Aroma compounds can interact with food matrix components e.g. proteins. This can be related to a chemical reaction between aroma compounds and the macromolecule or with not covalent interaction between aroma compounds and proteins. The free binding energy of a macromolecule/odorant-complex is influenced by steric, entropic and thermodynamic contributions. The binding strength of a ligand-protein complex is characterized by the experimentally measurable association constant (KA) or by the dissociation constant (KD).
The aim of the present research work was the determination of the binding constants of selected flavour compounds (γ-and δ-Lactones and aliphatic esters) to important food proteins (β-lactoglobulin, BLG and bovine serum albumin, BSA).
In this work, ultracentrifugation and micro dialysis technique were used for the determination of the binding constants. A headspace method was developed for the measurement of partition coefficient of odorant/protein solutions.
Using dialysis technique, the highest association constant was found for γ-undecalacton (BSA: KA= 1.6 x 10⁴ M-1; BLG: KA= 5.9 x 10³ M-1).
In the series of γ- and δ-lactones, respectively, the association constants decreased with decreasing molecular weights of the lactones. The determination of the air/protein-solution partition coefficient of odorants revealed a reduction of odorant in the headspace in the presence of a protein (e.g. the δ-decalactone concentration in the headspace decreased by a factor of about two in the presence of BSA in comparison to the water solution without protein).
To determine the influence of the lipophilicity of odorants on the binding behaviour, the Log P values octanol/water and cyclohexan/water were examined. The highest LogPOct/H2O values were found for γ-undecalacton (LogPOct/H2O= 3.30) and δ-undecalacton (LogPOct/H2O= 2.93). These both odorants also showed the highest binding affinities to the proteins.
The application of molecular modelling (AUTODOCK, GRAMM) led to the identification of a specific "lactone-binding position" to BLG, not described up to now in the literature. These findings were supported by competitive binding experiments with ligands with known binding positions to BLG. The experimental results from binding energies were compared with the data calculated by molecular modelling (van der Waals and electrostatic interactions of the protein-ligand complexes; intramolecular energy, rotational degrees of freedom and Log POct/H2O values of the ligands). A statistical investigation of experimental and calculated bindings energies by means of Partial Least Square (PLS) analysis showed a correlation coefficient for the calibration of 0.992 and for the cross validation of 0.980.
The molecular modelling method developed in this work is a suitable procedure for the prediction of binding constants of unknown compounds to macromolecules.