Lithium Promotes Longevity through GSK3/NRF2-Dependent Hormesis

1Lithium is the most commonly prescribed drug for the treatment of bipolar disorder. It also improves disease phenotypes in animal models of many clinical conditions including Alzheimer disease, depression, and stroke. The effects of lithium on aging have been documented in yeast and Caenorhabditis elegans, with lithium extending lifespan. The effects of lithium on Drosophila aging have previously been inconclusive, with demonstration of both positive and negative effects on survival. Moreover, lithium concentration in the drinking water of a large Japanese population has been associated with reduced all-cause mortality, suggesting that lithium may be a bona fide anti-aging drug. However, the mechanisms by which lithium acts in humans remain poorly understood.

2In vitro studies have reported that lithium can protect against several forms of oxidative and xenobiotic stressors, but in vivo evidence for such protective effects of lithium is lacking. Longevity has been extensively correlated with resistance to stress. Transcriptomic analysis of interventions known to extend lifespan have identified particular genes likely to be involved in stress resistance. Upregulation of the transcription factor cap’n’collar C (CncC, an NRF-2 homolog) has been shown not only to confer resistance to toxic compounds, but also to promote longevity in C. elegans and flies. In flies and mammals, NRF-2/CncC is negatively inhibited through cytosolic sequestration and proteasomal degradation by the canonical Keap1. However, a second emerging upstream regulator of NRF-2/CncC is GSK-3, a well-documented target of lithium. GSK-3 regulates NRF-2 by phosphorylation and nuclear exclusion, an effect that is evolutionarily conserved from invertebrates to mammals. Interestingly, GSK-3 inhibition has been shown to phenocopy the effects of lithium for protection against xenobiotic stress in vitro.

3Activation of NRF-2/CncC produces hormetic effects on lifespan, such that at low level NRF-2/CncC activity extends lifespan while higher levels of activation limit it. Interestingly a hormetic signature was recently reported for the survival of a mammalian cell line treated with lithium, suggesting that lithium and GSK-3 inhibition could influence animal lifespan and stress resistance through activation of NRF-2.

4Here we show that lithium supplementation in the diet can modulate longevity, stress resistance, and metabolism in Drosophila through the inhibition of GSK-3. Correspondingly, genetic downregulation of GSK-3 and lithium treatment are epistatic, suggesting a common molecular pathway. We also show that lithium and the genetic inhibition of GSK-3 promote xenobiotic stress resistance and lifespan extension through the activation of a transcriptional response mediated by CncC/NRF-2. Furthermore, lithium protects against a high-sucrose diet and acts through mechanisms that only partially overlap with those mediating lifespan extension by dietary restriction (DR). These findings demonstrate an alternative genetic and pharmacological target for the promotion of longevity and stress resistance, and emphasize the potential of pharmacological inhibitors of GSK-3 as viable anti-aging treatments.

5To assess the role of lithium in Drosophila aging, we treated adult female flies with lithium chloride (LiCl) by supplementation in their food. Lithium treatment in the range of 1 to 25 mM resulted in lifespan extension, whereas higher doses (50–100 mM) shortened lifespan (Figure 1A). These effects of lithium treatment on lifespan extension were also observed in an independent genetic background (Figure S1A) and in males (Figure S1B). Thus, lithium treatment extended Drosophila lifespan independently of genetic background and sex.

6To ensure that the increased lifespan observed with lithium supplementation was dependent on the addition of lithium itself, we treated flies with equivalent molar concentrations of sodium chloride (NaCl) and found no lifespan extension (Figures S1C and S1D). Thus, the pro-longevity effect of LiCl is specific to lithium and not its chloride counterion.

7Interestingly, we observed that, unlike with many other genetic and pharmacological interventions (e.g., DR, insulin/IGF downregulation, rapamycin, or trametinib treatment), lithium did not reduce fecundity at life-extending doses or compromise feeding behavior (Figures S1E and S1F). Moreover, it delayed locomotor decline at two concentrations that extend lifespan (Figure 1B). Thus, lithium promotes healthspan in adult Drosophila with limited side effects.

8To limit the side effects of long-term use, a drug that improves lifespan and healthspan will ideally do so with late-onset administration. We therefore assessed the effect of commencing lithium treatment at older ages. Flies were switched onto food containing a range of lithium concentrations (1–75 mM) at 32 days of age (Figure 1C). Lower doses (1–25 mM) of lithium extended lifespan, whereas higher doses (50 and 75 mM) significantly reduced lifespan, similar to the dose-dependent effects we observed in younger flies.

9We also tested whether transient lithium treatment early in life could increase lifespan. We therefore exposed young flies to 1 or 10 mM lithium for 15 days and then switched them to control food for the remainder of their lifespans. Early treatment with these doses of lithium extended lifespan (Figure 1D). Lithium treatment early in life, and for a transient period, can therefore increase survival later in life.

10Genetic and environmental interventions that extend lifespan often induce abnormalities in carbohydrate and lipid metabolism. We therefore examined the effects of lithium on whole body trehalose, glycogen, and triglyceride levels. Following 15 days of lithium treatment, and over a wide range of lithium concentrations, we were unable to detect a significant change in the levels of either trehalose or glycogen (Figures S1G and S1H). However, we observed a dose-dependent reduction in whole body triglycerides, the main lipid storage in flies (Figures 1E and S1I). In keeping with the lowered triglyceride levels, lithium treatment reduced survival under starvation conditions in a dose-dependent manner (Figures 1F and S1J). Moreover, lithium also extended lifespan under dietary conditions that promote triglyceride accumulation. Flies fed a high-sucrose diet were short lived and lithium was able to partially rescue this defect (Figure 1G) while completely blocking the increase in triglycerides observed with a sucrose-rich diet (Figure 1H). Therefore, lithium can extend lifespan under obesogenic dietary conditions.

11We next investigated whether lithium treatment was acting as a DR mimetic. DR is a well-established anti-aging intervention that extends healthy lifespan in diverse species, and some pharmacological and genetic interventions that extend lifespan have features of DR mimetics. To determine whether lithium and DR extend lifespan by similar mechanisms, we assessed whether lithium could extend lifespan beyond the maximum achievable by DR. To maximize lifespan under DR, we varied the yeast concentration in the food while maintaining a constant concentration of sucrose, resulting in a typical tent-shaped response, with peak lifespan at food containing a 1.0 yeast concentration (Figures 2A and S2A–S2D). If lithium treatment and DR share overlapping pathways, then lithium would not be able to further extend lifespan already maximized by DR. All lithium doses tested significantly extended median lifespan in both the yeast condition that maximized lifespan (1.0 yeast; Figures 2A and S2C) and under full feeding (2.0 yeast; Figures 2A and S2D), with greatest extension of median lifespan with 10 mM lithium under full feeding. However, under reduced yeast concentrations that shorten lifespan (0.2 and 0.5 yeast), 10 mM lithium either significantly reduced lifespan (Figure S2B) or did not confer a significant lifespan benefit (Figure S2A). Cox proportional hazards analysis showed a significant interaction between lithium and yeast concentrations for lifespan (interaction term p < 0.0001). The extension of lifespan from lithium increased with the level of yeast in the fly diet, suggesting partially overlapping mechanisms to those of DR.

12A well-known target of lithium is GSK-3. We therefore evaluated the phosphorylation status of the fly ortholog of GSK-3, Shaggy (Sgg), in response to lithium treatment. Lithium addition to the fly medium resulted in a dose-dependent increase in the inhibitory phosphorylation (Serine 9 or S9) of Sgg (Figure 2B). To evaluate the role of Sgg in lithium-mediated lifespan extension, we directly manipulated its activity in adult flies. Ubiquitous overexpression of wild-type or constitutively active Sgg (SggS9A) significantly reduced lifespan by ∼30% and 50%, respectively (Figures 2C and S2E). This reduction in lifespan was almost completely reversed by lithium treatment. Furthermore, RNAi-mediated reduction in sgg expression using two independent dsRNA-expressing transgenes significantly increased lifespan (Figures 2D and S2F). Importantly, lithium was unable to further increase the lifespan of these sgg RNAi knockdown mutants flies (Figure 2D). Taken together, these findings suggest that Sgg/GSK-3 inhibition and lithium treatment increase lifespan by acting on the same downstream targets.

13To identify downstream mediators of lifespan extension by lithium and of GSK-3 inhibition, we analyzed the genome-wide transcript profiles of lithium-treated flies using microarrays. Genes encoding ribosomal proteins were among the most upregulated (Figure 3A) and downregulated (Figure S3A) gene ontology (GO) categories in lithium-treated flies. This transcriptional response could underlie the translational repression following lithium treatment that has been previously observed in fission yeast and Drosophila heads. In addition, five GO terms for genes encoding enzymes in the detoxification pathway were also in the ten most upregulated categories (Figure 3A).

14The responses to xenobiotics and oxidative stress in Drosophila are regulated by the transcription factors dFOXO, CncC, and DHR96. We therefore assessed whether the transcriptional responses to activation of these transcription factors overlapped with that of lithium treatment. The transcriptomic response to lithium did not overlap with that of dFOXO-dependent or -independent transcriptional regulation downstream of IIS (Figures S3B and S3C). Furthermore, although we detected a significant overlap in the transcriptional signatures of lithium and DHR96, they did not share the same directionality (Figure S4). However, we found a significant overlap (Figure 3B) between the genes that were upregulated by lithium and cncC overexpression, but not between genes downregulated by both treatments (Figure S5A), suggesting that lithium might activate a CncC transcriptional response downstream of GSK-3. The barbiturate phenobarbital activates CncC and induces a similar transcriptional response to that of cncC overexpression. We therefore analyzed the overlap between the transcriptional profiles induced by lithium and phenobarbital treatment, and again found a significant overlap (Figure 3B) between upregulated, but not downregulated, genes (Figure S5B). The genes upregulated in common between lithium treatment, phenobarbital treatment and cncC overexpression (Figures 3B and S5C) encoded enzymes that participate in all three phases of xenobiotic metabolism (Figure 3C). To further confirm the activation of CncC by lithium, we used a previously generated CncC reporter that responds to both chemical and genetic inducers of CncC. Flies carrying the GstD-eGFP CncC reporter showed a dose-dependent increase in GFP expression with increasing concentrations of lithium (Figure 3D). Taken together, our results suggest that lithium activates CncC to upregulate the expression of genes in the detoxification pathway.

15We next assessed whether CncC activity is required for the pro-longevity effects of lithium. Ubiquitous, RNAi-mediated knockdown of cncC expression blocked the lifespan extension of 1 to 10 mM lithium, but was detrimental to survival in flies treated with 25 mM lithium, the highest dose that extends lifespan under basal conditions, albeit to a lesser extent (Figure 4A). Thus, lithium treatment requires CncC activity to confer its longevity benefits.

16Because CncC/NRF-2 can induce hormesis, we assessed whether lithium can also do this. To test for a hormetic effect of lithium at low doses, we pre-treated flies with a range of concentrations of lithium and then challenged them with a toxic dose of 500 mM. Most pre-treatment doses of lithium induced subsequent resistance to the toxic dose (Figure 4B). To assess whether the hormetic response of lithium was mediated by CncC, we knocked down expression of cncC using RNAi, and treated the flies with 1–25 mM lithium. Reduction in cncC expression completely blocked the hormetic response induced by 10 mM lithium pre-treatment, and significantly reduced the effect of 25 mM lithium (Figure 4C).

17We next assessed the ability of lithium pre-treatment to protect against other xenobiotics. Flies pre-treated with increasing concentrations of lithium ranging from 1 to 100 mM were significantly resistant to a toxic concentration of phenobarbital, with lithium doses between 1 and 75 mM almost doubling survival (Figure 4D). Lower doses of lithium also protected against a toxic dose of the anti-malarial drug, chloroquine (Figure S5D; 1–10 mM), and the pesticide paraquat (Figure 4E). Thus, low to intermediate concentrations of lithium protect against xenobiotic toxicity. To determine the role of CncC activity in lithium-mediated protection against phenobarbital, we used RNAi to knock down expression of cncC, which sensitized the flies to phenobarbital and completely abrogated the protection against phenobarbital afforded by lithium supplementation (Figure 4F). Thus, CncC is at least partly responsible for the hormetic effect induced by low-level treatment with lithium.

18To confirm that Sgg, upstream of CncC, is also necessary for the resistance to xenobiotic stress, we assessed the effect of ubiquitous overexpression of wild-type sgg or the constitutively active Sgg(S9A) on xenobiotic resistance. Both significantly sensitized flies to phenobarbital (Figures 5A and S5E). We confirmed that sgg or sgg(S9A) overexpression regulated CncC by showing significantly lower levels of MRP and keap1 (Figures 5B and S5F), both CncC target genes. Correspondingly, RNAi-mediated knockdown of sgg resulted in resistance to phenobarbital (Figure 5C), and paraquat (Figure S5G). An increase of mRNA levels of cncC, keap1, and gstD2 confirmed that CncC was active in sgg knockdown flies (Figure 5D). Thus, increased Sgg activity sensitizes against xenobiotic stressors, whereas its inhibition protects against them.

19In addition to activating CncC by repressing Sgg/GSK-3, lithium could potentially increase CncC activity by inhibiting its canonical repressor Keap1. Hence, we analyzed the interaction between lithium treatment and Keap1. Overexpression of Keap1, which inhibits CncC activity in vivo, was unable to prevent the lifespan-extending properties of lithium (Figure 6A), suggesting that the longevity effect of lithium treatment is independent of Keap1. Next, we analyzed the interaction of loss of Keap1 and lithium treatment. We generated a deletion of the keap1 coding sequence by P-element-mediated male recombination using a previously described P-element insertion line (Figure 6B). The keap1 deletion (keap1Del) was homozygous lethal, but activated CncC 4-fold in the heterozygous state, as measured by the CncC reporter (Figure 6C). Lithium treatment of the keap1Del flies further activated CncC (Figure 6C). We next tested whether this effect on CncC activation protected against paraquat and lithium toxicity. keap1Del flies were significantly resistant to both paraquat and lithium (Figures 6D and 6E), and pre-treatment with lithium further protected them. We confirmed these findings using a previously described heterozygous loss-of-function mutation in the keap1 gene (keap1EY5) (Figures S6A and S6B). Thus, the combination of loss of keap1 and lithium treatment further protected against paraquat and lithium-induced toxicity, suggesting that stronger CncC activation results in greater protection against these xenobiotics.

20We subsequently evaluated the interaction between loss of keap1 and lithium treatment for longevity. Survival analysis showed that the lifespan of keap1Del mutant flies was indistinguishable from controls, but that addition of 1 mM lithium marginally, yet significantly, extended lifespan (Figure 6F). Increasing the dose of lithium to 10 mM restored longevity to control levels. The keap1EY5 mutant flies showed a significant lifespan extension (Figure 6G). However, supplementation of either 1 or 10 mM lithium to the keap1EY5 mutant shortened lifespan in a dose-dependent manner. These results suggest that the level of activation of CncC that maximizes extension of lifespan is considerably lower than that which maximizes protection against toxic doses of lithium and paraquat.

21Activation of autophagy has been proposed as a mechanism for the beneficial effects of lithium. We therefore analyzed the induction of autophagy by LC3-I/LC3-II (Atg8 in Drosophila) levels without detecting statistically significant changes. Indeed, there was a tendency for lower LC3-I that did not reach statistical significance (Figure S7A). Moreover, lithium treatment was able to extend the lifespan of flies with autophagy defects due to heterozygous loss of atg1 (Figure S7B). Thus, taken together our results do not immediately support a role for autophagy in the pro-longevity effects of lithium treatment, and strengthen our conclusion that they are mediated through the inhibition of GSK-3 and the subsequent activation of CncC/NRF-2 (Figure 7). However, it remains possible that induction of autophagy occurs in atg1-deficient flies, or that lithium induces autophagy in a tissue-specific manner.

22Drug repurposing is the most promising approach for developing pharmacological agents to improve healthy aging. So far, two medically approved drugs, metformin and rapamycin, have been reported to promote longevity and provide health benefits across species from invertebrates to mammals. We and others have shown that lithium can extend lifespan in fission yeast, C. elegans, and Drosophila. We also showed that this effect was common between two different laboratory strains and, unlike other interventions that seem to be more effective in females, lithium similarly extended lifespan in both sexes.

23Lifespan-extending drugs can often act like DR mimetics; hence, it was important to determine whether lithium was acting in a similar manner. While low doses of lithium were able to extend lifespan at all dietary levels tested, median lifespan extension was greatest under full feeding conditions. Our data thus suggest that lithium and DR act via partially overlapping mechanisms and confirms the observation made in C. elegans that lithium extends lifespan of eat-2 mutants, a genetic model of DR in worms. Lithium also extended the lifespan of flies fed a diet enriched with sucrose, possibly by modulating lipid metabolism. However, the role of CncC in modulating the triglyceride phenotype of lithium remains to be explored. Overall, our observations strongly suggest that lithium is a pro-longevity drug capable of extending lifespan at low doses independent of sex and genetic background, and under a variety of dietary conditions.

24In humans, the therapeutic window for lithium treatment of bipolar disorder lies between 0.5 and 1 mM in serum, whereas concentrations of 1.5 mM and above severely increase the risk of tissue damage. Previous work in Drosophila suggests that the dose range at which we observed lifespan extension (0.5–25 mM) translates to Drosophila tissue concentrations below 0.5 mM. As previously reported for C. elegans and Drosophila, concentrations above 50 mM were highly toxic.

25Drug interventions to promote healthy lifespan are less likely to have side effects if started late in life. Only a handful of drugs approved by the US Food and Drug Administration, namely rapamycin, metformin, and the Ras inhibitor trametinib, induce lifespan extension when commenced at later ages in model organisms. We found that lithium extends lifespan when first administered in mid-late life. In humans, long-term treatment with lithium for psychiatric disorders is associated with progressive and permanent renal damage. We showed that short treatment periods in Drosophila, 15 days during early adulthood, are sufficient to prolong life. Taken together, our data suggest that when testing lithium as a pro-longevity drug in mammals, lower doses than those used in psychiatric disorders are likely to be sufficient, and other strategies such as alternate-day dosing or transient treatment periods (either early or late in life), may be sufficient to reduce undesirable side effects and maximize the potential health benefits.

26Interestingly, doses of lithium that shortened lifespan were protective against certain forms of xenobiotic stress. In vitro studies in mammalian cells have shown that lithium, and other GSK-3 inhibitors, protect against cell death caused by rotenone-induced oxidative stress, glutamate excitotoxicity, and H2O2. This is likely mediated through a hormetic response, in this case orchestrated by NRF-2 activation. We observed that while simultaneous activation of CncC by loss of Keap1 and lithium treatment is additive and confers greater stress resistance to xenobiotics, the threshold for lifespan extension is perhaps considerably lower. A similar situation has been observed in C. elegans in which strong activation of the endoplasmic reticulum unfolded protein response conferred stress resistance benefits, while shortening lifespan. Our findings thus suggest that while NRF-2 activation either by loss of Keap1 or inhibition of GSK-3 is beneficial for longevity and stress resistance, at low levels of activation, stronger induction is detrimental for lifespan. This suggests that the hormetic benefits of lithium are more likely to occur at low levels under basal non-stress conditions. Hence, when testing for GSK-3 inhibitors or NRF-2 activators in modulating animal (and especially mammalian) aging, the degree of NRF-2 activation within the hormetic curve will determine positive or negative longevity outcomes. Future work studying the convergence of the salutary and damaging effects of lithium will aid in understanding to what extent the molecular mechanisms are shared. Additionally, our microarray analysis was performed in heads and thoraces; therefore, it remains to be explored to what extent systemic or localized activation of NRF-2 modulates longevity, stress resistance, and lipid metabolism at the tissue level.

27Complete absence of GSK-3 in C. elegans, Drosophila, and mice shortens lifespan or prevents development, while moderate inhibition has been associated with most of its positive effects. GSK-3 is upregulated in many disease states, including neurodegeneration, diabetes, inflammatory conditions, and some cancers. We have shown that adult-specific genetic manipulation of the fly ortholog of GSK-3, Sgg, affects longevity. Downregulation of Sgg prolonged lifespan and lithium was unable to further extend the lifespan, suggesting that lithium and inhibition of Sgg act through a common molecular pathway to extend lifespan.

28In C. elegans and mammalian cells, GSK-3 directly interacts with NRF-2 to repress its activity, independently of Keap1. Therefore, we hypothesized that lithium might act via Sgg/GSK-3, to de-repress CncC, the fly ortholog of NRF-2 and activate the oxidative and xenobiotic stress transcriptional signature, which in turn would induce a CncC/NRF-2-dependent protective response. GO enrichment analysis identified a transcriptional signature that indeed suggested that lithium acts via CncC/NRF-2. CncC activity was indispensable for the lifespan extension conferred by lithium. In keeping with our results, work in rodents and mammalian cell lines has shown that lithium treatment and GSK-3 inhibition activate NRF-2. Because activation of CncC/NRF-2 modulates longevity in C. elegans and Drosophila, our results provide evidence that GSK-3 is a viable therapeutic target to promote longevity via activation of NRF-2.

29To date, the only GSK-3 inhibitor approved for human use is lithium. However, researchers and pharmaceutical companies have developed more selective GSK-3 inhibitors, some of which have already entered the early stages of clinical trials for obesity, Alzheimer disease, and progressive supranuclear palsy. Our results call for a reassessment of the potential use of GSK-3 inhibitors and NRF-2 activators as potential anti-aging compounds.

30The w1118 stock was obtained from Bloomington Drosophila Stock Center. The control white Dahomey (wDah) stock has been maintained in large population cages with overlapping generations since 1970. The wDah stock was initially derived by incorporation of the w1118 mutation into the outbred Dahomey background by backcrossing. Further details concerning fly mutants can be found in the Supplemental Experimental Procedures.

31LiCl (Sigma) or NaCl (Sigma) were dissolved in ddH2O at a concentration of 5 M before supplementing to the medium. Equivalent volumes of vehicle were supplemented to the medium to compensate for dilution.

32The DR protocol was performed as described previously.

33Statistical analyses were performed using Excel, GraphPad Prism, or JMP software version 9 (SAS Institute). Survival experiments were analyzed using log rank test. Other data were tested by one-way analyses of variance (ANOVA) and planned comparisons of means were made using Tukey-Kramer HSD test. Cox proportional hazards analysis was performed to compare interactions for survival.

34J.I.C.-Q. and I.B. conceived the experiments. J.I.C.-Q., I.B., L.L., K.J.K., L.S.T., T.N., and F.K. performed the experiments. D.K.I. analyzed the microarray data. C.S. and I.B. contributed reagents. J.I.C.-Q., I.B., J.T., J.H., and L.P. supervised experiments/project. J.I.C.-Q. and L.P. wrote the manuscript. All authors approved the final submission.