Taxonomy and species-groups of the subgenus Crematogaster ( Orthocrema) in the Malagasy region (Hymenoptera, Formicidae)

Abstract The species-level taxonomy of the subgenus Crematogaster (Orthocrema) in the Malagasy region is evaluated with both morphological data from worker and queen ants, and genetic data from three nuclear markers (long wavelength rhodopsin, arginine kinase and carbamoylphosphate synthase). These two types of data support the existence of six Orthocrema species: Crematogaster madecassa Emery, Crematogaster rasoherinae Forel, Crematogaster telolafy sp. n., Crematogaster razana sp. n., Crematogaster volamena sp. n. and Crematogaster mpanjono sp. n.. Two new synonyms of Crematogaster rasoherinae Forel are recognized, Crematogaster rasoherinae brunneola Emery, syn. n. and Crematogaster voeltzkowi Forel, syn. n., as these were not supported as distinct taxa by the data. A neotype is designated for Crematogaster rasoherinae; lectotypes are designated for Crematogaster madecassa, Crematogaster rasoherinae brunneola and Crematogaster voeltzkowi. Species descriptions, images, distribution maps and identification keys based on worker and queen ants are given for all six species. A diagnosis of the subgenus Orthocrema in the Malagasy region is presented for both workers and queens. Within the Malagasy Orthocrema, three distinct phylogenetic lineages are suggested by molecular and morphological data. Newly defined monophyletic species-groups are thus the Crematogaster madecassa-group (Crematogaster madecassa, Crematogaster telolafy and Crematogaster razana) and the Crematogaster volamena-group (Crematogaster volamena and Crematogaster mpanjono); Crematogaster rasoherinae represents an isolated lineage in the Malagasy region and its closest relatives remain unclear. Other interesting biological findings are the presence of an intermediate caste between workers and queens in Crematogaster rasoherinae and Crematogaster madecassa, and unusually large workers in Crematogaster volamena resembling a major caste.


introduction
The large and species-rich ant genus Crematogaster Lund (> 450 described species; Bolton 2011) has a global distribution throughout most forest and savannah habitats in warm-temperate to subtropical and tropical climates. Crematogaster species diversity is highest in the tropics and subtropics however, where these ants often form a dominant part of the local ant community. The majority of Crematogaster species nest arboreally, for example in twigs or under bark, but ground nesting seems to be equally common in temperate and cooler climates and also more prevalent in some species-groups in the tropics. The species-level taxonomy of Crematogaster ants is notoriously difficult and encumbered by synonyms and ambiguous subspecies names (Longino 2003;Ward 2010). Blaimer (2010) provided a comprehensive review on the natural history and taxonomic state of Crematogaster.
In the Malagasy region, here defined as Madagascar and the surrounding Indian Ocean islands, the taxonomy of Crematogaster is currently being revised in a series of publications, subdivided into the several distinct species-groups present in the region (see Blaimer 2010Blaimer , 2012. The total species diversity of Crematogaster in the Malagasy region is estimated to be approximately 33 species (Blaimer, unpublished data). This estimate represents a mixture of previously described species, species new to science, and reductions due to synonymy. Most of these species occur only in Madagascar, but five species also are found on the Comoros Islands, Mayotte and the Seychelles. Recent intensive inventories of arthropods and especially the ant fauna in the Malagasy region (see e.g. Fisher and Penny 2008) have immensely increased the extent of available specimens for revisionary work, generating much more complete distribution records for already described species and discovering numerous undescribed new species.
The present study is part of this larger revisionary work and treats all species associated with the subgenus Orthocrema Santschi (1918) in the Malagasy region. Recent molecular work (Blaimer, in prep.) has found Malagasy species placed in the subgenus Mesocrema Santschi (1928) to be closely related to the former, and these are therefore included with Orthocrema in the present revision. This altered classification follows anticipated changes in the subgeneric classification of Crematogaster in the near future, based upon a molecular phylogenetic framework (Blaimer in prep.).
Up to now, one species has been described from the Malagasy region for Orthocrema, C. madecassa Emery, whereas three species and subspecies have been described for Mesocrema: C. rasoherinae Forel, C. rasoherinae brunneola Emery and C. voeltzkowi Forel. The latter has been recorded exclusively from the Comoros Islands, whereas the other species were first described from Madagascar. My observations suggest that the Malagasy Orthocrema present an exception to the predominantly arboreal life habit of Crematogaster ants in this region. Most species in this group appear to be generalists, as they have been collected nesting both on the ground in rotten logs or branches, or arboreally in dead twigs or bark and canopy moss mats. A very interesting aspect of the biology of some of the Malagasy species in this group is the presence of intermediate workers in the colony. These possess morphological features that are intermediate between workers and queens, but their function and behavior in the colony remains unclear. Intermediates have also been reported in the North American Crematogaster (Orthocrema) species C. smithi, where they were denoted as 'large workers' (Heinze et al. 1999). In the case of the latter, it was shown that these had the ability to lay unfertilized trophic eggs, but were not capable of sexual reproduction (Heinze et al. 1999(Heinze et al. , 2000. The presence of this separate caste may be a more widespread phenomenon in Orthocrema species.
In the following, I focus on a reevaluation of the presently described Malagasy Orthocrema species with both morphological and molecular methods, and further describe new species that are supported by these two types of data. A second aim of this study is the delimitation of two morphologically and genetically distinct speciesgroups within Malagasy Orthocrema.

Morphological study
All morphological observations were made with a Leica MZ12.5 stereomicroscope. Standard measurements (in mm) were taken at 50× with a Wild M5A stereomicroscope and a dual-axis Nikon micrometer wired to a digital readout. Measurements are given to the second decimal place, and indices are presented as decimal fractions (also to the second decimal). Ranges are always expressed as minimum -maximum values. Measured specimens were chosen to represent the entire distribution range of a given species. The abbreviations used for measurements and indices below follow Blaimer (2010) and Longino (2003); for illustrations of these see Blaimer (2010).

Measurements and indices HW
Maximum head width including eyes, in full face view. HL Head length; perpendicular distance from line tangent to rearmost points of vertex margin to line tangent to anterior most projections of clypeus, in full face view. EL Eye length; measured along the maximum diameter. SL Scape length; length of scape shaft from apex to basal flange, not including basal condyle and neck. If scape is strongly arched, this measurement is taken as the chord length from the basal flange to the apex.

PTL
Petiole length; measured in lateral profile as the distance from dorsoposterior margin of segment to anterior inflection point where petiole curves up to condyle.

PTH
Petiole height; measured in lateral profile as vertical distance from ventral margin to highest point of dorsoposterior margin.
Color images were created with a JVC KY-F75U digital camera, a Leica MZ16A stereomicroscope and ZERENE STACKER (v1.02) software. The scanning electron microscope images were taken at the California Academy of Sciences using a Zeiss/LEO 1450VP SEM. All ant images presented here are also publicly available on AntWeb (www.antweb.org). Line drawings were produced by tracing color images in Adobe Illustrator CS5. Species distributions were plotted with ARCMAP (v9.3) within the software ARCGIS, based on coordinates (latitude and longitude) as given on the specimen labels of all material (see also supplementary table 1 for a species list with GPS coordinates). For material lacking this information, i.e. syntype specimens, the following sources were used to georeference collection sites: the GEOnet Names Server (National Geospatial-Intelligence Agency 2010) and the Gazetteer to Malagasy Botanical Collecting Localities (Schatz and Lescot 2003). Classification of major geographic regions in Madagascar throughout species descriptions follows Gautier and Goodman (2003). Common abbreviations within locality data are: P.N. = Parc National, R.S. = Réserve Spéciale, F = Forêt, P.C. = Parc Naturel Communautaire, R.N.I. = Réserve Naturelle Intégrale.
The International Commission on Zoological Nomenclature (1999) requires lectotypes designated after 1999 to "contain an express statement of deliberate designation" (amended Article 74.7.3). I use the statement 'lectotype by present designation' to fulfill this requirement. Lectotypes have been designated where a name lacks a holotype or lectotype and unambiguous syntypes have been identified. The purpose is to provide stability of nomenclature, and designation is done in a revisionary context in agreement with the amended Recommendation 74G of Article 74.7.3. Neotype designations have further been made for names with no extant name-bearing types that are in need of a namebearing type "to objectively define the nominal taxon" (Article 75.1, ICZN, 1999), and are in agreement with the qualifying conditions stated in Article 75.3 (ICZN, 1999

Molecular data collection and phylogenetic analyses
After sorting all available specimens to morphospecies, one to four individual worker ants for each of six putative Malagasy Crematogaster (Orthocrema) species were selected for genetic analysis. Four non-Malagasy Orthocrema species were chosen as outgroups, given their approximate relationships to the Malagasy taxa as known from a previous, larger phylogenetic analysis (Blaimer, in prep.). Two of these (C. sordidula Nylander and C. longipilosa Forel) represent distant relatives to all Malagasy Orthocrema, whereas the remaining two taxa (C. arcuata Forel and C. cf. dolens Forel) are closer relatives to the Malagasy taxa. For the distribution of the sampled taxa refer to Table 1. From these 17 specimens, DNA was extracted from either entire worker adults or pupae using a DNeasy Tissue Kit (Qiagen Inc., Valencia, California, U.S.A.), following the manufacturer's protocol but eluting the extract in sterilized water rather than the supplied buffer and at half the suggested volume. I used either a non-destructive method (cuticle of ant pierced prior to extraction, mostly used for adults), enabling me to retain and re-mount voucher specimens after extractions, or a destructive technique (entire ant pulverized, mostly used for pupae) in cases where multiple individuals from the same colony series were available. Three nuclear protein-coding genes were amplified: long wavelength rhodopsin (LW Rh, 856bp exon /255bp intron), arginine kinase (ArgK, 388bp exon/177bp intron) and carbamoylphosphate synthase (CAD, 529bp exon/252bp intron). The sequence lengths given here refer to the aligned sequence data included in phylogenetic inference and add up to a total of 2457bp. The three amplified genes are widely used for phylogenetic inference in ants and primers are available (Ward and Downie 2005;Brady et al. 2006;Moreau et al. 2006;Ward et al. 2010;Blaimer in prep.), and their usefulness in phylogenetic inference between closely related species has been demonstrated (Lucky 2011;Blaimer 2012). Amplifications were performed using standard PCR methods outlined in Ward and Downie (2005) and sequencing reactions were analyzed on an ABI 3730 Capillary Electrophoresis Genetic Analyzer with ABI BigDye Terminator v3.1 Cycle Sequencing chemistry (Applied Biosystems Inc., Foster City, CA). All sequences have been deposited in GenBank, with accession numbers listed in Table 1; the data matrix and tree used to create Fig. 1 have further been deposited in TreeBase (ID 12240; available at: http://purl.org/phylo/treebase/phylows/study/TB2:S12240). Sequence data were assembled and edited in the program SEQUENCHER 4.6 (Gene Codes Corporation, 2006, Ann Arbor, MI), aligned in CLUSTALX 2.0.12 (Thompson et al. 1997;Larkin et al. 2007), and corrected by eye in MACCLADE 4.08 (Maddison and Maddison 2000). Phylogenetic analyses within a Bayesian framework (BI hereafter) were performed using MRBAYES v3.1 (Ronquist and Huelsenbeck 2003), accessed through the CIPRES science gateway (Miller et al. 2010); analyses within a maximum likelihood framework (ML hereafter) used GARLI v2.0 (Zwickl 2006) and were performed on an IMac desktop computer. BI-and ML-analyses were based on a concatenated data matrix of the three loci. The data matrix was divided into nine data subsets by gene, translational pattern (exon, intron) and codon position (1 st + 2 nd vs 3 rd ). Best-fitting models of nucleotide sequence evolution were selected for each partition using the Akaike information criterion (AIC) in the program MRMODELTEST v2.3 (Posada and Crandall 1998;Nylander et al. 2004) for application in BI-analyses, and in MODELTEST v3.7 (Posada and Crandall, 1998) for specification in ML-analyses, both executed through PAUP* 4.0b10 (Swofford 2000). Selected models for each data subset can be found in Table 2. BI-analyses each employed two runs of Metropolis-coupled Markov Chain Monte Carlo (MCMCMC) consisting of four chains (temp=0.05) and sampling every 1000 generations. The model parameters transition-transversion ratio, gamma shape, proportion of invariable sites, rate matrix and state frequencies were unlinked across partitions, and a variable rateprior was employed to allow for rate variation among partitions. Convergence of chains and other diagnostic values were assessed in several ways. In MRBAYES I confirmed that the ASDSF had reached values well below 0.01 and PSRF values had approached 1.0 for all parameters. In TRACER v1.5 (Rambaut and Drummond 2007), convergence was confirmed visually and mixing of chains was evaluated with effective sample size (ESS) values. To assess whether tree topologies were sampled in proportion to their true posterior distribution, I further used the compare, slide and cumulative plotting functions on the AWTY-online server (Wilgenbusch et al. 2004). All the above indicators returned good values after MCMCMC-sampling for 20 million generations; consensus trees were summarized in MRBAYES after discarding 25% of samples as burnin. I further performed a ML-search for the best scoring tree (results not shown), as well as a bootstrap search with 100 replicates in GARLI. Program configuration settings were left at defaults. Trees resulting from the bootstrap search were summarized as majority-rule consensus tree in PAUP* 4.0b10 (Swofford 2000).
Ancillary genetic data supporting the results outlined below has been generated through the joint barcoding initiative of Malagasy ants by the California Academy of Sciences and the Biodiversity Institute of Ontario, Guelph, Canada (www.barcodinglife.org). The barcoding region of cytochrome oxidase I (COI) for ~130 specimens of five of the six below recognized species (with variable taxon sampling of 2-101 individuals per species) was thus available to guide taxonomic decisions. Analyses of these data are to be published elsewhere.

Molecular results
All molecular phylogenetic analyses (BI and ML) of the data strongly suggest that there are six species of Malagasy Orthocrema, namely the previously described C. rasoherinae and C. madecassa and four new species: C. telolafy sp. n., C. razana sp. n., C. volamena sp. n. and C. mpanjono sp. n. (Figure 1). The previously described C. voeltzkowi from the Comoros Islands shows little genetic differentiation from C. rasoherinae, warranting synonymy with the latter (as indicated in Figure 1A). The same applies to C. rasoherinae brunneola (as indicated in Figure 1A). Further supported is the presence of two distinct species-groups, the C. madecassa-group and the C. volamena-group, with members as listed below. Crematogaster rasoherinae is shown as quite distantly related to the C. madecassa and C. volamena species-groups, which in turn also clearly Table 2. Data partitions and selected substitution models. Information on data subsets, including number of bases, number of variable characters (VC), number of parsimony-informative characters (PIC) and substitution models selected for the respective partition using the Akaike information criterion in MRMODELTEST v2.3 (Posada and Crandall 1998;Nylander 2004) or MODELTEST v3.7 (Posada and Crandall 1998) for application in BI-or ML-analyses respectively. do not form a monophyletic grouping. This suggests these two species groups and C. rasoherinae have originated from separate ancestors and three colonizations of the Malagasy region took place within the Orthocrema lineage. The exact relationships of these two species-groups and of C. rasoherinae to each other, and to the non-Malagasy taxa C. arcuata and C. cf. dolens remain unclear as they receive only moderate support in the BI analysis ( Figure 1A), and are unresolved in the ML analysis ( Figure 1B).

Minimal diagnosis
A combination of characters 6, 7, 8 and 9 will unequivocally separate workers of Orthocrema species from the remaining Crematogaster species in the Malagasy region.
Worker description (Figures 24A-F). Very small species (HW 0.51-0.64, WL 0.52-0.68). Masticatory margin of mandibles with 4 teeth; clypeus with several weak vertical carinae; posterior margin of head in full face view usually laterally rounded, sometimes medially slightly depressed; occipital carinae well pronounced; antennal scapes usually just reaching, but not surpassing posterior margin of head; midline of eyes situated well above midline of head in full face view; eyes flush with head, not notably protruding.
Promesonotum laterally subangular, with mesonotum posterolaterally slightly marginate and metanotal groove bordered by weak carinae; in lateral view outline of promesonotum moderately convex; promesonotal suture usually absent; mesonotum with or without a distinct posterior face; metanotal groove with 2-3 median carinae of varying prominence; propodeal spines short (SPI < 0.17) or absent (most Comoros Isl. material), if present straight or upwards curved, in lateral view directed upwards, in dorsal view almost parallel and not diverging; dorsal face of propodeum very short; petiole in dorsal view rectangular, with dorsolateral margins weakly carinate or angular and small antero-and posterolateral denticles; subpetiolar process mostly developed as broad, rounded protuberance, sometimes as small angular dent; postpetiole more or less globular, merely impressed posteriorly, or with faint median impression; subpostpetiolar process often present as small, angular protrusion.
Several color variants. Most widespread in Madagascar is a light to dark brown form; less common is a bicolored form with light brown or reddish head and mesosoma and dark gaster. On the Comoros islands, the Seychelles and Mayotte, C. rasoherinae is most often yellow or pale yellow colored, often with the posterior half of the gaster black. The typical brown Madagascar color form seems to be only present on the Seychelles.
Intermediate worker measurements ( Intermediate worker description (Figures 25A-E). Intermediate between workers and queens in size. Head, petiole and postpetiole characters similar to queens; ocelli present, but smaller than in queens; the mesonotum is to various extent raised and fused dorsally over pronotum and has wing attachment sutures; otherwise mesosomal characters more similar to worker characters, especially propodeum, and propodeal spines are present.
Queen measurements ( (Figures 26 A-C). Very small (HW 0.80-0.89, WL 1.50-1.63). With worker characters, except as follows. Masticatory margin of mandibles with 5 teeth. Antennal scapes not surpassing posterior margin of head, reaching only to about level of lateral ocelli; eyes large (OI 0.30-0.34), situated at midline of head in full face view; head shape quadrate (CI 0.99-1.02), posterior margin of head straight.

Queen description
Mesosoma slender (MSNI 1.72-1.91, WL 1.50-1.63); mesoscutum in dorsal view oval, about half as wide as long; dorsal face of propodeum distinct, about half as long as posterior face; propodeal spines absent; petiole and postpetiole as in worker; anteroventral subpetiolar tooth present, but reduced with respect to worker.
Sculpture smooth and shiny throughout; erect pilosity generally more abundant, but finer than in workers: face with 4-6 longer erect setae and abundant shorter erect to suberect pilosity; mesonotum with abundant short, and scattered longer erect setae; petiole with one pair of long flexuous setae posterior to posterior denticles; postpetiole with flexuous pair of dorsoposterior setae and 2-4 additional long setae; petiole and postpetiole with abundant shorter pilosity throughout. Body color similar to respective workers.
Mesosoma fairly slender (MSNI 1.56-1.80, WL 0.64-0.80); mesoscutum in dorsal view slightly wider than long; scutellum with two distinct faces: anterior face short and steeply sloping from mesoscutum, posterior face long and flat; scutellum in dorsal view oval and posteriorly rounded, but dorsoposterior margin carinate; dorsal face of propodeum about as long as posterior face; propodeal spines absent; petiole in dorsal view more or less rectangular, but carinae or denticles absent and all margins rounded, in lateral view petiole anteriorly tapering; anteroventral subpetiolar tooth absent; postpetiole globular, but dorsally somewhat compressed and flat, median impression absent; wings clear.
Sculpture smooth and shiny throughout; face with 2 longer erect setae close to ocelli and sparse short suberect pilosity; mesoscutum with scattered short erect or suberect pilosity; longer erect pilosity present on posterior part of scutellum; petiole with one pair of fine, erect setae; postpetiole with fine dorsoposterior setae. Head medium brown, mesosoma pale yellow, metasoma light brown.

Distribution and biology. Crematogaster rasoherinae is the most widespread species of the Malagasy Orthocrema, and in fact it is one of the most abundant Crematogaster species in Madagascar and on a number of Indian Ocean islands (Figures 26 D-F). The species
is distributed throughout all native forest habitats in Madagascar -rainforest, dry or spiny forest alike -and is also found in disturbed habitats and urban areas. It has been collected at elevations up to 1225m, but appears to be more common at lower elevations. In natural habitats this species is predominantly arboreal nesting, both in dead twigs as well as in live plant parts. However, ground nesting in rotten logs or branches does occur occasionally.
Biologically C. rasoherinae is interesting because of the presence of intermediate workers (i.e., individuals intermediate between workers and queens) (Figures 25 A-E) of unknown function in the colony. I have found intermediates in all four colonies that I collected of this species, with the highest number hereby being eight individuals in one nest. In all cases a normal, dealate queen also was present in the nest. None of the intermediates observed was winged and it seems likely that they are either entirely wingless or brachypterous. Scanning electron micrographs of the lateral mesosoma ( Figures 25 C and D) show the presence of a rudimentary suture above the mesopleuron where in a normal queen the forewing attaches.
Discussion. An intriguing characteristic of C. rasoherinae is the morphological variability of this species on the Comoros Islands compared to the remainder of its distribution range. In Madagascar, the Seychelles and Mayotte this species always possesses propodeal spines. On the Comoros Islands in contrast, propodeal spines can be present (as in Figure 24B), reduced or entirely absent (as in Figure 24E). More specifically, all specimens examined from the island of Grand Comore have no, or very reduced propodeal spines, whereas on Anjouan and Moheli propodeal spines are mostly reduced or absent and present only in fewer individuals. This spine-polymorphism was presumably the basis of the description of the here synonymized C. voeltzkowi. Anjouan is the type locality for this species name and the syntype specimens represent the morphological form lacking the propodeal spines. Analysis of DNA sequence data from both the nuclear markers (see Figure 1) and ancillary mitochondrial data however clearly shows a lack of genetic divergence between the 'armed' and 'unarmed' forms in C. rasoherinae. The cause and maintenance of this intraspecific polymorphism remains to be investigated.
The syntype specimens of C. rasoherinae have been lost or destroyed during the times of World War II. Confirmation for this has been obtained via e-mail communication with the Naturhistorisches Museum Hamburg (F. Wieland, 19.vii.2011). I designate a neotype in this study to unequivocally ascertain the identity of the species C. rasoherinae, hereby selecting a worker specimen from or close to the original type locality in Madagascar, Tamatave [Toamasina, town]. In a large and taxonomically difficult genus such as Crematogaster type material is indispensable to clarify species identities. Although no closely resembling species is currently known, it is likely that a morphologically similar species could be discovered in the future, either in Madagascar or on the African mainland.
Pronotum laterally subangular; mesonotum laterally with distinct, raised carinae that are confluent with lateral carinae bordering metanotal groove and propodeum; in lateral view outline of promesonotum moderately convex; mesonotum transversely concave, without a distinct posterior face and gradually sloping into metanotal groove; metanotal groove in dorsal view constricted by bordering lateral carinae, propodeal spines short to medium-sized (SPI 0.10-0.26), form variable; dorsal face of propodeum very short; petiole in dorsal view ovo-rectangular, with dorsolateral margins increasingly carinate posteriorly, ending in small posterolateral denticles; subpetiolar process variable: from small, but distinct and acute tooth to reduced angular dent; postpetiole globular, faintly impressed posteriorly, no trace of median impression; subpostpetiolar process present or absent.
Queen diagnosis of the Crematogaster madecassa-group: C. madecassa, C. razana (C. telolafy unknown). Very small (HW 0.80-1.10, WL 1.28-1.74). With worker characters, except as follows. Masticatory margin of mandibles with 5 teeth; antennal scapes not, or just reaching posterior margin of head; eyes large (OI 0.29-0.37) and protruding, situated slightly above midline of head in full face view; head wider than long (CI 1.11-1.21) and widest just posterior to eyes, posterior margin of head straight.
Mesosoma more compact (MSNI 1.55-1.82, WL 1.28-1.74); mesoscutum in dorsal view almost or as wide as long; dorsal face of propodeum absent, and posterior face very sharply and almost vertically sloping; propodeal spines present, much shorter than in workers (SPI 0.02-0.14), sometimes reduced to minute dents; petiole and postpetiole as in workers.

Crematogaster madecassa
Diagnosis. Workers of C. madecassa can be distinguished from all other species treated here except C. telolafy by the presence of two distinct vertical carinae on the clypeus (Figure 16). From workers of C. telolafy, C. madecassa workers are distinguishable by their longer, more spiniform propodeal spines ( Figure 18A) and longer antennal scapes. Queens of C. madecassa can be easily identified from queens of most species (C. rasoherinae, C. volamena, C. mpanjono) by the presence of propodeal spines. C. madecassa queens can be distinguished from C. razana queens by the absence of a median clypeal notch (present in C. razana queens). Note however that queens of C. telolafy are currently unknown and could be very similar morphologically to C. madecassa queens.
Worker description ( Figure 26A-C). Very small species (HW 0.48-0.60, WL 0.44-0.63), with characters of the C. madecassa-group, in addition to the following. Clypeus with two distinct median vertical carinae; antennal scapes well surpassing posterior margin of head.
Metanotal groove constricted to less than half as wide as pronotal width; propodeal spines medium-sized (SPI 0.17-0.26) and straight, usually thin and acute, in lateral view directed upwards, in dorsal view moderately diverging; subpostpetiolar process usually present, often as acute minute tooth.
Promesonotum usually with 6 erect, long flexuous setae: 2 humeral setae, and 2 setae each at anterior and posterior end of mesonotal carinae. Color pale to medium yellow.
Intermediate worker description (Figures 27E-G). Intermediate between workers and queens in size. Head, petiole and postpetiole characters similar to queens; ocelli present, but smaller than in queens; mesonotum is to various extent raised and fused dorsally over pronotum and has wing attachment sutures; otherwise mesosomal characters more similar to worker characters.
Male unknown. Distribution and biology. Crematogaster madecassa is fairly widespread throughout the rainforests and littoral forests of northern, eastern and southeastern Madagascar ( Figure 27D) and is found up to medium elevations (highest record from 1175m). It occurs widely in sympatry with C. rasoherinae, and at a few localities also with C. volamena, C. mpanjono and C. telolafy. There is evidence that C. madecassa nests both on the ground and arboreally, as collections have been made from rotten logs, as well as from dead twigs above the ground. As in C. rasoherinae (see above), intermediate workers are found in C. madecassa ( Figure 27E-F).   -19.14194, 44.82800, 50m; -19.13222, 44.81467, 100m; -18.70944, 44.71817, 150m  Diagnosis. Workers of Crematogaster telolafy can be differentiated from all other Malagasy Orthocrema species except C. madecassa by the presence of two distinct vertical carinae on the clypeus (Figure 16). From C. madecassa, which shares this feature, C. telolafy workers are distinguished by the form (triangular) and the shorter length of their propodeal spines (Figures 18B and C), and by the shorter antennal scape. Queens of C. telolafy are unknown, but are expected to be morphologically similar to C. madecassa queens.
Worker description ( Figure 29A,C-D). Very small species (HW 0.50-0.58, WL 0.52-0.69), with characters of the C. madecassa-group, in addition to the following. Clypeus with two distinct median vertical carinae; antennal scapes reaching, or barely surpassing posterior margin of head.
Metanotal groove constricted to less than half as wide as pronotal width; propodeal spines short to medium-sized (SPI 0.10-0.19), usually in form of acute triangular points, sometimes more elongate and spiniform, distinctly directed upwards in lateral view, in dorsal view parallel or moderately diverging; subpostpetiolar process usually present, often as acute minute tooth.
Promesonotum usually with 6 erect, long flexuous setae: 2 humeral setae, and 2 setae each at anterior and posterior end of mesonotal carinae. Color pale to medium yellow.

Queen, male and intermediate (if existing) unknown.
Distribution and biology. No colony collections of C. telolafy have been made that could give clues on the biology of this species. It has been collected by methods of litter sifting, malaise, pitfall and pan traps, as well as beating low vegetation. Crematogaster telolafy is distributed mainly in the dry and spiny forests of southern and western Madagascar at low elevations ( Figure 29B), with some notable exceptions of records from remnant patches of western rainforest, e.g. Rés. Ambohijanahary, or gallery forest such as exists in the Isalo region. This species is allopatric with the closely related C. madecassa (see above), except for a narrow sympatry or parapatry in the Andohahela region, where one collection of C. telolafy has been made in low elevation rainforest. Otherwise C. telolafy occurs in sympatry only with C. rasoherinae among the species treated here.
Etymology. This species is named for the triangular form of its propodeal spines, as "telolafy" is the Malagasy word for triangle. This name should be treated as a noun in apposition.
Worker measurements ( Metanotal groove constricted to about half the width of pronotum; propodeal spines short (SPI 0.11-0.14), in form of acute triangular points, distinctly directed upwards in lateral view, in dorsal view moderately diverging; subpostpetiolar process absent.
Queen measurements (n=1 Queen description (Figures 30E-G). Very small (HW 1.10, WL 1.74), with characters of the C. madecassa-group, in addition to the following. Clypeus with a median notch; antennal scapes just reaching posterior margin of head.
Petiole with one pair of long flexuous setae posterior to denticles.

Male and intermediate (if existing) unknown.
Distribution and biology. Crematogaster razana has only been collected at two localities in southern Madagascar ( Figure 30B). One of these sites is a dry forest (west- ern slopes of P.N. Andohahela), whereas the other collections were made in a montane grassland habitat (R.S. Kalambatritra). Since the few collections of this rare species have been made mostly on the ground (under stone, pitfall trap or on low vegetation), I assume that C. razana is ground-nesting. This species is sympatric with C. rasoherinae and parapatric or narrowly sympatric with both C. madecassa and C. telolafy in the Andohahela massif.
Etymology. The Malagasy word "razana" means ancestor and alludes to the isolated distribution of this species and its phylogenetic position within the C. madecassagroup. This name should be treated as a noun in apposition.
Very small to medium sized species (HW 0.51-0.98, WL 0.56-0.92). Masticatory margin of mandibles with 4 teeth; clypeus with or without several irregular vertical carinae; posterior margin of head in full face view laterally subangular, often medially slightly depressed; occipital carinae indistinct; antennal scapes just barely (small workers) or not reaching (larger workers) posterior margin of head; midline of eyes situated well above midline of head in full face view; eyes small (0.18-0.22) and fairly flush with head.
Pronotum laterally subangular; in lateral view, anterior part of mesonotum often angular or denticulate, posteriorly at least weakly carinate until meeting metanotal groove; in lateral view outline of promesonotum fairly flat; dorsal face of mesonotum flat, posterior face distinct or indistinct; metanotal groove very constricted by bordering lateral carinae, a third as wide as pronotal width; propodeal spines short (SPI 0.06-0.12), upwards directed sharp points; length of dorsal face of propodeum about a third of posterior face; petiole in dorsal view ovo-rectangular, dorsolateral margins angulate, ending in small posterolateral denticles; subpetiolar process variable, from well pronounced acute tooth to reduced angular dent; postpetiole short and broad, appearing bilobed, with diffuse, broad median impression; subpostpetiolar process absent.
Sculpture overall reduced; head shiny to aciculate; mesosoma dorsally mostly shiny, meso-and metapleuron mostly shiny, rugulose in some parts; dorsal and posterior face of propodeum shiny with some carinulae; dorsal face of petiole shiny to carinulate; helcium dorsally reticulate; postpetiole dorsally shiny; lateral and ventral face of petiole and postpetiole feebly reticulate; face with very abundant silken erect to suberect pilosity of variable length, usually hereof 6-12 longer setae; promesonotum with at least 6 erect, long flexuous setae: 2 humeral setae, 2 setae at anterior end of mesonotum and 2 setae on mesonotal tubercles or denticles; additional long erect setae, and scattered shorter erect setae may be present dorsally on promesonotum; longer, erect pilosity present or absent from propodeum; petiole with a single flexuous setae on each posterolateral tubercle; postpetiole with a pair of long flexuous dorsoposterior setae, and several shorter setae dorsally and laterally; abdominal tergites and sternites 4-7 with dense erect pilosity (> 50 setae) of medium length, interspersed with a subdecumbent shorter pubescence. Color, pale or golden yellow, or medium brown.
Queen diagnosis of the Crematogaster volamena-group: C. volamena, C. mpanjono. Large (HW 1.48-1.72, WL 2.61-2.70). With worker characters, except as follows. Masticatory margin of mandibles with 5 teeth. Antennal scapes not surpassing posterior margin of head, reaching to about level of median or lateral ocelli; occipital carinae well pronounced or indistinct; eyes medium-sized to large (OI 0.23-0.27), situated at midline of head in full face view; head wider than long or slightly longer than wide, posterior margin of head straight.
Mesosoma compact (MSNI 1.93, WL 1.44); mesoscutum in dorsal view as wide as long; scutellum with only one long and flat dorsal face, in dorsal view oval-shaped and posteriorly rounded, dorsoposterior margin not carinate; dorsal face of propodeum about as long as posterior face; propodeal spines absent; petiole in dorsal view oval, carinae or denticles absent and all margins rounded, in lateral view petiole anteriorly tapering; anteroventral subpetiolar tooth absent; postpetiole fairly globular, median impression absent; wings clear.
Distribution and biology. Crematogaster volamena is currently known from about 10 localities along the eastern rainforest belt in Madagascar ( Figure 32F). Here the species occurs in low-to mid-elevation rainforest or littoral forest. Crematogaster volamena appears to be a generalist in terms of nesting preferences, as it has been collected both nesting arboreally in dead twigs and on the ground in rotten logs. Scant natural history information exists, but noteworthy is a size variation in workers of this species that is reminiscent of a major-minor distinction in other ant genera. The larger workers ( Figures 31D-F) have a distinctly enlarged head and more powerful mandibles compared to regular workers ( Figures 31A-C), and they are more rarely seen throughout collections. These individuals do not share the attributes of the intermediates described here for C. rasoherinae and C. madecassa, such as presence of ocelli, or a queen-like modified mesosoma.
Etymology. This species is named for the golden yellow coloration that most of its workers possess, as "volamena" means "gold" in Malagasy. This name should be treated as a noun in apposition.
Crematogaster mpanjono sp. n. urn:lsid:zoobank.org:act:7A4E5257-126C-460D-B9A4-7C52CB2D62F1 http://species-id.net/wiki/Crematogaster_mpanjono Figure 33 Type locality. MADAGASCAR: Antsiranana: R.S. Manongarivo: -13.96167, 48.43333, 400m, rainforest, beating low vegetation, 18.xi.1998, B.L.Fisher. Type specimen. holotype worker: pinned, CASENT0193889, BLF01998 (12) Diagnosis. Workers of C. mpanjono are diagnosed from all Malagasy Orthocrema, except the closely related C. volamena, by a combination of the following: indistinct occipital carinae (Figure 14), non-protruding and small eyes and the absence of raised, sharp lateral carinae on the propodeum ( Figure 15). All these characteristics are shared with the much more common C. volamena, from which it cannot be distinguished reliably based on the worker caste. Crematogaster mpanjono has slightly shorter propodeal spines than C. volamena, and often a long erect pilosity is present on the propodeum. The distributions of these two species (compare Figures 32F and 33D) aid in distinguishing between them, as they are not known to co-occur. Queens of C. mpanjono are easily separated from C. volamena queens by virtue of the scuto-scutellar suture that is acutely meeting the mesoscutum ( Figure 23, compare with Figure 21) and the indistinct occipital carinae. From the remaining Malagasy Orthocrema queens, C. mpanjono queens are diagnosable by the absence of propodeal spines and by their larger size (HW 1.48, WL 2.70).
Worker description (Figures 33A-C). Very small to small species (HW 0.51-0.78, WL 0.56-0.86), with characters of the C. volamena-group, in addition to the following. Clypeus with or without several short vertical carinae.
Head sculpture shiny to aciculate; propodeum often with longer erect pilosity. Color yellow to pale yellow.  Queen description (Figures 33E-G). Medium-sized (HW 1.48, WL 2.70), with characters of the C. volamena-group, in addition to the following. Antennal scapes reaching to about level of lateral ocelli; occipital carinae indistinct; eyes larger (OI 0.27); head slightly longer than wider (CI 0.96).
Head sculpture mostly aciculate, carinulate below eyes. Color reddish brown. Male unknown. Distribution and biology. Only one nest collection of this species (a queen and a nanitic worker) from an arboreal root pocket exists. Therefore next to nothing is known about the biology of C. mpanjono. The known distribution records of this rare ant show a macrohabitat preference for lowland rainforests of the north-western Sambirano region (Nosy Bé, R.S. Manongarivo) or eastern littoral rainforest (F Ambohidena) ( Figure 33D). These disjunct records from north-western Madagascar and the east coast island Ile St.Marie are peculiar and could point to incomplete distribution records for this species. In any case, although the Ambohidena population has not been sampled for nuclear genetic data, the conspecificity of these disjunct populations is supported by the COI barcoding data. Crematogaster mpanjono occurs in sympatry with the widespread C. rasoherinae and C. madecassa.
Etymology. The Malagasy word "mpanjono" means "fisher "or "fisherman". The name for this rare ant species is dedicated to B. L. Fisher and his ant diversity and conservation efforts in Madagascar. This name should be treated as a noun in apposition.

Discussion
The species diversity of Malagasy Crematogaster (Orthocrema) has been tripled in the context of this revisionary work, adding four new species to two already described and here well supported species. Another result that was strongly supported by the molecular part of this study is the presence of three phylogenetically distinct lineages of Malagasy Orthocrema (Figure 1). The relationships between the C. madecassa-and C. volamena-group remain weakly supported, similarly to results of a larger analysis (Blaimer, in prep.). The molecular data, as well as distinctive morphological differences indicate that these two species-groups do not constitute a monophyletic group. One can therefore assume they reached Madagascar through independent colonization events.
In analogy to previous taxonomic studies on the genus in Madagascar (Blaimer 2010;in prep.), a widespread species, Crematogaster rasoherinae, has been found to have synonymic names. Widespread species in Crematogaster are prone to this "over-naming" as they can show strong gradual geographic variation. Discrete polymorphisms, as in the case of the entirely unarmed form of C. rasoherinae on the Comoros Islands, are presumably much rarer, but another case of strong polymorphism has just been described for queens of Crematogaster ranavalonae Forel in Madagascar (Blaimer 2012). In either species the causes and selective forces maintaining these different morphotypes remain to be investigated.
Similar to previous work on Malagasy Crematogaster (Blaimer 2010), the new species described here have very restricted distributions and suggest adaptation to narrow environmental niches. Madagascar is well known for its highly endemic species assemblages (Goodman and Benstead 2005), but, similar to other biodiversity hotspots in the world, habitat destruction is posing a constant threat especially to these locally endemic taxa. For instance, one of the three localities at which the newly described C. mpanjono occurs is a currently unprotected parcel of littoral forest, Ambohidena on Ile de St. Marie (see Goodman 1993). Littoral forests represent one of the most threatened vegetation types in Madagascar (Consiglio et al. 2006) and forests in Madagascar that are not under protection are inexorably shrinking further (Brooks et al. 2002;Allnutt et al. 2008). Although one new endemic ant species will not be sufficient to motivate protection of forests such as Ambohidena, adding and tallying these numbers across a breadth of taxa, as was achieved recently by Kremen et al. (2008), will aid tremendously in ongoing efforts to define priorities for the expansion of the protected area network in Madagascar. In particular arthropod taxonomists are lagging behind in describing Madagascar's incredibly diverse fauna. The present study offers a small contribution to fill this gap.