Existing simulation-based studies on applying model predictive control (MPC) schemes for space heating operation to enable demand response (DR) make use of linear models for the heating system, usually by assuming convective electrical baseboard heaters. However, buildings connected to district heating networks are typically equipped with hydronic heat emitters, such as radiators, that behave nonlinear. This paper therefore investigates the effect of including the nonlinear dynamics of a hydronic heat emitter on the DR potential of MPC for space heating. Furthermore, the performance of a practical two-level control approach suitable for real application, in which a heating setpoint was determined by a linear MPC and communicated to a conventional proportional integral controller, was investigated. The simulation framework for the investigation was based on the application of an experimentally obtained hydronic radiator model applied in different co-simulation setups, featuring a model of a poorly and a highly insulated apartment, respectively. The results indicated that inclusion of the nonlinear thermal effects of hydronic radiators did not significantly affect the DR performance when compared to the results of an MPC scheme controlling convective electrical baseboard heaters. In general, both setup achieved operational cost savings of approx. 5% and 18% in an existing and retrofitted building, respectively, while restricting the amount of thermal comfort violations to a limited extent. This suggests that results obtained in previous studies featuring electrical baseboard heaters also apply to buildings equipped with hydronic heating systems, and that future simulation-based studies and practical implementation of MPC for space heating can continue to rely on the use of far less computationally demanding linear control-models. Furthermore, the results suggest that the two-level control scheme seems like an appropriate control setup suitable for real applications. (C) 2018 Elsevier B.V. All rights reserved.