I eco, therefore I evo.

share us on facebook!!! Facebook

logo_team_ecoevo.jpg

Running title

Do ecological interaction networks predict macroevolutionary trajectories of ecologically-relevant traits?

Key words

communities, ecological interactions, ecological networks, ecosystems, macroevolution, mutualism, phylogeny, traits

Members

María Aguilar (Post-doc, Real Jardín Botánico, CSIC, Spain)
Alessandro Alves-Pereira (PhD candidate, ESALQ/USP, Brazil)
Justin C. Bagley (PhD candidate, Brigham Young University, USA)
Emanuel C. Barbosa (Undergraduate student, UFPI, Brazil)
Eileen Butterfield (Masters student, Uppsala University, Sweden)
Anca-Neluta Dragu (Masters student, Emil Racovita Institute of Speleology, Romania)
Yang Li (PhD candidate, University of Oxford, UK)
Mirian Liza A. F. Pacheco (PhD candidate, IG-USP, Brazil)

Introduction

Ecologists and evolutionary biologists have long recognized the importance of ecological factors in driving evolution. For example, predatory fishes are known to have caused the rapid evolution of life history traits of Trinidadian guppies (Poecilia reticulata; Poeciliidae), including earlier development of reproductive maturity at smaller body size, multiple times (Reznick et al. 1996). Also, overfishing (size-selective predation) by humans caused a similar pattern of rapid evolution of maturation within the Atlantic cod (Gadus morula) fishery just prior to its collapse (Olsen et al. 2004). Aside from predation, other strong ecological interactions including competition, mutualisms and plant-pollinator interactions have led to the evolution of distinct patterns of phenotypic variation.

1fish.jpg

In recent years, the development of methods and databases (e.g. TreeBase, http://www.treebase.org/; Fishbase, http://www.fishbase.org/) relevant to phylogenetic comparative methods (PCM; Felsenstein et al., 1985; Harvey and Pagel, 1991; Garland and Ives, 2000; Garland et al., 2005) has generated many novel insights into the tempo and mode of the evolution of biotic lineages as well as their most important traits. For example, statistical analyses of trait data and phylogenies are yielding insights into unusual pulses of vertebrate radiation (Alfaro et al., 2009), as well as the role of traits and early bursts of morphological evolution which may be adaptive during radiations (e.g. Harmon et al., 2003, 2010), among other themes. Several novel methods now permit the incorporation of time-calibrated molecular phylogenies, as well as information from the fossil record (Slater et al., 2012), to test macroevolutionary predictions within explicit, phylogeny-based statistical frameworks. Additional theoretical work underlies these methods, and permits using frameworks based on likelihood (e.g. Pagel, 1999) as well as Bayesian (e.g. Pagel et al., 2004) and approximate Bayesian computation methods (e.g. Eastman et al., 2011).

Nonetheless, although evolutionarily important traits arise within species in the context of ecological communities and ecosystems (and their interacting biotic parts) over time, the effects of ecological interactions in shaping the evolution of traits has been frequently overlooked in phylogenetic comparative approaches. The contribution of several environmental parameters (e.g. habitat, physical barriers to gene flow, environmental/selective regime gradients, and catastrophic events) to evolution have been widely studied (Schluter, 2000; Alfaro et al., 2009). However, to our knowledge, no quantitative method is available for evaluating the evolution of ecologically-relevant traits in the context of interactions within ecosystems.

Previous studies have shown that the strength of the connections within protein protein interaction networks can predict the similarity of the rate of evolution of the involved proteins (see figure 1, Fraser et al., 2002). The evolutionary trajectory of a single gene has therefore been shown to be affected not only by direct selective pressures, but also by selective forces which act indirectly on interacting genes.

fig1.png

These results, in addition to a burst of recent work presenting new data on ecological interactions and computational methods for analyzing networks of ecological relationships (for more comprehensive reviews see Proulx et al., 2005; Ings et al., 2009), are revitalizing interest in defining ways to describe interactions between species/populations. While it has been recognized that such networks can display phylogenetic signal (Rezende et al., 2007), we propose that they also may be useful in predicting the evolutionary trajectories of various traits. The general idea is that strongly interacting species may share evolutionary trajectories in traits which reflect their interactions, or that simply being part of a tightly knit community has a significant impact on the evolutionary trajectory of the species.

Team ecoevo proposes to extend previous phylogenetic comparative approaches to the study of the effects of ecological interactions on macroevolutionary pattern and process (Eastman et al., 2011) by developing a method that integrates information about interaction between species/populations with phylogenetic. Our approach reconstructs ancestral interaction networks from extant networks using several models under the assumption that ecological interactions are evolutionarily conserved (Gomez et al., 2010). Parameters of ancestral interaction networks are then used in two different ways: 1) to fit an integrative model of macroevolutionary trait evolution, and 2) to define clusters of species/populations expected to share similar evolutionary trajectories. We test whether the novel integrative model (using ancestral network parameters) is significantly better than the simpler model (not accounting for ecological network relationships) at predicting the parameters of subsequent trait evolution. We also test whether members within distinguishable network clusters tend to evolve in more similar ways than members of a set of randomly sampled species.

We aim to test our novel approach using three case studies derived from different ecological systems.

RESEARCH GOALS

GOAL 1. Develop method for network evolution and ancestral ecological network reconstruction on phylogenies. Questions: "Which (ancestor of modern) species were most likely to interact with which other (ancestral) species in the evolutionary past? OR Which lineages were most likely to interact in the past, given a model of the phylogeny plus extant ecological network?"
GOAL 2. Development of datasets for application of method described in goal 1 to case studies.
GOAL 3. Development of new method for combining simulations of network evolution with models of trait evolution.
- What describes trait evolution better… model with 1, less or more potential OU optima for trait values of given trait (e.g. fish gape width)?
- How do we connect this to ecol network evolution? a. ad hoc approach seeing if/where shifts in optimum correlate with changes in interactions, or particular patterns of interactions…
GOAL 4. Implementation of new method developed in 3 to analyze the case studies we obtained data for from nature in goal 2.

PROPOSED RESEARCH

GOAL 1. Develop method for network evolution and ancestral ecological network reconstruction on phylogenies.

Developing models of ecological network evolution―This method requires several kinds of data and assumptions typically made in other phylogenetic comparative methods—a phylogeny, assumed to be known without error, for all or a select set of species within an ecosystem of defined area (e.g. 1000 m segment of river or stream habitat, 1 km2 of Amazonian forest); as well as phylogenetic branch lengths, in this case in units of time (thus the phylogeny should be a chronogram, a product of a time-calibrated analysis). However, in addition to these data, our method requires additional pieces of information in the form of a known (or hypothetical, e.g. based on a literature search) network of ecological interactions among the members of the study ecosystem (e.g. a supertree generated from phylogenetic analysis in BEAST of a set of GenBank DNA sequences from individuals of all organisms, or as many of the organisms as possible, that occur in the ecosystem of interest).

10model_iii.jpg
7input.jpg

To capture the effects of changes in the pattern of ecological interactions on the macroevolutionary trajectories of ecological traits, we simulate the evolution of ecological networks across our phylogenetic trees by estimating parameters of network evolution and obtaining distributions of the potential ecological network for a set of taxa at ancestral time points (e.g. branching/speciation points).

We evolve our network by comparing models with different numbers of parameters describing the rate at which gains (G)/losses (L) of interaction occur between pairs of species. Our simplest model, which we call M, consists of 2 parameters (G the rate of gaining an interaction with each un-interacting species, and L the rate of losing an interaction with each interacting species). The model M+G adds an extra parameter, G', to M allowing the evolution of the network to be governed by two rates of interaction gains, while the model M+L allows two rates of interaction losses. The model M+G_nL_m consists of n additional gain parameters and m additional loss parameters.

We assume that after speciation all interactions are inherited from parent to daughter species (see t1,t2,t3,t4 in last 2 figures of section). In theory, the model of network evolution is allowed to change at any time, but we only estimate the best model describing network evolution between speciation events in the phylogenetic tree.

Parameter and ancestral network estimation―The estimation of parameters and ancestral network is tricky. On one hand, we need a probability distribution of the ancestral network to choose the best model (i.e. n,m for M+G_nL_m) and estimate its parameters that best describe the extant network. On the other hand, we need the model and the distribution of its parameters to estimate the distribution of the ancestral networks. We outline below a simple expectation-maximization strategy to choose the model, estimate its parameters and produce a distribution of potential ancestral networks.

Update the distribution of ancestral network―Given fixed models of network evolution (M+G_nL_m) and their parameters,

8model_i.jpg
9model_ii.jpg

we update the ancestral network distribution by starting with a network and "proposing" small, local, changes and allowing the new network to evolve over time according different fixed parameters (see figure on left). The likelihood of the model is then computed by comparing the distribution of "evolved" networks obtained to the distribution of the extant network. More precisely, the bigger the overlap between the "evolved" network distribution and the extant network distribution, the more likely the ancestral network was the one we "proposed". After proposing many alternative ancestral networks, we update the distribution of ancestral network by comparing their likelihood.

Updating the evolutionary model and the distribution of its parameters―Given fixed ancestral networks, we update the evolutionary model and its parameters by testing different models separately and proposing small changes to their parameters. Once again, we let the network "evolve" and we look at the overlap of the distribution of "evolved" networks and the distribution of extant networks.

12model_v.jpg
11model_iv.jpg

Evolving the network―Given the extant phylogenies, we first estimate the birth/death parameters (B,D) best fitting each clade of the tree according to previously established methods (cite something). With these parameters, we simulate the birth and death of lineage by repeatedly sampling according to a Poisson distribution with rate B and D, respectively (see figure on the left).

At this point, we have a novel phylogeny with speciation and extinction events. We also have an ancestral network, a model of network evolution and its parameters (e.g. M, M+G, …). For every lineage that survive in the two phylogenies (see bold lineages on figure), we simulate the gain and loss of interaction between species X and Y (where X and Y are in different phylogenies) by drawing from a Poisson distribution with rate min{G_X,G_Y} and max{L_X,L_Y}, respectively (see figure). The rationale for modeling the gain of interactions between species X and species Y as the minimum rates of interaction gain of the two can be understood by considering a scenario where X is extremely promiscuous in creating interactions with other species (G_X is high), but the other species Y rarely gains interactions (G_Y is low). The waiting time for species X to start interacting with species Y should be longer, therefore it is intuitive to take the min of the two rate. The justification for using max in the case of loss rates is similar.

GOAL 2. Development of datasets for application of method described in goal 1 to case studies.

5phylogeny%281%29.jpg

Phylogenetic methods, taxon sampling — We are proposing a general model, and many study systems may be evaluated. Options include:
1. species form very diverse taxonomic groups. In this case, distinct groups of ecological interaction networks could be studied by evaluating different subgroups, or targeting only one specific subgroup of interest. One example could be the study of the Angiosperm plant species. Since Angiosperms have a worldwide distribution, one could use the method to study a specific community of interest, or even at the ecosystem level.
2. sampling only taxa which constitute a given ecosystem of interest. In this scenario, it is not necessary to sample specimens from all of the taxa which constitute the ecosystem, or individual communities within it. Rather, one could use only those species for which the ecological networks are well-know or may be easily described.

The latter approach would likely provide insights into the influence of interactions on evolution in other ecosystems/communities that share similar features.

DNA sequence acquisition – This may depend on the chosen group of organisms. If the taxa within the group are closely related it would be required genome-wide sequences, while for more distant related taxa the gene sequences that produced well-resolved phylogenies would be sufficient. The DNA sequences might also be available in public data bases (e.g. GenBank).

Data analyses — Sequence alignment will be performed in Geneious (; ). Evolutionary models of substitution for the data under the corrected Akaike information criterion (AIC) may be determined using the MrModeltest v.2.2 (Nylander, 2004). The reconstruction of phylogenetic relationships among taxa may be performed with Bayesian MCMC inference using BEAST v.1.5.3 (Drummond and Rambaut, 2007). The burn-in phase will exclude trees sampled before stable posterior probability (PP) values is reached.

Divergence dates estimation – The fossil record provides the best information with which to transform relative time estimates into absolute ages (Magallón, 2004). The inclusion of this information may guide at which points in phylogeny the model will be applied. The divergence dates may be estimated with BEAST, assuming a uniform distribution for the fossil calibrations with the lower hard bound of the distribution set to the youngest age of the fossil, and the upper hard bound set to the first fossil record. Convergence statistics for each single-gene run will be analyzed in Tracer (Rambaut and Drummond, 2003). TreeAnnotator v. 1.5.3 (Drummond and Rambaut, 2007) may be used to produce maximum clade credibility trees from the post–burn-in trees and to determine the 95% probability density of ages for the nodes in the tree.

Morphological ancestral state reconstructions — Morphological information may be obtained from literature, from observations of herbarium, museums etc. Ideally, the traits used for the reconstruction should be related to adaptive features that organisms should evolve in response to the interactions with each other. Additionally, traits which show a clear pattern of modification across the fossil record might also be included in the analysis. The character state reconstructions will be based on the tree with the highest PP, and will be performed with the Mesquite v.2.72 (Maddison and Maddison, 2010), using likelihood reconstruction methods based on the Mk1 model (Markov k-state one parameter – equal probability of change between states).

Constructing ecological networks – To the construction of networks, the modularity level and the number of modules per network may be determined using an algorithm based on simulated annealing (identification of modules, which are groups of species having most of their links within their own module) with ‘adonis’ in the R package VEGAN (Oksanen, 2008).

6networks%281%29.jpg

GOAL 3. Development of new method for combining simulations of network evolution with models of trait evolution. This will be used to analyze the datasets generated in GOAL 2.
(To be determined…)

GOAL 4. Implementation of new method developed in 3 to analyze the case studies we obtained data for from nature in goal 2.
(To be determined…)

SUMMARY AND FUTURE DIRECTIONS

We have laid the groundwork for developing methods to incorporate information on ecological networks in analyses of trait evolution. The next steps for this research consist of developing a new method for combining simulations of network evolution developed herein with models of trait evolution. In future work, it will be desirable to work towards implementation of the new method developed in goal 3, in order to analyze the case studies we obtained data for from nature in goal 2.

References

Alfaro M. E., Santini F., Brock C., Alamillo H., Dornburg A., Rabosky D. L., Carnevale G., Harmon L. J. 2009. Nine exceptional radiations plus high turnover explain species diversity in jawed vertebrates. Proceedings of the National Academy of Sciences USA, 106:13411-13414.

Drummond, A.J.; Rambaut, A. 2007 BEAST: Bayesian evolutionary analysis by sampling trees. BMC Evol Biol, 7(214),8 pp.

Eastman J. M., Alfaro M. E., Joyce P., Hipp A. L., Harmon L. J. 2011. A novel comparative method for identifying shifts in the rate of character evolution on trees. Evolution, 65(12):3578-3589.

Fraser H. B., Hirsh A. E., Steinmetz L. M., Scharfe C., Feldman M. W. 2002. Evolutionary rate in the protein interaction network. Science, 296:750-752.

Garland T. Jr., Bennett A. F., Rezende E. L. 2005. Phylogenetic approaches in comparative physiology. Journal of Experimental Biology, 208:3015-3035.

Garland T. Jr., Ives A. R. 2000. Using the past to predict the present: confidence intervals for regression equations in phylogenetic comparative methods. American Naturalist, 155:346-364.

Gomez J. M., Verdú M., Perfectti F. 2010. Ecological interactions are evolutionarily conserved across the entire tree of life. Nature, 465:918-922.

Harmon L. J., Schulte J. A. II, Larson A., Losos J. B. 2003. Tempo and mode of evolutionary radiation in Iguanian lizards. Nature, 301:961-964.

Ings T. C., Montoya J. M., Bascompte J., Blüthgen N., Brown L., Dormann C. F., Edwards F., Figueroa D., Jacob U., Jones J. I., Lauridsen R. B., Ledger M. E., Lewis H. M., Olesen J. M., van Veen F. F. J., Warren P. H., Woodward G. 2009. Ecological networks—beyond food webs. Journal of Animal Ecology, 78:253-269.

Maddison, W.P.; Maddison, D. R. 2010. Mesquite: A modular system for evolutionary analysis, version 2.72. http://mesquiteproject.org.

Magallón, S.A. 2004. Dating lineages: molecular and paleontological approaches to the temporal framework of clades. Internatnl J Plant Sci 165:S7–S21.

Nylander, J.A.A. 2004 . MrAIC.pl. Evolutionary Biology Centre, Uppsala University, Uppsala, Sweden. http://www.abc.se/~nylander/.

Oksanen, J. 2008. Multivariate Analysis of Ecological Communities in R: Vegan Tutorial.

Pagel M. 1997. Inferring evolutionary processes from phylogenies. Zoologica Scripta, 26:331-348.

Pagel M. 1999. Inferring the historical patterns of biological evolution. Nature, 401:877-884.

Pagel M., Meade A., Barker D. 2004. Bayesian estimation of ancestral character states on phylogenies. Systematic Biology, 53(5):673-684.

Proulx S. R., Promislow D. E. L., Phillips P. C. 2005. Network thinking in ecology and evolution. Trends in Ecology and Evolution, 20(6):345-353.

Rambaut, A.; Drummond, A.J. 2003.Tracer 1.4. Available at http://tree.bio.ed.ac.uk/software/tracer.

Rezende E. L., Lavabre J. E., Guimarães P. R. Jr., Jordano P., Bascompte J. 2007. Non-random coextinctions in phylogenetically structured mutualistic networks. Nature, 448:925-928.

Schluter, D. 2000. The ecology of adaptive radiation. Oxford University Press, Oxford, UK.

Slater G. J., Harmon L. J., Alfaro, M. E. 2012. Integrating fossils with molecular phylogenies improves inference of trait evolution. Evolution, in press.

Glossary

Character - an observed structure or behavior in organisms. In systematics, characters are used as common properties of different organisms and are regarded as being heritable and homologous. [Team ecoevo]

Character state - a variant form of a given phenotypic character (see also phenotypic characters). For example, pink, red, yellow or white petals may be traits of flower petal color charcater. [Team ecoevo]

Community - all actually or potentially interacting species present in a defined area at the same place and time. [Team ecoevo]

Diversification - in biology, the multiplication of biodiversity, in general. [Team ecoevo]

Diversification rate - rate of increase of biodiversity. This varies with unit/proxy; for example, we can understand diversification by studying lineages through time and in this case diversification is measured by a proxy or index, the number of evolutionary lineages accumlated through time within a clade. [Team ecoevo]

Ecological interaction - any relationship between two elements in an ecosystem. [Team ecoevo]

Ecological network - a set of interacting nodes (usually species) that are connected to one another edges. One example is food webs, in which links between species represent trophic interactions (Woodward et al., 2012 - TREE). [Team ecoevo]

Ecosystem - all actually or potentially interacting abiotic and biotic entities in a defined area, as well as their interactions. [Team ecoevo]

Edge - a connection between two nodes in a network. [Team ecoevo]

Interaction strength - the impact of one population (of one species) on the distribution, abundance, and/or body size of another population (species) important in the same community. Alternatively, this can be defined as the amount of flows of energy and matter interchanged between two or more elements within an ecosystem. [Team ecoevo]

Mutualism - an ecological interaction between at least two ecological entities whereby both entities benefit from the interaction (e.g. in terms of increases in fitness or population growth rates). [Team ecoevo]

Network - model for the relationships between the elements in a system defined by nodes and edges. [Team ecoevo]

Node - an individual element within a network. [Team ecoevo]

Phylogeny - a graphical representation of the hypothetical evolutionary relationships among a set of taxa derived from a set of characters. [Team ecoevo]

Phylogeography - the field of study concerned with the geographical distributions of genetic lineages within and among closely related species, studied to understand the processes driving population divergence and the impact of geographical phenomena on microevolution. [Team ecoevo]

Population - a set of individuals of the same species living in a defined area. [Team ecoevo]

Predation - an ecological interaction between two organisms whereby one organism consumes all or part of the other, so that it benefits at a cost to the other organism. [Team ecoevo]

Rate of evolution - formalized in molecular evolution as the number of DNA substitutions (mutations) per site per unit of time (e.g. years, millions of years). [Team ecoevo]

Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License