Systems genetics analysis of body weight and energy metabolism traits in Drosophila melanogaster

Abstract

Background: Obesity and phenotypic traits associated with this condition exhibit significant heritability in natural populations of most organisms. While a number of genes and genetic pathways have been implicated to play a role in obesity associated traits, the genetic architecture that underlies the natural variation in these traits is largely unknown. Here, we used 40 wild-derived inbred lines of Drosophila melanogaster to quantify genetic variation in body weight, the content of three major metabolites (glycogen, triacylglycerol, and glycerol) associated with obesity, and metabolic rate in young flies. We chose these lines because they were previously screened for variation in whole-genome transcript abundance and in several adult life-history traits, including longevity, resistance to starvation stress, chill-coma recovery, mating behavior, and competitive fitness. This enabled us not only to identify candidate genes and transcriptional networks that might explain variation for energy metabolism traits, but also to investigate the genetic interrelationships among energy metabolism, behavioral, and life-history traits that have evolved in natural populations.

Results: We found significant genetically based variation in all traits. Using a genome-wide association screen for single feature polymorphisms and quantitative trait transcripts, we identified 337, 211, 237, 553, and 152 novel candidate genes associated with body weight, glycogen content, triacylglycerol storage, glycerol levels, and metabolic rate, respectively. Weighted gene co-expression analyses grouped transcripts associated with each trait in significant modules of co-expressed genes and we interpreted these modules in terms of their gene enrichment based on Gene Ontology analysis. Comparison of gene co-expression modules for traits in this study with previously determined modules for life-history traits identified significant modular pleiotropy between glycogen content, body weight, competitive fitness, and starvation resistance.

Conclusions: Combining a large phenotypic dataset with information on variation in genome wide transcriptional profiles has provided insight into the complex genetic architecture underlying natural variation in traits that have been associated with obesity. Our findings suggest that understanding the maintenance of genetic variation in metabolic traits in natural populations may require that we understand more fully the degree to which these traits are genetically correlated with other traits, especially those directly affecting fitness.

Figure 1

Variation in body weight and energy metabolism traits in D. melanogaster. Distribution of trait means among 40 wild-derived inbred lines of D. melanogaster. Data represent means ± SEM for n = 10 independent replicates. The pink and blue bars in panels A-E depict females and males, respectively.

Figure 2

Metabolites with significantly different levels in P[GT1] and PiggyBac transposon insertional mutations as compared to control strains. Data represent least square means ± SEM of GLY (panel A), TAG (panel B), and GLYC (panel C) calculated using total protein content as a covariate in the analysis averaged across sexes (n = 20 independent replicates). Black and white bars represent mutant and control flies, respectively.

Figure 3

Modules of correlated transcripts associated with variation in body weight and energy metabolism traits. (A) Heat map of correlated probe sets after module formation for BW (13 modules), GLY (9 modules), TAG (5 modules), GLYC (13 modules), and MR (6 modules). Each point represents the correlation between two genes. The color scale bar indicates the value of the correlation. (B) Interaction network of correlated (|r| ≥ 0.7) transcripts for BW module 10. Each node represents a gene and each edge a significant correlation between a pair of genes. (C) Distribution of tissue-specific expression of transcripts in BW module 10 based on data from FlyAtlas http://www.flyatlas.org/[49]. (D) Interaction network of correlated (|r| ≥ 0.6) transcripts for GLY module 7. (E) Network of correlated (|r| ≥ 0.7) transcripts for TAG module 4. (F) Interaction network of correlated (|r| ≥ 0.9) transcripts for GLYC module 11. (G) Distribution of tissue-specific expression of all transcripts associated with MR. Nodes showed as pink in the interaction networks represent those candidate genes for which homozygous mutants were tested. The visualization of interaction networks was performed using Cytoscape 2.6.3. [84].