plements the bootstrap values. The BLAST program was
used for sequence alignment (http://www.ncbi.nlm.nih.
gov/BLAST).
The linearized NJ tree was also obtained for estimating
divergence times using the divergence between the
chicken and pheasant as a molecular calibration point
[10]. Times of species divergence are only rough esti-
mates. Thus, the time scale for the linearized NJ tree was
obtained by comparing cyt b of chicken [11] and pheas-
ant [12], two species that diverged about 20 million years
ago (MYA) [13] according to combined fossil and mole-
cular comparison calculations. The trees shown in this pa-
per were rooted with Fringilla coelebs.
Maximum parsimony (MP), NJ with maximun-likeli-
hood distances, linearized NJ with Kimura biparametric
distancesn and UPGMA with biparametric distance ma-
trices were obtained with the PAUP*4.0b2 program,
kindly provided by Swofford [14] and with the MEGA
package program in the case of the linearized tree [15].
The following calculations were carried out: number of
substitutions (synonymous and nonsynonymous), num-
ber of variable and phylogenetically informative sites,
and the base composition according to codon position.
Bootstrap values were calculated to test the topology ro-
bustness of trees [16]. Low bootstrap values have been
included in cases when the same topology was supported
by at least two different tree construction methodologies
[17]. To further assess node robustness, the statistical CP
of a particular sequence cluster was calculated using the
MEGA program [15]. Subsequent analyses were de-
signed to take into account levels of saturation (multiple
substitutions at single sites) in different partitions of the
data sets. Scatter plots were drawn to compare pairwise
percent sequence divergence to pairwise transversion and
pairwise transition divergence at first, second, and third
codon positions; both Gallus gallus and Fringilla coelebs
chaffinch were used in the saturation plots (fig. 1). Two
different estimates of percent divergence were used; these
serve as approximations of time since divergence:
Kimura’s two-parameter [18] genetic distance, and un-
corrected pairwise divergence (p = Nd/n, where p is the
percent sequence divergence, Nd is the number of nu-
cleotides that differ between two sequences, and n is the
total number of nucleotides compared [19, 20]).
Domestic chicken (G. gallus [11]) was used as a distant
outgroup in both UPGMA and NJ linearized trees. Simi-
lar tree topologies were obtained with either Gallus or the
more closely related F. coelebs (not shown).
A LINTRE test was also performed to test whether a mo-
lecular clock exists among all lineages used in our analy-
ses [9]; linearized trees were obtained using this test, re-
estimating the branch lengths under the assumption of a
constant rate of evolution (i.e., a molecular clock), and
examining whether these showed significant differences
from the trees illustrated in figure 3 (see below). The
computer software used for these calculations can be ob-
tained from the web sites: ftp://ftp.bio.indiana.edu/mol-
bio/evolve/lintr/ or http://cib.nig.ac.jp/dda/ntakezak.html.
Results and discussion
Patterns of DNA base substitution
Saturation plots for cyt b (fig. 1) indicated that only third-
position transitions showed a clear leveling-off associated
with saturation; this occurred at 13% uncorrected total
sequence divergence (crossbills, rosefinches, bullfinches,
and grosbeaks/Fringilla) and at sequence divergences of
more than 20% in the comparisons of crossbills,
rosefinches, bullfinches and grosbeaks/Gallus (see fig.
1). Assessment of saturation gave similar results using
Kimura’s two parameter distances. We concluded that
five out six data partitions (first, second, third codon po-
sition bases and transitions/transversions) were not satu-
rated and were thus available to calculate correct phylo-
genies. Variable and phylogenetically informative sites
were also calculated; there are 328 and 250 respectively,
when crossbills, rosefinches, bullfinches, and grosbeaks
are analyzed using F. coelebs as an outgroup, and 369 and
CMLS, Cell. Mol. Life Sci.
Vol. 58, 2001
Research Article
3
Figure 1. Saturation plots for the cytochrome b gene that relate un-
corrected sequence divergence to changes due to transitions (top)
and transversions (bottom) at first, second, and third codon posi-
tions. CB, crossbills; RF, rosefinches; BF, bullfinches; GB, gros-
beaks. Pinicola enucleator is included among bullfinches and
Haematospiza sipahi and Uragus sibiricus is included among
rosefinches.
F
