Thermal stability of a protein predicted from its three dimensional structure

Abstract

A native protein in aqueous salt solution usually unfolds reversibly upon heating at a transition temperature (T_m) with unfolding enthalpy (ΔH_m). The method to predict T_m and ΔH_m from the 3D structure assuming simple two-states transition has been developed using experimental values of T_m, ΔH_m and heat capacity change (ΔC_{p, u}) compiled in the database“ProTherm”. In order to account for dependence of thermostability on pH and salt concentration, the difference in electrostatic interactions between native and denatured states (Δh_{u, ele}) was computed applying Debye-Huckel type potential between partial charges located on ionizable side chain atoms. Standard unfolding free energy (ΔG⁰_u), enthalpy (ΔH⁰_u), and ΔC_{p, u} at 25°C after subtracting electrostatic energies were obtained using a data set of 63 experimental data on 32 uncharged proteins. Each of these quantities was assumed to be a sum of products of the difference in accessible surface area (ΔASA) of an i-th constituent atomic group between the native and unfolded extended structure times a corresponding adequate constant. Proportional constants for six (aliphatic, aromatic, hydroxyl, amide, carbonyl, and thiol) atomic groups were determined using the data set. Thus, ΔG⁰_u, ΔH⁰_u and ΔC_{p, u} for a protein at uncharged form of known 3D structure can be computed using its ΔASA value of every group, and ΔG_u(T) at T is determined according to thermodynamic relation, adding Δh_{u, ele}(T). Values of T_m and ΔH_m for the protein may, therefore, be evaluated as values at ΔG_u(T)=0. The agreement of calculated and experimental values for the data set used for the determination of the constants was practically the same as that for another data set of 33 experimental data on T_m and ΔH_m for 19 proteins with standard deviations of about 12°C and 38kcal/mol, respectively. The stability and instability of correctly and misfolded proteins (hemerythrin, and Ig-variable domain fragment) were successfully explained in terms of ΔG_u.