Bacterial quorum sensing (QS), where a secreted signal activates a cognate receptor, is thought to benefit bacteria by coordinating cooperative behaviors in a density-dependent manner. While a single QS system is sufficient to probe density, many bacteria contain complex multi-signal QS networks that regulate the same cooperative behavior. It is unclear what the benefits of a multi-signal system design are. Here we combine mathematical modeling and experiments to show that a multi-signal QS network can evolve from a simpler uni-signal network even without an apparent benefit under cooperative conditions, through facultative cheating. As a minority, a strain coding for the multi-signal network will exploit the cooperative behavior of a strain coding for a uni-signal network, but will resume cooperation as its frequency increases. Critically, evolution depends on the architecture of the complex network. Social advantage can arise only if the new QS signal forms an "AND" gate with the ancestral QS signal, in their regulation of cooperative behavior. We experimentally verify our predictions by studying the QS network regulating the B. subtilis ComA master regulator, which is controlled by a single comX-comP QS system and four paralogous rap-phr systems. We show that addition of each rap-phr paralog is selected for, while addition of comX-comP paralogs are selected against, due to the different circuit architecture these systems have. Together, our results demonstrate that social signaling complexity can evolve without intrinsic value, through a selfish mechanism and connect the architecture of social regulation to its evolutionary outcome.
