Drug discovery is an inherently inefficient process, particularly in oncology. The difficulty in matching the immense and complex chemical world with a desired physiological effect is illustrated by limitations such as harmful side effects and drug resistance, which defy the most powerful chemotherapeutics available. Novel therapeutic targets and new ways to identify, to characterize and to develop anti-cancer drugs are needed. Most drug discovery efforts by pharmaceutical companies concern the development and/or expansion of their pre-clinical and clinical pipeline primarily targeting G protein-coupled receptors, nuclear receptors, ion channels and the active sites of enzymes (eg kinases). Although this strategy is understandable for historical reasons and risk management, inhibitors of protein-protein interaction represents an alternative and almost unexplored reservoir for drug development in oncology. Thus, targeting protein-protein interaction interfaces appears a valuable and promising strategy, which is further underscored by an expected reduced occurrence of resistance that might arise due to mutations in the protein-protein interface. In this context, our objectives are to identify, to understand, to validate and to target protein-protein interaction interfaces critically involved in tumor cell signaling, with the specific purpose of facilitating the transfer of therapeutic and pharmacological targets into preclinical and clinical development programs in oncology.
The study of the mechanisms of tumorigenesis helps to understand how genetic and epigenetic changes can lead to different subtypes of cancer 1. Various molecular profiles have thus been described, with objectives such as the prognostic stratification of patients and the search for new therapeutic targets. The identification of enzyme targets has allowed the advent of targeted therapies, but protein-protein interactions (PPIs) controlling these processes provide a huge reservoir of additional potential targets, being either multiprotein complexes and/or intra-molecular interactions that regulate the enzymatic activity 2. However, PPIs have long been regarded as poorly, or even not 'druggable', due to the flexibility, the complex size and the topology of their interfaces, as well as their low compatibility with small molecules present in the libraries dedicated to high-throughput screening. Despite these limitations, a number of PPIs have been explored with success, resulting in the identification of small molecules under clinical development 3, and/or used as extremely useful chemical probes for the decryption of the molecular mechanisms of normal and pathological processes, as exemplified recently with JQ1, an inhibitor of protein bromodomains (BRDs) 4.
In this context, the ISCB team develops innovative approaches to define, identify, characterize and target "druggable" PPIs in oncology, and operates usefully designed chemical probes to decipher the targeted molecular processes, thus opening new therapeutic perspectives.
This database and its associated tools (Figure 1) developed by P Roche and X Morelli in the team, have been cited in almost 200 research articles, interrogated > 50K times in the last 3 years, and downloaded from more than 50 countries. The related work has been presented in 10 international conferences and the latest version (v2) of the database has recently been published8.
The screening of collections of chemical molecules to target PPIs has so far delivered a lower success rate of hit than expected, particularly because of their inadequacy to the specific chemical space that is found in protein-protein interfaces5. Using data available on the compounds that modulate protein interactions over fifty complexes, our team has firstly developed a database (2P2IDB, Figure 1) 6-8, and an associated website, defining what makes PPI targets 'druggable', and, secondly, a comprehensive list of descriptors of molecules successfully acting on these interfaces (http://2p2idb.cnrs-mrs.fr). Using a support vector machines (SVM) machine-learning approach, we then developed 2P2IHUNTER, an algorithm based on these criteria for filtering chemical libraries in silico in order to extract potential orthosteric modulators of protein-protein interfaces 9. This innovative tool was applied to 8.3 million available commercial compounds and allowed the construction of a diverse chemical library of 1664 compounds, dedicated to the inhibition of PPIs 10. The screening of this chemical library on 5 distinct PPIs complexes revealed a hit rate 10 to 50 times higher than expected for a non-filtered one, which correlated with the druggability of the targeted interfaces, thus validating our global approach 11.
Figure 1: The structural database 2P2IDB gathers all PPI families for which the 3D structure of protein-protein and protein-inhibitor complexes have been characterized experimentally. Various tools are available to analyze the properties of the protein-protein or protein-ligand interfaces (2P2I-INSPECTOR) or to predict their 'druggability' (2P2I-SCORE) and a machine-learning tool (2P2I-HUNTER) allows to conceive PPI-focused libraries. A resulting 1,664 compound 2P2I3D library has been successfully screened against several structurally diverse protein-protein targets 11.
The success obtained with this chemical library allowed the creation of a French consortium (3 academic laboratories, 5 academic screening platforms and 1 Biotech company, Hybrigenics) coordinated by X. Morelli and supported by the French National Research Agency (ANR 2015-2018) with the aim to create a focused chemical library of 10,000 compounds dedicated to the inhibition of protein-protein interactions. This library will be acquired, plated & evaluated on 12 PPI targets. Following this evaluation/validation, replicates of the chemical database will be ‘freely’ distributed to the French academic laboratories (and biotech’s or pharmaceutical companies, upon agreement).
The quality and value of our chemical library were particularly illustrated towards the targeting of two bromodomains (BRDs) present in each member of the BET family (BRD2, 3, 4 and BRDT) for which we have identified several orthogonally validated compounds, displaying differential selectivity towards distinct BRDs and, more remarkably, one compound being the only currently available one selective of the first BRD4 BRD 12.
Figure 2: Identification of a selective BRD4 (BD1) acetylated-mimic xanthine inhibitor 1. (a) 2D Structure. (b) Homogeneous Times Resolved Fluorescence (HTRF) selectivity assay on all BET members and ATAD2 as a non-BET contro. (c) 3D structure (d) HTRF on a non selective compound. (e) Isothermal Titration Calorimetry (ITC). (f) 3D structure alignment of different derivatives.
- 1- Rodriguez-Paredes et al. Nat Med 2011;17:330–339.
- 2- Vidal M. et al. Cell 2011;144:986–998.
- 3- Scott DE. et al. Nat Rev Drug Discov 15, 533-550.
- 4- Filippakopoulos P. et al. Nature. 2010 Dec 23;468(7327):1067-73.
- 5- Morelli X. et al. Curr Opin Chem Biol 2011;15(4):475–481.
- 6- Bourgeas R. et al. PLoS One. 2010 Mar 9;5(3):e9598.
- 7- Basse MJ. et al. Nucleic Acids Res. 2013 Jan;41(Database issue):D824-7.
- 8- Basse M., et al. Database Mar 15;2016.
- 9- Hamon V, et al. J. R. Soc. Interface. 2013 Nov 6;11(90):20130860.
- 10- Hamon V, et al. Med Chem Comm 2013 ; 4:797-809.
- 11- Milhas S, et al. 2016 Aug 19;11(8):2140-8
- 12- Raux, B., et al. J.Med.Chem. 2016 Feb 25;59(4):1634-41.
2 CRCM facilities are tighly associated with the iSCB research group. Trget a precinical facility, and INT3D a molecular modeling facility:
TrGET is a preclinical platform, performing and developping in vitro and in vivo testing of target genes involved in tumorogenesis or antitumoral therapeutics. Great efforts are provided towards technological developments, including assay, tool and both cell and animal proposed models. TrGET affers a variety of flexible and customizable services, with privileged development of bioluminescence-based technologies (in vitro or in vivo). Staff: Y. Collette (20%), A. Restouin, R. Castellano, A. Goubard.
The INT-3D platform has been created at the beginning of 2008 by Dr. Xavier Morelli and Dr. Philippe Roche to answer to an increasing demand of the academic laboratories regarding molecular modeling (virtual mutagenesis mainly) and virtual screening (hit finding and hit to lead optimization). Staff: X. Morelli (20%) - P. Roche (20%) - MJ. Basse (50%)