MORPHOLOGICAL VARIABILITY IN WING SIZE AND WING SHAPE AMONG SPECIES Aphidius absinthi Marshall, A. rosae Haliday AND A. urticae Haliday (HYMENOPTERA: BRACONIDAE: APHIDIINAE)

. In this paper, we applied traditional and geometric morphometric methods to analyze variability in wing size and wing shape among species Aphidius absinthi Marshall, A. rosae Haliday and A. urticae Haliday. These taxa represent closely related species with different biological and ecological characters. For the morphometric analyses, we used a sample of 52 female specimens that were collected during the period 2009-2013, on different localities in Serbia. Traditional morphometric analyses revealed statistical significance in stigma shape discrimination of analysed taxa. Our geometric morphometric analyses also confirmed that major contribution to the wing shape variation had the changes in length of the radial sector and stigma shape. Combining the traditional and geometric morphometric analyses, we confirmed the validity of the wing characters previously used in taxonomic studies of the genus Aphidius .


INTRODUCTION
The genus Aphidius Nees, with about 100 described species worldwide (TOMANOVIĆ et al., 2007) and about 35 detected species in Europe (STARÝ, 1970), is one of the the largest within the subfamily Aphidiinae (Hymenoptera: Braconidae). All species are solitary endoparasitoids of aphids. Many Aphidius species have a great potential as biocontrol agents in biological control programs (HAGVAR and HOFSVANG, 1991), so a success of these programs depends on their correct identification. Because of that, there are many papers relating to the taxonomy of Aphidius species (EADY, 1969;STARÝ, 1973;PUNGERL, 1983;PENNACCHIO, 1989; MESCHELOFF and ROSEN, 1990;TAKADA, 1998;TOMANOVIĆ and STARÝ, 2001;KAVALLIERATOS et al., 2001KAVALLIERATOS et al., , 2006TOMANOVIĆ et al., 2003TOMANOVIĆ et al., , 2004TOMANOVIĆ et al., , 2007TOMANOVIĆ et al., , 2013KOS et al., 2011;JAMHOUR et al., 2016). However, due to a great variability of morphological characters, many taxonomic problems were encountered in the genus Aphidius. One of them is taxonomical position of Aphidius absinthi Marshall, A. rosae Haliday and A. urticae Haliday, which represent closely related species with different biological and ecological characters. A. absinthi is parasitoid of Macrosiphoniella Del Guecio species, A. rosae represents a highly specialized species restricted to Macrosiphum rosae Linnaeus, while A. urticae has wide host range and parasitizes on Acyrthosiphon Mordvilko, Amphorophora Buckton, Macrosiphum Passerini and Microlophium Mordvilko species. According to the last revision, these taxa mostly differ from each other by the number of antennal segments, length of metacarpal vein, number of costulae on anterolateral area of petiole and host range (STARÝ, 1973).
The purpose of this study was to analyse morphological differentiation in the forewing size and shape among species A. absinthi, A. rosae and A. urticae by traditional morphometry and geometric morphometric analyses and to test the validity of morphological characteristics, such as wing venation, previously used for their identification (STARÝ, 1973;PENNACCHIO, 1989;TOMANOVIĆ et al., 2003TOMANOVIĆ et al., , 2007.

Traditional morphometry
For the morphometric analyses, we used a sample of 52 female specimens that were collected during the period 2009-2013, on different localities in Serbia (Table 1). Plant samples bearing both live and mummified aphid hosts, were collected for parasitoids rearing. Samples of live aphids were preserved in 90% ethanol and 75% lactic acid at a ratio of 2:1 (EASTOP and VAN EMDEN, 1972) for later identification. The remaining aphids were maintained in the laboratory until parasitoid emergence. Mummies, each attached to a small leaf piece, were placed separately in small plastic boxes and put inside a growth cabinet. On the lid of each box there was a circular opening covered with muslin for ventilation in order to maintain the conditions inside the boxes similar to those in the growth cabinet (22.5 o C, relative humidity 65%, 16L:8D) (KAVALLIERATOS et al., 2001). All analyzed specimens were boiled in 10% KOH, dissected, and mounted in Canada balsam (STARÝ, 1970). The external structure of emerged parasitoids was studied using a ZEISS Discovery V8 stereomicroscope.
Three continuous characters were used for the morphological characterization of the analyzed specimens, as follows: stigma length (STL), stigma width (STW) and the length of R1 vein = metacarpal (R1L) (Fig. 1). All the characters were presented in terms of a ratio in order to eliminate effect of size, also allowing direct comparison of the obtained results with other analyses ( Table 2). Morphological terminology for wing diagnostic characters used in this study is based on SHARKEY and WHARTON (1997).

Geometric morphometrics
The geometric morphometrics approach was applied to explore and quantify variations in wing size and wing shape of 52 female specimens (ZELDITCH et al., 2012). The same sample was used for both, traditional morphometry and geometric morphometric methods (Table 1). Left forewing of each specimen was detached, mounted in Canada balsam and photographed using a Leica System Microscope DM2500 with a Leica DFC490 Digital Camera. We selected 13 specific landmarks to describe the wing size and shape. The positions of the landmarks are given in Fig. 2, while their definitions are presented in Table 3. All landmarks were digitized using TpsDig software (ROHLF, 2005). Landmarks were superimposed by the Generalized Procrustes Analysis (ROHLF and SLICE, 1990;BOOKSTEIN, 1991). Procrustes coordinates were used as shape variables in following statistical analyses. To estimate wing size, we computed the centroid size (CS), a geometric measure of the size which reflects the amount of dispersion around the centroid of the landmark configuration (BOOKSTEIN, 1991). The variation in the wing size (CS) among different species from the genus Aphidius was analyzed by one-way ANOVA. Post-hoc pairwise comparison for wing size was done by Tukey's test. Multivariate analysis of variance (MANOVA) on the full set of the shape variables was performed to analyze a differences in the wing shape of parasitoids belonging to different species (ZELDITCH et al., 2012). All statistical analyses were performed with the Statistica 6 software package (STATSOFT, 2001). Canonical variate analysis (CVA), which reduces within group variances and increases between group divergences, was performed to explore a divergence of the wing shape among the three species using software MorphoJ (KLINGENBERG, 2011). Discriminant function analysis (DFA) was used to evaluate the accuracy of classsification by original and cross-validation percentages of the cases (MANLY, 1997).

Traditional morphometry
The Analysis of variance showed statistically significant differences in stigma shape Also, the results of Canonical Variate Analysis confirmed that ratio character STL/STW has higher contribution to the species discrimination ( Table 4). The results of Discriminant function analysis (DFA) based on analysed ratio characters, indicate about 50% correct assignment of specimens to the a priori designated species. The following percentages for the correct classification of individuals per species were found: A. absinthi 50.00% , A. rosae 47.37% and A. urticae 52.94%.
Correct classification of individuals per species based on wing shape was provided as the following percentages (the first and the second values in brackets represent the original and cross-validation, respectively): A. absinthi 100% (62.5%), A. rosae 100% (100%) and A. urticae 100% (65%).
Canonical variate analysis (CVA) revealed that the first canonical axis explained 90.60% of the total variability in wing shape. A. absinthi and A. urticae were clustered together and clearly discriminated from A. rosae by the position of radial sector and r-m vein (Fig. 3). Specimens of A. rosae have elongated radial sector (described by landmarks 4, 5 and 8) and shorter r-m vein (described by landmarks 5 and 6). However, A. absinthi and A. urticae specimens are separated along the second canonical axis, which explained 9.40% of the total variability in wing shape. The main shape changes that discriminate these species are related to the stigma shape (described by landmarks 3, 4 and 9) and R1 vein (described by landmarks 9 and 10). In contrast to A. urticae, the specimens of A. absinthi have wider stigma, shorter R1 vein and wider distal part of the wing (described by landmarks 11, 12 and 13). Previously, taxonomic studies within a genus Aphidius were based mainly on a few morphological characters such as: wing venation, sculpturing on the anterolateral area of petiole, number of antennal segments, shape and chaetotaxy of the female genitalia, tentorial index, number of maxillary and labial palpomeres (SMITH, 1944;EADY, 1969;STARÝ, 1973;TOMANOVIĆ et al., 2003TOMANOVIĆ et al., , 2007.
Based on traditional and geometric morphometric analyses our results confirmed that all three analyzed taxa, A. absinthi, A. rosae and A. urticae are true species. According to the results of traditional morphometry, we found that the stigma shape has statistically significant influence on the species discrimination. Geometric morphometric analysis also confirmed that the major variation in the wing shape consisted of changes in the length of the radial sector and r-m vein, as well as in the stigma shape itself. The shape of the stigma, the length of R1 vein and the ratio between the length of stigma and the length of the R1 vein were comonly used as valid characters for the morphological characterization and separation within the genus Aphidius (STARÝ, 1973;PENNACCHIO, 1989;KAVALLIERATOS et al., 2001;RAKHSHANI et al., 2008;PETROVIĆ et al., 2010;KOS et al., 2011;TOMANOVIĆ et al., 2003TOMANOVIĆ et al., , 2013TOMANOVIĆ et al., , 2014. In contrast to the last revision of the genus Aphidius, where A. absinthi and A. rosae belong to the A.rosae group (STARÝ, 1973), we found that A. absinthi has clustered together with A. urticae and clearly discriminate from A. rosae, which specimens have elongated radial sector and shorter r-m vein. However, we also found clear differences between A. absinthi and A. urticae in stigma shape, length of R1 vein, as well as in wing size.

CONCLUSIONS
We found clear differences among species A. absinthi, A. rosae and A. urticae in the wing size, as well as in the wing shape by the position of radial sector and r-m vein, stigma shape and the ratio between the length of stigma and the length of the R1 vein.
Combining the traditional and geometric morphometric analyses, we confirmed the reliability of previosly used wing characters for Aphidius identification, also indicating that rm vein could be used as a new character in identification keys.
This paper represents a contribution to the resolving of some taxonomical problems within the genus Aphidius, but the clarification of the status of many other aggregations or cryptic species requires further morphological and molecular researches.