Breeding strategies for integrating new and orphan crops in production systems

In recent decades, plant breeding has shifted toward catering for intensive monocropping based on a few staple crops. This shift has led to a reduction in biodiversity, environmentally-unsustainable agricultural practices, the increased exposure of food production to climate change risks, and nutritionally inadequate diets. More sustainable agricultural practices that promote species- and nutrient-rich crops are needed, and new and orphan crops are attractive candidates for inclusion. Many orphan crops are typically grown in diverse cropping systems with low inputs, cope well with adverse climatic conditions, and are nutrient-rich. However, they have received limited research attention and have generally not yet been intensively selected or domesticated. This thesis addresses several aspects of plant breeding research that could support new and orphan crop improvement. Chapter 1 introduces concepts essential for understanding the thesis as well as thesis objectives. The allotetraploid orphan crop finger millet Eleusine coracana is a promising crop for further cultivation due to its climate resilience and rich nutritional properties. My first results chapter (Chapter 2) presents a study of phenotypic and genomic diversity in finger millet based on 423 landrace accessions collected from major cultivation areas worldwide. The application of various statistical and bioinformatics tools, building on a recently separately published genome assembly, revealed strong population structure at both phenotypic and genomic levels, with differences at the sub-genomic level shedding light on potentially complex diversification history of the crop. Extensive phenotypic variation was found in almost all of the 13 agronomic traits analysed, while genomic diversity in landraces originating from East Africa was relatively low considering this area is the primary centre of domestication. The results of this study provide a better understanding of the crop that support context-specific breeding across major cultivation regions as well as germplasm conservation. Intercropping is a promising agricultural practice for promoting diversification that is a potential entry point for new and orphan crops, as they are traditionally adapted for diverse systems and have not yet been extensively selected for monocropping. My second results chapter (Chapter 3) develops new breeding strategies for intercropping, a topic that has been largely ignored because of the challenges with traditional phenotype-based selection (e.g., costly and budget-limited field trials). Stochastic simulations were used to test the potential utility of genomic-assisted tools, particularly genomic selection. Four intercrop breeding programs using genomic selection were simulated that produced one to three times higher genetic gain per unit cost compared to a phenotypic intercrop breeding program, depending on the simulated genetic correlation between monocrop and intercrop grain yields, and the total operating cost of the breeding program. The results of this study suggest that genomic selection could be used to revitalize intercrop breeding research embracing new and orphan crops. The use of appropriate statistical models that allow for analysis of data collected on multiple traits and multiple environments within plant breeding programs is critical for increasing genetic progress in resource-limited breeding programs for new and orphan crops. Although breeders select promising genotypes based on multiple traits measured multiple environments, current statistical approaches are typically limited to analysis of multiple environments for one trait or multiple traits for one environment. My third and final results chapter (Chapter 4) develops a new genomic selection approach for analysis of data consisting of multiple traits and multiple environments, which we call the multi-trait multi-environment factor analytic linear mixed model (MTME-FA-LMM). The MTME-FA-LMM uses a genomic relationship matrix for genotypes and a joint factor analytic model for environments and traits to appropriately model genotype × trait × environment interaction, while also allowing a different genotype × environment pattern for each trait as well as a different genotype × trait pattern for each environment. In addition, the factor analytic selection tools were extended to multiple traits to provide measures of genotype overall performance and stability for each trait, summarized across environments, to allow breeders to perform simultaneous selection of promising genotypes based on multiple traits. To demonstrate the application of the new approach, a multi-trait multi-environment trial dataset from preliminary yield trials of a finger millet breeding program is used. Overall, my thesis demonstrates how various aspects of plant breeding research, including understanding germplasm diversity, developing new breeding strategies, and developing advanced statistical approaches, can support increased genetic progress in new and orphan crops to integrate them into production. My thesis concludes by presenting opportunities and challenges for future orphan crop breeding work (Chapter 5).