Facing the Future of Plant-Insect Interaction Research: Le Retour à la “Raison d'Être”

The interaction between herbivorous insects and the angiosperm plants they consume, collectively constituting the majority of macroscopic species in terrestrial communities, has often been metaphorically likened to warfare (e.g. Gonzalez and Nebert, 1990), and the process of reciprocating defense and counter defense has been called a coevolutionary arms race (Whittaker and Feeny, 1971). Few people probably realize, however, that this particular field of study arose as a direct consequence of an actual, rather than metaphorical, war. When World War II broke out, the Department of Zoology and Applied Entomology at Imperial College, London, was relocated to Slough and the Pest Infestation Laboratory was founded to aid in the war effort. Insect physiologist Gottfried Fraenkel undertook a study of the nutrition of stored product pests in an effort to gain insight into how to control them. In the process, he determined that, with only a few exceptions, human and insect nutritional requirements are essentially the same and that, moreover, the majority of green plants are essentially nutritionally equivalent (Fraenkel, 1953). This observation in turn led to speculation that nonnutritious substances, the so-called secondary substances idiosyncratically distributed throughout the plant kingdom, determine patterns of host plant utilization—“the enormous variety in the distribution and composition of the secondary plant substances, for which no comprehensive and plausible explanation then existed, accounted for the equally staggering variety of insect-food-plant relationships, by their acting as repellents and attractants for insects and other organisms” (Fraenkel, 1984, p. 1).

Fraenkel's revolutionary new idea was first advanced in a lecture at an international zoological congress in 1953 held in Copenhagen, but it did not reach the mainstream press until 1959, when he authored a now-famous article titled “The raison d'être of secondary plant substances” in the journal Science. The idea that the reason plants manufacture such a diversity of secondary metabolites is to defend themselves against insects and other herbivores was in fact slow to catch on; the article was cited fewer than 12 times between 1959 and 1964. Attention was refocused on the article, however, by Ehrlich and Raven (1964), who expanded on the idea with a phylogenetic analysis of butterflies and plants and suggested that coevolution (the process Fraenkel had earlier called “reciprocal adaptive radiation”) was responsible not only for the tremendous diversification of plant secondary substances, but also for the diversification of angiosperm plants and the insects that eat them.

Technically, then, the 50th anniversary of the concept of chemical coevolution between plants and insects passed without notice in 2003. Inasmuch as the idea did not gain much traction until 1964, perhaps a more appropriate 50th anniversary would be 2014; yet 2009 marks 50 years of the concept of chemical coevolution in the refereed mainstream literature. Thus, as the year approaches it seems timely to evaluate the progress that has been made, particularly in recent years, and to identify the challenges that remain.

To make his case in 1959, Fraenkel concentrated his argument on a handful of plant families for which information was available relating to their secondary chemistry and to their ecological associations with insect herbivores. These included the Cruciferae, Umbelliferae, Leguminosae, Solanaceae, Moraceae, and Gramineae (although Cruciferae, Umbelliferae, Leguminosae, and Gramineae, for reasons of taxonomic consistency, are now known as Brassicaceae, Apiaceae, Fabaceae, and Poaceae, respectively). At the time, most of the evidence for chemical mediation of interactions between plants and insects was based on interactions involving species in these families. For the most part, five of these families have remained the focus of studies of plant-insect interactions since that time; the sole family whose star never fully reached its ascendancy was the Moraceae. By the end of the century, principal unresolved issues in chemical coevolution of plants and herbivores concerned the magnitude of costs of induced and constitutive defense, the relative importance of top-down and bottom-up selection pressures, the frequency of diffuse, specific, and escape-and-radiate coevolution (Thompson, 1994), the impacts of the abiotic environment on the outcome of plant-herbivore interactions, and the influence of the community matrix (including microbes and fungi) on plant-herbivore interactions. Studies of five of the families that inspired Fraenkel (1959) have contributed significantly to illuminating these issues.

The first definitive demonstration of a behavioral impact of a plant secondary chemical, cited by Fraenkel (1959) as the “first detailed description of a chemical insect-host plant relationship” (p. 1467), was the report by Verschaffelt (1911) of the ability of sinigrin, a mustard oil glycoside from Brassicaceae, to stimulate feeding by pierid caterpillars. The Cruciferae/Brassicaceae subsequently provided the foundation for much of coevolutionary theory—serving as the inspiration for the resource concentration hypothesis of Root (1973) and forming one end of the continuum in apparency theory as propounded by Feeny (1975). As more taxa of both crucifers and insect herbivores were examined, it became clear that chemicals other than glucosinolates also play an ecological role and that host plant assessment is a function of a “balance of positive and negative chemical stimuli” (Renwick and Radke, 1987, p. 1771) as Fraenkel (1959) had suggested. Among the first studies documenting costs of resistance and herbivore selection pressure on secondary metabolite production were those that involved glucosinolates in crucifers (Mauricio and Rausher, 1997; Mauricio, 1998).

Although ecologists have historically eschewed the use of model organisms, the crucifers and their insect associates occupy an unusual position within chemical ecology. Among other things, the powerful behavioral impacts of glucosinolates as allomones (antagonists or deterrents) for generalists and kairomones (cues) for specialists have been difficult to ignore and the availability of herbaceous species that are easy to cultivate (e.g. collards) and that grow rapidly made them experimentally manageable. Moreover, in the year 2000 the family Brassicaceae provided the first plant species with a sequenced genome, making experimental analysis of evolutionary phenomena more feasible. Arabidopsis (Arabidopsis thaliana), mouse-ear cress, was a logical choice for sequencing—its low chromosome number was determined in 1907 and due to the availability of mutants and ecotypes it had been promoted as a model organism as early as 1943 (Meyerowitz, 2001). The genome itself, comprising about 157 Mb on five chromosomes, made it an attractive candidate for sequencing, particularly once T-DNA-mediated transformation methods had been perfected.

Since the genome sequence was published in 2000, Arabidopsis has proved to be an outstanding model system for the study of plant-insect interactions at genetic and molecular levels (Mitchell-Olds, 2001). The fundamental signal transduction pathways involving jasmonate, salicylate, and ethylene have been documented to occur in this species (Schenk et al., 2000; Stotz et al., 2000), making the plant well suited for examining mechanisms characterizing early responses to herbivory that may be shared across many groups. Other fundamental findings include demonstrating a modular genetic system for regulation of selectable variation in glucosinolate production (Kliebenstein et al., 2001), genetic variation in multiple signal transduction pathways regulating glucosinolate biosynthesis (Kliebenstein et al., 2002), and differential biosynthetic and transcriptional responses to specialist and generalist herbivory by insects in at least two orders (Stotz et al., 2000; de Vos et al., 2007; Kusnierczyk et al., 2007).

Interactions between solanaceous plants and their herbivores, also important to Fraenkel (1959), have continued to feature prominently in studies of plant-insect interactions. The role of nicotine as an allomone was long established (by virtue of its longstanding use as a commercial insecticide) and Fraenkel himself, working with his students, demonstrated the importance of solanaceous chemicals as kairomones (Yamamoto and Fraenkel, 1960). Two domesticated plants in the family, petunia (Petunia hybrida) and tobacco (Nicotiana tabacum), were important in early molecular genetic studies because they could be easily transformed. Since the turn of the century, Baldwin and coworkers have developed new molecular techniques to create a genetically manipulable native species, Nicotiana attenuata (Krügel et al., 2002). Enormous strides have been made with this system (Kessler and Baldwin, 2002), including documentation of the defensive function of herbivore-induced plant volatile emissions in attracting natural enemies and repelling herbivores (Kessler and Baldwin, 2001), trypsin protease inhibitor production (Glawe et al., 2003), and nicotine production (Steppuhn et al., 2004), fitness costs of induced resistance (Heil and Baldwin, 2002), herbivore-specific transcriptional responses in plants (Voelckel and Baldwin, 2003, 2004), overlapping transcriptional responses to biotic and abiotic stresses (Izaguirre et al., 2003), costs and benefits of proteinase inhibitor production (Zavala and Baldwin, 2004; Zavala et al., 2004), individual variability in herbivore-specific elicitors (Roda et al., 2004), and impacts of nitrogen availability on defense chemistry (Lou and Baldwin, 2004; Pearse et al., 2006).

Species in the Poaceae, i.e. grasses and their relatives, have also been the subjects of groundbreaking studies in understanding the chemical mediation of interactions between plants and insects, notably the release of volatile organic compounds in response to herbivore damage that serve as attractants to natural enemies of those herbivores (Turlings et al., 1995), the role of insect-derived elicitors in stimulating the production of these signal compounds (Alborn et al., 2000; Turlings et al., 2000), the repellent function of damage-induced volatiles in conspecific herbivores (deMoraes et al., 2001), and the involvement of defense signaling substances in inducing damage volatiles (Schmelz et al., 2003).

The Fabaceae, yet another family used by Fraenkel (1959) to make his case, has figured prominently as well in some of the more spectacular advances in the field in the past half century. That many defenses of legumes are proteinaceous made these plants attractive for studies of chemical coevolution in that such defenses are the direct products of protein-coding genes; moreover, such forms of resistance were attractive candidates for genetically engineering pest-protected crop plants. A substantial body of literature accumulated documenting the production of proteinase inhibitors and the counter adaptations to these defenses (Murdock et al., 1988; Chrispeels and Raikhel, 1991; Shade et al., 1994). As well, studies of lima beans (Phaseolus lunatus) and spider mites compellingly demonstrated the role of plant volatiles, including widely conserved defense signaling substances, as signals to third trophic level enemies of herbivores (Dicke et al., 1990; Dicke and Sabelis, 1990; Vet and Dicke, 1992; de Boer et al., 2005).

Thus, great progress has been made in the past five decades in elucidating the ways in which secondary substances mediate interactions between insect herbivores and their host plants. Key to that progress were the organisms serving as models; these proved particularly useful in elucidating many fundamental processes, particularly in the past decade. Plant signaling is a case in point. Dozens of studies have documented that herbivores initiate a chain reaction that begins with damage and/or the introduction into the wound site of herbivore-specific elicitors, detection by the plant, and activation of one or more signaling cascades. These cascades typically involve oxylipin signaling, which influences the physiological plant response, leading to upregulation of defensive biosynthetic pathways. As well, on the arthropod side of the interaction, oxylipin signaling influences herbivore feeding behavior and effects ecological responses on the part of the natural enemies of those herbivores (Baldwin et al., 2001).

The problem with the use of model organisms, however, is that they can provide useful information about universal attributes of a phenomenon, such as the basic physiological mechanisms underlying early responses to herbivory, but they are limited in terms of their ecological value. In redirecting attention to the work of Verschaffelt (1910), for example, Fraenkel (1959) sent several generations of ecologists searching for “sign stimuli”—single compounds that can elicit oviposition or feeding in the way that the glucosinolate sinigrin appears to influence pierid butterflies. For species that sequester host plant toxins, such singular stimuli can be sufficient (Rees, 1969; Chyb et al., 1995; Roessingh et al., 1997; Bernays et al., 2002; Schoonhoven and van Loon, 2002). But in other interactions (and even in some interactions involving Brassicaceae; Nielsen et al., 1979), combinations of chemicals proved necessary for eliciting oviposition and feeding (Nishida, 2005). Even the mechanistic basis for defense signaling may not be as highly conserved as has been assumed (Agrawal, 2005).

That secondary chemicals are idiosyncratically distributed is key to their role in evolutionary diversification of plants and insects. As Ehrlich and Raven (1964) state, it is only in the context of plant-insect interactions that “the irregular distribution in plants of such chemical compounds of unknown physiological function as alkaloids, quinines, essential oils (including terpenoids), glycosides (including cyanogenic substances and saponins), flavonoids, and even raphides…is immediately explicable” (p. 602). Adding to the challenge of understanding “irregular distribution” is the fact that new technologies—notably, liquid chromatography-mass spectrometry and gas chromatography-mass spectrometry—have dramatically enhanced the capacity to detect, inventory, and catalog those irregularities in need of ecological explanation. Moreover, whereas molecular modeling of proteins and bioinformatics have made prediction of gene function and biochemical activity at least feasible, predicting ecological activity based on chemical structure, where there are tens of thousands of structures to account for in the context of a virtually limitless number of possible biotic interactions, remains, to say the least, a challenge.

Given the utter centrality of idiosyncracy to the nature of chemical coevoluton, the enormous literature of plant-insect interactions still has a precariously narrow base. The majority of articles in this field have been and still are based on a handful of plant families, of chiefly agricultural interest. The richness of the plant-herbivore literature bears no relationship to the species or chemical richness of plant families (Table I  

As a case in point, the evolution of glucosinolate biosynthesis is a textbook example of the coevolutionary process as proposed by Ehrlich and Raven (1964). Molecular phylogenetic studies have unambiguously determined that glucosinolate biosynthesis has arisen only twice (Rodman et al., 1998). Substantial progress has been made, based on the model species Arabidopsis, in understanding this pathway, including, notably, documentation of the roles of gene duplication, neofunctionalization, and positive selection on methylthioalkylmalate synthases encoded at the methylthioalkylmalate synthase gene cluster in generating glucosinolate diversity (Benderoth et al., 2006).

Glucosinolates, however, may not be the sole model for understanding the evolution and diversification of plant secondary metabolites. Some groups of secondary compounds of demonstrable defensive value may be distributed apparently erratically in a small number of not particularly closely related families. Furanocoumarins, for example, a group of phenylpropanoid derivatives, are distributed in a phylogenetically mystifying pattern (Fig. 1  

Figure 1.

A phylogeny of the major plant groups (APG II, 2003) with occurrences of two classes of secondary compounds (based on Rodman et al. [1998] and Murray et al. [1982]) superimposed. Families containing furanocoumarins include Fabaceae (Fabales), Meliaceae and Rutaceae (Sapindales), Moraceae and Rosaceae (Rosales), Apiaceae and Pittosporaceae (Apiales), and Asteraceae (Asterales). Asterisked orders represent individual reports of furanocoumarins from a single species. One report in Cyperales is not shown.

Open in new tabDownload slide

). These compounds have been reported throughout the Apiaceae (55 genera) and Rutaceae (31 genera), but they are found in only one or two species in 10 families and in two genera in two other families (Murray et al., 1982). As clear as it is that glucosinolate biosynthesis evolved only twice, it is equally clear that there are multiple, independent origins of furanocoumarins. These compounds occur in two major clades of flowering plants (rosids and asterids) and within each of these major clades they have evolved at least three times. So, while it can be argued that the ancestors of the Apiales and Asterales shared a common origin, the distribution of these compounds within each of these orders is such that they must have occurred multiple times even in lineages that are closely related and ostensibly by a process different from that underlying the diversification of glucosinolates.

Although rare occurrences in those 10 families may be attributable to errors of compound or plant identification, the fact that in three of these families more than one furanocoumarin was found (by more than one investigator) reduces this possibility. The question then arises as to why there has been no adaptive radiation within those families of furanocoumarin-containing species of the sort undergone by glucosinolate-synthesizing plants. One answer may be that furanocoumarins present a different set of challenges to the plants that produce them. Compartmentalization of inert substrates and activating enzymes—the “mustard oil bomb”—appears to be a shared mechanism for reducing risks of autotoxicity (Rodman et al., 1998). No such system appears to have accompanied the evolution of many other secondary substances. Unleashed from their typical confines in the plant, many plant secondary substances (e.g. nicotine and terpenes) are toxic to their producers (Baldwin and Callahan, 1993; Gog et al., 2005). Consequently, plants that synthesize these toxic materials must also have the means to handle them, which among other things involves transporting them into and sometimes, as during leaf senescence, out of containment structures (such as glands, oil tubes, or laticifers) brimming with highly nonpolar, densely concentrated, secondary compound soups. Little is known of such mechanisms; transporter proteins have been implicated in the handling of the innocuous glucosides and glucoronides of secondary compounds (Klein et al., 2001; Frangne et al., 2002), but the more perplexing question as to how plants retrieve the active forms of these metabolites remains unanswered, despite its importance in the context of understanding plant-insect interactions.

If in fact defense systems evolve in lockstep with systems for reducing autotoxicity, then it may be difficult to understand the evolution of plant defense, and attendant evolution of insect counter defense, by examining only a single class of compounds in isolation. A focus for future research, tremendously facilitated by such experimental innovations as microarray analysis and metabolic profiling, is to examine linkages among biosynthetic pathways in the context of plant-insect interactions.

With respect to furanocoumarins, the fact that such distantly related plant families have converged upon the same chemical structures suggests that the potential for elaborating novel compounds is constrained, possibly by efficacy. Thus, identifying mode of action and structure/activity relationships is necessary for understanding the distribution and abundance of idiosyncratically distributed groups of compounds. Understanding mode of action and efficacy against herbivores is also not easily accomplished with model organisms.

Nor, for that matter, is it a straightforward task to select a model organism to understand the evolution of herbivore responses to plant toxins. The argument against relying on a limited number of model species to understand plant-insect interactions applies perhaps even more emphatically to the herbivore side. Among Brassicaceae specialists, for example, while pierid caterpillars in at least two genera appear to rely on nitrile-specifier proteins to avoid formation of toxic isothiocyanates (Agerbirk et al., 2006), the plutellid Plutella xylostella relies on a glucosinolate sulfatase to circumvent the formation of toxic hydrolysis products (Ratzk et al., 2002). Whether a particular “key innovation” led to the diversification of either herbivore group awaits molecular characterization of the relevant detoxification enzymes and a robust understanding of phylogenetic relationships among the relevant groups.

Where molecular evidence does exist, it appears that identical metabolites can be produced via distinct evolutionary trajectories. Two distinct groups of lepidopterans, depressariine oecophorids and papilionine papilionids, have converged in utilizing a common group of chemically related, furanocoumarin-containing plant species and have arguably experienced similar selection pressures with respect to the evolution of detoxification mechanisms (Berenbaum, 2001). These two groups display different coping mechanisms at the molecular level, supporting the idea that the evolution of resistance to plant toxins is as idiosyncratic a process as is the evolution of plant defense. The cytochrome P450 monooxygenases involved in furanocoumarin detoxification in Depressaria pastinacella, the parsnip webworm, are in different subfamilies than are the P450s involved in metabolizing these same compounds in Papilio polyxenes, although in some cases they produce the same metabolites, activity and specificity differ dramatically (Li et al., 2007).

In contrast with Papilio species, which feed on a range of furanocoumarin-containing species in two plant families, the highly specialized D. pastinacella feeds only on two furanocoumarin-containing genera in the Apiaceae and is capable of metabolizing furanocoumarins at rates 10-fold higher than Papilio species (Berenbaum, 2001). This high activity is due in part to the ability of CYP6AB3 to selectively metabolize imperatorin, the furanocoumarin that is most abundant in its two host plant genera (Li et al., 2004; Mao et al., 2006, 2007), and myristicin, a methylenedioxyphenyl compound that, in other insect species, is an effective P450 inhibitor (Mao et al., 2007) and thus acts, as do many methylenedioxy-containing compounds, as a synergist, enhancing toxicity. CYP6AB3 epoxidizes the four-carbon isoprenoid chain on imperatorin and the three-carbon allyl group on myristicin but not any of the other tested furanocoumarins containing only small modifications of the core psoralen structure (Mao et al., 2006). With just these two demonstrated activities, CYP6AB3 is among the most specialized xenobiotic-metabolizing P450s yet characterized. That it is specialized for metabolizing compounds derived from two biosynthetically distinct pathways is yet another argument for examining coevolutionary interactions in the context of plant chemical complexity.

Gene-for-gene coevolution, characteristic of interactions between plants and pathogens, was in part an inspiration for models of plant-insect interaction (indeed, the first use of the term coevolution appeared in an article on flax [Linum usitatissimum] and flax rust; Mode, 1958). Technology has advanced to the point now that coevolution can be examined in the context of gene family for gene family. Within particular suites of coevolving species, active sites for evolution can be compared directly. Bioinformatics methods can be used to discover potential synergistic interactions. P450s, for example, are involved in the biosynthesis of furanocoumarins in Apiaceae and Rutaceae and also play a role in their detoxification in Oecophoridae and Papilionidae. Comparisons of sequences of active sites, which accommodate the same molecules, could provide insights into the genetic changes involved in resistance and counter resistance. As well, the principal signaling molecules in plants, including jasmonate, salicylate, and ethylene, are not only encountered by insects, but also play a role in mediating responses to plant defense. In the case of the corn earworm, Helicoverpa zea, these signal substances are effective inducers of P450 genes encoding enzymes that can detoxify a broad range of plant secondary metabolites (Li et al., 2002). Similar regulatory responses in plants and insects raise the possibility of coevolving regulatory regions in plants and insects.

Bioinformatics analyses of sequence and genome data are greatly complemented by tremendous advances in molecular modeling, a powerful tool for characterization of the range of potential substrates for a putative detoxification enzyme. In the case of P450s, the existence of known crystal structures allows for computer-based estimates of structure and ligand specificity (Baudry et al., 2006). Accompanied by several rounds of energy minimization, the best-scoring model built for each P450 with known functions can be “docked” in a low-throughput manner with a defined set of substrates and subjected to several more rounds of energy minimization for the substrate configurations having the lowest predicted interaction energies. These low-throughput docking modes allow prediction of both the site of modification on each plant chemical and key residues responsible for defining substrate specificity. P450s with undefined functions can also be docked in a high-throughput manner using in silico docking procedures that “virtually screen” three-dimensional chemical databases; databases are now available containing close to 3,000 molecules in three-dimensional format (from information at http://www.alanwood.net/pesticides/) and an additional 14,000 compounds from the KEGG Ligand Database (http://www.genome.ad.jp/kegg/). In a number of ongoing studies that have accurately predicted substrates and substrate-binding modes for Arabidopsis, Synechocystis, and insect P450s (Baudry et al., 2003; Rupasinghe et al., 2003), the compounds predicted to be the best binders could be ranked and successfully tested for binding using either heterologous expression systems or type-I substrate-binding analyses (Jefcoate, 1978) to determine whether P450s bind substrates displacing water as the sixth axial ligand coordinated with the heme. Such screening of insect metabolic capabilities could inform subsequent analyses of plant chemical diversity by indicating which biosynthetic pathways are likely to be of greatest relevance in an ecological context.

In conclusion, the future of plant-insect interactions (at least the short-term future) may lie in complementing the ongoing search for those mechanisms that universally affect interactions with herbivores with a new effort to identify evolutionary forces leading to chemical novelty. Herbivores encounter plant defense compounds in a complex matrix, the exact constituents of which likely differ even among plant families sharing a particular conspicuous group of defense compounds. This chemical complexity can be dissected by identifying genes whose expression is correlated with the production of known defense compounds unique to a family or order. Major advances in sequencing technology make this goal reachable and affordable. Pyrosequencing, for example, allows for rapid and inexpensive DNA sequencing—up to 100 million bp at lengths of 200 bp (soon, perhaps 500 bp) in less than a day. As lengths of these sequences increase, the ability to reconstruct intact genomes becomes feasible. Comparisons among genomes between the various families and orders of plants will enable investigators to identify the constellations of genes unique among plant families and thus likely the result of idiosyncratic interactions with other organisms. With candidate sequences in hand, it would then be possible by microarray experiments to determine patterns of coexpression among these genes instigated by herbivore feeding. Highly parallel sequencing approaches may soon provide an alternative to microarray construction and analysis for studies of gene expression by allowing investigators simply to “count” the number of genes expressed and sequenced (e.g. Weber et al., 2007). Ultimately, it may be possible to compare herbivore and plant genomes with a view to reconstructing the coevolutionary history of particularly intimate interactions in the classical gene-for-gene sense.

It is not altogether surprising that the same half-dozen plant families have occupied plant physiologists and insect ecologists for the past 50 years; even 50 years ago, these were the model systems. These families, however, are hardly random samples of plant diversity—much of the spectacular diversity of plants (including several of the largest plant families) is not represented in this literature. Fraenkel (1959) provided an answer to the question, “Why are there secondary metabolites in plants?” A challenge for the 21st century is to answer the question, “Why are there so many different kinds of secondary metabolites?” The answer to this question may be obtainable by using the genomic tools developed for understanding shared physiological attributes of model organisms as a springboard for understanding ecological idiosyncrasy.

ACKNOWLEDGMENTS

We thank Mary Schuler for serving as our patient mentor in molecular methodology and Anurag Agrawal, Kevin Wanner, Evan DeLucia, and three anonymous reviewers for helpful comments on this manuscript.

LITERATURE CITED

Author notes