Protein-protein interactions form the basis of many cellular processes. Disruption or deregulation of these complex interactions is the main cause of a significant number of human ailments. Consequently, there is intense research effort to design inhibitors that target specific protein-protein interactions. This places intricate protein-protein interactions in the heart of the development for novel drug leads. The emergence of ‘omic’ technologies, namely genomics, transcriptomics and proteomics, has greatly accelerated our understanding of the protein-protein interaction networks leading to the discovery of a number of proteins and their interaction interface as potential drug targets.
The “druggable proteins” are targeted by commercially viable, and preferably orally bioavailable, therapeutics. These drugs are usually small organic molecules that function as competitive or noncompetitive proteins inhibitors. On the other hand, many “undruggable” proteins are important targets in various disease states. These proteins are considered undruggable because they lack a cavity for the small organic inhibitors to bind and they interact with their protein partners through extensive and flat surfaces. The use of protein based therapeutics expands the repertoire of “druggable proteins” by targeting those proteins that cannot be inhibited by the available small molecules. Besides their improved specificity offered by their high compatibility with the target proteins, the major advantage of peptide therapeutics is their reduced immunogenicity and improved safety. On the other hand, low oral bioavailability, low protease/peptidase resistance, low cellular uptake, high rate of hepatic and renal clearance, high biodegradability and high flexibility are some limitations of peptides as therapeutics.
Peptide drugs take advantage of the highly specific and selective interaction between proteins. The peptide is usually based on the sequence of the binding region between the two proteins. The linear sequences might originate from a loop within a structured domain, or from a disordered region in protein termini or between defined domains. In order to achieve desired efficacy, following delivery to the correct cellular compartment, the designed peptide needs to target the appropriate site and bind it. A perfect therapeutic agent is then a short protein sequence that will reach, bind and modulate the function of a target protein for the required amount of time and efficacy. These linear sequences are difficult to discover due to their short length and a tendency to reside in disordered regions in proteins. Increase in the available structural information on the protein – protein interactions has spurred the design of novel peptide therapeuticals. Furthermore, it is now possible to screen and select high affinity peptides for these targets with the advent of peptide libraries and array techniques.
Peptide drugs may function by interacting with different targets such as proteins, lipids, nucleotides or metabolites. Particularly, there is significant research focused on antimicrobial peptides that target lipid cell membranes. This review focuses on the challenges and opportunities in the design and development of peptide based drugs that bind and inhibit some important protein targets.
Diverse cellular events such as protein and vesicle trafficking, gene expression, DNA repair, control of the cytoskeleton and targeted protein degradation as well as signaling cascades are regulated through dynamic protein interactions. Enhancing the efficacy of a peptide therapeutic addressing one of these processes is tightly bound to basic principles governing protein-peptide interactions. Despite their significance and estimated abundance, a large fraction of protein-peptide interactions lack detailed characterization and some questions of scientific and commercial interest remain: How does a peptide overcome the energetic cost involved in switching from an unstructured, flexible peptide to a rigid, well-defined bound structure? What is the recognition process for the binding event? What stabilizes these interactions? If a peptide binds to a protein, what is the spatial configuration and what is the strength of this interaction? If they don’t bind each other, can they be made to bind by modifications? The increasing number of resolved protein-peptide structures sheds light into the mechanistic details of binding.
The rational design of peptide drugs is stimulated by the availability of structural information on protein – protein complexes. Peptides derived from the binding region of an inhibitor protein usually serve as a starting point in the design of peptide inhibitors against the protein – protein interaction. Coordinate and distance information about the binding interface can be based on X-ray crystallography or NMR methods. Other experimental methods that identify interface residues include alanine scanning mutagenesis, chemical modification, mass spectrometry and phage display.
We have filtered the Protein Data Bank for the keyword “peptide” in the structure description and with chain length between 5 to 35 amino acids and found a total of 1816 crystal structures and 307 structures determined by solution NMR in March 2012. Additionally, two electron microscopy structures of the Escherichia coli 70S ribosome in the presence of the leader peptide were also reported. A 2010 study clustered the complex structures determined by crystallography in the Pep-X database (https://www.frankenthalerfoundation.org). This database contains 505 nonredundant protein-peptide interface complexes. 14% of these complexes are with the Major Histocompatibility Complex, 12% of them are with thrombin and 8% are with alpha-ligand binding domain. Another nonredundant protein – peptide database of 103 structures was reported in 2010 by the Schueler-Furman group.
The peptide binding site is usually a large and shallow pocket on the protein surface and it does not change its conformation upon peptide binding. In addition, hydrogen bonds with the peptide backbone and interactions with hot spot residues provide the enthalpic contribution to protein – peptide recognition. The protein – peptide interface is enriched in Leu and Ile as well as aromatic residues. The protein – peptide interface was shown to resemble the core of the protein, with more hydrophobic residues than the protein surface and with the structural motifs found in protein folds.
It has become apparent that a significant number of protein interactions are commonly formed between conserved protein recognition domains and short linear peptide motifs, often less than 10 amino acids in length. Members of a given protein domain family usually recognize a consensus motif but they may recognize different variations of this motif and they may possess unique binding specificities.
Peptides can interact with globular protein domains in very diverse ways. These include binding of a peptide onto a protein domain by forming an additional beta-sheet, binding to clefts in extended beta or proline type II helical conformations or adoption of a helical conformation. For example, SH2 and phosphotyrosine-binding (PTB) domains recognize phosphotyrosine motifs, while polyProline helices are recognized by SH3, WW and EVH1 domains. 14-3-3 proteins, FHA and WD40 domains recognize phosphothreonine/serine-containing elements; bromo and chromo domains recognize acetylated or methylated lysine; VHL proteins recognize hydroxyproline motifs. On the other hand, short amino acid motifs at the carboxyl termini of target proteins, such as ion channels, are important for recognition by PDZ domains.
Design of peptide based inhibitors against proteins with such modules is hampered by the similarity between the recognized peptide sequences. However the structural information available clarifies many ambiguities regarding protein-peptide interactions. The specificity and selectivity of the protein modules in the cell suggest the presence of a mechanism whereby a selective peptide drug can be designed that interferes with the binding of protein domains to their respective partners.
The primary steps involved in the design and development of peptide ligand drugs are illustrated. The initial step in peptide drug design is the identification of the protein target. This is usually a protein that is implicated in a disease state. If possible (and/or available), the interaction partners of the target protein are also determined. Information from structure-activity relationship studies is then used for rational design. Structural information of the protein – protein interface is fundamental for rational drug design. If there is no information about the interacting partner, combinatorial approaches, such as phage display, peptide arrays or peptide aptamers, should be used to screen tight binding peptide sequences. Rational design may follow combinatorial approaches to design a peptide sequence with improved specificity and higher affinity. Once a tight binding potential peptide sequence is identified, the peptide is usually modified to enhance stability, uptake and delivery. These may include alteration of amino acids to nonnatural amino acids, cyclization of the peptide or constraining the peptide so that it forms an alpha helix. This modified peptide is a peptidomimetic, which has the properties of the peptide with respect to binding mechanism but also has higher stability and uptake potential than a natural peptide ligand. After in vitro tests of the modified peptide, in vivo tests and clinical trials are performed. Peptide may undergo further modifications during these tests. The ones that pass clinical trials are then marketed.
Increase in the availability of crystallographic structures of protein complexes has conveyed valuable information for rational drug design efforts. Given a known (or predicted) protein – protein complex structure, inhibitors that target the interface between the two proteins can interfere with this interaction. Design of peptides based on the interface has also been an area of intense research.
Docking of small organic molecules to protein targets has shown good progress with the advent of docking, virtual screening and pharmacophore building algorithms. However, the prediction of the complex structure between a peptide ligand and its protein partner is not easy due to the flexible nature of peptides. The solution structure and the bound structure of peptides are usually different, with the peptide adopting its bound conformation only in the presence of the protein. In addition, protein – peptide docking studies are further complicated by the absence of a cavity for peptide binding, because protein – peptide interaction sites are usually shallow pockets on the surface. Several algorithms have been proposed for protein – flexible peptide docking. Three recent algorithms are the molecular dynamics based Dynadock, the Monte Carlo based FlexPepDock.
