Alternating between even and odd ploidy levels switches on and off the recombination control, even near the centromeres

Abstract

Meiotic recombination is a key biological process in plant evolution and breeding, as it generates genetic diversity in each generation through the formation of crossovers (COs). However, due to their importance in genome stability, COs are highly regulated in frequency and distribution. We previously demonstrated that this strict regulation of COs can be modified, both in terms of CO frequency and distribution, in allotriploid Brassica hybrids (2n = 3x = 29; AAC) resulting from a cross between Brassica napus (2n = 4x = 38; AACC) and Brassica rapa (2n = 2x = 20; AA). Using the recently updated B. napus genome now including pericentromeres, we demonstrated that COs occur in these cold regions in allotriploids, as close as 375 kb from the centromere. Reverse transcription quantitative PCR (RT-qPCR) of various meiotic genes indicated that Class I COs are likely involved in the increased recombination frequency observed in allotriploids. We also demonstrated that this modified recombination landscape can be maintained via successive generations of allotriploidy (odd ploidy level). This deregulated meiotic behavior reverts to strict regulation in allotetraploid (even ploidy level) progeny in the second generation. Overall, we provide an easy way to manipulate tight recombination control in a polyploid crop.

Manipulating the ploidy level of Brassica napus affects meiotic homologous recombination, even in centromeric proximal regions.

IN A NUTSHELL.

Background: Species evolution and adaptation relies on genetic diversity. Such diversity is shuffled at each generation during the gamete formation through “meiotic recombination”, where DNA is exchanged through “crossovers” (COs) between chromosome pairs called “homologs”. This mechanism is strictly regulated in sexually reproducing organisms. Normally, COs occur in the distal parts of chromosomes, preventing the combination of some favorable gene and the removal of genes with undesirable effects that are in the “cold” central regions. The identification of novel strategies modifying the recombination landscape and favoring the formation of COs in these “cold” recombining regions is so far lacking.

Question: We sought to determine the extent to which meiotic recombination deregulation in Brassica hybrids with odd ploidy levels may favor the formation of COs in cold chromosome regions, and gain knowledge of the putative underlying mechanisms. We also investigated whether this meiotic deregulation can remain or may be reversible in the following generation.

Findings: Using Brassica napus (oilseed rape) as a model organism, we demonstrate that individuals with an odd ploidy level present a massive deregulation of homologous recombination, with the formation of COs in “cold” regions close to the centromeres. We also demonstrate that this deregulation can be retained or erased in the following generation. Furthermore, we provide novel insights into the potential mechanisms involved in this massive deregulation of the meiotic recombination control, showing that some meiotic genes present an altered expression in Brassica hybrids presenting an odd ploidy level.

Next steps: The next step will be to deepen our understanding of the various molecular mechanisms (specifically histone marks, DNA methylation, or transcriptomic regulation) that are involved in this unprecedented meiotic recombination deregulation. Another major step will be to investigate whether there still are undiscovered benefits of this meiotic deregulation and if these findings may be translated to many other plant models.

Introduction

Meiotic recombination, which involves the formation of crossovers (COs) between homologous chromosomes, is a key biological process in sexually reproducing eukaryotes as it both ensures the faithful segregation of chromosomes and generates genetic diversity in each generation. However, modifying the natural regulation of meiotic recombination remains a major challenge.

COs are initiated in Prophase I by the formation of double-strand breaks (DSBs). Of the 250 DSBs occurring at each meiosis in Arabidopsis (Arabidopsis thaliana), only a few (∼10) are resolved as COs, with the remaining giving rise to non-COs (Mercier et al. 2015). These successful COs belong to 2 classes (I and II) depending on the meiotic proteins involved. Most COs (70% to 85%) formed derive from Class I, which is controlled by proteins of the ZMM group {the zipper proteins ZYP1, ZIP2, ZIP3 [or HUMAN ENHANCER OF INVASION 10 (HEI10)], ZIP4, MUTS HOMOLOG 4 (MSH4), and MSH5 and MEIOTIC RECOMBINATION 3 (MER3), along the MutLγ complex with the MUTL HOMOLOG 1 (MLH1) and MLH3}. These COs are subject to interference, a process that prevents the formation of 2 close-by Class I COs. The remaining COs (15% to 30%) derive from the Class II, which depends mainly on the MMS AND UV SENSITIVE 81 (MUS81) pathway, and are interference insensitive (Mercier et al. 2015; Dluzewska et al. 2018).

In most cases, meiotic recombination is strictly regulated with just 1 obligate CO forming per pair of homologous chromosomes to ensure the proper segregation of chromosomes. Rarely do >3 COs per pair of homologous chromosomes form during meiosis (Mercier et al. 2015; Fernandes et al. 2018).

The overall pattern of meiotic recombination at the chromosome level is consistent within a species but different between species. Some species that have small chromosomes (such as Capsella bursa-pastoris and A. thaliana) tend to present recombination near the centromeric region, whereas other species (such as crops: wheat, barley, maize, or Brassica) with longer chromosomes present a huge bias toward the telomere region, with almost no recombination near the centromeres (Brazier and Glémin 2022). In most organisms, it has been observed that 80% of the COs were concentrated in about 25% of the genome (Choi et al. 2013; Mercier et al. 2015; Darrier et al. 2017). This uneven distribution of COs is associated with the global chromatin organization on the chromosome: COs mainly occur in open chromatin, which harbors a low nucleosome density, a low DNA methylation and is enriched in H3K4me3 histone mark (Choi et al. 2013; Choi and Henderson 2015; Marand et al. 2017; Lian et al. 2022).

In contrast, COs are largely suppressed in the regions close to the centromeres, corresponding to the chromosomal region where the kinetochore platform assembles to allow the attachment of the spindle and the segregation of chromosomes at the opposite pole of meiocytes (McKinley and Cheeseman 2016). These centromeric regions are heterochromatic, repeat rich, dense in nucleosomes, heavily methylated, and enriched in H3K9me2 histone mark (Yelina et al. 2012; Underwood et al. 2018). Despite this tight control of meiotic recombination in most organisms, some natural variations may exist within a species, influencing the species evolution and adaptation (Coop and Przeworski 2022). These differences can be the result of allele or copy number variations of certain meiotic genes, such as HEI10 (Ziolkowski et al. 2017), but also can be related to sex (Lenormand and Dutheil 2005). Indeed, different CO landscapes have been observed between male and female meiosis in several species, a phenomenon referred to as heterochiasmy (Sardell and Kirkpatrick 2020; Capilla-Pérez et al. 2021; Cai et al. 2023).

Within the last 10 years, several methods have emerged to greatly increase the number of COs per meiosis, either by overexpressing pro-CO genes such as HEI10 (Ziolkowski et al. 2017), or by knocking-out (KO) anti-CO genes [e.g. FANCONI ANEMIA COMPLEMENTATION GROUP M (FANCM), RECQ HELICASE L4 (RECQ4), FIDGETIN-LIKE 1 (FIGL1), TOP3α, or ZYP1: Crismani et al. 2012; Séguéla-Arnaud et al. 2015; Girard et al. 2015; Durand et al. 2022; Capilla-Pérez et al. 2024] or by combining both approaches (Serra et al. 2018; Durand et al. 2022). As an example, the recent overexpression of the pro-COs HEI10 protein combined with KO of the ZYP1 gene involved in the synaptonemal complex gave rise to a massive increase in COs, but only in regions where COs normally occur (Durand et al. 2022). To date, very few cases have led to an increase in CO numbers near the centromeric regions. Most of them were obtained when performing a KO of genes involved in epigenetic marks. As an example, the loss of CG methylation via the KO of DNA HYPOMETHYLATED METHYLTRANSFERASE 1 (MET1) or DECREASED DNA METHYLATION 1 (DDM1) in Arabidopsis altered the CO distribution, with an increased COs number around the centromeres but also a decrease in pericentromeric regions (Melamed-Bessudo and Levy 2012; Yelina et al. 2012). Similarly, the disruption of Arabidopsis H3K9me2 and non-CG DNA methylation pathways [via the mutation of KRYPTONITE/SUPPRESSOR OF VARIEGATION 3–9 HOMOLOG 5 and 6 (KYP/SUVH5-6), and CHROMOMETHYLASE 3 (CMT3)], slightly increased meiotic recombination in proximity to the centromeres, including pericentromeres (Underwood et al. 2018).

A third factor affecting recombination corresponds to the ploidy level. In the case of allopolyploid species, which have at least 2 different genomes in the same nucleus, a 2-fold increase in recombination rate has been observed, as exemplified in Gossypium (Brubaker et al. 1999; Desai et al. 2006), Arabidopsis (Pecinka et al. 2011), Brassicoraphanus (Park et al. 2020), or wheat (Wan et al. 2022 ; Yang et al. 2022). In Brassica, it has been found that allotriploidy increases homologous recombination by 3.5-fold compared with diploid and even allotetraploid counterparts (Boideau et al. 2021). This increased recombination frequency in allotriploids is 1.7 times higher in female compared with male meiosis, despite the absence of heterochiasmy in the diploid B. rapa (Pelé et al. 2017). However, in both cases, this increased recombination rate was always associated with a modification of the recombination landscape and with a strong decrease in interference (Pelé et al. 2017). These results in allotriploid hybrids obtained by genetic mapping were confirmed by a 1.7-fold increase in Class I COs detected on pollen mother cells (PMCs) by MLH1 immunostaining (Leflon et al. 2010). However, the low density of genetic markers in centromeric and pericentromeric regions prevented us from unambiguously of KRYPTONITE determining to what extent allotriploidy may facilitate the formation of COs in these regions.

Additionally, it is still unknown what is causing this strongly modified recombination pattern. It has been proposed that the presence of C univalents in allotriploid Brassica hybrids (2n = 3x = 29; AAC) may cause the nucleus to linger in a recombination active state, resulting in an increased CO number and in a modified distribution in the remaining chromosomes that are correctly synapsed (Martinez-Perez and Moore 2008). Although more work is required to understand the origin of this phenomenon, it is important to note that this modified recombination landscape is not specific to Brassica, as a recent study replicating our work in wheat found similar results (Yang et al. 2022). Indeed, in a monocotyledon pentaploid wheat, deriving from a cross between a hexaploid wheat (Triticum aestivum, 2n = 6x = 42, AABBDD) and a tetraploid wheat (Triticum turgidum, 2n = 4x = 28), an increase (3- to 4-fold) in COs between either A or B homologous chromosomes and a modified distribution has been observed (Yang et al. 2022). The similar meiotic deregulations in both Brassica (dicotyledon) and wheat (monocotyledon) hybrids harboring an odd ploidy level suggest that this phenomenon may be common among flowering plants and may be used to efficiently improve the genetic diversity and breeding of many polyploid crops, such as wheat, cotton, oilseed rape, coffee, and strawberry (Leitch and Leitch 2008).

In this study, we wanted to address several remaining important questions linked to the massive modifications of the recombination pattern observed in allotriploids. (i) To what extent can allotriploidy facilitate the formation of COs near the cold recombining centromeric regions? (ii) Can this modified recombination pattern be kept for successive generations, further breaking some linkage drag? Inversely, can it revert back to a normal strict meiotic behavior, preventing putative long-term genomic instabilities? (iii) Can we shed light on some of the mechanisms that might be involved in this phenomenon? To address the first question on the potential ability of allotriploidy to allow the formation of COs close to the centromere regions, we took advantage of the recently updated Brassica napus cv. Darmor genome assembly (Rousseau-Gueutin et al. 2020) that now contains the repeat rich pericentromeric regions and designed 106 markers specific to 2 pericentromeric regions. Thereafter, we studied the recombination landscape within an allotriploid (AAC, odd ploidy level) and allotetraploid (AACC, even ploidy level) progeny of an AAC allotriploid hybrid to decipher if the modified recombination landscape observed in an allotriploid can be maintained by successive generations at the allotriploid level (odd ploidy level) or be reverted back to normal meiotic control by recovering an allotetraploid progeny (even ploidy level).

To answer these questions, we produced and genotyped a total of 6 segregating populations (different ploidy levels and different generations). To compare the recombination frequency and distribution between these hybrids, we performed comparative genetic mapping using the same anchored single nucleotide polymorphism (SNP) set (including pericentromere markers) and analyzed the Class I CO frequency by HEI10 immunostaining on PMCs associated with Genomic in situ hybridization like (GISH-like) chromosome painting. From these analyses, we demonstrated that (i) COs may occur as close as 375 kb from the centromere in an allotriploid compared with 4.5 Mb in an allotetraploid, (ii) the modified recombination landscape can be maintained by successive generations at the allotriploid level (odd ploidy), (iii) normal meiotic control is recovered in an allotetraploid progeny (even ploidy) of an allotriploid hybrid, (iv) the recombination rate is increased within hybrids of the second generation, whatever their ploidy levels, (v) some CO genes present an altered expression in allotriploids. This knowledge is important for both fundamental and applied research, as it increases our understanding on how allotriploidy may impact the global expression of some key meiotic genes and may play a major role in polyploid speciation, evolution, and adaptation. In addition, the method described here provides an efficient and easy way to switch on and off the strict control of recombination in polyploid crop breeding programs.

Results

Allotriploid hybrids have the expected meiotic behavior with lower pollen and seed fertility compared with allotetraploid hybrids

As an initial step to analyzing the evolution over successive generations of the recombination profile in Brassica hybrids harboring different ploidy levels, we investigated the meiotic behavior at Metaphase I of an allotetraploid (A_nA_rC_nC_o) and allotriploid (A_nA_rC_n) F1 hybrids presenting the exact same A genotype, as well as their backcross progenies. For the purposes of this study, we chose the following nomenclature to describe our hybrids: with first the ploidy level, then the generation, and finally, the maternal ploidy origin of each hybrid. Briefly, the F1 allotetraploid and allotriploid hybrids are referred to as 4x-F1 and 3x-F1, respectively. The descending allotetraploid and allotriploid hybrids deriving from the backcross of the 3x-F1 were entitled 4x-B1F1^3x and 3x-B1F1^3x, whereas the allotetraploid hybrid deriving from the backcross of the 4x-F1 was referred to as 4x-B1F1^4x. More details on the genotypes and on the methods used to create these hybrids are given in Fig. 1.

Figure 1.

Schematic representation of the production of the F1 and B1F1 hybrids and their corresponding segregating backcross populations. The names and the ploidy level of the parental lines are indicated and correspond to B. napus cv. Darmor (2n = 4x = 38, A_nA_nC_nC_n), to B. rapa cv. Chiifu (2n = 2x = 20; A_rA_r) and B. oleracea cv. HDEM (2n = 2x = 18; C_oC_o). The nomenclature A_r and A_n correspond to the A genome of B. rapa and B. napus, respectively, whereas C_o and C_n refer to the C genome of B. oleracea and B. napus, respectively. The index “rec” indicates that recombination occurred on the concerned genome. The allotriploid lineage is present on the left part of the diagram, while the right part corresponds to the allotetraploid lineage. The number of genotyped progeny for each of the 6 investigated segregating populations is indicated.

For each hybrid, the meiotic behavior was first established using both classical acetocarmine staining and a GISH-like technique using the Bob014O06 BAC probe that specifically hybridizes to all C chromosomes. The 3x-F1 and 4x-F1 hybrids showed the expected meiotic behavior, with a majority of PMCs with 10 A bivalents plus 9 C univalents for 3x-F1 and 10 A bivalents plus 9 C bivalents 4x-F1 (Table 1, Fig. 2). We also assessed the pollen viability and seed set of these hybrids and observed that the 3x-F1 presents a lower pollen fertility compared with the 4x-F1 hybrid (69% and 97%, respectively, Table 1). Similarly, there is a lower number of seeds per pollinated flower in the 3x-F1 compared with the 4x-F1 (2.3 vs. 10.2, respectively, Table 1). However, the 3x-F1 is fertile. At the second generation, we chose to study 2 different 3x-B1F1^3x hybrids carrying different heterozygous genomic regions in order to validate that any allotriploid hybrid would present the same meiotic recombination pattern. The 2 selected 3x-B1F1^3x hybrids had an improved meiotic behavior compared with the 3x-F1, with systematically 10 A bivalents and 9 C univalent chromosomes. They presented a male fertility and seed set similar to their parental mother plant 3x-F1 (Table 1, Fig. 2). Concerning the 4x-B1F1^3x and 4x-B1F1^4x, they also both shared a similar meiotic behavior, pollen viability, and seed set with the 4x-F1 plant (Table 1, Fig. 2).

Table 1.

Meiotic behavior, pollen fertility, and seed set of the hybrids used in this study

Hybrid	Genomic structure	Number of PMCs	Meiotic behavior	Percentage of cells at the expected meiotic behavior	Pollen viability (%)	Average seed set per pollinated flower	References
3x-F1	AnArCn, 2n = 3x = 29	21	9.38I + 9.81II	81	68.80	2.26	Boideau et al. (2021)
4x-F1	AnArCnCo, 2n = 4x = 38	20	0.2I + 18.8II + 0.05IV	85	96.70	10.16	Boideau et al. (2021)
3x-B1F1a3x	ArecAnCn, 2n = 3x = 29	20	9I + 10II	100	64.90	1.52	This study
3x-B1F1b3x	ArecAnCn, 2n = 3x = 29	20	9I + 10II	100	61.40	1.49	This study
4x-B1F13x	ArecAnCnCn, 2n = 4x = 38	17	0.2I + 18.8II + 0.05IV	85	93.80	13.22	This study
4x-B1F14x	ArecAnCrecCn, 2n = 4x = 38	20	0.3I + 18.75II + 0.05IV	80	96.70	14.42	This study

Figure 2.

Meiotic behavior on chromosome spreads in Metaphase I in the different allotriploid and allotetraploid hybrids using the C-genome-specific BoB014O06 probe. Chromosomes in gray are from the A genome and chromosomes painted in red are from the C genome. GISH-like analyses of meiotic chromosomes were carried out in the allotriploid (left) and allotetraploid (right) hybrids: 3x-F1 (A_nA_rC_n) (A and B), 3x-B1F1a^3x (A_recA_nC_n) (C and D), 3x-B1F1a^3x (A_recA_nC_n) (E and F), 4x-F1 (A_nA_rC_nC_o) (G and H), 4x-B1F1^3x (A_recA_nC_nC_n) (I and J), 4x-B1F1^4x (A_recA_nC_recC_n) (K and L). The nomenclature A_r and A_n correspond to the A genome of B. rapa and B. napus, respectively, whereas C_o and C_n refer to the C genome of B. oleracea and B. napus, respectively. The index “rec” indicates that recombination occurred on the concerned genome. The 9 C univalent in allotriploids (B, D, F) are indicated by asterisks, whereas the 9 C bivalents in allotetraploids are shown with white arrowheads (H, J, L). Bars = 5 µm. Original pictures from the 3x-F1 and 4x-F1 (A, B, G, H) derived from Boideau et al. (2021).

In the following sections, we compared the homologous recombination profile on the 10 A chromosomes of these different hybrids by performing genetic mapping using an identical set of polymorphic SNP markers for each pairwise comparison. These results were thereafter complemented by the quantification of Class I COs frequency on the A subgenome in PMC (male meiosis) at diakinesis, using HEI10 immunostaining coupled with GISH-like chromosome painting using a C-chromosome-specific BAC.

Allotriploidy increases recombination frequency and enables the formation of COs at the vicinity of centromeres, notably via deeply reducing the strength of interference

Backcross populations of the 4x-F1 and 3x-F1 hybrids (always used as females) were previously obtained and genotyped using a Brassica Illumina Infinium SNP array (SGS TraitGenetics GmbH, Gatersleben, Germany). However, the recently improved B. napus cv. Darmor genome using third-generation sequencing technologies (Rousseau-Gueutin et al. 2020) enabled the assembly of the complex and repetitive-rich pericentromeric regions (Supplementary Fig. S1) and revealed that almost no markers were in fact present in these regions. Indeed, in our previous study, the closest markers surrounding the A01 and A02 centromeres were located 5 and 7.6 Mb away from the centromere and thus did not allow us to identify the presence of COs in these regions. To circumvent this problem and properly investigate the presence of COs near the centromere, we took advantage of the recent improvements of both B. napus cv. Darmor and B. rapa cv. Chiifu genome assemblies (A parental donors of our hybrids) to design 106 polymorphic KASPar markers in both A01 (81 markers) and A02 (25 markers) pericentromeric regions (Supplementary Data Set 1). These 2 chromosomes were chosen as they were polymorphic in most of the hybrids investigated in this study, including most second-generation hybrids (3x-B1F1b^3x, 4x-B1F1^3x, 4x-B1F1^4x; see Fig. 1).

Using the same set of A chromosome polymorphic markers from the 15 K array and the newly designed KASPar markers (totaling 2,274 SNPs), we generated genetic maps for both 3x-F1 and 4x-F1. As previously observed (Boideau et al. 2021), the 3x-F1 hybrid showed a significant 3.7-fold increase in COs frequency compared with the 4x-F1 hybrid [Bonferroni-corrected χ² test (BcC test), P < 2.2E−16, Supplementary Data Set 2 and Fig. S2].

When investigating the pericentromeric regions, we found that the 4x-F1 hybrid had no COs, whereas the 3x-F1 hybrid presented 36 COs in total (Fig. 3, Supplementary Data Set 2). Interestingly, these COs arose as close as 375 kb from the centromeric borders in the 3x-F1, while the nearest CO detected for the 4x-F1 hybrid was localized 4.5 Mb from the centromeric border. We also compared the strength of interference between these hybrids, using the Gamma model, where the value Nu reflects the strength of interference (with Nu = 1 representing a total absence of interference and a greater value representing a stronger interference). We observed that the level of interference is significantly lower in the 3x-F1 compared with the 4x-F1 hybrid, with an estimated Nu value of 1.847 and 9.588, respectively (χ² test, P = 0.02). Therefore, interference is highly reduced but not totally abolished in the 3x-F1 hybrid (testing if Nu was close to 1, P < 0.001).

Figure 3.

Comparison of the homologous recombination landscape within the pericentromeric regions of A01 and A02 chromosomes within the 5 investigated hybrids. The investigated genotypes are as follows: 3x-F1 (A_nA_rC_n) (A), 3x-B1F1b^3x (A_recA_nC_n) (B), 4x-B1F1^3x (A_recA_nC_nC_n) (C), 4x-F1 (A_nA_rC_nC_o) (D) and 4x-B1F1^4x (A_recA_nC_recC_n) (E). The nomenclature A_r and A_n correspond to the A genome of B. rapa and B. napus, respectively, whereas C_o and C_n refer to the C genome of B. oleracea and B. napus, respectively. The index “rec” indicates that recombination occurred on the concerned genome. Regions shaded in gray correspond to common heterozygous regions between hybrids for either the A01 (left) and A02 (right) chromosomes. Under each plot, the position of each genetic marker is indicated by “|”. The dark gray rectangles correspond to the pericentromeric region and the black rectangle to the centromere.

The increased CO frequency in the 3x-F1 compared with 4x-F1 hybrids observed in female meiosis using genetic mapping was thereafter validated using HEI10 immunostaining, labeling Class I COs on male meiocytes at diakinesis [24 foci in the 3x-F1 vs. 18 in the 4x-F1 on average, Mann–Whitney Wilcoxon test (MWW test), P = 2.05E–08, Figs. 4 and 5, Supplementary Table S1]. As we previously reported (Pelé et al. 2017), the increased recombination frequency observed in Brassica allotriploids is higher in female compared with male meiosis, explaining the higher increase observed from genetic mapping (female meiosis) compared with HEI10 immunostaining (male meiosis).

Figure 4.

Immunolocalization of HEI10 and GISH-like labeling on diakinesis chromosomes in PMCs of the various allotriploid and allotetraploid hybrids. The investigated genotypes are as follows: 3x-F1 (A_nA_rC_n) (A), 3x-B1F1a^3x (A_recA_nC_n) (B), 3x-B1F1b^3x (A_recA_nC_n) (C), 4x-B1F1^3x (A_recA_nC_nC_n) (D), 4x-F1 (A_nA_rC_nC_o) (E), and 4x-B1F1^4x (A_recA_nC_recC_n) (F). The nomenclature A_r and A_n correspond to the A genome of B. rapa and B. napus, respectively, whereas C_o and C_n refer to the C genome of B. oleracea and B. napus, respectively. The index “rec” indicates that recombination occurred on the concerned genome. For each cell, chromosomes were counterstained with DAPI (white, A1 to F1), HEI10 immunolabeled (green, A2 to F2) and magenta painted chromosomes derive from the C subgenome (BoB014O06, A3 to F3). The overlay of the 3 signals is shown in the column “Merge” (A4 to F4) and a focus on some A and C chromosomes highlighted with a dotted white rectangle in “Merge” column is displayed at the end of each row in the column “Zoom” (A5 to F5). In the third column, the 9 C univalent chromosomes in the allotriploids are indicated by asterisks (A3, B3, C3) and the 10 C bivalents in the allotetraploid hybrids are indicated by white arrow heads (D3, E3, F3). HEI10 foci are highlighted by white arrows in the last column (A5 to F5). Bars represent 5 µm, except in the “Zoom” column, representing 2 µm.

Figure 5.

Average number of HEI10 foci per A bivalent in the hybrids. The 6 investigated hybrids are as follows 3x-F1 (A_nA_rC_n, n = 33), 3x-B1F1a^3x (A_recA_nC_n, n = 30), 3x-B1F1b^3x (A_recA_nC_n, n = 30), 4x-B1F1^3x (A_recA_nC_nC_n, n = 30), 4x-F1 (A_nA_rC_nC_o, n = 30), and 4x-B1F1^4x (A_recA_nC_recC_n, n = 38). Each data point is overlayed and represented by empty circles. The means are indicated by a black asterisk, the medians by the central line of the box. The lower and upper edges of each box represent the lower and upper quartile, respectively. The lower and upper whiskers represent the smallest and largest values within 1.5 time the interquartile range from the lower and upper quartile, respectively. Outliers are represented by filled circles outside of the whiskers.

The modified recombination profile observed in AAC allotriploids can be maintained by performing successive generations at the allotriploid level (odd ploidy) with even a slight increase in CO frequency

To determine whether the modified recombination landscape observed in the 3x-F1 may be conserved for successive generations at the allotriploid (odd ploidy) level, we searched for allotriploid plants in the backcross progeny of the 3x-F1 using flow cytometry. We identified 8 out of the 245 genotyped progeny that were potentially allotriploids. Their chromosome number was thereafter validated at 2n = 3x = 29 using molecular cytogenetics. It is important to note that as these progeny were obtained via a backcross, on average only half of the A genome is expected to be heterozygous. For the following analyses, we thus kept one 3x-B1F1b^3x hybrid, which was heterozygous for the A01 and A02 pericentromere regions, allowing a comparison of these regions in the successive generations. Comparative genetic mapping was only performed in the common heterozygous regions between the 3x-B1F1b^3x and 3x-F1 hybrids (distributed in 18 regions of the 10 A chromosomes, totaling 182.3 over the 346 Mb) using the same set of 1,205 polymorphic SNP markers (Table 2, Fig. 6, Supplementary Fig. S2), thus preventing biases when comparing their recombination profiles. It revealed that the 3x-B1F1b^3x presents a similar recombination profile with even a statistically significant 1.7-fold increase in recombination frequency compared with the 3x-F1 hybrid (BcC test, 3.82E−78). This significant increase was observed for both nonpericentromeric and pericentromeric regions (BcC test, P = 1.36E−14, 2.71E−65, respectively). It was also validated in male meiosis using HEI10 immunostaining, but as expected at a lesser extent due to heterochiasmy (1.18-fold increase, MWW test, P = 1.27E−05, Figs. 4 and 5).

Table 2.

Characterization of the hybrids and segregating populations

Hybrid	Genomic structure	Number of plants	Number of heterozygous regions	Number of polymorphic markers	Number of COs	Genetic size (cM)	Genome coverage (Mb)	Genome coverage (%)	Average recombination rate (cM/Mb)
3x-F1	AnArCn, 2n = 3x = 29	131	10a	2,274	3,946	3,045.3	327.58	94.55	9.30
3x-B1F1a3x	ArecAnCn, 2n = 3x = 29	130	22	980	2,216	1,739.7	146.68	42.33	11.86
3x-B1F1b3x	ArecAnCn, 2n = 3x = 29	130	18	1,205	3,365	2,673.3	182.27	52.61	14.67
4x-B1F13x	ArecAnCnCn, 2n = 4x = 38	172	17	1,128	946	574.6	125.18	36.13	4.59
4x-F1	AnArCnCo, 2n = 4x = 38	172	10a	2,274	1,408	829.6	327.58	94.55	2.53
4x-B1F14x	ArecAnCrecCn, 2n = 4x = 38	172	14	1,099	1,093	665.4	198.84	57.39	3.35

For each hybrid and associated segregating population, several parameters are indicated.

^aThe F1 hybrid is fully heterozygous for the A subgenome.

Figure 6.

Homologous recombination landscape on the A genome of the 6 hybrids. The first outer circle represents the 10 A chromosomes from B. napus cv. Darmor-bzh. The second circle represents the position of the polymorphic SNPs used to generate the genetic maps (A). The red polymorphic SNPs on chromosomes A01 and A02 correspond to the pericentromeric KASPar developed in this study. The third circle refers to the position of the pericentromeric and centromeric regions in gray and black, respectively (B). The lineplots correspond to the recombination rates in the 3x-B1F1a^3x (C), 3x-B1F1b^3x (E), 4x-B1F1^3x (G), 4x-B1F1^4x (I), 3x-F1 (K), and 4x-F1 (M) hybrids. The genotyping blocks bellow each lineplot indicate that the region is either heterozygous (blue) or homozygous (green) in the 3x-B1F1a^3x (D), 3x-B1F1b^3x (F), 4x-B1F1^3x (H), 4x-B1F1^4x (J), 3x-F1 (L), and 4x-F1 (N) hybrids. The hatched areas (as example: end of chromosome A03) correspond to regions without informative genotyping data. Finally, the 2 most inner circles represent the transposable elements density (O) and gene density (P), respectively.

To further validate this increased recombination frequency in the 3x-B1F1b^3x compared with the 3x-F1, similar experiments were performed on another 3x-B1F1a^3x plant. Comparative genetic maps also showed a significant increase in recombination frequency (1.35-fold, BcC test, P = 1.92E−19, Table 1, Fig. 6, Supplementary Data Set 2 and Fig. S2) compared with the 3x-F1, while HEI10 immunostaining revealed no significant differences of HEI10 foci (MWW test, P = 0.98, Figs. 4 and 5, Supplementary Table S1). Therefore, the modified recombination profile observed in the 3x-F1 is maintained in the 3x-B1F1 hybrid, with even a slight increase in the recombination frequency.

The deregulation of the recombination control provoked by allotriploidy can be switched off and reverted back to normal in a descending allotetraploid individual

To decipher whether the modified recombination landscape observed in the 3x-F1 hybrid may revert back to normal and strict meiotic control in the next generation, we first searched for an allotetraploid individual (even ploidy level) in the backcross progeny of the 3x-F1 using both flow cytometry and molecular cytogenetics. In total, we identified 2 potential allotetraploid progeny and we chose 1 individual (referred as 4x-B1F1^3x, Fig. 1) that was heterozygous for the A01 and A02 pericentromeric regions (Fig. 3). In this 4x-B1F1^3x plant, we could identify 17 polymorphic regions, totaling as expected about half of the A genome. We performed comparative genetic mapping between the 4x-B1F1^3x and 3x-F1 (Fig. 6, Table 2, Supplementary Fig. S2) and observed a significant 2.6-fold decrease in the recombination frequency in the 4x-B1F1^3x compared with the 3x-F1 hybrid (BcC test, P = 2.45E−19, Table 1, Supplementary Data Set 2). A significant decrease in Class I COs was also observed in male meiosis using HEI10 immunostaining on A bivalents (1.24-fold, MWW test, P = 5.92E−05, Figs. 4 and 5, Supplementary Table S1).

Interestingly, using both comparative genetic mapping or HEI10 immunostaining, the decrease in the recombination frequency observed in the 4x-B1F1^3x was smaller than expected, compared with the observed 3.7-fold observed between the 4x-F1 and the 3x-F1. The smaller decrease in the recombination frequency in the 4x-B1F1^3x may either be due to a residual effects from the allotriploid step or its backcross origin, with only half of the genome that contains heterozygous regions juxtaposed to homozygous regions. To test these hypotheses, we compared the recombination frequency between a 4x-B1F1^4x and the 4x-F1 hybrid using both genetic mapping and HEI10 immunostaining (Table 2, Fig. 5, Supplementary Fig. S2). We observed an increased recombination frequency in the 4x-B1F1^4x compared with the 4x-F1, as previously observed for the allotriploid lineage. We also compared the recombination frequency between a 4x-B1F1^3x and 4x-B1F1^4x using an identical set of markers, and surprisingly observed slightly more recombination in the 4x-B1F1^4x compared with the 4x-B1F1^3x (BcC test, P = 0.026). However, this difference was only due to 1 region on the chromosome A07 (BcC test, P = 0.085). Altogether, these results highlight an absence of residual effect from the allotriploid step and suggest that the increased recombination frequency observed in the second generation is a result of the backcrossing.

The global transcription level of some meiotic genes is modified in the AAC allotriploid hybrid

To determine whether some meiotic genes might present a modification of their global transcription level in reproductive tissues where meiosis takes place, RT-qPCR experiments were performed on flower buds containing a majority of PMC at the prophase stage for genes involved in double-strand break repair (DMC1), synaptonemal complex formation (ASY3, ZYP1A, ZYP1B), Class I (HEI10, MSH4), or Class II (MUS81) pro-COs and anti-COs (TOP3α) genes in the allotriploid 3x-F1 (2n = 3x = 29; AAC) and the F1's parental species: the diploid B. rapa cv. Chiifu (2n = 2x = 20; AA) and the allotetraploid B. napus cv. Darmor (2n = 4x = 38; AACC; Fig. 1). It is important to note that the diploid B. rapa (AA) and the allotriploid AAC hybrid will only form COs between the 10 A homologous chromosomes (the 9 C chromosomes remaining as univalents at Metaphase I in the allotriploid hybrid), whereas COs will be formed between the 19 pairs of homologous chromosomes in the allotetraploid AACC B. napus (between the 10 A and 9 C pairs of homologous chromosomes).

Interestingly, we observed that genes involved in CO precursors and Class I COs (DMC1, ASY3, ZYP1, HEI10, and MSH4) tend to be more highly transcribed in the allotriploid 3x-F1 AAC hybrid compared with the parental lines, but only a few of them show some statistically significant differences (Fig. 7). More precisely, the relative expression of MHS4 and HEI10, involved in Class I CO formation, was roughly 3-fold greater in the allotriploid 3x-F1 AAC hybrid compared with the diploid AA parent B. rapa cv. Chiifu (ANOVA and Tukey honestly significant difference (HSD) test, P = 0.011 and 0.017, respectively, Fig. 7), but the expression levels did not significantly differ between the allotriploid 3x-F1 AAC hybrid and the allotetraploid AACC B. napus cv. Darmor (ANOVA and Tukey HSD test, P-values in Supplementary Data Set 3, Fig. 7). Regarding the Class II CO gene, a significantly smaller relative abundance of MUS81 was observed in the allotetraploid AACC B. napus cv. Darmor compared with other ploidy levels (ANOVA and Tukey HSD test, P-values in Supplementary Data Set 3, Fig. 7). No difference was observed for the anti-CO gene TOP3α between the 3 different ploidy levels (ANOVA and Tukey HSD test, P-values presented in Supplementary Data Set 3, Fig. 7).

Figure 7.

Relative transcription levels of 7 meiotic genes in 3 ploidy levels. Relative transcription level of the genes DMC1, ASY3, ZYP1, MSH4, HEI10, MUS81, and Top3α, determined by RT-qPCR on flower buds, for the diploid B. rapa cv. Chiifu (AA, in green), the allotriploid 3x-F1 (AAC, in pink), and the allotetraploid B. napus cv. Darmor (AACC, in blue). Each bar corresponds to the average of 3 biological replicates. The different letters indicate that mean values are significantly different (P < 0.05, Supplementary Data Set 3) between genotypes using an ANOVA and Tukey post hoc tests. Error bars correspond to the Sd.

Allotriploidy does not alter chromatin compaction in leaves

To determine whether allotriploidy was associated with a modification of the chromatin state, thus potentially facilitating the formation of COs in the pericentromere regions, we compared the chromatin conformation (from leaves) of the F1 allotriploid AAC hybrid 3x-F1 and its 2 parental lines B. napus cv. Darmor and B. rapa cv. Chiifu using Hi-C (Supplementary Fig. S3). This technique can determine whether a chromosome region is in an open chromatin (euchromatin, compartment “A”) or closed chromatin (heterochromatin, compartment “B”) states. Between biological replicates within each genotype, 78.7% to 93.7% of the genomic 100 kb bins were assigned to the same chromatin compartment (Supplementary Fig. S3). The parental lines B. napus cv. Darmor and B. rapa cv. Chiifu present 88.0% of bins assigned to the same chromatin compartment and the allotriploid 3x-F1 hybrid presents, with B. napus cv. Darmor and B. rapa cv. Chiifu, 99.2% and 88.8% of bins assigned to the same chromatin compartment, respectively. Overall, no important modification of chromatin compaction was observed at the chromosome scale between these 3 genotypes, especially between the allotriploid 3x-F1 hybrid and its parent B. napus cv. Darmor.

Discussion

In this paper, we deepened our understanding of the important modification of the recombination landscape observed in Brassica AAC allotriploid hybrids. We notably showed that COs can arise next to the centromere and only in an AAC allotriploid hybrid, as close as 375 kb. Additionally, this modification of the recombination landscape can either be maintained or be reverted back to normal meiotic behavior in the following generation, either by maintaining an AAC allotriploid structure (odd ploidy level) or by recovering a AACC allotetraploid structure (even ploidy level) in the allotriploid's progeny, respectively. Based on these findings, together with those from our previous studies on allotriploid Brassica hybrids (Leflon et al. 2010; Pelé et al. 2017; Boideau et al. 2021) and the current knowledge about meiosis in Arabidopsis, we attempted to increase our knowledge on the underlying mechanisms. Our study revealed that transcriptomic deregulation of some meiotic genes might be involved in the modified recombination landscape of the allotriploids. Given the massive modification of the recombination pattern observed in allotriploid hybrids AAC, it is expected that numerous genes and molecular mechanisms are involved.

Allotriploidy massively increases the CO formation in the normally low recombining pericentromeric regions

Contrary to the hypothesis that the absence of COs in centromeric regions may be related to the key role of this chromosomic region in the proper segregation of chromosomes (McKinley and Cheeseman 2016), we demonstrated that COs occur in natural allotriploid hybrids at a high frequency in pericentromeric regions. Due to their key cellular function, the centromeric regions, which are composed of megabases of satellite repeat arrays (Naish et al. 2021), are typically lacking in COs (Fernandes et al. 2024). Even large regions (several Mb) surrounding the centromeres normally experience very low recombination rates, as recently exemplified in Arabidopsis (Fernandes et al. 2024) and in crops, such as barley (Dreissig et al. 2020) and wheat (Raz et al. 2021). Our previous research (Pelé et al. 2017) had limited capacity to explore the impact of ploidy level on pericentromere recombination due to the absence of pericentromeric markers within the first Brassica genome assemblies (Wang et al. 2011; Chalhoub et al. 2014). Here, we took advantage of the recent improvements of B. napus cv. Darmor (Rousseau-Gueutin et al. 2020) and B. rapa cv. Chiifu (Zhang et al. 2018) genome assemblies, which now include the pericentromeres, to determine whether allotriploidy facilitates the formation of COs in these regions.

To fulfill this goal, we designed several KASPar markers specific to the pericentromeric regions of chromosomes A01 and A02 spanning 8.2 and 7.1 Mb, respectively, and that were heterozygous in most of our investigated hybrids (5/6). As many as 36 out of 131 allotriploid individuals had a CO in the normally cold pericentromeres, while no COs were detected in the same chromosomal regions in an allotetraploid population of 172 individuals. Additionally, we observed that COs may occur as close as 375 kb from the centromere border in allotriploids compared with 4.5 Mb in allotetraploids. The natural method we present here, which switched from an even to an odd ploidy level, will be far more efficient for generating genetic diversity in these pericentromeric regions than any other methods tested yet. The ability to form COs in these pericentromere regions is crucial for plant breeding applications, as they are not deprived of expressed genes and include genes of interest for crop improvement (Rousseau-Gueutin et al. 2020; Boideau et al. 2021; Fernandes et al. 2024). As an example, the B. napus cv. Darmor A01 and A02 pericentromeric regions contain 662 and 505 genes, representing 15.64% and 11.31% of the genes present on the A01 and A02 chromosomes, respectively. According to Rousseau-Gueutin et al. (2020), 28 and 25 Resistant Gene Analogs were identified among these genes in the A01 and A02 pericentromeric regions. Recently, the particular interest of using Brassica allotriploidy to decrease the size of QTLs present in pericentromeric regions was highlighted by a modeling study (Tourrette et al. 2021) and exemplified for a pericentromeric QTL conferring blackleg resistance (Boideau et al. 2021), demonstrating the potential of this method for plant breeding.

A natural method enabling to switch on and off the strict meiosis regulation

In Brassica, allotriploidy was shown to deeply modify the recombination landscape (Pelé et al. 2017; Boideau et al. 2021). However, it remained to be determined whether this modification may be maintained for successive generations to take advantage of this meiotic deregulation or whether the recombination landscape would revert back to normal to prevent potential longer term genomic instabilities (Vincenten et al. 2015). To test this hypothesis, we identified AAC and AACC progeny from the first allotriploid hybrid. We were able to show that the maintenance of an allotriploid level in the second generation conserved this modified recombination pattern, including the presence of COs in the normally cold recombining pericentromeric regions. Contrastingly, we demonstrated that recovering an allotetraploid individual can immediately switch off this deregulation. This descending allotetraploid individual recovered a normal meiotic behavior and a seed set similar to most cultivated allotetraploid B. napus varieties (Siles et al. 2021). The observation of the presence of a modified recombination pattern only in allotriploid but never in allotetraploid hybrids (irrespective of generation) strongly indicates that the presence of an odd ploidy level or interploidy (i.e. the presence of 1 subgenome in a single copy remaining as univalent during meiosis) is at the origin of this modification of recombination control.

Recently, the identification of a similar meiotic deregulation in a wheat allopentaploid (2n = 5x = 35, AABBD; Yang et al. 2022) strongly supports this hypothesis and suggests that this phenomenon may be common to many polyploid flowering plants. Given the large number of polyploid crops, and the fact that many of them have experienced genetic diversity erosion due to domestication and subsequent crop improvement, this modification of the recombination control via interploidy could deeply increase genetic diversity and produce new allelic combinations (or haplotypes) in future crop cultivars. Indeed, many polyploid crops can be crossed to their parental progenitor of a smaller ploidy level and give rise to a still fertile interploid hybrid, as demonstrated in wheat (Vardi and Zohary 1967), coffee (Krug and Mendes 1940), strawberries (Yarnell 1931), tomato (Rick et al. 1988), or tobacco (East 1933). These examples highlight the real potential of applying this method in plant breeding programs.

The ability to easily recover allotetraploid B. napus individuals with stable meiosis and improved seed set among the progeny of allotriploid hybrids provides a proof of concept that the allotriploid pathway is of high relevance: (i) to speed up and reduce the costs of some breeding programs by reducing the size of segregating populations and the number of individuals to be genotyped, (ii) to reduce the size of B. rapa introgressed regions of interest (preventing the parallel introgression of deleterious alleles), and lastly, (iii) to reduce the linkage disequilibrium observed within B. napus.

Allotriploidy alters the global transcription of some meiotic genes but does not modify chromatin compaction

Given the massive deregulation of recombination observed in allotriploid hybrids, numerous mechanisms and genes may be involved. Here, we compared the global expression of major key meiotic genes by RT-qPCR between the allotriploid hybrid and its parental lines. We observed a 3-fold higher transcription level of 2 ZMM (Class I COs) genes (HEI10 and MSH4) compared with the diploid parent that presents the same number of bivalents on which COs may form. Interestingly, these 2 genes are also slightly more expressed in the allotriploid than in the allotetraploid, despite having a smaller number of copies. Thus, this increased expression of HEI10 in Brassica allotriploids seems to be involved in the increase in the number of Class I COs, as previously observed in A. thaliana (Ziolkowski et al. 2017). On the contrary, similar transcriptional levels were observed between the allotriploid and the diploid for the Class II pro-COs MUS81 gene or the anti-COs Top3α gene, suggesting that only Class I COs and not Class II COs are involved in the increased recombination frequency observed in allotriploids. It will be interesting to perform RNA-seq on similar material in the near future, despite the difficulties to obtain it in sufficient quantity, as it might bring additional knowledge of the complex transcriptional modifications occurring in the allotriploid meiosis.

The modification of the global transcription level of some meiotic genes may also be involved in the presence of heterochiasmy in Brassica allotriploid hybrid. In the latter, the presence of heterochiasmy was identified with an increased recombination rate in female compared with male, despite the absence of heterochiasmy in Brassica diploid and allotetraploid hybrids (Kelly et al. 1997; Pelé et al. 2017). The heterochiasmy observed in Brassica allotriploid differs from A. thaliana, as Arabidopsis presents, on the contrary, a higher recombination rate in male compared with female meiosis (Drouaud et al. 2007). Nevertheless, a higher increase in the recombination frequency in female compared with male meiosis (as found in Brassica allotriploids) has been observed in Arabidopsis mutants that were KO for zyp1 and overexpressed for HEI10 (Durand et al. 2022), or in mutants that are KO for recq4 and figl1 genes (involved in Class II anti-CO pathways: Fernandes et al. 2018). According to our RT-qPCR experiments, the ZYP1 gene seems to present a higher transcription level in the allotriploid compared with the diploid or the tetraploid, but these differences are not statistically significant due to the important variation existing between biological replicates for that particular gene. So far, it is unknown whether ZYP1 overexpression could generate a similar phenomenon and whether similar mechanisms are involved in both Arabidopsis and Brassica.

In Brassica allotriploids, the increased recombination rate in both male and female meiosis is associated with a near loss of interference. Such loss of interference was observed in Arabidopsis mutants for the synaptonemal complex: zyp1 (Capilla-Pérez et al. 2021) or scep1 and scep2 double mutants (Vrielynck et al. 2023). As allotriploid hybrids seem to have increased transcription of ZYP1, it thus seems unlikely that this gene is involved in the modified recombination control. In agreement with this hypothesis, in Arabidopsis ZYP1 mutants, a lack of interference generates a higher recombination in distal regions but a slightly lower recombination in regions close to the centromere regions, which is the opposite of what we observe in Brassica allotriploids.

In Brassica allotriploids, the modification of the recombination distribution, even in the vicinity of centromeres, could be explained by the modification of some epigenetic marks. It has been shown that constitutive modifications of DNA or histone methylation are associated with the recombination profile in Arabidopsis (Melamed-Bessudo and Levy 2012; Yelina et al. 2012; Serra et al. 2018; Underwood et al. 2018; Fernandes et al. 2024). However, the formation of pericentromeric COs in Arabidopsis mutants for epigenetic marks arises to a much smaller extent compared with what we observed in our Brassica AAC allotriploid, and thus may be the result of numerous epigenetic alterations. Here, we studied the chromatin compaction using Hi-C data obtained from leaves to determine whether allotriploidy may have modified the chromatin compaction (especially in pericentromeric regions), potentially facilitating the formation of COs in these latter regions. Comparative analyses of Hi-C data revealed that the allotriploid hybrid and its parental lines presented a similar euchromatin and heterochromatin compartmentation, indicating that allotriploidy does not importantly alter chromatin compaction in leaves. In the near future, it will be important to replicate such an experiment using reproductive cells (where meiosis occurs) as chromatin reorganization may specifically be altered during meiosis (Lenykó-Thegze et al. 2021). Additionally, the analysis of some key histone marks (such as H3K4me3 and H3K9me2) by ChIP-seq or DNA methylation by bisulfite sequencing on reproductive tissues might bring additional information.

The tremendous modifications of the recombination landscape observed in allotriploid Brassica are likely the result of the deregulation of many meiotic and epigenetic genes. Functional genetic studies testing some candidate genes individually remain the best approach for understanding the mechanisms involved in this intriguing deregulation observed in Brassica allotriploids. Despite the continuous improvement of the Crispr-Cas technologies in recent years, including in B. napus (Facon et al. 2023), testing the impact of a given gene on recombination in Brassica hybrids of different ploidy levels will remain difficult and time consuming, given (i) the duration of a single generation in Brassica, (ii) the presence of several copies for most genes (25% of meiotic genes are in duplicated copies in Brassica species), (iii) the necessity to generate mutants for 2 genotypes per species to evaluate their impact on the recombination landscape, (iv) the requirement to compare hybrid mutants from different ploidy levels, as well as (v) the ability to produce a segregating population from each hybrid mutant.

The juxtaposition effect might participate in the recombination frequency of the second generation

The comparison of the first- and second-generation hybrids (from the same ploidy level) revealed the presence of a slight increase in CO frequency in the second generation, which may be associated with the juxtaposition effect identified in Arabidopsis (Ziolkowski et al. 2015). In this latter species, a slight increase in 1.35-fold COs in the heterozygous region was juxtaposed to a homozygous region (Ziolkowski et al. 2015). The juxtaposition effect corresponds to an MSH2-dependent local redistribution of Class I COs toward polymorphic heterozygous regions to the detriment of the juxtaposed homozygous regions in A. thaliana (Blackwell et al. 2020). Similarly, a slight increase in recombination in BC2 compared with BC1 hybrids was observed in Brassica oleracea (1.66-fold: Kearsey et al. 1996). Therefore, this effect may be general at least to Brassicaceae and may be considered in some breeding programs.

Putative role of an allotriploid step in the speciation success of a newly formed polyploid

Allopolyploid species may arise by numerous routes. They may form via the merging of 2 divergent diploid genomes through the formation of an allodiploid (i.e. homoploid bridge), allotriploid (i.e. unilateral pathway) or allotetraploid (i.e. bilateral pathway) hybrids (Tayalé and Parisod 2013; Pelé et al. 2018). For several polyploid species, only a few allopolyploidization events are often at their origin. To increase its population size, a new polyploid may produce progeny through self-fertilization or via crossing with one of its parental species. The latter route will form allotriploid individuals presenting the deregulated recombination patterns described in this study. Thereafter, the selfing of this allotriploid and/or its backcrossing with an allotetraploid individual will thus give rise to more genetically diversified allotetraploid individuals. A recent study from Cao et al. (2023) revealed that within 3 generations of self-fertilization, allotriploids mainly developed into near-complete allotetraploid individuals by gradually increasing chromosome number and fertility. Indeed, natural selection strongly acts and favors the genomically stable allotetraploid progeny over interploid or aneuploid individuals. Nevertheless, during these few intermediate generations of aneuploidy, it is likely that these individuals also presented a modified recombination pattern, as previously observed in Brassica (Suay et al. 2014). Thus, allotriploidy, with aneuploidy as intermediate, can play a major role in polyploid genetic diversification and potential adaptation, facilitating their potential establishment and speciation success. These results further support the importance of allotriploidy as a bridge in polyploid speciation.

Materials and methods

Plant material and growth conditions

The plant material analyzed in this study is described in Fig. 1. To distinguish the different genomes, we used as nomenclature A_r and A_n for the A genome of B. rapa and B. napus, respectively, whereas we used C_o and C_n for the C genome of B. oleracea and B. napus, respectively. The term “Rec” as index indicates that recombination occurred on the genome concerned. To produce an allotriploid AAC and allotetraploid AACC hybrids presenting a genetically identical A genome sequence enabling an unbiased comparison of their recombination profile, we used the following strategy. To produce the allotetraploid hybrid (4x-F1), we first created a resynthesized oilseed rape, which was obtained by crossing the pure inbred line B. rapa cv. Chiifu (A_rA_r, 2n = 2x = 20) with the doubled haploid B. oleracea cv. HDEM (C_oC_o, 2n = 2x = 18; used as male) and then by performing embryo rescue on the obtained amphihaploid (A_rC_o, 2n = 19), as described earlier (Jahier et al. 1992). This hybrid spontaneously doubled its genomes and gave rise to the resynthesized allotetraploid A_rA_rC_oC_o (2n = 4x = 38). This latter plant (conserved as cuttings) was crossed with B. napus cv. Darmor (A_nA_nC_nC_n) as female, and the A_nA_rC_nC_o F1 hybrid referred as 4x-F1 gave rise to 172 B1F1 plants after backcrossing with recurrent parent B. napus cv. Darmor as male (Boideau et al. 2021). One of these B1F1 plants with ∼50% of A and C genome at the heterozygous stage was kept (A_RecA_nC_RecC_n, 2n = 4x = 38, referred as 4x-B1F1^4x) to produce a B2F1 segregating population (172 plants) by backcrossing it with B. napus cv. Darmor as male (Fig. 1).

For the allotriploid pathway, an allotriploid F1 hybrid (A_nA_rC_n, 2n = 3x = 29, referred as 3x-F1) was obtained by crossing the allotetraploid B. napus cv. Darmor A_nA_nC_nC_n (2n = 4x = 38) and B. rapa cv. Chiifu (A_rA_r, 2n = 2x = 20) as male (Nagaharu 1935). For the production of its B1F1 progeny, the F1 hybrid was crossed as a female with B. napus cv. Darmor as male. We selected 131 plants to generate the mapping population of the 3x-F1 hybrid (Boideau et al. 2021). Within this latter B1F1 population, we selected 2 AAC allotriploid hybrid progeny (A_RecA_nC_n, 2n = 3x = 29, hereafter referred to as 3x-B1F1a^3x and 3x-B1F1b^3x) and 1 allotetraploid hybrid progeny (A_RecA_nC_nC_n, 2n = 4x = 38, referred as 4x-B1F1^3x) based on their meiotic behavior and on the common polymorphic regions. Then, these 3 hybrids were backcrossed to B. napus cv. Darmor as male, giving rise to 3 mapping populations of 130, 130, and 172 plants, respectively (Fig. 1). All parental accessions were provided by the Biological Resource Center BrACySol (INRAE, Ploudaniel, France).

All of the plants used for the genetic analyses were grown in a greenhouse 16 h of light (natural daylight with supplementation by high pressure sodium lamps; 200 µmol m⁻² s⁻¹, 21 °C) followed by 8 h of dark (18 °C). Plants were grown in 2-L pots filled with a nonfertilized commercial substrate (Falienor, reference 922016F3) and irrigated twice a week with a commercial fertilized solution (Liquoplant Blue, 2.5% nitrogen, 5% phosphorus, 2.5% potassium, w/v).

DNA extraction and genotyping

Genomic DNA was extracted from lyophilized young leaves with the sbeadex maxi plant kit (LGC Genomics, Teddington, Middlesex, UK) on the oKtopure robot at the GENTYANE platform (INRAE, Clermont-Ferrand, France). Genotyping data were obtained using the Brassica 15 and 19 K Illumina Infinium SNP array (SGS TraitGenetics GmbH).

To assess recombination within the pericentromeric region of A01 and A02 chromosomes that contained a very low number of polymorphic SNPs between our parental lines, we then developed an additional set of markers to specifically densify these pericentromeric regions. The improved assembly of pericentromeric regions between B. napus cv. Darmor v10 (Rousseau-Gueutin et al. 2020) and v5 genome assemblies (Chalhoub et al. 2014) can be visualized in Supplementary Fig. S1, which was obtained by comparing their assemblies using SyRI (Goel et al. 2019) with default parameters.

To design A01 and A02 pericentromeric markers, we first took advantage of the presence of polymorphic markers between our parental lines using data from the Brassica 60 K Illumina Infinium SNP array (Clarke et al. 2016). This enabled us to identify 48 polymorphic markers that were absent from the 15 and 19 K Brassica Illumina Infinium arrays and for which we developed KASPar markers using their context sequences. We also designed 39 additional KASPar markers closer to the A01 and A02 centromeres by taking advantage of the whole-genome assembly of the parental lines (Zhang et al. 2018; Rousseau-Gueutin et al. 2020). To design these new SNP markers within our intervals of interest, we first retrieved the genes present in single copy in each B. napus cv. Darmor-bzh v10 subgenome (Rousseau-Gueutin et al. 2020) and in B. rapa cv. Chiifu v3 (Zhang et al. 2018), preventing the putative design of markers amplifying on different paralogous regions. The putative orthologous copy of these genes in B. oleracea cv. HDEM was also retrieved using BlastP. For each of these genes, we performed an alignment of the different copies (B. rapa, B. napus Copy A and Copy C, B. oleracea) and used an in-house Python script to specifically detect polymorphism between the A chromosomes of B. napus cv. Darmor and B. rapa cv. Chiifu. Overall, genotyping data were obtained for this set of 87 SNP markers (specific to the A01 and A02 pericentromeric regions, details for these markers are given in Supplementary Data Set 1) and revealed by Biomark HD system (Fluidigm Technology) and KASPar chemistry at the GENTYANE platform (INRAE, Clermont-Ferrand, France). The raw genotyping data deriving from the Brassica Illumina Infinium arrays or from the KASPar technology were analyzed using either the GenomeStudio v.2011.1 (Illumina Inc., San Diego, CA, USA) or the Fluidigm SNP Genotyping Analysis v4.1.2 softwares (Wang et al. 2009). In both cases, the raw genotyping data were processed with the auto-clustering option and validated manually. The polymorphic SNPs between the parental genotypes were selected for the establishment of genetic maps. SNPs showing >20% of missing data and plants showing >25% of missing data were removed for the downstream analyses. Potential double COs supported by only 1 genetic marker and with a physical distance between these 2 events <500 kb were corrected as missing data, as described in Rowan et al. (2019). To prevent the overestimation of recombination in the low recombining pericentromeric regions, plants showing >1 CO in the pericentromeric region of A01 and A02 were removed. Finally, to determine the physical position of the different markers, the context sequences of each SNP marker were physically localized on the reference genome B. napus cv. Darmor-bzh v10 (Rousseau-Gueutin et al. 2020) by using Blastn (ver. 2.9.0, Altschul et al. 1990) and by keeping the best blast hit obtained for a given subgenome (minimum percentage of alignment and identity: 80%).

Genetic maps

The genetic maps were produced, as described in Boideau et al. (2022). Succinctly, all SNP markers that were polymorphic between the parental lines B. rapa cv. Chiifu and B. napus cv. Darmor, heterozygous in both F1 hybrids (3x-F1 and 4x-F1) and showing a disomic segregation (1:1 ratio) in the 2 backcross populations were selected to only follow homologous recombination between A chromosomes. The first genetic maps were established separately but in the same manner for each population using the CarthaGene software (v. 1.2.3, De Givry et al. 2005). Establishment of linkage groups and SNP ordering were determined using a logarithm of odds score threshold of 4.0 and a threshold recombination frequency of 0.3, as previously described (Pelé et al. 2017). The few markers showing discordance between the genetic order and the physical position of the markers along the A chromosomes of the B. napus cv. Damor v10 genome assembly (Rousseau-Gueutin et al. 2020) were discarded in order to build the final uniform genetic maps. After these few corrections, the final genetic maps were created using the Kosambi function to evaluate the genetic distances in centimorgans (cMs) between linked SNP markers (Kosambi 1943). For the second-generation hybrids (3x-B1F1a^3x and 3x-B1F1b^3x, 4x-B1F1^3x and 4x-B1F1^4x), the same methodology as previously described was used to establish the linkage groups. Genotyping data used are available in Supplementary Data Set 1. The genetic landscapes in Fig. 6 were illustrated using Circos v0.65f (Krzywinski et al. 2009).

The pericentromeric and centromeric borders were retrieved from Boideau et al. (2022) and were based on gene density and centromeric-specific repeats, respectively. Precisely, pericentromeric regions were defined as regions surrounding the centromere and characterized by a gene density below the chromosome average. In this study, a pericentromeric interval refers to the interval between the pericentromeric borders, corresponding to both the centromeric and pericentromeric regions.

Interference

The interference parameters using the Gamma model were determined using the software CODA with the default parameters (Gauthier et al. 2011). Confidence intervals and statistical tests were not based on Fisher's Gaussian approximation as proposed in the graphical user interface of CODA but were performed based on 1,000 resimulations using the command-line interface via custom Perl and R scripts.

Flow cytometry and cytogenetic studies

Chromosome numbers of the allotriploid progenies were assessed in leaves by flow cytometry, as described in Leflon et al. (2006). Pollen viability was assessed from 3 independent flowers, using acetocarmine staining, as described in Jahier et al. (1992). For the establishment of meiotic behavior, samples of young floral buds were fixed in Carnoy's II solution (alcohol:chloroform:acetic acid, 6:3:1, Carnoy 1886) for 24 h at room temperature and stored until use in 50% v/v ethanol at 4 °C. Anthers were then squashed and stained with a drop of 1% w/v acetocarmine solution. Per plant, chromosome pairing was assessed from 20 PMCs at Metaphase I.

Meiotic protein immunolocalization

The immunolabeling of the protein HEI10 was carried out on meiotic chromosome spread prepared, as described in Chelysheva et al. (2013) with minor modifications. Briefly, the chromosome preparations were incubated in Rnase A (100 ng µL⁻¹, Sigma-Aldrich R4642) and pepsin (0.05% w/v, Sigma-Aldrich P6887) in 10 mmol HCL, dehydrated in an ethanol series (70%, 90%, and 100% v/v) and air-dried. The rabbit anti-HEI10 polyclonal antibodies (kindly provided by Mathilde Grelon, INRAE Versailles) were used at a dilution of 1% v/v in 1× PBS-T-BSA. The slides were incubated overnight at 4 °C. After 3 rinses in 1× PBS-T, slides were incubated for 1 h at room temperature with a labeled secondary antibody [labeled goat anti-rabbit IgG Alexa fluor 488-Invitrogen (Invitrogen, ref. A-11008) diluted at 1/200 in 1× PBS-T-BSA]. After 3 rinses in 1× PBS-T, the slides were mounted in Vectashield containing 2.5 µg mL⁻¹ of DAPI (Vector Laboratories, H-1200). Fluorescence images were captured using an ORCA-Flash4 (Hamamatsu, Japan) on an Axioplan 2 epifluorescence microscope (Zeiss, Oberkochen, Germany) using the following parameters: DAPI: excitation 353 nm, emission 465 nm, and exposure time 50 ms; Alexa 488: excitation 493, emission 517 nm, and exposure time 400 ms. Pictures were analyzed using Zen PRO software (version 2, Carl Zeiss, Germany).

GISH-like at meiosis using a genome-specific BAC clone

For observation of HEI10 foci on A bivalents only, the BAC clone BoB014O06 from B. oleracea (Howell et al. 2002) was used as a probe for the C genome on the same cells. This GISH-like BAC hybridized specifically to regions on every C-genome chromosome in B. napus (Książczyk et al. 2011). The BoB014O06 probe was labeled by random priming with biotin-14-dUTP (Invitrogen, Life Technologies, 10553853). The hybridization mixture, consisting of 50% w/v deionized formamide, 10% w/v dextran sulfate, 2× SSC, 1% w/v SDS, and labeled probes (200 ng per slide), was denatured at 92 °C for 6 min and transferred to ice. The denatured probe was placed on the chromosome preparation, and in situ hybridization was carried out overnight in a moist chamber at 37 °C. After hybridization, slides were washed for 5 min in 50% formamide in 2× SSC at 42 °C, followed by 2 washes in 4× SSC-Tween. Biotinylated probe was immunodetected by Texas Red avidin DCS (Vector Laboratories, A-2016), and the signal was amplified with biotinylated anti-avidin D (Vector Laboratories, BA-0300). The chromosomes were mounted and counterstained in Vectashield containing 2.5 µg mL⁻¹ DAPI (Vector Laboratories, H-1200). Fluorescence images were captured using an ORCA-Flash4 on an Axioplan 2 epifluorescence microscope (Zeiss) using the following parameters: DAPI: excitation 353 nm emission 465 nm, exposure time 150 ms; Texas Red: excitation 280 nm, emission 618 nm, and exposure time 150 ms. Picture were analyzed using Zen PRO software (version 2, Carl Zeiss).

Statistical analyses

Comparison of the CO rates between progeny was assessed for every interval between consecutive SNP markers using a 2-by-2 χ² analysis considering a significance threshold of 5%. Additionally, these comparisons were also performed for a given heterozygous region, at the chromosome and at the genome scales using 2-by-2 χ² tests. For these tests, a conservative Bonferroni-corrected threshold of 5% was applied, using either the number of regions, the number of chromosomes, or the number of intervals between adjacent SNP markers per A chromosome, or for the whole A subgenome (as described in Pelé et al. 2017 and Boideau et al. 2021). Statistical comparisons of the interference strength were assessed using the software CODA (Gauthier et al. 2011). Statistical differences for the number of HEI10 foci between the different hybrids were assessed using the MWW statistical test and considering a significance threshold of 5%, with a conservative Bonferroni-corrected threshold of 5%. For the analysis of RT-qPCR results, 1-way ANOVA with Tukey post hoc tests, with a threshold of 5%, was performed independently for each gene Supplementary Data Set 3.

All statistical analyses were realized using RStudio (version 3.6.1, R Core Team, 2021).

Plant material and RNA extraction used for RT-qPCR

In total, 9 plants of B. rapa cv. Chiifu, B. napus cv. Darmor, and the 3x-F1 allotriploid hybrid were placed in a controlled growth chamber [long day condition, 16 h of light at 21 °C (fluorescent lamp, 135 µmol m⁻² s⁻¹), followed by 8 h of dark at 18 °C] on the same day after vernalization for 4 (B. rapa cv. Chiifu) and 8 weeks (B. napus cv. Darmor and the allotriploid). Plants were in 2-L pots filled with a nonfertilized commercial substrate (Falienor, reference 922016F3) and were irrigated twice a week with a commercial fertilized solution (Liquoplant Blue, 2.5% nitrogen, 5% phosphorus, 2.5% potassium, w/v). Three blocks, each one containing 3 biological replicates (cuttings) of each genotype (further pooled), were placed randomly within a block to avoid positional bias. As the size of the floral buds differs between species and the allotriploid hybrid, we first determined at which meiotic stage, most cells were using acetocarmine staining of PMCs according to the floral bud size (testing buds of 0.5 to 1, 1 to 1.5, and 1.5 to 2 mm). This preliminary analysis revealed that bud size of 0.5 to 1 mm for B. rapa cv. Chiifu and of 1 to 1.5 mm for B. napus cv. Darmor and the F1 allotriploid hybrid 3x-F1 contained a majority of PMCs at the stage of Prophase I, which corresponds to the meiotic stage encompassing most of the recombination events from DSBs to CO maturation. For each replicate per genotype, we harvested 50 flower buds of the predefined size that we immediately froze in liquid nitrogen and stored at −80 °C upon further processing for RT-qPCR analysis. These floral buds were harvested between 9 and 11 AM to prevent potential physiological modifications due to the circadian clock of plants that can impact meiotic processes. Additionally, we collected 2 flower buds per genotype that were fixed in Carnoy II (6 vol. ethanol:3 vol. chloroform:1 vol. glacial acetic acid; Carnoy 1886) for a day at room temperature and then conserved in 50% ethanol at 4 °C. Acetocarmine spreading of 3 anthers per individual of a given genotype was then performed to confirm that the majority of the analyzed PMCs were at the Prophase I stage.

Total RNA was extracted using the NucleoSpin RNA Plus Kit (Macherey Nagel, 740984.50) from frozen tissues ground into a fine powder following the manufacturer's instructions. Potentially remaining DNA traces were removed using DNase I RNase-free (Thermo Fisher Scientific, EN0521). The quality of each RNA sample was assessed on a 1.5% denaturing agarose gel stained with ethidium bromide and using a NanoDrop 2000 Spectrophotometer (Thermo Fisher Scientific). After verifying the absence of DNA contamination in each sample by PCR, first-strand cDNA synthesis was performed. In each reaction, 2 μg of total RNA per sample was reverse-transcribed using oligo(dT)₁₈ primer (Thermo Fisher scientific, SO132) and Goscript reverse transcriptase M-MLV RT [H−] Point Mutant (Promega, M3681) following the manufacturer's instructions. Samples were then kept at −20 °C.

RT-qPCR assay

Four reference genes previously used for RT-qPCR experiments (Zhu et al. 2013; Wang et al. 2016) and shown to have highly stable expression levels (Czechowski et al. 2005) were tested for performance in this study. This included: F-box (F-box/kelch-repeat protein: AT5G15720), PEPKR1 (phosphoenolpyruvate carboxylase-related kinase 1: AT1G12580), SAND (SAND family protein: AT2G28390), and UNK1 (hypothetical protein: AT1G31300). To identify the putative orthologous sequences of these A. thaliana genes in B. rapa cv. Chiifu (v4: Zhang et al. 2023) and B. napus cv. Darmor (v10: Rousseau-Gueutin et al. 2020), we recovered the amino acid sequences of these 2 individuals (plus A. thaliana) and used orthofinder (Emms and Kelly 2019), as well as performed BlastP analyses (v 2.9.0, Altschul et al. 1990) for validation. Thereafter, we aligned the Brassica nucleotide sequences for each given gene using MAFFT (available in Geneious Prime 2022.2.2) to design primer pairs for the RT-qPCR experiment. As the purpose of this RT-qPCR experiment is to compare the global level of transcription of a given gene between Brassica individuals of different ploidy levels, it is critical that the primers share the same amplification efficiency in all individuals and for all gene copies. To that end, we designed primer pairs (using NetPrimer: https://www.premierbiosoft.com/netprimer/) in regions that were identical between B. rapa cv. Chiifu and B. napus cv. Darmor gene copies (absence of indels in the amplified region between copies or species), preventing any amplification bias. The geNorm method (Vandesompele et al. 2002) was used to validate whether the reference genes had stable expression under the conditions tested in this study. The efficiency of RT-qPCR primers was calculated from a serial dilution (1:20, 1:40, 1:80, 1:160, 1:320) of pooled biological replicates per genotype from our experiments, as described in Hellemans et al. (2007). Simply, a calibration curve for a dilution series of a pooled sample containing an equal fraction of all cDNA samples per genotype in the experiment was performed with Cp vs. log(dilution). The slope was determined by linear regression analysis and the efficiency of RT-qPCR primers “E” was determined as: E = 10^(−1/slope).

We then wanted to compare the global level of transcription (i.e. all copies from the A and C subgenomes) of several meiotic genes between the allotriploid with either B. rapa or B. napus: ASY3 (AT2G4680), DMC1 (AT3G22880), HEI10 (AT1G53490), MLH3 (AT4G35520), MSH4 (AT4G17380), MUS81 (AT4G30870), TOP3α (AT5G63920), ZYP1A and ZYP1B (AT1G22260 and AT1G22275). For this purpose, we followed the same methodology that is described above for the reference genes in order to identify the putative orthologous copies in Brassica and design the primer pairs. The putative orthologous copies in B. rapa cv. Chiifu and B. napus cv. Darmor of each A. thaliana genes used for the RT-qPCR experiment are given in Supplementary Data Set 4.

For RT-qPCR experiments, 4 μL of diluted cDNA (1:25) per sample was mixed with 5 µL of the 2× LightCycler 480 SYBR Green I Master mix (Roche, 4887352001) and 0.5 µL of each primer (10 µm) in a total volume of 10 µL. The RT-qPCR reactions were achieved on a Light Cycler LC480 (Roche) under the following conditions: an initial step at 95 °C for 10 min, then 50 cycles of 95 °C for 10 s, 60 °C for 10 s, and 72 °C for 20 s. All RT-qPCR experiments were carried out with 3 technical replicates of 3 biological replicates (apart from B. rapa cv. Chiifu which only had 2 biological replicates). The analysis of relative gene expression was conducted according to the methods described in Remans et al. (2014). Briefly, the relative expression of each gene was calculated using the following formula: E^{(Cp mean−Cp sample)}, where Cp mean corresponds to the average Cp value of all technical and biological replicates of all genotypes, and Cp sample refers to the average Cp value of a single biological replicate of a single genotype. The relative expression level of the genes of interest in each biological replicate was then normalized against 4 validated reference genes by dividing the gene of interest relative expression value by the reference gene normalization factor (i.e. the geometric mean of the 4 individual reference gene relative expression values, as calculated in geNorm: Vandesompele et al. 2002).

Plant material, nuclei isolation, and Hi-C assay for chromatin compaction

For the 2 replicates of B. rapa cv. Chiifu and the 3 replicates of B. napus cv. Darmor and the allotriploid hybrid, 1 g of fresh newly formed leaves below the inflorescences was collected. Crosslinking was performed by immediately immersing the leaves in a 2% v/v formaldehyde solution followed by 2 cycles of infiltration under vacuum for 15 min. Treated leaves were slightly dried with absorbent paper, frozen immediately in liquid nitrogen, and stored at −80 °C. For Hi-C sequencing, whole plant nuclei were isolated using the CelLytic Plant Nuclei Extraction kit (Sigma-Aldrich, CELLYTPN1), and DNA was digested using the restriction enzymes DpnII and HinfI from the Arima HiC kit (Arima Genomics, A510008) and the libraries were prepared using the Accel-NGS 2S Plus DNA Library Kit (Swift Biosciences, 21024), following the manufacturer instructions. Hi-C contact matrices were generated using nextflow (v23.10.1, Di Tommaso et al. 2017) nf-core (Ewels et al. 2020) and hic pipeline (v2.1.0-gfe4ac65, Servant et al. 2015). The mapping of the reads was carried out on the B. napus cv. Darmor v10 genome (Rousseau-Gueutin et al. 2020). The arima preset was selected for digestion step (–digestion arima) as the Arima-Genomic kit was used for DNA digestion. Compartments were called with a 100 kb resolution (–res_compartments 100000). All other parameters were set to default. Because the sign (positive or negative) of the eigenvector is arbitrarily defined for each chromosome, we defined a consistent sign across different chromosomes, replicates, and genotypes by correlating the eigenvector with the positions of the centromeres defined in Boideau et al. (2022), changing sign if necessary to ensure that negative values are associated with the closed (B) compartment. For each genotype, only the genomic 100 kb bins showing a concordant eigenvector sign across all the replicates were kept for further analysis. The comparison between genotypes was carried out by calculating the percentage of genomic bins with the same eigenvector sign between the 2 genotypes considered.

Accession numbers

Hi-C sequencing data are available in the European Nucleotide Archive under the project PRJEB76232. All other data analyzed in this study are provided as Supplementary Material. All gene accession numbers analyzed in this study are available in Supplementary Data Set 4.

Author contributions

F.B., A.-M.C., and M.R.-G. designed the study. F.B. performed the genetic maps and statistical analyses. L.M. contributed to the automatization of the genetic map establishment. A.B. and V.H. performed the HEI10 experiments. A.B., V.H., G.T., O.C., and F.B. analyzed the HEI10 images. G.T. and V.H. performed and analyzed the MI BAC FISH meiotic behavior. F.E. performed the classical meiotic behavior. M.L.-T. took care of the DNA samples. M.L.-T., G.D., C.F., and F.B. analyzed the genotyping data. M.R.-G. identified orthologous copies of Arabidopsis genes used for the RT-qPCR experiment and designed the primer pairs. A.-M.C., C.F., F.B., G.D., G.R., L.M., M.L.-T., M.R.-G., O.C., and V.H. were involved in the collection of flower buds for the RT-qPCR experiment. G.D., C.Ba., and C.F. participated in the generation of the RT-qPCR results that were analyzed by G.D., C.Ba., C.F., M.R.-G., and F.B. J.-M.A., K.L., and C.Be. generated the Hi-C data. L.M. and G.R. analyzed the Hi-C results. M.G. generated and took care of the plant material. J.B. designed the additional pericentromeric SNP. J.M. generated the Circos representations. M.F. and O.M. analyzed the interference. F.B., M.R.-G., and A.M.-C. wrote the manuscript. All authors approved the final version of the manuscript.

Supplementary data

The following materials are available in the online version of this article.

Supplementary Figure S1. Improvement of the B. napus cv. Darmor genome assembly.

Supplementary Figure S2. The pairwise comparison of the genetic map length between the different hybrids.

Supplementary Figure S3. Example of the identification of the A and B compartments along chromosome A02 in a Brassica allotriploid hybrid and its parental lines using Hi-C.

Supplementary Table S1. HEI10 foci quantification used for Fig. 4.

Supplementary Data Set 1. Sequences of the KASPar markers used.

Supplementary Data Set 2. Genotyping data used to generate the genetic maps.

Supplementary Data Set 3. Statistical analysis of the RT-qPCR results.

Supplementary Data Set 4. The list of RT-qPCR primers used in this study.

Funding

This work was made possible by the financial support of the Agence Nationale pour la Recherche (ANR) “Stirrer” (ANR-19-CE20-001 awarded to M.R.-G.). This project also received funding from Institut National de la Recherche Agronomique (INRAE) “Biology and Plant Breeding” department to perform part of the experiments presented here (projects “Recipe” and “Recoptic” awarded to M.R.-G. and A.-M.C., respectively), as well as to provide part of F.B.'s PhD salary. The authors also thank the “Conseil Régional de Bretagne” that financed the second half of F.B.'s PhD salary (project “Diverse”) and the post doc fellowship of G.R. through the “SAD” call. GQE and IPS2 benefit from the support of Université Paris-Saclay (SPS) (ANR-17-EUR-0007).