This may be treated as a general conclusion. However, the larger point is not that IR transgenes may be relatively easily contained. Rather, the ecological model provides a tool with which to determine how best to do this, as well as to assess how well it has been done. In the model, all terms are ratios of the character in question in the transformed versus untransformed plants. The risk of a transgene can thus be assessed under protected conditions and calibrated by the response of the untransformed plant under more natural conditions, as at least a first pass evaluation.

It may be possible to manipulate either costs or benefits in order to contain the transgene. Benefit, which is quantified in the model as growth in the absence of competition (), by definition is maximized for commercial return on the crop, and so may not be available for manipulation. Costs may therefore be the more likely target for manipulation: if you are going to have a magic helmet then you must have an Achilles heel. Some costs are a function of resistance itself, e.g., additional protein construction. It is not possible to have the transgene without these costs and the trade-off between costs and benefits in this case would be in choosing the transgene. As with any 'cost', its usefulness as a control will come in two parts: the extent to which it does not cut into commercial returns on the transgene; and the reliability of its genetic linkage to the transgene over time.

The model delineates control possibilities in addition to costs inherent to the transgene. Costs most amenable to manipulation are included in two composite variables: seedling competitive ability (β) and seeds viable in the first season after seed set (Υ). β includes factors such as overtopping, and this is where construction costs would be taken into account. Υ is not simply seed set; Υ determines not just seedlings in year t + 1, but the character of the seed bank. Υ interacts with the behavior of seeds in the seed bank (germination fraction, persistence from season to season) to produce seedlings in future seasons. It is an essential part of the population persistence of a species that has a chance of producing few or no successful offspring in any one year. Υ, the number of seeds that make it through the first winter viable, is the first level of control for any novel allele. If Υ is low, then there are few viable seeds to remain in the seed bank and the capacity to get through bad years is severely limited. If Ε is high, then control shifts to the germination fraction Ε. If Ε is high, then, again, there are few seeds in the seed bank to get through bad years. If Ε is low, control shifts to S, the fraction of seeds that survive from one year to the next. Persistence over hard times is then limited by a low S. Pollen viability, η, where male sterility manipulation may be applied, plays a similar role to Y.

The equations of the model may be iterated numerically but Kelly et al. also provide a useful calculation for doubling time to assess the rate of spread of the transgene at early stages (when hemizygotes, individuals with only one copy of the transgene, comprise approximately less than 20% of the population). Although it requires the same information as the more detailed calculations that gave rise to the above conclusions, the doubling time equation is easier to calculate and provides a reasonable estimate of hemizygote spread.

The focus on the natural population may seem to imply that the ecological interaction will be of importance only once the allele arrives there. In fact, it is likely that the ecological dynamic will have an effect at every step of the journey of the allele from crop to natural community. As introgression of the allele into the natural genome proceeds, the increment of protection that the allele offers will decrease in the context of the already well-protected wild genome;⁸ any cost it incurs is likely to have greater impact in a natural situation in which nutrient supplements become less and less available with increasing distance from the 'home' agricultural field. The model can be applied to determine this, by specifying the values of the factors that go in the applicable ratios.

The goal of the model is to clarify the components and importance of fitness in the interaction between individuals with and without IR alleles in nature. The model may be usefully modified to include pathogens and seed predators that work either on the seed while it is still retained on the parent plant or in the seed bank. However, the model is also of significance to the basic ecology of sexually compatible invasive-native pairs where the active difference is in vulnerability to local herbivores, pathogens, or seed predators, whether the invasive individual is a crop, unwanted alien, or new mutation³.

References

PROTEOMIC PROFILING AND UNINTENDED EFFECTS IN GENETICALLY MODIFIED CROPS
Sirpa O. Kärenlampi and Satu J. Lehesranta

It is generally accepted that traditional food is safe for the majority of consumers. For the introduction of a new variant or cultivar developed from a traditional crop plant, maximum limits have been set in some cases, e.g., for potato and oilseed rape, to the content of known toxins. The requirements are much more stringent if the crop is developed by using genetic engineering. Why is it so? In a majority of cases seen so far, a new gene, often derived from other plants or microbial species, has been introduced to a non-predetermined location in the plant genome. It is quite feasible to ask the question whether the new gene products are safe or not. Therefore, for all genetically modified crop plants, the safety of the newly introduced proteins needs to be demonstrated before the plants can be released into the market.

Another point of concern is the random integration of the new gene into the plant genome. Both the new gene itself and its site of integration may give rise to unintended adverse effects. For example, transgene integration might interrupt regulatory sequences or open reading frames leading to novel fusion proteins and, thereby, modify plant metabolism.¹ These modifications could compromise the safety of the food crops by, for instance, leading to the production of new allergens or toxins. Having the gene and the integration site well characterised should provide a good basis for the safety assessment. However, it is a common practice today to perform a large number of analyses, so-called targeted analyses, to demonstrate that the characteristics of the novel crop are comparable with those of the conventional counterpart, in addition to the intended alterations. Targeted analyses include key macronutrients, micronutrients, antinutrients, and toxins. In certain cases, toxicity studies on experimental animals are advised. And yet, the question about the unintended effects does not seem to be covered in a way that would escape all criticism.

Cellini et al.² have considered transgene integration in the context of naturally occurring DNA recombination. It is well known that genetic variation is the cornerstone of plant breeding. Natural chromosomal recombination plays a central role in generating new variation. Non-homologous end joining, which is the predominant form of recombination in plants, rarely occurs without any sequence alterations, and usually gives rise to deletions of up to more than 1 kb and introduction of new filler DNA. Since the double-strand break repair system involved in recombination is more error-prone in plants than in other organisms, errors that change the original sequence occur at a very high frequency. The fact that gene-rich regions (and genes) are hotspots for recombination has facilitated the emergence of novel characteristics in crop plants.

Integration of exogenous DNA (transgene) occurs via the same mechanism as natural recombination. Several types of rearrangements are thus observed, both in transgene integration sites and in natural recombination sites. While this mechanism provides a selection of natural variation for breeders, it is also a source of unintended effects similar to that in genetically engineered crop plants.

In the light of variation generated by natural recombination and by the repertoire of conventional breeding technologies exploited for decades, the question is how much variation in the overall genetic makeup of a crop plant might be generated by the transfer and integration of a single gene, compared to the variation already existing. A related question is how probable are the unintended effects that extend beyond this variation.

To answer these and other questions, we made a comparative analysis of eight GM lines of potato, including vector-only lines without the target gene.³ The parent cultivar, Desirée, and a line that had undergone tissue culture only, were included as non-GM comparators. Nine of 730 proteins showed statistically significant differences among the GM lines and controls. No new proteins that would be unique to the individual GM lines were observed. The conclusion from this study, supported by the EU-funded GMOCARE project, was that there was no evidence for any major changes in protein patterns of the GM lines tested.

It can be argued that proteomics is not sensitive enough to find differences between potato lines or varieties. The European breeders have developed a large number of very different potato cultivars, many of them with genes introgressed from other Solanum species. Of that diversity, we analysed 32 non-GM potato genotypes, including 21 conventional cultivars, eight landraces, and three lines of S. phureja. From that study it was obvious that there is a great deal of variation in the protein patterns of the different potato genotypes: out of 1111 protein spots analyzed, 1077 differed significantly among two or more genotypes. The protein profile of the diploid species S. phureja could be clearly distinguished from the ones of the tetraploid S. tuberosum genotypes.

These studies indicated that the variation between the non-GM cultivars/genotypes was much greater than the differences between the GM lines. This was further confirmed by direct comparison of some of the GM lines with two non-GM genotypes; there was no separation among the GM lines and their control, but the two non-GM genotypes separated very clearly from each other and from all Desirée-based lines. In other words, there were considerably fewer differences between the GM and non-GM lines of the same genetic background than between different non-GM cultivars. Many of the proteins that contributed to the separation of the non-GM genotypes appeared to be involved in disease and defense responses, sugar and energy metabolism, or protein targeting and storage, and are presently considered to convey no safety risk.

Our results have been corroborated recently by Catchpole et al.,⁴ who compared several GM potato lines and cultivars using metabolic profiling. The authors found differences between the GM lines only in those metabolites that were targets of the genetic modification; apart from those compounds, the GM lines could not be distinguished from their controls. On the other hand, all cultivars could be clearly distinguished from one another.

The results of both profiling studies are not surprising, considering what is now known about the nature of plant genome and its dynamics. Even though genetic modification does not generate major changes apart from the ones targeted, a protein identified at an increased level in the GM line compared to the conventional counterpart might be worth further attention if the level clearly falls outside the normal variation. This is to exclude any risks from, for example, potent allergens. As current profiling methods produce a huge amount of data, it is almost inevitable that some statistically significant differences will be found. Therefore the focus should be in truly consistent differences.

How feasible are profiling techniques in general as tools to provide additional data for the risk assessment of GM crops? Do they provide added value worth the investment? Do they give reassurance that unintended adverse effects have not occurred? Non-targeted methods, such as transcriptional, protein, and metabolite profiling, offer potentially unbiased approaches to the detection of unintended effects. Of these, transcriptomics is possibly the most comprehensive, with full genome arrays currently available for a limited number of plant species. While it is clear that a comprehensive coverage of all proteins and metabolites present in a given tissue is difficult to obtain with current technologies, proteins are the key molecules of interest, as they are potential allergens and catalyse the synthesis of metabolites, some of which are potential toxins.

To assess observed differences within the context of natural variation in composition, comparative data of 'normal' protein levels are needed to understand the effect of genetic background, developmental stages, physiological states, environmental conditions, and cultivation techniques, and to be able to set the criteria against which a determination of a significant difference worth considering as a possible safety risk can be made. Currently there is very little information publicly available on protein patterns in potato tubers or in any other crops.

As with other profiling methods, proteomic screening is not yet routine for assessing the safety of GM products. However, proteomic profiling has the potential to reduce uncertainty by providing much more information about crop composition than does targeted analysis alone, especially in combination with other profiling methods. In addition, multivariate statistical methods can give a much better overall picture of how the given samples relate to each other than does the comparison of single compounds. These facts may make proteomics increasingly important when developing second generation GM crops with multiple genes, engineered metabolic pathways, or edible pharmaceuticals.