The subfamily Microgastrinae (Hymenoptera: Braconidae) is one of the most speciose subfamilies of parasitoid wasps, with more than 2, 200 described species (Yu et al. 2012) and many thousands more awaiting description (Mason 1981; Rodriguez et al. 2012). Microgastrine wasps are significant in biological control because they attack the larvae of most families of Lepidoptera (Whitfield 1995, 1997).

The genus Apanteles was erected by Förster (1862) to include all species of microgastrines lacking a second submarginal cell in the fore wing (from the Greek: A – without, panteles – complete, entire; referring to the “incomplete” venation, i.e., missing cell, when compared with the other genera of Microgastrinae known at the time). As the study of Apanteles progressed, it became evident that it included a huge number of species, and many attempts to subdivide the genus have been made since 1880; there are summarized in Mason (1981) and Whitfield et al. (2002). During the last 150 years more than two dozen new genera have been created as a result of those splitting efforts, but still more than one thousand described species belong to Apanteles (Yu et al. 2012), and thousands more await discovery. It is worth mention that many of these species still belong to Apanteles sensu lato, and have not yet been assigned to currently recognised genera (sensu Mason 1981).

Area de Conservación Guanacaste (ACG) is a single decentralized unit of Costa Rica’s Ministerio del Ambiente, Energia (MINAE; Ministry of Environment and Energy) covering about 2% of Costa Rica in its northwestern corner, slightly south of the southeastern border of Nicaragua (http://www.acguanacaste.ac.cr). Comprising ~1, 200 km² of terrestrial habitat (centered at 10.8 latitude, -85.6 longitude), it is a swath from Pacific coastal mangroves across lowland dry forest (dry season deciduous), up the slopes of three volcanoes to cloud forest (1400–2000 m), and down into Caribbean lowland (90 m) rain forest. It is only 85 km from east to west, yet contains portions of eight Holdridge Life Zones within mosaics of them, some as small as 5 km in linear dimensions and 20 km². Nearly all of the ACG lowlands have been subjected to four centuries of light to intense cultivation, logging, burning, hunting, ranching, and other forms of habitat destruction, followed by explicit protection and restoration beginning in 1971 and intensifying after 1985 (Janzen 1988, 2000, 2002). The outcome is a mosaic of all imaginable ages and kinds of secondary succession intermingled with tiny to medium-sized fragments of approximations of intact forest (more intact in upper elevations than lower), as well as severe blurring and elimination of interdigitated boundaries between habitats and ecosystems (Janzen 1986-1988). All of the ACG region has also now experienced at least two decades of notable drying and increasing weather unpredictability, rendering it yet more difficult to know if the marked annual and decadal population changes are being generated by climate changes, successional changes, insularization of the ACG ecological island in the agroscape, species-by-species biological serendipity, and/or interactions among all of these (Janzen et al. 2011).

ACG has been the focus of 34+ years of inventory of wild-caught caterpillars, their food plants and their parasitoids, as described in detail in Janzen et al. (2009) and Janzen and Hallwachs (2011), and available in a rearing-by-rearing specimen-based public database at Janzen and Hallwachs (2013). The ACG is currently staffed and supported by about 180 Costa Ricans, nearly all of whom are honoured with patronynms in this paper.

Haphazardly placed Townes Malaise traps in all three major ACG terrestrial ecosystems have yielded another set of ACG Apanteles species, many of which have not yet been reared and are included here (and are so indicated as distinct from the species that have been reared, many of which have not yet been encountered by Malaise-trapping).

The rearing results have been complemented since 2003 by extensive DNA barcoding of one or more voucher specimens from each rearing, past and present (Janzen and Hallwachs 2011). This has provided an additional layer of data to study the ACG species of caterpillars, parasitoids, and food plants (e.g., Smith et al. 2006, 2007, 2008; Whitfield et al. 2012; Janzen et al. 2011, 2012).

DNA barcoding uses a short standardized region of the mitochondrial gene cytochrome c oxidase (COI) as a key character for species-level identification and discovery (Floyd et al. 2002, Hebert et al. 2003a and b, Janzen et al. 2009, Smith et al. 2006, 2007, 2008). Interspecific barcode variation can be used as part of a suite of characters for the discovery and description of new species (e.g., Hebert et al. 2004, Burns et al. 2008, Fisher and Smith 2008, Fernández-Triana 2010), and can speed the rate of taxonomic research by flagging otherwise cryptic diversity (e.g., Janzen et al. 2009, Fisher and Smith 2008, Smith and Fisher 2009, Smith et al. 2008). DNA barcoding has been extensively used in biodiversity and taxonomic studies of Microgastrinae during the past five years (e.g., Smith et al. 2008 and 2013, Janzen et al. 2009, Fernández-Triana 2010, Fernández-Triana et al. 2011, Rodriguez et al. 2012, Whitfield et al. 2012, Fernández-Triana et al. 2013).

Taxonomic studies of ACG Microgastrinae have been published elsewhere (e.g., Valerio et al. 2005, Grinter et al. 2009, Smith et al. 2008, Valerio et al. 2009, Janzen and Hallwachs 2011, Janzen et al. 2009, Whitfield et al. 2012, Arias-Penna et al. 2013, Fernández-Triana et al. 2013). However, the ACG species of Apanteles sensu stricto have never been treated in a taxonomic review.

The combination of this comprehensive inventory with the richness of biological, ecological and DNA barcoding data, allowed us to engage in the taxonomic study of ACG Apanteles as a whole, and within the context of the other hundreds of species of ACG Microgastrinae. In doing so, we also revised all 19 of the previously described Apanteles sensu stricto known from Mesoamerica and incorporate them here. However, no effort was made to study specimens representing undescribed species from areas outside ACG, areas that will certainly contain hundreds of other species of Apanteles as well as many of those in ACG. We hope that this study will be a foundation upon which future studies of tropical Apanteles and other microgastrine genera can be based.

In this study, Mesoamerica is defined as the region from (and including) Mexico through Panama, and all the Caribbean islands, following Gauld (1988).

We studied 4, 100+ specimens from 3, 200+ individual caterpillar rearings, and 2, 000+ DNA sequences (usually one sequence per rearing event) of Apanteles from ACG. Ecological, biological and distribution data for all of these records can be accessed at http://janzen.sas.upenn.edu/caterpillars/database.lasso (Janzen and Hallwachs 2013) by searching on the “DHJPARxxxxxxx” voucher code of the wasp, or the “yy-SRNP-xxxxx” voucher codes of the caterpillar. If a DHJPARxxxxxxx voucher code is cited, it is for a single specimen. If a yy-SRNP-xxxxx voucher code is cited, it is for 1 to N specimens reared from a single caterpillar and which are presumed siblings, but have not been individually vouchered, whether point-mounted or remaining preserved in ethanol. All holotypes bear a DHJPARxxxxxxx unique voucher code (and if there was more than one specimen in that rearing from that one caterpillar, all of them will bear the same yy-SRNP-xxxxx code). In this paper we refer to these voucher codes as “ACG database codes” when providing specimen details in the taxonomic treatment of species. In the case that a set of specimens reared from one individual caterpillar was not DNA barcoded, the vial containing those specimens has only the yy-SRNP-xxxxx code, while an individual wasp that has been barcoded from that sample bears both the SRNP code and the DHJPAR code. Each barcoded specimen also has an accession code from the Barcode of Life Data System (BOLD) and GenBank.

Type material for most of the 19 previously described Mesoamerican species was borrowed for study. However, no molecular data is available for any of those holotypes. It will not be surprising if some of their names are found to encompass complexes of species. Some members of such complexes may be some of the ACG species described here, but it would be premature to even speculate about that.

The following acronyms are used:

BMNH The Natural History Museum, London, United Kingdom

CNC Canadian National Collection of Insects, Arachnids and Nematodes, Ottawa, Canada

INHS Illinois Natural History Survey, Champaign, Illinois, United States

INBio Instituto Nacional de Biodiversidad, Santo Domingo de Heredia, Costa Rica

NMNH National Museum of Natural History, the Smithsonian Institution, Washington DC, United States

Morphological terms and measurements of structures are mostly as used by Mason (1981), Huber and Sharkey (1993), Sharkey and Wharton (1997), Whitfield (1997), and Valerio et al. (2009). However, we also incorporated a recent, comprehensive morphological treatment of Opiinae (Braconidae) by Karlsson and Ronquist (2012), which is part of a wider effort to standardize and homologize morphological terms and definitions across the order Hymenoptera, the Hymenoptera Anatomy Ontology (HAO) project (Yoder et al. 2010, Seltmann et al. 2012). As a result of adopting most HAO preferred terms (but see exceptions below), some of the morphological terms we apply have never been used in taxonomic papers treating Microgastrinae.

Karlsson and Ronquist (2012) named and numbered the first metasomal segment as “abdominal tergum/sternum 2”. Usually, that segment has been called (and numbered) “metasomal tergum/sternum 1” (Mason 1981, Whitfield 1997, 2006). Though both approaches are correct, we use “metasomal tergum/sternum 1” because we consider it is clearer and facilitates the counting of metasomal segments (as in Fig. 207). The same applies to its associate sclerites (mediotergites and laterotergites).

We considered that the “preferred label” (i.e., name) provided in the HAO website for “mesoscutellar arm” was better than the corresponding term, “posterior bar of mesoscutellum”, used by Karlsson and Ronquist (2012).

The terms “mesoscutellar trough” and “mesoscutellar arm” (Fig. 206) have been used extensively in taxonomy of Microgastrinae (e.g., Mason 1981, and many subsequent papers), usually under the name of “lateral face of scutellum”. Karlsson and Ronquist (2012) did not provide much detail for those areas because the two species of Opiinae they studied are relatively simple and non-differentiated in that body region.

We also calculated and compared many ratios between linear dimensions of structures (morphometric taxonomy), a common practice in the taxonomy of many groups of parasitoid wasps (Baur and Leuenberger 2011, and references cited there). However, most of the ratios presented here have not been used previously in Microgastrinae taxonomy.

To facilitate understanding of the traits and ratios, a detailed account of every morphological structure and measurement used in this study is provided in Appendix 1, including links to the HAO website and references to terms that have been commonly used previously in Microgastrinae taxonomy. The most important morphological characters used in this study are illustrated in Figs 206–209.

Throughout the text, especially in the keys, “body length” refers to the length of the anatomical line that is median and extends between the anteriormost point of the head and the posteriormost point of the metasoma (excluding ovipositor and ovipositor sheaths). “Fore wing length” refers to the length of the anatomical line that extends between the median margin of the first axillary sclerite and the distalmost point of the wing blade (Appendix 1).

The measurement of variables must be done as uniformly as possible, and special care must be taken when choosing the end points of any structure. It is also advisable to measure at the highest possible magnification to minimize errors. Some measurements that are particularly error-prone are discussed further in Appendix 1.

Throughout the keys the following acronyms are used for morphological terms: T1, T2, T3 (mediotergite 1, 2, 3). Whenever there is a “(N = a number)”, e.g., “(N = 4)” after a species name, it refers to the number of specimens studied morphologically for that species. It is only provided when the available number of specimens was less than 5.

Molecular analysis has revealed a large number of morphologically cryptic species, often possessing very subtle morphological differences that we found to correlate with ecological and host data. Certain features differ just slightly between species, and there may be overlap of values between individual specimens of different but very similar species. We studied as many specimens as were available. Our definition of a “species” is a postulated biological unit that differs from other species in its morphology (however subtle), COI barcode, and host use, and presumably represents a distinct breeding population. In the few cases where what we consider to be a species differs only in barcode and/or host, we indicate this. All the species in ACG are essentially fully sympatric to parapatric (the case when two ecosystems intergrade).

The dichotomous keys were built to accommodate, as much as possible, what appear to be potential natural groups, based on morphology, biology (host data), and DNA barcoding. However, in such a large assemblage of species there is likely to be considerable homoplasy and thus in some couplets we had to use logical characters (e.g., “if”, “then”, “and”, “or”, “and/or”). Those words are shown in bold and italic throughout the keys, to be explicit that in those cases more than one character system has to be considered.

The species descriptions are based on the holotype female, and we consider their DNA barcodes to be definitive when available (actual barcodes are available through the BOLD web site at http://www.boldsystems.org); when other specimens are available, their data are included to provide some idea of intraspecific variation. When the holotype was not examined or was lacking some body parts (some old holotypes), other specimens were used to complete the descriptions, and details were explained in the “Comments” section.

Males of Microgastrinae are difficult to key out under the present knowledge of the subfamily (Whitfield 1997, 2006), and may be difficult or impossible to identify unless associated with sibling females from the same host caterpillar, DNA barcodes, or host data. Thus, the keys are only intended for female specimens and use many characters only found in that sex (e.g., length of ovipositor sheaths).

Non-morphological characters are also provided whenever available, e.g., host species, whether there are one or many larvae per host caterpillar, microhabitat, microgeographical distribution, and molecular differences in the DNA barcode region, that may serve as diagnostic characters. Sometimes those features are included in brackets at the end of the corresponding couplet, intended as supplementary information that can help the user to correctly identify specimens. They are best not separated from the morphological features provided. However, in future, practical and routine identification may often depend heavily on these other traits because they are easier to assess than the morphology of wasps 1–5 mm in length.

Lucid 3.5.4 (http://www.lucidcentral.com/) software was used to automatically generate descriptions of the species and to prepare Lucid identification keys. A dataset of 41 characters and 239 character-states was used to provide uniform description formats for all species treated (except for the leucostigmus species-group, see next paragraph). The description format includes one sentence per character, with the character mentioned first and the character-state following after a colon, e.g., “Tarsal claws: simple”. Whenever a species scored more than one character-state, the description included all of the pertaining character-states separated by “or”, e.g., “Tarsal claws: simple or simple with single basal spine-like seta”. Whenever a character-state was coded as uncertain due to poor condition of the specimen(s), the description includes the details of the character-state as best assessed, followed by a question mark, e.g., “Tarsal claws: simple (?)”. Sometimes a character could not be coded due to missing body parts in the available specimens; in such instances the feature was left out of the description for that particular species.

The leucostigmus species-group was found to be exclusively composed of morphologically cryptic species (e.g., Smith et al. 2008). In this group we only provide “complete” classical descriptions (i.e., similar to other groups) for the two species previously described, Apanteles albinervis (Cameron, 1904) and Apanteles leucostigmus (Ashmead, 1900). Another 37 new species found to belong to this group are only described using a simplified set of 10 characters and 46 characters-states, and are most definitively defined by the combination of their barcodes and their hosts.

Accurate color description is a major difficulty in taxonomic works (Aguiar 2005), and often its variation within and between species can lead to confusion. In Microgastrinae, as in many other groups, the color pattern of body regions seems to be more important and taxonomically informative than the definition of the color per se. Accordingly, in the Lucid dataset, we used a simplified convention to code color, considering it as either pale (white, light yellow, orange-yellow, light brown-yellow) or dark (dark brown, black). For details on the exact color patterns on the body, we provide extensive illustrations for every species. When describing leg color (especially metacoxa, metafemur, and metatibia), we are referring to the outer side of the leg, e.g., sensu figure 1 of Goulet and Huber (1993).

Most of the photos were taken with a Canon EOS 60D with MPE-65 lenses (aperture: 4.0, ISO: 100, CR2 format images), and a 600EX-RT Speedlight (manual) flash. The camera was mounted on a Kaiser copy stand with a Z-stepper (Stackshot) to allow for taking of multiple images. A dozen species were photographed with a Keyence VHX-1000 Digital Microscope, using a lens with a range of 13–130 ×.

Multiple images through the focal plane were taken of a structure and these were combined to produce a single in-focus image. For the pictures taken with the Canon camera, the Zerene Stacker program (http://zerenesystems.com/cms/stacker) was used; the software associated with the Keyence System produced the focused images taken with that camera. Plates for the illustrations were prepared using Adobe Photoshop CS4.

Although the keys provided in this paper are based on morphological characters, in a few cases (mainly with species belonging to the leucostigmus group) we used molecular characters to differentiate species that are morphologically exceptionally similar to each other. In those cases we used characteristic loci in the DNA barcoding region in identification couplets. The bases are numbered from the start of the COI gene according to the reference sequence U37541 (Drosophila melanogaster). The bases noted are only diagnostic within the couplet and beneath that described split. A hypothetical example of the format used is: “A total of 11 diagnostic characters in the barcoding region: 12 A, 18 C, 22 A, 23 T, 44 C, 56 G, 120 G, 340 A, 488 T, 502 T, 601 A”. The letters A, C, G, and T correspond to adenine, cytosine, guanine, and thymine respectively.

DNA barcodes for all ACG inventory Microgastrinae (Appendix 2) were obtained using DNA extracts prepared from single legs using a glass fibre protocol (Ivanova et al. 2006). Extracts were re-suspended in 30 μl of dH2O, and a 658-bp region near the 5’ terminus of the COI gene was amplified using standard primers (LepF1–LepR1) following established protocols (Smith et al. 2006, 2007, 2008). If the initial 658 bp amplification was not successful, composite sequences were generated using internal primers. Primer information for individual sequences can be retrieved from the Barcode of Life Data System (BOLD) (Ratnasingham and Hebert 2007), but primers are as detailed in Smith et al. (2008). DNA barcoding data and related information for all specimens studied in this paper can be accessed at: https://doi.org/10.5883/DATASET-ACGAP1.

Sequences were considered as “barcode-compliant” when they had 500 or more base pairs and correctly placed the species with its conspecifics in an NJ tree, the function of which is simple identification (as well as species discovery). We did not follow the more formal, but less pragmatically useful, definition of also having less than 1% ambiguous characters (Barcode Compliance standards as in http://www.boldsystems.org/index.php/resources/handbook?chapter=6_managingdata.html&section=record_list), though in selecting the holotype we strive to choose a specimen with a “full” and “complete” DNA barcode. The number of sequences available in BOLD for each species that we treat is provided in the “Molecular data” section within the taxonomic treatment of every species and all sequences can be obtained from BOLD.

In the taxonomic treatment of species, “Material Examined” presents the specimen’s information in the following format: “Number of females/males, acronym of the storing institution between parenthesis, COUNTRY: State/Province, city, other locality details, coordinates (in Decimal Degrees), date, collector name, biological information on host (starting with “ex”), ACG database codes (in the format “yy-SRNP-xxxxxx” for the host or “DHJPARxxxxxxx” for the wasp)”. For states of the United States and for Canadian provinces/territories, acronyms consisting of two capital letters are used, following Canada Post (http://www.canadapost.ca/tools/pg/manual/PGaddress-e.asp).

For phylogenetic analyses conducted outside of BOLD, DNA barcode sequences from 180 Mesoamerican Apanteles species were downloaded from BOLD along with 29 sequences from species of other Microgastrinae genera (Microplitis, Pholetesor, Parapanteles, Dolichogenidea). All sequences were imported into Geneious Pro 6.1 (Drummond et al. 2011) and aligned using the default settings for MUSCLE. The first and last nucleotide positions of the aligned dataset were deleted to reduce the amount of missing data. All sequences downloaded were over 500 bp long and the final aligned dataset contained 657 characters of which 315 were parsimony informative. The dataset was partitioned into two partitions, the first containing first and second codon positions and the second containing third codon positions. Model testing done in JModelTest v0.1.1 (Posada 2008) using the Bayesian Information Criterion selected the GTR+I+G model for both partitions. Two independent Bayesian analyses with 4 chains each were run in MrBayes v3.2.1 (Ronquist and Huelsenbeck 2003) for 31 253 000 generations each. Trace files of all parameters were examined in Tracer v1.5 (Rambaut and Drummond 2009) to verify that the runs had converged on the same stationary distribution, and to select the percentage of samples to remove as burn-in. A 10% burn-in was removed from both tree files which were then combined and resampled at 10%. A Maximum Clade Credibility Tree (MCCT) was made from the combined resampled tree file in TreeAnotator v1.7.5 (Rambaut and Drummond 2013). The MCCT only shows posterior probability values over 0.50 (Fig. 1). When discussing the support to particular clades (species-groups), those values are mentioned within the text in the format: “PP: 0.95” (i.e., posterior probability of 0.95). While this tree is obviously better than a simple NJ tree for displaying presumed phylogenetic relationships, we stress that the purpose of this paper is alpha taxonomy and not phylogeny, and any phylogenetic analyses are highly preliminary until hundreds more of the ACG and Mesoamerican species of Apanteles can be included.

A neighbor-joining (Saitou and Nei 1987) tree was also constructed in Geneious Pro 6.1 (Drummond et al. 2011) using the TN93 model (Tamura and Nei 1993) (Fig. 2).