CYPRINIFORMES

Conway, Kevin W., 2011, Osteology of the South Asian Genus Psilorhynchus McClelland, 1839 (Teleostei: Ostariophysi: Psilorhynchidae), with investigation of its phylogenetic relationships within the order Cypriniformes, Zoological Journal of the Linnean Society 163 (5), pp. 50-154 : 140-142

ORDER CYPRINIFORMES

The most comprehensive morphological phylogenetic investigation of the interrelationships of the order Cypriniformes conducted prior to the present study is that of Siebert (1987). Based on his analysis of 41 morphological characters, Siebert (1987) proposed a classification in which the Cypriniformes was divided into two major subgroups: the superfamily Cyprinioidea [for the sole inclusion of the Cyprinidae (inclusive of Psilorhynchus )] and the Cobitoidea (including all non-cyprinid cypriniforms). The phylogenetic hypothesis of cypriniform interrelationships presented herein is largely congruent with Siebert (1987), differing only in the interfamilial relationships of the Cobitoidea .

In the phylogenetic hypothesis proposed by Siebert (1987) the family Gyrinocheilidae was recovered as the sister group to all other cobitoids ( Fig. 2B ). In the present analysis, Gyrinocheilidae was recovered as the sister group to the Catostomidae based on six derived characters, including: (1) presence of an anterolateral process on the lateral ethmoid (character 7, 0> 1); (2) presence of a pterotic fossa at the origin of the levatator operculi (character 18; 0> 1); (3) connection between infraorbital sensory canal and preopercular sensory canal (character 49, 1> 0); (4) absence of mandibular sensory canal (59, 1> 0); (5) absence of hypobranchial 3 (hypobranchial 3 cartilage present only) (character 76, 1> 0); and (6) fifth vertebral centrum with enlarged zygapophysis contacting neural complex (character 100, 0> 1). Siebert (1987) made note of numerous similarities between the gill arches of Gyrinocheilus and members of the Catostomidae , including the presence of pharyngobranchial 1 (pharyngobranchial 1 cartilage only present in catostomids), the presence of epibranchial uncinate processes, and the absence of hypobranchial 3 (hypobranchial 3 cartilage present), but considered these to be either plesiomorphic features or the result of convergence, and rejected a close relationship between the two ( Fig. 2B ). Although I concur with Siebert (1987) and consider the presence of pharyngobranchial 1 and epibranchial uncinate processes to be plesiomorphic features at the level of Cypriniformes , based on a broad survey of other ostariophysan fishes and basal teleosts, I cannot accept the absence of hypobranchial 3 to be a plesiomorphic feature, nor can I accept (based on the results of the present phylogenetic analysis) its absence in Gyrinocheilus and catostomids to be the results of convergence. In addition to this derived feature of the gill arches, Gyrinocheilus shares a number of other derived osteological traits with members of the Catostomidae (listed above) that further support a close relationship between the two. A sister-group relationship between the Gyrinocheilidae and Catostomidae has also been recovered in a number of molecular phylogenetic studies using mitochondrial genomic data ( Tang et al., 2005; Saitoh et al., 2006), but not in more recent molecular phylogenetic investigations using data from the nuclear genome (e.g. Šlechtová et al., 2007; Chen et al., 2008).

The relationships amongst the remaining families of the Cobitoidea presented by Siebert (1987), which is congruent with the hypothesis proposed earlier by Sawada (1982), differ markedly from the hypothesis presented herein. Both Sawada (1982) and Siebert (1987) proposed a sister-group relationship between the spined loaches of the families Cobitidae and Botiidae (recognized by both authors as subfamilies of a single family, the Cobitidae ). Sawada (1982) lists two characters in support of this relationship [mobility of the lateral ethmoid (equivalent to character 8 herein), and the presence of an articulation between the supraethmoid (= supraethmoid portion of the ethmoid complex) and the frontals (equivalent to character 4 herein)], and Siebert (1987) lists the same two characters plus an additional two [the presence of a rectis communis process on hypobranchial 1, and the absence of an anterolateral process on the lateral ethmoid (equivalent to character 7 herein; a reversal to a plesiomorphic condition in Siebert’s analysis)]. Although three of these four characters were included in the present phylogenetic analysis (ventral gill-arch musculature was not examined, and as such it was difficult to confirm the insertion point of the rectis communis muscle on hypobranchial 1 in the cleared and stained specimens examined), the spined loaches were not recovered as a monophyletic grouping, with members of the Cobitidae being more closely related to members of the Balitoridae and Nemacheilidae than to members of the Botiidae . This relationship is supported by three derived characters, including: the presence of gill-filament ossifications (character 66, 0> 1); (2) the presence of basibranchial 4 (character 74, 0> 1); and (3) the absence of hypural 6 (character 122, 0> 1). The distribution of an erectable suborbital spine formed by the lateral ethmoid (character 8) amongst the members of the Cobitoidea was interpreted to be the result of convergence in the present analysis, being gained independently once each at the level of clades O and Q. The presence of an articulation between the ethmoid complex and the frontals (character 4) is similarly interpreted to be the result of convergence. A recent molecular phylogenetic investigation of the Cobitoidea by Šlechtová et al. (2007), based on RAG1 sequence data, also failed to recover the spined loaches as a monophyletic group. In their phylogenetic hypothesis members of the Botiidae were recovered as the sister group to all remaining loach families, with members of the Cobitidae forming the sister group to a clade composed of members of the Nemacheilidae and Balitoridae . Rather interestingly, Šlechtová et al. (2007) found Serpenticobitis (a taxon with an erectible suborbital spine, considered to belong to the Cobitidae ; Roberts, 1997) to be more closely related to members of the Balitoridae than to the other spined loaches. Šlechtová et al. (2007) interpreted the presence of an erectable suborbital spine formed by the lateral ethmoid (which they incorrectly referred to as the subethmoidal head bone) as a plesiomorphic feature of loaches rather than the result of convergence between different lineages of loaches. Despite the results of the present analysis (two independent origins of the lateral ethmoid spine amongst cobitoid fishes), I find it hard to accept that the lateral ethmoid would have become modified in such a way as to form a complex erectable spine on three separate occasions throughout the evolutionary history of the loaches, and I favour Šlechtová et al.’s. (2007) interpretation, that this character may be a plesiomorphic feature of loaches in general, with its absence in other loaches explained by the re-expression of the cypriniform plesiomorphic condition rather than the result of convergence in three separate clades of loaches. Unfortunately, specimens of Serpenticobitis were not available for examination during the course of this investigation, and as such its suborbital spine could not be compared with the suborbital spine of members of the Botiidae and Cobitidae .

Sawada (1982) and Siebert (1987) both reached different conclusions regarding the phylogenetic placement of Vaillantella , an enigmatic loach possessing a longer dorsal fin than any other cypriniform fish. Based on an examination of radiographs, Sawada (1982) provisionally considered Vaillantella to be more closely related to members of the Nemacheilidae , whereas Siebert (1987), based on the examination of cleared and stained material, concluded that Vaillantella was more closely related to members of the Botiidae (a similar conclusion to that reached earlier by Nalbant & Bãnãrescu, 1977; and again recently by Nalbant, 2002). Neither Sawada (1982) nor Siebert (1987) provided character evidence in the form of synapomorphies to support their conclusions, with both authors simply listing a number of similarities between Vaillantella and members of the different groups to which they proposed a close relationship. In the present phylogenetic analysis Vaillantella was recovered as the sister group to the Botiidae , based on two derived characters: (1) absence of supraneural 2 (character 96, 0> 1); and (2) interdigitation between ventral pre- and postzygopophyses of caudal vertebrae (character 102, 0> 1). Neither of these characters represent unique support for the sister-group relationship between Vaillantella and the Botiidae , as both are exhibited by certain members of the Nemacheilidae ( Lefua costata also lacks supraneural 2 and Acanthocobitis cf. zonalternans and Mesonemacheilus triangularis exhibit strong interdigitation between the ventral pre- and postzygopophyses of caudal vertebrae). Although common in non-cypriniform otophysans ( Fink & Fink, 1981; Hoffman & Britz, 2006), the absence of supraneural 2 from the anterior edge of the neural complex is highly unusual amongst cypriniform fishes. It is unclear whether the absence of supraneural 2 in the aforementioned taxa arises from the complete absence of supraneural 2 or the result of fusion between supraneurals 2 and 3 during ontogeny. Recent molecular phylogenetic investigations have recovered Vaillantella either as the sister group to the Botiidae ( Saitoh et al., 2006) or as the sister group to a large clade of loaches including members of all families, but excluding Botiidae ( Šlechtová et al., 2007) .

One last point of incongruence between the phylogenetic hypothesis presented herein and the results of previous phylogenetic investigations on cypriniform fishes concerns the Nemacheilidae . Sawada (1982) provided the first evidence that members of the Nemachilidae (which he referred to as the Nemacheilinae and treated as a subfamily of the Balitoridae ) were more closely related to members of the Balitoridae than to the spined loaches ( Cobitidae and Botiidae ), as had been considered previously (e.g. Berg, 1940). All subsequent morphological investigations, including the present investigation, support a close relationship between the nemacheilids and balitorids ( Siebert, 1987; Conway & Mayden, 2007), as do the majority of molecular phylogenetic studies (e.g. Saitoh et al., 2006; Šlechtová et al., 2007). In the present analysis, however, the Nemacheilidae was not recovered as a monophyletic group, with Barbatula barbatula and Lefua costata being more closely related to members of the Balitoridae than to Acanthocobitis cf. zonalternans and Mesonemacheilus triangularis . Although it is not strongly supported by osteological characters ( Sawada, 1982; Conway et al., 2010), the Nemacheilidae have been recovered as monophyletic in all recent molecular phylogenetic studies that have included more than two species ( Šlechtová et al., 2007; Bohlen & Ŝlechtová, 2009).

Although the results of both morphological and molecular phylogenetic studies of cypriniform fishes provide strong support for a monophyletic loach clade, the relationships recovered within this clade by morphological and molecular data sets are in conflict ( Figs 2 , 43). A more in-depth study on the morphology of the loaches was beyond the scope of this study, but should be addressed by future morphological studies on cypriniform fishes.

Figures

Figure 2. Previous hypotheses of relationships amongst cypriniform fishes. A, Hora’s views on the evolution of the Balitoridae and Psilorhynchus; redrawn from Hora (1920). B, phylogenetic relationships of cypriniform fishes from Siebert (1987), based on parsimony analysis of 41 morphological characters. C, phylogenetic relationships of cypriniform fishes from Conway & Mayden (2007), based on parsimony analysis of 24 morphological characters. D, phylogenetic relationships of cypriniform fishes from Šlechtová et al. (2007), based on Bayesian analysis of RAG1 sequences. E, phylogenetic relationships of cypriniform fishes from He et al. (2008), based on Bayesian analysis of mitochondrial sequences. Grey and black vertical bars indicate the superfamilies Cyprinoidea and Cobitoidea, respectively, in B–E. Non-cyprinid clades collapsed to single terminals in D and E.