Parapanteles Ashmead (Braconidae: Microgastrinae) is a medium-sized genus of microgastrine wasps that was erected over a century ago and lacks a unique synapomorphic character, and its monophyly has not been tested by any means. Parapanteles usually are parasitoids of large, unconcealed caterpillars (macrolepidoptera) and have been reared from an unusually large diversity of hosts for a relatively small microgastrine genus. We used Cytochrome Oxidase I sequences (“DNA barcodes”) available for Parapanteles and other microgastrines to sample the generic diversity of described and undescribed species currently placed in Parapanteles, and then sequenced four additional genes for this subsample (wingless, elongation factor 1-alpha, ribosomal subunit 28s, and NADH dehydrogenase subunit 1). We constructed individual gene trees and concatenated Bayesian and maximum-likelihood phylogenies for this 5-gene subsample. In these phylogenies, most Parapanteles species formed a monophyletic clade within another genus, Dolichogenidea, while the remaining Parapanteles species were recovered polyphyletically within several other genera. The latter likely represent misidentified members of other morphologically similar genera. Species in the monophyletic clade containing most Parapanteles parasitized caterpillars from only five families - Erebidae (Arctiinae), Geometridae, Saturniidae, Notodontidae, and Crambidae. We do not make any formal taxonomic decisions here because we were not able to include representatives of type species for Parapanteles or other relevant genera, and because we feel such decisions would require a more comprehensive morphological analysis of the boundaries of these genera to be meaningful.

2.1 METHODS

2.2 Taxon & gene selection

2.2.1 COI-tree based within-group taxon sampling

To approximate the diversity of Parapanteles under its current morphological definition, we accessed all available COI sequences for Parapanteles and 16 other microgastrine genera: Alphomelon Mason, Apanteles, Clarkinella Mason, Cotesia, Diolcogaster Ashmead, Dolichogenidea, Exoryza Mason, Glypapanteles, Hypomicrogaster, Microplitis Förster, Pholetesor Mason, Prasmodon Nixon, Promicrogaster Bues & Richardson, Protapanteles Ashmead, Rhygoplitis Mason, and Xanthomicrogaster Cameron from the Barcode of Life Database (BOLD) (Ratnasingham & Hebert 2007; http://v3.boldsystems.org/) and GenBank (http://www.ncbi.nlm.nih.gov/genbank/) on 01/11/2017. The long-term caterpillar rearing project at Área de Conservación Guanacaste (ACG), Costa Rica (Janzen & Hallwachs 2009, 2016) has provided the majority of known yet undescribed Parapanteles specimens, which have been morphologically identified to genus by professional braconid taxonomists and DNA barcoded (Janzen & Hallwachs 2009, 2016), and is the source of approximately 43% of the DNA barcode sequences and putative species used in this study (Supplemental Materials 17). This dataset includes the most species-rich microgastrine genera, genera which Parapanteles has been recovered sister to in previous phylogenies, genera with morphological similarities to Parapanteles, and several small microgastrine genera. In addition, we sequenced and included COI for 110 more Parapanteles specimens reared at Yanayacu Biological Station in Ecuador (Dyer et al. 2017). We aligned this dataset with PASTA v1.6.3 (Mirarab et al. 2014) and edited it manually in Geneious v9.1.5 (http://www.geneious.com, Kearse et al. 2012), discarding sequences that were missing approximately 200 or more base pairs (bp) of the 658 bp COI barcoding region. Our dataset then contained 24,611 aligned sequences from at least 1,713 described or putative species (the number of species may be higher because many sequences were only identified to genus) (Supplemental Materials 2 & 3), which we used to make a phylogenetic tree with FastTree 2.1.8 (Price & Arkin 2010) using the GTR+I+G substitution model (Supplemental Materials 4). We used this tree to selected 56 Parapanteles specimens recovered from disparate clades in this tree to sequence additional genes. This subsample included representatives of Parapanteles specimens recovered in 12 of 15 major clades throughout our initial COI tree. Dolichogenidea specimens were frequently intermixed with Parapanteles specimens in this tree, so we also included 13 Dolichogenidea species in our subset.

2.2.2 Gene selection and outgroups

We sequenced portions of two mitochondrial genes [655 bp of cytochrome oxidase I (COI) (the “DNA barcoding” region) and 447 bp of NADH dehydrogenase subunit 1 (ND1)] and three nuclear genes [451 bp of wingless (WG), 418 bp of elongation factor 1-alpha (EF1a) and 666 bp of ribosomal subunit 28s (28S)] (Supplemental Materials 5) to construct a molecular phylogeny. We used the “DNA barcoding” region of cytochrome oxidase I because of the availability of other microgastrine COI sequences through the Barcode of Life Database (BOLD, Ratnasingham & Herbert 2007) because it is commonly used for species delimitation in this group (Whitfield et al. 2002, Banks & Whitfield 2006, Smith et al. 2008, 2013, Janzen & Hallwachs 2009, Fernández-Triana et al. 2014). Wingless has been used in generic-level phylogenies of microgastrines (Banks & Whitfield 2006; Murphy et al 2008), and sequences are available for all Parapanteles that attack Eois Hübner (Geometridae: Larentiinae) caterpillars to date (Wilson et al. 2012) and all outgroup taxa with pre-existing sequences (Banks & Whitfield 2006). Similarly, EF1a has been used extensively in insect systematics and sequences were available from the Parapanteles attacking Eois (Wilson et al. 2012). Previous microgastrine phylogenies have used ND1 and 28s (Dowton & Austin 1998, Michel-Salzat & Whitfield 2004, Kankare & Shaw 2004, Rodriguez 2009, O’Connor 2011), and we were able to incorporate existing outgroup sequences by including these genes in our dataset.

We included all available non-Parapanteles microgastrine specimens that have sequences available in GenBank for at least 3 of the 5 genes used in this study. This outgroup set includes 11 species from 8 other microgastrine genera: Apanteles, Cotesia, Dolichogenidea, Glyptapanteles, Hypomicrogaster, Microplitis, Pholetesor, Prasmodon, Promicrogaster and Rhygoplitis (Supplemental Materials 6). We used one Cheloninae species (Phanerotoma Wesmael), the earliest-diverging group in the microgastroid complex (including Microgastrinae and four other smaller Braconidae subfamilies: Cardiochilinae, Cheloninae, Mendesellinae and Miracinae) (Dowton & Austin 1998, Murphy et al. 2008), to root trees. Preliminary results placed many Parapanteles among several other genera, so we included additional specimens from Glyptapanteles and Apanteles, including sequences from two unpublished molecular phylogenies of these genera (Rodriguez 2009, Arias-Penna 2015). We used COI and WG sequences from 28 undescribed species of Glyptapanteles, to which we added sequences of EF1a, 28s, and ND1. We sequenced all 5 genes in our dataset for 17 additional Glyptapanteles specimens from Yanayacu, Ecuador (Arias-Penna 2015). We used existing sequences of all five genes for 19 Apanteles species (Rodriguez 2009).

2.3 Sequencing

Genomic DNA was extracted from adult microgastrines using Qiagen DNEasy Blood and Tissue kits (Qiagen, Valencia, CA, USA) following the manufacturer’s directions. For gregarious species (multiple conspecific larvae developing in the same host), we extracted DNA from whole specimens. For solitary species, we extracted DNA from one hind leg, removed above the coxa, or one mid- and/or foreleg if one or more hind legs were missing. We used New England Biolabs Taq DNA Polymerase with Standard Taq buffer and the primers and thermocycler protocols listed in Supplemental Materials 5. We purified PCR products with EXO SAP and performed sequencing reactions with ABI Prism BigDye Terminator v3.1 Cycle Sequencing Kits, typically using 1/8^th-1/16^th of the recommended amount of BigDye Terminator 3.1 Ready Reaction Mix (1µl-0.5µl) but otherwise following the manufacturer’s instructions. PCR products were sequenced at the W.M. Keck Center for Comparative and Functional Genomics at the University of Illinois. We edited sequences with Geneious v9.1.5 (http://www.geneious.com, Kearse et al. 2012). All novel sequences are deposited in GenBank (Supplemental Materials 6).

2.4 Alignment and Phylogenetic Analysis

We excluded from concatenated analyses any taxon for which we were unable to sequence at least three genes, but still included all available sequences in individual gene trees. We therefore included 142 species in our concatenated alignment, with the following numbers of species missing for each gene followed by the number of species included in each individual gene tree in parentheses: COI: 0/142 (295), WG: 4/142 (160), ND1: 50/142 (126), EF1a: 27/142 (135), 28s: 18/142 (139). We aligned sequences with MUSCLE v3.8.31 (Edgar 2004). Our concatenated alignment had 2626 characters total, with 169 invariable sites across all taxa. We used Partitionfinder v1.1.1 (Lanfear et al. 2012) to select appropriate models for phylogenetic analysis based on their Bayesian Information Criterion (BIC) score. In all analyses we partitioned COI and ND1 alignments into three partitions by 1^st, 2^nd, and 3^rd codon positions, WG and EF1a into two partitions by 1^st+2^nd and 3^rd codon positions, and 28s into two partitions, with the conserved regions flanking the D2 variable region in one partition and the variable region in the other, for a total of 12 partitions (Supplemental Materials 7). We constructed Maximum Likelihood (ML) trees in RAxML v8.1.15 (Stamatakis 2014) with 1000 bootstrap replicates for each gene independently and for all genes concatenated. For each analysis we selected either GTR+G or GTR+I+G depending on which model was favored by the majority of partitions. We constructed an additional tree for each analysis with MrBayes v.3.2.2 (Ronquist et al. 2012) using mixed models. We ran each Bayesian analysis for 10 million generations with 4 MCMC chains, and sampled trees every 1000^th generation. Appropriate burn-in values were estimated in Tracer v.1.5 (Rambaut & Drummond 2007). All trees except ND1 trees were rooted with the outgroup Phanerotoma, representing Cheloninae, the most closely related subfamily to Microgastrinae, as the most distant outgroup. ND1 trees were rooted with Microplitis demolitor, as Microplitis has been shown to be the most early-diverging genus of microgastrines in our sample (Whitfield et al. 2002, Banks and Whitfield 2006). We graphically edited all trees in FigTree v.1.4.2 (http://tree.bio.ed.ac.uk/software/figtree/) and poorly supported branches were manually collapsed in Adobe Illustrator CC 2015.3. All alignments and unedited tree files are deposited in Dryad (DOIs pending).

2.5 Specimen and provisional species naming conventions

Specimens collected by the ACG rearing project in Costa Rica are either assigned to described species or assigned interim provisional species epithets based on COI sequence similarity as displayed in a Nearest Neighbor Joining tree and host associations (Janzen et al. 2009, Janzen & Hallwachs 2009, 2016). Interim species names follow the convention of the last name of an ACG taxonomic collaborator and a number (e.g., Apanteles Rodriguez01, Parapanteles Whitfield113). Specimens of undescribed species from ACG used in this study are identified by these interim species epithets. The voucher codes for them, in the form DHJPAR#######, refer to the specific wasp specimen, while the individual caterpillar from which it was reared has its own voucher code in the form of yy-SRNP-#####. The voucher specimens are deposited in the Canadian National Insect Collection (CNC) in Ottawa, Canada and at the University of Illinois Urbana-Champaign to eventually be transferred to the CNC.

Specimens from the Yanayacu rearing project in Ecuador are not routinely sequenced or grouped into interim species as they are in the ACG inventory, so specimens of undescribed species from Yanayacu used in this study are identified by “yy” and their individual sample number (e.g. Parapanteles yy3653). This identifies the specific specimen used, not an informal species name. We grouped specimens from the Yanayacu rearing project into interim species based on COI sequence similarity, natural history, and then morphological similarity. We calculated the pair-wise distances of COI sequences with MEGA v7.0.26 (Kumar ­et al. 2016).