Michael Raatz, Svenja Schälicke, M. Sieber, Alexander Wacker, Ursula Gaedke

• Chemostat experiments are employed to study predator-prey and other trophic interactions, frequently using phytoplankton-zooplankton systems. These experiments often use population dynamics as fingerprints of ecological and evolutionary processes, assuming that the contributions of all major actors to these dynamics are known. However, bacteria are often neglected although they are frequently present. We argue that even without external carbon input bacteria may affect the experimental outcomes depending on experimental conditions and the physiological traits of bacteria, phytoplankton, and zooplankton. Using a static carbon flux model and a dynamic simulation model, we predict the minimum and maximum impact of bacteria on phytoplankton-zooplankton population dynamics. Under bacteria-suppressing conditions, we find that the effect of bacteria is indeed negligible and their omission justified. Under bacteria-favoring conditions, however, bacteria may strongly affect average biomasses of phytoplankton and zooplankton. The populationChemostat experiments are employed to study predator-prey and other trophic interactions, frequently using phytoplankton-zooplankton systems. These experiments often use population dynamics as fingerprints of ecological and evolutionary processes, assuming that the contributions of all major actors to these dynamics are known. However, bacteria are often neglected although they are frequently present. We argue that even without external carbon input bacteria may affect the experimental outcomes depending on experimental conditions and the physiological traits of bacteria, phytoplankton, and zooplankton. Using a static carbon flux model and a dynamic simulation model, we predict the minimum and maximum impact of bacteria on phytoplankton-zooplankton population dynamics. Under bacteria-suppressing conditions, we find that the effect of bacteria is indeed negligible and their omission justified. Under bacteria-favoring conditions, however, bacteria may strongly affect average biomasses of phytoplankton and zooplankton. The population dynamics may become highly complex, which may result in wrong interpretations when inferring processes (e.g., trait changes) from population dynamic patterns without considering bacteria. We provide suggestions to reduce the bacterial impact experimentally. Besides optimizing experimental conditions (e.g., the dilution rate) the appropriate choice of the zooplankton predator is decisive. Counterintuitively, bacteria have a larger impact if the predator is not bacterivorous as high bacterial biomasses and complex population dynamics arise via competition for nutrients with the phytoplankton. Only at least partial bacterivory minimizes the impact of bacteria. Our results help to improve the design of chemostat experiments and their interpretation, and advance the study of ecological and evolutionary processes in aquatic food webs.…