Keywords

1 PPI Targets in Cancer

The modern understanding of cancer biology is that tumour cells have acquired the characteristic ‘hallmarks of cancer’ through gain-of-function mutations to oncogenes and loss-of-function of tumour suppressor genes [1]. Detailed examination of signalling pathways associated with hallmark traits has allowed the identification of numerous potential targets for intervention by therapeutic agents, in many cases with small-molecule drugs. A number of pathways investigated thus far rely on signalling via nodes regulated by protein–protein interactions (PPIs). In particular, PPIs regulating the apoptotic response of a cell have been characterised, providing promising areas for drug discovery, for example, between members of the B-cell lymphoma 2 (Bcl-2) family, the mouse double minute 2 (MDM2)–p53 interaction, and the interaction of second mitochondria-derived activator of caspases protein with X-linked inhibitor of apoptosis protein (Smac-XIAP). This chapter will describe each of these PPI targets and the discovery of small-molecule PPI inhibitor drugs that have entered clinical trials.

1.1 PPI Chemotherapy Drugs

Historically, extensive screening of natural products was used to identify compounds with cytotoxic activity for use in chemotherapy regimens. For a number of compounds investigated, the molecular mechanism of action was identified to be the disruption of a PPI (Table 1).

Table 1 Natural product derived PPI drugs
Full size table

The regulation of microtubule cycling by the polymerisation and depolymerisation of α,β-tubulin is essential for the organisation of DNA during the M-phase of the cell cycle, and is targeted by a number of natural products from diverse sources. Isolated from plant sources, the vinca alkaloids e.g. vinblastine (1), and the taxoids e.g. taxol (2) were initially thought to share the same mechanism of action. However, detailed studies indicated that vinblastine (1) acts by binding into a hydrophobic groove at the interface of the α- and β-tubulin subunits at the ends of microtubules, and stabilises the microtubule at low concentrations, but induces depolymerisation at higher concentrations [2]. In contrast, taxol (2) binds to the β-tubulin subunit close to the GTP-binding site [3,4,5]. A conformational change is induced by drug binding that results in an unusually strong interaction between subunits and consequently microtubules grow in an uncontrolled manner. Isolated from marine sponges, the natural product Halichondrin B (3) and the simplified synthetic derivative Erubilin (4) also bind to microtubules close to the vinca binding site, and overlapping with the GTP-binding site, suppressing the growth of microtubules [6, 7]. Isolated from the myxobacteria Sorangium cellulosum, Epothilones e.g. 5 have a similar mechanism of action as the taxoids, and promote microtubule growth.

The semi-synthetic derivative of rapamycin, temsirolimus (6) inhibits the kinase activity of the mammalian target of rapamycin protein (mTOR) by forming a protein–protein complex with the FK506 binding protein 12 (FKBP12) [8].

1.2 Druggability of PPIs

Despite the serendipitous discovery of PPIs as cancer therapies, until recently the established view from medicinal chemists has been that PPI targets are significantly less ‘druggable’ than enzyme or receptor targets. The reasoning for this is based on the high surface area for PPI targets (1,500–4,660 Å2) compared with other drug–protein interfaces (1,200–2,000 Å2), and their relatively high binding affinities (ΔG = −7.0 to −17.2 kcal/mol) [9]. Other analyses suggested that PPIs are more circular, flatter, less polar, and present fewer opportunities for H-bonding than enzyme or receptor drug targets [10]. The negative outlook from these global analyses of PPIs has been contradicted by detailed investigations of PPI binding sites, including alanine-scanning mutagenesis, that have revealed that many PPIs include small regions of high binding affinity called ‘hot-spots’ that can have attractive features for small molecule binding [11, 12]. Hot-spots may include the binding sites for large amino acid side chains on one protein that bind to small pockets on the partner protein, and may be circled by polar residues [13]. ‘Hot segments’ are larger continuous epitopes that contribute the majority of the binding energy of a PPI [14]. The most tractable of these for drug discovery are the interactions between a continuous peptide sequence of one peptide that interacts with a defined domain of the partner protein. The interfaces between protein and α-helical epitopes have demonstrated good tractability as drug targets, as exemplified by the development of Bcl-2 family and MDM2-p53 PPI inhibitors (vide infra).

Targeting PPIs with small-molecules may have some advantages as they represent an under exploited class within the ‘druggable genome’ and may present opportunities to inhibit cellular pathways that would be difficult to approach through other target classes [15]. PPIs may offer better opportunities for selectivity in comparison to competitive inhibition of enzymes at their substrate of co-factor binding sites due to the size and structural diversity of the interactions targeted. Unfortunately, the proposal that PPI inhibitors will offer a lower likelihood of developing resistance has not been borne out in clinical experience, especially in the case of MDM2-p53 inhibitors [16, 17].

2 Bcl Family Inhibitors

2.1 Biology

The response of cells to a variety of signals resulting in programmed cell death or apoptosis is controlled by the PPIs of a family of proteins related to Bcl-2 [18, 19]. Within the family, proteins can be grouped according to their role (Fig. 1). The BH3-only proteins [e.g. Bcl-2-associated death promoter (Bad) and p53 upregulated modulator of apoptosis (PUMA)] act as initiators of apoptosis in response to cytokine deprivation, intracellular damage or oncogenic signalling, whereas truncated BH3 interacting-domain death agonist (t-Bid) is activated by the death receptor and signalling via caspase 8. Activation of BH3 proteins induces both the negative regulation of the pro-survival Bcl-2 proteins, and the activation of the pro-apoptotic effector proteins: either Bcl-2-associated protein (Bax) or Bcl-2-homologous antagonist killer (Bak), which disrupt the outer membrane of mitochondria and initiate the apoptotic cascade.

Fig. 1
figure 1figure 1

Apoptotic signalling via the Bcl-family proteins

Full size image

Given the pivotal role of the Bcl-2 proteins in regulating cell death, it is not surprising that dysregulation of the pathway is strongly linked to the development of cancer. Overexpression of pro-survival proteins is linked to tumours, e.g. Bcl-2 overexpression is found in 90% of follicular centre B cell lymphomas [20], and MCL-1 or BCL2L1 is amplified in a variety of tumours [21]. Similarly, deletion or downregulation of BH3 initiator genes, BIM and PUMA has been observed in tumours [22,23,24]. Abnormalities in the Bcl-2 pathway are also associated with resistance to chemotherapy and radiotherapy [25].

The prospect of activating apoptosis in tumour cells where the Bcl-2 pathway is defective provides an attractive approach to new cancer therapies. Compounds have been sought that mimic the action of pro-apoptotic BH3 proteins by binding to Bcl-2 family members to block their pro-survival action and so induce apoptosis directly.

2.2 Structural Biology and Peptides

NMR and X-ray studies with the B-cell lymphoma-extra large (Bcl-xL) protein provided a structural understanding of the domains involved in the anti- and pro-apoptotic activity of the protein [26]. The tBH1, BH2 and BH3 domains of Bcl-xL form a hydrophobic cleft, which is the site for interactions with the pro-death proteins, Bax and Bak. The NMR solution structure of Bcl-xL bound to a Bak BH3 domain peptide shows the Bak peptide forms an amphipathic α-helix, which binds via the hydrophobic side chains of the Val74, Leu78, Ile81 and Ile85 residues into a hydrophobic groove on the surface of Bcl-xL (Fig. 2) [27]. Investigations with mutant Bak peptides showed that the peptide binds as an amphipathic α-helix. Hydrophobic interactions between the side-chains of Val574, Leu578, Ile581 and Ile585 of Bak and a hydrophobic groove formed by Tyr101, Leu108, Val126 and Phe146 on the surface of Bcl-xL contribute the majority of the interaction affinity. In addition, Asp583 of the Bak peptide and Arg139 of Bcl-xL form an important electrostatic interaction. The identification of key interactions across a short peptide epitope encouraged the development of small molecule inhibitors of the interaction.

Fig. 2
figure 2figure 2

NMR solution structure of Bcl-xL (green) bound to a peptide of residues 572–587 of Bak (red) (pdb 1BXL) [27]. (a) Ribbons; (b) Bcl-xL surface (green) and Bak (red with grey surface) showing the key sidechain interactions

Full size image

2.3 Small Molecule Bcl Family Inhibitors

α-Helix mimics based on a terephthalamide scaffold have been designed to mimic the side chain interactions of the Bak peptide as Bcl-xL inhibitors. Compound 7 inhibits the BH3 domain of Bak binding to Bcl-xL with a similar affinity to the peptide (Ki = 0.78 μM) (Fig. 3) [28]. Such compounds offer additional validation of the tractability of the Bcl-xL PPI as a target for small molecule inhibitors. However, the poor physicochemical properties of the compound (e.g. cLogP 7.4) prevent their optimisation as drugs.

Fig. 3
figure 3figure 3

Terephthalamide α-helix mimic (7)

Full size image

Screening of a library of natural products using NMR and fluorescence polarisation (FP) assays identified Gossypol (8) and Purpurogallin (9) as competitive inhibitors of the binding of a labelled BH3 peptide to Bcl-xL (Ki = 0.3 and 2.2 μM, respectively) (Fig. 4). Binding to Bcl-xL BH3 binding pocket was confirmed by NMR. Gossypol induces multiple cellular effects, some of which can be attributed to it acting as a BH3 mimetic [29]. Despite known liabilities of undesirable redox chemistry and protein reactivity for polyphenolic compounds [30], Gossypol and (R)-gossypol (AT-101) have been evaluated in clinical trials, but with limited efficacy observed in breast and prostate cancer [31,32,33].

Fig. 4
figure 4figure 4

Bcl-2 inhibitors Gossypol (8), Purpurogallin (9), and TW-37 (10)

Full size image

A structure based approach was used to design a series of Bcl-2 inhibitors based on the docked structure of Gossypol (8) bound to Bcl-2 [34]. The most potent compound TW-37 (10) retains the extensive hydrogen-bonding of one triphenolic moiety combined with a larger hydrophobic group. The second half of 8 is replaced with a hydrophobic amide that allows more extensive contacts with the BH3 binding pocket resulting in improved potency. 10 inhibits Bcl-2, Bcl-xL, and induced myeloid leukaemia cell differentiation protein (Mcl-1) interactions (Ki = 0.29, 1.1, and 0.26 μM, respectively) and induces apoptosis in the PC3 prostate cancer cell line (GI50 = 200 nM).

Fragment based screening offers an alternative to the optimisation of relatively high molecular weight hits from screening library compounds or natural products. Bcl-2 family inhibitors have been discovered and optimised using “SAR by NMR”. This is a two-step method, in which the first step identifies low molecular weight fragments by observing chemical shift changes in the 15N- or 1H–HSQC NMR spectrum of the protein on binding of the ligand. In the second step, additional fragments occupying different binding sites close to the first fragment hits are identified. NMR or X-ray crystallography is used to determine binding site and orientation of both fragments prior to the design of a suitable linking group. The linked fragments should result in a higher affinity ligand that accesses both binding sites [35].

Fragments binding to two subpockets of the BH3 interacting-domain death agonist (Bid) BH3 domain were identified in an NMR screen [36]. The interaction sites of the fragments were mapped to specific amino acids on the Bid surface using additional NMR experiments and docking used to determine the likely binding modes. A small library of sulfonamides was prepared to evaluate linking of the fragments together, giving rise to sulfonamide 11 (Fig. 5). Binding of 11 to the Bid BH3 domain was confirmed by NMR. Compound 11 was able to inhibit t-Bid induced Smac release in a mitochondrial assay and t-Bid induced apoptosis in transformed HeLa cells. Although further developments to this series have not been reported, this study confirmed the tractability of BH3 domains for NMR based fragment screening and the utility of the fragment linking approach.

Fig. 5
figure 5figure 5

Linking of Bid BH3 fragment hits

Full size image

Another application of the SAR by NMR method gave rise to fragment hits for Bcl-xL that bound with millimolar potency. Modification of the fragment bound to Site 1 to an acylsulfonamide allowed linking to the second fragment site and maintained the acidic hydrogen (Fig. 6). Parallel synthesis identified compound 12 (Ki = 36 nM), incorporating the 3-nitro-4-(2-phenylthioethyl)amino phenol group that both acts as a linker and occupies the second fragment site. Structure-guided optimisation of 12 allowed the identification of modifications to reduce binding to human serum albumin (HSA) whilst maintaining affinity for Bcl-xL. The replacement of one phenyl ring with a piperazine ring reduced binding to HSA and allowed the introduction of a biphenyl group, as seen in compound 13, binding deep into a hydrophobic cleft on Bcl-xL that is not seen in the peptide bound structure. The addition of a basic group to the thioethylaminoethyl group further reduced HSA binding giving ABT-737 (13) [37, 38].

Fig. 6
figure 6figure 6

(a) Linking of Bcl-xL bound fragments and optimisation by linking and structure-based design; (b) solvent accessible surface of the NMR structure of Bcl-xL with 12 (green) bound (1YSI), overlaid with fragments (magenta) from 1YSG; (c) solvent accessible surface of the NMR structure of Bcl-xL with 13 (green) bound (2O2N); (d) solvent accessible surface of the X-ray structure of Bcl-xL with 14 (green) bound (4LVT); (e) chemical structure of ABT-737 (10)

Full size image

ABT-737 (14) binds potently (Ki ≤ 1 nM) to Bcl-xL, Bcl-2, and Bcl-2-like protein 2 (Bcl-w) proteins, but less well to family members with lower homology, e.g. B-cell lymphoma/leukaemia 11B protein (Bcl-B), Mcl-1 and Bcl-2-related protein A1 (Ki = 460 nM and >1,000 nM, respectively). In cells, 14 acts in a similar manner to an apoptosis sensitising Bad-BH3 peptide by binding to and inhibiting anti-apoptotic Bcl-2 proteins thus removing their protection from apoptosis. 14 does not directly cause cytochrome c release or directly activate the pro-apoptotic proteins Bak and Bax. 14 showed synergism in combination with chemotherapeutics (2–20-fold) and radiation (2–4-fold). 14 induced tumour regression in Bcl-2 overexpressing small cell lung cancer (SCLC) xenografts (100 mg/kg/d ip) and showed improved activity in combination with etoposide [39].

Promising results with ABT-737 prompted further rounds of structure-guided optimisation resulting in Navitoclax (ABT-263, 15) [40]. In this case modifications were designed that did not disrupt the core binding motif and aimed to optimise aqueous solubility, cellular activity, and ultimately oral bioavailability. Modulation of the acidity of the sulfonamide group by variation of the 3-nitro group enhanced oral bioavailability, and switching the basic dimethylamino group to a morpholine reduced metabolism and improved oral exposure. A 3-trifluoromethylsulfone group proved to be the optimal replacement for the nitro group. Modifications to the 4-chlorobiphenyl moiety gave improved metabolic stability and cellular activity when the internal phenyl ring was replaced with a cyclohexenyl or gem-dimethylcyclohexenyl group. ABT-263 (15) has a high molecular weight (974 Da) and falls outside of other commonly used drug-like parameters, indicating the difficulty of addressing relatively long lipophilic binding sites. However, 15 shows cellular activity including mechanism-based cytotoxicity in Bcl-2 or Bcl-xL dependent models, and has acceptable oral bioavailability and PK properties [41]. In vivo efficacy with complete regressions was observed in SCLC and acute lymphoblastic leukaemia (ALL) based xenografts (100 mg/kg qd p.o.). Reversible thrombocytopenia was noted as a mechanism related toxicity.

A selective Bcl-2 inhibitor was designed with the aim of avoiding the thrombocytopenia observed with 15, attributed to the inhibition of Bcl-xL, and so improving the therapeutic window [42]. In order to separate Bcl-2 binding from Bcl-xL, a series of compounds was prepared starting from 15 with the removal or replacement of key binding moieties. Removal of the thiophenyl group of 15 (e.g. 16) allowed the key hydrophobic interaction to be achieved from a different vector with an azaindole group, and optimisation of the gem-dimethylcyclohexenyl group resulted in the discovery of venetoclax (ABT-191, 17) (Fig. 7). Venetoclax (17) is a potent Bcl-2 inhibitor (Ki < 0.01 nM) with good selectivity over Bcl-xL and Bcl-w (Ki = 48 and 245 nM, respectively). Selective cellular activity was seen for 17 in Bcl-2 dependent lines but not in Bcl-xL lines. In vivo antitumour activity was demonstrated for 17 in a selection of haematological tumour xenografts on a single oral dose. Ex vivo studies with human platelets showed that 17 was significantly less toxic than 15.

Fig. 7
figure 7figure 7

Design evolution of ABT-191 (17) from ABT-263 (15)

Full size image

Obatoclax (18) was identified in an FP assay measuring the displacement of a BH3 peptide from Bcl-2 family proteins as an inhibitor Bcl-2, Bcl-xL, Mcl-1, Bcl-w, A1, and Bcl-B (Ki ≈ 1–7 μM) (Fig. 8) [43]. Docking to the Bcl-2 NMR structure predicted that 18 binds to the BH3 binding groove in a hydrophobic pocket. The hydrophobicity of 18 (cLogP >4) was expected to favour membrane localisation, so studies with isolated mitochondria were used to demonstrate disruption of Mcl-1-Bak interactions. In whole cells, treatment with 18 initiates apoptosis, and is able to overcome Mcl-1 overexpression induced resistance to the Bcl-2 inhibitor ABT-737 (14) and bortezomib. Single agent antitumour activity was demonstrated in SCID mice bearing human C33A cervical carcinoma tumors with IV dosing at 0.25 and 0.5 mg/kg. A Phase I dose finding study with obatoclax in various tumour types by IV infusion established the tolerability and pharmacokinetics of the drug with some indication of clinical activity [44].

Fig. 8
figure 8figure 8

Obatoclax (18) and bendamustine (19)

Full size image

2.4 Clinical Development of Bcl-2 Inhibitors

Three Bcl-2 family inhibitors have progressed to clinical trials. Given the central role Bcl-2 family proteins play in the development of many lymphomas, this has been a key focus of clinical studies. A Phase I study with the pan Bcl-family inhibitor Navitoclax (15) in relapsed or refractory lymphoid malignancies established the optimum dosing regimen, pharmacokinetics, pharmacodynamics and tolerability of the drug and demonstrated some efficacy [45]. Chronic lymphocytic leukaemia (CLL) is characterised by over-expression of Bcl-2, and a Phase I study in CLL with 15 demonstrated substantial efficacy [46]. In SCLC and other solid tumours, a Phase I clinical trial with 15, under intermittent or continuous oral dosing schedules, established a maximum tolerated dose and demonstrated the safety of the drug and antitumour activity was observed [47]. A Phase IIa single-agent trial of 15 in SCLC also demonstrated safety and biomarker modulation, however, efficacy was limited [48]. Clinical studies with 15 in solid tumours are ongoing as a single agent and in combination with other therapies.

Despite promising results with 15, the Bcl-xL induced thrombocytopenia observed was dose limiting and so attention switched to the Bcl-2 selective inhibitor venetoclax (17) [49]. A single agent Phase I study of 17 in CLL and small lymphocytic leukaemia (SLL) showed an 84% response rate in CLL patients without significant reports of thrombocytopenia [50]. A Phase II study of 17 in ALL demonstrated activity in this indication, particularly in patents with a poor prognosis [51]. A combination of 17 with the anti-CD20 antibody rituximab was evaluated to determine the maximum tolerated dose and safety profile, producing an overall response rate of 88% with good tolerability [52]. A Phase 1b combination study of 17 with bendamustine (19) in relapsed non-Hodgkin lymphoma established the safety of the combination, and showed good evidence of efficacy [53]. A Phase 2 study of 17 in relapsed or refractory CLL patients with a chromosome 17p deletion showed an 80% response rate, demonstrating the effectiveness single agent 17 in a patient group that responds poorly to immunotherapy [54]. On the basis of these results venetoclax (17) received accelerated clinical approval in the US and Europe.

In a phase I trial with obatoclax (18) in advanced CLL acceptable safety, and phamacodynamic responses (Bax and Bak elevation) were observed. Modest single agent activity was seen with 1/26 patients achieving a partial response [55].

3 MDM2-p53

3.1 Biology

The p53 tumour suppressor acts as a central node in the cellular response to stress and DNA-damage [56]. Activation of p53 produces an active tetramer that can transcribe around 150 genes, giving rise to a number of cellular effects including initiating apoptosis, cell-cycle arrest, DNA-repair, senescence or altered metabolism [57]. MDM2, itself a product of p53 mediated transcription, acts as a negative repressor of p53 by binding to and occluding the p53 transactivation domain and by acting as an E3-ubiquitin ligase, targeting the complex for proteasomal degradation [58, 59]. The MDM2–p53 interaction forms an autoregulatory feedback loop, ensuring that cellular activation of p53 is transient.

Amplification of the MDM2 gene and overexpression of MDM2 protein occurs in around 10% of common cancers. The highest incidence is found in hepatocellular carcinoma (44%), lung (15%), sarcomas and osteosarcomas (28%), and Hodgkin disease (67%) [60]. MDM2 overexpression overrides p53 activation and inactivates its tumour suppressor function, thus rendering the cells effectively p53 null.

Inhibitors of the MDM2-p53 PPI have been sought as they should act to reactivate p53-mediated cell death by removing MDM2 mediated repression of transcription. Tumour selectivity is predicted to result from the removal of an MDM2 block on pre-existing DNA-damage or oncogenic activation signals, thus allowing p53-mediated activation of apoptosis.

3.2 Structural Biology and Peptides

The NMR structure of MDM2 reveals that the protein adopts multiple conformations [61], with residues 16–24 of the N-terminal region forming a flexible lid that closes over the p53 binding site [62]. p53 easily displaces the lid to form a more stable complex [63]. The X-ray structure of a p53 peptide and MDM2 interaction shows the surface MDM2 forms a hydrophobic cleft into which the p53 peptide binds as an amphipathic α-helix (Fig. 5a) [64]. Epitope scanning determined that three p53 residues, Phe19, Trp23 and Leu26, form the principal interactions with MDM2, secured by two intermolecular hydrogen-bonds, between the backbone of Phe19 of p53 and the Gln72 side chain of MDM2, and between the p53 Trp23 indole NH and the backbone carbonyl of Leu54 from MDM2 (Fig. 9a). Potent inhibitors of the MDM2–p53 interaction were discovered from synthetic peptide libraries, including the AP peptide (IC50 = 5 nM) that incorporates 6-chlorotryptophan, and phosphonomethylphenylalanine residues [65]. The X-ray structure of the AP peptide bound to MDM2 (Fig. 9b) shows the 6-chlorotryptophan group occupies additional space below the Trp23 site, and makes additional van der Waals contacts with Phe86 and Ile99 of MDM2.

Fig. 9
figure 9figure 9

X-ray structure of MDM2 (ribbon – green, surface – light green) bound to (a) residues 15–29 of p53 showing the key hydrophobic sidechains (red) (pdb 1YCR), (b) AP peptide (magenta) (pdb: 2GV2) (c) SAH-p53-8 (purple) (pdb: 3V3B)

Full size image

Hydrocarbon stapled α-helical peptides e.g. SAH-p53-8, Ac-QSQQTF*NLWRLL*QN-NH2 (* indicates the position of the hydrocarbon staple), have shown excellent potency as MDM2-p53 inhibitors [66, 67]. The X-ray crystal structure of SAH-p53-8 bound to MDM2 shows that the key binding residues Phe19, Trp23 and Leu26 form part of a short α-helix that binds to MDM2 (Fig. 9c) [68]. SAH-p53-8 induces p53 mediated cytotoxicity in MDM2 overexpressing cells. Similarly, the stapled peptide ATSP-7041, Ac-LTP*EYWAQAba*SAA-NH2, is a potent dual inhibitor of both MDM2-p53 (Ki = 0.9 nM) and MDMX-p53 (Ki = 6.8 nM), which has been optimised for solubility, and cellular potency [69]. The X-ray structure of ATSP-7041 bound to MDMX shows that the hydrocarbon staple makes significant interactions with the protein and so contributes to binding affinity. ATSP-7041 showed activity in vivo in MDM2 and MDMX overexpressing human tumour xenografts. Other peptide derived MDM2-p53 inhibitors using alternative backbones have been investigated with varying success including reteroinverso d-peptides [70, 71], β-peptides [72,73,74], di-proline β-hairpins [75, 76] and peptoids [77].

3.3 Hit Discovery and Optimisation

A variety of peptide mimetic approaches have been employed, using differing structures to array side chains in the appropriate orientation to recapitulate the peptide interaction with the MDM2 binding cleft (Fig. 10) [78]. Terphenyls e.g. 20 demonstrated the validity of this approach [79]. α-Helix mimics based on an oligoamide scaffold (21 and 22) were weak MDM2-p53 inhibitors [80, 81]. The α-1,4 linked oligogalactose 23, bearing groups corresponding to the important p53 side chains from alternate C-6 and C-3 hydroxyl groups, showed weak MDM3-p53 inhibitory activity [82]. Although these approaches offer interesting insights into the binding requirements for MDM2, the compounds’ high molecular weights and poor drug-like properties prevent their development.

Fig. 10
figure 10figure 10

α-Helix mimetic MDM2-p53 inhibitors

Full size image

The Nutlins are cis-imidazoline containing compounds, which were identified by HTS as inhibitors of MDM2-p53 with IC50 values in the 100–300 nM range (e.g. 24, IC50 = 90 nM) [83]. The series yielded the first reported X-ray and NMR structures of small molecules in complex with MDM2, showing that the heterocyclic scaffold arrays its substituents to mimic the side-chains of an α-helical peptide (Fig. 11a, pdb 4J3E) [84]. Structure-guided optimisation enabled improvements to stability at the imidazoline heterocycle by the introduction of two cis-methyl groups and by modification of the metabolically labile 4-methoxy group to a tert-butyl group. Additional potency and favourable PK properties were gained by the addition of a hydrophilic sulfone chain to the piperazine moiety, giving rise to the clinical candidate RG-7112 (25, IC50 = 18 nM). The X-ray co-crystal structure of MDM2 with 25 (Fig. 11b) shows it has a similar binding mode to Nutlin-3a, with the methylsulfone side chain extending into solvent.

Fig. 11
figure 11figure 11

X-ray structure of MDM2 (electrostatic surface) bound to (a) Nutlin 3a 24 (green) (pdb 4J3E), (b) RG7112 25 (pink) (pdb: 4IPF)

Full size image

Modification of the Nutlin scaffold to a dihydroimidazothiazole resulted in the identification of a series of potent MDM2 inhibitors, such as 26 (IC50 = 90 nM). Addition of the 6-methyl group is sufficient to stabilise the imidazoline ring to oxidation. The bicyclic scaffold orientates the 3-isopropyl group into the MDM2 Phe19 pocket, and addition of an amide at the 4-position gave additional potency [85, 86]. The X-ray structure of 26 bound to MDM2 shows the imidazole ring and chlorophenyl groups occupying similar space to the Nutlin series (Fig. 12a). The thiazole and isopropyl groups fill space occupied by the alkoxyphenyl group of 24, and the amide side-chain occupies a surface groove formed by the movement of a flexible loop. Interactions in the Trp23 binding pocket were optimised by the addition of a 3-fluoro-4-chlorophenyl group, and the amide moiety was elaboration with a 4,7-diazaspiro[2,5]octane group giving DS-5272 (27, IC50 = 24 nM, Fig. 12b) [87]. DS-5272 showed good cellular activity, solubility, microsomal stability and acceptable in vivo PK properties. Anti-tumour efficacy was demonstrated in an MDM2-amplified mouse xenograft model (50 mg/kg/qdx4 p.o.). The authors suggest that DS-5272 is a possible candidate for clinical development, but the structure of the clinical candidate DS-3032b has not been disclosed.

Fig. 12
figure 12figure 12

(a) X-ray structure of MDM2 (electrostatic surface) bound to dihydroimidazothiazole 26 (green) (pdb 3VZV), (b) Design evolution of DS-5272 (27) from 26

Full size image

A series of spirooxindoles, e.g. 28 (Ki = 86 nM), were identified initially by substructure searching for oxindole containing natural products designed to mimic the tryptophan of p53 (Fig. 7) [88]. Structure-based elaboration of the amide group and addition of a 5-fluoro group resulted in improved potency, e.g. 29 (IC50 = 28 nM) (Fig. 13) [89, 90]. The X-ray co-crystal structure of 29 bound to MDM2 shows the oxindole ring occupying the Trp pocket, with hydrogen bonds between the oxindole NH and the backbone carbonyl of Val54, and His96 NH and the amide side-chain carbonyl group, the 3-chloro-2-fluorophenyl group fills the Leu pocket, and the t-butyl group is in the Phe pocket (Fig. 14). The diol MI-147 (31, IC50 = 16 nM) showed cellular activation of p53 and in vivo activity in an MDM2 amplified tumour xenograft. Importantly, spirooxindoles e.g. 30 were found to epimerise under oxidative conditions to diastereomeric mixtures and the most stable diastereoisomer proved to have improved potency and in vivo activity [91, 92].

Fig. 13
figure 13figure 13

Design evolution of SAR405838 (32) from 28

Full size image
Fig. 14
figure 14figure 14

X-ray structure of MDM2 (electrostatic surface) bound to oxindole 29 (green) (pdb 3LBL)

Full size image

SAR405838 (32, Ki = 0.88 nM) was selected from a series of amides as it showed improved chemical and metabolic stability, in addition to potent MDM2 binding affinity, selectivity, and growth inhibitory activity [93] (Fig. 13). Spirooxindole 32 induced complete tumour regression in SJSA-1 mouse xenografts with a single oral dose (100 mg/kg) and on repeat dosing. The in vivo efficacy of 32 was also demonstrated in a multiple tumour types without MDM2 amplification. Consequently, 32 was selected as a clinical candidate from the spirooxindole series.

Further structure-guided optimisation of the spirooxindole series aimed to remove instability due to epimerisation [94]. Replacement of the 2′-neopentyl group of 32 with a cyclohexyl ring removed one chiral centre and maintained interactions in the Phe19 pocket of MDM2. Introduction of a carboxylic acid to the amide side-chain improved cell-free potency with the bicyclo[2.2.2]octane moiety as a phenyl ring replacement conferring improved PK characteristics. The alkylation of the pyrrolidine nitrogen was necessary to improve the cellular PD response of the compound. The resulting compound APG115 (33) showed anti-tumour efficacy in a SJSA-1 mouse xenograft model (100 mg/kg/qd p.o.) and has entered phase I clinical trials (Fig. 15).

Fig. 15
figure 15figure 15

Design of APG115 (33) from 32

Full size image

The spirocyclic amide scaffold of spirooxindole 32 was simplified to a cyanopyrrolidine core allowing a “cis” configuration of aromatic groups e.g. 34 (IC50 = 196 nM) [95]. The X-ray co-crystal structure of 34 bound to MDM2 shows the 4-chlorophenyl ring occupying the Trp pocket, with a hydrogen bond between His96 NH and the amide side-chain carbonyl group, the 3-chlorophenyl group fills the Leu pocket, with the cyano group projecting into solvent and the t-butyl group occupies the Phe pocket (Fig. 16). The addition of fluoro substituents to both aromatic rings was found to improve MDM2 inhibitory potency. Optimisation of the PK profile was achieved at the amide side chain substituent, that was also found to modulate cellular potency. A 4-benzenecarboxylic group gave the desired properties, and SAR studies around the ring revealed that the ortho-methoxy group, 35 (RG7388), had the ideal combination of MDM2 binding affinity, cellular potency and selectivity. RG7388 (35) gave complete tumour regression in vivo in an SJSA-1 xenograft model (25 mg/kg p.o.), and a favourable PK profile. RG7388 (35) was selected as the second generation MDM2-p53 clinical trial candidate owing to its improved potency, cellular selectivity and PK properties over RG7112 (25).

Fig. 16
figure 16figure 16

(a) X-ray structure of MDM2 (electrostatic surface) bound to cyanopyrrole 34 (green) (pdb 4JRG); (b) Design evolution of RG7388 (35) from 34

Full size image

Substituted piperidines, e.g. 36 (IC50 = 600 nM), were identified as hits from an FP-based HTS (Fig. 17a) [96]. Introduction of a 2-methoxyethoxyphenyl substituent gave improved potency and the alkyl substituent on the piperidine ring was found to be essential to promote the optimum conformation of the ring e.g. 37 (IC50 = 40 nM) [97, 98]. The MDM2 bound X-ray structure of 36 identified the main interactions as the trifluoromethylpyridine occupying the Phe19 pocket, the trifluoromethylphenyl residue projecting into the Trp23 pocket and the Leu26 pocket filled by the alkoxyphenyl group (Fig. 17b). Structure-guided optimisation resulted in 38 (Fig. 17c) [99]. A combination of a 2-trifluoromethylthiophene replacement for the trifluoromethylphenyl residue and a butanoic acid substituent to the alkoxyphenyl group gave a fivefold improvement in potency and cellular activity (38, IC50 = 7 nM). Compound 38 showed single agent activity in an A549 tumour xenograft at 200 mg/kg qd and in combination with taxoids.

Fig. 17
figure 17figure 17

(a) Design of piperidine 38; X-ray structures of MDM2 (electrostatic surface) bound to (b) piperidine 37 (green) (pdb 5HMK) (c) piperidine 38 (green) (pdb 5HMH)

Full size image

A structure-based design approach identified the piperidone series of inhibitors (e.g. AM-8553, 39, IC50 = 1.1 nM) [100]. An X-ray structure of 39 bound to MDM2 showed that the (3R)-carboxyl group at C3 forms a key interaction with His96, the 4-chlorophenyl group occupies the Typ pocket, the 3-chlorophenyl group fills the Leu pocket and the N-alkyl chain projects into the Phe pocket (Fig. 18a). Modifications to the N-substituent aimed to optimise interactions with the ‘glycine shelf’ region, a hydrophobic region on the surface of MDM2 formed by Gly58 (Fig. 18b) [101, 102]. Sulfone 40 (AMG 232, IC50 = 0.6 nM) showed excellent cell-free potency and cell growth inhibitory activity. Complete tumour regression in SJSA-1 xenografts observed (60 mg/kg/qd p.o.), and so AMG 232 was selected for clinical trials [103]. Additional SAR studies around the carboxylic acid moiety resulted in compounds with equivalent potency [104, 105].

Fig. 18
figure 18figure 18

X-ray structure of MDM2 (electrostatic surface) bound to (a) AM-8553 39 (green) (pdb 4ERF) (b) Design of AMG232 40

Full size image

Rational design based on the X-ray structure of the AP peptide bound to MDM2 (Fig. 5b) gave rise to the ‘central valine’ concept, resulting in the identification of substituted imidazole (41, IC50 = 30 nM) (Fig. 19) [106]. Structure-guided modifications to optimise potency and solubility gave rise to compound 42 (IC50 = 2 nM), however, the cellular activity in this series is poor compared with other series displaying similar MDM2-p53 potency [107].

Fig. 19
figure 19figure 19

Substituted imidazole MDM2-p53 inhibitors (41 and 42)

Full size image

Virtual screening identified dihydroisoquinolinone 43 as a hit MDM2-p53 inhibitor (IC50 = 540 nM) (Fig. 20a) [108]. An X-ray structure of 44 bound to MDM2 showed an unexpected binding mode, significantly different to most MDM2 inhibitors, whereby the Leu26 sub-pocket is occupied by the C7 ethyl ether and not a substituted phenyl ring as seen in the majority of other inhibitor series (Fig. 20b). X-ray structure guided optimisation gave rise to 45 (IC50 = 8 nM) with excellent cellular activity (Fig. 20c). An X-ray structure of 45 showed that chlorophenyl group requires to be in the (S)-configuration at C1 and showed opportunities for optimisation around in the Phe pocket occupied by the 4-pyridinyl group (Fig. 20c). Replacement of the pyridyl group with a trans-cyclohexyl amine gave NVP-CGM097 (46, IC50 = 1.7 nM) with good cellular and ADME properties [109]. The X-ray structure of 46 bound to MDM2 shows the Trp23 pocket is filled by the C1 4-chlorophenyl group, the Leu26 pocket filled by the C7 isopropyl ether, and the Phe19 sub-pocket is occupied and extended by movement of the protein to accommodate the large amine side-chain (Fig. 20d). NVP-CGM097 showed good PK properties in a number of mammalian species and induced tumour regression in SJSA-1 xenografts (30 mg/kg/qd p.o.). NVP-CGM097 (46) has been selected for clinical trials.

Fig. 20
figure 20figure 20

(a) Design evolution of NVP-CGM097 (46); X-ray structure of MDM2 (electrostatic surface) bound to (b) dihydroisoquinolinone 43 (green) (pdb 4ZYC); (c) dihydroisoquinolinone 45 (green) (pdb 4ZYI); (d) NVP-CGM097 (46) (green) (pdb 4ZYF)

Full size image

In an alternative approach, based on hit compound 41 structural considerations gave rise to a new scaffold comprising a substituted delta-lactam fused to a five-membered ring. Substructure searching identified lactam 47 as a novel hit with modest potency (IC50 = 1.5 μM) (Fig. 21) [110]. Elaboration of hit 47 identified structural requirements for potency including, introduction of a 5-chloropyridin-2-one moiety to occupy the Leu23 pocket and stack efficiently with His96. The clinical trial candidate from this series HDM-201 (48) was reported recently as a potent inhibitor with favourable PK and PD profiles in animal studies [111].

Fig. 21
figure 21figure 21

Design of HDM-201 (48)

Full size image

3.4 Clinical Development Summary

A number of MDM2-p53 inhibitors have entered clinical trials in solid and haematological malignancies (Table 2). A Phase I clinical trial of RG7112 25 in liposarcoma patients showed responses to therapy in 16 from 20 patients as partial response or stable disease [112]. A poor PK profile was observed, PD markers for apoptosis and p53 activation showed a linear relationship with drug exposure and a food effect was noted [113]. Some toxicity was observed including neutropenia and thrombocytopenia. A Phase I trial of 25 in various leukaemias showed activity in patients with AML [114]. The second generation inhibitor RG7388 (35) is under clinical investigation in a Phase III trial in AML [115, 116]. SAR405838 (32) was selected as the clinical candidate from the spirooxindole series and entered Phase I clinical trials as a single agent and in combination with the MEK1/2 inhibitor pimasertib. APG-115 (33) has entered Phase I clinical trials in advanced solid tumours and lymphomas as a single agent.

Table 2 MDM2 inhibitors in clinical trials
Full size table

A Phase I/II trial of AMG232 (40) is ongoing in metastatic melanoma, in combination with the MEK1/2 inhibitor trametinib and the B-Raf inhibitor dabrafinib. DS-3032b has been evaluated in a Phase I trial with some evidence of efficacy, and on target toxicity [117]. Phase I trials of MK-8242 demonstrated some activity in liposarcoma and in AML [118, 119]. Results of trials of CGM097 (46) and HDM201 (48) have not been reported to date.

Overall, a number of agents have shown efficacy in both MDM2 amplified tumours e.g. liposarcomas and AML. Dose limiting p53 mediated on target toxicity, including thrombocytopenia, has been reported. To date, no MDM2-p53 inhibitor has been approved, and the second generation inhibitor RG7388 has advanced the furthest.

4 SMAC-XIAP

4.1 Biology

The inhibitor of apoptosis (IAP) family of proteins inhibits entry into apoptosis at the point of caspase activation [120]. XIAP binds to the N-terminal amino acids of caspases 3, 7 and 9 via its BIR domains and directly inhibits their activity. The activating Smac promotes apoptosis by binding to XIAP [121]. The ML-IAP proteins bind to Smac, preventing its association with XIAP and so inhibiting apoptosis (Fig. 22). The c-IAPs have a different mechanism, acting as negative regulators of the alternative NF-κB pathway by inhibiting NF-kappa-B-inducing kinase (NIK), and as activators of the classical NF-κB pathway by acting on Receptor-interacting serine/threonine-protein kinase 1 (RIPK1). Amplification or overexpression of XIAP has been observed in haematological and solid tumours, and is linked to poor prognosis and poor response to therapy. Similarly, amplification or overexpression of c-IAPs has been identified in a large variety of tumour types. The role of XIAP at the convergence of multiple apoptotic signaling pathways has made it an attractive target for therapeutic intervention.

Fig. 22
figure 22figure 22

The role of IAP proteins in caspase activation

Full size image

4.2 Peptides and Structural Biology

The X-ray structure of Smac bound to XIAP shows the N-terminal four residues of Smac bound to a groove on the surface of XIAP (Fig. 23a, b) [122]. Validation of targeting the Smac-IAP interaction as a therapeutic strategy was demonstrated using cell-penetrant synthetic peptides including seven N-terminal residues of Smac and a cell penetrating TAT domain were found to sensitise TNF-related apoptosis-inducing ligand (TRAIL)-induced apoptosis and cytotoxic drug treatment in cells, but did not initiate apoptosis as single agents [123]. N-Terminal Smac mimetic tetrapeptide libraries established the pharmacophore for the binding to BIR3 (Fig. 23c) [124]. Modifications of the alanine residue at position 1 are limited to small l-amino acids, e.g. Abu, and retention of the NH is essential. Substitution at position 2 (Val) is tolerated, consistent with this residue projecting into solution, but d-amino acids are not tolerated. The proline ring at position 3 was required for activity, only the four-membered azetidine substitution was tolerated. Replacement of the isoleucine at position 4, with aromatic rings e.g. Phe or Nal, was favoured. Non-amino acid substituents at position 4 reduced the peptidic nature of the compounds and improved activity e.g. 49 (KD = 12 nM). Substitution onto the proline ring added potency, e.g. 50 (KD = 5 nM). Structure based design of peptide inhibitors gave 51 [125, 126].

Fig. 23
figure 23figure 23

X-ray structure of the BIR3 domain of XIAP complexed to Smac (a) dimeric structure of XIAP (blue) and Smac (green); (b) X-ray structure of the BIR3 domain of XIAP (electrostatic surface) complexed to Smac (green) (pdb 1G73); (c) pharmacophore for tetrapeptide Smac-mimetic IAP inhibitors; (d) structures of peptide IAP inhibitors 49–51

Full size image

4.3 Hit Discovery and Optimisation

A number of structurally diverse of peptidomimetic scaffolds have been used to develop potent mono-valent Smac mimetics that show a range of selectivity between cIAP-1 and -2 and XIAP. Structure-based design of cIAP selective inhibitors revealed that variations at the P4 amino acid gave good selectivity, 2-pyrimidinyl compound (52) was a selective inhibitor of cIAP-1 and -2, but >2,000-fold less potent vs XIAP [127]. The selectivity was explained by modelling that showed an adverse lone–pair interaction in the P4 binding pocket between the pyrimidine nitrogens and the hydroxyl of Thr308 in XIAP, that is not seen for Arg313 in IAP-1. Structure-based design established the [7,5]-bicycle as a suitable dipeptide replacement for the second and third amino acid positions (Fig. 24) [128]. Compound 53 showed potent inhibition of BIR3-IAP interactions and was cytotoxic as a single agent and in combination with doxorubicin.

Fig. 24
figure 24figure 24

(a) X-ray structure of the BIR3 domain of XIAP complexed to 49 (green) (pdb 2I3I); (b) XIAP inhibitors 52 and 53

Full size image

Smac mimetic AT-406 (Fig. 25, 54) incorporates a [5,8]-fused bicyclic replacement for residues 2 and 3 and is predicted to make additional hydrophobic contacts from the amide sidechain to Trp323 in the XIAP BIR3 domain [129]. AT-406 (54) is a pan-selective IAP inhibitor (BIR3 Kd = 4.4, 0.7, 1.9 nM XIAP, cIAP1, cIAP2, respectively), and is cytotoxic in a number of cell lines. AT-406 (54) shows good oral PK in a number of mammalian species.

Fig. 25
figure 25figure 25

XIAP inhibitor AT-406 (54)

Full size image

In order to rapidly determine the optimal substitutions of tetrapeptide Smac mimetics, solid-phase peptide libraries were employed [130]. A scan of the P1-P3 positions identified MeAla, ChGly and Pro, for each position, respectively. Similarly, a systematic scan of the P4 position showed a diphenylethylamine moieity conferred potency. Additional SAR studies at the P4 position were restricted to amines connected to an aromatic ring by one or two carbons. The thiadiazole compound GDC-0152 (55) showed improved potency (BIR3 Ki = 28, 17, 43 nM XIAP, cIAP1, cIAP2, respectively). The X-ray structure of 55 bound to chimeric BIR3 domains (Fig. 26) shows the thiadiazole ring positions the phenyl residue into the P4 pocket. 55 showed efficacy in an MDA-MD-231 breast cancer xenograft, and a reasonable PK profile. On the basis of these results 55 was selected for clinical trials. GDC-0917 (56) is a second generation IAP inhibitor which is orally active and has improved PK [131].

Fig. 26
figure 26figure 26

(a) X-ray structure of cIAP1/XIAP chimeric BIR3 domain with 51; (b) GDC-0152 (55) and GDC-0917 (56)

Full size image

NMR and X-ray based fragment screening with XIAP-BIR3 domain identified weakly binding fragment 57, which bound with the piperazine ring in the P1 pocket (Fig. 27a) and the piperidine ring in the P3 pocket [132]. Elaboration of the fragment by the addition of a methyl group to access the alanine pocket and switching to a pyrrolidine amide gave 58 (XIAP-BIR3 IC50 = 52% at 1 mM) (Fig. 27b). Amide 55 (XIAP-BIR3 IC50 = 5.5 μM) was identified from a virtual screening library of amides and the gem-dimethyl group added to improve potency and the metabolic stability of the azaindoline. Introduction of a methoxymethyl group on the piperazine ring improved potency (60), and the addition of a 6-benzyl group improved interactions with the P4 pocket (61, XIAP-BIR3 IC50 = 0.16 μM). Elaboration of the piperazine substituent to an iso-indolinone resulted in ASTX660 (62), a potent inhibitor of XIAP and c-IAP1 that was selected for clinical trials.

Fig. 27
figure 27figure 27

Evolution of ASTX-660 (58). X-ray structure of XIAP-BIR3 (electrostatic surface, ligand green) with (a) 57 (pdb 5C3H); (b) 58 (pdb 5C7B); (c) 60 (pdb 5C84); (d) 61 (pdb 5C83); (e) Evolution of ASTX660 (62) from 57

Full size image

Birinapant (Fig. 28, TL32711, 63) is a bivalent Smac-mimetic that shows potent XIAP and cIAP antagonism, and induces the downregulation of specific pools of cIAPs in cells [133, 134]. 63 induced apoptosis in cell lines either as a single agent by inducing the formation of a RIPK1-caspase 8 protein complex. 63 has been shown to potentiate some chemotherapy drugs in a p53-independent manner. Single-agent anti-tumour activity was demonstrated in patient derived xenograft models of ovarian cancer, colorectal cancer and melanoma. Birinapant 63 is currently undergoing clinical trials in solid and haemaotological malignancies.

Fig. 28
figure 28figure 28

Clinical trial candidates Birinipant (63) and LCL161 (64)

Full size image

4.4 Clinical Development Summary

A number of XIAP and c-IAP inhibitors are in clinical trials currently in solid and haematological malignancies, as single agents and in combination with chemotherapy or anti-PD1 antibody therapies (Table 3) [138]. In Phase I trial GDC-0917 (52) was well tolerated and showed good PK and PD properties, 2/42 patients showed a complete response (one ovarian cancer and one MALT lymphoma), four patients had stable disease [131, 135]. A Phase I trial of LCL161 (Fig. 28, 64) showed the drug was well tolerated and significant degradation of cIAP1 indicated a PD effect [137]. Although clinical evaluation of IAP inhibitors is in the early stages, initial results are encouraging.

Table 3 IAP inhibitors in clinical trials
Full size table

5 Outlook and Conclusions

PPIs are a challenging class of molecular targets to address with small-molecule inhibitors. The importance of PPIs in the biology of cancer, and particularly the regulation of apoptosis, has stimulated considerable drug discovery effort. The application of modern technologies such as screening of virtual, fragment, and lead-like libraries, coupled with structure-based design has enabled the discovery of potent and selective PPI inhibitors. The inhibitors profiled in some cases stray outside the confines of the ‘rule-of-five’ emphasising the challenge of targeting large, lipophilic sites [139].

The first PPI inhibitor has been licenced for clinical use in CLL, demonstrating the efficacy of selective Bcl-2 inhibition. Clinical trials of MDM2-p53 and IAP inhibitors are at an earlier stage. Preliminary results with MDM2-p53 inhibitors show promise, but with the limitations of on-target toxicity. The IAP inhibitors show less toxicity and some efficacy.