Glycine increases lifespan in C. elegans depending on the methionine cycle

We previously measured amino acid levels throughout the life of C. elegans, including four larval phases (L1-L4) and ten days of adulthood from young worms to aged ones (days 1–10) [22]. We reported that the concentrations of most amino acids peaked at the later larval stage or early adult phase and then began declining at different adult stages, reaching low levels by the latest stages of the animals' life [22]. One stark exception was glycine, which continued to accumulate in aged worms [22].

To determine if the accumulation of glycine with age is due to increased synthesis or reduced degradation, we measured the expression levels of genes directly involved in glycine metabolism in worms collected at different ages (Fig 1A). Specifically, we observed that the expression levels of most genes in glycine degradation and consumption pathways including glycine decarboxylase (gldc-1), glycine cleavage system H protein (gcsh-1), phosphoribosylamine-glycine ligase (F38B6.4), and D-amino acid oxidase (daao-1) were dramatically lower at day 9 of adulthood (D9) (Fig 1B). In contrast, the expression levels of most genes involved in glycine synthesis including threonine aldolase (R102.4), serine hydroxymethyltransferase (mel-32), alanine-glyoxylate aminotransferase 2 (T09B4.8), and alanine-glyoxylate amino transferase (agxt-1) remained unchanged in aged worms (Fig 1C). These data suggest that the accumulation of glycine observed in aged worms is predominantly due to a reduction in the expression of genes required for its degradation.

To gain a better understanding of the significance of glycine accumulation in aged animals, we next asked whether this phenomenon is prevalent among long-lived worms such as daf-2(e1370) and eat-2(ad465), the C. elegans models of impaired insulin signalling [2] and dietary restriction [24], respectively. To characterize the changes in the levels of glycine during aging in the daf-2(e1370) and eat-2(ad465) mutant lines, we measured glycine levels in long-lived worms at young and old stages, specifically L3 (larval stage 3) and day 10 (D10) of adulthood. Interestingly, the levels of glycine in daf-2(e1370) and eat-2(ad465) at D10 were significantly increased relative to both their levels at L3 (S1 Fig). These results suggest that there might be a generic regulatory mechanism mediating the level of glycine during aging in both wild type and long-lived C. elegans.

To confirm the metabolic branch points of glycine metabolism in the control of glycine levels in C. elegans, we subjected worms to RNAi against the genes in glycine metabolism (Fig 1A) at the time of hatching, and then measured the level of glycine in D1 worms. Worms treated with RNAi against the genes in glycine synthesis pathways showed no effect on the levels of glycine (Fig 1D). This is perhaps because glycine from bacteria may compensate for the reduction of glycine synthesis in worms. Interestingly, a notable exception to this pattern was found for the knockdown of mel-32, encoding a worm homologue of mammalian SHMT1 and SHMT2, which acts to interconvert serine and glycine in one-carbon pathway (Fig 1A) [25]. mel-32 RNAi led to a strong increase in the level of glycine (Fig 1D) and a concomitant slight decrease in the level of serine (S2 Fig). Thus, these results indicate that MEL-32 is prone to synthesize serine from glycine in C. elegans (Fig 1D). Likewise, RNAi of T25B9.1 also led to a subtle increase of glycine in worms, suggesting that T25B9.1 tends to degrade glycine in C. elegans (Fig 1D). RNAi knockdown of genes in glycine catabolic pathways including daao-1, gss-1, gcsh-1, glycine cleavage system T-protein (gcst-1), and dihydrolipoamide dehydrogenase (dld-1), all significantly increased endogenous glycine compared to control (Fig 1E). Surprisingly, RNAi of phosphoribosylformylglycinamidine synthase (pfas-1), which is thought to block glycine being used in purine synthesis, was found to lower the level of glycine (Fig 1E). This implies that there are complex metabolic consequences from the perturbation of de novo purine synthesis. Collectively, these data suggest that the majority of glycine in C. elegans is influenced by two metabolic branches, namely one-carbon metabolism via glycine cleavage complex and serine synthesis via MEL-32.

We next verified the effects of glycine on lifespan by administering various concentrations of glycine to worms. To avoid influences of glycine on bacterial metabolism and vice versa, we killed E. coli OP50 with a combination of ultraviolet (UV)-irradiation and antibiotic (carbenicillin) supplementation.

In line with a previously reported observation [26], worms being fed UV- and carbenicillin-killed E. coli OP50 (referred to hereafter as "killed bacteria") live significantly longer compared to those being fed live E. coli OP50 (S3 Fig). Therefore, to confirm if glycine still accumulates in aged worms after switching to a killed bacteria diet, we again quantified amino acids levels in worms at different stages of C. elegans lifespan including L3 (larval stage 3), day 1 (D1), day 3 (D3), day 6 (D6) and day 9 (D9) of adulthood. We found that most amino acids remained unchanged from L3 to D9, including valine, tryptophan, lysine, isoleucine, glutamine, asparagine, aspartate, arginine, serine, and proline (S4A Fig). Some amino acids change with age such as leucine, methionine, ornithine, and glutamate (S4A Fig). The levels of these either peak at the L3 stage or at D3, then decrease with age (S4A Fig). Interestingly, the level of tyrosine peaked at both L3 and D9, and the level of alanine peaked at D3, then maintained stable from D3 to D9 (S4A Fig). Although the levels of some of these amino acids in worms fed killed bacteria were more stable with age compared to those in worms fed live OP50 [22], we consistently found that the levels of leucine, methionine, ornithine, and glutamate decreased and the level of glycine increased in aged worms (Fig 2A and S4A Fig). The results suggest that the changes of these amino acids during aging are robust phenotypes that are independent of the worms being fed live or killed bacteria.

On killed bacteria we tested how a range of glycine concentrations from 5 μM to 10 mM affects the lifespan of C. elegans. We observed a significant increase in median lifespan at concentrations of 5 μM, 50 μM, and 500 μM of dietary glycine as compared to untreated controls, with a 7.7% (p < 0.0001), 19.2% (p < 0.0001), and 19.2% (p < 0.0001) extension observed, respectively (Fig 2B). Higher concentrations, however, including 5 mM and 10 mM, failed to extend worm lifespan suggesting a dose response where only low concentrations of glycine between 5–500 μM are beneficial to lifespan (Fig 2B, S4B Fig). Our findings are in agreement with previous studies suggesting a dose-effect of amino acids on worm lifespan [12].

Next, we measured amino acid levels in C. elegans at D1 of the adult stage to investigate how glycine supplementations at different doses affect glycine levels in vivo. We did not detect obvious changes of glycine abundance itself or of the other amino acids at any concentration of supplementations such as serine and threonine, the substrates for glycine synthesis (S5A–S5E Fig). These results suggest that in worms, glycine immediately fuels metabolic pathways, likely through glycine cleavage complex and MEL-32 (Fig 1D and 1E). Moreover, the results suggest that the levels of amino acids are tightly controlled in the early adult stage when glycine metabolic genes are actively expressed (Fig 1B and 1C).

To test whether a specific timing is required for the beneficial effect of glycine on C. elegans lifespan, we administered glycine to worms at different times in the animal's life. These times included (a) the development phase from the time of hatching to L4, (b) the early adult period from D1 to D3, (c) the adulthood starting from D1 until death, and (d) during the entire life (Fig 2C). Intriguingly, the longevity-promotion function of glycine was observed to act exclusively during adulthood, especially the early adult phase, as supplementation from D1 to D3 was sufficient to prolong lifespan by 19.0% (p < 0.0001) (Fig 2D). In contrast, glycine supplementation during development did not exert a beneficial effect on lifespan (Fig 2D). These data suggest that the first three days of adulthood are important for glycine to confer its longevity effects in C. elegans.

RNAi knockdown of mel-32 or the components in glycine cleavage complex (e.g. gcst-1) resulted in pronounced increases of the endogenous levels of glycine in C. elegans (Fig 1D and 1E). As such, we wanted to determine whether such increase by blocking two distinct metabolic branches also impacts lifespan in C. elegans. We hence assayed the lifespan of animals fed RNAi targeting mel-32 and gcst-1 from the time of hatching. Intriguingly, we found that while RNAi of gcst-1 increased the endogenous level of glycine 6.9-fold (Fig 1E), it failed to increase lifespan in C. elegans (Fig 2E). In contrast, mel-32 RNAi not only increased the endogenous level of glycine robustly (6.7-fold) (Fig 1D), but also improved the survival of wild-type N2 C. elegans significantly (Fig 2F). Thus, these results suggest that elevated endogenous or exogenous glycine levels drive lifespan extension, and that an intact glycine cleavage complex is likely required for this extension.

As glycine is an important one-carbon donor and an essential substrate for de novo purine synthesis, we next sought to understand how one-carbon and purine metabolism are influenced by aging. We therefore quantified the expression levels of genes involved in one-carbon metabolism (Fig 3A) at five developmental distinct time points (L3, D1, D3, D6, and D9) in the worm lifespan. We found that with the exception of tyms-1 and dhfr-1 (thymidylate synthetase and dihydrofolate reductase) which were upregulated in aged worms, the expression of genes participating in transferring the one-carbon moiety of glycine to form SAMe including gcst-1 (glycine cleavage system T-protein), dld-1 (dihydrolipoamide dehydrogenase), gcsh-1 (glycine cleavage system H-protein), mthf-1 (methylene tetrahydrofolate reductase), metr-1 (methionine synthase), and sams-1 (S-adenosyl methionine synthetase), dropped dramatically in aged animals (Fig 3B) [27]. Furthermore, the de novo purine synthesis genes atic-1 (5-aminoimidazole-4-carboxamide ribonucleotide formyltransferase/IMP cyclohydrolase) (Fig 3B) and F38B6.4 (Fig 1C) were markedly reduced in aged worms compared to their expression levels in worms at D1. Together these data suggest a downregulation of both one-carbon and purine metabolism during aging in C. elegans.

To resolve in greater detail how glycine supplementation counteracts the age-related changes in one-carbon and purine metabolic genes, we performed next generation RNA-sequencing on D1 worms which were fed control diet (UV-killed E. coli OP50) and 500 μM glycine-supplemented diet from the time of hatching, respectively. In contrast to the changes occurring with age, glycine induced a marked increase in the expression of genes in purine metabolic pathways (Fig 3C). Concomitantly, several genes in one-carbon metabolism (indicated in red in S6 Fig) were differentially expressed, including upregulation of atic-1, F38B6.4, dhfr-1, tyms-1, mel-32, and gcsh-1, and downregulation of sams-1, mthf-1, and gldc-1 (S6 Fig). These results suggest that the expression of purine metabolism genes is subjected to an overall upregulation by exogenous supplementation of dietary glycine, while genes in one-carbon metabolism are under complex regulations upon glycine supplementation.

Having determined one of the effects of glycine on worm lifespan and its ability to partly counteract age-related declines in gene expressions in one-carbon and purine metabolic pathways, we next aimed to investigate whether similar gene regulatory events are also present in long-lived mutant worms. To test this, we turned to microarray datasets from three long-lived worm models, daf-2(e1370) and eat-2(ad465) [28] or mrps-5 RNAi worms (reported here). We specifically looked at glycine-associated metabolic pathways in the Kyoto Encyclopedia of Genes and Genomes (KEGG) database including 'glycine serine and threonine metabolism', 'one carbon pool by folate', 'cysteine and methionine metabolism', and 'purine metabolism'. Strikingly, although distinct longevity pathways are known to be active in these long-lived worms [2,29,30], all these longevity worm models consistently showed a transcriptional activation of glycine metabolism, folate-dependent one-carbon metabolism, and methionine metabolism (S7A–S7C Fig). This suggests that transcriptional activation of glycine and one-carbon metabolism is a prominent signature of longevity shared by these long-lived worms. In contrast, overall purine metabolism including de novo synthesis, the salvage pathways, and purine degradation was only mildly deactivated across the three long-lived strains (S7A–S7C Fig).

To identify prominent transcriptional features in the three long-lived strains, we next examined the expression profile of individual genes belonging to "glycine, serine and threonine metabolism", as well as folate-mediated one-carbon metabolism and purine metabolism. We found that genes involved in glycine anabolism including T09B4.8, agxt-1, C15B12.1, R102.4, and T25B9.1 were upregulated in daf-2(e1370), eat-2(ad465) [28] and mrps-5 RNAi worms (S7D–S7F Fig). Moreover, a concomitant rise in the expression levels of genes in glycine catabolism and de novo purine synthesis occurred, including gldc-1, gcst-1, gcsh-1, F38B6.4, and atic-1 (S7D–S7I Fig), suggesting a stimulation of metabolic activity of glycine-associated processes in these long-lived worms. To query the expression of genes in the production of SAMe in the long-lived worm models, we specifically checked the expression of five homologues of SAMe synthetases in C. elegans from the microarray data including sams-1, sams-2, sams-3, sams-4, and sams-5 (S8 Fig). Interestingly, the expression of sams-1, which encodes the enzyme accounting for the majority of overall SAMe production in C. elegans [31], was significantly upregulated in all these long-lived worms (S8 Fig). Moreover, sams-5 was increased in daf-2(e1370) and eat-2(ad465) [28], while sams-2, sams-3, and sams-4 were suppressed in mrps-5 RNAi and daf-2(e1370) (S8 Fig). Collectively, the data point to elevated methylation activities in these long-lived worms.

In contrast to an overall upregulation of purine metabolism genes in response to dietary glycine treatment, all three longevity worm models specifically induced the expression of genes in de novo purine synthesis compared to control, including F38B6.4, F10F2.2 (pfas-1, phosphoribosylformylglycinamidine synthase), B0286.3 (pacs-1, phosphoribosylaminoimidazole succinocarboxamide synthetase), and atic-1 (S9A–S9C Fig).

Taken together, our data reveal that transcriptional activations of glycine and glycine-associated pathways, including one-carbon and de novo purine synthesis, are present in three distinct longevity models.

To understand the mechanism of glycine-mediated lifespan extension on a more global scale, we returned to our next-generation RNA-sequencing dataset and performed unsupervised Principle Component Analysis (PCA) on the individual libraries. We found a clear separation between glycine-treated versus untreated samples (Fig 4A), corresponding to a large difference in gene expression (Fig 4B). Interestingly, more genes were transcriptionally repressed in response to glycine treatment in which 2629 genes were differentially down-regulated, and 983 genes were up-regulated compared to control (Fig 4B). This suggests an inhibition propensity of glycine on gene expression.

To probe the processes changed upon glycine supplementation, we performed gene ontology (GO) term enrichment analysis on the differentially expressed genes using the Database for Annotation, Visualization and Integrated Discovery (DAVID) bioinformatics resource [32]. A larger number of GO terms were found enriched in the downregulated gene list, among which were GO terms for body morphogenesis, growth regulation, post-embryonic development, molting cycle, cuticle development, multicellular organism growth, and positive regulation of growth rate (Fig 4C). These enrichments are all related to growth control, suggesting a decelerating effect of glycine upon development and growth which is a phenomenon known to be associated to longevity [33]. Additionally, unlike some mutations that confer longevity to the soma at the cost of a reduction in fecundity [34], supplementation of glycine mildly enhanced the expression of genes in reproduction-related biological processes such as the GO terms of vitellogenesis, meiosis cell cycle, and gamete generation (Fig 4D). While we did not observe clear differences in the number or the size of embryos, the progenies from glycine- and serine-supplemented worms seem to be healthy and normal. Overall, with the observations of the lifespan extending effects of glycine, these data suggest that the beneficial role of glycine slows down pathways that are traditionally ameliorated in healthy aging models.

MEL-32 and glycine cleavage complex are major gatekeepers responsible for the flux of glycine into metabolic pathways (Fig 1A). Particularly, although suppression of mel-32 or the components in glycine cleavage complex in C. elegans dramatically elevated the level of endogenous glycine (Fig 1D and 1E), only mel-32 RNAi showed a potent lifespan extension effect on worms likely by favoring the flux of glycine into one-carbon metabolism (Fig 2E and 2F). Additionally, by fueling the one-carbon metabolic network through glycine cleavage complex, glycine contributes to the synthesis of SAMe (Fig 3A), the availability of which has been implicated in the regulation of histone methylation patterns and subsequently gene expressions [35]. Collectively, this led us to hypothesize that the methionine cycle may be required for the lifespan-extending effect of glycine. To test if the methionine cycle is necessary for the longevity effect of glycine, we performed lifespan analyses with the methionine cycle-deficient mutants metr-1(ok521) and sams-1(ok3033). In these mutants, 500 μM glycine failed to promote lifespan (Fig 5A–5C), demonstrating that the effects of glycine on worm longevity depend on the methionine cycle.

Serine is another important one-carbon donor and the major precursor for glycine synthesis in vivo [14]. Thus, serine and glycine are closely related to each other in one-carbon metabolism. Given their similarities, we next investigated whether serine can also exert beneficial effects on worm lifespan. Serine supplementation at a concentration from 1 mM to 10 mM has been shown previously to extend worm lifespan [12]. We therefore administered 5 mM serine to worms and measured lifespan. Similar to glycine, we confirmed that serine prolonged the lifespan of worms (+20.8% in median lifespan) (Fig 6A).

Given that the adulthood is important for glycine-mediated lifespan extension, we investigated whether serine increases lifespan in the same fashion as glycine. Similarly, we treated worms with serine at different times in worm's lifespan including (a) developmental stage, (b) the beginning of adulthood from D1 to D3, (c) adulthood from D1 until death, and (d) during the whole lifetime. Similar to the effects of glycine on lifespan, serine treatment from D1 to D3 is sufficient to recapitulate the beneficial effects on lifespan as did the treatment throughout the entire life or during adulthood only (Fig 6B). In contrast, supplementation during the developmental stage failed to increase lifespan (Fig 6B). This further implies that serine acts on the same downstream longevity signalling pathways to influence aging as does glycine.

We further investigated whether the anti-aging effects of serine also rely on the methionine cycle. In agreement with the results observed with glycine supplementation, the lifespan extending effect of serine was also ablated by mutations of metr-1 and sams-1 (Fig 6C and 6D). These results further suggest that serine and glycine prolong C. elegans lifespan via a similar mechanism.

To determine the common regulators in both glycine and serine-mediated longevity, we performed RNA-sequencing on serine-supplemented worms. As expected, PCA analysis showed a clear separation between worms treated with either of the amino acids when compared to non-treated worms, and a strong similarity between glycine- and serine-treated worm groups (Fig 6E). Statistical analysis found one significantly differentially expressed gene, F38B6.4 (S10A Fig), an enzyme that consumes glycine for purine synthesis. In addition, visualizing the data as a volcano plot showed again a greater number of genes repressed by serine treatment (2865 downregulated genes vs 973 upregulated genes), in line with the same gene expression suppression pattern of glycine-supplemented worms (S10B Fig, Fig 4B). Likewise, we found a strong overlap between glycine- and serine-treated worms when looking at the up- and downregulated genes, as shown in the Venn diagrams where 82.8% (2335) of the downregulated and 72.3% (804) of the upregulated genes are shared (Fig 6E–6G). Taken together, we suggest a model whereby both glycine and serine supplementation stimulate longevity in a methionine cycle-dependent fashion and through common signaling pathways. Moreover, this seems to be dependent on the expression of sams-1 and metr-1. This model is illustrated in Fig 6H.

C. elegans strains N2 Bristol, RB2204 sams-1(ok3033)X, RB755 metr-1(ok521)II, eat-2(ad465), and daf-2(e1370) were obtained from the Caenorhabditis Genetics Center (CGC, University of Minnesota). Nematodes were grown and maintained on Nematode growth media (NGM) agar plates at 20°C as previously described [50].

E. coli OP50 and E. coli HT115 (DE3) with the empty vector L4440 was obtained from the CGC. Bacterial feeding RNAi experiments were carried out as described [51]. RNAi E. coli feeding clones used were mrps-5 (E02A10.1), T09B4.8, agxt-1 (T14D7.1), C15B12.1, R102.4, T25B9.1, gss-1 (M176.2), gcsh-1 (D1025.2), gcst-1 (F25B4.1), pfas-1 (F10F2.2), and dld-1 (LLC1.3). Clones of agxt-1 (T14D7.1), C15B12.1, R102.4, T25B9.1, gss-1 (M176.2), gcsh-1 (D1025.2), gcst-1 (F25B4.1), and pfas-1 (F10F2.2) were derived from the Ahringer RNAi library [52]; Clones of mrps-5 (E02A10.1), T09B4.8, and dld-1 (LLC1.3) were derived from the Vidal RNAi library [53]; Worms were fed RNAi bacteria from the time of hatching unless otherwise indicated.

Glycine and serine were purchased from Merck Millipore (no. 8.1603.0250) and Sigma (no. S4500), respectively. A stock of concentration of 1 M glycine and serine was made by dissolving glycine and serine in water. The pH of glycine and serine stock solution was adjusted to 6.0–6.5 with sodium hydroxide and then sterilized with 0.45 μm Millipore filter. The concentrations of 5 μM, 50 μM, 500 μM, 5 mM and 10 mM glycine, and 5 mM serine were used in the present study.

Overnight cultures of E. coli OP50 were seeded on standard NGM plates containing carbenicillin (25 μg ml^-1) to prevent the bacterial growth. After drying overnight at room temperature, the bacterial lawn was irradiated with 254 nm UV light using a Stratalinker UV Crosslinker model 1800 (Stratagene, USA) at 999900 μJ/cm² for 5 min. A sample of UV-exposed E. coli OP50 was collected and cultured in LB medium overnight at 37°C to confirm the bacteria were completely killed. Plates seeded with UV-killed bacteria were stored in 4°C and used within 1 week after seeding.

Lifespan experiments of amino acids supplementation were performed at 20°C without fluorouracil as described with the exception that UV-killed E. coli OP50 was used [54]. Briefly, for treatment throughout lifespan, worms were cultured on glycine- or serine- supplemented plates from the time of hatching until death; For the treatment during larval development only, worms were cultured on glycine- or serine- supplemented plates from egg until L4, and then transferred onto control plates until death; For the treatment from D1 to D3, worms were cultured on glycine- or serine- supplemented plates from D1 (one day after L4) to D3, and transferred on control plates. For amino acid treatment during adulthood only, worms were cultured on glycine or serine supplemented plates from D1 until death. During the reproductive period (≈ day 1–8), worms were transferred to fresh plates every other day to separate them from their progeny.

For RNAi lifespan experiments, worms were cultured on NGM plates containing 2mM IPTG and seeded with HT115 (DE3) bacteria transformed with either pL4440 empty vector or the indicated RNAi construct from the time of hatching. Worms were transferred to fresh plates containing 10 μM fluorouracil at L4 larval stage to prevent egg laying. 100–150 worms per condition were used for every lifespan. Survival was scored every other day throughout the lifespan and a worm was considered as dead when they did not respond to three taps. Worms that were missing, displaying internal egg hatching, losing vulva integrity, and burrowing into NGM agar were censored. Statistical analyses of lifespan were calculated by Log-rank (Mantel-Cox) tests on the Kaplan-Meier curves in GraphPad Prism.

Amino acids were extracted and analyzed as described before [22] and each experiment was performed in three biological replicates. About amino acids profiles change with age in wild type N2 related to Fig 2A and S4A Fig, worms were cultured on UV-killed E. coli OP50 and collected at the desired stage (L3, D1, D3, D6, and D9) for amino acids extraction; About amino acids profiles change with age in wild type N2, daf-2(e1370), and eat-2(ad465) related to S1 Fig, worms were cultured on alive E. coli OP50 and collected at L3 and D10 for amino acids extraction; About glycine supplementation experiments related to S5A–S5E Fig, worms were fed UV-killed E. coli OP50 and supplemented with glycine at the desired concentration (5 μM, 50 μM, 500 μM, 5 mM, and 10 mM) or with water control from the time of hatching. Then, D1 worms were harvested for amino acids extraction; About RNAi experiments related to Fig 1D and 1E, and S2 Fig, worms were cultured on NGM plates containing 2mM IPTG and seeded with HT115 (DE3) bacteria transformed with either pL4440 empty vector or the indicated RNAi construct from the time of hatching. Then, D1 worms were harvested for amino acids extraction.

Around 1500 synchronized worms at the desired stage were collected, freeze-dried and stored at room temperature until use. Worm lysates were obtained by homogenization and subsequent tip sonication. Amino acids were extracted from worm lysate containing 50 μg protein and measure by UPLC-MS/MS analysis.

In Fig 1B and 1C, worms were cultured on live E. coli OP50 and harvested at D1, D4, and D9 for mRNA extraction; In Fig 3B, worms were cultured on UV-killed E. coli OP50 and collected at L3, D1, D3, D6, and D9 for mRNA extraction. Approximately 500 worms were collected in three biological replicates per condition at the desired stage. Total RNA was isolated according to the manufacturer's protocol. Briefly, samples were homogenized in TRIzol (Invitrogen) with a 5 mm steel metal bead and shaken using a TissueLyser II (Qiagen) for 5 min at a frequency of 30 times/sec. RNA was quantified with a NanoDrop 2000 spectrophotometer (Thermo Scientific) and stored at -80°C until use. Genomic DNA was eliminated, and cDNA was synthesized using the QuantiTect Reverse Transcription kit (QIAGEN). The qPCR reaction was carried out in 8 μL with a primer concentration of 1 μM and SYBR Green Master mix (Roche) in a Roche LightCycler 480 system. In all analyses, the geometric mean of two reference genes, eif-3.C and F35G12.2, was used for normalization and the oligonucleotides used for PCR are listed in S2 Table.

daf-2(e1370) and eat-2(ad465) worms were fed HT115 E. coli expressing empty vector from the time of hatching. Wild type N2 worms were fed HT115 E. coli expressing dsRNA against mrps-5 from the larval stage 4 of parental worms, and this exposure was continued in the first filial population (F1). mrps-5 RNAi-treated F1 worms were used for total RNA extraction. Microarray experiment was performed as described [28]. Approximately 500 young adult worms were collected in four replicates per condition and total RNA was extracted as described above. RNA quality and quantity were assessed after DNase clean-up using a 2100 Bioanalyzer (Agilent Technologies). 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 [55].

Worms were cultured on UV-killed E. coli OP50 upon 500 μM glycine treatment or 5 mM serine treatment from the time of hatching. Approximately 500 worms at D1 were collected in quadruplicates per condition for total RNA extraction as described above. Genomic DNA residues were eliminated with RNase-Free DNase (Qiagen), followed with the cleaning up with the RNeasy MinElute Cleanup Kit (Qiagen). Samples were sent to GenomeScan B.V. (Leiden, The Netherlands) for RNA library preparation and sequencing at a 20 million read-depth (see methods below).

Samples were processed for Illumina using the NEBNext Ultra Directional RNA Library Prep Kit (NEB #E7420) according to manufacturer's description. Briefly, rRNA was depleted using the rRNA depletion kit (NEB# E6310). A cDNA synthesis was performed in order to ligate with the sequencing adapters. Quality and yield after sample preparation was measured with the Fragment Analyzer. Size of the resulting products was consistent with the expected size distribution (a broad peak between 300–500 bp). Clustering and DNA sequencing using the Illumina cBot and HiSeq 4000 was performed according to manufacturer's protocol with a concentration of 3.0 nM of DNA. HiSeq control software HCS v3.4.0, image analysis, base calling, and quality check was performed with the Illumina data analysis pipeline RTA v2.7.7 asnd Bcl2fastq v2.17.

Reads were subjected to quality control FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc↗), trimmed using Trimmomatic v0.32 [56] and aligned to the C. elegans genome obtained from Ensembl, wbcel235.v91 using HISAT2 v2.0.4 [57]. Counts were obtained using HTSeq (v0.6.1, default parameters) [58] using the corresponding GTF taking into account the directions of the reads. Statistical analyses were performed using the edgeR [59] and limma/voom [60] R packages. All genes with no counts in any of the samples were removed whilst genes with more than 2 reads in at least 4 of the samples were kept. Count data were transformed to log2-counts per million (logCPM), normalized by applying the trimmed mean of M-values method (Robinson et al., 2010) and precision weighted using voom [61]. Differential expression was assessed using an empirical Bayes moderated t-test within limma's linear model framework including the precision weights estimated by voom [61]. Resulting p-values were corrected for multiple testing using the Benjamini-Hochberg false discovery rate. Genes were re-annotated using biomaRt using the Ensembl genome databases (v91). RNA-seq samples were compared using principal component analysis (PCA) and Partial least squares discriminant analysis (PLS-DA) using mixomics [62]. Heatmaps, venn diagrams, and volcano plots were generated using ggplot2 [63] in combination with ggrepel (https://CRAN.R-project.org/package=ggrepel↗), and venneuler (https://CRAN.R-project.org↗). Data processing and visualization was performed using R v3.4.3 and Bioconductor v3.6.

Gene ontology (GO) analyses were conducted using DAVID bioinformatics resource [32]. Genes found to be significantly up- or downregulated with an adjusted p-value < 0.05 and | log₂ (fold change) | > 0.5 were subjected to functional annotation clustering. To retrieve significantly enriched GO terms, enrichment threshold (EASE score) was set as 0.05 for all analyses and the category of each annotation cluster generated by David was curated manually.

Heat maps of gene expression profile in S7–S9 Figs were plotted using "R2", Genomics analysis and Visualization platform (http://r2.amc.nl↗). Gene set map analyses were performed on "R2" with defined gene sets from the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (http://www.genome.jp/kegg/↗) [64].

Data were analyzed by two-tailed unpaired Student's t-test or by one-way ANOVA with Tukey's post hoc test for multiple comparisons, except for survival curves, which were calculated using the log-rank (Mantel-Cox) method. For all experiments, data are shown as mean ± SD and a p-value < 0.05 was considered significant.