ADRIANICHTHYIDAE TO OTHER ATHERINOMORPH FISHES

Parenti, Lynne R., 2008, A phylogenetic analysis and taxonomic revision of ricefishes, Oryzias and relatives (Beloniformes, Adrianichthyidae), Zoological Journal of the Linnean Society 154 (3), pp. 494-610 : 506-507

RELATIONSHIPS OF ADRIANICHTHYIDAE TO OTHER ATHERINOMORPH FISHES

Ricefishes had been placed traditionally in the order Cyprinodontiformes until they were reclassified as a sister group of the exocoetoids in the order Beloniformes , the cyprinodontiform sister group, by Rosen & Parenti (1981). This hypothesis of ricefish relationships has been well corroborated by morphologists ( Parenti, 2005; Fig. 4 ) and molecular systematists (M. Miya, pers. comm., 2003), and challenged recently only by Li (2001) who contended that morphological evidence supports the close relationship of ricefishes to cyprinodontiforms. Twenty-eight characters and their states in cyprinodontiforms, ricefishes and exocoetoids were tabulated by Li (2001: table 1) who calculated an undefined ‘specialized degree’ of each taxon. According to Li, the ‘specialized degree’ of Cyprinodontiformes is 66.5, that of ricefishes 65.7 and that of exocoetoids 54.3 ( Li, 2001: 585). This metric was offered as evidence that ricefishes are more closely related to cyprinodontiforms than to exocoetoids.

Li and I disagree not only in the description of characters, but also in how they should be used to interpret phylogenetic relationships. Some characters tabulated by Li and used to calculate the ‘specialized degree’ are irrelevant to the question of ricefish relationships. For example, the number of basibranchials was coded by Li (2001) as plesiomorphic (three basibranchials) in ricefishes, exocoetoids and aplocheiloid cyprinodontiforms, and apomorphic (two basibranchials) in cyprinodontoid cyprinodontiforms. This coding is correct, but the character is silent on relationships of ricefishes to either exocoetoids or cyprinodontiforms. Other characters were dismissed by Li (2001) using particular arguments. For example, absence of an interhyal bone is a cogent beloniform synapomorphy, following Rosen & Parenti (1981) and below; Li (2001: 26) argues that although the interhyal is present in all cyprinodontiforms, it is cartilaginous in some, and therefore shows a ‘degenerate tendency.’

Beloniform synapomorphies enumerated by Rosen & Parenti (1981) and Parenti (2005) include: (1) interhyal bone absent; (2) interarcual cartilage absent; (3) presence of only a single, ventral hypohyal bone; (4) relatively small second and third epibranchials; (5) vertical reorientation of the second pharyngobranchial bone; and (6) caudal skeleton characterized by the lower caudal lobe with more principal rays than in the upper caudal lobe. An additional beloniform synapomorphy proposed by Rosen & Parenti (1981), i.e. large, ventral flanges on the fifth certaobranchials, was interpreted by Stiassny (1990) as an atherinomorph synapomorphy, and I concur (see also Parenti, 2005).

To the above six beloniform synapomorphies, I add a seventh: parietals extremely small or absent. Parietal bones are absent from ricefishes at all stages of development ( Kulkarni, 1948; Yabumoto & Uyeno, 1984; Langille & Hall, 1987). In exocoetoid beloniforms ‘... the parietals, when present, are very small, separated by the supraoccipital’ ( Regan, 1911b: 328). This is in contrast to the cyprinodontiforms Aplocheilus , Xiphophorus and Gambusia , for example, in which parietals are present ( Ramaswami, 1946). Parietals may be present or absent in cyprinodontoid cyprinodontiforms, such as poeciliids (e.g. Rosen & Bailey, 1963). Likewise, parietals are present in atherinopsid and notocheirid atheriniforms, yet are absent in atherines ( Dyer & Chernoff, 1996). Li (2001: 585) argues that the absence of parietals among a variety of cyprinodontiform taxa and beloniforms is evidence of their close relationship. I disagree with this interpretation because well-formed parietals are present in most and basal cyprinodontiforms.

Another putative beloniform synapomorphy, ventral position of the lateral line neuromasts, is considered here and discussed further under monophyly of the family Adrianichthyidae (see character 17, below). Exocoetoids have long been known to have a well-developed ventral lateral line, but the polarity of this character in phylogeny reconstruction has been questioned (e.g. Rosen & Parenti, 1981: 16–17) largely because other atherinomorphs – atheriniforms, cyprinodontiforms and ricefishes – have been described incorrectly as having no lateral line because the scales may be pitted, but are rarely pored. The posterior lateral line nerve is well developed in these atherinomorphs (e.g. Ishikawa, 1994; material here of O. carnaticus, AMNH 20650), as in exocoetoids, and the body has numerous superficial neuromasts, but there are no pored lateral line scales. Differences in neuromast patterns noted by Yamamoto (1975: fig. 10-2a) include the absence of neuromasts from the mid-lateral trunk region in the medaka and their presence in the cyprinodontiform Fundulus , following Denny (1937). Recent study of lateral line development at the cellular level confirms the unique pattern of the medaka, compared with other model organisms such as the zebrafish, and offers an ontogenetic mechanism – ventral neuromast migration – to shift the lateral component in adults ( Sapède et al., 2002: 613): ‘At the time the hair cells differentiate... the L-PLL [lateral branch of the posterior lateral line] neuromasts have already migrated all the way to the ventral midline, towing their innervating axons along the way.’ Homology of the ventral position of the lateral line in exocoetoids and ricefishes may be tested with additional developmental studies; I consider it additional evidence of their sister group relationship.

The explicitly phenetic, rather than phylogenetic, arguments raised by Li (2001) do not reject the hypothesis of Rosen & Parenti (1981) that ricefishes are beloniform fishes. Furthermore, Li (2001) unfortunately did not have access to more recent literature (e.g. Parenti, 1993) in which some of the characters he dismisses, such as ‘prolonged embryonic development’, were re-evaluated as atherinomorph, rather than cyprinodontiform, synapomorphies (see Parenti, 2005).

Monophyly of the exocoetoids (sensu Rosen & Parenti, 1981; Collette et al., 1984) is well corroborated and need not be reviewed here. Ricefishes are hypothesized to be the sister group of exocoetoids and therefore polarity of the characters used to infer phylogenetic relationships among ricefish species will be hypothesized using comparison with exocoetoid, cyprinodontiform and atheriniform outgroup taxa. The relationships among the atherinomorph orders, reviewed extensively elsewhere ( Rosen & Parenti, 1981; Parenti, 1993, 2005; Dyer & Chernoff, 1996), will not be tested further here.

Figures

Figure 4. Strict consensus of two cladograms of relationships among 28 species of ricefishes, family Adrianichthyidae (shaded) as produced by the phylogenetic analysis based on the characters coded as listed (Appendix 1) and as presented in a data matrix (Appendix 2). Unsupported nodes were collapsed. Support at each node is discussed in the text. Coded outgroup taxa representing the major atherinomorph lineages are Menidia (Atheriniformes), Rivulus (Cyprinodontiformes) and Zenarchopteridae (Beloniformes, suborder Exocoetoidei). Cyprinodontea = Cyprinodontiformes plus Beloniformes (following Dyer & Chernoff, 1996). See Phylogenetic analysis for discussion and details.