Ecological, Genetic, and Social Factors Affecting Environmental Assessment of Transgenic Plants
DEVELOPING A TWENTY-FIRST-CENTURY VIEW OF AGRICULTURE AND THE ENVIRONMENT
At the beginning of the twentieth century U.S. farmers were in general free to use whatever agricultural practices best suited their needs. Although the agricultural extension system (Smith-Lever Act of 1914) was in place at the time, its role was to provide advice, not to enforce regulations. As the twentieth century progressed, the impacts of agricultural practices on human health and the environment became a focus of public attention (e.g., Carson 1962). Regulations and incentive programs were developed for agriculture and now have a major influence on farming, ranging from the choice of tillage practices to the choice of pest control techniques.
Another obvious change in agriculture during the twentieth century was in productivity. Some U.S. citizens see the last 50 years of the twentieth century as a time when hundreds of years of insecurity over food availability came to an end. In their eyes, innovative technologies such as plant breeding, water management, fertilizers, and synthetic pesticides played a heroic role in this drama. Others look back on the same events and see an era when for the first time in history human activity threatened the basic stability of global ecosystems on which all life, including human society, depends. In their eyes, modern agricultural science and technology are inimical to the natural environment.
There also are contrasting perceptions of the agricultural environment itself. In the United States, environmental protection became understood by most citizens to have two priorities: (1) protecting human popu-
lations from the toxic effects of pollution and (2) protecting natural areas from human impacts. Both priorities reinforce the view that farms are separate from the natural environment, which must be protected from agriculture. In Europe and some other parts of the world, farms are often seen as an integral part of the natural environment that should be protected (Frewer et al. 1997, Durant et al. 1998). Environmental protection can therefore encompass the protection of a partially domesticated countryside where appropriate farming techniques can foster biodiversity. There certainly are exceptions to this generalization because many U.S. citizens are concerned about the effects of agricultural practices on the flora and fauna of farms, and there are agricultural production regions in Europe that are as thoroughly industrialized as any in the United States (Durant and Gaskell, in press). It is nevertheless important to bear in mind that contrasting cultural values influence the way that different people understand the relationship between agriculture and environment, and this in turn influences their judgment of what constitutes a threat to the environment (Knowles, in press).
The molecular techniques for producing transgenic agricultural crops that came to fruition in the late 1980s arrived on a scene in which advocates for agricultural technology were already prepared to embrace them. Biotechnology was greeted by these advocates as a means of increasing agricultural efficiency, decreasing world hunger, and ameliorating environmental damage caused by previous agricultural technologies. But at the same time, critics of agricultural technology were prepared to view these new techniques with skepticism. Their skepticism was based both on their negative assessment of postwar technologies that contributed to the boom in agricultural productivity and their judgment that the science which could identify the environmental risks of these new technologies was not being adequately supported.
As farming enters the twenty-first century it faces a world in which there is increasing public pressure on governments to more actively protect the environment and conserve biological diversity. Currently, much public concern regarding agriculture and the environment is focused on the potential impacts of transgenic plants. However, it is clear that many of the environmental effects that could result from this specific technology could also occur due to changes in other agricultural technologies and farming practices (see “Environmental Effects of Agroecosystems” below). Concern over the impact of transgenic plants on the environment has led world governments to reassess the standards by which they judge what constitutes a significant negative effect of agriculture on the environment. For example, it is clear that environmental standards being developed for transgenic plant cultivars consider impacts that were rarely even measured when novel conventional crop cultivars were introduced
in the 1960s and 1970s (see Chapter 2). Government agencies charged with the regulation of transgenic plants find themselves in the difficult position of enforcing a much higher environmental standard for these plants than the standards currently used to regulate the impacts of other agricultural technologies and practices. If, as we move further into the twenty-first century, stresses on the environment and public concern about those stresses increase, it is likely that new standards developed for transgenic plants will be applied in some fashion to other agricultural technologies and practices. In that sense, decisions now being made with regard to transgenic plants could set a precedent for evaluating all of agriculture. Government agencies and the public must, therefore, keep an eye to the future when working to develop new environmental standards for transgenic plants.
ROLE OF THIS REPORT
Potential controversy over agricultural biotechnology was anticipated by the U.S. government in the early 1980s. Before any agricultural products of genetic engineering had been developed, the federal government began taking steps to develop a regulatory structure that would assure the safety of potential products. In the mid-1980s the U.S. Coordinated Framework for the Regulation of Biotechnology was developed. This framework (OSTP 1986) calls for the Environmental Protection Agency (EPA), U.S. Department of Agriculture (USDA), and Food and Drug Administration (FDA) to work together in assessing the safety of the process and products of genetic engineering. In its current form, the coordinated framework gives the USDA the lead role in assessing the potential effects of nonpesticidal transgenic plants on other plants and animals in both agricultural and nonagricultural environments. The EPA takes the lead role in assessing the health and environmental effects of plants engineered to produce pesticidal substances, and the FDA leads the review of potential health effects of nonpesticidal transgenic plants.
Over the past 15 years the USDA’s Animal and Plant Health Inspection Service (APHIS) has developed a system for examining potential environmental effects of transgenic plants. However, there has been concern that an agency with a mandate to promote U.S. agriculture may not be able to objectively assess the safety of new products of agricultural biotechnology. In July 1999, then Secretary of Agriculture, Dan Glickman, publicly expressed concern over this situation, which later resulted in a request for the National Academy of Sciences (NAS) to examine the scientific basis for, and the operation of APHIS regulatory oversight, to ensure that the commercialization of engineered plants is appropriately regulated.
Previous NAS committees have examined a number of issues related to the safety of genetically engineered organisms (NRC 1984, 1987, 2000c), but none specifically examined how the commercial use of all genetically engineered crops could affect agricultural and nonagricultural environments. The task set before this committee specifically included provision of guidance for assessment of the cumulative effects of commercialization of engineered crops on agricultural and nonagricultural environments. Therefore, in this report the committee examines potential effects on the environment that could result from the use of engineered crops on large spatial scales and over many years. In addition to evaluating the potential environmental impacts of single engineered traits in existing agricultural systems, the committee also examines how commercialization of engineered crops with single and multiple traits could actually change farming and thereby impact agricultural and nonagricultural landscapes of the United States. In this report the committee uses the current ecological and risk assessment literature in developing what it finds to be an appropriate framework for assessing the environmental effects of genetically engineered products, and then uses this framework to evaluate APHIS’s regulatory process.
The remainder of this introductory chapter presents background information on a set of topics that must be understood before a realistic environmental risk assessment framework can be developed. First, historic evidence is examined of how changes in agricultural technologies and practices have affected surrounding habitats so that readers can gain a sense of the extent of possible interactions. There has been much debate about the potential for genetic modification of crops to cause environmental impacts of a magnitude similar to that caused by the introduction of completely new species. Therefore, the next section of this chapter presents information on the history of environmental effects of conventional crop breeding compared to that of introduced species and also examines the hypothesis that the degree of environmental risk is related to the number of genetic changes introduced into an ecosystem. Next, an in-depth assessment is presented of predictable and unpredictable aspects of both conventional and transgenic processes used to add novel genes to plants. An understanding of the differences and similarities between these methods should give readers a basis for judging how different the side effects of the genetic engineering process can be compared to the side-effects of traditional processes of crop improvement with which we now live. The chapter ends with a brief description of the U.S. Coordinated Framework for the Regulation of Biotechnology and the USDA’s role and authority in regulating transgenic crops.
This introductory chapter leads to six detailed chapters. Chapter 2 uses the general principles of ecology and risk analysis to develop a frame-
work that addresses many of the concerns of the public and the scientific community regarding biological risks associated with commercializing genetically engineered plants. Chapters 3 and 4 provide in-depth reviews of how the USDA-APHIS regulates genetically engineered plants, and Chapter 5 assesses how well the USDA-APHIS approach functions and how well it matches the framework described in Chapter 2. Chapter 5 ends with recommendations for specific improvements in the APHIS precommercialization process. One general conclusion from Chapters 1 to 5, and from other published studies is that it will be impossible to assess some types of environmental effects of genetically modified plants based on the small-scale field testing that can be conducted prior to commercialization. Therefore, Chapter 6 examines the prospects and problems associated with developing post-commercialization monitoring programs to detect and measure such effects. The potential environmental impacts of future products of genetic engineering may differ from those of the engineered crops that have recently been commercialized. Therefore, the final chapter of this report (Chapter 7) is devoted to examining potential future products of genetic engineering and to how sets of novel traits in crops could alter the use of land, chemicals, and other resources. This concluding chapter also discusses how the public’s view of agriculture and its regulation may evolve in the future.
ENVIRONMENTAL EFFECTS OF AGROECOSYSTEMS ON SURROUNDING ECOSYSTEMS
In recent decades agriculture has been intensified by increases in the use of mechanization, irrigation, high-yielding crops, synthetic fertilizers, and pesticides. This has led to major changes in the structure, function, management, and purposes of agroecosystems (Swift and Anderson 1993, Swift et al. 1995, Matson et al. 1997). Typical structural changes include large reductions in plant, animal, and microbial biodiversity (Swift and Anderson 1993, Lacher et al. 1999) and simplified patterns of abiotic resource stocks and flows (Swift et al. 1995, Matson et al. 1997, Vitousek et al. 1997). Intensified agroecosystems are now predominant in developed countries and in highly capitalized export and commodity production of developing countries. In these latter countries there has also been expansion of less capitalized, smallholder agriculture onto marginal and ecologically sensitive lands, again typically accompanied by reductions in biodiversity and simplification of resource stocks and flows (Holloway 1991, Swift et al. 1995, Perfecto et al. 1996).
These changes have increased the environmental effects of agroecosystems on neighboring ecosystems, relative to those exerted by preintensified agroecosystems. These environmental effects are primarily
driven by outflows of energy, organisms, and materials from agroecosystems. Materials include resources such as nitrogen (Vitousek et al. 1997) and other substances such as pesticides (Carroll 1990, Soule et al. 1990).
Moreover, changes in agricultural land-use patterns appear to have increased the effects of these outflows on neighboring ecosystems. The most important landscape effect of agricultural development is great reductions in the area of nonagricultural ecosystems (Pimentel et al. 1992, Dale et al. 2000). Today, approximately 50 percent of U.S. land is used for crop and animal production. In agriculture-dominated landscapes, nonagricultural ecosystems have come to function as more or less isolated islands in an ocean of agricultural lands. In Western Europe and North America this habitat loss (and to a lesser degree, habitat fragmentation) has caused reductions in the diversity of birds, small mammals, and insects (Fahrig 1997, Trzcinski et al. 1999). In addition, the habitats that serve as a buffer or interface between agroecosystems and other ecosystems have commonly been degraded or eliminated (Carroll 1990, Risser 1995), as is the case when field borders are extended to the edge of streams or other wetlands. This landscape fragmentation and the degradation of interfaces facilitate the flow and penetration of energy, materials, and organisms into ecosystems adjacent to agroecosystems (Risser 1995).
The specific effects of agroecosystems on surrounding ecosystems have often been considered in isolation from each other (e.g., studies on changes in water quality or wildlife abundance). Now, a growing body of evidence suggests a strong need for a more integrative view of these effects (Carroll 1990, Carpenter and Cottingham 1997, Lacher et al. 1999). Specifically, there is growing evidence (Vitousek et al. 1997, Tilman 1999, Mack et al. 2000) that ecological effects of agroecosystems compound to exert simplifying and destabilizing effects on neighboring ecosystems. These effects are potentially of concern because they appear to weaken or destroy the neighboring ecosystem’s capacity for resilience—that is, an ecosystem’s ability to return to its original ecological structure and function despite disturbance (Carpenter and Cottingham 1997, Ludwig et al. 1997, Tilman 1999, McCann 2000).
For example, agricultural land-use changes and phosphorus enrichment combine to destabilize temperate lake ecosystems by degrading a number of negative feedback mechanisms that collectively provide ecosystem resilience to lakes (Carpenter and Cottingham 1997). In healthy lakes, floating microscopic plants called phytoplankton are generally held in check by limited nutrients and the microscopic predators (zooplankton) that feed on them. Increased inflows of phosphorus can result from greater agricultural use, increased soil erosion, and loss of riparian (i.e., water-edge) vegetation and wetland interfaces adjoining agroecosystems
or lakes. The loss of riparian vegetation and wetlands also decreases the inflow of humic substances that typically limit phytoplankton growth by effects on light and temperature in the lakes. Populations of large predatory fishes are often reduced by the biotic and physical changes in lakes, or by overfishing. Reductions in these predators allow zooplankton-feeding fishes to become more abundant. These changes cause reductions in the zooplankton and consequent increases in phytoplankton species, which outcompete larger aquatic plants. The lake food web is further simplified by reductions in crucial habitat provided by large aquatic plants and trees that fall into the lake from riparian areas. The losses of humic substances, adjoining wetlands and riparian forests, large predatory fishes, zooplankton, and large aquatic plants combine to remove a complex of negative feedback factors that act to buffer the lake’s response to disturbances and environmental change. As a result, the lake’s conditions and attributes become more variable and less desirable to people, as dense algal blooms with associated anoxia and algal toxins become more frequent. The lake also becomes more susceptible to additional disturbances, such as those caused by toxic pollutants or species invasions (Carpenter and Cottingham 1997).
As this example suggests, loss of resilience capacity can place ecosystems in a “poised” or “at-risk” position, increasing the likelihood of further change in ecosystem structure and function (Perry 1995, De Leo and Levin 1997, Ludwig et al. 1997, Gunderson 2000). In addition to affecting the aquatic systems discussed above (Nepsted et al. 1999, Gunderson 2000), intensified agriculture has been shown to reduce resilience of neighboring wetlands, rangelands, and even sections of the Amazonian rain-forest (Nepsted et al. 1999, Gunderson 2000).
The committee cannot presently judge whether extensive commercialization of transgenics—and other crops bearing novel traits—will significantly perturb agroecosystems or neighboring ecosystems because of major gaps in our knowledge of these systems. Below, the committee reviews some of the evidence of current agroecosystem effects on neighboring ecosystems as a means of conveying the extent and complexity of potential effects.
Flows of Materials and Organisms from Agroecosystems
Generally, simplified agroecosystems generate greater outflows of materials and organisms than less intensified systems (Swift et al. 1995). These agroecosystems feature both ecosystem processes (e.g., nutrient dynamics, water use) and organism population densities that vary widely in time because they are driven by temporal synchronization of activities of a few dominant plant, animal, and microbial species (Carroll 1990,
Haycock et al. 1993). Peak outflows of materials and organisms are associated with extremes of this pattern of wide variation in activities and abundances. For example, high-yield monocultural cropping systems receive large inputs of mobile nutrients to meet a single period of intense nutrient demand by the developing crop (Robertson 1997). At other times, crop nutrient uptake is modest, and uptake by other plant species is limited by stringent weed management and the absence of plant cover during fallow periods (Swift and Anderson 1993, Robertson 1997). Typically, a significant fraction of total nutrient inputs is not used by the crop, and this residual nutrient pool becomes available for outflow to surrounding systems (Robertson 1997). Outflow commonly occurs by wind or water vectors, and resources also can be carried by organisms, sometimes resulting in highly focused or long-distance transport of materials. For example, in New Mexico, geese feeding on agricultural fields vector large quantities of nitrogen into their roosting areas, which are managed wetlands (Jefferies 2000).
Effects of Outflows of Materials and Organisms on Neighboring Ecosystems
When agroecosystems discharge large flows of resources into surrounding systems, strong ecosystem effects can result. For example, outflows of nitrogen can cause major changes in plant communities, as species that were adapted to low nitrogen are replaced by a less diverse group of species (often mainly composed of exotic invasive species) adapted to higher nitrogen levels (Jefferies and Maron 1997, Vitousek et al. 1997). Plant community changes and resultant changes in landscape-level biodiversity, caused in large part by nitrogen deposition from intensified agriculture, have been particularly dramatic in the Netherlands (Aerts and Berendse 1988), where nitrogen deposition rates are among the highest observed rates globally (Vitousek et al. 1997). High levels of nitrogen deposition have been associated with serious degradation in forest ecosystems, including high rates of tree mortality and problematic soil acidification (Draaijers et al. 1989). Such nitrogen-related reductions in plant diversity appear to reduce ecosystem resilience. In experiments on temperate North American grasslands, nitrogen enrichment caused species loss, and grassland plant communities that had lost species were much less resilient to rainfall and climate variation than more diverse communities that were not nitrogen-enriched (Tilman and Downing 1994, Tilman 1996). Nitrogen deposition also reduces soil quality and fertility in surrounding ecosystems and causes freshwater acidification and coastal zone eutrophication (Jefferies and Maron 1997, Vitousek et al. 1997).
Resource outflows from agroecosystems can also affect population and community dynamics in recipient ecosystems (Polis et al. 1997). One notable response to resource input is a so-called trophic cascade, which results in severe reductions in numbers or biomass of organisms that are positioned in a food web below a consumer of some resource that is flowing from an agroecosystem. For example, in northeast Europe, cormorants have increased greatly in both average abundance and variability in abundance due to increased fish populations that have resulted from nutrient flows from agroecosystems into surface waters. After reaching high levels of abundance, these cormorant populations have greatly reduced populations of plankton-consuming fish in other lakes (Jefferies 2000). Also, a “paradox of enrichment” (Polis et al. 1997, Jefferies 2000) has been recognized, in which large inputs of formerly scarce resources cause local extinction of organisms that formerly coexisted in a food web. Again, the net effect of these changes often appears to be in the direction of simplification and reduction in resilience in ecosystems experiencing inflows of resources originating from agroecosystems (McCann 2000).
Nutrient overenrichment from human activities is one of the major stresses affecting coastal ecosystems. There is increasing concern that in many areas of the world an oversupply of nutrients from multiple sources is having pervasive ecological effects on shallow coastal and estuarine areas. These effects include reduced light penetration, loss of aquatic habitat, harmful algal blooms, a decrease in dissolved oxygen (or hypoxia), and impacts on living resources. The largest zone of oxygen-depleted coastal waters in the United States, and the entire western Atlantic Ocean, is found in the northern Gulf of Mexico on the Louisiana-Texas continental shelf. The freshwater discharge and nutrient flux of the Mississippi River system influence this zone (Rabalais et al. 1999). A series of reports that provide an integrated assessment of this hypoxia zone in the Gulf of Mexico document one of the most widespread and substantial outflows of nitrogen from agricultural sources in the United States (Diaz and Solow 1999, Goolsby et al. 1999, Rabalais et al. 1999).
On average, some 1.6 million metric tons of total nitrogen flow from the Mississippi-Atchafalaya River basin annually. Of the various nitrogen sources (atmospheric deposition, groundwater discharge, and soil erosion), agricultural sources of nitrogen such as fertilizers and animal wastes constitute the majority of the annual flux. Nitrogen input from fertilizers into the river basin has increased sevenfold since the 1950s (Goolsby et al. 1999). The impact of these increases in nitrogen outflows (similar patterns are noted with phosphorus) has been linked to the eutrophication of coastal areas and the seasonal hypoxic zone in the Gulf of Mexico (Diaz and Solow 1999). Except in areas of natural upwelling, coastal hypoxia is
not a natural condition. The size of the zone fluctuates annually, but in the mid to late 1990s the area averaged 16,000 to 18,000 km2 (Rabalais et al. 1999).
There are significant ecosystem impacts that result from hypoxia. Hypoxia in the northern Gulf of Mexico has become one of the many factors that at present control or influence the population dynamics of the regions of many pelagic and benthic species. Hypoxia exerts its control at two levels: the primary impact is the loss of bottom and near-bottom habitat through seasonal depletion of oxygen levels. From all indications, few, if any, mobile organisms stay on bottoms that are hypoxic. This forced movement could increase population losses due to predation of species dependent on the seabed for their survival. But when hypoxia dissipates, mobile organisms return. The second major impact is the alteration of energy flows. During hypoxia, significant amounts of the ecosystem’s energy are shunted to microbial decomposition, which is the primary biological process that creates and maintains the hypoxia. In the northern Gulf of Mexico the major sources of this energy seem to be organic matter produced in the surface waters and benthic biomass (mostly small worms, snails, and clams; Diaz and Solow 1999). Energy flows between trophic levels (mediated largely by predator/prey interactions) are also altered, as seen in the population movements of fish (Azarovitz et al. 1979) and shrimp species (Gazey et al. 1982, Renaud 1986, Zimmerman et al. 1996).
In addition to outflows of abiotic resources, agricultural activities often support increases in populations of organisms that then enter surrounding ecosystems, where they can exert a wide variety of effects (Carroll 1990). For example, populations of white (or snow) geese in North America have increased greatly since the 1950s, apparently driven by the growth of high-input cropping systems in the United States (Lacher et al. 1999). These large geese populations have altered vegetation on a regional scale in the eastern Canadian arctic by destroying coastal vegetation near their nesting sites (Kerbes et al. 1990, Jefferies 2000). In this case, organism flows from an agroecosystem have had an effect over distances of 3,000 to 5,000 km.
Invasive organisms entering surrounding ecosystems from agroecosystems frequently have properties that increase their colonization success, creating positive feedbacks that amplify their effects on the ecosystems they enter (Mack et al. 2000). For example, certain invasive weeds enter tropical forests from pasture and field crop agroecosystems (Carroll 1990). In addition to being highly competitive with tree seedlings, these species often increase the frequency of fires in forests, entraining a cycle of increasing fires, reductions in forest biodiversity, increased abundance of the invasive species, and more fires. This pattern of invasion driven by
a positive feedback cycle appears to be the basis of the success of a variety of grasses derived from agroecosystems that have become major invaders worldwide (D’Antonio and Vitousek 1992).
Landscape-Level Effects of Agriculture
Agricultural land use has always tended to fragment landscapes relative to their pre-agricultural conditions. This pattern of fragmentation has increased the importance of edge effects (i.e., effects of one ecosystem on a second that act directly on only the periphery of the second ecosystem).
Populations of many organisms that are resident in nonagricultural lands, including predators (Lacher et al. 1999), can be rapidly depleted by edge effects. For example, populations of migratory forest birds in the United States have been strongly affected by generalist predators (e.g., blue jays, common crows) that are active in peripheral portions of forest fragments in agricultural landscapes (Wilcove et al. 1986). Very few of these fragments are extensive enough to create refuges from predation for these migratory bird populations. These edge-related depletion effects can be particularly significant when they reduce populations of predatory organisms, (e.g., on grazing lands in the western United States; Knowlton et al. 1999). These species are often strong determinants of ecosystem structure and function; in their absence, herbivorous species can increase sharply and quickly deplete vegetation, setting off a wave of ecosystem change.
Fragmentation of nonagricultural lands also affects dispersal and movement of organisms that reside in these lands. In Northwest Europe, red squirrels occur in small populations in forest patches in agricultural landscapes. These small populations are prone to becoming extinct in individual forest patches. After extinction in a particular patch, successful recolonization of the patch by squirrels depends on the distance to other patches and the degree to which patches are interconnected by woody vegetation along field borders. For these reasons, highly isolated patches will usually not be occupied by squirrels (Verboom and van Apeldoorn 1990). Similar effects of fragmentation on the biodiversity of birds, mammals, and insects in nonagricultural lands have been observed (Forman 1997).
Finding 1.1: There is substantial evidence that ecological effects of agroecosystems compound to exert simplifying and destabilizing effects on neighboring ecosystems. These effects are potentially of concern because they appear to weaken or destroy the neighboring ecosystems’ capacity for resilience—that is, an ecosystem’s ability to maintain a certain state despite disturbance.
Finding 1.2: Potential ecological effects of transgenic crops—and other crops bearing novel traits—may be substantially heightened in the simplified and destabilized ecological milieu that has resulted from agricultural intensification. This prospect is a strong argument for a cautious approach to the release of any crop that bears a novel trait into agroecosystems and the less managed surrounding ecosystems. Equally, it is an argument for a cautious approach to any extensive change in agricultural practices.
Recommendation 1.1: There is a need for a cautious approach when making any extensive change in agricultural practices, including changes in the genetics of crops, because of potential ecological impacts on agricultural and surrounding ecosystems.
ENVIRONMENTAL IMPACTS OF THE DELIBERATE INTRODUCTION OF BIOLOGICAL NOVELTY: FROM GENES TO MINICOMMUNITIES
The practice of modern agriculture introduces biological novelty into the environment for a number of purposes. Conventional plant breeders introduce novel genes into crops to improve yield, taste, agronomic traits, and other characteristics. Pest management specialists often import natural enemies of pest species into the United States from other continents. And horticulturalists are continually seeking new varieties and species of plants that could enhance the aesthetic beauty of urban and rural landscapes.
The biological novelty introduced to agricultural systems by these varied activities can result from very small to very large changes in genetic information relative to preexisting organisms. The general degree of change in genetic information can be measured along two axes: the number of genetic changes and the taxonomic or phylogenetic distance between the source and the recipient of the new genetic information.
Biological novelty within a species’ genome can result from a change in a single DNA base pair to more complex changes affecting the constitution of one or more chromosomes. The introduction of new individuals to an ecosystem is another source of biological novelty. Adding an individual of a preexisting species to a given environment introduces biological novelty if that immigrant is genetically different from those already established in that population. The genetic difference can be as small as one novel allele (i.e., one form of a gene) or as large as an individual that represents a different race or subspecies. The introduction of an individual of a new species with preexisting close relatives represents a greater
addition in genetic information relative to the prior example. That species may hybridize with those close relatives, and the resulting hybrids represent another type of biological novelty (Abbott 1992). The most substantial addition of genetic information at the individual level would be the introduction or immigration of a species with no close relatives. Of course, an introduced species may arrive bearing other novel organisms in the form of mutualists, parasites, and other symbionts, effectively representing the arrival of a minicommunity.
Within the continuum from minor to major alterations in the genetic information in agroecosystems, the practice of conventional crop breeding has typically been considered to result in minor genetic changes and is not subject to formal regulatory review. On the other hand, the introduction of new species used for biological control, or as new horticultural plants, is considered to involve greater changes in genetic information, and these introductions have been subject to review. Some participants in the debate over genetically engineered crops compare them to the products of conventional breeding because of the small number of genetic changes involved and have concluded that the need for oversight is minimal. Others compare genetically engineered crops to the introduction of new species because the new genetic information often originates from taxonomically distant species. They call for stringent regulation. Both of these perspectives make the assumption that there is a relationship between the nature of biological novelty (number of genetic changes in the first example and phylogenetic distance in the second) and the risk of negative environmental impacts.
The remainder of this section examines the literature on the history of environmental impacts of conventional crop breeding and of species introductions to determine if the extent of genetic change or other general factors can be used as predictors of risk. While the available studies do not permit a precise quantitative comparison, they are informative and suggest that (1) changes at any level of genetic information can have profound environmental consequences, (2) the consequences of biotic novelty depend strongly on the specific environment into which they are released, (3) the significance of the consequences of biotic novelty depend on societal values, (4) the introduction of any type of biological novelty can have unintended and unpredicted effects on the recipient community and ecosystem, (5) it is not possible to qualitatively differentiate the general environmental risk associated with the release of conventionally bred crop cultivars and the introduction of new species.
Small or large genetic changes can have profound environmental consequences. The introduction of a single gene for short plant stature in rice (see BOX 1.1) enabled major changes in agricultural practices that resulted in much higher yields and substantial stresses on agricultural and
nonagricultural habitats. The introduction of T male-sterile cytoplasm into U.S. maize crops (a very small genetic change relative to the maize genome) permitted the rapid spread of Bipolaris maydis, the disease organism causing Southern corn leaf blight (NRC 1972). A single allelic replacement in rose clover (Trifolium hirtum) appears to account for a tremendous increase in its colonizing ability (Jain and Martins 1979), and only a few allelic differences appear to be responsible for weediness in Johnsongrass (Sorghum halepense), one of the world’s worst weeds (Paterson et al. 1995). Honeybees with European ancestors in the New World radically expanded their niche in the tropics after human-mediated, but unintended, introduction of a handful of genetic alleles from an African strain (essentially hybridization between subspecies; Camazine and Morse 1988). Natural hybridization between the introduced New World cordgrass Spartina alterniflora and the native S. maritima in Britain gave rise to S. anglica, an invasive species that rapidly altered the coastal ecosystems of the British Isles (Gray et al. 1991, Thompson 1991). Finally, some of North America’s worst exotic invasive plants have close indigenous relatives (purple loosestrife, Lythrum salicaria) and others do not (garlic mustard, Alliaria petiolata; kudzu, Pueraria montana). Clearly, there is potential for small and large changes in genetic information to result in environmental impacts.
In addition to attempts at explaining risk based on the number of introduced genes, studies have examined whether there is a relationship between the taxonomic or phylogenetic distance between the gene donor and recipient and the level of environmental risk. If level of risk increases with the phylogenetic distance between organisms exchanging genes, we might expect to find that documented cases of natural interkingdom horizontal gene transfer (the nonsexual transfer of the genetic material from one organism into the genome of another) correlating with a dramtic ecological change in the recipient organism. That does not appear to be the case (Palmer et al. 2000).
The consequences of biotic novelty depend strongly on the specific environment, including the genomic environment, physical environment, and biotic environment. “Experience with exotics shows overwhelmingly that an organism’s effect and ecological role can change in new environments” (U.S. Congress, Office of Technology Assessment 1993). And these effects and roles of a newly introduced genotype can change with time. Differences in physical environment can make a big difference. Abutilon theophrasti, native to Asia, is an important weed in much of the temperate world. In California it is only a weed where available moisture from rain or irrigation will allow its growth (Holt and Boose 2000).
Responses to the physical environment may vary with relatively small genetic differences. In Florida the South American tree Schinus terebinthi-
folius is a noxious invader (U.S. Congress, Office of Technology Assessment 1993). But the closely related Schinus molle is much more invasive in California’s natural ecosystems than S. terebinthifolius (Hickman 1993). The differences in invasiveness between the two species in the two locations appear to depend on differences in their germination and growth responses in the different physical environments (Nilsen and Muller 1980a, 1980b). While more subtle, the biotic environment may play the largest role in determining the impact of biological novelty. An obvious example is that many crops initially benefit by being grown in an appropriate abiotic environment that is far from their place of origin. But that benefit eventually erodes as local pests adapt to the new species or as the original pests arrive via unintended human-mediated dispersal (e.g., Strong 1974, Strong et al. 1977, Andow and Imura 1994).
Interactions among genes can be viewed as an internal environmental component. For example, no matter how many alleles a mammal possesses for increased melanin production, the predicted phenotype based on those loci will not be expressed if that individual is homozygous at another locus for alleles that cause albinism (Mange and Mange 1990). If a gene for pest resistance is added to a plant that lacks genes for reproduction without human intervention, it is unlikely to become weedy.
The significance of the consequences of biotic novelty depend on societal values, whether that novelty represents new genotypes or new species. Ornamental mutants of native American wildflowers are the darlings of those who buy them to grow in their backyards and the bane of restorationists who find them in their seed mix (Montalvo et al., in press). While Africanized honeybees create concern in the New World when they kill humans and livestock, prudent management of the bees has resulted in better honey yields in Brazil compared to the European genotypes of the same species (e.g., Camazine and Morse 1988). Monterey pine is an endangered species in its native range in California, but in Australia it is both valued as an important commercial forest tree and despised as an invader of native eucalypt forests (Burdon and Chilvers 1994). In many cases the majority of the public can agree qualitatively if not quantitatively on the desirability of a specific plant in a specific place, but it is the variation among the public in such value judgments that has made much of environmental regulation so contentious.
Introduction of biological novelty can have unintended and unpredicted effects on the recipient community and ecosystem. In the past few decades the planting of new varieties of maize in Mexico that require the tools of modern industrial agriculture has rendered such fields inhospitable to several native taxa, most notably teosinte, the progenitor of maize (Sánchez and Ruiz Corral 1997, Wilkes 1997). This outcome would be considered desirable relative to the goal of maximizing yields. Also, natu-
ral hybridization with cultivated rice has been implicated in the near extinction of the endemic Taiwanese wild rice, O. rufipogon ssp. formosana (Kiang et al. 1979). During the twentieth century, the cultivation of domesticated rice steadily increased on the island of Taiwan. During the same time, collections of O. rufipogon ssp. formosana showed a progressive shift toward characters of the cultivated species and a coincidental decrease in fertility of seed and pollen. By 1979 naturally occurring populations of this subspecies were on the verge of extinction. Similar substantial effects can occur for successfully established exotic species. The spectacular roosts of overwintering and migrating monarch butterflies along California’s coast are on eucalyptus, introduced from Australia in the last 200 years (prior butterfly roosting sites in California are unknown; Malcolm and Zalucki 1993). Another example is the introduction and establishment of European grassland species to California, which have altered natural fire regimes of its inland valleys, resulting in the conversion of native shrublands to European grasslands (Minnich 1998).
There is no strict dichotomy between environmental risk associated with releases of conventionally bred crop cultivars and introduction of new species. The majority of introduced cultivars and species do not result in long-term environmental establishment. Most registered crop varieties fail to become popular with farmers, and thus do not persist in agroecosystems. The half-life of accepted varieties is on the order of about five years. In the same way, on the order of 10% of intentionally introduced species persist after introduction (Williamson 1993). From this fraction of species that persist, very roughly 10% become an obvious problem in agricultural or non-agricultural environments (Williamson 1993). In sum then, only approximately 1% of species introductions are problematic.
The small fraction of exotics that do cause environmental effects can be tremendously disruptive and have received considerable recent attention (e.g., Pimentel et al. 2000). It is not clear what fraction of introduced crop and horticultural species that naturalize also become pests or otherwise create hardship. Although the fraction is low, it is not zero. Many plants intentionally introduced for agricultural purposes have become noxious weeds (e.g., see review by Williams 1980). A good example is bermudagrass, Cynodon dactylon, which is frequently introduced as a forage or as turfgrass but has also become one of the world’s worst weeds (Holm et al. 1977). Additionally, the natural hybridization of crops with wild relatives has led to the evolution of both new agricultural weeds and noxious invaders of natural ecosystems (see examples in Ellstrand et al. 1999, Ellstrand and Schierenbeck 2000). Hybridization with crops has also increased the risk of extinction of some native plants (as in the case of O. rufipogon ssp. formosana discussed above; see also examples in Ellstrand et al. 1999).
But are there any cases where the introduction of a new agricultural genotype to a previously extant species has resulted in economic or environmental change? The most dramatic recorded example is the development of semidwarf varieties of wheat and rice (see BOX 1.1). Other examples also illustrate that the introduction of a new crop genotype can have environmental impacts. As noted above, the introduction of modern maize genotypes in Mexico that have replaced traditional maize varieties require agronomic techniques that displace teosinte, leading to increased risk of its extinction. The problems created by new genotypes also can be agronomic, as illustrated in the case of T cytoplasm in maize mentioned above.
The introduction of cytoplasmic male sterile (CMS) genotypes of sugar beet had an indirect consequence that has been economically disruptive to Europe’s sugar industry. CMS genotypes are used to create hybrid varieties in this wind-pollinated, outcrossed species (Ford-Lloyd 1995). Natural pollination from wild sea beets growing near the seed production fields in both southern France and near the Adriatic Coast in Italy contaminates the seed with hybrids (Boudry et al. 1994, Mücher et al. 2000). The hybrids are annuals that bolt, resulting in a woody root that yields very little sugar and damages both harvesting and processing machinery. Before they bolt, the weeds are essentially identical to the crop and are therefore difficult to control (Longden 1993). If they are not removed before they set seed, their seed contaminates the soil, frustrating attempts at growing sugar beet for several years. The cumulative cost of three decades of weed beets is at least on the level of hundreds of millions of dollars. Presently there are some sugar beet fields in Europe that produce more weed beets than sugar beets (Mücher et al. 2000).
In summary, the vast majority of crop genotype and species introductions do not persist in the environment and cause environmental damage. A minority causes hardship to humans, agronomic ecosystems, and/or natural systems. The examples reviewed here indicate that general information on the amount or origin of genetic material cannot, on its own, be used to predict the risk associated with a new crop genotype or introduced species. In this regard the committee’s findings support those of other scientists who have examined this problem of predicting risk and concluded that risk assessment cannot depend on general characteristics such as the amount of new genetic information introduced but must focus on the ecology of the specific introduced organism (or both the donor and recipient in the case of transgenic organisms) and the characteristics of the accessible environment into which the organism will be released (e.g., NRC 1987, Tiedje et al. 1989, Scientists Working Group on Biosafety 1998). This does not mean that each introduction requires an equally intensive assessment. A number of scientific panels have developed outlines and
BOX 1.1 The Green Revolution
The goal of the Green Revolution was to increase crop yields to feed existing and growing populations in developing countries (Hanson et al. 1982, Conway 1998). Conventional plant-breeding technologies provided the crucial crop cultivars that allowed this revolution in agricultural production to begin. Prior to 1950, tropical cultivars of rice and wheat were about 1.5 to 2.0 times as tall as present-day cultivars. When high rates of fertilizer were applied to these tall cultivars, yields did not rise substantially and the plants tended to lodge (fall over). Lodging caused decreased yield, and made harvest difficult. A few genes were found in temperate zone cultivars of rice and wheat that resulted in what were called “semidwarf” plants (Hanson et al. 1982, Dalrymple 1986). When high rates of fertilizer were applied to these plant types, yield increased dramatically. When these semidwarf genes were transferred to tropical rice and wheat, they too showed very positive responses to fertilization. Thus, these semidwarf genotypes enabled a technology package to be assembled that led to revolutionary increases in yield in many developing countries.
Most semidwarf tropical rice cultivars derive their short stature from a single gene, sd-1 (Dalrymple 1986), and therefore all of these cultivars presumably carried many of the genes that were tightly linked to the sd-1 gene in the donor plant. Some of the early semidwarf cultivars were more heavily attacked by pathogens than traditional cultivars, so genes for disease resistance were transferred from traditional cultivars to what became known as high-yielding varieties or HYVs. With the financial backing of the international community and governments of some developing countries, rates of farmer adoption of these semidwarf HYVs were sometimes very rapid. In an extreme case, adoption of HYVs in the Philippines went from 0 to 50% in five years (Evenson and David 1993).
The semidwarf genes were typically transferred to tropical and subtropical cultivars of rice and wheat along with other non-linked genes that decreased time to harvest (Khush 1995). One of the impacts of these short-season HYVs was the potential, in some areas, for growing two or three crops of rice on the same land in a single year (Greenland 1997). In other areas farmers began to grow one crop of rice and one of wheat in the same year (Hanson et al. 1982). To extract high yields from these fast-growing HYVs, more fertilizer and water were needed. In many areas pest problems appeared to increase, possibly due to the continual availability of the crop in large monocultures (Greenland 1997). These pest problems were often responded to with application of pesticides (Gallagher et al. 1994, Matteson 2000).
As discussed in the text, runoff of fertilizer into wetlands can have major environmental impacts. The extent of these impacts in developing countries has not been measured carefully over time. In many tropical rice ecosystems nitrogen and phosphorus are typically not expected to cause eutrophication because of absorption of these nutrients to soil particles or loss through volatilization (Greenland 1997), but impacts on wetlands have been found (Ghosh and Bhat 1998). Additionally, a study conducted in the Philippines indicated that nitrate nitrogen had leached through the root zone toward the groundwater table. The researchers concluded that over time this leaching could contaminate ground
water (IRRI 1995). In temperate zone areas of Japan, which have the highest rates of fertilizer use in rice, impacts on wetlands have often been reported (Mishima 2001). Although some short-season rice cultivars can be produced with less water than long-season cultivars, it is difficult to raise yields without more water (Greenland 1997), and when an additional crop is grown each year, water resources can be stretched thin. In production areas not adjacent to river deltas, additional water was generally acquired by increased extraction of subsurface water (Bandara 1977). This has resulted in lowering of the water table in many areas (Bandara 1977, Greenland 1997). For example, in the Punjab area of India the groundwater table has been receding at a rate of 20 cm per year, and it has been proposed that rice cultivation in this area be limited (IRRI 1995). In some arid areas, salinization of water in the rice root zone has become more problematic (Greenland 1997).
Increased use of pesticides has sometimes been successful in controlling insects, but often insecticide use has reduced natural enemies of pest species and triggered worse outbreaks. This resulted in a classical “pesticide treadmill” (Gallagher et al. 1994, Matteson 2000). A few studies have documented the impacts of increased pesticide use on rice agricultural ecosystems (Gallagher et al. 1994, Heong and Schoenly 1998, Matteson 2000) and the effects of pesticide runoff on nonagricultural systems (Nohara and Iwakuma 1996).
As part of the research agenda accompanying the Green Revolution, long-term yield experiments were initiated on research stations to determine if double and triple cropping of rice was sustainable. A 1979 analysis of yields from a triple-cropping experiment begun in 1963 at the International Rice Research Institute showed a pattern of decreased yield (Ponnamperuma 1979). Yield continued to decrease at an annual rate of 1.4 to 2.0% through 1991, when yields were 38 to 58% lower than in the baseline year of 1968 (Cassman et al. 1995). Changes in agricultural practices were able to reverse this yield decline, but the specific causes for yield decline and its reversal are not well understood (Dobermann et al. 2000). A more broad-based analysis of 30 long-term experiments in a diverse set of rice-growing regions found statistically significant yield declines in eight experiments (Dawe et al. 2000). Similar declines have not been seen in farmers’ fields, and it is thought that the yield decline may only result when the production system is pushed to nearly 100% of the yield potential and soils rarely go through a drying period (Dawe et al. 2000). There are a number of lessons from these soil fertility studies. One is that these long-term trends could not have been revealed by precommercialization testing (see Chapter 6). A second lesson is that effects on the environment are typically complex, and it is often difficult to determine a specific cause of an observed event. A third lesson is that the monitored environments should closely reflect the diverse environmental conditions in commercial fields.
Because the Green Revolution involved many technical changes, it is sometimes difficult to establish a direct cause-and-effect relationship between the few genes that started the Green Revolution, and resulting environmental changes. However, the Green Revolution clearly demonstrates that commercialization of cultivars with relatively simple genetic changes can have major effects on farming practices that ultimately result in environmental change.
flowcharts for guiding the direction of risk assessments (e.g., Tiedje et al. 1989, Scientists Working Group on Biosafety 1998). With information on the ecology of the organism and the environment in hand, it is possible to quickly rule out some introductions as too risky and to determine that others are unlikely to pose any risks. More importantly, such information can be used to flag cases where only further testing will provide sufficiently accurate information on potential risks. Chapter 2 discusses how a regulatory system can be developed to efficiently utilize this approach. From the above assessment of impacts of adding biological novelty to ecosystems the committee has determined the following:
Finding 1.3: A comparison of environmental impacts from adding biological novelty to ecosystems indicates that:
small and large genetic changes have had substantial environmental consequences;
the consequences of biological novelty depend strongly on the specific environment, including the genomic, physical, and biological environments into which they are introduced;
the significance of the consequences of biological novelty depend on societal values;
introduction of biological novelty can have unintended and unpredicted effects on the recipient community and ecosystem;
a priori there is no strict dichotomy between the possibility of environmental hazard being associated with releases of cultivated plants with novel traits and the introduction of nonindigenous plant species. However, the highly domesticated characteristics of some cultivated plants decrease the potential of certain hazards.
COMPARING AND CONTRASTING CONVENTIONAL AND TRANSGENIC APPROACHES TO CROP IMPROVEMENT
In debates regarding the need for formal assessment of environmental impacts of transgenic plants, concerns about risks that result from the process of plant transformation and risks associated with the intended products of transgenically-based crop improvement are often not clearly differentiated. Crops produced by transgenic approaches could have risk potential because a novel gene in the transgenic crops makes an intended product with insecticidal properties that affect non-target organisms. In contrast, a transgenic crop could have unanticipated environmental effects if during the transgenesis process some of the normal plant genes were unintentionally disrupted, leading to a change in the products of those genes. In the following section, risks from the genetic processes
involved in traditional and modern conventional crop improvement are compared and contrasted with those used in transgenic-based approaches. Other chapters of this report discuss the risks associated with intended products of conventional and transgenic methods.
Traditional and Conventional Processes of Crop Improvement
Humans have been using genetic modification to improve crop plants for thousands of years. Their first experiments consisted of identifying various ecotypes or genetic variants and comparing their ability to produce a good crop when cultivated. Selections (genotypes) that performed well were saved and propagated, while inferior ones did not survive or were simply discarded. These early farmers discovered that crop yields could be increased by modifying the environment. Consequently, they developed methods of irrigating and administering soil fertilizers. Over millennia they learned to modify an important relationship that influences the productivity of all crop species—the interaction between the genotype of the plant and the environment in which it is grown (Simmonds and Smartt 1999).
The types of traits (see BOX 1.2) these early farmers searched for were similar to what is considered important today: higher yield; uniformity in germination, growth and development, flowering time, and shattering (seed dispersal); disease and drought tolerance; resistance to insects and pathogens; and improved functional characteristics of the products used for food (i.e., sugar, starch, protein, oil content, etc.). They had to rely on the genetic diversity that existed in a given ecotype as a consequence of naturally occurring mutations. No doubt there was some degree of allelic diversity among the plants grown, but early farmers had no easy way to identify it nor to deliberately select for the responsible loci. Consequently, their ability to genetically improve crops was limited to the existing genotypic variability and their capacity to discern the consequent phenotypes and select for them. As knowledge of breeding techniques developed, it became possible to improve traits in crop varieties through genetic recombination, which can bring together favorable mutations, genes, or blocks of genes and thereby create more useful varieties (Simmonds and Smartt 1999).
As discussed earlier, the genetic basis for an altered crop trait can simply involve a single gene or can be complex, involving multiple, potentially interacting genes. Simply inherited traits are generally easy for plant breeders to manipulate because they usually show dominant or recessive phenotypes and produce simple ratios of progeny with and without the trait in segregating crosses. Simply inherited traits are often controlled by key enzymes in biochemical pathways. A good example is
BOX 1.2 Traits and Characters
In the conventional genetic sense, a character is a feature, such as flower color, while a trait is a particular form of a character, such as red in the case of a rose. For this example, the rose would have a red phenotype or appearance, and this color is determined by the nature of the genes or alleles (forms of a gene at a particular chromosomal locus) that determine the red trait. The set of genes determining the red color is the genotype of this trait. Informally, the term trait is often broadly used for many of these terms, and this leads to confusion. Genes give rise to traits, but they are not the traits themselves.
The alternate states of a character are the traits associated with that character. For example, an individual plant that is resistant to glyphosate (the herbicide Roundup®) has the trait or phenotype of glyphosate-resistance. This trait is conferred on an individual by the gene for 5-enolpyruvyl shikimate-3-phosphate synthase (ESPS), which is of bacterial origin. The genotype of that individual refers to genes that it has at the location in the genome where the ESPS gene is found. The character, resistance to glyphosate, is the set of two traits, glyphosate-resistance and glyphosate-susceptibility.
A character can be narrowly or broadly defined. For example, insect resistance is a plant character that includes resistance in any plant part (roots, stems, leaves, flowers, etc.) to any insect species (caterpillars, beetles, etc.). This character could also be narrowly defined as the resistance to stem-boring Lepidoptera (moths and butterflies), which restricts the range of insect species under consideration. The insect resistance character could encompass traits such as leaf and stem toughness, production of toxic chemicals and growth inhibitors, and modifications in leaf and stem morphology, etc. We also could further narrow the scope of this character to toxin-based resistance in maize, which would focus it primarily on traits involving DIMBOA (a toxin produced naturally in corn) and various transgenic traits based on the production of Cry toxins from Bacillus thuringiensis. We also could define a character as Cry toxin-based resistance in maize to stem borers. Alternatively we could define a character based on the expression of a protein, so the character could be Cry toxin production in corn. Even this character could be defined
starch synthesis, which is partially controlled by the enzyme ADP-glucose pyrophosphorylase. In maize this enzyme is composed of two protein subunits encoded by genes that have been given the names Shrunken2 and Brittle2 (Hannah 1997). If either of these genes is defective, the result is a smaller seed with reduced starch content. Traits with more complex genetic control, often called quantitative traits, are controlled by many genes that interact with the environment (genotype by environment interaction) and with each other (epistasis) in producing a plant trait (phenotype). Thus, knowledge of each gene is necessary but not sufficient to predict the final plant traits in a specific environment (Lewontin 2000).
more narrowly: transgenic maize with Cry toxins effective against stem borers includes four different proteins, Cry1Ab, Cry1Ac, Cry1F, and Cry9C (see also BOX 2.1), each of which could be defined as a separate character, because they are different proteins and they affect stem borers in different ways. Indeed, even this character can be defined more narrowly. Resistance to late generation stem borers in the various Cry1Ab transformation events is variable. Thus, the character can be defined as full-season Cry1Ab resistance to stem borers in maize, which would include the transformation events Bt-11 and Mon 810; another character, partial late season resistance, would include the Event 176 transgene. A similar range of definitions exists for many other genetically engineered characters, including herbicide tolerance, delayed ripening, modified oil content, and virus resistance.
A clear definition of character facilitates scientific analysis and communication (see Chapter 2). For a genetic engineer, a specific transgene may be of interest, so a useful definition of character could be an immediate product of the transgene, such as RNA or protein. Use of this narrow definition could facilitate detection and development of products based on the transgene. For a farmer using the transgenic crop, resistance to stem-boring Lepidoptera may be a broad and sufficient definition of a useful character. Additional detail may be irrelevant to a farmer, but knowing which pest is controlled would be quite important.
From the perspective of a risk analyst, a character should be defined in a manner that enables establishment of scientifically rigorous comparisons (that is reference scenarios see Chapter 2). For example, using the broad definition—resistance to stem-boring Lepidoptera in maize—could result in concluding that environmental risks associated with Bt-maize are the same as risks associated with DIMBOA (that is, a conclusion of substantial equivalence, see Chapter 4), when the comparison itself may be indefensible scientifically. Instead, defining two characters more narrowly, Bt toxin expression, and DIMBOA expression in maize, would lead to a scientifically more rigorous environmental assessment. Thus the definition of character in a risk assessment should include scientific confirmation that the traits included in the definition are similar enough to allow scientifically valid comparison.
Yield is an example of a complex trait. It is easy to imagine that factors such as photosynthesis, water-use efficiency, disease resistance, flower development, and others all contribute to the accumulation of biomass and its deposition in seeds.
For sexually reproducing species, simply inherited or single gene traits are usually incorporated into existing elite cultivars by the process of backcrossing. Backcrossing is conducted by crossing the elite cultivar with a donor (source of the single gene) and is followed by repeated crossing (backcrossing) to the elite cultivar with the goal of recovering all of the genome of the elite cultivar plus the gene of interest. Backcrossing
is most useful when the phenotype of the trait can be unambiguously scored by visual, biochemical, or molecular methods. Complex or quantitative traits are typically integrated into cultivars by pedigree breeding. Pedigree breeding is initiated by crossing two lines or cultivars together that complement one another for the traits of interest. This initial cross is typically followed by a series of self-pollinating generations, with each generation evaluated in multiple environments for the traits of interest (Hallauer 1990).
Existence of genetic variation is critical for the genetic improvement of any crop. The ultimate source of genetic variation for both simple and complex traits is mutations. Mutations result from changes in DNA sequence, resulting from errors in replication and recombination, insertion of retrotransposons and unstable genetic elements (transposons)*, gene silencing, and environmental mutagens such as radiation. In recent years the molecular bases for some of the mutations used in crop development and improvement have been characterized. For example, high sugar content in sweet corn and peas results from mutations in genes encoding enzymes required for starch synthesis (Hannah 1997). Sticky rice, which is favored in East Asia, is also caused by a mutation affecting starch synthesis (Isshiki et al. 1998). One example of flower color variation in petunias results from a phenomenon called gene silencing, which occurred as a consequence of a duplicated gene sequence (Quideng and Jorgensen 1998). A large number of mutations, such as one influencing flower color in morning glories (Clegg and Durbin 2000) and another causing parthenocarpic fruit formation in apples (Yao et al. 2001), result from insertion of retrotransposons in genes. Retrotransposons and transposable genetic elements have had a dramatic effect on the size and evolution of plant genomes (Kalendar et al. 2000). A trait for male sterility in maize that was used for many years to produce hybrid corn resulted from a bizarre natural mutation in which part of the DNA in the mitochondrial genome was rearranged to create a novel protein-coding sequence using a
ribosomal RNA gene (Dewey et al. 1986). Normally, ribosomal RNA genes are not translated into protein.
While the nature of gene mutations can be simple or complex, it is not necessary for the plant breeder to understand their bases in order to use them in a breeding program. As long as the mutation creates a stable, heritable phenotype, it can be incorporated into a crop plant by conventional breeding methods. The challenge of improving a trait in a crop plant through breeding depends on whether the trait is simple or complex and the ease with which the phenotype can be measured. Many simple traits can be monitored visually or by measuring biochemical products, such as sugars, proteins, or lipids. For complex traits the end product, such as height, yield, or resistance to stress, is measured.
With sexually compatible species, a useful trait is transferred from one genetic background into a desirable so-called elite genotype by making a cross. In this process, not only are genes encoding the trait or traits of interest transmitted, but so are thousands of other genes, some of which can lead to inferior or undesirable traits. However, most of these inferior genes can be eliminated by making a series of recurrent backcrosses (usually six or more) to the elite parent in the case of simple traits or by selection of the best phenotypes in the case of complex traits. In this way, genes encoding the valuable trait are recovered in the nearly homogeneous elite genetic background. However, the process of backcrossing or recurrent selection cannot avoid introducing a large number of genes that are tightly linked with those encoding the trait or traits of interest. Even after 20 backcrosses there is a theoretical expectation that genes within 9 centimorgans (units of crossing over) of the selected gene will be carried in the new plant line (Naveira and Barbadilla 1992). This 9-centimorgan region could include over 100 genes.
For an empirical example, genes for insect and pathogen resistance are often transferred to cultivated tomatoes from related species. Young and Tanksley (1989) examined eight commercial tomato cultivars that contained the Tm-2 gene for resistance to tobacco mosaic virus. This resistance gene originated from the wild species Lycopersicon peruvianum and had been transferred to a variety of commercial tomato types (e.g., cherry tomatoes, large greenhouse tomatoes) by backcrossing for 5 to 20 generations. Using DNA markers specific to L. peruvianum, Young and Tanksley (1989) demonstrated that the introgressed DNA from L. peruvianum in the final cultivars encompasses a chromosomal region ranging from 4 centimorgans to over 51 centimorgans. In the extreme case, the cultivar Craigella-Tm-2 contained an entire arm of chromosome 9 from L. peruvianum.
The importance of these linked genes depends on how genetically different the original parents are that were used in the cross. These extra
genes, so-called genetic drag, may or may not have important phenotypic consequences for crop production. The plant breeder must, in the end, compare the overall performance of the backcross progeny with the original parent to determine if there has been an improvement in the crop. Sometimes there is no improvement, and it is generally difficult to predict how a given gene or set of genes will affect the overall performance of the elite genotype. Consequently, plant breeders maintain a continuous cycle of creating and testing various genetic combinations for those that improve phenotypic performance.
In some cases, plant breeders introgress completely new chromosomes into a crop from another species or genus, thereby introducing hundreds or thousands of new genes/genetic loci. For example, there are over 200 lines of wheat to which a chromosome of another species has been introduced (Islam and Shepherd 1991). Embryo rescue, which makes use of tissue culture techniques to propagate incompletely developed embryos, permits the great genetic distances between these species to be bridged when the interspecifically produced progeny would not otherwise survive. For example, oats and corn can be crossed and partial hybrids recovered by embryo rescue. A hybrid zygote is formed, but the corn chromosomes are most often eliminated in the early divisions of the embryo; however, corn chromosomes are retained along with the haploid set of oat chromosomes in about a third of the progeny and the resulting plants are fertile (Riera-Lizarazu et al. 1996). Such crosses between exotic plants from germplasm collections or wide hybridization, such as between corn and oats, introduce genes not usually found in the same nucleus, even though these genes arise in the same or related genus. These techniques provide the opportunity to introduce novel and sometimes useful genetic traits. Most breeding programs make relatively little use of such potentially valuable genetic materials. However, as described below, the development of gene transformation technology greatly increases the ability of the plant breeder to utilize diverse genetic resources to a much greater extent than was previously possible.
Traditional plant breeding is an effective way to improve the genotype of elite varieties, but it is largely done in ignorance of the genes involved. The process of making sexual crosses, often followed by backcrossing to a preferred parental line, provides a mechanism to shuffle and then fix gene combinations that can subsequently be evaluated for trait performance. Backcrossing is particularly important following radiation or chemically induced mutagenesis, as a number of defective genes can be simultaneously created and carried into the progeny. Most of these mutations are not obvious to the breeder. Even without mutagenesis, most plants carry a number of inferior alleles of genes that reduce performance when incorporated into an elite genetic background. These mutant genes
can be stable or unstable, and they can be responsible for traits that are disadvantageous or even dangerous.
There are many examples of breeding projects that have resulted in phenotypes that were not expected based on knowledge of the parental genotypes. In one cross between the cultivated potato, Solanum tuberosum, and a wild potato, Solanum brevidens, the hybrid offspring produced a novel steroidal alkaloid, demissine, which was not produced by either parent. A retrospective analysis revealed a potential biochemical explanation for this occurrence (Laurila et al. 1996), but it was certainly not predicted. In another case, potatoes bred for insect resistance were found to be toxic because of high levels of glycoalkaloids and had to be withdrawn from commercial production (Zitnak and Johnston 1970; see also Hellenas et al. 1995). A gene used to create corn hybrids was found several years later to make the plants susceptible to a toxin produced by the fungus Bipolaris madis, and this led to widespread destruction of the corn crop by Southern corn leaf blight in 1970 (Dewey et al. 1986; see also “Environmental Impacts of the Deliberate Introduction of Biological Novelty” above). The backcrossing and selection procedures help eliminate undesirable genetic material that is inadvertently introduced when crosses are made between genetically distinct parental lines. However, it is not possible to know the nature of the majority of genes in the genome, so the outcome of a typical genetic cross cannot be known.
Transgenic Techniques for Crop Improvement
The development of techniques that allow the isolation, sequencing, and transformation of genes into plants provides a novel mechanism for creating genetic diversity. Not only is it possible to isolate a gene (or genes) that control a valuable trait in a sexually compatible species and introduce it without simultaneously carrying along thousands of other genes (the so-called genetic drag), it is also possible to introduce genes from species that are not sexually compatible with the recipient (horizontal gene transfer). Thus, completely novel traits can be introduced into grain and fruit crops that will improve their yield, their resistance to biotic and abiotic stresses, their biochemical composition, their shelf life, and other traits.
Currently, there are two common transgenesis procedures by which purified genes are routinely introduced into plant cells (Birch 1997). One makes use of the Ti-plasmid of Agrobacterium tumefaciens to transfer the gene as part of this plasmid’s DNA (referred to as T-DNA). The Ti-plasmid is a natural vector for transforming genes into plant cells, and a great deal is known about how this occurs (Sheng and Citovsky 1996). Normally, the T-DNA contains genes encoding enzymes that cause the trans-
formed cell to become tumorigenic; however, these tumor-causing genes are removed from the T-DNA vectors used to genetically engineer crop plants.
The Ti-plasmid of A. tumefaciens is capable of transforming most dicot plant cells. However, in order to transmit a gene to the progeny, it must be introduced into a totipotent cell (i.e., one that can be regenerated into a whole plant). It is common, therefore, to cocultivate plant tissues, such as very young embryos, which contain totipotent cells with A. tumefaciens containing genetically engineered T-DNAs (Birch 1997). In the case of Arabidopsis, a model flowering plant used for basic research, developing flower buds can simply be soaked in liquid containing the bacteria in order for germ cells to become transformed with the T-DNA. Typically, the proportion of totipotent cells that become transformed is small. Therefore, a gene encoding a selectable marker, such as a gene for antibiotic or herbicide resistance, is incorporated into the Ti-plasmid to simplify the detection of transformed cells. This allows screening among the embryogenic transformants for those carrying the transgene because only the transformed cells will grow in media containing the antibiotic or herbicide.
The second method for plant cell transformation involves using a physical agent (metal particle or fiber) to pierce the cell wall and carry the cloned DNA into the nucleus. The first application of this technique involved coating tungsten particles with plasmid containing a cloned gene and using the discharge of a 22-caliber cartridge to accelerate the metal particles into the target tissue (Sanford et al. 1993). Thus, it is known as ballistic transformation, and the device used to transfer the DNA into the tissue is called a gene gun. Subsequently, it was found that a similar result is possible using silicon fibers, or whiskers, or electroporation (Asano et al. 1991). Physical insertion does not require any special proteins to transfer the DNA into the nucleus, as occurs with T-DNA-mediated transformation, so the technique is not restricted to species that can be transformed by A. tumefaciens. Ballistic transformation was the first technique used to transform monocots, such as cereals, as they appeared to be resistant to A. tumefaciens infection. However, in recent years, procedures for transforming cereals with A. tumefaceins have been developed, and it is now possible to transform many agronomically important cereals with T-DNA (Chan et al. 1992, Ishida et al. 1996). It is commonly thought that T-DNA-mediated transformation results in fewer, and simpler, transgene insertions than ballistic transformation; however, ballistic transformation can also result in single-gene insertions (Kohli et al. 1998).
As with T-DNA transformation, a gene introduced by physical insertion must be in a totipotent cell in order to end up in the germ line. For
this reason, tissues capable of generating somatic embryos, such the cotyledons of young embryos, are also routinely used for ballistic transformation. As only a few of the transformed cells develop into a plant, a selectable marker gene is also valuable in this case to identify cells carrying the transgene.
Ideally, one would target an engineered gene to a specific chromosomal locus, thereby replacing an existing allele (called homologous recombination). While this is possible with certain fungi and mammals (Doetschaman et al. 1987, Hynes 1996), it is not yet a routine procedure for crops. It is possible that directing gene replacement in higher plants is thwarted by the large amount of repetitive DNA contained in their nuclear genomes, or the enzymes required for homologous recombination might not be present in somatic cells. There is only one report of homologous recombination directing gene replacement in a flowering plant (Kempin et al. 1977), and this was in Arabidopsis, which contains only a small amount of repetitive DNA. In this case a very large number of transformants had to be examined before a gene replacement was detected.
While it is not possible to target a specific site for inserting a gene into a plant, it is possible to determine the exact location of the gene after it has been inserted. The mechanisms that determine the site of transgene insertion into the DNA of plant cells by either T-DNA or ballistic/physical transformation are not understood. It is widely believed that the transgene is targeted to regions of transcriptionally active chromatin, but this has not been widely investigated (Birch 1997). Since transformation is often done with cells from somatic embryos, it is possible they are expressing genes from a variety of regions throughout their genome. In the case of T-DNA transformation, it is reasonably well documented that insertions take place throughout most regions of the genome (Azpironz-Leehan and Feldmann 1997).
It is common for two or more foreign DNA molecules to combine end to end in the plant cell prior to insertion into the nuclear genome. Many transformed plants contain multiple copies of a transgene, which occur in tandem or inverted repeats. In some cases, fragments of host DNA can become interspersed between the duplicated copies of the transgene, creating complex loci (Kohli et al. 1998, Pawlowski and Somers 1998). Although the mechanisms that create complex transgenic loci are incompletely understood, some evidence suggests they are related to plant cell wound responses and DNA repair mechanisms (Sonti et al. 1995).
Multiple transgene insertions and complex insertions at single loci are highly correlated with transgene silencing (see BOX 1.3), which can lead to inactivation of the transgene and possibly other genes in the genome with sequence homology to regions of the transgenic DNA (Birch
BOX 1.3 Types and Consequences of Transgene Silencing
Two types of transgene silencing have been described: transcriptional and posttranscriptional (Chandler and Vaucheret 2001). Transcriptional gene silencing is often associated with the presence of duplicated DNA sequences in the genome corresponding to the transgene. However, it can also occur through transgene insertion into nonexpressed regions of chromosomes. DNA duplication-directed silencing can occur as a consequence of multiple or complex insertions of the transgene, or it can occur due to other genes in the genome with as little as 100 base pairs of DNA identical with that of the transgene. The consequence can be silencing of the transgene, the endogenous gene, or both. Posttranscriptional gene silencing is associated with sequence homology in gene-coding regions and appears to involve the formation of double-stranded (ds) RNA molecules (Mourrain et al. 2000). It can result either from antisense transcription of a gene, RNA-dependent RNA polymerase copying of an mRNA, or transcription of an inverted repeat sequence. These dsRNAs appear to trigger an antiviral response, resulting in ribonuclease-mediated degradation of homologous RNAs into small 25 base pair dsRNA fragments. These small dsRNA fragments spread throughout the plant via plasmodesmata and phloem tissue, leading to systemic destruction of transcripts containing identical sequences.
1997). Consequently, in order to avoid transgene silencing it is important to screen transformation events to identify those consisting of simple insertions at single loci within the genome.
One way to obtain single gene insertions is by replacing one gene with another. While it is not yet possible to obtain gene replacements by homologous recombination in higher plants, it is possible to target transgene insertion to preset sites by using the P1 bacteriophage Cre/lox or the FLP/FRT recombination system of the 2-i plasmid of Saccharomyces cerevisiae (Gates and Cox 1988, Hoess and Abremski 1990). These technologies require transformation of one plant with a transgene flanked by lox or FRT DNA sequences and a second plant transformed with the gene encoding the Cre or FLP recombinase, respectively. When a sexual cross is made between such plants, the enzymatic activity of the recombinase will splice out multimeric transgene insertions, leaving behind single copies of the transgene DNA flanked by lox or FRT sequences (Srivastava et al. 1999). Although these genetically engineered recombination systems require additional steps for creating a transgenic trait, they are useful for removing selectable marker genes that are required to propagate a transgene in a bacterial and plant host cell. Transgenes can also be targeted to lox or FRT insertion sites, which allows well-characterized trans-
genic lines to be reused for multiple types of transgene insertions, including introduction of stacked (multiple) transgenes.
Transgene silencing is a phenomenon that can occur with either T-DNA or biolistic/physical insertion of transgenes into the genome, and it may not become detectable until several generations beyond the initial transformation event (see BOX 1.3). Commonly, the goal of inserting foreign genes into plants is to obtain expression of that gene. In these cases gene silencing is a problem. However, there are cases when gene silencing can be used as a valuable tool to block expression of one or more of the existing genes in the crop. For example, this is a way to inactivate a key enzyme in an undesirable metabolic pathway or a means to eliminate expression of genes responsible for producing food allergens.
Because the phenomenon of gene silencing became apparent in the early days of plant genetic engineering, simple types of transgene insertions were often selected from hundreds of transformation events before advancing a few transgenic events for trait evaluation. Typically, these events were characterized by restriction enzyme mapping to determine the number and complexity of transgene insertions, and the nucleotide sequences flanking the transgene may have been sequenced to ensure the integrity of the transgene itself. Because of improvements in and cost reduction of DNA-sequencing technologies in recent years, it is now cost effective, and prudent, to sequence the entire transgene and its site of insertion, unless the objective is silencing a host gene.
In some crops the ability to transform and regenerate plants is restricted to a few genotypes. Typically, these are not the elite lines that have been bred for high agronomic performance. Consequently, a genetically engineered trait is usually transferred by sexual crossing followed by six or more generations of backcrossing into a more useful genetic background. This requirement has many of the same advantages and disadvantages of mutation breeding with conventional crops, as there is a certain amount of genetic drag associated with any nonelite genetic background. However, the process of backcrossing provides the plant breeder with an opportunity to select among a large number of transformation events for those with the best phenotypic outcome and agronomic performance. Through this screening mechanism, complex transgene loci with the potential for unwanted gene silencing or poor expression and heritability are eliminated. Should the genetically engineered trait prove to be unstable or lead to an unpredictable phenotype or a plant with poor agronomic or horticultural performance, it can be culled before being advanced to evaluation for commercial production. As with traditional plant breeding, newly created genotypes must be evaluated in a variety of environmental situations before it is possible to predict their favorable or unfavorable consequences. Thus, there is extensive selection and field
testing before a transgenic event becomes propagated to a sufficient degree to allow commercialization of a transgenic crop.
In summary, the above discussion indicates that the genetic processes used in conventional and transgenic approaches to crop improvement each have associated risks. In both processes there are two basic steps. One involves adding novel genetic material to the crop, and the second involves filtering out any of the added genetic material that is undesirable. In conventional approaches there are typically more unwanted genes that must be eliminated from the hybridized lines. But in both approaches there is always a possibility that some undesirable traits will remain in the final product. Furthermore, in both cases there is a possibility that the desired genes themselves will interact with existing genes in the parent line and create unexpected effects.
While transformation of genes into crop plants is a new technique for creating genetic diversity, it fundamentally accomplishes many of the same goals as genetic recombination and reassortment, or making wide genetic crosses between geographically isolated or related species. The key difference is the way in which the genetic variation is created. Once inserted into the genome, transgenes are subject to the same types of genetic regulatory mechanisms as conventional genes. As we are learning from the characterization of plant genomes (Kaul et al. 2000), the nuclear DNA is naturally in a constant state of flux as a consequence of single nucleotide changes and DNA rearrangements and insertions. Even though it is uncommon, horizontal gene transfers are not unprecedented. Even the human genome appears to contain a surprisingly large number of insertions of bacterial DNA without ill effect (Lander et al. 2001, Venter et al. 2001). Thus, the fact that transgenes can originate from another species is not novel. What makes the transgenic approach particularly new is the potential to incorporate novel traits into plants. Current and future transgenic traits are discussed elsewhere in this report.
The line between conventional crop breeding and the creation of transgenic crops has never been perfectly clear, but the distinction between the two approaches is likely to blur even further. The genetic engineering process, per se, presents no new categories of risk compared to conventional breeding, although this technology could introduce specific traits or combinations of traits that pose unique risks. However, crop varieties developed solely by conventional breeding could express traits that would need to be regulated as stringently as those developed by transgenic approaches.
Finding 1.4: Genetic improvement of crops typically involves the addition of genetic variation to existing cultivars, followed by screening for individuals that have only desirable traits.
Finding 1.5: Both transgenic and conventional approaches (i.e., hybridization, mutagenesis) for adding genetic variation to crops can cause changes in the plant genome that result in unintended effects on crop traits.
Finding 1.6: The transgenic process presents no new categories of risk compared to conventional methods of crop improvement, but specific traits introduced by both approaches can pose unique risks.
Finding 1.7: Screening of all crops with added genetic variation must be conducted over a number of years and locations because undesirable economic and ecological traits may only be produced under specific environmental conditions.
OVERVIEW OF CURRENT U.S. REGULATORY FRAMEWORK FOR TRANSGENIC ORGANISMS
In the early 1980s the U.S. government determined that there was a need to regulate transgenic organisms, including engineered plants, to assure health and environmental safety. However, it was decided that such regulation could be accomplished by using existing federal statutes instead of developing new ones specifically designed for transgenic organisms. The U.S. Coordinated Framework for the Regulation of Biotechnology (OSTP 1986) was developed to offer guidance for using existing federal statutes and the expertise of existing regulatory agencies in a manner that would ensure health and environmental safety while maintaining sufficient regulatory flexibility to avoid impeding the growth of the biotechnology industry. What follows is an abbreviated description of the coordinated framework, so that readers can understand the regulatory context of the USDA-APHIS regulation of transgenic organisms. The federal government, in lieu of creating a new law for transgenic organisms, asked that the USDA adapt its interpretation of the Federal Plant Pest Act (FPPA) and the Federal Plant Quarantine Act (FPQA) to ensure environmental safety while not impeding growth of the industry.
Two basic principles were intended to guide regulatory policy. First, agencies should seek to adopt consistent definitions of those transgenic organisms subject to review to the extent permitted by their respective statutory authorities. Second, agencies should use scientific reviews of comparable rigor. (Both products and research are regulated under this framework, but the review here will concentrate on “products,” which are defined as transgenic organisms that are released as commercial products into the environment).
The 1986 coordinated framework was constructed to be a flexible policy that would evolve as needed. In 1992 the policy’s scope was clarified. While each agency’s statute clearly specifies the scope of oversight that each agency can exercise, the diversity of statutes meant that regulatory oversight could vary in degree from agency to agency and statute to statute. This could create a situation where under one statute a transgenic organism is inadequately regulated while under another it is overregulated. To address this possibility, the Office of Science and Technology Policy (1992) issued two principles to restrict the scope of oversight. First, federal agencies were mandated to exercise oversight of transgenic organisms only when there is evidence of “unreasonable” risk. An unreasonable risk is one in which the value of the reduction in risk obtained by oversight is greater than the cost of oversight. The evidence of risk must include information about the transgenic organism, the target environment, and the type of application. The scope policy is vague as to the appropriate valuation system that should be used in making regulatory judgments. Second, federal agencies were required to focus on the characteristics and risks of the biotechnology product, not the process by which it is created.
Since 1992 there have been a few more changes in the policy, including shifts in the responsibilities assumed by each agency. For the purpose of this discussion, one significant shift has been the assumption of responsibility by EPA for a subgroup of transgenic plants that were designated as pesticidal transgenic plants, including insect-resistant crops, such as Bt corn, and virus-resistant plants.
In response to the coordinated framework, the Biotechnology, Biologics, and Environmental Protection unit (BBEP) of the USDA-APHIS published regulations for the release of transgenic organisms in 1987 (APHIS 1987), under the statutory authority of the FPPA and the FPQA. While the coordinated framework explicitly provided that federal agencies should focus on the characteristics and risks of the biotechnology product, not the process by which it is created, APHIS explicitly used the process of genetic engineering to trigger its oversight. However, in the 1987 Final Rule (APHIS 1987), APHIS argues that it is not treating transgenic organisms differently than so-called established plant pests or naturally occurring organisms, which may be plant pests. For example, APHIS regulates the movement and release of geographically separated populations of known plant pests because a new geographic population may have genetic characteristics absent in the recipient geographic population, which could increase the plant pest risk in the recipient population. In all cases a permit must be obtained from APHIS prior to importation and interstate movement of potential plant pests. For certain transgenic organisms, APHIS has determined that the release of these organisms
into the environment is tantamount to the introduction of a new organism.
The original 1987 APHIS regulations were later to exempt Escherichia coli strain K-12, sterile strains of Saccharomyces cerevisiae, asporogenic strains of Bacillus subtilis, and Arabidopsis thaliana from permitting requirements for interstate movement under certain specified conditions (APHIS 1988, 1990). Modifications with broader impact were made in 1993. These provided a more streamlined procedure called “notification” for enabling the field release of transgenic plants not expected to pose serious risks. They also provided a more formal process for applicants to petition the agency for nonregulated status of an engineered plant based on genetic and ecological knowledge of that plant (APHIS 1993). A detailed analysis of APHIS oversight of genetically engineered plants is the focus of Chapters 3, 4, and 5. Chapter 2 provides a more conceptual overview of risk analysis and its application to transgenic plants.