The responses of forests to nitrogen (N) deposition largely depend on the fates of deposited N within the ecosystem. Nitrogen-fixing legume trees widely occur in terrestrial forests, but the fates of deposited N in legume-dominated forests remain unclear, which limit a global evaluation of N deposition impacts and feedbacks on carbon sequestration. Here, we performed the first ecosystem-scale ¹⁵N labeling experiment in a typical legume-dominated forest as well as in a nearby non-legume forest to determine the fates of N deposition between two different forest types and to explore their underlying mechanisms. The ¹⁵N was sprayed bimonthly for 1 year to the forest floor in control and N addition (50 kg N ha⁻¹ year⁻¹ for 10 years) plots in both forests. We unexpectedly found a strong capacity of the legume forest to retain deposited N, with 75 ± 5% labeled N recovered in plants and soils, which was higher than that in the non-legume forest (56 ± 4%). The higher ¹⁵N recovery in legume forest was mainly driven by uptake by the legume trees, in which ¹⁵N recovery was approximately 15% more than that in the nearby non-legume trees. This indicates higher N-demand by the legume than non-legume trees. Mineral soil was the major sink for deposited N, with 39 ± 4% and 34 ± 3% labeled N retained in the legume and non-legume forests, respectively. Moreover, N addition did not significantly change the ¹⁵N recovery patterns of both forests. Overall, these findings indicate that legume-dominated forests act as a strong sink for deposited N regardless of high soil N availability under long-term atmospheric N deposition, which suggest a necessity to incorporate legume-dominated forests into N-cycling models of Earth systems to improve the understanding and prediction of terrestrial N budgets and the global N deposition effects.