As we have seen in equation 4.1, affinity can be expressed by the difference in free energy ∆G. A change in ∆G for an interaction may thus, in principle, influence its specificity. Hence, mechanisms that lead to changes in ∆G are potential mechanisms to manipulate specificity [128]. Greenspan [133] mentions three major causes for al-tered affinity, these are changes in: first, shape complementarity; second, chemical complementarity and third, molecular flexibility. In addition, Szwajkajzer et al. [128]
emphasise the importance of entropy that contributes to the free energy. Changes in entropy upon binding are expressed by changes in the flexibility of the interact-ing molecules but also by changes in their hydration shells. Upon bindinteract-ing, water molecules or ions can be released or sequestered by the interacting proteins. Thus, binding affinity can be increased by reducing the entropic costs of complex formation, e.g. by locking the interacting molecules into the binding conformation and by control-ling the solvation on the surface of the molecules. This is the basic idea of host-guest chemistry (see previous section) [128]. In such a system, optimal complementarity will lead to binding specificity.
Cellular protein complexes are very different from the complexes observed in host-guest chemistry regarding the molecular flexibility. Proteins usually encounter a loss of conformational flexibility and substantial reorganisation of the hydration shell when they bind to other proteins [128]. Disorder to order transitions of protein structures upon binding are common (see section 2.4). Such transitions cost free energy but on the other hand allow for induced-fit interactions that can lead to better fits between protein and ligand. Increases in shape and chemical complementarity between a pro-tein and a ligand do not necessarily lead to an increase in affinity if entropic costs outweigh the enthalpy gain [128]. This illustrates that changes in free energy upon binding are mostly very complex. Attempts to increase the binding specificity between two interacting proteins by studying solely the structure of the complex neglects the thermodynamic nature of protein interactions and are therefore much less likely to be successful [128].
5.4. Biological aspects of the specificity of protein interactions
Protein interaction specificity at the molecular level is often studied between minimal fragments of proteins, e.g. globular domains and SLiMs, that define the core bind-ing interface between the two interaction partners. This limitation mostly originates from the fact that larger proteins are more difficult to maintain in stable and active forms for biochemical or biophysical assays. Similar size limitations apply to molecular modelling approaches where bigger molecules exceed the calculation power of available computer clusters. Of course, we can gain important insights on the molecular
mecha-nisms of interaction specificity when focussing on minimal interaction fragments. Yet what can we learn from these results about the specificity of protein interactions in the cell?
5.4.1. The influence of sequence context on the specificity of protein interactions
Numerous studies provide increasing evidence that extended protein fragments or full length proteins display altered or/and more intermolecular contacts than those ob-served between the minimal interacting fragments. These changes in molecular con-tacts were shown to substantially alter the binding affinity and sometimes specificity of the interaction. In addition, it has been shown that the structure and/or dynamics of minimal interacting fragments can be dramatically changed in the extended or full length context (for more details, see our review presented in chapter 10). The term sequence context is used to describe these regions in protein sequences that influence the binding interface of the minimal interacting fragments. Sequence context can be extensions of the core interacting fragments, neighbouring domains, or other regions of the sequence that are not contiguous to the fragments under consideration (see Figure 5.1). Intramolecular allostery and cooperative binding can be seen as examples where sequence context influences the binding properties of two interacting proteins.
Intramolecular allostery includes cases where the binding of a region of a protein se-quence to a globular domain that it carries leads to conformational changes of the binding pocket of this domain that in turn alters binding to its target. Cooperative binding describes cases where several distinct binding interfaces, e.g. formed by several globular domains and linear motifs of the two interaction partners, cooperate to give an overall stronger interaction [43].
5.4.2. Specificity vs. multi-specificity vs. promiscuity
The study of large protein-protein interaction networks revealed that many proteins, called hub proteins, interact with a huge number of other proteins. Given these many interaction partners, it may seem that hub proteins are unspecific [136, 137]. Let us consider the example of the highly studied tumour suppressor protein p53. p53 has more than 230 identified interaction partners in the database STRING (see section 6.3) [138]. p53 is implicated in the regulation of important biological processes such as cell division, cell growth, apoptosis, and transcription [139]. Given these essential functions, it seems impossible that p53 will not specifically bind its targets.
The concept of modular protein architecture (see section 2) may partially serve in re-solving these at first sight contradictory observations [43,136,137]. Possessing different modules confers different interaction sites on a protein. Especially disordered regions allow a protein to have different interaction sites (e.g. different classes of SLiMs) lo-cated within a relatively short region of a protein. Half of the about 390 residues of p53 are predicted to be disordered [140]. They encode for numerous overlapping
A B C
Figure 5.1. Examples of sequence context in the PDZ domain family. PDZ domains, bound C-terminal peptides, and the parts of the structures that constitute the sequence context are coloured in light blue, dark blue, and red, respectively. A: PDZ3 of Par3 bound to an extended PBM derived from PTEN (PDB code: 2K20 [12]). Residues of the extended PBM and key interacting residues of the PDZ domain are shown in sticks. B: PDZ3 domain of PSD-95 bound to a PBM derived from CRIPT (PDB code: 1BE9 [8]). PDZ3 possesses an additional C-terminal αhelix that influences peptide binding [13, 14]. C: PDZ3 of ZO-1 bound to a PBM derived from JAM-A (PDB code: 3TSZ [15]). The neighbouring SH3 domain that is located C-terminal to PDZ3 influences peptide binding.
SLiMs including many sites for PTMs [43] (see Figure 5.2).
Using one particular interaction interface, a protein A will bind ligands of a par-ticular type X (e.g. a particular class of SLiMs). If protein A binds a few ligands of type X with much higher affinity than other ligands of type X, then protein A is specific for this type of ligands and the employed interaction interface. At the same time, protein Amight be able to bind as well to numerous other proteins using other types of interaction sites (e.g. other globular domains or SLiMs). Thus, the question whether a protein such as p53 is promiscuous or specific cannot solely be answered by looking at the total number of its interaction partners but rather by concentrating on groups of interaction partners that share similar binding interfaces. The term multi-specificity might be used to describe such cases where a protein binds specifically to different types of ligands via different interaction interfaces.
5.4.3. The influence of cellular context on the specificity of protein interactions
Pairwise interactions between proteins depend not only on their mutual binding energy but also on other parameters that influence their localisation, concentration and active state in the cell. Proteins are usually expressed at precise moments during the cell cycle or under certain physiological conditions. Proteins are actively transported and localised to certain cellular compartments or sub-locations in the cytosol. Proteins become activated or inhibited via PTMs. Individual proteins often carry out their functions in complexes composed of multiple proteins. Often, an interaction between
Figure 5.2. Binding interfaces of p53 with structural evidence. Protein modules of p53 are illustrated with red (globular domains) and yellow (SLiMs) boxes at the bottom of the figure. The tetramerisation domain of p53 is shown in purple. Published structures in the PDB involving these interaction sites are displayed in cartoon representation using the same colour code to highlight structure segments of p53. PDB codes of these structures are indicated in yellow ovals. This figure has been extracted from [43].
two proteins prerequisites their assembly into a multiple protein complex via a scaf-folding protein. Overlapping interaction sites in one protein (see Figure 5.2) allow for molecular switching, e.g. if one ligand is bound, interactions with other potential lig-ands may be disabled [141]. All these highly regulated processes point to a discrete and deterministic cellular system controlling that protein interactions are formed at the right moment and place [43]. In this regard, interaction specificity between proteins in biological systems might only be understood when investigating both the molecular and cellular mechanisms.