2. Protein-protein interactions: Large-scale analysis of peptide recognition modules & high-throughput generation of peptide-based binders.

• Analysis of peptide recognition modules binding specificities

Protein-protein interactions are crucial for the assembly of multiprotein complexes and cellular signaling pathways. One of the central branches of protein science is concerned with understanding the energetics responsible for affinity and specificity in protein-protein interactions. We have developed rapid combinatorial methods that can be used to quantitatively survey binding energetics across protein-binding surfaces using molecular biology and high throughput sequencing (Ernst et al, 2010). These combinatorial methods alleviate one of the major bottlenecks of protein research and represent a novel approach to the analysis of protein-protein interactions.
Protein-protein interactions are mediated by the recognition of peptide motifs by Peptide Recognition Modules (PRMs). These domains are present in many proteins and are expected to bind their target by the same general principles, yet possess the means to selectively bind a specific target. To better understand factors that impact both the generality and selectivity of peptide binding by PRM domains, we have mapped the peptide binding specificities of specific PRMs, and developed a systematic analysis of these module families on a genome-scale level (Tonikian et al, 2007; Gfeller et al, 2011). Our studies have revealed that most PRMs adhere to the known canonical binding modes, but that some also use additional, previously uncharacterized binding modes. These are overlayed by unique cognate ligand binding features. We used specificity profiling of PRM point mutants to test the robustness of the binding site, and better understand factors contributing to specificity. We also used complementary techniques (two-hybrid, peptide arrays) to identify binding partners, which allowed us to generate interaction maps for the PRMs (Sidhu et al, 2003). These maps allow prediction of protein partners and protein localization, and are used in further development of protein-protein interaction prediction algorithms (Sidhu & Kossiakoff, 2007).
Thus far, we have focused a lot of our effort on PDZ domains (which recognize specific C-terminal sequences). We constructed phage-displayed peptide libraries to define binding profiles for over 200 natural and mutant PDZ domains (Fuh et al, 2000; Tonikian et al, 2008). This enormous data set has been used to predict natural interaction networks (Laura et al, 2002; Ivarsson et al, 2014), to guide structural and functional analysis of key representatives of family subclasses, and to develop computational algorithms that can predict PDZ domain specificities from primary sequence (Lam et al, 2010; Yip et al, 2011; Kim et al, 2012).

To further understand the mode of action of PDZ domains, we used a combination of methods to investigate the molecular contribution of residues within specific PDZ domains to protein-protein interaction (Murciano-Calles et al, 2014). These include peptide libraries and affinity assays to define specificity, structural studies to view the molecular details of ligand recognition (Appleton et al, 2006; Zhang et al, 2007; Runyon et al, 2007), and alanine scanning mutagenesis to investigate the energetic contributions of individual residues to ligand binding (Weiss et al, 2000; Pal et al, 2003; Skelton et al, 2003; Pal et al, 2005 (2 articles); Zhang et al, 2006; Pal et al, 2006). These studies have revealed that two types of ligands are recognized by PDZ domains, the C-terminal peptide and an internal peptide, and that the details of these interactions differ. We have determined the individual contributions of specific amino acid residues to binding within peptide ligands, showing that some residues form a core motif, while other residues contribute to or modulate the interaction. These findings contribute to our understanding of both the common principles of binding of PDZ domains with their ligands, and highlight how subtle sequence changes that affect specificity can significantly alter the range of biological partners for PDZ domains.
In addition, the ability to manipulate protein-protein interactions is crucial for biological research and therapeutic intervention. Therefore, we are using protein engineering methods to modulate protein function by altering binding of PRMs to motifs. For example, we analyzed the binding specificity of PDZ domain variants engineered in vitro and defined synthetic repertoires of functional binders that are as specific and diverse as the natural family (Ernst et al, 2009). Our results are shedding light on the mechanisms of evolution of protein domain families and complex protein interaction networks.
The methods we have developed for high throughput protein purification, phage selections and data analysis have also been applied to a comprehensive analysis of SH3 domains (Alto et al, 2007; Tonikian et al, 2009; Xin et al, 2013) and WW domains (Gupta et al, 2007; Abu-Odeh et al, 2014) (both of which bind specific proline-rich motifs), and we are now extending our studies to other PRMs.

Binding variants as inhibitors of biological function and potential therapeutics

Peptide ligands are promising small-molecule therapeutic candidates for devastating diseases such as cancer, in which inhibition of critical signaling pathways holds much promise. In principle, some natural proteins could be used as therapeutic agents, but their target binding affinities often precludes their use in a clinical setting. Using a phage display strategy and libraries of variant proteins designed based on crystal structure information, we can evolve high affinity variants that show increased binding affinity and improved activity compared to the wild type protein (Deshayes et al, 2002; Li et al, 2009). These engineered proteins act with greater potency, and therefore may have greatly improved therapeutic efficacy. Our phage display strategy for isolating peptides with high affinity and specificity to a given PRM or full length protein not only allows validation of suitable domains to target for therapeutic intervention, but also yields candidate peptide inhibitors that can be tested for activity and affinity matured, and further tested or used for guiding the design of small-molecule inhibitors. The peptide-PRM or peptide-protein complex can be used for further understanding of the structural and molecular mechanisms of function of the particular motif or protein, and the peptides themselves are valuable tools for probing the cellular functions of the protein.
Using this technology, we have recently developed a novel and general strategy for developing potent and specific inhibitors of hundreds of ubiquitin ligases and deubiquitinases in the ubiquitin proteasome system (Ernst et al, 2013; Zhang & Sidhu, 2014). By using phage-displayed libraries of ubiquitin variants, we have been able to generate inhibitors that have been used for structural studies and as intracellular tools for the exploration of cell biology and for target validation.
We have also used the technology to generate peptide inhibitors of the Wnt/β-catenin pathway (Zhang et al, 2009), DR5 (Li et al, 2006), IGF-1 (Deshayes et al, 2002, Schaffer et al, 2003; Skelton et al, 2004), and Collagenase (Bahudhanapati et al, 2011) as well as SH2-based inhibitors of Fyn (Kaneko et al, 2012).