Evolution and instability in ring species complexes: An in silico approach to the study of speciation

https://doi.org/10.1016/j.jtbi.2010.03.017Get rights and content

Abstract

Ring species are a biological complex that theoretically forms when an ancestral population extends its range around a geographic barrier and, despite low-level gene flow, differentiates until reproductive isolation exists when terminal populations come into secondary contact. Due to their rarity in nature, little is known about the biological factors that promote the formation of ring species. We use evolutionary algorithms operating on two simple computational problems (SAW and K-max) to study the process of speciation under the conditions which may yield ring species. We vary evolutionary parameters to measure their influence on ring species’ development and stability over evolutionary time. Using the SAW problem, ring species consistently form, i.e. fertility is negatively correlated with distance (R-values between −0.097 and −0.821, p<0.001), and terminal populations show substantial infertility. However, all SAW simulations demonstrate instability in the complex after sympatric zones are established between terminal populations. Higher mutation rates and larger dispersal/breeding radii promote ring species’ formation and stability. Using a problem with a simple fitness landscape, the K-max problem, ring species do not form. Instead, speciation around the ring occurs before ring closure as good genotypes become locally dominant.

Introduction

This paper presents a computational simulation for studying ring species. These biological complexes are rare and thus difficult to evaluate in nature. Our study employs evolutionary computation, a technique that allows for the incorporation of gene flow, fertility and fitness (selection) into the resulting simulation, to approximate the biological process under study. We explore the effect of modifying parameters: gene length, mutation rate and dispersal/breeding radius (defined in Table 1). By applying the simulation to two different computational problems, the SAW and the K-max, we explore the effect of different fitness landscapes on ring species’ formation. The SAW is a problem with a more rugose fitness landscape with many neutral networks. Its goal is to visit every square in a grid following a string of instructions telling it to move up, down, left, or right. The K-max is a problem with a simple and highly symmetric fitness landscape. Its goal is to create a string entirely composed of one of the K possible characters (for detailed descriptions see 3.1 Self-avoiding walks (SAW) as an artificial simulation system for evolution, 3.2 The, and Appendix B The SAW problem, Appendix C The).

Speciation is a fundamental process of evolutionary biology. Although a universal definition of species remains elusive, the biological species concept (BSC) (Mayr, 1942) is usually applicable to macroscopic animal life. In its strictest interpretation, species are defined as groups of interbreeding individuals that are reproductively isolated from other such groups. A more relaxed definition allows for small amounts of gene flow between closely related species (Coyne and Orr, 2004). The major barrier to a universal species concept is that species are not static entities but subject to the processes of evolution, and for this reason, exceptions to species definitions are common (Mayden, 1997). Despite the dynamic nature of evolution, speciation is an adaptive process that creates reproductively isolating mechanisms (Bush, 1975) with the expected outcome that a “species” will represent a group of individuals that are connected by gene flow but isolated from other such groups (Baker and Bradley, 2006, Bush, 1975).

Ring species (Mayr, 1942, Mayr, 1963) are a special case of speciation in which hybridization is not a rare occurrence, but a systematic event in a complex of individuals isolated from other groups but also isolated in a predictable pattern from each other. Ring species theoretically develop when an ancestral population expands its range around a geographic barrier and undergoes isolation by distance (morphologically, genetically, behaviorally, etc.). Despite gene flow (Irwin et al., 2005) and intergradation of populations around the ring, terminal populations become reproductively isolated by the time their dispersal brings them into secondary contact. In this way, ring species spatially exemplify the temporal process of speciation (Irwin, 2000). There are several specific criteria for a biological population to be deemed a ring species including: evidence that the sub-populations around the ring are derived from a single ancestral population; one of the terminal populations is the most recently founded sub-population; no existence or pre-existence of a geographical barrier that interrupted the ring; and, evidence of extensive gene flow between adjacent sub-populations that decreases with geographical distance around the ring (Coyne and Orr, 2004; Irwin, 2009; Irwin et al., 2001a,). All of these criteria can be easily controlled in our simulations.

One might assume that the expansion of populations around geographic barriers would be a common occurrence and the development of ring species complexes should follow. The paucity of examples suggests that the criteria may be too stringent (e.g. island systems are usually excluded) or that these complexes may be inherently unstable and prone to collapse and rapid speciation in nature (Noest, 1997; Gavrilets et al., 1998, Liebers et al., 2004), a prediction tested in our simulations. Genetic and phenotypic differentiations have been investigated in an attempt to study putative ring species. For example, Crochet et al. (2002) and Gay et al. (2009) use genetic evidence to argue that the gull complex, though exhibiting a high level of phenotypic variation, differs only in restricted regions of the genome and may not be reproductively isolated. Similarly, Joseph et al. (2008) use genetic information to reject a simple ring species model for Australian parrots and propose alternative models including incomplete ring speciation. The lack of a wide variety of natural examples makes it difficult to investigate factors that promote the formation of these complexes and influence their stability. We propose an in silico simulation that can incorporate measures of stability, fertility, gene flow, fitness and speciation.

In previous analyses (Ashlock and von Königslöw, 2008a, Ashlock et al., 2008b) we used problems in evolutionary computation to study speciation under different topological constraints and explored computational strategies to simulate speciation. In evolutionary computation the term ‘species’ is rarely used and has no standard definition as these analyses are normally done with isolated populations, not ecologies. For the study of simulated speciation in Ashlock and von Königslöw (2008) and Ashlock et al. (2008), a definition of ‘fertility’ (Table 1) was constructed within the context of the computational simulation to evaluate the resulting complexes.

In this investigation we use the computational problem (SAW, see Section 3.1) that best demonstrated ring species’ formation in previous simulations (Ashlock and von Königslöw, 2008a, Ashlock et al., 2008b) and introduce a negative control (K-max, see Section 3.2) predicted to favor speciation. The experimental design used controls the first three ring species requirements (single founder population, terminal populations are the most recent derivatives of the complex and no barriers to gene flow in the ring). The simulation output is tested for gene flow and fertility between terminal populations of the complex. The resulting complexes are evaluated to determine how well they adhere to the ring species concept. The simulation parameters are varied and the impact on speciation is assessed. Finally, the complexes are allowed to evolve zones of sympatry between terminal populations to assess their stability.

Section snippets

Evolutionary computation

Evolutionary computation (Ashlock, 2006, De Jong, 2006) is a technique used in diverse fields (engineering, economics, business, game theory, biology, etc.) incorporating evolution as an algorithm to solve otherwise intractable problems. Most practitioners of evolutionary computation draw inspiration from biological theory for practical application. A subset of evolutionary computation, Artificial Life (Langton, 1989), attempts to model biological evolution itself. Although there are

Self-avoiding walks (SAW) as an artificial simulation system for evolution

The self-avoiding walk (SAW) is a computational problem in which an artificial organism is represented by a string of commands: up (U), down (D), left (L) and right (R) that denote the path it walks through a finite grid (Fig. 1). The goal of the problem is to visit the largest number of squares in the grid with grid size and string length as parameters of the problem. This name “self-avoiding walk” indicates that optimal solutions do not revisit squares. The string of commands comprising a SAW

Methods

We allowed SAW and K-max simulations to evolve in a ring shaped topology. We varied evolutionary parameters (gene length l, breeding/dispersal radius b, and mutation rate μ). We assessed the formation and stability of the resulting complexes on populations saved at two time points, ring closure and post-ring closure (after a substantial sympatric population had formed between terminal populations). An initial simulation with the SAW problem established the viability of the SAW problem for

Experiment 1: simulation of ring species

Under the moderate parameters of gene length l=35, breeding/dispersal radius b=7 and mutation rate μ=3, all 20 replicates showed a sharp drop in the fertility between adjacent members of the terminal populations when the ring closed. We calculated the fertility of 5000 randomly chosen pairs of individuals in each replicate from various distances around the ring. In all cases there was a significant negative Spearman Rank correlation between these variables (R-values between −0.232 and −0.557, p

Discussion

The goals of this study were: to determine if a ring species could form and persist in our simulation incorporating natural selection and genetic drift; to explore the parameters of the simulation; and, in the case that ring species formed, to explore the impact of the parameters on their stability. Using the SAW, our simulation consistently demonstrated a pattern of development analogous to the proposed evolution of natural ring species, and we showed that basic evolutionary factors, such as

Conclusion

This study uses a simulation incorporating evolutionary computation to study the formation and stability of ring species complexes. Using two different computational problems, we demonstrate that the fitness landscape has a large impact on whether ring species form. The K-max problem, with K global optima but no local optima, does not form ring species. The SAW problem, which is rich in both global and local optima, forms ring species consistently, but they are unstable when terminal

Acknowledgments

Funding for this project was provided by the Natural Sciences and Engineering Research Council (NSERC) of Canada through grants to D. Ashlock and through CGS-D awards to ELC and W. Ashlock. We appreciate the comments of three anonymous reviewers who provided suggestions enhancing the presentation of this manuscript.

Author's contributions: Daniel Ashlock designed and conducted the study and participated in the creation of the manuscript. Elizabeth Clare assisted in the design of the experiment,

References (46)

  • S. Bensch et al.

    Patterns of stable isotope signatures in willow warblers Phylloscopus trochilus feathers collected in Africa

    J. Avian Biol.

    (2006)
  • S. Bensch et al.

    Genetic, morphological, and feather isotope variation of migratory willow warblers show gradual divergence in a ring

    Mol. Ecol.

    (2009)
  • R.S. Bigelow

    Hybrid zones and reproductive isolation

    Evolution

    (1965)
  • G.L. Bush

    Modes of animal speciation

    Annu. Rev. Ecol. Syst.

    (1975)
  • S.A. Church et al.

    The evolution of reproductive isolation in spatially structured populations

    Evolution

    (2002)
  • J.A. Coyne et al.

    Speciation

    (2004)
  • P-A. Crochet et al.

    Systematics of large white-headed gulls: patterns of mitochondrial DNA variation in western European taxa

    Auk

    (2002)
  • M.A.M. De Aguiar et al.

    Global patterns of speciation and diversity

    Nature

    (2009)
  • Kenneth A. De Jong

    Evolutionary Computation: A Unified Approach

    (2006)
  • S. Gavrilets et al.

    Rapid parapatric speciation on holey adaptive landscapes

    Proc. R. Soc. London Ser. B

    (1998)
  • S. Gavrilets et al.

    Patterns of parapatric speciation

    Evolution

    (2000)
  • L. Gay et al.

    Speciation with gene flow in the large white-headed gulls: does selection counterbalance introgression?

    Heredity

    (2009)
  • R.M. Gibson et al.

    Sexual selection in Lekking sage grouse: phenotypic correlates of male mating success

    Behav. Ecol. Sociobiol.

    (1985)
  • Cited by (0)

    View full text