Forum

A Framework for Assessing the Risk of Transgenic Crops JAMES F. HANCOCK

The environmental risks of many transgenic crops can be evaluated without additional experimentation by using already available information on the biology of the crop, the presence of compatible relatives, and the transgene phenotype. The level of crop invasiveness and the location of compatible relatives can be determined by consulting local floras and the crop literature. Decisions about invasiveness can be bolstered by determining the number of weediness traits carried by the crop and its congeners. The potential impact of transgenes can be ranked by their likely effect on reproductive success, ranging from neutral to advantageous to detrimental. This scheme can identify not only the low-risk transgene–crop combinations that are safe to deploy but also those that either are too dangerous to release or require additional experimentation. Keywords: genetically modified crops (GMOs), genetic engineering, invasiveness, gene flow, introgression

T

here is considerable international concern about the release of transgenic crops. These concerns have stalled research, particularly in Europe, where the further development of genetically modified crops has been essentially stopped. At the forefront have been fears about the risk of transgene escape into natural populations and an alteration of the “natural balance” (Colwell et al. 1985, Ellstrand 1988, Rissler and Mellon 1996, Marvier 2001). These risks are very real, and in many instances knowledge is insufficient to allow deployment of transgenic crops; however, there are cases where environmental risk can be assigned on the basis of existing knowledge about the biology of the crop, the proximity of wild relatives, and the phenotypic effects of the transgene. Further delays of many biotechnological advances might be minimized if there were a consistent protocol that could be used to rank transgenic crops according to their environmental risk. Lengthy delays in release could be avoided by deciding at the onset of research and development what types of environmental research are needed for future deployment. In some cases, previous knowledge can be used to decide that a transgenic crop is of very low risk, and development and subsequent deployment can proceed rapidly with little additional environmental research. In other cases, transgenic crops might be deemed of such high risk that their development is foolhardy and should not progress at all. In many other cases, the transgenic crop might be considered intermediate in risk and would require varying levels of environmental research before release. Such early analyses could save consid512 BioScience • May 2003 / Vol. 53 No. 5

erable time at the point of release by identifying what environmental research should be done during the developmental stage. In this way, when the transgenic crop is ready for commercialization, it can indeed be released. This article presents a decisionmaking framework based on answers to three questions about risk factors: (1) Is a compatible relative present in the areas of deployment? (2) Is the native relative or crop highly invasive? (3) Will the engineered trait significantly affect the invasiveness of the crop or native relative? There are other risks involved with the release of transgenic crops, such as nontarget effects of novel toxins or development of resistance in native populations, but the emphasis here will be placed on fitness effects and the potential to increase invasiveness. The outlined framework is based on the assumption that the potential invasiveness of a plant species can be predicted if we have knowledge about its biology, its distribution, and the likely fitness impact of a transgene. The factors limiting gene flow between compatible relatives can be largely ignored, as transgenes will eventually escape into the natural environment if there is a compatible relative near the transgenic

James F. Hancock (e-mail: [email protected]) is a professor in the Department of Horticulture, Michigan State University, East Lansing, MI 48824. He is an evolutionary botanist who also breeds blueberries and strawberries, using both conventional and biotechnological approaches. He is the author of a forthcoming book, Plant Evolution and the Origin of Crop Species, which will be published by CAB International. © 2003 American Institute of Biological Sciences.

Forum crop (Hancock et al. 1996, Ellstrand 2001), unless the transgenic crop produces no viable gametes or has a system incorporated that prevents embryo viability. With acceptance of this reality, decisions on risk can focus on evaluating the invasive characteristics of potential recipient species and determining whether the transgene itself will have a significant impact on the fitness of those species. North American crops will be used as examples, but the protocol may work in any part of the world, depending on the array of crops and the native progenitors present.

Previous research on risk of gene escape and crop invasiveness Numerous authors have pointed out that transgene flow from engineered crops to their wild relatives could result in the evolution of increased invasiveness in wild relatives and in the evolution of pests that are resistant to newly developed control strategies (Dale 1992, Rissler and Mellon 1996, Snow and Palma 1997, Hails 2000). Invasive species are defined here as those that readily increase in numbers and aggressively spread, outcompeting other species for resources. This applies in both agronomic and native environments. A species is considered to have “increased invasiveness” if its abundance noticeably increases and it outcompetes a species that it theretofore could not. The likelihood of hybridization between crops and their wild relatives has been measured in numerous studies (Ellstrand 2001). Although the early consensus was that such hybridizations occurred infrequently, research in the last decade has shown that they are relatively common (Ellstrand et al. 1999, Desplanque et al. 2002) and that crop alleles can persist for long periods in natural populations (Klinger and Ellstrand 1994, Arriola and Ellstrand 1996, Linder et al. 1998). Factors such as breeding system, flowering time, hybrid viability, and isolation distance can alter the rate of gene escape (Hokanson et al. 1997, Hancock et al. 1996), but if compatible relatives are within the cloud of crop pollen, genes will escape. The first determination that should be made concerning the risk of transgene escape is whether compatible relatives exist in the area of deployment. The literature on breeding contains much information on what wild species are compatible with crops—breeders frequently want to find potential gene sources outside the immediate crop species—and considerable work has been done on elucidating the wild progenitors of crops. Several large compilations of crop histories provide portals to the literature (see Sauer [1993], Zohary and Hopf [1993], Smartt and Simmonds [1995], Hancock [2003]). Local floras can be used to provide information about the geographical ranges of species and their invasiveness. Most floras describe the location and relative abundance of native species and whether crops persist in the wild. When local floras representing California (Munz and Keck 1973) and the southern United States (Radford et al. 1968) and eastern parts of the United States (Gleason and Cronquist 1963) were examined, Hancock and colleagues (1996) found that

about 62 percent of the crops that had come under APHIS (Animal and Plant Health Inspection Service, US Department of Agriculture) oversight were not persistent in native environments and thus could be considered noninvasive. Another 21 percent persist for a few generations in native environments but eventually disappear. These have slightly higher native fitness than the nonpersistent crops, but they can still be considered noninvasive, because they do not spread outside the agroecosystem. About 17 percent fall into the persistent category; these can be ranked as invasive, because they readily reproduce outside the agroecosystem and spread. The nonpersistent species are probably of minimal risk in genetic engineering, unless the transgene incorporated has dramatic fitness effects; those in the persistent, invasive class fall into the high-risk category, both as pollen donors and as weeds. Those in the persistent but noninvasive category would be intermediate in risk, depending on the nature of the transgene and the presence or absence of compatible relatives. A set of traits associated with plants’ ability to readily invade disturbed habitats can be used to provide supplementary information on the potential invasiveness of a crop and its wild progenitors. In a widely cited pair of papers, Baker (1965, 1974) listed traits associated with the most successful weeds: broad germination requirements, seed dispersal over short and long distances, discontinuous germination, vigorous vegetative reproduction, rapid growth to flowering, brittle propagules, continuous seed production, vigorous competitors, self-pollination, polyploidy, unspecialized pollinators, long-lived seeds, very high seed output, and plastic seed production (see box 1). Although Baker was most concerned with identifying those traits associated with success in disturbed habitats, the same traits can contribute to invasiveness in natural environments. Keeler (1989) found that what are considered to be the world’s 17 worst weeds possessed on average 85.6 percent (0.12 Box 1. Traits associated with the most successful weeds (Baker 1965, 1974) • • • • • • • • • • • • • •

Broad germination requirements Discontinuous germination Long-lived seeds Rapid growth to flowering Continuous seed production Self-pollination Unspecialized pollinators High seed output Seeds produced in many habitats Seed dispersal over short and long distances Vigorous vegetative reproduction Brittle propagules Vigorous competitors Polyploidy

May 2003 / Vol. 53 No. 5 • BioScience 513

Forum standard error, or SE) of Baker’s weediness traits, whereas a group of 20 randomly selected species had 59 percent (0.20 SE) and crop plants had 42 percent (0.14 SE). This suggests that the invasive potential of a species can often be characterized by evaluating the number of weediness traits it contains. This classification system is not foolproof—Keeler observed considerable variability in each of the categories—but all of the serious weeds had more than 65 percent of the weediness traits. If this ranking scheme is combined with information on crop persistence in the wild, it can aid in the identification of high- and low-risk species. For example, it is unlikely that the addition of a transgene to a crop that does not persist in nature and has few weediness characteristics (i.e., less than 40 percent) will make it highly invasive, unless the transgene has very powerful effects on the phenotype; however, even a subtle change in the phenotype of an already invasive species with numerous weedy traits (more than 80 percent) could make it much more problematic. For those species that persist in native environments and carry intermediate numbers of weediness traits, risk would be difficult to predict; additional experimentation would be necessary.

The environmental impact of transgenes The potential impact of individual transgenes can be determined by evaluating their phenotypic effect. Although current information may be insufficient to rank the relative risk of many transgenes, transgenes can be grouped by the type of impact they have on reproductive fitness: The impact can be neutral; detrimental; variable, depending on the weediness of the recipient species; variable, depending on the degree of natural biological control; or advantageous. Transgenes that have a neutral effect on fitness might spread in natural populations through genetic drift, but they would have no subsequent impact on fitness. Few transgenes can now be clearly classified in this category, except for the reporter genes used to recognize transformants such as nptII and B-glucuronidase. These genes have been inserted into a wide array of genomes without any noticeable effects on phenotype, as long as their placement does not disrupt gene function or regulation. Some of the genes intended to remove toxic substances through phytoremediation, such as pentaerythritol tetranitrate reductase (French et al. 1999), can be considered neutral if they have no phenotypic effect on noncontaminated soils and no negative pleiotropic effects. Other genes, however, such as mercuric ion reductase (Rugh et al. 1996) and organomercurial lyase (Bizily et al. 2000), should be considered detrimental because they reduce plant fitness in the absence of heavy metal contamination. Genes with detrimental effects will be selected against in the natural environment and will not spread. Many of the traits associated with crop domestication fall into this category. The weediness of wild species has been increased in agronomic fields through crop hybridization (Ellstrand and Hoffman 1990), but there is little evidence of the reverse, where crop genes have affected the fitness of a wild species in its 514 BioScience • May 2003 / Vol. 53 No. 5

native environment. In spite of many substantial advances in breeding for resistance to pests, drought, cold, and salinity, studies have not yet shown that the native fitness of the wild species was noticeably changed through hybridization with the crop progenitor. However, genes with detrimental effects could become important if population sizes of the compatible relatives of the genetically modified crop are very small and endangered. In this case, the bombardment of the native populations by a deleterious gene could result in their extinction through “swamping,” whereby the fitness of the natural population would be reduced to the point that it could no longer persist (Rissler and Mellon 1996). I am not aware of any North American crop relatives that fall into this category, but the possibility exists in other parts of the world. Examples of transgenes that fit into the detrimental category are male sterility, altered fiber quality, changes in lignin biosynthesis, and altered fruit ripening and storage characteristics. Transgenes incorporated for biological production of proteins, such as serum albumin, could also fall into the detrimental category even though they do not have a direct negative effect, because such genes could have negative pleiotropic effects on overall plant vigor. Many of the transgenes used to alter the amino acid composition of proteins for improved nutrition must also be placed in the detrimental category, at least initially, until it can be proved that they do not have negative pleiotropic effects on growth and development. The transgenes that produce herbicide and pest resistance will vary in their fitness potential, depending on the invasiveness of the recipient species and the level of natural control. Genes for viral, fungal, and pest resistance fall into a group whose incorporation into natural populations could increase fitness, if the pest is currently controlling natural populations. The role of disease in controlling natural populations is just beginning to become clear, but there are at least a few examples where the removal of insect pests has been shown to have significant impacts on the reproductive success of native species (Marvier and Kareiva 2000). Some of the transgenes already deployed that fall into this category are Bt genes for insect resistance in a number of crops; phytophothora, PLRV, and PYV resistance in potato; PRSV resistance in papaya; and CMV, WMV2, and ZYMV in squash. Herbicide resistance genes fall into a separate category, because they are selectively neutral in the natural environment, though if they were incorporated into already weedy species, they could abolish a valuable method of control. Examples of these transgenes are those providing resistance to phosphinothricin, glyphosate, and sulfonylurea in a wide array of crops. Transgenes that change the environmental tolerance of a species or alter its patterns of growth and development could result in dramatic adaptive shifts and have a major impact on fitness. For example, juvenility in trees might be reduced by overexpression of a regulatory gene such as LEAFY (Peña et al. 2001), allowing for earlier reproduction and possibly greater overall reproductive success because of more frequent flowering. The cold tolerance of species might be altered

Forum by the overexpression of the transcription factor CBF1 (Jaglo-Ottosen et al. 1998, Jaglo et al. 2001), in which case the adaptive range of the species would shift to more northern climates. If they escaped, transgenes such as metallothioneins (Cai et al. 1995), which are used in phytoremediation to remove heavy metals, could also allow a native species to invade new, heavy-metal-laden habitats. It should be noted, however, that many of the transgenes that dramatically alter the physiological tolerances of crop species can also have negative fitness effects that would make them deleterious in the native environment. Stabilizing selection has operated in natural populations to optimize physiological function, and the kinds of genes sought by farmers will often be radically different from those that promote overall fitness in wild plants under intense competition. For example, overexpression of CBF1 might increase cold tolerance, but it dramatically alters the development of Arabadopsis thaliana and Brassica oleraceae, resulting in reduced vigor (Jaglo-Ottosen et al. 1998, Jaglo et al. 2001). Earlier reproduction in a woody species could result in reduced fitness, because the plant might be weakened by early reproduction or might ultimately produce far fewer seeds. The plants engineered for phytoremediation may not grow as rapidly as their antecedents and may be outcompeted in natural environments. If “super-genes” do emerge that are dramatically more powerful than conventionally derived ones, they may often have complex effects with negative impacts on overall fitness in nature. In such cases, the transgenes can be deemed safe for development. One fear that has been expressed is that transgenes will have unexpected epistatic or pleiotropic effects in natural populations. While this possibility cannot be completely excluded, it is highly unlikely; the constructs that have been selected for deployment in crops have already faced numerous phenotypic screens, from the initial transformations to the final field screens before release. If there is doubt about their overall effects in the crop, further research is indeed warranted before deployment, but it is improbable that the transgenes will act

differently in the crops than in their progenitor genome or that they will have more extensive epistatic or pleiotropic effects than native genes. Genes have been moved from native species to crop species by breeders without any unexpected ramifications, and as discussed earlier, substantial gene flow has occurred between conventionally bred crops and wild species without any apparent dramatic, unexpected alterations in phenotype.

A framework for estimating the environmental risk of trangenic crops The considerations outlined above suggest that there are two major factors that influence the environmental risk of transgenic genotypes—the potential invasiveness of the species and the relative impact of the transgene on fitness. Of course, species have differing geographical ranges, and estimates of invasiveness may vary from country to country. In addition, a transgene deemed safe in one area may be risky in another. The relative risk of a transgenic crop needs to be determined for all geographic regions in which it will be deployed. The availability of compatible relatives and the biology of the crop itself determine the degree of invasive potential the crop possesses. Various crops are grouped in table 1 into what will be referred to here as species categories, in accordance with their invasive potential. In addition, several degrees of risk are associated with particular transgenes, depending on whether the fitness impact of the genes is neutral or selectively advantageous. These are outlined in table 2 and will be referred to as transgene categories. The overall environmental risk of the transgenic crop can be determined by combining these two categories. The six species categories shown in table 1 are • S-1: Crops that have no compatible relatives, carry few weediness traits (less than 40 percent), and do not persist in natural environments • S-2: Crops that have no compatible relatives, carry intermediate numbers of weediness traits, rarely escape, and do not persist in natural environments

Table 1. Invasive biology of crop species and their compatible relatives in North America. Category

Biology

Examples of crops in North America

S-1

No compatible relatives. Crop carries only a few weediness traits and does not escape or persist.

Broccoli, cabbage, cauliflower, citrus, cucumber, cotton, eggplant, pea, potato, soybean, sugarcane, tomato, and watermelon

S-2

No compatible relatives. Crop carries an intermediate number of weedy traits but rarely escapes and does not persist.

Peanut and beans

S-3

No compatible wild relatives. Crop carries many weediness traits and can escape and persist.

Barley and wheat

S-4

Has compatible relative. Crop or relative carries only a few weediness traits. Crop can escape but does not persist. Native relative does not aggressively spread.

Celery, lettuce, maize, melon, pepper, squash, and tobacco

S-5

Has compatible relative. Crop or relative carries intermediate numbers of weedy traits. Crop can escape and persist. Native relative does not aggressively spread.

Apple, asparagus, beet, blueberry, carrot, cranberry, onion, pear, poplar, plum, radish, spruce, and strawberry

S-6

Has compatible wild relative. Crop or relative carries many weediness traits. Crop can escape and persist. Native relative spreads aggressively.

Oats, rapeseed, rice, sorghum, and sunflower

May 2003 / Vol. 53 No. 5 • BioScience 515

Forum • S-3: Crops that have no compatible wild relatives, carry many weediness traits, and can escape and persist in natural environments • S-4: Crops that have compatible relatives, carry few weediness traits, and can escape but do not persist in natural environments; their compatible relatives also carry few weediness traits and do not aggressively spread • S-5: Crops that have compatible relatives, carry intermediate numbers of weediness traits, and can escape and persist in natural environments; their compatible relatives also carry intermediate numbers of weediness traits but do not aggressively spread • S-6: Crops that have compatible wild relatives, carry many weediness traits, and can escape and persist in natural environments; their compatible relatives also carry many weediness traits and aggressively spread

The five transgene categories presented in table 2 are • T-A: Transgenes that are selectively neutral in the natural environment • T-B: Transgenes that have negative impacts on fitness • T-C: Transgenes for herbicide resistance • T-D: Transgenes for pest resistance • T-E: Transgenes that alter environmental tolerance or development

The relative risk of deployment of transgenic crops is shown in table 3. According to this scheme, any transgenes that do not have nontarget influences are safe for deployment in S-1 crops without further experimentation. Crops such as potato and soybean have no compatible relatives in the United States. Thus, transgenes cannot escape into native relatives, do not persist in native environments, and have so few weediness traits that even dramatic changes in their adaptations (T-E) are highly unlikely to make them invasive. In addition, it is relatively easy to control them without herbicides, so the incorporation of T-C transgenes is unlikely to cause major pest problems. All transgenes except those with major impacts on adaptation are probably safe to deploy in S-2 crops without further experimentation. These species would have no compatible relatives, so transgenes could not escape, but the crop could escape and would have enough weediness traits that a dramatic change in its adaptations (T-E) could make it invasive. T-C genes would be of low risk, because it is relatively easy to

control an S-2 crop without herbicides, and because to date these species have not evolved resistance to the available herbicides. Most pest resistance transgenes (T-D) would also be of low risk, as there are no reported incidences, to my knowledge, where a single pest has been shown to play a major role in preventing a crop from becoming a weed. More scrutiny would have to be placed on genes that confer broad rather than narrow resistance, because their spectrum of target organisms would be larger. Only T-A and T-B genes are safe to deploy in both S-3 and S-6 species. S-6 crops such as sorghum and rice already spread aggressively in native environments, so any change in their fitness might make them a greater pest, and beneficial transgenes associated with environmental tolerances (T-E) and pest resistance (T-D) could escape through pollen flow and further increase the invasiveness of the native species. S-3 species such as barley and wheat have no compatible relatives, so there is no risk of transgene escape. The crop itself, however, is persistent in native environments, so any increase in its fitness through increased adaptation (T-E) or pest resistance (T-D) might make it more invasive. Herbicides are used to control the S-3 and S-6 species as weeds, so the deployment of genotypes with T-C must be carefully evaluated. S-6 species are strong competitors and have large population sizes, so the escape of deleterious alleles is unlikely to have a negative impact on the species. Only T-A and T-C genes are safe to deploy in S-4 species. Crops such as maize, melon, and squash carry few weediness traits and do not persist in native environments. Even dramatic changes in these crops’ adaptations are unlikely to produce a greater agronomic pest; however, beneficial transgenes associated with environmental tolerance (T-E) and pest resistance (T-D) could escape into natural populations and alter their fitness characteristics. These would require a case-by-case evaluation of the role of these factors in limiting their range. S-4 species are at best only modest competitors, so the escape of deleterious alleles could have a negative impact on their natural populations. T-C genes are relatively safe for deployment in S-4 crops, because the crops and their relatives are easy to control without herbicides and to date they have not evolved resistance to the available herbicides. The T-A, T-B, and T-C genes are safe to deploy in S-5 crops such as apple or strawberry. Because these crops have intermediate numbers of weediness traits and do escape, a

Table 2. Relative fitness impact of transgenes. Category

Fitness impact

Examples of transgene use

T-A

Neutral in the native environment

Marker genes

T-B

Detrimental in the native environment

Male sterility, altered fiber quality, altered fruit ripening, and storage

T-C

Variable, depending on invasiveness of crop or native relative

Herbicide resistance

T-D

Variable, depending on level of biological control

Viral, fungal, and pest resistance

T-E

Potentially advantageous in the native environment

Cold, drought, and metal tolerance; improved nutrient uptake; altered development

516 BioScience • May 2003 / Vol. 53 No. 5

Forum Table 3. Transgenic crops safe to release without further experimentation. Crop category

Transgene category

Transgene characteristics

S-1

T-A, T-B, T-C, T-D, T-E

There are no compatible relatives, and the crop has so few weediness traits that even the most dramatic changes in phenotype are highly unlikely to make it invasive. The crop is easy to control without herbicides.

S-2

T-A, T-B, T-C, T-D

There are no compatible relatives, but the crop has enough weediness traits that a dramatic change in its adaptations could make it invasive. A lack of pest resistance is unlikely to be the primary factor preventing invsasiveness. The crop is relatively easy to control without herbicides.

S-3

T-A, T-B

There are no compatible relatives, but the crop itself is weedy, so any change in its fitness might make it more invasive. Herbicides are also needed for the crop’s control.

S-4

T-A, T-C

The crop or native relative has so few weediness traits that even the most dramatic changes in phenotype are highly unlikely to make it invasive in agronomic systems; however, advantageous and detrimental transgenes could escape into natural populations and significantly alter their fitness. It is relatively easy to control the crop and its relatives without herbicides.

S-5

T-A, T-B, T-C

The crop or native relative has enough weediness traits that a dramatic change in their adaptations could make it invasive; however, advantageous transgenes could escape into natural populations and positively alter their fitness. The escape of detrimental traits is unlikely to have long-term influences on native populations, because population sizes are large. It is relatively easy to control the crop and its relatives without herbicides.

S-6

T-A, T-B

The crop is already invasive, so any positive change in its fitness might make it a greater pest. Beneficial transgenes associated with environmental tolerances and pest resistance could escape into natural populations and alter their fitness. The escape of detrimental traits is unlikely to have long-term influences on native populations, because population sizes are large. In addition, herbicides are needed for the control of the crop and its relatives.

dramatic change in their adaptations could allow them to be more invasive. Beneficial transgenes associated with environmental tolerance (T-E) and pest resistance (T-D) are also risky, since they could escape into natural populations and alter species fitness. Detrimental traits (T-B) could also persist in native populations and diminish their fitness if native population sizes are low, but this is much less likely in S-5 than in S-4 species, which are stronger competitors. It is relatively easy to control the crop and its relatives without herbicides, so the deployment of T-C should not dramatically affect control.

Cases where additional experiments are needed The types of further experimentation that are needed will depend on the type of crop–transgene combination (table 4). The risk involved in the deployment of any transgene can be deemed low and needing no further experimentation if the phenotypic effects of the transgenes closely mimic conventionally deployed or native genes (Hokanson et al. 2000). This would fit a number of genes for pest resistance (T-D), many of which affect aspects of plant development such as flowering time, and alterations in environmental tolerance (TE). If similar genes are already in natural populations, the genes have already faced the sieve of natural evolution and are

unlikely to have a major impact. Examples of similar phenotypes include the pest resistance previously mentioned, developmental processes such as early flowering or photoperiod insensitivity, and physiological tolerances to cold or salt. If the native species already carries a phenotype that is being engineered into the crop, it is unlikely to have a significant impact on fitness. This approach, in which transgene phenotypes are used to assign risk, was recommended by the first group of scientists who evaluated the environmental risks of transgenic crops (Tiedge et al. 1989) and in the National Research Council report (CLS 1989) Field Testing Genetically Modified Organisms: Framework for Decisions. Unfortunately, although many of the transgenes that have an impact on resistance are similar to conventionally deployed resistance genes (Hokanson et al. 2000), many are not. Also, many of the other types of transgenes influencing physiological traits will produce phenotypes that are unique to the species or have broader effects than the native genes. These situations will require further experimentation, depending on the invasive characteristics of the transgenic crop species. Among those crops that are not invasive themselves and that do not have compatible relatives but do have intermediate numbers of weediness traits (S-2), it will be necessary to

Table 4. Various crop–transgene combinations that require further experimentation. Transgene category

Crop type

Further research needed

T-A

None

T-B

S-4, S-5, S-6

None.

T-C

S-3, S-6

Document that any wild recipient species can still be controlled as agronomic weed.

T-D

S-1, S-2

Show that a similar phenotype exists in wild populations. If not, measure levels of native biological control. If levels are significant, test fitness of transgenic crop and hybrids in representative native environments.

T-E

S-1

Document that wild recipient species is not endangered.

Show that a similar phenotype exists in wild populations. If not, test fitness of transgenic crop and hybrids in representative native environments.

May 2003 / Vol. 53 No. 5 • BioScience 517

Forum determine whether the genes causing dramatic change in the crop’s adaptations or development (T-E) could make them more invasive. If clearly deleterious pleiotropic effects are observed in the initial field and greenhouse trials of the transgenic crop, the transgene can be downgraded to T-B and further experimentation will not be necessary. However, if no clear negative interactions were observed, the fitness of the transgenic crop will need to be tested at multiple sites in a range of the natural environments found adjacent to crop fields. Crowley and colleagues (2001) have provided a model of how this can be done. In the other crops that are not invasive themselves but have proximal native relatives (S-4 and S-5), further research will be warranted for the genes associated with detrimental traits (T-B), pest resistance (T-D), and dramatic alterations in physiology and development (T-E). The fitness impact of T-E transgenes will have to be determined on the crop, as well as on F1 hybrids and backcrosses with native plants (Rissler and Mellon 1996). Likewise, beneficial genes associated with pest resistance might also escape from these transgenic crops into the native populations, so their fitness impact must also be examined (Bartsch et al. 1996, 2001). As previously noted, a similar phenotype might have been reported from native sources, which would eliminate the need for further experimentation, but if this is not the case, the transgene’s influence on fitness would have to be measured on F1 hybrids and backcrosses with native plants (Rissler and Mellon 1996). In the case of deleterious genes (T-B), before the transgenic crop can be released, it will be necessary to document that the population sizes of wild compatible relatives are large. If the population sizes of native relatives are very small and can be classified as endangered, the transgenic crop should not be released. Among the crops that can persist in nature with or without native relatives (S-3 and S-6), further research will need to be conducted on the impact of transgenes associated with pathogen or insect resistance and physiological or developmental shifts. In the case of S-3 crops without native relatives, the effect of the transgene on fitness will only have to be measured on crop genotypes across a wide array of natural environments (Crowley et al. 2001); however, with S-6 crops that have compatible relatives, the fitness of F1 and backcross individuals will have to be tested also in the native environment. Transgenes associated with herbicide resistance should not be incorporated in either S-3 or S-6 crops, because farmers will lose a valuable means of controlling them as agronomic weeds.

Conclusions The environmental risk of many transgenic crops can be evaluated with little further study than that conducted during their development. When the biology of the crop is known and the nature of the transgene is understood, low- and high-risk categories can be clearly identified without additional experiments. In the middle-risk category, further experimentation is necessary only if phenotypes similar to that of 518 BioScience • May 2003 / Vol. 53 No. 5

the transgene cannot be identified in native populations. Careful evaluation of the fitness characteristics of transgenic genotypes during the developmental stage also can obviate the need for additional experimentation. If such measures are employed early during research and development—before a company tries to release a transgenic crop—costly mistakes may be averted. Evaluation efforts should be aimed at identifying the low-risk transgenic crops for which commercialization can move rapidly; the medium-risk categories of crops, for which ecological research must be conducted alongside the biotechnology; and the high-risk transgenic crops, whose further development should be discouraged.

References cited Arriola PE, Ellstrand NC. 1996. Fitness of interspecific hybrids in the genus Sorghum: Persistence of crop genes in wild populations. Ecological Applications 7: 512–518. Baker HG. 1965. Characteristics and modes of origin of weeds. Pages 147–168 in Baker HG, Stebbins GL, eds. The Genetics of Colonizing Species. New York: Academic Press. ———. 1974. The evolution of weeds. Annual Review of Ecology and Systematics 5: 1–23. Bartsch D, Schmidt M, Pohl-Onf M, Haag C, Schuphan I. 1996. Competitiveness of transgenic sugar beet resistant to beet necrotic yellow vein virus and potential impact on wild beet populations. Molecular Ecology 5: 199–205. Bartsch D, Brand U, Morak C, Pohl-Onf M, Schuphan I, Ellstrand NC. 2001. Biosafety of hybrids between transgenic virus-resistant sugar beet and Swiss chard. Ecological Applications 11: 142–147. Bizily SP, Rugh CL, Meagher RB. 2000. Phytodetoxification of hazardous organomercurials by genetically engineered plants. Nature Biotechnology 18: 213–217. Cai XH, Traina ST, Logan TJ, Gustafson T, Sayre RT. 1995. Applications of eukaryotic algae for the removal of heavy metals from water. Molecular Marine Biology and Biotechnology 4: 338–344. [CLS] Commission of Life Sciences, Board of Biology, National Research Council. 1989. Field Testing Genetically Modified Organisms: Framework for Decisions. Washington (DC): National Academy Press. Colwell RE, Norse EA, Pimentel D, Sharples FE, Simberloff D. 1985. Genetic engineering in agriculture. Science 229: 111–112. Crowley MJ, Brown SL, Hails RS, Kohn DD, Rees M. 2001. Transgenic crops in natural habitats. Nature 409: 682–683. Dale PJ. 1992. Spread of engineered genes to wild relatives. Plant Physiology 100: 13–15. Desplanque B, Hautekèete N, Van Dijk H. 2002. Transgenic weed beets: Possible, probable, avoidable? Journal of Applied Ecology 39: 561–571. Ellstrand NC. 1988. Pollen as a vehicle for the escape of engineered genes? Pages S30–S32 in Hodgson J, Sugden AM, eds. Planned Release of Genetically Engineered Organisms. Cambridge (United Kingdom): Elsevier. ———. 2001. When transgenes wander, should we worry? Plant Physiology 125: 1543–1545. Ellstrand NC, Hoffman CA. 1990. Hybridization as an avenue of escape for engineered genes: Strategies for risk reduction. BioScience 40: 438–442. Ellstrand NC, Prentice HC, Hancock JF. 1999. Gene flow and introgression from domesticated plants into their wild relatives. Annual Review of Ecology and Systematics 30: 539–563. French CE, Rosser SJ, Davis GL, Nicklin S, Bruce NC. 1999. Biodegradation of explosives by transgeneics expressing pentaerythritol tetranitrate reductase. Nature Biotechnology 17: 491–494. Gleason HA, Cronquist A. 1963. Manual of Vascular Plants of Northeastern United States and Adjacent Canada. New York: Van Nostrand. Hails RS. 2000. Genetically modified plants: The debate continues. Trends in Ecology and Evolution 15: 14–18.

Forum Hancock JF. 2003. Plant Evolution and the Origin of Plant Species. Wallingford (United Kingdom): CAB International. Forthcoming. Hancock JF, Grumet R, Hokanson SC. 1996. The opportunity for escape of engineered genes from transgenic crops. HortScience 31: 1080–1085. Hokanson SC, Grumet R, Hancock JF. 1997. Direct comparison of pollenmediated movement of native and engineered genes. Ecological Applications 7: 1075–1081. Hokanson KD, et al. 2000. The concept of familiarity and pest resistant plants. Pages 15–20 in Traynor PL, Westwood JH, eds. Ecological Effects of Pest Resistance Genes in Managed Ecosystems. Blacksburg (VA): Information Systems for Biotechnology. Jaglo KR, Kleff S, Amundsen KL, Zhang X, Haake V, Zhang JZ, Deits T, Thomashow MF. 2001. Components of the Arabidopsis C-repeat/ dehydration-response element binding factor cold-response pathway are conserved in Brassica napus and other plant species. Plant Physiology 127: 910–917. Jaglo-Ottosen KR, Gilmore SJ, Zarka DG, Schabenberger O, Thomashow MF. 1998. Arabidopsis CBF1 overexpression induces COR genes and enhances freezing tolerance. Science 280: 104–106. Keeler KH. 1989. Can genetically engineered crops become weeds? Bio/ Technology 7: 1134–1139. Klinger T, Ellstrand NC. 1994. Engineered genes in wild populations: Fitness of weed-crop hybrids of radish, Raphanus sativus L. Ecological Applications 4: 117–120. Linder CR, Taha I, Seiler GJ, Snow AA, Rieseberg LH. 1998. Long-term introgression of crop genes into wild sunflower populations. Theoretical and Applied Genetics 96: 339–347. Marvier M. 2001. Ecology of transgenic crops. American Scientist 89: 160–167. Marvier M, Kareiva P. 2000. Extrapolating from field experiments that remove herbivores to population-level effects of herbivore resistance

transgenes. Pages 57–64 in Traynor PL, Westwood JH, eds. Ecological Effects of Pest Resistance Genes in Managed Ecosystems. Blacksburg (VA): Information Systems for Biotechnology. Munz PA, Keck DD. 1973. A California Flora. Berkeley: University of California Press. Peña L, Martin-Trillo M, Juárez J, Pina JA, Navarro L, Martínez-Zapater JM. 2001. Constitutive expression of Arabidopsis LEAFY or APETALA 1 genes in citrus reduces their generation time. Nature Biotechnology 19: 263–267. Radford AE, Ahles HE, Bell CR. 1968. Manual of the Vascular Flora of the Carolinas. Chapel Hill: University of North Carolina Press. Rissler J, Mellon M. 1996. The Ecological Risks of Engineered Crops. Cambridge (MA): MIT Press. Rugh CL, Wilde HD, Stack NM, Thompson DM, Summers AO, Meager RB. 1996. Mercuric ion reduction and resistance in transgenic Arabidopsis thaliana plants expressing a modified bacterial merA gene. Proceedings of the National Academy of Sciences 93: 3182–3187. Sauer JD. 1993. Historical Geography of Crop Plants. Boca Raton (FL): CRC Press. Smartt J, Simmonds NW, eds. 1995. Evolution of Crop Plants. 2nd ed. Essex (United Kingdom): Longman Scientific and Technical. Snow AA, Palma PM. 1997. Commercialization of transgenic plants: Potential ecological risks. BioScience 47: 86–96. Tiedge JM, Colwell RK, Grossman YL, Hodson RE, Lenski RE, Mack RN, Regal PJ. 1989. The planned introduction of genetically engineered organisms: Ecological considerations and recommendations. Ecology 70: 298–315. Zohary D, Hopf M. 1993. Domestication of Plants in the Old World. 2nd ed. Oxford (United Kingdom): Clarendon Press.

May 2003 / Vol. 53 No. 5 • BioScience 519