A Pyridine Dearomatization Approach to the Matrine-type Lupin Alkaloids

Abstract

(+)-Matrine and (+)-isomatrine are tetracyclic alkaloids isolated from the plant Sophora flavescens, the roots of which are used in traditional Chinese medicine. Biosynthetically, these alkaloids are proposed to derive from three molecules of (–)-lysine via the intermediacy of the unstable cyclic imine Δ¹-piperidine. Inspired by the biosynthesis, a new dearomative annulation reaction has been developed that leverages pyridine as a stable surrogate for Δ¹-piperidine. In this key transformation, two molecules of pyridine are joined with a molecule of glutaryl chloride to give the complete tetracyclic framework of the matrine alkaloids in a single step. Using this dearomative annulation, isomatrine is synthesized in four steps from inexpensive commercially available chemicals. Isomatrine then serves as the precursor to additional lupin alkaloids, including matrine, allomatrine, isosophoridine, and sophoridine.

Graphical Abstract

The lupin alkaloids are a structurally diverse class of quinolizidine-containing natural products isolated from plants in the Lupinus genus (Fig. 1A).¹ (+)-Matrine (2), the primary component of Chinese Kushen injection, inhibits proliferation in metastatic cancer cell lines and has also been investigated as a therapeutic agent against encephalomyelitis, asthma, arthritis, and osteoporosis.^2,3 (–)-Sophoridine (4) is an approved chemotherapeutic in China, which has also demonstrated antibiotic activity.⁴ Little is known about the pharmacological properties of (+)-isomatrine (1) and (+)-isosophoridine (5), which likely reflects their limited accessibility from commercial vendors.⁵

Figure 1.

(A) Chemical structures of matrine-type lupin alkaloids. (B) Proposed biosynthesis of matrine. (C) Retrosynthetic analysis of isomatrine.

Although the detailed enzymatic pathway has not been fully annotated, the biosynthesis of matrine is proposed to initiate with the enzymatic conversion of (–)-lysine (6) to Δ¹-piperidine (7) (Fig. 1B).^6,7 Subsequent dimerization of 7 followed by oxidation and isomerization is proposed to yield quinolizidine 8, a shared biosynthetic precursor to several lupin alkaloids.^8,9 Mannich addition of 8 to a third equivalent of 7 and cyclization with the pendant aldehyde is proposed to generate the oxidized tetracycle 9, which upon reduction gives (+)-matridine (10). Seminal studies by Abdusalamov demonstrated that feeding ¹⁴C-labeled (+)-10 to Goebelia Pachycarpa resulted in the isolation of radio-labelled (+)-2, suggesting that the final step in the biosynthesis of 2 is a site-selective C–H oxidation.^10,11

Synthetically, most of the work prior to 2022 had focused on matrine (2), with four reported total syntheses.^12,13,14,15 Synthetic access to the minor congeners is far more limited: until a recent report by Sherburn and coworkers,¹⁶ there was a single total synthesis each of allomatrine (3) and isosophoridine (5), and no reported total syntheses of isomatrine or sophoridine (4).^17,18 We sought to devise a unified synthesis that could provide access to the series of matrine-type alkaloids shown in Fig. 1A. Inspired by the proposed biosynthesis, it was envisioned that pyridine (14) could serve as a stable, inexpensive synthon for Δ¹-piperidine (7), and the remaining five carbons of the tetracyclic matrine framework could derive from glutaryl chloride (15, Fig. 1C). In the key step, we proposed a dearomative annulation via bis-acyl pyridinium salt 17 to form tetracycle 12, a molecule that contains all the carbon and nitrogen atoms of 1.¹⁹ Tetracycle 12 could be elaborated to 1 by global reduction followed by a site-selective oxidation of isomatridine (11) reminiscent of the proposed biosynthesis of matrine (2). Isomatrine (1) is the least thermodynamically stable lupin alkaloid and its isomerization to both 2 and 3 has been previously reported.²⁰ We therefore anticipated that access to 1 could enable the synthesis of additional lupin alkaloids.²¹ This type of late-stage isomerization strategy was also deployed in the 2022 Sherburn synthesis of several matrine alkaloids.¹⁶

Our studies commenced with the investigation of the dearomative annulation (Fig. 2A). Addition of glutaryl chloride (15) to pyridine (14) in dichloromethane at −50 °C followed by warming to 20 °C resulted in clean formation of (±)-tetracycle 12 in 62% yield (10 g scale). The reaction was highly robust and could be carried out on one mole scale to produce over 160 grams (67% yield) of (±)-tetracycle 12 in a single batch. The product was isolated by precipitation from the crude reaction mixture, alleviating the need for a workup or column chromatography. Given the cost of pyridine ($7/mol), glutaryl chloride ($211/mol), and all solvents ($31/mol), the raw materials cost $398/mol of product formed.²² Recrystallization of (±)-12 enabled single crystal X-ray diffraction, which confirmed the syn-syn relative stereochemistry.²³

Figure 2.

(A) Diastereoselective cyclization of pyridine and glutaryl chloride. (B) Diastereoselective cyclization of (R)-2-methylpentandioyl chloride. (C) Proposed mechanism for the diastereoselective cyclization. (D) Reaction coordinate diagram – performed with Gaussian using ωB97XD/def2-TZVP/SMD(DCM).

To elucidate the reaction pathway, mechanistic and computational studies were undertaken. Monitoring the reaction between 14 and 15 by ¹H NMR determined that the major species at −40 °C was bis-acyl pyridinium salt 17 (Fig. 2C). After warming to 25 °C, acid chloride 18 resulting from mono-cyclization was observed. Presumably the acyl pyridinium salt Int1a and acyl chloride 18 are in equilibrium, but the acyl chloride is the major species at 25 °C. A second, minor species assigned as the acid chloride resulting from Int1b (vide infra) was also observed; this species was consumed as the reaction progressed to full conversion.²³ Although deprotonation of acyl pyridinium salts or acyl chlorides can give rise to ketene intermediates,²⁴ no such species was detected by ¹H NMR or by reactIR. Attempts to calculate a pathway involving ketene intermediates failed to locate a transition state (TS) for a concerted [2+2] cycloaddition. Similarly, no TS for the concerted [4+2] cycloaddition of bis-acyl pyridinium salt 13 could be located. Investigation of a stepwise pathway determined that the lowest-energy TS for the first cyclization involves a boat-like conformation to form the syn product (TS1a, ΔG_TS = 7.3 kcal/mol) (Fig. 2D). Attempts to find the analogous chair-like TS were unsuccessful and led instead to conversion to the boat-like TS. The pathways leading to the anti mono-cyclization product (Int1b) are higher in energy (see TS1b and TS1c). The preference for the syn boat compared to the anti boat TS is likely due to favorable dispersive interactions between the heteroaryl ring and the oxygen-bearing carbon of the enolate, as well as minimization of the dipole moment in the syn TS. In order to test the importance of dispersive interactions in these TSs, the TSs were recomputed with B3LYP, a functional known to lack dispersion. Indeed, with this functional, the difference between the two transition states was only 0.1 kcal/mol. Inclusion of dispersion with Grimme’s D3 correction restored the energy difference to 1.4 kcal/mol in favor of the syn boat transition state.

These TSs lead to two intermediates: syn intermediate Int1a (–12.6 kcal/mol) and anti intermediate Int1b (–10.7 kcal/mol). The TS for the second C–C bond formation (TS2a) is most favorable for the syn-syn intermediate (Int2a), with a barrier of 20.2 kcal/mol. The second lowest-energy pathway proceeds via TS2b leading to Int2b, which gives rise to the anti-syn-anti configuration at the ring fusions. The transition states leading to the other four potential diastereomers are higher in energy. Formation of Int2a and Int2b is followed by deprotonation by pyridine. While Int2b is lower in energy than Int2a, the deprotonation of Int2a to give syn-syn (±)-12 follows the lowest-energy pathway. Thus, the selectivity-determining step is the final deprotonation (TS3a) and syn-syn (±)-12 is favored, even though it is thermodynamically less stable than anti-anti 19. These results are consistent with the experimentally observed formation of product (±)-12 as a single diastereomer, despite the initial mixture of monocyclization products. When enantiopure (R)-2-methylpentandioyl chloride was employed, C3-methyl tetracycle (+)-16 was obtained in 66% yield as a single diastereomer and in >99% ee (Fig. 2B). This stereochemical outcome is consistent with the calculations, where the pathway initiating with a syn boat transition state bearing the methyl group in a pseudo-equatorial position is favored.

At this stage, attention turned to elaborating (±)-12 to isomatrine (1) (Fig. 3A). Hydrogenation of tetraene 12 proceeded smoothly followed by reduction with alane to give (±)-isomatridine (11) in 60% yield over two steps. Purification was readily accomplished by generating the hydrogen oxalate salt followed by trituration in acetone, obviating the need for column chromatography. At this stage, resolution of diamine (±)-11 can be achieved by recrystallization of the di-p-toluoyl tartaric acid salt to give 24% recovery (46% theoretical yield) of the desired (+)-diamine 11 in 90% ee.

Figure 3.

(A) Complete total synthesis of isomatrine. (B) Attempted oxidation reactions.

Inspired by the proposed biosynthesis,²⁵ we initially investigated the enzymatic oxidation of (+)-11 to give (+)-1. Unfortunately, a screen of >180 bacterially derived P450 enzymes (both wild type and mutants) failed to produce any promising leads. As a result, our focus turned towards non-enzymatic methods for the selective oxidation of C15. Analysis of the X-ray structure of diamine 11 revealed that N1 points into the cavity of the molecule while the N2 lone pair points outwards. Selective oxidation of N2 was achieved by treatment of diamine 11 with peracetic acid to yield the expected N-oxide 20 in a 54% yield. However, attempts to advance 20 via Polonovski reaction (acetic anhydride/di-tert-butyl-4-methyl pyridine (DTBMP)) led to formation of the undesired enamine 21.²⁶ Analysis of the X-ray structure of 20 confirmed that antiperiplanar alignment of H17 with the N–O bond is ideally suited to regioselectively form the undesired elimination product 21.

We became interested in a report by Kessar and coworkers demonstrating that amine–BF₃ adducts could undergo deprotonation using mixtures of tert-butyl lithium (t-BuLi) and potassium tert-butoxide (t-BuOK).²⁷ Consistent with the selectivity in the N-oxide formation, treatment of diamine (+)-11 with BF₃·OEt₂ quantitatively formed the Lewis acid-base complex. Deprotonation of the BF₃ complex of (+)-11 with a mixture of t-BuLi and t-BuOK in N,N,N,N-tetramethylethylenediamine (TMEDA) occurred with good selectivity for the less sterically encumbered C15 over C17 (10:1), as determined by trapping with deuterated methanol (22, Fig. 3B).²⁸ Unfortunately, trapping of this anion with other electrophiles proved more challenging. For example, deprotonation followed by quenching with TMSCl provided silylated diamine 23 in only 35% yield, while trapping with methyl benzoate gave unstable phenyl ketone 25 in 27% yield. The best yield of C15-functionalized product was obtained when 11 was deprotonated and then trapped with trimethyl borate; oxidation with hydrogen peroxide and trapping of the resultant enamine with HCN gave aminonitrile 24 in 55% yield. Aerobic oxidation of 24 provided isomatrine in 46% yield (25% yield over three steps from 11).^{23, 29} Alternatively, deprotonation of (+)-11, trapping with methyl benzoate, and aerobic oxidation could be carried out in a single reaction flask to give (+)-1 directly in 18–26% yield, depending on the scale.³⁰ This route provides access to (±)-isomatrine in four steps, and (+)-isomatrine can be easily accessed by incorporating the resolution of diamine 11. To date, >1 gram of (+)-isomatrine has been prepared.

Initial attempts to reproduce Okuda’s Pt-catalyzed isomerization of (+)-isomatrine failed to provide the reported yields of (+)-matrine (2) and (+)-allomatrine (3), and instead produced a mixture of five compounds.²⁰ To improve the yield of 2 and 3, while also broadening the synthetic access to other congeners, an investigation of several isomerization catalysts was carried out. Use of Rh/C provided the best yields of (+)-2 (32% yield, Fig. 4), while (+)-3 could be obtained in 83% yield when Pd/C was used. Isomerization with Pt/C provided (+)-isosophoridine (5) in 55% yield. Finally, use of PtO₂ at 98 °C for 15 minutes furnished (–)-sophoridine (4) in 10% yield, together with the other isomers.^31,32 When the reaction with PtO₂ was conducted at 80 °C for 24 hours, (–)-isomer 26 was isolated in 40% yield. To our knowledge, 26 has not yet been isolated from natural sources.

Figure 4.

Isomerization of (+)-isomatrine to additional matrine-type lupin alkaloids. Highlighted values are the isolated yields of the indicated products.

The bioinspired dearomative annulation between pyridine and glutaryl chloride developed here has enabled the first total synthesis of the lupin alkaloid (–)-sophoridine, and the shortest syntheses of (+)-isomatrine, (+)-matrine, (+)-allomatrine, and (+)-isosophoridine reported to date. The brevity of this route results from the ability to construct the entire carbon framework of the matrine-type alkaloids in a single step. The diversity of matrine-type alkaloids prepared from commodity chemicals is anticipated to support future pharmacological investigations.