Phylogenetic relationships among the subfamilies of Dryinidae (Hymenoptera, Chrysidoidea) as reconstructed by molecular sequencing

Previously, the only published phylogenetic analysis of Dryinidae was a morphological analysis of just 32 characters. Herein, I present the first analysis of molecular sequence data examining the relationships among several of the major subfamilies of Dryinidae. A total of 77 specimens of Dryinidae from seven subfamilies, two specimens of Chrysis (Chrysididae), one specimen of Cleptes (Chrysididae), and one specimen of Sclerogibba (Sclerogibbidae) were examined utilizing molecular sequence data from nuclear 18S and 28S genes and mitochondrial Cytochrome Oxidase Subunit I (COI) and Cytochrome b (Cytb) genes. Dryininae were rendered nonmonophyletic due to the placement of Thaumatodryinus, which was sister to the remainder of Dryininae and Gonatopodinae. To establish monophyly of Dryininae, Thaumatodryininae were resurrected for Thaumatodryinus.

With only one or two world experts exclusively studying Dryinidae at any one time, the family has an interesting, but sparse, taxonomic history.Kieffer (1914) wrote the first world monograph of Dryinidae, with the first revisionary taxonomy for the family coming from Richards (1939Richards ( , 1953)).Outside of small agricultural studies and taxonomic descriptions, there was little focus on Dryinidae until the publication of Olmi (1984), a 1913-page world monograph that revised much of the taxonomy and provided keys throughout the family.Since then, there has been a growth in known dryinid diversity and host-records and the production of several large regional monographs (Olmi 1994a, b, Olmi 2005, Olmi 2007, Xu et al. 2013, Olmi and Virla 2014).
There is very little published on the phylogenetic relationships of the subfamilies within Dryinidae.Olmi (1994a) stated "we cannot discuss species affinities, because evolution has followed completely different paths in males and females, and female affinities are completely different from male affinities", and did not attempt to combine morphological data from both sexes to reconstruct a phylogeny.Olmi (1994a) presented a tree, but only included female specimens from four subfamilies found within Denmark and Fennoscandia and did not make clear how characters were coded and analyzed.In Carpenter (1999), a cladogram was reconstructed from 32 characters based on the taxonomic keys and descriptions of Massimo Olmi from both sexes.Given the growth in known dryinid diversity since then, neither study reflects the current subfamily classification and only addressed a small number of morphological features, although both placed Aphelopinae as the basal subfamily of Dryinidae and placed Gonatopodinae and Dryininae as sister groups (as in Olmi 1994a) or as closely associated in a polytomy that also contained Transdryininae and (Apodryininae + Plesiodryininae) (Carpenter 1999).There are no published molecular phylogenies, but DNA has been used to link the highly modified females of Gonatopus javanus Perkins, 1912 to males, which are similar looking throughout the genus, and to explore intraspecific genetic variation (Mita andMatsumoto 2012, Mita et al. 2013).Herein, I present the first analysis of molecular sequence data examining the relationships among several of the major subfamilies.

Materials
Phylogenetic relationships were inferred from 77 specimens of Dryinidae with one specimen of Sclerogibba Riggio & De Stefani-Perez, 1888 (Chrysidoidea: Sclerogibbidae), two species of Chrysis Linnaeus, 1761 (Chrysidoidea: Chrysididae), and Cleptes seoulensis Tsuneki, 1959 (Chrysidoidea: Chrysididae) as outgroup taxa.The majority of specimens came from two sources: Instituto Nacional de Pesquisas da Amazônia (INPA) and Canadian National Collections (CNC).Materials from the CNC were sorted from bulk alcohol materials from a variety of institutions and collectors, as detailed in Suppl.material 1.Additionally, several specimens were provided courtesy of Massimo Olmi, Toshiharu Mita and Pierre Tripotin.Specimens were stored in 95 percent ethanol and refrigerated prior to extraction.As these materials were acquired from unsorted bulk Malaise, yellow pan trap, and sweep net samples, they have not been accessioned in collections.Materials will be returned to their original institutions following the completion of this work and subsequent description of new species.

Analyses
Sequences were assembled and edited in Geneious 5.4 (Kearse et al. 2012).Mitochondrial genes COI and Cytb were checked for stop codons and numts and aligned using the translation alignment algorithm within Geneious.18S and 28S were aligned using MAAFT, using the E-INS-I algorithm as implemented in Geneious.This algorithm was chosen for its accuracy in difficult alignments (Morrison 2009) and its recent use in the Hymenoptera Tree of Life project, which provided sequences for outgroup taxon Chrysis cembricola Krombein, 1958(Klopfstein et al. 2013) The concatenated matrix was assembled in SeqMatrix (Vaidya et al. 2011), resulting in a final matrix of 6594 characters, with 13 percent missing data.
Phylogenetic analyses were performed using parsimony, Bayesian and maximum likelihood approaches.For parsimony, TNT (Goloboff et al. 2008) was used with the new technology search algorithms with the following parameters modified from default: 200 ratchet iterations, upweighting percentage 8, downweighting 4; 50 cycles of drift; minimum length hit 25 times with gaps treated as missing data.Jackknife resampling (Farris et al. 1996) support values were calculated using GC-values from a symmetric resampling of 1000 replicates.Separate analyses were performed using equal weighting and implied weighting as implemented by the setK script in TNT (courtesy of J. Salvador Arias).
PartitionFinder (Lanfear et al. 2012) was used to select models of molecular evolution for the RAxML (Stamatakis 2014) and MrBayes (Ronquist et al. 2012) analyses for each ribosomal gene, and each codon for COI and Cytb.For the models available for implementation in RAxML, each partition was returned as GTR+I+G.Using RAxML 8.1.11XSEDE on the Cipres server, 20 independent analyses were performed with different starting seed values and 1000 rapid bootstrapping (BS) replicates, choosing the tree with the best known likelihood (BKL) score amongst those independent searches (method adapted from Munro et al. 2011).Additionally, Garli 2.1 (Bazinet et al. 2014) on www.molecularevolution.org was utilized to see if the same topology was returned as the best tree, with 1000 bootstrap replicates.
For Bayesian analyses, Mr Bayes 3.2.3(Ronquist et al. 2012) XSEDE was utilized with the following partitions: K80+I+G for 18S and GTR+I+G for 28S, HKY+I+G for the 1 st positions in COI and Cytb, and GTR+I+G for the 2 nd and 3 rd positions in COI and Cytb.
In MrBayes, default parameters were used, with the exception of allowing enough time for 15,000,000 generations.

Results
The topologies of the equal weighting and implied weighting analyses in TNT (parsimony) were the same, with the equal weighting analysis recovering nine trees with a best score of 8562 steps (CI 0.287 RI 0.641) and the implied weighting (K = 20.527)analysis recovering nine trees with a best score of 200.The best RAxML tree from 20 separate analyses had a final optimization likelihood of -44251.166938(Fig. 2), and had the same topology as the tree produced by Garli, and the MrBayes analysis produced an average standard deviation of split frequencies (ASDSF) of 0.010179, with 25 percent of samples discarded as burn-in (Fig. 3).
Results were largely congruent for parsimony, likelihood, and Bayesian approaches in terms of higher-level topology (Figs 1-3), while species-level topologies were more variable.Apodryininae (as represented by Madecadryinus politus Olmi, 2007) were the sister taxon to all other Dryinidae in every analysis.The greatest difference among analyses were among Aphelopinae, Bocchinae and Conganteoninae.In all three trees, Aphelopinae and Conganteoninae were recovered as monophyletic, but since Bocchinae were represented by one species, its monophyly could not be tested.In the Bayesian analysis, Bocchinae were the sister group to the remainder of Dryinidae excluding Apodryininae, with Aphelopinae and Conganteoninae as sister groups.In the parsimony and likelihood analyses, Bocchinae were the sister group to Conganteoninae, with Aphelopinae sister to (Conganteoninae + Bocchinae).The remaining subfamily topologies were the same in all three analyses -Anteoninae, Aphelopinae, and Gonatopodinae were monophyletic, with Anteoninae as the sister subfamily to ((Thaumatodryinus + (Dryininae partim + Gonatopodinae).Dryininae were paraphyletic due to the placement of Thaumatodryinus merinus Olmi, 2004 andThaumatodryinus macilentus De Santis &Vidal Sarmiento, 1974, which were sister to a monophyletic Gonatopodinae and the remainder of Dryininae.
Many of the genera tested were found to be nonmonophyletic.Within Anteoninae, Lonchodryinus Kieffer, 1905 was the only genus found as monophyletic, as was Epigonatopus Perkins, 1905 in Gonatopodinae.Dryinus and Thaumatodryinus were the only genera from Dryininae tested, although all four of the Dryinus 'species groups' defined by Olmi (1993), were examined.Species groups were only defined for females, so undescribed male dryinid specimens could not be assessed.However, Dryinus Group 1 was found nonmonophyletic due to the placement of Dryinus striatus Fenton, 1927, although Dryinus Group 2 and Dryinus Group 4 were monophyletic.Dryinus Group 3 could not be assessed due to the sampling of a single specimen.Gonatopus was not monophyletic, nor were any of its species groups.

Discussion
The validity of Thaumatodryininae Olmi (1993) synonymized Thaumatodryininae with Dryininae, placing Thaumatodryinus close to the Dryininae genus Pseudodryinus.Olmi (1989) had originally attributed Pseudodryinus to Gonatopodinae on the basis of lacking a spur (1, 0, 2 tibial formula), but later examination of Pseudodryinus specimens by Olmi revealed a tibial formula of 1, 1, 2, allowing for the genus to be moved to Dryininae.At that time, previously unknown males of Pseudodryinus were discovered, and were shown to have quadridentate mandibles, as opposed to the tridentate mandibles found in all other male Dryininae (Olmi 1993).Olmi proposed that these males belonged to Thaumatodryininae, and then further noted that it would be unfeasible to have the females of Pseudodryinus within Dryininae and the males of Pseudodryinus within Thaumatodryininae.To preserve Pseudodryinus as a valid genus, Thaumatodryinus (the only genus within Thaumatodryininae) was synonymized within Dryininae.
In the molecular analyses presented here, the two different species of Thaumatodryinus were monophyletic and sister to Gonatopodinae + Dryininae.Molecular   data from Thaumatodryinus macilentus were taken from a female specimen, while molecular data from Thaumatodryinus merinus come from a male.Unfortunately, neither male nor female specimens of Pseudodryinus with viable DNA were available to test their placement within Dryininae or Thaumatodryininae.To establish the monophyly of Dryininae, and retain Gonatopodinae as a separate subfamily, I resurrect Thaumatodryininae, containing the genus Thaumatodryinus.The defining synapomorphy of Thaumatodryininae is the presence of long hairs on flagellomeres 3 -8 in females (Xu et al. 2013).

Evolution of the chela
The tree produced by Olmi only treated Aphelopinae, Anteoninae, Dryininae, and Gonatopodinae from Fennoscandia and Denmark (Olmi 1994a), and similarly found Dryininae and Gonatopodinae as sister groups (Thaumatodryinus was not included in the cladogram).Olmi (1994a) placed Anteoninae as sister to (Dryininae + Gonatopodinae), which was found in this study.Carpenter (1999) also found Anteoninae as the sister group to the clade that contained Dryininae and Gonatopodinae.This study diverges from these past two trees in the basal lineage of Dryinidae.In both Olmi (1994a) and Carpenter (1999), Aphelopinae were considered the basal lineage of Dryinidae on the basis of the lack of the characteristic pincer-like chela.Here, Apodryininae were found as the basal lineage of Dryinidae and while not all subfamilies of Dryinidae were considered, this suggests that the loss of the chela is a derived trait of Aphelopinae.Erwiniinae (known only from the type species) are also achelate, but were not included in this study.

Sampling of genera and species groups of Dryinus and Gonatopus
Several of the smaller subfamilies were not represented in this study because of their scarcity -Apoaphelopinae are known from two species, Erwiniinae from one species, Plesiodryininae from one species, and Transdryininae from two species.
Sampling of the genera of the subfamilies was also incomplete.Within Dryininae, only Dryinus was treated, although all four of the species groups were included.Megadryinus Richards, 1953 (known from three species), Gonadryinus Olmi, 1991 (known from one species), and Pseudodryinus (known from ten species) were absent.Given the shared characteristic of having quadridentate mandibles in males, Thaumatodryinus and Pseudodryinus might be related, but without a specimen from which viable DNA could be sequenced, the placement of Pseudodryinus could not be assessed.
Within Anteoninae, three out of four extant genera were included, with Metanteon Olmi, 1984 (known only from the type species) not included.Conganteoninae were only represented by one genus, Fiorianteon Olmi, 1984, and

Conclusion
In all analyses, Thaumatodryinus was well-supported and Thaumatodryininae were resurrected here, bringing the total subfamilies of Dryinidae to 16.
The validity of species groups within Dryinus and Gonatopus remains questionable.Some species groups, like Dryinus Group 4, which was originally a separate genus, Perodryinus, were easily recovered as monophyletic while Dryinus Group 1, which contains several synonymized genera, was not recovered as monophyletic.This may be because the larger species groups share synonymized genera -for example, Dryinus species groups 1, 2, and 3 all contain synonymized species from Mesodryinus.Shared synonymized genera are found within the Gonatopus species groups as well.
In continuing molecular studies, specimens from each of the species groups of Gonatopus and Dryinus should be included, as well as all of the genera of the subfamilies, where sampling permits.In particular interest would be to find morphological synapomorphies at the generic level for male Dryinidae.

Figure 1 .
Figure 1.Parsimony support tree.Jackknife support for nodes given in GC-values (frequency differences) from 1000 replicates.CI 0.287 RI 0.641.Scale bar in all images is 1.0 mm.

Figure 2 .
Figure 2. Likelihood support tree.Rapid Bootstrap support values shown at nodes.

Figure 3 .
Figure 3. Bayesian support tree.Support probabilities shown at nodes as a percent.
did not include the other genus, Conganteon Benoit, 1951.Bocchinae were only represented by Bocchus Ashmead, 1893, and did not include Mirodryinus Ponomarenko, 1972 and Mystrophorus Förster, 1856.Aphelopinae were only represented by Aphelopus Dalman, 1823, and did not include Crovettia Olmi, 1984.Apodryininae were only represented by Madecadryinus Olmi, 2007, and did not include the six other genera.