Hello Gergely, I have seen 2 articles where the authors have reported endothermic reactions from the ITC profiles. The references are: 

Mertins & Dimova, Langmuir 2011, 27, 5506-5515 & Huang et al. ACS Nano 2014, 8, 5402-5412. 

Hope you find them useful. 

There are so many publications where endothermic interactions have been reported that I would not dare to say what publications should be highlighted. A quick survey through the literature will result will be sufficient.

Regarding the assessment of the endothermic/exothermic character of ligand interactions, it is important to distinguish whether or not the buffer-independent binding enthalpy is reported. If a given protein-ligand interaction is coupled to proton exchange processes, the binding parameters will be pH-dependent. In addition, the binding enthalpy will be dependent on the ionization enthalpy of the buffer employed. In this case, it is important to perform a titration using a low-ionization enthalpy buffer or using different buffers to remove that contribution, following a well-known procedure. The observed enthalpy (using a given buffer) and the buffer-independent enthalpy may be quite different.

Other factors may also contribute to the observed binding enthalpy. For example, in many cases it is necessary to add some glycerol to the sample solutions in order to stabilize a given protein (against aggregation, precipitation...). In that case the binding enthalpy might be dependent on the glycerol concentration.

enthalpy of a process = heat released due to new bonds formation  -  heat required to break the old bonds.

The old bonds comprise of solvation spheres. if the formation energy is more than required for breaking the solvation sphere, the process will be exothermic, the other way would be endothermic. The strength of solvation depends on the ionic potential of metal ion and the polarity of the solvent. A hard acid with a soft base (Ex:  Lanthanides with sulphur or phosphorous ligands) can be endothermic in aqueous phase process.