An integrative taxonomic revision and redefinition of Gephyromantis (Laurentomantis) malagasius based on archival DNA analysis reveals four new mantellid frog species from Madagascar

Miguel Vences; Jörn Köhler; Angelica Crottini; Michael Hofreiter; Carl R. Hutter; Louis du Preez; Michaela Preick; Andolalao Rakotoarison; Loïs Rancilhac; Achille P. Raselimanana; Gonçalo M. Rosa; Mark D. Scherz; Frank Glaw

doi:10.3897/vz.72.e78830

Abstract

The subgenus Laurentomantis in the genus Gephyromantis contains some of the least known amphibian species of Madagascar. The six currently valid nominal species are rainforest frogs known from few individuals, hampering a full understanding of the species diversity of the clade. We assembled data on specimens collected during field surveys over the past 30 years and integrated analysis of mitochondrial and nuclear-encoded genes of 88 individuals, a comprehensive bioacoustic analysis, and morphological comparisons to delimit a minimum of nine species-level lineages in the subgenus. To clarify the identity of the species Gephyromantis malagasius, we applied a target-enrichment approach to a sample of the 110 year-old holotype of Microphryne malagasia Methuen and Hewitt, 1913 to assign this specimen to a lineage based on a mitochondrial DNA barcode. The holotype clustered unambiguously with specimens previously named G. ventrimaculatus. Consequently we propose to consider Trachymantis malagasia ventrimaculatus Angel, 1935 as a junior synonym of Gephyromantis malagasius. Due to this redefinition of G. malagasius, no scientific name is available for any of the four deep lineages of frogs previously subsumed under this name, all characterized by red color ventrally on the hindlimbs. These are here formally named as Gephyromantis fiharimpe sp. nov., G. matsilo sp. nov., G. oelkrugi sp. nov., and G. portonae sp. nov. The new species are distinguishable from each other by genetic divergences of >4% uncorrected pairwise distance in a fragment of the 16S rRNA marker and a combination of morphological and bioacoustic characters. Gephyromantis fiharimpe and G. matsilo occur, respectively, at mid-elevations and lower elevations along a wide stretch of Madagascar’s eastern rainforest band, while G. oelkrugi and G. portonae appear to be more range-restricted in parts of Madagascar’s North East and Northern Central East regions. Open taxonomic questions surround G. horridus, to which we here assign specimens from Montagne d’Ambre and the type locality Nosy Be; and G. ranjomavo, which contains genetically divergent populations from Marojejy, Tsaratanana, and Ampotsidy.

Keywords

Amphibia, Anura, archival DNA, Mantellidae, new species, phylogeography

Introduction

Among the hyperdiverse anuran fauna of Madagascar, with about 370 scientifically named species (Frost 2021; AmphibiaWeb 2021), many species are represented by only a few voucher specimens in collections. In some cases, this reflects the high proportion of microendemism in Madagascar’s fauna (Wilmé et al. 2006; Vences et al. 2009) where species may occupy small ranges, often in remote mountain massifs that are extremely difficult to access. In other cases, species may be relatively widespread, but are difficult to detect, as has been shown for some fossorial taxa among reptiles (Brown et al. 2016). They may be characterized by secretive habits or may occur at low densities across their range.

One such secretive group of Madagascar frogs is the subgenus Laurentomantis in the genus Gephyromantis. These frogs have long been so poorly known that the monograph of Blommers-Schlösser and Blanc (1991) contained no information on their color in life, and the first photograph of a living specimen was published by Blommers-Schlösser and Blanc (1993). The six recognized nominal species of this subgenus form a clade (Kaffenberger et al. 2012) known from across Madagascar’s eastern and northern rainforests, but the number of individuals known to science is very small for most species. For instance, Gephyromantis horridus (Boettger, 1880) was described based on a juvenile individual from the northern offshore island Nosy Be, but has never been collected there in recent times despite intensive surveys (e.g., Andreone et al. 2003); new collections include a few individuals from the Montagne d’Ambre and Tsaratanana massifs assigned to this species (Vences et al. 2002; Glaw and Vences 2007). Gephyromantis ventrimaculatus (Angel, 1935), originally described as Trachymantis malagasia ventrimaculatus from Isaka-Ivondro in extreme south-eastern Madagascar, is known from localities in the South East, Southern Central East and Northern Central East of Madagascar. It occurs in two of the most intensively studied rainforest sites in Madagascar, Ranomafana National Park and Analamazaotra-Mantadia National Park near Andasibe. Despite its conspicuous coloration with grey-bluish vermiculations on a black venter, only a few records of this species exist from Ranomafana and only a single specimen has ever been recorded from Andasibe (Vences et al. 2002; Glaw and Vences 2007; Randrianiaina et al. 2011; Strauß et al. 2013; Riemann et al. 2015). A further recently described species, G. ranjomavo Glaw and Vences, 2011, is known from the holotype collected in the Marojejy Massif in the North East and a second specimen from an unknown locality. Only G. malagasius (Methuen and Hewitt, 1913) and G. striatus (Vences, Glaw, Andreone, Jesu and Schimmenti, 2002) have been regularly reported from various sites in eastern or north-eastern Madagascar, respectively. However, as currently understood, G. malagasius (originally described as Microphryne malagasia Methuen and Hewitt, 1913), consists of several genetically deeply divergent lineages of uncertain taxonomic status (Vieites et al. 2009; Perl et al. 2014), and not all of them are commonly collected. Lastly, the recently described species G. marokoroko is known from only the type series collected at three nearby sites in the Northern Central East (Hutter et al. 2022).

Besides their apparent rareness, most species of the subgenus Laurentomantis are remarkable frogs by their very spiny and tubercular dorsal integument and have long been considered a separate genus (e.g., Blommers-Schlösser and Blanc 1991). These terrestrial frogs occur in the rainforest leaf litter, and calling males have been found on the ground or on moderately high perches in the vegetation, often at the bottom of slopes. The few known tadpoles in the subgenus (all of G. ventrimaculatus) were non-feeding (Randrianiaina et al. 2011) and have been collected in small rainforest streams, although it is uncertain if they were accidentally washed into the streams from putative terrestrial nests. Some Laurentomantis are characterized by tibial glands of unknown function found in males and females; others have a ventral pattern of conspicuous bluish vermiculations on dark or reddish color primarily on the ventral sides of their hindlimbs. A more comprehensive understanding of the ecology of these frogs and the biological functions of their morphological and chromatic characteristics is hampered by the scarcity of field observations and a lack of understanding of their systematics.

The present study provides a comprehensive molecular assessment of the frogs in the subgenus Laurentomantis, based on DNA sequences of all samples collected over the past 20 years. To clarify the taxonomic status of lineages currently included in G. malagasius, we applied targeted enrichment sequencing to genetically characterize the 110-year-old holotype of this species, which unexpectedly was found to be conspecific with G. ventrimaculatus; this surprising result indicates that all recently collected frogs assigned to G. malagasius instead belong to four species new to science. We herein formally name and characterize these new species, based on an integration of molecular, bioacoustic and morphological information.

Materials and Methods

This study is based on voucher specimens and call recordings collected during various field campaigns in Madagascar between 1994–2017, and tissue samples collected since the year 2000. Upon collection in the field, frogs were anesthetized and euthanized by immersion in tricaine methanesulfonate (MS222) or chlorobutanol solution. We removed tissue samples for molecular analysis and stored them separately in 1.5 ml vials with 95% ethanol. Vouchers were then fixed in 95% ethanol (or in 12.5% formalin), preserved in 70% ethanol, and deposited at the Museo Regionale di Scienze Naturali, Torino (MRSN); the Zoologisches Forschungsmuseum A. Koenig, Bonn (ZFMK); Zoological Museum Amsterdam (ZMA; collection now included in Naturalis, Leiden); Zoologische Staatssammlung München (ZSM); and the Université d’Antananarivo, Mention Zoologie et Biodiversité Animale (UADBA). Additional material was studied from the Transvaal Museum Pretoria (TMP), the Museum National d’Histoire Naturelle, Paris (MNHN), Naturhistorisches Museum Wien (NMW), Naturhistorisches Museum Bern (NMBE), and Senckenberg Museum Frankfurt (SMF), and reference is made to material hosted at the Biodiversity Institute and Natural History Museum of the University of Kansas (KU). FGZC, FGMV and ZCMV refer to field numbers of F. Glaw and M. Vences. FAZC and FN refer to field numbers of F. Andreone. RJS, DLR, PSG, CRH, APR, JCR, MSZC and ACZCV refer to field numbers of Jasmin E. Randrianirina, Dina Ramamonjisoa, Philip-Sebastian Gehring, Carl R. Hutter, Achille P. Raselimanana, Jana Riemann, Mark D. Scherz and Angelica Crottini, respectively. Geographic regions within Madagascar are named according to Boumans et al. (2007).

The holotype of Microphryne malagasia Methuen and Hewitt, 1913, TMP 10076 (original number ‘No. 1155’), was collected by Herschell-Chauvin in 1911 in a locality called Folohy (Methuen and Hewitt 1913b). The specimen is currently preserved in ca. 70% ethanol, but its precise preservation history is unknown.

Morphometric measurements were taken by MV with an accuracy of 0.1 millimeter with a manual caliper. The measurements and abbreviations used are: snout–vent length (SVL); maximum head width (HW); head length from tip of snout to posterior edge of mouth opening (HL); horizontal tympanum diameter (TD); horizontal eye diameter (ED); distance between anterior edge of eye and nostril (END); distance between nostril and tip of snout (NSD); distance between both nostrils (NND); forelimb length, from limb insertion to tip of longest finger (FORL); hand length, to the tip of the longest finger (HAL); hind limb length, from the cloaca to the tip of the longest toe (HIL); foot length (FOL); foot length including tarsus (FOTL); and tibia length (TIBL). We report webbing formula according to Blommers-Schlösser (1979) to ensure comparability with previous species descriptions of Malagasy frogs. Femoral gland terminology follows Glaw et al. (2000) and Vences et al. (2007).

We recorded vocalizations in the field using different types of tape recorders (Tensai RCR-3222, Sony WM-D6C) with external microphones (Sennheiser Me-80, Vivanco EM 238), digital recorders (Tascam DR05, Edirol R-09, Marantz PMD 661 MkII, Olympus LS-10, Zoom H5) with built-in microphones or external microphones attached (Sennheiser K6+ME-66, ME-67, MKH-8060), and in one case extracted sound files from a published audio CD (Rosa et al. 2011). Recordings were sampled or re-sampled at 22.05 kHz and 32-bit resolution and computer-analyzed using the software Cool Edit Pro 2.0. We obtained frequency information through Fast Fourier Transformation (FFT; width 1024 points) with Hanning window function. Spectrograms were drawn with Blackman window function at 256 bands resolution. In some cases sensitive filtering was applied to remove background sounds, applied only to frequencies outside the prevalent bandwidths of calls. Temporal measurements are summarized as range with mean ± standard deviation in parentheses and calls are described following note-centered scheme of Köhler et al. (2017). Calls have been deposited in the Macaulay Library (https://www.macaulaylibrary.org; reference numbers 274464–274474).

Two molecular dataset were assembled to examine the genetic variation and differentiation within the subgenus Laurentomantis:

(1) All available samples were DNA barcoded using a fragment of the mitochondrial 16S rRNA gene, which has previously been used as standard marker for Malagasy frogs (Vieites et al. 2009). DNA was salt-extracted and a fragment at the 3′ terminus of the 16S rRNA gene amplified using the primer pair: 16SFrogL1/16SFrogH1 (5′–CATAATCACTTGTTCTTTAAA–3′; 5′–GATCCAACATCGAGGTCG–3′) modified from Palumbi et al. (1991), and the following PCR protocol: initial denaturation for 90 s @ 94°C, followed by 36‒40 cycles of denaturation for 45 s @ 94°C, primer annealing for 45 s @ 50‒53°C and elongation for 90 s @ 72°C, followed by a final extension step for 5 min at 72°C.

(2) To understand the concordance between the variation in mitochondrial and nuclear-encoded genes, we amplified fragments of two nuclear-encoded genes: sacsin (SACS) and recombination-activating gene 1 (RAG1). For SACS, we applied the nested PCR approach of Shen et al. (2012) with primers SACSF2 (5′–AAYATHACNAAYGCNTGYTAYAA–3′) and SACSR2 (5′–GCRAARTGNCCRTTNACRTGRAA–3′) in the first round, and SACSNF2 (5′–TGYTAYAAYGAYTGYCCNTGGAT–3′) and SACSNR2 (5′–CKGTGRGGYTTYTTRTARTTRTG–3′) in the second round. PCR protocols for both rounds were identical, as suggested by Shen et al. (2012): 240 s @ 94°C, 45 × [45 s @ 94°C, 40 s @ 45°C, 120 s @ 72°C], 600 s @ 72°C. For RAG1, we used the primers Gephlut-RAG1-F1 (5′–ATGGAGAGCCAACCCCTATC–3′) and Gephlut-RAG1-R1 (5′–KCCAGACTCGTTTCCTTCRC–3′) (originally developed for a study of Gephyromantis luteus: Vences et al. 2021b) with the PCR protocol 120 s @ 94°C, 35 × [20 s @ 94°C, 50 s @ 53°C, 180 s @ 72°C], 600 s @ 72°C.

We purified PCR products with Exonuclease I and Shrimp Alkaline Phosphatase digestion, and the purified products along with sequencing primers were shipped to LGC Genomics (Berlin) for sequencing on automated capillary sequencing instruments. The 16S fragment was sequenced with the forward PCR primer only, SACS and RAG1 were sequenced with the PCR primers in both directions and the two strands combined. Chromatograms were checked for base-calling errors and edited with CodonCode Aligner 6.0.2 (Codon Code Corporation, Dedham, MA, USA) and newly determined sequences submitted to GenBank (accession numbers OM885271–OM885340 and OM897144–OM897211).

To obtain DNA sequences from the holotype of Gephyromantis malagasius, we applied targeted enrichment sequencing (Straube et al. 2021) following a ‘barcode fishing’ strategy that we have previously employed and described in previous studies with Malagasy frogs (Rancilhac et al. 2020; Scherz et al. 2020). This approach aims at obtaining sequences of the same 16S rRNA gene fragment that we also PCR-amplified and sequenced from the fresh samples. The strategy uses targeted enrichment with 5,962 baits of 70 nucleotides in length, after filtering based on melting temperature and collapsing 99% identical baits. The baits were designed by Arbor Biosciences using sequences from most Malagasy frog species, including various Gephyromantis species of the subgenus Laurentomantis. A thigh muscle tissue sample was extracted from the type of G. malagasius using DNA-free scissors and stored in 100% ethanol in a 1.5 ml tube filled in a lab naïve to Gephyromantis research. We then performed DNA extraction in a clean lab dedicated to museum specimen and ancient DNA analysis where no other sample of the subgenus Laurentomantis had been processed before. The sample was washed with a GuSCN-based Qiagen PE Buffer, DNA extracted following the protocol of Rohland et al. (2004), and purification following the protocol of Dabney et al. (2013). The GuSCN-based extraction buffer from Rohland et al. (2004) has been widely shown to be effective in releasing DNA from museum specimens, while the silica column-based purification from Dabney et al. (2013) is especially suitable for retaining short fragments. This combination was also found to perform best with wet-collection specimens in an extensive comparison of different approaches (Straube et al. 2021). Specifically, tissue samples were extracted with an Guanidine thiocyanate extraction buffer (5 M GuSCN, 50 mM Tris pH 8.0, 25mM NaCl, 20 mM EDTA, 1% Tween-20, 1% 2-mercaptoethanol) as described in Rohland at al. (2004). Samples were incubated for 18 hours rotating at room temperature. The supernatant was then added to 13 mL binding buffer (5 M guanidine hydrochloride, 40% Isopropanol, 0.05% Tween-20, 90 mM sodium acetate) as described in Dabney at al. (2013). DNA was purified using the MinElute silica spin columns (Qiagen).

For library preparation, we used a single-stranded (ss-DNA) approach optimized for ancient and archival DNA (Gansauge et al. 2013, 2017) with custom adapters from Gansauge et al. (2013), amplified with custom Illumina indexing primers described in Paijmans et al. (2017) after determining the optimal cycle number using qPCR (Gansauge et al. 2013; Basler et al. 2017).

The ss-DNA library of the G. malagasius holotype was then captured twice for the aforementioned target sequences using the Arbor Biosciences MyBaits kit (RNA-based in-solution sequence capture), with 14.5 μL of each indexed library in a 24 h reaction at a hybridisation temperature of 65°C, and following the MyBaits target enrichment protocol except reducing the bait volume to 2.75 μL and substituting the missing 2.75 μL in each reaction with nuclease-free water. After hybridization, the libraries were bound to streptavidin-coated magnetic beads, and the reactions washed and eluted according to the MyBaits kit protocol. We then performed PCR amplification in a reaction volume of 60 μL with the following PCR conditions: 120 s @ 95°C, then with an optimal cycle number determined using qPCR, 30 s@ 95°C, 45 s @ 60°C, 45 s @ 72°C, and final extension of 180 s @ 72°C. Amplifications were purified using a Min Elute PCR Purification Kit (Qiagen), with final elution in a total volume of 30 μL of 10 nM Tris-HCl, 0.05% TWEEN-20 solution (pH 8.0). This procedure was performed twice to increase target capture reactions success, as described in Li et al. (2015) and Paijmans et al. (2016). Qubit 2.0 and 2200 TapeStation (Aligent Technologies) assays were used to determine the final library concentration and length distribution. We sequenced the enriched library on an Illumina Next-Seq 500 sequencing platform using 500/550 High Output v2.5 (75 cycles SE, aimed at 3 million reads) with custom sequencing primers (Paijmans et al. 2017).

After quality-trimming and adapter removal, all reads (duplicates not removed to keep information of read frequency) were compared against reference sequences of various Laurentomantis species using a custom script described in Rancilhac et al. (2020), with a similarity threshold to the references of 90%, to reduce the data set for further analysis. The selected reads were then uploaded in CodonCode Aligner 6.0.2 (CodonCode Corp.) and three majority-based alignments performed to align reads to alternative reference sequences of Laurentomantis species (corresponding to G. “ventrimaculatus” sensu lato, G. “malagasius” lineage B, and G. striatus). For this purpose, we used the “Align to Reference” option building a majority rule consensus, discarding stretches with <50 reads, filling uncovered regions of the reference sequence with N, with a local alignment approach (minimum percent identity = 70%, word length = 8, match score = 1, mismatch penalty = –2, gap penalty = –2, additional first-gap penalty = –3, minimum overlap score = 40). The contigs resulting from the three individual alignment attempts were identical. The respective sequence was deposited in GenBank (accession number OM897120) and analyzed along with the set of 16S DNA sequences of fresh samples.

We aligned the sequences for each locus individually in MEGA7 (Kumar et al. 2016) with the Muscle alignment option. As the alignment was unambiguous and only required few single indels for 16S (especially in the outgroup), all sites were used for phylogenetic analysis. All alignments are available from Figshare, DOI: https://doi.org/10.6084/m9.figshare.19299884.

The 16S alignment was analyzed with a relatively simple (K2P) substitution model to avoid overparametrization for shallow branches, in a Maximum Likelihood analysis in MEGA 7 with NNI branch swapping, and 500 nonparametric bootstrap replicates to assess node support. We calculated uncorrected pairwise distances between 16S sequences using the program TaxI2, implemented in iTaxoTools (Vences et al. 2021a). For this purpose, we used a reduced set of 47 sequences that spanned a full length of 508 nucleotides, thus avoiding biases that could arise when excessively short sequences (consisting mainly of the hypervariable central stretch of the 16S fragment) are included (reduced alignment available from Figshare: DOI: https://doi.org/10.6084/m9.figshare.19299884).

The two nuclear-encoded genes (RAG1 and SACS) were analyzed separately from the mitochondrial gene and each other since our main interest was to understand concordance (or absence thereof) in the differentiation of unlinked genetic markers. We used a haplotype network visualization to graphically represent the relationship among alleles (haplotypes) of these genes. Haplotypes were estimated with the PHASE algorithm (Stephens et al. 2001) implemented in DnaSP (Version 5.10.3; Librado and Rozas 2009) with default parameters. The phased sequences were used to reconstruct Maximum Likelihood trees with the Jukes-Cantor substitution model in MEGA 7 (the simplest available model, to avoid overparametrization), and these were used as input for Haploviewer (written by G.B. Ewing; http://www.cibiv.at/~greg/haploviewer), a software that implements the methodological approach of Salzburger et al. (2011).

As in previous studies, we follow the general lineage concept (de Queiroz 1998, 2007) in combination with a relaxed biological species criterion, i.e., demanding reproductive isolation indicated by restricted gene flow among lineages (e.g., Speybroeck et al. 2020). Because reproductive barriers generated through time increase genealogical depth and agreement among unlinked loci (Avise and Wollenberg 1997), we use genealogical concordance (Avise and Ball 1990) between mitochondrial and nuclear loci, especially in populations occurring in sympatry or close geographical proximity, as an indicator for restricted gene flow. We then assigned species status to a lineage, along with concordance between genetic and morphological evidence (Padial et al. 2010).

Results

Molecular phylogenetics and genetic divergence

The phylogenetic analysis of the 16S sequences of 88 individuals of Laurentomantis (plus one sequence of Gephyromantis (Vatomantis) webbi used as outgroup; total alignment length 508 nucleotides) revealed nine major clades separated by uncorrected pairwise distances >4%, which we consider as candidate species-level lineages (Fig. 1). Some of these contain intra-lineage variation of up to 5%; these cases will be discussed below. All of the nine lineages are supported by bootstrap values of at least 90%, whereas nodes defining inter-lineage relationships received only in one case a support of 70%, and 63% or less for all other nodes. Our tree does not constitute a reliable hypothesis of relationships among the main lineages, but provides a reliable assignment of individuals into lineages.

Figure 1.

Maximum Likelihood phylogenetic tree inferred from an alignment of 508 nucleotides of the mitochondrial 16S rRNA gene, in 88 individuals of Gephyromantis belonging to the subgenus Laurentomantis. A species of the subgenus Vatomantis (G. webbi) was used as the outgroup. Numbers at nodes are bootstrap values in percent (500 replicates; not shown if <50%). Colors correspond to main species-level lineages as discussed in the text.

The archival DNA analysis of the holotype of Microphryne malagasia yielded 3,481,149 raw sequence reads, of which 11,118 reads (0.32%) were retained after the filtering pipeline. These were used for downstream analysis for the sample, which led to a consensus sequence built with 10,497 reads. In the phylogenetic analysis, this consensus sequence clustered among specimens of the lineage commonly considered G. ventrimaculatus (e.g., Vences et al. 2002; Glaw and Vences 2007; Vieites et al. 2009; Perl et al. 2014). This clustering is supported by a bootstrap value of 94%. This assignment of the malagasius type is supported by diagnostic SNPs in different parts of the 16S fragment (Supplementary Fig. S1); this confirms that the phylogenetic assignment of the consensus sequence is due to substitutions in various sequence fragments obtained separately by the capture approach.

The molecular evidence suggests that the name G. malagasius should take nomenclatural priority to refer to those frogs currently considered as G. ventrimaculatus. In order to assess whether the name Trachymantis malagasia ventrimaculatus Angel, 1935 represents a junior synonym of Microphryne malagasia Methuen and Hewitt, 1913, a comparison of the respective name-bearing types is necessary: (1) The lectotype of Trachymantis malagasia ventrimaculatus, examined previously by Vences et al. (2002), shares with other individuals previously referred to G. ventrimaculatus, a highly contrasted ventral pattern of light, more or less vermiculated spots and markings on dark background, which in this expression is unique within the subgenus Laurentomantis. Since this color pattern allows for an unambiguous definition of G. ventrimaculatus sensu lato, we did not attempt to sequence the ventrimaculatus lectotype. (2) The holotype of Microphryne malagasia lacks this typical contrasted ventral pattern; the color of this specimen was almost completely faded upon examination in 2021, and ventral vermiculation also was not mentioned in the original description (Methuen and Hewitt 1913b). It is however worth noting that the description also lacks a mention of reddish ventral coloration, the main pattern characterizing all those lineages commonly assigned in the past to Gephyromantis malagasius. We therefore suspect that already in 1913, two years after its collection in Folohy, the color of the specimen was largely faded. Hence, we here anticipate our taxonomic conclusions below, and redefine G. malagasius as the lineage containing frogs with the typical ventral pattern of ventrimaculatus; this implies G. ventrimaculatus becoming a subjective junior synonym of G. malagasius.

As a consequence, no scientific name is available for those lineages that include the frogs with red ventral color on limbs that were assigned to Gephyromantis malagasius by Blommers-Schlösser and Blanc (1993), Vences et al. (2002), Glaw and Vences (2007) and other authors. The newly defined G. malagasius is separated from the other eight Laurentomantis species-level lineages by uncorrected pairwise distances in the 16S rRNA gene of 6.4–9.1%. Intra-lineage distances reach 2.9% between specimens from Befotaka and Ranomafana.

The red-legged individuals traditionally assigned to G. malagasius are placed by our analysis into four deep mitochondrial lineages, in this section provisionally named A–D. Lineage A (bootstrap support BS=94%) occurs at mainly low-elevation sites in Madagascar’s Southern and Northern Central East, from Ambohitsara (close to Ranomafana National Park) to Betampona and Befanjana Forest. Lineage B (BS=98%) occupies roughly the same region but appears to occur at mid-elevations, ranging from Ranomafana National Park to the Andasibe region and Anjozorobe. Lineage C (BS=95%) is only known from low elevations at the north-eastern localities Masoala and Makira. And finally, lineage D (BS=96%) is known from the Northern Central East, specifically from Sahafina and Betampona. These four lineages form a clade in our tree, which however did not receive strong bootstrap support. The smallest 16S uncorrected pairwise distances of 4.0–5.7% are found between lineages A and B, whereas the other comparisons among these four lineages yielded distances of 4.3–8.1%. Intra-lineage distances are up to 3% in lineage A, 1.8% in lineage B, 1.4% in lineage C, and 3.9% in lineage D.

Two further lineages correspond to the nominal species G. marokoroko and G. striatus and are confirmed by the analysis, with 100% and 99% bootstrap support, respectively. For G. marokoroko only individuals from the type locality (Vohidrazana) were available, and intra-lineage divergences are therefore negligible (<1%).

The final two species-level lineages revealed by our 16S tree correspond to a complex of enigmatic and poorly known taxa from northern Madagascar. They comprise specimens from Manarikoba forest in the Tsaratanana Massif that previously were assigned to G. horridus (Vences et al. 2002); however, in our tree, the single sequence obtained from these specimens clusters with high support (90%) with the holotype of G. ranjomavo, and with a specimen from Ampotsidy that phenotypically is very similar to G. ranjomavo. The three sequences substantially differ from each other (pairwise 16S distances 3.0–4.9%) but we here subsume all three under G. ranjomavo preliminarily. The second lineage contains individuals from Montagne d’Ambre that phenotypically resemble G. horridus from its type locality Nosy Be, and that we assign to this species preliminarily. The G. horridus + G. ranjomavo clade is supported by a bootstrap value of 97%, and the two species are separated by 6.7–8.3% 16S divergence.

Analysis of the nuclear-encoded genes RAG1 (alignment length 475 nucleotides for 37 samples) and SACS (834 nucleotides for 30 samples) suggested genealogical concordance across unlinked markers in the differentiation of most lineages described in the previous paragraphs (no sequences of these two genes were available for G. morokoroko). The haplotype networks of both genes (Fig. 3) show a relatively limited amount of variation, with a maximum of 14 mutational steps in RAG1 and nine mutational steps in SACS, but without haplotype sharing between lineages except for lineages B and D which had intertwined haplotypes (without sharing) in RAG1, and had one shared haplotype in SACS.

Figure 2.

Maps showing locality records in Madagascar considered in this study, the majority of them based on molecular data. The left map shows records for G. marokoroko, G. horridus, G. malagasius, G. ranjamavo, and G. striatus in paler colors if they would benefit from further confirmation or taxonomic revision, as follows: pale yellow marks the type locality of G. ventrimaculatus (specimens not studied genetically) and records based on individuals phenotypically matching the types of ventrimaculatus (here considered to be a junior synonym of G. malagasius). Light green marks records assigned to G. striatus based on phenotype without genetic data. Light blue marks a record assigned to G. horridus tentatively due to the lack of genetic data from the type locality. Sky blue mark genetically divergent specimens here assigned to G. ranjomavo but in need of further study. The type locality of G. malagasius, Folohy, is here tentatively placed close to Toamasina, but its exact location is uncertain (north of Toamasina / east of Lake Alaotra, see species account).

Figure 3.

Haplotype networks based on sequences of the nuclear-encoded genes RAG1 (475 nucleotides; 74 phased sequences corresponding to 37 individuals) and SACS (834 nucleotides; 60 phased sequences corresponding to 30 individuals). Colors correspond to sequences of individuals assigned to lineages based on mitochondrial DNA (Fig. 1). Sequences were phased into haplotypes before analysis, and each individual is therefore represented with two sequences in each network.

Morphological differentiation

In this study we limit our morphological comparisons to those lineages for which we provide relevant novel taxonomic information compared to Vences et al. (2002) and Glaw and Vences (2011), namely G. horridus, G. ranjomavo, G. ventrimaculatus sensu lato, and lineages A–D. For additional information on the remaining taxa, see Vences et al. (2002).

The molecular data suggest a need to re-assess the identity of specimens from the Manarikoba Forest in the Tsaratanana Massif previously assigned to Gephyromantis horridus (Vences et al. 2002). Our tree suggests that the single specimen sequenced from Manarikoba is genetically close to an individual from Ampotsidy, and together they form the sister group of the holotype of G. ranjomavo. The Ampotsidy specimen, in life, shows a color pattern reminiscent of G. ranjomavo, with reddish brown dorsal color on hindlimbs and forelimbs (yellowish brown in the G. ranjomavo holotype), differing from the predominantly blackish and brown dorsum (Fig. 7). A further character shared by the male specimen from Ampotsidy with the known males from Manarikoba and the male type specimen of G. ranjomavo is the presence of a distinct tibial gland (Vences et al. 2002; Glaw and Vences 2011).

No photographs of the Manarikoba specimens in life were available, but we assign this population tentatively to G. ranjomavo based on the genetic similarity to the Ampotsidy individual and to the G. ranjomavo holotype. The two specimens of G. horridus from Nosy Be, from historical collections of the 19th century (Fig. 4), do not show a tendency of lighter color dorsally on the limbs, and instead have both a relatively distinct pattern of two dark brown / blackish markings on the dorsum, at the area between the forelimbs, and at the posterior dorsum. These markings are also visible in life in specimens from Montagne d’Ambre, where in the single known male individual (Fig. 5) they take the form of two distinct transverse bands. This supports assigning the Montagne d’Ambre population to G. horridus, pending new collections and molecular data from the type locality. The single known male from Montagne d’Ambre (Fig. 5) has no tibial glands, which allows a distinction from all males attributed to G. ranjomavo (see above). From Nosy Be, unfortunately, no males have been collected and this character therefore remains unverified for topotypical G. horridus.

Figure 4.

Preserved specimens of Gephyromantis horridus from the type locality Nosy Be. A, B Holotype SMF 7177 (juvenile) in dorsal and ventral views. C, D Specimen NMW 3643 (female) in dorsal and ventral views.

Figure 5.

Specimens assigned to Gephyromantis horridus from Montagne d’Ambre National Park in life. A, B Adult female photographed in March 1994 (ZFMK 57433) in dorsolateral and ventral views. C–E Adult male (ZSM 126/2018) in dorsolateral, dorsal and ventral views. F, G Adult female discovered under rotten wood in May 2014 (not collected) in dorsolateral and dorsal views. Note the absence of tibial glands in all specimens. Not to scale.

Figure 6.

Audiospectrogram and corresponding oscillogram of one advertisement call of Gephyromantis horridus (ZSM 126/2018) recorded on 11 December 2017 in Montagne d’Ambre National Park.

The synonymy of G. ventrimaculatus with G. malagasius as supported by the genetic data from the holotype of G. malagasius is surprising. However, examination of the holotype of malagasius (Fig. 14) revealed it is characterized by a remarkably tubercular dorsum, with large warts, several taking the form of short prominent ridges. Such a tubercular integument is also typical for the lectotype of G. ventrimaculatus (Fig. 14) and for individuals previously assigned to this species. Among the lineages that in the recent past have been assigned to G. malagasius, three occur near the malagasius type locality Folohy: lineages A, B and D. Of these, lineages A and B usually have a much less tubercular dorsum, whereas in lineage D the dorsum is also strongly tubercular.

Individuals of lineages A, B, C and D (all previously assigned to G. malagasius) are characterized by reddish color on the posteriormost portion of the belly and the ventral side of the thighs (Figs 17, 20, 22, 24). The reddish color partly extends on the ventral surface of shanks, covers a part of the inguinal region, and often also forms a small patch at the insertion of forelimbs. The reddish color appears more intense in lineage D than in lineage B. Males of lineage B can easily be distinguished from those of lineages A, C and D by the presence of a distinct tibial gland (lacking in the other three lineages). Individuals of lineage C are characterized by their head and dorsum being covered by prominent and distinct, partly spine-like tubercles (Fig. 22), which are less prominent in specimens of lineages A and B and coarser and less spiny in lineage D. Individuals of lineage D have the third toe distinctly longer than the fifth.

Bioacoustic differentiation

Advertisement calls in the subgenus Laurentomantis are multi-note calls, containing non-tonal notes. Calls are usually emitted in call series. The calls studied herein are all somewhat similar and simple in their structure, namely pulsatile or pulsed notes repeated at regular intervals, with only one exception where note repetition within calls is rather irregular. Despite the overall similar structure of advertisement calls in this species complex, detailed comparison of call parameters among the genetically identified clades reveals more or less pronounced differences in certain characters.

Among all calls analyzed, calls assignable to G. ranjomavo are longest in duration (1780–2526 ms) and differ from calls of G. horridus (call duration 543–618 ms) by notes consisting of a single pulse only (versus 2–4 pulses/note), low note repetition rate within calls (10.8–14.2 vs. 29.4–31.0 notes/second), and higher dominant frequency (2348–3204 vs. 1172–1369 Hz).

Calls of G. striatus contain very short single-pulse notes, but mainly differ from calls of G. ranjomavo by shorter call duration (439–1360 vs. 1780–2526 ms) and higher note repetition rate within calls (25.6–33.5 vs. 10.8–14.2 notes/second).

Calls of the recently described G. marokoroko contain multi-pulsed notes with clearly separated pulses. They are similar to calls of G. horridus in general character, but mainly differ by lower note repetition rate (18.2–19.6 vs. 29.4–31.0 notes/second) and higher dominant frequency (2916–3192 vs. 1172–1369 Hz).

Calls assignable to G. malagasius (previously reported under its synonym ventrimaculatus) contain multi-pulsed notes, with 2–10 pulses being partly fused and thus not always clearly distinguishable in oscillograms. Interestingly, calls from Vohiparara and Manombo generally agree in call duration, note duration and note structure, but differ considerably by a doubling number of notes per call and consequently much higher note repetition rate within calls of the Manombo population, which together with the sample from Befotaka Midongy forms a subclade within G. malagasius.

Among the calls assignable to the four unnamed lineages identified in the G. malagasius complex, calls of lineages A and C both contain multi-pulsed notes with a clear pulse structure, with the calls of lineage A containing a higher number of pulses per note (6–10 vs. 1–4) and exhibiting a lower note repetition rate within calls (20.8–21.6 vs. 43.5–54.1 notes/second) when compared to lineage C. Calls of lineage B from Andasibe, Ankeniheny and Vohidrazana contain single-pulse notes only and are unique among all calls analyzed in exhibiting rather variable note repetition rate within calls. Calls of lineage B from Ambatolahy (Ranomafana National Park) are generally in agreement with the latter, but differ slightly by exhibiting a regular note repetition within calls and more pronounced amplidute modulation. No call recordings are available from lineage D. For detailed call descriptions and comparison see species accounts and Table 2.

Taxonomy

Our results suggest the necessity of taxonomic changes in the subgenus Laurentomantis. We re-define G. malagasius by considering Trachymantis malagasia ventrimaculatus Angel, 1935 as a junior synonym of Microphryne malagasia Methuen and Hewitt, 1913. This leaves four genetic lineages (A–D) without scientific names. Of these, lineages A, B and C concordantly differed by a strong mitochondrial divergence (with uncorrected pairwise differences in the 16S gene > 4%), by a lack of haplotype sharing in two nuclear-encoded genes, and by temporal characteristics of advertisement calls. Only lineages B and D have limited haplotype sharing in one nuclear-encoded gene, but all four lineages have several morphological characters that enable their identification. This combined evidence suggests that lineages A, B, C and D represent biologically distinct species, which will be formally named in the respective species accounts in the following.

Gephyromantis horridus (Boettger, 1880)

Hemimantis horrida Boettger, 1880: 282

Note

This species was described based on a juvenile specimen from Nosy Be, and a second specimen (an adult female) from the type locality has previously been reported by Glaw and Vences (2011). Pictures of both specimens are shown in Fig. 4; measurements are given in Vences et al. (2002) and Glaw and Vences (2011), respectively. The Tsaratanana population included in this species by Vences et al. (2002) is assigned to G. ranjomavo based on genetic data, and the only locality beside Nosy Be currently accepted for G. horridus is therefore Montagne d’Ambre. Measurements of the first male known from the Montagne d’Ambre population are given in Table 1 and photos in life of various individuals from this locality are included in Fig. 5. The male is characterized by very prominent femoral glands of the same color as the surrounding skin of the ventral thigh, composed of approximately six large gland granules as recognizable in external view (Fig. 5). The male has no tibial gland. According to the measurements reported in Vences et al. (2002), Glaw and Vences (2011) and this study, SVL is 33.0–35.4 mm in two females and 33.5 in a male from Montagne d’Ambre, and 33.7 mm in a female from Nosy Be. The advertisement call of this species was previously unknown, and is described in the following for the Montagne d’Ambre population.

Table 1.

Download as

CSV

XLSX

Morphometric measurements (all in mm) of voucher specimens of Gephyromantis belonging to the subgenus Laurentomantis newly examined for this study and identified on the basis of DNA sequence data. See Vences et al. (2002) and Glaw and Vences (2011) for measurements of additional individuals. For abbreviations of measurements, see Materials and Methods. HT, holotype.

Voucher	Field number	Status	Sex	Locality	SVL	HW	HL	ED	HTD	END	NSD	NND	FORL	HAL	HIL	FOTL	FOL	TIBL	FGL	FGW
G. horridus
ZSM 126/2018	MSTEMP 1		M	Montagne d‘Ambre	33.5	13.2	13.8	4.8	2.3	3.7	2.3	3.5	24.3	11.5	53.6	23.8	15.6	15.8	7.7	4.8
G. ranjomavo
ZSM 60/2016	MSZC 0163		M	Ampotsidy	26.6	10.5	10.6	4.0	2.2	2.8	2.0	2.8	19.0	9.2	41.9	18.6	11.7	12.6	3.1	2.6
G. malagasius (=ventrimaculatus)
ZSM 2464/2007	ZCMV 5497		M	Manombo	27.0	10.7	11.4	4.4	1.8	3.3	1.8	2.2	19.0	8.6	44.1	18.9	12.0	14.0	5.8	2.4
ZSM 2463/2007	ZCMV 5486		F	Manombo	29.8	11.7	13.0	4.3	2.0	3.2	2.1	2.9	24.7	11.4	51.4	22.8	14.5	16.9	NA	NA
ZSM 537/2006	ZCMV 3362		M	Ranomafana	23.6	9.9	10.0	3.7	2.0	3.2	1.7	2.4	19.0	8.8	40.8	17.2	11.2	12.7	5.4	2.0
G. matsilo sp. nov. (lineage A)
ZSM 711/2009	ZCMV 7234	HT	M	Ambatoroma, Befanjana	21.9	8.8	8.9	3.4	1.7	2.7	2.0	2.2	17.0	8.0	37.1	16.1	10.8	12.2	5.1	2.0
ZSM 712/2009	ZCMV 7300	PT	F	Babitanety, Befanjana	24.9	9.6	10.9	4.0	2.0	2.8	1.8	2.6	18.6	8.5	42.1	17.8	11.1	13.1	NA	NA
ZMA 20247	ZCMV 90	PT	M	Ambohitsara	21.7	8.9	9.3	3.5	1.7	2.7	1.2	2.1	18.1	8.8	39.7	17.5	11.4	11.9	5.8	2.3
G. fiharimpe sp. nov. (lineage B)
ZSM 164/2016	FGZC 5181	HT	M	Mandraka	23.4	8.9	9.6	3.2	1.7	3.2	2.0	2.4	17.5	8.1	41.0	18.4	12.2	13.0	3.7	1.6
ZMA 19421	FGMV 2002.415	PT	F?	Ranomafana	25.8	9.2	10.7	3.4	1.9	2.7	1.7	2.0	18.5	9.2	43.5	19.3	12.5	13.3	NA	NA
G. oelkrugi sp. nov. (lineage C)
ZSM 271/2016	FGZC 5300	PT	F?	Masoala	21.1	8.4	8.7	3.3	1.8	2.8	1.5	2.3	14.8	7.1	32.6	14.5	9.6	10.1	NA	NA
ZSM 315/2010	NA	PT	M	Ambodivoangy	21.6	8.7	8.9	3.7	2.2	2.1	1.2	2.6	16.9	17.8	38.9	16.7	11.0	12.0	3.9	1.6
ZSM 314/2010	FGZC 4220	HT	M	Ambodivoangy	21.9	8.4	9.0	3.0	2.3	3.1	1.2	2.4	15.0	6.8	33.8	14.8	9.9	11.0	3.2	2.0
G. portonae sp. nov. (lineage D)
ZSM 115/2021	ACZCV 0032	HT	M	Betampona	22.9	9.0	10.7	3.4	1.7	2.9	1.6	2.7	17.0	8.2	36.1	16.1	10.5	11.5	5.5	2.5
ZSM 118/2021	ACZCV 0223	PT	F	Betampona	26.1	9.8	10.5	4.0	1.9	3.0	1.3	2.8	19.3	9.5	45.2	20.3	12.5	13.6	NA	NA
ZSM 116/2021	ACZCV 0030	PT	F	Betampona	25.1	9.8	10.5	3.4	1.5	3.4	2.0	2.6	18.9	8.7	40.5	17.7	11.3	12.5	NA	NA
ZSM 117/2021	ACZCV 0025	PT	M	Betampona	22.4	8.3	9.1	3.1	1.8	2.5	1.9	2.4	19.3	9.0	41.2	17.7	12.1	12.6	5.1	2.4

Table 2.

Download as

CSV

XLSX

Comparative numerical parameters of analyzed advertisement calls of frogs in the subgenus Laurentomantis. Values are provided as range followed by mean ± standard deviation in parentheses.

species	locality	call duration [ms]	note duration [ms]	notes per call	note repetition rate within call [notes/second]	pulses per note	dominant frequency [Hz]
G. horridus	Montagne d’Ambre	543–618 (585.1±27.4)	6–12 (9.4±1.7)	17–19 (18.1±0.8)	29.4–31.0 (29.9±0.7)	2–4 (3.0±0.5)	1172–1369 (1264±71)
G. ranjomavo	Manarikoba Forest	1780–2526 (2237.9±255.6)	10–19 (13.6±3.1)	23–33 (29.0±3.2)	10.8–14.2 (12.7±1.0)	1	2348–3204 (2909±355)
G. striatus	Marojejy	439–1360 (848.5±319.1)	3–7 (5.1±0.8)	17–45 (29.8±8.4)	25.6–33.5 (31.1±3.2)	1	3768–4153 (3983±161)
G. marokoroko	Vohidrazana	557–1203 (968.7±357.7)	10–26 (17.4±5.0)	12–28 (21.0±8.2)	18.2–19.6 (18.8±0.6)	2–5 (3.3±0.9)	2411–2476 (2444±28)
G. malagasius	Vohiparara/ Ranomafana	360–465 (428.4±37.2)	8–17 (12.6±2.9)	8–10 (9.6±0.8)	19.8–21.1 (20.5±0.5)	2–10 (6.2±2.1)	2606–3516 (3127±254)
G. malagasius	Manombo	348–446 (386.3±34.5)	9–16 (12.2±1.9)	16–20 (17.5±1.6)	42.7–45.5 (44.7±1.1)	2–9 (5.5±1.9)	2916–3192 (3041±107)
G. matsilo sp. nov. (lineage A)	Betampona	355–660 (552.3±88.0)	13–21 (16.2±2.5)	8–15 (12.6±1.9)	20.8–21.6 (21.2±0.3)	6–10 (6.9±1.1)	3090–3434 (3200±116)
G. fiharimpe sp. nov. (lineage B)	Ankeniheny	609–935 (798.8±110.8)	7–9 (7.8±0.8)	11–19 (15.5±2.4)	14.9–25.6 (18.3±3.9)	1	3638–3723 (3676±23)
G. fiharimpe sp. nov. (lineage B)	Andasibe	840–1503 (1179.3±232.7)	5–12 (7.6±2.2)	26–43 (36.9±7.3)	18.2–37.4 (25.5±6.5)	1	3649–4078 (3796±157)
G. fiharimpe sp. nov. (lineage B)	Vohidrazana	950–1764 (1405.6±290.4)	2–7 (4.3±1.7)	26–52 (40.7±11.5)	18.8–54.1 (32.1±13.1)	1	3726–3913 (3781±68)
G. fiharimpe sp. nov. (lineage B)	Ambatolahy/Ranomafana	916–1162 (1073.9±86.0)	4–7 (6.2±0.9)	27–34 (30.6±2.3)	25.6–28.0 (26.8±1.0)	1	3305–3736 (3523±145)
G. oelkrugi sp. nov. (lineage C)	Ambodivoangy	279–504 (350.9±88.4)	3–10 (6.8±1.3)	13–25 (16.9±4.4)	43.5–54.1 (46.5±4.0)	1–4 (2.9±0.7)	3266–3882 (3539±190)

Call

The advertisement call of specimen ZSM 126/2018 was recorded at 20:20–21:00 on 11 December 2017 on a muddy bank beside a very slow segment of a stream on the west slope of Montagne d’Ambre (12.5915°S, 049.1372°E, 939 m a.s.l.; air temperature not available). It consists of a series of short, pulsed notes (Fig. 6). They were extremely difficult to localize in the field due to their low amplitude. There is amplidude modulation within each call, with call energy being greatest at approximately 70% of the call’s duration, with initial notes being the least energetic ones. Within calls, notes are repeated at very constant intervals. Each note contains few clearly countable pulses repeated at an approximate rate of 500 pulses/second. Within regular call series, calls are emitted in rapid succession. Numerical parameters of 9 analyzed calls are as follows (range followed by mean ± standard deviation in parentheses): call duration 543–618 ms (585.1 ± 27.4 ms); note duration 6–12 ms (9.4 ± 1.7 ms); number of notes per call 17–19 (18.1 ± 0.8); note repetition rate within calls 29.4–31.0 notes/second (29.9 ± 0.7 notes/second); pulses per note 2–4 (3.0 ± 0.5); dominant frequency 1172–1369 Hz (1264 ± 71 Hz); prevalent bandwidth 900–3500 Hz. Calls were emitted in regular call series at a rate of approximately 41 calls/minute.

Gephyromantis ranjomavo Glaw and Vences, 2011

Gephyromantis ranjomavo Glaw and Vences, 2011: 122

Note

This species was previously known from two specimens (Glaw and Vences 2011), one from the Marojejy Massif and another without precise locality data. As redefined here, the species includes a second deep genetic lineage, expanding the species’ range into the southwestern and western foothills of the Tsaratanana Massif (localities Ampotsidy and Manarikoba Forest on the western slope). Morphometric measurements of one newly collected male specimen (ZSM 60/2016 from Ampotsidy) are given in Table 1, and color in life of this individual is shown in Fig. 7. Compared to its sister species G. horridus (see data above), both the holotype from Marojejy (Glaw and Vences 2011) and the male from Ampotsidy have relatively small femoral glands without externally visible large gland granules. A tibial gland is present in all specimens. According to the measurements reported in Vences et al. (2002), Glaw and Vences (2011) and this study, male SVL is 23.5 mm in Marojejy, and 25.8–28.1 mm in the remaining specimens; females are unknown.

Figure 7.

Adult males of Gephyromantis ranjomavo in life. A Dorsolateral and B ventral view of the holotype, ZSM 222/2005 (FGZC 2843) from Marojejy National Park. C–E Dorsolateral, dorsal and ventral views of specimen ZSM 60/2016 (MSZC 0163) from Ampotsidy, attributed to G. ranjomavo but genetically divergent. Not to scale.

Call

We redescribe the advertisement call genetically assignable to Gephyromantis ranjomavo recorded on 17 February 1997 at Manarikoba Forest, Tsaratanana Strict Nature Reserve (air temperature 17.5°C; Vences et al. 2006: CD2, track 27). The call (previously described by Vences et al. 2002) consists of a long multinote call, usually emitted in series at regular intervals and fast succession (Fig. 8). Slight amplitude modulation is recognizable within calls, with notes at the beginning and the end of the call having lower energy. Notes consist of a single pulse, but in a few notes some diffuse substructure is recognizable. Frequency modulation is absent in notes. Numerical call parameters of 7 newly analyzed calls are as follows (range followed by mean ± standard deviation in parentheses): call duration 1780–2526 ms (2237.9 ± 255.6 ms); note duration 10–19 ms (13.6 ± 3.1 ms); number of notes per call 23–33 (29.0 ± 3.2); note repetition rate within calls 10.8–14.2 notes/second (12.7 ± 1.0 notes/second); call repetition rate within call series approximately 14–15 calls/minute; dominant frequency 2348–3204 Hz (2909 ± 355 Hz); prevalent bandwidth 2000–3800 Hz.

Figure 8.

Audiospectrogram and corresponding oscillogram of a 1000 ms section of one advertisement call of Gephyromantis ranjomavo recorded on 17 February 1997 at Manarikoba Forest. Recording band-pass filtered at 1000–7500 Hz.

Gephyromantis striatus (Vences, Glaw, Andreone, Jesu and Schimmenti, 2002)

Mantidactylus striatus Vences, Glaw, Andreone, Jesu and Schimmenti, 2002: 203

Note

This species was described from lower elevations of the Marojejy Massif, and specimens from Tsararano and Masoala were assigned to the species based on morphology (Vences et al. 2002). We provide the first molecular data for specimens collected at two sites on the Masoala Peninsula, confirming that these populations are conspecific with those from Marojejy (Fig. 1). G. striatus is defined by its rather weakly tubercular dorsum and an incomplete (posterior) light vertebral stripe (Fig. 9). A tibial gland is absent. Based on morphometric data in Vences et al. (2002), SVL ranges from 22.2–23.8 mm in males and 23.9–26.9 mm in females.

Call

The advertisement call was recorded on 22 February 1995 at Marojejy National Park (air temperature 25.0°C; Vences et al. 2006: CD2, track 29). It consists of a moderately long multinote call of variable duration, usually emitted in series at rather regular intervals (Fig. 10). Slight amplitude modulation is recognizable within calls, with notes at the beginning and the end of the call having lower energy. Notes are very short and appear to consist of two pulses, which are largely fused. They are usually repeated at regular intervals within calls, but might be spaced more narrowly at the beginning and end of a call. Frequency modulation is absent. Numerical call parameters of 23 analyzed calls emitted by 2 individuals are as follows (range followed by mean ± standard deviation in parentheses): call duration 439–1360 ms (848.5 ± 319.1 ms); note duration 3–7 ms (5.1 ± 0.8 ms); number of notes per call 17–45 (29.8 ± 8.4); note repetition rate within calls 25.6–33.5 notes/second (31.1 ± 3.2 notes/second); call repetition rate within call series approximately 22–27 calls/minute; dominant frequency 3768–4153 Hz (3983 ± 161 Hz); prevalent bandwidth 2300–5200 Hz.

Figure 9.

Adult male of Gephyromantis striatus from Marojejy National Park in life: A dorsolateral and B ventral view; photographed in 1994 (individual not reliably attributable to a voucher specimen). Not to scale.

Figure 10.

Audiospectrogram and corresponding oscillogram of one advertisement call of Gephyromantis striatus recorded on 22 February 1995 at Marojejy National Park. Recording band-pass filtered at 1500–7500 Hz.

Gephyromantis marokoroko Hutter, Andriampenomanana, Andrianasolo, Cobb, Razafindraibe, Abraham and Lambert, 2022

Gephyromantis marokoroko Hutter, Andriampenomanana, Andrianasolo, Cobb, Razafindraibe, Abraham and Lambert, 2022: 486

Note

A recently described species from the Northern Central East of Madagascar (Hutter et al. 2022) that is genetically and morphologically distinct by a reddish iris, gray to whitish vocal sacs and large femoral glands with 7–8 distinct, externally visible gland granules, and a dorsal skin texture consisting mainly of sharp ridges rather than of rounded tubercles (Fig. 11). A tibial gland is absent. According to the original description, SVL is 24.0–27.0 mm in males and 23.9–24.6 mm in females. The species is so far only known from several sites near Andasibe: Belakato in Andasibe-Mantadia National Park, Vohimana, Vohidrazana (type locality) and Tavalobe near Vohidrazana. We reanalyzed the call of G. marokoroko to ensure full comparability with the call descriptions of all other Laurentomantis.

Call

The advertisement call was recorded on 6 January 2016 at Vohidrazana (air temperature 20.4°C; call voucher KU 343230 [CRH 1110]). It consists of a series of short distinctly pulsed notes of variable duration (Fig. 12). There is recognizable amplidude modulation within each call, with call energy being greatest at approximately 60% of the call’s duration, with initial notes being the least energetic ones. Within calls, notes are repeated at constant intervals. Each note contains a certain number of clearly separated pulses repeated at an approximate rate of 150–170 pulses/second. Numerical parameters of 3 analyzed calls are as follows (range followed by mean ± standard deviation in parentheses): call duration 557–1203 ms (968.7 ± 357.7 ms); note duration 10–26 ms (17.4 ± 5.0 ms); number of notes per call 12–28 (21.0 ± 8.2); note repetition rate within calls 18.2–19.6 notes/second (18.8 ± 0.6 notes/second); pulses per note 2–5 (3.3 ± 0.9); dominant frequency 2411–2476 Hz (2444 ± 28 Hz), with a second peak of almost equal energy at around 3600–3700 Hz; prevalent bandwidth 1000–5200 Hz. Call repetition rate in short call series (containing 3 calls) approximately 18–20 calls/minute.

Figure 11.

Adult male of Gephyromantis marokoroko (KU 343230) from Vohidrazana in life, in A dorsolateral, B dorsal and C ventral view.

Figure 12.

Audiospectrogram and corresponding oscillogram of a 1000 ms section of an advertisement call of Gephyromantis marokoroko recorded on 6 January 2016 at Vohidrazana.

Gephyromantis malagasius (Methuen and Hewitt, 1913)

Microphryne malagasia Methuen and Hewitt, 1913b: 55

Trachymantis malagasia ventrimaculatus Angel, 1935: 205; syn. nov.

Note

As discussed above, we redefine G. malagasius based on molecular data from the holotype as containing those frogs previously considered as G. ventrimaculatus (e.g., Blommers-Schlösser and Blanc 1991; Vences et al. 2002; Glaw and Vences 2007). These frogs are easily recognizable by their reddish-brown dorsum with highly expressed tubercular skin texture, and a highly contrasted ventral color with grayish to bluish reticulations on a deep black ground color (Fig. 13). A tibial gland is absent. As hypothesized in the Molecular Phylogenetics section, we assume this typical color pattern has faded in the holotype of Microphryne malagasia while its general morphology roughly agrees with the morphology of the other specimens assigned to the species, despite being of smaller size (Fig. 14). Vences et al. (2002) discussed the mention in the original description (Methuen and Hewitt 1913b) of “large white blotches” present on the “hidden surface of the thighs and tibiae”, and interpreted this pattern as indicative of the reddish areas in life present on those frogs they assigned to G. malagasius. However, the described pattern might as well correspond to the contrasted markings found on the ventral side of the specimens previously considered as G. ventrimaculatus, which might have persisted in the hidden (not light-exposed) parts of the limbs at the time of the original description.

Figure 13.

Specimens of Gephyromantis malagasius (previously considered as G. ventrimaculatus, herein considered to be a junior synonym of G. malagasius), in life. A, B Adult male from Ranomafana in dorsolateral and ventral view, photographed 2006, probably corresponding to voucher specimen ZSM 537/2006 (ZCMV 3362). C, D Adult male from Manombo, ZSM 2464/2007 (ZCMV 5497) in dorsolateral and ventral view, photographed 2007. E, F Adult male from Ranomafana (Vohiparara), photographed 1996, possibly corresponding to voucher specimen ZFMK 62281. G, H Adult male from Ambohitantely, not clearly assignable to a voucher specimen, photographed in 2017 in Ambohitantely Special Reserve. Not to scale.

Figure 14.

A, B Preserved holotype of Gephyromantis malagasius (originally named Microphryne malagasia; specimen TMP 10076) in dorsal and ventral view. C, D Preserved lectotype of Trachymantis malagasia ventrimaculatus (MNHN 1935.172), herein considered as a junior synonym of G. malagasius. Photos C and D by Antoine Fraysse, project MNHN-RECOLNAT (ANR-11-INBS-0004), photographed in 2015 (http://coldb.mnhn.fr/catalognumber/mnhn/ra/1935.173)

Vences et al. (2002) also provided information on the femoral gland of the G. malagasius holotype, which according to them consists of 1–2 granules (examined in external view), while the femoral glands of specimens assigned to G. ventrimaculatus were found to consist of nine granules. For this study, we unfortunately were not able to examine the gland of the malagasius holotype in internal view, but a detailed look in external view (Fig. 13) reveals a rounded, well-defined structure, which may consist of several granules (more than the 1–2 previously reported), and clearly differs from the corresponding gland structures typical for specimens of lineage B where one or two single granules are arranged longitudinally on the ventral thigh; however, at least lineages C and D also have glands composed of more (4–7) granules, suggesting that femoral gland characteristics cannot be used to unambiguously allocate the malagasius holotype to a genetic lineage. Summarizing morphometric measurements of Vences et al. (2002) and the present study, males measure 23.0–27.0 mm, females measure 29.1–29.8 mm, and the malagasius holotype measures 20.2 mm SVL. We here provide bioacoustic data from Ranomafana National Park (Vohiparara) and from a second locality, Manombo.

Call

The advertisement call of G. malagasius recorded at Vohiparara (Vences et al. 2006, CD2, track 30) from ZFMK 62281 consists of a series of very short pulsed notes and is emitted in series at regular intervals (Fig. 15). There is amplidude modulation within each call, with call energy increasing from the beginning of the call reaching the maximum amplitude at about 40% of its duration and from there decreasing towards its end. Pulses within notes are partly fused, but clearly countable. Within calls, notes are repeated at regular intervals. Numerical parameters of 7 analyzed calls are as follows (range followed by mean ± standard deviation in parentheses): call duration 360–465 ms (428.4 ± 37.2 ms); note duration 8–17 ms (12.6 ± 2.9 ms); number of notes per call 8–10 (9.6 ± 0.8); note repetition rate within calls 19.8–21.1 notes/second (20.5 ± 0.5 notes/second); pulses per note 2–10 (6.2 ± 2.1); call repetition rate within call series approximately 17–18 calls/minute; dominant frequency 2606–3516 Hz (3127 ± 254 Hz); prevalent bandwidth 1800–6000 Hz.

A second recording from 23 February 2007 (Manombo; air temperature estimated at 23–25°C) consists of a series of pulsed notes and is emitted in series at regular intervals (Fig. 16). There is amplidude modulation within each call, with call energy increasing from the beginning of the call reaching the maximum amplitude at about 60% of its duration and from there decreasing towards its end. Pulses within notes are partly fused, but in most notes distinct pulses are recognizable and countable. Within calls, notes are repeated at regular intervals at a high rate. Numerical parameters of 14 analyzed calls are as follows (range followed by mean ± standard deviation in parentheses): call duration 348–446 ms (386.3 ± 34.5 ms); note duration 9–16 ms (12.2 ± 1.9 ms); number of notes per call 16–20 (17.5 ± 1.6); note repetition rate within calls 42.7–45.5 notes/second (44.7 ± 1.1 notes/second); pulses per note 2–9 (5.5 ± 1.9); call repetition rate within call series approximately 39–44 calls/minute; dominant frequency 2916–3192 Hz (3041 ± 107 Hz); prevalent bandwidth 1700–5500 Hz.

Figure 15.

Audiospectrogram and corresponding oscillogram of one advertisement call of Gephyromantis malagasius recorded at Vohiparara. The oscillogram below shows a 100 ms section of the call figured above, showing two notes and their repective pulse structure. Recording band-pass filtered at 1500–7500 Hz.

Figure 16.

Audiospectrogram and corresponding oscillogram of one advertisement call of Gephyromantis malagasius recorded at Manombo. The oscillogram below shows a 100 ms section of the call figured above, showing four notes and their repective pulse structure. Recording band-pass filtered at 1500–6200 Hz.

Distribution

G. malagasius as redefined here is known based on genetically confirmed records from (1) the type locality Folohy, (2) Ranomafana, (3) Manombo, and (4) Befotaka-Midongy. Furthermore, morphologically identified specimens with the typical ventral pattern are known from (5) Andasibe, (6) Ambohitantely (based on phenotypically identified specimens collected by one of us [APR]; see Fig. 13G,H), and (7) the type locality of the junior synonym ventrimaculatus (“Isaka Ivondro, alt. 700 m”), which is located within or very close to the current Andohahela National Park. It is important to mention that the exact location of the type locality Folohy is uncertain. At this site, collections were made by “M. Herschell-Chauvin” in 1911 (Methuen and Hewitt 1913b). Methuen and Hewitt (1913a) name the collector “Monsieur Herschell-Chauvin”, probably referring to the English naturalist and photographer Charles Herschell-Chauvin who worked in Tamatave (=Toamasina) in the first years of the 20^th century. Methuen and Hewitt (1913a) place the locality Folohy “in the neighbourhood of Tamatave”, and also Blommers-Schlösser and Blanc (1991) plot Folohy as near-coastal locality close to Toamasina in their distribution maps. Barbour and Loveridge (1929) locate Folohy “east of Lake Alaotra” for a frog specimen exchanged from the Transvaal Museum. The catalogue of the Museum of Comparative Zoology, Harvard, includes a lemur specimen (MCZ 18740) collected by Frederick Roelker Wulsin in 1915, with the verbatim locality information “Folohy forest, 100 miles west of Tamatave” (https://www.idigbio.org/portal accessed 19 November 2021), which however is unlikely to be correct as it would place the site onto Madagascar’s high plateau, outside the main rainforest area. Along with Blommers-Schlösser and Blanc (1991) we here assume that Folohy refers to a low- or mid-elevation site close to Toamasina.

The occurrence of individuals morphologically corresponding to G. malagasius as redefined herein in Andasibe is supported by three records: one voucher specimen collected by Denis Vallan and reported in Vences et al. (2002); one specimen photographed by Daniel S. Moen; and one specimen photographed by Devin Edmonds in the Mitsinjo forest on 4 December 2014 (https://www.inaturalist.org/observations/2315204).

Gephyromantis matsilo sp. nov. (lineage A)

http://zoobank.org/55361232-0291-4652-80F3-B030F0A91FBB

Holotype

ZSM 711/2009 (field number ZCMV 7234), adult male, from campsite “Ambatoroma, S II “ (a campsite in the Manompana – Befanjana Forest area, approximately at coordinates 16.66°S, 49.59°E; precise elevation unknown), Analanjirofo Region, eastern Madagascar, collected on 19 May 2009 by J.E. Randrianirina.

Paratypes

ZSM 712/2009 (ZCMV 7300), adult female, from campsite “Babitanety, S III” (Manompana – Befanjana Forest), Analanjirofo Region, eastern Madagascar, collected on 20 May 2009 by J.E. Randrianirina; MRSN A5678 (FAZC 13344), adult male, from Sahavontsira (Miorimivalana, Fenerive Est), Analanjirofo Region, eastern Madagascar, collected on 23 January 2006 by F. Andreone, F. Mattioli and J.E. Randrianirina; ZMA 20247 (ZCMV 90), adult male, from Ambohitsara (21.3572°S, 047.8157°E, 294 m a.s.l.), Vatovavy-Fitovinany Region, eastern Madagascar, collected on 21 January 2004 by D.R. Vieites and I. de la Riva.

Etymology

The species epithet is derived from the Malagasy adjective matsilo (spiny) and refers to the spiny tubercles on the dorsum of this frog. The name is used as a noun in apposition.

Diagnosis

A member of the subfamily Mantellinae based on the presence of intercalary elements between terminal and subterminal phalanges of fingers and toes (verified externally), and on the absence of nuptial pads and presence of femoral glands in males. Assigned to the subgenus Laurentomantis in the genus Gephyromantis based on the strongly tubercular dorsal skin, absence of foot webbing, completely connected lateral metatarsalia, and molecular phylogenetic relationships. The new species differs from all nominal species of the subgenus Laurentomantis as follows: From G. horridus by smaller body size (male SVL 21.7‒21.9 mm vs. 33.5 mm) and absence of a dorsal pattern of two blackish transverse patches (vs. presence); from G. ranjomavo by slightly smaller body size (male SVL 21.7‒21.9 mm vs. 23.5–28.1 mm), absence of light brown to orange-brown color covering limbs dorsally (vs. presence), and absence of a tibial gland (vs. presence); from G. striatus by absence of a vertebral stripe posteriorly on dorsum (vs. presence) and a more strongly tubercular dorsal skin; from G. malagasius (as redefined herein) by absence of a distinct bluish gray pattern on a dark venter (vs. presence); and from G. marokoroko by a more coarsely tubercular dorsal skin, presence of red color ventrally on limbs (vs. absence), absence of orange spots and vermiculations on dorsum (vs. presence), and absence of gray to whitish color on vocal sac (vs. presence). Also distinguished from G. horridus and G. striatus by longer notes in advertisement calls (13–21 vs. 3–12 ms). Furthermore, from all nominal Laurentomantis species distinguished by the presence of bright red color in the inguinal region and probably ventrally on limbs and posterior belly in life (see. Fig. 17 where this color is recognizable in the inguinal region) (vs. absence), and by a substantial genetic divergence (>8% uncorrected pairwise distance in the 16S gene). For a distinction from the other new species described in the following (lineages B, C and D), see the diagnoses in the respective species accounts below.

Description of the holotype

Adult male in good state of preservation (Fig. 18). SVL 21.9 mm, for other measurements see Table 1. Body slender; head slightly longer than wide, wider than body; snout rounded in dorsal and lateral views; nostrils directed laterally, distinctly protuberant, much nearer to tip of snout than to eye; canthus rostralis concave; loreal region distinctly concave; tympanum distinct, rounded, 77% of eye diameter; no supratympanic fold except some elevated skin folds directly encircling the tympanum dorsally; tongue ovoid, distinctly forked posteriorly; vomerine teeth absent; choanae rounded; maxillary teeth present. Dermal fold along the posterior part of the lower jaws (the inflatable parts of the vocal sac) weakly expressed. Arms slender, subarticular tubercles single; outer and inner metacarpal tubercles weakly expressed, not prominent; fingers without webbing; relative length of fingers 1 < 2 = ≤ 4 < 3, second finger distinctly shorter than fourth finger on right hand, almost of same length on left hand; finger discs enlarged; nuptial pads absent. Hindlimbs slender; tibiotarsal articulation reaching nostrils when hindlimb is adpressed along body; lateral metatarsalia connected; inner metatarsal tubercle distinct, outer metatarsal tubercle small but recognizable; webbing between fingers and toes absent; relative toe length 1 < 2 < 5 = 3 < 4. Third toe of same length as fifth toe. Toe discs enlarged. Skin on upper surface granular, with rather indistinct ridges especially on anterior dorsum, and many smaller irregularly distributed tubercles on head, eyes, and dorsum. Ventral skin smooth on throat, chest and limbs, slightly granular on posterior belly. Femoral glands well delimited and distinctly recognizable from external view. No tibial glands.

After twelve years in preservative, dorsal coloration of head and body uniformly dark brown, with darker crossbands on hind- and forelimbs. Posterodorsal surface of thigh with a small pigmentless patch near the knee joint, this area presumably corresponding to reddish color in life. Ventrally cream, with a rather faint and irregular brown marbling, which is more contrasted on the ventral surface of the hindlimbs.

Figure 17.

Specimens of Gephyromantis matsilo sp. nov. (lineage A) in life in dorsolateral view. A Specimen from Ambodiriana probably corresponding to tissue sample PSG 1015. B Specimen from Antanambe, probably corresponding to tissue sample PSG 49.Vouchers not collected (photo by P.S. Gehring).

Figure 18.

Preserved holotype specimens of the four new Gephyromantis species in dorsal and ventral view. A, B G. matsilo sp. nov. (lineage A), ZSM 711/2009 (ZCMV 7234) from Ambatoroma, Befanjana Forest. C, D G. fiharimpe sp. nov. (lineage B), ZSM 164/2016 (FGZC 5181) from Mandraka. E, F G. oelkrugi sp. nov., ZSM 314/2010 (FGZC 4220) from Ambodivoangy. G, H G. portonae sp. nov. (lineage D), ZSM 115/2021 (ACZCV 0032) from Betampona.

Variation

All photographed and examined specimens of G. matsilo lack a tibial gland. The male ZMA 20247 from Ambohitsara agrees with the morphology of the specimens examined from the northern localities in the Befanjana forest, including body size, and a more spiny-granular dorsal skin compared to specimens of lineage B. The female ZSM 712/2009 is distinctly larger than the two measured males (SVL 24.9 mm vs. 21.7‒21.9 mm).

Call

The advertisement call (Fig. 19) was recorded on 12 December 2007 at Vohitsivalana, RNI Betampona (air temperature 20°C; Rosa et al. 2011: track 24; sequence HM364637 from specimen FAZC 13977). It consists of a series of short distinctly pulsed notes. There is considerable amplidude modulation within each call, with call energy being greatest at approximately the middle of the call, with initial notes being the least energetic. Within calls, notes are repeated at very constant intervals. Each note contains several clearly separated pulses repeated at an approximate rate of 500 pulses/second. In some notes, the terminal pulse is separated by a slightly larger interval from preceding pulses. Numerical parameters of 14 analyzed calls are as follows (range followed by mean ± standard deviation in parentheses): call duration 355–660 ms (552.3 ± 88.0 ms); note duration 13–21 ms (16.2 ± 2.5 ms); number of notes per call 8–15 (12.6 ± 1.9); note repetition rate within calls 20.8–21.6 notes/second (21.2 ± 0.3 notes/second); pulses per note 6–10 (6.9 ± 1.1); dominant frequency 3090–3434 Hz (3200 ± 116 Hz); prevalent bandwidth 2000–4700 Hz. Calls were emitted more or less isolated or in short call series (containing up to 12 calls) at a rate of 24–31 calls/minute within series.

Figure 19.

Audiospectrogram and corresponding oscillogram of one advertisement call of Gephyromantis matsilo sp. nov. recorded on 12 December 2007 at Betampona. The oscillogram below shows a 100 ms section of the call figured above, showing two notes and their respective pulse structure.

Distribution and natural history

Based on genetically verified records, the distribution area spans a south-north direction from: (1) Ambohitsara to (2) Betampona, (3) Ambodiriana, (4) two sites in Befanjana, and (5) Sahavontsira. The known elevational range of the species spans from near sea level (Ambodiriana, 53 m a.s.l.) to approximately 294 m a.s.l. (Ambohitsara). Very little is known on the natural history of this species. Despite intensive sampling, only three individuals of this species have been collected at Betampona, where the most commonly found Laurentomantis species is the lineage D. All individuals were collected in rainforest habitat.