This paper proposes a dimensionality reduction of interference subspace to improve the interference suppression capability of null-space expansion (NSE) for multiuser massive multiple-input-multiple-output (MIMO) systems. The fundamental concept of the NSE is to nullify the interference subspace generated by the combined multiple past channel state information (CSI) vectors. It can suppress inter-user interference (IUI) leaked due to outdated CSI in a mobility environment. It is enabled by plentiful degrees of freedom (DoF) brought by massive antenna elements the base station has, and hence NSE can outperform existing channel prediction based precoding approaches. However, additive noise-induced CSI estimation error limits its effectiveness; expanded dimension of null-space cannot be appropriately spanned to interference subspace. The key idea of our new proposal is to extract principal interference subspace channel. Null-space can be appropriately expanded with reduced its dimension, and remaining DoF can be assigned to beamforming. It results in improving overall signal-to-interference-plus-noise power ratio (SINR) performance which is verified via numerical simulation.