Phylogenomic Species Delimitation, Taxonomy, and ‘Bird Guide’ Identification for the Neotropical Ant Genus Rasopone (Hymenoptera: Formicidae)

Abstract Rasopone Schmidt and Shattuck is a poorly known lineage of ants that live in Neotropical forests. Informed by phylogenetic results from thousands of ultraconserved elements (UCEs) and mitochondrial DNA barcodes, we revise the genus, providing a new morphological diagnosis and a species-level treatment. Analysis of UCE data from many Rasopone samples and select outgroups revealed non-monophyly of the genus. Monophyly of Rasopone was restored by transferring several species to the unrelated genus Mayaponera Schmidt and Shattuck. Within Rasopone, species are morphologically very similar, and we provide a ‘bird guide’ approach to identification rather than the traditional dichotomous key. Species are arranged by size in a table, along with geographic range and standard images. Additional diagnostic information is then provided in individual species accounts. We recognize a total of 15 named species, of which the following are described as new species: R. costaricensis, R. cryptergates, R. cubitalis, R. guatemalensis, R. mesoamericana, R. pluviselva, R. politognatha, R. subcubitalis, and R. titanis. An additional 12 morphospecies are described but not formally named due to insufficient material. Rasopone panamensis (Forel, 1899) is removed from synonymy and elevated to species. The following species are removed from Rasopone and made new combinations in Mayaponera: M. arhuaca (Forel, 1901), M. becculata (Mackay and Mackay, 2010), M. cernua (Mackay and Mackay, 2010), M. conicula (Mackay and Mackay, 2010), M. longidentata (Mackay and Mackay, 2010), and M. pergandei (Forel, 1909).

Two major challenges to solving the taxonomic impediment for hyperdiverse insect groups are specimen rarity and morphological complexity, with the latter often coming in the form of continuous subtle variation among and within species. In ants, those genera that predominately occupy the leaf-litter microenvironment tend to suffer from both of these problems. Leaf-litter ants nest and forage in the ground, leaf-litter, and rotting wood and rarely expose themselves above cover where they can be readily detected by visual collectors. Consequently, routinely collecting these ants requires the use of bulk sampling techniques, such as leaf-litter sifting (Winkler and Berlese methods) and pitfall traps. Ants that are predominately subterranean or have small colonies tend to be especially rare (Wong and Guénard 2017). Many leaf-litter ant genera also tend to exhibit very subtle morphological differences among species, with many of the most obvious differences, e.g., size, varying continuously or tracking environmental variables like elevation. The Neotropical ant genus Rasopone exhibits both of these impeding characteristics (rarity and morphological complexity), making taxonomic revision and species identification very challenging in this group. Here, we aim to greatly improve Rasopone taxonomy and address these challenges. We do this by 1) examining new specimens collected during several large-scale sampling projects in Middle America, 2) integrating molecular phylogenomic and barcode data into our assessments of species boundaries, and 3) using a 'bird guide' approach to species identification.
The genus Rasopone was established by Schmidt and Shattuck (2014) to contain a small set of Neotropical species of the ant subfamily Ponerinae. The genus was diagnosed by a suite of morphological characters, none of which was definitively autapomorphic. The phylogenetic placement of the genus was not treated in Schmidt's (2013) phylogeny and affinities were uncertain. Prior to Schmidt and Shattuck the species were placed in a very broadly defined genus Pachycondyla. Mackay and Mackay (2010) revised the New World species of Pachycondyla, and the species that were later segregated in Rasopone were placed in two species complexes. The arhuaca complex contained six species (R. arhuaca, Insect Systematics and Diversity, (2020) 4(2): 1; 1-33 doi: 10.1093/isd/ixaa004 Research R. becculata, R. cernua, R. conicula, R. longidentata, and R. pergandei) and the ferruginea complex contained five (R. breviscapa, R. ferruginea, R. lunaris, R. minuta, and R. rupinicola).
Prior to routine use of mass sampling methods (Berlese and Winkler) very few collections of Rasopone were known. Over the last 25 yr, quantitative inventory projects in Middle America have greatly increased the number of specimens from this region, allowing for a much finer examination of species diversity and morphological variability. For Middle American taxa, the biodiversity landscape is one of many extremely similar species, each exhibiting its own geographic range and elevational specialization, and occurring in communities of locally sympatric species. The morphological similarity of species and the frequent convergence in species-level characters make purely morphological approaches to species delimitation and identification difficult, particularly among geographically separated populations.
To improve Rasopone taxonomy we rely on both morphological and molecular evidence, and take an iterative approach to resolving species boundaries, involving morphological species assignments, sequencing of multiple populations of putative species, and subsequent refinement of species concepts using tree-based results. Such an approach is commonplace in the era of mitochondrial (mt) DNA barcoding (Hebert et al. 2003), but instead of using single or only a few genetic markers, we employ a phylogenomic method: the targeted enrichment of ultraconserved elements (UCEs; Faircloth et al. 2012. The UCE method generates thousands of nuclear loci for samples at relatively low cost, and unlike DNA barcoding, which uses a single fast-evolving marker, can provide definitive information to resolve both very deep and very shallow phylogenetic relationships (Faircloth et al. 2015, Branstetter and Longino 2019. Resolving relationships above the species level, even in taxonomic studies, has value, because it makes it possible to confidently gauge phylogenetic distance among species or populations, which might have convergent morphologies. Another positive feature of the UCE method is that mitochondrial DNA, including the animal barcode gene cytochrome oxidase I (COI), can often be extracted as bycatch in the sequencing process (Pierce et al. 2017, Branstetter andLongino 2019), making it possible to generate complementary nuclear and mitochondrial datasets for testing species boundaries and to include additional samples from the Barcode of Life Database (Ratnasingham and Hebert 2007). The combination of detailed imagery, comprehensive sampling across the Middle American region, and molecular evidence of underlying genetic structure allows a major advance in our understanding of this group of ants and the ability of researchers to identify new material.
Identification with traditional dichotomous keys becomes more difficult with the increasing number and ever-finer resolution of cryptic species. Species exhibit very subtle differences in continuous characters, which stymies efforts to organize them in a hierarchical decision tree. An alternative approach is the 'bird guide' style: species are organized by one or few variables, with a page or figure for each species that encapsulates information and provides identification tips (e.g., Longino 2019). Non-specialists have always relied on this style and it continues to be favored in the digital age (Stevenson et al. 2003). For insect genera with large numbers of very similar species, the bird guide approach might be more useful for specialists and non-specialists alike. In this revision of Rasopone, we apply the bird guide style in lieu of the traditional dichotomous key.
The 'bird guide' approach includes range maps to facilitate identification. A traditional goal of taxonomy is to be geography-free. Species should be defined such that they can be identified regardless of where they are found. Such global definitions can be achieved using genetic sequence data and may also be achievable with morphological data that are sufficiently detailed (e.g., extensive morphometrics as in Seifert 2003). But UCE phylogenomics show that in most cases, especially among litter ants, species have relatively restricted ranges. At the same time, the most easily observed or measured morphological characteristics show strong convergence in geographically separate clades. Easily observable characters may readily differentiate locally sympatric species but fail to separate geographically distant species. Thus, geography becomes an important identification aid when sequence or detailed morphometric data are not available.
The sampling of Rasopone species is still a long way from complete. Even with the intensive sampling in Middle America, we were able to examine a total of only 549 species occurrences (i.e., not counting duplicate occurrences of the same species within individual collection events). Among these, we recognized 29 species, but eight of these, nearly 30%, are still known from single occurrences. Panama and South America in particular are terra incognita, where a combination of greater sampling and DNA sequencing would undoubtedly reveal many additional species.
Our primary objective in this work is to provide foundational descriptive taxonomy that will facilitate research on the biogeography, ecology, and evolutionary history of ants in the Middle American corridor, while also demonstrating the value of integrating phylogenomic data into species delimitation. We hope that this improved taxonomy will contribute to the general 'biodiversity map' of Middle America, revealing zones of endemism and contributing to conservation efforts in the region.

Material Examined
This study was based on 549 separate species occurrence records. Most of the examined material was from the Middle American corridor (Veracruz, Mexico to Nicaragua). Almost all the specimens were from Winkler or Berlese samples of sifted leaf litter and rotten wood from wet forest habitats. Most material was from large-scale biodiversity inventory projects in Central America and southern Mexico, spanning 25 yr (Projects ALAS, LLAMA, and ADMAC). All holotypes and paratypes associated with the new species described here have unique specimen-level identifiers ('specimen codes') affixed to each pin, and most dry-mounted non-type specimens do as well. Specimen codes should not be confused with collection codes, which are associated with particular collection events. When reported, collection codes follow the collector. Collection data are derived from a specimen database and are not direct transcriptions of labels. Latitudes and longitudes, when present, are reported in decimal degrees, as a precise point (five decimal places) followed by an error term in meters. Material examined is not listed in the species accounts, but instead is available as digital supplementary material to this article (Supp Table 1 [online only]), at the journal's website. Images of holotypes, distribution maps, and all specimen data on which this paper is based are available on AntWeb (www.antweb.org), where they are subject to future modification (data corrections and reidentifications).

DNA Sequence Generation
We selected 75 specimens for DNA sequencing (Table 1): 62 Rasopone, 1 Mayaponera constricta (the type species of this genus), 8 additional Mayaponera (formerly in Rasopone, transferred here), and single outgroup specimens from the genera Dinoponera, Neoponera, Pachycondyla, and Simopelta. All data were newly generated for this study, except for five samples, in which data were taken from Branstetter et al. (2017) (Table 1).
To examine species boundaries and phylogenetic relationships among species and populations, we employed the UCE approach to phylogenomics (Faircloth et al. 2012, Faircloth et al. 2015), a method that combines target enrichment of UCEs with multiplexed, next-generation sequencing. All UCE molecular work was performed following the UCE methodology described in Branstetter et al. (2017). Briefly, the process involves DNA extraction, sample QC, DNA fragmentation (400-600 bp), library preparation, library pooling (equimolar pools of 10 samples), UCE enrichment, qPCR quantification, final pooling (100-104 total samples per sequencing pool), and sequencing. All sequencing was performed on an Illumina HiSeq 2500 instrument (2x125 bp v4 chemistry; Illumina Inc., San Diego, CA) by the University of Utah genomics core facility. For a few samples, we followed a modified enrichment protocol, in which the day 1 procedure followed the Arbor Biosciences (Ann Arbor, MI) myBaits v4 protocol, and day 2 followed a standard UCE enrichment protocol (version 1.5; available from http://ultraconserved.org). To enrich UCE loci, we used an antcustomized bait set ('ant-specific hym-v2') that includes 9,898 baits (120 mer) targeting 2,524 UCE loci shared across Hymenoptera and a set of legacy markers (data not used) . The ability of this bait set to successfully enrich UCE loci and resolve relationships in ants has been demonstrated in several studies , Pierce et al. 2017, Blaimer et al. 2018, Branstetter and Longino 2019.

UCE Matrix Assembly
After sequencing, the University of Utah bioinformatics core demultiplexed the data using bcl2fastq v1.8 (Illumina 2013) and made the data available for download. Once received, the sequence data were cleaned, assembled and aligned using PHYLUCE v1.6 (Faircloth 2016), which includes a set of wrapper scripts that facilitates batch processing of large numbers of samples. Within the PHYLUCE environment, we used the programs ILLUMIPROCESSOR v2.0 (Faircloth 2013), which incorporates TRIMMOMATIC (Bolger et al. 2014), for quality trimming raw reads, TRINITY v2013-02-25 (Grabherr et al. 2011) for de novo assembly of reads into contigs, and LASTZ v1.0 (Harris 2007) for identifying UCE contigs from all contigs. All optional PHYLUCE settings were left at default values for these steps. For the bait sequences file needed to identify and extract UCE contigs, we used the ant-specific hym-v2 bait file. To calculate assembly statics, including sequencing coverage, we used scripts from the PHYLUCE package (phyluce_assembly_get_trinity_coverage and phyluce_as-sembly_get_trinity_coverage_for_uce_loci) that use the programs BWA (Li and Durban 2010) and GATK (McKenna et al. 2010).
After extracting UCE contigs, we aligned each UCE locus using a stand-alone version of the program MAFFT v7.130b (Katoh and Standley 2013) and the L-INS-I algorithm. We then used a PHYLUCE script to trim flanking regions and poorly aligned internal regions using the program GBLOCKS (Talavera and Castresana 2007). The program was run with reduced stringency parameters (b1:0.5, b2:0.5, b3:12, b4:7). We then used another PHYLUCE script to filter the initial set of alignments so that each alignment was required to include data for ≥90% of taxa. This resulted in a final set of 1,802 alignments and 1,542,220 bp of sequence data for analysis. To calculate summary statistics for the final data matrix, we used a script from the PHYLUCE package (phyluce_align_get_align_summary_ data). Information related to UCE sequencing and assembly results can be found in Supp Table S2 (online only). All steps, including the phylogenetic analyses described below, were performed on a multicore Linux workstation (40 CPUs and 512 Gb of memory) housed in the Branstetter lab.

Phylogenomic Analysis
To partition the UCE data for phylogenetic analysis, we used the Sliding-Window Site Characteristics based on entropy method (SWSC-EN; Tagliacollo and Lanfear 2018), which breaks UCE loci into three regions, corresponding to the right flank, core, and left flank. The theoretical underpinning of the approach comes from the observation that UCE core regions are conserved, while the flanking regions become increasingly more variable (Faircloth et al. 2012). After running the SWSC-EN algorithm, the resulting data subsets were analyzed using PARTITIONFINDER2 (Lanfear et al. 2012, Lanfear et al. 2017. For this analysis we used the rclusterf algorithm, AICc model selection criterion, and the GTR+G model of sequence evolution. The resulting best-fit partitioning scheme included 1,252 data subsets and had a significantly better log likelihood than alternative partitioning schemes (SWSC-EN: −11,959,697.15; By Locus: −12,331,121.88; Unpartitioned: −12,481,042.22). Using the SWSC-EN partitioning scheme, we inferred phylogenetic relationships of Rasopone with the likelihood-based program IQ-TREE v1.6.8 (Nguyen et al. 2015). For the analysis we selected the '-spp' option for partitioning and the GTR+G model of sequence evolution. To assess branch support, we performed 1,000 replicates of the ultrafast bootstrap approximation (UFBoot) (Minh et al. 2013, Hoang et al. 2018) and 1,000 replicates of the branch-based, SH-like approximate likelihood ratio test (Guindon et al. 2010). For these support measures, values ≥95% and ≥80%, respectively, signal that a clade is supported.

COI Barcode Analysis
Due to the high abundance of mitochondrial DNA in samples and the less-than-perfect efficiency of target enrichment methods, Cytochrome Oxidase I (COI) sequence data, and sometimes entire mitochondrial genomes (Ströher et al. 2016) are often generated as a byproduct of the UCE sequencing process. To provide a separate assessment of species identities, possibly with more samples included, we extracted COI sequences from our UCE enriched samples and combined them with Rasopone COI sequences downloaded from the BOLD database (Ratnasingham and Hebert 2007) (Accessed 13 May 2019). To extract COI from UCE data, we downloaded a complete 658 bp barcode sequence of Rasopone JTL-014 from BOLD (Acc.#ASNEI075-09) and used this as the bait input sequence for a PHYLUCE program (phyluce_assembly_match_contigs_to_ barcodes) that extracts COI sequence from bulk sets of contigs.
After extracting COI sequence from UCE sample data, we downloaded accessible barcode sequences from BOLD following a series of steps. First, using the BOLD workbench interface, we searched for all records matching the taxonomy search term 'Rasopone'. We then copied all of the resulting Barcode Index Numbers (BINs) and performed a second search using these numbers in the identifiers field. This approach recovers taxonomically mislabeled samples because BINs group sequences into units by sequence similarity, not name (Ratnasingham and Hebert 2013). All returned sequences were downloaded, examined, and filtered for quality. Because some of the sequences included private, unpublished data, we contacted data owners for permission to use the private sequences in our analyses.
We combined the final set of BOLD sequences with the successfully extracted COI sequences from UCE samples and aligned the data using MAFFT. We visually inspected the resulting alignment for signs of pseudogenes or other anomalies using MESQUITE v3.2 (Maddison and Maddison 2018). The final matrix was partitioned by codon position and analyzed with IQ-TREE using GTR+G, 1,000 ultrafast bootstrap replicates, and 1,000 SH-like replicates. Following a preliminary analysis of all samples, we pruned out many samples that we viewed as extraneous, and then repeated the analysis. In Supp  Table S3 (online only), the full set of specimens is reported, with a column indicating which specimens were pruned. The full COI tree with all specimens is also presented in Supp. Fig. S1 (online only).

Morphological Measurements
Measurements were made with a dual-axis micrometer stage with output in increments of 0.001 mm. However, variation in specimen orientation, alignment of crosshairs with edges of structures, and interpretation of structure boundaries resulted in measurement accuracy to the nearest 0.01 mm. All measurements are presented in mm. In species accounts, measurements are presented as mean (minimum-maximum, sample size) for each variable when more than one specimen was measured for the variable. Otherwise the single measured value is shown.
The following measurements and indices are reported: HW: Head width, maximum width of head capsule in full-face view (not including eyes). HL: Head length, maximum length of head capsule in full-face view, from anteriormost projection of clypeus to posteriormost projection of vertex. SL: Scape length, maximum length of scape not including basal condyle and neck. PTH: Petiole height, in lateral view, perpendicular distance from posteroventral lobe of petiolar tergite to dorsal margin of petiolar node (Fig. 1B). In lateral view, the posteroventral lobe and the dorsal margin of the node are in different focal planes, requiring refocusing during measurement. PTL: Petiole length, in lateral view, perpendicular distance from anterior face of petiolar node to posterior margin of node (Fig. 1B) Vouchers may be the same specimen (non-destructive DNA extraction) or with varying degrees of subjectivity (destructive extraction of specimen from same nest, same series, or same population). Full voucher specimen details are in Supp Table S1 (online only).  Insect Systematics and Diversity, 2020, Vol. 4, No. 2

Species Delimitation and Identification
In most cases, species were delimited by identifying sets of specimens that 1) were morphologically very similar, 2) formed monophyletic or paraphyletic groups (mtDNA only) based on sequence data, and 3) showed no evidence of having separate morphological or genotypic clusters within communities (separate clusters indicated the presence of multiple species). In one case a species was named for which the third criterion was not satisfied, and this particular case is discussed in the species account. New species were formally named when there were enough specimens from a local population for an adequate type series, such that specimens with high probability of conspecificity could be distributed to at least two institutions. Otherwise they were given informal, and unavailable, morphospecies codes, pending additional collections and better knowledge of the species. Diagnoses in species descriptions provide separatory characters for all species that are within the same geographic area and morphological size range.
Detailed text descriptions are omitted, relying instead on detailed imagery.
Identification should be carried out using diagnoses in individual species accounts, Figs. 6-13, and Table 2. These figures are standard images and a distribution map of each species, with species in order of mean HW. Specimen identification can be done by measuring the HW, going to that HW in the figures, and working up and down from there to locate potentially matching species. Table 2 provides a list of species in order of size, with key characteristics and geographic ranges.

UCE Sequencing and Matrix Assembly
After sequencing, assembly, and the extraction of contigs representing UCE loci from our set of 75 specimens (70 new to this study), we recovered an average per contig coverage of 41.0x (range: 10.1-78.2x) and a mean contig length of 872 bp (range: 367-1,108 bp). Following alignment trimming, and filtering of the UCE contigs, our UCE matrix consisted of 1,802 loci and 1,542,220 bp of sequence data, of which 339,246 bp were informative. The mean alignment length post-trimming was 855.8 bp (range: 223-2,268 bp). The final matrix included 12.3% missing data (including gaps). For additional sequencing assembly information see Supp Table S2 (online only).

COI Extraction and Matrix Assembly
We successfully extracted COI mtDNA sequence for 73 out of 75 UCE samples, but removed one sequence from the dataset (Dinoponera longipes) for being short and unnecessary for species delimitation. Except for two sequences, all were above 650 bp in length and most were 657 or 658 bp. No sequences had any obvious indications of being pseudogenes. From BOLD, we downloaded an initial set of 313 samples (Supp Table S3 [online only]) and pruned this set down to 22 key samples for the final analysis. The final aligned matrix included 94 COI sequences and was 658 bp in length.

Phylogenetics and Species Delimitation
Analysis of the UCE data recovered a robust phylogenetic hypothesis with most nodes receiving maximum support (Fig. 2). Rasopone was found to be non-monophyletic due to species in the arhuaca complex (R. arhuaca, R. becculata, R. pergandei) being 0.02 recovered as sister to the genus Mayaponera and phylogenetically very distinct from true Rasopone. Using our criteria for species delimitation, we identified 19 putative Rasopone species based on the phylogenetic results. Morphological separation of locally sympatric species was typically straightforward and correct, but attempts to make connections among widely separated populations generally failed. To demonstrate the challenge of separating species by morphology-alone, we left previous morphospecies codes as terminal labels in Fig. 2.
Although we save analysis of biogeographic patterns for later investigation, we note the existence of a core Rasopone clade that includes most species from Panama to Mexico and two more divergent clades, one consisting of a set of species related to the South American taxon R. lunaris and one consisting of only the new species R. cubitalis from Costa Rica and Nicaragua. The latter clade was found to be sister to all other Rasopone species.
COI results were mostly concordant with UCE results at the species level (Fig. 3). One striking exception was R. minuta. Rasopone minuta was monophyletic and phylogenetically distant from R. panamensis in the UCE tree, but paraphyletic with respect to R. panamensis in the COI tree. The COI data also included a couple of COI-only samples that might belong to additional new 0.2 Fig. 3. Phylogenetic relationships among a curated set of COI barcode sequences for Rasopone. Black samples were sequenced for UCEs. Red samples were downloaded from the BOLD database. The tree was inferred using IQ-TREE with the data partitioned by codon position. Black circles on nodes indicate high support, which we define as ≥95% ultrafast bootstrap support and ≥95% SH-like branch support. Terminal names match taxonomic changes proposed in paper and provide useful sample identifiers (e.g., extraction codes [EX#] or BOLD process IDs). A complete, unpruned COI tree is available in Supp Fig. S1 (online only). species since they did not cluster closely with any of our UCE samples. The relationships among species within Rasopone were largely discordant, although the deep split between R. cubitalis and the remaining Rasopone species was supported in both trees. At deeper levels, relationships were again congruent, although lacking strong node support in the COI tree.

Taxonomic Results
Redefinition of Rasopone and Mayaponera Schmidt and Shattuck (2014) provided the following diagnosis for Rasopone, based on workers: eyes present, mandibles relatively long, mandibular pit or groove absent, mesosomal profile nearly continuous, the metanotal groove shallow or absent, metapleural gland orifice without a posterior U-shaped cuticular lip, propodeal spiracle round or ovoid, mesotibiae dorsally without abundant stout traction setae, ventral apex of the metatibia with both a large pectinate spur and a smaller simple spur, fenestra absent from the petiolar process, prora present on anterior margin of first gastral sternite, and stridulatory organ absent from A4 pretergite. This should be emended to 'prora absent'. There is no true prora in any caste of Rasopone or Mayaponera as defined here. In some cases there is a denticle on the sternite of the helcium that can be misinterpreted as a prora.
To Schmidt and Shattuck's diagnosis we add the following characters: 1) petiolar sternite with deep transverse posterior groove such that posterior portion of sternite forms a shovel-like extension separate from tergite ( Fig. 1B) (this character applies to all castes); 2) anterior margin of clypeus truncate to emarginate, never entirely convex ( Fig. 1E and F); and 3) color light to dark red brown, never entirely black. These additional characters exclude the arhuaca complex. The arhuaca complex has a variously shaped anteroventral petiolar process, but the posterior portion of the sternite is closely appressed to the tergite. The anterior margin of the clypeus is convex, and the color is black.
Members of the arhuaca complex share many characters with Mayaponera constricta (currently the sole member of the genus), and the molecular data place them near each other (and distant from Rasopone). The main morphological difference between Mayaponera constricta and the arhuaca complex is that the former has an impressed metanotal groove and the species of the arhuaca complex do not. Mayaponera constricta is a surface forager found in a wide variety of habitats, from mature forest to open or disturbed areas, and from wet to seasonally dry areas. It also has a very large geographic range, from Honduras to southern Brazil. In contrast, species of the arhuaca complex are found only in shaded rainforest habitats, and the species have smaller geographic ranges. In spite of these differences, we place the arhuaca complex in Mayaponera, and we consider the impressed metanotal groove of M. constricta to be a species-level trait and not a diagnostic feature of the genus. A full diagnosis of Mayaponera is not attempted here and should await a true revision of the genus.

Rasopone Species-Level Characters
The species of Rasopone have a uniform habitus (Fig. 1A). Worker HW varies from 0.80 to 1.70. The coefficient of variation for metric characters is typically ~5%. The medial projection of the anterior clypeal margin varies from strongly square-cut, flat medially and with sharp lateral corners (Fig. 1E), to medially emarginate with rounded lateral lobes (Fig. 1F). These two states are referred to as truncate or sinuous in species accounts. The mandibles vary from striate to smooth and shining. The sides of the head show variable levels of pilosity, from bare to having numerous short setae uniformly distributed. Face sculpture is usually a uniform surface of dense minute puncta, but in some species these are overlain with larger, more widely spaced puncta. When present, the larger puncta are usually faint and only visible under particular orientation and lighting. The degree of expression of these larger puncta is a continuum across species, and their presence is only noted for species with relatively strong expression. The petiole in lateral view varies from scale-like to cuboidal. The anterior margin is always flat to slightly concave. In species with strongly cuboidal nodes the posterior margin is also flat, parallel to the anterior margin, and there is a differentiated dorsal face (Fig. 1D). In species with a scale-like node, the dorsal and posterior faces are not differentiated and form a single convex curve (Fig. 1B). Among species there is a continuum from cuboidal to scale-like nodes (Fig. 1B-D). Petiolar nodes also vary in height to length ratio, some being relatively tall and thin, others shorter and longer.
Queens are very similar to workers except for larger compound eyes, presence of ocelli, and queen-typical mesosomal sclerites (Fig. 4). The ocelli are always very small. In species for which queens and workers are known, HW is approximately the same for the two castes. Rasopone cubitalis may be an exception, with queens ~1.17× larger than workers.
Males are known for a few species. The type series of R. rupinicola was collected with one or more males. Males occur in Malaise samples and a few have been associated with workers using COI sequence data. A figure is provided of the associated male of R. mesoamericana sp. nov. (Fig. 5), and several males are figured in Supp Figs. S18, S21, and S31 (online only), but otherwise males are not examined in this study.

Nomenclature
This paper and the nomenclatural act(s) it contains have been registered in Zoobank (www.zoobank.org), the official register of the International Commission on Zoological Nomenclature. The LSID (Life Science Identifier) number of the publication is: urn:lsid:zoobank. org:pub:0DE2398D-199F-40A7-8207-91148630CD76.

Geographic Range
Southern Mexico to Bolivia, Paraguay, and southern Brazil. There are no records from the Caribbean.

Comments
This species is known only from the type queen. Rasopone breviscapa, R. rupinicola, and R. titanis are the largest species in the genus, with HW ~1.7. Rasopone breviscapa has the anterior clypeal margin truncate, mandibles striate, and petiole intermediate between cuboidal and scale-like. It is differentiated from R. rupinicola

Biology
This species complex occurs in cloud forest habitats, from 1,000 to 2,000 m elevation (although one enigmatic collection is from 250 m elevation on the Osa Peninsula). Workers, and in one case a dealate queen, are collected in Winkler samples of forest floor leaf litter and rotten wood. Workers have been collected beneath epiphytes in treefalls and beneath rotten wood on the ground. Workers have been found both in closed-canopy cloud forest and in open pastures near cloud forest edges (beneath wood on the ground). In Monteverde, in addition to specimens from multiple Winkler samples, a worker was found in soil of a clay bank in a steep-sided ravine, and a worker was found beneath a stone. At Estación Pittier in the Talamanca range, a dealate queen was found beneath a stone. It is likely that the most common forms (of the various cryptic species) nest in the soil, with workers foraging in the litter. One nests beneath epiphytes. Males can be common in Malaise traps (associated with COI results). An alate queen was collected by an INBio Parataxonomist, at Tapantí National Park, on 30 October 1991. Cryptic species in the complex may specialize on particular microhabitats (e.g., beneath epiphytes; see Comments).

Comments
DNA sequence data, both COI and UCE, support a clade endemic to Costa Rican and adjacent Panamanian cloud forest. Within the clade, genetic evidence also supports the occurrence of multiple sympatric species at two intensively sampled sites. But the genetic data are fragmentary and the relationships among species across sites is unclear. We take the approach of referring to the entire clade as R. costaricensis, with the acknowledgment that further study will almost certainly result in the further splitting of the clade into component species. We describe here what we currently know of the sympatric forms at particular sites. There is evidence that three sympatric species occur on the Barva transect, on Costa Rica's Caribbean slope. The three species differ in morphology, microhabitat, and COI sequence.
Form a (Fig. 11; Supp Fig. S5 [online only]): This is the largest of the three. The mandibles are smooth and shiny; the other two forms have striate mandibles. It is known from five separate collections: 1) 2 workers under epiphytes in an old treefall; 2) a worker and some brood in a rotten knot in a treefall; 3) a worker under epiphytes on dead wood at the edge between pasture and forest; 4) a Berlese sample of epiphytic material; and 5) a worker in a collection of mixed ants collected by hand. The first four collections were from a 1,500 m site and the last collection from an 1,100 m site. It is notable that no specimens were collected in the 350 miniWinkler samples of forest floor litter that were taken at the two sites.  Insect Systematics and Diversity, 2020, Vol. 4, No. 2 Form b ( Fig. 10; Supp Fig. S6 [online only]): This and Form c have striate mandibles. Form b is very similar to Form c, but the petiole is somewhat more tapering and scale-like. It occurred at the 1,500 m and 2,000 m sites on the Barva transect, where it was moderately abundant in miniWinkler samples. Parataxonomist Ronald Vargas collected a worker by hand at the 2,000 m site.
Form c (Fig. 8; Supp Fig. S7 [online only]): This form has a relatively less scale-like node than Form b. It occurred at the 1,500 m site, from two collections. One worker was collected by hand from under rotten wood. Two workers were collected by student Andy Boring in an open pasture area, beneath rotten wood. These two collections were united by DNA sequence data (see below), which also supported their distinctness from Form b. However, a few specimens from miniWinklers from the 1,500 m and 2,000 m sites lacked sequence data and were intermediate in petiole shape, and thus could not be assigned to one form or the other.
There is evidence for at least three sympatric species, based on COI clusters in the BOLD database, on the peak of Volcán Cacao, a cloud forest site in Guanacaste, Costa Rica. The evidence comes mostly or entirely from males in Malaise traps, sampled by Alex Smith and others. Barva Form a and Form b cluster with one of the Cacao clusters. Form a and Form b differ from each other by about 5%, while each differs from the Cacao specimens by about 3%. Barva Form c clusters with the largest Cacao cluster, which contains over 120 specimens, with much less than 0.5% sequence divergence among them. A third Cacao cluster is small, and currently unassociated with any other specimens ( Fig. 3; Supp Fig. S1 [online only]).

Biology
This species occurs in lowland rainforest, from sea level to about 500 m elevation. Only one worker is known, from a Winkler sample of forest floor litter and rotten wood. Multiple alate queens are known, from Malaise traps, flight intercept traps, and light traps. One alate queen is from a Berlese sample of litter and soil; it is possible the queen was a contaminant, attracted to the light bulb of the Berlese funnel. The queen records are from the months of January, February, March, May, and September.

Comments
There has been intensive Winkler sampling at La Selva Biological Station, and workers of the smaller species R. pluviselva occur moderately frequently in these samples, yet only one worker of R. cryptergates has been discovered. The alate queens are the reverse, with moderately abundant queens of R. cryptergates, and a single alate queen of R. pluviselva. Dealate queens of R. pluviselva occur occasionally in Winkler samples. It is possible that R. cryptergates is more subterranean than R. pluviselva, and workers hardly ever forage in the litter. Alternatively, R. cryptergates may prefer open habitats such as pastures and lawns, and thus be more abundant in the agricultural landscape adjoining La Selva. Rasopone cryptergates may produce more abundant alate queens, or queens that fly greater distances or higher above the ground. In contrast, R. pluviselva may rely on fewer or less vagile queens that fly close to the ground, rarely being captured by Malaise or light traps.
UCE and COI data are available for the single worker specimen, placing it in a clade with three other species known only from Panama. However, the worker was associated with the queens based on morphology alone: the cuboidal shape of the petiolar node and matching size. There are currently no genetic data definitively associating the sequenced worker with the holotype queen, so future confirmation is warranted. Geographic range. Southern Nicaragua to Costa Rica.

Diagnosis
Lowland; mandible smooth and shiny; anterior clypeal margin short, sinuous; side of head with abundant erect setae; face with abundant short erect setae; face sculpture of dense, minute puncta overlain with larger, more widely spaced puncta visible in particular orientation and lighting; petiole cuboidal.

Biology
This species occurs in lowland rainforest, with records from 160 to 880 m elevation. Workers have twice been collected in Winkler samples of forest floor litter and rotten wood. An alate queen was collected in a Malaise trap in October. Two workers and a larva were collected in a small chamber in a clay bank, in mature rainforest.

Comments
This species is known from three localities, one in southern Nicaragua and two in northern Costa Rica. UCE and COI data show a single cluster with little sequence divergence among the three populations.

Biology
This species occurs in a variety of wet forest habitats, across an elevational range of 180-1,340 m. Habitats include mature and second growth forests, from lowland rainforest to cloud forest. Cloud forest habitats include Liquidambar dominated forests in Mexico to mesophyll cloud forest in Nicaragua. Most specimens are workers from Winkler samples of forest floor litter and rotten wood. Student D. J. Cox collected a dealate queen and a worker while hand collecting at night. A worker was collected on a clay bank at a stream edge, and another one was collected at a cookie bait.

Comments
The image of the holotype queen on AntWeb shows the distinctive clypeal shape, with a sinuous, non-trapezoidal anterior margin. The size (measured on image), lack of setae on the side of the head, and the asymmetrical, scale-like node also match material identified here as R. ferruginea. North of the Isthmus of Tehuantepec, R. ferruginea is sympatric with R. JTL034 and R. JTL035. In the northern lowlands of Chiapas, it is sympatric with R. minuta and R. subcubitalis. In the Cordillera de Chiapas it is sympatric with R. politognatha. There are no records from Guatemala. In Honduras, it is known from the hills above Tela, where it is sympatric with R. minuta. In Nicaragua, it is known from Cerro Musún, where it is sympatric with R. mesoamericana and R. politognatha.

Biology
This species occurs in cloud forest habitats, with a known elevational range of 1,270-1,850 m. Habitats include pine and oak forests, and diverse mesophyll forests. All known specimens are from Winkler and Berlese samples of forest floor litter and rotten wood. Most specimens are workers, but two dealate queens are known.

Biology
This species occurs in lowland habitats, with records from sea level to 1,050 m elevation. Sampling methods, when indicated on specimen labels, are Winkler and 'hypogaeic Winkler'.

Comments
UCE results reveal that most Rasopone species are in a clade of relatively smaller species that is sister to the large species R. cubitalis. These smaller species occur throughout Middle and South America. Within this clade of smaller species, three South American specimens were sequenced for UCEs, and these form a clade sister to all the Middle American specimens (Panama northward). One of the sequenced South American specimens is from southern Brazil (Minas Gerais). The morphology closely matches the type of R. lunaris, and the locality is not too distant from the type locality. Thus, this specimen can be identified as R. lunaris with confidence. This specimen is sister to the two other sequenced specimens from South America. One of the latter is from Amazonian Colombia and was initially identified as R. lunaris, differing only in the lack of erect setae on the side of the head, and somewhat larger size. The other one is from French Guiana and has distinctive puncta on the face. The former is placed in the morphospecies JTL049, and the latter in JTL047. Given the sparse sampling from South America, we can expect continued discovery of cryptic species and geographic structuring there. The examined specimens of R. lunaris from southern South America are quite uniform and suggest a single widespread species at the southern range limit of the genus. However, the scattered records from northern South America are more variable and future sequencing work may identify them as multiple cryptic species, perhaps more related to the nearby morphospecies JTL047, JTL048, and JTL049. MacKay and MacKay had a broad view of R. lunaris, identifying material from Guatemala to Paraguay as Pachycondyla lunaris. Our definition is narrower and more geographically limited.

Diagnosis
Mandible striate; anterior clypeal margin truncate; side of head with variable pilosity, nearly bare to evenly distributed short erect setae; face sculpture of dense, minute puncta overlain with larger, more widely spaced puncta visible in particular orientation and lighting (very faint in some specimens); petiolar node tapering, scale-like. Two species are within geographic and size range of R. mesoamericana: Rasopone subcubitalis (

Biology
This species occurs across a range of wet forest habitats, from 310 to 1,750 m elevation. Most records are from cloud forest, but often at the lower edge, at the transition to lowland habitat. For example, on the Barva transect in Costa Rica, R. mesoamericana is known from an 1,100 m site, while R. costaricensis occurs at 1,500 and 2,000 m sites. BOLD specimens from Guanacaste, Costa Rica are males from Malaise traps at 1,080 m and 1,185 m, and workers from a litter sample and a bait at 1,000 m and 972 m, respectively. In contrast, none are known from the nearby higher elevation peak, where R. costaricensis is common. In some sites workers can be moderately abundant in Winkler samples of forest floor litter and rotten wood. Dealate queens occasionally occur in Winkler samples. Workers are occasionally collected at baits. Males occur in Malaise traps.

Comments
Molecular data indicate that Rasopone mesoamericana and R. subcubitalis are sister species and the two are quite similar. Rasopone mesoamericana has a more asymmetrical, scale-like petiolar node and is larger on average (but the size ranges overlap). The two have no known cases of sympatry in a local assemblage, but the geographic ranges broadly overlap. It may be somewhat arbitrary to treat them as two species at this stage, but they do form two separate clades, and the petiolar shape differences are consistent. BOLD COI sequences are available for many specimens in this clade, and the BOLD results are fully congruent with the UCE results. BOLD sequences segregate into five allopatric BINs, raising the potential for multiple cryptic species.

Diagnosis
Lowland; mandible smooth and shiny; anterior clypeal margin truncate; side of head bare or with a few inconspicuous erect setae; petiolar node moderately tapering, scale-like; color orange. The most similar species is R. pluviselva ( Fig. 6

Biology
This species occurs in lowland wet to seasonal moist forest, with records from sea level to 890 m elevation. The holotype queen was collected in a pitfall trap. Other specimens are from Winkler samples of forest floor litter and rotten wood. A dealate queen was collected in a Winkler sample.

Comments
This is one of the two smallest species, the other being the allopatric R. pluviselva, which occurs further south. These two species are very similar but differ in relative scape length: R. minuta, SI 76-80 (n = 7); R. pluviselva, SI 72-74 (n = 7). Rasopone minuta is a lowland species, typically without sympatric forms at low elevation, but overlapping with cloud forest species (such as R. subcubitalis) at the upper end of its elevational range. UCE and COI data support the delimitation of this species.
MacKay and MacKay included in the description a worker from Guatemala, but expressed doubts of its conspecificity. They indicated that the worker petiolar node was rectangular, rather than narrowed dorsally as in the holotype queen. Petiole shape shows little intraspecific variation, and the worker is undoubtedly not conspecific. They also identified a worker from Venezuela as R. minuta, but this is also unlikely to be conspecific.  Forel, 1899: 15. Neotype worker: Costa Rica, San José: Cerro Plano, 9.48059 −83.96402 ±10 m, 1,060 m, 4-vii-2015  Pachycondyla ferruginea panamensis : Brown, in Bolton, 1995: 308. Pachycondyla ferruginea: MacKay andMacKay, 2010: 319 (incorrect synonymy).

Biology
This species occurs in lowland wet to seasonal dry forest habitat, with records from sea level to 1,070 m elevation. Nearly all specimens are workers and occasional dealate queens from Winkler samples of forest floor litter and rotten wood.

Comments
This is a lowland species known from both coasts of Costa Rica. On the Pacific coast, it occurs from Cabo Blanco on the Nicoya Peninsula to the Osa, where it is the common species in litter samples. On the Caribbean side, it is the common species in litter at Hitoy Cerere, south of Limón, but is unknown north of there, in spite of intensive sampling at La Selva Biological Station and the Barva transect.
The original type of R. panamensis is missing. Longino searched for it during a visit to MHNG in 1990, Mackay and Mackay (2010) reported it missing, and Fisher did not find it when imaging MHNG types in 2013. The queen is described as being 6.5 mm long, the mandible with an oblique sulcus, and the petiolar node concave posteriorly but as thick at the top as at the base ('mais aussi épaisse en haut qu'en bas'). These characters match the relatively common lowland species found in the southern Pacific lowlands of Costa Rica, near the originally published type locality (Bugaba, Panama).

Diagnosis
Lowland; mandible smooth and shiny to faintly striate; anterior clypeal margin truncate; side of head bare or with a few inconspicuous erect setae; petiolar node moderately tapering, scale-like; color orange. The most similar species is R. minuta ( Fig. 6

Biology
This species occurs in lowland wet to seasonal dry forest habitats, with records from 50 to 1100 m elevation. BOLD data associate a male from a Malaise trap in Santa Rosa National Park, a dry forest site in Costa Rica. Most specimens are workers and the occasional dealate queen in Winkler samples of forest floor litter and rotten wood. One worker was hand collected beneath a stone. An inexplicable record is a worker in a vegetation beating sample from an 1,100 m site on the Barva transect in Costa Rica. An alate queen was collected in February in a Malaise trap.

Comments
This is one of the two smallest species, the other being R. minuta. These two species have mean HW < 0.9, while all other species have mean HW > 0.9. Both are lowland species. They are allopatric, R. pluviselva occurring east and south of the Sierra de Agalta in Honduras, and R. minuta occurring north and west of this range. They are extremely similar, but differ in relative scape length: R. minuta, SI 76-80 (n = 7); R. pluviselva, SI 72-74 (n = 7).
BOLD COI data unite specimens from Honduras (Las Marias), Nicaragua (Saslaya), Costa Rica (Barva transect, Santa Rosa National Park, Osa), and Panama (Barro Colorado Island). UCE data likewise unite two specimens, one from Saslaya and one from the Barva transect, and the COI sequence from these specimens is consistent with the BOLD results. We have not examined Panama specimens directly, but BOLD images of a dealate queen from the Barro Colorado Island match our concept of R. pluviselva. The COI data form four geographically structured BINs with Panama sister to Costa Rica northward, and Pacific slope Costa Rica sister to an Atlantic slope region that extends from Costa Rica to Honduras. Pacific slope Costa Rica further separates into two BINs, based on one specimen from Santa Rosa National Park and one specimen from the Osa Peninsula.