Spatial heterogeneity can dramatically impact evolution in bacterial communities, raising the question of whether spatial profiles of drug concentration can be tuned to slow the emergence of antibiotic resistance. In this work, we combine lab evolution experiments in spatially connected, computer-controlled chemostats with mathematical models to investigate resistance evolution in E. faecalis, an opportunistic bacterial pathogen. We find that both the rate of adaptation to doxycycline, a protein-synthesis inhibiting antibiotic, and the associated cost of resistance in the associated mutants depends strongly on drug concentration in spatially uniform populations. Interestingly, when spatially separated subpopulations are exposed to different concentrations of drug, adaptation can be dramatically slowed by tuning the rate of migration between habitats, leading to selection for phenotypically distinct resistant mutants. Our results highlight the rich evolutionary dynamics of adaptation in spatially connected habitats and indicate that resistance evolution can be slowed by balancing evolutionary trade-offs of migration and heterogeneity.