We present a molecular phylogeny of African stream frogs (genus Strongylopus), based on 12S rRNA, 16S rRNA, the nuclear recombination activating gene 1 (RAG-1) and tyrosinase exon 1 (tyr). Molecular data were supported by advertisement call analysis and morphology. We recognise six valid species: Strongylopus bonaespei (Dubois, 1981) from the southern and southwestern parts of the Western Cape Province, South Africa; Strongylopus fasciatus (Smith, 1849) from eastern South Africa to Zimbabwe; Strongylopus grayii (Smith, 1849) found throughout South Africa with older records in Naukluft, in central Namibia; Strongylopus rhodesianus (Hewitt, 1933) known from the eastern highlands of Zimbabwe and western Mozambique; Strongylopus wageri (Wager, 1961) from KwaZulu-Natal Province, South Africa and Strongylopus merumontanus (Lönnberg, 1907) from eastern Zambia, Malawi, northern Mozambique and Tanzania. Strongylopus fuelleborni (Nieden, 1911), S. kitumbeine Channing & Davenport, 2002 and S. kilimanjaro Clarke & Poynton, 2005 were shown to be junior synonyms of Strongylopus merumontanus. Strongylopus springbokensis Channing, 1986 is recovered as a junior synonym of Strongylopus grayii. Divergence ages were estimated, and discussed in terms of paleoclimatic events.
Stream frogs in the genus Strongylopus are widespread in the south of Africa and at high elevations of tropical East Africa (Poynton, 2013), reaching low elevations only in the high latitudes. They are assigned to the endemic African family Pyxicephalidae (van der Meijden et al., 2005), and presently 10 species are recognised (Frost, 2022).
Strongylopus bonaespei (Dubois, 1981), endemic to South Africa, is associated with the coastal lowlands and the higher mountainous regions running parallel with the Western Cape Province coast from west to east; Strongylopus fasciatus (Smith, 1849) is known from eastern South Africa to Mozambique, Zimbabwe, southern Zambia and the highlands of eastern Africa, including the Eastern Arc Mountains of Tanzania and the Nyika Plateau of Malawi; Strongylopus fuelleborni (Nieden, 1911) occurs on the isolated mountains in the south of Malawi and adjacent Mozambique; Strongylopus grayii (Smith, 1849) is widespread in South Africa at both lower and higher elevations, extending into Lesotho, and reaching the borders of Botswana in the north and an isolated population from Naukluft in central Namibia; Strongylopus kilimanjaro Clarke & Poynton, 2005 is only known from the high slopes (above 1700 m) of Mt. Kilimanjaro (Zancolli et al., 2013); Strongylopus kitumbeine Channing & Davenport, 2002 is restricted to an extinct volcano, Mt. Kitumbeine in northern Tanzania; Strongylopus merumontanus (Lönnberg, 1907) similarly occurs on the highlands of an extinct volcano, Mt. Meru in northern Tanzania; Strongylopus rhodesianus (Hewitt, 1933) is known from the Eastern Highlands of Zimbabwe and the Gorongosa massif in western Mozambique; Strongylopus springbokensis Channing, 1986 is known from the arid area of Namaqualand in South Africa; Strongylopus wageri (Wager, 1961) is found on the uplands of KwaZulu-Natal Province, including the slopes of the Drakensberg.
The taxon has an unsettled taxonomic history and was regarded as a ‘problematic’ genus by Poynton (2013) due to the similarity of many of the species. Strongylopus was described by Tschudi in 1838, although it was still placed in the genus Rana by most workers for the next century. The status of the genus Strongylopus was later confirmed based on tadpole morphology (Van Dijk, 1966), but subsequently placed as a subgenus of Rana (Dubois, 1988). The misidentification of an Amietia gene sequence as S. grayii resulted in the apparent polyphyly of Strongylopus (Bittencourt-Silva et al., 2016; Channing et al., 2016). Presently the genus Strongylopus is recognised as part of the endemic African family Pyxicephalidae, which is not closely related to Rana (Channing, 2001; Frost et al., 2006; Channing et al., 2016; Du Preez & Carruthers, 2017). Frost (2022) presents a history of the taxonomic changes applied to Strongylopus, and there are still questions concerning the taxonomy of many of the species.
One problem is the occurrence of colour variations. For instance, a common colour pattern is longitudinal lines of black, brown and yellow, which may have caused confusion in the identification of S. bonaespei, S. fasciatus and the three northern species (Poynton, 1964b; Greig et al., 1979; Clarke & Poynton, 2005). Colour patterns may distinguish some species, although these can be very variable. Further pattern elements that vary include ventral spotting, vertebral stripes, and cross-banding on legs. Some morphological criteria are useful in species diagnosis, for example, Strongylopus grayii can be distinguished from S. rhodesianus by the ratio of head width/length of fourth toe (Poynton, 1964a). Morphological characters previously used to differentiate between taxa (Poynton, 1964a; Poynton & Broadley, 1985) include relative head width, relative snout length, relative diameter of tympanum, relative foot length, relative tibia length, and amount of webbing on the fourth toe. Advertisement calls are useful taxonomic characters in most instances (Passmore, 1977), as morphologically similar species of Strongylopus can be easily distinguished acoustically (Channing, 1979).
While most species have small distributions, Strongylopus fasciatus (Smith, 1849) and S. grayii (Smith, 1849) are widespread and expected to show some genetic sub-structure. Strongylopus fuelleborni is known from Tanzania, Malawi and northeastern Zambia. Loveridge placed S. fuelleborni as a synonym of Rana fasciata merumontana (1933), and later as a subspecies of S. fasciata (1953). Poynton (1964b) compared six specimens from each of six localities, and showed that the ratio of head width/foot length revealed three groups, one in South Africa and Zimbabwe (as S. f. fasciatus), one on Mt. Mulanje and Nyika Plateau (as S. fasciatus fuelleborni) and the third in the north on the Usambara and Poroto Mountains (as S. fasciatus merumontana). The proportions of the central S. fuelleborni are between those of the southern S. fasciatus and the northern S. merumontanus. Strongylopus fuelleborni which occurs on many isolated highlands (Conradie et al., 2018) was elevated to a full species by Channing (2001) based on advertisement call differences. Channing & Davenport (2002) based on similarity in advertisement calls and morphology, subsequently placed S. fuelleborni as a junior synonym of S. merumontanus (Lönnberg, 1907), which was described from Mt. Meru in northern Tanzania. Poynton (2004) regarded S. merumontanus as restricted to high elevations on Mt. Meru. Pickersgill (2007) used the name S. fuelleborni for the northern species as it was geographically separate from S. fasciatus. However, the integrity of this species has not been tested phylogenetically.
The other common species, the Clicking Stream Frog Strongylopus grayii (Smith, 1849) is widespread in South Africa, known from many vegetation types and a range of elevations. Tolley et al. (2010) showed differences between southwestern (winter rainfall) and northern (summer rainfall) populations, which they suggested might represent cryptic species. Strongylopus kitumbeine Channing & Davenport, 2002 was described from Mt. Kitumbeine, an extinct volcano in northern Tanzania. It was diagnosed based on some unique characters in morphology and advertisement calls. Strongylopus kilimanjaro Clarke & Poynton, 2005 is the most recently described species, distinguished on small morphological differences from S. kitumbeine. It is only known from Mt Kilimanjaro in Tanzania. Strongylopus rhodesianus (Hewitt, 1933) was originally described as a subspecies of Rana grayi (Hewitt, 1933). However, small samples of typical S. grayii and S. rhodesianus showed significant differences in the ratio of head length/length of fourth toe (Poynton, 1964a). Strongylopus rhodesianus is not well known, occurring on the Eastern Highlands of Zimbabwe and the adjacent highlands of Mozambique, including Mt Gorongosa. Poynton & Broadley (1985) thought it very likely that there was genetic mixing with S. fasciatus. Strongylopus bonaespei (Dubois, 1981) was originally described as Rana fasciata montana FitzSimons, 1946, and later named Rana montana by Greig et al. (1979). However, these names were preoccupied by Rana temporaria var. montanus Koch, 1872. The species was renamed Rana (Strongylopus) bonaespei by Dubois (1981). It occurs along the coast and on the mountains in the southern and southwestern Cape Province, South Africa. The little-known Strongylopus springbokensis Channing, 1986 occurs in the montane, arid region of Namaqualand in South Africa. Strongylopus wageri (Wager, 1961) is known from the midlands of KwaZulu-Natal Province and the slopes of the Drakensberg Mountains in South Africa. Although the species was inadvertently described by Wager, there seem to be no taxonomic problems in this taxon (Poynton, 1963). In contrast the status of the taxa isolated on relatively small areas in northern Tanzania (Strongylopus kitumbeine, S. kilimanjaro and S. merumontanus) is in need of further study, i.e. to test their genetic isolation. Poynton (2004) showed that the northern S. fuelleborni is not related to the southern S. fasciatus. The early changes in nomenclature of these species are reviewed by Poynton (2013).
The purpose of this study is to re-evaluate the taxonomy of stream frogs, using nuclear and mitochondrial gene sequences, advertisement calls, and morphology.
MATERIALS AND METHODS
Sampling: The specimens used for the molecular study are listed in Appendix 1. This includes material from South Africa, Lesotho, Zimbabwe, Malawi, Mozambique, and Tanzania. Localities where advertisement calls were obtained are listed in Appendix 2. Finally, all material used for the morphological study is listed in Appendix 3.
Molecular: DNA extraction and amplification of fragments of mitochondrial 12S rRNA, 16S rRNA, the nuclear-coding tyrosinase exon 1 (tyr), and a fragment of the nuclear recombination activating gene 1 (RAG-1) followed standard protocols (Channing et al., 2016). We developed new nuclear recombination activating gene 1 (RAG-1) primers for Strongylopus, Rag1-Str-F1 (TGC TGC AAA CCC CTT TGC CTA) and Rag1-Str-R1 (TTC CCT TCG CTG GCC CA) and modified 12S primers: 12Sam F (TCCTRGCCTTRTCARCT) and 12Sam R (ATAGTGGGGTATCTAATCCCAGTTT). The annealing temperature for the RAG-1 primers was 55°C, and 51°C for the 12Sam primers. Some samples were sequenced in both directions, and some were sequenced on two separate occasions. Sequences were aligned using Sequencher 5.4. GenBank sequences and sources are listed in Appendix 1. Uncorrected p-distances were calculated using PAUP* (Swofford, 2002). The uncorrected p distances for the mitochondrial 16S rRNA, nuclear tyr and RAG-1 genes were compared between all available Strongylopus sequences. The genus Amietia is the appropriate outgroup for Strongylopus (Frost et al., 2006; Jetz & Pyron, 2018, 2019). The outgroup sequence is a chimaera of Amietia nutti from Kenya (12S KU69367, 16S KU693777, tyr KU694111), and Amietia delalandii from Lesotho (RAG-1 HQ014422).
The two most likely haplotypes for each individual for the nuclear genes were determined by first submitting the aligned sequences, including IUPAC polymorphism symbols, to SeqPhase step 1 (Flot, 2010), which produces an input file for PHASE (Stephens & Donnely, 2003). PHASE calculates the likelihood of possible haplotypes. The output is submitted to SeqPhase step 2, which produces full sequences for each haplotype (accessed at https://eeg-ebe.github.io/SeqPHASE). Haplotype networks for both nuclear gene fragments and the mitochondrial genes were determined using TCS 1.21 (Clement et al., 2000) and illustrated using PopArt (Population Analysis with Reticulate Trees; http://popart.otago.ac.nz).
Phylogenies based on each gene separately (phased nuclear genes), and then on all four genes combined, for samples where this was possible, were generated using the Maximum Likelihood (ML) software IQ-TREE (Nguyen et al., 2015; Trifinopoulos et al., 2016). Aligned sequences were submitted to the website W-IQ-TREE ( http://iqtree.cibiv.univie.ac.at). The ultrafast bootstrap UFBoot2 (Minh et al., 2013; Hoang et al., 2017) was determined using 1000 replicates. The software determined the appropriate model for the ML analysis: TIM2e+I+G4. Support values for the phylogeny were calculated as SH-aLRT (%) and the ultrafast bootstrap (%). The tree was rooted and drawn using FigTree 1.4.4 (Rambaut, 2018). Divergence dates for representatives of each species were estimated in BEAST 2.6.3 (Hasegawa et al., 1985; Heled & Drummond, 2012; Bouckaert et al., 2019), calibrating the tree based on a mean rate of change of 0.16% per million years for Pyxicephalidae 16S rRNA, after Bittencourt-Silva et al. (2016).
Museum acronyms: The museum acronyms used in the literature and their current equivalents follows Sabaj (2019). TMP – Transvaal Museum, Pretoria (now DNMNH – Ditsong National Museum of Natural History, Pretoria); BMNH – Natural History Museum, London; PEM – Port Elizabeth Museum, Bayworld complex, Gqeberha; NRM – Swedish Museum of Natural History, Stockholm; ZMB – Museum für Naturkunde, Leibniz Institute for Evolution- and Biodiversity Science, Berlin; CAS – California Academy of Sciences.
Advertisement calls: Most calls were obtained in the field using a Marantz PMD660 digital recorder with a directional microphone. Older recordings using a Sony cassette recorder were also included. Air temperature was noted. Call analysis was carried out using Raven Pro 1.6.1 (Center for Conservation Bioacoustics, 2019). Calls consist of brief clicks or whistles. Call parameters were determined following Köhler et al. (2017). The following call parameters were noted: dominant harmonic midpoint, other emphasised harmonics, pulses per note, note duration and note rate. Calls were classified into guilds following Emmrich et al. (2020).
Morphology: Measurements were taken with a digital calliper accurate to 0.01 mm, rounded to 0.1 mm, with the aid of a dissecting microscope. The following measurements were recorded: SUL – snout-urostyle length, HW – maximum head width, HL – head length measured from the back of the lower jaw to the snout tip, SL – snout length measured form the anterior corner of the eye to the snout tip, EW – horizontal eye diameter, EN – distance between eye and nostril, ET – distance between eye and tympanum, TYM – horizontal tympanum diameter, FOT – foot length measured to include the inner metatarsal tubercle to the tip of the fourth toe, TIB – length of tibiafibula, HND – hand length, measured to include the palmar tubercle to the tip of the third finger. In addition, notes were made of ventral, snout, back and limb patterns. The dorsal vertebral skin ridges may be absent, interrupted or continuous, and this was noted. Colour patterns were scored from photos of live animals. The following ratios were calculated: HW/FOT – relative foot length as used by Poynton (1964a), SUL/HW – relative head width, SUL/SL – relative snout length, HW/EW – relative eye size, ET/EW relative distance between eye and tympanum, TYM/ET – relative tympanum size, SUL/TIB – relative tibiafibula length, TIB/FOT – relative foot to tibia length, HW/HND – relative hand length, SL/EW – snout length relative to eye width, FOT/SUL – length of foot relative to body length, EN/TYM – relative size of tympanum.
Phylogeny: The software proposed TIM2e + I + G4 as the appropriate model for the ML analysis. The phylogeny based on four genes showed six well-supported clades. The same topology was recovered from the four genes when analysed separately. We show the 16S and tyr trees as they included most samples (Figs 1, 2, 3). The 16S rRNA sequences (Figs 1, 2) show slight variation between the highland populations of S. fuelleborni on Mt. Mulanje (Malawi), S. fuelleborni on the Mt. Namuli and Mt. Mabu complex (Mozambique), S. fuelleborni at Nyanga and the highlands of eastern Zimbabwe, S. fuelleborni on the Mbizi Hill highlands south of Lake Rukwa in Tanzania, the Livingstone Mountains in Tanzania, and the Udzungwa Mountains in Tanzania, S. kilimanjaro on Mt. Kilimanjaro (Tanzania), S. kitumbeine on Mt. Kitumbeine (Tanzania) and S. merumontanus on Mt. Meru (Tanzania)(Fig. 4). The nuclear tyr phylogeny (Fig. 3) shows that many sequences of S. kilimanjaro, S. fuelleborni, S. merumontanus and S. kitumbeine are identical.
Range of uncorrected p-distances as percentages for 16S rRNA, tyr and RAG-1 between Strongylopus taxa. GRA – S. grayii, FAS – S. fasciatus, MER – S. merumontanus, WAG – S. wageri, BON – S. bonaespei. RHO – S. rhodesianus. Top line –16S, middle line – tyrosinase exon 1, lowest line – RAG-1.
The 16S rRNA phylogeny (Fig. 2) shows that S. springbokensis sequences from Rooiberg and Studer Pass on the Namaqualand highlands are embedded within sequences of S. grayii from the southern and southwestern Cape in South Africa. Sequences of the nuclear tyr gene from the same localities (Fig. 3) places them basal to S. grayii but in a well-supported clade. Strongylopus springbokensis is thus regarded as a junior synonym of S. grayii.
Uncorrected p distances: The uncorrected p distances for 16S and the two nuclear genes, tyr and RAG-1 are presented in Table 1. The uncorrected interspecific p distances for 16S varied from 2.8-7.4%, for tyr 0.2-3.6%, and for RAG-1 0.2-6.7%. The intraspecific uncorrected p distances for 16S varied from 0-2.1%, with the largest differences revealed between isolated populations of S. merumontanus. Table 4 lists the uncorrected p distances between S. kitumbeine, S. kilimanjaro, S. merumontanus, and populations of S. fuelleborni on Mt. Mulanje, Mts. Namuli and Mabu, Nyanga, Mbizi Hill, and the Udzungwa Mountains. The distances between S. merumontanus and the populations of S. fuelleborni on Mt. Mulanje and Mts. Namuli and Mabu vary from 1.5 to 2.1%. The distances between S. merumontanus, S. kitumbeine, S. kilimanjaro, and populations of S. fuelleborni from Nyanga, Mbizi Hill, the Livingstone Mts. and the Udzungwa Mts. are very similar, varying 0-0.8%. The intraspecific distances for tyr and RAG-1 varied from 0-1.0%, and 0-1.4%, respectively.
Haplotype networks: Haplotype networks using the TCS algorithm are shown in Fig. 5. The mitochondrial 12S rRNA and 16S rRNA, as well as the two nuclear genes showed similar patterns. There was a single shared RAG-1 haplotype between S. rhodesianus and S. merumontanus from Nyanga in Zimbabwe. Due to low genetic differences, S. merumontanus, S. fuelleborni, S. kitumbeine, and S. kilimanjaro were combined and treated as a single species. The minimum base changes between closely related species are listed in Table 2. Discordance: Some recent studies have shown discordance between the phylogenies derived from mitochondrial and nuclear genes (Toews & Brelsford, 2012), which is often related to incomplete lineage sorting (Streicher & Day, 2020). The faster rate of divergence of mitochondrial genes compared to nuclear genes in amphibians (Crawford, 2003), should not change the relationships recovered where there is no discordance. Hybridization has been proposed to account for discordance in Polygonia butterflies (Wahlberg et al., 2009) and for many cases in amphibians (Toews & Brelsford, 2012). Gonçalves et al. (2007) suggested that a deep coalescence was responsible for the discordance in midwife toads Alytes, but could not discount the possibility of putative hybridization events causing a reticulated pattern of evolution. Although there may be a difference in the relationships recovered from mitochondrial and nuclear genes, there was no difference in the composition of the clades (species) recovered here. No discordance was found in Strongylopus, with the clades (species) strongly supported, similar to the findings of Stöck et al. (2008) in the species of Hyla found around the Mediterranean, and Pleuroderma in South America (Faivovich et al., 2012).
Minimum base differences between species of Strongylopus. See Table 1 for species codes. Data presented as 12S rRNA/16S rRNA/tyr/RAG-1.
Dating (Fig. 6): The split of Strongylopus from Amietia happened during the early Eocene, 53 Mya (95% HPD [highest posterior density] 83.4-26.8 Mya). The next split during the middle Oligocene was between the ancestor of southern S. bonaespei + S. grayii + S. wageri from the ancestor of northern S. merumontanus + S. fasciatus + S. rhodesianus at 27 Mya (95% HPD 41.3-15.6 Mya). Further splits during the Miocene included the Strongylopus bonaespei split from S. grayii + S. wageri around 22 Mya (95% HPD 34.3-11.7 Mya), followed by the split of S. grayii from S. wageri around 19 Mya (95% HPD 29.6-8.7 Mya). The S. rhodesianus split from the ancestor of S. fasciatus + S. merumontanus took place at around 16 MYA (95% HPD 22.6-10.8 Mya), followed by the split of S. fasciatus from S. merumontanus at around 11 Mya (95% HPD 11.7-10.2 Mya). The paleoclimate correlates are discussed below.
The calls could be divided into four types. A single amplitude modulated (AM) click, a brief high-pitched frequency modulated (FM) whistle, a chuckle, and a trill (Fig. 7). Strongylopus grayii usually produces clicking calls. Brief whistle calls are known for S. fasciatus and S. merumontanus. A trill is a series of repeated whistles, common in S. merumontanus. The chuckle call is part of the S. bonaespei vocalisation, often interspersed with croaks. The call of S. wageri is best described as a very variable cackle, with an infrequent loud clack or trill. The call of S. rhodesianus is unknown. It was previously described as a trill (Channing, 2001) but this was not based on a male whose identification was confirmed by DNA, and may have been a misidentified S. fasciatus. Species-specific call characteristics are given with the species summaries below, including the call guilds after Emmrich et al. (2020). The call parameters of S. fuelleborni, S. kitumbeine, and S. kilimanjaro are discussed under H. merumontanus below, as they are very similar and overlap in structure.
Greig et al. (1979) showed differences between the calls of S. fasciatus and S. bonaespei. Poynton (1964a) remarked that the southern and northern populations of S. grayii had slightly different advertisement calls with the northern population having slightly lower pitched calls. Our large sample (474 calls) shows that the range of dominant frequencies, 1464-3030 Hz was not related to latitude (Fig. 8) with a range of frequencies shown in both northern and southern populations.
The available frogs of all taxa showed remarkable similarity in most body proportions (Table 3), although there was considerable variation. The largest species in our sample was S. wageri with a snout-urostyle length of 50.6 mm. It was closely followed by S. fuelleborni (49.3), S. kilimanjaro (46.5) and S. bonaespei (46.5). Poynton (1964a) gives a maximum size for S. bonaespei of 47.5 mm.
The mean snout-urostyle lengths were very similar. Strongylopus merumontanus is the shortest (25 mm) while S. grayii, S. springbokensis, S. fasciatus, S. fuelleborni, S. kitumbeine, S. kilimanjaro, S. wageri, S. bonaespei and S. rhodesianus all varied 31.4-40.5 mm.
Strongylopus springbokensis had the shortest legs, as shown by the average value of 0.7 for HW/FOT, although there was considerable overlap shown by all other species of 0.4-0.6. Strongylopus springbokensis also had the shortest hands relative to head width (HW/HND=1.3-2.1) while S. fasciatus had the longest hands (HW/HND=0.8-1.2). The other eight species showed intermediate but overlapping values for relative hand length.
Head width as a proportion of body length varied evenly across the species, as shown by the overlap in the proportion SUL/HW (Table 3) for all species. The largest eyes, as indicated by HW/EW were found in S. merumontanus (mean HW/EW=12.0), with the smallest eyes (mean HW/EW=18.3) were found in S. springbokensis. However, all species have overlapping ranges in this proportion.
The largest distance between tympanum and eye, as ET/EW, occurred in S. bonaespei (mean ET/EW=0.4). All the other species had mean values of 0.3. Body proportions in general were not useful for species identification as there was considerable variation and overlap between all species (Table 3).
The amount of webbing, as indicated by the number of phalanges free of webbing on the fourth (longest) toe, varied from 3-4 within most species, with the most webbing shown by S. wageri, where there are 2-3 phalanges free in our sample, often with a margin of webbing to the toe tip. Poynton (1964a) reported 1-3 phalanges free of web in S. wageri. Breeding males tend to have more webbing, but this needs to be investigated with a larger sample, or with a long-term field study. The presence and completeness of dorsal ridges have been used as characters in most species descriptions. Our data revealed that S. fasciatus has unbroken vertebral dorsal ridges while S. grayii, S. springbokensis, S. rhodesianus and S. bonaespei have broken ridges, sometimes with a smooth back. Strongylopus merumontanus, S. fuelleborni, S. kitumbeine and S. kilimajaro have dorsal ridges which are usually continuous, but are often broken or absent. Strongylopus wageri has no dorsal ridges.
Colour patterns: Colour patterns have previously been used as diagnostic taxonomic characters by all authors (Figs 9-35). Variable pattern elements in most species include a thin vertebral stripe which may be present or absent, a wide vertebral band, the development of the ‘mask’ a dark stripe from the nostril through the eye to the arm insertion, a pale triangle on the snout, longitudinal dark stripes running parallel to the dorsal midline, the presence and arrangement of lateral stripes, sometimes broken into rows of spots, the presence and arrangement of dorsal spots or blotches, and the presence of a pale mark running from the anterior corner of the eye to the lower jaw. Overall colour varies from golden brown through dark brown, grey, yellow, orange to red. Stripes and spots are often contrasting in colour. Longitudinal dark stripes along the dorsal midline and laterally are common in S. fasciatus and S. merumontanus. The variation shown by the different species is detailed below.
The northern species (S. merumontanus, S. fuelleborni, S. kitumbeine, and S. kilimanjaro) are similar in genetics, morphology, and advertisement call. They are recognised here as a single species, with S. merumontanus being the senior synonym. Strongylopus springbokensis is recognised as a junior synonym of S. grayii.
Strongylopus bonaespei (Dubois, 1981)
Rana fasciata montana FitzSimons, 1946: 351 (type TMP 20223).
Rana montana. – Greig, Boycott & De Villiers, 1979: 2, figs 7, 10-12 .
Molecular: The mitochondrial 16S rRNA, nuclear tyr and RAG-1 sequences are remarkably uniform across the range of the species, from the Cape Peninsula in the west to Grootvadersbosch in the east (Table 1, Figs 1, 3).
Morphology: Maximum SUL is 35 mm for males and 48 mm for females (Poynton, 1964a). Snout-urostyle length 3.2 times head width (3.0-3.2). Snout length 1.9 times eye width (1.8-2.1). Eye-nostril distance 1.5 times tympanum width (1.3-1.6). Length of foot 0.8 of SUL (0.7-0.8). Head width 0.41 of foot length (0.40-0.44). Four phalanges of fourth toe free of web. The overall colour is golden brown, although very dark morphs are known. Snout mottled or a pale triangle. A thin dark line runs from the snout through the eye, over the tympanum to the arm. The dark pair of dorsal stripes usually has paired breaks (Poynton, 1964a); cross-banding present on the tibia. There are broken rows of dark paravertebral stripes, or silver and dark stripes (Figs 9, 10). A vertebral stripe is usually present. There are no brown stripes from behind eyes to urostyle. The ventral surface is immaculate.
Distribution: The distribution falls within the Cape Fold ecoregion of Abell et al. (2008). They are known from sea level to the high plateaux of the mountains, in the south-west, extending north along the Cederberg and east to the Tsitsikama Mountains in South Africa (Fig. 36).
Advertisement call: Strongylopus bonaespei produces a brief croaking call and a chuckle call (Fig. 7). The duration of the croaking note is 0.083-0.139 s, with 8 or 9 pulses, at a mean pulse rate of 60.8 s-1 (50.3-84.3 s-1, n=13). The mean emphasised frequency is 2115 Hz (1987-2231 Hz; n=14). Some calls have the first harmonic suppressed, with the second visible on the sound spectrogram. One individual emphasised both the fundamental frequency and the first harmonic. The chuckle call consists of an initial two notes, followed by three rapidly repeated notes, within 0.8 s (n=1). The call falls into guild E of Emmrich et al. (2020).
Strongylopus body proportions. Mean and range shown. GRA – Strongylopus grayii, SPR – S. springbokensis, FAS – S. fasciatus, FUE – S. fuelleborni, KIT – S. kitumbeine, KIL – S. kilimanjaro, MER – S. merumontanus, WAG – S. wageri, BON – S. bonaespei, RHO – S. rhodesianus. See text for morphological abbreviations.
Strongylopus fasciatus (Smith, 1849)
Rana fasciata Smith, 1849: 333, pl. 78, fig. 1; (neotype BMNH 188.8.131.52).
Rana fasciata fasciata. – Barbour & Loveridge, 1928: 197 (by implication).
Strongylopus fasciatus fasciatus. – Van Dijk, 1966: 259.
Rana (Strongylopus) fasciata. – Dubois, 1981: 250.
Strongylopus fasciatus. – Channing, 2001: 353, fig. 22.7.
Molecular: There are four 16S rRNA haplotypes, of which three overlap geographically, and differ by 1.8% or less. The samples from eastern Zimbabwe are 1.3-2.2% different from the southern population, but are all included in a well-supported clade (Fig. 1). The nuclear RAG-1 and tyr haplotypes show differences of 0-1.1% and show a well-supported clade (Fig. 3).
Morphology: Maximum SUL is 37 mm for males and 50 mm for females (Poynton, 1964a). Snout-urostyle length 3.7 times head width (3.1-4.4). Snout length 1.9 eye width (1.4-2.3). Eye-nostril distance 1.3 times tympanum width (1.0-1.9). Length of foot 0.8 of SUL (0.7-0.8). Head width 0.4 of foot length (0.3-0.4). Fourth toe with 3.5-4 phalanges free of web. Poynton (2013) described this species as possessing characteristic dark-edged skin ridging and stripes. Our samples are overall brown and yellow, with some darker morphs known. There is a pale triangle on the snout. There are two dark continuous paravertebral stripes with a pale vertebral band. A thin dark line runs from the snout through the eye, over the tympanum to the arm. A dark stripe runs from behind the eye to the urostyle, and there are two or three dark lateral stripes. The tibia has mottling or longitudinal stripes. The ventral surface is immaculate.
Distribution: This widespread species is found in six of the ecoregions of Abell et al. (2008): Cape Fold, Amatolo-Winterberg Highlands, Southern Temperate Highveld, Zambezian Lowveld, Eastern Zimbabwe Highlands and Zambezian Highveld (Fig. 36). It is known from Zimbabwe, to the South African highveld and east coast.
Advertisement call: Strongylopus fasciatus produces a very brief whistle that shows a rise in frequency, or sometimes a high-pitched click (Fig. 7). The mean emphasised frequency is 2854 Hz (1650-3200 Hz, n=361), although a few calls had the fundamental frequency suppressed, with the lowest emphasised frequency being the first harmonic, 5130-6140 Hz (n=30). In three cases, the sound spectrogram showed no harmonics, but only the fundamental frequency. Air temperature does not affect the frequency. The calls fall into guild B of Emmrich et al. (2020).
Strongylopus grayii (Smith, 1849)
Rana grayii Smith, 1849: 335, plate 78, fig. 2 (lectotype BMNH 184.108.40.206).
Strongylopus grayi. – Steindachner, 1867: 21.
Rana grayi var. dorsalis Werner, 1910: 297 (type unknown).
Rana grayi grayi. – Hewitt, 1933.
Dicroglossus grayi. – Deckert, 1938: 138.
Strongylopus grayii. – Van Dijk, 1966: 259
Rana grayii grayii. – Passmore & Carruthers, 1979: 138.
Rana (Strongylopus) grayii. – Dubois, 1981: 250.
Strongylopus springbokensis Channing, 1986: 128, fig. 1 (type PEM A963). New synonym
Rana (Strongylopus) springbokensis. – Dubois, 1992: 259.
Molecular: The 16S rRNA haplotype variation shows a maximum difference of 1.7%. These sequences fall into three groups (Fig. 37): a southwestern clade extending from Namaqualand in the northwest, to the southern Cape (blue square symbols); a southern Cape clade (red triangles); and a northern clade (yellow circles) which extends from northwestern South Africa to the south in the Cape Town area. Strongylopus springbokensis is embedded within the southwestern clade. These clades overlap geographically; see the discussion below. The nuclear RAG-1 and tyr sequences are 0-1.0% different. The tyr phylogeny places S. springbokensis basal but within a well-supported clade of S. grayii sequences (Fig. 3).
Morphology: Values are given for S. grayii and S. springbokensis separately in Table 3. Body proportions are listed here as S. grayii/S. springbokensis. Maximum SUL is 42/42 mm for males and 64/44 mm for females (Channing & Rödel, 2019). Snout-urostyle length is 3.0/2.7 times head width (2.6-3.3/2.5-2.8). Snout length 1.8/1.9 eye width (1.5-2.3/1.7-2.9). Eye-nostril distance 1.2/1.2 times tympanum width (0.8-1.8/0.8-1.8). Length of foot 0.6/0.5 of SUL (0.6-0.7/0.5-0.6). Head width 0.5/0.7 of foot length (0.4-0.6/0.7-0.7). Fourth toe with four phalanges free of web, infrequently three (Poynton, 1964a). These two species are remarkably similar in body proportions, supporting the placement of S. springbokensis as a synonym of S. grayii.
Our samples (Figs 15-26) are overall brown and grey with yellow, orange or red tinges. There is sometimes a pale triangle on the snout. A thin pale vertebral stripe or wide vertebral band is common. There are no paravertebral stripes, but irregular dark blotches, usually in pairs. A thin dark line runs from the snout through the eye, over the tympanum to the arm. There is usually a thin pale line from the anterior corner of the eye to the upper jaw. A dark stripe runs from behind the eye, through the tympanum to the arm. The tibia has dark cross-banding. An unusual morph with a red dorsum was found in the Langvlei dunes near Wilderness (Jacobsen, 2013). The ventral surface is immaculate, rarely with grey speckles.
Distribution: This widespread species is found in eight of the ecoregions of Abell et al. (2008): Karoo, Cape Fold, Amatolo-Winterberg Highlands, Drakensberg-Maluti Highlands, Southern Temperate Highveld, Western Orange (extended to Naukluft). Karoo and Zambesian Lowveld (Figs 36, 37). It is widespread in South Africa, excluding the arid interior. This species was introduced to St Helena Island around 1883 by Miss Phoebe Mary Moss (Barbour, 1934; Mertens, 1971; Ashmole & Ashmole, 2000; Lever, 2003) and they are still present. There are records from Naukluft in central Namibia from 1979 (Channing, 2020), which might represent an introduction. It has not been recently recorded from Naukluft.
Advertisement call: The call of S. grayii consists of a sharp, energetic click (Fig. 7). Sometimes the clicks are run together into a trill. The call of S. springbokensis from Springbok in Namaqualand is a trill or single click. The dominant frequency of S. springbokensis varies from 975 Hz to 1862 Hz (n males=12), while that of S. grayii varies from 1464 Hz to 3030 Hz (n males=61). The overlap in call structure and dominant frequency supports placing S. springbokensis as a junior synonym of S. grayii.
In summary, the duration of the click varies from 0.011-0.028 s. The mean emphasised frequency visible on the sound spectrogram is 2182 Hz (975-3030 Hz, n calls=560), with high energy visible up to 8 kHz. Air temperature does not affect the emphasised frequency. The calls are often produced in chorus, and frequently as a trill. The trills consist of 2-12 notes, at a pulse rate of 12.5-14.9 s-1. Females produce a twittering call during amplexus. The call is classified into guild A (Emmrich et al., 2020).
Strongylopus merumontanus (Lönnberg, 1907) Figs 27-33
Rana merumontana Lönnberg, 1907: 21, fig. 4 (type NRM 1367).
Rana (Ptychadena) merumontana. – De Witte, 1921: 7.
Rana fasciata merumontana. – Barbour & Loveridge, 1928: 197.
Strongylopus merumontanus. – Channing & Davenport, 2002: 140.
Rana fülleborni Nieden, 1911: 436 (type ZMB 21773). New synonym
Rana (Ptychadena) fuelleborni. – De Witte, 1921: 7.
Strongylopus fuelleborni. – Channing, 2001: 354, fig. 22.8.
Strongylopus kitumbeine Channing & Davenport, 2002: 137, figs 2-5 (type CAS 225064). New synonym
Strongylopus kilimanjaro Clarke & Poynton, 2005: 54, figs 1-2 (type BMNH 19220.127.116.11). New synonym
Molecular: The 16S rRNA sequences form a well-supported clade, with recognisable sub-structure with differences varying from 0-2.1%. Isolated populations on the highlands of Mt. Mulanje, Mt. Kilimanjaro, the eastern highlands of Zimbabwe, Mt. Kilimanjaro, Mt. Namuli and Mt. Mabu in Mozambique, and Mt. Kitumbeine (Fig. 4) each show slight differences, with the greatest p-distances between Mt. Mulanje in Malawi and Mts Namuli and Mabu in Mozambique and the Udzungwa Mountains in Tanzania (Table 4). The nuclear RAG-1 and tyr sequences differ by 0-1.4%.
Morphology: The body proportions of S. merumontanus, S. fuelleborni, S. kitumbeine and S. kilimanjaro are listed separately in Table 3. The combined values are presented as S. merumontanus here. Maximum SUL is 40 mm for males and 53 mm for females (Channing & Rödel, 2019). Snout-urostyle length 3.2 times head width (2.0-4.0). Snout length 1.8 eye width (1.5-2.2). Eye-nostril distance 1.3 times tympanum width (1.0-2.0). Length of foot 0.7 of SUL (0.6-1.0). Head width 0.4 of foot length (0.4-0.6). Fourth toe with four phalanges free of web, infrequently slightly less. Our samples (Figs 27-33) are overall brown to reddish orange. There is sometimes a pale triangle on the snout. A thin pale vertebral stripe or wide vertebral band is common. There is a pair of paravertebral stripes in many populations. A thin dark line runs from the snout through the eye, over the tympanum to the arm. The tibia has cross-banding or a thin longitudinal line. Material from the northern volcanic mountains lacks the ridged paravertebral stripes of the southern S. fasciatus (Poynton, 2013). The ventral surface is immaculate or with grey speckles.
Distribution: This widespread species is found in six of the ecoregions of Abell et al. (2008): Eastern Zimbabwe Highlands, Mulanje, Coastal East Africa, Lake Malawi, Lake Rukwa, Pangani. It is known from the eastern highlands of Zimbabwe, the isolated highlands and old volcanic mountains of Malawi, northern Mozambique inselbergs, southern Tanzania, and the Eastern Arc Mountains. It has not been found in the drier lowlands between the Eastern Arc highlands (Fig. 36).
Advertisement call: The advertisement call of S. fuelleborni consists of a brief whistle or series of whistles combined into a trill. Each note is frequency modulated, with a rise in pitch, and may contain 1-10 pulses. The dominant frequency varies from 2061 Hz to 3145 Hz (n males=31). The advertisement call of S. kilimanjaro consists of a brief whistle or series of whistles combined into a trill. Each note is frequency modulated, with a rise in pitch. The dominant frequency varies from 2667 Hz to 2940 Hz (n males=6). The advertisement call of S. kitumbeine consists of a trill with two to eight notes. The dominant frequency varies from 2135 Hz to 2325 Hz (n males=15). The only recorded call of S. merumontanus consists of a brief whistle with a dominant frequency of 2663 Hz. The advertisement calls of these northern species overlap in structure and dominant frequency, supporting placing them into a single species.
In summary, the call of S. merumontanus consists of a brief whistle or pulsed note, that may be repeated as a trill (Fig. 7). The mean emphasised frequency is 2493 Hz (2061-3145 Hz, n calls=298). The trill consists of two to eight notes at a mean pulse rate of 9.4 s-1 (7-11.6 s-1). Whistles are frequency modulated, rising in frequency. Both trills and whistles are heard in the same chorus. The function of the different calls is not known. These calls fall into guilds B and F of Emmrich et al. (2020).
Uncorrected 16S rRNA p distances between populations included in Strongylopus merumontanus. Species codes: MER – S. merumontanus, FUE – S. fuelleborni, KIT – S. kitumbeine, KIL – S. kilimanjaro.
Strongylopus rhodesianus (Hewitt, 1933)
Rana grayi rhodesiana Hewitt, 1933: 12, plate 1 (Type not traced).
Strongylopus grayii rhodesianus. – Van Dijk, 1966: 259.
Strongylopus rhodesianus. – Channing, 2001: 359, fig. 23.3.
Molecular: Only a single specimen was available for sequencing. The two RAG-1 haplotypes were 0.2% different.
Morphology: Maximum length 40 mm for males and 50 mm for females (Channing & Rödel, 2019). Snout-urostyle length 3.2 times head width (3.0-3.5). Snout length 1.6 eye width (1.5-1.7). Eye-nostril distance 1.5 times tympanum width (1.3-2.0). Length of foot 0.7 of SUL (0.7-0.8). Head width 0.4 of foot length (0.4-0.6). Fourth toe with 3.7-4 phalanges free of web. Our samples (Fig. 34) are overall brown and grey. There is usually a pale triangle on the snout, often continuous with a broad vertebral band. A thin vertebral stripe is absent. The back is usually plain, or with darker irregular blotches. There is a thin pale line from the anterior corner of the eye to the upper jaw. A thin dark line runs from the snout through the eye, over the tympanum to the arm. The tibia has dark cross-bands. The ventral surface is immaculate or with grey speckles.
Advertisement call: In light of the presence of three species in the area (S. merumontanus, S. fasciatus and S. rhodesianus) and the similarity in the calls, there is no confirmed recording of a call where the male has been positively identified by sequencing.
Strongylopus wageri (Wager, 1961)
Rana wageri Wager, 1961: 151, figs 151, 154 (type NMP 1145).
Strongylopus wageri. – Van Dijk, 1966: 259.
Rana (Strongylopus) wageri. – Dubois, 1981: 233.
Molecular: Three 16S rRNA sequences were available from the Sani Pass area, which were identical. The nuclear RAG-1 phased sequences varied from 0.7-1.5%. Only a single tyr sequence was available.
Morphology: Maximum SUL is 42 mm for males and 48 mm for females (Channing & Rödel, 2019). Snout-urostyle length 2.9 times head width (2.4-3.5). Snout length 1.6 eye width (1.4-1.9). Eye-nostril distance 1.5 times tympanum width (1.1-1.9). Length of foot 0.7 of SUL (0.6-0.8). Head width 0.5 of foot length (0.4-0.6). Fourth toe with one to three phalanges free of web. There are no dorsal skin ridges. Our samples (Fig. 35) are overall brown. There is a pale triangle or spotting on the snout. A thin pale vertebral stripe is often present. There are irregular dark blotches or fine spots on a uniform dorsal background. A thin dark line runs from the snout through the eye, over the tympanum to the arm. A dark lateral line runs from the arm to the leg. The tibia has dark cross-banding. The ventral surface is immaculate.
Distribution: This species is found in two of the ecoregions of Abell et al. (2008): Drakensberg-Maluti Highlands and Southern Temperate Highveld (Fig. 38). It is known from forested areas and mountain slopes.
Advertisement call: Strongylopus wageri produces a series of notes that can best be described as a cackle, as the notes are irregularly spaced (Fig. 7). The individual calls (n calls=36) may sound like a cough in some cases. The note rate for cackles varies from 3.6-13.6 s-1, while the coughs show individual pulses with pulse rates of 36.7-129.0 s-1. The mean emphasised frequency is 1319 Hz (682-1563 Hz). The calls fall into guild E of Emmrich et al. (2020).
Genetic variation: The 16S gene is widely used as a yardstick when discussing species boundaries with 3% difference a general indication of potential species difference (Vences et al., 2005; Vietes et al., 2009). In the Pyxicephalidae, the 16S uncorrected p-distances between species within a genus varied as follows: Amietia 1.3-10.0% (Channing et al., 2016), Anhydrophryne 6.3-7.6% (Dawood & Stam, 2016), Arthroleptella 1.1-6.6% (Turner & Channing, 2008), Cacosternum 1.1-7.3% (Channing et al., 2013), Nothophryne 3.0-6.5% (Bittencourt-Silva et al., 2016), and Tomopterna 1.0-9.2% (Wilson & Channing, 2019). The distances found between species of Strongylopus (2.8-7.4%) found in our study are thus in line with the other pyxicephalid genera.
Substructure in Strongylopus grayii: Tolley et al. (2010) suggests that gene flow may be restricted between the northern (summer breeding) and southern (winter breeding) populations of S. grayii. They recognise a southern clade, which was divided into Southwest (Group 1) and South (Group 2), and a northern clade divided into East (Group 3), Drakensberg (Group 4) and North (Group 5). They also had two samples from the far north, which were not classified. Our study recognised three clades, equivalent to groups 1 and 2 and the northern clade (groups 3+4+5) of Tolley et al. (2010) and supports the recognition of a southern and northern clade, with their far northern samples included in the northern clade. However, our study included more samples than were available to Tolley et al. (2010), and shows geographical overlap between the northern and southern clades (Fig. 37), with the ‘northern clade’ extending to Cape Town at the southern tip of Africa, and a greater p-distance between individuals of up to 1.7%.
The northern and southern populations of S. grayii showed a mean 0.6% difference between populations in 16S sequences (Tolley et al., 2010). Tolley et al. (2010) found this difference to be “comparable to the level of divergence found between other amphibian species (Turner et al., 2004; Vences et al., 2005; Measey et al., 2007; Channing & Schmitz 2008; Turner & Channing, 2008; Blackburn, 2009).” However, this is a misreading of the literature. Turner et al. (2004) found a mean difference between combined 12S and 16S sequences between Arthroleptella species of 3.9%; Vences et al. (2005) found a mean difference between sibling species of mantellid frogs of about 7% for 16S; Measey et al. (2007) reported a 16S difference between two species of Ptychadena of 5.2-5.6%; Channing & Schmitz (2008) found Cacosternum kinangopensis to be 4% different for 16S from C. plimptoni; Turner & Channing (2008) found 16S differences between species of Arthroleptella of 1.7-6.6%; and Blackburn (2009) reported a mean divergence between Arthroleptis species of 12.8%, and between Cardioglossa species of 8.8%, using the 12S to 16S fragment. The 0.6% difference in 16S sequences found between some northern and southern populations of Strongylopus (Tolley et al., 2010) or the 1.7% difference in our larger sample is the kind of variation that is expected in a widespread species and reflects intraspecific variation.
Substructure in Strongylopus merumontanus: Many of the populations are found on isolated highlands, for example the extinct volcanoes Mt. Kitumbeine (S. kitumbeine) Mt. Kilimanjaro (S. kilimanjaro) and Mt. Meru (S. merumontanus), while S. fuelleborni is widely distributed in northeastern Zambia, Malawi and Tanzania (Fig. 4). Although these four populations have previously been recognised as different species, this study has demonstrated a similarity in genetics, advertisement call and morphology. The 16S rRNA uncorrected p distances between populations are not large enough to maintain specific status for the northern volcanic species. Strongylopus merumontanus has uncorrected p values of 0-1.9% compared to all other species in this group. Strongylopus fuelleborni from Mt. Mulanje has uncorrected p values of 1.5-1.9% between all other populations (Table 4), while the other populations of S. fuelleborni have uncorrected p values of 0.2-1.9%. The nuclear tyr sequences between the populations vary 0-1.0%.
The greater uncorrected 16SrRNA p distance of the population of S. fuelleborni on Mt. Mulanje compared to other populations may be related to the age of the inselberg. While the volcanoes of northern Tanzania are less than 2 million years old, Mt. Mulanje was formed by erosion of softer soil around the granites of the inselberg that were formed 140 Mya. The long period of isolation is reflected in the presence of at least 69 endemic plant species on Mt. Mulanje (Strugnell, 2002).
New relationships: The relationship between Strongylopus fasciatus and S. merumontanus has been documented since Barbour & Loveridge (1928) recognised Rana fasciata merumontana. Our molecular phylogeny confirmed the relationship, placing these taxa as sister species. Strongylopus bonaespei was also described as a subspecies of Rana fasciata by FitzSimons (1946), but the molecular phylogeny places S. bonaespei basal to all the other species. Strongylopus rhodesianus was initially believed to be related to S. grayii when it was described by Hewitt (1933) as Rana grayi rhodesiana. The molecular phylogeny, however, places S. rhodesianus and S. grayii in different clades.
Origin of the species: Poynton & Broadley (1985) suggested that the genus shows a progression from primitive taxa in the south, with more derived taxa in the north. Poynton later doubted that there was support for this suggestion, stating that the genus does not easily fit the commonly held pattern of radiation from the south (Poynton, 2013). The southern African rainfall gradient from >1000 mm in the east to <20 mm in the west is reflected in the presence of Strongylopus only in the east of the continent, with the exception of a possible introduction of S. grayii in central Namibia (Channing, 2020). We estimated the split of the genus Strongylopus from Amietia to have happened 53 Mya (95% HPD 26.8-83.4 Mya), during the middle Eocene. At this time all major rifting, igneous intrusions and mountain building had ceased (McCarthy & Rubidge, 2005), when the vegetation in Africa was mostly woodland in areas with intermediate rainfall. The Orange River was formed by the Kalahari River that captured the Karoo River. In the late Eocene the climate began to cool, due to the opening of the Drake Passage, which permitted the circulation of cold oceanic water as the Antarctic ice cap reached a maximum 25-35 Mya (McCarthy & Rubidge, 2005). The cold Benguela current developed at this time (Feakins & de Menocal, 2010), leading to increased southern African aridity. Strongylopus is represented mostly by highland species, with the exception of the widespread S. grayii, which occurs in a range of habitats from coastal lowlands to the southern African inland plateau. This transition from the Eocene-Oligocene ice-free greenhouse world, to an icehouse climate with the first major glaciation of Antarctica occurring around 34 Mya, lasting only 0.5 Ma (Hutchinson et al., 2021).
The earliest split separated the ancestor of the southern species from the ancestral widespread species, which may have been limited to the north. The ancestor of S. bonaespei + S. grayii + S. wageri split from the ancestor of S. merumontanus + S. fasciatus + S. rhodesianus at 27 Mya (95% HPD 15.6-41.3 Mya), during the mid-Oligocene. At around 30 Mya grasslands developed in drier areas, possibly isolating the northern and southern clades, although frog fossils are not sufficient or varied enough to understand what effect, if any, the conditions leading to this climate change had on semi-terrestrial amphibians (McCarthy & Rubidge, 2005).
The early Miocene from about 20 Mya was characterised by warmer global climates (Jacobs, 2004), with grasslands expanding, along with deserts and tundra. During the early Miocene the southwest coast and the Cape Peninsula was an open riparian forest and swamp vegetation until the middle Miocene (Sciscio et al., 2016). Uplift of the eastern side of southern Africa started around 20 Mya (McCarthy & Rubidge, 2005). Rain forest trees showed a similar split around 17 Mya (95% HPD 10.5-23.4 Mya) (Couvreur et al., 2008). The beginning of the Miocene saw the evolution of mammals into niches left vacant by the Eocene extinction event (McCarthy & Rubidge, 2005). This period was also the start of the final evolution events leading to the six species recognised today.
The central high plateau was formed by uplifting around 20 Mya (Partridge, 1997). Rifting started in east Africa during the Miocene. The uplift of the eastern side of southern Africa reduced rainfall to the interior, leading to a rain shadow to the west of the escarpment (McCarthy & Rubidge, 2005), and a rain shadow in East Africa. An arid corridor started to develop between Namibia and northeast Africa (Bobe, 2006). The reduced rainfall led to the expansion of grasslands, which allowed S. grayii to occupy seasonal wetlands. These experienced a stronger rainfall over a shorter period (McCarthy & Rubidge, 2005).
The last two splitting events occurred close together. Strongylopus bonaespei split from S. grayii + S. wageri around 22 Mya, followed by the split between S. grayii and S. wageri around 19 Mya that may have been related to the expansion of grasslands and reduced rainfall that isolated the two species in highland forest, leading to allopatric speciation. The southwest coast and the Cape Peninsula were covered in riparian forest from the early to middle Miocene (Sciscio et al., 2016). The present S. bonaespei distribution largely coincides with the reconstructed Miocene distribution of tropical/subtropical forest patches (Roberts et al., 2013).
After 16 Mya there was a climax of warmth (Flower & Kennet, 1994), the mid Miocene Climatic Optimum (Feakins & de Menocal, 2010), followed by the Benguela current becoming intensified (Flower & Kennett, 1994), with the further expansion of grasslands, associated with the growth of the Antarctic ice cap and rapid cooling (McCarthy & Rubidge, 2005). This may have isolated the northern S. merumontanus from the southern S. fasciatus. Strongylopus rhodesianus split from the ancestor of S. fasciatus + S. merumontanus around 16 Mya. The populations presently on the recent volcanic mountains of northern Tanzania are presumably successful due to orographic rainfall (Poynton, 2013), surrounded by more arid lowlands.
There was a mid-Miocene climatic optimum around 15 Mya. Subsequent climate changes include a major cooling event at 14.1-14.8 Mya, leading to the increase of the East Antarctic ice sheet, with increased seasonality, and a mixture of environments (Hill, 1987). Forests were still common in East Africa at 12 Mya (Bobe, 2006). Mid-Miocene warming was followed by cooling, which led to the retreat of tropical ecosystems and increased seasonality. The split between Strongylopus fasciatus and S. merumontanus occurred around 11 Mya, with S. fasciatus distributed in the grasslands south of the moist mountains where S. merumontanus was found. They presently overlap on the Eastern Highlands of Zimbabwe, along with S. rhodesianus. This may be the result of recent northward expansion of S. fasciatus aided by a stable climate (Cohen et al., 2007).
Intense volcanic activity started north of Lake Tanganyika at 10 Mya, which continued until 4 Mya. In East Africa an increase in grassland around 5.4-8.4 Mya was implied by a study of Annonaceae rain forest trees (Couvreur et al., 2008; Senut et al., 2009). Drier conditions were indicated by the first appearance of C4 grasses in East Africa around 10 Mya (Uno et al., 2006). Strongylopus bonaespei and S. wageri occur in forest (or where forest was recently present), and their ancestor may have been restricted to moist high elevations. An arid corridor running from Namibia to northeast Africa continued to develop during the Miocene, separating the Central African forest block from the coastal and montane forests of East Africa (Bobe, 2006).
Between 7-8 Mya the Arctic ice cap expanded, shifting climatic belts to the south (Senut et al., 2009). The Sahara became arid although parts of southern Africa remained relatively tropical. In southern Africa a strong seasonal west coast rainfall regime developed (Hoffmann et al., 2015). The C4 grass expansion that occurred around 10 Mya in East Africa only occurred much later in southern Africa (Hopley et al., 2019). This is shown by the presence of predominantly C3 vegetation in the Hoogland cave in South Africa, 5.3-7.3 Mya (Hopley et al., 2019). During the late Miocene there was increased aridity around 6 Mya (Bobe, 2006), with expansion of grasslands, and evidence of local varied habitats.
The southern African coast was uplifted about 600 m during the last 5 Ma (Partridge, 1997), resulting in the development of the Great Escarpment. East Africa 3-5 Mya was warmer and wetter (Danley et al., 2012), while in southern Africa the cold water Benguela upwelling was initiated 3-5 Mya, resulting in aridification of the adjacent west coast (Hoffmann et al., 2015).
A shift from woodland to grassland occurred just before 4 Mya (de Menocal & Bloemendal, 1995). The populations of S. fasciatus have subsequently been able to expand to the south and north, overlapping the northern S. merumontanus on the Eastern Highlands of Zimbabwe, with 3-5 Mya characterised by warmer and wetter conditions (Danley et al., 2012). In the northern parts of its range, S. merumontanus is restricted to moist highlands. Many of these highlands, such as Mt. Kitumbeine, Mt. Meru and Mt. Kilimanjaro, are relatively recent in origin, having formed in the last two million years (Nonnotte et al., 2008).
The warmest phase of the Pliocene occurred between 3-4 Mya, with sea levels about 30 m higher than today, and with high latitudes experiencing mild climates (Partridge, 1997). A series of ice ages followed, resulting in desertification starting in East Africa (Senut et al., 2009). The present-day arid conditions along the west coast of southern Africa started around 2.8 Mya.
East Africa and the Lake Malawi basin experienced prolonged drought 135-127 ka when the surrounding watershed was a semidesert, followed by the expansion of grasslands between 120-75 ka around Lake Malawi (Cohen et al., 2007). These conditions would have continued the isolation of populations of S. merumontanus on the moist highlands of Mt. Mulanje and surrounding granite inselbergs, the Eastern Arc Mountains and the volcanic highlands of Mt. Kilimanjaro and Mt. Kitumbeine. In the last 70 ka conditions have become more humid and stable (Cohen et al., 2007).
Southern Africa has been called a ‘cradle of faunal diversity’ (Tolley et al., 2008). Strongylopus evolution took place during the Miocene, which is similar to the ages of splits found in Bradypodion chameleons (Tolley et al., 2008), Psammobates tentorius tortoises (Zhao et al., 2020), killifish in the genera Nothobranchius (van der Merwe et al., 2021) and Poropanchax (Bragança & Costa, 2019), lizards in the genus Agama (Matthee & Flemming, 2002), Platypleura cicadas (Price et al., 2007) and sengis in the genus Elephantulus (Smit et al., 2007).
Distribution: The distribution of our molecular, morphological and advertisement call samples is shown in Figs 36, 38. The presumed species ranges can be found based on museum records in Poynton (1964a, 2013), Poynton & Broadley (1985), Channing (1986), Channing & Davenport (2002), and Clarke & Poynton (2005). The overall distribution of the genus closely follows the ecosystem regions of Abell et al. (2008), which are based on freshwater fish ranges (Table 5). For example, Strongylopus species occupy many of the habitats where killifish (Nothobrachius) occur, such as headwater floodplains, and marshes, seepage areas and slow-flowing seasonal stream channels (van der Merwe et al., 2021). These ecoregions reflect drainage basins and may prove to be useful for examining distributions of amphibians. The present S. bonaespei distribution largely coincides with the reconstructed Miocene distribution of tropical/sub-tropical forest patches (Roberts et al., 2013). A broad division of Africa into biogeographical regions based on plants, birds, amphibians, reptiles and mammals (Linder et al., 2012) did not recognise the highlands occupied by Strongylopus. Poynton (2013) pointed out that the one degree scale of the Linder et al. (2012) study was too coarse to separate the Afrotemperate region of the high mountains of Tanzania where Strongylopus occurs.
The distribution of Strongylopus species in the ecoregions of Abell et al. (2008).
The distribution of genetically similar populations of S. merumontanus across the presently isolated mountain blocks of East Africa, including the West and East Usambaras, the Nguru and Nguu mountains, the Ulugurus, the Rubeho, Malundwe and Udzungwas is similar to populations of reed frogs Hyperolius that also occur on moist montane blocks of the Eastern Arc. Lawson (2010) showed that lineages within Hyperolius puncticulatus, the H. spinigularis complex, and H. mitchelli had separate histories with splits dating from 2.1-15.2 Mya, even though they showed similar distributions.
Field identification of S. fasciatus and S. merumontanus is not easy, and older museum records may be confused. Overlap in body proportions and colour patterns between species indicates that morphology alone is not a useful means of identification. The discovery of S. merumontanus on the Eastern Highlands of Zimbabwe close to S. fasciatus suggests that more detailed genetic studies are required where these two species overlap. As little as a single base difference separates some tyr haplotypes of S. fasciatus and S. merumontanus, suggesting relatively recent gene flow between these species. Further sampling may discover nuclear haplotypes shared between these two species.
The following are thanked for their contributions to this project: Matabaro Ziganira (KwaZulu-Natal Museum, Pietermaritzburg) for the loan of S. wageri material. Nkosinathi Mazungula and Roger Bills (South African Aquatic Biodiversity Institute, Makhanda) for the loan of specimens; Frank Tillack (Museum für Naturkunde, Berlin) and Simon Loader (Natural History Museum, London) are thanked for providing data and information for specimens in their respective collections. Additional tissue samples were provided by Michele Menegon, and Krystal Tolley (South African National Biodiversity Institute, Kirstenbosch). Additional field recordings were provided by Michele Menegon, Wilirk Ngalason, David Moyer, and Marius Burger. WC specifically thanks Jan Venter (ex-Eastern Cape Parks Tourism Agency – ECPTA), Brian Reeves (ECPTA) and Michael Cherry (Foundation Biodiversity Information Program Eastern Cape Forest project) for fieldwork support in ECPTA reserves, and Department of Economic Development, Environmental Affairs and Tourism for permits (nos. CRO 64/CR, CRO 84/CR). We acknowledge the Tanzanian Commission for Science and Technology (COSTECH; research permit 2010-363-NA-96-44), Tanzania National Parks (TANAPA), Tanzania Wildlife Research Institute (TAWIRI) and Kilimanjaro National Park for issuing the research permits.
We acknowledge financial support from the National Research Foundation (South Africa), the National Geographic Society grant 6475-99 to AC, and NorthWest University. GZ was financed by a grant from the German Excellence Initiative to the Graduate School of Life Sciences, University of Würzburg. Her field work was possible thanks to Ingolf Steffan-Dewenter, the DFG-Research Unit FOR1246 and the personnel at the research station in Nkweseko. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Two reviewers and the editor, Peter Schuchert, made valuable suggestions which greatly improved this report.