The hypoxia response pathway promotes PEP carboxykinase and gluconeogenesis in C. elegans

Abstract

Actively dividing cells, including some cancers, rely on aerobic glycolysis rather than oxidative phosphorylation to generate energy, a phenomenon termed the Warburg effect. Constitutive activation of the Hypoxia Inducible Factor (HIF-1), a transcription factor known for mediating an adaptive response to oxygen deprivation (hypoxia), is a hallmark of the Warburg effect. HIF-1 is thought to promote glycolysis and suppress oxidative phosphorylation. Here, we instead show that HIF-1 can promote gluconeogenesis. Using a multiomics approach, we reveal the genomic, transcriptomic, and metabolomic landscapes regulated by constitutively active HIF-1 in C. elegans. We use RNA-seq and ChIP-seq under aerobic conditions to analyze mutants lacking EGL-9, a key negative regulator of HIF-1. We integrate these approaches to identify over two hundred genes directly and functionally upregulated by HIF-1, including the PEP carboxykinase PCK-1, a rate-limiting mediator of gluconeogenesis. This activation of PCK-1 by HIF-1 promotes survival in response to both oxidative and hypoxic stress. Our work identifies functional direct targets of HIF-1 in vivo, comprehensively describing the metabolome induced by HIF-1 activation in an organism.

Fig. 1. Identifying genes directly regulated by HIF-1.

a Diagram of the hypoxia response pathway in C. elegans. HIF-1 is hydroxylated by EGL-9, ubiquitinated by VHL-1, and degraded by the proteasome. In egl-9(sa307) mutants, HIF-1 remains stable and regulates the transcription of target genes whose expression alters metabolism. b, c HIF-1::GFP fluorescence in the indicated genotypes under normoxia. Scale bar indicates 30 microns. Similar results were obtained in three independent trials. d Strategy for integrating HIF-1 ChIP-seq and RNA-seq data to identify directly regulated targets. e Volcano plot of RNA-seq FDR values versus log₂ fold-change expression for individual genes (blue squares) in egl-9 mutants relative to wild type. The direct targets identified by BETA are indicated with magenta diamonds. f Consensus HRE sequences identified by MEME-suite in humans and enriched in the C. elegans ChIP-seq sequences identified by BETA. g Graph of log₂ fold change in expression for individual genes in egl-9 mutants relative to wild type versus the number of HIF-1 binding sites for those genes identified by ChIP-seq. Target genes mentioned in the text are circled and labeled magenta. N = 216 total sites. Error bars indicate mean ± SEM.

Fig. 2. Characterizing HIF-1 binding sites.

a Diagram of the pck-1 locus and the different promoters used to drive Venus expression. The oval indicates HIF-1, and the black line beneath it indicates the HIF-1 binding site identified by ChIP-seq. The inverted triangle indicates the site of the HRE motif. Arrows indicate the TSS. Boxes indicate exons. The yellow arrow indicates sequences encoding the fluorescent Venus reporter. The black bar under the gene indicates sequences removed in the pck-1(ok2098) deletion. b–g Fluorescence images of animals expressing the transgenes indicated above the images and with the indicated genotype (wild-type, top row; egl-9 mutant, bottom row) under normoxic conditions. Venus fluorescence is indicated in yellow. Fluorescence from mCherry, used as an internal control in which fluorescence is not expected to change as it is not regulated by HIF-1, is magenta. Scale bar indicates 100 microns. Similar results were obtained in three independent trials. h Graph of Venus/mCherry fluorescence ratios for the indicated genotype under normoxic conditions. Dot color indicates a specific reporter as per panel (a) Error bars indicate mean ± SEM. ***P < 0.001, *P < 0.05 ANOVA/Bonferroni Multiple Comparison two-sided test for the indicated comparisons. N = 267 total animals (25–51/column). i Violin plot of histogram counting the number of HIF-1-binding sites (ChIP-seq peaks) against the number of other transcription factors (TFs) known to bind to each site’s region of the genome (within 400 bps). The top violin plot indicates all HIF-1 binding sites, whereas the middle (green) and bottom (red) violin plots indicate only the functional HIF-1 binding sites identified by BETA alone or the combined LOA and BETA approaches, respectively. The solid line indicates median, whereas dotted lines indicate quartiles. HOT sites (peaks near the binding site of 15 or more other transcription factors) are indicated by the blue dotted line. ****P < 0.0001, *P < 0.05 ANOVA/Kruskal–Wallis Multiple Comparison two-sided test for the indicated comparisons. j Graph of the fold change in enrichment (log2) of each indicated transcription factor bound within 400 bps of a HIF-1 binding site (ChIP-seq peak center). TFs with overrepresented or underrepresented binding near HIF-1 sites compared to chance (P < 0.0001 via a one-sided Bootstrap test adjusted for multiple comparisons via Benjamini–Hochberg using FDR values based on 1 × 10⁵ bootstrapped datasets) are highlighted in yellow or cyan, respectively. TF data are from the modERN/ModENCODE consortium.

Fig. 3. HIF-1 reprograms metabolism.

a Overview of major metabolic pathways impacted by HIF-1 activation. Yellow and cyan arrows overlaying the pathways indicate upregulation or downregulation, respectively, in egl-9 mutants relative to wild type. b–i Combined heatmap for metabolites (red text) and the enzymes (black text) and directional pathways (arrows) known to regulate their metabolism. The color index in (b) is applicable to all panels in the figure. On one side of each pathway arrow (above for most arrows except gluconeogenesis), differentially colored boxes indicate enzyme mRNA expression levels (log₂ fold change) in egl-9 mutants relative to wild type; enzymes are labeled in the black text by their indicated C. elegans gene name. On the other side of each pathway arrow (below for most arrows except gluconeogenesis), differentially colored boxes indicate metabolites levels (log₂ fold change) in egl-9 mutants relative to wild-type; metabolites are labeled in red text. Gene expression changes that exceed a log₂ fold change of 1.5 are marked with an asterisk showing the actual log₂ fold-change value. Yellow arrowheads indicate genes with HIF-1 binding sites and that show direct regulation by HIF-1. Glycolysis and gluconeogenesis are graphed together, in opposite direction, as they share multiple metabolites and enzymes.

Fig. 4. HIF-1 targets required for adaptive survival.

a Heatmap of row Z-scores for qRT-PCR measurements from L4-stage animals of transcript levels for the indicated genes (rows) as normalized to actin. Columns indicate the genotypes (either wild type or hif-1(ia4) mutants) and experimental conditions (4 h under either normoxia or hypoxia (0.5% O₂), with four replicates from each genotype/condition shown. b, d Percent of L4-stage animals, grown on fixed E. coli, to survive liquid culture without food and under hypoxia (0.5% O_2, 48 h at 25 °C, with 24 h recovery at 20 °C). Error bars indicate mean ± SEM. c, e Percent of embryos, obtained from animals grown on fixed E. coli, to survive 24 h of hypoxia at 25 °C, with 24 h recovery at 20 °C. Animals are indicated by colored symbols, and associated figure legends list the condition in which they were grown: on plates supplemented with either metabolites or antioxidants. Error bars indicate mean ± SEM. For graphs (b–e), ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05 ANOVA/Sidak’s multiple comparison two-sided test between each supplement and the no-supplement control. ^#P < 0.0001 ANOVA/Sidak’s multiple comparison two-sided test between the indicated mutant versus wild-type, both no-supplement controls. Data for each column in (b, c) represents four biological replicates, and in (d, e) represents six biological replicates, typically 20–100 animals per replicate. f Model illustrating how HIF-1 binds near the pck-1 promoter to upregulate PCK-1 expression and drive gluconeogenesis and the production of antioxidants.

Fig. 5. Differential gene expression for human orthologs of HIF-1 direct targets across multiple independent experiments.

Heatmap of mRNA expression across 31 independent experiments (columns, including accession numbers and hypoxia treatment indicated in vertical names) involving HIF1A activation. Fold change in expression (expressed as log2) is mapped for individual human orthologs (rows) of the C. elegans HIF-1 direct targets identified in this study. Only genes showing expression in at least 15 experiments are shown. Genes with similar patterns of expression across experiments are hierarchically clustered. Clusters of genes were analyzed for GO Reactome Pathway enrichment. While a cluster of genes (top of map) that were enriched for glycolysis and gluconeogenesis GO and reactome annotations showed near-universal upregulation across all the experiments, most putative orthologs clustered based on shared context- or tissue-dependent expression, including PCK2 (red arrowhead), an ortholog of pck-1. These latter clusters were enriched for sulfur and glutathione metabolism, metabolite transport, and chromatin modification.