New data on Garra makiensis (Cyprinidae, Labeoinae) from the Awash River (Ethiopia) with remarks on its relationships to congeners on the Arabian Peninsula

Abstract On the African continent, the genus Garra consists of several species often insufficiently separated from each other by diagnostic characters. Herein, a detailed morphological redescription of Garra makiensis from the Awash River drainage is presented, together with additional data on the type specimens of G. makiensis and G. rothschildi. Mitochondrial CO1 sequence data are also provided, including the historic paralectotype of G. makiensis, with a comparison to Garra species from Africa and the Middle East. Based on these sequences, G. makiensis clusters outside the group of African congeners and is a sister lineage to species from the south-east of the Arabian Peninsula. Although morphologically variable, G. makiensis is characterised by having a single unbranched pectoral-fin ray, a short distance between vent and anal-fin origin (7.3–19.7 % of pelvic – anal distance), chest and belly covered with scales, and a prominent axillary scale at base of pelvic fin (18.8–35.5 % of pelvic-fin length).


Introduction
The endorheic Awash River drainage in the northern part of the Main Ethiopian Rift (MER) is subdivided into two freshwater ecoregions, the Ethiopian Highlands and the Northern Eastern Rift (Abell et al. 2008). It originates close to Ethiopia's capital city Addis Ababa at an altitude of > 3,000 m a.s.l. Along its course (1,250 km in length), it flows from the highlands into the MER, and drains into saline Lake Abbe at the Ethiopian-Djibouti border. Numerous smaller sub-drainage systems, among them the Gotta River, belong to the Awash catchment. Biogeographically, the region is classified as part of the Abyssinian Highlands ichthyofaunal province (Roberts 1975) (a subprovince of the Nilo-Sudanic province according to Snoeks and Getahun (2013)) or the Ethiopian Rift Valley province (Paugy 2010). Evidence for ichthyofaunal affinities with the Nile River system and the Central MER were recently provided by Beshera and Harris (2014) for the Labeobarbus intermedius complex and by Englmaier et al. (2020a) for small-sized smiliogastrin barbs (Enteromius Cope, 1867).
The fish fauna of the Awash is commonly described as "impoverished" (Roberts 1975: 291) with 10-13 species belonging to five families (Golubtsov et al. 2002;Englmaier et al. 2020aEnglmaier et al. , 2020b. One of the poorly investigated groups of freshwater fishes in the region is the Afro-Asian genus Garra Hamilton, 1822. Menon (1964 and later Getahun (2000) and Stiassny and Getahun (2007) provided the first comprehensive morphological studies on African Garra. Twenty-four valid species of Garra are currently recognised in Africa (Moritz et al. 2019). They are distributed from North Africa (Nile River in Egypt) to drainage systems in West Africa (e.g., Senegal River) and central sub-Saharan Africa (Tanzania and Angola) (Daget et al. 1984;Getahun 2000;Stiassny and Getahun 2007;Habteselassie et al. 2010;Moritz et al. 2019). With 12 species, the genus was found to be particularly diverse in the Ethiopian Highlands and surrounding drainage systems (Stiassny and Getahun 2007). In the Awash River, recent surveys gave evidence for three well supported mitochondrial clades of Garra, identified as: G. aethiopica (Pellegrin, 1927), G. dembeensis (Rüppell, 1835) and G. makiensis (Boulenger, 1903) by Englmaier et al. (2020b).
Compared to Garra species in Asia and the Middle East (e.g., Krupp 1983;Yang et al. 2012;Sayyadzadeh et al. 2015;Esmaeili et al. 2016;Hashemzadeh Segherloo et al. 2016;Nebeshwar and Vishwanath 2017;Kirchner et al. 2020;Kottelat 2020), systematic relationships and diagnostic morphological characters of African taxa have not been well investigated (Stiassny and Getahun 2007). So far, no consistent opinion has been reached in assessing important diagnostic characters such as the presence/ absence of a red or black blotch behind the upper edge of operculum, the scale pattern on ventral side, the size and shape of the gular disc (also referred to as "mental adhesive disc" in Zhang et al. (2002)) or the tuberculation on snout and head (Menon 1964;Getahun 2000;Golubtsov et al. 2002Golubtsov et al. , 2012Stiassny and Getahun 2007). Together with recent reports of considerable intraspecific morphological variability Englmaier 2018), this complicates species-level determination and has led to different taxonomic opinions and inconsistent distribution records of many African species (see Getahun 2000;Golubtsov et al. 2002Golubtsov et al. , 2012Stiassny and Getahun 2007;Habteselassie 2009Habteselassie , 2012Moritz et al. 2019).
One such problematic species, G. makiensis, was described from the Meki River (endorheic basin of Lake Ziway) in the Central MER (Boulenger 1903). In the first comprehensive revision of the genus, Menon (1964) provided a redescription of this species and included G. rothschildi (Pellegrin, 1905), described from the Gotta River in the Northern MER, as a synonym. This opinion was later corroborated by Getahun (2000) and Stiassny and Getahun (2007), who extended the distribution range of G. makiensis to the Southern MER, the Blue and White Nile, and the Omo River drainage. However, during recent surveys in the Awash River, preliminary observations showed that diagnostic characters described for G. makiensis in recent literature and identification keys (e.g., Stiassny and Getahun 2007;Habteselassie 2012) contain uncertainty and did not allow reliable species identification (Englmaier 2018).
Therefore, as a first step towards resolving taxonomic inconsistencies among African Garra, we present a detailed redescription of G. makiensis based on specimens from the Awash River drainage with new data on the type specimens of G. makiensis and G. rothschildi. Moreover, we provide mitochondrial CO1 sequence data for Garra species from the Awash River (G. aethiopica, G. dembeensis, G. makiensis), and the first CO1 sequence of the historic paralectotype of G. makiensis (BMNH 1905.7.25.88) in order to evaluate their phylogenetic relationships. These considerations are complemented with a morphological comparison of G. makiensis with closely related species and remarks on biogeographical implications.

Materials and methods
Specimens of Garra were collected in the Awash River, including its major tributaries (Fig.  1 and Table 1, Suppl. material 1: Table S1). Collections were made during the dry seasons between 2017 and 2019. Sampling methods are described in Englmaier et al. (2020a). After anaesthesia with etheric clove oil (Eugenia caryophyllata) diluted in water, fish specimens were fixed in 6 % pH neutral formalin (later stored in 75 % ethanol) or 96 % ethanol.
Museum samples included specimens deposited in the collections of the Natural History Museum Vienna (NMW; Fig. 1: sampling sites S6-S16, T1, T3-T4); the British Museum of Natural History (BMNH; Fig. 1: sampling site M1); the American Museum of Natural History (AMNH; Fig. 1: sampling site G1); the Muséum national d'Histoire naturelle, Paris (MNHN); and the Musée royal de l'Afrique centrale, Tervuren (MRAC). Comparative material is listed in Table 1. In the present study, we refer to the species names Garra smarti Krupp & Budd, 2009 and Garra sindhi Lyon, Geiger & Freyhof, 2016, although the specific epithet of both species was recently 'corrected' to smartae and sindhae by Kirchner et al. (2020). Krupp and Budd (2009) and Lyon et al. (2016) dedicated the species names to two different women, using the masculine genitive ending -i, instead of the common feminine genitive ending -ae. However, the names smarti and sindhi are not to be considered incorrect according to Art. 32.5 of the International Code of Zoological Nomenclature (1999) and are therefore not to be modified (see also Dubois (2007) and Nemésio and Dubois (2012) for species names derived from personal names).  (Bonini et al. 2005). Study sites: S1-S6 and T1-T8 Awash River drainage, G1 Gotta River (type locality of G. rothschildi), M1 Meki River (type locality of G. makiensis), Z1-Z2 Lake Ziway, and L1 Lake Langano. Sampling sites in red showing localities where G. makiensis is known from examined material or was recorded during recent surveys. The small inserted map showing known distribution range of G. makiensis in the Awash and Meki River drainages.

Morphological analyses
In total, 124 specimens were examined, including type specimens of G. makiensis and G. rothschildi. A maximum of 43 measurements (seven for the gular disc), 22 external body counts, and nine axial skeleton counts (from x-rays) were taken. Type specimens of G. rothschildi were examined from photographs and radiographs, and only meristic counts were taken. Measurements and counts are defined in Suppl. material 1: Table  S2; and measurements illustrated in Fig. 2. Most measurements follow Hubbs and Lagler (1958) and Holčík et al. (1989) and were made point to point using a digital calliper to the nearest 0.1 mm. Head length (HL) excludes the skin fold on the operculum. Length of the axillary scale was measured from the anteriormost to the posteriormost extremity. Length of the dorsal-fin rays was measured from the visible base of the ray to the end of the uppermost flexible part. We refer to the postpelvic region as an area on the ventral side between the insertion of the pelvic fins and the anterior margin of anus. Scales in the postpelvic region were counted along midline. The terminology used for the external oral and gular morphology, including the gular disc (referred to as "disc" or "mental adhesive disc" in Zhang et al. (2002) and Stiassny and Getahun (2007)) follows Kottelat (2020). Measurements of the gular disc were done as follows ( Fig. 2 and Suppl. material 1: Table  S2) Maximum width of pulvinus between lateral extremities. The terminology used for nuptial tubercles and grooves on the snout follows Nebeshwar and Vishwanath (2017) as described for Garra.
External meristic counts follow Skelton (1980) and those summarised in Englmaier et al. (2020a) (Suppl. material 1: Table S2). The posterior two branched rays in the dorsal and anal fins, located on the last complex proximal pterygiophore of the fin, were counted as two. As the anteriormost unbranched rays of the dorsal and anal fins are usually deeply embedded, ray counts for those fins were taken from radiographs. Total number of lateral-series scales were counted from the first scale behind the opercular opening to the last scale on the caudal fin (bearing the lateral-line canal or without the canal). Counts and terminology of the axial skeleton follow Naseka (1996). Vertebral counts and supraneural bones were examined from radiographs.
For DNA extraction of historic museum material (BMNH 1905.7.25.88, G. makiensis, paralectotype) we used tissue from the branchial arches (right side of the specimen). DNA was extracted using the QIAamp DNA Mini and Blood Mini Kit (Qiagen) following the manufacturer's protocol. Final DNA concentration was 23.4 ng μl -1 . All lab work was performed in a DNA clean room with sterilised and UV radiated utensils. Because museum DNA is typically fragmented, we designed specific primers to amplify approximately 150 bp long fragments of the cytochrome c oxidase subunit 1 (CO1) ( Table 2). Primers were designed based on the CO1 alignment of the extant Garra samples included in this study using Primer-BLAST (NCBI), and were arranged in a way that adjacent fragments extensively overlap. PCR reactions were done in 50 μl, with 5 μl buffer, 4 μl MgCl 2 (2.0 mM), 2 μl Enhancer, 1 μl dNTPs (500 μM), 0.50 μl of each primer (50 pmol μl -1 , for primer sequences see Table 2), 0.4 μl of AmpliTaq Gold 360 DNA Polymerase (1 unit) and 2-3 μl of DNA. The same touch-down PCR protocol was used for all primer pairs, with initial denaturation at 95 °C for 10 min, followed by five cycles at 95 °C for 30 s, 63 °C for 2 min and 72 °C for 45 s, and 40 cycles at 95 °C for 45 s, 61 °C for 45 s and 72 °C for 45 s. Final extension was performed at 72 °C for 7 min. PCR products were purified with a Qiagen PCR purification kit, and purified PCR products were sequenced (in both directions) by Microsynth with PCR primers. After amplification, fragments were aligned with MEGA6 (Tamura et al. 2011) and composed into a single sequence (451 bp total length).
The dataset used for phylogenetic analysis comprised 13 original CO1 sequences (611 bp) from freshly collected samples of G. aethiopica, G. dembeensis and G. makiensis, and the historic paralectotype sequence (451 bp). Obtained sequences were deposited on GenBank under accession numbers MT946118-MT946130. Additionally, we included 46 sequences retrieved from GenBank, corresponding to different Garra species from Africa and the Arabian Peninsula. An Asian Garra species was included as outgroup following Yang et al. (2012) (Suppl. material 1: Table S6).
Sequences were edited in MEGA7 (Kumar et al. 2016) and aligned with ClustalW. PartitionFinder2 (Lanfear et al. 2017) was used to estimate the best partition scheme and best fit substitution model. Phylogenetic reconstruction was conducted using Maximum Likelihood (ML) and Bayesian Inference (BI) methods. ML was performed using RAxML-HPC2 Workflow on XSEDE (v. 8.2.12) through CIPRES Science Gateway (Miller et al. 2011), with 1,000 bootstrap replicates. For BI, MrBayes v. 3.2.6 (Ronquist et al. 2012) was used. Two independent Markov Chain Monte Car-lo (MCMC) were run simultaneously for 5 million generations with sampling trees every 500 generations. The first 25 % of obtained trees was discarded as 'burn-in'. All computed trees were visualised and edited with FigTree v. 1.4.4 (Rambaut 2012) and Inkscape v. 0.92.3. MEGA7 (Kumar et al. 2015) was used to calculate between group mean distances (uncorrected p-distances).
Two different sequence alignments were used for calculating the p-distances: the first alignment did not include the paralectotype sequence of G. makiensis and each sequence had a length of 611 bp; the second dataset was trimmed to the length of the paralectotype sequence (451 bp) and was used to compare the paralectotype with the remaining studied species.

Results
Identification of the Awash samples as G. makiensis is based on 1) the morphological comparison to type specimens of G. makiensis and G. rothschildi (Tables 3, 4), including multivariate statistical analyses (PCA, DFA) of 17 morphometric and six meristic characters with a comparison to G. aethiopica and G. dembeensis from the Awash River (Fig. 3); and 2) CO1 sequence data with a p-distance of 0.08 % between the paralectotype and the samples of G. makiensis from the Awash River, and 9.53-11.31 % between the paralectotype and other African species (including G. aethiopica and G. dembeensis from the Awash).

Primer Sequence (5'-3') Product (incl. primers)
A similar pattern of morphological differences is supported by DFA. Variables that contribute most for discrimination of the samples are vent distance (% pelvic -anal distance), dorsal-fin depth (% SL), total number of lateral-series scales, and number of branched pectoral-fin rays. Predicted classifications for the samples from the Awash (G. aethiopica, G. dembeensis, G. makiensis) and the type specimens of G. makiensis were 100 % correct, with the exception of one specimen identified as G. aethiopica falling within the group of G. dembeensis (Suppl. material 1: Table S5).

Taxonomy
Garra makiensis (Boulenger, 1903 Fig. 7 for tubercles on head; Fig. 8 for scales on chest and shape of gular  BMNH 1905.7.25.87, paralectotype: BMNH 1905. Analyses are based on 17 morphometric and six meristic characters as given in Suppl. material 1: Table S3. disc; and Fig. 9 for axial skeleton and shape of supraneural bones. Tubercles on scales and the pectoral fin, and scale pattern on ventral side are shown in Suppl. material 2: Figs S1, S2. Measurements and counts are given in Tables 3, 4. Longest examined specimen 147.9 mm SL (female, NMW 99504). Body elongated, moderately compressed, more in the caudal region. Shape of body and head very variable. Dorsal head profile slightly convex, its transition to back usually smooth, in few specimens with a slight nuchal hump. In most specimens, predorsal back outline rises gently, slightly convex or straight, to dorsal-fin origin. Postdorsal profile slightly concave to caudal-fin origin. Caudal peduncle almost twice as long as its minimal depth. The vent is close to the anal-fin origin (7.3-19.7 % of pelvic -anal distance). Head usually as long as body depth at dorsal-fin origin. Head depth at nape shorter than head length. Snout blunt and longer than orbit -operculum distance. Transverse groove weakly developed (absent in some specimens); transverse lobe separated from lateral field by a shallow groove (or without groove). Two deep grooves originating above anterior barbel, posteriorly not connected; at posterior end of upper groove a patch of few tubercles in some specimens (Fig. 7).
Tubercles on snout and head in both males and females (smallest specimen with tubercles: 45.4 mm SL, Awash River, S14), but often completely absent or rudimentary developed (Figs 6, 7). Transverse lobe with large conical tubercles; tubercles extending to lateral surface and the area between anterior rim of eyes and nostrils. Depressed rostral margin usually without tubercles; in some specimens few and irregularly placed. Anterior extremity of the ethmoid field often elevated from depressed rostral surface and covered with large tubercles, especially in anterior region. Small tubercles are commonly spread on the frontal and occipital regions, sometimes extending to the operculum ( Fig. 7 and Suppl. material 2: Fig. S1). In few specimens (n = 7), small circular tubercles on scales  in the predorsal region and the lateral side of the abdominal region (above lateral line). A single specimen (NMW 99231, male, 91.2 mm SL, Awash River, S10) with tubercles on dorsal side of the pectoral fins (at fin membranes) (Suppl. material 2: Fig. S1). Gular disc well-developed but often variable in size and shape (Fig. 8). Its width greater than its length. Width of torus less than disc length. Pulvinus wider than long and with few papillae. Labrum well-developed and longer than torus. Width of mouth usually less than snout length. Abundant papillae on rostral cap, torus, labellum and labrum. Rostral cap with invecked ventral margin. Two pairs of barbels, their length usually shorter than eye diameter; anterior barbel slightly longer or about equal to posterior barbel. Dorsal fin with 3 or 4, commonly 4, unbranched and 8 branched rays, its last unbranched ray is the longest (89.0 % of dorsal-fin depth); length of first branched ray 86.4 % of dorsal-fin depth; second branched ray much shorter (76.9 % of dorsal-fin depth). Pelvic fin with a single unbranched ray and 7-9, commonly 8, branched rays; pelvic splint present. Long axillary scale at base of pelvic fin, its length 18.8-35.5 % of pelvic-fin length. Pectoral fin with a single unbranched ray and 13-17, commonly 16, branched rays. Caudal fin forked with 2+17 principal rays. Upper procurrent rays 7 (9), 8 (49) or 9 (3), lower procurrent rays 6 (14), 7 (42) or 8 (6).
Morphological variability. Similar to other Garra species in Africa and the Arabian Peninsula (Krupp 1983;Golubtsov et al. 2012;Englmaier 2018), we found considerable morphological variability in G. makiensis (Fig. 6). Based on our data, we cannot confirm the presence of a sexual dimorphism, but the largest specimens were females (> 140 mm SL), and more males with prominent tuberculation on snout and head were found. Though few specimens were examined, and samples were collected during dry season only, our data suggest that body shape and tuberculation in G. makiensis might be (directly or indirectly) related to abiotic habitat characteristics. Specimens with a more slender body shape and without (or reduced) tubercles on snout and head (Fig.  6B) were caught in low flow velocity habitats, whereas deep bodied specimens with large conical tubercles on snout and head (Fig. 6C) exclusively occurred in high flow velocity habitats over coarse substrate. Intermediate morphs (Fig. 6A) and large growing specimens with reduced tubercles (Fig. 6D) were occasionally found.
Colouration. In life (Fig. 5): Body colour usually light grey, above lateral line often pale-brown or blueish iridescent and darker than below. Head yellowish brown, mouth and ventral side cream. Iris white and yellow. Some individuals show an indistinct, roundish, dark blotch at posteriormost caudal peduncle. At anteriormost lateral line (behind upper edge of operculum) a small dark (rarely blueish iridescent, never red) blotch, not extending on gill cover. Fin membranes usually hyaline, sometimes light grey or yellowish; on caudal fin often light orange. Fin rays hyaline or pale. Dorsal fin with four to six indistinct black blotches at base of branched rays (strongest between 3 rd and 6 th branched rays).
Distribution. Garra makiensis is endemic to Ethiopia where it is found in endorheic drainages (Awash (including the Gotta River sub-drainage) and Meki) of the Northern and Central MER (Fig. 1) (Golubtsov et al. 2002). It is absent from the headwaters and was found characteristic for the middle and lower sections of the Awash River (Englmaier et al. 2020b). In the current study, we cannot confirm the presence of G. makiensis in the Meki River, its type locality. The Meki drainage is highly altered by human impacts (e.g., water abstraction, sand mining) and the last records of G. makiensis in this drainage date back to 1984 (Golubtsov et al. 2002). The extended distribution range reported by Stiassny and Getahun (2007), including the southern part of the MER, the Blue and White Nile drainages and the Omo River drainage, contains uncertainty and needs clarification (Wakjira and Getahun 2017).

Molecular analyses
The alignment used for BI and ML phylogenetic reconstructions comprised 59 CO1 sequences of a length of 611 bp, and one sequence (BMNH 1905.7.25.88, G. makiensis, paralectotype) with a length of 451 bp. The alignment included nine individuals of G. makiensis, two individuals of G. dembeensis and two individuals of G. aethiopica, all from the Awash River. Forty-six other sequences of Garra species from Africa and the Middle East were included to resolve the phylogenetic relationships of the Awash species (Fig. 11). Garra makiensis clusters together with species from the south-western Arabian Peninsula (G. tibanica Trewavas, 1941, G. buettikeri Krupp, 1983, G. dunsirei Banister, 1987 forming a monophyletic group (Bayesian posterior probability, BPP 0.96; bootstrap value, bs 86; Clade A, Fig. 11). The lineage of G. Figure 6. General appearance and morphological variability of Garra makiensis from the Awash River. A NMW 99485, female, 85.8 mm SL, Adayitu (S13), B NMW 99485, female, 82.2 mm SL, Adayitu (S13), C NMW 99491, male, 106.1 mm SL, Dubti (S14), D NMW 99504, female, 147.9 mm SL, Jara River (T3). makiensis appears as a strongly supported sister lineage to all remaining species within Clade A (BPP 1; bs 100). Pairwise distances between G. makiensis and Arabian Peninsula species range from 9.51 % to 10.23 % (Suppl. material 1: Table S7). Garra makiensis is clearly distinct from congeners in the Awash River. Pairwise distance between G. makiensis and G. aethiopica is 9.90 % and 9.78 % between G. makiensis and G. dembeensis. The paralectotype of G. makiensis clusters together with our samples of G. makiensis from the Awash River (p-distance 0.08 %, Suppl. material 1: Table S7), corroborating the identification of the Awash population as G. makiensis.  Fig. 6B, B and E same specimen as in Fig. 6C, C and F same specimen as in Fig. 6D.
In summary, mitochondrial CO1 data provide support that G. makiensis is more closely related to Garra species in the south-west of the Arabian Peninsula than to congeners from Africa, as all of them belong to a different, well-supported, monophyletic clade (BPP 0.96; bs 77). Below we provide a morphological comparison of G. makiensis with closely related species of Clade A (Fig. 11).

Comparison of G. makiensis with congeners of Clade A.
Garra species of Clade A (Fig. 11) are currently known only from the north-east of Ethiopia (G. makiensis) and the south-west of the Arabian Peninsula (Trewavas 1941;Balletto and Spanò 1977;Krupp 1983;Banister 1987;Golubtsov et al. 2002;Krupp and Budd 2009;Lyon et al. 2016). For morphological comparison we combined all specimens of G. makiensis from Ethiopia, including the type specimens of G. makiensis and G. rothschildi, into one sample. No original material of G. buettikeri, G. dunsirei, G. smarti, and G. sindhi was examined and we refer to published data (original descriptions: Trewavas (1941), Krupp (1983), Banister (1987), Krupp andBudd (2009), andLyon et al. (2016)) for comparison. Garra tibanica includes several subspecies (Balletto and Spanò 1977) of unknown systematic relationship and taxonomic status. We, therefore, refer to G. tibanica as described by Trewavas (1941) and present new data on axial skeleton elements for the type specimens (Suppl. material 1: Table S8).  Fig. 6B, B same specimen as in Fig. 6C, C same specimen as in Fig. 6D.

Biogeographical aspects
The distribution ranges and systematic relationships of African Garra species are still poorly investigated. Our morphological and mtDNA data suggest 1) a palaeohydrolog- ical connection between the Awash River drainage and the lakes of the Central MER as G. makiensis is known from both drainage systems (see also Englmaier et al. 2020a and references therein); 2) the distinctiveness of G. makiensis in comparison to G. aethiopica and G. dembeensis from the Awash, suggesting a different evolutionary or colonisation history of Garra species in the region; and 3) a closer relationship between G. makiensis and Garra species in the south-west of the Arabian Peninsula than to African congeners.
Biogeographical similarities between the Horn of Africa and the Arabian Peninsula are evident for different animal groups (e.g., Pook et al. 2009;Zinner et al. 2009;Portik and Papenfuss 2012;Šmíd et al. 2013;Gilbert et al. 2014;Garcia-Alix et al. 2016;Yanai et al. 2020). Though dispersal routes across the Bab-al-Mandab Strait have been proposed by several studies (Šmíd et al. 2013;Stewart and Murray 2017), the exact timing is still controversial (Fernandes et al. 2006). Geological data provide evidence that the formation of the southern Red Sea rift section began in the early Oligocene with a first culmination from the upper Oligocene to the lower Miocene (30-23 Ma), followed by a geologically well-established reconnection period of Africa and Arabia at the Bab-al-Mandab-strait in the uppermost Miocene and Pliocene (10-5 Ma) (Bosworth et al. 2005;Autin et al. 2010). Postulated land bridge periods in younger times are not supported (Fernandes et al. 2006).
The restricted distribution of G. makiensis in the Northern and Central MER, and its close relationship to Garra species in the south-west of the Arabian Peninsula (based on CO1 sequence data) may support the hypotheses of dispersal events and vicariance around the southern Red Sea area. However, based on our mtDNA data, G. makiensis is currently the only known African species of Clade A (Fig. 11) and further investigations of the coastal drainages in the Horn of Africa are needed to clarify ichthyofaunal similarities across the Red Sea. Several examples can be found in literature: 1) a close affinity of G. tibanica with G. blanfordii (Boulenger, 1901) from coastal drainages in Eritrea was suggested by Trewavas (1941); 2) Menon (1958Menon ( , 1964 placed G. ethelwynnae Menon, 1958 from Salamona (Eritrea) close to G. tibanica; and 3) Stiassny and Getahun (2007) synonymised G. tibanica and G. brittoni Trewavas, 1941with G. quadrimaculata (Rüppell, 1835 from the Ethiopian Highlands. Furthermore, the Somalian cavefish G. andruzzii (Viciguerra, 1924) might reflect early dispersal events among Garra species in Africa (Yang et al. 2012).
The high genetic diversity and tree topology observed, not only within Clade A but in the whole studied dataset, suggest a complex evolutionary history and different evolutionary rates within the focal taxa. A more thorough sampling and deeper genome-level sequencing are needed to clarify the phylogenetic relationships and taxonomic status of several African Garra species.
In summary, we provide new data on morphology, mtDNA, and distribution of G. makiensis in Ethiopia. By introducing a wide set of morphological characters, we hope to support further morphological comparisons among Garra species in African and beyond. The CO1 sequence of the historic paralectotype of G. makiensis demonstrates that the use of historic museum material in phylogenetic analyses and species identification provides an invaluable potential for taxonomic studies, in particular in phenotypically variable groups. In the future, further research on African Garra is needed to clarify phylogenetic relationships, evolutionary history, and intraspecific morphological plasticity, including the variability of tubercles observed in the present study.