A distinctive new frog species ( Anura , Mantellidae ) supports the biogeographic linkage of two montane rainforest massifs in northern Madagascar

We describe a new species of the genus Gephyromantis, subgenus Vatomantis (Mantellidae, Mantellinae), from moderately high elevation (1164–1394 m a.s.l.) on the Marojejy, Sorata, and Andravory Massifs in northern Madagascar. The new species, Gephyromantis (Vatomantis) lomorina sp. n. is highly distinct from all other species, and was immediately recognisable as an undescribed taxon upon its discovery. It is characterised by a granular, mottled black and green skin, reddish eyes, paired subgular vocal sacs of partly white colour, bulbous femoral glands present only in males and consisting of three large granules, white ventral spotting, and a unique, amplitude-modulated advertisement call consisting of a series of 24–29 rapid, quiet notes at a dominant frequency of 5124–5512 Hz. Genetically the species is also strongly distinct from its congeners, with uncorrected pairwise distances ≥10 % in a fragment of the mitochondrial 16S rRNA gene to all other nominal Gephyromantis species. A molecular phylogeny based on 16S sequences places it in a clade with species of the subgenera Laurentomantis and Vatomantis, and we assign it to the latter subgenus based on its morphological resemblance to members of Vatomantis. We discuss the biogeography of reptiles and amphibians across the massifs of northern Madagascar, the evidence for a strong link between Marojejy and Sorata, and the role of elevation in determining community sharing across this landscape.


Introduction
In recent decades, the number of frog species that have been discovered in Madagascar, while steadily increasing (Köhler et al. 2005), often included species that were not immediately recognizable as new to science, though with occasional exceptions, e.g.Boophis lichenoides (Vallan et al. 1998), Scaphiophryne boribory (Vences et al. 2003), and Tsingymantis antitra (Glaw et al. 2006).The majority of newly discovered taxa are assignable to existing complexes and must be investigated closely before it can

Specimen collection and morphological measurement
Specimens were collected at night using head torches along montane streams, euthanized using MS222 anaesthesia and subsequent overdose, fixed in 96 % ethanol, and deposited in 75 % ethanol for long-term storage.Tissue samples were stored in 96 % ethanol.Field numbers refer to the zoological collections of Miguel Vences (ZCMV), Frank Glaw (FGZC), and Steven Megson (SM).Specimens were deposited in the amphibian collections of the Mention Zoologie et Biodiversité Animale, Université d'Antananarivo (UADBA-A) and the Zoologische Staatssammlung München (ZSM).
Morphological measurements were taken to the nearest 0.1 mm using a digital calliper.Measurement schemes followed generally previous work on the genus (e.g.Vences et al. 2017) with modifications to decrease the risk of damaging the fragile limbs of the specimens when ascertaining limb lengths: snout-vent length (SVL), maximum head width (HW), head length from posterior edge of tympanum to snout tip (HL), horizontal eye diameter (ED), horizontal tympanum diameter (TD), distance from eye to nostril (END), distance from nostril to snout tip (NSD), distance between nostrils (NND), upper arm length from the articulation of the arm with the trunk to the elbow (UAL), lower arm length from the elbow to the base of the hand (LAL), hand length from the base of the hand to the tip of the longest finger (HAL), forelimb length (FORL*, given by the sum of UAL, LAL, and HAL), forearm length (FARL, given by the sum of LAL and HAL), thigh length from cloaca to knee (THIL), tibia length from knee to heel (TIBL), tarsus length from heel to base of foot (TARL), foot length from base of foot to tip of longest toe (FOL), hindlimb length (HIL*, given by the sum of THIL, TIBL, TARL and FOL), and length and width of femoral gland (FGL, FGW).Asterisks in this list indicate measurements that have the same abbreviation as the analogous single-measurement of previous studies (e.g.Vences et al. 2017) but are cumulative here and therefore not necessarily equivalent; comparison of such values must be done cautiously.

Sequencing and analysis of DNA sequences
DNA was extracted from tissue samples using a Qiagen DNeasy blood & tissue kit (Qiagen, Hilden, Germany), or standard salt extraction protocols.For two samples from Sorata and one sample from Marojejy (ZCMV 15269), we amplified a fragment of the mitochondrial 16S rRNA gene (hereafter 16S) in 25 µl polymerase chain reactions with the primers 16Sra-L and 16Sb-H (Palumbi et al. 1991), 1 µl of template DNA, and the following steps: initial denaturation for 3 min at 94 °C, followed by denaturation with 35 cycles of 30 sec each at 94 °C, 30 sec of annealing at 55°C and 60 sec of elongation at 72 °C, and a final elongation step of 10 min at 72 °C.Sequencing was conducted using the BigDye Terminator v1.1 Cycle Sequencing Kit on ABI 3730 and ABI 3130xl capillary sequencers.Newly determined sequences were deposited in GenBank (accession numbers MG926811-MG926823).For an additional nine specimens from Marojejy, we sequenced a shorter, highly variable stretch of 250 bp of the same 16S region by an Illumina amplicon approach (Vences et al. 2016) to confirm their identification (data not shown).
For an exploratory analysis, we aligned the new sequences with 16S sequences used by Kaffenberger et al. (2012) for all nominal species of Gephyromantis.Because the obtained tree (not shown) confirmed the new species to be related to the Laurentomantis/Vatomantis clade as also strongly suggested by morphology, we focused our analysis on this subgroup, i.e., all nominal species of the subgenera Laurentomantis and Vatomantis, and G. klemmeri which is known to be related to these subgenera (Kaffenberger et al. 2012), as well as G. granulatus (subgenus Duboimantis) as outgroup.
We aligned sequences in MEGA 7 (Kumar et al. 2016), yielding an alignment of 532 positions of the sequenced stretch of the 16S rRNA gene.As only a few indels were found in this alignment, we did not exclude any positions for further analysis.We used the Bayesian Information Criterion in jModelTest 2.1.4(Darriba et al. 2012) to determine a SYM+G substitution model as best-fitting our data.We implemented this model in MrBayes 3.2 (Ronquist et al. 2012) and computed a Bayesian inference phylogenetic analysis, with two independent runs of 20 million generations, each comprising four Markov Chains (three heated and one cold), sampling every 1000 generations.Chain mixing and stationarity were assessed by examining the standard deviation of split frequencies and by plotting the -lnL per generation using Tracer 1.5 software (Rambaut and Drummond 2007).Results were combined to obtain a 50 %-majority rule consensus tree and the respective posterior probabilities of nodes, after discarding 25 % of the generations as burn-in (all compatible nodes with probabilities <0.5 kept).In addition, we computed a Maximum Likelihood (ML) tree in MEGA 7, with a GTR+G model (as the SYM model is not available in this program), SPR level 5 branch swapping, and 500 nonparametric bootstrap replicates.Genetic distances (uncorrected pairwise p-distances) were also calculated in MEGA 7.

Bioacoustic analyses
Recordings from Marojejy were made on a Marantz PMD661 MKII with a Sennheiser ME66/K6 supercardioid microphone, at a bandwidth of 44.1 kHz.Recordings from Sorata were made on an Edirol R-09 with its internal microphone.Call analysis was conducted in Cooledit 2.0 (Syntrillium Corp.).To obtain frequency information, the recording was transformed with Fast Fourier Transformation (FFT; width 1024 points).Spectrograms were created with a Hanning window of 512 or 256 bands.Measurements are given as mean ± one standard deviation, with range in parentheses.Terminology follows the recently-published recommendations of Köhler et al. (2017) with a note-centred approach.This definition is different from that of Vences et al. (2002) for Laurentomantis and Sabino-Pinto et al. (2014) for Vatomantis; the 'pulses' of those studies are here treated as notes, because each of these units in the new species described herein are distinctly pulsed, and therefore are treated as individual notes following Köhler et al. (2017).Recordings are deposited in the Animal Sound Archive of the Museum für Naturkunde, Berlin (DOI: 10.7479/nmx8-aq7v), and are available as Suppl.materials 1-2.

Taxonomic work
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 act 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:8A83DE58-A2EE-494F-A03C-820DC836CDDF.The online version of this work will be archived and made available from the following digital repositories: CLOCKSS and Zenodo.

Results
Based on 16S sequences, the newly collected specimens represent an undescribed and hitherto unknown species of Gephyromantis that is highly distinct from all others (≥10 % p-distance).Exploratory phylogenetic analyses including all species of Gephyromantis clearly suggested their relationships with the subgenera Laurentomantis and Vatomantis, which also is strongly supported by morphological affinities, in particular by the greenish dorsal colour, granular skin, riparian habits, and paired subgular vocal sacs of partly white colour in males (see Diagnosis below for more details).A phylogenetic analysis of 16S sequences (total alignment length 532 bp) for all described species of Laurentomantis and Vatomantis as well as G. klemmeri, which was related to these subgenera in the multi-gene analysis of Kaffenberger et al. (2012), places the newly collected specimens sister to a clade with all described species of Vatomantis.Gephyromantis klemmeri is placed sister to Laurentomantis, although these basal nodes did not receive relevant support from ML bootstrap values or Bayesian posterior probabilities (Fig. 1).Genetic distances of the new specimens to all other species were high: 10.9-15.4 % to the three described species of Vatomantis, 10.0-13.2% to species of Laurentomantis, and 12.2-12.5% to G. klemmeri.The newly collected specimens from Sorata and Marojejy differed by 2.9 %, while no sequence differences were detected within each of these two localities, except for two mutations observed in one Marojejy specimen (ZCMV 15219).
Phenotypically the new specimens bear resemblance to both Laurentomantis and Vatomantis.Their advertisement call is more similar to Laurentomantis, but their morphological resemblance to Vatomantis is greater (see the diagnosis below).We here tentatively assign them to Vatomantis due to their morphological affinities and preliminary phylogenetic relationships.Given their very high genetic divergence to all other Gephyromantis, isolated phylogenetic position (not placed as close sister group to any other species), and morphological and bioacoustic differences, there is no doubt that these specimens belong to a new species, which we describe below.

Diagnosis.
A species assigned to the genus Gephyromantis on the basis of its granular skin, moderately en- with 2-4 pulses per note, an inter-note interval of 41-75 ms, and a dominant frequency of 5124-5555 Hz.DNA sequence data from the 16S gene fragment supports the high divergence of this taxon to all other Gephyromantis, and is in agreement with its subgeneric assignment, albeit without statistical support (Fig. 1).
Within the genus Gephyromantis, G. lomorina sp.n. can be distinguished from all subgenera except Laurentomantis and Vatomantis on the basis of the combination of femoral glands composed of few large granules (vs.composed of many, small granules; note that G. klemmeri is here treated separately from all other subgenera, below, due to its unclear assignment), SVL < 26 mm (vs.> 27 mm in all other subgenera except Gephyromantis), absence of a white stripe along the upper lip (vs.general presence in subgenus Gephyromantis), and absence of distinctly enlarged supraocular spines (vs.presence in Asperomantis and some Duboimantis).It may be distinguished from all members of the subgenus Laurentomantis (G.ventrimaculatus (Angel), G. malagasius (Methuen & Hewitt), G. striatus (Vences, Glaw, Andreone, Jesu & Schimmenti), G. horridus (Boettger), and G. ranjomavo Glaw & Vences) by paired subgular vocal sacs (vs.single), absence of outer metatarsal tubercles (vs.presence), and at least partly greenish dorsal skin (vs.mostly yellowish to brown to reddish), and from several of these by the absence of tibial glands in males (vs.typical presence).Within the subgenus Vatomantis, G. lomorina sp.n. may be distinguished from all species by its more granular dorsal skin (vs.granular but not rough) and venter spotted with white (vs.generally without whitish spotting except on the chin and over the sternum); from G. rivicola (Vences, Glaw & Andreone) and G. webbi (Grandison) by its reddish iris colouration (vs.copper and greenish, respectively); from G. silvanus (Vences, Glaw & Andreone) by its smaller size (SVL 20.5-25.5 mm vs. 31 mm) and partly whitish vocal sacs (vs.yellowish); from G. webbi by femoral glands composed of few large granules (vs.composed of many, small granules) and large inner metatarsal tubercle (vs.small).Gephyromantis lomorina sp.n. may be distinguished from G. klemmeri by its roughly granular dorsal skin (vs.smooth to shagreened), greenish skin colour (vs.brownish), reddish iris (vs.gold), and strongly protruding inner metatarsal tubercle (vs.small and not protruding).
The call of G. lomorina sp.n. may be distinguished from all Vatomantis and Laurentomantis species in having notes that are clearly pulsed (vs.unpulsed notes in all species except G. ventrimaculatus); Gephyromantis ventrimaculatus has a higher number of pulses per note notes than G. lomorina sp.n. (ca.6 pulses per note vs. 2-4 in G. lomorina sp.n.).The call of G. lomorina sp.n. is somewhat similar to that of G. klemmeri, especially in having pulsed notes, but the call duration is much longer (1681-1827 ms vs. 626-982 ms), the call has a more distinct amplitude decay (vs.complex amplitude modulation, see Vences et al. 1997), the notes of the call are more homogeneous (vs.distinct components of the call), and it lacks frequency modulation (vs.frequency modulated toward the end of the call).Colouration in life (Fig. 2) dorsally mottled with greens, browns, blacks, and yellows.Particularly green over the eyes.Raised ridges on the back were mostly yellowish, but some also with an orange hint.Flanks and lateral head as dorsum.Legs dark brown with yellow-green cross-bands, three on the thigh, three on the shank, and two on the tarsus.The tarsus and dorsal foot were a more ruddy brown than the rest of the body, mottled with a tan orange on the toes and on the heel.A few tubercles on the legs were red.A whitish annulus was present before the terminal disc of each toe and finger.The forelimbs were as the shanks and foot, ruddy brown mottled with yellow-green and dark brown, with a few red tubercles.Whitish spots were present in the inguinal region and the ventral portion of the flank, and also two cream stripes were present below the eye that continued on the bottom lip.The tympanum was distinctly brownish.The venter was umber in base colour with more reddish portions of translucent skin on the ventral side of the arms.The chin had white portions along the lip and especially on the vocal sacs, but the jaw itself was blackish.The venter had distinct white spots.The ventral hindlimbs were umber with irregular pale olive and yellow patches on the ventral thigh and shank.The ventral tarsus, foot, and hand were umber.The femoral glands were fleshy in colour, and the area ventral to the cloaca was pinkish.The iris was copper above and below, and rusty anteriorly and posteriorly, with blackish reticulations and a blackish line above and below the centre of the pupil.

Description of the
After six months in preservative, the colouration of the holotype has faded to become more uniformly brownish, and areas that were greenish in life have become cream.White areas of the venter are still immaculately white.
Variation.All paratypes resemble the holotype in gross morphology; see Table 1 for morphological variation.Tympanum diameter ranges from 47-79 % of eye, without strong sexual dimorphism in tympanum size.Females are marginally but not significantly larger than males (t-test, t = -1.9215,df = 13, p = 0.07687).Several paratypes have smaller femoral glands than the holotype.Femoral glands are composed of 2 or 3 large granules (mean 2.875 ± 0.35, n = 8; all but one of eight examined specimens with 3 granules).Females have miniscule raised bumps in the femoral area.There is considerable variation in colouration of the specimens, with some individuals being much darker, and others being more green (Figs 3-4).The chin of females is more solidly dark than that of males, and they lack most white spots.A pair of cream stripes below the eye that continue on the lower lip is present in all specimens.Two specimens (UADBA-A 60299, and ZSM 1545/2012, Fig. 4) have a bright vertebral stripe.
Bioacoustics.Call recordings were made in Marojejy from the holotype ZSM 419/2016 at its collection locality at a distance of 0.5 m during light rain (Suppl.material 1, DOI: 10.7479/nmx8-aq7v).The call is interpreted as an advertisement call as it resembles the advertisement calls of the subgenus Laurentomantis, and was emitted without close proximity to other individuals, and while the frog was otherwise inactive (Köhler et al. 2017).Air temperature was not recorded.A strict FFT bandwidth filter was applied to the dataset to remove all sound below 400 Hz in order to remove wind artefacts.Two calls were recorded from the holotype, but numerous calls were heard whilst searching for this species along the river where it was found.Calls consisted of a rapid series of 24-29 extremely short notes (note duration 6.3 ± 1.9 ms, range 2-10 ms, n = 53; Fig. 5a), each of which had 2.6 ± 0.6 pulses (2-4 pulses, n = 50), the peak amplitudes of which were separated by 2.7 ± 0.6 ms (1-4 ms, n = 53; Fig. 5b).Notes were separated by silent inter-note intervals of 64.6 ± 5.5 ms (47-75 ms, n = 51).The call was amplitude modulated, increasing in amplitude quickly and slowly decaying toward the end of the call.Call duration was 1769-1827 ms (n = 2), with one inter-call interval recorded of 2399 ms.Generally, however, the calls appeared to be emitted rather irregularly.Dominant frequency was 5124-5512 Hz, and the 90 % bandwidth was from 2723-2759 to 6391-6462 Hz.
Similar calls were recorded in Sorata from ZSM 1549/2012 at its collection locality (Suppl.material 2, DOI: 10.7479/nmx8-aq7v).Air temperature was not recorded.The calls strongly resembled those recorded from the holotype.Three calls were recorded, but one was cut off and another had loud calls of Gephyromantis (Duboimantis) sp. in the background, so only one was analysed.The call consisted of a rapid series of 31 extremely short notes (note duration 6.9 ± 0.8 ms, range 6-10 ms, n = 27 analysed), each of which had 2.0 ± 0.2 pulses (2-3 pulses, n = 27), the peak amplitudes of which were separated by 3.0 ± 0.4 ms (2-4 ms, n = 27).Notes were separated by silent inter-note intervals of 46.3 ± 3.8 ms (41-55 ms, n = 27).The call was amplitude modulated in the same way as that of ZSM 419/2016.Call duration was 1681 ms, and one inter-call interval was ca.1900 ms.In general however calling was irregular.The dominant frequency was 5555 Hz, and the 90 % bandwidth was from 4979 to 6003 Hz.The call with a loud Gephyromantis (Duboimantis) sp. in the background was considerably shorter, and consisted of just 11 notes over a duration of 515 ms, but we suppose this call may have been disturbed as it lacked amplitude reduction toward its end.
Natural history.Specimens were collected near mountain streams in pristine montane riparian rainforest (Fig. 4g).In Marojejy National Park they were encountered during and after light rain, sitting in inconspicuous locations, especially on the fronds of tree ferns, but also on other low vegetation, between a few centimetres and up to 2 m above the ground.Specimens in Sorata were found in similar positions during dry weather, in the days just before the beginning of the rainy season.Males called irregularly and softly (see the call description above).Population density in Marojejy was remarkably high, with around three or four individuals being found along a 10 m stretch of stream.The observed density in Sorata was lower, possibly due to the absence of rain during the observation period.The species occurred in close sympatry with a number of other mantellids, but only few of these (especially Mantidactylus aff.femoralis) were found in the same microhabitat.Several specimens from Marojejy had pinkish mites (probably of the genus Endotrombicula; see Wohltmann et al. 2007) embedded within translucent whitish pustules on the skin of their fingers, toes, and bodies.Nothing is known about the reproduction of this species, but the calling sites suggest an association with lotic water.
Available names.There are no other, earlier names currently available (e.g., junior synonyms) that are assignable to the subgenera Vatomantis or Laurentomantis and that could apply to the new species.
Etymology.The specific epithet is the Malagasy word lomorina, meaning 'covered in moss', in reference to the green, mossy appearance of the species in life.It is used as an invariable noun in apposition to the genus name.
Conservation.The species occurs in two regions with very different conservation situations: the highly protected forests of Marojejy National Park, and the unprotected, isolated, and highly threatened forests of Sorata and Andravory.Maminirina et al. (2008) report a study site in the rainforest of Sorata at 970 m a.s.l., but in our surveys in 2012, we detected larger patches of forest only at elevations of ca.1270 m and above.The new species was collected at lower elevation in Andravory (1164-1179 m a.s.l.), where forest persists.Higher elevation levels of Sorata are covered by high-elevation forests different to those where G. lomorina sp.n. was found, and these therefore may not support this species.In this area, the species is therefore directly threatened by the loss of the only forests in which it has been detected.
By contrast in Marojejy, forest extends down to roughly 200 m a.s.l., is highly protected, and the high elevation forest where this species occurs does not seem to be facing any immediate threats.Although the tourist load to Marojejy is relatively high, and the area upslope from the collection locality of the holotype and several paratypes is somewhat polluted with refuse from the nearby tourist camp, the species was abundant around this stream during our survey there in 2016, and presumably inhabits other streams around the same elevation across the massif.
Accommodating this spread of risk is a challenge for the IUCN Red List status.However, G. (V.) lomorina sp.n. is not the first species to have almost exactly this distribution.Rhombophryne vaventy Scherz, Ruthensteiner, Vences & Glaw was recently recovered from Sorata (Peloso et al. 2016, Scherz et al. 2016, Lambert et al. 2017) after initially having been described from the same type locality as G. lomorina sp.n. (Scherz et al. 2014).In the case of this species, Scherz et al. (2017a) argued for a classification of Endangered under IUCN criterion B1ab(iii), i.e. an extent of occurrence under 5000 km 2 (B1), known from fewer than five threat-defined locations (a), and an observed, estimated, inferred, or projected decline (b) in the area, extent, and/ or quality of habitat (iii).Given the similar situation in G. lomorina sp.n., i.e., very similar, limited distribution and ongoing reduction and threat to a substantial part of its habitat (i.e., the forests of Sorata and Andravory), we propose that the same threat status and justification be given for this species.

Discussion
Gephyromantis (Vatomantis) lomorina sp.n. is a distinctive species, mostly due to its granular, greenish skin, which is rougher than in all other members of the subgenus Vatomantis, but not as rugose as in many species of the subgenus Laurentomantis.Indeed, it is in several aspects intermediate between these subgenera, having a call that sounds similar to both (Vences et al. 2006).Its phylogenetic position is at present basically unresolved between these two subgenera.However, its morphology is clearly more similar to Vatomantis than to Laurentomantis, as it lacks an outer metatarsal tubercle (present in Laurentomantis), has a distinct brown tympanum (less distinct in Laurentomantis), lacks a broadened head (usually distinctly broadened in Laurentomantis) and has paired subgular vocal sacs (single in Laurentomantis) (Glaw and Vences 2006).
Gephyromantis (Vatomantis) rivicola, G. (V.) silvanus, G. (V.) lomorina sp.n., and most Laurentomantis species share a unique femoral gland morphology with glands being composed of a small number of large, round granules (each granule representing a single gland within the femoral macrogland; Vences et al. 2007; Fig. 1).Glaw et al. (2000) interpreted these unusual glands as possible intermediate steps between Type 2 glands (sharply delimited groups of numerous granules of up to 0.9 mm diameter) toward Type 3 and 4 glands (a rounded structure composed of few, large granules and an external central depression).The position of G. (V.) lomorina sp.n. appears to make this situation more complicated; formerly, it seemed that granule size had increased and number decreased in G. (V.) rivicola and G. (V.) silvanus while G. (V.) webbi had retained Type 2 glands typical of most other Gephyromantis species (Glaw et al. 2000, Vences et al. 2007).However, given the split of G. (V.) lomorina from a more basal node in that clade (Fig. 1), and given the ubiquity of these unusual glands in the sister subgenus Laurentomantis (Glaw et al. 2000, Kaffenberger et al. 2012), it seems that Type 2 femoral glands may have independently originated one or more times in this clade.A better resolved phylogeny of the clade will be necessary to better understand the evolution of their femoral gland morphology.
The apparently highly divergent G. (V.) lomorina sp.n. sheds some light on questions regarding the relationships of G. klemmeri.Formerly, G. klemmeri was considered a member of the subgenus Gephyromantis, but Kaffenberger et al. (2012) showed that it has affinities between Laurentomantis and Vatomantis.They forestalled action on transferring it to one of these subgenera until more data become available, as single genes disagreed as to its position.Gephyromantis klemmeri shares femoral gland morphology with both Laurentomantis and Vatomantis, having large glands with a small number of large granules.This lends credence both to its phylogenetic position being close to these subgenera, and also to the hypothesis that smaller numbers of larger granules in the femoral glands may be ancestral in this clade.Kaffenberger et al. (2012) suggested three possible alternatives to dealing with the phylogenetic affinities of G. klemmeri: (a) including G. klemmeri in Laurentomantis (its position sister to Laurentomantis was supported with 94 % bootstrap support from maximum likelihood and >0.99 posterior probability, but was not supported in maximum parsimony analysis), (b) erecting a new monotypic subgenus, or (c) redefining a more inclusive subge-nus Laurentomantis that besides G. klemmeri would also include Vatomantis as a junior synonym (the clade containing Laurentomantis, Vatomantis, and G. klemmeri was supported with 100 % bootstrap support from maximum likelihood, >0.99 posterior probability, and 86 % bootstrap support from maximum parsimony).Determining the best course of taxonomic action will in part depend on the resolution of the phylogenetic relationships of G. klemmeri and of G. (V.) lomorina sp.n., in the framework of a more comprehensive revision of Laurentomantis and Vatomantis, as these subgenera still contain further candidate species requiring in-depth analysis (Vieites et al. 2009).
Gephyromantis (Vatomantis) lomorina sp.n. also sheds light on the biogeography of northern Madagascar, providing yet more evidence for a strong link between Sorata and Marojejy.The environmental conditions of these two regions are similar (Brown et al. 2016), and various species originally described from one of the two areas have subsequently been discovered in the other, e.g.Rhombophryne vaventy (Peloso et al. 2016, Scherz et al. 2016, 2017a, Lambert et al. 2017), Gephyromantis (Asperomantis) tahotra (Glaw et al. 2011, Vences et al. 2017), and G. (D.) schilfi (Glaw andVences 2000, Scherz et al. 2017b).These similarities are generally limited to species found above 1200 m, probably because forest below 1200 m in Sorata has been mostly eradicated.
We predict that similarities between faunal compositions of the mountainous massifs of northern Madagascar are limited by elevational connectivity.For instance, there is continued connectivity between regions of elevation up to 1400 m from Sorata to Marojejy and indeed roughly to the Manongarivo massif as well.There is no connectivity above this elevation however; areas of over 1400 m across the different massifs are separated by lower elevations, leading to island-like isolation of peak areas.Therefore, we predict that species occurring above 1400 m will show a greater degree of microendemism, and those below this elevation will have a greater probability of occurring more widely; the higher a species' centre of elevational distribution is located, the greater its chance of being microendemic.No absolute threshold of turnover is expected, because major climate fluctuations in the past will likely have blurred elevational boundaries over time.
So far, evidence appears to support this hypothesis; as already stated, several species from around 1300 m are shared between Marojejy and Sorata (and Andravory, though at present only limited and generally unpublished data are available from this forest), and some species known from higher elevations are so far thought to be microendemic to either region, e.g.Rhombophryne longicrus (Scherz et al. 2015), Gephyromantis (Duboimantis) tohatra (Scherz et al. 2017b), Calumma jejy, and C. peyrierasi.Assuming this hypothesis is correct, it raises questions about species that are microendemic at lower elevations, but opportunities to study and understand these taxa are increasingly limited by the fact that forest at lower elevations is disappearing outside of protected areas.Conservation efforts must be redoubled to ensure that these study systems may remain long enough to be investigated and understood.

Figure 1 .
Figure 1.Preliminary phylogenetic tree of Gephyromantis (Vatomantis) lomorina sp.n., based on Maximum Likelihood analysis of a 532 bp fragment of the mitochondrial 16S rRNA gene.Numbers at nodes indicate bootstrap values in percent (500 replicates, above) and posterior probabilities from a Bayesian Inference analysis (20 million generations, below), shown only if >50 % (bootstrap) or >90 % (posterior probabilities).Each specimen/species is followed by the corresponding GenBank accession number used in the alignment.Schematic drawings of femoral glands of all species in the subgenera Laurentomantis and Vatomantis as well as of G. klemmeri are shown to the right of the phylogeny, and coloured according to the subgenus to which they are assigned.

Figure 5 .
Figure 5. Spectrogram (above) and waveform (below) of a call of the holotype of Gephyromantis (Vatomantis) lomorina sp.n., ZSM 419/2016, from Marojejy.(a) A full call (spectrogram shown using FFT of 512 points to visualise call structure); and (b) a 250 ms section from the middle of a call, showing the degree of pulsation of each note (spectrogram shown using FFT of 128 points to visualise note structure).

Figure 6 .
Figure 6.Distribution of Gephyromantis (Vatomantis) lomorina sp.n. in northern Madagascar.Areas with diagonal lines are official protected areas.The dotted outline indicates the proposed area with the scope of the WWF protection plan for this part of Madagascar (Biodev Madagascar Consulting 2014, WWF Madagascar 2015).Three arc second SRTM data from Jarvis et al. (2008).

Table 1 .
holotype.A specimen in a good state of preservation, a piece of tissue taken from the left thigh.SVL 23.3 mm; for other body measurements see Table1.Body slender.Widest part of head marginally wider than widest part of body.Snout rounded in dorsal and lateral view, protruding slightly over upper jaw in lateral view.Nostrils not distinctly protruding, with lateral openings.Canthus rostralis distinct, concave.Loreal re-Morphological data on specimens of Gephyromantis lomorina sp.n.Abbreviations: m = male, f = female, sa = subadult; for measurement abbreviations, see the Materials and methods.The holotype is bolded.Additive measurements(FARL, FORL, and HIL)are not explicitly shown but can be deduced from these data.