Magnetic induction (MI) based communication is an important means for underwater data collection due to its less susceptibility to surrounding environments, predictable channel condition and high data rate. However, the communication rate is affected by the severe path loss in underwater channel and the orientation of the MI transceiver coils. In this case, allowing the coils to transmit collaboratively is an effective method to enhance the end-to-end communication performance. In this paper, we consider an autonomous underwater vehicle (AUV) collecting data from two seabed sensor nodes (SNs) using MI-based communication. In particular, we perform a comparative study to examine the performance of different cooperation schemes between the two nodes in MI-based underwater communication (MI-UWC) scenario. Specifically, we consider the three cooperation schemes: joint beamforming (JB), active relay (AR) and reactive relay (RR). We derive the communication rate of the two SNs under different cooperative schemes, and optimize the minimum throughput of the two nodes. We show through simulations that JB scheme and AR scheme are less sensitive to the orientation or the placement of the MI transceiver coils. In between, JB scheme significantly improves the throughput performance under different system setups.