Recent breakthroughs in metabolomics promise to reveal the cryptic chemical traits that mediate plant community composition, character evolution and lineage diversification

I. Introduction

Seminal hypotheses in community ecology and evolutionary biology propose a fundamental role for plant chemical defenses against herbivores and pathogens in the evolutionary origins and ecological coexistence of plant species. Whereas nearly all plants require a small number of shared resources (water, light, CO₂ and nutrients), plant interactions with natural enemies provide a highly multidimensional space within which species can carve out a distinct niche defined by the enemies they support, and those they avoid. Gillett (1962) first proposed that plants build up host-specific natural enemies where they are abundant, impeding their fitness relative to competitors that manage to avoid sharing the enemy. Janzen (1970) and Connell (1971) proposed that the resulting negative density-dependent recruitment is a primary driver of plant species coexistence in tropical forests. Ehrlich & Raven (1964) extended the concept to macroevolution, proposing that the ecological success conferred by novel defenses against natural enemies facilitates speciation, and hence lineage diversification, in plants (and in their enemies; see Box 1).

Many thousands of secondary metabolites influence interactions between plants and herbivores and pathogens. For example, the bean family (Fabaceae) alone synthesizes thousands of compounds from nearly 20 major chemical classes (Wink & Mohamed, 2003). The sheer number of secondary metabolites of unknown structure has long precluded comparative metabolomics, the comparison of small-molecule metabolite profiles, at the large taxonomic scales required for the study of macroevolution and community ecology (Hilker, 2014). However, recent advances in tandem mass spectrometry (MS/MS) and nuclear magnetic resonance (NMR) spectroscopy bioinformatics make chemical analysis at community and macroevolutionary scales possible by enabling comparison of the structure of unknown compounds from crude extracts of chemically complex biological samples.

Here, I illustrate the potential for MS/MS and NMR structural metabolomics to permit unprecedented tests of the hypotheses that plant secondary chemistry (1) determines host ranges of pathogens and herbivores and thereby (2) facilitates species coexistence in hyperdiverse plant communities such as tropical forest tree communities and (3) promotes the evolutionary diversification of plants due to the fitness advantage conferred by the evolution of novel chemical defenses.

II. Recent innovations in structural metabolomics

Until recently, untargeted metabolomic profiling of chemically complex samples has required the isolation of individual compounds for manual structural determination. By contrast, innovations in bioinformatics use NMR and MS/MS to compare the structures of unknown compounds (Table 1). Coupled, automated separation and detection methods, such as liquid chromatography (LC)-solid-phase extraction (SPE)-MS/NMR, have enhanced the efficiency of the collection of MS and NMR spectra from chemically complex samples (Moco & Vervoort, 2012). The interpretation of NMR spectra allows unequivocal determination of molecular structure, which allows compounds to be classified based on their structural scaffold or by metabolic pathway in well-studied organisms (Wetzel et al., 2007). However, the isolation of individual compounds of interest, such as with SPE, is a rate- and cost-limiting step even when automated at micro-volumes. For these reasons, applications of metabolomics to understanding the role of plant secondary chemistry in community ecology and macroevolution have been limited. Recent advances in both MS/MS and NMR bioinformatics, however, make high-throughput chemical analysis at the scale of a species-rich ecological community or phylogenetic clade possible by quantifying the structural similarity of samples that are complex mixtures of compounds of unknown structure.

Recent innovations in NMR bioinformatics enable the quantification of molecular structural diversity in complex samples of unknown compounds (Richards et al., 2015). The NMR method begins with ¹H-NMR spectra collected from crude tissue extracts. The diversity of ¹H-NMR resonances in a spectrum is indicative of molecular structural (scaffold and moiety) diversity. In pairwise comparisons of NMR spectra, an unique ¹H-NMR resonance indicates the presence of a unique moiety in one of the samples, and the similarity of crude-extract spectra for two plant species can be interpreted as the chemical structural similarity of their metabolomes (Table 1).

MS/MS enables comparisons of the chemical structure of compounds because molecules with similar structures fragment into many of the same sub-structures (Fig. 1a,b). Thus, structural similarity can be quantified for thousands of pairs of unknown compounds by measuring the cosine of the angle between vectors that represent the mass-to-charge ratio (m/z) of the constituent fragments (Table 1; Watrous et al., 2012). In addition, comparison of MS/MS spectra with publically available spectral libraries can identify unknown molecules (‘dereplication’; Allard et al., 2016; Wang et al., 2016). More importantly, this method enables the quantification of molecular similarity even for samples in which few compounds are unambiguously identified, permitting chemical ecology in understudied and species-rich plant communities such as tropical forests.

The structural comparison of unknown molecules using MS/MS is scalable to datasets containing hundreds of samples and tens of thousands of unique molecules (Watrous et al., 2012; Wang et al., 2016). Visualization and analysis is aided by the organization of pairwise MS/MS similarities into molecular networks, in which nodes represent compounds and links between nodes represent similarity scores (Fig. 1c). Furthermore, molecular networks aid in the classification of unknown molecules that are linked to compounds that match annotated spectra in public libraries (Wang et al., 2016).

The vast diversity of plant secondary metabolites and the rarity of any particular compound among species have long stymied ecologists and evolutionary biologists. Structural metabolomic methods address these problems by accounting for the structural similarity of unknown compounds that are not shared between species when quantifying chemical similarity (Table 1). Richards et al. (2015) developed a chemical diversity index (CDI) based on crude ¹H-NMR spectra that reflects both inter- and intramolecular moiety diversity. Sedio et al. (in press) developed a chemical structural-compositional similarity (CSCS) metric that weights the structural similarity of every pair of compounds in a network by their relative ion intensity in two plant species. Conventional methods of calculating similarity in ecology, such as Bray–Curtis similarity, consider the relative abundance of shared compounds, but ignore structural relationships between molecules. By contrast, both CDI and CSCS account for the presence of structurally similar compounds that are not shared between species or samples. A simple example illustrates the implications. Compounds x and y are structurally similar, species A contains compound x but not y, and species B contains y but not x. In this example, compounds x and y make no contribution to Bray–Curtis similarity, but make a positive contribution to CDI and CSCS.

Furthermore, the proximity of compounds in an MS/MS molecular network can be used to quantify structural scale, from pairs of highly structurally similar compounds that share a direct link, to subnetworks of compounds with shared structural scaffolds, to large clusters that may correspond to chemical classes with shared structural elements (Fig. 1h). By enabling the high-throughput structural comparison of unknown molecules, structural metabolomics promises unprecedented insight into the secondary-chemistry niches hypothesized to generate and sustain plant diversity.

III. Species coexistence

Can the number of niches defined by secondary metabolites and their impact on plant enemies approach the number of coexisting tree species in a tropical forest? Or, is much of the variation in secondary chemistry and other defenses redundant in the eyes of plant enemies? Biologists have accumulated 40 yr of evidence in support of the predictions of Gillett (1962), Janzen (1970) and Connell (1971) that seeds and seedlings experience reduced survival and recruitment in the vicinity of conspecific adults (reviewed in Comita et al., 2014). Density-dependent suppression of conspecific individuals suggests that natural enemy host ranges are sufficiently narrow to ensure that enemy-mediated competition (Holt, 1977) is greater among conspecific than between heterospecific individuals, the definition of niche differentiation and a prerequisite for coexistence (Chesson & Kuang, 2008).

However, many herbivores (Novotny et al., 2002) and pathogens (Gilbert & Webb, 2007) are not strict specialists. Generalist plant enemies, even those with narrow host ranges, are expected to mediate competitive exclusion of shared hosts, effectively limiting the co-occurrence of species that are chemically similar and promoting chemical diversity in the plant community (Fig. 1f,g; Sedio & Ostling, 2013).

Secondary chemistry exhibits phylogenetic signal at broad phylogenetic scales (Wink, 2003). However, secondary chemistry is often not conserved within species-rich genera, and closely related species can differ dramatically chemically in Bursera (Becerra, 1997), Inga (Kursar et al., 2009), Protium (Fine et al., 2013), Solanum (Haak et al., 2014), Piper (Richards et al., 2015; Salazar et al., 2016), Eugenia, Ocotea and Psychotria (Sedio et al., in press). Furthermore, co-occurring species of Bursera (Becerra, 2007), Inga (Kursar et al., 2009) and Piper (Salazar et al., 2016) are less similar chemically than by chance, suggesting that niche partitioning based on defense compounds stabilizes coexistence among species in these genera.

Structural metabolomic methods make it possible to move beyond individual genera to study chemical ecology at the scale of whole communities. In addition, recent developments in DNA barcoding (Garcia-Robledo et al., 2013) and microbial metagenomics (Barberán et al., 2015) enable determination of plant–insect and plant–microbe associations at large community scales. The standardized application of metabolomics and metagenomics across multiple sites can facilitate climatic, latitudinal and biogeographical comparisons. To implicate herbivores and/or pathogens in plant species coexistence, one might statistically infer the relative explanatory power of chemical traits in determining host use patterns of natural enemies (Fig. 1e), and similarly, infer the power of local neighborhood densities of those traits (or mutual natural enemies themselves) in determining plant performance and recruitment (Fig. 1f,g).

Structural metabolomic data can be used to infer the chemical traits that determine plant–enemy associations by modeling natural enemy host use as a multinomial distribution over potential host plants, with a probability vector that is a function of plant chemistry (Fig. 1e). Alternatively, models used in genome-wide association studies to identify loci associated with discrete phenotypes out of thousands of candidate loci could be modified to instead identify compounds (analogous to loci) associated with the presence or absence of particular natural enemies (analogous to phenotypes) out of thousands of candidate compounds. Similarly, the prediction that certain chemical traits are associated with density-dependent neighborhood effects on recruitment or mortality (Fig. 1f,g) can be tested by drawing on a deep literature relating traits to density-dependent dynamics and community structure (e.g. Kraft et al., 2008; Comita et al., 2010; Pollock et al., 2012; Sedio & Ostling, 2013; Lebrija-Trejos et al., 2016; Salazar et al., 2016).

Alternative models of natural enemy host associations or of density-dependent performance might consider chemical ‘traits’ that represent MS/MS subnetworks or NMR functional groups at various degrees of inclusiveness, from scales at which groups represent broad compound classes (e.g. xanthine-derivative alkaloids) to much narrower groups of structurally similar compounds (e.g. caffeine and theobromine), as illustrated in Fig. 1(h). Such model comparisons could reveal the scale at which chemical structural variation influences herbivore and microbe host associations and at what scale defensive compounds are functionally redundant.

IV. Character evolution and lineage diversification

Ehrlich & Raven (1964) first proposed that diversification in plants is often the result of innovation in defenses against natural enemies. Their hypothesis is referred to as the ‘Escape and Radiate’ Hypothesis because it envisions the evolution of a novel defense (e.g. the two-compound cluster in Fig. 2a), subsequent ecological success as the plant population grows unchecked by natural enemies, followed by speciation, and ultimately diversification of many species descended from the original plant species (Fig. 2b; Schluter, 2000). Two mutually compatible predictions follow. Species richness or the rate of diversification (speciation minus extinction) in phylogenetic clades should be associated with the evolution of key innovations in defense. And, variation in rates of lineage diversification over a plant phylogeny should be associated with variation in rates of defense evolution (Fig. 2b).

There have been two tests of the key innovation hypothesis. Farrell et al. (1991) demonstrated that plant lineages that exude latex from damaged tissue exhibit greater species richness than sister lineages that lack latex, suggesting that latex is a key innovation associated with adaptive radiations in plants. Similarly, Weber & Agrawal (2014) found that plant lineages in which some species use extra-floral nectaries to recruit ants to their defense show greater rates of diversification than sister lineages that lack nectaries. Latex and extra-floral nectaries are recognizable characters in distantly related lineages of plants. Secondary chemistry has proved a more challenging subject for macroevolutionary analyses. The phylogenetic rarity of any particular compound makes it difficult to identify potential key innovations. Likewise, the astonishing diversity of compounds that plants deploy as defenses makes it difficult to identify comparable characters to compare rates of character evolution across plant lineages. For these reasons, the predictions of Ehrlich & Raven (1964) have remained largely untested with respect to secondary chemistry at taxonomic scales beyond closely related congeners (e.g. Agrawal et al., 2009).

Structural metabolomics promises to open a new frontier in the study of chemical macroevolution by providing a common currency by which to measure character evolution of compounds of unknown structure in distantly related lineages. The hyperdiverse tree genera Inga, Piper and Psychotria deploy distinct chemical classes in defense (Kursar et al., 2009; Richards et al., 2015; Sedio et al., in press). In an MS/MS molecular network or an NMR metabolic profile, interspecific variation in unknown phenolic compounds of Inga, sesqui- and tri-terpenes of Piper, and indole and pyridone alkaloids of Psychotria are quantified on a comparable scale (e.g. using CDI or CSCS; Richards et al., 2015; Sedio et al., in press).

Rates of chemical evolution can easily be quantified by measuring phylogenetic independent contrasts (PICs; Felsenstein, 1985) in terms of CSCS (Sedio et al., in press) for every branch in a phylogeny. Bayesian phylogenetic comparative methods, such as the Bayesian Analysis of Macroevolutionary Mixtures (BAMM; Rabosky, 2014), can identify the location and number of key shifts in the rate of diversification or character evolution without a priori predictions. Because MS/MS data are increasingly shared through repositories such as the Global Natural Products Social (GNPS) Molecular Networking database (Wang et al., 2016), the availability of metabolomic data and appropriate phylogenetic comparative methods will allow unprecedented tests of the role of secondary chemistry in the evolutionary origins of plant diversity at global scales.

V. Conclusions

Seminal hypotheses that interspecific variation in secondary chemistry enables species coexistence and drives evolutionary diversification in plants have remained largely untested due to an inability to measure chemical traits comprehensively at scales appropriate for studies of community ecology and macroevolution. Advances in MS/MS (Watrous et al., 2012; Wang et al., 2016) and NMR (Richards et al., 2015) bioinformatics enable comparative structural metabolomics, that is, the ability to quantify the structural similarity of thousands of unknown secondary compounds in hundreds of species at a time. The tools now exist to reveal the cryptic chemical traits that were hypothesized to drive global patterns of diversity among communities and among evolutionary lineages of plants more than half a century ago (Gillett, 1962; Ehrlich & Raven, 1964; Janzen, 1970; Connell, 1971).