Introduction

Chiral 3-substituted and 3,4-disubstituted succinimide substructures are widely found in natural products and pharmaceuticals (Fig. 1)1,2,3,4,5,6. Succinimide derivatives such as phensuximide (PTS), methsuximide (MTS) and ethosuximide (ETS), are well-known antiepileptic drugs (AEDs). In recent years, chiral succinimides have received great attention from synthetic chemists due to their wide range of biological activities such as antibacterial7, antifungal8, analgesic9, anticonvulsant10, and antitumor effects11,12. The most well-developed methods include enantioselective cycloaddition reactions13,14,15 and hydrogenation reaction16,17,18 using maleimides as the substrates. Asymmetric catalytic addition 19,20of nucleophilic reagents to maleimides has also been reported (Fig. 2a). However, compared with well-developed synthetic methods of 3-substituted succinimides, few methodologies exist concerning the synthesis of 3,4-disubstituted succinimides19,21. The synthon, 3-hydroxy-4-substituted-succinimides can be easily converted to valuable chiral skeletons, such as chiral pyrrolidones and chiral lactams (Fig. 2c). Consequently, developing an efficient synthetic methodology for the construction of 3-hydroxy-4-substituted-succinimides from readily available starting materials is highly desirable.

Fig. 1: Biologically active compounds and drugs derived from succinimides.
figure 1

Phensuximide, methsuximide and ethosuximide are well-known antiepileptic drugs. Upadacitinib is a drug for the treatment of immune disorders, Echinocandins have antifungal activities, and Tivantinib is used for the treatment of advanced hepatocellular carcinoma.

Full size image
Fig. 2: Synthetic methods for the construction of chiral succinimide derivatives.
figure 2

a Catalytic asymmetric synthesis of chiral succinimides (previous work). b Stereodivergent synthesis of chiral succinimide derivatives via ATH (this work). c Obtained building blocks.

Full size image

Since the first asymmetric transfer hydrogenation (ATH) of ketones reported by Noyori and Ikariya using (S,S)-cat.122, a series of catalysts containing N-monotosylated 1,2-diphenylethylenediamine (TsDPEN) have been documented by Noyori22, Ikariya23, Wills24,25,26 and others27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42. In 1999, Baiker et al. reported an enantioselective hydrogenation of pyrrolidine-2,3,5-triones using Pt–cinchonidine systems, obtaining 3,4-disubstituted succinimides with a single stereocenter16. Inspired by this work, we envisioned that 3-hydroxy-4-substituted maleimides could be reduced through a DKR (dynamic kinetic resolution)-ATH process. Over the past decades, DKR-ATH is an extremely attractive approach for the reduction of α-functional ketones due to the assured enantiomeric purity43,44,45,46,47,48,49,50,51. In particular, base is necessary to facilitate the ATH process52,53. Compared with well-studied α-functional ketones, 3-hydroxy-4-substituted maleimide (1) exists mainly in its enol form45, and has low reducing activity under alkaline conditions. It can be reduced through the Pd/C-H2 system but cannot be reduced by NaBH4/MeOH system (for details, see Supplementary Fig. 9). Moreover, the corresponding reduction products are prone to epimerization under alkaline conditions. This also made it difficult to obtain the reduction products using traditional alkaline reduction systems. Therefore, the development of appropriate asymmetric reduction methods to solve such problems is of great significance.

Herein, we report the stereodivergent enantio- and diastereoselective ATH of maleimide derivatives catalyzed by a tethered rhodium catalyst. Anti-3-hydroxy-4-substituted-succinimides 2 (up to >99% ee, up to >99:1 dr) and syn-3-hydroxy-4-substituted-succinimides 3 can be obtained by adjusting the amount of base (up to >99% ee, up to >99:1 dr). In addition, both the enol and the adjacent imide can be reduced to obtain 4 with high chemoselectivity by control the loading of the catalyst (up to >99% ee, up to >99:1 dr) (Fig. 2b).

Results

Condition optimization

To investigate the proposed ATH process, model substrate 1a was subjected to reduction using the commercially available TsDPEN-derived Ru, Rh and Ir complexes (2 mol%) in an azeotropic mixture of formic acid and triethylamine at 25 °C in MeOH. (Table 1) The Noyori’s catalyst (R,R)-cat.1 and (S,S)-cat.2 cannot achieve any catalytic conversion (entries 1–2). Ir complexes, (S,S)-cat.3 can only achieve 23% conversion with 84% ee and 84:16 dr (entry 3). In contrast, the tethered-catalysts, cat.4~cat.5 showed more excellent conversion, enantioselectivity and diastereoselectivity. 96% ee and 90:10 dr were achieved with (R,R)-cat.4 while 95% ee and 92:8 dr were achieved with catalyst (R,R)-cat.5 (entries 4–5). To our great delight, the catalytic selectivity of Rh catalyst (R,R)-cat.6 and (S,S)-cat.6 outperformed that of the catalysts previously evaluated, providing corresponding product 2a with 95:5 dr and 96% ee in an anti-selective manner (entries 6–7). (S,S)-cat.6 proved to be potential catalysts for the DKR-ATH of 1a.

Table 1 Optimization for Rh(III)-catalyzed DKR-ATH of 1aa
Full size table

With (S,S)-cat.6 as a catalyst, we further explored the effects of solvents and the ratio of formic acid and triethylamine (Table 1). Several solvents, such as EtOH, DCM, THF, dioxane, toluene, hexane and EtOAc were examined with HCO2H/Et3N (5:2) for 12 h (entries 8–14). Except for a low conversion in hexane (<5% conv.), moderate to high conversion and high stereoselectivity were observed in other solvents. Pleasingly, when EtOAc was used, excellent enantioselectivity and diastereoselectivity were obtained in full conversion (entry 14, 99% ee, 98:2 dr). On the contrary, syn-product 3a was obtained by reducing the amount of trimethylamine (entry 15). When only formic acid was used, the enantioselectivity of the product decreased slightly, and extended reaction time was necessary to achieve full conversion (entry 16). In addition, different ratios of triethylamine and formic acid were investigated in EtOAc (Supplementary Table 1 presents details), indicating that the decrease in the proportion of Et3N resulted in a decrease in the proportion of anti-product. Finally, isopropanol and sodium formate were also investigated as hydrogen donors, and a very low reactivity was observed under the alkaline environment (entries 17–18, <5% conv.). So far, HCO2H/Et3N (5:2) was used as a hydrogen donor for anti-2a (entry 14) and HCO2H/Et3N (2.0 equiv./0.02 equiv.) for syn-3a (entry 15).

Substrate scope

Under the optimal conditions, the methodology using of HCO2H/Et3N (5:2) as a hydrogen source was successfully extended to a series of substrates, and the results were illustrated in Fig. 3. Substrates with the different protective groups on the nitrogen atom were subjected to the standard reaction conditions, and the reaction proceeded smoothly to provide the anti-product with high yields, excellent enantioselectivities and diastereoselectivities (2a–2g, 94–98% yield, 94:6–>99:1 dr, 88%–>99% ee). Next, we evaluated the effect of the substituents on the C4 position. Functional groups, such as halides (2h–2j), trifluoromethyl (2k), methyl (2l), methoxy (2m) at the para position of the phenyl group were compatible with this transformation (88–97% yield, 90:10–99:1 dr, 96%–>99% ee). Substrates with meta-substitution on the phenyl group were also tolerated, and 98% ee and 96:4 dr were obtained (2n). In addition, the ortho-methoxy and ortho-fluoro substrate were evaluated with lower dr values (2o–2p, 95–96% yield, 91:9 dr and 97% ee). Moreover, the product 2r with 3,4-methoxy groups on the phenyl ring was obtained with 99% ee and 99:1 dr. The current reaction also tolerates substrates bearing 2-naphthyl, 2-furanyl, and 3-indolyl groups (2r–2t, 92%–95% yield, 84:16–99:1 dr, 97–99% ee).

Fig. 3: Substrate scope.
figure 3

aConducted with catalyst/substrate (0.2 mmol) ratio of 1: 50 in 1 mL of solvent, HCO2H (2.0 equiv.)/Et3N (2 mol%) for syn–isomer 3 and HCO2H/Et3N azeotropic mixture (20 μL) obtained anti–isomer 2. Isolated yield including the minor enantiomer, the ee and dr value of were determined by HPLC analysis using a chiral stationary phase. b(S,S)-cat.6 (5 mol%) and HCO2H/Et3N azeotropic mixture (50 μL) were used. c(S,S)-cat.6 (5 mol%) and HCO2H (2.0 equiv.) were used for 24 h. d(S,S)-cat.6 (1 mol%) and HCO2H/Et3N azeotropic mixture (20 μL) were used for 1 h. e(S,S)-cat.6 (1 mol%) and HCO2H/Et3N azeotropic mixture (20 μL) were used for 0.5 h. f(S,S)-cat.6 (5 mol%) and HCO2H/Et3N azeotropic mixture (40 μL) were used for 24 h.

Full size image

Subsequently, a sharp contrast was exhibited using HCO2H (2.0 equiv.)/Et3N (2 mol%) as a hydrogen source, and the syn-products could be obtained selectively. (Fig. 3) High enantioselectivities and diastereoselectivities were obtained with a wide range of N-protected substrates (3a–3g, 93–98% yield, 93:7–>99:1 dr, 88–98% ee). The effect of substitution at the C4 position of the substrates was also evaluated, and products with halides, trifluoromethyl, methyl, and methoxy group at the ortho, meta, or para position of the phenyl group were smoothly produced in 92–98% yields with 95:5–99:1 dr and up to >99% ee (3h–3q). 2-Naphthyl-containing substrates could obtain excellent enantioselectivity and diastereoselectivity as well (3r, 98% yield, >99:1 dr, 98% ee). Heterocycle-containing substrates, such as indole (3s, 91% ee, 99:1 dr) and furan (3t, 94% ee, 96:4 dr) were compatible with the conditions. For substrates with alkyl substitution at the C4 position, using an azeotropic mixture of HCO2H/Et3N (5:2) as a hydrogen source, this transformation proceeded smoothly to provide the syn-product in excellent yields (up to 97%) and with high levels of diastereo- and enantioselectivities (3u–3x, 97–98% ee, >99:1 dr). What’s more, N-unprotected substrates were well tolerated (3y–3aa, 98%–>99% ee, >99:1 dr). Interestingly, for N-unprotected substrates, both the enol and imide groups can be reduced. The products 4,5-dihydroxy-3-alkyl-pyrrolidin-2-one, containing three contiguous stereocenters, can be obtained by increasing the loading of catalyst (4y-4aa, 98%–>99% ee, >99:1 dr).

Synthetic applications

To demonstrate the synthetic utilities of this methodology, two gram-scale transformations were conducted, and the results were summarized in Fig. 4. To our delight, when 0.05 mol% (S/C = 2000) catalyst loading was used, the gram-scale experiment proceeded smoothly to provide 3u with excellent results. According to the procedure in the literature, 3u can be further transformed into the important synthon 5u (3-hydroxy-4-methylproline), which is a key intermediate for Echinocandin (Fig. 4a)54. Similarly, the ATH of 1w could be performed with a catalyst loading of 0.1 mol% to obtain 3w with excellent results. Then it can be further transformed into pyrrolidine derivative 6w, lactam derivatives 7w55, 8w, and succinimide derivative 9w56 in high yield and excellent stereoselectivity (Fig. 4b).

Fig. 4: Gram scale synthetic utilities of the DKR-ATH process.
figure 4

a Gram-scale experiment of 1u. b Gram-scale experiment of 1w and transformations of product 3w.

Full size image

Mechanism study

To elucidate the detailed mechanism of this ATH process, a series of mechanistic investigations were conducted, and the results were summarized in Fig. 5. First, the reaction of hydroxyl-protected substrate 1a’ was conducted under the standard conditions, and 2a’ was obtained with only 37–38% ee and 85/15-87/13 dr, which means that the presence of the –OH group at the C3-position is necessary to ensure high enantioselectivity, and the C=C reduction pathway is also possible. (Fig. 5a). Second, the syn-product 3a can be smoothly transformed into the anti-product 2a in the presence of an azeotropic mixture of HCO2H/Et3N (5:2). This suggested that the anti-product was probably produced by epimerization of the syn-product. Interestingly, 3y can be transformed into the 4y with three contiguous stereocenters (Fig. 5b). Moreover, the reaction kinetic experiment showed that 1a can be completely reduced in preference to 1a’ under standard conditions (Fig. 5c). During the process, the reduction of 1a first produced the syn-product which isomerized to the anti-product along with the process of the reaction, whereas, the reduction of 1a’ obtained the anti-product mainly, and the anti/syn ratio changed slightly during the reaction process (Fig. 5d). In addition, the keto-enol equilibrium experiment revealed that the ratio of enol-1u increased with the increase of pH value in acidic environment (Fig. 5e). Our mechanistic investigations revealed that the current reaction probably proceeded via reduction of the keto form.

Fig. 5: Mechanistic investigations.
figure 5

a Reactions of ethyl-protected substrate under standard conditions. b Observed effect of HCO2H/Et3N (5:2) on the diastereoselectivities. c Kinetic experiments of 1a and 1a’. d Monitoring of the reaction process of 1a and 1a’ under standard conditions. e Effect of pH on the relative quantity of enol and keto form of 1u.

Full size image

The mechanistic investigations and theoretical data57,58 accumulated for the ATH of 1 point to two plausible catalytic cycles. Key results from DFT calculations, including the most favorable reaction pathways for C=O and C=C reduction, respectively, are shown in Fig. 6. Active catalyst cat1 was generated from decarboxylation of the formate complex of cat.6 (Supplementary Fig. 11 and Supplementary Data 1). In the catalytic cycle for C=O reduction, cat1 interacts with substrate keto-1g to generate the intermediate a-R-int1, followed by a direct hydride (H−) transfer via transition state a-RS-TS1, with an energy barrier of 3.1 kcal/mol relative to a-R-int1. This step generates an intermediate a-RS-int2, which is more stable than a-R-int1 by 11.3 kcal/mol. Subsequently, the proton transfer step is completed via a-RS-TS2 by overcoming the energy barrier of 12.9 kcal/mol relative to a-RS-int2. The active catalyst cat1 is regenerated after releasing of the syn-product RS-pro followed by proton transfer from formic acid. Finally, considering that active catalyst cat1 first more easily capture enol-1g than keto-1g to form a more stable intermediate int1 in an exergonic process of 2.6 kcal/mol, the free energy barrier of the C=O reduction pathway from cat1 to product should be 15.0 kcal/mol from int1 to a-RS-TS1. On the other hand, for the catalytic cycle of the C=C reduction, the hydride transfer step suffered from an energy barrier of 19.7 kcal/mol via transition state S-TS1 relative to int1. The following proton transfer step has an energy barrier of 21.1 kcal/mol relative to int1. Subsequent release of the anti-product SS-pro and interaction with formic acid regenerates cat1 and produces CO2. The results show that from cat1, the free energy barrier of the most favorable pathway for C=O and C=C reduction are respectively 15.0 kcal/mol (from int1 to a-RS-TS1) and 21.1 kcal/mol (from int1 to SS-TS2), and are consistent with the experimental observation that the formation of the anti-product is favored over the syn-product. The CO2 generated in the asymmetric reduction with formic acid can be effectively removed from the catalytic system. In addition, the theoretical study of the C=C reduction pathway for the hydroxyl-protected substrate was also conducted under standard conditions (Supplementary Fig. 12 and Supplementary Data 1 present the details).

Fig. 6: Proposal catalytic cycle for ATH of 1g with HCOOH.
figure 6

The C=C reduction pathway was depicted in the left catalytic cycle and the C=O reduction pathway was described in the catalytic cycle on the right. The Gibbs free energy of the compounds, intermediates and transition states are provided below the corresponding chemical structures and the energies are given in kcal/mol.

Full size image

The free energies and key lengths of the transition states leading to different stereoselective products were also shown in Fig. 7. For the C=O reduction pathway, it is found that the transition state a-RS-TS1 has a significant advantage relative to other three transition states a-SS-TS1, a-SR-TS1, and a-RR-TS1 by 6.2, 7.5, and 10.8 kcal/mol, respectively. The proton transfer step in the C=C reduction pathway is the enantio/rate-determining step, the transition state SS-TS2 is more stable than the other three transition states RS-TS2, SR-TS2, and RR-TS2 by 9.2, 12.7, and 2.6 kcal/mol, respectively. Comparing to the reduction of the C=C, a-RS-TS1 is more favorable than the corresponding transition states SS-TS2 by 6.1 kcal/mol. The result showed that reduction of the C=O bond is the more reasonable pathway, resulting in the formation of syn-product, which can transform into the anti-product in the presence of Et3N. The reduction of the C=C path need to overcome a higher energy barrier (21.1 kcal/mol vs 15.0 kcal/mol) and form the anti-product in the catalytic cycle, which is also inconsistent with the experimental results of the formation of syn-product.

Fig. 7: The free energies of the transition states of the enantio-determining steps.
figure 7

a Calculated transition states for the C=O bond reduction pathway. b Calculated transition states for the C=C bond reduction pathway. The energies are given in kcal/mol. The lengths are given in angstrom.

Full size image

Discussion

In conclusion, a highly asymmetric transfer hydrogenation of 3-hydroxy-4-substituted-maleimide derivatives was successfully developed using a tethered Rh catalyst under mild reaction conditions. Through strategic modulation of the amount of Et3N, a variety of 3-hydroxy-4-substituted-maleimide were transformed into the corresponding syn- and anti- chiral succinimides with excellent enantio- and diastereoselectivities. This method successfully breaks the inherent impression of single product in previously reported ATH methodologies. Comprehensive mechanistic studies revealed that the -OH group at the C3-position of 1 is crucial for driving the reaction to the high enantioselectivity, and the configuration of the product will undergo epimerization during the ATH process. This also leads to two possible reduction processes (C=O or C=C). Computational analysis revealed that the C=C reduction pathway may suffer a higher energy barrier than the C=O reduction pathway, and it cannot generate the syn-product. Thus, the C=O reduction pathway via dynamic kinetics resolution process is reasonable. In addition, gram-scale experiments and varieties of transformation at high TONs provides an efficient way to synthesis chiral pyrrolidine derivatives. The present findings demonstrate successful mechanistic control to realize the ATH of a challenging substrate, which can provide further insight into the development of ATH.

Methods

Representative procedure of asymmetric transfer hydrogenation of 1a

To a 10 mL Schleck tube charged with a magnetic stirring bar were added successively substrate 1a (0.2 mmol, 56 mg), formic acid/trimethylamine azeotropic mixture (5/2) (40 μL), cat.6 (3 mg, 0.004 mmol) and the solvent (2 mL). The mixture was then stirred at room temperature for the indicated reaction time. After completion, the reaction solution was concentrated and the residue was passed through a short column of silica gel (eluent: EtOAc:PE = 2:1) to produce 2a as a white solid (55 mg, 97% yield, 99% ee, 98:2 dr). The ee or dr values of compound 2a were determined by HPLC analysis on a chiral stationary phase (Chiralpak IE column, hexane/isopropanol = 80/20; flow rate = 1.0 mL/min; UV detection at 210 nm; t1 = 8.9 min, t2 = 9.9 min, t3 = 10.6 min, t4 = 13.4 min (major).