The 3GPP has proposed enhanced intercell interference coordination (eICIC) by making the macrocell silent in the almost blank subframes (ABSs) to mitigate interference for low-power base stations (BSs) in the downlink direction. However, two important challenges are introduced: 1) the underutilization for macrocell in ABS and 2) the optimal downlink association may not be optimal for uplink transmission, i.e., downlink/uplink decoupling (DUDe). In this paper, we first propose a novel method based on configuring uplink transmission in the macrocell in ABSs (UM-ABS) for eICIC to improve the efficiency of the ABS. Then, the UM-ABS and DUDe are formulated as an optimization problem of investigating the joint load balancing of downlink and uplink for eICIC in heterogeneous networks (HetNets) while considering different service classes. The problem is actually a mixed binary integer programming problem, which is difficult to solve. We provide a relaxed-rounding solution to the model. Numerical results illustrate that the proposed algorithm achieves superior performance in comparison to state-of-the-art methods in terms of system throughput and the rate gain for users.