Chromosome-scale genome assemblies of five differentBrassica oleraceamorphotypes provide insights in intraspecific diversification

Abstract

Brassica oleracea is an economically important vegetable and fodder crop species that includes many morphotypes exhibiting enormous phenotypic variations. Previously, a pan-genome study based on short reads mapping approach has shown extensive structural variations between B. oleracea morphotypes. Here, to capture more complete genome sequences of B. oleracea, we report new chromosome-scale genome assemblies for five different morphotypes, namely broccoli, cauliflower, kale, kohlrabi and white cabbage, which were created by combining long-read sequencing data and Bionano DLS optical maps. The five assemblies are the most continuous and complete B. oleracea genomes to date (contig N50 > 10 Mb). Comparative analysis revealed both highly syntenic relationships and extensive structural variants among the five genomes. Dispensable and specific gene clusters accounted for ~38.19% of total gene clusters based on a pan-genome analysis including our five newly assembled genomes and four previously reported genomes. Using the pan-genome of B. oleracea and B. rapa, we revealed their different evolutionary dynamics of LTR-RTs. Furthermore, we inferred the ancestral genome of B. oleracea and the common ancestral genome of B. oleracea and B. rapa via a pan-genome approach. We observed faster WGT-derived gene loss in B. rapa than in B. oleracea before intraspecific diversification. We also revealed continuing gene loss bias during intraspecific diversification of the two species and a strong bias towards losing only one copy among the three paralogous genes. This study provides valuable genomic resources for B. oleracea improvement and insights towards understanding genome evolution during the intraspecific diversification of B. oleracea and B. rapa.

Fine mapping of meiotic crossovers inBrassica oleraceareveals patterns and variations depending on direction and combination of crosses

“…We previously generated five chromosome‐scale genome assemblies of five homozygous B. oleracea morphotypes, including broccoli, cauliflower, kale, kohlrabi, and white cabbage (Cai et al., 2022). To explore the recombination landscapes in B. oleracea , we constructed five FwC populations using the above five genotypes as founders, each including four different parents representing four different morphotypes (Figure S1; Figure 1).…”

“…To investigate if SVs locally affect COs, we first focused on two inversions that were validated by Bionano optical maps (Cai et al., 2022); a 4.88‐Mb kale‐specific inversion in C3 and a 1.42‐Mb cauliflower‐specific inversion in C7 (Figure S13). We found only 1 CO inside this region in crosses involving kale, while 13 COs were detected from crosses between parents lacking this inversion (Figure 4a).…”

“…Similarly, we did not find any CO inside the 1.42‐Mb cauliflower‐specific inversion on C7 when crosses involved cauliflower, whereas 10 COs were observed inside this region in crosses between parents without this inversion (Figure 4b). To examine the local suppression effect on a broader scale, we compared all the SVs reported previously (Cai et al., 2022) (Table S4) against fine‐resolution COs in the corresponding four crosses (Broccoli_Cauliflower, Broccoli_Kohlrabi, Kale_Broccoli and White_Broccoli). In all cases, occurrence of large deletions (≥500 bp) and inversions, translocations and transpositions independent of size translate into significantly reduced numbers of COs compared with expected by chance based on 5000 permutations (Figures S14–S17).…”

“…The SelectVariant function was then used to select biallelic SNPs and further SNP filtering was performed using the VariantFiltration function with parameters "--filter-expression 'QD < 2.0 ¦¦ FS > 60.0 ¦¦ MQ < 40.0' --cluster-window-size 5 --cluster-size 2 --filter-name LowQual." SNPs among the five parental lines were called using the high-depth Illumina sequencing data generated by (Cai et al, 2022) with similar strategies as described above for the populations. Only homozygous parental SNPs were retained.…”

“…SNP densities between each combination of the parents were calculated using 2 Mb sliding windows with 50 kb steps. The positions of SVs between parental genomes were obtained from Cai et al (2022). Overlaps between fine-resolution CO intervals and the SV regions were studied using regioneR (Gel et al, 2016).…”