Hypothesis

We propose that the direction of the relationship between CDKN2A exon 2 DNA methylation and mRNA expression is determined by the predominant ANRIL isoform expressed in a given tissue. In vascular and blood cells, the full‑length ANRIL isoform recruits PRC2, maintaining a repressive chromatin state where promoter methylation correlates with silencing. In neurons and astrocytes, an alternatively spliced ANRIL isoform lacking the PRC2‑binding domain instead binds the H3K27 demethylase Jmjd3, creating a chromatin configuration where methylation of exon 2 recruits methyl‑CpG‑binding protein MeCP2, which in turn stabilizes Jmjd3 at the locus, leading to localized demethylation of H3K27me3 and transcriptional activation despite increased DNA methylation.

Mechanistic Basis

ANRIL transcription yields multiple isoforms through alternative splicing of exons 1‑3. The full‑length transcript (ANRIL‑FL) contains a conserved stem‑loop that interacts with SUZ12/PRC2, facilitating H3K27me3 deposition and compaction of the CDKN2A/B TAD 1. Splicing that excludes exon 2 removes this stem‑loop, generating a truncated isoform (ANRIL‑ΔEx2) that retains affinity for Jmjd3 through an exposed arginine‑rich motif. In Alzheimer’s disease brains, oxidative stress promotes splicing factor SRSF1 phosphorylation, shifting the isoform ratio toward ANRIL‑ΔEx2. This isoform scaffolds Jmjd3 to the CDKN2A promoter; MeCP2 binds methylated CpGs at exon 2 and interacts with Jmjd3, creating a positive feedback loop that lifts H3K27me3 repression, thereby increasing p16INK4a transcription 2 3. Consequently, higher exon 2 methylation coincides with higher CDKN2A mRNA, reversing the canonical methylation‑silencing pattern observed in blood where ANRIL‑FL dominates 4.

Testable Predictions

1. Isoform‑specific knock‑down: Silencing ANRIL‑ΔEx2 in primary human astrocytes will convert the positive correlation between exon 2 methylation and CDKN2A expression into a negative one, as measured by bisulfite sequencing and qPCR after forced methylation of exon 2 using dCas9‑DNMT3A.
2. Reciprocal overexpression: Ectopic expression of ANRIL‑FL in AD‑derived neuronal cultures will reduce CDKN2A mRNA despite high exon 2 methylation and increase H3K27me3 levels at the locus (ChIP‑seq).
3. Biomarker shift: The ratio of ANRIL‑ΔEx2 to ANRIL‑FL in circulating extracellular vesicles will inversely correlate with blood CDKN2A expression and directly correlate with CSF p16INK4a levels in longitudinal cohorts of mild cognitive impairment.
4. Pharmacologic modulation: Inhibiting Jmjd3 with GSK‑J4 will diminish the methylation‑dependent activation of CDKN2A in astrocytes, lowering p16INK4a expression and rescuing senescence‑associated secretory phenotype without altering DNA methylation levels.

Implications

If validated, this model explains tissue‑specific epigenetic dysregulation at 9p21, reconciles contradictory methylation‑expression data in Alzheimer’s disease, and highlights ANRIL splicing as a regulatory node linking genetic risk, environmental stress, and cellular senescence across ageing‑related pathologies.