“Evolutionary responses to conditionality in species interactions across environmental gradients”

Posted on

Anna M. O’Brien, Ruairidh J. H. Sawers, Jeffrey Ross-Ibarra, and Sharon Y. Strauss (Dec 2018)

Read the Article

(Co)evolutionary outcomes of conditionality in biotic interactions: prediction, experiment design, and statistical test

Phlox diffusa (flowering) and various other species growing in close proximity in high alpine. Mt. Rainier National Park, WA, 2016. Well-supported ecological theory suggests that some or all individuals benefit from this proximity, but at lower elevation sites, these same species would compete strongly for resources. Such divergent impacts on reciprocal fitness could have evolutionary consequences that we discuss here.
(Credit: Anna M. O’Brien)

While we often characterize species interactions as being either beneficial or detrimental to the participants, in reality, the fitness outcomes of interactions can vary substantially, especially across environments (a.k.a. conditionality). A well-known example of such conditionality is the outcome of interactions between plants in the high alpine, where sites are cold, dry, and windy, and the substrate is unstable. Here, plants promote each others’ growth by buffering temperatures, wind, and soil erosion, but when the same species are neighbors lower on the same mountain slope, they compete with one another for resources. Ecological theory suggests that physical stress or the lack of availability of resources can determine the outcome of interactions: species interactions may become more positive (shift to mutually beneficial outcomes) under stressful conditions and become more negative when resources are freely available or when the physical environment is benign.

Environmentally driven variation in interaction outcomes may also shape evolutionary responses of species. For example, selection may favor new mutations in interacting species that increase investment in beneficial interactions (mutualisms) at stressful ends of environmental gradients, leading to mutually beneficial coevolution between interacting partners. On the other hand, when stress is minimal or non-existent, one might predict no coevolution, or possibly antagonistic coevolution between interacting partners.

Here, the authors develop and demonstrate new methods to empirically test these predictions. Insights gained from these models can be applied to a wide range of species interactions and conditional outcomes. Understanding these ecological and evolutionary forces will become increasingly important as species find themselves in novel interactions and physical contexts with climate change.


The outcomes of many species interactions are conditional on the environments in which they occur. Often, interactions grade from being more positive under stressful or low-resource conditions to more antagonistic or neutral under benign conditions. Here, we take predictions of two well-supported ecological theories on conditionality—limiting resources models and the stress gradient hypothesis—and combine them with those from the geographic mosaic theory of coevolution (GMTC) to generate predictions for systematic patterns of adaptation and co-adaptation between partners along abiotic gradients. When interactions become more positive in stressful environments, mutations that increase fitness in one partner may also increase fitness in the other; because fitnesses are aligned, selection should favor greater mutualistic adaptation and co-adaptation between interacting species in stressful ends of environmental gradients. As a corollary, in benign environments, antagonistic co-adaptation could result in Red Queen or arms-race dynamics, or reduction of antagonism through character displacement and niche partitioning. Here, we distinguish between generally mutualistic or antagonistic adaptation (i.e., mutations in one partner that have similar effects across multiple populations of the other), versus specific adaptations to sympatric partners (local adaptation), which can occur either alone or simultaneously. We then outline the kinds of data required to test these predictions, develop experimental designs and statistical methods, and demonstrate these using simulations based on GMTC models. Our methods can be applied to a range of conditional outcomes, and may also be useful in assisted translocation approaches in the face of climate change.