Extensive gene duplication in Arabidopsis revealed by pseudo-heterozygosity

Abstract

BackgroundIt is becoming apparent that genomes harbor massive amounts of structural variation, and that this variation has largely gone undetected for technical reasons. In addition to being inherently interesting, structural variation can cause artifacts when short-read sequencing data are mapped to a reference genome. In particular, spurious SNPs (that do not show Mendelian segregation) may result from mapping of reads to duplicated regions. Recalling SNP using the raw reads of the 1001 Arabidopsis Genomes Project we identified 3.3 million heterozygous SNPs (44% of total). Given that Arabidopsis thaliana (A. thaliana) is highly selfing, we hypothesized that these SNPs reflected cryptic copy number variation, and investigated them further.ResultsWhile genuine heterozygosity should occur in tracts within individuals, heterozygosity at a particular locus is instead shared across individuals in a manner that strongly suggests it reflects segregating duplications rather than actual heterozygosity. Focusing on pseudo-heterozygosity in annotated genes, we used GWAS to map the position of the duplicates, identifying 2500 putatively duplicated genes. The results were validated using de novo genome assemblies from six lines. Specific examples included an annotated gene and nearby transposon that, in fact, transpose together.ConclusionsOur study confirms that most heterozygous SNPs calls in A. thaliana are artifacts, and suggest that great caution is needed when analysing SNP data from short-read sequencing. The finding that 10% of annotated genes are copy-number variables, and the realization that neither gene- nor transposon-annotation necessarily tells us what is actually mobile in the genome suggest that future analyses based on independently assembled genomes will be very informative.

Population genomic evidence of selection on structural variants in a natural hybrid zone

“…An additional 3842 SVs from the mate-pair data set were initially included in the nanopore data set but were removed during filtering because all individuals from one of the two parental species were heterozygotes. Most of the SVs with this pattern of excessive heterozygosity were deletions or insertions, consistent with expectations for pseudo-heterozygosity caused by errors in mapping transposable elements to the reference genome (Jaegle et al 2021). The proportion of SVs exhibiting excessive heterozygosity from the mate-pair sequencing data set (~32%) was much higher than for the nanopore sequence data set (~12.7%).…”

“…Importantly, Sniffles filters false SV signals by considering both minimum read support as well as consistency of the breakpoint position and size. Because loci with excessive htererozygosity are likely caused by erroneous mapping of transposable elements to the reference genome (Jaegle et al 2021), we excluded SVs with excessive htererozygosity, defined for this purpose as all individuals in either species initially being called heterozygotes because they had reads supporting the reference and SV allele. Additionally, we excluded SVs with sequence data for less than 20% of the individuals, which left a total of 290,276 SVs for downstream analysis.…”

Pushing the limits of HiFi assemblies reveals centromere diversity between twoArabidopsis thalianagenomes

“…To assemble contigs with the CLR dataset, we used Canu with a maximum input coverage of 200x, only using subreads larger than 10 kb, and polished the resulting assembly with Arrow (34), also using 200x of the initial long-reads. The resulting contigs had an NG50 of 14.82 Mb, which is on a par with the best published Arabidopsis thaliana CLR contigs (12–17).…”

“…To assemble contigs with the CLR dataset, we used Canu with a maximum input coverage of 200x, only using subreads larger than 10 kb, and polished the resulting assembly with Arrow (34), also using 200x of the initial long-reads. The resulting contigs had an NG50 of 14.82 Mb, which is on a par with the best published Arabidopsis thaliana CLR contigs (12)(13)(14)(15)(16)(17). With the HiFi dataset, we compared the performance of five different assemblers: FALCON (23), HiCanu (30), Hifiasm (31), Peregrine (32), and Pacbio's Improved Phased Assembler (IPA; (33)).…”

“…Notwithstanding their high error rate, long-read sequencing technologies, such as Oxford Nanopore Technologies (ONT) sequencing (reviewed in (5,6)) and PacBio's single-molecule real-time (SMRT) in the original continuous long read (CLR) sequencing mode (7), have significantly improved the contiguity of de novo assemblies. To date, there are 16 A. thaliana accessions sequenced with CLR technology (8)(9)(10)(11)(12)(13)(14)(15)(16)(17), and although these assemblies commonly achieved some chromosome-arm level contigs, they invariably stop short of assembling through centromeric and pericentromeric regions. Only very recently, it was possible to assemble gapless centromeres in the A. thaliana reference accession Col-0, primarily with ultra-long ONT reads and the addition of PacBio high-fidelity (HiFi) reads for gap closing and polishing (18).…”