KDM5B as a Nutrient-Sensitive Epigenetic Switch Governing Autophagy Gene Bivalency in Siege vs. Reprogramming States

Hypothesis

Chronic nutrient siege forces cells into a rationing mode where KDM5B demethylase activity is limited by low α‑ketoglutarate, causing loss of H3K4me3 and gain of H3K27me3 at selective autophagy promoters. This creates a repressive bivalent chromatin state that silences cargo‑specific autophagy while sparing core starvation‑response genes. In contrast, oncogenic metabolic reprogramming restores α‑KG levels, reactivates KDM5B, erases H3K27me3, and drives transcription of selective autophagy programs that support tumor survival.

Mechanistic Model

KDM5B is an Fe²⁺/α‑KG‑dependent histone demethylase. When glutamine‑derived α‑KG falls—as in aging, quiescence, or microenvironmental stress—its catalytic rate drops, reducing H3K4me3 maintenance. Concurrently, the PRC2 complex deposits H3K27me3 on nucleosomes lacking H3K4me3, establishing a bivalent domain that poises genes for repression. At autophagy loci such as SQSTM1/p62, BNIP3, and ATG5, this shift preferentially attenuates selective autophagy, aligning with the “siege” concept where non‑essential components are spared from self‑cannibalism. When oncogenic signaling (e.g., KRAS^G12D, PI3K/AKT) elevates glutaminolysis or expresses mutant IDH that produces α‑KG analogues, KDM5B regains activity, removes H3K27me3, and sustains H3K4me3, thereby unlocking selective autophagy flux needed for tumor adaptation.

Predictions

1. Under glutamine deprivation, CUT&RUN will show decreased KDM5B occupancy, reduced H3K4me3, and increased H3K27me3 at selective autophagy promoters compared to nutrient‑rich conditions.
2. Expressing a catalytically dead KDM5B (H614A) will phenocopy low‑α‑KG chromatin states and suppress selective autophagy flux without affecting LC3‑II conversion under starvation.
3. Restoring α‑KG via cell‑permeable dimethyl‑α‑KG or overexpressing wild‑type KDM5B will rescue H3K4me3 levels, diminish H3K27me3, and increase selective autophagy even in low‑glutamine media.
4. In KRAS‑mutant glioblastoma cells, KDM5B inhibition (JIB‑04) will increase H3K27me3 at SQSTM1 and BNIP3 promoters and reduce selective autophagy, contrary to the pan‑autophagy increase reported with KDM5A/6B inhibition.

Experimental Design

• Cell models: Human IMR‑90 fibroblasts (senescence model), primary mouse muscle satellite cells (quiescence), and U87MG glioblastoma lines with inducible KRAS^G12D.
• Treatments: Glutamine‑free medium ± dimethyl‑α‑KG (1 mM); Torin1 (100 nM) to induce autophagy; JIB‑04 (1 µM) for KDM5 inhibition.
• Assays: CUT&RUN for KDM5B, H3K4me3, H3K27me3 at SQSTM1, BNIP3, ATG5; RT‑qPCR for transcripts; GFP‑LC3 puncta and mCherry‑GFP‑LC3 flux assay; western blot for p62 turnover.
• Controls: Non‑targeting IgG, catalytically dead KDM5B rescue, and PRC2 inhibitor (GSK126) to test H3K27me3 dependence.

Potential Outcomes and Interpretation

If low α‑KG reduces KDM5B binding and increases H3K27me3 at selective autophagy genes, supporting the siege‑rationing model, we will observe decreased p62‑dependent flux despite elevated LC3‑II, indicating a shift toward bulk, non‑selective autophagy. Conversely, rescuing α‑KG or KDM5B activity should restore selective autophagy flux. Failure to detect chromatin changes would refute the hypothesis and suggest that KDM5B acts indirectly or that other demethylases (e.g., KDM6A/B) dominate autophagy gene regulation under these conditions.