The inhibition of protein–protein interactions using small molecules is a viable approach for the treatment of a range of pathological conditions that result from a malfunctioning of these interactions. Our strategy for the design of such agents involves the mimicry of side-chain residues on one face of the α-helix; these residues frequently play a key role in mediating protein–protein interactions. The first-generation terphenyl scaffold, with a 3,2′,2″-substitution pattern, is able to successfully mimic key helix residues and disrupt therapeutically relevant interactions, including the Bcl-XL–Bak and the p53–hDM2 (human double minute 2) interactions that are implicated in cancer. The second- and third-generation scaffolds have resulted in greater synthetic accessibility and more drug-like character in these molecules.
- helix mimetic
- medicinal chemistry
- protein–protein interaction
- rational design
The complex molecular machinery within a living cell operates through a web of interdependent pathways, many of which involve the interaction between two or more proteins. The proper functioning of an organism is dependent on the fidelity of these processes, and an aberration in the interaction pathway can lead to various diseased states. The association of a cytokine with its cell-surface receptor, for instance, is key to the regulation of signal transduction pathways. On the other hand, the aggregation of proteins to form insoluble fibres in the case of amyloidogenic diseases is undesirable and needs to be prevented. As a result, the understanding and modulation of protein–protein interactions for therapeutic applications has received considerable interest in recent years .
The most successful approach to targeting protein–protein interactions has been through the use of monoclonal antibodies . Although antibodies offer the advantage of having very high affinities for their target proteins, they are expensive to produce, have low oral bioavailability and undesired side effects .
An alternative approach involves the identification of small molecules that are capable of disrupting protein–protein interactions. However, this remains a formidable challenge [4,5], as small molecular scaffolds are often limited in their affinity and selectivity towards the shallow topological features and large surface areas that are characteristic of protein interfacial domains. The absence of a natural small-molecule binding partner, the inherent conformational flexibility of proteins and the lack of suitable binding assays provide additional challenges. Moreover, it is difficult to characterize the binding epitope as well as the protein partner to which an active small molecule binds. Traditional drug discovery methods, which, in recent years, have relied heavily on high-throughput screening of chemical libraries, have been less successful in these cases. The most likely reason for this is that classical drug molecules occupy a chemical space different from that needed to effectively disrupt natural affinity among proteins .
The intrinsic differences between the nature of interacting protein surfaces and enzyme active sites means that new strategies must be devised to target protein–protein interactions. In silico screening of many different compounds , for example, is increasingly being used to virtually identify lead molecules for the inhibition of protein–protein interactions. A new fragment-based lead discovery method based on the ‘tethering’  of molecular subunits to the target surface has been applied successfully to the inhibition of the interaction of interleukin-2, a key immunoregulatory cytokine, with its receptor, resulting in a lead compound with an IC50 of 60 nM . ‘Protein grafting’, which involves the insertion of a known binding epitope of a protein on to a small protein scaffold, has been shown to be an effective strategy in inhibiting the interaction of the KID (kinase-inducible domain) of CREB (cAMP-response-element-binding protein) with the KIX domain of the transcriptional co-activator protein, CBP (CREB-binding protein) . Another approach for targeting protein surfaces is through the use of large-surface-area scaffolds such as porphyrins, calixarenes and self-assembling G-quadruplexes [11–13].
We have reported an alternative design strategy for small molecules that are capable of mimicking short secondary-structural domains of proteins. Although protein–protein interactions bury a large surface area, a few residues on the interacting surfaces contribute a significant fraction of the binding energy . These regions are defined as ‘hotspots’, and their synthetic mimicry has been demonstrated to be a valid approach in the design of small-molecule inhibitors.
Pre-organized α-helix mimetics
Several years ago, we became interested in designing small-molecule mimics of α-helix structure and function, since a large number of key protein–protein interactions are mediated by α-helices. In designing such proteomimetic inhibitors, we reasoned that the central helical core, which is stabilized by backbone hydrogen-bonding, serves to project the amino acid side chains in a particular orientation. It should thus be possible to replace this cylindrical core by a synthetic scaffold capable of projecting functionality in a spatial orientation comparable with that of an α-helix. Early examples involved trisubstituted 3,2′,2″-terphenyl derivatives [15–19] in which the central aryl ring adopts a staggered conformation, effectively mimicking the position and projection of i, i+3 or i+4 and i+7 residues on one face of the helix. These residues are often key contributors to the recognition properties of the helix, and their successful mimicry offers an approach to inhibiting protein–protein interactions .
The effectiveness of this strategy was first shown by the design of mimics of the smMLCK (smooth muscle myosin light-chain kinase) that binds to calmodulin using an α-helical domain . A compound based on scaffold 1 (Figure 1a) with side chains that mimic Trp800, Thr803 and Val807 of smMLCK was shown to bind calmodulin and disrupt its interaction with other proteins that bind in the same region as smMLCK.
To assess the viability of the terphenyl scaffold in disrupting a therapeutically relevant protein–protein interaction, a series of derivatives was tested for its ability to inhibit the assembly of a key viral entry protein, gp (glycoprotein) 41 . In particular, gp41 and gp120 are proteins that mediate the fusion and entry of HIV-1 into human cells involving, in a key step, the assembly of a six-helix bundle of the C- and N-terminal domains of gp41. A derivative of terphenyl, 1a, was designed that incorporated leucine and isoleucine surrogates and was a structural mimetic of the α-helical heptad repeat sequences of the C-helical regions of gp41 (Figure 1a). The CD spectrum of a model system for gp41 showed that the terphenyl derivative was able to disrupt a pre-formed six-helix bundle as evident from a gradually decreasing α-helical signature peak at 222 nm. These results were corroborated further using an ELISA and a dye-transfer cell fusion assay on full-length and biologically relevant forms of gp41. Control molecules, where the terphenyl core was shortened to biphenyl or lacked the key hydrophobic side chains were not effective, suggesting an important role for all three of the side chains in the inhibition process.
Appropriately substituted derivatives of the terphenyl scaffold [16,18] were also tested for their ability to inhibit the Bcl-XL–Bak interaction, a key protein–protein complex that is up-regulated in cancer. Val74, Leu78 and Ile81 of the BH3 (Bcl-2 homology domain 3) of Bak, which, along with Ile85, make key contributions to the interaction, were effectively mimicked by derivatives of 1 (Figure 1a), as demonstrated in a fluorescence polarization assay. Scrambling the side chains or altering their hydrophobic character resulted in a lowering of affinity, suggesting a structural complementarity of the helix mimetic with the Bcl-XL surface. Significantly, the most active compound, 1b, showed selectivity for the Bcl-XL–Bak interaction over the related p53–hDM2 (human double minute 2) interaction, which is also mediated by an α-helical domain on p53. p53 is pro-apoptotic and its activity is suppressed on binding to hDM2; the overexpression of hDM2 in a number of cancers makes this interaction an important target for cancer therapeutics. A different terphenyl derivative, 1c (Figure 1a), was shown to disrupt the interaction of p53 with hDM2  selectively over the Bcl-XL–Bak interaction.
These examples demonstrate the versatility of the terphenyl scaffold in selectively targeting different protein–protein interactions as the side chains are varied. This is in analogy with α-helical domains in proteins, which have the same core structure, but assert their recognition properties through the use of different side chains. This early synthetic scaffold, however, suffers from several disadvantages, including a challenging synthesis and relatively high hydrophobicity. The latter can be overcome, to some extent, by replacing the benzene rings of the terphenyl with pyridines, as in the terpyridine scaffold .
To introduce additional contacts with residues that flank the hydrophobic face, the terphenyl scaffold was modified by replacing the middle benzene ring with a 1,6-disubstituted indane . The substituent at position 1 of the indane ring is able to mimic the corresponding i+3 residue of the helix, in addition to the i, i+4 and i+7 mimicry present in the terphenyl, and an overlay with a polyalanine α-helix showed a good match between them.
The terephthalamide scaffold, 2 (Figure 1b), was designed to address the demanding synthesis of the terphenyl and to introduce greater water solubility and drug-like character . Here, the benzene ring on either end of the terphenyl was replaced by a secondary or tertiary amide. An o-alkoxy substituent on the remaining aryl ring made hydrogen-bonding contact with the secondary amide, providing conformational rigidity. Molecule 2a has a lower relative hydrophobicity, as seen from its calculated log P (partition coefficient for n-octanol/water) value of 4.42 compared with the terphenyl value of 9.25, suggesting that this scaffold had improved ‘drug-like’ properties. Derivatives of 2 were able to compete with Bak for binding to Bcl-XL with an inhibition constant as low as 0.78±0.07 μM in a fluorescence polarization assay. HSQC (heteronuclear single-quantum coherence)–NMR experiments showed that this molecule, like the terphenyl, bound to the Bcl-XL surface in the same region as the helical domain of the Bak peptide. Assays using HEK-293 (human embryonic kidney) cells showed that terephthalamide 2a was able to intracellularly inhibit the association of Bcl-XL with Bax with an IC50 of 35 μM under the conditions of the experiment.
Improved aqueous solubility of the pyridyl group relative to the phenyl group was used in shorter bis-aryl motifs based on the pyridyl-pyridone scaffold 3 (Figure 1c) . This scaffold mimics the i, i+3 and i+4 residues in the helical domain of co-activator peptides that make important contacts with the oestrogen receptor. The interaction of the oestrogen receptor with an oestrogen molecule followed by co-activator peptides results in the activation of the transcriptional machinery, and its malfunction has been associated with various pathological conditions. An X-ray crystal structure of a derivative of 3 showed that, in the solid state, it adopts a conformation that should effectively mimic the required residues of the co-activator peptide GRIP1 (glutamate receptor-interacting protein 1). These molecules were tested in a fluorescence polarization assay and were found to inhibit the interaction of a model GRIP1 peptide and oestrogen receptor with inhibition constants in the low-micromolar range. Molecule 3a was shown to have very little affinity for the oestradiol-binding site on the oestrogen receptor, suggesting that it competes directly with the co-activator peptide for binding to the oestrogen receptor.
α-Helix mimetic foldamers
Instead of using a pre-organized scaffold such as the terphenyl, we have developed a related approach in which hydrogen-bonding propensities are built into the molecule that control its folding into a desired conformation. These molecules fall under the general category of ‘foldamers’  and offer a wide variety of synthetic approaches to structures that effectively disrupt protein–protein interactions . In this regard, oligomers based on β-peptides [26–28] and mixed α/β scaffolds  have been particularly successful.
The enaminone 4 (Figure 2a) is one such scaffold that, like the terphenyl, mimics the i, i+4 and i+7 residues of the α-helix . This molecule replaces the central aromatic ring of the first-generation terphenyl scaffold with a six-membered hydrogen-bonded ring and imparts it with greater synthetic accessibility. An X-ray crystal structure of 4a confirmed the presence of the intramolecular hydrogen bond, as well as the suitability of using the enaminone functionality as an aromatic ring isostere with a planar deviation of only ∼0.02 Å (1 Å=0.1 nm). In addition, variable-temperature NMR studies in CDCl3 as well as DMSO gave temperature coefficients ≤2.0 p.p.b./K, suggesting the existence of the intramolecular hydrogen bond in solution. A superimposition of 4a on a polyalanine α-helix showed good correspondence of the side chains to the i, i+4 and i+7 residues.
Most reports on synthetic mimicry of protein α-helices to date have focused on mimicking small structural domains, most commonly up to two to three turns of the helix. Longer α-helices are frequently found to play a critical structural and functional role in higher-order structures such as helix bundles, coiled coils and transmembrane domains of proteins. To mimic extended regions of an α-helix, we have designed a benzoylurea scaffold 5 (Figure 2b), that replaces the central aromatic ring of the terphenyl unit with another aromatic ring isostere, an acylurea motif . Ready synthesis and side-chain modification were achieved in the benzoylurea scaffold by sequential coupling between an appropriate secondary amide and an isocyanate. Using this strategy, significantly elongated helix mimetics containing four to five benzoylurea subunits have been obtained and can span lengths of up to 37.1 Å, corresponding to a helix of approx. 30 residues and seven turns. The modular synthesis allows for the incorporation of different amino acid side-chain surrogates and the desired recognition properties in these potentially membrane-spanning foldamers can be ‘dialed in’ with a reasonable synthetic effort. Examples of mimicry of such extended regions in totally synthetic oligomers are relatively rare .
Another foldamer scaffold based on trispyridylamide 6 (Figure 2c) uses an amide functionality to connect the pyridine rings of the terpyridine scaffold discussed above . The repeat unit is a pyridyl amino acid, and conformational control is achieved through bifurcated hydrogen-bonding of the amide proton to the pyridine nitrogen and o-alkoxy side chains on the pyridine ring. In this orientation, the o-alkoxy side chains mimic the projection of the i, i+4 and i+7 residues of an α-helix. The presence of the intramolecular bifurcated hydrogen bond, even in a polar solvent such as DMSO, is supported by variable-temperature NMR studies. An X-ray crystal structure further confirmed that the molecule indeed adopted the proposed conformation in the solid state. Derivatives of 6a with appropriately substituted o-alkoxy side chains were tested for their ability to inhibit the Bcl-XL–Bak interaction. A fluorescence polarization assay showed that these α-helix mimetics could successfully displace the BH3 domain of the Bak peptide from the Bcl-XL–Bak BH3 complex, with Ki values in the low-micromolar range, thereby validating the design. The same scaffold 6, with carboxylate side chains as mimics of aspartate residues can interact with the surface of a growing crystal, resulting in expression of faces and morphologies not seen in the control crystals . In doing so, molecule 6b mimics the action of acidic proteins in biomineralization , the process that gives form to much of the natural world.
The rational design of small organic molecules that mimic protein secondary structures has been shown to be successful in the inhibition of protein–protein interactions. As more molecular and structural details on interacting protein surfaces become available, a better understanding of the design principles should emerge. This strategy holds great promise in the development of potential therapeutics, and we expect to see a significant growth in this currently underexplored area of medicinal chemistry.
We thank the National Institutes of Health for financial support of this work (grant number GM69850).
2nd International Meeting on Molecular Perspectives on Protein–Protein Interactions: Independent meeting held at Hotel Croatia, Dubrovnic, Croatia, 27 June–1 July 2008. Organized and Edited by Colin Kleanthous (York, U.K.), Jacob Piehler (Frankfurt, Germany) and Gideon Schreiber (Weizmann Institute, Rehovot, Israel).
Abbreviations: BH3, Bcl-2 homology domain 3; CREB, cAMP-response-element-binding protein; gp, glycoprotein; GRIP1, glutamate receptor-interacting protein 1; hDM2, human double minute 2; smMLCK, smooth muscle myosin light-chain kinase
- © The Authors Journal compilation © 2008 Biochemical Society