Completely predatory development is described in a braconid wasp

Hymenopteran parasitoids are well known for their ubiquitous diversity, important ecological roles and biocontrol potential. We report the first detailed documentation of mite predation by a parasitoid wasp, Bracon predatorius Ranjith & Quicke sp. nov., (Insecta: Hymenoptera), first case of obligate predatory behaviour in the family Braconidae and first case of mite feeding within the superfamily Ichneumonoidea. Larvae of a new wasp species are shown to develop entirely as predators of eriophyid mites that induce leaf galls in a commercially important plant. They display highly modified head capsule morphology that we interpret as being associated with this atypical life style. We propose that the new feeding strategy evolved separately from recently described entomophytophagy in another species of the same genus. The divergent larval morphological adaptations of both species indicate a high degree of evolutionary developmental plasticity in the developmental stage.

The enormously diversified order Hymenoptera includes sawflies, bees, wasps and ants, and is one of the four largest insect orders 1,2 . The vast majority of hymenopterans are parasitoids of other insects and thus are often used by us humans as biological-management agents 3 . Although the parasitoid life-history strategy has evolved in diverse insect groups, recent estimates show that 10-20% of all known insects are parasitoid wasps 4 . The parasitic Hymenoptera exhibit either of two mode of development, ecto-or endoparasitoidism, the former generally being the plesiomorphic condition 5 . Ectoparasitoids lay their eggs on or close to the host after paralyzing it. Complete host paralysis leads to idiobiosis 6 in which the parasitoid larva's interaction with the host is short lived. Most idiobiont parasitoids attack hosts that are concealed 4 . Species of the highly diverse superfamily Ichneumonoidea (comprising Braconidae, Ichneumonidae and Trachypetidae) are predominantly parasitoids on other insects 7 , although a few display other biologies including gall induction and seed predation 8,9 , entomophytophagy 10 or predation within arachnid egg masses 11 . Most species of Braconidae are parasitoids of larval Lepidoptera, Diptera, and Coleoptera 12 . Egg predation has evolved on several occasions within the Ichneumonidae where it is clearly derived from the parasitoid ground plan, but predation of multiple mobile prey is only known from one cryptine species 13 .
We provide here the first detailed observations of acarophagy, a new larval feeding strategy, in the parasitoid Hymenoptera, the first such case in the Ichneumonoidea, and the first case of purely predatory behaviour in the Braconidae. In addition to this, we describe a new braconid wasp, Bracon predatorius Ranjith & Quicke sp. nov. (Insecta: Hymenoptera). The larvae of this wasp feed solely on the mite, Aceria (=Eriophyes) doctersi (Nalepa, 1909) (Acari: Eriophyidae) which induce galls on the leaves of Cinnamomum verum J.Presl (Lauraceae) 14 (Fig  1a-c) an economic crop in much of S.E. Asia 15 . Severe infestation by the mite leads to a high density of galls which reduces plant vigour (Fig 1a,b). The biology described here expands both host range and feeding strategy within the Ichneumonoidea.

Results
The presents study indicates that Bracon predatorius generally oviposits during early stages of gall development (Fig. 1d) on galls induced by Aceria doctersi mostly on tender leaves (Fig 1a-c) and rarely on petioles and stems 13 . The number of B. predatorius larvae in parasitized galls ranged from 1-27 (n=93). Eighty-five percent of the examined galls (n=109) were parasitized by B. predatorius. Different development stages of larvae (Fig. 1f, . 1i) of B. predatorius were found together in some large galls (n=31) (Fig. 1i), which suggests multiple oviposition at different stages of gall development. Dissection of leaf galls two hours after oviposition by B. predatorius always revealed only a single egg (n=8). No live A. doctersi individuals were found close to the parasitoid wasp pupae (Fig. 1h). Aceria doctersi galls parasitised by B. predatorius have also been found in Kodakara (Thrissur district, Kerala) about 100 km away from the type locality in Kozhikode. The larval stages of B. predatorius feed on both juvenile and adults of A. doctersi (Fig 2d-f, Supplementary Video 1) which usually remain close to the erineal hairs on which they feed 16 ; no egg predation occurs. Young larvae of B. predatorius wriggle through in between erineal hairs (Supplementary Video 1). They use their sickleshaped mandibles (Fig 3b- Unattacked galls were significantly smaller than those containing B. predatorius (means 217 and 595 respectively; p<0.0001) (Fig. 2a) as were galls containing only a single Bracon juvenile (p<0.0001). However, galls without Bracon larvae contained significantly more mites than either all attacked galls (means 503 and 194 respectively; p<0.0001) (Fig. 2b) or those attacked by only one Bracon larvae (p<0.0001). Considering only attacked galls with numbers of Bracon and gall volume as explanatory variables and including an interaction term, we found that the number of mites was highly significantly negatively correlated (p<0.0001) with the number of Bracon present (Fig. 2c) but gall volume and the interaction terms were both insignificant (p=0.94 and p=0.58 respectively).

Male. Similar to female.
Etymology. The specific epithet refers to its unusual feeding behaviour. Comparative affinities. The cosmopolitan genus Bracon 7 includes 17 subgenera and more than 800 described species 20 . For historical reasons, its subgeneric classification is largely based on the Palaearctic fauna and is unreliable especially for extralimital taxa [21][22][23] . Bracon predatorius sp. nov. comes close to the subgenus Orthobracon Fahringer although with some distinct differences in the sculpture of propodeum and metasoma, and therefore we provisionally include it under Orthobracon. Of the described Indian fauna, B. predatorius comes close to B. keralense reared from the leaf galls on Cinnamomum malabatrum (Burm.f.) J.Presl (Lauraceae) induced by an unnamed Cecidomyiidae 24 , but B. predatorius differs in having the face without medial longitudinal ridge, notauli indistinct posteriorly, scutellar sulcus divided by eight carinae, metanotum with midlongitudinal carina anteriorly, second metasomal tergite without smooth, parallel sided medio-basal area, ovipositor without dorsal nodus and ventral serrations.

Discussion
The present study gives the first detailed documentation of mite predation by a parasitoid wasp Bracon predatorius Ranjith & Quicke sp. nov. (Insecta: Hymenoptera), which is the first case of obligate predatory behaviour in the family Braconidae and first case of mite feeding within the superfamily Ichneumonoidea. Even though parasitoid associations of Hymenoptera with non-insect groups is well studied in the past 7,10,25,26 , host-parasitoid associations between the Hymenoptera and Acarina are known only from the superfamily Chalcidoidea 27,28 and have never been described in detail. Chalcidoid parasitism of non-insect arthropod groups is best known in the family Encyrtidae which includes species of Ixodiphagus which are endoparasitoids of ticks 29,30 . Acarophagy within the Hymenoptera has previously only been reported within the chalcidoid family Eulophidae 31,32 . Egg predation of arachnids is also known in a few Ichneumonidae 11,33 , but a rigorous search in the literature did not yield any report of predatory behaviour in the Braconidae nor utilisation of non-insect hosts. Significant reduction of mites in the galls parasitized by B. predatorius confirmed that the obligate acarophagy of B. predatorius is enhanced by pure predation of its host mites. This deviating lineage (from parasitoidism) of foraging behaviour (through predation) is likely to be a confirmation of the diversified feeding traits within the www.nature.com/scientificreports/ 'Parasitica' where aculeates are well known for this feeding trait 34 . Some tentative observations on the predatory behavior by the larvae of the chalcidoid wasp Aprostocetus (Eulophidae) within Eriophyes ribis (Nalepa) induced plant galls indicate that acarophagy probably also occur within the Chalcidoidea 28,31,32 , however, the larval behaviour of Aprostocetus eriophyes has not been described in detail 31 .
In the case of the new Indian predatory Bracon species described here, the larval head capsule morphology that we describe is highly derived compared with typical ectoparasitic, idiobiont species of the subfamily 17 . However, highly derived mandibles and other larval features have been described in both the purely phytophagous, seed-predating, B. phytophagus and in the entomophytophagous B. garugaphagae 9,10 . It is suggested that the absence of basal comb of teeth on the mandibles means the larvae feed on the mites without chewing. Most of parasitoid species are either attached onto or located inside their hosts 11 and some lay their eggs near to the paralysed hosts. It is possible that the pure predatory behaviour reported here for B. predatorius is derived from entomophytophagous behaviour as reported in B. garugaphagae 10 . Collectively, these observations suggest considerable potential for evolutionary plasticity in braconid larval anatomy as a result of changes in feeding habit.
Methods collection permits. Necessary  Species description and terminology. Alcohol preserved adult specimens were treated with hexamethyldisilazane to prevent collapse during drying and then card mounted. Final instar larval head capsules were prepared by macerating the larval heads in 10% KOH (aq. wt/vol.) to dissolve soft tissues, followed by washing in dilute acetic acid, dehydration through alcohol series and mounted in Canada balsam. Morphological terminology used in the description of B. predatorius follows van Achterberg 35 but with wing veins follows Quicke 7 . Terminology for describing sculpture follows Harris 36 . Terminology employed in the description of final instar larval head capsule follows Čapek 17 . Illustration and documentation of behaviour. Images from the fields (Fig 1a,b,d) were taken by Canon 7D equipped with 100 mm Canon macro lens. Light microscopic images of egg, larvae, pupa and adults of B. predatorius were taken by Leica DMC2900 camera connected with Leica M205A (Figs. 1c,g, 3b, 4 and 5) and Leica DFC295 camera connected with Leica S8 APO (Figs. 1e,f,h,i, 2d-f and 3e) respectively. Images of final instar larval head capsule were taken by Leica DMC4500 camera connected with Leica DM2000 compound microscope ( Fig. 3a-d). Image stacks were combined into a single image was done using Leica Application Suite V4.2. Images were edited using Photoshop CS8 (Version 6.1) (Adobe Inc.). Measurement of holotype and developmental stages of B. predatorius were done using AxioVision 4.8. Feeding behaviour of B. predatorius was recorded with a hand-held smart phone connected with Leica S8 APO stereozoom trinocular microscope. The rate of feeding of mites by B. predatorius larva was estimated from the recordings. The behavioural peculiarities of B. predatorius while feeding on the mites was also noted.
Statistical analysis. Firstly, we analysed whether galls differed in volume between attacked ones and those without Bracon larvae (n=109) using Student's t-test with unequal variances. Then we examined the relationship between the number of mites per gall (response variable), numbers of Bracon larvae present and gall size measured as diameter × height (explanatory variables). Visual inspection of an initial scatterplot of mite number versus Bracon number showed that galls with Bracon larvae all had far fewer mites than unattacked galls. Therefore, we performed three separate analyses. Firstly, we compared gall volume for attacked and unattacked galls as well as unattacked galls and those containing a single Bracon using two-tailed t-tests with unequal variances. Secondly, we compared mite numbers between unattacked galls and attacked galls as well as against galls containing a single Bracon larva. Thirdly, we tested whether in attacked galls there was an effect of numbers of Bracon larvae on mite number using ANOVA. To make model errors satisfactorily near normal, we used the transformation log(number of Bracon larvae) to reduce overdispersion. Further analyses excluding a few remaining outlier data points made no substantive difference to results. All data analyses were performed using R 37 .

Data availability
All data are available in the main text or the supplementary materials.