Animals in nature are capable of locomotion at high speed and with high efficiency by adapting their gait based on proprioceptive information from the body, even in dynamic environments. The muscle-tendon complex in the limb, an essential component of animal body movement, is used for adaptive actuation based on the sensory tension of the complex caused by the extension and contraction depending on the load applied to the foot. In this paper, we implemented a CPG-based gait control based on the passive behavior of a leg mechanism consisting of elastic elements inspired by a muscle-tendon complex of mammals, which we previously proposed, on a quadruped robot and verified its effectiveness through experiments. By determining appropriate mechanical and control parameters, we experimentally demonstrated that the proposed control improved running speed and autonomously transitioned the gait pattern.