For EGF for 48 hours, stimulated with media containing 20 ng/ml EGF and imaged for 24 hours. (G) Box plots show the distributions of lengths of trajectories travelled by MCF10A cells transfected with the indicated siRNA species between t = 1 h and t = 7 h of imaging. Data was obtained in four biological repeats of the experiment, and in each case ten cells were manually tracked. The green and pale yellow areas correspond to the second and third quartile of the distribution, respectively. The shaded area represents the distribution of distances covered in control 301353-96-8 manufacturer siGAPDH-transfected cells. P-values were obtained in a Smirnov-Kolomogorov test (*P,0.05). (H) Distribution of the trajectories travelled by cells plotted in (G). doi:10.1371/journal.pone.0049892.gseen between GABPA targets within this network (Fig. S3) thereby providing a molecular rationale for how GABPA might control the formation of the actin cytoskeleton and cell migration. Such interconnectivities are not necessarily expected, as individual GO term categories do not refer to genes which form known pathways or complexes but rather represent genes which impact on a broader common biological or molecular process, and each gene might do so in an independent manner. We also looked more broadly at the entire set of genes KDM5A-IN-1 deregulated by GABPA depletion and, strikingly, amongst genes positively regulated by GABPA, several subnetworks can be identified, one of which relates to known GABPA functions in controlling the cell cycle [9,14]. Additional subnetworks point to a role for GABPA in controlling different aspects of gene expression and also cytoskeletal 1480666 activities (Figs. S4). In contrast, fewer subnetworks were detectable amongst the genes negatively regulated by GABPA, with the most prominent one being associated mainly with transcriptional regulation (Figs. S5). To concentrate on the role of GABPA as a direct regulator of genes associated with the cytoskeleton, cell migration and adhesion, 1676428 we further probed the interconnectivities amongst target genes that are bound and regulated by GABPA and that are annotated with the relevant GO terms. We found that the majority of these genes also formed an interconnected network (Fig. 3A). Four genes from this network, RAC2, RHOF, RACGAP1 and KIF20A, were taken for further analysis due to their multiple interactions, and likely functional importance as nodes within the network. First we validated the microarray data for these targets by performing quantitative RT-PCR on MCF10A cells depleted of GABPA (Fig. 3B). Three of the four selected genes (RAC2, RACGAP1 and KIF20A) exhibited significant reductions in expression upon depletion of GABPA, while no statistically significant changes were seen on two control genes or RHOF, suggesting that the latter is probably a false positive. Similarly, we were able to detect specific binding of GABPA to the regulatory regions of RAC2, RACGAP1 and KIF20A in MCF10A cells but no binding to the RHOF regulatory region could be detected, reaffirming this as a likely false positive. Importantly, these results confirmed that RAC2, RACGAP1 and KIF20A are direct targets for GABPA in MCF10A cells as predicted from ChIP-seq data. Gene expression data showed that at least two of these genes, RAC2 and RACGAP1 are not regulated by ELK1, whereas RHOF and KIF20A require ELK1 for maximal activity (Fig. 2D). Previous ChIP-seq studies did not identify ELK1 occupancy at any of these genes [7] but we wished to confirm this.For EGF for 48 hours, stimulated with media containing 20 ng/ml EGF and imaged for 24 hours. (G) Box plots show the distributions of lengths of trajectories travelled by MCF10A cells transfected with the indicated siRNA species between t = 1 h and t = 7 h of imaging. Data was obtained in four biological repeats of the experiment, and in each case ten cells were manually tracked. The green and pale yellow areas correspond to the second and third quartile of the distribution, respectively. The shaded area represents the distribution of distances covered in control siGAPDH-transfected cells. P-values were obtained in a Smirnov-Kolomogorov test (*P,0.05). (H) Distribution of the trajectories travelled by cells plotted in (G). doi:10.1371/journal.pone.0049892.gseen between GABPA targets within this network (Fig. S3) thereby providing a molecular rationale for how GABPA might control the formation of the actin cytoskeleton and cell migration. Such interconnectivities are not necessarily expected, as individual GO term categories do not refer to genes which form known pathways or complexes but rather represent genes which impact on a broader common biological or molecular process, and each gene might do so in an independent manner. We also looked more broadly at the entire set of genes deregulated by GABPA depletion and, strikingly, amongst genes positively regulated by GABPA, several subnetworks can be identified, one of which relates to known GABPA functions in controlling the cell cycle [9,14]. Additional subnetworks point to a role for GABPA in controlling different aspects of gene expression and also cytoskeletal 1480666 activities (Figs. S4). In contrast, fewer subnetworks were detectable amongst the genes negatively regulated by GABPA, with the most prominent one being associated mainly with transcriptional regulation (Figs. S5). To concentrate on the role of GABPA as a direct regulator of genes associated with the cytoskeleton, cell migration and adhesion, 1676428 we further probed the interconnectivities amongst target genes that are bound and regulated by GABPA and that are annotated with the relevant GO terms. We found that the majority of these genes also formed an interconnected network (Fig. 3A). Four genes from this network, RAC2, RHOF, RACGAP1 and KIF20A, were taken for further analysis due to their multiple interactions, and likely functional importance as nodes within the network. First we validated the microarray data for these targets by performing quantitative RT-PCR on MCF10A cells depleted of GABPA (Fig. 3B). Three of the four selected genes (RAC2, RACGAP1 and KIF20A) exhibited significant reductions in expression upon depletion of GABPA, while no statistically significant changes were seen on two control genes or RHOF, suggesting that the latter is probably a false positive. Similarly, we were able to detect specific binding of GABPA to the regulatory regions of RAC2, RACGAP1 and KIF20A in MCF10A cells but no binding to the RHOF regulatory region could be detected, reaffirming this as a likely false positive. Importantly, these results confirmed that RAC2, RACGAP1 and KIF20A are direct targets for GABPA in MCF10A cells as predicted from ChIP-seq data. Gene expression data showed that at least two of these genes, RAC2 and RACGAP1 are not regulated by ELK1, whereas RHOF and KIF20A require ELK1 for maximal activity (Fig. 2D). Previous ChIP-seq studies did not identify ELK1 occupancy at any of these genes [7] but we wished to confirm this.