Results, Günther, 1859

Burridge, Christopher P., 2000, Molecular phylogeny of the Aplodactylidae (Perciformes: Cirrhitoidea), a group of Southern Hemisphere marine ® shes, Journal of Natural History 34 (11), pp. 2173-2185 : 2176-2182

Results

DNA sequences analysed were deposited in GenBank ( AF067084 , AF092140 , AF092155 ± AF092157 , AF092167 , AF133060 ± AF133069 ).

The mitochondrial partial cytochrome c oxidase subunit I and cytochrome b sequences analysed were 499 and 402 bp in length, respectively (fi gure 1). Among the aplodactylids 164 characters were variable, and 82 of these were phylogenetically informative. Transition nucleotide substitutions were observed at 153 sites, while TVs were observed at 25. Only six of the variable sites were not third codon positions. Length mutations were absent. This pattern of sequence evolution, also observed for the outgroup taxa, indicates that orthologous protein-coding sequences were obtained. The cytochrome b region was approximately 1.2 times more variable than the cytochrome oxidase I region, but both fragments contributed similar amounts of variation to the study due to the larger number of cytochrome oxidase I characters analysed.

The partition homogeneity test indicated phylogenetic congruence between cytochrome oxidase I and cytochrome b sequences, both for all characters (P = 0.36) and only third codon positions (P = 0.65), allowing the combination of genes during phylogenetic analyses. Tree length-frequenc y distributions were signi fi cantly skewed for all taxa (g 1 = — 0.60, P <0.01) and the aplodactylids alone (g 1 = — 1.09, P <0.01), suggesting the presence of phylogenetic signal. Visual inspection did not reveal any nucleotide substitution saturation among the aplodactylids, evidenced by the linear accumulation of transitions relative to transversions (fi gure 2). The TI:TV observed during comparisons between the aplodactylids and each of the three outgroups were signi fi cantly lower than that observed during comparisons among the aplodactylids (fi gure 2, P <0.001, df = 3, one-way ANOVA with Tukey HSD post hoc tests), indicating substitution saturation during the former.

Kimura (1980) two-parameter genetic distances among the aplodactylids ranged from 6.1 to 12.4% sequence divergence ( table 2 ). Distances between the aplodactylids and the three cirrhitoid outgroups, and among these outgroups, ranged from 18.3 to 23.1% ( table 2 ). The levels of intraspeci fi c variation were not assessed, but are typically less than 1.0% for mitochondrial protein-coding genes in other cirrhitoids (Burridge, 2000; Burridge, unpublished).

Neighbour-joining and unweighted maximum parsimony analyses clustered the aplodactylids as monophyletic, with high bootstrap support (fi gure 3A). Aplodactylus arctidens and A. punctatus were clustered together, with A. westralis as sister taxon to this clade. Aplodactylus lophodon and then A. etheridgii were successively removed. These relationships received moderate to high bootstrap support from neighbourjoining analysis, but low to moderate support (50±70%) from unweighted parsimony analysis. Increased weighting of TVs during parsimony analysis, according to the optimum TI:TV of 3.0 from maximum likelihood analysis, produced the same topology as unweighted analysis, with similar bootstrap values (not shown). Monophyl y of the aplodactylids was also supporte d by parsimony analysis restricted to transversion substitutions, but the inferred relationships among these taxa were not well supported due to the small number of informative character-state changes (not shown).

The maximum likelihood topology differed from the neighbour-joining and unweighted maximum parsimony topology in root placement among the aplodactylids (fi gure 3B). The root was placed on the branch leading to A. lophodon and A. etheridgii , rather than that leading to A. lophodon alone. Enforcing the maximum likelihood topology during parsimony analysis produced a tree only one step longer than the most parsimonious unconstrained topology. Neither topology was signi fi cantly superior according to the Templeton (P = 0.86) or Kishino and Hasegawa (P = 0.75) tests. Choosing Chironemus marmoratus or Cheilodactylus fasciatus as the most basal outgroup had no effect on the inferred relationships among aplodactylids or the levels of bootstrap support from any method of analysis. Removing one or two outgroup taxa occasionally altered aplodactylid relationships inferred by a given method of analysis, but only in such a manner as observed for different methods of analysis when all outgroups were analysed.

The two-cluster and branch-length tests did not reveal signi fi cant nucleotide substitution rate heterogeneity for third codon positions at the 5% level, when nodes and branches were analysed individually. Signi fi cant rate heterogeneity was also not observed when nodes were analysed simultaneously by the two-cluster test (x 2 = 8.03, df = 6, P> 0.10).

Discussion

Two topologies were recovered from phylogenetic analysis of aplodactylid and outgroup mitochondrial DNA cytochrome oxidase I and cytochrome b sequences. These varied in the placement of the root among the aplodactylids, and neither was signi fi cantly superior. The New Zealand and south-eastern Australian species Aplodactylus arctidens and the South American species A. punctatus were placed as sister taxa, in a monophyletic clade with the south-west Australian species A. westralis . The inconsistency in root placement influenced the relationships among the remaining aplodactylids, A. lophodon from eastern Australia and A. etheridgii from northern New Zealand and several south-west Paci fi c islands. These species were either clustered as sister taxa, or were successively removed from the other three aplodactylids with A. lophodon basal. Bootstrap support for aplodactylid monophyly and the inferred relationships were moderate to high. There was no evidence of nucleotide substitution saturation among the aplodactylids, although saturation was present during comparisons with the outgroups. This may explain the variability in root placement ( Smith, 1994).

Taxonomy

The genus Crinodus , monotypic for C. lophodon , is only distinguished from Aplodactylus by larger scales and the absence of vomerine teeth, and has been relegated to the synonomy of Aplodactylus by Russell (2000). Genetic data provides support for this revision. Crinodus lophodon does not appear su ffi ciently divergent from species of Aplodactylus to be given distinct generic status; it is only slightly more divergent from A. punctatus , A. arctidens and A. westralis than is A. etheridgii ( table 2 ). The placement of C. lophodon is also uncertain, as two topologies were recovered and neither was signi fi cantly superior. If C. lophodon is the sister taxon to Aplodactylus (fi gure 3A), retention of Crinodus at the generic level could be argued despite the limited molecular and morphological divergence. However, if C. lophodon and A. etheridgii are sister taxa (fi gure 3B), then Aplodactylus is paraphyletic. Given the absence of marked genetic distinction, and the possibility that Aplodactylus is paraphyletic, Russell’s (2000) scheme in which Crinodus is synonymized with Aplodactylus is supported.

Most aplodactylids recognized by Russell (2000) have widespread distributions and encompass several previously described species. While Russell (2000) has examined many specimens throughout the range of each taxon he recognized, the possibility exists that speciation has occurred in some of these without the development of readily apparent distinguishing features. While genetic studies often identify such instances of cryptic speciation ( Knowlton, 1993), this study has not been su ffi ciently intensive to examine the status of species recognized by Russell (2000).

Zoogeography

The recovered phylogenies suggest that the Aplodactylidae originated in the vicinity of Australia and New Zealand, with the majority of radiation occurring prior to this family achieving representation in South America. This is evidenced by the basal positions of the Australian, New Zealand and south-west Paci fi c island aplodactylids relative to the South American A. punctatus (fi gure 3). An Australian± New Zealand origin and subsequent movement east has also been proposed for the cheilodactylid genera Nemadactylus and Acantholatris ( Burridge, 1999) .

Mitochondrial third codon position molecular clock calibrations of 2.3% and 3.3% sequence divergence per million years ( Martin et al., 1992; Bermingham et al., 1997) suggest that the aplodactylids shared a common ancestor 12.8±18.5 million years ago (Mya). Similarly, the disjunct trans-Paci fi c species pair A. arctidens and A. punctatus diverged 6.2±9.0 Mya. While estimates of divergence time from molecular clock calibrations should be treated cautiously given the number of untested assumptions ( Rand, 1994), these values appreciably post-date the isolation of Australia and South America from Antarctica during the fragmentatio n of Gondwana (40±30 Mya; Lawver et al., 1992). In addition, the presence of conspeci fi c populations in Australia and New Zealand ( A. arctidens ), but their sister species in South America ( A. punctatus ), also argues against a distribution based entirely on vicariance accompanying Gondwana fragmentation, as New Zealand was the fi rst of these land masses to become isolated ( Lawver et al., 1992). Therefore, the disjunct transoceani c distribution of aplodactylids is best explained by chance dispersal. Chance dispersal rather than Gondwanan vicariance was also inferred for the similarly distributed diadromous species Galaxias maculatus (Jenyns 1842) , based on intraspeci fi c relationships and levels of molecular divergence ( Waters and Burridge, 1999).

Chance dispersal of aplodactylids across the Paci fi c was most likely undertaken during their larval phase, as juveniles and adults are restricted to nearshore habitats ( Hutchins and Swainston, 1986; Stepien, 1990; Cole et al., 1992; Francis, 1996). Although little is known of aplodactylid larval dispersal capabilities (B. Bruce, personal communication), high dispersal capabilities have been suggested for other cirrhitoid larvae. The cheilodactylids N. macropterus Bloch and Schneider 1801 and A. monodactylus Carmichael 1818 possess a 9±12-month offshore pelagic larval phase ( Annala, 1987; Andrew et al., 1995), and molecular genetic studies on members of these genera suggest geneflow across distances in excess of 1000 km ( Elliott and Ward, 1994; Grewe et al., 1994; Burridge, 1999). Particularly relevant is the species pair A. gayi Kner 1865 and Nemadactylus sp. , which are similar in distribution to A. arctidens and A. punctatus and exhibit negligible cytochrome b sequence divergence ( Burridge, 1999).