RNA triggers chronic stress during neuronal aging

Abstract

Neurodegenerative diseases are linked with dysregulation of the integrated stress response (ISR), which coordinates cellular homeostasis during and after stress events. Cellular stress can arise from several sources, but there is significant disagreement about which stress might contribute to aging and neurodegeneration. Here, we leverage directed transdifferentiation of human fibroblasts into aged neurons to determine the source of ISR activation. We demonstrate that increased accumulation of cytoplasmic double-stranded RNA (dsRNA) activates the eIF2α kinase PKR, which in turn triggers the ISR in aged neurons and leads to sequestration of dsRNA in stress granules. Aged neurons accumulate endogenous mitochondria-derived dsRNA that directly binds to PKR. This mitochondrial dsRNA leaks through damaged mitochondrial membranes and forms cytoplasmic foci in aged neurons. Finally, we demonstrate that PKR inhibition leads to the cessation of stress, resumption of cellular translation, and restoration of RNA-binding protein expression. Together, our results identify a source of RNA stress that destabilizes aged neurons and may contribute to neurodegeneration.

Keywords: aging; double-stranded RNA; integrated stress response; mitochondria; neurodegeneration; stress granules.

Figure 1:. Double-stranded RNA levels are elevated in aged neurons.

(A) Schematic of neuronal transdifferentiation approach. (B) Left, confocal immunofluorescence images of transdifferentiated neurons (Tdiff.1) stained for neuronal markers (cyan) and DAPI (blue). Right, quantification of the fraction of Tubβ3⁺ neurons where the box plot lines denote the 75^th, 50^th, and 25^th percentile of data values; error bars denote the range of non-outlier values, and the “X” denotes the mean. (C) Top, schematic of the eIF2α kinases that activate the integrated response. Bottom, confocal immunofluorescence images of G3BP1 (green) and DAPI (blue) in transdifferentiated neurons (Tdiff.1) treated with the indicated eIF2α kinase inhibitors. The yellow dashed box region denotes the inset, and yellow arrowheads denote chronic stress granules in the cytoplasm of transdifferentiated neurons. White Scale Bar = 10 μm, Yellow Scale Bar = 2 μm. (D) Quantification of the number of stress granules (SGs) per cell from the data in (C) (n=3 replicates). From top to bottom, the box plot lines denote the 75^th, 50^th, and 25^th percentile of data values; error bars denote the range of non-outlier values, the “X” denotes the mean, and any indicated points are outliers. Statistics were calculated using a two-tailed Welch’s t-test for each treatment condition compared to the control. (E) Left and middle, representative confocal immunofluorescence images of Map2 (cyan) and dsRNA detected by the J2 antibody (magenta) in transdifferentiated and iPSC-derived neurons. Right, quantification of relative J2 intensity (n=12); the box plot lines denote the 75^th, 50^th, and 25^th percentile of data values, error bars denote the range of non-outlier values, the “X” denotes the mean, and any indicated points are outliers. White Scale Bar = 10 μm. (F) Dot blot of dsRNA intensity detected by J2 antibody staining; total RNA was loaded in decreasing concentrations from left-to-right. Exogenous dsRNA was added to the indicated cells in culture. (G) Quantification of dsRNA intensity measured in (F) (n=3 replicates, one example shown). (H) Same as in (C) but for the indicated transdifferentiation lines and only treated with the PKR inhibitor. (I) Same as (D) but for the data in (H). (J) Same as (F) but for RNA isolated from mid-age (green) and old-age (blue) cohorts of brain frontal cortex. (K) Same as (G) but for the data in (J).

Figure 2:. dsRNA sequesters stress granule components in transdifferentiated neurons.

(A) Schematic of dsRNA (J2) immunoprecipitation and mass spectrometry. (B) dbSTRING network of proteins that were immunoprecipitated by dsRNA (J2) in transdifferentiated neurons (Tdiff.1) (n=3 replicates). Only significantly interacting proteins (log₂(FC)>5 for J2/IgG pulldown) that clustered within the main network are shown. The superposition of distinct groups of proteins (e.g. ribosome, transcription, etc.) is approximate. Nodes of dsRBPs (purple; GO:0003725) and stress granule proteins (green; GO:0010494) are labeled on the network. (C) Volcano plot of the log₂(FC) of dsRNA-interacting proteins in transdifferentiated neurons (Tdiff.1). The p-value was calculated using a two-tailed Welch’s t-test. Proteins of interest are labeled as in (B), and the canonical stress granule markers G3BP1 and Caprin1 are labeled. (D) Fold change versus fold change plot of J2/IgG pulldown in transdifferentiated neurons and isogenic iPSC-derived neurons. dsRBPs and stress granule proteins are highlighted as in (B). The gray dashed line denotes the y=x line whereas the light gray dashed lines denote log₂(FC)>5 for the Tdiff.1 pulldown and log₂(FC-Tdiff.1) −log₂(FC-iPSC.diff.1)>2. (E) Lollipop plot of the top 5 Gene Ontology terms for the proteins above the light gray dashed lines in (D); the size of the circle scales with the number of proteins within each term. (F) Co-immunofluorescence and intensity profile of G3BP1 and J2 (dsRNA) in the indicated neurons. White Scale Bar = 10 μm. (G) Same as (F) but for Caprin1.

Figure 3:. Aging leads to accumulation of mitochondrial dsRNA in neurons.

(A) Schematic of the dsRNA-seq protocol. (B) Boxplot of the fraction of transcripts identified as mitochondrial (MT-)RNA in the input (IN) and immunoprecipitation (IP) samples of the indicated cell lines (n=3 replicates). (C) Volcano plot of the log₂(Fold Change) of all MT-RNA transcripts identified by dsRNA-Seq in the input and IP samples of the Tdiff.1 transdifferentiated neuron line. (D) Visualization of representative mitochondrial sequencing reads detected by dsRNA-Seq (n=3 replicates) and PKR CLIP (n=2 replicates) in the Tdiff.1 transdifferentiated neuron line. The shaded region denotes the reverse-transcribed MT-ND6 gene. The range of the y-axis values is denoted on the right side of the genomic track and was fixed for each pair of input and IP samples. (E) Same as (D) but for the ~11–17 kb inset of the mitochondrial chromosome for the dsRNA-sequencing results from the indicated cell lines. (F) Same as (B) but for cohorts of human brain tissue (N=2). (G) Same as (C) but for cohorts of human brain tissue (N=2). (H) Same as (D) but for the indicated human brain tissue samples.

Figure 4:. Leaky mitochondria release mitochondrial dsRNA.

(A) Airyscan immunofluorescence-FISH images of the mitochondrial marker TOM70 (magenta) and MT-ND6 transcript (green) in transdifferentiated neurons (Tdiff.1/2/4/5). Green arrowheads denote cytoplasmic MT-ND6 puncta; the white arrows denote the intensity profiles in the adjacent plot. Yellow Scale Bar = 2 μm. (B) Schematic of typical mitochondrial stress test results. (C) Plot of oxygen consumption rate (OCR) in the mitochondrial stress test (1 μM of each compound) of iPSC-diff.1 neurons (n=6 replicates). Error bars are standard deviation. (D) Same as (C) but for Tdiff.1–5 neurons with 3 μM of each compound. (E) Fluorescence confocal microscopy images of representative mitochondria in the neurites of iPSC-diff.1 neurons. The mitochondria were visualized with TMRM (magenta) and MitoTracker (MT) Green (green) and are identified with arrowheads. Rotenone (Rot.) and antimycin A (AA) were added 6 min into the video as indicated by the white arrow. Yellow Scale Bar = 2 μm. (F) Same as (E) but for Tdiff.1 neurons. (G) Quantification of ψ_m as a function of time in the perfusion experiments in panels (E) and (F) (n=3 replicates). Error bars denote standard deviation, and statistics were calculated using Welch’s t-test. (H) Same as (A) but for iPSC-diff.1 treated with DMSO (control), 5 nM AA for 6 h, or 10 μM AA for 30 min (n=3 replicates). The green arrowheads denote cytoplasmic MT-ND6 puncta. (I) Representative genomic tracks of input and immunoprecipitation (IP) samples of PKR CLIP in iPSC-diff.1 neurons treated with AA as in panel (G) (n=2 replicates). The PKR CLIP track for the control condition is reproduced from Figure S4G. The shaded region denotes the MT-ND6 gene.

Figure 5:. PKR inhibition reverses cellular stress and restores ribosomal function.

(A) Schematic of the PKR inhibitor experiments in this figure. (B) Volcano plot of RNA-seq expression of all detected transcripts in transdifferentiated neurons (Tdiff.1) with or without PKR inhibitor treatment (n=3 replicates). The p-values were calculated using a two-tailed Welch’s t-test. Stress granule transcripts are highlighted in green (GO:0010494) and ribosomal transcripts are highlighted in blue (GO:0005840). (C) Quantification of phospho-eIF2α expression relative to total eIF2α expression in transdifferentiated neurons (Tdiff.1) with and without PKR inhibitor treatment as detected by Western blot (n=4 replicates). Statistics were calculated using a two-tailed Welch’s t-test. (D) Volcano plot of proteins detected by G3BP1 AP-MS in transdifferentiated neurons (Tdiff.1) with or without PKR inhibitor treatment (n=3 replicates). Stress granule proteins are highlighted in green (GO:0010494) and ribosomal proteins are highlighted in blue (GO:0005840). P-values were calculated using a two-tailed Welch’s t-test. (E) Lollipop plot of top gene ontology terms calculated by ShinyGO v0.82 among proteins that were enriched >4-fold in the G3BP1 AP-MS data from panel D. The size of the circle correlates to the number of genes in the corresponding term. Ribosomal pathways are highlighted in blue, and the stress granule pathway is highlighted in green. (F) Volcano plot of mRNA transcripts detected by Riboseq in Tdiff.1 neurons with or without PKR inhibitor treatment (n=2 replicates). Stress granule proteins are highlighted in green (GO:0010494) and cytoplasmic translation proteins are highlighted in blue (GO:0002181). P-values were calculated using a two-tailed Welch’s t-test. (G) Lollipop of normalized enrichment scores of top gene ontology terms from data in (F). The size of the circle correlates to the number of genes in the corresponding term. Ribosomal pathways are highlighted in blue, RBP metabolic pathways are highlighted in black.

Figure 6:. Leaky mitochondria cause chronic activation of the stress response in aged neurons.

Our results indicate that neuronal aging leads to (1) MT-dsRNA release, (2) PKR activation, and (3) stress granule assembly. The chronic stress response can be ablated in aged neurons with PKR inhibitor or induced in young neurons with Antimycin A.