Application of omics technologies for environmental risk assessment of genetically modified plants : arabidopsis and modified defence mechanisms as a model study

As a result of rapid biotechnological developments in the past century, genetically modified (GM) crops were developed and introduced for field application. Despite the advantages of these crops and the professional marketing policies, people also started questioning the safety of GM products for humans and the environment. In response to that, scientific advisory bodies (such as COGEM, The Netherlands Commission on Genetic Modification) suggested that, among other measurements, an environmental risk assessment (ERA) of a GM crop should be done before introduction into the field. Ecological knowledge about the possible effects was considered a vital component of that assessment. In 2007, the Dutch Government initiated the ERGO (Ecology Regarding Gene-modified Organisms) research programme to generate a scientific basis for a sound ecological risk analysis. The main objective of the ERGO-programme was to develop ecology-based guidelines for how to best assess the possible ecological side-effects of new GM crops. Also the European Food Safety Authority (EFSA) recognised the interaction of a GM crop with non-target organisms as a potential environmental risk and therefore they provided guidelines for selection of a range of non-target organisms and phenotypes to be studied under laboratory conditions as part of a GM crop risk assessment study. These guidelines formed the basis for the ERGO research themes. Parallel to the new biotechnological developments leading to the introduction of GM plants into the environment, new analytical techniques were also introduced that revolutionized the field of analytical biology. High throughput analytical platforms, collectively called omics technologies, created opportunities for untargeted analysis of cellular components with biological and ecological functions including mRNAs (transcriptomics), proteins (proteomics) and metabolites (metabolomics). These analytical platforms were recommended by several researchers in the field of GM food/feed safety for the analysis and comparison of a GM product with its safe counterpart. However, EFSA failed to formulate concrete rules about the application of the omics platforms in GM risk assessment perhaps due to a lack of consensus about where and how to employ these technologies in the whole ERA of GM plants. In the ERGO programme, exploration of the potential to apply omics platforms for ERA of GM crops was therefore one of the objectives. This PhD thesis originates from one of the ERGO themes, assessment of the effect of genetic modification on non-target organisms. Under this theme with three PhD students a multidisciplinary approach was pursued to provide guidelines for how to evaluate non-target effects of GM crops altered in insect resistance using ecological methods as well as omics platforms. In this PhD thesis, I set out to find solutions for some of the limitations in the application of omics platforms such as the lack of a statistical method to evaluate the differences between GM vs. wild type plants at the omics level and the question what would be a fair reference for the judgement about the effect of genetic modification. As a model for the evaluation of the impact of genetic modification on the omics phenotype we used three insect defence traits that we introduced using genetic modification into several different Arabidopsis thaliana accessions. The first trait, indirect defence, was the production of the volatile (E)-nerolidol which has been shown to attract predatory mites that can control spider mites. The other two traits were direct defence traits and consisted of overexpression of the transcription factor (MYB28) to boost aliphatic glucosinolate biosynthesis and the introduction of Cry1from Bacillus thuringiensis encoding the Bt toxin that is effective against lepidopteran insects (caterpillars). As a reference for comparison of the effects of the genetic modification, we used a panel of wild type A. thaliana accessions that were selected in this study and publically available data of different accessions and individuals of a RIL population that together constitute the baseline, the variation present in the non-GM background germplasm. To allow for comparison of large datasets with this baseline, in Chapter 2 a statistical measure was developed, which we coined hyper-plane distance and which was used to assess the non-target effects of our genetic modification in transcriptomics as well as metabolomics analyses. In omics untargeted analyses, multivariate, hyper-dimensional data are generated, making global comparison of samples or groups of samples very difficult. In chapter 2 a method was developed to calculate a distance between the metabolome - analysed on three different metabolomics platforms - of genotypes or environments. Hereto, we employed principal component analysis (PCA) to reduce the number of analysed metabolites to a series of principal components (PCs) or dimensions of a PCA plot. The scores of the samples on a number of PCs, representing the relative position of samples to each other on those PCs, were subsequently used in an analysis of similarity (ANOSIM). In this manner, we used the variation in the samples’ PC scores to derive a distance between groups of samples on a multi-dimensional plot, the hyper-plane distance, in the case of metabolites called the metabolic distance. This distance represents between-group differences as well as within-group differences and therefore is a measure of the overlap between groups in a multi-dimensional context. Furthermore, it was also possible to statistically test the calculated distance in ANOSIM by permuting the samples’ scores to produce a P-value for the calculated distance. Hyper-plane distance gives a single measure for the difference between groups of samples in a PCA hyper-plane, something that is impossible visually with many samples of many groups in a multi-dimensional context. The metabolic distance was used to select metabolically diverged accessions of A. thaliana and to determine the impact of the environment on the metabolome of A. thaliana. The accessions thus selected (An-1, Col-0, Cvi and Eri) are representative for the metabolome diversity across the set of analysed accessions, and hence represent the baseline metabolome. Engineering A. thaliana to produce the volatile (E)-nerolidol was used to alter indirect defence in A. thaliana. In Chapter 3 several genetic engineering strategies were used to generate transgenic lines that uniformly emit sufficient amount of the volatile. Combination of the gene responsible for (E)-nerolidol biosynthesis (FaNES1) with the gene responsible for biosynthesis of its precursor, farnesyl diphosphate synthase (FPS1L), both equipped with mitochondrial targeting signal, resulted in higher production of (E)-nerolidol than with FaNES1 alone. The transgenic production of (E)-nerolidol in Arabidopsis also resulted in the formation of non-volatile conjugates. Adding also 3-hydroxy-3-methylglutaryl CoA reductase 1 (HMGR1), a rate limiting enzyme of the mevalonate pathway, resulted in a further increase in the production of (E)-nerolidol as well as its non-volatile conjugates. Transgenic A. thaliana plants emitting (E)-nerolidol were more attractive to the insect Diadegma semiclausum, which is an important endoparasitoid of the larvae of Plutella xylostella (cabbage moth). In Chapters 4 and 5 the chemical changes in and effects of transgenic A. thaliana accessions altered in indirect or direct defence on insect behaviour were characterised. In Chapter 4 the mitochondrial-targeted nerolidol synthase (COX-FaNES1) and the gene encoding the enzyme for the substrate (FPP) biosynthesis in mitochondria (COX-FPS2) were introduced into three A. thaliana accessions. Transgenic plants also emitted (E)-DMNT and linalool in addition to (E)-nerolidol. The aphid, Brevicoryne brassicae, was repelled by the transgenic lines of two of the accessions, although its performance on the transgenic lines was not affected. The aphid parasitoid, Diaeretiella rapae, preferred aphid-infested transgenic plants over aphid-infested wild-type for two of the accessions. Although another aphid predator, Episyrphus balteatus, did not differentiate between aphid-infested transgenic or wild-type plants, the results suggest that genetically engineering plants to modify their emission of VOCs holds promise for improving control of herbivores. In Chapter 5, MYB28 was overexpressed in three A. thaliana accessions. MYB28 overexpression had different effects (positive as well as negative) on the total aliphatic glucosinolate level in different transformation events of the same genetic background, possibly as a result of tight post-transcriptional regulation of MYB28. Furthermore, enhancement of the aliphatic glucosinolate pathway seems to be genetic background specific. Leaf damage by Brassicaceae generalist Mamestra brassicae and specialist Plutella xylostella were negatively affected by MYB28 overexpression, giving promises for improvement of chewing pest damage control. Higher glucosinolate levels as a result of MYB28 overexpression affected insect performance positively in the specialist and negatively in the generalist. Statistical analysis revealed the differential influence of certain structural groups of aliphatic glucosinolates on the two different insects. Chapter 6 demonstrates the application of the hyper-plane distance for the assessment of GM-mediated effects on the transcriptome. In this case, publicly available meta data containing the natural transcriptome variation in A. thaliana were proposed as a reference. Using this approach we showed that GM Arabidopsis lines with a novel indirect defence trait display changes in the transcriptome due to introduction of pleiotropic transgenes. However, the observed changes were well within the range of variation and plasticity in gene expression occurring naturally in A. thaliana. We also showed that unintended changes in the transcriptome are the result of other factors than the novel trait itself. This is an important observation because it implies that untargeted effects could be avoided or changed by using other strategies for transformation. In Chapter 7 all the transgenic lines generated in my thesis work were included in a metabolomics approach to study the effect of genetic modification on the metabolome level. The primary selected accessions of A. thaliana (Chapter 2) formed the baseline metabolome and the hyper-plane distance measurement was employed for analysis of differences. Untargeted metabolomics analyses using GC-TOF-MS and LC-TOF-MS of shoot and root material showed that the metabolome of most of the transgenic lines was substantially equal to the baseline even though the baseline did not yet include environment-induced metabolome variation. We suggest that substantial equivalence of a GM line’s metabolome with the baseline can be used to infer a low or even no risk of the particular genetic modification for non-target organisms and can be used as a first-pass criterion in the assessment of non-target ecological effects. Chapter 8 was written in collaboration with the two other PhD students from the same ERGO project. It summarizes and discusses the most important conclusions of the research done by the three PhD students and integrates the results in the form of guidelines for assessing the non-target ecological effects of a new GM crop. These guidelines suggest rules that must be taken into consideration when a request for permission for field trials or commercialisation of a new GM crop is submitted to COGEM.