A hydrogen peroxide responsive prodrug of Keap1-Nrf2 inhibitor for improving oral absorption and selective activation in inflammatory conditions

Abstract

Transcription factor Nuclear factor erythroid 2-related factor 2 (Nrf2) and its negative regulator, the E3 ligase adaptor Kelch-like ECH-associated protein 1 (Keap1), control the redox and metabolic homeostasis and oxidative stress. Inhibitors of Keap1-Nrf2 interaction are promising in oxidative stress related inflammatory diseases but now hit hurdles. By utilizing thiazolidinone moiety to shield the key carboxyl pharmacophore in Keap1-Nrf2 inhibitor, a hydrogen peroxide (H₂O₂)-responsive prodrug pro2 was developed. The prodrug modification improved the physicochemical properties and cell membrane permeability of the parent drug. Pro2 was stable and stayed inactive under various physiological conditions, while became active by stimulation of H₂O₂ or inflammation derived reactive oxygen species. Moreover, pro2 exhibited proper pharmacokinetic profile suitable for oral administration and enhanced anti-inflammatory efficiency in vivo. Thus, this novel prodrug approach may not only provide an important advance in the therapy of chronic inflammatory diseases with high level of H₂O₂, but also offer a fresh solution to improve the drug-like and selectivity issues of Keap1-Nrf2 inhibitors.

Graphical abstract

Highlights

• •Pro2 was developed by utilizing H₂O₂-responsive thiazolidinone moiety to shield carboxyl group in Keap1-Nrf2 inhibitor.
• •Pro2 was stable and inactive under various physiological conditions, while became active under inflammatory conditions.
• •Pro2 exhibited proper pharmacokinetic profile for oral administration and enhanced anti-inflammatory efficiency in vivo.

1. Introduction

Transcription factor nuclear factor erythroid 2-related factor 2 (Nrf2), as the master regulator of multiple cytoprotective responses, is pivotal in redox and metabolic homeostasis, as well as the regulation of oxidative stress [1,2]. Nrf2 activation enhances the anti-oxidant capacity and provides cryoprotection against oxidative stress and inflammatory insults [3,4]. Recently, therapeutic targeting of protein-protein interaction (PPI) of Nrf2 and its principal negative regulator, the E3 ligase adaptor Kelch-like ECH-associated protein 1 (Keap1) is emerging as a new strategy for drug development of chronic diseases [[5], [6], [7], [8], [9], [10]]. Several types of PPI inhibitors with high potency in disrupting Keap1-Nrf2 interaction have been reported by both pharmaceutical companies and academic institutes [[11], [12], [13], [14], [15], [16], [17]]. However, concerns regarding pharmacokinetic properties and safety still remain. Most of the currently reported inhibitors are molecules with fairly high molecule weight and several polar functional groups [[18], [19], [20]], requirements for blocking the large and high polar Keap1-Nrf2 interface, and thus these compounds with high potency in vitro at present exhibit poor absorption, distribution, metabolism and excretion properties and relative low efficacy in vivo. Safety issues also challenge the further development of Keap1-Nrf2 inhibitors, since Nrf2 activation in normal cells can excessively strengthen the antioxidant system, resulting in the clearance of reactive oxygen species (ROS), which may disturb the pathological function of ROS [21]. Moreover, a concern about the increased cancer risk of Nrf2 activators also appears [[22], [23], [24], [25]]. Somatic mutations in KEAP1 and NFE2L2, resulting in high and unrestrained Nrf2 activity, have been regarded as driving factors of several tumors [[26], [27], [28]]. Therefore, the selective inhibition of Keap1-Nrf2 PPI in oxidative-stressed tissue is a key challenge, as it would dramatically benefit the therapeutic application. Designing a prodrug of Keap1-Nrf2 inhibitor that can reveal carboxyl acid group upon activation by ROS could kill two birds with one stone. It can not only improve the poor pharmacokinetic properties caused by the polar and ionizable characters of carboxyl acid group, but also shield the key group of Keap1-binding until reaching the target cells, specifically rebalancing the redox state in pathological cells while not affecting normal cells.

Hydrogen peroxide (H₂O₂), an uncharged molecule and a stable ROS, is endogenously produced and ubiquitous existed in living organisms [29,30]. Physiologically, H₂O₂ plays an active role in redox signaling through reversible redox post-translational modifications, and its level is fine-tuned by anti-oxidant system [[31], [32], [33]]. Nevertheless, high level of H₂O₂ has been closely associated with several pathological conditions, including inflammation [34], neurodegenerative disorders [35] and cancer [36]. On one hand, H₂O₂ overabundance, together with the aberrant oxidative stress, contributes to the pathology of these conditions. On the other hand, excessive level of H₂O₂ in pathological microenvironments could be an ideal trigger for targeting activation of therapeutic agents [37]. Chang's group reported the pioneer work of H₂O₂-responsive aryl boronate trigger [38,39], inspiring the rapid development of aryl boronate-based probes of H₂O₂ [[40], [41], [42], [43], [44], [45]]. Using aryboronates or boronic acids as the trigger units, Peng's group developed the first H₂O₂-activated DNA cross-linking agents [46], and various H₂O₂-activated anti-cancer agents have been identified [[47], [48], [49], [50], [51], [52], [53]].

More recently, increasing evidence from different studies supports the relation between oxidative stress and the pathogenesis of inflammation, and it inspired the research around the H₂O₂-responsive cytoprotective and anti-inflammation agents, including prodrugs of methotrexate and aminopterin for the treatment of rheumatoid arthritis [54,55], prodrugs of H₂S donors [56,57] and CO donors [58], and prodrug of angiogenin for neuroprotective activity [59]. The presence of high level of H₂O₂ in inflammatory conditions also provides an ideal trigger for the selective activation of Nrf2. To the best of our knowledge, an oral H₂O₂-responsive prodrug for Nrf2 activation is still unavailable, which limits therapeutic usage of Nrf2 activation agents in chronic inflammatory diseases. It is therefore envisioned that improvements could be made to develop new orally administered H₂O₂-responsive prodrug of Keap1-Nrf2 PPI inhibitors.

Here, we report a H₂O₂-reponsive prodrug of Keap1-Nrf2 inhibitor by shielding the key carboxyl pharmacophore with the H₂O₂-sensitive thiazolidinone moiety. Our study showed that prodrug modification of the parent drug can improve the drug-like properties and enable the molecule suitable for oral administration. We proved that the caged inhibitor lost its activity in disrupting Keap1-Nrf2 interaction and inflammation derived ROS allowed for the release of the active entity to antagonize the inflammatory conditions in both cellular and in vivo inflammatory models. To the best of our knowledge, it is the first example of H₂O₂-responsive prodrug suitable for oral administration, and this study highly stresses the in vivo anti-inflammation efficacy of small molecule Keap1-Nrf2 inhibitory agents.

2. Materials and methods

2.1. Chemistry

The synthesis of prodrugs is highlighted in Scheme 1. All chemicals purchased from commercial suppliers were used as received unless otherwise stated. All solvents were reagent grade and, when necessary, were purified and dried by standard methods. Reactions were monitored by thin-layer chromatography on silica gel plates (GF-254) visualized under UV light. Melting points were determined on a Mel-TEMP II melting point apparatus without correction. ¹H NMR and ¹³CNMR spectra were recorded in CDCl₃ or DMSO-d₆ on a Bruker Avance-300 instrument. Chemical shifts (δ) are reported in parts per million (ppm) from tetramethyl silane (TMS) using the residual solvent resonance (CDCl₃: 7.26 ppm for ¹H NMR, 77.16 ppm for ¹³C NMR; DMSO: 2.5 ppm for ¹H NMR, 39.5 ppm for ¹³C NMR). Multiplicities are abbreviated as follows: s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet). HR-MS spectra were recorded on a Water Q-Tof micro mass spectrometer. Flash column chromatography was performed with 100–200 mesh silica gel and yields refer to chromatographically and spectroscopically pure compounds. The purity (≥95%) of the compounds was verified by the HPLC study performed on an Agilent C18 (4.6 mm × 150 mm, 3.5 μm) column using a mixture of solvent methanol/water at a flow rate of 0.5 mL/min and monitored by UV absorption at 254 nm. The petroleum ether used in the study was the grade of analytical pure and boiled over the range 60–90 °C.

Scheme 1.

Reagents and conditions: (a) NH₂OH·HCl, 95% ethanol, MeOH, 60 °C, 2 h, yield 57%; (b) Pd/C, H₂, THF, rt, 4 h; (c) 4-Methoxybenzenesulfonyl chloride, Na₂CO₃, THF, 0 °C, 2 h, yield 83%; (d) 4-Methoxybenzenesulfonyl chloride, toluene, pyridine, 100 °C, 2 h, yield 64%; (e) DMF, K₂CO₃, ethyl bromoacetate, rt, 3 h, yield 84%; (f) LiOH, MeOH, H₂O, 2 h, yield 94%; (g) thiazolidin-2-one, DMAP, DCC, DMF, rt, 6 h, yield 76%.

N,N'-(naphthalene-1,4-diyl)bis(4-methoxy-N-(2-oxo-2-(2-oxothiazolidin-3-yl)ethyl) benzenesulfonamide) (pro1) To a solution of 1 (0.5 g, 0.81 mmol) in DMF (10 mL) was added DCC (0.37 g, 1.79 mmol) and DMAP (0.22 g, 1.79 mmol), stirring at room temperature. After 30 min, thiazolidin-2-one (0.37 g, 3.60 mmol) was added and the mixture was stirred overnight. After reaction finished, the mixture was poured onto water and extracted with Et₂O (3 × 20 mL). The organic extracts were combined, dried over Na₂SO₄, and concentrated in vacuo. The crude product was purified by column chromatography to give the pure product as a white solid (0.368 g, 58%). R_f = 0.42 (EA/PE 1:1); m.p. 234–235 °C; ¹H NMR (300 MHz, DMSO-d₆) δ 8.28 (dd, J = 6.9, 3.4 Hz, 2H; Ar–H), 7.57 (d, J = 8.5 Hz, 4H; Ar–H), 7.10 (d, J = 8.8 Hz, 4H; Ar–H), 7.03 (d, J = 8.5 Hz, 4H; Ar–H), 4.23 (s, 4H; CH₂), 3.87 (s, 6H; OCH₃), 3.81 (s, 4H; CH₂). 3.01–3.06 (t, J = 7.1 Hz, 4H; CH₂); ¹³C NMR (75 MHz, Chloroform-d) δ 173.57, 168.62, 163.32, 134.73, 132.71, 130.29, 128.91, 127.26, 124.77, 121.74, 114.44, 56.89, 55.81, 46.72, 26.03; HRMS (ESI): m/z calcd. for C₃₄H₃₂N₄O₁₀S₄+H⁺: 785.1074 [M+H]⁺, found: 785.1074.

4-methoxy-N-(4-((4-methoxyphenyl)sulfonamido)naphthalen-1-yl)-N-(2-oxo-2-(2-oxothiazolidin-3-yl)ethyl)benzenesulfonamide (pro2) To a solution of 2 (0.5 g, 0.90 mmol) in DMF (10 mL) was added DCC (0.41 g, 1.98 mmol) and DMAP (0.24 g, 1.98 mmol), stirring at room temperature. After 30 min, thiazolidin-2-one (0.186 g, 1.80 mmol) was added and the mixture was stirred overnight. After reaction finished, the mixture was poured onto water and extracted with Et₂O (3 × 20 mL). The organic extracts were combined, dried over Na₂SO₄, and concentrated in vacuo. The crude product was purified by column chromatography to give the pure product pro2 as a white solid (0.263 g, 46%).R_f = 0.31 (EA/PE 1:1); m.p. 229–230 °C; ¹H NMR (300 MHz, DMSO-d₆): δ = 10.23 (s, 1H; NH), 8.03 (d, J = 8.6 Hz, 2H; Ar–H), 7.65 (d, J = 8.5 Hz, 2H; Ar–H), 7.49 (d, J = 8.8 Hz, 4H; Ar–H), 7.07 (dd, J = 21.6, 8.4 Hz, 6H; Ar–H), 5.00–4.77 (m, 4H; CH₂), 3.95 (s, 2H; CH₂), 3.86–3.75 (m, 6H; OCH₃); ¹³C NMR (75 MHz, Chloroform-d) δ 173.40, 168.46, 163.24, 163.15, 134.57, 132.61, 132.54, 130.83, 130.38, 130.12, 129.54, 129.19, 128.74, 127.26, 127.09, 124.61, 121.57, 120.51, 114.27, 113.92, 56.72, 55.64, 46.55, 25.86; HRMS (ESI): m/z calcd. for C₂₉H₂₇N₃O₈S₃+NH₄⁺: 659.1299 [M + NH₄] ⁺; found: 659.1300.

2.2. Biology

2.2.1. HPLC assay for drug release study

Compounds were dissolved in DMSO as solutions (10 mM) and stored at −20 °C. The incubation was initiated by the addition of compound (10 mM) to phosphate buffer solution (PBS) (10 mM, pH 7.4) to obtain a final concentration of 50 μM and then added H₂O₂ followed by vortex mixing. The solution was incubated at 37 °C and conducted in triplicate. Samples were taken at appropriate time intervals and directly analyzed by HPLC analysis. Peak areas were recorded to calculate the percentage of compounds. Agilent 1260 HPLC and DAD detector with conditions: Agilent C18 column (4.6 × 150 mm, 3.5 μm); Mobile phase: methanol 70%; Flow rate: 0.5 mL/min. A standard curve for compounds was made to fit the measured concentrations.

2.2.2. Physicochemical property and cell membrane permeability determination

The pKa and partition coefficient (log D, pH 7.4) were determined according to the methods of Avdeef and Tsinman on a Gemini Profiler instrument (pION) by the “gold standard” Avdeef−Bucher potentiometric titration method. The pH-metric method was used to determine the intrinsic solubility. The potentiometric solubility data were obtained with the pSOL model 3 instrument (pION INC., Cambridge, MA, USA). Permeability coefficients were determined via double sink PAMPA on a PAMPA Explorer instrument (pION).

2.2.3. Cell culture and ARE-luciferase activity assay

The mouse RAW 264.7 cell line was obtained from Cell Bank of Shanghai Institute of Biochemistry and Cell Biology, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences. HepG2 cells stably transfected with a luciferase reporter (HepG2-ARE-C8) were kindly provided by Professor Dr. A. N. Tony Kong (Rutgers University, Piscataway, NJ) and Prof. Rong Hu (China Pharmaceutical University, Nanjing). The cells were maintained by regular passage in modified RPMI-1640 medium (GiBco, Invitrogen Corp., USA) supplied with 10% FBS, 100 units per mL penicillin and 100 μg/mL streptomycin, cultured at 37 °C in a water vapour saturated atmosphere with 5% CO₂. The experimental procedures were carried out as reported previously [60].

2.2.4. RNA extraction and qRT-PCR analysis

The experimental procedure of quantitative real-time RT-PCR was previously reported. Total RNA of RAW264.7 cells was extracted from the treated cells using TRIzol reagent (Invitrogen). Then the RNA was converted to cDNA by reverse transcriptase (PrimeScript RT reagent kit) according to the manufacturer's instructions. Quantitative real-time RT-PCR analysis of Nrf2, HO-1, NQO1 and GCLM were performed by using the StepOne System Fast Real Time PCR system (Applied Biosystems). The values are expressed as the fold of the control. All genes' mRNA expression was normalized against β-Actin expression.

2.2.5. Western blot analysis

Anti-Nrf2 (ab62352), anti-IL-1β (ab45692) antibodies were purchased from Abcam Technology (Abcam Technology, England). Anti–HO–1 (SC-136960) and anti-NQO1 (SC-271116) antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Anti-β-actin (60,008-1-lg), anti-GCLM (14241-1-AP) antibodies were purchased from Proteintech Group (Proteintech Group, USA). Briefly, the extracts were separated by SDS-PAGE and then electro-transferred to PVDF membranes (PerkinElmer, Northwalk, CT, USA). Membranes were blocked with 1% BSA for 1 h followed by incubation with a primary antibody at 4 °C overnight. Then they were washed and treated with a DyLight 800 labeled secondary antibody at 37 °C for 2 h. The membranes were screened through the odyssey infrared imaging System (LI-COR, Lincoln, Nebraska, USA).

2.2.6. IL-1β, IL-18, IL-6, TNF-α and NO production

Levels of IL-1β (IL-1β (m) ELISA kit, EK0394, Boster), IL-18 (IL-18 (m) ELISA kit, #EMC011, NeoBioscience), IL-6 (IL-6 (m) ELISA kit, EK0411, Boster), TNF-α (TNF-α (m) ELISA kit, EK0527, Boster) and NO production (Nitrate/Nitrite Assay Kit, S0023, Beyotime, China) were evaluated using commercially available kits according to the manufacturer's instructions.

2.2.7. Stability

• (1)pH Stability. PBS buffers with different pH values (pH 4–10) were to prepare by using 0.1 M HCl solution and 0.1 M NaOH solution. The prodrug pro2 (20 μM) was co-incubated with PBS buffer with different pH values at 37 °C and the assays were conducted in triplicate. After 24 h incubation, solution was filtered and adopted in direct injection analysis by HPLC. The injections were proceeded thrice, and peak areas were recorded to calculate the percentage of compounds;
• (2)Plasma stability. Sample of prodrug pro2 (20 μM) was co-incubated with mice plasma at 37 °C for different times and three parallel experiments was conducted. Protein was precipitated by adding methanol and samples were subjected to vortex mixing and then centrifugation for 5 min at 12,000 rpm to deproteinize. Samples of the resulting supernatants were withdrawn and analyzed by HPLC to record peak areas;
• (3)Microsome stability. The in vitro microsome stability of the compound was evaluated in isolated liver microsomes (from CD-1 male rat). Ketanserin was used as reference compounds. A solution of liver microsomes (20 mg/mL) was added to a microcentrifuge tube containing of PBS at 37 °C, and the mixture was shaken for 10 min before the actual assay was started. Then, a DMSO solution of test compound (0.5 mM) was added. For 0 min, add ice-cold acetonitrile to the wells of 0 min plate and then add NADPH stock solution (6 mM). Pre-incubate all other plates at 37 °C for 5 min. Add NADPH stock solution (6 mM) to the plates to start the reaction and timing. At 5 min, 15 min, 30 min, and 45 min, add ice-cold acetonitrile to the wells of corresponding plates, respectively, to stop the reaction. After quenching, shake the plates at the vibrator for 10 min and then centrifuge at 5000 rpm for 15 min. Transfer the supernatant from each well into a 96-well sample plate containing ultra pure water for LC-MS/MS analysis;
• (4)Stability in artificial gastric juice and intestinal juice. Artificial gastric juice and intestinal juice were purchased from commercial suppliers. Sample of pro2 (20 μM) was co-incubated with artificial gastric juice and intestinal juice respectively for different times at 37 °C and three parallel experiments was conducted. Zymoprotein was precipitated by adding methanol and samples were subjected to vortex mixing and then centrifugation for 5 min at 5000 rpm. Samples of the resulting supernatants were withdrawn and analyzed by HPLC to record peak areas; All the chromatographic condition is consistent with above-mentioned.

2.2.8. LPS challenge mouse acute inflammation model

Animal studies were conducted according to protocols approved by Institutional Animal Care and Use Committee of China Pharmaceutical University. All animals were appropriately used in a scientifically valid and ethical manner. After treatment with regular drinking water for 2 days for adaptation, female C57BL/6 mice (6–8 weeks of age, weighing 18–20 g) were randomized into eight groups: (A) Control group (n = 3); (B) LPS (Sigma-Aldrich, St. Louis, no. L4130) model group (300 μg/kg/day, n = 8); (C) LPS (300 μg/kg/day) + pro2 low-dose (10 mg/kg/day) group (n = 8); (D) LPS (300 μg/kg/day) + pro2 high-dose (40 mg/kg/day) group (n = 8); (E) LPS (300 μg/kg/day) + parent compound 2 high-dose (40 mg/kg/day) group (n = 8); (F) LPS (300 μg/kg/day) + dexamethasone low-dose (10 mg/kg/day) group (n = 8); (G) pro2 high-dose (40 mg/kg/day) group (n = 3); (H) parent compound 2 high-dose (40 mg/kg/day) group (n = 3).

Animals in control and (G, H) groups received a single IP injection containing 500 μL of saline (day −3, −2, −1). All LPS-challenged mice received a single IP injection containing 500 μL of LPS (day −3, −2, −1). Four hours after the injection, mice were pretreated with compound (ig, in 500 μL of saline, day −3, −2, −1). Animals were sacrificed 24 h after the last dose of compound and sera were collected (day 0). The level of the cytokines in sera was measured using ELISA kits.

3. Results and discussion

3.1. Design and synthesis of prodrugs

Previously, our group reported CPUY192002 (1), the first nanomole Keap1-Nrf2 inhibitor bearing two carboxylic acid groups important for the Keap1 binding [61], and further study showed that removal of one of the carboxylic acid groups induced moderate decrease in activity from an IC₅₀ of 28.6 nM–96.7 nM (compound 2, Fig. S1) in the fluorescence polarization (FP) competition assay, which has also been confirmed by the study of Moore's group [62]. However, for both of the two inhibitors, micromole range concentrations are needed for achieving obvious Nrf2 activation effects in various cellular models [[62], [63], [64]]. The offset between the in vitro potency and cellular activities may largely attribute to the poor permeability caused by the ionizable carboxyl group. Thus, we chose to develop a ROS-responsive prodrug by appending a ROS-cleavable protecting group to the carboxylic acid moiety. The pioneer work from Cohen et al. reported the first identified thiazolidinone protecting group that can reveal carboxyl acid moiety upon activation by H₂O₂ [65]. Compared to the widely used aryl boronic esters trigger, the promising characters of low molecular mass and no GSH deletion byproducts make the thiazolidinone-based trigger more suitable for developing anti-oxidant and anti-inflammation agents. With these considerations in mind, we designed and synthesized the thiazolidinone-based prodrugs of potent Keap1-Nrf2 inhibitors (pro1 and pro2, Fig. 1 and Scheme 1). These two prodrugs demonstrated no inhibition against Keap-Nrf2 interaction in the FP assay, indicating that the thiazolidinone promoiety did abolish the activity of parent drugs.

Fig. 1.

Proposed scheme for activation of the thiazolidinone prodrugs.

The synthesis of the prodrugs is shown in Scheme 1. Amination of commercially available 1-nitronaphthalene afforded 4-Nitronaphthalen-1-amine (4). The nitro group of 4 was reduced by hydrogen and Pd/C, and the subsequent condensation with 4-Methoxybenzenesulfonyl chloride gave compound 5. Compound 6 were obtained by nucleophilic substitution of NH by ethyl bromoacetate in the presence of K₂CO₃ in DMF, and the subsequent condensation with 4-Methoxybenzenesulfonyl chloride gave compound 7. Hydrolysis of the ester group of 7 resulted in the parent compound 2. The thiazolidinone moiety was then introduced through a DCC/DMAP-mediated coupling with 2 to give the desired prodrug pro2. The synthesis for CPUY192002 (1) had been previously reported [61]. Amide bond formation between the carboxylic acid compound 1 and the thiazolidinone promoiety was also performed via addition of DCC and DMAP in DMF, which finally gave the prodrug pro1.

3.2. Validation of H₂O₂-responsive parent drug release and evaluation of stability

To verify that H₂O₂ was able to deprotect the thiazolidinone group, H₂O₂-induced prodrug transformation experiments were performed and monitored by HPLC. However, in the presence of H₂O₂, the prodrug pro1, which contains two thiazolidinone pro-moieties, was hydrolyzed to generate the compound with only one free carboxylic acid, identified by UPLC/HRMS, and the production of compound 1 cannot be observed even after 12 h (Fig. S2). In the case of the prodrug pro2, as expected, after incubating 24 h, a H₂O₂ concentration-dependent release of the parent compound 2 was observed, and approximately 60% prodrug can be released upon treatment with 1 mM H₂O₂ (Fig. 2A). Upon treatment with 10 mM H₂O₂, pro2 was activated with increasing incubating time and completely released to the parent compound 2 within 3 h (>98% conversion determined by HPLC, Fig. 2B and C). Then, the H₂O₂-responsive activation of pro2 was further examined in live cells. Murine macrophage cells (RAW264.7) were stimulated with H₂O₂ for 24 h to induce the intracellular ROS and then culture medium was replaced with fresh medium. Pro2 (1 mM) was added to stimulated or non-stimulated cells for incubation and the cell lysates were analyzed by LC-MS. No peak of 2 was observed in non-stimulated cells, while obvious signal of 2 appeared in stimulated cell lysate (Fig. S3). The results demonstrated that pro2 can be taken up by cells and transformed to 2 by intracellular H₂O₂.

Fig. 2.

Activation of pro2 with H₂O₂. (A) Activation of pro2 under different concentrations of H₂O₂ after incubating 24 h. (B) Time release curve of pro2 with 10 mM H₂O₂. (C) HPLC chromatograms of the activation of pro2 by H₂O₂.

Then, the stability of the synthesized prodrug was evaluated in different relevant physiological conditions. It was observed that pro2 was relatively stable to hydrolysis in PBS of different pH values from 4 to 10 after incubating 24 h (Fig. 3A). In bio-relevant media, including simulated gastric fluid (SGF) and simulated intestinal fluid (SIF), more than 50% of pro2 remained after 8 h (Fig. 3B). Moreover, pro2 is quite stable in rat plasma (Fig. 3C) and in co-incubation with rat liver microsomes (Fig. 3D), indicating the high metabolic stability.

Fig. 3.

Stability of the prodrug pro2. Stability of pro2 (20 μM) in different relevant physiological conditions including different pH (4–10) (A), SGF (B), SIF (B), rat plasma (C) and liver microsomes (D).

3.3. Physicochemical properties and cell membrane permeability

Then, we determined the physicochemical properties of pro2 as well as the parent compound 2. As shown in Table 1, the pKa and log D_pH=7.4 of 2 are 4.31 and 1.88, while the pKa and logD_pH=7.4 of pro2 are 4.53 and 2.34, respectively. These results confirmed that pro2 is a less ionizable and less polar molecule than 2, which can enhance the passive cell membrane permeability of pro2. We then examined the cell membrane permeability by using a standard parallel artificial membrane permeability assay (PAMPA). The parent compound 2 showed a permeability coefficients (Pe) value of 0.80 × 10⁻⁶ cm/s, while pro2 gave a Pe value of 6.35 × 10⁻⁶ cm/s, which proved that pro2 can permeate the cell membrane more easily.

Table 1.

Physicochemical property and membrane permeability measurement.

Compd	pKa	LogD at pH 7.4	Aqueous solubility at pH 7.4 (μg/mL)	Pe at pH 7.4 (10−6 cm/s)
Pro2	4.53	2.34	879.6	6.35
2	4.31	1.88	428.2	0.80

3.4. Selective activation of prodrug by inflammation-derived ROS

Next, the cellular biological activity of pro2 was investigated. Initial studies aimed at demonstrating selective Nrf2 activation effects of pro2 upon ROS. The Nrf2/ARE luciferase reporter assay was applied to evaluate the cellular Nrf2 activity, and the methyl ester prodrug 3, which is not responsive to cellular ROS, was used as an unselective control to demonstrate the ROS-responsive selectivity of pro2. As shown in Fig. 4A, both of 2 and 3 concentration-dependently induced the ARE activity. The methyl ester prodrug 3 showed higher ARE-induction activity compared to the active compound 2. However, pro2 kept nearly inactive even at the highest concentration, which indicated that pro2 is stable and does not affect the Nrf2-ARE system under physiological conditions. In order to simulate the inflammation related high ROS microenvironment, cells were exposed to lipopolysaccharide (LPS), a widely used inflammation inducer which can enhance the production of intracellular ROS. Treatment with LPS can elevate ROS level with the increase of LPS concentration (Fig. S4). Then, Nrf2-ARE induction activites of these compounds were examied under the gradient concentrations of LPS. In order to exclude the LPS-induced Nrf2 activation effects, cells pretreated with gradient concentration of LPS alone were used as the control. As shown in Fig. 4B, LPS-stimulation before drug exposure (200 nM) activated pro2 and resulted in the LPS concentration-dependent enhancement of Nrf2-ARE inducing activity, while the potency of 2 and 3 did not show obvious changes. Under LPS stimulation, pro2 showed the lowest EC₅₀ value (0.32 μM) among them in the ARE induction activity (Fig. 4C). Together, the thiazolidinone-based prodrug pro2 can be selectively activated by LPS-induced intracellular ROS, enhancing the Nrf2-ARE system at a much lower concentration than the parent drug.

Fig. 4.

Comparison of activities of pro2, 2 and 3. (A) Relative ARE induction activity in HepG2-ARE-C8 cells under normal conditions. (B) Relative ARE induction activity under various concentrations of LPS in HepG2-ARE-C8 cells. Cells were pretreated by various concentrations of LPS for 8 h, and then treated with selected compounds (200 nM) or DMSO (for use as the control) for another 12 h. The values are expressed as the fold of the control. (C) The EC₅₀ curves of the relative ARE induction activity. Cells were pretreated by LPS (20 ng/mL) for 8 h, and then treated with various concentrations (0.001–50 μM) of selected compounds or DMSO (for use as the control) for another 12 h. The values are expressed as the fold of the control.

3.5. Activation of the Nrf2-ARE regulated antioxidant system in the RAW264.7 cells

To ascertain the effects of pro2 on the transcription of Nrf2-ARE-driven genes, the mRNA levels of Nrf2 downstream genes, Heme oxygenase 1 (HO-1), NAD(P)H: Quinone Oxidoreductase 1 (NQO1) and glutamate-cysteine ligase, modifier subunit (GCLM), were examined (Fig. 5A–C). The quantitative real-time PCR (qRT-PCR) analysis showed that exposure of RAW264.7 cells to 20 ng/mL LPS for 8 h slightly increased the transcription of Nrf2 targeted genes. Addition of pro2 (500 nM) highly enhanced the transcription, superior to the effects of 2 and its methyl ester prodrug at the same concentration. The protein levels of these genes were measured by immunostaining, and pro2 remarkably accumulated Nrf2 protein and elevated downstream antioxidant enzymes, more potent than the parent drug and its methyl ester prodrug (Fig. 5D).

Fig. 5.

Prodrug activated the Nrf2-ARE regulated antioxidant system in the RAW264.7 cells. RAW264.7 cells were pretreated with 20 ng/mL LPS for 8 h and then treated with compounds for another 16 h. PCR analysis of HO-1 (A), NQO1 (B) and GCLM (C) in the RAW264.7 cells. (D) Western blot analysis of the Nrf2-targeted proteins in the RAW264.7 cells. (E–H) Measurement of SOD, GSH-PX, GSH/GSSG and MPO level in the RAW264.7 cells. The values shown are the means ± SEM (n = 3 independent observations). *p < 0.05, **p < 0.01, ***p < 0.001, which were calculated with one-way ANOVA.

Subsequently, to explore the effects of pro2 on antioxidant capacity under inflammatory conditions, the activities of representative enzymes, superoxide dismutase (SOD) and glutathione peroxidase (GSH-Px), were determined. Mouse RAW 264.7 cells treated with LPS (20 ng/mL) alone showed the obvious decrease in activities of SOD and GSH-Px (Fig. 5E and F). Treatment with pro2 significantly restored the activities of SOD and GSH-Px, while the parent compound and its methyl ester analog were less efficient. We also measured the GSH/GSSG ratio and myeloperoxidase (MPO) activity (Fig. 5G and H), two important markers for indication of oxidative stress. LPS exposure caused huge decline in the GSH/GSSG ratio and sharp rise in the MPO activity, confirming that LPS could cause the oxidative stress. The addition of pro2 restored the levels nearly back to normal, the parent drug and its methyl analog showed similar trend but much lower activity.

3.6. Anti-inflammatory effects of pro2 in the LPS-induced RAW264.7 cells

Nrf2 activation has been proven to be an effective way to relieve the inflammatory conditions. We first examined the mRNA levels of IL-1β and IL-6, which can be transcriptionally repressed by Nrf2 [66]. After LPS (20 ng/mL) stimulation for 8 h, the mRNA levels of IL-1β and IL-6 were remarkably induced in the RAW264.7 cells. Treatment with 2 or the methyl ester prodrug 3 slightly inhibited the transcription of IL-1β and IL-6. Of note, treatment with pro2 resulted in significant suppression effects (Fig. 6A and B). Consistent with the results of the PCR analyses, 2 and 3 showed moderate effects on the inhibition of LPS-induced elevation of IL-1β and IL-6 protein, but pro2, at the same concentration, almost completely suppressed the elevation (Fig. 6C). Then, we quantitatively evaluated several inflammatory mediators that are closely related with ROS, including IL-1β, IL-6, TNF-α and NO. As expected, all these inflammatory factors increased markedly in the LPS-treated groups, and pro2 showed obvious superiority in the suppression of IL-1β, IL-6 and TNF-α production (Fig. 6D–F). The LPS triggered NO production was also remarkably diminished by pro2 with an EC₅₀ of 0.12 μM, while compound 2 and 3 showed higher EC₅₀ values (Fig. 6G).

Fig. 6.

Prodrug reduced inflammatory factors production induced by LPS in the RAW264.7 cells. RAW264.7 cells were pretreated with 20 ng/mL LPS for 8 h and then treated with compounds for another 16 h. PCR analysis of IL-1β (A) and IL-6 (B) in the RAW264.7 cells. (C) Western blot analysis of IL-1β and IL-6 protein levels in the RAW264.7 cells. Concentrations of IL-1β (D), IL-6 (E), TNF-α (F) and NO (G) in the RAW264.7 cell culture supernatants. The values shown are the means ± SEM (n = 3 independent observations). *p < 0.05, **p < 0.01, ***p < 0.001, which were calculated with one-way ANOVA.

3.7. In vivo efficacy of pro2 against the LPS-induced inflammatory conditions

After confirming the ROS-responsive Nrf2 activation effects of pro2 in live cells, we finally investigated the therapeutic potential of pro2 in vivo. To explore the suitability of thiazolidinone-based prodrug for oral administration, we evaluated the pharmacokinetics of pro2 for both IV and PO, and the oral pharmacokinetics with bioavailability of 68.1% and half-life of 2.09 ± 0.93 h (Table 2) indicated pro2 could be an effective oral medication.

Table 2.

Physicochemical Properties in vivo.

	P.O. (10 mg/kg)	I.V. (2 mg/kg)
T1/2 (hr)	2.09 ± 0.93	4.41 ± 3.01
Tmax (hr)	0.67 ± 0.29	–
Cmax (ng/mL)	2753 ± 15.3	–
AUC0-t (ng·hr/mL)	6203 ± 248	1842 ± 363
AUC0-∞ (ng·hr/mL)	6315 ± 273	1854 ± 366
Vz_F(L/kg)	4.79 ± 2.17	6.44 ± 4.09
Cl_F (mL/min/kg)	26.4 ± 1.14	18.5 ± 3.9
MRT (hr)	1.82 ± 0.35	1.25 ± 0.30

Bioavailability 68.1%.

Then, the murine LPS challenged acute inflammation model was used to evaluate anti-inflammation effects in vivo. C57BL/6 mice were challenged with LPS (300 μg/kg, IP) and then orally administrated with compound 4 h after the LPS challenge for 3 days. The blank group only received saline during the experiment. Dexamethasone (DXM), the widely used steroid anti-inflammatory drug, was used as the positive control. Animals were sacrificed 24 h after the last dose of compound and sera were collected. LPS challenge markedly elevated the pro-inflammatory cytokines in mice sera, including IL-1β, IL-6, TNF-α and IFN-γ. DXM treatment (10 mg/kg) diminished inflammatory response (Fig. 7A–D). Oral administration of 2 (40 mg/kg) showed moderate anti-inflammation activity, much less potent than DXM. Pro2 showed comparable therapeutic effects with DXM at the same dose (10 mg/kg), and high dose of pro2 showed more potent effects, indicating the dose-dependent behavior. Taken together, these results suggested that the thiazolidinone-based ROS-responsive prodrug of the Keap1-Nrf2 PPI inhibitor is an effective oral medication for oxidative stress related inflammatory conditions.

Fig. 7.

Prodrug reduced the LPS-induced production of the pro-inflammatory factors in vivo. Levels of serum IL-1β (A), IL-6 (B), TNF-α (C) and IFN-γ (D) were measured by Elisa kits. *p < 0.05, **p < 0.01, ***p < 0.001, which were calculated with one-way ANOVA.

4. Conclusions

In this study, a new ROS-activated prodrug pro2 was developed by utilizing H₂O₂-responsive thiazolidinone moiety to shield the key carboxyl pharmacophore in Keap1-Nrf2 inhibitor. Inflammation derived intracellular H₂O₂ can deprotect carboxyl group, producing the potent Keap1-Nrf2 inhibitor to selectively activate Nrf2-regulated antioxidant system in target cells. Moreover, the thiazolidinone modification of the carboxyl group improve the physicochemical properties and cell membrane permeability of the parent drug. Further cellular studies showed that pro2 can stay inactive at physiological conditions and be much more potent than the parent drug at inflammatory conditions. In addition, thiazolidinone-based drug displayed good stabilities at various physiological conditions and appropriate PK profile for oral administration. Finally, in vivo therapeutic activity was demonstrated by oral use of pro2 in LPS challenged acute inflammation model.

An increasing body of literature has revealed that over and unselective activation of Nrf2 may cause unexpected risks and particularly contribute to the initiation and progression of cancer [[67], [68], [69], [70]], inspiring the discovery of precision Nrf2 activators. Direct Keap1-Nrf2 PPI inhibitors are assumed to have higher target selectivity than electrophilic Nrf2 activators. However, one critical challenge for the Keap1-Nrf2 inhibitors has also emerged. Direct Keap1-Nrf2 inhibitors are designed to bind to Keap1 with a similar pattern of Nrf2 ETGE motif, which may also affect functions of other Keap1 substrates with a similar recognition motif [71,72]. Developing Keap1-Nrf2 inhibitors selectively activating Nrf2 in specific pathologic conditions is a new avenue. The carboxyl group, the key pharmacophore in the direct Keap1-Nrf2 inhibitors, can be utilized to design a targeted prodrug that not only improves the ADMET properties but also selectively activates Nrf2. The high concentrations of H₂O₂ in the inflammatory environment can serve as the stimulus for prodrug activation. This pathologic site-selective drug delivery system therefore restricts the effects of Keap1-Nrf2 inhibitors on Keap1-involved interactome of normal tissues.

Taken together, our study confirmed that this H₂O₂-activated prodrug can achieve selective Nrf2 activation and enhanced in vivo efficacy simultaneously, providing an attractive approach for the further development of Keap1-Nrf2 inhibitors.

Declaration of competing interest

All authors have given approval to the final version of the manuscript. The authors have no conflicts of interest to declare.

Abbreviations

Nrf2

Nuclear factor erythroid 2-related factor 2

Keap1

Kelch-like ECH-associated protein 1

H₂O₂

hydrogen peroxide

ROS

reactive oxygen species

ARE

antioxidant response element

PPI

protein-protein interaction

ITC

Isothermal Titration Calorimetry

PAMPA

parallel artificial membrane permeability assay

Pe

permeability coefficients

LPS

lipopolysaccharide

HO-1

Heme oxygenase 1

NQO-1

NAD(P)H dehydrogenase (quinone) 1

GCLM

glutamate-cysteine ligase regulatory subunit

qRT-PCR

quantitative real-time PCR

SOD

superoxide dismutase

GPx

glutathione peroxidase

MPO

myeloperoxidase

DXM

dexamethasone

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Characterization of the other intermediates, expression and purification of Keap1 Kelch Domain, FP competition assay to determine the inhibitory potency of the Keap1-Nrf2 interaction, the Isothermal titration calorimetry (ITC) assay, UPLC/HRMS and LC-MS assay for drug release study, LPS-induced ROS generation, primers used for qRT-PCR and in vivo pharmacokinetic assay.