We have studied 52 Chrysolina species representing 30 out of the ca. 65 subgenera currently recognized for the genus (Bieńkowski 2001, Kippenberg 2010), plus five Oreina species from two subgenera. Our sampling includes type species representatives regarding 13 of the studied Chrysolina subgenera and one type species for Oreina. In addition, several representatives of other genera of the subfamily Chrysomelinae were analysed as outgroups, including a species from the early-divergent genus Timarcha Latreille (Gómez-Zurita et al. 2008) (Table 1). Beetles were collected by us in the field or received from colleagues in absolute ethanol and stored in the laboratory at -20 °C before processing. Voucher specimens are deposited for long-term storage at the DNA and tissue collection of the Biodiversity, Systematics and Evolution group (Bio6Evo) of the University of the Balearic Islands.

Total DNA was purified from beetle head and pronotum using the DNeasy Tissue kit (Qiagen, West Sussex, UK) and following the manufacturer’s protocol. Elutions were done in 200 μL volume and one microliter was used in PCR reactions. Three different molecular markers were selected for the study, including a partial sequence of the mitochondrial 16S rDNA (rrnL; primers LR-N-13398 and LR-J-12887; Simon et al. 1994), a partial sequence of the mitochondrial cytochrome c oxidase subunit 1 gene (cox1; primers C1-J-2183 and TL2-N-3014; Simon et al. 1994), and a fragment from the nuclear histone 3 gene (H3; primers H3aF and H3aR; Colgan et al. 1998). PCR conditions used 0.2 μM of each primer and 3.5 mM MgCl₂ using a standard protocol of 35 cycles with annealing temperature ranging from 50 to 45 °C (60s) depending on the sample, and denaturation (94 °C) and elongation (72 °C) lasted 30 and 60s, respectively. PCR products were visualized by 1% agarose gel electrophoresis and subsequently purified using MSB Spin PCRapace (Invitek, Berlin, Germany). Sanger sequencing was performed with the same primers as above using the BigDye Terminator Cycle Sequencing kit (Applied Biosystems, Foster City, CA, USA). Sequences were edited and contigs were assembled using BIOEDIT v. 7 (Hall 1999), and deposited at GenBank under the accession numbers referred in Table 1.

Heterogeneity in base composition across taxa was explored for each codon position of the protein-coding genes and for rrnL using the chi-square test for base frequency differences implemented in PAUP*4.0b10 (Swofford 2003). Multiple sequence alignment was performed using MAFFT 7 online version (http://mafft.cbrc.jp/alignment/server/, Katoh and Standley 2013) under default parameters. Molecular markers were checked for combinability using the incongruence length difference (ILD) test (Farris et al. 1994) implemented in PAUP* v4.0b10 (Swofford 2002). The test was run using 100 random stepwise additions and 1000 replicates of heuristic search with tree bisection–reconnection (TBR) branch swapping. The optimal partitioning strategy and evolutionary models for the combined sequence matrix were assessed with PartitionFinder (Lanfear et al. 2012) under the Bayesian Information Criterion (BIC) and using the implemented greedy algorithm.

Bayesian phylogenetic inference was conducted using MrBayes 3.2 (Ronquist et al. 2012). Two independent analyses consisting of four chains each were run for 5·10⁶ generations specifying a sampling frequency every 100 generations, and setting a burn-in fraction of 10%. MCMC convergence and the effective sample sizes (ESS) estimates were checked with TRACER v. 1.5 (Rambaut and Drummond 2007). Additionally, a maximum likelihood search was done using GARLI v.2.01 (Zwickl 2006) and performing 100 bootstrap replicates.

Specific hypotheses of monophyly were tested using a ML framework and the Approximately Unbiased test (AU test, Shimodaira 2000) as implemented in the CONSEL program (Shimodaira and Hasegawa 2001). We compared our molecular phylogenetic hypothesis with some of the most relevant systematic proposals for the genus Chrysolina (see results). Prior to the evaluation of each taxonomic scenario, a ML phylogenetic analysis was performed in GARLI v.2.01 using the same partitioning scheme and models as in the phylogenetic searches described above, but enforcing the monophyly of the taxa of interest. Once the resulting ML trees were obtained, their per site log-likelihoods were calculated using RAxML v8.0.X program (Stamatakis 2014) and used as input data in CONSEL.

Ancestral host plant affiliations were reconstructed using BayesTraits v. 2.0 (Pagel and Meade 2013) selecting the MCMC mode and the “multistate” model of evolution (Pagel et al. 2004). To take into account phylogenetic uncertainty, reconstructions were based on 1000 randomly selected post-burnin Bayesian trees from the phylogenetic analysis in MrBayes 3.2. Following the manual’s recommendations (http://www.evolution.rdg.ac.uk/BayesTraitsV2.0Files/TraitsV2Manual.pdf), the reversible-jump (RJ) MCMC with a hyperprior approach was chosen, and the interval of 0–30 for the RJ-hyperprior implementing an exponential distribution was applied. The “addMRCA” command was used to calculate the posterior distribution of ancestral character states at selected nodes in the Bayesian Chrysolina tree. A total of 10·10⁶ generations were run, with samples taken every 100 iterations and discarding a burn-in fraction of 10%. Results of the MCMC runs including the ESS values were analysed in TRACER v. 1.5.

We also used BayesTraits to evaluate different ancestral host plant affiliations scenarios at the root of the Chrysolina tree. Analyses were conducted by enforcing the ancestral state of the most recent common ancestor (mrca) for the core Chrysolina node (excluding the divergent species Ch. vigintimaculata) to be one of the eight host plant families recorded for the studied Chrysolina species. MCMC was used to explore the samples and the space rate parameter of 1000 post-burnin trees generated in the MrBayes analysis. We performed two independent runs of 10·10⁶ generations for each one of the constrained searches, and sampling rate parameters every 100 generations. The constrained runs were then compared by calculating the Bayes factors between the best and second best models based on the harmonic mean of the likelihood from each analysis as indicated in BayesTraits manual.

Lengths of the amplified gene fragments ranged from 581 to 794 bp for cox1, 278 to 512 bp for rrnL, and 294 to 363 for H3. Total length of the concatenated DNA sequence matrix was 1682 bp. In cox1, 48.36% of the aligned positions were variable, indicating high divergence level among the studied sequences. Indeed, accumulation of mutations for cox1 was higher than for the other markers, as shown by the pairwise sequence divergence (p-distance), which ranged between 0.0063 and 0.2236 (average: 0.1331±0.0105) for cox1, 0.0012 and 0.1723 (average: 0.0924±0.0100) for rrnL, and 0.0027 and 0.1077 (average: 0.0641±0.0108) for H3. Also, cox1 and rrnL sequences showed the well-known A+T bias typical of insect mtDNA (69.9% and 76,4%, respectively), whereas base frequency was more balanced in the nuclear H3 marker (54,8%). Chi-squared tests for bias in base composition showed no significant heterogeneity in our datasets (P>0.99). On the other hand, ILD test revealed no evidence of incongruence among molecular markers (P= 0.24), and we therefore performed all subsequent phylogenetic analyses following a supermatrix approach.

The best-fit partitioning scheme selected by PartitionFinder under BIC divided the data into seven subsets, each with its own model of molecular evolution (Table 2). The effective sample size value for each parameter sampled from the MCMC analysis was always >200. Bayesian and ML searches resulted in almost the same topology (Figures 1 and 2), with few discrepancies affecting only unsupported relationships such as the placement of the species Chrysolina bicolor (Fabricius, 1775), the position of the subgenus Sulcicollis (Fairmaire, 1887), and the internal branching pattern of the three species of the subgenus Chrysolinas. str. Motschulsky. Both phylogenetic approaches also yielded similar results in terms of nodal support, differing mainly in the values associated to some of the basal nodes of the core Chrysolina clade, which were higher in the Bayesian analysis (e.g. nodes K, D and T). The resulting phylogenetic trees revealed the paraphyly of the genus Chrysolina as currently described, due to the inclusion of the Oreina representatives within the Chrysolina clade (Figures 1 and 2). The genus Oreina is also recovered as a paraphyletic clade that includes the species Chrysolina haemochlora (Gebler, 1823). The results showed the monophyly of the studied Chrysolina (plus Oreina) species [clade A, Bayesian posterior probability (pp)=1, bootstrap=100] excepting the African taxa Chrysolina (Polysticta) vigintimaculata, which showed a higher affinity with outgroup taxa. In addition, the monophyletic status of the subgenera with more than one species sampled in the study was recovered in all cases excepting Anopachys Motschulsky, Chalcoidea, TimarchopteraMotschulsky and Oreina subgenus Chrysochloa Hope. The inferred topology allowed for the identification of four main monophyletic subgenera assemblages within the core Chrysolina clade with high support values in at least one of the resulting trees (clades B, C, D and K). Within these main lineages, it was also possible to identify systematic relationships among subgenera at different phylogenetic levels. The inferred groups of phylogenetically related subgenera and their statistical supports are summarized in Table 3.

Figure 1. 

Bayesian phylogenetic tree obtained from the combined analysis of cox1, rrnL and H3. Node numbers represent Bayesian posterior probability values. Only support values higher than 0.9 are shown. Numbers accompanying the subgeneric classification of the Chrysolina species on the right correspond to the systematic groups defined by Bourdonné and Doguet (1991). Clades mentioned in the text are highlighted.

Figure 2. 

Maximum likelihood phylogenetic tree obtained from the combined analysis of cox1, rrnL and H3. Node numbers represent bootstrap support values. Only support values higher than 0.7 are shown. Numbers accompanying the subgeneric classification of the Chrysolina species on the right correspond to the systematic groups defined by Bourdonné and Doguet (1991). Clades mentioned in the text are highlighted.

Constrained ML searches were used to evaluate a number of taxonomic hypotheses for Chrysolina and Oreina using the AU test (Table 4). The phylogenetic scenarios that were rejected in the analyses included the systematic placement of Oreina as a different genus from Chrysolina (P=0.016), the synonymy of subgenera Paraheliostola L. N. Medvedev and Timarchoptera (Mikhailov 2002, P=0.001), the suggestion of a close relationship between Threnosoma Motschulsky and Ch. (Timarchoptera) haemochlora (Mikhailov 2005, P<0.001), the reciprocal monophyly of Colaphodes and Taeniochrysa (Hsiao and Pasteels 1999, P<0.001), the inclusion of Chrysolina timarchoides (Brisout, 1882) within the subgenera Maenadochrysa Bechyné (Bieńkowski 2001, P<0.001), the recognition of Craspeda sensu Bourdonné 2005 as a different genus from Chrysolina (P<0.01), the segregation from Chrysolina of the subgenera Allochrysolina Bechyné, Chalcoidea and Pezocrosita Jakobson (Bourdonné 2012, P<0.01), the monophyly of the Chrysolina species belonging to the “group 2” described by Bourdonné and Doguet (1991) (P<0.001) (Table 1), as well as the monophyly of the Chrysolina species feeding on hosts from the same plant family (Apiaceae, Asteraceae, Lamiaceae, Plantaginaceae, Ranunculaceae and Scrophulariaceae; P≤0.001 in all cases). Conversely, the molecular data could not reject the reciprocal monophyly of several taxa assemblages, such as Chrysolina vigintimaculata and the rest of the studied Chrysolina species (P=0.165), Chrysolina species belonging to the “group 6” described by Bourdonné and Doguet (1991) (P=0.527) (Table 1), subgenera Allochrysolina and Anopachys (Hsiao and Pasteels 1999, P=0.215), subgenera Chalcoidea and Hypericia (Pasteels et al. 2003, P=0.066), subgenera Allochrysolina, Chalcoidea and Pezocrosita (Bourdonné 2012, P=0.205), and the subgenera Palaeosticta Bechyné and Taeniosticha Motschulsky (Bourdonné 2005, P=0.198). Also, the monophyly of the sampled species concerning the subgenera Anopachys, Chalcoidea and Oreina subgenus Chrysochloa could not be rejected (P≥0.212 in all cases).

The Bayesian reconstruction of ancestral host plant associations showed an ancient affiliation with Lamiaceae at the root of the core Chrysolina clade (Figure 3, node A, P=0.98; Table 5). This plant family was also recovered as the most likely ancestral host for three of the main clades in our molecular phylogeny (nodes B, C and D; P=0.94, 0.99 and 0.95, respectively). Within clade D, a host shift from Lamiaceae towards Asteraceae (P=0.54) and/or Apiaceae (P=0.37) was detected for the mrca of Oreina and Chrysolina (Timarchoptera) haemochlora (clade G’). On the other hand, ancestral host plant reconstruction for node K was ambiguous, recovering associations with multiple families. However, it was possible to identify the occurrence of several host shifts for its derived lineages towards a new trophic association with (i) Apiaceae (node K’’, P=0.62), (ii) Hypericaceae (nodes P and Q, P=0.51 and 0.97, respectively), (iii) Asteraceae (node R, P=0.94), (iv) Plantaginaceae (node X, P=0.91), and (v) Scrophulariaceae (node Z’, PP=0.66). Nodes W and Y’ respectively showed a reversal shift from an ancestral Plantaginaceae host to the original Lamiaceae host family (P=0.5) as well as a new trophic link with Asteraceae (P=0.5).

Figure 3. 

Ancestral reconstruction of host plant affiliations in the studied species of Chrysolina and Oreina. Terminal taxa are coded according to the available host plants records from the literature (Table 1). Pie charts at selected nodes show probabilities of each state from the Bayesian analysis in BayesTraits. Clades mentioned in the text are highlighted.

Results from Bayes factor comparisons of the constraint hypotheses for the ancestral plant family at the root of the core Chrysolina clade (node A) corroborated MCMC ancestral state reconstruction, offering positive to very strong statistical support for an ancestral trophic association with Lamiaceae (Table 6).

The mitochondrial and nuclear genes used here provided an expanded and better-resolved tree topology for the genus Chrysolina, significantly improving previous phylogenetic hypotheses. Our results support the reciprocal monophyly of the studied species of Chrysolina (plus Oreina) including the divergent Ch. (Polysticta) vigintimaculata, whose relationship with the core Chrysolina-Oreina clade could not be rejected by the AU test. The inferred tree topologies recovered Ch. vigintimaculata as a well-differentiated lineage sister to the rest of the ingroup taxa. This species has been traditionally assigned to the subgenus Atechna Chevrolat (Bieńkowski 2001), a species of which was included in the phylogenetic analysis of Gómez-Zurita et al. (2008) based on three ribosomal genes and showing a clear divergence from the Chrysolina-Oreina clade. In addition, the same pattern was observed in a different phylogenetic study based on five molecular markers (Jurado-Rivera et al. in prep.) that included the species Ch. (Atechna) striata (Degeer, 1778). Although more data are needed, the available information indicates that these taxa may represent a lineage of early divergence within Chrysolina whose taxonomic status should be further investigated.

The inferred topology also supported most of the current subgeneric taxonomy of Chrysolina (Bieńkowski 2001, Kippenberg 2010), since the monophyly of the subgenera screened for more than one species could be demonstrated or alternatively could not be rejected by the AU test. The exceptions in this regard are the synonymy of the subgenus Paraheliostola with the subgenus Timarchoptera by Mikhailov (2002) and the combination of the species Ch. (Threnosoma) timarchoides with the subgenus Maenadochrysa by Bieńkowski (2001). In both cases the taxa in question were recovered with support as well-differentiated lineages, thus indicating that such taxonomic decisions could be wrong. Therefore, the subgenus Paraheliostola (type species Ch. soiota Jacobson, 1924) should be restored according to the present molecular phylogeny. Moreover, the available karyological evidence also conflicts with Bieńkowski’s (2001) proposal (Petitpierre 1975, 1981), and we thus agree with Daccordi and Ruffo (2005) and with Kippenberg (2010) in that Ch. timarchoides belongs in the subgenus Threnosoma.

The new molecular phylogeny also sheds light on the contentious issue of the taxonomic status of Oreina. Our analyses supported the inclusion of the studied Oreina species within the core Chrysolina clade, which was also backed up statistically in the AU test constraining these genera to be reciprocally monophyletic (Table 4). The sample included the type species of the genus, O. speciosa (Linnaeus, 1758), which further strengthens our findings and corroborates previous hypotheses that consider Oreina as part of the Chrysolina lineage (Chapuis 1874, Bourdonné and Doguet 1991, Daccordi 1994). Moreover, the species feeding on Apiaceae hosts, O. ganglbaueri (Jakob, 1953) and O. speciosa, were recovered as more closely related to the also Apiaceae feeding Ch. haemochlora than to the remainder of the Oreina species analysed here, reinforcing our conclusions and highlighting the need for a taxonomic revision for the group. On the other hand, the proposal of considering the genera Craspeda and Chalcoidea (sensuBourdonné 2005 and 2012, respectively) as separate lineages from the remainder of the Chrysolina species is not supported in our phylogenetic framework, although the monophyly of the taxa included in each of them could not be statistically rejected (Table 4). Thus, the recognition of Craspeda and/or Chalcoidea as valid genera would render Chrysolina paraphyletic.

Excluding the divergent species Ch. vigintimaculata, Chrysolina could be subdivided into four major clades (Figures 1 and 2, clades B, C, D and K). The clades B and C comprised species from the “group 2” defined by Bourdonné and Doguet (1991), all of them feeding on host plants belonging to the family Lamiaceae and with a diploid chromosome number of 2n=24 (Petitpierre 1975, 1981, 1983). The hypothetical monophyly of the aforementioned “group 2” was statistically rejected by the AU test, thus reinforcing our finding that such an assemblage of species does not constitute a natural group. The clade B included two monotypic subgenera (Chrysolinopsis Bechyné and Taeniochrysea, sensuBieńkowski 2001) that have been recently regarded as synonyms by Kippengberg (2010), a taxonomic decision that is strongly supported in our phylogenetic analyses. The monophyly of the species nested in clade C were also noted in the phylogenetic study of Garin et al. (1999), excepting the species Ch. cerealis (Linnaeus, 1767) that they recovered in a divergent clade as sister to Ch. fastuosa with maximum bootstrap support. Here we have analysed the subspecies Ch. cerealis cyaneoaurata (Motschulsky, 1860) inferring a clear relationship with the remainder of the members in clade C that is supported with maximum posterior probability and bootstrap values. Genetic distances (p-distance) between the sequences deposited in GenBank by Garin et al. (1999) regarding C. cerealis and our data for C. cerealis cyaneoaurata were unusually high for an intraspecific comparison (cox1: 0.14; rrnL: 0.08), thus suggesting that the taxa in question do not belong to the same species. It remains to be investigated whether their divergence is due to specimen misidentification or whether C. cerealiss. str. and C. cerealis cyaneoaurata really are different species. Meanwhile, the results about the systematic position of Ch. cerealis should be interpreted with caution.

Clade D defined the monophyletic origin of seven Chrysolina subgenera traditionally associated with the “group 2” proposed by Bourdonné and Doguet (1991) plus two Oreina subgenera included in “groups 5 and 6”, all of them with a karyotype 2n = 24 (Petitpierre 1975, 1981, 1983) excepting Ch. haemochlora (2n=27, Petitpierre and Mikhailov 2009). The affinity between the subgenera Colaphosoma Motschulsky and Maenadochrysa could be established with confidence agreeing with their shared feeding habits on Lamiaceae species of the tribe Mentheae (Jolivet and Petitpierre 1976, Jolivet et al. 1986, Bieńkowski 2010). On the other hand, the close relationship recovered in the present work among Ch. fastuosa and the studied Oreina species is consistent with the findings of Hsiao and Pasteels (1991) based on a different set of molecular markers. The authors concluded that such association was contradicted by strong morphological evidence, highlighting the need of further research on this issue. Our molecular phylogeny not only confirmed the monophyly of these taxa, but also revealed the inclusion of an additional Chrysolina species in this clade, Ch. haemochlora.

Interestingly, our results regarding the clade K were fully consistent with most species groupings established by Bourdonné and Doguet (1991) based on morphology, karyology and biology of the species (“groups 1, 3, 4, 7, 8, 9, 10 and 2 partim.”). Available molecular phylogenies of Chrysolina (Garin et al. 1999, Hsiao and Pasteels 1999) failed at recovering supported relationships among these groups, excepting the monophyletic origin of the species belonging in the “groups 1, 3 and 7” inferred by Garin et al. (1999). In contrast, our analyses allowed for the identification of their phylogenetic relationships at deep taxonomic level, and also extended the results to seven Chrysolina subgenera not studied by Bourdonné and Doguet (1991). The latter was the case of clade M, where the subgenera Crositops Marseul and Timarchoptera (more likely Paraheliostola, see above) were recovered as the sister lineage of the Threnosoma species regarded as “group 4”. Indeed, the subgenera Crositops and Threnosoma are known to share morphological attributes (Mikhailov 2005). Although no information is available for the species Ch. soiota, the remainder of the species in clade M feed on Apiaceae and also share a male karyotype 2n=47 (Petitipierre 1981, 1999, Petitpierre et al. 2004, Petitpierre and Mikhailov 2009), which is highly consistent with their close association recovered here. On the other hand, the existence of a relationship between the Mediterranean subgenus Threnosoma and the Siberian subgenus Timarchoptera proposed by Mikhailov (2005) was rejected by the AU test. Another subgenus that was not analysed by Bourdoneé and Doguet (1991) is represented in our sampling by the species Ch. (Pezocrosita) convexicollis (Jakobson, 1901), which appeared in the trees clearly nested within the species “group 9” (clade R) sharing with them a trophic link with Asteraceae. Our phylogenetic hypotheses also allowed for the identification of two main evolutionary lineages within “group 9”, on one hand the species belonging in the subgenera Anopachys [excluding Ch. aurichalcea (Gebler, 1825)], Chalcoidea and Pezocrosita, all of them feeding on closely related plant species in the family Asteraceae in the tribe Anthemideae (Achillea, Artemisia, Santolina, Tanacetum; Cobos 1953, Jolivet and Petitpierre 1976, Bieńkowski 2010, 2011, clade S) and sharing a karyotype of 2n=40 [cytogenetic data for Ch. eurina (Frivaldszky, 1883) and Ch. convexicollis are not available], and on the other hand the species in the subgenera Allochrysolina with a male karyotype 2n=42 (Petitpierre 1999) and feeding on closely related Asteraceae host plants in the subtribe Centaureinae (Centaurea, Mantisalca, Jolivet and Petitpierre 1976, Bourdonné and Doguet 1991). In turn, the species in “group 9” were recovered as the sister lineage of the species classified in the “group 10” (subgenus Hypericia; clade Q), thus contradicting Bourdonné and Doguet’s (1991) view that the subgenus Hypericia is so differentiated from the remainder of the Chrysolina subgenera that it deserves a generic status. Recognition of the genus Hypericia would render Chrysolina paraphyletic. Also regarding this lineage, Pasteels et al. (2003) found that the subgenera Hypericia, Chalcoidea and Sphaeromela are the only Chrysomelinae leaf beetles producing polyoxygenated steroids as defensive toxins, and suggested that they could be raised to a distinct genus. However, our inferred topologies were not compatible with this hypothesis, although the AU test could not reject the constrained monophyly of Chalcoidea and Hypericia. On the other hand, the well-supported and resolved clade T allowed for the identification of the phylogenetic relationships among four of the systematics groups defined by Bourdonné and Doguet (1991), and also expanded our knowledge regarding the systematic position of four subgenera not included before in any phylogenetic analysis. The species in the subgenera Chrysolinas. str. were placed in the “group 2” based on their trophic link with the plant family Lamiaceae but our results clearly contradict this association (clade U), agreeing with their unique male karyotype (2n=23; Petitpierre 1975, 1981, 1983). The common ancestry of Colaphodes, Ovosoma Motschulsky, Palaeosticta and Stichoptera Motschulsky demonstrated by Garin et al. (1999) was confirmed here, and in addition we show that the subgenera Allohypericia Bechyné, Arctolina Kontkanen, Pleurosticha Motschulsky and Taeniosticha also belong in this monophyletic lineage. The close relationship between the subgenera Arctolina and Pleurosticha has been previously proposed according to their morphology (Bieńkowski 2004) and their karyological resemblances [2n=26 (Xy_p), Petitpierre and Mikhailov 2009]. In this regard, our study contributes additional evidence confirming their phylogenetic relatedness (clade Y’). The monophyly of the species adapted to the plant family Plantaginaceae (subgenera Palaeosticta, Colaphodes and Ovosoma) could not be rejected, indicating that they could conform to a natural group, thus expanding Bourdonné and Doguet’s (1991) “group 7”. On the other hand, the Stichoptera species of the “group 1” sensuBourdonné and Doguet (1991) were demonstrated to be sister to the morphologically well-defined subgenus Taeniosticha (Bourdonné et al. 2013). Stichoptera species are characterized by their marked asymmetrical karyotypes (Petitpierre 1999) and their affiliation with Lamiaceae and Scrophulariaceae host plants, but unfortunately no data are available regarding the biology and the cytogenetics of the subgenus Taeniosticha to contrast with our molecular results.

The initial stages of the evolutionary history of the genus Chrysolina were closely related to the plant family Lamiaceae (Figure 3, node A), which is in line with the pioneering studies based on the karyology and the ecology of the species (Petitpierre and Segarra 1985, Bourdonné and Doguet 1991) and also on mtDNA sequences (Garin et al. 1999). The inferred ancestral association with Lamiaceae was highly favoured in our analyses compared to the alternative hypotheses, including an original affiliation with the family Asteraceae suggested by Crowson (1981).

The most basal clades in our Chrysolina phylogeny are those living on Lamiaceae. However, the phylogenetic uncertainty affecting this region of the tree prevents us for drawing firm conclusions about the number of lineages that have adapted to this plant family at the early stages of the evolution of the genus. In contrast, our phylogenetic analyses allowed for the identification of a minimum of eight host plant family shifts in the Chrysolina tree, thus indicating that the feeding spectrum of the extant Chrysolina species is the result of frequent and abrupt host shifts in their evolutionary history. While some of these shifts are between plant families belonging to the same order (Lamiaceae, Plantaginaceae, Scrophulariaceae; order Lamiales; APG 2009), others are between distant plant families from different subclasses [shift from families in the subclass Asterids to Hypericaceae (subclass Rosids); APG 2009] or even from more divergent lineages [shifts from Asterids to Ranunculaceae (basal Eudicot); APG 2009]. Three main hypotheses have been proposed concerning the macroevolution of insect–plant associations (Nyman 2010): (i) the ‘cospeciation’ or ‘parallel cladogenesis’ model (Fahrenholz 1913): matching of speciation events between insects and their host plants; (ii) the ‘escape and radiate’ model (Ehrlich and Raven 1964): plants ‘escape‘ from herbivory due to novel defences and radiate, followed by colonization of new insect taxa that then radiate on them; and (iii) the ‘sequential evolution’ model (Jermy 1984): insects have little effect on the speciation of their hosts, whereas the diversification of hosts increases possibilities of ecological speciation in insects. The hypothesis of ‘parallel cladogenesis’ between Chrysolina lineages and their host plant families can be discarded as the temporal origin of the more closely related host plant families recorded for Chrysolina (Lamiaceae and Scrophulariaceae: mrca >65Ma, Bremer et al. 2004) clearly pre-dates the diversification of the Chrysolina lineage itself [mrcaca. 40Ma, (ca. 20Ma excluding the divergent subgenera Atechna), Gómez-Zurita et al. 2007]. Consistently, this pattern of asynchronous diversification has been found among other phytophagous insect groups and their host plants (Lopez-Vaamonde et al. 2006, McKenna et al. 2009). Regarding the ‘escape and radiate’ model, the existence of coincident radiations at a large scale among host families and the Chrysolina lineages is also not possible due to this time lag in their respective origins. Conversely, the ancestral host plant family affiliations inferred for Chrysolina seem to fit better the ‘sequential evolution’ model, as deduced from the continuous host-shifting among pre-existing host families that characterizes the evolution of the genus (Nyman 2010). Indeed, some Chrysolina clades have experienced multiple host shifts from the ancestral affiliation with Lamiaceae. As an example we could cite the case of the preference for Lamiaceae observed in the derived lineages Allohypericia (clade W), Stichoptera (clade Z’), Arctolina and Pleurosticha (clade Y’), which seems to be a back-colonization of this family from ancestors previously adapted to Plantaginaceae. Another case of multiple shifts is illustrated by the transition from Lamiaceae to Asteraceae and then to Apiaceae inferred for the Oreina clade, which is highly consistent with previous results based on allozyme data (Dobler et al. 1996) and mtDNA sequences (Hsiao and Pasteels 1999). In addition, convergent shifts to the same host plant family in different Chrysolina lineages have also occurred (Apiaceae: clades G’ and K’’; Asteraceae: clades G’, R, W’ and Y’, Ch. sturmi (Westhoff, 1882) and Ch. cerealis cyaneoaurata; Ranunculaceae: Ch. costalis (Olivier, 1807) and Ch. silvatica (Gebler, 1823); Scrophulariaceae: clade Z’ and Ch. sturmi), thus suggesting the existence of evolutionary constraints in host shifts as it has been described in other phytophagous insects including Chrysomelidae (Futuyma et al. 1993, Futuyma and Mitter 1996, Janz et al. 2001, Nosil 2002). A possible explanation for the continuous and convergent shifts among restricted sets of plant taxa is the phytochemical similarity among the alternative hosts (Feeny 1992), and indeed this seems to be the underlying mechanism in other herbivorous beetle groups (Becerra 1997, Kergoat et al. 2005). It also has been suggested that convergent shifts may not be independent, in the sense that an ancestral trait allowing the colonisation of a given plant group might have been already present in the insect lineages (Janz and Nylin 2008).

Chrysolina leaf beetles are highly specialized herbivores feeding on a narrow range of host plants (Jolivet and Petitpierre 1976, Bourdonné and Doguet 1991). However, despite the high level of specialization, their diet breadth ranges from species feeding on few plant species from the same genus or family (i.e., monophagous or oligophagous, respectively) to more generalist species exploiting few species but from different plant families (i.e., polyphagous). In this regard, Garin et al. (1999) reported the subgenus Chrysolinas. str. as the only lineage within the genus experiencing a shift to a generalist feeding habit at the plant family level. Now, our expanded taxon sampling coupled with the availability of a more complete host plant record shows that polyphagy is distributed across the Chrysolina tree, although it occurs at a lower frequency than mono- and oligophagy. Moreover, our results suggest that niche widths have varied through time, since some Chrysolina clades include mixtures of species with different levels of diet breadth (clades E, G’, R, U, Y’ and Z’). Oscillations in host range over evolutionary time are thought to play an important role in the diversification of the phytophagous insects (oscillation hypothesis, Janz et al. 2006, Janz and Nylin 2008). Under this model, speciation is driven by successive cycles of expansion of the host-plant range and generation of new species through specialization on different hosts. The oscillations are maintained through the ability to retain essential parts of the genetic “machinery” to utilize ancestral hosts, and therefore the probability of a major host shift seems to be positively influenced by polyphagy (Janz 2011). Our results on Chrysolina are still too preliminary to offer any scenario for the evaluation of this hypothesis. However, as it has been shown here, the evolutionary history of the genus is deeply associated with the occurrence of frequent and abrupt host shifts giving rise to the specialization on a restricted set of divergent host plant taxa, which is consistent with the model predictions. Optimizing niche width on the Chrysolina phylogeny would help in elucidating whether the diet breadth of the extant polyphagous species indeed represent an event of host range expansion from specialized ancestors, and whether polyphagy has been a transitional stage during host shifts. However, ancestral host range reconstruction will require very detailed information on host plant records and a well-resolved phylogeny for all Chrysolina species (Janz and Nylin 2008). In this respect, future research will be directed towards the expansion of the taxonomic sampling and the exploration of additional molecular markers in order to improve phylogenetic resolution. The implementation of DNA-based techniques for the taxonomic identification of the host plants (Jurado-Rivera et al. 2009) would also contribute to our understanding on the evolution of the ecological associations in this large and highly diversified leaf-beetle genus.