Towards the new normal: Transcriptomic convergence and genomic legacy of the two subgenomes of an allopolyploid weed (Capsella bursa-pastoris)

Abstract

Allopolyploidy has played a major role in plant evolution but its impact on genome diversity and expression patterns remains to be understood. Some studies found important genomic and transcriptomic changes in allopolyploids, whereas others detected a strong parental legacy and more subtle changes. The allotetraploid C. bursa-pastoris originated around 100,000 years ago and one could expect the genetic polymorphism of the two subgenomes to follow similar trajectories and their transcriptomes to start functioning together. To test this hypothesis, we sequenced the genomes and the transcriptomes (three tissues) of allotetraploid C. bursa-pastoris and its parental species, the outcrossing C. grandiflora and the self-fertilizing C. orientalis. Comparison of the divergence in expression between subgenomes, on the one hand, and divergence in expression between the parental species, on the other hand, indicated a strong parental legacy with a majority of genes exhibiting a conserved pattern and cis-regulation. However, a large proportion of the genes that were differentially expressed between the two subgenomes, were also under trans-regulation reflecting the establishment of a new regulatory pattern. Parental dominance varied among tissues: expression in flowers was closer to that of C. orientalis and expression in root and leaf to that of C. grandiflora. Since deleterious mutations accumulated preferentially on the C. orientalis subgenome, the bias in expression towards C. orientalis observed in flowers indicates that expression changes could be adaptive and related to the selfing syndrome, while biases in the roots and leaves towards the C. grandiflora subgenome may be reflective of the differential genetic load.

Fig 1. Evolutionary history and sampling locations of the three Capsella species used in this study.

A Solid lines represent subgenomes segregation after the hybridization between C. grandiflora (CG) and C. orientalis (CO) ancestors. C. grandiflora and C. orientalis genetic backgrounds are marked with red and blue respectively. The ploidy levels (n) and the reproductive system are also indicated. Dashed and dotted lines represent the comparisons used to compute the gene expression convergence index (see Material and methods). B. CO, CG, ASI, EUR, ME, CASI correspond to C. orientalis, C. grandiflora, and four populations of C. bursa-pastoris, Cbp, (Asia, Europe, Middle East, and Central Asia) respectively. We shifted slightly population geographical coordinates when those overlapped to make all of them visible on the map.

Fig 2. Genomic variation patterns in three Capsella species.

Variation was visualized with principal component analyses based on the SNPs of C. grandiflora (CG), C. orientalis (CO), and four populations of C. bursa-pastoris (Cbp) (Asia (ASI), Central Asia (CASI), Europ (EUR), and Middle East (ME)). The left plot shows variation in the three species with lines connecting subgenomes of corresponding Cbp accessions and the dash-dotted circles highlighting two subgenomes of Cbp. The middle and right plots show only the variation within the subgenomes of C. bursa-pastoris (Cbp_Cg and Cbp_Co).

Fig 3. Transcriptomic variation patterns in three Capsella species.

Variation was visualized with principal component analyses of phased gene expression data (11,931 genes) for the three different tissues. CO, CG, ASI, EUR, ME, and CASI correspond to C. orientalis, C. grandiflora, and four populations of C. bursa-pastoris, Cbp, (Asia, Europe, Middle East, and Central Asia), respectively. The dash-dotted circles highlight the two different subgenomes of Cbp.

Fig 4. Levels of gene expression in C. bursa-pastoris relative to its parental species.

CO, CG, and Cbp correspond to C. orientalis, C. grandiflora, and C. bursa-pastoris, respectively. The y-axis indicates the level of expression. Expression levels were considered significantly different for the FDR < 0.05. In total, 16,032 genes were analyzed.

Fig 5. Relationships between the relative expression of the C. bursa-pastoris subgenomes and the relative expression of parental species.

The figure shows expression in flower as an example. A. Top-left panel is for all transcripts (11,931). B. Transcripts belonging to a specific category. The diagonal dashed lines indicate 100% cis-regulation divergence while the horizontal dashed lines indicate 100% trans-regulation. The solid lines give the slopes of the linear regressions between both ratios either for all transcript (black) or for transcript belonging to a specific category. β is the slope of the corresponding regression. For Transgressive category (bottom right panel), dark gray corresponds to categories #7a and b, light grey is for category #7c (see Fig 6).

Fig 6. Main categories of expression variation of C. bursa-pastoris subgenomes relative to expression in parental species.

The figure shows expression in flower as an example. Each transcript was assigned to one of seven main categories defined from the relative expression pattern of Cbp subgenomes (Cbp_Cg and Cbp_Co) and parental species (CG and CO). For each category, dashed lines correspond to single transcript relative expression to the maximal expression of this transcript in parental genomes or subgenomes. Solid lines indicate the average expression for each genome or subgenome. Colors discriminate alternative patterns in the same category.

Fig 7. Similarity and convergence indices for differentially expressed genes between subgenomes of C. bursa-pastoris.

A. For each tissue and each subgenome, the median of similarity indices for each subgenome (S_Co and S_Cg) are presented as well as the difference between the two indices (Δ_S) that indicates the dominance of one parental genetic background. Grey dotted lines (S = 0) indicate level of no bias. B. The proportion of transcripts showing convergence (C_i > 0) is reported for the whole genome (green plus signs) or each subgenome (Cbp_Co, Cbp_Cg). The significance of difference between the subgenome convergence indices is also depicted (binomial test,***, p < 0.001). The number of differentially expressed genes considered for each tissue are indicated with N.

Fig 8. Variation in deleterious mutations in the two subgenomes of C. bursa-pastoris.

A. Proportion of deleterious mutations in the subgenomes and in the parental species. CO, CG, ASI, EUR, ME, CASI correspond to C. orientalis, C. grandiflora, and four populations of C. bursa-pastoris, respectively. The two subgenomes are indicated with Co and Cg. Functional effects were annotated with the C. rubella SIFT database (the annotation with A. thaliana SIFT database is in the S10 Fig). B. Maximum likelihood estimates of parameters of the distribution of deleterious mutations on Cbp_Cg genes. Each box represents the estimates for one accession, with 1000 bootstrap replicates. The estimates are presented as the difference between the estimated parameter for deleterious mutations, DEL, and the estimated parameter for synonymous mutations, SYN (Δb = bDEL − bSYN, Δφ = φDEL − φSYN). Notches represent the median and the 95% confidence interval. The left axis refers to Δb (green boxes), and the right axis refers to Δφ (blue boxes). The estimated parameters (b and φ) for DEL and SYN are shown separately in S11 Fig.