Subgenome-aware analyses suggest a reticulate allopolyploidization origin in three Papaver genomes

Abstract

Hybridization and polyploidization are important driving forces in angiosperm evolution, resulting in novel phenotypes capable of prompting ecological diversification and invasion of new niches 1 . The genus Papaver (Papaveraceae) contains many taxa used in the pharmaceutical and culinary industries and as ornamental plants 2 . Yang et al. assembled de novo two chromosome-level genomes of P. rhoeas (common poppy, 2n = 14) and P. setigerum (Troy poppy, 2n = 44), and improved the P. somniferum genome assembly (opium poppy, 2n = 22) 3 . These high-quality, chromosome-scale genome assemblies represent a valuable resource for studying the early evolutionary history of eudicots and the evolution of morphinan biosynthesis. Based on synteny and phylogenomic analyses, the authors identified two rounds of whole-genome duplication (WGD), one in the ancestor to P. setigerum and P. somniferum (WGD-1) at ~7.2 million years ago (MYA), and one lineage-specific WGD-2 in P. setigerum at ~4.0 MYA. In the absence of effective subgenome-phasing techniques, they proposed complex models to explain the extensive genome reorganization and gene family evolution built upon the duplication of the genome itself (their Figs. 2–4 and Supplementary Figs 27 – 34 ). Leveraging the recent developments in subgenome-phasing method published by Jia et al. 4 , we propose an alternative model, i.e., reticulate allopolyploidization, to account for the evolution and the genomic diversity of these three Papaver species. Our hypothesis is supported by the following lines of evidence: 1. We extracted 4,791 anchor genes from the inter-genomic syntenic blocks at a ratio of 1:2:4 in P. rhoeas , P. somniferum and P. setigerum using OrthoFinder v2.3.1 5 and MCScanX 6 (Supplementary Figs. 1 A, 2 – 3 ). According to the WGD model proposed by Yang et al. 3 , P. setigerum should have two sister-pairs of homoeologous subgenomes appearing as sisters to the subgenomes of P. somniferum (Fig. 1A ). We inferred the maximum likelihood (ML) trees for each gene and the concatenated sequences of all genes in the same homoeologous chromosome sets (macro-synteny) using IQ-TREE v1.6.12 7 , with P. rhoeas as the outgroup. The top six gene tree topologies, supported by 4,231 (88%) of the 4,791 gene trees (Supplementary Fig. 4 ), show that orthologous gene pairs from P. somniferum and P. setigerum group together, and are sister to the homoeologous genes from P. setigerum . None of the topologies comprising at least 50 gene trees (Supplementary Fig. 4 ) agree with the WGD model shown in Fig. 1A , and most gene trees (43% of the 4791) support the hypothesis that P. somniferum and P. setigerum were derived from a reticulate origin (Supplementary Fig. 4 ; Fig. 1B ). In addition, we obtained 15 groups of concatenated gene trees (macro-synteny trees) with at least 100 syntenic genes, and the topologies of these macro-synteny trees are identical to the most gene trees (Supplementary Fig. 5 ), which further supports the model presented in Fig. 1B rather than that in Fig. 1A . Fig. 1: The origin and evolution of the subgenomes in the three studied Papaver species. A Phylogenomic relationships among the subgenomes assuming the whole-genome duplication (WGD) model of Yang et al. 3 . B Tree topology recovered by gene trees, macro-synteny trees, and species/subgenome trees (see Supplementary Figs. 4 , 5 , 8 for details). The four subgenomes of P. setigerum are designated PseA, PseB, PseC and PseD; the two subgenomes of P. somniferum are designated PsoA and PsoC, and their ancestors are designated A – D . C Circos plot of subgenome partitions of P. somniferum and P. setigerum genomes (more details in Supplementary Figs. 6 , 7 ) indicates that PseA and PsoA, and PseC and PsoC share almost all subgenomic exchanges except a segment in PseC-chr5 that shows exchange with PseB. (a) Subgenome assignment of chromosomes based on the k -means algorithm. (b) Significant enrichment of subgenome-specific k -mers (subgenome partitions). Partitions with the same color as that of a subgenome indicates significant enrichment of k -mers specific to that subgenome. The white areas are not significantly enriched. (c–d) Count (absolute) of each subgenome-specific k -mer set. (e) Homoeologous blocks between the two species. All statistics (b–d) were computed in sliding windows of 1 Mb. Exchanges between subgenomes, such as that in the middle regions of PseC-chr10 and PsoC-chr10, are inferred from inconsistencies between subgenome assignments calculated using chromosomes (ring a) and windows (rings b–d). D The mapping depth of Illumina sequencing reads from P. somniferum to P. setigerum subgenomes. E Insertion times of subgenome-specific LTR-RTs. The 95% confidence intervals (CI) of the insertion times are used to infer the time boundary of divergence to hybridization period. Full size image 2. We phased the subgenomes of P. somniferum and P. setigerum using SubPhaser v1.2 4 (Supplementary Fig. 1B, C , Supplementary Figs. 6 , 7 ), and extracted orthogroups across 23 species/subgenomes, including two subgenomes of P. somniferum , four subgenomes of P. setigerum , P. rhoeas , and representative lineages of other angiosperms, using OrthoFinder (Supplementary Fig. 8 ). We then inferred species/subgenome trees using the ML and coalescence-based methods (Supplementary Fig. 8 ). The topologies of these subgenome trees were consistent with those of the most gene trees (Supplementary Figs. 4 , 5 ), which support the model presented in Fig. 1B . We named the four subgenomes of P. setigerum as PseA, PseB, PseC and PseD, and the two subgenomes of P. somniferum as PsoA and PsoC according to their phylogenetic relationships (Fig. 1B ). Our data suggest that PseA and PsoA, and PseC and PsoC, are derived from separate common ancestors (designated A and C) (Fig. 1B ). The A subgenome is sister to PseB, and the combined A/B clade is sister to the C subgenome (Fig. 1B , Supplementary Fig. 8 ). PseD is sister to P. rhoeas and that clade is sister to the combined A + B + C clade (Fig. 1B , Supplementary Fig. 8 ). 3. We identified exchanges between homoeologous subgenomes in P. somniferum and P. setigerum using SubPhaser (Supplementary Figs. 6 , 7 ; Supplementary Tables 1 , 2 ). We found that the pattern of exchanges on each chromosome between PsoA and PsoC is almost identical to that between PseA and PseC (except for a single exchange between PseB and PseC; Fig. 1C , Supplementary Figs. 6 , 7 ). We then mapped the Illumina sequencing reads from P. somniferum to the P. setigerum subgenomes (Supplementary Fig. 1B ) using sppIDer 8 . The coverage depth plot showed that almost all the P. somniferum reads mapped to PseA and PseC, and very few reads mapped to PseB (i.e. the region exchanged between PseB and PseC) (Fig. 1D ). Syntenic dot plots between the subgenomes showed that PsoA and PsoC had greater similarity (lower Ks ) with PseA and PseC, respectively, but higher Ks with PseB and PseD (Supplementary Fig. 2 ). These results strongly suggest that P. somniferum and the two subgenomes PseA and PseC of P. setigerum were derived from a common allotetraploid ancestor (designated AC). This suggestion agrees with previous cytological evidence that hybrids between P. somniferum (2n = 22) and P. setigerum (2n = 44) had around 11 bivalents (mean 10.7II + 11.6I) at metaphase I 9 . 4. There are two possible processes that could lead to the genomic pattern observed in P. setigerum : (i) AC hybridized with the ancestors of PseB and PseD separately in a stepwise process; or (ii) the ancestors of PseB and PseD hybridized, forming an allotetraploid (designated BD), then BD hybridized with AC forming the allooctoploid progenitor of P. setigerum . To test these two scenarios, we first removed all the potential exchanges between subgenomes of P. setigerum , and identified the subgenome-specific long terminal repeat retrotransposons (LTR-RTs) using SubPhaser (Supplementary Fig. 1D ). Then we estimated the insertion times of subgenome-specific LTR-RTs in P. setigerum to represent the time boundaries from subgenomes differentiation to allohybridization. The estimated PseA- and PseC-specific LTR-RTs insertion times were similar, ranging from ~5 to ~0.5 MYA (95% confidence interval; Fig. 1E ). Similarly, the PseB- and PseD-specific LTR-RTs insertion times were also similar (ranging from ~7.3 to ~0.7 MYA) but distinct from those of PseA and PseC (Fig. 1E ), suggesting that PseB and PseD were more likely to have been introduced into the P. setigerum genome at the same time. Thus, we favored the second scenario, i.e. that the ancestors of PseB and PseD formed an allotetraploid BD, then BD hybridized with AC forming the allooctoploid progenitor of P. setigerum . 5. To test whether other potential progenitors were involved in the evolution of these three species, we downloaded all the available sequencing data of Papaver species from public databases (see Data Availability for details), and assembled the genes of each species using the HybPiper pipeline 10 . We then extracted 1,474 single-copy genes, and inferred a species tree using ASTRAL-MP v5.14.5 11 . The results suggested that subgenome PseD, P. rhoeas and P. dubium originated from a common ancestor (Supplementary Fig. 9 ). Similar to P. rhoeas 3 , P. dubium showed no evidence of recent WGD (Supplementary Fig. 10 ), suggesting it could not be a direct tetraploid progenitor (BD) of P. setigerum . We did not find closely related species for the subgenomes A, B and C, suggesting either the extinction of related ancestors or a sampling gap in taxon coverage. The tree inferred from the whole chloroplast genomes further suggested that P. somniferum was the most likely direct maternal progenitor of P. setigerum (Supplementary Fig. 11 ). Patterns of genome organization in P. setigerum and P. somniferum suggest that post-polyploidization diploidization is probably still ongoing within the two species as there was no largely biased gene fractionation observed in the subgenomes (Supplementary Fig. 12 ).

Topics & Concepts