Circumscription and biogeographic patterns in the eastern North American-east Asian genus Stewartia (Theaceae: Stewartieae): Insight from choloroplast and nuclear DNA sequence data

Castanea, Sep 2002 by Prince, Linda M

rpl16 Intron

MP analyses of the rp116 intron data matrix produced 3 shortest trees. The fewer characters produced far fewer trees than the ITS data matrix and higher index values suggesting fewer homoplastic characters in the intron data set. The strict consensus tree is shown in Figure 2B with bootstrap values indicated. This data set found strong support (100%) for the monophyly of Stewartieae and also identified a strongly supported (100%) Stewartia gemmata S. monadelpha S. serrata S. sinensis S. pseudocamellia S. rostrata clade. The inclusion of two partial sequences (S. malacodendron [LP124] and S. ovata [LP2001) did not appear to alter tree topology. Alignment of the rp116 intron matrix for all taxa required the inference of few insertion and deletion (INDEL) events. The inferred INDELs were virtually free of ambiguity and the inclusion or exclusion of those characters did not alter rooting or ingroup topology (data not shown).

ML analyses under the JC and GTR G I models data (see Table 3) produced a tree of -In = 2084.98239 and -In = 1960.71838 respectively. Topologies were identical to the strict consensus of the MP analyses and are not shown. The rp116 intron MP and ML analyses produced similar although less resolved relationships to the ITS ML analyses.

trnE-T Intergenic Spacer

MP analyses of the trnE-T data matrix produced a single shortest tree which is shown in Figure 2C with bootstrap values indicated. These data provided only 16 parsimony informative characters for tree building, thus limiting bootstrap support. These data provided strong support (100%) for the monophyly of Stewartieae, moderate support (84%) for a Hartia clade, and weak support (61%) for a sister relationship of S. malacodendron and S. ovata to Hartia. Alignment of the trnE-T matrix for all taxa required the inferrence of few INDEL events. The inferred INDELs were virtually free of ambiguity and the inclusion or exclusion of those characters did not alter rooting or ingroup topology (data not shown).

ML analyses under the JC and GTR G I model with all parameters estimated from the data (see Table 3) produced a trees of -In = 1646.28964 and -In =1537.22595 respectively. Tree topologies for these analyses were also identical to the strict consensus of the MP analyses and are not shown.

Combined Analyses

Partition homogeneity tests of pruned data matrices produced P-values of 0.5811 (ITS x rp116 intron) to 1.0000 (ITS X trnE-T igs) indicating no significant heterogeneity in the data partitions. The combined data set included one representative of each species (composite sequences for S. malacodendron and S. ovata) for a total of ten ingroup sequences and two outgroup sequences. MP analyses produced 2 shortest trees. The strict consensus tree is shown in Figure 2D along with bootstrap values. As expected, the combined analyses produced a more highly resolved tree with stronger bootstrap support than any of the individual data partition analyses. The only relationship not resolved in the combined analysis is whether S. ovata or S. malacodendron are sister to each other or whether one of the two is more closely related to all Hartia representatives included in this study. The ML tree (not shown) is identical to the MP tree except ML analyses produce a fully resolved tree of -In = 6242.85764 (JC) and -In = 6003.23897 (GTR G I), with S. malacodendron sister to Harta.