Shoot defoliation by grazers or mowing can affect root traits of grassland species, which may subsequently affect its aboveground traits and ecosystem functioning (e.g., aboveground primary production). However, experimental evidence for such reciprocal feedback between shoots and roots is limited. We grew the perennial grass Leymus chinensis–common across the eastern Eurasian steppe–as model species in a controlled-hydroponics experiment, and then removed half of its shoots, half of its roots, or a combination of both. We measured a range of plant aboveground and belowground traits (e.g., phenotypic characteristics, photosynthetic traits, root architecture) in response to the shoot and/or root removal treatments. We found the regenerated biomass was less than the lost biomass under both shoot defoliation and root severance, generating a under-compensatory growth. Root biomass was reduced by 60.11% in the defoliation treatment, while root severance indirectly reduced shoot biomass by 40.49%, indicating a feedback loop between shoot and root growth. This defoliation-induced shoot–root feedback was mediated by the disproportionate response and allometry of plant traits. Further, the effect of shoot defoliation and root severance on trait plasticity of L. chinensis was sub-additive. That is, the combined effects of the two treatments were less than the sum of their independent effects, resulting in a buffering effect on the existing negative influences on plant persistence by increased photosynthesis. Our results highlight the key role of trait plasticity in driving shoot–root reciprocal feedbacks and growth persistence in grassland plants, especially perennial species. This knowledge adds to earlier findings of legacy effects and can be used to determine the resilience of grasslands.