Multiple nuclear gene phylogenetic analysis of the evolution of dioecy and sex chromosomes in the genus Silene

Abstract

In the plant genus Silene, separate sexes and sex chromosomes are believed to have evolved twice. Silene species that are wholly or largely hermaphroditic are assumed to represent the ancestral state from which dioecy evolved. This assumption is important for choice of outgroup species for inferring the genetic and chromosomal changes involved in the evolution of dioecy, but is mainly based on data from a single locus (ITS). To establish the order of events more clearly, and inform outgroup choice, we therefore carried out (i) multi-nuclear-gene phylogenetic analyses of 14 Silene species (including 7 hermaphrodite or gynodioecious species), representing species from both Silene clades with dioecious members, plus a more distantly related outgroup, and (ii) a BayesTraits character analysis of the evolution of dioecy. We confirm two origins of dioecy within this genus in agreement with recent work on comparing sex chromosomes from both clades with dioecious species. We conclude that sex chromosomes evolved after the origin of Silene and within a clade that includes only S. latifolia and its closest relatives. We estimate that sex chromosomes emerged soon after the split with the ancestor of S. viscosa, the probable closest non-dioecious S. latifolia relative among the species included in our study.

Figure 1. Phylogeny of the Silene species studied.

Only bootstrap values >50% are shown. The breeding systems of the species are indicated in the figure. The outgroup for rooting the tree for the genus Silene was usually P. hispanica (see Tables S2 and S3 for details). For the sex-linked genes, only X sequences were included. A) Maximum likelihood tree obtained with PhyML using the concatenated alignment with 12 genes (LIP21 and the outgroup sequences for 2A10, clpP3, ELF were excluded by Concaterpillar). B) Dataset is as in A, but tree obtained by combining all 12 gene trees using the SDM method. C) Consensus tree built from the trees in A and B. Labels 1, 2 and 3 in the trees indicate the nodes used in the BayesTraits analysis (see main text).

Figure 2. Concatenate tree and supertree for sex-linked genes.

A) PhyML tree (GTR with gamma distribution) for the concatenated alignments of the three sex-linked genes. Bootstrap values >50% are shown. B) SDM tree from the three individual sex-linked gene trees.

Figure 3. Divergence between sex chromosomes and between Silene species.

Distribution of synonymous divergence values (mean d _S values, estimated using PAML) from pairwise comparisons between S. latifolia and the other species for the 13 genes. One such comparison is for S. viscosa, and the others are for the following groups of species: Dioecious 1 = S. dioica, S. diclinis, S. marizii, S. heuffelii, Dioecious 2 = S. otites, S. acaulis, S. nutans (although, as noted in the text, most S. nutans populations are gynodioecious), Others = S. vulgaris, S. conica, S. noctiflora, Lychnis and outgroups = Lychnis flos-jovis, Lychnis coronaria and outgroups. In addition, vertical arrows indicate the average d _S between the S. latifolia X and Y sequences for SlXY4 and SlXY7, which belong to the oldest stratum of X-Y divergence (symbolised by XY₁), and for SlCyp-XY (labelled XY₂), a gene which belongs to the “intermediate stratum” that stopped recombining considerably later than the first two genes , , . For each panel in Figure 3, we performed a sign-test (see Methods) of whether either the XY₁ or the XY₂ d _S value is significantly different from the mean in the panel (e.g. for the top panel, we tested S. latifolia vs. Dioecious 1). Significant differences are indicated as follows: * p<10⁻¹, ** p<10⁻², *** p<10⁻³; ns = non-significant differences.