Current Topics in Medicinal Chemistry - Volume 7, Issue 10, 2007

Protein-protein interactions (PPIs) are ubiquitous in biological systems, with essential functions in key of biological processes such as cell growth and differentiation, intracellular signaling and programmed cell death (apoptosis). Due to the pivotal role of PPIs, protein-protein interfaces have garnered attention as novel targets for therapeutic intervention in oncology, neuroscience and antiinfectives. However, the pharmaceutical industry has long regarded PPIs as intractable targets, which could only be disrupted by large macromolecules and peptides. Indeed, PPI targets are considerably more complex than classical enzyme targets, characterized by well-defined binding pockets and small molecule substrates, for a number of reasons. First, the natural protein ligand polypeptides of a PPI do not afford design opportunities for the development of small molecule leads and protein surfaces/residues which contribute to the binding interface are often unknown. Second, small molecule inhibitors appear to be at a disadvantage as the interacting protein-protein surfaces are large, featureless and often contain buried surfaces essential to the PPI. Finally, few small “drug like” leads have emerged from classical high-throughput screens aimed at identifying disruptors of PPIs, and alternative screening techniques are often required. Despite these significant challenges, progress has been made and small molecule disruptors of PPIs have been discovered. In large part, recent success in this field has been due to the pioneering work of Clarkson and Wells and the discovery of “hot spots”. Hot spots are small, high affinity regions at the protein-protein interface of many PPIs and can account for the majority (>80%) of the binding energy. As a result, disruptors of PPIs do not need to mimic the entire protein binding surface, but rather a smaller subset of key residues, suggesting small molecules can disrupt PPIs. There are now reports of a staggering diversity of small molecule chemotypes that disrupt PPIs and the concept of ‘privileged structures’, usually associated with GPCR ligands, can now be easily applied to small molecule PPIs. This issue contains a collection of reviews that describe state-of -the art screening techniques, medicinal chemistry strategies and case studies for various protein-protein targets. The authors have included the background biology needed to understand the target rationale and then have moved on to the screening paradigm, lead discovery and drug optimization stages as appropriate. The reader will find introductory reviews describing PPI chemotypes, screening techniques and system biology approaches followed by case studies on such important PPI targets as p53/HDM2, XIAP, Bcl-XL-BH3 and Nrf2-Keap1. I hope that this Issue serves to inspire and inform researchers interested in protein-protein interactions and accelerates scientific discovery in this arena.

Protein-protein interactions play crucial roles in a number of biological processes, and, as such, their disruption is becoming an area of intense research. Despite the many challenges associated with the development of protein-protein interaction inhibitors, such as the large and relatively featureless interfacial areas involved, there has been considerable success in recent years. Importantly, through the existence of protein “hot spots”, some of this success takes the form of small molecule inhibitors that have been identified from a variety of screening techniques.

Due to the pivotal roles that protein-protein interactions play in a plethora of biological processes, the design of therapeutic agents targeting these interactions has become an attractive and important area of research. The development of such agents is faced with a variety of challenges. Nevertheless, considerable progress has been made in the design of proteomimetics capable of disrupting proteinprotein interactions. Those inhibitors based on molecular scaffold designs hold considerable interest because of the ease of variation in regard to their displayed functionality. In particular, protein surface mimetics, α-helical mimetics, β-sheet/β-strand mimetics, as well as β- turn mimetics have successfully modulated protein-protein interactions involved in such diseases as cancer and HIV. In this review, current progress in the development of molecular scaffolds designed for the disruption of protein-protein interactions will be discussed with an emphasis on those active against biological targets.

Protein-protein interactions are increasingly becoming drug targets. This is understandable, since they are crucial at all levels of cellular expression and growth. In practice, targeting specific disease-related interactions has proven difficult, with success varying with specific complexes. Here, we take a Systems Biology approach to targeting protein-protein interactions. Below, we first briefly review drug discovery targeted at protein-protein interactions; we classify protein-protein complexes with respect to their types of interactions and their roles in cellular function and as being targets in drug design; we describe the properties of the interfaces as related to drug design, focusing on hot spots and surface cavities; and finally, in particular, we cast the interactions into the cellular network system, highlighting the challenge of partially targeting multiple interactions in the networks as compared to hitting a specific proteinprotein interaction target. The challenge we now face is how to pick the targets and how to improve the efficiency of designed partiallyspecific multi-target drugs that would block parallel pathways in the network.

This article describes recent progress in the development of small molecule protein-protein inhibitors of the p53-MDM2 (purine double minute 2, or HDM2 for the human congener) protein-protein interaction, with special attention to the diversity of chemotypes reported to disrupt this protein-protein interaction. In >50% of all human cancers, the tumor supressor 53 KDa phosphoprotein p53 is either mutated or deleted. The discovery that MDM2 (HDM2) negatively regulates p53 and therefore inhibits the tumorsupressor activity of p53 has instigated numerous drug discovery campaigns aimed at disrupting this protein-protein interaction as a potential cancer therapy. Once regarded as intractable targets disrupted by only large macromolecules, protein-protein interactions (PPI) are now mainstream targets due in large part to the intensive effort applied to the study of p53 and the surprising diversity of small molecules (peptides, natural products, terphenyl and other α-helix mimetics, chalcones, piperidines, piperazines, fused indoles, isoindolinones, spiro-oxindoles, cis-imidazolines (nutlins), quinolinol and benzodiazepines) capable of disrupting the p53-HDM2 PPI. In addition, drug discovery researchers have employed a number of screening approaches and technologies to identify SMPPIs of the p53- HDM2 interaction, and these discovery paradigms will be dicussed. This review will detail the biology of the p53-MDM2 interaction, the major classes of SMPPIs and key medicinal chemistry and in vitro/in vivo biological data reported through October 2006.

The Bcl-2 family of anti-apoptotic proteins are key regulators of programmed cell death. Bcl-2 and its closely related Bcl-XL counterpart are one of several pro-survival proteins which can share up to four highly conserved domains known as the BH1, BH2, BH3 and BH4 domains. These domains form the basis of a well defined groove whereupon a heterodimeric protein-protein interaction can occur with pro-apoptotic BH3 proteins such as Bad, Bid and Bim. Extensive evidence clearly indicates a strong correlation between neoplastic progression and deregulation of apoptotic pathways. Overexpression of Bcl-XL is associated with tumor progression, poor prognosis and resistance to chemotherapy. Antagonism of Bcl-XL is therefore viewed as a means to mimic the endogenous apoptotic pathways initiated by Bad, Bid and other pro-apoptotic proteins. Several successful approaches to block the Bcl-XL-BH3 binding groove have been reported but only recently have proteomimetics been found which could prove to be clinically useful as new anticancer agents capable of overcoming apoptosis resistance. ABT-737 is an example of one of the first small-molecule inhibitors of Bcl-2/XL proteins shown to be efficacious in vivo, causing complete regression in small-cell lung carcinoma tumour xenografts in mice. This review will focus on the recent advances surrounding the non-peptidic Bcl-2/XL inhibitor ABT-737 developed by Abbot laboratories and highlight the key structural characteristics found within this unique BH3 alpha-helical mimetic.

The X-linked inhibitor of apoptosis proteins (XIAP) is thought to play a key role in the unchecked proliferation of cancer cells by interfering with the signaling cascade leading to cell death. The structure and mechanism of XIAP has been widely investigated and characterized over the past few years, to the point where this may be the best understood apoptosis protein inhibitor. As a result, XIAP is viewed as an attractive target for the treatment of cancer. To date, several research groups have reported on the discovery of small molecule inhibitors of this protein. This review focuses on the discovery and optimization of these leads.

The transcription of antioxidant response element (ARE)-containing cytoprotective genes has been proposed as a means to combat oxidative stress-related disorders, such as cancer and Parkinson's disease. Transactivation of the ARE requires the transcription factor nuclear factor erythroid 2-related factor 2 (Nrf2). Cellular levels of Nrf2 protein are regulated by the Kelch-like ECH-associated protein 1 (Keap1), a substrate adaptor protein for the ubiquitin ligase machinery and subsequent proteasomal degradation. Recently, detailed studies have elucidated the structure and interactions of the Keap1-containing ubiquitin ligase complex. Here, we propose that small molecule modulation of Keap1 protein:protein interactions may permit Nrf2's nuclear accumulation and the transcription of AREdependent genes to enhance cellular resistance to oxidative insult.

The present CTMC issue entitled ‘Computational Approaches in Medicinal Chemistry and Drug Discovery’, covers a wide range of methodologies and applications that are having an increasing impact in the search for drug leads both in academic and pharmaceutical industry settings. These reviews deal with the wide gamut of methodologies used in this field ranging from the application of QSAR based techniques (see contributions by González-Diaz et al. and Volarath et al.) to receptor based binding predictions of Cavasotto and Orry, Villaverde et al. and Volarath et al. The latter algorithms have spanned the Structure-Based Virtual Screening (SBVS) of novel protein ligands, an in silico technique that can be used successfully in synergy with the wet lab High Throughput Screening protocols (HTS), in the search for drug leads with novel chemical moieties, as shown by Cavasotto and Orry. It is worth noticing that the computer assisted methods presented here could also have a bearing on a large number of disciplines (including proteomics) assisted and supported by the use of QSAR protocols based on topological indexes (see review by González-Diaz et al.). The drug targets are also well represented. They include the HIV-1 PR protease (see review by Volarath et al.), arguably the most successful receptor based rational lead design endeavor in the history of drug discovery. As it is well known, drugs based on inhibitors of this enzyme have made a huge impact in the life span and quality of life of AIDS patients. The experience gathered for an enzymatic target could be employed in targeting other enzymes of the same family. That has been the case of β-secretase (a member of the same protease family as the HIV-1 protease) which is crucial for the development of amyloid plaques in the brain of Alzheimer diseased patients. Although the inhibitor sequence specificities are quite different for both enzymes, similar kinds of isosteres, as those employed for HIV-1 PR, are currently being used for the design of β-secretase inhibitors. Based on our current involvement in a β-secretase inhibitor discovery program we reviewed this burgeoning field in the light of computer assisted approaches on drug discovery (see review by Villaverde et al.). No review issue on computer assisted drug design could be complete without a chapter on G-coupled protein receptors (GCPRs), one of the largest protein families whose functioning is activated by a wide range of external signals (odor, light, etc) as well as internal signals (ions, hormones, neurotransmitters, etc). These proteins are the target of about 40% of the prescribed drugs and of around 25% of the topselling drugs. GPCRs interact with an extraordinary diversity of ligands by means of their extracellular domains and/or the extracellular part of the transmembrane (TM) segment. Each receptor subfamily has developed specific sequence motifs to adjust the structural characteristics of its cognate ligands to a common set of conformational rearrangements of the TM segments near the G protein binding domains during the activation process. This adaptation has been achieved during evolution by customizing a preserved 7TM scaffold through conformational plasticity. Deupi et al. have contributed to this issue a review on this subject that helps to explain the functional versatility of these molecules. A large number of diseases can be explained nowadays in terms of protein aggregation. These include degenerative CNS diseases, some types of diabetes, etc. The group headed by Nussinov has been at the forefront of the computer assisted study of protein-protein interactions and has contributed many new insights to this field. In this issue Ma and Nussinov review their seminal work on the detection of residues that constitute hot spots for protein interaction and aggregation, a crucial issue in the design of drugs targeted against the formation of plaques that are thought to be the causal source of these diseases.

The inhibition of β-secretase has become a very promising approach to control the onset and progression of Alzheimer's disease due to its involvement in the generation of amyloid plaques. The main goal of the many drug discovery projects targeting this enzyme is the identification of highly specific, non-peptidic compounds with low molecular weight, characteristics that are desirable for drug leads with low toxicity that have to transverse the blood brain barrier. We describe the main approaches used in the discovery of novel inhibitors, including substrate specificity, target structure based design, and high throughput screening (HTS), both in vitro and in silico. We place special emphasis in the receptor based design and in silico HTS, two strategies that make wide use of computer assisted tools. To exemplify the usefulness and versatility of computer tools in this endeavor we use the computer generated ‘enzyme's binding site cast’ to rationalize qualitatively some salient structural features of known β-secretase second generation inhibitors, and for guiding the review of many of the ligands whose complexes with the enzyme have been studied by X-ray crystallography. We discuss the use made by other authors of molecular modelling for the understanding of the very special characteristics of ligand binding to β-secretase and for the design of new inhibitors. Finally, we review the quest for non-peptidic inhibitors that has led to the use of HTS in vitro and in silico. The screening of extensive libraries resulted in a few low affinity compounds that do not fit into the key S1/S1' pockets, indicating that this is not an easy target to block.

G protein-coupled receptors (GPCRs) interact with an extraordinary diversity of ligands by means of their extracellular domains and/or the extracellular part of the transmembrane (TM) segments. Each receptor subfamily has developed specific sequence motifs to adjust the structural characteristics of its cognate ligands to a common set of conformational rearrangements of the TM segments near the G protein binding domains during the activation process. Thus, GPCRs have fulfilled this adaptation during their evolution by customizing a preserved 7TM scaffold through conformational plasticity. We use this term to describe the structural differences near the binding site crevices among different receptor subfamilies, responsible for the selective recognition of diverse ligands among different receptor subfamilies. By comparing the sequence of rhodopsin at specific key regions of the TM bundle with the sequences of other GPCRs we have found that the extracellular region of TMs 2 and 3 provides a remarkable example of conformational plasticity within Class A GPCRs. Thus, rhodopsin-based molecular models need to include the plasticity of the binding sites among GPCR families, since the “quality” of these homology models is intimately linked with the success in the processes of rational drugdesign or virtual screening of chemical databases.

Protein-protein interactions are crucial to biological functions. Consequently, designing drugs to control protein-protein interactions is receiving increasing attention. Protein structures can associate in different ways. Analysis of the structures of proteinprotein complexes using amino acid sequence order-independent multiple structural comparison algorithms, led us to conclude that the amino acids Trp, Met, and Phe are important for protein-protein interactions. Hence, in principle, drug design targeting the Trp/Met/Phe should modulate protein functions effectively. Several clusters of the Trp/Met/Phe residues are involved in the p53 protein-protein interactions. The best example in this regard is the Phe19/Trp23 of p53, which binds to transcriptional factors and to the MDM2 protein. In the HIV related proteins, the Trp/Met/Phe residues have roles in the dimerization of the transcriptase (p51/p66) and in cell-fusion processes, including the gp120-CD4 interaction and the gp41 six-helix bundle formation. Trp/Met/Phe residues are preferred in ‘normal’ functional protein-protein interactions and they also appear to be exploited in amyloid formation, especially the phenylalanine. Comparison of binding propensity and amyloid formation preference reveals that apart from Lysine, Isoleucine is the least structurally conserved in protein binding sites and has a high propensity in sequences forming amyloids. Thus, this may suggest that nature tends to avoid Ile conservation in protein-protein interaction to avoid amyloid formation. In this regards, Trp/Met/Phe as well as Ile may be targeted to modulate protein-protein interaction.

Ligand-docking-based methods are starting to play a critical role in lead discovery and optimization, thus resulting in new ‘drug-candidates’. They offer the possibility to go beyond the pool of existing active compounds, and thus find novel chemotypes. A brief tutorial on ligand docking and structure-based virtual screening is presented highlighting current problems and limitations, together with the most recent methodological and algorithmic developments in the field. Recent successful applications of docking-based tools for hit discovery, lead optimization and target-biased library design are also presented. Special consideration is devoted to ongoing efforts to account for protein flexibility in structure-based virtual screening.

The numerical encoding of chemical structure with Topological Indices (TIs) is currently growing in importance in Medicinal Chemistry and Bioinformatics. This approach allows the rapid collection, annotation, retrieval, comparison and mining of chemical structures within large databases. TIs can subsequently be used to seek quantitative structure-activity relationships (QSAR), which are models connecting chemical structure with biological activity. In the early 1990's, there was an explosion in the introduction and definition of new TIs. The Handbook of Molecular Descriptors by Todeschini and Consonni lists more than 1500 of these indices. At the end of the last century, researchers produced a large number of TIs with essentially the same advantages and/or disadvantages. Consequently, many researchers abandoned the definition of TIs for a time. In our opinion, one of the problems associated with TIs is that researchers aimed their efforts only at the codification of chemical connectivity for small-sized drugs. As a consequence, recently it seems that we have arrived at “Fukuyama's End of History in TIs definition”. In the work described here, we review and comment on the “quo vadis” and challenges in the definition of TIs as we enter the new century. Emphasis is placed on new chiral TIs (CTIs), flexible TIs for unifying QSAR models with multiple targets, topographic indices (TPGIs), TIs for DNA and protein sequences, TIs for 2D RNA structures, TPGIs and drug-protein or drug-RNA quantitative structure-binding relationship (QSBR) studies, TIs to encode protein surface information and TIs for protein interaction networks (PINs).

Significant progress over the past decade in virtual representations of molecules and their physicochemical properties has produced new drugs from virtual screening of the structures of single protein molecules by conventional modeling methods. The development of clinical antiviral drugs from structural data for HIV protease has been a major success in structure based drug design. Techniques for virtual screening involve the ranking of the affinity of potential ligands for the target site on a protein. Two main alternatives have been developed: modeling of the target protein with a series of related ligand molecules, and docking molecules from a database to the target protein site. The computational speed and prediction accuracy will depend on the representation of the molecular structure and chemistry, the search or simulation algorithm, and the scoring function to rank the ligands. Moreover, the general challenges in modern computational drug design arise from the profusion of data, including whole genomes of DNA, protein structures, chemical libraries, affinity and pharmacological data. Therefore, software tools are being developed to manage and integrate diverse data, and extract and visualize meaningful relationships. Current areas of research include the development of searchable chemical databases, which requires new algorithms to represent molecules and search for structurally or chemically similar molecules, and the incorporation of machine learning techniques for data mining to improve the accuracy of predictions. Examples will be presented for the virtual screening of drugs that target HIV protease.

A new, non-SSRI mechanism of action for Prozac™. Unquestionably, Prozac™ was a landmark drug that redefined the treatment for depression, anxiety disorders, premenstrual syndrome and post-traumatic stress disorder as well as ushering in a wave of new therapies targeting specific serotonin (5-HT) reuptake inhibition (SSRI) [1]. In a recent manuscript (Psychopharmacology 2006, 186, 362-372), Pinna and co-workers report that fluoxetine and its N-desmethyl congener norfluoxetine stereospecifically and selectively increase brain neurosteroid content at doses that are inactive on 5-HT reuptake [2]. Pinna further states that the term ‘SSRI’ may be misleading in describing the pharmacological profile of fluoxetine and its congeners and suggests the term ‘selective brain steroidogenic stimulants’ (SBSSs) as a better descriptor [2]. Recently, it has been shown that neurosteroids, such as allopregnanolone (Allo), play critical roles in human brain pathophysiology such as anxiety disorders, premenstrual syndrome, posttraumatic stress disorder, postpartum depression and general depression. For the treatment of major depression, recent clinical trials have indicated that the pharmacological profiles of Prozac™ and related congeners correlate with an increase of Allo in the brain and cerebrospinal fluid. Importantly, Allo is a potent (nM affinity) positive allosteric modulator of gamma-aminobutyric acid (GABA) action at GABAA receptors, potentiating the intensity of GABAgated Cl- currents [3]. Thus, Prozac™ and related analogs increase Allo levels, and subsequent activation of GABAA receptors, at currently prescribed clinical doses which may not block 5-HT reuptake, but does effectively treat depression, premenstrual dysphoria and anxiety [2]. Pinna then explored this hypothesis in animal behavioral models of aggression and anxiety [2]. Their studies found that social isolation in mice is correlated with a down-regulation of Allo content and that pretreatment with Allo abolished aggressiveness as well as correcting the altered responses of GABAA receptors to specific agonists (pentobarbital) and antagonists (picrotoxin). Therefore, drugs capable of upregulating brain Allo content normalize behavioral abnormalities expressed in mice with a downregulation of brain Allo levels. Significantly, Pinna and co-workers discovered that the actions of fluoxetine and norfluoxetine on brain Allo content are stereoselective whereas 5-HT reuptake lacks stereospecificity. Prozac™ is a racemic compound. Pinna found that the EC50s of the (S)-enantiomers of fluoxetine and norfluoxetine to normalize brain Allo levels are 10- and 50-fold lower than those required to inhibit 5-HT reuptake [2]. Will this research herald new drug discovery campaigns aimed at identifying compounds capable of selectively increasing brain Allo levels under the moniker SBSSs? REFERENCES [1] Wong, D.T.; Bymaster, F.P.; Reid, L.R.; Mayle, D.A.; Krushinski, J.H.; Robertson, D.W. Norfluoxetine enetiomers as inhibitors of serotonin uptake in rat brain. Neuropsychopharmacology 1993, 8, 337-344. [2] Pinna, G.; Costa, E.; Guidotti, A. Fluoxetine and norfluoxetine stereospecifically and selectively increase brain neurosteroid content at doses that are inactive on 5-HT reuptake. Psychopharmaology 2006, 186, 362-372. [3] Lambert, J.J.; Belelli, D.; Peden, D.R.; Vardy, A.W.; Peters, J.A. Neurosteroid modualtion of GABBA receptors. Prog. Neurobiol. 2003, 71, 67-80.