Selection against hybrids in mixed populations of Brassica rapa and Brassica napus: model and synthesis


Author for correspondence:
Tom J. de Jong
Tel: +31 715275118


  • Pollen of the crop oilseed rape (Brassica napus, AACC) can cross-fertilize ovules of Brassica rapa (AA), which leads to an influx of unpaired C-chromosomes into wild B. rapa populations. The presence of such extra chromosomes is thought to be an indicator of introgression. Backcrosses and F1 hybrids were found in Danish populations but, surprisingly, only F1 hybrids were found in the UK and the Netherlands.
  • Here, a model tests how the level of selection and biased vs unbiased transmission affect the population frequency of C-chromosomes. In the biased-transmission scenario the experimental results of the first backcross are extrapolated to estimate survival of gametes with different numbers of C-chromosomes from all crosses in the population.
  • With biased transmission, the frequency of C-chromosomes always rapidly declines to zero. With unbiased transmission, the continued presence of plants with extra C-chromosomes depends on selection in the adult stage and we argue that this is the most realistic option for modeling populations.
  • We suggest that selection in the field against plants with unpaired C-chromosomes is strong in Dutch and UK populations. The model highlights what we do not know and makes suggestions for further research on introgression.


Potential gene flow from crops to wild relatives remains a central issue in discussions of the ecological risks involved in introducing genetically modified (GM) plants (Ellstrand et al., 1999; Andersson & de Vicente, 2010). Once a gene has dispersed from a crop to a wild relative it is beyond control, and this process may be irreversible. Care is thus required. Transgenes for herbicide resistance may hold no selective advantage outside agricultural fields but, for instance, a Bt gene that reduces herbivory may well be selected for in the wild (Vacher et al., 2004). The same issue pertains for transgenes that are currently developed to enhance plant growth under suboptimal conditions (Warwick et al., 2009). Introgression involves the incorporation of genes from one species into the gene pool of another. For introgression to occur, gene exchange between species must first result in viable hybrids. Secondly, these newly introduced alleles must be picked up by selection, perhaps aided by drift, and sweep through the population of the receptor species. Whether both aspects of introgression occur is still a matter of debate. Here, we review these outstanding issues using the genus Brassica as a model system.

Brassica napus L. (oilseed rape) has been cropped in Western Europe since the Middle Ages (Zohary & Hopf, 2000). The species crosses readily with Brassica rapa L. (Devos et al., 2009), a common self-incompatible weed occurring throughout Europe and North America. Hybrids resulting from such crosses and backcrosses to B. rapa have up to nine unpaired C-chromosomes (explained below) derived from B. napus. In a recent review, Andersson & de Vicente (2010) concluded that these unpaired C-chromosomes impose little or no fitness cost (see also Légère, 2005). If this view were correct, then wild B. rapa populations growing in an area where B. napus is frequently cropped should accumulate C-chromosomes. Surprisingly, population surveys of B. rapa in the UK (Wilkinson et al., 2000, 2003) and the Netherlands (Luijten & de Jong, 2010, 2011) only picked up F1 hybrids. By contrast, F1 and backcross hybrids were found in a Danish agricultural field converted to organic farming (Hansen et al., 2003). Hansen et al. (2001) suggested that backcrossing occurs much more frequently than hybridization between B. rapa and B. napus. Warwick et al. (2008) found F1 hybrids and backcrossed plants in B. rapa populations growing along the margins of GM canola fields in Canada; exact fractions were not given in this publication.

In this paper we ask under what scenarios C-chromosomes can persist in B. rapa populations, using a population genetic model that integrates detailed knowledge on hybrid crosses from the plant breeding literature, thereby highlighting discrepancies between different observations and interpretations. Our model has multiple hybrid classes that are characterized by the number of C-chromosomes that plants possess. Several earlier models in the field of crop-to-wild gene flow used F1, BC1, BC2 etcetera as different hybrid classes (for instance, Hooftman et al., 2007). Our model differs from those models in the sense that we link the presence of C-chromosome to reduced fitness rather than attributing an average fitness to BC1, BC2 etcetera.

We first outline some peculiarities about the Brassica system. An ancient cross between B. rapa (AA genome, 2= 20) and Brassica oleracea L. (CC, 2= 18) has resulted in the allopolyploid B. napus with the AACC (2= 38) genome (U, 1935). Crosses between wild B. rapa and B. napus produce AAC hybrids with 2= 29. Ten pairs of chromosomes belong to the A set and nine unpaired chromosomes to the C set. When an AAC hybrid is backcrossed to B. rapa, all chromosomes from the A genome are paired and each gamete contains a set of 10 A-chromosomes. The nine chromosomes from the C genome, however, are unpaired. These univalent C-chromosomes are passed on in crosses, resulting in plants with various C-chromosome numbers. Such plants can be grown and raised to maturity in the glasshouse. Homeologous recombination between A- and C-chromosomes can occur, starting with meiosis in the F1 hybrid, but its frequency is still under discussion (Leflon et al., 2006; Howell et al., 2008; Nicolas et al., 2009). In the model we ignore this aspect and keep A- and C-chromosomes strictly separate.

The chance that a specific unpaired C-chromosome is passed on to a gamete is theoretically 50%. An AAC hybrid contains all nine C-chromosomes, so the frequency of additional C-chromosomes in BC1 offspring (AA × AAC) is expected to follow a binomial distribution with a mean of 4.5. Although results from some studies have confirmed the binomial distribution (Mikkelsen et al., 1996), others have not (Table 1). Lu & Kato (2001) showed that the number of C-chromosomes per gamete was in close accordance with a binomial distribution, while in surviving embryos the distribution was skewed to the left (Table 1). They concluded that this biased transmission of C-chromosomes occurred during the embryo stage, as a result of either lower fertilization capacities of aneuploid gametes or higher abortion rates of aneuploid embryos. Indeed, seed set in AAC × AA crosses is typically very low (4.5%; Lee & Namai, 1992). However, Tomiuk et al. (2000) challenged the idea of gametic selection causing biased transmission. They argued that C-chromosomes form multivalents in the gametes and this clustering may also cause biased transmission, even without selective embryo abortion. Whether or not there is gametic selection against C-chromosomes can be ascertained by counting the total number of C-chromosomes passed on to the offspring in the AAC × AA cross. This involves summing all chromosome numbers in the BC1, weighted for the frequency at which they occur (Table 1). Instead of the expected transmission of 4.5 C-chromosomes (50%), only 3.6 C-chromosomes (40%) were passed on to the next generation in the data set of Lu et al. (2002). However, 50% were passed on in several other studies, which suggests no selection against gametes carrying C-chromosomes (Mackay, 1977; Fantes & Mackay, 1978; Leflon et al., 2006; see Table 1).

Table 1.   Unpaired C-chromosomes in BC1 offspring produced in various crosses between AAC Brassica napus and AA Brassica rapa plants
 0123456789C-transmission (%)Reference
  1. 1Calculated from Table 3 in Leflon et al. (2006). The authors presented chromosome number in the AAC plant minus chromosome number from recurrent gamete (in this case A with number 10). We deducted 10 to arrive at the number of C-chromosomes given here. Plants with –1 and zero C-chromosomes were pooled (0 C-chromosomes) as well as plants having nine or more C-chromosomes (9 C-chromosomes).

  2. 2Calculated as in Lu & Kato (2001); see the Description section.

  3. 3From Table 1 in Hansen et al. (2001).

Theoretical %0.21.8716.424.624.616.471.80.250 
Observed % et al. (2006)1
Relative fitness210.1250.0550.0150.0220.0240.0310.0410.0610.33
Observed %61310691716156349.8Fantes & Mackay (1978)
Relative fitness10.
Observed %16.113.910.413. et al. (2002)
Relative fitness10.0950.0180.0100.0050.0050.0060.0130.0420.335
Observed %28384334903826.3Hansen et al. (2001)3
Relative fitness10.1510.0040.0010.0010.0010.00400.0120.286

Using Table 1, Lu et al. (2002) computed, with and without biased transmission, the distribution of the number of C-chromosomes in backcrosses of F1 hybrids (AAC) to B. rapa. In this paper we extend their idea to a large outcrossing population of B. rapa in which there is continuous input of C-chromosomes through crossing with an oilseed rape crop. We first consider the simplest case with nonbiased transmission of unpaired C-chromosomes and a binomial distribution of C-chromosomes in the gametes. Next we consider biased transmission of univalent C-chromosomes, by using data including gametic selection (Lu et al., 2002) or only the clustering of C-chromosomes in gametes (Fantes & Mackay, 1978). Finally, we incorporate weak and strong selection against adult plants carrying C-chromosomes in our model. For all cases, we examine how the shape of the frequency distribution of C-chromosomes in the recipient B. rapa population develops and ask how much selection is needed to eliminate C-chromosomes from the population.



We consider a large well-mixed wild B. rapa population in which there is yearly repeated input of C genomes as a result of cross-fertilization by pollen from a nearby B. napus crop. B. rapa and F1 hybrids are strictly self-incompatible (Hansen et al., 2003). In addition to 10 A-chromosome pairs, F1 hybrids entering the population have nine unpaired C-chromosomes. We used the following procedure.

(1) Plants are assumed to carry between zero and nine C-chromosomes, and their frequencies are denoted as f0, f1, f2f9; the few plants having more than nine C-chromosomes are included in the last category.

(2) Next we compute the gamete frequencies P0, P1, P2P9 with different C-chromosome numbers, assuming that gamete C-chromosome numbers follow a binomial distribution. For a parent containing i C-chromosomes the probability Pk that its gamete contains exactly k C-chromosomes follows the binomial distribution inline image, which reduces with = 0.5 to inline image. Then, assuming random mating, these gametes form embryos with zero to nine C-chromosomes. Equations used for random mating are: inline image, inline image, inline image and so on up to f8; inline image. In theory, plants may contain more than nine C-chromosomes but the error made in pooling individuals with more than nine C-chromosomes in the highest category (9 C-chromosomes) is small when dealing with populations with low B. napus pollen input. The pooling of plants in the 9 C category results in a slight underestimate of the average number of C-chromosomes in the population.

We assumed that all chromosomes are unpaired and inherited independently of one another. In reality, when two plants with many C-chromosomes mate there is of course a chance that their offspring will have one or more matching pairs of C-chromosomes. In that case each gamete will contain one copy of that chromosome rather than zero (25% probability), one (50%) or two (25%) copies as the binomial distribution dictates. This slight change in the distribution of C-chromosomes over gametes again has little effect on our results.

(3) As an alternative to the binomial distribution we multiply the frequencies of gametes having i C-chromosomes by their relative fitness (Table 1), as in Lu & Kato (2001), where fertilization fitness Fi = (observed frequency of class i/theoretical frequency of class i) and relative fitness Ri = Fi/F0. The relative fitness of gametes with zero C-chromosomes is thus always 1.

There are initially Pi gametes (0 < < 9) with i C-chromosomes before selection. After biased transmission the new frequency Pi′ of surviving gametes with i C-chromosomes is P


This equation shows that it does not matter whether we use fertilization fitness or relative fitness, because if Fi/F0 is substituted for Ri, the term F0 is in both the numerator and denominator and thus cancels out. The equation also shows that how many copies of a certain class are passed on to the next generation depends on the frequency of other classes in the population. If a population consists of nearly 100% plants with zero C-chromosomes, then inline image, and the contribution to the next generation of a rare class with i chromosomes equals its relative fitness Ri. This is a severe restriction.

To illustrate this point and to clarify the calculation, first consider a cross between a parent with zero and another parent with, for example, four C-chromosomes. With unbiased transmission the offspring resulting from such a cross would differ in the number of C-chromosomes they possess, but the average would be two, that is, 50% transmission of C-chromosomes. Secondly, the same cross with biased transmission is outlined in Table 2. The initial frequency of gametes with zero to four C-chromosomes formed is calculated using the binomial equation (see paragraph (2)). Then, after biased transmission, the expected number of C-chromosomes in the surviving gametes is 0.747 instead of 2: 18.6% transmission instead of 50%. Selection against C-chromosomes is thus much stronger than expected when one compares the observed vs expected frequencies in Table 1. The strong effect is a result of the fact that gametes with zero chromosomes occur at a fairly high frequency in this biparental cross and have the highest relative fitness. We return to this issue in the Discussion. It would also be possible to select in the embryo instead of the gametic phase, but we did not pursue this option.

Table 2.   Example illustrating how using relative fitness affects the number of C-chromosomes passed on to the next generation in a cross between a parent with zero and a parent with four C-chromosomes
No. of C-chromosomes (i)Initial Pi in gametesRelative fitness (Ri)1After biased transmissionNew PiC-chromosomes passed on
  1. 1See the Leflon et al. (2006) data in Table 1.

   PiRi = 0.1195 Total = 0.747

(4) New hybrids, all containing nine C-chromosomes, are then added to the population each year in the seed stage. This is because estimates of hybridization rates in the literature are typically based on seeds and young plants, that is, after selection among gametes and embryos has taken place.

(5) Plants with zero C-chromosomes produce more offspring than all others. Plants carrying one to nine C-chromosomes have fitness 1 – s, whereas pure B. rapa plants have fitness 1. Thus s is the selection coefficient (0 < < 1) at the level of the adult plant. A high positive value of s means strong selection against plants carrying C-chromosomes.

(6) After normalization all frequencies add up to 1, and the cycle is repeated for 10 or 50 generations.

Model parameters

In a recent review, Devos et al. (2009) reported that between 0 and 69% of all seeds produced by B. rapa plants near an oilseed rape field were hybrid. Simard et al. (2006) reported a similar range, but noted that the hybridization rate rapidly declined with increasing density of B. rapa plants and distance to the crop. Warwick et al. (2003) reported 13.6% hybrid seeds near oilseed rape crops. In a mixed experimental population, Halfhill et al. (2002) reported between 0.7 and 16.8% hybrid seeds. Of course, under conditions where species distributions only partly overlap there is less opportunity for hybridization. Landbo et al. (1996) found no hybrids in seeds produced by autumn-flowering B. rapa and an average hybridization rate of 3%. We chose hybridization rates of 10% and 1% for our model.

F1 hybrids typically grow well (Hauser et al., 1998a; Halfhill et al., 2005), but their percentage seed set and pollen viability are lower than those of B. rapa (Mikkelsen et al., 1996; Ammitzbøll et al., 2005; Allainguillaume et al., 2006). For instance, Hauser et al. (1998b) documented that F1 hybrids produced 7.5 times more seed pods than B. rapa under field conditions, but because of reduced seed set this resulted in only a threefold fitness advantage of these hybrids over B. rapa. In a competition experiment, Sutherland et al. (2006) demonstrated that F1 hybrids produced c. 80% of the seed output of B. rapa. Allainguillaume et al. (2006) estimated that F1 hybrids yielded 46.9% of the seed output of B. rapa. Average pollen viability in F1 hybrids may vary between 54 and 84.5% (Warwick et al., 2003; Leflon et al., 2006). F1 hybrids thus generally do quite well.

Contrary to the strong performance of hybrids, BC1 plants generally produce very low seed numbers per pod and do worse than B. rapa, although Hauser et al. (1998b) noted that some BC1 plants were as fit as B. rapa and that survival after planting was high for all types. Rose et al. (2009) reported lower seed yield of some backcross lines but none of the differences were significant. Rose et al. (2009) regressed seed yield of hybrids under competitive conditions against the number of B. rapa markers in the genome and found a significant positive correlation. This could suggest that the number of C-chromosomes present is negatively correlated with seed yield, but regression data are currently not available.

Taking all data into account, Andersson & de Vicente (2010) concluded that overall fitness costs of carrying C-chromosomes may be low. This is certainly true for the F1 but debatable for the BC1, for which there are fewer data. The fit plants may be the ones that have already lost their C-chromosomes. Our model covers a range of selection regimes against hybrids: slightly reduced fitness of hybrids (= 0.1; i.e. 90% of the fitness of pure B. rapa), strongly reduced fitness (= 0.9) and intermediate cases (= 0.3, 0.5, 0.7).


To check our model, we first looked at a single hybridization event, that is, input of C-chromosomes, without selection. In that case C-chromosomes are diluted in the population and after a few generations the number of C-chromosomes per plant approximately follows a normal distribution (of course constrained between zero and nine; data not shown).

Unbiased transmission of univalent C-chromosomes

With some hybridization annually (10%) and no selection against hybrids (= 0), populations eventually consist of plants that all have nine C-chromosomes. This is because each year there is input of C-chromosomes into the population and this input is not counteracted by selection. When imposing weak selection (= 0.1), C-chromosomes continue to accumulate and pure B. rapa plants, with zero C-chromosomes, rapidly disappear from the population; their frequency declines from 100 to 1.7% in just 10 generations, after which it eventually falls to zero (Fig. 1). We find a similar pattern when selection is somewhat stronger (= 0.3), but here the zero class disappears at a slower rate. With stronger selection against adult plants with C-chromosomes (= 0.5) the class with zero C-chromosomes is no longer eliminated and all different types coexist within the population; after 50 yr the population consists of 50.0% pure B. rapa, 7.4% hybrids with nine C-chromosomes and 42.6% intermediate hybrids with one to eight C-chromosomes (Fig. 1). When selection is even stronger (s = 0.7 or 0.9), hybrids with intermediate C-chromosome numbers become very rare and the population essentially consists of pure B. rapa and F1 hybrids that are continuously formed de novo through cross-fertilization with B. napus. With = 0.9 the population consists stably of 98.7% pure B. rapa plants, 1.0% hybrids with nine C-chromosomes and only 0.3% intermediate hybrids (not shown in Fig. 1). This equilibrium is reached after just a few generations, because the simulations start with 100%B. rapa and 10% annual input of hybrids, which is already close to the equilibrium.

Figure 1.

Frequency of plants in a Brassica rapa population having zero to nine unpaired C-chromosomes. Each year, 10% of the B. rapa plants are fertilized by Brassica napus pollen, leading to an influx of nine C-chromosomes. Mating is random and transmission of C-chromosomes is unbiased. Results are given for the starting population (pure B. rapa at = 1; closed squares, solid line), after 10 generations (closed circles, broken line) and after 50 generations (open circles, dotted line). The selection coefficient s denotes selection against plants with C-chromosomes, ranging from 0.1 (weak) to 0.7 (strong selection against adult plants carrying C-chromosomes).

In Fig. 2 all intermediate (one to eight) numbers of C-chromosomes are pooled. When selection is weaker than = 0.5, plants with intermediate C-chromosome numbers are frequently found within the population; selection coefficients of = 0.7–0.8 or higher are needed to reduce the fraction of intermediates to a few per cent.

Figure 2.

The fraction of plants with one to eight C-chromosomes (after 50 yr) decreases with selection during the adult stage. The model calculations assumed unbiased transmission and 10% hybridization between Brassica rapa and Brassica napus. Zero C, dark gray; one to eight C, light gray; nine C, mid gray.

With a lower hybridization rate of 1% the patterns are essentially the same but changes occur at a slower rate (results not shown). With weak selection (= 0.1) pure B. rapa has steadily declined from 100 to 20.8% after 50 yr, and continues to do so until it eventually disappears. With = 0.3 we find coexistence of many types in the population: after 50 yr the population consists of 90.2% pure B. rapa, with 0.7% of the plants having nine C-chromosomes and the remaining 9.1% having an intermediate (between one and eight) number. A further increase in the strength of selection (= 0.5) essentially eliminates the intermediates from the population (1.8% remaining after 50 yr). With even stronger selection (= 0.9) < 0.02% intermediates persist.

Biased transmission of univalent C-chromosomes

Our simulations contrasted gametic selection with clustering only. The results are extremely similar for all data sets with biased transmission in Table 1. Intermediates with one to eight chromosomes are very rare after 50 yr, reaching a maximum of 0.6% without any adult selection and declining to 0.006% with strong selection (= 0.9) (10% hybridization) (Fig. 3). With only 1% hybridization, the fraction of pure B. rapa plants is even higher; there are essentially no intermediates with one to eight C-chromosomes in the population, only the new F1 hybrids that are added each year.

Figure 3.

Frequency of plants in a Brassica rapa population having zero to nine unpaired C-chromosomes. Each year, 10% of the B. rapa plants are fertilized by Brassica napus pollen. Mating is random, but now with biased transmission of the C-chromosomes using the data on relative fitness of Lu et al. (2002) from Table 1. Results are given for the starting population (pure B. rapa at = 1; closed squares solid line). After 10 generations (open circles, broken line) and 50 generations results are indistinguishable. The selection coefficient s denotes selection against plants with C-chromosomes. (a) = 0 (no selection); (b) = 0.9 (strong selection).

The reason for these results is simply the very high relative fitness of gametes with zero C-chromosomes, as explained in the model section. The effect is so dominant that additional gametic selection against plants with C-chromosomes, as in the data of Lu et al. (2002), hardly makes any difference.


Unbiased or biased transmission?

Extrapolating results from the AA × AAC cross to other crosses gives unlikely results (see also Table 1 in Lu et al., 2002). Biased transmission rapidly eliminates all C-chromosomes from the population and is therefore incompatible with the high frequency of plants with intermediate numbers of C-chromosomes found in the Danish population (Hansen et al., 2001). When clustering of chromosomes is a major factor, as several studies in Table 1 suggest, it is also not plausible to apply the relative fitness concept to crosses between plants with few C-chromosomes. The best way to proceed is of course to make crosses between plants with different numbers of C-chromosomes and test how well relative fitness predicts C-chromosome number in the offspring. The crosses should be made using wild accessions of B. rapa rather than cultivars, as few studies have used wild plants and results depend on the cultivar used (Leflon et al., 2006; Lu & Kato, 2001; Xiao et al., 2009). If we accept nonbiased transmission of C-chromosomes as the most plausible scenario at the population level, strong selection (> 0.7; Fig. 2) against plants carrying C-chromosomes is required to produce the UK and Dutch scenario of populations consisting of pure B. rapa, F1 hybrids and no intermediates. This challenges the commonly held view that C-chromosomes cause little or no decline in fitness (e.g. Légère, 2005; Andersson & de Vicente, 2010).

Performance of the BC1 in the field

We expect the BC1 to do poorly in the field because of low viability. Stace (1975, p. 36) already noted that: ‘… it has frequently been found that whereas F1 plants are easy to rear in cultivation, seemingly without special attention, they have not been found in nature.’Wei & Darmency (2008) also emphasized the difference between cultivation and nature. Smaller seed size of Brassica hybrids had no effect when plants were grown in the laboratory, but negatively affected survival under more stressful field conditions.

Viability selection may have influenced the results of Halfhill et al. (2003). These authors made a BC2 (offspring of an AA × AAC cross were again crossed to AA) by hand pollination. Halfhill et al. (2003) then placed the BC2 plants in a mixed bulk population in which they were pollinated by flies. Ten examined plants grown from the seeds produced in the bulk population had lost all C-chromosomes and the authors concluded that most C-chromosomes were already lost in the BC2. Alternatively, C-chromosomes were still present in the BC2 but plants with C-chromosomes had a large selective disadvantage in the bulk population with open pollination.

In the case of hybridization between B. napus and B. rapa the F1 is not the critical phase, for such plants grow well (see the Description section) and have been found flowering and producing pollen and seed in nature (Hansen et al., 2001; Wilkinson et al., 2003; Luijten & de Jong, 2010); the problem appears to be in the BC1. These ideas can further be tested by collecting seeds of flowering F1 hybrids in nature (apart from their appearance such plants can be identified by flow cytometry) and sowing these seeds under both noncompetitive and competitive conditions in the field.

Hybridization in the field

How does the Brassica system compare to other hybridization studies? Several recent studies have documented frequent occurrence of hybrids and, importantly, repeated backcrossing to one of the parental species, thus leading to gene flow between species and reticulate evolution (Marhold & Lihova, 2006; Arnold & Martin, 2010; Ducarme et al., 2010; Tang et al., 2010).

Introgression across different ploidy levels is less well documented and is thought to be rare. Chapman & Abbott (2010) attribute this to intermediates producing a high fraction of unfunctional nonbalanced gametes. Nonetheless, introgression across ploidy levels does sometimes appear to occur in nature. Kim et al. (2008) recently showed that a gene for an adaptive trait, the production of ray florets, had been transferred from the diploid Senecio squalidus to the tetraploid Senecio vulgaris. This is the only study in which introgressed genes affecting fitness have been isolated and characterized. Further evidence from gene exchange among ploidy levels comes from observation of the ploidy level of F1 hybrids and the occurrence, as established plants in the field or as seeds, of intermediates between the F1 and one of the parental species. Chapman & Abbott (2010) conclude, in accordance with Stebbins (1971), that the gene flow occurs most often from the lower to the higher ploidy level (as in the Senecio example). Only a handful of studies suggest gene flow from tetraploids to diploids (as in our Brassica model). Such cases, which are further discussed in Elkington (1984), include Euphrasia anglica/Euphrasia micrantha (Yeo, 1956), Cochlearia officinalis/Cochlearia danica (Fearn, 1977) and Dactylorhiza fuchsii/Dactylorhiza purpurella (Lord & Richards, 1977).

The presence of extra functional chromosomes could be a remnant of introgression. Most reports of aberrant chromosome numbers refer to deletion or duplication within the species (Levin, 2002). Aneuploids occurred in field-collected seeds of four clones of Fallopia complex that are involved in interspecific hybridization (Saad et al., 2011). Some lines of Plantago lagopus apparently contain an extra functional chromosome that does not match other chromosomes from the normal complement (Dhar & Koul, 1995). In the Boechera complex some lines are aneuploid and these lines are then often apomictic (Sharbel et al., 2004). Yet aneuploid lines are rare in natural populations and Dobes et al. (2006) suggest that this is a consequence of the need for chromosomal balance for embryo formation.

The addition of functional chromosomes, as singletons or in pairs, apparently has a greater impact on the phenotype than whole-genome changes (ploidy) (Birchler & Veitia, 2010). Birchler & Veitia (2010) argued that this is a result of dosage effects disturbing a balance at the genome level. This fits in with the idea that the F1 (AAC) genotype with all nine C-chromosomes does quite well but problems start for BC1 plants with different combinations of C-chromosomes. It would certainly be interesting, from both an applied and an evolutionary point of view, to detail the phenotypes and fitnesses of the different hybrid classes in the offspring of the AA × AAC crosses in Brassica.

Introgression of transgenes

The likelihood of introgression depends on where transgenes are inserted into the DNA of B. napus. This can be into the nuclear DNA, either the A or C genome, and in theory also into the cytoplasmic DNA. Under current legislation in the European Union the applicant does not need to specify where the transgene is inserted. This information is therefore unavailable for the GM lines of oilseed rape that have been approved for import into the EU. What are the implications of the insertion site for introgression?

A transgene on a chromosome of the A set in B. napus will be passed on to 6–28% (Table 1) of the offspring in the BC1 that have lost all their C-chromosomes. The transgene is then stably incorporated into the B. rapa genome. Warwick et al. (2008) have reported on a gene for herbicide resistance being transferred from B. napus to B. rapa. The simplest way for this to have happened is if the transgene were located on a chromosome of the A-set. Haider et al. (2009) found that two wild B. rapa populations that were sympatric with B. napus contained a chloroplast typical for the latter (see also Hansen et al., 2003). As chloroplasts are maternally inherited (Johannessen et al., 2005), this suggests that B. napus was the maternal parent in the crosses with B. rapa and C-chromosomes were subsequently lost.

With the transgene on C there are two routes for introgression. First, when the transgene has a positive effect 1 + sT on fitness, for instance because it makes the plant resistant to a herbicide, such an extra chromosome could be selected for (Tomiuk et al., 2000; Lu et al., 2002) when the selective advantage is greater than the disadvantage of being aneuploid. In the present model, plants with an extra chromosome carrying the transgene would have a fitness 1 + sTs and eventually all plants would contain the extra C-chromosome when sT > s. Secondly, transgenes on C could also be incorporated in the A genome by homeologous recombination. This can happen first during meiosis of the F1 hybrid (Tomiuk et al., 2000). If no BC1 plant with extra C-chromosomes survives, the transgene can still be inserted into the AA genome in this way. The chance of homeologous recombination occurring in the field can be evaluated by, again, collecting seeds of the BC1 plants in nature and screening the AA seeds without C-chromosomes for the presence of C-specific markers.

Insertion of transgenes on C-chromosomes, instead of A-chromosomes, results in a much lower chance of introgression. This has already been argued by several authors based on the transmission rate being < 50% in some of the studies in Table 1 (Lu et al., 2002; Zhu et al., 2004; but see Mikkelsen et al., 1996). When BC1 plants with C-chromosomes are unfit, there is only one opportunity for a transgene on C to be transferred to an A-chromosome: during meiosis of the F1 and subsequent survival of the AA seed. This probability of introgression is much lower than for a transgene that is already on A. The old recommendation of using C- rather than A-chromosomes for insertion of transgenes in B. napus should therefore be taken seriously by both bioindustries and European legislators. If the environmental risk assessment shows that BC1 plants are unfit under natural conditions and are unlikely to carry the transgene as a result of homeologous recombination, then the probability of gene flow from oilseed rape C-chromosomes to wild B. rapa may still be low.


We thank the Netherlands Organization for Scientific Research for funding through the ERGO program, grant 383.06.112. Nigel Harle brushed up our English.