We found that the vast majority of phosphocomplexes contain just a few phosphorylated sites whereas for some proteins up to half of their sites (Ser, Thr and Tyr sites) are potentially phosphorylated at some point, which is evident from the long tail of the probability distribution () for the fraction of phosphosites per protein. Several studies previously established that phosphosites may form clusters along specific regions of a protein sequence or on a protein surface (Schweiger and Linial, 2010
; Yachie et al., 2009
). Although the main reasons for these findings remain largely unknown, it was observed that in some cases the groups of sites can be phosphorylated simultaneously and cooperatively leading to certain advantages in terms of signal amplification and its strength modulation (Park et al., 2006
). In our study we showed that phosphosites have a tendency to be located on binding interfaces in protein complexes and this trend depends on the type of complex. This allows us to better understand the regulation of protein activity through phosphorylation within the framework of protein binding.
There are several reasons for such coupling between phosphorylation and binding. Phosphorylation may modulate the strength of interactions, bringing about changes in binding energy which may trigger the transitions between different conformer and oligomeric states. For the majority of proteins in our dataset the phosphorylation did not change the binding affinity significantly, which is consistent with several experimental studies pointing to the modest effect of phosphorylation on stability and protein conformation (Murray et al., 1998
; Serber and Ferrell, 2007
; Strickfaden et al., 2007
). At the same time in one third of our complexes the attachment of a phosphate group to interfacial Ser/Thr/Tyr sites which are expected to be phosphorylated caused a relatively large change in estimated binding energy. This in turn could lead to conformational changes or due to steric constraints preclude undesired interactions. Moreover, phosphorylation sites on interfaces significantly overlapped with the binding hot spots in heterooligomeric complexes and phosphorylation at binding hotspots could potentially disrupt the complex formation. In addition we showed that phospho Ser/Thr/Tyr on interfaces were more conserved than non-phosphorylated Ser/Thr/Tyr on interfaces. It should be mentioned that regulatory mechanisms of phosphorylation are quite diverse and in some cases phosphorylation might destabilize the complex and lead to protein activation or inactivation, while in others it may mediate complex formation and through competitive binding provide a negative control mechanism (as was shown for Smad example). In our study the phosphate group was attached to only one site at a time and since there can be several phosphosites per protein (on average there are about 2 phosphosites per protein in the set), we expect a greater effect if multiple sites are phosphorylated simultaneously.
Phosphorylation might not directly affect significantly complex stability, but rather provide diversity in recognition patterns and offer recognition sites for binding of certain domains and motifs (e.g. pTyr-binding by the SH2 domain, pSer/pThr binding by the SH2 and FHA domains) thereby modulating binding selectivity. Indeed, the reversibility of phosphorylation events allows decouple the binding selectivity and affinity thereby mediating selective binding even between proteins within transient and not very stable complexes. At the same time, phosphorylation of multiple sites on interfaces may amplify this signal and provide enhanced binding selectivity. Indeed such specific and reversible signaling at the residue level is a good indicator that a previous stage in cellular signaling networks has completed successfully. Many cellular control mechanisms operate at the level of protein-protein interactions and main signaling pathways involve dense networks of protein-protein interactions and phosphorylation events. Moreover, signaling pathways are quite often disrupted in cancer and it was recently shown that somatic cancer mutations are enriched with those that cause gain or loss of phosphorylation sites (Radivojac et al., 2008
). Similarly our study showed that the following signaling pathways “Hemostasis”, “Pathways in cancer”, “Cell Cycle, Mitotic”, and “Signaling in immune system” are enriched with phosphorylated heterooligomeric complexes.
Interestingly, we found that metabolic and hemostasis pathways are also enriched with phosphorylated homooligomeric complexes and phosphosites in weak transient homooligomers are considerably involved in binding. Previously we manually compiled a set of experiments which furnish evidence that phosphorylation at or near the homooligomer interface shifted the equilibrium between different oligomeric states with different protein activities (Hashimoto et al., 2011
). According to the classical model by Goldbeter and Koshland post-translational modifications may allow large activity changes with only moderate concentration changes to provide sensitive response to external stimuli (Goldbeter and Koshland, 1981
). To supplement this model our analysis offers additional new evidence for how reversible phosphorylation events may modulate reversible transitions between different discrete conformations or oligomeric states in homooligomeric and heterooligomeric complexes and might represent an important mechanism for regulation of protein activity.