Phylogenetic relationships of selected European Ennominae ( Lepidoptera : Geometridae )

This study reports the results of a molecular phylogenetic analysis of thirty three species of Ennominae (Lepidoptera: Geometridae). The aim of this analysis was to determine the phylogenetic affinities of 13 European species not previously studied using these methods. Fragments of seven nuclear genes, elongation factor 1 alpha (EF-1 ), wingless (wgl), isocitrate dehydrogenase (IDH), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), ribosomal protein S5 (RpS5) and expansion segments D1 and D2 of the 28S rRNA gene and fragment of one mitochondrial gene, cytochrome oxidase subunit I (COI), were used. In the analysis using Bayesian phylogenetic inference, original gene sequences of the target species were combined with a larger data matrix of 20 species of Ennominae, used in a previous study (Wahlberg et al., 2010, Mol. Phylogenet. Evol. 55: 929–938). Most notably, the results indicate that the representatives of the genera Cepphis, Plagodis, Pseudopanthera and Selenia form a well-supported monophyletic group which appeared as the sister clade to the rest of the “ennomine” group of tribes. On the other hand, Crocallis and Opisthograptis group together with Ennomos. These results conflict with previous tribal subdivisions of the subfamily pointing to the need to reconsider the concepts of Ennomini and Ourapterygini. Within the tribe Macariini, the genus Macaria appears to be more closely related to Itame (=Speranza) than to Chiasmia clathrata. The emerging phylogenetic tree of Ennominae suggests only a limited phylogenetic inertia in body size making this group a promising target for comparative studies on this central life history trait and its correlates.


INTRODUCTION
The largest subfamily of Geometridae, the Ennominae (about 45% of all Geometridae: Minet & Scoble, 1999), is a well defined and uncontroversial group.However, there is no consensus about the taxonomic affinities of the numerous currently recognized tribes within the subfamily (Heppner, 2003).Within Ennominae, the genera can be divided into the "ennomine" and "boarmiine" groups based on the structure of the cremaster in the pupal stage (Forbes, 1948;Holloway, 1993;Pato ka & Tur ani, 2005;Viidalepp et al., 2007;Wahlberg et al., 2010).Beyond this major subdivision, the relationships among the numerous traditionally recognised tribes of Ennominae have remained largely uncertain (see Holloway, 1993 for a recent morphology-based hypothesis) and await a reassessment using contemporary methods of phylogenetic analysis.
In fact, there still are only few molecular-based studies on the phylogeny of geometrids.The first molecular phylogeny of a sample of geometrid species was published as recently as 2001 (Abraham et al., 2001).To date, there appear to be just three large-scale studies of molecular relationships: the one by Young (2006), based on a sample of mainly Tasmanian species; an analysis of a limited set of Japanese taxa (Yamamoto & Sota, 2007) and the most recent one addressing the evolution of female flightlessness among holarctic Ennominae (Wahlberg et al., 2010).There are, however, an increasing number of more focused taxon-specific molecular studies on geometrids (e.g.Snäll et al., 2007;Viidalepp et al., 2007;Õunap et al., 2008, 2009).
The goal of the present study was to establish phylogenetic affinities of 13 European species of Ennominae not studied earlier using molecular methods.Various lifehistory traits of these species and a number of other geometrids included in previous phylogenetic studies (Snäll et al., 2007;Wahlberg et al., 2010) are currently being studied (Javoiš et al., in prep.).Knowing the position of these taxa in the phylogenetic tree of the subfamily is a precondition for subsequent phylogenetically explicit comparative analyses.In particular, the great variability in body size, characteristic of this subfamily, is most promising in the context of further studies on the evolutionary ecology of body size.Although body size is a trait of central importance in life history studies (e.g.Roff, 1992), the selective forces determining optimal size in insects are poorly understood (e.g.Blanckenhorn, 2000;Tammaru et al., 2002;Gotthard, 2004).Phylogenetic comparative analyses appear most promising in this context.
For the phylogenetic analysis, 7 nuclear gene fragments [partial sequences of elongation factor 1 alpha (EF-1 ), wingless (wgl), isocitrate dehydrogenase (IDH), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), ribosomal protein S5 (RpS5) and expansion segments D1 and D2 of 28S rRNA gene] were sequenced in addition to one mitochondrial gene fragment, a partial sequence of cytochrome oxidase subunit I (COI).The selection of markers follows that of Wahlberg et al. (2010), which allows the original sequence data obtained in this study to be combined with the larger data set used in Wahlberg et al's article.Based on the combined data matrix, a phylogenetic tree was derived using Bayesian phylogenetic inference.In addition to providing necessary information for forthcoming comparative analyses the results help to resolve several taxonomic ambiguities in the Ennominae.

Species studied
In total, 33 Ennomine species were included in the present analysis (Table 1), 20 of which were used earlier to construct a preliminary phylogenetic tree for the Ennominae (Wahlberg et al., 2010) and provide the necessary reference framework.The remaining 13 taxa, however, had not previously been subjected to a rigorous phylogenetic analysis (Table 1).These 13 newly studied species represent both the "ennomine" (tribes Ennomini and Ourapterygini, sensu Viidalepp, 1996) and "boarmiine" branch (Macariini, Abraxini) of the subfamily.The specimens used in this study were collected from Estonia (Table 1), voucher specimens are housed in the Museum of Zoology, University of Tartu.

DNA extraction, PCR and sequencing
DNA was extracted from parts of dried specimens using the High Pure PCR Template Preparation Kit (Roche Diagnostics GmbH, Mannheim, Germany) according to the manufacturer's instructions for isolation of nucleic acids from mammalian tissue.A fragment of one mitochondrial (COI) and five nuclear protein-coding genes (EF-1 , wgl, GAPDH, IDH, RpS5) and two fragments (expansion segments D1 and D2) of one nuclear ribosome gene (28S rRNA) were amplified.Primers for amplifying GAPDH, IDH and RpS5 had universal tails on their 5' ends (see Regier & Shi, 2005;Wahlberg & Wheat, 2008) that allowed us to sequence the respective gene fragments using common sequencing primers T7 Promoter and T3 (Table 2).All other gene fragments were sequenced utilizing the same primers that were used for PCR (Table 2).
PCR was performed in a total volume of 20 µl, with the reaction mixture containing 1X BD Advantage 2 PCR buffer, 1U BD Advantage 2 Polymerase mix (BD Biosciences, San Jose, USA), 0.2 mM dNTP (Fermentas, Vilnius, Lithuania), 4 pmol of primers and 20-80 ng of purified genomic DNA.PCR was carried out in a Biometra T1 Thermocycler (Biometra, Göttingen, Germany), its conditions were an initial denaturation at 94°C for 2 min followed by 35-40 cycles of 30 s at 94°C, 30 s at 50-63°C depending on the gene fragment and the primer pair (Table 2), and 1 min at 68°C, and a final extension at 68°C for 7 min.PCR products were visualised on a 1.6% agarose gel and 10 µl of the PCR solution was treated with fast alkaline phosphatase and exonuclease I (Fermentas).DNA cycle sequencing was performed in a total volume of 10 µl using the Big Dye Terminator v.3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, USA).Cycling conditions were: initial denaturation for 1 min at 96°C followed by 25 cycles of 10 s at 95°C, 15 s at 47-58°C and 4 min at 60°C.Both DNA strands were sequenced using 1.6 pmol of primers, and sequences were resolved using 3730xl DNA Analyzer (Applied Biosystems).For some taxa, PCR amplification products and sequence data could not be obtained for all gene fragments (see Table 1).

Phylogenetic analysis
Consensus sequences were created in Consed (Gordon et al., 1998) using sequence data from both DNA strands.Sequences were double-checked by eye, edited in BioEdit (Hall, 1999) and aligned in ClustalW (Thompson et al., 1994) using the default settings.The protein coding genes were trivial to align, with no indel events inferred except in the cases of wgl and RpS5, as one 3 bp deletion was found in both two genes of Plagodis pulveraria.The lengths of the fragments of protein-coding genes used in the phylogenetic analysis were 667 bp for COI, 1008 bp for EF-1 , 389 bp for wgl, 691 bp for GAPDH, 699 bp for IDH and 617 bp for RpS5.Alignment of the 28S fragments proved difficult, as had been shown earlier by Snäll et al. (2007), Õunap et al. (2008) and Wahlberg et al. (2010).The length of successfully sequenced fragments of D1 varied from 281-283 bp and the length of aligned data matrix was 283 bp.Three positions with indels were excluded from data matrix resulting in a 280 bp indel-free matrix.The alignment of D2 was more complicated, as the length of successfully sequenced fragments varied from 402-440 bp and the length of aligned data matrix was 450 bp.Of those positions 83 contained indels and were removed, resulting in a 367 bp indel-free data matrix.The total length of the combined data matrix was 4718 bp.GenBank accession numbers of both originally contributed and downloaded sequences are provided in Table 1.
All previous molecular phylogenetic studies have resolved Ennominae as a monophyletic entity (Abraham et al., 2001;Young, 2006;Yamamoto & Sota, 2007;Wahlberg et al., 2010).Therefore, monophyly of this group was assumed and as appropriate analytical methods are now available, no outgroups were used in this study in order to avoid possible long-branch attraction effects.
For phylogenetic analysis, data were divided into three partitions.First, COI as the single mitochondrial protein-coding gene was treated as a separate partition.Second, as expansion segments D1 and D2 of 28S are different regions of the same rRNA gene and therefore share a similar evolutionary history, data from these gene fragments were concatenated and treated as a single partition.Third, as sequencing nuclear protein-coding genes proved difficult and failed on a number of occasions (Table 1) the data from all respective gene fragments were treated as one partition and failed regions were defined as missing characters.Modeltest 3.06 (Posada & Crandall, 1998) was used in PAUP*4.0b10(Swofford, 1998) to search for the model of DNA substitution that best fitted the data for each partition.Beast v1.5.4 (Drummond & Rambaut, 2007) was used for the Bayesian estimation of phylogeny, implementing GTR + I + model selected by Modeltest for each of the tree partitions and using relaxed molecular clock allowing branch lengths to vary according to an uncorrelated lognormal distribution (Drummond et al., 2006).To obtain an ultrametric tree, the age of the Ennominae was calibrated according to Wahlberg et al. (2010), i.e. 37.5 million years with a standard deviation of 6.5 million years.The TMRCA of Hypomecis + Ematurga clade was given a uniform prior distribution from 4 to 10 million years according to the same article.The tree prior was set to the Birth-Death process and all other priors were left to defaults in Beast.First, Bayesian MCMC was run over 30 million generations, sampling every 1000th generation.Thereafter, suggestions by Beast for improving the analysis were taken into account and four further MCMC runs (one for 50 million and three for 30 million generations, sampling every 1000th generation) were performed.The results of these four analyses were combined and inspected with Tracer v1.5.The first 10% of the sampled trees were discarded as "burn-in" from each of the three analyses and the remaining Macaria wauaria Saare, 2004 Itame loricaria (Eversmann, 1837)  Parainen, 2002 Biston strataria (Hufnagel, 1767) Lomaspilis marginata (L., 1758) Ruohomäki FIN, Hanko, 2004 Calospilos sylvata (Scopoli, 1763) --HQ340216 HQ340174 HQ340234 HQ340207 HQ340197 HQ340186 J. Javoiš EST, Laelatu, 2006 Abraxas grossulariata  Viidalepp (1996) with the exception of using the name Boarmiini instead of Cleorini (following Holloway, 1993).Taxonomy of the generic and species levels was adopted from Müller, 1996.90% of trees were combined together with LogCombiner v1.5.4.Subsequently, a final tree file was created on the basis of the saved trees using TreeAnnotator v1.5.4 and the results were visualized with FigTree v1.3.1.
In this analysis the 95% credibility intervals for divergence time estimates of each node appeared to be very wide and thus not informative.Therefore, we chose to drop the information on the node ages from the following discussion and concentrated only on the phylogenetic relationships between the studied taxa.

RESULTS AND DISCUSSION
Adding new species to the previously derived phylogenetic tree of Ennominae (Wahlberg et al., 2010) did not result in any changes in the topology of the previously resolved parts of the tree (Fig. 1).In particular, the basal dichotomy into the "ennomine" and "boarmiine" groups of genera, first rigorously shown by Wahlberg et al. (2010), remained valid.No surprise, Lomographa temerata and Abraxas grossulariata appeared as sisters to L. bimaculata and Calospilos sylvata, respectively.Similarly, the placement of the three Macaria species as sisters to Itame was expected (Scoble & Krüger, 2002).Nevertheless, the data presented conflict with the frequently assumed close relationship between Chiasmia clathrata and Macaria spp.ences in the appearance and ecology of these moths, they are frequently treated as congeneric in European literature (though not so in the most recent publications: Müller, 1996, Scoble & Krüger, 2002), which plausibly follows Wehrli's (1939Wehrli's ( -1953) ) combining Chiasmia with Semiothisa as a subgenus.Moreover, the data presented here show that the recent suggestion to transfer wauaria from Itame (or Speranza, following Ferguson, 2008) to Macaria (e.g.Müller, 1996) is not justified, as it would make both Itame and Macaria paraphyletic.
In the "ennomine" branch, the genera Cepphis, Plagodis, Pseudopanthera and Selenia form a well supported monophyletic entity, which does not include any of the species from Wahlberg et al.'s (2010) data set.Close relationships between these genera is not surprising because all were included in the tribe Ennomini by Herbulot (1961Herbulot ( -1962)).On the other hand, this study indicates that Ennomos is not closely related to the four above men-tioned genera, but forms a common clade with Crocallis and Opisthograptis.The two latter genera are, however, also placed in Ennomini by Herbulot (1961Herbulot ( -1962) ) but the phylogenetic tree revealed by this study shows that Ennomini sensu Herbulot is not justified as it would be paraphyletic with respect to several other currently recognised tribes (Fig. 1).Moreover, the way Viidalepp (1996) divides these genera between Ennomini and Ourapterygini (Table 1) is also not supported.In contrast, the clade consisting of Cepphis, Plagodis, Pseudopanthera and Selenia matches the concept of Anagogini of Forbes (1948) -subsumed to Hypochrosini by Holloway (1993) -pointing to the conclusion that the idea of "reviving" this tribe should deserve more attention from taxonomists.
In the "ennomine" branch of the subfamily, the present analysis revealed 4 to 6 monophyletic groups, which could be considered to represent different tribes.Even if these clades received reasonably high support and the topology presented here does not conflict with Wahlberg et al. (2010), the currently available studies include only a fraction of the total diversity of Ennominae.Any suggestions for taxonomic rearrangements are therefore clearly premature.Nevertheless, phylogenetic relationships among a sufficient number of north European Ennominae are now known well enough to facilitate using this information in comparative studies.Notably, body size appears to be an evolutionarily plastic trait in this group.For instance, the similarly looking stout-bodied moths in the genera Colotois, Selenia, Ennomos and Crocallis are not closely related to each other, but all have small and slender-bodied relatives (Opisthograptis, Alsophila, Cepphis).Multiple independent evolutionary changes in body size should create a favourable situation for studies on morphological and ecological correlates of this important life history trait.

Fig. 1 .
Fig. 1.Bayesian phylogenetic tree (GTR + I + model) of selected European Ennominae based on a 4718 bp combined sequence of COI, 28S, EF-1 , wgl, GAPDH, IDH and RpS5 sequences.The numbers above or below branches are Bayesian posterior probabilities.The original contribution is highlighted: the species first examined using molecular methods in the current study are indicated in bold.

TABLE 1 .
Details of the specimens used in the molecular analysis.Collecting site (EST -Estonia, FIN -Finland) and date, collector's name and Gen-Bank accession numbers for 28S D1, 28S D2, EF-1 , wingless, COI, GAPDH, RpS5 and IDH sequences of the studied specimens are indicated.Tribal classification follows Despite considerable differ-270* species with gene sequences first determined in the present study; -not available;1 Viidalepp et al., 2007; 2 Wahlberg et al., 2010.