Systems-level response to point mutations in a core metabolic enzyme modulates genotype-phenotype relationship

Abstract

Linking the molecular effects of mutations to fitness is central to understanding evolutionary dynamics. Here, we establish a quantitative relation between the global effect of mutations on the E. coli proteome and bacterial fitness. We created E. coli strains with specific destabilizing mutations in the chromosomal folA gene encoding dihydrofolate reductase (DHFR) and quantified the ensuing changes in the abundances of 2,000+ E. coli proteins in mutant strains using tandem mass tags with subsequent LC-MS/MS. mRNA abundances in the same E. coli strains were also quantified. The proteomic effects of mutations in DHFR are quantitatively linked to phenotype: the SDs of the distributions of logarithms of relative (to WT) protein abundances anticorrelate with bacterial growth rates. Proteomes hierarchically cluster first by media conditions, and within each condition, by the severity of the perturbation to DHFR function. These results highlight the importance of a systems-level layer in the genotype-phenotype relationship.

Figure 1. The effect of mutations, media composition, and DHFR overexpression on bacterial growth

Growth rates of all studied strains under various conditions. Growth rates were determined from the exponential phase of growth using the three parameter fit of ln(OD/OD₀) vs time curves proposed in (Zwietering et al., 1990). Both metabolic (“folA mix”) and functional (pBAD WT DHFR overexpression) complementation brought the growth rates of the mutant strains very close to the WT levels. Error bars represent the standard deviation of three independent growth measurements.

Figure 2. Global statistical properties of proteomes correlate strongly with fitness

A. Distribution of LRPA. (Data shown are for W133V strain. LRPA distributions for all strains and conditions can be found in related Figure S1). B. Same as A for LRMA. C. S.D. of LRPA are directly correlated to the destabilizing effects of mutations. The change (relative to WT DHFR) in folding free energy (ΔΔG) for multiple mutants is calculated by the additive approximation using empirical measurements obtained for single DHFR mutants in (Bershtein et al., 2012). D. S.D. of LRPA is anti-correlated with growth rate. E. S.D. of LRMA is correlated with S.D. of LRPA. The slope is close to 2, suggesting that transcriptomes are more readily perturbed than proteomes. (See related Figures S1 and S2)

Figure 3. Correlation between proteomes and transcriptomes

A. Scatter plots of z-scores of LRPA between proteomes of the wild-type strain grown to various OD levels, and proteomes of mutant strains, and TMP-treated WT strain. Low correlations indicate that perturbations observed in response to DHFR mutations or functional inhibition by TMP is largely unrelated to natural variation rooted in different stages of the growth cycle. B. Scatter plots and correlation coefficients between proteomes of all strains under standard growth conditions (right panel), and in presence of the “folA mix” (left panel). Addition of the “folA mix” minimizes the variation between different proteomes. C. Transcriptomics data obtained for strains grown under standard conditions (identical to A). Correlations are overall higher for mRNA abundances, but similar classes of transcriptomes are discernible. (See related Figures S3 and S4 for correlation in biological repeats).

Figure 4. Proteomes cluster hierarchically according to media conditions and growth rates

A. Ward hierarchical clustering of proteomes. Values of the horizontal axis at split points indicate Ward distances between corresponding clusters (see Supplementary Methods). B. Same as A for null model proteomes generated to correlate according to growth rates. C. Correlation between Ward distance at branch point of a cluster and number of members of the cluster for real proteomes. D. Control: same as C for NMPs.

Figure 5. DHFR protein and mRNA abundances A

Changes in folA mRNA and DHFR protein abundances (relative to WT, in z-score), as detected by the high-throughput techniques for all strains. Expression of both TMP-treated WT and mutant folA genes is upregulated (transcriptome data, blue columns). The abundance of WT DHFR is also elevated with TMP addition. Conversely, all mutant DHFR proteins show a substantial drop in abundances under standard growth conditions (green columns), and upon addition of the “folA mix.” B. Direct measurements of the folA promoter activity using GFP reporter plasmid for all strains and for all conditions studied. Metabolic complementation (addition of the “folA mix”) eliminates the up-regulation of expression of the mutant folA genes, despite the drop in abundances of the mutant DHFR proteins (Fig.4). C. DHFR abundance determined by the western blot correlates with promoter activity but dramatically differs between mutant strains and TMP-treated WT. Variation in abundance measurements by Western Blot (standard deviation obtained in 4 independent experiments) constitutes app. 8%. D. The level of folA promoter activation is strongly anti-correlated with the bacterial growth rate. The color code of strains and conditions is the same as in Figure 1 and Figure 4.

Figure 6. Pleiotropic biological effects of DHFR mutation and DHFR inhibition by trimethoprim

A. Scatter plot showing the relative variation of cumulative average z-scores for transcriptomic changes (LRMA) and proteomic changes (LRPA) for overlapping groups of genes defined in (Sangurdekar et al., 2011). Data points corresponding to groups of genes whose cumulative average LRMA z-scores are positive (up-regulation) are shown in red, while transcriptionally down-regulated groups are shown in blue. The data shown are for the W133V strain (see related Figure S5 for the remaining mutant strains and TMP-treated WT). B. Variation for specific gene groups. Clustering global variation in mRNA and protein abundances as belonging to functional classes (upper panel) (see (Sangurdekar et al., 2011)) or co-regulated by a specific operon (lower panel) reveals the highly statistically significant variation of several functional groups. The color code indicates the direction of change (blue – down-regulation, red – up-regulation) and the color depth codes for the logarithm of p-values against the null model of independent variation within a group of genes (see Supplemental Information).