Using internally transcribed spacer 2 sequences to re-examine the taxonomic status of several cryptic species of Trichogramma ( Hymenoptera : Trichogrammatidae )

Identification and classification of Trichogramma species are difficult because of their small size. The closely related Trichogramma species or cryptic species complexes are extremely difficult or impossible to distinguish morphologically. For example, T. confusum and T. maidis have long been considered to be subspecies of T. chilonis and T. brassicae, respectively (Lin, 1994). Based on the examination of more than 10 thousand specimens on slides, mostly collected by sweeping from all over China, Lin (1994) described 39 genera and 128 species of the total of 142 species in 41 genera in China. Trichogramma Westwood is the type genus of the Trichogrammatidae family, most species of which parasitize the eggs of lepidopterous pests. The larvae are morphologically indistinguishable, and the adults very difficult to differentiate. The identity of these egg parasitoids was based almost exclusively on the morphology of the male genitalia (Nagarkatti & Nagaraja, 1971; Nagarkatti & Nagaraja, 1977; Sorokina, 1993), but male wasps occur in very low proportions or are absent in natural populations (Aeschlimann, 1990). While the taxonomy of the genus Trichogramma is still being studied (Pinto, 1992; Pinto & Stouthamer, 1994; Neto & Pintureau, 1995), mass releases of these egg parasitoids for the biological control of crop pests have gained increasing attention worldwide. Trichogramma species are used to control over 20 pest species on corn, cotton, rice, sugarcane, vegetables and fruit trees. We used not only the native species, T. dendrolimi, T. ostriniae, T. evanscens and T. confusum, but also commercially available Trichogramma such as T. brassicae supplied by Biocare (Einbeck, Germany) and T. maidis produced by BASF (Valbonne, France) to control the European corn borer, Ostrinia nubilalis Hübner (Lepidoptera: Pyralidae) in the corn fields of HengShui in HeBei province, China. In Germany and France, T. brassicae is considered to be a strain of T. maidis (Hassan & Zhang, 2001). With the extensive application of Trichogramma for biocontrol worldwide in the middle of 20 century, the identification of species and strains of Trichogramma became important (Smith & Hubbes, 1986). Quednau (1960) pointed out that only individuals reared in the same host at the same temperature could be differentiated. Lack of type specimens is a key factor affecting the accurate classification of Trichogramma species (Pang, 1999). For example, there is only one incomplete female specimen of T. evanescens. Pintureau & Voegele (1980) re-described T. evanescens when they described T. maidis Eur. J. Entomol. 101: 347–358, 2004 ISSN 1210-5759


INTRODUCTION
Identification and classification of Trichogramma species are difficult because of their small size.The closely related Trichogramma species or cryptic species complexes are extremely difficult or impossible to distinguish morphologically.For example, T. confusum and T. maidis have long been considered to be subspecies of T. chilonis and T. brassicae, respectively (Lin, 1994).Based on the examination of more than 10 thousand specimens on slides, mostly collected by sweeping from all over China, Lin (1994) described 39 genera and 128 species of the total of 142 species in 41 genera in China.
Trichogramma Westwood is the type genus of the Trichogrammatidae family, most species of which parasitize the eggs of lepidopterous pests.The larvae are morphologically indistinguishable, and the adults very difficult to differentiate.The identity of these egg parasitoids was based almost exclusively on the morphology of the male genitalia (Nagarkatti & Nagaraja, 1971;Nagarkatti & Nagaraja, 1977;Sorokina, 1993), but male wasps occur in very low proportions or are absent in natural populations (Aeschlimann, 1990).While the taxonomy of the genus Trichogramma is still being studied (Pinto, 1992;Pinto & Stouthamer, 1994;Neto & Pintureau, 1995), mass releases of these egg parasitoids for the biological control of crop pests have gained increasing attention worldwide.Trichogramma species are used to control over 20 pest species on corn, cotton, rice, sugarcane, vegetables and fruit trees.We used not only the native species, T. dendrolimi, T. ostriniae, T. evanscens and T. confusum, but also commercially available Trichogramma such as T. brassicae supplied by Biocare (Einbeck, Germany) and T. maidis produced by BASF (Valbonne, France) to control the European corn borer, Ostrinia nubilalis Hübner (Lepidoptera: Pyralidae) in the corn fields of HengShui in HeBei province, China.In Germany and France, T. brassicae is considered to be a strain of T. maidis (Hassan & Zhang, 2001).
With the extensive application of Trichogramma for biocontrol worldwide in the middle of 20 th century, the identification of species and strains of Trichogramma became important (Smith & Hubbes, 1986).Quednau (1960) pointed out that only individuals reared in the same host at the same temperature could be differentiated.Lack of type specimens is a key factor affecting the accurate classification of Trichogramma species (Pang, 1999).For example, there is only one incomplete female specimen of T. evanescens.Pintureau & Voegele (1980) re-described T. evanescens when they described T. maidis and found that the previously described T. evanescens were actually T. maidis.Pintureau (1987) classified T. maidis as T. brassicae and Lin (1994) accepted this nomenclature.T. chilonis was first described by Ishii (1941) but there is no holotypes for verification.Nagarkatti & Nagaraja (1979) chose a male specimen mounted on a glass slide from the syntypes of T. chilonis described by Ishii (1941) as its lectotype.Meanwhile they classified T. chilonis as T. confusum (Viggiani, 1976).Lin (1994) agreed with this classification.However, T. confusum is common in China and it is essential to reconsider its taxonomic status according to Pang (1999).Viggiani's description of T. confusum was very simple and there are no holotypes for comparison, but there is a detailed figure of the male genitalia for reference.It seems incorrect to classify T. confusum as T. chilonis on the basis of this figure (Pang, 1999).
Many methods were used to discriminate sibling species of Trichogramma in addition to morphological comparisons (Pinto et al., 1997), such as allozyme analyses (Pinto et al., 1992(Pinto et al., , 1993;;Pintureau, 1993) and reproductive compatibility tests (Pinto et al., 1991;Stouthamer et al., 1996Stouthamer et al., , 2000a, b), b).Recently closely related or cryptic species were characterized using DNA-based methods (Landry et al., 1993;Vanlerberghe-Masutti, 1994;Sappal et al., 1995;Landais et al., 2000).Ribosomal DNA (rDNA) consists of several regions (genes and spacers) that evolve at different rates, among which the internal non-coding transcribed spacer (ITS) region usually evolves faster than the coding regions (Hoy, 1994).Many of the phylogenetic relationships between Trichogramma species deduced from ITS2 sequences were recorded by previous studies (Orrego & Silva, 1993;van Kan et al., 1996van Kan et al., , 1997;;Pinto et al., 1997;Schilthuizen & Stouthamer, 1997;Stouthamer et al., 1999;Chang et al., 2001;Pinto et al., 2002), which showed that the DNA sequence of the internally transcribed spacer (ITS2) of Trichogramma wasps could be used for species identification.Consistent differences occur among species, whereas the spacer sequences show little variation within species.As ITS2 sequences can be used to identify cryptic species, we used them to distinguish the proposed cryptic species of Trichogramma.

Insects
We established 12 iso-female lines, six from different Trichogramma species and six from geographical populations of another species (Table 1).T. brassicae, T. maidis and T. embryophagum (Hartig) were identified and provided by Sherif A. Hassen of the Federal Biological Research Center for Agriculture and Forestry, Institute for Biological Control, Heinrichstr, Darmstadt, Germany.T. brassicae was purchased from Biocare (Einbeck, Germany) and T. maidis from BASF (Valbonne, France).Six geographical populations of T. dendrolimi were collected from noctuid eggs on corn, cotton, rice or fruit trees in China.We selected T. dendrolimi for a within-species between-population study because the largest number of collections are available for this species, and it has the largest geographical distribution in China.The lines of T. brassicae, T. maidis and T. embryophagum were maintained in the Department of Entomology, China Agricultural University in Beijing.Other lines were maintained at the Biological Control Centre, Beijing Academy of Agriculture and Forestry Sciences, Beijing, China, by rearing T. chilonis, T. confusum and T. dendrolimi in the eggs of tussore worm, Antheraea pernyi Guérin-Méneville and the other lines in the eggs of the rice moth, Corcyra cephalonica (Stainton), at 20-25°C, RH 75-80% and 16L : 8D.Individual rearing was used to avoid linebreeding.Individual neonate wasps from different cultures were placed in sterile 1.5 ml tubes at -80°C for further analysis.Meanwhile, 10 freshly emerged wasps from each line were soaked in acetic acid prior to morphological identification.Trichogramma specimens were identified using the procedure of Lin (1994).

PCR amplification of ITS2, cloning and sequencing
PCR was performed in 50 µl reaction volumes using a Hybaid thermocycler (Fisher Scientific Pte Ltd., Singapore) with 5 µl (10×) PCR buffer, 0.8 µl dNTP mixture (each in a 10 mM concentration), 0.5 µl forward and reverse primers (each in 0.25 µM), 2 µl genomic DNA, 0.2 µl Taq DNA polymerase (Gib-coBRL, Eggenstein, Germany, 5U/µl), and 41 µl sterile water.The ITS2 region was amplified using the following primers: forward, 5'-TTCTCGCATCGATGAAGAACG-3' (ITSN2) located in the 5.8S rDNA; reverse, 5'-TCCTCCGCTTATTGATA TGC-3' (ITSB) located in the 28S rDNA (Amornsak et al., 1998).The PCR cycling program was 3 min at 95°C followed by 35 cycles of 1 min at 94°C, 1 min at 53°C, and 1 min at 72°C with 7 min at 72°C after the last cycle.PCR products were subjected to electrophoresis on a 1.5% (w/v) agarose gel.Gels were stained with ethidium bromide.Molecular weight standards (100 bp DNA ladder) were run along with the samples for reference.The target bands (approximately 600~650 bp) were then excised from the agarose gel and recovered using a QIAquick ® DNA Purification kit (Qiagen, Hilden, Germany), and 4 µl of the final eluted solution (30 µl) were ligated into pGEM-T Vector System I (Promega, Madison, USA) following the protocol provided by the manufacturer.The ligated products were then transformed into E. coli DH5α competent cells (Life Technologies, Rockville, USA).The colonies containing an insert of correct size were checked by PCR using the same primers as described above.One to 3 positive clones from each line were selected and sent for sequencing (Sangon, Shanghai, China).To check the accuracy of the automatic sequencer, some clones were sequenced in two directions.Clones from different individuals of the same line were sequenced for a comparative analysis of the between individual (within line) sequence variation.

Alignment and sequence divergence analyses
The ITS2 regions sequenced in this study and additional 67 ITS2 sequences obtained from GenBank, representing a wide variety of Trichogramma species, were aligned in ClustalW (1.82) (Higgins et al., 1994).The sequences from GenBank are all previously published, except the sequence for T. confusum con_GZ (GenBank accession no.AY244461) (Table 2).This unpublished ITS2 sequence comes from an important geographical population of T. confusum in southern China.A fast pairwise alignment algorithm was chosen for global multiple alignment of ITS2 sequences.The alignment is progressive and considers the sequence redundancy.DNA Identity Matrix (Unitary Matrix) was selected to generate the alignment, which creates a positive score for a match, and a score of -10000 for a mismatch.The penalty for opening a gap is 10; the penalty for extending a gap is 0.05, and the gap separation penalty is 8.The end-unaligned sequences of the multiple alignments were carefully trimmed.
Sequence divergence analyses at the levels of interspecies and intraspecies were conducted using MEGA version 2.1 (Kumar et al., 2001).For ITS2 sequence divergence analysis, a total of 27 groups were set up from 79 Trichogramma taxa (12 identified in this study and 67 retrieved from GenBank), i.e. 6 strains of T. dendrolimi, 11 of T. deion, 6 of T. pretiosum), 5 of T. platneri, 4 of T. bourarachae, 4 of T. evanescens, 4 of T. cordubensis, 3 of T. alpha, 3 of T. turkestanica, 3 of T. kaykai, 3 of T. californicum, 3 of T. minutum, 2 of T. pratti, 2 of T. itsybitsi, 2 of T. brassicae , 2 of T. aurosum, 2 of T. sathon, 2 of T. sibericum, 2 of T. confusum, 2 of T. exiguum, 2 of T. cacoeciae, 1 of T. maidis, 1 of T. chilonis, 1 of T. ostriniae, 1 of T. embryophagum, 1 of T. oleae, 1 of T. nubilale.The ITS2 sequence of Nasonia vitripennis (GenBank accession no.U02960) (Camp-bell et al., 1993), a species of Pteromalidae in the same superfamily Chalcidoidea as Trichogrammatidae, and the ITS2 of Uscana semifumipennis (GenBank accession no.U74608) (Doutt & Viggiani, 1968), a member of a related genus, were incorporated into the 79 Trichogramma ITS2 sequences for net between groups distance analysis.We chose 7 groups from the 27 groups for within group analyses, because each of them contained at least 4 members (strains), the largest numbers available.Net Between Groups and Within Groups methods in MEGA were used to compute average distances.The net average distance between two groups is given by: where dXY is the average distance between groups X and Y, and dX and dY are the mean within-group distances.For each group, an arithmetic average is computed for all valid pairwise comparison.Distance algorithm Kimura 2-parameter (Kimura, 1980) in MEGA was used to compute genetic distances.The complete-deletion option and pairwise-deletion option were alternatively used.For the first option, sites containing missing data or alignment gaps are removed before the analysis begins; for the latter, sites containing missing data or alignment gaps are removed as the need arises during the analysis.Both transitions and transversions were included in the analyses, assuming that the substitution rates do not vary among sites.

Phylogenetic analyses
Phylogenetic analyses were based on the sequence alignments constructed by ClustalW with options and parameters as described above.Because different tree-building algorithms make different evolutionary assumptions, aligned sequences were evaluated by parsimony, maximum-likelihood and neighbour-joining methods.For parsimony, the branch-andbound method of DNA parsimony algorithm, version 3.572c of PHYLIP (Felsenstein, 1993) was used.Bootstrapping was performed with the heuristic search option for 1000 replications.To construct maximum-likelihood trees, the fastDNAml program of PHYLIP (Olsen et al., 1994), based in part on Joseph Felsenstein's nucleic acid sequence Maximum Likelihood method (Felsenstein, 1993), was used with a transition / transversion ratio of 2.0.The neighbour-joining method (Saitou & Nei, 1987) of MEGA was used to construct a distance tree.The number of nucleotide substitutions per site was estimated by distance model Kimura 2-parameter (Kimura, 1980).Gaps or missing data were treated using Complete Deletion option in MEGA, which removed the sites that contain missing data or alignment gaps before the analysis begins.Both transitions and transversions were considered in the substitution analyses, assuming the substitution rates do not vary among sites.Distance method Kimura 2-parameter (Pairwise distances) was used, and bootstrapping of 1000 replications was performed to test the reliability of the putative tree.A total of 79 Trichogramma ITS2 sequences (12 identified in this study and 67 retrieved from GenBank) were used in the analyses, using U. semifumipennis as the outgroup based on previous phylogenetic work (Schilthuizen & Stouthamer, 1997).

ITS2 sequences and alignment
The 18 ITS2 sequences identified in this study were registered in GenBank with the accession numbers listed in Table 3.The registered sequences are complete ITS2 sequences containing no flanking sequences of 5.8S and 28S.The ITS2 sequences showed little difference in length between different individuals of the same line and a The first 3 letters of Trichogramma species names suffixed with the corresponding strain names represent acronyms for strain designation as described in Table 1 and Table 2 between populations of the same species (0~7 bases in complete ITS2 sequences), while ITS2 sequences between species revealed inconsistent divergence.For example, the ITS2 sequences of T. embryophagum and T. dendrolimi strain den_RH have a difference of 71 bases, whereas those between T. dendrolimi strain den_JL and T. brassicae strain bra_GER clone02 do not differ in sequence length.Twelve of these ITS2 sequences were aligned with 67 ITS2 sequences taken from GenBank in ClustalW (1.82).DNA Identity Matrix was used, and the data matrices are available at TreeBASE (http://www.treebase.org/treebase/index.html).

Within groups and net between groups average distances
Within groups or within species, the ITS2 sequences showed little variation (Table 4).No distances could be detected in T. platneri using the Complete Deletion option, while a distance of 0.005 was found using the Pairwise Deletion option.In all cases, within groups average distances were consistently smaller than 0.02.
Between groups or at the interspecies level, the ITS2 sequences showed much greater divergence than within groups (Table 5).The overall pairwise distance between Trichogramma species is 0.23 (n = 27 groups).The largest pairwise ITS2 distances between Trichogramma species were found between T. bourarachae and other Trichogramma species, ranging from 0.634 to 0.826.There was no divergence between T. minutum and T. platneri, while the ITS2 distance between T. evanescens and T. maidis was only 0.003.The ITS2 sequences indicate that the species in these 2 pairs of Trichogramma species are identical.There are other very closely related Trichogramma species, such as T. alpha / T. aurosum (0.020), T. pretiosum / T. oleae (0.017), T. sathon / T. deion = (0.026) and T. cacoeciae / T. embryophagum (0.027), all of which have an ITS2 distance of approximately 0.02, the highest value for within groups average.TABLE 4. Within groups average distances (Kimura, 1980).
Number of groups -29; number of taxa -81; number of sites -681; gaps/missing data -Pairwise deletion; distance method: Kimura 2-parameter (Net between group average).Standard errors were estimated by bootstrap method (replications -1000).Average distances are shown below the diagonal while standard errors are shown above the diagonal.
-   (Kimura, 1980).U. semifumipennis was chosen as outgroup.Bootstrap values for 1000 replicates are shown on the branches.Strain names are as described in Table 1 and Table 2. Surprisingly, the distance between T. confusum and T. chilonis was 0.127, much greater than expected for within species distances.The overall average distance between U. semifumipennis and Trichogramma species, which could be regarded as an inter-genera ITS2-based distance, was 0.817, while the inter-family distance between Nasonia vitripennis and Trichogramma species was 1.177.

Phylogenetic analyses
Three different tree-making methods were used: Maximum Parsimony (MP), Neighbour-Joining of distances (NJ) and Maximum Likelihood (ML).All three methods gave similar results.A NJ tree was reconstructed for 80 taxa with U. semifumipennis as the outgroup (the sum of branch length, SBL = 2.2065).Different distance models were used and all produced similar phylogenetic topologies.The bootstrapped NJ tree is shown in Fig. 1.A search for the most parsimonious trees was performed by branch-and-bound analysis.A total of 58 mostparsimonious trees were found, which required a total of 2683 steps in each site (CI = 0.784566, RI = 0.581250, number of informative sites = 309).A bootstrapped search generated the 50% majority-rule tree in Fig. 2 and yielded 19 nodes with strong support (>90%).When evaluated by maximum likelihood (jumble 10×, global rearrangement, randomized input order), the ITS2 datasets produced 24068 ML trees (total weight of positions in analysis = 708; transition / transversion ratio = 2; transition / transversion parameter = 1.4765).The extended majority rule consensus tree (Fig. 3) clearly supported the phylogenetic relationship deduced by using the distance method and parsimony.
In general, all phylogenetic trees generated showed that the ITS2 sequence of T. maidis clustered within T. evanescens but not in the same group as T. brassicae, while T. confusum and T. chilonis sequences clustered in different branches.

DISCUSSION
As expected, our studies show that within group or intraspecies divergence is significantly smaller than between groups or interspecies divergence.The ITS2 sequences clearly separated T. maidis and T. confusum from T. brassicae and T. chilonis, respectively.The 6 populations belonging to T. dendrolimi formed a distinct and unique clade, and T. maidis is always in the same branch as T. evanescens populations in the topologies obtained using different methods.On the basis of the distance data, we concluded that T. confusum is not a subspecies of T. chilonis, and T. maidis is not T. brassicae but a cryptic or sibling species of T. evanescens.Our results provide the first molecular evidence of the taxonomic status of these previously proposed cryptic species complexes.
The utility of ITS2 as a means of identification was tested on the T. deion (Pinto & Oatman) and T. pretiosum (Riley) complexes (Stouthamer et al., 1999).This indicated it could be used for species identification in Trichogramma, because the sequence variation within species was minor relative to the differences found between species and all the morphologically distinct cryptic species were distinguished by sequence differences.As shown in our topologies, the populations of T. deion and T. pretiosum clustered as separate groups.Stouthamer et al. (2000) used ITS2 to separate T. minutum (Riley) and T. platneri (Nagarkatti), two North American species that cannot be distinguished morphologically (Pinto, 1999), but as no species-specific sequence differences were found the authors suggested that both species had recently diverged from a common ancestor.In all three trees presented in this paper, T. minutum and T. platneri always cluster together.However, as T. minutum and T. platneri are reproductively incompatible (Nagarkatti, 1975;Pinto et al., 1991), the general correlation between sequence variation and reproductive compatibility is complicated, because the biological species concept is based solely on the fact that they are reproductively compatible.However, as reproductive incompatibility is often associated with differences in morphology and ITS2 sequence structure in many Trichogramma species that have been investigated, the taxonomic status of T. minutum and T. platneri is questionable.In the case of T. maidis and T. brassicae, our experiments indicate they are reproductively incompatible (data not shown), and based on this and their ITS2 variation, we conclude they are reproductively isolated species.As for T. maidis and T. evanescens, they can successfully mate but do not produce offspring.In the future we shall focus on making additional crosses and molecular studies, including studies on mitochondrial DNA (mtDNA).Although the taxonomic position of cryptic species of Trichogramma is still disputable, we suggest that T. maidis is a cryptic species of T. evanescens, because their ITS2 sequences are nearly identical and in the phylogenetic trees T. maidis is embedded in the T. evanescens group with a bootstrap support of 84% in NJ tree or 79% in MP tree.Insects are notorious for evolving morphologically similar sibling species, so the above needs to be confirmed by additional mating studies and the gathering of other data.
It is important to note that the species status of T. confusum is so uncertain that Chinese research workers decided to vote on its relationship with T. chilonis at the 1999 National Symposium on Trichogramma (Nai-Quan Lin, pers.commun.),which confirmed that the taxonomic status of T. confusum is disputable.Because no crossing experiments have been made, it is difficult to apply the biological species concept.However, the diagnosable differences or genetic distance between T. confusum and T. chilonis (0.127 compared with <0.02 within species) and their positions in the trees indicate they are closely related sister species but not cryptic or sibling species.
The potential use of the ITS2 sequence for identifying Trichogramma species depend on sound morphological studies, because traditionally species are morphologically based.In addition to single rearing, the samples (lines) used here were all tested for consistency using hundreds of independent specimens, so the chance of mixing the lines was very unlikely.It should be noted that the Fig. 2. MP consensus tree for 79 Trichogramma species using U. semifumipennis as outgroup.A total of 58 parsimonious trees were found, which required a total of 2683 steps in each site.Bootstrap values for 1000 replicates are shown on the branches.Strain names are as described in Table 1 and Table 2. sequence data previously reported by different authors and used for phylogenetic analysis in this study all support the formerly discovered conclusions, namely minor within-species and distinct interspecies ITS2 sequence divergence.Moreover, based on comparatively largescale sequence sampling, a baseline that can be regarded as a species border for delineating Trichogramma populations was determined, namely a distance value of approxi-Fig.3. The extended majority rule consensus ML tree for 79 Trichogramma species using U. semifumipennis as outgroup.A total of 24068 ML trees were examined using fastDNAml version 1.2.2 (Olsen et al., 1994).Total weight of positions in analysis = 708; transition / transversion ratio = 2; transition / transversion parameter = 1.4765.Strain names are as described in Table 1 and Table 2. mate 0.02 calculated by the Kimura 2-parameter model.Can new taxa be erected on the basis of rDNA differentiation?Is it sufficient to have a distance of about 0.02 to define a new species?Such a baseline does not confirm the existence of a taxonomic relationship.This can only be done by performing crossing experiments as suggested by Pinto et al. (1991).

Fig. 1 .
Fig. 1.Rooted NJ tree for 79 Trichogramma species based on Kimura 2-parameter distance model(Kimura, 1980).U. semifumipennis was chosen as outgroup.Bootstrap values for 1000 replicates are shown on the branches.Strain names are as described in Table1 and Table 2.

TABLE 2 .
Reference ITS2 sequences retrieved from GenBank for phylogenetic analysis.
. b GB ID indicates GenBank accession numbers.c ITS2 sizes were trimmed from the original sequences as shown in the sequence alignment.

TABLE 3 .
Different clones represent sequences from different individuals of the same line.Strain abbreviations are as described in Table 1.b GB ID indicates GenBank accession numbers.c Length (bp) represents ITS2 without flanking sequences.PCR products (bp) include flanking regions.Length and GenBank accession numbers of ITS2 sequences identified in this study.

TABLE 5 .
Pairwise distances for ITS2 sequences between groups.