In October and November 2000, Ignacio Chapela and David Quist collected cobs of criollo corn, native varieties of maize grown in the Sierra Norte mountains of Oaxaca, Mexico. The cornfields they sampled, they noted, were “more than 20 kilometers from the main mountain-crossing road.” Chapela was an assistant professor in the Department of Environmental Science, Policy, and Management at the University of California, Berkeley. Quist was his graduate student. They collected the corncobs while working at the Mycological Facility in Oaxaca, described by Berkeley’s public relations office as “a locally run biological laboratory.” Chapela was trained in mycology, the study of fungi. Another press account identified them as advisors for a program that helps indigenous farmers.

Mountainous Oaxaca, along with Chiapas and parts of Guatemala, is recognized as the center of diversity of corn: it was in this southern part of Mexico that teosinte first gave birth to maize thousands of years ago. More different types of corn can be found here than anywhere else on Earth. They are grown in small plots by farmers who “spend a lot of time as breeders,” according to Wayne Parrot, a plant genetics professor at the University of Georgia. The farmers seek to maintain and

improve the criollo varieties, or landraces. As Masa Iwanaga, at the time the director general of the International Maize and Wheat Research Center in Mexico City (Centro Internacional de Mejoramiento de Maiz y Trigo, or CIMMYT), explains, “The landraces that farmers grow today are often somewhat different from those collected in the same communities decades ago, and they are certainly different from those grown centuries ago, precisely because they have continued to evolve under the combined influence of farmers and the environment. Mexico is not a center of diversity for maize simply because many landraces are ‘found’ in Mexico. In reality, those landraces are the products of farmers’ continuing desire to maintain a great deal of diversity in the maize they grow.”

Some of their criollo maize, Chapela and Quist contest, contains the CaMV 35S promoter and a Bt gene from across the border—in spite of the fact that growing genetically modified corn has been banned in Mexico since 1998 expressly to protect the native landraces. In September 2001 Chapela and Quist alerted the Mexican authorities. Government scientists, along with CIMMYT, began their own tests to see if native landraces had been “contaminated” by outlawed plantings. Chapela and Quist published their study in the journal Nature in November 2001. A month before the paper came out, the two authors appeared at a press conference with Mexican government officials. The New York Times trumpeted, “In a finding that has taken researchers by surprise and alarmed environmentalists, the Mexican government has discovered that some of the country’s native corn varieties have been contaminated with genetically engineered DNA.”

Rather than sounding surprised, however, one knowledgeable critic said, “It is probably inevitable that eventually engineered genes will be found in Mexican corn, as gene flow is a normal and natural phenomenon with maize.” Another said, “I’d be shocked if they didn’t find it there.” What puzzled the scientists—and inspired activists—were two additional claims in the Nature paper: that the DNA was “introgressed” into the landrace, and that it was “attached” to different sequences in different samples, even to different sequences in a single sample. What Quist and Chapela meant by these additional claims, translated a reporter for the Scientist magazine, was that the genes

“were behaving in a way never before observed: fragmenting into smaller bits of DNA, hopping along the genome like a skipping stone over placid water and potentially creating a tremendous opportunity for mutations. In other words, they claimed that the engineered genes were out of control.”

The DNA sequence purported to be so out of control was the CaMV 35S promoter from the cauliflower mosaic virus—a sequence whose job is to turn on the gene next to it. “Activists fears,” said a news story in Science, “centered on the promoter sequence…. If the promoter broke off during hybridization, it could conceivably take over other genes, with unknown consequences.” This kind of promoter instability was just what Mae-Wan Ho had predicted in 1999 when she called the use of CaMV 35S a “recipe for disaster.” The Science story quoted Peter Rosset, co-director of Food First, an advocate organization for small-scale farmers, as saying: “The spread of the promoter could prove to be worse than the spread of the genes for herbicide and insect resistance. If true, this would be a red flag that would call into question every other GM crop on the market.”

If true.

Five months later Nature took the highly unusual step of announcing that it should never have published the Quist and Chapela paper.

Quist and Chapela had collected ears from four Oaxaca fields and also obtained some corn from Diconsa, a government agency that distributes subsidized food. They pooled the kernels from each ear, ground them to a fine powder, and extracted DNA from the powder. They tested the DNA using PCR to see if it contained three sequences commonly found in genetically modified corn: the CaMV 35S promoter, the NOS terminator from Agrobacterium tumefasciens, and the Cry1Ab gene from Bacillus thuringiensis. To see the results of the PCR test, the fragments of DNA were sorted by gel electrophoresis. The scientists placed the DNA at one end of a tray of gel and briefly passed an electric current through it. The fragments, pulled toward the positive electrode, sorted themselves by size, because the smaller pieces could travel faster through the gel than the larger ones. The result was a distinctive pattern of bands. Quist and Chapela reported that their tests had detected a short 200-base-pair fragment of the CaMV promoter in

several of the Mexican maize samples they tested. On the gels, they had seen a weak PCR band of the right size in four samples from farmers’ fields and a strong band in the one obtained from the government. They said they also detected the NOS terminator sequence in two samples and the Bt gene in one.

The PCR bands were weak, they explained, because the transferred gene, or transgene, was present in just a few of the kernels. To a geneticist reading their Nature paper, that explanation sounds peculiar. The paper’s title was, “Transgenic DNA introgressed into traditional maize landraces in Oaxaca, Mexico.” The observation that the transgene was present in only a few kernels doesn’t fit with the assertion in the title that the DNA had introgressed. Introgression means that a gene (in this case the CaMV promoter sequence) has not only been introduced into the plant by crossbreeding a landrace with a genetically modified variety, but that the hybrid plants were then repeatedly backcrossed to the landrace. By using the term introgress, Chapela and Quist were claiming that the CaMV promoter sequence was still there, but that the nearby corn genes had been progressively eliminated through the normal crossing-over step that takes place when chromosomes divide. But an introgressed DNA sequence will not be present in only a few of the kernels on a cob; it will appear in all of them.

What other explanations might there be for the weak bands? One possibility is that the DNA samples used to run the PCR tests were contaminated with a bit of CaMV gene. In order to find any bands at all, Quist and Chapela had to use a technique called nested PCR, which entails performing consecutive PCR reactions to detect very small quantities of the target DNA. “This is a particularly risky approach,” said the editorial board of the journal Transgenic Research, “since extremely low levels of contamination introduced during the handling of samples can be the cause of a positive result.”

And Quist and Chapela did, in fact, handle samples that contained the CaMV gene. They used two types of controls to establish the validity of their experiments, to prove that what they saw wouldn’t just appear in any randomly chosen corncobs tested using this system. One type, the negative controls, were samples of blue corn from Peru and corn collected by CIMMYT from the same Oaxaca area in 1971. These

should—and did—give a negative reading: no sign of the CaMV promoter band. The second set of controls were samples of Monsanto’s Yieldgard Bt corn and its Roundup Ready corn. These positive controls should and did give very strong bands in the PCR experiment: a positive reading. Yet the presence of Monsanto’s transgenic corn in the same experiment means that there was a chance of contaminating the landrace samples with a minute amount of Monsanto DNA. These contaminated samples would produce a positive reading in the test, but it would be a false positive: a mistake. As several scientists pointed out in letters to Nature and, less politely, in the press and over the Internet, Quist and Chapela failed to run the standard follow-up tests to eliminate such false positives.

Quist and Chapela’s PCR results are not, however, necessarily wrong. There is still a third possible explanation for the weak bands they saw on the gels—an explanation that does not require introgression or several seasons of backcrossing and yet is not a false positive. The plants from which the cobs were collected could be ordinary landrace plants onto whose silks a few pollen grains from a genetically modified plant growing nearby had landed that same season. Each pollen grain would have germinated on the cornsilk, delivering its sperm cells to fertilize a single egg, which would then develop into the embryo of the kernel. If planted, that kernel would have grown into a new hybrid variety, a cross between a landrace and a transgenic variety. But Quist and Chapela did not plant any of their corn to see if this cross had, in fact, happened. They couldn’t because they had ground all their kernels into powder.

Again it was the lack of a double check—not the results themselves—that irritated many scientists. Double-checking would have taken time and patience. More corn would have had to be harvested and planted out the next year. If the gene were present in just a few of the kernels, Quist and Chapela would have had to test many—perhaps hundreds—of the plants grown from such kernels to identify the few plants that were hybrids (if their appearance didn’t give them away, which it likely would have). The editors of Transgenic Research complained, “Most frustrating is the total failure of the authors to do the easy and incontrovertible experiment of growing out the suspected

contaminated lines. Hybrids between Mexican landraces and transgenic commercial maize would be very obvious.”

And yet despite the New York Times’ assertion that researchers had been surprised, Quist and Chapela’s critics overwhelmingly agreed that Mexican landraces probably do contain traces of transgenes. Mexican farmers often plant corn kernels—increasingly shipped in from the United States, where genetically modified corn is widely grown—meant as food, not seed. What puzzled the critics most was the claim that the CaMV promoter popped into various places in the genome, that the transgene was “out of control.” The New York Times quoted a fellow scientist: “If real, that would have been a huge finding.”

It was this claim that precipitated Nature’s announcement that it should not have published the paper. Having received detailed critical letters from reputable scientists, Nature asked Chapela and Quist to produce more data in support of their claim. The evidence they supplied, Nature’s editor wrote in April 2002, “is not sufficent to justify the original paper.”

Chapela and Quist’s published data were obtained using a technique called inverse PCR, one that a reporter for the Scientist identified as “not only complex, but also cutting edge, so even a Ph.D. in another field—no less a layman—would have difficulty in making a reasoned judgment about who is right and who is wrong.” In inverse PCR, the primers are inverted. Rather than framing and amplifying the CaMV promoter, they look for what is attached to its two ends. The DNA sample (in this instance from the ground-up corn kernels) would first need to be cut into small bits. As their molecular scissors, scientists use restriction enzymes. When cut fragments are joined back together in a very dilute solution they tend to make circles because the closest matching end to reconnect to is the tail end of the same fragment. When the outward-pointing primers are used to run a PCR test on this solution of circular bits of DNA, what is amplified are the sequences that flank the CaMV promoter sequence they originally detected. These can then be sequenced and identified.

But there are two common problems with the inverse PCR technique. First, the circle-making step is far from perfect, even when the right restriction enzyme is used (and, according to their critics, Quist

and Chapela picked the wrong one). No matter how it is cut, the fragment of interest might not find its tail. If it doesn’t, there’s no telling what other fragment, from which part of the corn genome, it might attach itself to, for the ligase—the molecular glue—is added to a test tube in which thousands of fragments of corn DNA of all different sizes are mixed together. In a solution molecules move. Where a fragment finds itself in the solution bears no relationship to where it was in a corn chromosome. Two pieces that are glued together in solution might have been nowhere near each other originally. This new, double piece might itself form into a circle, making it difficult to figure out that the two fragments were not originally together in the chromosome.

The second problem is that it’s quite easy, during the amplification step, to make many copies of pieces that just happen to have some similarity—but are not a perfect match—to the primers. These create what PCR practitioners call artifactual (or bogus) bands. There are ways of making sure a band is not bogus. In Quist and Chapela’s case, the piece of DNA on which they based their primer was a 200-base-pair-long fragment of the CaMV promoter. This piece was the one they had detected as weak bands in the first PCR tests on their original corn samples. The whole CaMV promoter is much longer than 200 base pairs. In Bt corn it is more than 1,000 base pairs long. So when Quist and Chapela ran inverse PCR, looking for the bits of DNA just next to the ends of that 200-base-pair fragment, the first thing they should have found was more of the CaMV promoter sequence.

The scientists who published a rebuttal of Quist and Chapela’s work simply looked at the authors’ own data and asked the question the authors should have asked: are there additional CaMV and other transgene sequences adjacent to the 200-base-pair sequence? There were not. The critics found no similarity to the CaMV promoter except for that very short primer sequence. This result means that the fragments Quist and Chapela amplified in their inverse-PCR experiment were experimental artifacts: mistakes. Quist and Chapela should themselves have suspected this. The critics’ analysis was carried out on the authors’ own sequence data.

Quist and Chapela claimed that the CaMV promoter broke up into

little pieces and got stuck into various places in the corn genome. To

people who have created and studied genetically modified plants, this suggestion was quite at odds with experience. However, it would also have been easy to verify. Going back to their original criollo corn DNA samples, Chapela and Quist needed only to use standard PCR with new primers designed to pick up the sequences that the CaMV fragments had supposedly attached themselves to. They did not do this simple experiment. Nature’s reviewers should have required this additional confirmation prior to publication. They did not. If the review process had been a bit tougher, it might have saved Nature some embarrassment. As for the transgene being “out of control,” the New York Times reported in April 2002, “Dr. Chapela acknowledged technical problems and said he and Mr. Quist were ‘backing off a bit’”—as close as they have come to admitting they were wrong.

But Quist and Chapela stood by their conclusion that criollo corn from Oaxaca, in maize’s center of diversity, contained transgenic DNA. Studies by the Mexican government, although not yet published, have also detected transgenes in landraces, as reported at a news conference in Mexico City in February 2002. Yet far from being surprised or alarmed, as the New York Times’s original report had it, the researchers and environmentalists associated with CIMMYT have been working since 1995—when the first genetically modified maize reached the market—to understand and contain, if necessary, the effects of these new varieties on Mexican corn. In 1995 CIMMYT held an international workshop called “Gene flow among maize landraces, improved maize varieties, and teosinte: implications for transgenic maize.” The proceedings were published in 1997 and are available on CIMMYT’s website. CIMMYT has also begun sociological studies to learn how the region’s small farmers select seed “and thus influence how genes (including transgenes) flow into and between landraces,” said Iwanaga in February 2002.

“In Mexico there is a moratorium on planting transgenic maize,” he said. But quite a bit of the corn Mexico imports as food comes from the United States. Since, according to the USDA, a third of U.S. cornfields in 2002 grew transgenic, or genetically modified, corn, Iwanaga said, “It is quite possible that some of the maize imported into Mexico was transgenic.” It is also quite possible that some of it is growing in

Mexico today. Mexican farmers are dedicated corn breeders. One could easily have bought U.S. corn, Iwanaga said, “and, instead of eating it, planted it, just to see what might happen.”

Gene flow is popularly considered a hazard of genetically modified food plants. Said Klaus Ammann, curator of the Botanical Garden at the University of Bern in Switzerland, “The debate on genetic engineering ‘forces’ us to focus in an unfortunate way on gene flow as a basically negative effect, as if pollen would have learned to fly with the transgenes.” But gene flow, he said, “has always occurred between different old landraces and between different new varieties of crops. Despite this, varieties of apples or cereals have been stable over many years and specific traits have not disappeared. Pollen has always flown.”

Plant genes flow—that is, they move—primarily because pollen is carried by the wind and by insects, sometimes over long distances, to pollinate flowers. Each pollen grain carries two sperm cells, each having a set of the parent plant’s chromosomes, ready to fertilize the female. At the heart of each female flower is an embryo sac. The sac consists of a very few cells, all of which are genetically identical. Like the sperm cells, they are also haploid, that is, they contain only one of the two copies of each chromosome in the other cells of the plant. The most important cells in the embryo sac are the egg and the genetically identical “central cell.” The central cell gives rise to the endosperm, which nourishes the plant as it germinates and grows until it can produce its own food through photosynthesis. The endosperm, with its stored starch, proteins, and fats, is the primary food of humanity, as well. It’s in tortillas and tacos, cornpuffs and cornflakes. Wheat endosperm makes flour, bread, and pasta—only wheat germ is the embryos themselves. The animals we raise for meat and a growing fraction of farm-raised fish are nourished by endosperm in their grain- and corn-based feeds.

Pollen—the male contribution to flower sex and what makes people sneeze in spring—is made in huge quantities. A single corn plant makes 18 to 25 million pollen grains. Multiplied by 25,000 or

more, for the number of plants in an acre of corn, then by the number of acres in the field, and the amount of pollen released in the short two weeks that corn pollen sheds is truly staggering.

A pollen grain is as small as a grain of sand. Most are round. Some plants make smooth pollen grains, some make patterned ones, and some make spiky pollen. A pollen grain is a vehicle for getting genes from plant to plant. Its surface is designed to resist the drying wind, but to catch and hold tight to the tiny hairs on the female part of the plant. The pollen grain is itself a cell, powered by its own nucleus, called the vegetative nucleus. Inside the pollen grain are two twin sperm cells, each containing the same haploid complement of parental genes. One sperm cell will fertilize the egg. The other will enter the central cell. This process is called double fertilization. It means that both the fertilized egg (the zygote), which will develop into the new plant, and the central cell, which will develop into the endosperm, belong to the next generation.

The landing platform for pollen within the flower is called the style. The style can be a fuzzy central part of a flower or it can be the long silk of the corn plant. Styles are covered with small hair-like projections that trap the pollen carried on the wind and by insects. Pollen is generally quite heavy. Most of it lands within a few yards of the plant that produced it, although some is carried much farther. It lands on everything. Most of it dries up and gets washed off by the rain. Park your car under a tree in spring and it will be coated yellow with DNA-laden pollen cells, dessicated and impotent. Most pollen hits a dead end. Only if the pollen lands in the right place—on the style of a female flower—and only if the chemistry is right, will the pollen have a chance of transmitting its DNA to the next generation.

The chemistry depends on the style. Corn pollen will germinate only on corn silks. Corn pollen that lands on a daisy or on a milkweed plant has the wrong chemistry. The two plants are sexually incompatible. Nothing happens. The pollen dries up and dies. But when a corn silk traps corn pollen—and each long silk traps many pollen grains—then the chemistry is just right. The pollen grain swells and germinates, sending out a pollen tube that grows faster than any other known plant cell—as much as a foot in just a few hours—carrying along the

two sperm cells at its tip. The many grains that fall on a single corn silk begin a race down the silk to be the first to reach the egg. Guided by chemical signals inside the silk, the pollen tubes grow in parallel tracks down the silk, aiming steadily for the embryo sac nestled in a little bump on the tiny, unfertilized ear. The pollen grain whose pollen tube grows fastest wins the race, delivering its two sperm cells to the waiting egg and central cell. The next generation begins with the fusion of egg and sperm. The genes of the egg are combined with the genes of the sperm. This familiar process of plant reproduction through pollination is now referred to as “gene flow.”

The public’s concern about gene flow from genetically modified crops is that the CaMV promoter or a Bt toxin gene or an herbicide-resistance gene might migrate by way of the pollen into wild plants or—in the case of maize—into the landraces. The worry is that these new genes would flow into—newspapers tend to use the words “invade,” “contaminate,” “pollute,” or “taint”—the native corn’s genome and, as Chapela said, crowd out landraces that do not carry it.

Yet corn genes have been flowing into—and out of—Mexican landraces for many centuries. Said Masa Iwanaga of CIMMYT, “When transgenes are present in Mexican maize landraces grown by farmers, does this mean that an important resource is lost forever? As scientists, we would answer ‘no,’ because the landraces may have changed, as they do all the time, but they have not disappeared. On the contrary, with the addition of a transgene, they could actually be considered more diverse. This additional diversity may not be desirable, however. It is precisely this issue that the Mexican government must resolve.”

Several months before Quist and Chapela began collecting corncobs in Oaxaca, Juan Pablo Martinez-Soriano of the National Polytechnic Institute in Irapuato, Mexico, wrote a letter to Science in which he reminded its readers that corn, unlike its ancestor teosinte, cannot reproduce without human help. Its kernels are tightly fixed to the cob and “viable seeds can only be released by mechanical means (basically by humans). Maize does not disperse itself and therefore does not exist as a free species in nature.”

He continued, “Arguments stating that maize is genetically fragile are weak. It seems paradoxical to argue that it is necessary to protect

the genetic background of corn when, for 6,000 years of traditional breeding, we have protected only alleles important for humankind. Even if we decide to protect the actual genotypes, there should be no reason for concern. Any transgene transferred inadvertently to native maizes can be removed from the progeny by selecting against the incorporated trait. Maize is always under strong artificial selection, and therefore natural selection has no meaning for the species.”

Like corn, all of our food plants originated from wild plants—weeds—with natural means of dispersing their seeds. Some crops and their weedy relatives grow in the same places, some don’t. No weedy relatives of corn grow in the United States, Canada, or Europe. But teosinte, the ancestor of maize, grows in Mexico and Central and South America. Teosinte can still make fertile hybrids with maize. When maize and teosinte cross, the next generation is a hodge-podge of traits from the two parents: some plants look more like maize and some look more like teosinte. None is as good a food producer as the carefully bred corn, whether commercial or criollo. But the most important point is that those hybrid plants that receive the maize genes that prevent seed dispersal are sure to die out without humans to pick them, plant them, and tend them.

Long known as cross-hybridization, this problem was only recently dubbed “gene flow.” It is not just a problem of crops with weedy relatives. It is an ever-present problem for plant breeders and seed producers. It is most acute in a crop, like corn or canola, that is not a selfer—that does not use its own pollen to pollinate itself—and in crops whose edible parts are produced after pollination, like corn kernels and canola seed. To produce seed for white corn, for example, a farmer has to be sure any neighbors (within pollen-flying range) are not growing yellow corn. The yellow trait is dominant. If white corn and yellow corn are grown side by side, the ears of the white corn will have a mixture of yellow and white kernels. The resulting cross is not always bad. An attractive bicolor variety of sweet corn is sold under several names, including Butter n’ Sugar. Yet to keep even that bicolor variety breeding true, it needs to be kept safe from yellow corn pollen.

The same is true for any kind of hybrid corn. Said retired Pioneer Hi-Bred corn breeder Don Duvick, “Of course the problem of unwanted pollination has been around ever since hybrid seed corn pro-

duction started in the 1930s,” he said. “The isolation standards that were set up in those days recognized the fact that zero ‘contamination’ was a biological impossibility.” Corn breeders learned, by trial and error, how far apart two varieties needed to be planted so that crossing would be kept to a minimum.

In order to develop—and sell—pedigreed seeds, breeders had to devise protocols that account for the patterns of pollen dispersal among plants that can crossbreed. Even backyard gardeners had to keep it in mind—and they still do. Anyone hoping to grow the 2000 All-America Selections Winner sweet corn variety named Indian Summer, with its festive mix of yellow, white, red, and purple kernels, needs to be sure to read the fine print: “Requires isolation from other corn pollen.”

Whether gene flow between crop plants and their wild relatives is a problem depends on several things: the crop plant, where it is grown, and how it is used; whether or not the crop has weedy relatives nearby with which it can crossbreed; and how those weeds are managed. This long list of issues cannot be lumped—as people have tried to do—into one category. Whether gene flow is a problem also depends on your perspective. Whether you view transgenic corn pollen as a source of genetic pollution or as a source of new genes that could make the corn you’re growing more resistant to insects might well depend on whether you’re an organic farmer or a subsistence farmer. An environmental activist might see the transfer of genes from crops to wild species as an ecological threat, while a farmer perceives it as a problem in weed management.

What is a weed, after all? One view calls a weed any plant that is growing in the wrong place. It is a plant that is causing harm or being of no benefit. But, notes Klaus Ammann of the Botanical Garden in Bern, it is hard even for botanists to know what plants to list as weeds. “One and the same species may be considered in some parts of its area as a harmless component of natural vegetation, in others as a weed, and again in others even as a useful plant species.” Another botanist notes that 17 of the 18 “World’s Worst Weeds” are also cultivated. In the right farmer’s field, they are not weeds but a crop.

There are technical definitions of weediness, as well. Weeds are plants with characteristics that allow them to persist and spread with-

out human intervention. Weeds have one or more of the following: easily shattered seed pods; seeds that are long-lived, tough, and have structures such as hooks that cause animals to disperse them; toxic chemicals or physical means, such as thorns, of repelling insects and herbivores; or more than one mechanism for reproducing and spreading, such as both seeds and rhizomes. Not surprisingly, crop plants that have been cultivated by people for food over millennia are generally not weedy. They can’t readily become weedy because they have lost several weedy characteristics. They are often wholly dependent on humans for survival and propagation. They are, in a very real sense, wards of people. We might worry about milkweed invading the cornfield, but we don’t worry about corn taking over the milkweed patch. The lone corn plants towering over the soybean field, called volunteers, present certain weed-control problems, but they’re not invasive weeds like Johnson grass, a sorghum relative.

Many crop plants are not native to the places where they are cultivated and their wild relatives are not found in the same geographical areas. The genes of these crops can’t escape from the field through pollen, because there are no plants growing nearby that can be fertilized by that pollen. Because there are no wild relatives of corn in the United States, the transgenes the genetically modified varieties contain cannot flow. But that is not true in Mexico, where corn’s wild relative teosinte grows. When a crop’s weedy relatives live in the same geographical area, often in and near farmers’ fields, then the crop’s genes can move into the weedy species. Again, whether such gene flow is a problem depends on your occupation and your point of view—as well as on the plant and the trait in question.

Canola, for instance, is defined as the low-erucic-acid variety of oilseed rape, Brassica napus. Canola oil is a health food. Oil of ordinary Brassica napus is good for lubricating steam engines, but the erucic acid it contains has been linked to heart disease in people. When he was developing the variety that would become canola, Keith Downey of the Agriculture Canada Research Station in Saskatoon needed to be sure that other brassicas didn’t pollinate his breeding plots, introducing the genes that promote erucic acid synthesis. He studied the rate of gene transfer between various types of brassicas, both cultivated and weedy. He found that if a plot of weedy brassica grew within 1,000 feet

of a canola field, more than 3 percent of the canola would be fertilized by weedy pollen. The canola pollen itself was not as successful. With as little as 140 feet between plots, only 2 percent of the weeds were fertilized by canola pollen. How far the pollen could fly was not the only factor that had to be taken into account. Was the flower fertilized? Did it produce seed? Did the resulting hybrid grow to reproductive age? Was it fertile? In the late 1980s Downey did a series of experiments in greenhouses and in the field to answer these questions, as well as to see how easily canola crossbred with its weedy relatives. The various brassicas, whether cultivated or weed, he found, easily crossbred.

Downey discussed his experiments in 1990 at the Workshop on Safeguards for Planned Introduction of Transgenic Oilseed Crucifers held at Cornell University by the USDA. Twenty-three panelists from the United States, the United Kingdom, Canada, Belgium, and India were joined by 70 observers, including representatives from France, Japan, and Thailand. The participants represented universities, industry, the government, and public interest groups. In addition to ecologists and plant breeders, the speakers included experts on herbicides and scientists studying the behavior of honeybees.

The workshop took place three years before the first transgenic crop was approved for market. The participants at the workshop agreed that the question to answer was not “whether or not genes from transgenic species would move out” (they would), but rather when, where, and how it would happen and what would be the consequences.

Because genes that confer resistance to herbicides were among the first to be introduced into crops by molecular techniques, herbicide resistance was one of the first problems associated with gene flow to be brought to the attention of the public. But it isn’t a new problem. Herbicide-resistant varieties of canola were already being grown in the 1970s when Downey began his experiments crossbreeding various brassicas. In fact, he used herbicide resistance as his marker to detect crop-weed hybrids. Some of these varieties had gained their resistance to an herbicide through spontaneous or natural mutations; other mutations were induced by conventional plant-breeding methods using chemicals or radiation. Breeders developed them so that canola farmers could have a way to kill the wild brassicas in their canola fields.

When a new herbicide-resistant canola variety was introduced

into Australia in 2000, a study was done to measure gene flow on a very large scale and under real commercial conditions. The variety, which was not genetically engineered, was resistant to the herbicide chlorsulfuron, sold by DuPont as Glean, that inhibits the enzyme acetolactate synthase. The fields surveyed ranged from 25 to 100 hectares in size and were up to several kilometers away from the field planted with the herbicide-resistant variety. The investigators—reproductive ecologist Mary Rieger of the Cooperative Research Center for Australian Weed Management and the University of Adelaide and her colleagues at the universities of Adelaide and of Western Australia—grew the seeds they collected in fields planted with canola that was not herbicide-resistant and asked how many grew into plants that were tolerant of the herbicide. Herbicide-resistant plants were detected in about two-thirds of the tested fields, but their numbers were tiny, averaging 0.03 percent, or just three seeds in 10,000 tested. The herbicide-resistance trait could occasionally be detected at a considerable distance—more than a kilometer—from the field of origin, probably the result of insectborne pollen.

Several varieties of herbicide-resistant canola are grown in Canada and the United States. Some are resistant to glyphosate (Roundup), others to glufosinate (sold as Rely, Challenge, Finale, or Basta), still others to compounds of the imidazolinone family (Patriot, Lightning, On Duty, and others). Some are genetically modified in the contemporary sense that genes from other plants or from bacteria have been introduced into them by molecular techniques. Others were modified through mutation breeding.

When varieties that are resistant to different herbicides are grown in adjacent fields, they cross-hybridize. Some of their offspring are resistant to both herbicides. This process can continue for generations, leading to what has come to be called “gene stacking.” Stacking of herbicide-resistant genes is not a problem for the crop—after all, it’s the crop plant that is supposed to survive the herbicide spray. Canola plants have been identified in Canada that can survive being sprayed with three different herbicides. The field in which they were detected had been planted with glufosinate-resistant and imidazolinone-resistant canola in 1997. It was next to a field in which glyphosate-resistant

canola was grown. In 1998 plants were detected that had resistance to all three herbicides.

Gene stacking becomes a weed-management problem when the herbicide-tolerance genes accumulate in plants the farmers wish to kill. Because herbicide-tolerant canola was grown in Canada well before the genetically modified varieties were released, farmers already have had to manage herbicide-resistant volunteers that grow as weeds in crop rotations. The most successful strategies employ a long crop rotation cycle, together with plowing and spraying of preemergent herbicides.

But weeds related to canola grow in many of the areas where canola is cultivated. In western Canada, these include Sinapis arvensis (wild mustard), Raphanus raphanistrum (wild radish), and Erucastrum gallicum (dog mustard). There have been no reports from field or greenhouse crosses that these weeds have picked up herbicide-resistant genes from canola. The fact that such gene flow to weeds hasn’t been detected might be due to the genetic distance between the crop and its relatives. Experimental crosses between these weedy species and canola produced plants that were sterile or didn’t grow well.

In the U.K. as well, despite considerable opportunities for exchanging pollen, the cultivation of oilseed rape has not created hardier weeds. Studies there recently assessed the extent to which oilseed rape (Brassica napus) and two wild brassicas, B. rapa (wild turnip) and B. oleracea (wild cabbage) form hybrids in the wild. None of the wild cabbage plants sampled near oilseed rape fields turned out to be hybrids. Hybrids were detected in wild turnip populations. In the first study 1 hybrid was detected among the 505 plants sampled; in the second study 47 hybrids were detected among the 3,230 plants sampled.

The low percentage of hybrids detected shows that the B. napus-B. rapa hybrids that form every year that oilseed rape is grown have not come to dominate the wild turnip populations. Nor will the introduction of an herbicide-resistant variety of rape affect the wild turnip population. The herbicide-resistance gene is of little use to plants that are not treated with herbicides. Nonetheless, wild turnips are weeds in some rape fields in parts of the U.K. In these fields hybrid weeds will arise. If the hybrids produce seeds, the herbicide-resistant plants that

grow from these seeds will be herbicide-resistant volunteers. They will need to be managed by rotating crops and by using different herbicides.

Herbicide use selects for herbicide-resistant weeds just as surely as antibiotic use in humans selects for antibiotic-resistant bacteria. Bob Scott and Chris Tingle are a weed specialist and a soybean agronomist, respectively, for the University of Arkansas’s Extension service. They note that 2.5 million acres of soybeans grown in Arkansas contain the gene for resistance to the herbicide glyphosate. “The reason for the large number of acres in Arkansas is the genuine need for the technology,” they write. Many weeds commonly found in the state’s soybean fields, including Palmer amaranth, common cocklebur, pigweeds, and goosegrass, have developed resistance to one herbicide or another. “Couple that with a long growing season, plenty of rain or irrigation, crop rotational issues, and a large soil bank of weed seeds, and you have a perfect fit for using Roundup Ready soybeans and glyphosate.” It is “only a matter of time,” they acknowledge, before resistance to Roundup shows up in the weed population as well. When it does, they say, “you need new technology.”

Three months before Nature abandoned the Quist and Chapela paper, its sister journal Nature Biotechnology summarized the Mexican maize scandal. Rather than zeroing in on the technical difficulties, the journal stated that “the major point of divergence in the current discussion” about gene flow was not whether a transgene from a genetically modified variety had moved into the Mexican landraces, but how the so-called contamination would affect biodiversity. The journal wrote: “The Berkeley researchers have claimed that appearance of DNA from GM crops into criollo maize compromises biodiversity. ‘If the transgene makes the carrier any more fit,’ says Chapela, ‘you would expect to see the crowding out of landraces that do not carry the trait.’” A few paragraphs later, the journal gave the opinion of Val Giddings, who is a spokesperson for the industry group called BIO, headquartered in Washington, D.C. Giddings was quoted as saying, “We know what threatens biodiversity, and it is not the substitution of one variety

for another in an agricultural field. It is the conversion of native and wild land to agriculture in the first place.”

The journal failed to point out that the two sides in this debate were talking at cross purposes. Although each was responding to a question about biodiversity, they defined that term—as their answers make clear—in rather different ways. Chapela is concerned about the diversity of traditional varieties of corn, about the number of genetically different maize landraces planted by farmers in Oaxaca, Mexico. If a genetically modified Bt corn, for example, proved to give a higher yield, it might become more popular than the landraces, and farmers might stop planting these old-fashioned varieties. Whether this change from old to new varieties is bad or good might well depend on your point of view. If the Bt gene were truly introgressed into a landrace, as Quist and Chapela originally suggested, then a landrace would still have all its original genes and it would still grow and produce corn as did the original landrace—but it would give a higher yield because of its new gene for insect resistance.

Yet Chapela’s fear—that genetically modified varieties of agricultural crops could push landraces to extinction—is a valid one, though not in the way he is concerned about. According to a United Nations report, the success of commercial varieties of all kinds—not just those created using molecular techniques—has led to the disappearance from farms of more than 80 percent of old-fashioned apples, maize, tomatoes, wheat, and cabbages worldwide. The introduction of the improved varieties has had the positive effect of increasing yields. But the widespread use of just a few different varieties leads to genetic sameness in crops and this genetic uniformity can make them vulnerable to new pathogens or unusual environmental conditions on a very large scale, as evidenced by the wheat rust epidemics.

Preserving crop gene pools is increasingly the function of organizations specifically devoted to the task. In America more than 400,000 crop varieties are maintained in the USDA’s National Plant Germplasm System. IRRI in the Philippines has more than 80,000 varieties of rice alone. CIMMYT has collected and preserved hundreds of landraces of maize, periodically growing and harvesting them to preserve their viability. In response to the publicity generated by Quist and Chapela’s

report, CIMMYT researchers sampled 42 of the landrace populations in its seed bank, finding no evidence that the Bt gene was present in any of the plants tested.

But perhaps this sudden concern with genetic purity is misplaced. In their work with small-scale farmers, CIMMYT’s researchers have found that landraces are far from being museum pieces. Population geneticist Julien Berthaud argues that in Oaxaca, where Chapela and Quist gathered their corn, the landraces do not even meet the standard definition of a crop variety. They are not distinct or uniform or stable. Trying to maintain one in a static form could doom it. As a CIMMYT press release explains, “Small-scale farmers select their own seed. Often they choose the best ears at harvest and save seed from only a few cobs—a logical approach but one that increases deleterious mutations. As defects accumulate, the variety loses its genetic value.” Rather than preserving a landrace, Berthaud says, what needs to be preserved is “the active flow of genes.” If the Bt gene is useful for insect control in landraces, the Mexican farmer who succeeds in introgressing the gene into a landrace will hardly obliterate that landrace but on the contrary will increase its chances of survival.

The idea that transgenic maize will displace landraces is nonsense, says Major Goodman of North Carolina State University. Goodman spoke at a conference titled “Gene flow: what does it mean for biodiversity and centers of origin?” that was sponsored by the Pew Foundation and held in Mexico City in September 2003. Despite the availability of improved corn varieties since the 1930s and the intensive maize breeding that has been conducted in Mexico, Goodman said, there has been little impact on the indigenous landraces grown by 80 percent of Mexican farmers. The new varieties designed for the U.S. Corn Belt are far from well adapted to the subtropics, where they are disease-prone and stress-sensitive. Even though such transgenic varieties have undoubtedly been introduced in Oaxaca, Goodman doubts that it will matter much—they will fare so poorly next to the landraces.

It is the hard struggle for survival faced by the small-scale Mexican farmers who grow and consume the native varieties that is the greatest threat to preservation of the landraces. Said Goodman, “Their economic survival (and hence maize diversity’s survival) prospects are

bleak, and transgenic maize is probably one of the least of their problems.” Adds CIMMYT director general Iwanaga, “The perception that transgenic maize is reducing diversity must not obscure the very real need for research to mitigate the many confirmed threats to maize diversity. Every day, diversity is eroded by habitat destruction, human migration from rural to urban areas, and the irreparable loss of traditional maize seed and knowledge as the farming population ages. The present concern about transgenic maize may only add to these threats. If farmers and consumers are convinced that landraces are ‘contaminated’ by transgenes and therefore unsafe to grow or eat, farmers will have even fewer incentives to preserve landraces in their fields.”

Even if one variety of maize did crowd out another in a Mexican farmer’s field, as Chapela fears, it wouldn’t have any effect on biodiversity as Val Giddings of BIO was using the term. Giddings was speaking about biodiversity in the much larger sense of the diversity lost when acres of forest are cut down, plowed, planted, and turned into farmers’ fields. Rather than thinking about the diversity within a single species—maize—he is thinking about the estimated 10 to 30 million species (not counting bacteria) that inhabit the earth today. In his closing remarks at the Mexico City conference on gene flow, Peter Raven spoke of both meanings of the term. Raven, as director of the Missouri Botanical Garden, is a tireless crusader for biodiversity. He noted that the historical rate of extinction climbed from about 10 per year in 1600 to roughly 100 per year in 1950. It now stands at several thousand extinctions per year.

Biodiversity in Mexico, as elsewhere, says Raven, “can be preserved only in the context of a sustainable nation.” Maintaining biodiversity is not as simple as keeping genetically modified maize out of the country. The most serious threats to both maize biodiversity and to biodiversity overall, concludes Raven, are not from GM maize, but from “habitat destruction, urbanization and the abandonment of cultures, alien invasive weeds and pests, and insufficient attention to indigenous peoples and to agriculture in general.” In today’s world, with its still-growing human population, preserving natural biodiversity requires raising agricultural yields to reduce the demand for new acres to plant. The best modern techniques are needed to increase the nutritional content of

crops, while decreasing the impact of agricultural chemicals on wildlife. To Raven, preserving biodiversity also means “increasing the productivity of selected landraces” using contemporary molecular techniques. Far from being a dire threat, the introgression of selected genes that enhance insect and disease resistance might ensure the survival of maize landraces and preserve biodiversity—in the larger sense—by improving small-scale farmers’ yields and lives.