Synthetically established methods for methylation of phenols and demethylation of methyl phenyl ethers rely in general on hazardous reagents or/and harsh reaction conditions and are irreversible. Consequently, alternative regioselective methods for the reversible formation and breakage of C-O-ether bonds to be performed under mild and sustainable conditions are highly desired. Here we present a biocatalytic shuttle concept making use of corrinoid-dependent methyl transferases from anaerobic bacteria. The two-component enzymatic system consists of a corrinoid protein carrying the cofactor and acting as methyl group shuttle, and a methyltransferase catalyzing both methylation and demethylation in a reversible fashion. Various phenyl methyl ethers are successfully demethylated and serve in addition as sustainable methylating agents for the functionalization of various substituted catechols. Therefore, this methyl transfer approach represents a promising alternative to common chemical protocols and a valuable add-on for the toolbox of available biocatalysts.
From a synthetic point of view formation as well as cleavage of alkyl aryl ethers are valuable transformations contributing to the structural diversity of pharmaceuticals and natural products1,2,3. Beyond that, O-methylated phenol derivatives are highly demanded building blocks for the manufacture of antioxidants, flavoring agents, fragrances, dyes, agrochemicals, and fine chemicals4,5. Although numerous chemical procedures6 for methylation and demethylation are available (Fig. 1) and have been broadly applied, none of these protocols allows for reversible C-O-ether bond formation and breakage on synthetic scale. Conventionally, the addition of methyl moieties onto heteroatoms is accomplished with methyl iodide or dimethyl sulfide as methylating agent together with a strong base7,8. Besides toxicity and corrosiveness of the reagents needed in excess and the sizable amounts of inorganic salts formed as by-products, these traditional procedures are in general neither chemo- nor regioselective in case substrates possess more than one hydroxyl group or other reactive functional moieties (e.g., amine). Attempts to improve or even replace the classical Williamson ether synthesis led to the development of methods making use of less hazardous alkylating agents such as dimethyl carbonate9,10 or methanol11,12 which, however, still require high temperatures for reaction. Due to the high stability of O-methyl ethers against hydrolysis also their cleavage usually involves harsh reaction conditions and reagents such as strong Bronsted13,14 or Lewis acids15,16. Alternatively, the use of metal catalysts17,18 or nucleophilic methods employing for instance sodium thiolates19,20 belong to the most commonly applied chemical techniques.
In the last decades, biocatalysis has gained increasing attention as promising research field at the interface of organic chemistry and biotechnology, complementing or even replacing conventional organic synthesis methods by more selective, mild and sustainable procedures21,22,23. Due to nature’s manifold source of enzymes supported by the rapid development in molecular biology and gene technology, there is no limitation in accessing novel biocatalysts and thus, the repertoire of biocatalytic reactions is constantly increasing and broadening24,25,26. Since the transfer of methyl groups is an indispensable reaction in many aspects of life, methyl transferases (MTases) are ubiquitously distributed in nature27,28 and according to the required cofactor two main enzyme classes have evolved (Fig. 2). The most prominent one relies on S-adenosyl-l-methionine29,30 and typically catalyzes the methylation of C-, O-, N-, or S-nucleophiles through a SN2 mechanism by simultaneously generating S-adenosyl-homo-cystein (SAH) as by-product. Alternatively, methylation can take place at non-nucleophilic carbons involving a radical-based mechanism31,32. Even if well-understood and providing exceptional access to a range of methylated compounds, SAM-dependent methyl transfer occurs in an irreversible fashion and thus, the cofactor is required in stoichiometric amounts hindering its feasibility in synthetic processes. In spite of the substantial progress that has been made in this research area1,33,34,35 including the use of artificial SAM-analogs36 and an in-vitro SAM-regeneration system37, there is still need for a scalable alternative. Much less investigated is the class of corrinoid-dependent MTases38,39. Corrinoids share the structural motif of a corrin skeleton, a heterocyclic system composed of four pyrrole rings. The most prominent members of this group are cobalamins40,41 (B12-derivatives, Fig. 2b), in which the unique central cobalt-carbon bond is key to enable chemically challenging biotransformations42. Methyl cobalamin (Me-Cbl)-dependent MTases have been described to mediate the transfer of a methyl cation received upon heterolytic cleavage of the Co-C bond to nucleophilic substrates. The corrinoid cofactor cycles thereby between the CoIII and the highly reactive CoI state43 and thus functions as methyl group shuttle (Fig. 3a). Nature exploits this methyl transfer mainly for O-demethylation, playing an essential role in the amino acid metabolism of eukaryotes as well as in the energy generation of anaerobic organisms39,44,45. For instance, several acetogenic bacteria have been described to utilize methyl phenyl ethers as carbon and energy source, amongst others Acetobacterium dehalogenans46, Acetobacterium woodii47, Moorella thermoacetica48, and Desulfitobacterium hafniense49,50. These organisms encode a complex O-demethylation machinery constituted of four enzymes (Fig. 3b), mediating a methyl group transfer from a substrate onto tetrahydrofolate (FH4) via two half-reactions. In the first half reaction, MTase I binds the substrate and mediates both ether cleavage and transfer of the methyl group onto CoI of the super-reduced cofactor bound to the corrinoid protein (CP). Then, a second MTase (MTase II) transfers the methyl group from CoIII-CP onto FH4, being further metabolized by the organism. The fourth protein acts as an activating enzyme (AE) to catalyze the ATP-driven reduction of the oxidized cofactor in the inactive CoII redox state to the catalytically active CoI species. While most of the MTase systems described in the literature have primarily been studied with respect to their physiological role using crude cell extracts of wild-type organisms, just a few of them51,52 have been recombinantly produced and characterized from a biochemical point of view. Even though they displayed promising demethylation properties, their feasibility as biocatalysts for synthetic purposes has not been evaluated yet.
In the present study, we report the use of a cobalamin-dependent methyltransferase system for a shuttle catalysis concept enabling both methylation and demethylation reactions in a reversible manner. To the best of our knowledge, such a system has not previously been investigated for synthetic biocatalytic applications and may represent an alternative to SAM-dependent enzymes as well as related chemical protocols.
Concept of reversible methyl transfer
For the design of an alternative biocatalytic demethylation and methylation approach inspired by the corrinoid-dependent methyltransferase system from anaerobic bacteria, we envisioned a simpler concept than shown in Fig. 3b. It relies on the use of just one methyltransferase being the only biocatalyst together with a corrinoid protein (CP) carrying the cobalamin prosthetic group and acting as methyl group shuttle. In nature MTase I is described to transform more than one substrate, whereas MTase II is restricted to the methylation of tetrahydrofolate for the purpose of energy generation. Therefore, due to the reversibility of catalysis we propose to employ MTase I both for demethylation and methylation, coupling the cleavage of a methyl phenyl ether with the methylation of unnatural acceptor molecules in shuttle catalysis like fashion53,54,55 (Fig. 4). Thus, MTase I catalyzes the demethylation of a methyl donor molecule transferring the methyl group onto the CP bound cobalamin; in the subsequent methylation step the methyl group is transferred from the cofactor to the methyl acceptor again mediated by MTase I to yield the methylated acceptor.
Employing a single enzyme (MTase I) for both half reactions offers considerable benefits as the number of required enzymes is reduced. Furthermore, the system can be employed for the demethylation as well as the methylation of a substrate. An undesired oxidation of CoI to CoII by molecular oxygen and thus an inactivation of the catalytic shuttle system is prevented by working in an inert atmosphere (glove-box).
Enzyme selection and production
The methyltransferase MTase I and the corrinoid protein CP from Desulfitobacterium hafniense52 (D. hafniense) were selected as model proteins for investigation of the suggested concept. The proteins were heterologously expressed in E. coli BL21 (DE3) pLysS using synthetic codon-optimized genes cloned into the pASK-IBA3plus vector. The use of this vector enabled the production of C-terminal Strep-tag®-fusion-proteins, which could be purified by affinity chromatography. Similarly, MTase II from D. hafniense was recombinantly produced in E. coli in order to verify the natural reaction in vitro. For a detailed description of cloning strategy and expression results, see Supplementary Fig. 1–3 and Supplementary Table 1. Since E. coli is lacking the ability to synthesize any of the cobalamin derivatives56, the corrinoid cofactor was incorporated into the apo-protein after expression in order to yield active holo-CP51.
Characterization of the MTase system from Desulfitobacterium hafniense
Having the required enzymes in hand the MTase system from D. hafniense was tested for its applicability for biocatalytic methyl transfer. Based on the reported demethylation properties of MTase I52, guaiacol 1a was chosen as model methyl donor. Since it is obtainable from renewable resources57,58 it represents an attractive substrate to demonstrate the concept and serves as a cheap and sustainable methylating agent. Due to the oxygen and light-sensitive cobalt cofactor all biotransformations were performed under inert atmosphere and light exposure was kept to a minimum. Initially, titanium(III) citrate and methyl viologen were used to reduce inadvertently formed CoII to the active CoI species. However, in an oxygen-free environment the use of this O2-scavenger system became obsolete and was omitted if not indicated otherwise.
As a first step the natural reaction involving both MTases (I and II) and CP was successfully reconstituted in vitro by transferring the methyl group from 1a onto the natural substrate tetrahydrofolic acid (see Supplementary Fig. 4 and Supplementary Table 2). Then, MTase I was investigated to act as the only biocatalyst mediating both demethylation and methylation in a reversible fashion. For this purpose 3,4-dihydroxybenzoic acid 2b was chosen as non-natural model substrate for accepting the methyl group (acceptor). Detection of the demethylated product catechol 2a as well as the mono-methylated carboxylic acid 1b (regioisomeric mixture meta/para = 2/1, see Supplementary Fig. 5) proofed the feasibility of employing MTase I as the only biocatalyst for both methylation and demethylation. Initial conversions of 2–3% were improved by optimizing various parameters such as temperature, enzyme and substrate concentration (see Supplementary Fig. 6–9). An excess of the acceptor molecule instead of equimolar amounts pushed the demethylation of the model substrate 1a to conversions of up to 80%. Employing 3,4-dihydroxybenzaldehyde 2c as alternative model acceptor gave identical results, yielding vanillin and isovanillin 1c in a ratio of 3/1. The significant increase of conversion obtained by using an excess of one substrate indicated an equilibrium between demethylation and methylation caused by the reversible nature of the reaction (Fig. 5a). It can be assumed that the via demethylation of guaiacol 1a obtained product (catechol 2a) functions subsequently as methyl acceptor for a methylation in the reverse direction consuming the methyl group from the methylated product (vanillin/isovanillin 1c). Following mentioned transformations in both directions over time using equimolar amounts of substrates (Fig. 5b) proved this hypothesis. After having a fast methyl transfer at the beginning of the reaction, it rapidly reached an equilibrium after 4–6 h, ending up with a comparable ratio of substrates and products in any case. In contrast, employing a 5-fold molar excess of methyl acceptor 2c significantly pushed the demethylation of 1a by diminishing the reverse reaction at the same time (Fig. 5c). Hence, in order to exploit cobalamin-dependent methyl transfer either for demethylation or methylation of a substrate in an efficient way, pushing the reaction equilibrium toward either side is crucial.
Substrate examples for demethylation and methylation
To demonstrate the feasibility of the concept we focused at first on demethylation. Compounds turning out to be easily demethylated might then serve as suitable methyl donors for the methylation reaction in a second step. For initial experiments with MTase I from D. hafniense catechol 2a and derivatives 2b-2c were employed in excess as methyl acceptors. In addition to the model substrate 1a various other aromatic compounds (Fig. 6) fitting the catechol pattern of published natural substrates52 as well as a series of aromatic and aliphatic compounds with alternative substrate pattern such as veratrol derivatives as well as 2-(methylamino)- and 2-(methylthio)phenol were tested for demethylation (Supplementary Table 3). Aromatic substrates with a hydroxy moiety in ortho-position to the methoxy were efficiently demethylated. Furthermore, a number of anisol derivatives (Fig. 6, compounds 1a-h, 2h) bearing various substituents such as an amine-, hydroxymethyl-, aldehyde-, or acid-group in ortho-position were converted as well. Remarkably, o-anidisine 1d was the only substrate which was demethylated with higher conversions when the acceptor was present in equimolar amounts instead of excess (Supplementary Table 4). This indicated a reaction equilibrium lying on the demethylation reaction, most likely because the formed product 2-aminophenol 2d turned out to be not a suitable substrate for methylation. Generally, the choice of the proper acceptor is an important parameter to reach high conversion; interestingly, the most suitable acceptor seemed to vary depending on the substrate. Vanillin and vanillic acid as well their regioisomers (compounds 1b-c) were demethylated with moderate conversions ranging from 25–45%. Substrates such as guaiacol 1a, 3-methoxycatechol 3-2h and 2-methoxyresorcinol 2–2h were demethylated with up to 80% conversion depending on the used co-substrate. Even higher conversions than for 1a were achieved for 2,3-dimethoxyphenol 2,3-1h (up to 93%) and 2,6-1h (up to 97%). Featuring two methoxy moieties, 2,3-1h and 2,6-1h represent particularly useful methyl sources for a possible methylation reaction.
Transfer of one methyl moiety leads to the formation of 3-methoxycatechol 3-2h, which is further converted to pyrogallol 3h upon twofold demethylation. Therefore, they were employed together with 1a as methyl-donating co-substrates for exploration of corrinoid-dependent methylation in the next step (Fig. 7).
Besides catechol 2a, 3,4-dihydroxybenzoic acid 2b and related aldehyde 2c, also 3,4-dihydroxybenzyl alcohol 2i and caffeic acid 2j were successfully methylated. Although a number of aliphatic, cyclic and aromatic compounds including also S-,N-, and C-nucleophiles (see supplementary Table 5) turned out to be non-substrates, numerous enzyme engineering techniques59 that have emerged in recent years might tune the MTase I from D. hafniense to convert an extended range of different substrates with high activity and selectivity. When using the mono-methyl donor 1a as co-substrate, it had to be employed in excess to achieve an efficient reaction. For instance, the amount of methylated product doubled in case of 2b, 2i, and 2j by using a 5-fold excess of donor, reaching up to 84% conversion. Interestingly, when bis-methyl donors capable of donating two methyl groups such as dimethoxyphenols 2,3-1h and 2,6-1h were used as methylating reagent, conversions up to 75% were already achieved with equimolar amounts of donor, indicating their potential as efficient donors. It seemed that the availability of two transferable methyl groups in a single molecule had a similar effect on the reaction as using excess of donor. It is worth mentioning, that exclusive mono-methylation was observed for all substrates investigated. For those substrates where the formation of regioisomers was possible, product mixtures of meta- and para-methylated isomers were identified. Generally, meta-preference was observed for substrates 2b, 2c, and 2j with a regioisomeric excess of approximately = 65/35 or higher. Therefore, the corrinoid-dependent MTase I from D. hafniense displayed a similar regioselectivity than observed for various SAM-dependent catechol O-MTases (COMT)1,60,61. One exception was 3,4-dihydroxybenzyl alcohol 2i for which almost an 1/1 mixture of isomers with marginal p-preference was obtained. Since related compound 2c on the contrary was methylated with reasonable regioselectivity (m/p up to 76/24) it can be assumed that the side chain at the aromatic catechol chore has an effect on substrate binding within the enzyme and thus on discrimination of regioisomeric groups1,44. However, the regioselectivity of MTase I could be increased in favor of the m-methylated product by medium engineering. For instance, in the presence of an organic co-solvent such as ethanol (EtOH, 10% v/v) the ratio of ferulic/isoferulic acid obtained from substrate 2j was significantly increased from 71/29 to 85/15. An even more pronounced effect of EtOH addition on the regioselectivity was observed in case of substrate 2i. Here, the minimal preference for the p-methylated product (m/p = 48/52) turned into a surplus formation of the meta-isomer (m/p = 81/19) without any loss of activity. Finally, the corrinoid-dependent methylation was performed in semi-preparative reactions (24 mL). Both 3,4-dihydroxybenzyl alcohol 2i and caffeic acid 2j were methylated employing equimolar amounts of 2,6-2h giving the methylated products with 72% conversion (38% isol. yield) and 76% conv. (45% isol. yield), respectively.
This study demonstrates the feasibility of a methyl transfer concept based on corrinoid-mediated shuttle catalysis. Being sustainable, safe, scalable and not restricted by cofactor recycling issues, this protocol may lay the base for an alternative method for methylation as well as demethylation. With the substrate scope identified, the applicability and possible future impact has already been demonstrated. While the concept might enable (i) tailoring of natural products, (ii) valorization of waste materials or biomass (e.g. lignin), and (iii) deprotection by demethylation on the one hand, methylation on the other hand will allow the preparation of value-added products. The substrate scope can be expanded by common rational engineering techniques and furthermore, other MTase systems with different substrate specificities can be chosen from nature’s vast supply. Therefore, corrinoid-dependent MTases may represent a valuable and promising add-on to the enzymatic toolbox applicable for organic synthesis.
For DNA- and amino acid sequences of the enzymes used see Supplementary Note 1. Biocatalysts were recombinantly expressed in E. coli and used as lyophilized cell-free extract (CFE) or purified preparation. Since E. coli is not able to synthesize cobalamins56, the corrinoid protein (CP) was reconstituted with Me-Cbl prior to use in order to yield active holo-CP. General material and detailed experimental procedures related to protein expression and cofactor reconstitution are given in Supplementary Methods. Biocatalytic reactions were performed at least in triplicates in degassed buffers under inert atmosphere, using a Labstar MB10-compact Glove-box (MBraun, Germany) equipped with a special O2-sensor (MB-OX-EC) to be used with aqueous systems.
Determination of enzyme activity
Representative biotransformation procedure
Analytical biotransformation reactions were carried out on 120 µL scale as follows: MTase I (10–20 mU, 40 mg mL−1 CFE or 13 mg mL−1 pure enzyme, respectively) was dissolved in holo-CP solution (400 µL mL−1, containing 67 mg mL−1 CFE or 22 mg mL−1 pure enzyme, respectively). Then appropriate amounts of substrate (10–50 mM final concentration) and co-substrate (10–50 mM final concentration) dissolved in MOPS/KOH buffer (50 mM, pH 6.5, 150 mM KCl) were added and the reaction samples were shaken at 30 °C and 800 rpm. See Supplementary Note 2 for details with respect to the use of an O2-scavenger system during initial experiments. After 24 h an aliquot of the reaction was withdrawn and diluted (1:10) with MeCN (6 parts) and deionized water (3 parts). Precipitated protein was removed by centrifugation (14,000 rpm, 15 min) and the clear supernatant was filtered and directly analyzed by reversed-phase HLPC.
Preparative methylation reactions
Semi-preparative reactions were carried out on a 24 mL scale in accordance to the procedure described above using equimolar amounts of substrate and methyl donor. The reaction mixture was shaken at 30 °C and 120 rpm for 24 h. Work-up, isolation and purification were performed as follows.
Methylation of 3,4-dihydroxybenzyl alcohol
Substrate 2i (1.2 mmol, 168 mg) was methylated according to the described procedure using 2,6-1h (1.2 mmol, 185 mg) as methyl donor. The reaction mixture was quenched by adding MeCN (24 mL) and the precipitated catalyst was removed by centrifugation (4000 rpm, 30 min). Then the supernatant was extracted with EtOAc (2× 25 mL), combined organic phases were dried over Na2SO4 and the solvent evaporated under reduced pressure. Purification by column chromatography on silica (eluent: EtOAc/Hexanes = 1/1) afforded the product mixture (vanillyl/isovanillyl alcohol = 60/40) as colorless solid in 38% yield (70 mg, 0.45 mmol). Rf (EtOAc/Hexanes = 1/1) = 0.26.
Vanillyl alcohol m-2i. 1H-NMR (300 MHz, d4-MeOH): δ 6.94 (s, 1H), 6.76 (s, 2H), 4.50 (s, 2H), 3.85 (s, 3H); 13C-NMR (75 MHz, d4-MeOH): δ 148.9. 146.9, 134.2, 121.1, 116.0, 112.1, 65.3, 56.3. NMR spectra are in agreement with literature62.
Isovanillyl alcohol p-2i. 1H-NMR (300 MHz, d4-MeOH): δ 6.88 (s, 1H), 6.86 (s, 1H), 6.83 (s, 1H), 4.46 (s, 2H), 3.83 (s, 3H); 13C-NMR (75 MHz, d4-MeOH): δ 148.4, 147.5, 135.6, 119.6, 115.4, 112.5, 65.1, 56.4. NMR spectra are in agreement with literature63.
Methylation of caffeic acid
Substrate 2j (0.48 mmol, 87 mg) was methylated according to the described procedure using 2,6-1h (0.48 mmol, 74 mg) as methyl donor. The reaction mixture was basified with aqueous NaOH (10% w/v) to pH 14 and extracted with EtOAc (2 × 25 mL). After acidifying the aqueous phase with HCl (6N) to pH 1 the mixture was extracted with EtOAc (2 × 25 mL), combined organic phases were dried over Na2SO4 and the solvent evaporated. Column chromatography on silica (eluent: EtOAc/Hexanes = 1/3, AcOH 1% v/v) afforded the product mixture (ferulic/isoferulic acid = 60/40) as brownish solid in 45% yield (42 mg, 0.23 mmol). Rf (EtOAc/Hexanes = 1/1, AcOH 1% v/v) = 0.71.
Ferulic acid m-2j. 1H-NMR (300 MHz, d4-MeOH): δH [ppm] = 7.60 (d, J = 15.9 Hz, 1H), 7.18 (d, J = 1.9 Hz, 1H), 7.06 (dd, J = 8.2, 1.9 Hz, 1H), 6.81 (d, J = 8.2 Hz, 1H), 6.31 (d, J = 15.9 Hz, 1H), 3.89 (s, 3H); 13C-NMR (75 MHz, d4-MeOH): δC [ppm] = 171.0, 150.5, 149.4, 146.9, 127.8, 124.0, 116.4, 115.9, 111.6, 56.4. NMR spectra are in agreement with literature64.
Isoferulic acid p-2j. 1H-NMR (300 MHz, d4-MeOH): δH [ppm] = 7.54 (d, J = 15.9 Hz, 1H), 7.05 (d, J = 2.1 Hz, 1H), 6.94 (dd, J = 8.2, 2.1 Hz, 1H), 6.81 (d, J = 8.2 Hz, 1H), 6.27 (d, J = 15.9 Hz, 1H), 3.88 (s, 3H); 13C-NMR (75 MHz, d4-MeOH): δC [ppm] = 170.8, 151.4, 148.0, 146.6, 128.9, 122.7, 116.4, 114.7, 112.5, 56.3. NMR spectra are in agreement with literature65.
Characterization of products
The data generated and/or analyzed during this study are available within the article and the supplementary information file or obtainable from the corresponding author upon reasonable request.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This study was financed by the Austrian FFG, BMWFJ, BMVIT, SFG, Standortagentur Tirol and ZIT through the Austrian FFG-COMET- Funding Program (Austrian Centre of Industrial Biotechnology). Financial support by the Austrian Science Fund (FWF) within the DK “Molecular Enzymology” program (project no. W901) is gratefully acknowledged. Financial support by the Austrian Science fund is acknowledged (P30920-B21). The drawing of the shuttle bus of the graphical abstract was taken from vecteezy (www.vecteezy.com) under the common licence BY-SA and modified.