Although improved crop yields can be engineered by genetically modifying plants, there is ecological concern over whether these plants are likely to persist in the wild in the event of dispersal from their cultivated habitat. Here we present the results of a long-term study of the performance of transgenic crops in natural habitats. Four different crops (oilseed rape, potato, maize and sugar beet) were grown in 12 different habitats and monitored over a period of 10 years. In no case were the genetically modified plants found to be more invasive or more persistent than their conventional counterparts.

In the late 1980s, there were three conjectural risks associated with genetically modified (GM) crops: that they would become weeds of agriculture or invasive of natural habitats; that the introduced genes would be transferred by pollen to wild relatives, whose hybrid offspring would then become more weedy or invasive; or that GM plants would be a direct hazard to humans, domestic livestock or beneficial wild organisms, for example by being toxic or allergenic. Our study assesses the grounds for the first two of these fears.

We have shown previously that GM oilseed-rape plants did not perform better or persist for longer than conventional plants over a 3-year period¹. We now compare the results of monitoring conventional and GM lines of four crop species over 10 years^{2, 3} in a field experiment conducted in 12 different habitats (four in each of Cornwall, Sutherland and Berkshire in the United Kingdom) and repeated in the years 1990, 1991 and 1992 in order to determine whether GM crops would be more invasive or more persistent in natural habitats than non-GM crops. We tested all the crop species and GM constructs that were available in 1990: oilseed-rape and maize plants expressing tolerance of the herbicide glufosinate, sugar beet tolerant of glyphosate ('Roundup'), and two types of GM potato expressing either the insecticidal Bt toxin or pea lectin.

The sites were monitored each year to follow the fate of sown individuals, to measure recruitment onto unsown areas nearby, and to determine whether there was any resurgence following natural disturbance in later years. Figure 1 shows the fraction of seeds sown (or tubers planted) that produced mature plants at the end of the first growing season. We found that there were no significant differences in average recruitment between conventional and GM plants for any of the four crops (full details and significance tests are provided in Supplementary Information). None of the crops, GM or conventional, increased in abundance at any of the sites.

Figure 1 The performance of conventional (blue) and transgenic (red) crops in natural habitats.   Full legend   High resolution image and legend (34k)

Population sizes of all crops declined after the first year as a result of increased competition from native perennial plants. In no case did the GM lines persist for significantly longer than their conventional counterparts. For oilseed rape, seedling establishment was significantly lower for GM plants compared with conventional lines in six out of 12 cases, and were not significantly greater in any case. Subsequent survival was significantly lower for GM lines in three out of 12 cases and significantly higher in two cases. For maize, seedling establishment was significantly lower for the GM line in two out of 12 cases, and significantly greater in no case. Subsequent survival was significantly higher for the GM line in one case, but all lines were extinct by the beginning of the second year.

For potato, survival of planted tubers to the end of the first growing season was significantly lower for GM lines in one case out of eight, and significantly higher in one instance. The few cases of increased GM-plant survival that were significant in the short term did not translate into long-term differences in persistence (see Supplementary Information). Survival of perennating potatoes was significantly lower for GM lines in one case out of eight, and never significantly higher.

For sugar beet, the genetic background (inbred or outbred) was much more important than whether the plants were genetically modified; outbred lines outperformed inbred lines in three cases out of four. GM lines showed significantly lower recruitment from seed in two cases out of four, and significantly higher recruitment in no case. Survival of GM sugar-beet seedlings was significantly lower for one transition at one site, but all sugar-beet lines were extinct within two years at all four sites. Survival of sea beet was significantly higher than sugar beet at all four sites, but it too was extinct at all sites by the end of the third year. The survival of sea beet on open ground elsewhere in Silwood Park, where potted plants had stood in 1992, sounds the cautionary note that perennial plants can persist for extended periods in extremely odd places.

These experiments involved GM traits (resistance to herbicides or insects) that were not expected to increase plant fitness in natural habitats. Our results do not mean that other genetic modifications could not increase weediness or invasiveness of crop plants, but they do indicate that arable crops are unlikely to survive for long outside cultivation. The ecological impact of plants with GM traits such as drought tolerance or pest resistance that might be expected to enhance performance under field conditions will need to be assessed experimentally when such plants are developed.