Identification of key pathways and metabolic fingerprints of longevity in C. elegans.

Impaired insulin/IGF-1 signaling (IIS) and caloric restriction (CR) prolong lifespan in the nematode C. elegans. However, a cross comparison of these longevity pathways using a multi-omics integration approach is lacking. In this study, we aimed to identify key pathways and metabolite fingerprints of longevity that are shared between IIS and CR worm models using multi-omics integration. We generated transcriptomics and metabolomics data from long-lived worm strains, i.e. daf-2 (impaired IIS) and eat-2 (CR model) and compared them with the wild-type strain N2. Transcriptional profiling identified shared longevity signatures, such as an upregulation of lipid storage and defense responses, and downregulation of macromolecule synthesis and developmental processes. Metabolomics profiling identified an increase in the levels of glycerol‑3P, adenine, xanthine, and AMP, and a decrease in the levels of the amino acid pool, as well as the C18:0, C17:1, C19:1, C20:0 and C22:0 fatty acids. After we integrated transcriptomics and metabolomics data based on the annotations in KEGG, our results highlighted increased amino acid metabolism and an upregulation of purine metabolism as a commonality between the two long-lived mutants. Overall, our findings point towards the existence of shared metabolic pathways that are likely important for lifespan extension and provide novel insights into potential regulators and metabolic fingerprints for longevity.

Introduction

Average life expectancy has dramatically increased over the past few decades, resulting from improvements in healthcare systems in most economically developed countries (Salomon et al., 2012). Aging and life expectancy are affected by genetic predisposition and non-genetic and environmental risk factors, making the process of aging different between individuals (Lopez-Otin et al., 2013). Although the genetic and environmental causes of aging are not fully established, several processes have been identified as hallmarks of aging. Many of these hallmarks involve metabolism, such as deregulated nutrient sensing and mitochondrial dysfunction (Lopez-Otin et al., 2013). A prototypical example of how environmental factors are involved in the metabolic network of longevity is caloric stress. One example of this is caloric restriction (CR), a condition of 20–50% reduced caloric intake, that activates several metabolic pathways and thereby extends lifespan across species (Houtkooper et al., 2010). Besides environmental factors, genetics also play an important role in the metabolic control of aging. So far, several metabolic genes have been identified that influence longevity (Houtkooper et al., 2010). For instance, mammalian longevity is influenced by metabolic changes as a result of genetically determined insulin sensitivity and glucose handling (Riera and Dillin, 2015). Taken together, it is evident that both the environment and genetics play an important role in controlling metabolism and aging. However, it remains to be elucidated how both components interact to regulate metabolism and aging.

C. elegans is one of the most popular model organisms for aging studies due to its relatively short lifespan, fully annotated genome and its ease of manipulation (Gao et al., 2018). One of the first aging genes to be discovered was daf-2, which encodes a homolog of mammalian insulin/insulin-like growth factor (IGF), and its regulatory mechanisms involved in longevity are highly conserved across a wide range of species, including yeast, flies, mice and human beings (Kenyon, 2011). Lowered DAF-2 signaling leads to translocation of DAF-16, a fork-head transcription factor, that enters the nucleus and activates the expression of numerous genes involved in longevity, lipid metabolism, stress response and innate immunity (Kenyon, 2010). As a result, daf-2 mutant animals can live twice as long as control animals (Kenyon et al., 1993). In addition, C. elegans also serves as a tractable model to study molecular mechanisms underlying CR. Distinct fashions of CR exist, including genetic (i.e. eat-2 mutants) and dietary interventions (e.g. diluting bacterial food) (Lakowski and Hekimi, 1998; Greer and Brunet, 2009). eat-2 mutant animals have a dysfunctional pharynx, which slows down food intake, and results in a lifespan increase of approximately 50% (Gao et al., 2017a). Previous studies on daf-2 and eat-2 mutants have outlined the essential biological processes involved in the regulatory networks of aging, and have opened up novel avenues for treatments and biomarkers for aging-related diseases (Kenyon, 2010). In fact, longevity mediated mechanisms of either CR or impaired insulin/IGF-1-like signaling (IIS) seem to be partially overlap in terms of the downstream regulatory processes, including activation of autophagy, and inhibition of the target of rapamycin pathway (TOR) (Kenyon, 2010; Panowski et al., 2007; Walker et al., 2005; Yen and Mobbs, 2010). Although the IIS pathway, especially the activation of DAF-16, is associated with CR-induced longevity, it is dispensable for the longevity phenotype observed in eat-2 mutants (Kenyon, 2010; Lakowski and Hekimi, 1998).

In recent years, studies using high-throughput “omics” approaches have gained commendation as they have provided comprehensive information on the genomic, metabolomic, and proteomic changes that occur during the aging process (Wan et al., 2017). Among all the omics analyses, metabolomics or profiling of metabolites represents possibly the most diverse and complex level of biological regulatory processes, and is intimately linked to phenotype. For instance, metabolite profiles facilitate the generation of phenotypic data of complex metabolic alterations that incorporate messages from multiple levels of systemic regulation, including the genome, the transcriptome, the proteome, the environment and their interactions (Chan et al., 2010). Technological advancements in metabolomics have enabled numerous discoveries of perturbations in physiological networks and have further expanded our knowledge of longevity mechanisms underlying specific biological functions in different organisms, including worms, mice and humans (Houtkooper et al., 2011; Cheng et al., 2015; Gao et al., 2017b).

Although distinct age-related metabolic signatures have been reported in C. elegans, the relationship between gene transcription and metabolite levels remains to be fully elucidated. In this study, we aim to identify key metabolic signatures of longevity. We directly compared two long-lived mutant strains, namely daf-2(e1370) and eat-2(ad465) (from here on called daf-2 and eat-2 mutants) by whole-genome microarray and mass spectrometry-based platforms (Gao et al., 2017b). Next, we superimposed these two platforms in a cross-omics approach to visualize the actual changes at two regulatory levels, and identified shared modes of metabolic regulation of longevity.

Materials and methods

C. elegans strains and bacterial feeding strains

The C. elegans N2 (wild-type strain, Bristol), daf-2(e1370) and eat-2(ad465), and E. coli OP50 and HT115 strains were obtained from the Caenorhabditis Genetics Center (CGC). Both mutant strains were backcrossed three times and sequenced before use.

Worm growth conditions and worm pellet collection

C. elegans were routinely grown and maintained on nematode growth media (NGM) at 20 °C. Worms of each strain were cultured on plates seeded with E. coli OP50, then eggs were obtained by alkaline hypochlorite treatment of gravid adults and placed onto plates seeded with E. coli HT115. At the young adult stage, each strain was collected by washing them off the culture plates with M9 buffer and the worm pellet was washed with dH2O for three times before collection in a 2 mL Eppendorf tube, snap-frozen in liquid N2 and lyophilized overnight. For RNA extraction, 500 worms/tube was collected in four replicates per worm strain; for polar metabolite extraction, 2000 worms/tube was collected in five replicates for both long-lived mutants and seven for N2 strain; for fatty acid and amino acid extraction, 2000 worms/tube was collected in four replicates.

Microarray

Total RNA was extracted using the standard Trizol method, followed by phenol‑chloroform purification and further processing by OakLabs Gmbh (Germany). In short, RNA quality and quantity was assessed after DNase clean-up using a 2100 Bioanalyzer (Agilent Technologies). All samples had a RNA integrity value of ≥7.2. RNA was amplified and labeled using a Low Input QuickAmp Labeling Kit (Agilent Technologies) and hybridized using the Agilent Gene Expression Hybridization Kit (Agilent Technologies). An ArrayXS-068300 with WormBase WS241 genome build (OakLabs) was used and fluorescence signals were detected by the SureScan microarray Scanner (Agilent Technologies). Data of all samples were quantile normalized using the ranked median quantiles as described previously (Bolstad et al., 2003). The most significant coding-gene isoform was selected to represent expression for mentioned genes. Fold-change of gene expressions in the mutant strains was characterized as log2 (expression/expressionN2), and the significance (p-value) was calculated using a Student's t-test. The expression of a certain gene in each long-lived mutant with a p-value < 0.05 and a fold-change either less than −1 or higher than 1 was considered as a significant compared to N2 and selected for further analysis. The overlapping up- or down-regulated genes in the long-lived mutants were selected using Venny diagram 2.1.0 (Oliveros, n.d.). KEGG pathway analysis was performed with David (Huang et al., 2009). GO term enrichment analysis was performed with David and ReviGo (Supek et al., 2011). The result from ReviGO was downloaded as an R script and images were imported into Illustrator (Adobe) for further editing.

Metabolomics analysis: polar metabolite measurements

500 μL MQ water, 500 μL MeOH and 1 mL chloroform was added to the dry worm pellet. The following internal standards were added to the homogenate: a mixture of amino acids ((Cambridge Isotope Laboratories, UK) containing 68 nmol dl‑alanine‑2,3,3,3‑D4, 44 nmol dl‑glutamic acid‑2,4,4‑D3, 40 nmol l‑leucine‑5,5,5‑D3, 28 nmol l‑phenylalanine‑ring‑D5, 34 nmol l‑valine‑D8, 34 nmol l‑methionine‑(methyl‑D3), 26 nmol l‑tyrosine‑ring‑D4, 22 nmol l‑tryptophan‑(indole‑D5), 46 nmol dl‑serine‑2,3,3‑D3, 48 nmol proline‑D7, 24 nmol l‑arginine‑2,3,3,4,4,5,5‑D7, 28 nmol l‑glutamine‑2,3,3,4,4‑D5, 32 nmol l‑lysine‑4,4,5,5‑D4, 26 nmol l‑citrulline‑ureido‑13C, 28 nmol l‑ornithine‑3,3,4,4,5,5‑D6, 42 nmol l‑isoleucine‑D10, and 46 nmol dl‑aspartic acid‑2,3,3‑D3 (Gao et al., 2017b)), 13C3‑pyruvate, 13C6‑glucose, 13C6‑fructose‑1,6‑biphosphate, 13C6‑glucose‑6‑phosphate, adenosine‑15N5‑monophosphate and guanosine‑15N5‑monophosphate (5 μM). Thereafter, the worm suspension was homogenized using a TissueLyser II (Qiagen) for 5 min at a frequency of 30 times/s. After centrifugation at 4 °C, ~800 μL of the “polar” top layer was transferred to a new 1.5 mL Eppendorf tube and dried in a vacuum concentrator at 60 °C (GeneVac). Dried residues were dissolved in 100 μL MeOH:MQ water (6:4, v/v). For polar metabolite measurements, we used a Thermo Scientific ultra-high pressure liquid chromatography system (Waltman, MA, USA) coupled to a Thermo Q Exactive (Plus) Orbitrap mass spectrometer (Waltman, MA, USA). The auto-sampler was kept at 10 °C during the analysis and 5 μL sample was injected on the analytical column. The chromatographic separation was established using a Sequant ZIC-cHILIC column (PEEK 100 × 2.1 mm, 3.0 μm particle size, Merck, Darmstadt, Germany) and held at 15 °C. The flow rate was kept at 0.250 mL/min. The mobile phase was composed of (i) acetonitrile:water (9:1, v/v) with 5 mM ammonium acetate; pH = 6.8 and (ii) acetonitrile:water (1:9, v/v) with 5 mM ammonium acetate; pH = 6.8, respectively. The LC gradient program was: start with 100% (i) hold 0–3 min; ramping 3–20 min to 36% (i); ramping from 20 to 24 min to 20% (i); hold from 24 to 27 min at 20% (i); ramping from 27 to 28 min to 100% (i); and re-equilibration from 28 to 35 min with 100% (i). The MS data were obtained in negative ionization mode at full scan range at a mass resolution of 140,000. Interpretation of the data was performed in the Xcalibur software (Thermo scientific, Waltman, MA, USA). α-, β-, d-, & l-metabolites were not distinguished. Statistical analysis was performed using the statistical programming software R (http://www.r-project.org) combined with packages MixOmics (Rohart et al., 2017), ropls (Thevenot et al., 2015) and ggplot2. Metabolites in either daf-2 or eat-2 mutants were determined statistically different from N2 worms using a Student's t-test and used Benjamini and Hochberg's method to control for false positives with the False Discovery Rate (FDR) (Benjamini and Yekutieli, 2001). Q value refers to the FDR level of significance, which should be <0.05.

Metabolomics analysis: fatty acid and amino acid extraction and analysis

Fatty acids were extracted and analyzed as described (Gao et al., 2017b). In brief, a dry worm pellet was re-suspended in 250 μL ice-cold 0.9% NaCl solution. Worm lysate was obtained by adding a 5 mm steel bead to the worm suspension and mixed using a TissueLyser II (Qiagen) for 2 × 2.5 min at frequency of 30 times/s, followed by tip sonication (energy level: 40 J; output: 8 W) for two times on ice-cold water. Protein quantification was performed with a BCA assay. Worm lysate (containing 150 μg protein) was transferred in a 4 mL fatty acid-free glass vial, and 1 mL of freshly prepared 100% acetonitrile (ACN)/37% hydrochloric acid (HCl) (4:1, v/v) was added to the lysate, together with deuterium-labeled internal standards. Samples were hydrolyzed by incubating at 90 °C for 2 h. After the vials cooled down to room temperature, fatty acids were separated in a hexane layer and then evaporated at 30 °C under a stream of nitrogen. Fatty acid residues were dissolved in 150 μL chloroform-methanol-water (50:45:5, v/v/v) solution containing 0.0025% aqueous ammonia, and then transferred to a Gilson vial for ESI-MS analysis.

Deuterium-labeled internal standards were added to each sample, including 17,17,18,18,18‑D5‑C18:0 (5.04 nmol, CDN isotope, Canada), 3,3,5,5‑D4‑C24:0 (2.52 nmol, Organic Synthesis Laboratory of VU Medical Center, The Netherlands) and 3,3,5,5‑D4‑C26:0 (0.25 nmol, Organic Synthesis Laboratory of VU Medical Center, The Netherlands); we also ran an additional five samples with increasing amounts of a fatty acid mixture (C18:0, C24:0 and C26:0) to make a five-point calibration curve. In these additional five samples, above-mentioned internal standards were also added (Gao et al., 2017b).

Amino acids (AAs) were extracted by transferring worm lysate (contains 50 μg of protein) to a 2 mL Eppendorf tube, and 1 mL of 80% ACN plus 20 μL of internal standard mixture was added to the lysate and homogenized by vortexing (Gao et al., 2017b). Samples were centrifuged and the supernatant was transferred to a 4 mL glass vial and evaporated under a stream of nitrogen at 40 °C. After evaporation, AA residue was dissolved in 220 μL of 0.01% heptafluorobutyric acid (v/v in MQ water). Then the suspension was transferred to a Gilson vial for UPLC-MS/MS analysis.

Cross-omics data analysis

Microarray transcriptomics data and polar metabolite data were integrated using package Pathview (Luo and Brouwer, 2013) in R. Metabolic enzymes nodes compromised of multiple enzymes were colored using the sum of all related gene expressions. The first gene listed in KEGG for a certain enzyme was selected for use in the figure.

Gene expression profile visualization

Heat maps of gene expression profile were plotted using “R2” (Genomics analysis and Visualization platform) (http://r2.amc.nl).

Results

Gene expression analysis reveals the major signaling pathways important for longevity

To comprehensively quantify changes in global gene expression, we performed microarray analysis for the long-lived mutants daf-2 and eat-2 and used N2 as the control group. Differentially expressed genes (DEGs) in the long-lived mutants compared to N2 were analyzed using gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment. In daf-2 mutants, 3544 genes were significantly upregulated and 2336 genes were downregulated compared to those in N2 worms (Fig. 1a). In eat-2 mutants, we found fewer genes to be significantly changed, with 2352 and 983 genes being up- and down-regulated, respectively (Fig. 1b).

Fig. 1

Transcriptional alterations in long-lived daf-2 and eat-2 mutants compared to N2 worms. (a) Volcano plot of log2-transformed fold change of gene expression in daf-2(e1370) over N2 worms. (b) Volcano plot of log2-transformed fold change of gene expression in eat-2(ad465) compared to those in N2 worms. p-Value was calculated using Student's t-test. The vertical dashed lines represent the log2 fold change = −1 or 1, horizontal dashed lines denote p-value cutoffs. The number of genes that are significantly up- (red) or down- (blue) regulated (p-value < 0.05) in each panel is also mentioned. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

To find shared pathways that are important for regulation of longevity, we focused on genes that have a similar change in expression in both daf-2 and eat-2 using the Database for Annotation, Visualization and Integrated Discovery (DAVID) (Huang et al., 2009) and ReviGo (Supek et al., 2011) (Fig. 2). GO term enrichment analysis for biological processes was performed on 1722 genes that were upregulated and 617 genes that were downregulated in both long-lived mutants. Enriched GO terms for upregulated genes included phosphorus metabolic process, lipid localization, regulation of anatomical structure, morphogenesis, cell surface receptor signaling pathways, response to stress, regulation of cell proliferation, male meiosis, spermatogenesis and flagellated sperm motility (Fig. 2a). For the 617 genes that were downregulated in both long-lived mutants, we observed an enrichment of GO terms for aromatic compound biosynthesis, heterocycle biosynthesis, RNA metabolism, regulation of RNA stability, response to abiotic stimulus, positive regulation of nucleobase-containing compound metabolism, regulation of nitrogen compound metabolism, mesoderm development and regulation of cell differentiation (Fig. 2b).

Fig. 2

Venn diagram and GO term enrichment analysis comparing daf-2 and eat-2 mutants with N2. (a) Venn diagram of significantly upregulated genes in both long-lived mutants identified 1722 shared transcripts. GO term enrichment (biological process) of these genes using David and ReviGO highlighted several metabolic pathways such as lipid metabolism and phosphorus metabolic process. Venn diagram percentages in brackets indicate the number of genes that account for the % of the total up-/downregulated genes. (b) Venn diagram of significantly downregulated genes in both long-lived mutants identified 617 transcripts. GO term enrichment (biological process) of these genes using David and ReviGO shows an attenuation of process that are involved in biosynthesis of macromolecules such as aromatic compounds, nucleobase-containing compounds and nitrogen compounds. The size of the plot is indicative of the frequency of the GO term in the underlying Gene Ontology Annotation database; the plots are color-coded according to significance (log10-transformed); level of significance increases from red to blue. GO terms belonging to the same cluster were grouped and circled in dark grey dashed line. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Similar results were obtained when analyzing the upregulated genes for KEGG pathway analysis using DAVID (Table 1). We observed a significant enrichment in intermediary metabolism processes including fatty acid degradation, metabolic pathways, biosynthesis of antibiotics, metabolism of xenobiotics by cytochrome P450, drug metabolism, amino acid synthesis, lipid metabolism, tyrosine metabolism, glycine, serine and threonine metabolism, aromatic compound degradation, branched-chain amino acid synthesis, and cysteine and methionine metabolism. Downregulated genes mapped to the TGF-beta signaling pathway, Wnt signaling pathway, protein processing in the endoplasmic reticulum and ubiquitin mediated proteolysis pathways (Table 1).

Table 1

KEGG pathway enrichment analysis of the 1722 up-regulated genes and 617 down-regulated genes of both long-lived mutants compared to N2. Pathways that are significantly up-/downregulated (p < 0.05 after Benjamini correction) are marked in red.

Taken together, we observed an enrichment of cellular metabolic pathways and energy metabolism pathways in both long-lived mutants. Furthermore, pathways that regulate the biosynthesis of macromolecules, such as aromatic and nitrogen compounds, were attenuated in both long-lived mutants, suggesting an overall low energy-consuming state.

Metabolomics analysis uncovers distinct metabolic changes in both long-lived mutants

In order to provide novel insight into the relationship between longevity and metabolism, we characterized the metabolic fingerprints of the long-lived mutants by performing semi-targeted metabolomics with a focus on polar metabolites (Fig. 3). Polar metabolite profiles of the two long-lived mutants and N2 worms were analyzed using unsupervised PCA analysis which demonstrated marked differences in metabolite profiles between the strains (Fig. 3a). The long-lived mutants differed in a number of metabolic features compared to N2 worms (Fig. 3b). In total, we observed thirteen metabolites that were significantly altered in either one or both long-lived mutants. We observed a lowered level of three metabolites in both of the long-lived mutants, including the glycerolipid intermediate glycerol‑3P, the nucleobase adenine, and the purine base xanthine. Additionally, the nucleotide adenosine monophosphate (AMP) was increased in both daf-2 and eat-2. Two metabolites—the TCA intermediate succinate and adenosine diphosphate (ADP)—were significantly changed in both long-lived mutants, although both metabolites were lowered in daf-2 mutants and increased in eat-2 mutants. Furthermore, we observed distinct metabolites that changed in either of the two long-lived mutants. For instance, we detected four metabolites that are uniquely changed for daf-2 mutant, including reduced TCA cycle intermediate citrate/isocitrate and malate, reduced pentose phosphate pathway (PPP) intermediate gluconate, and elevated glycolytic intermediate phosphoenolpyruvate (Fig. 3b and Supplementary Table S2). Moreover, three metabolites showed an increased levels in eat-2 mutants, including two pyrimidine metabolism intermediates—uridine monophosphate (UMP) and uracil—and one purine metabolism intermediate—guanosine monophosphate (GMP) (Fig. 3b and Supplementary Table S2).

Fig. 3

Metabolite signatures of the long-lived mutants daf-2 and eat-2. (a) Principle Component Analysis (PCA) analysis score plot showing group separation based on polar metabolite profiles in both long-lived mutants compared to N2 worms, ellipses represent the 95% confidence interval. (b) Polar metabolite profiles of daf-2 and eat-2 mutants compared to N2 worms (metabolites that are significantly different (FDR < 0.05) are selected and the p-values from the t-test is indicated as *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). Several shared metabolic signatures were detected, including increased levels of AMP, decreased levels of glycerol‑3P, adenine, and xanthine. Unique metabolite signatures for each long-lived mutant were also observed, for the daf-2 mutant: increased levels of phosphoenolpyruvate, and decreased levels of citrate/isocitrate, gluconate, ADP, succinate and malate; for the eat-2 mutant: increased levels of succinate, UMP, uracil, GMP, and ADP. (c) Most amino acid species in both long-lived mutants were lowered compared to N2 worms. (d) Total amino acids in both daf-2 and eat-2 mutants were significantly lower compared to that in N2 worms. (e) A number of abundant fatty acids were increased in daf-2 mutants compared to N2 worms, whereas most fatty acids in eat-2 mutants were decreased compared to N2. (f) Total fatty acids in daf-2 mutants were significantly elevated while those in eat-2 mutants were significantly decreased. (g) Mono-unsaturated fatty acids (MUFA) were more abundant in daf-2 mutants and PUFAs were more abundant in eat-2 mutants. The bar graphs depict mean ± SD. Significance in panel c–f was calculated using one-way ANOVA; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

To expand on the characterization of metabolites in long-lived mutants, we measured amino acids and fatty acids using our targeted quantitative mass spectrometry (MS) platform (Gao et al., 2017b). We observed a decrease in most individual amino acids in the two long-lived mutants compared to N2 worms (Fig. 3c). The total abundance of amino acids was significantly decreased in both long-lived mutants (Fig. 3d). These results suggest the reorganization of amino acid metabolism in the two long-lived mutants.

From the transcriptional dataset analysis, we identified several pathways involved in lipid metabolism, including the KEGG pathway terms fatty acid degradation and fatty acid metabolism, and the GO terms lipid localization and lipid storage (Fig. 2, Table 1). We therefore examined the specific lipid signatures in the two long-lived mutants (Fig. 3e). We observed marked changes in fatty acid levels that were shared between both long-lived mutants, namely lowered levels of saturated fatty acid C18:0, C20:0, C22:0 and odd-chain monounsaturated fatty acid (MUFA) C17:1 and C19:1. In the daf-2 mutants only, we observed an elevated level for several saturated fatty acids and MUFAs, including C14:0, C15:0, C14:1, C15:1, C16:1, and C18:1 (Fig. 3e and Supplementary Table S2). Some fatty acids were decreased in daf-2 mutants, including saturated fatty acids C19:0, C20:0, C21:0, C22:0, MUFA C22:1, and polyunsaturated fatty acid (PUFA) C24:6. In eat-2 mutants, fatty acid levels were generally lower, such as the saturated fatty acids C14:0, C15:0, C16:0, and the MUFA C18:1. In addition, the total amount of fatty acids in daf-2 mutants was significantly increased, while this was opposite in eat-2 mutants (Fig. 3f). Interestingly, daf-2 and eat-2 mutants showed a distinct lipid saturation profile; the former has a higher proportion of MUFAs compared to the other two strains, and the latter has more PUFAs (Fig. 3g). Taken together, these results suggest a distinct utilization of lipids in both long-lived mutants and a tight regulation of energy metabolism in these animals.

Cross-omics analysis highlights important nodes in central carbon metabolism

Following the changes in polar metabolites, we then focused on elucidating key nodes of central carbohydrate metabolism and nucleotide metabolic pathways. We integrated the polar metabolite profiles together with the gene expression profiles for both long-lived mutants and plotted both omics datasets in the known pathways and interactions annotated in KEGG (Fig. 4) (Luo and Brouwer, 2013). In the glycolysis/gluconeogenesis pathway, we observed an upregulated expression for pfk-1.2 in both long-lived mutants (Fig. 4). This gene encodes phosphofructokinase, which is one of the major regulatory enzymes of glycolysis that irreversibly catalyzes β‑d‑fructose‑6P to β‑d‑fructose‑1,6P2. We detected downregulation of gpd-4 (glyceraldehyde‑3‑phosphate dehydrogenase) in both long-lived mutants, however, this had only a minor influence on the concentrations of glycolytic intermediates in the following steps. In addition, except pyk-1 (pyruvate kinase) that regulates the final step of glycolysis, expression levels of genes encoding the transitional glycolytic enzymes, including pgk-1 (phosphoglycerate kinase), ipgm-1 (phosphoglycerate mutase), and enol-1 (enolase) remained unaltered in both long-lived mutants. Interestingly, we detected an elevated level of phosphoenolpyruvate in daf-2 mutants whereas a slightly decreased level of phosphoenolpyruvate is observed in eat-2 mutants compared to N2 worms. This may be explained by the discovery that the expression level of pck-2 (phosphoenolpyruvate carboxykinase) was upregulated in daf-2 mutants while down-regulated in eat-2 mutants. Furthermore, we observed an increased expression of pyk-1 (pyruvate kinase), which may explain the elevated levels of pyruvate in daf-2 mutants while it remained unchanged in eat-2 mutants. Another gene that encodes a key glycolytic enzyme is hexokinase, hxk-1, which was downregulated in daf-2 mutants, possibly explaining our observations that β‑d‑fructose‑1,6P2 was lowered in daf-2 mutants whereas in eat-2 mutants hxk-1 expression was elevated compared to N2 worms.

Fig. 4

Central carbon metabolism pathways in daf-2 and eat-2 mutants. (a) Data from the gene expression profile (rectangles) and metabolite profile (circles) were color-coded and integrated in three major pathways of central carbon metabolism, including glycolysis/gluconeogenesis, the pentose phosphate pathway and TCA cycle. The majority of genes involved in glycolysis/gluconeogenesis remained unaltered in both long-lived mutants.

Glycolysis/gluconeogenesis intermediates remained low in both long-lived mutants, except the final product of this pathway, i.e. pyruvate, which accumulated in daf-2 mutants. Genes coding for PPP enzymes were mostly unchanged and PPP intermediates were lowered in both long-lived mutants. Alcohol fermenting genes also showed increased expression in both long-lived mutants. The majority of genes involved in the TCA cycle were unchanged and TCA intermediates were significantly lowered in daf-2 mutants. TCA intermediates were unchanged in eat-2 mutants except for elevated cis‑aconitate and succinate levels. White circles are metabolites that were not measured. (b) Gene expression profiles of genes that are involved in fermentative metabolism. Each column represents one biological replicate (four biological replicates for each worm strain), and the fold-change is represented as log2-transformed.

The PPP is also involved in cytosolic central carbon metabolism. Transcript levels of most PPP genes in both long-lived mutants showed no significant differences compared to those in N2 worms, including gdpd-1 (glucose‑6‑phosphate dehydrogenase), Y57G11C.3 (6‑phosphogluconolactonase), T25B9.9 (phosphogluconate dehydrogenase), tkt-1 (transketolase), rpia-1 (ribose 5‑phosphate isomerase A), tald-1 (transaldolase 1), R151.2 (phosphoribosylpyrophosphate synthetase 1), and Y43F4B.5 (phosphoglucomutase 2) (Fig. 4). We did, however, detect a significant upregulation in the expression of F07A11.5, an orthologue of human ribokinase (RBKS), in both long-lived mutants. Moreover, we observed an upregulation of F08F8.7, a gene encoding an enzyme that is predicted to have ribulose‑phosphate 3‑epimerase activity, in daf-2 mutants compared to N2. Most PPP metabolite concentrations in both long-lived mutants remained either unchanged or relatively lower compared to N2. Taken together, the selective increased transcription levels of PPP genes and low levels of PPP intermediates suggests a lowered flux in this pathway in both long-lived mutants compared to N2 worms.

In both long-lived mutants, we observed a strong upregulation of genes encoding enzymes involved in alcohol fermentation, including alh-5 (aldehyde dehydrogenase 3 family member B2), alh-12 (aldehyde dehydrogenase 9 family member A1), sodh-1 (sorbitol dehydrogenase family), and sodh-2 (sorbitol dehydrogenase family) (Fig. 4a and b, Supplemental Table S1). The substrates of these dehydrogenases remain unclear, although SODH-1 has been reported to catalyze the breakdown of ethanol (Alaimo et al., 2012). In addition, SODH-1 was predicted to have oxidoreductase function and its expression can be regulated by several essential metabolic genes, including daf-16 (FOXO), lbp-5 (lipid binding protein), mdt-15 (mediator subunit), and pept-1 (peptide transporter) (Zhang et al., 2013; Xu et al., 2014; Geillinger et al., 2014). Hence, the known biological activity of SODH-1 may include more than its function in fermentative metabolism alone. Furthermore, upregulation of alh-5 and alh-12 may lead to higher intracellular levels of acetate which, together with the observed unchanged expression of pdhb-1 (pyruvate dehydrogenase beta), may partly explain the low abundance of citrate that we observed in daf-2 mutants and unaltered levels in eat-2 mutants, respectively.

As mentioned above, transcript profiles of both long-lived mutants did not show pronounced changes in mRNA levels for TCA cycle enzymes, and this pathway was therefore not identified in the KEGG pathway and GO term enrichment analyses (Table 1, Fig. 4). However, a few genes that encode key enzymes in the TCA cycle were differentially regulated in daf-2 and eat-2 mutants compared to N2. Specifically, aco-2 (aconitase) and idhg-2 (isocitrate dehydrogenase gamma), were upregulated, and mdh-2 (malate dehydrogenase) and acly-2 (ATP citrate lyase) were downregulated in daf-2 mutants, and idh-2 (isocitrate dehydrogenase) was upregulated in eat-2 mutants (Fig. 4). Furthermore, we also observed low levels of TCA cycle intermediates in both long-lived mutants, except for cis‑aconitate and succinate that were relatively high in the eat-2 mutant compared to N2. These findings suggest that energy metabolism in both long-lived mutants is differentially regulated compared to that in N2 worms. Moreover, the TCA cycle may be particularly impaired in daf-2 mutants, as shown by the high level of pyruvate and reduced levels of TCA intermediate metabolites.

Pyrimidine metabolism is altered in both long-lived mutants

In characterizing the metabolic fingerprints of both long-lived mutants compared to N2 worms, we observed marked changes in nucleotides that are involved in the pyrimidine metabolic pathway. We observed a general decrease of pyrimidine metabolic gene transcripts in both long-lived mutants, especially in daf-2 mutants (Fig. 5). For instance, rpb-6, a gene encoding a human orthologue of the RNA polymerase II subunit F, was significantly downregulated in both long-lived mutants and likely resulted in the low levels we observed for UTP and CTP (Fig. 5). We also detected low expression of ndk-1 (nucleoside diphosphate kinase), which may explain the reduced levels of dTDP in the eat-2 mutants. Another downregulated pyrimidine metabolic gene was rnr-1, which encodes the large subunit of ribonucleotide reductase and catalyzes the irreversible conversion of CDP and UDP to dCDP and dUDP, respectively.

Fig. 5

Pyrimidine metabolism pathways in daf-2 and eat-2 mutants. Genes involved in the pyrimidine metabolism in both long-lived mutants were significantly down-regulated compared to those in N2 worms. Only a few genes involved in pyrimidine degradation were upregulated. In line with the expression, most pyrimidine intermediates were low abundant in both long-lived mutants. White circles are metabolites that were not measured.

Overall, we observed decreased concentrations of pyrimidine intermediates in daf-2 mutants that corroborates a downregulation of the majority of genes involved in pyrimidine metabolism. In contrast, although many pyrimidine metabolic genes have lowered expression in the eat-2 mutants, we observed elevated concentrations of the pyrimidine intermediate metabolites, including UDP, UMP, CMP, uracil, and thymidine (Fig. 5). Elevated levels of uracil may be associated with increased expression levels of dpyd-1 (dihydro‑pyrimidine dehydrogenase) and K02D7.1 (purine nucleoside phosphorylase). Overall, our cross-omics results suggest a down-regulation of pyrimidine metabolic pathways in both long-lived mutants.

Purine de novo synthesis pathways were enhanced in both long-lived mutants

We next explored the purine metabolic pathways, including de novo synthesis, the salvage pathways and purine degradation (Fig. 6). Most genes involved in de novo purine biosynthesis pathways, including ppat-1 (phosphoribosyl pyrophosphate amidotransferase), F38B6.4 (phosphoribosylamine‑glycine ligase and phosphoribosyl‑formylglycinamidine cyclo‑ligase), pacs-1 (phosphoribosylamino‑imidazole carboxylase and phosphor‑ribosylaminoimidazole‑succinocarboxamide synthase), and atic-1 (5‑aminoimidazole‑4‑carboxamide ribonucleotide formyltransferase/IMP cyclohydrolase), were upregulated in both long-lived mutants. We observed a low level of ribose‑5P in eat-2 mutants compared to N2 and daf-2 animals. Interestingly, the abundance of 5‑formamidoimidazole‑4‑carboxamide‑ribotide (FAICAR) and inosine monophosphate (IMP), the intermediates at the final two steps of purine ring synthesis, were decreased in daf-2 mutants but elevated in eat-2 mutants. Genes encoding enzymes involved in the salvage pathway are shared between purine and pyrimidine metabolism and we therefore observed a similar gene expression profile: most genes were the either downregulated or remain unchanged in both long-lived mutants. Similar to IMP, the levels of other purine nucleotides, including GDP, GMP, AMP, and ADP were elevated in eat-2 mutants, while decreased in daf-2 mutant. GTP and ATP remained at low levels in both long-lived mutants. Notably, genes involved in purine degradation were upregulated in both long-lived mutants. Most nucleotide metabolic intermediates displayed complex patterns: some showed comparable concentrations to their upstream substrates, such as xanthine and hypoxanthine; guanine showed an opposite level compared to guanosine and GMP, which was elevated in daf-2 mutants and lowered in eat-2 mutants (Fig. 6). Overall, this result suggests an up-regulation of purine metabolism, including de novo synthesis, salvage pathways and degradation.

Fig. 6

Purine metabolism pathways of daf-2 and eat-2 mutants. The majority of genes involved in the three major pathways (de novo synthesis, salvage, and degradation) of purine metabolism were significantly upregulated in both daf-2 and eat-2. Three metabolites showed similar changes in both mutants, i.e. AMP, adenine, and xanthine. Most purine metabolism intermediates were decreased in daf-2 mutants, except guanine; in eat-2 mutants the differences in purine intermediate abundance is more diverse. White circles are metabolites that were not measured.

Other altered metabolic pathways related to amino acids changes in the long-lived mutants

We also analyzed cross-omics data for the altered amino acid profiles in both long-lived mutants. Because we observed lower levels of methionine and glycine, which are involved in one-carbon metabolism, we checked the changes at the transcriptional level. Indeed, genes involved in one carbon metabolism, such as atic-1, dao-3, F38B6.4, gcst-1, and K07E3.4, were significantly upregulated in daf-2 mutants and alh-3, atic-1, F38B6.4, gcst-1, and K07E3.4 were upregulated in eat-2 mutants (Fig. 7a and Supplementary Table S1). This suggests that one-carbon metabolism may be enhanced in both long-lived mutants.

Fig. 7

Heat maps showing transcriptional changes in specific gene sets. Gene expression profiles of gene sets that are involved in one carbon metabolism (a), and branched-chain amino acid (BCAA) metabolism (b). Each column represents one biological replicate (four biological replicates for each worm strain), and the fold-change is represented as log2-transformed.

To gain insight into BCAA metabolism modulation, we evaluated the transcriptomic datasets and found that the majority of genes involved in the BCAA degradation pathway were altered in their expression levels in both long-lived mutants. In daf-2 mutants, Y44A6D.5, acdh-8, alh-12, F54C8.1, gta-1, hacd-1, and mce-1 were upregulated while in eat-2 mutants, approximately 50% of the DEGs were upregulated (e.g. bcat-1) and the rest were downregulated (Fig. 7b and Supplementary Table S1). The distinct alterations at the transcriptional level may explain why the level of BCAAs in eat-2 mutants were lowered, and especially drastically reduced in daf-2 mutants. From the functional annotation clustering for KEGG pathway terms, we observed a significant enrichment for tyrosine metabolism, glycine metabolism and degradation of aromatic compounds. Together with the previous observed lowered level of these amino acids in both daf-2 and eat-2 mutants, these data suggest a shift from carbon to amino acid catabolism as alternative energy source in both long-lived mutants.

Discussion

Impaired IIS and caloric restriction are two well-known longevity interventions that operate in parallel or through partially independent mechanisms (Houtkooper et al., 2010). Previous studies with whole-genome transcript profiling of long-lived mutants have discovered differential expression of gene sets that are involved in lipid metabolism, oxidative stress response and heat shock stress response (Zhang et al., 2013; Tullet, 2015; Heestand et al., 2013). Because most of these genes play important roles in energy metabolism, recent studies have put great effort into characterization of the metabolic consequences in these long-lived mutants (Wan et al., 2017; Butler et al., 2013; Depuydt et al., 2014). However, a correlation between the enzyme regulation at a transcriptional level and their metabolite concentrations remain unclear, especially with regard to energy metabolism (Depuydt et al., 2014). Although a number of essential regulators have been identified at the transcriptomic and metabolomic level of both long-lived mutants, the common mechanisms that underlie metabolic control of longevity in daf-2 and eat-2 mutants have not been elucidated. One possible reason is a lack of comprehensive metabolomics analysis and multi-layer omics platforms. In this study, we hence present an in-depth identification of changes at both the transcript and metabolite level to acquire comprehensive information of potentially common metabolic longevity pathways shared in two long-lived mutants. By analyzing the gene expression profile of daf-2 and eat-2 mutants, we have identified a number of genes in both animals that similarly regulate regions of the metabolic network. The analysis of KEGG pathways and GO terms on genes upregulated in both long-lived mutants showed a significant enrichment of genes involved in regulation of lipid metabolism (lipid storage and lipid localization) and amino acid metabolism (tyrosine, glycine, serine threonine, methionine and BCAAs). Conversely, biosynthesis of macromolecules, such as aromatic compounds and nucleobase-containing compounds, was downregulated in both long-lived mutants. Most of the transcriptomics findings are in a good agreement with previous studies in eat-2 (Heestand et al., 2013) or daf-2 (Murphy et al., 2003), and our work highlights the strong conservation of these pathways in these two longevity models. Furthermore, our work builds upon the studies independently assessing eat-2 and daf-2 transcriptional changes (Heestand et al., 2013; Murphy et al., 2003), and together suggests a phenotype that is conserved in each longevity model, namely, a metabolic shift in amino acid levels. Indeed, we observed lower amino acid levels in both long-lived mutants relative to controls. Although a previous study on the daf-2 mutant showed an increase in the amino acid levels (Fuchs et al., 2010), this may due to the differences in the experimental conditions, such as sample collecting time, culture temperature and bacterial feeding strains. There are three main reasons that may explain the lowered amino acids in both long-lived mutants: (1) overall reduced protein catabolism, (2) increased protein biosynthesis, and (3) increased amino acid-dependent energy expenditure. When analyzing the transcriptomics data, we observed downregulation of protein processing in ER and ubiquitin mediated proteolysis in KEGG pathway enrichment analysis. This suggests that the lowered amino acids in the longevity models may indeed be due to an overall reduced protein catabolism. In addition to these observations, we also observed an upregulation of genes involved in the degradation of aromatic compounds and downregulated gene expression for aromatic compound biosynthesis, suggesting that amino acids, especially tyrosine, tryptophan, and phenylalanine are utilized as primary source for energy production. Indeed, several amino acids feed into the TCA cycle, including alanine, glutamine, tyrosine, and phenylalanine, and were lower in both long-lived mutants. As previous studies showed that protein turnover, particularly protein synthesis, slowed down in daf-2 mutants (Visscher et al., 2016) and protein translation decreased in both daf-2 and eat-2 mutants (Essers et al., 2015), it is unlikely that increased protein biosynthesis is the cause for the lowered amino acids levels detected in both long-lived mutants. Taken together, these findings suggest a shift away from protein catabolism in both long-lived mutants and a strong involvement of amino acid utilization as an alternative energy source.

We observed several novel metabolite features shared in daf-2 and eat-2 mutants, including elevated levels of the purine salvage pathway intermediate (AMP) and reduced levels of the glycerolipid intermediate (glycerol‑3P), and purine degradation intermediates (adenine and xanthine). These findings suggest the existence of shared metabolite profiles and common alterations in biological processes that are required for longevity in both long-lived mutants. It is worth noting however that the metabolomes for both long-lived mutants showed a clear separation in our PCA analysis, indicating the presence of distinct polar metabolite features between the longevity mutants. Indeed, we identified novel metabolite features specifically present in either of these long-lived mutants. Several central carbon metabolism intermediates were lowered in the daf-2 mutants, including succinate, citrate/isocitrate, gluconate and phosphoenolpyruvate, confirming an altered central carbon metabolism in daf-2 animals. Specific metabolic signatures for the eat-2 mutant were elevated levels of succinate and both pyrimidine and purine metabolism intermediates, highlighting an altered regulation in nucleotide metabolism.

In the fatty acid profiles of these two long-lived mutants we observed several fatty acids that showed a similar profile, i.e. lowered concentration of C18:0, C17:1, C19:1, C20:0 and C22:0. In daf-2 mutants, we detected relatively higher MUFAs levels—especially due to high levels of C16:1 and C18:1—while in eat-2 mutants, we observed higher PUFAs levels. Interestingly, elevated level of MUFAs was also observed in rsks-1 and other long-lived mutants that are mediating longevity phenotype (Han et al., 2017). In addition, our findings on the enrichments of fatty acid species in daf-2 mutants are comparable to previous studies using different approaches (Perez and Van Gilst, 2008; Shmookler Reis et al., 2011). Because we identified a significant enrichment of the lipid storage pathway in upregulated genes in both long-lived mutants, we further analyzed the gene clusters involved in the lipid metabolism. We observed an upregulation of genes involved in fatty acid degradation and fatty acid metabolism in both long-lived mutants in the KEGG pathway enrichment analysis, which are in line with the previous studies that showing the activation of fatty acid oxidation in mice undergoing DR (Hallows et al., 2011).

To further uncover potentially shared metabolic pathways that are important for longevity in both daf-2 and eat-2 mutants, we integrated transcriptome and metabolome datasets based on known KEGG pathways involved in central carbon metabolism, pyrimidine and purine metabolism. We analyzed three major pathways involved in central carbon metabolism, including glycolysis/gluconeogenesis, the PPP, and TCA cycle. We detected a distinct regulation at both transcript and metabolite levels involved in glycolysis/gluconeogenesis in daf-2 and eat-2 mutants. Proteomics analysis previously identified a set of proteins involved in amino acid metabolism, carbohydrate metabolisms and ROS detoxification that was significantly altered in daf-2 mutants, although such proteins do not causally contribute to lifespan extension of daf-2 mutants (Dong et al., 2007). In eat-2 mutants, genes encoding enzymes that are involved in glycolysis and gluconeogenesis were downregulated, which is consistent with proteomics analyses (Yuan et al., 2012). Because we have observed lower amino acids levels and increased fatty acid levels in our metabolomics analysis, it is likely that amino acids serve as an alternative energy source for daf-2 mutants (Depuydt et al., 2014; Fuchs et al., 2010; Elle et al., 2012). Similarly, a previous proteomics study on eat-2 mutants showed a subtle shift from carbohydrate to fatty acid metabolism (Yuan et al., 2012), which is also partly reflected in our data. Additionally, our data pinpoint amino acids as an alternative nutrient source in eat-2 mutants. Overall, our data suggest that both long-lived mutants display a large degree of metabolic flexibility (Smith et al., 2018), which allows them to cope better with the available energy source and maintain metabolic homeostasis.

Our results revealed an upregulation of pyrimidine degradation in both long-lived mutants. Although regulation of pyrimidine anabolism at gene expression level was similar in both long-lived mutants, we especially detected high levels of several pyrimidine intermediates in eat-2 mutants. In contrast, purine metabolism, including de novo synthesis, salvage and degradation pathways, was upregulated in both long-lived mutants at the transcriptional level. At the metabolite level, although daf-2 and eat-2 mutants showed distinct features, two intermediates of purine degradation were lowered in both long-lived mutants (adenine and xanthine), denoting a novel metabolic signature shared between the two long-lived mutants.

In conclusion, the integration of transcriptomics and metabolomics analyses provides an extensive dataset of changes resulting from impaired IIS and CR in C. elegans. Importantly, although daf-2 and eat-2 mutants have extended lifespan through disparate mechanisms, impaired IIS and CR share transcriptome and metabolome signatures that likely facilitate longevity. These novel findings provide important insights into the relationship between metabolism and longevity and help pave the way towards metabolic and pharmacological interventions that may enhance longevity.

The following are the supplementary data related to this article.

Supplemental Table S1

Microarray datasets of genes that are differentially expressed between daf-2 mutant and N2 worms or between eat-2 mutant and N2 worms.

Supplemental Table S2

Metabolite profiles of daf-2 mutant, eat-2 mutant and N2 worms.

Author contributions

AWG, RLS, and RHH conceived and designed the project. AWG, RLS, RK, MvW, and GEJ performed experiments and interpreted data. AWG, RLS and MvW performed bioinformatics. AWG, RLS, GEJ and RHH wrote the manuscript, with contributions from all other authors.

Funding

We thank the Caenorhabditis Genetic Center, which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440), for supplying C. elegans strains. AWG is supported by an AMC PhD Scholarship. Work in the Houtkooper group is financially supported by an ERC Starting grant (no. 638290), and a VIDI grant from ZonMw (no. 91715305).

Competing financial interests

The authors declare no competing financial interests related to this work.

Data availability

The microarray data is available at Gene Expression Omnibus (accession number: GSE106672).