The phylogenetic position of Hyperolius sankuruensis (Anura: Hyperoliidae) reveals biogeographical affinity between the central Congo and West Africa, and illuminates the taxonomy of Hyperolius concolor

Abstract

African reed frogs (Hyperolius, Hyperoliidae) represent a hyperdiverse genus of arboreal frogs, known for their high intraspecific variation and interspecific similarity. Many recent publications have offered phylogenetic reconstructions and revisions of the genus or specific species groups or complexes. However, there are still many taxa known only from a limited number of localities and collected material that still await molecular evaluation and validation. Among these is H. sankuruensis, a species formally known only from the type locality in the central Congo Basin. The results of our phylogenetic analyses showed this central Congolian taxon to be related to an undescribed species from southwestern Gabon, and unexpectedly to a group of West African species. The West African species also include the widespread H. concolor, which presently comprises three geographically separated subspecies, H. c. concolor occurring west of the Dahomey Gap, H. c. ibadanensis known from Nigeria, and H. c. guttatus from southwestern Cameroon. Species delimitation based on both mitochondrial and nuclear data, together with morphological analyses, found sufficient differences among the three subspecies to elevate them to species status. In addition, the species groups of one of the three major clades within the genus Hyperolius are revised.

Introduction

African reed frogs of the genus Hyperolius Rapp 1842 are a widespread and diverse group of mostly arboreal species (Schiøtz 1999). The genus encompasses a large number of recognized species that exhibit similarity in colouration among species and also high variability, even among conspecific populations, of which some are recognized at the subspecific level [e.g. H. balfouri (Werner 1908); Schiøtz 1975, 1999; Amiet 2012]. Sexual dichromatism, which is relatively rare in anurans, is prevalent in this genus (Veith et al. 2009, Channing and Rödel 2019, Portik et al. 2019, 2020). Using a phylogenomic approach, Portik et al. (2019) recognized three major clades within the genus: the Hyperolius nasutus complex, Hyperolius Clade 1, and Hyperolius Clade 2. Although recent research activities have been carried out on all three clades (e.g. Channing et al. 2013, Ernst et al. 2021, Channing 2022), there are still many unanswered questions regarding the identity or phylogenetic position of some taxa. Such, for example, is the little known Hyperolius sankuruensis Laurent, 1979 from the central Democratic Republic of the Congo (Laurent 1979). In previous publications, it was considered that H. sankuruensis may be closely related to H. platyceps (Boulenger 1900) and H. langi Noble 1924 (Laurent 1979), H. diaphanus Laurent, 1972, and H. frontalis Laurent 1950 (Laurent 1981) or Congolius robustus (Laurent, 1979) (formerly Hyperolius; Schiøtz 2006). However, our unpublished first preliminary results of phylogenetic analyses placed H. sankuruensis in a close relationship with the West African H. concolor (Hallowell, 1844) and related taxa, hereinafter referred to as the Hyperolius concolor group.

The Hyperolius concolor group, as defined later in this study, consists of three described and one undescribed species: H. concolor, H. bobirensis Schiøtz, 1967, H. zonatus Laurent, 1958, and ‘Hyperolius sp. 1’ sensuBurger et al. (2004) [or ‘Hyperolius sp. B’ sensuVeith et al. (2009)]. Hyperolius bobirensis is known only from four forested areas in south-western Ghana [Bobiri, type locality (Schiøtz 1967); Ankasa National Park (Rödel et al. 2005, Hillers et al. 2009); Atewa Reserve (Kouamé et al. 2007); and Kakum National Park (this study); Fig. 1B]. Hyperolius zonatus is a more widespread species. Its area of occurrence covers most of Liberia, south-eastern Guinea and Sierra Leone, and south-western Côte d’Ivoire (Schiøtz 1967, 1999, Rödel and Branch 2002, Rödel et al. 2004, Rödel and Glos 2019; collections of Museum für Naturkunde, Berlin, Germany; Fig. 1B). Hyperolius zonatus, similarly to H. bobirensis, prefers primary forests (Schiøtz 1967, 1999, Channing and Rödel 2019, Rödel and Glos 2019). The undescribed species of Hyperolius is known from a single locality in the Moukalaba-Dougoua Faunal Reserve (Moukalaba-Doudou National Park), Ogooué-Maritime Province, south-western Gabon, consisting of a series of swamps in forested habitat near a forest/savanna ecotone (Burger et al. 2004). Veith et al. (2009) showed a close relationship of this taxon (‘Hyperolius species B Gabon’) to H. concolor and H. zonatus.

Unlike the previous—forest—species, H. concolor favours degraded forests (farmbush sensuSchiøtz 1967) in the forest zones, and gallery forests and humid savanna in the savanna zone (Schiøtz 1967, 1999, Rödel 2000). It is a widespread species occurring from south-western Cameroon and eastern Nigeria in the East, to Guinea in the West (Fig. 1C; Channing and Rödel 2019, Nneji et al. 2019a, b, Rahman et al. 2020). Three subspecies are presently recognized (Frétey et al. 2014). Hyperolius concolor concolor (Hallowell, 1844) is the westernmost of them, the inaccurate type locality is given as ‘Liberia, W. Africa’ (Fig. 1C; Hallowell 1844). In the East, it is separated by the Dahomey Gap from H. concolor ibadanensis Schiøtz, 1967 (Schiøtz 1967, 1999). Rödel et al. (2007) reported H. concolor ibadanensis from Lokoli and Lama Forests in southern Benin in the Dahomey Gap. The area of occurrence of H. concolor ibadanensis spans from the above-mentioned forests in the West to the Cross River, south-eastern Nigeria, in the East, the type locality being the University Campus in the city of Ibadan, Oyo State, western Nigeria (Fig. 1C; Schiøtz 1967). Amiet (1978) reported a potential presence of this subspecies in the vicinity of Mamfe (or Mamfé), South-West Region, Cameroon. Portik et al. (2019) later included a specimen from Mamfe in their phylogenomic study as H. concolor ibadanensis, although Amiet (2012) previously raised the question of whether H. concolor ibadanensis should be considered valid. His doubt was based on overlapping geographical variation in the distribution of colour morphs. Schiøtz (1967) originally distinguished the two subspecies on the basis of male colouration. Some males of H. concolor ibadanensis possess distinct light yellow dorsolateral stripes, while these are absent in the nominotypical subspecies (Schiøtz 1967). However, Schiøtz (1967) observed these dorsolateral lines only in about a third of all examined male H. concolor ibadanensis. The third subspecies is H. concolor guttatus Peters, 1875. It was previously named ‘H. concolor ssp. indet.’ or ‘H. concolor ssp.’ (e.g. Schiøtz 1967, 1999) and originally described as H. guttatus (Peters 1875). Its area of occurrence extends from the Cross River in the West, along the Atlantic coast as far south as Eboundja (15 km south of Kribi), Sud Region, Cameroon, and east to Modé, Centre Region, Cameroon (Fig. 1C; Schiøtz 1967, 1999, Amiet 2012, Frétey et al. 2014; this study). Schiøtz (1967, 1999) did not consider this taxon because at that time it was classified as a synonym of H. ocellatus Günther 1858. However, Frétey et al. (2014) removed this taxon from H. ocellatus, where it was moved by Laurent (1956), placed it under H. concolor as a subspecies, and reported on a new colour morph present in specimens from the vicinity of Douala, Littoral Region, Cameroon. These frogs have a dorsum with dark spots and a dark canthal line. Frétey et al. (2014) also provided a detailed nomenclatural history of H. concolor guttatus, re-described the lectotype from ‘Cameruns’, which was designated by Laurent (1961), following the discussion of Laurent (1958), and specified its geographical origin as Douala, Cameroon. Frétey et al. (2014) noted that H. concolor concolor and H. concolor guttatus may well represent distinct species, while H. concolor ibadanensis seemed closer to the latter taxon. However, they also mentioned that further bioacoustic and genetic studies are needed to resolve this question.

Interestingly, Amiet (2012) reported on males with light dorsolateral lines from Mamfe, Cameroon, while he did not find this colour morph in males from Bafut, North-West Region, Cameroon. There, most of the specimens exhibited reduced but distinguishable typical ‘concolor’ dorsal colouration of an ‘hourglass’ pattern covered with large, black diluted spots. This suggested the possibility of a distinct population in the Mamfe area. Recently, Nneji et al. (2019a, b) and Rahman et al. (2020) reported a genetically distinct population, relative to H. concolor from nearby south-eastern Nigeria, which they referred to as a candidate taxon, H. cf. concolor. Another potentially related taxon is Hyperolius stenodactylus Ahl 1931, which is known only from the holotype collected in southern Cameroon (Ahl 1931, Tillack et al. 2021; Supporting Information, Fig. S7). This taxon is currently listed as a valid species (Frost 2024), although Perret (1966) suggested that this nominal taxon may represent a synonym of H. concolor.

At the time of its description, the above-mentioned Hyperolius sankuruensis was known only from a single locality in Omaniundu, Sankuru Province (former Kasaï-Oriental), Democratic Republic of the Congo (3.35°S, 23.27°E; Laurent 1979; Fig. 1A). There it was collected in 1959 together with the sympatric Congolius robustus (Laurent 1979). After its description, H. sankuruensis had not been found until 2005, when its identity, however, was not recognized (Schiøtz 2006). Hyperolius sankuruensis was, therefore, classified by the IUCN Amphibian Specialist Group and Conservation International, with support from Global Wildlife Conservation, as one of the so-called ‘Lost Frogs’. It was officially rediscovered in 2010 (Kielgast and Lötters 2011, Moore 2011, 2014). However, H. sankuruensis has never been included in any molecular analysis and, thus, its systematic position and biogeographic affinities are not well understood.

Schiøtz (2006) reported on the genus Hyperolius from the central Democratic Republic of the Congo (DRC), and mentioned an adult female specimen (ZMUC R771205) of a large ‘Hyperolius sp.’ from Mabali in the vicinity of Lake Tumba, Équateur Province, DRC (0.88°S, 18.13°E), collected in December 2005. Schiøtz (2006) thought that the specimen from Mabali represented an undescribed species (Supporting Information, Fig. S1), but was hesitant to describe it on the basis of a single female. This specimen, however, possessed the same size, colouration, and similar webbing as the female paratype of H. sankuruensis. Its identity was later confirmed as H. sankuruensis when compared to the original type specimens (Supporting Information, Fig. S2), and further newly collected material of this species (Kielgast and Lötters 2011; Supporting Information, Fig. S3).

Our study aims to clarify three main objectives, using molecular and morphometric approaches: (i) to investigate the phylogenetic position of H. sankuruensis, an endemic from the central Congo Basin; (ii) to assess whether West African H. concolor constitutes more than one species; and (iii) to revise the species groups of Hyperolius Clade 2 (sensuPortik et al. 2019), based on new and recently published data.

Materials and methods

Taxon sampling

Three specimens of H. sankuruensis were investigated genetically. Two specimens from near the Lokoro River, Salonga National Park (South sector, Mai-Ndombe Province, DRC, 2.7531°S, 20.3781°E), male ZFMK 95069 and female ZFMK 95074, were together with a further 14 males and two females from the same locality identified as H. sankuruensis on the basis of a morphological comparison with the male holotype and female paratype (Supporting Information, Appendix S1). The female specimen ZMUC R771205 from Mabali near Lake Tumba (Équateur Province, DRC, 0.88°S, 18.13°E), originally labelled as Hyperolius sp. (Schiøtz 2006; Supporting Information, Fig. S1), was also recognized as H. sankuruensis based on a morphological comparison with the type and newly collected material (Supporting Information, Fig. S3). Collection acronyms follow Sabaj (2020). The only exception is IVB for the Institute of Vertebrate Biology of the Czech Academy of Sciences, Brno, Czech Republic. This institution is registered with the Global Biodiversity Information Facility (https://www.gbif.org/grscicoll), and IVB-H denotes the herpetological collection located in the Studenec research facility. All three genetic samples of H. sankuruensis were sequenced for a single mitochondrial marker, 16S rRNA (hereinafter as 16S), while the sample ZMUC R771205 was also sequenced for five nuclear markers: FIC domain-containing gene (FICD), KIAA2013 gene (KIAA2013), proopiomelanocortin gene (POMC), recombination activating protein 1 gene (RAG1), and tyrosinase gene (Tyr).

Our unpublished preliminary phylogenetic analyses showed close affinity of H. sankuruensis to species from the H. concolor group. Therefore, additional 16S and nuclear marker sequences were prepared for the species included in this group. These species are namely: H. bobirensis (one new specimen sequenced for 16S and nuclear markers, 1/1), H. concolor (22/7), H. zonatus (4/4), and Hyperolius sp. from Gabon. However, the latter was not available for nuclear marker sequencing. Among the newly generated sequences of H. concolor, all of its three presently described subspecies are included. Additional new sequences were also prepared for specimens of H. laurenti Schiøtz, 1967, H. schoutedeni Laurent, 1943, and divergent lineages of H. balfouri (Werner 1908) and H. cinnamomeoventris Bocage 1866, as these taxa/lineages have not yet been phylogenetically investigated using multilocus data. Homologous sequences (16S and the five nuclear markers) of 52 other species from Hyperolius Clade 2 (sensuPortik et al. 2019) were downloaded from GenBank online database to cover available species richness of the whole Clade 2. For details, see Supporting Information, Table S1.

DNA extraction, gene amplification

Total genomic DNA was extracted from the tissue samples using a commercial DNA extraction kit (GeneAll Biotechnology). One mitochondrial (mtDNA) and five nuclear (nDNA) markers were amplified using single-step and two-step (Shen et al. 2013) polymerase chain reactions (PCRs). A ~535 bp-long fragment of the 16S marker was amplified using the 16SL1 (forward) and 16SH1 (reverse) primers (Palumbi et al. 1991; modified). Nuclear markers were amplified using the primers: FICD F1, FICD F2 (forward) and FICD R1, FICD R2 (reverse) and were used to amplify FICD (529 bp; Shen et al. 2013); KIAA2013 F1, KIAA2013 F2 (forward), and KIAA2013 R1, KIAA2013 R2 (reverse) for KIAA2013 (540 bp; Shen et al. 2013); POMC-1 (forward; Wiens et al. 2005) and POMC-7 (reverse; Smith et al. 2005) for POMC (595 bp); RAG1 DCB1Fi (forward) and RAG1 DCB1R (reverse) for RAG1 (748 bp; Portik and Blackburn 2016); and Tyr 1C (forward) and Tyr 1G (reverse) for Tyr (532 bp; Bossuyt and Milinkovitch 2000). All newly generated sequences were deposited in the GenBank under the accession numbers PP536132–PP536166 for 16S and PP551135–PP551218 for nuclear markers. See Supporting Information, Table S1 for an overview of analysed sequences, and Supporting Information, Table S2 for sequences of the primers.

Phylogenetic analyses and dated phylogeny

Sequences were checked by eye and assembled using GENEIOUS PRIME v.2023.1.1 (https://www.geneious.com). MAFFT v.7.450 plug-in for GENEIOUS PRIME was used for aligning the sequences (Katoh et al. 2002, Katoh and Standley 2013). Uncorrected p-distances using the 16S mtDNA marker were generated using MEGA v.11.0.13 (Tamura et al. 2021) with the standard error estimation method set to bootstrapping for 500 replicates.

All maximum likelihood analyses (ML) were performed using IQ-TREE v.2.2.0 (Minh et al. 2020) and all Bayesian inference (BI) analyses were done in MrBayes v.3.2.7 (Ronquist et al. 2012). ModelFinder (Chernomor et al. 2016, Kalyaanamoorthy et al. 2017), as incorporated in IQ-TREE v.2.2.0, was used to find best-fit partitioned substitution models for both tree-building approaches. See Supporting Information, Table S3 for selected best-fit substitution models. All BI analyses were run twice with four Markov chains Monte Carlo for 10 million generations with sampling every 1000th generation and 25% ‘burn-in’. ML analyses were run using the standard nonparametric bootstrap for 100 replicates (Felsenstein 1985).

Hyperolius phantasticus (Boulenger 1899), H. adspersus Peters 1877, and Cryptothylax greshoffii (Schilthuis 1889) were selected as outgroups for BI and ML phylogenetic analyses. Hyperolius phantasticus is a representative of a sister-clade within the genus Hyperolius [Clade 1 sensuPortik et al. (2019)]. Hyperolius adspersus represents another major clade within the genus, the H. nasutus group. Cryptothylax greshoffii is a member of the tribe Cryptothylacini Dubois, Ohler and Pyron, 2021, which is closely related to the tribe Hyperoliini Laurent, 1943, which consists solely of the genus Hyperolius (Portik et al. 2019, Nečas et al. 2022). Hyperolius occidentalis Schiøtz, 1967 was selected as an outgroup for species delimitation analyses and time-calibrated coalescent-based analyses focusing on the H. concolor group. For summary of sampling data from outgroup samples see Supporting Information, Table S1.

Two coalescent-based analyses were set up in BEAUTi v.2.6.7 and performed in parallel using the StarBeast3 package (Douglas et al. 2022) for BEAST v.2.6.7 (Bouckaert et al. 2019) to infer dated species-tree for the H. concolor group. The analyses (StarBeast3) based on five nuclear genes utilized the uncorrelated relaxed clock (Drummond et al. 2006, Douglas et al. 2021) and the calibrated birth–death model (Heled and Drummond 2015). Two calibration points were used based on the dated phylogenomic tree of Portik et al. (2019). First for the split of the outgroup H. occidentalis at 13.0 (±2.0) Mya and second for the initial split in the H. concolor group at 7.5 (±2.5) Mya. Both analyses were run for 300 million generations with each 10 000th generation sampled. The first 10% of the sampled trees were discarded as a ‘burn-in’ which resulted in a total of 54 000 trees. TRACER 1.7.2 was used to inspect stationarity and effective sample size (ESS) values (Rambaut et al. 2018). Maximum clade credibility tree produced using LogCombiner v.2.6.7 and TreeAnnotator v.2.6.4 (Bouckaert et al. 2019) was visualized in FigTree v.1.4.4 (http://tree.bio.ed.ac.uk/software/figtree/) and edited using GNU Image Manipulation Program v.2.10.28 available at https://gimp.org. All best-fit substitution models were selected using ModelFinder of IQ-TREE 2.2.0. See Supporting Information, Table S3 for details on selected models for all above-mentioned analyses.

Species delimitation

The nuclear DNA dataset, composed of five nuclear markers, was analysed in BEAST 2.6.7 using the Standard BEAST template and coalescent constant population tree prior. The resulting ultrametric tree and tree topology were used as inputs to the multilocus delimitation methods Delineate 1.2.3 (Sukumaran et al. 2021) and BP&P 4.3 (Flouri et al. 2018), respectively. The A10 option of BP&P was used. Prior investigations of the dataset showed that setting the variable θ to 0.021 and the variable τ to 0.076 best reflects the tested dataset (Yang and Rannala 2010, Rannala and Yang 2013). Delineate 1.2.3 was run using the PYTHON programming language v.3.10.5 (Python Software Foundation, https://www.python.org). Two different approaches were employed in the Delineate analyses to assess how closely related species may affect delimitation results. The first analysis included samples of H. sankuruensis, H. zonatus, and H. bobirensis assigned to their respective species. In the second analysis, samples of the closely related H. zonatus and H. bobirensis (estimated divergence time 1.4 Mya, see Results) were assigned as a single ‘species’.

The mitochondrial DNA dataset of 16S sequences was analysed in BEAST 2.6.7 using the Standard BEAST template and implementing the coalescent constant population tree prior. The resulting ultrametric tree was used as an input to the single-threshold version of generalized mixed Yule coalescent (GMYC; online server available at https://species.h-its.org/gmyc/) species delimitation method (Fujisawa and Barraclough 2013). The ML phylogenetic tree of the 16S mtDNA produced in IQ-TREE was used as an input to the PTP (Poisson tree processes; https://species.h-its.org/) species delimitation method (Zhang et al. 2013). An alignment of 16S sequences was used as an input to ABGD (automatic barcode gap discovery; https://bioinfo.mnhn.fr/abi/public/abgd/abgdweb.html) and ASAP (assemble species by automatic partitioning; https://bioinfo.mnhn.fr/abi/public/asap/asapweb.html) delimitation methods (Puilliandre et al. 2012, 2021). In order to identify potential candidate species, we applied the 3% threshold in 16S genetic distances, as is regularly applied in anurans (Vieites et al. 2009). Best-fit substitution models for all BEAST analyses were selected using ModelFinder implemented in IQ-TREE v.2.2.0. See Supporting Information, Table S3 for selected substitution models.

Morphology

Snout–urostyle length (from snout tip to posterior edge of urostyle, SUL) and the following 13 taxonomically potentially important variables (Watters et al. 2016, Nečas et al. 2021) were taken with a digital calliper Insize 1108-150W to the nearest 0.1 mm: head width (HW), at greatest head width in close proximity to posterior edge of jaw; interorbital distance (IOD), shortest distance between upper eyelids; eye diameter (ED), between anterior and posterior corners of eye; eye–nostril length (ENL), from anterior corner of eye to centre of nostril; internarial distance (IND), between centres of nostrils; snout length (SL), from anterior corner of eye to snout tip; upper arm length (UAL), from body to outer edge of elbow; lower arm length (LAL), from elbow to outer edge of medially flexed wrist; hand length (HL), from outer edge of medially flexed wrist to tip of fourth manual digit; thigh length (TL), from centre of vent to outer edge of knee; lower leg length (LLL), from outer edge of knee to outer edge of tibiotarsal articulation; foot length (FL), from proximal edge of inner metatarsal tubercle to tip of fourth pedal digit; disc width (DW4), at greatest width of adhesive disc of fourth pedal digit. The above variables were measured in a total of 270 adults (191 males/79 females) of H. zonatus (21/3), H. bobirensis (14/7), H. sankuruensis (15/4), and H. concolor (141/65), including holotypes of H. argentophthalmus Ahl, 1931 (ZMB 36092), H. depressus Ahl, 1931 (ZMB 43554), H. guineensis Ahl, 1931 (ZMB 77464), H. hildebrandtii Ahl, 1931 (ZMB 8378), H. maximus Ahl, 1931 (ZMB 36113), H. narinus Ahl, 1931 (ZMB 36090), H. petersi Ahl, 1931 (ZMB 5573), H. pulcher Ahl, 1931 (ZMB 36088), H. togoensis Ahl, 1931 (ZMB 39009), and the lectotype of H. guttatus Peters, 1875 (ZMB 8378), to which the nomen H. hildebrandti is an objective junior synonym—all these latter nomina are presently treated as synonyms of H. concolor (Tillack et al. 2021, Frost 2024). The holotype of H. stenodactylus Ahl, 1931 (ZMB 85834) was also examined (Supporting Information, Fig. S7). The material is available in the collections of the Institute of Vertebrate Biology of the Czech Academy of Sciences, Brno/Studenec, Czech Republic (IVB), Zoological Research Museum Alexander Koenig—Leibniz Institute for the Analysis of Biodiversity Change, Bonn, Germany (ZFMK), Museum für Naturkunde—Leibniz Institute for Evolution and Biodiversity Science, Berlin, Germany (ZMB) and Natural History Museum of Denmark, University of Copenhagen, Copenhagen, Denmark (ZMUC).

Measured variables were statistically treated following Mosimann (1970) to remove the effect of body size; see Gvoždík et al. (2008) and Dolinay et al. (2021). A total of four datasets were analysed. Two datasets for the whole H. concolor group: Dataset 1 comprised of only males, and Dataset 2 of only females. Two datasets for H. concolor only: Dataset 3 comprised of only males, and Dataset 4 of only females. Multivariate analyses of variance (MANOVAs), which were used to assess variance among the taxa (including the candidate taxon H. cf. concolor) (Hand and Taylor 1987, Krzanowski 1988), were followed by analyses of variance (ANOVAs) for each of the 12 measured variables to find which of them show significant differentiation among the groups (Chambers et al. 1992). Tukey’s honest significant differences tests (Tukey’s HSD tests) were applied to results of the above ANOVAs to investigate group to group differences in specific morphological variables (Miller 1981, Yandell 1997). Principal component analyses (PCAs) and canonical variate analyses (CVAs) were used to investigate differentiation of the taxa in the morphospace. Both types of analyses were performed using the programming language R v.4.2.2 (R Core Team 2022) and R-packages ‘vegan’ v.2.6.4 (Oksanen et al. 2022) and ‘Morpho’ v.2.11 (Schlager 2017), respectively. Two linear discriminant analyses (LDAs) were performed on datasets composed of males and females of H. concolor concolor and H. concolor ibadanensis to identify the species of ungenotyped specimens collected in the Dahomey Gap. The analyses were performed using the R-package ‘MASS’ v.7.3.60 (Venables and Ripley 2002). Morphological variables identified by Tukey’s HSD tests as significantly differentiating the two subspecies were used. Body size variation of the taxa within the H. concolor group was visualized using the R-package ‘ggplot2’ v.4.2.3 (Wickham 2016).

Graphics

Maps (Fig. 1) were created in ArcGIS v.10.8.1 (Esri Inc., 2020) using occurrence data available from literature and museum collections (see Appendix S1 in Supporting Information). One occurrence point of H. sankuruensis from the Lotulu outpost, Yenge River, Salonga NP, Tshuapa Province, DRC (1.1468°S, 20.8001°E), was provided by Eli Greenbaum (University of Texas at El Paso, Texas, USA). Five localities of H. cf. concolor from Eshobi (5.7844°N, 9.3580°E), Nyang (5.9540°N, 9.4217°E), and Tinta (approx., 6.2720°N, 9.5110°E), South-West Region, and Bambili (approx., 5.9770°N, 10.2640°E) and Lake Wum (6.4078°N, 10.0523°E), North-West Region, Cameroon, were provided by Thomas Doherty-Bone (Royal Zoological Society of Scotland, Edinburgh, Scotland, UK). Land cover was visualized by implementing results of the GlobCover project (Arino et al. 2012). Country boundaries and shaded relief background were downloaded from https://naturalearthdata.com. All figures were composed and edited in GNU Image Manipulation Program v.2.10.28 available at https://gimp.org.

Nomenclatural note

The electronic version of this article will represent a published work according to the International Commission on Zoological Nomenclature (ICZN) and, therefore, the nomenclatural acts contained in the electronic version are effectively published under that Code from the electronic edition alone. This published work and the nomenclatural acts 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:004579A5-F55C-4429-97C6-A52A23F4501D.

Results

Phylogeny

Mitochondrial DNA

Results are presented with a focus on H. sankuruensis and the H. concolor group, but see Supporting Information, Figure S4 for further details of the mtDNA (16S) phylogeny of Hyperolius Clade 2. In H. sankuruensis, the sample ZMUC R771205 from near Lake Tumba forms a common clade with two samples from southwestern Salonga, DRC (1.00 BI posterior probabilities/100% ML bootstrap support; Supporting Information, Fig. S4). The uncorrected p-distance between the samples from the two geographic areas is 1.05% and 1.26%. For uncorrected pairwise distances between species and subspecies of the H. concolor group, see Supporting Information, Table S4. The two samples from Salonga differ from each other in a single site (0.21%).

The clade of H. sankuruensis is placed with intermediate support (1.00/67) in a sister-position to the undescribed Hyperolius sp. from south-western Gabon (Supporting Information, Fig. S4). The clade comprised of these two lineages forms a sister-lineage to the highly supported clade containing H. c. concolor (1.00/100). The clade composed of H. sankuruensis, Hyperolius sp., and H. c. concolor is sister to the clade consisting of H. c. guttatus, H. c. ibadanensis, and H. cf. concolor (0.98) in the BI analysis. The ML analysis places this clade with low support (35) in a sister-position to the clade containing H. zonatus and H. bobirensis. Samples of H. c. guttatus form a highly supported clade (1.00/97) in a sister-position (1.00/93) to H. c. ibadanensis (1.00/98). These two taxa are placed into a highly supported sister-relationship (1.00/100) with a clade composed of H. cf. concolor samples (1.00/100). All of the above-mentioned taxa and lineages form a common clade with H. zonatus and H. bobirensis, which together form a clade (0.96/58).

The H. concolor group forms a highly supported clade (1.00/96), which the BI analysis reveals to be sister (0.94) to the H. cinnamomeoventris group (as defined below; 1.00/100; Supporting Information, Fig. S4). The ML analysis places it with low support (58) in a sister-relationship to the clade containing the H. cinnamomeoventris group, H. balfouri group (as defined below; 1.00/96), and H. quinquevittatus (1.00/100).

Nuclear DNA

Results of the BI and ML phylogenetic analyses of five nuclear markers (Fig. 2) substantially differ from the mtDNA tree topology (Supporting Information, Fig. S4), mainly in combining all H. concolor subspecies and the candidate taxon (H. cf. concolor) into a single highly supported clade. Both phylogenetic analyses place H. sankuruensis in the common highly supported clade with H. concolor, H. zonatus, and H. bobirensis (1.00/90). Hyperolius sankuruensis is in a sister-position to the three other taxa. Although the three taxa form together only a weakly supported clade (0.63/29), Hyperolius zonatus and H. bobirensis form together a highly supported common clade (1.00/100) and H. concolor (1.00/81) consist of two main clades. One corresponds to H. concolor concolor (1.00/57), while the other clade (1.00/85) is formed by H. c. ibadanensis in a sister-position to a clade (0.99/60) comprising H. c. guttatus (1.00/89) and H. cf. concolor. It is not possible to reveal the phylogenetic position of Hyperolius sp. from Gabon as no nuclear markers were available.

In the BI analysis, the H. concolor group is placed in a polytomy (with high support, 1.00) with the H. montanus group (1.00/87), H. quinquevittatus, and a clade (1.00/76) consisting of the H. balfouri group (1.00/81) and the H. cinnamomeoventris group (1.00/99). The ML analysis places the H. concolor group in a sister-position but with low support (39) to a clade (49) comprised of the H. balfouri group, the H. cinnamomeoventris group, and H. quinquevittatus.

Species delimitation

Mitochondrial DNA

Single-locus species delimitation analyses performed using the 16S mtDNA marker on the H. concolor group provide relatively similar results (Fig. 3A). The ABGD and ASAP analyses give the best scores to delimitation schemes that separate all presently recognized and candidate species and subspecies, although the ABGD best-scheme proposes to split H. c. concolor into several ‘species’. The PTP analysis shows similar results with the exception of H. c. guttatus, H. c. ibadanensis, and H. cf. concolor combined together, while H. c. concolor split into more ‘species’. The GMYC analysis supports only three groupings. The first includes H. c. concolor, H. sankuruensis, and Hyperolius sp. The second groups the two remaining subspecies of H. concolor and H. cf. concolor, similarly to PTP. The third group consists of H. zonatus and H. bobirensis. The 3% threshold separates all currently recognized and candidate taxa, except of H. c. guttatus and H. c. ibadanensis, which show uncorrected p-distance in 16S smaller than 3% (Supporting Information, Table S4).

Nuclear DNA

The multilocus species delimitation analysis BP&P, performed on five nuclear markers, supports the species status of H. sankuruensis, H. zonatus, and H. bobirensis (Fig. 3B), but splits H. concolor into three groups. One group includes H. c. concolor, the second group includes H. c. guttatus, and the third group consists of H. c. ibadanensis and H. cf. concolor. Two different analyses were run using the Delineate software (Fig. 3B). The first analysis, in which H. sankuruensis, H. zonatus, and H. bobirensis samples were assigned to their respective species, shows similar results to BP&P (Delineate 1 in Fig. 3B). However, similarly to ABGD and PTP (mtDNA), BP&P further splits some taxa into more ‘species’ (H. c. concolor and H. c. guttatus). The second analysis, which had the closely related H. zonatus and H. bobirensis assigned as a single ‘species’, supported all lineages of H. concolor (subspecific and candidate taxa) as a single species (Delineate 2 in Fig. 3B).

Morphology

The average body size (SUL in mm) in males/females of all analysed taxa is as follows (Fig. 4A): H. c. concolor (26.7/32.3), H. c. ibadanensis (27.2/31.9), H. cf. concolor (30.6/34.9), H. c. guttatus (27.0/30.7), H. zonatus (26.0/37.2), H. bobirensis (31.0/44.4), and H. sankuruensis (28.8/35.1). MANOVA analyses of the four datasets show all recognized and candidate taxa across the four datasets to be significantly different from each other (Supporting Information, Table S5).

Analyses of variance (ANOVAs) of the Dataset 1 (H. concolor group, males) found most of the variables significant in differentiating the four species, with the exception of internarial distance (IND) and width of disc of fourth finger (DW4; Supporting Information, Table S5). Further investigation of individual ANOVAs using Tukey’s HSD tests revealed that H. concolor differs significantly from the other species in head width (HW) and lower leg length (LLL; Supporting Information, Table S5). Hyperolius concolor further differs from H. zonatus and H. bobirensis in head shape variables (ED, ENL, and SL), and from H. bobirensis further in upper arm length (UAL) and hand length (HL). Hyperolius zonatus and H. bobirensis differ from each other only in lower arm (LAL) and lower leg lengths (LLL). Hyperolius sankuruensis differs from H. zonatus in eye diameter (ED), from H. zonatus and H. bobirensis in eye to naris length (ENL), snout length (SL) and upper arm length (UAL), from H. bobirensis and H. concolor in foot length (FL), and from all the other three species in interorbital distance (IOD). PCA of the Dataset 1 shows an overlap of portion of the morphospace of H. concolor with the other three species, while H. zonatus overlaps with the other two species, but H. bobirensis and H. sankuruensis do not (Fig. 4B). Head shape variables (HW, ED, ENL, and SL) contribute most to the principal component 1 (PC1 22.1%), and hindlimb variables (TL, LLL, and FL) contribute most to the principal component 2 (PC2 17.5%). CVA of the Dataset 1 finds head width (HW), interorbital distance (IOD), eye–naris length (SL; CV1 79.1%), internarial distance (IND), and hand length (HL; CV2 16.3%) to be the most differentiating variables (Supporting Information, Fig. S5A). See Supporting Information, Table S5 for more details.

The Dataset 2 (H. concolor group, females) of the four species shows significant difference only in four head shape variables (HW, IOD, IND, and SL) and three limb variables (HL, LLL, and DW4), when tested by ANOVAs. Hyperolius concolor significantly differs from H. sankuruensis in head width (HW), and from H. bobirensis in snout length (SL), hand length (HL), and lower leg length (LLL). Hyperolius sankuruensis and H. bobirensis differ only in snout length (SL), and H. sankuruensis and H. zonatus vary only in interorbital distance (IOD). Females of H. zonatus are not distinguished from H. concolor or H. bobirensis in any variable. PCA of the Dataset 2 shows that the morphospaces of species overlap substantially, with H. bobirensis being the most differentiated species (Fig. 4C). Eye–naris length (ENL), snout length (SL), and foot length (FL) contribute most to PC1 (23.4%), while hindlimb variables (TL and DW4) contribute most to PC2 (14.7%). CVA plots the variables of head shape (HW, IOD, ED, and IND) and forelimbs (UAL and LAL) as the most differentiating variables (Supporting Information, Fig. S5B). See Supporting Information, Table S5 for more details.

Analyses of variance (ANOVAs) of the Dataset 3, males of three recognized subspecies and the unnamed lineage of H. concolor (H. concolor s.l.), show that eye–naris length (ENL) and variables of limbs (UAL, LAL, TL, FL, and DW4) bear no significant difference among the recognized and candidate taxa. Tukey’s HSD tests of the ANOVA results show that H. c. concolor is significantly different from H. c. ibadanensis in head variables (HW, IOD, IND, and SL) and hand length (HL), from H. c. guttatus in head variables (IOD, ED, and SL) and lower leg length (LLL), and from H. cf. concolor in only two variables (HW and LLL). Hyperolius c. ibadanensis shows significant difference from H. c. guttatus in the shape of head (HW, IOD, and ED) and two limb variables (HL and LLL), and from H. cf. concolor in only two variables (IOD and LLL). Hyperolius c. guttatus displays significant difference from H. cf. concolor in only one variable (IOD). Similar to previous PCAs, the PCA of the Dataset 3 (Fig. 4D) shows that head shape (IOD, ENL, IND, and SL) and limb (LAL and LLL) variables contribute most to PC1 (22.2%) and PC2 (16.0%), respectively. The PCA again shows that the morphospaces of all recognized and candidate taxa overlap to a large extent, the only exception being H. c. guttatus, for which there is only marginal overlap. CVA identifies head (HW, IOD, ENL, IND, and SL) and forelimb (UAL and HL) variables as the most contributing and, similarly to the PCA, identifies H. c. guttatus as the taxon most distinct from the other two subspecies and the unnamed lineage (Fig. 4E). In addition, H. c. concolor is also shown to be distinct from H. c. ibadanensis and H. cf. concolor. See Supporting Information, Table S5 for more details.

ANOVAs of the Dataset 4 (H. concolor s.l., females) show that only head width (HW) exhibits significant difference among the recognized subspecies and the unnamed lineage (Supporting Information, Table S5). A Tukey’s HSD test further reveals that head width (HW) is significantly different only when H. c. ibadanensis is compared to H. c. concolor or to H. cf. concolor (Supporting Information, Table S5). PCA shows head shape (IOD, ENL, and SL) and thigh (TL) and lower leg (LLL) lengths as most contributing variables, and also identifies a large overlap in the morphospaces of the recognized and candidate taxa (Supporting Information, Fig. S5C). According to the CVA, head variables (HW, IOD, ED, and SL) and upper arm length (UAL), hand length (HL) and width of disc of 4th finger (DW4) differentiate the taxa most (Supporting Information, Fig. S5D), with H. c. ibadanensis and the single analysed female of H. c. guttatus differing from H. c. concolor and H. cf. concolor (the latter two overlap to a large extent). See Supporting Information, Table S5 for more details.

The linear discriminant analysis, trained on a dataset of males of H. c. concolor and H. c. ibadanensis (LDA 1), was used to classify three male specimens collected in the Dahomey Gap (presumed contact zone of these two taxa). Five variables identified by a Tukey’s HSD test performed on these two taxa were used in the discriminant formula (Supporting Information, Tables S5, S6). The LDA analysis identifies two of the three specimens as belonging to H. c. concolor, probabilities of 70.9% and 84.8% (Supporting Information, Table S7). One specimen is assigned to H. c. ibadanensis, but with lower probability of only 54%. Similarly, six female specimens collected in the Dahomey Gap were subjected to a discriminant analysis, trained on a dataset of females of H. c. concolor and H. c. ibadanensis (LDA 2). However, a Tukey’s HSD test revealed that only a single variable (head width, HW) shows significant difference between these two taxa (Supporting Information, Table S5). The analysis classifies all six females as H. c. concolor, probabilities between 64.9% and 97.7% (Supporting Information, Table S7). Basic descriptive statistics of the measured variables are given in Table S8 in the Supporting Information.

Dated phylogeny

A dated analysis (StarBeast3; Fig. 5), based on five nuclear DNA genes, places H. sankuruensis in a common clade with H. zonatus and H. bobirensis, albeit with low support (0.47 posterior probabilities), and estimates the divergence time of H. sankuruensis to approximately 5.0 (2.1–7.8 95% HPD) Mya, and H. zonatus and H. bobirensis to approximately 1.4 (0.02–3.8) Mya. The H. zonatus–H. bobirensis–H. sankuruensis clade is sister to the highly supported clade (0.98) composed of H. concolor s.l. The split between the two clades is estimated at 6.8 (4.5–9.3; calibrated to 7.5) Mya. The divergence between H. c. concolor and the clade (0.91) comprising H. c. guttatus, H. cf. concolor, and H. c. ibadanensis is estimated to have occurred around 3.3 (1.6–5.5) Mya. The split between H. c. guttatus and the clade containing H. cf. concolor and H. c. ibadanensis (0.79) is estimated to have happened around 2.2 (0.8–3.9) Mya and between the latter two around 0.6 (0.0001–2.2) Mya.

Discussion

Phylogenetic position of Hyperolius sankuruensis: linking the central Congo and West Africa

Our study unambiguously reveals the central Congolian forest endemic H. sankuruensis as a member of a clade containing H. concolor (including the subspecies H. c. guttatus, H. c. ibadanensis, and the candidate taxon H. cf. concolor), H. zonatus, and H. bobirensis. Mitochondrial DNA analyses also supported the undescribed Hyperolius sp. from south-western Gabon as a representative of this clade (Figs 2, 3, 5; Supporting Information, Fig. S4). Except for the latter, all other species are known primarily from West Africa. Previous to our study, it was assumed that H. sankuruensis was potentially related to either another central Congolian endemic, Congolius robustus (Schiøtz 2006), or to either of the Central African species H. platyceps or H. langi, or to the eastern Congolian montane species H. diaphanus and H. frontalis (Laurent 1979, 1981). None of these proposed relationships have been confirmed, and our discovery of a rather unexpected phylogenetic relationship highlights a biogeographical link between the central Congo and West Africa. However, the exact phylogenetic position of H. sankuruensis within the H. concolor group (as defined in this study) remains unclear. Different analytical approaches have provided different results, although the undescribed species Hyperolius sp. 1 or sp. B from the Doudou Mountains of Gabon in western Central Africa (Burger et al. 2004, Veith et al. 2009), appears to be the closest relative of H. sankuruensis (Fig. 3A; Supporting Information, Fig. S4). Unfortunately, the undescribed Gabonese species was not available for nDNA analysis, and its inclusion in the future may provide a better resolution of the interrelationships among species of the H. concolor group. However, based on the depth of divergence in mtDNA between the undescribed Gabonese Hyperolius sp. and H. sankuruensis, it seems likely that these two species diverged during the Pleistocene (Fig. 3A), similar to the H. zonatus–H. bobirensis pair (cf. Fig. 5). The divergence may have occurred in the context of climatic oscillations and allopatric speciation, linked to geographically isolated forest refugia, which were probably located in the central Congo and the area around the Doudou Mountains, in present-day Gabon (Colyn et al. 1991, Sosef 1994, Burger et al. 2004).

In addition to the undescribed Gabonese species, the species-tree analysis reveals the Upper Guinean forest species H. zonatus and H. bobirensis as potential closest relatives of the central Congolian forest species H. sankuruensis (Fig. 5). Interestingly, the potentially closest relative of H. sankuruensis, the undescribed Hyperolius sp. from Gabon, is also morphologically similar to H. zonatus and H. bobirensis (Burger et al. 2004). The time to the most recent common ancestor of H. sankuruensis, H. zonatus, and H. bobirensis is estimated to be approximately at the Miocene/Pliocene boundary, when it is thought that Late Miocene cooling triggered gradual aridification with forest retreat and the spread of open habitats (Couvreur et al. 2021). This process may have separated populations of the common ancestor of this clade into the Upper Guinean and Congolian forests, respectively.

The phylogenetic analyses confirm that the newly collected specimens of H. sankuruensis from Salonga, and the single female collected at Lake Tumba are conspecific (~1.2% difference in the 16S mtDNA fragment). The latter, Schiøtz (2006) referred to as a potentially undescribed Hyperolius species. These two localities extend the known range of H. sankuruensis, previously known only from the type locality in the Sankuru Province, westward. However, all presently known localities lie within the Central Congolian Lowland Forests ecoregion (Burgess et al. 2004, Dinerstein et al. 2017). None of the fieldwork conducted north of the Congo River reports the occurrence of H. sankuruensis (e.g. Jackson and Blackburn 2007, Jackson et al. 2007, Badjedjea et al. 2015, 2016, Masudi et al. 2019). It is, therefore, possible that the area of occurrence of H. sankuruensis is restricted to the area south of the wide arc of the Congo River, similar to that of Congolius robustus (Nečas et al. 2021).

West African Hyperolius zonatus and Hyperolius bobirensis

The sister-relationship between the two West African forest species H. zonatus and H. bobirensis is supported in all analyses. These two species are found relatively deeply divergent in mtDNA (~4.1% uncorrected distance in 16S; Fig. 3A; Supporting Information, Table S4, Fig. S4), whereas the divergence is relatively shallow in nDNA dated to c. 1.4 Mya (Figs 2, 3B, 5). Portik et al. (2019) estimated this divergence time to be >5 Mya, but this deeper divergence was probably estimated because the analysis included only mtDNA of H. zonatus. The divergence time 1.4 Mya corresponds to a period of increased aridity on the African continent between ~1.5 and 1 Mya, which pushed rain forests into a series of small refugia (Fig. 5; Mayr and O’Hara 1986, deMenocal 2004, Plana 2004, Ségalen et al. 2007, Trauth et al. 2009). The areas presently occupied by the two allopatric forest species correspond to areas of putative forest refugia (Maley 1987, 1996). The same pattern has been observed in other West African amphibians as well (Rödel et al. 2012).

Our morphological analyses show that H. zonatus and H. bobirensis have similar body shapes in both sexes, despite the considerable difference in body size, with H. bobirensis being the largest species of the H. concolor group (Fig. 4A–C; Supporting Information, Fig. S5A, B, Tables S5, S8). Hyperolius zonatus has also relatively larger head proportions than H. bobirensis (Supporting Information, Fig. S6). Although Schiøtz (1967) already discussed the possible subspecies status of these two taxa, he ultimately preferred to distinguish them at the species level. Despite their relatively recent evolutionary history and their general morphological and ecological similarity, we agree with the species-level distinction, especially with respect to morphological characters that were not included in our analyses (e.g. the small male gular gland and the almost absent vocal sac in H. zonatus vs. large gular gland and some dilatable skin in H. bobirensis), some acoustic differences, and particularly their geographically isolated distributions, suggesting that the two taxa represent vicariant species (Schiøtz 1967, 1999, Channing and Rödel 2019).

Phylogeny and mitonuclear discordance in Hyperolius concolor

The mitochondrial phylogeny (Fig. 3A; Supporting Information, Fig. S4) of mostly West African H. concolor differs from the nuclear phylogeny (Figs 2, 3B) in two main aspects. First, the species is paraphyletic in respect to the position of H. sankuruensis and Hyperolius sp. from Gabon in mtDNA (although the topology has only intermediate to low support), whereas it is monophyletic with high support in nDNA. The mtDNA paraphyly of the predominantly West African H. concolor in relation to the central Congolian and Gabonese representatives can be explained in two ways. One is that the observed mtDNA phylogeny indeed mirrors mitochondrial evolutionary history. The hypothesis would be that mitochondrial capture from a distantly related species occurred in the past, and this ancient lineage has persisted in western or eastern populations. For a similar scenario in lizards, see Gvoždík et al. (2023). An alternative explanation is that mtDNA paraphyly is an analytical artefact caused by the analysis of a relatively short fragment of mtDNA. This would need to be verified in the future, using a longer mtDNA fragment, multiple fragments, or entire mitogenomes. In both datasets, however, H. concolor is divided into two main clades. One consists exclusively of individuals of the nominotypical subspecies H. c. concolor, occurring west of, and probably within, the Dahomey Gap—a savanna region extending from the central Ghana into Togo and Benin, being surrounded by forested areas (see below and Supporting Information, Table S6). The second clade consists of the two subspecies H. c. ibadanensis and H. c. guttatus, and the candidate taxon H. cf. concolor, all occurring east of the Dahomey Gap (Fig. 1).

The second mitonuclear discordance is related to the position of H. cf. concolor, which is relatively deeply divergent in mtDNA (>3%; Supporting Information, Table S4), but with only a shallow divergence in nDNA (Figs 2, 3). The mtDNA analysis is based on three specimens from two distant hilly areas in south-eastern Nigeria, and one specimen from south-western Cameroon, which together form a clade sister to H. c. guttatus and H. c. ibadanensis. In nDNA, only one identical specimen from south-western Cameroon (near Mamfe) was available. This was previously identified as H. c. ibadanensis (Portik et al. 2019), but in mtDNA it clearly belongs to H. cf. concolor (sensuNneji et al. 2019a, b, Rahman et al. 2020). As in the case of mitonuclear discordance (see above), this can be explained as an ancient mitochondrial introgression from a distantly related species, with this ancient mtDNA lineage persisting in the population geographically restricted to the north-western slopes of the Cameroon Volcanic Line. Otherwise, however, this population (H. cf. concolor) would not differ significantly from H. c. ibadanensis in the nuclear genome (Figs 3B, 5). Alternatively, the Mamfe individual may be a hybrid between H. c. ibadanensis and H. cf. concolor, from which mitochondrial introgression may have occurred. Hyperolius cf. concolor may represent a distinct species, but apart from the presumably hybrid individual from Mamfe, it has not yet been investigated in nDNA. Similarly, Portik et al. (2017) found a genetically distinct population in the arthroleptid frog Scotobleps gabonicus Boulenger 1900 north of the Cross River, but individuals south of the river showed a similar pattern of mitochondrial introgression from the north. Further research with denser sampling in this area is needed to explain the evolutionary relationships between H. c. guttatus, H. c. ibadanensis, and H. cf. concolor. Considering the relatively deep divergences among several lineages, the complex evolutionary history with some mitonuclear discordances, and the results of the species delimitation analyses (Fig. 3 and see below), we hereafter consider ‘H. concolor’ as a species complex.

The split between the H. concolor complex living predominantly in semi-open bushland habitats and the forest species in the Late Miocene at ~6.8 Mya (Fig. 5) coincides with the abrupt change in tree cover in West Africa caused by a rapid decline in humidity between approximately 7 to 5 Mya (Bonnefille 2010, Couvreur et al. 2021). This could have initiated a shift in habitat preferences in the ancestor of the H. concolor complex towards more open habitats. Divergence between the western H. concolor concolor and the two eastern subspecies and the candidate taxon has been estimated to occur around 3.3 Mya, twice as deep as the estimated divergence between H. zonatus and H. bobirensis (Fig. 5). Portik et al. (2019) estimated that this divergence occurred even earlier, around 6 Mya. Repeated forest/savanna expansions and retreats between ~5.5 and ~2.7 Mya (Brouat et al. 2009, Bonnefille 2010, Couvreur et al. 2021) may have isolated some populations of the ancestor of the H. concolor complex and triggered diversification into the subspecies that exist today.

Taxonomy of the Hyperolius concolor complex

Species delimitation analyses performed on the H. concolor group, on both mitochondrial and nuclear datasets, generally support all currently recognized species, including the candidate species Hyperolius sp. from Gabon (Fig. 3). For H. concolor, Frétey et al. (2014) hypothesized that H. c. concolor and H. c. guttatus may represent distinct species, and mentioned the need for a detailed investigation of H. c. ibadanensis, because this subspecies appeared to be morphologically closer to H. c. guttatus. A majority of our species delimitation analyses support H. c. concolor, H. c. guttatus, and H. c. ibadanensis as three separate species. Morphological analyses also show significant differences among the current and candidate taxa (Fig. 4D, E; Supporting Information, Fig. S5C, D, Table S4). Some species delimitation analyses provide support for the subdivision of some lineages of H. c. concolor and H. c. guttatus. However, given the relatively small morphometric differentiation, and keeping a rather conservative approach in the current phase of research, we prefer the results of species delimitation analyses supporting the current taxa (i.e. BP&P in multilocus nDNA analyses; ASAP and 3% threshold in single-locus mtDNA analyses). Therefore, we propose to elevate the subspecies of H. concolor to full species as follows:

Hyperolius concolor (Hallowell, 1844) s.s. (Benin, Togo, Ghana, Burkina Faso, Côte d’Ivoire, Liberia, Sierra Leone, Guinea);

Hyperolius guttatus Peters, 1875 stat. nov. (Cameroon, possibly coastal south-eastern Nigeria east of the Cross River), LSID: urn:lsid:zoobank.org:act:C2BB82F9-F679-4951-9CA5-517CD4BF40F9;

Hyperolius ibadanensis Schiøtz, 1967 stat. nov. (Nigeria, possibly Benin), LSID: urn:lsid:zoobank.org:act:1F99FAA9-6529-492E-86A0-1C67A9C4E5F2.

Taking into account the mitonuclear discordance in the candidate taxon H. cf. concolor and the need for further study to understand the evolutionary relationships among populations in south-eastern Nigeria and south-western Cameroon (see above), and given the close relationship in nDNA with H. ibadanensis in the Mamfe individual, we tentatively propose to name these populations Hyperolius cf. ibadanensis (hilly areas of south-eastern Nigeria and adjacent Cameroon). For distribution ranges, see Figure 1.

A detailed synonymy of H. guttatus was presented by Frétey et al. (2014) under the trinomen Hyperolius concolor guttatus. No synonyms are available for H. ibadanensis. Yet, we would like to point out that available names for H. cf. ibadanensis exist. Rappia sordida Fischer, 1888 with type locality ‘Kamerun’ (Cameroon) was previously applied to populations of ‘H. concolor’ in the vicinity of Mamfe as H. sordidus (Parker 1936, Sanderson 1936). Another possibility is Hyperolius maximus Ahl, 1931 with type locality ‘Ossidinge’ (=Mamfe). However, we recommend keeping both nomina conservatively listed as synonyms of H. guttatus for the time being, given the uncertainty of the exact geographical origin of Rappia sordida and the fact that the Mamfe population may be of a hybrid origin. The geographical origin of Rappia sordida needs to be properly investigated using historical records and/or ancient DNA sequencing. However, it is possible that the types originate from the Douala region, as they were sent to the ‘Lübecker Museum’ by Johannes Voss, who headed a factory in Douala in 1875–91 (Fischer 1888, Todzi 2023), and thus it may indeed be a synonym of H. guttatus, rather than an available name for H. cf. ibadanensis.

Note: Schiøtz (1967) studied what is now H. concolor s.s., H. ibadanensis, and his then ‘H. concolor ssp. indet’. The latter now corresponds partly to H. guttatus and partly to H. cf. ibadanensis. He personally field-studied a population in Osomba, Nigeria, which now corresponds to H. cf. ibadanensis.

Morphologically, H. cf. ibadanensis is the largest member within the H. concolor complex, and H. guttatus is the most distinct taxon in body shape, whereas H. concolor, H. ibadanensis, and H. cf. ibadanensis have relatively similar body shapes (Fig. 4; Supporting Information, Figs S5, S6, Table S4). The results of our analyses are mainly consistent with those of Frétey et al. (2014), who focused particularly on H. concolor and H. guttatus. Also consistent with our findings, Amiet (2012) found that the population from Bafut in Cameroon, now probably H. cf. ibadanensis, exhibited larger body size and had a less wide head than present-day H. guttatus from the lowland plains of southwestern Cameroon.

Schiøtz (1967) found no substantial differences in parameters of advertisement calls between H. concolor, H. ibadanensis, and H. cf. ibadanensis. Similarly, Amiet (2012) did not detect any acoustic differences between what is now H. guttatus and H. cf. ibadanensis. However, it is known that closely related species of frogs that live in allopatry may have only slight, if any, differences in their advertisement calls (e.g. Schneider 2004, Gvoždík et al. 2015, Köhler et al. 2017).

Hyperolius concolor and H. ibadanensis presumably meet in the Dahomey Gap, which is a known distribution barrier for many animal and plant taxa and may have played a role in speciation of some closely related species in other anuran amphibians [e.g. Leptopelis macrotis and L. millsoni (Jaynes et al. 2022)]. However, the extent of contact between H. concolor and H. ibadanensis in the Dahomey Gap is unknown. Rödel et al. (2007) identified specimens collected in the Lokoli and Lama Forests of Benin, located in the Dahomey Gap, as H. (c.) ibadanensis. However, this assignment was based on colour pattern only, and the specimen ZMB 80719 collected in Lokoli Forest and analysed in this study was assigned to a common clade with H. concolor s.s. in both mitochondrial and nuclear DNA analyses (Figs 2, 3; Supporting Information, Fig. S4). Morphological linear discriminant analyses of specimens collected in the Dahomey Gap classified most of them as H. concolor, whereas only a single specimen from Lama Forest was approximately half assigned to H. concolor and half to H. ibadanensis, and is thus a potential hybrid (Supporting Information, Table S7). These results suggest that the population inhabiting the Dahomey Gap belongs primarily to H. concolor. This may be supported by the reported occurrence of this species further north in drier savanna (Nago et al. 2006, Ayoro et al. 2020), suggesting that H. concolor could be more tolerant of drier conditions than, for example, H. ibadanensis. However, further sampling in the region and genetic analyses are needed to clarify whether these two taxa have a contact zone and potentially hybridize.

Hyperolius stenodactylus, known only from the holotype from southern Cameroon (Supporting Information, Fig. S7), which is usually reported as valid (e.g. Frost 2024), but which Perret (1966) identified as a possible synonym of H. concolor, we rather consider as a possible synonym of H. acutirostris Buchholz and Peters, 1875, based on examination of the holotype. However, we hesitate to draw a definitive conclusion and retain this taxon with doubts as valid, pending availability of new material and/or results of ancient DNA analysis. See Supporting Information, Appendix S2 for more details.

Species groups in Hyperolius Clade 2 and discrepancies in GenBank data

Based on mtDNA, Dehling and Sinsch (2019) divided the genus Hyperolius into six species groups, three of which corresponded to Hyperolius Clade 2 sensuPortik et al. (2019). Ernst et al. (2021) established eight species groups (A–H) across the genus, again based on mtDNA, seven of which can be assigned to Clade 2 (B–H). The species groups proposed in this study are based on the results presented here (multilocus nDNA, and mtDNA), as well as on the results of the phylogenomic analysis of Portik et al. (2019) and works on specific species groups (see Appendix  1 for details). All groups are monophyletic clades that are recognizable in analyses of both nuclear and mitochondrial datasets and that have high supports in most cases. The only exceptions are the H. occidentalis group, which has only moderate support in phylogenetic analyses of the nuclear dataset (Fig. 3), and the H. obstetricans group, with moderate support in the 16S dataset (Supporting Information, Fig. S4). Names were assigned to species groups based on the oldest available name of the included species. The 11 species groups of Hyperolius Clade 2 are as follows (in alphabetical order): Hyperolius balfouri group; H. castaneus group; H. cinnamomeoventris group; H. concolor group; H. langi group; H. montanus group; H. mosaicus group; H. occidentalis group; H. obstetricans group; H. platyceps group; H. spinigularis group. Hyperolius quinquevittatus Bocage 1866, H. koehleri Mertens 1940, and H. tuberilinguis Smith 1849 were not assigned to any species group. See Appendix  1 for more details on included species and distributions of the species groups.

Our phylogenetic analyses also revealed a number of discrepancies in some GenBank sequences. See Appendix  2 for details.

Conclusion

The first phylogenetic investigation of mitochondrial and nuclear DNA of Hyperolius sankuruensis, an endemic from the central Congo Basin, shows its close but unresolved relationship to the West African Hyperolius concolor species group, including an undescribed species from the Doudou Mountains in south-western Gabon. Further focus on H. concolor and its three subspecies reveals their profound molecular differences. Based on these results, together with morphological investigations, Hyperolius guttatus stat. nov. and H. ibadanensis stat. nov. are elevated from subspecies to species. It is further highlighted that the area around the central Cameroon Volcanic Line is inhabited by a potentially undescribed taxon, tentatively named H. cf. ibadanensis. Furthermore, based on current and published research eleven species groups are proposed for Hyperolius Clade 2 (sensuPortik et al. 2019).

Acknowledgements

We would like to thank P.R. Møller, M.D. Scherz, and D. Klingberg Johansson (Natural History Museum of Denmark, University of Copenhagen), C. Koch, M. Flecks, and U. Bott (Zoological Research Museum Alexander Koenig, Bonn), F. Tillack (Museum für Naturkunde, Berlin), D. Meirte and G. Cael (Royal Museum for Central Africa, Tervuren), and E.B. Fokam (University of Buea) for assistance and/or providing access to the material. T. Doherty-Bone and C. Liedtke are acknowledged for providing photographs of H. cf. ibadanensis and Hyperolius ukaguruensis, respectively. T. Doherty-Bone and E. Greenbaum are acknowledged for providing distribution data for H. cf. ibadanensis and H. sankuruensis, respectively.

Conflict of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding

This work was funded by the Czech Science Foundation (grant number 23-07331S) and Ministry of Culture of the Czech Republic (DKRVO 2024–2028/6.I.a, National Museum of the Czech Republic, 00023272). TN was further supported by the Department of Botany and Zoology, Faculty of Science, Masaryk University, Brno, Czech Republic (MUNI/A/1348/2022) and Synthesys + project of the European Commission (DE-TAF-057).

Data availability

The data supporting this article are available in the GenBank online database (https:/ncbi.nlm.nih.gov/genbank/), and can be accessed with following accession numbers: PP536132–PP536166 for 16S; PP551135–PP551151 for FICD; PP551152–PP551168 for KIAA2013; PP551202–PP551218 for POMC; PP551169–PP551184 for RAG1; PP551185–PP551201 for Tyr.

Appendix 1

Species groups in Hyperolius Clade 2

A list of the 11 species groups of Hyperolius Clade 2 established in this study, with included species, is provided below.

Hyperolius balfouri group

Hyperolius concolor superspecies of Schiøtz (1975) [partim]; Group II of Amiet (2012) [partim].

Contents (based on genetic data): H. balfouri (Werner 1908)—H. balfouri balfouri, H. balfouri viridistriatus Monard 1951; H. kivuensis Ahl, 1931; H. schoutedeni Laurent, 1943 (Bamba-Kaya et al. 2019, Portik et al. 2019; this study).

Area of occurrence: Central Africa from western Cameroon in the North-West, South Sudan and south-western Ethiopia in the North-East, Malawi in the South-East, to central Angola in the South-West, with the exception of the Lower Guinean Forests.

Hyperolius castaneus group

Hyperolius castaneus group of Dehling and Sinsch (2019); Group D of Ernst et al. (2021).

Contents (based on genetic data): H. castaneus Ahl, 1931; H. constellatus Laurent, 1951; H. cystocandicans Richards and Schiøtz 1977; H. discodactylus Ahl, 1931; H. frontalis Laurent 1950; H. jackie Dehling 2012; H. lateralis Laurent 1940 (Greenbaum et al. 2013, Dehling and Sinsch 2019, Portik et al. 2019, Ernst et al. 2021).

Area of occurrence: Eastern Democratic Republic of the Congo, Rwanda, Burundi, Uganda, western Tanzania, and western Kenya. Hyperolius cystocandicans is known from central Kenya.

Other potential members: Hyperolius atrigularis Laurent 1941 and H. kibarae Laurent, 1957 are possibly related to this group (Laurent 1957, Schiøtz 1999). Laurent (1951), based on morphological data, found H. leleupi Laurent, 1951 to be closely related to H. castaneus (and H. atrigularis). Laurent (1972) suggested that the single specimen of H. xenorhinus Laurent, 1972 from Virunga NP, eastern Democratic Republic of Congo may be an aberrant individual of H. discodactylus. Schiøtz (1999) suggested that further investigation of H. frontalis, H. diaphanus Laurent, 1972, H. chrysogaster Laurent 1950, H. leucotaenius Laurent 1950, and H. xenorhinus might show them conspecific.

Hyperolius cinnamomeoventris group

Group II of   Amiet (2012) [partim]; Hyperolius cinnamomeoventris complex of Bell et al. (2019); H. cinnamomeoventris group of Dehling and Sinsch (2019); Group H of Ernst et al. (2021).

Contents (based on genetic data): H. cinnamomeoventris Bocage 1866; H. drewesi Bell 2016; H. molleri (Berdriaga 1892); H. olivaceus Buchholz and Peters, 1876; H. thomensis Bocage 1886; H. veithi Schick, Kielgast, Rödder, Muchai, Burger and Lötters, 2010 (Schick et al. 2010, Bell et al. 2015, 2019, Portik et al. 2019).

Area of occurrence: Central Africa from central Cameroon in the North-West, the islands of the Gulf of Guinea in the West, South Sudan and western Kenya in the North-East, to south-western Angola in the South-West.

Other potential members: Channing and Rödel (2019) considered H. vilhenai Laurent 1964 from northern Angola to be part of the H. cinnamomeoventris complex based on unpublished molecular data. Hyperolius polli Laurent, 1943 was mentioned to be similar to H. cinnamomeoventris (Laurent 1954).

Hyperolius concolor group

Hyperolius concolor superspecies of Schiøtz (1975) [partim]; Group II of Amiet (2012) [partim]; Hyperolius cinnamomeoventris group of Dehling and Sinsch (2019) [partim]; Group H of Ernst et al. (2021) [partim].

Contents (based on genetic data): H. bobirensis Schiøtz, 1967; H. concolor (Hallowell, 1844); H. guttatus Peters, 1875 stat. nov.; H. ibadanensis Schiøtz, 1967 stat. nov.; H. sankuruensis Laurent, 1979; Hyperolius sp. (Hyperolius sp. 1 sensuBurger et al. 2004, Hyperolius sp. B Gabon sensuVeith et al. 2009); H. zonatus Laurent, 1958 (Burger et al. 2004, Veith et al. 2009, Portik et al. 2019; this study).

Area of occurrence: From south-eastern Sierra Leone in the West to south-western Cameroon in the East. Doudou Mts. in south-western Gabon and in the central Congo Basin south of the Congo River.

Other potential members: Perret (1966) suggested that H. stenodactylus Ahl, 1931 from southern Cameroon (type locality: Bipindi) potentially represents a synonym of ‘H. concolor’. Considering the geographical origin of H. stenodactylus, it may be a synonym of H. guttatus, but see Discussion.

Hyperolius langi group

Group II of Amiet (2012) [partim].

Contents (based on genetic data): H. adametzi Ahl, 1931; H. kuligae Mertens 1940; H. langi Noble 1924; H. major Laurent, 1957 (Portik et al. 2019; this study).

Area of occurrence: Disjunct distribution in Cameroon and Gabon, north-western Democratic Republic of the Congo and Uganda, and southern Democratic Republic of the Congo.

Hyperolius montanus group

Hyperolius mitchelli clade of Conradie et al. (2018); Groups F and H of Ernst et al. (2021).

Contents (based on genetic data): H. mitchelli Loveridge 1953; H. montanus (Angel 1924); H. pictus Ahl, 1931; H. rubrovermiculatus Schiøtz, 1975; H. stictus Conradie, Verburgt, Portik, Ohler, Bwong and Lawson, 2018; H. substriatus Ahl, 1931 (Conradie et al. 2018, Portik et al. 2019, Bwong et al. 2020, Ernst et al. 2021).

Area of occurrence: East Africa from central Kenya in the North to central Mozambique in the South and eastern Zambia and south-eastern Democratic Republic of the Congo in the West.

Other potential members: Hyperolius atrigularis Laurent 1941 and H. kibarae Laurent, 1957 are possibly related to this group (Schiøtz 1999).

Hyperolius mosaicus group

Group II of Amiet (2012) [partim]; Central African forest species group of Dehling and Sinsch (2019); Group B of Ernst et al. (2021).

Contents (based on genetic data): H. endjami Amiet 1980; H. mosaicus Perret 1959 (Portik et al. 2019, Ernst et al. 2021).

Area of occurrence: South-western Cameroon to northern Gabon and Republic of the Congo.

Other potential members: Schiøtz (1999) found H. acutirostris Buchholz and Peters, 1875 from south-western Cameroon very similar and possibly related to H. endjami and H. mosaicus. Hyperolius stenodactylus may potentially be related to or conspecific with H. acutirostris (see Discussion).

Hyperolius occidentalis group

Group II of Amiet (2012) [partim]; Group G of Ernst et al. (2021).

Contents (based on genetic data): H. baumanni Ahl, 1931; H. chlorosteus (Boulenger 1915); H. laurenti Schiøtz, 1967; H. occidentalis Schiøtz, 1967; H. picturatus Peters, 1875; H. sylvaticus Schiøtz, 1967; H. torrentis Schiøtz, 1967 (Schick et al. 2010, Portik et al. 2019, Ernst et al. 2021; this study).

Area of occurrence: From eastern Sierra Leone in the West to south-western Cameroon in the East.

Other potential members: Schiøtz (1967) commented on the resemblance of colour patterns of H. viridigulosus Schiøtz, 1967 to H. chlorosteus and H. laurenti.

Hyperolius platyceps group

Group II of Amiet (2012); Group B of Ernst et al. (2021).

Contents (based on genetic data): H. cinereus Monard 1937; H. chelaensis Conradie, Branch, Measey and Tolley, 2012; H. platyceps (Boulenger 1900); H. raymondi Conradie, Branch and Tolley, 2013 (Conradie et al. 2012, 2013, Portik et al. 2019, Ernst et al. 2021).

Area of occurrence: Southern Cameroon in the North, south-western Angola in the South and western Democratic Republic of the Congo and north-eastern Angola in the East.

Other potential members: Hyperolius atrigularis, H. kibarae, and H. polli were mentioned as potentially related to H. platyceps (Badjedjea et al. 2022).

Hyperolius spinigularis group

Group C of Ernst et al. (2021); Spiny-throated reed frog complex/group of Loader et al. (2015), Barratt et al. (2017) and Lawson et al. (2023); H. spinigularis complex/group of Lawson (2010) and Barratt et al. (2017).

Contents (based on genetic data): H. burgessi Loader, Lawson, Portik and Menegon, 2015; H. davenporti Loader, Lawson, Portik and Menegon, 2015; H. minutissimus Schiøtz, 1975; H. ruvuensis Barratt, Lawson and Loader, 2017; H. spinigularis Stevens 1971; H. tanneri Schiøtz 1982; H. ukaguruensis Lawson, Loader, Lyakurwa and Liedtke, 2023; H. ukwiva Loader, Lawson, Portik and Menegon, 2015 (Loader et al. 2015, Barratt et al. 2017, Portik et al. 2019, Ernst et al. 2021).

Area of occurrence: Mountains of central, eastern, and southern Tanzania (East and West Usambara Mts., Nguru Mts., Ukaguru Mts., Uluguru Mts., Rubeho Mts., Udzungwa Mts., Southern Highlands). Hyperolius spinigularis is known from southern Malawi (Mt. Mulanje) and adjacent regions of north-eastern Mozambique (Mt. Namuli). Hyperolius ruvuensis is known from the coastal forests of eastern Tanzania (Ruvu Forest).

Hyperolius obstetricans group (subgenus Alexteroon Perret, 1988)

Group E of Ernst et al. (2021).

Contents (based on genetic data): Hyperolius hypsiphonus (Amiet 2000); H. jynx (Amiet 2000); H. obstetricans Ahl, 1931 (Ernst et al. 2021).

Area of occurrence: From southern and southwestern Cameroon in the North to northwestern Angola in the South (Channing and Rödel 2019, Ernst et al. 2021).

Taxa not assigned to a species group

Hyperolius quinquevittatus Bocage 1866 is not assigned to any species group because the nuclear DNA analyses place it in polytomy with the H. balfouri group, the H. cinnamomeoventris group, the H. concolor group, and the H. montanus group (Fig. 2). Phylogenetic analyses of 16S mtDNA show that H. quinquevittatus is sister to the H. balfouri group with high support (Supporting Information, Fig. S4), which was also obtained by Portik et al. (2019). However, the phylogenomic species-tree analysis of Portik et al. (2019) showed a relatively deep divergence between the H. balfouri species group and H. quinquevittatus. Similarly, H. koehleri Mertens 1940 and H. tuberilinguis Smith 1849 are not assigned to any species group as both species show unstable phylogenetic positions between the two datasets. Portik et al. (2019) found them to be sister to all other taxa in Hyperolius Clade 2.

Appendix 2

GenBank discrepancies

Jetz and Pyron (2018) placed H. bobirensis in a sister-relationship to H. picturatus in a clade including taxa assigned here to the H. montanus group. At the time of the Jetz and Pyron (2018) study, only a single 16S sequence of H. bobirensis (GU443982, voucher ANK101) was available in the GenBank online database. We checked the sequence using the BLAST tool (Altschul et al. 1990), which showed 100% similarity to H. sylvaticus (MK509693, MVZ 245057). Our unpublished preliminary phylogenetic analysis included this sequence and placed it in a clade formed by species of the H. occidentalis group, suggesting that this sequence was assigned to H. bobirensis erroneously. Therefore, the sister-relationship between H. bobirensis and H. picturatus presented in Jetz and Pyron (2018) and later adopted by Dubois et al. (2021) should be rather interpreted more as a sister-relationship between H. picturatus and H. sylvaticus. However, Jetz and Pyron (2018) already placed H. sylvaticus in a sister-position to H. baumanni. In this case, the same erroneous identification of the H. sylvaticus sequence (GU443981, BOB37) occurred. BLAST searches showed that H. picturatus is 100% identical. Further investigation of the sequences of H. concolor (GU443984, AM40), H. picturatus (GU443983, HLOM26) and H. sylvaticus (KU166829, UWBM 5730) showed similar mismatches and placed them in close relationship to H. torrentis, H. sylvaticus and H. torrentis, respectively.

References