.
Unlike the other PMM genes we analysed in detail, ARC (activity-regulated cytoskeleton-associated protein) and ZRANB1 are each unique genes without closely related isoforms in mammals. Depletion of each of them strongly inhibits migration of PC3 cells and, like LIMD1 knockdown, increased stress fibres. ARC was originally identified as a gene rapidly induced by neuronal activity , and mice lacking ARC have deficits in long-term memory [39, 40]. ARC has also been implicated in endocytosis through interactions with endophilin and dynamin . Our data indicate for the first time that it plays an important role in cancer cell migration and actin cytoskeletal organization, which could be linked to endocytosis, although it does not specifically localize to vesicular structures.
ZRANB1 (also known as TRABID) has three zinc-finger domains and a ubiquitin thioesterase domain, and has been reported to de-ubiquitinate APC and, thereby, regulate Wnt signalling . It also binds to TRAF6, which in turn associates with TNF-receptor family members . It is possible that it regulates cell migration and stress fibres through effects on APC ubiquitination, which is known to affect microtubules and the actin cytoskeleton , although it could also deubiquitinate other proteins involved in cell migration.