Chapter One - Biomonitoring for the 21st Century: Integrating Next-Generation Sequencing Into Ecological Network Analysis

https://doi.org/10.1016/bs.aecr.2017.12.001Get rights and content

Abstract

Ecological network analysis (ENA) provides a mechanistic framework for describing complex species interactions, quantifying ecosystem services, and examining the impacts of environmental change on ecosystems. In this chapter, we highlight the importance and potential of ENA in future biomonitoring programs, as current biomonitoring indicators (e.g. species richness, population abundances of targeted species) are mostly descriptive and unable to characterize the mechanisms that underpin ecosystem functioning. Measuring the robustness of multilayer networks in the long term is one way of integrating ecological metrics more generally into biomonitoring schemes to better measure biodiversity and ecosystem functioning. Ecological networks are nevertheless difficult and labour-intensive to construct using conventional approaches, especially when building multilayer networks in poorly studied ecosystems (i.e. many tropical regions). Next-generation sequencing (NGS) provides unprecedented opportunities to rapidly build highly resolved species interaction networks across multiple trophic levels, but are yet to be fully exploited. We highlight the impediments to ecologists wishing to build DNA-based ecological networks and discuss some possible solutions. Machine learning and better data sharing between ecologists represent very important areas for advances in NGS-based networks. The future of network ecology is very exciting as all the tools necessary to build highly resolved multilayer networks are now within ecologists reach.

Introduction

Traditionally, community ecology tends to focus on patterns of species richness and community composition, while ecosystem ecology focuses on fluxes of energy and materials. Ecological networks (sometimes called food webs for trophic interactions), however, provide a quantitative framework to combine these approaches and unify the study of biodiversity and ecosystem function (Thompson et al., 2012). Ecological networks, which describe which species are interacting with which (i.e. qualitative networks) as well as the strength of their interactions (i.e. quantitative networks), are now routinely used to understand ecosystem ‘robustness’ to species extinctions (Evans et al., 2013; Säterberg et al., 2013), quantify ecosystem services (Derocles et al., 2014a; Macfadyen et al., 2009) or examine the impacts of environmental change (Morris et al., 2015; Thompson and Gonzalez, 2017; Tylianakis et al., 2007). By using a burgeoning range of metrics to describe network structure, complexity and stability (see Arnoldi et al., 2016; Bersier et al., 2002; Donohue et al., 2013; Dunne et al., 2002a, Dunne et al., 2002b), ENA is considerably improving our understanding of ecology and evolution, with a growing number of applications for biomonitoring (Bohan et al., 2017; Gray et al., 2014). Indeed, ENA is increasingly being used to assess ecosystem response to environmental changes (e.g. climate change, pollution, invasive species; Aizen et al., 2008; Blanchard, 2015; Bohan et al., 2017; Thompson et al., 2016). There is consequently a growing shift in biodiversity monitoring away from conventional species and community-level descriptions towards a more comprehensive and mechanistic approach using species interaction networks (Bohan et al., 2013; Derocles et al., 2014a; Evans et al., 2013; Fontaine et al., 2011; Gray et al., 2014; Ings et al., 2009; Kéfi et al., 2012; Macfadyen et al., 2009; Pocock et al., 2012; Wirta et al., 2014).

Nevertheless, ecological networks can be difficult to construct with conventional approaches and suffer some major pitfalls mainly centred on sampling issues, taxonomic misidentification and/or incorrect species interactions (Evans et al., 2016, Gibson et al., 2011). Major errors occurring in either of these steps could ultimately affect network-level structural metrics and thus our understanding of ecosystem functioning (Novak et al., 2011). DNA-based methods (based on combined taxonomic identification and interaction data from DNA sequences) have the potential to overcome many of these issues, providing large, highly resolved, phylogenetically structured networks suitable for rapid and reliable biomonitoring (Bohan et al., 2017; Evans et al., 2016; Vacher et al., 2016, Valentini et al., 2009b).

Today, next-generation sequencing (NGS) or high-throughput sequencing (see Goodwin et al., 2016 for a review) can rapidly generate millions of DNA sequences. Sequences can describe, very precisely, not only the biodiversity present within an ecosystem, but also species interactions, the data from which can then be used to construct ecological networks (Evans et al., 2016). Recently, ecological network studies have taken advantage of NGS to successfully construct networks (e.g. Toju et al., 2014 for a plant–fungus network). Advances in statistical modelling and machine learning approaches bring a new opportunity to predict species interactions and rapidly build multilayer ecological networks from DNA sequences data generated with NGS (Vacher et al., 2016).

Despite species identification from DNA sequences commonly being seen as a universal way to identify species (Hebert et al., 2003), the NGS technology to build food webs is not applied uniformly in network ecology. Experimental designs (field sampling and molecular protocols) and the construction of ecological networks are heavily dependent on the ecosystem studied, and particularly on the type of interactions (see Box 1). Here, we distinguish two cases in particular. First, NGS can be directly used to build quantitative ecological interactions between organisms by resolving species interactions (e.g. Evans et al., 2016; Kitson et al., 2016; Piñol et al., 2014; Toju et al., 2013, Toju et al., 2014). This use of NGS data is, however, only possible in ecosystems in which relationships between organisms can clearly be established, such as host–parasitoid interactions where the parasitoid can be detected within the host (Derocles et al., 2014a, Derocles et al., 2015; Wirta et al., 2014), prey–predator interactions by detecting prey in gut contents (e.g. Piñol et al., 2014; Tiede et al., 2016) or faeces (Clare et al., 2014; Zeale et al., 2011; see Symondson and Harwood, 2014) and plant–pollinator interactions by using high-throughput sequencing to identify the pollen carried (Bell et al., 2017; Galimberti et al., 2014; Pornon et al., 2016; Sickel et al., 2015). Second, there are systems in which it is impossible (or logistically very problematic) to detect interactions between organisms and assessing whether these interactions are positive or negative, such as those within microbial (Jakuschkin et al., 2016) or planktonic communities (Lima-Mendez et al., 2015). For these systems, NGS approaches can only identify cooccurring species and their relative abundance. NGS data then need to be combined with theoretical approaches, including statistical modelling (Faust and Raes, 2012) or machine learning (Bohan et al., 2011a), to predict species interactions from their abundance patterns and finally to build ecological networks (Bohan et al., 2017; Kamenova et al., 2017; Vacher et al., 2016). These two ways of building ecological networks have their own specificities and challenges to overcome but also share common problems. These problems are related to (1) the qualitative and quantitative reliability of NGS data (i.e. polymerase chain reaction (PCR) bias and errors, sequencing bias and estimation of species abundances and frequency of interactions with number of NGS reads; Sommeria-Klein et al., 2016); (2) the identification of nodes and interactions in the network (inferring species interactions with statistical models when interactions are not directly resolved by molecular tools); (3) the costs of the sequencing technology and the expertise needed to process the data (Toju et al., 2013, Toju et al., 2014; Vacher et al., 2016).

Here, we bring new insights on how to integrate NGS and ENA into biomonitoring (Fig. 1). We first consider why ecological networks provide a suitable framework for a better understanding of biodiversity and ecosystem functioning and how they can be used to complement or supersede conventional biomonitoring approaches. Second, we underline the challenges that ecologists face in building ecological networks when DNA-based tools are not available (which represent the vast majority of food web studies in the literature). Third, we demonstrate how molecular methods, NGS in particular, can overcome (at least partially) the numerous constraints inherent in conventional network construction methodologies (e.g. taxonomic identification, insect rearing, fieldwork issues), while considering the challenges of using NGS tools for building networks. Fourth, we give insights on how to overcome NGS data issues and efficiently build networks through machine learning and data sharing. Finally, we discuss new areas of research and development centred on ENA of multilayer networks to ultimately create more resilient ecosystems.

Section snippets

Traditional Biomonitoring Is Typically Descriptive and Rarely Provides an Understanding of the Underlying Mechanisms Behind Ecosystem Functions

Biomonitoring of change lies at the core of ecosystem conservation, management and restoration. As biomonitoring is an obligation today, biomonitoring programs are framed by government organizations (e.g. European Commission, Joint Nature Conservation Committee in the United Kingdom). In its simplest form, biomonitoring consists of recording species diversity and abundances across different locations and times using a range of ecological census techniques and taxonomic identification. Most

Ecological Networks Can Be Challenging to Build Using Conventional Approaches

Despite their proven value in ecological research, networks are nevertheless limited by the difficulties of building them. These difficulties are centred around three major issues: (i) the sampling effort required to capture a significant range of species interactions; (ii) the reliable identification of specimens; and (iii) the adequate description of interactions between the organisms (see Box 1).

First, detecting the majority of species and their interactions within a network requires

Using NGS to Construct Ecological Networks

Currently, ecological networks constructed using DNA-based approaches are not used to regularly monitor ecosystems. This may be partially due to the historical reliance on classic field survey methods in network ecology, which rely on observation, specimen sampling, laboratory rearing and morphological identification to construct bipartite networks. Recent work has demonstrated that NGS can be rapid, universal and relatively cheap, in comparison to conventional (i.e. ‘traditional’ taxonomy

Learning Ecological Networks From Data

A basic premise of ecology is that the variation in the observed data, through sampling ecosystems, contains information about past and current ecological interactions between species in the ecosystem. Thus, the abundance and variation of any one species are in part a consequence of past interactions between the individuals of previous generations, such as sexual reproduction, and of the current generation, such as competition, cannibalism or migration. Across a community the observed

NGS Network Data Sharing

With NGS approaches, network ecologists will be able to rapidly generate a large amount of data, including both DNA sequences and putative species interactions (interactions derived from machine learning must be validated through observation and experiment, see Lima-Mendez et al., 2015). This means that data curation and sharing between researchers must be done systematically for both DNA sequences and species interactions. Currently, sequences data are available publicly to researchers through

Towards Larger, Highly Resolved Networks

The recent conceptual revolution in ecology brought forth by the development of metacommunity and metacosystem theory (Holyoak et al., 2005; Leibold et al., 2004; Loreau et al., 2003) has stressed the importance of spatial processes for the functioning, diversity, complexity and dynamics of ecological systems (Massol et al., 2011). Recent theoretical findings have highlighted the role of the spatial structure and scales of communities, food webs and ecosystems to understand properties such as

Acknowledgements

We are grateful to Frédéric Barraquand (BIOGECO, INRA, Univ. Bordeaux, Pessac, France) for his contribution to the manuscript and the figures. We acknowledge the support of the projects FACCE SURPLUS PREAR and ANR-17-CE32-0011. We thank Michael J.O. Pocock (Centre for Ecology & Hydrology, Wallingford, United Kingdom) for his help in defining the direction of the manuscript. We thank Eleanor Collinson (School of Natural and Environmental Sciences, Newcastle University, United Kingdom) for her

Glossary

Molecular Biology

Amplicon
a DNA fragment amplified by primers during the polymerase chain reaction (PCR).
Diagnostic PCR
specific primer pairs are designed for each targeted species or higher taxonomic group to produce amplicons of different sizes. These primers are then used in a multiplex and/or several singleplex PCRs for taxonomic identification. Identification of taxonomic groups in samples is then based on the presence/absence of bands on an electrophoresis gel (taxonomic group is present or not) as well as

References (300)

  • M. Elias et al.

    Evolutionary history and ecological processes shape a local multilevel antagonistic network

    Curr. Biol.

    (2013)
  • K. Faust et al.

    Metagenomics meets time series analysis: unraveling microbial community dynamics

    Curr. Opin. Microbiol.

    (2015)
  • S. Gaba et al.

    Weed species richness in winter wheat increases with landscape heterogeneity

    Agric. Ecosyst. Environ.

    (2010)
  • N.A. Abreu et al.

    Decoding molecular interactions in microbial communities

    FEMS Microbiol. Rev.

    (2016)
  • M.T. Agler et al.

    Microbial hub taxa link host and abiotic factors to plant microbiome variation

    PLoS Biol.

    (2016)
  • M.A. Aizen et al.

    Invasive mutualists erode native pollination webs

    PLoS Biol.

    (2008)
  • T.J. Albert et al.

    Direct selection of human genomic loci by microarray hybridization

    Nat. Methods

    (2007)
  • S. Allesina et al.

    A competitive network theory of species diversity

    Proc. Natl. Acad. Sci. U.S.A.

    (2011)
  • S. Allesina et al.

    Stability criteria for complex ecosystems

    Nature

    (2012)
  • S. Allesina et al.

    A general model for food web structure

    Science

    (2008)
  • M. Almeida-Neto et al.

    A consistent metric for nestedness analysis in ecological systems: reconciling concept and measurement

    Oikos

    (2008)
  • P. Amarasekare et al.

    Mechanisms of coexistence in competitive metacommunities

    Am. Nat.

    (2004)
  • M. Andrello et al.

    Low connectivity between Mediterranean marine protected areas: a biophysical modeling approach for the Dusky Grouper Epinephelus marginatus

    PLoS One

    (2013)
  • J. Astegiano et al.

    The robustness of plant-pollinator assemblages: linking plant interaction patterns and sensitivity to pollinator loss

    PLoS One

    (2015)
  • M. Bálint et al.

    Millions of reads, thousands of taxa: microbial community structure and associations analyzed via marker genes

    FEMS Microbiol. Rev.

    (2016)
  • M.A. Barnes et al.

    The ecology of environmental DNA and implications for conservation genetics

    Conserv. Genet.

    (2015)
  • J. Bascompte et al.

    Plant-animal mutualistic networks: the architecture of biodiversity

    Annu. Rev. Ecol. Evol. Syst.

    (2007)
  • J. Bascompte et al.

    The nested assembly of plant-animal mutualistic networks

    Proc. Natl. Acad. Sci. U.S.A.

    (2003)
  • K.L. Bell et al.

    Applying pollen DNA metabarcoding to the study of plant–pollinator interactions

    Appl. Plant Sci.

    (2017)
  • K.C. Beng et al.

    The utility of DNA metabarcoding for studying the response of arthropod diversity and composition to land-use change in the tropics

    Sci. Rep.

    (2016)
  • L.F. Bersier et al.

    Quantitative descriptors of food-web matrices

    Ecology

    (2002)
  • J. Binladen et al.

    The use of coded PCR primers enables high-throughput sequencing of multiple homolog amplification products by 454 parallel sequencing

    PLoS One

    (2007)
  • S. Biswas et al.

    Learning microbial interaction networks from metagenomic count data

    J. Comput. Biol.

    (2016)
  • J.L. Blanchard

    Climate change: a rewired food web

    Nature

    (2015)
  • N. Blüthgen et al.

    Measuring spaecialization in species interaction networks

    BMC Ecol.

    (2006)
  • N. Blüthgen et al.

    What do interaction network metrics tell us about specialization and biological traits?

    Ecology

    (2008)
  • D.A. Bohan et al.

    Automated discovery of food webs from ecological data using logic-based machine learning

    PLoS One

    (2011)
  • D.A. Bohan et al.

    National-scale regulation of the weed seedbank by carabid predators

    J. Appl. Ecol.

    (2011)
  • L. Bolchoun et al.

    Spatial topologies affect local food web structure and diversity in evolutionary metacommunities

    Sci. Rep.

    (2017)
  • F. Briand et al.

    Community food webs have scale-invariant structure

    Nature

    (1984)
  • A.W. Brooks et al.

    Phylosymbiosis: relationships and functional effects of microbial communities across host evolutionary history

    PLoS Biol.

    (2016)
  • V. Bucci et al.

    MDSINE: Microbial Dynamical Systems INference Engine for microbiome time-series analyses

    Genome Biol.

    (2016)
  • V. Calcagno et al.

    Constraints on food chain length arising from regional metacommunity dynamics

    Proc. R. Soc. B

    (2011)
  • E. Canard et al.

    Emergence of structural patterns in neutral trophic networks

    PLoS One

    (2012)
  • J. Cavender-Bares et al.

    The merging of community ecology and phylogenetic biology

    Ecol. Lett.

    (2009)
  • K. Cazelles et al.

    A theory for species co-occurrence in interaction networks

    Theor. Ecol.

    (2016)
  • N.P. Chacoff et al.

    Evaluating sampling completeness in a desert plant-pollinator network

    J. Anim. Ecol.

    (2012)
  • A. Chao

    Nonparametric estimation of the number of classes in a population

    Scand. J. Stat.

    (1984)
  • M. Charlery de la Masselière et al.

    Changes in phytophagous insect host ranges following the invasion of their community: long-term data for fruit flies

    Ecol. Evol.

    (2017)
  • V.B. Chaudhary et al.

    MycoDB, a global database of plant response to mycorrhizal fungi

    Sci. Data

    (2016)
  • Cited by (0)

    View full text