Transcriptional profiles of Type I dInr^E19/dInr⁷⁴ and Type II dInr³⁵³/dInr^wt

We conducted RNA-seq from somatic tissue of 10-day-old mated females of dInr^E19/dInr⁷⁴ (Type I), dInr³⁵³/dInr^wt (Type II), and coisogenic wildtype (dInr^wt/wt line ‘29B’) (Fig 1A and S1 Table). 862 transcripts were differentially expressed among the three genotypes (iDEP http://bioinformatics.sdstate.edu/idep96/). 191 differentially expressed genes relative to wildtype were common to dInr^E19/dInr⁷⁴ and dInr³⁵³/dInr^wt. Relative to wildtype, adult dInr^E19/dInr⁷⁴ uniquely altered 439 genes, while 233 were uniquely expressed by dInr³⁵³/dInr^wt. The first principal component axis separated all the genotypes, while on principal component axis 2, the two longevity genotypes were similar but distinct from wildtype (Fig 1B).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Transcriptional profiles of dInr^E19/dInr⁷⁴ and dInr³⁵³/dInr^wt.

(A) Venn diagram of transcripts altered in Type I dInr^E19/dInr⁷⁴ and Type II dInr³⁵³/dInr^wt genotypes. (B) Principal component analysis of transcripts profiled in dInr^E19/dInr⁷⁴ and dInr³⁵³/dInr^wt. (C) Heatmap of transcripts altered in dInr^E19/dInr⁷⁴ and dInr³⁵³/dInr^wt. (D) Gene set enrichment analysis of transcripts altered in dInr³⁵³/dInr^wt. (E) Gene set enrichment analysis of transcripts altered in dInr^E19/dInr⁷⁴. (F) Survival of dInr^E19/dInr⁷⁴, dInr³⁵³/dInr^wt, and coisogenic wildtype (dInr^wt/wt) infected with ECC15 bacteria. (G) Down- and up-regulated transcription factor binding sites in dInr^E19/dInr⁷⁴ and dInr³⁵³/dInr^wt.

https://doi.org/10.1371/journal.pgen.1011640.g001

K-means clustering and gene set enrichment describe how these genotypes differ based on function (Fig 1C–1E and S2 Table). Type I dInr^E19/dInr⁷⁴ elevates defense response genes and reduces transcripts for egg production, consistent with their low fecundity [4]. In contrast, defense response is weakly amplified in dInr³⁵³/dInr^wt, while this genotype increases transcripts associated with somatic activity to support egg production, consistent with their high fecundity. To evaluate the impact of altered defense gene expression, we orally infected adults with ECC15 bacteria. Type I dInr^E19/dInr⁷⁴ adults resisted infection and survived as well as uninfected controls. In contrast, dInr³⁵³/dInr^wt suffered infection-induced mortality to the same extent as infected wildtype adults (Fig 1F). Thus, altered innate immunity is not necessary for dInr mutations to extend longevity.

Differences in mRNA among the genotypes arise in part from how dInr regulates transcription factors. We therefore used iDEP to identify transcription factor binding sites (Fig 1G) that are enriched in the mutants relative to wildtype. Four TF motifs were under-represented (down) in dInr³⁵³/dInr^wt including daughterless (da), which was the only site also under-represented in dInr^E19/dInr⁷⁴. The dInr mutants shared many over-represented (up) TF binding motifs relative to wildtype exception for tango, which was under-represented in dInr^E19/dInr⁷⁴ yet over-represented in dInr³⁵³/dInr^wt, and four sites that were unique to either dInr^E19/dInr⁷⁴ (HLH106, dl) or dInr³⁵³/dInr^wt (Eip75B, sim). Notable for their absence were binding sites for the transcription factors FOXO and Aop (Anterior open), which are recognized to function downstream of insulin signaling in Drosophila to control aging [14,15]. Genetic epistasis analyses will be required to robustly evaluate if these expected transcription factors are required for the Type I or Type II dInr genotypes to extend lifespan.

Both dInr types altered the mRNA of genes previously associated with extended Drosophila lifespan. Up-regulated longevity genes in both dInr³⁵³/dInr^wt and dInr^E19/dInr⁷⁴ include tequilla, Sodh-1, and Fbp1 [16–18]. Shared down-regulated mRNA transcripts include ImpL2, which binds insulin ligands and extends lifespan when over-expressed [19–21], and la costa (lcs), which is suppressed by AOP when the activity of this transcription factor contributes to extended lifespan [14,22].

Adults of dInr^E19/dInr⁷⁴ depress dawdle (daw) and adipokinetic hormone (Akh). Dawdle is modulated by FOXO when reduced insulin signaling extends lifespan [15]. Adipokinetic hormone has glucagon-like functions that enhance lifespan by regulating water homeostasis and metabolism [8,15,23]. Glycine N-methyltransferase (gnmt) was induced in dInr^E19/dInr⁷⁴ but not in dInr³⁵³/dInr^wt. GNMT uses S-adenosylmethionine as a cofactor to methylate glycine into sarcosine [24]. Notably, over-expression of gnmt extends lifespan of wildtype adults, and the gene is required for reduced insulin activity (via a transgene for a dominant negative dInr) to slow aging [25,26]. We likewise find dInr^E19/dInr⁷⁴ increases insulin-like peptides ilp3 and ilp5 mRNA, indicating that these adults are hyper-insulinemic.

Few transcripts with recognized impact on aging change exclusively in dInr³⁵³/dInr^wt. Rather, many DEGs of dInr³⁵³/dInr^wt are associated with egg production; however the transcription factor daughterless (da) is reduced only in dInr³⁵³/dInr^wt, although its transcription factor binding site is under-represented in both types of dInr. dInr³⁵³/dInr^wt also upregulates several Tot secreted stress-response proteins (TotA, TotC, and TotX), while it downregulates the transcripts for the innate immune proteins DptA, Drosomycin, and Edin.

Overall, while Type I and Type II dInR mutants have several common downstream transcriptional impacts, the dInr³⁵³/dInr^wt mutant drives a program distinct from insulin-resistant Type I adults.

Metabolomic profiles of dInr^E19/dInr⁷⁴ and dInr³⁵³/dInr^wt

To study the metabolic consequences of altered insulin receptor function, we performed targeted steady-state metabolite profiling of the insulin-resistant dInr^E19/dInr⁷⁴ and the insulin-sensitive dInr³⁵³/dInr^wt at different ages. Somatic tissues of 15-day-old and 30-day-old-females were analyzed, with six biological replicates (each of 10 females without ovaries) of both genotypes. We first considered the 15-day-olds alone. Here, the three genotypes can be distinguished with principal component analysis (PCA) (S1 Fig) based on the 170 detected metabolites within the target set (361 metabolites). Relative to wildtype, 26 metabolites were uniquely altered in dInr³⁵³/dInr^wt and 38 were differentially abundant just in dInr^E19/dInr⁷⁴ (FDR < 0.05). Seventeen metabolites were similarly affected in the same direction in both genotypes (Fig 2A and 2B and S3 Table).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Steady state metabolomic profiles of 15-day old females of dInr^E19/dInr⁷⁴ and dInr³⁵³/dInr^wt.

(A) Heatmap of the most altered metabolites in dInr^E19/dInr⁷⁴ and dInr³⁵³/dInr^wt. (B) Venn diagram of metabolites significantly altered in dInr^E19/dInr⁷⁴ and dInr³⁵³/dInr^wt. (C) Metabolite set enrichment of metabolites altered in dInr³⁵³/dInr^wt. (D) Metabolite set enrichment of metabolites altered in dInr^E19/dInr⁷⁴. (E) Scheme of methionine metabolism. Significantly (FDR < 0.1) altered metabolites relative to wildtype are marked with “a” for dInr³⁵³/dInr^wt and “b” for dInr^E19/dInr⁷⁴. (F) Relative levels of methionine in dInr^E19/dInr⁷⁴, dInr³⁵³/dInr^wt, and coisogenic wildtype (dInr^wt/wt). (G) Relative levels of cysteine in dInr^E19/dInr⁷⁴, dInr³⁵³/dInr^wt, and coisogenic wildtype (dInr^wt/wt).

https://doi.org/10.1371/journal.pgen.1011640.g002

Metabolite Set Enrichment Analysis distinguishes pathways affected by each mutant. Type II dInr³⁵³/dInr^wt impacts ‘aspartate metabolism’ and ‘methionine metabolism’. Insulin-resistant dInr^E19/dInr⁷⁴ mutants affect ‘taurine and hypotaurine metabolism’, and ‘phosphatidylcholine biosynthesis’ (Fig 2C and 2D). Notably, these longevity genotypes intersect at methionine metabolism (Fig 2E) but in strikingly different ways. At 15 days of age, dInr³⁵³/dInr^wt has reduced steady-state methionine, methionine sulfoxide, and cystathionine (Figs 2E, 2F and S1), while dInr^E19/dInr⁷⁴ has elevated cysteine, hypotaurine, taurine, and adenosine, and reduced glutathione (Figs 2E, 2G and S1). Sarcosine is also strongly enriched (7-fold) in dInr^E19/dInr⁷⁴ yet is less abundant in dInr³⁵³/dInr^wt adults. Sarcosine is notable because this metabolite is generated by GNMT, and we found gnmt mRNA is markedly greater in dInr^E19/dInr⁷⁴ (S1 Table). Previous work has established that gnmt plays a role in how repressed insulin signaling extends lifespan [25,26]. Overall, Type I and Type II dInr mutants produce different patterns of reprogrammed methionine metabolism.

Age dynamics of dInr^E19/dInr⁷⁴ and dInr³⁵³/dInr^wt metabolic profiles

Metabolic profiles change with age [27], including the profile of methionine metabolism in Drosophila [28,29]. We therefore analyzed how aging affects metabolic changes in dInr^wt/dInr^wt, dInr^E19/dInr⁷⁴, and dInr³⁵³/dInr^wt females. Fig 3 (Fig 3C–3J) normalizes metabolites to their level at 15-days old; day 15 is set at 100%, and the value at day 30 reflects the change relative to day 15. The direction of these changes within the methionine network is summarized in Fig 3B. Relative to young adults, methionine, sarcosine, and cystathionine decreased with age in wildtype dInr^wt/dInr^wt (Fig 3C, 3D and 3E), as previously reported for wildtype OregonR, yw, and w¹¹¹⁸ flies [28,30]. The elevated methionine and sarcosine seen in young dInr^E19/dInr⁷⁴ adults decreased with age (Fig 3C and 3D). In contrast, the steady-state level of methionine was low in young dInr³⁵³/dInr^wt adults and remained low in aged adults (Fig 3C). SAM increased with age in wildtype flies but remained constant across age in dInr^E19/dInr⁷⁴ and dInr³⁵³/dInr^wt (Fig 3F). The level of methionine sulfoxide was relatively low in young dInr³⁵³/dInr^wt adults and significantly increased as this genotype aged, but the metabolite remained constant in dInr^E19/dInr⁷⁴ and wildtype adults as they aged (Fig 3G).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Age-associated metabolic profiles of dInr^E19/dInr⁷⁴ and dInr³⁵³/dInr^wt.

(A) Heatmap of age-associated metabolites among wildtype (dInr^wt/wt), dInr^E19/dInr⁷⁴, or dInr³⁵³/dInr^wt. (B) Methionine metabolism network where significant age-associated changes at day 30 are indicated for each genotype. (C-J) The level of methionine metabolism at day 30 relative to the level within each genotype at day 15 (all values normalized to the mean at day 15): (C) methionine, (D) sarcosine, (E) cystathionine, (F) SAM, (G) methionine sulfoxide, (H) cysteine, (I) taurine, and (J) hypotaurine.

https://doi.org/10.1371/journal.pgen.1011640.g003

There are notable differences among genotypes in how metabolites of the transsulfuration pathway change with age. Aged wildtype flies have reduced levels of cystathionine but elevated hypotaurine (Fig 3E and 3H–3J), while these metabolites were initially elevated in young dInr^E19/dInr⁷⁴ adults and remained high in aged adults (Fig 3E and 3H–3J). In contrast, all assayed transsulfuration metabolites were initially low but then markedly increased with age in dInr³⁵³/dInr^wt adults (Fig 3E and 3H–3J).

The dInr genotypes have different age-associated changes in methionine metabolism. Female dInr^E19/dInr⁷⁴ mutants enhance the transsulfuration pathway as young adults and maintain this level with age. Young insulin-sensitive dInr³⁵³/dInr^wt adults produce few transsulfuration metabolites but increase these molecules as they age.

Methionine cycle dynamics of dInr^E19/dInr⁷⁴ and dInr³⁵³/dInr^wt

Tracking ¹³C can estimate the transition rates across steps in the methionine cycle and quantify the relative flow of methionine derivatives toward nucleotides, polyamines, and the transsulfuration pathway [31,32]. In this approach, carbon isotope-labeled nutrients are fed to adults, and the source of carbons within specific metabolites is estimated from the quantity of labeled and unlabeled carbons [33–35].

We fed young dInr adults with stable isotope-labeled methionine (¹³C5-methionine, five labeled carbons, m + 5) and measured how these carbons were incorporated into methionine that is regenerated from homocysteine, which is seen as methionine m + 4 (¹³C4-methionine, four labeled carbons: m + 4) [28,30] (Fig 4A). We find that the ratio of methionine m + 4/m + 5 is higher in dInr^E19/dInr⁷⁴ than in wildtype (Fig 4B), indicating that the activity of the central methionine cycle is greater in these insulin-resistant females than in wildtype. This high activity can account for the elevated steady-state sarcosine and transsulfuration pathway products of the Type I adults and implies that somatic cells of these mutants may need to cope with abundant methionine. In contrast, the ratio of m + 4/m + 5 methionine is reduced in dInr³⁵³/dInr^wt relative to wildtype (Fig 4B), indicating the somatic tissue of these adults has lower activity through their central methionine cycle. Consequently, they produce less Hcy and recycle less methionine.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Methionine dynamics from isotope tracking in dInr^E19/dInr⁷⁴ and dInr³⁵³/dInr^wt.

(A) Scheme of methionine labeling. Blue marks the labeled carbons in each metabolite derived from the labeled dietary methionine, which has five ¹³C (m + 5). (B) The ratio between methionine (m + 4) recovered from the methionine cycle and methionine (m + 5) derived from the diet. Differences among the ratios reflect significant differences (ANOVA test with posthoc correction for multiple hypothesis testing) in the activity of the central methionine cycle among wildtype (dInr^wt/wt), dInr^E19/dInr⁷⁴, and dInr³⁵³/dInr^wt. ***, p < 0.001; ****, p < 0.0001. (C) The distribution of labeled methionine in young and old female wildtype (dInr^wt/wt), dInr^E19/dInr⁷⁴, and dInr³⁵³/dInr^wt, referenced to the methionine cycle, polyamine cycle and transsulfuration pathway. The number of labeled carbons in each metabolite is indicated as M0-M5. Bars reflect the relative proportion of each labeled isoform (labeling enrichment) within metabolites at each age and genotype.

https://doi.org/10.1371/journal.pgen.1011640.g004

Although we cannot directly estimate flux to the transsulfuration pathway, it seems unlikely that methionine in dInr³⁵³/dInr^wt is routed through that pathway because we do not observe elevated taurine or glutathione in the steady-state metabolome. Likewise, roles for the salvage pathway are also difficult to evaluate because metabolites were not well represented in our steady-state data. Some insight is gained, however, from methionine carbons traced into the salvage pathway of young and aged flies. Dietary methionine is first converted to S-adenosyl-L-methionine (SAM), which can be decarboxylated by AdoMet decarboxylase into decarboxylated SAM (dcSAM), an aminopropyl group donor. Spermidine synthase and spermine synthase use this aminopropyl donor to convert putrescine to spermidine and spermine, with dcSAM converted to MTA. MTA then undergoes six enzymatic steps to regenerate methionine, with 4-methylthio-2-oxobutanoic acid (MTOB) as one intermediary. In dInr³⁵³/dInr^wt adults, we see that MTA and MTOB are labeled with one carbon (m + 1) derived from methionine, which can arise from m + 4 or m + 5 labeled methionine (Fig 4C). Furthermore, we see a significantly elevated fraction of m + 1 labeled MTA (in young and old flies) and of m + 1 labeled MTOB (in old flies), suggesting the dInr³⁵³/dInr^wt genotype may indeed route methionine towards polyamine production.

Our technical platform did not detect differential labeling of SAM and cystathionine. Likewise, we could detect labeled SAH only in old flies, perhaps because SAH is intrinsically elevated with age [28]. SAH generated directly from m + 5 or m + 4 methionine is labeled as m + 4 SAH. SAH generated from methionine that passed through the salvage pathway is labeled as m + 1 SAH (via m + 1 methionine). Notably, we found an elevated fraction of m + 1 SAH in dInr³⁵³/dInr^wt adults and almost no fraction of m + 4 SAH. These results imply that carbons from methionine in dInr³⁵³/dInr^wt adults move through the salvage pathway first and then enter the methionine cycle, consistent with the m + 1 labeling of salvage pathway intermediaries (MTA and MTOB).

Overall, dInr³⁵³/dInr^wt adults have two unique features of methionine metabolism: their central methionine cycle has reduced activity, as occurs in dietary methionine restriction, and they activate the methionine salvage pathway, which is expected to generate polyamines.

Interactions between dInr³⁵³/dInr^wt and genes of the methionine cycle

Previous work established that reduced insulin signaling confers longevity in part because it induces gnmt, which itself is required for dInr genotypes to extend lifespan [25,26]. Consistent with those reports, we find that dInr^E19/dInr⁷⁴ increases gnmt mRNA and sarcosine, the product of GNMT enzymatic activity. In contrast, these features were not seen in dInr³⁵³/dInr^wt, and correspondingly, knockdown of gnmt mRNA in dInr³⁵³/dInr^wt did not inhibit the exceptional longevity of the Type II mutant (Fig 5A). We therefore tested other components of the methionine cycle. As noted, S-adenosyl-methionine (SAM) was reduced in dInr³⁵³/dInr^wt. SAM has been recently shown in mammalian cells to modulate TOR activity by disrupting the SAMTOR-GATOR1 complex [36], although a more recent report suggests a gene called ‘unmet’, not SAMTOR, is the SAM sensor in Drosophila [37]. In mammals, reduced SAM promotes the association of SAMTOR with GATOR1 and inhibits TORC1. Thus, reduced SAM may promote longevity by activating SAMTOR. Based on the initial report of SAMTOR, we investigated the interaction of samtor and dInr³⁵³/wt in Drosophila by depleting samtor in dInr mutant and wildtype adults. This did not diminish the extended longevity of dInr³⁵³/dInr^wt females (Fig 5B) while in males, samtor-RNAi alone repressed survival in wildtype flies (Fig 5E). To test if dInr³⁵³ otherwise interacts with TOR, we overexpressed activated S6K (UAS-S6K-TE) in dInr³⁵³ and wildtype adults. This manipulation depressed survival in all genotypes but did not eliminate the relative longevity benefit of dInr³⁵³ (Fig 5C and 5F). Downstream in the central methionine cycle, S-adenosyl-homocysteine (SAH) is generated when SAM contributes a methyl group to substrates [38]. SAH itself is an inhibitor of SAM-dependent methylation and is therefore robustly regulated by the activity of AHCY [39]. Here, we find that depletion of Ahcy13 in the dInr³⁵³/wt background eliminates the survival advantage of the dInr mutant relative to similarly treated wildtype. While this depletion reduces the lifespan of all genotypes, as anticipated from previous work [28], we conclude that dInr³⁵³/wt requires Ahcy13 to extend lifespan relative to dInr wildtype. Reduced flux through SAH is required for the dInr³⁵³/dInr^wt mutant to extend lifespan: Type II dInr³⁵³/dInr^wt slows aging by reprogramming the methionine cycle.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Epistasis analysis for interactions between elements of the methionine cycle and the dInr KID mutant that affect survival.

Kaplan-Meier plots and Cox proportional hazard to evaluate the epistatic impact on survival of the dInr genotype, the targeted methionine cycle-associated gene and the interaction between these factors. (A) Survival of female wildtype (dInr^wt/wt) and dInr³⁵³/dInr^wt combined with gnmt RNAi. (B) Survival of female wildtype (dInr^wt/wt) and dInr³⁵³/dInr^wt combined with samtor RNAi. (C) Survival of female wildtype (dInr^wt/wt) and dInr³⁵³/dInr^wt combined with S6K-activated transgene. (D) Survival of female wildtype (dInr^wt/wt) and dInr³⁵³/dInr^wt combined with Ahcy13 RNAi. (E) Survival of male wildtype (dInr^wt/wt) and dInr³⁵³/dInr^wt combined samtor RNAi. (F) Survival of male wildtype (dInr^wt/wt) and dInr³⁵³/dInr^wt combined with S6K-activated transgene. (G) Survival of male wildtype (dInr^wt/wt) and dInr³⁵³/dInr^wt combined with Ahcy13 RNAi.

https://doi.org/10.1371/journal.pgen.1011640.g005

Type I and II dInr

Type I and II insulin-like receptor mutants extend lifespan, yet these genotypes differ in many ways. Type I dInr genotypes are manipulations that reduce insulin signaling and extend lifespan but retard growth, slow development, repress fecundity, and cause insulin resistance. Genetically, they are homozygous lethal and extend longevity only as transheterozygotes. Yamamoto et al. [4] identified three alleles with these characteristics: dInr^E19, dInr²¹¹, and dInr⁷⁴. The dInr^E19 allele was originally reported to increase lifespan and to reduce insulin receptor phosphorylation when combined with an enhancer trap insertion in the first coding exon [40]. The dInr^E19 allele substitutes aspartic acid for valine in the linker sequence between the L2 and FnIII-1 ectodomains. dInr⁷⁴ generates a missense mutation in the conserved activation loop of the intracellular kinase domain and likely reduces the ability of the kinase to be transactivated. The dInr²¹¹ allele substitutes glycine with arginine at residue 1598 of the kinase domain C-terminal lobe. It extends lifespan when combined with dInr^E19 or with dInr⁷⁴, where all transheterozygotes are insulin resistant.

In contrast, the Type II dInr mutant of the KID, when present as a heterozygote with a wildtype allele (dInr³⁵³/dInr^wt), has normal growth, high fecundity, robust insulin sensitivity, and extended lifespan. KIDs are unstructured peptide sequences that link the kinase domain N- and C-lobes in tyrosine kinase receptors [13]. KIDs often contain binding motifs for adaptor proteins, although none are reported in the KID of Drosophila dInr, which is relatively small (27 residues), or in the KIDs of mammalian IR and IGF1R (both with 15 residues). These insulin/IGF KIDs have conserved residues at the start of the KID, including the arginine in Drosophila (R1466) that extends longevity in dInr³⁵³ when substituted for cysteine. No function to date has been identified for the KID of any insulin/IGF-like receptor. We therefore characterized transcriptional and metabolomic profiles of the Type II dInr relative to wildtype and the Type I mutants. We found the altered KID of dInr produces a methionine restriction-like state and showed that this mutant receptor genetically interacts with a key enzyme of the methionine cycle to slow aging.

Transcriptional profiles of insulin-like receptor genotypes

C. elegans and Drosophila have been characterized for how mutant insulin receptors (daf-2, dInr) impact mRNA profiles and enrich for Daf-16/FOXO transcription factor binding sites [15,41]. Working with variants of daf-2, Patel et al. [11] described how different alleles impact mRNA levels. The weak allele daf-2(m577) modestly increases C. elegans lifespan and mildly affects the temperature at which animals enter dauer, while the often-studied daf-2(e1370) allele robustly increases lifespan, strongly affects dauer, and reduces feeding and movement [10]. Patel et al. [11] discriminated many differentially expressed genes among these longevity alleles and how these patterns depended on daf-16. Transcripts upregulated by daf-2(e1370) that were not affected in daf-2(m577) included small heat shock proteins, the insulin ligand ins-22, and the nuclear hormone receptor nhr-206. These genes may contribute to how daf-2(e1370) slows aging in ways that are distinct from the action of the daf-2(m577) allele.

Here, we discriminate among the differentially expressed genes of Drosophila dInr longevity mutants. Type I dInr^E19/dInr⁷⁴ and Type II dInr³⁵³/dInr^wt together expressed 862 different genes relative to wildtype. The Type II KID mutant dInr³⁵³/dInr^wt uniquely altered 233 transcripts that were not seen to change in the Type I mutants. Principal components axis 1 (PC1) partitioned all the dInr genotypes, while the mutant genotypes were similar on PC2 but together distinct from wildtype. Type I and Type II dInr mutants produce common and unique transcriptional changes that may contribute to extended lifespan.

Tequila is one transcript elevated in both dInr genotypes. Previous work found that a tequila hypomorph extended Drosophila lifespan, perhaps because it reduces neuronal insulin secretion [17]. In contrast, the dInr mutants may elevate tequila to increase insulin secretion as a mechanism to compensate for their mutated receptor function. Compensation may also explain why ImpL2 is reduced in both genotypes: ImpL2 is a peptide-binding protein that inhibits the biological availability of Drosophila insulin-like peptides [19]; limiting this inhibitor may increase the bioavailability of insulin ligand. Consistent with these explanations, mRNA for insulin-like ligands ilp3 and ilp5 is elevated in Type I, although not in Type II. Overall, Type I dInr adults in particular have three features associated with human pre-diabetes: hyperinsulinemia, mechanisms to increase insulin peptide bioavailability, and impaired insulin receptor sensitivity [4].

Enzymes that mediate oxidative stress are elevated in many longevity-conferring manipulations of Drosophila and C. elegans [18], and we see elevated Sod-1 in both types of long-lived dInr mutants. However, whether increased Sod-1 is sufficient to slow aging in Drosophila or mice is unresolved because its over-expression does not consistently extend lifespan [42]. Finally, fat body protein (fbp) was elevated in both types of dInr mutants. Fat body protein was recently implicated as a factor that limits lifespan when induced by juvenile hormone in young adults, while transiently blocking fbp in young adults improves life expectancy [16]. Thus, increased fbp in long-lived dInr genotypes, as seen here, is somewhat counter to expectations.

Few previously recognized aging-associated genes changed only in the dInr³⁵³/dInr^wt adults. One exception involves the transcription factor daughterless (da): da expression decreased, and we found few binding sites for this transcription factor among the differentially expressed genes of dInr³⁵³/dInr^wt. Notably, depletion of the daughterless ortholog HLH-2 in C. elegans extends nematode lifespan [43].

There is a striking difference in innate immune-related genes among the dInr longevity genotypes. In previous studies, the expression of antimicrobial peptides and peptidoglycan recognition proteins declines with age, while dietary and genetic manipulations that extend Drosophila longevity often maintain these defenses [44]. These observations suggest that elevated innate immunity across age favors Drosophila longevity. Here, we see innate immune-associated genes are elevated in Type I dInr^E19/dInr⁷⁴ but not in Type II, and these expression differences are functional because they correspond with the ability of adults to resist oral infection with ECC15 bacteria. This outcome demonstrates that elevated innate immunity is not necessary for altered insulin signaling to extend lifespan.

Overall, while both dInr^E19/dInr⁷⁴ and dInr³⁵³/dInr^wt extend lifespan, the genotypes express sets of distinct and similar genes. Remarkably, their differences are caused by the placement of single amino acid substitutions within the insulin-like receptor. Deeper analysis will be required to understand how these variants differentially transduce through signaling cascades, including the roles of the transcription factors FOXO and Aop.

Methionine metabolism of insulin-like receptor mutants

Methionine restriction and aging

Dietary methionine restriction (MetR) slows aging across many species, including rodents and Drosophila [53–56]. Likewise, knockdown of C. elegans S-adenosyl methionine synthetase (sams-1) increases adult lifespan [57], and Drosophila lifespan is increased by 25% by expressing an exogenous Methioninase that degrades methionine [29]. MetR decreases levels of intermediates in the methionine metabolism pathway, including methionine, methionine sulfoxide, SAM, SAH, cystathionine, and 5’-methylthioadenosine (MTA) [29,30,58]. Profiles of methionine metabolism are also altered in long-lived species and strains, such as naked mole-rats [59], flies selected for delayed reproductive senescence [28,60], and Ames mice [61]. Across these cases, methionine restriction is thought to slow aging through many potential avenues, including protein and nucleotide methylation, polyamine production, protein translation, and energetic or oxidative status [51].

Steady-state metabolomics and tracing analysis demonstrate that young Type II adults have attributes of methionine restriction. The long-lived dInr³⁵³/dInr^wt adults have less methionine and less steady-state methionine sulfoxide. Type II adults also have increased m + 1 labeling in the salvage pathway and at SAH, and elevated transsulfuration metabolites as they age.

Genetic epistasis analysis confirms that dInr³⁵³/dInr^wt extends longevity because it modifies methionine metabolism. Unlike gnmt, which has been reported to be required for repressed insulin signaling to slow aging [25,26], the extended longevity of dInr³⁵³/dInr^wt is independent of gnmt. As well, dInr³⁵³/dInr^wt did not interact with SAMTOR or S6K to slow aging, suggesting its impact is independent of TOR (although the utility of testing SAMTOR in Drosophila is ambiguous). Rather, dInr³⁵³/dInr^wt requires ahcy13. AHCY-13 encodes SAHH (S-adenosylhomocysteine hydrolase), which converts SAH into homocysteine. Emerging work shows that AHCY itself modulates aging: Drosophila lifespan was increased by downregulating genetic repressors of AHCY (AhcyL1, AhcyL2) [28], and C. elegans lifespan was increased by a deficiency of AHCY-1 that increases the ratio of SAH to SAM [62]. Cells normally maintain low concentrations of SAH because it is an inhibitor of SAM-dependent methylation. Inhibition of AHCY blocks SAH clearance and prevents the flux of methionine through the methionine cycle and into the transsulfuration pathway. We have previously demonstrated that downregulation of Ahcy13 in Drosophila upregulates SAH levels [28], which strongly inhibits methionine flux. By inhibiting Ahcy13 in wildtype and dInr³⁵³ flies, we asked if the observed changes in methionine metabolism are required for the extended lifespan of dInr³⁵³/dInr^wt. Depleting ahcy13 shortens dInr³⁵³/dInr^wt lifespan to the level of similarly manipulated wildtype controls. We conclude that dInr³⁵³/dInr^wt extends longevity, in part, because it alters methionine metabolism. This metabolic programming has three impacts: i) methionine is routed towards the salvage/polyamine pathway; ii) the flux of the central via methionine cycle is reduced, and iii) old adults can upregulate products of the transsulfuration pathway.

Alternative modes of longevity control by insulin/IGF signaling

Canonical Type I insulin mutants elevate the products of the TSP relative to wildtype at both young and older ages. These mutants have reduced fecundity, which perhaps directs methionine from eggs into the TSP of somatic cells. Thus, the dInr^E19/dInr⁷⁴ mutant elevates cysteine, taurine, and reduced-glutathione, which may favor somatic maintenance by managing ROS [49,63–65]. The Type II mutant at the Kinase Insert Domain is substantially different. The females are very fecund, and young adults have reduced somatic methionine activity and few products of the TSP. Yet, their mortality is remarkably low [4]. This result may arise through longevity assurance mechanisms associated with polyamine synthesis. As well, the methionine metabolism of these flies is flexible, where aged Type II adults increase their production of TSP. Remarkably, the different Type I and Type II metabolic syndromes arise from the placement of a single amino acid substitution within the insulin-like receptor. Mutation of the kinase insert domain, which produces insulin sensitive receptors, reduces somatic methionine cycle activity in young adults, while canonical, insulin resistant longevity alleles of dInr have excess methionine.

The ability to efficiently adapt metabolism by substrate utilization in response to nutrient intake and timing is a form of metabolic flexibility. Metabolic flexibility was first described in lean humans as they alter the fuels they burn when fasting when simultaneously they are infused with insulin, in contrast to the static fuel use seen in obese individuals [66,67]. Metabolic flexibility may explain why mild insulin resistance, which is seen in a variety of long-lived mice, is not necessarily an indicator of poor health [68]. While the concept of metabolic flexibility is typically applied on the scale of daily physiology as tissues transit between carbohydrate and lipid metabolism, here we envision metabolic flexibility on the scale of aging. In the Type II KID mutant, this flexibility permits adults to adjust methionine metabolism from a MetR-like state when young to a state where methionine feeds the transsulfuration pathway as they age.

Fly stocks and culture

The insulin receptor mutants were generated in Yamamoto et al [4]: dInr^+[HR] (wildtype, accession ‘29B’), dInr^353[HR], dInr^E19[HR], dInr^74[HR]. All alleles were generated by ends-out gene replacement. Wildtype accession ‘29B’ is a gene replacement of the wildtype allele back into its originating stock (wDahomey, wDah, source: L. Partridge UCL, UK). Mutant alleles were produced by nucleotide substitutions of the wDah wildtype allele to alter the targeted amino acid. All replacemtent dInr alleles were maintained as third chromosome (TM6) balanced stocks. Mutant genotypes for all analyses were generated by crossing dInr³⁵³/TM6B Tb,Sb,e to dInr^wt, recovering F1 dInr³⁵³/dInr^wt, and by crossing dInr^E19/Tb,Sb,e to dInr⁷⁴/Tb,Sb,e, recovering trans-heterozygote dInr^E19/dInr⁷⁴. Other stocks used in the epistasis analysis: gnmt-RNAi: w[1118]; P{GD10563}v25983 (VDRC); samtor-RNAi: P{KK107414}VIE-260B (VDRC); Ahcy13-RNAi: y¹ sc* v¹ sev²¹; P{TRiP.HMS05799}attP40 (BDSC #67848); S6K: w[1118]; P{w[+mC]=UAS-S6k.TE}2 (BDSC #6912). tubulinGal4; tubulinGal80ts stock is a gift from the lab of Dr. Norbert Perrimon.. Each of the VDRC and BDSC lines were backcrossed into wDah; + / + ; + /+ for five generations. Flies were reared at 25°C, 40% RH, 12L:12D on standard food media: cornmeal (5.2%), sugar (11.0%), autolyzed yeast (2.5%; SAF brand, Lesaffre Yeast Corp., Milwaukee, WI, USA.), agar (0.79%) (w/v in 100 mL water) with 0.2% Tegosept (methyl4-hydroxybenzoate, Sigma, St Louis, MO, USA), with baker’s yeast supplemented to the food surface.

Infection survival assay

We followed the protocols of Troha and Buchon [69] to orally infect adults with Pectinobacterium carotovora carotovora (strain ECC15). ECC15 was grown in shaking culture at 29C overnight, pelleted by centrifugation (10 min at 2400g, 4C) and resuspended in sterile PBS to OD600. Treatment conditions were initiated with 8-to-9-day-old adult females, collected over a period of 48 hours post-eclosion in the presence of males. The beginning of the treatment period is designated as survival age zero. Cohorts of 100 adults for each genotype were evenly distributed across 10 food vials per treatment group (PBS control, ECC15). A volume of 100 μL of either sterile PBS or Ecc15 suspension was added to a round filter paper placed on top of the food within each vial. Prior to the treatment, flies were starved for 2 hours in empty vials, then placed in ECC15 or control treated vials and retained for 18–24 hours. Subsequently, adults were flipped into fresh, uninfected food daily, at which time dead flies were removed and counted. Kaplan-Meier survival curves are plotted with right-censored data at survival age 47 days. Cox-proportional survival analysis evaluated genotype, treatment and genotype-by-treatment interaction.

RNA-seq data and analysis

Flies eclosed and mated over a 48-hour period and females of each genotype (wildtype, dInr^E19/dInr⁷⁴, dInr³⁵³/dInr^wt) were then collected into six biological replicates of 20 females. Each replicate was lysed in Trizol using a bead Tissuelyzer (Qiagen) at room temperature. RNA was extracted from each sample by NH₄OAc and EtOH precipitation. RNA concentration (ng/μl) and purity (absorbance: 260nm/280nm) was confirmed with a Nanodrop spectrophotometer (ThermoFisher Scientific). Samples were sent to Genewiz for library preparation (Illumina, RNA with PolyA selection) and Illumina HiSeq 2x150 bp sequencing. Returned reads were processed on the Basepair platform for RNA-Seq QC, alignment, and counts. Differential expression, gene set enrichment, K-means enrichment and transcription factor binding site enrichment were generated with iDEP (http://bioinformatics.sdstate.edu/idep96/).

Steady state metabolomics

Tissue for analysis of steady state metabolomic profiles were collected from females at age 15d and 30d. Females eclosed in the presence of males and were flipped into female only vials from age 2 days until sample collection. Six biological replicates of each genotype (dInr^wt/dInr^wt, dInr^E19/dInr⁷⁴, dInr³⁵³/dInr^wt) were generated with 10 females per sample. Females aged to 15 and 30d were collected between 9 and 10 AM, anesthetized with CO₂ and dissected to remove the abdomen to avoid signal from eggs. The somatic tissue was flash frozen in liquid nitrogen and stored at -80. Samples were shipped on dry ice to the Northwestern Metabolomics Research Center (NWMRC) at the University of Washington, Seattle. Protein from each biological sample was extracted at the NWMRC and quantified by Targeted LC-MS across a 360-feature set (https://northwestmetabolomicsorg.wpcomstaging.com/300-aqueous-metabolites/). The analysis returns Quality Control statistics and relative concentration of each detected metabolite based on isotope-labeled internal standards. Data processing and analyses were performed using metaboanalyst (http://www.metaboanalyst.ca/MetaboAnalyst/). Metabolites that were not detected in 40% of the samples were excluded from the analysis. Data were normalized to the median (per sample) and processed through log transformation. Heat map and hierarchical clustering was generated using Pearson correlations and Ward’s method. Metabolite Set Enrichment Analysis (MSEA) was performed using Metaboanalyst which uses the KEGG (http://www.genome.jp/kegg/pathway.html) pathway database. Metabolite sets containing at least five compounds were employed in the analysis. MSEA uses globaltest to test associations between metabolite sets and the outcome [70]. The algorithm uses a generalized linear model to compute a ‘Q-stat’ for each metabolite set. The Q-stat is calculated as the average of the Q values calculated for the each single metabolites; while the Q value is the squared covariance between the metabolite and the outcome [70].

Labeled metabolite profile

Tissue for analysis of methionine labeling was collected from females at 10 days of age. Females eclosed in the presence of males. Flies were transferred to vials with fresh food for two days, after which females alone were separated into vials with chemically defined food (CDF) containing 1 mM unlabeled methionine for two days. At day five, females were transferred to CDF with 1 mM ¹³C5-methionine for 5 days and flipped every 2 days onto fresh CDF containing labeled methionine, or in parallel, maintained on CDF with unlabeled methionine (to correct for the natural abundance of ¹³C isotopes).

On day ten, we generated six biological replicates of each genotype (dInr^wt/dInr^wt, dInr^E19/dInr⁷⁴, dInr³⁵³/dInr^wt), with 15 females per sample. Between 9 and 10 AM, the flies were transferred to empty vials containing a piece of Kimwipe soaked with 800 µL of 5% sucrose solution and kept for four hours (to clean their intestines of the labeled food). After four hours, whole (non-dissected) flies were anesthetized with CO₂, placed into pre-weighed vials, quickly weighed, flash-frozen in liquid nitrogen, and stored at −80 °C.

Intracellular metabolites were extracted using 1.2 mL of cold (−80 °C) 80% (v/v) aqueous methanol. Flies were homogenized using 0.5 mm zirconium beads (via three consecutive rounds of homogenization in a Bullet Blender in the cold room), and the insoluble material in lysates was centrifuged at 10,000 rpm for 5 minutes. The resulting supernatant was evaporated using a SpeedVac.

For metabolomic analysis, samples were mixed with −20 °C 80%:20% methanol:water (extraction solvent), vortexed, and immediately centrifuged at 16,000 x g for 20 min at 4°C. The supernatant was collected for LC-MS analysis. LC was performed on an Xbridge BEH amide HILIC column (Waters) with a Vanquish UHPLC system (Thermo Fisher). Solvent A was 95:5 water: acetonitrile with 20 mM ammonium acetate and 20 mM ammonium hydroxide at pH 9.4. Solvent B was acetonitrile. The gradient used for metabolite separation was 0 min, 90% B; 2 min, 90% B; 3 min, 75%; 7 min, 75% B; 8 min, 70% B, 9 min, 70% B; 10 min, 50% B; 12 min, 50% B; 13 min, 25% B; 14 min, 25% B; 16 min, 0% B, 21 min, 0% B; 21 min, 90% B; and 25 min, 90% B. MS analysis was performed on a Orbitrap Exploris 480 mass spectrometer (Thermo Fisher) by electrospray ionization with the parameters as: scan mode, full MS; spray voltage, 3.6 kV (positive) and −3.2 kV (negative); capillary temperature, 320 °C; sheath gas, 40 arb; aux gas, 7 arb; resolution, 120,000 (full MS); scan m/z range, 70–1,000.

Raw LC–MS data files were converted to mzXML format files by ProteoWizard and analyzed using El-MAVEN Software (Elucidata; elucidata.io). Isotope labeling was corrected for the natural abundance of ¹³C isotopes.

Demographic epistasis analysis

Females for survival analysis were collected as F1 progeny over a 48-hour emergence period in the presence of males, then separated into all female demography cages with 125 adults in each of three cages per genotype. Demography cages were made from 1 L clear food service containers with a ventilated lid, a gasket-covered aperture to provide access to remove dead flies and a port near the cage bottom that opened to a plastic tube affixed a standard glass media vial with 3 ml of standard Drosophila diet. Every two days, dead flies were removed by aspiration and counted, and fresh food vials were provided.

Genotypes for survival analyses were produced by replacing the balancer third chromosome of dInr³⁵³ with the tubulin-Gal4.Gal80 third chromosome, followed by replacing the second chromosome of offspring with the targeted UAS-RNAi or over-expression. Females from these crosses were set up in controlled density bottles to lay eggs. Hatching larvae were reared at 25°C to suppress transgene expression. Emerging adults were shifted to the permissive temperature (29°C) and sorted into demography cages, maintained at 29°C. Wildtype F1 control and dInr³⁵³ single mutant genotypes were generated by replacing the third chromosome balancer with the tubulin-Gal4,Gal80 third chromosome. Cox proportional hazard analysis evaluated the impact of each genotype and their interaction.