bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

1 A C. elegans model of C9orf72-associated ALS/FTD uncovers a conserved role

2 for eIF2D in RAN translation

3

4

5 Yoshifumi Sonobe 1, 2, 3, Jihad Aburas 1, 3, 4, Priota Islam 5, 6, Tania F. Gendron 7, André E.X.

6 Brown 5, 6, Raymond P. Roos* 1, 2, 3, Paschalis Kratsios* 1, 3, 4

7

8 * These authors contributed equally to this study.

9

10 Correspondence:

11 R.P.R ([email protected]), P.K ([email protected])

12

13 Affiliations: 14 1 University of Chicago Medical Center, 5841 S. Maryland Ave., Chicago, IL 60637 15 2 Department of Neurology, University of Chicago Medical Center, 5841 S. Maryland Ave., 16 Chicago, IL 60637 17 3 The Grossman Institute for Neuroscience, Quantitative Biology, and Human Behavior, 18 University of Chicago, Chicago, IL, USA 19 4 Department of Neurobiology, University of Chicago, Chicago, IL, USA

20 5 MRC London Institute of Medical Sciences, London, UK

21 6 Institute of Clinical Sciences, Imperial College London, London, UK

22 7 Department of Neuroscience, Mayo Clinic, Jacksonville, FL, USA 23

1 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

24 ABSTRACT (150 words) 25 26 A hexanucleotide repeat expansion GGGGCC in the noncoding region of C9orf72 is the most

27 common cause of inherited amyotrophic lateral sclerosis (ALS) and frontotemporal dementia

28 (FTD). Potentially toxic dipeptide repeats (DPRs) are synthesized from the GGGGCC sequence

29 via repeat associated non-AUG (RAN) translation. We developed C. elegans models that express,

30 either ubiquitously or exclusively in neurons, a transgene with 75 GGGGCC repeats flanked by

31 intronic C9orf72 sequence. The worms generate DPRs (poly-glycine-alanine [poly-GA], poly-

32 glycine-proline [poly-GP]) and display neurodegeneration, locomotor and lifespan defects.

33 Mutation of a non-canonical translation-initiating codon (CUG) upstream of the repeats blocked

34 poly-GA production and ameliorated disease, suggesting poly-GA is pathogenic. Importantly,

35 eukaryotic translation initiation factor 2D (eif-2D/eIF2D) was necessary for RAN translation.

36 Genetic removal of eif-2D increased lifespan in both C. elegans models. In vitro findings in

37 human cells demonstrated a conserved role for eif-2D/eIF2D in RAN translation that could be

38 exploited for ALS and FTD therapy.

39

40

41 42 43 44 45 46 47 48 49 50 51 52 53

2 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

54 INTRODUCTION 55 56 The GGGGCC (G4C2) hexanucleotide repeat expansion in the non-coding region of C9orf72 is

57 the most common monogenic cause of inherited amyotrophic lateral sclerosis (ALS) and

58 frontotemporal dementia (FTD)1, 2, and also causes up to 10% of what appears to be sporadic

59 ALS. The G4C2 repeat expands in patients to hundreds or thousands of copies that vary in

60 number in different cells of the same individual. This genetic insult is thought to cause ALS/FTD

61 pathogenesis via three non-mutually exclusive mechanisms: (1) loss of function due to decreased

62 expression of C9ORF72 protein, (2) RNA toxicity from bidirectionally transcribed sense

63 (GGGGCC) and anti-sense (CCCCGG) transcripts, and (3) proteotoxicity from dipeptide repeat

64 (DPR) proteins produced from the expanded nucleotide repeats3.

65 This study focuses on the molecular mechanisms underlying DPR toxicity. Strong

66 evidence suggests that DPRs are toxic in both cell culture and animal models of disease4, 5.

67 Despite the presence of the expanded G4C2 repeat in the non-coding region of the RNA and the

68 absence of an AUG initiating codon, DPRs are translated in all three reading frames from both

69 sense and anti-sense transcripts through a process called repeat-associated non-AUG (RAN)

70 translation4. Poly-glycine-alanine (poly-GA), poly-glycine-proline (poly-GP), and poly-glycine-

71 arginine (poly-GR) are produced from sense transcripts, whereas poly-proline-arginine (poly-PR),

72 poly-proline-alanine (poly-PA), and poly-GP are generated from antisense transcripts6, 7, 8. DPRs

73 are present in neural cells of patients with C9orf72-associated ALS or FTD, indicating that RAN

74 translation occurs in vivo6, 7, 9, 10. A number of potential causes for DPR toxicity have been

75 proposed, including protein aggregation, impaired nucleocytoplasmic transport, and proteasome

76 dysfunction5; however, the relative contribution of each DPR to disease pathogenesis remains

77 unclear.

3 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

78 Recent in vitro and in vivo studies have identified non-canonical translation initiation

79 factors as critical mediators of DPR production. For example, a recent study showed that the

80 ribosomal protein eS25 (RPS25), which is required for efficient translation initiation at the

81 internal ribosome entry site (IRES) of a number of viral and cellular RNAs11, 12, is also required

82 for RAN translation of poly-GP from G4C2 repeats in yeast, fly, and induced pluripotent human

83 stem cell models13. In addition, the eukaryotic translation initiation factor eIF3F, which is

84 thought to control translation of hepatitis C viral (HCV) RNA through binding to the IRES of the

85 viral genome14, regulates poly-GP production from G4C2 repeats in HEK293 cells15. Moreover,

86 a fly study found that eIF4B and eIF4H are important for RAN translation of poly-GR from

87 G4C2 transcripts16. Lastly, knockdown of eIF2A, an unconventional translation initiation factor

88 used by cells under stress conditions17, 18, 19, had a partial role in poly-GA production in HEK293

89 cells and neural cells of the chick embryo20. Collectively, these and other studies21, 22 suggest that

90 multiple factors are involved in RAN translation of DPRs from G4C2 transcripts; yet, their

91 identity and molecular mechanisms of action remain elusive.

92 To identify novel factors involved in RAN translation of DPRs in vivo, we developed

93 two C. elegans models for C9orf72-associated ALS/FTD. C. elegans provides a powerful system

94 for the study of molecular and cellular mechanisms underlying neurodegenerative disorders due

95 to: (1) its short lifespan (~3-4 weeks), (2) its compact nervous system (302 neurons in total), (3)

96 the ability to study neuronal morphology and function with single-cell resolution, (4) the fact that

97 most C. elegans genes have human orthologs, and (5) the ease of genomic engineering and

98 transgenesis, which enables the rapid generation of worms that express human genes, permitting

99 in vivo modeling of neurodegenerative disorders.

100 In this study, we initially developed C. elegans transgenic animals that carry, under the

4 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

101 control of a ubiquitous promoter, 75 copies of the G4C2 repeat flanked by intronic C9orf72

102 sequences. Compared to controls, these animals produce DPRs, namely poly-glycine-alanine

103 (poly-GA) and poly-glycine-proline (poly-GP), and display evidence of neurodegeneration, as

104 well as locomotor and lifespan defects. A second set of transgenic worms expressing the 75

105 G4C2 repeats exclusively in neurons displayed similar phenotypes, indicating that DPR

106 production in this cell type is sufficient to cause disease phenotypes. These C. elegans models

107 provide an opportunity to study in vivo the molecular mechanisms underlying DPR production.

108 Through a candidate approach, we identified a novel role for the non-canonical translation

109 initiation factor eif-2D (ortholog of human eIF2D) in RAN translation of G4C2 repeats. Genetic

110 removal of eif-2D significantly decreased DPR production, and improved locomotor activity and

111 lifespan in both C. elegans models. Supporting the phylogenetic conservation of our C. elegans

112 findings, knock-down of eIF2D in human cells led to decreased DPR production. Together, these

113 observations suggest that eIF2D plays an important role in RAN translation of DPRs from G4C2

114 repeats, opening new directions for treatment of C9orf72-associated ALS and FTD.

115

116

117

118

119

120

121

122

123

5 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

124 RESULTS 125 126 Generation of a new C. elegans model for C9orf72-mediated ALS/FTD

127 The C. elegans genome contains a C9orf72 ortholog (alfa-1), which is involved in lysosomal

128 homeostasis, cellular metabolism and stress-induced neurodegeneration23, 24, 25. However, alfa-1

129 does not contain G4C2 repeats. To study the molecular mechanisms underlying DPR production

130 from G4C2 repeats, we generated worms that carry a transgene encoding 75 copies of the G4C2

131 sequence under the control of a ubiquitous (snb-1) promoter26 (Fig. 1A, see Materials and

132 Methods). These 75 copies are flanked with intronic sequences normally found upstream (113

133 nucleotides) and downstream (99 nucleotides) of the G4C2 repeats in the human C9orf72 gene

134 (Fig. 1A). To monitor expression of poly-GA, the most amyloidogenic DPR27, 28, the transgene

135 carries a nanoluciferase (nLuc) reporter in the poly-GA reading frame. Hereafter, we will refer to

136 these transgenic animals as C9 ubi (Fig. 1A). In parallel, we generated two control C. elegans

137 strains: (a) the UAG ubi worms carry an identical sequence to the C9 animals, but the upstream

138 translation initiation codon CUG, which is required for translation of poly-GA in vitro20, 29, 30, 31,

139 is mutated to UAG, and (b) the ΔC9 ubi worms lack the G4C2 repeats and intronic sequence of

140 human C9orf72 (Fig. 1A). Three independent transgenic lines for C9 ubi, ΔC9 ubi, and UAG ubi

141 worms were generated and tested as described below.

142 First, we performed quantitative PCR with reverse transcription (RT-qPCR) and found

143 that the G4C2 repeats (C9orf72 intronic mRNA) are only transcribed in C9 ubi and UAG ubi

144 animals (Fig. 1B). Next, we tested whether DPRs (poly-GA, poly-GP, poly-GR) from the sense

145 transcript (GGGGCC) are produced in the C9 ubi animals. We focused on these because the 75

146 G4C2 repeats are downstream of the snb-1 promoter, suggesting that the sense transcript may be

147 the one primarily transcribed and translated. Western blotting showed that poly-GA was robustly

6 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

148 expressed in C9 ubi worms when compared to ΔC9 ubi and UAG ubi control animals (Fig. 1C, D).

149 Importantly, mutating the non-canonical initiation codon CUG to UAG did not affect C9orf72

150 mRNA levels, but led to a dramatic decrease in poly-GA expression (Fig. 1B-D). Similar results

151 were obtained with a luciferase assay (Fig. 1E). Corroborating these observations,

152 immunofluorescence staining revealed the presence of poly-GA aggregates in head neurons of

153 C9 ubi animals (Fig. 1F). Although we did not detect poly-GP through Western blotting, a more

154 sensitive ELISA assay detected a modest but statistically significant increase of poly-GP in the

155 C9 ubi worms above the background level observed in ΔC9 ubi controls (Fig. 1G). These

156 background levels likely reflect the abundance of GP motifs in the C. elegans proteome.

157 Although we were not able to detect poly-GR in the C9 ubi worms (Supplementary Figure 1A),

158 our overall results demonstrate that RAN translation occurs in C. elegans.

159 Importantly, we found that the median survival of C9 ubi worms is significantly shorter

160 when compared to ΔC9 ubi and UAG ubi controls (P < 0.0001) (Fig. 1H), presumably due to poly-

161 GA- and/or poly-GP-associated toxicity. Together, these findings suggest that at least two

162 distinct DPRs (poly-GA and poly-GP) are generated from the 75 G4C2 repeats in the C9 ubi

163 animals and likely contribute to the reduced survival of these animals, thereby establishing a new

164 C. elegans model for C9orf72-mediated ALS/FTD.

165

166 C9 ubi animals display progressive motor neuron degeneration and neuromuscular synapse

167 defects

168 At the level of animal behavior, we observed striking locomotion defects in C9 ubi worms, which

169 are presented later in Results. Such defects could be attributed, at least partially, to degeneration

170 of ventral nerve cord motor neurons, which are essential for C. elegans locomotion. To test this

7 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

171 possibility, we fluorescently labeled the cholinergic motor neurons present in the nerve cord of

172 C9 ubi and ΔC9 ubi animals with the unc-17/VAChT::gfp transgenic reporter, which allows for

173 single-cell resolution in our analysis (Fig. 2A-B). Next, we quantified the number of cholinergic

174 motor neurons (unc-17/VAChT::gfp positive) in C9 ubi and ΔC9 ubi animals at juvenile (larval

175 stage 4, L4) and adult (Day 2 and 5) stages. Although no differences were observed at the

176 juvenile stage, we found a significant reduction in the number of motor neurons that populate the

177 nerve cord of C9 ubi animals (Fig. 2C), arguing for progressive motor neuron degeneration. Of

178 note, we cannot exclude the possibility that, in addition to cholinergic motor neurons, other

179 neuron types also degenerate in C9 ubi animals.

180 We next sought to determine whether the observed motor neuron degeneration is

181 accompanied by neuromuscular synapse defects, an early hallmark in ALS disease progression32.

182 To this end, we crossed the C9 ubi worms to animals carrying a fluorescent reporter (Punc-

183 3::rab-3::eGFP) that specifically labels the presynaptic boutons of cholinergic motor neurons

184 (Fig. 2D). We found that the number of presynaptic boutons is significantly decreased in C9 ubi

185 animals during adult (Day 2 and 4) stages, but no differences were observed between C9 ubi and

186 ΔC9 ubi animals at a juvenile stage (L4) (Fig. 2E). In summary, these findings provide evidence

187 for motor neuron degeneration and neuromuscular synapse defects in C9 ubi animals.

188

189 DPR production selectively in neurons is sufficient to cause disease phenotypes in vivo

190 The DPRs (poly-GA, poly-GP) in C9 ubi animals are produced ubiquitously because the snb-1

191 promoter is active in all C. elegans cell types (Fig. 1A)26. To test whether DPR production from

192 the G4C2 repeats exclusively in neurons is sufficient to cause disease-associated phenotypes, we

193 generated a second set of transgenic animals by replacing the snb-1 promoter with a fragment of

8 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

194 the unc-11 promoter (unc-11c) known to be exclusively active in all C. elegans neurons26. We

195 refer to these animals as C9 neuro, ΔC9 neuro, and UAG neuro (Fig. 3A). First, we confirmed by RT-

196 qPCR that the G4C2 repeat RNA (with C9orf72 intronic mRNA) and the nLuc mRNA are

197 transcribed in C9 neuro and UAG neuro animals (Fig. 3B-C). Similar to C9 ubi animals, we found

198 through Western blotting that poly-GA is robustly produced in C9 neuro animals, but not in

199 control animals lacking the G4C2 repeats (ΔC9 neuro) or having the non-canonical translation

200 initiation codon CUG mutated to UAG (UAG neuro) (Fig. 3D-E). However, we were not able to

201 detect poly-GR by ELISA in the C9 neuro animals (Supplementary Figure 1B). Importantly, we

202 found a dramatic decrease in the median survival of C9 neuro animals when compared to ΔC9 neuro,

203 and UAG neuro controls (p value < 0.0001) (Fig. 3F). The effect on median survival in C9 neuro

204 animals appears stronger when compared to C9 ubi animals (Fig. 1H, 3F); this is likely due to

205 differential strengths of the snb-1 and unc-11c promoters, resulting in different amounts of DPRs.

206 Our findings suggest that DPR (poly-GA) production from the G4C2 repeats exclusively in

207 neurons is sufficient to trigger disease-associated phenotypes, establishing C9 neuro animals as a

208 model for C9orf72-mediated ALS/FTD.

209

210 Genetic removal of the eukaryotic translation initiation factor 2D (eif-2D/eIF2D) reduces

211 poly-GA production and ameliorates lifespan defects in C9 ubi and C9 neuro animals

212 The C9 ubi and C9 neuro worm models for C9orf72-mediated ALS/FTD offer a unique opportunity

213 to uncover the molecular mechanisms responsible for RAN translated DPRs from G4C2 repeats.

214 Our results strongly suggest that RAN (non-AUG dependent) translation of poly-GA in both in

215 vivo models (C9 ubi and C9 neuro worms) requires the non-canonical initiation codon CUG. We

216 therefore surveyed the literature for factors known to initiate translation via this codon. Two

9 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

217 proteins, the eukaryotic translation initiation factors eIF2A and eIF2D, are known to deliver

218 tRNAs to the CUG codon and initiate translation in vitro33, 34. A single ortholog for each factor is

219 present in the C. elegans genome, that is, eif-2A for human eIF2A and eif-2D for human eIF2D.

220 We therefore questioned whether eif-2A and/or eif-2D are necessary for RAN translation of

221 DPRs from the G4C2 repeats. To test this, we obtained mutant animals that carry strong loss-of-

222 function (putative null) alleles for eif-2A (gk358198) or eif-2D (gk904876). Since these animals

223 are viable and fertile, we crossed them to the C9 ubi worms and generated eif-2A (gk358198); C9

224 ubi and eif-2D (gk904876); C9 ubi animals. Next, we found via RT-qPCR that the G4C2 repeat

225 RNA with flanking C9orf72 intronic sequence, as well as the nLuc mRNA are normally

226 transcribed in both eif-2A (gk358198); C9 ubi and eif-2D (gk904876); C9 ubi animals (Fig. 4A-B),

227 indicating that loss of eif-2A or eif-2D does not affect transcription or stability of the repeat RNA.

228 However, we found a significant decrease in DPR (poly-GA) production in eif-2D (gk904876);

229 C9 ubi animals but not in eif-2A (gk358198); C9 ubi animals (Fig. 4C-D), suggesting that eif-2D is

230 required for RAN translation. eif-2D does not appear to play a role in canonical translation

231 because loss of eif-2D gene activity in worms did not affect expression of an AUG-initiated

232 green fluorescent protein (GFP) transgene (Supplementary Figure 2).

233 Since eif-2D is necessary for translation of poly-GA, which is thought to be toxic, loss of

234 eif-2D gene activity may improve survival in C9 ubi animals. Indeed, we found that the eif-2D

235 (gk904876); C9 ubi animals display a normal lifespan, similar to ΔC9 ubi animals (Fig. 4E).

236 Eukaryotic translation initiation factors are expressed broadly in many tissues, and this appears

237 to be the case for eif-2D in C. elegans

238 (https://bgee.org/?page=gene&gene_id=WBGene00016113) and its human ortholog

239 (https://www.proteinatlas.org/ENSG00000143486-EIF2D/tissue). To test whether neuronal eif-

10 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

240 2D gene activity is critical for poly-GA production and animal survival, we crossed the eif-2D

241 (gk904876) mutants to the C9 neuro animals that selectively produce poly-GA in neurons (Fig.

242 3D). Although transcription of the G4C2 repeats (C9orf72 intronic mRNA) increased in eif-2D

243 (gk904876); C9 neuro animals for unclear reasons (Fig. 3F-G), we found a dramatic decrease in

244 poly-GA production in these worms compared to C9 neuro worms (Fig. 3H-I). Of note, genetic

245 removal of eif-2D in C9 neuro animals resulted in a more prominent decrease in poly-GA levels

246 than in C9 ubi animals (compare Fig. 4C-D with Fig. 4H-I), suggesting that eIF2D may have a

247 greater impact on RAN translation in neurons than in other cell types. Importantly, loss of eif-2D

248 restored survival of C9 neuro animals (P value < 0.0001 between eif-2D (gk904876); C9 neuro and

249 C9 neuro) to levels similar to ΔC9 neuro and eif-2D (gk904876) control animals (Fig. 3J). In

250 summary, we found that eif-2D/eIF2D is critical for poly-GA production, and its genetic removal

251 mitigates the shortened lifespan otherwise observed in both C9 ubi and C9 neuro animals.

252

253 Genetic removal of eif-2D/eIF2D ameliorates the locomotor defects of C9 ubi animals

254 Because loss of eif-2D reduces poly-GA production and improves animal survival (Fig. 4), we

255 next asked whether the locomotor defects of C9 ubi animals can be ameliorated upon genetic

256 removal of eif-2D. To this end, we performed a high-resolution behavioral analysis of freely

257 moving adult (day 2) C. elegans using automated worm tracking technology35, 36. This analysis

258 can quantitate several features related to locomotion (e.g., velocity, crawling amplitude, body

259 curvature). We found multiple locomotor defects in C9 ubi animals, such as reduced velocity and

260 impaired body curvature (Fig. 5, Supplementary Figure 3). These defects were not present in

261 ΔC9 ubi and UAG ubi controls indicating that a hallmark of ALS, namely motor deficits, is

262 recapitulated in the C9 ubi animal model. Importantly, we performed the same tracking analysis

11 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

263 on eif-2D (gk904876) mutants and eif-2D (gk904876); C9 ubi animals. This analysis revealed that

264 loss of eif-2D significantly improved the locomotor defects of C9 ubi animals (Fig. 5,

265 Supplementary Figure 3). Of note, eif-2D (gk904876) single mutants did not display any

266 prominent locomotor defects (Fig. 5). In summary, this behavioral analysis complements our

267 molecular analysis (poly-GA detection) and survival assays (Fig. 4), providing compelling

268 evidence that loss of eif-2D ameliorates the pathogenic phenotype associated with RAN

269 translation of C9orf72 G4C2 nucleotide repeats.

270

271 Knock-down of eIF2D suppresses DPR production in human cells

272 We next sought to determine whether the function of eif-2D/eIF2D in RAN translation and DPR

273 production is conserved from worms to human cells. To this end, we transfected HEK293 cells

274 with a bicistronic construct, in which the firefly luciferase (fLuc) gene is located in the first

275 cistron followed by a second cistron containing the 75 G4C2 repeats along with intronic

276 sequences that normally flank the repeats in the human C9orf72 locus (Fig. 6A). To monitor

277 expression of different DPRs, we generated three different versions of this construct by inserting

278 a nanoluciferase (nLuc) reporter gene in the poly-GA, poly-GP, or poly-GR reading frame (Fig.

279 6A) (see Materials and Methods). Importantly, previous studies showed that DPR translation can

280 occur when the G4C2 repeats are located in the second cistron of a bicistronic construct20, 37.

281 Each bicistronic construct was co-transfected with plasmids carrying either control short hairpin

282 RNA (shRNA) or shRNA against eIF2D. First, we assessed the eIF2D protein levels via Western

283 blotting and observed a close to 50% reduction in HEK293 cells that received the eIF2D shRNA

284 compared to cells transfected with control shRNA (Fig. 6B-C). Next, we performed Western

285 blotting and luciferase assays and found that, similar to our C. elegans model, poly-GA and poly-

12 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

286 GP peptides are produced in HEK293 cells upon transfection (Fig. 6B, D-E). Moreover, the

287 increased sensitivity of the luciferase assay enabled us to detect poly-GR in these cells, albeit at

288 lower levels compared to poly-GA (Fig. 6E). Importantly, knock-down of eIF2D by shRNA

289 dramatically reduced (by at least 75%) the expression of all three DPRs (poly-GA, poly-GP,

290 poly-GR), suggesting eIF2D is required for RAN translation of GA, GP, and GR (Fig. 6B, D-E).

291 In contrast, the protein levels of fLuc, which is expressed by the first cistron via canonical AUG-

292 dependent translation, were similar in cells transfected with either control or eIF2D shRNAs (Fig.

293 6B, F), suggesting a specific role for eIF2D in unconventional translation in human cells. In

294 summary, these results suggest that eIF2D is necessary for DPR production (poly-GA, poly-GP,

295 poly-GR) in human cultured cells, uncovering a conserved role for eIF2D in RAN translation of

296 DPRs from G4C2 repeat-containing RNA.

297

298

299

300

301

302

303

304

305

306

307

13 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

308 DISCUSSION

309 The discovery of RAN translation of C9orf72 nucleotide repeats combined with the reported

310 toxicity of DPRs in model systems have focused the ALS/FTD field on this unconventional form

311 of translation, as well as on DPRs as potential therapeutic targets. To study the molecular

312 mechanisms underlying RAN translation, we initially developed a C. elegans model that carries

313 a transgene encoding 75 copies of the G4C2 repeat under the control of a ubiquitous promoter.

314 These C9 ubi animals generate DPRs, namely poly-GA and poly-GP, show signs of

315 neurodegeneration, and display locomotor and lifespan defects. Such phenotypes are likely due

316 to DPR toxicity because mutation of a non-canonical CUG initiation codon decreased poly-GA

317 production and ameliorated locomotor and lifespan defects in C9 ubi animals without affecting

318 C9orf72 mRNA levels. Moreover, a second C. elegans model (referred to as C9 neuro) that

319 generates poly-GA exclusively in neurons displayed a similar phenotype. Importantly, we found

320 that knock-out of the eukaryotic translation initiation factor eif-2D/eIF2D suppressed poly-GA

321 expression and ameliorated lifespan defects in both C9 ubi and C9 neuro models. Lastly, our in

322 vitro findings in human cells combined with our C. elegans data argue for a conserved role of

323 eif-2D/eIF2D in RAN translation that could open up new therapeutic directions for ALS and

324 FTD.

325

326 The C9 ubi and C9 neuro worm models offer insights into DPR toxicity

327 An unresolved issue in the field of ALS and FTD research is the relative contribution of each of

328 the five DPRs (poly-GA, poly-GP, poly-GR, poly-PR, poly-PA) to disease pathogenesis. In this

329 paper, two lines of evidence indicate that poly-GA, the most readily visible DPR in the brain and

330 spinal cord of ALS/FTD patients with C9orf72 expanded repeats38, 39, significantly contributes to

14 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

331 disease phenotypes in C. elegans: (1) poly-GA aggregates and appears to be the most abundantly

332 expressed DPR in our worm models, and (2) mutation of the non-canonical CUG initiation

333 codon decreased poly-GA production and ameliorated locomotor and lifespan defects in C9 ubi

334 and C9 neuro animals. Importantly, these in vivo results significantly extend previous reports

335 demonstrating that this same CUG is required in vitro for poly-GA translation 20, 29, 30, 31. In

336 support of our C. elegans findings, recent studies in mice reported neurodegeneration, motor

337 deficits, and cognitive defects following ubiquitous expression of poly-GA40, as well as

338 behavioral abnormalities following spinal cord expression of poly-GA41. A prevailing hypothesis

339 is that poly-GA causes cellular toxicity through proteosomal impairment and protein

340 sequestration5. A recent publication reports that in an in vitro system poly-GA spreads cell-to-

341 cell and inhibits proteasomal function both cell-autonomously and non-cell-autonomously,

342 leading to ubiquitination of the nuclear localization signal of TDP-43 and its subsequent

343 cytoplasmic mislocalization42. Importantly, treatment with an antibody against poly-GA

344 ameliorated TDP-43 mislocalization42, as well as toxicity in mouse models for C9orf72-

345 associated ALS/FTD43. However, a zebrafish model that produces poly-GA did not exhibit motor

346 axon toxicity44, presumably due to low poly-GA expression.

347 Although our C9 ubi worm model produces a small amount of poly-GP, this DPR is

348 thought to be relatively nontoxic5. In contrast, the toxicity of arginine-containing DPRs,

349 especially poly-GR, has been demonstrated in mouse neurons45. In C. elegans, codon-optimized

350 constructs driving expression of poly-GR or poly-PR caused toxicity in muscles and neurons46.

351 Furthermore, initiation of poly-GR translation has been reported to occur at the same upstream

352 CUG codon used for poly-GA production, but with subsequent frameshifting into the GR reading

353 frame30. However, two recent papers showed that translation initiation of poly-GR did not occur

15 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

354 from the upstream CUG codon31, 47. Although we could not detect poly-GR by Western blotting

355 and ELISA (Supplementary Figure 1), it is possible that small amounts of poly-GR are

356 produced in our C. elegans models and contribute to the observed disease phenotypes.

357 Supporting this notion, very low levels of poly-GR in neurons have been reported to impair

358 synaptic function and induce behavioral deficits in both Drosophila and mice45.

359 In addition to the possibility of DPR toxicity in our C. elegans models, there remains

360 the question as to whether the repeat RNA itself rather than, or in addition to, its protein product

361 could be toxic48. A role for RNA toxicity does not appear to be the case in our nematode

362 experiments since mutation of the non-canonical initiation codon CUG to UAG led to a

363 significant decrease in poly-GA expression and rescue of worm survival to wild type levels, but

364 had no effect on C9orf72 mRNA levels. However, it is possible that the CUG > UAG mutation,

365 albeit minimal in nature, can affect RNA structure.

366 One question that arises is whether our transgenic C9 ubi and C9 neuro animals constitute

367 an appropriate model for the C9orf72 patient. It seems clear that the RNA template (repeat RNA

368 plus flanking intronic sequence) produced in these animals differs from the RNA template

369 actually expressed in the central nervous system of patients with a C9orf72 expanded repeat

370 because it is likely that there are varying RNA templates of different lengths and sequences in

371 these patients. If this is indeed the case, it suggests that a number of potential toxicities besides

372 one related to poly-GA production may be present in the human disease state, including toxicity

373 related to repeat RNA and/or synthesis of other DPRs. Toxicity may also occur as a consequence

374 of a decrease in the authentic C9orf72 gene product. Supporting this possibility, a recent study

375 reported that a decrease in C9ORF72 protein leads to a decrease in autophagic clearance of toxic

376 accumulations of the DPRs31.

16 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

377 A novel role for eif-2D/eIF2D in RAN translation of C9orf72-associated nucleotide repeats

378 eIF2D is known to initiate translation in vitro at both AUG and non-AUG initiation codons33, 49.

379 The present study demonstrates a key role of eIF2D in RAN translation of poly-GA since eIF2D

380 knock-down in HEK293 cells, as well as eif-2D knock-out in C. elegans led to a prominent

381 decline in poly-GA expression. It is noteworthy that genetic removal of eif-2D in the worm had

382 no effect on: (a) canonical translation of an AUG-initiated GFP reporter, (b) animal locomotion,

383 and (c) lifespan. The absence of detrimental phenotypes in C. elegans animals that globally lack

384 eif-2D/eIF2D make this factor an attractive therapeutic target. Moreover, we found that removal

385 of eif-2D led to a greater decrease in DPR (poly-GA) production in C9 neuro animals versus C9 ubi

386 animals (Figure 4D, I), suggesting that eIF2D may be regulating RAN translation in varying

387 ways in different cell types.

388 Our previous study demonstrated that knock-down of eIF2A in HEK293 cells and neural

389 cells in the chick embryo led to a modest (30%) decrease in poly-GA expression from G4C2

390 expanded repeats20. In the present study, however, eif-2A/eIF2A deficiency had no effect on

391 poly-GA expression in C. elegans, while eIF2D knock-down in HEK293 cells and eif-2D knock-

392 out in C. elegans had a prominent effect on poly-GA expression, suggesting eIF2D plays a more

393 important role in RAN translation than eIF2A. In contrast to the slight decrease of a single DPR

394 (poly-GA) we previously found with eIF2A knock-down20, knock-down of eIF2D in HEK293

395 cells dramatically reduced (by at least 75%) the expression of multiple DPRs (poly-GA, poly-GP

396 and poly-GR), indicating a critical role for eIF2D in RAN translation in all three reading frames.

397 It is possible, however, that eIF2A does play a significant role in RAN translation, but only upon

398 cellular stress since it can translate a subset of mRNAs when eIF2α is phosphorylated17, 18, 50.

17 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

399 In addition to eIF2D, RPS25 is the only other factor known to be required for the

400 expression of these three DPRs, albeit to variable degrees, i.e., RPS25 knock-down in vitro

401 reduced poly-GA by 90%, poly-GP by 50%, and poly-GR by 30%13. Collectively, these and

402 other studies16 suggest that a number of different factors play critical roles in RAN translation.

403 Although the present study uncovered a requirement for eif-2D/eIF2D in RAN translation,

404 this protein has additional functions besides translation initiation. For example, eIF2D was

405 implicated in recycling 40S ribosomal subunits51, 52, 53, which enabled ribosomes and mRNAs to

406 participate in multiple rounds of translation, leading to efficient protein expression. It is thus

407 possible that genetic removal of eif-2D in our worm models caused inefficient DPR production

408 by blocking recycling 40S ribosomal subunits.

409

410 RAN translation occurs in C. elegans and offers the opportunity to discover conserved

411 regulators

412 Two previous studies reported DPR-associated toxicity in C. elegans by using either a codon-

413 optimized construct that carries no G4C2 repeats, or a construct that carries 66 G4C2 repeats22, 46.

414 The latter study strongly indicated that repeat associated non-AUG (RAN) translation occurs in

415 the worm. However, the non-canonical (non-AUG) initiating codon and translation initiation

416 factor remained elusive. By generating C. elegans animals that carry 75 G4C2 repeats flanked by

417 intronic C9orf72 sequences and subsequently blocking poly-GA expression by mutating a non-

418 canonical initiation codon (CUG) upstream of the repeats, we advanced the molecular

419 understanding of how RAN translation occurs in the worm. This finding in C. elegans, together

420 with previous reports in yeast and flies 16, 22, strongly suggest that the molecular machinery

18 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

421 required for RAN translation is evolutionary conserved, offering an opportunity to use simple

422 model systems for the discovery of RAN regulators.

423 In summary, we leveraged the specific strengths of the C. elegans model and

424 discovered that the non-canonical translation initiation factor eif-2D/eIF2D is required for RAN

425 translation of expanded G4C2 repeats in the C9orf72 gene. Our in vitro findings argue for a

426 conserved role of eif-2D/eIF2D in RAN translation, potentially offering new directions for

427 C9orf72-associated ALS and FTD therapy. In fact, because RAN translation of other nucleotide

428 repeat expansions (not G4C2) occurs in several neurodegenerative diseases4, eIF2D may also

429 prove to be a promising therapeutic target in these disorders. Determining how broadly eIF2D

430 influences RAN translation is thus an important future endeavor.

431

432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455

19 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

456 METHODS 457 458 C. elegans strains

459 Worms were grown at 15˚C, 20˚C or 25˚C on nematode growth media (NGM) plates supplied

460 with E. coli OP50 as food source54. The mutant strains used in this study are eif-2A (gk358198)

461 and eif-2D (gk904876). The PCR primers used to genotype these mutant strains are shown in

462 Supplementary Table 1. In addition, we used animals carrying the vsIs48 [unc-17::gfp] and

463 otIs437 [Punc-3::eGFP::rab-3cDNA::unc-10_3'UTR, + ttx-3::mCherry] transgenes.

464

465 Generation of C9 ubi and C9 neuro transgenic animals

466 The template for the generation of the plasmids used in the C9 ubi, ΔC9 ubi, and UAG ubi animals

467 have been described in an in vitro study20. In C9 ubi animals, the plasmid has 75 G4C2 repeats

468 with 113 nucleotides 5' and 99 nucleotides 3' of intronic sequences that normally flank the G4C2

469 repeats in the human C9orf72 gene; this sequence is followed by nanoluciferase (nLuc) reporter

470 in the reading frame of poly-GA, which is followed by the unc54 3' UTR. In ΔC9 ubi animals, the

471 construct is identical but lacks the G4C2 repeats. In UAG ubi animals, the construct is identical

472 but the non-canonical translation initiating codon CUG is mutated to UAG. In C9 ubi, ΔC9 ubi,

473 UAG ubi animals, the following promoter sequence of the snb-1 gene was used:

474 cagttcgggtatctcagcaaccatgacgtgtgttcaagtatttcattttttctccctttctaggaaaattcaagttgaatttttattcattttgaaacttc

475 ctctcatctctcgaagatatgcctcgggcttcaaattcgtagtccacccgttgcacgtaacctgctgaaaacttccttgtattcagttctctcgtttc

476 gttttgtctgcaaaagaatttagctccatttgtgtgaaatcgagcaaaatcatgcggctagagaaaataagaaaaaaaaacgaaagagaactg

477 gagcgtgtgcatgggacattctcctttcttcgccttattcttcgtctcccattattagtttcgcctcccacaattctggctgcaaagcaataaaaagt

478 ggcttcgttatttgtctggaatgcgtcccactgtatgtgttgtcttctatttaatacaagttataacctccactcgcttttttttttattttttgactgcctct

479 tggtaaaaacgtaatcatacaacagcagtgaaaaccagttttttgaaaaccgtcccgtcattttttttctatttatcatttcaatttatttattaatcgat

20 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

480 gattgaaagtgaatggatgacggtcatgaccgattatcgattatcccgaaatagagatgcgcgtaggtcataatgcccagtacgcaaaatgtt

481 ttatcggtgtttgcacagatttcgcaacatctctcattgaatttccattcatcgcttcgtcatctgaccccatttcttattttttcatccttttccctgttct

482 catcgttccttactattttcctaatttcagaacatcgcgattttataatttcgttaaatattcgtaatcccgttatacaaaaatagctaaattttctagtc

483 gttctcgtttttgagagggcactttagtccgtcatcgtgtcgcttgtcgtgctcaatttttcatgcataaatgggcgtcgccgtcccccctgtcgttt

484 tcttcctttacctcactttccagttctgaattccgatacgaatttttaaatttttctaactcgcttcatttcagggaacagccggataagaccatcttga

485 cgac.

486 In C9 neuro, ΔC9 neuro, UAG neuro animals, the following promoter sequence of the unc-11

487 gene was used:

488 ctgtgtctctccgtcctatccactatctttcatccagattcccggttttcccgagaaaaagtgagagagagaagaggctactgcgcggcgaca

489 ctggctgcggacaccctcactcctctacttgtctccatcgattttcgctggttcgtcgaagcctcgtcgccaacacaacttcttcttctttttcgtcc

490 tcttcttttctgccacacgcttgcacagcttcttcttctcattttacgctgtctacttttttcggaggggccatacccccacacacagctccttttattt

491 gg.

492 The unc54 3' UTR was cloned into the XhoI-SacI site of the plasmids using the primer

493 sets shown in Supplementary Table 1. Fig. 1A shows the constructs with the snb-1 promoter and

494 the upstream CUG, which is the translation initiation codon for poly-GA (see Results).

495 The C9, UAG, or ΔC9 plasmids (100 ng/μl), along with a myo-2:GFP plasmid (3

496 ng/μl), and pBlueScript plasmid (20 ng/μl) were microinjected into the C. elegans gonad.

497 Multiple transgenic lines per construct were established. The resulting C9 ubi, ΔC9 ubi and UAG

498 ubi animals, as well as the resulting C9 neuro, ΔC9 neuro and UAG neuro animals, carried these

499 transgenes as extrachromosomal arrays. The exact extrachromosomal lines used in each

500 experiment are indicated in Figure legends. The following extrachromosomal lines were

501 generated for this study: kasEx153 [Psnb-1::C9 + myo-2::GFP line 3], kasEx154 [Psnb-1:: ΔC9

502 + myo-2::GFP line 1], kasEx155 [Psnb-1::UAG + myo-2::GFP line 1], kasEx156 [Punc-11::C9

21 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

503 + myo-2::GFP line 2.1], kasEx157 [Punc-11::C9; myo-2::GFP line 11.1], kasEx158 [Punc-11::

504 ΔC9 + myo-2::GFP line 3.1], kasEx159 [Punc-11:: ΔC9 + myo-2::GFP line 4.2], kasEx160

505 [Punc-11::UAG + myo-2::GFP line 4.1], and kasEx161 [Punc-11::UAG + myo-2::GFP line

506 8.1]. For the C9 ubi and ΔC9 ubi extrachromosomal lines, we also established integrated versions

507 into the C. elegans genome: kasIs6 [Psnb-1::C9 + myo-2::GFP line 1.9], kasIs7 [Psnb-1::C9 +

508 myo-2::GFP line 12.3], kasIs8 [Psnb-1:: ΔC9 + myo-2::GFP line 2.3], and kasIs9 [Psnb-1::

509 ΔC9 + myo-2::GFP line 7.7].

510

511 Lifespan assay

512 The lifespan assay was performed as previously described22. Worms were transferred at the L4

513 larval stage to NGM plates containing 50 μM 5-fluoro-2'deoxyuridine (FUDR, Sigma). The day

514 of the timed egg laying was considered day 0 for lifespan analysis.

515

516 Microscopy

517 Worms were anesthetized using 100mM of sodium azide (NaN3) and mounted on 5% agarose on

518 glass slides. Images were taken using an automated fluorescence microscope (Zeiss, Axio Imager

519 Z2). Acquisition of several z-stack images (each ~1 μm) Zeiss Axiocam 503 mono using the

520 ZEN software (Version 2.3.69.1000, Blue edition). Representative images are shown following

521 max-projection of 2-5 µm Z-stacks using the maximum intensity projection type. Image

522 reconstruction was performed using Image J software55.

523

524 Immunocytochemistry

22 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

525 C9 ubi and ΔC9 ubi animals were grown at 20 °C on nematode growth media (NGM). We followed

526 an immunocytochemical staining procedure described previously56. In brief, worms were

527 prepared for staining following the freeze-crack procedure and they were subsequently fixed in

528 ice-cold acetone (5 min) and ice-cold methanol (5 min). Worms were transferred using a Pasteur

529 pipette from slides to a 50mL conical tube that contained 40mL 1XPBS. Following a brief

530 centrifugation (2min, 3000rpm), worms were pelleted and 1xPBS was removed. Next, the worms

531 were incubated with 300 μl of blocking solution (1XPBS, 0.2% Gelatin, 0.25% Triton) for 30min

532 at room temperature (rolling agitation). Following removal of the blocking solution, worms were

533 incubated over-night with a mouse anti-poly-GA antibody [1:25 in PGT solution, EMD

534 Millipore]. Next, the primary antibody solution was removed and worms were washed five times

535 with washing solution (1XPBS, 0.25% Triton). Worms were incubated with a Alexa Fluor 594

536 goat anti-mouse IgG secondary antibody (1:1000 in PGT solution, A-11020, Molecular Probes)

537 for 3 hours at room temperature. Following 5 washes, worms were mounted on a glass slide and

538 examined at an automated fluorescence microscope (Zeiss, AXIO Imager Z1 Stand).

539

540 Worm tracking

541 Worms were maintained as mixed stage populations by chunking on NGM plates with E. coli

542 OP50 as the food source. Worms were bleached and the eggs were allowed to hatch in M9

543 buffer to arrest as L1 larvae. L1 larvae were refed on OP50 and allowed to grow to day 2 or

544 adulthood. On the day of tracking, five worms were picked from the incubated plates to each of

545 the imaging plates (see below) and allowed to habituate for 30 minutes before recording for 15

546 minutes. Imaging plates are 35 mm plates with 3.5 mL of low-peptone (0.013% Difco Bacto)

547 NGM agar (2% Bio/Agar, BioGene) to limit bacteria growth. Plates are stored at 4oC for at least

23 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

548 two days before use. Imaging plates are seeded with 50μl of a 1:10 dilution of OP50 in M9 the

549 day before tracking and left to dry overnight with the lid on at room temperature.

550

551 Behavioral feature extraction and analysis

552 All videos were analyzed using Tierpsy Tracker to extract each worm’s position and posture over

553 time35. These postural data were then converted into a set of behavioral features selected from a

554 large set of features as previously described35. For each strain comparison, we performed

555 unpaired two-sample t-tests independently for each feature. The false discovery rate (FDR) was

556 controlled at 5% across all strain and feature comparisons using the Benjamini Yekutieli

557 procedure. The p-value threshold for this analysis (FDR 5%) is 0.0014.

558

559 In vitro studies

560 Bicistronic plasmids were constructed with firefly luciferase (fLuc) in the first cistron and C9 in

561 the second cistron, as previously published20.

562 shRNA plasmids against eIF2D were prepared using previously published methods20. In

563 brief, oligonucleotides with an siRNA sequence (Supplementary Table 1) were cloned into the

564 BamHI and Hind III sites of psilencer 2.1-U6 neo vector (Thermo Fisher Scientific) according to

565 the manufacturer’s protocol. The control shRNA vector was provided by this vector kit.

566 For Western blotting (see below), HEK293 cells were planted in 6-well plates at 2 x 105

567 per well and cotransfected with 0.5 μg bicistronic plasmids along with either 2.5 μg control

568 shRNA or anti-eIF2D shRNA using Lipofectamine LTX (Thermo Fisher Scientific). For

569 luciferase assays (see below), the cells were plated in 24-well plates at 5 x 104 per well and

570 cotransfected with 0.1 μg bicistronic plasmids along with either 0.5 μg control shRNA or anti-

24 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

571 eIF2D shRNA using Lipofectamine LTX (Thermo Fisher Scientific). The cells were cultured for

572 48h in DMEM supplemented with 10% FBS, 2 mM L-Glutamine, 100 U/ml Penicillin and 100

573 μg/ml Streptomycin, and then processed.

574 575 Western blotting

576 Lysates of worms were prepared by sonication in buffer consisting of 4% urea and 0.5% SDS

577 with 1 x HaltTM Protease and Phosphatase inhibitor Cocktail (Thermo Fisher Scientific).

578 HEK293 cell lysates were prepared using RIPA buffer consisting of 150 mM NaCl, 1% Nonidet

579 P-40, 0.5% sodium deoxycholate, 0.1% SDS, 25 mM Tris-HCl (pH7.4). These lysates were

580 subjected to electrophoresis on 4-20% SDS polyacrylamide gels (Mini-PROTEAN TGX Gels,

581 BIO-RAD), and then transferred to Amersham Hybond P 0.45μm PVDF membranes (GE

582 Healthcare). The membrane was blocked with 1% non-fat skim milk in Tris-buffered saline

583 containing 0.05% Tween-20 for 1h at room temperature, and then incubated overnight at 4°C

584 with primary antibodies against poly-GA (1:1000, MABN889, EMD Millipore), fLuc (1:1000,

585 G7451, Promega), eIF2D (1:2000, ab108218, Abcam) or tubulin (1:5000, YL1/2, Abcam).

586 Following washing, the membrane was incubated for 1h at room temperature with anti-mouse

587 (1:5000, GE Healthcare), anti-rabbit (1:5000, GE Healthcare) or anti-rat horseradish peroxidase–

588 conjugated secondary antibodies (1:1000, Cell Signaling Technology). The signal was detected

589 using SuperSignal West Dura Extended Duration Substrate (Thermo Fisher Scientific) and

590 analyzed using ChemiDoc MP Imaging System (BIO-RAD).

591

592 Luciferase assay

25 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

593 The cells were lysed with 1x passive lysis buffer (Promega). Levels of nLuc and fLuc were

594 assessed using Nano-Glo Dual-Luciferase Reporter assay system (Promega) and a Wallac 1420

595 VICTOR 3 V luminometer (Perkin Elmer) according to the manufacturer’s protocol.

596

597 RT-PCR

598 Total RNA was extracted using TRIzol (Thermo Fisher Scientific) according to the

599 manufacturer's protocol, treated with TURBO DNase (Thermo Fisher Scientific) for 30min at

600 37°C, and RNA was then isolated using RNeasy Mini kit (Qiagen). cDNA was generated using

601 SuperScript IV First-Strand Synthesis System (Thermo Fisher Scientific). Quantitative RT-

602 PCR was performed using Power SYBR Green PCR Master Mix (Thermo Fisher Scientific) and

603 primer sets (Supplementary Table 1) in a CFX96 Real-Time System (Bio-Rad).

604

605 ELISA

606 The worm pellets were suspended in cold Co-IP buffer consisting of 50 mM Tris-HCl, pH 7.4,

607 300 mM NaCl, 5mM EDTA, 1% Triton-X 100, 2% sodium dodecyl sulfate, 1 x HaltTM Protease

608 and Phosphatase Inhibitor Cocktail and sonicated with 10 cycles of 0.5 second pulse/0.5 second

609 pause. Protein lysates were cleared by centrifugation at 16,000 g for 20 min at 4°C. The protein

610 concentration was determined using a BCA protein assay kit (Thermo Fisher Scientific). Lysates

611 were diluted to the same concentration using Co-IP buffer, and the levels of poly-GP in lysates

612 were measured with a Meso Scale Discovery-based immunoassay as described previously57.

613

614 Statistical analysis

26 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

615 Statistical analysis was performed by an unpaired t-test, one-way ANOVA with Dunnett's

616 multiple comparisons test, and two-way ANOVA with the Šídák multiple comparison test using

617 GraphPad Prism version 8.2.1. P values for the worm lifespan assay were calculated by the

618 Mantel-Cox log-rank test using GraphPad Prism 8.2.1. A P-value of <0.05 was considered

619 significant. The data are presented as mean±standard error of the mean (SEM).

620

621

622

623

624

27 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

625 Acknowledgements

626 We thank the Caenorhabditis Genetics Center (CGC), which is funded by NIH Office of

627 Research Infrastructure Programs (P40 OD010440), for providing strains. This work was

628 supported by grants from the Lohengrin Foundation and the Association for Frontotemporal

629 Degeneration (AFTD).

630

631 Author Contributions

632 Study design: Y.S., P.K., R.P.R. Literature research: Y.S., P.K., R.P.R. Experimental studies: Y.S.,

633 J.A., P.K., T.F.G., P.I. Data analysis/interpretation: Y.S., A.E.X.B., P.K., R.P.R. Statistical

634 analysis: Y.S., A.E.X.B. Manuscript preparation: Y.S., P.K., R.P.R. Manuscript editing: Y.S., P.K.,

635 R.P.R.

636

637 Competing Interests 638 639 The authors declare no competing interests. 640

28 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

641 FIGURE LEGENDS 642 643 Figure 1. A C. elegans model for C9orf72-mediated ALS/FTD.

644 (A) Schematic diagram showing ΔC9 ubi, C9 ubi, and UAG ubi constructs with the snb-1 promoter.

645 (B) The G4C2 repeat RNA (with C9orf72 intronic RNA) and act-1 mRNAs were assessed by

646 RT-PCR (n = 3 different transgenic lines in each transgenic worm, mean ± s.e.m.). (C) Worm

647 lysates were processed for Western blotting and immunostained with poly-GA and α-tubulin

648 antibodies. (D) Quantification of poly-GA expression on Western blots (n = 3 different

649 transgenic lines in each transgenic worm, mean ± s.e.m.) (E) Luciferae assay was performed in

650 worm lysates (n = 3 different transgenic lines in each transgenic worm, mean ± s.e.m.). (F)

651 Immunofluorescent staining for poly-GA (red) in ΔC9 ubi and C9 ubi adult worms. (G)

652 Quantification of poly-GP in worms ΔC9 ubi (kasIs8) and C9 ubi (kasIs7) as determined by ELISA

653 (n = 4, mean ± s.e.m.). (H) Lifespan of ΔC9 ubi, C9 ubi, or UAG ubi worms. Three different

654 transgenic lines per genotype were used and the results are combined per genotype. ΔC9ubi: n =

655 66, C9ubi: n = 68, UAGubi: n = 67). * P < 0.05, *** P < 0.001, N. S: not significant.

656

657 Figure 2. Evidence for motor neuron degeneration and synaptic defects in C9 ubi animals.

658 (A) Schematic showing cholinergic motor neurons (green) in the C. elegans ventral nerve cord.

659 Dashed boxes indicate the regions analyzed in panels B-C and C-D. (B) Motor neuron cell

660 bodies are visualized with the vsIs48 [unc-17/VAChT::gfp] marker in ΔC9 ubi and C9 ubi animals

661 at the indicated stages. Arrowheads indicate motor neuron cell bodies. (C) Quantification of

662 panel B. N >10 per genotype. *: p <0.05; **: p < 0.01; N.S: not statistically significant. (D) The

663 presynaptic boutons of cholinergic motor neurons on dorsal body wall muscle are visualized with

664 otIs437 [Punc-3::eGFP::rab-3cDNA] in ΔC9 ubi and C9 ubi animals at the indicated stages. (E)

29 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

665 Quantification of panel D. N >10 per genotype. *: p <0.05; **: p < 0.01; N.S: not statistically

666 significant.

667 668 Figure 3. DPR expression exclusively in neurons reduces C. elegans lifespan (A) Schematic

669 diagram showing ΔC9neuro(kasEx159), C9neuro (kasEx157), and UAGneuro (kasEx160) worms under

670 the unc11c promoter. (B) The G4C2 repeat RNA (with C9orf72 intronic RNA) and act-1

671 mRNAs were assessed by RT-PCR. (C) The nLuc and act-1 mRNAs were assessed by RT-PCR

672 (The experiments were repeated 4 times using 1 transgenic line per genotype, mean ± s.e.m.).

673 (D) Worm lysates were processed for Western blotting and immunostained with poly-GA and

674 tubulin antibodies. (E) Quantification of poly-GA expression on Western blots. The experiments

675 were repeated 3 times using 1 transgenic line per genotype, mean ± s.e.m.) (F) Lifespan of

676 ΔC9neuro (kasEx159), C9neuro (kasEx157), or UAGneuro (kasEx160) worms. The experiments were

677 repeated 3 times using 1 transgenic line per genotype, mean ± s.e.m. ΔC9neuro (kasEx159): n = 75,

678 C9neuro (kasEx157): n = 75, UAGneuro (kasEx160): n = 75). ** P < 0.01, **** P < 0.0001, n.s. not

679 significant.

680 681 Figure 4. Genetic removal of eif-2D/eIF2D in C9 ubi and C9 neuro animals suppresses poly-

682 GA expression and ameliorates lifespan defects. (A and F) The G4C2 repeat RNA (with

683 C9orf72 intronic RNA) and Act-1 mRNAs were assessed by RT-PCR. (B and G) The nLuc and

684 Act-1 mRNAs were assessed by RT-PCR. The experiments were repeated 4 times in C and 3

685 times in G, respectively, using 1 transgenic line per genotype (mean ± s.e.m). (C and H) Worm

686 lysates were processed for Western blotting and immunostained with poly-GA and tubulin

687 antibodies. (D and I) Quantification of poly-GA expression on Western blots. The experiments

688 were repeated 4 times in B and 3 times in F, respectively, using 1 transgenic line per genotype

30 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

689 (mean ± s.e.m.) (E and J) Lifespan of ΔC9ubi (kasEx154), C9ubi, and eif-2D (gk904876); C9 ubi

690 animals. One transgenic line was used per genotype. ΔC9ubi (kasEx154): n = 75, C9ubi: n = 72,

691 eif-2D (gk904876); C9 ubi: n = 75) and ΔC9neuro (kasEx159), C9neuro (kasEx157), eif-2D

692 (gk904876); C9neuro (kasEx157), and eif-2D (gk904876) worms. One transgenic line was used per

693 genotype. ΔC9neuro: n = 62, C9neuro: n = 63, eif-2D (gk904876); C9neuro: n = 67, eif-2D

694 (gk904876): n = 68). * P < 0.05, ** P < 0.01, **** P < 0.0001.

695

696 Figure 5. Genetic removal of eif-2D/eIF2D ameliorates the locomotor defects of C9 ubi

697 animals.

698 Examples of locomotion features significantly affected in C9 ubi animals. Genetic removal of eif-

699 2D ameliorated the locomotion defects of C9 ubi animals [eif-2D (-); C9 ubi]. Tracking analysis

700 was performed at Day 2 adult animals (N = 12). The p-value threshold for this analysis (FDR

701 5%) is 0.0014 (see Methods). Panel A shows three locomotion features related to animal velocity.

702 d rel to body ang vel tail tip abs 10th: 10th percentile of the derivative of the absolute value of

703 the angular velocity of the tip of the tail relative to the centroid of the mid-body points. d ang vel

704 abs 10th: 10th percentile of the derivative of the absolute value of the angular velocity of the

705 worm. d rel to tail base radial vel tail tip w backward 90th: 90th percentile of the derivative of

706 the radial velocity of the tip of the tail relative to the centroid of the tail base points, while the

707 worm is moving backwards. Panel B shows three locomotion features related to body curvature

708 while the worm is moving forward. d curvature mean neck w forward abs IQR: interquartile

709 range of the derivative of the absolute value of the mean curvature of the neck.

710 d curvature mean hips w forward abs IQR: interquartile range of the derivative of the absolute

711 value of the mean curvature of the hips. d curvature mean tail w forward abs 10th: 10th percentile

31 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

712 of the derivative of the absolute value of the mean curvature of the tail.

713

714 Figure 6. Knockdown of eIF2D suppresses DPR expression in HEK293 cells. Bicistronic

715 constructs were cotransfected along with either control or anti-eIF2D shRNA and cultured for

716 48h. (A) Schematic diagram showing bicistronic constructs. (B) The cell lysates were processed

717 for Western blotting and immunostained with poly-GA, fLuc, eIF2D, and tubulin antibodies. (C,

718 D, and F) Quantification of (C) eIF2D, (D) poly-GA, and (F) fLuc on Western blots (The

719 experiments were repeated 3 times. mean ± s.e.m.) (E) The levels of luciferase activity was

720 assessed by dual luciferase assays (The experiments were repeated 3 times. mean ± s.e.m.) * P <

721 0.05, *** P < 0.001, **** P < 0.0001, n.s. not significant.

722

723

32 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

724 SUPPLEMENTARY FIGURE LEGENDS

725 Supplementary Figure 1. Assays to detect poly-GR in worm lysates. (A) The lysates from

726 ΔC9ubi and C9ubi worms were processed for Western blotting and immunostained with poly-GR

AUG 727 and a-tubulin antibodies. The lysate from HEK293 cells transfected with GR (G4C275-nLuc)

728 plasmid, in which AUG start codon was inserted before G4C2 repeats in the frame of poly-GR,

729 was used as a positive control. (B) The lysates from ΔC9neuro, C9neuro, and UAGneuro worms were

730 processed for GR ELISA (n = 4, mean ± s.e.m.).

731 Supplementary Figure 2. Genetic removal of eif-2D/eIF2D does not affect expression of an

732 AUG-initiated green fluorescent protein (GFP) transgene. The ΔC9 ubi transgene also carries

733 a pharyngeal gfp marker (myo-2::gfp), but its expression is not affected in eif-2D (gk904876)

734 animals. Representative images of larval stage 4 (L4) worms are shown. Differential interference

735 contrast (DIC) images of the worm head are shown on top. Dashed white line demarcates the

736 pharynx. Fluorescent images are shown below in the GFP channel. N = 10.

737

738 Supplementary Figure 3. Defects in velocity and body curvature are ameliorated in C9 ubi

739 animals upon genetic removal of eif-2D/eIF2D.

740 Examples of seven additional locomotion features significantly affected in C9 ubi animals.

741 Genetic removal of eif-2D ameliorated the locomotion defects of C9 ubi animals [eif-2D (-); C9

742 ubi]. Tracking analysis was performed at Day 2 adult animals (N = 12). The p-value threshold for

743 this analysis (FDR 5%) is 0.0014 (see Methods). Panel A shows features related to animal

744 velocity. d ang vel head base w forward abs 90th: 90th percentile of the derivative of the absolute

745 value of the angular velocity of the base of the head, while the worm is moving forwards.

746 rel to hips radial vel tail tip w forward IQR: interquartile range of the radial velocity of the tip of

33 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

747 the head relative to the centroid of the hips points, while the worm is moving forwards.

748 d rel to tail base radial vel tail tip 50th: 50th percentile of the derivative of the radial velocity of

749 the tip of the tail relative to the centroid of the tail base points.

750 d rel to body radial vel tail tip w backward 10th: 10th percentile of the derivative of radial

751 velocity of the tip of the tail relative to the centroid of the midbody points, while the worm is

752 moving backwards. Panel B shows a feature related to body curvature. d curvature mean hips

753 abs IQR: interquartile range of the derivative of the absolute value of the mean curvature of the

754 hips.

755

756

757

758

759 SUPPLEMENTARY INFORMATION 760

761 Supplementary Table 1: Sequences of oligonucleotides used in this paper.

762

763

764

765

766

767

768

34 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

769 REFERENCES 770

771 1. DeJesus-Hernandez, M. et al. Expanded GGGGCC hexanucleotide repeat in noncoding 772 region of C9ORF72 causes chromosome 9p-linked FTD and ALS. Neuron 72, 245-256 773 (2011). 774 775 2. Renton, A. E. et al. A hexanucleotide repeat expansion in C9ORF72 is the cause of 776 chromosome 9p21-linked ALS-FTD. Neuron 72, 257-268 (2011). 777 778 3. Taylor, J. P., Brown, R. H., Jr. & Cleveland, D. W. Decoding ALS: from genes to 779 mechanism. Nature 539, 197-206 (2016). 780 781 4. Nguyen, L., Cleary, J. D. & Ranum, L. P. W. Repeat-Associated Non-ATG Translation: 782 Molecular Mechanisms and Contribution to Neurological Disease. Annu Rev Neurosci 42, 783 227-247 (2019). 784 785 5. Freibaum, B. D. & Taylor, J. P. The Role of Dipeptide Repeats in C9ORF72-Related 786 ALS-FTD. Front Mol Neurosci 10, 35 (2017). 787 788 6. Mori, K. et al. The C9orf72 GGGGCC repeat is translated into aggregating dipeptide- 789 repeat proteins in FTLD/ALS. Science 339, 1335-1338 (2013). 790 791 7. Zu, T. et al. RAN proteins and RNA foci from antisense transcripts in C9ORF72 ALS and 792 frontotemporal dementia. Proc Natl Acad Sci U S A 110, E4968-4977 (2013). 793 794 8. Gendron, T. F. et al. Antisense transcripts of the expanded C9ORF72 hexanucleotide 795 repeat form nuclear RNA foci and undergo repeat-associated non-ATG translation in 796 c9FTD/ALS. Acta Neuropathol 126, 829-844 (2013). 797 798 9. Mann, D. M. et al. Dipeptide repeat proteins are present in the p62 positive inclusions in 799 patients with frontotemporal lobar degeneration and motor neurone disease associated 800 with expansions in C9ORF72. Acta Neuropathol Commun 1, 68 (2013). 801 802 10. Mackenzie, I. R. et al. Quantitative analysis and clinico-pathological correlations of 803 different dipeptide repeat protein pathologies in C9ORF72 mutation carriers. Acta 804 Neuropathol 130, 845-861 (2015). 805 806 11. Fuchs, G. et al. Kinetic pathway of 40S ribosomal subunit recruitment to hepatitis C virus 807 internal ribosome entry site. Proc Natl Acad Sci U S A 112, 319-325 (2015). 808 809 12. Olivares, E. et al. The 5' untranslated region of the human T-cell lymphotropic virus type 810 1 mRNA enables cap-independent translation initiation. J Virol 88, 5936-5955 (2014). 811 812 13. Yamada, S. B. et al. RPS25 is required for efficient RAN translation of C9orf72 and other 813 neurodegenerative disease-associated nucleotide repeats. Nat Neurosci 22, 1383-1388

35 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

814 (2019). 815 816 14. Zhou, M. et al. Mass spectrometry reveals modularity and a complete subunit interaction 817 map of the eukaryotic translation factor eIF3. Proc Natl Acad Sci U S A 105, 18139- 818 18144 (2008). 819 820 15. Ayhan, F. et al. SCA8 RAN polySer protein preferentially accumulates in white matter 821 regions and is regulated by eIF3F. EMBO J 37, e99023 (2018). 822 823 16. Goodman, L. D. et al. eIF4B and eIF4H mediate GR production from expanded G4C2 in 824 a Drosophila model for C9orf72-associated ALS. Acta Neuropathol Commun 7, 62 825 (2019). 826 827 17. Kim, J. H., Park, S. M., Park, J. H., Keum, S. J. & Jang, S. K. eIF2A mediates translation 828 of hepatitis C viral mRNA under stress conditions. EMBO J 30, 2454-2464 (2011). 829 830 18. Starck, S. R. et al. Translation from the 5' untranslated region shapes the integrated stress 831 response. Science 351, aad3867 (2016). 832 833 19. Sendoel, A. et al. Translation from unconventional 5' start sites drives tumour initiation. 834 Nature 541, 494-499 (2017). 835 836 20. Sonobe, Y. et al. Translation of dipeptide repeat proteins from the C9ORF72 expanded 837 repeat is associated with cellular stress. Neurobiol Dis 116, 155-165 (2018). 838 839 21. Cheng, W. et al. CRISPR-Cas9 Screens Identify the RNA Helicase DDX3X as a 840 Repressor of C9ORF72 (GGGGCC)n Repeat-Associated Non-AUG Translation. Neuron 841 104, 885-898 (2019). 842 843 22. Kramer, N. J. et al. Spt4 selectively regulates the expression of C9orf72 sense and 844 antisense mutant transcripts. Science 353, 708-712 (2016). 845 846 23. Therrien, M., Rouleau, G. A., Dion, P. A. & Parker, J. A. Deletion of C9ORF72 results in 847 motor neuron degeneration and stress sensitivity in C. elegans. PLoS One 8, e83450 848 (2013). 849 850 24. Ji, Y. J. et al. C9orf72/ALFA-1 controls TFEB/HLH-30-dependent metabolism through 851 dynamic regulation of Rag GTPases. PLoS Genet 16, e1008738 (2020). 852 853 25. Corrionero, A. & Horvitz, H. R. A C9orf72 ALS/FTD Ortholog Acts in Endolysosomal 854 Degradation and Lysosomal Homeostasis. Curr Biol 28, 1522-1535.e1525 (2018). 855 856 26. Stefanakis, N., Carrera, I. & Hobert, O. Regulatory Logic of Pan-Neuronal Gene 857 Expression in C. elegans. Neuron 87, 733-750 (2015). 858 859 27. Guo, Q. et al. In Situ Structure of Neuronal C9orf72 Poly-GA Aggregates Reveals

36 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

860 Proteasome Recruitment. Cell 172, 696-705 (2018). 861 862 28. Chang, Y. J., Jeng, U. S., Chiang, Y. L., Hwang, I. S. & Chen, Y. R. The Glycine-Alanine 863 Dipeptide Repeat from C9orf72 Hexanucleotide Expansions Forms Toxic Amyloids 864 Possessing Cell-to-Cell Transmission Properties. J Biol Chem 291, 4903-4911 (2016). 865 866 29. Green, K. M. et al. RAN translation at C9orf72-associated repeat expansions is 867 selectively enhanced by the integrated stress response. Nat Commun 8, 2005 (2017). 868 869 30. Tabet, R. et al. CUG initiation and frameshifting enable production of dipeptide repeat 870 proteins from ALS/FTD C9ORF72 transcripts. Nat Commun 9, 152 (2018). 871 872 31. Boivin, M. et al. Reduced autophagy upon C9ORF72 loss synergizes with dipeptide 873 repeat protein toxicity in G4C2 repeat expansion disorders. EMBO J 39, e100574 (2020). 874 875 32. Moloney, E. B., de Winter, F. & Verhaagen, J. ALS as a distal axonopathy: molecular 876 mechanisms affecting neuromuscular junction stability in the presymptomatic stages of 877 the disease. Front Neurosci 8, 252 (2014). 878 879 33. Dmitriev, S. E. et al. GTP-independent tRNA delivery to the ribosomal P-site by a novel 880 eukaryotic translation factor. J Biol Chem 285, 26779-26787 (2010). 881 882 34. Kearse, M. G. & Wilusz, J. E. Non-AUG translation: a new start for protein synthesis in 883 eukaryotes. Genes Dev 31, 1717-1731 (2017). 884 885 35. Javer, A., Ripoll-Sánchez, L. & Brown, A. E. X. Powerful and interpretable behavioural 886 features for quantitative phenotyping of Caenorhabditis elegans. Philos Trans R Soc Lond 887 B Biol Sci 373, 20170375 (2018). 888 889 36. Yemini, E., Jucikas, T., Grundy, L. J., Brown, A. E. & Schafer, W. R. A database of 890 Caenorhabditis elegans behavioral phenotypes. Nat Methods 10, 877-879 (2013). 891 892 37. Cheng, W. et al. C9ORF72 GGGGCC repeat-associated non-AUG translation is 893 upregulated by stress through eIF2alpha phosphorylation. Nat Commun 9, 51 (2018). 894 895 38. May, S. et al. C9orf72 FTLD/ALS-associated Gly-Ala dipeptide repeat proteins cause 896 neuronal toxicity and Unc119 sequestration. Acta Neuropathol 128, 485-503 (2014). 897 898 39. Zhang, Y. J. et al. Aggregation-prone c9FTD/ALS poly(GA) RAN-translated proteins 899 cause neurotoxicity by inducing ER stress. Acta Neuropathol 128, 505-524 (2014). 900 901 40. Zhang, Y. J. et al. C9ORF72 poly(GA) aggregates sequester and impair HR23 and 902 nucleocytoplasmic transport proteins. Nat Neurosci 19, 668-677 (2016). 903 904 41. Schludi, M. H. et al. Spinal poly-GA inclusions in a C9orf72 mouse model trigger motor 905 deficits and inflammation without neuron loss. Acta Neuropathol 134, 241-254 (2017).

37 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

906 907 42. Khosravi, B. et al. Cell-to-cell transmission of C9orf72 poly-(Gly-Ala) triggers key 908 features of ALS/FTD. EMBO J 39, e102811 (2020). 909 910 43. Nguyen, L. et al. Antibody Therapy Targeting RAN Proteins Rescues C9 ALS/FTD 911 Phenotypes in C9orf72 Mouse Model. Neuron 105, 645-662 (2020). 912 913 44. Swinnen, B. et al. A zebrafish model for C9orf72 ALS reveals RNA toxicity as a 914 pathogenic mechanism. Acta Neuropathol 135, 427-443 (2018). 915 916 45. Choi, S. Y. et al. C9ORF72-ALS/FTD-associated poly(GR) binds Atp5a1 and 917 compromises mitochondrial function in vivo. Nat Neurosci 22, 851-862 (2019). 918 919 46. Rudich, P. et al. Nuclear localized C9orf72-associated arginine-containing dipeptides 920 exhibit age-dependent toxicity in C. elegans. Hum Mol Genet 26, 4916-4928 (2017). 921 922 47. Almeida, S. et al. Production of poly(GA) in C9ORF72 patient motor neurons derived 923 from induced pluripotent stem cells. Acta Neuropathol 138, 1099-1101 (2019). 924 925 48. Swinnen, B., Robberecht, W. & Van Den Bosch, L. RNA toxicity in non-coding repeat 926 expansion disorders. EMBO J 39, e101112 (2020). 927 928 49. Skabkin, M. A. et al. Activities of Ligatin and MCT-1/DENR in eukaryotic translation 929 initiation and ribosomal recycling. Genes Dev 24, 1787-1801 (2010). 930 931 50. Kwon, O. S. et al. An mRNA-specific tRNAi carrier eIF2A plays a pivotal role in cell 932 proliferation under stress conditions: stress-resistant translation of c-Src mRNA is 933 mediated by eIF2A. Nucleic Acids Res 45, 296-310 (2017). 934 935 51. Skabkin, M. A., Skabkina, O. V., Hellen, C. U. & Pestova, T. V. Reinitiation and other 936 unconventional posttermination events during eukaryotic translation. Mol Cell 51, 249- 937 264 (2013). 938 939 52. Young, D. J. et al. Tma64/eIF2D, Tma20/MCT-1, and Tma22/DENR Recycle Post- 940 termination 40S Subunits In Vivo. Mol Cell 71, 761-774 (2018). 941 942 53. Zinoviev, A., Hellen, C. U. T. & Pestova, T. V. Multiple mechanisms of reinitiation on 943 bicistronic calicivirus mRNAs. Mol Cell 57, 1059-1073 (2015). 944 945 54. Brenner, S. The genetics of Caenorhabditis elegans. Genetics 77, 71-94 (1974). 946 947 55. Schneider, C. A., Rasband, W. S. & Eliceiri, K. W. NIH Image to ImageJ: 25 years of 948 image analysis. Nat Methods 9, 671-675 (2012). 949 950 56. Gendrel, M., Rapti, G., Richmond, J. E. & Bessereau, J. L. A secreted complement- 951 control-related protein ensures acetylcholine receptor clustering. Nature 461, 992-996

38 bioRxiv preprint doi: https://doi.org/10.1101/2020.06.13.150029; this version posted June 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

952 (2009). 953 954 57. Andrade, N. S. et al. Dipeptide repeat proteins inhibit homology-directed DNA double 955 strand break repair in C9ORF72 ALS/FTD. Mol Neurodegener 15, 13. doi: 956 10.1186/s13024-13020-00365-13029. (2020). 957 958

39 Figure 1

A B C9orf72 mRNA 2.0 ΔC9 ubi Psnb-1 ATG nLuc unc-54 3’ UTR N. S ) 1.5 * CUG In frame with GA 1.0 ubi Psnb-1 (GGGGCC)75 ATG nLuc unc-54 3’ UTR C9 (113 bp 5’) (99 bp 3’)

C9orf72/Act-1 0.5 (

UAG In frame Relativeexpression with GA 0.0 ubi ATG nLuc UAG Psnb-1 (GGGGCC)75 unc-54 3’ UTR (113 bp 5’) (99 bp 3’)

C D Poly-GA E Luciferase assay ΔC9 ubi C9 ubi UAG ubi 2.0 800800

*** 1 2 3 1 2 3 1 2 3 1.5 *** *** *** 600600 Poly-GA 1.0 gprotein 400400 µ

0.5 200200

Tubulin nLuc/ (Poly-GA/Tubulin) (Poly-GA/Tubulin) Relativeexpression 0.0 00

F anti-GA staining G H ΔC9 ubi Poly-GP P < 0.0001 C9 ubi ubi nerve ring ΔC9 2.0 P < 0.0001 UAG ubi *** 100 1.5

1.0 50 nerve ring ubi 0.5 C9 Foldchange

0.0 Percent(%)survival 0 0 10 20 30 40 Days from egg Figure 2 A D - E

pharynx

B - C

B Juvenile (L4 stage) Adult (Day 5)

ΔC9 ubi

C9 ubi

25 N. S * C ** 20 ΔC9 ubi 15

10 C9 ubi

positive motor neurons 5 Number of unc-17/VAChT::gfp 0

Adult Adult (Day 2) (Day 5) Juvenile (L4 stage) D Juvenile (L4 stage) Adult (Day 4) ΔC9 ubi

C9 ubi

E 18 16 ** N. S 14 ubi * ΔC9 12

10 ubi 8 C9

6

pre-synaptic boutons 4

Number of eGFP::RAB-3+ 2

0

Adult Adult (Day 2) (Day 4) Juvenile (L4 stage) Figure 3

A ΔC9 neuro Punc-11 ATG nLuc unc-54 3’ UTR

CUG In frame with GA neuro C9 Punc-11 (GGGGCC)75 ATG nLuc unc-54 3’ UTR (113 bp 5’) (99 bp 3’)

UAG In frame with GA neuro nLuc UAG Punc-11 (GGGGCC)75 ATG unc-54 3’ UTR (113 bp 5’) (99 bp 3’)

B C9orf72 mRNA nLuc mRNA 1.5 C 1.5 ** N. S. N. S. N. S. ) ) 1.0 1.0

0.5 nLuc/Act-1 0.5 ( C9orf72/Act-1 ( Relativeexpression Relativeexpression 0.0 0.0

D E Poly-GA 1.5 *** *** 1.0 Poly-GA

0.5 Tubulin (Poly-GA/Tubulin) (Poly-GA/Tubulin) Relativeexpression 0.0

F ΔC9 neuro 100 P < 0.0001 C9 neuro P < 0.0001 UAG neuro 50 Percent(%)survival 0 0 10 20 30 40 Days from egg Figure 4

A B C C9orf72 mRNA nLuc mRNA N.S. 1.5 2.5 N.S. N.S. N.S. ) *** 2.0 ) 1.0 N.S. Poly-GA 1.5

1.0 0.5 nLuc/Act-1 C9orf72/Act-1 ( Tubulin ( 0.5 Relativeexpression Relativeexpression 0.0 0.0

D Poly-GA E C9 ubi 2.0 100 Δ * P < 0.0079 N.S. C9 ubi 1.5 ** P < 0.0017 eif-2D (-); 50 1.0 C9 ubi

0.5 Percent(%)survival (Poly-GA/Tubulin) (Poly-GA/Tubulin)

Relativeexpression 0 0.0 0 10 20 30 Days from egg

F C9orf72 mRNA G nLuc mRNA H

2.0 *** 2.0 N.S.

) N.S.

1.5 ) 1.5 *** Poly-GA 1.0 1.0 nLuc/Act-1

( 0.5

C9orf72/Act-1 0.5

( Tubulin Relativeexpression Relativeexpression 0.0 0.0

I Poly-GA J ΔC9 neuro 1.5 100 P = 0.0009 C9 neuro P < 0.0001 P = 0.0678 ** ** eif-2D (-); 1.0 C9 neuro 50 eif-2D (-) 0.5 (Poly-GA/Tubulin) (Poly-GA/Tubulin) Percent(%)survival Relativeexpression 0 0.0 0 10 20 30 40 Days from egg Figure 5

A Velocity features

N. S N. S N. S N. S N. S N. S N. S N. S N. S *** *** 0.05 *** 0.045 55

0.045 0.04 50

0.04 0.035 45 0.035 0.03 40 0.03 0.025

d ang vel abs 10th 35 0.025 0.02

0.02 0.015 30 d rel to body ang vel tail tip abs 10th

0.015 0.01 25

ubi ubi d rel to tail base radial vel tip w backward 90th ubi ubi ubi ubi ubi ubi ubi ubi ubi ubi ΔC9 C9 C9 C9 UAG ΔC9 UAG ΔC9 UAG eif-2D (-) eif-2D (-) eif-2D (-) eif-2D (-); C9 eif-2D (-); C9 eif-2D (-); C9

B Curvature features N. S N. S N. S N. S N. S N. S N. S N. S N. S x 10 -3 x 10 -3 x 10 -4 *** *** *** 7.5 5.5 8

7 5 7 6.5 4.5 6 6 4 5.5 3.5 5 5 3 4.5 4 2.5 4 3 3.5 2 d curvature mean hips w forward abs IQR d curvature mean tail w forward abs 10th d curvature mean neck w forward abs IQR 3 1.5 2

ubi ubi ubi ubi ubi ubi ubi ubi ubi ubi ubi ubi ΔC9 C9 C9 C9 UAG ΔC9 UAG ΔC9 UAG eif-2D (-) eif-2D (-) eif-2D (-) eif-2D (-); C9 eif-2D (-); C9 eif-2D (-); C9 Figure 6

A CUG In frame with STOP GA, GP, or GR ATG nLuc CMV fLuc (GGGGCC)75 BGH polyA (113 bp 5’) (99 bp 3’)

eIF2D Poly-GA B Con eIF2D C 1.5 D 1.5 shRNA shRNA * **** eIF2D 1.0 1.0

Poly-GA 0.5 0.5 (eIF2D/Tubulin) (eIF2D/Tubulin) (Poly-GA/Tubulin) (Poly-GA/Tubulin) Relativeexpression fLuc Relativeexpression

0.0 0.0 Tubulin Con eIF2D Con eIF2D shRNA shRNA shRNA shRNA

E F fLuc 20 **** Control shRNA 1.5 N.S.

15 eIF2D shRNA ) 1.0 fLuc 10 **** /Tubulin) /Tubulin) (nLuc/

fLuc 0.5 5 *** ( Relativeexpression Relativeexpression

0 0.0 Con eIF2D GA GP GR shRNA shRNA