J.-K. PARK AND D. O´ FOIGHIL
Figure 2a shows the single most parsimonious unrooted phylogram (9 steps; 2 parsimony-informative characters; CI 5 1.0; RI 5 1.0) obtained by exhaustive PAUP analysis of the 10 Bermudan haplotypes. This unrooted tree had a starlike topology in which there was a distinct lack of structure; 7 of the 10 haplotypes had only a single connec- tion. A central topological position was occupied by BDA1, the most common Bermudan genotype, which was present in all three sampled sites and was one step away from 7 of the 9 other island CO I genotypes. Heuristic estimation of outgroup weighting (Castelloe and Templeton, 1994) for each of the Bermudan mt genotypes yielded a maximum value (0.445) for BDA1 (Fig. 2a), indicating that this is probably the oldest haplotype in the Bermudan dataset.
Phylogenetic analysis of the combined Floridian/Bermu- dan dataset (Fig. 2b) produced a single most parsimonious tree (11 steps; three parsimony-informative characters; CI 5 1.0; RI 5 1.0) that maintained a starlike topology and a central position for BDA1. Floridian haplotypes were peripherally positioned in a single branch of the tree topol- ogy. BDA1 was one step removed from eight western Atlantic haplotypes, including the predominant mainland lineage (FL1), and two steps away from the remaining three lineages. In contrast, FL1 was respectively, one, two and three steps removed from two, seven, and two western Atlantic lineages. Heuristic outgroup weighting of the hap- lotypes in the combined tree topology (Fig. 2b) identified BDA1 (0.440) as the oldest haplotype in the western North Atlantic dataset.
Caribbean outgroup populations were unavailable to root our western North Atlantic gene tree, so we generated homologous CO I sequence for an Australian direct-devel- oping polyploid Lasaea lineage (GenBank# AF153064) that is sister to western Atlantic congeners in global phyloge-
´ netic trees (O Foighil and Jozefowicz, 1999; Taylor and
O´ Foighil, 2000). The Australian haplotype differed by
4.4% from the western Atlantic lineages, and its utiliza- tion as an outgroup taxon produced a single most parsimo- nious tree (not shown; 39 steps, CI 5 0.667, RI 5 0.555) in which the ingroup topology (11 steps total length) was identical to that of Figure 2b and the outgroup taxon con- nected to BDA7, a peripheral Bermudan haplotype (Fig. 2b), by a relatively enormous branch length of 28 steps. The basal positioning of BDA7 resulted from it sharing a thy- midine at position 344 with the outgroup, whereas other western Atlantic lineages had either an adenine or guanine residue (Table 1).
The COI data set corroborates O´ Foighil and Jozefo- wicz’s (1999) preliminary mt 16S characterization of Flo- ridian and Bermudan Lasaea populations: they are geneti- cally very similar, yet distinct, and genetic diversity levels
are higher in the oceanic island samples. Surprisingly, the parsimony network generated was a mirror image of a priori source and founder topological expectations; conti- nental populations contained a small number of endemic genotypes that positioned peripherally in the western North Atlantic Lasaea network (Fig. 2b). Rooting of this network is necessary to test dispersal hypotheses for these two re- gional populations; however, rooting is complicated by the unavailability of Caribbean outgroup genotypes.
Two rooting procedures were used: phylogenetic analy- ses utilizing an Australian sister lineage, and calculation of relative outgroup weights for individual haplotypes based on neutral coalescence expectations. BDA7, a relatively rare Bermudan haplotype, was identified as the root of the west- ern North Atlantic clade when an Australian sister taxon was utilized as an outgroup. This result should be viewed with caution because of the 28:1 ratio of outgroup-to-in- group branch lengths in the analysis and because of the peripheral positioning of BDA7 in unrooted analyses (Fig. 2b). Closely related species frequently fail to accurately root intraspecific trees if intraspecific differences are much less than interspecific differences (Templeton, 1992), and co- alescence theory predicts that rare haplotypes occupying cladogram tips are likely to be recently derived lineages (Donnelly and Tavare´, 1986).
According to coalescence theory, the probability that a given haplotype is the oldest in a population is a function of its frequency and the centrality of its positioning in topo- logical networks (Donnelly and Tavare´, 1986; Crandall and Templeton, 1993, 1996; Beckenbach, 1994). Castelloe and Templeton’s (1994) heuristic method weights interior hap- lotypes more heavily than tip haplotypes and differentially weights interior haplotypes by their mutational connected- ness and their proximity to high-multiplicity haplotypes. It unambiguously identified BDA1, the centrally positioned (Fig. 2a) and most common Bermudan haplotype (Table 2), as the most likely candidate for the oldest Bermudan mito- chondrial genotype (outgroup weight 5 0.445) in our sam- ple. BDA7 is two steps away from the central topological position (Fig. 2a), received a much lower outgroup weight- ing (0.008), and cannot be accepted as the putative ancestral Bermudan lineage.
Haplotype BDA1 maintained its topological centrality and its high (0.44) outgroup weighting (Castelloe and Templeton, 1994) in combined analyses of Bermudan and Floridian lineages (Fig. 2b). Although the mainland haplo- type, FL1, is by far the most common haplotype in the combined data set, its peripheral positioning in the phylo- gram and lack of connectedness to most of the other western North Atlantic haplotypes give it a relatively low outgroup weighting (0.297). Surprisingly, our analyses based on co- alescence theory expectations indicate that the Bermudan lineage BDA1 is the oldest in our data set and represents the