The identity of Barbus capensis Smith , 1841 and the generic status of southern African tetraploid cyprinids ( Teleostei , Cyprinidae )

The identity of Barbus capensis, as described by Andrew Smith (1841), is reviewed following a careful examination of the lectotype in the Natural History Museum, London. Evidence shows clearly that it represents a specimen of the Berg-Breede River whitefish or ‘witvis’ and not the species known as the Clanwilliam yellowfish, to which it was attributed until recently. The original illustration of the species is shown to be a composite of these two different species. A replacement name for the Clanwilliam yellowfish is drawn from the earliest described synonym, Labeobarbus seeberi (Gilchrist & Thompson, 1913). Following widespread recognition that the genus Barbus Daudin, 1805 does not occur in subSaharan Africa, the generic status of the Berg-Breede River whitefish (witvis) and other tetraploid cyprinines of southern Africa is reviewed, taking genetic and morphological characters into account. Five distinct lineages, each representing a genus, are recognized, including the genera Pseudobarbus Smith, 1841 and Cheilobarbus Smith, 1841, and three new genera described herein: Amatolacypris gen. nov., Sedercypris gen. nov. and Namaquacypris gen. nov.


Introduction
The advent of molecular phylogenetic analyses and genomics has rapidly improved the understanding of interrelationships of fishes in large diverse orders like the Cypriniformes (Mayden et al. 2008(Mayden et al. , 2009;;Conway et al. 2010;Saitoh et al. 2011).At this time, the inter-familial relationships within the order are reaching consensus (Yang et al. 2015).Higher level intra-family relationships within the Cyprinidae, one of the largest and most complex vertebrate families in existence, have also received attention (Shunping et al. 2008;Wang et al. 2012).Cyprinids are widespread stream and river fishes that feature considerable convergent phenotypes across the continents.Early scientific explorations in colonial times often resulted in weak taxonomy and large genera that mask species relationships under a cover of general similarity.Outsized genera such as Barbus, for hundreds of species from Europe to Africa and even Asia, simply make no sense of the concept of a genus that seeks to portray the closest phylogenetic relationships of species.The type specimens and their attached African species names, derived from early colonial-era explorations, reside mostly in the large museums of European nations.This has obstructed the easy resolution of such outsized genera, as well as the identities and relationships of numerous species, by subsequent generations of researchers in home countries.The emergence of modern technologies for analytics and communication is helping to overcome these limitations, and the resultant taxonomic convulsions are now widespread.This situation is typical for a number of African freshwater fish genera, and in this study we seek to disentangle the deep-rooted and historical taxonomic confusion of a distinct clade of African cyprinids.
Andrew Smith (1841: unnumbered 2 nd page of description) described and illustrated a large cyprinid fish that he named Barbus capensis, from the "rivers of the western coast of South Africa, more particularly the Breede and Oliphants rivers".In the description Smith referred to "four grown specimens", of which only one is known to survive, preserved in the Natural History Museum, London (NHMUK).The species was initially recognized as the Berg-Breede River whitefish or witvis (Afrikaans for whitefish) (Boulenger 1911;Gilchrist & Thompson 1913;Barnard 1937).However, on information that he received from J.R. Norman at the NHMUK about the form of the scales of the type specimen, Barnard (1937) attributed the name to the Clanwilliam yellowfish, a well-known species from the Olifants River system (Skelton 2001).He accordingly renamed the witvis as Barbus andrewi, for Sir Andrew Smith.
Upon re-examination of the type specimen, Vreven et al. (2016) reversed this action by concluding that the correct identity of the species is the witvis of the Berg and Breede rivers, and not the Clanwilliam yellowfish.Accordingly, they restored the name of the witvis to Barbus capensis (Smith, 1841) and that of the Clanwilliam yellowfish to Labeobarbus seeberi (Gilchrist & Thompson, 1913).As both the witvis and the Clanwilliam yellowfish are large and well-known species, these scientific name changes are undoubtedly disruptive to the community and need a clear explanation (Skelton 2016).This is especially relevant because, as reflected in the genera of both aforementioned species, the name Barbus, formally attributed to many African species, now has been restricted to a lineage of European species.The many African species that were formally part of this genus have been dispersed into different genera or left in a basket of unclassified 'Barbus' following the findings of Berrebi et al. (1996Berrebi et al. ( , 2014)), Machordom & Doadrio (2001), Tsigenopoulos et al. (2002Tsigenopoulos et al. ( , 2010) ) and Yang et al. (2015).The taxonomic recommendations made by Yang et al. (2015) with regard to a lineage of tetraploid species from southern Africa, including species of the genus Pseudobarbus, was to place all species other than Pseudobarbus in a temporary genus 'Pseudobarbus'.This suggestion was accepted by Skelton (2016) pending further analysis of the tetraploid clade.The witvis is a member of the tetraploid lineage, and therefore both its specific taxonomy and its generic status are in question.

Material examined
Comparative material, additional to specimens studied by Skelton (1980), used for determining the identity of the lectotype of Barbus capensis in this study is listed below.Morphological measures taken are based on Hubbs & Lagler (1964) as interpreted by Skelton (1988) and Armbruster (2012).
Bayesian posterior probabilities and branch lengths were estimated with the generalised time reversible model of DNA substitution with gamma-distributed rate variation across sites.One cold and three heated Monte Carlo Markov chains were run for a million generations.Resulting log-likelihood scores were plotted and showed that runs became stable before 5000 generations.To ensure that only optimal trees were being sampled, we discarded the first 10 000 generations as burn-in.Optimal trees were sampled every 100 generations, yielding 10 000 trees.Posterior probabilities and branch lengths were based on these optimal trees.

Characters of the lectotype of Barbus capensis Smith, 1841
The type specimen of Barbus capensis (NHMUK 1845.7.3.99) (Fig. 1) was identified by Barnard (1937) as the lectotype of the species (see ICZN 1999: Art. 74.6).Smith (1841) indicated that he used four adult specimens for his original description of B. capensis but, apart from the type of B. marequensis (see Greenwood & Crass 1959), no other specimens of large cyprinids available to Smith in the 1830s are known to exist.
The evidence for identifying this specimen as a witvis is as follows.First, the scale striae pattern of the lectotype is difficult to determine accurately, because the scales are fixed and have been lacquered.As such, the striae cannot be inspected under a microscope.However, the pattern of striae on the witvis (SAIAB 52691, 165 mm SL) (Fig. 2A) is typically with few primary (reaching from the radial centre to the edge of the scale) radiate striae and more numerous secondary (not originating in the radial centre) striae in the posterior (exposed) field of the scale (Fig. 2A).These secondary striae could easily be interpreted as being parallel or longitudinal in form, especially in larger specimens where they are more numerous.This is evident in the scales of the lectotype (Fig. 2B), and was likely responsible for Norman's initial misleading advice to Barnard.Second, the lectotype is uninformative regarding the form of the last unbranched dorsal-fin ray, or spine, as this ray is broken close to its base.In the witvis (e.g., Fig. 3A), this ray is bony and serrated, but relatively weakly so, at least in comparison with its sister species 'Pseudobarbus' (formerly Barbus) serra (Peters, 1864) (Fig. 3C).The form of the ray in the Clanwilliam yellowfish is simple and nonspinous (Fig. 3B).
The lectotype is unambiguously informative with regard to a number of other significant diagnostic characters that identify it as a witvis, and separate the witvis from the Clanwilliam yellowfish.These characters are as follows (witvis values given first): (2) Number of branched anal-fin rays: six (vs five).
(3) Shape of the anal fin in larger specimens (± ≥ 300 mm SL): square to moderately trapezoid vs extended anteriorly in larger specimens (see Barnard 1943: 164-165).(4) Length of the anal fin in larger specimens (± ≥ 300 mm SL): 11.5-15.6(mean: 13.5) vs 15.7-18.9(mean: 17.6) % SL.The divergent trend in the length of the anal fin between these species is clearly evident in the scatterplot (Fig. 4).( 5) Head length: 26.3-30.9(mean: 28.2) vs 23.7-25.6 (mean: 24.9) % SL. (6) Reach of the mouth: reaches to below the nostril vs reaches to below the anterior margin of the orbit (7) Length of the barbels: length of anterior and posterior barbels greater than orbit diameter vs less than orbit diameter.

Molecular phylogeny
Our molecular phylogenetic analysis (Fig. 5) reflects a polytomy in the tetraploid species and indicates 'P.' trevelyani as a distinct lineage, the sister group to all other southern African tetraploids.The sister clade to 'P.' trevelyani presents a set of two sub-clades, each dichotomous, as follows: Pseudobarbus and 'P.' hospes as a distinct lineage of its own, and a dichotomy with 'P.' capensis and 'P.' serra on the one hand, and 'P.' calidus and 'P.' erubescens on the other.Each of the sub-clades and their sub-lineages are strongly monophyletic and are recognized as new genera in the taxonomic descriptions below.
A summary of the morphological characters that further informs the clusters identified by the molecular phylogeny presented in Fig. 5, derived from Skelton (1980Skelton ( , 1988)), is presented in Table 1.

The identity of Andrew Smith's Barbus capensis
The lectotype of Andrew Smith's Barbus capensis (NHMUK 1845.7.3.99) is a mounted specimen of ± 345 mm SL (Fig. 1).While the NHMUK label only reports "Cape Colony" (no additional information in old NHMUK catalogue [James Maclaine: pers.comm., 2013]) as the type locality, Smith (1841: unnumbered 2 nd page of description) indicated the species was found in "… rivers of the western coast of South Africa, more particularly the Breede and Oliphants rivers".Smith's statement was not correct concerning the Breede River as being of the "western coast", because it actually drains to the south coast and the Indian Ocean.The Berg River and Olifants River are the two largest systems draining to the west coast in the south-west Cape.Further, Smith (1841: Pisces, plate X, fig. 1) provided an illustration of the ) that, in general, reflects the lectotype specimen but differs in two significant respects: (1) it shows many more and smaller scales (around 61 in the lateral line versus 37 on the lectotype); (2) the shape of the anal fin in the illustration is extended, as in a Labeobarbus, and not trapezoid as on the lectotype.
It is now known that there are four large cyprinid species (known colloquially as the witvis, sawfin, Clanwilliam yellowfish and Clanwilliam sandfish) in the rivers mentioned by Smith (1841) (see Skelton 2001).It is possible, therefore, that the four specimens mentioned in the original description of Barbus  8 capensis included at least two species and the illustration could be a composite drawing of the four specimens mentioned by Smith (1841) and not just of the existing lectotype.
Subsequent to the original description, Günther (1868: 98-99) reported B. capensis to be a species "without [a] strong osseous ray".However, Boulenger (1911: 123, fig. 100) included Barbus capensis in his division of radially striated species with a serrated dorsal-fin ray, and provided an illustration of a non-type specimen from the Burg (sic) R. (Fig. 3A).The account given by Boulenger ( 1911)
In the 1930s Dr. K.H. Barnard realized that Smith's description of Barbus capensis could apply to at least two species: one, the Cape whitefish or witvis, in the Berg and Breede Rivers, the other the Clanwilliam yellowfish in the Olifants River system.He contacted Dr. J.R. Norman of the Natural History Museum, London, to resolve the identity of the type specimen.Barnard's query directed Norman to the nature of the scales, whether they were striated in a longitudinal pattern, or with radiating striae as identified by Boulenger (1911).Norman responded that the striae of the scales on the type were longitudinal in form (Barnard 1937).This caused Barnard (1937) to re-assign the identity of Smith's Barbus capensis to the Clanwilliam yellowfish and to rename the witvis from the Berg and Breede rivers, with the serrated dorsal spine and radiating striae on the scales, as Barbus andrewi.Barnard (1937) designated the specimen illustrated by Boulenger (1911: fig. 100) (Fig. 3A: NHMUK 1901.2.11.9), as the holotype of Barbus andrewi, the witvis.Barnard's taxonomy has been followed since (e.g., Barnard 1943;Jubb 1965Jubb , 1967;;Skelton 1993Skelton , 2001)).Skelton (2001), in his second edition of the freshwater fishes of Southern Africa, placed the Clanwilliam yellowfish, B. capensis (sensu Barnard 1937), with all other South-African longitudinally striated large barbs, into the genus Labeobarbus Rüppell, 1835.During a recent study visit to the NHMUK, however, one of us (EV 2013) re-examined the mounted lectotype of B. capensis and determined that it is, without doubt, not a Labeobarbus, and that it is not conspecific with Labeobarbus seeberi as previously established by Barnard (1937) (Vreven et al. 2016).Smith's (1841) illustration of the type (Fig. 6) indicates about 60 lateral line scales, 11 transverse body rows, 23 predorsal rows, and eight lateral rows around the caudal peduncle (i.e., about 16 caudal peduncle rows).The type specimen (Fig. 1) equivalently indicates about 38+2 lateral-line, 10 transverse body, 14 predorsal and eight lateral caudal peduncle scale rows.This marked difference and discrepancy in scale size has been discussed previously by Barnard (1943: 117) and by Greenwood & Crass (1959) in reference to the second illustration of Barbus marequensis by Smith (1841: Pisces, plate X) (Fig. 7).Furthermore, this illustration of B. marequensis does not reflect the type specimen of B. marequensis (NHMUK 1845.7.3.95) in details of scale size and other aspects.Greenwood & Crass (1959) concluded that the discrepancy in scale size between illustration and specimen was a case of artistic license, but further information on this matter is to be gleaned from an examination of the plate proofs as annotated by Smith, in the Günther portfolio at the University of Witwatersrand.Whilst there are no original illustrations of either of these species in the Günther portfolio of illustrations by George Ford and other artists in the Cullen Library, University of Witwatersrand, there is, however, a proof of Plate X, with Andrew Smith's annotations on it (Fig. 7).The annotation reveals that Smith intended to name the lower species as 'Barbus gariepensis' nomen nudem and not B. marequensis as it was published.'Gariep' is the early indigenous name for the Orange River.This seems to suggest that the specimen on which that illustration was based was an Orange-Vaal River yellowfish, Labeobarbus aeneus (Burchell, 1822), taken from the Orange River and not, as is currently considered, from the type of Barbus marequensis (NHMUK 1845.7.3.95) that is derived from the Limpopo River system.The illustration is more like a Vaal-Orange River smallmouth yellowfish (Labeobarbus aeneus) than it is to a large-scale yellowfish Labeobarbus marequensis.
If this deduction is correct, it would explain to a large extent the scale size discrepancy discussed by Greenwood & Crass (1959) for Labeobarbus marequensis.It also suggests that Smith, when finally compiling the fish volume with his illustrations, was aware of, and prepared to tolerate, discrepancies between the illustrations and the specimens he had on hand to describe the species.We have no further information as to when Andrew Smith reversed his decision to use the name Barbus gariepinus (n.n.) Fig. 6.Andrew Smith's 1841 Plate X illustrating Barbus capensis (above) and Barbus marequensis (below).Photo credit: Cory Library, Rhodes University, Grahamstown.
European Journal of Taxonomy 410: 1-29 (2018) 12 and published the illustration as Barbus marequensis, except that it must have been shortly before the plate was published in 1841.The conclusion to be reached from this observation is that Smith either did not crosscheck his descriptions closely with the illustrations and specimens he had deposited in the NHMUK, or, simply tolerated the discrepancies.
The Clanwilliam yellowfish is endemic to the Olifants River system (see Skelton 1993Skelton , 2001)).There are two other large-sized cyprinids in the Olifants (but not in the Berg or the Breede).One is the Clanwilliam sandfish Labeo seeberi Gilchrist & Thompson, 1911 and the second is 'Pseudobarbus' serra (Peters, 1864), known as the sawfin (Skelton 1993(Skelton , 2001)).Labeo seeberi is distinct within that genus and need not be considered further.The sawfin is sister species to the witvis (Figs 5, 8) and therefore has relevance.
Part of the argument put forward by Barnard (1937: 305-306) for erecting a new name, i.e., B. andrewi, for the Berg River witvis, lay in the fact that no large cyprinid with a simple, non-serrated dorsal-fin ray was known from the Berg River, whereas one such species, i.e., the Clanwilliam yellowfish, was known to occur in the Olifants River.
The similarities of the sawfin and the witvis such as radiately striated scales and characters of the head, including an elongated snout, is well illustrated in the scatterplot in Fig. 9.Although the last unbranched dorsal-fin ray is bony and serrated in both species, it is distinctly more heavily developed and serrated in the sawfin than in the witvis (see Fig. 3C and 3A, respectively).Moreover, the two species differ significantly in two diagnostic characters, firstly the number of branched rays in the anal fin (five in the sawfin vs six in the witvis) and the origin of the dorsal fin (behind the origin of the pelvic fins in the sawfin vs above the origin of the pelvic fins in the witvis).The lectotype of Barbus capensis reflects the witvis condition in both these characters and we therefore conclude that it is a witvis and not a sawfin derived from the Olifants River system.
Part of the existing confusion seems to have been due to the fact that the last unbranched dorsal-fin ray of the lectotype of B. capensis has, probably since it was mounted, been broken off almost at the base (see also Barnard 1937: 305-306), making it impossible to verify whether the specimen had a serrated dorsal spine as in both B. andrewi and B. serra.Smith (1841) is not clear on this point, as he does not make any mention on the possible serrations of the last unbranched dorsal-fin ray and neither does the drawing (Fig. 6) show this typical character.At the time when Smith (1841) wrote his account, no other cyprinid with a serrated dorsal-fin ray was known from Southern Africa.Apart from the witvis, all known cyprinid species with a serrated dorsal-fin ray from Southern Africa were described after Smith's (1841) paper.
The last unbranched dorsal-fin ray of Smith's Barbus capensis might have been broken at its base before he got the specimen in hand, as fishermen frequently break the spine of such species when it holds the specimen fast in a net.Smith (1841), for instance, reported that B. capensis is "frequently taken in nets, and also occasionally by hooks".This possibility is supported by the fact that the base of the last unbranched dorsal-fin spine of the lectotype is not mounted exactly parallel to the other rays, but rather, is clearly inclined to the left-hand side of the specimen, a feature that most probably would have been avoided by the taxidermist if the spine were still intact before mounting the specimen.This is the case, for instance, with the holotype of B. marequensis.
The lectotype of B. capensis has a dorsal-fin ray formula of IV + 8 (when the last two rays that meet at their base are counted as one) and an anal-fin ray formula of iii + 6 (when the last two rays that meet at their base are counted as one).Such a combination of fin-ray formulae, i.e., with a low number of branched dorsal-fin rays (eight) and more than five branched anal-fin rays, is unknown in Labeobarbus.These counts fit perfectly with that of the witvis.The number of branched anal-fin rays for this species (six), is considered specifically different from B. serra, which has five branched anal-fin rays.Furthermore, the lectotype of B. capensis has 40 lateral-line scales, fitting perfectly the range given for the witvis, i.e., 38-41 LL scales, but falls outside the range given for the sawfin, i.e., 41-44 LL scales (Skelton 1993(Skelton , 2001)).Finally, the lectotype of B. capensis (± 345 mm SL, Fig. 1) has an unmistakably elongated snout (i.e., 56.0%HL), as reflected in Smith's ( 1841) illustration (Fig. 6).This is a longer, more pointed, snout than that of the Clanwilliam yellowfish, as illustrated by Gilchrist & Thompson (1913: 398, fig. 57) (Fig. 3C) for the holotype of B. seeberi (NHMUK 1936.8.4.6, 185 mm SL).The Clanwilliam yellowfish has a clearly shorter and more rounded snout, i.e., 29.5% HL (Fig. 9), than the witvis.

Diagnosis
Pseudobarbus is distinct from all other southern African tetraploid cyprinine genera in having a flexible, non-serrated dorsal-fin unbranched ray (vs serrated), in adults sexual dimorphism expressed in nuptial tubercle development and fin size and shape (Pseudobarbus males develop conical tubercles on the head in a distinct pattern, on the scales in a row along the free edge, and in bands on the pectoral fins vs adults of both sexes having small erupted tubercles scattered over the head and scales; males of Pseudobarbus have longer and more expansive fins than females vs similar finnage in both sexes), a reduced or absent mandibular lateral line canal (vs normally developed mandibular canal), and in having weakly ossified characteristics of the skeleton, especially the supraneural and intra-muscular bones (vs regularly ossified supraneural and intramuscular bones).In addition to these characters Pseudobarbus differs from all these genera except Sedercypris gen.nov. in having red pigmentation at the base of the fins (vs no red pigmentation); from Cheilobarbus in adult size (< 150 mm SL vs > 150 mm SL); from Amatolacypris gen.nov. in having slender third and fourth infraorbital bones (vs broad third and fourth infraorbitals); and from Namaquacypris gen.nov. in the position of the dorsal fin (Pseudobarbus dorsal fin in midbody, origin over or just behind the origin of the pelvic fins vs posteriorly, origin behind the pelvic fin), and in not having a membrane connecting the inner pelvic fin rays to the body (vs a membrane connecting the inner pelvic rays to the body).Pseudobarbus differs further from Sedercypris gen.nov. in the position of the mouth (subterminal vs terminal), and the number of anal-fin branched rays (five vs six or seven).

Etymology
Pseudobarbus is derived from 'pseudes' (Greek) meaning false, deceptive, and 'barba' (Latin) a beard; in reference to a deceptive similarity with the genus Barbus from Europe.

Description
Species of Pseudobarbus are moderate-sized (< 150 mm SL) fusiform or terete, tetraploid smiliogastrin minnows, with one or two pairs of simple circum-oral barbels; lips variably developed, pharyngeal teeth in two or three rows; tooth formulae 0-2,3,3,4+5-5-4,3,0-2; pharyngeal tooth crowns variable with offset major cusp; intestine variable in length from 1:1 in SL to 3-4:1 SL; scales radially striated, from small to moderate in size, nape and breast scales reduced or embedded; no pectoral or pelvic axil scale; lateral line interrupted or complete, in mid-body; cephalic lateral line system with pre-opercular branch disconnected and reduced or absent on the mandible.Dorsal fin with simple flexible unbranched ray and normally seven branched rays.Anal fin with five branched rays.Pectoral fins sexually dimorphic in mature adults, males with longer more expansive pectorals.Adults with bright red patches at the base of the fins.Mature males develop conical tubercles on the head, in a distinctive pattern as illustrated in Skelton (1988: fig. 31), in single rows along the free edge of scales and in bands over anterior pectoral fins.Axial skeleton without ossified supraneural or intramuscular bones.

Diagnosis
Cheilobarbus is distinguished from all other southern African tetraploid cyprinine genera by attaining a relatively large size (adult > 150 mm SL vs < 150 mm SL), and by having an extended snout as reflected in an elongated lachrymal bone (length 3 × depth vs rectangular-shaped lachrymal length 2 × depth).
Cheilobarbus also differs from Sedercypris gen.nov.by an absence of red colour at the base of the fins, in the position of the mouth (subterminal vs terminal) and in the length of the gut (> 1.5 × SL vs < 1.5 × SL).Cheilobarbus differs from Amatolacypris gen.nov. in overall size, colour (silvery to olive-bronze vs grey with double black lateral stripe), number of barbels (two pairs vs one pair), number (five vs four) and form of the 3 rd and 4 th infraorbital bones (narrow and slender vs broad).Cheilobarbus differs from Namaquacypris gen.nov. in the position of the dorsal fin (entirely before anal fin base vs reaching above anal fin base), form of the scales (regular vs thin), and the lack of a membrane between the innermost pelvic rays and the body (vs present).Cheilobarbus differs from Pseudobarbus in having a serrated dorsal-fin unbranched ray (vs simple dorsal-fin unbranched ray), an absence of red patches at the base of the fins (vs presence), no sexual dimorphism in fin size and shape and in the form and expression of nuptial tubercles as adults (vs clear sexual dimorphism in fin size and shape [males with larger fins] and nuptial tubercles [males with large conical tubercles on snout and head, small tubercles on scales and fins]).Cheilobarbus can be separated from all other large sized southern African cyprinids by karyology (tetraploid vs diploid or hexaploid) the form of the scales (radiate striae vs parallel striae) and a serrated unbranched dorsal-fin ray (vs simple spinous unbranched dorsal-fin ray).

Etymology
The name Cheilobarbus is derived from the Greek 'cheilos' meaning a margin, lip or brim, and the Latin 'barba' meaning a beard and referring to the oral barbels.Smith (1841) mentioned the "lips" of these species as being "full and firm".Gender masculine.

Description
Cheilobarbus is a genus of tetraploid cyprinine fishes with moderately sized, radiately striated scales; an elongated snout with elongated lacrymals, mouth subterminal with firm, well-developed lips, two pairs of simple oral barbels, pharyngeal bones with three rows of hooked pharyngeal teeth, tooth formula 5,3,2-2,3,5; intestine involuted and longer than the SL; dorsal fin positioned over or slightly behind the origin of the pelvic fins, the last simple ray bony weakly or strongly serrated, and eight branched rays; anal fin with three simple rays and five or six branched rays; mature breeding adults of both sexes develop small erupted nuptial tubercles densely scattered over head dorsum and in single rows over the pectoral fin rays; both species breed in male dominated nuptial shoals over gravel and cobbles.

History
Smith (1841) attributed both the large Barbus species he described to a new subgenus Cheilobarbus, i.e., Barbus (Cheilobarbus) capensis and Barbus (Cheilobarbus) marequensis.While the latter species is identified as a Labeobarbus, the former, a tetraploid species, does not belong in Labeobarbus.The type species of the subgenus, B. capensis, was designated by monotypy by Jordan (1919: 244) as first revisor and, therefore, is not a junior synonym of Labeobarbus.The subgenus Cheilobarbus was described by Smith (1841) as follows: "Mouth opening forwards; lips full, and firm; intermaxillary bones slightly extensible; nostrils double; four cirri, two from snout, and one from each angle of mouth; lateral line consisting of a series of small tubes; scales large; dorsal fin short, and commencing slightly in front of base of ventral fins; commencement of anal fin about midway between ventral and caudal fins."However this description is not diagnostic, e.g., it does not mention the nature of the unbranched dorsal-fin ray nor the nature of the scale radii.In order to institute Cheilobarbus as the genus for these species, an expanded diagnostic definition is required and the species to be included identified along phylogenetic grounds.

Diagnosis
Species of Sedercypris gen.nov.are distinct from all other southern African tetraploid genera in the combination of having six or seven branched rays in the anal fin (vs five -or six in Cheilobarbus capensis), and a red base to the fins (only Pseudobarbus also has a red base to the fins).The genus is further distinguished from Cheilobarbus in overall body size (adults ≤ 150 mm SL vs > 150 mm SL); from Namaquacypris gen.nov.by fewer pre-dorsal vertebrae (11-14 vs 13-15) the position of the dorsal fin (in advance of the origin of the anal fin vs to over the origin of the anal fin), a difference in mouth position (terminal vs inferior), in not having a membrane between the inner pelvic rays and the body (vs presence of such a membrane Namaquacypris gen.nov.); from Amatolacypris gen.nov. in number and size of barbels (Sedercypris gen.nov.with two pairs of well developed barbels that equal the orbit diameter vs one pair that is less than half an orbit diameter), in the number and size and shape of the infraorbitals (five, all slender in Sedercypris gen.nov.vs four, 3 rd and 4 th broad in Amatolacypris gen.nov.); and from Pseudobarbus in a serrated unbranched dorsal-fin ray (Pseudobarbus has a simple unbranched dorsal-fin ray), the position of the mouth (terminal vs subterminal) and the absence of strong sexual dimorphism (vs sexual dimorphism with males having conical tubercles on the head, body and fins and larger fins compared to females).

Etymology
Endemic to and named for the Sederberg (Cedarberg), Western Cape, South Africa, a Cape Fold mountain range in which rise streams and rivers tributary to the Olifants River system.The Afrikaans spelling of Sederberg is adopted for the name to avoid the possible confusion with the genus Cheilobarbus when the genus name is abbreviated to an initial in text.Masculine.

Description
The genus Sedercypris gen.nov.includes medium sized (adults < 120 mm SL) tetraploid cyprinine species from southern Africa with radiately striated scales; mouth terminal, lips slender, two pairs of well-developed simple oral barbels, pharyngeal bones with three rows of hooked teeth, formula 2,3,5 or 4-4 or 5,3,2; a simple s-folded intestine about equal to the SL in length; dorsal fin positioned over or behind the origin of the pelvic fins, with the last simple ray bony with posterior margin weakly or strongly serrated and 8 branched rays; anal fin with six or seven branched rays; mature adults with bright red flashes at the bases of fins.Breeding adults develop small erupted nuptial tubercles scattered over head dorsum and anterior body, single rows over anterior pectoral rays.Breeding takes place in male dominated nuptial schools over creviced rock faces.Amatolacypris gen.nov.urn:lsid:zoobank.org:act:0DD940C2-952C-4770-A0FC-C61104CD7599Barbus Daudin 1805: 58 (Barbus trevelyani Günther, 1877).'Pseudobarbus ' -Yang et al. 2015: 99.

Included species
Type species only.

Diagnosis
Among the southern African tetraploid cyprinine genera Amatolacypris gen.nov. is unique in colouration (silvery-grey with a thin double mid-lateral line) and in having only four infraorbitals, and broadly flanged 3 rd and 4 th infraorbital bones.It is further distinguished from Cheilobarbus on account of size (adults < 150 mm SL vs adults > 150mm SL), from Pseudobarbus and Sedercypris gen.nov.by an absence of red pigmentation at the base of the fins (vs red at the base of the fins); from Pseudobarbus by lacking clear sexual dimorphism in nuptial tubercles and fin size (vs strong sexual dimorphism in these characters); from Sedercypris gen.nov.by the position of the mouth (subterminal vs terminal) and barbels (one pair, short < 50% orbit diameter vs two pairs, long > 50% orbit diameter); from Namaquacypris gen.nov.by the number of predorsal vertebrae (10-12 vs 13-15) and position of the dorsal fin (origin above or slightly behind the origin of the pelvics vs well behind the origin of the pelvics), by the barbels (one pair, short, 50% orbit diameter vs two pairs, long > 50% orbit diameter) and by lacking a membrane between the inner pelvic rays and the body (vs presence of such a membrane in Namaquacypris gen.nov.).

Etymology
Endemic to and named for the Amatola mountains in the Eastern Cape, South Africa.Masculine.

Description
Amatolacypris gen.nov. is a monotypic genus of medium sized (< 120 mm SL) tetraploid smiliogastrin minnows from southern Africa, with radiately striated scales; four infraorbital bones, infraorbital three and four broadly flanged, covering space between orbit and preoperculum; mouth subterminal with one or two pairs of short oral barbels; pharyngeal bones with three rows of hooked teeth, formula 2,3,4 -4,3,2; intestine involuted, longer than SL; dorsal fin origin over origin of pelvics, last simple dorsal-fin ray flexible, usually with small, weak serrations along posterior edge, seven branched rays; pelvic fin with reduced axillary scale; anal fin with five branched rays; mature adults of both sexes with minute erupted tubercles over head and body, single spaced rows over pectoral rays.Breeding biology not known.

Included species
Type species only.

Diagnosis
Namaquacypris gen.nov.may be separated from all other southern African tetraploid genera by the relatively posterior position of the dorsal fin (behind the pelvic fin and base to over the origin of the anal fin vs over the pelvic fin and base before the origin of the anal fin), by a high number of predorsal vertebrae (13-15 vs 13 or fewer), the inner pelvic rays partly attached to the body by a membrane (vs an absence of such attachment), and by having a pair of barbels sub-equal in length (vs anterior barbels shorter than posterior barbels).Namaquacypris gen.nov.differs further from both Pseudobarbus and Sedercypris gen.nov. in an absence of red pigment at the base of the fins (vs bright red base to the fins), from Pseudobarbus in an absence of sexual dimorphism in fin length and nuptial tubercles (vs strong sexual dimorphism in fin length and nuptial tubercles), and from Sedercypris gen.nov. in the position of the mouth (subterminal vs terminal) and in having five branched anal-fin rays (vs six or seven branched anal-fin rays).Namaquacypris gen.nov.differs from Amatolacypris gen.nov. in colour and pigmentation (silvery, white below vs grey with a dark thin stripe and stripe over the lateral line), the number of infraorbitals (five vs four), the number and length of the barbels (two pairs, as long as the eye diameter vs one pair, shorter than half an eye diameter), and the size and shape of the 3 rd and 4 th infraorbital (slender vs broad).It differs from Cheilobarbus in size (< 150 mm SL vs > 150 mm SL), head shape (short snout vs long snout) and colour (adults plain silvery vs olive-bronze).

Etymology
Named for Namaqualand, a semi-desert region of the Northern Cape, South Africa and Namibia through which the Lower Orange River flows.The genus is endemic to the lower Orange river below the Augrabies waterfall.Masculine.

Description
Namaquacypris gen.nov. is a monotypic, tetraploid smiliogastrin genus of medium to small (< 100 mm SL) minnows from southern Africa, with small, light (thin) radiately striated scales; small, inferior mouth, with two pairs of equivalently long (equal or longer than the orbit diameter) oral barbels; pharyngeal bones with three rows of hooked teeth, formula 2,3,5, -5,3,2; intestine with simple s-fold, short, slightly less than SL; a relatively high number (13-15) of predorsal-fin vertebrae; dorsal-fin origin nearer caudal base than tip of snout, entirely behind the pelvic fins, dorsal-fin unbranched ray serrated along posterior edge, basal serrations directed distally, seven branched rays; pelvic fins with inner rays attached to body by a membrane; anal fin origin below hind margin of dorsal base.Mature adults of both sexes develop minute nuptial tubercles scattered over head dorsum and upper body anterior to the dorsal fin, with bands 2-3 tubercles deep over pectoral fin rays.The breeding biology is not known.

Taxonomic implications and conclusions
This re-identification of the lectotype of Barbus capensis requires nomenclatural changes as made by Vreven et al. (2016), and elaborated on here as follows.Considering that B. capensis is, in fact, not a large hexaploid cyprinid, i.e., Labeobarbus, the species name B. seeberi, a junior synonym of B. capensis according to Barnard (1937), becomes available, and, as the earliest available name, is assigned to the Clanwilliam yellowfish as Labeobarbus seeberi.The name Barbus andrewi Barnard, 1937, becomes a junior synonym of B. capensis.Barnard (1937) subsequently identified the NHMUK syntype of B. seeberi that had been illustrated by Boulenger (1916: 241) as the "type", leaving the two SAIAB (134867) specimens as "cotypes".This decision holds as a correct lectotype designation (ICZN 1999: Art.74.5) for the species.In addition, following Yang et al. (2015), the genus in which the witvis, B. capensis (= B. andrewi), is now placed recently changed to 'Pseudobarbus'.This placement receives further discussion and taxonomic determination below.et al. (2015) recommended that the genus Pseudobarbus be expanded to embrace the entire lineage of tetraploid cyprinids in southern Africa using quotation marks, 'Pseudobarbus', until the situation could be considered in detail.All Pseudobarbus and 'Pseudobarbus' species are tetraploid (Naran 1997;Naran et al. 2006).In addition to Yang et al. (2015) the phylogeny of the southern African tetraploid cyprinids was investigated, in part, by Tsigenopoulos et al. (2002), de Graaf et al. (2007de Graaf et al. ( , 2010) ) and Berrebi et al. (2014).All these studies consistently show that the genus Pseudobarbus, i.e., the redfins with a flexible, simple or unbranched dorsal-fin ray, is a strongly supported monophyletic lineage that is sister to the tetraploid species with a serrated last unbranched dorsal-fin ray species, here 'Pseudobarbus' (Figs 5,8).All the aforementioned studies also show a strongly supported clade of serrated-rayed species that includes 'P.' capensis, 'P.' serra, 'P.' calidus and 'P.' erubescens.The position of 'P.' trevelyani (Günther, 1877) is, however, unsettled.'Pseudobarbus' trevelyani is potentially linked as the sister species the serrated-rayed clade (Fig. 8C, as in de Graaf et al. 2007), or, as an independent lineage in an unresolved polytomy with the other two lineages (Fig. 8B), as in Tsigenopoulos et al. (2002), de Graaf et al. (2007de Graaf et al. ( , 2010)), Berrebi et al. (2014) and Yang et al. (2015).

Yang
It should be noted that the configuration of the lineages with respect to the position of 'P.' trevelyani, using the same data set, differs when analysed as a Neighbour joining tree (de Graaf et al. 2007; Fig. 8B) or using Bayesian analysis (de Graaf et al. 2007; Fig. 8C).Furthermore, none of the above-mentioned studies included samples of 'P.' hospes (Barnard, 1938).Thus the phyletic positions of 'P.' trevelyani and 'P.' hospes are currently uncertain.
From a morphological perspective, Skelton (1976) compared the post-cranial meristics of certain southern African cyprinids and drew tentative conclusions on relationships.He pointed to several possible links between species, including 'P.' capensis (as Barbus andrewi) and 'P.' serra (as Barbus serra), and these with 'P.' calidus and 'P.' erubescens, based on the high number of branched rays in the anal fin of all but 'P.' serra.In addition, Skelton (1976: 406) indicated that 'P.' calidus and 'P.' erubescens were similar to 'P.' hospes and 'P.' trevelyani by having the relatively high vertebral counts of all the smaller Barbus species examined.Skelton mentioned that both Farquharson (1962) and Gaigher & Pott (1973) regarded these species as derivatives of an early invasion of the Cape region, but cautioned that this did not necessarily indicate a common ancestry.Skelton (1980) considered the morphology, osteology, and vertebral meristics of a wide range of southern African cyprinine species, in order to use the characters for a phylogenetic and taxonomic analysis of the redfin lineage.The genus Pseudobarbus was based on these data that, together with molecular data, were incorporated in the phylogenetic analysis by Swartz et al. (2009) of the redfin lineage.Several characters such as number of barbels, nature of the last simple dorsal-fin ray (serrated vs simple) were equivocally assigned, and therefore uninformative at the higher level of universality in terms of this lineage.Synapomorphies of the Pseudobarbus species, known as redfin minnows, include such prominent morphological characters as the development and pattern of nuptial tubercles, squamation, cephalic lateral line development, reductive osteology in both the cranial and post-cranial skeleton, pharyngeal teeth morphology and gut development (Skelton 1980;Swartz et al. 2009).
In contrast to Pseudobarbus, all other tetraploid smiliogastrin species are differentiated by having a serrated unbranched ray in the dorsal fin (Table 1).In 'P.' trevelyani and 'P.' erubescens the ray is generally weakly serrated and flexible and thus non-spinous.The soft-rayed redfin species of Pseudobarbus are sexually dimorphic, with males developing conical tubercles on the head, rows of small tubercles along the edge of scales and bands over the pectoral fins rays.The males establish and defend a nuptial territory, breeding one-on-one with individual females.Mature adults of other tetraploids in breeding condition all show small, scattered "pimple-like" tubercles over the head and body of both males and females (Skelton 1980).The differences in development and pattern of nuptial tubercles reflects a difference in the breeding biology of the lineages.The breeding behavior of 'P.' trevelyani and 'P.' hospes has not yet been reported, but all the other serrated rayed species are group spawners, with males and females assembling in nuptial swarms and breeding over specific spawning beds or sites (Skelton 2001;Impson 2008;Paxton 2008;Paxton & King 2009).
These considerations together with the molecular phylogenetic results (Fig. 5) all indicate that a set of related genera is justified for the monophyletic clades of serrated-ray tetraploid cyprinids in southern Africa.The evidence favors a set of related genera because of the very distinctive and exclusive phenotype of each clade as defined by autapomorphies.Thus the large bodied sister species 'P.' capensis and 'P.' serra are sufficiently distinct on the grounds of attainable size, osteology and skeletal meristics from the other species to warrant their own genus.The sister clade to these large-bodied species, 'P.' calidus and 'P.' erubescens, are also sufficiently similar to each other but distinct, both morphologically and genetically, from other lineages, to warrant their own genus.'Pseudobarbus' trevelyani and 'P.' hospes are each separate and independent lineages on morphological and genetic grounds, and each should be allocated a monospecific genus, justified by distinctive autapomorphies (Skelton 1980).Diagnoses of these new genera are provided in this paper.Appendix 1 (continued on next page).Specimens and their origin analysed for the mitochondrial cytochrome b gene, using the nomenclature proposed in the present study.

Fig. 2 .
Fig. 2. A. A scale of the witvis Cheilobarbus capensis Smith, 1841 (SAIAB 52691) drawn by camera lucida to show the pattern of striae.Scale bar = 1 mm.Arrow indicates anterior (embedded field) to posterior (exposed field) orientation.Primary radii reach from radial centre to scale edge; secondary radii do not reach radial centre.B. Close up of the scales from the right flank of the lectotype of Barbus capensis (NHMUK 1845.7.3.99).Scale bar = 5 cm.Photograph by E. Vreven.

Fig. 5 .
Fig. 5. Bayesian phylogram, showing the phylogenetic relationships among southern African tetraploid barbs based on the mitochondrial cytochrome b gene.Bayesian posterior probabilities are shown above branches.Allele codes are indicated at terminal branches.NS: Not significant.The Bayesian posterior probability at higher NS node is 53%; at the subsequent node 77%.