Community function landscape and steady state species composition shape the eco-evolutionary dynamics of arti1cial community selection

Abstract

Microbial communities often perform important functions that arise from interactions among member species. Community functions can be improved via artificial selection: Many communities are repeatedly grown, mutations arise, and communities with the highest desired function are chosen to reproduce where each is partitioned into multiple offspring communities for the next cycle. Since selection efficacy is often unimpressive in published experiments and since multiple experimental parameters need to be tuned, we sought to use computer simulations to learn how to design effective selection strategies. We simulated community selection to improve a community function that requires two species and imposes a fitness cost on one of the species. This simplified case allowed us to distill community function down to two fundamental and orthogonal components: a heritable determinant and a nonheritable determinant. We then visualize a "community function landscape" relating community function to these two determinants, and demonstrate that the evolutionary trajectory on the landscape is restricted along a path designated by ecological interactions. This path can prevent the attainment of maximal community function, and trap communities in landscape locations where community function has low heritability. Exploiting these observations, we devise a species spiking approach to shift the path to improve community function heritability and consequently selection efficacy. We show that our approach is applicable to communities with complex and unknown function landscapes.