The Brassicaceae family comprises about 3000 species, the majority of which are found in the Northern Hemisphere. Many common agricultural weeds, such as Brassica nigra (L.) Koch, Brassica rapa L., Cardaria draba (L.) Desv., Raphanus raphanistrum L., and Sinapis arvensis L., belong to this family. The most important crop species from this family are the oilseed Brassicas; Brassica napus L., B rapa L., and B juncea Coss., which are generally referred to as rapeseed, oilseed rape, or canola. Other widely cultivated species in this family are: B. oleracea L. (cabbage, kale, kohlrabi, Brussels sprouts, cauliflower, and broccoli), B. chinensis L. (syn. B. napus var. chinensis; Chinese cabbage), Raphanus sativus L. (radish), and Armoracia rusticana Gaertn. (horseradish). We agreed to limit our discussion to the rapeseed species because they are the focus of most of the gene transfer technology. Worldwide, rapeseed is grown on more than 20 million hectares; it is the third most important oil plant after palm oil and soybean. Major producers are China, India, Canada, the European Community, and Australia.

In weedy relatives of rapeseed, seed dormancy is a very powerful survival mechanism. Dormant seeds may germinate over a period of several weeks during the growing season, which greatly improves the chance that at least some plants will be successful in replenishing the seedbank. The presence of flowering individuals during a large portion of the growing season increases the opportunity for cross-pollination with populations that produce only one cohort per season. Furthermore, seed dormancy may allow a plant population to survive one or several seasons of complete reproductive failure. Whether or not a particular hybrid will exhibit seed dormancy is rather unpredictable (Landbo and Jørgensen 1997).

Pest resistance genes are bred into cultivated varieties to prevent yield loss. In agricultural systems, pest resistant varieties generally have a significant advantage over non-resistant varieties, and this advantage typically translates into a rapid spread of the resistant variety at the expense of the non-resistant varieties. Humans mediate the spread of pest resistant crop varieties. Whether introgression of a pest resistant gene into a feral population would increase the spread of a wild population is open to question. We, as a group, were not aware of any example where introgression of a pest resistance gene, or the consequences of such an event, had actually been documented.

There is, however, ample evidence that hybridization between closely related species occurs spontaneously (OECD 1997); introgression of resistance genes is therefore quite plausible. Available commercial varieties already have varying degrees of resistance to two of the most common fungal pathogens, Leptosphaeria maculans and Albugo candida. These traits, identified in related germplasm, were introduced into cultivars by conventional breeding. It is reasonable to assume that these resistance traits already exist within the genome of the closely related Brassica species complex, so their reintroduction via gene flow from cultivated transgenic varieties is not likely to have much influence on the fitness of feral populations.

The situation may be different with respect to many of the engineered resistance traits likely to be available over the next few years. Such new traits, imparted for example by a set of stacked genes that confer broad insect and fungus resistance, may protect plants from a wide range of pests. It is conceivable that this situation could trigger an increase in the size and range of the population in question.

There is little doubt that genes from transgenic rapeseed have the potential to escape into related varieties and species. Chèvre et al. (1997) documented the introgression of glufosinate ammonium resistance from B. napus to Raphanus raphanistrum under experimental conditions. Spontaneous production of crop-weed hybrid seeds under field conditions was reported from Denmark and the Netherlands (Jørgensen et al. 1996; De Vries et al. 1992). Consequently, the escape of transgenes is certainly a cause for concern.

Data from a field experiment with Brassica napus containing a Bacillus thuringiensis cry1Ac transgene suggest that this pest resistance gene is not likely to have a significant impact on weediness (Stewart et al. 1997). The Bt gene conferred increased fitness under moderate selection pressure by Plutella xylostella, however, this did not translate into increased competitiveness, nor did the transgenic plants exhibit greater weediness (Stewart et al. 1997).

During the plenary session of this workshop, it was argued that scientists have numerous questions they would like to have answered, but what really needs to be known is a much more restricted set of information. We partially disagree with this statement because essential knowledge can only be identified once we have a sufficient general understanding of the system under consideration.

Our current knowledge tells us that pest resistant transgenic rapeseed varieties are likely to hybridize with a number of weedy species (OECD 1997) and that there is a good chance that transgenic traits will eventually introgress into populations of weedy species (Mikkelsen et al. 1996; Metz et al. 1997). We know from ecological work that herbivory by insects and other organisms is the principal factor in many environments that limits the abundance of plant species (Louda and Potvin 1995). Hence, there is a definite possibility that the newly acquired pest resistance will result in greater fitness of weedy Brassica species, which could be expected to become more abundant. Such an event could threaten biodiversity by displacing other plant species as well as their associated fauna. From a strictly agronomic standpoint, the increased abundance of a weed is also likely to be undesirable.

It may therefore be argued that transgenic pest resistant rapeseed varieties should not be released because potential negative effects due to the spread of pest resistance genes are unacceptable. However, before making such a decision we should also consider potential positive effects of this technology, including the health, environmental, and economic benefits due to a reduction in pesticide use. In order to weigh the positive effects against the negative, a more detailed analysis is required.

Current knowledge of the mechanisms that determine population dynamics of a weedy species is at best sketchy, even for the most extensively studied weeds. We have little quantitative information on the influence of insects, pathogens, and other organisms on the relative abundance of weeds. Data from ecological research, as well as from biocontrol studies, show that reductions in population growth caused by insects and/or pathogens are often dependent on environmental factors (Louda and Rodman 1996). A geographic information system that matches the environment with pest and host species would provide valuable information about the potential impact of increased pest resistance. Introgression of a pest resistance gene has the potential to increase plant fitness only in areas where the pest is present and where environmental conditions are favorable to the pest.

We also need to know how frequently introgression would occur, how quickly such populations would spread, and the fate of the gene within the population. The frequency of hybridization events and subsequent introgression may depend on the time of flowering, presence of pollination vectors, and degree of sexual compatibility. Spread at the landscape level will largely depend on seed dispersal mechanisms. Seeds that rely on wind, water, animals, or humans for dispersal may travel long distances and spread their genes much faster than seeds that are only locally dispersed. Introgression and dispersal can easily be simulated, but actual data is needed to parameterize the model and to test underlying hypotheses.

Information on the frequency of hybridization events and on the rate at which pest resistance genes may spread could be obtained with the use of marker genes. Furthermore, such genes could help reduce the risk associated with gene escape by facilitating identification of hybrid individuals. The green fluorescent protein (GFP) gene is a commonly used transgenic marker that could easily be inserted along with a pest resistance gene (Stewart 1996).

A complete record of wild species that have the potential to hybridize with the rapeseed crops is needed. We should also be concerned with compatible species outside the US because transgenic varieties will eventually cross national, and therefore regulatory, boundaries. Furthermore, we need some measure of the likelihood of specific hybridization events and the level of fertility of subsequent generations.

Questions also remain on how transgenes are expressed when moved from one species to another. Current evidence shows that transgenes are expressed similarly in hybrids and parent plants (Chèvre et al. 1997; Stewart et al. in press). It is, however, conceivable that situations could exist in which a pest resistance gene may not function as expected when transferred to a different host. Unexpected behavior must be taken into consideration in order to evaluate the risks properly.

Filling these knowledge gaps should allow us to build simulation models that can quantify the impact of introgression of pest resistance traits on fitness and abundance of sexually compatible species. Based on this information it would be possible to make a more accurate risk assessment that would consider costs as well as benefits of a proposed technology.

Because of the above knowledge gaps, we recommend support for the following kinds of research projects: