Oilseed rape (Brassica napus) is sexually compatible with its wild and weedy relative B. rapa, and introgression of genes from B. napus has been found to occur over a few generations. We simulated the early stages of transgene escape by producing F1 hybrids and the first backcross generation between two lines of transgenic B. napus and two populations of weedy B. rapa. Transgene expression and the fitness of the hybrids were examined under different environmental conditions. Expression of the transgenes was analyzed at the mRNA level by quantitative PCR and found to be stable in the hybrids, regardless of the genetic background and the environment, and equal to the level of transcription in the parental B. napus lines. Vigor of the hybrids was measured as the photosynthetic capability; pollen viability and seed set per silique. Photosynthetic capability of first generation hybrids was found to be at the same level, or higher, than that of the parental species, whereas the reproductive fitness was significantly lower. The first backcross generation had a significantly lower photosynthetic capability and reproductive fitness compared to the parental species. This is the first study that examines transgene expression at the mRNA level in transgenic hybrids of B. napus of different genetic background exposed to different environmental conditions. The data presented clarify important details of the overall risk assessment of growing transgenic oilseed rape.

Concerns have been raised in Europe about the efficiency, sustainability, and environmental impact of the first genetically modified crops. The committees and regulators in charge of approving procedures have encouraged a field trial approach for safety assessment studies under current agronomic conditions. We describe the gene flow from sugar beet (Beta vulgaris L.) in a multi-year and multi-crop monitoring study on farmers' fields at two locations that has been carried out since 1995. We analyzed two sugar beet lines that have been genetically transformed for herbicide resistance. One sugar beet has resistance to glufosinate and the other to glyphosate. Large differences among lines, years and locations were observed. These differences provided a broad range of situations to estimate the risks. Sugar beet bolters produced the majority (86%) of the herbicide-resistant seeds harvested in the field. Direct pollen flow from sugar beet bolters to weed beets that were growing within the same field as well as in a neighboring field that was left fallow accounted for only 0.4% of the resistant seeds released over the years and locations. Descendants of the hybrids between the sugar beet and the weed beet produced the remaining 13.6% of resistant seeds. Herbicide-resistant seeds from the progeny of the weed beet were recorded up to 112 m away from the closest transgenic pollen donor. Indications were observed of non-randomness of the weed beet producing resistant progeny. We also analyzed pollen flow to male-sterile bait plants located within and outside of the sugar beet field. Herbicide-resistant pollen flow was recorded up to 277 m, and fitted with an inverse power regression. Using sugar beet varieties with no, or very low, sensitivity to bolting and destroying bolters are two necessary measures that could delay gene flow.

Development of transgenic triticale as a platform for novel bio-industrial products is predicated on an environmental biosafety assessment that quantifies the potential risks associated with its release. Pollen-mediated gene flow to related species and conventional triticale varieties is one pathway for transgene movement. A tier 1 quantification of triticale hybridization was conducted by emasculating and hand pollinating flowers under greenhouse conditions. Approximately 2000 manual pollinations were conducted for each cross and its reciprocal between two triticale genotypes: a modern triticale cultivar (AC Alta) and primary triticale (89TT108), and common wheat, durum wheat and rye. The frequency of outcrossing, hybrid seed appearance and weight, and F1 emergence and fertility were recorded. Outcrossing, F1 emergence and fertility rates were high from crosses between triticale genotypes. Outcrossing in inter-specific crosses was influenced by the species, and the genotype and gender of the triticale parent. In crosses to common and durum wheat where triticale was the male parent, outcrossing was ≥73.0% and ≥69.5%, respectively, but ≤23.9% and ≤3.0% when triticale was the female parent. Overall, outcrossing with rye was lower than with common and durum wheat. F1 hybrid emergence was greater when triticale was the female parent. With the exception of a single seed, all wheat-triticale F1 hybrid seeds were non-viable when triticale was the male parent in the cross. Only seven durum wheat-triticale F1 hybrids emerged from 163 seeds sown, and all were produced with triticale 89TT108 as female parent. With rye, 8 F1 hybrids emerged from 38 seeds sown, and all were produced from crosses to AC Alta; five with AC Alta as the female parent and three as the male. Interspecific F1 hybrids were self-sterile, with the exception of those produced in crosses between common wheat and triticale where triticale was the female parent. Tier 2 hybridization quantification will be conducted under field conditions.

The seemingly inexorable expansion of global human population size, significant increases in the use of biofuel crops and the growing pressures of multifunctional land-use have intensified the need to improve crop productivity. The widespread cultivation of high-yielding genetically modified (GM) crops could help to address these problems, although in doing so, steps must also be taken to ensure that any gene flow from these crops to wild or weedy recipients does not cause significant ecological harm. It is partly for this reason that new GM cultivars are invariably subjected to strict regulatory evaluation in order to assess the risks that each may pose to the environment. Regulatory bodies vary in their approach to decision-making, although all require access to large quantities of detailed information. Such an exhaustive case-by-case approach has been made tractable by the comparative simplicity of the portfolio of GM crops currently on the market, with four crops and two classes of traits accounting for almost all of the area under cultivation of GM crops. This simplified situation will change shortly, and will seriously complicate and potentially slow the evaluation process. Nowhere will the increased diversity of GM crops cause more difficulty to regulators than in those cases where there is a need to assess whether the transgene(s) will enhance fitness in a non-transgenic relative and thereafter cause ecological harm. Current practice to test this risk hypothesis focuses on attempting to detect increased fitness in the recipient. In this paper we explore the merits and shortcomings of this strategy, and investigate the scope for developing new approaches to streamline decision-making processes for transgenes that could cause unwanted ecological change.

The effect of food components on degradation of DNA by DNase I (EC 3.1.21.1) was monitored by electrotransformation of Escherichia coli, making it possible to determine the number of plasmid molecules capable of giving rise to transformed cells. The transformation frequency increased linearly with the plasmid number within the range of 2 × 106 to 2 × 1010. DNA degradation was reduced by one order of magnitude in the presence of 0.05% (w.v–1) maltol or 1 mM putrescine. Complete inhibition of degradation was observed with ≥0.2% (w.v–1) maltol, ≥0.01% (w.v–1) octyl gallate or ≥0.5 mM of spermine. To monitor degradation of plant DNA during food processing, a real-time PCR system was established. The ratio of copy numbers of a potato gbss DNA fragment of 325 bp and a nested 96 bp fragment was determined. The latter served as internal reference for normalization. The system made it possible to exclude process-dependent changes of DNA concentration in the food matrix. Processing of genetically modified potatoes to dried potato sticks, crisps or flakes was studied and drying steps were shown to exert the strongest effect on DNA degradation, resulting in a drop of the ratio from 0.73 to 0.16.

Biological risk assessment of food containing recombinant DNA has exposed knowledge gaps related to the general fate of DNA in the gastrointestinal tract (GIT). Here, a series of experiments is presented that were designed to determine if genetic transformation of the naturally competent bacterium Acinetobacter baylyi BD413 occurs in the GIT of mice and rats, with feed-introduced bacterial DNA containing a kanamycin resistance gene (nptII). Strain BD413 was found in various gut locations in germ-free mice at 103-105 CFU per gram GIT content 24–48 h after administration. However, subsequent DNA exposure of the colonized mice did not result in detectable bacterial transformants, with a detection limit of 1 transformant per 103-105 bacteria. Further attempts to increase the likelihood of detection by introducing weak positive selection with kanamycin of putative transformants arising in vivo during a 4-week-long feeding experiment (where the mice received DNA and the recipient cells regularly) did not yield transformants either. Moreover, the in vitro exposure of actively growing A. baylyi cells to gut contents from the stomach, small intestine, cecum or colon contents of rats (with a normal microbiota) fed either purified DNA (50 µg) or bacterial cell lysates did not produce bacterial transformants. The presence of gut content of germfree mice was also highly inhibitory to transformation of A. baylyi, indicating that microbially-produced nucleases are not responsible for the sharp 500- to 1 000 000-fold reduction of transformation frequencies seen. Finally, a range of isolates from the genera Enterococcus, Streptococcus and Bifidobacterium spp. was examined for competence expression in vitro, without yielding any transformants. In conclusion, model choice and methodological constraints severely limit the sample size and, hence, transfer frequencies that can be measured experimentally in the GIT. Our observations suggest the contents of the GIT shield or adsorb DNA, preventing detectable exposure of feed-derived DNA fragments to competent bacteria.

Accumulation of information about natural hybridization between GM soybean (Glycine max) and wild soybean (Glycine soja) is required for risk assessment evaluation and to establish biosafety regulations in Japan. This is particularly important in areas where wild relatives of cultivated soybean are grown (i.e. East Asia including Japan). To collect information on temporal and spatial factors affecting variation in hybridization between wild and GM soybean, a two year hybridization experiment was established that included one wild soybean and five GM soybean cultivars with different maturity dates. Hybridization frequencies ranged from 0 to 0.097%. The maximum hybridization frequency (0.097%) was obtained from wild soybean crossed with GM soybean cv. AG6702RR, which were adjacently cultivated with wild soybean, with 25 hybrids out of 25 741 seedlings tested. Cultivar AG6702RR had the most synchronous flowering period with wild soybean. Ten hybrids out of 25 741 were produced by crossing with cv. AG5905RR, which had the second most synchronous flowering period with wild soybean. Most hybrids were found where GM and wild soybeans were adjacently cultivated, whereas only one hybrid was detected from wild soybean plants at 2 m, 4 m and 6 m from a pollen source (GM soybean). Differences in flowering phenology, isolation distance and presence of buffer plants accounted for half of the variation in hybridization frequency in this study. Temporal and spatial isolation will be effective strategies to minimize hybridization between GM and wild soybean.

The likelihood of gene transfer from transgenic plants to bacteria is dependent on gene number and the presence of homologous sequences. The large number of transgene copies in transplastomic (transgenes contained in the chloroplast genome) plant cells as well as the prokaryotic origin of the transgene, may thus significantly increase the likelihood of gene transfer to bacteria that colonize plant tissues. In order to assess the probability of such transfer, the length of homologous DNA sequences required between the transgene and the genome of the bacterial host was assessed. In addition, the probability that bacteria, which co-infect diseased plants, are transformable and have sequences similar to the flanking regions of the transgene was evaluated. Using Acinetobacter baylyi strain BD143 and transplastomic tobacco plants harboring the aadA gene (streptomycin and spectinomycin resistance), we found that sequences identical to the flanking regions containing as few as 55 nucleotides were sufficient for recombination to occur. Consequently, a collection of bacterial isolates able to colonize tobacco plant tissue infected by Ralstonia solanacearum strain K60 was obtained, screened for DNA sequence similarity with the chloroplastic genes accD and rbcL flanking the transgene, and tested for their ability to uptake extracellular DNA (broad host-range pBBR1MCS plasmids) by natural or electro-transformation. Results showed that among the 288 bacterial isolates tested, 8% presented DNA sequence similarity with one or both chloroplastic regions flanking the transgene. Two isolates, identified as Pseudomonas sp. and Acinetobacter sp., were able to integrate exogenous plasmid DNA by electro-transformation and natural transformation, respectively. Our data suggest that transplastomic plant DNA recipients might be present in soil bacterial communities.

More than 10 years after registration of the first Bt maize cultivar in Europe, there still exists a remarkable lack of data on effects on Lepidoptera which would be necessary for a complete and comprehensive environmental risk assessment. So far only very few European butterfly species have been tested in this aspect. In our study the effect of transgenic Bacillus thuringiensis (Bt) maize pollen (event Bt-176) on the development and survival of neonate larvae of the Peacock butterfly, Inachis io (L.) was for the first time shown. The results of our study suggest that the Peacock butterfly may serve as a model organism for assessing potential side effects of new developed transgenic Bt crops on non-target butterflies in a GMO environmental risk assessment. The study was done under laboratory conditions by exposing larvae of the Peacock butterfly to various pollen doses of transgenic maize event Bt-176 (cv. PACTOL CB) or the conventional isogenic maize (cv. PACTOL) using a no-choice test. Larvae feeding for 48 h on nettle plants (Urtica dioica) that were contaminated with higher pollen concentrations from Bt-176 maize (205 and 388 applied pollen.cm-2) suffered a significantly higher mortality rate (68 and 85% respectively) compared to larvae feeding on leaves with no pollen (11%), or feeding on leaves with pollen from conventional maize (6 to 25%). At lower Bt maize pollen doses (23–104 applied pollen.cm-2), mortality ranged from 11–25% and there were no apparent differences among treatments. The corresponding LC50-and LC90-values for neonate larvae of the Peacock butterfly were 187 and 448 applied pollen grains.cm-2 of Bt-176, respectively. Weight of larvae surviving consumption of Bt-176 maize pollen declined between 10 and 81% with increased pollen doses (r =  −0.95). The highest weight reduction (81%) corresponded to the highest pollen concentration (388 pollen grains applied.cm-2). Ingestion of pollen from the conventional maize hybrid did not have negative effects on larval weight gain or survival rate.

This paper is the third in a series designed to demonstrate the application of rigorous, systematic hazard identification techniques to ecological systems. Here we use Hierarchical Holographic Modelling to identify the potential ecological hazards associated with the commercial release of herbicide tolerant oilseed rape. Hierarchical Holographic Models decompose complex systems into a series of sub-systems and consider interactions between the components and processes of these sub-systems in order to identify hazards. In this example we considered 1356 potential interactions between the biological, chemical and physical components and processes of the herbicide tolerant oilseed rape environment, and identified 152 potential hazards, grouped into 14 categories. The hazards were subsequently scored for degree of concern and plausibility, and then compared with an equivalent list of hazards generated independently by a checklist approach. The incidence of herbicide tolerant volunteers (and weeds) both on and off the farm had the highest average score of all the ecological hazard categories. The checklist based approach identified or implied 44% of the hazards identified in the Hierarchical Holographic Model, including nine of the ten hazards ranked most important. The checklist approach focussed almost exclusively on the phenotypic and genotypic hazards associated with herbicide tolerant oilseed rape and largely ignored the hazards associated with the circumstances surrounding its use. As a result the checklist identified only 6 out of the 79 potential hazards associated with changes to farming practice. The commercial release of herbicide tolerant oilseed rape will be associated with changes in tillage and the application of post-emergent herbicides. It may also lead to changes in spray schedules of insecticide and fungicide. Many of the environmental hazards identified with these changes are plausible and may warrant further investigation or targeted monitoring.

Several biological containment methods have been developed to reduce pollen dispersal; many of them only have a proof of concept in a model plant species. This review focuses on biological containment measures which were tested for their long-term efficiency at the greenhouse or field scale level, i.e. plastid transformation, transgene excission, cleistogamy and cytoplasmic male sterility (CMS). Pollen-mediated gene transfer in transplastomic tobacco could occur at very low frequencies if the predominant mode of inheritance is maternal. Transgene excision from tobacco pollen can be made highly efficient by coexpression of two recombinases. For cleistogamous oilseed rape it was shown that some flowers were partially open depending on genotypes, environment and recording dates. Reports on the stability of CMS in maize and sunflower indicated that there is a high variability for different genotypes under different environmental conditions and over successive years. But for both crop types some stable lines could be selected. These data demonstrate that the biological containment methods discussed are very promising for reducing gene flow but that no single containment strategy provides 100% reduction. However, the necessary efficiency of biological containment methods depends on the level of containment required. The containment level may need to be higher for safety purposes (e.g. production of special plant-made pharmaceuticals), while much lower containment levels may already be sufficient to reach coexistence goals. It is concluded that where pollen-mediated gene flow must be prevented altogether, combinations of complementary containment systems will be required.

Endotoxins from Bacillus thuringiensis (Bt) produced in transgenic pest-resistant Bt crops are generally not toxic to predatory and parasitic arthropods. However, elimination of Bt-susceptible prey and hosts in Bt crops could reduce predator and parasitoid abundance and thereby disrupt biological control of other herbivorous pests. Here we report results of a field study evaluating the effects of Bt sprays on non-target terrestrial herbivore and natural enemy assemblages from three rice (Oryza sativa L.) fields on Luzon Island, Philippines. Because of restrictions on field-testing of transgenic rice, Bt sprays were used to remove foliage-feeding lepidopteran larvae that would be targeted by Bt rice. Data from a 546-taxa Philippines-wide food web, matched abundance plots, species accumulation curves, time-series analysis, and ecostatistical tests for species richness and ranked abundance were used to compare different subsets of non-target herbivores, predators, and parasitoids in Bt sprayed and water-sprayed (control) plots. For whole communities of terrestrial predators and parasitoids, Bt sprays altered parasitoid richness in 3 of 3 sites and predator richness in 1 of 3 sites, as measured by rarefaction (in half of these cases, richness was greater in Bt plots), while Spearman tests on ranked abundances showed that correlations, although significantly positive between all treatment pairs, were stronger for predators than for parasitoids, suggesting that parasitoid complexes may have been more sensitive than predators to the effects of Bt sprays. Species accumulation curves and time-series analyses of population trends revealed no evidence that Bt sprays altered the overall buildup of predator or parasitoid communities or population trajectories of non-target herbivores (planthoppers and leafhoppers) nor was evidence found for bottom-up effects in total abundances of non-target species identified in the food web from the addition of spores in the Bt spray formulation. When the same methods were applied to natural enemies (predators and parasitoids) of foliage-feeding lepidopteran and non-lepidopteran (homopteran, hemipteran and dipteran) herbivores, significant differences between treatments were detected in 7 of 12 cases. However, no treatment differences were found in mean abundances of these natural enemies, either in time-series plots or in total (seasonal) abundance. Analysis of guild-level trajectories revealed population behavior and treatment differences that could not be predicted in whole-community studies of predators and parasitoids. A more conclusive test of the impact of Bt rice will require field experiments with transgenic plants, conducted in a range of Asian environments, and over multiple cropping seasons.

Genetically modified (GM) maize expressing the insecticidal protein Cry1Ab from Bacillus thuringiensis (Bt maize) is the only GM crop planted commercially in the European Union (EU). Cultivation in accordance with Directive 2001/18/EC demands post-market environmental monitoring (PMEM) to ensure the detection and prevention of adverse effects on the environment possibly deriving from commercial cultivation. Based on a seminar organized in Berlin, Germany, in April 2008 by the EU-funded Biosafenet project, the present paper reflects on experiences and hurdles faced during the implementation of PMEM for Bt maize. It reviews and reconsiders PMEM programs of Bt maize in view of existing experiences from cultivation, current monitoring activities initiated by Member States and applicants, proposed monitoring strategies and methods as well as potential environmental impacts of cultivation. Future challenges will arise from large-scale and cumulative cultivation of various events. This will demand optimized organization structures for data collation and integration to support further decision-making and management.

This study assessed the impact of Pinus radiata (D. Don) genetically modified (GM) by biolistic insertion of the LEAFY and nptII genes on rhizosphere microbial communities of field grown trees. Rhizosphere soil was sampled quarterly for two consecutive years. A culture-independent approach was used to characterise the microbial communities based on PCR and denaturing gradient gel electrophoresis (DGGE) of 16S/18S rDNA gene fragments, and internal transcribed spacer (ITS) fragments amplified from total rhizosphere DNA. Trees from two independent transformation events were sampled, together with non-modified control trees of the same parental genotype. DGGE profiles of rhizosphere general Bacteria did not differ between GM and control trees with one exception (summer 2006 sample). For Alphaproteo- and Actinobacteria, significant differences between treatments were detected in one out of eight samplings. Small seasonal shifts could be detected in all bacterial communities. General fungal and ectomycorrhizal communities did not differ significantly between GM and control trees with the exception of summer 2006, when ectomycorrhizal communities associated with GM trees from one transformation event differed from those associated with control trees. Small seasonal shifts of general fungal and ectomycorrhizal communities were seen over the two-year sampling period. More detailed analysis of microbial communities at one sampling date (using amplified rDNA restriction analysis (ARDRA) and 16S/18S rDNA sequencing) revealed significant differences in four ARDRA groups between one GM treatment and the control (bacteria), and significant differences in one ARDRA group between the two GM treatments (fungi). When data from all sampling dates are considered together, the low incidence of statistical differences in the microbial communities associated with the genetically modified and control trees suggests that there was no significant impact of this genetic modification on rhizosphere microbial communities.

Over the past decades, genes of Bacillus thuringiensis var. kurstaki (Berliner) (Bt) coding for protein toxins have been engineered into maize for protection against the European Corn Borer (Ostrinia nubilalis (Hbn.)). However, these transgenic plants may have an impact on non-target organisms. In particular, a potential hazard was identified for non-target lepidopteran larvae, if they consume Bt maize pollen on their host plants. Risk can be defined as a function of the effect of an event (hazard) and the likelihood of this event occurring. Although data on toxicity (hazard) are available from many lab and field studies, knowledge about the environmental exposure of European lepidopteran larvae is incomplete at the population level. Therefore we studied the distribution of small tortoiseshell caterpillars (Aglais urticae (L.)) and its host plant in an agricultural landscape in Germany, to estimate the potential population exposure to maize pollen. The results showed that larvae of the small tortoiseshell developed primarily on freshly sprouted nettle stands (Urtica dioica (L.)) in field margins, rather than adjacent to hedges and groves. However, the main distribution was at margins of cereal (non-maize) fields, where 70% of all larvae were found. This may be due the fact that cereals covered 54% of the survey area, while maize only covered 6.1%. On the other hand, maize fields seem so show higher food plant densities than cereal crops. The results must be interpreted carefully, as the data basis of the present study is very small, and the situation can vary between years due to crop rotation or other changes in agricultural practices. Therefore it is still questionable whether the small tortoiseshell is significantly exposed to maize pollen. For a conclusive risk assessment, more replications and surveys of larger areas in different intensively managed agricultural landscapes over several years are needed.

Genetically engineered transgenic crop varieties (TGVs) have spread rapidly in the last 10 years, increasingly to traditionally-based agricultural systems (TBAS) of the Third World both as seed and food. Proponents claim they are key to reducing hunger and negative environmental impacts of agriculture. Opponents claim they will have the opposite effect. The risk management process (RMP) is the primary way in which TGVs are regulated in the US (and many other industrial countries), and proponents claim that the findings of that process in the US and its regulatory consequences should be extended to TBAS. However, TBAS differ in important ways from industrial agriculture, so TGVs could have different effects in TBAS, and farmers there may evaluate risks and benefits differently. To evaluate some potential impacts of TGVs in TBAS we used the RMP as a framework for the case of Bt maize in Mesoamerica and Cuba. We interviewed 334 farmers in Cuba, Guatemala and Mexico about farming practices, evaluations of potential harm via hypothetical scenarios, and ranking of maize types. Results suggest high potential for transgene flow via seed, grain and pollen; differences in effects of this exposure in TBAS compared with industrial agriculture; farmers see some potential consequences as harmful. Perceptions of harm differ among farmers in ways determined by their farming systems, and are different from those commonly assumed in industrial systems. An RMP including participation of farmers and characteristics of TBAS critical for their functioning is necessary to ensure that investments in agricultural technologies will improve, not compromise these agricultural systems.

Brassica rapa grows as a wild and weedy species throughout the world and is the most likely recipient of transgenes from GM oilseed rape. For transgene introgression to occur, the critical step which must be realized, is the formation of an F1 hybrid. Concerns exist that hybrid populations could be more vigorous and competitive compared to the parental species. This study examines the effect of simulated herbivory and interspecific competition on the vegetative and reproductive performance of non-transgenic F1 hybrids and their parental lines. Several vegetative and reproductive performance measures were used to determine the effect of simulated herbivory and competition on the Brassica lines, including leaf length and biomass for herbivory and seedling height and biomass for competition. For defoliation experiments, B. rapa showed little response in terms of leaf length but B. napus and the F1 hybrid responded negatively. Brassica rapa showed elevated biomass responses, but B. napus and the hybrid demonstrated negative responses to defoliation. Defoliation at the cotyledon stage had a slight effect upon final biomass with the F1 hybrid performing significantly worse than B. napus, although seed counts were not significantly different. For the series of competition experiments, hybrids seemed to be more similar to B. rapa in terms of early seedling growth and reproductive measures. The underperformance of hybrid plants when challenged by herbivory and competition, could potentially decrease survivorship and explain the rarity of hybrids in field surveys. However, should transgene introgression occur, the dynamics of hybrids could change radically thus increasing the risk of gene flow from a transgenic oilseed rape crop to the wild recipient.

In Europe, there seems to be widespread, morally based scepticism about the use of GMOs in food production. In response to this scepticism, the revised EU directive 2001/18/EC on the deliberate release into the environment of genetically modified organisms stresses the importance of respecting ethical principles recognized in the Member States. However, the directive fails to reflect the critical role of value judgements in scientific risk assessment and any subsequent approval procedure. In this paper we argue that it is important to make all ethically relevant assumptions involved in the approval procedure transparent and thus available for public scrutiny. Mapping the value judgements that are made in an environmental risk assessment and approval procedure, we describe the political liberal nature of the EU legislation. We then look more closely at the prescriptions for environmental risk assessment and approval of GMOs outlined in the directive. An environmental risk assessment views the world through a “risk window” that only makes visible that which has been predefined as a relevant risk. The importance of the value judgements that define the risk window consists in limiting the information the risk assessment can provide. In the penultimate section of the paper, the significance of the risk window is demonstrated through a case study of the approval of glyphosate resistant fodder beets (Beta vulgaris L. ssp. vulgaris) in Denmark.