mTORC2-SGK-1 acts in two environmentally responsive pathways with opposing effects on longevity

The nematode worm Caenorhabditis elegans provides a powerful system for elucidating how genetic, metabolic, nutritional, and environmental factors influence aging. The mechanistic target of rapamycin (mTOR) kinase is important in growth, disease, and aging and is present in the mTORC1 and mTORC2 complexes. In diverse eukaryotes, lifespan can be increased by inhibition of mTORC1, which transduces anabolic signals to stimulate protein synthesis and inhibit autophagy. Less is understood about mTORC2, which affects C. elegans lifespan in a complex manner that is influenced by the bacterial food source. mTORC2 regulates C. elegans growth, reproduction, and lipid metabolism by activating the SGK-1 kinase, but current data on SGK-1 and lifespan seem to be conflicting. Here, by analyzing the mTORC2 component Rictor (RICT-1), we show that mTORC2 modulates longevity by activating SGK-1 in two pathways that affect lifespan oppositely. RICT-1/mTORC2 limits longevity by directing SGK-1 to inhibit the stress-response transcription factor SKN-1/Nrf in the intestine. Signals produced by the bacterial food source determine how this pathway affects SKN-1 and lifespan. In addition, RICT-1/mTORC2 functions in neurons in an SGK-1-mediated pathway that increases lifespan at lower temperatures. RICT-1/mTORC2 and SGK-1 therefore oppose or accelerate aging depending upon the context in which they are active. Our findings reconcile data on SGK-1 and aging, show that the bacterial microenvironment influences SKN-1/Nrf, mTORC2 functions, and aging, and identify two longevity-related mTORC2 functions that involve SGK-regulated responses to environmental cues.

Work in C. elegans on the Rictor ortholog rict-1 has begun to elucidate processes that are regulated by mTORC2. rict-1 loss-offunction mutants exhibit developmental abnormalities (slow growth, small body size, impaired reproduction, and increased fat storage), and their lifespan is markedly affected by the type of bacterial food they consume (Jones et al., 2009;Soukas et al., 2009). Individual bacterial strains vary in how well they support C. elegans growth and reproduction, a parameter that is not understood but has been defined as food 'quality' (Avery & Shtonda, 2003;Shtonda & Avery, 2006). In wild-type (WT) worms, lifespan is increased modestly by feeding of higher-quality food such as the Escherichia coli strain HT115 (Soukas et al., 2009;Maier et al., 2010). This difference is magnified in rict-1 mutants, which live slightly longer than WT when fed HT115 but have a markedly reduced lifespan on the standard laboratory strain OP50 (Soukas et al., 2009). An understanding of this food source effect may reveal a paradigm for how an organism responds to differences in nutrient intake or to signals from the bacteria that are present in its body space (microbiome). The second possibility would be of particular interest, given that the gut microbiome profoundly influences human metabolism and immunity (Tremaroli & B€ ackhed, 2012).
To address this problem, we have investigated how rict-1 and sgk-1 affect SKN-1 activity, stress resistance, and lifespan under different food source and temperature conditions. Our findings reconcile the data on sgk-1 and longevity, by showing that mTORC2 and SGK-1 act in two pathways that have opposing effects on lifespan. In the first pathway, mTORC2 directs SGK-1 to inhibit SKN-1. Depending upon cues from the bacterial microbiome, release of this inhibition may allow SKN-1 to increase stress resistance and lifespan. In the second pathway, mTORC2 activity in neurons is required for SGK-1 to promote longevity at lower temperatures. mTORC2 and SGK-1 therefore determine how mechanisms that increase lifespan are modulated by two distinct environmental inputs, bacterially produced signals and temperature.

Results and discussion
Food source, temperature, and SKN-1 determine how RICT-1/ mTORC2 affects stress resistance and lifespan SKN-1 accumulates in intestinal nuclei in response to various stresses (An & Blackwell, 2003), but does so constitutively when rict-1 mutants are fed the 'quality' strain HT115 (Robida-Stubbs et al., 2012). This does not occur with OP50 feeding, however, suggesting that loss of RICT-1/ mTORC2 function activates SKN-1 only when animals are maintained on quality food. To test this idea further, we asked whether loss of rict-1 increases stress resistance in a skn-1-dependent manner in animals that are fed HT115 and OP50 as typical examples of quality and standard food, respectively (Avery & Shtonda, 2003;Shtonda & Avery, 2006). Exposure to rict-1 feeding RNAi (in HT115) increased resistance to oxidative stress from sodium arsenite (As) in WT animals, but not skn-1 mutants ( Fig. 1A and Table S1, Supporting information). A probable null rict-1 mutant was also more resistant to As treatment than WT, but only when propagated on HT115 rather than OP50 (Fig. 1B,C, Table S2, Supporting information). Again, this increased stress resistance required skn-1. rict-1 mutation also conferred resistance to tert-butyl hydrogen peroxide (TBHP) in a manner that was largely dependent upon skn-1 and HT115 feeding (Fig. 1D,E, Table S3, Supporting information).
We next investigated the involvement of SKN-1 in the effects of food source on rict-1 lifespan, which had been examined only at 25°C (Soukas et al., 2009). Our experiments at 25°C reproduced these earlier findings: rict-1 animals fed HT115 lived slightly longer than WT, but on OP50, their lifespan was significantly reduced ( Fig. 2A,B, Table S4, Supporting information). By contrast, when skn-1 was inactivated, the rict-1 mutation failed to increase lifespan in animals that were fed HT115 (compare rict-1; skn-1 double mutants to skn-1 mutants, Fig. 2B and Table S4, Supporting information). Moreover, a comparison of Fig. 2A,B reveals that rict-1 but not rict-1; skn-1 mutants lived far longer when fed HT115 instead of OP50 (+60% vs. +15% mean lifespan). Both WT and skn-1 mutant animals lived slightly longer on 'quality' food (+10% and +14%, respectively), but skn-1 was required for the dramatically increased benefits of quality food enjoyed by rict-1 mutants. This scenario in which quality food increases SKN-1 activity seems to be unique. For example, SKN-1 accumulation in intestinal nuclei and skn-1dependent oxidative stress resistance were dramatically increased in daf-2 (insulin/IGF-1 receptor) mutants that were propagated on OP50 (Tullet et al., 2008).
Given that C. elegans and other ectotherms live longer at lower temperature (Lee & Kenyon, 2009;Xiao et al., 2013), we analyzed rict-1 lifespan at 20°C. In contrast to the lifespan increase seen with rict-1 RNAi (Robida-Stubbs et al., 2012), at 20°C, rict-1 mutation dramatically shortened lifespan compared with WT on either food source (Fig. 2C,D, Table S5, Supporting information). rict-1 mutants lived longer on HT115 than OP50, but the absolute and relative differences were less pronounced at 20°C than at 25°C (15.8 days on HT115 vs. 12.5 days on OP50 at 20°C (+26%), compared with 14.2 days on HT115 vs. 8.9 days on OP50 at 25°C (+59%); Fig. 2). Similar results were obtained at 15°C (Fig. 2E,F, Table S6, Supporting information). Thus, while WT animals lived considerably longer at lower temperatures when raised on either food, rict-1 mutants benefitted much less from the reductions in temperature. Although reproduction is impaired in rict-1 mutants (Jones et al., 2009;Soukas et al., 2009), the effects of rict-1 on lifespan do not derive from a pathway in which lifespan is extended by a reduction in germline stem cell (GSC) number (Kenyon, 2010). GSC ablation extends lifespan at either 20°C or 25°C (Kenyon, 2010), in contrast to rict-1 mutation ( Fig. 2A-D), and daf-16 is required for lifespan extension from GSC loss (Kenyon, 2010), but not rict-1 RNAi (Robida-Stubbs et al., 2012). Moreover, rict-1 RNAi increased lifespan in a long-lived mutant that lacks GSCs (Fig. S1, Table S18, Supporting information). Our data show that the impact of rict-1 on longevity is profoundly influenced by temperature as well as food source and by whether rict-1 function is impaired by mutation or RNAi.
RICT-1/mTORC2 modulates SKN-1 activity, stress resistance, and lifespan through SGK-1 lifespan are beset with apparent contradictions (see Introduction). Having observed that rict-1 effects on lifespan are profoundly influenced by the bacterial food source, temperature, and the mode by which rict-1 function is disrupted (mutation or RNAi), we incorporated these variables into an analysis of how sgk-1 influences stress resistance and lifespan.

Survival
Days Days

Parallel regulation of SKN-1 by RICT-1/SGK-1 and bacterial signals
Our results raise the question of how the effects of RICT-1/mTORC2 on SKN-1 regulation and lifespan might be influenced by the bacterial food source. rict-1 mutants that were raised on the quality strain HB101 on plates were previously found to spend a greater percentage of time away from the bacterial lawn than WT animals (10.8% percent vs. 3.1% percent as day-one adults) (Soukas et al., 2009). This suggested that when rict-1 mutants are provided with quality food, they live longer because they then consume less food and experience a DR-like state (Soukas et al., 2009). However, we have obtained various lines of evidence suggesting that propagation on quality food may influence rict-1 lifespan primarily through a mechanism other than DR. Under our conditions, in a microbial avoidance assay (Melo & Ruvkun, 2012), rict-1 mutants spent a similar proportion of time away from either an HT115 or OP50 lawn (Fig. S5, Supporting information). This raised the possibility that rict-1 mutants might consume a reduced amount of food in either case, but be unable to respond to DR when fed OP50. However, we observed that rict-1 mutants provided with OP50 lived much longer when subjected to DR in a liquid-feeding regimen, arguing against this idea (Fig. 5A,B, Tables S10 and S11, Supporting information). Moreover, the reduction in food availability needed for robust DR lifespan extension (≥6-fold, Fig. 5A,B) was greater than would be expected to arise from the proportion of time that rict-1 mutants spent off their food in plate culture. Unexpectedly, in liquid cultures at 20°C, rict-1 animals lived nearly as long as WT when fed either HT115 or OP50 ad libitum ( Fig. 5C and Table S12, Supporting information), in striking contrast to results obtained on plates at this temperature (Fig. 2C,D). This last result suggests that when rict-1 mutants are maintained in liquid, neither temperature nor food source influences their lifespan to the dramatic extent seen in plate culture. Together, our data suggest that the effects of quality food on SKN-1 activity and lifespan in rict-1 mutants are less likely to derive from differences in nutrient intake than from signals produced by 'quality' bacteria and that these signals are perceived on plates but not in liquid. Propagation on quality food was required for either rict-1 or sgk-1 mutation to promote SKN-1 nuclear accumulation and increase oxidative stress resistance (Fig. 1B-E, 3B, and Fig. S2D, Supporting information) (Robida-Stubbs et al., 2012), suggesting that signals from the bacterial food source influence SKN-1 at a step downstream of or in parallel to SGK-1 (Fig. 6A). Consistent with this notion, propagation on HT115 did not reduce SGK-1 phosphorylation of SKN-1 in either WT or rict-1 animals ( Fig. 4F (lanes 5 and 6), 4G, and Fig. S4, Supporting information). Perhaps signals produced by quality food lower a threshold for SKN-1 activation, so that when either rict-1 or sgk-1 function is lost under these conditions, SKN-1 accumulates in nuclei and increases stress resistance and lifespan (Fig. 6A).
Recent findings indicate that bacterial metabolites of L-Tryptophan (L-Trp) mediate effects of quality food on C. elegans reproduction (Gracida & Eckmann, 2013). Animals that lack NHR-114, a homolog of the hepatic transcription factor HNF4, were sterile when fed OP50 but not HT115. Remarkably, supplementation of OP50 with L-Trp rescued this phenotype. Rescue was mediated by L-Trp metabolites, which induced intestinal expression of enzymes that detoxify xenobiotics and other small molecules and are otherwise nhr-114-dependent. Apparently, quality food produces chemical signals that direct C. elegans to increase detoxification enzyme expression (Gracida & Eckmann, 2013).
SKN-1 regulates many of the same detoxification gene classes that were induced by L-Trp supplementation (Oliveira et al., 2009), suggesting that it might mediate this effect of L-Trp metabolites. Accordingly, L-Trp supplementation of OP50 increased SKN-1 nuclear accumulation and activity (Fig. S6A,B, Supporting information). This effect was greatly magnified in rict-1 mutants, in which L-Trp supplementation of OP50 activated SKN-1 even more dramatically than HT115 feeding (Fig. S6C, D, Supporting information). Like HT115 feeding, L-Trp supplementation did not alter SGK-1 activity (Fig. 4F (lanes 3 and 4), 4G, and Fig. S4, Supporting information). However, L-Trp supplementation decreased the lifespan of either WT or rict-1 animals ( Fig. S7 and Table S20, Supporting information), in contrast to maintenance on HT115. Our data suggest that L-Trp metabolites induce a SKN-1-mediated detoxification response that might mediate rescue of germ cell production in OP50-fed nhr-114 mutants, but also that L-Trp supplementation alone does not recapitulate the benefits of HT115 feeding for longevity. Perhaps quality food generates additional signals that activate SKN-1 in rict-1 and sgk-1 mutants and mediate its effects on lifespan, but are less toxic long term than L-Trp metabolites.

Dual longevity functions of RICT-1/mTORC2 and SGK-1
The above models explain many of our findings, but not the profound effect of temperature on lifespan seen in rict-1 and sgk-1 mutants (Fig. 2). For example, on OP50 WT, animals lived almost twice as long at 20°C compared with 25°C (20.7 days vs. 11.5 days), but rict-1 mutants did not benefit nearly as much from the decrease in temperature (12.5 days vs. 8.9 days)( Fig. 2A,C). Similar effects were seen in sgk-1 mutants (Fig. 2C-F, 4D,E). The previous finding that SGK-1  ). A 600 = 3 is designated as ad libitum feeding. (C) rict-1(mg451) lifespan is not increased by HT115 vs. OP50 feeding in liquid culture (Student's ttest P < 0.0001). Ad libitum feeding at 20°C is shown. Note that under these conditions, rict-1(mg451) lifespan is only modestly reduced compared with wild-type (WT). Statistical analysis is presented in Tables S10 (A, B), S11 (A, B), and S12 (C) (Supporting information).   Fig. 6 RICT-1/mTORC2 and SGK-1 act in two pathways that affect lifespan oppositely. (A) A working model for how RICT-1/mTORC2 and SGK-1 influence Caenorhabditis elegans longevity. RICT-1/mTORC2 directs SGK-1 to activate DAF-16 in the TRPA-1 temperature-sensing pathway and inhibit SKN-1 in an interaction that is modulated by bacterially produced signals. These signals regulate SKN-1 downstream or in parallel to its phosphorylation by SGK-1. A single interconnected pathway is arbitrarily shown, but while mTORC2 and SGK-1 inhibit SKN-1 in the intestine, neuronal mTORC2/RICT-1 is required for longevity mediated by the TRPA-1 pathway. See text for details. (B, C) The sgk-1(gf) mutation blocked lifespan extension from adulthood rict-1 RNAi. (D, E) sgk-1(gf) dramatically reduced the lifespan extension that rict-1 mutants experienced when fed HT115 instead of OP50. (F, G) sgk-1(gf) failed to increase lifespan in rict-1(mg451) mutants, which are short-lived at 20°C. (H, I) Effects of rict-1 and sgk-1 RNAi on lifespan in sgk-1(gf) mutants. Compare to (G) to observe the dramatic difference between effects of rict-1 mutation and RNAi, and note that sgk-1 RNAi did not interfere with the lifespan increase in sgk-1(gf) relative to wild-type (WT) (H, I). (J) Lifespan is reduced by neuronal rict-1 RNAi. RNAi was performed in a strain in which the dsRNAi transporter mutation sid-1 is rescued specifically in neurons (Calixto et al., 2010). A composite of three replicates is shown in B-E, H, and I, with mean lifespans indicated in parentheses. Statistics and all individual experiments are described in Tables S13 (B, C), S14 (D, E), S15 (F, G), S16 (H, I), and S17 (J) (Supporting information). These lifespan experiments are compared with others performed at the same temperatures in Tables S21 and S22 (Supporting information).
is required in the TRPA-1 pathway that increases C. elegans lifespan at low temperature (Xiao et al., 2013) could explain the relative sensitivity of sgk-1 mutants to temperatures of 20°C or below (Fig. 6A). Our evidence that rict-1 and sgk-1 mutations affect lifespan and stress resistance essentially identically, and that mTORC2 promotes SGK-1 activity, predicts that RICT-1/mTORC2 also acts in the TRPA-1 pathway, by activating SGK-1 (Fig. 6A). Taken together, our findings suggest the working model that mTORC2 and SGK-1 act in two pathways that affect lifespan oppositely (Fig. 6A). They inhibit the pro-longevity and stressresponse effects of SKN-1, but on the other hand promote longevity in the context of the TRPA-1 low-temperature pathway. The latter pathway appears to function independently of SKN-1 (Xiao et al., 2013). The balance between these two pathways would determine how mTORC2/ SGK-1 influences longevity. If the pathways suggested by our model are linear, it might be possible to compensate for lack of mTORC2 activity by activating SGK-1. To test this idea, we examined how lifespan is affected by the sgk-1 gain-of-function allele ft15, which was identified in a rict-1 suppressor screen (Jones et al., 2009). This sgk-1(gf) mutation suppressed the effects of a rict-1 hypomorph on fat accumulation, body size, and development and partially suppressed the effects of a rict-1 null mutation on body size, suggesting that sgk-1(gf) allows the SGK-1 kinase to be partially active in the absence of mTORC2 activity. Importantly, sgk-1(gf) prevented rict-1 RNAi from increasing lifespan (Fig. 6B,C, Table S13, Supporting information) and substantially reduced the response of rict-1 mutants to propagation on quality food that is seen at 25°C (Fig. 6D,E, Table S14, Supporting information; compare rict-1 and rict-1; sgk-1(gf) mutants). This strongly supports the model that RICT-1/mTORC2 inhibits SKN-1 by activating SGK-1 and that this pathway mediates the food-source-dependent effects of rict-1 or sgk-1 loss on lifespan and stress resistance.
Constitutive SGK-1 activation might not only compensate for the effects of reducing mTORC2 activity, but also increase lifespan through constitutive activation of the TRPA-1 temperature-sensing pathway (Fig. 6A). Consistent with this idea, at 20°C, the sgk-1(gf) mutation modestly increased lifespan in a daf-16-dependent manner . We also found that sgk-1(gf) mutants lived slightly longer than WT and observed lifespan extension not only at 20°C but also at 25°C with HT115 feeding (Fig. 6B-G and Tables S13-S15, Supporting information). The explanation for the latter result is not clear, but it could imply that under some circumstances, the TRPA-1 pathway can increase lifespan at higher temperatures. At 25°C with HT115 feeding, lifespan can therefore be increased modestly by either complete loss of SGK-1 (Fig. 4E) or partial SGK-1 activation (sgk-1(gf); Fig. 6B,C,E). In the former case, SGK-1-mediated inhibition of intestinal SKN-1 is released, and in the latter case, any possible reduction in SKN-1 activity is presumably outweighed by benefits of increased SGK-1 function through the TRPA-1 pathway or a different mechanism. Interestingly, simultaneous loss of rict-1 by mutation prevented sgk-1(gf) from increasing lifespan (Fig. 6D-G, Tables S14 and S15, Supporting information). Apparently, the partial constitutive SGK-1 activation that occurs in sgk-1(gf) mutants (Jones et al., 2009) is not sufficient to overcome the need for rict-1 in this context.

Complex homeostatic functions of mTORC2 in vivo
By showing that SGK-1 mediates the effects of mTORC2 on C. elegans stress resistance and lifespan, we have further demonstrated the importance of SGK-1 as a biological target of this little-understood but critical kinase complex. In mammals, the SGK kinases regulate numerous ion channels but may have many other overlapping functions, and have been implicated in hypertension, metabolic diseases, and cancer (Bruhn et al., 2013;Lang & Shumilina, 2013). Together with other C. elegans studies (Jones et al., 2009;Soukas et al., 2009), our new findings suggest models for how mammalian SGK kinases could be involved in mTORC2regulated functions in growth, metabolism, stress defense, and interactions with the environment. Our data have also reconciled seemingly conflicting observations with respect to how C. elegans SGK-1 influences lifespan. They are consistent with all published findings on SGK-1 and longevity, except for the report that the lifespan increase from sgk-1 RNAi was daf-16-dependent (Hertweck et al., 2004), a discrepancy that could derive from differences in RNAi conditions. Most importantly, we have shown that by activating SGK-1, mTORC2 regulates two distinct longevity pathways that have opposing effects (Fig. 6A).
In one pathway, mTORC2/SGK-1 inhibits SKN-1 (Fig. 6A). While much remains to be learned about mTORC2 functions, it is clear that this kinase complex is involved in growth (Jones et al., 2009;Soukas et al., 2009;Zinzalla et al., 2011;Wang et al., 2012). Other growth-related pathways also inhibit SKN-1 (mTORC1 and IIS) (Tullet et al., 2008;Robida-Stubbs et al., 2012), suggesting that it might be advantageous to suppress SKN-1-regulated protective mechanisms under anabolic conditions and to mobilize these stress defenses when growth signals and resources are low. IIS increases activity of the AKT kinase, which inhibits SKN-1 directly through phosphorylation (Tullet et al., 2008), but also activates SGK-family kinases in C. elegans and mammals (Hertweck et al., 2004;Bruhn et al., 2013;Lang & Shumilina, 2013). This makes it of interest to determine whether the mTORC2 and IIS pathways might converge at SGK-1 and SKN-1 under certain conditions. During the earliest embryonic stages in C. elegans, SKN-1 specifies endomesoderm development (Bowerman et al., 1992), a function that is antagonized by mTORC2/SGK-1 but not mTORC1 (Ruf et al., 2013), suggesting that the effects of these growth regulatory pathways on SKN-1 are both context and tissue dependent.
The striking influence of food source on rict-1 mutant stress resistance and lifespan involves regulation of SKN-1 at a step downstream of or in parallel to SGK-1 (Fig. 6A). Apparently, signals produced by quality food lower a threshold for SKN-1 activity that determines whether reduction in mTORC2/SGK-1 signaling is sufficient to release SKN-1 from inhibition. The difference in rict-1 lifespan that resulted from OP50 vs. HT115 feeding disappeared with liquid culture (Fig. 5C), suggesting that critical signals from these bacterial strains may be volatile, or that culture conditions influence their production or perception. Recently, neuronal signaling was implicated in a dietary adaptation to alterations in proline catabolism (Pang & Curran, 2014) and in a transcriptional response to bacterial signals that resulted in C. elegans developing at different rates when propagated on OP50 and the soil bacteria Comamonas (MacNeil et al., 2013). We have determined that signals from the bacterial microenvironment also modulate intestinal SKN-1 activity, particularly when mTORC2/SGK-1 activity is low. Each of these findings provides a paradigm for applying this simple model organism to investigate how interactions between a host organism and its bacterial microbiome lead to changes in gene expression, metabolism, and health. Our results also raise the exciting possibility that in mammals, the microbiome might influence the activity of Nrf proteins.
In a second longevity pathway that involves an environmental input, we found that RICT-1/mTORC2 acting in neurons is required for the TRPA-1/SGK-1 pathway to increase lifespan at low temperature (Fig. 6A). The importance of RICT-1/mTORC2 in this mechanism suggests the exciting possibility that mTORC2 might respond to either temperature or Ca 2+ signaling (Fig. 6A). Surprisingly, under liquid culture conditions, rict-1 mutants lived nearly as long as WT animals at 20°C, in striking contrast to results obtained on plates (Fig. 2C,D, and 5C). Perhaps C. elegans adapts to lower temperatures independently of the TRPA-1 pathway in liquid culture, or a particular aspect of the liquid culture environment (reduced oxygen availability, constant thrashing) might largely circumvent the need for this pathway. Alternatively, the temperature signal itself might be influenced by immersion in liquid. Our results identify an unexpected function for RICT-1/mTORC2 in this adaptation to temperature and provide an exciting new entry point for probing the TRPA-1/SGK-1 longevity pathway and other processes that involve mTORC2. They also demonstrate how environmental inputs from culture conditions, temperature, and signals from bacteria may substantially alter how genetic pathways influence C. elegans aging, illustrating the value of the organism for uncovering these regulatory networks.

Experimental procedures
Full detailed methods and experimental procedures are available in Supporting Information.

Supporting Information
Additional Supporting Information may be found in the online version of this article at the publisher's web-site.

Table S1
Oxidative stress assays shown in Fig. 1A.

Table S2
Oxidative stress assays shown in Fig. 1B,C.

Table S3
Oxidative stress assays shown in Fig. 1D,E.