Reversible Age-Related Phenotypes Induced during Larval Quiescence in C. elegans

Summary

Cells can enter quiescent states in which cell cycling and growth are suspended. We find that during a long developmental arrest (quiescence) induced by starvation, newly-hatched C. elegans acquire features associated with impaired proteostasis and aging: mitochondrial fission, ROS production, protein aggregation, decreased proteotoxic-stress resistance, and at the organismal level, decline of mobility and high mortality. All signs of aging but one, the presence of protein aggregates, were reversed upon return to development induced by feeding. The endoplasmic reticulum receptor IRE-1 is completely required for recovery, and the downstream transcription factor XBP-1, as well as a protein kinase, KGB-1, are partially required. Interestingly, kgb-1(−) mutants that do recover fail to reverse aging-like mitochondrial phenotypes and have a short adult lifespan. Our study describes the first pathway that reverses phenotypes of aging at the exit of prolonged quiescence.

Introduction

Many cell types are able to suspend proliferation in a reversible manner by entering a state known as quiescence. Quiescence can be induced by withdrawal of growth factors or by other adverse conditions; in fact, many cells are thought to spend the majority of their lifetime in quiescence (Valcourt et al., 2012; Yao, 2014). Very long periods of quiescence reduce cell-proliferation potential and viability in every species studied (Baugh, 2013; Delaney et al., 2013; Soprano, 1994). The mechanism underlying long-term quiescent cell survival and the return to growth and proliferation is of particular interest for understanding stem cell dysfunction during aging (Cheung and Rando, 2013; Sousa-Victor et al., 2014).

Caenorhabditis elegans larvae have the ability to survive long periods of food deprivation by entering a quiescent state. During this developmental arrest, cells exit the cell cycle and cease growth, they slow their metabolism and they increase their stress resistance (Baugh, 2013). Newly hatched, so-called “arrested L1 larvae” can survive in quiescence for weeks, almost twice the normal lifespan of never-arrested animals. When fed, arrested L1s commence development and grow to adulthood. Interestingly, the adult lifespans of these animals are not short; instead they are normal (Johnson et al., 1984). This observation was interpreted to mean that the process of aging had been suspended during L1 arrest (Johnson et al., 1984). Recently, however, mutations that extend or shorten adult lifespan have been found also to extend or shorten the time arrested L1s can survive (Baugh, 2013). For instance, mutations affecting insulin/IGF-1 signaling (Baugh and Sternberg, 2006; Kenyon et al., 1993), translation (Hansen et al., 2007; Lee et al., 2012; Syntichaki et al., 2007), heat-shock factor (Hsu et al., 2003; Morley and Morimoto, 2004), AMP kinase (Apfeld et al., 2004; Baugh and Sternberg, 2006) or sensory perception (Apfeld and Kenyon, 1999; Lee and Ashrafi, 2008) either extend or shorten both adult lifespan and L1-arrest survival. Because the lifespans of arrested L1s are affected by genes that affect adult aging, we considered an alternative to the “suspended-animation” hypothesis; namely that arrested L1s experience a cellular decline that is similar to that of normal aging, but that can be reversed by feeding. This way, the animals could initiate the aging process afresh, and have a normal adult lifespan.

We found that long-arrested L1s do, in fact, acquire several phenotypes characteristic of aging, including proteostasis decline. We found the endoplasmic reticulum (ER) unfolded-protein sensor inositol-requiring enzyme 1 (IRE-1) to be completely required for L1 recovery. In contrast, the downstream transcription factor X-box binding protein 1 (XBP-1) was only partially required. Instead, we found that IRE-1 could act, at least in part, through the mitogen-activated protein (MAP) kinase KGB-1 to promote recovery. Together these findings suggest that the ER is involved in a quality control mechanism that can re-establish youthful cellular homeostasis in animals that exhibit phenotypes resembling those of adult aging.

Results

Long-term L1 quiescence induces signs of cellular stress seen in old adults

Prolonged and reversible developmental arrest of C. elegans L1 larvae did not change their subsequent adult longevity (Figures 1A, B). The first study that described this phenomenon concluded that their aging might be arrested (Johnson et al., 1984); however, an examination of their appearance and behavior suggested otherwise. First, over time the tissues acquired a granular appearance (Figure 1C) similar to that of old adults (Garigan et al., 2002). Their movement slowed and their mortality rate increased (Figures 1D and S1A) as in aging adults (Herndon et al., 2002). In old unicellular fungi, and in C. elegans, mitochondria lose their tubular morphology and undergo fission (Hughes and Gottschling, 2012; Scheckhuber et al., 2007; Yasuda et al., 2006). We found that this phenotype arose during aging in adult muscle cells, and was delayed in long-lived daf-2 insulin/IGF-1-receptor mutants (Figures 1E and S1B–E). A similar, independent, finding was reported recently (Hahm et al., 2015). We found that mitochondrial fragmentation also occurred over time in the muscles of arrested L1s (Figures 1E and S1F–H). We also observed myosin-fiber disorganization in long-arrested L1s (Figure S2), though to a lesser extent than in aging adults (Herndon et al., 2002). Additional phenotypes characteristic of aging appeared in arrested L1s. Both adults and arrested L1s displayed increased levels of reactive oxygen species (ROS) over time, as measured with the ROS-sensitive dye dihydroethidium (DHE) (Figures 1F and S3A). Auto-fluorescence of granules in the posterior intestines increased over the course of the arrest (Figure S3B). Intense auto-fluorescence of cellular granules is a well-known correlate of aging in C. elegans and mammals (Gerstbrein et al., 2005; Höhn and Grune, 2013).

Figure 1. Appearance of age-related phenotypes during C. elegans L1 developmental arrest.

(A) Timeline depicting the experimental protocol. (B) Lifespans of adults previously L1-arrested for 1 or 17 days as L1s were the same. n(blue)=74 and n(red)= 88; p = 0.95. (see Table S1). (C) Nomarski images after 1 day or 30 days of C. elegans L1 arrest. Scale bar, 10 μm. (D) Survival declines during L1 arrest. Animals first lose the ability to develop when fed, and subsequently, as assayed by their failure to move when prodded, they die. Ratio comparing developing ability relative to survival, and replicates are shown in Table S2. (E) Representative confocal images showing fragmentation of the mitochondrial network in muscle cells of aging adults and arrested L1s. Mitochondria were visualized using a mitochondrial-targeted GFP protein (Pmyo-3::mito-gfp). The strain also expressed nuclear-localized GFP (Pmyo-3::NLS-gfp), to facilitate cell identification. Scale bar, 10 μm. n, nucleus; M, mitochondria. Quantifications and additional pictures are shown in Fig. S1C, D, F, G. (F) Increased ROS levels in aging adults and arrested L1s, visualized by microscopy using DHE (quantification in Fig. S3A). Scale bar, 10 μm. G, intestinal granules, also labeled with DHE. (G) DTT resistance declined during L1 arrest. The graph shows the percentage of L1s that developed after stress, normalized to the percentage alive before stress. n > 100 in each of 3 independent cultures, **p = 0.009. (H) Heat resistance declined during L1 arrest. The graph shows the percentage of L1s that developed after stress relative to the percentage of L1s that developed without stress. n > 100 in 3 independent cultures, *p = 0.01. (I) Increased KIN-19::tagRFP aggregation in aging adults and arrested L1s. Scale bar, 10 μm (quantification in Fig. S3C). (J, K) Muscle-specific polyglutamine (Pmyo-3::polyQ35::yfp) aggregated in aging adults and arrested L1s. Each dot represents a worm. L2, L3 and L4, larval stages. Blue bar, average (pictures in Fig. S3E).

During aging, protein quality control is impaired (Demontis and Perrimon, 2010; Ghazi et al., 2007) and adult worms become sensitive to heat and protein-folding stress (Ben-Zvi et al., 2009; Henis-Korenblit et al., 2010). Compared to L1s arrested for 1 day, L1s arrested for 15 days were twice as sensitive to heat and to dithiothreitol (DTT), a reducing agent that causes ER stress, suggesting impaired proteostasis (Figures 1G, H). We and others have shown that many endogenous proteins become detergent-insoluble with age and, when fluorescently tagged, can form visible aggregates (David et al., 2010; Reis-Rodrigues et al., 2012; Walther et al., 2015). We analyzed one such protein by microscopy, KIN-19::tagRFP, and found that it also aggregated during L1-arrest (Figures 1I and S3C). More generally, we found a significant increase in detergent-insoluble proteins in day-22 compared to day-1 arrested L1s (Figure S3D). Age-dependent aggregation of polyglutamine-containing proteins (Poly-Q) is associated with several neurodegenerative diseases (Zoghbi and Orr, 2000). When proteins consisting of 35 Poly-Q repeats fused to YFP are expressed in C. elegans muscle, they remain soluble during development, but form aggregates over time during adulthood (Morley et al., 2002) (Figure 1J). We also found that soluble (diffuse) Poly-Q35 levels began to fall, and punctate Poly-Q35::YFP aggregates began to accumulate in arrested L1s starting at day 5 (Figures 1K and S3E). In summary, phenotypes reminiscent of adult aging also occur during a long period of quiescence in juvenile worms.

Age-related phenotypes, but not residual fat content, predict the ability of arrested L1s to recover

Concomitant with their systemic decline, L1s starved for long periods of time progressively lost their ability to resume development when they were fed (Figure 1D). Conceivably, their failure to develop could be caused by a depletion of energy reserves. However, we observed that worms failing to return to development nevertheless were able to move (data not shown) and to ingest food (Figure S4A), suggesting that energy reserves might not be the limiting factor. We tested this hypothesis quantitatively. Fluorescently-labeled fatty acid (C1-Bodipy-C12) allows one to visualize and quantify fat-granule volume, which correlates with triglyceride levels (Barros et al., 2012; Zhang et al., 2010). Using this indicator, we observed that stored-fat levels decreased during L1 arrest (Figures S4B–D). However, in longitudinal studies of individual animals, we found no correlation between the level of fat and the ability to resume development upon feeding (Figures 2A, S4E, F). In contrast, both the extent of mitochondrial fragmentation in the muscle, as well as KIN-19::tagRFP aggregation in the head, did predict the recovery of long-starved individual L1s (Figures 2B, 2C and S4G). These observations suggest that a decline of proteostasis manifested by the appearance of phenotypes characteristic of aging, rather than depletion of energy stores, limits the animal’s ability to recover from L1 arrest. Consistent with this interpretation, a reduction-of-function mutation in the heat-shock transcription factor HSF-1, which affects protein aggregation, the rate of tissue aging, and adult lifespan in C. elegans (Baird et al., 2014; Garigan et al., 2002; Hsu et al., 2003; Morley and Morimoto, 2004), compromised both survival (Baugh and Sternberg, 2006) and recovery during L1 arrest (Figure S5A).

Figure 2. Age-related phenotypes, but not residual fat content, predict the ability of arrested L1s to recover when fed.

(A) The abundance of fat granules in L1s arrested 20–21 days, measured with Bodipy-C12 fluorescent dye in singled animals, did not correlate with the ability, or the time, to develop to the L2/L3 stage after feeding. a.u.: arbitrary unit of green fluorescence intensity measured in the entire L1 (p = 0.97). (Supporting data, Figure S4.) (B) The degree of mitochondrial fragmentation in the muscle cells of singled L1s arrested for 18 days predicted their ability to develop when fed (left panel). A score of 0, 1 or 2 represents no, intermediate or high fragmentation levels per cell, and was used to calculate a total fragmentation score per animal. Averages are shown (right panel). **p = 0.003 (see Fig. S1F–H). (C) The level of KIN-19::RFP aggregates in singled L1s arrested 10 days predicted their ability to develop when fed. Scale bar, 10 μm. ***p < 0.0001 (see also Fig. S4G).

Multiple age-related phenotypes are reversed when arrested L1s are fed

To investigate the fates of age-related phenotypes upon recovery, we examined arrested L1s before and after feeding. When fed, larvae that had been arrested for about 2 weeks resumed growth with a lag time ranging from one day to several days (data not shown and (Lee et al., 2012)). During that same period, prior to the entering the L2 larval stage, their movement improved (Figure 3A) and reactive oxygen species returned to a low level (Figure 3B). Likewise, about a day later, the fragmented mitochondrial morphology was replaced by a normal tubular network (Figures 3C, S5B), possibly via thin new tubular structures that appeared to sprout from round mitochondrial fragments (Figure S5C).

Figure 3. Feeding erased all signs of aging in arrested L1s, except for protein aggregates, which persisted.

(A) Movement speed was restored after feeding L1s arrested 18 days. ***p < 0.0001. L1*, some worms started to grow. (B) ROS-indicator DHE signal in the heads of singled 15-day arrested L1s, each examined before and after 2–3 days feeding. Quantification is shown on right panel. G, intestinal granules, these exhibited a ROS signal even in day-1 arrested L1s; the levels over time may increase but were not quantified. (C) Mitochondrial morphology in individual 18-day arrested L1s over time, before and after feeding (quantification in Fig. S5B). M: mitochondria, n: nucleus. Scale bar, 10 μm. Images are projections of 5–10 frames. (D) Viability of recovering larvae increased markedly after feeding. L1 populations arrested for 1 or 16 days were fed for 0, 24 or 48 h and their further development was blocked with the DNA synthesis inhibitor 5-Fluoro-2′-deoxyuridine (FUdR), in order to measure their survival at different larval stages. Upper panel: experimental timeline. Lower graphics: feeding increased the length of time long-arrested larvae can survive. Survival curves of 1-day L1s (left panel) and 16-day L1s (right panel) shifted to plates with food. FUdR was added at the time indicated (0h, 24h and 48h after feeding). n > 90 animals per curve, survival repeated once with similar findings (supporting material in Fig. S6, replicates in Table S1). (E) The number of polyQ35-YFP aggregates in 16-day arrested L1s remained the same after 2 days of feeding, and increased after 3 days of feeding, at the L4 stage. Each day assays a different fraction of the same population. Blue bar, average. ***p < 0.0001. (See also Fig. S5D, E for KIN-19::RFP aggregates.)

The normal adult longevity of previously long-arrested L1s (Figure 1B) suggested that either feeding per se, or, alternatively, a process linked to growth and development, improves their viability. The DNA synthesis inhibitor FUdR blocks development but not feeding (Figures S6A, B), allowing us to test these two possibilities. We allowed long-arrested L1s to feed for various times and assessed their ‘longevity’, measured as the length of time they remained alive after FUdR treatment (Figure 3D). L1 larvae arrested for 16 days had a high probability of dying when FUdR was added to block development concurrently with food. However, giving the animals some time to develop by waiting for 24 or 48 hours after feeding before adding FUdR markedly increased their chances of survival (Figure 3D). Moreover, the worms that had grown the most prior to FUdR treatment survived the longest (Figures 3D, S6C). Thus we concluded that a process coupled to growth, and not simply feeding, is required for improved viability.

Unexpectedly, one correlate of aging, the presence of protein aggregates, was not erased during recovery (Figures 3E, S5D, E). However, later in life, the abundance of KIN-19::RFP aggregates in the adults previously L1-arrested for a long period of time was similar to adults previously L1-arrested only 1 day (Figures S5F). Likely because the aggregates that form during L1 arrest are greatly outnumbered by the new aggregates that accumulate during development (Figures S5D, E). In contrast, animals previously L1-arrested for a long period start adulthood with more Poly-Q35::YFP aggregates than do animals L1-arrested only 1 day; and maintain this difference during aging (Figures S5G).

Notably, cell division was not required to erase age-related phenotypes. None of the new muscle cells generated during post-L1 larval development reside in the anterior body region of the animal (Sulston and Horvitz, 1977). However, by following individual animals, we observed the re-appearance of youthful, tubular mitochondrial networks in (non-dividing) anterior as well as posterior muscle cells (Figures 3C and S5B). Thus, many age-related phenotypes were reversed upon feeding and growth, in both dividing and non-dividing cells.

The ER-UPR gene ire-1 is necessary for the recovery of long-arrested L1s

We speculated that a genetic program might be activated specifically to restore proteostasis and erase aging-like phenotypes that accumulate during long periods of L1 arrest. In a pilot screen, we tested genes encoding chaperones or proteins known to be required for adult longevity (Table S3) by feeding arrested L1s bacteria expressing dsRNA. RNAi-bacteria that impaired the development of L1s arrested for long, but, importantly, not short periods of time were selected for further investigation (Table S3). The most effective hit was the endoplasmic reticulum (ER) chaperone gene hsp-4/BIP (Figures 4A, S7A, B). A subsequent screen of genes functionally related to hsp-4 yielded a second predicted ER chaperone (Kapulkin et al., 2005), stc-1 (Figure S7A). hsp-4/BIP is involved in the ER (UPR) protein response, which is known to be activated through the ER receptors ATF-6, PERK-1 and IRE-1 (Hetz, 2012). One of these, IRE-1, was completely required for the growth of long-arrested L1s but not for L1s arrested only 1 day (Figure 4B). The striking failure of ire(ok799) null mutants to resume development was not due to a failure to ingest bacteria because, like control long-arrested animals (Figure S4A), ire-1(−)-arrested animals ate normally (data not shown). Nor was failure to develop due to a more rapid loss of viability during the arrest itself, as wild-type and ire-1-mutant L1s lost viability at similar rates when food-deprived (Figures 4C, Table S2).

Figure 4. The ER-UPR gene ire-1 is required for arrested L1s to recover and regain youthful cellular phenotypes.

(A) The ER chaperone hsp-4/BIP was identified in a screen for RNAi clones that impaired the growth of long-arrested L1s, but not 1-day-arrested L1s, after feeding (see also Fig. S7A, B and Table S3). RNAi control, vector-only (L4440) bacteria. ***p < 0.0001 is the interaction p-value for the 2-way ANOVA test. (B) Effect of the three major ER-UPR pathways on the recovery of L1s arrested for 10 days. Average of three experiments (n > 100 in each). ***p < 0.0001. (C) Compared to wild-type, L1-arrested ire-1(ok799) mutants soon lost the ability to develop when fed, independently of their survival. n > 100 for each time point (Table S2). (D) The movement speed of ire-1(ok799) arrested L1s did not increase after feeding. Thrashing speed was measured in single 15-day arrested L1s (0 h) and after 12 h and 24 h feeding. (E) Representative mitochondrial morphology of an ire-1(ok799) L1 photographed after 14 days of arrest and again after 4 days on food. Scale bar, 10 μm. n, nucleus. M, mitochondria. (F) Longitudinal quantification of mitochondrial fragmentation shown in E. No reversal was observed.

Because long-arrested ire-1-mutant L1s did not recover when fed, we asked whether they also failed to erase markers of aging. We found that, unlike wild-type arrested L1s, ire-1(−)-arrested L1s did not increase their ability to move when fed (Figure 4D). Likewise, unlike wild-type (Figure S5B), individual arrested ire-1(−) L1s with fragmented mitochondria did not synthetize tubular networks after feeding but instead accumulated even more fragmented mitochondria (Figures 4E, F). To ask whether their high probability of dying could be reversed by the addition of food, we allowed long-arrested ire-1 mutants and wild-type controls to feed for 48 hours and then blocked further growth with FUdR. We found that allowing the animals to feed for 48 hours markedly extended the long-term survival of wild-type L1s but not that of ire-1(−) L1s (Figure S7C). Thus ire-1(+) was required to erase all of the aging-like phenotypes that we examined.

IRE-1 acts in part through XBP-1 to promote L1 recovery

Consistent with the importance of ire-1 in the ER-UPR, we found that during the first 12 hours with food, prior to growth, the long-arrested L1 population strongly increased its resistance to the ER-stressor DTT (Figure 5A) and, to a lesser extent, to heat (Figure 5B). This increased stress resistance could potentially be part of the recovery mechanism of these long-arrested animals, as it was quite specific: the stress resistance of newly-hatched L1s arrested only briefly (12 h) dropped after feeding [see day-1 L1s in Figures 5A, B and (Jobson et al., 2015)].

Figure 5. IRE-1’s ability to control L1 recovery partially requires XBP-1.

(A, B) The stress resistance of long-arrested L1s increased after feeding, prior to growth, particularly resistance to the ER stressor DTT. In contrast, the stress resistance of L1s starved for a short period of time, which is known to be increased relative to continuously-fed L1s, was reduced after feeding (Jobson et al., 2015). (A) DTT resistance (2.5h at 5mM) of L1s arrested for 1 day and 16 days and after 12h feeding. n > 100, 3 times per condition, ** p < 0.001, *** p < 0.0001. After 12 hours feeding day-1 L1s initiated growth, day-16 L1s did not. (B) Heat-shock resistance (6h at 35 °C) of L1s arrested for 1 and 16 days and then fed for 12 hours. *p = 0.029, ***p < 0.0001. 24 h after feeding, heat-shock resistance increased slightly for both 1-day L1s and 16-day L1s (not shown). (C) Phsp-4::gfp fluorescence level in L1s arrested 1 day, 10 days and 10 days-then fed 6 hours. The basal level of expression of Phsp-4::gfp decreases during the arrest (as with aging adults, data not shown) and goes back up during the recovery of long-term arrested L1s (in the intestine only). Arrows: auto-fluorescence of intestinal granules. (D) An xbp-1 null mutation had a significant but relatively mild effect on the recovery of L1s arrested 10 days compared to that of ire-1(−) mutants. n > 100 in each of three experiments (average shown here). ***p < 0.0001, *p = 0.014. (These xbp-1(−) animals were examined in parallel with the strains shown in Fig. 4B; the same ire-1 data are shown again here for clarity.) (E) L1-arrested mutants carrying ire-1(zc14), a partial loss-of-function mutation that strongly reduces hsp-4 induction in response to tunicamycin (Calfon et al., 2002), are only partially defective in recovery. n > 100 for each time point (Table S2). (F) Treatment of wild-type day-10 L1s with 150 μM of the IRE-1 RNase inhibitor 4μ8C during the arrest and after feeding did not affect recovery. Development (body size) was assessed after 4 days of feeding. n=50 in each of three trials. (G) RT-PCR showing the splicing of xbp-1 mRNA after tunicamycin treatment (TM, 2 μg/mL). The splicing was inhibited in the presence of 150 μM 4μ8C inhibitor. 2 independent samples are shown for each condition. PCR control: unc-15 primers. (H) Phsp-4::gfp expression levels measured by microscopy in the head of day-1 L1s increased after treatment with 2 μg/mL tunicamycin. The response was inhibited by 150 μM 4μ8C, as was (I) the ability of worms to develop to adulthood. ***p < 0.0001. HS: heat-shock.

ER stress activates IRE-1’s RNase activity. This leads to the splicing and translation of xbp-1 mRNA, which in turn encodes a key ER-UPR transcription factor (Hetz, 2012). We found that the hsp-4/BIP ER-chaperone gene, a target of XBP-1, was up-regulated during larval recovery (Figure 5C), and that deleting hsp-4/BIP slightly delayed the development of long-starved L1s (Figures S7B, D). In addition, the putative xbp-1 null mutation xbp-1(zc12) reduced the percentage of long-starved L1s that were able to recover (Figure 5D), and extended the time that it took growth to begin (Figure S7D). However unexpectedly, inactivating xbp-1 (or hsp-4) by mutation had a smaller effect on recovery and viability than did ire-1 mutation (Figures 5D and S7E, F). A kinase-domain point mutation, ire-1(zc14), impairs the response to ER stress and fails to induce hsp-4 (Calfon et al., 2002). Interestingly, when compared to the ire-1 null mutant, this mutant still maintained half of its recovery potential after long periods of arrest (Figure 5E). This observation shows that UPR induction (which is largely defective in this mutant) and L1 arrest recovery can be partially uncoupled. Likewise, treating the animals with 4μ8c, an inhibitor that selectively targets the IRE-1 RNase domain (Cross et al., 2012), either just before feeding or during the entire course of larval arrest, did not inhibit larval recovery (Figure 5F). [We confirmed that 4μ8c inhibited xbp-1 mRNA splicing, prevented Phsp-4::gfp up-regulation, and reduced larval survival (Figures 5G–I).]

Together these findings indicate that the canonical IRE-1/XBP-1 branch of the UPR plays only a partial role in the reversal of aging phenotypes; thus, a different ire-1-dependent pathway participates as well.

The ire-1-regulated MAP kinase KGB-1 is partially required for recovery of long-arrested L1s

Two studies in cultured mammalian cells have shown that IRE-1 can phosphorylate and activate c-JUN mitogen-activated protein kinase (JNK1) (Ogata et al., 2006; Urano et al., 2000). Thus, we tested L1 recovery efficiency in mutants of the closest JNK1 homologs in C. elegans. We found that kgb-1(−) and pmk-1(−) mutations significantly reduced the recovery of 10-day arrested L1s (Figure S7G). pmk-1 mutants exhibited very poor survival during the L1 arrest itself, complicating the interpretation (Figure S7H). In contrast, the kgb-1(−) mutant survived short L1 arrest as well as wild-type, but nevertheless exhibited an early loss of recovery potential when arrested for long periods of time (Figure 6A, Table S2). JUN kinases, including KGB-1, are regulated by upstream kinases, so we analyzed the level of phosphorylated KGB-1 during the L1 arrest and after recovery using Western blot analysis (Figure 6B). In the wild-type, KGB-1 was phosphorylated throughout the period of L1 arrest and upon subsequent feeding. This finding contrasts with the situation in adults, in which KGB-1 is phosphorylated only during starvation (Uno et al., 2013). Interestingly, KGB-1 phosphorylation levels were significantly decreased after 10 days of L1 arrest in an ire-1(−) background. Because ire-1(−) L1s arrested for 10 days are not able to recover, this finding suggests that one mechanism by which wild-type IRE-1 promotes recovery is by maintaining, directly or indirectly, the phosphorylation of KGB-1.

Figure 6. The Jun-N-terminal kinase KGB-1 regulates L1 recovery after prolonged arrest and is controlled by ire-1.

(A) kgb-1(um3) null mutants had an impaired ability to recover from L1 arrest (less than half of WT, see Table S2). Like ire-1 mutants, their survival time was comparable to wild-type. n > 100 for each time point. (B) KGB-1 phosphorylation was decreased in ire-1(ok799) mutants following a 10-day arrest. Protein extracts of L1s from wild-type and ire-1(ok799) mutants were prepared following either a brief or long L1 arrest before and after feeding. Pictures show a representative western blot of the same membrane probed with antibodies targeting KGB-1, phospho-KGB-1 or actin. (C) kgb-1 is required for previously long-arrested L1s, but not briefly-arrested L1s to subsequently have normal adult lifespans. Lifespan of WT and kgb-1(um3) adults previously L1-arrested for, respectively, 1 and 23 days or 1 and 16 days. n (blue) = 48 and n (red) = 62 ; p = 0.34. n (orange) = 67 and n (green) = 99 ; p <0.0001 (see Table S1). (D) Muscle cells of kgb-1(um3) pre-adults (L4s) previously arrested 15 days as L1s had higher levels of mitochondrial fragmentation relative to those of kgb-1(um3) L4s arrested 1 day. (E) Representative images of L4 worms from L1s arrested 1 day or 15 days showing fragmented mitochondria. Scale bar, 10 μm. n, nucleus. M, mitochondria. Scores “A–D” refer to the degree of mitochondrial fragmentation (see also Figure S1C, D). (F) Each cell was attributed a mitochondrial fragmentation score A to D [like on panel D, described in Fig. S1C, D)] and a score of 0, 1, 2 or 3 was respectively associated to it to calculate a fragmentation score per animal (details are in Methods). Averages are shown. **p < 0.001.

Because the recovery defect of kgb-1(−) mutants was not as dramatic as that of ire-1(−) mutants, we were able to ask whether lifespans of kgb-1(−) adults that had been arrested for prolonged periods as L1s but did recover might be shorter than wild-type controls. Remarkably, the adult lifespans of kgb-1(−) mutants arrested for 16 days as L1s could be up to 50% shorter than were those of kgb-1(−) mutants arrested for only 1 day (Figure 6C, S7I, Table S1). We then questioned whether the reduced lifespans of post-starved kgb-1(−) adults correlated with the persistence of cellular phenotype of aging, mitochondrial fission. We found that the degree of fission was abnormally high at the L4 (pre-adult) stage of kgb-1(−) mutants compared to that of control wild-type animals (Figures 6D–F and S1C, D). Thus, the KGB-1 kinase is required for normal cellular restoration of L1s starved for a long period of time, which in turn likely ensures their normal longevity.

Finally, we constructed the xbp-1(−); kgb-1(−) double null mutant to determine whether its phenotype might recapitulate that of the much stronger ire-1 mutant. We found that this was not the case; the recovery in the double mutant remained more efficient than in the ire-1(−) mutant (Figure S7J, Table S2). Thus ire-1 likely acts through additional, yet unknown, effectors to promote recovery.

Discussion

Phenotypes reminiscent of aging in arrested C. elegans larvae

Because long-arrested L1s, when fed, go on to have normal adult lifespans (Johnson et al., 1984) (Figure 1B), food limitation has long been thought to “pause the aging clock”, arresting growth but maintaining viability until energy reserves are depleted and the animals die. However, mutations that increase adult lifespan also increase the lifespans of quiescent L1s. In addition, mutations selected to increase survival of arrested L1s have a high probability (80%) of extending adult lifespan as well (Muñoz and Riddle, 2003). These findings suggested to us that a failure of homeostasis and proteostasis maintenance, which is thought to promote aging, might also drive the decline and death of arrested L1s.

Our data show that like aging adults, quiescent L1s progressively suffer a decline in motility. At the cellular level, their mitochondrial networks fragment and their ROS levels increase, as with aging adults. In addition, over time, the heat- and DTT-resistance of arrested L1 larvae declines, and protein aggregation and detergent-insolubility increase. The heat-shock transcription factor HSF-1 a key regulator of proteostasis and longevity (Hsu et al., 2003), appears to be required for proper survival and recovery of L1s as well. Together these observations support the hypothesis that quiescent L1s undergo a progressive decline in proteostasis, which is a hallmark of adult aging (Labbadia and Morimoto, 2014). Interestingly, we found that two measures of aging-like stress (mitochondrial fission and protein aggregation), but not residual fat storage, predicts recovery. Thus, age-related cellular decline rather than stored energy depletion might be responsible for the loss of recovery of quiescent L1s (for more details see Supplemental Methods). In the future, it will be interesting to ask whether levels of other types of energy stores, or metabolites linked to energy charge, such as ATP and NAD, predict which animals will recover and thus could be in the causal pathway.

Age-related phenotypes that arise during L1 quiescence can be erased

When long-arrested L1s were fed, remarkably, signs of aging were erased, as if a form of somatic restoration was occurring. The animals regained most of their youthful behaviors and cellular morphologies. Upon feeding, mitochondrial morphology, ROS levels, movement speed and viability were all returned to levels comparable to ‘younger’ L1s, starved only one day.

Protein aggregation is associated with many age-related diseases (Powers et al., 2009), and, in addition, many endogenous proteins aggregate with age (David et al., 2010; Demontis and Perrimon, 2010; Peters et al., 2012; Walther et al., 2015). In our study, we found that the endogenous kinase KIN-19 (fused to tagRFP), and the exogenous polyglutamine repeat protein Poly-Q35 both formed aggregates during L1 arrest just as they do during adult aging (David et al., 2010; Morley et al., 2002), and, more generally, that the detergent-insoluble protein fraction increases in long-arrested L1s. We also observed that the extent of KIN-19::RFP aggregation was inversely correlated with recovery efficiency, suggesting that phenomena linked with aggregation might inhibit recovery. Unexpectedly, we found that protein aggregates that accumulated during L1 arrest were not removed upon feeding. Whether protein aggregates promote aging (for example, by interfering with key cellular processes); or whether they may actually protect against aging is not clear (David et al., 2010; Reis-Rodrigues et al., 2012; Walther et al., 2015). Our findings are most simply explained by the idea that the conditions that promote aggregation in starved L1s are toxic, but that the aggregates themselves, which persist during recovery, are not.

ire-1 controls L1 recovery through xbp-1-dependent and xbp-1-independent mechanisms

In our search for genes regulating L1 recovery, we identified the ER membrane receptor IRE-1 as being necessary for long-arrested L1s to develop to adulthood when fed. IRE-1 was also required to reverse the age-related phenotypes we examined; namely, mitochondrial fission, slow movement and high mortality. It is important to note that this mutant did not show decreased survival during the L1 arrest or pumping defects when subsequently fed, suggesting that the inability of these animals to develop to adulthood was not a consequence of poor viability or incapacity to feed. Instead, IRE-1 appears to be part of a regulatory switch that specifically restores L1 larvae. ER-stress resistance increases dramatically when long-starved (but not briefly-starved) wild-type L1s are placed on food, suggesting that IRE-1 may promote recovery by mechanisms that specifically restore ER homeostasis. However, the specific nature of IRE-1 activation and function during L1 recovery remains to be determined.

We found that inactivating the UPR transcription factor XBP-1 by mutation (Calfon et al., 2002), or inhibiting the RNase domain of IRE-1 responsible for xbp-1 mRNA splicing with a drug had, respectively, a lower, or no impact on L1 recovery. We concluded that the canonical UPR pathway downstream of IRE-1 was only partially required for the recovery. In addition, we could exclude a role in recovery for IRE1-dependent degradation of mRNAs encoding ER proteins (RIDD), which also requires the IRE-1 RNase (Hollien and Weissman, 2006). Together these findings indicate that a different ire-1-dependent pathway is involved. It is possible that a similar pathway could counter adult aging, as is suggested by our previous finding that a non-canonical ire-1 pathway contributes substantially to the increased adult longevity of the daf-2 insulin/IGF-1-receptor mutant (Henis-Korenblit et al., 2010).

KGB-1, a MAP kinase regulated by IRE-1, promotes L1 recovery

We speculated that upon feeding, IRE-1 could transmit a recovery signal through its kinase domain. After testing several candidate MAP kinases, we found that a kgb-1 null mutant partially mimicked the ire-1 phenotype for the premature loss of larval recovery. Interestingly, this mutation also prevented the restoration of tubular mitochondrial morphology to a normal level during development, arguing for a role in re-establishing cellular homeostasis. Our data do not explain the mechanism that links ire-1, kgb-1 and mitochondrial fusion. However, western blotting showed that KGB-1 phosphorylation was lost in the ire-1 null mutant after a long period of arrest. Thus, IRE-1 seems likely to act upstream of KGB-1 activity to promote recovery of arrested L1s, though whether it phosphorylates KGB-1 directly is not known.

kgb-1 is known to make a substantial contribution to the lifespan extension obtained after intermittent food restriction in adult worms (Uno et al., 2013), again suggesting that common mechanisms influence adult longevity after fasting and L1 recovery. In this study, we found that kgb-1 is required for the normal lifespan of adults previously arrested for long (but, importantly, not short) periods of time as newly-hatched L1s. This result, that the adult has a memory of events that happened when it was significantly much smaller [1/150 of adult volume (Watanabe et al., 2005)], is consistent with the interpretation that a process regulated in part by the IRE-1/KGB-1 axis is turned on by feeding, allowing the removal of defects accumulated during prolonged quiescence of L1 larvae. It also raises the possibility that the failure to restore mitochondrial homeostasis at a young age has lasting consequences that trigger premature adult aging at the end of life. How might KGB-1 enhance L1 recovery? The downstream effectors of KGB-1 in this process are unknown; however in fasting adults, kgb-1 was shown to promote the expression of genes that regulate proteostasis, including oxidoreductases and xenobiotic-response genes (Uno et al., 2013). In animals subjected to hyperosmotic stress, KGB-1 promoted expression of genes encoding chaperones as well as genes that mediate biosynthesis (Gerke et al., 2014). Thus KGB-1 could potentially enhance a variety of molecular processes that reestablish proteostasis in recovering L1s, including de novo protein synthesis.

A potential for the reversal of aging phenotypes at other times of life

If instead of starving, C. elegans L1s are given low levels of food, they develop into morphologically specialized, resilient L3s called “dauers”. Animals held in dauer for long periods experience changes in the levels of certain metabolites that are similar to those seen in old adults, and this decline is reversed when dauers resume development (Houthoofd et al., 2002). Moreover, like long-arrested L1s (Johnson et al., 1984), these dauers become relatively infertile adults (Kim and Paik, 2008). This finding makes one wonder whether the aging-like features that we see in long-arrested L1s might also arise in long-arrested dauers, and if so, whether a single restorative mechanism may able to reverse them at different times in life. A previous study showed that young adults, too, can undergo long periods of quiescence in absence of food and recover from it with a normal lifespan for their remaining life time (Angelo and Van Gilst, 2009). It raises the possibility that the mechanisms to restore somatic tissues could also be activated later in young adults.

Conclusions

To the best of our knowledge, such an apparent reversal of age-related phenotypes on the scale of a whole animal, occurring naturally in a multicellular species, has never been demonstrated. Many aspects of C. elegans biology are conserved evolutionarily, and similar phenomena could potentially exist in other species. For example, nutrient limitation arrests growth and cell division in yeast and induces over time many signs of aging (Longo et al., 2012; Roux et al., 2010). It would be interesting to investigate whether re-entry into the cell cycle is linked to an IRE1-dependent reversal of aging phenotypes. Notably, subjecting old mice to repeated cycles of fasting and re-feeding can reverse age-dependent changes in hematopoietic stem cell lineage composition and white blood cell number (Cheng et al., 2014). This apparent rejuvenation was dependent on reduced activities of IGF1 and PKA, two conditions linked to increased longevity. Likewise, providing young blood or serum to an old mouse stimulates neurogenesis in the brain and “rejuvenates” muscle stem cells (Conboy et al., 2003, 2005; Villeda et al., 2014, 2011). It can also reverse age-related hypertrophy of non-dividing heart muscle cells (Loffredo et al., 2013). However, the molecular mechanisms that underlie these observations are poorly understood. We hope that the understanding of how age-related phenotypes can be reversed in somatic tissues in C. elegans, a highly tractable experimental organism with conserved mechanisms for longevity maintenance, will ultimately suggest new ways to maintain vitality and improve the quality of human life during aging.

Experimental Procedures

L1 isolation, survival and recovery assays

To induce L1 larval arrest, C. elegans eggs were isolated from a well-fed adult population, washed to eliminate any remaining E. coli OP50, and incubated overnight at 20°C for hatching in 5-10 mL S-basal with 5 μg/ml cholesterol, 0.01 % PEG and 0.1 μ/mL amphotericin B. The synchronized larvae obtained by this protocol were considered 1-day arrested L1s and were kept at 20°C for the time indicated with slow rotation. During the L1 arrest, survival was assayed every few days as the percent of worms moving and reacting to gentle prodding in a drop of solution on plate. To induce re-entry into development, arrested L1 larvae were pipetted from liquid culture to petri plates with food. The percent developing (also referred to as “recovering”), was measured as the percent of worms progressing to stage L3-L4 after feeding. Approximately 120 L1 larvae were tested for each time point.

Quantification of age-related phenotypes and their reversal by microscopy

Age-related phenotypes are described individually in the figure legends and in Supplemental Methods. Mitochondrial morphology, ROS, KIN-19::tagRFP aggregates, Poly-Q35::YFP aggregates, autofluorescence of intestinal granules and muscle fiber morphology were observed by confocal fluorescence microscopy. First, briefly- and long-arrested L1 larvae were mounted between two glass slides and imaged by microscopy. For studying recovery within a large population, an aliquot of L1s was analyzed by microscopy before inducing development by feeding, and the imaging was then applied to another aliquot from the same tube of arrested L1s, but this time after recovery. For the longitudinal studies of recovery in single animals, L1s were individually recovered from the glass slides after the first measurement, transferred to a plate with food to induce development, and observed again after 2–4 days.

All microscopy experiments were performed on living worms using spinning-disc confocal fluorescence microscopy and non-invasive immobilization with 0.5 μL polystyrene microsphere solution on fresh 1 mm-thick 4% agarose pads. Mitochondrial morphology and Poly-Q35::YFP aggregate number were scored visually in a blinded manner. KIN-19::tagRFP aggregates, ROS and auto-fluorescence were scored using image quantification software.

Larval survival on food using FUdR to block development

Arrested L1s were transferred on a petri dish and counted before E. coli OP50 was added to the medium. FUdR (5-Fluoro-2′-deoxyuridine) at 90 μM final concentration was then added 0, 24, or 48 hours after feeding to block further development of the larvae. We verified that FUdR prevented growth but not feeding. For the survival assay, scoring started 5 days after reintroduction of E. coli OP50, after censoring L1s that died earlier during the arrest. Animals were scored as dead when they failed to move after gentle prodding. The survival of the remaining larvae was scored every 2 to 3 days. For each survival curve, at least 100 worms were used, and the experiments were repeated at least twice independently. The size of the animals was measured 5 days after feeding to control the efficiency of FUdR, and after death during the survival analysis.

RNA interference screen

To screen for genes required for recovery, wild-type L1s arrested for 1, 13 or 20 days were fed dsRNA-expressing bacteria, and their development after feeding was analyzed. In total, RNAi clones for 71 genes were tested. RNAi clones that specifically prevented recovery of 13-day and 20-day arrested L1s were selected for further validation (see Table S3). Surprisingly, ire-1 and xbp-1 were not identified in the RNAi screen (where the XBP-1 target hsp-4 was identified), but were shown to be involved in recovery in follow-up experiments using ire-1(−) and xbp-1(−) mutants. An independent RNAi experiment targeting gfp indicated that the RNA interference machinery only became effective 2 days after feeding (not shown). This may explain why the upstream, early regulators of the UPR pathway were not present in our RNAi screen hits (Table S3). The hsp-4 gene, which is regulated by the transcription factor XBP-1, was found by RNAi knockdown likely because it normally acts after 48 h of feeding, as shown in Figure S7B.

Western Blot

Proteins were extracted from worms by sonication with a Diagenode Bioruptor. The western blot was carried out as described in (Mizuno et al., 2008). Also see Supplemental Methods.

Statistics

For lifespan analysis, statistics were performed using STATA/IC 10.0. p values were calculated with the log rank/Mantel Cox test. Other statistics were performed using GraphPad Prism 6. If not specifically mentioned, p values were calculated by the unpaired two-tailed student’s t-test. The R² value describes the Pearson correlation coefficient and p is the probability of obtaining this value by chance. To measure statistical significance between 2 differences, the interaction between variables was tested with a 2-way ANOVA test (measures compiled as not matching). The corresponding p value then describes whether the interaction between the effect of RNAi and the effect of age (or mutation and time on food before FUdR) was statistically significant. In the figures, the error bars represent standard deviations.