Long-lived mutants have distinctive metabolic profiles
Of the various pathways known to regulate longevity in worms, the best known is the Insulin/Insulin-Like signalling (IIS) pathway [32
]. Many mutations that disrupt components of this pathway affect the ability of larval worms to enter and leave the dauer stage, but they also increase the longevity and stress resistance of adults as well as reduce their fecundity [2
We began by studying m41
, a hypomorph mutation that disrupts daf-2
which encodes a tyrosine kinase that is expressed throughout the worm and is thought to act as a receptor for many of the 37 insulin-like ligands present in the C. elegans
hermaphrodites are 10 to 90% longer-lived than wild-type worms [3
] (our data not shown). Since m41
is a dauer-constitutive temperature-sensitive mutation we grew these worms at the permissive temperature, 15°C, until L4, transferred them to 22.5°C, and assayed their metabolites as old adults (240 hours). We did this by freezing the worms instantly in liquid nitrogen, extracting polar metabolites, and then acquiring 1
H NMR spectra. The spectra showed a range of resonances from small molecule metabolites, typical of tissue extracts. We then divided the spectra into bins each chosen to represent as far as possible a single metabolite resonance. Principal components analysis (PCA) and hierarchical cluster analysis (HCA) of the reduced data showed that daf-2(m41)
and wild-type samples have distinct metabolic profiles with little overlap between the two groups of samples in PC1 (Figure ). The loadings along this axis showed that many NMR-detectable metabolites contribute to the difference between the genotypes (Figure ).
Figure 1 The insulin-signalling mutant daf-2(m41) has a distinctive metabolic profile. In this experiment m41 and N2 wild-type worms were initially raised at 15°C, transferred to 22.5°C at L4, and assayed at 240 hours post-bleaching. A. PCA (above) (more ...)
Many IIS mutations exist and they differ in the severity and kind of their phenotypic effects [3
]. So, in a separate experiment we simultaneously studied three daf-2
mutations: m41, e1370
as well as daf-28(sa191)
which disrupts an insulin-like ligand thought to bind DAF-2. DAF-28 is thought to activate DAF-2 and so promote normal, reproductive growth and longevity, but sa191
is a dominant negative gain-of-function allele [35
]. Like daf-2
is partly dauer constitutive, has long-lived adults, and can be repressed by mutations in daf-16
]. In this experiment, we used fewer samples of worms per genotype than in previous experiments, but sampled three ages, L1 (15 hours), middle-age (144 hours) and old age (240 hours), and raised them first at 15°C and then at 25°C. Considering just the old-age samples we found that all of these mutants have distinctive metabolic profiles, with e1370
having respectively the least and most distinctive metabolic phenotypes relative to wild-type (Figure ). The rank order of allele strength for longevity at 22.5°C and 25°C is e1370
(our data not shown; [3
]), but daf-2
alleles have a variety of phenotypic effects which do not all show the same rank order of severity [3
]. The three daf-2
mutations affect different parts of the receptor: e1370
disrupts the intracellular tyrosine kinase domain while m41
disrupt, respectively, the Cysteine Rich and Leucine Rich L2 extracellular domains [34
]. Some L2 domain mutations in the human Insulin receptors have very low ligand binding affinity [40
]; a similar property of daf-2(m596)
may explain why its metabolome resembles that of daf-28(sa191)
which disrupts a putative ligand.
Figure 2 Longevity mutants and dauer larvae have distinctive metabolic profiles. A. PCA of four IIS mutants and a long-lived translation-defective mutant, ife-2(ok306); cluster analysis separates the mutants into distinct groups. daf-2(e1370) is the most similar (more ...)
When surveying the four daf-2(-)
mutants, we also looked at another kind of long-lived mutant, ife-2(ok306)
, which disrupts a gene encoding an isoform of the eukaryotic translation initiation factor, eIF4E [9
]. Since this mutation does not require DAF-16 to confer increased longevity, it is thought that IFE-2 works either downstream or in parallel to DAF-2 to regulate longevity. We found that the metabolic profile of ife-2(ok306)
is very similar to that of the ILS mutants: cluster analysis and PCA do not clearly separate ife-2(ok306)
worms from ILS mutants (Figure ).
Since we sampled all of our mutants at three ages we were also able to study, at least crudely, when the mutant worms acquired their distinctive metabolic profiles. PCA and cluster analysis shows that all mutants had distinctive metabolic profiles even as larvae, but in all cases the metabolic profiles became increasingly different from wild-type with age (Figure ). Since reproduction has not yet begun in L1 larvae, which do not even have gonads, the distinctive profiles of the long-lived mutants cannot be entirely due to a decrease in metabolic resources allocated to reproduction.
Finally, in this same experiment, we also studied dauer larvae. Dauers form when L2 worms are crowded or deprived of food. They do not feed, have very distinctive transcriptional profiles, physiologies and morphologies, are very stress resistant and also do not age [1
]. We raised dauers at two temperatures, 20°C (n = 3) and 25°C (n = 5), and compared them to old adults (240 hours) raised at the same temperatures (n = 5 for both). Clustering and PCA showed that dauers and adults have unambiguously distinct metabolic profiles as do worms raised at different temperatures, with temperature nested within the two stages (Figure ). Comparing dauers to L1s or young adults gave very similar results (data not shown). We found that metabolite levels showed strong stage × temperature interactions. This is reflected in the reversal of the relative positions of dauers and adults along the PC 2 axis depending on the temperature and also in analysis of variance on individual bins (data not shown).
The metabolic signature of long life in worms
What are the metabolic features of long-life? In C. elegans, many mutants and environmental treatments confer increased longevity, but the devices by which they do so, or whether or not they are the same, remains unclear. One reason for this is that few studies assay more than one class of long-lived worm using the same technology and experimental conditions. Since we have studied three classes of long-lived worm, dauers, IIS defective and a translation-defective mutant, using the same metabolic profiling technique, we can directly compare them and ask what, if anything, do they have in common?
To do this we determined the relative concentrations of 26 metabolites (Figure ; Additional file 1
). Then, to identify the metabolic signatures of long-life we ranked them by consistency and direction of response in long-lived worms relative to wild-type (Figure ). Although this figure shows the results of all of our experiments, here we focus on the single experiment in which worms were raised at 25°C and sampled at 10 days after hatching.
Figure 3 The metabolic signature of long-life. A. Relative concentrations of 26 metabolites in worms sampled at 25°C and 240 hours post-hatching. B. Summary of metabolic responses. Here we show, for all experiments, the observed response in long-lived (more ...)
We found that the metabolic responses of our long-lived worms were strikingly similar. More than half of the examined metabolites show qualitatively similar changes in dauers, IIS mutants, and ife-2 mutant worms. This result was surprising since IIS and ife-2 mutations ostensibly influence very different aspects of the worm's physiology. We propose that these metabolites constitute a minimal metabolic signature of long-life in worms.
One of the signature metabolites was the disaccharide trehalose. An important carbohydrate storage molecule in nematodes, trehalose is thought to confer stress resistance in many invertebrates [41
]. Previous studies have shown that the expression of genes involved in its synthesis are elevated in dauers and IIS mutants so our finding that trehalose pool sizes are elevated in dauers and IIS mutants was expected; indeed, trehalose has been proposed as a longevity assurance sugar
worms show high trehalose levels as well implying that a deficiency in protein synthesis can affect carbohydrate metabolism as well.
This similarity across different classes of long-lived worms is also seen in amino acid levels. Of the 12 amino acids we studied, 11 are regulated in IIS mutants. Of these, 10 are regulated in the same way in ife-2(oK306) and 5 are in dauers. We also found the following metabolites consistently regulated across dauer, IIS, and ife-2(0) mutants: choline, phosphocholine, and glycerophosphocholine (GPC), which are associated with lipid metabolism; acetate, malate and succinate, which are associated with carbohydrate metabolism; propanoate and NAD+. Some of these longevity-signature metabolites are expected from previous studies of long-lived worms. This is particularly true of metabolites that have a role in carbohydrate metabolism and we consider them in greater detail below. Others, such as altered pool sizes of amino acids, choline, phosphocholine, propanoate and NAD+, could point to new mechanisms of longevity assurance in C. elegans. We also note that, although we have ignored metabolic responses peculiar to particular worm strains, we cannot exclude the possibility that they influence longevity as well.
Autophagy and the dauer metabolome
Although the dauer metabolome resembles that of long-lived mutants in many ways, we also found that it has some unique features. We detected pools of two post-translationally modified amino acids, phosphoserine and hydroxyproline, in dauers and only dauers. Since hydroxylation and phosphorylation generally take place on peptides rather than free amino acids, these pools are likely the result of protein degradation. The most obvious source of free hydroxyproline is collagen: the C. elegans
genome contains approximately 175 genes encoding collagens which are used in the basement membranes and cuticle, two prolyl 4-hydroxylases, and at least five peptidases that are required for the processing or turn-over of cuticle collagens [45
]. The source of free phosphoserine is less obvious, however, given that we studied high concentration metabolites, they probably are not derived from the phosphorylated serines found in signalling pathways but rather represent structural components. We suspect that the phosphoserine pool seen in dauers is derived from serine phosphoglyceride lipids freed by turnover of membrane phospholipids. This is supported by the observation that choline compounds (choline, phosphocholine, and GPC) are also strongly increased in dauers, and altered choline compound concentrations are frequently observed in mammalian tumours where they mark the membrane turnover characteristic of rapidly proliferating cells [47
]. Since dauers do not feed they rely on energy stores such as fats and glycogen to survive, and much evidence shows that the beta-oxidation and glyoxylate pathways that metabolize fatty acids are upregulated in dauers [19
]. One explanation, then, for the elevated levels of these modified amino acids is that dauers are utilizing spare extracellular matrix, and other proteins, as yet another energy store. Consistent with this idea, dauer morphogenesis requires extensive autophagy [51
]. A mutation that abolishes autophagy also decreases longevity specifically in daf-2(e1370)
which implies that high levels of autophagy promote long-life [51
]; if so we did not detect any obvious signatures of this.
Some, but not all, signature metabolite responses require DAF-16
Much evidence shows that the longevity prolonging effects of IIS mutants are mediated by the FOXO transcription factor, DAF-16 [2
]. Down-regulation of DAF-2 signalling results in nuclear localization and hence activation of DAF-16 which, in turn, activates or represses many genes which contribute to longevity [14
]. Since the longevity prolonging effect of daf-2
mutants is repressed by null mutations of daf-16
, one way of disentangling the phenotypes of IIS mutants that contribute causally to longevity from those that do not, is to ask whether they, too, depend on DAF-16 activity. The reasoning, previously applied to transcriptomic and proteomic data, is that any molecular phenotype that contributes to longevity should be abolished by inhibiting DAF-16 activity just as longevity itself [15
To find out whether IIS control over metabolism was also DAF-16 dependent we compared the metabolomes of 144 hour-old wild-type worms to worms carrying either another daf-2 allele, e1370, or the null daf-16(m26) mutation, or both. Unsupervised methods (PCA and HCA) divide these samples into two major groups: a cluster which contains 7/8 daf-2(-) samples and a cluster that contains the rest (Figure ). Thus the metabolic phenotype of daf-2 is at least partially DAF-16 dependent; however the double mutants form a sub-cluster clearly distinct from wild-type implying that not all of the daf-2 phenotype is so.
Figure 4 DAF-16 dependence of metabolites. A. PCA of metabolic profiles of wild-type, single mutant daf-2(e1370) and daf-16(m26) and double-mutant daf-2(e1370);daf-16(m26) samples. PCA and cluster analysis shows that daf-2(e1370) have the most divergent metabolism (more ...)
By measuring metabolite levels directly in single mutant daf-2(e1370)
and double mutant daf-2(e1370);daf-16(m26)
worms we were able to apply this test to 11 metabolites (Figure ). Of these only four, phosphocholine, and the closely related amino acids isoleucine, valine, and leucine, showed the classic pattern of DAF-16 dependence: concentrations of each were substantially elevated or repressed in daf-2(-)
, but not daf-16(0)
or double mutant samples. Several other metabolites showed more complex patterns of epistasis. For example, lysine, lactate and glycerol concentrations were all significantly different in either daf-16(0)
worms or both compared to wild-type worms (P =
0.05; two-tailed t-test) and so fail the classical test of DAF-16 dependence. Surprisingly, trehalose showed no sign of DAF-16 dependence: daf-16(0)
samples have wild-type levels of the sugar, but double mutants are not different from daf-2(-)
. This is in contrast to previous results shown by transcriptomic studies [15
], which may perhaps reflect the fact that changes in gene expression levels alone do not necessarily equate to functional differences [56
Classical DAF-16 dependence make isoleucine, valine, leucine and phosphocholine strong candidates for having a causal role in long life, particularly as all four are signature metabolites as defined above. Conversely, the absence of DAF-16 dependence in trehalose suggests that DAF-2 regulates it via another transcription factor parallel to DAF-16 and that it may not contribute to long life. The interpretation of non-classical daf-16
epistasis is less clear. The classical test supposes that DAF-16 is fully repressed in normal worms, but activated in the absence of DAF-2 signalling by translocation of the transcription factor from cytoplasm to nuclei [55
]. This model is certainly too simple since daf-16(0)
mutants have a variety of subtle phenotypes such as rapid growth, early reproduction and a slightly reduced lifespan, and normal worms have at least some DAF-16 visible in their nuclei [59
]. Some metabolites are, then, also apparently sensitive to low levels of DAF-16 activity.
Metabolic targets of DAF-2 signalling
We have shown that the pool sizes of many metabolites differ between long-lived and normal worms and that some of these differences are shared by various long-lived mutants, in particular the several daf-2
alleles that we studied. But what genes does daf-2
regulate that result in these changes? In order to investigate this, we mapped some of our signature metabolites onto a standard metazoan metabolic network so that we could identify those parts of the network that are altered in daf-2
worms. We then asked whether the genes encoding metabolic enzymes that work in the same parts of the network are regulated as well, and if so, whether the two sources of data could be used to give an economical account of how the metabolism of C. elegans
is altered in daf-2
worms. As an initial guide we used NEMAPATH [61
] to identify, for each metabolite, the pathways in which they might work and the C. elegans
genes that might act in them. We then interrogated a previously published global expression dataset based on daf-2(e1370)
worms raised under conditions similar to ours [20
] and examined the expression patterns of these genes for patterns of co-regulation.
We found that at least five of the metabolites regulated in daf-2(-)
worms, malate, acetate, succinate, glucose and trehalose, were linked by three major pathways: the glyoxylate shunt, gluconeogenesis, and starch metabolism. Consistent with previous studies [15
] we found that genes encoding enzymes in these pathways are up-regulated in daf-2(-)
worms. In all, the C. elegans
genome contains 38 genes encoding components of 20 enzymes that work in these pathways. Of these 38 genes, 14 are significantly up-regulated, 5 are significantly down-regulated, 18 are not regulated, and 1 has no data (Figure , Additional file 2
Figure 5 Carbohydrate metabolism in daf-2 worms. Five signature metabolites - malate, acetate, succinate, glucose and trehalose - are linked by three major pathways: the glyoxylate shunt, gluconeogenesis, and starch metabolism. Expression data  shows that (more ...)
While the observed shifts in carbohydrate metabolism might have been expected from the results of gene expression studies, the changes in amino acid metabolism shown by our data were not. Most amino acid pools are upregulated in long-lived worms (8/12 in the IIS mutants; 10/12 in ife-2
). One possible explanation for this is that protein synthesis is generally repressed in IIS mutants as it is in ife-2
] and that the amino acid pool represents a surplus. daf-2(e1370)
worms may have reduced protein synthesis since the expression of their t-RNA synthetases are generally repressed (Additional file 2
). However, there is no correlation between amino acid pool size and tRNA synthetase expression (data not shown). It is more likely then that amino acid pool sizes are dictated by catabolic pathways that direct them to energy production or other uses. Consistent with this, many genes that encode components of phenylalanine and tyrosine catabolic pathways, among them tyrosinase and phenylalanine hydroxylase, are regulated in daf-2
(Additional file 2
]. Intriguingly, melanin, a tyrosine metabolite, has recently been discovered in worm cuticles where it is thought to have a protective function [63
The most striking response among the amino acids, however, is the upregulation of the branched chain amino-acids (BCAAs) isoleucine, leucine and valine. The pool sizes of these amino acids are positively correlated across long-lived mutants (Figure ). Furthermore, unlike most other metabolites, their upregulation in daf-2
is entirely DAF-16 dependent, making them strong candidates for being causally involved in longevity (Figure ). Like other animals, C. elegans
cannot synthesize these amino acids [64
], and so any difference in their relative concentrations must be due to a change in either protein turnover or their catabolism. In fact, BCAA pool sizes are co-regulated in many circumstances such as growth in worms [31
] or obesity in humans [65
]. This co-regulation is a consequence of them sharing the first two steps in their catabolic pathways: transamination by BCAT and oxidative carboxylation by the mitochondrial BCKD enzyme complex [65
]. In daf-2(e1370)
worms, BCAT expression is wild-type, but all four genes encoding components of the BCKD complex are strongly downregulated (Additional file 2
). We hypothesize that downregulation of the BCKD complex is responsible for the increased BCAA pool sizes of daf-2
worms. This hypothesis also suggests a way to manipulate BCAA pool sizes to test their contribution to long life. Strong inactivation of BCKD-complex genes in worms causes severe embryonic and larval phenotypes ([67
]; wormbase.org) and, in humans, maple syrup urine disease, a metabolic disorder resulting in encephalopathy and death [68
]; however, it remains possible that more subtle elevation of BCAA levels by diet or partial downregulation of the BCKD complex will confer long life.