Lifespan extension by peroxidase and dual oxidase-mediated ROS signaling through pyrroloquinoline quinone in C. elegans

The free radical, or oxidative stress, theory of aging proposed that reactive oxygen species (ROS), produced as byproducts of metabolism, contribute to aging and certain pathologies by damaging lipids, proteins and DNA (Harman, 1956, 1972). It has since become apparent that ROS also serve as signal molecules to regulate various biological processes, such as cell proliferation, differentiation, inflammation and regeneration (Finkel, 2011; Reczek and Chandel, 2015; Schieber and Chandel, 2014). The classic oxidative stress theory has been a broadly accepted explanation for aging and is the theoretical backbone for using antioxidants, such as vitamins and N-acetyl cysteine (NAC), as anti-aging supplements.

However, ROS at low levels are now recognized as important signal molecules for promoting longevity. Recent reports suggested that ROS from mitochondria, known as mtROS, extend lifespan (Kawagishi and Finkel, 2014; Lee et al., 2010; Reczek and Chandel, 2015; Ristow, 2014; Wang and Hekimi, 2015; Yang and Hekimi, 2010a; Yun and Finkel, 2014). Various stimuli, such as mitochondrial dysfunction, calorie restriction, acute impairment of insulin signaling, glucose restriction or absence of a germline, can release mtROS to extend lifespan (Lee et al., 2010; Schaar et al., 2015; Schulz et al., 2007; Wei and Kenyon, 2016; Xie and Roy, 2012; Yang and Hekimi, 2010a,b; Zarse et al., 2012). The hypoxia inducible factor 1 (HIF-1), the stress-activated transcriptional factor SKN-1 (Nrf2 or NFE2L2 in mammals), AMP-activated protein kinase (AMPK) and molecules in the intrinsic apoptosis pathway including CED-4 (Apaf1 in mammals) also participate in mtROS-mediated longevity signaling (Castillo-Quan et al., 2016; Hwang et al., 2014; Lee et al., 2010; Wei and Kenyon, 2016; Weimer et al., 2014; Xie and Roy, 2012; Yee et al., 2014; Zarse et al., 2012).

Dual oxidases (DUOXs) are NADPH oxidases (NOXs) that produce the ROS H₂O₂ at the plasma membrane (Bedard and Krause, 2007; De Deken et al., 2014, 2000; Donkó et al., 2005; Dupuy et al., 1999; Sumimoto, 2008). H₂O₂ generated by DUOXs plays important roles in multiple biological processes such as thyroid-hormone biosynthesis, tyrosine-crosslinking in the cuticle and fertilization envelope, innate immunity and wound healing (Chávez et al., 2009; De Deken et al., 2014, 2000; Dupuy et al., 1999; Edens et al., 2001; Fortunato et al., 2010; Grasberger et al., 2007; Ha et al., 2005; van der Hoeven et al., 2011; Kumar et al., 2010; Lipinski et al., 2009; Moribe et al., 2012; Moribe and Mekada, 2013; Niethammer et al., 2009; Song et al., 2010; van der Hoeven et al., 2015; Wong et al., 2004). Because of its crucial role in gut immunity, DUOXs are likely to be involved in lifespan regulation (van der Hoeven et al., 2015). Very recently, ROS produced by the C. elegans DUOX BLI-3 was found to extend lifespan in C. elegans (Ewald et al., 2017).

Pyrroloquinoline quinone (PQQ) is a polyphenolic compound with a quinone group and the third redox cofactor, after nicotinamide and flavin, identified in bacterial dehydrogenases (Anthony and Zatman, 1967; Hauge, 1964; Ikemoto et al., 2012; Salisbury et al., 1979). PQQ was shown to have anti-oxidative effects and stimulate various cellular signaling and defensive responses, such as neuro- and cardio-protection (Akagawa et al., 2015; Rucker et al., 2009). Classic studies revealed the importance of PQQ in nutrition. Mice fed PQQ-deficient food grow poorly, failed to reproduce and had bone and skin abnormalities, similar to those observed in early aging mice, such as alpha-klotho mutants (Killgore et al., 1989; Kuro-o et al., 1997; Rucker et al., 2009). Consistent with these reports, PQQ extended lifespan and increased oxidative stress defensive responses in C. elegans (Wu et al., 2016). However, the mechanisms through which PQQ mediates lifespan extension are largely unknown.

In this study, we showed that PQQ activates BLI-3 to generate H₂O₂, which confers longevity on C. elegans. We demonstrated a molecular trick for generating low levels of ROS at the plasma membrane, in which H₂O₂ generated by BLI-3 was subsequently degraded by its coupled peroxidase (MLT-7) to achieve the low levels. We propose that such fine-tuning of ROS at the plasma membrane is essential for regulation of longevity.

Before investigating mechanisms underlying PQQ-mediated lifespan extension, we characterized the general effects of PQQ on lifespan (Fig. 1A,C; Fig. S1C–I, Table S1). PQQ extended the lifespan of wild-type C. elegans adults in a dose-dependent manner (Fig. 1C; Table S1) (Wu et al., 2016). PQQ was most effective at 5 mM when administered during adult stages, with mean values for lifespan, the age at 90% survival and age at 10% survival increased by 31, 55 and 13%, respectively (Table S1). These results indicate that PQQ strongly prevented early death and extended lifespan. Imidazolopyrroloquinoline (IPQ), a PQQ derivative lacking the quinone structure (Fig. 1B; Fig. S1A,B), did not extend lifespan (Fig. 1D; Table S1). This suggests that lifespan extension was caused specifically by PQQ.

Unlike with 1, 3 and 5 mM PQQ, a lower concentration (0.1 mM) was ineffective, while higher concentrations (10, 15 and 20 mM) decreased lifespan in a dose-dependent manner (Fig. S1C,D; Table S1). These results are consistent with PQQ acting as an antioxidant, with an optimal concentration range being more pharmacologically effective than lower or higher concentrations (Wu et al., 2016).

Experiments administering PQQ during different periods of life suggest that PQQ treatment during the adult stage was necessary and sufficient for prolonging lifespan. PQQ treatment from adult days (AD) 1 to 10, and especially AD 1 to 5, was important for lifespan extension (Fig. S1E–H; Table S1).

Using UV-killed E. coli as a food source, we eliminated the possibility that PQQ affected live E. coli, thereby indirectly inducing lifespan extension. PQQ-treated wild-type animals lived longer than control animals on UV-killed E. coli (Fig. S1I; Table S1), suggesting that PQQ directly affects C. elegans to extend lifespan. Previous studies have reported that wild-type animals on UV-killed E. coli lived longer than on live E. coli, probably due to the metabolic change into calorie restriction mode by recognizing dead food through a chemical sense (Apfeld and Kenyon, 1999; Garigan et al., 2002). Consistent with theses reports, wild-type animals lived longer on UV-killed E. coli in our lifespan experiments.

To address whether PQQ affects other physiological activities, such as behaviors, we investigated the pharyngeal pumping and body bends as locomotion activity at AD 3, 6, 9 and 12 of wild-type animals with and without 5 mM PQQ. We found that PQQ did not significantly affect these behaviors (Fig. S1J,K).

To clarify the molecular mechanism of PQQ-mediated lifespan extension, we undertook a molecular genetic approach by searching for mutants insensitive to PQQ. Insulin and IGF-1 signaling (IIS) plays a central role in lifespan regulation across species (Antebi, 2007; Gems and de la Guardia, 2013; Kenyon, 2005, 2010; Kim, 2013; Shore and Ruvkun, 2013). Reduction-of-function mutations in daf-2, which encodes the sole insulin and IGF-1 receptor (Kenyon et al., 1993; Kimura et al., 1997), and age-1 encoding phoshatidylinositol 3-kinase, which acts at downstream of daf-2 (Johnson, 1990; Morris et al., 1996) extend lifespan. DAF-16 (a FOXO protein) is a central transcriptional factor regulating IIS-mediated lifespan in C. elegans, since the prolonged lifespans of daf-2 and age-1 mutants are suppressed by mutations in daf-16 (Dorman et al., 1995; Kenyon et al., 1993; Lin et al., 1997; Ogg et al., 1997).

To address whether PQQ-induced lifespan extension is regulated by IIS, we examined the effect of PQQ on daf-2, age-1, and daf-16 mutants (Fig. S2A–E; Table S2). Lifespans of daf-2(e1370) and age-1(hx546) longevity mutants were further extended with the administration of PQQ (Fig. S2A,C; Table S2). A particular reduction-of-function allele of daf-2(e1368) did not induce PQQ-mediated lifespan extension (Fig. S2B; Table S2). Different results between daf-2(e1370) and daf-2(e1368) may be attributed to the different mutation sites in the DAF-2 receptor (Gems et al., 1998; Patel et al., 2008) (see Discussion). Lifespans of daf-16(mu86) loss-of-function mutants and daf-16(m26) reduction-of-function mutants were extended by PQQ (Fig. S2D,E; Table S2). However, the degrees of lifespan extension in daf-16 mutants were smaller than those in control wild-type animals (Fig. S2D and E; Table S2). Taken together, the results with daf-2(e1368) and daf-16(mu86), daf-16(m26) mutants imply that PQQ-mediated lifespan extension may be partially dependent on IIS (see Discussion).

We also tested other mutants known to regulate lifespan. mtROS represents the low level of ROS released from the mitochondria in response to several stimuli and can extend lifespan (Kawagishi and Finkel, 2014; Ristow, 2014; Wang and Hekimi, 2015; Wei and Kenyon, 2016; Yang and Hekimi, 2010a,b; Zarse et al., 2012). The intrinsic apoptosis factor CED-4 and the hypoxia inducible factor HIF-1 are reported to mediate mtROS signaling (Feng et al., 2001; Hwang et al., 2014; Lee et al., 2010; Xie and Roy, 2012; Yang and Hekimi, 2010b). The lifespans of ced-4(n1162) and hif-1(ia4) reduction-of-function mutants were extended by PQQ (Fig. S2F,G; Table S2), suggesting that CED-4- and HIF-1-mediated mtROS signaling were not directly involved in PQQ-mediated lifespan extension. The eat-2 reduction-of-function mutants are defective in food intake (McKay et al., 2004) and considered to be a model of calorie restriction (Lakowski and Hekimi, 1998). We showed that PQQ extended the lifespan of eat-2 mutants (Fig. S2H; Table S2), suggesting that the calorie restriction is not a major cause for PQQ-mediated lifespan extension.

Because PQQ regulates the activity of redox enzymes (Akagawa et al., 2015; Anthony and Zatman, 1967; Hauge, 1964; Rucker et al., 2009; Salisbury et al., 1979), we focused on BLI-3, a C. elegans ortholog of the mammalian dual oxidases, DUOX1 and DUOX2 (Brenner, 1974; Edens et al., 2001). PQQ did not extend the lifespans of three reduction-of-function bli-3 mutants, bli-3(im10), bli-3(e767) and bli-3(n529) (Fig. 2A; Fig. S3A,B, Table S2), suggesting that BLI-3 is required for PQQ-induced longevity.

Dual oxidases are NADPH oxidases that produce the ROS H₂O₂ at the plasma membrane (Bedard and Krause, 2007; De Deken et al., 2014, 2000; Donkó et al., 2005; Dupuy et al., 1999; Sumimoto, 2008) (Fig. S3C–E). BLI-3 is expressed in intestine and hypodermis, and BLI-3-dependent H₂O₂ is required for cuticle formation and innate immunity in C. elegans (Chávez et al., 2009; Edens et al., 2001; van der Hoeven et al., 2011, 2015; Moribe and Mekada, 2013). Previous studies showed that ROS production from BLI-3 requires tetraspanin (TSP-15) and dual oxidase maturation factor (DOXA-1), both of which are also present at the plasma membrane (Charrin et al., 2014; Hemler, 2005; Moribe et al., 2012, 2004; Moribe and Mekada, 2013; Xu et al., 2015) (Fig. S3E). PQQ did not extend the lifespan of the tsp-15(sv15) reduction-of-function mutant (Fig. 2B and Table S2), suggesting that TSP-15 is critical for PQQ-mediated lifespan extension. We further confirmed in rescue experiments that bli-3 and tsp-15 genes were responsible for PQQ-induced lifespan extension (Fig. 2C,D; Table S2).

Heterologous expression of BLI-3, DOXA-1 and TSP-15 in human HT1080 cells induced H₂O₂ production (Moribe et al., 2012) (Fig. 3A; Fig. S3E). Using this system, we investigated whether PQQ directly altered H₂O₂ production. PQQ dramatically enhanced H₂O₂ production in a dose-dependent manner (Fig. 3A). In contrast, IPQ had only a limited effect on H₂O₂ production, consistent with its minimal influence on lifespan (Fig. 1D; Table S1). PQQ-induced elevated H₂O₂ production was dependent on the NOX enzymatic activity of BLI-3, based on its complete suppression by the NOX inhibitor diphenyleneiodonium (DPI) (Fig. 3A). We examined H₂O₂ production in cell lines expressing a mutant bli-3 gene (Fig. 3B; Fig. S3C,D). H₂O₂ production was markedly attenuated in these cells, further suggesting that BLI-3 is responsible for PQQ-induced elevated H₂O₂ production. Consistent with the catalytic effect of PQQ on BLI-3 activity, H₂O₂ production increased in an exponential manner over time (Fig. 3C). PQQ treatment did not alter protein expression levels of BLI-3, DOXA-1 or TSP-15 at the plasma membrane (Fig. 3D). Thus, we conclude that PQQ catalytically activated BLI-3 to accelerate H₂O₂ production in this heterologous expression system. Similar to BLI-3, human DUOX1 and DUOX2 were enzymatically activated by PQQ at the plasma membrane (Fig. 3D,E), suggesting conservation of PQQ-mediated DUOX activation across species.

We postulated that if ROS produced by BLI-3 were signals for lifespan extension in vivo, elimination of ROS with antioxidants might inhibit lifespan extension, and conversely overproduction of ROS might extend lifespan. Indeed, PQQ-induced lifespan extension was abolished upon treatment with the antioxidant NAC (Benrahmoune et al., 2000; Desjardins et al., 2017) (Fig. 4A; Fig. S4A, Table S2). Furthermore, transgenic strains overexpressing bli-3, doxa-1 and tsp-15 (these DNAs were introduced at 1 ng/µl each) had slightly longer lifespans than wild-type animals (Fig. 4B; Fig. S4B; Table S2), and strains overexpressing bli-3, doxa-1 and tsp-15 (introduced at 10 or 25 ng/µl each) had significantly longer lifespans than wild-type animals without PQQ, in accordance with transgene dosage (Fig. 4C,D, Fig. S4C–E; Table S2). These results suggest that BLI-3-dependent ROS are acting as signals for prolonging lifespan in vivo.

When strains overexpressing bli-3, doxa-1 and tsp-15 (introduced at 25 ng/µl each) were treated with 5 mM PQQ, they had shorter lifespans than untreated animals, suggesting that the high level of ROS adversely affected lifespan (Fig. 4D; Fig. S4D; Table S2). Consistent with this result, PQQ concentrations higher than 10 mM shortened the lifespan of wild-type animals (Fig. S1D; Table S1). Previous studies showed that iodide induced a hyperproduction of BLI-3-dependent ROS, leading to death (Xu et al., 2015). Thus, we hypothesized that low levels of ROS generated by BLI-3 induced lifespan extension through the effect of hormesis (Cypser and Johnson, 2001; Gems and Partridge, 2008; Kourtis and Tavernarakis, 2011; Ludovico and Burhans, 2014), while excess levels of ROS are harmful, resulting in a short-lived phenotype.

To test this hypothesis, we analyzed the effects of PQQ in mlt-7 reduction-of-function mutants (Fig. 5A; Table S2). During collagen tyrosine-crosslinking in the extracellular matrix, MLT-7 peroxidase coupled to BLI-3 converts BLI-3-generated H₂O₂ into H₂O (Moribe et al., 2012; Moribe and Mekada, 2013; Thein et al., 2009) (Fig. S3E). Therefore, the mlt-7 mutants incapable of degrading BLI-3-generated H₂O₂ may have relatively high H₂O₂ levels. Similar to the strains overexpressing bli-3, doxa-1 and tsp-15, mlt-7(im39) mutants lived longer than wild-type animals, even without PQQ (Fig. 5A; Table S2). In fact, PQQ treatment decreased the lifespan of mlt-7 mutants in a dose-dependent manner (Fig. 5A; Table S2). These results indicate that an adequate level of ROS is maintained in mlt-7 mutants inducing longevity in the absence of PQQ, whereas the higher level of PQQ-induced ROS in mlt-7 mutants shortens their lifespans (Fig. 5B–F). These results suggest that fine-tuned ROS levels regulated by the peroxidase and dual oxidase (MLT-7 and BLI-3) system are essential for lifespan regulation.

SKN-1 (Nrf2) is the central transcriptional factor for oxidative stress responses and longevity across species (An and Blackwell, 2003; Blackwell et al., 2015; Inoue et al., 2005; Kubben et al., 2016; Maher and Yamamoto, 2010; Park et al., 2009; Tullet et al., 2008). Some studies showed that SKN-1 transduces BLI-3-dependent the ROS signal during innate immunity and that the impairment of a negative regulator of BLI-3 activates SKN-1, thereby conferring longevity (Ewald et al., 2017; van der Hoeven et al., 2011), and another study showed that PQQ-mediated lifespan extension requires both SKN-1 and DAF-16 (Wu et al., 2016). We thus addressed whether SKN-1 is involved in PQQ-mediated lifespan extension in our assay system. Two alleles of skn-1 reduction-of-function mutants, skn-1(zu67) and skn-1(ok2315), completely abolished PQQ-mediated lifespan extension (Fig. 6A,B; Table S2). Consistent with the previous report (Wu et al., 2016), these results suggest that SKN-1 is a crucial transcriptional factor for PQQ-mediated lifespan extension.

In mammals, c-JUN and c-FOS constitute the AP-1 transcription factor complex that can be activated by various kinds of stress including oxidative stress (Eferl and Wagner, 2003). In C. elegans, JUN-1 and FOS-1 constitute the AP-1 complex and it plays an important role in stress responses (Hattori et al., 2013; Hiatt et al., 2009). JUN-1 is also required for the lifespan extension induced by intermittent fasting (Honjoh et al., 2009; Uno et al., 2013). We tested the effect of PQQ on jun-1(gk557) loss-of-function mutants. Similar to mlt-7 loss-of-function mutants, jun-1(gk557) mutants died earlier with 5 mM PQQ than without PQQ (Fig. 6C; Table S2). However, jun-1 mutants were more short-lived than wild-type animals without PQQ, and this phenotype of jun-1 mutants is different from long-lived phenotype of mlt-7 mutants without PQQ (Fig. 6C; Table S2). These results imply that JUN-1 is involved in PQQ-mediated ROS signaling, although its action on the lifespan extension is thought to be rather complex. The requirement for the stress-responsive transcriptional factors SKN-1 and JUN-1 suggest that the fine-tuned ROS at plasma membrane strengthen the anti-stress network in the body, thereby contributing to longevity.

In this study, we showed that a highly specific mechanism of promoting longevity, in which ROS generation confined to the plasma membrane can regulate lifespan. Our results are consistent with the model that PQQ activates the membrane-associated protein BLI-3 to generate ROS, which are, in cooperation with the peroxidase MLT-7, fine-tuned to a low level, thereby leading to lifespan extension through SKN-1 and JUN-1 by virtue of hormesis (Fig. 7).

We demonstrated that the low levels of ROS generated by the peroxidase and dual oxidase system induce longevity. Since the administration of high concentrations of PQQ (10, 15 and 20 mM) instead decreased the lifespan of wild-type animals (Fig. S1D, Table S1), and the administration of 5 and 10 mM PQQ drastically decreased lifespan of mlt-7(im39) mutants (Fig. 5A; Table S2), high levels of BLI-3-dependent ROS are disadvantageous for the animals. Thus, regulation of ROS levels at the plasma membrane orchestrated by the peroxidase and dual oxidase system is likely to be critical for lifespan. We speculate that DUOX-mediated ROS would be elicited by the activation of ROS generation mechanism, which organisms potentially possess for other biological functions such as tyrosine-crosslinking and innate immunity. Consistent with the importance of BLI-3/DUOX-mediated ROS for lifespan regulation, bli-3(im10) mutants carrying a mutation in the NOX domain did not show PQQ-mediated lifespan extension. However, bli-3(e767) mutants carrying a mutation in peroxidase-like domain showed a peculiar phenotype: compared with wild-type animals, bli-3(e767) mutants exhibited longer lifespan without PQQ and a shorter lifespan with 5 mM PQQ (Fig. S3A, Table S2), and this lifespan phenotype is similar to that of mlt-7(im39) mutants (Fig. 5A; Table S2). Given that the peroxidase-like domain of DUOX is believed to interact with peroxidase (Fig. S3E) (Donkó et al., 2005; Fortunato et al., 2010; Song et al., 2010; Thein et al., 2009), we hypothesize that the bli-3(e767) mutation in the peroxidase-like domain (Fig. S3C and D) caused effects similar to the mlt-7(im39) reduction-of-function mutation, for example, by preventing the interaction with MLT-7 protein.

Recent reports suggest that mtROS extend lifespan (Hwang et al., 2014; Lee et al., 2010; Ristow, 2014; Schulz et al., 2007; Wang and Hekimi, 2015; Wei and Kenyon, 2016; Yang and Hekimi, 2010a; Yee et al., 2014; Yun and Finkel, 2014; Zarse et al., 2012). The DUOX-mediated ROS-generating pathway may be distinct from that for mtROS. While mtROS signaling requires CED-4/Apaf1 and HIF-1, the lifespan of ced-4 (Yee et al., 2014) and hif-1 (Hwang et al., 2014; Lee et al., 2010) mutants were extended by PQQ (Fig. S2F,G; Table S2). These results are consistent with the different cellular sites of ROS production: mtROS is generated at mitochondria and DUOX-mediated ROS at the plasma membrane. Although mtROS is induced by various stimuli as shown in many studies, DUOX-mediated ROS is specifically generated by the activation of BLI-3 or DUOX proteins at plasma membrane. Because of this specificity, we predict that exploration of natural products or chemical synthesis of compounds that activate DUOX proteins will be valuable for anti-aging and lifespan intervention.

Two reduction-of-function mutants in the skn-1 gene, skn-1(zu67) and skn-1(ok2315), did not show any lifespan with PQQ at all (Fig. 6A,B; Table S2). We thus concluded that SKN-1 is a critical transcriptional factor for PQQ-mediated lifespan extension. Our conclusion can be supported by two previous studies showing that SKN-1 mediates BLI-3-dependent ROS in innate immunity (van der Hoeven et al., 2011) and that the impairment of a negative regulator of BLI-3 confers longevity by activating SKN-1 (Ewald et al., 2017).

The previous report showed that both SKN-1 and DAF-16 are required for PQQ-mediated lifespan extension (Wu et al., 2016). Our results suggest that, in contrast to SKN-1, the role of IIS on PQQ-mediated lifespan extension is complex. We found that four alleles of three genes involved in IIS, daf-2(e1370), age-1(hx546), daf-16(mu86) and daf-16(m26), caused lifespan extension with PQQ (Figs. S2A–E, Table S2). However, the lifespan extensions in daf-16(mu86), and daf-16(m26) mutants are not as large as those in daf-2(e1370) and age-1(hx546) mutants, implying the partial dependence of PQQ-mediated lifespan extension on IIS. Furthermore, we observed a peculiar phenotype for daf-2(e1368), in which no lifespan extension was observed for this reduction-of-function allele (Fig. S2B, Table S2). daf-2 mutations lead to two phenotypic classes: e1368 belongs to class I where mutations mainly exist in extra-cellular ligand-binding domain, whereas e1370 belongs to class II where mutations mainly exist in intra-cellular tyrosine-kinase domain (Gems et al., 1998; Kimura et al., 1997; Patel et al., 2008). Different results between e1368 and e1370 mutants may be attributed to different signaling owing to the different mutation sites in DAF-2 receptor. Similar to our results, differently classified mutations are known to result in different outcomes: the lifespan of class II daf-2(e1370) mutants is further extended by diet restriction, whereas that of class I daf-2(e1368) mutants was not (Iser and Wolkow, 2007). These phenotypes of daf-2(e1368) mutation imply that PQQ-mediate lifespan extension can utilize IIS. Since SKN-1 and IIS are reported to function cooperatively for lifespan regulation (Blackwell et al., 2015; Ewald et al., 2015; Tullet et al., 2008), DAF-16 could partly participate in PQQ-mediated lifespan extension.

Numerous reports have shown that the SKN-1 transcriptional factor contributes to longevity across species (An and Blackwell, 2003; Blackwell et al., 2015; Inoue et al., 2005; Kubben et al., 2016; Maher and Yamamoto, 2010; Park et al., 2009; Tullet et al., 2008). The chronic activation of SKN-1 would strengthen the antioxidant network and elicit hermetic effects to induce longevity. However, the role of JUN-1 for lifespan regulation still remains poorly understood, although JUN-1 is known to be required for the lifespan extension induced by intermittent fasting (Honjoh et al., 2009; Uno et al., 2013). We showed in this study that the lifespan of jun-1(gk557) loss-of-function mutants was not extended, but rather shortened with PQQ (Fig. 6C; Table S2). Although this phenotype is similar to mlt-7(im39) reduction-of-function mutants (Fig. 5A; Table S2), jun-1(gk557) mutants live for a shorter time than wild-type animals without PQQ, which is different from the long-lived phenotype of mlt-7(im39) mutants without PQQ (Figs 5A and 6C; Table S2). It is possible that JUN-1 might be necessary for buffering the possible toxic effects of ROS when animals are exposed to PQQ. Hence, JUN-1 might upregulate the expression of redox enzymes such as MLT-7 upon exposure to PQQ.

PQQ is known to exhibit antioxidant effects both in vivo and in vitro, and exert beneficial effects such as neuro- and cardio-protection (Akagawa et al., 2015; Rucker et al., 2009). We showed here the unexpected efficacy of PQQ, in which lifespan can be extended by generating ROS at the plasma membrane. Similar to how a vaccine works, the low levels of ROS may strengthen the defensive mechanism to extend lifespan.

Large-scale epidemiological studies have shown that certain vitamins do not promote longevity and rather shorten lifespan in some cancer patients (The Alpha-Tocophenorol, Beta Carotene Cancer Prevention Study Group, 1994; Bjelakovic et al., 2007; Moyer, 2013). Given that low levels of ROS, namely mtROS and DUOX-mediated ROS, extend lifespan, the intake of vitamins could eliminate such low levels of ROS and cancel their longevity-promoting effects (Desjardins et al., 2017). We propose that PQQ is an effective chemical in lifespan extension acting through generating longevity-promoting ROS at the plasma membrane by the peroxidase and dual oxidase system, which is a fundamentally different action from those of existing antioxidants.

Dozens of chemicals have been identified as lifespan-extension molecules. Some of them generate ROS to extend lifespan. The botanical pesticides juglone and plumbagin are thought to extend lifespan of C. elegans by generating ROS (Hartwig et al., 2009; Hunt et al., 2011; Przybysz et al., 2009). Low concentrations of paraquat, which is toxic at high concentrations, were reported to extend lifespan of C. elegans by generating ROS from mitochondria (Yang and Hekimi, 2010a). D-Glucosamine extends the lifespan of C. elegans by impairing glucose metabolism, and increases the mitochondria biogenesis that triggers mtROS generation (Weimer et al., 2014). The mood stabilizer lithium extends the lifespan of Drosophila through inhibition of GSK3 and activation of Nrf2 (also known as Cap-n-collar in flies) (Castillo-Quan et al., 2016). Importantly, while these chemicals mainly generate ROS in mitochondria, PQQ generates ROS in the plasma membrane. Nevertheless, even if the sites of ROS generation in cells are different, the final longevity outcomes are the same when ROS levels are low (Schaar et al., 2015).

Previous studies have reported that BLI-3 is expressed in intestine and hypodermis (Edens et al., 2001; van der Hoeven et al., 2015) and SKN-1 is expressed in intestine and ASI chemosensory neurons (An and Blackwell, 2003). These results suggest that the intestine may be the organ responsible for PQQ-mediated lifespan extension. We also demonstrated that PQQ activated human DUOX1 and DUOX2 in a heterologous expression system, and that DUOX2 was more highly activated by PQQ than DUOX1 (Fig. 3E). Intriguingly, and consistent with the importance of intestine as the target of PQQ, DUOX2 is expressed in thyroid glands and the intestine, while DUOX1 is expressed in thyroid glands, salivary glands and the respiratory tract (An and Blackwell, 2003; Bedard and Krause, 2007; Geiszt et al., 2003; Sumimoto, 2008). The intestine is indeed shown to be a key organ to determine lifespan in C. elegans (Libina et al., 2003) and is proposed to be important for maintaining health during aging in human (Flint et al., 2012). Alternatively, a recent report showed that collagen remodeling of ECM in hypodermis contributes to longevity (Ewald et al., 2015). Thus, the BLI-3-mediated ROS signal might strengthen the integrity of ECM, thereby contributing to longevity.

C. elegans strains were maintained using a standard method with slight modifications (Brenner, 1974; Mohri et al., 2005). C. elegans animals were cultivated on 6-cm nematode growth medium (NGM) plates. The composition of the medium on the NGM plates was, per 1 liter: NaCl (3 g), Bacto peptone (2.5 g; Becton, Dickinson and Company, Sparks, MD, USA), agar (20 g; Wako Pure Chemical Industries Ltd., Osaka, Japan), 1 M KH₂PO₄ (25 ml), 1 M CaCl₂ (1 ml), 1 M MgSO₄ (1 ml) and 5 mg/ml cholesterol in ethanol (1 ml). The strains used in this study were: wild-type N2 Bristol, OB140 bli-3(im10), CB767 bli-3(e767), MT1141 bli-3(n529), SP2275 tsp-15(sv15), OB254 bli-3(gk141);imEx144[bli-3] (Moribe et al., 2012), OB14 tsp-15(sv15);imEx14 [tsp-15::gfp] (Moribe et al., 2004), OB274 mlt-7(im39), CF1038 daf-16(mu86), DR26 daf-16(m26), TJ1052 age-1(hx546), IK38 daf-2(e1370), DR1572 daf-2(e1368), MT2547 ced-4(n1162), ZG31 hif-1(ia4), DA465 eat-2(ad465), EU1 skn-1(zu67)/nT1[unc-?(n754) let-?](IV;V), VC1772 skn-1(ok2315)/nT1[qIS51] (IV;V), VC1200 jun-1(gk557).

To generate transgenic strains, a DNA mixture containing bli-3, doxa-1::venus, tsp-15::gfp, and lin-44::gfp (Herman et al., 1995; Moribe et al., 2012) was injected using standard methods (Mello et al., 1991). The concentrations of the three genomic DNAs (bli-3, doxa-1::venus, tsp-15::gfp) were varied at 1, 10 or 25 ng/µl each. The concentration of injection marker [lin-44::gfp (Herman et al., 1995; Sawa et al., 1996)] was 50 ng/µl. The transgenic strains constructed in this study included IK2989 njEx1211 (OE 1 line 1) and IK2990 njEx1212 (OE 1 line 2), both overexpressing bli-3, doxa-1::venus, and tsp-15::gfp at 1 ng/µl each. IK2866 njEx1150 (OE 10 line 1) and IK2988 njEx1210 (OE 10 line 2) are strains in which the three DNAs are overexpressed at 10 ng/µl each. IK2991 njEx1213 (OE 25 line 1) and IK2992 njEx1214 (OE 25 line 2) are strains in which the three DNAs are overexpressed at 25 ng/µl each. IK2994 njEx1216 is the strain in which only the injection marker (lin-44::gfp) is expressed at 50 ng/µl.

Nuclear magnetic resonance (NMR) measurements were performed using a 500 MHz NMR, JNM-ECA500 spectrometer (JEOL Ltd, Tokyo, Japan). High-performance liquid chromatography-mass spectrometry (LC-MS) measurements were performed using a LTQ Orbitrap Discovery (Thermo Fisher Scientific, Waltham, MA, USA). Reagents were from Wako Pure Chemical Industries Ltd., unless otherwise specified.

Pyrroloquinoline quinone {PQQ; 4,5-dioxo-4,5-dihydro-1H-pyrrolo[2,3-f]quinoline- 2,7,9-tricarboxylic acid} is generally used as a disodium salt. Accordingly, PQQNa₂ (BioPQQ; Mitsubishi Gas Chemical Co. Inc., Japan) was designated as ‘PQQ’ for treatments administered in this study (Ikemoto et al., 2012). Single-crystal X-ray structure analysis showed that this compound was PQQ disodium trihydrate. Purity of this material was nearly 100% (specification >99%) as determined by high-performance liquid chromatography analysis.

To synthesize 7-oxo-7,10-dihydroimidazo[4,5,1-ij]pyrrolo[2,3-f]quinoline-1,3,9-tricarboxylic acid trisodium (IPQNa₃), PQQ disodium (BioPQQ, 100 g) was mixed with glycine in water (0.5 l) and heated at 340 K for 3 days. The reaction mixture was then added to 10% (w/v) aqueous NaCl and the precipitate filtered. The solid was crystalized in 20% (w/v) aqueous NaOH aqueous (150 ml) at 340 K overnight. The solid was washed with 2-propanol and dried under vacuum. The resulting product was IPQNa₃ (65.8 g). The detailed data for IPQNa₃ are: ¹H-NMR (D₂O, TSP): 7.24, 8.20, 9.35 ppm; ¹³C-NMR (D₂O, TSP): 112.45, 118.56, 123.79, 128.59, 130.81, 133.17, 134.10, 135.05, 135.09, 136.94, 138.66, 169.63, 170.90, 174.54, 178.89 ppm; MS (ESI positive): 342.03537 (M+1=IPQ+1).

A sodium ion meter (LAQUA twin B-722Na+; Horiba Scientific, Kyoto, Japan) was used for sodium analysis. These data indicated presence of the trisodium salt.

Solutions were prepared at 30 or 40 mM. PQQNa₂ was poured into distilled water and 1 M NaOH was added to increase the pH to ∼5. Next, the solution was heated to ∼90°C for at least 8 h. After the solution cooled to room temperature, 1 M NaOH was slowly added to raise the pH to 6.0. The solution was then autoclaved and the pH was again measured. If the value was not close to pH 6.0, 1 M NaOH was added to increase the pH to 6.0.

PQQ-containing NGM medium (2.4 l) was prepared in an Erlenmeyer flask (3 l). NaCl (7.2 g), Bacto peptone (6.0 g) and agar (48 g) were poured into distilled water and autoclaved. Next, 1 M KH₂PO₄ (60 ml), 1 M CaCl₂ (2.4 ml), 1 M MgSO₄ (2.4 ml) and 5 mg/ml cholesterol in ethanol (2.4 ml) were added. Because PQQ easily reacts with amino acids at high temperature (Mitchell et al., 1999), the PQQ and the NGM medium were mixed at a controlled temperature. The surface temperature of NGM medium-containing flasks was measured using a digital thermo-probe (HA-100K; Anritsu, Kanagawa, Japan). A PQQ solution pre-incubated at 37°C was added to flasks when the surface temperature was between 40 and 42°C. The mixture was poured into 6-cm-diameter plates (14 ml/plate). PQQ plates were stored at 4°C.

PQQ-containing NGM plates and control plates were maintained at room temperature for 1–2 days to remove moisture. To produce PQQ-containing OP50, PQQ was mixed with E. coli OP50. The final PQQ concentration of the OP50 was the same as that of the plates. PQQ-containing OP50 (100 µl) was spotted on PQQ-containing NGM plates and left for 1–2 days at room temperature for propagation of the E. coli OP50.

We designated IPQNa₃ as ‘IPQ’. IPQ stock solution (30 mM) was made from IPQNa₃. IPQ lifespan assay plates were generated as described for PQQ assay plates.

UV-killed bacteria plates were prepared as described previously (Garigan et al., 2002). Control and PQQ lifespan assay plates were placed on a UV-transilluminator (DT-35MCP; ATTO, Tokyo, Japan) and irradiated with 302 nm UV for >25 min. To ensure that bacteria had been killed, E. coli OP50 was streaked on Luria–Bertani (LB) agar plates, and this testing performed several times. No colonies were detected on LB agar plates, demonstrating successful killing of the bacteria.

NAC (Wako Pure Chemical Industries Ltd.) was dissolved in autoclaved distilled water to produce stock solution (500 mM). Appropriate volumes of 500 mM NAC were added to medium to make 1 mM or 10 mM NAC plates. These were seeded with E. coli OP50. Since OP50 growth was very poor on 10 mM NAC plates, a lawn of live OP50 was transplanted to 10 mM NAC plates and lifespan assays were performed as previously described (Yang and Hekimi, 2010a).

Lifespan assays were performed as described previously (Mair et al., 2011; Sutphin and Kaeberlein, 2009) with slight modifications. Hermaphrodite animals were used and all lifespan assays were performed at 20°C. Animals were cultivated from the egg to L4 stage on NGM plates. L4 stage animals (as judged by gonad morphology) were transferred to fresh NGM plates and cultivated until the adult stage. Adult animals were divided onto plates corresponding to experimental groups (for example, onto control and 5 mM PQQ plates). About 10–25 animals were placed on each plate, and the time of this placement was defined as adult day (AD) 1. AD 1 of wild-type animals corresponds to the fourth day after hatching. Adult Days is displayed in all graphs and tables. Animals were transferred to new plates every day during the reproduction period (until ∼AD 10). After the reproduction period, animals were transferred to new plates every few days. Animals were scored as dead when they did not respond to mechanical stimuli. Animals that crawled off the plate, exploded or died as internal hatchlings (bag) were excluded from the analysis. Since PQQ-mediated lifespan experiments can be inevitably influenced by uncontrolled factor(s), we always performed the control experiments using wild-type animals without PQQ or with 5 mM PQQ (unless samples could not be used, for example, because of mold growth). When the significant differences were not observed in the control experiments, we did not use data. All strains except IIS mutants were cultivated at 20°C for lifespan experiments. To avoid dauer traits (Hu, 2007), daf-2(e1370), daf-2(e1368) and age-1(hx546) mutants were cultivated at 15°C from eggs to young adults before being transferred to new plates and being cultivated at 20°C for >10 h. Adult animals were divided onto control and 5 mM PQQ plates for lifespan assays. We did not use 5-fluoro-2′-deoxyuridine (FUdR) to prevent reproduction in lifespan assays.

Mean adult lifespan was calculated by averaging the adult lifespan of all animals from one experimental group. Mean 90% survival was calculated by averaging the days when the fraction of surviving animals reached 90% in each experimental group. Mean 10% survival was calculated by averaging the days when the fraction of surviving animals reached 10%. All values are means±standard error of the mean (s.e.m.). Student's t-test, Welch's t-test and the Mann–Whitney U-test were used for statistical comparisons between two independent groups. Student's t-test was used if the data had a normal distribution, and equality of two variances was determined by F-tests. Welch's t-test was used if the data had a normal distribution but unequal variance was determined by F-tests. The Mann–Whitney U-test was used if the data did not have a normal distribution. For multiple comparisons, statistical analyses of variance were performed using Bartlett's test with either single-factor ANOVA employing the post hoc Tukey–Kramer test, or the Kruskal–Wallis test employing the post hoc Steel–Dwass test, if the data had equal or unequal variance across groups, respectively. Statistical and other data analyses were performed using Excel 2010 (Microsoft) with the add-in software Statcel3 (OMS Publishing Inc., Saitama, Japan). Statistical data are displayed in supplementary material (Tables S1 and S2).

Pharyngeal pumping and body bends with ages were measured based on the previous reports (Kenyon et al., 1993; McKay et al., 2004; Mohri et al., 2005). Animals that were close to dying, or with difficult to identify pharyngeal structures, or that stop pumping were excluded for pumping measurements. Animals close to dying were excluded for body bends measurements.

HT1080 stable transfectants expressing C. elegans bli-3, doxa-1 and tsp-15 were as described previously (Moribe et al., 2012). The mutants bli-3^G246D, bli-3^D392N and bli-3^P1311L were generated by site-directed mutagenesis, as described previously (Moribe et al., 2012). Stable transfectants expressing HA-tagged human DUOX1 and FLAG-tagged human DUOXA1α (D1A1), or HA-tagged human DUOX2 and FLAG-tagged human DUOXA2 (D2A2), were established because DUOX1–DUOXA1 and DUOX2–DUOXA2 were reported to be the best partner pairings for proper H₂O₂ production (De Deken et al., 2014). The HA tag was inserted between Ala²¹ and Gln²² of DUOX1 and between Asp²⁷ and Ala²⁸ of DUOX2 (Grasberger and Refetoff, 2006; Luxen et al., 2009; Morand et al., 2009). The FLAG tag was attached to the C-terminus of DUOXA1α and DUOXA2. H₂O₂ production by tagged human DUOXs and DUOXAs was comparable to that by the non-tagged forms. Cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Equitech-Bio, Inc, Kerrville, TX).

Extracellular H₂O₂ production in stable cell lines was detected as described previously (Moribe et al., 2012). Briefly, cells were plated onto 96-well plates at 5×10⁴ cells/well and cultured for 24 h. Cells were incubated at 37°C for 1 h in 100 µl assay buffer containing PBS with 0.9 mM, CaCl₂, 0.49 mM MgCl₂, 1 g/l glucose, 50 µM Amplex Red and 0.1 U/ml horseradish peroxidase (HRP; Nacalai Tesque, Kyoto, Japan), supplemented with PQQ or IPQ. Because a Ca²⁺ stimulus is required for human DUOX activation (Deme et al., 1985; Gorin et al., 1997; Grasberger et al., 2007), 1 µM ionomycin (IM; LKT Laboratories, St. Paul, MN) was added to the assay buffer. NOX enzymatic inhibition by DPI was evaluated as described previously (Moribe et al., 2012). Briefly, to inhibit flavoprotein activity, cells were preincubated for 10 min in 10–25 µM DPI (Sigma-Aldrich, St. Louis, MO) and assay buffer was added. Fluorescence (544 nm excitation, 590 nm emission) was measured using a Fluoroskan Ascent FL (Thermo Fisher Scientific). For measurement of time-dependent H₂O₂ production, fluorescence was measured at 0, 2, 5, 10, 15, 30, 45 and 60 min. H₂O₂ concentrations were calculated using control standard curves for each experiment. Means±s.e.m. were determined from at least six independent experiments. Significance was determined using the two-tailed Student's t-test.

Cells were incubated at 37°C for 1 h in PBS containing 0.9 mM CaCl₂, 0.49 mM MgCl₂ and 1 g/l glucose, with or without 10 µM PQQ. Cells were lysed in lysis buffer (20 mM Tris-HCl, 150 mM NaCl and 2 mM EDTA, pH 7.4) with 1% CHAPS or 1% Triton X-100. For cell surface labeling, cells were incubated with 0.1 mg/ml sulfo-NHS-LC-biotin (Thermo Fisher Scientific) at 4°C for 30 min prior to cell lysis. The cell lysates were centrifuged at 20,000 g for 15 min at 4°C. The supernatants were incubated with streptavidin–agarose beads (Solulink, San Diego, CA) and then blotted with rabbit anti-BLI-3 [N1, as described previously (Moribe et al., 2012), 1:1000], mouse anti-FLAG (M2; Sigma-Aldrich, 1:1000), rabbit anti-Omni (Xpress) (M21; Santa Cruz Biotechnology, Santa Cruz, CA, USA, 1:1000), mouse anti-HA (F7; Santa Cruz Biotechnology, 1:1000), mouse anti-DYKDDDDK (1E6; Wako Pure Chemical Industries Ltd., 1:1000), mouse anti-GAPDH (5A12; Wako Pure Chemical Industries Ltd., 1:1000) or mouse anti-Na⁺/K⁺ ATPase (464.6; Abcam, Cambridge, UK, 1:5000) antibodies. HRP-conjugated donkey anti-rabbit-IgG (Merck Millipore, Billerica, MA, USA, 1:5000) or donkey anti-mouse-IgG (Merck Millipore, 1:5000) were used as secondary antibodies. Prior to immunoblotting of TSP-15, samples were treated with N-glycosidase F (PNGase F; New England Biolabs, Beverly, MA) to remove N-glycans from TSP-15 (Moribe et al., 2012).

We thank N. Hisamoto, K. Matsumoto, E. Nishida, E. Mekada and the Caenorhabditis Genetics Center (funded by National Institutes of Health Office of Research 362 Infrastructure Programs P40 OD010440) for providing C. elegans strains. We thank L. Xu, R. Kondo, W. Ohata, N. Ozawa, R. Yamaguchi, T. Tamada, Y. Oshikoshi, S. Iwasaki and M. Yoshida for assistance with lifespan assays; Y. Murakami, F. Takeshige, J. Okada and K. Sawayama for technical assistance; and L. Avery, C. Bargmann, K. Nishiwaki, S. Shibata, Y. Young-Jai, H. Komatsu, S. Nakano, T. Kohashi, Y. Tsukada and I. Aoki for critical reading of the manuscript.