Hypothesis

Epigenetic noise at enhancer regions drives stochastic transcription factor (TF) binding variability, which in turn causes the observed decline in mutual information (MI) between TFs and target genes and the centralization of gene regulatory networks (GRNs) during aging.

Rationale

• et al. show that loss of GRN fidelity stems from input distribution mismatch rather than TF incompetence [1].
• Epigenetic drift (DNA methylation, H3K9me3) creates stochastic gene expression that silences DNA repair genes, forming self‑reinforcing noise cycles [4][5].
• Transcription elongation factors modulate senescence via isoform switching, linking core transcription machinery to cell‑cycle arrest [6].
• Splicing and processing alterations compensate for aberrant RNAs from degraded GRN fidelity [7][8].

We propose that the primary source of the input distribution mismatch is enhancer‑level epigenetic noise, which randomly alters chromatin accessibility and thus the probability of TF binding at specific sites. This stochastic binding reduces coordinated TF activity, increases network centralization, and erodes stabilizing feedback motifs, manifesting as the measured MI decay.

Novel Mechanistic Insight

If enhancer epigenetic noise is causal, then targeted restoration of a permissive epigenetic state at a defined set of aging‑sensitive enhancers should:

1. Increase TF binding coherence (higher ChIP‑seq signal correlation across cells).
2. Raise MI between TFs and their targets.
3. Decentralize the GRN (reduce hub‑node degree, increase motif richness).
4. Improve functional readouts (e.g., contractility in cardiomyocytes, mitochondrial respiration).

Testable Predictions

• Prediction 1: In aged human iPSC‑derived cardiomyocytes, single‑cell ATAC‑seq will show increased variance in accessibility at enhancers of sarcomere and mitochondrial genes compared with young cells.
• Prediction 2: CRISPR‑dCas9‑p300 mediated acetylation of those enhancers will reduce accessibility variance and increase average accessibility.
• Prediction 3: Consequently, scRNA‑seq will reveal higher MI between key TFs (e.g., MEF2C, GATA4, SRF) and their target genes, approaching youthful levels.
• Prediction 4: Network inference (e.g., PIDC or GENIE3) will demonstrate decreased centralization (lower betweenness centrality of hub TFs) and reinstatement of feedback loops (e.g., MEF2C‑HDAC5).
• Prediction 5: Functional assays will show improved calcium handling and ATP production.

Experimental Design

1. Cell model: Generate iPSC lines from young (20‑30 yr) and aged (>70 yr) donors; differentiate to cardiomyocytes.
2. Baseline profiling: Perform scRNA‑seq + scATAC‑seq (multiome) on n=3 biological replicates per age; compute MI, enhancer accessibility variance, and GRN metrics.
3. Intervention: Use CRISPR‑dCas9‑p300 guided by sgRNAs targeting the top 50 variable enhancers identified in step 2. Include controls: dCas9‑dead, non‑targeting sgRNA.
4. Post‑intervention profiling: Repeat multiome after 7 days; assess predictions 1‑5.
5. Validation: Perturb predicted hub TFs (siRNA) to test whether network improvements depend on restored TF coordination.

Falsifiability

If enhancer acetylation fails to reduce accessibility variance, does not increase MI, or does not alter GRN centralization/function, the hypothesis that enhancer‑level epigenetic noise drives aging‑related GRN decline would be falsified. Conversely, a positive result would support a causal role and suggest a rejuvenation strategy.

References

[1] https://pubmed.ncbi.nlm.nih.gov/41278781/ [2] https://arxiv.org/html/2601.04016v1 [3] https://pubmed.ncbi.nlm.nih.gov/41542164/ [4] https://pmc.ncbi.nlm.nih.gov/articles/PMC12402629/ [5] https://www.jci.org/articles/view/158446 [6] https://news.feinberg.northwestern.edu/2025/12/18/exploring-the-connection-between-gene-expression-and-aging/ [7] https://www.aging-us.com/article/206347/text [8] https://www.aging-us.com/news-room/rna-splicing-and-processing-emerge-as-central-features-of-human-aging-across-tissues