Comparisons between closely related species in different habitats provide a window into understanding how biotic factors shape evolutionary pathways. Weevils in the genus Curculio have radiated extensively across the Palearctic, where similar ecomorphs have evolved independently on different hosts. We examined ecological and morphological data for 31 Curculio species using multivariate statistics to determine which morphological traits covary and which correlate with the host seed size. A subset of 15 taxa for which phylogenetic relationships were known were used for comparative analyses and inferring historical patterns of trait evolution. The morphological analyses suggest rostrum size increased proportionately to body size in both males and females and that both rostrum and body size correlate with host seed size but that rostrum shape does not correlate with any of the seed traits used in the analyses. Host shifts from small seeds to considerably larger seeds or vice versa have occurred several times independently and historical trait evolution indicates that these host shifts were accompanied by morphological changes in rostrum size. These patterns suggest that seed size is an important selective agent for changes in rostrum length along with body size and thus may be a key factor promoting morphological differentiation in the genus Curculio.
Consistent with Darwinian thought, the framers of the modern synthesis accorded a central role to ecological adaptation in the speciation process (Simpson 1944, 1953; Lack 1947; Mayr 1947). When ecological opportunities abound, divergent natural selection causes differentiation in phenotypic traits influencing the use of the environment, and speciation ultimately may follow. Alternatively, speciation is driven by mechanisms having little to do with differences between environments or ecological opportunity (e.g., genetic drift, hybridization, chromosome rearrangement). Phenotype divergence may occur independently of this process or it may occur later, possibly in response to interspecific competition or other interactions (Schluter 2000).
Ecomorphological analyses attempt to link the structure and function of organisms with relevant features of the environment and have been an integral tool of ecologists and evolutionary biologists in elucidating patterns and interpreting processes (Losos and Miles 1994). With respect to structure-function correlations, studies of morphology may reveal selective factors in the environment and constraints on the response of the phenotype to these factors. Many studies assume that the patterns revealed in an ecomorphological analysis reflect adaptation to prevailing selective pressures, but these interpretations can be problematic in the absence of phylogenetic information (Harvey and Pagel 1991). To date, phylogenetic patterns of morphological adaptation to the environment have been studied in very few invertebrates, including Cicindela mandible length and prey size (Vogler and Goldstein 1997) and Cancer crab body size and habitat (Harrison and Crespi 1999).
This study will attempt to add to our knowledge of ecomorphological adaptations of insects using the example of weevils in the genus Curculio. These beetles are of particular interest as the females lay their eggs in the host seeds and as a result the fitness of the host and the weevil are intimately related. This type of life-history strategy (i.e., seed predation) is relatively widespread among the Coleoptera. Indeed, within the superfamily Curculionoidea, Marvaldi et al. (2002) showed that feeding on fruit and/or seeds has evolved at least seven times and that it appears never to have been reversed. This feeding habit has also independently evolved in the Chrysomeloidea such as the Bruchinae. Understanding the details of these seed-feeding radiations is important for studying the potential role of natural selection in diversification, which in turn would inform models of speciation (Marvaldi et al. 2002).
The genus Curculio is part of the large family of Curculionidae, which contains more than 47,930 described species (Kuschel 1995). The diversity of the Curculionidae is hypothesized to be a consequence of phytophagy mostly on angiosperms (Farrell 1998) and the use of the rostrum in oviposition site preparation is considered the key adaptation that facilitates circumventing of the physical defenses of the plant (shells, spines), avoidance of desiccation of the larvae, and initiation and maintenance of attachment to the host (Anderson 1995).
The species in the genus Curculio attack a wide range of host plants from various families including Fagaceae, Juglandaceae, Betulaceae, Theaceae, and Moraceae. The Ficus-associated species are believed to be ancestral and are distributed from India to Africa (Fig. 1; Perrin 1992); the other species found in the Palearctic are principally associated with the family Fagaceae, Quercus being the most common host (Hoffmann 1954; Gibson 1969). A few species have also been found to attack various wasp galls on oaks (Quercus sp.) and willows (Salix sp.) and the latter species have occasionally been grouped into the subgenus Balanobius. All species of Curculio oviposit in the seed while still on the tree. Females of all Curculio species use the rostrum to excavate a hole in the nuts of various plant species (Gibson 1969), they then turn around and oviposit one to three eggs into the hole. The larvae feed on the nuts before leaving the seeds to overwinter in earthen cells in the soil, where they pupate before emerging in spring.
The various Curculio species are specialized with respect to particular plant genera, but might attack up to 30 species within that genus (see Appendix Table 1 and Table 2 available online at http://dx.doi.org/10.1554/04-119.1.s1 (10.1554_04-119.1.s1.doc)). A phylogenetic tree derived from two mitochondrial and two nuclear markers was used to infer the history of host associations within the genus (Hughes and Vogler 2004). As in several other insect herbivore studies (reviewed in Mitter et al. 1991), there is evidence that the species have switched their affinities from one plant group to another, for example, the Quercus-feeding and the gall-feeding strategies have evolved in two separate lineages (Fig. 1).
Large directed changes in morphology are expected in such a genus because a shift to a new habitat often involves substantial changes in selection pressures and, subsequently, the rapid evolution of new adaptations (McPeek 1995). Weevils of the genus Curculio are ideal subjects with which to evaluate phylogenetic hypotheses of ecomorphological adaptation because the genus comprises approximately 345 morphologically diverse species found on a variety of host plants (Von Dalla Torre and Schlenkling 1932; Hoffmann 1954; Gibson 1969; Anderson 1995). Their most conspicuous phenotypic differences are in rostrum size and shape, especially length and curvature. Within the genus Curculio, previous researchers have noted that species that are relatively large in body size and rostrum length predate large seeds such as Castanea sativa and have suggested that an increasingly long rostrum has been selected for in some species because of its value in excavating deeper oviposition holes, thus allowing for the exploitation of the larger, thicker-shelled, or spinose (e.g., Castanea) nuts with which these species are associated (Anderson 1995). However, no study to date has evaluated the hypothesis that the morphological diversity of Curculio is the result of adaptation to specific seed size or shape— that is, a major assumption that the morphological space maps onto the ecological space—has not been adequately studied.
In virtually all Curculionoidea with a long rostrum there is sexual dimorphism in rostrum length. Females have a longer rostrum than males and perhaps this reaches an extreme in species of the genus Curculio (Anderson 1995). Comparisons between males and females help provide a window into understanding the factors that shape evolutionary pathways. The rostrum if under selection for oviposition should vary disproportionately in females versus males, as opposed to a body size–related factor alone (i.e., isometry) that would result in similarly increased rostrum length in larger species without regard to sex. Studying the rostrum and other functional body parts involved in oviposition should therefore enable discrimination between different pathways of morphological evolution and reveal the effect of the selective regime on the buildup of species diversity.
The aims of this study are three-fold: (1) testing the correlation between the seed characteristics and the weevil morphology; (2) determining differences in the morphology of the males and the females; and (3) testing whether the correlation between seed characteristics and Curculio morphology persists when phylogenetic relationships among taxa are taken into account.
Materials and Methods
Data Collection and Quantification of Body Shape
Light scanning tomography was used to take three-dimensional images of Curculio specimens from the Natural History Museum, London, and C. W. O'Brien's personal collection (Tallahassee, FL). This technique enables the surface of the specimen to be optically cut into thin slices, which allows the three-dimensional reconstruction of the scanned specimen. Exact morphometric measurements were then performed on the virtual 3-D image on a Macintosh computer using the public domain NIH Image program (developed at the U.S. National Institutes of Health and available at http: //rsb.info.nih.gov/nih-image/).
Up to eight specimens were measured for each species (see supplementary material available online only at: http://dx.doi.org/10.1554/04-119.1.s2 (10.1554_04-119.1.s2.doc)). A total of 31 species from Europe and North America for which ecological data are known were measured. The species included in the study are listed online only in Appendix Table 1 and 2 with the range of hosts they predate. The host of a weevil is considered in the following discussion as the plant on which the larvae develop. Morphological measurements of host seeds were gathered from the literature (See supplementary material available online only). As seed sizes within a species are normally distributed (Harper et al. 1970), the median between the minimum and the maximum seed size from the literature was taken as being the average seed size. Four variables for the seeds were used: seed height, seed diameter, seed volume, and thickness of the cup or length of spines.
The morphology of each species was described by nine variables. Three measurements of body size (elytra length, thorax width, and abdomen) were made because changes in body size often track environmental changes and because body size is an important correlate of many aspects of life history, ecology, and behavior (Peters 1983). The length of the legs (tibia and femur length) is believed to vary with substrate size as relatively longer legs will maximize locomotor capabilities on broad surfaces and on small seeds relatively shorter legs maximize agility. Due to the role of the rostrum in preparing the oviposition site, four measurements of the rostrum were made (rostrum length, eye-antennal insertion distance, length of the straight part of the rostrum, angle of the curved part of the rostrum). The rostrum should be proportionately longer for species predating thick seeds or seeds with needles, so that the embryo and/or the endosperm of the seed can be reached, and should be more curved for species predating small seeds, because the curvature of the plane is more pronounced in a smaller seed.
Analysis of Morphological Data
The structure of the morphological space occupied by the Curculio species was investigated by principal component analysis (PCA) for males and females. The eigenvalues are the variances of the species positions on each of the orthogonal axes. Each axis accounts for a significantly larger portion of the variance than the next smaller axis. The eigenvectors contain loadings, which are the directional cosines that indicate the rotation of the principal component axes relative to the original variables (Morrison 1967; Blackith and Reyment 1971; Venables and Ripley 1997; StatSoft 2002).
Ratios of measurements were not calculated because they are not linearly related to other measurements or ratios and, in such cases, the PCA axes may reflect spurious correlations among variables. Instead, a nine-dimensional morphological space was defined by the logarithms of each variable. Log transformation serves two purposes: it reduces the skewness of the original variables and makes their variances homogeneous; it scales variables in terms of the ratios between values, which are generally thought to be more relevant to ecological position (Hespenheide 1973). As a result, linear combinations of logarithms represent products and ratios of variables, depending on the signs of their coefficients (Miles and Ricklefs 1984).
Canonical Correlation Analysis
A canonical correlation analysis (CCA) was used to describe the relationship between positions of species in ecological and morphological space and the degree to which the points correspond. The squared canonical correlation coefficients are interpretable as the proportions of variance in one dataset that are common to the other. The null hypothesis that the ith canonical correlation and all that follow are zero, that is, that there is no correlation between the ecological and morphological space, can be tested by Wilk's likelihood ratio, Λ:
where rk is the kth canonical correlation. Λ is the product of the proportion of variance left unexplained by the s canonical correlations. Λ can be transformed to an approximate chi-square distribution by Barlett's formula, where N is the number of species and p and q are the number of variables in the two datasets; the value for the ith axis has (p + 1 − i)(q + 1 − i) degrees of freedom (Miles and Ricklefs 1984; Venables and Ripley 1997; StatSoft 2002).
The eigenvectors resulting from the CCA are equivalent to partial regression coefficients of a multiple regression analysis, but interpretation of the canonical variates using these weights is difficult when the variables within each set are correlated. In this case, correlation of the original variables with each canonical axis both within and between data sets provides a useful basis for interpretation. The between-set correlations are the products of within-set correlations and the canonical correlations. The square of each intraset correlation coefficient represents the proportion of variance in the original variable that is associated with each canonical variate. In this study the PCA and the CCA were performed in S-Plus using the function princomp and cancor (Venables and Ripley 1997). Comparisons between male and female correlations were executed using linear modelling in S-Plus (Crawley 2002).
Comparative Analysis
The method of phylogenetically independent contrasts was used to statistically test the hypothesis of joint evolution of Curculio morphology and host seed characteristics for a subset of 15 taxa (Felsenstein 1985; Harvey and Pagel 1991). The null hypothesis is that there is no correlation between changes in traits at the nodes. Contrasts were produced using the CAIC computer package for Apple Macintosh (Purvis and Rambaut 1995). The branch lengths of the phylogeny, providing the expected variance of evolutionary change, were estimated in PAUP using the maximum likelihood model HKY85 + Γ with the mitochondrial sequence data (cyt b and COI), which was more complete in terms of taxonomic coverage, and constrained to the topology of one of the most parsimonious trees from the total evidence (COI, cyt b, EF1-α, Pglym). While most of the interspecific clades of the total evidence phylogeny have high bootstrap values and Bremer support, two internal nodes exhibit bootstrap values of less than 50%. However, the elephas + pellitus + venosus clade and the caryae + nasicus + iowensis + confusor clade are well supported. But, as for many phylogenies, this phylogenetic tree is only a preliminary hypothesis of the relationships between the Curculio species and may be open to reinterpretation in the light of further sampling (Hughes and Vogler 2004). This phylogeny was used to calculate three sets of contrasts using the positions of females from 15 species along the first three principal components and seed height, diameter, volume, and cup thickness logarithmically transformed to produce standardized contrasts. Contrasts were analyzed using multiple regressions through the origin.
The independent contrast approach has come to dominate comparative analyses of continuously varying characters but new maximum likelihood models of correlated trait evolution of continuously varying characters as implemented in Continuous (Pagel 1997, 1999) offer advantages in their ability to scale phylogenetic path lengths in response to patterns in the data by estimating three scaling parameters. The parameter κ differentially stretches or compresses individual branch lengths and can test for punctuational versus gradual modes of trait evolution. δ scales the overall path length of the phylogeny, that is, distance from the root to the species and the shared path length, which helps to detect whether the rate of trait evolution has changed from the root to the tips. The third parameter, λ, indicates whether a trait is phylogenetically associated, that is, whether the phylogeny correctly predicts the trait covariance among species (Pagel 1997, 1999). The significance of these parameters can be tested using a likelihood ratio test to compare the goodness of fit of a model with a simpler null model that lacks parameters.
The scaling parameters λ and κ were estimated simultaneously followed by δ for each principal component separately. If trait evolution has not followed the topology or branch lengths, these values will depart from 1.0. First, the correlation between the principal components and the seed traits were calculated and the significance of the correlation tested comparing a model with zero covariance to an alternative model with covariance without using scaling transformation (all parameters set to 1.0). This is broadly equivalent to an independent contrast analysis when there are no polytomies present in the phylogeny. Second, the correlation was tested using the scaling parameters estimated by maximum likelihood, which should improve the fit of the data to the model when estimating the correlation between traits.
Results
Morphological Space
The shape and dimensions of the morphological space occupied by the female of 30 species are revealed by PCA (Fig. 2A,B). The first three PC axes explain 96% of the variance among species (Table 1). The remaining six axes have substantially smaller eigenvalues. PC1 which alone explains 81.1% of the variation among species, is primarily an overall size dimension because all loadings except one are negative and have very similar values. This axis separates the larger species (e.g., C. caryatripes, C. caryae) from the smaller species (e.g., C. rubidus, C. cerasorum, Fig. 2). PC1 contains some shape information because the coefficient for curvature of rostrum is positive and different in magnitude to all the other high negative loadings, that is, small weevils have more curvature in their rostrum. PC2 (11%) exhibited high negative loadings for curvature of the rostrum and positive loadings for the straight part of the rostrum, that is, it separates species based on the constructed value (straightness of rostrum/curvature of rostrum tip). At one end of the axis are straight-rostrum species with a small curvature at the tip (C. caryatripes, C. elephas), at the opposite end are species with almost the whole rostrum curved (C. sayi, C. victoriensis, C. strictus). PC3 (4%) is principally a rostrum size component with high positive loadings for rostrum length and rostrum straightness.
The PCA for males distributes the species along similar axis (Fig. 2C,D). PC1 explains 82% of the variation among species. Unlike in females, this axis does not contain a measure of curvature and is purely a size axis (Table 2). The cumulate of PC1, PC2, and PC3 explain 95% of the variance. PC2 separates the species on exactly the same constructed value as for the females (straightness of rostrum/curvature of the rostrum), and PC3 separates the species on the rostrum length/elytra length construct.
Ecomorphological Relationships
A canonical correlation analysis was used to compare the positions of species in the ecological space and the morphological space described above. The coordinates of the space were the four log-transformed ecological measurements and the nine log-transformed morphological measurements. The canonical correlation coefficients and associated tests of significance are presented in Table 3. Bartlett's chi-square approximation of Wilk's Λ to test overall significance rejects the null hypothesis of no association between the ecological and morphological data sets (females: Wilk's Λ = 0.045, F36,65 = 29.61, P = 0.001; males: Wilk's Λ = 0.049, F36,69 = 28.07, P = 0.001). However, only one canonical variate exceeds a correlation factor of 0.80 in each analysis and differs significantly from zero (P = 0.005). The small sample size in the CCA makes statistical tests very conservative. Because the second correlation is high (0.75 for females and 0.74 for males), we are justified in retaining it in subsequent analyses.
To interpret patterns of covariance between ecological and morphological variables, the correlations between positions of species on the canonical axes and their values for the original variables were calculated (Tables 4, 5). These correlations are more readily interpretable than the canonical vectors themselves because they show how much each variable contributes to the canonical structure and reveal affinities and contrasts among variables. The results of two analyses are presented: (1) within-set correlations describe the contribution of each morphological and each ecological variable to its own canonical axes; and (2) between-set correlation describe the relation of each morphological variable to the canonical axes in ecological space and vice versa. For the ecological variables, the first canonical variate is associated primarily with cup thickness and needle length (r2 = 0.99 for females and r2 = 0.82 for males). Seed diameter is also an important variable for the females (r2 = 0.61). The second canonical axis is related primarily to seed diameter in females and cup thickness contributes least, while no particular variable is correlated strongly with the second axis in the male dataset. For the morphological variables in males and females, the first canonical axis relates mainly to overall size but does include shape information as previously explained for the PCA. The second canonical axis on the other hand differs for males and females. In females it has contributions from rostrum length (r2 = 0.11) and length of the straight part of the rostrum (r2 = 0.13) whereas in males the rostrum is not strongly correlated to the canonical axis. This axis contains the shape information of the various species.
The between-set correlations reveal the direct contribution of each variable in one dataset to the canonical variates of the other data set. Because of the high correlation between the first canonical variate in ecological and morphological space, the between-set correlations for the first canonical variates are similar to the within-set correlations. To a lesser extent, the same is true of the between-set correlations for the second canonical variates. In females, the first morphological canonical variate (CV1) is strongly correlated with cup thickness, while morphological CV2 is correlated primarily with seed diameter. The ecological CV1 exhibits correlations with relative body size and the ecological CV2 exhibits the highest correlation with straightness of rostrum. In males, morphological CV1 correlated with seed volume and morphological CV2 correlates with seed height while ecological CV1 correlated with overall size and ecological CV2 correlates with elytra length. Thus, the rostrum traits do not explain as much of the variation in the male morphology as in the females. The position of the females along morphological CV2, which contains variation related to shape (length, straightness, and curvature of rostrum), was used to determine whether there were significant relationships between shape of the various weevils and the various ecological variables for both males and females. The morphological CV2 was significantly correlated to all ecological variables except to the cup thickness, but in all cases the correlation coefficients were low (Table 6).
Another CCA was conducted on the female dataset including only the Curculio species predating acorns, that is, excluding the species feeding on chestnuts and pecans and other non-Fagaceae. Similarly to the latter analysis, the first canonical correlation was significant (Wilk's Λ = 0.015, F36,65 = 62.68, P = 0.001) but the second had a low correlation value (r2 = 0.49). Thus, within the acorn-feeding Curculio as in all Curculio species studied, the shape of the weevil rostrum does not appear to correlate strongly to any of the acorn characteristics but the body and rostrum variables are important factors correlated mainly with seed diameter and cup thickness.
Morphological Comparisons between Males and Females
Ecomorphological relationships that were substantial among males and females were selected. Both the ecological and morphological variables that have 60% of their variation respectively associated with the morphological and ecological canonical variates qualified as significant (Table 7). Based on this criterion, eight relationships were extracted, four significant only among females, four significant only among males. Only two relationships were significantly different between males and females. Females have a longer straight part to their rostrum that correlates with the seed volume (Fig. 3A) and have a greater eye-antennal insertion distance that correlates highly with cup thickness (Fig. 3B), although the slopes of these relationships are not significantly different between males and females, suggesting that there is not a proportional change in the length of the rostrum in females compared to the males. Similarly, other morphological traits such as the thorax and elytra length do not show significant differences in their slopes or intercepts in males and females (Table 7).
Independent Phylogenetic Comparisons
The results of the comparative analysis by independent contrasts show that PC3 (rostrum size) is positively correlated with seed diameter, cup thickness, and seed volume (Table 8, Fig. 4). This is also the case, when the data are analyzed in Continuous with no scaling transformation. Importantly, the correlation of PC3 (rostrum size) and the seed diameter stands to scrutiny under all analyses (Table 8). However, although PC1 (inverse body size) is correlated with seed diameter in the CCA (Fig. 4) and in Continuous, when this trait is analyzed by independent contrasts in CAIC, the relationship is not significant. On the other hand, none of the seed traits correlate with PC2 (rostrum shape) in the various phylogenetic analyses.
The scaling parameter λ (Table 9), which is not significantly different than zero, suggest that all three PCs are evolving among species as if they were independent, that is, they do not covary in direct proportion to their shared evolutionary history, and thus justify the results of the CCA. Although the value of κ remains undefined for PC2 (not significantly different than zero or one), κ is significantly greater than one for PC1 and PC3, thus indicating that longer branches contribute more to changes in the traits of PC1 and PC3, that is, body size and rostrum size evolution is greater in longer branches. For example, C. caryae and C. pellitus have longer phylogenetic branches and longer rostra than their sister taxa (Fig. 5). Simultaneously, the values of δ, which are not significantly different than one for all three PCs, imply that the rate of trait evolution for the PCs has been gradual from the root to the tips as well as for the shared path lengths.
These results, which take into account the phylogenetic relationships of 15 Curculio species, support the previous results found in the CCA of 30 species, namely that rostrum size and body size (except in CAIC) correlate with seed diameter, and suggest that the rostrum size is an important morphological characteristic for the adaptation to various seed sizes but that the shape of the rostrum does not seem to play a role. These analyses, which took into account the phylogenetic relatedness of species, show that the historical changes in seed use have coincided with evolutionary changes in rostrum size of the various species within the genus (Fig. 5).
Discussion
The distribution of Curculio species within morphological space exhibits distinctions between the various feeding categories both for males and females. The species that predate pecans (C. caryatripes, C. caryae) form an ecomorph with large bodies and long straight rostra whereas the species that oviposit in galls (C. rubidus, C. cerasorum, C. salicivorus), the gall ecomorph, have small bodies; these species have previously been classified in the subgenus Balanobius. The remaining species, which mainly predate acorns, are grouped together and can be differentiated mainly along PC2 (straightness of rostrum/curvature of rostrum). On the edge of the oak ecomorph space, various species with divergent life strategies can be found. Curculio elephas, found both on oaks and chestnuts in Europe, has a different morphology than that of species strictly found on oaks, its rostrum is longer and straighter. Curculio villosus found in galls of Biorrhiza pallida on oak trees has a morphology that tends toward that of the other gall-feeding species but due to its larger size it has been grouped within the oak ecomorphs in the PCA. For this reason, taxonomists have not classified C. villosus into the subgenus Balanobius with other gall-feeding species. The selection pressure imposed by similar ecological resources of the hazelnut and the acorn has probably driven the hazelnut weevils, C. nucum and C. neocorylus, toward similar morphological characteristics as the oak ecomorphs. The males appear to cluster into these same ecomorphs, that is, their morphology shows variation for the same traits as the females.
The hypothesis that morphological differences among Curculio species are associated with ecological differences was assessed in this study by determining the concordance between ecological and morphological distributions of species. In our analysis, a significant association was revealed between the ecological variables and morphological variables of each species for both males and females. Significant relationships were observed between the overall size of the species and the cup thickness as well as the seed volume. Thus, Curculio species appear to adapt to differences in resources by changes in body size. As holometabolous insects, Curculio body size is influenced by the larval environment. Differences in nutrient availability during larval development influence how large a larva grows to be, and hence partially determines the final adult size of the insect and perhaps, in turn, the adults performance. These circumstances suggest a strong physical interdependence between the body size and the size of the nut in which the species develops. Evolutionary change in seed choice should cause relatively rapid reciprocal adjustment in body size. Indeed, species attacking chestnuts and pecans have larger bodies than species found in smaller nutrient-poor galls. Ecological evidence suggests that lower larval survival in small seeds represents the main selective agent for increased body size in Curculio (Harris 1976; Forrester 1991; Desouhant 1998; Desouhant et al. 2000). Literature reviewed for other insect species predating seeds such as the Bruchinae (Chrysomeloidea), also suggested that larger seeds conferred a higher offspring survival (Mitchell 1975), and in some cases the larvae developed faster or into larger adults or adults that produced eggs at a faster rate (Moegenburg 1996). For example, the smaller the seed, the smaller the resultant bruchid (Dickason 1960; Center and Johnson 1974). As such, size-dependent resource use, mediated by larval survival, may be one of the primary adaptive processes driving the diversification of Curculio weevils and other seed-feeding insects. However, the seed characteristics measured in this study are probably not the only ecological dimensions differentiating the various Curculio species; plant chemistry, for example, is probably also a fundamental ecological factor.
Among females, ecological variation correlated with body length and weakly with rostrum shape, whereas in males it correlated with body length alone. Moreover, the eye-antennal insertion and the straight part of the rostrum are significantly longer in females, whereas the correlation between overall size measurements and seed volume or cup thickness did not differ significantly between males and females, that is, males and females are morphologically similar except for the rostrum of females having a longer straight section. This is likely due to the function of the rostrum in oviposition site selection and this dimorphism probably evolved prior to the diversification of the genus. Contrasting males and females in these datasets suggests that rostrum length has increased at a similar rate in both sexes. Although, the rostrum trait in the genus appears to be exaggerated in certain species and always longer and straighter for females, this change in size appears to be largely allometric because it does not vary disproportionately in females as opposed to males. The lack of correlation between any seed characteristics and rostrum shape could be due to the type of variables measured for the seed. Perhaps a correlation between the rostrum shape (i.e., curvature) and the seed could have been found had additional seed characteristics been taken into account. Seed hardness, for example, might have been an interesting variable to correlate to rostrum shape, although it would be more complex to measure than the four variables used in this study.
This study does not necessarily imply that the rostrum does not have a different function in males and females. On the contrary, it is quite clear that the rostrum plays a vital role in oviposition site preparation for the females. This work does, however, suggest that there is no selection component for a disproportionately longer rostrum in females. Static allometry for rostrum size does not appear to be the case across all the Curculionidae. Evidence for selection for increased rostrum length due to an oviposition-associated role comes from a study of variation in structural features of Anthonomus quadrigibbus (Burke and Anderson 1989). Rather than comparing rostrum length between species, this example contrasts within-species variation in rostrum length of males and females. The data on structural variation of A. quadrigibbus demonstrate that adult beetles collected or reared from different hosts differ largely in absolute size and other features dependent on the variation in size. These are attributes undoubtedly due, at least in part, to variation in size or quality of the host fruit in which the larvae developed and are thus, in part, ecophenotypic. However, there are also differences in relative rostrum length in larger females. These differences do not appear attributable solely to variation in body size but rather suggest selection for a longer rostrum, which, because of its role in oviposition, is associated with variation in available host fruit size (Burke and Anderson 1989; Anderson 1995). Similarly, Tychius soltaui and T. prolixus use hosts, Astragulus mollissimus and A. utahensis, respectively, which have densely pubescent pods, and females of weevil populations associated with these plants have longer, more slender rostra than do populations of the same species associated with plants that lack the pilosity of the pods. With these long rostra the weevils are apparently able to penetrate the pilosity (Clark and Burke 1977). However, whether this trait varies disproportionately to body size is not emphasized.
The pattern observed in the genus Curculio support the hypothesis that morphological diversification is strongly related to seed size use. However, it is well known that shared evolutionary history can explain correlations between character sets, and hence similarities between closely related species may not reflect adaptive processes (Felsenstein 1985; Harvey and Pagel 1991). Therefore, the observed association between insect morphology and seed characteristics was analyzed in the context of phylogeny to determine whether the observed correlation is consistent with shared evolutionary history. Our results show that ecological substrate choice and morphological parameters are correlated even in independent comparative analyses. Indeed, the evolutionary comparison between seed diameter and body size and seed diameter and rostrum length indicates that host shifts toward larger seeds were accompanied by increased morphological changes toward longer rostra and at the same time larger body sizes in the genus Curculio, independent of the phylogenetic relationships of the taxa and that across the phylogeny longer branches contributed more to the variation in rostrum size and body size. For example, C. caryae and C. elephas both have long straight rostra with short curved parts, predate large seeds, and have longer phylogenetic branches than their sister taxa, and yet C. caryae and C. elephas are not sister taxa, suggesting that the shift to a host with large seeds results in the adaptation of a longer straighter rostrum and a proportionately larger body. The seed size selection by the female thus is likely an important selective pressure on larval survival underlying the ecomorphological diversification of the genus. Direct observations of seed size selection and improved performance with a longer rostrum would have to confirm this conclusion.
Changes in shape are not as evident within the genus; for example, there does not appear to be selection for disproportionate changes in rostrum size, despite this parameter being an important female morphological characteristic in oviposition site preparation. While a long rostrum and large body size are closely correlated, it is plausible that the selection is primarily exerted on the rostrum, but this parameter cannot easily be uncoupled from other body proportions, such as overall size. If this is correct, selective forces acting to maintain the correlation of both body size and rostrum length preserve these allometric relations. It is unclear if these forces reflect ecological and functional constraints or represent a correlation at the level of developmental pathways. Perhaps, evolutionary shifts in shape are more likely to be found at taxonomic levels higher than the species and genus boundaries.
The cause of species diversity in the Curculionidae has been attributed to the evolution of the rostrum as an oviposition tool (Zwölfer 1975; Anderson 1995) and the resulting opportunities for diverse associations with their angiosperm food plants (Farrell 1998; Marvaldi et al. 2002). However, the mechanism of species diversification remains to be explained. Ecomorphological studies can quantify the differences between species and correlate these to function. In the genus Curculio, evolutionary changes in rostrum traits and overall body size appear to be correlated with the variation in the substrate. As these changes are largely uncoupled from phylogenetic history, this corroborates the notion that selection is shaping the morphological diversity. The precise path of evolution remains to be established with better taxon sampling for a complete species-level analysis, but the current data support a scenario unlike classical examples of adaptive radiation where rapid early evolution has been followed by slower rates of change among closely related species. Instead, in Curculio, variation is reconstructed to occur at the species-specific level where longer branches have contributed more to trait evolution. In part, these differences represent the effects of random switches to different food resources, which trigger the functional-morphological variation. Within the main class of substrates, the acorns of oak trees, these adjustments in size as a result of substrate size are mediated by a straightforward mechanism, variation in total body size, which also greatly affect the rostrum size, and hence there is a direct correlation of larval food source and rostrum size used to attack a particular size substrate. Hence, the diversification process in Curculio is likely mediated by the variation of the substrate itself, rather than driven by competition of beetle populations and resource segregation.
In conclusion, we have presented a refined example of correlated divergence of phenotype and use of the environment, providing a mechanism for the evolution of the diversity in a seed-feeding lineage where mechanical rather than chemical plant defenses drive diversification. Viewed in the context of the evolution of all the weevils (Marvaldi et al. 2002), this study adds evidence for the adaptive radiation undergone by the Curculionoidea.
Acknowledgments
We thank T. G. Barraclough and C. Lyal for comments on earlier drafts of the manuscript and C. W. O'Brien for lending his personal collection of Curculio specimens. We are also grateful to the associate editor and the two anonymous reviewers for their constructive comments. This work was funded by the Natural Environment Research Council and the Natural History Museum, London.
Literature Cited
Fig. 1.
The host-association phylogeny based on the consensus of 5795 most parsimonious trees (length = 2094, CI = 0.53, RI = 0.73) from the combined analysis of two mitochondrial (cyt b and COI) and two nuclear (EF1-α and Pglym) genes. Support for the branches is provided by bootstrap support above the branch and Bremer support below. The branch shading mapped onto the phylogeny by parsimony indicates the host plants on which the Curculio weevils feed (Hughes and Vogler 2004)
Fig. 2.
Principal component analysis: (A) PC2 against PC1 and (B) PC3 against PC1 for 30 females; and (C) PC2 against PC1 and (D) PC3 against PC1 for 31 males. The nine morphological variables are represented by the eigenvectors: elyt, elytra length; tho, thorax width; abdo, abdomen width; tib, tibia length; fem, femur length; rostrum, rostrum length; ante, eye-antennal insertion distance; strai, length of the straight part of the rostrum; ang, angle of the curved part of the rostrum
Fig. 3.
Regressions (A) between the length of the straight part of the rostrum and the seed volume; and (B) between the eye-antennal insertion distance and the cup thickness for males (white squares) and females (black circles). All values are ln-transformed
Fig. 4.
Linear regressions analysis of the comparative analysis data: (A) the PC1 (inverse body size) is negatively correlated to seed diameter in CAIC (F = 2.07, df = 14, n.s. but significant in Continuous); (B) the PC3 (rostrum length) is positively correlated to the seed diameter in CAIC (F = 20, df = 14, P < 0.001) and (C) the PC3 (rostrum length) is positively correlated to the cup thickness and needle length (F = 5.7, df = 14, P < 0.05). The seed measurements were ln-transformed and the trend line is forced through the origin
Fig. 5.
Evolution of rostrum size and seed association in the genus Curculio using the topology of one of the most parsimonious trees from the total evidence (COI, Cyt b, EF1-α, Pglym) with relative branch lengths estimated in PAUP using the maximum likelihood model HKY85 + Γ with the more complete mitochondrial data (Cyt b and COI). The rostrum pictures are scaled relative to one another; the seed diameter and rostrum length have been drawn at an approximate scale of 1:48. Bootstrap values are above the branch and Bremer support are below
Table 1.
Loadings from a principal components (PC) analysis of nine log-transformed morphological characteristics for females of 30 Curculio species. Substantial loadings (i.e., > 0.80) are marked in bold
Table 2.
Loadings from a principal components (PC) analysis of nine log-transformed morphological characteristics for males of 31 Curculio species. Substantial loadings are marked in bold
Table 3.
Summary of canonical correlation analyses for the females of 30 and males of 31 Curculio species.
Table 4.
Loadings of nine morphological and four ecological variables for the first two canonical variates for females of 30 Curculio species. Numbers in parentheses are proportions of variation of each variable associated with an ecological or morphological canonical variate. Correlations that are substantial (i.e., proportions of variation > 0.60) are marked in bold
Table 5.
Loadings of nine morphological and four ecological variables for the first two canonical variates for males of 31 Curculio species
Table 6.
Linear regression between the position of each species on the second morphological canonical variate (CV2) and the four ecological variables. Degrees of freedom for the F-value are 3 and 57
Table 7.
Pearson correlation coefficients between ecological and morphological variables for 30 females (df = 28) and 31 males (df = 29). All slopes tests have 1 and 58 df, while intercepts tests have 1 and 59 df
Table 8.
Analysis of phylogenetic dependence of the morphological and ecological data. Correlation (r-values) of the seed characteristics (log-transformed values) against the respective principal component using (a) CAIC and (b) Continuous with no scaling transformations (λ = 1.0, δ = 1.0, κ = 1.0), (c) Continuous using the scaling parameters from Table 9. The significance of the correlation was calculated by comparing a model with zero covariance between the traits to an alternative model with covariance
Table 9.
Scaling parameters estimated in Continuous and significance of the likelihood ratio test. The parameters λ and κ were estimated simultaneously followed by δ using each principal component (PC) in turn against the seed characteristics. The values in parentheses represent the 95% confidence interval of the scaling parameter
