Printer Friendly

Teleostean phylogeny based on osteological and myological characters/Filogenia de teleosteos basada en caracteristicas osteologicas y miologicas.

INTRODUCTION

The Teleostei, with an estimated of about 28000 living valid species, is the most speciose group of vertebrates (Nelson, 2006). The extraordinary taxonomic diversity of teleosts is accompanied by a remarkably variety of morphological features and adaptations to very different freshwater, brackish, and marine habitats, from high elevation mountain springs over 5000 meters above sea level to the ocean abyss almost 8500 meters below (e.g. Arratia, 2000; Stiassny et al., 2004).

Due to restrictions of size and to the high number of characters described and of clades diagnosed in the present paper, the Introduction, as the other Sections of the paper, will be short and concise and, thus, it is not possible to provide here a detailed historical account of all the works dealing with teleostean phylogenetic relationships. Such a detailed account was provided, for instance, by Arratia (2000). Springer & Johnson (2004) have recently presented, in their figure 3 (i.e. in the introduction of their work), a cladogram that summarizes what they consider to be, in face of a comprehensive review of the literature, the most supported scenario regarding teleostean higher-level phylogeny (see Fig. 1). However, as stressed by those authors, this scenario is far from being consensual among teleostean specialists. For instance, some authors argue that the most basal extant teleostean group is the Elopomorpha and not the Osteoglossomorpha (e.g. Arratia, 1997, 1999), while other authors argue that the Elopomorpha is not even a monophyletic unit (e.g. Filleul, 2000; Filleul & Lavoue, 2001). Researchers such as Ishiguro et al. (2003) defend that some otocephalans are more closely related to the 'protacanthopterygian' alepocephaloids than to other otocephalans (e.g. Ishiguro et al.). This has major implications for teleostean higher-level phylogeny, because this would mean that two of the four major extant teleostean groups, i.e. the Neoteleostei and Otocephala (the two other groups being the Osteoglossomorpha and Elopomorpha: Fig. 1) are in fact not natural groups. These latter researchers also defend that the remaining, non-alepocephaloid 'protacanthopterygian' groups (i.e. their Esociformes, Salmoniformes, Osmeriformes and Argentinoidea) do form a monophyletic 'Protacanthopterygii' clade, contrary to what is accepted by most authors nowadays (see Fig. 1). These are just a few examples to illustrate that, despite the progresses done in the field of teleostean phylogeny in the last decades, recent studies continue to raise controversial questions about the higher-level relationships of this remarkably diverse group of fishes. A more extensive analysis of the major current controversies concerning teleostean phylogeny will be provided in the discussion below.

[FIGURE 1 OMITTED]

The main aim of the present work is to help to clarify the higher-level relationships of lower teleosts, because a basic understanding of this subject is crucial to pave the way for analyses on the evolutionary history of these fishes.

MATERIAL AND METHOD

In order to clarify the higher-level phylogeny of teleosts we undertook a cladistic analysis based in numerous osteological and myological characters. The inclusion of a large number of myological characters is one of the main differences between the present study and previous morphological cladistic analyses on teleostean relationships, as these latter were almost exclusively based on osteological features (see e.g. Stiassny et al.). The use of numerous myological characters in the cladistic analysis allows testing if these characters support, or not, the results of previous studies based essentially on osteological characters. With these myological characters, the present cladistic embraces a total of 271 phylogenetic characters, a number that is significantly higher than that used in previous morphological cladistic analyses of Teleostei higher-level phylogeny. The other main difference with previous analyses is the attempt to include, in a same cladistic analysis, representatives of all the major non-Neoteleostei groups (see Fig. 2). The necessity of including representatives of taxa such as the Anguilliformes, the Saccopharyngiformes or the Alepocephaloidea in a cladistic analysis of teleostean higher-level phylogeny been stressed by authors such as Forey et al. (1996), Arratia (1999, 2004), Filleul (2000), Belouze (2002), Filleul & Lavoue, Ishiguro et al., Diogo (2004) and Stiassny et al.

Although it is not possible to explain with much detail the reasons for the choice of each of the 70 terminal taxa included in the cladistic analysis, the most relevant points concerning this choice are briefly summarized. First of all it is important to note that the great majority of these terminal taxa concern extant groups, although some fossils are also included. The main reason for this is precisely the fact that we use a great number of characters referring to the configuration and presence of muscles, ligaments and cartilages. Thus, the fossil taxa that were included in the cladistic analysis of the present work, i.e. [dagger] Chanoides macropoma, [dagger] Clupavus maroccanus, [dagger] Santanichthys diasii, [dagger] Lusitanichthys characiformis and [dagger] Sorbininardus apuliensis were chosen for a major, precise reason: authors such as Gayet (1981, 1985, 1986), Taverne (1977a, 1995, 1999) and Filleul & Maisey (2004) have argued that these are particularly 'problematic' fossil taxa that, if included in an explicit cladistic analysis together with other ostariophysan and non-ostariophysan taxa, could well show that the four extant otophysan orders, and possibly the clade Ostariophysi, as currently recognized, do not form monophyletic groups. As the ostariophysan fishes play an important, central role for a proper understanding of the higher-level phylogeny and evolution of lower teleosts (e.g. Fink & Fink, 1981, 1996), and as the testing of the monophyly of the otophysan, ostariophysan and otocephalan fishes was precisely among the main aims of the present work, we decided to include these five fossil taxa in the cladistic analysis. The inclusion of these fossils in an explicit cladistic analysis is also crucial to clarify a major question in the evolution of teleosts: if the characteristic Weberian apparatus of extant otophysans was, or not, acquired just once in the evolutionary history of these fishes (see e.g. Gayet, 1981, 1985; Fink & Fink, 1981, 1986; Taverne, 1995; Filleul & Maisey). The reason for including these five fossil taxa and not, for example, other 'problematic' otophysan fossil taxa sensu e.g. Gayet (1981, 1985, 1986), such as [dagger] Salminops ibericus, is that these five taxa are particularly well-conserved, what is clearly not the case of taxa such as [dagger] S. ibericus.

Concerning the 65 extant taxa included in the cladistic analysis (see Fig. 2), Amia and Lepisosteus were chosen in order to have two extant outgroup taxa in the analysis representing both the Halecomorphi and the Ginglymodi (see Fig. 1), which can thus help to polarize both the osteological and myological characters used. Representatives of each of the four osteoglossiform extant families sensu Hilton (2003) are included in the cladistic analysis: Hiodon (Hiodontidae), Pantodon (Osteoglossidae), Xenomystus (Notopteridae) and Mormyrus (Mormyridae). Representatives of all five extant elopomorph orders (see Fig. 1) are also included: Elops and Megalops (Elopiformes), Albula (Albuliformes), Notacanthus (Nothacanthiformes), Anguilla and Conger (Anguilliformes) and Eurypharynx (Saccopharyngiformes). The five extant ostariophysan orders are also covered in the analysis, including all extant gonorynchiform genera: Chanos, Gonorynchus, Phractolaemus, Kneria, Parakneria, Cromeria and Grasseichthys (Gonorynchiformes), Opsariichthys, Barbus, Danio, Cobitis and Catostomus (Cypriniformes), Xenocharax, Distichodus, Citharinus and Brycon (Characiformes), Sternopygus, Gymnotus and Brachyhypopomus (Gymnotiformes) and Diplomystes, Nematogenys, Trichomycterus, Callichthys, Cetopsis, Silurus, Pimelodus, Bagrus and Chrysichthys (Siluriformes). The four major extant clupeomorph groups (see Fig. 1) are also represented: Denticeps (Denticipitoidei), Ilisha (Pristigasteroidea), Ethmalosa (Clupeoidea) and Thryssa and Engraulis (Engrauloidea). All the major extant groups of 'Protacanthopterygii' sensu Ishiguro et al. (see Fig. 1) are represented: Coregonus, Thymallus and Salmo (Salmoniformes), Stokellia, Retropinna and Galaxias (Galaxioidea), Osmerus and Plecoglossus (Osmeroidea), Searsia, Xenodermichthys and Alepocephalus (Alepocephaloidea), Argentina and Bathylagus (Argentinoidea) and Umbra and Esox (Esociformes). Lastly, in order to test the monophyly versus paraphyly of the 'Protacanthopterygii' (see above), we have included in the cladistic analysis four representatives of two of the most basal neoteleostean orders, i.e. Stomias and Astronesthes (Stomiiformes) and Aulopus and Chlorophthalmus (Aulopiformes) (see Fig. 1).

[FIGURE 2 OMITTED]

With exception to the five fossil taxa included in the cladistic analysis, which were not directly examined by us and are thus coded following exclusively their descriptions in the literature (i.e. Taverne, 1977a, 1995: [dagger] Clupavus maroccanus; Gayet, 1981, 1985: [dagger] Lusitanichthys characiformis; Patterson, 1984: [dagger] Chanoides macropoma; Taverne, 1999; [dagger] Sorbininardus apuliensis; Filleul & Maisey: [dagger] Santanichthys diasii), we have personally checked the characters listed in the Appendix 2 for all extant taxa included in the analysis. The phylogenetic procedure employed for proposing hypotheses of relationships is the cladistic methodology: parsimony was employed to find the hypothesis best supported by the analyzed data, using both the Hennig86 (Farris, 1988) and the Nona & Winclada (Nixon, 2002) computer programs. The Implicit Enumeration algorithm (ie*) was employed in the search for the most parsimonious cladograms, with Nona & Winclada used to check the most parsimonious results found with this algorithm. Tree manipulations and diagnostics were done with the help of Nona & Winclada. Multistate characters were ordered, following Diogo (2004). As explained above, Amia and Lepisosteus were used as outgroups; the descriptions given in the literature concerning numerous other basal actinopterygians, either fossil or living, were also taken into account in the polarization of the characters. Autapomorphies for the different taxa examined were actively searched for and included in the analysis. The complete list of the 271 morphological characters included in the cladistic analysis is given in the Appendix 2.

RESULTS AND DISCUSSION

Diagnoses for clades obtained and comparison with previous hypotheses.

The characters listed in Appendix 2 were coded for each of the taxa included in the cladistic analysis, resulting in the data matrix shown in Table I. The phylogenetic analysis of these characters resulted in 48 equally parsimonious trees with a length of 680, CI=0.39, and RI=0.76. Fig. 2 shows the phylogenetic relationships between the teleostean groups included in the analysis according to the results obtained by using the 'majority fools' option of Nona & Winclada, which shows all clades that are supported by more than 50% of these 48 most parsimonious trees and thus provides more information than that given by the use of the 'strict consensus' option (Nixon). The terms from 'C1' to 'C59' indicate the number of the clades, following the order given in the synapomorphy list provided below. The numbers 100 and 66 (not followed by a "%") shown above the numbers of the clades indicate the percentage of most parsimonious trees supporting the respective clades. Bootstrap values for 1000 replicates are shown (below or in front of the number of the respective clade, in percentage, i.e. followed by a "%") on branches for which these values are [greater than or equal to] 50%. As can be seen in Fig. 2, of the 59 clades obtained, 43 (i.e. about 73%) have bootstrap values [greater than or equal to] 50%, and, within these 43 clades, 26 (i.e. about 60%) have bootstrap values [greater than or equal to] 70% (these values need to be interpreted in the context of a morphological cladistic analysis including 271 characters; when a higher number of phylogenetic characters is used to perform a bootstrap analysis-e.g. genomic level sequence cladistic analyses may include several thousands of characters-the bootstrap values obtained tend to be higher: e.g. Hillis & Bull, 1993; Rokas et al., 2003). The synapomorphy list provided below includes, for most clades, a commentary and a comparison with previous hypotheses; the numbering for diagnostic characters follows that of Appendix 2. Character state changes mentioned in this list are restricted to those unambiguous character states changes occurring in the different nodes, and can be included in two main categories: 1) state changes occurring exclusively in a certain node (in bold); 2) state changes subsequently reversed in a more terminal node and/or independently acquired in another node (non-bold).

Clade C1: [2:0 [right arrow] 1], [16:0 [right arrow] 1], [17:0 [right arrow] 1], [20:0 [right arrow] [right arrow] 1], [22: [right arrow] 1], [23:0 [right arrow] 1], [40:0 [right arrow] 1], [66:0 [right arrow] 1], [105:0 [right arrow] 1], [110:0 [right arrow] 1], [132:0 [right arrow] 1], [163:0 [right arrow] 1], [167:0 [right arrow] 1], [181:0 [right arrow] 1], [236:0 [right arrow] 1], [244:0 [right arrow] 1], [245:0 [right arrow] 1]

As expected, the teleostean taxa included in the cladistic analysis appear more closely related to each other than to Amia and Lepisosteus. In all the 48 most parsimonious trees obtained, the teleostean taxa examined appear more closely related to Amia than to Lepisosteus (the clade including Amia and these teleostean taxa is supported by a bootstrap value of 82%). However, the relationships between the Teleostei, the Halecomorphi and the Ginglymodi are of course beyond the scope of the present work: such relationships can only be seriously addressed in an analysis including a great number of other neopterygian and non-neopterygian fishes (e.g. Arratia, 2004; Cloutier & Arratia, 2004). Among the unambiguous synapomorphies of the clade including all the teleosts examined in the present work, which is supported by a bootstrap value of 100% (Fig. 1), those concerning characters 2 (posterior intermandibularis integrated in protractor hyoidei, but also deeply associated with anterior intermandibularis), 16 (anteroventromesial portion of hypoaxialis continuous with posteroventromesial portion of sternohyoideus), 17 (sternohyoideus consolidated into a single median muscle), 20 (presence of distinct muscle arrector ventralis), 40 (prevomer unpaired), 66 (ossification of supraoccipital), 105 (mesocoracoid arch ossified), 110 (first pectoral ray articulating directly with scapula and/or possibly coracoid), 163 (premaxillae not markedly ankylosed with neurocranium), 181 (articulation, either direct or indirect, between autopalatine/dermopalatine and maxilla), 236 (presence of ossified urohyal/parurohyal), 244 (coronoid bones absent as independent ossifications) and 245 (prearticulars absent as independent ossifications) have been proposed as potential Teleostei synapomorphies by authors such as Schaefer & Rosen (1961), Lauder (1980), Lauder & Liem (1983), Jollie (1986), De Pinna (1996), Arratia & Schultze (1990) and Arratia (1997, 1999). However, the results of the present work pointed out four potential Teleostei synapomorphies that, at least according to our knowledge, have not been previously proposed: those concerning characters 22 (arrector dorsalis subdivided into two well-developed sections), 23 (arrector dorsalis attaching on both the first and second pectoral rays), 132 (absence of distinct adductor mandibulae A3') and 167 (primordial ligament attaching posteriorly on posterolateral surface of mandible). In addition to the unambiguous synapomorphies listed above uniting the teleostean taxa examined in the present work, there are other features that may represent potential Teleostei synapomorphies. For example, the loss of the muscle branchiomandibularis (14: 0 [right arrow] 1): according to the results of the cladistic analysis, this feature may have occurred in Lepisosteus + Amia + the teleosts examined and then reverted in Amia or may have occurred independently in Lepisosteus and in teleosts. Although the two hypotheses appear as theoretically equally parsimonious, the independent acquisition, in Amia, of a muscle that is strikingly similar to the characteristic muscle branchiomandibularis of other actinopterygians seems rather unsound (see e.g. Lauder, 1980; Wilga et al., 2000). The loss of muscle protractor pectoralis (24: 0 [right arrow] 1) is also a feature that may have occurred in Lepisosteus + Amia + the teleosts examined and then reverted in Amia or that may have occurred independently in Lepisosteus and in teleosts. Other features are e.g. the presence of distinct, strong ligaments connecting the anterior surface/anterior cartilage of autopalatines and/or dermopalatines and maxilla and/or premaxillae (160: 0 [right arrow] 1) and the presence of an ossified interhyal (223: 0 [right arrow] 1). According to the results of the cladistic analysis, these two latter features might represent synapomorphies of the Clupeocephala (see Fig. 1) and of the Elopomorpha, or, instead, might be synapomorphies of the clade including all the teleostean taxa examined, that were subsequently lost in the Osteoglossomorpha (as well as in other, more derived taxa: see below). These two features could, thus, possibly be interpreted as synapomorphies of the Elopomorpha + Clupeocephala, if the Osteoglossomorpha were accepted as the most basal extant teleostean clade examined (see e.g. Fig. 1). However, as can be seen in Figure 2, all (100%) the parsimonious trees obtained in the cladistic analysis of the present work support the Elopomorpha, and not the Osteoglossomorpha, as the most basal teleostean group examined (see 'Clade C7' below).

[FIGURE 3 OMITTED]

[FIGURE 3 OMITTED]

Clade C2: [247:0 [right arrow] 1], [268:0 [right arrow] 1]

As explained in the Introduction, the monophyly of an Elopomorpha clade including elopiforms, albuliforms, notacanthiforms, anguilliforms, and the peculiar saccopharyngiforms has been recently questioned by authors such as Filleul and Filleul & Lavoue. As stressed these authors, no published morphological cladistic analysis has included representatives of all these taxa and supported their grouping in a monophyletic clade. Some recent molecular cladistic analyses supported the inclusion of these taxa in a monophyletic clade (e.g. Wang et al., 2003; Inoue et al., 2004), but others have contradicted this view (e.g. Obermiller & Pfeiler, 2003). The elopiform, albuliform, notacanthiform, saccopharyngiform and anguilliform fishes included in the present cladistic analysis do appear in a monophyletic clade in all the 48 most parsimonious trees obtained (Fig. 2). In this respect, this is thus the first published cladistic morphological analysis supporting the monophyly of these fishes. It should however be stressed that the bootstrap value obtained for this clade C2 is not [greater than or equal to] 50% (Fig. 2), i.e. this clade is not strongly supported by a bootstrap analysis. Interestingly, the monophyly of the Elopiformes is not supported by the tree of Figure 2, although it is also not contradicted by this tree: Elops, Megalops and the remaining elopomorphs are placed in an unresolved trichotomy. A close relationship between Elops and Megalops has been defended in recent molecular studies (e.g. Obermiller & Pfeiler; Wang et al.; Inoue et al., 2004). It has been also defended in the past in studies such as Greenwood et al. (1966), Nelson (1973) and Forey et al. However, in some other studies Megalops was placed as the sister-group of Elops plus the remaining elopomorphs (e.g. Forey, 1973b) or was placed together with Elops and the other elopomorphs in a trichotomy such as that shown in Figure 2 (e.g. Patterson & Rosen, 1977).

[TABLE I OMITTED]

The first synapomorphy listed above concerns the absence of the retroarticular as an independent ossification (char. 247). Within the fishes included in the cladistic analysis, this is a rather rare feature, being only found in elopomorphs, in catfishes and in Mormyrus (Hiodon was coded as '?': see char. 247), although it should be noted that it is also found in some teleostean fishes that were not included in the analysis (see e.g. Nelson, 1973). The second synapomorphy, which is homoplasy free within the fishes examined, concerns the presence of a 'leptocephalus larva' (see char. 268). Apart these two unambiguous synapomorphies, there are some features that exhibit an ambiguous distribution in the tree but that may possibly also represent synapomorphies for this clade C2. One of these features concerns the 'anteroventral margin of prevomer situating well posteriorly to anteroventral margin of mesethmoid' (32: 0 [right arrow] 1). It is found in the specimens examined of the genera Elops, Albula and Notacanthus, and cannot be discerned in the anguilliform (due to a complete fusion between the prevomer and the mesethmoid) and saccopharyngiform (such a fusion might also occur, but this is not clear) fishes analyzed. Thus, this feature might be interpreted either as acquired independently in Elops and in the clade C3 or as acquired in the elopomorph clade and then reverted in Megalops (other hypothesis would be to interpret the acquisition of the feature in the clade uniting the non-Megalops elopomorph fishes examined, if Megalops were to be placed in the most basal position within clade C2). It should however be noted that this feature is rather homoplasious, being found in some other teleostean groups examined (see below). Two other features with ambiguous distributions that may possibly also represent elopomorph synapomorphies are the presence of distinct, strong ligaments connecting the anterior surface/anterior cartilage of the autopalatines and/or dermopalatines and the maxillae and/ or premaxillae (160: 0 [right arrow] 1) and the presence of an ossified interhyal (223: 0 [right arrow] 1) (see 'Clade C1' above).

[FIGURE 5 OMITTED]

[FIGURE 6 OMITTED]

Elops: [102:0 [right arrow] 1], [131:0 [right arrow] 1], [206:0 [right arrow] 1]; Megalops: [101:0 [right arrow] 1]

Clade C3: [34:0 [right arrow] 1]

As expected (see Fig. 1) the albuliform, notacanthiform, anguilliform and saccopharyngiform fishes examined are grouped together. Apart the synapomorphy listed above, there are various other features with an ambiguous distribution that may possible represent synapomorphies of this clade, such as those concerning character 32 (0 [right arrow] 1) (if a 'slow' optimization is chosen) and concerning characters 1 (1 [right arrow] 0), 111 (0 [right arrow] 1), 115 (0 [right arrow] 1), 155 (0 [right arrow] 1), 190 (0 [right arrow] 1) and 232 (1 [right arrow] 0) (if a 'fast' optimization is chosen).

Albula: [2:1 [right arrow] 0], [49:0 [right arrow] 1], [166:0 [right arrow] 1], [195:0 [right arrow] 1], [228:0 [right arrow] 1]

Clade C4: [2:1 [right arrow] 2], [44:0 [right arrow] 1], [61:0 [right arrow] 1], [64:0 [right arrow] 1], [85:0 [right arrow] 1], [104:0 [right arrow] 1], [133:0 [right arrow] 1], [158:0 [right arrow] 1], [193:0 [right arrow] 1]

The grouping of notacanthiform, anguilliform and saccopharyngiform fishes is expected (see Fig. 1). The posterior intermandibularis forming the protractor hyoidei and not being deeply mixed with the anterior intermandibularis (char. 2) and the absence of adductor mandibulae Aw (char. 133) have not been previously proposed in the literature as synapomorphies of this group.

Notacanthus: [42:0 [right arrow] 1], [102:0 [right arrow] 1], [128:0 [right arrow] 1], [141:0 [right arrow] 1], [215:0 [right arrow] 1]

Clade C5: [75:0 [right arrow] 1], [109:0 [right arrow] 1], [157:0 [right arrow] 1], [200:0 [right arrow] 1], [201:0 [right arrow] 1], [239:0 [right arrow] 1], [246:0 [right arrow] 1]

The grouping of anguilliform and saccopharyngiform fishes is expected (see Fig. 1) and is well-corroborated, being supported by a bootstrap value of 95%.

Eurypharynx: [3:0 [right arrow] 1], [7:0 [right arrow] 1], [9:0 [right arrow] 1], [15:0 [right arrow] 1], [35:0 [right arrow] 1], [48:0 [right arrow] 1], [66:1 [right arrow] 0], [68:1 [right arrow] 0], [86:0 [right arrow] 1], [91:0 [right arrow] 1], [97:0 [right arrow] 1], [118:0 [right arrow] 1], [126:0 [right arrow] [right arrow] 1], [127:0 [right arrow] 1], [148:1 [right arrow] 0], [208:0 [right arrow] 1], [210:0 [right arrow] 1], [213:0 [right arrow] [right arrow] 1], [235:0 [right arrow] 1], [236:1 [right arrow] 0]

Clade C6: [125:0 [right arrow] 1], [206:0 [right arrow] 1]

The order Anguilliformes is usually considered a monophyletic group, and the well-supported grouping of the anguilliform genera Conger and Anguilla (with a bootstrap value of 91%) is thus expected (see e.g. Greenwood et al., 1966; Forey, 1973b; Nelson, 1973; Patterson & Rosen; Obermiller & Pfeiler). But authors such as Forey et al., Belouze, Wang et al. and Inoue et al. (2004) have defended that some anguilliforms (e.g. congroids or, alternatively, anguilloids) are more closely related to saccopharyngiforms than to other anguilliforms, a view that is not supported by the present work. It is however obvious that only a study including numerous anguilliform and saccopharyngiform taxa, and, it is important to stress this, also numerous other elopomorph and non-elopomorph fishes, can help to address this question in a more conclusive way.

Anguilla: [228:0 [right arrow] 1]; Conger: [133:1 [right arrow] 0]

Clade C7: [75:0 [right arrow] 1], [102:0 [right arrow] 1], [109:0 [right arrow] 1], [131:0 [right arrow] 1], [158:0 [right arrow] 1], [241:0 [right arrow] 1]

As referred above (see 'Clade C1') all (100%) the most parsimonious trees obtained place the Elopomorpha, and not Osteoglossomorpha, as the most basal teleostean group examined (Fig. 2). However, it is important to stress that this clade C7, including the osteoglossomorph and the remaining non-elopomorph teleosts examined, is not supported by a bootstrap value [greater than or equal to] 50% (Fig. 2). Six unambiguous synapomorphies support this clade in the analysis: fusion of at least some parapophyses of the two first free vertebrae to the respective centra (75: 0 [right arrow] 1, subsequently reverted in some taxa of this clade and independently occurring in the elopomorph anguilliforms + saccopharyngiforms); mesial limb of coracoids (or scapulocoracoids) broad and anteroposteriorly elongated (101: 0 [right arrow] 1, subsequently reverted in some taxa of this clade and independently occurring in the elopomorph Megalops); absence of pectoral splints (109: 0 [right arrow] 1, not reverted in any taxa of this clade and independently occurring in the elopomorph anguilliforms + saccopharyngiforms); absence of well-differentiated, separated section A3' of adductor mandibulae (131: 0 [right arrow] 1, subsequently reverted in some taxa of this clade and independently occurring in the elopomorph Elops; the condition in Eurypharynx is not clear); supramaxillae absent as independent ossifications (158: 0 [right arrow] 1, subsequently reverted in some taxa of this clade and independently occurring in the elopomorph notacanthiforms + anguilliforms + saccopharyngiforms); absence of ossified gular plate (241: 0 [right arrow] 1, not reverted in the taxa examined belonging to this clade but independently occurring in the elopomorph notacanthiforms + anguilliforms + saccopharyngiforms; the condition in Albula is not clear). It should however be noted that some of these features characterizing the clade including the osteoglossomorph and the remaining non-elopomorph teleosts examined might actually not diagnose a clade including all non-elopomorph teleosts known. For example, supramaxillae and gular plates have been described in some fossil osteoglossomorphs (e.g. Taverne, 1972, 1977b, 1978; pers. comm.). If these structures were in fact plesiomorphically present in osteoglossomorphs, their absence would thus not constitute a valid feature to diagnose a clade including all osteoglossomorph + remaining non-elopomorph teleosts.

[FIGURE 7 OMITTED]

As referred in the Introduction, it is usually accepted that osteoglossomorphs occupy a more basal position within Teleostei than elopomorphs (e.g. Patterson, 1977; Patterson & Rosen; Lauder & Liem; Ishiguro et al.; Obermiller & Pleifer; Inoue et al., 2003, 2004; Wang et al.) (see Fig. 1). However, a more basal position of elopomorphs, such as that suggested in the present work, has also been suggested by various authors, e.g. Greenwood et al., Li (1996), Shen (1996), Arratia (1997, 1999).

[FIGURE 8 OMITTED]

[FIGURE 9 OMITTED]

Clade C8: [96:0 [right arrow] 1], [180:0 [right arrow] 1], [261:0 [right arrow] [right arrow] 1]

As stated in the recent paper of Lavoue & Sullivan (2004), although the Osteoglossomorpha is widely accepted as a monophyletic unit, the only cladistic analyses that have tested the monophyly of this group by including representatives of its four families sensu Hilton and an appropriate sample of other teleostean taxa are essentially molecular ones. In the present cladistic analysis, the representatives of these four osteoglossomorph families (Hiodon: Hiodontidae; Pantodon: Osteoglossidae; Xenomystus: Notopteridae; Mormyrus: Mormyridae) do appear grouped in a monophyletic clade (C8), which is supported by the three unambiguous synapomorphies listed above but is not supported by a bootstrap value [right arrow] 50% (Fig. 2). The first synapomorphy concerns the peculiar anteroventrolateral bifurcation of the posttemporal in a shorter, lateral arm carrying a sensorial canal and a longer, mesial arm corresponding to the ossified 'ligament between the posttemporal and the posterior margin of the neurocranium' of the present work (96: 0 [right arrow] 1). This feature is exclusively found in the osteoglossomorphs examined except Pantodon, in which the latter ligament is not ossified. The second synapomorphy concerns the poorly ossified, or completely unossified, autopalatine (180:0 [right arrow] 1), a feature that is only also found in a few fishes examined in the present work, such as gymnotiforms. The third synapomorphy concerns the presence of a 'tongue-bite mechanism' with dorsal teeth on parasphenoid, a feature that is homoplasy free within the fishes examined (261: 0 [right arrow] 1). Contrary to the first feature, the latter two have been listed as potential osteoglossomorph synapomorphies in previous works such as Lauder & Liem, Li & Wilson (1996), Arratia (1997, 1999) and Hilton. As mentioned above, there are two features with ambiguous distributions that might possibly constitute synapomorphies of the osteoglossomorphs examined: the absence of distinct, strong ligaments connecting the anterior surface/anterior cartilage of the autopalatines and/or dermopalatines to the maxillae and/or premaxillae (160: 1 [right arrow] 0) and the absence of an ossified interhyal (223: 1 [right arrow] 0) (see 'Clade C1').

[FIGURE 10 OMITTED]

Hiodon: [42:0 [right arrow] 1], [140:0 [right arrow] 1], [143:0 [right arrow] 1], [206:0 [right arrow] 1]

Clade C9: No unambiguous synapomorphies

Although a detailed discussion of the relationships between the four osteoglossomorph families (see above) is beyond the main scope of the present work, it is worthy to note that in the majority (66%) of the most parsimonious trees obtained Hiodon appears as the sister-group of the clade including the other osteoglossomorphs examined (Fig. 2), as expected (see Fig. 1). However, no features can be unambiguously interpreted as synapomorphies of this latter clade in the 'majority fools' tree shown in Fig. 2, and this clade is not supported by a bootstrap value [greater than or equal to] 50%. In the majority of the most parsimonious trees in which the clade appears, it is however diagnosed by an unambiguous synapomorphy (124: 1 [right arrow] 0), and can be possibly diagnosed by other features if a 'fast optimization' is chosen (e.g. 21: 0 [right arrow] 1; 34: 0 [right arrow] 1; 68: 0 [right arrow] 1; 208: 1 [right arrow] 0) or, alternatively, if a 'slow optimization' is chosen (e.g. 246: 0 [right arrow] 1). In the most parsimonious trees in which the clade does not appear, Hiodon is grouped with Xenomystus, with a single feature supporting this grouping (96: 0 [right arrow] 1), and this only if a 'slow optimization' is chosen. Some authors have suggested that some notopterids (the group in which Xenomystus is nowadays included) might be more related to Hiodon than to some other extant osteoglossomorphs (e.g. Greenwood et al.; Nelson, 1968; Greenwood, 1973; Lauder & Liem). It can thus be said that this latter view is not completely contradicted by the results of the present work, although it is important to remind that, in the overall, the majority of the most parsimonious threes obtained in this work do support the grouping of the non-hiodontid osteoglossomorph fishes examined (Fig. 2).

Xenomystus: [16:1 [right arrow] 0], [27:0 [right arrow] 1], [166:0 [right arrow] 1], [179:0 [right arrow] 1], [228:0 [right arrow] 1]

Clade C10: [45:0 [right arrow] 1], [65:1 [right arrow] 0], [99:0 [right arrow] 1], [111:0 [right arrow] 1], [133:0 [right arrow] 1], [162:0 [right arrow] 1], [201:0 [right arrow] 1]

Pantodon: [2:1 [right arrow] 2], [95:0 [right arrow] 1], [96:1 [right arrow] 0], [107:0 [right arrow] 1], [163:1 [right arrow] 0], [223:0 [right arrow] 1]; Mormyrus: [2:1 [right arrow] 0], [10:0 [right arrow] 1], [32:0 [right arrow] 1], [42:0 [right arrow] 1],[61:0 [right arrow] 1], [119:0 [right arrow] 1], [143:0 [right arrow] 1], [155:0 [right arrow] 1], [178:0 [right arrow] 1], [190:0 [right arrow] 1], [193:0 [right arrow] 1], [233:0 [right arrow] [right arrow] 1], [247:0 [right arrow] 1]

Clade C11: [2:1 [right arrow] 2], [18:0 [right arrow] 1], [60:0 [right arrow] 1], [64:0 [right arrow] 1]

[FIGURE 11 OMITTED]

The assembly of the non-elopomorph and non-osteoglossomorph teleosts examined in clade C11 is expected (see Fig. 1: Clupeocephala). Although this clade is not supported by a bootstrap value [greater than or equal to] 50% (Fig. 2), it is supported by four unambiguous synapomorphies, one of them being actually homoplasy free within the fishes examined: posterior intermandibularis included in protractor hyoidei and not deeply mixed with anterior intermandibularis (2: 1 [right arrow] 2, not reverted in fishes of this clade C11 but independently occurring in Pantodon and in Nothacanthiformes + Saccopharyngiformes + Anguilliformes); presence of distinct muscle 'arrector 3' (18: 0 [right arrow] 1, homoplasy free within the fishes examined); main bodies of parietals (or of parietoextrascapulars) widely separated from each other in dorsal view (60: 0 [right arrow] 1, not independently acquired within the taxa examined of other clades, but subsequently reverted in some more terminal groups of this clade C11: see below); absence of parasphenoid teeth (64: 0 [right arrow] 1, not subsequently reverted within the fishes examined of this clade C11, and only occurring independently, within the taxa analyzed, in Nothacanthiformes + Saccopharyngiformes + Anguilliformes). Contrary to the two latter features, the two former ones have not been previously proposed as clupeocephalan synapomorphies. Some other features with an ambiguous distribution may be interpreted as synapomorphies of this clade if a 'fast optimization' is chosen (155: 0 [right arrow] 1; 193:0 [right arrow] 1) or, alternatively, if a 'slow optimization' is chosen (160:0 [right arrow] 1; 223:0 [right arrow] 1; 246:0 [right arrow] 1; 248:0 [right arrow] 1) (see e.g. 'Clade C1' above).

Clade C12: [35:0 [right arrow] 1], [87:0 [right arrow] 1], [228:0 [right arrow] 1]

The grouping of the euteleostean fishes analyzed in clade C12 is expected (see Fig. 1). It is not supported by a bootstrap value [greater than or equal to] 50%, but is supported by three unambiguous synapomorphies in the cladistic analysis: 'ethmoid endoskeleton' ossification markedly reduced (35: 0 [right arrow] 1, only occurring independently in a few taxa outside this clade C12 and only being reverted in a few taxa examined of this clade as e.g. aulopiforms); main body of posttemporal (or posttemporo-supracleithrum) lying far from neurocranium, with almost no contact between these two structures (87: 0 [right arrow] 1, not reverted within the fishes examined of this clade C12, and only occurring independently in a few taxa outside the clade, e.g. Gonorynchus); mandibulohyoid and mandibulo-interopercular ligaments not wellseparated from each other (87: 0 [right arrow] 1, not reverted within fishes examined of this clade C12, and only occurring independently in a few taxa outside the clade as e.g. Anguilla, Albula, Denticeps and Xenomystus). Contrary to the first feature, the latter two have not been previously proposed as synapomorphies of the Euteleostei. There is one feature with an ambiguous distribution that may be interpreted as synapomorphy of this clade, if a 'slow optimization' is chosen (193: 0 [right arrow] 1).

Clade C13: [34:0 [right arrow] 1], [47:0 [right arrow] 1], [153:0 [right arrow] 1], [167:1 [right arrow] 0], [188:0 [right arrow] 1], [264:0 [right arrow] 1]

The grouping of the argentinoid and alepocephaloid fishes examined is strongly supported by six unambiguous synapomorphies (two of them being homoplasy-free within the fishes examined), and by a bootstrap value of 74% (Fig. 2): posterodorsal portion of mesethmoid (or of supraethmoid) being markedly compressed transversally when seen in dorsal view (34: 0 [right arrow] 1, not reverted within taxa examined of this clade C13, but occurring independently in some groups outside the clade, e.g. characiforms, gymnotiforms and siluriforms); both autopterotic and dermopterotic bones present as independent, distinct ossifications (47: 0 [right arrow] 1, homoplasy free within the teleostean taxa examined); fibers of hypaxialis and/or epaxialis peculiarly covering great part of neurocranial floor (153: 0 [right arrow] 1, not reverted inside this clade C13 and only occurring independently in the aulopiforms + stomiiforms examined); primordial ligament attaching posteriorly on dorsal surface of coronoid process (167: 0 [right arrow] 1, occurring in some groups outside this clade C13 but not reverted inside it; Bathylagus, Xenodermichthys and Searsia were coded as 'Inapplicable' for this character as they seemingly do not have a distinct primordial ligament); peculiar dorsoventral enlargement of posterior portion of autopalatine (188: 0 [right arrow] 1, not reverted inside this clade C13 and only occurring outside of it in the osmeroid specimens examined); presence of peculiar 'accessory cartilage of the fifth ceratobranchial' (264: 0 [right arrow] 1, homoplasy free within the taxa examined). Contrary to the other features listed above, the first and forth features have not been previously proposed as synapomorphies of the Argentiniformes. Various features with an ambiguous distribution may be interpreted as synapomorphies of this clade if a 'fast optimization' is chosen (16: 1 [right arrow] 0; 102: 0 [right arrow] 1; 111: 0 [right arrow] 1; 124: 1 [right arrow] 0; 142: 0 [right arrow] 1; 215: 0 [right arrow] 1).

[FIGURE 12 OMITTED]

It should be noted that, contrary to what is usually accepted (e.g. Greenwood & Rosen, 1971; Rosen, 1973, 1974, 1985; Fink & Weitzman, 1982; Lauder & Liem; Fink, 1984; Begle, 1991, 1992; Nelson, 2004; Johnson & Patterson, 1996; Sanford, 2000; Stiassny et al.), some recent molecular cladistic analyses (e.g. Ishiguro et al.; Lavoue et al., 2005) have contradicted the monophyly of Argentiniformes (sensu this work: see Figs. 1, 2). According to these molecular analyses, the Alepocephaloidea is the sister-groups of either the Clupeomorpha or the Ostariophysi, but not of the Argentinoidea. This particular aspect makes us very reticent about the conclusions of these molecular analyses. The Alepocephaloidea + Argentinoidea clade appears as the most basal euteleostean group in the present cladistic analysis, and, in this sense, to learn that this clade was placed closer to certain Otocephala taxa than to other euteleostean groups would perhaps not seem too unsound. But to learn that the Alepocephaloidea are placed inside the Otocephala and the Argentinoidea not, this does seem rather unsound in face of the large amount of data (obtained by various authors and concerning various types of morphological characters) supporting the monophyly of the Argentiniformes (e.g. Greenwood & Rosen; Rosen, 1973, 1974; Begle, 1991, 1992; Johnson & Patterson; Sanford, 2000; this study).

[FIGURE 13 OMITTED]

Clade C14: [140:0 [right arrow] 1], [166:0 [right arrow] 1], [168:0 [right arrow] 1], [174:0 [right arrow] 1], [226:0 [right arrow] 1]

The grouping of the argentinoid fishes examined is expected (see Fig. 1) and is supported by a bootstrap value of 83%.

Argentina: [35:1 [right arrow] 0], [60:1 [right arrow] 0], [101:1 [right arrow] 0], [125:0 [right arrow] 1], [242:0 [right arrow] 1]; [259:0 [right arrow] 1]; Bathylagus: [25:0 [right arrow] 1], [48:0 [right arrow] 1], [104:0 [right arrow] 1], [133:0 [right arrow] 1]

Clade C15: [22:1 [right arrow] 0], [23:1 [right arrow] 0], [33:0 [right arrow] 1], [158:1 [right arrow] 0]

The grouping of the alepocephaloid fishes examined is expected (see Fig. 1) and is supported by a bootstrap value of 88%.

Alepocephalus: No unambiguous features

Clade C16: [68:0 [right arrow] 1], [105:1 [right arrow] 0], [178:0 [right arrow] 1]

Xenodermichthys: [101:1 [right arrow] 0], [222:0 [right arrow] 1]; Searsia: [65:1 [right arrow] 0], [170:0 [right arrow] 1], [193:1 [right arrow] 0], [260:0 [right arrow] 1]

Clade C17: No unambiguous synapomorphies

The grouping of the non-argentiniform euteleosteans examined in a same clade is supported in two thirds of the 48 most parsimonious trees obtained in the cladistic analysis, but is not supported by a bootstrap value [greater than or equal to] 50% (Fig. 2). In half of the most parsimonious trees in which this clade appears, the clade is diagnosed by an unambiguous synapomorphy: the attachment of a mainly undivided A2 on the mesial surface of the mandible being accomplished by means of two well-distinguished, thick tendons (124: 0 [right arrow] 1). Two other features may be interpreted as synapomorphies of this clade if a 'fast optimization' is chosen: rostrodermethmoids ossified and not fused with median supraethmoid (29: 0 [right arrow] 1) and orbitosphenoid not present as independent ossification (44: 0 [right arrow] 1). In the most parsimonious trees in which this clade does not appear, the argentiniforms appear as the sister-group of a clade including the salmoniform + neoteleostean fishes examined. In those trees no unambiguous synapomorphies support such a sister-group relationship: only the choosing of a 'fast optimization' (33: 0 [right arrow] 1; 158: 1 [right arrow] 0; 168: 0 [right arrow] 1) or of a 'slow optimization' (193: 0 [right arrow] 1) can reveal, in these trees, potential synapomorphies to support such a relationship. It can thus be said that this latter view is not completely contradicted by the results of the present work, although it is important to remind that, in the overall, the majority of the most parsimonious threes obtained in this work do support the grouping of the non-argentiniform euteleosteans examined (Fig. 2). However, the evidence supporting this clade C17 is rather week.

[FIGURE 14 OMITTED]

The relationships between basal euteleosteans have been widely discussed, and, as seen above, remain problematic (e.g. Greenwood et al.; Gosline, 1969; Rosen, 1973, 1974, 1985; Fink & Weitzman; Lauder & Liem; Fink, 1984; Begle, 1991, 1992; Johnson & Patterson; Sanford, 2000). For instance, authors such as Ishiguro et al. consider that the osmeriforms are closely related to the Argentinoidea, and that the esociforms are closely related to the salmoniforms. Authors such as Fink & Weitzman, Lauder & Liem and Begle (1991, 1992) consider that the osmeriforms are closely related to the argentiniforms, and that the esociforms are not closely related to the salmoniforms. Authors such as Johnson & Patterson, in turn, defend a close relationship between osmeriforms and salmoniforms, and between the clade formed by these two latter groups and the argentiniforms. A brief, up-to-dated summary of these and other hypotheses concerning basal euteleostean relationships has been provided by Ishiguro et al.. In short, it can be said that, together with the phylogenetic hypothesis shown in Fig. 2 of the present work, almost all possible combinations between the major basal euteleostean groups have already been proposed in the literature. The present work provides strong support for the monophyly of the Alepocephaloidea, of the Argentinoidea and of the Alepocephaloidea + Argentinoidea (see above), and some support for the monophyly of the Galaxioidea + Osmeroidea and of the Esociformes (see below), but does not provide strong evidence to resolve the relationships between the Argentiniformes, the Salmoniformes, the Neoteleostei, the Esociformes and the Osmeriformes (see below).

Clade C18: [33:0 [right arrow] 1], [75:1 [right arrow] 0], [158:1 [right arrow] 0], [161:0 [right arrow] 1], [168:1 [right arrow] 0], [185:0 [right arrow] 1]

The assembly of the salmoniform and neoteleostean fishes examined is this clade is supported by six unambiguous synapomorphies, but is not supported by a bootstrap value [greater than or equal to] 50%. These synapomorphies are: presence of anterolateral processes of mesethmoid supporting and/or articulating with premaxillae (33: 0 [right arrow] 1, occurring independently outside this clade C18 in a few groups, and reverted, inside of this clade, in the aulopiforms; the members of the genus Coregonus might display either CS0 or CS1 of this character); parapophyses of two first free vertebrae not fused to centra (75: 0 [right arrow] 1, occurring in a few groups outside of this clade C18, and reverted, inside this clade, in Aulopus and Astronesthes); supramaxillae present as independent ossifications (158: 1 [right arrow] 0, occurring in various groups outside of this clade C18, but not reverted inside of it, although it should be noted that Astronesthes and Stomias were coded as '?' for this character); presence of well-developed 'rostral' cartilaginous or cartilaginous-like structures associated with the posterior surface of well-developed premaxillary dorsomedial processes attached to/articulating with the ethmoid region (161: 0 [right arrow] 1, within the fishes examined by us, found exclusively in this clade C18, with the single exception of Osmerus); presence of strong, well-defined ligament between premaxilla and proximal surface of maxilla (168: 0 [right arrow] 1, occurring in some groups outside of this clade C18, but not reverted inside of it); anterior portion and/or anterior cartilage of autopalatine forming peculiar 'broad hook' covering a great portion of proximal portion of maxilla in lateral view (185: 0 [right arrow] 1, found exclusively in the taxa of this clade C18, and only reverted, within the fishes examined, in Coregonus and Stomias). There is a feature with an ambiguous distribution that may be interpreted as a synapomorphy of this clade, if a 'fast optimization' is chosen (155: 1 [right arrow] 0). It should be noted that, although nowadays many authors consider the Esociformes as the probable sister-group of the Neoteleostei (see Fig. 1), some studies have defended, in the past, that salmoniforms are closely related to neoteleosteans (e.g. Lauder & Liem; Fink, 1984).

[FIGURE 15 OMITTED]

Salmo: No unambiguous features; Coregonus: [29:0 [right arrow] 1], [42:0 [right arrow] 1], [178:0 [right arrow] 1], [185:1 [right arrow] 0]; Thymallus: No unambiguous features

Clade C19: [68:0 [right arrow] 1], [84:0 [right arrow] 1], [111:0 [right arrow] 1], [153:0 [right arrow] 1], [262:0 [right arrow] 1]

The grouping of the neoteleostean fishes studied in this clade is expected (see Fig. 1). It is supported by a bootstrap value of 62% (Fig. 2) and by five synapomorphies: anteroposterior elongation of anterior neural arches (68: 0 [right arrow] 1, not reverted in the fishes we have examined from this clade C19, but occurring in some taxa outside of it); peculiarly large, distinct 'precervical gap' filled mainly with connective tissue between first free vertebra and neurocranium (84: 0 [right arrow] 1, homoplasy free within the fishes examined); adductor mandibulae attaching not only on mandible and/or primordial ligament, near its mandibular insertion, but also on other structures (111: 0 [right arrow] 1, not reverted inside this clade C19, but occurring in some taxa outside of it); fibers of hypaxialis and/or epaxialis peculiarly covering great part of neurocranial floor (153: 0 [right arrow] 1, not reverted inside this clade C19 but also occurring in the argentiniform fishes analyzed); presence of peculiar muscle retractor dorsalis (262: 0 [right arrow] 1, homoplasy free within the fishes examined). There are some features with an ambiguous distribution that may be interpreted as synapomorphies of this clade C19 if a 'fast optimization' is chosen (22: 1 [right arrow] 2; 60: 1 [right arrow] 0; 104: 0 [right arrow] 1; 140: 0 [right arrow] 1; 166: 0 [right arrow] 1). It should be noted that, although some features listed above might effectively reveal to be potential synapomorphies of the Neoteleostei, this can evidently only be examined appropriately in a study including numerous other representative neoteleostean taxa.

[FIGURE 16 OMITTED]

Clade C20: [19:0 [right arrow] 1], [24:1 [right arrow] 0], [33:1 [right arrow] 0], [34:0 [right arrow] 1], [35:1 [right arrow] 0], [74:0 [right arrow] 1], [101:1 [right arrow] 0], [114:0 [right arrow] 1], [134:0 [right arrow] 1]

The grouping of the aulopiform fishes examined in this clade is expected (see Fig. 1). It is supported by the nine features listed above and by a bootstrap value of 99% (Fig. 2). A detailed discussion of the synapomorphies of the Eurypterygii (Aulopiformes + Ctenosquamata), is clearly beyond the main scope of this work, as it includes a single eurypterygian group (see Fig. 1). In fact, it should be stressed that some of the nine features listed above may well be synapomorphies of the Eurypterygii as a whole, and not of the Aulopiformes. For instance, the consistent presence of the coracoradialis (19: 0 [right arrow] 1), of the protractor pectoralis (24: 1 [right arrow] 0) and of the adductor mandibulae A1 (114: 0 [right arrow] 1) have been proposed by some authors as synapomorphies of eurypterygian fishes (e.g. Winterbottom, 1974; Greenwood & Lauder, 1981; Lauder & Liem; Gosline, 1986; Wu & Shen, 2004). Another example concerns the attachment of the Aw on the suspensorium and/or opercular series (134: 0 [right arrow] 1). As explained in the description of character 134, this feature is found in many non-aulopiform eurypterygian fishes and may well constitute a potential eurypterygian synapomorphy. The taxonomic distribution of these features can only be appropriately examined in a study including numerous other representative eurypterygian taxa.

[FIGURE 17 OMITTED]

[FIGURE 18 OMITTED]

[FIGURE 19 OMITTED]

[FIGURE 20 OMITTED]

Aulopus: [75:0 [right arrow] 1], [193:1 [right arrow] 0], [206:0 [right arrow] 1]; Chlorophthalmus: [143:0 [right arrow] 1]

Clade C21: [94:0 [right arrow] 1], [115:0 [right arrow] 1], [238:0 [right arrow] 1], [240:0 [right arrow] 1], [270:1 [right arrow] 0]

The grouping of the stomiiform fishes examined is expected (see Fig. 1). It is supported by the five features listed above and by a bootstrap value of 94% (Fig. 2).

Astronesthes: [75:0 [right arrow] 1]; Stomias: [85:0 [right arrow] 1], [185:1 [right arrow] 0], [224:0 [right arrow] 1]

Clade C22: [104:0 [right arrow] 1], [196:0 [right arrow] 1]

The assembly of the esociform and osmeriform fishes examined by us in this clade C22 is supported by two synapomorphies and is not supported by a bootstrap value [greater than or equal to] 50% (Fig. 2). These synapomorphies are: mesocoracoid arch absent (104: 0 [right arrow] 1, occurring in some groups outside of this clade C22 and reverted inside of it in Osmerus + Plecoglossus); presence of peculiar, prominent hyomandibular lateral spur at or below the level of the opercular process (196: 0 [right arrow] 1, within the fishes examined, occurring exclusively in the taxa of this clade C22 and only reverted in Stokellia + Retropinna). Some features with ambiguous distributions may be interpreted as synapomorphies of this clade if a 'fast optimization' is chosen (45: 0 [right arrow] 1) or, alternatively, if a 'slow optimization' is adopted (44: 0 [right arrow] 1; 155: 0 [right arrow] 1). It is important to note that although many authors (e.g. Rosen, 1973, 1974; Johnson & Patterson; Springer & Johnson) consider the Salmoniformes as the probable sister-group of the Osmeriformes (see Fig. 1), some studies have partially supported a closer relationship between esociforms and osmeriforms than between these latter fishes and the salmoniforms (see e.g. Waters et al., 2000: Figs. 4, 5).

Clade C23: [29:0 [right arrow] 1], [74:0 [right arrow] 1], [101:1 [right arrow] 0], [131:1 [right arrow] 0]

The grouping of the esociform fishes examined is expected (see Fig. 1). It is supported by the four synapomorphies listed above and by a bootstrap value of 68% (Fig. 2).

Esox: [68:0 [right arrow] 1], [158:1 [right arrow] 0], [190:0 [right arrow] 1]; Umbra: [33:0 [right arrow] 1], [140:0 [right arrow] 1], [167:1 [right arrow] 0]

Clade C24: [48:0 [right arrow] 1], [60:1 [right arrow] 0], [166:0 [right arrow] 1], [193:1 [right arrow] 0]

This clade is expected (see Fig. 1). It is supported by the four synapomorphies listed above, and by a bootstrap value of 51% (Fig. 2). Interestingly, the results of the cladistic analysis do not support, but also do not contradict, the monophyly of the Galaxioidea: the galaxiid Galaxias appears in a trichotomy that also includes the galaxioid retropinnids Retropinna + Stokellia and the osmeroids Osmerus + Plecoglossus. This is very likely the result of not having included more osmeriform representatives in the present study (see e.g. the strong evidence presented by Johnson & Patterson, to support the monophyly of the Galaxioidea). However, it should also be noted that in the recent molecular analysis of Lopez et al. (2004) some galaxioid fishes do also appear more closely related to certain osmeroids than to other galaxioids (see e.g. their Fig. 2).

Galaxias: [42:0 [right arrow] 1], [144:0 [right arrow] 1], [145:0 [right arrow] 1], [178:0 [right arrow] 1]

Clade C25: [74:0 [right arrow] 1], [93:0 [right arrow] 1], [179:0 [right arrow] 1], [196:1 [right arrow] 0]

The grouping of the retropinnid galaxioids examined is expected (see e.g. Patterson & Johnson, 1996: Fig. 19). It is supported by the four synapomorphies listed above and by a bootstrap value of 94%.

Retropinna: [155:1 [right arrow] 0]; Stokellia: [16:1 [right arrow] 0]

Clade C26: [104:1 [right arrow] 0], [188:0 [right arrow] 1], [216:0 [right arrow] 1]

The grouping of the osmeroid fishes examined is expected (see Fig. 1).

Osmerus: [60:0 [right arrow] 1], [155:1 [right arrow] 0], [158:1 [right arrow] 0], [161:0 [right arrow] 1]; Plecoglossus: [42:0 [right arrow] 1], [133:0 [right arrow] 1], [171:0 [right arrow] 1], [172:0 [right arrow] 1], [173:0 [right arrow] 1],[178:0 [right arrow] 1], [242:0 [right arrow] 1], [251:0 [right arrow] 1]

Clade C27: [53:0 [right arrow] 1], [81:0 [right arrow] 1], [82:0 [right arrow] [right arrow] 1], [206:0 [right arrow] 1]

The clade including the clupeomorph and ostariophysan fishes examined is supported by a bootstrap value of 50% and by four unambiguous synapomorphies, of which 3 are seemingly homoplasy free within the fishes examined : position of sacculi and lagenae more posterior and principally nearer to midline (53: 0 [right arrow] 1, homoplasy free within the fishes examined in which this character could be discerned); swimbladder with a silvery peritoneal tunic covering at least part of its anterior portion (81: 0 [right arrow] 1, homoplasy free within the fishes examined in which this character could be discerned); swimbladder markedly divided into peculiar anterior and posterior chambers (82: 0 [right arrow] 1, homoplasy free within the fishes examined in which this character could be discerned); hyomandibula exhibiting two articulatory heads for neurocranium (206: 0 [right arrow] 1, independently occurring in some fishes outside this clade C27 and reverted in some taxa of this clade: see below). The three first features were considered by authors such as Rosen & Greenwood (1970) to be potential synapomorphies of the Ostariophysi. However, Grande & De Pinna (2004) have recently defended that these features are also found in many clupeomorphs and that they may well constitute, in fact, otocephalan synapomorphies. This latter view is supported by the present work. There are three features with an ambiguous distribution that may be interpreted as potential synapomorphies of this clade C28 if a 'fast optimization' is chosen (42: 0 [right arrow] 1; 174: 0 [right arrow] 1; 242: 0 [right arrow] 1).

[FIGURE 21 OMITTED]

Although the otocephalan clade is nowadays accepted by most researchers (e.g. Lecointre, 1995; Johnson & Patterson; Arratia, 1997, 1999; Filleul & Lavoue; Inoue et al.; Elmerot et al., 2002; Wang et al.; Zaragueta-Bagils et al., 2002; Stiassny et al., 2004) (see Fig. 1), some authors, based on molecular cladistic analyses, have argued that the Otocephala should be enlarged in order to also include the Alepocephaloidea (Ishiguro et al.; e.g. Lavoue et al., 2005). This subject was already discussed above (see 'Clade C13'). These latter authors argue that the results of their analyses do not directly contradict the results of most morphological cladistic analyses, as these morphological analyses almost never included, in a same matrix, representatives of the Clupeomorpha, of the Ostariophysi, of the Alepocephaloidea, and of other teleostean taxa to which these three groups should be compared. We fully agree with this point. However, it should be noted that, with the present work, there are already two extensive morphological cladistic analyses that did include these three groups and many other teleost taxa in a same matrix and that did contradict the inclusion of the Alepocephaloidea in the otocephalan clade (Patterson & Johnson; this work).

[FIGURE 22 OMITTED]

Clade C28: [34:0 [right arrow] 1], [52:0 [right arrow] 1], [56:0 [right arrow] 1], [271:0 [right arrow] 1]

The grouping of the clupeomorph fishes examined is expected (see Fig. 1). It is supported by a bootstrap value of 67% and by the four synapomorphies listed above, three of which are homoplasy free within the fishes examined.

Denticeps: [88:0 [right arrow] 1], [140:0 [right arrow] 1], [146:0 [right arrow] 1], [160:1 [right arrow] 0], [217:0 [right arrow] 1], [228:0 [right arrow] 1]

Clade C29: [142:0 [right arrow] 1], [158:1 [right arrow] 0], [166:0 [right arrow] 1]

[FIGURE 23 OMITTED]

As expected, Denticeps appears as the sister-group of the remaining clupeiforms examined (see Fig. 1).

Ethmalosa: [24:1 [right arrow] 0], [33:0 [right arrow] 1], [63:0 [right arrow] 1], [144:0 [right arrow] 1]

Clade C30: [77:0 [right arrow] 1]

As seen in Figure 1, many authors consider the relationships between the Clupeoidea, the Engrauloidea and the Pristigasteroidea as still unresolved. This subject was recently revised by Di Dario (2002), who defended a sister-group relationship between the Clupeoidea and the Engrauloidea. The present work does not support this view, since the engrauloid and pristigasteroid fishes examined are grouped together in this clade C30. Apart the synapomorphy listed above (70: 0 [right arrow] 1), the sister-group relationship between these fishes may be supported by other four synapomorphies, if a 'fast optimization' is chosen (68: 0 [right arrow] 1; 155: 1 [right arrow] 0; 174: 1 [right arrow] 0; 242: 1 [right arrow] 0). However, it should be stressed that this clade is not supported by a bootstrap value [greater than or equal to] 50% (Fig. 2). Obviously, the relationships between the Clupeoidea, the Engrauloidea and the Pristigasteroidea can only be appropriately examined in a cladistic analysis including many other representatives of these three groups.

[FIGURE 24 OMITTED]

Ilisha: [143:0 [right arrow] 1], [160:1 [right arrow] 0]

Clade C31: [32:0 [right arrow] 1], [111:0 [right arrow] 1], [209:0 [right arrow] 1]

As expected (see Fig. 1), the two engrauloid taxa examined are grouped together. This clade is supported by a bootstrap value of 66% (Fig. 2).

Engraulis: [144:0 [right arrow] 1]; Thryssa: [88:0 [right arrow] 1], [166:1 [right arrow] 0]

Clade C32: [24:1 [right arrow] 0], [45:0 [right arrow] 1], [77:0 [right arrow] 1], [111:0 [right arrow] 1], [113:0 [right arrow] 1], [167:1 [right arrow] 0], [178:0 [right arrow] 1]

The assembly of the ostariophysan fishes examined in this clade is expected (see Fig. 1). It is supported by a bootstrap value of 78% and by the following unambiguous synapomorphies (one of them, 113: 0 [right arrow] 1, being homoplasy free within the fishes examined): presence of protractor pectoralis (24: 0 [right arrow] 1); absence of basisphenoid (45: 0 [right arrow] 1); ribs/parapophyses of third free vertebra highly modified (77: 0 [right arrow] 1); adductor mandibulae attaching not only on mandible and/or primordial ligament, near its mandibular insertion, but also on other structures (111: 0 [right arrow] 1); presence of adductor mandibulae A1-OST (113: 0 [right arrow] 1); primordial ligament attaching posteriorly on dorsal surface of coronoid process (167: 1a0); absence of toothed dermopalatine (178: 0 [right arrow] 1). It is important to note that the scoring of the first feature as an ostariophysan synapomorphy might well be an artificial result associated with the using of the specific clupeomorph taxa included in the present cladistic analysis. This is because various clupeomorphs other than Ethmalosa (the only clupeomorph analyzed with CS0) actually have a protractor pectoralis (see e.g. Greenwood & Lauder). Also, one should be cautious about the feature concerning the attachment of the primordial ligament on the dorsal surface of the coronoid process. This feature was considered by Fink & Fink (1981, 1996) as a synapomorphy of gymnotiforms + siluriforms. However, the cypriniforms examined in the present work do also exhibit a primordial ligament attaching posteriorly on the dorsal surface of the coronoid process (except Danio, in which this feature could not be appropriately discerned). Since the condition present in the gonorynchiform and in the fossil taxa examined is also not clear, either because it was difficult to discern this feature or because a distinct primordial ligament is seemingly missing, such an attachment of this ligament on the dorsal surface of the coronoid process was scored in the cladogram as a potential ostariophysan synapomorphy. However, precisely because the condition in these taxa is not clear, one should be reticent regarding the acceptance of this feature as a synapomorphy of the Ostariophysi as a whole. The other feature listed above that was not listed by Fink & Fink (1981, 1996) as an ostariophysan synapomorphy is the presence of an A1-OST. This feature is exclusively, and consistently, found in ostariophysans, and does seem to constitute a well-grounded synapomorphy of these fishes (e.g. Gosline, 1989; Diogo & Chardon, 2000; Diogo; this work). The monophyly of ostariophysans is therefore well supported (Fig. 2), thus supporting some recent molecular studies (e.g. Lavoue et al.) and contradicting some others, in which some or all gonorynchiforms were placed as the sister-group of some or all clupeiforms included in those studies (e.g. Ishiguro et al.; Saitoh et al., 2003; Wang et al., 2003; Inoue et al., 2004).

[FIGURE 24 OMITTED]

Clade C33: [44:0 [right arrow] 1], [166:0 [right arrow] 1]

The grouping of the gonorynchiform fishes examined is expected (see Fig. 1). It is supported by a bootstrap value of 55% (Fig. 2) and by the two synapomorphies listed above. Interestingly, the genus Gonorynchus, the genus Chanos, and the Kneriidae sensu Grande & Poyato-Ariza (1999), appear in an unresolved trichotomy (Fig. 2). In the most recent and extensive morphological cladistic analysis of the Gonorynchiformes (Grande & Poyato-Ariza) the Gonorynchidae, in which Gonorynchus is included, is placed as the sister-group of the Kneriidae. However, in the most recent and extensive molecular cladistic analysis of this order (Lavoue et al.) Gonorynchus appears as the sister-group of the remaining extant gonorynchiform genera. As stressed about 20 years ago by Howes (1985), the phylogenetic position of Gonorynchus thus continues to be a particularly problematic issue. As stated by Howes (1985), at least some members of the genus Gonorynchus share some peculiar derived anatomical features with the Kneriidae (e.g. 133: 0 [right arrow] 1, absence of adductor mandibulae Aw), with the genus Phractolaemus (249: 0 [right arrow] 1, mandible highly modified, the dentary bone being roughly perpendicular to the main body of the mandible), and with the genus Chanos (216: 0 [right arrow] 1, presence of prominent, thin, dorsally/anterodorsally oriented anterodorsal projection of opercle). It is hoped that work in progress, together with T. Grande and F. Poyato-Ariza, which focuses on the osteology of all the gonorynchiform fossil taxa as well as on both the myology and osteology of all the extant taxa of this order, could help to clarify this question.

[FIGURE 26 OMITTED]

Chanos: [57:0 [right arrow] 1], [75:1 [right arrow] 0]; Gonorynchus: [6:0 [right arrow] 1], [50:0 [right arrow] 1], [87:0 [right arrow] 1], [95:0 [right arrow] 1], [112:0 [right arrow] 1], [125:0 [right arrow] 1], [139:0 [right arrow] 1], [140:0 [right arrow] 1], [169:0 [right arrow] 1], [177:0 [right arrow] [right arrow] 1], [184:0 [right arrow] 1], [193:1 [right arrow] 0], [197:0 [right arrow] 1], [211:0 [right arrow] [right arrow] 1], [249:0 [right arrow] 1], [269:0 [right arrow] 1]

Clade C34: [37:0 [right arrow] 1], [123:0 [right arrow] 1], [227:0 [right arrow] 1]

The grouping of the kneriid taxa examined is expected (e.g. Grande & Poyato-Ariza, 1999). It is supported by a bootstrap value of 66% and by the three synapomorphies listed above (Fig. 2).

Phractolaemus: [4:0 [right arrow] 1], [36:0 [right arrow] 1], [41:0 [right arrow] 1], [86:0 [right arrow] 1], [88:0 [right arrow] 1], [98:0 [right arrow] 1], [112:0 [right arrow] 1], [120:0 [right arrow] 1], [121:0 [right arrow] 1], [156:0 [right arrow] 1], [160:1 [right arrow] 0], [175:0 [right arrow] 1], [176:0 [right arrow] [right arrow] 1], [181:0 [right arrow] 1], [183:0 [right arrow] 1], [194:0 [right arrow] 1], [199:0 [right arrow] [right arrow] 1], [218:0 [right arrow] 1], [221:0 [right arrow] 1], [229:0 [right arrow] 1], [249:0 [right arrow] 1], [251:0 [right arrow] 1], [263:0 [right arrow] 1]

Clade C35: [61:0 [right arrow] 1], [164:0 [right arrow] 1], [258:0 [right arrow] 1]

The assembly of these four taxa examined is expected (e.g. Grande & Poyato-Ariza, 1999). It is supported by a bootstrap value of 71% and by the three synapomorphies listed above (Fig. 2).

Clade C36: [46:0 [right arrow] 1], [223:1 [right arrow] 0]

The grouping of the two taxa examined belonging to the Cromeriini is expected (e.g. Grande & Poyato-Ariza). It is supported by a bootstrap value of 78% and by the two synapomorphies listed above (Fig. 2).

Grasseichthys: [221:0 [right arrow] 1]; Cromeria: No unambiguous features

Clade C37: [108:0 [right arrow] 1], [198:0 [right arrow] 1], [250:0 [right arrow] 1]

The grouping of the two taxa examined belonging to the Kneriini is expected (e.g. Grande & Poyato-Ariza). It is supported by a bootstrap value of 90% and by the three synapomorphies listed above (Fig. 2).

Parakneria: No unambiguous features; Kneria: [145:0 [right arrow] 1]

Clade C38: [32:0 [right arrow] 1]

In Taverne's (1999) paper describing [dagger] Sorbininardus apuliensis, that author considered probable (based on a handmade tree made by him) that this taxon is an ostariophysan, and namely the sister-group of Gonorynchiformes. The results of the present study support the first hypothesis, but do not provide support for the second: within ostariophysans, [dagger] Sorbininardus apuliensis appears more closely related to the non-gonorynchiform fishes examined than to gonorynchiforms (Fig. 2). However, it should be stressed that this clade 38 is not supported by a bootstrap [greater than or equal to] 50% (Fig. 2), and is supported by a single synapomorphy: anteroventral margin of prevomer situates well posteriorly to anteroventral margin of mesethmoid (32: 0 [right arrow] 1, only occurring independently in a few taxa outside this clade C38 and only reverted, inside of it, in cypriniforms; the condition of [dagger] Santanichthys diasii and of [dagger] Clupavus maroccanus is not clear). Fink & Fink (1981, 1996) considered a similar feature as a synapomorphy of the clade including siluriforms + gymnotiforms + characiforms. Because this feature, as defined in the present work, is seemingly also found in [dagger] Sorbininardus apuliensis (e.g. Taverne, 1999: Fig. 3), in [dagger] Chanoides macropoma (e.g. Patterson, 1984: fig. 6B) and in [dagger] Lusitanichthys characiformis (e.g. Gayet, 1985: fig. 19; p. 114) (the condition of the two other fossil taxa included in the analysis is not clear), it was scored in the tree of Figure 2 as a synapomorphy of this clade C38. One should however keep in mind that, as explained above, this is the only unambiguous synapomorphy supporting this clade C38 and, thus, that the evidence provided in the present work to support the clade is not strong, although it is stronger than that supporting a close relationship between [dagger] Sorbininardus apuliensis and gonorynchiforms (Fig. 2).

[FIGURE 27 OMITTED]

[FIGURE 28 OMITTED]

[dagger] Sorbininardus apuliensis: [49:0 [right arrow] 1], [101:1 [right arrow] 0], [246:1 [right arrow] 0]

Clade C39: [60:1 [right arrow] 0], [76:0 [right arrow] 1], [78:0 [right arrow] 1]

This clade is supported by three unambiguous synapomorphies and by a bootstrap value of 76% (Fig. 2). In principle, it can be named Otophysi, although this places the Sorbininardiformes in a kind of limbo (because they are not Otophysi but, according to the scenario proposed in Fig. 2, they cannot be considered Anatophysi, as this will rend the Anatophysi paraphyletic). The three unambiguous synapomorphies supporting this clade C39 are: main bodies of parietals (or of parieto-extrascapulars) not widely separated from each other in dorsal view (60: 1 [right arrow] 0); presence of 'rudimentary tripus' (76: 0 [right arrow] 1; the condition in [dagger] Clupavus maroccanus is not clear); presence of 'rudimentary os suspensorium' (78: 0 [right arrow] 1; the condition in [dagger] Santanichthys diasii is not clear). Some features with ambiguous distributions may be interpreted as synapomorphies of this clade if a 'fast optimization' is chosen (e.g. 101: 0 [right arrow] 1, mesial limb of coracoids or scapulo-coracoids broad and anteroposteriorly elongated) or, alternatively, if a 'slow optimization' is adopted (e.g. 69: 0 [right arrow] 1, presence of 'rudimentary scaphium').
COPYRIGHT 2008 Universidad de La Frontera, Facultad de Medicina
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2008 Gale, Cengage Learning. All rights reserved.

 
Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Part 1
Author:Diogo, Rui; Doadrio, Ignacio; Vandewalle, Pierre
Publication:International Journal of Morphology
Date:Sep 1, 2008
Words:11397
Previous Article:Study on cranial and caudal mesenteric arteries in opossum (Didelphis albiventris)/Estudio de las arterias mesentericas craneal y caudal del zorrillo...
Next Article:Teleostean phylogeny based on osteological and myological characters/Filogenia de teleosteos basada en caracteristicas osteologicas y miologicas.

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters