Human DNMT1 transition state structure

Significance

DNA methyltransferase 1 (DNMT1) is the major enzyme responsible for maintenance of DNA CpG methylation marks in human cells. The enzyme is a validated target for cancer, but current treatments are mutagenic. Knowledge of the transition state (TS) structure of DNMT1 will inform the chemical reaction mechanism and provide information for TS analog design. Here, we report the subangstrom geometric and electrostatic character of the TS for the DNMT1 methylation of hemimethylated DNA. The experimental and computational TS analysis indicates methyl transfer is the rate-limiting chemical step for the reaction. Methyl group transfer can be characterized as a loose nucleophilic substitution TS. TS analysis of DNMT1 demonstrates an approach to understand a complex epigenetic enzyme.

Abstract

Human DNA methyltransferase 1 (DNMT1) maintains the epigenetic state of DNA by replicating CpG methylation signatures from parent to daughter strands, producing heritable methylation patterns through cell divisions. The proposed catalytic mechanism of DNMT1 involves nucleophilic attack of Cys¹²²⁶ to cytosine (Cyt) C6, methyl transfer from S-adenosyl-l-methionine (SAM) to Cyt C5, and proton abstraction from C5 to form methylated CpG in DNA. Here, we report the subangstrom geometric and electrostatic structure of the major transition state (TS) of the reaction catalyzed by human DNMT1. Experimental kinetic isotope effects were used to guide quantum mechanical calculations to solve the TS structure. Methyl transfer occurs after Cys¹²²⁶ attack to Cyt C6, and the methyl transfer step is chemically rate-limiting for DNMT1. Electrostatic potential maps were compared for the TS and ground states, providing the electronic basis for interactions between the protein and reactants at the TS. Understanding the TS of DNMT1 demonstrates the possibility of using similar analysis to gain subangstrom geometric insight into the complex reactions of epigenetic modifications.

Human DNA methyltransferases (DNMTs) catalyze the formation of 5-methylcytosine (5mC) at CpG sites on DNA, a key epigenetic mark present in the human genome (1). DNA methylation is involved in transcriptional silencing, cellular differentiation, genomic imprinting, and X-chromosome inactivation. In addition, hypermethylation of CpG islands at gene promoter regions has been associated with carcinogenesis (2). Maintenance of DNA methylation patterns is conducted by human DNMT1, a multidomain protein of 1,616 amino acids. The C-terminal methyltransferase domain shows sequence similarities to the bacterial methyltransferases (3). Crystal structures of mouse and human DNMT1 complexed with different substrates have provided a structural basis for DNMT1-mediated maintenance DNA methylation (4, 5). Domain interactions and large conformational changes are responsible for properly positioning hemimethylated DNA within the active site and catalyze methyl transfer from S-adenosyl-l-methionine (SAM) to DNA. Site-directed mutations have offered insights into the structure–function relationship of DNMTs (6, 7), but their transition state (TS) structures have remained unknown.

DNMT1 has been proposed to follow a catalytic mechanism shared by bacterial DNA-(cytosine C5)-methyltransferases (4, 8–10): nucleophilic attack of cytosine (Cyt) C6 by Cys¹²²⁶ of DNMT1, methyl transfer from SAM to Cyt C5, and β-elimination of H5 to produce 5mC in the final step (Fig. 1). Recent quantum mechanics (QM)/molecular mechanics (MM) and molecular dynamics (MD) simulations of the bacterial M.HhaI methyltransferase suggested that Cys¹²²⁶ attack is concerted with methyl transfer (11, 12), and that β-elimination of H5 is the rate-limiting step (12). The combination of kinetic isotope effects (KIEs) and computational chemistry can test predicted reaction mechanisms and can provide a model of the TS structure.

Fig. 1.

DNA methylation catalyzed by DNMT1. (A) Proposed catalytic mechanism for DNMT1 involves three chemical TSs (TS1, TS2, and TS3). SAH, S-adenosylhomocysteine. Cys attack at TS1 brings a negative charge (−1) to the Cyt ring, whereas Cys withdrawal after TS3 restores the aromaticity of the Cyt. (B) Based on the KIE analysis presented here, methyl transfer is chemically rate-limiting for DNMT1 and has a higher energy barrier than the thiol-attack and β-elimination steps. Small forward commitments demonstrate the chemical steps to have a higher energy barrier than the binding and release of substrates.

Enzymes catalyze reactions by forming short-lived TSs from their reactants held in Michaelis complexes (13). The lifetime of a chemical TS is typically around 10⁻¹⁴ s, on the time scale of chemical bond vibrations. No spectroscopic method is generally available to observe the chemical structure of TSs directly for enzymatic reactions (14). TS analysis based on experimental KIEs has provided detailed chemical insights into the catalytic mechanisms of enzymes acting mostly on small molecules and has led to the design of some of the most powerful enzyme inhibitors (15, 16). Enzymes in epigenetic regulations often involve large and complex substrates, creating experimental challenges in both KIE measurements and computational models. Nonetheless, TS analysis can be applied to complex enzyme systems, including the 50S ribosomes (17), as long as the chemical steps can be interrogated with the appropriate isotope labels.

Here, we measured 10 experimental KIEs to investigate the TS and catalytic mechanisms of human DNMT1. By combining these experimental KIE values with QM calculations, we established the subangstrom TS structure for human DNMT1. Our results also show methyl transfer to be the major chemical barrier in the reaction coordinate, rather than the Cys attack, β-elimination from the C5-position, or departure of the 5-methyl Cyt from the catalytic site Cys. The work demonstrates an experimental approach to analyze the TS structures of complex epigenetic enzymes, for unraveling their catalytic mechanisms, and for advancing target-specific drug designs.

Results

Isotope Labeling of DNA and SAM as DNMT1 Substrates.

A library of six hemimethylated DNA and eight SAM substrates with site-specific isotope labels was synthesized to measure the respective KIEs. The isotopic labels were placed in chemical bond positions such that all atoms directly involved in the chemical steps of DNMT1-catalyzed reaction were represented. Six isotopically labeled dCTPs were prepared as the building blocks for the DNA substrates [5-²H, 5′-¹⁴C]-, [5-¹³C, 5′-¹⁴C]-, [6-³H]-, [6-¹⁴C]-, [5′-³H₂]-, and [5′-¹⁴C]-dCTPs through coupled reactions using up to 14 different enzymes (18) (SI Appendix, Figs. S1 and S2). Each isotopically labeled dCTP was incorporated into a 26-bp DNA by in vitro replication using Klenow fragment extension (Fig. 2 A–C). The DNA molecules synthesized as labeled reactants all contained one hemimethylated CpG site, in which the unmethylated 2′-deoxycytidine (dC) residue is enriched with specific isotopes. This substrate design provides a single methylation site per DNA for methyltransferase; therefore, the observed KIEs are not complicated by processive DNA methylations. In addition, eight species of isotopically labeled SAM substrates were synthesized enzymatically from labeled ATP and methionine using SAM synthetase (SI Appendix, Fig. S3), including [Me-¹³C, 8-¹⁴C]-, [Me-¹⁴C]-, [Me-³H₃]-, [5′-¹⁴C]-, [5′-³H₂]-, [³⁶S, 8-¹⁴C]-, [1′-³H]-, and [8-¹⁴C]-SAMs. The ³⁶S-labeled SAM was synthesized from elemental ³⁶S (19). The percentages of enrichment for stable heavy isotopes (i.e., ²H, ¹³C, ³⁶S) in our isotopically labeled reactants were measured by MS. Those values, together with the lists of isotopic starting materials, are summarized in SI Appendix, Tables S1 and S2.

Fig. 2.

Development of chemical tools for KIE measurement on DNMT1. (A) Examples of reactant radiolabels: [6-³H]- and [5-²H, 5′-¹⁴C]-dCTPs and [Me-¹⁴C]- and [Me-¹³C, 8-¹⁴C]-SAMs. Remote labels (5′-¹⁴C for DNA, 8-¹⁴C for SAM) were used as radioactive reporters for stable isotopes (e.g., ²H, ¹³C). (B) Incorporation of specifically labeled dCTPs (red) into a 26-bp hemimethylated DNA by Klenow (exo-) extension. Note the premethylated CpG dinucleotide in the template strand. (C) Analysis of DNA synthesis by nondenaturing polyacrylamide gel and fluorescent staining. Lanes 1–3: 10-bp DNA ladder, DNA template and primer (upper and lower bands), and 26-bp DNA product. (D) Radiometric quantitation of hemimethylated DNA and its fully methylated product was analyzed by the dC and 5m-dC nucleosides in hydrolyzed DNA.

DNA and SAM Show Low Substrate Commitments in Vitro.

In enzymatic reactions, the rate of chemical bond changes can be similar to the rates of substrate binding and product release. These rate similarities (called commitment factors) can obscure the values of the intrinsic chemical isotope effects and must be quantitated to permit calculation of intrinsic isotope effects. We measured the forward commitment factor (C_f) values for DNA and SAM by isotope trapping experiments. These experiments used pulse–chase analysis to trace radiolabeled product formation over the course of the reaction (SI Appendix, Fig. S4 and Table S3). The C_f values for DNA and SAM bound to DNMT1 were found to be small and insignificant (0.016 and 0.0013, respectively). The small C_f values establish that DNA and SAM bind to and release from DNMT1 63 and 770 times, respectively, before each catalytic turnover and are not highly committed to the chemical steps. Small C_f values demonstrate that DNMT1-catalyzed methyl transfer is much slower than substrate binding and release steps. DNA binding requires Cyt base-flipping by DNMT1, and with dsDNA, multiple excursions into the catalytic site are required to achieve the proper catalytic site geometry for methylation. As a result of small C_f values, the experimental KIE values are not significantly reduced by the C_fs and are within experimental error of intrinsic KIEs reporting on the rate-limiting chemical step(s) (Fig. 1B). Our measured C_f values are indicative of the behavior for isolated DNMT1 in vitro. The processivity of DNMT1 may increase in the context of nuclear DNA replication machinery, where DNMT1 is part of a multiprotein complex at the replication foci.

Ten KIEs at Nine Atomic Positions Define the TS Parameters.

We measured 10 KIEs at nine atomic positions in experiments to determine the rate-limiting TS of DNMT1-catalyzed DNA methylation (Fig. 3). Competitive assays with light and heavy substrates in the same reaction give high experimental accuracy (Table 1). The KIEs were determined by the change in isotope ratios traced by ³H or ¹⁴C, comparing the initial substrate with the residual substrate after DNMT1 conversion. KIEs for SAM were directly measured by isotopic depletion of the SAM substrate (Eq. 1), where f is the fraction of conversion and R_s and R₀ are the isotope ratios found in the remaining substrate and the initial substrate. Control reactions without DNMT1 corrected for the spontaneous decomposition of SAM (20) during reactions. Cyt KIEs from DNA were obtained by resolution of isotopes in hemi-(substrate) and fully methylated (product) DNAs following hydrolysis to dC and 5-methyl-2′-deoxycytidine (5m-dC) and separation by HPLC (Fig. 2D and SI Appendix, Fig. S5). The ³H and ¹⁴C isotopes in dC and 5m-dC originated from the unreacted substrate and the fully methylated product, respectively:

$$KIE=\frac{\text{log}\left(1-f\right)}{\text{log}\left[\left(1-f\right)R_s/R_0\right]}.$$ [1]

Intrinsic KIEs report on the chemical structure of the dominant TS for DNMT1. The KIEs from the methyl carbon (Me-C) and sulfur of SAM report directly on the extent of methyl transfer to C5 of Cyt at the TS. The KIEs of [Me-¹⁴C] (1.107), [Me-¹³C] (1.069), and [³⁶S] (1.019) for SAM are near theoretical limits for their respective isotopic masses and establish a TS dominated by the methyl transfer step (Fig. 1B). In addition, the relatively large [6-¹⁴C] Cyt KIE (1.038) and inverse [6-³H] Cyt KIE (0.945) establish the sp³ (tetrahedrally coordinated) character at C6 of Cyt, demonstrating the presence of a covalent bond between the Cys¹²²⁶ thiol and Cyt C6 at the TS. The 10 experimental KIEs provide a comprehensive dataset for resolving the DNMT1 kinetic mechanism and rate-limiting TS at subangstrom resolution.

Fig. 3.

Intrinsic KIEs and TS bonds for DNMT1-catalyzed DNA methylation. [5′-¹⁴C] on DNA and [8-¹⁴C] on SAM are remote control labels (KIE = 1.000) for the substrates. Bond distances and bond orders are listed for the final ONIOM model of DNMT1 TS (Fig. 5): d₁ and d₂ are the bond distances from the Me-C to its donor and acceptor, and d₃ is the bond distance between the Cys sulfur and C6. The bond orders of d₁, d₂, and d₃ establish that the CH₃ transfer occurs while the C6-S bond is nearly complete (Fig. 1A).

Table 1.

Experimental and theoretical KIEs

Position	Experiment V/K KIEs*	KIEs predicted from simple model analysis†			Theoretical KIEs using ONIOM‡
		TS1	TS2	TS3
5-13C	1.010 ± 0.004	1.007	1.013	0.997	1.009
5-2H	0.991 ± 0.005§	1.048	0.973	4.036	0.955
6-14C	1.038 ± 0.005	1.073	1.029	0.994	1.036
6-3H	0.94 ± 0.01	0.971	0.874	1.066	0.901
Me-14C	1.107 ± 0.006	ND¶	1.106	0.980	1.110
Me-13C	1.07 ± 0.01	ND	1.055	0.989	1.057
Me-3H3	0.980 ± 0.006	ND	0.814	0.993	0.964
5′-14C	0.979 ± 0.008	ND	1.030	ND¶	0.996
5′-3H2	1.073 ± 0.003	ND	1.318	ND	1.066
36S	1.019 ± 0.006	ND	1.016	ND	1.017

ND, not determined; V/K, maximum catalytic rate/Michaelis constant.

^*Intrinsic KIEs measured experimentally, represented by an average ± SD.

^†KIEs predicted with the traditional TS theory, based on the vibrational differences between the GS structures and TS structures of each chemical step (Fig. 4).

^‡KIEs predicted for the methyl transfer TS structure by ONIOM (Gaussian's “our own N-layered integrated molecular orbital and molecular mechanics”) simulations (Fig. 5). No geometry constraints were necessary in the final model to reach the agreement with experimental KIEs.

^§Representative primary data from six individual experiments, with the values 0.98798, 0.98422, 0.99363, 0.99452, 0.99435, and 0.98974, were used to calculate the experimental KIE for 5-²H shown in the table.

^¶These atoms do not participate in this chemical step and will not contribute KIE values.

Methyl Transfer Is the Rate-Limiting Chemical Step.

A quantitative demonstration of the TS for the DNMT1 reaction was obtained by comparing experimental KIEs with computationally predicted KIEs for each chemical TS as the highest barrier on the reaction coordinate, namely, the Cys¹²²⁶ attack, methyl transfer, and β-elimination (TS1, TS2, and TS3 in Figs. 1 and 4). Theoretical KIEs were predicted according to the Bigeleisen equations using ISOEFF (21), based on vibrational frequencies calculated for the ground state (GS) and TS structures optimized in Gaussian 09 (22). β-Elimination (TS3) has been proposed as the rate-limiting step for bacterial M.HhaI methyltransferase by QM/MM and MD simulations (12). For human DNMT1, if TS3 were the highest barrier preceded by reversible steps, the [5-²H]Cyt would show a large normal KIE, predicted to be 4.036, and the Me-C (Me-¹³C and Me-¹⁴C) KIEs would be slightly inverse (TS3 column in Table 1). These predictions contradict experimental KIEs (Fig. 3); therefore, the β-elimination step cannot be rate-limiting. Similarly, if Cys¹²²⁶ attack were rate-limiting, the largest KIE would be observed on [6-¹⁴C]Cyt, whereas Me-C KIEs would be in unity, because those atoms are not involved in the Cys¹²²⁶ attack step (TS1 column in Table 1). The KIEs predicted for the methyl transfer step are consistent with experimental KIEs (TS2 column in Table 1), confirming that methyl transfer is rate-limiting for DNMT1. Small deviations from theory are observed for α- and β-secondary hydrogen KIEs. Secondary hydrogen KIEs often involve contributions from binding isotope effects (BIEs) arising from changes in the bonding environments for the free and enzyme-bound substrates. The vibrational modes altered by these binding environments are not captured in the simple models (Fig. 4), and we used Gaussian's ONIOM method to account for changes in the bonding environments (Fig. 5A).

Fig. 4.

Simple models used to predict KIEs for each chemical step. The models included a methane thiolate to mimic the Cys¹²²⁶ residue, acetic acid to mimic the Glu¹²⁶⁶ residue, and a water molecule as the proton acceptor in the third step. The potential energy (E) surface of each step was scanned by varying the bond distance(s) that drive the chemical reaction: d₃ is the C6-S distance between Cyt C6 and Cys¹²²⁶ S atoms (A); d₁ is the distance between the Me-C and sulfur of SAM, and d₂ is the distance between the Me-C and C5 of Cyt (B); and d₄ is the distance between the water oxygen and the H5 of Cyt (C). The structures indicated by the red circles in the energy plots (Top) were subsequently optimized as first-order saddle points to obtain the TS structures (Bottom) for each chemical step. The KIEs predicted for these theoretical TS structures are listed in Table 1.

Fig. 5.

Structure and electrostatic character of the DNMT1 TS. (A) TS of DNMT1 solved by QM predictions, in agreement with experimental KIEs. (B) Electrostatic potential surface (ESPS) calculated for this TS structure. (C) Close-up view of the calculated ESPSs at the reaction center of the TS, in comparison to reactant SAM and its analog sinefungin (a methyl transferase inhibitor).

TS Structure Reveals Stepwise DNMT1 Catalysis and Symmetrical Nucleophilic Substitution Methyl Transfer.

A subangstrom structure of the DNMT1 TS was obtained by including close-contact protein residues and water molecules from the DNMT1 catalytic site (Fig. 5 and SI Appendix, Fig. S6C). We optimized a TS structure that reproduced experimental KIEs without applying geometry constraints (column of theoretical KIEs using ONIOM in Table 1). This TS structure has a single imaginary frequency corresponding to methyl transfer in the reaction coordinate (Movie S1).

The TS of DNMT1 is defined by 10 experimental and QM-predicted KIEs, and provides excellent agreement for all primary atomic positions (Table 1). Both the GS of DNA-Cyt and the TS of DNMT1 were simulated by a two-layer ONIOM [M062X/6-31+G(d,p):PM6] method. Cyt in free DNA was modeled by double-strand TCG (thymine, cytosine, guanine) base stacking (23, 24) (SI Appendix, Fig. S6A), and DNMT1 TS was simulated by a model, including SAM, deoxycytidine 5′-monophosphate, eight DNMT1 residues, and three water molecules (196 atoms in total). Heavy atoms ([5-¹⁴C]- and [6-¹⁴C]-Cyt and [Me-¹³C]-, [Me-¹⁴C]-, and [³⁶S]-SAM) KIEs are only influenced by covalent bond changes, not by binding interactions, and therefore reliably define the TS geometry. The final TS model describes a near-symmetrical nucleophilic substitution (S_N2) TS for methyl transfer from the sulfur of SAM to C5 of Cyt of DNA. The bond distances from the Me-C to its donor and acceptor are d₁ = 2.29 Å and d₂ = 2.22 Å, with bond orders of 0.52 and 0.34, respectively. It is significant that these bond orders sum to less than 1.0 (0.86), indicating a loose (noncompressed) nucleophilic substitution reaction.

The DNMT1 methyl transfer TS involves a nearly complete covalent bond between the sulfur of Cys¹²²⁶ and C6 with a bond distance of d₃ = 1.93 Å and a bond order of 0.92. This TS structure is consistent with the stepwise mechanism depicted in Fig. 1, where methyl transfer occurs after formation of the covalent bond between the sulfur of Cys¹²²⁶ and C6 and the methyl transfer step is chemically rate-limiting for DNMT1. These features of human DNMT1 are different from the mechanism of M.HhaI DNMT suggested from previous QM/MM simulations (11, 12).

Natural bond orbital analysis (22) of the methyl transfer TS permits construction of an electrostatic potential surface that visualizes the polarity, electronegativity, and bond characteristics of the TS. The TS of DNMT1 shows a distribution of the positive charge in the direction of methyl transfer, originating from the SAM sulfonium ion and extending toward the Cyt ring (Fig. 5B). The electrostatic character of the reaction center at the TS is different from the SAM substrate and sinefungin, a relatively weak binding analog of SAM (Fig. 5C).

Discussion

DNA Cyt methyltransferases have been identified in organisms ranging from bacteria to humans. The bacterial enzymes perform DNA methylation to protect themselves from endogenous restriction enzymes. In mammalian cells, DNMTs have extensive N-terminal regulatory domains in addition to the catalytic domain, and the genomic methylation patterns are generally associated with gene regulation. Despite their differences in overall size and biological roles, the bacterial enzymes (exemplified by M.HhaI methyltransferase) and mammalian DNMT1 share a conserved set of active site residues in the covalent complex (4). They also catalyze similar steps (Fig. 1A) for methylating Cyt, including Cyt base-flipping, Cys attack on C6, methyl transfer, and β-elimination. Although previous computational studies predicted the β-elimination to be rate-limiting for M.HhaI (12), our experimental KIEs, together with QM analysis on DNMT1, demonstrate a mechanism where methyl transfer has the largest energy barrier among the chemical steps. Whether the difference is due to the inherent natures of the bacterial enzyme and DNMT1 remains an open question.

The TS formed by DNMT1, resolved here to subangstrom resolution, is the most complex enzyme system yet analyzed by a combination of KIEs and quantum chemistry. In this TS structure, a nearly full bond is formed between Cys¹²²⁶ and Cyt C6 (bond order = 0.92), whereas the sum of two bond orders on the methyl transfer path (d₁ and d₂) is less than 1 (0.86). Whether the Cys attack occurs before or simultaneously with methyl transfer during DNMT catalysis has been debated. Our results suggest a stepwise mechanism for DNMT1, where the Cys attack is not fully synchronized with methyl transfer. This evidence is also the first combined experimental and computational evidence, to our knowledge, that addresses concertedness for a DNMT.

Compression of the methyl donor and acceptor at the TS has been proposed for catechol-O-methyltransferases (COMTs) (25, 26), but has been disputed in theoretical studies (27, 28). In DNMT1 TS, the bond orders on the methyl transfer path (d₁ and d₂) sum to less than 1, indicating a loose S_N2 substitution for the methyl transfer, in contrast to a compression-type TS claimed on COMT (25, 26). This discrepancy is not entirely surprising, given the major differences in the methyl acceptors and the catalytic mechanisms involved. Comparative results from COMT and DNMT1 suggest that diverse TS mechanisms exist among the SAM-dependent methyltransferases.

Inhibitors of human DNMT have been used in cancer therapy (29), because elevated CpG methylation in tumor repressors can result in carcinogenesis. However, all US Food and Drug Administration-approved DNMT inhibitors are cytotoxic and mutagenic. Those inhibitors (i.e., 5-aza-cytidine, 5-aza-deoxycytidine) promote DNA demethylation by incorporation into host DNA to form covalent adducts between DNMTs and DNA. Other DNMT inhibitors have been identified from chemical library screening, but they often lack specificity against DNMTs (30). The subangstrom geometry and electrostatics details of the DNMT1 TS described here will assist the design of mechanism-based TS analog inhibitors for DNMT1. Solving the TS of DNMT1 provides proof of concept to gain mechanistic and chemical insights into complex enzyme reactions involved in epigenetic control (31, 32) and other macromolecular modifications.

Materials and Methods

Supplementary materials and methods are available in SI Appendix. These materials and methods include the synthesis and purification of all isotope-labeled small-molecule and macromolecule substrates, including six labeled as dCTP, their incorporation into DNA, and eight different SAM labels; the measurement of C_f values for DNA and SAM; and further computational details.

Human DNMT1.

Full-length human DNMT1 protein (amino acids 1–1,616) containing an N-terminal His6-tag was expressed in a Sf9 insect cell line as described previously (33). The expressed human DNMT1 was purified by nickel-nitrilotriacetic acid column chromatography on FPLC. The concentrated DNMT1 was further purified through a gel filtration column (200 pg, Hiload 16/600 Superdex; Amersham) in 50 mM Hepes (pH 7.4), 150 mM NaCl, and 1 mM DTT to remove low-molecular-weight contaminants. Purified DNMT1 appeared as a single band on SDS/PAGE with an apparent size of 190 kDa. Aliquots were frozen and stored at −80 °C.

Forward Commitment Values.

The C_f values were measured by substrate trapping procedures. The formation of radiolabeled product was quantitated after chasing enzyme-bound labeled substrate with a large excess of unlabeled substrate. The procedures involved HPLC separation and liquid scintillation counting (LSC) of the [5′-³H]-labeled 5m-dC or the [1′-³H]-labeled S-adenosylhomocysteine product in respective experiments (details are available in SI Appendix, Supplementary Materials and Methods and Table S3).

Measurement of DNA Cyd KIEs.

KIEs were measured by internal competition using a mixture of isotope-labeled (heavy) and remote-labeled (light) DNA substrates. ¹⁴C- and ³H-labeled hemimethylated 26-bp DNA and 1.0 mM nonlabeled SAM were incubated in a buffer of 20 mM Tris⋅HCl (pH 7.4), 100 mM KCl, and 1 mM DTT at 37 °C. The methylation reaction was initiated by the addition of human DNMT1. Reactions were quenched at different reaction intervals by placing tubes in a 95 °C heat block for 10 min, followed by cooling on ice. The quenched reactions (100 μL) were treated with 20 μL of 10 mM Tris⋅HCl (pH 7.9), 50 mM NaCl, 10 mM MgCl₂, 1 mM DTT, 10 units of exonuclease III, 0.1 unit of snake venom phosphodiesterase I, and 0.5 unit of alkaline phosphatase. The DNA digestion (37 °C overnight) converted all DNA to mononucleosides (Fig. 2D).

The nucleoside mixture from each reaction aliquot was separated by a C18 column on an HPLC system equipped with a photodiode array detector (SI Appendix, Fig. S5). Cold carriers of dC and 5m-dC were added to samples to facilitate detection and collection. The dC and 5m-dC were collected in glass scintillation vials and dried on a vacuum concentrator. The samples were dissolved in water and mixed thoroughly with 10 mL of scintillation fluid. The ¹⁴C and ³H radiation levels in both dC and 5m-dC from each reaction were measured by LSC (five or more cycles). ¹⁴C-labeled dC nucleoside was used to standardize the scintillation channel energy crossover with ³H in the low-energy range of 0–25 kiloelectron volts. Specific counts of the ¹⁴C and ³H isotopes were calculated by using the observed counts in both channels and the crossover ratio.

During each KIE experiment, control reactions that convert 100% of the labeled species to product were also conducted. Complete reactions (f = 1) were achieved by extended incubation periods (24 h) with additional enzyme. These reactions were used to confirm the purity of radiolabels in the substrate (f = 0) and to detect any nonreactive labels.

Measurement of SAM KIEs.

KIEs were measured by internal competition using a mixture of isotope-labeled (heavy) and remote-labeled (light) SAM substrates. ¹⁴C- and ³H-labeled SAM and nonlabeled 26-bp hemimethylated DNA substrate were incubated in 100-μL solutions with 20 mM Tris⋅HCl (pH 7.4), 100 mM KCl, and 1 mM DTT at 37 °C. Methylation reactions were initiated by the addition of human DNMT1. Enzyme-free control reactions (n ≥ 5) were incubated simultaneously to the enzyme reaction samples. Control samples quantitated the slow chemical degradation of SAM during the reaction and sample handling. Reactions were quenched by adding 20 μL of 0.5 mM cold SAM and 50 mM H₂SO₄, and were stored immediately at −80 °C until HPLC purification.

Remaining unreacted SAM substrate from each reaction and control sample was purified by HPLC in ammonium formate buffer as described in SI Appendix. The ds-DNA was trapped on the C18 guard column and was not observed as a defined peak. The ¹⁴C and ³H counts in the purified SAM were measured by LSC. Experimental KIEs were calculated from the isotope ratios present in the unreacted SAM substrate.

Computational Methods.

We followed the general procedures previously established (16) to perform QM computational TS analysis of DNMT1. All of the geometry optimizations and frequency calculations were performed with Gaussian 09. In our final QM models, we simulated the GS of SAM using a density functional theory [with M06-2X functional (34) and 6-31+G(d,p) basis set; SI Appendix, Fig. S6B], and we simulated the GS of DNA-Cyt (SI Appendix, Fig. S6A) and the TS of the methyl transfer using a two-layer ONIOM method (35, 36) [M062X/6-31+G(d,p):PM6; SI Appendix, Fig. S6C] as implemented in Gaussian 09. The calculated vibrational frequencies were scaled by 0.967 to reproduce the true vibrational zero-point energies (37) for calculations of theoretical KIEs in the ISOEFF program (21) at the experimental temperature (37 °C). More details of the computational methods are available in SI Appendix.