Phylogenetic Relationships among Populations of Northern Swordtails (Xiphophorus) as Inferred from Allozyme Data

Twenty-nine populations of Xiphophorus fishes representing nine species of northern swordtails, one southern swordtail and a platyfish were assayed electrophoretically for allozyme variation. Phylogenetic relationships were inferred using parsimony and likelihood analysis of gene frequency characters, as well as Fitch-Margoliash, minimum evolution and neighbor-joining analyses of genetic distances. The phylogenetic relationships among species that were well supported in all analyses included (1) monophyly of the northern swordtails, (2) the pygmaeus clade of Xiphophorus nigrensis, X. multilineatus, and X. pygmaeus, and (3) the clade of X. nigrensis and X. multilineatus. Of those species represented by more than one population, all analyses supported monophyly of X. montezumae and weakly supported monophyly of X. nezahualcoyotl and X. birchmanni. Only the distance analyses supported monophyly of X. cortezi, and the support was weak. Finally, all analyses supported a clade including X. nezahualcoyotl from the Río Tamesí drainage and some populations from the Río Pánuco drainage, that is, nonmonophyly of the set of populations from the Río Pánuco drainage. Previously published trees based on morphology, behavior and randomly amplified DNAs were generally congruent with the optimal trees for the allozyme data and were not rejected by those data; in contrast, trees based on DNA sequences were more incongruent with the optimal trees for the allozyme data and were rejected by those data.

Fig. 1.

Map of the Río Pánuco drainage system showing most major drainages. Localities are marked with D. C. Morizot's collection numbers (see Appendix 1)

Fig. 2.

Strict consensus tree of 60 shortest trees found using MANOB on the full dataset with bootstrap values for the various nodes. Length of optimal trees equals 61.96. Gray branches are supported by bootstrap values of < 50%

Fig. 3.

Strict consensus tree of 10 shortest trees found using MANOB on the reduced dataset with bootstrap values for the various nodes. Length of optimal trees equals 53.625

Fig. 4.

The single best tree from continuous character maximum likelihood analysis on the reduced dataset. Ln-likelihood equals 1167.69

Fig. 5.

The single best tree from Fitch-Margoliash analysis of the reduced dataset converted to Nei genetic distances with bootstrap values for the various nodes. Sum of squares equals 3.74291

Fig. 6.

The single best tree produced by the minimum evolution analysis of the reduced dataset converted to Nei genetic distances. S = 1.80852

Fig. 7.

Strict and semistrict consensus trees of the optimal trees for the allozyme data based on parsimony, likelihood, Fitch-Margoliash, neighbor-joining and minimum evolution methods. The tree as illustrated is the semistrict consensus tree; the branch in gray is missing from the strict consensus tree

Accepted: 29 Jun 2000