Figure 1.
Basal gene expression is not sufficient for HIF transactivation.
(A) A list of well-characterized HIF target genes (from ref. 7) present in individual gene expression profiling (microarray) datasets (see B for GEO IDs) were categorized according to their basal (normoxic) expression level into two groups: Null, no detectable basal expression; Significant, detectable basal expression. In addition, for each microarray experiment, HIF-target genes were further classified into those whose expression was significantly induced by hypoxia (ratio hypoxia/normoxia greater than 2.6Sd above the mean) and non-responsive genes. The graph represents the percentage of HIF target genes in each category that were induced by hypoxia. Each pair of joined dots represents the data from a single microarray experiment. Box and whisker plots represent the distribution of values in each category. **: p<0.01 (Wilkoxon matched test) (B) For each of the indicated microarray datasets (GEO identifiers in horizontal axis), we identified all the genes showing a significant basal (normoxic) expression. Then, we classified them according to their response to hypoxia (“Induced” and “notInduced”, see A) and the presence of conserved RCGTG motifs in their regulatory regions (“HBS” and “NoHBS”, respectively). The graph depicts cumulative percentages (vertical axis) of genes in each of the four combinations of the two categories: no conserved HIF binding motifs and no hypoxic induction (blue, NoHBS_notInduced), no conserved HIF binding motifs but hypoxic induction (green, NoHBS_Induced), conserved HIF binding motifs but no hypoxic induction (yellow, HBS_notInduced) and conserved HIF binding sites and hypoxic induction (red, HBS_Induced).
Figure 2.
Comparative analysis of HIF1A ChIP-chip data in cell lines of different tissue origin.
(A) Visualization of HIF1A ChIP-chip readouts for the GAPDH promoter region in HeLa cells. The plot represents normalized intensities (log fold change hypoxia/normoxia, vertical axis) along the hg19 genomic coordinate (horizontal axis). Vertical black bars (top of the graph) mark the center position of array probes. The signal of four independent biological replicates is indicated in different colors (BR1 to 4). The smoothed black line corresponds to the averaged signal across replicates. The horizontal blue line indicates the intensity threshold for bound probes. (B) Volcano plot of HeLa ChIP-chip data (linear model across the four biological replicates). Spots in the plot correspond to individual probes in the array. Probes significantly enriched in hypoxic samples (p<0.02, FDR) are highlighted as red spots. (C) Quantitative PCR validation of HeLa ChIP-chip candidates. The official Gene Symbol corresponding to each region is indicated in the horizontal axis. Four control regions (one positive and three negative) were used as reference to estimate the successful validation of the indicated ChIP-chip candidates (candidate ranks in parenthesis). Bars represent the average fold enrichment in hypoxic versus normoxic ChIP samples (vertical axis), as obtained in three independent experiments. Error bars represent the standard deviation. The horizontal black line indicates the threshold for successful validation (90% confidence interval). (D) 4-way Venn diagram indicating the overlap of HIF1A bound regions as reported by ChIP-chip studies in hepatoma (HepG2 cells, ref. 9), mammary gland (MCF-7 cells, ref. 10), glioma (U87 cells, ref. 11) and cervix (HeLa cells, this study) origin.
Figure 3.
Integrative strategy for prediction of cooperativity in HIF binding regions.
(A) HIF1 binding locations common to at least two out of four different ChIP-chip studies in HeLa, HepG2, MCF-7 and U87 cells (left), mammalian sequence conservation of the HIF binding regions (center) and regions close to genes robustly induced in hypoxia in a meta-analysis of 16 gene expression experiments (right) were integrated into a set of bona-fide core HIF binding regions (B) Stepwise diagram for prediction and validation of TFBSs enriched in core HIF binding regions: collection of a set of core HIF binding regions and a background set of control sequences (left), in silico prediction of transcription factor binding sites present or absent in the sequences of core and background sets (center), statistical analyses of enriched TFBSs in sequences from the core set (right, top) and experimental validation of these predictions (right, bottom).
Table 1.
Enriched TFBSs in core HIF binding regions (Fisher’s exact test).
Table 2.
Enriched TFBSs in core HIF binding regions (variable selection).
Figure 4.
Effect of stress-responsive transcription factor binding sites in the proximity of hypoxia-response elements on hypoxic induction of HIF-responsive promoters.
HeLa cells were transfected with reporter plasmids containing promoter regions of mouse LDHA (A), human GYS1 (B) and human CA9 (C) in their wild-type form or harbouring the indicated mutations. Diagrams to the left of each graph indicate the location of the different mutations in the employed promoter constructs (grey blocks, highlighted with red border). Effects on reporter induction by hypoxia and the hypoxia mimetic DMOG are summarized in the central columns: -, no hypoxic/DMOG induction; +, ++, +++: increasing strength of hypoxic/DMOG induction. Graphs represent the fold induction over normoxia of the wild-type promoter (WT) upon hypoxia or DMOG treatment, compared to that of promoter versions harbouring mutations in the hypoxia response element (HRE), in CREB (A), CEBPB (B), or AP1 (C) binding sites proximal to the HRE, or in control genomic regions (CONTROL). Bars represent average values in four to six independent experiments, and error bars the standard deviation. Statistical significance of observed activity compared to the wild-type promoters are indicated: n.s.: not significant, *: p<0.05, **: p<0.01 (repeated measures ANOVA with Dunnet post-hoc correction).
Figure 5.
Transcriptional targets of stress-responsive transcription factors are enriched among core hypoxia-responsive genes.
(A) Gene-set enrichment analysis on a set of 11673 genes sorted by their response to hypoxia according to a meta-analysis of hypoxia gene expression experiments (ref. 13). The graph depicts the normalized enrichment score of 3174 gene sets from the curated collection (C2) of the GSEA molecular signatures database v3.0, that includes sets of transcription factor target genes. Solid circles highlight gene-sets derived from studies on HIF1 (black, ELVIDGE_HYPOXIA_UP and SEMENZA_HIF1_TARGETS), CEBPA/B (purple, GERY_CEBP_TARGETS, HALMOS_CEBPA_TARGETS_UP and TAVOR_CEBPA_TARGETS_UP), CREB1/ATF5 (orange, GHO_ATF5_TARGETS_DN and MCCLUNG_CREB1_TARGETS_UP) and AP1 (green, OZANNE_AP1_TARGETS_UP) transcriptional targets. The vertical blue line corresponds to an FDR-adjusted p-value of 0.05. (B) GSEA analysis of hypoxia-responsive genes (see A) against the GERY_CEBPA_TARGETS (M12338, derived from the GEO dataset GSE2188) gene-set. Hypoxic response is rank-ordered in the horizontal axis (Rank in ordered dataset). Black bars indicate the position of individual targets in the CEBPA gene-set. The graph on top (green curve) represents enrichment scores of CEBPA targets across hypoxia responsive genes, indicating positive correlation between the two. The gradient color bar indicates positive (red) and negative (blue) correlation boundaries.