Mesocosm experiment
The mesocosm experiment was conducted to simulate three-level food chains (algae-zooplankton-shrimp) at the Taihu Laboratory for Lake Ecosystem Research, located at the eastern side of Taihu Lake (31︒03’N, 120︒42’E). Lake seston was used as basal resources to allow a regional assortment of phytoplankton species representative of local water bodies. We implemented a factorial design by crossing three levels of nutrient supply (HP, total phosphorus at 0.08 mg L-1; MP, total phosphorus at 0.058 mg L-1; LP, total phosphorus at 0.028 mg L-1, Fig. S1) with three types of zooplankton communities (Daphnia magna + Daphnia brachyurum ,Sinocalanus dorrii + D. brachyurum , and D. magna +S. dorrii + D. brachyurum ). Nutrient manipulation was achieved by adding K2HPO4 solutions (0.5 mol L-1) to 150 L tanks that contained mixtures of lake water and aerated tap water (at the proportion of 1:1). Tanks receiving no P additions represents the LP treatments with total phosphorus concentration at 0.028 mg L-1 (Fig. S1c). In addition, KNO3 solutions were added to all tanks to compensate for the dilution effects of tap water (Fig. S1d-f). Zooplankton communities were manipulated by manually adding pre-cultured zooplankton species (see details in SI Appendix 1 ) to each tank at a total density of around 30 ind. L-1. Among these species, D. magna possesses larger body size, slower locomotion (thus lower anti-predation ability), and higher nutrient concentrations than the other two species (Sterner et al. 1992). We therefore expected D. magna to be higher-quality food for predators. The inclusion of D. brachyurum in all treatments was to prevent the extinction of zooplankton communities due to predator exploitation. Before filling the tanks, animals originally contained in lake water were eliminated via repeatedly screening through a 15 µm mesh. For each of the nine treatments (3 nutrient levels×3 zooplankton communities), we created six mesocosms, half with and the other half without predators (i.e., two individuals of juvenile shrimp Exopalaemon Modestus ) in order to quantify the strength of trophic cascades (He et al.2021). Therefore, a total of 54 experimental units (i.e., 3 nutrient levels × 3 zooplankton community compositions × 2 predator treatments × 3 replicates) were created. All the mesocosms were fixed in a pond to ensure a relatively constant ambient temperature and checked three times a day to remove unexpected allochthonous materials. The experiment lasted for 6 weeks from 14 July to 25 August, 2021.