Modulation of lipid biosynthesis contributes to stress resistance and longevity of C. elegans mutants

Construction and characterization of outcrossed strains

A series of congenic C. elegans longevity-mutant strains was created by repeated out-crossing of strains that had been reported to be long-lived, although they were initially isolated in diverse genetic backgrounds and characterized relative to different N2 control stocks [3,7,18-20]. Survivals were conducted for these strains (e.g., Figure 1); median and mean lifespans from several assays, normalized to their isogenic control (N2DRM), are summarized in Table 1. Adult survival times for these strains varied slightly between experiments, whereas relative lifespans (normalized to wild-type controls, assessed simultaneously) were highly reproducible. Several departures from the literature values were observed, which could be due to strain outcrossing into the N2DRM genetic back-ground, use of a different “wild-type” reference strain than originally employed to evaluate life extension, or inter-laboratory differences in culture conditions. Among the strains utilized herein, unc-31(e928) and eat-18(ad820) conferred less life extension than previously reported, and strain SR803 (carrying the old-1(zIs3000) transgene integrated in an N2DRM background [21]) was indistinguishable from N2DRM controls.

The wild-type progenitor strain N2DRM, near-isogenic with all mutant strains studied, served as their common control. Double mutants with daf-16(mu26) provided additional controls that largely or entirely revert life-span and stress-resistance traits of single-gene insulin/IGF-1 signaling (IIS) mutants [7,22,23], although it cannot be assumed that all affected traits are identically reversed. By this comparative design, we sought features of fatty-acid metabolism that correlate with strain longevity, while controlling for mutation-specific phenotypes not associated with longevity. Because all three double-mutant controls and one of the reportedly long-lived mutants, old-1(zIs3000), had longevities very close to that of N2DRM controls, in log-linear regressions they serve chiefly as additional controls for the consequences of the more effective IIS mutants. Although relatively modest, the lifespan extensions by unc-31 and eat-18 mutations were reproducible and sometimes significant; they thus provide informative assay points that contribute to all measures of correlation, and in particular to rank-order correlations.

Identification and quantitation of fatty-acid chains

In total, 23 fatty acids (FA) were detected, positively identified, and quantified for the worm samples examined, which included most of the fatty acids previously reported for wild-type C. elegans [24,25]. These comprised two cyclopropane fatty acids (9,10-methylene 16:0 and 11,12-methylene 18:0), two monomethyl branched-chain fatty acids (15:0-iso and 17:0-iso), five saturated FAs (14:0, 16:0, 18:0, 20:0 and 22:0), five mono-unsaturated fatty acids (16:1∆9, 18:1∆9, 18:1∆11, 20:1∆11, and 22:1∆13), and eight poly-unsaturated fatty acids, of which five are n-6 (or “omega-6”) fatty acids (18:2∆9,12, 18:3∆6,9,12, 20:3∆8,11,14, 20:4∆5,8,11,14, and 20:2∆11,14), while three are n-3 or “omega-3” chains (18:3∆9,12,15, 20:4∆8,11,14,17, and 20:5∆5,8,11,14,17). Of these, fatty acids 20:0, 22:0, 20:2∆11,14(n-6) and 22:1∆13(n-9) do not appear to have been previously reported in this species [24,25]. The first three are presumed to be +2C elongation products of the fatty acids 18:0, 20:0 and 18:2∆9,12(n-6), respectively, whereas 22:1(n-9) would require two +2C additions to 18:1∆9(n-9). The fatty acid 18:4(n-3), shown in schemas as a possible intermediate to formation of 20:4(n-3), was not observed; it has been observed in elo-1 mutants but not in worms with normal lipogenesis [25,26]. Supplemental Table S3a lists the percent content by strain, of 19 of the above fatty acids, excluding several that were less reliably identified or quantified in one or more strains.

The omega-terminal branched-chain fatty acids 15:0-iso and 17:0-iso were found in all C. elegans mutant and control strains. These odd-carbon-number fatty acids are produced in C. elegans by de novo synthesis [27,28], by ELO-5 and ELO-6 extension of branched primers, with essentially none (<1%) arising from ingested bacteria [25]. In contrast, fatty acids containing a cyclopropane ring are believed to be entirely of bacterial origin and vary in quantity with the bacterial diet [29]; they were not included in the analyses described. However, because total cyclo-propane fatty acids varied rather little among strains (6.3 -10.6% of total lipids), their inclusion or exclusion had very little effect on the percentages calculated for other FAs.

Quantitations of fatty acid classes for most of the congenic strains are summarized in Table 2. Double-mutant controls, in which daf-2 and age-1 mutants are reverted by a daf-16 mutation [7,22,23], agreed well with other “normal-lifespan” controls and thus were omitted from this table and Figure 2, although they contributed to the correlations. Columns at the right indicate the strengths of linear (Pearson) and rank-order (Spearman) correlations with the logarithm of strain longevity. Several of the strongest correlations are shown graphically in Figure 2, in which the mean percent of total fatty acids in each class was plotted as a function of the log₍₁₀₎ of strain lifespan. It is apparent from this figure that most log-linear correlations are dominated by the three longest-lived mutants: age-1(hx546), daf-2(e1370) and age-1(mg44). Because these three mutations affect components of the insulin/IGF-1 signaling (IIS) pathway, no conclusions based on correlation analyses alone will necessarily apply outside that pathway. Nevertheless, they serve to dissect the properties of IIS attenuation that bear most directly on longevity, from those that are peculiar to specific gene mutations or to other sequence changes tightly linked to an identified mutation.

The first correlation column, headed “N=10”, treats each of the ten strains as a single data point, defined by the means of its lifespan and GC-MS measurements. The other two correlation columns (“N=36”), consider each GC-MS measure on an independent cohort of nematodes as one data point, since these lipid quantitations had far higher variance than did relative lifespan. Rank-order correlations, shown in the right-most column, are independent of all assumptions regarding data distribution or transformation, and are also conservative in that they are influenced equally by strains of moderate or extreme longevity. To the extent that a proximal factor impacting lifespan is expected to vary in abundance in proportion to longevity, however, Pearson log-linear correlations have the merit of preserving that information. Of 21 fatty-acid parameters quantified in Table 2, 19 had Pearson log-linear correlations that were nominally significant at P<0.05, and 13 were highly significant at P<0.0024 (i.e. <0.05/21). This was true whether 10 or 36 data points are considered. In the description that follows, R and P values are given only for 10-point regressions, to maintain consistency with Figure 2. We note, however, that the 36-point analyses better reflect the extent of uncertainty due to biological variation within each group; they usually result in lower correlation coefficients, but improved significance.

Seventeen of these longevity-dependent trends met a more stringent criterion of having nominally significant Spearman rank-order correlation coefficients, P<0.05, of which seven were highly significant at P<0.0024. The data of Table 2 were compiled from a more complete listing that comprises quantitative data for 19 FA species (Supplemental Table S3a) and 18 derived measures of biological interest (Table S3b).

A positive correlation was observed between longevity and the abundance of short-chain saturated fatty acids C14:0, C16:0 and C15-iso, a trend that was reversed for the longer-chain classes 18:0, 20:0 and 22:0 (Figure 2A and Table 2). The same pattern was repeated among the monounsaturated fatty acids (MUFAs, Figure 2B), with positive correlations to lifespan for C16:1∆9 and C18:1∆11, but weaker negative correlations for C20:1 and C22:1 (Tables 2 and S3a). As expected from these trends in individual lipid species, the average chain length of all fatty acids declined in association with lifespan (Figure 2C), with log-linear correlation coefficient R = −0.87 (P< 0.002), and rank-order R_S = −0.49 (P<0.003).

Of five major polyunsaturated fatty acid (PUFA) species observed, four fell significantly in amount as longevity increased: 18:3(n-6), 20:4(n-6), 20:4(n-3) and 20:5(n-3), whereas 20:3(n-6) did not (Figure 2, B and E; Tables 2 and S3). Unlike mammals, which require dietary PUFAs to maintain health, nematodes possess all of the enzymes necessary for their biosynthesis, and thus their lipid composition depends on genetics and gene-expression as well as diet (see below). Although it remains controversial whether lifespan is indeed limited by reactive oxygen species [31,32], there is broad agreement that lipoperoxidation products can amplify via chain reactions [32,33], and thus pose a serious biological hazard for all cells. In PUFAs, the carbons situated between double bonds are especially vulnerable to peroxidation, whereas saturated and monounsaturated chains are several hundredfold less susceptible [34]. The Double Bond Index (DBI) is a simple weighted average of the number of double bonds per fatty acid molecule: Σ (1·Σ(1-DB %) + 2·Σ(2-DB %) + …. + 5·Σ(5-DB %) [35]. The Peroxidation Index (PI) is calculated in similar fashion, except that the contributions of MUFAs and PUFAs are weighted to reflect their susceptibilities to lipoperoxidation [34].

DBI and PI displayed quite similar monotonic declines with longevity, giving correlation coefficients of −0.84 and −0.85 (each P<0.0024) and rank-order correlation coefficients of −0.40 and −0.39 (P<0.02; Table 2 and Figure 2F).

C. elegans, unlike mammals, possesses omega-3 (i.e., n-3) desaturase enzymes. They are thus able to convert arachidonic acid, 20:4∆5,8,11,14(n-6), to eicosapen-taenoic acid (EPA), 20:5∆5,8,11,14,17(n-3), and the level of ω-3 desaturase activity can be inferred from the ratio of those chains [30]. Estimated ω-3 desaturase activity declined with lifespan (Figure 2G and Table 2), with log-linear correlation coefficients R of −0.71 (P<0.02) and −0.88 (P<10⁻³). Among all PUFAs, the ratio of ω-6 to ω-3 fatty acids also declined (Figure 2G) with a log-linear correlation coefficient of −0.81 (P< 0.005). Although many aspects of human health are affected by dietary intake of ω-3 fatty acids, and tend to benefit from reduction of our high ω-6/ω-3 intake ratio, this reflects constraints on mammalian biosynthesis of lipids and of lipid-derived eicosanoid signaling molecules, rather than lipoperoxidation potential (which is somewhat lower for ω-6 than ω-3 fatty acids [34]). Because the same constraints do not apply to nematode biosynthesis of lipids [25], the impact of the ω-6/ω-3 ratio on C. elegans health is not known.

The levels of other desaturase activities can be inferred in like manner from their product/substrate ratios [30]. Based on these ratios (Tables 2 and S3b), the imputed levels of Δ6 desaturase showed a moderate inverse correlation to longevity (R= −0.75, P<0.02), whereas two ratios indicative of Δ9 desaturase activities (16:1Δ9/16:0 and 18:1Δ9/18:0) displayed strongly positive correlations with lifespan (Figure 2H), with a combined log-linear R value of +0.94 (P<5×10⁻⁵). These results are fully consistent with a study of the effects of mutations to Δ9 desaturases, encoded by fat-5, -6 and -7 genes in C. elegans [24]. The Δ9 desaturases are responsible for the initial introduction of a double-bond in a saturated fatty acid chain, and are thought to serve as “pacemaker” enzymes for lipid desaturation pathways [24,25]. It is thus paradoxical that the longest-lived mutant strains had the highest apparent Δ9 desaturase activity (Figure 2H) and the highest amounts of both 16:1Δ9 and 18:1Δ9 (Table S3), and yet the lowest PUFA levels (Figure 2B). This implies that additional regulation must occur downstream of the Δ9 desaturases, to favor or impede further desaturation. The declines with increasing lifespan, seen here for Δ6 and ω-3 desaturases (Table 2), support that interpretation.

Because the longest-lived strain, SR808, differs from the other strains with respect to fertility (age-1(mg44)-F2 homozygotes being completely infertile [7]), we were concerned that the longevity trends observed in FA content may have been confounded by variation in quantities of egg lipid stores. We therefore ran two post-hoc comparisons by ANOVA. In the first, three control strains were analyzed for FA content, comparing them at day 3 (egg-laying adults) vs. day 6 of adulthood, by which time they were post-gravid, pre-senescent adults, devoid of internal eggs or embryos by microscopic examination. None of the 20 FA measures showed a significant effect of eggs that could have contributed to the observed trends with longevity. The effect of infertility would often have opposed the shift seen in the long-lived SR808 [age-1(mg44)] strain, but in those instances where it could have contributed to that strain's exceptional FA content, the estimated effect size ranged from 3-26% of SR808's shift (data not shown). The second comparison simply re-evaluated the association between lifespan and FA content, considering only postgravid samples from four strains plus nongravid age-1(mg44)-F2 samples (SR808). The resulting correlations, summarized in the two right-hand columns of Table 3, closely mirrored the overall trends with lifespan observed previously (columns 2 and 3, taken from Table S3). Although the nongravid analyses carried less statistical significance, as expected in view of their much smaller sample size, the correlation coefficients for these 12 parameters were either similar to, or greater than, the corresponding coefficients from the full data set. The observed longevity trends thus cannot be attributed to strain differences in their egg or embryo content.

Transcriptional regulation of lipid biosynthesis

Transcript steady-state levels were assessed for a panel of genes known to be involved in nematode biosynthesis of lipids. Independent biological samples (N = 3 - 8) were prepared from the four longest-lived groups: F1 and F2 homozygotes for the age-1(mg44) mutation (median lifespans 2.5- and 10-fold longer than wild-type, respectively); age-1(hx546) (1.8-fold); and daf-2(e1370) (2.2-fold). First-generation homozygotes for mg44 receive oocyte contributions from their heterozygous parent (termed “maternal rescue”), blunting the more extreme traits associated with their F2 progeny. Additional groups tested include N2DRM adults (wild-type, lifespan defined as 1.0), N2DRM dauer larvae (developmentally arrested at an alternative larval stage 3), and double mutants of each age-1 allele with daf-16 (not shown), which largely or entirely revert longevity and other traits of those mutations [7,36]. Shifts in transcript levels, such as those shown in Figure 3, were highly reproducible; similar results were obtained in three replicate experiments, for all changes that were either significant or suggestive (P<0.1) in individual experiments.

The longest-lived group, age-1(mg44) F2 adults, showed downregulation for the elongase genes elo-1, elo-2, and elo-5, by ~4-fold [P<10⁻⁵], 2.5-fold [P<10⁻²], and 13-fold [P<10⁻⁴], respectively, each relative to N2DRM controls (Figure 3, A-C). However, far smaller decreases were seen in first generation (F1) age-1(mg44) homozygotes. A fourth elongase gene, elo-6, did not alter significantly in any strain (not shown). Transcript levels of elo-1,-2 and -5 were also low in N2 dauer larvae. Neither daf-2(e1370) nor age-1(hx546) adults showed significant decreases for any of the elo genes, and daf-2 mutants instead exhibited a modest (1.7-fold) increase in elo-2 expression. Thus, transcriptional attenuation of elongase expression appears to be a mechanism of reducing lipid chain length (ACL, Figure 2C) peculiar to age-1(mg44) mutants and dauer larvae.

Transcript levels were also assessed for seven nematode fatty-acid desaturase genes. Expression of the fat-1, fat-2 and fat-3 genes (respectively encoding ω-3, ∆12 and ∆6 desaturases), did not alter significantly in any of the strains examined (data not shown). Transcripts of fat-4, encoding a Δ5 desaturase that converts C20:3Δ8,11,14 (n-6) to arachidonic acid, C20:4Δ5,8,11,14(n-6), declined only in F2 age-1(mg44) worms, by 2- to 7-fold in three experiments (e.g., Figure 3 D), consistent with the decline in PUFAs observed in longer-lived strains (e.g. Figure 2B). In contrast, two genes encoding Δ9 desaturases, fat-6 and fat-7, both increased similarly in first- and second-generation age-1(mg44) homozygotes: two-fold for fat-6 and 1.6- to 1.8-fold for fat-7 (Figure 3 E,F). Lesser shifts in the same direction were also seen for fat-6 in age-1(hx546)and daf-2(e1370) mutants, and for fat-7 only in daf-2(e1370) adults. In marked contrast to our observations for elongase genes, the desaturase gene shifts observed in age-1(mg44) F2 adults either did not occur in dauer larvae (D, fat-4), or levels shifted in the opposite direction (E, fat-6, and F, fat-7). These expression changes seen in longer-lived strains are presumed to underlie their striking increases in Δ9 MUFAs (Table 2; Figure 2, B and H).

Functional consequences of altering lipid-biosynthesis gene expression

Because all three elongase genes were expressed at reduced levels in the longest-lived (IIS-mutant) worms, while desaturase gene transcript levels varied less systematically, we tested the functional importance of such shifts in expression, for nematode survival of oxidative stress, a rather robust surrogate for longevity effects of IIS disruption [7,37,38]. Wild-type (N2DRM) worms were fed for 3 days, starting at the L4/adult molt, on bacteria expressing gene-specific siRNA constructs [39], and were then assessed for their duration of survival in medium containing a toxic level of hydrogen peroxide. Typical results from two independent experiments are shown in Figure 4 A & B.

Knock-down of elo-1 or elo-2 consistently and significantly extended survival of this oxidative stress (each P<0.01), whereas RNAi againstelo-5 (encoding an elongase specific for monomethyl branched-chain fatty acids) was no more effective than empty vector (Figure 4). Knock-down of fat-4, encoding a Δ5 desaturase, increased peroxide survival time by at least 60% in multiple experiments (each P<10⁻⁶), while depletion of fat-6 transcripts extended survival by 24% (each P<10⁻⁵). With the exception of the latter result, the effects of RNAi were entirely consistent with expectations based on RNA transcript abundance shifts in the longest-lived strain, age-1(mg44).

Effects of RNAi knockdown on longevity (Figure 4, C & D), largely paralleling conferral of peroxide resistance, agreed even more closely with predictions based on transcript changes in age-1(mg44) (see Table 4). Knockdown of either elo-1 or elo-2 elicited moderate and significant extensions of C. elegans lifespan (each P<0.04). RNAi to elo-5 is not expected to affect longevity, since ELO-5 extends only monomethyl branched-chain fatty acids involved in developmental signaling [27,28]. Combined knock-down of both elongases (Figure 4D) was more effective than depletion of either elongase gene alone, but less than their sum; this could reflect the known redundancy between their functions [40], or simply the dilution of each bacterial strain. The greatest life extension (25%, P<10⁻⁴) followed depletion of fat-4 transcripts, which also produced the greatest peroxide resistance (60%). In contrast, knockdown of fat-6 or fat-7, both upregulated in long-lived mutants, effected small and insignificant reductions in longevity (Table 4). Life-span changes presented here agree remarkably well with previous data on survival of worms exposed to RNAi's targeting elo-2, fat-3, fat-4, fat-6 and fat-7 [10,12].

Hypomorphic mutations to genes implicated in the present studies have been demonstrated previously to alter lipid profiles in the directions we found to be associated with lower transcript levels [24-26,41]. For example, mutation to elo-1 (which shows substantial functional redundancy with elo-2 [40]) has markedly reduced long-chain FAs and hence PUFAs [26]. A mutation in fat-4 (encoding the Δ5 desaturase) effectively eliminates specific PUFAs such as 20:5(n-3) and 20:4(n-6) [26]; whereas individual mutations in fat-5, -6 or-7 (encoding Δ9 desaturases) lead to significant reductions in either 16:1 or 18:1 Δ9 fatty acids [41]. A schematic of lipid biosynthetic pathways (Figure 5), illustrating alterations to FA abundance and transcript levels in longevity mutants, is thus congruent with the corresponding diagrams based on data from lipogenesis mutants [24-26,41].

Another way to perturb the fatty-acid composition of worms is through diet. We fed worms on the standard monoxenic bacterial diet, either unmodified or supplemented with specific fatty acids. As illustrated in Figure 6, the longevity of wild-type Bristol-N2 [DRM] adult worms was significantly reduced (P<0.01) by addition of 40 μM eicosapentaenoic acid [EPA, C20:5(ω-3)], the longest-chain PUFA we observed in this nematode, relative to worms given an isocaloric supplement of palmitic acid (C16:0), an unsaturated, shorter-chain fatty acid. Lifespan was also reduced relative to worms not given any supplement. These results are entirely consistent with the trends observed among strains, and the life extensions produced by interference with expression of genes involved in PUFA synthesis.

Caenorhabditis elegans strains

Reputedly long-lived strains examined in this study were CB928, DA820, TJ3014, CB1370, DR26, MQ514, and GR1168 (provided by the Caenorhabditis Genetics Center [CGC], Minneapolis MN). These strains were made congenic (near-isogenic) by six consecutive out-crossings of each against a single male-enriched stock of Bristol-N2 wild-type worms (CGC “N2 male” stock; the Riddle-laboratory designation is N2DRM). The daf-16(m26) mutation largely or completely reverses life extension by age-1(mg44) [7], similar to its reversion of traits arising from daf-2 and other age-1 mutations [5,6,59]. Double mutants combining age-1 or daf-2 with daf-16(m26) were created by crossing several of the above outcrossed strains to DR26 [daf-16(m26)], similarly outcrossed, and identified among progeny of those crosses by genotyping via polymerase chain reaction (PCR) after propagation. The age-1(mg44)mutation is maintained as a mixed heterozygous stock in its original strain (GR1168) and in our N2DRM-outcrossed strain (SR808), with a wild-type age-1 allele provided on the mnC1 balancer chromosome to suppress homologous recombination [60]. From the SR808 stock, age-1 homozygotes were segregated as needed, and their “F2” adult progeny (which are slow-developing and infertile) were harvested after attaining approximately normal adult size (day 8 - 11 after the L4/adult molt). The congenic strains constructed, and their relative longevities, are summarized in Table 1.

Nematode maintenance and harvesting

Worms were cultured at 20°C on “NG” plates (supplemented with peptone to 2% w/v) seeded with lawns of E. coli var. OP50 [61]. At harvest, worms were held without food for 30 min at 20° C, which we found sufficient for digestion of gut bacteria and degradation of bacterial DNA to below detectable limits [62]. Worms were harvested as young adults (unless noted otherwise) to avoid confounding of lifespan effects with senescence effects. Worm pellets were promptly frozen on dry ice and stored at −80°C. Replicate assays of C. elegans strains used independent worm expansions and cultures to assess inter-cohort variance, which could also contribute to inter-strain variation.

Determination of lifespan

Nematodes, grown as described above, were rinsed from plates with S buffer (0.1 M NaCl, 0.05 M potassium phosphate, pH 6.0) [63-65]. Adults, enriched and recovered by settling, were resuspended in alkaline hypochlorite (0.5-N NaOH, 1.05% hypochlorite) and lysed during 5 min incubation at 20° C. The recovered eggs were rinsed in S buffer and transferred to fresh agar plates seeded with E. coli (strain OP50). Survival cultures were begun on 60-mm agar plates, one day after the L4/adult molt, transferring 50 adults (25 per plate) to 60-mm dishes containing NGM agar and a central lawn of OP50 E. coli. Cultures were maintained at 20°C and live worms counted daily upon transfer to fresh dishes. Worms that failed to move, either spontaneously or in response to touch, were scored as dead. Worms lost (stranded on Petri-dish walls or within the agar), or killed by internal hatching of progeny, were censored at the midpoint of the time span between observations, whereas worms inadvertently killed were censored at the time of the event. Adult lifespans of C. elegans strains (Table 1) are calculated as mean or median days of survival after the L4-adult molt, normalized to the lifespan of the N2DRM strain, identically assessed as a simultaneous control. This departs from the more common practice of expressing lifespan as days of survival after eclosion of larvae from eggs, but avoids confounding the effects of mutations on developmental timing and longevity [66].

Reagents

Chemicals used in GC sample preparation (1-docosanol, chloroform, pyridine, methanol, and hydrochloric acid) were purchased from Regis Technologies Inc. (Morton Grove, IL). Chemicals for buffers, and a Supelco 37-FAME set of GC-MS standards, were obtained from Sigma-Aldrich (St. Louis, MO).

GC-MS sample preparation

Frozen pellets were pulverized to fine powders under a dry nitrogen gas atmosphere in mortars chilled on dry ice. Frozen powder samples were lyophilized at −60° C under reduced pressure (<100 mTor) for 27 h followed by 1 h at −20° C and then 30 min at 15° C. Lyophilized samples weighing 4.00-4.05 mg were soaked in 1 ml of chloroform overnight at −80° C. A solution of 1-docosanol in 15 μL of chloroform was added at 15 μg per sample as an internal standard to monitor the efficiency of recovery through all processing and analytic steps. Samples were brought to 0° C for 1 h and sonicated repeatedly in an ice-water bath for a total of 10 min. Two successive liquid-liquid extractions were performed after additions of de-gassed, phosphate-buffered water (0.5-mM, pH 7.4). The organic phases were dried under a nitrogen stream at room temperature, whereas aqueous phases (not analyzed here) were deproteinized, dried and stored. Organic extract residues were dissolved in a solution of 1.25-M HCl in methanol. Incubation for 4 h at 50° C resulted in methyl-transesterification of esters, thus generating fatty acid methyl esters (FAMEs) from lipids. After desolvation, the residues were reacted with at least a 2:1 excess of the trimethylsilyl donor MSTFA (including 1% TMCS as catalyst; Pierce, Rockford IL), dissolved in pyridine. After incubation for 1 h at 50° C, reactions were chilled and subjected to GC-MS analysis within 24 h.

GC-MS analysis

GC-MS was performed on an Agilent 5890 II+ gas chromatograph coupled to an Agilent 5972 mass spectrometer. Each sample was introduced into a DB-5MS capillary column (J&W Scientific), 60-m long × 0.25-mm ID with 0.25-μm film thickness, integrated with a 5-m guard column. Automatic injections of 0.5-μl samples were made, without splitting, into the GC inlet set to 280°C. The thermal program began at 80° C for 2 minutes, then increased linearly to 210° C at a ramping rate of 6° C/min, further increased to 310° C at the slower ramping rate of 2.5°C/min, and held this temperature for 8 min. Helium was used as the carrier gas with a constant flow rate of 1 ml/min under electronic pressure control. Positive 70-mV electron-impact quadrupole ion scans were acquired every 0.75 sec, from 650 to 50 m/z. Each run was initiated with an 11.5-minute solvent delay before ionization commenced, to bypass the pyridine peak. The transfer-line temperature was set to 280°C and the indicated detector temperature was 172° C. Detection parameters were tuned using the maximum-sensitivity autotune mode. A blank sample (pyridine only), a blank supplemented with four long-chain alkane retention-index (RI) standards, and a sample containing the same RI standards and a derivatized recovery standard used in quantitation, were run at the beginning, the middle and the end of each sequence of runs, for quality-control purposes.

Quantitation of transcript abundances by real-time polymerase chain reaction

Gene expression was assessed by real-time quantitative polymerase chain reaction (RT-qPCR) to amplify cDNAs reverse-transcribed from total RNA, as described [37]. Total RNA was purified (RNeasy, Qiagen) from each group, copied to DNA by reverse transcriptase (SuperScript III, Invitrogen), and the products quantified by RT-qPCR on an Opticon2 thermal cycler (MJ Research), using SYBR Green stain (Roche).

Data analysis and statistics

The first step in quantitation of fatty acids is the alignment of peaks across all GC-MS runs, utilizing both chromatographic information (retention times) and mass-spectral data (m/z components) to establish the chemical identity of peaks being compared. To correct retention time variation between runs, we applied an iterative block-shift alignment method [67] to produce precisely aligned data with minimal peak distortion (thus preserving quantitative information), resulting in 77 three-way arrays (m/z × time × samples). Each array was independently decomposed into three matrices (relative concentrations, elution profile, and mass spectrum), by PARAFAC analysis [68] using a MATLAB script (Version 3.10) from the N-way Toolbox [69]. Compounds corresponding to the deconvoluted mass spectra were identified by matching to the NIST-08 Mass Spectral Library, with minimal criteria of >80% for both AMDIS (Version 2.66) peak purity and the NIST matching index. All compounds identified as fatty acid methyl esters (FAMEs) agreed closely, in retention times and spectra, to corresponding analytes in a reference set comprising 37 FAMEs.

Fatty acid quantities obtained by GC-MS were normalized in two ways: (1.) to an internal standard (1-docosanol, 15 μg added to each sample prior to sonication and lipid extraction), or (2.) to the sum of all lipid peaks identified. Expression of peak areas as ratios to the area of the internal standard peak, observed on the same chromatogram, adjusts for run-to-run variation in the efficiencies of extraction, reaction with derivatizing agents, and detection. At the same time, this normalization preserves strain-to-strain differences in the absolute amounts of individual or summed FA components, since dry mass was accurately measured for lyophilized samples, and loads were thus standardized prior to GC-MS analysis. In the second normalization procedure, relative FA amounts were obtained by dividing each peak area by the sum of areas for all FA peaks appearing on the same chromatogram, which thus serves as an unbiased, endogenous internal standard (Table 2).

Associations between lifespan or gravidity and the amounts of FAs were analyzed for data organized by strain, using parametric and nonparametric analysis of correlation (Table 2) or one-way ANOVA (Table 3). In pairwise post-hoc comparisons, Dunnett's procedure was used to compare other strains to N2DRM as the normal-lifespan reference group. For correlation analysis of trends in young-adult worms as functions of lifespan, strains were grouped into bins if their longevities could not be unambiguously distinguished. Lifespans relative to wild-type (N2DRM) are summarized in Tables 1 and S1.

Correlations between the various lipid parameters and longevity were evaluated in two ways: (1.) as Pearson correlation coefficients (R), which describe the linear relationship between parameters; or (2.) as Spearman (Rank-Order) correlation coefficients (R_S), which are nonparametric and hence independent of the distributions of variables and of any assumptions regarding the relationships among variables. In either case, each correlation coefficient R was calculated from N data points, equal to the number of independent populations harvested and analyzed in separate GC-MS runs. The corresponding P value were ascertained by t-transformation,t = |R| ∙ [(N - 2) / (1 - R2)]^½, treating the result as coming from a t distribution with N-2 degrees of freedom [70].

To test for a possible confounding influence of the infertility phenotypes of strain SR808, age-1(mg44) (reduced sperm and absence of mature oocytes or retained embryos) on its FA profile, two post-hoc analyses were made. (a.) Three control strains were assessed both as fertile young adults (days 2-3 of adult-hood), and as post-gravid, pre-senescent adults (approximately day 6 of adulthood) with no eggs or mature oocytes observed by microscopic examination; the relative contribution of fertility to FA content, estimated by ANOVA, was not statistically significant, ranging from −33 to +26% of the difference between SR808 the mean of all control samples. (b.) The trend with longevity was reassessed by ANOVA for the subset of nongravid (“eggless”) samples, i.e. strain SR808 and post-gravid (day-6) samples which comprised the above controls plus SR807, age-1(hx546). Equal weight was assigned to each lifespan grouping in this analysis.

In considering the significance of trends in FA levels with longevity, strict Bonferroni adjustment for multiple endpoints shown in Table 2 would require P < 0.05/21, or 0.0024. By this stringent criterion, 13 of the 21 linear correlations would be considered significant, while 7 of the nonparametric correlations qualify. Such a strict threshold is appropriate for measures that are independent of one another, but is overly conservative for lipid biosynthesis wherein successive enzymatic steps of elongation and desaturation produce a series of FA-chain types. The abundance of each FA type depends strongly on the level of its precursors; e.g., C18:0 depends on C16:0 levels, and C18:1(n-9) depends on the level of C18:0, and thus also of C16:0 (see Figure 5). In Table 2, numerous ratios are derived as functions of a subset of 19 measured FA levels, and hence are entirely dependent on those data. Uncorrected P values are given here, to allow the reader to choose between strict Bonferroni adjustment, and a less conservative significance threshold such as α = 0.01, commonly applied in similar situations to avoid overinflating type-II errors.