Finnerty and Block: Evolution of cytochrome b in the Scombroidei 



87 



Table 3 



Variable sites in the cytochrome 6 nucleotide alignment 

 (Fig. 2) according to codon position. Variable sites are fur- 

 ther characterized according to how many nucleotide states 

 are present: 2 states=twofold variable, 3 states=threefold 

 variable, fourstates=fourfold variable. 



Total sites 



Variable sites 



A) twofold variable sites 



B) threefold variable sites 



C) fourfold variable sites 



Minimum inferred 

 substitutions 

 = [(A) + 2(B) + 3(C)] 



Phylogenetically informative 

 variable sites 



42 18 188 248 



beromorus), and for the monophyly of the family 

 Istiophoridae. Interrelationships within the family 

 Istiophoridae and the genus Thunnus are well resolved. 

 No other nodes are supported by more than fifty per- 

 cent of bootstrap replicates (Felsenstein, 1985). Fur- 

 thermore, there is a substantial polychotomy. 



Weighted nucleotide analysis 



The lack of resolution in the unweighted nucleotide 

 analysis is not entirely unexpected. From Table 3, 

 we can surmise that many nucleotide sites have in- 

 curred multiple substitutions and therefore the like- 

 lihood of convergent substitutions or reversals is 

 high. In order to minimize the confounding effects of 

 these homoplasious base substitutions, we have 

 weighted infrequent substitution types more heavily 

 using a modification of the method of Knight and 

 Mindell (1993). If we disregard the direction of char- 

 acter change, we can place all nucleotide substitu- 

 tions into six classes : A<=>G, C<=>T, G<=>T, G<=>C, A<=>T, 

 and A<=>C. Through pairwise sequence comparisons 

 we obtained observed counts for each of these sub- 

 stitution types (Table 2; also see Methods section). 

 We observe a nearly 50-fold difference between the 

 most common (Cc=>T) and the least common (G<=>T) 

 substitution types. Then, from the total number of 

 observed substitutions and the observed frequency 

 of each base, we derived the expected number of oc- 

 currences for each substitution type. The ratios of 

 expected occurrences to observed occurrences for each 



substitution type (EOR's) were then used to weight the 

 six types of base substitutions. The result of this weight- 

 ing scheme is that substitution types that occur less 

 frequently than expected are weighted more heavily. 



A phylogeny based on EOR weighting of nucleotide 

 substitutions is presented in Figure 4. It retains all 

 of the strongly supported nodes that appear in the 

 unweighted topology. In addition, the weighted to- 

 pology contains three more basal nodes that are 

 strongly supported by the bootstrap analysis (>50%): 

 the node uniting Gempylidae, Scombridae, and 

 Trichiuridae, the node uniting Xiphiidae and 

 Istiophoridae, and the node uniting Auxis and 

 Euthynnus. This suggests that the character weight- 

 ing scheme has accomplished its goal to some extent: 

 we have retained the phylogenetic signal from rap- 

 idly evolving substitutions while emphasizing the phy- 

 logenetic signal from slowly evolving substitutions. 



According to the weighted cladogram (Fig. 4), all 

 scombroids fall into two clades. The billfishes com- 

 prise one clade consisting of a monophyletic 

 Istiophoridae and its sister group, Xiphiidae. All 

 other scombroids (Gempylidae, Scombridae, and 

 Trichiuridae) fall into a separate clade. This major 

 split within the suborder Scombroidei is in agree- 

 ment with our previous study (Block et al., 1993). 

 However, in contrast with our previous study, the 

 use of character weighting and the inclusion of more 

 distant outgroups leads to the result that the subor- 

 der Scombroidei is not monophyletic. On the most 

 parsimonious tree, Sphyraena and Coryphaena share 

 a common ancestor with the gempylid-scombrid- 

 trichiurid clade to the exclusion of billfishes, though 

 this node does not receive particularly strong sup- 

 port from the bootstrap analysis. This result indi- 

 cates some support for the hypothesis that billfishes 

 are not scombroids. More importantly, the cladogram 

 excludes the possibility that billfishes and scombrids 

 comprise a monophyletic group within the 

 Scombroidei, as required by the scombrid subgroup 

 and scombrid sister group hypotheses. In summary, 

 the weighted analysis agrees with the nonscombroid 

 hypothesis and conflicts with the scombroid subgroup 

 and scombroid sister group hypotheses. 



Amino acid analysis 



Amino acid substitutions occur far less frequently 

 than nucleotide substitutions owing to the strong 

 functional constraints on many regions of the mol- 

 ecule. Cytochrome b is a component of the electron 

 transport chain and spans the inner mitochondrial 

 membrane. The portion of the gene sequenced in this 

 study encodes 195 amino acids corresponding to resi- 

 dues 46 through 240 of the human cytochrome b (Fig. 5). 



