DNA molecules carrying a Mu end(s) are inefficient targets in the Mu DNA strand-transfer reaction. This target immunity is due to preferential dissociation of Mu B protein from DNA molecules that have Mu A protein bound to the Mu end; free DNA is a much poorer target than DNA with Mu B protein bound. We show that Mu B protein, which binds nonspecifically to DNA, is immobile once bound. An encounter between Mu A and Mu B proteins, bound some distance apart along DNA, is necessary to facilitate the Mu B dissociation. Experiments which show that DNA without a Mu end can acquire immunity, by catenation to DNA with a Mu end(s), are consistent with a model of Mu A-Mu B interaction by DNA looping, but not by linear movement of protein(s) along DNA.