Engineering Orthogonal Methyltransferases to Create Alternative Bioalkylation Pathways

Abstract S‐adenosyl‐l‐methionine (SAM)‐dependent methyltransferases (MTs) catalyse the methylation of a vast array of small metabolites and biomacromolecules. Recently, rare carboxymethylation pathways have been discovered, including carboxymethyltransferase enzymes that utilise a carboxy‐SAM (cxSAM) cofactor generated from SAM by a cxSAM synthase (CmoA). We show how MT enzymes can utilise cxSAM to catalyse carboxymethylation of tetrahydroisoquinoline (THIQ) and catechol substrates. Site‐directed mutagenesis was used to create orthogonal MTs possessing improved catalytic activity and selectivity for cxSAM, with subsequent coupling to CmoA resulting in more efficient and selective carboxymethylation. An enzymatic approach was also developed to generate a previously undescribed co‐factor, carboxy‐S‐adenosyl‐l‐ethionine (cxSAE), thereby enabling the stereoselective transfer of a chiral 1‐carboxyethyl group to the substrate.


Introduction
Methylation is one of the simplest and most widely used reactions in nature.T he majority of methylation reactions in cells are catalysed by S-adenosyl-l-methionine (SAM)-dependent methyltransferase (MT) enzymes that transfer am ethyl group from SAM to nucleophilic centres in aw ide array of substrates,i ncluding small metabolites,p roteins, nucleic acids,a nd other biological molecules. [1][2][3][4][5] Ty pically, MTs transfer methyl groups with exquisite chemo-and regioselectivity,which often has aprofound effect on the properties of biological molecules,t hus making them ah ighly diverse and important class of enzymes.F or example,methylation of proteins and DNAi sakey mechanism in the epigenetic control of gene expression. [1,2] MT-catalysed methylation is also ac ommon theme in the biosynthesis of therapeutically important secondary metabolites,including antibiotics,where methyl substituents can affect the properties and bioactivity of the natural products (NPs). [3,5] Several SAM-dependent MTs have been shown to accept synthetic SAM analogues,with alternative S-alkyl, S-allyl, or S-propargyl substituents,f acilitating non-native alkylation reactions. [5][6][7][8][9][10][11][12][13][14][15][16] Thea bility to transfer functional groups regioselectively using MTs and SAM derivatives has been exploited for the site-selective labelling of proteins and DNA, [5][6][7][8][9][10] as well as in the structural diversification of clinically important NPs,i ncluding the anticancer agent rebeccamycin and rapamycin immunosuppressive agents (Figure 1a). [11,13] In most cases,t he SAM analogues are prepared synthetically through the alkylation of S-adenosyl-l-homocysteine (SAH); this process however gives rise to am ixture of sulfonium epimers,w ith the natural S epimer functioning as ac ofactor and the R epimer potentially inhibiting activity.S ome SAM analoguesc an be prepared from syntheticm ethionined erivatives andATP usingamethionine adenosyltransferase(MAT) enzyme to producethe biologically relevant sulfoniumepimer as the sole product. [11,13] MAThas also been used to produce related seleno-SAM analogues. [14,15] Notwithstanding this, SAM analogues are expensive to produce,u nstable,a nd cannot penetrate the cell membrane,t hus largely limiting their utility to small scale in vitro applications. [16][17][18][19][20] New enzymatic approaches to produce SAM derivatives from metabolically available,r ather than synthetic,p recursors would be highly desirable.S uch approaches,c ombined with engineered orthogonal MTs that are selective for enzymatically derived SAM analogues,c ould open up new bioalkylation pathways in vitro and in vivo.T ot his end, we were particularly interested in the discovery of an aturally occurring SAM derivative,carboxy-S-adenosyl-l-methionine (cxSAM), which is generated from SAM and prephenate by the cxSAM synthase (CmoA) enzyme ( Figure 1b). [21,22] CmoA is found in bacterial species and functions in tandem with ac arboxymethyl transferase (CmoB), which transfers the carboxymethyl group from cxSAM to at RNAs ubstrate. [23] Whilst another pathway was recently discovered that utilises cxSAM in the carboxymethylation of ap eptide substrate, [24] such bioalkylation pathways are extremely rare in nature.H owever,i fC moA could be coupled with orthogonal variants of the more common MTs,e ngineered to have higher selectivity for cxSAM, then al arge array of diverse carboxymethylated products could be generated with new functionality and bioactivity.I nt his paper,w ed emonstrate how CmoA can be utilised in tandem with MT enzymes to generate carboxymethylated, rather than methylated, products.Furthermore,wedemonstrate how active-site mutagenesis can lead to orthogonal MTs with improved selectivity for cxSAM over SAM, thereby opening the way to engineering new carboxymethylation pathways in vitro and in vivo. Finally,w es how how CmoA can also be used to generate an ovel carboxy-S-adenosyl-l-ethionine (cxSAE) cofactor, thereby facilitating the stereoselective MT-catalysed transfer of ac hiral 1-carboxyethyl group.

Results and Discussion
Activity of Methyltransferases (COMT &CNMT) with Synthetic cxSAM.
To assess the possibility of using cxSAM as an alternative cofactor to SAM, we chose to explore reactions catalysed by catechol-O-methyltransferase (COMT) from Rattus norvegicus and coclaurine-N-methyltransferase (CNMT) from Coptis japonica. COMT was previously shown to accept ar ange of catechol substrates as well as synthetic SAM analogues. [25] Accordingly,W TC OMT was incubated with 3,4-dihydroxybenzaldehyde 2 and synthetic cxSAM, prepared through alkylation of SAH with bromoacetic acid, [21] resulting in amixture of meta-and para-carboxymethylated regioisomers 2a (40 %) and 2b (29 %), respectively (Figure 2a and Figure S1 in the Supporting Information). These results are consistent with the relaxed regioselectivity observed for methylation reactions catalysed by COMT. [25] Previously we showed that aC OMT Y200L mutant has significantly improved regioselectivity,a ffording predominately metamethylation of various substituted catechol derivatives. [25] In light of this,c atechol 2 and cxSAM were similarly incubated with COMT Y200L, affording significantly improved regioselectivity for meta-o ver para-carboxymethylation (2a 64 % vs. 2b 3%). Similar results were obtained for 4-nitrocatechol 3 and cxSAM, with COMT Y200L providing 76 % meta-and 5% para-carboxymethyl products 3a and 3b,r espectively Figure 1. a) Alkyl diversificationo fnatural products:rapamycin and rebeccamycinderivatives generated using MTsand SAM analogues. [11,12] b) CmoA-catalysed formation of cxSAM. In the native pathway,c xSAM functions as aco-factor for the carboxymethyltransferase CmoB in the modification of atRNA substrate. In this work (grey), we show how CmoA can be combined with an engineered methyltranferase (MT*), thereby opening up new pathwayst ocarboxymethylated products. CNMT was pre-incubated with THIQ 4 or 5 for 1hto remove copurified SAM and then transferred to fresh buffer.Assays were then conducted with synthetic cxSAM (1.6 mm), substrate 4-7 (0.25 mm) and CNMT (100 mm)and incubated for 17 h. Assays were done in triplicate and standard errors calculated.
( Figure 2a). Forb oth catechol substrates 2 and 3,s mall amounts of the corresponding methylated products were observed (e.g., 2 with WT and Y200L COMT gave 21 AE 1% and 16 AE 1% methylated products,r espectively), which is attributed to co-purification of SAM with COMT and background decarboxylation of cxSAM to SAM, which is apparent during the incubation period.
In our previous studies,w ea lso showed that CNMT accepts avariety of tetrahydroisoquinoline (THIQ) substrates as well as SAM analogues. [26] To further explore its cofactor selectivity,W TC NMT was incubated with synthetic cxSAM and various THIQ substrates (4-7;F igure 2b and Figure S2). THIQs 4 and 5 showed significant carboxymethylation (82 and 83 %r espectively), whilst THIQs 6 and 7 gave only low conversions with cxSAM (8-12 %). Using the crystal structure of CNMT (PDB ID:6 GKV), [26] we infer that the close proximity of the carboxymethyl group of cxSAM and the C1substituent of the THIQs (6 and 7)i sl ikely to cause steric hindrance leading to decreased activity.

Site-Directed Mutagenesis of CNMT.
Having established that CNMT can utilise cxSAM as ac ofactor, we sought to develop CNMT mutants with increased activity and selectivity for cxSAM, since SAM will ultimately be present in significant levels within tandem CmoA-CNMT cascade reactions.C moB,w hich naturally functions in tandem with CmoA, exhibits approximately 500-fold higher affinity for cxSAM over SAM, which presumably ensures that no significant competing methylation of its tRNAsubstrate occurs in vivo. [21,23] In light of this, we used the reported X-ray crystal structure of CmoB (PDB ID:4 QNU) to guide mutagenesis of CNMT. [26] Due to the low sequence similarity between CNMT and CmoB (17.5 %i dentity matrix, CLUSTAL 2.1), only those residues in close proximity to the cofactor were considered. Within the CmoB active site,L ys91, Ty r200, and Arg315 residues were identified, which form interactions with the S-carboxymethyl group of cxSAM, likely contributing to the high selectivity of CmoB (Figure 3a). [23] Am odel of CNMT in complex with cxSAM, which possesses an atural S-configured sulfonium centre,i ndicates that Glu204, Glu207, and Ty r81 residues would be in closest proximity to the S-carboxymethyl group. These residues were each subjected to site-directed mutagenesis,introducing either smaller residues that could provide more space to accommodate the larger carboxymethyl group, or basic residues which may form favourable electrostatic (salt-bridge) interactions with the carboxylic acid moiety.
Eleven CNMT mutants in total (E204A/S/K, E207G/S/K/ R, and Y81F/H/K/R,) were assessed for activity with THIQ 4 and cxSAM. In order to reduce competing methylation The CmoB active site highlights three residues thought to interact with the carboxymethyl group (PDB ID:4QNU). Bottom:WTCNMT with cxSAM positioned in the active site (based on PDB ID:6 GKV,which has SAH bound), highlightingthree residues predicted to be in proximity to the carboxymethyl group. b) CNMT time courses with THIQ 4 (0.25 mm)and SAM or cxSAM were conducted using either 1mm SAM (top/blue) or 1.6 mm cxSAM (bottom/red), with commercial SAM comprising only the active S enantiomer at 80 %purity and cxSAM being synthesiseda samixture of enantiomers. c) COMT activity assays with catechol 2 (0.2 mm)a nd SAM or cxSAM were conducted for 5hat 37 8 8Ctocompare %methylationa nd carboxymethylation, using either 1mm SAM (top/blue) or 1.6 mm cxSAM (bottom/red). Assays were done in triplicate and standard errors calculated. occurring due to co-purification of SAM with CNMT,t he enzymes were pre-incubated for an hour with THIQ 5 and washed with reaction buffer prior to assay with cxSAM and 4.
All E204 and E207 mutants tested showed complete loss of activity with cxSAM. TheE204 residue is situated close to the sulfonium centre of the cofactor and ammonium ion of the substrate,w hich suggests it may be important for substrate and/or cofactor binding via electrostatic interactions. [26] In contrast, all of the Y81 mutants showed comparable or higher activity with cxSAM and significantly lower activity with SAM compared with the WT,with the Y81R mutant showing greatest cxSAM selectivity (Figures S3, S4). At this stage,w e have not pursued determination of accurate kinetic parameters because the synthetic cxSAM produced is am ixture of diastereoisomers,the separation of which is problematic due to the instability of the cofactor. In addition, as stated above, CNMT co-purifies with SAM, which would further complicate determination of accurate kinetic parameters.Weinstead looked to demonstrate enzyme selectivity through separate time course assays for WT and Y81R/K CNMT with SAM or cxSAM and substrate 4 (Figure 3b). This showed that methylation of 4 with SAM by the WT CNMT is complete in 30 min, whilst both Y81R/K mutants afford less than 25 % methylation over the same time period. In contrast, carboxymethylation of 4 with cxSAM is 23 %a nd 28 %h igher than the WT for Y81K and Y81R, respectively,a cross the 0-4 h time period. Theo bservation that Y81K affords lower levels of carboxymethylated product (4a)compared with the WT in 17 ha ssays described above (Figure 2a)m ay be due to instability of the Y81K mutant over the longer time course.
In addition, competition assays,w ith varying ratios of SAM/cxSAM and THIQ substrate 4,also reveal that mutants Y81R/K have improved selectivity for cxSAM ( Figure S5). We hypothesise that the basic side chains of R/K81 mutants may be suitably positioned to form polar interactions with the carboxyl group of cxSAM similar to the interactions observed in CmoB. [19,21] Taken together,t he data presented here provides ap roof-of-concept that the selectivity of MTs for cxSAM can be significantly improved by rational targeted mutagenesis.

Site-Directed Mutagenesis of COMT.
As imilar approach was adopted to engineer COMT variants with higher selectivity for cxSAM. Within the active site of COMT,w eo bserved am ethionine residue (M40) in close proximity (2.7 )t ot he methyl group of SAM (Figure S6). Accordingly,s ix point mutants were produced (M40K/R/H/A/S/C) in order to alter the electrostatic and steric interactions of this residue with cxSAM. Of these mutants,M 40A showed the most significant increase in activity with cxSAM (Figure 3c and Figure S7). We propose that this is due to increased space surrounding the carboxymethyl group of cxSAM within the active site,l eading to reduced steric hindrance.Whilst M40A has increased activity with cxSAM, the regioselectivity of carboxymethylation was poor.S ince the COMT Y200L mutant is known to achieve ahigh regioisomeric excess (re)ofmore than 90 % meta, [25] we produced an M40A/Y200L double mutant. Assays conducted over five hours showed that the double mutant has increased regioselectivity with cxSAM, affording meta-carboxymethylation with high re (94 %; Figure 3c). Moreover,M 40A/ Y200L also showed a2 1% decrease in methylation yields when assayed with SAM, compared to the WT.A lthough both the M40A and M40A/Y200L mutants exhibited improved selectivity for cxSAM versus SAM, further mutagenesis may be necessary to deliver COMT variants with higher selectivity for more efficient in vitro or in vivo carboxymethylation reactions.

CmoA-MT Coupled Assay.
We next sought to couple CmoA with CNMT to generate cxSAM enzymatically for carboxymethylation of the THIQ substrate 4 in ac ascade reaction (Figure 4a). Initially,w e envisaged using chorismate mutase (CM) to generate prephenate from chorismate,which can be isolated in high yield from an E. coli KA12 CM-deletion strain. [27,28] However, after optimising conditions for the production of cxSAM from SAM and chorismate with CM and CmoA, it was apparent that equivalent and in some cases higher yields of cxSAM could be generated if CM was omitted from the reaction ( Figures S8, 9). Chorismate is well known to undergo aspontaneous Claisen rearrangement to prephenate,a nd although CM accelerates this reaction am illion-fold, the rate for the non-enzymatic rearrangement is sufficient for efficient CmoA-catalysed SAM carboxylation. Following optimisation, at andem coupled assay was developed that utilises chorismate and CmoA to give cxSAM (ca. 80 %), with some residual SAM (ca. 20 %) remaining,f ollowed by addition of CNMT and the substrate 4.Through this method, CmoA and the more cxSAM-selective CNMT Y81R mutant afforded 70 AE 1% conversion of THIQ 4 to carboxymethylated product 4a,w ith only 16 AE 1% methylated THIQ 4b produced (Figure 4b,F igures S10, 11). Whilst not completely selective, the final 4.5:1 ratio of carboxymethylation to methylation observed with CmoA-CNMT (Y81R) compares favourably with the coupled assays with CmoA-WT CNMT,w hich give a1 :1 mixture of carboxymethylated to methylated products. Since both SAM and prephenate are present in bacterial cells, we envisage that the combination of CmoA and an engineered MT,with higher selectivity for cxSAM over SAM, may have potential to generate carboxymethylated products in vivo as well as in vitro.
In order to further expand the synthetic utility of CmoA-MT cascade reactions,w ee nvisaged generating the new cofactor carboxy-S-adenosyl-l-ethionine (cxSAE), which may enable MT-mediated transfer of ac hiral 1-carboxyethyl group (Figure 5a). To test this,s ynthetic cxSAE, generated from alkylation of SAH with (AE)-2-bromopropionic acid (BPA), was incubated with CNMT WT as well as the Y81K and Y81R mutant enzymes ( Figure S12). TheWT, Y81K, and Y81R enzymes showed significant activity with synthetic cxSAE, giving 25, 13, and 32 %o f1 -carboxyethyl THIQ 4c, respectively.T od etermine the configuration of the product 4c,t he substrate THIQ 4 was also separately alkylated with (S)-and (R)-BPA. However,c hiral HPLC analysis revealed that the alkylation of 4 proceeds with significant racemisation. This is likely due to enolisation of BPAa tt he elevated temperature of the reaction (> 40 8 8C). [29] In addition, BPAcan form an a-lactone,w hich could lead to the alkylation of 4 proceeding with overall retention of configuration, whilst direct alkylation with BPAp roceeds with inversion of stereochemistry.I nl ight of this, 4 was alkylated with (R)and (S)-BPAm ethyl esters and then hydrolysed to afford 1carboxyethyl THIQ standards,( S)-4c and (R)-4c,i nh igh enantiomeric excesses (ee;95% ee in each case). Using these standards,i tw as apparent from chiral HPLC that CNMTcatalysed carboxyethylation of THIQ 4 with synthetic cxSAE, generated from racemic BPA, is completely stereoselective giving only (R)-4c.T of urther probe the stereochemistry of this process,s ynthesis of cxSAE was also attempted using homochiral BPAm ethyl esters,however alkylation reactions in water or mixed aqueous/organic solvents proved problematic.T herefore,S AH was separately alkylated with (R)-and (S)-BPAa tl ow temperature to minimise racemisation, and the resulting cxSAEs were similarly incubated with WT, Y81K, and Y81R CNMT and THIQ substrate 4.W hilst the yields of 1-carboxyethyl THIQ were similar for (AE)-and (S)-BPA-derived cxSAE, the yields with (R)-BPA-derived cxSAE were significantly reduced. Surprisingly,h owever, the R-configured carboxyethyl THIQ product (R)-4c was formed in all cases (Figure 5b).
Four diastereoisomers of cxSAE can be produced upon alkylation of SAH with BPA. However, the separation of these diastereoisomers and subsequent determination of their absolute configurations is very difficult, particularly given their instability.Consequently,itwas not possible to reach any conclusions about the stereochemical course of the process using synthetic cxSAE. To avoid the complication of using diastereoisomeric mixtures of cxSAE, we attempted to generate cxSAE enzymatically,w hich should afford only one stereoisomer of cxSAE. Firstly,ahuman methionine adenosyltransferase (hMAT2A) mutant (I322V), which is known to show promiscuity with methionine derivatives, [11,17] was used to form S-adenosyl-l-ethionine (SAE) from ATP and ethionine (Figure 5a). CmoA was then added along with prephenate to form cxSAE in 13 %y ield. Thee nzymatic cxSAE was then incubated with CNMT and THIQ 4,leading to production of carboxyethyl THIQ 4c.Aswith the synthetic cxSAE, only (R)-4c was evident from chiral HPLC-MS ( Figure S13). Based on the crystal structure of CmoA in complex with cxSAM (PDB ID:4 GEK) and computational predictions of likely positions of precursors SAM and prephenate in the CmoA active site, [21] we predict that cxSAE would possess 1-(S)-carboxyethyl group and an S-configured sulfonium centre ( Figure S14). In light of this,w ec onclude that the CNMT carboxyethylation of THIQ 4 with cxSAE proceeds with inversion of configuration, which is also consistent with the native methylation reactions of MT,which have been shown to proceed with inversion using SAM possessing ac hiral methyl group. [28] Taken together, these results suggest that CxSAE possessing a1 -(R)-carboxyethyl group,g enerated synthetically from BPA, is not turned over by CNMT.

Conclusion
In summary,wehave shown that two typical members of the ubiquitous Class Imethyltransferase,COMT and CNMT, can utilise cxSAM as an alternative cofactor to generate carboxymethylated products.S tructure-guided mutagenesis was used to engineer COMT and CNMT mutants with improved selectivity for cxSAM. This enabled at andem enzyme reaction to be developed with the cxSAM synthase (CmoA) and CNMT,d elivering carboxymethyl-THIQ in good yields.T othe best of our knowledge,there are no other reports describing the combination of CmoA with MTs to create new pathways to carboxymethylated rather than methylated products.P revious approaches to produce SAM analogues as co-factors for MTs have relied on multistep chemical or enzymatic synthesis for in vitro applications, which are limited by the costs of precursor and the instability of SAM analogues.Our approach opens up the possibility of combining CmoA and engineered MTs to create new structural diversity,invitro and in vivo,from entirely natural precursors,thereby obviating the need for multistep chemical or enzymatic synthesis and purification steps.
MTs are amongst the most common enzymes in nature, methylating avast array of substrates,from small metabolites to biomacromolecules.A ddition of ac harged and polar carboxymethyl group,rather than asmall hydrophobic methyl substituent, could therefore lead to many new products with significantly altered physiochemical properties,b iological activities,a nd potentially new functions.M oreover,t he additional carboxyl group provides ah andle for further selective derivatisation using conventional amide coupling chemistry or powerful metallaphotoredox reactions, [31][32][33][34] that can be performed under mild aqueous (biocompatible) conditions. [35] Such downstream chemistry could be used to fine-tune the properties of bioactive molecules and can provide further opportunities for regioselective bioconjugation and labelling. [5][6][7][8][9][10] Finally,w eh ave demonstrated that CmoA can be used to generate anew co-factor,cxSAE, which was accepted by CNMT,thereby enabling the stereoselective transfer of achiral 1-carboxyethyl group to aTHIQ substrate. Several C-methyltransferases,including GlmT and MppJ, [36][37][38] have been described that transfer am ethyl group to ap rochiral substrate and create an ew stereogenic centre. However,weare unaware of any other examples where aMT has been shown to catalyse the stereoselective transfer of ac hiral functional group from the co-factor to the substrate.