Phylogeny of the Channichthyidae (Notothenioidei, Teleosteil
289
end region. Noncorrected pairwise distances between the 14 ingroup
species ranged from 4.07% to 19.97% (not shown). Hence, D-Ioop is
potentially more variable than cytochrome b gene for which pairwise
differences ranged from 1.86% to 9.31%. Intraspecific distances were
evaluated by comparing our individual of Chionodraco hamatus from
Terre Ad6lie with three individuals of C. hamatus from the Ross Sea
(sequences taken from L. Bargelloni's unpublished data). The pairwise
differences ranged from 0.82% to 1.73% within the populations from the
Ross Sea and from 1.36% to 1.64% between the populations from two
different localities. These values of intraspecific diversity are smaller than
those of interspecific ones. Anyway, the topologies obtained from these
two different mitochondrial genes showed very few differences, suggesting
that possible bias from intraspecific variability should be negligible here
and permit us to reconstruct a robust phylogeny for channichthyids at the
species level. Moreover, no saturation was detected when the pairwise
observed transitional differences were plotted against the pairwise inferred
transitional substitutions (the regression coefficient is 0.93).
Phylogenetic Analyses
Branch and Bound searches were performed on the 15 species for the two
mitochondrial genes (cytochrome band D-Ioop respectively) separately,
each yielding only one most-parsimonious tree [tree length (1) = 299,
retention index (RI) = 0.546, consistency index (Cl) = 0.639 for the
cytochrome b; I = 354, R.l. = 0.612, C.l. = 0.698 for the D-Ioop]. These
two trees shared the same topology excepting the positions of three
species: Chaenodraco wilsoni,
Chaenocephalus aceratus and
Channichthys rhinoceratus, but all having their branching point in the
same area. The NJ method yielded the same topologies as above. It must be
noted that differences in topology only affected nodes with short internal
branches and low bootstrap proportions. Anyway, we consider that, when
differences in topology occur, the best way to evaluate the relative strength
of the two phylogenetic signals is to perform a total evidence approach,
because the congruence of characters is more important than the
congruence of trees [16]. Thus, we combined the two data sets in a single
one. An interesting feature of this study is that the phylogenetic analysis
based on the 1135 sites from the two mitochondrial genes resulted in a
better resolution of the channichthyid relationships in terms of bootstrap
proportions. A Branch and Bound search yielded two equiparsimonious
trees (659 steps, Cl = 0.665, RI = 0.570, Fig. 1). The NJ method yielded
one of these two (Fig. 1, left). These trees were very similar to the ones
derived from the two genes separately, but bootstrapping showed higher
289
end region. Noncorrected pairwise distances between the 14 ingroup
species ranged from 4.07% to 19.97% (not shown). Hence, D-Ioop is
potentially more variable than cytochrome b gene for which pairwise
differences ranged from 1.86% to 9.31%. Intraspecific distances were
evaluated by comparing our individual of Chionodraco hamatus from
Terre Ad6lie with three individuals of C. hamatus from the Ross Sea
(sequences taken from L. Bargelloni's unpublished data). The pairwise
differences ranged from 0.82% to 1.73% within the populations from the
Ross Sea and from 1.36% to 1.64% between the populations from two
different localities. These values of intraspecific diversity are smaller than
those of interspecific ones. Anyway, the topologies obtained from these
two different mitochondrial genes showed very few differences, suggesting
that possible bias from intraspecific variability should be negligible here
and permit us to reconstruct a robust phylogeny for channichthyids at the
species level. Moreover, no saturation was detected when the pairwise
observed transitional differences were plotted against the pairwise inferred
transitional substitutions (the regression coefficient is 0.93).
Phylogenetic Analyses
Branch and Bound searches were performed on the 15 species for the two
mitochondrial genes (cytochrome band D-Ioop respectively) separately,
each yielding only one most-parsimonious tree [tree length (1) = 299,
retention index (RI) = 0.546, consistency index (Cl) = 0.639 for the
cytochrome b; I = 354, R.l. = 0.612, C.l. = 0.698 for the D-Ioop]. These
two trees shared the same topology excepting the positions of three
species: Chaenodraco wilsoni,
Chaenocephalus aceratus and
Channichthys rhinoceratus, but all having their branching point in the
same area. The NJ method yielded the same topologies as above. It must be
noted that differences in topology only affected nodes with short internal
branches and low bootstrap proportions. Anyway, we consider that, when
differences in topology occur, the best way to evaluate the relative strength
of the two phylogenetic signals is to perform a total evidence approach,
because the congruence of characters is more important than the
congruence of trees [16]. Thus, we combined the two data sets in a single
one. An interesting feature of this study is that the phylogenetic analysis
based on the 1135 sites from the two mitochondrial genes resulted in a
better resolution of the channichthyid relationships in terms of bootstrap
proportions. A Branch and Bound search yielded two equiparsimonious
trees (659 steps, Cl = 0.665, RI = 0.570, Fig. 1). The NJ method yielded
one of these two (Fig. 1, left). These trees were very similar to the ones
derived from the two genes separately, but bootstrapping showed higher