Imaging genetics of seed performance

The Netherlands has a long history of plant breeding which has resulted in a leading position in the world with respect to the sales of vegetable seeds. Nowadays high-tech methods are used for crop-production which demands high standards for the quality of the starting materials. While breeding has mainly focused on crop yield and disease resistance in the past, it now becomes equally important to create seeds that rapidly and uniformly germinate under a wide range of production environments. A better understanding of the molecular processes that are underlying seed quality is a crucial first step to enable targeted breeding. In this thesis we describe the results of new methods that were used to map the genetics of seed germination. For this research we have used the leading plant science model species Arabidopsis thaliana which has a short generation time and a fully sequenced genome. Further, the large scientific community working on this model species is providing a wealth of resources ranging from large collections of worldwide accessions, genetic mapping populations, mutants and knowledge about gene, protein and metabolite action. A disadvantage of using Arabidopsis is the small size of the seeds, which requires evaluation of the germination of individual seeds with the use of magnifying glasses. This problem has been solved by using image analysis to create an automated procedure to obtain detailed information for parameters such as rate, uniformity and maximum germination. This procedure, called ‘the Germinator’, is described in Chapter 2 and has been enthusiastically adopted by the seed community. Plants cannot walk away from the environment at which the seed is dispersed. To survive and to enable reproduction, plants adapt to the prevailing environment which results in considerable genetic variation. This ‘natural variation’ is a great resource to study the mechanisms of adaptation. In Chapter 3 we have used two distinct Arabidopsis accessions, one originating from Germany (Bayreuth) and the other from high altitude in the Pamiro-Alay Mountains in Tadjikistan (Shahdara). In contrast to the Bayreuth accession, the Shahdara accession is well adapted to survive harsh conditions and is known to be stress tolerant to a range of environments. A genetic mapping (recombinant inbred line; RIL) population, consisting of 165 lines, that was derived from these two accessions is therefore particularly suitable to locate the genomic regions with genetic differences that influence seed germination. Such genomic regions are commonly referred to as quantitative trait loci (QTL). With help of the Germinator system we were able to evaluate germination of this RIL population under many different conditions. This resulted in a description of the ‘genetic landscape of seed performance’ in which we identified many QTLs for Arabidopsis seed germination. QTL regions are often large and identification of the causal gene requires intensive follow up research. We therefore aimed for a high throughput analysis using modern ‘omics’ techniques to analyze differences in metabolite levels and gene expression between the lines. A method to classify and visualize the vast amount of data derived from such an approach is described in Chapter 4. The so called genetical ‘omics’ experiments are expensive and therefore often force researchers to limit their study to a single developmental stage or environment only. A novel generalized setup overcomes this limitation and was tested for metabolite level changes in Chapter 5. This setup offers a unique reduction of experimental load with minimal effect on statistical power and is of great potential in the field of system genetics. Four different developmental stages of seed germination were tested in the RIL population. This approach resulted in a large dataset for which efficient analytical procedures were lacking. Thus, Chapter 5 also includes a description of a newly developed statistical procedure to analyze this type of data. The same approach and material were used in Chapter 6 to evaluate the genetics of genome wide gene expression.  Another approach to zoom in on the molecular mechanisms underlying seed performance is described in Chapter 7. Here, the genetic diversity was maximized by using 360 different Arabidopsis accessions which had been subjected to ultra-high density genotyping. In potential, such a genome wide association (GWA) study can provide high resolution mapping of genetic variation resulting in only a few candidate genes per association for the phenotype under study. Although we were able to replicate experiments over two years with a high level of heritability, no significant associations were found. This emphasizes the need to critically review the power of such an approach for traits that are expected to be determined by many small effect loci. Finally, closing in on the molecular mechanisms underlying the seed traits that we studied might be possible by a full integration of the datasets that were described in the different chapters. Two examples that show the potential and the complexity of such integration are described in the General Discussion (Chapter 8). Research focused on seed quality does not end here but has gained an impulse by the described new methods and hypotheses to continue on both the fundamental and applied level in the coming years