Coevolutionary patterns caused by prey selection

Highlights

• Prey selection represents the predator individual behavior in searching for prey.

• We investigated the effect of prey selection on the predator and prey coevolution.

• Prey selection can cause coevolutionary patterns that have not yet been reported.

• Prey selection can increase the likelihood of extinction of both predators and prey.

• An analytical approach allowed us to show the robustness of our results.

Abstract

Many theoretical models have been formulated to better understand the coevolutionary patterns that emerge from antagonistic interactions. These models usually assume that the attacks by the exploiters are random, so the effect of victim selection by exploiters on coevolutionary patterns remains unexplored. Here we analytically studied the payoff for predators and prey under coevolution assuming that every individual predator can attack only a small number of prey any given time, considering two scenarios: (i) predation occurs at random; (ii) predators select prey according to phenotype matching. We also develop an individual based model to verify the robustness of our analytical prediction. We show that both scenarios result in well known similar coevolutionary patterns if population sizes are sufficiently high: symmetrical coevolutionary branching and symmetrical coevolutionary cycling (Red Queen dynamics). However, for small population sizes, prey selection can cause unexpected coevolutionary patterns. One is the breaking of symmetry of the coevolutionary pattern, where the phenotypes evolve towards one of two evolutionarily stable patterns. As population size increases, the phenotypes oscillate between these two values in a novel form of Red Queen dynamics, the episodic reversal between the two stable patterns. Thus, prey selection causes prey phenotypes to evolve towards more extreme values, which reduces the fitness of both predators and prey, increasing the likelihood of extinction.

Introduction

Antagonistic interactions are an ubiquitous phenomenon in nature, in which one species gains resources at the expense of another. Victim-exploiter relationships encompass a range of interspecific interaction modes such as plant-herbivore, prey-predator and host-pathogen interactions. These interactions can exert reciprocal selective pressures resulting in genetic changes in populations; what is defined as coevolution (Janzen, 1980). In the last decades, many theories have been formulated to explain and better understand how interacting species affect the evolution of others (e.g. Thompson (2005) and Abrams (2000)).

One of the most influential hypotheses in coevolution is the Red Queen Dynamics (van Valen, 1973), which proposes that species must constantly adapt as a response to the unceasing adaptations of the organisms with which it interacts. As the fitness of the exploited species is reduced, selection will favor those organisms with a better capability of defending themselves or evadiploiters, whereas exploiters will be selected to evolve countermeasures (Hochberg and van Baalen, 1998). Many empirical examples and systems that attain such dynamics are already well known and studied, as between birds and avian brood parasites (Avilés et al., 2006, Martín-Gálvez et al., 2006, Refsnider and Janzen, 2010, Noh et al., 2018) and between herbivorous insects and their host plants (Ehrlich and Raven, 1964, Chew, 1977, Nylin et al., 2005, Merrill et al., 2013).

Two frequent ingredients present in coevolutionary antagonistic models are stabilizing and interaction selections. Stabilizing selection favors an optimum phenotype that the population would evolve towards in absence of any other selective pressure. The interaction selection depends on both victim and exploiter phenotypes and, once the interaction is antagonistic, it does not favor the convergence between victim and exploiter phenotypes. Although the intensity of the interaction selection is able to limit the range of phenotypes that an exploiter can succeed in attacking, it does not limit the phenotype range that an exploiter will try to attack. It means that the individual behavior in searching victims is not under selection when no other pressure aside from stabilizing and interaction selections is considered (Matsuda, 1985). Some models on optimal foraging strategies have already highlighted the possibility of prey selection (victim choice) may have important consequences on the population dynamics (Stephens and Krebs, 1986, Berec, 2000), however they do not incorporate evolutionary dynamics. On the other hand, most models for predator-prey coevolution do not impose any priority on predators’ attacks, which imply that they attack at random in their predation neighborhood (see e.g, Murdoch, 1969, May, 1974, Hutson, 1984, Matsuda, 1985, Gleeson and Wilson, 1986, Gendron, 1987, Fryxell and Lundberg, 1994, Abrams and Abrams, 1999, van Baalen et al., 2001). An exploiter can increase its fitness if it can choose the victim that maximizes its chances of success instead of interacting randomly. A classical empirical example is brood parasites, known to coevolve with their hosts (Rothstein, 1990). Brood parasites can choose to parasitize a host when their eggs match their host’s egg colour (Resetarits, 1996, Avilés et al., 2006, Refsnider and Janzen, 2010, Soler et al., 2014). The choice of host has consequences for offspring success and therefore it is subject to strong selection (Resetarits, 1996, Refsnider and Janzen, 2010). There is also a vast literature on insect oviposition patterns suggesting that host plant preference and selection has a significant role in the coevolutionary history of these species (Chew, 1977, Jorge et al., 2014, Nylin et al., 2005, Merrill et al., 2013).

The evolutionary patterns predicted by coevolutionary models are a result of different pressures in which coevolutionary interactions occur (Thompson, 2005) Specifically, when the interaction selection is determined by the similarity of the interacting species phenotypes - phenotype matching (Brown and Vincent, 1992, Berenbaum and Zangerl, 1998, Gomulkiewicz et a., 2000, Abrams, 2000, Gandon and Michalakis, 2002, Nuismer and Thompson, 2006, Calcagno et al., 2010, Yoder and Nuismer, 2010, Gokhale et al., 2013, Andreazzi et al., 2017) – prey phenotype can evolve to values adjacent to the phenotype predators aim at. When stabilizing selection and non-directional interaction selection are combined and assume a symmetrical shape (e.g., as a Gaussian function), the predicted temporal population phenotype distributions are also symmetrical with respect to the optimum phenotype favored by stabilizing selection, resulting insymmetrical coevolutionary branching(Brown and Vincent, 1992, Abrams, 2000, Calcagno et al., 2010, Yoder and Nuismer, 2010) orsymmetrical coevolutionary cycling (Gomulkiewicz et a., 2000, Abrams, 2000, Dieckmann et al., 1995). In the former, the population phenotypes evolve towards a set of stable phenotypes symmetrically distributed around the optimum phenotype favored by stabilizing selection while in the latter the phenotypes oscillate around it. On the other hand, when the interaction selection is directional (for example when the outcome of the interaction is determined by phenotype differences among interacting species), prey have a preferential evolution pathway imposed by the interaction (Abrams, 2000). As a consequence, the symmetry is broken (this is clearly shown by Yoder and Nuismer (2010)). Even though there are abundant studies modeling co-evolutionary patterns, we still lack in understanding how the individual behavior in victim preference by the exploiters modulates these patterns.

To investigate the effects of prey selection in the coevolutionary dynamics of antagonistic populations, we approach the evolutionary predator-prey system where the individual phenotypes related to the interaction can evolve subject to both interaction (phenotype matching) and stabilizing selections. These selective pressures are assumed to have symmetrical shape, which is expected to promote symmetrical coevolutionary patterns: symmetrical coevolutionary branching and symmetrical coevolutionary cycling. We explore the coevolutionary outcomes in phenotypic evolution considering two scenarios: (i) Predation occurs at random; (ii) predators select which prey to attack among those present in the predator’s ‘attack neighborhood’, according to phenotype matching. We then analyze the individuals’ payoffs and also propose an individual-based model where individual phenotypes are explicitly modeled and predators have a limited predation neighborhood. Both approaches agreed that prey selection can promote an asymmetrical stable pattern. Moreover, simulation outcomes allowed us to assess the robustness of our findings and also its sensitivity under different strengths of interaction selection and carrying capacity.