Grassland ecosystems worldwide not only provide many important ecosystem services but they also function as a major source of the greenhouse gas nitrous oxide (N2O), especially in response to nitrogen deposition by grazing animals. To explore the role of plants as mediators of these emissions, we tested whether and how N2O emissions are dependent on grass species richness and/or specific grass species composition in the absence and presence of urine deposition. We hypothesized that: (i) N2O emissions relate negatively to plant productivity; (ii) four-species mixtures have lower emissions than monocultures (as they are expected to be more productive); (iii) emissions are lowest in combinations of species with diverging root morphology and high root biomass; and (iv) the identity of the key species that reduce N2O emissions is dependent on urine deposition. We established monocultures and two- and four-species mixtures of common grass species with diverging functional traits: Lolium perenne L. (Lp), Festuca arundinacea Schreb. (Fa), Phleum pratense L. (Php) and Poa trivialis L. (Pt), and quantified N2O emissions for 42 days. We found no relation between plant species richness and N2O emissions. However, N2O emissions were significantly reduced in specific plant species combinations. In the absence of urine, plant communities of Fa+Php acted as a sink for N2O, whereas the monocultures of these species constituted a N2O source. With urine application Lp+Pt plant communities reduced (P < 0.001) N2O emissions by 44% compared to monocultures of Lp. Reductions in N2O emissions by species mixtures could be explained by total biomass productivity and by complementarity in root morphology. This study shows that plant species composition is a key component underlying N2O emissions from grassland ecosystems. Selection of specific grass species combinations in the context of the expected nitrogen deposition regimes may therefore provide a key for mitigation of N2O emissions.