In this paper, the impact of secondary MAC cooperation on the sum-throughput of multichannel cognitive radio networks is studied. The main goal is twofold: Given the primary and secondary users' duty cycle, (i) investigate the amount of spectrum sharing (i.e., number of secondary users) that maximizes the sum-throughput in the presence of secondary MAC cooperation; (ii) assess the performance gains attainable with cooperation. First, analysis is provided for the idealistic case of perfect sensing, with a simple model for secondary cooperation. Then, for the more realistic case of imperfect sensing, novel cooperative secondary strategies are proposed that are shown to provide relevant performance gains in terms of sum-throughput. Finally, numerical simulation results are provided to evaluate the performance of the cooperative schemes relative to other non-cooperative schemes.