In this paper, we develop a novel joint transmit and receive design for multiuser MIMO (MU-MIMO) systems. The proposed scheme exploits the channel temporal correlation and incorporates the matched filter (MF) combining vector into the beamforming design. Considering the beamforming is optimized with the outdated channel state information (CSI) feedback, we design the beamformer by maximizing the conditional expectation of signal-to-leakage-and-noise ratio (SLNR) given that a delayed version of CSI feedback is available. In this way, the optimized beamforming is able to alleviate the performance degradation due to the CSI delay. Numerical results show that our proposed scheme achieves almost the same performance as conventional iterative algorithms but with much lower complexity.