Local adaptation is the process by which local populations evolve traits that increase their fitness relative to individuals of foreign populations. Traditional maize landraces across the Americas have long been observed to exhibit local adaptations to abiotic environmental variables, chief among which is elevation. However, aside from field observations and studies of limited geographic and/or phenotypic scope, the strength and breadth of these local adaptations have remained unexplored.

To investigate local adaptations in maize landraces broadly, I conducted a broad-scale reciprocal transplant experiment with four populations of maize landraces (Mexican highland, Mexican lowland, South American highland, and South American lowland). I found that landraces grown at or near their native elevation have higher fitness than landraces from distant elevations. Several traits show higher between-population variance than expected given neutral genetic variance, suggesting that those traits are under divergent selective pressures. Furthermore, putatively highland-adaptive traits have strong, elevation-dependent correlations with fitness, further supporting their role in highland adaptation.

To investigate patterns of local adaptation across Mexico with greater resolution, I conducted Ecological Niche Modeling (ENM), Environmental Association Analyses (EAA), environmental prediction using mixed effect models. I characterized landrace niches by associations with numerous environmental variables, and found that not all landraces are equally locally adapted; highland landraces have narrower niches and higher niche overlap than lowland landraces. Also, niche models of genetic groups on average had wider niches, suggesting that landrace groups group adaptively similar accessions into categorical groups at least as well as Bayesian clustering algorithms. However, additive genetic variance estimated with a kinship matrix was a better predictor of environment than was landrace group, demonstrating the genetic foundation of local adaptation. Finally, I identify SNPs strongly associated with environmental variables that constrain landrace distributions, and assert that these SNPs are good local adaptation candidates.

In my third chapter, I use genotypic data collected in the reciprocal transplant experiment and further investigate the partitioning of adaptive genetic variation into groups across geographic space. I find again that landrace genetic population structure is largely shaped by geographic distance, latitude, and elevation, and is consistent with demographic historical migration and bottleneck events. To incorporate these complex patterns of relatedness across the Americas, I looked for phenotypic trends across the axes of variation of an eigen-decomposition of a genetic distance matrix which capture elements of genetic structure between populations that correlate to latitude and elevation. Traits that varied more or less than expected given neutral genetic variance were identified as under divergent or stabilizing selection, receptively. I also identified SNPs with high FST between highland and lowland populations in either Mexico or South America, or both, to find genes involved in highland adaptation unique to one continent or the other or parallel adaptation.

In summary, my research identifies strong phenotypic and genetic evidence of local adaptation in maize landraces across the Americas, and characterizes the local niches that have evolved.