Tau is an m 6 A-associated RNA binding protein that causes RNA dysregulation in disease

TL;DR

Tau binds RNA in an m6A- and hnRNPA2B1-dependent manner, and the FTD-causing MAPT V337M mutation disrupts hnRNPA2B1 phosphorylation, localization, and RNA binding, contributing to RNA dysregulation. These findings support a direct RNA-regulatory role for tau relevant to tauopathy.

Key findings

• Proteomics of postmortem MAPT coding-variant brains and MAPT V337M iNeurons identified 253 well-characterized RBPs with differential levels in both datasets, with enrichment for RNA-regulatory pathways.
• MAPT V337M iNeurons show increased total RNA and widespread transcript abundance/isoform changes; mutation correction reversed transcript levels for 7 of 8 tested transcripts (INPP5F was not confidently rescued). A CRISPRi screen identified HNRNPA2B1 and YTHDC1 as top modifiers, and HNRNPA2B1 knockdown reversed increased RNA levels in MAPT V337M but not WT iNeurons.
• eCLIP showed tau binds many neuron-specific transcripts, with ~47% of WT- or V337M-tau-bound transcripts also bound by hnRNPA2B1. Tau binding overlapped strongly with m6A peaks, and HNRNPA2B1 knockdown or METTL3 inhibition with STC-15 reduced tau binding.
• In MAPT V337M iNeurons, hnRNPA2B1 was hyperphosphorylated at two RRM2 sites, nuclear-enriched, and showed decreased RNA binding, especially at exonic features. Nuclear/cytoplasmic RNA-seq showed compartment-specific dysregulation; 76 of 176 differentially expressed genes from a MAPT V337M patient dataset overlapped hnRNPA2B1-bound and/or decreased nuclear abundance transcripts, including NRXN1.

Methods

• Proteomics on postmortem brain samples from healthy donors and patients with MAPT coding variants (V337M, P301L, L284L, S305I, S305S), plus iPSC-derived iNeurons (WT, MAPT V337M, MAPT knockdown, mutation-corrected M337V, MAPT knockout, HNRNPA2B1 knockdown).
• Long-read RNA sequencing (Oxford Nanopore), short-read and nuclear/cytoplasmic RNA sequencing, CRISPRi screen targeting characterized RBPs with SYTO RNA Select sorting, eCLIP for tau and hnRNPA2B1, MeRIP-seq for m6A, phosphoproteomics, western blot, immunofluorescence, and qPCR.
• Analyses included differential expression/isoform usage (edgeR, satuRn), eCLIP binding windows (Skipper, Flipper), motif analysis (HOMER), cyto/nuc abundance ratio modeling (DESeq2 interaction), and Fisher’s exact tests for overlaps.

Limitations

• Stated in the paper: long-read differential expression used n=4 per group and no covariates due to low sample size; nuclear and cytoplasmic effects are difficult to completely decouple; the reason for binding depletion immediately at the m6A peak is unknown; more extensive RNA-profiling patient datasets are needed; class C transcripts are not explained by the proposed model.
• Evident from the text: the iPSC/iNeuron model and transgenic FLAG-tagged eCLIP may not fully reflect endogenous tau/RBP biology; postmortem patient brains represent late-stage disease while iNeurons model early-stage toxicity; some findings rely on small replicate numbers (e.g., n=2 for fractionation western blots); causality of hnRNPA2B1/m6A in human disease remains partly correlative.