Optimization of 3-aminotetrahydrothiophene 1,1-dioxides with improved potency and efficacy as non-electrophilic antioxidant response element (ARE) activators

Abstract

Activating NRF2-driven transcription with non-electrophilic small molecules represents an attractive strategy to therapeutically target disease states associated with oxidative stress and inflammation. In this study, we describe a campaign to optimize the potency and efficacy of a previously identified bis-sulfone based non-electrophilic ARE activator 2. This work identifies the efficacious analog 17, a compound with a non-cytotoxic profile in IMR32 cells, as well as ARE activators 18 and 22, analogs with improved cellular potency. In silico drug-likeness prediction suggested the optimized bis-sulfones 17, 18, and 22 will likely be of pharmacological utility.

Continual oxidative insult is a characteristic of several age-related conditions, including autoimmune disease and neurodegeneration.^1,2 To combat runaway inflammatory damage during oxidative challenge, the mammalian cell has evolved nuclear factor erythroid 2-related factor 2 (NFE2L2, NRF2 throughout), a transcription factor that induces a pro-protective transcriptional program of gene products containing antioxidant response element (ARE) sites within their genomic loci.³ In normal physiological states, NRF2 is continually ubiquitinated by cytoplasmic repressor and CUL3 adaptor protein, Kelch-like ECH-associated protein 1 (KEAP1), which promotes its proteasomal degradation. However, in response to oxidative challenge, several nucleophilic cysteine sensor residues within KEAP1 can be oxidized or alkylated, resulting in the accumulation and transcriptional activity of NRF2.⁴

Forced activation of NRF2 has been shown to be ameliorative in numerous animal models of disease. Consequently, an expansive set of small molecules that activate NRF2 have been described in the literature.^5–10 Most of these molecules act by covalently modifying key sensor cysteines on KEAP1. Although selective covalent drugs have been used to great effect clinically,¹¹ NRF2-activating molecules, such as the clinical candidate bardoxolone methyl, are unselective¹² and cytotoxic¹³ and have failed to display durable clinical efficacy. Indeed, no approved drug that activates NRF2 as its primary mechanism is available, necessitating the development of alternative strategies to activate this central protective transcriptional program.

In addition to responding to exogenously derived oxidative stressors, KEAP1 also responds to altered levels of endogenous metabolites. For example, the intrinsically reactive mitochondrial metabolites, fumarate¹⁴ and itaconate¹⁵, covalently modify KEAP1 residues and alter cellular physiology by activating NRF2. As such, one potential mechanism for pharmacologically activating NRF2 involves co-opting natural signaling mechanisms by augmenting the levels of naturally occurring KEAP1-reactive metabolites. We recently described the identification of CBR-470–1, a bis-sulfone containing small molecule that inhibits the glycolytic enzyme phosphoglycerate kinase (PGK1).¹⁶ Decreased flux through glycolysis in this context results in the accumulation of a glycolytic degradation product methylglyoxal (MGO), a dicarbonyl that modifies cysteine and arginine residues to crosslink adjacent KEAP1 molecules.¹⁶ Despite altering a central metabolic pathway, CBR-470–1 treatment reportedly promotes cellular survival in several models of oxidative challenge.^17,18

In an initial attempt to characterize bis-sulfone series, we recently identified non-cytotoxic and potent activators of ARE transcriptional activity, indicating that targeting PGK1 through this mechanism does not result in obligate cytotoxicity to cells.¹⁸ One plausible explanation for this observation comes from the metabolism of molecules like 2, which like other molecules of this series, including 1, is oxidatively dealkylated by CYP enzymes, yielding 3 (Fig. 1). 3 displays considerably decreased potency in activating NRF2 and is considerably more cytotoxic than other analogs. As such, optimizing the substituents of the aminoalkyl moiety may afford analogs with decreased metabolism and improved cellular activity. To further explore this hypothesis and to improve the pharmacological properties of the scaffold, an additional optimization of 2 was performed.

Fig. 1.

Fine-tuning strategy for 3-aminotetrahydrothiophene 1,1-dioxide-based NRF2 activators.

Linkage-modified analogs 10 and 12 were successfully obtained following the strategy outlined in Scheme 1. For the efficient synthesis of the desired compounds, the common intermediate 5 was constructed first by the Michael addition of isobutylamine with commercially available methyl 3-sulfolene-3-carboxylate (4). Protection of the secondary amine of 5 with a Boc group, followed by reduction of the protected methyl ester 6 afforded aldehyde 7. Compound 7 reacted with commercial 3,4-dichlorophenylmagnesium bromide to provide secondary alcohol 8, which was subsequently oxidized by Dess–Martin oxidation to obtain ketone 9. Lastly, deprotection of 9 with TFA successfully produced the desired ketone 10. Amide derivative 12 was derived from the same common intermediate 5. After hydrolysis of 5 and subsequent HATU coupling of the resulting acid 11 with 3,4-dichloroaniline in the presence of Hünig’s base, the final compound 12 was readily synthesized.

Scheme 1.

Synthesis of analogs 10 and 12. Reagents and conditions: (a) i-BuNH₂, CH₃CN, rt, 91%; (b) Boc₂O, DMAP, CH₂Cl₂, rt, 92%; (c) DIBAL-H, CH₂Cl₂, −78 °C, 90%; (d) 3,4-dichlorophenylmagnesium bromide, THF, 0 °C; (e) DMP, CH₂Cl₂, 0 °C, 46% for 2 steps; (f) TFA, CH₂Cl₂, rt, 73%; (g) LiOH·H₂O, THF/H₂O (2:1), 0 °C, 95%; (h) 3,4-dichloroaniline, HATU, i-Pr₂NEt, DMF, rt, 35%.

Subsequent structural modifications focused on the metabolically labile aminoalkyl groups of 2. The synthesis of bis-sulfone analogs 17–28 was achieved according to our previously established procedure, as depicted in Scheme 2.¹⁸ Briefly, commercially available 3-sulfolene (13) was reacted with 3,4-dichlorobenzenethiol in the presence of N-bromosuccinimide (NBS) to produce β-bromosulfide 14. Treatment of 14 with pyridine in refluxing CH₂Cl₂ gave the dehydrobrominated product 15. Consecutive mCPBA oxidation of 15, followed by conjugate addition with diverse amines under basic or neutral conditions, successfully afforded the desired bis-sulfones 17–28.

Scheme 2.

Synthesis of analogs 17–28. Reagents and conditions: (a) 3,4-dichlorobenzenethiol, NBS, CH₂Cl₂, rt, 50%; (b) pyridine, CH₂Cl₂, 70 C, 95%; (c) mCPBA, CH₂Cl₂, rt; (d) 3,3,3-trifluoropropylamine, 1,4-diaminobutane, or HMDA, CH₃CN, rt, 21–78% for 2 steps (17, 23, and 24); (e) β-alanine, 1 N NaOH, CH₃CN, rt, 30% for 2 steps (18); (f) γ-aminobutyric acid or aminocaproic acid, 2 N NaOH, CH₃CN, rt, 16–23% for 2 steps (19 and 20); (g) 3-amino-2,2-difluoropropanoic acid hydrochloride, β-alanine methyl ester hydrochloride, 3-aminopropanamide hydrochloride, 2-aminoethanesulfonic acid, 2-aminoethanesulfonamide hydrochloride, or 2-aminoethylphosphonic acid, i-Pr₂NEt, CH₃CN, rt, 11–29% for 2 steps (21, 22, and 25–28).

The fourteen compounds 10, 12, and 17–28 obtained from the described synthetic route were evaluated for potency (EC₅₀) and efficacy (E_max) at inducing NRF2-dependent ARE-LUC reporter activity in IMR32 cells (Table 1). The cytotoxicity of the analogs against IMR32 cells after 24 h of treatment was also evaluated using a CellTiter-Glo-based assay. First, the effect of altering the linkage between the sulfolane and the aromatic ring was evaluated to assess the essentiality of the bis-sulfone core structure. As expected, the replacement of sulfone with a ketone displayed decreased potency and efficacy of ARE induction (Table 1, 10). A bioisosteric change of the sulfone to an amide also resulted in complete loss of ARE-LUC activity (Table 1, 12).^19,20 Together, these brief data confirm that the bis-sulfone core is essential for ARE activation.

Table 1.

ARE-LUC inducing activity and cytotoxicity of compounds 10, 12, and 17–28.

Comp.	ARE-LUC EC50 [μM]a (95% CI)	ARE-LUC Emax [FI]b	IMR32 cytotoxicity IC50 [μM]c
2	3.6 (0.9–5.6)	20.0	5.6
10	4.5 (0.9–5.8)	5.6	6.8
12	> 20	1.7	> 20
17	1.5 (1.0–4.1)	70.7	> 20
18	0.4 (0.3–2.0)	61.5	5.2
19	> 20	10.4	> 20
20	> 20	3.6	> 20
21	2.9 (0.8–5.7)	31.2	1.3
22	0.4 (0.3–0.5)	31.9	7.4
23	0.9 (0.4–1.8)	37.3	8.0
24	0.9 (0.8–3.1)	38.5	7.4
25	0.9 (0.4–1.3)	53.3	10.0
26	4.2 (2.1–5.6)	18.4	> 20
27	1.9 (1.6–5.1)	27.5	> 20
28	4.2 (1.4–4.8)	43.0	0.04

^aEC₅₀ values are the mean of three experiments and correspond to the concentration resulting in half-maximal induction for each compound.

^bFI, fold induction relative to a DMSO neutral stimulation control.

^cIC₅₀ values are the mean of three experiments and correspond to the concentration of each compound which results in 50% cytotoxicity.

The effect of altering the substituents on the aminoalkyl group was then analyzed (Table 1, 17–28). The designed analogs generally showed similar or enhanced potency compared to 2, except for compounds 19 and 20. The most efficacious analog was 17, a 3,3,3-trifluoropropylamine substituted compound, demonstrating 3.5 times higher FI (70.7) than that of 2. Of note, 17 displayed no cytotoxicity toward IMR32 cells (< 20 μM). Replacement of the isobutylamine group of 2 with β-alanine resulted in an approximately 10-fold improvement of ARE-inducing potency (Table 1, 18). In contrast, substitution with γ-aminobutyric acid or aminocaproic acid led to the inability to activate ARE-driven transcription (Table 1, 19 and 20). For terminal carboxylic acids, an optimal alkyl chain length seems to exist. Among the tested compounds, five analogs, 18 and 22–25, possessed submicromolar potency for ARE activation. 22 is anticipated to be hydrolyzed to the carboxylic acid 18 in cells, which we attribute to the observed equipotency of these analogs. Like 17, the active analogs 26 and 27 also had no cytotoxic potential below 20 μM. Together, these data support the notion that cytoprotective NRF2 activation can be induced in cells by modulating PGK1 activity without obligate cytotoxicity.¹⁸

To investigate the likelihood of the bis-sulfones as candidates for therapeutic development, the most potent and efficacious compounds 17, 18, and 22 were evaluated using the web-based SwissADME tool for their in silico physicochemical and pharmacokinetic properties and drug-likeness (Table 2).²¹ The calculated scores forecasted that compounds 17, 18, and 22 possess favorable drug-likeness. They contain 7–8 hydrogen bond acceptors (HBA), 0 or 1 hydrogen bond donor (HBD), and 6–7 rotatable bonds (RB), which are within the range of lead-likeness criteria (HBA ≤ 9, HBD ≤ 5, and RB ≤ 10).²² Then, the topological polar surface area (TPSA) value and Abbot bioavailability score (ABS) were evaluated.²³ The tested analogs met the criteria of TPSA (≤ 140 Å), and their ABS was determined as 0.55, indicating a high potential of oral bioavailability.^23–25 The investigated compounds had reasonable physicochemical properties of ClogP and LogS values within the lead-likeness range²² and were moderately soluble (MS) based on Ali solubility class. Moreover, the analogs were estimated to have a high plausibility of gastrointestinal absorption (GA) and to not be p-glyco-protein substrates. Plotting these results within Bioavailability Radar plots enabled the visualization of these properties (Fig. 2). The optimal range for six physicochemical properties, including lipophilicity (LIPO, 0.7 < XLOGP3 < +5.0), size (SIZE, 150 < MW < 500 g/mol), polarity (POLAR, 20 < TPSA < 130 Å), solubility (INSOLU, 6 < log S < 0), saturation (INSATU, 0.25 < Fsp³ < 1), and flexibility (FLEX, 0 < rotatable bonds < 9), are depicted as the pink area. Although only the polarity of 18 was slightly out of the radar range, the plots of all compounds were constrained within the hexagonal area and thus considered to be drug-like.²⁴ Lastly, analogs 17, 18, and 22 were evaluated for drug-likeness (Lipinski and Veber) and Pan Assay Interference Structures (PAINS).^27,28 No such structural alerts were found in these analogs.

Table 2.

SwissADME physicochemical, pharmacokinetic, and drug-likeness prediction for compounds 17, 18, and 22.

Comp.	HBA and HBDa	RBb	TPSAc (Å2)	ABSd	CLogPe	Ali LogSf	Ali class	GAg	Pgph	Drug-likeness (# viol.)i	PAINS
17	8 and 1	6	97.07	0.55	3.32	−5.00	MS	High	No	Yes(0)	0 alerts
18	7 and 2	7	134.37	0.55	1.13	−1.38	MS	High	No	Yes (0)	0 alerts
22	7 and 1	7	123.37	0.55	2.09	−3.84	MS	High	No	Yes (0)	0 alerts

^aHBA: number of hydrogen bond acceptors; HBD: number of hydrogen bond donors.

^bNumber of rotatable bonds.

^cTopological polar surface area.

^dAbbot bioavailability score.

^eConsensus LogP calculated by average of iLOGP, XLOGP3, WLOGP, MLOGP, and SILICOS-IT Log P.

^fAli topographical method LogS.²⁶

^gGastrointestinal absorption.

^hP-glycoprotein substrates.

ⁱNumber of violations of Lipinski and Veber filters.

Fig. 2.

Bioavailability Radar plot of compounds 17, 18, and 22.

In summary, a series of 3-aminotetrahydrothiophene 1,1-dioxides was designed, synthesized, and biologically evaluated as a part of campaign to efficiently target NRF2 with non-electrophilic small molecules. The systematic studies on scaffold modification confirmed that the sulfone linkage is essential for ARE-inducing activity. Structural fine-tuning of the aminoalkyl moiety of 2 allowed us to identify the most efficacious and potent analogs 17, 18, and 22 in this chemical series. The drug-likeness of these advanced analogs was evaluated using the in silico SwissADME tool, and the results suggested that these analogs likely provide a chemical basis for future preclinical development, providing a roadmap for further optimization of this series and the chemical matter for in vivo investigation of therapeutic activity.

Data availability

No data was used for the research described in the article.