Revalidation of Enteromius alberti and presence of Enteromius cf. mimus (Cypriniformes: Cyprinidae) in the Lake Edward system, East Africa

A revision was done on the species of Enteromius Cope, 1867 (Cypriniformes: Cyprinidae) from the Lake Edward system with a smooth, flexible third unbranched dorsal fin ray without serrations. Specimens with these characteristics had previously been attributed to E. perince and E. stigmatopygus. A combination of a genetic (COI, mtDNA) and a morphometric approach was used. Based on the COI gene, we found two groups with a distance of 8.5%, though neither of the two corresponded to E. perince or E. stigmatopygus. One group revealed to be conspecific with E. alberti, previously a synonym of E. stigmatopygus, described from the Rutshuru River, May-Ya-Moto (DRC, Lake Edward system), and revalidated here. In addition, E. cercops, described from the Nzoia River (Kenya, Lake Victoria basin), is put in synonymy with E. alberti. The second group was most similar to E. mimus, but differed morphologically somewhat from the types of E. mimus. Therefore, specimens of this group were identified as E. cf. mimus. Morphologically, E. alberti can be separated from E. cf. mimus based on a higher number of lateral line scales and smaller values for interorbital width, pre-pelvic distance, body depth, maximum and minimum caudal peduncle depth, head width and head depth.


Introduction
With an estimated 367 genera and 3006 species, Cyprinidae is the largest family of freshwater fishes in the world (Nelson et al. 2016). One of these genera, Enteromius Cope, 1867, occurs only on the African continent and contains about 211 valid species, making it the second largest fish genus in Africa, and the third largest in the world, only exceeded by Schistura McClelland, 1838 (228 valid species), and Haplochromis Hilgendorf, 1888 (229 valid species) (Froese & Pauly 2019). The genus Enteromius, formerly referred to as 'Barbus' (Berrebi et al. 1996), represents a non-monophyletic group that includes most small-sized African diploid barbs (Yang et al. 2015;Hayes & Armbruster 2017).

Taxonomic problems in the genus Enteromius
There is a general opinion among ichthyologists that cyprinids are a taxonomically difficult group and that many species 'look the same', having a rather uniform body shape (Lowe-McConnell 1975;Howes 1991). This specifically holds for Enteromius. Specimens of this genus can be attributed to three groups based on the morphology of the third unbranched dorsal fin ray: bony with serrations, bony without serrations, or slender and flexible without serrations (Skelton et al. 1991). While these characters seem to be indicative of a common descent on a small geographic scale, as demonstrated in studies on the species from the basins of the Congo River (Van Ginneken et al. 2017) and of Lake Edward , on a larger geographic scale, they do not seem to give a reliable phylogenetic signal (Yang et al. 2015;Ren & Mayden 2016;Hayes & Armbruster 2017), and point at convergent evolution at higher levels of hierarchy in the trees. Furthermore, the identification of species of Enteromius is challenging because of morphological similarities, a limited representation in the literature and because of a lack of identification keys and large-scaled revisions (Van Ginneken et al. 2017;Decru et al. 2019). Consequently, many specimens in museum collections, as well as the voucher specimens of several sequences on GenBank are misidentified (Hayes & Armbruster 2017). In addition to the paucity of taxonomic data, recent studies on species of Enteromius from various regions revealed the presence of many undescribed species (Schmidt et al. 2017;Van Ginneken et al. 2017). This also implies that the species diversity in Enteromius might be of an even much larger magnitude than currently thought. Indeed, if these findings of hidden diversity can be extrapolated to other regions, then an enormous increase of species numbers among the small African barbs is to be expected, which could render Enteromius the by far most species-rich, although currently nonmonophyletic, fish genus in the world.
It is intriguing to find these small, inconspicuous fish to be so hyperdiverse. However, the mechanisms driving speciation in Enteromius have never been investigated. The study of Van Ginneken et al. (2017) on some groups of Enteromius from the Congo basin, suggested a mainly allopatric mode of speciation, as most lineages are confined to a certain river stretch. Allopatric speciation is probably the mode of speciation by which the majority of riverine fishes arose (Seehausen & Wagner 2014). Still, in the case of Enteromius, it is difficult to envisage how allopatric speciation alone can explain some cases of morphological highly similar species occurring in sympatry, as was found in the study of Van Ginneken et al. (2017). Furthermore, hybridisation events can also be important in the evolutionary history of Enteromius. This has already been documented for several cichlid and cyprinid lineages in Africa available (Supplementary File SM.01) (voucher specimens are stored in the National Institute For Oceanography & Fisheries (NIOF), Alexandria, Egypt). The genetic divergence between our sequences and those from GenBank has been assessed.

Morphometric analyses
Twenty-four morphometric measurements were taken with a dial caliper and 13 meristics were counted on each specimen. Most measurements and all counts were done under a binocular microscope (Wild Heerbrugg M3). Measurements were based on Bamba et al. (2011) with some modifications. Post-dorsal distance I, post-anal distance I and body depth I were not measured, because of the strong correlation with other measurements. The following measurements were added: head width (width of the head at the level of the posterior margin of the pre-operculum), head depth (depth of the head at the level of the posterior occipital margin), anal fin base length (distance between the articulation base of the first and last anal fin rays), anal fin length (distance from the articulation of the first anal fin ray to the tip of the longest anal fin ray), pectoral fin length (distance from the articulation of the first pectoral fin ray to the tip of the longest pectoral fin ray) and pelvic fin length (distance from the articulation of the first pelvic fin ray to the tip of the longest pelvic fin ray). For all meristics, Bamba (2012) was followed.
The multivariate dataset was explored by Principal Component Analyses (PCA) to reduce the large number of variables into a few meaningful axes that are linear combinations of the original variables (Manly 1994;Snoeks 2004;Decru et al. 2012). Measurements and meristics were analysed separately. Log-transformed data were used for the PCA on the measurements and analysed with a covariance matrix. The first axis of the PCA on log-transformed measurements can be interpreted as a proxy for size (Bookstein et al. 1985). Hence, subsequent axes can be plotted against the first axis to verify possible size effects. The raw data were used for the PCA on the meristics based on the correlation matrix. Invariable meristics were excluded from the PCAs. Data for the barbels were not included in the analyses, as the barbels were small and difficult to measure and hence their measurements were prone to errors.
A first PCA was performed on the 72 specimens from the Lake Edward system that were analysed genetically to evaluate the morphological differences among specimens from different genetic clades. To find possible significant differences between these groups, univariate comparisons were performed with non-parametric Mann-Whitney U-tests (MWU), with sequential Bonferroni corrections (Rice 1989). For measurements, these comparisons were restricted to subsets of specimens that belong to a similar size class (46.4-53.8 mm SL; with p = 0.6888 for SL). This was done to avoid allometric interference. Scatterplots of individual measurements (% SL) were made to visualize between-group differences taken into account possible size-effects. Based on this first set of analyses, additional specimens for which no genetic data were available, were measured and assigned to one of the groups in order to enlarge the size ranges of these groups. This also allowed for comparison with the type series of similar species. Hence, in a second series of analyses, PCAs were performed to compare each group separately with all of the studied type specimens of E. perince, E. stigmatopygus, E. alberti, E. mimus and E. cercops, and with the specimens from Tshambi identified as E. alberti.

Results
We successfully obtained 137 COI sequences (trimmed to 651 bp) (Supplementary File SM.01). The haplotype network of the specimens from the Lake Edward system (Fig. 1) showed two clear groups with a genetic divergence of 8.5%. No geographical structuring was observed. We refer to these groups as group A, the smallest group (n = 36), and group B, the largest group (n = 101). The sequences of E. perince from Egypt, obtained from GenBank, had a genetic difference of 12% from the sequences of our groups A and B, and were therefore not integrated in the haplotype network.

Morphometric analyses
Of the 137 specimens that were analysed genetically, 71 specimens were also analysed morphometrically. A first PCA on the log-transformed measurements of these specimens resulted in a separation on PC2 (Fig. 2) of the two groups that were delineated based on the genetic data. PC1 represented an inverted proxy of size, with larger specimens having lower values on PC1 than smaller specimens. The most important loadings for PC2 were for the post-anal distance and the maximum and minimum caudal peduncle depth (Supplementary File SM.02). In the PCA on meristics, the two groups partially overlapped (not illustrated).
Based on Mann-Whitney U (MWU) tests with sequential Bonferroni correction, 13 measurements and three meristics were significantly different between the two groups, of which the three meristics were highly significant (Table 1). Individual scatterplots were made for the most significant measurements (% SL) (Fig. 3): interorbital width, pre-pelvic distance, body depth, maximum caudal peduncle depth, minimum caudal peduncle depth, head width, and head depth. For all of these measurements, the ranges per se overlapped because of allometry, but for specimens of similar size classes, the two groups could be distinguished from each other, with specimens of group B having higher values for all these measurements. Based on the results of the first series of analyses discussed above, we were able to identify 16 additional specimens as members of group A, and 22 additional specimens as members of group B. These specimens were added to the dataset to enlarge the size ranges of the two groups, allowing for a better comparison of the groups with the type series of the relevant species. Subsequently, each Cope, 1867 with a smooth, flexible last unbranched dorsal fin ray from the Lake Edward system. Each circle represents a haplotype, with the size of the circles indicating the number of individuals with this haplotype and the colour indicating the different parts of the basin. Each bar represents a mutation between two haplotypes. A PCA on log-transformed measurements was performed on the specimens assigned to group A, all the type specimens included in this study and the specimens from Tshambi (Fig. 4a). On a scatterplot of PC2 vs PC1, specimens belonging to group A were separated from E. perince, E. stigmatopygus and the specimens from Tshambi. They were also separated from the types of E. mimus, except for one specimen. The holotype and paratypes of E. cercops and two of the three types of E. alberti fell, however, into the polygon of group A. The most important loadings for PC2 were the maximum caudal peduncle depth, the interorbital width and the body depth (Supplementary File SM.03).
In a scatterplot of the two main PCs of a PCA on the meristics, the type specimens of E. alberti fell within group A and E. cercops overlapped with group A (Fig. 4b).
Specimens belonging to group B were also compared with the type specimens of the selected species and the specimens from Tshambi. In the PCA on the log-transformed measurements, group B was separated from most of the type specimens, but partially overlapped with those of E. mimus and with one paratype of E. cercops on PC2 for specimens with similar sizes (Fig. 5a). Furthermore, one of the specimens from Tshambi was situated within the polygon of group B, while the other one was located just outside of the polygon. The most important loadings for PC2 were the anal fin base length, the interorbital width and the anal fin length (Supplementary File SM.05).
Another PCA was performed on the meristics of group B and the selected type specimens. Here, both specimens from Tshambi fell within group B. Almost all types of E. mimus were situated within the polygon of group B, except for one, which was located just outside of the polygon. Group B partially overlapped with E. cercops (Fig. 5b). The most important loadings for PC1 were the number of scales on the lateral line, around the caudal peduncle and below the lateral line and, for PC2, the total number of gill rakers, the number of gill rakers on the lower part of the first gill arch and the number of pre-dorsal scales (Supplementary File SM.06).
These PCAs (Fig. 5) illustrated that the specimens from Tshambi, identified as E. alberti, were most probably conspecific to our group B. This latter group seemed to be most similar to E. mimus. Therefore, additional PCAs on meristics and log-transformed measurements were performed including only group B, with the inclusion of the specimens from Tshambi, and the types of E. mimus. For the meristics, the syntypes of E. mimus fell completely within the range of group B for the first two axes (not illustrated).
For the log-transformed measurements, however, only a partial overlap was observed on PC2 ( Fig. 6).

Fig. 6.
Scatterplot of PC2 against PC1 of the PCA (n = 83) on 24 log-transformed measurements. For group B, the filled squares (■) represent the specimens used for the genetic analysis, the open squares (□) indicate the additional specimens of the Lake Edward system and the two specimens from Tshambi. Specimens of E. mimus (Boulenger, 1912) (lectotype) (▼) and E. mimus (paralectotypes) (▼).
Here, the most important loadings for PC2 were for the snout length, the pelvic fin length and the interorbital width (Supplementary File SM.07).
In the analysis of measurements restricted to group B, one of the types of E. alberti was situated within the polygon of the types of E. mimus (Fig. 5a). In that of the measurements of group A (Fig. 4a), the same specimen was situated closest to the polygon of the types of E. mimus. This specimen had an aberrant value for the number of lateral line scales (23)  The results confirmed the occurrence of two syntopic species of Enteromius with a smooth dorsal fin ray in the Lake Edward system. Specimens of group A were conspecific with the types of E. alberti, except for one type specimen, which was assigned to group B, and with the holotype and paratypes of E. cercops. Enteromius alberti used to be considered a synonym of E. stigmatopygus. Based on our results, and following the principle of priority for species names, we propose to revalidate E. alberti with E. cercops as its junior synonym. A redescription of E. alberti is provided based on the lectotype (here designated) (Fig. 7a) and one paralectotype of E. alberti, the type series of E. cercops, here synonymised with E. alberti, and the additional specimens from the Lake Edward system examined in this study and identified as E. alberti (n = 37). The other paralectotype of E. alberti, which was revealed not to be conspecific with E. alberti, was not included in the redescription. The measurements and meristics can be found in Table 2. An illustration of a fresh specimen of E. alberti, captured during the HIPE expeditions is provided in Fig. 7b.
The specimens of group B were most similar to E. mimus, though the PCA on log-transformed measurements (Fig. 6) suggested some differences. Therefore, we decided to assign the specimens of group B to E. cf. mimus with "cf" indicating specimens that are comparable to E. mimus but whose identification is uncertain (Decru et al. 2016). Because of the unsuitable original description of E. mimus for modern taxonomic purposes, we decided to make a redescription based on the lectotype (here designed) (Fig. 8) and eight of the nine paralectotypes of E. mimus. We found one paralectotype to have a serrated dorsal fin ray; therefore, it was considered not to be conspecific with the other types and excluded from the redescription. Specimens from the Lake Edward system identified as E. cf. mimus, the specimens from Tshambi and the aberrant lectotype of E. alberti were not included in the redescription, but their measurements and meristics are given in Table 2. A picture of a fresh specimen of E. cf. mimus is given in Fig. 9.
For both redescribed species, a lectotype was chosen from the syntypes. For E. alberti, this is the specimen illustrated in the original description by Poll (1939), which is also the largest specimen. The lectotype of E. mimus is the specimen that resembles most the original figure of E. mimus in the Catalogue of the Fresh-water Fishes of Africa in the British Museum (Natural History), Volume IV, by Boulenger (1916).  (Boulenger, 1912)    The sampling sites of the two species in the Lake Edward system and the type localities of E. mimus, E. alberti and E. cercops are illustrated in Fig. 10. The specimens from Tshambi, initially identified as E. alberti, are indicated as E. cf. mimus on the map, as found in the present study.

Diagnosis
Enteromius alberti (Fig. 7a-b) belongs to the group of species of Enteromius with a flexible last unbranched dorsal fin ray that lacks serrations along its posterior edge. A comparison was made with other species of Enteromius from this group from the East Coast and Nilo-Sudan ichthyofaunal provinces (list made based on information available in, e.g., Greenwood 1962;Okaronon et al. 1997;Seegers et al. 2003;Schmidt et al. 2017;Froese & Pauly 2018). Enteromius alberti can easily be distinguished from the other species of this group from the East Coast and Nilo-Sudan ichthyofaunal regions by the following combination of characteristics (data from original description, unless otherwise stated): a complete lateral line vs an incomplete lateral line in E. atkinsoni (Bailey, 1969), E. pumilus (Boulenger, 1901), E. serengetiensis (Farm, 2000), E. tongaensis (Rendahl, 1935) and E. toppini (Boulenger, 1916); two pairs of barbels vs one pair in E. pseudotoppini (Seegers, 1996), and no barbels in E. anema (Boulenger, 1903) and E. profundus (Greenwood, 1970); one to three dark spots on the flanks, which sometimes fuse into a mid-lateral line in preserved specimens, starting posterior to the operculum vs a dark line running from the tip of the snout to the caudal fin base in E. bifrenatus (Fowler, 1935) and E. yongei (Whitehead, 1960), and a thin dark line from the beginning of the operculum to the caudal fin base in E. viviparus (Weber, 1897); 12 scales around the caudal peduncle, with one aberrant specimen in the paratypes of E. cercops (16), vs 8 in E. leonensis (Boulenger, 1915), 9-10 in E. venustus (Bailey, 1980), and 10 in E. magdalenae (Boulenger, 1906) and E. yeiensis (Johnsen, 1926); 4.5 scales between the dorsal fin base and the lateral line vs 3.5 in E. radiatus (Peters, 1853) (Greenwood 1970), 5.5 in E. unitaeniatus (Günther, 1867), and 6 in E. usambarae (Lönnberg, 1907); a dorsal fin length which is larger than the head length vs a dorsal fin length equal to the head length in E. innocens (Pfeffer, 1896); a body depth which is larger than the head length vs a body depth which is equal to the head length in E. nigeriensis (Boulenger, 1903) and E. trispilopleura (Boulenger, 1902); a pectoral fin length which is ⁵⁄₆ of the head length vs ¾ in E. lineomaculatus (Boulenger, 1903), and ⅔ to ¾ in E. neglectus (Boulenger, 1903) Fig. 3).

Etymology
Etymology not explained in the original description. Probably, the species epithet alberti is derived from the name of the former 'Parc National Albert' (now Virunga National Park) in which the type locality is located.

Description
Based on the lectotype and one paralectotype of E. alberti, the type series of E. cercops, which is synonymised here as E. alberti, and 37 additional specimens from the Lake Edward system, identified as E. alberti. The lectotype is illustrated in Fig. 7a. Maximum observed size: 74.3 mm SL. Morphometric and meristic data are given in Table 2. Body fusiform, largest depth anterior to dorsal fin. Dorsal profile from tip of snout to origin of dorsal fin slightly convex, slightly tapering to base of caudal fin. Ventral profile from operculum to origin of pelvic fin slightly convex, slightly tapering to posterior end of anal fin base, then slightly concave to caudal fin. Head small. Eye large and round, located dorso-laterally, closer to tip of snout than to distal margin of operculum, interorbital profile slightly convex. Snout rounded and mouth subterminal. Two pairs of barbels, anterior barbel usually slightly shorter than posterior with anterior one usually reaching up to vertical through posterior margin of eye, while posterior one usually up to vertical trough posterior margin of pre-operculum. Dorsal fin with 3 unbranched and 7 to 9 branched rays, distal margin straight to slightly concave, origin located slightly anterior [1 or 2 lateral line scale(s)] to vertical through pelvic fin insertion. First branched dorsal fin ray longest, posterior rays decreasing progressively in size. Pectoral fin with 1 unbranched and 13 to 16 branched rays, distal profile straight to slightly convex, not reaching anterior base of pelvic fin. Pelvic fin with 1 unbranched and 6 to 8 branched rays, distal margin slightly convex. Anal fin with 3 unbranched and 5 to 6 branched fin rays, distal margin slightly concave. Caudal fin forked with outer rays twice as long as median ones and both lobes rounded and of similar size. Anus and urogenital opening situated immediately in front of anal fin base. Scales cycloid, rounded and radially striate. Lateral line completely pored with many lateral line scales [27-34 (median 31)], and gently curved downwards over abdomen but running straight along middle of caudal peduncle and ending at base of caudal fin. Lateral-line scales smaller on caudal peduncle than below dorsal fin.

Colour pattern
No sexual dimorphism observed. In life, overall background colour of body silvery, darker toward dorsum and lighter towards belly (Fig. 7b). Thick yellowish midlateral band, often with one to three darker spots. All fins translucent. In preserved specimens, overall background colour of body yellowish, greyish dorsally (Fig. 7a). One to three dark spots on flanks on thick silvery midlateral band; spots fused into mid-lateral line in some specimens, overlying silvery band and starting posterior to operculum. First spot situated anterior to dorsal fin origin, median one below last ray of dorsal fin and last one at caudal fin base. Specimens with more than three spots are rare (1 specimen with 4 spots and another with 6 spots). Distribution (Fig. 10) Enteromius alberti occurs in river systems throughout the Lake Edward basin. For the Congolese part of the basin, it is known from its type locality, the Rutshuru River, May-Ya-Moto (Poll 1939). Although the Congolese part of the system was not sampled during the recent expeditions, we identified specimens from the collection at the RMCA (86-01-P-479 to 500 and 86-09-P-70 to 71) from Vitshumbi, located in the southern Congolese part of the Lake Edward system, as E. alberti. In the Ugandan part, we collected the species in the Mahoma, Mpanga, Nchwera, Nyamweru and Kyambura (an affluent of the Rusangwe) Rivers, in Lake Edward offshore at Kayanja, and in Lake George at Kashaka Bay. For the Lake Victoria basin, E. alberti is known from the Nzoia River, the Malawa River and the Middle Akagera system (Whitehead 1960;Greenwood 1966;De Vos & Thys van den Audenaerde 1990;Seegers et al. 2003). (Boulenger, 1912) Fig. 8, Table 2 Diagnosis Enteromius mimus (Fig. 8) belongs to the group of species of Enteromius with a flexible last unbranched dorsal fin ray that lacks serrations along its posterior edge. A comparison was made to other species of Enteromius from this group from the East Coast and Nilo-Sudan ichthyofaunal provinces (list made based on information available in, e.g., Greenwood 1962;Okaronon et al. 1997;Seegers et al. 2003;Schmidt et al. 2017;Froese & Pauly 2018). Enteromius mimus can easily be distinguished from the other species of this group from the East Coast and Nilo-Sudan ichthyofaunal regions by the following combinations of characteristics (data from original description, unless otherwise stated): a complete lateral line vs an incomplete lateral line in E. atkinsoni, E. pumilus, E. serengetiensis, E. tongaensis and E. toppini; two pairs of barbels vs one pair in E. pseudotoppini, and no barbels in E. anema and E. profundus; one to three dark spots on the flanks, which sometimes fuse into a mid-lateral line in preserved specimens, starting posterior to the operculum vs a dark line running from the tip of the snout to the caudal fin base in E. bifrenatus and E. yongei, and a thin dark line from the beginning of the operculum to the caudal fin base in E. viviparus; 11-12 scales around the caudal peduncle vs 8 in E. leonensis, 9-10 in E. venustus, and 10 in E. magdalenae and E. yeiensis; 24-27 lateral line scales (median 25) vs 30 in E. lineomaculatus, and 31 E. innocens; 3.5-4.5 scales between the dorsal fin base and the lateral line vs 5.5 in E. unitaeniatus, and 6 in E. usambarae; 3-4 scales between the lateral line and the pelvic fin (median 3) vs 2 in E. trispilopleura, and 2-2.5 in E. neglectus (Boulenger 1907); 7-8 branched dorsal fin rays vs 9 in E. quadripunctatus; a body depth which is larger than the head length vs a body depth which is equal to the head length in E. nigeriensis; a smaller snout length [4.1-7.1% SL (average 6.0) vs 7.1-10.1] in E. radiatus (Greenwood 1970 (Table 2).

Description
The redescription of E. mimus is based on the specimens of the type series only (the lectotype and eight paralectotypes). The lectotype is illustrated in Fig. 8. Maximum observed size: 45.6 mm SL. Morphometric and meristic data are given in Table 2. Body fusiform, largest depth anterior to dorsal fin. Dorsal profile from tip of snout to origin of dorsal fin slightly convex, slightly tapering to base of caudal fin. Ventral profile from operculum to origin of pelvic fin slightly convex, slightly tapering to posterior end of anal fin base, then slightly concave to caudal fin. Head small. Eye large and round, located laterally, closer to tip of snout than distal margin of operculum, interorbital profile slightly convex. Snout rounded. Mouth subterminal. Two pairs of barbels, anterior barbel shorter than posterior with anterior one usually reaching up to vertical through the middle of the eye, while posterior one can reach the vertical trough posterior margin of pre-operculum. Dorsal fin with 3 unbranched and 7 to 8 branched rays, distal margin slightly concave, origin located on vertical through pelvic fin insertion or slightly anterior (1 lateral line scale). First branched dorsal fin ray longest, posterior rays decreasing progressively in size. Pectoral fin with 1 unbranched and 12 to 14 branched rays, distal profile straight to slightly convex, not reaching anterior base of pelvic fin. Pelvic fin with 1 unbranched and 7 branched rays, distal margin convex. Anal fin with 3 unbranched and 5 to 6 branched fin rays, distal margin concave. Caudal fin forked with outer rays twice as long as median ones and both lobes rounded and of similar size. Anus and urogenital opening situated immediately in front of anal fin base. Scales cycloid rounded and radially striate. Lateral line completely pored with many lateral line scales (total) [24-27 (median 25)], gently curved downwards over abdomen but running straight along middle of caudal peduncle and ending at base of caudal fin. Fig. 9. Fresh specimen of E. cf. mimus (Boulenger, 1912) (RMCA 2016.035.P.0125-0139 HP547) with 45.2 mm SL.

Colour pattern
No sexual dimorphism is observed. No information of in vivo colour pattern is available. In preserved specimens, overall background colour of body brown, darker dorsally. One to three dark spots on flanks on thick silvery midlateral band; spots fused into mid-lateral line in some specimens, overlying silvery band and starting posterior to operculum. First spot situated anterior to dorsal fin origin, median one below or somewhat anterior to last ray of dorsal fin and last one at caudal fin base.

Distribution
Enteromius mimus is known from its type locality, the Euasso Nyiro River below the falls, east of Lake Baringo, Kenya, and from the Tana River system (Seegers et al. 2003). For the time being, the specimens of the Lake Edward system are referred to as E. cf. mimus and are not included in the distribution of E. mimus.

Enteromius alberti and Enteromius cf. mimus
Although the occurrence of E. perince and E. stigmatopygus in the Lake Edward system was mentioned in several studies (e.g., Poll 1939;Greenwood 1966;Decru et al. 2019), we found that the specimens from the Lake Edward system, previously assigned to these species, clearly differed from the syntypes of these species in morphology. The genetic distance between the specimens from the Lake Edward system and sequences of E. perince from GenBank was large (12%). Based on our study, we can conclude that E. perince and E. stigmatopygus do not occur in the system. Instead, the specimens of Enteromius from the Lake Edward system with a smooth unbranched dorsal fin ray are assigned to either E. alberti ("group A") or E. cf. mimus ("group B"), which share a very similar colour pattern. However, no correspondence was found between specimens previously identified as E. stigmatopygus or E. perince and their current identification as E. alberti or E. cf. mimus. Specimens from the system, which were historically assigned to E. stigmatopygus, include both specimens of E. alberti and E. cf. mimus. The same applies to specimens historically assigned to E. perince. Banister (1987) considered E. alberti to be a synonym of E. stigmatopygus, although he did not examine the syntypes of E. alberti. The synonymy was based on a similar position of the gill rakers, the alignment of pharyngeal tooth rows and the position of the flank spots. However, the number of lateral line scales was different (Banister, 1987); 24 to 26 in E. stigmatopygus, according to Banister (1987), compared to 31 in E. alberti, according to Poll (1939. We found similar differences in the present study: E. alberti has 27 to 34 and the type series of E. stigmatopygus has 20 to 25 lateral line scales. Furthermore, in the original description of E. stigmatopygus, Boulenger (1903) mentioned that the syntypes had no barbels. However, Banister (1987) found one pair of posterior barbels in the three largest syntypes, and no sign of anterior barbels, but at the same time emphasized the poor condition of these syntypes. Also in the present study, we did not find distinct anterior barbels. Posterior barbels were found in the larger type specimens of E. stigmatopygus. In his redescription, Banister (1987) included additional specimens (amongst which the types of E. alberti) and stated that the anterior barbels appear later in development. The two pairs of barbels in E. alberti are clearly visible.
In the present study, we found the holotype and the examined paratypes of E. cercops to be conspecific with E. alberti. Enteromius cercops was described in 1960 from the Nzoia River (Nyanza Province, Kenya, Lake Victoria basin) by Whitehead. In addition to its type locality, E. cercops is also known from other rivers in the Lake Victoria basin, such as the Malawa River (Uganda) and rivers in the Middle Akagera system (Rwanda) (Whitehead 1960;Greenwood 1966;De Vos & Thys van den Audenaerde 1990;Seegers et al. 2003). We revalidated E. alberti with E. cercops as its junior synonym.
We found the specimens of "group B" to be most similar to E. mimus. The syntypes of E. mimus originated from the Euasso Nyiro River (Kenya), which drains to the Indian Ocean and is situated further east in the East Coast ichthyofaunal province (Fig. 10). Furthermore, E. mimus is only known from this type locality and from the Tana River system (Seegers et al. 2003). The type series of E. mimus is polyspecific as it included a specimen with serrations on the third unbranched dorsal fin ray (not included in the analyses), though Boulenger (1912) clearly mentioned the lack of serrations as a feature of the species. Further comparative research on the species of Enteromius from the coastal river systems in Kenya is required in order to identify this specimen. Since some small differences were observed between "group B" and E. mimus in the results of the PCA on the measurements (Fig. 6, Table 2), we assigned "group B" to E. cf. mimus, with "cf." indicating specimens that are comparable to E. mimus but whose identification is uncertain (Decru et al. 2016). Furthermore, "group B" and E. mimus occur in distant river systems. Therefore, it is unlikely that the specimens of the Lake Edward system are conspecific with the geographical distant species E. mimus, though this possibility  (Boulenger, 1912) (■), including the specimens from Tshambi, in the Lake Edward system. The location of the lectotype and the paralectotypes of E. alberti (♦). The approximate location of the lectotype and the paralectotypes of E. mimus (Boulenger, 1912) (▼) and the holotype and paratypes of E. cercops (Whitehead, 1960) (▲) are indicated in the inset. cannot be excluded. Genetic studies already revealed that widespread species of Enteromius often consist of multiple species that are actually confined to smaller distribution areas (Schmidt et al. 2017;Van Ginneken et al. 2017).While we found E. perince and E. stigmatopygus not to occur in the Lake Edward system, it is remarkable that we found a comparable set of similarly looking species in this system: E. alberti and E. cf. mimus, which mirror E. perince and E. stigmatopygus in the Nile River. Whether or not this implies sister-group relationships between species of Enteromius of the two ichthyofaunal provinces, is difficult to say without a more comprehensive genetic study. Furthermore, COI data of E. stigmatopygus is lacking. The occurrence of sister relationships between other species of Enteromius from the East Coast and other ichthyofaunal provinces has been suggested by Ndeda et al. (2018). These authors found sister relationship between E. yongei (from Lake Victoria, East Coast) and a lineage containing a specimen identified as E. sp. (region of Conakry) and one as E. stigmatopygus (Mafou River, Niger River) with a 12% genetic divergence in the cytochrome b (Cytb) gene (mtDNA). However, in view of the small sample size, especially of the West African lineage, the vast geographical distance and the use of only one mitochondrial marker, these results need further confirmation.

Unknown diversity in Enteromius
In a recent study on species of Enteromius from the north-eastern part of the Congo basin (Van Ginneken et al. 2017), DNA barcodes (COI) indicated the presence of 23 mitochondrial lineages within what initially were assumed to be only four different morphospecies. Genetic distances were mostly well over 5% and even up to 20% between lineages of morphologically similar specimens, clearly exceeding 2% which is the cut-off value that is commonly used to denote intraspecific variation (Hebert et al. 2003). In the study of Van Ginneken et al. (2017), clear-cut diagnostic characteristics could, however, not be delineated. Yet, morphologic differences could be demonstrated with multivariate morphometric analyses, which were consistent with the COI lineages. This unrecognised diversity was explained by the occurrence of a cryptic diversity within the species from the Congo basin. In the present study, we found seven characteristics to distinguish the two morphologically similar species of the Lake Edward system. Hence, although, a priori, we failed to distinguish the two species from each other, the concept of cryptic species is not applicable here. Another study (Schmidt et al. 2019) also revealed unrecognised diversity in E. foutensis from the Fouta Djallon (Upper Guinea region) highlands, both genetically (Cytb) and morphologically. Schmidt et al. (2017) also found evidence of high levels of genetic divergence and unrecognized diversity (Cytb) in specimens identified as E. kerstenii, E. apleurogramma and E. cf. paludinosus from the East Coast ichthyofaunal province in Kenya. Also, patterns of introgression seem to occur in both closely and distantly related species of Enteromius (Schmidt et al. 2017), which suggest hybridisation events. Whether this is also the case in the specimens of E. alberti and E. cf. mimus remains to be studied.
The results of these recent integrative studies on other systems raised the expectation to find more species in the Lake Edward system as well. This was, however, not the case for the group with an unbranched third dorsal fin ray. Initially, two species, though misidentified, were reported for this group, and during the present study no additional species were found.
The unrecognised diversity found within species complexes in various systems (e.g., Schmidt et al. 2017Schmidt et al. , 2019Van Ginneken et al. 2017), the revalidation of species as found in our study and the study of Schmidt et al. (2018), and the description of new species, such as E. pinnimaculatus from Gabon (Mipounga et al. 2019) and E. thespesios from the Upper Congo River (Katemo Manda et al. 2019) all suggest that the species richness of Enteromius is much larger than currently thought. Hence, the present number of 211 valid species of Enteromius (Froese & Pauly 2019), is certainly a considerable underestimation of the real species diversity.