SEARCH

SEARCH BY CITATION

Keywords:

  • coevolution;
  • cospeciation;
  • divergence times;
  • evolutionary rates;
  • fossil record;
  • host shift;
  • leaf-mining moth;
  • penalized likelihood;
  • plant–insect interactions;
  • sequential evolution

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Coevolution has been hypothesized as the main driving force for the remarkable diversity of insect–plant associations. Dating of insect and plant phylogenies allows us to test coevolutionary hypotheses and distinguish between the contemporaneous radiation of interacting lineages vs. insect ‘host tracking’ of previously diversified plants. Here, we used nuclear DNA to reconstruct a molecular phylogeny for 100 species of Phyllonorycter leaf-mining moths and 36 outgroup taxa. Ages for nodes in the moth phylogeny were estimated using a combination of a penalized likelihood method and a Bayesian approach, which takes into account phylogenetic uncertainty. To convert the relative ages of the moths into dates, we used an absolute calibration point from the fossil record. The age estimates of (a selection of) moth clades were then compared with fossil-based age estimates of their host plants. Our results show that the principal radiation of Phyllonorycter leaf-mining moths occurred well after the main radiation of their host plants and may represent the dominant associational mode in the fossil record.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

The associations between vascular plants and insects or their early arthropodan relatives have a fossil record dating back to the first known terrestrial ecosystems of the early Devonian, 415–400 Ma (Kevan et al., 1975; Labandeira, 2002). More recently, during the Cretaceous, from 145 to 65.5 Ma, flowering plants and their insect herbivores have dominated terrestrial biomes and currently constitute a major fraction of the earth's biodiversity, perhaps accounting for as much as 50% of all described species (excluding micro-organisms; Strong et al., 1984). Since the pioneering work of Ehrlich & Raven (1964) on the coevolution of butterflies and their host plants, there has been great interest in trying to detect and understand macroevolutionary patterns in insect–plant associations (Farrell, 1998b, 2001; Sequeira & Farrell, 2001; Percy et al., 2004; Kergoat et al., 2005). This research has received added impetus in recent years due to the increasing availability of molecular phylogenies of speciose groups of plants and their herbivores.

There are now many established examples of clades of herbivores that have radiated on phylogenetically related groups of plants (e.g. Farrell, 1998a; Ronquist & Liljeblad, 2001; Weiblen & Bush, 2002). Less clear is the temporal relationship between the plant and insect diversifications. At least three patterns are possible, but these should be seen as parts of a continuum rather than discrete categories. First, the insects might speciate contemporaneously with their host plants (cospeciation) so that their phylogenies are congruent. Secondly, insect diversification may follow rapidly (in geochronological terms) after plant diversification (‘fast colonization hypothesis’), but with the pattern of plant colonization influenced by some factors in addition to host-plant phylogeny, resulting in an insect radiation that is slightly younger and incongruent with that of the plants. Thirdly, colonization of plants may require an evolutionary innovation and occur only when this arises, some time after plant diversification (‘delayed colonization hypothesis’). Again, an imperfect match between the phylogenies is predicted, but now we also predict a time lag before colonization.

Testing the above alternative scenarios ideally requires phylogenies of both taxa, along with fossil data for both groups, to provide absolute time calibrations. Recently, Farrell (1998a) studied the evolution of host-plant use among herbivorous Phytophaga beetles (weevils, longhorned and leaf beetles), using phylogenetic and fossil data. He showed two bursts of beetle speciation: one during the mid-Cretaceous (90 Ma) and the other in the later-Cainozoic (10 Ma). This bimodal distribution may be attributable to three events: (1) a major spurt of herbivore diversification coinciding with the mid-Cretaceous ecological expansion of angiosperms, (2) extinction of many specialized associations at the K/T boundary and (3) a gradual rebound to late-Cretaceous species diversity levels during the earlier Cainozoic (Wilf et al., 2001; Labandeira et al., 2002).

More recently, Percy et al. (2004) compared fossil-dated molecular phylogenies of both jumping plant-lice (Psylloidea) and their legume host plants. They found that there was significant topological congruence of the phylogenies, but that the psyllids radiated after their legume hosts, except in the geologically recent Canary Islands where psyllids and genistoid legumes radiated contemporaneously.

So far, attempts to time the diversifications of insect and plant groups have focused mainly on external feeders. Here, we explore this issue in a speciose genus of leaf-mining moths, Phyllonorycter (Lepidoptera: Gracillariidae). Radiations of internal feeders, such as leaf miners, may differ from those of external feeders. For example, the larvae of leaf-mining moths have very intimate associations with their host plants and this may mean that there are more host-imposed constraints on their ecology and evolution compared with externally feeding herbivores. Several authors have suggested that internal feeders show both greater host-plant specificity and greater phylogenetic conservatism in colonizing new species of plants (e.g. Powell, 1980; Mitter & Farrell, 1991; Lopez-Vaamonde et al., 2003).

There are also some particular advantages to studying these issues with leaf-mining microlepidoptera. First, the earliest fossil evidence for leaf-mining moth clades extends to at least 99 Myr, almost to the start of the major explosion of angiosperm diversity during the Cretaceous (130–90 Myr; Powell, 1980; Powell et al., 1999). Secondly, the fossil record preserves important host-plant associations of leaf-mining insects, often in exquisite detail, and replete with larval behavioural and developmental characters that often are assignable at mid-level taxonomic ranks to extant taxa (Labandeira et al., 1994). This facilitates the temporal calibration of the insect phylogeny, which is more difficult in other insect groups.

In a previous study, we found no evidence for cospeciation between Phyllonorycter and their host plants (Lopez-Vaamonde et al., 2003). However, Phyllonorycter species from the same host-plant genus frequently form a monophyletic clade. Closely related moth species often feed on closely related plants, which we interpret as due to host switching (colonization followed by speciation) being more likely to happen amongst phylogenetically related plants.

The aim of this paper was to test whether the main diversification of Phyllonorycter moths occurred near-contemporaneously (on a geochronological time scale) with that of their host plants, or after a significant time lag. To do this we generated a molecular phylogeny of 100 species of Phyllonorycter leaf-mining moths using 28S rDNA sequence data. We then used a penalized likelihood method with a Bayesian approach to estimate divergence times, a method that takes into account phylogenetic uncertainty. We converted the relative ages of the moths into absolute dates using a calibration point from the fossil record. The absolute ages of the moths were then compared with fossil-based estimates of the dates of diversification of their host plants.

Material and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Insect sampling

Moths were reared and collected as described by Lopez-Vaamonde et al. (2003). We used several other genera of gracillariids (from the subfamilies Gracillariinae and Phyllocnistinae as well as the Lithocolletinae to which Phyllonorycter belongs), one species of Bucculatricidae (in the same suborder) and one species of Micropterigidae (a basal group of lepidopterans) as outgroups (see Appendix S1).

Data generation and laboratory procedures

We estimated phylogenies using sequence data from an approximately 1000 bp fragment of the 28S nuclear ribosomal subunit gene. We added 21 new outgroup species and 22 new Phyllonorycter species to the published 28S rDNA phylogeny of 77 Phyllonorycter species in Lopez-Vaamonde et al. (2003). All new sequences have been deposited in GenBank (accession numbers: AY521484-AY521528; see Appendix S1). We used the polymerase chain reaction and sequencing procedures described by Lopez-Vaamonde et al. (2001). 28S sequences showed substantial length variation and were aligned following the procedure of Lopez-Vaamonde et al. (2003). The DNA alignments are available from treebase (http://www.treebase.org/treebase/; study accession number = S1423).

Estimating phylogenies

We used Markov Chain Monte Carlo (MCMC) methods (Larget & Simon, 1999) within a Bayesian framework to estimate the posterior probabilities of the phylogenetic trees. By adopting this Bayesian approach, we take account of phylogenetic uncertainty, both with respect to topology and to branch lengths, in our estimates of divergence age.

We used hierarchical likelihood ratio tests in the computer program modeltest 3.0b6 (Posada & Crandall, 1998) to select the best model of nucleotide substitution, which was a general time reversible (GTR) model, allowing for rate heterogeneity across sites assuming a discrete γ-distribution and for a proportion of sites to be invariable (GTR + I + Γ). The phylogenies were then estimated using mrbayes v3.0b4 (Huelsenbeck & Ronquist, 2001) and the MCMC was run for 106 cycles, sampling one tree every 1000th cycle. After filtering the trees for our defined nodes of interest, a total of 8368 trees remained for use in the age estimation analyses.

Estimating divergence times

We estimated node ages for the moths using a penalized likelihood approach in the computer program r8s v. 1.50 (Sanderson, 2003; http://ginger.ucdavis.edu/r8s/). For each node of interest (see below) the posterior distribution of the divergence time, conditioned on both the phylogenetic model and the calibration point obtained from the fossil record, was obtained by local density estimation using the program locfit (Loader, 1999), implemented in the ‘R’ statistical package (Ihaka & Gentlemen, 1996 see script in Fig. S1). To summarize the fitted distributions, we report their modes and 90% highest posterior density (HPD) limits (Table 1 and Fig. S1). The mode represents the most likely value for the divergence time of a particular node, and the HPD provides a confidence interval for this estimate. It is important to emphasize that we have inferred age from the entire posterior probability distribution of the 8368 trees sampled rather than using a single topology, so our age estimates account for both branch length and topological uncertainties.

Table 1.   Estimates of the age when particular plant groups were colonized by Phyllonorycter leaf-mining moths (column 3).
TaxonNumber of nodeMode (Myr)90% LHPD90% UHPDProbability (%)
  1. Number of node is indicated on the cladogram (see Fig. 3).

  2. LHPD, 90% lower highest posterior density; UHPD, 90% upper highest posterior density; probability (%), the figures represent the probability (as averaged over the 100 randomly selected trees from the posterior distribution) that a particular plant group is reconstructed for our node of interest and not reconstructed for the node below (i.e. the colonization event has occurred along the stem lineage to our node of interest).

Fagales237.827.350.832
Malpighiales237.827.350.827
Populus329.921.142.576
Ulmus432.624.548.479
Acer513724.579
Fagaceae624.414.634.151
Lyonia718.810.927.268
Malus811.47.617.743
Alnus925.718.136.297
Corylus1011.37.417.598
Salicaceae1115.911.124.465
Prunus129.66.216.298
Acer131812.127.873
Carpinus1414.89.822.281
Quercus1515.81124.447

The ages of the host-plant groups were obtained from the literature based on the fossil record (Table 2).

Table 2.   Earliest fossil occurrences of plant taxa that currently are hosts for Phyllonorycter leaf-mining moths.
Host-plant taxonEarliest occurrenceAbsolute age (Myr)*TypeReferences
EpochStage
  1. *To nearest 0.5 Myr; stage-interval data are given as midpoints; some are radioisotopic age dates. Time scale is from Gradstein & Ogg (2004).

  2. †The assignment is ‘aff. Carpinus’; of which the same deposit has the very closely related Paleocarpinus, also a member of the subfamily Coryleae.

  3. ‡Fagales comprises the Betulaceae, Casuariniaceae, Fagaceae, Juglandaceae, Myricaceae, Nothofagaceae and Rhoipteleaceae (Judd et al., 1999).

  4. §Malpighiales comprises the Malpigiaceae, Clusiaceae, Euphorbiaceae, Chrysobalanaceae, Rhizophoraceae, Salicaceae, Passifloraceae, Violaceae and Flacourtiaceae (Judd et al., 1999).

  5. ¶Leaves of Malus are difficult to distinguish from Pyrus.

AcerPaleoceneThanetian56.5Leaves, fruitCrane et al. (1990)
AlnusEocene[Middle] Lutetian48.5Leaves, fruitsCrane (1989)
CarpinusEocene[Middle] Lutetian49.0FruitWilde & Frankenhäuser (1998)
CorylusEocene[Middle] Lutetian48.5FruitPigg et al. (2003)
Fagales‡GallicSantonian85.0PollenDoyle & Robbins (1977), Romero (1986), Hill & Dettmann (1996), Sims et al. (1998), Herendeen et al. (1999)
FagaceaeGallicUpper Santonian84Staminate inflorescences, flowers, cupules and fruitsSims et al. (1998)
LyoniaMiocene[Middle] Langhian15.5Fruits, seedsFriis (1985)
Malpighiales§Late-CretaceousTuronian91.5FlowersCrepet & Nixon (1998)
MalusOligocene[Early] Rupelian31.0LeavesMeyer & Manchester (1997)
PopulusPaleocene[Late] Thanetian55.0LeavesS. R. Manchester, personal communication to CCL
PrunusEocene[Middle] Lutetian48.0FruitsCevallos-Ferriz & Stockey (1991); Manchester (1994)
QuercusEocene[Middle] Lutetian48.5AcornsManchester (1994)
SalicaceaeMiddle-Eocene[Middle] Lutetian45Reproductive foliage and leavesManchester et al. (1986)
UlmusEocene[Middle] Lutetian48.0Leaves, fruitManchester (1989, 1999)

Character mapping and phylogenetic uncertainty

We mapped the character ‘host-plant taxa’ onto the moth phylogeny using Bayesian methodology in order to incorporate phylogenetic and mapping uncertainty. Unfortunately, current software (mrbayes and simmap, http://www.simmap.com) does not allow sufficient character states for this analysis. Consequently, we reconstructed the evolution of host use manually on 100 trees drawn randomly from the Bayesian posterior distribution (the 8368 trees sampled, filtered for our nodes of interest) using mesquite version 1.0 (build e58; http://mesquiteproject.org; see Table S1). We selected clades that grouped closely related moths feeding on closely related plants (same host-plant genus, family, order). On this basis, 15 nodes of interest, supported by posterior probabilities of at least 92%, were chosen for insect and plant clade age comparisons (Tables 1 and 2).

Crown-group and stem-group ages

The distinction between crown and stem-group ages is crucial when dealing with colonization events. Consider a clade of insects that all feed on the same host-plant taxon. A colonization event is inferred to have occurred somewhere along the stem lineage to this clade. However, the date of this colonization event can be estimated either as the date of origin of the stem group or as the date of origin of the crown group (see Fig. 1). We have consistently used stem-group ages for both moth and host-plant groups in our analyses. In terms of testing for a significant delay between plant and insect radiations, insect stem-group ages are the conservative option as they provide the older date. For the host plants, crown-group ages (younger) would be more conservative for the hypothesis test. However, palaeobotanical reports of the first occurrence of particular plant groups rarely identify the taxon as a stem-group lineage or part of the crown group of living species. Generating crown-group age estimates would therefore require an extensive taxonomic evaluation of both fossil and living plant taxa, clearly beyond the scope of this study. Therefore, we use stem groups wherever possible; one exception being the estimate for the origin of Phyllonorycter. This taxon was used to establish a diversification rate estimate based only on numbers of extant species, so a crown-group estimate is more appropriate.

image

Figure 1.  Distinction between crown-group and stem-group ages (based on Magallón, 2004).

Download figure to PowerPoint

Fossil record and calibration points

To convert the relative ages obtained through the penalized likelihood analysis into dates, we used an absolute calibration point from the fossil record. Mines of the gracillariid subfamily Phyllocnistinae have been recorded from magnoliid dicot leaves originally reported to be at least 97 Myr old (Labandeira et al., 1994; Fig. 2), but this date has now been revised to at least 99 Myr, based on new evidence (Brenner et al., 2000; Wang, 2002; Gradstein & Ogg, 2004).

image

Figure 2.  Gracillariidae (Phyllocnistinae) leaf mine on host plant Rogersia parlatorii Fontaine, whose affinities is closest to extant Lauraceae within the order Laurales cf. Lauraceae. (a–c) Specimen UF7351, showing a camera lucida drawing of the entire leaf (a), terminal mine chamber (b) and mine loop adjacent to midrib at the leaf base (c). (d–g) Specimen UF4818, displaying a camera lucida drawing of the entire leaf (d), mine course near leaf apex (e), terminal chamber (f) and oviposition site with adjacent frass trail from early instars (g). Both specimens are from the Cenomanian-Albian boundary interval of the Dakota Formation and were found at the Braun's Ranch Locality, Cloud Co., Kansas, USA.

Download figure to PowerPoint

These mines represent the oldest fossil evidence for the family Gracillariidae (Kristensen & Skalski, 1999; see Table S2). However, it is unclear whether they should be assigned to the stem or crown group as the adult morphology is not preserved, and therefore cannot be compared with extant species. Consequently, we applied a 99 Myr calibration point to the Phyllocnistinae stem-group node (the time of divergence of the Phyllocnistinae from its sister taxon).

Estimating diversification rates

We estimate rates of diversification (r) for the genus Phyllonorycter as a whole by using Magallón & Sanderson's (2001) eqn (7) for a crown-group age

  • image

where n is number of species, e is the extinction rate and t the crown-group age.

In addition, we estimate r using two extinction rate values at opposite ends of the spectrum of possible values: one corresponds to zero extinction (r0) and the other to a very high extinction rate (r0.9; for a similar approach see: Strathmann & Slatkin, 1983; Magallón & Sanderson, 2001). We previously estimated the number of extant Phyllonorycter species as 400 (Lopez-Vaamonde et al., 2003) and used this estimate for the age of the crown group obtained in this study.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Bayesian analysis

We obtained 28S rDNA sequences for 100 Phyllonorycter species and 36 outgroup taxa. The aligned D1 + D2 + D3 region sequences comprised 1130 bp. Results of the Bayesian analysis are presented as a 50% majority rule consensus tree in Fig. 3. As in a previous study (Lopez-Vaamonde et al., 2003), Cameraria is the sister group of Phyllonorycter, and Phyllonorycter species are arranged in clades that reflect their host-plant relationships (Fig. 3). The r8s analysis suggests that the genus Phyllonorycter originated during the early Palaeocene (mode =62.3 Myr, lower and higher 90% posterior density limits = 50.3 and 76.3 Myr).

imageimage

Figure 3.  Bayesian tree with posterior probability values above the branches. Nodes of interest (where host-plant colonization events are inferred) are numbered from 1 to 15 and their stem-group age estimates are reported in Table 1.

Character evolution and phylogenetic uncertainty

The mapping of host-plant use onto the 100 randomly chosen trees from the Bayesian posterior distribution (Fig. 3) enabled us to identify 15 host-plant colonization events. Nodes, indicating where these events occurred, are numbered from 1 to 15 on Fig. 3.

When host use was mapped onto Fig. 3, node 2 was matched to a host shift onto the plant order Fagales whereas two other nodes were identified as colonizations of the plant family Fabaceae (results not shown). When phylogenetic uncertainty was taken into account, the robustness of these host-plant colonization events decreased. Thus, the mapping of host use on the 100 posterior trees showed node 2 as a colonization event of Fagales and Malpighiales (Table 1), but there was not a single tree out of the 100 sampled that supported the two nodes as colonization events of Fabaceae (see Table S1). This clearly indicates that using a single topology for reconstructing character evolution can yield misleading results.

Comparing moth and plant age estimates

When we compare the age estimates of the 15 nodes (Table 1) with those of the associated host plants (Table 2), we observe, as expected, that the moths are younger than the plants. For example, we placed the origin of the genus Phyllonorycter in the Palaeocene (62.3 Myr), considerably postdating the origin of both extant angiosperm clades (Early Cretaceous, 130–110 Myr) and extant eudicot subclades (mid-Cretaceous, 110–90 Myr; Stuessey, 2004). Similarly, the colonization of the order Fagales (Fig. 3a, node 2), a major group of hosts for Phyllonorycter, occurred during the late-Eocene (37.8 Myr) whereas the order Fagales originated during the late-Cretaceous (84 Myr; Herendeen et al., 1999; Table 2). From these data, on average, moths are 32.3 ± 16.6 Myr younger than their host plants and this difference is significant (paired t-test: t14 = −7.52, P < 0.0001).

The ages of the host plants were obtained from the fossil record (Table 2) and compared with the sequence-derived ages of the moths. An alternative would be to estimate the ages of the plants from molecular data as well (with fossil data providing minimum age constraints). However, it has been shown that fossil-based ages are nearly always younger than the corresponding sequence-derived ages (Rodriguez-Trelles et al., 2002; Wikström et al., 2003) and hence this would simply increase the time span between host-plant origin and moth colonization.

Rate of diversification

Given that there are about 400 Phyllonorycter species extant today (Lopez-Vaamonde et al., 2003), we can use our estimate of the age of the crown group (62.3 Myr) to obtain a measure of the net diversification rate as a function of the relative extinction rate (e). The values we obtain are 0.04 net speciation events per million years in the absence of extinction (e = 0), and a minimum of 0.03 under the assumption of a high relative extinction rate (e = 0.9).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Contemporary or sequential radiations?

This is the first study to make a rigorous quantitative comparison of the timing of radiations by internally feeding insects and their host plants. Our results provide strong evidence for sequential radiation of the insects onto an already diversified group of plants. The analyses suggest that the principal Phyllonorycter radiation was at least 27.3–50.8 Ma, when these moths colonized plants in the orders Fagales and/or Malpighiales (Fig. 3a, node 2). By contrast, the main radiation of Fagales occurred at least 84 Myr (Hill & Dettmann, 1996; Sims et al., 1998) and that of Malpighiales at least 90 Ma (Crepet & Nixon, 1998).

On average moths took 32.3 ± 16.6 Myr to colonize new plant clades. However, this figure is an underestimate of the true figure because we have no information about the number and age of suitable but as yet uncolonized plant clades. Moreover, fossil calibration dates are underestimates that are more likely to increase rather than decrease with additional fossil data. Regardless, our findings clearly suggest substantial delays (15–52 Myr) between the radiation of host-plant taxa and their colonization by Phyllonorycter moths, rejecting the hypothesis of ‘fast colonization’ and supporting the hypothesis of ‘delayed colonization’ or sequential evolution (Jermy, 1976; Percy et al., 2004). Interestingly, the moth phylogeny is poorly resolved around the time of colonization of Fagales (internal branches are very short, Lopez-Vaamonde et al., 2003) and this probably reflects rapid speciation following this key host shift.

Based on recent studies of the three major types of insect herbivore radiation on new host-plant clades – cospeciation, fast colonization and delayed or sequential radiation (see Labandeira, 2002) – it appears that the sequential mode is the most common. This pattern is supported by several other studies focusing on extant plant host and their insect herbivore clades (Schoonhoven et al., 1998) as well as by our own study. Detection of sequential evolution (Jermy, 1976) in the fossil record requires geochronologically measurable time lags between the origin of a plant clade and its subsequent colonization by insects. In contrast, the rapid radiation of angiosperms that occurred from the later-, early- and mid-Cretaceous is normally thought to be associated with cospeciation or rapid insect diversification (Grimaldi, 1999), based exclusively on insect fossil data. If further studies support this conclusion then our data and the other studies discussed above indicate a different mode of colonization in subsequent colonization events. Consequently, we suggest that delayed sequential colonization may represent the norm for insect herbivores during the overwhelmingly longer durations of geological time that are characterized by more typical levels of plant cladogenesis.

Our data suggest a substantial delay between the radiation of host plants and their colonization by host-specific Phyllonorycter moths. The moths, in turn, also harbour host-plant-specific parasites, namely parasitoid waSPS of the genus Achrysocharoides (Hymenoptera: Eulophidae; Lopez-Vaamonde et al., 2005). It is clear that these waSPS have not cospeciated with either the plants or the moths (Lopez-Vaamonde et al., 2005), but there has not been a formal comparison of the ages of moth and parasitoid radiations. However, comparisons of both nuclear and mitochondrial DNA reveal greater sequence divergences between pairs of moth species than matched pairs of their parasitoids, suggesting strongly that the parasitoids radiated across an already diversified group of moth hosts (Lopez-Vaamonde et al., 2005). Consequently, this plant-leafminer-parasitoid system seems to be characterized by two waves of sequential colonization, which we might call ‘bottom–up sequential radiation’.

Dealing with uncertainty

Our analyses are subject to three potentially important sources of error. These occur in the procedures of phylogeny estimation, character mapping and node dating respectively.

Phylogeny estimation is difficult when based on a single genetic marker due to the stochastic nature of coalescence events (Hudson, 1990). Indeed, the use of a single locus to estimate divergence times can be problematic (Edwards & Beerli, 2000; Arbogast et al., 2002). To address this issue, at least partly, we used a Bayesian framework for phylogenetic reconstruction. It is very probable that obtaining data from one or more additional genes would decrease the error bounds of our age estimates. However, our main result about the timing of speciation is sufficiently marked that it is very unlikely to change.

Any phylogeny generated from data is only an estimate of the true phylogeny and may be little better than some alternative estimates. Conclusions from mapping exercises using a single tree can depend crucially on a few influential (but not necessarily well supported) nodes. Our comparison between mapping on a single topology and on a sample of 100 posterior trees shows that uncertainty in the estimate of a phylogeny estimate can lead to spurious mapping results. Mapping onto 100 trees sampled randomly from a posterior distribution is similar to using nonparametric bootstrapping to account for phylogenetic uncertainty (Ronquist & Liljeblad, 2001). However, neither approach addresses mapping uncertainty – i.e. the error associated with reconstructing the evolution of a character on a given tree – adequately (Ronquist, 2004). Only a full Bayesian approach can account simultaneously for both mapping and phylogenetic uncertainty, but such a procedure has yet to be applied in any empirical study. We hope that future software development will permit analysis of characters with more than 10 states, a constraint that prevented us from using the full Bayesian approach here.

Molecular dating and uncertainty

It is well established that molecular rates vary across lineages, and that this makes it difficult to estimate divergence dates accurately using DNA sequence data (Arbogast et al., 2002). There are two major approaches to dealing with this problem (Welch & Bromham, 2005). The first, nonparametric rate smoothing (NPRS) was developed by Sanderson (1997) and allows different rates on different branches of the tree. It assumes that rates are constant on individual branches and that branch length estimates are their true values. The penalized likelihood method is a superior modification of NPRS that relies on a parameter-rich model to estimate rates on branches and, like NPRS, employs a numerical penalty to avoid rapid changes between neighbouring branches. The relative importance of the model vs. the penalty function is determined by a smoothing parameter, the choice of which is based on the data using cross-validation (Sanderson, 2002).

A second approach was developed by Thorne et al. (1998), and uses Bayesian statistics to model explicitly the rate of evolution in a fully parametric setting. This is a more sophisticated technique, but it would be impractical to apply it to our data for two reasons. First, it is too computationally intensive to apply this approach to the full set of >8000 trees. Secondly, although in theory we could apply it to a suitable consensus tree, there is no satisfactory consensus tree for this data set, because of poor resolution (<50% bootstrap support) at many nodes. We thus chose to use the penalized likelihood method for practical reasons, but also note that this does not necessarily perform less well than the alternative approach (Welch & Bromham, 2005).

Age estimates may be poor if the fossil calibration is inaccurate, or if there is substantial variation in the rate of sequence evolution (Magallón, 2004). As we have shown a significant delay between host-plant radiation and moth colonization, we are most concerned here with factors that might lead us to accept the hypothesis of sequential radiation erroneously, i.e. to underestimate the age of insect clades or overestimate the age of plant clades.

The low estimate of moth diversification rate (see below) might indicate that our calibration point is too old, resulting in Phyllonorycter crown-group age being overestimated. However, any younger age would further strengthen the result of sequential evolution, so this is not a great concern. To favour contemporaneous radiation, a considerably older Phyllonorycter origin would be necessary and there is no fossil evidence to support this. However, this might only reflect the general paucity of phyllocnistine leaf mines in the Cretaceous fossil record, a general concern in paleobiology (Marshall, 1997) and it has been shown recently that divergence dates based on single calibration points tend to be underestimates (Yoder & Yang, 2004). Unfortunately, no other calibration points are currently available, although further studies of plant–insect associations from the Dakota Formation and other Cretaceous biotas may provide new calibration points in the future. Particularly promising are a Cenomanian (approximately 97 Ma) deposit from Israel (Krassilov & Bacchia, 2000) and a younger Turonian (approximately 92 Ma) flora from Kyzl Zhar, Kazakhstan (Kozlov, 1988), both yielding dicot leaf mines, and possibly older strata such as the Barremian (approximately 125 Ma) Yixian deposits from Hebei Province, China (Ren, 1995; Sun et al., 2001) and the Hauterivian (approximately 135 Ma) Baissa deposits from Transbaikalia, Russia (Zherikhin et al., 1999).

Additionally, variation in the rate of sequence evolution can bias both relative and absolute age estimates in a nonlinear way. The direction and magnitude of errors can only be evaluated on an individual node basis by direct comparison with an independent age estimate, typically a fossil. A recent comparison between fossil- and sequence divergence-based age estimates of land-plant lineages indicated that rate variation can impose large error bounds on age estimates (Soltis et al., 2002).

When taken overall, the fossil record supports the hypothesis that the colonization of Fagales by Phyllonorycter moths occurred significantly after the diversification of major fagalean lineages (Table 2).

Interpreting diversification rates

The estimated rate of diversification (r) for Phyllonorycter leaf-mining moths (0.03–0.04) is very slow compared with estimates for other insect groups (i.e. Coleoptera, r ≤ 0.20; Diptera, r ≤ 0.23; Hymenoptera, r ≤ 0.25; Wilson, 1983) and also slower than typical rates in angiosperms [r ≤ 0.077 (e = 9) to r ≤ 0.089 (e = 0); Magallón & Sanderson, 2001]. However, this could be an artefact due to either an underestimate of the number of species (overlooked cryptic species, poor sampling in tropics) or a phylogeny calibration error. Nevertheless, such a slow rate is consistent with other plant host and dependent insect clades that have associations extending to the Cretaceous, including leaf-rolling hispine beetles and their ginger hosts (Wilf et al., 2000) as well as water lilies and their scarab beetle pollinators (Ervik & Knudsen, 2003) and seed-beetles and their legume hosts (Kergoat et al., 2005).

In conclusion, determining the precise timing of phylogenies has become increasingly important to disentangle the evolution of interactions between insects and plants (Percy et al., 2004; Kergoat et al., 2005). Our dated phylogeny supports the hypothesis that the diversification of Phyllonorycter leaf-mining moths occurred well after the diversification of their host plants. This scenario of asynchronous radiations might apply to other predominantly monophagous insect–plant systems and has implications for the generation of biodiversity at geochronological time scales.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

The authors thank the colleagues listed in the Appendix S1 for providing specimens for DNA sequencing. CLV is most grateful to Tosio Kumata, Ohshima Issei and Kazuhiro Sugisima for help during field collections in Hokkaido. We also thank David Dilcher of the Florida Museum of Natural History for the loan and identification of the host-plant specimens (UF7351, UF4818) upon which one of our dates is based. The authors thank Elisabeth Herniou and Tim Barraclough for comments on earlier versions of this manuscript, and Finnegan Marsh for assistance with Fig. 2. This study has been supported by the National Environment Research Council Centre for Population Biology. This is contribution no. 121 of the Evolution of Terrestrial Ecosystems consortium at the National Museum of Natural History.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information
  • Arbogast, B.S., Edwards, S.V., Wakeley, J., Beerli, P. & Slowinski, J.B. 2002. Estimating divergence times from molecular data on phylogenetic and population genetic timescales. Annu. Rev. Ecol. Syst. 33: 707740.
  • Brenner, R.L., Ludvigson, G.A., Witzke, B.J., Zawistoski, A.N., Kvale, E.P., Ravn, R.L. & Joeckel, R.M. 2000. Late Albian Kiowa-Skull Creek marine transgression, Lower Dakota Formation, eastern margin of Western Interior Seaway, USA. J. Sediment. Res. 70: 868878.
  • Cevallos-Ferriz, S.R. & Stockey, R.A. 1991. Fruits and seeds from the Princeton Chert (middle Eocene) of British Columbia: Rosaceae (Prunoideae). Bot. Gaz. 152: 369379.
  • Crane, P.R. 1989. Early fossil history and evolution of the Betulaceae. In: Evolution, Systematics, and Fossil History of the Hamamelidae, Vol. 2: ‘Higher’ Hamamelidae (P. R.Crane & S.Blackmore, eds), pp. 87116. Systematics Association and Clarendon Press, Oxford, UK.
  • Crane, P.R., Manchester, S.R. & Dilcher, D.L. 1990. A preliminary survey of fossil leaves and well-preserved reproductive structures from the Sentinel Butte Formation (Paleocene), near Almont, North Dakota. Fieldiana 20: 163.
  • Crepet, W.L. & Nixon, K.C. 1998. Fossil Clusiaceae from the late Cretaceous (Turonian) of New Jersey and implications regarding the history of bee pollination. Am. J. Bot. 85: 11221133.
  • Doyle, J.A. & Robbins, E.I. 1977. Angiosperm pollen zonation of the continental Cretaceous of the Atlantic coastal plain and its application to deep wells in the Salisbury embayment. Palynology 1: 4378.
  • Edwards, S.V. & Beerli, P. 2000. Perspective: gene divergence, population divergence, and the variance in coalescence time in phylogeographic studies. Evolution 54: 18391854.
  • Ehrlich, P.R. & Raven, P.H. 1964. Butterflies and plants: a study in coevolution. Evolution 18: 586608.
  • Ervik, F. & Knudsen, J.T. 2003. Water lilies and scarabs: faithful partners for 100 million years? Biol. J. Linn. Soc. 80: 539543.
  • Farrell, B.D. 1998a. ‘Inordinate fondness’ explained: why are there so many beetles? Science 281: 555559.
  • Farrell, B.D. 1998b. The timing of insect/plant diversification: might Tetraopes (Coleoptera: Cerambycidae) and Asclepias (Asclepiadaceae) have co-evolved? Biol. J. Linn. Soc. 63: 553577.
  • Farrell, B.D. 2001. Evolutionary assembly of the milkweed fauna: cytochrome oxidase I and the age of Tetraopes beetles. Mol. Phylogenet. Evol. 18: 467478.
  • Friis, E.M. 1985. Angiosperm fruits and seeds from the middle Miocene of Jutland (Denmark). Biol. Skrif. Kon. Dan. Videns. Selsk. 24: 1165.
  • Gradstein, F.M. & Ogg, J.G. 2004. Geologic time scale 2004 – why, how and where next! Lethaia 37: 175181.
  • Grimaldi, D.A. 1999. The co-radiations of pollinating insects and angiosperms in the Cretaceous. Ann. Mo. Bot. Gard. 86: 373406.
  • Herendeen, P.S., Magallón-Puebla, S., Lupia, R., Crane, P.R. & Kobylinska, J. 1999. A preliminary conspectus of the Allon Flora from the late Cretaceous (late Santonian) of central Georgia, USA. Ann. Mo. Bot. Gard. 86: 407471.
  • Hill, R.S. & Dettmann, M.E. 1996. Origin and diversification of the genus Nothofagus. In: The Ecology and Biogeography of Nothofagus Forests (T. T.Veblen, R. S.Hill & J.Read, eds), pp. 1124. Yale University Press, New Haven, USA.
  • Hudson, R.R. 1990. Gene genealogies and the coalescent process. Oxford Surv. Evol. Biol. 7: 144.
  • Huelsenbeck, J.P. & Ronquist, F. 2001. MRBAYES: Bayesian inference of phylogeny. Bioinformatics 17: 754755.
  • Ihaka, R. & Gentlemen, R. 1996. R: a language for data analysis and graphics. J. Comp. Graph. Stat. 5: 299314.
  • Jermy, T. 1976. Insect-host-plant relationships – coevolution or sequential evolution? Symp. Biol. Hung. 16: 109113.
  • Judd, W.S., Campbell, C.S., Kellogg, E.A. & Stevens, P.F. 1999. Plant Systematics: A Phylogenetic Approach Sinauer. Sunderland, Massachusetts, USA.
  • Kergoat, G.J., Alvarez, N., Hossaert-Mckey, M., Faure, N. & Silvain, J.-F. 2005. Parallels in the evolution of the two largest New and Old World seed-beetle genera (Coleoptera, Bruchidae). Mol. Ecol. 14: 40034021.
  • Kevan, P.G., Chaloner, W.G. & Savile, D.B.O. 1975. Interrelationships of early terrestrial arthropods and plants. Palaeontology 18: 391417.
  • Kozlov, M.V. 1988. Paleontology of lepidopterans and problems of the phylogeny of the order Papilionida. In: The Mesozoic-Cenozoic Crisis in the Evolution of Insects (A. G.Ponomarenko, ed.), pp. 1669. Academy of Sciences, Moscow.
  • Krassilov, V.A. & Bacchia, F. 2000. Cenomanian florule of Nammoura, Lebanon. Cret. Res. 21: 785799.
  • Kristensen, N.P. & Skalski, A.W. 1999. Phylogeny and palaeontology. In: Lepidoptera, Moths and Butterflies. Vol. 1: Evolution, Systematics, and Biogeography (N. P.Kristensen, ed.), Handbuch der Zoologie, Band IV (Arthropoda: Insecta), Teilband 35, pp. 725. Walter de Gruyter, Berlin, Germany.
  • Labandeira, C.C. 2002. The history of associations between plants and animals. In: Plant-Animal Interactions – An Evolutionary Approach (C. M.Herrera & O.Pellmyr, eds), pp. 2674, 248–261. Blackwell Science, Oxford, UK.
  • Labandeira, C.C., Dilcher, D.L., Davis, D.R. & Wagner, D.L. 1994. Ninety-seven million years of angiosperm-insect association: paleobiological insights into the meaning of coevolution. Proc. Natl Acad. Sci. U S A 91: 1227812282.
  • Labandeira, C.C., Johnson, K.R. & Lang, P. 2002. A preliminary assessment of insect herbivory across the Cretaceous/Tertiary boundary: major extinction and minimal rebound. In: The Hell Creek Formation and the Cretaceous-Tertiary Boundary in the Northern Great Plains – An Integrated Continental Record at the End of the Cretaceous (J. H.Hartman, K. R.Johnson & D. J.Nichols, eds), Geol. Soc. Am. Spec. Pap. 361: 297327.
  • Larget, B. & Simon, D.L. 1999. Markov Chain Monte Carlo algorithms for the Bayesian analysis of phylogenetic trees. Mol. Biol. Evol. 16: 750759.
  • Loader, C. 1999. Local Regression and Likelihood. Springer Mathematics, New York, USA.
  • Lopez-Vaamonde, C., Rasplus, J.Y., Weiblen, G.D. & Cook, J.M. 2001. Molecular phylogenies of fig waSPS: partial co-cladogenesis between pollinators and parasites. Mol. Phylogenet. Evol. 21: 5571.
  • Lopez-Vaamonde, C., Godfray, C. & Cook, J.M. 2003. Evolutionary dynamics of host-plant use in a genus of leaf mining moth. Evolution 57: 18041821.
  • Lopez-Vaamonde, C., Godfray, C., West, S.A., Hansson, C. & Cook, J.M. 2005. The evolution of host use and unusual reproductive strategies in Achrysocharoides parasitoid waSPS. J. Evol. Biol. 18: 10291041.
  • Magallón, S. 2004. Dating lineages: molecular and paleontological approaches to the temporal framework of clades. Int. J. Plant Sci. 165: 721.
  • Magallón, S. & Sanderson, M. 2001. Absolute diversification rates in angiosperm clades. Evolution 55: 17621780.
  • Manchester, S.R. 1989. Systematics and fossil history of the Ulmaceae. In: Evolution, Systematics, and Fossil History of the Hamamelidae, Vol. II: ‘Higher’ Hamamelidae (P. R.Crane & S.Blackmore, eds), pp. 221252. Clarendon Press, Oxford, UK.
  • Manchester, S.R. 1994. Fruits and seeds of the middle Eocene Nut Beds Flora, Clarno Formation, Oregon. Palaeontol. Am. 58: 1205.
  • Manchester, S.R. 1999. Biogeographical relationships of North American Tertiary floras. Ann. Mo. Bot. Gard. 86: 472502.
  • Manchester, S.R., Dilcher, D.L. & Tidwell, W.D. 1986. Interconnected reproductive structures and vegetative remains of Populus (Salicaceae) from the middle Eocene Green River Formation, northeastern Utah. Am. J. Bot. 73: 156160.
  • Marshall, C.R. 1997. Confidence intervals on stratigraphic ranges with nonrandom distributions of fossil horizons. Paleobiology 23: 165173.
  • Meyer, H.W. & Manchester, S.R. 1997. The Oligocene Bridge Creek flora of the John Day Formation, Oregon. Univ. Calif. Public Geol. Sci. 141: 1195.
  • Mitter, C. & Farrell, B.D. 1991. Macroevolutionary aspects of insect-plant relationships. In: Insect-Plant Interactions (E.Bernays, ed.), pp. 3578. CRC Press, Boca Raton, Florida, USA.
  • Percy, D.M., Page, R.D.M. & Cronk, Q.C.B. 2004. Plant-insect interactions: double-dating associated insect and plant lineages reveals asynchronous radiations. Syst. Biol. 53: 120127.
  • Pigg, K.B., Manchester, S.R. & Wehr, W.C. 2003. Corylus, Carpinus, and Palaeocarpinus (Betulaceae) from the middle Eocene Klondike Mountain and Allenby formations of northwestern North America. Int. J. Plant Sci. 164: 807822.
  • Posada, D. & Crandall, K.A. 1998. MODELTEST: testing the model of DNA substitution. Bioinformatics 14: 817818.
  • Powell, J.A. 1980. Evolution of larval food preferences in microlepidoptera. Annu. Rev. Entomol. 25: 133159.
  • Powell, J.A., Mitter, C. & Farrell, B. 1999. Evolution of larval food preferences in Lepidoptera. In: Lepidoptera, Moths and Butterflies, Vol. 1: Evolution, Systematics, and Biogeography (N. P.Kristensen, ed.), Handbuch der Zoologie, Band IV (Arthropoda: Insecta), Teilband 35, pp. 403422. Walter de Gruyter, Berlin, Germany.
  • Ren, D. 1995. Insects. In: Faunae and Stratigraphy of Jurassic-Cretaceous in Beijing and Adjacent Areas (D.Ren, L.Lu, Z.Guo & S.Ji, eds), pp. 54197. Geological Publishing House, Beijing, China.
  • Rodriguez-Trelles, F., Tarrio, R. & Ayala, F.J. 2002. A methodological bias toward overestimation of molecular evolutionary time scales. Proc. Natl Acad. Sci. U S A 99: 81128115.
  • Romero, E.J. 1986. Fossil evidence regarding the evolution of Nothofagus Blume. Ann. Mo. Bot. Gard. 73: 276283.
  • Ronquist, F. 2004. Bayesian inference of character evolution. Trends Ecol. Evol. 19: 475481.
  • Ronquist, F. & Liljeblad, J. 2001. Evolution of the gall-wasp-host plant association. Evolution 55: 25032522.
  • Sanderson, M.J. 1997. A nonparametric approach to estimating divergence times in the absence of rate constancy. Mol. Biol. Evol. 14: 12181231.
  • Sanderson, M.J. 2002. Estimating absolute rates of molecular evolution and divergence times: a penalized likelihood approach. Mol. Biol. Evol. 19: 101109.
  • Sanderson, M. 2003. r8s: inferring absolute rates of molecular evolution and divergence times in the absence of a molecular clock. Bioinformatics 19: 301302.
  • Schoonhoven, L.M., Jermy, T. & Van Loon, J.J.A. 1998. Insect-Plant Biology: From Physiology to Evolution. 409 p. Chapman and Hall, London, UK.
  • Sequeira, A.S. & Farrell, B.D. 2001. Evolutionary origins of Gondwanan interactions: how old are Araucaria beetle herbivores? Biol. J. Linn. Soc. 74: 459474.
  • Sims, H., Herendeen, P.S. & Crane, P.R. 1998. New genus of fossil Fagaceae from the Santonian (late Cretaceous) of central Georgia, USA. Int. J. Plant Sci. 159: 391404.
  • Soltis, P.S., Soltis, D.E., Savolainen, V., Crane, P.R. & Barraclough, T.G. 2002. Rate heterogeneity among lineages of tracheophytes: integration of molecular and fossil data and evidence of molecular living fossils. Proc. Natl Acad. Sci. U S A 99: 44304435.
  • Strathmann, R.R. & Slatkin, M. 1983. The improbability of animal phyla with few species. Paleobiology 9: 97106.
  • Strong, D.R., Lawton, J.H. & Southwood, T.R.E. 1984. Insects on Plants: Community Patterns and Mechanisms. 313 p. Harvard University Press, Cambridge.
  • Stuessey, T.F. 2004. A transitional-combinational theory for the origin of angiosperms. Taxon 53: 316.
  • Sun, G., Zheng, S., Dilcher, D.L., Wang, Y. & Mei, S. 2001. Early Angiosperms and their Associated Plants from Western Liaoning, China. 227 p. Shanghai Scientific and Technological Education Publishing House, Shanghai, China.
  • Thorne, J.L., Kishino, H. & Painter, I.S. 1998. Estimating the rate of evolution of the rate of molecular evolution. Mol. Biol. Evol. 15: 16471657.
  • Wang, H.-S. 2002. Diversity of angiosperm leaf megafossils from the Dakota Formation (Cenomanian, Cretaceous), north western interior, USA. 395 p. PhD thesis, University of Florida, FL, USA.
  • Weiblen, G.D. & Bush, G.L. 2002. Speciation in fig pollinators and parasites. Mol. Ecol. 11: 15731578.
  • Welch, J.J. & Bromham, L. 2005. Molecular dating when rates vary. Trends Ecol. Evol. 20: 320327.
  • Wikström, N., Savolainen, V. & Chase, M.W. 2003. Angiosperm divergence times: congruence and incongruence between fossils and sequence divergence estimates. In: Telling the Evolutionary Time: Molecular Clocks and the Fossil Record (P. C. J.Donoghue & M. P.Smith, eds), pp. 142165. CRC Press, Boca Raton, USA.
  • Wilde, V. & Frankenhäuser, H. 1998. The middle Eocene plant taphocoenosis from Eckfeld (Eifel, Germany). Rev. Palaeobot. Palynol. 101: 728.
  • Wilf, P., Labandeira, C.C., Kress, W.J., Staines, C.L., Windsor, D.M., Allen, A.L. & Johnson, K.R. 2000. Timing the radiations of leaf beetles: hispines on gingers from latest Cretaceous to recent. Science 289: 291294.
  • Wilf, P., Labandeira, C.C., Johnson, K.R., Coley, P.D. & Cutter, A.D. 2001. Insect herbivory, plant defense, and early Cenozoic climate change. Proc. Natl Acad. Sci. U S A 98: 62216226.
  • Wilson, M.V.H. 1983. Is there a characteristic rate of radiation for insects? Paleobiology 9: 7985.
  • Yoder, A.D. & Yang, Z. 2004. Divergence dates for Malgasy lemurs estimated from multiple gene loci: geological and evolutionary context. Mol. Ecol. 13: 757773.
  • Zherikhin, V.V., Mostovski, M.B., Vršanský, P., Blagoderov, V.A. & Lukashebvich, E.D. 1999. The unique lower Cretaceous locality Baissa and other contemporaneous fossil insect sites in north and west Transbaikalia. In: Proceedings of the First Palaeoentomological Conference, Moscow (1998) (P.Vršanský, ed.), pp. 185191. AMBA Projects, Bratislava, Slovakia.

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Figure S1. Distributions of age estimates. Age = mode of the distribution. It represents the most likely value fo the divergence time of a particular node. LHPD = 90% Lower Highest Posterior Density limit. UHPD = 90% Upper Highest Posterior Density limit. Nodes of interest are numbered from 1 to 15 and indicated on the cladogram (see figure 3 a, b). Table S1.Probability (as averaged over the 100 randomly selected trees from the posterior distribution) that a particular plant group is reconstructed for our node of interest and not reconstructed for the node below. Table S2.Body and trace-fossil record of the Gracillarioidea (Lepidoptera: Ditrysia, Gracillariidae and Bucculatricidae). Appendix S1.Specimens used in this study. They have been ordered according to the plant classification (APG, 1998).

FilenameFormatSizeDescription
supplementarydata.pdf2235KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.