Integration of phenotype, genotype and gene expression to unravel flower colour biosynthesisand complex plant quality traits in azalea (Rhododendron simsii hybrids)

Azalea breeding is a slow process, from the moment of crossing until the first flower can be evaluated, it already takes 3 years. At that time, selection is primarily based on the flower quality. Only when the seedlings have been cloned, other plant quality traits can be evaluated thoroughly. As a result, plants with attractive flowering are kept too long in the breeding cycle. Information on the inheritance of cultivation related quality traits is lacking, but the inheritance of flower colour has been well studied. Therefore flower colour was selected as a model system for genetical genomics in azalea. In this approach, phenotypic and genetic information is combined with gene expression data of candidate genes on a genetic map to unravel the regulation of the desired traits. In case phenotypic data are co-located with mapped candidate genes, these genes are proven to be directly involved in the creation of the phenotypic variation of the trait. Nevertheless, it is very likely that not the genes themselves but transcription factors are the switches that regulate the phenotype of the trait. In that case, phenotype and genotype will be mapped at different positions, but phenotype is then expected to be mapped together with the true regulators, the transcription factors. eQTL mapping of gene expression profiles from candidate genes will integrate this information on the genetic map.
Ultimately, our goal was to implement this model on the more complex plant quality traits as leaf morphology and plant architecture. However, candidate genes for these traits are lacking in azalea so far. Therefore, we limited ourselves to study the segregation of these complex traits in diverse genetic backgrounds and to evaluate the advantages of multi-population QTL mapping. For this purpose, leaf morphology (both colour and shape) and plant architecture have been scored in four unrelated populations. QTL mapping being the purpose, image analysis was used whenever possible to generate continuous, highly informative data. Both classical parameters and symmetrical elliptic Fourier descriptors performed well in describing leaf morphology. Image analysis resulted in a massive amount of data that had to be combined in principal component analysis. Leaf colour was split up in RGB values but was not always informative to discriminate between populations. Plant architecture was the most complex trait under study. Division in well-defined sub-traits as e.g. shoot length, plant area and number of shoots allowed to draw some preliminary conclusions on the inheritance of these complex traits. However, QTL mapping was the ultimate purpose with these data.
Genetical genomics requires candidate genes to be mapped. In azalea this was only possible for the flower colour biosynthesis genes of the anthocyanin pathway. However, EST data in azalea could also be explored in order to develop non-characterised but functional markers in conserved genomic regions. Both intron-spanning EST-based markers and EST-SNPs were developed and the advantages of both marker types were discussed. HRM-based EST-SNP mapping is clearly the method of choice for future candidate gene mapping. The EST-based functional markers were, together with myb-based markers as candidate genes for transcriptional regulation of flower colour, scored in a single mapping population. AFLP and SSR markers were analysed in all four populations as a reference backbone. An integrated framework map on the four individual linkage maps was constructed. A combination of regression mapping (JoinMap) and multipoint-likelihood maximisation (Carthagène) enabled the alignment of the four maps on the basis of framework markers. This facilitated in turn the alignment of QTL regions detected in individual populations for the complex traits under study. The focus was on the detection of multi-population QTLs, since these QTLs were assumed to be more conserved in the overall azalea gene pool and therefore more apt for breeding purposes. No such QTLs for branching were detected, plant architecture resulted in one combined QTL. Several leaf morphology QTLs were found, many of them were related to the leaf width to length ratio which is the most discriminative factor for azalea leaves. Overall, populations ‘CxD’ and ‘GxH’ were more appropriate for leaf morphology studies, whereas population ‘ExF’ resulted in the highest number of QTLs related to plant architecture.
Flower colour was selected as a model system for genetical genomics. In the same way as for the other traits, image analysis was used to turn flower colour into a variable suitable for QTL mapping. Gene expression profiles of the anthocyanin biosynthesis pathways were also to be added onto the genetic map by means of eQTL mapping. This required the establishment of a reliable RT-qPCR protocol for transcriptional profiling. RT-qPCR is a highly sensitive method that required intensive optimisation. Unfortunately, the expression profiles could not discriminate between the flower colour groups. However, these data could be used to identify the genetic regulatory loci explaining the observed variation in flower colour by eQTL mapping. The minimal population size required to have sufficient power for eQTL mapping was set at 70 plants. The gene expression profiles were therefore determined on a subset of 70 siblings of the mapping population segregating for flower colour. eQTLs in combination with QTLs for the flower colour phenotype were positioned on the genetic linkage map. Since the map also contained functional markers for flower colour biosynthesis, the integration of phenotypic, genetic and transcriptional information allowed to elucidate partly the process of flower colour biosynthesis in azalea. Coordinated expression of the early pathway genes was confirmed by eQTL mapping of these genes in the region of a myb-fragment. RT-qPCR proved to be a good method for expression profiling in a genetical genomics approach and even has some advantages compared to conventionally applied micro-arrays. The co-localization of the FLS gene expression with a QTL for the co-pigmentation phenotype at the locus Q served as a proof of concept. No solid conclusions could be drawn yet concerning pink and white. However, the first is most likely a gene-dosage effect and the latter seems to be regulated at the transcriptional level.