Molecular phylogeny of the Aphidiinae ( Hymenoptera : Braconidae ) based on DNA sequences of 16 S rRNA , 18 S rDNA and ATPase 6 genes

Phylogenetic relationships among 16 genera of the subfamily Aphidiinae (Hymenoptera: Braconidae) were investigated using sequence data from three genes: the mitochondrial large ribosomal subunit (16S), 18S ribosomal DNA and mitochondrial ATPase 6. All sequences were downloaded from the GenBank database. A total of 2775 base pairs of aligned sequence were obtained per species from these three genes. The results support the existence of three-tribes: Ephedrini, Praini and Aphidiini, with the Ephedrini occupying the basal position; Aphidiini could be further subdivided into three subtribes: Monoctonina, Trioxina and Aphidiina. The genus Aphidius is a paraphyletic group. The taxonomic status of the subfamily Aphidiinae within the Braconidae is probably closer to the non-cyclostome than the cyclostome subfamilies. 133 * Corresponding author; e-mail: xxchen@zju.edu.cn addition, whether there are three or four main clades within this subfamily was tested and the phylogenetic trees inferred here and those based on other characters compared. MATERIAL AND METHODS


INTRODUCTION
Aphidiinae is one of the subfamilies of the family Braconidae (Insecta: Hymenoptera) with approximately 50 genera and 400 species (Mackauer & Starý, 1967;Starý, 1988).They are exclusively solitary endoparasitoids of aphids.Several species have been used successfully in biological control programs throughout the world (Carver, 1989).Because of their importance as biological control agents, many aspects of their biology have been studied (Starý, 1970).
Aphidiines have often been treated as a separate family, the Aphidiidae, because of their specialization on aphids, the presence of a flexible suture between the second and third mesosomal tergites and reduced wing venation.However, recent phylogenetic studies have shown aphidiines to be a lineage within the Braconidae (Quicke & van Achterberg, 1990, 1992;Wharton et al., 1992), but it still remains unclear that to which of the many braconid subfamilies the aphidiines are most closely related.
Although the Aphidiinae is a coherent group defined by a number of synapomorphies, significant differences exist in morphology, biology and behaviour among tribes, genera and species, and the phylogenetic relationships within this subfamily remain to be resolved.Several phylogenies, based on adult and larval morphology, embryology and DNA sequences, have been proposed for Aphidiinae (Mackauer, 1961;Tremblay, 1967;Tremblay & Calvert, 1971;Chou, 1984;Gärdenfors, 1986;Quicke & van Achterberg, 1990, 1992;Whitfield, 1992;Belshaw & Quicke, 1997;Dowton et al., 1998;Smith et al., 1999;Kambhampati et al., 2000;Sanchis et al., 2000).The most widely accepted classification scheme for Aphidiinae is that of Mackauer (1961) who divided the subfamily into four tribes: Aclitrini, Aphidiini, Ephedrini and Praini.The Aphidiini is the largest of the four tribes, includes the majority of genera and species, and is further subdivided into two subtribes, Aphidiina and Trioxina.Because the Aclitini is poorly represented and hardly available (Kmabhampati et al., 2000 are the only authors to have included them in a molecular analysis) most authors accept the existence of four natural groups: Ephedrini, Praini, Trioxini and Aphidiini.Trioxini and Aphidiini are treated as independent tribes, forming a four-tribe hypothesis (Ephedrini + (Praini + (Trioxini + Aphidiini))) (Belshaw & Quicke, 1997) or they are placed in the same tribe, resulting in a three-tribes hypothesis: Ephedrini, Praini and Aphidiini (Smith et al., 1999;Sanchis et al., 2000).However, Sanchis et al. (2000) claimed that their results favour either the three-tribes system or a new classification of at least five tribes (Ephedrini, Praini, Monoctonini, Trioxini and Aphidiini).
Therefore, the aim of our study was to determine which tribe might be basal within the Aphidiinae.This was done using three different molecular markers, the mitochondrial ATPase 6, the ribosomal 18S rDNA and the mitochondrial 16S rRNA genes, whose sequences for the taxa studied are already available in the GenBank database.In addition, whether there are three or four main clades within this subfamily was tested and the phylogenetic trees inferred here and those based on other characters compared.

Sampling of taxa
Twenty three species belonging to 16 genera were examined in this study.The species are listed in Table 1 and the arrangement of the tribes is based on morphological and biological characters.DNA sequences of the three genes used in this study were downloaded from the GenBank database with accession numbers listed in Table 1.

Outgroup selection
Three outgroups were selected for the phylogenetic analysis: the genera Jarra (Doryctinae) and Mesostoa (Mesostoinae) of the cyclostome lineage and genus Schizoprymnus (Helconinae) of the non-cyclostome lineage.Helconinae is widely recognized as a sister group of the Aphidiinae, and the Doryctinae and Mesostoinae are postulated to occupy a relatively basal position within Braconidae (Quicke & van Achterberg, 1990).

Sequence alignments
Sequences were aligned using CLUSTAL X version 1.81 (Thompsom et al., 1997) with default parameters.The manual alignment was followed to remove some regions with high variation.The lengths of the resulting alignments of 18S rDNA ranged between 1752 to 1820 bp, of 16S rRNA between 394 to 486 bp and of ATPase 6 between 618 to 624 bp.

Phylogenetic analysis
Following alignment, three different methods of phylogenetic analyses were performed using PAUP* 4.0 (beta 10 version) (Swofford, 2001).First, maximum parsimony (MP) was used to find the most parsimonious tree(s), and heuristic parsimony search (Hillis et al., 1996) were performed using 100 replicates 134 1 Sequences from Kambhampati et al., 2000; 2 Sequences from Sanchis et al., 2000; 3 Sequences from Dowton et al., 1998; 4  of random addition sequences and TBR option for branch swapping followed by additional rounds of branch swapping on the resulting trees with restriction on the number of trees to one.Each base was treated as an unordered character of equal weight, with gaps treated as missing data.Where more than one most parsimonious tree was found, a strict consensus tree was calculated.Downweighting transitions or treating gaps as a fifth base did not markedly affect the results.Statistical support for each node was evaluated by bootstrap analysis (Felsemstein, 1985) with 1000 replications.Second, a distance-based method based on the neighbor-joining algorithm (NJ) with Tamura-Nei correction (Saitou & Nei, 1987;Tamura & Nei, 1993) was used for obtaining a minimum-evolution tree and bootstrapping evaluation of each node was performed as above.Third, maximum likelihood (ML) trees were generated under the HKY85 model, using base frequencies estimated by PAUP, default number of substitution type (2, HKY85 variant) and transition/transversion ratio (2).Heuristic search were used with 100 replicates of random addition sequence and TBR branch swapping.Bootstrap analysis was performed with 100 replicates.The Bayesian approach to phylogenetic reconstruction (Yang & Rannala, 1997;Huelsenbeck et al., 2001) was implemented using MRBAYES 3.0B4 (Huelsenbeck & Ronquist, 2001).Each run was performed using default staring parameters and comprised 5 000 000 generations.Bayesian pos-terior probabilities (Pbay) were calculated from majority-rule consensus of trees sampled every 100 generations once the Markov chain reached stationary (determined by empirical checking of likelihood values).

RESULTS AND DISCUSSION
We tested alignment using the Clustal X program with different gap opening and gap extension values, and resulted in different length of aligned sequences.This result is identical with that of Morrison & Ellis (1997).They conclude that the multiple alignments, using different procedures, vary greatly in length and those produced using the Clustal W program with different gap weights are at least as different from each other as those produced by different alignment algorithms (Morrison & Ellis, 1997).Because the default parameters in version 1.81 (gap opening 15, gap extension 6.66) were optimized using the balibase multiple alignment in the 142 alignment test in balibase (J.Thompson, pers. comm.), we used the alignments with default parameters for the analysis presented here.
The trees resulting from PAUP* and MrBayes analyses are presented in Figs 1-4.We also show the bootstrap values and Bayesian posterior probabilities obtained from the identical analysis.
The topology of all trees inferred from molecular data using different methods was similar.They confirmed the existence of two of the four traditionally accepted tribes, Ephedrini and Praini, but questioned the existence of the Trioxini and Aphidiini s. str.Our analyses support the three-tribe hypothesis: ((Ephedini + Praini) + Aphidiini s. lat.), as do the results of Smith et al. (1999) and Sanchis et al. (2000).Because our analyses support the monophyletic nature of the tribe Aphidiini s. lat.(tribal defintion of three-tribe system) we do not accept the classification system of five tribes proposed by Sanchis et al. (2000) and merge the two tribes, Trioxini and Aphidiini s. str., into one tribe -Aphidiini s. lat.
As shown in the figures the clade Praini seems to be the sister group of the Aphidiini, with the Ephedrini occupying the basal position, which is supported by the results of Belshaw & Quicke (1997) and Sanchis et al. (2000) based on molecular data, and Mackauer (1961) and Gärdenfors (1986) based on adult morphology, but not by  those of Dowton et al. (1998), Smith et al. (1999) and Kambhampati et al. (2000).Belshaw & Quicke (1997), using sequence data from three genes (elongation factor-1 , cytochrome b and the second expansion segment of the 28S rRNA), suggested the following tribal relationships: (Ephedrini + (Praini + (Aphidiini + Trioxini))), while Sanchis et al. (2000), using only the 18S rDNA gene, established the basal position of the tribe Ephedrini within the subfamily.Kambhampati et al. (2000) considered that the basal lineage of Aphidiinae was Aclitus (Aclitini) and that the Praini was basal relative to Ephedrini based on the sequence data of the 16S rRNA gene alone while similar topologies were inferred from a combined analysis that included DNA sequences of 16S rRNA, NADH1 dehydrogenase and 28S rRNA (Aclitus was not included), resulting in that Praini was basal.At the same time, they also pointed out that Aclitus possesses both presumed synapomorphic and plesiomorphic characters with Aphidiina and Trioxina on the one hand, and its phylogenetic position is further complicated by the presence of characters that are related to its parasitization of rootfeeding aphids on the other hand (Takada & Shiga, 1974).Therefore, the proposed basal position of Aclitus requires further study (Kambhampati et al., 2000).
The tribe Aphidiini s. lat., can be further subdivided into three subtribes, Monoctonina (containing the genus Monoctonia), Trioxina and Aphidiina, with the genus Monoctonia occupying a basal clade within the tribe, as proposed by Sanchis et al. (2000).Within the tribe Xenostigmus has a close relationship with Aphidius (Figs 1-4) while it was closer to Protaphidius in the trees produced by Sanchis et al. (2000), suggesting that Xenostigmus might be a member of the subtribe Aphidiina, and not the Protaphidina as defined by Sanchis et al. (2000).Our results also indicate that Aphidius is a paraphyletic group, as suggested by Smith et al. (1999) and Sanchis et al. (2000), but the nature of this genus needs to be validated.

Fig. 4 .
Fig. 4. Phylogeny of the Aphidiinae based on 3 genes using MrBayes.Jarra, Mesostoa and Schizoprymnus were used as outgroups.Numbers at nodes are Bayesian posterior probabilities.