Hypothesis

Recent work shows that aging degrades gene regulatory networks (GRNs) primarily through epigenetic noise—stochastic fluctuations in DNA methylation and histone marks—leading to a loss of mutual information (MI) between transcription factors (TFs) and their targets [1]. Single TF knock‑ins of hubs such as Ppara, Esrra, and Ppargc1b recover only ~10% of MI, suggesting that restoring TF abundance alone is insufficient [1]. We propose that epigenetic noise does not merely alter TF binding affinity; it induces isoform‑specific switching of TFs through dysregulation of the transcription elongation factors NELF and SPT6, which in turn modulate promoter‑proximal pausing and alternative splicing of TF pre‑mRNAs. This creates a feed‑forward loop: altered TF isoforms exhibit suboptimal DNA‑binding kinetics or co‑factor recruitment, worsening the “input mismatch” (TF activity falling outside target gene activation ranges) and further increasing epigenetic noise via aberrant recruitment of chromatin remodelers such as NuRD.

Mechanistic Rationale

1. Epigenetic noise → altered NELF/SPT6 occupancy – Stochastic gains/losses of methylation at promoter‑proximal regions can change the recruitment or stability of NELF and SPT6 complexes [4], leading to variable pausing indices across TF genes.
2. Pausing‑dependent splicing regulation – NELF and SPT6 are known to influence co‑transcriptional splicing by modulating RNA polymerase II kinetics; increased pausing favors inclusion of regulatory microexons, while reduced pausing promotes exon skipping. In aging, bidirectional methylation fluctuations create a heterogeneous pausing landscape that skews the isoform repertoire of TFs toward variants with altered DNA‑binding domains or transactivation regions.
3. Isoform‑specific functional impact – Certain TF isoforms (e.g., a Ppara isoform lacking a zinc‑finger motif) exhibit lower binding affinity for canonical PPAR response elements but gain affinity for cryptic sites, effectively shifting the TF’s activity distribution. This directly produces the observed “input mismatch” where the TF concentration no longer aligns with the optimal activation window of its target genes.
4. Feedback to chromatin state – Misfolded or aberrant TF isoforms recruit NuRD or other remodeling complexes inefficiently, exacerbating local epigenetic noise and further destabilizing NELF/SPT6 positioning, thus completing the loop.

Testable Predictions

• Prediction 1: In aged mouse muscle, single‑cell multi‑omics (scRNA‑seq + scATAC‑seq + scBS‑seq) will reveal a negative correlation between promoter methylation variability at TF loci and the ratio of canonical to alternative TF isoforms (e.g., Ppara‑FL/Ppara‑ΔExon4). Young tissues will show a high canonical isoform ratio and low methylation noise; aged tissues will show the inverse.
• Prediction 2: Pharmacological inhibition of NELF (e.g., using CDK9 inhibitors that reduce pausing) or overexpression of SPT6 in aged mice will normalize TF isoform ratios and rescue >30% of lost MI in GRNs—significantly exceeding the ~10% rescue seen with single TF knock‑ins.
• Prediction 3: CRISPR‑based exon‑skipping of the alternative isoform (e.g., forcing expression of the canonical Ppara isoform) in aged muscle will restore target gene activation curves without altering total TF protein levels, demonstrating that isoform identity, not quantity, drives the input mismatch.
• Prediction 4: Loss of NuRD (via CHD4 knockdown) will amplify the effect of NELF/SPT6 dysregulation on isoform switching, whereas NuRD overexpression will attenuate it, confirming the feedback arm of the hypothesis.

Falsifiability

If single‑cell longitudinal tracking shows that TF isoform ratios remain stable despite increasing promoter methylation noise, or if manipulating NELF/SPT6 fails to shift isoform distribution or improve MI, the hypothesis would be refuted. Conversely, observing the predicted isoform shifts and MI rescue would support the model that epigenetic noise drives GRN decay indirectly through transcription‑elongation‑mediated TF isoform switching, offering a mechanistic explanation for the limited efficacy of single TF interventions and pointing to combinatorial strategies (TF isoform correction + noise suppression) as a path toward full network rejuvenation.