KDM5as a redox-sensitive rheostat governing bivalent domain fate in aged hematopoietic stem cells

Chloe Vargas (PhD)· Mar 23, 2026

Mechanism: KDM5 acts as a bifunctional switch in aged HSCs, with low levels promoting R-loop accumulation and cGAS/STING activation under hypoxia, while high levels suppress R-loops but risk tumor suppressor silencing. Readout: Readout: Low KDM5 conditions show elevated R-loops, high SASP, increased senescence, and low leukemia risk, whereas high KDM5 reduces R-loops and senescence but increases leukemic potential.

Hypothesis

In aged hematopoietic stem cells (HSCs) the coordinate gain of H3K4me3 and H3K27me3 creates bivalent domains that are uniquely vulnerable to KDM5 activity. We propose that KDM5 acts as a bifunctional switch: its demethylase activity resolves bivalency by removing H3K4me3, while its R-loop suppression activity prevents immunostimulatory nucleic acid accumulation. Low KDM5 levels in aged HSCs allow R-loops to form at these domains, triggering cGAS/STING-driven inflammation and senescence; high KDM5 levels erase protective H3K4me3 marks at tumor suppressor loci, facilitating oncogenic transformation. This switch is modulated by cellular oxygen tension via KDM6A, which senses O2 and influences KDM5 recruitment to chromatin.

Mechanistic Model

Bivalent domain substrate: Aged HSCs show ~4‑fold increase in bivalent domains through concurrent H3K4me3 and H3K27me3 gains (PMC7534803). These domains harbor low‑level transcription of both active and repressed genes, increasing the chance of transcription‑replication collisions.
KDM5 demethylase function: By removing H3K4me3, KDM5 reduces RNA Polymerase II pausing and nascent RNA output, decreasing the substrate for R-loop formation. Simultaneously, KDM5 binds RNA:DNA hybrids and recruits RNase H2, directly suppressing R-loops (elifesciences.org/articles/106249).
R-loop consequence: When KDM5 activity falls, unresolved R-loops accumulate at bivalent loci, activate the cGAS/STING pathway, and drive SASP‑mediated senescence—a tumor‑suppressive barrier.
Oncogenic consequence: When KDM5 is overexpressed, H3K4me3 loss at bivalent promoters of tumor suppressors (e.g., p16INK4A, p14ARF) diminishes their expression, while retained H3K27me3 maintains a repressed state that is resistant to reactivation, permitting leukemic transformation (PMC12465163).
Oxygen sensing link: KDM6A functions as an O2 sensor, modulating H3K27me3 levels (science.org/doi/10.1126/science.aaw1026). Under hypoxia, KDM6A activity drops, leading to H3K27me3 accumulation at bivalent domains, which may enhance KDM5 binding (via increased chromatin affinity) and shift the balance toward demethylation and R-loop suppression.

Experimental Design

Model system: Use aged (18‑month) Vav‑CreERT2;Rosa26‑LSL‑KDM5B overexpression and KDM5B knockout mice; include young (3‑month) controls.
Oxygen manipulation: House cohorts in normoxia (21% O2) or chronic hypoxia (8% O2) for 4 weeks to test KDM6A‑dependent effects.
Readouts:
- ChIP‑seq for H3K4me3, H3K27me3, and KDM5B occupancy.
- DRIP‑seq (S9.6 antibody) to quantify R-loops genome‑wide, focusing on bivalent domains.
- RNA‑seq to assess expression of tumor suppressors, SASP genes, and replication stress markers.
- Phospho‑TBK1/IRF3 staining and IFN‑β ELISA for cGAS/STING activation.
- Senescence biomarkers (SA‑β‑gal, p16INK4A protein) and colony‑forming unit assays for leukemogenic potential.
Rescue experiments: Treat aged HSC cultures with RNase H2 overexpression or STING inhibitor (C‑176) to dissect R-loop versus inflammation contributions.

Expected Outcomes

Low KDM5 / hypoxia: Increased H3K27me3, elevated R‑loops at bivalent loci, heightened cGAS/STING signaling, and robust senescence without increased colony formation.
High KDM5 / normoxia: Reduced H3K4me3 at tumor suppressor promoters, decreased R‑loop accumulation, low SASP, and enhanced clonogenic/leukemic potential.
KDM5B KO: Phenocopy of low KDM5 condition, confirming demethylase dependence.
RNase H2 rescue: Suppression of cGAS/STING activation despite low KDM5, indicating R‑loops as the upstream trigger.

Potential Pitfalls & Alternatives

Compensation by other KDMs: Simultaneous loss of KDM5B may upregulate KDM5A or KDM6 family members; we will quantify their expression and include double‑knockout controls if needed.
Antibody specificity for DRIP‑seq: Validate with RNase H treatment controls and S9.6 dot‑blot assays.
Systemic effects of hypoxia: Use ex‑vivo HSC cultures under controlled O2 to isolate cell‑intrinsic effects.

This hypothesis directly tests whether the dual enzymatic and nucleic‑acid‑handling activities of KDM5 create a fate decision in aged HSCs that is tuned by oxygen sensing, offering a mechanistic bridge between chromatin aging, inflammaging, and oncogenic transformation.

Hypothesis

Mechanistic Model

Bivalent domain substrate: Aged HSCs show ~4‑fold increase in bivalent domains through concurrent H3K4me3 and H3K27me3 gains (PMC7534803). These domains harbor low‑level transcription of both active and repressed genes, increasing the chance of transcription‑replication collisions.

KDM5 demethylase function: By removing H3K4me3, KDM5 reduces RNA Polymerase II pausing and nascent RNA output, decreasing the substrate for R-loop formation. Simultaneously, KDM5 binds RNA:DNA hybrids and recruits RNase H2, directly suppressing R-loops (elifesciences.org/articles/106249).

R-loop consequence: When KDM5 activity falls, unresolved R-loops accumulate at bivalent loci, activate the cGAS/STING pathway, and drive SASP‑mediated senescence—a tumor‑suppressive barrier.

Oncogenic consequence: When KDM5 is overexpressed, H3K4me3 loss at bivalent promoters of tumor suppressors (e.g., p16INK4A, p14ARF) diminishes their expression, while retained H3K27me3 maintains a repressed state that is resistant to reactivation, permitting leukemic transformation (PMC12465163).

Oxygen sensing link: KDM6A functions as an O2 sensor, modulating H3K27me3 levels (science.org/doi/10.1126/science.aaw1026). Under hypoxia, KDM6A activity drops, leading to H3K27me3 accumulation at bivalent domains, which may enhance KDM5 binding (via increased chromatin affinity) and shift the balance toward demethylation and R-loop suppression.

Experimental Design

Model system: Use aged (18‑month) Vav‑CreERT2;Rosa26‑LSL‑KDM5B overexpression and KDM5B knockout mice; include young (3‑month) controls.

Oxygen manipulation: House cohorts in normoxia (21% O2) or chronic hypoxia (8% O2) for 4 weeks to test KDM6A‑dependent effects.

Readouts:

ChIP‑seq for H3K4me3, H3K27me3, and KDM5B occupancy.
DRIP‑seq (S9.6 antibody) to quantify R-loops genome‑wide, focusing on bivalent domains.
RNA‑seq to assess expression of tumor suppressors, SASP genes, and replication stress markers.
Phospho‑TBK1/IRF3 staining and IFN‑β ELISA for cGAS/STING activation.
Senescence biomarkers (SA‑β‑gal, p16INK4A protein) and colony‑forming unit assays for leukemogenic potential.

Rescue experiments: Treat aged HSC cultures with RNase H2 overexpression or STING inhibitor (C‑176) to dissect R-loop versus inflammation contributions.

Expected Outcomes

Low KDM5 / hypoxia: Increased H3K27me3, elevated R‑loops at bivalent loci, heightened cGAS/STING signaling, and robust senescence without increased colony formation.

High KDM5 / normoxia: Reduced H3K4me3 at tumor suppressor promoters, decreased R‑loop accumulation, low SASP, and enhanced clonogenic/leukemic potential.

KDM5B KO: Phenocopy of low KDM5 condition, confirming demethylase dependence.

RNase H2 rescue: Suppression of cGAS/STING activation despite low KDM5, indicating R‑loops as the upstream trigger.

Potential Pitfalls & Alternatives

Compensation by other KDMs: Simultaneous loss of KDM5B may upregulate KDM5A or KDM6 family members; we will quantify their expression and include double‑knockout controls if needed.

Antibody specificity for DRIP‑seq: Validate with RNase H treatment controls and S9.6 dot‑blot assays.

Systemic effects of hypoxia: Use ex‑vivo HSC cultures under controlled O2 to isolate cell‑intrinsic effects.