In a K-user multiple-input multiple-output (MIMO) interference channel where each N-antenna transmitter communicates with its corresponding M-antenna receiver by using d degrees of freedom (DoF), the feasibility condition for conventional interference alignment (IA) implies that M+ N > d(K + 1). In this paper, we consider a K-user MIMO interference channel where averagely half of the total decoded data can be shared by receive nodes via partial coordination. We propose IA algorithms such that first, their feasibility condition can now be expressed as M+ N > d(K + 4)/2, which further implies less number of antennas is needed in comparison with conventional IA techniques to achieve the same number of DoF; and second, even with this reduced number of antennas, their throughput is still comparable to that of conventional IA methods. This is particularly pronounced for asymptotically large K where the required number of antennas for the proposed IA schemes can be half of that of conventional IA techniques to secure the same number of DoF. Moreover, we also examine the performance of the proposed IA algorithms in the presence of both error propagations and channel estimation errors. Simulation results show that in these cases, the throughput of the proposed algorithms can be comparable to that of conventional IA techniques.