This paper presents an input-mapping based distributed MPC approach for linear systems, where each subsystem is coupled with the inputs of other subsystems, and the coupling matrices are unknown but bounded. By mapping the reference input transmitted by neighbours to the past inputs, the unknown input coupling term can be represented as a linear combination of past online measured input and state trajectories of the subsystem itself. The input-mapping scheme is combined with distributed MPC approach to stabilize the linear system with unknown input coupling. The recursive feasibility of the optimization problems and the asymptotic stability of the closed loop system are established. Finally, the simulation results show that the proposed strategy works well.