3. Origin of Euteleostean Biodiversity
43
gence times of hundreds of million years ago, when distances are appropriately
estimated using the gamma-correction of the rate heterogeneity over sites (Fig. 3;
Kumazawa and Nishida 1999). In Figs. 3A and 3B, we examined the performance
of the ND2/cytb sequences as the molecular clock for long periods of time using
mammals (plus some birds) and sharks considered to have relatively reliable fossil
records. For mammals, reliable divergence-time estimates are also available from
an independent molecular data set using multiple nuclearly-encoded protein sequences, and these time estimates are largely in agreement with Cenozoic and Paleozoic fossil records (Kumar and Hedges 1998).
As shown in Fig. 3A, gamma-corrected ML distances among the mammalian
(plus avian in part) ND2/cytb sequences accumulate almost linearly (R=0.99) by
time. The linear relationship was similarly observed when gamma-corrected Poisson distances were plotted, but not in the case of simple ML or Poisson distances
(Kumazawa and Nishida 1999). The evolutionary rate of ND2/cytb amino acid
sequences for sharks (Fig. 3B) appears to be much slower (approximately a fourth)
than that of mammals (Fig. 3A). The reduction of molecular evolutionary rates in
sharks is in agreement with previous work (Martin et al. 1992) using DNA sequences. Taken together, these results support the assumption that ND2/cytb amino
acid sequences can perform as a molecular clock even for divergences of a few
hundreds of million years ago in time or 0.5-1.0 substitutions per site in pairwise
distance. These genes were therefore expected to be useful for estimating divergence times among teleosts.
3.3 Rate Calibration in Bony Fishes
Fragmentation processes of Gondwanaland have been well-elucidated from recent
geological evidence (Smith et al. 1994; The Plates Project 1998). Figure 4 shows
three possible models for the evolutionary history of cichlids based on both the
molecular and geological evidence. For the reasoning described in the legend of
Fig. 4, the vicariance or dispersal model leads to the reasonable assumption that
extant African and neotropical cichlids separated from each other at approximately
100 MYA (80-120 MY A). When molecular evolutionary rates for bony fishes were
calibrated under this assumption (Fig. 3C), the line calibrated at 100 MY A nearly
runs through another data point for the divergence between coelacanth and
actinopterygians (teleosts plus Amia in this case) plotted against an independent
time estimate using multiple nuclear protein sequences (450 MY A, Kumar and
Hedges 1998). Although the other data point for the divergence between bony fishes
and sharks appears to deviate slightly from this calibration line, it is placed well
within the possible range of calibration (see the dark-shaded area in Fig. 3C). If
cichlids had dispersed across the paleo-Atlantic or originated from labroid ancestor(s)
at a time which is less incompatible with the fossil record (i.e., 35-65 MY A from
the Eocene to the Paleocene), the molecular evolutionary rate for bony fishes would
have to be increased by at least 50% from the estimate calibrated at 100 MY A,
43
gence times of hundreds of million years ago, when distances are appropriately
estimated using the gamma-correction of the rate heterogeneity over sites (Fig. 3;
Kumazawa and Nishida 1999). In Figs. 3A and 3B, we examined the performance
of the ND2/cytb sequences as the molecular clock for long periods of time using
mammals (plus some birds) and sharks considered to have relatively reliable fossil
records. For mammals, reliable divergence-time estimates are also available from
an independent molecular data set using multiple nuclearly-encoded protein sequences, and these time estimates are largely in agreement with Cenozoic and Paleozoic fossil records (Kumar and Hedges 1998).
As shown in Fig. 3A, gamma-corrected ML distances among the mammalian
(plus avian in part) ND2/cytb sequences accumulate almost linearly (R=0.99) by
time. The linear relationship was similarly observed when gamma-corrected Poisson distances were plotted, but not in the case of simple ML or Poisson distances
(Kumazawa and Nishida 1999). The evolutionary rate of ND2/cytb amino acid
sequences for sharks (Fig. 3B) appears to be much slower (approximately a fourth)
than that of mammals (Fig. 3A). The reduction of molecular evolutionary rates in
sharks is in agreement with previous work (Martin et al. 1992) using DNA sequences. Taken together, these results support the assumption that ND2/cytb amino
acid sequences can perform as a molecular clock even for divergences of a few
hundreds of million years ago in time or 0.5-1.0 substitutions per site in pairwise
distance. These genes were therefore expected to be useful for estimating divergence times among teleosts.
3.3 Rate Calibration in Bony Fishes
Fragmentation processes of Gondwanaland have been well-elucidated from recent
geological evidence (Smith et al. 1994; The Plates Project 1998). Figure 4 shows
three possible models for the evolutionary history of cichlids based on both the
molecular and geological evidence. For the reasoning described in the legend of
Fig. 4, the vicariance or dispersal model leads to the reasonable assumption that
extant African and neotropical cichlids separated from each other at approximately
100 MYA (80-120 MY A). When molecular evolutionary rates for bony fishes were
calibrated under this assumption (Fig. 3C), the line calibrated at 100 MY A nearly
runs through another data point for the divergence between coelacanth and
actinopterygians (teleosts plus Amia in this case) plotted against an independent
time estimate using multiple nuclear protein sequences (450 MY A, Kumar and
Hedges 1998). Although the other data point for the divergence between bony fishes
and sharks appears to deviate slightly from this calibration line, it is placed well
within the possible range of calibration (see the dark-shaded area in Fig. 3C). If
cichlids had dispersed across the paleo-Atlantic or originated from labroid ancestor(s)
at a time which is less incompatible with the fossil record (i.e., 35-65 MY A from
the Eocene to the Paleocene), the molecular evolutionary rate for bony fishes would
have to be increased by at least 50% from the estimate calibrated at 100 MY A,