2.3.4.3 Fertility and fitness of hybrids

This section discusses hybrids that have been shown to backcross spontaneously with at least one of their parental species.

a. B. rapa x Brassica napus = (B. x harmsiana)

The PGS application states that "[B. x harmsiana] is capable of producing a small amount of viable seeds under open pollination". This may be a reference to spontaneous F₂ production in experiments reported by Kajanus in 1917 (see Scheffler & Dale, 1994). Scheffler & Dale reported that many studies produced backcross (BC) and F₂ seed by manual pollinations.

New information on the fertility of B. x harmsiana comes from the work of Jørgensen and her colleagues. Jørgensen & Andersen (1994) collected seed from a weedy population of B. rapa in a field of winter oilseed rape. Among 78 plants grown from this seed were two B. rapa-like plants expressing a B. napus C genome-specific isozyme (but with presumably otherwise normal B. rapa isozyme phenotypes). One plant had a chromosome number of 2n=21-22 (compared with 2n=29 for most B. x harmsiana) and the other died before a count was made. These data suggest that the plants were at least first generation backcrosses (BC₁) of B. x harmsiana with B. rapa. Jørgensen et al. (1996) studied a "self-maintained" population of oilseed rape, weedy B. rapa and B. x harmsiana in the ratio 10:5:1 at the time of observation. In the field, the hybrids had lower pollen fertility (12-54%, mean 54%, N = 16) than the parental species (B. rapa 89-99%, mean 96%, N = 10; B. napus 99-100%, mean 100%, N = 6) and much lower seed set (102 seeds per plant compared with 3000+ for the parents). Jørgensen et al. (1996) germinated seed from 16 B. x harmsiana parents and selected 20 plants with B. rapa-like morphology. All plants had chromosome numbers of 2n > 25, most had high (> 70%) pollen fertility and the number of oilseed rape-specific molecular markers per plant was between 0-16. Most of these plants, therefore, were probably backcrosses to B. rapa.

Mikkelsen et al. (1996) produced B. x harmsiana on both parental species by spontaneous hybridisation. The B. napus parent was glufosinate-resistant. The hybrids were grown in small plots with B. rapa and produced more than 450 seeds per plant. Over 4,000 seeds from 32 hybrids were germinated and 865 randomly-chosen glufosinate-resistant plants were grown to maturity. Of these, 44 with B. rapa-like morphology were analysed and several were found to have 2n=20 and high pollen fertility (>90%). Four of these plants were backcrossed to B. rapa, and an average of 6.4 second generation backcross (BC₂) plants were produced per pollination. The BC₂ plants showed strong seed dormancy (like B. rapa) and 42% of the 416 plants were glufosinate-resistant.

Metz et al. (1997) studied the inheritance of the glufosinate-resistance transgene through several generations of backcrosses using manual pollinations. Two lines of oilseed rape hemizygous for the bar gene were crossed with Chinese cabbage and Pak Choi (both B. rapa cultivars). Herbicide-tolerant plants in F₁ and backcross generations were selected for further rounds of crossing. The F₁ between oilseed rape line TP3 and Chinese cabbage showed 1:1 segregation of the herbicide-resistance phenotype, as did F₁s between TP3 and Pak Choi. Backcrosses between resistant F₁s and their respective B. rapa parent also showed 1:1 segregation. Crosses between oilseed rape line TP2 and Pak Choi also showed no significant difference from 1:1 segregation in the F₁ generation. However, the resistant phenotype was inherited by only 26% of BC₁ plants, 5% of BC₂, 11% of BC₃ and 9% of BC₄ (in all cases significantly different from 1:1). Metz et al. (1997) inferred that the bar gene was integrated into the A (B. rapa) genome of TP3 and the C (B. oleracea) genome of TP2.

The work of Jørgensen et al. and Metz et al. shows that herbicide-tolerance genes can introgress into B. rapa from B. napus by spontaneous hybridisation. However, the likelihood of introgression is strongly dependent on whether the gene is integrated into the A or the C genome. With high resolution linkage maps of oilseed rape (e.g. Parkin et al., 1995), it should be possible to find out whether the transgene is in the A or C genome without having to infer it from phenotype data (which may be sensitive to factors that alter gene expression).

There appear to be no insurmountable fertility or compatibility barriers to transgene introgression from B. napus to B. rapa. However, if the F₁ or BC generations do not survive in the field, introgression will not occur. Hauser et al. (1998a) compared the fitness of B. x harmsiana with that of its parents. Three (non-transgenic) varieties of B. napus and three populations of B. rapa were crossed in a factorial design to give all 36 possible reciprocal crosses (18 hybrids, 9 B. rapa, 9 B. napus). From each B. rapa population 12 and 6 parents were used, and from each B. napus variety 6 and parents. This gave 648 unique maternal-paternal combinations. After pooling of seed from B. rapa () x B. napus (), mature seeds were available from 357 combinations. Seeds were sown in pots and placed in a field with (after thinning) a single plant from each combination in each of three randomised blocks of 25 pots m_-2. The trial was weeded, watered and sprayed with insecticide, but not fertilised.

Several measures of plant performance were recorded, and the following were multiplied to give a combined estimate of plant fitness: proportion of fully developed seed on the maternal plant (zygote survival); survival from day 17 to harvest (pre-day 17 deaths were excluded because of dormancy and drought); number of pods per offspring plant and number of seeds per pod on offspring plants. Considering the overall fitness only, the results can be summarised very simply (Br = B. rapa; Bn = B. napus):

• The reciprocal hybrids were not significantly different from one another
• The hybrids were fitter than Br
• The hybrids were less fit than Bn

The pattern for the individual components varied:

• There were no significant differences in the proportion of fully developed seeds among Br, Bn x Br and Bn. Br x Bn produced a lower proportion of developed seeds than the others
• Survival from day 17 showed no significant differences among all parents and hybrids
• Br x Bn and Bn x Br each produced over twice the number of pods per plant than Bn, but the differences were not significant. Br produced significantly fewer pods per plant than the others
• Both Br x Bn and Bn x Br produced significantly fewer seeds per pod than Br. Bn produced significantly more seeds per pod than Br

Hauser et al. (1998b) carried out a similar experiment to compare the fitness of F₂ and BC₁ generations with the parent species. Two B. rapa populations and B. napus varieties and their four hybrid combinations were used in the crossing design. As parents, Hauser et al. (1998b) used 16 B. rapa plants per population (two plants from each of eight full-sib families from the F₁ trial), 22 F₁s (two each of 11 full-sib families) and 16 of each B. napus family (from the original seed lots). Each B. rapa and B. napus parent received pollen from three plants from its own population/ variety and four plants from each of the two F₁ hybrid populations derived from that population/variety (to give BC₁: Br x F₁ and Bn x F₁). Each F₁ parent received pollen from four plants from its own F₁ population (to give F₂) and from the B. rapa and B. napus population/variety from which it was derived (to give BC₁: F₁ x Br and F₁ x Bn). Plants raised from 644 fully-developed pods were grown as in the F₁ trial and the same fitness components were estimated.

The results for the combined fitness components were:

• The parental species and BC₁ with Bn as parent were the fittest (although the Bn estimate was much the highest, it was not significantly different from the other two lines)
• BC₁ with Bn as parent was fitter than both reciprocal BC₁s with Br
• F₂ hybrids were the least fit

Again, the pattern for individual components was variable:

• F₂ hybrids were always the least fit, apart from survival from day 25 to harvest
• All hybrid generations showed significantly lower proportions of fully developed seeds compared with the parental lines
• Both reciprocal BC₁s with Bn had higher numbers of pods than other lines, although they were not significantly higher than the parental lines
• All hybrid lines showed significantly lower numbers of seeds per pod than the parental lines

In general the results showed large variation within lines. Although mean values for some lines were more than twice the value of other lines, they were not always significantly different (e.g. the number of pods on F₁ hybrids). Also, while on average F₁s were less fit than B. napus, and F₂s and BC₁s were generally less fit than both parents, some individual hybrid plants were as fit as parental lines. Consequently, Hauser et al. suggested that low fitness in F₁ and subsequent generations will not completely prevent introgression from B. napus to B. rapa.

Under conditions that were similar to cultivation, hybrids were generally less fit than B. napus. This implies that herbicide-resistant hybrids will be no more of a weed problem than volunteers of resistant oilseed rape. There is, however, one possible route by which glufosinate-resistant plants with greater weediness than oilseed rape could be produced. Linder (1998) showed that B. x harmsiana, like B. napus, completely lacks dormancy. Therefore, the occurence of B. x harmsiana will cause no greater persistence of the transgene in the seedbank than the oilseed rape itself. However, if transgenic B. x harmsiana is crossed with B. rapa, BC₂ plants having both pronounced seed dormancy and the transgene can arise (Mikkelsen et al. 1996). While this scenario is unlikely in oilseed rape fields, if it did happen, herbicide-tolerant volunteers might survive in the seedbank for longer than if the transgene was present in the pure oilseed rape variety, B. x harmsiana, or backcrosses with oilseed rape as the recurrent parent.

That F₁ hybrids are fitter than B. rapa suggests that they could establish in natural populations of B. rapa. However, Scott & Wilkinson (1998) found no establishment of F₁ hybrids in B. rapa populations adjacent to oilseed rape fields. As discussed above, the frequency of hybridisation in natural populations is low because most pollinations are intraspecific. Also, in natural populations there are very high rates of seedling mortality. It is possible that under these conditions the relative fitness of the hybrids is much lower than B. rapa. Experiments of Linder & Schmitt (1995) suggest that seedling emergence and growth rate of B. rapa and B. x harmsiana are similar, although seedlings were not grown in competition. Linder (1998), however, found that B. x harmsiana completely lacked dormancy, whereas its B. rapa parent had a high proportion of dormant seeds. Therefore, the hybrid may germinate during unfavourable conditions and have much higher mortality than B. rapa. It is possible, of course, that recruitment of hybrids and B. rapa occurs randomly, and the likelihood of recruitment of hybrids is simply proportional to their frequency. Taken together, the results of Hauser et al. (1998 a & b) and Scott & Wilkinson (1998) imply that while introgression can occur, the lower fitness of the hybrids (especially the F₂ and backcross generations) and their infrequent formation mean that gene flow to natural populations of B. rapa will be rare and erratic.

b. Raphanus raphanistrum x B. napus

Darmency et al. (1998) looked at the fertility of radish x oilseed rape hybrids. Plants with radish as parent were produced by in vitro culture. Fifteen hybrids were grown with 90 Wild Radish plants and 10 seeds were collected from the hybrids (i.e. <1 per plant). Hybrids with male-sterile oilseed rape as parent were grown under similar conditions and produced 3.2 seeds per plant. Wild Radish plants produced an average of 3,600 seeds in these experiments.

Four seeds from the in vitro hybrids germinated; three were sterile and one produced 70 seeds. Most seedlings from the male-sterile parent hybrid were accidentally lost. The remaining plant produced six seeds.

Chèvre et al. (1998) studied spontaneous backcrossing between oilseed rape x wild radish hybrid plants (Bn x Rr) and wild radish parents. The F₁ hybrids were generated by crossing each of six oilseed rape lines with a mixture of two wild radish populations. Three of the oilseed rape lines were male-sterile varieties with Ogu-INRA cytoplasm. The other three were F₁s between these varieties and Westar T5, which contains the bar gene. Male-sterile F₁Bn x Rr plants were sown in a 1:1 ratio with wild radish. Seed production was very low on all Bn x Rr lines, with on average fewer than one seed per plant.

Backcross seed from herbicide-tolerant F₁ interspecific hybrid parents were used to study the inheritance of the transgene. Eighty-two percent of BC₁ seed were Basta-resistant. About half of the BC₁ plants were produced from unreduced gametes on the F₁ side (i.e. BC₁ plants were ACRrRr, 2n=37). About 16% had fewer than 37 chromosomes. Most of these had 2n=28 - the same chromosome number as the F₁ (ACRr). Of the 34% with more than 37 chromosomes, most had 2n=56, twice the F₁ number (i.e. ACRrACRr). This suggests some male fertility in at least some of the F₁ lines.

The BC₁ plants had low male fertility, but improved female fertility compared with the F₁s. In a field trial with same design as the F₁ x Rr trial, BC₁ plants produced 11 seed per plant. In general, the BC₂ plants had lower chromosome numbers than their respective BC₁ parents and about 10% had 2n=18 - the same number as wild radish. The proportion of 2n=18 progeny was 37% from the 2n=28 BC₁ parents. Only 54% of BC₂ progeny of herbicide-tolerant mothers were tolerant themselves. Over half the BC₂ plants were male-fertile and seed production in crosses with radish was 229 seeds per plant.

Unlike B. rapa, wild radish does not share a common genome with oilseed rape. Therefore, introgression of transgenes into wild radish by repeated backcrossing is difficult. Some multivalent formation did occur in BC₁ plants, which means that recombination could have occurred between the oilseed rape and radish genomes. However, is not known whether gametes produced from meioses with multivalents were viable. The lower proportion of herbicide-tolerant plants, and lower chromosome numbers in the BC₂ generation compared with BC₁, suggest successive BC generations lose oilseed rape chromosomes, without recombination introducing oilseed rape genes into the radish genome. Although experiments with male-fertile hybrids backcrossing to wild radish parents have yet to be done, the available data indicate hybridisation and backcrossing with fully-fertile parents is rare, and transgenes are likely to be lost rapidly from backcrosses to wild radish.

c. Hirschfeldia incana x B. napus

Lefol et al. (1995) studied the fitness of B. napus x H. incana hybrids in competition with the parental species. Experiments were done in seed trays (0.12m₂), with ten plants per tray. One plant each of oilseed rape (male-sterile cv. Brutor), wild hoary mustard and their F₁ hybrid were allocated random positions in the tray. The remaining positions were taken by seven oilseed rape or seven mustard plants, to give two "environments". Some trays received nutrient solution, while others were just watered.

On average over the two environments, significant differences were observed for all vegetative traits measured (total leaf area, total fresh weight and total dry weight). In all cases the hybrids were intermediate to the parents. Hoary mustard and hybrid plants were smaller when grown with oilseed rape than with hoary mustard. No measure of reproductive performance was made.

Lefol et al. (1996a) found that hybrids between male-sterile glufosinate-resistant oilseed rape and wild hoary mustard had low fertility. Hybrid plants grown with hoary mustard at a ratio of 4:30 produced almost no fertile pollen. Compared with hoary mustard, hybrids produced 40% fewer flowers per unit inflorescence length, 20 times fewer pods per flower, 50 times fewer seeds per pod, and 200,000 times fewer seeds per plant (0.5 - 0.6 cf. 96,500 - 92,000). In separate experiments, hybrids and hoary mustard plants were grown in ratios of 16:18, 12:18 and 4:30. A total of 168 hybrids produced 32 seeds, of which only five germinated. Chromosome counts (2n=40) and isozyme phenotypes suggested that at least some of the progeny resulted from unreduced gametes of the hybrid (2n=26) and hoary mustard (2n = 14).

These experiments suggest that transgene introgression from oilseed rape to hoary mustard will be a very slow process, if it is even possible. Hybrids can compete vegetatively with hoary mustard; however, as reciprocal crosses were not made, the competitive ability of the hybrids against the hoary mustard parent may be a maternal effect of the oilseed rape. If hybrids do establish in natural populations, their fertility will be low. In addition, recombination between the oilseed rape and hoary mustard genomes has not been proved. Therefore, backcross plants will be very rare and the likelihood of transgene loss in subsequent generations is high.

2.3.4.4 Will genetically modified oilseed rape outcross with greater frequency than conventional varieties?

A crucial assumption in discussions of gene flow is that genetic modification does not change the frequency with which a crop hybridises with its wild relatives. PGS compared the results of work with transgenic and non-transgenic rape and concluded that genetic modification did not change outcrossing ability.

However, a recent paper by Bergelson et al. (1998) showed that Arabidopsis thaliana, genetically modified with a gene for herbicide-tolerance, outcrossed more than plants with the same gene produced by mutation of wild type plants.

There is no reason to suppose that the PGS oilseed rape is more "promiscuous" than unmodified oilseed rape. First, the Arabidopsis experiment did not compare the transgenic plants with wild-type plants - the mutation to herbicide-tolerance may have had pleiotropic effects that reduced outcrossing ability. Second, there were large differences in outcrossing rate between the two transgenic lines (1.2% and 10.8%). Therefore, generalisations about outcrossing rates of transgenic Arabidopsis lines are not possible and extrapolation to oilseed rape is invalid.

Finally, work done since the PGS application has produced no data that suggest higher outcrossing in transgenic rape (Table 2). Comparison of data collected with various experimental designs in different years is not ideal. Nevertheless, separate field trials in Denmark found 3.0% hybrid seed on B. rapa grown in a 1:10 ratio with unmodified oilseed rape, and 2.6% hybrid seed on B. rapa grown in a 1:11 ratio with the male-fertile form of the PGS oilseed rape. Other crosses (e.g. B. juncea x B. napus) also suggest that transgenic oilseed rape outcrosses with the same frequency as conventional oilseed rape.