Transgenerational programming of longevity and reproduction by post-eclosion dietary manipulation in Drosophila

Introduction

Accumulating studies have established a strong link between early-life nutrition and adult health and disease [1-6]. In particular, maternal and postnatal malnutrition, among other environmental factors, can profoundly influence adult health outcomes and increase the subsequent risk of noncommunicable diseases (NCDs, including metabolic syndrome, diabetes, cardiovascular disease, etc.). The World Health Organization has recognized unhealthy diets as one of the leading causes of NCDs and such chronic diseases kill over 38 million people annually [7, 8]. Strikingly, 44% (or 16 million) of the deaths are premature, manifesting shortened average lifespan, one key characteristic of NCDs (cf. [9]). A very recent study, for example, indicates that combination of diabetes and heart disease may lead to a substantially lower life expectancy by over a decade [10]. The Organization has further recognized that maternal and child under-nutrition increases the subsequent risk of NCDs, accounts for 11% of the global burden of diseases, and is the underlying cause of 35% of early child death [11, 12].

It is now widely accepted that early-life nutrition programs the long-term health of an individual and his/her offspring [5, 6, 13, 14]. Such nutrition-mediated programming effects in fact have often been shown to be inheritable and transgenerational. Aiken and Ozanne recently performed a systematic and exhaustive search of the literature about transgenerational developmental programming for both epidemiological and animal studies [15]. Out of 45 primary research papers using rodent models that bred and examined a phenotype in offspring at least as far as the F2 generation, roughly 2/3 of them have employed early-life nutritional intervention (see Table 1 from [15]), highlighting “early-life nutrition” and “transgenerational effects” as key elements of developmental programming. However, the underlying mechanisms are only starting to emerge, with epigenetics as perhaps the most important mechanism through which diet and nutrition can directly influence the genome and response to diet itself [15-17]. Epigenetic modifications including DNA methylation, histone modifications, and non-coding RNA-based mechanisms are heritable but reversible changes that affect gene expression without altering the underlying DNA sequence. Such epigenetic “marks” influenced by early-life nutrition may therefore influence the subsequent health later in life and even across generations [5, 6, 15-19]. Importantly, trans-generational inheritance of nutritional programming of metabolic status has been recently demonstrated in flies [20], supporting “Drosophila as a valid model for studies of the epigenetic inheritance of metabolic state” [21] and the proof-of-concept of studying nutritional programming of other disease conditions in this much simpler while genetically tractable system.

Several recent studies have also started to implicate early-life nutrition in the regulation of longevity [14, 19]. Flies had a shortened lifespan when raised on a high-sugar (HS) diet for 3 weeks [22]. Considering that the same HS diet elicited transgenerational metabolic alternations up to F2 offspring [20], the shortening effect of longevity may propagate across generations. Consistently, an earlier epidemiologic study demonstrated a negative correlation between the lifespan of grandsons and food consumption of parental grandfather during their slow growth period [23]. However, it remains unexplored whether immediate postnatal nutritional manipulations may program longevity, and whether such potential programming effects may propagate across generations.

Drosophila serves as a promising model for studying ageing and longevity, especially considering its short lifespan and usefulness for rapid gene discovery. A short lifespan makes it possible to complete one longevity experiment in about 3 months, rather than several years with a typical rodent model. Furthermore, Drosophila has been increasingly used for modeling human diseases and for drug discovery and development [24-26]. Importantly, Buescher et al. have employed a 7-day post-eclosion dietary treatment to demonstrate trans-generational metabolic programming in flies [20], and several diets have been used to show nutritional effects on learning and memory across multiple generations [27, 28]. Therefore, Drosophila presents itself as an excellent system to model transgenerational inheritance of nutritional programming of longevity and then to rapidly identify and characterize the underlying genetic and epigenetic pathways.

In this study, we examined whether specifically-defined PDMs may program longevity, and, if so, whether such programming effects may be long-lasting and inheritable across generations in Drosophila. The LP, IP and HP diets were adapted, because different protein-carbohydrate intakes appear to be essential for longevity in Drosophila [29, 30] and because similar diets have been used to demonstrate nutritional effects on learning and memory across multiple generations [27, 28]. We also assayed the effects of the same PDMs on lifetime fecundity, as reproduction is generally considered to trade off against longevity, with increased reproduction frequently associated with shortened lifespan [31]. We found that the IP diet improved longevity and lifetime fecundity while the LP and HP diets tended to decrease longevity and fecundity across three generations. Interestingly, no clear evidence was observed to support any trade-off between longevity and reproduction, suggestive of the feasibility of elevating both longevity and reproduction with proper nutrition across generations.

Results

Transgenerational inheritance of nutritional programming of longevity

Newly-born F0 flies were subjected to a 7-day PDM with one of three different diets (LP, IP and HP; see Table S1 for detailed description). Our longevity study was performed with these treated F0 flies and their F1, F2, and F3 progenies while being maintained on the CD food (i.e., without any additional exposure to the LP, IP or HP food across the F0–F3 generations) throughout their developmental and whole adult lives (see Figure S1 and Table S2 for detailed procedures and experimental design). The data were collected with 4 types of flies (i.e., virgin males and females, mated males and females) simultaneously. Such a design allowed us to assay whether the nutritional effects on longevity were similar among all 4 types of flies and further evaluate whether nutrition may affect longevity due to gender, mating, and/or reproduction. Overall, rather similar effects on longevity were observed for virgin males, virgin females, mated males, and mated females across the F0–F2 generations with the LP and IP diets but not with the HP diet (Figure 1 and Table 1).

Figure 1. Survival curves of various flies across the F0–F2 generations after 7-day PDMs of the F0 parents with the LP, IP, or HP diet. (A) Virgin male (1st row), (B) virgin female (2nd row), (C) mated male (3rd row), and (D) mated female (4th row) flies; and (0) F0 parents (1st column), (1) F1 offspring (2nd column), and (2) F2 offspring (3rd column). N = 178–185 for the F0 and F1, and 199–205 for the F2 generation (see supplemental Table S2 for specific sample size for any given group).

Table 2. Statistical analyses of effects of gender, mating and reproduction on nutritional reprograming of longevity, using the same survival data as shown in Figure 1

Comparison	PDM diets	F0 parents: subjected to 7-day PDMs			F1 offspring			F2 offspring
		Mantel-cox	Median lifespan		Mantel-cox	Median lifespan		Mantel-cox	Median lifespan
		P value	Value	Change	P value	Value	Change	P value	Value	Change
Virgin males vs. females	CD	< 0.0001	36 vs. 42	14% ↓	0.0018	40 vs. 46	13% ↓	< 0.0001	40 vs. 48	17% ↓
	LP	< 0.0001	26 vs. 34	24% ↓	< 0.0001	30 vs. 35	14% ↑	0.0005	34 vs. 38	11% ↑
	IP	< 0.0001	48 vs. 56	14% ↑	0.0002	50 vs. 56	11% ↑	< 0.0001	48 vs. 56	14% ↑
	HP	0.023	34 vs. 40	15% ↑	0.14	38 vs. 42	10% ↔	0.07	40 vs. 44	9% ↔
Mated vs. virgin males	CD	0.41	36 vs. 36	0% ↔	0.066	36 vs. 40	10% ↔	0.08	40 vs. 40	0% ↔
	LP	0.0005	24 vs. 26	8% ↑	0.0064	30 vs. 30	0% ↔	0.062	32 vs. 34	6% ↔
	IP	0.15	46 vs. 48	4% ↔	0.081	48 vs. 50	4% ↔	0.45	48 vs. 48	0% ↔
	HP	0.97	35 vs. 34	3% ↔	0.061	38 vs. 38	0% ↔	0.071	38 vs. 40	5% ↔
Mated vs. virgin females	CD	0.31	40 vs. 42	5% ↔	0.078	44 vs. 46	4% ↔	0.11	46 vs. 48	4% ↔
	LP	0.015	32 vs. 34	6% ↑	0.023	32 vs. 35	9% ↑	0.31	36 vs. 38	5% ↔
	IP	0.0039	52 vs. 56	7% ↑	0.0066	54 vs. 56	4% ↑	0.11	54 vs. 56	4% ↔
	HP	< 0.0001	36 vs. 40	10% ↑	0.0006	36 vs. 42	14% ↑	0.15	42 vs. 44	5% ↔
The transgenerational longevity experiments were designed so that all 4 types of flies (i.e., virgin males and females, mated males and females) were assayed for longevity simultaneously. Such a design allowed us to ask whether nutrition may affect longevity differently because of gender, mating, and reproduction. For the potential effect of gender, survival curves and median lifespan for virgin males and females were compared within CD or each PDM diet across the F0–F2 generations (i.e., first 4 rows). For the potential effect of mating alone, survival curves and median lifespan for virgin and mated males were compared within CD or each PDM diet across the F0–F2 generations (i.e., next 4 rows). For the potential effect of reproduction (and mating), survival curves and median lifespan for virgin and mated females were compared within CD or each PDM diet across the F0–F2 generations (i.e., last 4 rows). ↑: significantly shortened lifespan; ↑: significantly improved lifespan; ↔: no significant lifespan change.

For the F0 generation, longevity was shortened after the LP PDM of the F0 parents for all 4 types of flies (Figure 1A0–D0, red diamonds; Table 1). Correspondingly, the median lifespan was decreased by 19–33% (Table 1). The HP PDM decreased longevity of the females significantly (P = 0.034 for virgin females and P < 0.0001 for mated females) or by 5–10% at the median lifespan, but did not affect longevity of males (P = 0.61 for virgin males and P = 0.47 for mated males; Figure 1A0–D0, purple squares; Table 1). In contrast, The IP PDM extended longevity for all 4 types of flies significantly, or by 28–33% at the median lifespan (Figure 1A0–D0, green circle; Table 1). These observations demonstrated that longevity was programmed after the 7-day PDMs in flies, and the HP treatment influenced longevity of the males and females differently.

For the F1 generation (Figure 1A1–D1), longevity was shortened after the LP PDM of the F0 parents for all 4 types of flies (red diamonds; P < 0.0001, Table 1). Consistently, the median lifespan was decreased by 17–27% (Table 1). The HP PDM decreased longevity of the females significantly (P = 0.016 for virgin females and P=0.0005 for mated females), or by 9–18 % at the median lifespan, but did not affect longevity of males (P = 0.85 for virgin males and P = 0.44 for mated males). In contrast, The IP PDM extended longevity for all 4 types of flies significantly (green circle; P < 0.0001, Table 1), or by 22–33% at the median lifespan. Considering that the germ cells (future gametes giving rise to F1 offspring) from the F0 males and females were also exposed to various PDMs while the F1 flies had never been exposed to LP, IP or HP diets, these data suggest that the nutritional programming of longevity was propagated from the F0 to F1 generation through parental effects.

For the F2 generation (Figure 1A2–D2), longevity was shortened after the LP PDM of the F0 parents for all 4 types of flies (red diamonds; P < 0.0001, Table 1), and the median lifespan was decreased by 15–22% (Table 1). The HP PDM decreased longevity of the females significantly (P = 0.0059 for virgin females and P=0.015 for mated females), or by 8–9 % at the median lifespan, but did not affect longevity of males (P = 0.67 for virgin males or P = 0.21 for mated males; Figure 1A2–D2, purple squares, and Table 1). The IP PDM again extended longevity for all 4 types of flies significantly (P = 0.0009 for virgin males and P < 0.0001 for the other three groups; Figure 1A2–D2 and Table 1), or by 17–20% at the median lifespan. Such observations clearly suggest that the nutritional programming of longevity may be transmitted to the F2 offspring through transgenerational inheritance, as both F1 (including their germ cells) and F2 flies have never been exposed to the LP, IP or HP diets. Such a conclusion was further supported by our observation of the F3 offspring (Figure S2) whereby longevity was still shortened after the LP PDM of F0 parents while improved after the IP PDM.

Figure 2. Lifetime fecundity and accumulative differences of egg production across the F0–F2 generations after 7-day PDMs of F0 parents with the LP, IP, or HP diet. (A0–A2)Average lifetime fecundity (i.e., total eggs produced by 1 female fly in its lifetime) was shown for the 100 mated females across three generations. The F0 females showed lower fecundity, as egg-production data collection was initiated when these flies were 11-day-old (because of the 7-day PDMs), while the F1 and F2 flies were 4-day-old. One-way ANOVA indicated a significant treatment effect among the control and 3 PDM groups in the F0 parents (P < 0.0001) and in their F1 and F2 offspring (P < 0.0001). Fisher's LSD (least significant difference) tests confirmed (P = 0.05) that lifetime fecundity was significantly reduced across all 3 generations after the LP PDM of the F0 flies, while increased across all 3 generations after the IP PDM or for the F0 and F1 generations after the HP PDM. However, lifetime fecundity for the F2 offspring was not increased after the HP PDM of the F0 flies (P > 0.05 for HP vs. CD). (B0–B2) Accumulative differences of egg production for the same 100 flies were plotted between CD and CD (black circle; 0 all the time in the graphs), LP (red diamond), IP (green circle), or HP (purple diamond) treatment over their whole “egg-laying” lives. The total eggs laid by the 100 control flies were 10933, 21081, and 20409 for the F0, F1, and F2 generations, respectively. The accumulative differences for the F1 and F2 generations were therefore normalized to the F0 generation (i.e., x 0.5186 or 10933/21081 for F1 and x 0.5357 or 10933/20409 for F2) for “straightforward” comparisons (but see supplemental Table S3 for details of egg-production results).

We also compared virgin males and virgin females within the CD or each PDM diet across 3 generations (Table 2, first 4 rows) to assess how gender affected longevity and transgenerational inheritance of longevity changes. Within the CD, virgin females lived about 15% longer than virgin males (14% for F0, 13% for F1, and 17% for F2 generations) at the median lifespan. This difference was maintained across the F0–F2 generations after both LP (P < 0.0001 for F0 and F1, P = 0.0005 for F2) and IP (P < 0.0001 for F0 and F2, P = 0.0002 for F1) PDMs of the F0 flies. Consistently, the median lifespan differences were 11–24% within the LP PDM and 11–14% within the IP PDM, suggesting that these PDMs affected males and females similarly. In contrast, such a difference was only maintained for the F0 generation for the HP PDM (P = 0.023, 15%). The F1 (P = 0.14) and F2 (P = 0.07) generations showed no significant difference, suggesting that the HP PDM affected males and females differently. Such analyses suggest that transgenerational inheritance of nutritional programming of longevity may be gender-independent, at least for the LP and IP PDMs, although female flies lived longer than males.

We next compared mated and virgin males within CD or each PDM diet across 3 generations (Table 2, 2nd 4 rows) to assess how mating affects longevity and nutritional programming of longevity changes. Mating did not produce a significant difference in lifespan between the mated and virgin males across F0-F2 generations for the control flies (P ≥ 0.066). Similarly, no significant difference was observed across the F0-F2 generations for the IP (P ≥ 0.081) and the HP (P ≥ 0.061) PDMs, suggesting that nutrition-induced longevity changes may be independent of mating for the IP and HP PDMs. A significant difference was present after the LP PDM between mated and virgin males across the F0–F1 generations (P ≤ 0.0064). However, such a difference was not maintained to the F2 generation (P = 0.062), supporting the notion that transgenerational longevity changes may be independent of mating for the LP PDM. Such analyses together suggest that transgenerational inheritance of nutritional programming of longevity may be independent of mating across all 3 PDMs, although the LP diet may affect the mating behavior of the F0 flies, and the effect appeared to propagate to the F1 generation likely through parental effects.

We finally compared virgin and mated females within CD or each PDM diet across 3 generations (Table 2, the last 4 rows) to assess how mating and reproduction affect longevity and nutritional programming effects of longevity. The mating and egg production within the CD medium did not produce a significant difference between mated and virgin females across the F0–F2 generations (P ≥ 0.078). In contrast, longevity was significantly shortened when comparing mated with virgin females for the LP, IP and HP diets across the F0 and F1 generations (P ≤ 0.023 for all comparisons). Correspondingly, the median lifespan decreased by 6–9 % for the LP PDM, 4–7% for the IP PDM, and 10–14% for the HP PDM. However, no difference was observed in the F2 generation for the LP, IP, and HP PDMs (P ≥ 0.11), suggesting that transgenerational longevity changes may be independent of mating and reproduction in the females. These analyses indicate that transgenerational inheritance of nutritional programming of longevity may be independent of mating and reproduction for all 3 PDMs, although all three PDMs may confound the mating and reproduction behavior of the F0 flies, and the effect appeared to propagate to the F1 generation likely through parental effects.

Taken altogether, our data demonstrate that nutritional programming of longevity may occur in the same F0 flies after PDMs with appropriate diets and then propagate to the F1 offspring through parental effects and further to the F2–F3 generations through transgenerational inheritance. The effects were very striking and appeared to be independent of gender, mating, and reproduction, at least for the LP and IP diets.

Transgenerational inheritance of nutritional programming of reproduction

To assay whether the same PDMs may program reproduction of the F0 flies and their F1 and F2 progenies, 100 mated female flies were evenly divided into 10 subgroups, and their egg production was counted throughout their whole “egg-laying” lives (see Figure S1 and Table S3 for detailed procedures). The F0 females showed lower fecundity than their F1 and F2 offspring, as egg-production data collection was initiated when the flies were 11-days-old for the F0 (because of PDMs), while 4-days-old for the F1 and F2 generations.

For the F0 mated females, lifetime fecundity was greatly reduced after the LP PDM, while mildly increased after both IP and HP PDMs (Figure 2A0). The LP PDM appeared to decrease egg production early on (Figure 2B0, red diamond; see also Figure S3A0), and the trend continued as these flies died out faster than the “CD” controls. The IP PDM also decreased egg production early on, but the trend was reversed since these flies started to produce similar eggs per day while surviving better than the CD controls. In contrast, the HP PDM greatly increased egg production early on (Figure 2B0 and S3A0, purple circles), and the differences remained over time. Consequently, the 100 “LP” F0 flies laid 1984 fewer eggs, while the “IP” and “HP” flies 913 and 802 more eggs, as compared to the “CD” F0 flies (Figure 2B0). These observations indicate that the same PDMs programmed reproduction of the F0 parents.

Figure 3. No evidence for trade-off between longevity and reproduction. (A0–A2) Correlation analyses, using total egg production and median lifespan of the 100 flies from the control (CD) and three PDMs, indicated that longevity and reproduction were positively correlated across the F0-F2 generations (r = 0.64, 0.68, and 0.75 for F0, F1, and F2). (B0–B2) Correlation analyses, using total egg production and median lifespan of the 10 subgroups (10 flies per group) within each dietary treatment, again indicated no obvious negative correlation across four dietary treatments and three generations (−0.01 ≤ r ≤ 0.41 for F0, −0.21 ≤ r ≤ 0.50 for F1, and −0.55 ≤ r ≤ −0.03 for F2).

Similar to that of the F0 generation, lifetime fecundity of the F1 offspring (Figure 2A1) was significantly reduced after the LP PDM, while mildly improved after the IP and HP PDMs of the F0 parents. As we observed for the F0 generation, the “LP” F1 flies produced fewer eggs early on, and the trend continued throughout their whole “egg-laying” lifetime; The “IP” F1 flies produced fewer eggs early on, but the trend was reversed over time; in contrast, the “HP” F1 flies showed increased egg production early on, and the differences remained over time (Figure 2B1 and S3A1). Accumulatively, the 100 F1 “LP” flies laid 1218 fewer eggs, while the F1 “IP” and “HP” flies 533 and 422 more eggs (after “normalization”, see Figure 2 legend for details). These observations indicate that the nutritional programming of reproduction propagated from the F0 flies to the F1 progeny through parental effects, as the F1 flies have never been exposed to the LP, IP, or HP diets.

Similar to that of the F0 and F1 generations, lifetime fecundity of the F2 offspring (Figure 2A2) was significantly reduced after the LP PDM of the F0 flies, while increased after the IP PDM of F0 flies. However, lifetime fecundity for the F2 offspring was not increased after the HP PDM of the F0 flies (P>0.05 for HP vs. CD). Again as for the F0 and F1 generations, we found that the “LP” F2 flies produced fewer eggs early on, and the trend continued throughout their whole “egg-laying” lifetimes; and the “IP” F1 flies produced fewer eggs early on, but the trend was reversed. In contrast, “HP” F1 flies showed increased egg production early on, and the differences remained over time (Figure 2B2 and S3A2). Accumulatively, the 100 F2 “LP” flies laid 1040 fewer eggs, while the F2 “IP” and “HP” flies 509 and 378 more eggs (after “normalization”). Such observations demonstrate that the nutritional programming of reproduction may be transmitted to the F2 offspring through transgenerational inheritance, as both the F1 and F2 flies have never been exposed to LP, IP or HP diets.

The differences in lifetime fecundity and total egg production between the “CD” and various “PDM” flies across the F0–F2 generations suggest that the nutritional programming of reproduction (1) was induced after PDMs of the F0 flies; (2) might be transmitted from the F0 to F1 generation through parental effects; and (3) further transmitted at least to F2 offspring through transgenerational inheritance.

Discussion

Our observations demonstrate that the nutritional programming of longevity may occur in the same F0 flies after PDMs with multiple diets and then propagate to the F1 offspring through parental effects and further to the F2–F3 offspring through transgenerational inheritance. The programming effects were obvious and appeared to be independent of gender, mating, and reproduction (at least for the LP and IP diets). Interestingly, no clear evidence was observed for any potential trade-off between longevity and reproduction, suggestive of the feasibility of improving both longevity and reproduction with proper nutrition across generations. Altogether, these findings establish the first animal model to study transgenerational nutritional programming of longevity through early-life experience, and would facilitate investigations to identify the underlying molecular and epigenetic mechanisms and eventually translational research to combat the burden of NCDs and extend human health and longevity through optimizing the early-life nutritional environment.

Postnatal nutritional programming has been much less studied, as compared to prenatal or fetal programming. Adjustment of litter size (e.g., large litter for undernourishment or small litter for overnourishment), modification of milk formula (e.g., carbohydrate-enriched), and lactational environment in the dam (e.g., maternal calories or protein restriction, maternal diabetes or obesity) have been used for altered nutritional experiences in the immediate postnatal life of rodents [15, 32, 33]. Many of the programming effects appear to be long-lasting and persist into adulthood, but transgenerational inheritance has not been formally explored in this context (cf. [15, 32]). Nevertheless, epigenetic regulations have been revealed for some postnatal nutritional treatments, such as early postnatal exposure to the high-carbohydrate or high-fat diets [34, 35], suggestive of involvement of transgenerational inheritance. Consistently, certain aspects (i.e., increased glucose and trehalose levels in larval stages) of the programmed metabolic status after a 7-day post-eclosion feeding of an HS diet to the female flies persist and propagate through the F2 generations, demonstrating that Drosophila display transgenerational inheritance of metabolic state after altered immediate post-eclosion nutritional experience [20].

It is worth noting that the above-mentioned transgenerational effect of metabolic status to the F2 generation represents a true cross-generation transmission, rather than a direct consequence of the original HS dietary feeding [20]. Here, virgin female flies (F0), after being subjected to an HS diet for 7 days, are transferred to a low-sugar (LS) control diet for mating with untreated male flies, and the F1 virgin females are then cultured on the LS food for 7 days prior to mating to generate the F2 population which are also raised on the LS diet. The germ cells (future gametes) from the F0 females therefore would also be directly exposed to the HS dietary influence, while those from the F1 females would not. As the F2 flies are generated from F1 germ cells never exposed to the HS diet, any displayed programming effects must result from transgenerational inheritance or transmission.

Such a paradigm was thus adapted for our study, with a few important modifications that may potentially improve the transgenerational inheritance effects. First, virgin flies (including both females and males) were collected within 4 hours after eclosion — in fact usually 2-3 hours for most virgin collections; each collection was quickly divided into 4 equal groups and then placed immediately on the LP, IP, HP, or CD for 7 days. Instead, virgin females were subjected to the HS diet within 24 hours of eclosion previously [20]. As in most mammalian species, neural and organ development in Drosophila is not complete at eclosion (birth) and continues in days (the immediate post-eclosion period). Fiber number in the adult mushroom bodies (MB; sensory integration centers of the insect brain), for example, markedly increases during the first week after eclosion [36]. Glial cell outgrowth also appears at this time, with its resulting processes forming a mesh-like network inside the MB *03b1*β lobes within 10 days of eclosion [37]. The lamina grows in volume during the first days after eclosion, and it grows more in the light than in darkness. The lamina is similarly small when flies are kept in the dark only for the first 12 hours of their adult lives or raised for 4 days in constant darkness, highlighting that the first day of adulthood appears to be particularly critical for development [37, 38]. Furthermore, epigenetic programming continues with significant dynamism across the early post-eclosion period (cf. [39]). With the consideration that greatest sensitivity may occur during the period of most rapid growth and maturation, a dietary treatment delivered within the shortest possible time after eclosion may induce the maximal long-lasting programming effects. Developmental plasticity during the immediate post-eclosion period then affords the offspring the adaptive ability to respond to an altered nutritional environment [38, 40, 41].

The second important modification was that both newly-born males and females were subjected to the immediate PDMs, instead of only the females as in Buescher et al. [20], or either males or females as in most rodent studies (reviewed in [15]). The majority of rodent studies have focused on the maternal transmission to demonstrate transgenerational programming effects of early-life nutritional experience [15]. Although much less studied, the transgenerational inheritance effects through the paternal transmission have recently been established for paternal over-nutrition, undernutrition, high-fat diet, and low-protein diet [42-46]. In particular, transgenerational glucose intolerance, induced by intrauterine hyperglycemia or by in utero undernutrition in mice, may be transmitted via either the maternal or paternal line while potentially through different mechanisms [45, 46], suggesting that the transgenerational nutritional programming effects may potentially be additive when induced in both father and mother. Therefore, our design would allow us to induce potentially the largest possible alternation of longevity and reproduction across generations, with potential contributions from both treated males and females. The limitation was in not being able to distinguish potentially different contributions from males and females, something that we did not intend to address in this study.

The third important consideration was that several food diets from other literature were used, with recipes slightly modified to produce isocaloric food media at various protein-carbohydrate contents different from that of CD, a food medium routinely used in our laboratory (see methods and Supplemental Table S1 for details). Our main considerations were that (1) these diets are either widely used across the fly community or used elsewhere to examine cross-generation effects on fly behavior [27, 28], (2) they contain a wide range of protein and carbohydrate contents, very different from CD, and (3) our flies have been adapted to the CD food over hundreds of generations, justifying the use of CD as the control diet. In contrast, using the LS and HS diets may cause several complications. These two diets are semi-defined food media, very different from our CD or any other routinely-used medium. Considering that chemically-defined food media affect longevity differently from routinely used ones across a wide range of protein-carbohydrate contents [29, 30], combined use of the LS and HS diets with our CD may be problematic. Importantly, the HS diet has a protein-carbohydrate content close to that of CD ([20], Supplemental Table S1), and the caloric value is dramatically different for the LS and HS diets [20]. These two semi-defined food media were therefore not used in our study.

Using this modified PDM paradigm, longevity was assayed with the PDM-treated F0 flies and their F1–F3 offspring (Figure 1). The PDM with the LP diet shortened longevity of all 4 types of flies (virgin males and females, mated males and females) across the treated F0 flies to the F3 generation. The PDM with the IP diet extended longevity of all 4 types of flies across the treated F0 flies to the F3 generation. The PDM with the HP diet shortened longevity of the virgin and mated females (while not males) across the treated F0 flies to the F2 generation (but not the F3 offspring). As discussed above, the nutritional programming effects, induced in the PDM-treated F0 flies and propagated to the F1 generation, would reflect parental effects, as the germ cells from the F0 flies were under influence of these PDMs; the effects propagated from the F1 to F2 generation represented a transgenerational transmission or inheritance, as the germ cells from the F1 offspring were not exposed to LP, IP, or HP diets. Taken together, our data demonstrate that the nutritional programming of longevity occurred in the same F0 flies after PDMs with several diets and was then propagated to the F1 offspring through parental effects and further to the F2–F3 offspring through transgenerational inheritance (Figure 1).

Interestingly, the transgenerational programming effect was similar among all 4 types of flies within each PDM for the LP and IP diets. Further data analysis confirmed that the transgenerational effect to the F2 generation was independent of gender, mating, and egg reproduction after PDMs of the F0 flies with the LP and IP diets (Table 2). Our PDM paradigm, with both newly-born males and females receiving the same PDMs simultaneously before mating, prevented us from distinguishing the potentially different contributions from the treated males and females to their offspring, and thus determining the existence of sexual dimorphism in the transgenerational inheritance of nutrition-induced longevity programming for the LP and IP diets. In contrast, the PDM with the HP diet induced the transgenerational programming effect only in virgin and mated females, but not in virgin and mated males of the F2 offspring, supporting the existence of sexual dimorphism, at least for this particular HP diet. Consistently, sexual dimorphism is present in the transgenerational inheritance of certain metabolic syndromes in rodent studies [45, 46].

The differences in lifetime fecundity and total egg production between the “CD” and “PDM” flies across the F0–F2 generations revealed the nutritional programming of reproduction and its transgenerational inheritance at least to the F2 generation. The effects were generally mild for all three diets used (Figure 2). Nevertheless, these data allowed us to perform correlation analyses between reproduction and longevity, and thus to conclude that there may not be a trade-off between longevity and reproduction. The no-trade-off observation strengthened the idea that transgenerational nutritional programming of longevity may be independent of mating and reproduction, and supported the feasibility of elevating both longevity and reproduction with proper nutrition across generations.

Our observations about transgenerational nutritional programming of longevity (Figure 1, Table 1), which especially may be independent of gender, mating, and reproduction at least for 2 out of 3 diets used (Figure 1; Table 2), thus establish the first animal model to study this newly-recognized phenomenon [19]. As recently surveyed, there is no such reported study yet on transgenerational inheritance of programmed longevity from early-life nutritional experience or any other developmental programming interventions [15, 19]. Nevertheless, with a strong link between early-life nutrition and the long-term health of an individual and his/her offspring well established (cf. [1, 3, 5, 6, 15]), recent studies have implicated that early-life nutrition may program longevity across generations [19]. Our results have confirmed this implication and, more generally, have established an animal model system for further studies.

This model offers several tractable advantages, in particular to identify and characterize the epigenetic mechanisms underlying the nutrition-mediated cross-generation programming of longevity and NCDs. The relatively short rearing time and lifespan of Drosophila facilitate longevity experiments over multiple generations in a reasonable time scale (this study). In addition, various dietary (e.g., particular nutrient-depleted or enriched, semi- or chemically-defined) manipulations and well-conserved (e.g., insulin/IGF, TOR and sirtuin) signaling pathways have been described and characterized for studies of longevity and aging-related diseases (e.g., obesity, cardiomyopathy and memory disability) in flies [20, 22, 27, 29, 30, 47-55]. These have been necessary and critical for rapid identification and characterization of any epigenetic mechanisms. Practical dissection of various tissues and diverse choices of genetic manipulations may also be readily applied to explore the relationship among diet, corresponding disease and underlying epigenetic mechanism (cf. [22, 47, 56]). Furthermore, major epigenetic mechanisms (e.g., DNA methylation, histone modifications and non-coding RNA interference) are present in the model system [57], with clear evidence for histone modifications (e.g., methylation, acetylation and biotinylation; [47, 58-61]) and at least 2 microRNAs [62, 63] participating in the regulation of longevity. However, there is no report yet implicating DNA methylation in longevity regulation, because it was long believed to be absent in adult flies (cf. [64]). Importantly, recent studies have reported convincing support for the existence of low-level of DNA methylation in adult Drosophila [65, 66], and an active yet-to-be-identified DNA methyltransferase, with sequence specificity confirmed by the presence of asymmetric methylation at corresponding sites in the genomic DNA [66]. Considering that transgenerational programming through DNA methylation has been documented in rodents (cf. [15, 42, 67], it would be likely that early life dietary manipulations may program longevity by influencing DNA methylation in flies. Finally, our study and the earlier one, using the same 7-day post-eclosion dietary treatment to demonstrate nutrition-mediated metabolic programming up to the F2 generation [20], support the use of post-eclosion adult stage to assay the epigenetic mechanisms underlying transgenerational programming of longevity or aging-related diseases in Drosophila.

We expect that our model system would facilitate studies to identify the underlying molecular and epigenetic mechanisms and eventually translational research to combat the burden of NCDs and extend the human health and longevity through optimizing the early-life nutritional environment. Such longevity studies may also reveal a common mechanism for preventing many NCDs. In fact, the link between early-life nutrition and adult health and disease has gradually been recognized as a cornerstone of public health nutrition programs globally. The World Health Organization recently published global targets and a comprehensive implementation plan for the nutrition of mothers, infants, and young children, aiming to alleviate the double burden of malnutrition in children, starting from the earliest stages of development [11]. Noticeably, Drosophila may also be easily adapted to model various exposures such as stress or environmental contaminants/toxicants [27, 68] with similar treatment paradigms and thus to help address the growing issue of food security related to industrialization and globalization [69, 70].

Methods

Supplementary Materials

Supplementary Figures and Tables