Did RNA editing in plant organellar genomes originate under natural selection or through genetic drift?

Abstract

Background: The C<-->U substitution types of RNA editing have been observed frequently in organellar genomes of land plants. Although various attempts have been made to explain why such a seemingly inefficient genetic mechanism would have evolved, no satisfactory explanation exists in our view. In this study, we examined editing patterns in chloroplast genomes of the hornwort Anthoceros formosae and the fern Adiantum capillus-veneris and in mitochondrial genomes of the angiosperms Arabidopsis thaliana, Beta vulgaris and Oryza sativa, to gain an understanding of the question of how RNA editing originated.

Results: We found that 1) most editing sites were distributed at the 2nd and 1st codon positions, 2) editing affected codons that resulted in larger hydrophobicity and molecular size changes much more frequently than those with little change involved, 3) editing uniformly increased protein hydrophobicity, 4) editing occurred more frequently in ancestrally T-rich sequences, which were more abundant in genes encoding membrane-bound proteins with many hydrophobic amino acids than in genes encoding soluble proteins, and 5) editing occurred most often in genes found to be under strong selective constraint.

Conclusion: These analyses show that editing mostly affects functionally important and evolutionarily conserved codon positions, codons and genes encoding membrane-bound proteins. In particular, abundance of RNA editing in plant organellar genomes may be associated with disproportionately large percentages of genes in these two genomes that encode membrane-bound proteins, which are rich in hydrophobic amino acids and selectively constrained. These data support a hypothesis that natural selection imposed by protein functional constraints has contributed to selective fixation of certain editing sites and maintenance of the editing activity in plant organelles over a period of more than four hundred millions years. The retention of genes encoding RNA editing activity may be driven by forces that shape nucleotide composition equilibrium in two organellar genomes of these plants. Nevertheless, the causes of lineage-specific occurrence of a large portion of RNA editing sites remain to be determined.

Figure 1

Editing frequency {(A [number of edited codons]/B [number of codons in the genome]) × 1000} versus Grantham index[ 47 ]for all edited codons in protein coding genes in chloroplasts of A. formosae (a, forward editing; b, reverse editing), and A. capillus-veneris (c, forward and reverse editing). Amino acid residues are shown as pre-edited to edited below all codons. The vertical axis indicates Grantham index value and editing frequency. Edited codons are grouped into nonsynonymous 1^st, 2^nd, and silent site changes. Asterisks indicate arbitrarily assigned values of 200 for Grantham index values in premature stop codons. The dotted line denotes an upper limit of editing frequency across silent editing sites. The values obtained for A/B were multiplied by 1000 to make these data plottable against Grantham's values.

Figure 2

Editing frequency {(A [number of edited codons]/B [number of codons in the genome]) × 1000} versus Grantham index[ 47 ]for all edited codons in protein coding genes in mitochondria of A. thaliana (a), B. vulgaris (b), and O. sativa (c). Amino acid residues are shown as pre-edited to edited below all codons. The vertical axis indicates Grantham index value and editing frequency. Edited codons are grouped into nonsynonymous 1^st, 2^nd, and silent site changes. The dotted line denotes an upper limit of editing frequency across silent editing sites. The values obtained for A/B were multiplied by 1000 to make these data plottable against Grantham's values.

Figure 3

Frequencies of amino acids in membrane-bound and soluble proteins in the chloroplast of A. formosae (a) and the mitochondrion of B. vulgaris (b). The "n" indicates the number of genes in each class. Plots i contain membrane-bound protein genes. Plots ii contain genes encoding soluble proteins. Amino acid residues are ordered from the most hydrophobic (left) to the most hydrophilic (right) according to the Kyte-Doolittle hydrophobicity scale [48].

Figure 4

The correlation of editing frequency and T-A distance at the 2^nd and 1^st codon positions in chloroplast genomes of A. formosae (a) and A. capillus-veneris (b, c). Membrane-bound protein coding genes are indicated by red dots, while soluble protein coding genes are presented as blue triangles. Regression results are significant for all presented analyses (p < 0.01) with details shown in each figure. Values in brackets indicate data points not included in the analyses due to excessive deviation from the mean (> 3 standard deviations). For clarity, the data of Figure 4c were re-analyzed using only the 19 genes containing reverse editing. However, the significant result obtained using all genes are presented in Additional file 1.

Figure 5

The correlation of editing frequency and T-A distance at the 2^nd and 1^st codon positions in the mitochondrial genome of B. vulgaris (a, b). Membrane-bound protein coding genes are indicated by red dots, while soluble protein coding genes are presented as blue triangles. Regression results are significant for all presented analyses (p < 0.01) with details shown in each figure.

Figure 6

The correlation of total editing frequency and gene specific rates of molecular evolution at nonsynonymous (d_N) (a, c) and synonymous (d_S) (b, d) sites in chloroplast genomes of A. formosae and A. capillus-veneris (a, b), and mitochondrial genomes of A. thaliana, B. vulgaris, and O. sativa (c, d). Solid points (red) signify membrane-bound proteins, while open points (blue) represent soluble proteins. Data for A. formosae are shown as triangles, while A. capillus-veneris is presented as circles (a, b). Data for A. thaliana, B. vulgaris, and O. sativa are presented as squares, triangles, and circles respectively (c, d).

Figure 7

Nucleotide frequencies across streptophytes for whole genomes of chloroplasts (a) and mitochondria (b), and combined codon positions for the conserved proteomes of chloroplasts (c) (57 genes) and mitochondria (d) (13 genes). Nucleotides are represented by filled circles for T, open circles for A, open triangles for G, and filled triangles for C. The convergence of nucleotide composition can be observed across whole genomes and conserved proteomes. Taxa with available RNA editing data are duplicated and designated as "edit-repaired" by an asterisk. Taxa are arranged according to a phylogenetic sequence of [46] and divergence times are not considered. Broad taxonomic groupings are indicated on the X-axis below brackets. Taxonomic abbreviations are described in Additional file 3.