Three new species of spiny throated reed frogs (Anura: Hyperoliidae) from evergreen forests of Tanzania

The East African spiny-throated reed frog complex (Hyperolius spinigularis, H. tanneri, and H. minutissimus) is comprised of morphologically similar species with highly fragmented populations across the Eastern Afromontane Region. Recent genetic evidence has supported the distinctiveness of populations suggesting a number of cryptic species. We analyse newly collected morphological data and evaluate the taxonomic distinctiveness of populations. We find three new distinct species on the basis of morphological and molecular evidence. The primary morphological traits distinguishing species within the Hyperolius spinigularis complex include the proportions and degree of spinosity of the gular flap in males and snout-urostyle length in females. Other features allow the three species to be distinguished from each other (genetics). We refine the understanding of H. minutissimus which can be found in both forest and grassland habitats of the Udzungwa Mountains, and provide more details on the call of this species. Further details on ecology are noted for all species where known. Three new species are described and we narrow the definition and distribution of Hyperolius spinigularis and H. minutissimus in East Africa. The spiny-throated reed frogs have highly restricted distributions across the fragmented mountains of the Eastern Afromontane region. Given the newly defined and substantially narrower distributions of these spiny-throated reed frog species, conservation concerns are outlined.


Background
The East African spiny-throated reed frog complex (Hyperolius spinigularis, H. tanneri, and H. minutissimus) is comprised of morphologically similar species occupying isolated mountaintops across the Eastern Afromontane region. Lawson [1] provided molecular evidence supporting the recognition of these three taxa and their distinctiveness from one another. The validity of these species has never been seriously questioned, though suggestions of more than one species in Hyperolius spinigularis and H. minutissimus have been remarked upon in the literature [1,2].
Hyperolius spinigularis was described by Stevens [3] based on material collected from Mulanje in Malawi. Subsequently, Schiøtz [2] reported the presence of this species ca. 1300 km north in East Usambara Mountains in Tanzania, though questioned its taxonomic placement. Schiøtz [2] noted (p. 166) "it is questionable whether the two forms should be separated subspecifically." Schiøtz expanded on morphological differences between the populations stating "the males are of the same size, the females seem smaller in the northern sample. The breadth of the protective flap is greater than the length and often weakly bilobed in the type material, circular or almost circular in the northern sample" [2], p. 166). This initial observation of geographically distinct morphological variation cast doubt on whether the Malawi and northern Tanzanian populations were part of the same lineage or better regarded as distinct species.
Since its initial description, Hyperolius spinigularis has been documented to occur across a much larger range in East Africa beyond Malawi and East Usambara. This includes Udzungwa [4], Nguru [5] and Uluguru [1]. However, the record provided by Schiøtz and Westergaard [4] of H. spinigularis in the forests of the Udzungwa was incorrect. The Udzungwa species record is Hyperolius minutissimusa species Schiøtz and Westergaard [4] described. H. minutissimus is found across forest and grassland sites in the Udzungwa. Tellingly, Schiøtz and Westergaard [4] (p. 8) noted about their Udzungwa forest material: "A chirping voice, acoustically similar to that of H. minutissimus, was noted from specimens kept in plastic bags." Also noting on their collection (p. 8), "The new material seems in these characters closest to the southern population; the gular flap is slightly heart-shaped with width greater than length in most specimens" which is also in line with the broader proportions of the gular flap in H. minutissimus. Beyond the Udzungwas, Portik et al. [6] reported H. cf. spinigularis from the Namuli massif in Mozambique, and Lawson [1] detailed the occurrence of H. spinigularis from the Uluguru in Tanzania. Overall the range of species referable to H. spinigularis has been recorded from fragmented and distant locations across the elements of the Afromontane region of East Africa. The species H. minutissimus was described by Schiøtz [2] 10 km West from Njombe in the vicinity of the Southern Highlands with populations also recorded further north from the grasslands [4] and forests [1] of the Udzungwa.
Lawson [1] outlined considerable genetic diversity in the species H. minutissimus and H. spinigularis above the species level. This included samples from recent surveys in Tanzania in the Rubeho and Livingstone Mountains and in Mozambique (Namuli) by the authors of this paper. Analyses by Lawson et al. [7] provided further evidence of these divergences, which sampled across known populations and extended the geographic scope of these analyses.
These recently discovered genetically divergent lineages also prove to be diagnosable using morphological characters, and here we provide formal species descriptions and a revised identification key for these lineages. We reassess the geographic distributions of the spinythroated reed frogs based on these discoveriesclarifying previous uncertainty in the distribution of species, and address the conservation implications based on the newly defined and narrowed ranges.

Molecular data and analysis
Specimens collected from our fieldwork were fixed in either 95% ethanol or 5% formalin, and subsequently stored in 70% ethanol. Samples of muscle and/or liver were taken from representative individuals and preserved in 95% ethanol, these specimens are listed in Additional file 1. Lawson [1] and Lawson et al. [7] provide details on the approaches and genes used in this study, which include one mitochondrial gene and three nuclear loci (mitochondrial: NADH dehydrogenase subunit 2; nuclear: POMC, C-myc and Rag-1). In Lawson et al's [7] publication their Table 1 provide details on samples included in this study, including their origin and associated Genbank numbers. Phylogenetic relationships were estimated between all individuals using likelihood and Bayesian methods, including BEAST, RAxML, and BPP [8][9][10], using data from Lawson et al. [7]. Species trees were constructed based on these trees and by using species delimitation methods in *BEAST and BPP [10,11]. To examine species boundaries across the reconstructed phylogeny we applied three species delimitation methods: a Bayesian implementation of the General Mixed Yule-Coalescent model ("bGMYC" package v. 1.0.2 for R, [12]) using trees from the BEAST analysis, a Bayes Factor species Delimitation (BFD; [13]) to compare alternative scenarios for the H. minutissimus, H. spingularis, and H. tanneri species complex using alternate *BEAST species trees (See Additional file 1: Table S1), and a joint estimation of the species tree and species delimitation in BPP3 [10]. Each of the coalescent species tree analyses were run twice.

Morphology
Material was examined in the following institutions: The Natural History Museum, London (BMNH); Field Museum, Chicago (FMNH); California Academy of Sciences, San Francisco (CAS); Museo Tridentino di Scienze Naturali, Trento (MTSN), and University of Dar es Salaam (UDSM) (see in Additional file 1: Table S1). Morphological measurements were taken using dial callipers, to the nearest 0.1 mm using Mitutoyo Absolute Digimatic Calipers (CD-6"C) with the aid of a Leica MZ8 stereo microscope (Leica Microsystems GmbH, Wetzlar, Germany). Only fully-grown specimens (adult color pattern and adult size) were measured. Sex was determined by the presence or absence of gular flap in adult specimens. . Furthermore, qualitative characters were investigated: gular shape, proportions, and spinosity to assess species differences. In order to assess the morphometric distinctness of these species we also conducted Principal Component analyses on log-transformed data using various packages in R [14][15][16]. See Lawson et al. [7] for further explanation of these results.

Acoustic information
Advertisement calls were recorded with an Olympus LS-10 PCM digital stereo audio recorder equipped with a Sony directional microphone. The calls were analyzed using the software package Raven 1.2 [17].

Phylogeny and species delimitation
All methods agreed on the optimal resolution of the evolutionary relationships within the clade of spinythroated frogs, which consists of a well-supported monophyletic assemblage ( Figure 1 Table S1) (see also Lawson, et al. [7].

Morphology
Measurements for specimens are given in Additional file 2. Summary statistics of each species and characters are given in Table 1 Table 2). PCA analysis of males shows largely overlapping results, though H. burgessi sp. nov. and H. minutissimus are distinct (see Lawson et al. [7]). PC1 is evenly representative of all traits. PCA analysis of females shows large areas of overlap, though the two H. ukwiva sp. nov. and two H. tanneri individuals are outside of the centroids of overlap for other species. PC1 represents 62% of variance and is even across traits with strongest influence from head width, snout-urostyle length, and leg measurements (see Supplementary Data Figure S3 and Table three in Lawson et al. [7]). Discriminant function analysis of males showed complete segregation between species, and females were distinguished for all species except for H. spinigularis and H. burgessi sp. nov., which were entirely overlapping (Supplementary Table S3, Lawson et al. [7]). Differentiation in males is largely driven by height of the gular flap (HGF). In females, differentiation is strongly dominated by tibia length and snouturostyle length.
Paratypes We restrict paratype material to localities within the East Usambara on the basis that further  , and is minimally 2% pairwise divergent from its closest relative based on mtDNA (ND2. Table 3; see Figure 1). Hyperolius burgessi also has a largely allopatric distribution with respect to other species in the complex ( Figure 5).
Description of holotype Small to moderate sized hyperoliid. Pupil horizontal. Snout blunt slightly rounded. Canthus rostralis angular, being slightly convex on the horizontal plane and slightly concave on the vertical plane. Distance between eyes is 4.1 mm and the inter orbital distance is 2.4 mm. The inter-narial distance is 2.0 mm, greater than narial distance to the eye (1.8 mm). The nostril to snout is 1.0 mm. The width of head (6.9 mm) equaling 0.36 length of body (19.2 mm). The gular flap width is less than (5.0 mm) the height (5.1 mm). The gular flap is rounded, thickened and not bilobed, anteriorly narrowing so that overall shape is wide based hexagon. It is marked by black asperities (ca. 80) evenly distributed across the whole of the gular flap. Some asperities, sparsely distributed, though more concentrated, on the apex of the chin (mentum). Tibiotarsal articulation of the adpressed hind limb reaching the eye. Tibio-tarsal (9.1 mm) is almost equal to thigh length (8.5 mm). The tibiable fibulare length is 5.9 mm. The toes have expanded fleshy discs with the foot being 7.8 mm. Webbing is extensive reaching the base of the fleshy discs on all toes apart from the first toe where it only reaches the first tubercle. The forelimb length is 4.2 mm, less than the hand length (5.0 mm). The hands have expanded, rounded fleshy discs. Webbing just reaching distal subarticular tubercle of the outer finger, reaching distal subarticular tubercle of the 4th toe on both sides. Dorsal skin surface granular with a single minute black asperites surmounting many of the granules. Ventral skin surface strongly granular with black asperities restricted to the mentum, gular disc, abdomen and undersurfaces of the femur. Ventral asperities much more prominent than those of the dorsum.
Paratypes Head and body proportions in close agreement with those of the holotype (Figure 2, Table 1). The distribution of the asperities of the males are in close agreement with that of the holotype. The proportions of the gular flap, diagnostic for the species, shows some variation which means care needs to be taken in applying this character ( Figure 2). Webbing of all the material conforms to that of the holotype. Material from the Uluguru mountains (FM 274258-60) is strongly dehydrated and this might have impacted the morphometric measurements. Uluguru material show extreme values for gular flap proportions. Freshly collected material will be necessary to assess the morphometric variation among populations that might potentially recognize one of these populations as being distinct. Given the large amount of molecular difference in Uluguru populations (1.7% mtDNA (ND2) pairwise divergence from joint East Usambara and Nguru populations) the population may be a candidate for a new species.
Colour patterning of adults in life See Figure 6 for photo in life. Generally, the females and males resemble the holotype in basic colorationnot showing dichromatic patterns from samples collected. The brown dorsal chromatophores varied in intensity from specimen, and varied within and between population (Nguru and East Usambara for which large series exist) (see Additional file 1: Figure S1 and S2). It should be noted that the Nguru material was prepared differently and much more recently than most of the East Usambara material compared. Some differences might reflect preservation differences. The intensity of chromotophores sometimes resulted in dark brown mottling (particularly in East Usambara materialsee Additional file 1: Figure S1). The majority of specimens either had lateral dark edged white stripes (either thin or irregular in size and outline) ending anteriorly in a narrow stripe meeting at the snout or a triangle covering the snout. The ventral side is of a lighter cream coloration.
Sexual dimorphism Females attaining a much larger size than the males (Figure 2). Asperities of the dorsum are less visible in the female and absent from the ventral side in females. Males are easily distinguished from the females during the breeding season by their characteristic rounded, slightly narrowing anteriorly gular sac (Figure 4).

Advertisement call No advertisement call is known but
Stevens [3] reported the males making a "weak, rasping high-pitched "tcheek-tcheek" call, believed to serve a territorial function in the close relative H. spinigularis occurring in Malawi and Mozambique. Stevens [3] questioned whether this might also be an advertisement call but speculated that the restricted breeding area and season in this species might have "obviated the necessity of a mating call." Vonesh (in litt.) conducted intensive survey of this species over a two-year period in Amani Nature Reserve and never heard calling males, which is suggestive of a lack of advertisement or territorial calls in H. burgessi.
Etymology The species is named for Prof. Neil Burgess, who has made and continues to make enormous efforts towards conserving Tanzania's forests, which this new species survival depends upon. The species is also restricted to the Eastern Arc Mountain regionan area which Neil has particularly devoted considerable time and energy to understand and preserve.

Distribution and conservation
The species is known to occur in East Usambara, Nguru and Uluguru Mountains in Tanzania ( Figure 5, Table 4). The species has been collected at high altitudes in, and on the edges of submontane forests. Vonesh [18] (Figures 2 and 4). Furthermore, the shape of the gular flap in males is different from H. burgessi in being wider than the height and shaped more equally in the anterior and posterior ends of the flap (Figures 2 and 7, Tables 1  and 2). Based on molecular comparisons the species is also genetically distinct from close relatives, and is minimally 5.7% pairwise divergent from its closest relative, based on mtDNA (ND2) ( Table 3; see Figure 1). Hyperolius davenporti has an allopatric distribution with respect to all other species in the complex ( Figure 5).
Description of holotype Small to moderate sized hyperoliid. Horizontal pupil. Snout blunt slightly rounded. Canthus rostralis angular, being slightly convex on the horizontal plane and slightly concave on the vertical plane. Distance between eyes is 4.0 mm and the inter orbital distance is 2.4 mm. The inter-narial distance is 1.9 mm, almost subequal to the narial distance to the eye (2.0 mm). The nostril to snout is 1.0 mm. The width of head (7.1 mm) equaling 0.37 length of body (19.1 mm). The gular flap is wider (5.3 mm) by 1.10, than it is in height (4.8 mm). The gular flap is rounded, thickened, and not bilobed. It is marked by black asperites (ca. 50) evenly distributed across the whole of the gular flap. Some asperites, sparsely distributed, though more concentrated, on the apex of the chin (mentum). Tibio-tarsal articulation of the adpressed hind limb reaching the eye. Tibio-tarsal (9.3 mm) is almost equal to thigh length (9.0 mm). The tibiable fibulare length is 5.8 mm. The toes have expanded fleshy discs and the foot is 6.2 mm. Webbing is extensive reaching the base of the fleshy discs on all toes apart from the first toe where it only reaches the first tubercle. The forelimb length is 4.6 mm, less than the hand length (5.5 mm). The hands have expanded, rounded fleshy discs. Webbing just reaching distal subarticular tubercle of the outer finger, reaching distal subarticular tubercle of the 4th toe on both sides. Dorsal skin surface granular with a single minute black asperities surmounting many of the granules. Ventral skin surface strongly granular with  black asperities restricted to the mentum, gular disc, abdomen and undersurfaces of the femur. Ventral asperities much more prominent than those of the dorsum.
Paratypes Head and body proportions are in close agreement with those of the holotype (Figure 2, Table 1).
The distribution of the asperities of the males are in close agreement with that of the holotype. The proportions of the gular flap, diagnostic for the species, show some variation which mean the boundaries of diagnosing this species are in some cases slightly overlapping ( Figure 2). Webbing of all the material conforms to that Colour patterning of adults in life See Figure 6 for photo in life. Generally the females and males resemble the holotype in basic coloration. The brown dorsal chromatophores varied slightly in intensity amongst specimens (see Additional file 1: Figure S3). The intensity of chromotophores sometimes resulted in dark brown mottling. The majority of specimens either had lateral dark edged white stripes (either thin, broad and regular in shape or irregular in size and outline) ending anteriorly in a narrow stripe meeting at the snout or a triangle covering the snout. The ventral side is of a lighter cream colouration.
Sexual dimorphism Females attaining a much larger size than the males (Figure 2). Asperities of the dorsum weaker in the female and absent from the ventral side in females. Males are easily distinguished from the females during the breeding season by their characteristic rounded and wide gular sac ( Figure 4).

Advertisement call
Neither advertisement or territorial calls in H. davenporti were recorded or heard, however survey time in the area (5 days) was relatively short. Little can be concluded on whether it resembles its congeners H. burgessi or H. spingularis as no call is known from these species.
Etymology The species is named after Dr. Tim Davenport, who has made substantial contributions towards conserving Tanzania's forests, in particular the Southern Highlands and Livingstone Mountains of Tanzania. The Livingstone Mountains are the only known locality of this species.

Distribution and conservation
The species is only known from Sakara Nyumo Forest Reserve, Livingstone Mountains, Southern Highlands ( Figure 5, Table 4). Specimens were collected in shallow ponds on the forest   Tables 1 and 2). Based on molecular comparisons the species is also genetically distinct from close relatives, and is minimally 7.0% pairwise divergent from its closest relative, based on mtDNA (Table 3; see Figure 1).   Hyperolius ukwiva has an allopatric distribution with respect to all other species in the complex ( Figure 5). Paratypes Head and body proportions in close agreement with those of the holotype (see Figure 2; Table 1; Additional file 1). The distribution of the asperities of the single male (see comment below) are medially and anteriorly concentrated on the gular flap (see Figure 4). Webbing of all the material conforms to that of the holotype.

Description of holotype
Colour patterning of adults in life See Figure 6 for photo in life. Generally the female and male resemble the holotype in basic coloration. The dorsum is described in field notes as being "brown with two light grey-brown stripes from nose to the hindlegs" for all three specimens. The ventrum is described as sunshine "yellow". The legs and arms are similarly colored dorsally and ventrally.
Sexual dimorphism Females attaining a much larger size than the males (Figure 2). Asperities of the dorsum weaker in the female and absent from the ventral side in females. Males are easily distinguished from the females during the breeding season by their characteristic gular sac ( Figure 4).
Advertisement call No calls were detected or recorded during collection of these three specimens, only one of which was a male.
Etymology The species is named after the forest area (Ukwiva) from where the type series was collected. The specific epithet is considered to be a noun in apposition.
Distribution, ecology and conservation The species is only known from Ukwiva Forest Reserve, in the Rubeho Mountains ( Figure 5; Table 4). Specimens were collected in and around the edge of montane forest. Collecting across the Rubeho Mountains, although only relatively recent, has been quite extensive (Rovero,et al. [19] so its localized distribution might not just be a function of restricted sampling. Comment It should be noted that we would have preferred to describe the holotype as a male (e.g. KMH 36053), in line with the designation of males for all other members of this group of hyperoliids. However, the only known male paratype (KMH 36053) currently cannot be located at UDSM or MTSN, and thus measurements were not possible. Data on this male were taken from photographs and observations made in Tanzania at the time (LL and MM).

Further remarks on Hyperolius minutissimus
Distribution and conservation Commenting on the distribution and ecology of Hyperolius spingularis Schiøtz [20] (p.180) stated "a search for it at the Udzungwas revealed it both in forest and in very open farmland, and it may be found wherever suitable habitats exist in the eastern Tanzanian-Malawi highlands." According to our records and based on the clustering of Schiøtz's material with our samples of H. minutissimus, we can confirm that these comments refer only to H. minutissimus, with H. spingularis not recorded from the Udzungwas. Therefore the records given by Schiøtz and Westergaard [4] refer to H. minutissimus. On p.167 Schiøtz [20] reports samples collected between Kilosa and Dabaga and these refer to localities reported in Schiøtz and Westergaard [4] and are limited to Udzungwa areaand not beyond, towards Kilosa, that might suggest potential presence in Rubeho of H. minutissimus. Currently, we only record the presence of H. ukwiva from Rubeho mountain region. H. minutissimus is found in forest and grassland habitats in the Udzungwa Mountains and Njombe, the latter in the region of the Southern Highlands. A wider distribution of H. minutissimus in the Southern Highlands and beyond into Malawi is as yet unconfirmed and future sampling will be required to establish if it occurs in these areas.