Figure 2

Validation of TFIIB target promoters. (a) Signal distribution of TFIIB enrichment. I, II and III denote groups of promoters from the upper 10th, 60th to 80th, or lower 10th percentile and correspond to high, mid-to-low, or no TFIIB occupancy. (b) Target gene validation. Selected genes from groups I to III were analyzed by ChIP-qPCR using TFIIB or IgG control antibody in a third chromatin sample from LCL721 cells in which two independent ChIP reactions were performed (upper panel). Genes are ordered from left to right according to TFIIB levels on the promoter arrays. The relative ChIP recovery is expressed as percentage of input (y-axis). The bars represent the mean, error bars the range of the two ChIP experiments. Corresponding gene expression levels in LCL721 B cells are shown in the lower panel. These were determined using Affymetrix U133 Plus 2.0 microarrays. They represent normalized hybridization signals of gene-specific microarray probesets. N.A., not analyzed. (c) Assembly of an active PIC at the GAPDH promoter. ChIP-qPCR was conducted with eight primer pairs spanning the human GAPDH locus (numbered boxes in top scheme). Results of ChIPs in LCL721 B cells with antibodies for TFIIB, TBP or the initiating form of RNAPII (CTD S5-P) are graphed as relative occupancy levels at the different amplicon locations (lower panel). (d) Scatter plot showing the genome-wide correlation of TFIIB binding to promoters (x-axis, log₂ scale) and steady-state mRNA levels (y-axis, log₂ scale) of the corresponding genes. The median of all expression array probesets with present calls is indicated by the dotted horizontal line. The red dots indicate the average expression in gene groups with increasing TFIIB occupancy. They were determined by moving a sliding window (step size 0.1) over the TFIIB data points and calculating the mean expression value for each increment. (e) Distribution of ranked gene expression quantiles (color-coding indicated to the right) in genes with increasing TFIIB occupancy levels. The difference in distributions was statistically evaluated using a Kolmogorov-Smirnov test (***P < 2e-16).