Hypothesis

In Alzheimer’s disease (AD) peripheral blood, a disease‑enriched splice variant of the long non‑coding RNA ANRIL (hereafter ANRIL‑ε) binds methylated CpG sites within CDKN2A exon 2 and recruits transcriptional co‑activators (e.g., YY1/p300) instead of the canonical Polycomb Repressive Complex 2 (PRC2). This converts the typical repressive effect of intragenic methylation into an activating signal, producing the observed positive correlation between exon‑2 methylation and CDKN2A mRNA levels.

Mechanistic rationale

1. ANRIL isoforms and chromatin context – ANRIL transcription generates multiple isoforms that differ in their 5′ exons and RNA secondary structures. Recent work shows that specific ANRIL isoforms can either scaffold PRC2 to deposit H3K27me3 or act as enhancer‑like RNAs that recruit activator complexes (see Frontiers in Endocrinology)[https://www.frontiersin.org/journals/endocrinology/articles/10.3389/fendo.2018.00405/full].
2. Methyl‑binding switch – Methyl‑CpG binding proteins such as MeCP2 or MBD2 can function as transcriptional repressors or activators depending on interacting partners. In AD blood, oxidative stress‑induced post‑translational modifications of these proteins may favor binding to activator complexes when anchored to methylated DNA (Science Advances)[https://www.science.org/doi/10.1126/sciadv.adk9373].
3. Splice‑variant specificity – RNA‑seq data from AD versus control peripheral mononuclear cells reveal an increase in an ANRIL isoform retaining exon 3 and skipping exon 4 (ANRIL‑ε). This isoform contains a conserved YY1‑binding motif absent in the canonical transcript, providing a plausible mechanism for methylation‑dependent activation (bioRxiv)[https://doi.org/10.1101/2024.06.04.597161].
4. Feedback to histone marks – Activation of CDKN2A transcription via ANRIL‑ε would lead to local loss of H3K27me3 (through recruitment of demethylases such as Jmjd3) and gain of H3K4me3, reinforcing the active state and explaining why histone‑modification clocks based on H3K27me3 may fail in AD blood.

Testable predictions

• Prediction 1: ANRIL‑ε expression will be significantly higher in AD blood than in age‑matched controls, and its levels will positively correlate with both exon‑2 methylation and CDKN2A mRNA (Pearson r > 0.3). Falsification: No difference or inverse correlation.
• Prediction 2: RNA immunoprecipitation (RIP) using antibodies against methyl‑CpG binding proteins (MeCP2/MBD2) will pull down ANRIL‑ε preferentially from AD blood lysates, and mass spectrometry will reveal enrichment of YY1 and p300 in the complex. Falsification: No specific pull‑down or only repressive partners (EZH2, SUZ12) are found.
• Prediction 3: Antisense oligonucleotides (ASOs) targeting the splice junction unique to ANRIL‑ε will reduce its expression in primary AD peripheral blood mononuclear cells, leading to decreased exon‑2 methylation‑associated transcription (measured by nascent RNA‑seq) and a restoration of the negative methylation‑expression relationship seen in controls. Falsification: ASO treatment does not alter CDKN2A transcription or methylation correlation.
• Prediction 4: CRISPR‑dCas9‑KRAB targeting of the ANRIL‑ε promoter will suppress its expression without affecting total ANRIL levels, recapitulating the control methylation pattern (negative correlation) in AD-derived cells. Falsification: No change in methylation‑expression correlation.

Broader implication

If validated, this model would redefine intragenic CpG methylation as a context‑dependent regulatory switch mediated by disease‑specific non‑coding RNAs, offering a novel biomarker (ANRIL‑ε/exon‑2 methylation ratio) and a therapeutic avenue (splice‑modulating ASOs) for AD and potentially other age‑related diseases where similar methylation paradoxes appear.