Multienzyme One‐Pot Cascades Incorporating Methyltransferases for the Strategic Diversification of Tetrahydroisoquinoline Alkaloids

Abstract The tetrahydroisoquinoline (THIQ) ring system is present in a large variety of structurally diverse natural products exhibiting a wide range of biological activities. Routes to mimic the biosynthetic pathways to such alkaloids, by building cascade reactions in vitro, represents a successful strategy and can offer better stereoselectivities than traditional synthetic methods. S‐Adenosylmethionine (SAM)‐dependent methyltransferases are crucial in the biosynthesis and diversification of THIQs; however, their application is often limited in vitro by the high cost of SAM and low substrate scope. In this study, we describe the use of methyltransferases in vitro in multi‐enzyme cascades, including for the generation of SAM in situ. Up to seven enzymes were used for the regioselective diversification of natural and non‐natural THIQs on an enzymatic preparative scale. Regioselectivites of the methyltransferases were dependent on the group at C‐1 and presence of fluorine in the THIQs. An interesting dual activity was also discovered for the catechol methyltransferases used, which were found to be able to regioselectively methylate two different catechols in a single molecule.


Introduction
THIQs are an important group of bioactive alkaloids. Alongside their applications as analgesics and antitussives, [1][2][3][4] THIQs show promise as therapeutics towards cancer, neuropathologies and multi-drug resistant bacteria. [3,[5][6][7][8] However, supply of these compounds for studies and clinical use is limited. Natural THIQs can be harvested from mixtures from certain plants, but their chemical syntheses are complicated by the presence of chiral centres and the density of functional groups. [9] An enzyme of particular interest has been norcoclaurine synthase (NCS), which catalyses a Pictet-Spengler (PS) reaction between dopamine and 4-hydroxyphenylacetaldehdye, to initiate benzylisoquinoline alkaloid (BIA) biosynthesis. Most usefully, using natural and engineered NCS variants, [10][11][12] the b-arylethylamine and carbonyl substrates can be varied to generate a number of natural and nonnatural THIQs. [13][14][15][16] The diversity of THIQs generated has been further extended by forming in vitro cascades with enzymes upstream of NCS. [17] In plant BIA biosynthesis, methyltransferase (MT) enzymes operate downstream of NCS to methylate norcoclaurine. [18] Such natural product methylations involve transfer of a methyl group from Sadenosylmethionine (SAM) to specific nucleophilic sites on the scaffold, and is widely associated with activation of the compound or improvement of its biological properties. [19][20][21][22][23][24] MTs are therefore an essential addition for the diversification of THIQs. However, while the O-MTs involved in producing natural BIAs are regioselective, [25,26] they lack the substrate scope in vivo needed to be components of a flexible enzyme cascade to larger sets of THIQs. While in vitro biocatalytic cascades take inspiration directly from natural biosynthetic pathways, they allow more flexibility over reaction design, offering control in the order and quantity of enzymes added to the cascade as well as the enzyme used. Importantly, the in vitro approach reduces background reactions such as the phosphate catalysed PS, [27] and can simplify product isolation and purification. Here we present the use of three O-MTs in vitro, catechol-O-MTs from Rattus norvegicus (RnCOMT), [28] Coptis japonica (Cj-6-OMT), [25,29] and Myxococcus xanthus (MxSafC), [30] and an N-MT from Coptis japonica (CjCNMT), [31] to establish potential reactivities towards a range of THIQs. We report the integration of these enzymes into up to 7-step cascades for the strategic diversification of bioactive compounds ( Figure 1). The cost of SAM has also previously limited the scale of MT-catalysed reactions, but this challenge is resolved with the incorporation of a reported modular in vitro cofactor supply system, [32] demonstrating the capacity of such an approach in ambitious biocatalytic applications.

Results and Discussion
Initial O-MT Screening In the initial screening, a variety of THIQs were prepared using Thalictrum flavum NCS (TfNCS), in high enantiomeric excesses (ees) where the single isomer is indicated, and purified. [11,16,17] The substrate panel (Figure 2 a) included BIAs (1-5) and C-1 substituted THIQs with linear aliphatic or cyclic moieties (6)(7)(8)(9)(10)(11). BIAs lacking the catechol group on the isoquinoline scaffolds or those bearing a halide group (12)(13)(14) were also included. All substrates 1-14 were tested with MT enzymes RnCOMT and MxSafC. Cj-6-OMT was also used for comparison as it is known to selectively methylate in vivo the 6-OH of norcoclaurine (S)-2. [29] Initially, purified enzymes were used for reactions with substrates 1-3. With the aim of scaling up the process and making it more viable in an industrial setting, the same reactions were also carried out with clarified lysates, showing regioselectivities that were comparable to the purified O-MTs (SI section 3.2). Clarified lysates were then used with all the substrates. We were delighted to see that all THIQs bearing a catechol were readily accepted and new peaks were detected by HPLC    Figure S5). In general, RnCOMT and MxSafC showed broader substrate scope relative to Cj-6-OMT, with excellent activities towards THIQs with an aliphatic side chain (6-10), which were only slightly or not converted by Cj-6-OMTunder these conditions. Notably, complete conversions were obtained within the first 90 min with substrates 1-11 with RnCOMT, while MxSafC showed some starting material left only in the case of substrates 9-11. As expected, none of the THIQs lacking a catechol moiety (12 and 14) were accepted by the MTs, including derivative 13 where a catecholic hydroxyl group was replaced with a bromine.

Building the Enzyme Cascades and O-MT Regioselectivities
Establishment of the regiochemistries was also key, which required scale-up of the reactions for product characterisation. Several factors were considered. Firstly, the high cost of the cofactor SAM restricts its use in larger scale biocatalytic reactions in vitro: although it has been reported in the N-methylation of THIQs, [31] this is not practicable for wider applications. In previous work, integration of MTs into a linear cascade with two further enzymes overcame this issue using a methionine adenosyltransferase (MAT E.C. 2.5.1.6) and a methylthioadenosine nucleosidase (MTAN, E.C. 3.2.2.9). [33] MATs generate SAM from ATP and L-methionine, both of which are less expensive and importantly more stable than SAM. When SAM is used for the methylation reaction, S-adenosylhomocysteine (SAH) is formed as a by-product. SAH, which would otherwise inhibit the methylation, can then be cleaved by the MTAN into Sribosyl-L-homocysteine (SRH) and adenine (Figure 1 a). For the SAM supply/SAH degradation system, enzymes MAT and MTAN, both from E. coli (EcMAT and EcMTAN, respectively) were selected. [33] Secondly, clarified lysates were used instead of pure enzymes to avoid the requirement for enzyme purification steps. Thirdly, to further streamline the process, the methylation cascade reaction was coupled directly to the NCS-catalysed PS reaction in a one-pot, two-step cascade, obviating the need for purification of the THIQ intermediates. The methylation of representative substrates (S)-1, (S)-6, 9, 10 and (S)-11 was explored first (Scheme 1. ees for (S)-1 (> 97 %) and (S)-6 (> 98 %) are shown in the SI). Here it was essential that all the dopamine was consumed in the TfNCS catalysed reaction before the MT step to avoid the methylation of unreacted dopamine, [33] which could interfere with the regiochemical outcome.
With the architecture of the cascades defined, a biocatalytic scale-up of the methylated THIQs was attempted in a single pot. Starting from dopamine (5-20 mM) and an excess of the corresponding carbonyl compound, the THIQs (S)-1, (S)-6, 9,10 and (S)-11 were generated with high enantiopurity (for single isomer compounds) and the methylation step initiated by the addition of RnCOMT clarified lysate (10 % v/v) in the presence of ATP, L-methionine and the SAM supply/SAH degradation system enzymes EcMAT (10 % v/v) and EcM-TAN (2.5 % v/v). Using RnCOMT the methylated products (S)-6-OMe-1, (S)-6-OMe-6, 7-OMe-9, 7-OMe-10, and (S)-6-OMe-11 were isolated with good yields and high regioselectivities with only 5-10 % of the opposite regioisomer detected by 1 H NMR spectroscopy in most cases ( Table 1). As expected, the regiochemistry of the reaction was strongly dependent on the nature of the side chain, which directed the Scheme 1. Four-enzyme cascades for the synthesis of: (a) THIQs (S)-6-OMe-1, (S)-6-OMe-6, and (S)-6-OMe-11, and (b) THIQs 6-OMe-9, 7-OMe-9, 6-OMe-10, and 7-OMe-10 from 9 and 10 respectively.  methylation preferentially at the 6-OH for substrates (S)-1, (S)-6, and (S)-11, and surprisingly preferably on the 7-OH for substrates 9 and 10. Interestingly, when 9 or 10 underwent the same reaction conditions with MxSafC, the 6-OH was methylated instead preferentially and derivatives 6-OMe-9 or 6-OMe-10 respectively were obtained in good yield and excellent regioselectivities (Table 1). This unique performance with a switch in the regioselectivities for both RnCOMT and MxSafC further confirms the complementarity of these two enzymes. [33] With the challenge to demonstrate the potential of this approach, the methylation with RnCOMT was further coupled with a multistep in vitro enzyme cascade for the in situ generation of norcoclaurine (S)-2 (generated in > 97 % ee) using Candidatus Nitrosopumilus salaria BD31 tyrosinase (CnTYR) and Enterococcus faecalis DC32 tyrosine decarboxylase (EfTyrDC), together with a versatile transaminase from Chromobacterium violaceum [34] (CvTAm) and wild-type TfNCS enzyme. [17] A total of seven enzymes as clarified celllysates were used in the same pot, achieving the conversion of L-tyrosine into (S)-coclaurine (S)-6-OMe-2 in an unique onepot three step cascade (Scheme 2 a).
The methylation with RnCOMT was almost completely regioselective with no appreciable regioisomeric peak detected by HPLC. When the cascade was scaled up to 50 mL (L-tyrosine 20 mM), (S)-6-OMe-2 was isolated in 55 % yield and small amounts (5 % of total product) of the 7-Omethylated regioisomer were detected by 1 H NMR spectroscopy. The lower isolated yields compared to the complete conversion of the starting material and the almost quantitative yield measured by HPLC, reflects again the challenge of purifying such alkaloids. [35] Catechol O-MTs such as RnCOMT have not been shown to methylate adjacent hydroxy groups or phenols, even in the presence of an excess of methyl equivalents. Interestingly, the structurally related MxSafC, proved to be an exception. This consideration derives from the fact that methylation of (S)-2 with MxSafC and an excess of SAM, led to a regioisomeric mixture of (S)-6-OMe-2 and (S)-7-OMe-2 (ratio 60:40), which were converted over time into a third product. This unexpected product was isolated and characterized confirming the structure of the 6,7dimethoxy tetrahydroisoquinoline (S)-6,7-(OMe) 2 -2 (see the SI).
A one-pot cascade was also established for the synthesis and methylation of (S)-5 (formed in 90 % ee) (Scheme 2 b). CnTYR and EfTyrDC were used for the hydroxylation and decarboxylation of 3-F-L-tyrosine, [17] then reaction with phenylacetaldehyde and TfNCS followed by methylation with RnCOMT on the crude reaction mixture, gave the 6-Omethylated derivative (S)-6-OMe-5 which was isolated in 19 % yield (27 % HPLC yield) over the 7-enzyme steps. Similarly, using MxSafC gave (S)-6-OMe-5 in 20 % yield (31 % HPLC yield). The limiting step is most likely the initial hydroxylation with CnTYR. The preferential methylation at the 6-OH position by both MTs is likely a consequence of substrate-active site interactions. However, the change in MxSafC regioselectivity between (S)-1 and fluorinated (S)-5 indicates electronic factors are also important. The fluorine adjacent to a catechol moiety decreases the pK a of the neighbouring -OH and has previously been shown to direct the methylation on the hydroxyl group to a position equivalent to that at 7-OH. [36] Here, the fluorine at C-8 in the THIQ scaffold directs the methylation unexpectedly towards (S)-6-OMe-5 with MxSafC. Notably, MxSafC led to a mixture of the two regioisomeric products when (S)-1, lacking the fluorine, was used as starting material. We hypothesized that a subtle substrate conformational change together with an intramolecular hydrogen bonding between the fluorine and the adjacent -OH, can make methylation at that position more difficult.

Extending the Enzyme Cascades for Further O-and Nmethylations
A case of particular interest is norlaudanosoline (S)-3. This is a substrate for Cj-6-OMT and it is converted to the corresponding 6-O-methylated derivative (S)-6-OMe-3. [29] When the assay was performed with either RnCOMT or MxSafC and an excess of SAM, new product peaks not attributable to the 6-or 7-O-methylated products were identified by HPLC. Interestingly, when only one equivalent of SAM was used with RnCOMT, the peak corresponding to 6-O-methylated 3 was identified as the main product together with some starting material. If MxSafC was used instead, a mixture of two regioisomeric methylated products was identified (SI Figure S5). With the aim of elucidating the structure of the new products, a two-step, five-enzyme cascade was established for the synthesis and methylation of norlaudanosoline (S)-3 (formed in > 97 % ee) in one pot (Scheme 3) starting from dopamine (20 mM). The wild-type TfNCS enzyme and CvTAm [34] were used in the first step, [37] Scheme 2. (a) Seven-enzyme cascades for the synthesis of THIQs (S)-6-OMe-2, (S)-7-OMe-2 and (S)-6,7-(OMe) 2 -2 and (b) six-enzyme cascade for the synthesis of THIQs (S)-6-OMe-5 on a laboratory preparative scale. and the methylation step was then coupled and scaled up using RnCOMT together with the SAM supply/SAH degradation system. Product isolation confirmed that a second methylation occurred on the catechol moiety of the C-1 side chain and RnCOMT proved to be exceptionally selective, with methylation preferentially occurring at the 6-OH and the 3'-OH. Complete conversion into nororientaline (S)-6,3'-(OMe) 2 -3 was observed, which was isolated in 33 % yield. This double methylation is completely unprecedented as it is known to be carried out by two different MTs in the BIA biosynthetic pathway, with RnCOMT combining the activity of a 6-OMT and a 3'-OMT. [38] MxSafC showed a similar performance, leading to a mixture of 6-OMe and 7-OMe-(S)norlaudanosoline which were then converted into the corresponding dimethylated analogues (S)-norreticuline (S)-6,4'-(OMe) 2 -3 and (S)-norprotosinomenine [39] (S)-7,4'-(OMe) 2 -3 in the presence of an excess of methyl equivalents. [ * ] Although less selective, MxSafC could also combine the activity of two methyltransferases, including a 4'-OMT. This reactivity was used for the selective synthesis of (S)-norreticuline (S)-6,4'-(OMe) 2 -3 by coupling the Cj-6-OMT and the MxSafC in a one-pot three-step cascade (Scheme 3), highlighting the ability to mix and match MTs to directed products.
Generally, RnCOMT is highly selective and the 6-OH is preferentially methylated except in the case of spiro-THIQs 9 and 10, where a complete inversion of the selectivity was observed. To understand this pattern of regioselectivity, a selection of THIQs were docked into models of the RnCOMT and MxSafC active sites using AutoDock Vina. [42] The catechol-binding pocket forms a deep groove with the magnesium ion proximal to highly conserved lysine residues (K144 and K145 in RnCOMT and MxSafC respectively) involved in the deprotonation of the hydroxyl to be methylated. The catechol moiety coordinates with the magnesium ion, while the side chain establishes interactions in the outer region pointing either towards the so called "hydrophobic wall" or towards the solvent, [35,43] defining the catechol position which is preferentially methylated. Nevertheless, none of our models could completely support the regioselectivity observed, presumably due to the high degree of reorientation of the active site around the specific substrate. [30] However, some docking analysis with (S)-6-OMe-3 and RnCOMT (which exhibited high regioselectivity for the second methylation at 3'-OH) and MxSafC (which favored the second methylation at 4'-OH) provided useful insights consistent with the experimental findings (SI Figure S6).

Conclusion
In summary, we have developed in vitro enzymatic cascades for the synthesis of a wide range of methylated natural and non-natural THIQ alkaloids, starting from commercially available and inexpensive building blocks. The application of an efficient SAM supply/SAH degradation system, together with the use of clarified lysates coupled with the NCS step, contribute to make the scale-up of the whole process viable. Finally, methylated THIQ alkaloids in high enantiopurities such as (S)-norreticuline, (S)-coclaurine, (S)nororientaline, and (S)-orientaline together with novel THIQs are readily available in high purity from a one-pot cascade. The methylation step showed exceptional regioselectivity bearing in mind there is a degree of non-enzymatic methylation (generally around 4 %) which has also been reported in the literature. [30,44] An unprecedented dual activity has also been discovered for the methyltransferases RnCOMT and MxSafC, which were found to be able to regioselectively methylate two different catechols in a single molecule. This unique activity permitted the regioselective synthesis of (S)-nororientaline in one single step from the precursor (S)-norlaudanosoline.