Phylogeny of the Channichthyidae (Notothenioidei, Teleostei)
295
Divergence Times
Our cytochrome b sequences do not exhibit mutational saturation, neither
within the Channichthyidae nor between the perciform outgroup Trachurus
and the channichthyids. There are two techniques to infer divergence times
from nonsaturated sequences. Both postulate the molecular clock. First,
one can use published mitochondrial rates and apply it to one's pairwise
distance matrix to infer the divergence time of a particular node. The
second technique consists in calculating its own rate using a time
calibration given by one species for which paleontological data provide a
divergence time. Gymnodraco is the sister-group of channichthyids [4,5].
Let us call the node X the divergence of Gymnodraco from the
channichthyid ancestor and node Y the divergence of Champsocephalus
from the rest of the channichthyids. According to the first technique,
assuming a rate of mitochondrial genes between 0.3% lMa and 0.7%lMa
[4,24-27] provides a divergence time of35 to 15 Ma for the node X and 27
to 12 Ma for Y. Assuming a transversional rate of 0.14%lMa [4], this
divergence time is 23 Ma for X and 15 Ma for Y. According to the second
technique, we have taken the perciform diversification dated back to 55 Ma
[28,29] as a calibrating point. Whatever the nonnotothenioid outgroup
chosen to represent a perciform, this rate, so calculated, is 0.47%lMa. This
is due to the fact that perciforms probably underwent a radiation [30], so
that taking one or another perciform outgroup did not change the pairwise
distances used to calculate the rate. The divergence time obtained for the
node X is 23 Ma and 17.5 Ma for the node Y.
This period of time when the first channichthyid ancestor must have
lived can be correlated to the development of unrestricted Antarctic
circumpolar current and the initial development of the Antarctic polar front
[1]. This second technique of calculation seems to be better than the former
because it does not impose any extrinsic rate to our cytochrome b
sequences, although it requires a correct paleontological time calibration. It
is well known that 16S mtDNA sequences evolve slower than cytochrome
b ones (for example see Meyer [30]). Therefore, it is questionable to apply
the same transversional rate of 0.14% to both genes. This is why we prefer
the second technique, which follows the cytochrome b intrinsic rate. This
may also be the reason why Bargelloni and Lecointre (this Vol.) also found
from 12S and 16S mtDNA a divergence time of 23 Ma for the emergence
of Eleginops, a very early notothenioid lineage. Applying an extrinsic rate
to 12S and 16S sequences may have underestimated the real divergence
time of Eleginops. A paleontological time is needed to confirm this point
of view. The other hypothesis is that the nonbovichtid notothenioid
radiation may have occurred in a very short time span.
295
Divergence Times
Our cytochrome b sequences do not exhibit mutational saturation, neither
within the Channichthyidae nor between the perciform outgroup Trachurus
and the channichthyids. There are two techniques to infer divergence times
from nonsaturated sequences. Both postulate the molecular clock. First,
one can use published mitochondrial rates and apply it to one's pairwise
distance matrix to infer the divergence time of a particular node. The
second technique consists in calculating its own rate using a time
calibration given by one species for which paleontological data provide a
divergence time. Gymnodraco is the sister-group of channichthyids [4,5].
Let us call the node X the divergence of Gymnodraco from the
channichthyid ancestor and node Y the divergence of Champsocephalus
from the rest of the channichthyids. According to the first technique,
assuming a rate of mitochondrial genes between 0.3% lMa and 0.7%lMa
[4,24-27] provides a divergence time of35 to 15 Ma for the node X and 27
to 12 Ma for Y. Assuming a transversional rate of 0.14%lMa [4], this
divergence time is 23 Ma for X and 15 Ma for Y. According to the second
technique, we have taken the perciform diversification dated back to 55 Ma
[28,29] as a calibrating point. Whatever the nonnotothenioid outgroup
chosen to represent a perciform, this rate, so calculated, is 0.47%lMa. This
is due to the fact that perciforms probably underwent a radiation [30], so
that taking one or another perciform outgroup did not change the pairwise
distances used to calculate the rate. The divergence time obtained for the
node X is 23 Ma and 17.5 Ma for the node Y.
This period of time when the first channichthyid ancestor must have
lived can be correlated to the development of unrestricted Antarctic
circumpolar current and the initial development of the Antarctic polar front
[1]. This second technique of calculation seems to be better than the former
because it does not impose any extrinsic rate to our cytochrome b
sequences, although it requires a correct paleontological time calibration. It
is well known that 16S mtDNA sequences evolve slower than cytochrome
b ones (for example see Meyer [30]). Therefore, it is questionable to apply
the same transversional rate of 0.14% to both genes. This is why we prefer
the second technique, which follows the cytochrome b intrinsic rate. This
may also be the reason why Bargelloni and Lecointre (this Vol.) also found
from 12S and 16S mtDNA a divergence time of 23 Ma for the emergence
of Eleginops, a very early notothenioid lineage. Applying an extrinsic rate
to 12S and 16S sequences may have underestimated the real divergence
time of Eleginops. A paleontological time is needed to confirm this point
of view. The other hypothesis is that the nonbovichtid notothenioid
radiation may have occurred in a very short time span.