INTRODUCTION

The caudal skeleton has been a rich source of taxonomically informative variation for the Teleostei and its subgroups. Numerous papers have documented and interpreted the structure of the internal supports of the caudal fin in bony fishes, both before and after the advent of phylogenetic thinking (see Monod, 1968; Schultze and Arratia, 1989 and references therein). This long record of investigation has made the caudal skeleton one of the most important sources of characters for elucidating phylogenetic relationships among teleostean fishes. Lundberg and Baskin (1969) were the first to study in detail the large degree of variation in the caudal skeleton of the order Siluriformes, and found many morphological patterns diagnostic for monophyletic groups therein. Since then, caudal skeleton characters have been regularly used in studies of relationships among catfishes. Despite relatively intense research focusing on the caudal skeleton and its variations within Siluriformes, it appears to us that a number of additional potentially informative characters still await discovery in that complex.

The aim of this paper is to report on a survey of a specific part of the caudal skeleton of siluriforms, namely the degree of development of the second ural centrum (U₂). The full development of the anterior half of that centrum, where it forms a complete intervertebral joint anteriorly and has a well- defined ossification center, occurs consistently in the adults of a few groups of catfishes, as first reported by Lundberg and Baskin (1969). This feature is unusual in siluriforms and unique among other Recent ostariophysans, yet it remains poorly understood as to its exact distribution and phylogenetic implications. Accordingly, we here conduct a comparative survey of the U₂ in various representatives of all siluriform families, aiming at documenting the distribution of the trait in as much detail as possible. On the basis of that, we propose that the degree of development of the second ural centrum is a well- defined character informative at various levels in the phylogeny of siluriforms.

This study started as an evaluation of the significance of the second ural centrum for the monophyly of the superfamily Sisoroidea, defined by de Pinna (1996) to include the Asiatic catfish families Amblycipitidae, Akysidae, Erethistidae, and Sisoridae, plus the neotropical Aspredinidae. The possible affinities of Aspredinidae with Asiatic taxa has important biogeographic implications, since it is the first case of a trans-Pacific relationship in South American freshwater fishes. Although a number of trans-Pacific relationships are known for the North American freshwater fish fauna, including both Recent and Fossil taxa (Grande, 1994: 68–74), no such case was known for South America. In the process of the study, taxonomic coverage was expanded to encompass the whole Siluriformes, and revealed data relevant also for understanding relationships of other subgroups within the order.

MATERIALS AND METHODS

Comparative material representing all currently recognized families of Siluriformes was examined as cleared and stained preparations, listed in appendix 1. Representatives chosen included the genera considered primitive or basal in their respective families, whenever that information is known and specimens were available. Most specimens were prepared according to the method of Taylor and Van Dyke (1985), which stains cartilage and bone, but some were prepared previously and were stained for bone only. The embryonic specimen of Pterobunocephalus was prepared inside the egg, by puncturing the egg case before the clearing and staining procedure. Illustrations were prepared with the aid of a microvideo system attached to a stereomicroscope. Images taken by the video system were printed with a color video printer and then traced by hand, against comparison with the actual specimen. Anatomical terminology follows Lundberg and Baskin (1969) and, for terms not treated therein, Monod (1968).

THE SECOND URAL CENTRUM IN SILURIFORMES AND OTHER TELEOSTEI

The ural centra in the caudal skeleton of adult teleosts are often compound structures formed ontogenetically by the fusion of auto- and chordacentra precursors corresponding to multiple individual centra. It has been argued that the so-called second ural centra in adult teleosts can be formed ontogenetically in markedly different ways and are not necessarily homologous in various lineages (Schultze and Arratia, 1988). A structure identifiable as a second ural centrum (U₂) is well developed in adults of the majority of lower teleosts, and this is probably the primitive condition for the group. Such is the case in most, or at least basal members of, osteoglossomorphs, elopomorphs, clupeomorphs, and esocoids (Monod, 1968; Schultze and Arratia, 1988; Arratia and Schultze, 1992), as well as most basal fossil teleosts (Patterson and Rosen, 1977). In Recent clupeomorphs, U₂ is most markedly developed in Denticeps clupeoides, sister group to all other Clupeiformes and sole Recent representative of suborder Denticipitoidei. The centrum is also well differentiated in the majority of other clupeomorphs, though usually smaller in relative size than that observed in Denticeps. The degree of development of U₂ is variable in most groups of lower Neoteleostei, including groups such as aulopiforms and protacanthopterygians (sensu Johnson and Patterson, 1996). A second ural centrum is not differentiated (or not independent) in adult acanthopterygians. Among ostariophysans, a well-developed autogenous second ural centrum is present in some fossil gonorynchiforms, such as Dastilbe, Tharrhias, and Parachanos (Poyato-Ariza, 1996), but not in Recent adult gonorynchiforms. Chanos has bilateral flanges extending dorsoposteriorly from its complex centrum (illustrated in Poyato-Ariza, 1996: fig. 19 and labeled “postero-lateral process of caudal terminal complex”). Topological similarities suggest that these structures may be remnants of the urostyle (sensu Monod, 1968) of a primitive U₂, itself fused or greatly reduced and adpressed to the posterior margin of PU₁ + U₁. A well-developed U₂ is also present in some siluriforms, as detailed in this paper, but not in other otophysans.

Lundberg and Baskin (1969) recognized for the first time that a few Siluriformes have a well-developed U₂ as adults. In those cases, the centrum forms a normal vertebral joint anteriorly with the compound caudal centrum (PU₁ + U₁), with a clearly recognizable conical articular surface intermediated with cartilage. Since its discovery, the presence of a well-developed U₂ has been considered to be of phylogenetic significance. Lundberg and Baskin observed that all other adult ostariophysans lack a U₂ autocentrum, and that its presence in some catfishes indicates that “the potential for its independent redevelopment has not been lost, at least in Siluriformes” (Lundberg and Baskin, 1969: 17). In light of the phylogeny currently accepted for ostariophysans (Fink and Fink, 1996) and lower teleosts (Johnson and Patterson, 1996; Arratia, 1997a, 1997b), it seems certain that the presence (or the redevelopment) of a well- developed U₂ in some catfishes should be interpreted as apomorphic. The Diplomystidae, considered as sister group to all other fossil and Recent siluriforms (Lundberg and Baskin, 1969; Grande, 1987; Arratia, 1987; Mo, 1991; de Pinna, 1998), do not have a well- developed U₂ as adults. Likewise, species of the fossil family Hypsidoridae, considered as basal siluriforms, also lack an identifiable U₂ (Grande, 1987; Grande and de Pinna, 1998). Finally, no instances of a well-developed U₂ are known in gymnotiforms, characiforms, or cypriniforms, closest relatives of siluriforms. Within Siluriformes, Lundberg and Baskin reported the presence of a well-developed U₂ in Helogeninae (Cetopsidae), Aspredinidae, Amblyceps (Amblycipitidae), and some Chiloglanis (Mochokidae). The authors consider that each of those have closest relatives lacking a full U₂, indicating that the structure has been redeveloped four independent times in siluriforms.

The U₂ in siluriforms, when well developed and in adults, is always entirely fused and continuous with the upper hypural plate. No siluriforms have a U₂ independent of hypural ossifications like in the primitive condition of other lower teleosts (which also include many cases of fusion). The second ural centrum in siluriforms, when present, tapers posteriorly, is fused with one or more of the upper hypurals, and does not have an articular surface on its posterior end. It should, thus, be more rigorously called a half-centrum, though for brevity in this paper it is just called a centrum. Also, it always lacks the peculiar dorsoposterior process (often tubular or semitubular) called a urostyle (in the terminology of Monod, 1968). A vestige of the urostyle may be represented by a ridge along the middorsal line of the centrum and of the associated hypural. In some aspredinids (such as Pterobunocephalus), the ridge may be quite deep anteriorly, indeed resembling an attenuated urostyle. This ridge accommodates the ventral margin of the uroneurals and seems to form a floor for the opisthural cavity. Overall, the condition of the caudal skeleton in catfishes with a well developed U₂ is markedly similar to that in some osteglossomorphs, such as Pantodon, Scleropages, and Heterotis (cf. Monod, 1968). Basal members of osteoglossomorphs, such as Hiodon, however, have an autogenous U₂ (Hilton, 2002) similar to that in basal Elopomorphs (such as Elops) and quite different from the siluriform condition. Some paracanthopterygians, such as percopsiforms (Rosen and Patterson, 1969: fig. 16) also have a condition similar to that in siluriforms, though there may be some question as to the actual homology of the second ural centrum in paracanthopterygians.

The well-developed condition of the U₂ in catfishes can be associated with some interesting structural modifications. In some taxa, such as aspredinids and akysids, the U₂ and associated upper hypural plate are extremely well developed, and are oriented nearly on the horizontal axis. In such cases, it seems like the upper hypural plate is the actual caudal-fin support, while the lower hypural plate appears superficially to be a large hemal spine. This arrangement must have important, yet unexplored, biomechanical implications in the movement of the upper lobe of the caudal fin in those fishes.

The degree of U₂ development in adults varies markedly among various catfish taxa. Three basic types can be recognized: Type 1: there is little or no trace of U₂ ossification. Hypural 3 (or the upper hypural plate) has no obvious proximal thickening and fits into a cavity on the posterodorsal end of PU₁ + U₁, immediately ventral to the base of the uroneural (fig. 1). This is the condition present in most catfishes and other ostariophysans. Type 2: the proximal end of hypural 3 or upper hypural plate has a small bony thickening, often having a tight contact, or even forming an incipient articulation, but without a well-defined cartilage disk, with the posterior portion of PU₁ + U₁ (fig. 2). Type 3: a well-formed, obvious vertebral centrum is attached to the base of hypural 3 or a compound upper hypural plate. This centrum has a well-defined articulation anteriorly with the posterior portion of PU₁ + U₁, intermediated by cartilage, and a conical ossification center visible by transparency (fig. 3). In anterior view, the similarity in structure between the U₂ and a normal vertebral centrum is evident (fig. 4).

Type 1 U₂ is present in outgroup ostariophysans, and is considered as the primitive state for catfishes. Type 2 seems to be subject to considerable degree of intraspecific variation; its delimitation as a discrete character state is difficult and we consider its phylogenetic significance to be uncertain at this point. In fact, the homology of the bony thickening at the base of the upper hypural with a ural centrum remains to be directly demonstrated by ontogenetic data. Type 3 is a condition derived within siluriforms; it seems to form a well-defined character state, with few cases of intermediate conditions and little or no relevant intraspecific variation in the samples examined. Although more in-depth studies may find a basis for considering U₂ types 1, 2, and 3 as a multistate character ordered in that sequence, we do not consider that to be warranted at present. Only type 3 can currently be considered as objective evidence of relationship.

ONTOGENETIC DATA

The ontogenetic fate of the second ural centrum seems to differ in various catfish taxa. Lundberg and Baskin (1969) showed that the most common condition in siluriforms is to have a U₂, or some remnant thereof, fused to the bases of hypurals 3 and 4, but noticed that in trichomycterids it is fused to the base of hypural 3 only. According to those authors, any remnants of the second ural centrum are associated with the base(s) of upper hypurals, and not with the compound centrum. Arratia (1983) confirmed that U₂ is fused to the base of hypural 3 in some Trichomycterinae, but argued that in Nematogenyidae it is instead fused to the posterior portion of the compound centrum. An alternative interpretation, similar to Arratia's on Nematogenyidae, was advanced by Fujita (1992) and Ichiyanagi and Fujita (1995), on the basis of observations on species of Clariidae (Siluriformes) and Cobitidae (Cypriniformes), respectively. In their view, the structure called PU₁ + U₁ by Lundberg and Baskin (1969) was considered to include the second ural centrum as well (as PU₁ + U₁₊₂). Thus, the vertebral centrum material associated with hypural 3, considered by Lundberg and Baskin to be the U₂, would actually be homologous to a more posterior ural centrum (U₃). This idea was not based on direct observation of fusion, but apparently inferred on the basis of the joint association of parhypural, hypural 1, and hypural 2 to a single centrum element during ontogeny, thereby implying that this element is compound from three primitive centra. The compound nature of the teleostean U₁ has been suggested in the past (Nybelin, 1971) on the basis of its association with hypurals 1 and 2 in the vast majority of teleosts, fossil and Recent (Patterson and Rosen, 1977; de Pinna, 1996; Arratia, 1997b). Schultze and Arratia (1988) went further, and proposed not only that U₁ was formed by fusion of two centra (as U₁₊₂), but also that U₂ was formed by fusion of three other centra (as U₃₊₄₊₅). Such complex fusions have not been corroborated by other authors (Hilton, 2002), and have also not been confirmed in the material examined for this paper. While some of the individual ural and preural centra may indeed be compound in origin, evidence for that is controversial. Therefore, we prefer to maintain the terminology of Lundberg and Baskin until more direct evidence of the exact composition of the compound ural centrum in ostariophysans and other teleosts is available. We would further note that hypural 1 originates, in at least one siluriform (Fujita, 1992) and one cypriniform (Ichiyanagi and Fujita, 1995), as a basal anteroventral offshoot of the cartilaginous hypural 2, rather than as an independent element (this information disagrees with Arratia, 1997b: 309, who stated that hypurals 1 and 2 are not connected by cartilage in juvenile ostariophysans). This is perhaps also the case in juveniles of many other teleosts, where cartilaginous hypurals 1 and 2 are fused at their bases at a stage when all other hypurals are independent (cf. Monod, 1968: figs. 30, 32, 61; Patterson and Rosen, 1977: figs. 26, 28). Although separate chondrification and subsequent cartilage fusion in early development has been documented for some taxa (cf. Fujita, 1994, in the cichlid Tilapia and Arratia and Schultze, 1992, in the Salmonid Onchorhynchus), it seems possible that hypural 1 in many, perhaps most, teleosts is just an ontogenetic offshoot from hypural 2. If so, it should not be expected to be associated with its own centrum in the vertebral series, and the centrum that supports hypurals 1 and 2 in the majority of teleosts would indeed be a single element. In that case, the compound centrum for the parhypural plus hypurals 1 and 2 in siluriforms and other ostariophysans would not involve three or more centra, but only two (PU₁ + U₁), as normally accepted. However, general conclusions along these lines need further information on early chondral differentiation of the elements of the caudal skeleton, not yet available for most relevant taxa.

The development of the second ural centrum in catfishes in general is still poorly documented, and current ideas are based on observations on few taxa. Ontogeny of U₂ in groups where it displays a type 3 degree of development is virtually unknown, in great part due to the difficulty in obtaining growth series of the relevant taxa. Juveniles of the superfamily Sisoroidea are rare in collections. For this study, we obtained juvenile sisoroid material of Gagata (Sisoridae) and Aspredo (Aspredinindae), plus late embryonic material of Pterobunocephalus (Aspredinidae). Although no growth series were available, observation of juvenile conditions is nonetheless informative about the formation of the second ural centrum in those taxa. In the smallest specimen examined, a Pterobunocephalus embryo still inside an eggshell, hypurals 3 and 4 are already fused to U₂ (fig. 5A). Still, there is little doubt that the structure is actually a ural centrum, because it forms a continuous series with other centra anterior to it. Concomitantly, its homology with the respective centrum in adults is also obvious (fig. 5B). Similar situations are seen in later-stage juveniles of Gagata (fig. 6) and Aspredo (MZUSP 77694). The configurations in these specimens rule out the possibility that the structure identified as U₂ might be a centrum-like specialization of the proximal part of the upper hypural complex, that is, not really a centrum, but simply a hypural modification gross-morphologically similar to a centrum in adults.

In catfishes without a well-developed U₂, very little centrum material, if any, is associated with the bases of hypurals 3 and 4. In very small juvenile Trichogenes longipinnis (Trichomycteridae) examined, no ossifications can be clearly identified as U₂ material during the ossification of the caudal skeleton. At the earliest stage available (fig. 7A), the limits between centra can be seen as membranous separations, and the region corresponding to U₂ is independent of any hypurals. It remains so as ossification progresses (fig. 7B). Whatever little U₂ material that gets ossified seems to be incorporated into the posterior region of the compound centrum and/or the mesial surface of the uroneural. The bone thickening at the base of hypural 3 in adults is entirely composed of hypural material, since an evident thickening is already evident in the cartilaginous phase of the hypural, before the onset of ossification. The same situation holds for juvenile Clarias gariepinnus (NRM 15319). Thus, the generally held assumption that U₂ is fused to the bases of hypurals 3 (trichomycterids) or 3 and 4 (all other catfishes) does not seem ontogenetically corroborated in the taxa examined. The fate of U₂ material may not be uniform among various catfish taxa.

Also, we found only little evidence of fusion between PU₁ and U₁. The region corresponding to the two centra before the onset of ossification is always continuous in all juveniles examined. On the other hand, a larger juvenile of Trichogenes (fig. 7B) shows a tenous line of suture at the expected limit between PU₁ and U₁ which was not evident before ossification (fig. 7A). Perhaps fusion of the precursors of the two centra occurs very early in development, yet beginning of ossification still reveals a vestige of the primitive separation between them. In adults, fusion is again complete.

THE SECOND URAL CENTRUM IN SILURIFORM FAMILIES

We were able to examine several taxa not available to other authors and to verify that U₂ forms a full vertebral joint anteriorly in several additional catfish taxa not previously recorded. Appendix 1 summarizes our results, with a list of the material examined and an indication of the degree of development of the second ural centrum for Siluriforms (types 1, 2, or 3; see above). Below we provide a discussion on each of the catfish families known to have at least one instance of a well-developed U₂.

PHYLOGENETIC IMPLICATIONS OF THE PRESENCE OF A WELL-DEVELOPED U₂

The presence of a type 3 second ural centrum in the cases discussed above can be mapped onto phylogenetic schemes currently accepted for various siluriform clades. It also provides additional evidence for some unsettled or controversial questions on the relationships of different subgroups of catfishes, at widely divergent levels of generality.

The most important phylogenetic implications of the second ural centrum are in the superfamily Sisoroidea, specifically the inclusion of the neotropical Aspredinidae therein, the only non-Asiatic taxon in the group. The first time aspredinids were aligned with Asiatic taxa was in Ferraris (1989), who proposed the family as the sister group to Akysidae. Chen (1994) placed aspredinids as the sister group to asiatic sisoroids (Sisoridae—including Erethistidae— Amblycipitidae, and Akysidae). Friel (1994), on the other hand, suggested that aspredinids were more closely related to doradoids (comprising the neotropical families Doradidae and Auchenipteridae—including Centromochlidae and Ageneiosidae—plus the African Mochokidae) than to any other siluriforms. De Pinna (1996) hypothesized that aspredinids were related to sisoroids, as the sister group to Erethistidae (members of which were formerly included in Sisoridae). The evidence for the monophyly of the suborder so far includes three relatively homoplasy- free synapomorphies (de Pinna, 1996): (1) the compressed and vertically expanded posterior center of ossification of the palatine; (2) the articular region of the lateral ethmoid laterally produced, with the articular facet for the palatine at the tip; and (3) the presence of a humero-vertebral ligament (connecting the humeral process—or the soft tissue in the humeral region—to the anterior portion of the vertebral column). Other potential characters exist, but these depend on more extensive ad hoc hypotheses of reversal within sisoroids and convergence in other catfish groups (for these, see de Pinna, 1996). All families included in Sisoroidea by de Pinna (1996) can be hypothesized as primitively having a fully formed U₂. The few exceptions therein are parsimoniously interpreted as secondary reversals, based on congruence with other character data. The presence of a well-developed, type 3 U₂ can be interpreted as additional support for the monophyly of Sisoroidea, including the neotropical Aspredinidae.

The presence of a type 3 U₂ in helogenines is the most well documented in the literature (cf. Lundberg and Baskin, 1969; Vari and Ortega, 1986). It seems to constitute an autapomorphy for the subfamily. It has been demonstrated (de Pinna, 1993; de Pinna and Vari, 1995) that former Cetopsidae and Helogenidae constitute a monophyletic group, now recognized as an expanded Cetopsidae (comprising monophyletic subfamilies Cetopsinae and Helogeninae). Cetopsines do not show a well-developed U₂, as confirmed by our own observations and those of other authors. There is some indication that the phylogenetic position of cetopsids may be close to the base of the siluriform cladogram (Mo, 1991, in part; de Pinna, 1993; de Pinna and Vari, 1995), perhaps as the sister group to all other non-diplomystid catfishes. If that is so, the immediate outgroups to helogenines are, in sequence, cetopsines, a clade containing most other catfishes, and diplomystids. This indicates that the condition of U₂ in helogenines can only be parsimoniously optimized as a transition at the base of the subfamily, and is therefore autapomorphic for it. A well-developed U₂ should be included among the diagnostic features of the subfamily Helogeninae.

Among Amphiliidae, the type 3 U₂ is present in Zaireichthys zonatus and a subgroup of Leptoglanis that includes forms similar in general aspect to Zaireichthys (L. rotundiceps and L. brevis). Therefore, it may be a synapomorphy indicating that those species are more closely related to Z. zonatus than to remaining species of Leptoglanis and that the latter genus may not be monophyletic. The type species of Leptoglanis, L. xenognathus, has the plesiomorphic type 1 U₂. Considering that amphiliids (including Leptoglanis) are a monophyletic group (de Pinna, 1993), the type 3 U₂ that occurs in some members of the family is convergent with that in other catfishes.

Remaining occurrences of a Type 3 U₂ seem to be synapomorphic for very small clades. In mochokids, it is likely a synapomorphy for a subgroup of Chiloglanis. Similarly, the occurrence in Pseudopimelodus is either autapomorphic for P. raninus or synapomorphic for a subset of the genus. In Auchenipteridae, type 3 U₂ occurs in two taxa which are not close relatives (see above), and which must be hypothesized as results of separate events.

Lundberg and Baskin (1969) estimated that a fully developed U₂ developed independently four times within siluriforms: in Helogenidae, Aspredinidae, Amblycipitidae, and Mochokidae. Considering the new descriptive data within an updated phylogenetic context, it seems that the conditions in Amblycipitidae and Aspredinidae are homologous. On the other hand, other occurrences imply additional cases of convergence, such as in some amphiliids, some pseudopimelodines, and twice in auchenipterids. A total of six events can thus be hypothesized: (1) Helogeninae (Cetopsidae); (2) Pseudopimelodus raninus (Pimelodidae); (3) Sisoroidea (Amblycipitidae, Akysidae, Sisoridae, Aspredinidae, and Erethistidae); (4) Zaireichthys and some Leptoglanis (Amphiliidae); (5) Entomocorus (Auchenipteridae); and (6) Asterophysus (Auchenipteridae).