AMPK and FoxO are necessary for the lifespan extension due to peptone dilution

We next asked how general the role of AMPK and FoxO was in longevity induced by different methods of restricting nutrients. Dilution of peptone in bacterial plates extends worm lifespan because it is considered to reduce bacterial growth on the plates (Hosono et al., 1989). We find that the dilution of peptone in bacterial plates extends lifespan in an AMPK/aak-2 and FoxO/daf-16 dependent manner (P < 0.0001 by two-way anova) (Fig. 2A; Table S2). This result indicates that AMPK and FoxO are also necessary for lifespan extension induced by another method of restricting nutrients.

Open in a separate window

Fig. 2

AMPK and FoxO mediate longevity induced by some but not all DR methods. (A) Longevity induced by dilution of peptone (DP) is dependent on AMPK/aak-2 and FoxO/daf-16. Dilution of peptone in the plates extends WT (N2) worm lifespan (25.4%, P < 0.0001) but does not extend aak-2(ok524) mutant lifespan (–1.0%, P = 0.6371) or daf-16(mu86) mutant lifespan (5.7%, P = 0.1402). Two-way anova revealed that the lifespan extension of WT (N2) worms across a peptone gradient was significantly different from that of aak-2(ok524) mutant worms (P < 0.0001) or daf-16(mu86) mutant worms (P < 0.0001). Mean, standard errors, and statistical analysis for two independent experiments performed in triplicate are presented in Table S2. (B) AMPK/aak-2 and FoxO/daf-16 are not completely necessary for bDR lifespan extension. The average and SEM of three independent experiments indicates that bDR increases WT (N2), aak-2(ok524), and daf-16(mu86) mutant lifespan but appears to increase WT (N2) worm lifespan to a greater extent than aak-2(ok524) or daf-16(mu86) mutant worm lifespans. For each individual experiments, see Figure S2 and Table S3. (C) As previously reported (Curtis et al., 2006), the eat-2(ad1116) mutation extends WT (N2) worm lifespan (19.8%, P < 0.0001) and aak-2(ok524) mutant lifespan (19.4%, P < 0.0001). Mean, standard errors, and statistical analysis for two independent experiments performed in triplicate are presented in Table S4A. (D) As previously reported (Lakowski & Hekimi, 1998), the eat-2(ad1116) mutation extends WT (N2) worm lifespan (12.9%, P < 0.0001) and daf-16(mu86) mutant lifespan (32.3%P < 0.0001). Mean, standard errors, and statistical analysis for two independent experiments performed in triplicate are presented in Table S4B.

AMPK and FoxO are not completely necessary for bDR and eat-2 induced longevity

FoxO/daf-16 was shown not to be entirely necessary for longevity induced by the dilution of bacteria in liquid cultures (bDR) (Houthoofd et al., 2003; Panowski et al., 2007). However, the importance of AMPK in bDR-induced lifespan extension has never been examined. We tested whether AMPK and FoxO were required for lifespan extension in response to a gradient of bacteria concentrations in liquid culture. Combining the results of three independent experiments (Fig. 2B), we found that bDR significantly extended the lifespan of WT (N2), aak-2(ok524), and daf-16(mu86) mutant worms (P < 0.0001), indicating that AMPK and FoxO are dispensable for the entire lifespan extension induced by bDR. In two of these experiments, bDR extended WT (N2) lifespan to a larger extent than aak-2 and daf-16 mutant worm lifespan (P < 0.0001 by two-way anova) (Figure S2A and S2B; Table S3). In one experiment however, bDR extended the lifespan of WT (N2) worms to a similar extent as that of aak-2 and daf-16 mutant worms (P = 0.1528 and 0.6643 respectively by two-way anova) (Figure S2C; Table S3). Note that in one of the assays (Figure S2A), starvation was not reached, which may alter the interpretation of the experiment. However, when this assay was omitted for statistical analysis, we still found that bDR extended WT (N2) lifespan to a larger extent than aak-2 and daf-16 mutant worm lifespan (P < 0.001 by two-way anova). Although the parameters for these variations are unknown, these results suggest that AMPK and FoxO are dispensable for bDR to extend lifespan, but that they play a modulatory role in the extension of lifespan by this regimen.

Finally, as previously reported, we confirmed that AMPK/aak-2 and FoxO/daf-16 were completely dispensable for the lifespan extension of eat-2(ad1116) mutant worms, a genetic way to mimic DR (Lakowski & Hekimi, 1998; Curtis et al., 2006) (Fig. 2C,D; Table S4). Together, these results indicate that AMPK and FoxO are required for longevity induced by some (sDR and DP) but not all (bDR and eat-2) DR methods.

Resveratrol extends lifespan in an AMPK-dependent, but FoxO-independent, manner

Resveratrol extends lifespan in many species and has been suggested to act as a DR mimetic (Sinclair, 2005). Resveratrol has recently been found to activate AMPK in cultured cells and in mice (Baur et al., 2006; Zang et al., 2006; Dasgupta & Milbrandt, 2007; Hwang et al., 2007). We thus asked whether AMPK was necessary for increased longevity induced by resveratrol in C. elegans. Resveratrol induced a modest, but statistically significant, increase in WT (N2) worm lifespan. In contrast, resveratrol did not increase the lifespan of aak-2(ok524) mutant worms, suggesting that AMPK is necessary for the beneficial effects of resveratrol on lifespan (Fig. 3A; Table S5), as was proposed by (Bass et al., 2007). Similar to what was previously reported (Viswanathan et al., 2005), we found that resveratrol still extended the lifespan of daf-16(mu86) mutant worms (Fig. 3B; Table S5), indicating that resveratrol extends lifespan by eliciting an AMPK-dependent, FoxO-independent pathway. Therefore, activation of AMPK is not always coupled to that of FoxO, even though AMPK's ability to extend lifespan is dependent on the presence of FoxO (Greer et al., 2007). This result suggests that AMPK also regulates other substrates to extend lifespan, which is consistent with published findings (Apfeld et al., 2004; Narbonne & Roy, 2006). Therefore, these results suggest that different ways of manipulating nutrients and chemical compounds extend lifespan via independent and non-linear genetic pathways.

Open in a separate window

Fig. 3

Resveratrol extends lifespan in an AMPK-dependent, but FoxO-independent, manner. (A) Resveratrol (100 µm) extended WT (N2) worm lifespan (14.2%, P = 0.0005), but did not significantly extend aak-2(ok524) mutant worm lifespan (2.2%P = 0.5485). Mean, standard errors, and statistical analysis for three independent experiments performed in triplicate are presented in Table S5. (B) Resveratrol (100 µm) extended both WT (N2) worm lifespan (14.6%P < 0.0001) and daf-16(mu86) mutant worm lifespan (13.7%P < 0.0001). Res, resveratrol. Mean, standard errors, and statistical analysis for one experiment performed in triplicate are presented in Table S5.

sir-2.1, pha-4, skn-1, and hsf-1 are not necessary for sDR-induced lifespan extension

Having shown that FoxO and AMPK were important for some, but not all, DR methods, we next tested if conversely, genes that have been implicated in lifespan extension by resveratrol or by other DR regimens were necessary for sDR to extend lifespan.

We first asked if sir-2.1, which encodes a Sirtuin family protein deacetylase, was required for lifespan extension in response to sDR. sir-2.1 has been found to mediate lifespan extension by resveratrol in some studies (Wood et al., 2004; Viswanathan et al., 2005), but not others (Bass et al., 2007). We used a mutant strain carrying a deletion in the sir-2.1 gene, which is predicted to be a null mutant (sir-2.1(ok434)) (Wang & Tissenbaum, 2006). We found that sDR increased the lifespan of both WT and sir-2.1 mutant worms to the same extent (Fig. 4A; Table S6). There was no statistically significant difference between the effects of sDR on the longevity of WT worms vs. sir-2.1 mutant worms (P = 0.1240 by two-way anova). These results indicate that sir-2.1 is not necessary for lifespan extension by sDR.

Open in a separate window

Fig. 4

sir-2.1, FoxA/pha-4, skn-1 and hsf-1 are not entirely necessary for sDR to extend lifespan. (A) A serial dilution of bacteria on plates (5 × 10¹² to 5 × 10⁷ bacteria mL⁻¹) extends WT (N2) (26.1%, P < 0.0001) and sir-2.1(ok434) (16.6%, P < 0.0001) mutant worm lifespan but does not extend aak-2(ok524) (1.3%, P = 0.6330) mutant worm lifespan. Two-way anova revealed that the lifespan extension of WT (N2) worms across a bacterial gradient was significantly different from that of aak-2(ok524) mutant worms (P < 0.0001) but not statistically different from sir-2.1(ok434) mutant worms (P = 0.1240). Mean, standard errors, and statistical analysis for three independent experiments performed in triplicate are presented in Table S6. (B) A serial dilution of bacteria on plates extended smg-1(cc546ts) and smg-1(cc546ts); pha-4(zu225) mutant worms to a similar extend (P = 0.3724 by two-way anova). Note that this experiment was performed at 15 °C after worms reached adulthood. Mean, standard errors, and statistical analysis for two independent experiments performed in triplicate are presented in Table S7A. (C) A serial dilution of bacteria on plates extended WT (N2) (24.9%, P < 0.0001) worm lifespan, skn-1(zu135) (23.1%, P < 0.0001) mutant worm lifespan, and hsf-1(sy441) (33.5%, P < 0.0001) mutant worm lifespan but did not extend aak-2(rr48) (4.6%, P = 0.2566) mutant worm lifespan. Two-way anova revealed that the lifespan extension of WT (N2) worms across a bacterial gradient was significantly different from that of aak-2(ok524) mutant worms (P < 0.0001) but not statistically different from skn-1(zu135) mutant worm lifespan (P = 0.5567) or hsf-1(sy441) mutant worm lifespan (P = 0.2843). Mean, standard errors, and statistical analysis for two independent experiments performed in triplicate are presented in Table S8 and Table S9.

We then tested the importance of the Forkhead transcription factor FoxA/pha-4, a gene involved in eat-2 and bDR-induced longevity (Panowski et al., 2007), in lifespan extension induced by sDR. We used a temperature sensitive allele of FoxA/pha-4, smg-1(cc546ts); pha-4(zu225), and its control counterpart smg-1(cc546ts) (Gaudet & Mango, 2002). When shifted to the restrictive temperature smg-1(cc546ts); pha-4(zu225) display very little PHA-4 protein and the phenotype of smg-1(cc546ts); pha-4(zu225) is similar to that of the pha-4(q490) null mutant (Kaltenbach et al., 2005; Kiefer et al., 2007). We found that sDR extended lifespan of both the control strain (smg-1(cc546ts)) and the inducible pha-4 mutant strain (smg-1(cc546ts); pha-4(zu225)) (Fig. 4B; Table S7A) (P = 0.3724 by two-way anova). These results indicate that FoxA/pha-4 is not necessary for sDR induced lifespan extension. Consistent with this observation, we found that sDR still increased the lifespan of worms in which FoxA/pha-4 was knocked-down by RNAi, but did not extend the lifespan of worms in which FoxO/daf-16 was knocked-down (Figure S3; Table S7B). These findings suggest that FoxO, but not FoxA, is required for sDR-induced longevity.

We also examined if skn-1, a gene encoding a transcription factor necessary for lDR to extend lifespan (Bishop & Guarente, 2007), was required for longevity in response to sDR. We used a loss of function mutant strain of skn-1, skn-1(zu135), which displays a premature stop codon in all three isoforms of SKN-1 (a, b, and c) (Bishop & Guarente, 2007; Tullet et al., 2008). We showed that sDR extended lifespan of WT worms and skn-1(zu135) mutant worms to a similar extent (P = 0.5567 by two-way anova). Although it is formally possible that some SKN-1 activity remains in the skn-1(zu135) mutant strain, these findings nevertheless suggest that skn-1 is not necessary for sDR-induced longevity (Fig. 4C; Table S8).

Finally, we asked if hsf-1, which encodes a heat-shock responsive transcription factor involved in longevity in response to DD (Steinkraus et al., 2008), was necessary for sDR-induced lifespan extension. We used a mutant strain of hsf-1 (hsf-1(sy441)) that contains a premature stop codon that eliminates the transactivation domain of HSF-1 and is likely to be a null mutant (Hajdu-Cronin et al., 2004). We found that sDR still extended the lifespan in hsf-1(sy441) mutant worms similarly to WT worms (P = 0.2843 by two-way anova), indicating that hsf-1 is not necessary for sDR-induced longevity (Fig. 4C; Table S9).

Together, these data indicate that four genes (sir-2.1, pha-4, skn-1, and hsf-1) that have been previously implicated in longevity in response to a variety of DR methods and DR mimetics do not mediate lifespan extension by sDR. These findings further corroborate the observation that different DR regimens evoke independent pathways.

clk-1 is necessary for sDR-induced lifespan extension

The clk-1 gene encodes a demethoxyubiquinone hydroxylase that is necessary for the biosynthesis of ubiquinone, a component of the electron transport chain (Ewbank et al., 1997; Miyadera et al., 2001). clk-1 mutant worms live longer than their WT counterparts (Lakowski & Hekimi, 1996) and their long lifespan is not further extended by the eat-2 mutation (Lakowski & Hekimi, 1998), suggesting that clk-1 is necessary for eat-2 induced lifespan extension. Although the clk-1 allele, clk-1(e2519), is unlikely to be a null mutant (Lakowski & Hekimi, 1996), we tested if clk-1 was important for sDR-induced lifespan extension. We found that clk-1(e2519) mutant worms, similarly to aak-2(ok524) and aak-2(rr48) mutant worms, no longer responded to sDR (Fig. 5; Table S9). These results suggest that clk-1 is necessary for sDR-induced longevity and are compatible with the observation that clk-1 longevity like sDR-induced lifespan is dependent on daf-16. Although the interpretation of these results is difficult because of the lack of a null allele for clk-1 (Gems et al., 2002), clk-1 may mediate two independent methods of DR, eat-2 and sDR. Thus, in addition to the genes that are specific to DR methods, there may also exist overlapping mechanisms underlying DR-induced longevity.

Parameters that may affect the mechanisms of DR-induced longevity

Different DR regimens may evoke distinct genetic pathways because some nutrients may be more limiting than others depending on the DR method. For example, sDR might reduce carbohydrates more prominently than amino-acids, which would render it dependent on AMPK and FoxO, while other methods (eat-2, bDR) may reduce amino-acids more readily, which would evoke TOR, a well-known amino-acid responsive pathway (Avruch et al., 2006), and the FoxA/pha-4 transcription factor, which has recently been found to be downstream of TOR (Sheaffer et al., 2008). DR could also trigger different genetic mechanisms depending on the time at which DR is initiated or the tissues/cells by which it is sensed.

Alternatively, different DR regimens may evoke independent pathways because the way in which DR is initiated may trigger other non-DR signaling pathways that in turn have beneficial effects on lifespan. For example, diluting bacteria in sDR might also lead to the reduction in bacterial pathogenicity, which could in turn affect the insulin-FoxO pathway. However, if pathogenicity were the sole cause for premature death, concentrating bacteria above the ad libitum concentration would be detrimental, which we have found not to be the case. We have also shown that sDR is not simply due to a reduction in bacterial pathogenicity because sDR reduces worm fertility (data not shown) and because sDR performed with UV-killed bacteria also extends lifespan (Greer et al., 2007). The way sDR is achieved (by adding diluted amounts of E. coli every 2 days) may induce cycles of feeding and fasting that may cause this method to be AMPK and FoxO dependent whereas other DR methods (eat-2) maintain a more constant bacteria concentration throughout lifespan. sDR may also induce a mild oxidative stress response that could trigger the activation of the AMPK and FoxO pathways, as these pathways are known to transduce stress stimuli (Henderson & Johnson, 2001; Apfeld et al., 2004; Schulz et al., 2007; Lee et al., 2008). Although such stress may be part of DR itself (see the ‘hormesis theory’ of DR (Sinclair, 2005)), it is possible that the other DR methods do not require oxidative stress to extend lifespan.

Conversely, other methods of inducing DR could trigger non-DR signaling pathways. For example, the eat-2 mutation, which affects the acetylcholine receptor, may have other effects in addition to reducing pharyngeal pumping. Liquid methods of DR may also evoke other parameters (e.g. altered oxygen consumption (Honda et al., 1993) or forced exercise (Retzlaff et al., 1966) since worms actively swim in liquid) that may interact with reduction in nutrients to result in lifespan extension in a manner that is dependent on specific genes. Identifying the parameters that are embedded in or interact with DR methods will be important to gain complete insight into the mechanisms by which DR extends lifespan.

Mechanisms of DR in other species

An important question is whether the genetic mechanisms identified to regulate longevity in response to DR in C. elegans are conserved between species. The existence of different DR regimens that extend lifespan in yeast, flies, and mice raises the possibility that some of the mechanisms sensing the restriction of specific nutrients are conserved throughout evolution.

For example, in flies, a number of different methods of restricting diet extend lifespan (for a comprehensive review see (Piper & Partridge, 2007)). Two methods of DR in flies, dilution of yeast (Min et al., 2008) and dilution of both sugar and yeast (Giannakou et al., 2008) have been found to be independent of the Drosophila FoxO transcription factor, dFOXO, although dFOXO modifies the response to DR (Giannakou et al., 2008). While the role of AMPK in DR has not been tested yet, the histone deacetylases Rpd3 and Sir2, and the tumor suppressors Dmp53 and Tsc2, a member of the TOR pathway, have all been shown to be necessary for lifespan extension in different methods of DR in flies (Rogina et al., 2002; Kapahi et al., 2004; Rogina & Helfand, 2004; Bauer et al., 2005). The main method in which DR is initiated in flies, which reduces amino-acids, may primarily evoke the TOR and FoxA pathway rather than the AMPK or FoxO pathway (Kapahi et al., 2004).

In mice, reduction of the total amount of food which mice receive every day (Weindruch et al., 1986) or alternating days of feeding and fasting (EOD) both extend lifespan (Goodrick et al., 1990). The specific genetic components involved in DR in mice have not been as extensively analyzed yet, although the deletion of Sirt1, the mouse Sir2 ortholog, abrogates the beneficial effect of DR on behavioral activity (Chen et al., 2005) and mice expressing additionally copies of the Sirt1 gene have metabolic parameters similar to those induced by DR (Bordone et al., 2007). A reduction of the protein concentration (Goodrick, 1978) or of the amount of methionine in the diet also extend mouse lifespan (Orentreich et al., 1993), raising the possibility that the TOR pathway might be critical in longevity induced by these methods. While the importance of AMPK or FoxO in longevity induced by DR in mammals is still unknown, emerging evidence suggests that these pathways play some role in DR. First, EOD and a 40% restriction of food activate AMPK in the liver of rats (Pallottini et al., 2004). Short term DR (60% reduction of food for 5 days) also activated AMPK in the hippocampus of mice (Dagon et al., 2005). Second, the lifespan of mice that are deficient for the growth hormone receptor is not extended by a 30% reduction of food (Bonkowski et al., 2006). The deficiency in growth hormone receptor is thought to act via a reduction in IGF-1 and insulin levels, which would lead to FoxO activation. These observations raise the possibility that FoxO may mediate longevity in response to DR in mice.

Role of AMPK in longevity induced by ‘DR mimetics’

Chemical compounds that mimic some of the beneficial effects of DR include 2-deoxyglucose (2DG) (Roth et al., 2001), metformin (Dhahbi et al., 2005), and resveratrol (Howitz et al., 2003). Metformin decreases the tumor incidence and extends the lifespan of tumor prone HER-2/neu transgenic mice (Anisimov et al., 2005). Resveratrol extends the lifespan of yeast (Howitz et al., 2003), worms (Wood et al., 2004; Viswanathan et al., 2005; Gruber et al., 2007), flies (Wood et al., 2004), fish (Valenzano et al., 2006) and mice on a high fat diet (Baur et al., 2006), although resveratrol does not extend lifespan under all circumstances (Bass et al., 2007; Pearson et al., 2008). Intriguingly, 2DG, resveratrol, and metformin have all been shown to activate AMPK (Baur et al., 2006; Zang et al., 2006; Dasgupta & Milbrandt, 2007; Hardie, 2007; Hwang et al., 2007). Resveratrol has been proposed to act through sir-2.1 (Wood et al., 2004; Viswanathan et al., 2005), although this was not found in all studies (Bass et al., 2007), and AMPK/aak-2 (this study) to affect lifespan, while 2DG was shown to act through AMPK/aak-2, but not sir-2.1, in C. elegans (Schulz et al., 2007). Thus, AMPK may play a pivotal role in mediating the lifespan extension induced by DR mimetics. However, the ‘DR mimetics’ have less of an impact on lifespan than DR itself. This observation suggests that multiple pathways need to be activated concomitantly to achieve optimal effects on lifespan and healthspan through chemical treatments.

Lifespan assays

Worm lifespan assays were performed at 20 °C unless noted differently. Worm populations were synchronized by placing young adult worms on NGM plates seeded with the E. coli strain OP50–1 (unless otherwise noted) for 4–6 h and then removed. OP50–1 is a strain derived from OP50 that contains a streptomycine resistance gene (Kawli & Tan, 2008). The hatching day was counted as day 1 for all lifespan measurements. Worms were changed every other day to new plates to eliminate confounding progeny, and were marked as dead or alive. Worms were scored as dead if they did not respond to repeated prods with a platinum pick. Worms were censored if they crawled off the plate or died from vulval bursting. For each lifespan assay, 90 worms per condition were used in three plates (30 worms per plate). The data were plotted in StatView 5.0.1 using the Kaplan–Meier Survival curves and statistical significance was determined by Log-rank (Mantel-Cox) tests. Lifespan assays were repeated at least once unless otherwise mentioned. Representative Kaplan-Meier survival curves are shown in the figures. For lifespan assays performed on a gradient of bacterial concentrations mean and standard error values were taken from Kaplan–Meier Survival curves plotted in StatView 5.0.1 and plotted in Excel. Representative curves are presented unless noted otherwise. For smg-1 and smg-1; pha-4 lifespan assays, worms were grown at permissive temperature 24 °C until the first day of adulthood when they were switched to 15 °C which allows for the degradation of pha-4(zu225) (Gaudet & Mango, 2002).

DR assays

OP50–1 bacteria were serially diluted from 5 × 10¹² to 5 × 10⁴ bacteria mL⁻¹. Bacteria were resuspended in S Medium to inhibit bacterial growth. Adult worms were placed on these various concentrations of bacteria starting at day 7 of life (day 4 of adulthood). Worms placed on a gradient of bacterial concentrations ranging from 5 × 10⁴ to 5 × 10⁷ bacteria mL⁻¹ died 2 days after being placed on these extreme DR diets. For specific assays, sDR was considered 5 × 10⁸ bacteria mL⁻¹ and ad libitum was 5 × 10¹¹ bacteria mL⁻¹ (Greer et al., 2007).

Peptone dilution assays

Assays were performed as in (Hosono et al., 1989). Synchronized populations of worms were obtained by placing adult worms on plates with decreasing concentrations of peptone (from 2.5 to .0025 g L⁻¹ with 150 µL of 5 × 10¹² bacteria mL⁻¹ seeded on each plate) and removing the worms after 4–6 h. Worms were switched every other day to fresh plates and were scored as alive as described above.

Liquid DR assays

Assays were performed as in (Panowski et al., 2007). Briefly worms were grown until day 1 of adulthood on NGM plates seeded with 150 µL of 5 × 10¹¹ bacteria mL⁻¹. They were then transferred for 1 day to NGM plates with FUdR (100 mg L⁻¹) seeded with 150 µL of 5 × 10¹¹ bacteria mL⁻¹. Approximately 20 worms were then placed in each well of a 12-well plate with 1 mL of S-basal supplemented with cholesterol (5 mg L⁻¹), Ampicillin (50 µg L⁻¹), Kanamycin (10 µg L⁻¹), Tetracycline (1 µg L⁻¹) and FUdR (100 mg L⁻¹) containing OP50–1 bacteria at various concentrations (from 3.33 × 10⁹ to 5 × 10¹¹ bacteria mL⁻¹). Approximately 90 total worms were used for each condition (4 wells containing approximately 22 worms per well). Plates were gently shaken at 20 °C and worms were scored as alive as described before. Live worms were switched to fresh liquid medium containing the appropriate dilution of bacteria every third day.

We thank Andy Dillin, Cynthia Kenyon, Susan Mango, and Man-Wah Tan for generously sharing protocols, reagents, and for helpful discussion. We thank Keith Blackwell for helpful suggestions. We thank Alex McMillan, Art Owen, and Dario Valenzano for help with statistical analysis. We thank Max Banko and Hirokazu Fukui for experimental help. We thank members of the Brunet laboratory as well as Paul Greer, Stuart Kim, Judy Lieberman, and Adolfo Sanchez-Blanco for critically reading the manuscript. We are grateful to Theresa Stiernagle and the Caenorhabditis Genetics Center funded by the National Center for Research Resources of the NIH for providing worm strains. This research was supported by an NIH grant (NIA AG026648–01), an American Institute for Cancer Research grant, and by a gift from the Glenn Foundation to A.B. E.L.G. was supported by a PHS grant (CA 09302) from the NCI and by an NSF graduate fellowship.