Independent and combined analyses of sequences from all three genomic compartments converge on the root of flowering plant phylogeny

Abstract

Plant phylogenetic estimates are most likely to be reliable when congruent evidence is obtained independently from the mitochondrial, plastid, and nuclear genomes with all methods of analysis. Here, results are presented from separate and combined genomic analyses of new and previously published data, including six and nine genes (8, 911 bp and 12,010 bp, respectively) for different subsets of taxa that suggest Amborella + Nymphaeales (water lilies) are the first-branching angiosperm lineage. Before and after tree-independent noise reduction, most individual genomic compartments and methods of analysis estimated the Amborella + Nymphaeales basal topology with high support. Previous phylogenetic estimates placing Amborella alone as the first extant angiosperm branch may have been misled because of a series of specific problems with paralogy, suboptimal outgroups, long-branch taxa, and method dependence. Ancestral character state reconstructions differ between the two topologies and affect inferences about the features of early angiosperms.

Figure 1

(A) Four-color traces for base pairs 113–121 of two clones and total PCR product for Amborella atpA. Top shows sequence obtained from 18 clones of the total PCR product shown in Middle. Bottom shows sequence from one clone obtained from the total PCR product. (B) NJ analyses clearly revealed that the sequence in Bottom of A is orthologous and groups with the basal-most angiosperms, whereas the sequence in Top of A was likely a paralog of all other basal angiosperm atpA sequences, grouping here with the asterid lineage.

Figure 2

(A) Optimal outgroup analyses (43) using the six-gene dataset indicate that Ginkgo and Pinus are the best species to use to identify the root node of angiosperm phylogeny. (B) Taxon-variance ratio plot showing that Gnetales (Gnetum and Welwitschia), Podocarpus, and Metasequoia are outliers relative to all other sampled taxa. Taxon-variance outliers have been shown to lead to inconsistent tree estimation when included in phylogenetic analyses (44).

Figure 3

NJ tree estimated with p distances calculated from the noise-reduced dataset of six combined genes. Bootstrap support (BP) is given for nodes >50 and major lineages are labeled. The topology is generally consistent with estimates obtained from raw data and other datasets or methods, except for the first branch of Amborella + Nymphaeales (see text).

Figure 4

(A) Alternative roots (marked by *) for basal angiosperm phylogeny based on six-gene analyses. (B) Comparisons of BP for estimates of roots A and B in A, using multiple methods of analysis from the six-gene combined dataset before and after noise reduction. (C) Comparison of BP for both NJ-P and UW parsimony estimates of roots A and B by using individual genomic compartments from the six-gene dataset before and after noise reduction (number of shared variable sites is given in bp).

Figure 5

(A) NJ-P tree estimated by using the nine-gene combined noise-reduced dataset. (B) Comparison of method-dependent bootstrap support (BP) for the two alternative roots of basal angiosperm phylogeny (see Fig. 4A) the nine-gene combined dataset before and after noise reduction was used. (C) Comparison of bootstrap support for the two roots when independent genomic compartments before and after noise reduction were used.

Figure 6

Bootstrap power curve showing BP support for root A (Fig. 4A) when NJ and UW are used on both the raw and noise-reduced nine-gene dataset. Various proportions of the nine-gene dataset were resampled (10–1000% of the original data) to investigate the effect of numbers of characters on BP for root A (Fig. 4A). Note that if 10 times (1000%) the amount of data (of the same quality) were sampled, NJ and UW (using noise-reduced data) would unambiguously support root A, whereas UW (using raw data) would provide no support for this root.