Allergies to soy would be nutty.

Hoping to improve the nutritional value of a protein supplement for hogs and poultry, agricultural scientists spliced a gene from Brazil nuts into soybeans. The gene codes for production of a protein rich in methionine, an amino acid important for livestock growth but produced in only small quantities by soy.

While the research proved "very promising," notes Tim Martin of Pioneer Hi-Bred International, in West Des Moines, Iowa, his company has just killed plans to develop the new soy. The reason is that a study Pioneer financed shows that the transferred protein is one of the major allergens in Brazil nuts. While posing no risk to livestock, this protein could trigger life-threatening reactions in susceptible people if it ever reached the dinner table. "We knew that one of the proteins in Brazil nuts was probably allergenic. But we had no idea which," says Steve L. Taylor, of the University of Nebraska at Lincoln. So he tested the protein Pioneer had used against blood from nine people allergic to Brazil nuts.

The protein, extracted directly from nuts, bound like an allergen to antibodies in the blood from eight of the allergic individuals. When taken from transgenic soy, the same protein bound to antibodies in seven of those eight people, Taylor's team reports in the March 14 New England Journal of Medicine. The group then confirmed the allergenicity of the new soy in skin-prick allergy tests on three volunteers with Brazil nut allergies. Ironically, while this study was underway, other scientists published animal data suggesting the protein is not a major allergen. Concludes Taylor, "this just emphasizes why we cannot rely exclusively on animal-based tests for these determinations."

Four years ago, the Food and Drug Administration instructed companies that were developing transgenic crops to test their new products for allergenicity if they carry genes from a material, such as nuts, to which many people have allergies. Taylor says his new data now suggest "that the FDA policy was pretty prudent."

That policy is also too limited, perhaps dangerously so, argues Marion Nestle of New York University, in a commentary accompanying the new study. "The real problem," she told Science News, "is that most of the transgenic work being done in food doesn't involve known allergens. It involves substances that haven't been in the food supply before." And, she notes, FDA doesn't require testing the allergenicity of these.

For now, Martin says Pioneer will look at redirecting its genetic engineering program to boost soy's methionine content with proteins from other cereals and grains already in the food supply.