The value of the ITS 2 region for the identification of species boundaries between Alloxysta hyperparasitoids ( Hymenoptera : Charipidae ) of aphids

1) Alloxystinae are major secondary parasitoids of aphids, important in both their ecology and pest management. 2) Two radically differing views of alloxystine taxonomy exist in the literature, in one of which the group is very diverse, in the other it con­ sists of a few variable species. 3) We sequenced a variable nuclear gene region (ITS2) for 28 specimens of a morphologically clearly defined group which, in one view belong to a single species and in the other to four species. We find that the four putative species each carry a different unique allele with no intraspecific variation. We show that the probability of the observed distribution of alleles under the assumption of a single interbreeding population is very small and we reject the view that all specimens belong to a single biological species. 4) We discuss the implications of our results for aphid parasitoid community ecology and the biological control of aphids with parasitoids.


INTRODUCTION
Taxonomic problems can be a severe hindrance to eco logical research.This is particularly true in community ecology where the basic unit of analysis is often the bio logical species and the presence of cryptic species and/or spurious species can cause huge problems in correct data inference.The difficulties often arise because the species concerned are small and have a limited number of diag nostic morphological characters.In these circumstances setting species boundaries can be hard and subject to dif ferent interpretations.The use of modern DNA-based molecular techniques can be an important tool for resolving these issues (Avise, 1994;Knowlton, 2000;Waters et al., 2001).
For the last eight years we have been studying quantita tively the community of aphids and their primary and sec ondary parasitoids at a site in the south of England (Müller et al., 1999).We are particularly interested in how pairs of species that do not directly interact may nev ertheless have linked population dynamics by virtue of their shared natural enemies, a process termed apparent competition (Holt, 1977).We construct quantitative food webs to assess the potential importance of these indirect effects, but the success of this approach depends on cor rectly identifying the biological species present.
Aphid primary parasitoids are insects whose larvae develop as parasites of aphids eventually pupating inside the mummified skin of the dead aphid, creating the socalled mummy (Godfray, 1994).Some of the major taxo nomic problems in this group have been resolved with electrophoretic studies (Pungerl, 1986;Atanassova et al., 1998).Secondary parasitoids attack the primary parasi toid larvae and can be thought of as the top predators in the aphid-parasitoid community.Of the 28 secondary parasitoid species listed by Müller et al. (1999) in our community, 18 belong to the Alloxystinae (Hymenoptera, Cynipoidea, Charipidae).There are two radically dif ferent opinions about the diversity of this group.In the only modern comprehensive review of the British species, Fergusson (1986) recognised few species, most of which were morphologically very variable and with a wide host range.The opposite view is represented by Evenhuis and co-workers (Evenhuis & Barbotin, 1987 and references therein) who recognised many more species, each of which were relatively invariable with a narrow host range.Our studies of the biology of the alloxystines within our community have suggested that species have narrow host ranges and we have provisionally accepted this second view, also because this leads to the most con servative estimate of the potential for hyperparasitoid mediated indirect interactions.However, if the alternative is in fact correct, then this would make a major difference to how we interpret the structure of our food web.
Here we use molecular techniques to determine the spe cies boundaries in one well-defined group of Alloxystinae, where we recognise three species in our community, but which according to Fergusson (1986) are all members of a single variable species.We sequenced the ITS2 region of the rDNA of specimens of each spe cies collected in or around our field site.To assess geo graphical variation, for two species we also sequenced specimens from Germany.Finally we also studied British and German specimens of a fourth taxon that does not appear in our food web but hyperparasitises the economi cally significant cabbage aphid and which also belongs to the same variable species according to Fergusson (1986).The Charipidae is a morphologically uniform group of insects about 0.5-2 mm in total body length, all of which develop as secondary parasitoids of Homoptera.It is divided into two sub families, the Charipinae which attack parasitoids of psyllids, and the Alloxystinae, which attack parasitoids of aphids.In Europe the Alloxystinae comprise two genera, Phaenoglyphis and AHoxysta.Andrews (1978) catalogues the world species of Alloxystinae and lists 120 names for Palaearctic Alloxysta although a large fraction of these are likely to be synonyms.Alloxystines parasitise the primary parasitoids of aphids (which belong to the chalcidoid family Aphelinidae and the braconid subfamily Aphidiinae) and attack their hosts while the aphid is still alive, prior to mummification (in contrast to other secon dary parasitoids that attack the mummy).
We worked with a group of morphologically very similar taxa that we refer to as the victrix group.They can be defined by the following set of characters: (i) closed radial cells; (ii) absence of carinae on propodeum; (iii) carinae present on pronotum, with area between them hairless; (iv) male antennae with segments 3, 4 and 5 curved or emarginate (see Evenhuis, 1974 andMenke &Evenhuis, 1991 for further details of alloxystine morphology).Fergusson (1986) treats the group as a single species, victrix Westwood, while we currently recognise four species by applying the criteria of Evenhuis and co-workers.Within the victrix group, individuals reared from different hosts can be dis tinguished by subtle characters such as the ratio of the lengths of antennal segments and body colour.Whether these differences are species diagnostic, or whether they merely reflect hostinduced phenotypic plasticity can only be resolved by studying genetic characters that are independent of the wasp's host.Fer gusson also synonymised A. circumscripta Hartig with A. victrix but the structure of its pronotum, a character not used by Fer gusson, shows it to be outside the victrix group and not closely related (Menke & Evenhuis, 1991).

Specimens analysed
In our long-term study of an aphid-parasitoid community (Müller et al., 1999; unpublished data) we have reared three taxa of the A. victrix group that we have provisionally considered to be species.Specimens of each are included in this study as well as a fourth species found nearby but not in our study site.Table 1 is a full list of the specimens sequenced, their hosts, host aphids and host plants.(i) victrix Westwood sensu stricto.The limited conception of this species is due to Evenhuis (1972, pers. comm.) but it is still the most polyphagous member of the genus, though the aphids whose parasitoids it attacks are all Macrosiphini.In our study site we have reared this species from 10 different hosts and found it to be quite variable, especially in colour.However, the colour variation seems to be correlated with host size.We included 11 specimens from three hosts span ning this variation.(ii) leunisii Hartig.Evenhuis (1982) reared an alloxystine from the aphid Uroleucon which he identified as this species by comparison with the type.We in the UK, and Wolfgang Völkl and Gerhard Hubner (pers.comm.) in Germany, have reared morphologically similar forms from Uroleucon species that differ, especially in colour, from rare victrix sensu stricto attacking the same host.We sequenced specimens from both countries.(iii) tscheki Giraud.Alloxystines reared from Cryptomyzus spp.
on Ribes have been called tscheki.We obtained specimens from Germany reared from this host which we found to be morphologically identical to the most common alloxystine in our study site which in Müller et al. ( 1999) is called Alloxysta v2 and is reared exclusively from the aphid Capitophorus carduinis on Cirsium.We also sequenced a specimen reared from a very different host, the spruce aphid Elatobium abietinum, that again we could not distinguish morphologically.A. tscheki was excluded from the British faunal list by Fergusson (1986).(iv) fuscicornis Hartig.One of the most frequently studied allox ystines is the hyperparasitoid of the cabbage aphid (Brevicoryne brassicae) through the primary parasitoid Diaeretiella rapae.In the older literature it was known as Charips bras sicae Ashmead but it was synonymised with infuscata Kieffer (originally described as a variety of victrix) by Evenhuis (1972) and then later both names were synonymised with^uscicornis (Evenhuis, 1982).Fergusson (1986) synonymised fuscicornis with victrix.We sequenced specimens from two localities in the UK, and one in Germany.Parasitised aphids were collected in the field and placed in gelatine capsules for rearing in the laboratory.Specimens for sequencing were placed in 100% ethanol within 24 hours of emergence and came from temporally and/or spatially separated samples to ensure they were not siblings.Molecular techniques and primers DNA was extracted from the metasoma of single specimens using the DNeasy kit (Qiagen) with final elution into 30 pl water.Following initial amplifications using primer sequences in Campbell et al. (1993), we obtained complete ITS2 sequences using the following primers situated in the 3' end of the 5.8S sub-unit and the 5' end of the 28S sub-unit respectively: forward = ATT CCC GGA CCA CGC CTG GCT GA; reverse = CGC CTG ATC TGA GGT CGT C (written 5' to 3').
PCRs were carried out in a GeneAmp 9700 thermal cycler in 50 pl reactions containing 1.0 pl DNA extract, 20 pmol primers, 10 nmol dNTPs (Amersham Pharmacia Biotech; APB), 1.5 units Taq polymerase (Roche) and 5.0 pl Taq buffer (containing 1.5 mM MgCl2).Cycle conditions were 94°C for 30 sec; 44°C for 30 sec and 72°C for 1 min (35 cycles, plus an initial denatura tion for 2 min and a final extension for 7 min).Products were cleaned using GFX gel band purification (APB) and then sequenced directly using dye terminators on an ABI 3700 auto mated sequencer using 1/4 recommended volumes (PE Biosys tems).

RESULTS
Four different alleles were identified among the 28 specimens sequenced (EMBL accession numbers: AJ309962-5).The aligned sequences of these alleles are shown in Fig. 1.The species provisionally classified as A. victrix, A. leunisii, A. tscheki and A. fuscicornis each pos sessed a unique allele with no intraspecific variation.In the three cases where we studied specimens from both the UK and Germany, the wasps from the two countries had identical sequences.
Is it possible that these results could have been obtained by chance if the null hypothesis is that the specimens belonged to a single interbreeding population?Consider only the 11 A. victrix (10 females, 1 male) and 5 A. tscheki (4 females, 1 male) collected at Silwood Park.Because the differences between the "victrix" and "tscheki" alleles are mainly indels (Fig. 1) diploid hetero zygote females would have been identified (the pherograms would become unreadable after the site of the first indel) but none were found.If they were part of a single interbreeding population then our best estimate from these data of the frequency of the "victrix" allele is 0.7 (this estimate takes into account that in Hymenoptera females are diploid and males haploid).The probability that a female is homozygous "victrix" is thus 0.49 and homozygous "tscheki" 0.09.The probability of the observed data based on the assumption of a single inter breeding population is (0.49)10(0.09)4= 5 x 10A Applying this calculation to the other five pair wise com parisons reveals that the probability always falls well below the 0.05 significance level (Table 2).It is important to note at this point that this analysis is based on the bio logical species concept and is possible because these four species occur sympatrically.This approach cannot be used to infer species status for geographically isolated populations.

DISCUSSION
Our molecular analyses firmly support the more diverse view of alloxystine taxonomy championed by Evenhuis and co-workers.The group appears to be relatively spe cies rich and composed of taxa that differ only in subtle morphological and colour characters.Although the appli cation of traditional morphological species concepts can work in this group, the molecular data is critical for pro viding independent evidence for their validity.It is clear that A. fuscicornis (Hartig), A. leunisii (Hartig) and A. tscheki (Giraud) are valid names and should not be con sidered synonyms of A. victrix (Westwood).Our results confirm that the potential for apparent competition medi ated by Alloxystinae was not underestimated by Müller et al. (1999), as would have been the case had the Fergusson (1986) concept of A. victrix proven correct.
Of the species we studied, A. fuscicornis and A. leunisii appear to be specialists on D. rapae attacking B. brassicae, and A. funebris attacking Uroleucon spp., respec tively.We have reared no other alloxystine from D. rapae and just a single A. victrix from A. funebris, which was sequenced.Confirmation of the distinctness of these two species suggests that it is worth testing the hypothesis that they have specialised on these hosts and competitively exclude other hyperparasitoids (these species are attacked by other secondary parasitoids, but only those with a dif ferent life history that attack the primary parasitoid after it has mummified the aphid).B. brassicae is a major pest of brassicas hence the interest shown in A. fuscicornis biology (Nahif & Madel, 1990;Ayal & Green, 1993).It is common throughout the world wherever brassicas are grown (Evenhuis, 1974;Andrews, 1978;Carver, 1992) and may be responsible for the poor performance of D. rapae as a biological control agent.Our results showing A. fuscicornis to be a specialist demonstrate that nearby non-cruciferous crops and weeds are unlikely to act as a source of hyperparasitoids that may affect D. rapae popu lations.
A. tscheki and A. victrix are more polyphagous species, and while we cannot exclude the possibility that they con tain more cryptic taxa that are not revealed by the ITS2 sequence, the available molecular and morphological evi dence clearly points to each being a distinct species.A. tscheki is a common hyperparasitoid on two closely related aphid genera (Capitophorus feeding on Cirsium and Cryptomyzus on Ribes) but its third host, Elatobium, though also in the Macrosiphini is not considered closely related (Heie, 1992;Heie, 1995) and its host plant (Picea) is very different.The primary parasitoid attacking Elato bium (Aphidius schimitscheki) is however a close relative of A. matricariae, the primary parasitoid of Capitophorus (Starý, 1973).A. victrix has the widest host range in the genus and has also been studied closely because of its role as a hyperparasitoid of aphids feeding on various crops (Micha et al., 1993;Grasswitz & Reese, 1998;Petersen et al., 2000).Unlike the case of A. fuscicornis, our results do suggest that neighbouring crops and weeds may harbour populations of the secondary parasitoid that might interfere with aphid biological control.Despite the consistent intraspecific variation in A. victrix associated with host size, we found no evidence that these forms were distinct species (though we have only studied speci mens from a small fraction of the species' host range).Further evidence for the unity of the species is the obser vation that wasps collected from the aphids Microlophium carnosum and Macrosiphum rosae readily mated and accepted Aphidius ervi attacking Acyrthosiphon pisum as a host (unpublished data).Although A. tscheki and A. vic trix are both polyphagous on larger spatial scales, in our long-term study community only A. victrix has multiple hosts (Müller et al., 1999) and is thus a candidate species to mediate indirect population effects amongst its hosts.
The ITS gene regions have been commonly used to used to detect species boundaries where morphology is suspect (Cornel et al., 1996;Van Oppen et al., 2000).Our study further demonstrates the usefulness of this region for this purpose, particularly when combined with testing a priori hypotheses derived from population genetic models (Sites & Crandall, 1997).Our results are clear-cut because of the observed lack of intraspecific variation in ITS2 among our insects.This pattern, possibly due to concerted evolution (Brown et al., 1972;Zimmer et al., 1980;Elder & Turner, 1995), combined with the rela tively fast rate of evolution of ITS2 (Navajas et al., 1998) makes this region ideally suited for detecting cryptic spe cies.However, intraspecific and even intragenomic het erogeneity have been reported in the ITS regions of some taxa (Vogler & Desalle, 1994;Onyabe & Conn, 1999;Beebe et al., 2000).What these different patterns of varia tion may be telling us about the population structures of the various organisms is unclear and requires a better understanding of the molecular mechanisms involved.

Table 1 .
Host details and localities of the AHoxysta specimens that were analysed.As = Ascot, Berkshire, UK; Ho = Hope, Derbyshire, UK; He = Henley, Oxfordshire, UK; Ba = Bayreuth, Germany.The location of our long-term community study is Ascot.

Table 2 .
Pair wise probabilities of the observed data under the null hypothesis of interbreeding populations.See text for a full explanation.