Mutations in global regulators lead to metabolic selection during adaptation to complex environments

Abstract

Adaptation to ecologically complex environments can provide insights into the evolutionary dynamics and functional constraints encountered by organisms during natural selection. Adaptation to a new environment with abundant and varied resources can be difficult to achieve by small incremental changes if many mutations are required to achieve even modest gains in fitness. Since changing complex environments are quite common in nature, we investigated how such an epistatic bottleneck can be avoided to allow rapid adaptation. We show that adaptive mutations arise repeatedly in independently evolved populations in the context of greatly increased genetic and phenotypic diversity. We go on to show that weak selection requiring substantial metabolic reprogramming can be readily achieved by mutations in the global response regulator arcA and the stress response regulator rpoS. We identified 46 unique single-nucleotide variants of arcA and 18 mutations in rpoS, nine of which resulted in stop codons or large deletions, suggesting that subtle modulations of ArcA function and knockouts of rpoS are largely responsible for the metabolic shifts leading to adaptation. These mutations allow a higher order metabolic selection that eliminates epistatic bottlenecks, which could occur when many changes would be required. Proteomic and carbohydrate analysis of adapting E. coli populations revealed an up-regulation of enzymes associated with the TCA cycle and amino acid metabolism, and an increase in the secretion of putrescine. The overall effect of adaptation across populations is to redirect and efficiently utilize uptake and catabolism of abundant amino acids. Concomitantly, there is a pronounced spread of more ecologically limited strains that results from specialization through metabolic erosion. Remarkably, the global regulators arcA and rpoS can provide a "one-step" mechanism of adaptation to a novel environment, which highlights the importance of global resource management as a powerful strategy to adaptation.

Figure 1. Population genomics and population proteomics were used to identify the biochemical basis for phenotypic convergence and parallel evolution during adaptation to novel resource rich environments.

Clones of two species, E. coli RU1 and C. freundii RU2, were isolated de novo from human stool. Single clones were used to inoculate twelve replicated populations and evolved in LB or BHI for 500 or 765 generations respectively by daily transfers into fresh media. Adaptive mutations were identified as mutations that evolved repeatedly across species and media. We performed a mutation accumulation experiment by transferring a single colony of twelve replicated lines for 200 days as control experiment. The relaxed selection allowed the random accumulation of mutations, and makes the parallel evolution of adaptive mutations unlikely.

Figure 2. The underlying genetic diversity of the adapted populations is readily observed as phenotypic diversity, when plated on tetrazolium arabinose plates.

When arabinose is provided as the carbon source, the diversity of genotypes in the adapted populations is seen as a marked increase in phenotypic diversity. Colony size, morphology and the ability to use arabinose (as indicated by the darkness of the colony) varied widely in the adapted populations compared to the ancestor.

Figure 3. The total number of mutations in coding regions was substantial in the evolved populations.

The accumulation of hundreds of mutations is consistent with weak selection, where many mutations can have small consequences for the fitness of organism. Overall, more mutations evolved in LB (A, B) than in BHI (C, D), both in the E. coli (A, C) and the C. freundii (B, D) populations. Two LB- and two BHI-evolved E. coli populations acquired considerably more mutations than the remaining populations evolved in the same media, suggesting that these populations evolved to become mutators. Total number of mutations in coding regions (black) and non-synonymous mutations (red) are shown for each population. Populations BHI5 and BHI20 were not included in the genomic analyses.

Figure 4. Parallel evolution was observed as mutations that evolved in many populations.

Only a few genes acquired mutations consistently across species and media (indicated by the arrow). Importantly, none of these genes acquired mutations in the control experiment, the mutation accumulation lines (MA, (E)). As expected from our conditions of weak selection, most mutations occurred only in a few populations. We analyzed 24 LB-evolved (A and B), 22 BHI-evolved populations (C and D) and 12 mutation accumulation lines as a control experiment (MA) (E). The number of mutations found in only one population was capped at 120 (LB) or 150 (BHI) to better discern the number of genes that evolved in parallel.

Figure 5. Mutations in arcA and rpoS evolved repeatedly both within and among E. coli populations evolved in LB (A) and BHI (C), and C. freundii populations evolved in LB (B) and BHI (D).

The number of mutations observed in the evolved populations is plotted over the number of populations with mutations in that gene. The scatterplots show all the genes with mutations for the twelve populations evolved in one environment (black circles). Selected genes that evolved in parallel are identified and marked individually, following the legend. None of these evolved genes had mutations in the mutation accumulation lines of E. coli evolved on LB agar (E).

Figure 6. Frequencies of arcA and rpoS mutants in the evolved populations arcA mutations (black bars) reached high frequencies in all LB-evolved populations (A, B) and reach fixation in LB5, while rpoS mutations (red) were more common in BHI-evolved populations (C, D).

The frequencies represent the total frequencies of all arcA or respectively, identified in a particular population.

Figure 7. Mutations in arcA evolved repeatedly and with remarkable diversity both within and among populations of E. coli evolved in LB (A) and BHI (C) and C. freundii populations evolved in LB (B) and BHI (D).

Specific mutations to arcA identified in the evolved populations are indicated. The red dots represent the number of populations with that specific mutation (out of twelve LB and eleven BHI populations for each strain). The red star indicates the mutation that was fixed in LB5. No mutations in arcA were identified in the BHI-evolved C. freundii populations. The receiver domain that includes the site of phosphorylation (Asp-54) is indicated in blue and the DNA binding domain in green.

Figure 8. The proteomic analyses shows consistent changes across the evolved populations.

We observed 488 proteins across the twelve evolved and the ancestral populations. Out of those, 166 proteins were significantly different between ancestor and evolved populations as judged by LC-MS based proteomics and then were grouped by function (S1 Text). The heat map shows the change in expression for the groups of proteins that had consistent changes with respect to the ancestor. The complete list of significantly changing proteins is given in S2 Table. Columns represent the BHI-evolved E. coli populations; rows represent individual proteins, grouped by functional categories. The color scale shows the average base-2 logarithm transformed fold change of each protein with respect to the ancestor sample, based on the average of triplicate measurements. Gray indicates that the protein was not detected in that population. Positive values (red) indicate higher expression in the evolved population than the ancestor; negative values (green) represent lower expression in the evolved population than the ancestor. The asterisk indicates the global regulator ArcA. The circles on the top indicate whether we identified mutations in arcA (black) or rpoS (red) in a population (at a minimum frequency of 0.05), and whether the population evolved a mutator phenotype (blue). ‘UND’ stands for undetermined as population BHI5 was excluded from the genomic analyses (see Results).

Figure 9. Global regulators arcA and rpoS provide a comprehensive metabolic shift during adaptation that circumvents epistatic bottlenecks.

Adaptation to the amino acid rich conditions of BHI by E.coli are consistent with a ‘metabolic selection’ that provides a facile strategy for shifting environments. Up-regulated systems (green) are associated with the movement of abundant amino acids into the cell coupled with an increased capacity for catabolic metabolism through the TCA cycle with excess nitrogen being secreted as putrescine. Down-regulated systems (red) include components of the starvation stress response consistent with the maintenance of a new nutrient rich homeostasis.