E associated with GABPA binding DNA regions. The network was clustered using the k-means algorithm GSK2126458 site provided by the STRING portal, with the number of clusters pre-set to 7 (empirically estimated as optimal). The functions of the proteins within circled clusters were determined through literature- and database mining. (TIF) Figure SExpression microarray analysisExpression array experiments were performed in triplicate and analysed as described previously [7] with the following modifications: only MCF10A cells grown in the absence of EGF for 48 hours were used, and filtering of probes with signal lower than background was not applied. One repeat was performed with an ON-TARGET SMARTpool siGABPA and two were performed with the SantaCruz duplex. Data are shown in Table S1 and are deposited with ArrayExpress (E-MEXP-3682).Chromatin immunoprecipitationChIP experiments using antibodies against ELK1 (Epitomics), GABPA (SantaCruz, sc-22810) and normal rabbit IgG (Millipore) were carried out as described previously [7].Bioinformatic and statistical analysisAll overlaps of lists of gene names were performed using an online tool available at http://jura.wi.mit.edu/bioc/tools/ compare.php. Networks of protein-protein interactions were created in STRING [13] using physical interaction, coexpression, database and literature mining as proximity criteria at medium stringency. Clustering of STRING networks was performed using an embedded k-means algorithm, with numbers of expected 1313429 clusters determined empirically. Z-score analysis and the statistical analysis of qPCR and imaging results were carried out as described previously [7].Genes negatively regulated by GABPA form several small clusters and code for stress-associated proteins. Image shows a STRING-derived network of proteins encoded by all genes which exhibit a statistically significant upregulation of expression in MCF10A cells depleted of GABPA and which are associated with GABPA binding DNA regions. The network was clustered using the k-means algorithm provided by the STRING portal, with the number of clusters pre-set to 4 (empirically estimated as optimal). The functions of the proteins within circled clusters were determined through literature- and database mining. Several subnetworks of proteins which are not discovered by STRING as clusters share partial functional associations. (TIF)Supporting InformationFigure S1 The effect of GABPA depletion on MCF10A cell phenotype is specific. (A and B) Wound healing assays were performed as in Figure 1C and D, with the use of an alternative siRNA duplex. Instead of time-lapse imaging, cells were fixed 15 hours after EGF buy GW788388 stimulation and stained with crystal violet. Shown are representative images of wounds (A) and quantification of three biological repeats of the experiment (average values with standard deviations) (B). (TIF) Figure S2 Overlaps between GABPA regulated genes and direct ELK1 targets. Table shows numbers of genes exhibiting a change of expression upon depletion of GABPA (B);Table S1 Lists of GABPA regulated genes. Summary of expression microarray data of gene expression changes in MCF10A cells following GAPBA depletion. Direct targets are inferred by comparing to GABPA occupancy as inferred from ChIP-seq analysis (see text for details). (XLS) Table S2 Oligonucleotides used for ChIP- and RTqPCR. List of all oligonucleotides used in this study. (DOCX)AcknowledgmentsWe thank Karren Palmer and Michael Smiga for excellent technical assistance; A.E associated with GABPA binding DNA regions. The network was clustered using the k-means algorithm provided by the STRING portal, with the number of clusters pre-set to 7 (empirically estimated as optimal). The functions of the proteins within circled clusters were determined through literature- and database mining. (TIF) Figure SExpression microarray analysisExpression array experiments were performed in triplicate and analysed as described previously [7] with the following modifications: only MCF10A cells grown in the absence of EGF for 48 hours were used, and filtering of probes with signal lower than background was not applied. One repeat was performed with an ON-TARGET SMARTpool siGABPA and two were performed with the SantaCruz duplex. Data are shown in Table S1 and are deposited with ArrayExpress (E-MEXP-3682).Chromatin immunoprecipitationChIP experiments using antibodies against ELK1 (Epitomics), GABPA (SantaCruz, sc-22810) and normal rabbit IgG (Millipore) were carried out as described previously [7].Bioinformatic and statistical analysisAll overlaps of lists of gene names were performed using an online tool available at http://jura.wi.mit.edu/bioc/tools/ compare.php. Networks of protein-protein interactions were created in STRING [13] using physical interaction, coexpression, database and literature mining as proximity criteria at medium stringency. Clustering of STRING networks was performed using an embedded k-means algorithm, with numbers of expected 1313429 clusters determined empirically. Z-score analysis and the statistical analysis of qPCR and imaging results were carried out as described previously [7].Genes negatively regulated by GABPA form several small clusters and code for stress-associated proteins. Image shows a STRING-derived network of proteins encoded by all genes which exhibit a statistically significant upregulation of expression in MCF10A cells depleted of GABPA and which are associated with GABPA binding DNA regions. The network was clustered using the k-means algorithm provided by the STRING portal, with the number of clusters pre-set to 4 (empirically estimated as optimal). The functions of the proteins within circled clusters were determined through literature- and database mining. Several subnetworks of proteins which are not discovered by STRING as clusters share partial functional associations. (TIF)Supporting InformationFigure S1 The effect of GABPA depletion on MCF10A cell phenotype is specific. (A and B) Wound healing assays were performed as in Figure 1C and D, with the use of an alternative siRNA duplex. Instead of time-lapse imaging, cells were fixed 15 hours after EGF stimulation and stained with crystal violet. Shown are representative images of wounds (A) and quantification of three biological repeats of the experiment (average values with standard deviations) (B). (TIF) Figure S2 Overlaps between GABPA regulated genes and direct ELK1 targets. Table shows numbers of genes exhibiting a change of expression upon depletion of GABPA (B);Table S1 Lists of GABPA regulated genes. Summary of expression microarray data of gene expression changes in MCF10A cells following GAPBA depletion. Direct targets are inferred by comparing to GABPA occupancy as inferred from ChIP-seq analysis (see text for details). (XLS) Table S2 Oligonucleotides used for ChIP- and RTqPCR. List of all oligonucleotides used in this study. (DOCX)AcknowledgmentsWe thank Karren Palmer and Michael Smiga for excellent technical assistance; A.
