Ribosomal quality control factors inhibit repeat-associated non-AUG translation from GC-rich repeats

Abstract

A GGGGCC (G4C2) hexanucleotide repeat expansion in C9ORF72 causes amyotrophic lateral sclerosis and frontotemporal dementia (C9ALS/FTD), while a CGG trinucleotide repeat expansion in FMR1 leads to the neurodegenerative disorder Fragile X-associated tremor/ataxia syndrome (FXTAS). These GC-rich repeats form RNA secondary structures that support repeat-associated non-AUG (RAN) translation of toxic proteins that contribute to disease pathogenesis. Here we assessed whether these same repeats might trigger stalling and interfere with translational elongation. We find that depletion of ribosome-associated quality control (RQC) factors NEMF, LTN1 and ANKZF1 markedly boost RAN translation product accumulation from both G4C2 and CGG repeats while overexpression of these factors reduces RAN production in both reporter assays and C9ALS/FTD patient iPSC-derived neurons. We also detected partially made products from both G4C2 and CGG repeats whose abundance increased with RQC factor depletion. Repeat RNA sequence, rather than amino acid content, is central to the impact of RQC factor depletion on RAN translation-suggesting a role for RNA secondary structure in these processes. Together, these findings suggest that ribosomal stalling and RQC pathway activation during RAN translation inhibits the generation of toxic RAN products. We propose augmenting RQC activity as a therapeutic strategy in GC-rich repeat expansion disorders.

Graphical Abstract

Figure 1.

NEMF, LTN1, and ANKZF1 act as genetic modifiers of RAN translation from both G4C2 and CGG repeats. (A) Stalled or collided ribosomes are sensed by ZNF598 and RACK1 to separate 80S ribosome to 60S and 40S. The 40S subunit with truncated mRNA is then released by PELO in concert with HBS1L and ABCE1 and subsequently degraded by XRN1. NEMF CAT-tails partially generated peptides within the 60S subunit, allowing for degradation by the proteasome through a process that involves ubiquitination by LTN1 and VCP. ANKZF1 removes the tRNA, allowing for ribosomal recycling. Schematic adapted with permission from (63). (B) Schematic of transiently expressed nano-luciferase (NLuc) reporters used to assess RAN translation product abundance. Either the first intron of C9orf72 including 70 GGGGCC (G4C2) repeats or the 5′UTR of FMR1 containing 100 CGG repeats were placed 5′ to a modified NLuc with the AUG initiator codon mutated to GGG. (C) Results from a targeted screen of key RQC pathway factors. The relative expression of NLuc was normalized to cell titer. These normalized values were then expressed as a fold change compared to a non-targeting siRNA control (NTC). The effect of transient transfection of each siRNA or luciferase reporter on cell titer is shown in Supplemental Figure S1 (Supplementary Figure S1). Data represent mean with error bars ± SEM of 6 biological replicates from at least 2 independent experiments. *P< 0.05; one-way ANOVA with Dunnett's multiple comparison test compared to NTC.

Figure 2.

Depletion of NEMF, LTN1, and ANKZF1 enhances RAN translation in a repeat length-dependent manner and across all reading frames. (A) Schematic of C9orf72 RAN G4C2 repeat length and reading frame reporters. Single nucleotide insertions to shift the reading frame and repeat contractions allowed for measurement of products from all potential reading frames and across 2 repeat sizes. (B–E) Luciferase assays after NEMF, LTN1, both NEMF and LTN1 or ANKZF1 depletion. All graphs show mean with error bars ± SD. Each N is shown as an open circle (n = 6–9/group across at least two independent experiments). Asterisks above each bar are comparisons of expression between NTC and gene(s) knockdown. ns = not significant; *P< 0.05; **P≤ 0.01; ***P≤ 0.001; ****P≤ 0.0001, as determined with two-way ANOVA with Sidak's multiple comparison test. Asterisks placed inside each bar are comparisons between AUG-driven no-repeat control and different repeat lengths of the GA frame of cells treated with gene(s) knockdown. ns = not significant; **P≤ 0.01; ***P≤ 0.001; ****P≤ 0.0001, represent unpaired t-test.

Figure 3.

Detection of partially made products translated from GC-rich transcripts. (A, B) Dual-tagged AUG initiated constructs were generated to allow for capture of N-terminal fragments generated during translation through the repeats. These were generated to detect AUG-V5-(G4C2)₇₀-NLuc-3xFLAG in GA frame and AUG-V5-+1(CGG)₁₀₀-NLuc-3xFLAG in polyG frame. A dual FLAG and V5 IP workflow was used to enrich for incomplete products generated from GC-rich transcripts. 3xFL: 3xFLAG. Green, pink, and purple elements represent other proteins from cell lysate that were cleared from the flow-through. (C, D) V5 antibody IP reveals truncated/stalled products (line next to the blot) generated from G4C2 (panel C) or CGG repeats (panel D) that do not contain the carboxyl-terminal FLAG tag. FT: Flow-through. Representative images from at least two independent experiments.

Figure 4.

Depletion of NEMF, LTN1, and ANKZF1 enhances generation of truncated GA and polyG products from GC-rich transcripts. (A–D) Immunoblots of HEK293 cells transfected with in vitro synthesized G4C2 (GA frame) and CGG repeat RNA reporters (polyG frame) in the presence and absence of NEMF + LTN1, and ANKZF1. Blots represent two biological replicates. Lines next to the blots indicate short/stalled product products. (E–H) Quantification of V5 positive truncated products generated by GA or + 1CGG translation reporters after depletion of NEMF/LTN1 (E and G) or ANKZF1 (F and H). Data represents fold change from non-template control stalled product abundance. Error bars ± SEM of n = 6–8 from at least three independent experiments, *P≤ 0.05; ***P≤ 0.005. t-test with Welch's correction.

Figure 5.

Enhancement of polyG production with NEMF, LTN1 and ANKZF1 depletion requires the CGG repeat RNA structure. (A, B) Prediction of the optimal RNA secondary structure and calculation of the minimum free energy in CGG₁₀₀ and GGN₁₀₀ repeats. Only the repeat region from each repeat was used to predict the RNA secondary structure and calculate the minimum free energy. The results were computed by RNAfold 2.5.1. (C)Schematics of AUG-V5-NLuc-3xFLAG, AUG-V5-+1(CGG)₁₀₀-NLuc-3xFLAG, and AUG-V5-(GGN)₁₀₀-NLuc-3xFLAG transcripts. (D) Knockdown of NEMF, LTN1, and ANKZF1 in HEK293 with no-repeat control, polyG from GGN repeats, and CGG repeats RNA transfection. Data represent means with error bars ± SD of n = 12, ns = not significant; **P≤ 0.01; ***P≤ 0.001; ****P≤ 0.0001. The statistic result placed in the legend is group comparisons by two-way ANOVA with Sidak's multiple comparisons tests.

Figure 6.

Depletion of NEMF enhances repeat-associated toxicity in a fly model of C9 ALS/FTD and DPR accumulation in human neurons. (A) Representative images of Drosophila eyes expressing (G4C2)28 repeats under the GMR-GAL4 driver in the presence or absence of NEMF at 25°C [BDSC36955 #1 and BDSC25214 #2]. Rough eye phenotypes quantified using an established nominal scoring system are shown as violin plots on the right. Individual flies are represented by single data points. n = 30–33/genotype. (B) (G4C2)28 repeats expressed with a GMR-GAL4 driver at 29°C show decreased eye width that is enhanced by the depletion of NEMF, as quantified on the right. Graphs represent the mean with error bars ± SD, n = 21–24. For A and B, *P< 0.05; ****P≤ 0.0001 by one-way ANOVA with Dunnett's multiple comparison test. (C) Schematic workflow for studies with C9ALS patient-derived iNeurons (iN). (D–F) RNA Expression of NEMF, LTN1, and ANKZF1 transcripts from C9ALS and isogenic control iN lysates. Data represent means with error bars ± SD. n = 3–6/gene, ns = not significant; *P< 0.05 by Student's t-test. (G–I) Quantification of GP by MSD assay from C9ALS and isogenic control iNs treated with lenti-empty vector or lenti-shRNA of NEMF, LTN1 or ANKZF1. Data represent mean ± SD; n = 3–6, ns = not significant; **P≤ 0.01; ***P≤ 0.001 by two-way ANOVA with Sidak's multiple comparison test.

Figure 7.

Overexpression of NEMF, LTN1 and ANKZF1 decreases RAN translation from G4C2 and CGG repeats. (A, B) Relative expression of AUG-driven no repeats, (G4C2)70 repeats in the GA frame, and (CGG)100 repeats in the polyG frame when overexpressing empty vector (EV) versus hNEMF or hLTN1. (C) Relative expression of AUG-driven no-repeat control and (G4C2)70 repeats in the GA, GP, and GR frames when overexpressing empty vector (EV) versus hANKZF1. (D) Relative expression of AUG-driven no-repeat control and (CGG)100 repeats in the polyG frame when overexpressing empty vector (EV) versus hANKZF1. Data represent means with error bars ± SD of n = 9–12, ns = not significant; **P≤ 0.01; ***P≤ 0.001; ****P≤ 0.0001 by two-way ANOVA with Sidak's multiple comparison test. (E–G) Relative GP response of C9 and isogenic control iN treated with lenti-empty vector or lenti-hNEMF, hLTN1, and hANKZF1. The level of GP was measured by MSD. Data represent means with error bars ± SD of n = 3–5, ns = not significant; *P< 0.05; ****P≤ 0.0001 by two-way ANOVA with Sidak's multiple comparison test.