Effects of reduced crossover recombination on quantitative trait analysis

During meiosis, i.e. the production of the reproductive cells, the number of chromosomes in the cells is halved to ensure that after fertilization the typical quantity of chromosomes is restored. Crossover recombination is an essential step in this process that serves a dual function. On the one hand crossovers ensure the proper segregation of homologous chromosomes and thus ensure the reduction in chromosome number. In addition, crossovers result in the exchange of genetic material between homologous chromosomes. Reshuffling of genetic material is useful for identifying the effects of certain chromosomal segments on phenotypes when specific chromosome segments co-segregate with specific traits. But if this happens at multiple locations and chromosomes simultaneously it can also obscure the detection of those genetic effects. This thesis addresses the question as to how crossover recombination affects the detection of quantitative traits.

The effects of crossover recombination on the detection of quantitative trait loci (QTLs) are explored in this thesis by using doubled haploid (DH) populations in Arabidopsis thaliana. As opposed to more traditional mapping populations like recombinant inbred lines, these doubled haploids show lower numbers of crossover events since a doubled haploid experienced only one round of meiosis. In this thesis, some doubled haploids were derived from a modified meiosis because of which they have reduced crossover recombination, or show the complete absence of crossovers. DH derived from a hybrid without crossover recombination are so-called chromosome substitution lines (CSLs). In multiple other species, CSLs are credited to detect QTLs with relative ease due to their simple genetic architecture. Additionally, combinations of CSLs can be used to detect genetic interactions. In this thesis it is explored how genetic analysis can be performed with different sets of CSLs and doubled haploids with reduced crossover recombination. Chapter 1 further elaborates on the potential use of CSLs and some of the basics of quantitative trait analysis and QTL mapping are explained.

When doubled haploids are made in Arabidopsis, one first produced monoploid (haploid) plants that then give rise to doubled haploids. In Chapter 2 monoploids and their subsequent doubled haploids were generated, and their phenotypes compared in a genetic analysis to assess the effect of ploidy on their phenotypes. This resulted in detection of QTLs specific for a single ploidy level and QTLs that were expressed in both mono- and diploid generations. The ploidy specific QTLs indicated a different response of the genotypes to the change in ploidy, which was hypothesized to be a response to the sterility of the monoploids. The DH population itself was also used in a QTL mapping approach for flowering time measured in vernalized and non-vernalized conditions. This resulted in the detection of genotype-by-environment QTLs. Although this confirmed that reduction of crossover recombination (in Doubled haploids) does not per se influence genetic mapping, in the following chapters active measures were taken to repress the number of crossover recombinations.

In 2014 reverse breeding was presented as a method in which the parental lines of an F1 hybrid could be recreated by the combined use of complete suppression of crossover recombination through a stable knock-down of the meiotic recombinase DMC1 followed by the generation of doubled haploid plants from F1 hybrid. In Chapter 3 the same approach was used to obtain offspring with no crossover recombination to collect all possible CSL genotypes of a biparental cross. Additionally backcross populations of single chromosome substitution lines (sCSL) with their respective recurrent parent were created. Each of these populations represent a family of lines that segregate for a single chromosome. While the CSLs serve to identify the QTL effects, these single chromosome segregating families were used to finemap major QTL effects. In this thesis it is shown that also for Arabidopsis QTLs can be identified with relative ease in CSLs. Additionally, by acquiring all the different combinations of CSL genotypes, one can also look for genetic interactions between chromosomes. Genetic interaction effects are assumed to occur when the phenotype of an individual in which two independent loci are substituted cannot be predicted based on the individual effects of those two loci. In Chapter 3 such interaction effects between chromosomes are identified for two traits (flowering time and main stem length). Such genetic interaction effects are shown to have effect sizes equal to typical QTLs effect sizes. This demonstrates the advantage of discovering possible genetic interactions using such CSL populations.

The experiment to detect epistasis (i.e. genetic interactions) using a panel of all possible CSL genotypes in Chapter 3 was limited to only two phenotypes. This did not allow generalizations on how frequent epistasis is observed. This question was in part answered in Chapter 4, where a shotgun-proteomics approach was used to detect main effects and interaction effects between chromosomes for the abundance of proteins. Especially a small but genetically diverse panel of CSLs allows the detection of genetic effects at the proteome level, which was so far not technologically feasible in a cost-efficient approach. Only a subset of CSLs was used, which illustrates that a limited set of fourteen genotypes with single and double substitutions in a recurrent background can be used to detect genetic effects. Using protein abundancy as a phenotype, more than a thousand QTLs divided over the different chromosomes were identified. Especially chromosomes 2 and 5 were identified as contributors to the proteome variation. Furthermore, it was shown that the abundance of approximately 20% of the measured proteins was significantly dependent on the presence of a combination of two chromosomes (i.e. an interaction effect). Furthermore, for several proteins their abundance was not influenced by the allelic state of any single chromosome, but they were only significantly influenced by a combination of chromosomes. This shows that interaction effects can also be detected when a limited number of CSLs are used.

The previous chapters deal with confirming the utility of CSLs, Chapter 5 addresses several practical difficulties of the application of reduced crossover recombination to advance plant breeding practice. Because proper homolog segregation is compromised in the absence of crossover recombination, plants produce high numbers of non-viable spores and have a very low fertility. In Chapter 5 the effect is studied of the knock down of MSH5 instead of DMC1. With a dysfunctional MSH5 crossover recombination is not completely absent but suppressed to retain still a few crossovers per meiosis. Therefore, proper homolog segregation occurs more often, and plants remain more fertile. Additionally, instead of stable suppression via transformation of a parental line, virus-induced-gene-silencing (VIGS) is used to transiently downregulate the gene MSH5 in the meiosis of a hybrid directly. Both revisions together ensure transgene-free progeny in only three generations with a few or no recombination events per offspring. Using such progeny, the original starting hybrid could be recreated using complementing non-crossover offspring which did not differ phenotypically from the original hybrid. Additional intercrosses were made between non-crossover and offspring containing few crossovers that were complementary for most of the genome to produce hybrids with small homozygous segments. In general, such near-full hybrids did not differ phenotypically from the original hybrids with only few exceptions. This suggests that one could improve hybrids directly by fixating specific segments containing for instance recessive QTLs. The improvements to reverse breeding make this breeding approach more efficient and flexible for application to higher chromosome number species.

In Chapter 6 a discussion follows on the previous chapters. Also new and additional experiments are proposed based on the observations described in the different chapters. It is argued that a complete CSL panel with all possible CSL genotypes can give a better overview of the genetic architecture controlling quantitative traits, especially when multiple interacting QTLs are to be expected. Smaller panels of CSLs might be informative enough for the purpose of genetic mapping in high chromosome number species, and practical considerations should include the type of genetic effects one wants to identify and the known genetics of the considered trait. Taken together, the work in this thesis shows that with different approaches reduced crossover recombination can additional advantages on the analysis of quantitative traits.