This paper describes an approach to save development costs and time by frontloading of engineering and starting calibration tasks already in early component conception phases. This requires a consistent and parallel virtual development and calibration methodology. The interaction between vehicle subcomponents physically available and those only virtually available at that time is considered by highly accurate real-time models on closed-loop co-simulation platforms (HiL-simulator) providing the appropriate response to components in the hardware test. This paper shows results of a heterogeneous testing scenario containing a real internal combustion engine on a test facility and a purely virtual vehicle using two different automatic transmission calibration and hardware setups. The first constellation is based on a validated vehicle model (A), including a physical dual clutch transmission, a semi-physical tire and a vehicle dynamics model. With this standard configuration the real-time model accuracy is initially illustrated by comparing the operating points distribution and the tailpipe emissions (diluted vs. undiluted) in WLTC tests for the closed-loop setup at the engine test bench with the real vehicle on a chassis dynamometer. Furthermore the achievable reproducibility with this in-the-Loop approach regarding gaseous and particulate emissions is shown. Finally the sensitivity and high reproducibility of tailpipe emissions related to changes in the calibration set of the virtual TCU are pointed out for this configuration in WLTC and RDE emission tests. In a second step, another vehicle model (B) is set up and also validated using extensive vehicle measurements. In contrast to model A, model B is equipped with an eight speed automatic transmission model based on physical relations and an all-wheel drive powertrain model. During the validation process of model B several drivability and emission tests are performed in a Model-in-the-Loop simulation environment. Afterwards the validated transmission and TCU models are installed into the vehicle model A virtually, resulting in vehicle variant C as an outcome. This physically nonexistent vehicle variant is tested at the Engine-in-the-Loop test facility. The conceptually different results at the test bench are compared and discussed regarding the vehicle A setup. The potential and reproducibility of the Engine-in-the-Loop approach are shown by a compilation of the results for the variants B and C.