In this paper, a cognitive radio network with multiple licensed users is considered. A dirty paper coding (DPC) based cooperation scheme for the Gaussian multiple-in-multiple-out (MIMO) cognitive radio network (CRN) with partial transmitter side information is studied. The problem of maximizing the sum-rate of MIMO CRN with multiple primary users over the transmitter covariance matrices, which is formulated as an optimization problem, is dealt with. Such an optimization, which needs to be performed jointly with the inflator factors under the DPC-based strategy, is solved by an iterative suboptimal solutions. Simulation results show that the proposed scheme is able to greatly improve the achievable rate performance in CRN.