PREreview of RNA triggers chronic stress during neuronal aging

by sean lim

Published: August 10, 2025
DOI: 10.5281/zenodo.16790781
License: CC BY 4.0

This expert review synthesizes evidence and provides author-actionable feedback on a preprint investigating whether mitochondrial double-stranded RNA (mt-dsRNA) activates PKR to chronically engage the integrated stress response (ISR) in aging human neurons.

Summary

Question. What upstream stressor/kinase chronically engages the ISR in aging human neurons?

Design. “Aged” neurons via fibroblast transdifferentiation (NGN2+ASCL1) were compared to isogenic “young” iPSC-derived neurons (NGN2). Assays included kinase perturbations (PKR, PERK, GCN2), dsRNA detection (J2 IF/dot blot/dsRNA-seq), protein interactomes (J2 IP–MS; G3BP1 AP–MS), PKR eCLIP-seq, and mitochondrial function (Seahorse OCR; TMRM). Limited human frontal cortex RNA provided orthogonal support.

Key results. PKR inhibition (1 µM, 24 h) reduced chronic G3BP1+ stress granules by ~50% (≈6→≈3 per cell; n=3; reproduced across Tdiff.2/4/5), whereas PERK/GCN2 inhibitors had no effect (Fig. 1C–D,H–I). dsRNA levels were higher in aged vs young neurons and in old vs mid-age cortex (Fig. 1E–K). J2 IP–MS in aged neurons enriched >800 proteins including canonical dsRBPs and stress-granule RBPs, with stronger enrichment than in iPSC-derived neurons (Fig. 2B–E). dsRNA-seq showed that >80% of immunoprecipitated dsRNA was mitochondrial; PKR eCLIP peaks localized prominently to MT-ND6 and nearby intergenic regions in aged neurons and old cortex (Fig. 3B–H; Fig. S4). IF-FISH revealed cytosolic MT-ND6 puncta in multiple aged lines, while Seahorse and live-cell imaging indicated proton leak and ψm insensitivity; antimycin A in young neurons phenocopied cytosolic MT-ND6 and increased PKR–mitochondrial RNA binding (Fig. 4A–I). Functionally, PKR inhibition lowered phospho-eIF2α, increased G3BP1–ribosome interactions, and enhanced translation efficiency of RNA-metabolism/translation gene sets by Ribo-seq (Fig. 5B–G; Fig. S7).

Implication. Mitochondrial dsRNA accumulates and leaks into the cytosol with neuronal aging, binds and activates PKR, and sustains ISR-driven stress granules; acute PKR blockade reverses core stress/translation defects, positioning the PKR/dsRNA axis—and mitochondrial dsRNA processing/containment—as actionable therapeutic entry points.

List of major concerns and feedback

PKR specificity and causality are not yet proven. Relying on a single small-molecule inhibitor leaves room for off-target effects, HRI was never tested, and there is no genetic PKR loss-of-function to establish necessity. Action: Add PKR siRNA or CRISPRi in at least two Tdiff lines; quantify number of stress granules (SGs) per cell, p-eIF2α, and translation efficiency; include ISRIB as a downstream ISR control that should restore translation without lowering J2 signal or PKR–RNA binding.
Cytosolic mitochondrial dsRNA has not been demonstrated biochemically. IF-FISH alone cannot rule out closely apposed mitochondrial signal or probe bleed-through. Action: Perform rapid cytosol/mitochondria fractionation (validate with TOM20/COXIV for mitochondria and GAPDH for cytosol) and RT-qPCR for MT-ND6/ND1 to compute cytosol:mitochondria ratios, add ± RNase III/T1 controls, run J2 IP-qPCR on the cytosolic fraction, and repeat ± PKR inhibitor.
Young–aged comparisons are confounded by substrate, media, and programming. iPSC-derived neurons (Matrigel; NGN2) are not directly comparable to transdifferentiated neurons (PDL/PLO/laminin; NGN2+ASCL1), so age effects may be conflated with culture conditions. Action: Replate iPSC-derived neurons on PDL/PLO/laminin for 48 hours and repeat J2 IF and SG counting, analyzing with mixed-effects models that include donor line and plate as random effects.

List of minor concerns and feedback

J2 specificity in neurons. The manuscript does not yet establish that the J2 signal reflects bona fide dsRNA in neuronal preparations (Just seeing a J2 signal doesn’t guarantee it’s actually dsRNA); please include enzymatic specificity controls by pre-treating matched samples with RNase III (dsRNA-specific) versus RNase T1 (ssRNA-biased) before J2 IF and dot blots to confirm selective loss of signal with RNase III.
NUMTs (Nuclear Mitochondrial DNA segmants) mis-mapping risk in mitochondrial analyses. Reads from nuclear mitochondrial pseudogenes could inflate mitochondrial signals (when you’re actually just seeing NUMTs from nucleus); remap the dsRNA-seq and PKR eCLIP data to the mitochondrial genome using a NUMT blacklist/filter (that excludes NUMTs) and report the estimated mis-mapping rate.

Details for the reproducibility of the study

Randomization and blinding. While SG counting was blinded, extend blinding to all image-derived measurements and randomize plate positions to minimize observer and layout bias, not just for SG counting
Biological replication. Aggregate data across independent donor lines and analyze with donor line as a random effect; explicitly report the number of lines and plates included per assay.
Data calibration. For dsRNA dot blots, include a poly(I:C) standard curve so results can be expressed quantitatively as fmol dsRNA per µg total RNA.

Conclusions and limitations discussed

The dataset supports a working model in which mt-dsRNA accumulates and escapes damaged mitochondria, binds PKR, and chronically activates the ISR in aged neurons, with partial rescue by PKR inhibition. However, specificity (PKR genetic LOF), biochemical proof of cytosolic mt-dsRNA, and rigorous randomization and replication are recommended to convert the model from plausible to compelling. Addressing these points will materially strengthen causal inference and translational relevance.

Concluding remarks

A clear, promising mechanistic story that is a few decisive controls away from being definitive. The fastest path: (1) PKR LOF ± ISRI and (2) cytosol/mitochondria fractionation ± RNase controls.