eIF2B extends lifespan through inhibition of the integrated stress response

Protein homeostasis is modulated by stress response pathways and its deficiency is a hallmark of aging. The integrated stress response (ISR) is a conserved stress-signaling pathway that tunes mRNA translation via phosphorylation of the translation initiation factor eIF2. ISR activation and translation initiation are finely balanced by eIF2 kinases and by the eIF2 guanine nucleotide exchange factor eIF2B. However, the role of the ISR during aging remains unexplored. Using a genomic screen in Caenorhabditis elegans, we discovered a role of eIF2B and the eIF2 kinases in longevity. By limiting the ISR, these mutations enhanced protein homeostasis and increased lifespan. Consistently, full ISR inhibition using phosphorylation-defective eIF2α or pharmacological ISR inhibition prolonged lifespan. Lifespan extension through ISR inhibition occurred without changes in overall protein synthesis, and depended on enhanced translational efficiency of the kinase KIN-35. Evidently, lifespan is limited by the ISR and its inhibition may provide an intervention in aging.

Aging is defined as the progressive loss of physiological integrity accompanied by 15 reduced cellular, organ, and systemic performance. It is characterized by cellular 16 hallmarks such as stem cell exhaustion, genomic instability, deregulated nutrient 17 sensing and loss of protein homeostasis 1 . Thus, aging is the main risk factor for 18 neurodegenerative disorders, cancer and metabolic syndrome. The aging process can 19 be modulated by environmental and genetic factors, and several evolutionary 20 conserved biological processes have been implicated in lifespan regulation 2 . Failure 21 of protein homeostasis is an early event during aging and various interventions that 22 promote or maintain protein homeostasis beneficially affect lifespan in model 23 organisms 3-5 . During stressful conditions, the maintenance of protein homeostasis by 24 cellular stress response pathways is an essential feature of cellular integrity and 25 organismal fitness. Internal and external stimuli trigger evolutionarily conserved 26 cellular stress pathways such as the heat shock response (HSR), organelle-specific 27 stress response pathways such as the endoplasmic reticulum or mitochondrial 28 unfolded protein responses (ER-UPR/mito-UPR) and the Integrated Stress Response 29 (ISR). Multiple lines of evidence show that longevity ultimately relies on the fidelity of 30 cellular stress response mechanisms 6 . 31 The biological function of the ISR is to restore cellular homeostasis upon stress. The 32 activation of the ISR relies on the eukaryotic initiation factor 2 (eIF2) kinases: the 33 heme-regulated inhibitor (HRI), protein kinase R (PKR), general control 34 nonderepressible 2 (GCN2), and PKR-like endoplasmic reticulum kinase (PERK). 35 They are activated, respectively, by iron deficiency, viral infection, amino acid 36 deprivation and accumulation of misfolded protein in the ER. The kinases converge on 37 the phosphorylation of the a subunit of eIF2. eIF2 is a key regulator of translation 38 initiation, the limiting step of protein synthesis 7 . For translation initiation to occur, the 39 eIF2.GTP.tRNA met ternary complex together with other initiation factors and the 40S 40 ribosomal subunit forms the 43S pre-initiation complex. The 43S complex binds to the 41 5'-cap structure and scans along the mRNA until it recognizes the AUG start codon. 42 Then, GTP hydrolysis releases eIF2 and other initiation factors from the mRNA-40S-43 complex, allowing the 60S ribosomal sub-unit to bind and proceed to elongation 8 . The 44 exchange of GDP to GTP is necessary for recycling of eIF2 back to its active form and 45 for further rounds of translation initiation. This exchange is catalyzed by the 46 heterodecameric guanine nucleotide exchange factor eIF2B. The phosphorylation of 47 eIF2a at serine 51 by the stress sensitive kinases represents the core event of the ISR. 48 Phospho-eIF2a is a strong inhibitor of eIF2B leading to attenuated ternary complex 49 formation and therefore to a reduction of 5'-cap-dependent protein synthesis. 50 Decreasing the abundance of the ternary complex paradoxically de-represses 51 translation of specific mRNAs that are regulated by upstream open reading frames 52 (uORFs) such as ATF4, ATF5, and CHOP. While the ISR and translation initiation are 53 finely balanced to provide robustness during acute challenges to protein homeostasis, 54 the role of this pathway during aging and in longevity remains largely unexplored. 55 Forward genomic screens in C. elegans have shed light on numerous pathways whose 56 activity extend lifespan 9,10 . These approaches used systematic mRNA knockdown, and 57 did not have the resolution to investigate consequence of other genetic alterations, 58 including gain-of-function mutations, in longevity. Unbiased forward screens using 59 chemical mutagenesis coupled with whole genome sequencing are a powerful tool to 60 reveal new longevity loci. We therefore set out to perform a large-scale mutagenesis 61 screen for increased survival in C. elegans. 62

Gcn(-) mutations extend C. elegans lifespan 86
We used RNAi to investigate the effect of ppp-1 silencing. ppp-1 knockdown did not 87 affect the lifespan of WT animals (Fig. 2a) Next, we tested whether the gcn-2(wrm4) and pek-1(wrm7) mutants also belong to the 102 Gcn(-) class. The gcn-2(wrm4) mutant displayed a 50% reduction of baseline eIF2a 103 phosphorylation suggesting that this mutant can be classified as Gcn(-) (Fig. 2d). The 104 reduction of eIF2a phosphorylation in the pek-1(wrm7) mutant did not reach 105 significance. To mechanistically address whether Gcn(-) mutations lead to longevity, 106 we engineered a phospho-defective eIF2aS51A mutant (eIF2a(syb1385)), abolishing 107 the ISR (Fig. 2e). Homozygous eIF2aS51A mutants were viable and displayed regular 108 pharyngeal pumping rates (Extended Data Fig. 2c), generation time (Extended Data 109 were hypersensitive to ER stress induced by tunicamycin, likely because 111 phosphorylation of eIF2a by the pek-1/PERK kinase is required to promote the ER 112 stress response and survival (Fig. 2f). Notably, eIF2aS51A mutants showed a robust 113 lifespan extension compared to WT animals demonstrating that Gcn(-) mutations lead 114 to longevity in C. elegans and that the genetic inhibition of the ISR can extend lifespan 115 (Fig. 2g). Consistently, eIF2aS51A mutants were heat resistant (Extended Data 116 Fig. 2f). Finally, we assessed survival during pharmacological ISR inhibition. For this, 117 we used a set of compounds that were previously described as UPR modulators in 118 worms 16 . Estradiol valerate reduced GFP induction of the atf-5 reporter during 119 tunicamycin treatment whereas propafenone hydrochloride further elevated GFP 120 expression (Extended Data Fig. 2g). Consistent with the Gcn(-) phenotype, estradiol 121 valerate significantly extended C. elegans lifespan (Fig. 2h) and suppressed eIF2a 122 phosphorylation upon DTT treatment (Fig. 2i). Surprisingly, treatments initiated at day 123 5 or day 10 of adulthood equally increased survival (Fig. 2h) suggesting that late ISR 124 inhibition might be sufficient to promote lifespan extension. ISR induction with 125 propafenone hydrochloride shortened lifespan (Extended Data Fig. 2h). 126

Gcn(-) longevity is independent of attenuated translation 127
As eIF2B is a key regulator of translation initiation, we monitored protein synthesis in of day 1 adult animals and did not observe any differences between WT animals and 130 ppp-1 mutants (Fig. 3a). To corroborate these results, we used surface sensing of 131 translation (SUnSET) as an alternative measurement of protein synthesis rates. This 132 technique is based on the incorporation of puromycin into newly synthesized peptides 133 followed by the detection of the labelled peptides with monoclonal antibody 17 . No 134 changes in protein synthesis were observed between ppp-1 mutants and WT animals 135 whereas control rsks-1/S6K mutants showed a drastic reduction of puromycin-labelled 136 peptides (Fig. 3b). Finally, we performed polysome profiling to evaluate the distribution 137 of the ribosomal subunits and complexes after separation on sucrose gradients. We 138 found no differences in the overall ribosome distribution and abundance ( Fig. 3c-d). 139 Likewise, we found no differences in polysome abundance at day 1 of adulthood 140 between WT animals and the eIF2aS51A mutant ( Fig. 3e-f). Since eIF2 activity is 141 regulated by phosphorylation, we also evaluated the level of phospho-eIF2a on day 1 142 and day 6 of adulthood. We found that the phosphorylation of eIF2a was increased 143 upon aging in WT animals (Extended Data Fig. 3a). However, we did not observe any 144 differences between ppp-1 mutants or WT control at day 1 or day 6. Together, our 145 results support the idea that the longevity of ppp-1 mutants is uncoupled from reduced 146 protein synthesis. 147

kin-35 translation is required for Gcn(-) longevity 148
As we did not observe any changes in global protein synthesis, we asked whether 149 translational efficiency of specific mRNAs could be causative for the lifespan extension 150 of the ppp-1 animals. We compared the ratio of polysome-associated mRNAs (>3 151 ribosomes/mRNA) normalized to total mRNA levels between WT and ppp-1 animals 152 (Fig. 4a). We found a significant de-enrichment of 336 mRNAs and an enrichment for 153 72 mRNAs in ppp-1 polysome fractions (Fig. 4b). GO Term analysis revealed an 154 enrichment for genes involved in phosphorylation (Fig. 4c). 155 Several studies have demonstrated that translation efficiency of specific mRNAs is a 156 key regulator of lifespan under different longevity paradigms in C. elegans 18,19 . 157 Therefore, we hypothesized that some of the enriched mRNAs define ppp-1 158 phenotypes. We used resistance to polyQ35 proteotoxicity of ppp-1 animals as a proxy 159 for longevity and knocked down the candidate mRNAs in ppp-1(wrm10) mutants with 160 RNAi. At day 8 of adulthood, all polyQ35 transgenic animals were paralyzed. with polysomes was enhanced in ppp-1 mutants without increased allover abundance 171 (Extended Data Fig. 4b). Together, these data suggest that increased translation of 172 kin-35 mRNA is required for ppp-1 longevity. C01A2.5 also significantly reduced ppp-173 1 longevity but also shortened WT lifespan suggesting general toxicity (Extended Data 174 Through an unbiased genomic screen for longevity in C. elegans, we identified the ISR 178 as a longevity pathway. We provide evidence that genetic inhibition of the ISR via 179

What is the link between Gcn(-) mutations and lifespan extension? 198
The regulation of translation initiation and the ISR are intimately linked. Our data 199 suggest that a shift in the translatome, and not the loss of the ISR per se, is responsible   Unbiased forward longevity screen 248 The longevity screen was performed with the temperature sensitive sterile strain 249 CF512 fer-15(b26) II; fem-1(hc17) IV. L4 larvae were exposed to 0,3% ethyl methane 250 sulfonate (EMS, Sigma) in M9 buffer for 4 h at room temperature. After recovery 251 overnight, young P0 adult animals were transferred to new plates. Singled F1 progeny 252 were allowed to lay eggs overnight. In the next generation, singled F2 progeny were 253 allowed to lay eggs for 16 h. After egg-laying, F2 worms were stocked at 15ºC. F3 eggs 254 were heat shocked at 25ºC for 48 h to induce sterility and adult animals were scored 255 twice a week for preliminary lifespan analysis. Mutants that outlived the non-256 mutagenized control by 20% (maximum lifespan) were selected for regular 257 demographic lifespan analyses to confirm the longevity phenotype. After the lifespan 258 assays, mutants with a mean lifespan extension above 18% compared to non-259 mutagenized CF512 controls were selected for whole genome sequencing. 260

Mutant mapping and sequence analysis 261
Genomic DNA of select long-lived strains was prepared using the QIAGEN Gentra 262 Puregene Kit. Whole genome sequencing was conducted on the Illumina HiSeq2000 263 platform. Paired-end 100 bp reads were used; the average coverage was larger than 264 16-fold. Sequencing outputs were analyzed using the CloudMap Unmapped Mutant 265 Workflow pipeline on Galaxy 44 . The WS220/ce10 C. elegans assembly was used as 266 reference genome. 267

Lifespan assays 268
Gravid day 1 adults were allowed to lay eggs for 5 h. The offspring was used for 269 lifespan analysis. The L4 stage was defined as day 0 and more than 100 worms were 270 used per strain and condition. Worms were kept at 20°C on NGM plates seeded with 271

Thermotolerance assays 288
After an egg-lay, synchronized day 1 animals were transferred to 6 cm NGM plates 289 containing OP50 and placed at 35ºC. Survival was scored for (touch-provoked) 290 movement and pharyngeal pumping every two hours until no survivors were left. to plates seeded with HT115 bacteria expressing luciferase or candidate RNAi clones. 300 On day 8 of adulthood, motility was tested by transferring single worms to M9, where 301 they were allowed to acclimatize for 30 sec, followed by the counting of body bends 302 over 30 sec. At least 12 worms were scored per experiment, genotype and/or 303 treatment. Throughout the experiment, strain and/or treatment was unknown to the 304 researcher. Unless stated otherwise, at least 3 independent experiments were 305 performed, error bars represent means ±SD and assays were analyzed by one-way 306 ANOVA, Dunnett's post hoc test. ice for 30 min, followed by a centrifugation step (12000 g, 10 min, 4°C) for clearance. 395 The pellet was discarded and the RNA concentration of the supernatant was estimated 396 by absorbance measurement at 260 nm. 397 To prepare sucrose gradients, 15% (w/v) and 60% (w/v) sucrose solutions were 398 prepared in basic lysis buffer (20 mM Tris pH 8.5, 140 mM KCl, 1.5 mM MgCl2, 1 mM 399 DTT, 1 mM cycloheximide). Linear sucrose gradients were produced using a Gradient 400 Master (Biocomp). Equivalent amounts of sample (around 400 µg RNA) were loaded 401 on the gradient and centrifuged at 39000 g for 3 h at 4°C, using an Optima L-100 XP 402 Ultracentrifuge (Beckman Coulter) and the SW41Ti rotor. To analyze the sample on 403 the gradient during fractionation, absorbance at 254 nm was measured and recorded 404 (Econo UV monitor EM-1, Biorad) using the Gradient Profiler software (version 2.07). RNAi treatments rescuing the ppp-1 polyQ35 motility phenotype to at least 50% 454 compared to the ppp-1 polyQ35 control on luciferase RNAi were validated by full 455 motility assays (without usage of FUDR) counting body bends over 30 seconds in 456 liquid. In a counter screen, the effect of the RNAi treatments on WT polyQ35 animals 457 was tested. To this end, young worms were treated as described before and the motility 458 on day 6 of adulthood was scored as described above. If motility of WT polyQ35 worms 459 treated with RNAi against candidate mRNAs was significantly lower compared to 460 animals treated with luciferase RNAi, candidates were excluded from further analysis.    total mRNA levels between WT and ppp-1 mutants. All displayed mRNAs were found 745