Degrees of freedom (DoF) studies of wireless networks have contributed many fundamental insights into their capacity limits. One of the most critical determinants of these capacity limits is the amount of channel state information at the transmitters (CSIT). In this dissertation, we consider the MIMO interference networks with both perfect CSIT and CSIT uncertainty. In particular, a novel class of replication-based outer bounds will first be presented for arbitrary rank-constrained MIMO interference networks with perfect CSIT. It creates a new perspective of the capacity problem, so that even simple arguments such as user cooperation become quite powerful when applied in the replicated network, giving rise to stronger outer bounds, than when applied directly in the original network. Then, we prove that when CSIT is not perfect, signal space partitioning schemes can be DoF optimal. An interesting idea that emerges from this study is ``elevated multiplexing'' where the signals are split into streams and transmitted from separate antennas at elevated power levels, which allows these signals to be jointly decoded at one receiver which has fewer spatial dimensions with lower interference floors, while another receiver is simultaneously able to separately decode these signals with a higher interference floor but across a greater number of spatial dimensions. Finally, we explore the compatibility of various approaches under CSIT uncertainty which only been studied in isolation before, such as Blind Interference Alignment (BIA) and partial zero-forcing. Coding schemes are proposed that jointly exploit partial channel knowledge and reconfigurable antennas, demonstrating synergistic DoF gains over what is achievable with either BIA or beamforming by itself.