Considerable attention has been paid to potential risks and risk assessment approaches for genetically engineered (GE) organisms. In recent years, a number of regulatory agencies around the world have been formulating assessment procedures to ensure the acceptable environmental safety of GE organisms. Although there are exceptions, these procedures typically follow conventional risk assessment steps, including formulation of the problem, assessment of the effects, assessment of the exposures, and characterization of the risks.

Risk assessments for GE organisms, especially GE crops, have ranged from simple, qualitative characterizations to complex, quantitative characterizations. Regardless of the type or complexity of the assessment, nearly all risk assessments of GE crops address the risk of the crop as a standalone system, without formally and systematically considering the ecological or human-health risks posed by alternative crops and cropping systems.

Risks for crops and cropping systems posed by organic, conventional, and mutagenic approaches typically are not compared to GE crops and systems. This is interesting because many other environmental risk assessments and decisions in the U.S. are presented as part of governmentally mandated Environmental Impact Statements or Environmental Assessments. A hallmark of those assessments is the analysis of alternatives (albeit not typically formalized, comparative risk assessments) to the proposed action. Formal comparative risk analyses of alternatives would be valuable for decision-making about, and public communication of, genetic engineering technologies. Comparative risk analysis can provide a broader perspective from which to consider risks and benefits posed by genetic engineering.

In Peterson and Hulting¹ and Peterson and Shama², we compared multiple aspects of risk associated with different wheat production systems in the U.S. and Canada using the risk assessment paradigm. We chose to examine risk issues associated with wheat because wheat varieties using genetic engineering are just now emerging. Wheat varieties produced using these biotechnologies have lagged behind other crop species, but are now being developed in the case of genetic engineering and are being grown commercially in the case of mutagenic techniques. Therefore, this provided us with a unique opportunity to assess comparatively the potential environmental risks (human health, ecological, and livestock risks) associated with the different biotechnology and conventional wheat production systems.

For our assessments, we used tier 1 quantitative and qualitative risk assessment methods to compare specific environmental risks associated with genetically engineered, mutagenic, and conventional wheat production systems (specifically herbicide and protein risks) in Canada and the U.S. Risk assessment typically utilizes a tiered modeling approach extending from deterministic models (Tier 1) based on conservative assumptions to probabilistic models (Tier 4) using refined assumptions³. In risk assessment, conservative assumptions in lower-tier assessments represent overestimates of effect and exposure; therefore, the resulting quantitative risk values typically are conservative and err on the side of environmental safety.

Herbicide-tolerant wheat varieties have been produced using both genetically engineered (DNA recombination) and chemically induced DNA mutation techniques. Replacement of traditional herbicides with glyphosate in a glyphosate-tolerant (genetically engineered) wheat system or imazamox in an imidazolinone-tolerant (mutagenic) wheat system may alter environmental risks associated with weed management. Additionally, because both systems rely on plants that express novel proteins, the proteins and plants themselves may impose risks.

For the conventional wheat system, herbicides were considered as the only stressor in our assessment. Therefore, the conventional wheat production system served as a baseline in our analysis. For the glyphosate-tolerant wheat system, herbicides and the transgenic protein were the stressors. For the imidazolinone-tolerant wheat system, herbicides and the mutated protein were the stressors. The primary stressors then potentially affected the systems through human health (dietary exposure for the biotech proteins and herbicides, and applicator risk for the herbicides), livestock, and ecological effects. The effects we considered in this assessment reflected primary impacts. Therefore, we presented only direct effects of the stressors on the human and ecological receptors. We did not consider economic risks or agronomic risks, such as pollen-mediated and mechanical mixing of wheat grain from different production systems, pollen-mediated gene flow to wild or weedy relatives of wheat, fallow management with herbicides, herbicide resistance in target weeds, and herbicide rotation risks to alternate crops.

The herbicide active ingredients evaluated in this study included 2,4-dichlorophenoxy acetic acid (2,4-D), bromoxynil, clodinafop, clopyralid, dicamba, fenoxaprop, flucarbazone, MCPA, metsulfuron, thifensulfuron, tralkoxydim, triallate, triasulfuron, tribenuron, and trifluralin. These active ingredients were chosen because they are used on a relatively large percentage of spring wheat acres in the U.S. and Canada. Risk associated with glyphosate and imazamox also was evaluated because of their use in glyphosate-tolerant and imidazolinone-tolerant wheat.

We characterized risks to the following ecological receptors: wild mammals, birds, nontarget terrestrial plants, nontarget aquatic plants, aquatic vertebrates, aquatic invertebrates, and groundwater. Ecological risks were assessed by integrating toxicity and exposure. To do this, risks to ecological receptors were assessed using the Risk Quotient (RQ) Method. For each ecological receptor, an RQ was calculated by dividing the Estimated Environmental Concentration (EEC) by the appropriate toxicity endpoint (e.g., the LC50).

Risks for the glyphosate-tolerant CP4 EPSPS protein were determined primarily using a qualitative weight-of-evidence approach. Effect and exposure information for humans, livestock, and wildlife (such as mammals, birds, and fish) was obtained from the scientific literature. Other information was obtained from regulatory reports and submissions.

As with the CP4 EPSPS protein, risks for the mutated imidazolinone-tolerant AHAS protein were determined using a qualitative weight-of-evidence approach. However, effect and exposure information for the mutated AHAS protein is not available in the scientific literature. Further, because it is a mutagenic trait and not a genetically engineered trait, regulatory approvals are not required in the U.S. The regulatory status of imidazolinone-tolerant wheat also limits the availability of public information. In Canada, imidazolinone-tolerant wheat is regulated as a novel trait by the Canadian Food Inspection Agency (CFIA) and Health Canada. Therefore, we used the decision documents produced by these two agencies for our risk assessment.

Both glyphosate and imazamox presented lower human health and ecological risks than many other herbicides associated with conventional wheat production systems. The differences in risks were most pronounced when comparing glyphosate and imazamox to herbicides currently with substantial market share. Current weight-of-evidence suggests that the transgenic CP4 EPSPS protein present in glyphosate-tolerant wheat poses negligible risk to humans, livestock, and wildlife. Risk for mutated AHAS protein in imidazolinone-tolerant wheat most likely would be low, but there were not sufficient effect and exposure data to adequately characterize risk. Environmental risks for herbicides were more amenable to quantitative assessments than for the transgenic CP4 EPSPS protein and the mutated AHAS protein.

An important caveat emerges from our work: Tier 1 risk assessment approaches have limited value for accurate quantifications of risk because of their simplistic hazard and exposure assumptions. These assumptions, which are highly conservative and err on the side of environmental safety, typically are used for highlighting significant versus negligible risks during preliminary decision-making and not for determining actual site-specific risks. Therefore, our results should not be used as representations of "actual" risks. To determine more realistic risks, higher-tier assessments for these technologies should be used. However, we believe quantitative and qualitative tier 1 approaches are valuable for making direct comparisons between environmental stressors.

In our work, environmental risks for herbicides were more amenable to quantitative assessments than for the transgenic CP4 EPSPS protein and the mutated AHAS protein. Because of specificity and familiarity with their native counterparts, evolving regulatory requirements, and the fact that they are not pesticides, these proteins do not have the same completeness of toxicity testing data as herbicides. We believe it is important that minimum effect and exposure data (such as acute mammalian toxicities to altered or inserted proteins) are generated and made publicly available for all novel plant traits, including non-genetically engineered approaches. These data would allow for independent, third-party tier 1 assessments of risk and proper communication of those risks to the public.

Although our assessment was not comprehensive, we believe the approach demonstrates the potential risk-risk tradeoffs when implementing the newer biotechnologies. We are currently applying similar comparative risk approaches to plant-based pharmaceuticals. These types of comparative biological risk assessments add valuable information, which subsequently aid regulatory and public decision-making about biotechnology.

1. Peterson RKD and Hulting AW. (2004) A comparative ecological risk assessment for herbicides used on spring wheat: the effect of glyphosate when used within a glyphosate-tolerant wheat system. Weed Sci. 52, 834-844

2. Peterson RKD and Shama LM. (2005) A comparative risk assessment of genetically engineered, mutagenic, and conventional wheat production systems. Transgenic Res. 14, 859-875

3. [NRC] National Research Council (1983) Risk Assessment in the Federal Government: Managing the Process. National Academy Press, Washington, DC.

Since the first large-scale introduction of genetically modified (GM) crops a decade ago, the global area cultivated with these crops has undergone a continuous increase, amounting to a total of 90 million hectares in 2005.¹ For comparison, this area equals the national sizes of Portugal, Spain, and Italy together. Many of the "foreign" genes that have been introduced into these crops, i.e., the transgenes, are derived from microbial sources. As explained below, the issue of their potential transfer to other organisms was addressed in a recent article published by our group.²

Long before the first introduction of GM crops, international organizations like the Food and Agriculture Organization (FAO), World Health Organization (WHO), and Organization for Economic Co-operation and Development (OECD), had been promoting international consensus on how to assess the safety of such crops. An internationally harmonized approach of comparative safety assessment was thus formulated in which the GM crop is compared to a conventional counterpart with a known history of safe use (reviewed in ³).

Usually, this comparison entails a description of the genetic modification, such as the nature of the DNA used and the function of the transgenes and encoded proteins, as well as of agronomic and phenotypic traits and composition. Based upon the differences thus identified, a strategy for further safety assessment can be chosen. Given the wide variety in characteristics of both the host crops and the transgenes, this approach entails decisions on a case-by-case basis, rather than a "cook book" with standard recipes.

Issues that are commonly addressed during the regulatory safety assessment of GM crops include:

In 2003, the activities on international consensus building culminated into the establishment of Codex Alimentarius' guidelines on the conduct of safety assessment of foods derived from genetically modified plants and micro-organisms.⁴ Codex Alimentarius standards, guidelines, and other documents are important because they serve as reference for international trade disputes over the safety of internationally traded foods under the international agreement on sanitary and phytosanitary standards (SPS).

Horizontal gene transfer is one of the important issues addressed during the safety assessment of GM crops. In the Codex Alimentarius guidelines, the focus of the assessment of this topic is restricted to the potential transfer of antibiotic resistance marker genes and the consequences thereof. These marker genes are used to facilitate the process of genetic modification. This is done by co-introducing the gene of interest with an antibiotic resistance gene into the DNA of a crop cell. Those cells that have been successfully modified can be selected based upon their ability to sustain on culture media containing the pertinent antibiotic, to which non-modified cells are sensitive. Antibiotic resistance marker genes therefore do not serve a purpose in the GM crop itself.

Antibiotic resistance currently is a matter of great priority to health care, as evidenced, for example, by the attention devoted to this issue by organizations like the WHO. For example, popular media give accounts of the dissemination in hospitals of antibiotic-resistant pathogens, such as methicillin-resistant Staphylococcus aureus (MRSA). In general, the spread of antibiotic resistance is considered to be linked to the way that antibiotics are used, among other factors.

During the safety assessment of GM crops, the possibility of the transfer of antibiotic resistance genes that have been introduced into GM crops is considered. The European Food Safety Authority’s Scientific Panel on Genetically Modified Organisms recently issued an opinion on antibiotic resistance genes.⁵ This opinion, among others, proposed a categorization of the antibiotic resistance genes into three categories based on the clinical importance of the antibiotic, the natural prevalence of resistance to the same antibiotic in nature, and the likelihood of transfer. Only antibiotic resistance genes that fall into the first category of this scheme, such as the kanamycin resistance gene nptII, are recommended to be allowed for use in GM crops that are to enter the market.

In practice, however, regulatory safety assessments do not limit the scope of potential transfer of transgenes from GM crops only to antibiotic resistance. These assessments also address other potential effects of transgenes, including pathogenicity. The potential impacts of gene transfer on health and the environment in a broad sense are considered by European Union guidelines.^6,7

Similar to antibiotic resistance, literature reports indicate that characteristics associated with pathogenicity have been exchanged between microorganisms like Escherichia coli and Salmonella enterica, such as through transfer of DNA fragments containing "pathogenicity islands." A wide array of biochemical characteristics are known to be involved in the pathogenicity of microorganisms, such as the formation of adhesion molecules that bind to host cells, enzymes that facilitate entrance into host cells, self-sufficiency for some nutritional compounds, and "quorum sensing" within groups of micro-organisms.

Various mechanisms by which DNA is horizontally transferred between microorganisms are known to exist in nature, including transfer after conjugation between bacteria, transduction by bacteriophages, and transformation by free DNA. Potential transfer of transgenes from GM crops to microbes in the gastro-intestinal tract likely proceeds through a process in which competent cells are transformed with free DNA. As stated above, this can occur after the DNA of the GM crops has been released from its host cells, for example during digestion.

Various factors influence the likelihood that transfer of DNA from a GM crop to a recipient bacterium will occur and become productive. One of these factors is the level of the bacterium’s competence, i.e., the physiological state of a bacterial cell during which it can bind, take up, and recombine DNA molecules. The outcomes of a number of studies indicate that the most likely mechanism by which DNA is transferred from GM crops to microorganisms is by homologous recombination. This means that the recipient microorganisms should already contain sequences that are sufficiently similar ("homologous") to the incoming foreign DNA, such that they can align with each other and allow for integration of the latter.

Finally, plant genes and microbial genes differ with respect to preferred base composition of the codons. Plant genes also have other features that differ from microbial genes, such as introns, which do not occur in bacterial sequences, and different types of regulatory sequences.

On the one hand, based on these considerations, which have been reviewed in more detail elsewhere,⁸ it appears that transgenes of microbial origin carry an enhanced likelihood of being transferred from GM crops to microorganisms. Genetic modification allows for the introduction of foreign genes from one organism into another, unrelated organism. As a result of this, many of the GM crops currently on the market contain transgenes of microbial origin, such as enzymes metabolizing herbicides obtained from soil microorganisms or insecticidal proteins obtained from Bacillus thuringiensis.

In our review,² we focused on transgenes of microbial origin other than antibiotic resistance genes that are present within GM crops approved by the regulatory authorities of the European Union, United States of America, Canada, Australia, and New Zealand. A number of factors that influence the transfer of these transgenes, as well as the potential impact of such a transfer on the health of consumers, were considered. For each gene studied, these factors, if applicable and information available, included:

None of these single items can be considered completely predictive for adverse effects and therefore a combination of factors has to be considered in a "weight of evidence"-based approach. Based upon these considerations, a conclusion was formulated for each gene as to whether its transfer from GM crops would be likely to have any adverse health effects in consumers. In total, 20 microbial transgenes were considered, including five that are linked with herbicide resistance, three with hybrid breeding through male sterility, two with prolonged fruit ripening, two linked with markers for genetic modification, and eight with insecticidal properties. The genes with insecticidal properties all encoded Cry proteins from B. thuringiensis.²

Another case of selective advantage in soil conditions was that of the 1-aminocyclopropane-1-carboxylate (ACC) deaminase gene, which has been isolated from a soil isolate of Pseudomonas and introduced into GM tomatoes to suppress ethylene synthesis and thereby delay ripening. It has been observed that this gene is expressed in soil microorganisms colonizing plant roots and that its activity is associated with increased root formation.¹⁰ We therefore postulated that the transfer of this gene may confer a selective advantage to recipient microorganisms in the vicinity of plants producing ACC.

It should be noted that the data on the original sequences from the native hosts may represent a "worst case" situation. This is because in GM crops the transgene sequences may have been optimized for expression in plants. As stated above, plant genes have a number of features that are different from bacterial genes, which decrease the likelihood of effective transfer and expression of plant genes to bacteria.

In conclusion, it was recommended to include the abovementioned considerations in safety assessments of GM crops carrying transgenes other than the ones already reviewed in the current survey².