Hypothesis

Rapamycin creates a non‑physiological signaling state (mTORC1-low/mTORC2-high) that simultaneously activates catabolic TFEB/FOXO programs and sustains mTORC2‑AKT survival signaling. This hybrid condition remodels the epigenetic landscape at lineage‑specific enhancers, increasing transcriptional noise without fully engaging the canonical starvation program. The resulting rise in enhancer‑level stochasticity erodes the coherent gene‑regulatory networks that underlie cellular specialization, effectively convincing cells to abandon their "civilizational" commitments while preserving core survival circuitry.

Mechanistic Rationale

1. mTORC2‑AKT phosphorylates epigenetic modifiers – AKT, hyper‑activated via relief of IRS‑1 negative feedback, directly phosphorylates EZH2 (inhibiting its methyltransferase activity) and HDAC4 (promoting nuclear export). Reduced H3K27me3 and altered histone acetylation at poised enhancers increase their accessibility in a stochastic manner.
2. TFEB/FOXO compete for limited co‑activators – Nuclear TFEB and FOXO3 sequester co‑activators such as CBP/p300 away from lineage‑defining TFs (e.g., MYOD, HNF4A). This competition creates a tug‑of‑war that produces intermittent bursts of both catabolic and lineage transcripts, measurable as increased cell‑to‑cell variance in scRNA‑seq.
3. mTORC1‑low reduces phosphorylation of S6K1, easing repression of Pol III transcription – Elevated tRNA and 5S rRNA synthesis perturbs nucleolar stress sensing, feeding back to chromatin remodelers (e.g., CHD4) and further destabilizing enhancer‑promoter loops.

These layers converge on a prediction: the rapamycin‑induced state raises enhancer‑specific noise (quantified as variance in ATAC‑seq signal or H3K27ac intensity) more than equivalent mTORC1 suppression by nutrient starvation, while preserving AKT‑dependent survival outputs.

Testable Predictions

• Prediction 1: In young murine liver, rapamycin treatment will produce a higher coefficient of variation (CV) for ATAC‑seq peaks at hepatocyte‑specific enhancers compared with 24‑h fasting, despite similar mTORC1 inhibition levels (p‑S6K).
• Prediction 2: scRNA‑seq will show increased CV of mRNA counts for lineage genes (Alb, Cyp2e1) and decreased CV for canonical autophagy genes (Lc3, Gabarapl1) under rapamycin versus fasting.
• Prediction 3: Pharmacologic AKT inhibition (e.g., capivasertib) combined with rapamycin will normalize enhancer CV and rescue lineage‑gene coherence without abolishing LC3‑II lipidation, indicating AKT’s role in noise generation.
• Prediction 4: ChIP‑seq for EZH2 and H3K27me3 will reveal focal loss of repression at enhancers coinciding with increased noise; CRISPR‑dCas9‑EZH2 targeting to those sites should suppress the noise increase.

Falsifiability

If rapamycin does not produce greater enhancer noise than starvation, or if AKT inhibition fails to reduce noise while preserving autophagy readouts, the hypothesis is refuted. Conversely, confirming the predicted epigenetic and transcriptional variance shifts would support the notion that mTOR acts as a dial that tunes the signal‑to‑noise ratio of enhancer activity, trading off specialized network fidelity for a survival‑biased, noise‑tolerant state.

References

• mTOR complexes and signaling
• IRS‑1 feedback and AKT activation by rapamycin
• TFEB/FOXO cooperation under mTORC1 inhibition
• mTORC1 control of metabolic gene networks