Nonlinear fitness consequences of variation in expression level of a eukaryotic gene

Abstract

Levels of gene expression show considerable variation in eukaryotes, but no fine-scale maps have been made of the fitness consequences of such variation in controlled genetic backgrounds and environments. To address this, we assayed fitness at many levels of up- and down-regulated expression of a single essential gene, LCB2, involved in sphingolipid synthesis in budding yeast Saccharomyces cerevisiae. Reduced LCB2 expression rapidly decreases cellular fitness, yet increased expression has little effect. The wild-type expression level is therefore perched on the edge of a nonlinear fitness cliff. LCB2 is upregulated when cells are exposed to osmotic stress; consistent with this, the entire fitness curve is shifted upward to higher expression under osmotic stress, illustrating the selective force behind gene regulation. Expression levels of LCB2 are lower in wild yeast strains than in the experimental lab strain, suggesting that higher levels in the lab strain may be idiosyncratic. Reports indicate that the effect sizes of alleles contributing to variation in complex phenotypes differ among environments and genetic backgrounds; our results suggest that such differences may be explained as simple shifts in the position of nonlinear fitness curves.

Fig. 1.

The cellular sphingolipid synthesis pathway and corresponding responses to changes in LCB2 expression. On left, the sphingolipid biosynthesis pathway in Saccharomyces cerevisiae is shown. Lcb2, the product of the titrated gene LCB2, catalyzes the first committed step in this pathway. Sphingolipid compounds were measured in response to changing levels of LCB2, and sphingolipids that displayed a significant response are shown to the right of their corresponding step in the pathway. All comparisons shown between sphingolipid levels in cells with wild-type LCB2 expression and with reduced (−2.7-fold) expression are significant according to a two tailed t test at a P value less than 0.05. Bars indicate medians and standard errors of nine biological replicates. Means, standard errors, and P values for all measured sphingolipid compounds at three LCB2 levels are provided in supplementary table S1, Supplementary Material online.

Fig. 2.

Titrated levels of gene expression reveal a fitness cliff for the gene LCB2 in Saccharomyces cerevisiae strain TetO₇–LCB2. We measured the relationship between doxycycline concentration and LCB2 expression level in (A) standard media and (B) osmotic stress media. Expression levels were determined by qPCR, and each curve was fitted with a five-parameter log-logistic regression. Box and whisker plots (gray bars and dashed lines) indicate error estimates for qPCR measurements from at least three biological replicates, and capped solid lines indicate error estimates based on a regression model. (C) We assessed cellular fitness relative to levels of LCB2 expression for TetO₇–LCB2 grown in two environments. The sharp fitness cost for decreases in LCB2 expression shows that the wild-type level of expression is on the edge of a fitness cliff. The shape of the fitness function in osmotic stress media (dashed line) is similar to its shape in standard media (solid line), although shifted toward higher levels of LCB2 expression in the former. Gene expression is relative to the level of expression in the parental wild-type (R1158) in standard media; arrows indicate wild-type expression levels in each environment. Fitness is normalized to the wild-type equivalent in each environment. Semitransparent boxes indicate error boundaries for fitness (standard error of at least three, and on average five, biological replicates) and for expression estimates (regression model-based error).

Fig. 3.

LCB2 and LCB1 gene expression levels covary among wild strains of Saccharomyces cerevisiae and between growth environments. LCB2 gene expression is upregulated in osmotic stress media (blue points) in comparison with levels of expression in standard media (orange points). LCB2 levels are significantly lower in wild strains YJM145 and RM11-1a than in the auxotrophic lab strain R1158, suggesting that the expression–fitness function has evolved in the lab strain (asterisk; two-tailed t test, P = 0.028 and P = 0.047, respectively). Fold change was ascertained by qPCR and is relative to the expression level of the same gene in the lab strain (R1158) when grown in standard media. Standard errors of three or more biological replicates are shown.