HIFα isoform specific activities drive cell-type specificity of VHL-associated oncogenesis

Authors

Joanna D. C. C. Lima

Madeleine Hooker

Ran Li

Ayslan B. Barros

Norma Masson

Christopher W. Pugh

David R. Mole

Julie Adam

Peter J. Ratcliffe

Samvid Kurlekar

Doi

10.1038/s41467-025-64214-3

PMID: 23797736 · DOI: 10.1038/s41467-025-64214-3 · Journal: Nature Communications (2025)

TL;DR

Lima et al. use an “oncogenic cell tagging” mouse model (tdTomato-linked Vhl recombination driven by Pax8-CreERT2) crossed with conditional Hif1a and/or Epas1 alleles to resolve the contribution of each HIFα isoform to the earliest events of VHL-associated clear cell renal cell carcinoma (CCRCC). Single-cell RNA-seq of 147,045 cells from VKO, VHKO, VEKO, and VHEKO mice (4 mice per genotype) shows that HIF1A and HIF2A drive distinct, largely non-overlapping transcriptional programs: HIF1A controls early glycolytic upregulation, while HIF2A drives late, “adaptive” dedifferentiation of proximal tubular (PT) cells. Both isoforms are required for the proliferative expansion of Vhl-null cortical/outer-medullary PT cells, whereas Hif1a loss specifically rescues the loss of Vhl-null cells in the renal papilla. The HIFα-isoform–specific and time-resolved programs identified in the mouse are also active in human ccRCC (TCGA), supporting early therapeutic targeting of HIF2A with belzutifan in VHL disease.

Cohort & data

  • Model system: Conditional mouse genetics — Pax8-CreERT2–driven biallelic Vhl inactivation in the renal tubular epithelium (RTE), coupled to an oncogenic cell-tagging Vhlpjr.fl allele expressing tdTomato upon recombination; tamoxifen-induced (5 × 2 mg).
  • Genotypes compared (one Vhl allele in each row paired with floxed HIF-isoform alleles):
    • VKO: Vhljae.KO/pjr.fl; Pax8-CreERT2 — biallelic Vhl loss only.
    • VHKO: VKO + Hif1afl/fl — biallelic Vhl + Hif1a deletion.
    • VEKO: VKO + Epas1fl/fl — biallelic Vhl + Epas1 (HIF2A) deletion.
    • VHEKO: VKO + Hif1afl/fl; Epas1fl/fl — Vhl + both HIFα isoforms.
    • Plus four Vhl-competent controls (ConKO, ConHKO, ConEKO, ConHEKO).
  • scRNA-seq: 147,045 high-quality cells from 7 male + 5 female mice (4 mice per genotype) collected 4–12 months after Vhl inactivation; average 12,254 ± 6438 (SD) cells/mouse, 4933 ± 2037 mapped reads/cell, 1594 ± 408 genes/cell. New data are deposited at GEO GSE282887; prior ConKO/VKO data at GSE253168.
  • Time points: Early (1–3 weeks post-recombination) and late (4–12 months) IHC + scRNA-seq.
  • Assays: Single-cell RNA-seq (FAC-sorted tdTomato+ cells), dual IHC for tdTomato and Ki67, RNAscope multiplex ISH for marker validation, automated tdTomato+ cell quantification.
  • Human validation: Aggregate expression of HIF1A/HIF2A “early” and HIF2A “adaptive” gene sets compared in ccRCC vs normal kidney TCGA data (accessed 14 April 2025), plus GSEA on three published murine ccRCC datasets (Hoefflin 2020; Harlander 2017; Nargund 2017).
  • Study slug (frontmatter): ccrcc_utokyo_2013 is retained from the orchestrator manifest; the actual primary data in this paper are mouse scRNA-seq (GSE282887 / GSE253168) with secondary analysis of kirc_tcga.

Key findings

  • Cell-type–specific consequences of Vhl loss are HIFα-dependent. In the renal papilla, Vhl-null tdTomato+ cells are lost over 4–12 months in VKO mice (p = 0.020) and VEKO mice (p = 0.006); loss is abrogated by Hif1a co-deletion (VHKO p = 0.311; VHEKO p > 0.999). Thus papillary cell elimination requires HIF1A but not HIF2A.
  • Cortex/outer medulla: opposite phenotype. Vhl-null PT cells in cortex/outer medulla expand over time in VKO (p = 0.016) but not in VEKO, VHKO, or VHEKO mice (p > 0.999 for VEKO/VHEKO; p = 0.335 for VHKO). Both HIFα isoforms are required for proliferation in this compartment.
  • Proliferation marker concordance. Dual IHC shows the proportion of tdTomato+ cells co-staining for MKI67 (Ki67) is increased in VKO vs ConKO (p = 0.049) but not in VHEKO vs ConHEKO (p > 0.999), VHKO vs ConHKO (p = 0.094) or VEKO vs ConEKO (p = 0.138).
  • scRNA-seq clustering. Across all six PT identities (S1, S2, S3 × Class A/B), VHKO cells co-cluster with VKO (HIF1A loss has minor transcriptional impact), while VEKO cells diverge sharply from VKO (HIF2A loss has greater impact). VHEKO cells do not collapse back to ConKO clusters, indicating partial residual Vhl-loss effects.
  • Isoform-specific gene sets are largely non-overlapping. Genes regulated by HIF1A and HIF2A (EPAS1) show “almost never” cross-regulation across PT identities. GO terms are entirely non-overlapping between HIF1A- and HIF2A-specific gene sets.
  • HIF1A → glycolysis; HIF2A → dedifferentiation. HIF1A-specific upregulation is dominated by glycolytic genes. HIF2A-specific upregulation encompasses cellular structure/motility (Igfbp5, Apela), secretion/transport (Fabp5, Slc3a1) and ECM organization (Col4a1, Npnt); HIF2A-specific downregulation hits transmembrane transport and PT differentiation markers — Slc5a12 (PT-S1), Inmt (PT-S2), Cyp2a4 (PT-S3) — validated by RNAscope ISH.
  • HIF2A–PT8 differentiation axis. Loci of HIF2A-specific (but not HIF1A-specific) genes are enriched for binding sites of PT-restricted TFs HNF4A, HNF1B, and FOXA2 by both LISA and ChEA3 (top-50 by both tools). This nominates a mechanism for tissue-restricted HIF2A action via cooperation with renal lineage TFs (extending earlier observations that HIF2A cooperates with PAX8 to upregulate CCND1 cyclin D1 in ccRCC).
  • Temporal dissection. HIF1A-dependent transcriptional changes reach their full magnitude within 3 weeks of Vhl inactivation and remain stable. HIF2A-dependent changes are only partially manifest early and continue to evolve over months, accounting for the late dedifferentiation phenotype.
  • Translation to human ccRCC. Both early (HIF1A- and HIF2A-driven) and adaptive (HIF2A-driven) gene programs identified in mouse PT cells are significantly regulated in human ccRCC vs normal kidney in TCGA, and in three independent murine ccRCC genetic models (refs. 11, 43, 44; Hoefflin 2020 Nat Commun, Harlander 2017 Nat Med, Nargund 2017 Cell Rep).

Genes & alterations

  • VHL — Biallelic conditional inactivation is the central perturbation; mimics the truncal event in human ccRCC. The cell-tagging Vhlpjr.fl allele links tdTomato expression to recombination so that Vhl-null cells can be retrieved before any morphological abnormality.
  • HIF1A — Encodes HIFα isoform whose stabilization after Vhl loss drives anti-survival effects in the renal papilla and is required for early proliferation in cortical/outer-medullary PT cells. HIF1A-specific upregulated targets are dominated by glycolytic genes.
  • EPAS1 — Encodes HIF2A. Required for cortical/outer-medullary PT proliferation but dispensable in the papilla. HIF2A-specific upregulation drives ECM/motility/secretion programs; HIF2A-specific downregulation drives time-dependent dedifferentiation of PT cells. The paper notes that human ccRCC almost universally bears VHL mutation but rarely if ever bears HIF2A-activating mutations (refs 27, 47, 48), and offers a possible explanation: HIF2A activation alone cannot regulate the HIF1A-specific metabolic program required for early proliferation.
  • PAX8 — Renal lineage TF; Pax8-CreERT2 drives RTE-restricted recombination in this model (95.7% of tdTomato+ cells were PT S1/S2/S3). Prior work cited (Patel 2022, ref 22) showed HIF2A cooperates with PAX8 to upregulate cyclin D1 (CCND1) in ccRCC.
  • HNF4A, HNF1B, FOXA2 — Identified by LISA and ChEA3 (top 50 in both) as TFs whose binding sites are enriched at HIF2A-specific (but not HIF1A-specific) Vhl-dependent gene loci; all three have established roles in proximal tubule / renal development. Proposed to mediate HIF2A’s tissue-restricted, dedifferentiation-linked transcriptional output.
  • MKI67 — Used as a cell-cycle marker by dual IHC; VKO cells show increased Ki67 positivity vs ConKO (p = 0.049), an effect absent when both HIFα isoforms are co-deleted.
  • TP53, RB1 — Not perturbed in this paper but discussed: prior models combining Vhl + Trp53 ± Rb1 inactivation (refs 11, 12, 43) showed dual HIFα dependence for ccRCC tumor formation, consistent with the current single-cell findings on PT proliferation.

Clinical implications

  • Therapeutic case for early HIF2A inhibition in VHL disease. The data show that HIF2A drives the dedifferentiation and adaptive transcriptional program that becomes prominent months after Vhl inactivation but before any morphological abnormality. The authors argue this supports using clinically approved HIF2A antagonist belzutifan (Jonasch 2021 NEJM, ref 24) early in VHL disease to prevent expansion of Vhl-null PT cancer cells-of-origin, rather than only at the established-tumor stage.
  • Mechanistic basis for the absence of activating EPAS1 mutations in ccRCC. Because HIF2A alone is insufficient to drive the HIF1A-controlled glycolytic program needed for early proliferative advantage, somatic EPAS1 gain-of-function mutations would not confer the same fitness benefit as biallelic VHL loss — offering a hypothesis for why VHL inactivation, not EPAS1 activation, is the truncal event in ccRCC.
  • Cell-of-origin model. Loss of Vhl-null cells in the renal papilla via HIF1A-driven mechanisms is consistent with the clinical observation that VHL-associated cancers do not arise from this region; proliferation is restricted to cortical/outer-medullary PT cells, matching the accepted PT origin of ccRCC.

Limitations & open questions

  • Mouse model, not human disease. No clinically detectable tumors form in this model within the observation window — phenotypes are early, pre-morphological proliferation/dedifferentiation. Translation to human ccRCC initiation is supported only indirectly via TCGA cross-validation.
  • Incomplete HIF co-deletion. The authors note that not every tdTomato+ cell harbored complete Hif1a/Epas1 recombination, which may underestimate the magnitude of HIFα-dependent effects in VHEKO mice (Supplementary Fig. 1c, 5a).
  • VHEKO cells do not return fully to ConKO transcriptome, raising the open question of whether residual HIFα activity, basal HIFα activity in controls, or genuinely HIFα-independent Vhl-dependent effects explain this gap.
  • Mechanism of delayed HIF2A action unresolved. Why HIF2A-dependent transcriptional changes take months to manifest is unclear; the authors speculate that new HIF2A targets become accessible in chromatinized DNA as PT cells dedifferentiate.
  • Causality of dedifferentiation for oncogenesis. The HIF2A-driven adaptive programs are associated with the proximal tubule’s tissue-specific susceptibility to ccRCC but the work does not formally test whether dedifferentiation is required for malignant transformation.
  • Single-allele HIF studies in cysts/tumors. Findings differ from earlier mouse work in which Hif1a co-deletion failed to block Vhl-driven cyst formation (Rankin 2006, ref 26) or where HIF1A/HIF2A overexpression alone produced lesions (refs 8, 52). Differences in tissue context, allele design, and timing remain unresolved.

Citations from this paper used in the wiki

  • “Biallelic mutation of the VHL ubiquitin ligase leads to constitutive activation of hypoxia inducible factors HIF1A and HIF2A and is generally a truncal event in clear cell renal carcinoma.” (abstract / introduction)
  • “Inducing Vhl inactivation in papillary RTE cells resulted in their loss from the renal papilla… This cell loss was rescued completely and specifically by Hif1a co-deletion, marking HIF1A as potentially ‘anti-tumorigenic’ and HIF2A as agnostic to tumorigenesis in cells in this region of the kidney.” (Discussion)
  • “The proportion of tdTomato-positive cells that were also Ki67-positive was increased in VKO versus ConKO cells (p=0.049). This difference was absent (p>0.999) when comparing VHEKO to ConHEKO cells.” (Results, Fig 2d)
  • “HIF1A-regulation was clearly linked to early upregulation of glycolytic genes, HIF2A-mediated gene regulation encompassed a range of pathways including lipid and amino acid metabolism, extracellular matrix reorganization, and dedifferentiation.” (Discussion)
  • “The substantial dependency on HIF2A of gene expression and proliferation of Vhl-null cells in the proximal tubule provides support for the early use of HIF2A antagonism (e.g., with the clinically approved inhibitor Belzutifan) to prevent the expansion of potential cancer cells-of-origin early in VHL disease.” (Discussion)
  • “HIF2A activating mutations are not observed in ccRCC… suggesting a more complex relationship between VHL inactivation and oncogenic HIFα activation during cancer development.” (Discussion)
  • “scRNA-seq data… available in the Gene Expression Omnibus (GEO) database under accession code GSE253168… [and] GSE282887.” (Data availability)

This page was processed by crosslinker on 2026-05-09.