Fig 1.
The publicly available yeast protein structures were combined with PTM sites data in order to assess the effects of PTMs on protein binding. In addition, conservation analysis aided in identifying the subset of predictions that are transferable to protein-protein interactions in other organisms.
Fig 2.
Comparison of local binding contributions between conditions and PTM types for the interface-located PTMs.
ΔΔGbind,contribution values in two conditions for each type of PTM (acetylation in blue, phosphorylation in orange), where NC stands for “normal conditions” and SC for “stress conditions”. Acetylation more frequently contributes to binding in a stabilizing way, as reflected in the majority of the violin plot being below zero, while the opposite is true for phosphorylation. No significant differences were found between ΔΔGbind,contribution for PTMs of the same type between normal and stress conditions (p-values indicated on the plots). Comparison of ΔΔGbind,contribution values between different types of PTMs within a given condition shows several statistically significant differences, connected by the lines (full line for NC and dashed for SC) and marked with stars above the plots. The corresponding p-values are: p(Lys-Ser, NC) = 1.3e-76, p(Lys-Thr, NC) = 7.7e-39, p(Lys-Tyr, NC) = 2.3e-03, p(Ser-Thr, NC) = 2.0e-02, and p(Ser-Thr, SC) = 2.9e-02. Cohen’s d-values indicate that the effect sizes vary from small (d(Ser-Thr, SC) = 0.29, d(Ser-Thr, NC) = 0.45) to large (d(Lys-Ser, NC) = 1.43, d(Lys-Thr, NC) = 1.82, d(Lys-Tyr, NC) = 2.23).
Fig 3.
Correlation between the subunit ΔΔGbind values and the sum of ΔΔGbind,contribution of PTMs located in the respective protein chains.
Points above the grey line (y = x, with the surrounding grey area describing the error in subunit ΔΔGbind values) belong to systems in which the overall effect of PTMs on subunits’ binding is more destabilizing than could be expected based solely on local contributions of the PTMs, and the other way around for points below the grey line, both thanks to long-range effects. For systems described by points lying within the grey area, the local effect of PTMs does explain the overall effect on binding.
Fig 4.
Overall effect of PTMs on the protein-protein binding.
Chain ΔΔGbind values are significantly more stabilizing for PTMs found in normal when compared to those identified in stress conditions, though the effect size is small (Cohen’s d = 0.32).
Fig 5.
PTM sites in the dataset used in this work appear to be widely conserved, with Eukaryota being the most common lowest common ancestor. The plot shows data for all unique PTM sites analyzed in this work, both acetylation and phosphorylation and independent of the conditions (normal or stress), where unique means that the redundancy due to PTMs appearing in multiple chains, complexes, or in both conditions was removed.
Fig 6.
A. The initial structure of the Kapα homodimer (PDB ID: 1BK5), with chain A shown in magenta and chain B in cyan. Green space-filling representation is used to denote the backbone atoms of lysine acetylation sites, while red is used for serine phosphorylation sites detected in normal conditions. B. Per-residue decomposition of the Kapα free energy of binding (ΔΔGbind,contribution). Amino acids of homodimer are shown on the x-axis, where the two chains are separated by a dashed cyan line. For each residue, the difference of binding contribution between modified and non-modified complex is shown as a vertical line ending with a dot, where negative values denote residues with a more stabilizing contribution in the modified than in the non-modified complex, and vice versa for the positive. Residues with contributions larger than 5 kcal/mol are labeled. C. Glu460 in chain A is interface located only in the modified complex, and therefore has a large stabilizing ΔΔGbind,contribution, mainly due to interactions with Arg120 from chain B. D. Glu506 has a significantly less destabilizing contribution to binding in the modified complex, where chain B is more distant. The inserted frames depict Glu506 and residues of chain B which are within 5 Å from it in a space-filling representation.
Table 1.
Overview of the PTMs count in the Swiss-Prot entries for human and model organisms.
[18].