A quantitative genetics framework for understanding the selection response of microbial communities

Heritability, a quantity that reflects the degree of resemblance between parent and offspring traits, is measured during plant and animal breeding because it predicts selection success during artificial selection of individuals. However, when whole microbial communities are under artificial selection to improve their traits, high heritability of the community trait does not necessarily predict selection success. To better understand the relationship between heritability and success during community selection, we establish a quantitative genetics framework, and in doing so, we obtain practical recommendations. Specifically, we decompose a community trait into "trait determinants": genotype compositions and species compositions that impact the community trait and that vary among communities. This allows us to interpret heritability of a community trait in terms of the heritability of its determinants. We then use the Price equation to partition the selection response of a community trait into three phenomena: inter-community selection (heritability multiplied by selection intensity), transmission infidelity (the change in community trait from parent to offspring), and nonlinearity (due to a nonlinear relationship between parent and offspring traits). We illustrate that evolution within a community can cause the three terms to covary: in addition to the known effect of worsening transmission infidelity, intra-community evolution can lead to inflated heritability values greater than one (through an effect whereby "the poor get poorer"), and simultaneously magnify nonlinearity. As a consequence of these effects, heritability no longer predicts the selection response of a community trait. We propose effective selection strategies that improve heritability without accelerating intra-community evolution.