Discovery and Development of Kelch-like ECH-Associated Protein 1

Oct 3, 2016 - Biography. Meng-Chen Lu earned her Bachelor of Science degree in medicinal chemistry from China Pharmaceutical University. Since 2014, s...
0 downloads 4 Views 9MB Size
Perspective pubs.acs.org/jmc

Discovery and Development of Kelch-like ECH-Associated Protein 1. Nuclear Factor Erythroid 2‑Related Factor 2 (KEAP1:NRF2) Protein− Protein Interaction Inhibitors: Achievements, Challenges, and Future Directions Zheng-Yu Jiang,†,‡ Meng-Chen Lu,† and Qi-Dong You*,†,‡ †

State Key Laboratory of Natural Medicines and Jiang Su Key Laboratory of Drug Design and Optimization, China Pharmaceutical University, Nanjing 210009, China ‡ Department of Medicinal Chemistry, School of Pharmacy, China Pharmaceutical University, 24 Tongjiaxiang, Nanjing 210009, China ABSTRACT: The transcription factor Nrf2 is the primary regulator of the cellular defense system, and enhancing Nrf2 activity has potential usages in various diseases, especially chronic age-related and inflammatory diseases. Recently, directly targeting Keap1−Nrf2 protein−protein interaction (PPI) has been an emerging strategy to selectively and effectively activate Nrf2. This Perspective summarizes the progress in the discovery and development of Keap1−Nrf2 PPI inhibitors, including the Keap1−Nrf2 regulatory mechanisms, biochemical techniques for inhibitor identification, and approaches for identifying peptide and small-molecule inhibitors, as well as discusses privileged structures and future directions for further development of Keap1−Nrf2 PPI inhibitors. diseases,21 and inflammation.22 To date, numerous Nrf2 activators have been reported, and some of these have been used in clinical trials to explore clinical applications. Among them, Tecfidera (dimethylfumarate, DMF) has been the most successful Nrf2 activator that has been approved by the FDA to treat patients with relapsing forms of multiple sclerosis.23 Another potent Nrf2 activator, the synthetic triterpenoid [2cyano-3, 12-dioxo-oleana-1,9(11)-dien-28-oic acid, methyl ester] (CDDO-Me), is now being assessed in a phase 3 study of patients with connective tissue disease-associated pulmonary arterial hypertension. Sulforaphane, an isothiocyanate compound extracted from broccoli sprouts, has also been evaluated in several clinic studies.24,25 This progress has inspired further development of therapeutic agents around the cytoprotective role of Nrf2. In this Perspective, we focused on the discovery and development of Keap1−Nrf2 protein−protein interaction (PPI) inhibitors as a selective and effective way to explore Nrf2targeting therapies.

1. INTRODUCTION Transcription factor Nrf2 (nuclear factor erythroid 2-related factor 2) is the primary regulator of cytoprotective genes in cells,1,2 and it activates transcription in response to electrophiles and reactive oxygen species (ROS) under cellular stress.3 Nrf2 binds DNA as a heterodimer with small musculoaponeurotic fibrosarcoma (Maf) proteins, MafF, MafG, and MafK.4 The binding site of Nrf2 in the DNA sequence is the antioxidant/ electrophile response element (ARE/EpRE, 5′TGACNNNGC-3′).5 Previous studies have determined that Nrf2 controls the expression of numerous cytoprotective proteins, including phase I and phase II drug-metabolizing enzymes, drug transporters, and antioxidants.6 However, there has been increasing evidence that the target genes of Nrf2 are involved in a much wider range of activities, including lipid and glucose metabolism, heme and iron metabolism, and gene transcription.7−11 Thus, Nrf2 not only plays an important role in cellular defense but also influences energy metabolism, inflammation, and cell growth.12 These important biological functions of Nrf2 have attracted interest from a broad range of researchers in various fields, especially the drug discovery community.13,14 Many studies have demonstrated that Nrf2 plays important roles in protection against various diseases, especially chronic agerelated and inflammatory diseases, including cancer,15−17 neurodegenerative diseases,18 cardiovascular diseases, acute lung injury,19 chronic obstructive pulmonary disease,20 kidney © 2016 American Chemical Society

2. REGULATORY PPIS THAT MODULATE NRF2 ACTIVITY Given the central role of Nrf2 in the cellular defense system, Nrf2 activity is tightly regulated. Clearly understanding the regulatory mechanism of Nrf2 activity is the basis for Received: April 15, 2016 Published: October 3, 2016 10837

DOI: 10.1021/acs.jmedchem.6b00586 J. Med. Chem. 2016, 59, 10837−10858

Journal of Medicinal Chemistry

Perspective

Figure 1. Schematic diagram of the regulation of Nrf2 activity.

mimic the endogenous process of Nrf2 activation.39 These Nrf2 activators enhance the transcriptional activity of Nrf2 via Salkylation of reactive cysteines in Keap1.5,40 The mechanism of electrophilic Nrf2 activation has been further confirmed by the crystal structure of a well-known electrophilic Nrf2 activator, CDDO with Keap1.41 Issues regarding this type of Nrf2 activator have been summarized by another excellent Perspective.42 The results mentioned above provide the Keap1−Cul3 E3 ligase-mediated Nrf2 ubiquitination model with a unique sensor-switch regulatory mechanism mediated by the Keap1 cysteine residues. 2.2. Unique Mechanism Underlying Keap1−Nrf2 PPI. The case of Nrf2 is much more complex. Existing achievements have proved that Nrf2 contains two Keap1 binding motifs, called ETGE and DLG. Two Keap1 protein molecules form a homodimer and interact with these two motifs in one molecule of Nrf2 protein. The early experimental proof for this PPI model was from an ITC (isothermal titration calorimetry) experiment of the Keap1 DC domain and the Nrf2 Neh2 domain.43 The obtained titration curve exhibits a biphasic curve, which is fitted best with a two-site binding model. After that experiment, the co-crystal structures of the Keap1−Nrf2 ETGE peptide and the Keap1−Nrf2 DLG peptide44 were reported, which confirmed that Nrf2 contains two Keap1 binding sites. The available crystal structures of Keap1 and Keap1 with ligands are listed in Table 3. In addition, the experimental results showed that the Keap1 protein exists as a stable dimer rather than a monomer in mammalian cells, and the dimerization of Keap1 is indispensable for its interaction with Cul3.45 The direct structural evidence for the Keap1 homodimer is the result of single particle electron microscopy that clearly showed two large spheres attached by short linker arms to the sides of a small forked-stem structure, resembling a cherry-bob.46 These research results clearly supported the twosite molecular recognition model of Keap1−Nrf2 and were also in line with the “hinge and latch” model that previously obtained wide acceptance.44 In that model, the high binding affinity ETGE motif acts as the hinge to tightly link Nrf2 with the Keap1 dimer and the weak binding affinity DLG motif acts as the latch to switch ubiquitination. Only when both of the

developing targeting modulators. Despite the identification of an increasing number of regulatory mechanisms, Keap1involved PPIs play a central role in the precise regulation of Nrf2 in accordance with the cellular oxidative state (as shown in Figure 1).26,27 2.1. General Model of Keap1−Cul3 E3 LigaseMediated Nrf2 Ubiquitination. First, Keap1 (Kelch-like ECH-associated protein 1) is an adaptor in the Cul3 (cullin3)based ubiquitin E3 ligase that is responsible for Nrf2 ubiquitination. E3 ligases take part in ubiquitin transfer to a target protein and mediate substrate specificity in the highly heterogeneous ubiquitination system (there are >600 E3s in humans).28 The cullin-based E3 ligase is composed of multiple subunits that are assembled on a cullin scaffold, binding a RING-box protein at the N terminus and an adaptor protein as a substrate receptor (responsible for substrate specificity) at the C terminus.29 In the Keap1−Nrf2−Cul3 complex, Keap1 acts as the substrate receptor and selectively recognizes Nrf2, linking Nrf2 to the scaffold protein Cul3. The recognition of the substrate by Keap1 is the key step for the selectivity. Therefore, Keap1−Nrf2 PPI ensures ubiquitination substrate selectivity. Cul3 is the scaffold protein in the assembly of the E3 ligase complex, and Keap1−Cul3 PPI ensures that specific lysines of Nrf2 are located in the proper position at the catalytic center. This PPI-mediated substrate ubiquitination is the typical mechanism in the ubiquitin system. In addition, Keap1 also acts as the sensor and switch for the Nrf2 ubiquitination machine.30 Keap1 is a cysteine-rich protein that contains 25 (mouse) or 27 (human) cysteine residues, some of which have been identified as sensors of electrophilic and/or oxidative assault.31−35 Under stress conditions, these sensitive cysteine residues are oxidized to disulfides or conjugated to electrophiles.36 These covalent modifications affect the precise assembly of the E3 ligase complex and suppress the ubiquitination of Nrf2, which finally leads to the activation of Nrf2.37,38 Activation of the Nrf2-regulated cytoprotective system promotes cytosolic reductase activities, elevates the reduced glutathione level, and accelerates the metabolism and export of xenobiotics. These findings have also inspired the development of electrophilic Nrf2 activators that 10838

DOI: 10.1021/acs.jmedchem.6b00586 J. Med. Chem. 2016, 59, 10837−10858

Journal of Medicinal Chemistry

Perspective

Figure 2. Conformation cycling model of Keap1−Nrf2 regulation.

two motifs bind to the Keap1 dimer will the ubiquitination of Nrf2 be carried out. The intracellular oxidative stress signals mainly induce the disassociation of the Keap1−Nrf2 DLG motif and block the ubiquitination of Nrf2. However, recent studies have provided new insights into the Keap1−Nrf2-Cul3 regulation pattern. Dinkova-Kostova’s group developed a quantitative Förster resonance energy transferbased system using multiphoton fluorescence lifetime imaging microscopy to investigate the interaction between Nrf2 and Keap1 in single live cells.47 Their research results showed that the complex of the Keap1 dimer with Nrf2 adopts the following two distinct conformations under homeostatic conditions: the “open” conformation, in which Nrf2 binds to the Keap1 dimer through the high binding affinity ETGE motif, and the “closed” conformation, in which both of the two motifs in Nrf2 bind with the Keap1 dimer. The conformation of the Keap1−Nrf2 complex switches between the two states, and ubiquitination of Nrf2 only occurs when the “closed” conformation is formed, which is termed the “conformation cycling model” (as shown in Figure 2).48 Under induced conditions, the accumulation of the complex in the closed conformation, not the open conformation, has been observed. This accumulation most likely occurs because the inducers have disrupted the ubiquitination step, and the closed conformation of the complex cannot begin cycling. Thus, intracellular Keap1 is anchored to Nrf2, and newly translated Nrf2 accumulates and translocates into the nucleus, which finally turns on the expression of cytoprotective genes. Then, inducers activate

Nrf2 by inhibiting the regeneration of free Keap1 rather than releasing Nrf2 from the Keap1 dimer. Recent studies on the Nrf2 DLG motif have further confirmed the conformation cycling model. The previously obtained minimal DLG motif-derived peptide has shown weaker binding affinity and is poorly tolerated with further optimization.49 This ambiguity was revealed by the further investigation of Masayuki Yamamoto’s group of the minimum Keap1-binding sequence of the Nrf2 DLG motif. The authors defined a new DLGex motif that covers a sequence much longer than what was previously defined.50 The binding affinity of the DLGex motif-derived peptide, rather than the DLG motif peptide, is consistent with the second binding constant of the Nrf2 Neh2 domain. Moreover, in kinetic analyses, Keap1− DLGex binding follows a fast association and fast dissociation model, whereas Keap1−ETGE binding contains a slow-reaction step that leads to a stable conformation. This research suggested that the dissociation of Keap1−DLGex may occur spontaneously, which indicates that the “hinge and latch” behavior is a dynamic process. These studies together confirmed PPIs’ decisive role in modulating the activation of Nrf2. Keap1 bridges Nrf2 with Cul3 through PPIs, which enables the Keap1-mediated selective ubiquitination of Nrf2. Besides, the Keap1 dimer interacts with Nrf2 through a unique pattern, which thereby benefits the precise modulation of Nrf2. Among these PPIs, Keap1−Nrf2 PPI is closely associated with Nrf2 activity, which provides a great way to discover Nrf2-targeting modulators. 10839

DOI: 10.1021/acs.jmedchem.6b00586 J. Med. Chem. 2016, 59, 10837−10858

Journal of Medicinal Chemistry

Perspective

Figure 3. Subpocket analysis of Keap1 substrate binding cavity. On the basis of the Keap1−Nrf2 crystal structure, the Keap1 substrate binding cavity can be divided into six subpockets. (A) Subpocket analysis of Keap1 substrate binding cavity based on the Keap1−Nrf2 ETGE complex (PDB code 1X2R). (B) Subpocket analysis of Keap1 substrate binding cavity based on the Keap1−Nrf2 DLGex complex (PDB code 3WN7). (C) Summary of key residues in each subpocket.

Figure 4. Co-crystal structure of Keap1−Nrf2 PPI. (A,D) Binding surface and key polar residue interactions of the Keap1−Nrf2 ETGE motif (PDB code 1X2R). (B,E) Binding surface and key polar residue interactions of the Keap1−Nrf2 DLG motif (PDB code 2DYH). (C,F) Binding surface and key polar residue interactions of the Keap1−Nrf2 DLGex motif (PDB code 3WN7). Hydrogen bonds are represented by green dashed lines, and electrostatic interactions are represented by yellow dashed lines. The carbon atoms of Nrf2 residues and Keap1 residues are colored cyan and purple, respectively.

structure of the Keap1−Nrf2 ETGE motif (Figure 3A).56 However, given the Keap1−Nrf2 DLGex interaction,50 a P6 subpocket should be included in Keap1 (as shown in Figure 3B). The contact surfaces involved in Keap1−Nrf2 PPI are more typical of those involved in protein−small-molecule interactions (300−1000 Å2) than of PPIs (1500−3000 Å2).57,58 It is reasonable that the two regions in Nrf2 that are responsible for Keap1 dimer binding are adjacent in the primary sequence, which does not allow for a large binding surface. However, the binding strength between Keap1 and Nrf2 is quite potent. The 9-mer Nrf2 ETGE peptide can reach a Kd of approximately 20 nM.59 This information indicates that the substrate binding cavity of Keap1 has more well-characterized hot-spots on the one hand and is precisely specific with ligands on the other hand. This feature ensures substrate selectivity but raises difficult issues for developing modulators. In general, PPIs are classified into the following two categories: a domain−domain pattern that is mediated by the interaction between two protein domains and a peptide-domain pattern that is mediated by the interaction between a linear sequence of residues of one of the partners and a domain of the other one.60,61 The Keap1−Nrf2 ETGE interaction and the Keap1−Nrf2 DLGex interaction belong to the two types,

3. CHARACTERISTICS OF KEAP1−NRF2 PPI PPIs are involved in almost every biological function. Hundreds of thousands of PPIs51,52 together constitute the huge interactome that affects cellular behaviors.53,54 Thus, it is not surprising that extraordinary diversity and complexity exist in PPIs. Although PPIs may adopt quite different sizes, shapes, and electrostatics,55 PPIs also share some similar characteristics, such as a large interface area, shallow interaction sites, and multiple binding interactions. These characteristics together present a challenge for agents targeting PPIs. Thus, it is imperative to obtain a thorough understanding of the mechanism of action of a given PPI target before discovery of modulators. The Keap1−Nrf2 interaction is an especially unique PPI. The binding pocket in Keap1 is open, which is similar to most PPIs, but it is well-defined. The size of the binding site is relatively small. The interface area of the Keap1−Nrf2 ETGE motif is approximately 529 Å2 (calculated from PDB code 1X2R), and that of the Keap1−Nrf2 DLGex motif is a bit larger, approximately 820 Å2 (calculated from PDB code 3WN7). Previously, we divided the Keap1 substrate binding cavity into five subpockets (P1−P5) based on the co-crystal 10840

DOI: 10.1021/acs.jmedchem.6b00586 J. Med. Chem. 2016, 59, 10837−10858

Journal of Medicinal Chemistry

Perspective

Table 1. Summary of Assays for Evaluating Keap1−Nrf2 PPI Inhibitors assay type In Vitro System biolayer interferometry (BLI)

assay principle

biological effect

notes

ref 56,62

biophysics-optics

Keap1 binding

binding kinetic analysis quantitative methods medium throughput

isothermal titration calorimetry (ITC)

physical chemistry-calorimetry

Keap1 binding

label free; quantitative methods; binding thermodynamics analysis low throughput

62

competitive fluorescence polarization (FP)

molecular size-based luminescence

Keap1−Nrf2 ETGE peptide disruption

high throughput

71

quantitative methods SPR-based competition assay

biophysics-optics

Keap1−Nrf2 ETGE peptide disruption

wide applicability for compounds; competition assay; medium throughput

59

fluorescence resonance energy transfer assay

proximity-based luminescence

Keap1−Nrf2 ETGE peptide disruption

high throughput

78

quantitative methods tracer displacement assay

fluorescence anisotropy and Van’t Hoff analysis

Keap1−Nrf2 ETGE peptide disruption

medium throughput

94

binding thermodynamics analysis differential scanning fluorimetry (DSF) assay Cellular and in Vivo System ARE-luciferase reporter gene assay ARE-β-lactamase reporter gene assay

melting temperatures of Keap1 protein

Keap1 binding

high throughput

luciferase enzymatic activity β-lactamase enzymatic activity

Nrf2 transcription activity

high throughput, high false positives

Nrf2 transcription activity

high throughput

69

83,84

provided by Invitrogen Nrf2−MafK or Keap1−RunX2 luciferase enzyme fragment complementation

luciferase enzymatic activity

Nrf2 translocation and Nrf2−MafK or Nrf2−RunX binding

high throughput

85

low false positives Nrf2−Keap1 luciferase enzyme fragment complementation (EFC) assay

luciferase enzymatic activity

Keap1−Nrf2 disruption

works in vivo

FRET and multiphoton fluorescence lifetime imaging microscopy

fluorescence lifetime

Keap1−Nrf2 disruption

direct detection of Keap1−Nrf2 interaction

enzymatic assay of Nrf2 dependent protein β-gal enzyme fragment complementation

enzymatic activity (NQO1, HO-1, etc.) β-gal enzymatic activity

protein level Nrf2-regulated genes

high throughput

Nrf2 translocation

high throughput low false positives provided by DiscoverX

gene expression

qRT-PCR detection of mRNA

mRNA level

cells and in vivo tissue;

Nrf2-regulated genes

low throughput

protein level

cells and in vivo tissue;

Nrf2-regulated genes

low throughput

protein expression

antigen−antibody reaction detection of protein

82

47,75

75

87

93

side chains and the peptide backbone. Glu79 and Glu82 insert into the P1 and P2 subpockets and form multiple electrostatic interactions with key arginine residues in Keap1, including Arg483, Arg415, and Arg380 (Figure 4D). The Keap1−Nrf2 DLGex interaction belongs more to the domain−domain pattern. The co-crystal structure shows that the DLGex region constitutes three helices, helix 1 (Leu19 to

respectively (as shown in Figure 4A,C). The Keap1−Nrf2 ETGE interaction is a typical example of peptide-domainmediated PPI. The Nrf2 ETGE peptide possesses a tight fourresidue β-hairpin conformation that comprises the conservative DXETGE motif, specifically the residues Asp77, Glu78, Glu79, Thr80, Gly81, and Glu82. This conformation is stabilized by intramolecular hydrogen bonds involving the Asp77 and Thr80 10841

DOI: 10.1021/acs.jmedchem.6b00586 J. Med. Chem. 2016, 59, 10837−10858

Journal of Medicinal Chemistry

Perspective

in the conformation cycling model. By using quantitative Förster resonance energy transfer-based methods, Baird et al. investigated Keap1−Nrf2 interactions in a single live cell and proposed a “conformational cycling” model for the Keap1mediated degradation of Nrf2.47 The authors found that the Keap1−Nrf2 complex sequentially takes on the following two distinct conformations: an “open” conformation in which the high-affinity ETGE motif of Nrf2 interacts with a single molecule of Keap1, followed by a “closed” conformation in which Nrf2 binds to both members of the Keap1 dimer.48 Subsequently, Masayuki Yamamoto group’s recent study proposed a two-state binding model of Keap1−ETGE in which Nrf2−ETGE binds to Keap1 by two steps. In the first step, Keap1 DC recognizes Nrf2 ETGE by a fast association and dissociation step, and then the complex adopts the fully optimized conformation by a slow conformational adjustment reaction in the second step.64 The current findings are reasonable but are far from a thorough confirmation. In contrast, the kinetic analysis results from an SPR experiment showed that Keap1−DLGex binding follows a fast association and fast dissociation model,50 which allows other signaling pathways to regulate Nrf2 activity. Without a doubt, the indepth investigation of Keap1−Nrf2 PPI, coupled with the cocrystal structure, can facilitate the discovery of Keap1−Nrf2 PPI antagonists.

Arg25), helix 2 (Ile28 to Leu30), and helix 3 (Arg34 to Phe37), which form a U shape that is distinctly different from the βhairpin conformation of the ETGE region.50 Helix 1 and helix 3 cover the outer surface of the P1 and P2 subpockets in Keap1, and helix 2 and the surrounding linking residues are located on the central part of the cavity. The overall binding pattern of DLGex is significantly different from that of the previously reported short DLG peptide (as shown in Figure 4). In the case of the DLGex motif, the side chain of Gln27 moves from the inside to the outside of the P1 subpocket and occupies a shallow groove composed of Ser431, Gly433, Cys434, His436, and Arg380. The side chain amide group of Gln26 supplies hydrogen bonds to the imidazole group of His436 and the main chain carbonyl group of Gly433 in Keap1, which is not found among other Keap1 ligands. Moreover, the side chain carbonyl oxygen atom of Gln26 can form hydrogen bonds with Arg380 in Keap1. Asp29 is inserted into the P1 subpocket and forms multiple hydrogen bonds with Arg415 and Arg483 simultaneously, while the short DLG peptide occupies this cavity with Gln27. Because of its negative charge, Asp29 can make a firm salt bridge with Arg415 in Keap1. Closer examination reveals that the positions of Asp27 also adopt a displacement, which allows the side chain carboxyl group to form multiple interactions with Arg415, Arg380, Asn382, and Asn414. Only the interaction with Arg415 can be observed in the Keap1− Nrf2 DLG complex. This result indicates that the DLGex peptide can form more polar interactions in the P2 subpocket than the DLG peptide. Overall, the incorporation of two helixes on both sides of the DLG peptide significantly changes the positions of the polar residues, especially Asp27 and Asp29, which form more polar interactions in the P1 and P2 subpockets. Compared with the Nrf2 ETGE peptide, the DLGex peptide can form similar polar interactions in the P1 and P2 subpockets with a completely different sequence (the DLD and ETGE sequence). Not only does this result stress the importance of polar interactions in Keap1 substrate recognition but also provides new ideas for both novel natural ligand discovery and the design of potent modulators. Moreover, the binding thermodynamics and kinetics profiling of the two regions also show significant differences. Several studies have elucidated that Keap1−ETGE binding is a single enthalpy-driven process,43,62 which is consistent with the key role of polar interactions in Keap1−ETGE binding. Our computational studies also showed that the polar subpockets, P1 and P2, contribute the major part of the total binding energy.63 The ITC results of Keap1−DC with a deletion mutant of Nrf2 Neh2, in which the residues of ETGE were removed, showed that the total entropy effects were beneficial for binding.43 These data indicated that the second binding site may have distinct binding thermodynamic behavior. The identification of the DLGex motif further confirmed this suggestion. The ITC experiment with the DLGex peptide and Keap1 clearly showed a synergistic enthalpy/entropy-driven process.50 This finding shows that even though the binding cavity of Keap1 has many polar residues, the overall entropy effects can benefit binding and stresses that both polar and hydrophobic interactions contribute to anchoring small molecules and peptides. The binding kinetics of Keap1− ETGE and Keap1−DLGex also have great differences. The kon and koff determined from both surface plasmon resonance (SPR)59 and the bio-layer interferometry (BLI)62 assays show that the Nrf2 ETGE peptide has a very slow dissociation rate, confirming the recognition and anchoring effects of this region

4. ASSAYS FOR EVALUATING KEAP1−NRF2 PPI INHIBITORS To modulate a PPI of interest, well-established agents must be identified first. However, the identification of the PPI inhibitors is another hurdle in PPI drug discovery. Unlike traditional drug targets, especially enzymes, PPIs do not involve enzymatic activity in most cases. Currently, various interdisciplinary techniques, especially biochemical and biophysical techniques, can be used to develop screening assays to find active hits from small arrays and large compound libraries. Cell-based high content screening, cellular downstream phenotype examination, and in vivo imaging to monitor PPIs have also been used to further evaluate the biological effects of PPI modulators. For the Keap1−Nrf2 PPI, various methods have been developed to meet the needs of drug discovery, including high-throughput screening (HTS) of hits, improvement of PPI inhibition activity, and optimization for drug-like properties. Hereafter, we will describe currently available assays for the identification and evaluation of Keap1−Nrf2 inhibitors (Table 1). Basic principles and advantages as well as strengths and weaknesses of these different assay types will be included. 4.1. In Vitro Assays. Accordingly, depending on the general propose of in vitro methods, they can be divided into the following two categories: Keap1 binding evaluation and Keap1−Nrf2 interaction inhibition. 4.1.1. Keap1 Binding Assay. The most straightforward parameter reflecting binding strength is the equilibrium dissociation constant Kd. Therefore, assays that can determine Kd and precisely measure direct binding are valuable in evaluating Keap1−Nrf2 PPI inhibitors. Because the estimation of Kd also plays an important role in PPI biology, these methods are often available before the discovery of PPI modulators. The Kd can be obtained via two distinct ways, the binding thermodynamic method and the binding kinetic method. The most commonly used technique for the binding thermodynamic method is ITC assay, which has been used in the evaluation of Keap1−Nrf2 interactions. The most attractive 10842

DOI: 10.1021/acs.jmedchem.6b00586 J. Med. Chem. 2016, 59, 10837−10858

Journal of Medicinal Chemistry

Perspective

which can induce a strong difference in FP. For this reason, the main disadvantage of the FP assay is that one protein in a PPI must be minimized to a short peptide size to meet the need for the FP difference. It is more suitable for the peptide-domain PPI, to which the Keap1−Nrf2 ETGE PPI belongs. Longqin Hu’s group first reported using the FP assay for the discovery of Keap1−Nrf2 PPI inhibitors by an optimized fluorescently labeled Nrf2 ETGE peptide.71 The authors also used this assay to conduct HTS and discovered a first-in-class small-molecule Keap1−Nrf2 inhibitor.72 Since those reports, the FP assay has been widely used in the discovery and optimization of Keap1− Nrf2 inhibitors.56,69,73−75 The main advantage of the FP assay is that it can quantitatively evaluate the activity of inhibitors with a biological reagent and compound saving system, simple operation process (“mix-and-read assay”), and high-throughput screening. The FP assay also has some weaknesses. In addition to the size-dependent nature and the need for the fluorescent tracer peptide mentioned above, the FP assay cannot be conducted in a time-resolved manner and lacks time stability. A further improvement of the FP method, the two-dimensional fluorescence intensity distribution analysis (2D-FIDA),76 has also been used to perform HTS of Keap1−Nrf2 PPI inhibitors, and it has also identified active hits.77 To overcome the drawbacks of FP assays, some proximitybased fluorescent methods have also been established. Among them, a fluorescence resonance energy transfer (FRET)-based assay is the most widely used. FRET is the radiationless transmission of energy from a donor molecule that initially absorbs the energy to an acceptor molecule to which the energy is subsequently transferred. The efficiency of this energy transfer is inversely proportional to the sixth power of the distance between donor and acceptor, making FRET extremely sensitive to small changes in distance. Thus, it is a proximitybased method, which can be adapted to detect the distance between two proteins and avoids the process of minimizing one binding partner to a short peptide, which facilitates the setup of the screening assay. Moreover, the time stability of FRET over FP can facilitate HTS. This type of assay also works in the Keap1−Nrf2 PPI system, and Geoff Wells’ group has successfully developed a steady-state FRET-based assay to identify inhibitors of the Keap1−Nrf2 PPI.78 In addition to evaluating PPI inhibition, the fluorescent tracer displacement assay can also be used to rapidly assess thermodynamic parameters for target-ligand binding.79 The main advantage of this type of thermodynamic binding assay is that it can reach medium throughput, which can be used in hit-to-lead programs.80 A research group from Evotec AG reported a protocol for the determination of the thermodynamic parameters of Keap1−Nrf2 PPI inhibitors and tested the small-molecule inhibitors obtained from HTS.81 The Keap1−Nrf2 inhibition assays mentioned above are all fluorescence related, which may not work on compounds that fluoresce. The SPR-based solution competition assay can be used to examine the Keap1−Nrf2 PPI inhibition activity of fluorescent compounds.59 In this assay, a biotin-labeled Nrf2 ETGE peptide is captured on a streptavidin chip and competitively binds Keap1 with the inhibitors in the solution. Because of the nature of the SPR technique, it is not disturbed by the compound’s fluorescence and throughput is also limited. 4.2. Cellular and in Vivo Assays. 4.2.1. Assays for the Keap1−Nrf2 PPI Inhibition. Compared with the determination of the Keap1−Nrf2 PPI inhibition activity in vitro, direct evaluation of Keap1−Nrf2 PPI inhibition activity in live cells or

advantage of ITC is that it can give the quantitative contribution of enthalpy and entropy simultaneously, which can benefit further drug-like optimization of the inhibitors.65,66 Values for both ΔG and ΔH can be obtained from the ITC titration profiles, and entropic contribution (TΔS) can be calculated using the fundamental equation ΔG = ΔH − TΔS. Our group has successfully used this method to optimize Keap1−Nrf2 peptide inhibitors and obtained the most potent peptide inhibitor with a Kd value of approximately 10 nM.62 For the binding kinetic method, SPR and BLI are the commonly used techniques. In addition to the resulting Kd value, a binding kinetic assay can monitor the association and dissociation process in real time and obtain the association rate constant kon and the dissociation rate constant koff. As described above, these kinetic results further enhance the understanding of the Keap1−Nrf2 interaction mechanism. In the optimization of inhibitors, the kinetic profile is also an important drug-like index, which directly impacts drug efficacy and safety.67 Longqin Hu’s group used SPR to carry out kinetic analyses of the Keap1−Nrf2 interaction and determine the minimal Nrf2 ETGE peptide sequence required for Keap1 binding.59 By using the BLI technique, we assessed the kinetic profile of the potent small-molecule Keap1−Nrf2 inhibitor and found that the inhibitor followed the binding behavior of the ETGE peptide, which has quite a slow dissociation rate.56 Generally, the binding kinetic and thermodynamic profile together ensure the precise assessment of Keap1−ligand binding and provide useful tips for further optimization of inhibitors. The advantages of these methods are obvious, such as labelfree detection, real-time monitoring of binding interactions, and quantitative evaluation of binding strength. The shortcomings are also conspicuous. All of these assays need a separate operation for each test sample, which restricts the throughput of the assay. These assays can only be used for validation and step-by-step optimization of the inhibitors rather than highthroughput screening to find novel hits. In recent years, differential scanning fluorimetry (DSF) has been developed to detect ligand−receptor binding interactions that promote protein stability.68 The main advantage of DSF is that it can be performed with higher throughput without requiring large amounts of protein. Chunlin Zhuang et al. used this method to confirm the binding interaction of lead compounds with Keap1.69 4.1.2. Keap1−Nrf2 Inhibition Assay. Compared to Keap1 binding strength, Keap1−Nrf2 interaction inhibition activity is a more direct index reflecting the potency of the compounds. Because the PPI itself cannot induce biological effects directly in most cases, extra reporter groups, especially fluorescent groups, must be introduced. Among these fluorescent-based techniques, fluorescence polarization (FP) is the most widely used. The basic principle of FP is that moving fluorescent molecules emit light on a different plane when excited with linear polarized light.70 This effect is induced by the movement of the molecules in solution between excitation and emission, which can be significantly affected by the molecular size. The difference in polarization can be observed for fluorescently labeled species of different sizes, which essentially relies on a mass difference (preferably ∼10×). Thus, the FP assay is often used in a competition format, i.e., a short peptide derived from the binding motif of one protein will be fluorescently labeled and the binding partner will be treated as the receptor. Only in this way can we obtain a sufficient difference in the mass of the unbound and bound state of the fluorophore-bearing peptide, 10843

DOI: 10.1021/acs.jmedchem.6b00586 J. Med. Chem. 2016, 59, 10837−10858

Journal of Medicinal Chemistry

Perspective

Compared with these medium- to high-throughput assays for evaluating the activation of Nrf2, there are a number of lowthroughput but direct methods to validate the activation of Nrf2 and its function at the cellular level and in vivo, including Western blot and qRT-PCR analysis of Nrf2 and its downstream genes. As the prototypical Nrf2-target genes, NAD(P)H:quinone oxidoreductase-1 (NQO-1)87−89 and heme-oxygenase-1 (HO-1)90−92 are the most widely evaluated. However, it has been noticed that the basal expression of NQO-1 is considerably high in certain cell lines, especially tumor cell lines such as HCT116 cells.56,74,92,93 Thus, a variety of downstream genes should be used to accurately evaluate the compound’s effects on Nrf2-regulated genes. The current suite of available assay technologies, some of which are available as commercial kits, allow scientists to efficiently identify activators, confirm mechanism of action in vitro, and demonstrate pathway engagement in vivo in a tissue of interest. As such, the tools are available for robust drugdiscovery efforts.

in vivo is hard to achieve. Recently, a cell-based luciferase enzyme fragment complementation (EFC) assay has been reported to measure the dissociation of the Keap1−Nrf2 complex.82 The authors generated HEK293T cells that stably express the Nrf2 protein with the N-terminal luciferase fragment (CLuc−Nrf2) and the Keap1 protein with the Cterminal luciferase fragment (NLuc−Keap1). When Keap1 interacts with Nrf2, the complemented luciferase enzymes achieve their active form. Once the Keap1−Nrf2 PPI is disrupted, the complemented luciferase enzymes are disassembled, leading to a decrease in luciferase activity. Thus, the disruption of the Keap1−Nrf2 PPI can be associated with the decrease in luciferase activity, which can be easily detected and used in a high throughput intact cell-imaging platform. Moreover, with the help of a noninvasive molecular imaging approach, this assay can also be used to evaluate compounds in small animal models. The FRET-based assay system is an alternative to evaluate Keap1−Nrf2 inhibition activity in live cells. As mentioned above, Geoff Wells’ group has successfully established and used this type of assay to investigate the Keap1−Nrf2 PPI regulatory mechanism.47,48 They also applied it to the evaluation of small-molecule inhibitors.75 4.2.2. Assays for Nrf2 Activation. Because of the transcriptional activation activity of Nrf2, it is much easier to construct an assay to estimate the potency of Nrf2 activation qualitatively and quantitatively. Among the available methods, the most straightforward way is the Nrf2−ARE-luciferase reporter gene assay. Nrf2 induces a battery of cytoprotective genes by binding to ARE located in the promoter regions; thus, a luciferase gene under the control of ARE promoters can reflect Nrf2 activation. Moreover, the luminescence signal affected by the luciferase activity can be easily detected, which provides a rapid and convenient quantification of Nrf2−ARE induction. Therefore, a cellular ARE-luciferase reporter gene assay can be used in HTS, and several HTS studies using this assay have been reported.83,84 Apart from using luciferase as the reporter gene, some other ARE-controlled reporter systems are also commercially available. For example, Invitrogen provides the Cell Sensor ARE-bla HepG2 Cell-Based Assay, which uses βlactamase as the reporter. The main disadvantage of AREluciferase reporter gene assays is the frequent occurrence of false positives. The cell-based luciferase enzyme fragment complementation (EFC) assay, to a certain extent, can overcome this shortcoming. This method depends on the reconstitution of a “split” luciferase. On the basis of the interaction of Nrf2 with its nuclear partner MafK or runtrelated transcription factor 2 (RunX2), a research group from GlaxoSmithKline developed such an approach in which firefly luciferase is split into two fragments, which are genetically fused to Nrf2 and MafK or RunX2, respectively.85 Thus, a nontranscriptional readout is needed to minimize false positives, and the luminescence signal can provide information directly about the level of Nrf2 protein and its function, which is a clear advantage over the classical ARE-luciferase reporter assay. The commercially available PathHunter U2OS Keap1− Nrf2 Nuclear Translocation Cell Line from DiscoverX uses a very similar mechanism. Nrf2 also controls the transcription of several enzymes, and thus the enzyme activity of these downstream enzymes can be used as reporters directly. The examination of NQO1 enzyme activity is widely used in both classic Nrf2 activators86 and Keap1−Nrf2 PPI inhibitors.75 This type of assay is a good complement to the ARE-luciferase reporter gene assay for eliminating false positives.

5. MEDICINAL CHEMISTRY STRATEGIES IN THE DISCOVERY OF KEAP1−NRF2 PPI INHIBITORS Generally, Keap1−Nrf2 PPI antagonists often represent a challenging target for drug development approaches.95 That is especially true for the Keap1−Nrf2 PPI, as there is a high surface area and open binding site at the interface of the Keap1−Nrf2 PPI. Despite this challenge, with the remarkable progress that has been made in discovery techniques and the characterization and development of PPI inhibitors, several different types of Keap1−Nrf2 PPI inhibitors have been recently reported.96,97 5.1. Peptide Keap1−Nrf2 Antagonists. The high surface area and lack of characterized interactions are the main hurdles for targeting PPIs. Peptides are the ideal candidates to tackle this limitation and match the PPI target as they are able to closely mimic the principle features and proper conformation of a protein but can be easily synthesized and modified for functional fine-tuning.60 In addition, the sequence of the natural protein can directly guide the design of peptides. It is feasible to discover first-in-class PPI inhibitors from peptides and to explore the mechanisms of action of PPIs using peptides,61 as has been the process of discovering the Keap1−Nrf2 PPI inhibitors.98 Representative Keap1−Nrf2 PPI peptide inhibitors are listed in Table 2. 5.1.1. Exploration of the Minimal Nrf2 Peptide Sequence Required for Keap1 Binding. As mentioned previously, the strong binding site (Nrf2 ETGE region), which is a typical peptide-domain PPI pattern, provides the primary sequence for discovering peptide antagonists. Searching for the minimal peptide sequence from the primary sequence is a general approach for peptide PPI inhibitors in early stages. The process for identification of the minimal sequence can help researchers to understand the hot-spots of PPI. Moreover, the minimal peptide can lead to useful tools to investigate the PPI chemical biology and set up the screening assays previously mentioned in section 4.1.2. The reported co-crystal structure of Keap1−Nrf2 ETGE provides direct guidance for this research. In the two reported structures, the Nrf2 ETGE regions are the 9-mer and 16-mer ETGE-containing peptides, respectively. The structure indicated that the 9-mer peptide (LDEETGEFL) may be a good starting point for the search for the minimal peptide. Two different groups carried out similar research using different evaluation methods. Using a competitive SPR assay, Lonqin 10844

DOI: 10.1021/acs.jmedchem.6b00586 J. Med. Chem. 2016, 59, 10837−10858

Journal of Medicinal Chemistry

Perspective

leucine at the N-terminus of the 9-mer peptide, and thus the FITC-labeled 7-mer peptide should be treated as an 8-mer peptide. Therefore, the main difference occurs at the leucine of the C-terminus. In the co-crystal structure, this residue has an obvious interaction with the Keap1 protein, and only its terminal carboxyl acid can form an intramolecular hydrogen bond with the amine at the N-terminus, which may have slight effects on maintaining the peptide conformation. It is reasonable that it can be simplified to obtain the minimal fluorescent peptide tracer for Keap1 binding. Using this tracer, Hancock et al. found that another 7-mer Nrf2 peptide (peptide 6, Ac-DEETGEF-OH) also has moderate peptide inhibition activity, which indicates that this type of 7-mer Nrf2 peptide is the minimal sequence for Keap1 binding.49 Taken together, we can conclude that the 7-mer Nrf2 peptide 6 is the shortest Keap1 binding peptide that can retain the proper conformation. Moreover, for most Keap1 binding interactions, the FITClabeled “8-mer” ETGE containing peptide 2 is the simplest fluorescent peptide tracer for the FP assay and the 9-mer peptide 5 (Ac-LDEETGEFL-OH) is the optimal peptide inhibitor with potent Keap1−Nrf2 inhibition activity. 5.1.2. Structure-Based Optimization of the Nrf2 ETGE Peptide. Although the 9-mer peptide showed fair Keap1 binding affinity and Keap1−Nrf2 inhibition activity, its Keap1 binding affinity is still lower than that of the Nrf2 protein. Moreover, the four acidic residues strongly disfavor cellular penetration. Further optimization is necessary to improve the activity, shorten the length of the peptide, and replace the unnecessary acidic residues. An SAR study of the AcDEETGEF-OH sequence was carried out. The first glutamic acid in the peptide can be optimized, and changing it to alanine improves the IC50 to 0.73 μM. Moreover, inspired by the primary sequence of another Keap1 substrate protein, p62, replacement of glutamic acid with proline further increased the activity to an IC50 of 0.248 μM. The core ETGE motif is highly conserved for the activity, and changing the glutamic acid to glutamine or asparagine is not conducive to activity. Even changing the second glutamic acid in ETGE to aspartic acid can abolish activity, and if only the first glutamic acid in ETGE is replaced by aspartic acid, there is a 3-fold loss in activity. The threonine that can form a hydrogen bond with Keap1 is also sensitive to residue alterations, and only the serine derivative retains some activity. The phenylalanine on the C-terminal side of ETGE has a high tolerance for residue changes. Replacement with tryptophan, leucine, or tyrosine can maintain activity, and leucine is the optimal and can result in the most potent peptide (peptide 7, Ac-DPETGEL-OH), with an IC50 of 0.115 μM in this study.49 In a subsequent study, Wells’ group found that transforming the first aspartic acid to asparagine is also acceptable, although it will induce an 8-fold activity loss (peptide 9). Moreover, they found that terminal modifications have dramatic effects on activity. To promote the lipophilicity of the peptide, the C18 fatty acid stearic acid was introduced at the N-terminal. This conjugate (peptide 10) showed 5-fold greater activity (0.022 vs 0.115 μM), whereas the introduction of a benzoyl group had almost no effect on activity. Their research also showed that the C-terminal carboxyl group is important for Keap1−Nrf2 inhibition activity, and changing the carboxyl group to carboxamide caused a 6−10-fold decrease in activity (peptides 7 and 8 as an example).99 Our group further investigated this activity loss by molecular dynamics (MD) simulation and found that the C-terminal carboxyl group can form hydrogen bonds with Arg483, whereas

Table 2. Representative Keap1−Nrf2 PPI Peptide Inhibitors

a b

Fluorescent tracers used in FP assays: FITC-LDEETGEFL-NH2. Fluorescent tracers used in FP assays: FITC-βAla-DEETGEF-OH.

Hu’s group found that the 10-mer to 14-mer Nrf2 peptides had similar binding affinities to the Keap1 Kelch domain, whereas the 9-mer Nrf2 peptide had a moderate binding affinity of 352 nM, and Nrf2 peptides