Bacterial diets differentially alter lifespan and healthspan trajectories in C. elegans

Diet is one of the more variable aspects in life due to the variety of options that organisms are exposed to in their natural habitats. In the laboratory, C. elegans are raised on bacterial monocultures, traditionally the E. coli B strain OP50, and spontaneously occurring microbial contaminants are removed to limit experimental variability because diet—including the presence of contaminants—can exert a potent influence over animal physiology. In order to diversify the menu available to culture C. elegans in the lab, we have isolated and cultured three such microbes: Methylobacterium, Xanthomonas, and Sphingomonas. The nutritional composition of these bacterial foods is unique, and when fed to C. elegans, can differentially alter multiple life history traits including development, reproduction, and metabolism. In light of the influence each food source has on specific physiological attributes, we comprehensively assessed the impact of these bacteria on animal health and devised a blueprint for utilizing different food combinations over the lifespan, in order to promote longevity. The expansion of the bacterial food options to use in the laboratory will provide a critical tool to better understand the complexities of bacterial diets and subsequent changes in physiology and gene expression.


INTRODUCTION
Since the discovery that aging can be improved by dietary, genetic, and pharmacological interventions, there has been an increase in studies aiming to elucidate how each of these interventions can promote long life and healthy aging. Many dietary interventions including calorie restriction, intermittent fasting, and dietary restriction [1] have been shown to not only increase maximal lifespan, but average lifespan of the cohort and healthspan as well. New diets and fads have become popularized within society [2,3], however, these dietary fads focus on removing one aspect of diet (carbohydrates, sugars, fats, proteins, etc.) instead of focusing on the overarching complexity of food.
Food is imperative for all organisms in order to provide nourishment to fuel growth and essential cellular functions. Diet is one of the more variable aspects in life due to the vast options that organisms are regularly exposed to in their natural habitats. Survival within these environments require organisms to select high quality food sources from a range of nutritiously diverse alternatives. For centuries, society has accepted that diet is important for health and longevity; "You are what you eat." However, our understanding of why diet holds such profound influence over our health and how we can use this knowledge to improve overall health and longevity requires further study.
Many diet studies have employed the use of the model organism Caenorhabditis elegans [4][5][6] due in part to shared metabolic pathways with mammals including AMPK, TOR, and insulin/IGF-1 (IIS) signaling [6]. These bacterivore nematodes can be used for the identification of dietary effects due to their invariant and short developmental and reproductive periods, which are followed by an averaged 3-week lifespan [7].
The impact of diet on physiology is vast and certainly integrates into multiple life history traits including development [8], reproduction [9], healthspan [10], and longevity [4,[11][12][13]. Additional studies have identified diet-induced phenotypic effects that are accompanied by measurable differences in metabolic profiles [11,14,15], fat content [5,6,8,16], and feeding behavior [8,13,17,18]. Under normal laboratory conditions, C. elegans are cultured using the standardized bacterial species Escherichia coli OP50. This diet was not chosen because of its association with nematodes in the wild, but because of availability in the laboratory setting [16]. Maintenance of C. elegans in the laboratory includes the routine removal of spontaneously occurring bacterial contaminants in order to limit experimental variations. Surprisingly, very few options to study the impact of different diets on C. elegans healthspan and lifespan are defined in the laboratory setting [16,19].

Identification and characterization of three distinct bacterial diets
In their natural environment, C. elegans need to adapt to a variety of microbes they might encounter as a source of food. The necessity to cope with varied food sources have likely shaped multiple aspects of their biology, which are masked in the artificial laboratory environment where they are fed the unnatural diet of E. coli [7]. In order to elucidate the underlying connection between diet and physiology, it is important to evaluate how different diets influence behavior and physiology. Occasionally, contaminating microbes "pop up" spontaneously on culture media and are often quickly removed to maintain a standardized monoculture diet. We routinely noticed that C. elegans are found feeding on contaminants, when present, rather than the E. coli food source supplied. With this in mind, we looked at these "contaminants" as an opportunity to examine the impact that different food sources have on physiology.
We isolated and cultured three bacterial species as novel C. elegans diets that we call "Red", "Orange", and "Yellow" due to their pigments when grown on plates ( Figure 1A). Using 16S ribosomal sequence alignment, we identified the genus of each new diet: Red is Methylobacterium; Orange is Xanthomonas, and Yellow is Sphingomonas. Intriguingly, these three novel bacterial diet genera can be found in C. elegans natural environments [16]. We next identified optimal growth parameters to ensure monoculture growth for each diet in LB broth and plates (figure supplement 1A). Although most species were able to grow at 37°C on LB solid media and in LB liquid media, Red/Methylobacterium required a growth temperature of 30°C on LB solid media, while Yellow/Sphingomonas required a growth temperature of 26°C in LB liquid media. The growth rate of each bacterial species was similar, except Red/Methylobacterium, which was markedly reduced ( Figure  1B).
To assess the nutritional value of each diet, we first performed bomb calorimetry to define the total caloric composition of each bacteria. Surprisingly, the total caloric composition was not significantly different in any of the diets in comparison to the E. coli OP50/B diet. In order to better assess the differential nutrient composition of each bacteria, we next measured the concentrations of glucose, glycerol, glycogen, triglyceride, and water to define a nutritional profile for each diet ( Figure 1C and figure supplement 1B-G). The Orange/Xanthomonas diet was the most significantly different diet among the three novel diets because glucose, glycerol, glycogen, and water content were all higher than the levels found in OP50/B. The Red/Methylobacterium diet displayed a similar trend with an increase in glycerol, glucose, and water content, but not triglycerides. Finally, the Yellow/Sphingomonas diet only carried more glycerol and water content, compared to OP50/B. Interestingly, HB101/B&K-12 was the only diet with an increase in triglyceride content relative to OP50/B. We also noted that glucose and glycogen content were highest in Orange/Xanthomonas, glycerol and triglyceride content were highest in HB101/B&K-12, and water content was highest in Yellow/Sphingomonas. Taken together, Red/Methylobacterium, Orange/Xanthomonas, and Yellow/Sphingomonas represent three new potential C. elegans diets, with similar calorie content, but differential nutritional composition to OP50/B that could be used to investigate dietary effects on animal physiology.

Novel diets direct unique transcriptional signatures
We next confirmed that worms could be cultured on each diet as the sole source of nutrition to enable the systematic characterization of the impact of these diets on C. elegans health. After maintaining animals on each diet for more than 20 generations, we then compared animals reared on these new laboratory diets alongside populations grown on the three most common E. coli diets (OP50/B, HT115/K-12 and HB101/B&K-12).
Because diet can impact multiple cellular processes, we first assessed the steady state transcriptional response to each diet by RNA-seq (Figure 2A  , except for animals fed the Yellow/Sphingomonas diet, which resulted in the depletion of intestinal lipids in late reproductive adults, a phenotype known as age-dependent somatic depletion of fat (Asdf) at day 3 of adulthood ( Figure 4M,N) [32,33].
The fact that animals fed each of the diets developed faster than OP50/B-reared animals indicates animals were not nutritionally deprived, which can result in developmental delay [20,[34][35][36][37][38]. Nevertheless, different microbes have been shown to alter rates of pharyngeal pumping, which regulates ingestion of bacteria [39]. Pharyngeal pumping rates were examined in both the L4 stage and day 1 adult animals, but none of the diets significantly altered pharyngeal pumping ( Figure 4O and figure supplement 3).
Because each diet represented a specific metabolic profile and C. elegans reared on these diets differed in the amounts of stored intracellular lipids, we examined the expression of genes that regulate lipid metabolism and homeostasis ( Figure 4P). Of the total number of genes in each diet that were differentially expressed, about 1/3 of the genes in each diet were related to metabolism. Feeding animals the Yellow/Sphingomonas diet evoked the largest number of metabolism-related gene changes, which correlated with our phenotypic observation that animals fed this diet store the lowest levels of fat at the L4 stage ( Figure 4G) and it was the only diet to drive the Asdf phenotype later in life ( Figure 4M). Among the genes differentially expressed on these new diets were fat-5 and fat-7, which may suggest changes in lipid biosynthesis pathways, as well as multiple lipases (Lipl-class), suggesting changes in lipid utilization ( Figure 4G).

Red/Methylobacterium and Yellow/Sphingomonas reduce reproductive output
We next investigated how each diet influenced fertility by measuring the total number of viable progeny laid by individual animals over a ten-day reproductive span. The total number of progeny was reduced ~25% when animals were fed the Red/Methylobacterium diet, modestly reduced (~10%) in animals fed the Yellow/Sphingomonas diet, and unchanged on other microbial diets ( Figure 5A). Although animals reached peak reproductive output at approximately the same time, animals fed the Red/Methylobacterium and HB101/B&K-12 diet ceased reproduction sooner than all other groups ( Figure 5B and figure supplement 4). Reproductive output of C. elegans hermaphrodites is determined by the number of spermatids generated at the L4 larval stage [40,41]. To determine if the decreased progeny production and early loss of reproductive output were due to diminished sperm availability in Red/Methylobacterium-reared worms, we mated hermaphrodites with males, which increases total reproductive output ( Figure 5C). Males raised on either the OP50/B or Red/Methylobacterium diets could increase total reproductive output similarly when mated to OP50/B-reared or Red/Methylobacterium-reared hermaphrodites, which indicated that sperm are functional when in excess. Surprisingly, when mated, hermaphrodites raised on Red/Methylobacterium have significantly more progeny as compared to hermaphrodites raised on OP50/B. We also counted the number of unfertilized oocytes laid by unmated hermaphrodites, which revealed HT115/K-12 and HB101/B&K-12, Red/Methylobacterium, and Orange/Xanthomonas diets resulted in markedly reduced expulsion of unfertilized gametes ( Figure 5D) while hermaphrodites on Yellow/Sphingomonas were similar to animals fed the OP50/B diet. An analysis of the reproduction-related genes that were differentially changed in animals fed each of the diets revealed several diet-specific changes ( Figure 5E).

Novel diets differentially alter organismal life expectancy and health
The quantity of diet ingested has been shown to influence lifespan across organisms [42,43]. Similarly, diet quality and composition can also influence lifespan, but the underlying mechanisms of healthspan improvement, and its relationship to diet, remain underdeveloped. As such, additional models to explore how diet can impact life-and healthspan are of great interest. To this end, we examined impact of each new diet on organismal lifespan. Similar to previous studies [13] animals raised on HB101/B&K-12 had a modest increase in mean lifespan, with the most significant impact on the last quartile of life ( Figure 6A and supplementary file 2). In contrast, worms raised on HT115/K-12, Red/Methylobacterium, and Yellow/Sphingomonas displayed significantly increased lifespan (Figure 6B-D and supplementary file 2) while animals raised on Orange/Xanthomonas were short-lived ( Figure 6E and supplementary file 2). Taken together, our results show that Red/Methylobacterium, Yellow/Sphingomonas, and Orange/Xanthomonas represent three diets of sufficient nutritional quality to accelerate development, but differentially alter life expectancy in a dietdependent manner.
We examined the transcriptional profiles of genes previously annotated to be involved in lifespan and discovered several interesting trends that may explain the lifespan-altering effects of each diet ( Figure 6F). Feeding the Yellow/Sphingomonas results in an increase in lifespan and, correspondingly, genes in the insulinlike signaling pathway (daf-2/Insulin receptor and age-1/PI3K) were downregulated, while DAF-16/FoxO target genes (Dao and Dod) were upregulated. Similarly, although less significantly, animals fed the Red/Methylobacterium diet have increased expression of several genes regulated by DAF-16. Moreover, the HT115/K-12, Red/Methylobacterium and Yellow/Sphingomonas all extend lifespan and share 11 genes with altered expression, all of which are up regulated on these diets. In previous studies, these 11 genes (abu-1, abu-11, abu-14, gst-10, kat-1, lys-1, pqn-2, pqn-54, lpr-5, glf-1, bus-8) lead to a shortened lifespan phenotype when the expression is reduced by RNAi [30]. Finally, feeding worms the Orange/Xanthomonas diet significantly decreased lifespan and although we did not identify any known longevity genes to display transcriptional changes, we found that kgb-1, pha-4, and vhp-1 were all up regulated on this diet, while previous studies have observed extended lifespan when these genes are targeted by RNAi [30]. Taken together, while longevity is a complex phenotype, feeding any of these diets can evoke transcriptional changes in genes associated with life expectancy.
Since lifespan can be uncoupled from healthspan [10,44] we examined the impact of each diet on animal muscle function via thrashing as a surrogate for health ( Figure 7A-E and figure supplement 4). We measured thrashing rate at five specific timepoints that were selected for their significance in relation to other phenotypes: L4, based on the RNAseq data and fat staining analysis; day 1 adult, based on the pharyngeal pumping assay; day 3 adult, based on the presentation of the Asdf phenotype; day 8 adult, due to 50% of the Orange/Xanthomonas population being dead; and day 11 adult when 50% of the OP50/B population had perished. Surprisingly, animals raised HT115/K-12, HB101/B&K-12, Red/Methylobacterium and Yellow/Sphingomonas had faster muscle movement while Orange/Xanthomonas were indistinguishable from OP50/B-fed animals at L4 stage ( Figure 7A  . Day 11 adults raised on HT115/K-12, HB101/B&K-12, Yellow/Sphingomonas, and Orange/Xanthomonas moved half as fast as they did at the end of development, similar to OP50/B-raised control animals, but Red/Methylobacterium-fed animals we significantly slower (~75% reduction) ( Figure 7E  and supplementary file 3). Taken together, despite some diets enhancing early life muscle function, agerelated decline remained similar, and was even enhanced in animals fed Red/Methylobacterium (Figure 7F). It is notable that when examining the GO-term analysis of the 30 shared genes that are deregulated on all five of the diets, as compared to OP50/B, all 30 genes have been annotated as causing movement defects in C. elegans when the expression of that gene is altered [45][46][47].

Diet as a nutraceutical to alter lifespan trajectories
Previous studies have demonstrated that effects of calorie restriction (CR) can be realized even when initiated later in life [48,49]. Moreover, when fed ad libitum after a period of CR, mortality is shifted as if CR never occurred. With this model in mind, we asked if switching diet type, rather than abundance, could be wielded to alter lifespan outcomes. To address these questions, we switched growth conditions at major life stages (development, reproduction, post-reproduction) and measured lifespan (Figure 8A). We used both the Orange/Xanthomonas diet, which decreases lifespan, and Red/Methylobacterium diet, which increases lifespan, as the basis of our model. Remarkably, after assessing eight combinations of diet switching, we discovered that the diet fed during development (diet 1) and reproductive period (diet 2) contributed little to overall lifespan and the last diet exposed (diet 3) exerted the most impact. In brief, when animals were exposed to the Red/Methylobacterium diet after experiencing the Orange/Xanthomonas diet, the normally shortened lifespan resulting from ingestion of Orange/Xanthomonas is suppressed (Figures 8B-D and  supplementary file 4). Conversely, exposure to the Orange/Xanthomonas diet suppresses the normally extended lifespan linked to whole-life feeding of the Red/Methylobacterium diet (Figures 8E-G and  supplementary file 4). Intriguingly, when compared to animals raised on the OP50/B diet, ingestion of the Red/Methylobacterium post-reproductively, regardless of ingestion of the Orange/Xanthomonas diet at any other life stage, results in a relatively normal lifespan; not shortened ( figure supplement 6). Moreover, if animals that eat the Red/Methylobacterium diet post-reproductively have an extended lifespan, which is further enhanced if the diet is introduced after development. Taken together, our studies reveal that expanding the available diets to feed C. elegans is a powerful tool to study the impact of diet on lifespan and healthspan. Future studies to integrate genetic analyses to define new gene-diet pairs [6,50] and gene-environment interactions in general, will be of significant interest.

DISCUSSION
C. elegans is a well-established model to study diet and aging. Here we augment that model by introducing a comprehensive phenotypic analysis of C. elegans fed three novel laboratory diets: Red/Methylobacterium, Orange/Xanthomonas, and Yellow/Sphingomonas. Interestingly, these microbes have been found in the normal C. elegans environments [16] and when compared to the three most common E. coli diets OP50/B, HT115/K-12, HB101/B&K-12, our study is a critical tool to aid in our understanding of how these two distinct environments influence physiological and transcriptomic responses over the lifespan [11,16]. Previous works have identified how C. elegans react to the three standard E. coli diets provided in the lab [5,8,[50][51][52]] and yet, to our knowledge, a "head-to-head" comparison of the age-related and healthspan-relevant outcomes that result from feeding these diets, is lacking. Because of this, we decided to investigate the effects of diet on both physiological attributes and transcriptional signatures of C. elegans raised on different bacterial species for over twenty generations, to avoid acute stress responses to the diet.
Recent studies have shown that diet can alter transcriptional responses in C. elegans [53,54]. Our work supports this observation with three new menu options for C. elegans culture. Although each of these diets evokes a unique transcriptional signature, several specific classes of genes are shared among multiple, or even all of the, diets. Our RNAseq analysis was limited to gene expression changes observed as animals enter adulthood, prior to reproduction. Given the extent of changes in reproductive capacity, health (movement and fat), and aging, future work to examine how gene expression is altered on each diet over the lifespan will be of great interest. Moreover, a fine-tuned analysis of transcription at each development stage will be informative based on our observation that different diets can accelerate the transition of animals across specific developmental stages. Nevertheless, it is clear that the standard laboratory E. coli diets and our three new diets induce a transcriptional response of phenotypically relevant genes.
Bacteria serve as a live food source for C. elegans both in their natural environment and laboratory setting. Many aspects of the bacterial communities themselves have been shown to influence multiple attributes in C. elegans, including food preference, feeding rates, brood size, and lifespan [55,56]. Previous studies have shown animals pumping at similar rates in the presence of bacteria they can and cannot eat [12,57]. The rates of pharyngeal pumping were not significantly different on any of the diets in L4 stage and day 1 adult animals ( Figure 4O and figure supplement 4A). Although actual bacterial ingestion can be measured, it is technically challenging, and the faster developmental timing observed (Figure 3) and large-sized broods (Figure 5A), demonstrate that each diet is providing sufficient nutrition to the worms feeding on them. Nevertheless, based on the similarities of animals eating Red/Methylobacterium and animals undergoing dietary restriction, it remains possible that this bacterium allows ad libitum ingestion with the physiological benefits of reduced eating.
Clearly, changing diets can potently impact multiple phenotypic attributes and several of these aging-relevant phenotypes are interrelated. For example, reproduction, fat, and stress resistance are intrinsically tied to overall life expectancy [9]. In support of this model, the Red/Methylobacterium and Yellow/Sphingomonas diets, which reduce overall reproduction, indeed increase lifespan. However, our study reveals that this relationship is more complex as these two diets have different effects on lipid storage and the Yellow/Sphingomonas diet evokes an Asdf response, which is a phenotype observed in animals exposed to pathogens [32,33]. It may be that C. elegans perceive the ingested Yellow/Sphingomonas diet as a pathogen, however the increased lifespan that follows on this diet, suggests this diet is health-promoting. Lastly, genetic and environmental mechanisms that delay developmental timing have been tied to increased longevity in adulthood [36,58,59].
Each of the diets we tested results in faster development into a reproductive adult, but only HB101/B&K-12, HT115/K-12, Red/Methylobacterium, and Yellow/Sphingomonas increase lifespan, while Orange/Xanthomonas decreases lifespan. Taken together our study reveals that each diet can exert a specific life history changing response, but also questions previously-established models of aging.
Inspired by previously described dietary interventions that are capable of extending lifespan in multiple species [43,60,61], we wanted to investigate whether, instead of keeping animals on one diet their entire life, altering dietary exposure at critical life stages could affect overall lifespan. We asked whether these diets could alter lifespan without any previous generational exposure. Strikingly, acute exposure to these diets was capable of altering lifespan (Figure 8), but intriguingly the magnitude of the response differed from animals that have been on these diets for 20+ generations (Figure 6). Furthermore, our study revealed that the diet fed during the post-reproductive period exerted the strongest influence over the lifespan of the cohort. This result, however, is potentially confounded by the amount of time (longitudinally) that is spent on each diet. Regardless, our study reveals that any potential early-life (development and reproductive span) exposure to diets that normally shorten lifespan can be mitigated by eating a lifespan-promoting food option. This result is reminiscent of previous studies showing the mortality rate of dietary restricted (DR)-treated animals is accelerated when switched to an ad libitum diet while mortality rate is reduced when ad libitum-fed animals are switched to a DR diet [62]. In our case, calories are perhaps not different on the Red/Methylobacterium or the Orange/Xanthomonas diets, but rather the overall nutritional composition of the diet is different. Given that in our model animals are chronically exposed to as much food as they can eat throughout their life, we posit that if a similar human diet were discovered that this treatment protocol would be more accessible as food restriction is difficult, socially and psychologically [63][64][65].
Taken together, our study expands the menu of bacterial diets available to researchers in the laboratory, identifies diets with the ability to drive unique physiological outcomes, and provides a diet quality approach to better understand the complexities of gene-diet interactions for health over the lifespan.  Gene Ontology (GO) term enrichment analysis of RNAseq data in larval stage 4 animals. (A) Volcano plot of all differentially-expressed genes in each diet relative to OP50/B-reared worms. All genes considered to be significant have a p-value <0.01. (B-F) Volcano plots of each individual diet with all significant genes that are differentially expressed. (G) Venn diagram of all significant genes. Shows the number of genes shared between two, three, four and five of the diets, along with the number of genes unique to each diet. GO-terms for uniquely-expressed genes are noted below and split between genes that were up-regulated and downregulated. Further information on genes unique to each diet and specific GO terms can be found in supplementary file 1.        Reproductive output of individual worms (each line represents a different worm) at each day of its reproductive span. The thicker green line (A) and black line (B-F) on the graph represents the average output per day of reproduction. Worms were moved every 24 hours and progeny were counted. Each diet had a reproductive peak from day 1 to day 2 of adulthood and then started producing fewer progeny per day until day 5 of adulthood. HB101/B&K-12 (C) and Red/Methylobacterium (D) worms have a significant decline in reproductive output by day 3 of adulthood.    Table S1. RNAseq analysis in L4 C. elegans.