Search

International audience; Centromeres are critical for cell division, loading CENH3 or CENPA histone variant nucleosomes, directing kinetochore formation and allowing chromosome segregation(1,2). Despite their conserved function, centromere size and structure are diverse across species. To understand this centromere paradox(3,4), it is necessary to know how centromeric diversity is generated and whether it reflects ancient trans-species variation or, instead, rapid post-speciation divergence. To address these questions, we assembled 346 centromeres from 66 Arabidopsis thaliana and 2 Arabidopsis lyrata accessions, which exhibited a remarkable degree of intra- and inter-species diversity. A. thaliana centromere repeat arrays are embedded in linkage blocks, despite ongoing internal satellite turnover, consistent with roles for unidirectional gene conversion or unequal crossover between sister chromatids in sequence diversification. Additionally, centrophilic ATHILA transposons have recently invaded the satellite arrays. To counter ATHILA invasion, chromosome-specific bursts of satellite homogenization generate higher-order repeats and purge transposons, in line with cycles of repeat evolution. Centromeric sequence changes are even more extreme in comparison between A. thaliana and A. lyrata. Together, our findings identify rapid cycles of transposon invasion and purging through satellite homogenization, which drive centromere evolution and ultimately contribute to speciation.

Additional file 1: Fig S1. Distribution of heterozygosity across individuals and the genome. Fig S2. Tract-length distribution. Fig S3. SNP filtering scheme. The SNPs matrix we started with contains 7 million SNPs. Of those, 3.3 million were detected as heterozygous in at least one line. We selected the 48,799 SNPs that we called heterozygous in at least 5 % of the lines, and focused on the 26,647 found in coding regions (according to the TAIR10 annotation). Fig S4. The distribution of distances between the “position” of a pseudo-SNPs and the corresponding GWAS peak. Out of the 411688 peaks we found that 236494 (57%) are on a different chromosome. The plots show the distribution of distances for the 175194 peaks that were on the same chromosome as the corresponding pseudo-SNPs. Roughly 50% of there are more distance than 50kb. Fig S5. Gene ontology analysis of the putatively duplicated genes. Fig S6. Pipeline to confirm GWAS peaks. Fig S7. Mapping reads tagging singleton pseudo-SNP. First, all reads overlapping the position of a specific pseudo-SNP were extracted based on mapping to the reference genome (TAIR10). This set of reads were then re-mapped to the appropriate Pacbio genome. Reads mapping to multiple regions indicate the presence of a duplicated segment. A decrease in coverage compared to the mapping to the reference genome is also a confirmation that reads map at different positions. An example is presented in Fig S9. Fig S8. Comparison of mapping coverage between reference genome and PacBio genomes for all regions surrounding pseudo-heterozygous doubleton positions. Fig S9. Mapping of reads overlapping singletons (example). Fig S10. Position of AT1G31390 in the reference (accession 6909) and each of the six newly assembled genomes. BLAST-thresholds of 70% identity were used, and only fragments of length greater than 50% of the original gene length are shown. Fig S11. Position of AT1G20400 in the reference (accession 6909) and each of the six newly assembled genomes. Cf. Fig S10. Fig S12. Dot-plots of the end of insertion B. LTR repeats can be detected on each side of the insertion. Cf. Fig S13. Fig S13. Dot plot of the ends of all insertions. Cf. Fig S12. Fig S14. Methylation profile across all copies of the TE+gene insertions. Bisulfite reads were mapped to the appropriate genomes and profiles extracted based on the inferred locations. Colors distinguish the different insertions found in each genome. Fig S15. Frequency of insertions of the new element in the 1001 Genomes. Fig S16. Mapping of the AT1G20400 region in multiple species. The rectangle corresponds to annotated genes around AT1G20400 in the A. thaliana reference genome. Fig S17. Chromosomal position of putatively duplicated genes that could not be confirmed using full genome sequence. The figure shows the distribution of these genes compared to the distribution of all genes, and RepeatMasker annotation. Fig S18. Illustration of GWAS peak calling. The p-values from GWAS were used to run the filterFFT function from the NucleR package in R, using 0.05 for pcKeepComp option. Fig S19. The relationship between the number of pseudo-SNPs and sequencing coverage. Each dot represents an accession.

Background It is apparent that genomes harbor much structural variation that is largely undetected for technical reasons. Such variation can cause artifacts when short-read sequencing data are mapped to a reference genome. Spurious SNPs may result from mapping of reads to unrecognized duplicated regions. Calling SNP using the raw reads of the 1001 Arabidopsis Genomes Project we identified 3.3 million (44%) heterozygous SNPs. Given that Arabidopsis thaliana (A. thaliana) is highly selfing, and that extensively heterozygous individuals have been removed, we hypothesize that these SNPs reflected cryptic copy number variation. Results The heterozygosity we observe consists of particular SNPs being heterozygous across individuals in a manner that strongly suggests it reflects shared segregating duplications rather than random tracts of residual heterozygosity due to occasional outcrossing. Focusing on such pseudo-heterozygosity in annotated genes, we use genome-wide association to map the position of the duplicates. We identify 2500 putatively duplicated genes and validate them using de novo genome assemblies from six lines. Specific examples included an annotated gene and nearby transposon that transpose together. We also demonstrate that cryptic structural variation produces highly inaccurate estimates of DNA methylation polymorphism. Conclusions Our study confirms that most heterozygous SNP calls in A. thaliana are artifacts and suggest that great caution is needed when analyzing SNP data from short-read sequencing. The finding that 10% of annotated genes exhibit copy-number variation, and the realization that neither gene- nor transposon-annotation necessarily tells us what is actually mobile in the genome suggests that future analyses based on independently assembled genomes will be very informative.