Edited by Andy Dillin, University of California, Berkeley, CA, and accepted by the Editorial Board April 28, 2014 (received for review November 28, 2013)
Recently it has been suggested that metformin, the most commonly used antidiabetic drug, might also possess general health-promoting properties. Elucidating metformin’s mode of action will vastly increase its application range and will contribute to healthy aging. We reveal a signaling cascade in which metformin is able to extend lifespan by increasing the production of reactive oxygen species (ROS). This allowed us to further work at the crossroads of human disease and aging research, identifying a key molecule that is able to translate the ROS signal into a prolongevity cue: an antioxidant peroxiredoxin is also able to activate a lifespan-promoting signaling cascade, here described in detail. Continued research efforts in this field lead toward a targeted improvement of aging-related complications.
The antiglycemic drug metformin, widely prescribed as first-line treatment of type II diabetes mellitus, has lifespan-extending properties. Precisely how this is achieved remains unclear. Via a quantitative proteomics approach using the model organism Caenorhabditis elegans, we gained molecular understanding of the physiological changes elicited by metformin exposure, including changes in branched-chain amino acid catabolism and cuticle maintenance. We show that metformin extends lifespan through the process of mitohormesis and propose a signaling cascade in which metformin-induced production of reactive oxygen species increases overall life expectancy. We further address an important issue in aging research, wherein so far, the key molecular link that translates the reactive oxygen species signal into a prolongevity cue remained elusive. We show that this beneficial signal of the mitohormetic pathway is propagated by the peroxiredoxin PRDX-2. Because of its evolutionary conservation, peroxiredoxin signaling might underlie a general principle of prolongevity signaling.
Metformin, an antiglycemic biguanide drug and the most common treatment of type II diabetes mellitus, has life-extending capabilities (1, 2). Several other human diseases, such as cancer (3) and nonalcoholic fatty liver disease (4), are also potentially alleviated by metformin treatment. This suggests that metformin acts on common pathways involved in a spectrum of aging-related disorders. Because of its demonstrated beneficial effect on lifespan in the nematode Caenorhabditis elegans (5, 6) and in the rodents Rattus norvegicus (1) and Mus musculus (2), these models facilitate research into the underlying mode of action.
It has been hypothesized that metformin elicits its beneficial effects by mimicking dietary restriction (DR) (7), a regimen wherein a physiological response is triggered by reducing the uptake of nutritive calories. The physiological response to DR causes lifespan extension and delays age-dependent decline from yeast to primates (8). The idea of similarity sprouts from the observed low blood glucose and insulin levels combined with increased glucose utilization in both calorically restricted and metformin-treated animals (7). In addition, metformin-treated worms show phenotypes similar to DR worms (5), and transcript profiles of metformin-treated and DR mice also overlap significantly (9).
Caution is due, however, in referring to DR, because different methods to induce DR in C. elegans act through different genes to elicit corresponding effects on lifespan (10⇓–12). Glucose restriction, a specific type of DR, requires AMP-dependent kinase (AMPK) to extend lifespan. Activation of AMPK leads to increased mitochondrial production of reactive oxygen species (ROS), which induces stress defense and results in a net increase in longevity. This process of lifespan extension based on mitochondrial oxidative stress is known as mitohormesis (13).
Despite its similarities to DR and its widespread use as an antiglycemic drug, the actual mode of action of metformin is largely unknown and a subject of much debate. In mammals, metformin is generally believed to act through the activation of AMPK, one of the main regulators of cellular energy homeostasis (2, 14⇓–16), but recent research suggests the existence of AMPK-independent mechanisms as well (17, 18). More upstream, metformin is thought to activate AMPK through partial inhibition of complex I of the electron transport chain (ETC) and a resultant increase in the AMP/ATP ratio (19⇓–21), although again, not all data support this theory (16, 22). Important players in metformin-induced lifespan extension in C. elegans are the liver kinase B1 ortholog PAR-4, the AMPK ortholog AMP-activated kinase-2 (AAK-2), and the skinhead-1 (SKN-1) transcription factor, which is involved in activating phase II detoxification mechanisms (5). It has recently been demonstrated that in C. elegans, metformin partly elicits its effects through altering the folate metabolism of its microbial food source (6), but questions about its direct effects are largely unaddressed. It is, for instance, unclear how metformin induces AAK-2 and SKN-1 activity. Overall, many gaps remain in our knowledge of metformin-induced lifespan extension.
To study the targets of metformin, we performed a proteomic analysis on metformin-treated C. elegans and used the results as a framework for follow-up experiments. We observed striking similarities between metformin-treated and glucose-restricted worms and discovered several factors involved in mitohormetic regulation of lifespan, including a previously unidentified role for peroxiredoxin-2 (PRDX-2), which seems to be responsible for translating oxidative stress into a downstream prolongevity signal. In addition to lifespan extension, we also report features that contribute to the healthspan of metformin-treated worms.
To gain a deeper understanding of the physiological changes elicited by metformin exposure, a differential gel-based proteomics experiment was performed. A total of 164 spots with differential abundances were detected, 134 of which could be identified by mass spectrometry. After removal of duplicate identifications, the final analysis resulted in a list of 58 up-regulated and 30 down-regulated proteins (Dataset S1).
Enrichment analysis (23) of the up-regulated proteins resulted in the detection of several overrepresented pathways (Table 1), of which the branched-chain amino acid (BCAA) degradation pathway was most markedly enriched. All other enriched pathways, including glycolysis and the tricarboxylic acid (TCA) cycle, are involved in general energy metabolism. No significant pathway enrichment was observed in the analysis of down-regulated proteins. Combined analysis of all differential proteins (both up and down) did not reveal additional pathways of interest.
Functional analysis of differential proteins
Further functional clustering was used to subdivide the up- and down-regulated proteins into meaningful groups (Table 1). For each functional cluster detailed information, including subterms and proteins in each group, was gathered to assist in data interpretation (Dataset S1). These results were used as the backbone of hypothesis-driven experiments into the mechanisms induced by metformin treatment.
Because of the marked clustering of mitochondrial proteins (Table 1) and metformin’s putative inhibitory effect on complex I of the ETC, it needs to be tested whether metformin is capable of inducing the mitochondria-specific unfolded protein response (UPRmt). This is essential because the UPRmt increases lifespan in response to either misfolding of mitochondrial proteins or stoichiometric abnormalities in the ETC complexes (24).
We did not observe up-regulation of the UPRmt marker hsp-6 in worms exposed to metformin (Fig. 1). Additional experiments with higher concentrations of metformin and different exposure times yielded similar negative results (Fig. S1). As such, lifespan extension of metformin via the induction of the UPRmt is unlikely, which suggests that the drug affects the mitochondria in another way.
Metformin treatment does not induce the mitochondria-specific unfolded protein response (UPRmt). Day-1 adult metformin-treated hsp-6::GFP worms show no difference in fluorescence compared with untreated control worms. hsp-6::GFP worms were exposed to cco-1 RNAi as a positive control. cco-1 encodes a cytochrome c oxidase subunit, an integral part of the mitochondrial electron transport chain (24).
Metformin has been put forward as a possible DR mimetic (7), and because DR has been linked to mitohormesis (13, 25), we looked into evidence for a role for mitohormesis in metformin-induced lifespan extension. Clustering analysis of the proteomics experiments revealed a significant overrepresentation of several mitochondrial proteins and proteins involved in catabolism (Table 1). These clusters might point to an increase in respiration, one of the hallmarks of mitohormesis (13). Indeed, upon metformin treatment, we observed a significant increase in respiration (45%) and metabolic heat production (38%) (Fig. 2A) in wild-type worms. These changes, in combination with the AMPK-dependence of metformin-mediated longevity (5), strongly resemble the mitohormetic pathway (13) and might point toward an increase in ROS production.
Metformin increases lifespan according to the principle of mitohormesis, for which it requires PRDX-2. (A) Metformin treatment increases metabolic heat production (**P < 0.01; n = 3 for untreated and n = 4 for treated worms) and respiration (*P < 0.05; n = 3). Bars represent mean ± SEM. (B) Metformin induced a significant increase in H2O2 release in day-1 adult worms after both continuous exposure during development (*P < 0.05; n = 7) and after 24 h of exposure, starting from the young adult stage (*P < 0.05; n = 7 for untreated and n = 5 for treated worms). Exposing the worms for 4 h before measurement did not result in a significant increase (n.s.P > 0.05; n = 7). Bars represent mean ± SEM. (C) The antioxidant NAC abolishes the lifespan-extending effect of metformin (***P < 0.001; n ≥ 169 for each curve; Table S1). (D) prdx-2 is required for metformin-induced lifespan extension. Metformin treatment significantly reduces lifespan of prdx-2 mutants (***P < 0.001; n ≥ 127 for each curve; Table S1). (E) Metformin treatment promotes the formation of PRDX-2 disulfide dimers (**P < 0.01), implied to function in signal transduction. Bars represent mean ± SEM (n = 4). (F) PRDX-2 is required for metformin-induced phosphorylation of the p38 MAP kinase PMK-1. Metformin treatment of wild-type worms induced phosphorylation of PMK-1, inferred from a larger band observed on the Western blot. This metformin-mediated induction of PMK-1 phosphorylation is absent in prdx-2 knockout worms. Histone H3 levels were used as a loading control.
To verify whether mitochondrial ROS production was affected by metformin treatment, we measured hydrogen peroxide levels in metformin-treated worms. ROS production was indeed increased (Fig. 2B), further supporting the mitohormesis hypothesis. Induction of ROS production was already clear after 24 h of exposure to metformin (Fig. 2B). This increase in ROS seems to be an integral part of metformin-induced lifespan extension, because treatment with the potent antioxidants N-acetylcysteine (NAC) (13) and butylated hydroxyanisole (BHA) (26) abolished the positive effect of metformin on lifespan (Fig. 2C and Fig. S2A). Finally, the critical phase for metformin-mediated lifespan extension clearly resembles the critical phase for other mitohormetic stressors (13), because treatment with metformin starting from adulthood onward or during the first few days of adulthood only were sufficient to increase lifespan. This is opposed to treatment with metformin during larval development only, which had no detectable effect on lifespan (Fig. S2B).
The mitohormetic pathway as seen during glucose restriction is induced when low availability of glucose causes low energy levels, which in turn activates AMPK (13). AMPK activity increases catabolism and respiration, which results in the increased production of ROS and a resultant activation of hormetic protective mechanisms (13). The largest hiatus in this pathway is the step between the increased ROS production and the induction of stress defense: no molecule was put forward that might translate the ROS signal into further downstream defense. We therefore set out to reveal this missing link and to further complete the hormetic signaling pathway.
Peroxiredoxins are known hydrogen peroxide scavengers, and their oxidized dimeric form is involved in the direct activation of kinases in mammalian cells (27). Because the peroxiredoxin PRDX-2 was up-regulated during metformin treatment (Dataset S1), this protein is of particular interest as a potential inductor of mitohormesis. Deletion of the prdx-2 gene results in an extreme decrease in lifespan upon metformin treatment (Fig. 2D). Not only did the positive effect of metformin on lifespan disappear, the prdx-2 experimental group collapsed when exposed to metformin. Treatment with NAC partially rescued this deleterious effect, implying that excessive ROS production is at least partly responsible for the detrimental effect of metformin on these mutants (Fig. S2C). In support of these results, we observed increased formation of PRDX-2 dimers after metformin treatment (Fig. 2E and Fig. S3). Because these oxidized dimers are likely involved in cellular signaling (27), we propose that PRDX-2 induces prolongevity signaling during the mitohormetic response to metformin treatment.
One of the potential downstream targets of PRDX-2 is the p38 MAP kinase family-1 (PMK-1) protein, which is involved in the activation of the SKN-1 transcription factor (28). This transcription factor is in turn required for metformin-mediated longevity (5). Western blot analysis revealed a marked increase in phosphorylation of PMK-1 after metformin treatment. In contrast, deletion of prdx-2 resulted in an absence of metformin-induced phosphorylation of PMK-1 (Fig. 2F). These data strongly imply that PRDX-2 is required for PMK-1 activation after metformin treatment.
In sum, all these findings subscribe that metformin extends lifespan via mitohormesis in C. elegans and that PRDX-2 is an integral part of the mitohormetic pathway.
Metformin is generally believed to act through inhibition of complex I of the ETC (19⇓–21), although some recent findings cast doubt on this (16, 22). Treatment of C. elegans with rotenone, another complex I inhibitor, at a concentration that extends longevity, results in a decrease in total oxygen consumption (25). Our finding that metformin increases respiration in worms therefore raises the question of whether it is truly capable of inhibiting complex I of the ETC.
We tested whether metformin is able to affect electron flow in mitochondria extracted from C. elegans and observed a clear and specific inhibition of electron flow from complex I, whereas the electron flow from complex II was unaffected (Fig. 3A and Fig. S4 A and B). These results, clearly mimicking the inhibitory action of rotenone (Fig. S4A), complement our previous data only if metformin inhibits complex I in a distinct way. To this end, we tested whether metformin and rotenone had different effects on ROS production in mitochondria. At concentrations at which both completely inhibit complex I respiration and after feeding only complex I, metformin increased ROS production, whereas rotenone decreased it (Fig. 3B), implying a fundamental difference between rotenone’s and metformin’s inhibitory action on complex I.
Metformin inhibits complex I of the ETC in a way distinct from rotenone. (A) Metformin inhibits complex I- but not complex II-based respiration. Higher values for O2 flux indicate higher respiration, negative values indicate a small influx of O2, usually due to the injection of a compound. The effect of metformin on mitochondrial respiration was measured by sequentially adding compounds to stimulate different parts of the ETC. Green arrows indicate compounds that were added in both the control and metformin-treated cells; red arrows indicate compounds that were only added in the treated cell. After adding an equal volume of mitochondria (↓start), metformin was added to one of the cells (↓metformin). Subsequently, the complex I substrates pyruvate and malate (↓P+M) were added, followed by the addition of ADP (↓ADP), initiating electron transport from complex I. Metformin-treated mitochondria clearly fail to initiate complex I-based respiration, thus indicating that metformin inhibits electron transport from complex I in vitro. Finally, the complex II substrate succinate (↓S) was added to the metformin-treated cell, which led to a marked increase in mitochondrial respiration, indicating that metformin does not block electron transport from complex II. Two variations of this experiment were executed to also use the potent complex I inhibitor rotenone as a positive control (Fig. S4A) and to add succinate to the negative control condition as well (Fig. S4B). (B) Mitochondria treated with metformin produce H2O2 at a higher rate than untreated mitochondria (***P < 0.001; n = 3). Treatment with the complex I inhibitor rotenone had the opposite effect (***P < 0.001; n = 3).
The BCAAs leucine, isoleucine, and valine display a certain duality in relation to health and longevity. On one hand their effectiveness has been noted in the treatment of liver and cardiac diseases (29, 30), and up-regulation of BCAAs is one of the metabolic signatures of the long-lived daf-2 mutant (31, 32). On the other hand, BCAAs have also been causally linked to the development of insulin resistance, type II diabetes (33), and neuropathologies (34). We set out to confirm the marked up-regulation of the BCAA degradation pathway (Table 1 and Fig. S5) at the level of free BCAAs in metformin-treated worms. Metformin treatment resulted in a significant decrease (−55%) in the amount of free BCAAs, lending further support to the validity of the proteomics data (Fig. 4A).
Metformin induces the BCAA degradation and β-oxidation pathways but the β-oxidation enzyme ACDH-1 is not required for metformin-mediated longevity. (A) Metformin treatment stimulates the BCAA degradation pathway (Fig. S5) and in turn reduces the concentration of free BCAAs in C. elegans (**P < 0.01). Bars represent mean ± SEM (n = 6). (B) Metformin-treated worms show reduced fat storage (***P < 0.001; n = 30 for untreated and n = 33 for treated worms), possibly indicating increased flux through the β-oxidation pathway. Bars represent mean ± SEM. (C) Deletion of β-oxidation enzyme acdh-1 results in a proportionally larger effect of metformin on longevity (***P < 0.001; n ≥ 54 for each curve; Table S1).
One of the most striking properties of metformin as an antidiabetic in humans is its ability to cause weight loss in patients through, among others, activation of the β-oxidation pathway (2, 35, 36). We observed a 4.3-fold up-regulation of mitochondrial acyl-CoA dehydrogenase-1 (ACDH-1) after metformin treatment (Dataset S1), defining it as the strongest up-regulated protein by a margin. Acyl-CoA dehydrogenases catalyze the first step in the β-oxidation of fatty acids.
We questioned whether the shift in β-oxidation suggested by ACDH-1 induction truly occurs and whether it is involved in metformin-induced lifespan extension. If so, we would grossly expect overall fat levels to drop in treated worms. Measuring fat content in L4 worms, we found evidence for increased β-oxidation because metformin-treated worms showed significantly lower (−11.4%) fat stores (Fig. 4B). A similar reduction in fat content was found in metformin-treated C. elegans using different methods (5).
To our surprise, acdh-1 knockout worms showed an even stronger metformin-induced lifespan extension than wild-type worms (Fig. 4C). This implies a more complex interaction between β-oxidation and metformin (Discussion).
Aging worms start to show several morphological defects (37) and shrink in size (38), but metformin-treated worms seemed less affected by this phenomenon. After measuring several formalin-fixed worms of various ages (Fig. 5A) it became clear that nontreated worms started losing volume after day 6 of adulthood, whereas metformin-treated worms still retained their normal volume on day 9. Although long-lived mutants are often smaller than their wild-type siblings (39), our results fit the recent observations that within isogenic populations the larger animals are generally the longer-lived ones (38).
Metformin attenuates the morphological decline with aging in C. elegans. (A) Metformin-treated worms retain a stable volume, whereas control worms older than 6 d start shrinking. (B) Electron micrograph of the cuticle of a day-9 adult nontreated wild-type worm. Some deformations of the cuticle (seen as “wrinkling,” marked with an arrow) are starting to manifest. (C) No structural abnormalities can be seen in a metformin-treated day-9 wild-type adult.
Muscle and cuticle are known to show severe morphological defects with increasing age in C. elegans (37), making them prime targets to question whether these tissues are better maintained in metformin-treated worms. Our proteomics data already pointed to an increase in muscle mass and several changes in the cytoskeleton (Table 1), including up-regulation of intermediate filament proteins (Dataset S1). Electron micrographs of 9-d-old control and metformin-treated worms displayed no difference in body wall muscle volume, but there was a clear difference in cuticle morphology (Fig. 5 B and C). Whereas the cuticle of 9-d-old adult nontreated worms started showing signs of age-related “wrinkling” and disorganization, the cuticle of metformin-treated worms resembled that of a young animal. Considering the cuticle’s known role in maintaining the shape and size of nematodes (40), it can be assumed that it is this amelioration of cuticle deterioration that allows metformin-treated worms to retain a healthy, young morphology.
We profiled the effect of metformin in C. elegans using a differential proteomics approach and used the resulting data for further examination of its beneficial effects and mode of action.
We observed many changes in the mitochondrial proteome, which might point toward mitonuclear protein imbalance (41) and an altered mitochondrial metabolism. It is therefore probable for metformin to increase lifespan through both the process of mitohormesis and the induction of the UPRmt. However, metformin administration during the larval stages had no effect on lifespan, whereas the critical phase for UPRmt induction in C. elegans is during larval development (24, 42). Because metformin also proved unable to induce the UPRmt, it is not a likely in vivo contributor to lifespan extension by metformin in C. elegans. Contrary to the UPRmt, the mitohormetic pathway is important for metformin-mediated longevity in this worm.
Hormesis in aging is defined as the process by which a short-term and nonlethal stressor induces the stress response mechanisms of an organism and thereby increases stress resistance and overall life expectancy (43). The mitohormetic pathway, first described in glucose-restricted C. elegans (13), states that a low availability of ATP—due to low glycolytic activity—causes activation of AMPK. AMPK in turn promotes general catabolism and mitochondrial respiration, leading to increased production of ROS (the hormetic “stressor”), which subsequently act as messengers to activate further stress defenses, prolonging lifespan. This contradicts the classic oxidative stress theory of aging, which postulates that ROS, because of their ability to damage biomolecules, would be the causative factor of aging (44). We were able to fully reproduce the mitohormetic phenomenon in metformin-treated worms and showed that metformin increases respiration, metabolic heat production, and ROS generation in C. elegans. Inhibition of ROS signaling abolishes the lifespan-extending effects of metformin, proving that metformin-mediated lifespan extension is dependent on the mitohormetic pathway and most closely resembles the C. elegans response to glucose restriction.
Although well studied, some major missing links remained in the mitohormetic pathway, in particular how the ROS signal can be translated to increased lifespan. We were able to demonstrate that the C. elegans peroxiredoxin PRDX-2, previously shown to be involved in peroxide and heavy metal resistance (45), is of major importance to this function. Peroxiredoxins are a class of antioxidant proteins characterized by their high susceptibility to cysteine oxidation (46, 47). PRDX-2 is a typical 2-Cys peroxiredoxin (45), of which the active, oxidized form exists as a homodimer (48). It has recently been shown that these dimers are subsequently able to oxidize and activate specific substrates, such as the MAP kinase ASK1 (27), thus resolving how these proteins might be able to translate ROS signals into downstream signaling. In the same vein, our experiments revealed increased formation of PRDX-2 dimers after metformin treatment. These oxidized dimers might activate ASK1’s closest ortholog in C. elegans, the MAP kinase NSY-1, which functions upstream in the same signaling pathway as the MAP kinases SEK-1 and PMK-1 (49). This evolutionarily conserved MAP kinase cascade could ultimately activate several downstream targets that mediate stress defense, including SKN-1 (28). We showed that prdx-2 is required for the downstream activation of PMK-1 in response to metformin exposure, completing the pathway (Fig. 6). Therefore, PRDX-2 may well be the protein responsible for translating oxidative stress into longer lifespan in C. elegans. Because the longevity-promoting effect of exercise in humans also depends on ROS signaling (50), this pathway may be evolutionarily conserved.
The mitohormetic signaling cascade as induced by metformin. Metformin induces an increase in activity in several catabolic pathways (Fig. S6), including the TCA cycle and β-oxidation, with β-oxidation producing a relatively higher amount of FADH2 per cycle. This increase in substrate allows for an increase in mitochondrial respiration, which in turn leads to an increase in ROS production, possibly through metformin-mediated perturbation of electron transport (marked in red). These ROS oxidize PRDX-2 peroxiredoxins, which subsequently dimerize and enter their active state. Active PRDX-2 will activate a conserved MAPK cascade containing the p38 MAPK PMK-1, likely leading to the activation of SKN-1 (28) and a concomitant increase in longevity and stress protection (dashed arrows: literature-based evidence; Fig. S6 provides more information on the pathways induced by metformin treatment).
AMPK activation is one of the main factors involved in mitohormetic lifespan extension (13, 26). Although metformin is known to activate AMPK, exactly how this occurs remains elusive. The most common hypothesis is that metformin is able to partially inhibit complex I of the mitochondrial ETC, which in turn would lead to energy depletion and activation of AMPK (19⇓–21). Recently this has been debated (16, 22). One of the main arguments against metformin-mediated inhibition of complex I is the increased activity of pathways that generate energy after metformin treatment. All these pathways require NAD+ to function, and without efficient recycling of NADH to NAD+, these pathways are limited in activity. Because NADH is mainly recycled at complex I of the ETC, blocking complex I should lead to an increased NADH/NAD+ ratio and lower catabolism (Fig. 6). As a case in point, the complex I inhibitor rotenone indeed increases the NADH/NAD+ ratio (51) and lowers the activity of the β-oxidation pathway (16). We were able to show that metformin specifically inhibits complex I of the ETC in vitro, albeit in a way that seems fundamentally distinct from rotenone: metformin increased mitochondrial ROS production, whereas rotenone decreased it. High concentrations of rotenone reduce the number of electrons that are transferred from NADH to complex I (high NADH/NAD+ ratio), hence less electrons can leak out to form ROS. Our results imply that upon metformin treatment, NADH is still able to transfer its electrons, but these are subsequently lost, resulting in increased ROS production (Fig. 6). On the basis of these data, we suggest that metformin activates AMPK through inhibition of complex I (Fig. S6), which leads to an increase in respiration and a concomitant up-regulation of several catabolic pathways (e.g., β-oxidation, glycolysis, BCAA catabolism, and others) to provide the necessary substrate for the ETC. Likely exacerbated by metformin-mediated perturbation of electron transport, the increase in respiration ultimately leads to a mitohormetic increase in ROS production in vivo (Fig. 6). Because elevated respiration under physiologically normal conditions often leads to a reduction in ROS production (31), this indeed implies a direct involvement of metformin in the observed increase in ROS.
Treatment of C. elegans with metformin results in an overall up-regulation of several pathways and processes (including glycolysis and the TCA cycle), most of which are involved in catabolism, supporting the shift toward increased respiration. Two catabolic pathways stood out after metformin treatment: β-oxidation, through the short chain acyl-CoA dehydrogenase ACDH-1, and BCAA degradation. In light of metformin’s inhibitory action on complex I but not complex II of the ETC, the observed increase in β-oxidation seems a logical adaptation because it produces a relatively high amount of the complex II substrate FADH2 compared with other catabolic pathways. This might explain why knocking out acdh-1 adds to the lifespan increase induced by metformin: high activity of the β-oxidation pathway could temporarily increase ATP levels and inactivate AMPK (Fig. 6 and Fig. S6). As such, deletion of acdh-1 could cause a more stringent activation of the mitohormetic pathway. Further experiments will be needed to fully explore the role of β-oxidation in metformin-induced lifespan extension. Our results suggest that the metformin-mediated increase in β-oxidation is likely a compensatory mechanism that is unrelated to metformin-induced longevity.
As for the BCAA degradation pathway, no clear singular correlation between BCAAs and longevity has yet been found in any organism. Although some evidence points toward BCAAs as a metabolic signature of long life in C. elegans insulin receptor mutants (31, 32), other studies have causally linked BCAAs to the development of insulin resistance, diabetes (33), and neuropathologies (34). Interestingly, BCAAs, leucine in particular, are potent activators of the target of rapamycin (TOR) kinase. Leucine deprivation has previously been associated with reduced TOR signaling (52), which can in turn prolong lifespan and is required for DR-mediated lifespan extension (53). However, it remains unclear whether the drop in BCAA levels is necessary for metformin-induced lifespan extension. The reduction in BCAA concentration might result in lower RNA translation into protein, both through inactivation of TOR and a reduction in the amount of substrate necessary for translation (Fig. S6). This in turn may lead to an increase in longevity. There is some precedence for this hypothesis: amino acid imbalance has been associated with increased longevity in Drosophila melanogaster (54), among others (55). The recently described bacteria-specific effect of metformin on longevity in C. elegans (6) might similarly depend on an amino acid imbalance that occurs through the lowered production of methionine in metformin-treated bacteria (Fig. S6). Our results suggest a hitherto unexplored role for the BCAA degradation pathway in longevity.
Metformin not only increases lifespan but also healthspan of C. elegans: treated worms retain a youthful morphology for a longer time. How metformin is able to attenuate morphological decline of the cuticle in C. elegans remains elusive, but the intermediate filament protein IFC-2 (up-regulated after metformin treatment; Dataset S1) seems particularly promising because it is required for normal body shape and cuticle strength (56). Integrity of the cuticle and epidermis might play a more important role in longevity than is generally thought. Deterioration in cuticle structure leads to a loss in barrier function, which may be one of the causes of death of older nematodes (37, 40).
In conclusion, this work reveals insights in the process of aging and shows that metformin extends lifespan through mitohormesis. A missing link in the mitohormesis pathway in C. elegans has now been assigned: the peroxiredoxin PRDX-2, a protein that translates a ROS signal into a prolongevity cue. Because peroxiredoxin signaling is evolutionarily conserved (27), peroxiredoxins might hold a similar function in humans.
The following strains were obtained from the Caenorhabditis Genetics Center (University of Minnesota): wild-type N2, VC289 prdx-2(gk169), VC1011 acdh-1(ok1489), and SJ4100 zcIs13[hsp-6::GFP]. GA507 glp-4(bn2ts) daf-16(mgDf50) was provided by the Gems laboratory. Strains were cultivated on standard nematode growth medium (NGM) seeded with Escherichia coli OP50. All strains were outcrossed at least four times, with the exception of SJ4100, which was outcrossed three times. The outcrossed prdx-2(gk169) and acdh-1(ok1489) strains were renamed LSC555 and LSC556, respectively. All experiments were performed at 20 °C unless stated otherwise.
The protein samples for two-dimensional difference gel electrophoresis (2D-DIGE) were taken from glp-4(bn2ts) daf-16(mgDf50) worms grown in liquid cultures. The glp-4(bn2ts) mutation confers sterility at the permissive temperature of 24 °C, avoiding contamination with progeny. Preventing germ-line development also removes abundant contaminating proteins that have no bearing on lifespan, facilitating the analyses. Because glp-4 mutations cause a slight DAF-16–dependent lifespan increase in C. elegans cultured on dead E. coli (57), glp-4(bn2ts) daf-16(mgDf50) double mutants were used. This does not interfere with the envisaged results, because lifespan extension due to metformin is independent of DAF-16 (5). Cultivating the worms in liquid cultures allowed full control over metformin dosage while ensuring that all worms were fully fed. Additionally, the high-density samples obtained from liquid cultures ensure high protein concentration in 2D-DIGE experiments.
Worms were synchronized by isolating eggs from gravid adults through hypochlorite and NaOH treatments (58) and a subsequent sucrose density centrifugation to separate eggs from dead worms and bacterial debris. L1 worms were added to Fernbach flasks containing S medium (59) and constantly shaken at 24 °C. Flash frozen E. coli K12 bacteria (Artechno) were added to the cultures as food source. The concentration of bacteria was checked twice daily, and new K12 bacteria were added accordingly to maintain the cultures at optimal food levels (OD550 = 1.8). The density of worms in the culture never exceeded 1,000 worms/mL to prevent large fluctuations in food availability. When worms reached the L4 stage, cultures were supplemented with 2′-deoxy-5-fluorouridine (FUdR; Sigma-Aldrich) at a final concentration of 100 mg/L. This ensures complete sterility, because glp-4(bn2ts) daf-16(mgDf50) worms rarely manage to produce a few eggs. Once worms reached the adult stage, the test group was exposed to 25 mM of metformin (1,1-dimethylbiguanide hydrochloride; Sigma-Aldrich) for 24 h (SI Materials and Methods describes the determination of the optimal metformin concentration in liquid media; Fig. S7), and protein samples were subsequently taken (SI Materials and Methods provides a complete protocol on protein extraction).
Samples were labeled, separated, and analyzed as described by Bogaerts et al. (60). In short, metformin-treated and nontreated protein extracts were differentially labeled with either Cy3 or Cy5 fluorescent dyes (GE Healthcare). A possible dye bias was taken into account by integrating a dye swap into the experimental design. Differentially labeled samples were pooled, an internal standard labeled with Cy2 was added, and the pooled samples were separated in two dimensions. Gels were scanned using an Ettan DIGE Imager (GE Healthcare), and DeCyder 7.0 (GE Healthcare) was used to statistically analyze the images. Spot intensity was compared using a standard Student t test with false positive rate correction.
Differential spots were excised using an automated spotpicker (GE Healthcare). The proteins in each spot were digested using a standard in-gel trypsin digestion protocol and subsequently identified using peptide mass fingerprinting (SI Materials and Methods provides a complete protocol).
Wormbase gene IDs of differential proteins were uploaded to the bioinformatic tool DAVID (23) to look for enrichment in functional clusters. Functional clustering was performed with high stringency. Pathway enrichment was determined using DAVID by querying the Kyoto Encyclopedia of Genes and Genomes.
All lifespan experiments were performed on NGM agar plates or NGM plates supplemented with or without 50 mM of metformin. NAC was used at a concentration of 5 mM and BHA at a concentration of 25 µM, when applicable. For the experiment using BHA, the compound was added from a 1,000× concentrated stock in DMSO, and an equal volume of DMSO was added to the respective controls. The worms used in the BHA lifespan experiment were pretreated with 25 µM of BHA for one generation before the start of the experiment. For lifespan experiments, 20 L4 stage worms from a synchronous culture were transferred to a plate (either regular NGM or NGM containing one or multiple of the tested compounds), allowed to lay eggs for 24 h, and were then removed from the plate. The progeny was transferred to a new plate when they reached the young adult stage and were transferred to new plates during each day of their reproductive period and every 3 d thereafter. Lifespan was monitored every day starting from day 1 of the adult stage: animals that did not move when gently prodded were marked as dead. Animals that crawled off the plate or died of vulval bursting or internal hatching were censored. Survival curves were statistically analyzed using a Cox proportional hazard (to compare the effect of metformin between strains and/or conditions) or log–rank test (to compare a single curve to control; corrected for multiple testing using the Benjamini-Hochberg method). Complete information regarding lifespan experiments can be found in the SI (Table S1).
Synchronized L1 worms of the SJ4100 zcIs13[hsp-6::GFP] strain were transferred to agar plates containing 0 mM, 50 mM, or 100 mM of added metformin. As a positive control, worms were transferred to plates seeded with HT115 Escherichia coli containing an RNAi construct targeted against cco-1 (24). To rule out a specific effect of HT115, plates seeded with HT115 containing an empty vector were used, which caused no fluorescence. Images were acquired using a Zeiss Axio Imager Z1 microscope.
Metabolic activity and respiration measurements were carried out as previously described (61). In short, respiration and metabolic heat production of synchronized day-1 adult worms were measured using a Clark-type electrode respirometer (Strathkelvin) and a Thermal Activity Monitor (Thermometric), respectively. Treated worms were exposed to 50 mM metformin starting from the L1 stage. Oxygen consumption rates were measured over a span of 15 min. After thermal equilibration of the calorimeter, live heat output of each sample was averaged over a 3-h period. All data were normalized according to the protein concentration in each sample, because this directly correlates to worm biomass. Protein concentrations were determined using a bicinchoninic acid kit (Thermo Scientific). Student t tests were used to statistically analyze the data.
Endogenous hydrogen peroxide production was quantified using the Amplex Red hydrogen peroxide kit (Invitrogen, catalog no. A22188). Adult worms were washed three times with S basal (59), after which the worms were pelleted through centrifugation (800 × g, 3 min). From this dense pellet, 50 µL of worms were transferred to a new microcentrifuge tube containing 450 µL of reaction buffer (sodium phosphate buffer supplied with the kit). After centrifugation (800 × g, 1.5 min), the reaction buffer was replaced with fresh reaction buffer, and 500 µL of Amplex Red working solution (prepared according to the manufacturer’s instructions) was added. The worms were allowed to rotate in the dark for 1 h, after which they were pelleted through centrifugation (800 × g, 1.5 min), and 100 µL of supernatant was transferred to a fluorescence-compatible 96-well plate. Fluorescence (Ex 550 nm, Em 590 nm) was measured, and the worm pellet was subjected to protein extraction for normalization. Welch’s t tests with multiple testing correction (Benjamini-Hochberg) were used to statistically analyze the heteroscedastic data.
Mitochondria were extracted from worms as previously described (62). Oxygen consumption of the extracted mitochondria was determined using a Clark-type electrode (Oxygraph 2k) and DatLab software (Oroboros Instruments). An aliquot of 25 µL of mitochondria was incubated in 2 mL of air-saturated MiR05 (0.5 mM EGTA, 3 mM MgCl2, 60 mM K-lactobionate, 20 mM taurine, 10 mM KH2PO4, 20 mM Hepes, 110 mM sucrose, and 1 g/L BSA, pH 7.1) at 20 °C. The effect of 25 mM of metformin on electron transport from complex I and complex II was measured as follows: (i) metformin was added from a 1.5-M stock (dissolved in MiR05) until a final concentration of 25 mM was reached; (ii) the complex I substrates pyruvate and malate were added from a 1-M stock until a final concentration of 5 mM was reached; (iii) ADP was added from a 10-mM stock to initiate complex I respiration (final concentration 25 µM); (iv) when applicable, rotenone was added as a positive control at a final concentration of 1 µg/mL from a 400-µg/mL stock; (v) to initiate complex II respiration, succinate was added from a 1-M stock until a final concentration of 5 mM was reached.
Mitochondrial hydrogen peroxide production was quantified using a protocol based on the Amplex Red hydrogen peroxide kit (Invitrogen, catalog no. A22188). A fluorescence-compatible 96-well plate was prepared, and several wells were filled with 96 µL of incubation medium (4 mM ADP, 10 mM pyruvate, 10 mM malate, 10 U Cu/ZnSOD) either containing no additional substance, 25 mM of metformin, or 1 µg/mL of rotenone. Subsequently, 100 µL of Amplex Red reaction buffer (100 µM Amplex Red, 4 U/mL horseradish peroxidase dissolved in the sodium phosphate buffer supplied with the kit) was added, followed by the addition of 4 µL of freshly extracted mitochondria. The plate was shaken in the dark, and every 2 min fluorescence (Ex 550 nm, Em 590 nm) was measured over the span of 1 h to accurately plot the increase in fluorescence over time resulting from hydrogen peroxide production. The resulting curves were statistically analyzed using analysis of covariance.
A nonreducing anti-PRDX-2 Western blot was performed as described by Oláhová et al. (45). In short, synchronized L1 worms were grown on NGM plates with or without 50 mM of added metformin. Proteins were extracted from young adult worms and separated on a nonreducing SDS/PAGE gel. Rabbit anti-PRDX-2 antibodies were kindly provided by Elizabeth Veal (Newcastle University, Newcastle upon Tyne, UK). Intensity of PRDX-2 bands was normalized against total protein content of each lane, which was visualized using a Deep Purple (GE Healthcare) total protein stain. A Welch’s t test was performed to analyze the data.
N2 and prdx-2 worms were synchronized and grown on regular NGM plates. Protein samples of day-1 adult worms were taken as described for the 2D-DIGE experiment. Proteins were separated on an SDS/PAGE gel, and rabbit anti-phospho-p38 antibodies were used for visualization (Cell Signaling Technology). Blots were reincubated with rabbit anti-histone H3 antibodies (Abcam) as a loading control (63).
Synchronized L1 worms were cultivated on NGM plates with or without 50 mM of added metformin. L4 stage worms were harvested and stained with Oil red O. The staining intensity was acquired for at least 30 worms for each condition (SI Materials and Methods). A Student t test was performed to analyze the data.
Wild-type worms were synchronized through hypochlorite treatment, and L1 worms were transferred to normal NGM plates or NGM plates supplemented with 50 mM of metformin. FUdR was added to the plates when worms reached the L4 stage at a final concentration of 50 µM. For sampling, worms were washed twice with S basal and subsequently fixed in 4% (wt/vol) formaldehyde (3 min at 65 °C). Images of the formaldehyde fixed worms were captured, and worm width (D) and length (L) were analyzed using a RapidVUE Particle Shape and Size Analyzer (Beckman Coulter). Worm volume was approximated using the cylinder formula [V = πL(D/2)2].
Wild-type worms were synchronized and sterilized as described for the volume measurements. Worms were harvested when they reached day 9 of adulthood and prepared for electron microscopy (SI Materials and Methods).
Wild-type worms were synchronized, and L4 worms were harvested from the plates. BCAA concentration was measured using a BCAA assay kit (Sigma-Aldrich, catalog no. MAK003) according to manufacturer’s instructions. Total protein content of the extracts was quantified using the Qubit Protein Assay (Invitrogen) and used for normalization. A Welch’s t test was performed to analyze the data.
All statistical analyses, apart from the DeCyder and DAVID analysis, were carried out using R (64). All bar graphs show the mean of biologically independent samples, and error bars show SEM. P values <0.05 were considered significant.
We thank Dr. Elizabeth Veal for providing the anti-PRDX-2 antibodies; Prof. Dr. David Gems for providing the GA507 strain; and Sven Van Bael, An Vandoren, Caroline Vlaeminck, and Ineke Dhondt for technical assistance. Several strains were provided by the Caenorhabditis Genetics Center, which is funded by National Institutes of Health Office of Research Infrastructure Programs Grant P40 OD010440. This work was supported by a grant from the Fund for Scientific Research–Flanders (Fonds voor Wetenschappelijk Onderzoek Vlaanderen; FWO-Flanders) (G.04371.0N). L.T. is an FWO-Flanders postdoctoral fellow.
Author contributions: W.D.H., B.P.B., L.S., and L.T. designed research; W.D.H., L.F., R.V.A., A.S., G.D., and J.B. performed research; W.D.H., L.F., R.V.A., A.S., and G.D. analyzed data; and W.D.H. and L.T. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission. A.D. is a guest editor invited by the Editorial Board.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1321776111/-/DCSupplemental.
Freely available online through the PNAS open access option.
