Function-as-a-Service (FaaS, serverless) computing systems use an actor-like model that executes a function asynchronously, atomically and in an isolated context. However, a function must often also access state, e.g., memory or a database. This mixed model can break the actor guarantees, leading to bugs, crashes and data loss. To avoid this, we define Transactional-Turn Causal Consistency (TTCC). TTCC unifies the Turn of the actor model with the Transaction of the database model, under asynchronous, atomic and isolated execution, and guarantees mutual consistency of messages and memory. We define the model formally and present a reference implementation, along with preliminary experimental evaluation.