In this paper, we propose an energy efficient iterative source localization scheme in wireless sensor networks (WSNs). Instead of sending data from all the sensors to the fusion center, a coarse location estimate is first obtained from a set of anchor sensors. Then, a few non-anchor sensors are activated at a time to refine the location estimate in an iterative manner. We assume that the channels between sensors and the fusion center are subject to fading and noise. The fusion center is assumed to either have the complete or partial channel knowledge. Based on the received information at each iteration, the minimum mean squared error (MMSE) estimate of the source location is approximated using a Monte Carlo method. Then, in order to activate the non-anchor sensors for the next iteration, we develop a mutual information (MI)-based sensor selection scheme. Simulation results for the partial channel knowledge (PCK) and the complete channel knowledge (CCK) are presented to show the performance of the proposed approach.