Sir

Introgression from genetically modified crops to wild populations: getting the details right

'Transgene introgression from genetically modified crops to their wild relatives' by Stewart et al. in the October 2003 issue of Nature Reviews Genetics deals with an important and timely topic1. Regrettably, the article has a substantial number of inaccuracies, especially in the first half. To set the record straight, I provide a representative sample of such misstatements below. The Correspondence word limit prevents me from exhaustively correcting the whole article.

“Introgression between species is of more concern than introgression between subspecies because of the possibility of creating more aggressive transgenic weedy or invasive species” is false. Increased weediness can evolve from hybridization with close wild relatives just as easily as from hybridization with distant ones2. In fact, the very worst new weed that results from hybridization between a crop and a wild relative is weed beet, a hybrid descendant of two subspecies of Beta vulgaris (sugar beet and sea beet), which has resulted in hundreds of millions of dollars of damage to Europe's sugar industry (see Refs 3,4).

“Introgression from a crop to its wild relative is generally regarded as more difficult than from the wild relative to the crop”. The opposite is generally regarded as more difficult, at least by breeders (see, for example, Ref. 5). This observation makes sense because part of the domestication process would involve selection of crop varieties that do not bear progeny sired by wild plants that bear the undesirable traits of their fathers.

“There is no molecular evidence of crop-to-wild introgression in soybean, ... pearl millet, ... [and] common bean” is false. There is substantial molecular evidence of introgression in all four species (see Refs 3,6).

The backcrossing of F1s to the wild parent is not a prerequisite for crop genes to enter wild populations. For example, the intermating of F1s to create F2s that subsequently mate among themselves or backcross to the wild parent is one of many other pathways for crop genes to end up in natural populations3.

I do not want to give the impression that the article does not contain some good information. Furthermore, I do not disagree with many of its conclusions, such as the need for case-by-case analysis or the need to proceed with caution. However, taken collectively, most of the article's inaccuracies tend in the direction of more optimism than is warranted — giving the impression that crop gene flow to wild relatives occurs for only a tiny subset of crops and, when it does occur, it has little consequence. The fact is that most crops mate with wild relatives somewhere in the world, and that hybridization — and subsequent introgression — has, on occasion, created problems in terms of the evolution of increased weediness in some species3. Those new weeds have resulted in hundreds of millions of dollars of economic damage. Similarly, such crop-to-wild hybridization and introgression has occasionally resulted in an increased extinction risk for crop relatives that are already rare3. In at least one case, such hybridization has nudged a taxon to extinction in the wild. There is no reason to assume that engineered crops should pose any less risk.