The poor lonesome A subgenome of Brassica napus var. Darmor (AACC) may not survive without its mate

Transcription

1 Research The poor lonesome A subgenome of Brassica napus var. Darmor (AACC) may not survive without its mate Alexandre Pele*, Gwenn Trotoux*,Frederique Eber, Maryse Lode, Marie Gilet, Gwenaelle Deniot, Cyril Falentin, Sylvie Nègre, Jer^ome Morice, Mathieu Rousseau-Gueutin and Anne-Marie Chèvre IGEPP, INRA, Agrocampus Ouest, Universite de Rennes 1, Le Rheu, France Author for correspondence: Anne-Marie Chevre Tel: anne-marie.chevre@inra.fr Received: 26 April 2016 Accepted: 12 July 2016 doi: /nph Key words: allopolyploidy, evolution, oilseed rape, structural rearrangements, subgenome extraction. Summary Constitutive genomes of allopolyploid species evolve throughout their life span. However, the consequences of long-term alterations on the interdependency between each original genome have not been established. Here, we attempted an approach corresponding to subgenome extraction from a previously sequenced natural allotetraploid, offering a unique opportunity to evaluate plant viability and structural evolution of one of its diploid components. We employed two different strategies to extract the diploid AA component of the Brassica napus variety Darmor (AACC, 2n =4x = 38) and we assessed the genomic structure of the latest AA plants obtained (after four to five rounds of selection), using a 60K single nucleotide polymorphism Illumina array. Only one strategy was successful and the diploid AA plants that were structurally characterized presented a lower proportion of the B. napus A subgenome extracted than expected. In addition, our analyses revealed that some genes lost in a polyploid context appeared to be compensated for plant survival, either by conservation of genomic regions from B. rapa, used in the initial cross, or by some introgressions from the B. napus C subgenome. We conclude that as little as c yr of coevolution could lead to subgenome interdependency in the allotetraploid B. napus as a result of structural modifications. Introduction Polyploidy, or whole genome duplication (WGD), is recognized as a major driving force in eukaryote evolution and speciation, particularly in plants (Comai, 2005; Cui et al., 2006; Van de Peer et al., 2009; Soltis & Soltis, 2012). It is now established that all angiosperms have experienced at least one event of WGD in their evolutionary history (Soltis et al., 2009; Jiao et al., 2011). These WGD events occurred by genome doubling, derived from the same (autopolyploidy) or related (allopolyploidy) species (Stebbins, 1947). Thus, the immediate and important genomic consequence of polyploidy is the simultaneous duplication of nuclear genes (gene redundancy) which, associated with a diploidization process, leads to the occurrence of a new species and contributes to diversification and adaptation of flowering plants (Wolfe, 2001). This diploidization process is responsible for nonadditive parental patterns identified for some duplicated genes through extensive structural and functional genome reorganization, particularly in allopolyploids (Otto, 2003; Albertin et al., 2005, 2006; Tayale & Parisod, 2013). Thus, duplicated genes may be subjected to physical loss (nonreciprocal exchanges, genic conversion and fractionation), functional loss (pseudogenization), *These authors contributed equally to this work. and/or functional diversification (sub- and neofunctionalization) resulting from an evolution in a polyploid context (Ohno, 1970; Wendel, 2000; Adams, 2007; Woodhouse et al., 2010; Sankoff et al., 2012; Yoo et al., 2014). Although structural and functional alterations were highlighted in allopolyploid flowering plants, the dynamic undergone by merged genomes (i.e. subgenomes) throughout their evolution remains unclear. In this way, short-term alterations observed in the first few generations following WGD events and long-term alterations taking place during the life span of the polyploid must be distinguished (Feldman et al., 2012). Short-term alterations have received considerable interest in recent years and have been studied by analyzing neopolyploids, either resynthesized artificially (e.g. Brassica napus, Triticum aestivum) or naturally formed (e.g. Spartina anglica, Tragopogon mirus), enabling the immediate impacts of allopolyploidy to be deciphered (Tayale & Parisod, 2013; Ainouche & Wendel, 2014). In all cases, functional modifications were detected after genome merging, in the presence of either a global genomic stasis (e.g. Gossypium hirsutum, Spartina anglica) or major structural rearrangements (e.g. B. napus, Nicotiana tabacum) (Tayale & Parisod, 2013; Ainouche & Wendel, 2014). Unlike neopolyploids, study of long-term alterations does not allow an assessment of the stabilization process in allopolyploids, but instead offers a final view of structural and 1886

2 New Phytologist Research 1887 functional genome reorganization, by comparing sequence and/ or expression in a natural allopolyploid and the current forms of its progenitors (Chalhoub et al., 2014; He et al., 2015). However, the progenitors of allopolyploids have evolved independently since the polyploid formation and may not truly reflect what happened in natural allopolyploids. To circumvent this problem, another approach corresponding to subgenome extraction offers a unique opportunity to understand long-term subgenome evolution in an allopolyploid context. The use of such an approach seems possible in recent allopolyploids only under the assumption that evolution of the associated subgenomes does not lead to interdependency. So far, this strategy has been undertaken in the hexaploid bread wheat (T. aestivum, AABBDD, 2n = 6x = 42) by extracting the tetraploid component (AABB, 2n = 4x = 28). It reveals that the evolutionary history at the allohexaploid level in this species did not alter karyotype structure but induced gene expression modifications (Kerber, 1964; Galili & Feldman, 1984; Zhang et al., 2014; Liu et al., 2015). However, using a current parental species, such an approach has never been applied to the extraction of a diploid component originating in an allotetraploid, and even less so in an allopolyploid species presenting major structural rearrangements. In this context, the allotetraploid B. napus (A n A n C n C n, 2n = 4x = 38), which was formed at a period similar to the hexaploid bread wheat (< yr ago), is particularly relevant (Nesbitt & Samuel, 1995; Cheung et al., 2009; Chalhoub et al., 2014). B. napus results from the natural hybridization of B. rapa (A r A r,2n = 2x = 20) and B. oleracea (C o C o,2n = 2x = 18), which both derived from a common hexaploid ancestor (Nagaharu, 1935; Tang et al., 2012; Cheng et al., 2013; Jenczewski et al., 2013). The ability to artificially resynthesize B. napus has helped to make it a model for the study of allopolyploidy (Olsson, 1960). B. napus is known to have experienced frequent structural and functional modifications in the first few generations following allopolyploidization (Song et al., 1995; Pires et al., 2004; Albertin et al., 2006; Gaeta et al., 2007; Xu et al., 2009; Gaeta & Pires, 2010; Marmagne et al., 2010; Szadkowski et al., 2010, 2011; Xiong et al., 2011). In addition, the recent genome sequencing of B. napus cv Darmor and the current form of its progenitors revealed a high frequency of genic conversion, nonreciprocal exchanges between homoeologous chromosomes, as well as gene loss and gain during its evolutionary development (Wang et al., 2011; Chalhoub et al., 2014; Liu et al., 2014; Parkin et al., 2014). However, the question still remains of the impact of these long-term alterations on the interdependency of each original genome in a polyploid context. In the current study, the diploid A n A n component of the recently sequenced allotetraploid B. napus cv Darmor was extracted, enabling us to determine if its evolution since the polyploidization event may impact plant viability at the diploid stage when deprived of the homoeologous C n subgenome. From an initial cross between B. napus and B. rapa, extraction of the diploid A n A n component was achieved by a process involving triploid steps (AAC, 2n = 3x = 29). Two different strategies were utilized, but only one enabled the successful production of diploid hybrids for which the latest generations (after four or five rounds of selection) were characterized using a high-density single nucleotide polymorphism (SNP) array (60K). We determined that the diploid AA plants had a lower proportion of B. napus A n subgenome than expected, as well as some introgressions from the C n subgenome. Additionally, we showed that this result may be due to the compensation of genes lost since the polyploidization event, suggesting that c yr of coevolution is sufficient to prevent the diploid A n A n component of B. napus from surviving on its own. Materials and Methods Strategies developed to extract the diploid A n A n component of Brassica napus Two strategies were used to extract the diploid A n A n component of the allotetraploid Brassica napus L. var. oleifera Darmor (A n A n C n C n, 2n = 4x = 38), a winter cultivar that has been recently sequenced by Chalhoub et al. (2014). Both of these strategies, presented in Fig. 1, were initiated by crossing the diploid B. rapa var. rapifera C1.3 (A r A r,2n = 2x = 20), which is an old nonhomogeneous French forage variety, with B. napus cv Darmor (male) from seeds provided by the CRB BrACySol (Rennes, France). This resulted in F 1 triploid hybrids (A n A r C n, 2n = 3x = 29), for which the presence of 29 chromosomes, corresponding to 10 A bivalents and nine C univalents, was verified by the establishment of the meiotic behavior. In the first strategy, an F 1 triploid hybrid was backcrossed as female four times to B. napus cv Darmor, following the backcrosses procedure of Kerber (1964). These steps enabled a decrease in the amount of genetic material contributed by B. rapa cv C1.3 in the initial cross. At each generation, the chromosome number of each progeny was determined using flow cytometry and the putative triploid plants were confirmed to be AAC hybrids (2n = 3x = 29) in metaphase I of meiosis. The last triploid hybrid was self-pollinated in order to remove the C subgenome and obtain a diploid AA hybrid, theoretically presenting an almost pure B. napus A n subgenome (96.9%). In the second strategy, an F 1 triploid hybrid was backcrossed as female to B. rapa cv C1.3. Progenies were also characterized cytogenetically to select diploid AA hybrids with 10 A bivalents, which were supposed to represent 25% of the B. napus A n subgenome. In order to increase the proportion of B. napus A n subgenome, a two-step method was employed. The first step corresponded to crossing a diploid hybrid obtained earlier with B. napus cv Darmor (male), producing triploid hybrids. In the second step, a newly generated triploid hybrid was crossed to the selfed diploid hybrid selected from the previous generation. This two-step method was performed over a total of five successive generations, with the different diploid (hereafter named A n/r A n/r ) and triploid hybrids characterized cytogenetically. Crosses were always performed by hand pollination, and seed number per pollinated flower was assessed.

3 1888 Research New Phytologist Fig. 1 Schematic representation of the two strategies used to extract the diploid A n A n (2n = 20) component of Brassica napus cv Darmor (A n A n C n C n,2n = 38) from an initial cross with Brassica rapa cv C1.3 (A r A r, 2n = 20). Circles and squares represent one copy of the A and C subgenomes, respectively. The red color represents the theoretical proportions of B. napus extracted, whereas the blue color indicates the theoretical proportions of B. rapa remaining. Dashed arrows represent selfing events. Cytogenetic characterization of hybrids The chromosome number of each of the AACC 9 AAC (strategy 1) and AAC 9 AA (strategy 2) progenies was determined at the seedling stage by flow cytometry following the protocol of Leflon et al. (2006) with a precision of two chromosomes. Subsequently, young floral buds from either AAC or AA putative hybrids were harvested in order to characterize their meiotic behavior as described in Suay et al. (2014) from 20 pollen mother cells (PMCs) per plant at metaphase I of meiosis. DNA preparation and genotyping Genomic DNA was extracted from the young leaves of the parents initially crossed (B. napus cv Darmor and B. rapa cv C1.3), an F 1 triploid hybrid (A n A r C n ), as well as three A n/r A n/r diploid hybrids produced in the second strategy from the fourth (A n/r A n/r G4, one plant) and fifth generations (A n/r A n/r G5, two plants), following the protocol described in Lombard & Delourme (2001). Three technical replicates were included per initial parent and genotyping was performed using the Brassica 60K Illumina Infinium SNP array ( This array contains SNPs distributed across the A and C chromosomes. Hybridizations were run according to the standard procedures provided by the manufacturer. The genotyping data obtained were visualized with GENOME STUDIO v (Illumina Inc., San Diego, CA, USA) and processed using a manually adapted cluster file. Identification of the B. napus A n subgenome proportion extracted The genotyping data obtained with the 60K SNP array offered the opportunity to identify the proportion of B. napus extracted for each of the diploid A n/r A n/r hybrids generated from the fourth and fifth generations of the second strategy. For that purpose, the sequence contexts (size ranging between 96 and 301 bp) of polymorphic SNPs between the parents crossed initially (B. napus cv Darmor and B. rapa cv C1.3) were retrieved (7846 SNPs). These sequence contexts were then blasted (Altschul et al., 1990) against the B. napus cv Darmor genome sequence version 4.1 (Chalhoub et al., 2014). Only those presenting a single hit on either the A or

4 New Phytologist Research 1889 C subgenomes with a minimum of 90% global overlap and 90% identity were retained. This last step enabled the removal of SNPs that may be amplified by paralogous sequences. The physical locations of the SNPs were then compared with their genetic positions obtained from a B. napus Darmor-Yudal genetic map (268 doubled haploid lines, C. Falentin, unpublished data) generated using the same 60K SNP array. Only those SNPs having a concordant physical and genetic location on the same A chromosome were kept, allowing us to discard SNPs that may have been amplified by homoeologous C subgenome sequences. Using this method, a total of 2290 polymorphic SNPs localized on the A subgenome were retrieved, allowing us to determine the homozygous (A n /A n or A r /A r ) or heterozygous (A n /A r ) state of each SNP marker within each diploid A n/r A n/r hybrid genotyped. The presence of at least two consecutive SNPs on a chromosome belonging to the same type (either A n /A n or A r /A r or A n /A r ) was considered as a single block. These blocks, as well as the centromere positions (Mason et al., 2016), were graphically represented along the 10 A chromosomes using CIRCOS (Krzywinski et al., 2009). For each A n /A n, A r /A r and A n /A r block within an A n/r A n/r hybrid, the physical size as well as the number of polymorphic SNPs were determined. Similarly, the number of genes present in each block was identified using the B. napus annotation v.4.1 (Chalhoub et al., 2014). The proportion of B. napus A n subgenome and B. rapa A r genome per A n/r A n/r hybrid was calculated for the A genome and per A chromosome using the physical size of the blocks as well as the polymorphic SNP and B. napus gene content in the blocks. Finally, from the SNP and gene content calculated per type block, the hypothesis H0 The proportion of B. napus A n subgenome extracted in A n/r A n/r hybrids is similar to that expected in case of a random segregation without biases (68.4% and 76.3% in fourth and fifth generations, respectively) was tested by a v² test. The hypothesis H0 The proportion of B. napus A n subgenome extracted in A n/r A n/r hybrids is homogeneously spread within the 10 A chromosomes was also tested by a v² test. Impact of the subgenome evolution on the B. napus A n subgenome extraction The subgenome evolution in a polyploid context, and notably gene losses, may prevent some genomic regions from being fixed in a B. napus homozygous state (A n /A n ) in the diploid A n/r A n/r hybrids genotyped. The distribution of the 1925 putative genes lost on A n subgenome (hereafter called A n genes), identified by Chalhoub et al. (2014) (corresponding to i.e. gene loss, truncation, pseudogenization), was studied. These A n genes as well as their adjacent genes were first identified in the B. rapa A r genome. The physical locations of the adjacent genes were then determined on the B. napus A n subgenome in order to infer the approximate location of the A n genes. Consequently, the number of A n genes present in all the homozygous B. napus regions (A n /A n ) identified in at least one of the A n/r A n/r hybrids was determined. Similarly, the number of A n genes present in all the B. rapa regions (including A r /A r and A n /A r ) common to the A n/r A n/r hybrids was identified. The hypothesis H0 The genomic regions fixed in a B. napus homozygous state (A n /A n ) during the production of A n/r A n/r hybrids are distributed independently of the distribution of A n genes was tested by a v² test. Identification of putative B. napus C n subgenome introgressions The genotyping data provided by the 60K SNP array, for each of the diploid A n/r A n/r hybrids genotyped, offer the opportunity to detect putative introgressions of the B. napus C n subgenome that occurred during the creation of the plant material. Within each A n/r A n/r hybrid, the SNPs amplifying in B. napus cv Darmor but not in B. rapa cv C1.3 were identified from the genotyping data. Their physical locations were compared with their genetic positions obtained from a B. napus Darmor Yudal genetic map (C. Falentin, unpublished data). The amplification of the 4807 SNPs presenting a concordant physical and genetic location on the same C chromosome was examined in each A n/r A n/r hybrid. Molecular and sequencing validations of putative B. napus C n subgenome introgressions To validate the presence of the putative B. napus C n subgenome introgressions in each diploid A n/r A n/r hybrid genotyped, several primer pairs were designed using Primer 3.0 (Rozen & Skaletsky, 2000). Subsequently, using Blastn (Altschul et al., 1990) against the B. napus cv Darmor genome sequence (version 4.1, Chalhoub et al., 2014), only the primer pairs yielding a single PCR product specific to each putative introgressed region were retained. The specificity of these primers to the B. napus C n subgenome was validated molecularly using two B. rapa (C1.3 and Chiifu), two B. oleracea (HDEM and RC34) and a B. napus (Darmor). Only the primers amplifying in B. oleracea and B. napus but not in B. rapa were conserved (Supporting Information Table S1), and thereafter tested on each A n/r A n/r hybrid. Each PCR amplification was performed in a total volume of 50 ll containing 10 ll of59 GoTaq Flexi buffer (Promega), 4 llof25mmmgcl 2, 0.5 llof 25 mm deoxyribonucleotide mix (Promega), 2.5 ll of each primer (10 mm), 0.2 ll of GoTaq G2 Hot Start polymerase (5 U ll 1 ) and 50 ng of template DNA. Cycling conditions were 94 C for 2 min, followed by 30 cycles of 94 C for 30 s, 60 C for 30 s and 72 C for 60 s, and a final extension of 72 C for 5 min. The PCR products were analyzed by agarose gel electrophoresis (2%) and then sequenced directly in both directions (Genoscreen, Lille, France). Results The long way for the extraction of B. napus A n subgenome The initial F 1 A n A r C n hybrids, produced by crossing B. rapa cv C1.3 (A r A r,2n = 2x = 20) and B. napus cv Darmor (A n A n C n C n, 2n = 4x = 38), exhibited regular meiotic behavior close to that expected, with 64 83% of cells with 10 bivalents and nine univalents in metaphase I of meiosis (Fig. 2; Table S2). These hybrids produced seeds after backcrossing to B. napus as well as after crosses to B. rapa (Fig. 3a).

5 1890 Research New Phytologist Fig. 2 Evolution of the percentages of pollen mother cells (PMCs) in Brassica allotriploid hybrids showing the expected regular meiotic behavior, with 10 bivalents (II) and nine univalents (I), in metaphase I along the different generations in strategies 1 (in blue) and 2 (in red). Bars represent SE. For the first strategy (Fig. 1), as developed for the extraction of the AABB genomes from the hexaploid bread wheat (AABBDD) by Kerber (1964), an F 1 A n A r C n hybrid was backcrossed four times with B. napus to select AAC hybrids at each generation. The plants produced at each generation from backcrosses were screened by flow cytometry, and putative AAC plants were checked in metaphase I of meiosis. We determined that the percentage of gametes with only the 10 A chromosomes, able to produce AAC hybrids after backcrosses to B. napus, was always low and ranged from 1.8% to 3.5%. For that reason, a large number of the available seeds was grown (from 57 to 175), enabling the detection of only two to four AAC plants per generation. Their meiotic behavior was less regular in advanced generations, with a decrease in the percentage of cells with the expected structure: 10 bivalents and nine univalents (Fig. 2; Table S2). Similarly, in spite of the numerous backcrosses to B. napus, the number of seeds produced by the different AAC plants was lower in advanced generations (from 355 seeds per 100 pollinated flowers for the F 1 to 49 for the G3, Fig. 3a), indicating a decrease in female fertility of AAC hybrids. Fertility was drastically reduced in the last (fourth) generation, in which AAC plants produced only four seeds (five seeds per 100 flowers pollinated). These plants had poor vigor and were totally sterile after selfing, preventing the elimination of C chromosomes for the production of AA plants with a practically pure B. napus A n genome (96.9% expected). Because of the inability to extract the pure diploid A n A n component of B. napus using the first strategy, we developed another approach, with the generation of AA plants at each generation initiated by crossing an F 1 A n A r C n hybrid with the recurrent B. rapa (Fig. 1). We observed that the gametes with the 10 A chromosomes from AAC hybrids, able to generate AA plants, were rare, ranging from 0.4% to 3.4% in an analysis of 53 to 231 plants per generation. Among the only one to two AA plants obtained at each generation, we showed that they exhibited regular meiotic behavior with always 10 bivalents (Fig. 4). In contrast with the first strategy, the AAC hybrids had a more regular meiotic behavior for each generation, with at least 77% of cells having 10 bivalents and nine univalents (Fig. 2; Table S2). In advanced generations, it was increasingly difficult to produce seeds from AAC hybrids (Fig. 3a) but also from AA plants by selfing or crosses with B. napus (Fig. 3b). In the fifth generation, AA plants had poor vigor and flowers had narrower petals than the initial B. rapa and B. napus parents, combined with a poor pollen fertility (25 73%) (Fig. 4). Fig. 3 Evolution of seed number per 100 pollinated flowers among the different generations from: (a) triploid AAC hybrids obtained with strategies 1 (in blue) and 2 (in red); and (b) diploid plants after selfing (in black) or backcrosses to Brassica napus (in purple) for strategy 2. Bars represent SE. The proportion of B. napus A n subgenome extracted is lower than expected and unevenly spread per chromosome We assessed the proportion of B. napus A n subgenome extracted in the diploid A n/r A n/r hybrids successfully produced from the fourth (one plant) and fifth generations (two plants) with the second strategy. For that purpose, these A n/r A n/r hybrids were genotyped and the 2290 polymorphic SNPs between the parents crossed initially were used to design B. napus and B. rapa homozygous (A n /A n and A r /A r ) as well as heterozygous (A n /A r )

6 New Phytologist Research 1891 Fig. 4 Characterization of one diploid A n/r A n/r hybrid produced from the fifth generation of the second strategy for: (a) meiotic behavior showing 10 bivalents; (b) pollen fertility (fertile grains stained in red by Aceto- Carmine); and (c) flower morphology (left, Brassica napus; middle, A n/r A n/r hybrid; right, Brassica rapa). (a) (b) (c) blocks physically localized along the 10 A chromosomes (Table S3; Fig. 5). These blocks, covering c. 80% of the B. napus A n subgenome sequence, allowed us to identify that 60.4% of the B. napus A n subgenome was extracted in the fourth generation and from 69.0% to 70.8% in the two plants of the fifth generation (Table 1). The proportion of B. napus A n subgenome extracted was lower in the A n/r A n/r hybrids from the fourth than from the fifth generation, indicating the continuous decrease of B. rapa genetic material. Similar results were found for the proportion of B. napus A n subgenome extracted using the block content in polymorphic SNP and B. napus genes (Table 1). We compared the observed proportions of B. napus A n subgenome extracted and B. rapa A r genome remaining in each A n/r A n/r hybrid genotyped with the expected values after four and five generations (68.4% and 76.3% of the B. napus A n subgenome expected in G 4 an G 5 respectively; Fig. 1). For each A n/r A n/r hybrid, we noted a significant difference in both SNP and gene block contents between observed and expected values (v² test, df = 1, P < 0.001; Table 1). This result involved a nonrandom segregation of parental A genomes during the production of the A n/r A n/r hybrids characterized by a lower proportion of B. napus A n subgenome extracted than expected in both the fourth and fifth generations (8% lower in G 4 and from 5.5% to 7.3% in G 5 ) (Table 1). We also evaluated the proportions of B. napus A n subgenome extracted and B. rapa A r genome remaining, using the designed blocks, within each A chromosome per A n/r A n/r hybrid genotyped. Testing the division of the parental A genomes in each A n/r A n/r hybrid, we identified uneven proportions between the 10 A chromosomes from both SNP and gene block contents (v² test, df = 9, P < 0.001). Indeed, large variations (up to a factor of 3) were noticed between the 10 A chromosomes for the B. napus A n subgenome extracted, particularly between the A05 and A06 chromosomes (Fig. 5). Evolution in a polyploid context may compromise the extraction of B. napus A n subgenome for plant survival at the diploid stage Based on the earlier results, we undertook to clarify why the extraction of B. napus A n subgenome may be prevented in the diploid A n/r A n/r hybrids genotyped. We first observed that B. napus and B. rapa homozygous (A n /A n and A r /A r ) as well as heterozygous (A n /A r ) blocks varied in size from a few hundred to several millions base pairs in each of the A n/r A n/r hybrids (Table S3; Fig. 5). As a consequence of the lack of polymorphic SNPs in centromeric regions, only seven centromeres were entirely included in the blocks designed (Fig. 5). Among these centromeres, five were fixed in a B. napus homozygous state (A n /A n ), regardless of their location along the chromosomes, indicating that the extraction of B. napus A n subgenome was independent of the chromosome conformation (Fig. 5). We noticed that a large proportion of the A genome in each A n/r A n/r hybrid was fixed in a B. rapa homozygous state (A r /A r ): 21.5% in the fourth generation and from 12.7% to 15.5% in fifth generation, while 8.9% and 5% were expected in G 4 and G 5, respectively. The higher proportion of B. rapa A r genome remaining in each A n/r A n/r hybrid genotyped, particularly in a homozygous state (A r /A r ), suggested that some B. rapa genomic regions could be essential for plant survival at the diploid stage. Chalhoub et al. (2014) established that the gene content of the B. napus A n subgenome differed from the B. rapa A r genome with the putative loss of 1925 genes since B. napus formation, hereafter called A n genes. Under the assumption that these A n genes may prevent the fixation in a B. napus homozygous state (A n /A n ) in diploid A n/r A n/r hybrids, we successfully localized 1575 of these genes along the A genome of B. napus (Table S4; Fig. 5), of which 1235 were included in the blocks identified. Then, we compared the division of A n genes in blocks fixed in a B. napus homozygous state (A n /A n ) in at least one of the A n/r A n/r hybrids, and in blocks containing B. rapa (including A r /A r and A n /A r ) common to all A n/r A n/r hybrids. Through this approach, we demonstrated a nonrandom distribution of A n genes along the A subgenome of A n/r A n/r hybrids (v²=5.29, df = 1, P = 0.021), with a higher proportion than expected located in blocks containing B. rapa. Detection of B. napus C n subgenome introgressions We attempted to detect whether introgressions from the B. napus C n subgenome occurred during the creation of the diploid A n/r A n/r hybrids from the fourth and fifth generations of the second strategy. We found that among the 4807 SNPs identified as specific to the B. napus C n subgenome, < 1% amplified in each A n/r A n/r hybrid genotyped. From their physical positions along the B. napus C n subgenome and considering at least the amplification of two consecutive SNPs (i.e. without disruption between each one by nonamplifying SNPs), we determined the presence of three putative introgressions (Table 2). We noticed that 17 consecutive SNPs, covering at least 2.27 Mbp of the C02 chromosome, amplified in each A n/r A n/r hybrid. Similarly, we

7 1892 Research New Phytologist Fig. 5 Circos diagram illustrating the homozygous Brassica napus A n /A n (red) and Brassica rapa A r /A r (blue) as well as the heterozygous A n /A r (light purple) states, of blocks identified along the 10 A chromosomes in A n/r A n/r hybrids produced in the second strategy from the fourth (A n/r A n/r G4, one plant) and fifth generations (A n/r A n/r G5, two plants). The 10 A chromosomes are represented in the first outer circle. The number of polymorphic single nucleotide polymorphisms (SNPs) for each chromosome is indicated in brackets. The size of each chromosome is indicated by a ruler drawn underneath each chromosome with larger and smaller tick marks every 10 and 2 Mbp, respectively. In the second outer circle, the density of genes present in B. napus (A n genes, in red) and lost since its formation (A n genes, in blue) are shown (Chalhoub et al., 2014). Different scales were applied to represent these densities according to the highest number of genes identified within a 250 kb window frame (see values indicated on y-axis in red for A n genes and in blue for A n genes). Active centromeres are indicated in dark grey using the positions established by Mason et al. (2016). In the third outer circle, the distribution of the 2290 polymorphic SNPs, which allowed the identification of homozygous (A n /A n and A r /A r ) and heterozygous (A n /A r ) states of blocks, is shown. determined that 11 SNPs covering 0.53 Mbp of the C01 chromosome, and eight SNPs covering 1.21 Mbp of the C02 chromosome, amplified specifically in the hybrid from the fourth generation and in one plant of the fifth generation, respectively (Table 2). In order to validate the introgression of these B. napus C n subgenome regions, we designed one primer pair specific for

8 New Phytologist Research 1893 Table 1 Characterization of A n/r A n/r hybrids produced from the fourth (one plant) and fifth (two plants) generations of the second strategy with polymorphic single nucleotide polymorphism (SNP) and Brassica napus gene content as well as physical size provided by either A n /A n,a r /A r or A n /A r blocks identified each one. For these primer pairs yielding a PCR product in B. napus and B. oleracea but not in B. rapa, we confirmed their ability to amplify a product in the corresponding A n/r A n/r hybrids. In addition, using a primer pair that sat at the very beginning of the C02 chromosome (Table S1), we redefined the size of the putative introgression common to the A n/r A n/r hybrids from 2.27 to 4.64 Mbp. The amplicons obtained were sequenced and compared with the B. napus C n subgenome sequence (Chalhoub et al., 2014). These sequences presented 100% homology, allowing validation of the presence of all these introgressions from the C01 and C02 chromosomes in the diploid A n/r A n/r hybrids. Discussion A n/r A n/r G4 A n/r A n/r G5 pl1 A n/r A n/r G5 pl2 % A n df v² test % A n Subgenome extraction offers the opportunity to determine if the long-term evolution in an allopolyploid context results in the interdependency of constitutive genomes. In this context, the extraction of a diploid component from the allotetraploid B. napus var. Darmor (A n A n C n C n, 2n = 38), which has been sequenced recently (Chalhoub et al., 2014), is particularly relevant. The best strategy for the extraction of B. napus A n subgenome from crosses with a current B. rapa progenitor The extraction of B. napus A n subgenome at the diploid stage is possible, first, because of the AAC hybrids fertility. This fertility df v² test % A n df v² test SNP content * * * Gene * * * content Physical size Chi-squared P-values are indicated: *, P < Table 2 Identification of putative introgressions from the Brassica napus C n subgenome into the A n/r A n/r hybrids produced from the fourth (one plant) and fifth (two plants) generations of the second strategy (X, B. napus C n subgenome introgression detected) Putative introgressions from C n subgenome Chromosome No. of SNPs Size (Mbp) Start (Mbp) End (Mbp) Detection in A n/r A n/r G4 G5 pl1 C X C X X X C X SNP, single nucleotide polymorphism. G5 pl2 has been reported by several authors either for production of monosomic addition line after backcross with B. rapa (McGrath & Quiros, 1990; Chen et al., 1992; Heneen et al., 2012) or for gene flow assessment from B. napus to B. rapa (Mikkelsen et al., 1996; Metz et al., 1997; Tomiuk et al., 2000) or for oilseed rape improvement (Crouch et al., 1994; Qian et al., 2006). Second, the AAC hybrids produced viable gametes with a binomial distribution of the chromosome number, which ranged from 10 (A genome) to 19 (A and C genomes) as a result of random segregation of the C chromosomes (Lu & Kato, 2001; Leflon et al., 2006). However, only the gametes deprived of C chromosomes can be used for the B. napus A n subgenome extraction. Lu & Kato (2001) reported that their frequency ranged from 10% to 15%, whereas Leflon et al. (2006) never observed more than 5% of such gametes. Our results are in agreement with these data, as we never obtained more than two AA and four AAC plants with strategies 1 and 2, respectively, corresponding to less than 5% of seeds screened in each generation. We first used strategy 1 (Fig. 1), which was originally developed in bread wheat (AABBDD) to extract the AABB components from the initial cross between T. aestivum (AABBDD) and T. durum (AABB). Seven successive backcrosses of AABBD selected hybrids by AABBDD were performed followed by a selfing at the last generation to eliminate D chromosomes (Kerber, 1964). In contrast with this species, we showed a progressive decrease in PMC number in the AAC Brassica hybrids having the expected meiotic behavior (10 bivalents and nine univalents) after a limited number of backcrosses (four) (Fig. 2). Simultaneously, the fertility of AAC hybrids decreased along the generations (Fig. 3a) and the sterility of the AAC hybrid in the last (fourth) generation after selfing excluded the possibility of producing AA plants. An alternative strategy could be to generate AA plants by selfing the AAC hybrids produced in strategy 1. This approach might only be interesting in advanced generations, as our initial objective was to extract the diploid A n A n component of B. napus as purely as possible. However, the seed set after AAC selfing was very limited (zero to six seeds per 100 pollinated flowers), with a decrease in fertility along the generations combined with poor hybrid vigor. Additionally, the probability of extracting AA plants from selfing of AAC hybrids is extremely low, ranging between 0% and 1%, in agreement with previously published results (Lu & Kato, 2001; Leflon et al., 2006), because this requires an elimination of C chromosomes from both the male and female gametes. Given these challenges, we developed a step-by-step method with the generation of AA plants in each generation (Fig. 1). Compared with strategy 1, the meiosis of AAC hybrids in strategy 2 was more regular, with a higher percentage of cells with 10 bivalents and nine univalents (Fig. 2). However, the fertility of AAC hybrids decreased at each generation (Fig. 3a), whereas the fertility of AA plants always remained low, with less than two seeds per pollinated flower (Fig. 3b). Key bottlenecks for the B. napus A n subgenome extraction The main challenge of this work was to extract one component at the diploid stage from crosses with current parental species. Two

9 1894 Research New Phytologist main hypotheses can be proposed to explain the difficulties encountered to extract viable plants. The first hypothesis corresponds to the occurrence of homoeologous exchanges between A and C chromosomes during the meiosis of AAC hybrids, which may prevent the production of A gametes but also the AA plants viability. This hypothesis is supported by the presence of multivalents in the F 1 A n A r C n hybrids in a low percentage of cells (3.7%) (Table S2), as previously reported by several authors (Chen et al., 1992; Hasterok et al., 2005; Leflon et al., 2006). If gametes carrying homoeologous exchanges contribute to the subsequent generations, we can expect a decrease in the number of cells with 10 bivalents and nine univalents. This is relevant to our results in advanced generations, where we observed higher percentages of cells with univalents, trivalents and quadrivalents, revealing the risk of cumulating rearrangements in strategy 1 (Table S2). This situation is different from the one described in pentaploid AABBD wheat hybrids, where the Ph1 locus prevents homoeologous pairing between genomes, ensuring karyotype stability (Riley et al., 1959). Nonetheless, the impact of meiotic stability was not evident with strategy 2, as the AAC hybrids were more regular with a higher percentage of cells having 10 bivalents and nine univalents in advanced generations (Fig. 2). Additionally, the creation of AA plants in each generation may be a good way to eliminate C chromosome rearrangements. This method was not totally successful, as we identified three low size introgressions from the C01 and C02 chromosomes in A n/r A n/r hybrids from the fourth and fifth generations (Table 2). Both C01 and C02 B. napus introgressions were different from those detected by Chalhoub et al. (2014), supporting the idea that these events occurred during the generation of the plant material. These two chromosomes have frequently been described as involved in homoeologous exchanges (Song et al., 1995; Gaeta et al., 2007; Szadkowski et al., 2010, 2011; Xiong et al., 2011; Nicolas et al., 2012; He et al., 2015) because of the highest percentage identity between A01-C01 and A02-C02. We determined that the introgression of 1.21 Mbp from the C02 chromosome was formed in one plant between the fourth and fifth generations, whereas the introgression of 0.53 Mbp from the C01 chromosome was eliminated in the fifth generation for both A n/r A n/r hybrids. The elimination of this introgression, combined with the consistently low fertility in the A n/r A n/r hybrids from G 4 and G 5, supports the hypothesis that homoeologous recombination was not responsible for the fertility decrease in the A n/r A n/r hybrids. By contrast, the largest introgression of 4.64 Mbp from the C02 chromosome was present and maintained in each A n/r A n/r hybrid. By looking at the amplification of specific SNP markers on the homoeologous region of the A02 chromosome, it appears that for one G 5 A n/r A n/r hybrid, this introgression was fixed and results in the loss of 0.94 Mbp from the A subgenome. This deletion is the only one identified in our hybrids, whereas all plants had a similarly poor fertility, indicating that this A deletion might not affect viability. However, the C02 introgression present in all plants could be essential in plant survival owing to its putative role in gene compensation. Indeed, we showed that this introgression may compensate 30 genes lost in the B. napus A02 homoeologous regions during its evolution in a polyploid context (Chalhoub et al., 2014). This last observation supports the second hypothesis, which is that a pure A n A n plant is not viable because of its structural evolution in a polyploid context. Indeed, in wheat, the presence of both genomes in AABB plants from AABBDD wheat may compensate for the impact of subgenome evolution in a polyploid context in spite of functional alterations (Kerber, 1964; Zhang et al., 2014; Liu et al., 2015). This hypothesis is supported by the inability to generate pure B. napus A n A n plants in strategy 1 as well as by the fertility decrease of A n/r A n/r hybrids in strategy 2 through generations (Fig. 3b). Additionally, using another approach, Tu et al. (2010) reported results in agreement with this assumption by attempting to extract a pure A n A n component of B. napus. These authors did not use B. rapa but performed intertribal crosses with 3000 pollinations between B. napus and Isatis indigotica and generated one mixoploid plant (2n = 25 30). After a second backcross with I. indigotica and selfing, they produced plants with 2n = 20, but none of them presented the pure B. napus A subgenome expected, as A n deletions as well as introgressions from I. indigotica were always observed. Here, compared with Tu et al. (2010), different varieties but also different cytoplasms (B. rapa here) were analyzed. In both cases, the same difficulties were encountered but we cannot exclude the possibility that these factors have an impact on the extraction of pure A n A n plants. The extraction of a pure B. napus A n subgenome may be difficult because of the subgenome evolution in a polyploid context In order to test the second hypothesis related to the increase of B. napus A n subgenome purity, we conducted a molecular analysis of the diploid A n/r A n/r hybrids produced from the fourth and fifth generations of the second strategy. Using the 60K SNP array we designed B. napus and B. rapa homozygous (A n /A n and A r /A r ) as well as heterozygous (A n /A r ) blocks covering c. 80% of each A n/r A n/r hybrid genome (Fig. 5). With this approach, we clearly demonstrated that the proportion of B. napus A n subgenome extracted is lower than expected in each A n/r A n/r hybrid that has a higher proportion of B. rapa A r genome remaining from the initial cross (Table 1). We can hypothesize that some B. rapa A r genomic regions could be essential for plant survival at the diploid stage. In fact, we noticed that the proportions of B. napus and B. rapa are spread unevenly per A chromosome in each A n/r A n/r hybrid, suggesting a bias at the chromosome scale for the B. napus A n subgenome extraction (Fig. 5). The blocks physical size varied from a few hundred to several millions base pairs in each A n/r A n/r hybrid (Table S3), probably as a result of the high frequency of homologous recombination in AAC hybrids (Leflon et al., 2010; Suay et al., 2014) used at each generation. We detected a high proportion of blocks fixed in a homozygous B. rapa state (A r /A r ) in each A n/r A n/r hybrid (21.5% in G 4 and from 12.7% to 15.5% in G 5 ), including one centromere (Fig. 5). This result was

10 New Phytologist Research 1895 unexpected, as a backcross with B. napus was performed at each generation in order to restore the heterozygous state in all of the fixed A r /A r genomic regions in the AAC hybrids. Thus, these heterozygous genomic regions return preferentially in a homozygous B. rapa state (A r /A r ), as only 8.9% and 5.0% of these such regions were expected in the A n/r A n/r hybrids from the fourth and fifth generations, respectively. We tried to determine why the B. rapa A r genomic regions could be essential for plant survival when deprived of the homoeologous C n subgenome. Since Chalhoub et al. (2014) reported that the B. napus A n subgenome differed from the B. rapa A r genome in its gene content, the putative loss of some of these genes since the formation of B. napus could prevent plant survival at the diploid stage. This hypothesis is supported by analyzing the positions of the 1235 A n genes included in designed blocks, which were more highly represented in blocks containing B. rapa (A r /A r and A n /A r ) than in blocks fixed in a homozygous B. napus state (A n /A n ) in the A n/r A n/r hybrids. Thus, the B. rapa A r genomic regions remaining could allow compensation of 595 genes lost since B. napus formation, for which some may be essential for the survival of AA plants. Of particular importance are the genes present in single copy in B. napus (despite numerous successive WGDs) and the fact that most presumably correspond to essential housekeeping genes (De Smet et al., 2013). Within the 595 A n genes lost that were compensated by B. rapa A r genomic regions, 25.3% were in single copy in B. napus and thus most probably essential for plant survival. In the present paper, we have shown that in a B. napus model, structural subgenome evolution in a polyploid context may impact plant survival at the diploid stage when deprived of its homoeologous subgenome. The material generated in this study offers for the first time the opportunity to identify genes essential for plant development, through the comparison of the extracted diploid AA component of B. napus with current B. rapa and B. napus via functional analyses. Acknowledgements We acknowledge the Genetic Resource Center (BrACySol, UMR IGEPP, Ploudaniel, France) for providing seeds and the technical staff for their technical assistance in glasshouses (especially L. Charlon, P. Rolland, J. P. Constantin, J. M. Lucas, F. Letertre and C. Guerin). We also thank the UMR 8199 genotyping service (Lille, France; A.P. was supported by a fellowship from BAP INRA and the Brittany region. This work was funded by the BAP INRA department. We thank Dr Joseph Jahier (UMR IGEPP, France), Dr Eric Jenczewski (UMR IJPB, France), Dr Angelique D Hont (UMR AGAP, France) and Prof. Stephen Strelkov (University of Alberta, Canada) for their critical review of the manuscript. Author contributions A.P., G.T., F.E., M.R-G., A-M.C. designed the research; G.T., F.E., M.L., M.G., A-M.C. carried out plant production and cytogenetic analyses; A.P., M.L., G.D., C.F., S.N., J.M., M.R-G. performed the molecular characterization; and A.P., G.T., M.R- G., A-M.C. carried out the data analyses and wrote the manuscript. References Adams KL Evolution of duplicate gene expression in polyploid and hybrid plants. Journal of Heredity 98: Ainouche ML, Wendel JF Polyploid speciation and genome evolution: lessons from recent allopolyploids. In: Pontarotti P, ed. Evolutionary biology: genome evolution, speciation, coevolution and origin of life. Springer: New York, Albertin W, Balliau T, Brabant P, Chevre A-M, Eber F, Malosse C, Thiellement H Numerous and rapid nonstochastic modifications of gene products in newly synthesized Brassica napus allotetraploids. Genetics 173: Albertin W, Brabant P, Catrice O, Eber F, Jenczewski E, Chevre A-M, Thiellement H Autopolyploidy in cabbage (Brassica oleracea L.) does not alter significantly the proteomes of green tissues. Proteomics 5: Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ Basic local alignment search tool. Journal of Molecular Biology 215: Chalhoub B, Denoeud F, Liu S, Parkin IAP, Tang H, Wang X, Chiquet J, Belcram H, Tong C, Samans B et al Early allopolyploid evolution in the post-neolithic Brassica napus oilseed genome. Science 345: Chen B, Simonsen V, Lanner-Herrera C, Heneen W A Brassica campestrisalboglabra addition line and its use for gene mapping, intergenomic gene transfer and generation of trisomics. Theoretical and Applied Genetics 84: Cheng F, Mandakova T, Wu J, Xie Q, Lysak MA, Wang X Deciphering the diploid ancestral genome of the mesohexaploid Brassica rapa. Plant Cell 25: Cheung F, Trick M, Drou N, Lim YP, Park JY, Kwon SJ, Kim JA, Scott R, Pires JC, Paterson AH et al Comparative analysis between homoeologous genome segments of Brassica napus and its progenitor species reveals extensive sequence-level divergence. Plant Cell 21: Comai L The advantages and disadvantages of being polyploid. Nature Reviews Genetics 6: Crouch J, Lewis B, Mithen R The effect of A genome substitution on the resistance of Brassica napus to infection by Leptosphaeria maculans. Plant Breeding 112: Cui L, Wall PK, Leebens-Mack JH, Lindsay BG, Soltis DE, Doyle JJ, Soltis PS, Carlson JE, Arumuganathan K, Barakat A et al Widespread genome duplications throughout the history of flowering plants. Genome Research 16: De Smet R, Adams KL, Vandepoele K, Van Montagu MCE, Maere S, Van de Peer Y Convergent gene loss following gene and genome duplications creates single-copy families in flowering plants. Proceedings of the National Academy of Sciences, USA 110: Feldman M, Levy A, Chalhoub B, Kashkush K Genomic plasticity in polyploid wheat. In: Soltis SP, Soltis ED, eds. Polyploidy and genome evolution. Berlin, Heidelberg, Germany: Springer, Gaeta RT, Pires JC Homoeologous recombination in allopolyploids: the polyploid ratchet. New Phytologist 186: Gaeta RT, Pires JC, Iniguez-Luy F, Leon E, Osborn TC Genomic changes in resynthesized Brassica napus and their effect on gene expression and phenotype. Plant Cell 19: Galili G, Feldman M Intergenomic suppression of endosperm protein genes in common wheat. Canadian Journal of Genetics and Cytology 26: Hasterok R, Wolny E, Kulak S, Zdziechiewicz A, Maluszynska J, Heneen WK Molecular cytogenetic analysis of Brassica rapa Brassica oleracea var. alboglabra monosomic addition lines. Theoretical and Applied Genetics 111: He Z, Cheng F, Li Y, Wang X, Parkin IA, Chalhoub B, Liu S, Bancroft I Construction of Brassica A and C genome-based ordered pan-transcriptomes for use in rapeseed genomic research. Data Brief 4: