S-Adenosyl-L-methionine (AdoMet) is the major methyl donor for biological methylation reactions catalyzed by methyltransferases. We report the first chemical synthesis of AdoMet analogs with extended carbon chains replacing the methyl group and their evaluation as cofactors for all three classes of DNA methyltransferases. Extended groups containing a double or triple bond in the β position to the sulfonium center were transferred onto DNA in a catalytic and sequence-specific manner, demonstrating a high utility of such synthetic cofactors for targeted functionalization of biopolymers.

AdoMet (1) is a biological sulfonium compound¹ known as the major methyl donor for enzymatic methylations of various biopolymers such as DNA, RNA, proteins and small biomolecules. These reactions are catalyzed by numerous methyltransferases (Scheme 1a) and have a critical role in the cell^2,3. However, enzymatically obtained cofactor analogs with an ethyl (2) or propyl (3) group replacing the methyl group show a marked decline of alkyl transfer rates^4,5, suggesting that methyltransferase-catalyzed transfer of larger saturated alkyl chains is probably not possible. In nature, AdoMet 1 is formed by AdoMet synthetases from L-methionine and ATP¹, but methyl group replacements larger than ethyl or propyl are not well tolerated^4,5. Therefore, we chemically synthesized a series of AdoMet analogs (2–5) with extended methyl replacements carrying alkenyl and alkynyl groups, in addition to alkyl groups. We accomplished this by alkylation of S-adenosyl-L-homocysteine (AdoHcy, 6; Scheme 1b) similar to the chemical synthesis of AdoMet 1 (ref. 6). The AdoMet analogs were obtained as approximately 1:1 diasteromeric mixtures at the sulfonium center with 40–90% conversion (Supplementary Fig. 1 online). The molecular structures were confirmed by ESI-MS and ¹H-NMR analysis (Supplementary Methods online). As only the S-epimer of AdoMet functions as cofactor for methyltransferases⁷, we separated the diasteromeric mixtures chromatographically for use in enzymatic reactions.

Reactions catalyzed by AdoMet-dependent methyltransferases and chemical synthesis of AdoMet analogs. (a) General reaction scheme for enzymatic transfer of methyl or extended groups from the natural cofactor AdoMet (1) or its analogs (2–5) onto various nucleophiles (Nu⁻) in DNA, RNA, proteins or small biomolecules. (b) Chemical synthesis of AdoMet analogs 2–5. (c) General structures of modified nucleobases in DNA formed by M.TaqI, M.HhaI and M.BcnIB.

Full size image

The synthesized AdoMet analogs 2–5 were tested with representatives of all three classes of DNA methyltransferases (Scheme 1c): namely, the DNA adenine-N6 methyltransferase M.TaqI (ref. 8; recognition sequence: TCGA (target base underlined)), the DNA cytosine-5 methyltransferase M.HhaI (ref. 9; GCGC) and the DNA cytosine-N4methyltransferase M.BcnIB (ref. 10; CCSGG). Enzymatic activities were investigated using a DNA protection assay, which makes use of the fact that methyltransferase-modified DNA is protected against fragmentation by cognate restriction enzymes. By varying the amount of enzyme, one can assess the reactivity of AdoMet analogs with different DNA methyltransferases or their variants (examples shown in Supplementary Fig. 2 online). Consistent with previous observations^4,5, we detected little to no transfer of the propyl group from 3 under various reactions conditions (Fig. 1a). Notably, the AdoMet analogs 4 and 5 with methyl group replacements containing a carbon-carbon double (allylic system) or triple bond (propargylic system) in the β position to the sulfonium center proved much more efficient. These cofactors render full modification of long natural DNA at substoichiometric amounts of methyltransferases over their recognition sites (Fig. 1a), demonstrating that the poor reactivity of AdoMet analogs with saturated groups can be overcome by placing an activating group (unsaturated carbon-carbon bond) next to the carbon unit that is attacked. Enzymatic transmethylation reactions with AdoMet 1 are thought to proceed via an S_N2 mechanism with inversion of configuration at the transferred methyl group¹¹. The chemical rationale for such reactivation apparently derives from conjugative stabilization of the p orbital at the reactive carbon (hybridization change from sp³ to sp²) formed during the transition state (Fig. 1b).

(a) Estimated enzymatic turnover rates of different DNA methyltransferases with cofactors 1–5. Estimation is based on the minimal molar ratio of DNA methyltransferase to target site that is required for complete modification of DNA in a DNA protection assay. (b) Proposed mechanism for transition state stabilization of enzymatic allyl group transfer from cofactor analog 4 to a nucleophile. S_N2 reactions at saturated carbon atoms proceed via a hybridization change from sp³ to sp², and conjugation of the transiently formed p orbital with the neighboring antibonding π orbital (shown by arrows) facilitates the reaction.

Full size image

Besides electronic factors, steric limitations imposed by the architecture of the catalytic sites in methyltransferases may also affect the efficiency of enzymatic transalkylations. Available crystal structures show that the AdoMet binding pockets in DNA methyltransferases are located near the protein surfaces, and further elongation of the allylic and propargylic cofactors should be possible without substantial loss of reactivity. Side chain shortening of certain residues within the cofactor binding pocket (Q82A mutation in M.HhaI) allows a further improvement of the alkyltransferase efficiency (Supplementary Fig. 2), suggesting that steric adjustments should be possible in many other AdoMet-dependent methyltransferases owing to their high structural similarity¹².

To characterize the modified nucleosides that are produced with the AdoMet analogs 2–5, the nucleoside composition of the modified DNA was determined. All newly formed compounds were identified by ESI-MS and UV absorption as corresponding alkylated derivatives of 2′-deoxyadenosine and 2′-deoxycytidine (Supplementary Table 1 and Supplementary Fig. 3 online). The enzymatic transalkylations of DNA occurred in a highly sequence-specific manner, as indicated by unaltered fragmentation patterns of noncognate restriction endonucleases. Notably, removal of the essential catalytic cysteine residue in M.HhaI (C81S mutation)^9,13 completely abolishes the transalkylation reactions (data not shown), indicating that transfer of the extended groups proceeds via the natural catalytic mechanism. The methyltransferase-mediated transalkylations are truly catalytic; that is, the enzyme is released after the reaction for subsequent turnovers. The transalkylation rates of up to 32 turnovers per hour (Fig. 1a) indicate that the sequence-specific derivatization of DNA is a convenient technique suitable for routine laboratory use. This is in contrast to the recently described AdoMet analogs containing a reactive aziridine group¹⁴: in those compounds, the whole cofactor is coupled to the DNA, preventing further turnovers.

In conclusion, the AdoMet analogs with extended allylic and propargylic groups function as efficient cofactors for DNA methyltransferases, which provides a new method for sequence-specific covalent derivatization of DNA. As exemplified for restriction endonucleases (Supplementary Fig. 4 online), bulky groups can be deposited at specific sites to interfere with the action of DNA-modifying enzymes or DNA-binding proteins. Furthermore, functional groups could be appended to the side chains of the cofactors, sequence-specifically transferred to DNA and then modified with chemical entities in chemoselective ligation reactions. Alternatively, a desired label could be directly attached to the side chain of the allylic and propargylic AdoMet analogs and sequence-specifically transferred to DNA. Besides the described three enzymes, many more DNA methyltransferases are active with the synthetic AdoMet analogs (data not shown). The REBASE¹⁵ database currently lists about 800 DNA methyltransferases that recognize over 200 different DNA recognition sequences spanning two to eight base pairs, offering an unprecedented experimental control over sequence-specific manipulation of DNA with many potential applications, ranging from probes for genetic screening technologies to molecular building blocks in DNA-based nanobiotechnology. Moreover, the newly developed cofactors should, in principle, be suitable for sequence-specific transfer of functional groups or other chemical entities to RNA and proteins using appropriate methyltransferases as catalysts.

Note: Supplementary information is available on the Nature Chemical Biology website.

The authors are grateful to E. Merkiene and M. Čaikovskij for constructing the Q82A variant of M.HhaI, F. Grygas and R. Gerasimaite for technical assistance and K. Glensk for preparing M.TaqI. We thank V. Gabelica and F. Rosu for performing high-resolution ESI-MS measurements at the Center for Analysis of Residues in Traces (CART), laboratory of E. De Pauw, University of Liège, Belgium. C.D. thanks the Deutsche Forschungsgemeinschaft for a stipend within the Graduiertenkolleg 440. This work was supported by grants from VolkswagenStiftung, the Howard Hughes Medical Institute and the Ministry of Science and Education of Lithuania.

Author notes

1. • Christian Dalhoff
   •  & Gražvydas Lukinavičius

   These authors contributed equally to this work.

Affiliations

1. Institute of Organic Chemistry, RWTH Aachen University, Landoltweg 1, D-52056 Aachen, Germany.

   • Christian Dalhoff
   •  & Elmar Weinhold

2. Laboratory of Biological DNA Modification, Institute of Biotechnology, V. Graičiūno 8, LT-02241 Vilnius, Lithuania.

   • Gražvydas Lukinavičius
   •  & Saulius Klimas̆auskas

Competing interests

The authors declare no competing financial interests.

Corresponding authors

Correspondence to Saulius Klimas̆auskas or Elmar Weinhold.

To obtain permission to re-use content from this article visit RightsLink.

Publication history

DOI

https://doi.org/10.1038/nchembio754