A low-complexity multimode transmission technique for downlink multiuser multiple-input-multiple-output (MIMO) systems with block diagonalization (BD) is proposed. The proposed technique adaptively configures the number of data streams for each user by adjusting its number of active receive antenna and switching between single-stream beamforming and multistream spatial multiplexing, as a means to exploit the multimode switching diversity. We consider a highly loaded system where there are a large number of users, hence a subset of users need to be selected. Joint user and antenna selection has been proposed as a multiuser multimode switching technique, where the optimal subset of receive antennas and users are chosen to maximize the sum throughput. The brute-force search, however, is prohibitively complicated. In this paper, two low-complexity near-optimal user/antenna selection algorithms are developed. The first algorithm aims at maximizing a capacity lower bound, derived in terms of the sum Frobenius norm of the channel, while the second algorithm greedily maximizes the sum capacity. We analytically evaluate the complexity of the proposed algorithms and show that it is orders of magnitude lower than that of the exhaustive search. Simulation results demonstrate that the proposed algorithms achieve up to 98% of the sum throughput of the exhaustive search, for most system configurations, while the complexity is substantially reduced.