Comment from GM-Free Cymru: We need to study this article more closely, but it appears to show that GM wheat is unstable, non-uniform, and sometimes incapable of coping with environmental stress in the field. This means it is increasingly vulnerable to disease as compared with the isogenic or non-GM lines from which it is bred. It also appears that yield is sharply lower than the yield predicted from greenhouse experiments; in two of the GM lines tested, yield dropped by a massive 56% -- maybe because of the 40-fold increase in the incidence of ergot disease.
We think this study might be of very great significance, since consents are often given for GM varieties on the assumption that they are stable, uniform and associated with yield increases........... and if none of these assumptions is born out in reality, why would anybody want to grow GM wheat such as the controversial wheat currently on trial at Rothamsted? Above all, why would any committee of scientists want to give it a clean bill of health and recommend an EU approval?
Once again, we have a demonstration that the concept of substantial equivalence is total nonsense which has nothing do with either good science or the real world. It exists only in the minds of the very silly people who work in EFSA's GMO Panel and in committees such as ACRE and ACNFP in this country.
Zeller, S. L., O. Kalinina, et al. (2010).
BACKGROUND: The introduction of transgenes into plants may cause unintended phenotypic effects which could have an impact on the plant itself and the environment. Little is published in the scientific literature about the interrelation of environmental factors and possible unintended effects in genetically modified (GM) plants. METHODS AND FINDINGS: We studied transgenic bread wheat Triticum aestivum lines expressing the wheat Pm3b gene against the fungus powdery mildew Blumeria graminis f.sp. tritici. Four independent offspring pairs, each consisting of a GM line and its corresponding non-GM control line, were grown under different soil nutrient conditions and with and without fungicide treatment in the glasshouse. Furthermore, we performed a field experiment with a similar design to validate our glasshouse results. The transgene increased the resistance to powdery mildew in all environments. However, GM plants reacted sensitive to fungicide spraying in the glasshouse. Without fungicide treatment, in the glasshouse GM lines had increased vegetative biomass and seed number and a twofold yield compared with control lines. In the field these results were reversed. Fertilization generally increased GM/control differences in the glasshouse but not in the field. Two of four GM lines showed up to 56% yield reduction and a 40-fold increase of infection with ergot disease Claviceps purpurea compared with their control lines in the field experiment; one GM line was very similar to its control. CONCLUSIONS: Our results demonstrate that, depending on the insertion event, a particular transgene can have large effects on the entire phenotype of a plant and that these effects can sometimes be reversed when plants are moved from the glasshouse to the field. However, it remains unclear which mechanisms underlie these effects and how they may affect concepts in molecular plant breeding and plant evolutionary ecology.
NB Text of the article is freely available on the web.
Transgene × Environment Interactions
This study demonstrates that GM plants can differ in morphological, fitness- and pathogen-related traits from their control plants. We found several significant transgene (GM vs. control) × environment interactions; that is, depending on the environmental conditions the studied transgene against mildew infection had beneficial or detrimental effects on most of the investigated plant traits. GM plants generally benefited from glasshouse conditions with high mildew infection pressure when compared with control plants but showed a stress reaction when powdery mildew was absent due to fungicide spraying. It is possible that the GM plants lacked the energy to cope with the stress caused by this treatment or the chemical itself could have interacted with the transgene or with pathways involved in Pm3b-mediated resistance. It is conceivable that the high fungicide dose increased the extent of the stress reaction of GM plants.
Similar to the fungicide treatment in the glasshouse, the natural conditions outdoors seemed to have stressed the GM plants in the field to the extent that their fitness was significantly reduced. Possible causes of environmental stress in the field were drought and neighbour competition. The only deliberately manipulated factor, i.e. fertilizer application, modified the transgene effects only in the glasshouse but not in the field. Apparently the transgene only offered a relative fitness benefit to GM plants growing under conditions of high mildew incidence but low levels of other stresses. These were exactly the conditions met in the glasshouse but not in the field (nor in the glasshouse after fungicide application). Under less beneficial conditions, the GM plants may have paid a physiological cost for the high intrinsic mildew resistance .
Differences among GM Lines
The four GM lines, which each contained a single copy of the identical transgene in homozygous condition, differed significantly from each other. There are several potential reasons for these differences. It is possible that cell culturing caused somaclonal variation among the four offspring pairs which subsequently might have interacted differentially with the transgene . Although theoretically possible  we would not expect that such interactions would be stably inherited over five plant generations as we found it here. It seems unlikely that random somaclonal events would cause similar effects in two of the four independently transformed lines (Pm3b#2 and #4). A more plausible explanation for the differential effects of the inserted transgene among the four offspring pairs may be that positional effects caused the line-specific differences. Several processes are known to cause such effects . Firstly, an inserted transgene may disrupt native genes. Because spring wheat is hexaploid, consists of more than 80% repetitive, non-genic DNA sequences and each GM line was created by a single insertion event, it is unlikely that the disruption of coding genes or their regulatory sequences could have caused these differential effects , . Secondly, the insertion position of a transgene into the genome may have affected its expression level. Studies have shown that transgene expression rates and activity patterns of independently transformed wheat lines with constitutive ubiquitin promoters can vary . Depending on the insertion site, flanking DNA regions may partially silence the inserted promoter. Head-to-tail arrangements of the transgenes, in our case of the Pm3b and the selectable marker gene, could also have a negative influence on the promoter activity . It is also possible that in some lines the transgene was inserted into a region of the genome with low transcription activity .
The semi-quantitative expression analysis (Figure S1) indicated that the expression of the Pm3b transgene did differ between the four GM lines. Thus, although we lack confirmation by quantitative expression data, it appears that the two GM lines Pm3b#2 and #4, where the transgene showed the strongest phenotypic effects, also had the strongest transgene expression. Obviously, this hypothesis should be tested with a much larger number of lines differing in expression levels. However, such a study currently would be beyond our capacities to obtain funding and permissions for field trials. If the hypothesis could be confirmed, there would still be the question whether the overexpression of the transgene led to an overabundance of its protein product and the subsequent phenotypic effects or if other mechanisms would be involved.
Besides the quantitative reduction of fitness, we observed that some spikes of the two lines Pm3b#2 and #4 also differed in their morphology during flowering time and that the same two lines were also more heavily infected by ergot fungus than the other two GM lines and the four control lines. The altered spike morphology may have increased the likelihood of ergot spores entering the florets . However, no indications of altered spike morphology were observed in the glasshouse.
Implications for Molecular Plant Breeding
Although transgenic plant lines with unintended phenotypes commonly arise during molecular plant breeding ,  they can usually be detected earlier and more easily and are thus not further investigated  and published. The development of commercial GM plants is based on long selection processes that start in the glasshouse and end in the field. Enormous numbers of seedlings are already discarded before they are exposed to realistic field settings. Our results may have implications for molecular plant breeding: some of the best GM lines in the glasshouse may still show aberrant performance in the field and some not so promising GM lines in the glasshouse may actually be the best for the field. They would likely be lost at early stages of a selection process only targeted at maximum performance under a particular environment. Based on our glasshouse findings, line Pm3b#1 would have suffered this fate yet was the best in the field. One lesson from our study and from genotype × environment studies in general , , ,  is that lines which perform particularly well in a specific environment may pay a cost of specialization and perform poorly in other environments.
Our study demonstrates that inserting a single transgene into the hexaploid wheat genome, along with the desired target effect such as mildew resistance in the present case, can significantly affect other phenotypic traits and thus, as in our case, change the ecological behaviour of the species (hypothesis (i) in Introduction). Such unintended effects of single genes to our knowledge are always smaller in experiments using naturally occurring genetic variation and wild plants , . Even when we included crop plants, we could not find any publications where single genes reduced quantitative fitness traits in a plant as strongly as in the present case, yet only in the field and not in the glasshouse . Commercial glyphosate-resistant soybean cultivars were found to suffer from a 5% yield depression that might be caused by the transgene or its insertion process . One study tested wheat varieties with introduced resistance genes against leaf and stripe rust and reported a 12% reduction of yield , which was considered to be a very large effect . Compared with these, the yield reductions of 48 and 56% observed in our two GM lines of wheat expressing the Pm3b transgene are much larger (Figure 3B).
We found that the level of mildew resistance as well as the magnitude of other phenotypic effects varied significantly between different GM lines (hypothesis (ii) in Introduction). We hypothesize that this variation in phenotypic effects may be due to different expression levels of the Pm3b transgene which in turn might have been caused by different insertion positions of the transgene in the genome. Some plant breeders suggest not selecting for plant lines with complete pathogen resistance because costs of such a resistance often outweigh benefits . In our case this would speak for selecting GM lines with relatively low expression levels yet still increased mildew resistance, i.e. line Pm3b#1 . However, to test the hypothetical correlation between expression level and phenotypic effects would require specific experiments with a larger number of GM lines as used here. With regard to risk assessment our findings are in agreement with the view that a each GM line should be tested in a case-by-case approach .
Finally, our results show that even if desired phenotypic effects of a transgene are found across a range of environments in a glasshouse experiment, some of these effects can be reversed if GM lines are exposed to natural environmental variation in the field (hypothesis (iii) in Introduction). Although it is likely that commercial plant breeders know of the presence of transgene × environment interactions, it seems that such observations so far have not found their way into the scientific literature. Breeding trials to select lines for further investigation do not need full replication and randomization, yet for an assessment of the ecological behaviour of such lines, replicated and randomized ecological experiments would be required. Our study may serve as an example of potential results that can be obtained in such experiments. We believe that such experiments can help us to gain a deeper understanding of single-gene effects in plant ecology and evolution.