Genome of tetraploid sour cherry (Prunus cerasus L.) 'Montmorency' identifies three distinct ancestral Prunus genomes

Abstract

Sour cherry (Prunus cerasus L.) is a valuable fruit crop in the Rosaceae family and a hybrid between progenitors closely related to extant Prunus fruticosa (ground cherry) and Prunus avium (sweet cherry). Here we report a chromosome-scale genome assembly for sour cherry cultivar Montmorency, the predominant cultivar grown in the USA. We also generated a draft assembly of P. fruticosa to use alongside a published P. avium sequence for syntelog-based subgenome assignments for 'Montmorency' and provide compelling evidence P. fruticosa is also an allotetraploid. Using hierarchal k-mer clustering and phylogenomics, we show 'Montmorency' is trigenomic, containing two distinct subgenomes inherited from a P. fruticosa-like ancestor (A and A') and two copies of the same subgenome inherited from a P. avium-like ancestor (BB). The genome composition of 'Montmorency' is AA'BB and little-to-no recombination has occurred between progenitor subgenomes (A/A' and B). In Prunus, two known classes of genes are important to breeding strategies: the self-incompatibility loci (S-alleles), which determine compatible crosses, successful fertilization, and fruit set, and the Dormancy Associated MADS-box genes (DAMs), which strongly affect dormancy transitions and flowering time. The S-alleles and DAMs in 'Montmorency' and P. fruticosa were manually annotated and support subgenome assignments. Lastly, the hybridization event 'Montmorency' is descended from was estimated to have occurred less than 1.61 million years ago, making sour cherry a relatively recent allotetraploid. The 'Montmorency' genome highlights the evolutionary complexity of the genus Prunus and will inform future breeding strategies for sour cherry, comparative genomics in the Rosaceae, and questions regarding neopolyploidy.

Figure 1

Syntenic relationships between ‘Montmorency’ subgenomes A, A', and B and other Prunus species. (a) A syntenic dotplot showing the results of a synteny comparison of the 24 superscaffolds (chromosomes) of P. cerasus ‘Montmorency’ and the eight chromosomes of P. persica “Lovell” v2.0 [47]. The figure was generated using unmasked coding sequences for each species with the CoGe platform. The prominent linearity for chr #[A, A', B] for P. cerasus vs the respective chr # in P. persica highlights the collinearity of this genus and supports the integrity of the assembly. (b) Macrosynteny determined with coding sequences shows the three subgenomes of sour cherry are highly syntenic with each other and the published P. avium ‘Tieton’ v2.0 genome [23]. Each gray line represents a syntenic block between the genomes. Small rearrangements between P. avium and each of the ‘Montmorency’ subgenomes are evident.

Figure 2

Linearity comparison of linkage group 1 and a published sour cherry genetic map [16]. A total of 545 markers from an F1 sour cherry cross in which ‘Montmorency’ was the female parent were mapped to the assembly, and the results demonstrate the high collinearity between the linkage map and assembly. Green lines connect the markers on the genetic map (left) to the physical location in the assembly (right). Each horizontal black line on the genetic map represents one marker. Postfiltering, 426 of the 545 markers mapped exactly once to each subgenome. Subgenome B is a representative of two possible haplotypes. This figure was generated with ALLMAPS [103]. Other chromosome sets (2–8) are shown in Figure S5.

Figure 3

Chr8A and chr8A' affect the k-mer groupings that differentiate subgenomes. (a) Hierarchal clustering of all 24 chromosomes based on 25-mer abundance (present 10 times or more on every chromosome and at least twice as abundant on one homoeolog compared to its sisters). The star indicates chr8A and chr8A'. A corresponding heat map is shown in Supplementary Figure S6. (b) An enlarged section of the Hi-C matrix for homoeologous chromosome set 8. The dark red signal in the dotted black box indicates an example of sequence on chromosome 8A' that could have been placed on chromosome 8A in the region indicated with a blue arrow. This could be due to an assembly artifact (collapse of highly similar sequences in these regions), homoeologous exchange between these two chromosomes, or a combination of both. (c) Same hierarchal clustering analysis as in a) but excluding the two 8-chromosome groups.

Figure 4

25-mer group densities differentiating the ‘Montmorency’ subgenomes peak at approximate centromeres. Only chr1 is shown for clarity. (a) Gene and transposable element (TE) densities plotted along the three chromosome 1 homoeologs. The centromeres are estimated to be regions that coincide with relatively low gene and high TE densities. (b) Group 2 and Group 3 25-mer densities (from Figure S6) plotted along the length of chromosome 1. These distinguish the A/A' subgenomes from subgenome B, when all 24 chromosomes are included in this clustering (Figure 3a). (c) Group 5 and Group 6 25-mer densities (from Figure S7) along the length of chromosome 1. These distinguish A and A' from one another. In both b) and c), the region between the vertical lines along the density plots designates the approximate location of the centromeres. Corresponding figures for the seven other chromosome sets are in Figure S8.

Figure 5

Subgenome assignment using syntelogs reveals little-to-no recombination has occurred between progenitor genomes in ‘Montmorency’. 24 093 syntelogs that were identified with phylogenomic ortholog comparisons and synteny analyses are plotted along the lengths of all eight chromosome sets and colored by the progenitor they are most likely derived from. Window size for each tick mark ranges between 130 and 133 Kb and is automatically optimized in chromoMap [142] based on the largest chromosome's size. Mbp, Megabase pairs.

Figure 6

Synteny and phylogeny of the Dormancy Associated MADS-box (DAM) genes in ‘Montmorency’ and Prunus fruticosa. (a) The genomic regions where full DAM haplotypes were identified in both species show high macrosynteny. (b) Phylogeny of the DAM gene coding sequences of ‘Montmorency’, P. fruticosa, and P. avium [31]. Arabidopsis thaliana SEP3 was used as an outgroup. The clustering of DAMs together by number suggests correct identification of these genes. The clustering shows the DAM genes from subgenomes A and A' are most closely related to P. fruticosa DAM genes while DAM genes from subgenome B are most closely related to P. avium DAMs. This agrees with each subgenome's prior assignment to a P. avium-like or P. fruticosa-like progenitor. Nodes below an 80% bootstrap value were collapsed.