No, because the latter theories has parametrically definable automorphisms that swap two equivalent classes, but the former theory is definably rigid. If the theories were bi-interpretable, then we could fix a copy of the model with two equivalent classes (having the same elements), and inside this model we could define a copy of MK, and inside that define a copy of the original model, but because the automorphism would have to fix the definable interpretation but swap the classes, it would ruin the bi-interpretability feature.