Prediction of protein-protein interaction sites between Russell's viper PLA2 and the γ-phospholipase inhibitor PIP from the amino acid frequency distribution of a bio-panned peptide set.

Abstract

We screened a random peptide phage display library using Russell's viper venom phospholipase A2 (RV-PLA2) as bait. Sequence information from the resulting set of bio-panned heptapeptides was analyzed and mined to determine likely sites of interaction between two subunits of RV-PLA2 homo dimers and between RV-PLA2 and the γPLA2 inhibitor PIP from Malayopython reticulatus. This was accomplished in part by sequence alignment of the affinity-selected peptides with the sequences of RV-PLA2 and PIP. Because similarity scores calculated from sequence alignments proved inadequate to determine interaction interfaces accurately for RV-PLA2 dimers, we explored the use of amino acid frequency-based interactions scores (SFI/SFIN) for a more accurate prediction of protein-protein interaction sites. Heptamers with elevated SFI(N) scores were compared to interfaces of interaction observed in crystal structures of RV-PLA2 homodimers and to sites of interaction predicted by protein-protein docking between structures of RV-PLA2 and model of PIP. Segments with a high density of protein-protein contacts coincided with heptamer sequences exhibiting SFI and/or SFIN scores significantly above average, in both RV-PLA2 homodimers and in RV-PLA2 γPLI heteromeric structures. Elevated SFI and SFIN scores were associated with peptide function since the heptamers with some of the highest SFI and SFI(N) scores, LPGLPLS, GLPLSLQ and SLQNGLY constitute the known PLA2 inhibitor P-PB.I (LPGLPLSLQNGLY) while KLGRVDI, and WDGVYIR, constitute PIP-17 (LGRVDIHVWDGVYIRGR), IC50 for hsPLA2: 5.3 μM. A graph showing the alignment of maxima between SFI scores and average solvent accessibility (per heptamer) suggests that solvent accessibility is a major driver of both protein-protein interaction and phage selection. Insights We show by computational methods that in sets of small phage-displayed peptides of the same length selected for binding to the same target protein, amino acids contributing to binding at a particular position occur at higher frequencies than in random peptides. This position-specific selection of particular amino acids can be detected in the position-specific amino acid frequency distribution of that set of selected peptides. Therefore, when this position-specific amino acid frequency is mapped back onto a particular amino acid sequence of the same length, the sum of these frequencies can function as a measure of enrichment of selected amino acids.