Taxonomic revision of the Malagasy Aphaenogaster swammerdami group (Hymenoptera: Formicidae)

Background Madagascar is famous for its extremely rich biodiversity; the island harbors predominantly endemic and threatened communities meriting special attention from biodiversity scientists. Continuing ongoing efforts to inventory the Malagasy ant fauna, we revise the species currently placed in the myrmicine genus Aphaenogaster Mayr. One species described from Madagascar, Aphaenogaster friederichsi Forel, is synonymized with the Palearctic A. subterranea Latreille syn. nov. This species is considered neither native to Madagascar nor established in the region. This revision focuses on the balance of species in the A. swammerdami group which are all endemic to Madagascar. Methods The diversity of the Malagasy Aphaenogaster fauna was assessed via application of multiple lines of evidence involving quantitative morphometric, qualitative morphological, and DNA sequence data. (1) Morphometric investigation was based on hypothesis-free Nest Centroid clustering (NC-clustering) combined with PArtitioning based on Recursive Thresholding (PART) to estimate the number of morphological clusters and determine the most probable boundaries between them. This protocol provides a repeatable and testable approach to find patterns in continuous morphometric data. Species boundaries and the reliability of morphological clusters recognized by these exploratory analyses were tested via confirmatory Linear Discriminant Analysis (LDA). (2) Qualitative, external morphological characteristics (e.g., shape, coloration patterns, setae number) were subjectively evaluated in order to create a priori grouping hypotheses, and confirm and improve species delimitation. (3) Species delimitation analyses based on mitochondrial DNA sequences from cytochrome oxidase I (COI) gene fragments were carried out to test the putative species previously delimited by morphological and morphometric analyses. Results Five species can be inferred based on the integrated evaluation of multiple lines of evidence; of these, three are new to science: Aphaenogaster bressleri sp. n., A. gonacantha (Emery, 1899), A. makay sp. n., A. sahafina sp. n., and A. swammerdami Forel, 1886. In addition, three new synonymies were found for A. swammerdami Forel, 1886 (A. swammerdami clara Santschi, 1915 syn. n., A. swammerdami curta Forel, 1891 syn. n. and A. swammerdami spinipes Santschi, 1911 syn. n.). Descriptions and redefinitions for each taxon and an identification key for their worker castes using qualitative traits and morphometric data are given. Geographic maps depicting species distributions and biological information regarding nesting habits for the species are also provided.


INTRODUCTION
The Malagasy zoogeographical region is known to harbor extremely diverse and unique biota, and has recently been the focus of extensive biodiversity and systematic research (Fisher, 2009;Hita Garcia & Fisher, 2014;Csősz & Fisher, 2015). Beyond understanding the history of the region and the evolution of the fauna and ecological processes, accurate information on the Malagasy fauna is indispensable for creating plans to halt extinction in this highly vulnerable region. These efforts to explore Malagasy biodiversity has considerably increased our knowledge of the island's myrmecofauna. The latest findings report extremely high species diversity in many genera of the region. The goal of the current paper is to contribute to this endeavor and prepare a species-level taxonomic revision of the myrmicine genus Aphaenogaster Mayr on Madagascar.
The native Malagasy Aphaenogaster species were either first described or combined in the genus or subgenus Deromyrma Forel. The type species of Deromyrma, A. swammerdami Forel, 1886, has been included in a number of molecular phylogenetic studies of ants. These studies have consistently shown that Aphaenogaster is not monophyletic and that swammerdami is of a different lineage than Aphaenogaster sensu stricta (Brady et al., 2006;Moreau, 2006;Moreau & Bell, 2013;Ward et al., 2015;Demarco & Cognato, 2015;DeMarco & Cognato, 2016).
Here we refer to the native species on Madagascar as members of the A. swammerdami group even though evidence indicates that they represent a different lineage from Aphaenogaster sensu stricta. Though placing the native species in a separate genus would more accurately reflect the different evolutionary history of the Malagasy species, we refrain from classifying them in the first available genus group name Deromyrma at this time. Though all members of Malagasy swammerdami group would belong to this clade, it is not clear which members from outside the Malagasy region would also belong. To address this point, future morphological and molecular studies would need to include species from Central America, Australasia, Indomalaya, and the southern Palearctic regions.
On Madagascar, the swammerdami-group species are abundant and dominant in many habitats. The genus includes some of the largest ant species in the region, which typically nest in soil or hollow wood cavities in lower vegetation. Foragers can also be found alone on the ground, in leaf litter, or in debris. The most striking morphological feature of this group is the presence of a cephalic constriction that forms a conspicuous "neck" on the posterior portion of the head capsule. The neck is present in workers, queens

MATERIALS AND METHODS
In the present study, 16 continuous morphometric traits were recorded in 176 worker individuals from 96 collecting events in Madagascar. Specimens evaluated for this study were from the following institutions: CASC (California Academy of Sciences, San Francisco, CA, USA), MHNG (Muséum d'Histoire Naturelle, Geneva, Switzerland), MSNG (Museo Civico di Storia Naturale "Giacomo Doria", Genova, Italy), NHMB (Naturhistorisches Museum, Basel, Switzerland), PSWC (Phil S. Ward's collection, University of California Davis, Davis, CA, USA). Type material and samples that were morphometrically investigated are presented in the "Type material investigated" and "Material examined" sections in the following format: collection code (in bold), unique identifying CASENT code, verbatim locality, longitude, latitude, elevation in meters, collector, date in MM.DD.YYYY (number of individuals measured, abbreviation of depository). Additional information on habitat and microhabitat is discussed in the Distribution and Biology sections for each species. Raw data, including indices and classification results for material morphometrically examined in this work, is given in Table S1.
All specimen information and relevant images used in this study are available online on AntWeb (antweb.org). The images were taken by staff at the California Academy of Sciences and the person taking the images is provided in the figure caption. The images are copyrighted by the California Academy of Sciences and licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0).
Digital color montage images were created using a JVC KY-F75 digital camera and Syncroscopy Auto-Montage software (version 5.0), or a Leica DFC 425 camera in combination with the Leica Application Suite software (version 3.8). Distribution maps were generated in R (R Development Core Team, 2015) via the "phylo.to.map" function using package phytools (Revell, 2012).

Morphometric character recording
Measurements were taken with a Leica M165C stereomicroscope equipped with an ocular micrometer at a magnification of ×25 to ×100. Body size dimensions are expressed in µm. Due to rarity or lack of queen and male specimens, the present revision is based only on the worker caste. All measurements were made by SC. Abbreviations of morphometric traits, explanations for measured characters, and the magnification applied for each certain trait, are given in Table 1. The complex morphometry-based statistical framework, including hypothesis formation and testing, follows the protocol published in detail in Csősz & Fisher (2016a, 2016b.

Qualitative morphology-based species delimitation
This workflow involves external morphological characteristics regularly considered in conventional verbatim species descriptions. Shapes, color patterns, setae number, and sculpture characteristics were subjectively evaluated during this work phase. A Leica LED3000 SLI LED Spotlight Illuminator with gooseneck cold-light source equipped with two flexible, focally mounted light-cables were used to characterize color patterns.

DNA based-species delimitation analysis
We analyzed 658 base pairs (bp) of the mitochondrial cytochrome oxidase I (COI) gene from 89 Malagasy Aphaenogaster specimens (Appendix 1). DNA was extracted using specimen legs, allowing preservation of vouchers. DNA extraction and COI sequencing were performed at the University of Guelph (Ontario, Canada), following the protocol described in Fisher & Smith (2008). All sequences are available at GenBank and BOLD (Table S2). We also included 2 sequences from GenBank as outgroup, Cephalotes texanus (KC335803.1) and Tapinoma magnum (KY426518.1). Sequences were aligned using Geneious 11.1.5 (Biomatters Ltd., Auckland, New Zealand; Masters, Fan & Ross, 2011). Final alignment is available as Supplemental File S3. To exclude redundancies in the matrix, we removed duplicated haplotypes using the Alter (Glez-Peña et al., 2010) online platform (http://sing.ei.uvigo.es/ALTER/). The final simplified data matrix had 74 sequences consisting of 72 unique Aphaenogaster haplotypes and 2 outgroups.
PartitionFinder2 (Lanfear et al., 2017) was used to establish the best nucleotide substitution model and partition scheme for the data under the corrected Akaike Information Criteria (AICc). Models that estimated a proportion of invariant sites ("+I" parameter) were not considered to avoid overparameterization when selecting the gamma shape parameter ("+G") (Mayrose, Friedman & Pupko, 2005;Stamatakis, 2006). The best fit model selected for each codon position (1 + 2 + 3) was GTR+G. Bayesian phylogenetics were estimated in BEAST v2.5.1 (Bouckaert et al., 2014) using a strict molecular clock, empirical frequency base, nucleotide substitution model, and codon partition established by PartionFinder2. Analyses ran for 10 9 generations, with one tree SPST Propodeal spine length. Distance between the center of propodeal stigma and spine tip ×100 SPTI Propodeal spine apical distance. The distance between spine tips in dorsal view; if spine tips are rounded or truncated, the centers of spine tips are taken as reference points ×100 sampled every 10 5 generation, resulting in 10 4 trees. MCMC chain stationarity and likelihood ESS values greater than 200 were verified in Tracer 1.7.1 (Rambaut et al., 2018). Consensus tree was computed in TreeAnotator 2.5.1 (BEAST package) after discarding the first 10% of trees as burn-in.

Species delimitation
Bayesian consensus tree was used as input to test the putative species designated a priori based on morphological and morphometric data. Analyses were conducted with the Species Delimitation Plugin (SDP; Masters, Fan & Ross, 2011), implemented in Geneious, using the following metrics: (i) monophyly; (ii) average intraspecific uncorrected pairwise distance (Intra Dist); (iii) average uncorrected pairwise distance between a putative species and its sister species (Inter Dist); (iv) PID Liberal (Ross, Murugan & Li, 2008); and (v) Rosenberg's P AB statistics (Rosenberg, 2007). PID Liberal measures the probability of a correct identification of an unknown specimen as a member of the putative species or its sister species, while Rosenberg's P AB is the probability of reciprocal monophyly by random chance. Thus, we expect valid species to be monophyletic, with overall intraspecific distance smaller than interspecific, high PID Liberal values, and small values of Rosenberg's P AB .

Field study permissions
The following information was supplied relating to field study approvals (i.e., approving body and any reference numbers): Ant samples used in this study comply with the regulations for export and exchange of research samples outlined in the Convention of Biology Diversity and the Convention on International Trade in Endangered Species of Wild Fauna and Flora. For field work conducted in Madagascar, permits to research, collect, and export ants were obtained from the Ministry of Environment and Forest as part of an ongoing collaboration between the California Academy of Sciences and the Ministry of Environment and Forest, Madagascar National Parks, and Parc Botanique et Zoologique de Tsimbazaza. Authorization for export was provided by the Director of Natural Resources.

New Zoological Taxonomic Names
The electronic version of this article in Portable Document Format (PDF) will represent a published work according to the International Commission on Zoological Nomenclature (ICZN), and hence the new names contained in the electronic version are effectively published under that Code from the electronic edition alone. This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The ZooBank LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix http://zoobank.org/. The LSID for this publication is: urn:lsid: zoobank.org:pub:A66CA7F2-7334-40CA-A368-E522BA54ACBF. The online version of this work is archived and available from the following digital repositories: PeerJ, PubMed Central and CLOCKSS.

RESULTS AND DISCUSSION
Both clustering methods "hclust" and "kmeans" using function "part" based on NC-clustering returned six clusters (Fig. 1). Five clusters were confirmed via subjective evaluation of qualitative traits and mtDNA; three clusters recognized by NC-PART clustering (A. bressleri, shown by green bars in Fig. 1) were not convincingly supported via Bayesian phylogeny using mtDNA data, hence these three clusters were lumped and described as a single species, A. bressleri. Another cluster of specimens (marked by orange bars in Fig. 1) that was not recognized via NC-PART clustering was outlined via descriptive traits and mtDNA. This cluster of specimens possesses a unique combination of traits that would qualify for a species description, but the small number of samples (a total of 3 workers were available) were below the minimum cluster size threshold (minsize = 5) set for the clustering methods. Hence this cluster-though their position is somewhat separated from the bulk of the swammerdami cluster-remained unrecognized by the gap statistics. We added this species, A. makay, to our morphological species hypothesis as a recognized entity.
When the five-species hypothesis was tested by LDA and by LOOCV, the overall classification success was 100%. All putative species with COI sequences available also showed satisfactory results for all DNA species delimitation criteria used ( Fig. 2; Table 2). All clusters proved to be monophyletic, with intraspecific distance varying from 0.1% (A. makay) to 2.0% (A. gonacantha and A. sahafina) falling outside the range of interspecific divergence (from 3.6% to 3.9%), average PID Liberal values ranging from 0.91 (A. gonacantha) to 0.98 (A. swammerdami) and Rosenberg's P AB values lower than 0.0001.
We found complete agreement between monophyly of clusters in the molecular phylogeny and the quantitative morphology-based cluster delimitation protocols. Taking multiple lines of evidence into account, we describe the five clusters as five species: Aphaenogaster bressleri, A. gonacantha (Emery, 1899), A. makay, A. sahafina, and A. swammerdami Forel, 1886. These species also differ in qualitative diagnostic features and body ratios (for body ratios, see Table 3).
Classification of measured type specimens was set to wildcard in LDA; i.e., no grouping label was added for type specimens, as their classifications were predicted by the LDA that assigns posterior probabilities to each individual in the analysis reflecting the uncertainty of assessing an observation to a particular class. The geometric mean of posterior p values was calculated for syntype series. Type material (series of 2 workers) of A. gonacantha (Emery, 1899) was placed in a cluster named A. gonacantha (Emery, 1899), with posterior probability p = 0.988. The four other type materials, A. swammerdami Forel, 1886 (3 syntype workers), A. swammerdami clara Santschi, 1915 (single worker), A. swammerdami curta Forel, 1891 (2 syntype workers) and A. swammerdami spinipes Santschi, 1911 (single worker) were all placed in a morphological cluster named A. swammerdami Forel, 1886 with posterior probability p = 1.0, and we hereby propose the three subspecific names are junior synonyms of A. swammerdami.
According to the material available, A. bressleri, A. gonacantha and A. sahafina are restricted to a north/south strip of the eastern humid forests of the island (Fig. 3). Only one species, A. swammerdami, can be considered widely distributed and extremely abundant in Madagascar, with a distribution stretching along the western coast and central part of the island but completely absent in eastern regions. One species, A. makay, is only  known from humid forest in the Makay Massif. It was collected from bamboo forest on sandy soil at the headwaters of canyon streams at the base of cliffs.
Malagasy Aphaenogaster swammerdami-group species Diagnosis within Madagascar. Among the Malagasy myrmicines, Aphaenogaster swammerdami-group species have 12-segmented antennae terminating in a weakly  defined 4-segmented club or with apical 4 segments gradually increasing in size towards the apex. Queens and males are alate and characterized by the absence of the 2rsm vein in the forewing of males and gynes (Fig. 4A). The posterior portion of the head capsule of workers, gynes, and males is always drawn out into a strongly constricted neck, behind which the head capsule flares out again to form a pronounced collar (Fig. 4B).
The neck is less developed in some species but the collar is always present. Some Malagasy Pheidole species also have a neck and collar (e.g., Pheidole grallatrix Emery, 1899) but in these Pheidole the antenna terminates in a strongly defined three-segmented club while in the swammerdami group, the club is less apparent, four-segmented, and gradually increases in size towards the apex. The first gastral tergite (tergite of A4) broadly overlaps the sternite on the side of the gaster, and the sting is very reduced, never visible. The workers are large and elongate, with long, spindly legs. Workers and queens are dark red to black.
Distribution, life history, and ecology: Aphaenogaster is widespread on Madagascar and constitute important residents of drier regions and open habitats (Fig. 5A). Workers show impressive cooperation in carrying large prey items back to the nest. Their presence in the west of Madagascar can impede the use of bait, such as tuna on white paper cards, along an ant sampling transect. Instead of being attracted to the bait plastered to the paper card, A. swammerdami will recruit nest mates to carry the entire card back to the nest. Hence, when you return to check the baits, the cards are gone, only to be found stuffed into the entrances of nearby nests. In Madagascar, A. swammerdami disperses the seeds of Commiphora guillaumini (family Burseraceae) in the dry deciduous forests of the west (Böhning-Gaese, Burkhardt & Schmid, 1996). The fruits of C. guillaumini are bicolored, with a fleshy red aril and a black seed. Ants carry the seeds into their colony, remove the arils, and discard the seeds undamaged on the refuse pile at the edge of the colony. Birds may be the primary seed Throughout western Madagascar, A. swammerdami is known locally for its association with snakes (family Colubridae), most notably Dromicodryas bernieri, A. quadrilineatus, and Madagascarophis colubrinus (Preston-Mafham, 1991;Cadle, 2003). Oral tradition also notes this association between snakes and ants. For example, according to people living in the vicinity of the Bezà Mahafaly protected area, A. swammerdami hosts a snake that they call rembitiky (mother of ants). The story goes that the ants provide a nest for the snake and feed it during the cool dry season. The snake gets bigger and bigger, while the ants reduce the size of the entrance hole until the snake is no longer able to leave. The ants then eat the fattened snake during the rainy season, when it is supposedly difficult for the insects to forage outside. There is no evidence to support the idea that the ants eat the snakes. Aphaenogaster swammerdami builds large, deep nests in the soil, which undoubtedly provide good habitat for large snakes seeking shelter (Figs. 5C-5D). Jono, Kojima & Mizuno (2019) conclude that the association with M. colubrinus helps defend the ant host larvae from predation by the blind snake Madatyphlops decorsei, a specialized predator of ant larvae. Eastern species such as A. gonacantha are occasionally found nesting in the ground but more often nest in hollow cavities in dead wood on or near the ground. Species nesting in wood use carton to modify the nest entrance (Figs. 5E-5F). Key to Aphaenogaster workers in the Malagasy Region If legs are missing, the PEW/NOH and SPST/PPH ratios help separate this species from gonacantha (see Fig. 7C   Etymology. The specific epithet is a patronym referring to the late Dr. Barry Lee Bressler, retired physicist, former adjunct professor of physics at Virginia Polytechnic Institute and State University, and amateur naturalist, in recognition of his interest in myrmecology and his support for research on ants. The orthography of a patronym is unchangeable and does not depend on the generic name in which the epithet is used.

Synopsis of Malagasy
Diagnosis. Apical flanges of hind femora acute. In lateral view, sides of hind femur taper from midpoint to distal end. Distribution and Biology. This species is collected in mid-elevation humid forests (20-855 m above sea level) and in humid forests in northeastern Madagascar; its southern range overlaps with A. sahafina. According to collection event information, available ground nests of this species can be found under stones or in rotten logs, while workers can be found foraging on the ground and can be sifted from leaf litter.
Aphaenogaster gonacantha (Emery, 1899) Figures 9A-9C, Table 3 Diagnosis. Apical flanges of hind femora acute. In lateral view, sides of hind femur parallel, not tapering from midpoint to distal end, slight taper occurs just before joint with tibia. Distribution and Biology. This species is collected in humid forests at elevations from 60 m to 470 m above sea level in northern Madagascar. According to collection event information, this species nests in dead branches, rotten logs, and carton nests (Fig. 5F). Foraging workers can also be found on ground or can be sifted from leaf litter. Its distribution overlaps with A. gonacantha in its southern range near Antongil Bay.
Aphaenogaster makay Csósz & Fisher sp. n.  Distribution and Biology. The species was collected from gallery forest with bamboo on sandy soil in the Makay Massif. The bamboo forest was located at the base of cliffs at the headwaters of a canyon stream.
Aphaenogaster sahafina Csósz & Fisher sp. n.  Etymology. This species is named after its type locality, Sahafina forest. The orthography of a proper noun is unchangeable and does not depend on the generic name in which the epithet is used. Distribution and Biology. This species is collected in humid forests between elevations of 140 m and 225 m above sea level in northeastern Madagascar. The distribution overlaps in part with A. bressleri. According to collection event information, this species nests in the ground, in rotten logs, and carves out carton nests; in two cases, nests were collected from Platycerium stag-horn ferns above the ground. Foraging workers can also be found on the ground or on low vegetation.

Aphaenogaster swammerdami Forel, 1886
Figures 12A-12C, Table 3 Aphaenogaster (  Etymology. The specific epithet is a patronym referring to the Dutch biologist Jan Swammerdam (1637-1680), who carried out the first systematic anatomical studies of social insects. Swammerdam never visited Madagascar. The orthography of a patronym is unchangeable and not depend on the generic name in which the epithet is used.
Diagnosis. Apical flange of femur rounded. Tarsal  Distribution and Biology. This species is widely distributed in dry forests, coastal forests, and spiny forests as well as in cultivated lands of the Western and central highlands of Madagascar. This species typically occurs at lower altitudes between 10 and 200 m above sea level but has been found up to 1,670 m in elevation in the Central Highlands. This species nests in the ground, while foraging workers can also be found on the ground or on low vegetation. Dittmann, Dammhahn & Kappeler (2014) investigated the density, size, and feeding ecology of colonies at three different sites within the Kirindy Forest (CNFEREF) in central western Madagascar. They observed that the home ranges of A. swammerdami extended more or less circularly around the nest entrance and were established by solitary foragers. Colonies maintain exclusive territories and home ranges do not appear to overlap. Diet consisted mainly of very small items below 1 cm, including insects such as termites, ants, and larvae. They measured colony metrics at three sites. The average diameter of mounds at each site ranged between 80 and 125 cm. The estimated number of workers ranged from 200 to 1,000 workers per colony and the home range for foragers averaged 70-250 m 2 , with the largest colony having a home range of 350 m 2 . In a rice baiting experiment, A. swammerdami carried seeds to their colony, with a mean dispersal distance of 4.4 ± 1.5 m (Voigt et al., 2002).