How Villains are Made: The Translation of Dipeptide Repeat Proteins in C9ORF72-ALS/FTD

Abstract

A hexanucleotide repeat expansion in the C9ORF72 gene is the most common genetic alteration associated with amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). These neurodegenerative diseases share genetic, clinical and pathological features. The mutation in C9ORF72 appears to drive pathogenesis through a combination of loss of C9ORF72 normal function and gain of toxic effects due to the repeat expansion, which result in aggregation prone expanded RNAs and dipeptide repeat (DPR) proteins. Studies in cellular and animal models indicate that the DPR proteins are the more toxic species. Thus, a large body of research has focused on identifying the cellular pathways most directly impacted by these toxic proteins, with the goal of characterizing disease pathogenesis and nominating potential targets for therapeutic development. The preventative block of the production of the toxic proteins before they can cause harm is a second strategy of intense focus. Despite the considerable amount of effort dedicated to this prophylactic approach, it is still unclear how the DPR proteins are synthesized from RNAs harboring repeat expansions. In this review, we summarize our current knowledge of the specific protein translation mechanisms shown to account for the synthesis of DPR proteins. We will then discuss how enhanced understanding of the composition of these toxic effectors could help in refining disease mechanisms, and paving the way to identify and design effective prophylactic therapies for C9ORF72 ALS-FTD.

1. Introduction

The C9ORF72 gene, located on the short arm of chromosome 9, is in a susceptibility locus for both amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD) (Vance et al., 2006). The underlying mutation was identified in 2011 as a hexanucleotide GGGGCC (G₄C₂) expansion in the first intron of the gene, which lies between non-coding exons 1a and 1b (DeJesus-Hernandez et al., 2011; Renton et al., 2011). In healthy individuals there are typically less than 30 G₄C₂ repeats, while individuals with the mutation may have from several hundred to thousands of these repeats. The hexanucleotide repeat expansion of the C9ORF72 gene is predominantly found in the Caucasian population and is considered to be a major genetic risk factor for both ALS and FTD. Accordingly, the mutation accounts for about 30 – 40% of familial cases of ALS and FTD and approximately 4 – 8% of sporadic cases of ALS and FTD (Majounie et al., 2012; van der Ende et al., 2021). In their pure forms, these two neurological diseases are characterized by degeneration of distinct brain regions: primarily of the motor cortex and upper and lower motor neurons in ALS and the prefrontal cortex in FTD. However, overlap of symptoms is observed in 15–50% of individuals initially diagnosed with only one of the two diseases (van der Ende et al., 2021). Knowing how the hexanucleotide repeat expansion causes neurodegeneration will be of instrumental importance for the development of effective therapies.

Three potential pathogenic mechanisms have been implicated for the G₄C₂ repeat expansion: (1) loss of function caused by downregulation of C9ORF72 protein expression; (2) toxic gain of function of sense and antisense RNAs containing the expanded repeats forming RNA foci that sequester important RNA-binding proteins; and (3) production of dipeptide repeat (DPR) proteins through translation of the sense and antisense RNAs containing expanded repeats. Translation of the repeat-containing sense strand RNA produces poly-glycine-alanine (poly-GA), poly-glycine-arginine (poly-GR) and poly-glycine-proline (poly-GP), and translation of the antisense strand generates poly-proline arginine (poly-PR), poly-proline-alanine (poly-PA) and poly-proline-glycine (poly-PG or poly-GP). The production and consequences of these six distinct DPR proteins are areas of intense research.

Here we discuss the evidence that underpins our current understanding of the translational mechanisms involved in the generation of DPR proteins. We will summarize the literature describing how different factors of the translation machinery impact the production of these proteins. Lastly, we will present a description of how the different DPR proteins are thought to contribute to neurodegeneration in C9ORF72 ALS-FTD.

2. Translational Mechanisms Associated with the C9ORF72 Repeat Expansion

2.1. The RNA Template for DPR Protein Production

The human chromosome 9 open reading frame 72 (C9ORF72) gene encodes two non-coding exons (1a and 1b) and 10 coding exons (from 2 to 11) and can be transcribed into one of three variants (Variants 1 – 3). Variants 1 and 3 (V1, NM_145005; V3, NM_001256054) mRNAs start on exon-1a, while the Variant 2 transcript (V2, NM_018325) starts on exon-1b. Since the hexanucleotide G₄C₂ expansion is found in the intron between exons-1a and −1b, V2, the major isoform, does not contain the repeat expansion. In contrast, V1 and V3 transcripts harbor the expanded repeats that can be translated into DPR proteins. Introns by their very definition are normally spliced out of mRNAs and do not get translated into protein. However, translation of RNAs containing expanded G₄C₂ repeats into DPR proteins has been reported in studies using patient-derived samples (Ash et al., 2013, Mori et al., 2013, Zu et al., 2013). A fundamental question in the field is therefore which RNAs containing the expanded repeats serve as template for DPR protein production. RNAs containing unspliced intron-1, spliced intron-1 RNA itself, and aborted RNA transcripts have all been proposed as the possible templates for translation that generate sense DPR proteins.

RNAs containing unspliced intron-1.

Niblock and colleagues proposed that defects in the splicing of intron-1 could result in the intron being retained in an otherwise mature mRNA (Niblock et al., 2016). Polyadenylated RNA species with an enlarged 5’- Untranslated region (UTR) containing the G₄C₂ repeats, the result of intron-1 retention, with the remaining downstream exons correctly spliced out were found in lymphoblast cell lines and brain samples from C9ORF72 G₄C₂ expansion carriers. Interestingly, retention of intron-1 was not dependent on the presence of the expanded G₄C₂ repeats as this was also observed for the wild-type allele (Niblock et al., 2016). Nevertheless, higher levels of RNA with the retained intron were detected in the frontal cortex of expansion carriers as compared to control individuals, suggesting that the long repeat expansion may further inhibit intron-1 splicing. It was subsequently reported that GC-rich intronic microsatellite expansions such as the G₄C₂ in C9ORF72 are likely to be localized near splice sites and selectively cause retention of their host intron (Sznajder et al., 2018). More recent studies have reported the retention of intron-1, as well as intron-2, in iPSC lines derived from C9ORF72 expansion carriers (Ben-Dor et al., 2021). An important mechanistic implication for the accumulation of mature mRNA transcripts containing intron-1 is that export of these RNA species to the cytoplasm would enable their availability for translation into DPR proteins. In line with this idea, an upstream Open Reading Frame (uORF) near the end of C9ORF72 exon-1 was shown to influence DPR protein synthesis (Tabet et al., 2018; van ‘t Spijker et al., 2021, and discussed in section 2.2.4). This finding suggests that unspliced RNA containing the repeat expansion is a source for DPR proteins. Examination of C9ORF72 RNAs in pelleted polysomes isolated from C9ORF72 iPSCs further supports the conclusion that the unspliced RNA is a primary substrate for translation of RNAs containing the G₄C₂ repeat expansion (van ‘t Spijker et al., 2021).

RNAs of spliced circular intron-1.

Independent of the existence of unspliced/retained intron-1 RNA transcripts, most studies find that the majority of the repeat-containing intron-1 RNA is properly spliced out of the mRNA, yet is detectable in the cytoplasm (Niblock et al., 2016; Cheng et al., 2018; van ‘t Spijker et al., 2021), raising the possibility that the spliced intron-1 may also serve as a template for translation. To investigate this potential mechanism, Cheng and colleagues designed a biscistronic splicing reporter in which intron-1 with 70 G₄C₂ repeats was placed in the context of exons-1a and −2 and thus should be correctly spliced out (Cheng et al., 2018). They found that the spliced RNAs accumulated in the cytoplasm to a higher extent and were more likely to be associated with translating ribosomes than unspliced pre-mRNAs (Cheng et al., 2018). An important caveat is that the reporter strategy results in the spliced intron RNA also containing the Nano-luciferase coding sequence downstream of the G₄C₂ repeats. Thus, it is unclear whether that coding sequence may be influencing the localization and translatability of the host RNA. In Wang et al. the group went one step further and used single molecule labeling of both the intron-1 and exon RNAs in cells expressing a splicing reporter and in fibroblasts from C9ORF72 expansion carriers (Wang et al., 2021). In both cell systems, the majority of signals from introns were found in the cytoplasm and did not colocalize with the exon signals, implying the introns were spliced out and not being exported to the cytoplasm in unspliced pre-mRNAs. Consistent with this, the spliced introns appeared to be in a circular form, likely resulting from a stable lariat structure that escaped debranching (Wang et al., 2021). In an elegant extension of their experiments, Wang et al. showed by labelling the active translation sites in addition to the intron-1 and exon RNAs that the circular spliced intron is the template for protein synthesis in cells expressing a reporter construct. An important next step will be to determine whether the same process is at work in cells and tissues from C9ORF72 expansion carriers.

RNAs from aborted transcripts.

Other RNA species containing the G₄C₂ repeats that could contribute to translation of DPR proteins are aborted or truncated transcripts. Using an in vitro transcription assay, Haeusler et al. found that RNA polymerase becomes impaired in the region encompassing the repeats. The presence of unique structural features such as R-loops and G-quadruplexes formed by the G₄C₂ repeats are postulated to be responsible for this transcriptional impairment (Haeusler et al., 2014). The authors found an accumulation of abortive transcripts containing the G₄C₂ repeats, with concomitant loss of full-length transcripts (Haeusler et al., 2014). Specialized elements of the transcriptional machinery, such as SUPT4, AFF2 or the PAF1 complex, can promote RNA polymerase to successfully transcribe through the long G₄C₂ repeat expansions (Kramer et al., 2016, Yuva-Aydemir et al., 2019, Goodman et al., 2019). As these were absent in the in vitro assay, additional studies will be important to confirm that these species are produced in cellular contexts, and to what extent they accumulate in disease tissue.

While determining the exact template(s) for DPR protein production could help to elucidate the translation mechanism(s) at play and identify the specific factors involved that may be therapeutically targeted, a challenge is that the template sources may change over time or in response to external factors such as different cellular stresses. Additionally, the different templates may not be mutually exclusive and could be contributing to DPR translation in the same cell at the same time. Additional experiments using C9ORF72 expansion carrier-derived cells and tissue are necessary to further investigate under which experimental conditions RNAs containing the repeat expansion are substrates for translation.

2.2. Initiation of Translation of the C9ORF72 Repeat Expansion

2.2.1. Cap-dependency

The first step in the initiation of canonical translation is the binding of the eukaryotic initiation factor 4E (eIF4E) to the m7G cap at the 5’ end of the mRNA. An open question is whether this step is required for the initiation of translation of the C9ORF72 G₄C₂ repeat expansion. Green et al. found that the translation initiation of the repeat expansion is cap-dependent (Green et al., 2017). A reporter with a m⁷G-cap analog (A-cap), which cannot bind to eIF4E, was found to express far less DPR protein compared to that of the normal m⁷G-cap reporter, indicating that the translation of the repeat expansion depends on the binding of eIF4E to the cap-end of the RNA. In line with this, a study by Tabet et al. reported that translation of the G₄C₂ repeats requires an initiator Met-tRNA (Tabet et al., 2018). Met-tRNA is part of the 43S preinitiation complex (PIC), which is recruited to the cap-bound eIF4F complex during canonical translation. However, initiation without cap-binding may also require the Met-tRNA (Haizel et al., 2020).

In contrast, Cheng et al. demonstrated that the translation of the C9ORF72 G₄C₂ repeat expansion is not cap-dependent (Cheng et al., 2018). This was shown using a bicistronic reporter system in which firefly luciferase was translated only via a canonical 5’ cap and 70 G₄C₂ repeats fused to nano luciferase were translated only through a cap-independent manner. Because the nano luciferase activity produced was higher than the firefly luciferase signal produced, the authors concluded that the G₄C₂ repeat expansion can be translated in a cap-independent fashion. Further support for divergence in cap dependency was shown using siRNAs targeting eIF4E to investigate the role of this factor in the translation of the G₄C₂ repeat expansion in the second ORF. While eIF4E siRNA reduced translation from the first cistron containing the firefly luciferase, the translation from the second cistron containing the nano luciferase was not affected, indicating that G₄C₂ repeat translation is not dependent at least on the cap-binding initiation factor eIF4E levels. More recently, van ‘t Spijker et al. investigated cap-dependency by inserting a hairpin at the start of an RNA construct containing 70 G₄C₂ repeats fused to nano luciferase. The reporter includes a firefly luciferase ORF following exon 2 of the C9ORF72 gene as control. Translation of firefly luciferase was reduced significantly by the hairpin as would be expected for canonical cap-dependent translation. However, nano luciferase translation remained unchanged, indicating the translation of the G₄C₂ repeat expansion was not dependent on cap-binding that was impacted by the hairpin insertion (van ‘t Spijker et al., 2021).

The apparently conflicting results between the studies of Green et al. and those of Cheng et al. and van ‘t Spijker et al. may result from differences in the experimental models employed. Experiments were performed with different reporter strategies, the interrogation of cap-binding dependency was explored by different approaches and the process inhibited by different methods. While Green et al. performed their experiments in Rabbit Reticulocyte Lysates (RRL), The experiments of Cheng et al., Tabet et al., and van ‘t Spijker et al. were carried out in cell lines expressing the reporter constructs. Resolution of the disparate conclusions may thus potentially be achieved if cross-validation of the strategies is applied among the groups’ chosen experimental systems, where feasible.

One additional initiation mechanism that has yet to be tested for involvement in the translation of the G₄C₂ repeat expansion is that of cap-independent translation enhancers (CITE), an alternative to canonical translation initiation (Shatsky et al., 2018). Initiation on a CITE is dependent on binding of the initiation factor eIF4G and subsequent scanning of the 5’ RNA end without a requirement for binding of the m7G cap (Kraft et al., 2013). Further investigation into the first steps of translation initiation on the G₄C₂ repeat expansion may elucidate the mechanisms involved. Since both cap-dependent and cap-independent translation may rely on different initiation factors, the identification of such factors may provide critical insight into how translation initiates for the intronic C9ORF72 repeat expansion.

2.2.2. Initiation Factors

A large group of eukaryotic initiation factors and ribosomal proteins are required to initiate translation through the canonical mechanism. As described above, the first step in the process is the binding of eIF4E to the m⁷G cap at the 5’ end of the mRNA. eIF4E is part of the eIF4F complex, which also consists of the eIF4G, eIF4A and eIF4B. Once bound to the cap, the eIF4F complex recruits the ribosomal 43S PIC, which consists of eIF3, eIF1 and eIF1A, as well as the eiF2 GTP with the Met-tRNA. Using cellular models, the role of several of these proteins has been investigated for potential involvement in the translation mechanism for the C9ORF72 G₄C₂ repeat expansion. A number of initiation factors appear to have no role in G₄C₂ repeat expansion translation: knock down of both eIF4E (Cheng et al., 2018) and eIF3D (van ‘t Spijker et al., 2021) show no effect on its translation. In contrast, other translation initiation factors do appear to regulate translation of the G₄C₂ repeat expansion. For example, phosphorylation of eIF2α, which forms the eIF2 complex with eIF2β and eIF2γ, results in upregulation of translation of the G₄C₂ repeat expansion (Cheng et al., 2018). Expression of a phosphomimetic eIF2α mutant increased DPR protein synthesis, while inhibition of PRKR-like ER kinase (PERK), a protein involved in the downstream signaling pathway of eIF2α phosphorylation, reduced it. In addition, eIF4B, which is part of the eIF4F complex, and eIF4H, a protein that can perform the same function, both bind to the G₄C₂ RNA (Satoh et al., 2014) and mediate DPR protein synthesis from the C9ORF72 repeat expansion in a Drosophila model (Goodman et al., 2019). Finally, another initiation factor found to regulate translation of the G₄C₂ repeat expansion is Ribosomal Protein Subunit 25 (RPS25), a constituent of the 40S ribosomal subunit. RNAi targeting RPS25 in a human cell line expressing 66 repeats, in C9ORF72 iPSCs and in a Drosophila model were used to interrogate its role. DPR protein synthesis was reduced by 50%, while global protein synthesis levels were unaffected. Furthermore, the knock down of RPS25 extended life span in Drosophila (Yamada et al., 2019), potentially an indication of a detrimental role of RPS25 in the production of the potentially toxic DPR proteins.

Several non-canonical translation initiation factors have also been proposed to regulate translation of the C9ORF72 G₄C₂ repeat expansion. Sonobe et al. reported that the presence of eIF2A is required (Sonobe et al., 2018). EIF2A, which is distinct from eIF2α, is a non-essential protein that can act in place of the eIF2 complex in delivering initiator tRNAs. The role of eIF2A in the translation of the G₄C₂ repeat expansion was demonstrated in an eIF2A HEK293 knock out cell line in which translation of the repeat expansion was significantly reduced compared to that of the WT cells (Sonobe et al., 2018). This result was confirmed by another group in HAP1 cells, but not in HEK293T cells (Green et al., 2022). Green et al. hypothesized that the difference between the findings in the two lines could have been caused by different levels of endoplasmic reticulum (ER) stress, which alters the level of eIF2 available for translation initiation. Indeed, when the binding between eIF2 GTP and the Met-tRNA was blocked by NSC119893 (a small molecule that blocks Met-tRNA binding), translation of the repeat expansion was also inhibited. Altogether, the findings suggest that either eIF2 or eIF2A must bind Met-tRNA to enable G₄C₂ repeat expansion translation. Knockdown of another non-canonical initiation factor, Death Associated Protein 5 (DAP5), was shown to increase translation of the G₄C₂ repeat expansion (van ‘t Spijker et al., 2021). DAP5 is an eIF4G homolog that can bind eIF3 and eIF4A to initiate translation and facilitates translation of approximately 20% of cellular mRNAs (de la Parra et al., 2018). MCTS1 Reinitiation and Release Factor (MCTS1) and Density Regulated Reinitiation and Release Factor (DENR), two factors involved in the re-initiation of the ribosome after release from mRNA, were also found to regulate translation of the G₄C₂ repeat expansion. Similar to the case with RPS25, knockdown of DENR prolonged the survival rates in a G₄C₂ repeat expansion drosophila model (Green et al., 2022).

In summary, while some initiation factors do not play a role, several factors appear to regulate the initiation of translation of C9ORF72 intronic repeat expansion, including: phosphorylated eIF2α, eIF4B, eIF4H, RPS25, eIF2A and DAP5, as well as re-initiation regulators MCTS1 and DENR. The involvement of a number of initiation factors leads to the question of where on the RNA the ribosomes are then initiating the translation of the intronic repeat.

2.2.3. Dependence on Initiation Sites Upstream of the Repeat Expansion

In canonical translation, the initiation of a polypeptide synthesis occurs at an AUG codon. Ribosome profiling studies, which are designed to map the positions and relative amounts of ribosomes on mRNA, have demonstrated that initiation of translation can also take place at near-cognate initiation codons (Ingolia et al., 2009; Ingolia et al., 2011). These codons differ from the AUG sequence by one nucleotide and also require Kozak motifs for efficient translation initiation, similar to canonical translation (Ichihara et al., 2021). Most non-AUG initiation codons are found upstream of an AUG-dependent ORF, and are less efficient than AUG codons in initiating translation; CUG is often the most efficient, followed by GUG, ACG, and AUU (Kearse et al., 2017). Downstream stimulators such as RNA secondary structures, RNA binding proteins and site-specific ribosomal pauses can enhance the efficiency of the non-AUG initiation, even, in some cases, to a comparable level to that of AUG initiation (Andreev et al., 2022). Evidence for initiation of translation at canonical and non-canonical initiation sites upstream of the C9ORF72 repeat expansion for each reading frame will be discussed below.

Poly-GA reading frame.

The identification of initiation sites in the sequence upstream of the CGG repeats in the untranslated region of the fragile X mental retardation 1 (FMR1) mRNA first highlighted the importance of the RNA sequence preceding repeat sequences for the production of repeat proteins (Kearse et al., 2016; Sellier et al., 2017). The context of the CGG repeats, located in the 5’ UTR of the FMR1 gene and therefore subject to typical ribosomal scanning, is quite distinct from the intronic C9ORF72 G₄C₂ repeats. Nevertheless, the finding raised at least the possibility that a similar mechanism could be responsible for the generation of DPR proteins from C9ORF72. Several groups noted that there is a CUG near-cognate initiation codon in a strong Kozak sequence 24 nucleotides upstream of the G₄C₂ repeats in the poly-GA reading frame (Green et al., 2017; Tabet et al., 2018; Sonobe et al., 2018; Almeida et al., 2019; Boivin et al., 2020). Mutation of this CUG in reporter constructs containing the intron 1 sequence largely decreased poly-GA translation levels (Green et al., 2017; Tabet et al., 2018; Sonobe et al., 2018; Lampasona et al., 2021; van ‘t Spijker et al., 2021) and mass spectrometry analysis confirmed that the CUG codon is used for initiation of the poly-GA protein synthesis (Boivin et al., 2020). Intriguingly, mutating this CUG codon in the poly-GA frame also affects the translation of the poly-GP and/or poly-GR reading frames, but the effects seem to be cell line and reporter construct dependent. In motor neurons derived from a C9ORF72 iPSC line carrying more than 1,000 G₄C₂ repeats, deletion of an intronic region containing the CUG initiation site eliminated poly-GA protein production without affecting poly-GP or poly-GR levels (Almeida et al., 2019). These results are particularly striking because they show that in human neurons 1,000 copies of the G₄C₂ repeats on their own are not sufficient to drive ribosomes to initiate translation in the poly-GA reading frame. In support of this, ribosome profiling studies revealed that the ribosome footprints on the CUG codon used to initiate translation of the poly-GA reading frame were still detected even when there were no G₄C₂ repeats present, indicating that the repeat expansion does not drive positioning of ribosomes in the intron (van ‘t Spijker et al., 2021). Collectively, these studies demonstrate that the poly-GA protein production relies on the RNA sequence upstream of the G₄C₂ repeat expansion.

Poly-GR reading frame.

Tabet et al. reported that mutating the CUG in the poly-GA reading frame also eliminated poly-GR translation in HEK293 cells, leading the authors to propose that the same CUG codon initiates poly-GR production through a frameshifting mechanism (Tabet et al., 2018). However, two other groups performing a similar experiment in the same cell line did not detect any significant effect on the translation of the poly-GR reading frame (Green et al., 2017; Lampasona et al., 2021). Similarly, in C9ORF72 motor neurons, elimination of the CUG region did not impact poly-GR protein levels (Almeida et al., 2019). Therefore, additional studies are needed to establish whether DPR proteins are synthesized involving a ribosome frameshifting mechanism. Mutation of a second near-cognate initiation codon of AGG in a weak Kozak sequence immediately adjacent to the G₄C₂ repeats in the poly-GR reading frame was shown to eliminate poly-GR protein production (Boivin et al., 2020). However, the authors were unable to confirm the finding using mass spectrometry analysis was as had been demonstrated for the poly-GA protein production (Boivin et al., 2020). In contrast, deletion of the AGG codon in a splicing reporter construct did not decrease translation in the poly-GR reading frame (Lampasona et al., 2021). Nevertheless, poly-GR protein production may still follow a near cognate initiation mechanism like poly-GA, as suggested by elimination of the 111 nucleotides right before the G₄C₂ repeats largely decreasing translation in the poly-GR frame, while removing only the repeats themselves did not affect translation levels (Lampasona et al., 2021). Sequential deletion analysis of the region preceding the repeats did not result in identification of a strong initiation codon and the translation efficiency appeared to show dependency on the length of the intronic sequence. Two intronic AAGAAAA motifs were found to regulate translation in the poly-GR frame (Lampasona et al., 2021). Thus, further studies are needed to determine whether poly-GR frame translation is indeed dependent on an as of yet unidentified initiation codon.

Poly-GP reading frame.

The poly-GA and poly-GR proteins can only be derived from sense strand transcripts. In contrast, the poly-GP protein can be synthesized from both the sense and the anti-sense strands of the C9ORF72 repeat expansion. In experiments investigating the sense direction, ribosome profiling revealed a footprint over a CUG codon 83 nucleotides upstream of the G₄C₂ repeats in the poly-GP frame, but this CUG is also in frame with a stop codon immediately prior to the repeat expansion (van ‘t Spijker et al., 2021). Mutation of the CUG codon decreased poly-GP frame translation, indicating that the ribosomes starting from this CUG can read through this termination codon directly preceding G₄C₂ repeats, at least in the reporter construct used for the study (van ‘t Spijker et al., 2021). Interestingly, on the anti-sense RNA, there are 3 putative canonical AUG initiation sites that are in the poly-GP frame (at 212, 194 and 113 nucleotides from the repeats), without any in-frame termination codons in between the AUGs and C₄G₂ repeats. Any or all of these could initiate the production of the poly-GP protein. Mass spectrometry analysis of the poly-GP protein generated by expression of the anti-sense sequence in cells revealed that the AUG start codon located 194 nucleotides upstream of the C₄G₂ repeats was used to initiate poly-GP protein production (Boivin et al., 2020). Deletion of the antisense sequence containing this AUG codon precluded poly-GP protein production (Boivin et al., 2020). Recently, another study showed that CRISPR/Cas9 mediated excision of exon 1a eliminated the sense direction-dependent poly-GA, but did not eliminate poly-GP protein in a C9ORF72 iPSC line (Sckaff et al., 2022). These results are consistent with studies in human autopsy tissues from C9ORF72 expansion carriers showing that the poly-GP protein is predominantly generated from antisense RNAs (Zu et al., 2013). Studies with C9-antisense oligonucleotides (ASOs) that target the sense RNA have found that these ASOs are very effective at eliminating poly-GP protein (Gendron et al., 2017), paradoxically indicating that poly-GP is produced from the sense RNA. Anti-sense RNAs that form the RNA foci do not seem to be affected by treatments with ASOs (Lagier-Tourenne et al., 2013; Jiang et al., 2016). An interesting feature of the C-rich repeats such as antisense C₄G₂ is that they do not induce export of the host intron to the cytoplasm in the same way that the sense G₄C₂ repeats (G-rich) do (Wang et al., 2021). Therefore, one could speculate that the important step for translation, that of cytoplasmic localization, may depend on the sense RNA facilitating the export of the antisense RNA to the cytoplasm and subsequent production of poly-GP protein.

Poly-PR and poly-PA reading frames.

Similar to the case for poly-GP, the context suggests poly-PR protein is likely to be made by a canonical translation mechanism from antisense RNA strands. Analysis of the antisense sequence revealed that there is a canonical AUG initiation codon 273 nucleotides from the start of the C₄G₂ repeats, without any intervening termination codon. A recently developed prediction model, used to identify translation initiation sites in genes associated with neurologic repeat expansion disorders, identified ten codons in which the AUG codon had the best prediction score for initiating translation of the poly-PR reading frame (Gleason et al., 2022). Confirmation of these findings experimentally using reporter constructs and cells from C9ORF72 expansion carriers is warranted. As for the poly-PA protein, it is still unclear whether a specific initiation codon is required for translation of the protein. In the absence of good prediction modeling or an assay to measure endogenous poly-PA protein levels, this species can currently only be characterized using reporter constructs in cellular assays.

2.2.4. Regulation by an Upstream Open Reading Frame

Translation of the C9ORF72 G₄C₂ repeat expansion has been proposed to be regulated by an uORF that starts at the end of exon-1a and extends 180 nucleotides into intron-1, ending at an in-frame termination codon (Tabet et al., 2018). uORFs are defined by having their start codons located in the 5′ UTRs of a host mRNA. They are highly prevalent in vertebrates and frequently inhibit translation of downstream ORFs (Johnstone et al., 2016). This inhibitory effect is caused by the inability of the ribosome to re-initiate translation after being released from the RNA following translation of the uORF (Young and Wek 2016). To further investigate the role of the uORF in the initiation of translation of the repeat expansion, van ‘t Spijker et al. mutated its AUG initiation codon to UAA in a reporter construct (van ‘t Spijker et al., 2021). The mutation led to an increase in translation of the poly-GA and poly-GP frames, indicating that the uORF indeed exerts an inhibitory effect on translation of the downstream G₄C₂ repeat expansion (Figure 1). Mass spectrometry established that this uORF is translated to produce a 6kDa polypeptide, predicted to form a helix-loop-helix structure (van ‘t Spijker et al., 2021). The polypeptide amino acid sequence is conserved among higher primates, consistent with the possibility that the polypeptide has a biological function in addition to its C9ORF72 inhibitory action. Further investigation into the biological activity of the C9ORF72 uORF polypeptide may be important for understanding the potential effects of therapeutic interventions targeting expression of the C9ORF72 repeat expansion that impinge on the polypeptide’s expression.

Figure 1. Translation Initiation of the C9ORF72 Repeat Expansion and Diversity of DPR proteins.

(A) A number of initiation factors are likely needed to recruit ribosomes to initiate translation of the repeat expansion. Both m7G cap-dependent and -independent binding mechanisms have been proposed to be involved in the synthesis of the DPR proteins. (B) DPR protein production is dependent on initiation sites located upstream of the repeat expansion. Canonical and non-canonical initiation codons have been identified in both the sense and antisense strands of RNA. (C) Translation of an uORF that starts at the end of exon-1a and extends 180 nucleotides into intron-1 has been proposed as inhibitory to DPR synthesis. (D) RAN translation initiation is proposed to occur by direct binding of ribosomes to a secondary structure formed by the repeat RNA sequence and to forego requirement for an AUG start site. (E) Owing to the involvement of varied translation mechanisms, DPR proteins are predicted to have unique N-termini and divergent C-termini. (F) Chimeric DPR proteins may be generated by ribosome frameshifting and interruptions in the repeat sequence. (G) Post-translational modifications (PTMs) can occur on DPR proteins, in particular dimethylation of arginine-containing species, providing yet another source of DPR protein diversity.

2.2.5. RAN Translation Initiation

Translation of the C9ORF72 repeat expansion has also been proposed to be the result of repeat-associated non-AUG (RAN) translation, a mechanism initially described in studies characterizing spinocerebellar ataxia type 8 and myotonic dystrophy type 1 (Zu et al., 2011; 2013). The key features of the RAN translation mechanism are the direct binding of ribosomes to a secondary structure that is formed by repeat RNA sequences, and initiation of translation without a requirement for scanning of the 5’ end of the RNA or an AUG start site, similar to how internal ribosomal entry sites (IRES) in viruses work (Abeyrathne et al., 2016). Zu et al. also demonstrated that translation of the repeat expansion is length-proportionate, suggesting that the longer the repeat expansion is, the more opportunity ribosomes have to associate on the repeat expansion RNA and initiate translation (Zu et al., 2011). This same laboratory also reported that the C9ORF72 repeat expansion can be translated in both the sense and antisense direction, leading to the production of six DPR proteins each with unique carboxyl terminal regions (Zu et al., 2013).

The RAN translation mechanism has been the leading hypothesis in the field of C9ORF72 repeat expansion research. However, a number of the recent studies discussed in this review challenge the idea that the ribosomes are initiating translation by direct association with the repeat expansion sequence of the RNA. For instance, the role played by the initiation factors eiF2α, eIF4B and eIF4H, as well as that of the uORF in the translation of the G₄C₂ repeat expansion indicate that the ribosomes scan the 5’ end of the RNA before initiating translation. Likewise, the dependence of DPR protein synthesis on the presence of initiation sites located upstream of the repeat expansion indicates that ribosomes initiate translation before reaching the beginning of the repeat expansion sequence. The study that provided evidence for the RAN translation mechanism in C9ORF72 ALS-FTD used a reporter containing 6 stop codons (2 per reading frame) followed by the repeat expansion, a strategy based on the expectation that termination codons are sufficient to on their own prevent ribosomes from read through. However, read through of stop codons has been demonstrated for several mamalian mRNAs (Manjunath et al., 2022), and translational termination and recruitment of release factor require more context as evidenced by studies characterizing premature termination codons and their read through (Dabrowski et al., 2018). Whether RAN translation indeed represents a mechanism involved in ALS-FTD caused by the repeat expansion ultimately needs to be convincingly established in C9ORF72 human cells.

3. DPR Proteins and their Contribution to Neurodegeneration in C9ORF72 ALS-FTD

3.1. Evidence from Human Studies

Since the discovery of DPR proteins originating from the translation of both the sense and the anti-sense strand of the C9ORF72 repeat expansion in ALS-FTD brain tissues (Zu et al., 2013; Mori et al., 2013; Gendron et al., 2013), many studies have reported that these proteins exert toxic effects upon expression in cells and animal models (Balendra and Isaacs 2018). These observations spurred studies to investigate whether a correlation exists between the presence of the various DPR proteins and the areas of degeneration in C9ORF72 expansion carriers’ brains. Early clinicopathological studies found no correlation between the presence of poly-GA, the most abundant of the DPR proteins, and the degree of neurodegeneration, while TDP-43 pathology was present in the clinically affected regions (Mackenzie et al., 2013). A negative correlation between the extent of poly-GA pathology and age of disease onset was noted (Davidson et al., 2014). Similarly, a more compressive analysis involving scoring of the different types of intracellular inclusions for all DPR proteins in relevant brain regions revealed only a moderate association between the presence of poly-GA inclusions in dystrophic neurites with degeneration in the frontal cortex and timing of disease onset, but no other association between DPR proteins and neurodegeneration (Mackenzie et al., 2015). Using a different approach, in which the brain tissues were subdivided into clinically related and unrelated regions, and by comparing to phosphorylated TDP-43, Saberi et al. found a correlation between the abundance of poly-GR inclusions in dendritic structures and neuroanatomical pathology (Saberi et al., 2018). This correlation was not seen for other DPR proteins. A later study using quantitative digital microscopic methods provided evidence that the density of poly-GR inclusions correlates with neurodegeneration more robustly than any other DPR proteins. However, the study found little evidence for the presence of poly-GR in dystrophic neurites (Sakae et al., 2018). An important limitation of these immunohistochemistry-based studies is that the analyses were all performed following fixation of the tissues, which could result in loss of the more soluble DPR protein species. To overcome this, two studies using protein fractionation and an immunoassay sought to quantitatively measure both soluble and insoluble DPR proteins in brain homogenates from C9ORF72 repeat expansion individuals (Quaegebeur et al., 2020; Gendron et al., 2015). The common observation from both studies was that the various DPR proteins have different solubility profiles across the brain tissue. Additional studies are needed to reveal whether and how the solubility properties of each DPR protein influences their toxic potential. In the study by Gendron et al., poly-GP and poly-GA levels largely did not correlate with neurodegeneration. However, in the cerebellum, higher poly-GP levels were associated with cognitive impairment (Gendron et al., 2015). In Quaegebeur et al., the most compelling observation was that soluble DPR proteins were less abundant in brain regions that are more affected by neurodegeneration, contrasting to the higher levels of these proteins found in the cerebellum. The authors speculate this could be an indication that soluble DPR proteins might be involved in the early stages of disease or may not play a significant role at all. The fact that DPR protein loads are consistently found to be higher in the cerebellum, an area observed to be less affected in the disease, raises the intriguing possibility that alterations in the cerebellum could contribute to the neurodegeneration seen in clinically affected areas of the brain. In the study, insoluble poly-GA and poly-GR levels also did not corelate with neurodegeneration, with these DPR protein concentrations being similar across the different brain areas analyzed. The study did find a negative correlation between insoluble poly-GR levels and clinical parameters of severity, potentially indicating that poly-GR does play a toxic role in the disease (Quaegebeur et al., 2020).

In addition to assessments in brain tissue, DPR proteins can also be quantitatively measured in the cerebrospinal fluid of C9ORF72 mutation carriers. Poly-GP levels were shown to be stable over time and not to correlate with neurodegeneration or disease progression (Meeter et al., 2018; Gendron et al., 2017). Likewise, poly-GA and poly-GR levels do not correlate with disease onset or duration, and their levels do not discriminate between symptomatic and pre-symptomatic C9ORF72 mutation carriers (Krishnan et al., 2022). Nevertheless, the level of these DPR proteins may still have the potential to be useful biomarkers for evaluating the effectiveness of therapies aimed at reducing the toxic burden of DPR proteins.

It is currently unclear why there is a disconnect between human post-mortem studies, which show little to no correlation between the localization or abundance of DPR proteins and neurodegeneration, and model systems where DPR proteins have robustly demonstrable toxic effects. One important aspect to consider is that studies with patient samples must necessarily be post-mortem. In addition to the risk of artifacts associated with the collection and storage of such samples, bias may also be a consequence, as these will be exclusively end-stage disease samples, thus failing to provide representation of the state and distribution of DPR aggregates from earlier stages of disease where these may be more consequential. An additional layer of complexity arising from the diverging results between such studies is more solvable. Differences in the methodologies used, the size of the study and clinicopathological phenotypes presented by the subjects are contributing factors to the difficulties in interpreting end-stage post-mortem data, that may be addressable by alignment between groups or with additional studies or methods. The exact role of individual DPR proteins in the cascade of events causing neurodegeneration in C9ORF72 ALS-FTD has not been definitively established. Additional functional studies and additional human post-mortem studies will both likely be required to more fully elucidate the role and impact of these mutation-specific proteins and to characterize how therapeutics may be developed to address their involvement.

3.2. Emerging Diversity of DPR Proteins

Advances in the understanding of how DPR proteins are synthetized have led to the conclusion that there may be more DPR species than initially suspected. The particular amino acid compositions of these proteins are likely to depend on the translation mechanisms at play. In landmark work by Zu et al., DPR proteins were found to have distinct C-terminal amino acid sequences (Zu et al., 2013). Subsequent studies characterizing the influence of initiation codons upstream of the repeat expansion indicated that DPR proteins also likely have unique N-termini in addition to divergent C-termini (section 2.2.3 and Figure 1). These findings are relevant because the particular amino acid composition will in all likelihood affect the structure, localization, interaction partners and ultimately toxic potential of these proteins. In line with this, He et al. found that inclusion of the native C-terminal sequence reduced toxicity and shifted the subcellular localization of the DPR proteins in comparison to the same proteins lacking this sequence (He et al., 2020). Future studies evaluating the toxic potential of DPR proteins must therefore consider the native C- and N-termini to ensure any findings are relevant to the endogenous context present in the disease.

Further diversity in DPR protein species is contributed by ribosome frameshifting that generates chimeric repeat proteins (Tabet et al., 2018; McEachin et al., 2020). Likewise, interruptions identified in the repeat expansion sequence could be another potential source of chimeric DPR proteins. Two studies using long-read sequencing technologies found that the C9ORF72 repeat region is mainly composed of G₄C₂ repeats (>96–99%). However, rarer motifs such as the pentanucleotide G₃C₂ and small interruptions could not be ruled out (DeJesus-Hernandez et al., 2021; Ebbert et al., 2018). Since even a small interruption could significantly affect the amino acid composition of the protein, for instance by changing the reading frame, precisely determining the repeat pattern and any of its interruptions or deviations would be informative. Improved methodologies are needed to overcome the challenges associated with the sequencing of long, repetitive, GC rich expansions. A new approach was recently developed to enable sequencing to the single nucleotide level of the long CCTG repeat expansion responsible for causing Myotonic dystrophy type 2 (Alfano et al., 2022). Applying a similar approach to C9ORF72 ALS-FTD may help establish how chimeric proteins arise, and whether a correlation exists between the composition of the expansion and the clinical phenotype.

Post-translational modifications (PTMs) of DPR proteins brings another layer of complexity to the diversity of these proteins on top of their potentially intricate amino acid compositions (Figure 1). Such modifications can substantially affect protein function and may impact their toxic potential. A recent study exploring the contribution of PTMs to disease pathogenesis in ALS-FTD showed that methylation of arginine residues in poly-GR that occurs in brain tissues of C9ORF72 expansion carriers appears to be protective against clinical severity (Gittings et al., 2020). Similar findings in primary neuronal cultures showed uptake of dimethylated poly-GR resulted in less toxicity than uptake of unmethylated poly-GR (Gittings et al., 2020). In contrast, a separate study showed that reduction of methylation through inhibition of type I protein arginine methyltransferases protects from poly-GR and poly-PR-induced toxicity in mouse spinal cord neuroblastoma hybrid cells (NSC-34, (Premasiri, Gill, and Vieira 2020). These apparently conflicting results may be reconciled by the fact that symmetric dimethylation (in Gittings et al., 2020) averts the occurrence of asymmetric dimethylation (in Premasiri et al., 2020) of a given substrate, and it is the symmetry of the methylated arginine species that determined toxicity in these studies. Both studies used relatively small peptides (15–20 residues) in comparison to the anticipated DPR lengths present in C9ORF72 repeat expansion carriers, and neither study considered possible flanking sequences. Additional studies using additional models will be needed to better understand how arginine methylation status of DPR proteins impacts their toxic potential and their contribution to neurodegeneration.

3.3. Linking DPR Proteins to Neurodegeneration

Although C9ORF72 expansion carrier cells are likely to simultaneously produce the diverse set of DPR proteins, most of our understanding of how DPR proteins cause toxicity is derived from studies investigating the consequences of the expression of individual DPR protein sequences. DPR proteins in these studies are purposefully produced using codon-optimized constructs to exclude the potential contribution of any RNA-mediated G₄C₂ repeat toxicity (Mizielinska et al., 2014 Jovičić et al., 2015). From this reductionist approach, the evidence suggests that DPR proteins containing arginine (poly-GR and poly-PR) are the most toxic, followed by poly-GA (Moens et al., 2019; Mizielinska et al., 2014; May et al., 2014). Poly-GP and poly-PA are considered to be relatively non-toxic (Freibaum et al., 2015; Lee et al., 2017). Further studies will be required to characterize the level of interplay between the different DPR proteins present together in carrier cells.

The pathogenic contribution of DPR proteins to disease progression is still unclear. DPR proteins have been shown to impact a number of cellular mechanisms whose dysfunction are capable of leading to neurodegeneration. Nucleocytoplasmic transport, DNA damage, mitochondrial dysfunction, ER stress, stress granule formation, axonal transport defects, dysfunction of calcium signaling, impaired global protein translation and protease inhibition have all been shown to be impinged on by DPR proteins (reviewed in Gao et al., 2017; Schmitz et al., 2021). Seen in 97% of ALS and up to half of FTD patients, a pathological hallmark of neurodegneration is the mislocalization of TDP-43 to the cytoplasm, where it can form inclusions (Neumann et al., 2006). Therefore a number of studies have searched for links between TDP-43 pathology and the expression of DPR proteins. Mislocalization of TDP-43 is thought to occur by perturbations at the nuclear pore complex that affect nuclear-cytoplasmic transport, but the mechanism is still not fully understood. The observation that poly-GR causes the formation of RNA granules in the cytoplasm that sequester NUP62, a nuclear pore complex protein, along with TDP-43, identifies one way in which DPR proteins are involved in this pathology (Gleixner et al., 2022). NUP62 colocalizes with phosphorylated TDP-43 in C9ORF72 and sporadic ALS-FTD brain tissue, suggesting NUP62 may also interact with TPD-43 independent of a poly-GR influence on this interaction. Independently, poly(ADP-ribose) (PAR), a PTM that plays an essential role in stress granule assembly, was shown to promote the formation of stress granules by poly-GR and poly-PR with subsequent aggregation of TDP-43 (Gao et al., 2022). Poly-GA is able to indirectly promote TDP-43 aggregation by inhibiting the proteosome (Khosravi et al., 2020). Two recent proteomic studies showed that poly-GA aggregates associate with proteosomal components, molecular chaperones and factors involved in protein folding and degradation (Božič et al., 2022; Liu et al., 2022). Additionally, poly-GA was proposed to cause alterations in autophagic processes by sequestering Valosin Containing Protein (VCP) into its aggregates (Božič et al., 2022). VCP loss is thought to impair autophagy via inhibition of the fusion of autophagosomes with lysosomes. Another protein that plays a role in the endolysosomal pathway, TBK1, was shown to link TDP-43 pathology and poly-GA aggregation with this pathway (Shao et al., 2022). The authors showed that loss of TBK1 function after sequestration into poly-GA inclusions resulted in disruption of endosome maturation with subsequent induction of TDP-43 aggregation (Shao et al., 2022). Interestingly, TBK1 was previously shown to mediate the function of the C9ORF72/SMCR8/WDR41 complex in autophagy through phophorylation of SMCR8 (Sellier et al., 2016). It may be that, in disease cells, poly-GA further decreases the function of C9ORF72 below its already reduced levels owing to the G₄C₂ repeat expansion. An interesting and novel aspect of the Shao et al study is the linkage of two ALS-FTD disease proteins, TBK1 and C9ORF72, with TDP-43 pathology. A number of ALS-FTD disease proteins including VCP, Optineurin, TMEM106B, progranulin, p62, CHMP2B, all affect endosomal or lysosomal functions (Almeida and Gao 2016). Taken together, these studies provide further support for the role of compromised protein-clearance pathways in the pathogenesis of ALS-FTD.

4. Conclusion

Ever since the unexpected discovery that intronic hexanucleotide repeat expansion in C9ORF72 can be translated into DPR proteins, a multitude of studies have sought to elucidate the molecular mechanisms responsible for their synthesis. Such mechanistic characterization may prove particularly important in enabling the design of more specific therapeutic strategies aiming to reduce or eliminate these proteins and their potential toxic effects. Some DPR proteins appear to be produced by an unconventional translational mechanism that involves 5’ end scanning of RNAs containing the repeat expansion, with initiation at near-cognate start sites. Other DPR proteins utilize a canonical AUG-dependent mechanism. Many factors are likely to be needed to initiate and promote synthesis of DPR proteins, and how these factors work together and influence each other is still not understood. More studies characterizing these processes in C9ORF72 repeat expansion carrier cells are needed to implicate the translational mechanisms that are relevant in the endogenous context that is present within cells in the disease state.

A further important aspect is the amino acid composition of the DPR proteins. The possibility of ribosome frameshifting and the observance of interruptions in the repeat sequence indicates that far more DPR species may be generated than initially suspected. Such chimeric proteins may potentially possess one or more of different biochemical properties, subcellular localization or toxic potential. Future studies will hopefully address whether chimeric DPR proteins can be detected in repeat expansion carrier cells, and how these contribute to disease pathogenesis.

Toxic effects have been demonstrated for a subset of DPR proteins when exogenously expressed in cellular and animal models. However, there is discordance of these findings with studies using post-mortem brain tissues showing DPR proteins are present in areas not thought to be of particular clinical relevance to the disease. If and how DPR proteins interact with each other, and any resultant effect on toxic potential of these interactions, is not an aspect that has been sufficiently explored: cellular and animal models only express one single specific DPR protein at a time. An emphasis on the development of more sophisticated disease models in which multiple DPR proteins, complete with their flanking regions, are expressed together at levels commensurable with physiological conditions would help to improve our understanding of the interplay of distinct DPR proteins and how that interplay may contribute to disease severity and progression.

Determining how DPR proteins exert their toxic potential and impact cellular functions is an important step in characterizing the pathogenesis of C9ORF72 ALS-FTD, with the eventual goal of identifying new therapeutic targets. The recently identified link between expression of DPR proteins and a compromised endosomal-lysosomal pathway suggest tantalyzing hypotheses, but more work is needed. A number of other ALS-FTD disease proteins impact this same pathway, suggesting that the development of strategies that alleviate dysfunction in this pathway could hold great promise for the development of therapies targeting a shared underlying cause for ALS and FTD.

Abbreviations list

ALS

amyotrophic lateral sclerosis

ASOs

antisense oligonucleotides

FTD

frontotemporal dementia

DPR

dipeptide repeat

UTR

5’- Untranslated region

G₄C₂

GGGGCC

uORF

upstream Open Reading Frame

eIF4E

eukaryotic initiation factor 4E

PIC

preinitiation complex

CITE

cap-independent translation enhancers

RPS25

Ribosomal Protein Subunit 25

DENR

Density Regulated Reinitiation and Release Factor

DAP5

Death Associated Protein 5

FMR1

fragile X mental retardation 1

RAN

repeat-associated non-AUG

IRES

ribosomal entry sites

PTMs

Post-translational modifications

VCP

Valosin Containing Protein

ER

endoplasmic reticulum