To obtain a reasonably comprehensive profile of lipids in the Drosophila soma, we examined 10 lipid classes: cardiolipin, cholesterol ester, diacylglycerol, free fatty acid, lysophosphatidylcholine, phosphatidylcholine, phosphatidylethanolamine, phosphatidylserine, sphingomyelin, and triacylglycerol by mass spectrometry (Lipomics Technologies, Sacramento CA). We made lipid determinations on mated sexed adult flies of the genotype tud1 bw1 sp1/CyO
at 5-7 days after eclosion. To eliminate direct germline contributions to the lipid profiles, we removed the gonads prior to extraction. This also results in the loss of hemolymph and therefore most of the circulating lipids. To determine if lipid profiles differed due to indirect effects of the germline on somatic physiology, we examined flies from homozgyous tud1
mothers. The progeny of homozygous tud1
mothers do not form a germline, while progeny of heterozygous tud1
mothers have a fully functional germline. This allowed us to examine the effect of the germline on somas with the same zygotic genotype. This is one of the same maternal/zygotic genotypes we previously described for expression profiling [8
]. Flies were grown on a standard rich cornmeal/sugar/yeast/agar media (<https://stockcenter.ucsd.edu/info/food_cornmeal.php
>, Drosophila Species Stock Center, Tucson AZ); at 22°C; with 60% relative humidity; under constant light. We obtained lipid profiles from 8 samples, 4 from each sex, further stratified by germline status (Additional File 1
). Note that statistical power was strongest for overall lipid profiles in adult flies where sample size was 8 and weakest for germline status within sex where sample size was 2. Because of the limited differences in lipid levels observed, collapsing germline classes to increase power was statistically justified by homogeneity.
To compare all of the fatty acids profiled, we plotted the data for individual lipid species within the ten major classes by germline state and by sex. We found remarkably little difference in lipids between flies with or without a germline (r > 0.97; Figure ) and only slightly increased scatter between the sexes (r = 0.96; Figure ). We did observe a few data points outside the 95% confidence interval limits, but outliers are expected among the 400 lipids measured. However, this exploratory analysis showed that all the outliers are members of the sphingomyelin, cholesterol ester, and lysophosphatidylcholine classes, suggesting that the outliers represent more than random measurement error. Our suspicions were raised further by the pattern of saturation states among the outliers. The species higher in males were saturated 16 or 18 carbon fatty acids (16:0 or 18:0) and the species higher in females were monounsaturated 18 carbon fatty acids (18:1n7) and polyunsaturated fatty acids (18:3n6, 20:3n3, or 20:3n9). These data suggested that saturation status within a lipid class might be sex-biased.
Figure 1 Relationships of fatty acids between germline states and between sexes. (A, B) Within sex comparisons between flies with a germline or with no germline. (C, D) between sex comparisons of flies with a germline or with no germline. Each data point represents (more ...)
To test this hypothesis, we grouped the lipids by class. By abundance, the storage molecule triacylglycerol was the dominant class of lipid in the adult soma, followed by the lipid bilayer components lysophosphatidylcholine, phosphatidylcholine, and phosphatidylethanolamine. As suggested by plotting the abundances of the individual lipids, there were no significant differences in the abundances of the 10 major classes in non-gonadal soma between the sexes, or in flies with or without a germline (Figure ).
Figure 2 Distribution of fatty acids in the sexes with a germline or with no germline. (A) Fatty acids in nmoles per gram tissue by lipid class. Varied fatty acid components of lipids can be grouped into three categories: (B) saturated fatty acids (SFA); (C) monounsaturated (more ...)
We then binned lipid classes into saturated, monounsaturated, and polyunsaturated fatty acids. Again, we observed no significant differences between the flies with or without a germline within each sex, but we did observe sex-bias in the saturation states of cholesterol esters and lysophosphatidylcholines (Figure ). Since we observed no significant differences due to germline status (p > 0.05, t-test), we treated these within-sex samples as an additional level of replication in order to increase the power of statistical tests for the differences in lipid saturation between sexes. As suggested by the initial exploratory analysis, we observed significantly higher saturated cholesterol ester and lysophosphatidylcholine levels in males (p < 0.005, t-test) and an increase in polyunsaturated and/or monounsaturated cholesterol ester and lysophosphatidylcholine levels in females (p < 0.05, t-test). Given that lecithin:cholesterol acyltransferase transfers fatty acids from phosphatidylcholine to form cholesterol ester and lysophosphatidylcholine, these differences in saturation states may be linked.
We were interested in further examining the relationships between the different lipids to determine if particular lipids co-vary, or cluster, among the samples. This type of analysis is particularly useful with limited sample sizes as there are many more measurable relationships between lipid species than between samples. We used nonnegative matrix factorization (NMF), an unsupervised, parts-based learning paradigm, to explore these relationships [28
]. The fatty acid profiling data was input as a matrix with cells representing the 40 combinations of samples and lipid classes that is decomposed into weight and pattern via a multiplicative updates algorithm [28
] to estimate that there were 4 meta-fatty acid clusters (not shown). On the basis of k = 4, we generated four consensus clusters of the 40 combinations of samples and lipid classes (Figure ). These distinct clusters are associated with specific biochemical functions. We observed only a single lipid class that mapped to different positions in the matrix as a result of sex. In males, lysophosphatidylcholine clustered with the sphingomyelins, while in females lysophosphatidylcholine clustered with the other membrane lipids. These data provide additional evidence that there is sexual dimorphism for this specific lipid class. However, the most striking finding is that the relationships between lipids are quite similar between the sexes and between flies with or without a germline.
Figure 3 A heatmap of sample-lipid class combinations. Given the matrix partition factor k = 4, we ran the NMF algorithm 100 times to generate consensus clustering of the combinations of samples and lipid classes. Each run resulted in a 40 × 40 connectivity (more ...)