Statistical analysis
All analysis was performed in R (version 4.0.5; R Core Team 2018) and plots were made using the R package ‘ggplot2’ (Wickham 2016). Population density data were log-transformed and bacteria resistance data were arcsine-transformed before analysis. Bacterial population size (density) during the evolution experiment was analysed using mixed-effect linear model in the ‘nlme’ package (Pinheiro et al. 2021). Treatment and time (transfer number) were included as categorical and continuous explanatory variables respectively, with microcosm ID as a random effect. Significance of each explanatory variable was estimated using the ‘Anova’ function provided by the ‘car’ package (Fox et al.2021). Pairwise multiple comparison between treatments was performed using Tukey’s HSD with the ‘glht’ function in ‘multcomp’ package (Bretzet al. 2022). To further estimate population dynamics, separate analyses were conducted for each treatment using linear models with time as an explanatory variable.
One-tailed t-tests were used to analyse the growth performance (density data) measured at 10°C and the top-down effect by phages. Wilcoxon rank tests were used when data were not normally distributed. The effects of treatment and transfer time on bacteria resistance were estimated using mixed-effect linear models.