Plant material used in the present study. (a) Five groups of Capsella plants. Diploid species (Co2 and Cg2 groups) and the second generation of resynthesized allotetraploids (Sd and Sh groups) were from Duan et al. (2023). Samples of natural allotetraploids, C. bursa-pastoris, were added to the present study. (b) Phylogenetic relationship of the three natural species used in the present study, modified from Douglas et al., (2015); C. bursa-pastoris originated from the hybridization between the ancestral population of C. orientalis and the (C. grandiflora + C. rubella) lineage, and C. rubella were omitted from the figure; kya: thousand years ago. (c) Geographic origin of the Capsella samples.

Phenotypic traits of the five Capsella groups. Co2: diploid C. orientalis; Cg2: diploid C. grandiflora; Sd: WGD-first resynthesized allotetraploids; Sh: hybridization-first resynthesized allotetraploids; Cbp: natural allotetraploid C. bursa-pastoris. The measured traits were (a) petal length, (b) sepal length, (c) pistil length, (d) stamen length, (e) petal width, (f) sepal length, (g) pistil width, (h) length of the longest stem, (i) number of pollen grains per flower, (j) number of seeds in ten fruits, (k) number of days from germination to the opening of the first flower, (l) proportion of viable pollen grains, and (m) proportion of normal seeds in ten fruits.

Multi-dimensional scaling (MDS) analyses of gene or homoeolog expression in two tissues. Down-sampled gene expressions were used to compare gene expression patterns of the five plant groups in either flowers (a) or leaves (b). Separated homoeolog expressions in allotetraploids were then compared with rescaled gene expression of diploid groups in both flowers (c) and leaves (d). All the MDS analyses used genes with count-per-million (CPM) > 1 in at least two samples, and expression levels were normalized with the trimmed mean of M-values (TMM) method.

Additive and non-additive expression in allotetraploid groups. a) Number of genes that showed additive expression (ADD, including partial expression level dominance), complete expression level dominance (ELD), and transgressive expression (TRE) in each allotetraploid group. b) Genes with complete ELD or TRE were further classified by whether they were up- or down-regulated in allotetraploids, and whether the expression level in allotetraploids was similar to C. grandiflora (Cg-ELD) or C. orientalis (Co-ELD). c) Venn diagram of genes with complete ELD of the three allotetraploid groups in flowers, separated by directions of ELD.

Variation of homoeolog expression bias (HEB) of the three allotetraploid groups in flowers. Gene HEB was calculated as the expression level of cg-homoeolog divided by the total expression level of both cg- and co-homoeologs (HEB = cg/(cg+co)). For each individual, HEB was calculated for 18,255 genes, which had count-per-million > 1 in all flower samples. The distribution of gene HEB was shown by (a) individuals and (b) chromosomes. (c) Gene HEB was also plotted along chromosome positions to show the sudden change of mean HEB between genomic blocks, taking individual Sd-6-4 as an example. The number of cg-homoeologs at each gene estimated by the five-state Hidden Markov Model (HMM) was indicated by five colors. Dark green, light green, grey, light purple and dark purple represent (0, 1, 2, 3, 4) cg-homoeologs and (4, 3, 2, 1, 0) co-homoeologs, respectively.

Homoeolog expression bias (cg/(cg+co)) along chromosomes of natural C. bursa-pastoris in flowers, taking four pairs of chromosome quartets with typical patterns as an example. The number of cg-homoeologs estimated by the five-state Hidden Markov Model was indicated by five colors. The two chromosome quartets in the same row are from the two individuals of the same major genetic cluster of natural C. bursa-pastoris (Kryvokhyzha et al., 2016, 2019b), showing that some estimated segments with an unbalanced number of cg- and co-homoeologs were shared between the individuals from the same genetic cluster.

Loss of homoeolog expression in resynthesized (Sd and Sh) and natural allotetraploids (Cbp). The number of genes with homoeolog expression loss per individual was compared among the three groups of allotetraploids in flowers (a,b) or leaves (c,d). Homoeologous genes that had obvious expression (count per million > 5) in all individuals of the corresponding diploid species but almost no expression (count per million < 0.5) in one allotetraploid individual were considered cases of homoeolog expression loss.

Relationships between homoeolog expression change and non-additive gene expression in flowers. (a) Expected patterns of homoeolog expression change in each scenario that may explain the cause of ELD in resynthesized allotetraploids. (b) Observed homoeolog expression change among genes with ELD in resynthesized allotetraploids. (c) Observed homoeolog expression change among genes with Cbp-specific non-additive expression.

Classification of additive and non-additive gene expression pattern in allotetraploids