In this paper, we consider a multiuser multiple-input multipleoutput (MU-MIMO) heterogeneous network (HetNet) in which a macro base station (MBS) and multiple small-cells (SCs) coexist to serve multiple user terminals. We propose a joint linear precoder design problem to maximize the energy efficiency of the considered network. To tackle the cross-tier interference in the HetNets, we exploit zero-forcing precoding where the interference at the users is cancelled out by block diagonalization scheme. A novel group sparsity promoted as group Lasso is proposed using the weighted 1 norm minimization, where the group sparsity pattern indicates those SCs that can be switched off and non-associated users. Simulation results show that the proposed algorithm outperforms many existing algorithms in terms of the total energy efficiency in the HetNets.