We have performed extensive cultivation work and isolated bacteria from the model plant Arabidopsis. Individual plants as well as individual leaves were sampled at different European sites to determine their core leaf community and to establish a reference strain collection. After identifying approximately 3,000 isolates using a high-throughput DNA sequencing-based method, we selected more than 200 representative strains belonging to 52 genera of the major phyla, covering the majority of the culture-independent taxonomic diversity in the phyllosphere. Draft genome sequences were generated of all selected isolates. Recolonization experiments with synthetic communities in a gnotobiotic model system showed reproducible colonization patterns and provided a valuable starting point to identify mechanisms of community formation and function. Examination of plant's responses to its microbiota revealed that the plant reacts differently to members of its natural phyllosphere microbiota and in a hierarchical manner. We also identified a conserved core set of genes that were consistently induced by the presence of a majority of strains. Remarkably, they also comprised the most differentially regulated genes across members of the microbiota and were predictive of the overall transcriptional reprogramming in the plant. We also found that these core genes are required for resistance to a bacterial pathogen, indicating that they constitute a defense adaptation strategy. In complementary work, we used the resource of leaf isolates and their genome to systematically probe the interaction of strains and mine for uncharacterized natural products. Genome analysis revealed more than 1,000 predicted natural product biosynthetic gene clusters, hundreds of which were unknown compared to a database of characterized clusters. We identified several novel natural product scaffolds that contribute to the observed antibiotic activity. We also analyzed the extent to which bacteria in complex communities in planta interact with each other or occupy their own independent niche. To study microbial interactions in planta, we performed drop-out and late introduction experiments with synthetic communities. In particular, we tested how the order of strain arrival shapes community structure. Our results showed that community assembly is historically contingent and subject to priority effects. Strains were able to invade, to varying degrees, an already established microbiota that was itself resistant and largely unaffected by latecomers. Additionally, our results identified key species that have the great potential to affect the plant community structure.