Rampant host and defensive ... - Wiley Online Library

doi: 10.1111/j.1420-9101.2012.02576.x

Rampant host- and defensive phenotype-associated diversification in a goldenrod gall midge J. O. STIREMAN III*, H. DEVLIN* & P. ABBOT† *Department of Biological Sciences, Wright State University, Dayton, OH, USA †Department of Biological Sciences, Vanderbilt University, Nashville, TN, USA

Keywords:

Abstract

adaptive radiation; Asteromyia; Cecidomyiidae; ecological speciation; gall morphology; Host-associated differentiation; multitrophic interactions; Solidago.

Natural selection can play an important role in the genetic divergence of populations and their subsequent speciation. Such adaptive diversification, or ecological speciation, might underlie the enormous diversity of plant-feeding insects that frequently experience strong selection pressures associated with host plant use as well as from natural enemies. This view is supported by increasing documentation of host-associated (genetic) differentiation in populations of plant-feeding insects using alternate hosts. Here, we examine evolutionary diversification in a single nominal taxon, the gall midge Asteromyia carbonifera (O.S.), with respect to host plant use and gall phenotype. Because galls can be viewed as extended defensive phenotypes of the midges, gall morphology is likely to be a reflection of selective pressures by enemies. Using phylogenetic and comparative analyses of mtDNA and nuclear sequence data, we find evidence that A. carbonifera populations are rapidly diversifying along host plant and gall morphological lines. At a broad scale, geography explains surprisingly little genetic variation, and there is little evidence of strict co-cladogenesis with their Solidago hosts. Gall morphology is relatively labile, distinct gall morphs have evolved repeatedly and colonized multiple hosts, and multiple genetically and morphologically distinct morphs frequently coexist on a single host plant species. These results suggest that Asteromyia carbonifera is in the midst of an adaptive radiation driven by multitrophic selective pressures. Similar complex community pressures are likely to play a role in the diversification of other herbivorous insect groups.

Introduction Due to their remarkable diversity, insects have long attracted attention from evolutionary biologists seeking to understand the underlying causes of speciation (Dobzhansky, 1937; Ehrlich & Raven, 1964; Mitter et al., 1988). In recent years, there has been increasing attention paid to the roles of ecological interactions and natural selection (ecological speciation; Rundle & Nosil, 2005; Funk et al., 2002, 2006; Schemske, 2010). This has been the case in particular for insects that feed on plants because their hosts often differ in traits that are Correspondence: John O. Stireman, Department of Biological Sciences, Wright State University, Dayton, OH 45435, USA. Tel.: +1 937 775 3192; fax: +1 937 775 3320; e-mail: [email protected]

linked ecologically and physiologically to performance (e.g. nutritional quality, chemical or physical defences, phenology, recognition cues), resulting in fitness trade-offs and divergent selection between plants when host shifts occur (e.g. see Berlocher & Feder, 2002; Matsubayashi et al., 2010). When coupled with a tendency to mate on or near the host plant, ecological speciation may be a common outcome of host shifts or expansions in herbivorous insects, even in the absence of geographical isolation or morphological character variation (Coyne & Orr, 2004; Rundle & Nosil, 2005). The importance, but not necessarily the generality, of ecological speciation in phytophagous insect diversification is supported by studies demonstrating host-associated genetic differentiation in insect populations (e.g. Feder et al., 1988; Funk, 1998; Emelianov et al., 2003). Most studies of host-associated genetic

ª 2012 THE AUTHORS. J. EVOL. BIOL. 25 (2012) 1991–2004 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2012 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY

1991

1992

J. O. STIREMAN ET AL.

differentiation in insects have tended to focus on closely related herbivore populations occupying pairs of host plant species (although see e.g. Funk, 1998; Berlocher, 2000). Although useful in understanding the process of ecological speciation, it is difficult to extrapolate from these pairs in order to judge the overall prevalence of host-associated differentiation in insect lineages. It may be that most cases studied to date are exceptional examples and host-associated diversification is uncommon, although this seems unlikely (e.g. Stireman et al., 2005). The extent to which widespread, morphologically cryptic insect species conceal undiagnosed differentiation and the ecological mechanisms mediating patterns of host use remain important areas of inquiry (Thompson, 2008). Most studies of host-associated differentiation have not taken tri- or multitrophic interactions between herbivores, their enemies and their host plants into account, even though these interactions can foster host shifts and potentially exert strong divergent selection across host plants (Jeffries & Lawton, 1984; Lill et al., 2002; Singer & Stireman, 2005; Heard et al., 2006; Nosil & Crespi, 2006; Craig & Itami, 2011). There is a general need for studies of patterns of divergence in broadly distributed herbivorous insect species and the selective pressures leading to host-associated differentiation (Bailey et al., 2009). In this study, we analyse diversification within a nominal species of North American gall midge within the genus Asteromyia. At the generic level, Asteromyia exhibits strong host plant–related phylogenetic structure, suggesting the likelihood of cryptic, genetically differentiated populations within species (Stireman et al., 2010). The focus of the present work, Asteromyia carbonifera, attacks a diverse array of Solidago species in North America. On one host plant, Solidago altissima, A. carbonifera appears to be undergoing a ‘radiation within a radiation’ and expresses genetically distinct gall morphotypes (Stireman et al., 2008, 2010). The polymorphism in gall morphology is visually conspicuous and has attracted the attention of various researchers over the past few decades (Weis, 1982a,b; Crego et al., 1990). One hypothesis has been that A. carbonifera galls confer protection from parasitoids, and the variation in gall shape is related to selective variation in parasitoid attack (as in Craig et al., 2007). The defensive benefit of A. carbonifera galls has been supported by experimental study (Weis, 1982b), but whether gall morphs have been selectively shaped by parasitoids remains to be determined, because patterns of genetic and gall variation in A. carbonifera across goldenrods have not yet been described systematically. Potentially, A. carbonifera harbours undescribed cryptic variation in host-associated genetic structure and gall polymorphism. If so, A. carbonifera may be well suited for asking how plants and enemies shape ecological speciation. First, we ask whether genetic structure in A. carbonifera is associated with host plant use and evaluate

evidence for co-diversification with host plant lineages. Strong genetic structuring along host plant lines would constitute positive evidence that midge–plant interactions drive diversification. Alternatively, strong phylogeographical structure would leave open the possibility of nonadaptive modes of diversification. Second, we evaluate genetic and phenotypic differentiation with respect to gall morphology beyond that established on the host plant Solidago altissima. We do not focus on current variation in parasitism to understand its role in shaping divergence of midge populations, but rather on the mark it has left on the midge’s extended phenotype. We hypothesized that there has been a form of character displacement in A. carbonifera gall phenotypes in response to enemy-mediated divergent selection. We thus expected to find gall phenotype–related genetic structure in this species complex that is at least partly independent of host plant, and we expect to see phylogenetic evidence of selection on gall form. Our results reveal an unusual degree of cryptic variation within a single, morphologically described species. Asteromyia carbonifera has undergone a recent, rapid evolutionary radiation in host plant use and defensive phenotypes, with little evidence of large-scale geographical isolation. While studies are needed to account for alternatives, these patterns are consistent with selection associated with host plant use and escape from parasitoid enemies. Similar tritrophic selection pressures may help to explain the extensive radiation of other gall-forming lineages as well as other specialized insect groups. Experimental system Asteromyia carbonifera is a North American gall midge (Diptera: Cecidomyiidae) that forms characteristic ‘blister’ galls on the leaves of its hosts in the genus Solidago (Asteraceae) (Fig. 1, also see Stireman et al., 2008, 2010). An unusual twist in the A. carbonifera story is that the galls themselves are actually composed of the hyphae of a fungal mutualist rather than host plantderived tissue. Thus, unlike other gall-inducing insects, the diversity in functionally important gall phenotypes in Asteromyia applies to fungi, not to plants. Larvae of A. carbonifera apparently feed on this mutualistic fungus, Botryosphaeria dothidea (Ascomycota: Dothideomycetes; Janson et al., 2009, 2010; Heath & Stireman, 2010), which may provide a buffer from variation in nutrition or defences among host plants. Females actively transport fungal conidia in specialized abdominal pockets (mycangia) and deposit them with their eggs on the underside of leaves (Borkent & Bisset, 1985; Heath & Stireman, 2010). From one to several generations are completed each year, depending upon the population, and mature larvae overwinter in galls, eclosing the following spring/early summer (Gagne´, 1989). Fungal proliferation and consequently, gall morphology, is likely controlled by larval secretions (Janson et al., 2010).


Rampant gall midge diversification

79/24

altissima

NY NY OH OH

100/99

98/72 ONT ONT 72/49 NY 100/89 ONT 54/-OH 84/86 OH 100/99 OH 62/--

55/33

ME IL NY MO NY

100/94

MA OK

96/93 49/63

100/98

IL PA OH

99/94

90/-66/--

73/61

flexicaulis

NY CO

CO 100/63

caesia

lepida MO

rugosa

altissima

IL NY

NY KY ME

ME

gigantea

100/98 98/78

IL

NY

rugosa

ME ME MO

OH

altissima nemoralis

OH NY NE NE OK OK OK 40/-OK 51/86 NY NY 55/-NY NY OH 51/37 NY NE 100/70 NE 100/97 WV WV 100/97 WV ME 100/81 MA MA MA IA OH KY 70/60 NY NY 100/100 NY NY NY

gigantea patula rugosa gigantea altissima rugosa 98/73 100/67

OH KY

MO OH OH OH NY MA 100/90 OH 74/62 NY NY KY KY 67/44 92/54

altissima

97/59

100/67

98/80 100/100

ME

ME 100/98 ME 98/99 MA

rugosa NY NY NY

NY

altissima

MA ME MA 88/72 MA 100/99 ME

100/99

89 41

100/100

40/--

100/100 100/99 50/80

ME ME

100/91 65/99

50/78 50/72

95/62

MA MA NC 96/91 NC IA IA IA IA IA IA

100/100

OH

OH OH OH

NY

NY NY NY NY NY

delicatula

ulmifolia velutina (californica)

sempervirens

petiolaris

ulmifolia nemoralis sempervirens

AR

OH

100/50

bicolor (or hispida) OH OH CT

NE

NE NY NE NE NE NE NE NE 100/86

NE NE

NY

juncea

missouriensis

NE NE NE NE

NE NE NE

98/62

94/35

AR AR AR AR

rugosa OK OK OK AR

rigida

61/48

100/100

shortii uliginosa gigantea

houghtonii

OK AR AR AR AR AR AR ME NY ME

100/99

Fig. 1 Bayesian consensus tree of Asteromyia carbonifera populations based on concatenated mtDNA and wg sequences. Geographical origin (state/province) of samples species are indicated by abbreviations. Host plants are indicated by branch colour and names to the right. A selection of gall forms is illustrated with wedges indicating to which taxa they belong. Support for branches is indicated by posterior probabilities/ML bootstraps above or below branches. Outgroup taxa (A. modesta and A. laeviana) are omitted in this figure.

100

AR AR AR

100

CA CA

42/--

50/100

1993

NY

NY

juncea

odora NY NY IL NM NM NM OH IL IL IL

94/55

100/48

100/84

54/44

55/33

UT UT UT UT UT UT CO NM NM NM CT NM 100/91

100/94

100/70 100/57

MA 100/100

UT

NC NC

juncea velutina

lepida speciosa velutina juncea sempervirens

100/100

OK

VA VA

Unknown sp. nemoralis

0.5

Of the eight recognized species of Asteromyia (Gagne´, 1968), A. carbonifera is of particular interest due to its abundance, broad distribution (across most of North

America) and association with the diverse plant genus Solidago (Gagne´, 1968). At least 65 species of Solidago are recorded as hosts of A. carbonifera (Gagne´, 1968;


1994


Appendix S1). The genus Solidago itself exhibits profound differentiation in life history traits, growth form and distribution and along environmental gradients (Semple et al., 1999; Semple & Cook, 2006). Like other gall midges, Asteromyia is attacked by a diverse array of natural enemies, particularly hymenopteran parasitoids (Weis, 1982b). At least seven hymenopteran parasitoids are known from A. carbonifera galls. As in many galling taxa, parasitism rates can be high (approaching 80%), suggesting that parasitoids exert significant selective pressure on Asteromyia behaviour and life history. Observations that gall tissue protects larval A. carbonifera from parasitoids (Weis, 1982b), that parasitism levels vary significantly among co-occurring gall morphs on S. altissima (Stireman et al., 2008) and that these gall morphs differ in their susceptibility to different parasitoid species (J. O. Stireman, J. J. Heath & B. L. Wells, unpublished) all strongly suggest that divergence in gall morphology is related to selection by parasitoids, as has been reported in other galling systems (e.g. Cooper & Rieske, 2010).

Materials and methods Collection of Asteromyia samples Galls of A. carbonifera were collected from 24 taxa of Solidago across much of the United States. Collections were made by JOS and colleagues from 2006 to 2009, with focus on the US Midwest. Additional samples, primarily from the northeast United States, were acquired from T.G. Carr who collected them from 1999 to 2004. When a patch of host plants bearing Asteromyia galls was discovered, 5–50 galls were haphazardly collected. Where possible, we attempted to spread the sampling across available individual plants to maximize diversity. Sampling locations were recorded, and ramets of the host plant were collected and pressed for later identification (see online Table S1 for a list of all Asteromyia samples used in this study, their hosts and their collection location). Plants sampled by Carr were identified with reference to specimens in the herbarium of the L.H. Bailey Hortorium of Cornell University and United States National Museum. Plants sampled by JOS were primarily identified by J. Semple (U. of Waterloo), although some specimens were identified using available keys (Semple et al., 1999; Semple & Cook, 2006). Vouchers of host plants collected by JOS are currently housed in Wright State University’s herbarium collection. Galls were placed in individual plastic bags and stored in an ice-filled cooler or refrigerator (4 °C) until they could be dissected under a microscope. For each gall, host plant, parasitism status, number of larvae and developmental stage were recorded. Immature midges dissected from these galls were stored in a ( 20 °C)

freezer in 95% ethanol until DNA was extracted. Representatives of preserved midges were selected for DNA extraction to maximize geographical and host plant diversity. Galls that were not mature and contained only early larval stages were excluded from estimates of parasitism rates and measurements of gall phenotype. Morphological data A subset of collected galls were photographed or scanned before they were dissected. These images were used to take measurements of gall morphology including gall area, perimeter, circularity (4p (area/perimeter)2), aspect ratio (major axis/minor axis), solidity (area/convex area) and mean grey value. All measurements were taken using ImageJ (Rasband, 1997–2011). In addition, the number of locules or ‘cells’ per gall was recorded. When multiple gall forms appeared to occur on a single host, galls were assigned to rough classes based on their appearance (e.g. ‘pimple’, ‘cushion’, ‘large’, ‘small’). Population gall measurements are summarized in online Table S2. Plots of galls on the first two axes of PCAs were used to assess whether these apparently different morphs were distinguishable based on the trait measurements taken above. Separation of groups of morphs identified by visual inspection of PCA plots was confirmed with MANOVAs of principal components 1–3 and post hoc pairwise comparisons (Tukey’s tests) of ANOVAs for the first principal component. Means of all traits were calculated for each recognized host plant–/gall morph–associated population for use in subsequent analyses. Molecular methods DNA were extracted from Asteromyia immatures using Puregene DNA Purification System, Cell and Tissue Kit (Gentra Systems Inc., Minneapolis, MN). Standard polymerase chain reactions (PCRs) were conducted following Stireman et al. (2008). Primers Barb1 and S1718 (Simon et al., 1994) were used to amplify an approximately 1800-bp fragment spanning much of COI and COII, as well as the intervening leucine tRNA gene. An approximately 700-bp fragment of the gene wingless (wg) was sequenced with primers 5′wg1 and 3′wg2 (Ober, 2002). PCRs were purified using an EXO/SAP enzymatic cleanup or Millipore PCR purification blocks (University of Arizona Genetics Core (UAGC)) (Dugan et al., 2002). Sequences were run on a 3730 DNA Analyzer from Applied Biosystems, Inc., at the UAGC. Both primers (forward and reverse) were used for sequencing each gene. Sequences were manually edited using CodonCode Aligner. Alignment was performed using ClustalW with further manual alignment in MEGA 5.05 (Tamura et al., 2011). The resulting alignment lengths, after truncation of ambiguous ends, were 1734 bp for the mtDNA and 537 bp for wg. A total of



195 and 96 A. carbonifera samples were successfully sequenced for mtDNA (also see Stireman et al., 2010) and wg, respectively. In addition, three samples of the related species A. modesta from different Solidago hosts and one from A. laeviana were used as outgroup taxa. Phylogenetic analysis Maximum-likelihood (ML) phylogenetic analysis of the concatenated data set was conducted using MEGA 5.05 (Tamura et al., 2011). For this analysis, a general time-reversible (GTR) model of substitution with gamma distributed rates (Γ) and assuming a proportion of invariant sites (Pinv) was selected as the best model of sequence evolution based on AIC scores in MEGA. Analysis began with an initial neighbour-joining tree (BIONJ) using the close neighbour interchange (CNI) method. Branch support was evaluated with 500 bootstrap samples of the data set (with nucleotide models as above). Bayesian phylogenetic analysis was conducted using MrBayes version 3.1 (Ronquist & Huelsenbeck, 2003). The data set was partitioned by gene (COI, tRNA, COII, wg), and a GTR evolutionary model was used with Γ and Pinv estimated for each partition. All likelihood priors were set to default values. Two MCMC runs were conducted simultaneously with 4 chains for 2,000,000 generations, sampled every 1000 generations. Stationarity of tree likelihoods was evaluated using Tracer version 1.5 (Rambaut & Drummond, 2009). Based on this evaluation, burn-ins of 0.5 and 1.1 million trees were used for each run, respectively. The two MCMC runs resulted in very similar consensus trees, tree likelihoods and branch posterior probabilities, and were combined to create a single consensus tree. Many of the geographical and host plant populations in our data set were represented by multiple samples with little or no genetic divergence among them. To speed up and simplify subsequent phylogenetic and comparative analyses, this data set was trimmed to minimally include at least one representative from each host plant, at least one individual of the known gall morphs on S. altissima and any additional isolated lineages with a distance of at least 0.01 from neighbouring clades. This resulted in a ‘skeletal’ data set of 38 OTUs for which sequences for both amplicons were available. Subsequent comparative analyses employed this reduced data set, with some additional trimming of taxa if data were unavailable for particular analyses. Maximum-likelihood and Bayesian phylogenetic analyses (as above but with 1 million generations, and burn-ins of 250,000) of the two gene regions separately indicated that, although most higher-level clusters of OTUs were similar, the deeper relationships inferred among these groups varied between gene trees. We therefore used Bayesian estimated consensus gene trees to estimate the species tree using the maximum tree

1995

(MT) method of Liu et al. (2010) in the R package ape version 2.7-3 (Paradis et al., 2004), a maximum-likelihood method, and using average coalescence times in Phybase (i.e. ‘coal.sptree’; Liu et al., 2009; Liu & Yu, 2010). Penalized likelihood (Sanderson, 2002) was used to obtain clock-like trees for analysis in ape with k smoothing parameters estimated by cross-validation. Host-associated genetic structure We used AMOVA in the R package pegas (Paradis, 2010) and GenAlEx (Peakall & Smouse, 2006) to assess the degree of genetic structure in A. carbonifera attributable to host plant and geography. Genetic distances were calculated using a TN93 model of molecular evolution (Tamura & Nei, 1993). Host plants with fewer than two A. carbonifera samples were excluded from analysis. Potential geographical effects on genetic structure were examined at multiple scales of population lumping. The finest level categorized samples by state (although neighbouring states with few samples were lumped, e.g. CT and MA), the intermediate level categorized populations by geographical subregions (north-east, mid-Atlantic, eastern mid-west, mid-west, plains west, mountain west, Appalachian, south-central and southeast), and the coarsest level categorized populations by major geographical regions (NE, EM, SE, MW, W). Each factor, host plant and geography, was examined separately at the finest geographical scale (state) and at broader regional scales using nested models, with both host nested within region and region nested within host plant. We also examined the effect of removing host plants that were only present in single geographical region, as geography and host plant are confounded in these samples. Analyses were conducted for each gene region separately as well as for the concatenated data set. Significance was evaluated with 999 permutations of host and geographical associations. Host–parasite coevolution To test for the possibility of co-phylogenesis of A. carbonifera lineages and Solidago species, we employed the reduced skeletal data set minimally including representative taxa from each Solidago host species as well as any additional genetically distinct host-associated A. carbonifera lineages. A pairwise (TN93 + Γ) distance matrix was computed from this data set. Sequence for an approximately 820 bp of ribosomal DNA spanning 18S, ITS1, 5.8S, ITS2 and 26S was obtained for 19 Solidago species via NCBI GenBank (largely from Laureto & Barkman, 2011). Solidago fistulosa sequences were included in place of the unavailable species S. delicatula, as they are members of the same subsection and are thought to be closely related (Semple & Cook, 2006). For S. shortii and S. fistulosa, only a smaller, 653-bp sequence spanning ITS1, 5.8S and ITS2 was available.


1996


The phylogenetic concordance of hosts (Solidago species) and parasites (A. carbonifera populations) was analysed using parafit (Legendre et al., 2002) implemented in the R package ape. This method compares principal coordinates of phylogenetic distance matrices with respect to a host–parasite link matrix to assess independence of host and parasite phylogeny. Significance of the global test of evolutionary independence as well as that of particular midge–plant links (see Legendre et al., 2002) was evaluated with 999 permutations of the associations. We used parsimony and ML ancestral state reconstruction modules in Mesquite version 2.74 (Maddison & Maddison, 2009) to examine evolutionary patterns of host plant use in A. carbonifera, assuming unordered transitions. We also evaluated phylogenetic autocorrelation in host plant use by assigning each host a number and calculating Moran’s I and Abouheif’s test (999 permutations) in the R package adephylo (Jombart et al., 2010; see Gittleman & Kot, 1990; Abouheif, 1999). Gall morphology To analyse the evolution of gall traits, we calculated mean and median values for each gall trait for each population (host-associated lineage or morph) for which we had morphological data (24 taxa). This included populations from each host plant as well as multiple morphotypes within host plants that were identified with the ordinations described previously. We conducted a PCA ordination of these mean values of gall traits in the R package vegan (Oksanen et al., 2010) and saved the PCA scores for the first three axes. We examined the morphological distribution of all gall morphs by plotting gall morph positions on the first two principal components. The evolution of gall traits and PCA axes was examined in Geiger (Harmon et al., 2008) by assessing the fit of several evolutionary models to the data including Brownian, Lambda, Kappa, Delta (see Pagel, 1999), Ornstein-Uhlenbeck (‘O-U’; see Butler & King, 2004) and Early Burst (‘EB’; Harmon et al., 2008), by calculating their likelihoods and AIC scores and by estimating phylogenetic signal via Blomberg’s K (Blomberg et al., 2003). A K value of one is expected under a Brownian motion model of evolution along branches, values > 1 indicate strong phylogenetic signal and trait conservatism, and values < 1 indicate trait value divergence among related taxa (or convergence among unrelated taxa; Blomberg et al., 2003). Significance of K was determined by comparing the observed variance in phylogenetic contrasts to that generated with 999 tip shufflings of the phylogeny in picante (Kembel et al., 2010). These analyses used a reduced taxon and maximum species tree (MT) based on the two gene trees. Finally, for comparison with host plant use, we examined phylogenetic autocorrelation in gall morph type (coded numerically) with Moran’s I and Abouheif’s test.

Results Phylogenetic structure of Asteromyia carbonifera Maximum-likelihood ( LnL = 21917.8) and Bayesian analyses ( LnL = 10477.72) of the concatenated mtDNA and wg data sets resulted in similar trees for A. carbonifera with comparable support for branches (Bayesian results are shown in Fig. 1). Monophyly of the A. carbonifera clade on Solidago was moderately supported (ML bootstrap, 97%, posterior probability, 50%) with respect to A. modesta and A. laeviana outgroups. Two aberrant sequences from collection of the host S. nemoralis are highly divergent from other A. carbonifera and may represent mistaken host identifications due to absence of flowers. This placement is driven by the mtDNA data, as wg sequences place at least one of these samples with other S. nemoralis, well nested within A. carbonifera (see Fig. S1). We obtained midge samples for genetic analysis from more than a third of the Solidago species known to host Asteromyia carbonifera (24 of 65; Appendix S1). The resulting phylogenetic reconstructions indicate strong, widespread genetic structuring related to host plant use and reveal the existence of sometimes distantly related ‘gall morph’ clades coexisting on the same host plant species (Fig. 1). Genetic divergence among populations ranges up to approximately 6% (uncorrected distance), indicating relatively deep splits. In analyses of each gene region separately, both data sets show similar strong host-associated phylogenetic structure, but inferred relationships between host- and morph-associated clades vary between them (Figs S1, S2). Using the reduced 38 taxon set (see Materials and Methods), both methods of species tree estimation, MT and coal.sptree using average coalescence times, resulted in qualitatively similar phylogenies (Fig. 2). The maximum tree exhibited branch lengths more consistent with the individual gene trees and was used in subsequent evolutionary analyses of traits. AMOVAs confirm the visual impression of host- and gall-associated differentiation. There is significant hostassociated genetic structure and relatively weak geographical structure regardless of the scale used (Table 1). When host was nested within region, geography failed to explain any significant variation, whereas region explained up to 18% of genetic variation when nested within host. This indicates that host is the superior grouping factor within which geography is nested. Across all analyses, host explained more than twice as much variation as geography, with estimates ranging up to 68% of total genetic variation (Table 1). This result holds even with a reduced data set containing only widespread host species and excluding distinct hostassociated populations with geographically limited collections. As is expected, geographical structure was more apparent at larger geographical scales than at



nemoralis OH altissima IL irr. altissima IL flat rugosa WV gigantea OK altissima NY cush. lepida CO cush patula OH caesia OH pimp. altissima OH flat flexicaulis NY caesia NY flat rugosa NY sempervirens NC flat nemoralis ME nemoralis VA sempervirens ME houghtonii NY shortii KY uliginosa NY altissima OH flat rugosa MA cres. ulmifolia AR cush. ulmifolia AR cush. ulmifolia AR fl delicatula OK rigida IA bicolor/hispida AR velutina NM irreg. juncea IL cush. lepida CO speciosa NM sempervirens NC pimp. odora AR missouriensis NE juncea NY flat juncea NY flat

1997

smaller ones. Interestingly, much more genetic structure in the mtDNA data was explained by geography than in the nuclear data set when nested with host (Table 1), suggesting male-biased dispersal. In the latter case, geography was not a significant predictor of genetic structure in any analysis. Relationships between Solidago host species based on rDNA sequences were generally poorly resolved. Despite this, a significant overall association between genetic distances of hosts and midges was found (P = 0.021, 999 permutations). This association is largely driven by the use of S. juncea/missouriensis and S. sempervirens, by several related midge lineages (gall morphs). If these populations are considered as a single lineage, the significant association disappears. Neither parsimony nor ML ancestral state reconstruction methods provide a clear picture of the ancestral Solidago host of the A. carbonifera clade. Ancestral states at deeper nodes are equivocal with many Solidago species having similar likelihoods (Online Fig. S4). There is some suggestion from state likelihoods that the host plants S. altissima, S. juncea and possibly S. sempervirens have served as important sources for the colonization of other host taxa. However, each of these taxa appears to harbour multiple gall morphotypes, suggesting that they may have been colonized repeatedly rather than acting as a reservoir for colonization of other host taxa. Small, negative values of Moran’s I suggest little if any autocorrelation in host use (Moran’s I: 0.070, Abouheif’s test, P = 0.644), despite the overall significant association of midge and host genetic distances found with parafit.

0.2

Fig. 2 Maximum tree ‘species tree’ of mtDNA and nuclear gene trees for recognizable host- or gall morph–associated lineages within Asteromyia carbonifera with branches coloured by host as in Fig. 1.

Gall phenotype Principal components analysis confirmed the existence of distinct variation in key gall morphological traits

Table 1 Results of nested AMOVAs examining molecular variance (% Var) explained by host plant and geographical region and their significance (bold-faced) for each locus separately and combined. Results for models with both host nested within region and region nested within host are given. Only results for the largest geographical scale are shown; results were similar at finer geographical scales although geography tended to explain relatively less variance. ‘Within’ indicates variance explained within regional host populations (i.e. error). See Materials and Methods for details. Both genes

mtDNA

wingless

Model

Source

% Var

P

% Var

P

% Var

P

Region (host)

Region Host Within Host Region Within Host Region Within

1.6 68.2 30.1 53.8 16.6 29.6 37.7 17.7 44.6

< 0.484 < 0.001

0.6 69.2 30.2 53.2 17.2 29.6 36.9 18.4 44.7

< 0.492 < 0.001

7.8 63.8 28.4 67.3 4.5 28.2 66.7 5.8 27.6

< 0.501 < 0.001

Host (region)

Host (region)*

< 0.001 < 0.002 < 0.001 < 0.005

< 0.001 < 0.001 < 0.001 < 0.003

*Reduced data set only containing host plant samples found in > 1 geographical region. ª 2012 THE AUTHORS. J. EVOL. BIOL. 25 (2012) 1991–2004 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2012 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY

< 0.001 < 0.163 < 0.001 < 0.149

1998


that previous studies have hypothesized to be ecologically important (Weis, 1982a,b; Stireman et al., 2008). The first three PCA axes explained 82.5% of the variance in the gall traits measured (PC1: 37.9%, PC2: 30.2%, PC3: 14.4%; Table S3). The first axis has relative strong loadings of gall traits associated with shape (circularity, aspect ratio, solidity), whereas the second is dominated by size-related traits (area, perimeter and number of gall cells). Size and shape variation in gall morphology is thought to be related to enemy protection for many gall-forming insects (Craig et al., 2007; Bailey et al., 2009). The third axis is dominated by coloration (mean grey; Table S3). The ordination plot of gall morphs on the first and second PCA axes reveals that morphologically dissimilar galls frequently cooccur on the same host plant (Figs 3 and 4). Seven of the host plants we examined demonstrated clear evidence of harbouring at least two gall morphotypes

(Fig. 3, Table 2). Not all gall morphs are clearly separated on these axes, and these were lumped in subsequent analyses. In other cases, morphs were further differentiated by examination of PC3. Nevertheless, we have undoubtedly underestimated gall diversity in A. carbonifera. Several additional host plants showed suggestive evidence of multiple morphs (e.g. S. velutina), but lacked adequate sample size for statistical analyses. Other morphs are widely separated phylogenetically from each other on the same host (e.g. S. nemoralis, S. sempervirens), but were not included in the analysis due to the lack of images. Given that about two-thirds of Solidago hosts have not yet been surveyed for A. carbonifera, phenotypic diversity in gall morphology is likely far greater than the present results indicate. Interestingly, there are more midge lineages than distinct gall shapes, indicating both multiple colonization of host plants by single morphs and the possibility of

Fig. 3 Gall morphospace (PC1 vs. PC2) for six Solidago species in which multiple gall morphs were observed. Grey ovals indicate 95% standard errors around the centroid, and dashed lines indicate convex hulls for galls that were visually categorized as a particular morphotype. Not all presumed gall morphs were clearly separated on these axes. See Table 2 for statistical analyses. Gall morphs are indicated as follows: S. caesia: open circles = ‘black pimple’, asterisks = ‘flat’, closed circles = ‘small cushion’, S. rugosa: asterisks = ‘flat’, closed circles = ‘crescents’, S. sempervirens: open circles = ‘brown pimple’, asterisks = ‘flat’, closed circles = ‘small cushion’, S. ulmifolia: open circles = ‘irregular’, asterisks = ‘flat’, closed circles = ‘crescent’, S. juncea: open circles = ‘brown pimple’, asterisks = ‘cushion’, closed circles = ‘flat’, S. lepida: open circles = ‘brown pimple’, asterisks = ‘flat’. ª 2012 THE AUTHORS. J. EVOL. BIOL. 25 (2012) 1991–2004 JOURNAL OF EVOLUTIONARY BIOLOGY ª 2012 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY

1999


classified by appearance, the Moran’s I value of phylogenetic autocorrelation was marginally nonsignificant (I:0.069, P = 0.052) based on permutation tests, as was Abouheif’s test (P = 0.062), suggesting some degree of phylogenetic conservatism. Parasitism frequencies of sampled populations ranged from seven to 65%, with well-sampled populations (> 50 galls) typically experiencing 30–50% parasitism (Table S2). Considering that these values are underestimates due to removal of galls from further parasitism risk, they suggest that parasitoids are a potent selective force in this system. However, it is unclear how diagnostic these frequencies are as most estimates of parasitism across host and morph-related populations are based on few collections with limited sample sizes.

Discussion Fig. 4 Mean gall morphology scores for PCA axes 1 and 2 for populations of Asteromyia carbonifera. Labels indicate host plants followed by gall morphotype (if noted).

convergent evolution in gall morphology. The first two PCA axes exhibit significantly nonrandom phylogenetic signal, with relatively low K values of 0.313 (P < 0.002) and 0.233 (P < 0.007), respectively. The third axis was also characterized by a low, but nonsignificant K value (P < 0.12). The best fitting evolutionary model was O-U for PCA axes 1–3; however, in no cases was this fit significantly better than the model employing Pagel’s lambda, and for axes 1 and 2, only the Brownian motion model was a significantly poorer fit (Table 3). Finally, when gall morphs were simply

Uncovering patterns of cryptic host–associated variation in insect herbivores is important for assessing the generality of speciation by natural selection. Gall-forming insects are particularly interesting in this regard because their galls often represent both nutritional and defensive structures that reflect or mediate ecological interactions with host plants, mutualists and natural enemies (Bailey et al., 2009). Correspondence between cryptic genetic variation and ecologically significant traits suggests linkages between adaptive trait variation and population differentiation (Hajibabaei et al., 2006). Our results demonstrate that A. carbonifera represents an actively radiating species complex, diversifying across host plant lineages. We also show that gall morphology, a defensive phenotype, represents another axis of diversification. We find evidence of both divergent gall

Table 2 Results from MANOVA of morphology using PC1-3 as response variables, and pairwise post hoc Tukey’s tests of gall morph pairs for PC1 for plants hosting more than two gall morphs. Pairwise comparisons

MANOVA

Host plant

Wilk’s lambda

Solidago caesia

0.339

2,86

Solidago rugosa Solidago sempervirens

0.358 0.438

Solidago ulmifolia

d.f.

F

P

Morph1

Morph2

P

20.1

< 0.00001

flat sm. cushion sm. cushion

b. pimple b. pimple flat

< 0.00001 0.25738 < 0.00001

1,40 2,360

22.714 60.883

< 0.00001 < 0.00001

0.409

2,108

19.915

< 0.00001

Solidago juncea

0.329

2,118

28.712

< 0.00001

flat sm. cushion sm. cushion irregular flat irregular cushion flat cushion

b. pimple b. pimple flat crescent crescent flat b. pimple b. pimple flat

< 0.00001 0.00812 < 0.00001 0.4222 < 0.00005 < 0.00001 0.0264 < 0.00001 < 0.00001

Solidago lepida Solidago gigantea

0.202 0.788

1,16 2,98

18.391 4.0427

0.00004 < 0.00001


2000


Table 3 Relative AIC scores (dAIC) of alternative models of the evolution of gall morphology (PC1–PC3) in Asteromyia carbonifera based on the maximum species tree (MT; Fig. 2). The best-fitting AIC model(s) is indicated by grey shading. P values indicate whether each model is a significantly worse fit to the data relative to the best-fitting model (based on LnL). (PC1)

(PC2)

(PC3)

Gall Morphology Model

dAIC

P

dAIC

P

dAIC

P

Brownian OU Pagel’s k Pagel’s j Pagel’s d Early burst

2.4 0 2.0 1.0 1.8 0.12

0.03 NA 0.15 0.31 0.12 0.72

2.3 0 2.8 3.5 1.7 0.09

0.03 NA 0.10 0.06 0.19 0.76

11 0 0.16 7.5 9.4 4.7

< 0.001 NA 0.69 0.006 0.002 0.03

morphs within hosts and convergent gall morphs across host plant species. Given that A. carbonifera galls on Solidago are largely composed of a fungal phytopathogen, these results suggest that not only the physiological compatibility between herbivores and their hosts but also other community interactions have the potential to play important roles in facilitating niche shifts and phenotypic divergence (Singer & Stireman, 2005; Janson et al., 2008). In A. carbonifera, fungal mutualists mediate trophic interactions by conferring defensive extended phenotypes, and ultimately both natural enemies and mutualists either enable or drive the colonization and adaptation to novel ecological niches. The role of mutualists and enemies in shaping how insect herbivores both explore adaptive landscapes and specialize ecologically on host plants may help to explain broad patterns of host plant use and diversification in other phytophagous insects (Bernays & Graham, 1988; Singer & Stireman, 2005). Asteromyia carbonifera diversity Asteromyia carbonifera consists of a striking number (> 20) of distinct lineages, ranging from cryptic species to, possibly, host races still experiencing gene flow. Many discrete lineages are associated with specific species of host plants, suggesting a process of host-associated genetic divergence (e.g. Abrahamson & Weis, 1997; Stireman et al., 2005). In contrast, geography, even at the broadest scales, accounts for remarkably little genetic variation for an organism with such a short adult lifespan (a few days) and weak directed flight. Despite the strong structuring effect of host plant, A. carbonifera populations have not diverged in strict congruence with their hosts. Shifts between relatively distant hosts have occurred, often repeatedly, supporting the view that A. carbonifera likely radiated onto their Solidago hosts following diversification of the plants themselves (Stireman et al., 2010). Still, several host shifts have occurred between related taxa, indicating that if Solidago species were colonized after they diversified, they were not colonized randomly. Interestingly, the nuclear gene wingless shows similar,

strong host-related genetic structure relative to the mtDNA sequences despite its larger effective population size. It is likely that additional host- or morphrelated population structure exists that is not apparent from analyses of these two genes. AFLP data, for example, clearly distinguish gall morph populations on S. altissima that do not form monophyletic clades in either of the current gene trees (Stireman et al., 2008). Given the likelihood of further fine-scale population structure, it is possible that all of the populations we sampled are primarily or completely restricted to a single host plant species. Evolution of gall phenotypes The strong host-associated genetic structure in Asteromyia carbonifera represents one form of phenotypic divergence among lineages. Another axis of phenotypic divergence involves the gall structure itself. The presence of multiple phenotypically and genetically distinct gall forms on a single host plant, as well as similar gall forms on different host plants, suggests that gall morphology is relatively unconstrained by host plant. The low, significant values of Blomberg’s K for gall morphology principal components 1 and 2 may be indicative of adaptive evolution accompanied by shifts between selective environments (Blomberg et al., 2003; Ackerly, 2009), such that related taxa differ more than expected. However, low values of K can also result from bounded Brownian motion or stabilizing selection around a single optimum, as might be characterized by an O-U process (Revell et al., 2008; Ackerly, 2009). In the current case, it may reflect some of both; O-U models were consistently the best fit for the evolution of gall morphology, but they were not significantly better than models employing Pagel’s k, which, for low values, is consistent with a rapid initial radiation followed by relative stasis (Pagel, 1999). Furthermore, it is clear that in some areas of the phylogeny, there is apparently rapid divergence in gall traits (e.g. the S. altissima–S. gigantea–S. rugosa clade) and in others relatively little (e.g. the S. juncea–S. lepida–S. missouriensis clade; Figs 1 and 4).



Diversification of A. carbonifera appears to involve both ecological opportunity and key innovation. Ecological opportunity is provided by the existence of a diverse array of closely related plant species that may serve as hosts, while the association with and use of fungi may represent a key innovation, enabling them to colonize these diverse hosts (Janson et al., 2008, 2010) and which they can mould for defence against specific natural enemies (the ‘enemy hypothesis’, which posits that the specific morphology of galls reflects adaptive shapes or positions on host plants that reduced predation in an ancestral population; Stone & Schonrogge, 2003). In general, most of the variation in gall morphology among galling insects is thought to be associated with defence (Stone & Schonrogge, 2003; Cooper & Rieske, 2010). Likewise, we found that parasitism frequency varies strongly among co-occurring Asteromyia gall morphs (Stireman et al., 2008; Wells, 2010; J. O. Stireman, J. J. Heath & B. L. Wells, unpublished), suggesting that enemies represent a key selective force on gall morphology. Our estimates of parasitism vary widely among A. carbonifera host- and morph-related populations, but these data are based on limited collections and sample sizes for most populations. Parasitism can vary considerably over space and time on a particular host (Wells, 2010; J. O. Stireman, unpublished), and we are hesitant to over-interpret differences in parasitism among populations. Just as past competition may have shaped the traits and distributions of populations that today show little evidence of competing (Connell’s (1980) ‘ghost of competition past’), so too may parasitism shape the traits (gall morphology) and distribution (hosts) of midges, although today the fitness benefits of these evolutionary changes may not be apparent. Rapid, ‘Red Queen’ processes of host–parasitoid coevolution hinder our ability to make inferences about past selective pressures from present variation in parasitism (e.g. Heard et al., 2006). However, we suggest here that selection by enemies has left an imprint in the form of divergent gall phenotypes in this system. A particularly promising avenue for future work would be focused studies of closely related populations to determine whether shifts in hosts and gall traits allow transient release from parasitism (Jeffries & Lawton, 1984). If so, we should expect to see relatively little phylogenetic conservatism in such traits, with the greatest variation among the most recently diverged lineages. Our findings are at least partly consistent with this expectation, suggesting rapid evolution of gall morphology and frequent host plant shifts. It appears that on Solidago taxa, in a manner reminiscent of a classic archipelago radiation, there are replicate instances of functional divergence in gall traits, overlain by a complex history of dispersal (host colonization) and vicariance (host-associated differentiation), a phenomenon thought to be rare among mobile,

2001

continental organisms (Losos, 2009, 2010).The genus Asteromyia is part of a much larger clade, the Cecidomyiinae (gall-forming members of the Cecidomyiidae), that appears to be prone to adaptive radiation (Price, 2005), probably due to the same processes suggested here: the ecological opportunity presented by diverse host plants and the divergent and coevolutionary selection pressures provided by multitrophic interactions (Singer & Stireman, 2005). Several lineages of gall midges have experienced recent explosive radiations on angiosperms (e.g. Joy & Crespi, 2007; Dorchin et al., 2009). Interestingly, this has occurred both in groups that possess fungal mutualists (e.g. the genus Asphondylia 247 spp.) and in groups that lack them (e.g. the genera Rhopalomyia, 250 spp. and Dasineura, 448 spp.; Gagne´, 2004). Mechanisms of isolation While the strong host- and morph-associated genetic structure in this group indicates that host plant traits and enemy attack may select for divergence, it is unclear how gene flow between populations is interrupted. Field observations provide no evidence of mating on the host (Heath & Stireman, 2010), but where mating does occur and how mates are selected remains a mystery. Given the short adult lifespan, micro-spatial and/or temporal isolation could play a role, particularly if plants and enemies exhibit strong spatiotemporal structure, as has been found for populations on S. altissima (Wells, 2010). However, AMOVAs presented here suggest extensive gene flow over vast distances. Other mechanisms are intriguing, but speculative. For example, a common element in both plant and midge divergence may be the heightened tendency for ploidy variation in Solidago (Semple et al., 1999; Halverson et al., 2008a; Schlaepfer et al., 2008), and such variation can have a strong influence on insect communities (Thompson et al., 2004; Halverson et al., 2008b). Another possibility is that the fungal symbiont may influence direct or indirect interactions between A. carbonifera and Solidago. The trophic relationship with the fungal mutualist may buffer the midge from physiological variation among Solidago species and thus facilitate host shifts and expansions. On the other hand, the midge–fungal association may encourage specialization if the ability of the midge to fine-tune gall morphology in response to enemy pressure is dependent on biochemical interactions with host plant tissues.

Conclusions In summary, our work reveals a large degree of cryptic host–associated genetic structure in a gall midge, highlighting the extent to which host plants can drive selective divergence between insect lineages. We also show that gall morphology appears to be relatively


2002


unconstrained by host plant: genetically and phenotypically distinct gall morphs co-occur on the same host plants. Likewise, similar morphs can occur on different hosts. We do not yet know the causes of morphological variation in galls in Asteromyia. However, most of the variation in gall morphology among galling insects is thought to be associated with defence (Stone & Schonrogge, 2003), and parasitism frequency varies strongly among co-occurring Asteromyia gall morphs (Stireman et al., 2008; Wells, 2010; J.O. Stireman, J.J. Heath & B.L. Wells, unpublished). It is likely then that enemies exert selection on gall morphology. In these respects, this system may be representative of many other taxa occupying intermediate trophic levels (e.g. phytophagous groups) that are sandwiched between diverse and often opposing selective forces (Price et al., 1980). Such groups often display the greatest levels of diversity (e.g. Mitter et al., 1988) and represent the bulk of adaptive radiations throughout Earth’s history. Given the diversity of specialized interactions that midges like A. carbonifera express (plants, fungi, and predators), as well as the surprising number of lineages we uncovered in this work, a general implication is that there is a need in the study of ecological speciation for increased attention towards the multitrophic communities in which species are imbedded. It is likely that appreciation of the effect of the broader community on adaptive diversification will lead to a more general and robust understanding of the causes of ecological speciation.

Acknowledgments We would like to thank Tim Carr for providing a large number of samples and sequence data from the northeast United States and John Semple for identifying many of the Solidago host plant species. We thank Jeremy Heath, Eric Janson, Brenda Wells, Seth Jenkins and Annie Doyle for aiding in gall collecting and dissections. Five Rivers Metroparks (Dayton, Ohio), Ohio State Parks and Greene County parks (Ohio) provided permits for sample collection and field studies. This work was supported by an NSF DEB 0614433 Grant to P.A. and J.O.S. and by a Wright State University Research Challenge award to J.O.S. None of the authors have any conflicts of interest.

References Abouheif, E. 1999. A method for testing the assumption of phylogenetic independence in comparative data. Evol. Ecol. Res. 1: 895–909. Abrahamson, W.G. & Weis, A.E. 1997. Evolutionary Ecology Across Three Trophic Levels: Goldenrods, Gallmakers and Natural Enemies. Princeton Monographs in Population Biology. Princeton University Press, Princeton, New Jersey.

Ackerly, D.D.. 2009. Phylogenetics and comparative methods. In: Guide to Ecology (S. Levin, ed.), pp. 117–125. Princeton University Press, Princeton, New Jersey. Bailey, R., Schonrogge, K., Cook, J.M., Melika, G., Csoka, G., Thuroczy, C. et al. 2009. Host niches and defensive extended phenotypes structure parasitoid wasp communities. PLoS Biol. 7: e1000179. Berlocher, S.H. 2000. Radiation and divergence in the Rhagoletis pomonella species group: Inferences from allozymes. Evolution 54: 543–557. Berlocher, S.H. & Feder, J.L. 2002. Sympatric speciation in phytophagous insects: moving beyond controversy? Annu. Rev. Entomol. 47: 773–815. Bernays, E.A. & Graham, M. 1988. On the evolution of host specificity in phytophagous arthropods. Ecology 69: 886– 892. Blomberg, S.P., Garland, T. & Ives, A.R. 2003. Testing for phylogenetic signal in comparative data: Behavioral traits are more labile. Evolution 57: 717–745. Borkent, A. & Bisset, J. 1985. Gall midges (Diptera: Cecidomyiidae) are vectors for their fungal symbionts. Symbiosis 1: 185–194. Butler, M.A. & King, A.A. 2004. Phylogenetic comparative analysis: a modeling approach for adaptive evolution. Am. Nat. 164: 683–695. Connell, J.H. 1980. Diversity and the coevolution of competitors, or the ghost of competition past. Oikos 35: 131–138. Cooper, W.R. & Rieske, L.K. 2010. Gall structure affects ecological associations of Dryocosmus kuriphilus (Hymenoptera: Cynipidae). Environ. Entomol. 39: 787–797. Coyne, J.A. & Orr, H.A.. 2004. Speciation. Sinauer Associates, Sunderland, MA. Craig, T.P. & Itami, J.K. 2011. Divergence of Eurosta solidaginis in response to host plant variation and natural enemies. Evolution 65: 802–817. Craig, T.P., Itami, J.K. & Horner, J.D. 2007. Geographic variation in the evolution and coevolution of a tritrophic interaction. Evolution 61: 1137–1152. Crego, C.L., Weis, A.E., Polans, N.O. & Bretz, C.K. 1990. Sympatric sibling species from 3 phenotypically distinct Asteromyia (Diptera, Cecidomyiidae) galls on the same host plantspecies. Ann. Entomol. Soc. Am. 83: 149–154. Dobzhansky, T. 1937. Genetic nature of species differences. Am. Nat. 71: 404–420. Dorchin, N., Scott, E.R., Clarkin, C.E., Luongo, M.P., Jordan, S. & Abrahamson, W.G. 2009. Behavioural, ecological and genetic evidence confirm the occurrence of host-associated differentiation in goldenrod gall-midges. J. Evol. Biol. 22: 729–739. Dugan, K.A., Lawrence, H.S., Hares, D.R., Fisher, C.L. & Budowle, B. 2002. An improved method for post-PCR purification for mtDNA sequence analysis. J. Forensic Sci. 47: 811–818. Ehrlich, P.R. & Raven, P.H. 1964. Butterflies and plants: a study in coevolution. Evolution 18: 586–608. Emelianov, I., Simpson, F., Narang, P. & Mallet, J. 2003. Host choice promotes reproductive isolation between host races of the larch budmoth Zeiraphera diniana. J. Evol. Biol. 16: 208–218. Feder, J.L., Chilcote, C. & Bush, G. 1988. Genetic differentiation between sympatric host races of the apple maggot fly Rhagoletis pomonella. Nature 336: 61–64.



Funk, D.J. 1998. Isolating a role for natural selection in speciation: Host adaptation and sexual isolation in Neochlamisus bebbianae leaf beetles. Evolution 52: 1744–1759. Funk, D.J., Filchak, K.E. & Feder, J.L. 2002. Herbivorous insects: Model systems for the comparative study of speciation ecology. Genetica (Dordrecht) 116: 251–267. Funk, D.J., Nosil, P. & Etges, W.J. 2006. Ecological divergence exhibits consistently positive associations with reproductive isolation across disparate taxa. Proc. Natl. Acad. Sci. U. S. A. 103: 3209–3213. Gagne´, R.J. 1968. A taxonomic revision of the genus Asteromyia (Diptera: Cecidomyiidae). Misc. Pubs. Entomol. Soc. Am. 6: 1–40. Gagne´, R.J. 1989. The Plant-Feeding Gall Midges of North America. Cornell University Press, Ithaca, New York. Gagne´, R.J. 2004. A catalog of the Cecidomyiidae (Diptera) of the world. Mem. Ent. Soc. Wash. 25: 1–408. Gittleman, J.L. & Kot, M. 1990. Adaptation – statistics and a null model for estimating phylogenetic effects. Syst. Zool. 39: 227–241. Hajibabaei, M., Janzen, D.H., Burns, J.M., Hallwachs, W. & Hebert, P.D.N. 2006. DNA barcodes distinguish species of tropical lepidoptera. Proc. Natl. Acad. Sci. U. S. A. 103: 968–971. Halverson, K., Heard, S.B., Nason, J.D. & Stireman, J.O. 2008a. Origins, distribution, and local co-occurrence of polyploid cytotypes in Solidago altissima (Asteraceae). Am. J. Bot. 95: 50–58. Halverson, K., Heard, S.B., Nason, J.D. & Stireman, J.O. 2008b. Differential attack on diploid, tetraploid, and hexaploid Solidago altissima L. by five insect gallmakers. Oecologia 154: 755–761. Harmon, L.J., Weir, J.T., Brock, C.D., Glor, R.E. & Challenger, W. 2008. GEIGER: Investigating evolutionary radiations. Bioinformatics 24: 129–131. Heard, S.B., Stireman, J.O., Nason, J.D., Cox, G.H., Kolacz, C.R. & Brown, J.M. 2006. On the elusiveness of enemy-free space: Spatial, temporal, and host-plant-related variation in parasitoid attack rates on three gallmakers of goldenrods. Oecologia 150: 421–434. Heath, J.J. & Stireman, J.O. 2010. Dissecting the association between a gall midge, Asteromyia carbonifera, and its symbiotic fungus, Botryosphaeria dothidea. Entomol. Exp. Appl. 137: 36–49. Janson, E.M., Grebenok, R.J., Behmer, S.T. & Abbot, P. 2009. Same host-plant, different sterols: variation in sterol metabolism in an insect herbivore community. J. Chem. Ecol. 35: 1309–1319. Janson, E.M., Peeden, E.R., Stireman, J.O. & Abbot, P. 2010. Symbiont-mediated phenotypic variation without co-evolution in an insect-fungus association. J. Evol. Biol. 23: 2212– 2228. Janson, E.M., Stireman, J.O., Singer, M.S. & Abbot, P. 2008. Phytophagous insect–microbe mutualisms and adaptive evolutionary diversification. Evolution 62: 997–1012. Jeffries, M.J. & Lawton, J.H. 1984. Enemy free space and the structure of ecological communities. Biol. J. Linn. Soc. 23: 269–286. Jombart, T., Balloux, F. & Dray, S. 2010. Adephylo: New tools for investigating the phylogenetic signal in biological traits. Bioinformatics 26: 1907–1909. Joy, J.B. & Crespi, B.J. 2007. Adaptive radiation of gall-inducing insects within a single host-plant species. Evolution 61: 784–795.

2003

Kembel, S.W., Cowan, P.D., Helmus, M.R., Cornwell, W.K., Morlon, H., Ackerly, D.D. et al. 2010. Picante: R tools for integrating phylogenies and ecology. Bioinformatics 26: 1463–1464. Laureto, P.J. & Barkman, T.J. 2011. Nuclear and chloroplast DNA suggest a complex single origin for the threatened allopolyploid Solidago houghtonii (Asteraceae) involving reticulate evolution and introgression. Syst. Bot. 36: 209– 226. Legendre, P., Desdevises, Y. & Bazin, E. 2002. A statistical test for host-parasite coevolution. Syst. Biol. 51: 217–234. Lill, J.T., Marquis, R.J. & Ricklefs, R.E. 2002. Host plants influence parasitism of forest caterpillars. Nature 417: 170–173. Liu, L. & Yu, L. 2010. Phybase: An R package for species tree analysis. Bioinformatics 26: 962–963. Liu, L., Yu, L., Kubatko, L., Pearl, D.K. & Edwards, S.V. 2009. Coalescent methods for estimating phylogenetic trees. Mol. Phylo. Evol 53: 320–328. Liu, L., Yu, L. & Pearl, D.K. 2010. Maximum tree: A consistent estimator of the species tree. J. Math. Biol. 60: 95–106. Losos, J.B.. 2009. Lizards in an Evolutionary Tree. University of California Press, Berkeley. Losos, J.B. 2010. Adaptive radiation, ecological opportunity, and evolutionary determinism. Am. Nat. 175: 623–639. Maddison, W.P. & Maddison, D.R. 2009. Mesquite: a modular system for evolutionary analysis, Version 2.72 http:// mesquiteprojectorg. Matsubayashi, K.W., Ohshima, I. & Nosil, P. 2010. Ecological speciation in phytophagous insects. Entomol. Exp. Appl. 134: 1–27. Mitter, C., Farrell, B. & Wiegmann, B. 1988. The phylogenetic study of adaptive zones – Has phytophagy promoted insect diversification. Am. Nat. 132: 107–128. Nosil, P. & Crespi, B.J. 2006. Experimental evidence that predation promotes divergence in adaptive radiation. Proc. Natl. Acad. Sci. U. S. A. 103: 9090–9095. Ober, K.A. 2002. Phylogenetic relationships of the carabid subfamily Harpalinae (Coleoptera) based on molecular sequence data. Mol. Phylo. Evol 24: 228–248. Oksanen, J., Guillaume Blanchet, F., Kindt, R., Legendre, P., O’Hara, R.B., Simpson, G.L. et al. 2010 Vegan: Community Ecology Package R package version 117-2 http://CRANRprojectorg/package=vegan. Pagel, M. 1999. Inferring the historical patterns of biological evolution. Nature 401: 877–884. Paradis, E. 2010. Pegas: An R package for population genetics with an integrated-modular approach. Bioinformatics 26: 419–420. Paradis, E., Claude, J. & Strimmer, K. 2004. APE: Analyses of phylogenetics and evolution in R language. Bioinformatics 20: 289–290. Peakall, R. & Smouse, P.E. 2006. Genalex 6: genetic analysis in Excel. Population genetic software for teaching and research. Mol. Ecol. Notes 6: 288–295. Price, P.W. 2005. Adaptive radiation of gall-inducing insects. Basic Appl. Ecol. 6: 413–421. Price, P.W., Bouton, C.E., Gross, P., McPheron, B.A., Thompson, J.N. & Weis, A.E. 1980. Interactions among 3 trophic levels influence of plants on interactions between insect herbivores and natural enemies. Annu. Rev. Ecol. Syst. 11: 41–65. Rambaut, A. & Drummond, A.J.. 2009. Tracer v1.5. Available from http://beast.bio.ed.ac.uk/Tracer


2004


Rasband, W.S.. 1997-2011. ImageJ. US National Institutes of Health, Bethesda, Maryland, USA, see http://imagejnihgov/ij/. Revell, L.J., Harmon, L.J. & Collar, D.C. 2008. Phylogenetic signal, evolutionary process, and rate. Syst. Biol. 57: 591– 601. Ronquist, F. & Huelsenbeck, J.P. 2003. MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19: 1572–1574. Rundle, H.D. & Nosil, P. 2005. Ecological speciation. Ecol. Lett. 8: 336–352. Sanderson, M.J. 2002. Estimating absolute rates of molecular evolution and divergence times: a penalized likelihood approach. Mol. Biol. Evol. 19: 101–109. Schemske, D.W. 2010. Adaptation and the origin of species. Am. Nat. 176: S4–S25. Schlaepfer, D.R., Edwards, P.J., Widmer, A. & Billeter, R. 2008. Phylogeography of native ploidy levels and invasive tetraploids of Solidago gigantea. Mol. Ecol. 17: 5245–5256. Semple, J.C. & Cook, R.E. 2006. Solidago Linnaeus. In: Flora of North America Vol 20 Asteraceae Part 2 Astereae and Senecioneae (Flora North America Editorial Committee, eds), pp. 107– 166. Oxford University Press, New York. Semple, J.C., Ringius, G.S. & Zhang, J.J. 1999. The Goldenrods of Ontario: Solidago L. and Euthamia Nutt. 3rd ed. University of Waterloo Biological Series 39: 1–90. Simon, C., Frati, F., Beckenbach, A., Crespi, B., Liu, H. & Flook, P. 1994. Evolution, weighting, and phylogenetic utility of mitochondrial gene sequences and a compilation of conserved polymerase chain reaction primers. Ann. Entomol. Soc. Amer. 87: 651–701. Singer, M.S. & Stireman, J.O., III. 2005. The tri-trophic niche concept and adaptive radiation of phytophagous insects. Ecol. Lett. 8: 1247–1255. Stireman, J.O., III, Nason, J.D. & Heard, S.B. 2005. Host-associated genetic differentiation in phytophagous insects: General phenomenon or isolated exceptions? Evidence from a goldenrod-insect community Evolution 59: 2573–2587. Stireman, J.O., III, Janson, E.M., Carr, T.G., Devlin, H. & Abbot, P. 2008. Evolutionary radiation of Asteromyia carbonifera (Diptera: Cecidomyiidae) gall morphotypes on the goldenrod Solidago altissima (Asteraceae). Biol. J. Linn. Soc. 95: 840–858. Stireman, J.O., III, Devlin, H., Carr, T.G. & Abbot, P. 2010. Evolutionary diversification of the gall midge genus Asteromyia (Cecidomyiidae) in a multitrophic ecological context. Mol. Phylogenet. Evol. 54: 194–210. Stone, G.N. & Schonrogge, K. 2003. The adaptive significance of insect gall morphology. Trends Ecol. Evol. 18: 512–522. Tamura, K. & Nei, M. 1993. Estimation of the number of nucleotide substitutions in the control region of mitochondrial-DNA in humans and chimpanzees. Mol. Biol. Evol. 10: 512–526. Tamura, K., Peterson, D., Peterson, N., Stecher, G., Nei, M. & Kumar, S. 2011. MEGA5: Molecular evolutionary genetics analysis using maximum likelihood, evolutionary distance, and maximum parsimony methods. Mol. Biol. Evol. 28: 2731–2739.

Thompson, J.N.. 2008. Coevolution, cryptic speciation, and the persistence of plant-insect interactions. In: Specialization, Speciation, and Radiation (K. Tilmon, ed.), pp. 216–224. University of California Press, Berkeley. Thompson, J.N., Nuismer, S.L. & Merg, K. 2004. Plant polyploidy and the evolutionary ecology of plant/animal interactions. Biol. J. Linn. Soc. 82: 511–519. Weis, A.E. 1982a. Resource utilization patterns in a community of gall-attacking parasitoids. Environ. Entomol. 11: 809–815. Weis, A.E. 1982b. Use of a symbiotic fungus by the gall maker Asteromyia carbonifera to inhibit attack by the parasitoid Torymus capite. Ecology 63: 1602–1605. Wells, B.L.. 2010. Ecological Speciation in a Multi-Trophic Complex: Gall Midges, Goldenrods, and Parasitoids. MS Thesis, Wright State University, Dayton, Ohio.

Supporting information Additional Supporting Information may be found in the online version of this article: Appendix S1 Species of Solidago that are known hosts of Asteromyia carbonifera. Table S1 Sampling locations and hosts for Asteromyia with NCBI accession numbers for mtDNA and nuclear sequences Table S2 Gall parasitism rates and gall trait means for A. carbonifera populations Table S3 PCA loadings for gall morphology employing population trait means. See text for an explanation of variables. Figure S1 Bayesian consensus tree of A. carbonifera populations based on the nuclear gene wingless. Figure S2 Linearized phylogenetic reconstructions based on mtDNA and the nuclear wingless gene with lines connecting taxa. Figure S3 Linearized bootstrap consensus trees of midge and Solidago host phylogenies based on ML analysis of trimmed data sets, with associations indicated by links. Figure S4 The Maximum Species Tree of A. carbonifera populations with pie charts showing the likelihood of ancestral states for host plants (by color; from ML analyses with unequal transition rates). As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Received 23 March 2012; revised 16 May 2012; accepted 25 June 2012
