Polyploids broadly generate novel haplotypes from trans-specific variation in Arabidopsis arenosa and Arabidopsis lyrata

Abstract

Polyploidy, the result of whole genome duplication (WGD), is widespread across the tree of life and is often associated with speciation and adaptability. It is thought that adaptation in autopolyploids (within-species polyploids) may be facilitated by increased access to genetic variation. This variation may be sourced from gene flow with sister diploids and new access to other tetraploid lineages, as well as from increased mutational targets provided by doubled DNA content. Here, we deconstruct in detail the origins of haplotypes displaying the strongest selection signals in established, successful autopolyploids, Arabidopsis lyrata and Arabidopsis arenosa. We see strong signatures of selection in 17 genes implied in meiosis, cell cycle, and transcription across all four autotetraploid lineages present in our expanded sampling of 983 sequenced genomes. Most prominent in our results is the finding that the tetraploid-characteristic haplotypes with the most robust signals of selection were completely absent in all diploid sisters. In contrast, the fine-scaled variant 'mosaics' in the tetraploids originated from highly diverse evolutionary sources. These include widespread novel reassortments of trans-specific polymorphism from diploids, new mutations, and tetraploid-specific inter-species hybridization-a pattern that is in line with the broad-scale acquisition and reshuffling of potentially adaptive variation in tetraploids.

Fig 1. Hypotheses about allelic sources in a diploid-autotetraploid system.

A: As compared to diploids, autotetraploid lineages may acquire alleles via increased introgression potential, higher level of retainment of ancestral standing variation, and increased population-scaled mutational input. Shown are two diploid species (red), each of which gave rise to an autotetraploid lineage (blue). B: Higher variability in the possible sources of potentially beneficial alleles suggests that positively selected haplotypes in tetraploids might be more likely to form from a mosaic of sources (right), in contrast to the traditionally assumed homogeneous (single-source) scenario (left).

Fig 2. Evolutionary relationships and genomic signatures of selection in tetraploid populations of A. lyrata and A. arenosa.

A: Locations of the focal 11 A. lyrata populations (AL Austria, AL Czechia, and AL Germany) and six A. arenosa populations (AA CEur) in Central Europe. Red and blue dashed lines show the ranges of diploid and tetraploid A. lyrata, respectively; tetraploids of A. arenosa occur throughout the entire area. For the map, the base layer ‘World Imagery’ was taken from https://www.usgs.gov/products/, which is granted to be published under CC BY 4.0 license. B, C: Phylogenetic relationships among the focal populations of A. arenosa (B) and A. lyrata (C) inferred by TreeMix analysis assuming no migration. Asterisks show bootstrap support = 100%. D: Introgression among tetraploid but not diploid populations of both Arabidopsis species. ABBA-BABA statistics demonstrating excess allele sharing between tetraploids (bottom tree), but not diploids (top tree) of A. arenosa and each of the three A. lyrata tetraploid lineages. P1 is BDO, the earliest diverging and spatially isolated diploid population of A. arenosa. E: A set of 14 unlinked genes showing significant evidence (p < 0.01) of positive selection (positively selected genes, PSGs) in the four tetraploid lineages identified using PicMin. Additional three genes, HEI10, SDS, and SYN1, were identified using a screen for candidate SNPs. F: Functional characterization of the 17 PSGs by STRING analysis. The network shows predicted protein-protein interactions among the PSGs. The width of each line corresponds to the confidence of the interaction prediction. PSGs were annotated into four processes, each represented by a bubble of different color. The 12 PSGs with names written in bold had enough candidate SNPs for the reconstruction of tetraploid haplotypes (see the main text).

Fig 3. Distribution of haplotypes of the 12 positively selected genes (PSGs) involved in adaptation to WGD.

A: Populations in this analysis (504 diploids and 479 tetraploids). Arabidopsis congeners are A. halleri, A. croatica, A. cebennensis, and A. pedemontana. The map was vectorized and modified according to the base layer ‘Dark Gray canvas’, https://www.usgs.gov/products/, which is granted to be published under CC BY 4.0 license. B: Frequency of shared tetraploid (blue), A. arenosa diploid (light red), and A. lyrata diploid (red) haplotypes in each of the 12 PSGs. All other haplotypes present (including possible recombinants of the above) are shown in grey. Note the absence of tetraploid haplotypes in diploids.

Fig 4. Mosaic of allelic sources of tetraploid haplotypes corresponding to the 12 positively selected genes (PSGs) associated with adaptation to WGD.

A: The 232 candidate SNPs marking the 12 tetraploid haplotypes were categorized into one of the seven source scenarios based on their allele distribution in the 983 samples. The most parsimonious origin of each pattern is provided in italics. Four scenarios (outlined by orange frame) involve trans-specific standing variation shared among diploids of at least two Arabidopsis species (‘standing in diploids’). Six scenarios (outlined by blue frame) require introgression between tetraploid lineages. Boxes ‘min 1’ show that the tetraploid standing variation is present in at least one outgroup species. Phylogenetic relationships according to [51], Fig 4A. B: Variable source scenarios for each of the 12 PSGs. Barplots show the proportion of candidate SNPs representing each of the seven source scenarios, numbers above bars indicate the number of candidate SNPs per each PSG. Two example PSGs detailed in panel C are highlighted in bold. C: Illustration of diploid and tetraploid haplotypes for two PSGs. Bold capital letters represent haplotype-defining alleles, and two letters at a position indicate the presence of an alternative minor allele. Coding sequences are highlighted as black boxes as part of the gene model above the haplotypes (only regions overlapping with candidate SNPs are shown). The bottom line ‘allelic source’ displays the SNP assignment to its source scenario as defined in panel A. Possible recombination breakpoints required to construct the haplotype from the different allelic sources are marked with asterisks. Top: example from the CYCA2;3 endoreduplication gene, displaying 19 candidate SNPs spanning 5371 bp. Bottom: example from the PDS5b meiosis gene, depicting candidate SNPs spanning 8904 bp.