RNA destabilization in ALS patient-derived fibroblasts
We asked if global or specific alterations in RNA stability are characteristic of ALS by applying Bru-seq and BruChase-seq24 to 14 fibroblast cell lines obtained from individuals with C9orf72-linked familial ALS (C9ALS, four lines), sporadic ALS (sALS, five lines), or controls (five lines; Supplementary Table 1). These methods capture newly synthesized RNA by metabolic labeling of nascent RNA transcripts with bromouridine (BrU), followed by immunoprecipitation of BrU-labeled RNA and deep sequencing (Fig. 1a). The stability of each transcript was assessed by comparing the corresponding reads after 0.5 h of labeling (Bru-seq) to that after a 6 h chase in uridine (BruChase-seq). Out of 22,977 annotated transcripts, we identified 333 RNAs whose stability was altered ≥1.5-fold in C9ALS fibroblasts (Fig. 1b–e; Supplementary Data 1), 56% of which were destabilized (Fig. 1f). Gene set enrichment analysis using gene ontology (GO)25 revealed that the ribosome and oxidative phosphorylation pathways were highly enriched among destabilized transcripts (false discovery rate (FDR) < 0.05; Fig. 1g), while no pathways were enriched among stabilized RNAs.
In sALS fibroblasts, 324 transcripts demonstrated a change in stability ≥1.5-fold compared to controls (Fig. 1h; Supplementary Data 1). For a subset of transcripts tested, changes in RNA stabilization were confirmed by quantitative RT-PCR (qPCR) (Supplementary Fig. 1a). Few pathways were enriched among stabilized or destabilized RNAs in sALS fibroblasts (Fig. 1i). Approximately 1/3 of stabilized transcripts and 1/5 of destabilized transcripts were common to sALS and C9ALS fibroblasts (Supplementary Fig. 1b, c). Even so, no pathways were enriched among the overlapping transcripts, arguing against conserved patterns of altered RNA stability in C9ALS and sALS fibroblasts.
From the Bru-seq data, we identified multiple transcripts whose synthesis differed ≥1.5-fold in controls vs. C9ALS or sALS fibroblasts (Fig. 2a–d; Supplementary Data 2). Nearly all (95%) of the 65 transcripts exhibiting altered synthesis in C9ALS fibroblasts displayed increased rates of production (Fig. 2e)—these RNAs were significantly enriched for inflammatory signaling pathways by GO analysis (Fig. 2f). Only 6 transcripts exhibited increased production in sALS fibroblasts compared to controls, while 22 displayed reduced synthesis (Fig. 2g). No pathways were significantly enriched in either group by GO. Only two transcripts showed increased synthesis in both C9ALS and sALS fibroblasts (MMP1 and NR4A2), while one transcript exhibited a common reduction in synthesis (DACT1; Supplementary Fig. 2 and Supplementary Table 2).
A conserved pattern of RNA destabilization in ALS iPSCs
Given the substantial heterogeneity of fibroblasts in culture26, we were concerned that relevant differences between C9ALS, sALS, and control cells could be obscured. We also questioned whether the observed changes in RNA synthesis and stability would be maintained in cell types other than fibroblasts. Therefore, we reprogrammed a subset of fibroblasts (Supplementary Table 1) into iPSCs using integration-free approaches27,28, verified their pluripotency (Supplementary Fig. 3), and assessed the production and turnover of RNA transcripts by Bru-seq and BruChase-seq.
From 22,984 annotated transcripts, we identified several hundred demonstrating a change in stability ≥1.5-fold over controls in ALS iPSCs (Supplementary Fig. 4a−d; Supplementary Data 3). In C9ALS iPSCs, 956 transcripts demonstrated differences in stability (Fig. 3a), with 36% destabilized ≥1.5-fold. For a subset of transcripts, changes in RNA stability were verified by qRT-PCR (Supplementary Fig. 5a). GO analysis highlighted a profound enrichment in ribosomal and oxidative phosphorylation pathways among destabilized transcripts in C9ALS iPSCs (Fig. 3b), reinforcing the results from C9ALS fibroblasts (Fig. 1g). Network analysis using STRING, a method for illustrating interactions among genes and proteins25, emphasized the strong enrichment for ribosomal and mitochondrial processes in this dataset (Fig. 3c–e).
Nearly all (95%) of the 865 transcripts demonstrating ≥1.5-fold change in sALS iPSCs were stabilized (Fig. 3f; Supplementary Data 3). These were modestly enriched in phagocytic and miRNA pathways, as determined by GO analysis (Fig. 3g); the latter pathway was also enriched in transcripts stabilized in C9ALS iPSCs (compared to Fig. 3b). No pathways were significantly enriched among the 44 destabilized RNAs in sALS iPSCs. Venn diagrams highlighted the 30−40% overlap of stabilized RNA transcripts in C9ALS and sALS iPSCs (Supplementary Fig. 5b), but no pathways were enriched among these transcripts by GO. Only 17 RNAs were commonly destabilized in both groups (Supplementary Fig. 5c), showing no enrichment for specific pathways.
To validate these results and explore their consistency across distinct datasets, we compared RNA abundance in C9ALS and control iPSCs using publicly available RNA-seq data generated by the NeuroLINCS consortium29. Transcripts downregulated ≥1.5-fold in C9ALS iPSCs were highly enriched for the ribosome pathway and mildly enriched for the mineral absorption pathway by GO analysis (Fig. 3h). We also detected a trend towards enrichment for the oxidative phosphorylation pathway, indicating consistent dysregulation of ribosomal and oxidative phosphorylation transcripts in C9ALS iPSCs.
To determine if this pattern was conserved in the human central nervous system (CNS), we examined select transcripts in postmortem cortex and spinal cord from ALS and FTD patients using qRT-PCR (Supplementary Table 3). In spinal cord from C9ALS patients, and in frontal cortex from C9orf72-mutant FTD patients (C9FTD), we detected a significant reduction of ribosome protein-encoding RNAs and oxidative phosphorylation RNAs (Fig. 3i, j). sALS spinal cord exhibited analogous changes (Fig. 3k), suggesting that the distinct patterns of RNA destabilization observed in fibroblasts and iPSCs are reflected in the CNS of human ALS and FTD patients.
In addition to RNA stability, we also investigated RNA synthesis in C9ALS and sALS iPSCs by Bru-seq (Supplementary Fig. 6a−d; Supplementary Data 4). We observed 834 transcripts showing a change in synthesis ≥1.5-fold over controls in C9ALS iPSCs (Supplementary Fig. 6e). Pathways involving focal adhesion and actin cytoskeleton were significantly enriched among RNAs displaying increased synthesis, while transcripts implicated in oxidative phosphorylation and Parkinson’s disease were enriched among RNAs exhibiting reduced synthesis in C9ALS iPSCs (Supplementary Fig. 6f). Among the 541 RNAs exhibiting altered synthesis in sALS iPSCs (Supplementary Fig. 6g), only those involved in rap1 signaling were significantly enriched (Supplementary Fig. 6h). Despite the substantial overlap of transcripts showing increased synthesis in C9ALS and sALS iPSCs (Supplementary Fig. 7a), we observed only moderate enrichment for the focal adhesion pathway. As with fibroblasts, no pathways were enriched among transcripts exhibiting reduced synthesis in C9ALS and sALS iPSCs (Supplementary Fig. 7b).
Transcript characteristics associated with RNA stability
In fibroblasts and iPSCs, transcript length was proportional to stability—shorter transcripts were unstable, while longer transcripts were more stable (Supplementary Fig. 8a, b). Similar relationships were observed for 3′UTR length, which is independently associated with susceptibility to NMD and RNA interference30 (Supplementary Fig. 8c, d). Transcripts exhibiting a change in stability, either increased or decreased, exhibited fewer introns in C9ALS and sALS fibroblasts; although an analogous trend was noted in iPSCs, the effect was less clear (Supplementary Fig. 8e, f).
RBPs regulate RNA stability by shuttling transcripts to processing bodies for degradation, or sequestration and stabilization of transcripts within cytoplasmic stress granules31,32. We therefore asked whether the relative abundance of specific RBP recognition sites among affected transcripts might drive altered RNA stability in ALS patient-derived cells. Motifs recognized by a series of RBPs that form stress granules were enriched among the 3′UTRs of transcripts exhibiting altered stability in ALS iPSCs (Supplementary Fig. 9). Moreover, many of the corresponding RBPs, including TIA1, FUS, and TDP-43, are genetically or functionally linked with ALS, highlighting their implicit role in disease pathogenesis8.
TDP-43 expression mimics RNA instability in control iPSCs
Given the enrichment for TDP-43 motifs among transcripts showing altered stability in ALS iPSCs (Supplementary Fig. 9), the critical function of TDP-43 in RNA splicing10,11,18, and the central connections between TDP-43 deposition and neurodegeneration in both sALS and C9ALS3,5, we surmised that TDP-43 accumulation could contribute to RNA instability in ALS patient cell lines. To test this, we overexpressed TDP-43 fused to enhanced green fluorescent protein (TDP43-EGFP) or EGFP itself in control iPSCs (Fig. 4a), and compared RNA stability in the two groups by BruChase-seq. TDP43-EGFP overexpression induced profound RNA destabilization—of the 1330 transcripts displaying ≥1.5-fold change in stability, 75% were destabilized (Fig. 4b; Supplementary Data 3). As in C9ALS fibroblasts and iPSCs, the ribosomal and oxidative phosphorylation pathways were highly enriched among TDP43-EGFP destabilized transcripts by GO (Fig. 4c). Network analysis confirmed this distinct pattern of RNA destabilization (Fig. 4d–g), outlining well-demarcated clusters centering on ribosomal, mitochondrial, and nucleosomal pathways. GO analysis also highlighted moderate enrichment for RNA transport and nucleotide excision repair pathways among stabilized transcripts (Fig. 4c). Previous studies link both pathways to ALS33,34, testifying to the relevance of TDP-43 and its downstream targets for disease pathogenesis.
One hundred and eighty transcripts were commonly destabilized in C9ALS and TDP43-EGFP-expressing iPSCs (Fig. 4h), accounting for 56% of RNAs destabilized in C9ALS iPSCs. Among these, the ribosome and oxidative phosphorylation pathways were highly and significantly enriched by GO. In the set of commonly destabilized RNAs in C9ALS and TDP43-EGFP overexpressing iPSCs (Supplementary Fig. 10), STRING illustrated clear clusters within the ribosomal and oxidative phosphorylation pathways. These data provide strong support that TDP43-EGFP expression is sufficient to produce a characteristic pattern of RNA instability—one also observed in C9ALS fibroblasts and iPSCs—involving the prominent destabilization of ribosomal and oxidative phosphorylation transcripts. Remarkably, TDP43-EGFP expression had virtually no effect on RNA synthesis as determined by Bru-seq (Supplementary Fig. 11; Supplementary Data 4), suggesting a relatively selective role for TDP-43 in post-transcriptional regulation of gene expression.
Consequences of RNA instability for proteins in iPSCs
To assess the impact of RNA destabilization for ribosomal and oxidative phosphorylation proteins, we took advantage of tandem mass spectroscopy (MS) and simultaneously assessed all measurable components of these pathways in control, C9ALS and sALS iPSCs (Fig. 5; Supplementary Data 5). A total of 61 oxidative phosphorylation proteins were detected in iPSCs by MS. Of these, 52% were significantly reduced in C9ALS iPSCs, and 16% were reduced in sALS iPSCs (Fig. 5a–c). Concordantly, the cumulative abundance of oxidative phosphorylation components was significantly reduced in C9ALS iPSCs, and less so in sALS iPSCs, in comparison to controls. These data are consistent with the destabilization of oxidative phosphorylation transcripts in C9ALS fibroblasts and iPSCs (Figs. 1, 3), and further suggest that oxidative phosphorylation abnormalities may exist in sALS iPSCs independent of RNA instability.
We next asked if ribosomal proteins are likewise affected in ALS iPSCs. For these studies, we separated ribosome-related proteins into two groups: cytoplasmic and mitochondrial ribosomal proteins. Contrary to our expectations based on the observed destabilization of ribosomal protein-encoding RNAs in C9ALS cells, 45% of the 95 cytoplasmic ribosome proteins identified by MS exhibited a significant but subtle increase in abundance in C9ALS iPSCs (Fig. 5d–f), a finding confirmed for a subset of proteins by immunoblotting (Supplementary Fig. 12). Only 3% showed a similar increase in sALS iPSCs, compared to controls. When the abundance of all cytoplasmic ribosomal proteins was measured in aggregate, both C9ALS and sALS iPSCs demonstrated significant but modest increases in comparison to controls. In contrast, C9ALS iPSCs demonstrated a selective reduction in the abundance of mitochondrial ribosomal proteins (Fig. 5g–i). Of the 72 mitochondrial ribosomal proteins detected by MS, 69% were significantly reduced in C9ALS iPSCs, and 3% were reduced in sALS iPSCs, compared to controls.
We also conducted an unbiased overview of all proteins showing a significant change in C9ALS and sALS iPSCs vs. controls. We identified 806 proteins that were significantly and consistently reduced ≥10% in C9ALS iPSCs (Fig. 6a). GO and network analysis using STRING highlighted oxidative phosphorylation constituents and mitochondrial ribosomal subunits among this set of proteins (Fig. 6b–e), confirming the results of investigations that focused specifically on these pathways (Fig. 5). We also noted a significant enrichment for components of the RNA exosome35 among proteins that were significantly reduced in C9ALS iPSCs, suggesting abnormal RNA decay machinery in these cells.
Because previous studies noted direct associations between mitochondrial ribosomes and glycine-arginine (GR) dipeptide repeat proteins encoded by the C9orf72 repeat expansion36, we wondered whether the observed changes in mitochondrial proteins might be related to mitochondrial GR deposition in C9ALS iPSCs. However, we detected no accumulation of GR dipeptides within mitochondria of C9ALS iPSCs by immunocytochemistry (Supplementary Fig. 13), suggesting that the reductions in mitochondrial proteins in C9ALS iPSCs are independent of direct binding by GR. These findings are also consistent with the destabilization of ribosomal and oxidative phosphorylation RNAs in TDP43-overexpressing iPSCs (Fig. 4) that lack GR dipeptides.
The mild but significant upregulation of cytoplasmic ribosomal proteins in C9ALS iPSCs (Fig. 5b) contrasted with the relative instability of ribosome protein-encoding RNAs noted in these cells (Figs. 1, 3). We therefore asked if the observed increase in cytoplasmic ribosome proteins might represent a compensatory change intended to preserve the cell’s capacity to synthesize proteins. In support of this hypothesis, we detected significant upregulation of the protein biosynthesis machinery in C9ALS iPSCs (Fig. 6f). GO and STRING network analysis confirmed the relative enrichment of the amino acid synthesis pathway and further highlighted the proteasomal and RNA transport pathways (Fig. 6g–j). These data suggest that reductions in the stability of ribosome protein-encoding transcripts may be balanced in C9ALS iPSCs by upregulation of ribosomes and elements of the protein synthesis pathway.
We next asked if similar pathways are affected in sALS iPSCs. Unbiased assessment of proteins downregulated ≥10% in sALS iPSCs demonstrated enrichment for components of the oxidative phosphorylation pathway and tricarboxylic acid (TCA) cycle (Supplementary Fig. 14a). Of the 316 proteins exhibiting reduced abundance in sALS iPSCs, 75% were also downregulated in C9ALS iPSCs (Supplementary Fig. 14b). Within this conserved set, GO analysis highlighted both oxidative phosphorylation and the TCA cycle, indicative of shared deficiencies in energy production machinery in C9ALS and sALS patient-derived cells (Supplementary Fig. 14c). We also detected significant upregulation of proteins involved in ubiquitin-mediated proteolysis and amino acid biosynthesis in sALS iPSCs (Supplementary Fig. 14d), similar to what we observed in C9ALS iPSCs (Fig. 6). In fact, 68% of the proteins increased ≥10% in sALS iPSCs were also upregulated in C9ALS iPSCs (Supplementary Fig. 14e). Both amino acid synthesis and ubiquitin-mediated proteolysis pathways were significantly enriched among commonly upregulated proteins in C9ALS and sALS (Supplementary Fig. 14f). Taken together, these data reveal conserved downregulation of energy production pathways coincident with upregulation of protein synthesis and ubiquitin-mediated proteolytic pathways in C9ALS and sALS iPSCs.
Correlating RNA stability and protein abundance in ALS iPSCs
To examine the relationship between RNA stability and protein levels, we compared RNA stability indices for each transcript to the abundance of the corresponding protein measured by MS. A significant and positive correlation between RNA stability and protein abundance was noted for all datasets (Fig. 7a–c), implying that RNA stability predicts protein concentration, particularly for more stable transcripts. Nevertheless, we were struck by the discrepancy between the relative instability of ribosome protein-encoding RNA in C9ALS iPSCs and the abundance of ribosomal proteins as detected by MS. Specifically, we wondered whether feedback mechanisms may be operating to maintain the concentration of ribosomal proteins37,38. In support of this, we did not detect a significant relationship between RNA stability and the abundance of cytoplasmic ribosomal proteins (Fig. 7d–f). In contrast, the stability of oxidative phosphorylation RNAs was clearly tied to the abundance of the corresponding proteins (Fig. 7g–i), consistent with reductions in oxidative phosphorylation proteins in connection with RNA instability in C9ALS fibroblasts and iPSCs.
One hundred and seventy proteins demonstrated concordant changes in RNA stability and protein abundance in C9ALS iPSCs (Supplementary Fig. 15a), representing 18% and 10% of the total changes in RNA stability and protein abundance detected in these cells, respectively. Within this defined set of proteins, we observed a highly significant enrichment for the oxidative phosphorylation pathway (FDR < 1×10−25), supporting the proportional relationship between RNA stability and protein concentration for these candidates. In sALS iPSCs, we found 121 proteins displaying concordant changes (Supplementary Fig. 15b), accounting for 14% and 17% of all differences in RNA stability and protein abundance, but no significant enrichment for proteins within a specific functional pathway.
Given the substantial overlap in RNA stability between C9ALS iPSCs and those overexpressing TDP43, we also compared the set of RNAs stabilized or destabilized by TDP-43 overexpression to the set of proteins whose abundance was altered in C9ALS iPSCs. In doing so, we identified 194 proteins with concordant changes in RNA stability and protein concentration (Supplementary Fig. 15c), representing 20% and 11% of the total RNAs and proteins affected in C9ALS iPSCs. Within this set of 194 proteins, components of the oxidative phosphorylation pathway were highly enriched by GO (FDR < 1×10−21), consistent with what we observed in C9ALS iPSCs (Fig. 7).
To determine if TDP-43 overexpression could recapitulate the pattern of RNA stability and corresponding changes in protein abundance in C9ALS iPSCs, we separately compared the abundance of this defined set of proteins in C9ALS iPSCs to the stability of their corresponding transcripts in (a) C9ALS iPSCs, or (b) TDP43-overexpressing iPSCs (Supplementary Fig. 15d, e). Linear regression analysis illustrated an analogous relationship between protein abundance in C9ALS iPSCs and RNA stability in C9ALS or TDP43-overexpressing cells, with a slope that was nearly identical in the two groups. This is in contrast to the distinct and steep relationship between RNA stability and protein abundance detected in sALS iPSCs (Supplementary Fig. 15f). These data show that TDP-43 overexpression destabilizes many RNAs whose stability is likewise dysregulated in C9ALS iPSCs, and the proteins corresponding to these RNAs. Furthermore, the transcripts most affected by TDP-43 are highly enriched in components of the mitochondrial oxidative phosphorylation pathway.
Mitochondria and ribosome function in patient-derived cells
To determine the functional implications of these observations, we assessed mitochondrial and ribosomal activity in ALS patient samples using a variety of methods. First, since mitochondrial function is closely tied to morphology39,40, we examined mitochondrial morphology in control and C9ALS fibroblasts by live-cell microscopy (Fig. 8a). Although overall mitochondrial content and length were unchanged (Supplementary Fig. 16a, b), C9ALS fibroblasts displayed relatively simple and rounded mitochondria in comparison to control cells41, as determined by mitochondrial form factor (an estimate of irregularity or branching, Fig. 8b) and aspect ratio (representing circularity, Fig. 8c). Accordingly, a composite measure of form factor and aspect ratio clearly distinguished mitochondria from control and C9ALS fibroblast (Fig. 8d). Fixed fibroblasts from sALS and C9ALS patients likewise demonstrated fewer and less distinct mitochondrial puncta in comparison to control cells (Supplementary Fig. 17).
To investigate mitochondrial morphology in iPSCs, we used tetramethylrhodamine (TMRE), a cell-permeable dye that is concentrated within active mitochondria42. As in fibroblasts, C9ALS iPSCs demonstrated a reduction in mitochondrial form factor and aspect ratio (Fig. 8e–g) but not mitochondrial content or length (Supplementary Fig. 16c, d), in comparison with control iPSCs. These results closely parallel those obtained in fibroblasts, revealing consistent abnormalities in mitochondrial morphology in C9ALS patient-derived fibroblasts and iPSCs.
We also measured mitochondrial respiratory chain function in control, C9ALS and sALS fibroblasts by Seahorse bioenergetic profiling43. Baseline ATP production, coupling efficiency, and ATP-coupled respiration were unchanged in ALS patient fibroblasts (Supplementary Fig. 16e-g). However, we observed differences in spare respiratory capacity when the cells were energetically challenged with higher concentrations of carbonyl cyanide p-[trifluoromethoxy]-phenyl-hydrazone (FCCP) (Fig. 8h). FCCP uncouples the mitochondrial inner membrane by allowing free exchange of ions, thereby depleting the mitochondrial membrane potential44. At 900 nM FCCP, both C9ALS and sALS fibroblasts show a significant decrease in spare respiratory capacity, suggesting that these cells are energetically fragile compared to controls and unable to maintain energy production to meet increased energy demands.
To estimate protein synthesis in ALS patient-derived cells, we took advantage of SUnSET (surface sensing of translation)45, a nonradioactive and quantitative method that measures the incorporation of puromycin into nascent polypeptides. We were unable to detect a difference in protein synthesis between ALS and control fibroblasts using this method (Supplementary Fig. 16h), which we attributed to the heterogeneity of fibroblasts in culture. However, we detected significantly greater puromycin antibody reactivity in C9ALS iPSCs than in control iPSCs (Fig. 8i), consistent with an increase in overall protein translation in C9ALS iPSCs. To verify these observations, we focused on the kinetics of an exogenous reporter protein (mCherry) expressed under the control of an integrated eIF2 promoter. After a brief photobleach, we measured the fractional return of mCherry fluorescence to estimate the rate of protein synthesis within control and C9ALS iPSCs46 (Supplementary Fig. 18). In doing so, we noted a significantly faster rate of return in C9ALS iPSCs compared to control iPSCs (Fig. 8j). These data complement the assessment of protein synthesis by SUnSET and confirm the upregulation of protein synthesis in C9ALS iPSCs that was suggested by proteomics (Fig. 6).