Polyglutamine-mediated ribotoxicity disrupts proteostasis and stress responses in Huntington’s disease

uORF dynamically regulates HTT translation

Human HTT messenger RNA harbours a uORF (Fig. 1a) that suppresses translation of a downstream reporter when ectopically expressed¹¹. This uORF is evolutionarily conserved from mouse to human in size and distance from the main open reading frame (ORF) (Fig. 1b), and features an optimal translation initiation context¹². uORFs are generally known to inhibit the translation of downstream ORFs¹³. To determine whether this uORF inhibits HTT translation in vivo, we analysed previously published ribo-seq datasets. In cultured human cells treated with either lactimidomycin (LTM) or harringtonine (Harr.)¹⁴, which arrest ribosomes at translation start sites, we find initiating ribosomes on both the uORF and main ORF of HTT mRNA (Fig. 1c and Extended Data Fig. 1a). In cells treated with cycloheximide (CHX), which blocks translation elongation, ribosomes occupy the entire length of the uORF but not flanking regions (Fig. 1c and Extended Data Fig. 1a), suggesting that ribosomes translate the uORF and terminate ahead of the main HTT ORF. Similar ribosome traces were also found in mouse hippocampus and cultured primary neurons¹⁵ (Fig. 1d), as well as other tissues and cell types (Extended Data Fig. 1b).

If the uORF suppresses translation of the main ORF in vivo, this should be reflected by low translation efficiency (TE), calculated as ribosome occupancy on the main ORF divided by RNA abundance. Indeed, the TE of Htt in mouse hippocampus is lower than that of highly expressed proteins and similar to that of Atf4 and Ppp1r15a (GADD34) (Fig. 1e), which are translationally repressed by their uORFs under basal conditions¹⁶. Taken together, these observations suggest that the endogenous uORF inhibits translation of HTT in vivo by reducing initiation on the main ORF.

uORF-regulated transcripts, including Atf4 and Ppp1r15a, are known to increase their translation during conditions that trigger the integrated stress response (ISR), including endoplasmic reticulum (ER) stress and neuronal differentiation. This occurs through a mechanism that involves uORF bypassing and/or re-initiation^16–18. To determine whether Htt translation is upregulated during stress, we analysed ribo-seq datasets from mouse embryonic fibroblasts undergoing acute ER stress¹⁹ or SH-SY5Y neuroblastoma cells undergoing differentiation¹⁷. In response to stress, more ribosomes bypass the uORF on both ATF4 and HTT (Extended Data Fig. 1c). To confirm this, we monitored the effect of thapsigargin (Tg)-induced ER stress on Htt translation in neuron-like cells isolated from mouse striatum (STHdh striatal cells)²⁰—the region of the brain most affected in patients with HD¹. We measured the association of specific transcripts with translating polysomes by sucrose gradient fractionation and quantitative polymerase chain reaction (qPCR). Tg treatment reduced global translation, seen as ribosome run-off in polysome profiles (Fig. 1f), and displaced the housekeeping β-actin (Actb) mRNA from polysomes (Fig. 1g), consistent with a general attenuation of translation initiation during stress. In contrast, Htt and Atf4 mRNAs responded to Tg treatment by shifting to heavier polysome fractions (Fig. 1g), as expected for uORF-repressed transcripts that are translationally induced during stress. Furthermore, immunoblot analysis confirmed that Htt protein levels increase in Tg-treated striatal cells (Fig. 1h).

uORF modulates aggregation of mHTT

Because HTT aggregation is concentration dependent², we next tested whether the presence of a functional uORF in HTT mRNA affects the solubility of the encoded protein. Expression of HTT-ex1–GFP variants with and without the endogenous 5′ untranslated region (UTR) of human HTT revealed that the uORF delays formation of inclusions in transfected 293T cells (Fig. 2a–c and Extended Data Fig. 2a). This effect was lost by mutating the uORF AUG to AAG, preventing translation initiation (Fig. 2a–c and Extended Data Fig. 2a). Immunoblot analysis of the soluble and insoluble fractions of transfected cells confirmed that the uORF reduces the fraction of insoluble mHTT-ex1 in both 293T and mouse striatal cells (Fig. 2d and Extended Data Fig. 2b,c).

Interestingly, the uORF also reduced formation of truncated mHTT-ex1 fragments lacking the C-terminal GFP, which were detected only in the insoluble fraction of transfected cells (Fig. 2d, pink arrowheads, and Extended Data Fig. 2b). Such neurotoxic mHTT-ex1 truncated fragments are also detected in brains of patients with HD and mouse models¹. To test whether mHTT-ex1 truncations originate from premature translation termination on mHTT-ex1, we used a previously established translation stalling reporter, GFP–K^AAA(20)–mCherry²¹. Upon transfection, the ratio of mCherry to GFP, translated from a single ORF and measured by flow cytometry, provides a proxy for abortive termination. We replaced the polylysine stalling sequence with HTT-ex1 and found that translation of polyQ(97) incurs a higher rate of abortive termination compared with polyQ(25) (Fig. 2e and Extended Data Fig. 2d). Inhibitors of translation initiation reduce premature termination on the polylysine reporter by lowering ribosome density on the coding region^22,23. Abortive termination of polyQ(97) was also reduced by mild suppression of translation initiation using 4EGI-1 (Fig. 2e and Extended Data Fig. 2e,f), an eIF4A1 inhibitor that ameliorates motor symptoms in a mouse model of HD²⁴. Mild suppression of initiation also reduced the ratio of truncated to full-length mHTT-ex1 found in the insoluble fraction of transfected 293T cells (Extended Data Fig. 2e). We conclude that attenuating mHTT-ex1 translation initiation, either pharmacologically or via its uORF, reduces premature translation termination and release of aggregation-prone fragments.

Ribosome collisions on mHTT cause premature translation termination

Since persistent ribosome collisions can lead to abortive termination²⁵, we asked if CAG expansions increase the risk of ribosome collisions on mHTT-ex1. Indeed, disomes can form on glutamine repeats in yeast²⁶ and polyproline tracts, such as the one just downstream of the polyQ in HTT, are known to cause collisions^27,28. We were unable to test this on endogenous HTT mRNA due to its low abundance and low TE, as well as the inability to confidently map CAG repeats by ribo-seq. To address this gap, we developed a viral delivery platform to express, label and detect HTT nascent chains (NCs) in cells. We introduced the coding sequence for wild-type (WT) polyQ(8) or mutant polyQ(73) human HTT-ex1 into the single ORF of poliovirus (Fig. 3a, top). Upon infection, poliovirus rapidly shuts off cellular mRNA translation and reprogrammes ribosomes to translate viral RNA from an internal ribosome entry site²⁹. In the engineered virus, HTT-ex1 is translated as part of the viral polyprotein and then released by proteolytic cleavage once the downstream viral protease is fully synthesized.

Consistent with a biosynthetic defect caused by polyQ expansion, viruses harbouring polyQ(73) replicated more slowly than their polyQ(8) counterparts (Extended Data Fig. 3a). To monitor ribosome distribution on HTT-ex1, we labelled NCs with puromycin, which becomes covalently incorporated by elongating ribosomes and can be detected using a puromycin-specific antibody³⁰. If ribosomes are evenly distributed on the translated RNA, the size of puromycylated NCs should also be evenly distributed; in contrast, local accumulation of ribosomes on a specific region of the RNA will generate puromycylated NCs of a discrete size range. As predicted, expression of mHTT-ex1 in either Huh7 (Fig. 3b and Extended Data Fig. 3b) or SH-SY5Y neuroblastoma (Extended Data Fig. 3c) cells produced a strong smear of puromycylated NCs (Fig. 3b, top and middle), indicative of local ribosome build-up. The size of these puromycylated products is consistent with ribosome stalling on the expanded polyQ tract, before cleavage of N- and C-terminal viral sequences by the downstream viral protease.

Collided ribosomes form a tight interface that is resistant to RNase digestion³¹. If ribosome build-up on mHTT generates such an interface, the above polyQ(73)-puromycylated NCs should be associated with RNase-resistant disomes or higher-order polysomes. To test this, we infected cells as above, digested lysates with RNase I and performed sucrose gradient fractionation. The presence and size of NCs in each fraction were assessed by puromycin labelling. As predicted, more RNase-resistant trisomes were formed on polyQ(73) than polyQ(8) (Fig. 3c, top). Importantly, these trisomes were enriched for the puromycylated NCs specific to polyQ(73) (Fig. 3c and Extended Data Fig. 3d).

To isolate ribosomes that collide on mHTT-ex1, we infected cells as above, fractionated lysates on sucrose gradients and labelled NCs using puromycin. Based on puromycin immunoblots (Fig. 3d), we extracted ribosomes from fractions with polyQ(73)-specific collision events, and analysed their protein content by liquid chromatography–tandem mass spectrometry (LC–MS/MS). Consistent with premature termination on CAG expansions, less viral NCs were detected downstream of polyQ(73) as compared with polyQ(8) (Fig. 3e). This resulted in lower levels of the viral protein encoded downstream of mHTT-ex1 (Extended Data Fig. 3e). Taken together, we conclude that polyQ expansions in HTT increase the likelihood of ribosome collisions, abortive termination and release of aggregation-prone mHTT fragments.

mHTT depletes soluble eIF5A

Collisions recruit ribosome quality control (RQC) and protein quality control (PQC) factors to promote either resolution and continued elongation or premature termination and clearance^25,32,33. To determine whether polyQ(73) translation attracts specific RQC and PQC factors, we compared the proteins that interact with polysomes translating either polyQ(73) or polyQ(8) HTT-ex1 (Supplementary Table 1). We find that ribosomes translating polyQ(73) mHTT recruit more RQC and PQC factors, including higher levels of proteasomes (Fig. 3f, left); collision-associated factors GCN1, ABCE1 and UFL1; and PQC factors involved in the clearance of mHTT, including UBR5 (ref. 34), VCP/p97 (ref. 35), HUWE1 (ref. 36), ADRM1 (ref. 37) and HSPB1/Hsp27 (ref. 38) (Fig. 3f, right).

Interestingly, despite similar interaction of translation elongation factors eEF1A1 and eEF2, we find that elongation factor eIF5A is markedly depleted from ribosomes translating polyQ(73) (Fig. 3f, pink, and Fig. 3g). Given the critical role of eIF5A in translation, proteostasis and neurodevelopment^39–41, as well as its involvement in TDP-43 aggregation⁴², we hypothesized that its depletion from ribosomes—potentially mediated by interactions with mHTT—may play a role in the aetiology of HD. To determine whether mHTT interacts with eIF5A, we transfected 293T cells with plasmids encoding Htt-ex1–GFP variants followed by GFP affinity purification. More eIF5A co-purified with polyQ(97) mHTT as compared with WT control (Fig. 4a). Similarly, more eIF5A interacted with mHTT-ex1 in mouse N2a cells⁴³ (Fig. 4b) and in yeast⁴⁴. Finally, eIF5A was also highly enriched in the insoluble fraction of N2a cells expressing mHTT-ex1 (ref. 45) (Fig. 4c), suggesting that its interactions with soluble mHTT lead to aggregation.

To relate these observations to HD pathology, we next examined a longitudinal dataset of soluble and insoluble brain proteomes from pre-symptomatic and symptomatic R6/2 HD mice, expressing mHtt with polyQ(>140), compared with age-matched WT controls⁶ (Fig. 4d). eIF5A was highly abundant in the insoluble fraction of aged, symptomatic R6/2 HD brains, but not that of young animals (Fig. 4e). Analysis of soluble translation elongation factors revealed that eIF5A—but not other elongation factors—becomes anti-correlated with soluble mHtt as R6/2 mice get older and develop symptoms (Fig. 4f and Extended Data Fig. 4a). Interestingly, at the pre-symptomatic age of 8 weeks, we observed an increase in the soluble levels of eIF5A and RQC factors Gcn1, Ltn1 and Drg2 (Extended Data Fig. 4b,c), potentially reflecting an early adaptive response to ribotoxic stress.

eIF5A facilitates translation elongation across ribosomal stall sites, particularly proline-rich sequences^27,46. Knockdown of eIF5A increases ribosome pausing and collisions in yeast and human cells^46,47. Therefore, we hypothesized that the reduction in soluble eIF5A in older HD mice should result in increased ribosome collisions throughout the transcriptome. Since ribosome collisions lead to ubiquitination of specific ribosomal subunits^48,49, we searched for these hallmark ubiquitination events in brain proteomics and ubiquitomics data of another HD mouse model (zQ175 (ref. 50)). This model showed a similar inverse correlation between soluble mHTT and eIF5A levels (Fig. 4g, left), as well as increased collision-related ubiquitination of ribosomal proteins^48,49 in aged HD animals (Fig. 4g).

Loss of eIF5A affects translation elongation dynamics

To better define how loss of eIF5A affects translation and cellular responses in HD, we used striatal cells from mice expressing two copies of full-length Htt with either polyQ(7) or polyQ(111) tracts²⁰. First, we confirmed that expression of mHtt is associated with lower levels of soluble eIF5A (Fig. 5a) and higher levels of 40S ribosomal protein ubiquitination (Fig. 5b), similar to HD mouse brains. Next, we treated lysates with RNase I, isolated RNase-resistant polysomes and analysed their content by LC–MS/MS (Supplementary Table 1). As expected, more RNase-resistant polysomes were recovered from cells expressing mHtt (Fig. 5c, black). Furthermore, polyQ(111) polysomes were depleted for eIF5A and enriched for the collided ribosome recognition complex Gcn1–Gcn20–Rbg³³ (Fig. 5c, pink). We next performed ribosome profiling and pause score analysis. As reported⁵¹, we find mHtt increases the rate of ribosome stalling (Fig. 5d,e). These stalling events were indeed enriched for PPX amino acid triplet motifs within the ribosomal active site (Fig. 5e and Supplementary Table 1). PPX motifs are known to depend on eIF5A for efficient elongation³⁹, confirming that eIF5A deficiency is linked to ribosome stalling in HD. We next examined the mRNAs affected by increased stalling in polyQ(111) cells. These disproportionately encode proteins involved in proteostasis, including ribosomal and proteasomal proteins, as well as the chaperonin TRiC/CCT and factors involved in myelination and glycolysis (Fig. 5f,g and Supplementary Table 1). Consistent with ribosome pausing leading to stalled NC aggregation⁵², the protein products of mRNAs encountering increased stalling in polyQ(111) cells were enriched in the insoluble brain proteome of aged R6/2 mice (Extended Data Fig. 5a). Stalling and aggregation were also accompanied by disruption of major proteostasis machines. Thus, the soluble brain proteome of 8- and 12-week-old R6/2 mice exhibits an early and persistent loss of 1:1 stoichiometry among core ribosomal proteins and proteasomal subunits (Extended Data Fig. 5b–d). The stoichiometry loss cannot be explained by increased aggregation of specific proteins, because neither ribosomal nor proteasomal proteins were enriched in the insoluble brain proteome of 8-week-old HD mice (Extended Data Fig. 5e). These analyses indicate that the increase in aberrant translation pausing due to eIF5A depletion disproportionally affects mRNAs encoding key proteostasis complexes, thereby impacting their biogenesis. We propose that these events seed a vicious cycle of proteostasis dysfunction that contributes to HD symptoms. Supporting this notion, mutations affecting ribosomal and proteasomal stoichiometry were linked to neurodegeneration⁵³. Taken together, we conclude that mHTT-ex1 protein interacts with and sequesters eIF5A, altering translation elongation dynamics on hundreds of mRNAs and disrupting proteostasis.

Loss of eIF5A impedes recovery from acute stress

ER stress and activation of the unfolded protein response are common features of HD, and their modulation can affect neurotoxicity⁵⁴. Indeed, we find that polyQ(111) striatal cells are hypersensitive to Tg-induced ER stress (Fig. 6a), as reported⁵⁵. One important aspect of the stress response is phosphorylation of translation initiation factor eIF2a (EIF2S1), leading to ISR induction. eIF2a phosphorylation induces translation of uORF-repressed factors such as ATF4 and GADD34/PPP1R15A, which promote recovery from stress¹⁶. Because the yeast eIF5A plays a central role in translation of uORF-repressed stress effectors^18,40, we hypothesized that depletion of eIF5A by mHTT may impair the translational induction of these protective factors, such as ATF4. Indeed, translation of Atf4 at steady state is lower in polyQ(111) cells, despite similar levels of Atf4 mRNA and higher basal phosphorylation of eIF2a (Extended Data Fig. 6a). This is associated with higher ribosome occupancy on the Atf4 uORF (Extended Data Fig. 6b), suggesting that translation of the main ORF is not induced despite eIF2a phosphorylation. Importantly, Atf4 accumulation in response to Tg was slower and weaker in polyQ(111) cells (Fig. 6b), potentially accounting for their increased susceptibility to stress (Fig. 6a).

To further assess if eIF5A is required for ISR-mediated induction of protective Atf4 in striatal neurons, we next treated WT polyQ(7) cells with eIF5A inhibitor GC7, followed by stress induction by Tg. On its own, GC7 caused a slight elevation in Atf4 levels (Fig. 6c, left). Importantly, when combined with Tg treatment, GC7 dampened Atf4 production in a dose-dependent manner (Fig. 6c, right). Next, we tested whether supplementing eIF5A boosts the weak Atf4 response to Tg stress in polyQ(111) cells. We overexpressed eIF5A and its essential modifying enzyme deoxyhypusine synthase (Dhps) in polyQ(111) cells followed by treatment with Tg. The eIF5A supplementation increased Atf4 protein levels during ER stress in HD cells (Fig. 6d), albeit to a modest extent. We conclude that eIF5A activity is required for the upregulation of ISR-regulated protective factors in response to ER stress in striatal cells and that mHtt expression dampens this response. Supporting this conclusion, symptom onset in HD mice is associated with eIF5A depletion from the soluble proteome and with increased collision-related ubiquitination of ribosomal proteins (Fig. 4e–g). Taken together, our data indicate that interventions to reduce the loss of eIF5A or otherwise increase its activity could be attractive therapeutic strategies in HD.

Reducing translation initiation mitigates ribotoxicity in HD

If mHTT sensitizes cells to ribosome collisions, HD cells should be more sensitive to drugs that increase collisions by slowing down elongation and less sensitive to drugs that decrease collisions by slowing down initiation. Indeed, polyQ(111) cells were hypersensitive to low doses of drugs that trigger collisions, for example, emetine (EME), anisomycin (ANS) and CHX⁵⁶ (Fig. 7a). By contrast, polyQ(111) cells were less sensitive to low doses of translation initiation inhibitors Rocaglamide A (RocA) and 4EGI-1 (Fig. 7b). This is consistent with our earlier observation that 4EGI-1 reduced abortive translation termination in 293T cells transfected with mHTT-ex1 (Fig. 2e and Extended Data Fig. 2e), as well as the ameliorative effects of 4EGI-1 in R6/1 mice²⁴. Finally, we speculated that the lower rate of translation initiation in polyQ(111) cells⁵¹ (Extended Data Fig. 6a) could reduce ribosome collisions and abortive termination. We transfected polyQ(7) and polyQ(111) cells with GFP–mCherry stalling reporters harbouring either K0 or K20 sequences and measured the relative fluorescence. As predicted, premature termination on the K20 motif—a hard staller—was lower in polyQ(111) cells (Fig. 7c). We propose that the protective effects of reducing translation initiation in HD are mediated by lower risk of collisions, which alleviate ribotoxic stress and premature termination.

Cell cultures

Huh7 human hepatocellular carcinoma cells were a kind gift from Raul Andino. HEK293T human embryonic kidney cells (CRL-3216) and SH-SY5Y human neuroblastoma cells (CRL-2266) were from ATCC. STHdh Q7/Q7 (CH00097) and Q111/Q111 (CH00095) mouse striatal neurons were from Coriell Institute for Medical Research. Huh7 and SH-SY5Y cells were grown in Dulbecco’s modified Eagle medium (DMEM)/F-12 1:1 medium (Thermo Fisher) supplemented with 10% foetal bovine serum (FBS), 100 units ml⁻¹ penicillin and 100 mg ml⁻¹ streptomycin. HEK293T human embryonic kidney cells and STHdh Q7/Q7 and Q111/Q111 striatal neuron cells were grown in DMEM high-glucose medium (Thermo Fisher) supplemented as above. Cells were grown at 37 °C (Huh7, 293T and SH-SY5Y) or 32 °C (STHdh) in a 5% CO₂ incubator.

Plasmids and oligonucleotides

Plasmids and oligonucleotide sequences are provided in Supplementary Table 1.

Analysis of published ribosome profiling data

gWIPS-viz⁶³ and RPFdb⁶⁴ webservers were used to search for relevant ribosome profiling datasets with coverage on Htt uORF and main coding sequence (CDS). Relevant mapped reads were downloaded from gWIPS-viz in BigWig format and visualized using Integrated Genome Viewer version 2.8.2.

Polysome profiles

A total of 1–5 × 10⁷ striatal neurons or Huh7 cells were collected by scraping in ice-cold phosphate-buffered saline (PBS), centrifuged 1,000g for 5 min at 4 °C, resuspended in PBS and centrifuged again. Cell pellets were flash frozen in liquid nitrogen. Pellets were thawed on ice and resuspended in 200 μl polysome buffer (25 mM Tris–HCl pH 7.5, 25 mM KCl, 10 mM MgCl₂, 2 mM dithiothreitol (DTT) and cOmplete EDTA-free protease inhibitor cocktail (Millipore Sigma)). Triton X-100 and sodium deoxycholate were added to a final concentration of 1% each, and the samples were incubated on ice for 20 min, passed ten times through a 26 G needle and centrifuged at 20,000g for 10 min at 4 °C to remove cell debris. For RNase-resistance assays, 100 U RNase I (Thermo Fisher) was added per 100 μg total RNA in clarified lysates. Digestion was performed for 45 min at room temperature and terminated by addition of 200 U Superase-In (Thermo Fisher). Lysates were loaded on 10–50% sucrose gradients in polysome buffer (or polysome buffer with 500 mM KCl, for RNase-treated samples) and subjected to ultracentrifugation at 36,000 rpm (average 160,030g) in an SW41.Ti swinging bucket rotor (Beckman Coulter) for 150 min at 4 °C. Equal volume fractions were collected using Gradient Station (BioComp) with continuous monitoring of ribosomal RNA (rRNA) at UV254.

qRT–PCR

To extract RNA from sucrose gradient fractions, 1 μl pellet paint (Millipore Sigma) and 500 μl phenol:chloroform:isoamyl alcohol (25:24:1) were added to each 500 μl fraction and incubated for 5 min at room temperature. After centrifugation at 12,000g for 15 min, 4 °C, the top phase was removed and subjected to another round of extraction as above. Four-hundred microlitres of the top phase was combined with 600 μl isopropanol and centrifuged at 12,000g for 30 min, 4 °C. The pellet was washed with 1 ml 75% ice-cold ethanol, air dried and resuspended in 20 μl RNase-free water. Complementary DNA was synthesized using the High Capacity cDNA Reverse Transcription Kit (Thermo Fisher), according to the manufacturer’s instructions, using 5 μl of RNA from each fraction. Quantitative real-time PCR (qRT–PCR) analysis was performed using SensiFast SYBR (BioLine) and gene-specific primers (Supplementary Table 1), according to the manufacturer’s instructions. To estimate relative abundance of specific mRNAs in each gradient fraction, each Ct value was divided by the sum of Ct values across all gradient fractions.

SDS–PAGE and immunoblotting

Cells were washed twice with ice-cold PBS and either lysed on plate with RIPA buffer (Thermo Fisher) or scraped in PBS, pelleted by centrifugation for 5 min at 1,000g, 4 °C, and resuspended in polysome buffer. Each lysis buffer was supplemented with fresh 2 mM DTT and cOmplete EDTA-free protease inhibitor cocktail. RIPA buffer was also supplemented with 50 units ml⁻¹ benzonase (Millipore Sigma) to remove DNA. For ubiquitination analysis of ribosomal proteins, 100 μM PR-619 and 5 mM N-ethylmaleimide (NEM, Selleck Chemicals) were included in the lysis buffer. For detection of phosphorylated proteins, PhosSTOP phosphatase inhibitor cocktail (Roche) was added at a ratio of one tablet per 10 ml buffer. Lysis was performed on ice for 20 min, and lysates were clarified by centrifugation for 10 min at 12,000g, 4 °C. Then, 4× Laemmli sample buffer (Bio-Rad) supplemented with fresh 10% 2-mercaptoethanol was added to a final concentration of 1×. Afterwards, 15–30 μg of each sample was resolved on 4–20% (Bio-Rad) or 12% (homemade) SDS–PAGE, and transferred to 0.2 or 0.45 μm polyvinylidene fluoride membranes pre-soaked in methanol for 30 s. Membranes were blocked with 4% bovine serum albumin (BSA; Millipore Sigma) in Tris-buffered saline supplemented with 0.1% Tween-20 (Millipore Sigma) (TBST) for 1 h at room temperature, and then probed with specific primary antibodies for 2 h at room temperature. Primary antibodies were diluted in 4% BSA/TBST supplemented with 0.02% sodium azide as follows: rabbit anti-HTT D7F7 (Cell Signaling 5656S, 1:1,000, used to detect full-length HTT), mouse anti-HTT 2B4 (Millipore Sigma MAB5492, 1:1,000, used to detect HTT-ex1), mouse anti-PolyQ MW1 (Developmental Studies Hybridoma Bank, 1:1,000), mouse anti-puromycin (Millipore Sigma MABE343, 1:2,500), mouse anti-eIF2a (Abcam ab5369, 1:1,000), rabbit anti-p-eIF2a S51 (Cell Signaling 9721, 1:1,000), rabbit anti-RPS2 (Bethyl Laboratories A303–794, 1:1,000), rabbit anti-RPS3 (Cell Signaling 9538, 1:1,000), rabbit anti-Atf4 (Cell Signaling 11815, 1:1,000), goat anti-Poliovirus VP1 (Thermo Fisher PA1–73124, 1:2,500), mouse anti-Poliovirus 3CD (homemade hybdridoma, 1:1,000), mouse anti-eIF5A (BD Sciences 611977, 1:1,000), rabbit anti-hypusine (Millipore Sigma ABS1064, 1:1,000) and rabbit anti-RPL13a (Cell Signaling 2765, 1:2,500). Secondary antibodies were diluted 1:10,000 in TBST. Western blot detection was done using ECL Plus Western Blotting Substrate (Thermo Fisher), and images were taken either by film radiography or Bio-Rad GelDoc imager. Alternatively, fluorescent western blot detection was performed using Li-Cor Odyssey Infrared imager. Densitometry analysis was performed using ImageJ version 1.52a.

Immunofluorescence analysis of HTT aggregation

Htt-ex1 Q25 or Q97 (ref. 65) was amplified by PCR and ligated into pcDNA3–GFP using NheI and XhoI. The endogenous 5′ UTR of human Htt (NM_001388492.1), either with a functional uORF (WT, ATG) or non-functional uORF (single nucleotide mutation, AAG), was synthesized by GenScript with flanking BmtI and HindIII sites and ligated into pcDNA3-mClover. For live cell imaging, 3 × 10⁵ HEK293T cells were seeded in six-well plates. At 24 h, 2 μg DNA was transfected using Lipofectamine 2000 (Thermo Fisher), according to the manufacturer’s instructions. Transfected cells were monitored once every hour starting from 12 h post-transfection, using a Zeiss Axio Vert.A1 inverted microscope. For inclusion body counting, fixed cells were used. A total of 0.8 × 10⁵ HEK293T cells were seeded on poly-L-ornithine (Millipore Sigma) pre-coated 12-mm coverslips. At 24 h, 400 ng DNA was transfected using Lipofecatmine 2000 (Thermo Fisher), according to the manufacturer’s instructions. Cells were fixed with 4% paraformaldehyde for 15 min at room temperature and washed twice with PBS for 5 min each. Cells were permeabilized with 0.1% Triton X-100 for 10 min at room temperature and washed twice with PBS. To label plasma membrane, coverslips were incubated with HCS CellMask Deep Red stain (Thermo Fisher) for 5 min at room temperature and washed twice with PBS. Coverslips were mounted with ProLong Glass Antifade Mountant with NucBlue Stain (Thermo Fisher), dried overnight and stored at −20 °C until image acquisition using a Zeiss LSM 700 laser scanning confocal microscope.

Generation of RIPA-soluble and RIPA-insoluble fractions

A total of 5 × 10⁵ HEK293T cells were plated on poly-D-lysine-coated six-well plates. The following day, cells were transfected with 0.5–2 μg of the above plasmids using Lipofectamine 2000 (Thermo Fisher), according to the manufacturer’s instructions. At 8, 24 or 48 h post-transfection, cells were washed twice with ice-cold PBS and scraped into the same buffer. Cells were pelleted by centrifugation at 300g for 5 min at 4 °C and washed again with the same buffer. Pellets were resuspended in 50 μl RIPA lysis buffer (Thermo Fisher) and triturated through a 25 G syringe for ten times. Lysates were clarified by centrifugation at 1,500g for 10 min at 4 °C. RIPA-insoluble fractions were washed with RIPA buffer twice, followed by centrifugation at 1,500g for 10 min at 4 °C. Pellets were incubated in 20 μl formic acid at 56 °C for 1 h, with mixing at 1,000 rpm, and then dried overnight using a SpeedVac. Dried pellets were resuspended in 8 M urea, 150 mM NaCl and 20 mM Tris pH 8.0 and subjected to immunoblot analysis.

NC puromycylation

A total of 2.5 × 10⁵ STHdh, Huh7 or SH-SY5Y cells in six-well plates were incubated with 1 μM puromycin (Thermo Fisher) for 15 min at 32 °C (STHdh) or 37 °C (Huh7 and SH-SY5Y), washed twice with PBS and lysed on-plate as described above. For sucrose gradients, each fraction was incubated with 1 μM puromycin for 15 min at 37 °C, followed by incubation with 7 μl StrataClean beads (Agilent) for 16 h at 4 °C with constant tumbling. Beads were pelleted by centrifugation at 1,000g for 5 min at 4 °C and resuspended in 30 μl 2× Laemmli buffer. Puromycylated samples were subjected to immunoblotting using an anti-puromycin antibody (Millipore Sigma).

Abortive termination reporters

The dual-fluorescence abortive termination reporters were a generous gift from Eric Bennett. pmGFP-P2A-K0-P2A-RFP was linearized by EcoNI and KpnI, and PCR-amplified human HTT-ex1 Q25 or Q97 was cloned in between GFP and mCherry, separated by self-cleaving P2A sites, using In-Fusion kit (Takara Bio). A total of 1 × 10⁶ STHdh cells were plated on six-well plates. The following day, cells were transfected in triplicates with 2 μg dual-fluorescence plasmids using Lipofectamine 2000 (Thermo Fisher), according to the manufacturer’s instructions. At 24 h post-transfection, cells were washed twice with ice-cold PBS, trypsinized and washed with 1% BSA in ice-cold PBS. Cells were stained with LIVE/DEAD Fixable Violet Dead Cell Stain Kit (Thermo), diluted 1:1,000 in PBS, for 30 min on ice in the dark. Cells were washed twice with 1% BSA in ice-cold PBS and filtered using a 35-μm strainer mesh before flow cytometry analysis. A total of 50,000 cells per sample were analysed using a BD LSR II Flow Cytometer. The medium of GFP and mCherry intensity and mCherry-to-GFP ratio were calculated from the live singlet population. Flow cytometry data analysis was performed using FlowJo Ver. 9.

GFP co-immunoprecipitation

A total of 3 × 10⁶ WT STHdh cells were plated in 10-cm plates. The following day, cells were transfected with 12 μg pcDNA3-Htt-ex1-GFP variants using Lipofectamine 2000 (Thermo Fisher), according to the manufacturer’s instructions. Sixteen hours post-transfection, cells were washed twice with ice-cold PBS, scraped and pelleted at 300g for 5 min at 4 °C. Pellets were lysed with immunoprecipitation (IP) lysis buffer (25 mM Tris–HCl pH 7.5, 50 mM NaCl, 10 mM MgCl₂, 5% glycerol, 1% NP40, 1 mM DTT and complete protease inhibitor) for 10 min on ice, and then passed 10× through 25 G syringes. Lysates were clarified by centrifugation at 1,500g for 10 min at 4 °C. Three-hundred micrograms of total protein and 10 μl ChromoTek GFP-Trap magnetic agarose beads (Proteintech) were used per sample. Beads were washed three times with ice-cold IP lysis buffer. Lysates were added and incubated at 4 °C overnight, with shaking. Beads were washed 3× with ice-cold IP wash buffer (50 mM Tris–HCl pH 7.5, 150 mM NaCl, 10 mM MgCl₂, 5% glycerol and 0.05% NP40). After the final wash, beads were resuspended in 1× Laemmli sample buffer and subjected to immunoblotting.

Ribosome profiling analysis

Ribosome footprints were prepared essentially as described⁶⁶. Briefly, adherent cells were washed and collected in ice-cold PBS by scraping. Cells were pelleted by centrifugation at 1,000g for 5 min at 4 °C and pellets were resuspended in lysis buffer (50 mM HEPES–KOH pH 7.5, 140 mM KCl, 5 mM MgCl₂ and 0.1% NP40) supplemented with 0.5 mM DTT and cOmplete EDTA-free protease inhibitor cocktail. Lysates were passed 10× through a 25 G needle and clarified by centrifugation at 20,000g for 10 min at 4 °C. Then, 25 U RNase I were added per 100 μg total RNA, and digestion was performed for 45 min at room temperature, and then terminated with 200 U Superase-In (Thermo Fisher). RNase-treated samples were fractionated on 10–50% sucrose gradients, and 80S monosome fractions were collected. RNA was extracted from each fraction using phenol:chloroform:isoamyl alcohol, as described above, and size selected for 16–34 nt on an 8% urea–PAGE. rRNA depletion was performed using Ribo-Zero rRNA Removal Kit (Illumina), according to the manufacturer’s instructions. All downstream steps were performed as described⁶⁶. Libraries from two independent repeats were sequenced on a HiSeq 4000 (Illumina). After demultiplexing, sequencing reads were trimmed of adapter sequences and quality filtered using cutadapt (-a CTGTAGGCACCATCAAT -m1 -q20). rRNA was removed using Bowtie2 (--un). Remaining reads were aligned to Mus musculus genome assembly GRCm38 (mm10) using Hisat2 (--trim5 1). Read count matrices were generated using featureCounts and BigWig files were generated using bamCoverage and visualized on Integrated Genome Viewer. For ribosome stall site analysis, we used the standalone version of PausePred⁶⁷ with the following parameters: window_size = 1,000, foldchange for pause = 20, read_lengths = 28,29,30, window coverage% = 10, upstream_seq and downstream_seq = 25, and offset = 12,12,12.

Generation and use of HTT-expressing virus

HTT-ex-1 Q8 or Q73 were amplified by PCR from gWizQ8 or gWizQ73 (kind gift from Donald Lo⁶⁸) using primers with flanking XhoI and EcoRI sites, and cloned into prib(+)XpA-GFP⁶⁹, encoding type 1 poliovirus Mahoney, by replacing the GFP.

Ten micrograms of plasmids were linearized with 50 U MluI for 2 h at 37 °C, followed by phenol:chloroform:isoamyl alcohol (25:24:1 v/v) extraction and in vitro transcription using MEGAscript T7 (Thermo Fisher), according to the manufacturer’s instructions. Ten micrograms of viral RNA was electroporated into 4 × 10⁶ Huh7 cells in 0.4 ml PBS, 4-mm electroporation cuvette, at 270 V/960 μF (GenePulser, Bio-Rad). Cells were grown for 24 h at 37 °C, followed by 3× freeze–thaw cycles, and the medium was centrifuged at 2,500g for 5 min, 4 °C, to remove cell debris. Clarified virus-containing media (passage 0, P0) were aliquoted and frozen at −80 °C. P0 viruses were amplified once in Huh7 cells to generate P1 virus stock, which was processed and frozen as above. P1 stock was used for all experiments.

To determine virus titres, 1 × 10⁵ Vero cells were seeded in six-well plates. The following day, tenfold serial dilutions of virus stocks/samples were prepared in media without FBS. Cells were washed with PBS, and 0.5 ml inoculum was added for 45 min at 37 °C. Two millilitres of overlay solution (1% agarose combined 1:1 v/v with 2× DMEM supplemented with 2% FBS) was added to each well and allowed to gel at room temperature. Two days later, 2 ml 2% formaldehyde was added for 1 h at room temperature. Cells were stained with 1 ml 0.1% crystal violet in 20% ethanol for 1 h at room temperature. For infections, Huh7 or SH-SY5Y cells were washed once with PBS and incubated with FBS-free DMEM supplemented with virus at a multiplicity of infection of 10. After 45 min at 37 °C, inoculum was replaced with fresh DMEM supplemented with 10% FBS.

Sample preparation for proteomic analysis

Proteins were extracted from isolated polysomes using methanol–chloroform precipitation. Four-hundred microlitres of methanol, 100 μl chloroform and 350 μl water were added sequentially to each 100 μl sample, followed by centrifugation at 14,000g for 5 min at room temperature. The top phase was removed and the protein interphase was precipitated by addition of 400 μl methanol, followed by centrifugation at 14,000g for 5 min at room temperature. Pellets were air dried and resuspended in 8 M urea and 25 mM ammonium bicarbonate (pH 7.5). Two to four micrograms of protein were subjected to reduction and alkylation with 10 mM DTT for 1 h at room temperature followed by 5 mM iodoacetamide for 45 min at room temperature, in the dark. The samples were then incubated with 1:50 enzyme-to-protein ratio of sequencing-grade trypsin (Promega) overnight at 37 °C. Peptides were desalted with μC18 Ziptips (Millipore Sigma), dried and resuspended in 10 μl 0.1% formic acid in water.

LC–MS/MS acquisition and data processing

Proteomic data were collected using Xcalibur 4.2 SP1 on LTQ-Elite (virus polysomes) and QE (RNase-digested polysomes) mass spectrometers (Thermo Fisher) coupled online to a nanoAcquity UPLC system (Waters Corporation) through an EASY-Spray nanoESI ion source (Thermo Fisher). Peptides were loaded onto an EASY-Spray column (75 μm × 15 cm column packed with 3 μm, 100 Å PepMap C18 resin) at 2% B (0.1% formic acid in acetonitrile) for 20 min at a flow rate of 600 nl min⁻¹. Peptides were separated at 400 nl min⁻¹ using a gradient from 2% to 25% B over 48 min (QE) or 220 min (LTQ) followed by a second gradient from 25% to 37% B over 8 min and then a column wash at 75% B and reequilibration at 2% B. Precursor scans were acquired in the Orbitrap analyser (QE: 350–1,500 m/z, resolution 70,000@200 m/z, automatic gain control (AGC) target 3 × 10⁶; LTQ: 300–1,800 m/z, resolution 60,000@400 m/z, AGC target 2 × 10⁶. The top ten or six (QE or LTQ) most intense, doubly charged or higher ions were isolated (4 m/z isolation window) and subjected to high-energy collisional dissociation (QE: 25 normalized collision energy (NCE); LTQ: 27.5 NCE), and the product ions were measured in the Orbitrap analyser (QE resolution: 17,500@200 m/z, AGC target 5 × 10⁴; LTQ resolution: 15,000@400 m/z, AGC target 9 × 10⁴).

Raw mass spectrometry (MS) data were processed using MaxQuant version 1.6.7.0 (ref. 70). MS/MS spectrum searches were performed using the Andromeda search engine⁷¹ against the forward and reverse human and mouse Uniprot databases (downloaded 28 August 2017 and 25 November 2020, respectively). Cysteine carbamidomethylation was chosen as fixed modification and methionine oxidation and N-terminal acetylation as variable modifications. Parent peptides and fragment ions were searched with maximal mass deviation of 6 ppm and 20 ppm, respectively. Mass recalibration was performed with a window of 20 ppm. Maximum allowed false discovery rate (FDR) was <0.01 at both the peptide and protein levels, based on a standard target–decoy database approach. The ‘calculate peak properties’ and ‘match between runs’ options were enabled.

All statistical tests were performed with Perseus version 1.6.7.0 using either ProteinGroups or Peptides output tables from MaxQuant. Potential contaminants, proteins identified in the reverse dataset and proteins identified only by site were filtered out. Intensity-based absolute quantification (iBAQ) was used to estimate absolute protein abundance. Two-sided Student’s t-test with a permutation-based FDR of 0.01 and S0 of 0.1 with 250 randomizations was used to determine statistically significant differences between grouped replicates. Categorical annotation was based on Gene Ontology Biological Process (GOBP), Molecular Function (GOMF) and Cellular Component (GOCC), as well as protein complex assembly by CORUM.

Cell viability

A total of 1 × 10⁴ STHdh cells were seeded per well in a black-wall, clear-bottom 96-well plate. At 24 h, Tg (AG Scientific) was added for another 24 h. Cell viability was assayed using CellTiter-Glo (Promega), according to the manufacturer’s instructions. Luminescence was recorded using a BMG Labtech CLARIOstar Microplate Reader.

Statistics and reproducibility

No statistical methods were used to pre-determine sample sizes. Sample sizes (n) reflect independent replicates, except when otherwise indicated in the figure legends. For all proteomic analyses of R6/2 mouse brains (Hosp et al., related to Fig. 4e,f and Extended Data Figs. 4a–c and 5a,c–e), data were acquired from four, three and three WT mice and four, three and four R6/2 mice at weeks 5, 8 and 12, respectively. Data were acquired separately from four different brain regions of each animal at each age. Separate areas of the brain are plotted as independent datapoints. No data were excluded from the analyses, and the experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.