The pathogen virulence is traditionally thought to co-evolve as a result of reciprocal selection with its host organism. In natural communities, pathogens and hosts are typically embedded within a web of interactions with other species, which could affect indirectly the pathogen virulence and host immunity through trade-offs. Here we show that selection by predation can affect both pathogen virulence and host immune defence. Exposing opportunistic bacterial pathogen Serratia marcescens to predation by protozoan Tetrahymena thermophila decreased its virulence when measured as host moth Parasemia plantaginis survival. This was probably because the bacterial anti-predatory traits were traded off with bacterial virulence factors, such as motility or resource use efficiency. However, the host survival depended also on its allocation to warning signal that is used against avian predation. When infected with most virulent ancestral bacterial strain, host larvae with a small warning signal survived better than those with an effective large signal. This suggests that larval immune defence could be traded off with effective defence against bird predators. However, the signal size had no effect on larval survival when less virulent control or evolved strains were used for infection suggesting that anti-predatory defence against avian predators, might be less constrained when the invading pathogen is rather low in virulence. Our results demonstrate that predation can be important indirect driver of the evolution of both pathogen virulence and host immunity in communities with multiple species interactions. Thus, the pathogen virulence should be viewed as a result of both past evolutionary history, and current ecological interactions.