Taxonomic revision of Madagascan Rhantus (Coleoptera, Dytiscidae, Colymbetinae) with an emphasis on Manjakatompo as a conservation priority

Abstract We review the diving-beetle genus Rhantus Dejean of Madagascar (Coleoptera, Dytiscidae, Colymbetinae) based on museum collection holdings and recently collected expedition material. Both morphology and DNA is used to test species boundaries, in particular whether newly collected material from the Tsaratanana mountains in the north represent a new species or are conspecific with Rhantus manjakatompo Pederzani and Rocchi 2009, described based on a single male specimen from the central Ankaratra mountains. DNA of the holotype of R. manjakatompo was successfully extracted in a non-destructive way and sequenced. The general mixed Yule coalescent model applied to an ultrametric tree constructed from mitochondrial cytochrome c oxidase subunit I (COI) sequence data delimited three species. Morphological characters supported the same species unambiguously. We therefore recognise three species of Rhantus to occur in Madagascar: R. latus (Fairmaire, 1869), R. bouvieri Régimbart, 1900 and R. manjakatompo Pederzani and Rocchi, 2009. All three species are endemic to Madagascar and restricted to the highlands of the island. Rhantusstenonychus Régimbart, 1895, syn. n., is considered a junior synonym of R. latus. We designate lectotypes for R. bouvieri and R. goudoti Sharp, 1882, the latter a junior synonym of R. latus. We provide descriptions, a determination key, SEM-images of fine pronotal and elytral structures, distribution maps, habitus photos, and illustrations of male genitalia and pro- and mesotarsal claws. We discuss the role of the Manjakatompo forest as a refugium for Madagascan Rhantus diversity and other endemics of the montane central high plateau.


Introduction
Rhantus Dejean is a large cosmopolitan genus of medium-sized aquatic diving beetles. Several studies have pointed out that the genus as presently defined is clearly paraphyletic and will likely be split in the future (Balke et al. 2007). Rhantus has a track record of colonizing oceanic islands and forming island endemics (Balke et al. 2007(Balke et al. , 2009, which indicates capacity for infrequent long-distance dispersal. It has colonized several islands in the Pacific with island-endemics on Fiji (Balke et al. 2007), New Caledonia (Balke et al. 2010), Galapagos (Peck and Balke 1993) and Hawaii (Balke 1993) to name a few. The Afrotropical region harbours only a modest part of the global Rhantus diversity (Guignot 1961). Balke (1992a) revised the Rhantus species of the Mascarene Islands outside of Madagascar and concluded that previous records were erroneous and that three species were endemic for this area, each to one of the three islands Mauritius, La Reunion and Ile Rodrigues. The Mascarene Islands are of rather recent volcanic origin (8-15 mya) and are known to have been on the receiving end of flora and fauna from Madagascar. Madagascar itself is a large island with ecosystems spanning from rainforests to semi deserts. The level of endemism is extremely high which has granted Madagascar a top placement among biodiversity hotspots (Myers et al. 2000). But the fauna is also known for its microendemic patterns -species distributions are often restricted to smaller geographical areas, almost like "islands within the island". Some colonizations have led to rich allopatric species radiations, but the seemingly low species diversity of Rhantus perhaps indicates that their relatively high dispersal capacity has prevented the genus from a significant radiation. The first Rhantus species described from Madagascar was Rhantus latus (Fairmaire 1869, in the genus Colymbetes Clairville). Since then four additional species have been described from the island, but the validity of some of these are questionable (Balke 1992b, Pederzani andRocchi 2009) and no modern revision exist. In addition, apart from R. latus, Madagascan Rhantus are very poorly represented in collections and are seemingly rare. Recent expeditions by the Swedish Museum of Natural History in collaboration with the Entomology department of the University of Antananarivo, have unearthed significant new material of Madagascan Rhantus, especially from Manjakatompo forestry station in the Ankaratra mountain massif. The new material both enables the assessment of intraspecific character variation and the extraction and sequencing of DNA data to test species limits. The purpose of this study is to revise the Madagascan Rhantus species based on this material, type material and other museum holdings, notably from the collections at the Museé National d'Historie Naturelle in Paris. In particular, the discovery in 2004 of what seemed to be a new undescribed Rhantus species from the Tsaratanana mountain massif in the north of Madagascar warranted a revisonary treat-ment of the group. In 2009 however, Pederzani and Rocchi (2009)

described a new
Rhantus species from the Ankaratra mountains in central Madagascar based on a single male specimen. The description showed both clear similarities but also some differences to the new species discovered in 2004 and it was hypothesized that DNA data and explicit species delimitation tests could aid in resolving this taxonomic question. Results of all these studies are summarized in the presented paper.

Material and methods
All specimens examined in this study are registered in the NHRS collection objects database (interface via www.naturarv.se) but are deposited in the following collections and referred to by the abbreviations: Measurements were taken on specimens in a horizontal position. The following measurements were taken: ML, maximum length from head to tip of elytra; MW, maximum width; LP, length of pronotum medially; WPB, pronotal width at base; LE, length of elytra from tip of scutellum to apex. The measurements were taken using an Olympus SZX12 stereomicroscope with an Infinity X camera and a calibrated ruler tool in the software DeltaPix Insight 2.0. Environmental scanning electron microscopy was done using a FEI/Philips XL30 FEG ESEM at the Institute for Surface Chemistry, Stockholm, Sweden. The images were generated at 350× magnification, with a gaseous secondary electron detector in low vacuum mode; the accelerating voltage of the electron beam was 17 kV. In opposite to a backscattered detector, a gaseous secondary electron detector depicts depressions of the surface brightly.

Molecular data
DNA was extracted from legs of ethanol-preserved material collected after 2004 using Qiagen DNeasy 96 Tissue kit (see Table 1 for specimen information). For the dry-mounted holotype of R. manjakatompo, (collected 2001) the QIAamp® DNA Micro kit was used, following the protocol for animal tissue with incubation at 56°C overnight. A single hindleg was carefully removed and after incubation re-glued to the body. Two fragments of the gene cytochrome c oxidase subunit I (COI) were sequenced for analysis. Primers used for amplification and sequencing were derived from several sources ( Table 2). DNA fragments were PCR amplified using "Ready-to-go" PCR Beads from Pharmacia Biotech and Phire Hot Start II PCR Master mix following the manufacture's standard protocols. The thermal cycling profile for "Ready-to-go" PCR was 95°C for 5 min, followed by 40 cycles of 95°C for 30 s, 50°C for 30 s, 72°C for 1 min and finally 72°C for 8 min. PCR cycling for Phire Hot Start PCR was 98°C for initial denaturation 30 s, followed by 40 cycles of 98°C for 5 s, 53°C for 5 s, 72°C for 15 s, 72°C for 1 min. Product yield, specificity of amplification and contamination were investigated using agarose gel electrophoresis. PCR products were purified using ExoFAP Cleanup mix and cycle sequenced using the same primers used to amplify. For sequencing reactions the ABI BigDye Terminator kit ver. 3.1 was used. Sequencing products were purified using the DyeEx 96 kit and fragments were analysed on an ABI377xl analyser from Applied Biosystems. Gene regions were sequenced in both directions. The contigs were assembled from the forward and reverse reads and primers were trimmed in Sequencher 5.0 (Gene Codes Corporation 2011). All sequences are deposited in Genbank under the accession numbers KF548613 -KF548663 (Table 1).

Analyses
Sequence data for 26 specimens were aligned in ClustalX 2.1 (Larkin et al. 2007) using default settings. The resulting combined matrix based on the two gene fragments was gap-free and had a total length of 1486 base pairs. An ultrametric tree was obtained from Bayesian MCMC analysis conducted in MrBayes 3.2 (Ronquist et al. 2012). We used a GTR+I+Γ model jointly for all codon positions but allowed each position a separate rate multiplier. Branch lengths were estimated under an autocorrelated relaxed clock model (TK02). Two separate MCMC analyses were run for 200000 generations with chains sampled every 100 generations, and a burn-in of 10% was chosen after visual examination of the -lnL plotted against generations and the convergence statistics provided by the program. A species delimitation test was conducted using the General Mixed Yule-Coalescent (GMYC) approach (Pons et al. 2006) and the splits package (Ezard et al. 2009), implemented in R 2.14.0 (R Development Core Team 2011). The single threshold method was used on the tree to find the maximum likelihood solution of the transition point between coalescent and interspecific branching patterns on the tree. Genetic distances were calculated under the Kimura-2-parameter model using Mega 5.0 (Tamura et al. 2011).

GMYC species delimitation
The GMYC model was significantly better than a single coalescence model with a likelihood ratio test (lgL GMYC: 158.9, lgL single coalescence: 149.4, p = 0.0003), and divided the terminals into tree separately coalescing units: R. manjakatompo, R. latus and R. bouvieri ( Figure 1). No other solution had a log-likelihood value within 2 units (an approximate confidence interval) of the maximum likelihood solution. The holotype of R. manjakatompo (NHRS-JLKB000030412) from Ankaratra mountains was nested within the single coalescing unit represented by the four specimens from the Tsaratanana mountains ( Figure 1). The genetic distance between the holotype from Ankaratra and the specimens from Tsaratanana was 0.008-0.011, which was within the range among the Tsratanana specimens alone (0-0.012). The within-species genetic variation was 0.015 for R. bouvieri and 0-0.019 (mean = 0.009) for R. latus. The distances between the three species were almost an order of magnitude greater: 0.11-0.13 between R. bouvieri and R. latus, 0.17 between R. bouvieri and R. manjakatompo and 0.16-0.17 between R. latus and R. manjakatompo.   Description. Size: ML 11.5-12.5 mm; MW 5.7-5.9 mm; Lp 1.4 mm; Epb 4.7-4.9 mm; Le 8.4-9.0 mm (n = 5).

Rhantus manjakatompo Pederzani & Rocchi, 2009
Head. Testaceous with black areas posteriorly and inside eyes. Interocular black pattern narrowly separated medially (Figure 2a-b). Dense reticulation of well impressed meshes, very fine punctuation within and at edges of meshes.
Pronotum. Testaceous to ferrugineous with two small dark spots medially ( Figure  2b), which may be absent (Figure 2a). Rim at lateral margin clearly visible to indistinct. Dense reticulation of well impressed meshes, very fine punctuation within and at edges of meshes (Figure 3a).
Elytra. Testaceous with black irrorations, leaving paler sides and yellow sutural lines (Figure 2a-b). Black irrorations somewhat confluent subapically. Reticulation of polygonal meshes simple anteriorly and double at middle and posteriorly, meshes are somewhat less impressed than on pronotum and larger posteriorly. Very fine punctuation within and at edges of meshes (Figure 3b-c).
Remarks. In 2004 David Lees and Ravomiarana Ranaivosolo collected material of what seemed to be a new Rhantus species from the mountain massif of Tsaratanana, North Madagascar. R. manjakatompo was described in 2009 based on a single male specimen from Ankaratra mountains, 70km south of Antananarivo in central Madagascar. Despite variation in color and impression of a lateral rim of pronotum, molecular data indicate that the studied material from Ankaratra and Tsaratanana are conspecific or at least not an old enough divergence to be recognised based on COI sequence (genetic distance 0.008-0.011). As also male tarsal claws, aedeagus and parameres were identical, we consider the specimens from Tsaratanana mountains conspecific with the holotype of R. manjakatompo.
Elytra. Testaceous with black irrorations, leaving paler sides and yellow sutural lines (Figure 2c). Somewhat confluent to form two small black areas subapically. Microreticulation double but somewhat indistinct. Very fine punctuation within and at edges of meshes (Figure 3e-f).
Ventral side. Dark brown to black, abdominal segments with testaceous spots along the lateral edges. Epiplura and legs, including metafemur, testaceous (Figure 4b).
Male. Protarsal claws equally long, medially straight in lateral view (Figure 5cd). Mesotarsal claws curved in lateral view, equally thin, the anterior claw somewhat longer than the posterior (Figure 6c-d). Penis in dorsal view evenly tapering towards apex (Figure 7d). In lateral view not evenly curved but almost angulate at base ( Figure  7c). Parameres with inner margin undulating (Figure 7h).
Remarks. In the original description, Régimbart refers to specimens from Fianarantsoa collected by Grandidier, but the holotype is not designated. The three specimens labelled "Fianarantsoa Grandidier" at MNHN are therefore considered to be syntypes and a male with genitalia extracted is designated here as the lectotype. Lectotype and paralectotypes were labelled as such.
Habitat. Known from a river in Andringitra and streams and grassy ponds in Manjakatompo, in both cases at altitudes between 1420 to 2070 m a.s.l.
Head. Testaceous with black areas inside and posterior of eyes. Interocular black pattern often rather broadly separated medially (Figure 2d). Dense micropunctuation but no reticulation.
Pronotum. Somewhat infuscated and with one wide dark spot which is not medially divided (Figure 2d). No rim at lateral margin. Micropunctation and vague microreticulation (Figure 3g).
Elytra. Testaceous ferrugineous with black irrorations, leaving paler sides and yellow sutural lines (Figure 2d). Black irrorations regular throughout, even posteriorly with no sign of preapical black areas. Microreticulation double, meshes are well impressed posteriorly but vague anteriorly. Micropunctation within and at edges of meshes (Figure 3h-i).
Male. Protarsal claws long and slender, almost twice the length of last tarsal segment, somewhat sinuated and unequal, anterior claw longer (Figure 5e Remarks. The interpretation of Fairmaire's name Colymbetes latus is unambiguous following common usage (e.g. Régimbart 1895, Guignot 1961, even though typematerial with Coquerel as collector could not be found in Fairmaire's collection in Paris, or elsewhere. Neither Guignot (1961) nor Balke (1992b) were able to localize the holotype of R. stenonychus and the validity of the name, based on a single specimen, has therefore not been evaluated before but the name lingered in the literature. The details for the type in Régimbart's original description are as follows: "Madagascar: Lac Ambodinandohalo (R. P. Camboué), un seul male (coll. R. Oberthür)". In Régimbart's collection at MNHN there was a pin without a specimen but with the handwritten labels by Régimbart "stenonychus Rég." and "Madag. Coll. Oberthür". We interpret the pin with the label as a reference to the Oberthür collection for the unique type. In Oberthür's collection a single male specimen was found bearing the label "Lac Ambodo [or Ambod°], R.P. Camboue". We believe "Lac Ambodo [or Ambod°]," is an abbreviation for Lac Ambodinandohalo and that this is the holotype. We are not able to find any records of another lake named "Ambodo" in Madagascar. Lake Ambodinandohalo was a lake in the haute-ville in Antananarivo and the French Jesuit priest and missionary Paul Camboué lived just west of Antananarivo in Arivonimamo and Ambohibeloma. Peltodytes quadratus Régimbart, 1895 was described in the same article from the same locality (Lac Ambodinandohalo, collected by R. P. Camboué) and the type specimen in Paris bore a label with the same abbreviated locality name (see Vondel 2010;Vondel and Bergsten 2012).
The holotype is as judged by morphological characters conspecific with R. latus. Already in the original description similarities with R. latus are obvious (also see Guig-not 1961) and it is odd that Régimbart did not compare the species with R. latus in his original description. We therefore synonymize the name R. stenonychus with R. latus.
Habitat. Occurs in a quite wide range of habitats, like streams and rivers, muddy waterpools and grassy ponds. Of the three Rhantus species it is the only one that can be found below 1000 m altitude and known range include 530 to 2070 m.
Distribution. Possibly endemic to Madagascar as we have not been able to verify the records from mainland Africa by Régimbart (1895) "Cafrerie, Cap" and by Guignot (1961) "Sud d'Afrique". In Madagascar rather widespread over the central high plateau (Figure 8). Known from Andasibe-Mantadia NP, Antananarivo and Ambohijanahary NP, which are the northernmost records. Further known from several localities south of Antananarivo along RN7 from Manjakatompo forestry station to Andringitra NP, and further southwest to Isalo NP and Zombitse-Vohibasia NP. Seemingly lacking from the northern third of Madagascar.

Identification key on habitus of females and males
1a Smaller ( Larger (ML: 11.5-12.5 mm), Male mesotarsal claws equally long but posterior claw distinctly thicker than anterior claw (Figure 6a-b). Pronotum with two small dark spots (Figure 2b), or spots are absent (Figure 2a). Penis in lateral view short robust and evenly curved (Figure 7a). Parameres evenly curved and tapering to apex (Figure 7g) (Balke 2001), the genus is lacking from the lowland tropics in Madagascar. The central highland plateau of Madagascar (about 40% of the island) however, is almost completely degraded and very little of the original forests remains. Manjakatompo forestry station is one of three small forest relics remaining of the high plateau together with Ambohitantely Special Reserve and the Anjozorobe Forest. In November of 2011 and January 2012 we carried out fieldwork in Manjakatompo forestry station, investigating the aquatic beetle fauna. Although we have sampled aquatic beetles across Madagascar at hundreds of localities, nowhere else was the Rhantus fauna richer both in individuals and species than in Manjakatompo. All three endemic species of Madagascar existed here in healthy populations, two of which are only known from one other locality outside of Manjakatompo. Manjakatompo is clearly a small and fragile but important forest refugium which Rhantus and other highland fauna may depend on. Manjakatompo is located in the province of Antananarivo, region Vakinakaratra and district of Ambatolampy, at 17 km to the west of the city Ambatolampy (19°22'S, 47°16'E). It lies on the eastern slope of the Ankaratra mountain massif of Quaternary volcanic origin. The altitude ranges between 1550 and 2643 m with the highest peak, Tsiafajavona, being the third highest on Madagascar. Forests are humid and the climate follows a pattern of cold and dry austral winter and a warm and wet austral summer (annual rainfall around 2000mm) (Vences et al. 2002). The average temperature of the coldest month is 5-10°C, but can drop below zero at higher altitudes. The station covers an area of 8320 ha, with only 650 ha of natural forest and 2300 ha replanted with exotic trees (Goodman et al. 1996). Even the natural forest portion is composed of largely secondary forest mixed with exotic trees. Manjakatompo forestry station was established in 1923 to preserve the relic of primary forests that remained at the time (Andriampenitra 2007).
The forest relics have been kept partly thanks to legal protection more or less effectively exerted by the agents of the Forestry Station, with its status as Integral Reserve (Andriampenitra 2007). However, no part of the Ankaratra massif is part of the protected area network of reserves and national parks with a higher level of protection. According to our personal observation, and local information, the Manjakatompo forest is constantly exposed to serious incidents such as commercial operation, fire and slash and burn agriculture, so called "tavy". In 2008, about 70 ha of the forest surface was burnt (Martin 2008). Inventories of the herpetofauna in Manjakatompo also found a specialised montane fauna with some 10-15% being endemics of the Ankaratra massif (Vences et al. 2002). Vences et al. (2002) were concerned that changes like increased use of pesticide, increased cattle (zebu) grazing, or aquaculture and release of fish, could cause a serious threat to the montane fauna. All these three factors are also known causes by which the aquatic insect fauna can drastically change, leading to the replacement of endemics by widespread opportunists. Kremen et al. (2008) used distribution data of various organism groups to model and optimize where additional protected areas on the island would come best to use for conserving additional components of the endemic fauna and flora not already under protection. The Ankaratra massif was part of the proposed new areas. While the flora and fauna show similarities to the Andringitra massif, which we can confirm based on the Rhantus fauna, it also has unique components and characteristics (Vences et al. 2002, Goodman et al. 1996, and remains a highly prioritized area for increased protection.