Site-specific bioalkylation of rapamycin by the RapM 16-O-methyltransferase

Characterisation of a rapamycin O-methyltransferase (RapM) and its utilisation in coupled reactions, with an improved variant of the human methionine adenosyl transferase (hMAT2A), results in new regioselectively alkylated rapamycin derivatives.


Introduction
The natural product rapamycin 1, rst isolated from Streptomyces hygroscopicus NRRL 5491 (reclassied as Streptomyces rapamycinicus sp. nov.), 1,2 is a polyketide macrolide antibiotic and a potent immunosuppressant that is widely used in the clinic. 3,4 Rapamycin's mechanism of action derives from its binding to the FKBP12 immunophilin, with the resultant rapamycin-FKBP12 complex acting to inhibit downstream signalling pathways involving the protein kinase mTOR (mammalian target of rapamycin). 5 Rapamycin analogues (rapalogs) have been approved for the treatment of a variety of diseases, demonstrating the versatility and wide-ranging biological activity of this natural product. For instance, the semisynthetic derivative Everolimus possesses efficacy against a number of disorders 6 such as renal angiomyolipoma, 7 neuroglial and neuroendocrine tumours, 8,9 and breast cancer. 10,11 Other rapalogs possess a shorter half-life for reduced systemic immunosuppression, 12,13 or have displayed neuroprotective abilities and show promising potential as treatment for ischemic stroke victims. 14 In light of this, there has been considerable interest in the diversication of the rapamycin scaffold to create novel rapalogs with altered or improved physicochemical and pharmacological properties.
Rapamycin is a highly complex molecule and although total syntheses of rapamycin have been reported, 15,16 these syntheses are unlikely to deliver the quantities of rapalogs that would be required for drug development. While semi-synthetic rapalogs have received FDA-approval, the scope and exibility of this approach is limited with regards to amenable diversication of the rapamycin scaffold. The elucidation of the rapamycin biosynthetic gene functions 17,18 has enabled methods including precursor-directed biosynthesis and mutasynthesis, to deliver rapalogs rapidly and in higher titres by fermentation using engineered strains. [19][20][21][22][23][24] However, the modications introduced by these approaches have so far been largely constrained to the cyclohexane starter unit or the pipecolic amino acid, with only a few examples of modications made to the remainder of the rapamycin scaffold. 18,25 The polyketide synthase (PKS) enzymes, RapA, B and C along with the nonribosomal peptide synthetase pipecolate incorporating enzyme RapP are responsible for assembly of the macrocyclic intermediate prerapamycin 2. 17 Post-PKS tailoring enzymes, including three O-methyltransferases RapM, I and Q, and two P450 monooxygenases RapJ and N, are required to transform prerapamycin, the rst PKS free intermediate, into the mature macrocycle 1 (Fig. 1). 18 Gene knockout experiments suggest that RapI, M and Q are regiocomplementary AdoMetdependent methyltransferases that methylate C39, C16 and C27 hydroxyl groups during the biosynthesis of rapamycin respectively. Given that some AdoMet-dependent methyltransferases have been shown to transfer alternative alkyl groups from AdoMet analogues, [26][27][28][29][30][31] we envisaged that the regiocomplementary rapamycin methyltransferases might be utilised in the diversication of distal positions of the rapamycin structure.
The natural co-factor AdoMet, utilised by the majority of methyltransferases, is biosynthesised from ATP and L-methionine by the enzyme methionine adenosyl transferase (MAT). 32 Production of AdoMet analogues in vivo carrying alternative alkyl groups on the sulphonium centre has been demonstrated by feeding methionine analogues to engineered Escherichia coli (E. coli) cells possessing MAT, with the resultant alkyl-donors subsequently utilised by methyltransferases to alkylate proteins. 33 During the completion of the work described in this manuscript, another study was reported describing the use of MAT enzymes in vitro, to generate AdoMet analogues that were used in coupled enzyme reactions with an O-methyltransferase that catalysed subsequent alkylations of the C4 0 -hydroxyl group of the sugar moiety of a biosynthetic desmethyl intermediate of rebeccamycin. 34 Other reports have utilised synthetic AdoMet analogues as alkyl-donors. [26][27][28][29]35,36 However, the chemoenzymatic route via MAT offers several advantages. Firstly, AdoMet analogues produced enzymatically using MAT are diastereomerically pure; 37 conversely the synthesis of AdoMet analogues, from S-adenosyl-homocysteine and alkyl halides, generates a mixture of diastereomers 37,38 at the sulphonium centre, of which only the (S)-congured diastereoisomer is accepted by methyltransferases. 37,39 Synthetic AdoMet analogues also need careful purication to remove the unwanted (R)-congured diastereoisomer and residual S-adenosyl-homocysteine (SAH), both of which can inhibit methyltransferases. 40,41 Additionally, some AdoMet analogues are unstable 29,42,43 so an in situ enzymatic preparation, using MAT, can minimise the formation of degradation products. Finally, unlike AdoMet, methionine analogues are membrane permeable and can therefore be fed to cells possessing MAT to generate alternative AdoMet derivatives in vivo. 33 Here we describe the rst in vitro characterisation of a rapamycin O-methyltransferase, and its utilisation in coupled reactions with an improved variant of the human enzyme hMAT2A, creating alkylated derivatives of rapamycin at the C16-Oposition.

Characterisation of the RapM in vitro
The wild-type rapM gene sequence was amplied from a pUC18-rapM construct (Biotica) 18 and sub-cloned into the E. coli expression vector pET28(a) generating the expression construct pET28a-rapM (Fig. S1A †). The RapM protein was heterologously expressed in E. coli BL21 (DE3) cells using standard methods, and the soluble N-terminal His 6 -RapM fusion protein (Fig. S1B †) was subsequently puried using Ni-NTA affinity chromatography.
The in vitro activity of the RapM enzyme was evaluated using AdoMet (Sigma Aldrich UK) as a co-factor and prerapamycin (BC150) 2, which was previously isolated from an S. rapamycinicus deletion strain decient in the genes encoding the post-PKS tailoring enzymes. 18 In addition, 16-, 27-, 39-tri-O-desmethyl rapamycin (BC231) 7 (Fig. 2), derived from an engineered S. rapamycinicus strain lacking the methyltransferaseencoding genes rapI, M & Q, 44 was also evaluated as a substrate for RapM. Analysis of the BC231 7 reaction by C 8 -reverse phase high performance liquid chromatography (RP-HPLC) (Fig. 3A) revealed formation of a new product with a longer retention time, t R ¼ 7.13 min, compared with the starting material t R of 5.99 min. The new product was isolated and high resolution mass spectrometry (HRMS) analysis revealed an observed m/z 908.5156 consistent with the [M + Na] + ion for a mono-methylated derivative of BC231 (Table S1 †). Based on the proposed post-PKS tailoring steps ( Fig. 1), 18 this new product was expected to be 16-O-methyl-BC231 7m. When prerapamycin (BC150) 2 was used as a substrate with RapM, C 18 -RP-HPLC analysis ( Fig. 3B) revealed two peaks for the substrate (t R ¼ 6.77, t R ¼ 7.67 min) due to the presence of interconverting rotamers about the amide bond, 45 along with new products with longer retention times which were predicted to be rotamers of 16-O-methyl-BC150 2m (t R ¼ 8.22, t R ¼ 9.36 min). The two product peaks for 2m were separated and HRMS revealed the same m/z for each, consistent with mono-methylated prerapamycin rotamers (Table S1 †). Additionally, the individual rotamers of both prerapamycin 2 and product 2m were separated by HPLC, le at 21 C for 16 h, and then re-analysed again by HPLC, which revealed reequilibration back to the original ratio of rotamers ( Fig. S2 †).

Conrmation of the regioselectivity of RapM
Based on gene deletion and complementation experiments with S. rapamycinicus the three methyltransferases RapM, Q and I, are predicted to methylate at the 16-, 27-and 39-O-positions respectively during rapamycin biosynthesis. 18 However, as of yet there has been no in vitro characterisation of these methyltransferases. Therefore, to conrm the regioselectivity of RapM, the mono-methylated products 2m and 7m were subjected to tandem mass spectrometry revealing a fragmentation pattern consistent with rapalogs described previously (Fig. 4). 18 In addition to the key MS/MS product 'f' ions of 2, 2m, 7 and 7m ( Table 1) a series of ions denoted as 'e' ions which result from the loss of CO 2 , H 2 O, and in the case of methylated products CH 3 OH, can also be used to assign structures and identify the methylation site of RapM ( Fig. S3 and S4 †). The f 2 product ion containing only the C39-OH site was identical for 2 and 2m (m/z [M + Na] + 614.2), as well as both 7 and 7m (m/z [M + K] + 644.2), indicating that RapM has not methylated the C39-hydroxyl. The f 1 and f 3 ions were both observed with a +14 mass difference between 7 and 7m, which would be consistent with methylation at either the C16-or C27-hydroxyl groups of compound 7. However, the MS/MS analyses of 2 and 2m also revealed a +14 increase in the f 1 product ion. As prerapamycin 2 lacks the C27hydroxyl, the presence of an additional methyl group in the f 1 fragment from 2m suggests that RapM is indeed the 16-Omethyltransferase for rapamycin (Fig. S4 †). Furthermore, reactions of RapM with previously isolated 46 compounds BC204 8 (16,27-bis-O-desmethyl-39-demethoxy-rapamycin) and BC207 9 (27-O-demethyl-39-demethoxy-rapamycin), showed conversion of 8 into 9, i.e. RapM catalyses methylation at the C16-hydroxyl ( Fig. 2 and S5A †). The absence of activity of RapM with 9, a rapalog where the C16-hydroxyl is blocked by methylation but the C27-hydroxyl remains available, further conrms that RapM does not accept the C27-O-methylation site and is exclusive for the C16-position (Fig. S5B †).  Finally, the 16-O-methyl-BC231 compound 7m was analysed by proton-NMR (600 MHz), which revealed that 7m exists as a mixture of amide rotamers, in line with previous NMR studies of rapamycin 1. 47 From COSY and TOCSY experiments in DMSO-d 6 , most of the 1 H signals for the major trans-isomer of 7m could be assigned (Table S2 and Fig. S6-S8 †) and are consistent with NMR data reported previously for rapamycin 1 as described by Pagano. 47 Notably a 3H singlet in the 1 H NMR of 2m is evident at 3.06 ppm, which corresponds with the C16-Omethyl signal assigned previously at 3.04 ppm, for rapamycin in DMSO-d 6 . This, along with the MS/MS experiments conrms the regioselectivity of the 16-O-methyltransferase RapM.

Alkylation of rapalogs using MAT and RapM in tandem
To effect regioselective alkylation of rapamycin, a one-pot coupled enzyme system with the human methionine adenosyl transferase hMAT2A and RapM was envisaged. The hMAT2A enzyme has been reported to be promiscuous, 33,34 and further mutagenesis of active site residues around the binding pocket for the AdoMet methyl substituent has been shown to improve activity with methionine analogues possessing larger S-alkyl substituents. 33 Based on this precedent a series of active site mutants of hMAT2A were created by site-directed mutagenesis of a hMAT2A expression vector pNIC28-Bsa4-hMAT2A that was kindly provided by Udo Oppermann (Oxford). Selected mutants were overproduced in E. coli Rosetta™ 2, puried by Ni-NTA affinity chromatography, and evaluated with L-ethionine and Sallyl-L-homocysteine (Sahc) as substrates, revealing that the I322V mutation provides the highest efficiency with the methionine analogues (Fig. S9 †).
The mutant hMAT2A (I322V) and RapM were then used in tandem to effect alkylation of prerapamycin 2 and BC231 7 with ATP and L-methionine, L-ethionine or S-allyl-L-homocysteine, forming in situ AdoMet, S-adenosyl-L-ethionine (SAE) or S-adenosyl-L-allyl-homocysteine (SAAH) respectively. The subsequent transfer of methyl, ethyl and allyl groups onto the rapalogs was analysed by RP-HPLC (Fig. 5). The alkylation of BC231 7 in coupled assays with hMAT2A (I322V) and methionine analogues resulted in two new products ethyl-BC231 7e (t R ¼ 7.56 min) and allyl-BC231 7a (t R ¼ 7.73 min) with longer retention times than the methylated product 7m. HPLC analysis of the coupled reactions with prerapamycin 2 revealed new products ethyl-BC150 2e (t R ¼ 8.90 min and 9.99 min) and allyl-BC150 2a (t R ¼ 9.18 min and 10.28 min). As with the corresponding methylated product 2m, the ethyl and allyl derivatives also result in dual peaks on HPLC analysis consistent with the presence of rotamers.
The minor amounts of methylated products observed in reactions utilising L-ethionine or S-allyl-L-homocysteine are presumably due to the presence of residual active site-bound AdoMet from co-elution during hMAT2A (I322V) and RapM protein purication. HRMS of the new alkylated products 2e, 2a, 7e and 7a was consistent with mono-alkylated derivatives (Table S1 †) and UV analysis showed that all the new products possessed the distinctive rapamycin l max at 267, 277 and 288 nm due to the triene chromophore. Finally partial assignment of the 1 H and 13 C NMR signals for the major rotamer of ethyl-BC231 7e using COSY, TOCSY, NOESY and HSQC experiments was also consistent with the expected structure of 7e (Tables S3  and S4 and Fig. S10-S14 †). The efficiency of alkylation of the prerapamycin 2 and BC231 7 in the coupled hMAT2A/RapM assays is clearly reduced with increasing size of the alkyl group on the sulphonium centre of the AdoMet analogue. Full conversion of 7 and 75% conversion of 2 to the methylated rapalogs was achieved for the assays with L-methionine, whereas ethylation and allylation reactions proceed to 19% (2e), 51% (7e), and 10% (2a), 33% (7a), under the same conditions (Fig. S15 †).  In order to test the efficiency of alkylation with RapM independently of hMAT2A, enzymatically prepared AdoMet, SAE and SAAH (Fig. S16 †) were isolated and then separately incubated with RapM and BC231 7. Whilst 7m and 7e were produced at similar levels to that observed in the tandem reactions, the production of the allylated product 7a was over ve-fold lower compared with the tandem assay (Fig. S17 †). To explore this further, competitive assays with AdoMet and equimolar SAE or SAAH were performed (Fig. S18 †), which showed that SAE has no effect on the methylation reaction of BC231 7. On the other hand, SAAH, and/or the SAAH degradation product 5 0 -S-allyl-5 0thioadenosine 29,42,43 (Fig. S16 †), appear to inhibit RapM methylation of 7 (Fig. S18 †). These observations were used to improve the productivity of the tandem reaction with SAAH (Fig. S19 †); by decreasing the concentration of hMAT2A (I322V), with RapM in 7.5-fold excess, yields of 7a can be increased to 72%. Presumably the excess RapM, relative to hMAT2A, prevents the accumulation of SAAH, minimising degradation and possible subsequent inhibition of RapM.
Finally, the RapM methyltransferase was assayed against a range of concentrations of the best substrate BC231 7 (2.5-250 mM) with a set concentration of AdoMet (500 mM), to give the kinetic constants K m ¼ 48.9 AE 7.1 mM and k cat ¼ 1.47 AE 0.072 min À1 . The relatively low catalytic rate of RapM is unsurprising given the nature of the methylation reaction; RapM is a tailoring enzyme for a complex polyketide natural product, and there is unlikely to be an evolutionary drive for a catalytically efficient methyltransferase, given that the intricate assembly process of the polyketide is more likely to be the limiting factor in overall rapamycin production. The lower relative activity of RapM with AdoMet and the natural precursor BC150 2 compared with BC231 7, an engineered compound possessing a C9 keto group that is not a natural intermediate (Fig. S15 †), is also consistent with the proposed timing of tailoring steps in rapamycin biosynthesis. 18 The proposed optimum pathway (Fig. 1), 18 suggests RapM methylation occurs aer oxidation at C9 by RapJ. Whilst a preferred route may have evolved to ensure each tailoring enzyme acts on its optimal substrate, the fact that BC150 2 is a substrate for RapM indicates that multiple minor parallel pathways are most likely operative, resulting in a diverse range of intermediary compounds all leading to the nal most highly modied product rapamycin.

Conclusions
In summary, we have characterised the O-methyltransferase RapM, conrming its specicity for methylation of the 16-OH position of rapamycin. The ability of RapM to transfer alkyl groups onto substrates including a biosynthetic precursor prerapamycin and an engineered rapalog BC231 has been shown, leading to alkyl-diversication of this clinically important polyketide. In addition, the catalytic constants for the enzyme when assayed with BC231 have been determined with AdoMet.
The enzymatic alkyl-diversication of rapamycin offers an attractive route to generating rapalogs with altered physicochemical and biological activity. Indeed, previous studies 48, 49 have noted that structural changes to rapamycin at the C16position can be well tolerated, and in addition, that groups attached to the C16-hydroxyl group can be localised to a specic region of the rapalog bound FKBP-mTOR complex. Moreover, the regioselective attachment of orthogonal chemical handles to the rapamycin scaffold opens up the possibility of in vivo labelling of the rapamycin:FKBP complex, facilitating studies of the interactions of rapamycin with protein targets. Alternatively, the immobilisation of rapamycin offers a new route to coating clinical apparatus such as stents, or allows functional assays of rapamycin in complex with its protein binding partner(s). Given that methyltransferases are one of the most common classes of enzymes found in secondary metabolism, the methods described in this paper could be applied to a wide range of other complex bioactive natural product scaffolds.

Construction of pET28a-rapM
The wild-type rapM gene sequence was PCR amplied using Phusion HF DNA polymerase (New England Biolabs) from a pUC18-rapM construct (Biotica) 18 with the primers His 6 -rapM F and R (Table S5 †) to introduce NdeI and XhoI restriction sites at the 5 0 and 3 0 ends respectively. The gel-puried PCR product was doubly digested with NdeI and XhoI restriction enzymes (New England Biolabs) and ligated into a linearised pET28a(+) vector (Novagen) which had been similarly digested. The resultant construct pET28a-rapM was veried by nucleotide sequencing (GATC Biotech).

Site-directed mutagenesis of hMAT2A
The construct pNIC28-Bsa4-hMAT2A (provided by Udo Oppermann, Oxford) was used as a template for site-directed mutagenesis to generate hMAT2A mutants with expanded methionine-binding pockets as described by Wang and coworkers. 33 Primers carrying the degenerate codon KYT (Table  S5 †) were designed to generate mutants at the I117, V121 and I322 amino acid positions with either alanine, serine, valine or phenylalanine replacements. The mutants were veried by nucleotide sequencing (GATC Biotech).

Overexpression and purication of His-tagged RapM and hMAT2A
Transformant cells of E. coli BL21 (DE3) (pET28a-rapM) or Rosetta™ 2 (pNIC28-Bsa4-hMAT2A) were cultivated (37 C, 200 rpm agitation) in LB supplemented with kanamycin (pET28a-rapM, 50 mg mL À1 ) or kanamycin and chloramphenicol (pNIC28-Bsa4-hMAT2A, 50 mg mL À1 and 25 mg mL À1 ). The cells were allowed to reach a density of OD 600 0.6-0.8 before protein expression was induced with the addition of isopropyl b-D-1thiogalactopyranoside (IPTG, 0.5 mM). The cells were further cultivated for 4 h at 30 C, 200 rpm before the cells were harvested by centrifugation. The pelleted cells were resuspended in lysis buffer (5 mL per pellet from 800 mL culture; RapM lysis buffer: 50 mM Tris-HCl pH 8.5, 500 mM NaCl, 5% glycerol (v/v); hMAT2A lysis buffer: 50 mM Tris-HCl pH 8, 50 mM NaCl, 10% glycerol (v/v)). The cells were lysed by sonication and the lysates claried by centrifugation. The soluble lysate was then applied over a Ni-NTA column (pre-equilibrated in lysis buffer) and washed with lysis buffer containing 20 mM imidazole (7 column volumes, CV) followed with a 60 mM imidazole containing buffer (5 CV). The protein was eluted with 5 CV buffer containing 250 mM imidazole. All fractions were analysed by SDS-PAGE. The eluate fraction containing the RapM or hMAT2A protein was simultaneously buffer exchanged into lysis buffer without imidazole and concentrated using a Vivaspin 20 centricon, MWCO 10 000 (Sartorius Stedim Biotech). The puried enzymes were stored at À80 C.
The hMAT2A assays were run as follows: 1 mM DTT, 3 mM MgCl 2 , 1.5 mM ATP, 1.5 mM L-methionine (or analogue), 10 mM hMAT2A, in 20 mM phosphate buffer pH 7.4. The reactions were incubated at 30 C, 800 rpm for 30 min, quenched with an equal volume of methanol and claried by centrifugation (13 000 Â g, 4 C, 10 min). The assays were monitored by HILIC HPLC and the AdoMet analogues veried by ES + LC-MS.

NMR of 16-O-methyl BC231 and 16-O-ethyl BC231
Compounds 7m and 7e were prepared by purifying over C 18 RP-HPLC and collecting fractions containing the product peaks. The fractions were dried under a stream of N 2 and then in vacuo. The dried samples were reconstituted in DMSO-d 6 in a glove bag lled with nitrogen, and transferred to a DMSO-matched Shigemi tube (300 mL sample volume). The NMR spectra were recorded at a temperature of 298 K on a Bruker Avance AMX 600 MHz spectrometer equipped with a 5 mm inverse triple resonance cryoprobe. Spectra processing was performed using Topspin 3.1 soware (Bruker).