We report on the measurement of the physical properties (rest-frame K-band luminosity and total stellar mass) of the hosts of 89 broad-line (type-1) active galactic nuclei (AGNs) detected in the zCOSMOS survey in the redshift range 1 < z < 2.2. The unprecedented multi-wavelength coverage of the survey field allows us to disentangle the emission of the host galaxy from that of the nuclear black hole in their spectral energy distributions (SEDs). We derive an estimate of black hole masses through the analysis of the broad Mg II emission lines observed in the medium-resolution spectra taken with VIMOS/VLT as part of the zCOSMOS project. We found that, as compared to the local value, the average black hole to host-galaxy mass ratio appears to evolve positively with redshift, with a best-fit evolution of the form , where the large asymmetric systematic errors stem from the uncertainties in the choice of initial mass function, in the calibration of the virial relation used to estimate BH masses and in the mean QSO SED adopted. On the other hand, if we consider the observed rest-frame K-band luminosity, objects tend to be brighter, for a given black hole mass, than those on the local M BH-MK relation. This fact, together with more indirect evidence from the SED fitting itself, suggests that the AGN hosts are likely actively star-forming galaxies. A thorough analysis of observational biases induced by intrinsic scatter in the scaling relations reinforces the conclusion that an evolution of the M BH-M * relation must ensue for actively growing black holes at early times: either its overall normalization, or its intrinsic scatter (or both) appear to increase with redshift. This can be interpreted as signature of either a more rapid growth of supermassive black holes at high redshift, a change of structural properties of AGN hosts at earlier times, or a significant mismatch between the typical growth times of nuclear black holes and host galaxies. In any case, our results provide important clues on the nature of the early co-evolution of black holes and galaxies and challenging tests for models of AGN feedback and self-regulated growth of structures.