DEGREE OF SPECIALIZATION IS RELATED TO BODY SIZE IN HERBIVOROUS INSECTS: A PHYLOGENETIC CONFIRMATION

Authors

  • Robert B. Davis,

    1. Department of Zoology, Institute of Ecology and Earth Sciences, University of Tartu, Vanemuise 46, EE-51014 Tartu, Estonia
    2. E-mail: davis@ut.ee
    Search for more papers by this author
  • Erki Õunap,

    1. Department of Zoology, Institute of Ecology and Earth Sciences, University of Tartu, Vanemuise 46, EE-51014 Tartu, Estonia
    2. Institute of Agricultural and Environmental Sciences, Estonian University of Life Sciences, Riia 181, EE-51014 Tartu, Estonia
    Search for more papers by this author
  • Juhan Javoiš,

    1. Department of Zoology, Institute of Ecology and Earth Sciences, University of Tartu, Vanemuise 46, EE-51014 Tartu, Estonia
    Search for more papers by this author
  • Pille Gerhold,

    1. Department of Botany, Institute of Ecology and Earth Sciences, University of Tartu, Lai 40, EE-51005 Tartu, Estonia
    Search for more papers by this author
  • Toomas Tammaru

    1. Department of Zoology, Institute of Ecology and Earth Sciences, University of Tartu, Vanemuise 46, EE-51014 Tartu, Estonia
    Search for more papers by this author

Abstract

Numerous studies have suggested a general relationship between the degree of host specialization and body size in herbivorous animals. In insects, smaller species are usually shown to be more specialized than larger-bodied ones. Various hypotheses have attempted to explain this pattern but rigorous proof of the body size–diet breadth relationship has been lacking, primarily because the scarceness of reliable phylogenetic information has precluded formal comparative analyses. Explicitly using phylogenetic information for a group of herbivores (geometrid moths) and their host plant range, we perform a comparative analysis to study the body size–diet breadth relationship. Considering several alternative measures of body size and diet breadth, our results convincingly demonstrate without previous methodological issues—a first for any taxon—a positive association between these traits, which has implications for evaluating various central aspects of the evolutionary ecology of herbivorous insects. We additionally demonstrate how the methods used in this study can be applied in assessing hypotheses to explain the body size–diet breadth relationship. By analyzing the relationship in tree-feeders alone and finding that the positive relationship remains, the result suggests that the body size–diet breadth relationship is not solely driven by the type of host plant that species feed on.

The relationship between body size and diet breadth in herbivores has been much discussed over the last few decades (e.g., Jarman 1974; Wasserman and Mitter 1978; Rosenberger 1992, Brandl et al. 1994; Fa and Purvis 1997; Brändle et al. 2000). Attention surrounding this question is not surprising as it relates to two major themes in evolutionary ecology: evolutionary dynamics of niche breadth (e.g., Jaenike 1990; Perry and Garland 2002), and understanding the costs and benefits of large versus small body size (e.g., Barclay and Brigham 1991; Blanckenhorn 2000). Understanding the relationship between body size and diet breadth can also contribute in identifying factors influencing community composition (Summerville et al. 2006), which may also subsequently inform conservation efforts (Gehring and Swihart 2003).

For insects, the observation that smaller herbivores tend to be more specialized is supported principally by evidence from studies on Lepidoptera (e.g., Wasserman and Mitter 1978; Gaston 1988; Gaston and Lawton 1989; Reavey 1992; Lindström et al. 1994; Loder et al. 1998), also through some other taxa ranging from Orthoptera, Phasmatodea, Coleoptera (Novotny and Basset 1999) to Hemiptera (Novotny and Basset 1999; Brändle et al. 2000). However, the pattern is not universal. Indeed, within Lepidoptera alone some studies on particular groups have found no clear relationship (Gaston 1988; Reavey 1992), or even reported a negative relationship (Reavey 1992). A trend that is strong and consistent across taxa may indicate a common and thus a general underlying mechanism, whereas exceptions must be valuable as means of testing the validity of any proposed explanation.

While providing valuable information and ideas, all aforementioned studies have suffered from a common fundamental weakness: the among-species comparisons have not been performed in an explicit phylogenetic framework. Indeed, most studies focused on Lepidoptera (e.g., Wasserman and Mitter 1978; Gaston 1988; Reavey 1992; Lindström et al. 1994; Loder et al. 1998) have not considered phylogenetic relatedness of the studied taxa in any way. Those that have attempted relied on taxonomic information to infer a phylogeny (Lindström et al. 1994; Loder et al. 1998), but naturally, traditional taxonomic schemes at best can only approximate phylogenetic trees (Gittleman and Luh 1992).

Another obstacle for quantitative analyses has been the complexity of measuring diet breadth. The number of host plant species, certainly for more polyphagous feeders, is hard to discern as various unrecorded host plant species likely exist. This has resulted in authors classifying diet breadth into discrete variables (e.g., monophagous/oligophagous/polyphagous or similar groupings) (e.g., Gaston and Lawton 1989; Lindström et al. 1994; Loder et al. 1998), but host plant use across species forms in reality a continuous spectrum (Bernays and Chapman 1994). A more complete list of host plant genera is less problematic to obtain than a species list and would still provide a useful measure of diet breadth on a continuous scale. Indeed, for Lepidoptera, comprehensive lists of host plant genera are readily available (e.g., Seppänen 1970; Crafer 2005).

Furthermore, any measure of diet breadth based on just counting plant taxa may similarly be criticized on the basis of ignoring phylogenetic information (Brändle et al. 2000). Indeed, it appears reasonable to consider a herbivore feeding on three distantly related genera to be “more polyphagous” than a herbivore feeding on three closely related genera. Today, comprehensive molecular phylogenies of higher plants are available such as that of Kühn et al. (2004), which makes it possible to incorporate these into measures of diet breadth (i.e., to produce indices in which host taxa are weighed by their phylogenetic distances (reviewed in Symons and Beccaloni 1999; Lewinsohn et al. 2005). Taking into account phylogenetic relationships between host plant genera also helps to overcome the bias that might arise when taxonomic ranks are not commensurate with plant lineages (Novotny et al. 2002).

If body size is related to diet breadth, then it is pertinent to ask why, and indeed, numerous hypotheses have been proposed over the last few decades (e.g., Loder et al. 1998; Brändle et al. 2000 for a review). Among these, several are based on the assumption that larger species may be better placed to withstand more stressful environmental conditions (Wasserman and Mitter 1978; Lindström et al. 1994; Loder et al. 1998). Higher tolerance to stress enables larger animals feed on a broader host plant range including hosts of low quality. In particular, larger herbivores should better tolerate the physical and physiological stresses of feeding on tougher plant tissues (Mattson 1980; Reavey 1992) or presence of hard-to-digest secondary plant compounds used in quantitative defense such as tannins (Feeny 1975; Mattson 1980; Gaston and Reavey 1989; Lindström et al. 1994; Müller-Schärer et al. 2004). Quantitative defenses are more typical of woody plants, whereas herbaceous plants tend to defend themselves against herbivores using toxic qualitative defense chemicals (Feeny 1975; Müller-Schärer et al. 2004). A herbivore that feeds on a plant species producing a particular toxin is likely to be specialized to protect itself against the harmful effects of this toxin (Gilbert 1979; Müller-Schärer et al. 2004), and large body size should not offer an immediate advantage in this respect. Any clade comprising herbivorous species that feed on either quantitatively or qualitatively defended plants may display an association between body size and diet breadth for this reason alone. However, before taking the effort to test any of the causal hypotheses, the underlying notion that body size is related to diet breadth itself must first be tested by appropriate means.

The principal aim of the present study is to provide a rigorous proof of a relationship between body size and diet breadth in a group of herbivorous insects. Using the Lepidopteran family Geometridae as our example taxon, we are able to overcome various methodological issues inherent in previous work. In particular, as the basis of our comparative analysis, we make an explicit use of a recent molecular phylogeny of geometrid moths, and additionally apply a phylogenetic measure of host plant diversity, which makes our study the first to use explicit phylogenetic information to tackle this question. Furthermore, we test the robustness of our conclusions by using alternative measures of body size and diet breadth. As an addition, we use available data to demonstrate how we can use the methods at our disposal to consider a previously proposed causal hypothesis regarding the relationship between body size and diet breadth.

Materials and Methods

SPECIES DATA

The Geometridae are an attractive Lepidopteran family to form the focus of such a study. Species of this group show considerable morphological and ecological variation in traits including body size and breath of diet. In terms of geographic range, our sample (105 species) was restricted to northern Europe. This is the region for which the best host plant data exist, and geometrid species from this region are also best represented in phylogenetic studies.

Two alternative measures of moth body size were used; female body mass and male wing length. To obtain data on female dry body mass, moths (57 species) were reared in the laboratory on their preferred host plants (see Javoiš et al. (2011) and Davis et al. (2012) for practical details), sample sizes of female individuals per species ranged from one to 48 (median = 5). Female rather than male body mass was considered because the former is more likely to be causally related to host plant use. For example, egg size, and therefore offspring size, is causally related to female body size in geometrid moths, which bears consequences for larval host plant use (Davis et al. 2012). Male wing length data (105 species) were taken from Leraut (2009). We used male rather than female wing because females of some species have reduced flight ability, and disproportionately short wings (Snäll et al. 2007), whereas the males of all geometrids are active flyers. Furthermore, for those geometrids with fully winged females, there is often very little difference between male and female wing length (Leraut 2009). While male wing length was used on a linear scale only, female body mass was additionally tested on a log scale in case of potential allometric scaling.

We considered two alternative sources of diet breadth data: (1) the Finnish host plant list of Seppänen (1970) (providing data on 97 species of moths from our sample), and (2) the British host plant list of Crafer (2005) (97 species, which partially overlap with those covered by Seppänen) individually. As a third measure of diet breadth, we combined the Seppänen and Crafer lists (all species with body size data covered). Diet breadth was additionally recorded in two alternative ways: (1) as the absolute number of host genera on which a species had been observed to feed, and (2) applying a measure of phylogenetic diversity of host genera. Phylogenetic diversity was expressed as the index PD (Faith 1992), which is computed as the sum of the total branch lengths on a phylogenetic tree (e.g., used in Beccaloni and Symons 2000; Morse and Farrell 2005). We used the phylogeny for vascular plants in Central Europe (Kühn et al. 2004), which covers, at genus level, the host plant range of the moth species in this study. We used the picante package (Kembel et al. 2010) for the R statistical environment (R Development Core Team 2012) to calculate phylogenetic diversity.

Diet breadth was also tested on a linear scale as well as logarithmically transformed scale, the latter having the advantage of weighing down the differences when the number of host genera was large: the variation which is perhaps less meaningful both biologically, and due to being more prone to recording bias. Female body mass and male wing length showed a strong positive correlation (R= 0.89) as do Seppänen and Crafer host plant lists (R= 0.56), as do absolute and phylogenetic measures of host plant diversity (R= 0.98, Fig. 1).

Figure 1.

Correlation of absolute number of host plant genera versus PD index of host plant phylogenetic diversity. Example here is for the combined Seppänen + Crafer host plant list. R= 0.98.

GEOMETRIDAE PHYLOGENY AND PHYLOGENETIC COMPARATIVE ANALYSIS

The comparative analysis was based on a phylogeny, which was built on DNA sequence data for Geometridae principally provided by Snäll et al. (2007), Viidalepp et al. (2007), Õunap et al. (2008, 2011), and Wahlberg et al. (2010), but also a small portion of unpublished original sequences (Fig. 2). The final data matrix comprised 116 species and 4643 positions. To construct an ultrametric tree from this, we implemented Bayesian phylogenetic inference using BEAST software (Drummond and Rambaut 2007) (see Supporting Information for further details). Eleven species for which relevant ecological data were not available were subsequently pruned from the tree.

Figure 2.

Phylogeny of Geometridae used in this study, see Supporting Information for technical details. Values/states for life-history traits given.

Phylogenetic independent contrasts (Felsenstein 1985) analysis was used to test for the association between body size and diet breadth, using the ape package for the R statistical environment (R Development Core Team 2012).

To provide an example of how our analysis can be extended to address particular hypotheses, we evaluated the role of tree- versus herb-feeding as a causal factor behind the relationship between body size and diet breadth (e.g., Lindström et al. 1994; Loder et al. 1998). In particular, we tested the relationship between body size and diet breadth on tree-feeders alone (79 species). Indeed, if the type of host plant (i.e., tree or herb) explains fully the relationship between body size and diet breadth (see above), then the relationship between these two traits should disappear when tree- and herb-feeders are analyzed separately. Unfortunately, in our dataset, the number of herb-feeding species was too low for a meaningful statistical analysis within this group.

Results and Discussion

In a dataset corrected for phylogenetic nonindependence, we found a consistent trend of positive correlations between body size and diet breadth in geometrid moths (Table 1A). The correlations are robust: it makes no difference which measure of body size is used, or whether the variables are on a linear or log scale, or for diet breadth, whether a measure of phylogenetic diversity of the hosts is included or not. Indeed, species of larger body size do have a broader range of host plants, and this pattern cannot be attributed to any confounding effect of common ancestry. This positive relationship between the two traits supports previous findings in Lepidoptera (Wasserman and Mitter 1978; Gaston 1988; Gaston and Lawton 1989; Reavey 1992; Loder et al. 1998; Lindström et al. 1994), but does so without the methodological problems (see introduction) inherent to earlier works. Our result thereby solidifies the empirical basis of the observation of an association between body size and diet breadth.

Table 1.  (A) Correlation coefficients between phylogenetic independent contrasts of body size and diet breadth in geometrid moths. (B) Correlation coefficients between phylogenetic independent contrasts of body size and diet breadth in tree-feeding geometrids only. Results are based on analyses with outliers removed (maximum three in any single analysis) by visual inspection. Significant results in bold and nonsignificant results in italics (significance at P < 0.05). Correlations marked with asterisk were nonsignificant before removing the outliers. For diet breadth measures, “Ab” indicates that absolute numbers of host plant genera were used, whereas “Ph” indicates that this analysis took account of host plant phylogenetic diversity.
(A). All Geometridae in sampleMale wing lengthFemale body massFemale body mass (log)
Ab. Diet Seppänen 0.30 0.45 0.35
Ph. Diet Seppänen 0.31 0.37 0.38
Ab. Diet Seppänen (log) 0.4 0.35 0.31
Ph. Diet Seppänen (log) 0.37 0.35 0.33
Ab. Diet Crafer 0.20* 0.53 0.51
Ph. Diet Crafer 0.02* 0.46 0.48
Ab. Diet Crafer (log) 0.34* 0.42 0.45
Ph. Diet Crafer (log) 0.09* 0.44 0.47
Ab. Diet Combined Seppänen-Crafer 0.29 0.35 0.37
Ph. Diet Combined Seppänen-Crafer 0.27 0.33 0.36
Ab. Diet Combined Seppänen-Crafer (log) 0.33 0.31 0.29
Ph. Diet Combined Seppänen-Crafer (log) 0.29 0.32 0.30
(B). Tree-feeding Geometridae only   
Ab. Diet Seppänen 0.34* 0.49 0.48*
Ph. Diet Seppänen 0.35* 0.37 0.38
Ab. Diet Seppänen (log) 0.37* 0.41 0.39*
Ph. Diet Seppänen (log) 0.35* 0.35 0.33
Ab. Diet Crafer 0.40* 0.53 0.56
Ph. Diet Crafer 0.31* 0.46 0.48
Ab. Diet Crafer (log) 0.38* 0.45 0.60
Ph. Diet Crafer (log) 0.32* 0.44 0.47
Ab. Diet Combined Seppänen-Crafer 0.29* 0.50 0.40
Ph. Diet Combined Seppänen-Crafer 0.22* 0.33 0.36
Ab. Diet Combined Seppänen-Crafer (log) 0.23* 0.32 0.36
Ph. Diet Combined Seppänen-Crafer (log) 0.29* 0.32 0.30

When we test tree-feeders separately, the pattern remains qualitatively the same: the positive relationship we observed in the total dataset is evident within this ecological group alone. Indeed, most of our alternative analyses within tree-feeders alone show a significant positive correlation comparable to the one in the total dataset (Table 1B). This finding implies that the overall correlation cannot solely be based on a herb- versus tree-feeding dichotomy. Naturally this analysis should not be taken as proof that the type of host plant is not related to the relationship between body size and diet breadth. As a minimum, we should test if the association between the two variables does differ within herb- versus tree-feeders, and when calculated across these groups: an analysis for which our dataset is still too limited. Nevertheless, we believe that the present example illustrates that phylogenetic comparative approaches can be used to address causal hypotheses behind the pattern detected, which should become increasingly feasible as reliable phylogenetic and comparative-ecological information accumulates.

In summary, our results provide robust support for the hypothesis that body size is positively related to diet breadth in Lepidoptera, in the first study to overcome previous problems surrounding a comparative approach to this question: (1) it is based on a solid molecular phylogeny of the herbivores, (2) diet breadth coded as a continuous character, and accounting for phylogenetic relationships among the plants. Our data failed to support one particular causal hypothesis that food type (i.e., tree or herb host plants) drives the body size–diet breadth relationship, but further analysis regarding this hypothesis is required. This is certainly feasible given enough data, and the methods used in this study can be used to address different causal hypotheses proposed to explain the relationship between body size and the degree of specialization. We hope that our results encourage such studies across different taxa, which would have an obvious potential to contribute to solving of various central problems of evolutionary ecology.


Associate Editor: P. Lindenfors

ACKNOWLEDGMENTS

The study was supported by Estonian Science Foundation grants 7682, 7699, and 8613, targeted financing projects SF0170160s08 and SF0180122s08, and by the European Union through the European Regional Development Fund (Center of Excellence FIBIR). We thank A. Kaasik for statistical advice and M. Pärtel for formatting the plant phylogeny for use in our analyses.

Ancillary