Binding thermodynamics and kinetics guided optimization of potent Keap1–Nrf2 peptide inhibitors

Abstract

Activation of Nrf2 by directly inhibiting the Keap1–Nrf2 Protein–Protein Interaction (PPI) has gained research interest with regard to developing novel agents for treating inflammatory related diseases. In this study, computational methods were used to guide the rational activity improvement of peptide Keap1–Nrf2 inhibitors and to explore the Keap1 binding cavity. Terminal hydrophobic residues and Tyr residue replacements were introduced into the novel peptides. The experimental values of these peptides further confirmed the potential binding sites explored by the MD simulation. Finally, the most promising, peptide 5 with an intracyclic conformation showed a K_d value of 2.8 nM when binding to Keap1 and an IC₅₀ value of 9.4 nM in the Keap1–Nrf2 PPI fluorescence polarization assay. It proved that the active conformation locking strategy is quite useful in Keap1–Nrf2 PPI inhibitor design. It also provided a useful probe to investigate the Keap1–Nrf2 PPI system.

---

Introduction

Nuclear factor erythroid 2-related factor 2 (Nrf2) has gained a lot of research interest in the drug discovery community because of its important role in the cell defense mechanism.¹ Nrf2 is a key transcription factor that regulates oxidative stress. It can induce the transcription of various antioxidant and protective genes with the antioxidant response element (ARE) in the promoter region, which can counteract oxidative insults, such as reactive oxygen species (ROS) and electrophilic chemicals.² Discovering agents targeting the Keap1–Nrf2–ARE signaling pathway is being considered as a promising strategy for the treatment of oxidative stress and inflammation related diseases, including cancer, neurological disease, and diabetes.³ Recently, a growing body of studies has highlighted the neuroprotective effects of Nrf2 not only in Parkinson’s and Alzheimer’s disease, but also in multiple sclerosis, amyotrophic lateral sclerosis and epilepsy.⁴

It is now widely accepted that Kelch-like ECH-associated protein (Keap1) is the primary regulator of Nrf2. Under basal conditions, Keap1 co-localizes with Nrf2 in the cytosol and mediates its ubiquitination, which results in the proteasomal degradation of Nrf2. Upon oxidative stress, Keap1 can sense the cellular redox state using its reactive cysteines. ROS and electrophilic chemicals can covalently modify the Keap1 cysteine residues, leading to conformational changes of Keap1. The conformational change may alter the positioning of key lysine residues in Nrf2, which comprise the conjugating site for ubiquitin. Thus, Nrf2 can no longer be ubiquitinated by the Keap1-involved E3 ligase complex, and the free Keap1 will not be regenerated.^5–7 Subsequently, newly synthesized Nrf2 will not be locked by Keap1, and free Nrf2 can translocate into the nucleus, bind the ARE and induce transcription by forming heterodimers with other transcriptional regulatory proteins.⁸ Generally, Keap1–Nrf2 protein–protein interaction (PPI) plays an important role in modulating the activity of Nrf2, and competitive inhibition of Keap1–Nrf2 interaction provides an alternative way for Nrf2 activation. To date, most Nrf2 activators target this pathway through covalent binding of the cysteine residues of the Keap1 protein.⁹ Compared with covalent mechanisms, directly disrupting Keap1–Nrf2 PPI to active Nrf2 may have potential advantages such as lower toxicity and higher selectivity.¹⁰ Recently, exciting progress has been achieved in this field. Several potent small molecule PPI inhibitors have been reported by us and other groups.^11–14

However, the structural diversity of the Keap1–Nrf2 PPI inhibitors is deficient^10,13,15 and the understanding of the Keap1 binding interface is still limited.¹⁶ These limitations restrict further target validation using small molecules. On the contrary, the discovery of potent peptide PPI modulators is easier. Peptides based on regions of Nrf2 with varying lengths and sequences have been systemically investigated.¹⁷ The anti-inflammatory effect of the peptide Nrf2–Keap1 PPI inhibitor has been validated.¹⁸ The sequences of the other two Keap1 substrates, sequestosome-1 (p62) and prothymosin-α, have also been investigated.¹⁹ These reported peptides provide the basis for the further study of peptide modulators. The development of novel peptide PPI modulators can enable further exploration of the binding cavity of Keap1 and provide the privileged fragment for Keap1 binding, which can significantly facilitate the effort to design small molecules to control PPIs.

In this work, with the help of computational methods, we explored the sub-pocket of the Keap1 binding interface, discovered the potential polar binding site and evaluated the binding conformation constraint of the peptide. By means of comparing the residue sequences, the terminal Leu residues were added to the template (peptide 2) and the activity of the resulting peptide 3 was significantly improved. The phenylalanine of the peptide 3 was located in the potential polar binding site and tyrosine replacement of the corresponding phenylalanine (peptide 4), which occupied this pocket, significantly strengthened the binding. Moreover, in order to restrict the binding conformation, we also used the terminal disulfide linkage to cyclize the peptide. The resulting peptide 5 showed significant improvements in both the competitive fluorescence polarization (FP) assay and K_d value, and is the most potent Keap1–Nrf2 inhibitor to date.

Results and discussion

A detailed understanding of the structural basis of the Keap1 and Nrf2 PPI is critical to develop PPI inhibitors. The crystal structures of the Keap1 DC domain in complexes with the Nrf2 ETGE and DLG peptide have been determined by different groups,^20–23 which provided detailed information about the Keap1–Nrf2 PPI. In our previous studies, the Keap1 cavity was divided into five sub-pockets which played different roles in binding.^14,16

For the reported peptide inhibitors, 1 and 2, the Pro replacement of the Glu and the terminal Leu residues were the major differences. The Pro residue has proved helpful to mimic the β-hairpin secondary conformation for a long time.²⁴ While the β-hairpin secondary conformation is quite important for the binding of peptides to Keap1. The stable β-hairpin secondary conformation of the peptide can ensure the two key glutamate residues in the ETGE motif (denoted for short as E79 and E82 correspondingly in the following section) are located on the appropriate site and form multiple electrostatic interactions with key arginines of Keap1. These electrostatic interactions have been proved important for Keap1 binding by our²⁵ and others’ studies.²¹ For the terminal Leu residues, our previous studies have proved that these can occupy the outside hydrophobic sub-pockets and provide a positive contribution to Keap1 binding (as shown in Fig. 1). These results together indicate that the terminal hydrophobic residues and the key Pro residue should be incorporated into peptide Keap1–Nrf2 inhibitors.

Fig. 1 Keap1–Nrf2 ETGE binding pattern. Sub-pocket analysis of the Keap1 DC domain cavity depending on the crystal structures of the Keap1–Nrf2 complex (PDB ID: 1X2R and 2FLU). It contains five sub-pockets: polar sub-pockets P1 and P2, hydrophobic sub-pockets P4 and P5 and the central sub-pocket P3. Three-letter abbreviations represent the residues of Keap1 and single-letter abbreviations represent the residues of Nrf2.

Considering the absence of the terminal hydrophobic residues, the peptide 2 was optimized using peptide 1 as a template, resulting in the novel peptide 3, Ac-LDPETGEFL-OH. In order to validate these SAR analyses, we also synthesized and evaluated the activity of peptide 3 using the FP assay. The peptide 1 was selected as a control and demonstrated an IC₅₀ of 333 nM in our FP experiment, similar to the reported value (389 nM). The peptide 3 showed an IC₅₀ of 42.6 nM, much more potent than peptide 1 (Table 1). The isothermal titration calorimetry (ITC) assay, which is widely used to assess the thermodynamics and affinity of ligand–receptor interaction,²⁶ was applied to quantify the ligand binding. The resulting K_d values also produced a consistent result. The ITC profiles clearly fit a reversible 1 : 1 binding model, suggesting that one peptide molecule bound to one molecule of Keap1. We first assessed the thermodynamic parameters of peptide 1 and 3. The K_d value of peptide 3 is about 4 fold smaller than peptide 1 (as shown in Table 2 and Fig. S3†). The Pro replacement of the Glu resulted in the formation of a stable β-hairpin secondary conformation of the peptide, resulting in a favorable change of entropy value. Compared with peptide 2, peptide 3 also showed a significant improvement in PPI inhibition activity, which further confirmed the importance of the terminal hydrophobic residues in binding Keap1.

Table 1 The IC₅₀ of the peptides in disrupting the Keap1 Kelch domain–Nrf2 interaction determined by the FP-based competition assay^14,15,17

No.	Sequence	IC50 ± SE^a^,^b (nM)
a Reported IC50 value.^19b Experimental IC50 value.
1	Ac-LDEETGEFL-OH	389 ± 33^a
		333 ± 65^b
2	Ac-DPETGEL-OH	115 ± 13^a
3	Ac-LDPETGEFL-OH	42.6 ± 6.4^b
4	Ac-LDPETGEYL-OH	29.6 ± 6.1^b
5	Ac-c[CLDPETGEYLC]-OH	9.45 ± 1.2^b
6	Ac-CLDPETGEYLC-OH	22.6 ± 33^b
7	Ac-c[CDPETGEYC]-OH	45.9 ± 28^b
8	Ac-CDFETGEYC-OH	65.8 ± 41^b

---

Table 2 Thermodynamic parameters of the interactions of seven peptides with Keap1 as determined by ITC assay^a

No.	Sequence	N	Kd (nM)	ΔH (kcal mol^−1)	TΔS (kcal mol^−1)	ΔG (kcal mol^−1)	Mw (Da)
a N is the stoichiometric coefficient. Kd is the binding constant. ΔH, ΔS, and ΔG refer to the changes in binding enthalpy, entropy, and total Gibbs free energy, respectively. ΔG is calculated according to the equation ΔG = ΔH − TΔS, with T the absolute temperature set up for the ITC experiment.
1	Ac-LDEETGEFL-OH	1.0 ± 0.03	271	−15.9	−7.0	−8.9	1094.11
3	Ac-LDPETGEFL-OH	1.0 ± 0.01	46.5	−10.9	−0.9	−10.0	1062.11
4	Ac-LDPETGEYL-OH	1.0 ± 0.01	42	−17.8	−7.7	−10.1	1078.11
5	Ac-c[CLDPETGEYLC]-OH	1.0 ± 0.02	10.4	−13.1	−2.2	−10.9	1282.38
6	Ac-CLDPETGEYLC-OH	1.0 ± 0.01	15.9	−14.4	−3.8	−10.6	1284.38
7	Ac-c[CDPETGEYC]-OH	1.0 ± 0.01	53.7	−17.6	−7.7	−9.9	1056.06
8	Ac-CDFETGEYC-OH	1.0 ± 0.02	69.9	−12.7	−2.9	−9.8	1058.06

---

Previously binding determinants analysis using validated Keap1 substrates has provided useful information about the Keap1 binding mechanism, e.g., that P1 and P2 mainly recognize the amino acid. And it has been successfully used to design a potent small molecule PPI inhibitor.¹⁴ In this study, we mainly focused on the outer sub-pockets, P4 and P5. Overall, the P4 and P5 sub-pockets are much less polar than the inner sub-pockets P1 and P2, but they both contain polar residues, for example, Tyr334 in the P5 sub-pocket. Moreover, the P5 sub-pocket is also adjacent to the positively charged residue, Arg336. To seek whether these residues can be further utilized to improve Keap1 binding, the optimized peptide 3 was used to carry out a MD simulation.

As shown in Fig. 2, the Phe of the peptide 3 was adjacent to the Arg336 during the MD simulation, and its phenyl ring can form a cation–pi interaction with the guanidino group of the Arg336. Considering the highly positively charged nature of the guanidino group, the tyrosine replacement of the corresponding phenylalanine may strengthen the interaction at this site, for it can be negatively charged. The MD simulation was also used to evaluate this hypothesis. The representative binding mode of the peptide 4 is shown in Fig. 2B. The Tyr replacement of the Phe resulted in significant changes of the binding interactions. The phenol group of the Tyr can form a hydrogen bond with the backbone carbonyl oxygen, and this hydrogen bond resulted in the opening conformation of the Arg336 side chain. This movement of the Arg336 side chain led to the exposure of the Tyr334 and allowed the interaction between the Tyr residue in the peptide and the Tyr334. These computational results were further evaluated by experimental results. The binding thermodynamics results indicated that peptide 4 showed a stronger binding with Keap1 than peptide 3. Tyrosine replacement of the corresponding phenylalanine significantly strengthened the interaction, resulting in a strong enthalpic component to binding, which could exceed the increase of unfavorable entropy. Next, we also examined the binding kinetics of the peptides using the biolayer interferometry (BLI) assay. Overall, the binding kinetics of the peptides 3 and 4 fitted the MD-based binding mode well. The peptide 4 showed no advantage in the k_on value (determined by BLI) partly because it induced a binding conformational change upon binding. In contrast, an approximately 12-fold lower k_off was observed with Ac-LDPETGEYL-OH. Thus, Ac-LDPETGEYL-OH binds more tightly to Keap1 than Ac-LDPETGEFL-OH possibly because of additional interactions of the Tyr residue with Keap1. Peptide 4 is also more potent than peptide 3 in the FP assay, consistent with the difference in Keap1 binding affinity. Briefly, the binding mode of peptide 4 gives a novel insight into the P5 sub-pocket and indicates the privileged fragment in Keap1–Nrf2 inhibitor design (Table 3).

Fig. 2 Model of the Keap1–peptide interaction. The hydrogen bonds are represented as green dashed lines and the electrostatic interactions are represented as yellow dashed lines. The carbon atoms of the Nrf2 peptide and Keap1 residues are colored as cyan and purple, respectively. The last frame was saved to carry out the analysis.

Table 3 Kinetics of the interactions of four peptides with Keap1 as determined by BLI

	Kd (nM)	kon (M^−1 s^−1)	kdis (s^−1)
Peptide 1	256.0	4.56 × 10^5	1.17 × 10^−1
Peptide 3	65.3	5.09 × 10^5	3.33 × 10^−2
Peptide 4	16.6	1.62 × 10^5	2.68 × 10^−3
Peptide 5	2.8	2.26 × 10^6	6.23 × 10^−3

---

On the basis of the crystal structure of the Keap1–Nrf2 ETGE peptide, the peptide has a stable type I β-turns conformation which is stabilized by the intramolecular hydrogen bonds formed by the peptide backbone and the side chains of D77 and T80. The finely tuned secondary structure of the peptide is critical for Keap1 binding. In order to further stabilize the appropriate conformation of the peptide, the terminal disulfide bridge was introduced to the original sequence. The cyclic peptide was designed based on the peptide 4, resulting in the cyclic peptide 5, Ac-c[CLDPETGEYLC]-OH. We also synthesized the peptide 6 with unbonded cysteine residues to fully confirm the binding conformation locking effects. Replacing terminal Leu residues with Cys residues resulted in peptides 7 and 8, which were used to investigate the position of the disulfide linkage. Firstly, we examined the PPI inhibition activity of these peptides. The results clearly showed that cyclic peptides can be more potent in disrupting the Nrf2–Keap1 PPI (9.4 nM for cyclic peptide 5, 29.6 nM for peptide 4 and 22.6 nM for peptide 6). Peptide 7 showed a significant decrease in PPI inhibition activity, further indicating that the terminal Leu residues play an important role in the hydrophobic interaction of the peptide with Keap1.

The cyclic peptide 5 was found to have a significant improvement in Keap1 binding affinity as shown by ITC analysis, with a K_d of 10.4 nM compared with the linear template peptide 4 which only gave a K_d of 42 nM and the non-cyclizing peptide 6 which gave a K_d of 15.9 nM. As shown in Table 2, the change in Gibbs free energy mainly originated from the entropy component. The decreased entropy component presumably resulted from the conformation locking effects caused by the terminal disulfide linkage. The BLI result also confirmed the binding conformation locking effects from the perspective of binding kinetics. The changes in the K_d value between peptide 4 and 5 mainly originated from the k_on. As depicted in Fig. 3 and Table 2, cyclic peptide 5, Ac-c[CLDPETGEYLC]-OH, bound to Keap1 with an approximately 14-fold faster k_on than Ac-LDPETGEYL-OH. In contrast, the k_off value showed much less of a difference. Thus, cyclic peptide 5 bound more tightly to Keap1 than peptide 4, presumably because of the conformational constraint induced by the terminal disulfide linkage.

Fig. 3 BLI dose–response association and dissociation curves reflecting direct binding to Keap1. BLI association and dissociation curves of peptide 1 , peptide 3, peptide 4 and peptide 5 using various concentrations. Concentrations: 333 nM, 111 nM, 37.0 nM, 12.3 nM.

Using a MD simulation method, we further investigated whether the terminal disulfide bridge could lock the precise conformation of the peptide and recognize the receptor without remarkable conformational changes. Indeed, during the MD simulation, peptide 5 showed a significantly smaller fluctuation of the binding motif compared with peptide 4 without the disulfide linkage (Fig. 4A). The more stable conformation of peptide 5 reduced the conformational changes upon binding which partly explained its good binding affinity. In order to further evaluate the terminal strain effects caused by the disulfide bridge, the distance between the terminal atoms (the nitrogen atom of the C-terminal Leu residue and the oxygen atom of the N-terminal acetyl group) was monitored along the trajectory. The disulfide linkage significantly reduced both the value and the fluctuation of the distance (Fig. 4B). These modelling studies proved that the terminal disulfide bridge can lock the peptide conformation. The cyclic peptide study confirmed that the appropriate and stable conformation of the ligand is quite important for Keap1 binding and it gave a good implication for potent Keap1–Nrf2 PPI inhibitor design.

Fig. 4 Effects of the terminal disulfide linkage on peptide conformation. (A) Atomic fluctuations (RMSF) per residue during MD trajectories of the peptide. (B) Terminal atom distance along the MD trajectory.

Conclusions

In this study, terminal hydrophobic residues and a Tyr residue replacement are introduced into these novel peptides. The experimental values of these peptides further confirm the potential binding sites explored by the MD simulation. The peptide 5, which represents the first intracyclic constrained Keap1–Nrf2 PPI inhibitor to date, proved that the active conformation locking strategy is quite useful in Keap1–Nrf2 PPI inhibitor design. It also serves as a useful probe to investigate the Keap–Nrf2 PPI system. The biological data, together with the computational results, indicated that the introduction of a terminal disulfide linkage can stabilize the peptide β-turn conformation which is important for Keap1 binding. Moreover, these potent peptides could be used as the starting points and excellent templates for the generation of potent PPI inhibitors for further target validation of Keap1–Nrf2 PPI.

Acknowledgements

This work is supported by the project 81230078 (key program), 81202463 (youth foundation), 81173087 and 91129732 of National Natural Science Foundation of China, 2014ZX09507002-005-015, 2013ZX09402102-001-005 and 2010ZX09401-401 of the National Major Science and Technology Project of China (Innovation and Development of New Drugs), Specialized Research Fund for the Doctoral Program of Higher Education (SRFDP, 20130096110002).

Notes and references

1. T. Suzuki, H. Motohashi and M. Yamamoto, Trends Pharmacol. Sci., 2013, 34, 340–346.
2. J. D. Hayes and A. T. Dinkova-Kostova, Trends Biochem. Sci., 2014, 39, 199–218.
3. S. Magesh, Y. Chen and L. Hu, Med. Res. Rev., 2012, 32, 687–726.
4. M. Mazzuferi, G. Kumar, J. van Eyll, B. Danis, P. Foerch and R. M. Kaminski, Ann. Neurol., 2013, 74, 560–568.
5. L. Baird and A. T. Dinkova-Kostova, Biochem. Biophys. Res. Commun., 2013, 433, 58–65.
6. L. Baird, D. Lleres, S. Swift and A. T. Dinkova-Kostova, Proc. Natl. Acad. Sci. U. S. A., 2013, 110, 15259–15264.
7. L. Baird, S. Swift, D. Lleres and A. T. Dinkova-Kostova, Biotechnol. Adv., 2014, 32, 1133–1144.
8. S. K. Niture, R. Khatri and A. K. Jaiswal, Free Radical Biol. Med., 2014, 66, 36–44.
9. A. J. Wilson, J. K. Kerns, J. F. Callahan and C. J. Moody, J. Med. Chem., 2013, 56, 7463–7476.
10. C. Zhuang, Z. Miao, C. Sheng and W. Zhang, Curr. Med. Chem., 2014, 21, 1861–1870.
11. D. Marcotte, W. Zeng, J.-C. Hus, A. McKenzie, C. Hession, P. Jin, C. Bergeron, A. Lugovskoy, I. Enyedy, H. Cuervo, D. Wang, C. Atmanene, D. Roecklin, M. Vecchi, V. Vivat, J. Kraemer, D. Winkler, V. Hong, J. Chao, M. Lukashev and L. Silvian, Bioorg. Med. Chem., 2013, 21, 4011–4019.
12. L. Hu, S. Magesh, L. Chen, L. Wang, T. A. Lewis, Y. Chen, C. Khodier, D. Inoyama, L. J. Beamer, T. J. Emge, J. Shen, J. E. Kerrigan, A. N. Kong, S. Dandapani, M. Palmer, S. L. Schreiber and B. Munoz, Bioorg. Med. Chem. Lett., 2013, 23, 3039–3043.
13. C. Zhuang, S. Narayanapillai, W. Zhang, Y. Y. Sham and C. Xing, J. Med. Chem., 2014, 57, 1121–1126.
14. Z.-Y. Jiang, M.-C. Lu, L. L. Xu, T.-T. Yang, M.-Y. Xi, X.-L. Xu, X.-K. Guo, X.-J. Zhang, Q.-D. You and H.-P. Sun, J. Med. Chem., 2014, 57, 2736–2745.
15. H.-P. Sun, Z.-Y. Jiang, M.-Y. Zhang, M.-C. Lu, T.-T. Yang, Y. Pan, H.-Z. Huang, X.-J. Zhang and Q.-d. You, MedChemComm, 2014, 5, 93–98.
16. Z.-Y. Jiang, L.-L. Xu, M.-C. Lu, Y. Pan, H.-Z. Huang, X.-J. Zhang, H.-P. Sun and Q.-D. You, J. Comput.-Aided Mol. Des., 2014, 28, 1233–1245.
17. D. Inoyama, Y. Chen, X. Huang, L. J. Beamer, A. N. Kong and L. Hu, J. Biomol. Screening, 2012, 17, 435–447.
18. R. Steel, J. Cowan, E. Payerne, M. A. O’Connell and M. Searcey, ACS Med. Chem. Lett., 2012, 3, 407–410.
19. R. Hancock, H. C. Bertrand, T. Tsujita, S. Naz, A. El-Bakry, J. Laoruchupong, J. D. Hayes and G. Wells, Free Radical Biol. Med., 2012, 52, 444–451.
20. S.-C. Lo, X. Li, M. T. Henzl, L. J. Beamer and M. Hannink, EMBO J., 2006, 25, 3605–3617.
21. K. I. Tong, B. Padmanabhan, A. Kobayashi, C. Shang, Y. Hirotsu, S. Yokoyama and M. Yamamoto, Mol. Cell. Biol., 2007, 27, 7511–7521.
22. T. Fukutomi, K. Takagi, T. Mizushima, N. Ohuchi and M. Yamamoto, Mol. Cell. Biol., 2014, 34, 832–846.
23. B. Padmanabhan, K. I. Tong, T. Ohta, Y. Nakamura, M. Scharlock, M. Ohtsuji, M.-I. Kang, A. Kobayashi, S. Yokoyama and M. Yamamoto, Mol. Cell, 2006, 21, 689–700.
24. Y. Che and G. R. Marshall, Expert Opin. Ther. Targets, 2008, 12, 101–114.
25. Z.-Y. Jiang, H.-X. Chu, M.-Y. Xi, T.-T. Yang, J.-M. Jia, J.-J. Huang, X.-K. Guo, X.-J. Zhang, Q.-D. You and H.-P. Sun, PLoS One, 2013, 8, e75076.
26. M. Citartan, S. C. Gopinath, J. Tominaga and T. H. Tang, Analyst, 2013, 138, 3576–3592.

---

Footnote

† Electronic supplementary information (ESI) available: Detailed procedures of the computational experiments, detailed information of the peptide synthesis, LC-MS profiles of the synthesized peptides and extra information for the biology test. See DOI: 10.1039/c5ra16262a

---