Extended Data Fig. 5: Features shaping the resistance level during evolution.
From: ESKAPE pathogens rapidly develop resistance against antibiotics in development in vitro
(A) Correlation of the initial MIC and increase in resistance levels during adaptive laboratory evolution for all species. The scatterplot shows the correlation between initial resistance level (MIC of the ancestor) and the increase in MIC (both on log10-scale) during adaptive laboratory evolution (ALE) for the four bacterial species. Increase in MIC was calculated by subtracting the initial MIC from the final MIC. Each point corresponds to an adapted line-antibiotic pair. Spearman’s rank correlation coefficients (calculated using a two-sided test) and corresponding p-values between the two variables across all adapted lines of each species are displayed in the figure. Colours indicate the 4 bacterial species studied. (B) Correlation analysis between the initial MIC and the increase in MIC during adaptive laboratory evolution for all genomic backgrounds. The scatterplots show the initial MIC and the increase in MIC (both on log10-scale) during adaptive laboratory evolution for each 8 studied bacterial strain (indicated in the top of each panel). Increase in MIC was calculated by subtracting the initial MIC from the final MIC. Each point corresponds to an adapted line-antibiotic pair. Spearman’s rank correlation coefficients (calculated using a two-sided test) and corresponding p-values are indicated within each panel. Error bars represent 95% confidence intervals. Colours indicate the 4 bacterial species studied. For abbreviations, see Supplementary Table 4. (C) Correlation analysis between the initial MIC and the increase in MIC during adaptive laboratory evolution for all tested antibiotics. The scatterplots shows the initial MIC and the increase in MIC (both on log10-scale) during ALE for all antibiotic studied (indicated in the top of each panel). Increase in MIC was calculated by subtracting the initial MIC from the final MIC. Each point corresponds to an adapted line-antibiotic pair. Spearman’s rank correlation coefficients (calculated using a two-sided test) and corresponding p-values are indicated within each panel (absence of values in certain panels is due to the lack of variability in the initial MIC). Error bars represent 95% confidence intervals. Colours indicate the 4 bacterial species studied. For abbreviations, see Table 1. (D) Multiple linear regression (MLR) analysis on features shaping the resistance level reached during evolution. The analysis focused on three main features i) the MIC level of the ancestor strain (MICa), ii) the antibiotic employed (AB) and the iii) genetic background (strain). The adjusted coefficient of determination (adjusted R-square) was used as a statistical metric to measure the explanatory power of the models, ie. how much of the variation in the absolute increase in MIC (log2) can be explained by the variation in these features and combinations thereof, while adjusting for the number of parameters used in the fitted model. Additivity (indicated with + sign in axis labels) and interaction (* sign in axis labels) between explanatory variables are marked with orange and red colours, respectively. The predictors included in the models are also depicted in the right panel. We found a significant increase in adjusted R-square in all cases when a more complex model was compared to a simpler one (ANOVA, two-sided, P < 0.0001).
