Nature of epistasis exhibited by two mutations is contingent on the genetic background.

(A) The nature of epistasis exhibited by two mutations (indicated by red circle and green square) in all 47 genetic backgrounds was quantified. Numbers f1 to f6 indicate the fraction of genomes in which the two mutations exhibit Positive epistasis (PE), Negative epistasis (NE), Reciprocal Sign epistasis (RSE), Single Sign epistasis (SSE), Other Sign epistasis (OSE), and No epistasis (No), and were calculated for this pair of mutations. The process was repeated for every possible pair of mutations. Distribution of the six fractions f1 to f6 in (B) high and (C) low fitness backgrounds is plotted.

Different sites on the landscape dictate change in epistasis differently.

(A) To determine the impact of mutation at locus X in changing the nature of epistasis between two mutations (indicated by a red circle and a green square), we took the following approach. For a particular pair of mutations, the six sites (indicated by N) were fixed, and the nature of epistasis recorded before and after a mutation at X. This process was repeated for all 46 backgrounds, and the fraction of genomes in which, upon introduction of a mutation at X, epistasis between the mutations (red circle and green square) changed, was noted. This gives us f, as indicated in the Figure. This process was repeated for all possible pairs of mutations (red circle and green square); giving a distribution of f for locus X. This distribution of f is shown in (B) for high fitness backgrounds and (C) for low fitness backgrounds. Functionally important sites (4 and 5) cause change in epistasis more frequently than others.

Introduction of a synonymous mutation changes the nature of epistasis between two interacting mutations.

(A) For a given sequence, two mutations (red circle and green square) could occur in either one or two codons, resulting in two or one unaltered codons (underlined in figure). All possible synonymous mutations were introduced in the unaltered codon(s) and change of nature of epistasis (if any) between the two mutations was noted. This was done for all backgrounds and all pairs of two mutations. Probability that a synonymous mutation at locus X leads to a change in nature of epistasis between two mutations in (B) high and (C) low fitness backgrounds is shown. Mutations resulting in TAA ↔ TGA and TAA — TAG are considered as synonymous mutations.

The folA fitness landscape exhibits weak patterns of global epistasis.

(A) Cartoon to illustrate statistical patterns of global epistasis. The beneficial effect of a mutation decreases with an increase in background fitness (Diminishing Returns Epistasis). Beyond a certain background fitness (Pivot Point), the mutation becomes deleterious and its deleterious effects increase as the fitness of the genetic background increases (Increasing Costs Epistasis). (B) Mutational effects vs. the background fitness show weak correlation for low fitness (R2 = 0.1236) (left of dotted line) and high fitness (R2 = 0.1284) (right of dotted line) backgrounds. The red lines show the best linear fits for the two groups.

A small fraction of mutations follow global epistasis.

Only 16 (14 of which are at position 4 and 5) (highlighted in red boxes) out of 108 possible mutations exhibit statistically significant patterns (R2 > 0.4) of global epistasis. See Supplement Table 1 for statistics for each mutation.

Most mutations switch from being beneficial to deleterious at the pivot point.

Background fitness (y-axis) at which each of the 108 mutations (x-axis) switches from being beneficial to deleterious. The black solid line shows the average fitness (−0.657) (or pivot) at which a mutation changes from being beneficial to deleterious. Individual bars show the deviation from the mean pivot point for each mutation. More than 80% (86 out of 108) of the mutations exhibit the pivot fitness in the range -0.657 +/-0.0657. The dotted line in red indicates the growth rate used in Papkou et al40 to differentiate between high and low fitness variants.

“Phenotypic DFE” is a predictor of DFE.

(A) Backgrounds with near identical fitness were randomly distributed in two groups comprising 90% and 10% of all backgrounds. The first group was used to define a “phenotypic DFE” (a mean DFE of all genomes in the group). The phenotypic DFE was compared with individual DFEs in the 10% group, and the p-value distribution of this comparison was obtained. The process was repeated for genotypes with other fitness. (B) Phenotypic DFE of high fitness backgrounds exhibited two peaks. The first peak was at fitness effect ∼ 0. The second peak comprised of deleterious mutations, whose magnitude increased with increasing background fitness. (C) Phenotypic DFE of low fitness backgrounds exhibited a single peak, whose mean decreased with increasing background fitness. (D) Percent of DFEs in a fitness window with p-value < 0.05 when compared with the Phenotypic DFE. As fitness increases, Phenotypic DFE becomes a better predictor of DFE of a genotype.