Non-enzymatic methylcyclization of alkenes

In nature methyltransferases (MT) play an integral role in cell signalling, gene regulation and the biosynthesis of secondary metabolites^{1,2,3,4,5,6,7,8,9,10,11,12,13,14}. For the latter, the selective methylation of alkenes by S-adenosyl methionine (SAM)-dependent C-methyltransferases provides access to new chemical space and allows escape from the canonical C₅-isoprene pattern of terpenoid carbon scaffolds. In most cases, C-methyltransferases catalyse the methylation of simple linear building blocks, such as isopentenyl pyrophosphate and geranyl pyrophosphate^{6,7,8,11,12,15,16}. Sterol methyltransferases introduce methyl groups onto linear alkenes that are linked to a polycyclic backbone^9,10. Bifunctional methyltransferase–cyclases (bMTC) stand out due to their ability to induce cyclization by an initial methyltransfer (methylcyclization), which allows for the rapid build-up of molecular complexity. So far, only two bMTCs have been identified: an MT in the sodorifen biosynthesis found in Serratia plymuthica and TleD in the teleocidin biosynthesis^13,14,17. TleD catalyses the SAM-dependent conversion of lyngbyatoxin A to teleocidin B-4 (Fig. 1a). Another methyltransferase–cyclase was proposed by Jaenicke and co-workers in 1988 for the biosynthesis of the iridals, which release irones through oxidative degradation¹⁸. These valuable degradation products are found in exclusive fragrances and are the subject of numerous patents and publications^{19,20,21,22,23,24,25}. While the responsible methyltransferase still remains elusive, the groups of Chen and André recently engineered TleD to efficiently convert psi-ionone to cis-α-irone in a process similar to the proposed biosynthesis (Fig. 1b)²⁶. Although there have been substantial advances in the C methylation of alkenes^{27,28,29,30,31,32,33,34,35,36,37,38,39}, especially hydromethylation, non-enzymatic methyltransfer-initiated cyclization reactions have remained elusive and, until now, access to methylated, non-canonical terpenoids has required lengthy chemical syntheses or knowledge of a suitable methyltransferase–cyclase. To the best of our knowledge, there are only two reports describing transformations akin to those catalysed by C-methyltransferases. In 1984, Barradas and co-workers disclosed an electrochemical protocol employing sodium acetate as the methyl donor to convert oct-1-ene to a complex mixture of mono- and dimethylated products³⁹. Notably, cyclohexene, an alkene with a lower oxidation potential, was not compatible. Shortly thereafter, Julia and Marazano disclosed conditions employing diaryl methyl sulfonium salts at high temperatures (135–170 °C) to convert 2-methyloct-2-ene to a mixture of eight mono- and dimethylation products³⁶. Further investigations of heteroaromatic sulfonium salts by the groups of Julia and Shiraishi failed to improve the low efficiency and poor selectivity^37,38. The reported methods are limited to pure hydrocarbons void of any functional group other than a single alkene. This is in stark contrast to C-methyltransferases in nature, which operate undisturbed in the presence of various nucleophilic functionalities, such as heteroatoms. To fill this longstanding gap in the synthetic toolbox, we envisioned the development of an electrophilic methylation system, which exhibits high chemoselectivity towards alkenes over arenes and other nucleophiles—a challenging endeavour owing to the comparably low nucleophilicity of alkenes (Fig. 1c). Carbocations generated in this way would be primed for a subsequent nucleophilic attack (cyclization) by a nearby C(sp²) system. Herein, we report the development of a non-enzymatic methyltransfer–cyclization (NEMTC) reaction, which proceeds not only at ambient temperatures with commercially available reagents, but also exhibits a wide functional group tolerance, including the presence of oxygen-, nitrogen- and sulfur-based functionalities. Products of this bioinspired transformation feature unique, semisaturated structural motifs with two new C–C bonds along with a quaternary carbon centre, which would otherwise require step-intensive preparations. This powerful methodology enables access to a wide area of unexplored chemical space (methyl-substituted polycycles, including bi-, tri-, tetra- and spirocycles) through variation of the initiating group (for example, substitution degree of the alkene), the terminating groups ((hetero)arenes, alkenes) and the linker chain (for example, mono- versus bicyclizations).

a, The natural bifunctional methyltransferase–cyclase TleD transforms lyngbyatoxin A to teleocidin B-4. The natural methyl donor SAM is converted to S-adenosyl homocysteine (SAH) upon losing a methyl group^13,17. b, An engineered bifunctional methyltransferase–cyclase converts psi-ionone to cis-α-irone, the major olfactory component of Irone Alpha²⁶. c, This work discloses a non-enzymatic methyltransfer–cyclization reaction that uses silver hexafluorophosphate and methyl iodide instead of enzymes and SAM. Numbers in transition states indicate selected distances in angstrom. d, Development of the NEMTC reaction. NMR yields are given. Choice of the silver salt and base are pivotal for product selectivity and reaction efficiency (see Supplementary Sections 2.1.2 and 2.1.3). Temp., temperature.

Full size image

For reaction development we subjected alkene 1 to a variety of classical electrophilic methylating reagents (Fig. 1d). While reagents reminiscent of SAM, namely trimethylsulfonium iodide (Me₃SI) and its sulfur(VI) analogue trimethylsulfoxonium iodide (Me₃S(O)I), remained unsuccessful, the use of methyl triflate (MeOTf) and trimethyloxonium tetrafluoroborate (Me₃OBF₄, Meerwein’s salt) was encouraging as, for the first time, traces of methylcyclized product 2 (2–5%) were obtained (Fig. 1d, entries 1 and 2). In an effort to improve the yield of 2, we turned our attention to the seminal work of Meerwein, which describes the synthesis of trialkyloxonium tetrafluoroborates from dialkylethers, alkyl halides and silver(I) tetrafluoroborate^40,41. Subjection of alkene 1 to a combination of MeI and AgBF₄ increased the yield of methylcyclized product 2 to 9% together with protocyclized product 3 (73%) (Fig. 1d, entry 3). Introduction of 2,6-di-tert-butyl-pyridine (B2), an exceptionally hindered base resistant to methylation⁴², prevented the competing protocyclization and improved the yield for the desired product 2 to 62% (Fig. 1d, entry 4). Unexpectedly, dimethylcyclized product 4 (19%) was also formed under these conditions. We speculate that 4 originates from a premature deprotonation allowing for a second methylation and subsequent cyclization. To probe the influence of the base on premature deprotonation, B2 was exchanged for 2,4,6-tri-tert-butyl-pyrimidine (B1) (Fig. 1d, entry 5). Although the exchange from B1 to B2 had no substantial influence on the formation of the dimethylcyclized product 4 (22%), we were pleased to obtain 2 in a modestly increased yield of 70%. Other bases, such as B3, B4, B5 and B6 suffered from low conversions due to competitive N methylation. Finally, we investigated the role of the silver(I) counterion on the formation of undesired product 4. From this screen, silver(I) hexafluorophosphate emerged as ideal, affording methylcyclized product 2 in 80% yield without detectable traces of dimethylcyclized product 4 (Fig. 1d, entry 6) while various other silver(I) salts exhibited low solubility and slow conversion rates. Lastly, reducing the reaction temperature to 0 °C afforded methylcyclized product 2 in 85% yield (Fig. 1d, entry 7, 84% yield after isolation).

With the optimized conditions in hand, we investigated the scope of this reaction while varying the temperature to account for substrate-dependent reaction rates. We found that para-, meta– and ortho-methoxy substituents on the arene provided the desired tetralins 5–7 in 38–71% yield (Fig. 2a). Unexpectedly, unselective Friedel–Crafts-type arene methylation was observed as a side reaction, especially for 6 and 7. Gratifyingly, the use of B2 was able to partially suppress this reactivity, increasing the yield for 6 from 38% to 47% (see Supplementary Section 2.5.1). The presence of an additional F or TfO substituent was well tolerated providing 8 and 9 in 67% and 94% yield, respectively. To our delight, a 1,3-dioxole moiety gave the cyclized product 10 in 72% yield without apparent arene methylation. A para-bromo substituent afforded tetralin 11 in excellent yield (89%). An aniline required prior protection with a sulfonyl group (2,4,6-(i-Pr)₃PhSO₂), but otherwise underwent the desired NEMTC reaction with concomitant N methylation to 12 (42% yield). Exchanging the arene for a naphthalene gave methoxy-substituted tetrahydroanthracene 13a/b in 66% yield. The use of cyclic alkylidenes enabled access to the tricyclic spiro-motifs 14 (79%), 15a/b (71%) and 16a/b (71%) in good yields. Introduction of a protected piperidine led to formation of spirocycle 17 (23%) but was accompanied by several unidentified side products. Subjecting cyclopropylidene 18 to the reactions conditions provided tricycle 21 (30%), whose 6/4-subunit is featured in the protoilludane scaffold (22) of natural products⁴³. The formation of 21 can be rationalized through formation of the most stable cyclopropyl cation 19, followed by Wagner–Meerwein shift to 20 and nucleophilic attack by the arene. For the cyclobutylidene, both methylation and cyclization to the spirocycle 23a/b and ring expansion to tricycle 23c were observed. Interestingly, similar reactivity was observed when formation of a seven-membered ring was attempted (Fig. 2b). Exposure of alkene 24 to the reaction conditions led to exclusive formation of tetralin 26, which features a quaternary centre substituted with an isopropyl and a methyl group. We assume that the initially formed tertiary cation 25 undergoes a 1,2-hydride shift followed by six-membered ring closure. This sequence is in accordance with the proposed biosynthesis of teleocidin B-4 (28)¹³. For the analogous para-methoxy substrate the same transformation leading to 27a/b (76%) was observed. With regard to the degree of alkene substitution, a 1,1-disubstituted alkene was suitable for productive methyltransfer leading to the formation of tetralin 29a/b (34%) via a methylation/1,2-hydride shift/cyclization cascade. A tetrasubstituted alkene also underwent NEMTC reaction to give 30a/b in 63% yield (for the comprehensive substrate scope, including limitations such as ring sizes, alkene substitution and different initiating groups, see Supplementary Sections 2.2 and 2.3).

a, Investigation into arene substituents and cyclic alkylidenes to access methylated bi-, tri- and spirocyclic compounds. For cyclopropylidenes and cyclobutylidenes ring expansions are observed. b, Investigation into substrates favouring the formation of six-membered rings via 1,2-migration, as well as an exploration of alternative alkene substitution patterns. Reactions were performed on 0.125 mmol scale and all yields are isolated. For substrates with additional arene methylation, the direct product is designated with ‘a’, whereas the arene-methylated product is designated with ‘b’. ^aB1 used. ^bB2 used. ^cFrom free N–H with increased reagent equivalents.

Full size image

Further investigation of the potential of this reaction prompted us towards substrates with increased complexity. Prenylated phenols and anilines were well tolerated affording chromane 31 (63%) and tetrahydroquinoline 32 (68%) in good yields (Fig. 3a). To our delight, both a furan and an N-sulfonylated pyrrole were competent terminating groups, affording the corresponding bicycles 33 (48%) and 34 (21%). N-Sulfonylated indoles, a benzofuran and a benzothiophene were also suitable substrates for the NEMTC reaction, furnishing the desired tricycles 35, 36, 37 and 38 in yields ranging from 58 to 79%. We were pleased to find that 2-geranyl- and 3-geranyl-substituted indoles underwent bicyclization to afford tetracycles 39a/b and 40a/b in good yields (53–66%) (Fig. 3b). Pure 40a could be obtained through crystallization, allowing structure validation by single-crystal X-ray analysis. Based on the importance of the family of irones for the perfume industry, we also explored multiple functionalized geranyl cyclization precursors to access the irone motif (Fig. 3c). To our surprise the choice of the functional group had a dramatic effect on the relative stereochemistry (cis versus trans) and the alkene regioselectivity (α, trisubstituted; β, tetrasubstituted and γ, 1,1-disubstituted). Subjecting 2,4,6-triisopropylbenzoyl-protected geraniol to the NEMTC conditions resulted in a mixture of cis– and trans-α-cyclohexenes 41 (30%). Notably, the sterically encumbered ester is not completely tolerated as methyl 2,4,6-triisopropylbenzoate was also isolated. Surprisingly, a phthalimide moiety was superior in the reaction and furnished methylated β-cyclohexene 42 in good yield (65%) and high selectivity, along with cis-γ-cyclohexene 42 (11%). Employing a phenyl sulfone instead afforded a mixture of products 43 (α, β, γ), including the cis and trans isomers in a combined yield of 66%. The formation of β-43 represents a shortened formal synthesis (five steps) of β-irone (46), which was previously synthesized in eight steps from geraniol (44) (Fig. 3d)²⁰. While the previously established protocol requires substantial skeleton editing before cyclization, the NEMTC platform makes direct use of canonical terpenoid buildings block (two steps to 43 from geranyl bromide (45)).

a, Construction of heterocycles and termination by heteroarenes. b, Bicyclization of indoles. c, Synthesis of functionalized methylated cyclohexenes bearing the irone core motif. For simplicity only the major product is depicted. d, Streamlined formal synthesis of β-irone (46). Reactions were performed on 0.125 mmol scale and all yields are isolated. ^aB1 used. ^bB2 used. ^cInseparable.

Full size image

The observed ring expansions and 1,2-hydride shifts during our substrate scope investigations led us to propose a cationic reaction pathway. To test this hypothesis and gain further insights into the reaction mechanism, we conducted density functional theory calculations (wB97XD⁴⁴/def2-TZVP^45,46(IEFPCM(CH₂Cl₂))). Starting from an implicitly solvated silver(I) ion (I), we compared various silver(I) complexes containing at least one methyl iodide ligand, which are expected to resemble reactive methylating species (Fig. 4a). Indeed, for complex [Ag(IMe)]⁺ (II) and complex [Ag(B1)(IMe)]⁺ (III) we were able to identify energetically comparable transition states, TS-II and TS-III, both of which are presumably operative (the energy difference is below the accuracy of density functional theory). A comprehensive overview of additional silver(I) complexes and transition states can be found in Supplementary Section 3.2 and Supplementary Figs. 1 and 2. Methyltransfer proceeds with an S_N2-type mechanism with the methyl group adopting a trigonal planar conformation with elongated perpendicular ‘bonds’ to the alkene and the iodine. Assuming a Curtin–Hammett-type scenario with comparatively fast interconversion between II and III, the difference in reaction rate is only determined by ΔΔG^‡ indicating approximately twofold faster conversion via TS-III compared to TS-II (ΔΔG^‡ = 0.5 kcal mol⁻¹). Of note, during the reaction development a combination of AgBF₄ and MeI was found to be sufficient for product formation, indicating that B1 is not essential for methyltransfer. Dissociation of silver(I) iodide (and B1) gives the tertiary cation IV, which is poised to undergo nucleophilic attack by the arene (TS-IV, ΔG^‡ = 5.8 kcal mol⁻¹) to form the Wheland intermediate V (Fig. 4b). Search for an alternative concerted methyltransfer–cyclization transition state resulted only in identification of a higher lying methyltransfer transition state (see Supplementary Section 3.2 and Supplementary Fig. 2). The profound influence of the silver(I) counterion on the formation of dimethylcyclized product 4 (compare Fig. 1) indicated that the counterion may play a crucial role for the deprotonation rather than B1. Indeed, deprotonation of V by pyrimidine B1 to rearomatized tetralin 2 and pyrimidinium [HB1]⁺ (VI) exhibits an exceedingly high transition state (TS-V, ΔG^‡ = 18.1 kcal mol⁻¹) compared to deprotonation from the contact ion pair VII (adduct of PF₆⁻ and V) via TS-VII (ΔG^‡ = 7.4 kcal mol⁻¹). TS-VII collapses to tetralin 2, PF₅ and HF (VIII), which undergoes proton exchange with B1 to regenerate PF₆⁻ (VI). In agreement with our reaction development, BF₄⁻ exhibits a substantially lower deprotonation barrier (see Supplementary Section 3.2 and Supplementary Fig. 3), which presumably causes premature deprotonation of tertiary cation IV leading to dimethylcyclized product 4. Overall, the proposed mechanism involves three distinct steps: (1) rate-limiting methyltransfer; (2) cyclization; and (3) deprotonation/rearomatization. Notably, pyrimidine B1 has little effect on the methyltransfer step and is primarily required to act as the terminal proton acceptor while being resistant to N methylation.

a, Complexation of the silver(I) ion (I) with either methyl iodide to complex II or methyl iodide and B1 to complex III generates electrophilic methylating agents. Methyltransfer to alkene 1 proceeds via TS-II and TS-III through an S_N2-type mechanism and was found to be rate limiting. Tertiary cation IV and silver(I) iodide are formed after complex dissociation. b, Tertiary cation IV undergoes nucleophilic attack by the arene via TS-IV to give Wheland intermediate V. Deprotonation of Wheland intermediate V by base B1 via TS-V is disfavoured compared to deprotonation via contact ion pair VII (adduct of PF₆⁻ and V) and TS-VII. The choice of the counterion is crucial for controlling the deprotonation kinetics. The conformational space was explored using CREST⁴⁹ and low energy conformers were further refined with wB97XD/def2-TZVP(IEFPCM(CH₂Cl₂)). For transition states selected distances are given in angstrom.

Full size image

A non-enzymatic alternative to bMTCs found in nature has been developed, directly enabling the generation of the corresponding naturally occurring substitution patterns. The NEMTC reaction reported here represents a fundamental advancement in (poly)cyclization methodology and provides simple and selective access to methylated spiro-, bi-, tri- and tetracyclic compounds using commercially available reagents. The unique products of this transformation may be especially interesting with regard to the profound efficiency boost often observed in drug discovery programmes upon introduction of methyl groups (‘magic methyl effect’)^47,48. The impact of this methodology is further emphasized by its compatibility with a diverse array of substrates and its complementarity to previously established modes for C methylation (hydromethylation of alkenes and acidic C–H alkylation). Given its simplicity, ease of operation and uniqueness, we expect widespread application and further research of the NEMTC reaction.

General procedure for the non-enzymatic methyltransfer–cyclization reaction

A vial was charged with silver hexafluorophosphate (AgPF₆, 63.2 mg, 250 µmol, 2.00 equiv.) in the glovebox and sealed under argon atmosphere using a rubber septum. To this vial was added, in succession, a solution of base B1 or B2 (100 mM in dichloromethane, 2.50 ml, 250 µmol, 2.00 equiv.), a solution of the alkene (50.0 mM in dichloromethane, 2.50 ml, 125 µmol, 1 equiv.) and a solution of methyl iodide (625 mM in dichloromethane, 600 µl, 375 µmol, 3.00 equiv.) at 0, 10 or 23 °C. The reaction was monitored by thin-layer chromatography or NMR spectroscopy and stopped by addition of triethylamine (100 µl, 717 µmol, 5.74 equiv.) after no further conversion was observed. The reaction mixture was stirred for 10 min at 23 °C, after which the solvent was removed under reduced pressure. The residue was purified either by flash column chromatography on silica gel or by semipreparative normal-phase high-performance liquid chromatography to afford the methylcyclized product.