Résumé : Auto/paracrine cell-to-cell communications via diffusive messengers can be coupled to a positive feedback loop in which cell stimulation by a messenger results in the production of new messengers. This yields a potential mechanism for relay transmission of the emitted message. This paper investigates the influence of noise on this mutual coupling of the cells with their environment, using numerical simulations of a stochastic minimal model. The results demonstrate that the deterministic (mean-field) approximation of this stochastic process fails short of predicting its behaviour because of the presence of strong noise-induced fluctuations. Instead, the behaviour of the model can be explained by the occurrence of a nonequilibrium phase transition, which is found to be in the universality class of directed percolation. This provides a theoretical framework to understand signal transmission in these stochastic systems.