Neuronal TORC1 modulates longevity via AMPK and cell nonautonomous regulation of mitochondrial dynamics in C. elegans

Target of rapamycin complex 1 (TORC1) and AMP-activated protein kinase (AMPK) antagonistically modulate metabolism and aging. However, how they coordinate to determine longevity and if they act via separable mechanisms is unclear. Here, we show that neuronal AMPK is essential for lifespan extension from TORC1 inhibition, and that TORC1 suppression increases lifespan cell non autonomously via distinct mechanisms from global AMPK activation. Lifespan extension by null mutations in genes encoding raga-1 (RagA) or rsks-1 (S6K) is fully suppressed by neuronal-specific rescues. Loss of RAGA-1 increases lifespan via maintaining mitochondrial fusion. Neuronal RAGA-1 abrogation of raga-1 mutant longevity requires UNC-64/syntaxin, and promotes mitochondrial fission cell nonautonomously. Finally, deleting the mitochondrial fission factor DRP-1 renders the animal refractory to the pro-aging effects of neuronal RAGA-1. Our results highlight a new role for neuronal TORC1 in cell nonautonomous regulation of longevity, and suggest TORC1 in the central nervous system might be targeted to promote healthy aging.

abrogation of raga-1 mutant longevity requires UNC-64/syntaxin, and promotes 23 mitochondrial fission cell nonautonomously. Finally, deleting the mitochondrial fission 24 factor DRP-1 renders the animal refractory to the pro-aging effects of neuronal RAGA-1. 25 Our results highlight a new role for neuronal TORC1 in cell nonautonomous regulation of 26 longevity, and suggest TORC1 in the central nervous system might be targeted to 27 promote healthy aging. 28 29

Introduction 30
Aging is the single biggest risk factor for the majority of non-communicable complex 31 diseases, including some of those with the greatest negative impact on human health 32 outcomes worldwide (Escoubas, Silva-Garcia, & Mair, 2017). Work over the last two 33 decades has uncovered molecular mechanisms that can be manipulated in model 34 organisms to modulate the aging process and reduce overall disease risk in old age 35 (Fontana, Partridge, & Longo, 2010). Many of these interventions have been linked to 36 nutrient and energy sensing pathways, whose modulation genetically or 37 pharmacologically mimics the effects of dietary restriction on healthy aging (Fontana & 38 Partridge, 2015). One classic example of a nutrient sensor linked to longevity is TORC1, 39 which promotes anabolic processes such as protein translation to provide 40 macromolecules for growth and proliferation while inhibiting catabolic activities such as 41 autophagy (V. Albert & Hall, 2015). TORC1 is activated by growth factors and amino 42 acids, the latter of which act through sensors such as the sestrins to facilitate heterodimer 43 formation of the Rag proteins (consisting of RagA, B, C and D in mammals), an essential 44 step for TORC1 activation (Saxton & Sabatini, 2017). Suppression of TORC1 both 45 genetically and pharmacologically, via rapamycin feeding, promotes longevity in multiple 46 species from yeast to mice (Kennedy & Lamming, 2016). In contrast to TORC1, the 47 conserved kinase AMPK is activated under low energy conditions. AMPK activation 48 promotes catabolic processes that generate ATP, including the TCA cycle, fatty acid 49 oxidation and autophagy (Burkewitz, Zhang, & Mair, 2014), and extends lifespan in C. Here, we elucidate the relationship between AMPK and TORC1 in the modulation of aging 63 by discovering a critical role for neuronal AMPK in lifespan extension resulting from 64 suppression of TORC1 components in C. elegans. We show that neuronal TORC1 65 pathway activity itself is critical for healthy aging. Restoring either raga-1 or rsks-1 66 To identify mechanisms specifically coupled to neuronal RAGA-1 regulation of lifespan, 212 we examined the transcriptomes of wild type ("WT"), raga-1 mutant ("mutant") and raga-213 1 mutant with neuronal raga-1 expression ("rescue") C. elegans by RNA-Seq. We 214 performed a cluster analysis that takes into account trends in gene expression across all 215 three conditions to identify genes that change in the raga-1 mutant and are reversed by these clusters reveals an enrichment for Gene Ontology (GO) terms related to organelle 223 organization, organelle fission, and unfolded protein/ER stress pathways that are 224 upregulated in the raga-1 mutant but not in rescue, whereas GO terms related to neuronal 225 function, including synaptic structure and signaling as well as regulation of dauer entry 226 are enriched in the cluster of genes that show reduced expression in raga-1 but not in 227 rescue (Figure 3b). These GO terms are also revealed in pairwise comparisons designed 228 to identify biological processes that differ between wild type and raga-1, but not between 229 wild type and neuronal rescued animals ( Figure 3-figure supplement 1b, c). Neuropeptides are released from dense core vesicles (DCVs) (Li & Kim, 2008). If 245 neuronal RAGA-1 suppresses raga-1 longevity via expression of insulin-like peptides or 246 other neuropeptide signals, we reasoned the lifespan of raga-1 neuronal rescue worms 247 might be de-repressed by blocking neuropeptide release. We utilized mutants for unc-64, 248 a homolog of mammalian syntaxin, an essential plasma membrane receptor for DCV 249 exocytosis to test whether impairing neuronal function in this way would block the ability 12 of neuronal RAGA-1 to rescue. Hypomorphic unc-64(e246) mutant animals are defective 251 for dense core vesicle docking (Zhou et al., 2007) and remarkably completely remove 252 suppression of raga-1 lifespan by neuronal RAGA-1. raga-1; unc-64 double mutant 253 animals with the neuronal rescue array live more than 40% longer than raga-1 neuronal 254 rescue animals (Figure 3e). The longevity effects by unc-64 mutation on raga-1 neuronal 255 rescue animals, combined with the RNA seq results, strongly suggest that neuronal 256 TORC1 actively causes the release of neuropeptide signals to limit longevity cell 257 nonautonomously. 258 259 Neuronal RAGA-1 drives peripheral mitochondrial fragmentation in aging animals 260 Since genes linked to organelle organization were upregulated in the raga-1 mutant but 261 not in rescue, we explored whether this might be causal to longevity of the raga-1 mutants. 262 Mitochondria can dynamically move between fused and fragmented networks in response 263 to changes in the cellular environment, mediated in C. elegans by the GTPases: FZO-1 264 (fusion) and DRP-1 (fission) (Wai & Langer, 2016). We examined mitochondrial networks 265 in young and old C. elegans in multiple tissues using reporters expressing GFP fused 266 with a fragment from the mitochondrial outer membrane protein TOMM-20, which 267 includes a transmembrane domain that anchors the fusion protein to mitochondria (Weir 268 et al., 2017). For neurons and muscle, we developed an ImageJ/FIJI macro, 'MitoMAPR', 269 to characterize and quantify changes in mitochondrial architecture and morphology in C.  supplement 5). Together, these data suggest that loss of raga-1 specifically in neurons 301 can maintain youthful mitochondrial network states with age in non-neuronal tissues. We sought to determine whether the changes we observed in mitochondrial network state 315 were causally associated with RAGA-1 longevity. First, we asked whether raga-1 mutant 316 animals require a fused mitochondrial network to extend lifespan. We crossed the raga- has also been shown to inhibit AMPK in the hypothalamus in mice (Dagon et al., 2012). 387 Together these data support a hypothesis that AMPK mediates TORC1 longevity in These results suggest that neuronal TORC1 might modulate lifespan through a specific 398 mechanism that is uncoupled from the broad effects of TORC1 on growth and anabolism, 399 for example via its regulation of insulin-like or other neuropeptides that act systemically 400 to regulate longevity. Understanding where and how RAGA-1 acts in the nervous system 401 and whether suppressing TORC1 signaling only in neurons either genetically or 402 pharmacologically is sufficient to promote healthy aging is now a key future goal.

Worm strains 457
Worms were grown at 20°C on nematode growth media (NGM) plates seeded with E. 458 coli strain OP50-1(CGC) with standard techniques (Brenner, 1974). Information for all 459 strains used is in Supplemental file 4. 460 461

RNAi feeding 462
Feeding RNAi clones were obtained from the Ahringer or Vidal RNAi libraries and 463 sequence-verified before using. To use, bacteria were grown overnight in LB broth with 464 100 μg/mL carbenicillin and 12.5 μg/mL tetracycline, seeded on NGM plates with 100 465 μg/mL carbenicillin (NG Carb) and allowed 48 hours to grow at room temperature. At 466 least 4 hours before use, 0.1M IPTG solution with 100 μg/mL carbenicillin and 12.5 467 μg/mL tetracycline was added to the bacterial lawn to induce dsRNA expression. 468

Lifespan experiments 470
All worms were kept fed for at least two generations on OP50-1 bacteria. Before the 471 start of each lifespan experiment, gravid adult worms were bleached and eggs were fed 472 HT115 bacteria until adulthood to either start the lifespan experiment or a timed egg lay 473 to obtain synchronized populations. Day 1 of lifespan marks the onset of egg laying. In 474 the cases where FUDR is used, plates were seeded with bacteria, allowed 24 hours to 475 grow and 100 μl of 1 mg/mL FUDR solution was seeded on top of the bacteria lawn for 476 each plate containing 10 mL NGM. FUDR was allowed 24 hours to diffuse to the whole 477 plate before plates were used. When combining FUDR with RNAi treatments, to 478 overcome potential inhibition of FUDR on dsRNA expression in bacteria, plates were 479 induced with IPTG solution 18 hours after seeding; FUDR was applied 24 hours after 480 seeding; IPTG was applied again 4 hours before use. Image Lab software (Version 4.1). 500 501

Genotyping of deletion alleles 502
Worms were individually lysed in single worm lysis buffer and lysates were used as 503 templates for PCR reactions with a combination of 2-3 primers that will produce bands 504 of different sizes for wild type and mutant alleles. Primers and PCR conditions for each 505 deletion allele are listed in Supplemental file 5. 506 507

Generation of transgenes 508
To generate transgenic animals expressing raga-1 in neurons, raga-1 cDNA was PCR 509 amplified and cloned using standard techniques into a plasmid where a 3X FLAG tag 510 was added to the N terminus and a gpd-2 SL2 sequence with mCherry ORF was added 511 between raga-1 stop codon and unc-54 3'UTR. rab-3 promoter was subsequently PCR 512 amplified and inserted. To express wild type and mutated forms of aak-2, both 3kb 513 promoter region before the aak-2 gene and the 6.7kb coding region were amplified from 514 N2 genomic DNA and cloned into a plasmid, where a 3X FLAG tag and unc-54 3'UTR 515 were added to the C terminus. Serine-to-alanine mutation was generated using 516 Worms of desired stage/age were anesthetized in 0.5 mg/mL tetramisole (Sigma, 566 T1512) diluted in M9 and mounted to 2% agarose pads. For imaging of the 567 mitochondrial TOMM-20 reporter in the muscle, images were taken using a Zeiss 568 Imager.M2 microscope. Apotome optical sectioning was used to acquire fluorescence 569 and one picture with best focus was chosen for each worm for quantification (as were selected as ROIs, processed and filtered using the CLAHE plugin (Zuiderveld, 1994) 582 with median filter and unsharp mask to increase the local contrast and particle 583 distinctiveness. The ROI is then converted to a binary image to generate a 2D skeleton 584