RNA translation goes something like this: Ribosomes glom onto a cap at the 5' end of a neatly spliced mRNA strand, scan along until they bump into a well-positioned AUG, then start spooling off protein. But rules are there to be broken. Despite being buried in an intron and having no start codon, hexanucleotide expansions in the C9ORF72 gene are somehow translated into the dipeptide repeats (DPRs) implicated in ALS/FTD. Now, a trio of papers in Nature Communications proposes mechanisms for this so-called RAN (repeat-associated non-AUG) translation. Though their authors came to somewhat different conclusions about the particulars of the process, cell stress emerged as a common trigger for this atypical mode of protein synthesis. In particular, phosphorylation of the major translation initiation factor, eIF2α, turned on translation of DPRs, much as it does stress response genes.

  • Three studies describe how dipeptide repeats are translated from an intron in the C9ORF72 gene.
  • Two report that a 5' cap is crucial for translation from non-AUG start codons; another finds the cap dispensable.
  • All three agree that stress responses spur DPR translation.

One of the studies, led by Clotilde Lagier-Tourenne of Massachusetts General Hospital in Boston and Franck Martin of CNRS in Strasbourg, reported that translation of expanded C9ORF72 starts out with the ribosome finding the 5' cap and subsequently scanning for AUGs. Things get a little wonky after that, as translation commences at a CUG codon, and ribosomes skip from one reading frame to another to produce three different DPRs. Another study led by Peter Todd of the University of Michigan in Ann Arbor came to similar conclusions regarding the cap-dependency of the process. He added that the cellular stress response promoted initiation at the CUG and potentially other non-AUG codons, and that the dipeptides, in turn, also ramped up the stress response. In a third paper, Shuying Sun and colleagues at Johns Hopkins University in Baltimore also implicated stress in the translation of DPRs; however, they found that DPRs could be translated from an excised intron lacking a 5' cap.

“Despite some major discrepancies, the current papers have unraveled important aspects of the process [of RAN translation] in the context of C9ORF72,” wrote Ludo Van Den Bosch and Bart Swinnen of KU Leuven in a joint comment to Alzforum.

Stressed Out.

Stress granules (green) emerge in HEK293 cells (nuclei are blue) expressing 70 G4C2 repeats. [Courtesy of Green et al., Nature Communications, 2017.]

Hexanucleotide expansions within the first intron of the C9ORF72 gene are the most common genetic cause of ALS/FTD, where anywhere from 30 to thousands of copies of the GGGGCC sequence can wreak havoc on neurons. The field was shocked to learn that despite their location within the supposedly noncoding intron 1 of the gene, the repeats were in fact being translated into aggregation-prone dipeptide repeats (DPRs) (Feb 2013 news). What’s more, the DPRs were translated from three different reading frames, and from both the sense and antisense strands, making for a total of five different polypeptides: glycine-alanine (GA), glycine-arginine (GR), proline-arginine (PR), proline-alanine (PA), and glycine-proline (GP). RAN translation was evoked to explain DPR production, as it does for repeat peptides in other neurodegenerative diseases, including Huntington’s and spinocerebellar ataxias (Nov 2015 conference news). Exactly how this noncanonical translation happens, how it is regulated, and whether it might serve as a therapeutic target are all focuses of intense research.

A Nucleotide Away from Normal
Lagier-Tourenne’s study, published January 11, investigated how translation can occur in multiple reading frames. First author Ricardos Tabet and colleagues used DNA constructs containing 66 G4C2 repeats, which previously had been shown to produce peptides from three forward reading frames in cultured cells and in transgenic mice (Nov 2013 news on Gendron et al., 2013, and May 2015 news). Inserted into an expression vector, the 66 repeats were preceded by 113 nucleotides from the human C9ORF72 intron 1, which included no AUG start codon. To compare the efficiency of translation in the three forward reading frames from this single construct, the researchers inserted a protein tag—HA, His, or FLAG—downstream of the repeats in each reading frame. In rabbit reticulocyte lysate, a commonly used cell-free translation system, the researchers found that translation of poly-GA (in the +1 reading frame) was far more efficient than poly-GP (+2) and poly-GR (+3). Using uncapped mRNA decreased the translation of poly-GA fivefold, while poly-GP and poly-GR became undetectable. 

Atypical Translation. In canonical translation (a), ribosomes, tRNA-Met, and other proteins scan mRNA until they encounter the AUG start codon. The same occurs in RAN translation (b), except it initiates at a CUG instead. Subsequent frameshifting may produce proteins products from other reading frames. [Courtesy of Tabet et al., Nature Communications, 2018.]

Using a series of mutated constructs, the researchers found that translation of all three peptides, regardless of reading frame, started at the same CUG codon located 24 nucleotides upstream of the repeat expansion. Translation depended on canonical translation machinery, including a cadre of cap-binding proteins and an initiator tRNA loaded with methionine, and the cap itself. Tabet determined that following initiation at this CUG, the translating ribosomes frameshifted, hopping over a nucleotide or two to generate peptides in the +2 and +3 reading frames. Ribosomal frameshifting can be induced by structural anomalies in mRNA, including the G-quadruplexes formed by these hexanucleotide expansions (Nov 2013 conference news). 

While these experiments were not designed to detect translation of the reverse strand, Lagier-Tourenne said that antisense mRNA strands can be capped, too, so it is possible their translation is subject to similar rules. However, she noted that in human samples, antisense DPRs are exceedingly rare.

Stress Makes Things Less Stringent
As reported already December 8, Todd and colleagues also compared the efficiency of RAN translation in different reading frames. Using reticulocyte lysates and cellular models, they found, much as Tabet and colleagues had, that the ribosome more efficiently translated poly-GA than it did poly-GP, which in turn edged out production of poly-GR. PolyGA lies in-frame with the CUG codon. Only by mutating this CUG was Green able to shift expression of poly-GA in favor of poly-GP. Hence, Todd thinks that frameshifting does not fully explain translation of the three DPRs. Instead, he believes the ribosome may occasionally skip over the CUG for some reason, and then initiate translation at another codon in a different reading frame. However, he acknowledged to Alzforum that the DNA constructs he used were not designed to detect ribosomal frameshifting, and that it could play a role. In agreement with Lagier-Tourenne, Todd and colleagues reported that translation initiation was strongly cap-dependent.

Green and colleagues next set out to understand how cellular stress affected this cap-dependent, non-AUG translation. Previous studies have reported that the integrated stress response (ISR), which switches on in reaction to accumulating misfolded proteins, viral infection, and other insults, affects the stringency and kinetics of translation initiation. Specifically, the ISR triggers the phosphorylation of elongation initiation factor 2α (eIF2α), which favors noncanonical translation mechanisms, including cap-independent translation and non-AUG starts. Using stressors, eIF2α phosphorylation inhibitors, and other molecular tricks, the researchers found that stress-induced eIF2α phosphorylation suppressed global translation in HEK293 cells and in primary rat cortical neurons. However, these treatments had the opposite effect on expression of the DPRs, boosting expression of poly-GA, GP, and GR peptides in the cells. The same was true when the researchers analyzed translation of CGG repeats associated with Fragile X-associated tremor/ataxia syndrome.

Feed-Forward Frenzy.

The integrated stress response inhibits global translation but promotes RAN translation, which promotes stress granule formation and reduces global translation. The ISR also directly boosts granules and saps translation. [Courtesy of Green et al., Nature Communications, 2017.]

Finally, Green and colleagues uncovered a connection between RAN translation and stress granules. These membraneless organelles have been implicated in multiple neurodegenerative diseases, and may serve as a hotbed for protein aggregation (Oct 2015 newsOct 2016 news; Nov 2017 news). The DPRs boosted eIF2α phosphorylation and triggered the formation of stress granules. In sum, the researchers proposed a model in which cellular stress promotes toxic RAN translation, unleashing DPRs that induce the formation of stress granules and reduce global translation, thus implicating them in a feed-forward loop that promotes neurodegeneration.

“The stimulation of DPR translation by the integrated stress response is an exciting finding, and the induction of a feed-forward response in which both the stress response and DPR translation induce each other may prove to be an important step in the pathogenesis of C9ORF72-mediated disease,” commented Brian Freibaum of St. Jude Children’s Research Hospital in Memphis, Tennessee. 

Caps Off to C9ORF72!
In the third paper, published January 4, researchers led by Sun also finger stress-driven eIF2α phosphorylation in RAN translation, but add a surprising twist. First author Weiwei Cheng and colleagues designed bicistronic constructs that allowed them to compare the efficiency of cap-dependent versus cap-independent translation of the repeats, as expressed in HeLa cells. Essentially, they found that cap-independent translation did occur, albeit 20–30 percent less efficiently than cap-dependent translation. They discovered that eIF2α phosphorylation induced by the ISR boosted cap-independent translation; this could be triggered by cytoplasmic accumulation of another protein linked to ALS/FTD—none other than TDP-43. 

Cheng and colleagues wondered whether translation of the repeats occurred following excision of intron 1 of C9ORF72. Notably, introns lack 5' caps and form a lariat structure when spliced out of RNA; this typically triggers their degradation in the nucleus. However, it appears that these RNA clippings sometimes make it out into the cytoplasm. Indeed, when the researchers expressed constructs containing the repeat-bearing intron flanked by the first two exons of C9ORF72, they found that not only did splicing occur, but the intron made its way into the cytoplasm, where it hooked up with ribosomes that made poly-GA and poly-GP. Again, eIF2α phosphorylation enhanced this translation.

Toxic Trimmings.

Dipeptide repeats are translated from a cap-less intron following its excision. Stress and eIF2α phosphorylation promotes this form of RAN translation. [Courtesy of Cheng et al., Nature Communications, 2018.]

Sun proposed that the structure of the excised, repeat-containing intron somehow facilitates cap-independent translation. She pointed out that although ribosomes translate capped transcripts more efficiently than uncapped, excision of C9ORF72 intron 1 is far more common than its aberrant retention, and thus the excised intron may be the most common C9ORF72 repeat substrate available for translation in the brain. She plans to investigate how the excised intron manages to escape degradation in the nucleus and gain entry into the cytoplasm for translation.

Lagier-Tourenne, on the other hand, thinks efficiency of translation is the most important factor, and therefore cap-dependent translation predominates (Niblock et al., 2016). However, when she tested this in the cell-free system, she found yet another twist. An upstream open reading frame with two consecutive stop codons in intron 1, just before the expansion, snagged scanning ribosomes before they could reach the repeats, nearly abolishing RAN translation. She told Alzforum that this does not necessarily imply that the excised intron hypothesis trumps the cap-dependent translation idea, because small upstream ORFs are established regulatory motifs. Cellular stress, which inhibits AUG-initiated translation, and even neuronal activation, might suppress translation of the uORF, enabling RAN, she said.

Todd thinks all of these mechanisms could happen in cells to some extent. “There is a lot of noise in nature, and this is especially true of translation,” he said. Lagier-Tourenne agreed, noting that because the studies each used different DNA constructs and expression systems, a direct comparison is difficult. Todd noted that despite different conclusions on cap dependence, both his and Sun’s studies converged on the role of the stress response in promoting RAN translation. “Stress both weakens codon fidelity, and enhances cap-independent translation,” he said.

Van Den Bosch and Swinnen also focused on this convergence. “The observation that stress induces a feed-forward loop enhancing RAN translation is extremely interesting and fits in a broader perspective,” they wrote. “It suggests a self-sustaining mechanism in which stress induced by either arginine-containing DPRs (PR and GR), or by the repeat RNA itself, fuels a vicious circle,” they added (Swinnen et al., 2018). 

Todd believes something as essential as translation likely makes a poor therapeutic target. “But intervening in the stress response could have therapeutic legs,” he said. Indeed, Sun reported that small-molecule inhibitors of PRKR-like ER kinase (PERK), which phosphorylates eIF2α in response to accumulating unfolded proteins, substantially blocked RAN translation. These compounds have been tested in animal models of neurodegenerative disease, including ALS/FTD, but turned out to be toxic (Oct 2013 news). Researchers have begun screening for less-toxic leads (Halliday et al., 2017). High levels of activated PERK and p-eIF2α have been reported in AD and PD brains, as well (Hoozemans et al., 2007; Hoozemans et al., 2009). 

Trying another approach, Lagier-Tourenne and colleagues took a hint from the uORF that suppresses RAN, generating a series of antisense oligonucleotides (ASOs) to adhere to the mRNA upstream from the repeats, to block scanning ribosomes. These constructs hindered RAN from transcripts with no uORF. The researchers proposed that blocking ribosomal scanning in this way could serve as a therapeutic strategy to prevent translation of the DPRs. ASOs that suppress proteins with repeat expansions reduce toxicity in animal models of ataxia and ALS (Apr 2017 news). In people with Huntington’s disease, this strategy curbs production of expanded huntingtin, and nusinersen, an ASO that corrects aberrant splicing of the spinomuscular atrophy gene, was approved by the U.S. Food and Drug Administration in 2016 (Dec 2017 newsNov 2016 news). 

All three study leaders emphasized that it will be crucial to examine how RAN translation actually occurs in neurons from C9ORF72 repeat expansion carriers. Lagier-Tourenne proposed monitoring translation of the DPRs after using CRISPR to mutate the putative CUG start codon in induced pluripotent stem cell (iPSC)-derived neurons from patients. Freibaum added that it would be interesting to visualize the 5' cap on the G4C2 transcript in such cells, while Sun aims to investigate the transport and translation of the excised intron.—Jessica Shugart 


  1. I consider all three papers to be significant contributions to our understanding of the role of DPR translation in ALS. The stimulation of DPR translation by the integrated stress response is an exciting finding, and the induction of a feed-forward response in which both the stress response and DPR translation induce each other may prove to be an important step in the pathogenesis of C9ORF72-mediated disease.

    While the primary discrepancy between the publications is whether DPR translation requires a 5' cap, all three publications show that cap-dependent translation is more efficient than cap-independent translation of the DPRs. It would be very interesting to see if the 5' cap could be visualized on the G4C2 transcript from IPS neurons derived from C9 patients. The three publications agree that the GA dipeptide is more efficiently translated than either the GP or GR dipeptides. Additionally, it would be interesting to know the relative translation rates of the antisense DPRs in relation to the sense DPRs and whether the antisense DPRs are also stimulated by the integrated stress response. 

  2. Very interesting that three research groups investigated the intriguing mechanism behind non-ATG-mediated translation of the hexanucleotide repeats in C9ORF72, a process also known as RAN translation. We find the observation by Cheng et al. and Green et al. that stress induces a feed-forward loop to enhance RAN translation extremely interesting and it fits in a broader perspective. These insights suggest a self-sustaining mechanism by which stress, induced by arginine-containing DPRs (PR and GR), which can phase-separate, or by the repeat RNA itself (Swinnen et al., 2018), fuels a vicious circle. This stress-induced process is mediated by phosphorylation of the alpha subunit of eukaryotic initiation factor-2 (eIF2). Accordingly, the formation of DPR-induced stress granules is dependent on eIF2α phosphorylation (Boeynaems et al., 2017). The immediate consequence of stress granule formation in general is that canonical cap-dependent translation is downregulated and that non-canonical cap-independent translation (e.g., making use of internal ribosomal entry sites, or IRESes, and non-AUG start sites) is increased in order to maintain proteostasis. Despite some major discrepancies, the current papers have unraveled important aspects of this process in the context of C9ORF72. Most consistently, the C9ORF72 repeat expansion undergoes cap-independent RAN translation, however with a 70–90 percent lower efficiency compared to cap-dependent RAN translation. Also, a near-cognate CUG codon upstream of the repeat, which is in frame with poly-GA, seems to be important for RAN translation.

    A number of questions remain. It still needs to be investigated how antisense repeats are translated into DPRs (PA and PR), although recent pathological data indicate that these are less abundant and that especially GR, translated from the sense strand, correlates with neurodegeneration and co-localizes with TDP-43 (Saberi et al., 2017). In addition, the exact nature of the RNA species present in patients who will evince RAN translation is still unknown. In relation to this, it is intriguing that the RAN translation process is mostly cap-dependent according to Green et al. and Tabet et al. and proposed to be cap-independent, although significantly enhanced by capping, in Cheng et al. Cap-independent translation would not only be expected based on the fact that the hexanucleotide repeats are spliced out of an intron, it would also fit better with the hypothesis that translation of non-capped RNA is still ongoing under stress conditions, while cap-dependent translation is silenced. Future research is needed to clarify this issue. Also, as only poly-GA is in frame with the near-cognate CUG codon, it is not clear how the other sense-derived DPRs are produced. One possibility is frameshifts during translation that subsequently generate poly-GP and poly-GR, which might explain why GA is the most abundant DPR in postmortem tissue. However, more complicated mechanisms could be at play as removing the CUG in the GA frame has variable effects on poly-GP and poly-GR production, depending on the cellular context.

    It is clear that if a “gain-of-function” mechanism is responsible for initiating the disease, prevention of the formation of repeat-containing RNAs is the way to go. This strategy not only prevents the formation of potentially toxic RNAs, but also the translation of the different DPRs.

    Ludo van den Bosch 


    . Phase Separation of C9orf72 Dipeptide Repeats Perturbs Stress Granule Dynamics. Mol Cell. 2017 Mar 16;65(6):1044-1055.e5. PubMed.

    . Sense-encoded poly-GR dipeptide repeat proteins correlate to neurodegeneration and uniquely co-localize with TDP-43 in dendrites of repeat-expanded C9orf72 amyotrophic lateral sclerosis. Acta Neuropathol. 2018 Mar;135(3):459-474. Epub 2017 Dec 1 PubMed.

    . A zebrafish model for C9orf72 ALS reveals RNA toxicity as a pathogenic mechanism. Acta Neuropathol. 2018 Mar;135(3):427-443. Epub 2018 Jan 4 PubMed.

Make a Comment

To make a comment you must login or register.


Alzpedia Citations

  1. C9ORF72

News Citations

  1. RNA Twist: C9ORF72 Intron Expansion Makes Aggregating Protein
  2. Unexpected Polypeptides Pop Up in Huntington’s Disease
  3. Sense, Antisense: C9ORF72 Makes Both Forms of RNA, Peptides
  4. First C9ORF72 Mice Mimic Key Pathology, Behavior
  5. Researchers Revel in C9ORF72 Advances at RNA Symposium
  6. Do Membraneless Organelles Host Fibril Nucleation?
  7. ALS Research ‘Gels’ as Studies Tie Disparate Genetic Factors Together
  8. Stress Granule Protein Stabilizes Tau Oligomers, Hastens Neurodegeneration
  9. PERKing Up Protein Synthesis May Prevent Neurodegeneration
  10. Two For One? ASOs for Ataxin Allay ALS and SCA2 in Mice
  11. Antisense Oligonucleotide Squelches Huntingtin Protein in Phase 1/2a Trial
  12. Positive Trials of Spinal Muscular Atrophy Bode Well for Antisense Approach

Paper Citations

  1. . Antisense transcripts of the expanded C9ORF72 hexanucleotide repeat form nuclear RNA foci and undergo repeat-associated non-ATG translation in c9FTD/ALS. Acta Neuropathol. 2013 Oct 16; PubMed.
  2. . Retention of hexanucleotide repeat-containing intron in C9orf72 mRNA: implications for the pathogenesis of ALS/FTD. Acta Neuropathol Commun. 2016 Feb 25;4:18. PubMed.
  3. . A zebrafish model for C9orf72 ALS reveals RNA toxicity as a pathogenic mechanism. Acta Neuropathol. 2018 Mar;135(3):427-443. Epub 2018 Jan 4 PubMed.
  4. . Repurposed drugs targeting eIF2α-P-mediated translational repression prevent neurodegeneration in mice. Brain. 2017 Jun 1;140(6):1768-1783. PubMed.
  5. . Activation of the unfolded protein response in Parkinson's disease. Biochem Biophys Res Commun. 2007 Mar 16;354(3):707-11. Epub 2007 Jan 17 PubMed.
  6. . The unfolded protein response is activated in pretangle neurons in Alzheimer's disease hippocampus. Am J Pathol. 2009 Apr;174(4):1241-51. PubMed.

Further Reading

Primary Papers

  1. . CUG initiation and frameshifting enable production of dipeptide repeat proteins from ALS/FTD C9ORF72 transcripts. Nat Commun. 2018 Jan 11;9(1):152. PubMed.
  2. . RAN translation at C9orf72-associated repeat expansions is selectively enhanced by the integrated stress response. Nat Commun. 2017 Dec 8;8(1):2005. PubMed.
  3. . C9ORF72 GGGGCC repeat-associated non-AUG translation is upregulated by stress through eIF2α phosphorylation. Nat Commun. 2018 Jan 4;9(1):51. PubMed.