Increased mitochondrial fusion allows the survival of older animals in diverse C. elegans longevity pathways

Mitochondria are dynamic organelles that undergo fusion and fission events. Mitochondrial dynamics are required for mitochondrial viability and for responses to changes in bioenergetic status. Here we describe an insulin-signaling and SCFLIN-23-regulated pathway that controls mitochondrial fusion in Caenorhabditis elegans by repressing the expression of the mitochondrial proteases SPG-7 and PPGN-1. This pathway is required for mitochondrial fusion in response to physical exertion, and for the associated extension in lifespan. We show that diverse longevity pathways exhibit increased levels of elongated mitochondria. The increased mitochondrial fusion is essential for longevity in the diverse longevity pathways, as inhibiting mitochondrial fusion reduces their lifespans to wild-type levels. Our results suggest that increased mitochondrial fusion is not a major driver of longevity, but rather is essential to allow the survival of older animals beyond their normal lifespan in diverse longevity pathways.

M itochondria play diverse roles in signaling, physiology, and metabolism 1 . Mitochondrial dynamics regulate the morphology, number, and function of mitochondria to allow adaptation to cellular needs 1 . Mitochondrial fission is required for: mitophagy; the mitotic segregation of mitochondria to daughter cells; and the distribution of mitochondria to subcellular locations, such as neuronal axons 2 . Mitochondrial fusion is required for maintaining mitochondrial membrane potential and respiratory capacity, and to protect against apoptosis in mammalian cells 3 .
Cells adjust mitochondrial morphologies to coordinate between the cellular demand for energy and the availability of resources 3 . Elongated morphology is associated with increased efficiency of ATP production and reduced generation of reactive oxygen species (ROS), while fragmented morphology is linked to reduced ATP production and mitochondrial uncoupling.
Mitochondrial fusion and fission events are tightly regulated and require the activity of evolutionarily conserved GTPases 4 .
Mitochondrial fission in yeast, invertebrates, and mammals requires the dynamin-related protein DNML1 (Dnm1 in budding yeast and DRP-1 in C. elegans). Mitochondrial fusion requires mitofusins for outer membrane fusion (MFN1 and MFN2 in mammals, Fzo1 in budding yeast, and FZO-1 in C. elegans), and inner membrane fusion (OPA1 in mammals, Mgm1 in budding yeast, and EAT-3 in C. elegans) [4][5][6] . In mammalian cells, overexpression of the outer membrane fusion proteins MFN1 and MFN2 can lead to either clustering of spherical mitochondria [7][8][9] or elongated mitochondria 9,10 . It is likely that the different outcomes result from differences in the level of expression, as it was shown that modest overexpression of the inner membrane fusion protein OPA1 induces mitochondrial fusion, while higher levels of expression induce fragmentation 11 . In C. elegans, the overexpression of either the outer membrane mitofusin FZO-1 or the inner membrane fusion EAT-3 was reported to induce mitochondrial fragmentation 6     regulates mitochondrial shape via spg-7 and ppgn-1. a Real-time PCR quantification of spg-7 mRNA levels in adults of the indicated genotypes, presented in arbitrary units (a.u.) normalized to control rpl-19 (ribosomal protein L19) mRNA. Error bars denote s.e.m. from four biological replicates, each with at least two technical replicates. P values were determined by Student's t-test. b Images of tubular, elongated, and fragmented mitochondria visualized by mitochondria-targeted GFP expressed in body-wall muscle cells. Scale bar, 10 µm. c The percentages of muscle cells with tubular, elongated, and fragmented mitochondria for the indicated genotypes and RNAi treatments. All animals were fed either gene-specific RNAi or control RNAi so that the feeding conditions were matched. P values were determined by χ 2 -test. Sample size (n) of muscle cells from left to right are: 729; 126; 210; 124, 210; 201; 234; 274; 209; 297; 318; 457. Mitochondrial morphology was scored blinded. d Real-time PCR quantification of ppgn-1 mRNA levels normalized to control rpl-19 mRNA in adults of the indicated genotypes. Error bars denote s.e.m. from three biological replicates, each with at least two technical replicates. P values determined by Student's t-test. For all panels, asterisks above bars denote P value comparisons to wild-type/controls; asterisks above lines denote comparisons under the lines: *P < 0.05; **P < 0.01; ***P < 0.001; ns = not significant mammalian results, the failure to generate elongated mitochondria could have resulted from excessive expression levels.
Mitochondrial dysfunction, impaired energy homeostasis, and increased production of ROS are associated with aging in both invertebrates and vertebrates 12 . In large-scale C. elegans RNA interference (RNAi) screens for genes regulating lifespan, one of the largest classes of genes were those encoding mitochondrial proteins affecting the electron transport chain (ETC) 13 , the RNAi depletion of which can either shorten or extend lifespan. Lifespan extension in response to ETC impairment occurs, at least in part, from the activation of the mitochondrial unfolded protein response (UPR mt ) 14 . In this study, we present evidence that increased mitochondrial elongation, which occurs independently of UPR mt , allows the survival of older animals in diverse longevity pathways.
Insulin/IGF-1 signaling (IIS) is an evolutionarily conserved pathway that controls lifespan 15 . In C. elegans, IIS reduces lifespan predominantly by inhibiting the activity of the FOXO transcription factor DAF-16 via inhibitory phosphorylation that blocks its nuclear localization. In the absence of IIS activity, DAF-16 enters the nucleus and regulates gene expression to extend lifespan.
In this study, we describe a pathway for the control of mitochondrial fusion in C. elegans that is regulated by IIS and a cullin-RING ubiquitin ligase (CRL). CRLs are multisubunit ubiquitin ligases (E3s) that ubiquitylate substrate proteins to induce proteasome-mediated degradation or post-translational regulation 16,17 . CRLs include a cullin protein as a scaffold, a RING finger protein, which binds the ubiquitin-conjugating enzyme, a substrate receptor that binds the substrate, and (generally) an adaptor that links the substrate receptor to the core complex. Substrate receptors are variable components, and core CRL complexes function with multiple substrate receptors. The binding of a different substrate receptor to the core CRL complex changes the substrates that are targeted and the cellular function. The most widely studied class of CRLs contains the cullin CUL1 and is designated SCF for its core components: the Skp1 adaptor; CUL1; and F-box protein substrate receptors.
The CAND1 protein functions as an exchange factor for CRL substrate receptors [18][19][20] . CAND1 affects the steady-state levels of different substrate receptors with the core CRL components [18][19][20] . In diverse organisms, the loss of CAND1 selectively affects a subset of CRLs, suggesting that certain substrate receptors are particularly reliant on CAND1 activity 16,17 .
Here, we describe a C. elegans mitochondrial fusion pathway that is regulated by CAND-1, the E3 SCF LIN-23 , and IIS. We show that this pathway is responsible for an increase in elongated mitochondria in response to physical exercise. Additionally, we show that increased levels of elongated mitochondria are associated with diverse lifespan extension pathways, and that the increase in mitochondrial elongation is required to allow the survival of older animals during extended lifespans.
CAND-1 and LIN-23 regulate mitochondrial morphology. Since the spg-7(ek25) suppressor regulates the expression of a mitochondrial m-AAA protease, we wanted to determine whether cand-1 mutants have a mitochondrial phenotype. To characterize mitochondrial morphology, we used a transgenic strain expressing mitochondria-targeted green fluorescent protein (GFP) in muscle cells 23 . Mitochondrial morphology was scored blinded, and the scoring correlates with a quantitative assessment of mitochondrial area ( Supplementary Fig. 4). The majority of wild-type hermaphrodite body-wall muscle cells have longitudinally arrayed tubular mitochondria (Fig. 1b, c). Smaller percentages of muscle cells exhibit elongated mitochondria in an interconnected mesh-like network, or fragmented mitochondria (Fig. 1b, c). Inactivation of cand-1 significantly increases the percentage of cells exhibiting elongated mitochondria; moreover, this phenotype is rescued by the spg-7(ek25) suppressor mutation in cand-1, spg-7(ek25) animals ( Fig. 1c and Supplementary Fig. 5). Inactivation of the COP9/Signalosome component CSN-3, which, like CAND-1, is required for CRL function 16,17 , also increased the proportion of elongated mitochondria ( Fig. 1c and Supplementary Fig. 5). This suggests that the increased mitochondrial elongation phenotype in cand-1 mutants arises from loss of CRL activity. The E3 SCF  , which contains the substrate receptor LIN-23, is particularly reliant on CAND-1 for activity 21 . We found that inactivation of lin-23 similarly reduces spg-7 expression and has increased levels of elongated mitochondria (Fig. 1a, c and Supplementary Fig. 5).
CAND-1 and SCF LIN-23 regulate mitochondria through DAF-16. spg-7(ad2249) loss-of-function and spg-7(tm2312) deletion (Δ) mutants exhibit mitochondrial elongation, but the proportion of mitochondria with elongated morphology is not as large as that observed in cand-1(RNAi) or lin-23(RNAi) animals ( Fig. 1c and Supplementary Fig. 5). Mammalian m-AAA proteases form hexameric complexes in the inner mitochondrial membrane that exist as hetero-oligomeric complexes of AFG3L2 and SPG7, or homo-oligomeric complexes of AFG3L2 24 . We considered whether the partial mitochondrial fusion phenotype of spg-7/ AFG3L32 mutants could arise from functional redundancy between SPG-7 and PPGN-1, the C. elegans ortholog of mammalian SPG7.
We found that inactivation of ppgn-1 alone increased mitochondrial fusion, while inactivating both ppgn-1 and spg-7 further increased the level of elongated mitochondria, implying that both m-AAA proteases negatively regulate mitochondrial elongation (Fig. 1c). CAND-1 and LIN-23 also promote ppgn-1 expression (Fig. 1d), suggesting that the increase in mitochondrial elongation in cand-1 and lin-23 mutants arises from the failure to adequately express both spg-7 and ppgn-1.
Unexpectedly, the reduction of ppgn-1 mRNA levels observed in cand-1 mutants was also rescued by the spg-7(ek25) suppressor allele. The mechanism for this effect is not known but could involve feedback loop(s) or the co-regulation of spg-7 and ppgn-1 expression. We wanted to determine whether CAND-1 and SCF LIN-23 promote spg-7 and ppgn-1 expression by inhibiting DAF-16 activity. One of the primary mechanisms to control DAF-16 activity is by regulating its nuclear localization 15 . We observed that inactivation of cand-1, cul-1, and lin-23 significantly increased DAF-16::GFP nuclear localization, suggesting that CAND-1 and SCF LIN-23 normally act to inhibit DAF-16 nuclear localization (Fig. 2a, b). To determine whether the CAND-1mediated inhibition of mitochondrial fusion is dependent on DAF-16 activity, we combined daf-16 and cand-1 loss-of-function mutations. The addition of the daf-16 mutation rescued the cand-1-elongated mitochondria phenotype, suggesting that CAND-1 inhibits mitochondrial elongation by negatively regulating DAF-16 activity (Fig. 1c).
To determine how SCF LIN-23 inhibits DAF-16 nuclear localization, we analyzed the kinase AKT-1, which phosphorylates DAF-16 to prevent its nuclear localization 15,27 . Inactivating cand-1 or the SCF LIN-23 components cul-1 or lin-23 leads to the accumulation of a faster-migrating form of AKT-1:: GFP without affecting its overall levels ( Fig. 2c and Supplementary Fig. 6). AKT-1 is activated by PDK-1 phosphorylation 15,27 . RNAi depletion of pdk-1, or the inhibition of the IIS pathway upstream of PDK-1 and AKT-1, results in the accumulation of a faster-migrating form of AKT-1::GFP that is consistent with the unphosphorylated, inactive form ( Fig. 2c) 28 . pdk-1 RNAi does not produce a further shift in the migration of AKT-1::GFP when combined with lin-23 RNAi (Fig. 2c). This suggests that the lower molecular weight form of AKT-1::GFP in lin-23(RNAi) animals reflects the loss of PDK-1-dependent activating phosphorylation.

Percentage of cells Tubular Elongated Fragmented
SPG-7 and PPGN-1 regulate mitochondrial fusion protein EAT-3. In yeast and mammals, the inner mitochondrial fusion protein, Mgm1 or OPA1, respectively, is cleaved into long (L) and short (S) isoforms, both of which are required for mitochondrial fusion 29,30 . In mammals, there are two cleavage sites, S1 and S2. S1 is cleaved by the protease OMA1, and S2 is cleaved by the protease YME1L1 29 . C. elegans lacks an OMA1 homolog, and the corresponding S1 cleavage site is not conserved in EAT-3/OPA1. In contrast, C. elegans has a YME1L1 ortholog, ymel-1, and its target S2 site is conserved in EAT-3. The expected sizes of EAT-3 are consistent with an L-isoform (after removal of the mitochondrial targeting sequence) and an S-isoform cleaved at the conserved S2 site (Fig. 2f). Mammalian m-AAA components AFG3L2 and SPG7 can cleave OPA1 when overexpressed or expressed ectopically in yeast 31,32 . We tested whether inactivating C. elegans spg-7 and ppgn-1 affects EAT-3 levels. Co-inactivation of spg-7 and ppgn-1 significantly increased the overall level of EAT-3, indicating that both m-AAA proteases negatively regulate EAT-3 levels (Fig. 2f, g). cand-1 mutants, which have reduced expression of spg-7 and ppgn-1, have higher levels of EAT-3 protein compared to wild type ( Fig. 2f, g). Significantly, the cand-1; spg-7(ek25) strain, which restores expression of spg-7 and ppgn-1 mRNA, does not have elevated EAT-3 levels (Fig. 2f, g). The levels of eat-3 mRNA remain unchanged in the above conditions, indicating that the negative regulation of EAT-3 protein is post transcriptional ( Supplementary Fig. 7). It is not known whether the negative regulation of EAT-3 is direct or indirect. EAT-3 appears to function downstream of CAND-1, LIN-23, SPG-7, and PPGN-1, as the eat-3(RNAi) mitochondrial fragmentation phenotype is epistatic to the mitochondrial fusion associated with inactivation of these genes (Fig. 3a, b).
The levels of FZO-1::GFP (outer mitochondrial membrane fusion protein) and DRP-1::GFP (inner mitochondrial fission protein) were not affected by inactivation of the mitochondrial  Fig. 8a, b). We also did not observe a major change in the levels of mitochondria-targeted GFP upon RNAi depletions of the mitochondrial fusion pathway genes ( Supplementary Fig. 8c). This suggests that the mitochondrial fusion pathway does not significantly affect the level of mitophagy, as changes in mitophagy affect the level of mitochondria-targeted GFP 33 . Our data support the model shown in Fig. 4, in which CAND-1 activates SCF LIN-23 , and SCF LIN-23 activates AKT-1, thereby reducing DAF-16-mediated transcriptional repression of spg-7 and ppgn-1. SPG-7 and PPGN-1 inhibit mitochondrial fusion, at least in part, by negatively regulating the level of EAT-3.
Regulation of mitochondrial fusion in physical exertion. Acute physical exertion in mice induces mitochondrial fusion in skeletal muscle cells 34 , presumably to increase the efficiency of ATP production 3 . We wanted to determine whether physical exertion similarly increases mitochondrial fusion in C. elegans, and whether the fusion is DAF-16-dependent. To induce physical exertion, we utilized swimming behavior. Animals were kept in a state of constant swimming by gently rocking them in M9 buffer solution that contained OP50 bacteria. Swimming, which manifests as an intense thrashing motion (Supplementary Movies 1 and 2), appears to be vigorous exercise based on the observation that ATP levels decrease during swimming (Fig. 5a). Wild-type animals and daf-16(mu86) mutants expressing mitochondria-targeted GFP swam vigorously at an equivalent swimming rate ( Fig. 5b and Supplementary Movies 1-3). Wild-type animals showed an increase in elongated mitochondria that reached an elevated plateau at 60 min that persisted through the remainder of the 5-h test period ( Fig. 5c and Supplementary  Fig. 9a). Significantly, daf-16 mutants did not exhibit increased levels of elongated mitochondria, implying that DAF-16 is required for this physiological response ( Fig. 5c and Supplementary Fig. 9b). Consistently, the nuclear localization of DAF-16::GFP increased in response to swimming in wild-type animals ( Fig. 5d and Supplementary Fig. 10). Notably, cand-1; spg-7(ek25) mutants, which have stabilized spg-7 and ppgn-1 expression, exhibited a lack of mitochondrial fusion in response to physical exertion despite a slightly faster swim rate (Fig. 5b, c and Supplementary Fig. 9c). Similarly, we did not observe an increase in mitochondrial fusion in spg-7(ek25) mutants, which also swam at a rate equivalent to wild-type animals ( Supplementary Fig. 11). This suggests that the DAF-16-mediated Percentage of cells    Fig. 1c was analyzed at the same time and is shown here for comparison. Asterisks above bars denote P value comparisons to wild type/control; asterisks above lines denote comparisons under the lines: *P < 0.05; **P < 0.01; ***P < 0.001; ns = not significant inhibition of spg-7 and ppgn-1 regulates mitochondrial dynamics in response to physical exertion.
Mitochondrial fusion is required for IIS lifespan extension. DAF-16 is required for the extended lifespan of IIS pathway mutants 27 . Because CAND-1 and SCF LIN-23 inhibit DAF-16 nuclear localization, we hypothesized that inactivation of these regulators would extend lifespan. As expected, RNAi depletion of cand-1 and lin-23 (begun in the late-L4 larval stage to ensure normal larval development) extended lifespan ( Fig. 6a; see Supplementary Table 2 for statistics on lifespan data). Interestingly, the cand-1; spg-7(ek25) strain, which rescues the cand-1 increase in elongated mitochondria, had normal lifespan, suggesting that the extended lifespan of cand-1 mutants requires increased levels of elongated mitochondria (Fig. 6b).
The mitochondrial fusion pathway includes activating AKT-1 to repress the DAF-16 transcription factor. The IIS pathway has the same downstream pathway, and so mutation of the IIS receptor DAF-2 would also be expected to activate DAF-16-mediated repression of spg-7 and ppgn-1 to activate mitochondrial fusion. Consistent with this, daf-2 mutants exhibit increased levels of elongated mitochondria (Fig. 6g). To determine the extent to which increased levels of elongated mitochondria contribute to the extended daf-2 mutant lifespan, we inactivated the mitochondrial fusion gene eat-3 in daf-2 mutants. Strikingly, eat-3 RNAi reduced the extended lifespan of daf-2 mutants to wild-type levels (Fig. 6e). To test an alternate method to reduce elongated mitochondria in daf-2 mutants, we RNAi-depleted the outer membrane fusion protein FZO-1; moreover, this also significantly reduced the extended lifespan of daf-2 mutants (Fig. 6f). These results suggest that increased levels of mitochondrial elongation are required for lifespan extension in daf-2 mutants.
Elongated mitochondria are known to produce ATP more efficiently 3 . We analyzed ATP levels in several key mutants, and found a correlation between increased levels of elongated mitochondria and increased ATP/protein ratios ( Supplementary  Fig. 12). The daf-2 mutant is known to have significantly increased ATP levels 35 , and had the highest ATP levels in our assay. Additionally, statistically significant increases in ATP levels were observed in lin-23(RNAi), cand-1(lf), and ppgn-1(RNAi) animals, all of which have increased levels of elongated mitochondria. spg-7(RNAi) animals, which do not exhibit increased levels of elongated mitochondria (Fig. 3b), did not have significantly increased ATP levels. The observation that cand-1(lf); spg-7(ek25) mutants had significantly lower ATP levels than cand-1(lf) mutants suggests that a significant portion of the increase in ATP levels in cand-1(lf) mutants can be attributed to the reduction of spg-7 expression. That ATP levels in cand-1(lf); spg-7(ek25) remained higher than in wild type may reflect the impact of the loss of CAND-1 on other CRL-regulated processes.
Physical exertion extends lifespan. We observed that during swimming, DAF-16 nuclear localization increases, which is associated with lifespan extension. We wanted to determine whether exercise would increase the median lifespan. Making animals swim initially for 30 min per day, followed by consecutive reductions in swim time of 1 min per day, produced significant lifespan extension. However, this swimming regimen (A) caused a rapid drop-off in the viability of older animals (Fig. 7a, see Supplementary Table 3 for statistics). To limit the stress on older animals, we tested two different swimming regimens (B and C)  Fig. 7 Exercise extends lifespan that is dependent on the mitochondrial fusion pathway. a-d Survival curves for adults that were kept continuously on agar plates or removed from the plates for brief periods for the described swim regimens. a Comparison of wild-type kept on agar plates continuously or subject to swim regimen A (30 min of swimming per day). b Test of swim regimens B and C. Swim regimen B was the following swimming times per day on the listed days: 30 min on days 1-5; 25 min on days 6 and 7; 20 min on days 8 and 9; 15 min on day 10; 10 min on day 11; and 5 min on day 12. Swim regimen C was 30 min of swimming on day 1, followed by a reduction in swim time of 2 min per day on subsequent days. c, d Test of swim regimen C with wild type, cand-1(tm1683); spg-7(ek25) c, or daf-16(mu86) d. The wild-type control and wild-type swim regimen C survival curves were analyzed at the same time and are shown in c, d for comparison. All swimming regimens significantly increased the mean lifespan of wild type a-d. Swim regimen C did not increase the mean lifespan of cand-1(tm1683); spg-7(ek25) c or daf-16(mu86) mutants d relative to the control, non-swimming condition. All lifespan experiments were performed with four biological replicates. See Supplementary Table 3  that further reduced the extent of swimming as the animals aged: an initial rate of 30 min swimming/day was tapered via arbitrary reductions (regimen B, Fig. 7 legend), or reductions of two min per day (regimen C). These age-moderated swim regimens produced further lifespan extensions relative to regimen A (Fig. 7b).
Notably, cand-1; spg-7(ek25) and daf-16 mutants, which failed to exhibit increased mitochondrial fusion in response to swimming, did not exhibit lifespan extension in response to swimming (Fig. 7c, d). This suggests that the induction of mitochondrial fusion during exercise contributes to the observed lifespan extension.
Because of how the swim experiments were carried out, the lifespans of the control animals are not directly comparable to the lifespans of control animals in other (non-swim) experiments. Control animals in the swim experiments had shorter lifespans than the control animals in non-swim experiments. Presumably, this difference arose because all animals in the swim experiments (including controls) were kept at room temperature during the swim periods (rather than the lower temperature of 20°C) and were transferred daily to new plates. Notably, the swim regimens B and C produced a significant extension of lifespan even when compared to control animals from the non-swim lifespan experiments (Supplementary Table 3).

Mitochondrial fusion is broadly required for longevity.
To determine whether mitochondrial fusion is more broadly correlated with longevity, we analyzed the mitochondrial morphologies of animals that exhibit longevity from five distinct mechanisms: age-1(RNAi) and pdk-1(RNAi) animals have extended lifespan from loss of IIS 15 ; eat-1(RNAi) and eat-6(RNAi) animals have extended lifespan linked to caloric/dietary restriction 36 ; clk-1(RNAi) extends lifespan because of mitochondrial ETC dysfunction; glp-1(RNAi) extends lifespan as a result of the loss of the germline 37 ; and vhl-1, the von Hippel Lindau tumor suppressor ortholog, mutants/RNAi animals have lifespan extension due to deregulation of the hypoxia transcriptional program 38 .
We analyzed the distribution of mitochondrial morphologies in the seven long-lived strains. Strikingly, six of the seven long-lived mutants/RNAi animals had increased levels of elongated mitochondria, with vhl-1(RNAi) animals the lone exception, with predominantly tubular mitochondria (Fig. 8a, b and Supplementary Fig. 5). Caloric restricted mutants, such as eat-6, have lifespan extension independent of DAF-16 36 . We found that the increase in elongated mitochondria in eat-6 mutants was not affected by daf-16 RNAi, suggesting that mitochondrial morphology in these mutants is regulated through a DAF-16independent pathway (Fig. 8a). In other organisms, nutrient limitation increases mitochondrial fusion 3 , and it is possible that a similar mechanism operates in C. elegans.
Strikingly, the increase in elongated mitochondria in these diverse mutants is essential for their longevity. RNAi depletion of the mitochondrial fusion gene eat-3 abolished the extended lifespans of diverse longevity mutants/RNAi: glp-1(e2141ts); eat-2(ad1116) and eat-6(ad467); and clk-1(RNAi) (Fig. 9a-d). Significantly, eat-3 RNAi depletion in a wild-type background or in vhl-1(ok161) mutants, neither of which have significantly elevated elongated mitochondria, did not affect their lifespans (Fig. 9e). This suggests that eat-3 RNAi does not have a negative For all panels, asterisks above bars denote P value comparisons to wild-type/control; asterisks above lines denote comparisons under the lines: *P < 0.05; **P < 0.01; ***P < 0.001; ns = not significant impact on lifespan directly, but rather acts indirectly by reducing the beneficial effects of increased levels of elongated mitochondria. Elongated mitochondria have decreased production of ROS 3 . We measured mitochondrial ROS levels using the dye MitoSOX Red 39 in daf-2, glp-1, and eat-6 mutants, all of which have elevated levels of elongated mitochondria. Each of these long-lived mutants had reduced ROS levels, and ROS levels increased significantly upon treatment with eat-3 RNAi to block mitochondrial fusion ( Supplementary Fig. 13 and Supplementary      We considered the possibility that the effect on lifespan by the mitochondrial fusion pathway was primarily due to changes in energy levels that were a secondary consequence of the changes in mitochondrial morphology. We analyzed three energy sensors that have an impact on lifespan: LET-363/TOR; SIR-2.1; and AAK-2/AMPK. LET-363/TOR responds to metabolic inputs to regulate lifespan through inhibition of the transcription factors DAF-16, SKN-1, and PHA-4 40,41 . SIR-2.1 is a NAD + -dependent histone deacetylase whose overexpression increases lifespan by activating DAF-16, UPR mt , and autophagy 42 . The AMP-activated protein kinase AAK-2 extends lifespan in response to ROS signals and impaired glycolysis, such as inactivation of GPI-1, glucose phosphate isomerase 35,43 . Overexpression of SIR-2.1 or AAK-2, and RNAi depletion of let-363/TOR or gpi-1, all exhibited increased mitochondrial elongation (Figs. 8b, c and 9h, i and Supplementary Fig. 5). eat-3 RNAi reduced the extended lifespan of these animals, suggesting that the elongated mitochondrial morphology contributes to their lifespan extension (Fig. 9f-i and Supplementary Fig. 5).
The induction of UPR mt is mediated by the transcription factor ATFS-1, which is required for the expression of hsp-6 and other UPR mt -regulated genes 46 . Counterintuitively, a gain-of-function mutation of ATFS-1, atfs-1(et17gf) 47 , with constitutively activated UPR mt , exhibits reduced lifespan, not lifespan extension 48 . We used the atfs-1(et17) gain-of-function mutant to ask whether activation of the mitochondrial fusion pathway reduces lifespan further, as would be expected if the pathway further activated the UPR mt . We found that RNAi depletion of cand-1, lin-23, spg-7, and ppgn-1 increased the lifespan of atfs-1(et17) mutants, suggesting that the lifespan extension operates independently of the UPR mt (Fig. 10c, d). Induction of UPR mt by loss of spg-7 is abolished upon the loss of atfs-1 46 . However, we did not observe significant differences in mitochondrial morphology between spg-7(Δ) mutants and spg-7(Δ); atfs-1(RNAi) mutants (Fig. 10e). The UPR mt is The quantification was performed as for b and the two graphs can be compared directly. No significant differences in Phsp-6::GFP were observed relative to 0 min control animals. For all panels, asterisks above bars denote P value comparisons to wild type/controls; asterisks above lines denote comparisons under the lines: *P < 0.05; **P < 0.01; ***P < 0.001; ns = not significant also not induced upon swimming, unlike the mitochondrial fusion pathway (Fig. 10f). These results suggest that the increase in mitochondrial fusion associated with inactivation of the IIS/SCF LIN-23 -regulated pathway occurs independently of UPR mt .

Discussion
Our study has revealed a pathway by which the ubiquitin ligase SCF LIN-23 and IIS control mitochondrial fusion in C. elegans (Fig. 4a). We found that the CRL regulator CAND-1 and SCF LIN-23 are required to activate AKT-1. Active AKT-1 inhibits the nuclear localization of DAF-16, which inhibits the expression of the mitochondrial m-AAA proteases SPG-7 and PPGN-1.
Decreased expression of SPG-7 and PPGN-1 is associated with increased levels of the mitochondrial fusion protein EAT-3, whose activity is essential for the increased mitochondrial fusion that is observed upon inactivating the pathway components CAND-1, LIN-23, and PPGN-1. Our analysis does not indicate whether SPG-7 and PPGN-1 directly reduce EAT-3 levels or whether the impact on EAT-3 is indirect. Given the additional functions of the yeast and mammalian orthologs of spg-7 and ppgn-1, it is likely that reducing their activity affects multiple mitochondrial pathways 29 . Additionally, SCF E3s are known to target multiple substrates 16,17 , and we cannot rule out the possibility that CAND-1 and SCF LIN-23 have an impact on mitochondria through other pathways. Our work identifies the SCF LIN-23 -mediated activation of AKT-1 as a new regulatory mechanism for IIS, for which AKT-1 is a critical component. Currently, it is unclear how SCF LIN-23 promotes AKT-1 phosphorylation. Ubiquitylation has been linked to AKT activation in mammalian cells, wherein K63-linked poly-ubiquitylation of AKT is required for its translocation to the plasma membrane, where it is activated 49,50 . In mammalian cells, the K63 linkage is only detectable with anti-K63 ubiquitin antibodies immediately after stimulation of the relevant signaling pathways, but not in unstimulated cells 49,50 . We have been unable to detect K63 poly-ubiquitin on immunoprecipitated AKT-1::GFP. However, the inability to rapidly induce signaling in intact animals limits the implications of this negative result.
In mouse skeletal muscle cells, physical exercise increases mitochondrial fusion 34 . We have shown a similar result in C. elegans using swimming as a source of exercise. C. elegans swimming is associated with a reduction in the level of ATP, indicating that the "thrashing" swim strokes involve a greater expenditure of energy than the normal sinusoidal movement on plates. Swimming induces an increase in the level of elongated mitochondria, and this increase appears to be controlled by the IIS/SCF LIN-23 -regulated pathway, as the increase in elongated mitochondria is blocked in daf-16 mutants as well as cand-1; spg-7(ek25) mutants, which are no longer responsive to DAF-16-mediated inhibition of spg-7 expression.
In rodents, modest physical exercise correlates with increases in mean lifespan and healthspan 51 , and we have found that exercise regimens increase C. elegans lifespan. The lifespan extension in response to exercise is not observed in daf-16 mutants or cand-1; spg-7(ek25) mutants, which are unable to activate the IIS/SCF LIN-23 pathway to increase mitochondrial fusion, suggesting that increased mitochondrial elongation positively contributes to exercise-induced longevity.
We observed an increase in elongated mitochondria in diverse longevity pathways, including: IIS inactivation; caloric restriction; germline depletion; ETC dysfunction leading to UPR mt ; TOR inactivation; Sirtuin overexpression; AMPK overexpression; impaired glycolysis; and exercise. RNAi depletions of six genes that cause ETC dysfunction leading to extended lifespan were previously shown to increase mitochondrial fusion 52 . These observations suggest that increased levels of elongated mitochondria are broadly associated with longevity.
Notably, the extended lifespans for the longevity pathway mutants with increased mitochondrial elongation were significantly reduced by eat-3 RNAi, which reduces mitochondrial fusion. eat-3 RNAi does not affect the lifespans of wild-type animals or long-lived vhl-1 mutants, neither of which exhibit increased levels of elongated mitochondria. This suggests that inactivation of the mitochondrial fusion protein EAT-3 does not directly reduce lifespan, but rather indirectly reduces lifespan by countering the beneficial effects of elongated mitochondria. RNAi depletion of the outer membrane mitochondrial fusion protein FZO-1 also reduces daf-2 mutant-extended lifespan.
ppgn-1 mutants have a substantial increase in elongated mitochondria, yet have only modestly longer lifespans. It has been reported that blocking fission by inactivating DRP-1 reduces 53 or has no effect on the lifespan of wild-type animals 54 , but increases the extended lifespans of IIS pathway mutants 54 . These observations suggest that increasing mitochondrial elongation, by itself, does not greatly extend lifespan. Rather, increased levels of elongated mitochondria appear to be a requirement for the survival of older animals in diverse longevity pathways. Thus, while increased levels of elongated mitochondria contribute to older animal survival, the precise level of the increase would not be expected to be predictive of the extent of longevity.
Elongated mitochondria have increased mitochondrial efficiency 3 ; in addition, we have observed increased ATP levels in mitochondrial fusion pathway mutants that have increased mitochondrial elongation. This leads to the hypothesis that increased mitochondrial efficiency resulting in sufficient ATP production is required in diverse longevity pathways to allow the survival of long-lived adults.
We observed that vhl-1 mutants do not have increased levels of elongated mitochondria and their extended lifespan is not reduced by eat-3 RNAi. Interestingly, a large-scale Cas9 inactivation screen in human cells found that inactivation of VHL allowed a bypass of the complete block on mitochondrial respiration 55 . VHL functions as a substrate receptor for the CRL2 VHL ubiquitin ligase, which degrades the hypoxia-inducible factor HIF1Α. During hypoxia, the HIF1Α transcriptional response is associated with a major reorganization of energy pathways to allow energy generation in the absence of oxidative mitochondrial metabolism 55 . Thus, unregulated HIF1Α transcriptional activity can functionally replace the requirement for mitochondrial respiration in human cells 55 .
C. elegans VHL-1 similarly targets the degradation of the HIF1Α ortholog HIF-1 56 . HIF-1 is required for the lifespan extension of vhl-1 mutants 38 . If the role of HIF-1 in providing alternate energy pathways under hypoxic conditions is conserved, then vhl-1 mutants would have increased energy capacity via HIF-1-mediated induction of energy metabolism. It is possible that increased energy capacity is a requirement for survival beyond the normal lifespan, and that most longevity mutants increase energy capacity via mitochondrial elongation, while vhl-1 mutants utilize HIF-1-mediated transcriptional activity. It will be interesting to determine whether increased mitochondrial fusion and/or energy metabolism contributes to the survival and healthspan of older animals in other species.
The isolation of the ek25 cand-1 suppressor mutation. cand-1(tm1683) mutants were synchronized as L1-stage larvae by isolating eggs by hypochlorite treatment 61 and allowing the eggs to hatch overnight in M9 buffer 61 supplemented with 5 µg ml −1 cholesterol. The synchronized L1-stage larvae were grown on NGM plates with OP50 bacteria until the L4/young-adult stage, when they were mutagenized with 0.5 mM N-nitroso-N-ethylurea (ENU) 62 for 4 h. The mutagenized animals were cultured on NGM plates with OP50 for 1 day, at which time F1 mutant eggs were isolated from the gravid adults and synchronized as L1 larvae. The F1 mutant animals were cultured on NGM plates with OP50 bacteria until they became gravid adults. F2 mutant eggs were isolated from the F1 gravid adults and synchronized as L1 larvae. In all, 10,000 F2 mutant L1 larvae were grown in a 500 ml liquid culture 61 supplemented with OP50 bacteria for 11 days, and then collected by centrifugation and regrown in fresh liquid culture for 3 days. Propagation in liquid culture is a hardship for C. elegans, and this culturing condition selects for healthier cand-1 suppressors, as cand-1(tm1683) mutants have lower brood sizes, impenetrant embryonic lethality, and impenetrant larval arrest 21 . Animals from the liquid culture were placed on 3xNGM plates. Twelve healthy L4 larvae that did not exhibit cand-1 mutant phenotypes (e.g., no bobtail phenotype) were cloned onto separate NGM plates with OP50 bacteria. One of these cloned animals (cand-1(tm1683); ek25) had 100% healthy progeny without visible cand-1 mutant phenotypes.
SNP mapping. The cand-1(tm1683); ek25 strain was outcrossed 10 times into the Hawaiianized cand-1(tm1683) strain ET335. The resulting Hawaiianized cand-1; ek25 strain appears to be largely Hawaiian throughout the genome with the exception of the +7.5 to +11.5 region of chromosome V, which contains the cand-1 gene and remains N2 Bristol. cand-1(tm1683) was crossed with the Hawaiianized cand-1(tm1683); ek25, and confirmed F1s were allowed to lay eggs. In all, 273 L1 larvae from the F2 generation were transferred onto separate csn-3 RNAi plates. cand-1 mutants subjected to csn-3 RNAi have an enhanced phenotype of 100% arrested embryos; in contrast, cand-1; ek25 is highly resistant to csn-3 RNAi (Supplementary Table S1). Using the resistance to csn-3 RNAi as a measure of cand-1 suppression, we identified 40 "suppressed" and 40 "non-suppressed" plates from the 273 F2 cross progeny. We performed whole-genome SNP mapping using previously described sets of primers 63 . Pooled DNA from the suppressed and nonsuppressed progeny were used to generate PCR amplicons that were digested with the Dra I restriction enzyme (as the SNPs differed in the presence or absence of a Dra I site) in order to distinguish Hawaiian from N2 Bristol genomic DNA. We observed an enrichment of Hawaiian regions in the Chromosome I region −4 to −2 in the "suppressed" population ( Supplementary Fig. 1a). We then looked at the distribution of Hawaiian vs. N2 regions in the each of the 40 "suppressed" clones ( Supplementary Fig. 1b). Most clones were Hawaiian in the −4 to −2 chromosome I region, but 12 clones were heterozygous in parts of this region. The only SNP that was Hawaiian in all clones was −3.18, which narrowed the location of the suppressor to the −3.96 to −2.07 interval. By sequencing SNPs in this region, we narrowed the region containing the suppressor to −3.48 to −3.18 and found a 26 bp insertion mutation in the major intron of spg-7 (located at −3.22 on chromosome I). The insertion mutation segregated completely with the rescue of the cand-1 (tm1683) mutant phenotypes through additional outcrossings, suggesting that it is the cand-1 suppressor mutation.
DAF-16 and PQM-1 data meta-analysis. DAF-16 has been found to predominantly function as a transcriptional activator via binding regulatory regions through DBE sites 64 . In contrast, genes that are repressed in a DAF-16dependent manner are predominantly repressed by the PQM-1 transcription factor via binding to DAE sites 25 . However, an analysis of DAF-16-reponsive genes shows that, of 2196 genes repressed in a DAF-16-dependent manner at p < 0.01 significance, 563 (26%) were shown to be bound by DAF-16 but not by PQM-1, and 131 of the genes have position-specific affinity matrices affinity scores of 0 for DAE sites but positive scores for DBE sites 25 . This suggests that DAF-16 can repress gene expression independently of PQM-1. The activating spg-7(ek25) mutation disrupts a DBE site, and both spg-7 and ppgn-1 regulatory regions have been found to be bound by DAF-16 but not by PQM-1 25 . This suggests that spg-7 and ppgn-1 are repressed by DAF-16 in a PQM-1-independent manner.
Epifluorescence microscopy. RNAi-treated and mutant animals were maintained at 20°C, and L4-stage animals to young-adult-stage animals were used for imaging. FZO-1::GFP expressing L4-stage animals were subjected to heat-shock treatment at 33.5°C for 10 h to ensure stable expression prior to imaging. Animals were mounted on slides with 0.5 mM levamisole (Sigma) to induce muscle paralysis. Animals were visualized with a Zeiss Axioskop microscope, and images were taken with a Hamamatsu ORCA-ER digital camera using Openlab 4.0.2 software (Agilent Technologies). Images were processed and analyzed with Adobe Photoshop CS6 or CC software. Matched images were taken with the same exposure and were processed and analyzed identically. Images from 20 animals were analyzed for AKT::GFP ( Supplementary Fig. 6b), FZO-1::GFP ( Supplementary Fig. 7a), DRP-1:: GFP ( Supplementary Fig. 7b), and mitochondrial GFP ( Supplementary Fig. 7c) levels.
The expected molecular weight of EAT-3 isoforms was determined as follows. The mitochondrial targeting sequence (MTS), which is cleaved from the precursor protein to form the L-isoform, was identified using the MitoProt program 65 . The predicted MTS in EAT-3 (NP495986) encompasses residues 1-59. The predicted L-isoform therefore encompasses amino acids 60-964 of EAT-3. The mammalian S2 cleavage site has been identified between amino acids 217-223 of human OPA1 (NP_570849) 66 . The human S2 cleavage site corresponds to residues 215-221 in EAT-3, cleavage of which would produce an S-isoform of between 216-964 and 222-964 amino acids.
Lifespan analysis. Survival assays 67 were performed at 20°C. Eggs were isolated from gravid adults by hypochlorite treatment and allowed to hatch on RNAifeeding bacteria plates, with the exception of cand-1, cul-1, lin-23, clk-1, and let-363 RNAi, where wild-type hermaphrodites were transferred to RNAi-feeding bacteria plates as late-L4-stage larvae to ensure that prior larval development was normal. Animals were transferred to fresh plates every alternate day throughout the lifespan study. For all lifespan studies, animals in the L4 larval stage were picked on day 0. Lifespan analysis of wild-type adult hermaphrodites subjected to a swimming regimen was performed by adding 3 ml of M9 buffer to plates, and the plates were placed on a rocking platform at room temperature to prevent the animals from settling down. The daily swim regimens are described in the main text. After swimming, animals were transferred to fresh agar plates with OP50 bacteria. The non-swim control animals were placed at room temperature for the duration of the swim period, and also transferred to new OP50-seeded NGM plates at the same time as swim animals. Animals that crawled off the plate or had ruptured vulvae were censored. Immobile adults were counted as dead when they failed to respond to prodding. P values were calculated by the Log-rank (Mantel-Cox) and Wilcoxon tests using GraphPad Prism software (version 6.0). ATP quantification. ATP levels were measured using the EnzyLight TM ADP/ATP Ratio Assay Kit (ELDT-100) from BioAssay Systems as per the manufacturer's protocol with the following modifications. Eggs were isolated from gravid adults by hypochlorite treatment and allowed to hatch on RNAi-feeding bacteria plates, or OP50 plates for mutant animals, and maintained at 20°C. L4-stage animals were collected by centrifugation and washed three times in M9 buffer +5 mg ml −1 cholesterol, flash-frozen in liquid nitrogen, and stored at −80°C. Frozen samples were boiled for 15 min to release ATP, and centrifuged at 15,000 × g to pellet debris. The supernatant was diluted twofold in RNase/DNase-free water, and used for ATP measurement. Protein concentration was measured using the Bradford reagent (Sigma). ATP levels were normalized to protein concentration.
Analysis of mitochondrial morphology. Adult hermaphrodites expressing mitochondrial matrix-targeted GFP [Pmyo-3::GFP(mit)] in their body-wall muscle cells 59 were imaged for mitochondrial morphology. Cells were categorized as harboring tubular, elongated, or fragmented mitochondria, as described 5 . Mitochondrial morphology was assessed with at least 100 muscle cells analyzed for each condition, and was scored blinded where noted. To quantitatively categorize cells harboring tubular, elongated, or fragmented mitochondria, the twodimensional areas of mitochondria were measured as previously described 68 . The areas (in pixels) of the five largest mitochondria per cell in~40 body-wall muscle cells were measured from micrographs using Photoshop software (Adobe).
Analysis of mitochondrial ROS. Eggs were placed on RNAi plates and L4-stage larvae were used for the analysis of mitochondrial ROS with the dye MitoSOX Red (Life Technologies), which was performed as previously described 39 . Animals were incubated in 10 µM MitoSOX Red in M9 buffer with cholesterol and RNAi bacteria in the dark at 20°C for 24 h. Animals were washed two times in M9 buffer and incubated in M9 buffer with OP50 bacteria for an hour to clear their intestines of residual dye. ROS levels were analyzed in the posterior pharyngeal bulb by measuring the MitoSOX Red epifluorescence intensity with a 300 ms exposure. Epifluorescence was analyzed with optical filters of 470/40 excitation wavelength and 500 long-pass emission wavelength. Mitochondrial ROS levels were measured in adult hermaphrodites 1 day-post-L4 stage.
Physical exertion swimming assay. L4-stage larvae and young adults expressing the Pmyo-3::GFP(mit) transgene were placed on one 6-cm OP50-seeded NGM plate per time point. To induce the swimming behavior, 3 ml of M9 buffer was added to each plate, and the plates were placed on a rocking platform to prevent the animals from settling down. At each time point, animals were collected by centrifugation, and visualized by epifluorescence microscopy to determine mitochondrial morphology (obtaining data from at least 100 cells). To determine the swim rate, 12 animals were analyzed for the number of swim strokes per minute for each genotype.
Data availability. The authors declare that all data supporting the findings of this study are available within this article, its Supplementary Information files, or are available from the corresponding author upon reasonable request.