Evolution of sequence-specific anti-silencing systems in Arabidopsis

Abstract

The arms race between parasitic sequences and their hosts is a major driving force for evolution of gene control systems. Since transposable elements (TEs) are potentially deleterious, eukaryotes silence them by epigenetic mechanisms such as DNA methylation. Little is known about how TEs counteract silencing to propagate during evolution. Here, we report behavior of sequence-specific anti-silencing proteins used by Arabidopsis TEs and evolution of those proteins and their target sequences. We show that VANC, a TE-encoded anti-silencing protein, induces extensive DNA methylation loss throughout TEs. Related VANC proteins have evolved to hypomethylate TEs of completely different spectra. Targets for VANC proteins often form tandem repeats, which vary considerably between related TEs. We propose that evolution of VANC proteins and their targets allow propagation of TEs while causing minimal host damage. Our findings provide insight into the evolutionary dynamics of these apparently "selfish" sequences. They also provide potential tools to edit epigenomes in a sequence-specific manner.

Fig. 1

Sequence-specific hypomethylation by VANC proteins. a Schematic diagram of structures of VANDAL transposons and the modified transgenes used. Boxes indicate exons. Vertical red lines show the positions of two nonsense mutations in VANC21 (Supplementary Fig. 2). b DNA methylation levels of VANDAL21 copies in ∆AB transgenic plants and parental wild-type plants (WT). Broken lines show TE ends. Each point represents proportion of methylated cytosine for a sliding window with seven fractions after separating each TE for 100 fractions. Right and left flanking regions are also analyzed by the same conditions. c, d Comparison of DNA hypomethylation between full-length Hi and ∆AB transgenic plants for CHG (c) and CHH sites (d). VANDAL21 copies are colored red. e, f Hypomethylation effects of VANC6 transgene. In e, conditions are as in b. In f, DNA hypomethylation is shown for each TE at CHG sites and CHH sites. VANDAL6 copies and VANDAL8 copies are colored green and gray, respectively. g Comparison of DNA hypomethylation between ∆AB and VANC6 transgenic plants at CHG sites. Results at CHH sites are shown in Supplementary Fig. 1f–h. In the panels c, d, f, and g, TEs more than 1 kb long are plotted (N = 5866). The significance of decrease in DNA methylation was assessed by the value $\left(Mn∕Cn-Mt∕Ct\right)∕\left(1∕\sqrt{Cn}+1∕\sqrt{Ct}\right)$, where Mn, Cn, Mt, and Ct are methylated cytosine (M) and total cytosine (C) counts mapped for each TE in the non-transgenic (n) and transgenic (t) plants, respectively. This value shows the significance by weighing the change in the methylation ratio with root of the count number. Effects of ∆AB and VANC6 transgenes on DNA methylation status of TEs longer than 1 kb are also shown in Supplementary Data 1

Fig. 2

Genomic localization of the VANC21 protein. a A genome-wide view showing the enrichment of FLAG-VANC21 signal. Each dot represents signal in a 10 kb region. Red dots indicate the regions with VANDAL21 copies more than 1 kb long. b Genome browser views showing the FLAG-VANC21 signals with normalized coverages (per million mapped reads) and DNA methylation profiles (0–100%) of WT and ∆AB transgenic plants at VANDAL21 copies. Each point represents proportion of methylated cytosine counted within five successive cytosine residues. VANC21 exons are colored red. Arrowheads indicate probe sites used in EMSA (Fig. 3). c–e Three contexts of DNA methylation level around VANC21 biding loci. Around the summits of VANC21-binding loci within VANDAL21 (N = 89), 500 bp-binned averages of DNA methylation profiles were plotted for WT and ∆AB transgenic plants with steps of 50 bp

Fig. 3

VANC21-binding motifs and their distribution among VANDAL families. a DNA sequence motif most commonly found at FLAG-VANC21-binding sites in VANDAL21 TEs. Localizations of the motif are shown in Fig. 2b and Supplementary Fig. 6, with green and orange bars indicating positions of the C-type (“YAGTATTAC”) and T-type (“YAGTATTAT”) motifs, respectively. b Electrophoretic mobility shift assay (EMSA) by VANC21 protein for double-stranded DNA of the sequences shown in c. The probe sites are also shown by white arrowheads in Fig. 2b. 1*, 2*, and 3* have single-base substitutions within the motif relative to the original sequences. 4 and 5 are control sequences from the exons, where VANC21 localization was not detected. Results for dependence of the shift to the protein amount and a competition assay are shown in Supplementary Fig. 7. c Sequences of the dsDNA probe used in EMSA (b). d Numbers of C- and T-type motifs within VANDAL21 and related VANDAL family members within the genomes of A. thaliana and A. lyrata. A. lyrata-specific lineages are shown with red lines. IDs for these TE copies and bootstrap probabilities are shown in Supplementary Fig. 8c

Fig. 4

Evolution of VANC21-binding regions. a Schematic diagram of dot plot for sequences with tandem repeat structures. Blue arrows indicate tandem repeats. Because of sequence identity, tandem repeat structure generates parallel lines. Difference in copy number can also be detected; in this case, three and two repetitions for X- and Y-axes, respectively. b Structures of VANDAL21 copies, with regions analyzed in c–f shown with shadow. Green and orange bars show C-type and T-type motifs. c–f Dot plots comparing the VANDAL21 sequences upstream of VANA21 (c), VANB21 intron (d), and VANC21 intron (e, f). Regions with 10 bp exact match are shown by dots. Green and orange indicate regions with C- and T-type motifs, respectively. Sequence alignments in these tandem repeat regions are shown in Supplementary Fig. 9

Fig. 5

Dot plot analyses of the VANC proteins. a Comparison of amino-acid sequences of the VANC proteins. Homologous regions were plotted with dotmatcher program (window size: 10, threshold: 23). Amino-acid sequences (Nʹ–Cʹ) were ordered from top to bottom and left to right. Scale bar for 200 a.a. was shown in the right of plots. Two domains (DUF1985 and DUF287) are shown by shaded and filled areas, respectively. VANA (transposase), another protein encoded in these VANDAL members, is much more conserved (Supplementary Fig. 10). b Comparison of the nucleotide sequences of VANC genes. VANC genes were plotted with YASS program under default parameters. DNA sequences are ordered from top to bottom and left to right. Gray boxes indicate exons. Divergence between these VANDAL families as well as divergence between the VANDAL copies in A. thaliana and A. lyrata is shown in Supplementary Tables 3, 4 and Supplementary Fig. 14