Protein complexes – challenges and opportunities for drug discovery

Introduction
The new discipline of proteomics has begun to profoundly change the drug discovery process. Large databases and novel platform technologies now allow for accelerated and more systematic identification of targets, biomarkers and compounds. Nevertheless, important implications are still neglected, and overall, proteomic studies have delivered large amounts of data but few results that are relevant for commercial development. Above all, the high complexity of biomolecules and their interactions poses major analytical and conceptual challenges. Protein protein-interactions have emerged as being crucial for physiological function and potentially affecting pharmacological behaviour, questioning the classical definition of targets. But which technologies are most successful in their identification? Can they be affected e.g. by small molecule drugs? And if so, what are the benefits that can concretely be expected for the drug discovery process?

Re-defining drug targets
The traditional gene-based view on targets that still governs the industry predicts a maximum of 2000-3000 target genes. Consequently, discovery focuses on a limited set of genes and compounds are optimized for interaction with individual proteins. However, the high attrition rates observed when going from in vitro to living systems suggest that the physiological targets are indeed different. Evidence has accumulated over the last years that proteins exert their function as part of larger assemblies like protein complexes or so-called microdomains. Modern proteomic approaches have revealed the prevalence and complexity of such assemblies in biological systems: more than 50% of proteins form stable complexes with five or more partners as reported by studies in yeast. In a number of examples the multiple implications of the modular organization of the proteome have been demonstrated. Association of proteins not only alters biochemical and structural properties, but may regulate the primary function, provide links to intracellular pathways ensuring functional specificity, determine trafficking and targeting, influence protein half-life and stability, and even change the pharmacological profile. Moreover, protein complex composition is often specific for cellular compartments or individual cell types. In conclusion, physiological functions are carried by protein complexes and interaction networks, and these form the targets for potential pharmacological intervention. This new definition of targets also implies new drug discovery strategies.

Interfering with protein-protein interactions
The idea of developing compounds acting on protein complexes still faces strong scepticism. Two main arguments, both missing important biochemical and physiological facts, are

(a)“Why should one do it?” Although directing compounds against the primary target has been successful in a lot of cases, there is a long list of examples where this has failed. Among the most prominent examples are ion channels, which were regarded as a very promising target class several years ago. Meanwhile, many compounds acting on pore-forming alpha subunits have been developed, but few of them have been successful in the clinic due to lack of specificity and critical side effects. Ironically, many marketed drugs that have originally been developed against individual targets were later shown to owe their success to unanticipated selectivity for target (complex) subpopulations or effects on different targets. In addition, complexes offer a chance to develop drugs with improved properties as described below.

(b)“It is not possible.” It is still a common notion in drug development that small molecules cannot interfere efficiently with protein-protein interactions. This is largely based on misconception of how protein interactions work. Even in constitutive complexes, there is conformational flexibility of the protein interfaces. In addition, the allosteric nature of proteins implicates that ligands can induce structural changes at protein-protein interfaces even when their binding occurs at distant sites. Allosteric ligands can be as effective as classical competitive (ant)agonist, but their binding site is not sterically restricted. Such compounds are therefore promising candidates for drugs affecting protein complexes, and a growing number of examples is described in the literature. Of course, their targeted screening and development remains demanding. Modern technologies like high-content screening and cell-based assays open promising perspectives.

Resolving the challenge of complexity
It is estimated that up to 250.000 protein products are encoded by our genome, and even if only a small portion is expressed at relevant levels, the number of potential complexes and isoforms thereof is beyond imagination. Thus, their systematic identification is a major technical challenge. Two distinct strategies are currently used:

(a) Recombinant methods like yeast-two hybrid screens or co-purification analysis of tagged proteins offer the advantage of standardized procedures and high throughput. Both are based on expression of gene construct libraries in yeast, but even when applied to the proteome of this rather simple organism error rates are quite high. More than 50% of reported interactions are estimated to be false-positive, and the majority – especially those involving membrane proteins – remains undetected. In vitro binding assays like protein arrays use recombinant proteins and can also achieve very high throughput. However, they are restricted to soluble proteins or domains and prone to artefacts.

(b) Native source based approaches couple biochemical fractionation or isolation of protein assemblies with mass spectrometric identification. Preparation of source material can be tedious and sample amounts become a limiting factor. The biochemical method of choice is affinity-based purification of native protein complexes. Strong and selective enrichment of target protein(s) considerably reduces complexity and at the same time allows the detection of very low abundant proteins. High-affinity ligands like specific antibodies are required as well as optimization of conditions including suitable solubilisation buffers for isolating membrane proteins and adequate controls. This has so far prevented large-scale application of this approach. However, when carried out properly, it can deliver highly reliable and comprehensive results.

It is evident that we are still far from understanding the complex protein interaction networks in living organisms. Current studies deliver snapshots that often show little overlap and – a key problem – lack independent verification, for example by functional experiments. Targeted approaches like affinity-based purification from native tissue provide the most reliable results, that can be used for drug discovery in several ways:

1. Target validation and subunit composition
The search for novel targets often starts with genes identified in patients or disease models. Detailed information on tissue distribution, regulatory and functional mechanisms, cell-type specificity and molecular structure of the corresponding proteins is required for validation but often not readily accessible. A third of our genome is still functionally unassigned and about the same fraction is annotated on a preliminary basis. Thus, identification of associated proteins can provide direct access to molecular pathways and function and complement genetic knockout strategies that are often inconclusive due to lethality or compensatory effects. 

Furthermore, several proteins are only functional when they form heteromultimers. Associated subunits may be mandatory for proper folding, stability, trafficking or primary function. Prominent examples are found among G-protein coupled receptors and ion channels.

But even when interaction partners have been described, the subunit composition in the cell type or tissue of interest is often unknown. Identification of cell-type specific assemblies provides a basis for developing more specific drugs.

Co-targets and modulators
Associated proteins can be highly specific for a given target or confer regulatory mechanisms that are relevant under pathological conditions. Such co-targets offer the chance to develop compounds with new functional properties, like conditional modulators or blockers/mimics of specific regulatory input pathways. Resulting drugs may have improved selectivity and safety profiles. In addition, co-targets can offer alternative access to the development of drugs in cases where the original target is protected by competitors’ patents.

Structural design
Protein 3D structures elucidated by crystallography or nuclear magnetic resonance spectroscopy draw a rather static picture of proteins biased towards independently folding domains. Unstructured regions – that actually represent a major portion of proteins – have been largely neglected. Studies of protein complexes have shown that unordered domains play an important role in protein assembly and often adopt structures within complexes. Although it may be technically challenging, more meaningful and novel target structures may be obtained by analysis of native protein complexes.

Structure-based design and screening focuses on known ligand binding sites since much less is known about conformational dynamics and sites for modulatory input. Thus, allosteric compounds are rather found by chance. Proteomic analysis of target complexes can provide clues for regulatory domains or allosteric sites, for example by revealing determinants for functional interactions or protein modifications such as phosphorylations.

Screening assays
Cellular assays provide the opportunity to study the target protein in a cellular context and with multiple readouts. Main challenges are to reconstitute the physiological target phenotype and to find meaningful and readily detectable readout signals. One possibility is to select a cell line that is closely related to the native target cell and provides a suitable protein background, but this requires additional validation and is not always possible. Alternatively, established cell lines with known background can be stably transfected with the target gene. Expression levels, stability and functionality of the target may be improved by co-transfection of accessory subunits. In addition, native complexes provide insight into potential effector pathways that can be used as readouts. Ultimately, differential screening of cell lines expressing different target subunit compositions would provide a means for systematic development of complex-specific compounds.