Reactive oxygen species extend insect life span usingcomponents of the insulin-signaling pathwayXiao-Shuai Zhanga, Tao Wanga, Xian-Wu Lina, David L. Denlingerb,c,1, and Wei-Hua Xua,1

aSchool of Life Sciences, Sun Yat-Sen University, Guangzhou 510006, China; bDepartment of Entomology, Ohio State University, Columbus, OH 43210;and cDepartment of Evolution, Ecology and Organismal Biology, Ohio State University, Columbus, OH 43210

Contributed by David L. Denlinger, August 3, 2017 (sent for review June 20, 2017; reviewed by Subba R. Palli and Marla B. Sokolowski)

Reactive oxygen species (ROS) are well-known accelerants of aging,but, paradoxically, we show that physiological levels of ROS extendlife span in pupae of themoth Helicoverpa armigera, resulting in thedormant state of diapause. This developmental switch appears tooperate through a variant of the conventional insulin-signalingpathway, as evidenced by the facts that Akt, p-Akt, and PRMT1are elevated by ROS, but not insulin, and that high levels of p-Aktfail to phosphorylate FoxO through PRMT1-mediated methylation.These results suggest a distinct signaling pathway culminating inthe elevation of FoxO, which in turn promotes the extension of lifespan characteristic of diapause.

insulin signaling | Akt | PRMT1 | insects | diapause

Diapause in insects is akin to dauer, the slowed develop-mental phase in Caenorhabditis elegans (1) and hibernation

in vertebrates (2). These dormant stages share a similar pheno-type, characterized by suppressed metabolic activity and arresteddevelopment. The brain functions as both a receptor and pro-grammable center to perceive environmental signals to inducediapause. When reared under long daylength and low tempera-ture (20 °C) the cotton bollworm Helicoverpa armigera quicklyprogresses through the pupal stage and develops immediatelyinto an adult, but when reared under short daylengths at thesame temperature the bollworm enters pupal diapause (3). Thelife span of this “nonaging” diapause pupa is extended manymonths (4), making H. armigera an excellent model for life spanresearch.Numerous reports indicate that reactive oxygen species (ROS)

promote aging processes (5) and are associated with diversemedical disorders, including Alzheimer’s disease, Parkinson’s dis-ease, cancer, diabetes, and others (6, 7). However, pioneering workin yeast, C. elegans, and Drosophila melanogaster has shown thatincreased ROS from chemical inhibition or mutations that affectmitochondrial function or allotopic expression can also lengthen lifespan (8–11). What is unclear is whether naturally occurring, physi-ological levels of ROS can also regulate life span. It is also unclearhow ROS play a dual function in these species and what molecularpathway is evoked by ROS to extend life span. ROS appear to elicitdistinct responses at different developmental stages: inducing agingin older individuals while promoting life span extension in youngerindividuals. To evaluate this dichotomy of function, we focusedon ROS in diapausing pupae, a nonaging stage that is locked intoa developmental arrest.The insulin-signaling pathway plays a critical role in regulating

the dauer state in C. elegans (1), life span extension in adults ofD. melanogaster (12), and adult diapause (reproductive arrest) inthe mosquito Culex pipiens (13). Insulin activates Akt, which inturn phosphorylates FoxO, promoting FoxO degradation (14)and elevation of metabolic activity (13, 15). Low insulin signalingresults in down-regulation of Akt, leading to activation of FoxO,which in turn promotes life span extension by regulating tran-scription in a number of critical downstream genes (16). Modu-lation of the insulin-signaling pathway by ROS through PI3K/Akt/FoxO (17) implicates ROS as a potential regulator of life spanthrough components of the insulin-signaling pathway.

Numerous genes, proteins, and metabolites that are differen-tially expressed in diapause-destined individuals are involved incarbohydrate metabolism (18–21). Diapause-destined pupae ofH. armigera are characterized by low glucose levels in the bloodbut high glucose levels in the brain, and we know that restriction ofglucose by 2-deoxy-D-glucose (DOG) injection can delay devel-opment of nondiapausing individuals (22). In C. elegans, reducedglucose metabolism increases life span in an ROS-dependentmanner through an impaired insulin pathway or glucose restric-tion (23). These results suggest that insulin signaling and ROSmay play important roles in diapause regulation.Here, we monitored insulin-like peptide (ILP) levels in pupal

blood and found that high ILP levels led to insect development,whereas low levels were associated with diapause. However,surprisingly, Akt, p-Akt, and FoxO levels were higher in brains ofdiapause-destined pupae of H. armigera, compared with theirnondiapausing counterparts. This result was not consistent withthe observations of (i) low ILP levels and (ii) negative regulationof FoxO by p-Akt. Further experiments showed that ROS, butnot ILPs, elicit high expression of Akt and abundant p-Akt indiapause-destined pupal brains. We conclude that high p-Aktlevels activate the target protein glucose transporter (Glut) tosense low glucose in the blood and enhance its uptake by thebrain as an energy resource. Elevation of protein arginine meth-yltransferase 1 (PRMT1), a predominant member of the PRMTfamily (24), blocks FoxO phosphorylation to reduce FoxO proteindegradation, thus promoting accumulation of FoxO in brainsof diapause-destined pupae, leading to life span extension (dia-pause). The results suggest a mechanism by which the brainnaturally controls life span extension through a distinct ROS-mediated insulin-signaling pathway, indicating that physiological

Significance

Oxidative damage is frequently associated with aging and aging-related disease, but, paradoxically, several recent studies haveshown that artificial boosts of reactive oxygen species (ROS) canalso extend life span in young individuals. Here, we show thatphysiological levels of ROS promote diapause, thereby extendinglife span in pupae of the moth Helicoverpa armigera. Insect dia-pause, like the dauer stage of nematodes, is a period of devel-opmental rest that results in a profound extension of life span.ROS appears to contribute to this life span extension by actingthrough components of the insulin-signaling pathway. Ourresults thus suggest a new molecular mechanism regulatinglife span and help to explain the dual nature of ROS actionin animals.

Author contributions: X.-S.Z., D.L.D., and W.-H.X. designed research; X.-S.Z., T.W., andX.-W.L. performed research; and X.-S.Z., D.L.D., and W.-H.X. wrote the paper.

Reviewers: S.R.P., University of Kentucky; and M.B.S., University of Toronto.

The authors declare no conflict of interest.1To whom correspondence may be addressed. Email: xuweihua@mail.sysu.edu.cn ordenlinger.1@osu.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1711042114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1711042114 PNAS Early Edition | 1 of 9

PHYS

IOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

May

1, 2

020

levels of ROS are exploited by insects to extend life span inyoung individuals.

ResultsNondiapause- and diapause-destined pupae of H. armigera werecultured at 20 °C, differing only in the photoperiod they received.Reared under long daylengths, pupae failed to enter diapauseand emerged as adults 21–23 d after pupation. By contrast, pu-pae reared under short daylengths entered diapause 8–10 d afterpupation, and their pupal life span was extended beyond 3 mo(Fig. 1A). We focused on the diapause initiation phase from day0 to day 10 after pupation. To elucidate the relationship betweeninsulin signaling and diapause, we measured ILP titers in bloodfrom nondiapause- and diapause-destined pupae using compet-itive ELISA; ILP titers in nondiapause-destined pupae graduallyincreased from days 0 to 10 after pupation, whereas titers indiapause-destined pupae remained low through day 10 (the on-set of diapause) (Fig. 1B). Injection of exogenous insulin intoday-1 diapause-destined pupae averted diapause and prompteddevelopment (Fig. 1C), suggesting that high insulin elicits de-velopment, while low insulin levels lead to diapause.To clarify how insulin signaling regulates this developmental

switch we monitored brain abundance of two downstream geneproducts, Akt and FoxO, in pupae by Western blots (Fig. 1D andSI Appendix, Fig. S1). Akt levels in nondiapause- and diapause-destined pupae were similar, but the active form of Akt, p-Akt,was significantly higher in diapause-destined pupae. FoxOabundance in diapause-destined pupae was significantly higher,but p-FoxO levels were lower. These results are inconsistent withthe predicted and anticipated low ILP levels, high Akt and p-Aktlevels, and low p-FoxO levels observed in diapause-destined in-dividuals, thus indicating an unexpected but distinct use of com-ponents of the insulin-signaling pathway in regulating diapause inthis moth.The known role of ROS in modulating insulin signaling to

extend life span (23) prompted our examination of ROS activityin the two types of pupae (Fig. 1E). ROS levels were significantlyhigher in diapause-destined pupal brains than in brains fromnondiapause-destined pupae. To further test this relationship,we injected the mitochondrial superoxide generator paraquat(PQ) into day-1 nondiapause-destined pupae to elevate ROSlevels. In control pupae, 50% completed stemmata migration (amarker for development) in 4 d, whereas individuals injectedwith PQ showed delayed development and required 5–6 d tocomplete stemmata migration (Fig. 1F). We also injected theROS scavenger N-acetyl-L-cysteine (NAC) into day-1 diapause-destined pupae to decrease ROS levels: significantly morepupae were channeled into nondiapause than in the controlsthat did not receive NAC (Fig. 1G). Injection of PQ into day-1nondiapause-destined pupae elevated Akt and p-Akt levels inthe brain, accompanied by increased ROS activity (Fig. 1H andSI Appendix, Fig. S2A). When NAC was injected into day-7diapause-destined pupae, ROS were significantly lower, accom-panied by decreased brain levels of Akt and p-Akt (Fig. 1I and SIAppendix, Fig. S2B). Collectively, these results indicate that ROScan elevate Akt and p-Akt levels, suggesting that high Akt andp-Akt in diapause-destined individuals are dependent on ROS,but not ILPs, and that ROS are important regulators of insectdiapause.The high p-Akt and low p-FoxO levels in diapause-destined

individuals imply that Akt does not phosphorylate FoxO. Toclarify the role of Akt in regulating FoxO phosphorylation inresponse to insulin and ROS, we performed cell experimentswith an HzAm1 cell line from Helicoverpa zea, a close relativeof H. armigera. Cell culture experiments clearly showed thatboth p-Akt and p-FoxO respond respectively to insulin (Fig. 2Aand SI Appendix, Fig. S3A) or oxidative stress (Fig. 2B and SIAppendix, Fig. S3B); increased p-FoxO was accompanied by

elevated p-Akt. In addition, insulin or oxidative stress increasednuclear export of FoxO (Figs. 2C and 2D). These results suggestthat Akt phosphorylates FoxO as previously reported (25). Thestructure of FoxO in H. armigera revealed two consensus Aktphosphorylation motifs (RXRXXS) at amino acids 186–191 and250–255, corresponding to Akt phosphorylation motifs of FoxOsin other taxa, including mammals, C. elegans, D. melanogaster,and Bombyx mori (SI Appendix, Fig. S4). The RXRXXS motif ofFoxO undergoes Akt-mediated phosphorylation at Ser, but thiscan be prevented by prior methylation at Arg from the PRMT1(26). Using coimmunoprecipitation, we found that Akt andFoxO specifically bind to each other in vitro (Fig. 2E). We thenperformed in vitro phosphorylation assays using GST-fused Aktand FoxO fragment containing the RXRXXS motif, and theresult showed that Akt specifically interacts with FoxO tophosphorylate FoxO (Fig. 2F), implying that Akt-mediatedphosphorylation of FoxO in diapause-destined individuals maybe abolished by prior PRMT1 methylation.We constructed three FoxO fragments fused with a V5 tag and

expressed these fragments in HzAm1 cells; only the 147–334amino acid fragment containing the RXRXXS motif was able tobind to PRMT1 (Fig. 3A). PRMT1 expression was higher inbrains from diapause-destined pupae, a pattern consistent withthat of FoxO (Fig. 3B); this suggests that PRMT1 is involved inFoxO methylation. We then performed an in vitro methylationassay using GST-PRMT1 and GST-FoxO fragment. The resultsshowed that PRMT1 specifically interacts with FoxO to meth-ylate FoxO (Fig. 3C).We further investigated changes in PRMT1 expression in re-

sponse to ROS signaling. Cell experiments indicated that PRMT1responds to oxidative stress (SI Appendix, Fig. S5). We thenoverexpressed Akt or PRMT1 in HzAm1 cells and treated thecells with a generator of oxidative stress, H2O2. Oxidative stressincreased binding of both Akt and PRMT1 to FoxO, but not toGFP (SI Appendix, Fig. S6). However, when Akt and PRMT1were cotransfected into HzAm1 cells and subsequently treatedwith H2O2 we observed increased PRMT1 and decreased Aktbinding to FoxO (Fig. 3D). We thus conducted sequential meth-ylation and phosphorylation assays using GST-FoxO (147–334amino acids) as substrate and observed that Akt-mediated phos-phorylation of FoxO was blocked by prior PRMT1 methylation(Fig. 3E), as previously reported (26). Furthermore, we conductedphosphorylation assays in HzAm1 cells using dsRNA againstPRMT1; silencing PRMT1 expression by 50–88% resulted in in-creased levels of p-FoxO (Fig. 4A). When we transfected PRMT1or GFP into HzAm1 cells and treated with H2O2, PRMT1 over-expression resulted in down-regulation of p-FoxO (Fig. 4B). In-jection of a selective PRMT1 inhibitor 2818500 (27) into day-1diapause-destined pupae increased p-FoxO levels in the brains(Fig. 4C), and treatment of HzAm1 cells with the PRMT1 in-hibitor resulted in an increase in nuclear export of FoxO (Fig. 4D),as reported (17). Taken together, this evidence suggests thatPRMT1 actively responds to ROS signaling and that FoxO ismethylated by PRMT1 to prevent phosphorylation.Insulin acts to increase glucose uptake through an Akt-

activated glucose transporter (Glut) (28). To probe this portionof the pathway we examined changes in p-Akt and Glut proteinlevels after treatment of HzAm1 cells with DOG, a glucose de-rivative known to elevate ROS levels (SI Appendix, Fig. S7), andfollowing injection of DOG into day-6 nondiapause-destinedpupae. Both p-Akt and Glut proteins were significantly ele-vated in both cells and brain in response to this oxidative stress(Fig. 5A). When HzAm1 cells were treated with the Akt inhibitorAI4, both p-Akt and Glut proteins significantly declined (Fig.5B), indicating that Glut can respond to p-Akt and that Glut isa likely target of Akt in insect diapause.

2 of 9 | www.pnas.org/cgi/doi/10.1073/pnas.1711042114 Zhang et al.

Dow

nloa

ded

by g

uest

on

May

1, 2

020

DiscussionDiapause is a complex physiological response, with many sig-naling pathways participating in the process. Although we haveknown about roles for prothoracicotropic hormone, ecdysone,the juvenile hormones, and diapause hormone for quite some

time (3), many additional signaling pathways are also likely in-volved. One such prominent pathway is the insulin-signalingpathway (29). In the insulin-signaling pathway, insulin activatesAkt, and high p-Akt levels repress FoxO activity and activateother genes that promote development. In contrast, low insulin

Fig. 1. Roles of insulin-like peptides (ILPs) and ROS in diapause and associated changes in Akt and FoxO. (A) A schematic representation of pupal diapauseregulation in the moth, Helicoverpa armigera. (B) ILP titers in pupal hemolymph detected by competitive ELISA. Hemolymph from 10 pupae was collected andmixed as a sample for competitive ELISA using an insulin antibody. DP, diapause-destined pupae; NP, nondiapause-destined pupae. (C) Developmental fate ofdiapause-destined pupae following an insulin injection. Diapause-programmed pupae were injected with insulin on day 1 following pupation, and the onsetof development was determined by observing timing of the disappearance of the pupal stemmata. (D) Western blot showing patterns of abundance for Akt,p-Akt, FoxO, and p-FoxO. Proteins from brains were extracted and detected with corresponding antibodies. (E) ROS levels in the brain. (a) Brains were in-cubated with the ROS detector CM-H2CFDA for 1 h. (b) Quantification of ROS levels in the brains. Relative ROS levels indicate the highest value as 100.(F) Developmental delay in nondiapause-destined pupae caused by injection of the ROS generator, paraquat (PQ). PQ was injected into day-1 pupae, anddevelopmental delay was determined by examining location of the pupal stemmata on different days after injection. (G) Developmental fate of diapause-destined pupae following injection of the ROS scavenger, N-acetyl-L-cysteine (NAC). Day-1 pupae were injected with NAC or H2O as a control, and the onset ofdevelopment was determined by observing disappearance of the pupal stemmata. Each point represents the mean ± SD of three independent replicates. *P <0.05; **P < 0.01 (determined by an independent t test). Effects of (H) PQ and (I) NAC on levels of ROS, Akt, and p-Akt in the brain. Nondiapause-destinedpupae were injected on day 1 with PQ for 48 h, and diapause-destined pupae were injected on day 7 with NAC for 48 h. ROS were detected as above, andproteins from brains were extracted and detected with corresponding antibodies.

Zhang et al. PNAS Early Edition | 3 of 9

PHYS

IOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

May

1, 2

020

signaling decreases p-Akt levels, resulting in active FoxO, whichregulates life span extension through activation of a variety ofdownstream genes that generate the diapause phenotype (30).FoxO is thus a well-known regulator of life span extension (31).In the present study, we found that the blood of H. armigeracontains low ILP levels, and the brain expresses high FoxO levels

in diapause-destined pupae, suggesting that ILPs regulate thedevelopmental timing of insects through genes downstream ofFoxO. Interestingly, the upstream signals Akt and p-Akt, whichare negative FoxO regulators, are also highly expressed in dia-pause individuals, indicating that high Akt and p-Akt levelsare regulated by factors other than ILPs. High p-Akt levels in

Fig. 2. P-Akt and p-FoxO levels in response to insulin and oxidative stress, and Akt binding to and phosphorylation of FoxO. P-Akt, and p-FoxO levels inresponse to (A) insulin and (B) H2O2. (a) Dose-related response to insulin or H2O2. (b) Time-related response to insulin or H2O2. HzAm1 cells were cultured withvarious doses of insulin or H2O2 for 25 min and with 500 nM insulin for 0, 5, 15, or 25 min or 500 μM H2O2 for 0, 5, 10, or 25 min. (C and D) Nuclear FoxOlocalization in response to insulin or H2O2. HzAm1 cells were transfected with a GFP-FoxO plasmid for 48 h and then treated with distilled water as a control(Con) or with 500 nM insulin or with 300 μM H2O2 for 30 min. Hoechst 33342 labels the nuclei. GFP-FoxO, recombinant GFP-FoxO protein. (Scale bar, 10 μm.)(E) Akt physically associates with FoxO as shown by coimmunoprecipitation. (F) Akt binds to and phosphorylates FoxO. (a) Akt interacts with FoxO as shown ina pull-down assay. (b) In vitro phosphorylation assay. CBB, Coomassie brilliant blue.

4 of 9 | www.pnas.org/cgi/doi/10.1073/pnas.1711042114 Zhang et al.

Dow

nloa

ded

by g

uest

on

May

1, 2

020

Fig. 3. PRMT1 binds to and methylates FoxO and blocks Akt-mediated phosphorylation. (A) FoxO interacts with PRMT1. (a) A schematic representationof FoxO. FH, forkhead box domain; NLS, nuclear localization signal. (b) Coimmunoprecipitation of FoxO fragments and PRMT1. (B) Abundance of PRMT1 inthe brain by Western blot. DP, diapausing pupae; NP, nondiapausing pupae. Protein bands were quantified and normalized to the levels of actin, using 1 asthe highest value. Each point represents the mean ± SD of three independent replicates. *P < 0.05 (determined by an independent t test). (C) PRMT1 binds toand methylates FoxO. (a) PRMT1 interacts with FoxO as shown in a pull-down assay. (b) In vitro methylation assay. SAM, S-adenosyl-methionine; CBB,Coomassie brilliant blue. (D) PRMT1 diverts Akt to associate with FoxO in HzAm1 cells. WCE, whole-cell extracts. (E) In vitro sequential methylation andphosphorylation assays.

Zhang et al. PNAS Early Edition | 5 of 9

PHYS

IOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

May

1, 2

020

Fig. 4. Effects of PRMT1 on FoxO phosphorylation and localization. (A) PRMT1 knockdown increased endogenous FoxO phosphorylation in HzAm1 cells.(a) Dose-dependent response to PRMT1 RNAi. HzAm1 cells were treated for 60 h with PRMT1 dsRNA. GFP dsRNA treated with 2 μg for 60 h was used ascontrol. (b) Time-dependent response to RNAi. HzAm1 cells were treated with 2 μg PRMT1 dsRNA. Histograms indicate quantification of the protein bandsusing image software (Gel-Pro Analyzer) and normalized to the levels of actin, using 1 as the highest value. (B) PRMT1 decreased p-FoxO levels underoxidative stress in HzAm1 cells. Recombinant PRMT1 or GFP were transfected into HzAm1 cells, and cells were treated with H2O2 for 0, 12, or 25 min.(C) Effects of PRMT1 inhibitor 2818500 on p-FoxO levels in the brain. Diapause-destined pupae were injected on day 1 with 2818500 and evaluated 48 hlater. Each point represents the mean ± SD of three independent replicates. *P < 0.05; **P < 0.01 (determined by an independent t test). (D) Nuclear exportof FoxO in response to a PRMT1 inhibitor 2818500. DMSO treatment was the control. (Scale bar, 10 μm.)

6 of 9 | www.pnas.org/cgi/doi/10.1073/pnas.1711042114 Zhang et al.

Dow

nloa

ded

by g

uest

on

May

1, 2

020

diapause individuals are inconsistent with low p-FoxO levels,suggesting that p-Akt does not regulate FoxO phosphorylation inthis case. These conflicting data indicate that the life span ex-tension phenotype, diapause, has a distinct regulatory mecha-nism in the brain, defined by the following observations.

ROS Increase Akt Expression and p-Akt Levels. ROS generation byexogenous sources such as H2O2 triggers a number of pathways

including the ROS-activated PI3K/Akt-signaling pathway (32).The association of aging with ROS has been observed through-out the animal kingdom (33). However, glucose restriction ac-tivates ROS (8), and impaired respiration elevates ROS (34) toextend life span in C. elegans. We now suggest a role for ROS ininsect diapause. Metabolic depression and glucose restriction arecommon in diapausing individuals (3, 22), suggesting that ROSare possible regulators of insect diapause. In this study, we

Fig. 5. P-Akt activation of the glucose transporter (Glut). (A) ROS increased Glut expression by elevating p-Akt levels. ROS increased Glut expression by p-Akt in(a) HzAm cells and (b) the brain. HzAm1 cells were cultured with DOG (2-deoxy-glucose) for 18 h. Nondiapause-destined pupae were injected on day 6 with DOGfor 48 h. Proteins were extracted from cells or brains for Western blots with p-Akt, Glut, and actin antibodies. (B) P-Akt and Glut levels in response to the p-Aktinhibitor AI4. Histograms indicate quantification of protein bands using image software (Gel-Pro Analyzer) and normalized to levels of actin, using 1 as the highestvalue. Each point represents mean ± SD of three independent replicates. *P < 0.05; **P < 0.01 (determined by an independent t test). (C) A schematic repre-sentation depicting a central role for ROS in regulation of pupal diapause in Helicoverpa armigera. High insulin signaling promotes continuous developmentthrough Akt-mediated phosphorylation of FoxO and activation of other downstream genes, including Glut. Low insulin signaling results in low metabolic levels,which induce high ROS levels in diapause individuals. ROS then lead to high levels of Akt, p-Akt, and PRMT1; PRMT1 methylates FoxO to inhibit Akt-mediatedFoxO-phosphorylation. High p-Akt levels activate Glut to sense and absorb low glucose levels from the hemolymph into the brain as an energy resource for slowutilization during the long diapause phase. Abundant FoxO leads to life span extension (diapause).

Zhang et al. PNAS Early Edition | 7 of 9

PHYS

IOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

May

1, 2

020

showed that ROS levels are higher in brains of diapause-destinedpupae than in their nondiapausing counterparts. Glucose re-striction can increase ROS activity, and increased ROS activity innondiapausing pupae delays development. These results suggestthe low metabolic rates characteristic of diapause result in in-creased ROS activity and consequently imply a regulatory rolefor ROS in insect diapause.Akt is widely recognized as a key component of the insulin-

signaling pathway and is highly responsive to insulin signaling(35). ILP levels are lower in diapause-destined individualscompared with their nondiapausing counterparts. This result isconsistent with the finding that a low insulin signal results indiapause through FoxO activation. However, diapause-destinedpupal brains express high Akt and p-Akt levels, suggesting thatfactors other than ILPs regulate Akt expression and phosphor-ylation in diapause individuals. ROS have been reported to in-duce increased Akt activity and p-Akt levels (36–38). In thepresent paper, Akt expression and p-Akt levels in diapause in-dividuals responded to oxidative stress, but not to ILPs, sug-gesting that ROS function as important regulators of insectdiapause by elevating Akt and p-Akt levels.

Inhibition of FoxO Phosphorylation by PRMT1. Akt-dependent phos-phorylation is crucial for the regulation of FoxO function. Aktphosphorylates FoxO to exclude it from the nucleus and promoteits degradation (39). However, the observation that high p-Akt andlow p-FoxO levels are synchronously present in diapause-destinedpupal brains indicates that other mechanisms regulate FoxO activityduring diapause in H. armigera. Two recent reports demonstratedthat FoxO methylation by PRMT1 can directly block Akt-mediatedphosphorylation (26, 40). Thus, we focused on FoxO methylation ininsect diapause. Our main findings are as follows: (i) Brains fromdiapause-destined pupae expressed high levels of ROS-inducedPRMT1, and the PRMT1 expression pattern was similar to thatof FoxO; (ii) PRMT1 could bind to FoxO and methylate FoxO invitro; (iii) PRMT1, but not Akt, can effectively bind to FoxOunder high oxidative stress; (iv) the down-regulation of PRMT1levels by RNAi resulted in increased p-FoxO levels, and a meth-ylation inhibitor increased p-FoxO levels in vivo and the nuclearexport of FoxO in vitro; and (v) FoxO methylation can counteractAkt-mediated phosphorylation. These results show that FoxOmethylation decreases Akt-mediated phosphorylation and thatFoxO accumulates, leading to the induction of diapause.

Physiological Significance of High Akt and p-Akt Levels in theRegulation of Diapause. As described above, strong insulin sig-nals usually promote development, not diapause (13). Our datashowed that both Akt and p-Akt levels were abundant in brainsof diapause-destined pupae, but FoxO cannot be phosphorylatedby p-Akt. This suggests that Akt and p-Akt may have some otherbiological significance in diapause-destined individuals. Basedon the facts that glucose is expressed at low levels in blood ofdiapausing individuals, while high levels of glucose accumulate inthe brain (22), and p-Akt can activate Glut to increase glucoseuptake (28), we speculate that high p-Akt levels may be corre-lated with glucose uptake. Our results showed that Glut is one ofthe p-Akt target proteins, suggesting that high p-Akt levels areresponsible for sensing and absorbing low glucose levels in theblood of diapause individuals by activating Glut, as previouslyreported (41, 42). In addition, it may be necessary to maintainabundant Akt until postdiapause development is initiated. Spe-cifically, key developmental regulators such as Akt may accu-mulate before diapause entry so that the animal is capable ofrapidly restarting pupal–adult development when diapause iscompleted (3). The exact function of Akt in diapausing individ-uals, however, requires further investigation.

High ROS Levels in Pupae Can Lengthen Life Span. Metabolism-induced ROS production and oxidative damage are considereda primary cause of aging and aging-related diseases (43, 44).However, increasing evidence suggests that the opposite mayalso be true: chemical inhibition of metabolism (2-deoxy-D-glucose,paraquat) during young adulthood or mutations that affect mito-chondrial function can extend life span in an ROS-dependentmanner in C. elegans (8, 9). ROS can also increase FoxO levelsand act through transportin-1 to extend life span in C. elegans (45).However, many details of the molecular mechanism that extendslife span by ROS remain unclear.We speculate that this paradox, an apparent dual role for

ROS, may reflect distinct actions of ROS at different life stages:ROS induce the aging process in old individuals, but extend lifespan in young individuals. For example, in D. melanogaster,consistent increased ROS levels were observed in old individualsand are presumed to contribute to the aging process, but trans-genic overexpression of NDI1 (a rotenone-insensitive alternativeNADH dehydrogenase from fungi) increased ROS levels in thebrain of young adults and extended life span (46). This resultindicates that increased ROS levels in the early adult stage canextend life span, although the mechanism is unknown. In addi-tion, epidermal mitochondrial oxidative damage delays featuresof aging in young mice; however, in older mice, this damageaccelerates features of aging (47).In the present study, we showed that diapause pupae, which

are a “nonaging” natural physiological state, produce high levelsof ROS in the brain and that ROS increase life span througha distinct signaling pathway using components of the insulin-signaling pathway, as described above. A model for the regula-tion of developmental timing is proposed in Fig. 5C. In developingindividuals, strong insulin signals lead to the progression of de-velopment through Akt-mediated phosphorylation of FoxO andactivation of downstream genes, including Glut. By contrast, lowinsulin signals found in diapause-destined pupae promote lowmetabolic activity, resulting in increased ROS activity. High ROSactivity then leads to high levels of Akt, p-Akt, and PRMT1, butPRMT1 inhibits Akt-mediated phosphorylation through FoxOmethylation, and abundant FoxO induces pupal life span exten-sion by regulating expression of select downstream genes. High p-Akt levels activate Glut expression, as previously reported (41, 42),to sense and absorb low blood glucose (22) into the brain as anenergy resource for slow utilization throughout the long diapausephase, as suggested by the fact that expression and activity ofhexokinase, which converts glucose to glucose-6-phosphate (thefirst rate-limiting enzyme in glycolysis), are low in brains of dia-pausing pupae and result in low metabolic activity (48).In summary, ROS have long been considered a primary cause

of aging and aging-related diseases (5–7), but only recently havestudies suggested that the opposite is also true: that elevatedROS at nonphysiological levels can extend life span at certainphases of the lifecycle (8–11). Our results support the conclusionthat physiological levels of ROS are beneficial for extending lifespan in young individuals, a viewpoint consistent with recentfindings in D. melanogaster (46), mice (47), and humans (49).

Materials and MethodsLarvae of Helicoverpa armigera were reared on an artificial diet at 20 ± 1 °Cunder a light–dark cycle of 14 h light/10 h dark (nondiapause) or under acycle of 10 h light/14 h dark (diapause). All nondiapausing pupae developedwithout entering diapause, whereas over 95% of the diapause-programmedpupae entered diapause. Developmental stages were synchronized by col-lecting pupae on the day of pupation. Pupal brains were dissected in ice-cold0.75% NaCl and stored at −80 °C until used.

For insulin injection experiments shown in Fig.1C, diapause-programmedpupae were injected on day 1 with differing concentrations of bovine insulin(Sigma) in a volume of 5 μL (0 mM insulin, n = 98; 0.5 mM insulin, n = 94;1 mM insulin, n = 94; 2 mM insulin, n = 92) and held at 22.5 °C. Pupaldevelopment was determined by observing disappearance of the pupal

8 of 9 | www.pnas.org/cgi/doi/10.1073/pnas.1711042114 Zhang et al.

Dow

nloa

ded

by g

uest

on

May

1, 2

020

stemmata, a marker for development. For PQ injections shown in Fig. 1F,nondiapause-destined pupae were injected with PQ on day 1 and incubatedat 22.5 °C. Developmental delay was determined by examining the location ofthe pupal stemmata on different days after injection (0 μg PQ, n = 73; 2 μg PQ,n = 71; 4 μg PQ, n = 70; 6 μg PQ, n = 56). For NAC injections shown in Fig. 1G,diapause-programmed pupae were injected on day 1 with a 3-μL NAC solutionor H2O as a control (H2O, n = 77; 150 μg NAC, n = 70; 300 μg NAC, n = 53;450 μg NAC, n = 53) and then kept at 22.5 °C. Pupal developmental statuswas determined by observing disappearance of the pupal stemmata. For DOGinjections shown in SI Appendix, Fig. S7B, nondiapause-programmed pupaewere injected on day 6 with 3 μL DOG solution (2 mM) for 48 h or H2O as acontrol (H2O, n = 45; DOG, n = 45).

Additional details on polyclonal antibody generation, competitive ELISA,protein extraction and Western blot, construction of overexpression plasmids,vector transfection and cell treatments, coimmunoprecipitation and immu-noblot analysis, measurement of ROS generation, RNA interference, immu-nofluorescence assay, GST pull-down assay, in vitro phosphorylation assay, invitro methylation assay, in vitro sequential methylation and phosphorylationassay, and gene-specific primers are included in SI Appendix, SI Materialsand Methods.

ACKNOWLEDGMENTS. This study was supported by National Natural Scien-tific Foundation of China Grant-in-Aid 31230066 (to W.-H.X.) and US De-partment of Agriculture (USDA)-National Institute of Food and Agriculture(NIFA) Grant 2015-67013-23416 (to D.L.D.).

1. Hu PJ (2007) Dauer. WormBook: The Online Review of C. elegans Biology, 10.1895/wormbook.1.144.1.

2. Aloia RC, Raison JK (1989) Membrane function in mammalian hibernation. BiochimBiophys Acta 988:123–146.

3. Denlinger DL, Yocum GD, Rinehart JP (2005) Hormonal control of diapause.Comprehensive Molecular Insect Science, ed Gilbert LI (Elsevier, Amsterdam), pp615–650.

4. Lu YX, Denlinger DL, Xu WH (2013) Polycomb repressive complex 2 (PRC2) protein ESCregulates insect developmental timing by mediating H3K27me3 and activating pro-thoracicotropic hormone gene expression. J Biol Chem 288:23554–23564.

5. Balaban RS, Nemoto S, Finkel T (2005) Mitochondria, oxidants, and aging. Cell 120:483–495.

6. Levine RL (2002) Carbonyl modified proteins in cellular regulation, aging, and disease.Free Radic Biol Med 32:790–796.

7. Dalle-Donne I, Giustarini D, Colombo R, Rossi R, Milzani A (2003) Protein carbonyla-tion in human diseases. Trends Mol Med 9:169–176.

8. Schulz TJ, et al. (2007) Glucose restriction extends Caenorhabditis elegans life span byinducing mitochondrial respiration and increasing oxidative stress. Cell Metab 6:280–293.

9. Yang W, Hekimi S (2010) A mitochondrial superoxide signal triggers increased lon-gevity in Caenorhabditis elegans. PLoS Biol 8:e1000556.

10. Yang W, Hekimi S (2010) Two modes of mitochondrial dysfunction lead in-dependently to lifespan extension in Caenorhabditis elegans. Aging Cell 9:433–447.

11. Pan Y, Schroeder EA, Ocampo A, Barrientos A, Shadel GS (2011) Regulation of yeastchronological life span by TORC1 via adaptive mitochondrial ROS signaling. CellMetab 13:668–678.

12. Tatar M, et al. (2001) A mutant Drosophila insulin receptor homolog that extends life-span and impairs neuroendocrine function. Science 292:107–110.

13. Sim C, Denlinger DL (2013) Insulin signaling and the regulation of insect diapause.Front Physiol 4:189.

14. Hedrick SM, Hess Michelini R, Doedens AL, Goldrath AW, Stone EL (2012) FOXOtranscription factors throughout T cell biology. Nat Rev Immunol 12:649–661.

15. Broughton SJ, et al. (2005) Longer lifespan, altered metabolism, and stress resistancein Drosophila from ablation of cells making insulin-like ligands. Proc Natl Acad SciUSA 102:3105–3110.

16. Manning BD, Cantley LC (2007) AKT/PKB signaling: Navigating downstream. Cell 129:1261–1274.

17. de Keizer PLJ, Burgering BMT, Dansen TB (2011) Forkhead box o as a sensor, medi-ator, and regulator of redox signaling. Antioxid Redox Signal 14:1093–1106.

18. Bao B, Xu WH (2011) Identification of gene expression changes associated with theinitiation of diapause in the brain of the cotton bollworm, Helicoverpa armigera.BMC Genomics 12:224.

19. Lu YX, Xu WH (2010) Proteomic and phosphoproteomic analysis at diapause initiationin the cotton bollworm, Helicoverpa armigera. J Proteome Res 9:5053–5064.

20. Zhang Q, Lu YX, Xu WH (2012) Integrated proteomic and metabolomic analysis oflarval brain associated with diapause induction and preparation in the cotton boll-worm, Helicoverpa armigera. J Proteome Res 11:1042–1053.

21. Zhang Q, Lu YX, Xu WH (2013) Proteomic and metabolomic profiles of larval he-molymph associated with diapause in the cotton bollworm, Helicoverpa armigera.BMC Genomics 14:751.

22. Xu WH, Lu YX, Denlinger DL (2012) Cross-talk between the fat body and brain reg-ulates insect developmental arrest. Proc Natl Acad Sci USA 109:14687–14692.

23. Zarse K, et al. (2012) Impaired insulin/IGF1 signaling extends life span by promotingmitochondrial L-proline catabolism to induce a transient ROS signal. Cell Metab 15:451–465.

24. Bedford MT, Richard S (2005) Arginine methylation an emerging regulator of proteinfunction. Mol Cell 18:263–272.

25. Kops GJ, et al. (1999) Direct control of the Forkhead transcription factor AFX byprotein kinase B. Nature 398:630–634.

26. Yamagata K, et al. (2008) Arginine methylation of FOXO transcription factors inhibitstheir phosphorylation by Akt. Mol Cell 32:221–231.

27. Dillon MBC, et al. (2012) Novel inhibitors for PRMT1 discovered by high-throughputscreening using activity-based fluorescence polarization. ACS Chem Biol 7:1198–1204.

28. Leto D, Saltiel AR (2012) Regulation of glucose transport by insulin: Traffic control ofGLUT4. Nat Rev Mol Cell Biol 13:383–396.

29. Sim C, Denlinger DL (2008) Insulin signaling and FOXO regulate the overwinteringdiapause of the mosquito Culex pipiens. Proc Natl Acad Sci USA 105:6777–6781.

30. Sim C, Kang DS, Kim S, Bai X, Denlinger DL (2015) Identification of FOXO targets thatgenerate diverse features of the diapause phenotype in the mosquito Culex pipiens.Proc Natl Acad Sci USA 112:3811–3816.

31. Martins R, Lithgow GJ, Link W (2016) Long live FOXO: Unraveling the role of FOXOproteins in aging and longevity. Aging Cell 15:196–207.

32. Finkel T (2000) Redox-dependent signal transduction. FEBS Lett 476:52–54.33. Shigenaga MK, Hagen TM, Ames BN (1994) Oxidative damage and mitochondrial

decay in aging. Proc Natl Acad Sci USA 91:10771–10778.34. Lee SJ, Hwang AB, Kenyon C (2010) Inhibition of respiration extends C. elegans life

span via reactive oxygen species that increase HIF-1 activity. Curr Biol 20:2131–2136.35. Zheng WH, Kar S, Quirion R (2000) Insulin-like growth factor-1-induced phosphory-

lation of the forkhead family transcription factor FKHRL1 is mediated by Akt kinase inPC12 cells. J Biol Chem 275:39152–39158.

36. Zhang J, Zhang LL, Shen L, Xu XM, Yu HG (2013) Regulation of AKT gene expressionby cisplatin. Oncol Lett 5:756–760.

37. Le Belle JE, et al. (2011) Proliferative neural stem cells have high endogenous ROSlevels that regulate self-renewal and neurogenesis in a PI3K/Akt-dependant manner.Cell Stem Cell 8:59–71.

38. Chetram MA, et al. (2013) ROS-mediated activation of AKT induces apoptosis viapVHL in prostate cancer cells. Mol Cell Biochem 376:63–71.

39. Accili D, Arden KC (2004) FoxOs at the crossroads of cellular metabolism, differenti-ation, and transformation. Cell 117:421–426.

40. Takahashi Y, et al. (2011) Asymmetric arginine dimethylation determines life span inC. elegans by regulating forkhead transcription factor DAF-16. Cell Metab 13:505–516.

41. Cheng CM, et al. (2000) Insulin-like growth factor 1 regulates developing brain glu-cose metabolism. Proc Natl Acad Sci USA 97:10236–10241.

42. Kupriyanova TA, Kandror KV (1999) Akt-2 binds to Glut4-containing vesicles andphosphorylates their component proteins in response to insulin. J Biol Chem 274:1458–1464.

43. Harman D (1956) Aging: A theory based on free radical and radiation chemistry.J Gerontol 11:298–300.

44. Morais VA, et al. (2014) PINK1 loss-of-function mutations affect mitochondrial com-plex I activity via NdufA10 ubiquinone uncoupling. Science 344:203–207.

45. Putker M, et al. (2013) Redox-dependent control of FOXO/DAF-16 by transportin-1.Mol Cell 49:730–742.

46. Scialò F, et al. (2016) Mitochondrial ROS produced via reverse electron transport ex-tend animal lifespan. Cell Metab 23:725–734.

47. Velarde MC, Demaria M, Melov S, Campisi J (2015) Pleiotropic age-dependent effectsof mitochondrial dysfunction on epidermal stem cells. Proc Natl Acad Sci USA 112:10407–10412.

48. Lin XW, Xu WH (2016) Hexokinase is a key regulator of energy metabolism and ROSactivity in insect lifespan extension. Aging (Albany NY) 8:245–259.

49. Ristow M, et al. (2009) Antioxidants prevent health-promoting effects of physicalexercise in humans. Proc Natl Acad Sci USA 106:8665–8670.

Zhang et al. PNAS Early Edition | 9 of 9

PHYS

IOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

May

1, 2

020