Genetic basis of metabolome variation in yeast

Abstract

Metabolism, the conversion of nutrients into usable energy and biochemical building blocks, is an essential feature of all cells. The genetic factors responsible for inter-individual metabolic variability remain poorly understood. To investigate genetic causes of metabolome variation, we measured the concentrations of 74 metabolites across ~ 100 segregants from a Saccharomyces cerevisiae cross by liquid chromatography-tandem mass spectrometry. We found 52 quantitative trait loci for 34 metabolites. These included linkages due to overt changes in metabolic genes, e.g., linking pyrimidine intermediates to the deletion of ura3. They also included linkages not directly related to metabolic enzymes, such as those for five central carbon metabolites to ira2, a Ras/PKA pathway regulator, and for the metabolites, S-adenosyl-methionine and S-adenosyl-homocysteine to slt2, a MAP kinase involved in cell wall integrity. The variant of ira2 that elevates metabolite levels also increases glucose uptake and ethanol secretion. These results highlight specific examples of genetic variability, including in genes without prior known metabolic regulatory function, that impact yeast metabolism.

Figure 1. Distribution of significant linkages across the genome.

Metabolite linkages that exceeded the 0.1 FDR significance threshold are plotted based on their most significant marker's genome location (indicated with a dot) with a 95% confidence interval. Continuous vertical lines represent chromosome ends. Numerals are placed at chromosomes' center. Genes investigated in this study are shown at top. mQTLs for ions of unknown identity were combined into a single class.

Figure 2. Similarities between metabolite and transcript linkage distributions.

Significant linkages are binned in 10-eQTL hot spots are colored green. Dotted blue lines show chromosome ends. Red lines show the hot spot cutoff (see Methods for calculation).

Figure 3. Levels of pyrimidine intermediates and products differ based on the ura3 allele inherited.

The relevant portions of the pathway are shown, with measured metabolites in red. The location of ura3 in the pathway is shown in green. The accompanying plots show phenotype distribution of the segregants based only on the allele of ura3 inherited: RM in purple, BY in orange. The ura3 gene is defective in RM. All metabolite levels are log2(Segregant/RM). Compounds that were significantly linked to ura3 locus (via LOD scores) are shown in bold.

Figure 4. Levels of sulfur-assimilation intermediates differ based on the slt2 allele inherited.

The relevant portions of the pathway are shown, with measured metabolites in red. The accompanying plots display the phenotypic distribution of the segregants based only on the allele of slt2 inherited: RM in purple, BY in orange. All metabolite levels are log2(Segregant/RM). Compounds that were significantly linked to slt2/Erc1 locus (via LOD scores) are shown in bold.

Figure 5. RM-inheriting segregants for slt2 and erc1 show significantly higher levels for SAM.

Intensities (mean standard error) of SAM are plotted based upon the allele of slt2 (top) and slt2 and erc1 (bottom). Mass spectrometer ion counts for BY background (diamonds) and RM background (squares) are shown on the left axis while segregants' log2 relative abundances (triangles) are indicated on the right axis.

Figure 6. Levels of glycolysis, pentose phosphate pathway and TCA intermediates differ based on the ira2 allele inherited.

The relevant portions of the pathway are shown, with measured metabolites in red and significant linkages shown in bold. The accompanying plots show phenotype distribution of the segregants based only on the allele of IRA2 inherited: RM in purple, BY in orange. All metabolite levels are log2(Segregant/RM). LOD score for the closest marker is also shown. *includes analytically indistinguishable isomers.

Figure 7. RM-inheriting segregants for ira2 show significantly lower levels for fructose-1,6-bisphosphate.

Intensities (mean standard error) of FBP are plotted based upon the allele of ira2. Mass spectrometer ion counts for BY background (diamonds) and RM background (squares) are shown on the left axis while segregants' log2 relative abundances (triangles) are indicated on the right axis.

Figure 8. Distribution of broad sense heritability (

) across measured metabolites. each circle represents a single metabolite, colored according to how many QTLs are associated with its abundance. 114 metabolites are shown: 74 known metabolites with 52 detected mQTL and 42 unknown metabolites (with known m/z, but unknown identity) associated with 20 additional mQTLs.

Figure 9. Fraction of broad-sense heritability explained by identified mQTLs.

Each stacked bar represents a single metabolite which was significantly associated with at least one locus. The height of the bar is the broad-sense heritability of the metabolite's abundance, and the coloration partitions this heritability into unexplained heritability (gray), and the effects of each mapped QTL (colors). Three examples are given to demonstrate the variable effect sizes observed across metabolites. The distribution of metabolite abundances for a genotype is shown as a violin plot, and a 95% confidence interval for the median of each genotype is reported with error bars. This confidence interval was determined using a percentile bootstrapping method .