Hypothesis

During sleep, nuclear acetyl‑CoA generated by ACSS2 does more than activate waste‑clearance genes; it creates a bidirectional epigenetic signal that simultaneously promotes the transcription of synaptic‑plasticity genes earmarked for consolidation and represses those marked for elimination. In this model, the same acetyl‑CoA‑dependent histone acetylation wave functions as a molecular ‘selector’: high acetylation at promoters of BDNF, Arc and PSD‑95 stabilizes selected synapses, while concurrent deacetylation (via HDAC2) at tags of weakly activated synapses permits their removal via microglia‑mediated pruning. Chronic sleep disruption uncouples this bistable switch, leading to either excessive synaptic retention (producing noisy networks) or premature pruning (driving cognitive decline), independent of extracellular waste load.

Mechanistic Rationale

1. Acetyl‑CoA flux bifurcation – Sleep‑associated fasting‑like metabolism raises nuclear acetate, fueling ACSS2‑dependent acetyl‑CoA synthesis. This pool acetylates H3K27 at promoters of both autophagy genes (TFEB, LC3) and immediate‑early‑gene (IEG) enhancers, creating a permissive chromatin state.
2. HDAC2 gating – In parallel, sleep reduces HDAC2 nuclear retention via phosphorylation‑dependent export, allowing selective deacetylation of H3K9 at synaptic loci that experienced low CaMKII activity during prior wakefulness. The balance between acetyl‑CoA‑driven acetylation and HDAC2‑mediated deacetylation determines whether a synapse is tagged for stabilization or elimination.
3. Coupling to clearance – The acetyl‑CoA wave also upregulates aquaporin‑4 polarization and lysosomal biogenesis, ensuring that the structural debris from pruned synapses is efficiently cleared by glymphatic flow and autophagy.
4. Failure mode – Sleep loss diminishes nuclear acetyl‑CoA (lower ACSS2 nuclear import) while sustaining or elevating HDAC2 activity, shifting the equilibrium toward hypoacetylation at IEG promoters and hyper‑deacetylation at pruning tags. Result: weakened synaptic tagging, aberrant microglial engulfment, and accumulation of both protein aggregates and dysfunctional synapses.

Testable Predictions

• Prediction 1: In mice, optogenetic activation of cholinergic basal forebrain nuclei during wakefulness will mimic the sleep‑associated nuclear acetyl‑CoA surge, increasing H3K27ac at TFEB and BDNF promoters and enhancing both amyloid‑β clearance and spine retention after learning.
• Prediction 2: Pharmacological inhibition of ACSS2 nuclear translocation (using a selective ACSS2‑NLS blocker) during sleep will reduce H3K27ac at autophagy and IEG genes, leading to decreased glymphatic influx (measured by MRI‑contrast tracer) and increased persistence of low‑activity synapses (quantified by synaptic‑protein turnover via puromycin‑labeling).
• Prediction 3: Chronic sleep fragmentation will elevate nuclear HDAC2 occupancy at PSD‑95 promoters, correlating with increased microglial phagocytic markers (CD68) and reduced LTP magnitude; HDAC2 inhibition (e.g., with SB939) should rescue spine density without altering extracellular amyloid levels.
• Prediction 4: Human PET‑derived acetate utilization in the brain (using ^11C‑acetate) during nocturnal sleep will predict next‑day performance on a declarative memory task, independent of glymphatic CSF‑influx measured by diffusion‑weighted imaging.

Experimental Outline

1. Acetyl‑CoA imaging – Employ a genetically encoded nuclear acetyl‑CoA sensor (e.g., ACE‑NLS) in cortical neurons of sleeping vs awake mice; correlate fluorescence intensity with ChIP‑seq for H3K27ac at TFEB, BDNF, and PSD‑95 loci.
2. Manipulate ACSS2 – Viral delivery of a NLS‑mutant ACSS2 or a DREADD‑controlled ACSS2 construct; assess nuclear acetyl‑CoA, histone marks, autophagic flux (LC3‑II/I), glymphatic clearance (intralymphatic tracer), and spine dynamics (two‑photon imaging).
3. HDAC2 modulation – Use HDAC2‑specific shRNA or conditional knockout in forebrain excitatory neurons during sleep deprivation; evaluate synaptic tagging (pCREB, Arc), microglial engulfment (Iba1+ phagocytic cups), and behavior (novel object recognition, fear conditioning).
4. Human translation – Simultaneous ^11C‑acetate PET and CSF‑amyloid PET in healthy volunteers after normal sleep vs 24‑h sleep deprivation; relate nuclear acetate synthesis rates to next‑day memory scores and CSF neurofilament light as a proxy for synaptic injury.

Falsifiability

If nuclear acetyl‑CoA levels during sleep do not predict concurrent changes in both autophagy/glymphatic gene expression and the acetylation state of synaptic plasticity promoters, or if manipulating ACSS2 nuclear import fails to alter either waste clearance or synaptic selection as outlined, the hypothesis would be refuted. Conversely, observing the predicted bidirectional epigenetic coupling would substantiate the view of sleep as an active nuclear‑metabolite‑driven editing session that decides which neural architectures persist.