Taxonomic revision of imitating carpenter ants, Camponotus subgenus Myrmopytia (Hymenoptera, Formicidae) of Madagascar, using morphometry and qualitative traits

Abstract The ant genus Camponotus (Mayr, 1861) is one of the most abundant and species rich ant genera in the Malagasy zoogeographical region. Although this group is commonly encountered, its taxonomy is far from complete. Here, we clarify the taxonomy of the Malagasy-endemic Camponotus subgenus Myrmopytia (Emery, 1920). Species delimitation was based on traditional morphological characters and multivariate morphometric analyses, including exploratory Nest Centroid clustering and confirmatory cross-validated Linear Discriminant Analysis. Four species are recognized: Camponotus imitator (Forel, 1891), Camponotus jodina sp. n., Camponotus karaha sp. n., and Camponotus longicollis sp. n. All four species appear to mimic co-occurring Aphaenogaster species. A diagnosis of the subgenus Myrmopytia, species descriptions, an identification key based on minor and major subcastes of workers, and the known geographical distribution of each species are provided.


Introduction
Malagasy Camponotus species are known to mimic other ant species or genera, such as Tetraponera, Catalaucus and Aphaenogaster, highlighting the incredible potential for morphological adaptation in Camponotus species (Forel 1886, Ward 2009. In this study, we assess the diversity of the Malagasy subgenus Myrmopytia. Minor workers of this subgenus all appear to mimic Aphaenogaster. In 1891, when describing Camponotus imitator, Forel noted that the minor workers exhibited strikingly similar morphology to the myrmicine ant Aphaenogaster swammerdami Forel, 1886 (Formicidae: Myrmicinae) (see Figure 1). Little is known about the evolutionary advantage of this phenomenon, although where A. swammerdami occurs, it is a dominant species and there might be some level of protection in imitating a dominant ant. The genus Camponotus (Mayr, 1861) is among the most diverse and abundant ant lineages in Madagascar. Of the more than 1500 valid species and subspecies worldwide, 78 Camponotus species are known from the Malagasy region (Antcat.org) with an additional estimated 100 undescribed taxa (AntWeb.org). Several complications have hindered progress in the refinement of Camponotus taxonomy. The first is morphological variation, especially the high level of intraspecific and intranidal polymorphism; the presence at times of three morphological castes (minor, media and major); and the poorly tested diagnostic value of morphological traits that make species delimitation using conventional approaches extremely difficult. The second difficulty is linking past works to current taxon concepts, as these same morphological challenges complicate the association of previously named type specimens to current species hypotheses. The third complication is that the existing subgenus classification of Camponotus is an amalgamation of past taxonomic works that have tried to "fix" the problem of taxa delimitation by adding new names rather than by completely revising the global classification (Bolton 1995, Clouse et al. 2015, Ward et al. 2016. We have attempted to overcome the problem of morphological variation in the Malagasy-endemic Camponotus subgenus Myrmopytia (Emery, 1920) by combining traditional morphology (i.e. evaluation of qualitative traits) with a numeric morphology-based approach. For quantitative analyses of morphometric data we follow the protocol introduced by Csősz and Fisher (2016) using a combination of NC-clustering (Seifert et al. 2014) and PART (Nilsen and Lingjaerde 2013). In this protocol, boundaries of operational taxonomic units (OTUs) are tested via cross-validated Linear Discriminant Analysis (LOOCV-LDA). In order to obtain high quality resolution of morphometric analyses we extended the existing set of morphometric characters by including many new traits. Our extended character set includes 19 traits of which four are defined for the first time.
In terms of overcoming the challenges of historical names, our work builds on 15 years of tracking down type material of all Malagasy taxa from across European collections to assess the identity of previous descriptions in relation to more recently collected material. Although we are still missing a small percentage of Camponotus type specimens, the identities of most names have been clarified. Lastly, this work does not address the issue of the unnatural subgeneric classification for the genus which is better addressed with molecular phylogenetic methods and global sampling. Instead, for this revision, we retain the historic subgenus Myrmopytia, for Camponotus imitator (Forel, 1891), and include for the convenience of this revision, species with morphologically similar workers.
All species are described here based on the worker caste, and the key provided combines qualitative characters and morphometric ratios that will help resolve the most problematic cases. Distribution maps are also provided.

Materials
The morphometric analysis was based on 130 workers principally collected by B.L. Fisher and the Madagascar Biodiversity Center team in 74 collecting events across Madagascar. Specimen data and images for material examined in this study is available on AntWeb (http://www.antweb.org) and can be accessed using the unique specimen identifiers (e.g. CASENT0101365).

Imaging
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).

Mapping
Distribution maps for all species were generated by importing specimen distribution records into the Diva-GIS program (Hijmans et al. 2011). For material lacking locality data such as older type material, georeferenced coordinates are placed in brackets. Georeferencing was completed with the aid of online maps and the Missouri Botanical Garden's Gazetteer to Malagasy Botanical Collecting Localities (2015).

Morphometric character recording and terminology
Measurements were taken with a Leica MZ 9.5 stereomicroscope equipped with a crossscaled ocular micrometer. Each worker was evaluated using 19 continuous morphometric traits, measured as in Rakotonirina et al. (2016) for the Camponotus edmondi group. The morphometric data are expressed in µm. All measurements were made by NR allowing to achieve the highest consistency in character recording throughout the entire data collection procedure. Definitions and abbreviations for the measured characters are as follows (see Figure 2):

CL
Maximum cephalic length. The maximum median length of the head in fullface view, measured from the midpoint of the posterior margin of head to the midpoint of the anterior margin of the clypeus ( Fig. 2A).

ClyL
Clypeal length. The maximum median length of the clypeus measured from the posterior margin to the anterior margin in frontal view, in which the anterior and posterior clypeal margins are aligned to the same focus. Median concavity on either or both margins reduces the length of the clypeus ( Fig. 2A).

CS
Cephalic size. This derived character is used as body size indicator and is calculated from the arithmetic mean of head length (CL) and maximum head capsule width (CWb).

CW
Maximum cephalic width. The longest distance between the lateral margins of the compound eyes in full-face view ( Fig. 2A).

CWb
Maximum head capsule width. The maximum width of the head capsule excluding the compound eyes ( Fig. 2A).

EL
Eye length. Maximum diameter of the compound eye ( Fig. 2A).

FR
Frontal carina distance. The longest distance between the frontal carinae ( Fig. 2A).

GPD
Maximum tentorial pit distance. The longest distance between the centers of the fossae located at or very close to the posterolateral margin of the clypeus ( Fig. 2A).

HTL
Maximum hind tibia length. Straight line length of the hind tibia measured from the constriction immediately before its proximal insertion to its distalmost point, excluding the bristles or spines (Fig. 2B).

ML
Mesosoma length: The longest anatomical line that connects the posteriormost point of the propodeal lobe with the anteriormost point of the pronotal collar; preferentially measured in lateral view, but if one of the reference points is not visible, dorsal view may help (Fig. 2B).

MPD
Mesothoracico-propodeal distance. With the promesonotal suture and the anterior petiolar foramen margin in the same plane of focus in dorsal view, the maximum length between the promesonotal suture and the posteriormost point of the propodeal process dorsal to the petiolar insertion (Fig. 2C).

MPH
Mesothoracico-propodeal height. Measured in lateral view, using as a diagonal reference line that connects the anteriormost point of the pronotal shield and the posteriormost point of the propodeal process dorsal to the petiolar insertion, MPH is the perpendicular distance between two lines parallel to this line, one of which just touches the anteroventral corner of the mesopleuron, dorsal to the insertion of the mesocoxa, and other which the dorsalmost point of the propodeum (Fig. 2B).

MW
Mesosoma width. Maximum width of the pronotum in dorsal view, which in the Myrmopytia is also the maximum mesosomal width (hence "mesosoma width") ( Fig. 2C).

NOH
Petiolar node height. The maximum distance between the petiolar spiracle and the dorsalmost point of the petiolar node (Fig. 2B). OMD Oculo-mandibular distance. The minimum distance between the anterior margin of the compound eye and the mandibular insertion to the head (Fig. 2B).

PEW
Petiolar width. The maximum width of the petiole in dorsal view (Fig. 2C). PoOC Postocular distance. The longest distance between the posteromedian margin of the head and the reference line set on the posterior margins of the two compound eyes. Measured at a right angle to the reference line in full-face view ( Fig. 2A).

PrOc
Preocular distance. The longest distance between the anteromedian margin of the clypeus and the level of the anterior margin of the compound eyes as reference line. Measured at a right angle to the reference line in full-face view ( Fig. 2A).

SL
Scape length. Straight line length of the first antennal segment excluding the basal condyle (Fig. 2B). TCD Torular carina distance. The minimum distance between the torular arches that surround the antennal insertion ( Fig. 2A).

Morphometric data analysis
Altogether 130 worker individuals (96 minor and 34 major workers) were measured and analyzed using multivariate statistics based on the protocol described by Seifert et al. (2014) and extended by Csősz and Fisher (2016). Camponotus species may produce two to three subcastes of workers exhibiting quite different allometric properties. The Linear Discriminant Analysis (LDA) cannot properly analyze non-linearly scaled data. Therefore, all specimens were sorted into two subcastes (minor and major were found in the present subgenus) via visual inspection of scaling properties using pairwise matrix scatterplots (see also Rakotonirina et al. 2016). With this procedure, linear within-class allometries were achieved for each trait. Thanks to the higher number of minor workers (96 were minors out of 130 total) this subcaste was selected for the purpose of morphometric data analyses. Morphometric data and descriptions for major workers are also provided, but their morphometric data were not included in statistical processing.
Generation of species hypotheses by combined application of NC-PART clustering and confirmatory Linear Discriminant Analysis (LDA) We evaluated the morphometric data recorded from 96 minor workers belonging to 66 samples. All samples represent nest series or individuals from one collecting event. We used a combined application of NC-clustering (Seifert et al. 2014) and PART algorithm implemented in 'clustergenomics' package (Nilsen and Lingjaerde 2013) under R environment (R Core Team 2014).
The optimal number of clusters and the partitioning of samples done by Partitioning Based on Recursive Thresholding (PART) were accepted as improved species hypothesis worth testing if the two clustering methods, 'hclust' and 'kmeans' via PART arrived at the same conclusion (see Seifert et al. 2014, Rakotonirina et al. 2017. Confirmatory Linear Discriminant Analysis with leave-one-out cross-validation (LOOCV-LDA) was run to confirm species boundaries. The classification returned by the quantitative analyses were further verified by searching for qualitative differences of morphological traits between clusters.

Results and discussion
The NC-PART clustering found three clusters (see numbered bars 1 (black), 2 (red) and 3 (blue) in Figure 3). Qualitative investigation of morphological traits found overall support for these three clusters. However, in cluster 2 (red bar in Figure 3) two samples remarkably differ in morphology from the other specimens in the cluster. Based on the distinct phenotype of the two individuals of the satellite cluster (purple specimens in Figure 3), we propose that these two individuals represent an additional species. These individuals also appear slightly separated in the NC-clustering dendrogram. A larger sample size, however, would be required to statistically evaluate the independence of the cluster using 'kmeans' or 'hclust'. Based on the morphometric and qualitative analysis, four species are hereby named as C. imitator Forel, 1891, C. longicollis sp. n., C. karaha sp. n. and C. jodina sp. n. (see Figure 3).
The confirmatory Linear Discriminant Analysis with leave-one-out cross-validation (LOOCV-LDA) yielded complete classification success and every specimen was correctly classified to one of the four species. It is important to note that even though the two specimens of C. jodina were correctly classified, the small sample size (2 specimens) is below the sample size criterion for LDA. As supporting evidence for C. jodina, univariate comparison of body ratios, PEW/CS and MPD/CS of samples of C. jodina and the most similar species, C. karaha yield non-overlapping ranges ( Table 1).

Type-species. Camponotus imitator
Minor workers of this subgenus can be recognized by the combination of the following characters: maxillary palps particularly long with respect to head length, extending past a point on line with posterior margin of eye; mesosoma slender, mesonotum elongated and constricted at midlength, propodeum a protruding rounded hump; mesothoracic spiracles prominent, pointing upward; propodeum protuberant. The major worker can be recognized by the distinct form of the mesosoma. The propodeum is dome-like or subrectangular in lateral view and the metanotum is elongate and impressed, and the upper mesopleuron is laterally pinched (see Figures 10A, 14A, 17A). Minor and major: in profile, petiole relatively low and nodiform, with the anterodorsal angle higher than the petiolar spiracle, NOH/CS <0.146 (Fig.  6A); anterior tentorial pit slightly impressed so that the posterior portion of clypeus protrudes weakly in lateral view (Fig. 7A).    propodeum more rectangular, with a raised anterodorsal transverse carina; integument of mesonotum and propodeum strongly sculptured (Fig. 8B)     Color: head, mesosoma, petiole, and base of first gastral segment reddish brown, remainder of gaster dark brown to black. Scattered appressed pubescence generally present. Setae light brown.

Synoptic species list of the subgenus
Distribution and biology. The minor worker of Camponotus imitator is thought to mimic the myrmicine ant Aphaenogaster swammerdami due to its color and the form of its constricted mesonotum and shape of propodeum, which could appear as a petiole in dorsal view (Forel 1891) (Fig. 11). This myrmicine nests underground and shares its nests with snakes, Madagascarophis colubrinus (Schlegel, 1837) and Leioheterodon modestus (Günther, 1863); it is an important secondary seed disperser of Commiphora guillaumini (Burseraceae) (Böhning-Gaese et al. 1999).
Camponotus imitator is distributed in the dry forest and woodland of western and southern Madagascar at elevations ranging from 25 m to 990 m (Fig. 11). Its distribution is sympatric with A. swammerdami through most of its range (Fig. 11). It has been collected by litter sifting, Malaise and pitfall traps, as well as beating low vegetation and from the ground in rotten logs. This species nests underground.
Comment. We propose that Camponotus imitator resinicola (Santschi, 1911) is synonymized with Camponotus imitator Forel. In the original descriptions, the former differs from the latter by the presence of reddish patches on the first gastral segment near the petiolar insertion. Examination of material from 10 collection events of C. imitator colonies indicates that this trait is highly variable within colonies, and no other reliable characters were found to separate the subspecies from imitator. Moreover, no other qualitative trait or biogeographic evidence exists that would underpin the subspecies status of resinicola.
Additional  Diagnosis. Camponotus jodina is easily distinguishable from the other species of Myrmopytia on the basis of the following character combination: petiole surmounted by a conical node terminated with a blunt spine, PEW/CS 0.152 [0.149, 0.156], propodeum coarsely reticulate with a short vertical anterior face.
Description of minor worker. Head longer than wide, rear portion of head extended into a long neck, sides of head narrowed in front of eyes. CS 1.84 mm [1.80, 1.88] (n=2). Posterior margin of head glabrous, suberect genal setae present on sides of head in full-face view. Eyes located anterior to the midlength of head capsule in full-face view,  Entire body including antennae dark brown. Coxae, femora, and tibiae brown. Head, pronotum, fore-coxae and petiole microreticulate, mesonotum and propodeum reticulate-costate.
Major worker. Unknown. Queen. Unknown. Male. Unknown. Distribution and biology. C. jodina is known from only two specimens collected in Parc National de Zahamena, in rainforest, at two different localities: Onibe River and Sahavorondrano River (Fig. 18A). Workers were found foraging on lower vegetation. C. jodina occurs sympatric with C. karaha at the Sahavorondrano River.
Etymology. This species is named for the shape of its petiole in profile; the Malagasy word "jodina" means directed upward.
Diagnosis. Workers of Camponotus karaha can be differentiated from the other three species by the triangular form of the petiole in lateral view, and the protruding clypeus.
Entire body generally black, pronotum may vary to brown, coxae a lighter color than mesosoma, basitarsus and up to 5 basal funiculi whitish. Mesonotum and pro- Entire body black; femora and tibia dark brown, basitarsus of second and third legs light brown, funicular segments light brown, becoming dark apically. Pronotum and mesonotum microreticulate, sides of propodeum finely punctate, standing filiform setae present in all surfaces.
Distribution and biology. Camponotus karaha is currently known from 14 localities along the eastern rainforest and montane rainforest of Madagascar at elevations ranging from 175 to 1325 m (Fig. 18B). Specimens have been collected on the ground in rotten logs and on low vegetation.
Etymology. The Malagasy word "karaha" means similar, look-alike. Notes on morphological variability. Workers of C. karaha exhibit morphological variability in qualitative traits such as sculpture, color of mesosoma, and profile of propodeum that differ between populations. This divergence is not, or barely, supported by multivariate analyses involving 19 quantitative traits. For this reason, we conclude that all populations examined represent a single species and ascribe the variation to intraspecific variability of populations occupying diverse sites, making geographic (e.g. elevation) or ecological factors as possible explanations for the variance. Populations from the north of its distribution range differ notably in shape of propodeal dorsum and petiolar node and further research on additional samples are needed to further evaluate species status of these populations. For now, we note the differences of the northern populations from the more typical karaha.
Variant 1. This variant is the typical C. karaha, and is fairly widespread throughout the eastern rainforest of Madagascar. It can be recognized by having a propodeum dorsum smoothly convex in lateral view, declivitous face of petiole distinctly flat with defined lateral margins, petiolar apex forms a right angle (Fig. 16A).  Entire body dark brown to black; femora and tibiae dark brown, basitarsus of second and third legs light brown, funicular segments light brown becoming dark brown apically. Head and pronotum microreticulate, mesonotum smooth and shiny, propodeum and fore-coxae moderately reticulate-punctate.
Description of major worker. Anterior margin of clypeus projecting to an obtuse angle, dorsal outline of clypeus smoothly convex in profile; masticatory margin with 6 teeth, palps long with respect to head size. Similar appearance as minor worker. Anterior region of mesonotum (immediately posterior to the pro-mesonotal suture) in profile rising above pronotum. Mesonotum straight and elongate, distinctly compressed laterally anterior to mesothoracic spiracles. Suberect filiform setae present, propodeum covered with dense pubescence, pubescence generally sparse elsewhere. Mesothoracic spiracles feebly produced laterally; Overall color black; basal funicular segments yellowish brown and becoming dark apically; basitarsus of second and third legs light brown. Color of filiform setae whitish. Head and pronotum finely reticulate, mesonotum smooth and shiny, clypeus, gena, propodeum and fore-coxae moderately reticulate punctate.
Queen. Unknown. Male. Unknown. Distribution and biology. C. longicollis has only been collected at two localities in the northwest Sambirano region of Madagascar, Réserve Spéciale de Manongarivo and Mont Kalabenono and Mont Galoko on the Galoko mountain chain (Fig. 18C). It has been collected via litter sifting and hand collecting on a rotten log.
Etymology. This species name is based on the Latin terms for long, "longi", and collis, "neck".