SKEMPI 2.0: An updated benchmark of changes in protein-protein binding energy, kinetics and thermodynamics upon mutation

Justina Jankauskaite; Brian Jiménez-García; Justas Dapkunas; Juan Fernández-Recio; Iain H. Moal

Within SKEMPI some entries can be combined to construct mutant cycles, which quantify the interactions between residues, the dependence of these interactions on other residues, and other higher order effects. The most common instances are double mutant cycles, where affinities are available for the wild type, A, B and AB mutations, of which there are 610 examples. Of these, 53 involve at least one mutant for which binding was not observed, or only an inequality is available, and 235 involve mutations reported in the same reference, and thus the affinities are likely to have been measured using the same technique and conditions. A further 218 double mutant cycles can be constructed in the background of a third mutation (i.e. C, AC, BC and ABC mutations are available), of which 209 are not composed of non-binding mutations or mutations with inequalities, and 131 involve affinities coming from the same reference. Of the 766 double mutant cycles containing neither inequalities nor non-binding mutants, a number of parameters can be calculated, including ∆∆G ab→Ab , ∆∆G ab→aB and ∆∆G ab→AB , the binding free energy change of both single and the double mutation respectively, as well as ∆∆G aB→AB and ∆∆G Ab→AB , the energy of a single mutation within the context of the other mutation, and ∆∆G int = ∆∆G aB→AB -∆∆G ab→Ab = ∆∆G Ab→AB − ∆∆G ab→aB , the interaction energy of the two mutations [26] . From these, it can be deduced that 345 are additive (∆∆G int < 0.5kcal.mol −1 ). Of the non-additive cycles, 293 exhibit tighter binding in the double mutant than the sum of the single mutants (positive epistasis), of which 6 result in even tighter binding than individual effects of two single mutations that strengthen the interaction (synergistic positive), while 273 correspond to double mutants which reduce binding by less than the sum of two single mutants which reduce binding (antagonistic positive). Similarly, 128 cycles have double mutants exhibiting weaker binding than the sum of the two single mutants (negative epistasis), of which 58 contain two destabilising single mutations (synergistic negative) and 26 contain two stabilising single mutations (antagonistic negative). The range of ∆∆G int values rarely fall outside of the -5 to 3 kcal.mol −1 range. Of the 421 nonadditive cycles, 151 show noticeable sign epistasis, in which the sign of the effect of either the A or B mutation flips depending on the presence or absence of the background mutation (i.e., for the A mutation, |∆∆G ab→Ab | > 0.2 kcal.mol −1 and |∆∆G aB→AB | > 0.2 kcal.mol −1 and |∆∆G ab→Ab − ∆∆G aB→AB | > 0.4 kcal.mol −1 ). Of these, 38 correspond to mutations which destabilise the complex in the presence of the background mutation, but stabilise in its absence (destabilising sign epistasis), while 113 correspond to mutations which stabilise the complex in the mutant background but otherwise destabilise the complex (stabilising sign epistasis). Only 8 cycles exhibit the more extreme reciprocal sign epistasis, which in 6 cases are where both single mutations are stabilising (< −0.2 kcal.mol −1 ), but the double mutant is destabilising (> 0.2 kcal.mol −1 ), and the remaining two correspond to two destabilising mutations (> 0.2 kcal.mol −1 ) for which the double mutation is stabilising (< −0.2 kcal.mol −1 ). The types of substitutions that can give rise to extreme effects such as stabilising reciprocal sign epistasis can be illustrated with the Mlc-IIB Glc interaction in E. coli [52] ,