RAN proteins in neurodegenerative disease: repeating themes and unifying therapeutic strategies

Abstract

Microsatellite-expansion mutations cause >50 neurological diseases, but there are no effective treatments. Mechanistic studies have historically focused on protein loss-of-function and protein or RNA gain-of-function effects. It is now clear that many expansion mutations are bidirectionally transcribed producing two toxic expansion RNAs, which can produce up to six mutant proteins by repeat associated non-AUG (RAN) translation. Multiple types of RAN proteins have been shown to be toxic in cell and animal models, to lead to common types of neuropathological changes and to dysregulate key pathways. How are RAN proteins produced without the canonical AUG or close-cognate AUG-like initiation codons is not yet completely understood but RNA structure, flanking sequences and stress pathways have been shown to be important. Here we summarize recent progress in understanding the role of RAN proteins, mechanistic insights into their production and the identification of novel therapeutic strategies that may be applicable across these neurodegenerative disorders.

Introduction

Repeat associated non-AUG (RAN) translation, a process in which protein translation occurs across microsatellite repeat expansion RNAs in the absence of AUG or AUG-like close cognate start codons, was first reported in 2011 [1]. Because many of these mutations express both sense and antisense expansion transcripts [2] and RAN proteins in all three reading frames, each repeat expansion mutation can produce up to six different polymeric proteins, three from the sense transcripts and the other three from the antisense transcripts. Depending on the repeat motif, homo-, di-, tetra- and penta-peptide RAN proteins can be produced. And based on the location of a given expansion mutation within its corresponding gene, toxicity may non-exclusively involve different mechanisms including protein loss- or gain-of-function, RNA gain-of-function, and RAN protein toxicity. We discuss our current understanding of RAN proteins and their contributions to disease, the molecular mechanisms of RAN translation and therapeutic strategies that target RAN translation or RAN proteins themselves and may be applicable to a wide range of microsatellite expansion disorders.

RAN translation defines new category of diseases in which unexpected and toxic repeat protein accumulate.

RAN translation was first described in spinocerebellar ataxia type 8 (SCA8) and myotonic dystrophy type 1 (DM1) [1], diseases that are caused by non-coding CAG repeat expansions in the ataxin 8 (ATXN8) and DMPK genes, respectively. To better understand the role of RNA vs. polyGln in SCA8, Zu et al., mutated the only ATG start codon upstream of the polyglutamine encoding CAG repeat track to AAG to block the expression of the polyGln protein but surprisingly this did not prevent the production of CAG coded polyGln. After ruling out potential artifacts, Zu et al. developed a frame-specific epitope tag system to test if proteins were made in multiple reading frames and showed that in addition to polyGln, polyAla and polySer were also expressed from an ATXN8 minigene in transfected HEK293T cells without an AUG initiation codon in any reading frame. Zu et al., demonstrated that this repeat associated non-AUG (RAN) translation: occurs when expansion RNAs lacking upstream AUG-initiation codons are transfected into cells; does not require frameshifting or RNA editing. Zu et al., went on to show that RAN proteins are also expressed in vivo. C-terminal specific antibodies showed the predicted polyAla RAN protein accumulates in Purkinje cells in SCA8 mice and human autopsy tissue, and that polyGln protein encoded by the CAG repeat accumulates in DM1 patient skeletal muscle, myoblasts and blood [1]. A later study in SCA8 showed polySer RAN proteins primarily accumulate in white matter regions while AUG-initiated polyGln showed prominent Purkinje cell accumulation in SCA8 mice and patient autopsy tissue [3].

RAN translation is not limited to repeat expansion mutations located in non-coding regions. Although most research into the molecular mechanism of Huntington’s disease (HD) has focused on the effects of the polyGln-expansion containing mutant huntingtin (mHTT) protein, the CAG•CTG expansion also produces polySer, polyAla, polyLeu and polyCys RAN proteins [4,5]. Using antibodies that recognize the unique C-terminal regions, Banez et al. detected sense (polyAla, polySer) and antisense (polyLeu, polyCys) RAN protein staining in human HD autopsy brains [6]. RAN proteins were most abundantly found in brain regions most affected by HD including the caudate/putamen. Additionally, abundant RAN staining was seen in white matter regions that were negative for polyGln, suggesting RAN proteins contribute to previously reported HD white matter abnormalities [7–10]. These data demonstrate that expanded repeats located within protein coding regions can undergo RAN translation and that relatively short expansion mutations (~40–100 repeats) can produce RAN proteins. The recent failure of a phase 3 antisense oligonucleotide (ASO) clinical trial, which did lower mHTT and was predicted to lower sense RAN proteins expressed from CAG expansion transcripts highlights the need to consider alternative mechanisms including strategies to reduce both sense and antisense RAN proteins [6].

Another RAN protein disease is myotonic dystrophy type 2 (DM2), a multisystemic neuromuscular disorder caused a tetranucleotide CCTG•CAGG repeat expansion in the first intron of the cellular nucleic acid binding protein (CNBP) gene. When the DM2 mutation was first identified, the molecular mechanisms were thought to mainly involve the nuclear sequestration and loss of function of muscle-blind like family of proteins (MBNLs) by CCUG expansion transcripts [11–13]. However, in 2017, Zu et al. showed that sense (polyLPAC) and antisense (polyQAGR) tetra-peptide RAN proteins accumulate in DM2 autopsy brains. These RAN proteins accumulate in brain regions that show pathologic changes. PolyLPAC was found mainly in grey matter regions while polyQAGR showed abundant accumulation in white matter regions. Interestingly, cellular studies showed that overexpression of MBNL1 protein sequestered CCUG expansion RNAs in the nucleus and reduced LPAC RAN levels [14]. These data suggest MBNL1 sequestration is an early pathologic event which is followed by RAN protein accumulation as CCUG RNA levels increase and MBNL fails to sequester expanded transcripts inside the nucleus. Because DM1 and DM2 share the RNA gain of function disease mechanism of MBNL sequestration, it would be interesting to study if RAN proteins accumulate in DM1 patient brains and if RAN protein pathology explains any differences in the clinical features of these two diseases.

RAN translation has been most extensively studied in C9orf72 amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD), a disease caused by a G4C2•G2C4 repeat expansion in intron 1 of the C9orf72 gene [15–17*]. Similar to other repeat-expansion diseases, proposed mechanisms for C9 ALS/FTD include C9orf72 protein haploinsufficiency, RNA gain-of-function effects, and C9 RAN protein toxicity. A number of studies demonstrated that dipeptide repeat proteins (sense poly-GA, poly-GP, poly-GR, and antisense poly-GP, poly-PR, and poly-PA) accumulate in patient autopsy tissues [18–20]. The accumulation patterns of C9 RAN proteins are different between the various dipeptide proteins and between brain regions [20]. C9 RAN proteins were also detected in animal models of disease. AAV vector mediated G4C2 overexpression mice showed the accumulation of RNA foci, RAN proteins, phosphorylated TDP43 inclusions and developed neuronal degeneration and behavioral deficits [21]. To understand the effects of expanded G4C2 repeats at endogenous levels, several groups developed C9orf72 BAC transgenic mice [22–25]. While RNA foci and sense C9 RAN proteins were detected in all BAC transgenic mouse models, only two showed neurodegeneration and behavioral deficits with different levels of severity [24,25]. These and other studies highlight the complexity of C9orf72 ALS/FTD and the importance of mouse genetic background, repeat length, upstream flanking sequences and experimental conditions [26,27*].

RAN proteins have also been shown to accumulate in diseases caused by repeat expansion mutations located in 5’ UTRs. While CGG expansions with >200 repeats in the 5’UTR of the FMR1 gene cause Fragile X Syndrome, premutation alleles with 55–200 repeats can cause Fragile X tremor ataxia syndrome (FXTAS) or Fragile X premature ovarian insufficiency (FXPOI) [28]. In FXTAS and FXPOI, the FMR1 CGG repeat expansion was found to undergo RAN translation, producing a polyGly expansion protein that accumulates in ubiquitin-positive inclusions in patient tissues [4,29–31]. While polyGly RAN proteins are the most studied RAN protein expressed from GCC•GGC repeats, antisense (AS) polyAla and AS polyPro proteins have also been shown to accumulate in FXTAS autopsy brains and warrant further study. A recent study of patients with neuronal intranuclear inclusion disease (NIID) [32**] showed polyGly expansion proteins expressed from mutant CGG repeats in an AUG-initiated upstream ORF in the 5’ UTR of NOTCH2NLC form nuclear inclusions.

In summary, a variety of homo-, di-, tetra- or penta-peptide expansion proteins have been reported in DM1 [1], DM2 [14], C9 ALS/FTD [18–20], HD [6], SCA8 [1], SCA31 [33], SCA36 [34*], FXTAS [29,35] and FXPOI [30], FECD [36], forming a group of RAN protein diseases. In each of these diseases RAN proteins are found in affected patient tissues and accumulate in different patterns and levels which can vary from patient to patient. The accumulation of RAN proteins in repeat expansion diseases has raised important questions about the molecular mechanisms of RAN translation and whether and how RAN proteins contribute to disease, topics which are discussed in the following sections.

Mechanistic insights into RAN translation

Effects of structured RNAs and repeat length

Although detailed mechanisms and precise machineries required for RAN translation are still unknown, the involvement of hairpin forming structures of the repeat regions has been described. For example, in 2011 Zu et al. reported that the polyGln RAN proteins are expressed from hairpin forming CAG but not non-hairpin forming CAA repeats [1]. In 2019, Wang et al. provided additional data supporting the importance of hairpin forming RNAs in RAN translation by showing that a small molecule targeting hairpin-forming r(G4C2)₆₆ but not the G-quadruplex G4C2 structures inhibits C9 RAN poly-GP production [37]. Hairpin structures may facilitate RAN translation through internal ribosome entry sites (IRES) or other similar mechanisms [38,39].

Repeat length is also an important factor in RAN translation. In the initial report by Zu et al., CAGs with ≥ 73 repeats produced RAN proteins in all reading frames, while those ≤ 20 repeats did not express detectable RAN proteins from any frame. Repeat length thresholds for RAN protein expression for CAG expansions was shown to differ by reading frame with polyGln, polySer and polyAla detected at repeat lengths ≥ 42, 58 and 73 repeats, respectively. The general observation that longer repeats favor RAN translation has held up for other repeat expansion mutations including those at the C9orf72 [20,40], HD [6] and DM2 [14] loci, although specific effects of different repeat motifs and flanking sequences have also been reported [41–43*]. Frame specific repeat length effects may be mechanistically linked to the molecular machinery required for their expression. For example, RAN proteins for which translation is initiated within the repeat tract (e.g. polyAla) may depend more on repeat length than those in which expression initiate at or near the 5’ end of the repeat (e.g. polyGln) [1]. Although somatic mosaicism makes it difficult to understand the effects of repeat length on disease in C9orf72 ALS/FTD patients, isogenic sublines of C9-BAC mice with 800, 500 and 50 repeats show that longer repeat lengths increase disease penetrance and severity as well as the levels of RNA foci and RAN protein aggregates [27*].

Effects of integrated stress response kinases on RAN translation

The integrated stress response (ISR), which acts to repress global protein translation under conditions of stress, is activated in a wide range of neurological diseases and considerable efforts have focused on understanding the role of the ISR in disease [44**]. The ISR is regulated by four kinases, protein kinase R (PKR), PKR -like endoplasmic reticulum kinase (PERK), general control nonderepressible 2 (GCN2) and heme-regulated inhibitor (HRI), each of which phosphorylates the α-subunit of the eukaryotic translation initiation factor 2 (eIF2α), which in turn prevents the formation of the ternary complex required for AUG-initiated protein synthesis. Stress conditions that activate the ISR include viral infection, protein homeostasis abnormalities, nutrient deprivation, and oxidative stress. Increased levels of p-eIF2α decrease the synthesis of most proteins but upregulate the synthesis of specific proteins that restore cellular homeostasis [45–47].

In cell culture models of FXTAS and C9orf72 ALS/FTD, oxidative and ER stress have been shown to increase the production of RAN proteins [48–51]. The accumulation of RAN proteins, in turn, is thought to further increase stress by activating PERK, which in turn leads to increased eIF2α phosphorylation and additional RAN protein production. Pharmacological inhibition of PERK or eIF2α has been shown to reduce C9 RAN protein levels and stress granule formation in cells overexpressing G4C2 repeats [49,50]. Similarly, RAN protein production from DM2 minigenes is significantly increased by oxidative and ER stress and RAN protein levels are reduced in PERK KO cells [42*]. Another ISR kinase that has been associated with RAN translation is PKR. The activation of PKR by single and double stranded RNA (dsRNA) viruses leads to RNA-mediated PKR dimerization followed by PKR auto-phosphorylation [52]. In 2000, Tian et al. reported that CUG repeat expansion RNAs, which form imperfect double-stranded RNA hairpin structures, bind to and activate PKR in a repeat length dependent manner in a cell free in vitro system [53]. Building on this observation Zu et al., showed PKR is activated by several additional types of disease-causing microsatellite repeat expansion RNAs (CAG, CCUG, CAGG and G4C2) in transfected cells [54**] and p-PKR levels are elevated in C9orf72 BAC transgenic mice containing 500 G4C2 repeats and in human C9orf72 autopsy tissue. Additionally, these authors showed that inhibiting PKR activation with a dominant negative form of PKR (PKR-K296R), PKR KO and PKR inhibitors, dramatically reduced p-eIF2α and RAN protein levels for multiple types of repeat expansion mutations including C9–500 BAC transgenic ALS/FTD mice (see reference [54**] and below). Phosphorylation of eIF2α is an important regulator of RAN translation through the ISR [48,49]. However, activation of eIF2α by thapsigargin treatment or eIF2α-S51D (a constitutively activated form of eIF2α) in PKR-KO cells only partially rescued of RAN protein production. Taken together, these data show PKR regulates RAN translation through both eIF2α–dependent and eIF2α–independent pathways and that PKR is an upstream regulator of RAN translation and an important therapeutic target [54**].

Translation initiation factor and locus-specific sequence effects on RAN translation

The eukaryotic translation initiation factor eIF3F has also been reported to modulate RAN translation [3]. Overexpression of eIF3F in cells increases RAN protein levels, whereas knocking down eIF3F significantly reduced polySer, polyAla, and polyGP RAN proteins [3]. Interestingly, eIF3F levels were found to be elevated in white matter brain regions where SCA8 poly-Ser RAN protein levels are also elevated. The accumulation of RAN proteins in white matter brain regions has also been found in HD and DM2 [6,14], raising the possibility that eIF3F may also modulate HD and DM2 RAN translation.

In a separate study Yamada et al., used a yeast genetic screen to identify a non-essential protein of the small ribosomal subunit, RPS25A, as a regulator of RAN translation [55]. Using Drosophila, human cell lines and patient derived induced-motor neurons (iMNs) these authors showed, RPS25 knockdown reduced the levels of multiple types of C9orf72 and other RAN proteins without affecting RNA foci or global protein expression. RPS25 knockdown also improved survival in flies overexpressing 36 G4C2 repeats and in C9 iMNs stressed with glutamate.

RAN proteins have been reported in microsatellite expansion diseases with a range of different types of expansion motifs and flanking sequences [4,5] but how these differences affect the translational regulation of these proteins is just beginning to be understood. For example, Sellier et al. showed that the production of FMR1 polyGly occurs mainly through initiation at a close cognate AUG-like ACG codon located upstream of the expanded CGG repeats [56]. In C9orf72 ALS/FTD, a CUG codon upstream of the expanded G4C2 repeat in the GA reading frame was reported to be important for sense C9 RAN protein expression [41,48,57*]. Additionally, AUG and CUG close-cognate codons are expressed throughout the repeat tracts in SCA31 TGGAA●TTCCA [33] and DM2 (CCTG●CAGG) [14], respectively. Thus, a combination of AUG, close-cognate AUG-like and non-cognate translation initiation mechanisms may occur in different reading frames of repeat expansion transcripts. In DM2, Tusi et al., [42*] showed the alternative, non-canonical, translation initiation factor eIF2A, which can initiate protein translation under conditions of eIF2 inhibition, is required for QAGR expression from CAGG expansion RNAs lacking efficient close-cognate AUG-like initiation codons. In contrast, LPAC RAN proteins expressed from CCUG expansion transcripts are reduced, but not as substantially in eIF2A knock-out cells. These data suggest AUG-like initiation using the canonical eIF2 translation machinery occurs at efficient close-cognate alternative CUG-codons while non-canonical mechanisms requiring eIF2A are used in the alternative CCU, UGC, GCC reading frames that lack AUG-like alternative initiation codons [42*]. Similarly, polyGA RAN protein levels have also been shown to be reduced in eIF2A knockout cells [41,51].

Additionally, Green et al. and Tabet et al. showed evidence that RAN translation of C9orf72 GGGGCC repeats is cap- and eIF4E-dependent [41,48]. In contrast, Cheng et al. demonstrated a cap-independent translation in their reporter system and showed translation efficiency varies for different reading frames [49]. Additional studies are required to elucidate the cap requirements for RAN translation and whether specific repeat motifs, flanking sequences, or physiological conditions favor specific translation mechanisms.

Taken together, RAN translation is a complex process in which multiple factors including RNA structure, repeat motif, flanking sequences, translation initiation factors, and stress kinases can contribute. Additional factors like RNA binding proteins that regulate localization of expanded transcripts could also play roles in regulating RAN translation [14]. Given disease-specific pathological conditions, RAN translation and accumulation of RAN products may be regulated differently between diseases. Understanding how variations in repeat motif and flanking sequences affect RAN protein expression across microsatellite expansion diseases will be important for understanding the molecular mechanisms of these diseases and for the development of drugs to treat these disorders.

Contribution of RAN proteins to disease: insights from human pathology, cell culture and animal studies

Since RAN proteins were first detected in tissues from patients with repeat expansion diseases, there has been a tremendous effort to study the potentially toxic effects of these proteins. Since this time, RAN proteins have been correlated with neuropathological changes in human autopsy tissue and have been shown to alter multiple cellular processes and drive disease independent of RNA gain-of-function effects in animal models.

RAN proteins and neuroinflammation

Overexpression of SCA8, HD and DM2 RAN proteins in glioblastoma (T98) cells leads to increased cell death [3,6,14]. In human autopsy brains for each of these diseases, RAN proteins are associated with markers of neuroinflammation and pathology (e.g. CD68 (DM2), Iba1 and activated Casp3 (HD), astrocytosis (HD and SCA8), were found in brain regions with abundant RAN protein aggregates. Additionally, in SCA8 patient and mouse brains, RAN proteins accumulate in regions with white-matter lesions characterized by demyelination and axonal degeneration [3]. In DM2, sense-strand polyLPAC proteins are found in grey matter regions that show dying cells with pyknotic nuclear staining and in brain regions with organizing necrosis and infiltrating CD68-positive macrophages [14]. In contrast, antisense polyQAGR RAN proteins were primarily found in white matter regions associated with increased astrocytes and activated microglia [14]. The abundant accumulation of RAN proteins in white matter regions in DM2, SCA8, and HD suggests that RAN proteins may underlie the white matter abnormalities found in these disorders [3,6,14]. Additionally, the molecular or environmental conditions in white matter regions may favor the expression of specific types of RAN proteins.

RAN proteins impair UPS, increase stress and induce neurodegeneration and cell death

A number of studies have shown that overexpression of various types of RAN proteins disrupt the ubiquitin-mediated protein degradation system (UPS) and protein homeostasis [56,58,59]. In FXTAS, ubiquitin positive FMR1 polyGly inclusions accumulate in patient tissue and animal models [56] and recruit p62 and 20S proteasomes [58]. Expression of FMR1 CGG expanded repeats were shown to impair the UPS system mainly through the accumulation of polyGly. Overexpression of FMR1 polyGly was shown to increase UPS deficits and exacerbate disease-related phenotypes in Drosophila models, while reducing polyGly production attenuates UPS deficits [59].

In C9orf72 ALS/FTD, AAV delivered GA overexpressing mice develop neurodegenerative phenotypes with motor and cognitive defects [60]. A study in adult flies shows that polyGA but not polyGR or polyPR spreads in fly brains, and the spreading efficiency increases with GA repeat length and age [61]. In a study by LaClair et al., congenic mice that overexpress polyGA but not polyPR trigger neuronal loss, muscle denervation, TDP43 inclusion, and microglial pro-inflammatory responses [62*]. PolyGA proteins have been shown to form ribbon-like structures that trap 26S proteasome complexes [63] and impair UPS-mediated protein degradation [64**]. This leads to protein homeostasis abnormalities, which likely contribute to the accumulation of other C9 RAN proteins [5,63,64**]. GA aggregation also induces ER stress and activation of the PERK pathway, conditions which favor the expression of RAN proteins [48].

While less abundant compared to polyGA and polyGP in C9orf72 ALS/FTD patient tissue, polyPR and polyGR proteins have been shown to be highly toxic in overexpression models [65–68]. Although not under physiological conditions and in the absence of other C9orf72 RAN proteins, a GFP-GR₁₀₀ mouse model showed the accumulation of diffuse, cytoplasmic poly-GR that co-localized with ribosomal subunits and the translation initiation factor eIF3η [65]. These GR overexpression mice develop motor and memory deficits and age-dependent neurodegeneration [65]. PolyGR overexpression in HEK293T cells causes stress granule formation and delayed stress granule disassembly [65]. Similarly, overexpression of GFP-PR₂₈ in homozygous mice led to decreased survival, while heterozygous mice had motor imbalance, reduced motor neuron number, and increased inflammation in the cerebellum and spinal cord [66]. In a separate study, polyPR overexpression in mice was reported to dysregulate histone posttranslational modifications and to alter heterochromatin structure in mice [69]. In a cell culture model, polyPR proteins were shown to localize to nuclear pores and affect their function [70]. Arginine rich RAN proteins have also been shown to interact with disease linked RNA binding proteins and disrupt the formation of phase-separated compartments [71].

Recent studies show the toxicity of C9 RAN proteins is enhanced under the conditions of C9orf72 protein loss [57*,72**]. C9orf72 protein is downregulated in brain tissue from C9orf72 expansion carriers, but the function of this protein is largely unknown. C9orf72 knockout mice show a phenotypes consistent with a compromised immune system including an enlarged spleen and increased cytokines, chemokines and autoantibodies, but these mice do not develop motor neuron dysfunction or other ALS/FTD related features [73]. McCauley et al., provide additional evidence that the C9orf72 protein plays a role in immune function. Myeloid cells from C9orf72^−/− mice have increased type I interferon levels [74*]. Additionally, a type I IFN signature is found in myeloid cells, whole blood, and brain tissue from C9-ALS expansion carriers [74*]. However, a lack of similar phenotypes in immune cells from heterozygous C9orf72^+/− mice suggests that the inflammation seen in C9-patients and C9orf72^−/− mice have different etiologies.

While C9orf72 loss-of-function is not sufficient to cause ALS/FTD, several studies suggest C9orf72 protein loss may be a modifier of disease. For example, Shi et al. showed that C9orf72 overexpression enhanced the clearance of PR₅₀ aggregates in induced pluripotent stem cell-derived motor neurons (iMNs) [75]. Additionally, reduction of C9orf72 protein levels was shown to exacerbate degenerative phenotypes in iMNs overexpressing PR₅₀ and in a C9 mouse model [75]. Similarly, studies by Boivin et al. and McCauley et al., showed that impaired autophagy caused by decreased levels of C9orf72 protein enhanced the accumulation of C9 dipeptide proteins and their toxicity [57*,74*].

RAN proteins recruit trafficking components and disrupt cellular transport

Disruption of nucleocytoplasmic transport is a common feature of a number of microsatellite expansion disorders and may contribute to disease pathogenesis [31,76–79]. In HD, mutant Htt protein was shown to co-localize with nuclear aggregation of RanGAP1 (a nucleocytoplasmic transport factor) and nuclear pore proteins (Nup62 and Nup88) in HD mouse brains [80]. Disruption of nuclear envelope architecture and miscolocalization of transport factors were also observed in HD patients [81]. In FXTAS, FMR1 polyGly was found to recruit LAP2β and disrupt nuclear laminin structure and function, leading to neuronal death [56,82]. In C9orf72 ALS/FTD, poly-GA was shown to recruit trafficking protein Unc119 and other nucleocytoplasmic transport proteins [60]. Zhang et al. reported that overexpression of (GR)₅₀ or (PR)₅₀ induced the formation of stress granules and altered localization of nucleocytoplasmic transport factors resulting in nucleocytoplasmic transport defects in cells [83]. Using cell culture models, Hayes et al. showed polyPR and polyGR proteins disrupt nuclear transport by interfering with importin β, which affects cargo loading of karyopherin mediated nuclear import [84]. A recent study by Fumagalli et al. showed that polyGR and polyPR associate with microtubules and motor proteins in motor neurons and postmortem patient tissues, causing the disruption of microtubule-based transport in cells [85].

RAN proteins and RAN translation as therapeutic targets in disease

Although abundant evidence shows RAN proteins accumulate in patient tissues and cause toxic effects when overexpressed in cells, mice, and other model systems, but a key question is whether or not RAN proteins themselves contribute to disease when expressed under physiologically relevant conditions. Two recent studies have addressed this question using C9orf72 BAC transgenic mice. These single-copy BAC transgenic mice (C9–500), which have ~500 G4C2 repeats and use the endogenous human sense and antisense promoters to drive expression, develop key molecular and clinical features of ALS/FTD including sense and antisense RNA foci, RAN proteins, gait abnormalities and motor neuron loss [25–27*].

Using these C9–500 mice, Nguyen et al., showed that that peripherally delivered antibodies can cross the blood brain barrier, enter cells, and target RAN protein aggregates. The anti-GA1 antibody reduced GA proteins without affecting sense or antisense RNA levels and this passive immunotherapy strategy rescued behavioral deficits, decreased motor neuron loss and increased survival. Interestingly, poly-GP and poly-GR levels were also reduced with anti-GA1 antibody treatment, which these authors showed is likely due to improvements in proteasome function [64**]. These data provide strong support that RAN proteins and not toxic effects of the sense or antisense RNAs are sufficient to drive C9orf72 ALS/FTD. Using a similar immunological approach, Zhou et al., showed poly-GA vaccination prevented motor deficits and reduced microglia activation in a C9orf72 mouse model overexpressing (GA)₁₄₉ [86**]. While the mouse model used in the Zhou et al. study overexpresses only the GA proteins, these findings combined with the Nguyen et al., BAC transgenic mouse study that expresses the cocktail of RAN proteins found in the human disease, suggest both passive immunotherapy and vaccination may be attractive therapeutic strategies for C9orf72 ALS/FTD.

Zu et al., used two alternative approaches to test the effects of inhibiting RAN translation in C9orf72 BAC transgenic mice. In the first set of experiments, these authors showed that inhibiting RAN translation using AAV-delivered PKR-K296R to inhibit PKR improved open-field and DigiGait behavior and reduced RAN protein levels without affecting sense or antisense RNAs. In a separate series of experiments, Zu et al., showed metformin, an FDA approved type II diabetes drug also decreased both p-PKR and RAN protein levels in cells. In C9orf72 BAC transgenic mice, metformin treatment improved Digigait and open-field behavioral deficits, reduced RAN protein levels, decreased neuroinflammation and importantly, improved motor neuron survival compared to untreated C9(+) BAC transgenic controls. Metformin treatment did not affect the levels of sense or antisense RNAs [54**].

Taken together, the demonstration that targeting C9 RAN proteins using antibodies, or decreasing RAN protein production by PKR inhibition, improves disease in C9–500 BAC transgenic mice without affecting sense or antisense RNAs, shows that RAN proteins themselves are an important driver of C9orf72 ALS/FTD and establishes novel therapeutic approaches for C9orf72 ALS/FTD and other RAN-protein diseases.

RAN proteins as disease biomarkers

To date, only a few studies have reported using RAN proteins as disease biomarkers, probably due to the challenges in developing sensitive assays for RAN protein detection in patient samples and limited access to patient bio-fluid samples. In C9orf72 ALS/FTD, polyGP is detectable in C9 patient peripheral blood mononuclear cells (PBMCs), lymphoblastoid B-cell lines (LCLs), induced pluripotent stem cells (iPSCs), and cerebrospinal fluid (CSF) [87,88]. Although no correlation has been found between poly-GP levels and disease progression, poly-GP has been successfully used as a pharmacodynamic biomarker in preclinical studies testing the efficacy of an antisense oligonucleotide approach [87]. Future results from the Biogen/Ionis Pharmaceuticals ASO BIIB078 clinical trial, which is aimed at reducing sense G4C2-containing expansion transcripts will demonstrate if poly-GP is a useful biomarker in C9 clinical trials.

Conclusion and Perspectives

Since the surprising discovery of RAN translation in 2011, RAN proteins have been reported in 11 diseases and this number is likely to continue to grow. Understanding the mechanisms and impact of RAN translation in different disease contexts is leading to a better understanding of the role and impact of RAN proteins in disease (Figure 1). The molecular mechanisms of RAN translation are complex and, depending on sequence context, can involve both canonical and non-canonical translational machinery. The interplay of expansion RNAs, RAN translation, the integrated stress response, and other disease factors are likely to favor RAN protein production and exacerbate disease.

Figure:

Molecular mechanisms and therapeutic strategies for repeat expansion disorders. Molecular deficits in repeat expansion disorders include the accumulation of RAN proteins and impaired protein clearance. These deficits lead to imbalances in protein homeostasis. Promising therapeutic strategies include targeting repeat expansion transcripts, targeting RAN proteins or reducing RAN protein production and increasing protein clearance.

As RAN translation is tightly linked with repeat expansion mutations and nearly two thirds of the human genome is made up of repetitive DNA [89], the existence of novel RAN-related diseases and new types of RAN proteins are almost certain. Additionally, it is possible that repetitive sequences in the genome produce short or long RAN proteins with normal functions, perhaps acting as inhibitory proteins which are specifically expressed under conditions of stress. Although most studies on RAN translation thus far have focused on the central nervous system, RAN proteins have also been reported in peripheral tissues [1,20] and their contributions to disease warrant further study. The accumulation of RAN proteins induces chronic stress and the ISR increases RAN translation, which raises the question of whether conditions of stress including viral infection increase RAN translation and can trigger disease onset.

Highlights.

• RAN translation of a single expansion mutation can produce 6 mutant proteins

• RAN proteins are toxic and accumulate in patient tissues in 11 expansion diseases

• RAN translation is affected by stress and flanking sequence context

• Targeting stress pathways or RAN proteins are attractive therapeutic approaches

• RAN proteins drive disease in C9orf72 mice without changing sense or antisense RNAs