In this paper, we consider a spectrum-sharing cognitive radio network (CRN), in which the secondary user (SU) is an active-load assisted symbiotic radio (SR) system. Specifically, the SU transmitter (SU-Tx) exploits multiple antennas to simultaneously support the secondary transmission from the SU-Tx to the SU receiver (SU-Rx) and the backscatter transmission from the active-load assisted backscatter device (BD) to the SU-Rx. As a result, the PU receiver (PU-Rx) is interfered by not only the SU-Tx but also the BD. For such a scenario, the SR system needs to design the transmit beamforming at the SU-Tx and the amplification gain at the BD to balance two conflicting goals, namely, the rate maximization for SU and the interference control to PU-Rx. We formulate an SU rate maximization problem under its own transmit-power constraint, the interference-power constraint as well as some practical constraints introduced by the SR system. This non-convex problem is solved by an alternating optimization based method, which iteratively optimizes the beamforming vector and transmit power at the SU-Tx, and the amplification gain at the BD. Simulation results show that the proposed method outperforms the equal-gain allocation method.