Data uploading time constitute a large portion of mobile data gathering time in wireless sensor networks. By equipping multiple antennas on the mobile collector, data uploading time can be greatly shortened. However, previous works only treated wireless link capacity as a constant and ignored power control on sensors, which would significantly deviate from the real wireless environments. To overcome this problem, in this paper we propose a new data gathering cost minimization framework for mobile data gathering in wireless sensor networks by considering dynamic wireless link capacity and power control jointly. Our new framework not only allows concurrent data uploading from sensors to the mobile collector, but also determines transmission power under elastic link capacities. We study the problem under constraints of flow conservation, energy consumption, elastic link capacity, transmission compatibility and sojourn time. We employ the sub gradient iteration algorithm to solve the minimization problem. We first relax the problem with Lagrangian duplication, then decompose the original problem into several sub problems, and present distributed algorithms to derive data rate, link flow and routing, power control and transmission compatibility. For the mobile collector, we also propose a sub algorithm to determine sojourn time at different stopping locations. Finally, we provide extensive simulation results to demonstrate the convergence and robustness of proposed algorithms. The results reveal 20% shorter data collection latency on average with lower energy consumptions compared to previous works as well as lower data gathering cost and robustness in case of node failures