DNA/DNA hybridization studies of carnivorous marsupials. III. Relationships among species ofDidelphis (Didelphidae)

We report three sets of DNA hybridization experiments conducted to determine relationships among species ofDidelphis (D. albiventris, D. marsupialis, D. virginiana). The 1989 and 1991 sets had fewer replicates per cell than the 1990 series (3.4 and 5.4 vs 9), but in 1991 we distinguished two populations ofD. marsupialis and utilized several individuals for each heterologous comparison. BothPhilander opossum andLutreolina crassicaudata were used as outgroups in 1989, but onlyLutreolina was included in subsequent sets. For each set, we calculated all four standard indices of thermal stability (T _mode,T _m,T ₅₀ H, and NPH) and constructed trees by least-squares (FITCH) and neighbor-joining methods, both before and after correction for asymmetric reciprocal cell values. Subsets of the 1989 data lacking eitherPhilander orLutreolina were analyzed similarly. To explore measurement imprecision, the corrected and uncorrected matrices for each of the four indices were bootstrapped 100 times for the 1989 set and subsets and 1000 times for the 1990 and 1991 sets. Again, for the 1990 and 1991 data, an additional 100 bootstrapped distances were fitted to user trees representing the three possible pairings ofDidelphis spp. to determine the significance of the FITCH branch lengths. The successive experimental sets generated increasingly consistent evidence for pairingD. marsupialis withD. albiventris. The 1989 experiments involved just 85 comparisons, and only T _mode's pairedD. marsupialis withD. albiventris (at bootstrap percentages of 70% or above), but did so whetherPhilander orLutreolina or both were included as the outgroup(s). FITCH and neighbor-joining trees had identical topologies for T _mode's but sometimes differed for the other measures. In contrast, all but one of the corresponding FITCH and neighborjoining trees matched for the 1990 and 1991 data, and three of the four distance measures (T _mode, T _m, and T ₅₀ H) unitedD. marsupialis withD. albiventris at bootstrap percentages averaging 81%; NPH gave a different result for 1990, associatingD. marsupialis withD. virginiana. Further, all but 2 of the 16 matrices for 1990 and 1991 gave mean bootstrapped branch lengths for a consensus pairing that were positive and at least one standard deviation from zero, despite the very short internodes recovered. These results illustrate that the potential of DNA hybridization for resolving very close relationships depends on both the index and the experimental design employed. We conclude that of the three species,D. albiventris andD. marsupialis shared a more recent common ancestor and estimate thatDidelphis spp. have diverged at about 0.39% in nucleotide sequence per myr.Deceased.