We sequenced pooled RNA samples from a three-diet by three-tissue by six-replicate experiment of outbred mated female D. melanogaster in the DSPR. Our aim was to understand diet-induced patterns of gene expression influencing plastic nutrient allocation in different diet conditions in a multiparent population resource. Our results suggest that: 1) global expression patterns are dominated by tissue and diet-tissue interaction effects, while the effects of diet alone are subtle but significant, 2) patterns of gene expression are generally similar in low-protein and carbohydrate-rich diets relative to the control diet, 3) multiple pathways, co-expressed gene modules, and biological processes are invoked that affect transcription in different diet conditions, especially in the head and body tissues, and 4) expression results did not suggest a single regulatory variant underlying QTL, but narrowed down to a few possible causal genes. Overall, our results suggest that multiple networks are involved in phenotypic changes in response to nutrient availability, rather than just a few key genes. We advocate a broader, genome-wide approach to studying the genetic mechanisms underlying diet effects on phenotypic change.
It is established that nutrient signaling pathways and hub genes in those pathways play a crucial role in how organisms adjust to changing conditions in availability and quality of nutrients to optimize fitness traits [50–52]. Analysis of differential gene expression in a population presented with diets differing in nutritional richness provides an ‘omic’ alternative to study intermediate processes that connect genetic architectures to phenotypic outcomes such as allocation patterns. In fruit flies, studies typically focus on whole-body or head tissue transcription (e.g. [64]); one or a few gene pathways known to affect diet responses at a time (e.g. [63]); one or two diet manipulations (e.g. [64–66]), but scarcely integrate over more than two organs and conditions at a time, or explore expression outside a few gene pathways. Further, despite costs trending downwards recently, sequencing of more highly replicated experiments remains unaffordable for many laboratories. By and large, studies in model organisms focus on genes in a few endocrine pathways, so called nutrient sensors, as critical players in coordinating growth, reproduction, stress resistance and somatic maintenance responses to changing diet conditions. Components of the IIS/TOR, growth, and ecdysone hormones; and sirtuin deacetylases are deemed some of the major players in this respect. Our results suggest an expanded scope of mechanisms underlying flexible responses to nutrient limitation (DR studies) or oversupply (high sugar and high fat diet studies) in natural populations, which has also been suggested by [67–69] and reviewed by [58, 70]. We discuss our major results and their implications below.
First, we observed a large global effect of tissue type and a more subtle, but significant, effect of diet treatment. Previous studies in flies have also found relatively small effects of diet on transcription. Previously, we characterized the genetic basis of standing genetic variation for 55 genes of the IIS/TOR pathway following treatment with the same diets used here [63], and found only small changes in gene expression associated with diet treatment although most of those genes were differentially expressed. Similarly, Reed et al, [71] measured transcriptional and metabolomic changes for 20 inbred lines (North Carolina and Maine population) of D. melanogaster treated with four diets varying in sugar and fat content and observed a small dietary component in gene expression profiles, with much larger contributions of genotype by diet interactions. Musselman et al. [72] investigated expression differences in D. melanogaster fed with two different forms of sugar and found small but significant changes in gene expression. Overall, diet seems to produce fairly small magnitude changes in expression in many genes across the genome, which in concert presumably can lead to large phenotypic changes.
Secondly, comparisons between DR and HS diets relative to C revealed a similar pattern of expression. This result mirrors our earlier finding in an eQTL mapping experiment using DSPR lines in which gene expression in DR and HS relative to C generally trended in the same direction [63]. This result is in spite of the fact that the DR and HS diets lead to very different outcomes in median lifespan and fecundity in our population [63, 73]. Nutritional geometry studies which measure traits in a series of concentrations of liquid media suggest that traits such as lifespan and reproduction (which differ significantly across our diets) are influenced primarily by the diet protein to carbon (P:C) ratio, not its caloric content (e.g., [8, 74, 75]. Thus, calorie limitation alone does not drive phenotypic patterns in these studies. While our HS diet has a high concentration of sugar per liter of food, the P:C ratio is a lot lower (i.e. ~1:2 yeast to sugar ratio). Lifespan and fecundity in nutritional geometry studies are maximized at a much higher P:C ratio (i.e. 2- to 4-fold higher than our HS diet) [74, 75]. Thus, our results are consistent with similarity in expression pattern between DR and HS. It is also possible that diet macromolecules serve only as a cue that lead to optimal allocation of resources in natural populations. Thus, it is possible for expression measures to trend in the same direction in DR and HS treatments. Further, Dobson et al. [76] found that excess sugar diets in young adult flies inhibited foxo and reduced survival in middle and old age. While we did not measure lifespan and fecundity here, our data showed mixed pattern of foxo activation across diet tissue combinations, and these mapped to many coexpression modules which may be due to the relatively young age (i.e. 21 days) of our experimental flies.
Thirdly, an overall take away from our GSEA on the full dataset, and also on gene modules coming from hierarchical clustering was that diet effects could not be attributed to a particular genetic mechanism. GSEA highlighted metabolism, oxidative phosphorylation and protein processing at pathway level, but showed a broad spectrum of processes in GO term enrichment encompassing metabolism, cell signaling, structural development and organization, and defense. Similarly, GO analysis of gene modules yielded a broad range of biological processes. From both these analyses, IIS, TOR, and FOXO signaling were not significantly enriched. However, several genes had significantly reduced expression in the IIS/TOR (e.g. Ilp5, Rheb, Atg1, Myc, and eIF4E1) and significantly higher expression in FoxO downstream effectors (e.g. AMPK, orct2, Gadd45, cdk2 and p38) in most DR-tissue combinations consistent with indicating canonical activity. With the exception of Ilp5 [41] however, undetectable differential expression of hub genes (such as Ilp2, S6K, chico, InR, Akt1, Torc, Thor, and foxo) suggest that diet induced effects may involve many more pathways/genes than have been traditionally studied in this context.
Evidence in C. elegans suggest that the worm ortholog of foxo, daf–16 is not required in the DR response [77–80]. On the other hand, sir–2.1 isa worm ortholog of the fruit fly Sirt 2 (which was significantly differentially expressed in this study), and is required for lifespan extension in adult worms by diet deprivation was independent of the daf–2/insulin-like signaling [80].In D. melanogaster, similar evidence is emerging that suggest that foxo is not required for the response to DR [81], but is involved in the normal response to DR [82]. When dFOXO was removed, DR treatment still resulted in significant lifespan extension in null flies [81]. Another study testing a novel DR assay in C. elegans found that DAF–16, but not DAF–2 (the worm lnR) was required when DR was performed on solid media, and concluded that AMPK-FOXO signaling resulted in lifespan extension on solid food [83]. Our data provides further evidence supporting these studies in suggesting a broader mechanism in which IIS, TOR and FOXO play a role, but in concert with other pathways.
A potential limitation of our study is the heterogeneity in tissue types present in our samples, which may affect the level and nature of gene expression [84]. For instance, assuming fewer cell types are available in the ovaries or head samples compared to body samples, we could expect the range of biological processes triggered by nutrient levels to scale with cell types to some degree. Further, DR and HS protocols vary tremendously across laboratories which can result in different studies detecting only subsets of the gene network distribution which responds to nutrient change [81]. Patterns of phenotypic expression (fecundity and lifespan) and patterns of gene expression have held stable across several studies using the same set of diets in our population, suggesting that the effects we find are biologically relevant in the DSPR. However, in addition to a broader view of the potential mechanisms causing phenotypic changes in response to diet, a broader consideration of different diets would also benefit the field.