Identification of a novel small-molecule Keap1–Nrf2 PPI inhibitor with cytoprotective effects on LPS-induced cardiomyopathy

Abstract A new Keap1–Nrf2 protein–protein interaction (PPI) inhibitor ZJ01 was identified from our compound library by fluorescence polarization assay, surface plasmon resonance, molecular docking and molecular dynamics simulation. ZJ01 could in vitro trigger Nrf2 nuclear translocation, subsequently resulting in increased mRNA levels of Nrf2 target genes HO-1 and NQO1. Meanwhile, ZJ01 suppressed LPS-induced production of ROS and the mRNA levels of pro-inflammatory cytokines TNF-α, IL-1β and IL-6 in H9c2 cardiac cells. Moreover, in an in vivo mouse model of septic cardiomyopathy induced by intraperitoneal injection of lipopolysaccharide, ZJ01 demonstrated a cytoprotective effect, upregulated Nrf2 protein nuclear accumulation, and remarkably suppressed the abovementioned cytokine levels in cardiomyocytes. The results presented herein provided a novel chemotype for the development of direct Keap1–Nrf2 PPI inhibitors and suggested that compound ZJ01 is a promising drug lead for septic cardiomyopathy treatment. ZJ01 was identified as a new Keap1–Nrf2 PPI inhibitor and drug lead for septic cardiomyopathy treatment by in vitro and in vivo experiments.


Introduction
Sepsis, a major cause of death worldwide, is a systemic inflammatory syndrome caused by serious infections 1 . According to the statistics, up to 70% of patients with sepsis are affected by septic cardiomyopathy, and the dysfunction of cardiomyocytes contributes to a high mortality rate 2 . Currently, septic cardiomyopathy remains a great challenge for clinical treatment. Lipopolysaccharide (LPS), the major component of the outer membrane of gram-negative bacteria, has proven to be a key stimulator of sepsis 3 . Under septic conditions, LPS can evoke an inflammatory response in innate immune cells, as well as non-immune cells (e.g. cardiomyocytes) expressing LPS-pattern recognition receptors 4 . More importantly, cardiac dysfunction frequently occurs in patients with sepsis and animals injected with LPS 5,6 . Increasing evidence indicates that control of the oxidative stress and inflammatory response in cardiomyocytes could alleviate LPS-induced cardiac injury [7][8][9] .
Nrf2 activity is regulated mainly by Keap1 14 . Under unstressed conditions, Nrf2 is negatively regulated by Keap1 via proteasomal degradation to remain at low cellular concentrations. Upon oxidative stress, reactive oxygen species (ROS) and/or electrophilic agents modify the reactive cysteine residues in Keap1, resulting in conformational changes and deactivation of Keap1. Subsequently, Nrf2 is activated after escaping from Keap1-mediated degradation and then translocate into the nucleus where it induces the transcription of the aforementioned downstream genes by binding to ARE 14 . More interestingly, a recent study demonstrated that Nrf2 could inhibit pro-inflammatory cytokine transcription through direct DNA binding 15 . Interestingly, other studies have highlighted the important role of Nrf2 activation in protecting cardiomyocytes from LPS-induced injuries 16 . Thus, to activate Nrf2 via disrupting the Keap1-Nrf2 protein-protein interaction (PPI) represents an attractive strategy to develop drugs for diseases involving oxidative stress and inflammation, such as cardiomyopathy 16,17 .
Recently, small molecules ( Figure 1) directly targeting the Keap1-Nrf2 PPI have been reported by several research groups [18][19][20][21][22] . However, the structural diversity of current Keap1-Nrf2 PPI inhibitors is still limited. In our project to discover Keap1-Nrf2 PPI inhibitors, the novel small molecule ZJ01 with an iminocoumarinbenzothiazole scaffold ( Figure 2) was recently screened out from our in-house compound library based on fluorescence polarization (FP) assay. In the present study, we describe the identification of ZJ01 as a new Keap1-Nrf2 PPI inhibitor and further investigate its in vitro and in vivo cytoprotective effects on LPS-induced cardiomyopathy.

General experimental procedures
Commercially available reagents were used without further purification. Organic solvents were evaporated with reduced pressure using a Buchi rotary evaporator. Reactions were monitored by TCL using Yantai Jiangyou (China) GF254 silica gel plates. Silica gel column chromatography was performed on silica gel (300-400 mesh) from Qingdao Haiyang (China). The NMR spectra were measured on Bruker Avance 600 spectrometer. Chemical shifts were expressed in d (ppm) and coupling constants (J) in Hz using solvent signals as internal standards (CDCl 3 , d H 7.26 ppm; d C 77.0 ppm). ESI-MS was recorded on an Agilent 6460 Triple Quard LC/MS, and HRMS spectrum were recorded on an Agilent Q-ToF 6520.

FP assay
Determination of equilibrium dissociation constants K D2 for the tested compound was performed using the FP assay by fitting the displacement curves as described in our previous study 20 . The FP was measured on Biotek Syngergy 2 microplate reader with 485 nm excitation and 535 nm emission after a 60 min incubation at room temperature. The K D1 value is determined with FITC À bAla À DEETGEF À OH as the fluorescence probe.

Surface plasmon resonance (SPR) analysis
Surface plasmon resonance analysis was performed on a Biacore TM T200 instrument (GE Healthcare, Salem, CT, USA). Keap1 proteins were immobilized on a CM5 chip via EDC/NHS-mediated crosslinking reaction. Tested compound was diluted in PBS at concentrations ranging from 0.25 to 64 mM. The analysis was then performed according to the protocol provided by GE Healthcare. In each analysis, the middle concentration was duplicated at the end of the wash run to confirm the stability of the sensor surface. The parameters of SPR were set as follow: flow rate, 30 mL/min; contact time, 120 s; disassociation time, 120 s. Affinity curve fitting was performed with the Biacore T200 Evaluation Software using a steadystate affinity model (1:1) to calculate the disassociation constant (K d ). Detailed SPR measurement data are shown in Table 1.

Cell culture and treatment
H9c2 cells, a rat ventricular myoblast cell line, were purchased from the Cell Bank of Chinese Academy of Sciences (Shanghai, China). The Cells were cultured in Dulbecco's modified Eagle's medium with 10% bovine calf serum and then were maintained at 37 C under humidified conditions with 5% CO 2 . Cells were seeded in appropriate dishes (30,000 cells/ml). After culture for 24 h, the cells were treated with different concentration of inhibitors and incubated with or without 1 lg/ml of LPS for indicated time.

Mice and treatments
Male C57BL/6 mice (eight weeks old) were purchased from the Experimental Animal Center of Shandong University (Jinan, Shandong, China). All in vivo experiments followed the ARRIVE guidelines 24 . Inhibitors or LPS was dissolved in DMSO: normal saline (1:100). The control group was injected intraperitoneally with equal DMSO and normal saline. C57BL/6 mice were challenged with different concentrations of ZJ01 or S47 overnight for approximately 12 h after being treated intraperitoneally with or without 4 mg/kg of LPS. At the end of treatment, all mice were euthanized by intravenous lateral tail vein injection of keta-mine/ xylazine (Sigma-Aldrich, Saint Louis, MO, USA, 150 mg/kg ketamine combined with 10 mg/kg xylazine). The left ventricles were collected for western blotting or real-time PCR assay.
Western blotting assay Protein Extraction Kit (Beyotime, China) was used to isolate the nuclear and cytosol protein of H9c2 cells and left ventricular cells of C57BL/6 mice according to the protocol. Then the collected protein was stored at À80 C until use. Equal amounts of protein were applied to 12% SDS-polyacrylamide gel. Proteins in gels were electroblotted onto poly-vinylidene difluoride membranes. After blocking at room temperature for 1 h, the membranes were probed with primary antibodies overnight at 4 C. After three washes in TBST, membranes were incubated with peroxidase-conjugated secondary antibodies for 1 h at room temperature, and proteins were detected by use of an enhanced chemiluminesence detection kit.

Molecular dynamics simulations
Molecular dynamics simulations were performed on the Keap1/ ZJ01 complex obtained from molecular docking. Before simulations, the protonation states of ionisable residues were determined using the Hþþ program 31 . The complex model was surrounded by a periodic box of transferable intermolecular potential 3P 32 water molecules that extend 12 Å from the protein atoms.
Counter-ions were added to neutralize the simulation system. Molecular dynamics simulations were performed using the GROMACS 5.0.4 package 33 with isothermal-isobaric (NPT) ensemble and periodic boundary condition. The CHARMM36-CAMP force field 34 was used for the protein, ions and water molecules. Atom charges of ZJ01 were calculated using the restrained electrostatic potential method 35 encoded in the AMBER suite of programs 36 at the RHF/6-31G Ã level. Covalent and non-bonded parameters for ZJ01 atoms were assigned by analogy or interpolation from those already present in the AMBER force field. Energy minimizations were first performed to relieve unfavourable contacts in the system, followed by equilibration steps of 1 ns in total to equilibrate the complex. Subsequently, 50 ns production runs were performed. The temperature of the system was maintained at 300 K using the v-rescale method 37 with a coupling time of 0.1 ps. The pressure was kept at 1 bar using the Parrinello-Rahman 38 with s p ¼ 1.0 ps and a compressibility of 4.5 Â 10 À5 bar À1 . SETTLE 39 constraints and LINCS 40 constraints were applied to the hydrogen-involved covalent bonds in water molecules and in other molecules, respectively, and the time step was set to 2 fs. Electrostatic interactions were calculated with the particle-mesh Ewald (PME) algorithm 41 with a real-space cutoff of 1.2 nm.

Binding free energy calculation by MMPBSA method
Based on the equilibrated dynamic trajectory, the binding free energy of the complex was calculated using the MM-PBSA method encoded in the AMBER 14.0 program. A total of 2500 snapshots from the trajectory were extracted every 20 ps, and the MM-PBSA calculation was performed on each snapshot using the MMPBSA.py module in AMBER 14.0.

Hits identification and confirmation
To identify Keap1-Nrf2 PPI inhibitors, our in-house library of 569 compounds was screened for their activity (K D2 , K D1 is the equilibrium dissociation constant of the fluorescent probe) in FP assay 20 at the concentration of 100 lM, with the previously identified Keap1-Nrf2 PPI inhibitor S47 20 as a positive control. The assay results revealed synthetic compound ZJ01 to have the best inhibitory activity, demonstrating 81.25% inhibition of Keap1-Nrf2 PPI interaction at 100 lM. However, an analogue (ZJ02, Scheme 1) with the same core structure displayed only 20% inhibition in the initial testing, indicating that the side chain play a vital role in their activity. Compound ZJ01 was then retested to determine its K D2 , which was found to be 5.1 lM (Figure 3(A)), slightly better than that of S47 (K D2 % 7.6 lM, close to the reported 2.9 lM) 20 . Later, a surface plasmon resonance (SPR) competition assay 42 was employed to confirm the interaction between ZJ01 and the Keap1 Kelch domain. As shown in Figure 3(B), ZJ01 displayed an extracellular binding affinity with Keap1 with an equilibrium dissociation constant (K d ) of 48.1 lM, somewhat weaker than that of S47 (K d % 39.4 lM). The equilibrium binding curve fits are shown in Figures S14 and S15 (ESI †). These results indicated that ZJ01 can directly bind to Keap1 and disrupt the Keap1-Nrf2 interaction.

Molecular modelling study
To investigate the possible binding mode of ZJ01 to the Keap1 Kelch domain, molecular docking was conducted using Autodock-4 25 . The binding mode of ZJ01 to Keap1 is shown in Figure 4. A 50 ns molecular dynamics (MD) study, discussed in the following section, did not indicate any instability of the docked Keap1/ZJ01 complex. Similar docking was also performed for Keap1/ZJ02. Compared with ZJ01, the result ( Figure S16, ESI †) indicated that ZJ02 had a totally different spatial orientation at the Keap1-Nrf2 interaction surface, which possibly resulted from the protonated nitrogen atom of amide in ZJ02, which remained unprotonated in ZJ01. This was speculated to be the reason that these two compounds exhibited different activities in the FP assay.

Steady RMSDs values indicate that the MD trajectory is reliable
To explore the dynamic traits of the Keap1/ZJ01 complex, a 50 ns MD simulation was performed on the complex model obtained by molecular docking. To examine the structural stability of the complex during MD simulations, the time evolution of weighted rootmean-square deviations (RMSDs) for backbone atoms of the Keap1 protein and heavy atoms of ZJ01 from their initial positions (t ¼ 0) was calculated. As illustrated in Figure 5, RMSD values of the protein backbone were found to be between 0.5 and 1.2 Å during simulations. Steady RMSD values for the heavy atoms of the protein and ZJ01 indicated well-equilibrated states of the system. The steady RMSD values showed the reliability of the MD trajectories and suggested the suitability for post analysis.

H-bond and hydrophobic interaction analyses revealed a well-defined substrate pocket
To probe the molecular interactions between Keap1 and ZJ01, the H-bond and hydrophobic interactions of the complex were analysed based on MD trajectories. As shown in Table 2, there is only one conserved H-bond between Keap1 and ZJ01 in the complex. Additionally, the non-polar contacts between active site residues and ZJ01 cannot be ignored due to the fact that Keap1 has a relatively hydrophobic pocket. Thus, the occupancy rates of hydrophobic interactions between active site residues and ZJ01 were calculated based on MD trajectories. The result demonstrated that residues Tyr334, Arg415, Tyr525, Ala556, Tyr572, and Ser602 have more intense hydrophobic interactions with ZJ01 (Table 3). Taken together, these results indicated extensive polar and non-polar interactions between Keap1 and ZJ01.

Binding free energy analysis
To further explore the molecular interactions between Keap1 and ZJ01, the binding free energy values of the Keap1/ZJ01 complex were calculated using the MM-PBSA method encoded in the AMBER 14 program (in Table 4) 43 . The binding free energy (DG) value of -26.10 kcal/mol suggested a stable binding between ZJ01 and Keap1.

Nrf2-ARE pathway activation in H9c2 cardiac cells
The above results demonstrated that ZJ01 was an effective direct Keap1-Nrf2 PPI inhibitor. Previous studies have confirmed that Keap1-Nrf2 PPI inhibitors can make Nrf2 detach from Keap1 and subsequently translocate to the nucleus to activate the anti-oxidant defense system 20,21 . Therefore, the effect of ZJ01 on the distribution of Nrf2 in H9c2 cardiac cells was investigated. Western blotting analysis indicated that ZJ01 increased the nuclear Nrf2 protein levels and decreased the non-nuclear Nrf2 protein levels at 8 lM (Figure 6(A)). Meanwhile, the immunofluorescence assay demonstrated that Nrf2 translocated into the nucleus after the cells were treated with ZJ01 ( Figure 6(B)) or S47 ( Figure S17).
The nuclear Nrf2 could bind to the ARE of target genes, which then resulted in transcriptional induction and chemoprotection effects. Thus, the mRNA levels of Nrf2 target anti-oxidant genes, such as HO-1 and NQO1, were further examined in H9C2 cells after exposure to ZJ01. It was found that ZJ01 could significantly increase the HO-1 and NQO1 mRNA levels (Figure 7(A)). Altogether, these results revealed that ZJ01 activated the Nrf2-ARE pathway in H9c2 cardiac cells and might be a protective agent in H9c2 cardiac cells.
Anti-inflammatory response and anti-oxidative stress ability assay in H9c2 cardiac cells. Oxidative stress and inflammatory responses are responsible for the deterioration of sepsis-induced cardiomyopathy. LPS has been used to induce an inflammatory response and oxidative stress in H9c2 cells, which are typically used for in vitro investigations 4,44 . Inflammatory cytokines, including tumour necrosis factor (TNF)-a, interleukin (IL)-1a, IL-1b, and IL-6, promote the development of sepsis-induced cardiomyopathy 45 . Previously published studies indicated that LPS could increase the production of IL-1b, IL-6, and TNF-a in H9c2 cardiomyocytes and primary cardiomyocytes 4,46 , and additional evidence postulated that LPSinduced ROS could lead to the activation of intracellular signalling pathways and transcription factors, and subsequently induce the production of inflammatory mediators, including TNF-a, IL-1b and IL-6 7,47-49 . Therefore, inhibiting oxidative stress and blocking inflammatory signalling may produce beneficial effects in the dysfunctional heart. The effects of ZJ01 on the production of TNF-a, IL-1b, IL-6 and ROS were thus examined in LPS-stimulated H9c2 cells as well as the in vitro model. Consistent with previously published data from other groups, our results showed that 1 lg/ml LPS induced a significant increase in mRNA levels of TNF-a, IL-1b and IL-6 in H9c2 cardiac cells. The 8 lM volume of ZJ01 or S47, which could effectively activate Nrf2, was applied to test its inhibitory effect on the production of TNF-a, IL-1b and IL-6 expression after exposure to LPS. It was observed that ZJ01 and S47 markedly alleviated LPS-induced production of these pro-inflammatory cytokines at the mRNA level (Figure 7(B); Figure S18A). Intracellular ROS levels were measured by dichlorodihydro-fluorescein diacetate (DCFH-DA). The results showed that the ROS levels were significantly upregulated after treatment with 1 lg/ml LPS, and this increase could be blocked by 8 lM ZJ01 (Figure 7(C)). However, 8 lM S47 could not inhibit the elevated ROS levels induced by LPS ( Figure S18B). The above results revealed that ZJ01 and S47 had the same anti-inflammatory effect, while ZJ01 had better anti-oxidant effects on LPS-treated myocardial cells. As ZJ01 is an effective Keap1-Nrf2 PPI inhibitor, we speculated ZJ01 might inhibit LPS-induced cardiomyopathy through disrupting      Keap1-Nrf2 PPI which could induce Nrf2 nuclear accumulation and thus blocking oxidative and inflammatory signalling.
In vivo studies. As left ventricular inflammation is an important feature of cardiac injury induced by LPS 16 , to examine whether ZJ01 and S47 were able to confer a cardiac protective effect in vivo, C57BL/6 mice were challenged with different concentrations of ZJ01 or S47 overnight for approximately 12 h after being treated intraperitoneally with or without 4 mg/kg of LPS, a dose commonly used for experimental models of septic cardiomyopathy in animals 16 . Then, the left ventricles were collected for further studies. Consistent with the in vitro experimental results, western blotting assay demonstrated that ZJ01 increased the nuclear Nrf2 protein levels and decreased the non-nuclear Nrf2 protein levels at 5 mg/kg or 10 mg/kg concentration suggesting ZJ01 could induce Nrf2 nuclear accumulation in left ventricular cells (Figure 8(A)). Real-time PCR results showed that inhibit LPS-induced inflammatory cytokines including TNF-a, IL-1b and IL-6 at the mRNA level in in left ventricular cells in vivo (Figure 8(B)). Compared to ZJ01, S47 could induce the Nrf2 nuclear accumulation at a 10 mg/kg concentration ( Figure S19), but did not exhibit obvious effects on Nrf2 nuclear accumulation at the lower 5 mg/kg concentration. This result suggested that ZJ01 might be more effective than S47 in activating Nrf2 of cardiomyocytes in vivo.

Conclusions
In this study, a new Keap1-Nrf2 PPI inhibitor ZJ01 with an iminocoumarin-benzothiazole core was identified via FP and SPR assays. The amide side chain of ZJ01 was found to have an important role in its activity. ZJ01 could inhibit LPS-induced production of pro-inflammatory cytokines (TNF-a, IL-1b and IL-6) and ROS through activating the Nrf2-ARE pathway in H9c2 cells. In vivo, ZJ01 also induced Nrf2 nuclear accumulation and inhibited LPSinduced inflammatory cytokine production in cardiomyocytes. The biological data, together with computational results, indicated that ZJ01 could serve as a new template for generating direct Keap1-Nrf2 PPI inhibitors and possessed great potential in septic cardiomyopathy therapy.

Disclosure statement
No potential conflict of interest was reported by the authors.   The nuclear and non-nuclear Nrf2 protein levels of left ventricular cells were determined by Western blotting technology. (B) C57BL/6 mice were treated intraperitoneally with 4 mg/kg of LPS and then with 5 mg/kg ZJ01 for 12 h. RT-PCR analysis of the expression of pro-inflammatory cytokines IL-1b, IL-6, and TNF-a. ÃÃ p < .01 vs. control, #p < .05 vs. LPS group. n ¼ 3.