Morphometric analysis and taxonomic revision of Anisopteromalus Ruschka (Hymenoptera: Chalcidoidea: Pteromalidae) – an integrative approach

We use an integrative taxonomic approach to revise the genus Anisopteromalus. In particular, we apply multivariate ratio analysis (MRA), a rather new statistical method based on principal component analysis (PCA) and linear discriminant analysis (LDA), to numerous body measurements and combine the data with those from our molecular analysis of Cytb and ITS2 genetic markers (on a subset of species) and all available published data on morphology, karyology, behaviour, host associations and geographic distribution. We demonstrate that the analysis of quantitative characters using MRA plays a major role for the integration of name-bearing types and thus for the association of taxa with names. Six species are recognized, of which two are new: A. cornis Baur sp.n. and A. quinarius Gokhman & Baur sp.n. For Anisopteromalus calandrae (Howard), a well-known, cosmopolitan parasitoid of stored-product pests, we have selected a neotype to foster continuity and stability in the application of this important name. The species was sometimes confused with the related A. quinarius sp.n., another cosmopolitan species that is frequently encountered in similar environments. We also show that several species originally described or later put under Anisopteromalus actually belong to different genera: Cyrtoptyx camerunus (Risbec) comb.n.; Meraporus glaber (Szelényi) comb.n.; Dinarmus schwenkei (Roomi, Khan & Khan) comb.n. Neocatolaccus indicus Ayyar & Mani is confirmed as a junior synonym of Oxysychus sphenopterae (Ferrière) syn.n. and Anisopteromalus calandrae brasiliensis (Domenichini) stat.rev. must be considered as a valid but doubtful taxon. This published work has been registered in ZooBank, http://zoobank.org/urn:lsid:zoobank.org:pub:BDFE96D3-D0F4-4012-90F5-9A087F7F5864.


Introduction
In systematics, data from multiple sources are becoming more and more easily available to taxonomists working at the species level. This has led to interest in a method for combining these data, which recently has been termed integrative taxonomy (for a review and the distinction from iterative taxonomy, see Yeates et al., 2011). The term is adopted when species delimitation is based on results from a variety of disciplines, for instance morphology, DNA analysis, cytogenetics, behaviour or biochemistry. Detailed procedures have been devised for integrating conflicting datasets into sound hypotheses (e.g. Schlick-Steiner et al., 2010).
In most studies adopting an integrative approach, attention is usually paid just to species limits per se, based on the evaluation of the sample at hand (e.g. Chesters et al., 2012). Only rarely is the main task of a taxonomic revision also considered (e.g. Steiner et al., 2010), that is, the association of taxa with names. This leads to cases where information is often very limited and data integration becomes a real challenge: for example, when name-bearing types are damaged and character sets are thus incomplete; when types are too fragile or old for molecular analyses; and when types are lost (or inaccessible) and nominal taxa are based on nothing more than a set of qualitative characters. In combination such challenges cause major, and widespread, problems in evaluating available evidence to support necessary taxonomic revisions.
Although focusing on morphometry, we are here pursuing such an integrative approach for establishing species boundaries and identities within Anisopteromalus Ruschka (Hymenoptera: Chalcidoidea: Pteromalidae: Pteromalinae), a small genus of parasitic wasps currently comprising seven species (Noyes, 2013). The genus was erected for Anisopteromalus mollis Ruschka which is the type species by monotypy, a species synonymized with A. calandrae (Howard) by Graham (1969). The genus is easily recognized by a combination of characters, most notably the female antenna with three anelli, the structure of the propodeum and the extended hind margin of the first gastral tergite (Graham, 1969;Bouček & Rasplus, 1991). Species of Anisopteromalus occur mainly in the Old World, where they were recorded from tropical Africa (Risbec, 1956;Rasplus, 1988), Asia (Roomi et al., 1973;Sureshan, 2010) and Western Europe (Szelényi, 1981). They usually parasitize beetle larvae (e.g. Chrysomelidae: Bruchinae, Anobiidae, Curculionoidea) feeding on stored grain and legume seeds (Fabaceae: Faboideae and Caesalpinioideae), but have sometimes been reared also from lepidopteran hosts (e.g. Gelechiidae, Pyralidae) (Noyes, 2013).
While for most Anisopteromalus species hardly anything has been published beside the original description or an occasional host record, A. calandrae is a well-known, cosmopolitan parasitoid of various stored-product pests. It has been the subject of numerous studies spanning a wide variety of topics, such as biological control (Hou et al., 2004;Ngamo et al., 2007;Ni et al., 2008;Chaisaeng et al., 2010), impact of pesticides and herbicides (Perez-Mendoza et al., 1999;Lacoume et al., 2009;Yoon et al., 2009), life-history traits Lebreton et al., 2009aLebreton et al., , 2010Chaisaeng et al., 2010), behaviour including learning (Ryoo et al., 1996;Lebreton et al., 2009b;Belda & Riudavets, 2010;Ishii & Shimada, 2010) and physiology (Zhu et al., 1999;Howard & Baker, 2003). The ISI Web of Science database (Thomson Reuters) cites more than 100 papers in this respect and many of the older works can be retrieved from the Universal Chalcidoidea Database (Noyes, 2013).
Despite the multitude of studies, little critical attention has been paid to the systematics of Anisopteromalus. As mentioned before, most species have not been re-considered since their description. Furthermore, the taxonomic status of A. calandrae is doubtful. Gokhman et al. (1998) first suspected that two sibling species might be hidden under that name. In fact, investigation of the karyotypes of two laboratory strains revealed a marked difference in the number and shape of chromosomes. One strain originating from Moscow, Russia (called the MSU strain) showed five chromosomes in its haploid set (n = 5). All of those chromosomes were metacentric except for the last submetacentric one. Moreover, they could be subdivided into two size classes. In the other strain, cultured at Imperial College at Silwood Park, Ascot, Berkshire, UK (origin Slough, Berkshire; called the ICSP strain) with n = 7, all chromosomes were metacentric and demonstrated a continuous gradation in length. Such differences in the karyotype are usually a clear indication of different species (Gokhman, 2009). Gokhman et al. (1998Gokhman et al. ( , 1999 also found significant differences in morphology, courtship behaviour and several important life-history traits. Because the crossing between MSU and ICSP strains also failed, there could be little doubt that more than one taxon was involved. Later on, similar differences between some other populations presumably belonging to A. calandrae were also found (Gokhman & Timokhov, 2002) and their different host preferences were revealed also (Timokhov & Gokhman, 2003). However, the presence of the two closely related and widely distributed species with contrasting life-history strategies in the genus Anisopteromalus got relatively little support from the expert community for a number of years (but see Quicke, 2002), although it is becoming increasingly accepted now (see, e.g., Sasakawa et al., 2012Sasakawa et al., , 2013. Of course, taxonomic ambiguity within such an important taxon needs to be solved. As mentioned above, we focused on morphometric data and their integration with molecular data as well as all relevant published information on morphology, karyology, behaviour, distribution and ecology. We also considered all valid names and their junior synonyms (in total 16 nominal taxa) and checked their name-bearing types whenever possible. Beside the type material, more than one thousand specimens from various collections and from several cultured strains of supposed 'A. calandrae' were studied. For the morphometric study we applied multivariate ratio analysis (MRA), a recently developed method that allows the interpretation of results from principal component analysis (PCA) and linear discriminant analysis (LDA) in terms of body ratios (Baur & Leuenberger, 2011) and that is thus especially suited for analysing body measurements in a taxonomic context (reviewed in László et al., 2013). Furthermore, the MRA algorithms offer separate analyses of shape and size as well as an estimation of the extent of shape change with size (i.e. allometry in the sense of Gould, 1966). Allometric variation of body ratios was first observed for Pteromalidae by Janzon (1986) who discussed its impact on species delimitation in the Pteromalus albipennis group. MRA also allowed the inclusion of the available name-bearing types, which were not usable for DNA sequencing because the age of specimens (usually more than 20, up to 100 years old) is likely to hamper successful extraction and amplification. We also used the mitochondrial cytochrome b (Cytb) and the nuclear internal transcribed spacer 2 (ITS2) for differentiating entities within the 'A. calandrae' complex.

Specimens and character selection
For the morphological investigation we used a total of more than 1300 specimens. Morphometric analysis was based on a subset of 289 dry-mounted females. We focused on females because they are usually easier to separate and more readily available (e.g. Graham, 1969;Bouček & Rasplus, 1991 Gibson (1997) is followed for terminology of morphological structures. The list of morphometric characters used in the analyses is given in Table 1. Table S1 gives an overview of the basic descriptive statistics for each measurement (in μm) and species as well as the sample sizes. The selected characters correspond to those used in the taxonomy of Pteromalidae for calculating standard ratios (e.g. Graham, 1969). Most measurements were made with an Olympus SZ11 stereomicroscope at different magnifications using a calibrated eye-piece micrometer (12 mm subdivided into 120 units) and were taken by Y. Kranz-Baltensperger. For some of the name-bearing types, each character was photographed with a Keyence VHX 2000 digital photo-microscope and a VH-Z20R/W zoom lens at a magnification of 200× (i.e. 1000 μm corresponded to 888 pixels) and was measured by H. Baur using ImageJ v1.47v (Schneider et al., 2012); body parts on the images were enlarged 3-4 times before measuring. For all measurements, we ensured that the points of  (Graham, 1969) 150× ool.l OOL Shortest distance between posterior ocellus and eye margin, dorsal view (Graham, 1969) 150× pdl.flg Pedicel + flagellum Combined length of pedicel plus flagellum, outer aspect (Graham, 1969) 100× pol.l POL Shortest distance between posterior ocelli, dorsal view (Graham, 1969 reference were equidistant from the lens of the microscope and that the diaphragm of the lens was fully open. To avoid additional variation due to fluctuating asymmetry (e.g. Palmer & Strobeck, 1986;Bechshøft et al., 2008), measurements of paired characters were usually taken on the left-hand side. The Keyence microscope with a VH-Z100UR/W zoom lens was used for making stack-images of body parts, except for the forewings, which were removed and embedded in Hoyer's medium on slides prior to stack-imaging using a Leica DFC420 camera on a Zeiss Axioskop 40 light microscope and the ImageAccess software (Imagic AG, Glattbrugg, Switzerland).

Morphometric analysis
We applied multivariate ratio analysis (MRA) of Baur & Leuenberger (2011) to our data. MRA comprises a set of tools for analysing size and shape of body measurements in a multivariate mathematical framework that is entirely consistent with the customary usage of body lengths and ratios in taxonomic works (e.g. descriptions, diagnoses). In systematic studies, MRA offers several advantages over conventional explorative multivariate methods, such as principal component analysis (PCA) and linear discriminant analysis (LDA). We refer to László et al. (2013) who gave an overview of the various MRA tools and applied them to a particular taxonomic problem in some other Hymenoptera. Following Baur & Leuenberger (2011) we first calculated an isometric size axis (isosize), defined as the geometric mean of all variables. We then performed a shape PCA (i.e. a PCA in the space of all ratios) for evaluating how well the morphometric pattern corresponds to the groups obtained by qualitative morphology and karyology. In order to decide how many shape components to retain we inspected the scree graph (Rencher, 2002: 398-399). We also plotted isosize against shape PCs, because the correlation of size with shape is a measure of the amount of allometry in the data (e.g. Klingenberg, 1998). Estimation of the extent of allometric variation is important, because body ratios sometimes correlate with size. Depending on the magnitude of this correlation, the use of ratios may then hold the risk that discrimination of groups is -indirectly -more or less based on size rather than shape (e.g. Janzon, 1986;Seifert, 2008). For this reason, we also employed two graphical tools, the PCA ratio spectrum and allometry ratio spectrum, respectively. Finally, we used the LDA ratio extractor to extract the best ratios for use in the identification key and diagnoses, and calculated the standard distance as well as the measure .
The R language and environment for statistical computing was used for data analysis (R Core Team, 2013;v3.0.2). In particular, we employed slightly modified versions of the R-scripts provided by Baur & Leuenberger (2011, under 'Supplementary material'). Pearson product-moment correlation coefficients were calculated with the function 'cor()' using the default settings. Scatterplots were generated with the package 'ggplot2' (Wickham, 2009). A few specimens, especially some of the name-bearing types, lacked one body part or another. In order to be able to include all specimens in the multivariate analyses, missing values were imputed with the R package 'mice', using the default settings of the function 'mice()'. For the calculation of ratios used in the description and Table 2, specimens with missing values were excluded, because imputed values may sometimes produce outliers when calculating ratios.

Taxonomic sampling
We sampled six specimens of Anisopteromalus calandrae and three specimens of A. quinarius belonging to the main strains used in laboratories (Table S2). Specimens were collected alive and stored in 95% ethanol. Adult specimens were identified to species by V.E. Gokhman, A.V. Timokhov and J.-Y. Rasplus. DNA vouchers are deposited at the CBGP collection, Montferrier-sur-Lez, France. Four outgroup taxa belonging to the pteromalid genera Nasonia (three species) and Pachycrepoideus (one species) were included to root phylogenetic trees. Cytochrome b (Cytb) and ITS2 sequences from Nasonia species were downloaded from GenBank (Table S2).

DNA extraction, amplification and sequencing
Total DNA was extracted using standard phenol-chloroform methods (Sambrook et al., 1989). Due to repeated failures of amplification of the Cytochrome c oxidase I Folmer fragment (COI, standard barcode) for most specimens, we used a long fragment of the Cytochrome b gene (Cytb) instead. To investigate potential mitochondrial introgression we also sequenced one nuclear gene, the internal transcribed spacer ITS2 rRNA.
PCR products were purified with QIAquick PCR purification kit (Qiagen, Venlo, The Netherlands) and directly sequenced on an ABI 377 automated sequencer using TaqFS and dye-labeled terminators (Perkin-Elmer). CP1 and Tser were used as sequencing primers for Cytb. Both strands for each overlapping fragment were assembled using the sequence-editing software Geneious v5.5.7 (Drummond et al., 2012). All sequences were deposited in GenBank (Table S2).

Sequence data analyses
All gene regions were aligned with MAFFT 6.864 (Katoh et al., 2005) using the L-INS-i option. Cytb alignment was translated to amino acids using MEGA 4 (Tamura et al., 2007) to detect frame-shift mutations and premature stop codons, which may indicate the presence of pseudogenes. Pairwise nucleotide sequence divergences were calculated using a Kimura 2-parameter (K2P) model of substitution (Kimura, 1980) in MEGA 4, using the 'pairwise-deletion' of gaps option. The most appropriate model of evolution for each gene region was identified using the Akaike information criterion implemented in MrAIC.pl 1.4.3 (Nylander, 2004). We performed Maximum likelihood (ML) analyses of the two gene regions using MPI-parallelized RAxML 7.2.8. (Stamatakis, 2006a). GTRCAT approximation of models was used for ML bootstrapping (Stamatakis, 2006b) (1000 replicates). Analyses were conducted on a 150-core Linux Cluster at CBGP.

Data resources
Morphometric raw data files, R-scripts used for calculating the MRA, alignments and photographs of measured characters, as well as of type specimens and their labels, are deposited in the Dryad Data Repository at doi: http://doi.org/10.5061/dryad.km728.

Morphometric analysis
For the MRA we focused on six groups that could be separated by qualitative characters (A. apiovorus, A. caryedophagus, A. ceylonensis, A. cornis) or by karyotype (A. calandrae, A. quinarius). At this stage prior to the morphometric and molecular analysis, we deliberately avoided the concept of species and rather interpreted the groups in the sense of operational taxonomic units (OTU). We first performed a series of shape PCAs to see how well the OTUs were supported by variation in shape (Figs 1-3). A PCA type of analysis is convenient here because it does not require a priori assignment of OTUs to particular groups but assumes instead that all OTUs belong to one single group. A PCA thus avoids bias with respect to particular groupings (e.g. Pimentel, 1979;Peters & Baur, 2011). According to the scree graph (not shown), only the first and second shape PC were relevant in all analyses reported below. Figure 1A shows a scatterplot of the first two shape PCs including all twenty variables and six OTUs. Only the OTU A. apiovorus was clearly distinct while the others were at least slightly overlapping. This result was surprising, as all OTUs could be rather well separated by qualitative morphology. We therefore suspected a significant amount of error variance in some of the variables and checked the variables with matrix

Fig. 2.
Size and shape analysis of females of OTUs Anisopteromalus calandrae and A. quinarius using all variables except gaster breadth. (A, B) Shape PCA, scatterplot of first against second shape PC (A), scatterplot of isosize against first shape PC (B). Symbols: blue dots, A. calandrae; red squares, A. quinarius; in parentheses the variance explained by each shape PC. (C, D) Ratio spectra, PCA ratio spectrum (C), allometry ratio spectrum (D); horizontal bars in the ratio spectra represent 68% bootstrap confidence intervals based on 1000 replicates. scatterplots and Pearson product-moment correlation coefficients. Correlation among body measurements of related taxa should usually be positive and high (e.g. Hills, 1978;Pimentel, 1979), hence its lack may indicate measurement error or morphological artefacts. Indeed, gaster breadth correlated much less strongly than all other variables (Table S3 and Figure  S1), with coefficients ranging from −0.18 to 0.30, including many around 0. Closer inspection revealed many specimens with a moderately to strongly deformed gaster (dorso-ventrally or laterally collapsed), mainly in specimens of A. apiovorus, A. caryedophagus, A. ceylonensis and A. cornis. The reasons for these artefacts were unclear, as the preservation history of the specimens was unknown. The observed damage nevertheless led us to remove gaster breadth from all further analyses. As a result, in the new graph (Fig. 1B) of a shape PCA with the remaining 19 variables all groups were neatly separated. The plot also shows the position of the available name-bearing types (marked with bold plus signs). Five of them represent some cornis were quite distinct based on qualitative morphology alone and were also well supported by morphometric analysis. Therefore, we focused on A. calandrae and A. quinarius that differ in their karyotypes with the haploid chromosome number of either n = 7 or n = 5, respectively, but were otherwise generally similar in qualitative morphology. The question thus was how well they were distinguished, not only in shape, but also in size. The results of a shape PCA including only these two OTUs confirmed the pattern in the first plot (Fig. 1B), as they were well separated along the first shape PC ( Fig. 2A). In a scatterplot of isosize against the first shape PC (Fig. 2B) A. quinarius was on average slightly larger than A. calandrae, although the size ranges were broadly overlapping. The plot thus revealed a moderate amount of allometric variation. A similar trend was visible by comparing the PCA ratio spectrum and the allometry ratio spectrum. In a PCA ratio spectrum, only ratios calculated with variables lying at the opposite ends of the spectrum are relevant for a particular shape PC (see also Baur & Leuenberger, 2011;László et al., 2013 for the interpretation of ratio spectra). In the same manner, the most allometric ratios are found in an allometry ratio spectrum. Now, the PCA as well as the allometry ratio spectrum were dominated by the same ratio, eye.b : ool.l (Fig. 2C, D), that is, the most important ratio concerning the first shape PC was also the most allometric one.
The LDA ratio extractor is a tool for finding the best ratios for separating some groups (see Baur & Leuenberger, 2011: 816-818 for how this algorithm works). In contrast to a PCA, group membership had to be specified beforehand. The results are compiled in Table 2 showing various comparisons. The ranges of first and second best ratios were often not or only just overlapping between the respective groupings. We were able to integrate such ratios in the identification key and diagnoses (see below), as they represent important diagnostic features.
The best ratio for discrimination of A. calandrae from A. quinarius, eye.b : ool.l, happened to be the same as the one that dominated the PCA and allometry ratio spectrum (Fig. 2C, D). Note that this is by coincidence, as the best separating ratios must not necessarily correspond to the most important ratios of a PCA ratio spectrum (see also Peters & Baur, 2011 for the conceptual difference between a PCA and a LDA based type of analysis). For instance, the second best ratio, mss.l : eye.d, that still separated most of the specimens in this comparison (compare Table 2), was neither in the PCA nor in the allometry ratio spectrum among the dominant ratios (Fig. 2C, D). In fact, the respective variables were lying rather closely to each other in the spectra and thus the ratio had a negligible influence. Now, of the two best discriminating ratios, eye.b : ool.l and mss.l : eye.d, only the former showed allometric behaviour (i.e. correlated with isosize), while the latter did not. This furthermore demonstrates that allometric variation generally had a marginal impact on the discrimination of the OTUs -in other words, separation of A. calandrae and A. quinarius was attributable to true shape differences, not merely to an indirect size effect.
For all comparisons, -a measure of how well shape discriminates in comparison to size (see Baur & Leuenberger, 2011: 818, formula 14) -was close to 0 (0.02-0.31), again indicating that separation of OTUs was mainly due to shape rather than size. Furthermore, standard distances (Baur & Leuenberger, 2011: 817, formula 12) were relatively high, ranging from 5.46-13.66, which reflects the good separation of OTUs observed in the shape PCA (Figs 1B, 2A).
For the exploration of within-group variation we performed a MRA of the three reared strains of A. calandrae (Fig. 3A) and A. quinarius (Fig. 3B). The scatterplots were in both cases rather homogenous, indicated by the low variation explained by the first two shape PCs (37.3 and 52.9%, respectively). In each OTU, two strains were almost distinct with respect to the first shape PC, but they were partly covered by the third. The analysis did thus not allow any further subdivision of the two OTUs.

Molecular analysis
Cytb (980 bp) and ITS2 (655 aligned bp) were successfully amplified from all specimens of A. calandrae and A. quinarius.  Alignment of Cytb was straightforward due to a lack of length variation, and no stop codons or frame shifts were detected. The intraspecific K2P distance range for Cytb was 0.001-0.024 (mean 0.014) for A. calandrae and 0.006-0.036 (mean 0.025) for A. quinarius. The intraspecific K2P distance range for ITS2 was 0.000-0.002 (mean 0.001) for A. calandrae and 0.002-0.006 (mean 0.004) for A. quinarius. Even though these OTUs were morphologically difficult to discriminate, minimum interspecific divergences between A. calandrae and A. quinarius (Cytb: 0.169, ITS2: 0.271) were very high, largely exceeding maximum intraspecific divergences. Models chosen by MrAIC were as follows: GTR + I for Cytb and SYM + Γ for ITS2. As the SYM model is not implemented in RAxML we used GTR instead.
The results of the molecular analysis confirmed the discrimination of A. calandrae from A. quinarius by MRA. Indeed, phylogenetic analyses of Cytb and ITS2 (Fig. 4) recovered the same well-supported clusters of sequences, which corresponded to both morphologically delineated OTUs ( Fig. 2A). Furthermore, comparison between Cytb and ITS2 genetic clusters revealed no mitochondrial introgression between these closely related species.

Status of OTUs
Our morphometric and molecular analyses unambiguously revealed that all OTUs formed distinct and well-supported taxa. We can thus conclude that the six OTUs examined in this study represent valid species, A. apiovorus, A. calandrae, A. caryedophagus, A. ceylonensis, A. cornis sp.n. and A. quinarius sp.n. The morphometric analysis furthermore confirmed the synonymy of five nominal taxa with A. calandrae (see Fig. 1B). Below, we provide a key to females of all species and descriptions for the two new species as well as for A. calandrae. Information on the other species and a discussion of a few doubtful nominal taxa hitherto associated with Anisopteromalus can be found in Appendix S1. Nomenclatural changes discussed therein are as follows: Cyrtoptyx camerunus (Risbec)  Key to females 1. Head breadth equal to or more than 1.53× metatibia length and eye height equal to or less than 1.1× scutellum length. Forewing speculum medially with a patch of about 10-30 setae (Fig. 5A), setea on wing disc whitish. Scutellum projecting beyond anterior margin of dorsellum (Fig. 5C). First funicular segment subcylindrical, proximally distinctly broader than third anellus and provided with 2-3 rows of longitudinal sensilla (Fig. 5E)

A. quinarius (strain MSU) Russia
Nasonia longicornis ( bare (Fig. 5B), setea on wing disc dark. Scutellum projecting at level of anterior margin of dorsellum (Fig. 5D). First funicular segment subconical, proximally at most slightly broader than third anellus and provided with 1-2 rows of longitudinal sensilla (Fig. 5F) (Fig. 5J). Head and mesosoma blue-green, covered with relatively long, conspicuous whitish setae (Fig. 5L -Gena terete, not carinate near mouth margin (Fig. 6D). Propodeum with anterior plica short and strongly bent inwards, joining a strong costula (Fig. 6F) (Bouček et al., 1979: 435), but see also Cotes (1896: 11); type locality: INDIA. Synonymized with A. calandrae by Bouček et al. (1979: 435). Because the original description is inconclusive (Cameron mainly described some colour characters that apply to many Anisopteromalus species), we are inclined to accept the synonymy of Bouček et al. (1979). Moreover, the species described here as new can be ruled out: first, A. cornis     Mani (1938: 103), synonymy confirmed by MRA (see above). We consider the entire series of 2♂ and 9♀ housed in the USNM to be syntypes, because no specimen was fixed as holotype in the original description (compare Crawford, 1913: 252-253). One female belonged to Dinarmus, as also stated by an identification label of Gahan. The rest of the specimens were clearly conspecific. We have taken as lectotype a female with most body parts intact, and not the female with the additional label 'Aplastomorpha pratti Type Crfd [hand]'. This specimen was in a slightly worse condition, lacking for instance the metatibiae that constitute one of the characters used in our morphometric analyses. Neocatolaccus australiensis Girault, 1913 Bouček (1988: 414), synonymy confirmed by MRA (see above). As pointed out by Dahms (1983: 102), Girault (1913) Tachikawa (1966: 99). The type material could not be traced, even with the help of Japanese colleagues. The original description clearly rules out that N. mamezophagus is the same as any of the new species described here. First, females of A. cornis sp.n. have the pedicel distinctly shorter than first funicular segment, while it is stated by Ishii & Nagasawa that the pedicel is slightly longer than the first funicular segment; second, males of A. quinarius sp.n. have the first funicular segment at least as long and usually longer than the second, while the first segment of N. mamezophagus is said to be slightly shorter than the second. The character states of N. mamezophagus match specimens of A. calandrae as defined here, hence we accept the synonymy of Tachikawa (1966).
Diagnosis, female. Head and mesosoma olive-green with slight bronze tinges in places, setae whitish, inconspicuous (Fig. 5K). Gena terete, not carinate near mouth margin. Flagellum distinctly clavate (Fig. 5I), first funicular segment subconical, basally slightly broader than third anellus, provided with 1-2 rows of longitudinal sensilla (Fig. 5F). Scutellum projecting at level of anterior margin of dorsellum (Fig. 5D), in lateral view weakly curved. Forewing with setae on wing disc dark. Speculum bare, closed in distal and proximal part. Anterior plica of propodeum short and evenly curved, joining sometimes an indistinct costula. Posterior margin of first gastral tergite curving backwards and strongly produced (Fig. 5G) Head in frontal view 1.10-1.30× as broad as high. Antenna with scape 0.77-1.00× as long as eye height. Combined length of pedicel plus flagellum 0.86-1.03× head breadth. First and second anellus strongly transverse, third transverse, about as long as second anellus.
Mesosoma 1.16-1.36× as long as broad. Mesoscutum 1.84-2.42× as broad as long. Hind margin of scutellum broadly rounded. Upper mesepimeron strongly narrowing below, reaching at most basal third of mesopleuron. Basal setal line complete. Costal cell with dorsal surface with at most a short row of setae distally, lower surface with a patch of setae in distal half and a single row of setae running to proximal part, costal setal line complete. Forewing disc moderately pilose. Marginal vein 1.35-1.96× as long as stigmal vein. Stigma subcircular to oval, medium-sized. Propodeum 0.42-0.60× as long as scutellum. Median carina fine, straight. Median area in anterior part evenly reticulate with inner corner of anterior plica with a rather deep and reticulate depression along the anterior plicae. Nucha subglobose, not distinctly separated from rest of propodeum, often at least with some traces of alutaceous sculpture.
Remarks. The females of A. calandrae can be most easily recognized by the characters given in the key and diagnosis. The species is most similar to A. cornis sp.n. From A. quinarius sp.n. it is further distinguished by the pilosity of the forewing, for which we refer to the description of each species.
In order to assess the taxonomic status of A. calandrae, it is necessary to concentrate on the other cosmopolitan species, A. quinarius sp.n. Not only have the two species been confused over a long time, but they both are also likely to occur in human dwellings. However, these species usually occupy somewhat different habitats due to their different host preferences (Gokhman & Timokhov, 2002;Timokhov & Gokhman, 2003): A. calandrae inhabits mills and grain bins where it is usually associated with Sitophilus spp.; A. quinarius sp.n. inhabits houses and warehouses (e.g. containing stored fruit or tobacco) being associated with Stegobium or Lasioderma beetles there. We believe that these differences were an important factor that hampered the discovery of A. quinarius sp.n. (see Discussion).
A further complication is that the male holotype of A. calandrae is now lost (Peck, 1963: 733), which was recently confirmed by the curator of the Chalcidoidea collections at the USNM, M. Gates (personal communication). We have therefore studied characters of the males of the two species that were relevant with respect to Howard's original description. They can be distinguished as follows: • A. calandrae: First funicular (third flagellar) segment always shorter than the second (Fig. 7A), in small specimens only about as long as combined length of first and second anellus (Fig. 7B). Pale band in proximal half of gaster dirty yellow (Fig. 7D). • A. quinarius sp.n.: First funicular (third flagellar) segment about as long as the second, always much longer than combined length of first and second anellus (Fig. 7C). Pale band in proximal half of gaster bright yellow (Fig. 7E). Howard in Comstock (1881: 273), who apparently examined a very small specimen, noted in his brief but concise description 'joint 5 [corresponding to the third flagellar segment] small, equal in length to the two ring joints' and 'abdomen [corresponding to the gaster] yellow-brown at base'. A comparison of Howard's statements with our description of the male characters should make it clear that both concepts of A. calandrae match almost perfectly.
With regard to the overall morphological similarity with the other cosmopolitan species, A. quinarius sp.n., we designated a neotype for A. calandrae in order to guarantee stability in the application of the name. This is even more important, because A. calandrae is the type species of Aplastomorpha Crawford, the only synonym of Anisopteromalus (see Noyes, 2013). We selected a specimen from Savannah (Georgia), which might appear to be quite far away from Hempstead, Waller County (Texas), the place where the holotype of A. calandrae was collected (Howard in Comstock, 1881: 273). Given that the species is very easily disseminated from one site to another by transportation of stored products, the distance between the two localities is certainly not a problem here. Taking material of the laboratory culture from Savannah also had the advantage that a specimen in perfect condition coming from the centre of the analysed size and shape space was available (cf. Fig. 1B). The Savannah strain could furthermore be included in our genetic analyses (Fig. 4). Finally, in both cases the specimens were originally reared from Sitophilus oryzae-infested crop -that is, both demonstrably had the same host preference.
Gena terete, not carinate near mouth margin. Flagellum filiform (Fig. 5J), first funicular segment subconical, basally about as broad as third anellus, provided with 1-2 rows of longitudinal sensilla. Scutellum projecting at level of anterior margin of dorsellum, in lateral view weakly curved. Forewing with setae on wing disc dark. Speculum bare, open below but sometimes closed in proximal part. Anterior plica of propodeum short and evenly curved, joining an indistinct costula. Posterior margin of first gastral tergite curving backwards and strongly produced. Head breadth 1.15-1.26× metatibia length and 1.37-1.40× eye distance; head height 3.15-3.59× eye breadth; eye height 1. Mesosoma 1.20-1.30× as long as broad. Mesoscutum 1.92-2.16× as broad as long. Hind margin of scutellum broadly rounded. Upper mesepimeron strongly narrowing below, reaching at most basal third of mesopleuron. Basal setal line complete. Costal cell with dorsal surface with a short row of setae distally, lower surface with a patch of setae in distal half and a single row of setae running to proximal part, costal setal line complete. Forewing disc moderately pilose. Marginal vein 1.41-1.78× as long as stigmal vein. Stigma subcircular to slightly elongate, small. Propodeum 0.51-0.57× as long as scutellum. Median carina fine, straight. Median area in anterior part evenly reticulate with inner corner of anterior plica with a rather deep and reticulate depression along the anterior plicae. Nucha subglobose, not distinctly separated from rest of propodeum, alutaceous.
Etymology. The specific name 'cornis' is derived from Latin and means 'horned'. It refers to the relatively long antennae.
Remarks. The species is very close in habitus to A. calandrae from which it can be most easily separated by its relatively long flagellum. Diagnosis, female. Head and mesosoma olive-green with slight bronze tinge in places, setae whitish, inconspicuous. Gena compressed, with a short carina near mouth margin (Fig. 6C). Flagellum almost filiform, first funicular segment subconical, basally about as broad as third anellus, provided with 1-2 rows of longitudinal sensilla. Scutellum projecting at level of anterior margin of dorsellum, in lateral view weakly curved. Forewing with setae on wing disc dark. Speculum bare, open below but sometimes closed in proximal part (Fig. 5B). Anterior plica of propodeum consisting of small deep pits, costula indistinct (Fig. 6E). Posterior margin of first gastral tergite curving backwards, not produced medially (Fig. 5H) (Linnaeus, 1758), laboratory culture (ZIN; ZMMU). Further, non-type material is listed in Appendix S1.

Distribution. Cosmopolitan.
Remarks. The females of A. quinarius sp.n. can be recognized by the characters given in the key and diagnosis. For the separation of males from A. calandrae, see 'Remarks' in that species description.

Morphometry and data integration
Multivariate ratio analysis (MRA) revealed six distinct clusters that fully corresponded to the species delineation based on qualitative characters (Fig. 1B). For A. calandrae and A. quinarius sp.n. our DNA analyses as well as the published data on karyotype, behaviour and life-history strategies (summarized in Gokhman & Timokhov, 2002) also confirmed the morphometric differentiation of the two species (see Fig. 2 for interspecific and Fig. 3 for intraspecific variation). Concerning the molecular markers we would like to point to the fact that the comparison between ITS2 and Cytb genetic clusters showed no mitochondrial introgression (Fig. 4). Consequently, Cytb should be valuable for identification of the cultured strains of 'A. calandrae' and accurate assignment of specimens to either A. calandrae or A. quinarius sp.n.
The associations of names with species as delimited here was straightforward for A. apiovorus, A. caryedophagus, A. ceylonensis, A. cornis sp.n. and A. quinarius sp.n., because type specimens were available in good condition and thus could be measured and included in the morphometric analyses. The same was true for other nominal taxa where name-bearing types were available (Fig. 1B). All of these names could be confirmed as synonyms of A. calandrae. Cases where the type material was lacking posed rather more difficulties. Here, we took the available evidence from the original description and integrated it with our data. Once again morphometric character ratios proved to be most helpful. For instance measurements taken from drawings of figures published in the original description of A. calandrae brasiliensis (Domenichini) allowed us to exclude that this taxon belongs to any of the new species described here (see section on doubtful taxa in Appendix S1). In contrast, where morphometric data were unavailable, the establishment of the identity of a taxon was usually less certain (see, for instance, Pteromalus oryzae Cameron, a synonym of A. calandrae).
Finally we would like to stress two points with respect to the MRA approach as applied herein. First, its utility in finding the best discriminating ratios for separating various groups could be demonstrated (Table 2). Such best ratios functioned as diagnostic characters and formed an essential part of the key and diagnoses. The application of MRA thus allowed a seamless integration of the output from multivariate analysis with the descriptive part of our study. This was also the case in other studies using the MRA technique (László et al., 2013;Huber et al., 2013;Neumeyer et al., 2014;Schweizer et al., 2014;Shirihai et al., 2014) and is clearly advantageous compared to the application of the standard morphometric methods (e.g. Polaszek et al., 2004;Peters & Baur, 2011).
A second point should be mentioned here because we think it often goes unnoticed. It concerns the fact that a single, error-prone variable may have a devastating effect on a morphometric analysis. We demonstrated such an effect in our Fig. 1A where the deformation of the gaster in some specimens led us to remove the variable gaster breadth which resulted in a much improved analysis ( Fig. 1B; see László et al., 2013, for why gaster length is much less vulnerable to damage in parasitic wasps). We therefore suggest that the quality of specimens and individual measurements should always be carefully checked by visual inspection, matrix scatterplots and calculation of correlation coefficients (Table S3 and Figure S1).

Taxonomic implications
The present study obviously demonstrates how little we know about parasitoid biodiversity even in densely populated areas of developed countries, not to mention remote (e.g. tropical) ones. It also raises a question that has general ramifications: why was A. quinarius sp.n. not discovered earlier? One reason is that because this species is superficially similar to the well-known A. calandrae, its presence was not detected until now, although A. quinarius sp.n. is likely to have also been inhabiting human dwellings for at least a few millennia, as its preferred hosts, including Stegobium paniceum, were first recorded there as early as >4000 years ago (see King, 2010). Another reason is that the different host preferences of these two parasitoid species (Timokhov & Gokhman, 2003) led to the situation where quite a few researchers, especially those working in the field of practical pest management, came across A. calandrae and A. quinarius sp.n. simultaneously. Consequently, very few institutions kept both species in culture at the same time, thus preventing the discovery of their reproductive isolation. Furthermore, even in those cases, the observed isolation could be treated as a by-product of continuous adaptation to different host species. Subtle morphological differences between these lineages could also be explained by host influence, both in terms of their genetic background and modification by the environment. A final reason is that the discovery of A. quinarius sp.n. was hampered by pressure of the taxonomic tradition: all authoritative catalogues and manuals that treat the genus Anisopteromalus (see Peck, 1963;Graham, 1969;Bouček, 1988;Bouček & Rasplus, 1991;Noyes, 2013) list A. calandrae as its only cosmopolitan species, and strong evidence was therefore needed to call it into question.
Because karyotypic study became a starting point for the present work, we would also like to stress the importance of chromosomal analysis for studying parasitoid taxonomy. Karyotype structure, together with molecular characters, does not depend on environmental conditions, at least not directly. The results obtained, together with a number of similar cases (see Gokhman, 2009 for review), call for wider application of chromosomal analysis to parasitoid stocks cultivated for both industrial and laboratory use. This kind of analysis can provide a means of rapid identification of particular strains, as well as reveal certain karyotypic features that could affect particular genetic and life-history traits.

Supporting Information
Additional Supporting Information may be found in the online version of this article under the DOI reference: 10.1111/syen.12081 Figure S1. Matrix scatterplot of eight variables (in μm) of Anisopteromalus. Table S1. Overview of measurements of Anisopteromalus females, showing minimum, maximum, mean and standard deviation in μm (except for A. ceylonensis with n = 1). Table S2. List of Anisopteromalus specimens and outgroups included in the molecular analysis. Table S3. Pearson's product-moment correlation coefficients for all measurements of Anisopteromalus.
Appendix S1. Descriptions of species and lists of further material examined.