Transgenes have the potential to improve crop nutrition, increase yield, reduce inputs of chemical herbicides and insecticides, and lower fuel use by farmers (Baenziger et al. 2006; Brookes & Barfoot 2006). These benefits are not trivial; for example, the use of herbicide resistant genetically modified (GM) canola has reduced herbicide applications by an estimated 22% and increased farm income by ca. 10% in North America over the last 10 years (Brookes & Barfoot 2006). Nevertheless, transgenes introduce novel traits with unknown consequences for both the consumer and the environment (Gillett et al. 1986). Indeed, transgenic crops are employed precisely because they possess traits that do not exist in the gene pools of crops and their close relatives.
One of the concerns about the use of transgenes noted most often is the potential for transgenes to spread from crops into natural populations. We know that many plant species can be promiscuous—pollen shed on the wind may indiscriminately fertilize eggs of close relatives (Arnold 1992). For many years, this has led to efforts to minimize the inadvertent contamination of crops by pollen or seeds of wild relatives (or other cultivars) with less desirable traits. Concerns about gene flow in the other direction, from crops into the wild, have escalated with the widespread deployment of genetically modified organisms (GMOs) (Ellstrand et al. 1999). Many GM traits are likely to be favoured in natural populations, where they may alter the abundance and distribution of crop relatives (Pilson & Prendeville 2004). For example, herbicide resistance would allow weeds to colonize transgenic fields (defeating the purpose of the transgene), and both disease and insect resistance might allow weeds to escape from their natural enemies, sparking invasion of both agricultural and natural areas (Keane & Crawley 2002; Pilson & Prendeville 2004). Risk management in the GM era is now focused on eliminating this gene flow. This task is far more challenging because the ultimate aim is to prevent (not minimize) transgene escape into the wild.
Warwick et al. (2003) had previously documented hybridization between transgenic canola, Brassica napus, and its wild relative B. rapa at two study sites in Quebec, Canada. At one site, some plants produced in excess of 50% hybrid progeny. In their latest study (Warwick et al. 2008), they demonstrate that hybrid lineages generated in 2000/2001 persisted for the 3–5 years of study at each site. They assessed evidence for hybridization and introgression using species-specific molecular markers [amplified fragment length polymorphism markers (AFLPs)], a transgene marker (herbicide resistance), and ploidy (B. napus is tetraploid, while B. rapa is diploid). They found that, while hybrid lineages declined dramatically over time, B. napus markers persisted in weedy B. rapa. One individual in 2005 displayed herbicide resistance in a B. rapa diploid genetic background. Although it suffered from reduced pollen fertility, this individual was still able to produce hundreds of offspring, most of which showed high pollen fertility and half of which showed herbicide resistance. Importantly, this transgene persisted in the weedy B. rapa populations despite a lack of selection by the appropriate herbicide after 2002.
Two key points emerge from these results: (i) introgression occurs despite genetic barriers; and (ii) transgenes can persist in the environment without the aid of strong selection. Matings between diploids and tetraploids typically produce triploids with low fertility (e.g. Burton & Husband 2000), and indeed this is the case in B. rapa × B. napus (Warwick et al. 2003). This barrier is likely to be critical for maintaining species boundaries. However, low fertility is not equivalent to sterility, permitting the type of introgression seen in this study. Crops and their weedy relatives frequently co-occur over North America's vast agricultural areas, creating ample opportunities for these rare events (Ellstrand et al. 1999). In this case, weedy B. rapa is relatively uncommon in eastern North America, but it does occur in B. napus growing areas (Simard et al. 2006). If B. rapa occurs at low densities near canola fields this may in fact enhance rates of gene flow from the crop; because B. rapa is self-incompatible, the absence of compatible conspecific mates may favour successful pollination by B. napus. Given that partially fertile hybrids are produced, the likelihood of persistence of individual traits will then depend on their genetic architecture and fitness effects. Simple genetic constructs such as the herbicide resistance transgene in this canola, which appears to carry no inherent fitness cost in the genetic background of B. rapa (Warwick et al. 2008), may persist but are unlikely to become widespread following single bouts of gene flow. However, if hybridization occurs repeatedly across many populations or the trait is selectively favoured (e.g. through the use of the herbicide), its spread is virtually assured.
For transgenic crop production, such news is sobering. Evidence is mounting that crop × weed hybridization is the rule rather than the exception for some crops and that transgenes regularly escape (Ellstrand et al. 1999; Pilson & Prendeville 2004; Andow & Zwahlen 2006). Transgenes that confer a direct fitness cost to weedy species, or that can be tightly linked to such a disadvantageous gene, might hold the most promise for ‘safe’ deployment (Andow & Zwahlen 2006). However, simulation studies suggest that these improved mechanisms will likely only slow transgene escape, and containment is likely to be achieved only if the rate of escape becomes low relative to the expected market lifespan of a cultivar (Haygood et al. 2004). Furthermore, fitness costs change with the genetic and ecological environment, and it will be hard to ensure that even an apparently disadvantageous crop gene will not be favoured or neutral in some weedy populations (Bergelson & Purrington 1996). Indeed, introgression is a key source of variation in the evolutionary history of plants (Arnold 1992); ironically this may prove to be the most natural feature of transgenic crops.