Dietary restriction (DR), also referred to as calorie restriction, is an intervention that extends life span and delays the onset of age-related phenotypes in nearly all eukaryotic organisms in which it has been tested 1. Simplistically, it is defined as a significant reduction in dietary intake in the absence of malnutrition. Many different approaches can be used to achieve DR. In mice and rats for example, life span extension is observed by either reducing the amount of food consumed daily (compared to an ad libitum control group) or by imposing an intermittent fasting regimen 23.

In addition to simply reducing the amount of food intake, the effects of altering dietary composition has also been examined in different organisms. Methionine-restricted mice 4 and rats 5 have an extended life span, suggesting that the nutritional composition of the diet can influence longevity in mammals. In fruit flies, reducing either the yeast extract or sugar composition of the food supply extends life span, although yeast extract appears to have the greatest impact on longevity 6. Restriction of amino acids also increases life span in flies 7. In yeast, reducing either the amount of glucose or amino acids in the growth media increases replicative life span 8910.

Multiple methods for DR have been used in C. elegans, as in other model organisms 11. One commonly used method in C. elegans is a genetic model, mutation of eat-2. The eat mutants were originally identified in a screen for defects in feeding behavior – they have reduced food consumption due to pharyngeal pumping defects 1213. Several alleles of eat-2 (and other eat mutants) increase life span, with longevity generally correlating with the degree to which pumping rate is decreased 14. Life span extension from mutation of eat-2 is independent of the FOXO-family transcription factor DAF-16 and is additive with a longevity-enhancing allele of the insulin/IGF-1-like receptor daf-2 14. This has led to the generally accepted model that DR acts to modulate longevity in a genetic pathway distinct from insulin/IGF-1-like signaling (IIS) 1516. As with any genetic model of DR, however, eat mutants are not suitable for certain studies. The relative activity associated with different eat-2 alleles has not been completely characterized, so it is unclear whether life span extension from DR is maximized in studies using these alleles. In addition, eat mutants are food restricted from hatching, and DR during development may have secondary effects on adult physiology that are not fully understood.

DR by reducing food availability also increases life span in C. elegans. The standard method for culturing C. elegans in the laboratory is to maintain the nematodes on the surface of nematode growth medium (NGM) nutrient agar with E. coli OP50 as the food source. DR on NGM nutrient agar has been achieved by reducing the amount of peptone in the media so as to limit bacterial growth 17 or by reducing the amount of live bacteria present on the surface of the NGM nutrient agar 18. Two methods of DR have also been described using non-standard, liquid-based growth conditions: bacterial dilution in S basal medium and axenic growth 1920212223. Both liquid-based methods increase life span relative to animals fed a diet of E. coli OP50 in S basal, and axenic growth behaves similarly to mutation of eat-2 in epistasis experiments with IIS 24. Nonetheless, neither method is widely used for aging studies. This may be due to the potential differences in growth under standard conditions on an agar surface versus growth in liquid culture. Axenic growth is reported to cause delayed development and poor growth 2225, and feeding animals E. coli in S basal shortens life span relative to growth on NGM agar 19. In addition, DR during development has been shown to be sub-optimal, since animals switched to DR just before adulthood are longer-lived than those maintained on DR from hatching 26.

Recently, two groups independently reported an novel reduced bacterial feeding DR protocol carried out under standard conditions 2728. By measuring adult life span as a function of E. coli food concentration, both studies determined that complete removal of bacterial food early in adulthood (BD, bacterial food deprivation; also referred to as dietary restriction through food deprivation or dietary deprivation) optimally increases median and maximum life span. BD differs from previous studies describing DR by axenic growth in the culture method (solid versus liquid media) and the time of initiation (reproductively mature adults versus hatchlings)27. Although bacterial food is completely absent during adulthood in the BD regimen, BD-treated animals do not suffer from malnutrition (as evidenced by their increased life span and stress resistance 2728), thus meeting the commonly accepted definition for DR. Life span extension from BD has been recently validated in multiple wild-derived C. elegans strains, as well as in a second closely related nematode species C. remanei 29.

Like mutation of eat-2 or axenic growth 1424, BD appears to modulate longevity by a mechanism distinct from reduced IIS 2728. While BD results in a more robust life span increase than mutation of eat-2, eat-2 mutants subjected to BD are not longer-lived than wild type animals on BD 2728. Taken together, these observations indicate that BD and mutation of eat-2 are likely to increase life span by similar or overlapping mechanisms. Thus, BD represents a simple plate-based method for DR that maximizes longevity from food restriction under standard nematode growth conditions and does not require additional genetic manipulation.

In order to better characterize the mechanism by which BD slows aging in C. elegans, we determined the effect of BD on survival as a function of the age at which BD is initiated. Previously, our group and an independent study 2728 reported that BD increases life span in C. elegans when initiated at the last stage of larval development (L4) through day 11 of adulthood. Here, we expand this analysis to more advanced ages, after which mortality has become significant in the population. Similar to prior observations in Drosophila 30 and in mice 31, we find that BD enhances survival even when imposed late in life. Transient exposure to BD confers a life span benefit that correlates with stress resistance, as measured by thermotolerance. We further demonstrate that age-independent life span extension in response to BD is not solely due to reduced caloric intake, but is also influenced by a diffusible, cell-free component of the bacterial diet.

To determine the relationship between BD and age, animals were switched from control fed to BD diets, or vice versa, at different ages, and survival was monitored. A significant increase in survival was observed when animals were transferred to BD at day 4, 8, 14, 20, or 24 of adulthood (Figure 1A, Table 1). This was true even when BD was initiated after more than 50% of the control-fed population had died. Animals placed on BD at the 2^nd day of adulthood and returned to a control-fed diet at the 8^th, 14^th, or 28^th day of adulthood lived longer than animals fed a control diet for their entire life (Figure 1B, Table 2). By considering the survival of only those animals alive at the time of return to control fed conditions, it was apparent that transient BD induced a significant long-term protective effect resulting in enhanced survival after return to a control fed diet (Figure 1C). This is demonstrated by the significant increase in survival following return to a control fed diet, relative to animals maintained on a control fed diet for life.

We have previously reported that BD induces a robust increase in thermotolerance 27. We wanted to determine whether worms exposed to transient BD retain this stress resistant phenotype, in correlation with the life span benefit. We observed an age-dependent decrease in thermotolerance of both control fed and BD animals; however, BD significantly increased thermotolerance at every age-point examined (Figure 2A; day 7, p = 4.3 × 10^-16; day 18, p = 3.9 × 10^-7; day 22, p = 1.7 × 10^-7), consistent with prior observations 28. Interestingly, enhanced thermotolerance was also maintained in animals subjected to transient BD (four days on BD then returned to the control diet for four days) (Figure 2B; p = 0.02), demonstrating that transient BD provides a long-term enhancement of thermotolerance as well as survival.

During our analysis of BD and the age at which it is initiated, we noted that older control fed animals qualitatively consumed less food than young animals, based on the rate at which bacterial food disappeared from the surface of the agar plates. This is consistent with prior reports that the rate of pharyngeal pumping, which is necessary for food to enter the intestine, decreases dramatically with age 203233. In a longitudinal study of age-related phenotypes, it was shown that no significant pumping can be measured in the last quartile of the mean life span 32. This suggested to us the possibility that BD might increase life span when initiated in animals that were no longer consuming a substantial amount of bacterial food.

To assay feeding behavior as a function of age, we quantified both food consumption and defecation. Food consumption was measured by mixing fluorescent beads comparable in size to bacterial cells with the E. coli food source and observing the presence of food within the gut by fluorescence microscopy. Within one hour of exposure, beads were observed in the intestine of 100% of young adult control-fed animals (Figure 3A). The percent of control-fed animals that are consuming beads decreased dramatically with age, however, and by 24 days of age, beads could not detected in the gut lumen of any animals assayed (Figure 3B). Defecation rate also decreased as a function of age and defecation cycles were not observed in control-fed animals older than 24 days (Figure 3C). Thus, as suggested by prior reports measuring pharyngeal pumping 203233, we conclude that consumption of bacterial cells by the control fed population decreases over the first 2–3 weeks of life and largely ceases by 20–24 days of adulthood.

Since BD extends life span in animals no longer consuming bacteria (see Figure 1, control fed animals transferred to BD at day 24 of adulthood), we considered the possibility that the bacterial food source produces a longevity-limiting signal that is sensed by the worms. Prior studies demonstrated that some sensory-defective C. elegans mutants have increased life span 343536. The possibility that dietary cues directly mediate longevity was not examined in these studies, however, and based on physiological assays it was concluded that the long-lived chemosensory mutants were not mimics of DR 35. Consistent with this conclusion and a recent study using a liquid bacterial dilution DR protocol 37, we observed that BD significantly increased the life span of five long-lived chemosensory mutants examined by Kenyon and colleagues 35: che-3(p801), che-11(e1810), daf-10(e1387), osm-3(p802), and tax-4(p678) (Figure 4A–E). Three of these mutants (daf-10(e1387), osm-3(p802), and tax-4(p678)) were previously demonstrated to be at least partially dependent on daf-16 35; whereas life span extension by BD is daf-16 independent 2728. BD significantly extends life span in each of these chemosensory mutants compared to control fed animals (Figure 4, Tables 3 and 4). Since the magnitude by which BD increased life span in long-lived chemosensory mutants was comparable to the percent life span increase in N2 animals, we suggest that life span extension from BD is likely to be mediated by a food sensing pathway that is at least partially distinct from the chemosensory mutants examined here.

To test the hypothesis that decreased sensing of a dietary component underlies a portion of the life span extension from BD, we asked whether cell-free supernatant from an overnight culture of E. coli OP50 would influence longevity. Addition of E. coli OP50 conditioned supernatant significantly decreased the life span of BD animals (Figure 5A, Table 5). This suppression of BD-induced longevity was not observed when fresh culture media was added to the diet, demonstrating that bacterial metabolism is required for production of the longevity-limiting signal.

As a further test of our hypothesis, we determined the longevity of animals that were physically separated from the bacterial food by an agar barrier (Figure 5B). Animals were grown on NGM + OP50 until the 2^nd day of adulthood, then transferred to the surface of either control fed, BD, or agar barrier plates. Similar to the effect of adding E. coli OP50 conditioned supernatant to the diet, animals maintained on agar barrier plates had a life span that was intermediate between control fed and BD (Figure 5C, Table 5). Thus, we conclude that a diffusible component of the bacterial diet limits longevity, and the absence of this dietary signal accounts for a portion of the life span extension associated with BD.

The health and longevity benefits of DR initiated early in life are well established in a variety of organisms 3. Studies in flies and rodents have indicated that at least some of these benefits can still be attained if DR is initiated during middle age 3031. Our work further supports this idea by demonstrating that BD significantly increases survival in C. elegans at any point during adulthood, at least up to the median population life span. Importantly, transient BD early in life results in a long-term protective effect, increasing both survival and stress resistance even after return to a control diet. Finally, we report the surprising observation that BD increases survival by a mechanism that is at least partially distinct from reduced food consumption, and involves reduced sensing of a diffusible signal produced by the bacterial component of the diet. Based on these observations, we propose that DR increases life span through two distinct mechanisms: calorie restriction and reduced food sensing (Figure 6).

The idea that dietary cues could influence nematode longevity via sensory signaling is supported by prior observations. C. elegans is attracted to both soluble and volatile compounds produced by the bacterial food source, and animals can sense and respond to food through chemotaxis 38. Although muscle tissue undergoes age-related deterioration, the nervous system remains remarkably intact in worms of advanced age, 39. This is consistent with our model that BD extends life span in worms of advanced age, not because of reduced food consumption, but because of the loss of a life span-limiting signal produced by the bacteria and sensed by the worms. In addition, several mutants defective in chemosensation are long-lived. The additive life span extension from combining long-lived chemosensory mutants with BD is consistent with our model, as these mutants were previously shown not to be genetic mimics of DR 35. Indeed, since life span extension in chemosensory mutants is largely dependent on daf-16 35, and multiple epistasis studies have placed DR and daf-16 in different genetic pathways, an additive life span extension is the predicted outcome of these experiments.

How can a sensory mechanism of longevity determination by BD be reconciled with other models of DR in C. elegans? Mutation of eat-2 causes a pharyngeal pumping defect that results in reduced food consumption 1213, but has not been shown to cause sensory defects. Thus, the life span extension from mutation of eat-2 is likely due to a reduction in caloric intake. Interestingly, BD increases life span to a greater extent than mutation of eat-2, and eat-2 mutants subjected to BD are not longer-lived than N2 animals subjected to BD 2728. One interpretation of these data is that reduced food intake (e.g. in the eat-2 mutant) is sufficient to increase life span, while reduced food intake coupled with reduced food sensing (e.g. in the case of BD) leads to an even greater life span extension. Axenic culture is somewhat more complicated because, although the animals are maintained in the absence of food, as they are in BD, the growth media is very nutrient rich (compared to NGM) and contains additional signals that may be sensed by C. elegans. eat-2 mutants subjected to axenic culture live longer than N2 animals on axenic growth 23, suggesting that mutation of eat-2 and axenic growth are not equivalent methods of DR.

Previous studies have demonstrated a role for the bacterial food source in life span determination – through bacterial pathogenesis. A number of bacterial strains have been shown to be pathogenic in C. elegans 40, and worms cultured on a non-pathogenic species, B. subtilis, were shown to have extended longevity compared to those cultured on the normal food source, E. coli OP50 41. OP50 has also been shown to affect life span through bacterial proliferation in the gut 42. In our studies, we maintain worms on UV-irradiated OP50 as a food source to eliminate bacterial pathogenesis as a complicating factor in life span analysis. However, we do find that a soluble product made by OP50 partially suppresses the life span extension by BD, independent of bacterial proliferation, supporting our model that reduced food sensing contributes to the longevity observed in animals subjected to BD. Formally, it is possible that the bacteria produce a toxin that is life span-limiting, independent of the pathogenic effects of bacterial proliferation in the gut. This does not change the interpretation of our data, however, which is that a diffusible, cell-free component of the diet can limit the life span extension from BD.

Recent studies in the fruit fly have indicated that a volatile component of the diet can also suppress life span extension from DR in that organism 43. Thus, modulation of longevity through food sensing may represent an evolutionarily conserved mechanism linking diet and aging. It will be of interest to determine which are the precise chemical cues in the diet that influence longevity in each of these organisms. At this time, there is little direct evidence that food sensing modulates aging in mammals; however, it is reasonable to speculate that food could induce relevant physiological changes through olfactory or taste cues, independent of food consumption. This raises the intriguing possibility that one class of DR mimetics may be achieved by interfering with food sensing, but without requiring reduced caloric intake.

EDS participated in the conception and design of the study, coordinated the data acquisition and analysis, and drafted the manuscript. TLK participated in the conception and design of the study, data acquisition and analysis. BTL contributed to experimental design, data acquisition, and data analysis of feeding and defecation assays. JS contributed to experimental design, data acquisition, and data analysis of thermotolerance assays. KLW contributed to experimental design, data acquisition, and data analysis of life span experiments. BKK participated in the conception and design of the study, and drafted the manuscript. MKK conceived of the study, participated in its design and coordination, and drafted the manuscript. All authors read and approved the final manuscript.