Abiotic stress is a major force of selection that organisms are constantly facing. While the evolutionary effects of various stressors have been broadly studied, it is only more recently that the relevance of interactions between evolution and underlying ecological conditions, that is, eco-evolutionary feedbacks, have been highlighted. Here, we experimentally investigated how populations adapt to pH-stress under high population densities. Using the protist species Tetrahymena thermophila, we studied how four different genotypes evolved in response to stressfully low pH conditions and high population densities. We found that genotypes underwent evolutionary changes, some shifting up and others shifting down their intrinsic rates of increase (r₀). Overall, evolution at low pH led to the convergence of r₀ and intraspecific competitive ability (α) across the four genotypes. Given the strong correlation between r₀ and α, we argue that this convergence was a consequence of selection for increased density-dependent fitness at low pH under the experienced high density conditions. Increased density-dependent fitness was either attained through increase in r₀ , or decrease of α, depending on the genetic background. In conclusion, we show that demography can influence the direction of evolution under abiotic stress.

Data was collected from video analysis of 20s videos made from Tetrahymena thermophila cells under a light microscope, and processed using BEMOVI video analysis.

Further downstream model estimates were generated using Beverton-Holt model fitting and further analysis of trait estimates using Bayesian linear mixed models.

-Included data contains:

1) Raw density data, from ancestral lines, evolved lines and of density measurements over the course of the evolution experiment

2) Beverton-Holt model fits (raw posteriors and summarized posteriors)

3) Trait data used for PCA fits

4) Data on the relation between HCl concentration and pH of the medium

-Code is also available from Github, but for further downstream analyses may require to also run previous steps, to recreate the Bayesian model posteriors used in these downstream analyses. All raw data is available to repeat the analyses.