Insects on Plants: Macroevolutionary Chemical Trends in Host Use

See allHide authors and affiliations

Science  11 Apr 1997:
Vol. 276, Issue 5310, pp. 253-256
DOI: 10.1126/science.276.5310.253


Determining the macroevolutionary importance of plant chemistry on herbivore host shifts is critical to understanding the evolution of insect-plant interactions. Molecular phylogenies of the ancient and speciose Blepharida (Coleoptera)–Bursera(Burseraceae) system were reconstructed and terpenoid chemical profiles for the plant species obtained. Statistical analyses show that the historical patterns of host shifts strongly correspond to the patterns of host chemical similarity, indicating that plant chemistry has played a significant role in the evolution of host shifts by phytophagous insects.

What factors have directed the evolution of host shifts by phytophagous insects? This has been a central question in the field of plant-insect interactions for the last 30 years (1). Ehrlich and Raven (2) postulated that shifts to new hosts are mediated by the chemical similarity between old and new hosts and that host plant chemistry should leave its trace on phylogenetic patterns of host shifts at a macroevolutionary level. However, demonstrating a role for plant chemistry in the macroevolution of host use has been difficult (3). Detailed quantitative investigations have had to await the development of modern molecular and phylogenetic techniques to reconstruct accurate host and herbivore trees. Also, an evolutionary association of host shifts with plant chemistry could be spurious: Related plants have similar chemistry, and plant and herbivore phylogenies may correspond for a variety of biogeographic or ecological reasons unrelated to chemistry. In fact, some studies have shown a close correspondence of host and insect phylogenies (4, 5), suggesting that the pattern of host cladogenesis may be important, and that host chemical similarity may be overemphasized. Here, a quantitative investigation of the chemical trace in the evolution of insects and their plant hosts is presented, using insect and plant phylogenies reconstructed from DNA sequences. These reconstructions of the ancient and diverseBursera-Blepharida system were used to test the relative importance of chemical similarity and host cladogenesis in the evolutionary history of host shifts.

Burseras are the New World frankincense and myrrh. The genus comprises about 100 species distributed from the southwestern United States to Peru. It reaches its maximum diversity in the tropical dry forests of Mexico, where about 80 species occur (6). The New World members of the beetle genus Blepharida (Chrysomelidae: Alticinae) include about 45 species, many of them monophagous, which feed mainly on Bursera and a few other members of Burseraceae and its sister family Anacardiaceae (7). Although New World Blepharida has been combined with the Afrotropical genus Blepharidina, there are well-marked morphological differences between Old and New WorldBlepharida, suggesting that the New WorldBlepharida form a monophyletic group (8). TheBursera-Blepharida interaction is old. In the New and Old World tropics Blepharida feeds on Anacardiaceae and Burseraceae, suggesting that their interaction probably started before the separation of Africa and South America, more than 100 million years ago (5, 9).

Bursera produces an array of terpenes, including alpha and beta pinene, camphene, phelandrene, and limonene (10), distributed in a reticulating network of resin canals in the cortex of the stems and throughout the leaves. In some species these resins may be under considerable pressure and squirt out when leaves are damaged (11). In many plant groups terpenes are toxic or repellent to insect herbivores (12), and in Bursera they decrease Blepharida survival and growth rate (11,13). The fact that this is an old and specialized interaction, with many species in both genera and well-known ecology and systematics, provides a valuable opportunity for testing macroevolutionary hypotheses.

The molecular phylogenies of Bursera andBlepharida (14, 15) are shown in Fig.1. All Blepharida species were collected in the field directly from the species they were attacking, and their host relationships were confirmed by multiple site visits over 3 to 5 years. A dendrogram of Bursera species based on their chemical similarity (chemogram) was constructed to test the importance of host chemistry in host shifts (16) (Fig. 2A). For reconstruction of the evolution of the chemistry of Bursera,chemical classes were parsimoniously traced onto Bursera’s phylogeny (17). Most clades of Bursera include plants that are in different classes, suggesting that chemical similarity is partially independent of plant phylogeny (Fig.2B).

Figure 1

Feeding associations ofBlepharida beetles (right) on Bursera hosts (left). According to the molecular phylogeny of Bursera(strict consensus tree), the genus is monophyletic and consists of two principal groups known as sections Bullockii and Bursera. The four most parsimonious trees differed only in resolving the positions ofBursera heteresthes, Bursera palmeri, andBursera mirandae. They had a consistency index of 0.57 and a retention index of 0.74. The two most parsimoniousBlepharida trees differed in the position ofBlepharida unknown sp. 11. They had a consistency index of 0.58 and a retention index of 0.72. Beetle species were determined by D. Furth of the Smithsonian Institution and many of them are undescribed. Sequences of insects identified as Bursera flavocostata present many genetic divergences and are very probably different species. For clarity, the hosts of polyphagousBlepharida alternata are not indicated (but see Fig. 3 for its host plants). Asterisks indicate outgroups and the numbers above the main branches of the trees are bootstrap percentages.

Figure 2

(A) Dendrogram ofBursera based on chemical similarity. Numbers and color coding indicate the four main chemical groups according to Ward’s clustering method. (B) The four classes of chemically similar species of Bursera are traced ontoBursera’s phylogeny. Most lineages of Burserainclude mixtures of plants that are in different classes, suggesting that chemical similarity is partially independent of plant phylogeny. Gray lines indicate that character reconstruction is undetermined.

To investigate whether plant chemical similarity facilitated host shifts by Blepharida, I compared the topology ofBlepharida’s phylogeny with the topology of the chemogram. Also, to investigate the importance of host cladogenesis in insect shifts, I compared the topologies of the phylogenies ofBursera and Blepharida. Three techniques were used for the comparisons: character tracing, which is graphical and not statistical (17), and tree mapping (18) and Farris’ distortion coefficient (19), which are statistical techniques. The latter two methods were selected because they do not depend on the operational division of clades and clusters, and their indices of similarity are sensitive even when the topologies compared are only loosely congruent, an expected scenario for phytophagous insects, which disperse rather freely among hosts. Tree mapping modifies one of two trees or cladograms until their differences are reconciled. In the present case, it modifies the plant tree by duplicating branches until a “reconciled” tree is obtained. This technique provides two measures of fit between host and associate trees. “Leaves added” is the difference between the number of nodes in the insect and reconciled tree, and “losses” is the number of instances in which an insect species is absent where it is predicted to occur on the reconciled tree. Both parameters decrease with increasing similarities of plant and insect trees [see (18) for details]. Farris’ distortion coefficient provides a measure of the discordance of the branching topology of two trees A and B by estimating how distorted each clade of A is on B. For each monophyletic cluster of A, one counts how many times the cluster is fragmented on the B tree. This number is divided by the number of taxa of the cluster minus one. For example, if a cluster of A includes three taxa, and they are all separate in tree B, then the coefficient for that cluster of A is 2/2 = 1. The distortion coefficient is the average of the values for all clusters of A. Perfect congruence yields a coefficient of 0, and complete distortion, a value of 1. Both approaches were tested statistically by comparing observed indices to the distribution of indices obtained by repeatedly randomizing one of the trees (Markovian model).

Tracing the chemical classes onto Blepharida’s phylogeny shows few shifts of Blepharida between chemically dissimilar plants (Fig. 3). Subclades of Blepharidaappear to have colonized species of chemically similar plants. For example, the lineage that includes Blepharida sparsadiversified using burseras that belong only to one chemical group (in blue). Similarly, the lineage of Blepharida flavocostata andBlepharida unknown spp. 1 and 2 evolved exploiting burseras from only two chemical groups. An interesting exception is the highly polyphagous Blepharida alternata, which can feed on burseras from all the chemical groups. The congruence is significant with tree mapping (“leaves added,” P < 0.006; “losses,”P < 0.0002) and the distortion coefficient (0.73,P < 0.05), which do not depend on the operational delimitation of chemical classes used in the figures. Character tracing graphically demonstrates that Blepharida has shifted host use from one of the two major clades of Bursera (the two sections of the plant genus) to the other several times.Blepharida has also shifted between hosts belonging to different subclades several times (Fig. 4). For example, the lineage of B. flavocostata and Blepharidaunknown spp. 1 and 2 attacks burseras from four terminal clades, and one clade is fairly distantly related to the others (pink line). With tree mapping, the congruence of the two phylogenies is significant for “leaves added” (P < 0.05), but not for “losses” (P < 0.26). The distortion coefficient for Bursera and Blepharida cladograms is 0.86 and congruence is not significant (P = 0.1).

Figure 3

Comparison ofBlepharida’s phylogeny and the dendrogram ofBursera chemical similarity using character tracing. The names on top of Blepharida’s names are the species ofBursera they feed on. The polytomies illustrate polyphagy. Tracing the four chemical classes onto Blepharida’s phylogeny show few shifts of Blepharida between chemically dissimilar plants.

Figure 4

Comparison of Blepharida andBursera phylogenies. The eight major clades ofBursera are traced onto Blepharida’s phylogeny. For analyses, Bursera mirandae and Bursera heteresthes were situated according to two of the four maximally parsimonious trees. Blepharida has shifted numerous times among different Bursera lineages.

Because chemical similarity in Bursera is partially independent of its phylogeny (Fig. 2B), it was possible to look at host shifts among plant chemical groups that were not host shifts among plant clades. In the same way, host shifts among plant clades that were also host shifts among chemical groups could be ignored. To do this, I modified the distortion coefficient. As mentioned before, the disagreement between trees A and B is measured by the number of fragments into which each cluster of tree A is broken on tree B. The same applies to trees A and C. But now, to measure the distortion between tree A and tree B, for each cluster of A, the value of the distortion coefficient between A and C is added to the value of the distortion coefficient between A and B. This mathematical procedure removes shifts among chemical groups that are also shifts among plant clades. With this modification the coefficient remained statistically significant for the comparison betweenBlepharida’s phylogeny and the chemogram (0.84,P < 0.05). However, for the comparison ofBlepharida and Bursera phylogenies, the modified distortion coefficient increased to 0.94 (P = 0.25). This suggests that the relationship between the two phylogenies is due in large part to the correlation between plant phylogeny and plant chemistry, whereas the relationship between Blepharida’s phylogeny and the chemogram of Bursera does not depend on the correlation between plant phylogeny and plant chemistry. Thus, comparisons ignoring the correlation between plant phylogeny and plant chemical variation, as well as comparisons controlling for this correlation, indicated a greater influence of host plant chemistry than host plant phylogeny in the evolution of host use inBlepharida and Bursera.


Stay Connected to Science

Navigate This Article