Selection can result in populations becoming more adapted, yet even in the presence of selection other effects can constrain adaptation. Two of the effects that can cause this discrepancy are genotype by environment interactions (GxE) and the accumulation of deleterious alleles. GxE is observed when an allele has different effects depending on the environment and can result in the maintenance of genetic variation, particularly when no genotype is best adapted to all environments. Deleterious alleles are generally considered as those which are deleterious across the conditions members of a species might encounter. Although they experience negative selection, they can still contribute a substantial proportion of genetic variance.

In this dissertation, I analyzed GxE and deleterious alleles in maize. First, I investigated GxE in a maize mapping population. We find that GxE contributes a substantial amount to the phenotypic variance for many traits. While we identify loci contributing to GxE, overall most of the GxE variance may be due to unidentified polygenic effects. Estimating the genetic covariances between traits in each environment reveals large differences in the genetic variance-covariance matrix between environments and in particular shows that differences in selection on flowering time may be contributing to the observed GxE for yield. Second, we analyzed the distribution of structural variants in maize inbred lines along the genome. We find that structural variants are more depleted in constrained regions of the genome than single nucleotide polymorphisms, possibly indicating that structural variants are more likely to be deleterious. Finally, we apply a machine learning method to identify constrained regions of the maize genome based on population genetic data. These predictions allow us to assess more recent evolutionary constraint in maize and find regions where mutations are more likely to be deleterious.

This dissertation explores the ecology and evolution of polyploid species, or organismswith multiple sets of chromosomes from whole genome duplication (WGD). Polyploidy is a widespread phenomenon across the tree of life, significantly impacting both animal and plant evolution. Recent advancements in genomics technology has made the study of polyploid species substantially more feasible, but the complexity of polyploid biology is still limiting. As a result, many questions on the evolutionary role of polyploidy remain unanswered. Here, I study the origin and ecological role of polyploidy in Andropogon gerardi Vitman, a mixedploidy and ecologically dominant prairie grass, using a novel reference genome and whole genome sequencing data. We found mixed-ploidy in A. gerardi is a result of recurrent polyploid formation, or happy little accidents. Further, we found WGD in A. gerardi confers immediate adaptive phenotypic changes. In the course of this research, we fortuitously assembled a reference genome for Poa pratensis, an economically valuable and popular turfgrass, instead of a second A. gerardi genome. Finally, I synthesize the challenges of variant calling in polyploids due to extensive genomic diversity and a lack of genomic resources and I propose a variant calling pipeline to addresses key challenges. Overall, this dissertation enhances our understanding of polyploidy’s role in plant evolution and environmental adaptation.

The search for quantitative trait loci (QTL) that explain complex traits has been ongoing in many species.Here we discuss the results of using a MAGIC population of maize lines created from 16 diverse founders to perform QTL mapping for agronomic and gene expression traits. We compare three models that assume bi-allelic, founder, and ancestral haplotype allelic states for QTL and show they have differing power to detect QTL for a variety of complex traits. Although the founder approach finds the most QTL, all methods are able to find unique QTL, suggesting that each model has advantages for traits with different genetic architectures. A closer look at a well-characterized flowering time QTL, qDTA8, which contains vgt1, highlights the strengths and weaknesses of each method and suggests a potential allelic series. Overall, our results reinforce the importance of considering different approaches to analyzing genotypic datasets, and shows the limitations of binary SNP data for identifying multi-allelic QTL.

Gene expression is a crucial intermediary between genetic variation and complex traits, and many studies have sought to incorporate gene expression in QTL mapping and GWAS to determine the underlying mechanisms of QTL for complex traits.However, the success of these integrative eQTL-QTL studies in identifying candidate genes and explaining genetic variation in complex traits has been mixed. We tested for association between effect sizes of overlapping local-eQTL and QTL and identified 5 promising candidate genes for two flowering time QTL. Generally, we found that local-eQTL for single genes tended to explain QTL for complex traits better than eQTL for factors of correlated genes, although we did identify a factor trans-eQTL moderately associated with a QTL for grain moisture. We show evidence of both historical and recent stabilizing selection acting on gene expression traits and find a significant relationship between the correlation of local-eQTL and complex traits and the significance of the local-eQTL. These findings provide support for the strong impact of stabilizing selection on gene expression traits, particularly for genes associated with complex traits.