Crossover rate between homologous chromosomes and interference are regulated by the addition of specific unpaired chromosomes in Brassica



  • Recombination is a major mechanism generating genetic diversity, but the control of the crossover rate remains a key question. In Brassica napus (AACC, 2= 38), we can increase the homologous recombination between A genomes in AAC hybrids. Hypotheses for this effect include the number of C univalent chromosomes, the ratio between univalents and bivalents and, finally, which of the chromosomes are univalents.
  • To test these hypotheses, we produced AA hybrids with zero, one, three, six or nine additional C chromosomes and four different hybrids carrying 2= 32 and 2n = 35 chromosomes. The genetic map lengths for each hybrid were established to compare their recombination rates.
  • The rates were 1.4 and 2.7 times higher in the hybrids having C6 or C9 alone than in the control (0C). This enhancement reached 3.1 and 4.1 times in hybrids carrying six and nine C chromosomes, and it was also higher for each pair of hybrids carrying 2n = 32 or 2n = 35 chromosomes, with a dependence on which chromosomes remained as univalents.
  • We have shown, for the first time, that the presence of one chromosome, C9, affects significantly the recombination rate and reduces crossover interference. This result will have fundamental implications on the regulation of crossover frequency.


Recombination is one of the main mechanisms generating diversity in all sexual organisms through crossovers (COs) occurring during the first division of meiosis. The frequency of COs is thus a key factor in increasing the variability in natural populations and in breeding programmes. These COs allow exchanges between homologous non-sister chromatids. They ensure appropriate segregation of chromosomes based on at least one obligate CO per pair of homologous chromosomes and produce new combinations of alleles at different genetic loci. COs are subject to stringent controls, as typically no more than one or two are formed between homologous chromosomes (Mézard et al., 2007). Their positions on each chromosome are usually affected by a phenomenon called interference, that is, a CO in one region reduces the probability that a second CO occurs simultaneously in an adjacent region. As a consequence, the distance between COs is less variable than would be expected were they to occur independently (Jones, 1984). The interfering COs (class I) are catalysed by specific proteins (including the ZMMs), whereas class II (non-interfering) COs are controlled by different proteins (such as MUS81). In plants, class I represents 70–85% of the COs and most of the remaining ones seem to be of class II (Mézard et al., 2007).

Previous reports have indicated that changes in ploidy level can be associated with an increase in recombination rate, suggesting that this could be a general trend (Desai et al., 2006; Leflon et al., 2010; Pecinka et al., 2011). In Brassica spp. and in Arabidopsis, the rates of meiotic recombination are higher in polyploids than in diploids with an identical genomic background (Pecinka et al., 2011). It has also been shown that karyotype composition could have an effect on the CO rate in Brassica spp. (Leflon et al., 2010; Mason et al., 2011). In these previous studies, the analysis of diploid (AA), allotriploid (AAC) and allotetraploid (AACC) hybrids, with the same genomic background, revealed that the CO rate depends on the genomic structure of the hybrids, with the highest recombination rate observed in the triploid hybrid. Leflon et al. (2010) also indicated that COs remain subject to interference in the allotriploid (AAC) hybrid, despite the high frequency of multiple COs per chromatid observed in this hybrid. In B. napus, it has also been demonstrated that homologous recombination rates are different in the triploid (AAC) and tetraploid (AACC), indicating that the dosage of factors carried by the C genome could affect the recombination rates (Nicolas et al., 2009). Previous work in Caenorhabditis elegans has shown that the presence of a single pair of univalents (unpaired chromosomes) during meiosis induces a compensatory increase in COs on the recombining chromosomes involved in bivalents (pairs of homologous chromosomes) of the same nucleus (Carlton et al., 2006). This observation is in agreement with at least two older studies in plants, in Crepis capillaris and in Hypochoeris radicata (Parker, 1975; Tease & Jones, 1976).

Given these results, we can postulate that the number of univalents or the ratio between the number of univalents and bivalents can modulate the frequency of COs on the recombining chromosomes. An alternative hypothesis is that the choice or ‘identity’ of univalent chromosomes matters most because every chromosome carries different genes that may affect recombination in a very specific way. Brassica napus (AACC, 2n = 38) is a relevant model to test these different hypotheses. This young allopolyploid species results from multiple independent hybridization events between the ancestors of B. oleracea (CC, 2n = 18) and B. rapa (AA, 2n = 20) diploids (Palmer et al., 1983; Song & Osborn, 1992; Allender & King, 2010). The fertile AAC hybrids have a regular meiosis, mainly showing the formation of bivalents between A chromosomes, the C chromosomes remaining unpaired as univalents and segregating in their progeny (Leflon et al., 2006). This allows the production of different hybrids with the A genome remaining diploid and with different numbers and identities for the C chromosomes remaining unpaired or paired (as univalents and/or bivalents). In addition, it has been shown that some loci carried by the C genome, specifically a major quantitative trait locus (QTL) PrBn on the chromosome C9, as well as two minor QTLs on C1 and C6, and even epistatic interactions (Liu et al., 2006), affect homoeologous recombination between A and C genomes in AC oilseed rape haploids; however, their potential roles in the control of homologous recombination are unknown (Nicolas et al., 2009).

In this article, we determine, for the first time, the effect of additional C chromosome numbers and identity, either unpaired or paired, on the frequency of COs between A homologous chromosomes. The extent of the variation in CO frequency was measured by comparing genetic map lengths of the A linkage groups. We analysed the progeny of F1 hybrids (AAC) with the A genome in the disomic condition and different numbers and choices of C chromosomes as univalents with the same genetic background. Unfortunately, only a few of these were fertile and, as a result, we were able to obtain genetic maps for only five progenies. We also evaluated the impact of the relative number of bivalents and univalents by studying the progeny of four hybrids (AAC) with AA genomes and with C chromosomes forming bivalents or as univalents. Our results indicate that the identity of chromosomes as univalents can modulate the recombination rate and reduce CO interference. This effect also differs according to the number of bivalents and univalents. These results will have fundamental implications for the regulation of CO frequency and will allow the optimization of breeding programmes for the introduction of new variability within crops.

Materials and Methods

Plant materials

The strategy used to produce hybrids and segregated backcross progenies is detailed in Fig. 1. One single plant, B. oleracea var. alboglabra (‘RC34’, homozygous doubled haploid lines, CoCo, 2n = 18) was crossed to C1.3 (A1A1, 2n = 20), selected within the Brassica rapa Chicon variety (an old non-homogeneous French forage variety). Both diploid varieties are available at the Genetic Resource Center, BrACySol (UMR IGEPP, Ploudaniel, France). The resulting A1Co amphihaploid hybrid was treated with colchicine to obtain the B. napus allotetraploid hybrid, as described by Chevre et al. (1989). This plant was then backcrossed as female to C1.3 (A1A1, 2n = 20) to synthesize A1A1Co triploid hybrids. Their meiotic behaviour was established from 219 pollen mother cells (PMCs) in metaphase I (MI), and revealed that 77% of PMCs had the expected meiotic behaviour: 10 AA bivalents and nine C univalents (Supporting Information Table S1).

Figure 1.

(a, b) Schematic representation of the production of hybrids carrying the A genome of Brassica rapa and C additional chromosomes of Brassica oleracea. (c, d) Schematic representation of the chromosomic composition of hybrids. A1A1 and A2A2 represent B. rapa cultivars C1.3 and Chiifu, respectively, CoCo designates the Brassica oleracea cv RC, ADADCDCD represents the B. napus cv Darmor and Cn represents the displayed C univalent chromosome. In the chromosome composition of F1 hybrids, X and I denote paired and unpaired chromosomes, respectively.

To analyse the effect of the number and choice of additional C chromosomes on the frequency of COs on A chromosomes, these A1A1Co hybrids were crossed to the B. rapa Chiifu cultivar (A2A2; kindly provided by Y. P. Lim, Chungnam National University, Daejeon, South Korea) whose genome sequence has just been released (Wang et al., 2011). By flow cytometry, 202 progeny plants were screened, showing the segregation of Co chromosomes. The screening of Co chromosomes was established using specific molecular markers of each Co chromosome (Table S2) from 106 plants selected on their assessed chromosome number with 2n = 20, 21, 23, 26 or 29 (Fig. S1). Combining cytometer and molecular data, 29 plants were selected and crossed to the B. napus Darmor variety. Finally, only six F1 hybrids, in spite of their low vigour, produced enough seeds after crosses to the B. napus variety Darmor (Genetic Resource Center, BrACySol, UMR IGEPP) to establish the A genetic map; their genome structures were A1A2 + 0C (control), 1C (C6 or C9), 3C (C1, C2, C3), 6C (C1, C5, C6, C7, C8, C9) and 9C, and they led to segregation progenies (65 plants for 0C, 102 plants for 1C6, 108 plants for 1C9, 105 plants for 3C, 90 plants for 6C and 70 plants for 9C).

The same A1A1Co hybrids were crossed with B. napus variety Darmor (ADADCDCD). Eighty-two plants were analysed by flow cytometry to assess the segregation of Co chromosomes (Fig. S1), and the identifications of these chromosomes were verified using specific simple sequence repeat (SSR) markers of Co chromosomes (Table S2). Each Co chromosome retained in the hybrids was able to pair with the corresponding CD chromosome. Among the 11 hybrids chosen, four were selected holding either 2n = 32 with A1ADCD + 3 Co chromosomes or 2n = 35 with A1ADCD + 6 Co chromosomes including or not C9. They were crossed with B. napus (Yudal) variety in order to obtain large segregation progenies (115 plants for 13II + 6I, 95 plants for 13II + 6I C9, 109 plants for 16II + 3I, 107 plants for 16II + 3I C9).

Cytogenetic characterization

Flow cytometry was used at the seedling stage to assess the chromosome number of each F1 hybrid, as described by Leflon et al. (2006).

For the establishment of meiotic behaviour, MI was analysed from 20–30 PMCs of young floral buds. They were fixed in Carnoy's solution (alcohol : chloroform : acetic acid, 6 : 3 : 1) for 24 h at room temperature and stored in 50% ethanol at 4°C. Anthers were squashed and stained in a drop of 1% acetocarmine solution.

Two BAC clones, B. oleracea BAC Bob014O06 (Howell et al., 2002) and B. rapa BAC KBrB021P15 (Kim et al., 2009), were labelled by random priming with Alexa 488-5-dUTP and biotin-14-dUTP (Invitrogen–Life Technologies), respectively. The clone BAC KBrB021P15 hybridizes to A7 and C6 chromosome pairs in B. napus. The long arm of A7 is homoeologous to chromosome C6 (Parkin et al., 2005). The BAC Bob014O06 was used as ‘genomic in situ hybridization (GISH)-like’ to distinguish specifically all C-genome chromosomes in B. napus (Szadkowski et al., 2010).

Chromosome preparations were incubated in Rnase A (100 ng μl−1) and pepsin (0.05%) in 10 mmol HCl, fixed with paraformaldehyde (1%), dehydrated in an ethanol series (70%, 90% and 100%) and air dried. The hybridization mixture consisted of 50% deionized formamide, 10% dextran sulfate, 2 × Saline Sodium Citrate (SSC), 1% sodium dodecylsulfate (SDS) and labelled probes (100 ng per slide). Chromosome preparations and pre-denatured (92°C for 6 min) probes were denatured at 82°C for 2 min. 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 several washes in 4 × SSC–Tween. For indirect detection of BAC KBrB021P15 DNA marked with biotin, we visualized the probe using avidin–Texas-red antibodies (Vector Laboratories, Burlingame, CA, USA) and the signal was amplified with biotinylated anti-avidin D (Vector Laboratories). The chromosomes were mounted and counterstained in Vectashield (Vector Laboratories) containing 2.5 μg ml−1 4′,6-diamidino-2-phenylindole (DAPI). Fluorescence images were captured using a CoolSnap HQ camera (Photometrics, Tucson, AZ, USA) on an Axioplan 2 microscope (Zeiss, Oberkochen, Germany) and analysed using MetaVue™ (Universal Imaging Corporation, Downington, PA, USA).

Molecular analyses

Genomic DNA was extracted from young leaves as described by Lombard & Delourme (2001). Specific markers for each C chromosome and for A2, A4 and A7 linkage groups were selected from published maps (markers prefixed with cnu and nia (Piquemal et al., 2005; Kim et al., 2009), CB, OI and Bras (Piquemal et al., 2005)) and from other sources, such as TraitGenetics, Germany (markers prefixed with NMS) and Agriculture and Agri-Food Canada (markers prefixed with sN or sR). Physical functional markers (PFMs) were developed from the coding sequence of Arabidopsis or Brassica sequences by B. Chalhoub (URGV Evry, France; unpublished data, primer sequences are available on request to Genoplante). JLP markers were obtained from Prieto et al. (2005). PCR assays were conducted and analysed on a 16 Capillary ABI Prism 3130xl, essentially as described by Esselink et al. (2004). Forward primers were tailed with an M13 extension to be revealed by fluorescent technology (Schuelke, 2000).

Construction of genetic linkage maps and statistical analyses

Distortion of molecular markers in the progenies was analysed using χ2 tests, and all significantly distorted markers were excluded from the analysis of each progeny. Linkage analysis was performed with MAPMAKER/EXP version 3.0 (Lincoln et al., 1992). Linkage groups were established with a Logarithm of Odds (LOD) threshold of 4.0, and the Kosambi function was used to evaluate the genetic distances in centimorgan (cM) between linked markers. The heterogeneity of CO rates among progenies was assessed for every interval between adjacent markers using χ2 tests.

One- and two-pathway Gamma modelling of CO interference

The single-pathway Gamma model, which assumes that the genetic distances between successive COs are independent identically distributed random variables following the Gamma distribution of shape parameter nu, was used to evaluate the CO interference (McPeek et al., 1995). Using the value of nu which maximizes the likelihood of the observed gametes, the confidence interval on nu was obtained by 1000 resimulations.

The two-pathway Gamma model (Copenhaver et al., 2002) assumes that there is a fraction p of the COs arising from a second non-interfering pathway P2 in addition to the COs arising from the first pathway P1. The model thus has two adjustable parameters, nueff and p. We denote by LG the genetic length in morgans of the chromosome (Gauthier et al., 2011). To obtain the confidence intervals, we used a resimulation approach as described in Falque et al. (2009) employing 1000 resimulations.

Pair-wise comparisons of the effective interference strength

The effective interference strength nueff is the interference parameter inferred when using the single-pathway modelling. If there are two datasets, the confidence intervals on each nueff allow for a qualitative comparison, but, for a quantitative test of similarity, we implemented the following procedure. First, we fitted each population separately, obtaining the interference strengths nu(1exp)eff and nu(2exp)eff. Second, we fitted the two populations jointly, leading to one consensus interference strength nu*eff. Third, using this consensus value, we generated 1000 simulated datasets for each of the two populations. The fitting of these simulated datasets produces a long list of simulated nu(1)eff and nu(2)eff values. The P value for the hypothesis that one common nu*eff describes the two populations is then obtained by calculating the number of these pairs nu(1)eff and nu(2)eff that differ by more than the difference between nu(1exp)eff and nu(2exp)eff. It is possible to generalize this test to the two-pathway framework. Nevertheless, for this to be meaningful, pathway P1 must be present and interfering. All populations except 0C and 1C6 were compatible with P1 being either absent or non-interfering.

Test for the presence of P1 COs at the level and strength of the 0C control

We introduced the hypothesis H0 that the additional C chromosomes solely increase the contribution of the non-interfering P2 pathway, leaving all characteristics of P1 unchanged (nu and the average number of COs produced via P1). Fitting the 0C control population with a two-pathway framework leads to nu = 23.8 and = 0. This result specifies the characteristics of P1: the average number of COs is L0C = 0.547 (genetic map length in morgans) and nu = 23.8. Under H0, any other population i of genetic map length Li is characterized by (nu; pi) with pi = (Li − L0C)/Li. To test H0, we took the summary statistic equal to the goodness-of-fit score used for fitting the two-pathway model. Based on 1000 simulated datasets produced under H0, we determined the distribution of the summary statistic. Placing the value obtained for the experimental dataset allowed us to provide the P value for H0.


Effect on recombination rate of the number and identity of additional C chromosomes

To assess the variation in the CO rate between homologous chromosomes as a result of the number and identity of univalent chromosomes, we produced hybrids with the same A genome at the diploid stage and different C genomic content. Unfortunately, the constraint of fertility of such hybrids severely restricted the cases for which we were able to produce genetic maps: only cases with zero, one (C6 or C9), three, six or nine additional C chromosomes were exploitable, as described in Fig. 1(a). All hybrids had a meiotic behaviour close to that expected with 10 bivalents and additional C chromosomes remaining as univalents at MI (Table S1, Fig. 2a–g). BAC-fluorescence in situ hybridization (FISH) analysis, using a specific BAC of A7/C6 chromosomes, revealed that A7 chromosomes always formed a bivalent in MI and never formed CO with the C genome (Fig. 2d–f). The CO rates in these hybrids were calculated from the genetic maps established in their progeny for A7, A4 and A2 chromosomes with 12, 13 and 15 molecular markers that cover at least 70%, 80% and 95%, respectively, of the total length of each chromosome.

Figure 2.

(a–l) Chromosome associations at metaphase I in pollen mother cells (PMCs) in hybrids carrying the A genome of Brassica rapa and C additional chromosomes of Brassica oleracea. (a) A1A2 plant (10II). (b) A1A2 + 3C plant (10II + 3I). (c) A1A2 + 6C plant (10II + 6I). (d–f) Metaphase I cell shown for A1A2 + 9C: nine C univalent chromosomes (green, e), one C6 univalent and A7 bivalent (red, f). (g) A1A2 + 1C plant (10II + 1I). (h) A1ADCD + 3C plant (13II + 6I). (i) A1ADCD + 6C plant (16II + 3I). (j–l) Metaphase I cell shown for A1ADCD + 6C plant; six C univalents and three C bivalents indicated by an arrow (green, k), one C6 univalent and A7 bivalent (red, l). Univalents are indicated by an orange star. BAC-fluorescence in situ hybridization (FISH)/genomic in situ hybridization (GISH)-like was carried out using BAC KBrB021P15 and Bob014O06 which identify the A7 and C6 chromosomes (red) and all the C chromosomes (green), respectively. Bars, 5 μm.

In a first step, we assessed the effect of the number of C chromosomes in addition by comparing genetic maps established from F1 hybrids, AA plus zero, three, six or nine C chromosomes in addition. The total lengths of linkage groups A7 were 54.7 and 63.4 cM in the progenies of hybrids with zero and three C chromosomes in addition, respectively (Fig. 3a). At the scale of the linkage group, considered as a combination of intervals with or without COs, the differences between 0C and 3C hybrids were not statistically significant (Table 1). By contrast, genetic map sizes were much higher in the progenies of hybrids 6C (167.6 cM) and 9C (> 238.2 cM). In the 9C hybrid, the CO rate was so high that the A7 linkage group was split into two, summing to > 238.2 cM (128.1 cM + at least 50 cM between Bras023 and nia_m63a markers which were unlinked + 60.1 cM; Fig. 3a, Table 1). Similar results were observed when comparing map lengths of chromosomes A2 and A4 in the same 0C, 3C and 6C progenies (Fig. S2). In view of this similarity, further analyses were performed on the A7 linkage group only.

Table 1. Comparison between each linkage group for the recombination rate and the number of crossovers (COs) per chromatid from hybrids carrying the A genome of Brassica rapa and 0C, 3C, 1C6, 1C9, 6C or 9C chromosomes of Brassica oleracea in the top half of the table and for hybrids with different numbers of bivalents (II) and univalents (I) in the bottom half of the table
  P values for CO rates
 0C 0C 0.1842E-41.60E-156.60E-142.60E-27
1CO3C0.361 3C 0.1511.50E-107.50E-111.30E-24
> 2COs0.038 1C 6 6.60E-64.30E-81.30E-19
1CO1C60.1770.010 1C 9 7.4E-48.5E-4
> 2COs0.2062.50E-07 6C 1.35E-8
> 2COs3.27E-085.44E-061.7039E-08
> 2COs1.44E-090.2194.0493E-100.2603
> 2COs2.43E-130.0023.3859E-080.00270.0575
 P value for the numbers of COs per chromatid
  P values for CO rates
13II + 6I16II + 3I13II + 6IC916II + 3IC9
  1. For each half of the table, the top right side of the matrix shows the P values of the χ2 tests (for the top half of the table, significant at = 0.004, for 11 intervals; for the bottom half of the table, significant at = 0.005, for nine intervals) comparing CO rates among populations for every interval between adjacent markers. The bottom left side of the matrix of each half of the table shows the P values of the χ2 tests comparing the numbers of CO per chromatid among populations with either one CO/chromatid (1CO) or two or more COs/chromatid (> 2COs).

 13II + 6I 13II + 6I 2.90E-54.70E-121.40E-8
1CO16II + 3I0.021 16II + 3I 1.90E-70.0073
> 2COs9E-4 13II + 6IC 9 0.017
1CO13II + 6IC94.124E-080.56E-4 
> 2COs5.46E-173.26E-08
1CO16II + 3IC90.01240.8340.0011
> 2COs1.46E-060.1213.12E-05
 P value for the numbers of COs per chromatid
Figure 3.

(a) Maps of the A7 Brassica rapa linkage group in the progenies of F1 hybrids carrying in addition 0, 3, 6 or 9 C chromosomes of Brassica oleracea. (b) Maps of the A7 linkage group in the progenies of 1C6 and 1C9. Genetic distances, indicated on the left of the linkage group, are expressed in cM and represent the distances between the marker and the first marker on the top. The total length of the linkage group is indicated in the red box.

As the AA + 6C and AA + 9C hybrids contain the C9 and C6 chromosomes that carry QTLs involved in the control of homoeologous pairing (Liu et al., 2006), it is appropriate to assess the effect of these specific chromosomes and, more generally, to ask whether the choice of chromosomes that are univalents affects the recombination rate.

Therefore, in a second step, we analysed the progeny of hybrids carrying only 1C6 or 1C9. The genetic map size was much higher in the progenies carrying C9 alone (1C9, 153.4 cM) than in the 1C6 progeny (74.1 cM). In the case of 1C9 vs 6C progenies, the map sizes of both linkage groups were quite similar: 153.4 cM for 1C9 and 167.6 cM for 6C. However, they showed a highly significant difference in the CO distribution as a result of a different pattern of COs along the chromosome: for some intervals, progeny from the 1C9 hybrid exhibited significantly more COs than 6C, whereas the contrary was observed for other intervals (Table 1, Fig. 3b). The ratio of the recombination rate between each hybrid and the control hybrid (0C) was calculated for each interval of the linkage group. The recombination rate in the 1C6 hybrid was, on average, 1.4-fold higher than in the control hybrid, whereas, in the 1C9 hybrid, it was 2.7-fold higher. In the 6C and 9C hybrids, this ratio was, on average, 3.1 and 4.1, respectively, with a large variation along the linkage group, most being statistically significant. In all hybrids, the ratios presented the same pattern, showing two peaks of maximum increase in CO rate around the sR282R and cnu_m273a markers (Fig. 4). The overall recombination rate in the 1C9 hybrid was similar to that in the 6C hybrid, but significantly lower than that in the 9C hybrid (Table 1).

Figure 4.

Ratio of recombination rate between each hybrid carrying the A genome of Brassica rapa with additional chromosomes (1C6, 1C9, 3C, 6C and 9C) of Brassica oleracea and the control hybrid (0C) along chromosome A7.

Concerning the number of COs per chromatid, the 1C9, 6C and 9C hybrids showed more multiple COs than the 0C hybrid (Table 1, Fig. 5). It can be highlighted that the number of total COs in the 9C hybrid was higher than that in the 6C hybrid (Table 1). As both hybrids shared six univalent C chromosomes (C1, C5, C6, C7, C8, C9), the data confirm that there was an effect coming from the additional C chromosomes (C2, C3 or C4) and/or from interactions between additional C chromosomes.

Figure 5.

Frequency of crossovers (COs) per progeny from F1 hybrids carrying the A genome of Brassica rapa and 0C, 1C6, 1C9, 3C, 6C and 9C chromosomes of Brassica oleracea.

Our genetic analyses clearly demonstrated that CO formation between A homologous chromosomes received a statistically significant boost in the hybrids that presented the C9 chromosome at single dosage. However, we cannot exclude the possibility that the effect could be caused by a threshold on the number of univalent vs bivalent chromosomes, or by epistatic effects between other additional chromosomes.

Effect on recombination rate of the number of univalents and bivalents in the presence of the C9 chromosome paired or unpaired

We wanted to test whether it is possible to ‘dilute’ the effect of the univalent chromosomes by adding more bivalent chromosomes. Therefore, we quantified the impact of the univalent and bivalent chromosome number on the CO rate between A homologous chromosomes. We selected hybrids having either 2n = 32 with A1ADCD + 3 Co chromosomes or 2n = 35 with A1ADCD + 6 Co chromosomes, as described in Fig. 1(b). They showed the expected meiotic behaviour: 13 bivalents + 6 CD univalents for 2n = 32 hybrids (13II + 6I) and 16 bivalents + 3 CD univalents (16II + 3I) for 2n = 35 hybrids (Table S1, Fig. 2h–l). BAC-FISH analyses revealed that A7 chromosomes always formed a bivalent (Fig. 2j–l). After crosses with another B. napus variety (Yudal), we produced large progenies to establish the genetic maps of A7 chromosomes from 10 polymorphic markers that cover at least 75% of the total length of the chromosome.

The map size of the A7 linkage group was significantly larger in the progeny of the 16II + 3I hybrid (237 cM) than in the 13II + 6I hybrid (140.6 cM; Fig. 6a). The total number of COs was significantly different between the two hybrids (Table 1). Comparing the number of recombinant intervals per chromatid between both hybrids, the 16II + 3I hybrid showed higher values than the 13II + 6I hybrid (Table 1, Fig. 7).

Figure 6.

(a) Maps of the A7 linkage groups in the progenies of F1 hybrids carrying the A genome of Brassica rapa and Brassica napus  + C genome of Brassica napus and additional chromosomes of Brassica oleracea with 2n = 32 (13II + 6I and 16II + 3I). (b) Maps of the A7 linkage groups in the progenies of F1 hybrids with 2n = 35 (13II + 6C9 and 16II + 3IC9). Genetic distances, indicated on the left of the linkage group, are expressed in cM and represent the distances between the marker and the first marker on the top. The total length of the linkage group is indicated in the red box.

Figure 7.

Frequency of crossovers (COs) per progeny from hybrids carrying the A genome of Brassica rapa and Brassica napus  + C genome of Brassica napus and additional chromosomes of Brassica oleracea with 2n = 32 (13II + 6I and 13II + 6IC9) and 2n = 35 (16II + 3I and 16II + 3IC9).

To assess the effect of C9 in single dosage, two progenies were derived from hybrids with 2n = 32 and 2n = 35, respectively, but with C9 in the univalent state (13II + 6IC9 and 16II + 3IC9). Their meiotic behaviours were close to those expected (Table S1). These progenies were compared with the progenies of 13II + 6I and 16II + 3I hybrids, respectively, in which C9 was at double dosage and thus in a bivalent state (C9CD9). The map lengths had a minimum size of 238.4 cM (55.7 cM + 82.7 cM + at least 50 cM from each part of Bras023 remaining unlinked) in the 13II + 6IC9 hybrid and 269.7 cM (78.6 cM + 91.1 cM + at least 50 cM from each part of Bras023 remaining unlinked) in the 16II + 3IC9 hybrid. In both cases, the presence of C9 in addition split up the linkage group into two different linkage groups plus a single marker, Bras023 (Fig. 6b). At the scale of the linkage group, the total number of COs was statistically higher in the 13II + 6IC9 hybrid than in the 13II + 6I form (Table 1). Therefore, the map sizes were higher in the hybrids with C9 in the univalent state than in the hybrids with the same genomic structure but with C9 in the bivalent state. Nevertheless, the proportions of CO for each interval of 13II + 6IC9 and 16II + 3IC9 linkage groups were, on average, only 1.59- and 1.14-fold higher than in the 13II + 6I and 16II + 3I hybrids, respectively.

In the 13II + 6I hybrid, the majority of chromatids carried one CO, whereas, in the hybrid with the same genomic structure, but carrying C9 unpaired (13II + 6IC9), the most frequent occurrence was for three and four COs per chromatid (Table 1, Fig. 7).

These results show that COs receive a boost in the hybrids which present the C9 chromosome at single dosage. However, an impact of the other additional chromosomes cannot be excluded.

Analysis of CO interference in the progenies

We investigated to what extent some differences in CO numbers observed with the addition of C chromosomes may correlate with differences in CO interference intensities. We first used the single-pathway Gamma model (hereafter referred to as P1; McPeek et al., 1995) to consider the interference parameter nueff; this was followed by a two-pathway Gamma model (hereafter referred to as P2; Copenhaver et al., 2002) to account for a proportion p of non-interfering COs. In the absence of additional C chromosomes (0C), COs showed a high level of effective interference (Table 2). Furthermore, using the two-pathway Gamma model, we found a very high value of nu (23.8) and P2 was either absent or contributed very little (small values for p; Table 2). By contrast, in the hybrids with additional C chromosomes, the effective CO interference was almost completely suppressed, with the exception of the 1C6 hybrid, in which some interference remained, although much less than in the 0C hybrid. Two-pathway analyses confirmed this trend: the additional C chromosomes significantly modified the interference strength in nearly all cases (Table 2).We performed tests to determine whether interference strengths could be the same in pairs of populations when using the single-pathway Gamma model. The global outcome, also seen from looking at the two-pathway analyses, confirmed that 0C was significantly more interfering than all other datasets, except 1C6 (Table S3, top part), but also showed that the other populations seemed to be compatible with no interference at all.

Table 2. Interference parameters estimated using the Gamma model for the different populations carrying the A genome of Brassica rapa with 0C, 3C, 6C, 9C, 1C6 or 1C9 chromosomes of Brassica oleracea or carrying the A genome of Brassica rapa and Brassica napus + C genome of Brassica napus and additional C chromosomes of Brassica oleracea (13II_6I, 13II_6IC9, 16II_3I, 16II_3IC9)
Cross nu eff nueff_infnueff_sup nu nu_infnu_sup p p_infp_sup
  1. nueff, overall effective interference intensity estimated with the single-pathway model; nu, interference intensity in pathway 1 COs, estimated with the two-pathway model; p, proportion of COs formed via pathway 2; nueff_inf, nueff_sup, nu_inf, nu_sup, p_inf, p_sup, lower and upper boundaries of the 95% confidence intervals for nueff, nu and p, respectively, estimated from 1000 resimulations.


We then asked whether the data for the crosses with additional C chromosomes were compatible with just an increase in the number of P2 COs, leaving the average number of COs in P1 and their interference strength untouched compared with that which arises in the 0C control. Table S3 (bottom part) shows that, in most cases, the test leads one to reject this hypothesis, suggesting that either the total number of P1 COs is reduced or the interference strength of P1 is reduced.

Let us consider now the effect on interference of the number of C univalents and bivalents via the hybrids 13II + 6I, 13II + 6IC9, 16II + 3I and 16II + 3IC9. From Table 2, we see that interference drops dramatically compared with the control 0C and, in fact, as for the other hybrids with no C bivalents, interference seems to disappear almost completely.


COs created during meiotic recombination not only generate new combinations of alleles, but, in most organisms, are essential to ensure the correct segregation of the homologues into the gametes. In the current study, we show that different factors greatly affect the variation of CO frequency between homologous chromosomes and interference strength, such as the presence of the C9 chromosome as univalent and the number of bivalents in the cell.

The choice of additional unpaired chromosomes affects homologous recombination on paired chromosomes

We showed that, in B. napus triploid hybrids (AAC), the homologous recombination rate was increased significantly between A chromosomes in AA hybrids carrying specific C chromosomes as univalent. Several previous studies have already pointed out this effect. In the plant H. radicata, a desynaptic mutant has been described in which the bivalent chiasma frequency in cells exhibiting a pair of univalents is higher than in cells without univalents (Parker, 1975). In an experimental population of C. capillaris, a similar phenomenon was described (Tease & Jones, 1976). The same situation has been published more recently in C. elegans, in which mutations that impair synapsis increase autosomal recombination (Carlton et al., 2006). There is an important difference between our hybrids and the mutants described in C. elegans, C. capillaris and H. radicata: these mutants have a complete set of chromosomes, but they present a mutation that affects the pairing of the other chromosomes in the nucleus. Our hybrids, except for 9C, present an incomplete set of the C genome and, for that reason, they must display univalents. Our results unexpectedly pointed out that the CO rate could be affected by both the number of unpaired chromosomes and by which chromosomes are unpaired.

One possible explanation for this increase in CO rates could be the level of homoeology between unpaired and paired chromosomes with a specific effect on the paired chromosomes (Parker, 1975; Naranjo & Corredor, 2008). We assessed the homologous recombination rate from different A chromosomes showing different relationships with C chromosomes with either homoeology over their full length (A2/C2), over one arm (A4 homeologous to one arm of C4) or over multiple genomic regions (A7 sharing homoeology with C6 and C7; Parkin et al., 2005). We observed similar results for these three A2, A4 and A7 chromosomes, whatever the additional C chromosomes, indicating that the enhanced recombination rates found in the present work arise as a general effect, presumably for all paired chromosomes in the nucleus. Thus, the increase in CO rates observed in our study is not a result of the level of homoeology between unpaired C chromosomes and paired A chromosomes or the size of the chromosomes paired (A2 > A7 > A4; Wang et al., 2011). For this reason, we focused our study on one chromosome, the A7 chromosome, to test the new genomic composition of F1 hybrids.

The most significant result shown in this study is the effect of the C9 chromosome. All hybrids carrying Co9 as univalent, alone or together with other C additional chromosomes, displayed a higher frequency of CO than did the hybrids without it. This chromosome carries a major QTL, PrBn, involved in the control of homoeologous recombination (Jenczewski et al., 2003; Liu et al., 2006). It is thus tempting to assume that PrBn could be responsible for the observed increase in CO frequency reported here. This hypothesis is supported by the fact that variable CO frequencies were observed between A genomes in two different AAC hybrids produced by crossing the same B. rapa plant with two B. napus varieties showing different PrBn activities (Nicolas et al., 2009).

However, there are also two minor QTLs on C1 and C6 and different epistatic relationships between C chromosomes (Jenczewski et al., 2003; Liu et al., 2006). Therefore, their presence, alone or combined, could very well affect the homologous recombination rate. In the 3C hybrid, we added three chromosomes (C1, C2 and C3), but this did not contribute to increase by much the CO rate in spite of the QTL carried by the C1 chromosome. On the contrary, the A7 map length in the 1C6 hybrid is higher than that in the control; this difference might be explained by the effect of a QTL located in C6, as such a QTL was found in Darmor (Liu et al., 2006). It is interesting to observe that the 6C hybrid carrying both C6 and C9 chromosomes showed a slightly higher rate of recombination for some intervals than did the 1C9 hybrid, indicating that the QTLs carried by C6 and C9 could have additive effects. The inflated CO rate arising in 9C is the net result of individual and epistatic (positive and negative) effects of all QTLs.

In view of our results, the unexpected huge recombination rate reported in the allotriploid hybrid (ArAr′Co; Leflon et al., 2010) may be explained by the same phenomenon. It must be remarked that the allotriploid hybrid (ArAr′Co) analysed in Leflon et al. (2010) showed a mean number of COs per chromatid higher (6.1) than the hybrid with the same genomic structure analysed in the present work, 9C (4.1). As the same C genotype was used in both studies, one explanation could be that the coverage of the linkage group in the present study is more complete than in the previous work. However, we cannot exclude an effect of the A genomes involved in paired chromosomes: in both studies, one of the two A genomes originated from B. rapa C1.3, whereas the other A genome originated from B. rapa Z1 in the previous study and from B. rapa Chiifu in the present work.

The number of bivalents vs univalents affects the homologous recombination rate

Our results reveal another important source of CO regulation. The hybrids with more bivalents (16II + 3I vs 13II + 6I) displayed more COs than the hybrids presenting fewer bivalents. This result can be associated with the general trend in polyploid species to increase the CO rate. In cotton, the recombination rates are similar between the two diploid species. However, the allotetraploid species shows an increase in recombination relative to its diploid progenitors (Desai et al., 2006; Suwabe et al., 2008). Similarly, Pecinka et al. (2011) showed in Arabidopsis that the frequency of recombination is higher in auto- or allotetraploids than in diploids. The same result was also reported by Leflon et al. (2010), who observed twice as much recombination in ArAr′CoCo hybrids than in the ArAr′ hybrid. We can hypothesize that an active mechanism, associated with the presence of more bivalents, increases the recombination rate. However, we cannot exclude the possibility that the QTLs carried by the additional chromosomes are involved in recombination regulation (Liu et al., 2006). The recombination rate is higher in the hybrid 16II + 3I (C2, C7 and C8 as univalents) carrying C8, but this chromosome is absent in the hybrid 13II + 6I. In addition, C8 and C9 chromosomes can interact as they are both present in hybrids with C9 as univalent.

Nevertheless, we observed that the effect of bivalent number can be partially additive with the effect of QTLs carried by the C9 chromosome; all hybrids with C9 as a univalent showed a higher rate of homologous recombination. We have confirmed that the choice of the univalents and the relative number of univalents and bivalents have an important influence on the number and distribution of COs. Furthermore, these different effects add up to achieve a maximum threshold to increase the recombination rate. This maximum limit could be the reason why there are no significant differences in the CO rate between the two hybrids 2n = 35.

CO interference is affected by the genomic structure of hybrids

We showed that interference is affected in all hybrids, except possibly in 1C6, compared with the control 0C. The strongest reduction in interference arose in the 1C9 hybrid. Interestingly, there was a rather clear negative correlation between the presence of C9 and the strength of interference. When ordering maps according to increasing genetic length, we have 0C < 3C < 1C6 < 1C9 < 6C < 9C, whereas, when we order according to decreasing interference strength, we have 0C > 1C6 > 3C > 6C > 9C > 1C9. The order is not perfectly correlated between the two cases, but we see a clear pattern whereby genetic length and interference strength are strongly negatively associated. As CO interference is believed to maintain homeostasis of CO number, reducing the strength of CO interference in the interfering pathway P1 could very well lead to an increased recombination rate. This allows us to conclude that the choice of the additional C chromosomes modifies significantly the properties of pathway P1, a result that is very encouraging both for fundamental insights into CO control and for practical breeding programmes. This scenario may also be at work in the hybrids 13II + 6I, 13II + 6IC9, 16II + 3I and 16II + 3IC9. However, the interference in these four hybrids is particularly low and provides no statistically significant evidence for an effect of the number of additional bivalents or univalents.

Our interference results appear to be different from those found in the study by Leflon et al. (2010). However, a closer look reveals that these differences can be minimized. First, the way in which the interference was analysed was different in the two studies. In the previous work, the interference was calculated by comparing the theoretical and real distributions of genetic lengths by assuming that the positions of COs on the linkage group were independently, uniformly distributed random variables, as proposed by Drouaud et al. (2006). However, in this study, we used single- and two-pathway Gamma modelling (McPeek et al., 1995; Copenhaver et al., 2002). These models allow better analysis of the interference when there is a higher frequency of COs over smaller distances. When we analysed the data of Leflon et al. (2010) with these models, there were no differences from the present work (data not shown). The second source of difference between the work of Leflon et al. (2010) and ours is that the coverage of the A7 linkage group in the present study is more complete than in the previous work. Again, when we analysed our data, but used less complete coverage, the interference results were equivalent (data not shown).

To add further evidence for all the effects pointed out in this work, it would be interesting to analyse other complementary F1 hybrids with different combinations of C chromosomes in addition or numbers of paired and unpaired chromosomes. The selection of such plants with a precise genomic structure is difficult because of the size of the segregating population which must be tested to obtain the expected combination (511 combinations from 1C to 9C), as a result of the dissimilar transmission rates of each C chromosome from AAC hybrids, as demonstrated previously (Hasterok et al., 2005; Leflon et al., 2006), and, most importantly, because numerous combinations are not fertile.

The A7 linkage group was split in the hybrids that displayed the highest CO rates. The recent publication of the A genome sequence (Wang et al., 2011) will allow the development of new, physically anchored markers, possibly revealing a correspondence between chromosomal structure and enhancement of CO rates, and perhaps allowing the detection of recombination hot spots (Drouaud et al., 2006).

Previous studies have shown that the presence of univalents can trigger checkpoints during meiotic recombination, causing the nucleus to linger in a recombination active state, and this may result in an increase in CO number and in a change in their distribution in the remaining chromosomes that are synapsed correctly (Martinez-Perez & Moore, 2008). The study of meiotic progress is necessary to obtain the precise role of the different factors detected. Overall, these different insights, together with the present work, should aid in our understanding of evolutionary dynamics involving genetic recombination. It may be of interest to check, in different polyploid species, whether the same model can be applied by assessing the level of recombination between homologous genomes from F1 hybrids carrying one genome at the diploid stage and a related one at the haploid stage. Our results should also spur research to better control the rate of homologous recombination, any advances therein leading to important practical applications, for instance, in breeding programmes.


D.Z. and L.S. were supported by fellowships from Foundation Louis D. – Institut de France, and from the Plant Breeding Departement – INRA. We thank M. Gilet for technical assistance. We acknowledge the Genetic Resource Center (BrACySol, UMR IGEPP, Ploudaniel, France) for providing seeds of the B. napus varieties used and Gentyane platform (INRA, Clermont Ferrand, France) for genotyping. This work was carried out with financial support from the ‘ANR – Agence Nationale de la Recherche – The French National Research Agency’ under the programme BLANC07-3_188863188863, project ‘Unravelling crossover pathways with Arabidopsis thaliana and crop relatives’ and from the Plant Breeding Department – INRA under the programme ‘Régulation de la fréquence de crossing-over entre chromosomes homologues chez le colza’.