These are in every paper about anti-windup or systems with actuator saturation. I haven't find any prove for that. In every paper they say they are equivalent.
K = [K1; K2] is the Controller
G = [G1; G2] is the plant
M = is the chosen transfer function matrix as anti windup
This image is taken from the paper:
"Accounting for uncertainty in anti-windup synthesis" by Matthew C. Turner, Guido Herrmann and Ian Postlethwaite