MU-MIMO transmission relies on exploiting multi-user diversity among a multitude of users. In this paper, to enhance MU-MIMO transmission, techniques for designing and exploiting reconfigurable antennas (RAs) are developed. A single-element RA is capable of generating modes with different radiation pattern and polarization states resulting in multiple different channel states for each user. This capability expands the search space for user-pairing optimization. To reduce complexity, an iterative user and antenna mode selection algorithm is proposed. Based on parasitic tuning, an RA optimized for MU-MIMO transmission and capable of creating four pattern types is designed using a genetic algorithm and full-wave electromagnetic analyses. Pattern- and channel-aware mode-set generation methods are developed, where mode groups are determined to reduce channel estimation overhead and mode selection complexity while achieving superior performance. Simulation results indicate that the proposed schemes need a lower number of users and/or relax user selection requirements to benefit from MU-MIMO transmissions. It is also seen that more legacy antennas are needed to achieve similar MU-MIMO performance provided by the RA system. Simulations show up to 16-dB signal-to-noise ratio gains for a single cell with 12 users, where the base station has four RAs and four single-antenna users are simultaneously served via MU-MIMO transmission.