Suspended mummies in Aleiodes species (Hymenoptera: Braconidae: Rogadinae) with descriptions of six new species from western Uganda based largely on DNA sequence data

A group of species of the rogadine braconid genus Aleiodes are shown to produce a distinctive mummy, which is “J”‐shaped and is formed after the host larva, in all cases an ennomine geometrid moth, has dropped from a plant suspended in midair by a silk thread. The group includes one described species, A. buzurae He & Chen from China, and a species complex from tropical Africa (W. Uganda). All the African specimens reared from suspended mummies looked morphologically virtually indistinguishable, though there was considerable colour variation that segregated the specimens into five groups. Three gene fragments (nuclear 28S D2‐3 rDNA, the nuclear ITS2 region and part of the mitochondrial cytochrome oxidase 1 gene (CO1)) were sequenced to assess if these specimens represented a single variable species or a complex of morphologically cryptic species. Results show variation in all three gene fragments, with strong signal in the CO1 gene, parsimony analysis of which revealed six well supported groups corresponding to the colour variants, except that two specimens with nearly identical colour differed considerably in their CO1 sequences. Large, and difficult to align, variation was found in the ITS2 fragments, which by eye also supported the same six groupings. Limited variation was found in the 28S fragment, but one position supported monophyly of the two specimens belonging to one of the species circumscribed by the other genes. These groups are considered to correspond to separate species, which are described as new: A. barnardae Quicke & Shaw, A. basutai Quicke & Shaw, A. kanyawarensis Quicke & Shaw, A. kasenenei Quicke & Shaw, A. mubfsi Quicke & Shaw and A. trevelyanae Quicke & Shaw. The possible function of the specialised mummification behaviour is discussed and some observations on rates of hyperparasitism are presented.

Introduction suppressaria (Guenée) (5Biston, according to Scoble (1999)) that clearly has close affinity with the species found in western Uganda. Neither in the original description (He and Chen 1990) nor in the revision (Chen and He 1997) is mention made of the mummy being suspended from a thread but, in view of the date of collection (1954) of the original material by another person, it seems likely that this feature had not been appreciated by these authors rather than that it was not so. Indeed, the configuration of the host mummy illustrated by Chen and He (1997) really leaves little other possibility.

Molecular methods
DNA was extracted from single legs preserved in absolute ethanol using an ethanol precipitation method with final elution into 30 ml of water. PCR was carried out in 20 ml reactions containing 1.0 ml of DNA extract, 10 pmol of primers (Table I), 10 nmol of dNTPs (Amersham Pharmacia Biotech: APB), 1.0 U of Taq polymerase (Bioline) and 2 ml of 106 reaction buffer (2.0 mM MgCl 2 ). PCR conditions were 94uC for 30 s, 50uC for 30 s and 72uC for 60 s (35 cycles with an initial denaturation for 2 min and a final extension for 7 min). PCR products were purified using GFX gel band purification kit (APB) and sequenced directly using BigDye terminators.
Primer sequences are given in Table I. The COI primers LCO/HCO (also called Folmer primers, after Folmer et al. (1994)) were COI forward5LCO 1490 and COI  ATG CTT AAA TTT AGG GGG T reverse5HCO 2198. The forward ITS2 primer was designed based on the 5.8S rDNA sequence of Trichogramma minutum Riley (GenBank accession numbers U36235 and U36236) anchoring between the 63rd and 81st positions. The reverse primer, which anchors at the beginning of the 28S rDNA sequence, was modified (terminal base removed) from that of Porter and Collins (1991).

Materials
Aleiodes apiculatus (Fahringer) and A. testaceus (Telenga) were included as outgroups because, in a larger study of the phylogeny of Aleiodes species based on analysis of the 28S D2-D3 rDNA gene region (Mori et al. in preparation), these appeared in groups on either side of the clade, including the suspended mummy taxa. One additional unidentified Afrotropical individual, AL0468, from the east shore of Lake Naivasha, Kenya, was included because it appears, on the basis of DNA sequence data and morphology, also to belong to this group. For comparison of variation in the CO1 gene fragment of the African specimens with that within known and well-supported European species, sequences were obtained from seven specimens of A. pictus (Herrich-Schäffer), five specimens each of A. coxalis (Spinola) and A. ruficornis (Herrich-Schäffer), and from four of A. dissector (Nees), collected from a wide range of localities in the UK and Europe.
Four additional species, representing A. compressor (Herrich-Schäffer), A. unipunctator (Thunberg), an unidentified species from Las Cuevas, Belize (AL0005) and one from the Amani Hills, Tanzania (AL0044), were also included to provide broad representation of the genus.
DNA sequences are deposited in the EMBL/GenBank database; accessions numbers, provenances and voucher numbers are given in Appendix A.

Cladistic methods
Sequence data were analysed using maximum parsimony with PAUP* (Swofford 1999). Bootstrapping on the 15 taxon data set used 500 bootstrap replicates, each search of the pseudoreplicate using branch-and-bound searching. Maximum parsimony analysis of the 38 taxon data set used 1000 random additions, tree bisection-reconnection branch swapping, with only one tree saved each time: trees of the most parsimonious length obtained were found in more than 90% of random additions.

Comparison with Aleiodes buzurae
The western Ugandan material is similar to A. buzurae in having a distinctive and strong rugulose-reticulate sculpture of the first four metasomal tergites and mid-dorsal and lateral sinuate emargination of the posterior of the fourth metasomal tergite. Although obviously belonging to the same compact species group, A. buzurae differs from all of the Ugandan material examined in its more sharply-defined and deeper postero-lateral emargination on the 4th metasomal tergite, its pattern of metasomal markings (Figure 6), and its somewhat slenderer legs in females ( Figure 175 in Chen and He 1997).

Molecular results
Analysis of the 28S D2+D3 sequence data for the suspended mummy specimens revealed no phylogenetic structure, a strict consensus of the .1 million equally parsmonious trees being completely unresolved. However, a small number of substitutions were apparent. Both specimens of A. trevelyanae sp. n. differed from all others by a single substitution in the D3 region, the single specimen of A. mubfsi sp. n. differed from all others by two bases in the D2 region (corresponding to positions 107 and 199 of the alignment presented by Belshaw et al. (1998; Figure 1 loc. cit.), and A. kasenenei sp. n. differed from all others at one base in the D3 region.
Although intraspecific variation can be found in both CO1 and ITS sequences (Alvarez and Hoy 2002), the variation observed in the CO1 sequences was greatly in excess of that observed between multiple conspecific individuals (even from widely different localities) for a number of other Aleiodes species. For example, Figure 2 shows a phylogram, derived from maximum parsimony analysis of CO1 sequence data, for the A. buzurae complex specimens and multiple individuals of four well-supported European species of Aleiodes, and within these the total CO1 variation on the tree corresponds to at most 10 base changes among the A. pictus individuals from six widely separated localities, seven base changes in each of A. coxalis and A. ruficornis and six in A. dissector. The variation between the individuals of A. barnardae sp. n. (six bases) and A. trevelyanae sp. n. (one base) is, therefore, equivalent to that found within the European species, given that they are all from the same small region of forest. Of the Ugandan species we recognise, the closest (on the basis of their COI Figure 2. Phylogram from analysis of CO1 DNA sequence data for individuals of the A. buzurae-group and related species, and also multiple representatives of four European species for comparison. sequences) are A. kanyawarensis sp. n. (one female) and A. kasenenei sp. n. (one male) (see Figures 2 and 3), which have widely differing colour patterns. The COI sequences of these two specimens, differing at 27 base positions, were still more than twice as different from one another as are conspecific members of any of the other species (Figure 4), and their ITS2 sequences have markedly different inserts ( Figure 5). In contrast, the ITS2 regions of the four individuals of A. barnardae n.sp. were identical, as were those of both of the A. trevelyanae sp. n. specimens. We, therefore, conclude that the specimens reared in Kibale represent a complex of morphologically practically identical, but genetically isolated distinct species.
Monophylies of both A. trevelyanae sp. n. and of A. barnardae sp. n. are indicated by 100% bootstrap support in the analysis of their COI sequence data ( Figure 3).  Furthermore, a sister group relationship between A. kanyawarensis sp. n. and A. kasenenei sp. n. obtained 99% bootstrap support. Other relationships between the buzurae group species were equivocal.
None of the species described here is known from more than one sex. The two species known only from males, A. kasenenei n.sp. and A. kanyawarensis n.sp. share a largely pale pterostigma ( Figure 14) and a more rectangular second submarginal cell of the fore wing. These might be secondary sexual features because the molecular phylogenetic analyses (see Figures 2 and 3) indicate that these are not particularly closely related despite their similar pterostigmal colour pattern.
Antenna with 40-42 segments (41 in holotype), 1.35 times longer than fore wing. Terminal flagellomere strongly acuminate, 3.4 times longer than wide. Median flagellomeres 2.25 times longer than wide. Sculptured parts of 1st and 2nd flagellomeres equally long. Third segment of maxillary palp 1.3 and 1.6 times longer than the 4th and 5th segments, respectively. Inter-tentorial distance 1.46 times tentorio-ocular distance. Width of clypeus:width of face51.0:2.3. Width of head:width of face:height of eye53.1:1.0:1.8. Face with small elongate median bulge, lateral to this with distinctly transverse rugose striae. Frons depressed and with distinct carina bordering anterior two-thirds of depression laterally, close to but separate from margin of eye. Stemmaticum coarsely rugose. Occipital carina broadly effaced medially.

Molecular features
The ITS2 sequence of A. barnardae n. sp. is virtually identical to that of A. kasenenei n. sp ( Figure 5), but these two species differ in their CO1 sequences at many 3rd codon positions (see Figure 4).

Etymology
Named after Sue Barnard for her friendship and help during the 2002 Kibale field trip.

Molecular features
Displays nine unique base substitutions in the CO1 gene fragment, of which one is shown in Figure 4. The ITS2 fragment shows five unique substitutions in the length-conserved part (  As for A. barnardae sp. n. except for colour. Largely pale honey-yellow, stemmaticum black, antennae except small ventral mark on scape, occiput, propodeum except narrowly laterally and posteriorly, first metasomal tergite except anterior semicircular area and narrowly medio-posteriorly, second metasomal tergite except broadly medially and narrowly laterally, third and fourth tergites except narrowly laterally, apex of hind tibia and hind tarsus brown or brown-black; malar region paler yellow; fore and mid coxae and trochanters yellowwhite; wings clear with dark brown venation and entirely black pterostigma.

Molecular features
Displays two unique sequences in the indel regions of the ITS2 gene ( Figure 5).

Etymology
Named after the type locality.

Description
Length of body 4.5 mm, of fore wing 4.1mm. Antenna with 40 segments. As for A. barnardae sp. n. except for colour. Pterostigma largely pale buff with borders and apical quarter grey. Metasomal tergites 1-4 largely pale yellow-white, narrowly more ochreous-yellow laterally.

Molecular features
Differs from all other species in the group by a single base substitution in the D3 region of the 28S gene. In terms of the ITS2 region, it has similar inserts and deletions to A. barnardae sp. n. (Figure 5).

Etymology
Named after Dr John Kasenene of Makerere University Biological Field Station, for his knowledge of Kibale and support for the Tropical Biology Association.

Molecular features
Both sequenced individuals possessed a unique substitution in the D3 region of the 28S gene. The CO1 gene fragment sequenced was identical for both individuals and displays 15 unique substitutions, four of which are shown in Figure 4. Both length variable parts of the ITS2 sequences had indels of unique length and sequence. Named after Dr Rosie Trevelyan, the 'chief mzungu female' of the Tropical Biology Association.

Use of DNA in tropical insect identification
Despite our extensive study of the specimens reared from the suspended mummies at Kibale, we have been unable to discern any morphological differences among them, though there is clear discontinuous variation in colour pattern. Without molecular evidence, we would simply have considered this as a variable or colour-polymorphic species. However, the high level of support for multiple clusters based on analysis of COI sequence data (Figures 2 and 3), and the congruence between these and the colour pattern and visuallyrecognised clusters of ITS2 sequences ( Figure 5), indicates that these clusters are reproductively isolated even though sympatric and, therefore, we consider them to represent discrete species. Importantly, COI is mitochondrial and ITS2 is nuclear and, therefore, in sympatric, sexually reproducing species congruence in haplotypes of these two markers provides strong evidence that these are reproductively isolated species.
These results not only illustrate the use of both ITS2 and CO1 genes for discriminating species, something which has attracted a lot of attention recently (Porter and Collins 1991;Paskewitz et al. 1993;Hebert et al. 2003a,b;van Veen et al. 2003), but also indicates that estimates of species diversity and global species richness, based purely on morphological assessment, might be considerable underestimates. Furthermore, although most of the A. buzurae-group species recognised here are distinguishable on the basis of colour, two are virtually identical so, even if colour had been used as an indicator, at least one cryptic species pair would have been missed. Apart from the academic interest in knowing what proportion of morphologically defined species are actually complexes of biologically and genetically delimited cryptic species, it will also be important to the understanding of food webs, especially in the tropics, where this approach is being used to try to understand why species diversity is generally so tropico-centric but where the taxonomy is least well known.
The data presented here indicate that in the genus Aleiodes there is some variation in the CO1 sequence among conspecific individuals, typically five or six base changes (within the approximately 650 base pair fragment amplified) separating individuals on a most parsimonious tree, whereas more than 15 changes distinguished even the two most closely related of the species described as new in this paper (A. kanyawarensis sp. n. and A. kasenenei sp. n.). In addition, little intraspecific variation was found in the ITS2 fragment but different species showed moderate to large differences in the length-variable zones ( Figure 5).
In many Aleiodes species, including some of the A. buzurae complex studied here, reading the ITS2 sequences after direct sequencing was very hard, because most individuals had some intragenomic variation and, more particularly, intragenomic length variation. The sequences presented here ( Figure 5; Appendix A) concentrate on the obviously dominant signal (probably representing the variant that was commonest among the multiple genomic copies of the region), but even with experienced human pherogram reading it was not possible accurately to determine all bases in this variant in the variable regions (see Xs in Figure 5 sequences). Thus, whereas CO1 sequencing might be reliably automated, there will sometimes be arbitrary decisions about the level of variation that is assumed to represent interspecific variation. In contrast, ITS2 (or ITS1) sequences might be more reliable indicators of species boundaries (different nuclear gene pools), but they may also be less practicable as intragenomic variants can compromise automated sequencing in some cases.

Mummification strategy
Many geometrid larvae escape from danger by dropping from their food plant on a silken thread, through which they regain access to the feeding site once danger is perceived to have passed. As Aleiodes species often exploit the latent behaviours of their hosts in order to pupate in greater safety, it is perhaps not surprising to find species that exploit this dangeravoiding reflex of certain geometrid larvae by causing the host to drop on a thread before being mummified. However, the host behaviour noted in the Aleiodes species-group sampled by us in western Uganda is not effective at completely preventing attack by pseudohyperparasitoids. In 2002, we collected a total of 19 suspended mummies, of which seven subsequently produced Aleiodes, eight produced hyperparasitoids, and four failed to emerge. Three of the hyperparasitoids belonged to groups known to behave as true hyperparasitoids, i.e. attacking the primary parasitoid while the latter is still feeding. These were two specimens of one species of Mesochorus (Ichneumonidae) and an Afroperilampus sp. (Perilampidae), and both of these made emergence holes like those of Aleiodes. The remaining five mummies produced three species of Eulophidae (some gregarious) that all belonged to groups likely to behave only as pseudohyperparasitoids in the context (one species each of the genera Pediobius and Tetrastichus, and a further unplaced species of Tetrastichinae). Because pseudohyperparasitism is generally an on-going process affecting the primary parasitoid throughout its cocooned period, and because some of the mummies were collected before this period was over, only a minimum level of hyperparasitism (about 50% overall) can be estimated from the above small collection (and even then that would presume that the overall level did not vary at other dates). In particular, it appears that mummification at the end of the thread was not preventing at least three Chalcidoidea species from exploiting the mummies as strongly presumed pseudohyperparasitoids (i.e. ,33% of the mummies collected).