ABSTRACT
4"-Thio-2"-deoxycytidine was synthesized as a 5"- protected phosphoramidite compatible with solid phase DNA synthesis. When incorporated as the target cytosine (C*) in the GC*GC recognition sequence for the DNA methyltransferase M.HhaI, methyl transfer was strongly inhibited. In contrast, these same oligonucleotides were normal substrates for the cognate restriction endonuclease R.HhaI and its isoschizomer R.HinP1I. M.HhaI was able to bind both 4"-thio-modified DNA and unmodified DNA to equivalent extents under equilibrium conditions. However, the presence of 4"-thio-2"-deoxycytidine decreased the half-life of the complex by >10-fold. The crystal structure of a ternary complex of M.HhaI, AdoMet and DNA containing 4"-thio-2"-deoxycytidine was solved at 2.05 Å resolution with a crystallographic R-factor of 0.186 and R-free of 0.231. The structure is not grossly different from previously solved ternary complexes containing M.HhaI, DNA and AdoHcy. The difference electron density suggests partial methylation at C5 of the flipped target 4"-thio-2"-deoxycytidine. The inhibitory effect of the 4" sulfur atom on enzymatic activity may be traced to perturbation of a step in the methylation reaction after DNA binding but prior to methyl transfer. This inhibitory effect can be partially overcome after a considerably long time in the crystal environment where the packing prevents complex dissociation and the target is accurately positioned within the active site.
The DNA cytosine-5 methyltransferases (C5-Mtases) are model systems for enzymes that integrate catalytic activity with molecular recognition of nucleic acid substrates and have been the focus of a renewed interest in DNA methylation (reviewed in 1 ,2 ). These enzymes are found in both prokaryotes and eukaryotes. In bacteria, C5-Mtases function as a component of a simple immune system in which cytosine modification is used to distinguish `self' from foreign DNA. Host sequences that are methylated are resistant to digestion by the host's restriction endonucleases, whereas unmodified infecting viral DNA is left susceptible to degradation (see 3 for a recent review). In eukaryotes, the role of cytosine methylation is more complex and less well understood. DNA methylation has been implicated in a variety of processes including genomic imprinting (4 ), embryonic development (5 ) and gene expression (reviewed in 6 ). 5-Methylcytosine and C5-Mtases themselves appear to be endogenous mutagens and may play a role in disease (7 ,8 ).
The bacterial enzyme M.HhaI, which recognizes the sequence GCGC and methylates the inner cytosine, is one of the most thoroughly examined C5-Mtases. The gene for this protein has been overexpressed in Escherichia coli to allow large-scale production and purification of the protein and subsequent crystallization (9 ). The structure of this enzyme has been solved in the absence of DNA (10 ) as well as in complex with both modified and unmodified DNA substrates (11 -13 ). The latter DNA-protein co-crystal structures revealed that binding is accompanied by a novel distortion of the DNA at the binding site: complete extrusion of the base to be methylated out of the DNA helix. This process, termed base flipping, has subsequently been confirmed in the structure of another C5-Mtase, M.HaeIII (14 ) as well as two DNA repair enzymes, T4 endonuclease V (15 ) and uracil-DNA glycosylase (16 ). Base-flipping may be a fundamental mode of DNA-protein interaction as several other enzymes are likely to use this mechanism (17 ,18 ).
An understanding of the catalytic mechanism of C5 methylation used by M.HhaI has emerged from the structures cited above as well as a variety of biochemical studies which have identified residues required for catalysis (19 ,20 ). Methylation by M.HhaI and other C5-Mtases requires the target base to be inserted into the catalytic pocket of the enzyme where the reaction proceeds via the formation of a transient covalent bond between the sulfhydryl group of a conserved cysteine (Cys81 in M.HhaI) and C6 of the target cytosine (Fig. 1 ). The initial covalent bond enhances the reactivity of C5 allowing nucleophilic attack by C5 on the methyl group of the methyl-donating cofactor AdoMet. The high energy carbanion generated at C5 is either resonance stabilized (19 ,21 ) or entirely avoided by protonation of N3 by the nearby protonated glutamic acid residue (Glu119 in M.HhaI) (22 ). Both alternatives allow for the formation of a reactive enamine. AdoMet is converted to AdoHcy by the methyl transfer and elimination of the C5 proton by a water molecule allows dissolution of the covalent complex (12 ,13 ).
N4-Benzoyl-3",5"-di-O-benzyl-4"-thio-2"-deoxycytidine and the [alpha]-anomer. To a stirred suspension of N4-benzoylcytosine (2.50 g, 11.63 mmol) in acetonitrile (50 ml), under a nitrogen atmosphere, was added bis-(trimethylsilyl)acetamide (4.91 g, 6.0 ml, 23.26 mmol). The reaction mixture was stirred at room temperature for 1 h, during which time a clear solution was formed. Additional acetonitrile (100 ml) was added, followed by the addition of crushed 4 Å molecular sieves (3.5 g) and a solution of benzyl 3,5-di-O-benzyl-2-deoxy-1,4-dithio
[beta]-anomer: Rf [0.36 EtOAc/n-hexane (3:1)]. [delta] (1H-NMR; 300 MHz, CDCl3): 8.85 (1H, br.s, 4-N
[alpha]-anomer: Rf [0.28 EtOAc/n-hexane (3:1)]. [delta] (1H-NMR; 300 MHz, CDCl3): 8.78 (1H, br.s, 4-N
N4-Benzoyl-4"-thio-2"-deoxycytidine. To a stirred solution of boron trichloride in dichloromethane (1.0 mol/dm3) (30 ml, 30 mmol), at -85oC under a nitrogen atmosphere, was added dropwise a solution of N4-benzoyl-3",5"-di-O-benzyl-4"-thio-2"-deoxycytidine (2.00 g, 3.80 mmol) in dichloromethane (75 ml), after which the reaction mixture was left stirring for 4 h. The reaction was then quenched, at -85oC, by the addition of methanol/dichloromethane (1:1) (50 ml). The reaction mixture was then immediately neutralized by the addition of methanolic ammonia (2.0 M) (~40 ml). Then the reaction mixture was allowed to warm to room temperature and was filtered to remove the ammonium chloride formed during neutralization. The solvent was then removed in vacuo to yield a white solid. The crude product was purified by flash column chromatography, using an eluent of dichloromethane/ methanol (9:1). Combination of the appropriate fractions yielded the product (0.78 g, 59%) as a white solid. Rf [0.40 DCM/MeOH (9:1)]. [delta] (1H-NMR; 300 MHz, d6-DMSO): 11.28 (1H, br.s, 4-N N4-Benzoyl-5"-O-(4,4"-dimethoxytriphenylmethyl)-4"-thio-2"- deoxycytidine. To a stirred solution of N4-benzoyl-4"-thio-2"- deoxycytidine (508 mg, 1.68 mmol) in pyridine (20 ml), under an argon atmosphere at room temperature, was added triethylamine (0.31 g, 0.42 ml, 3.03 mmol) 4-dimethylaminopyridine (10 mg, 0.08 mmol) and 4,4"-dimethoxytriphenylmethyl chloride (0.97 g, 2.86 mmol). The reaction mixture was left stirring at room temperature for 42 h and was then quenched by the addition of methanol (5 ml) and stirring was continued for a further 10 min. The solvent was removed in vacuo and the crude product dissolved in dichloromethane (100 ml). The organic layer was washed with a saturated aqueous solution of sodium hydrogen carbonate (2 * 100 ml) and a saturated aqueous solution of sodium chloride (1 * 100 ml). The organic layer was dried (MgSO4), filtered and the solvent removed in vacuo, followed by co-evaporation with toluene and then methanol, to yield the crude product as a yellow foam. The product was purified by silica gel column chromatography, using an initial eluent of dichloromethane/ methanol (99:1) (+0.5% TEA), followed by an eluent of dichloromethane/methanol (98:2) (+0.5% TEA). Combination of the appropriate fractions yielded the product (0.84 g, 89%) as a cream foam. Rf [0.46 DCM/MeOH (97:3) +0.5% TEA]. [delta] (1H-NMR; 300 MHz, CDCl3): 8.37 (1H, d, 3J = 7.6 Hz, H-6), 7.89 (2H, m, o-benzoyl), 7.65-7.24 (14H, m; 9H, DMTr aromatic; 3H, benzoyl aromatic; 1H, H-5; 1H, N N4-Benzoyl-5"-O-(4,4"-dimethoxytriphenylmethyl)-3"-O-[2-cyanoethyl(diisopropylamino)
The following oligonucleotides were used in the experiments (HhaI target sequence in bold; M, 5-methyl-2"-deoxycytidine; S, 4"-thio-2"-deoxycytidine; F, 5-fluoro-2"-deoxycytidine). Oligonucleotides A-E were 37mers with the sequence:
GACTGGTACAGTATCAGGXGCTGACCCACAACATCCG
TGACCATGTCATAGTCCGYGACTGGGTGTTGTAGGCT
where X and Y varied as follows: A, X = S, Y = M; B, X = C, Y = M; D, X = C, Y = C; and E, X = S, Y = C; Oligo C contained the unmethylated sequence GGCC in place of GCGC.
The 5"-protected phosphoramidite of 4"-thio-2"-deoxycytidine was synthesized as described above and incorporated into the desired oligonucleotides by standard solid phase synthesis methodology [Roger Knott, New England Biolabs (NEB)] and purified using HPLC and polyacrylamide gel electrophoresis. Individual strands were annealed in 10 mM Tris-HCl (pH 8.0), 1 mM Na2EDTA (TE) to generate double-stranded DNA. Annealed oligonucleotides were 5" end-labeled using T4 polynucleotide kinase (NEB) and [[gamma]-32P]ATP [New England Nuclear (NEN)] using the protocol suggested by the manufacturer. Unincorporated nucleotide was removed by spun column chromatography using Quick-Spin G-25 columns (Boehringer Mannheim) or MicroSpin S-200 columns (Pharmacia).
M.HhaI was purified as described previously (9 ) and exhaustively dialyzed to remove endogenous bound S-adenosyl-l-methionine (AdoMet).
DNA binding was studied by gel retardation, essentially as described previously (27 ). The target oligonucleotide concentration ranged from ~5 to 200 nM. The cofactor S-adenosylhomocysteine (AdoHcy) was included in some reactions at a concentration of 1 mM. After autoradiography the portions of each gel lane corresponding to bound and free oligonucleotide were excised and radioactivity measured by scintillation counting in Opti-fluor (Packard). Binding constants were determined by Scatchard analysis (28 ).
Off-rates for DNA binding were measured by pre-forming complexes in the presence of 1 mM AdoHcy, 6.4 nM normal or 4.8 nM thio-substituted labeled oligonucleotide and 6.4 nM M.HhaI. The association reactions were quenched by adding excess unlabelled oligo at 1 [mu]M final concentration. Dissociation of the labeled oligo was allowed to occur for various time intervals. Samples were then loaded onto a running 6% polyacrylamide gel and processed in the same manner as the equilibrium binding experiments. Off-rates were calculated by monitoring the disappearance of the bound labeled DNA.
Methylation reactions were performed in [20 mM Tris-acetate (pH 7.9) 10 mM magnesium acetate, 50 mM potasium acetate, 1 mM DTT, 200 [mu]g/ml BSA] at 37oC in the presence of S-[methyl-3H]adenosyl-l-methionine (NEN) with a specific activity adjusted to either 9.7 or 15 Ci/mmol. The concentrations of enzyme, DNA and cofactor used are listed in the appropriate figure legends. All components except the DNA were mixed and pre-warmed at 37oC for 3 min, at which point the DNA was added to initiate the reaction. Reactions were stopped by the addition of excess S-adenosyl-l-homocysteine (Sigma) to a final concentration between 1 and 10 mM and/or flash freezing in a dry ice-EtOH bath. Methyl transfer to DNA was measured by trapping DNA on Whatman DE-81 filters as previously described (29 ). Bound radioactivity was determined by scintillation counting in Opti-fluor (Packard).
Oligonucleotides at a concentration of 1 [mu]M were digested with 14 U restriction endonuclease in 20 [mu]l reactions using the buffer suggested by the enzyme supplier (NEB). Digestion products were separated by electrophoresis on 4% Metaphor agarose (FMC) gels in 1* TBE buffer and visualized by ethidium staining and UV illumination.
The structure of M.HhaI in a ternary complex with AdoMet and a short DNA duplex (referred to as the S13 structure) was determined by X-ray crystallography. Two 13mer oligonucleotides (NEB) were annealed to form a 12 bp asymmetric duplex with 5" single base overhangs:
5'-TCCATGMGCTGAC -3'
3'- GGTACGSGACTGT-5'
where M = 5-methyl-2"-deoxycytidine and S = 4"-thio-2"- deoxycytidine. Crystals of the S13 complex were grown by mixing the M.HhaI-AdoMet-DNA complex, as described previously (12 ,13 ), with an equal volume of 1.4-1.8 M ammonium sulfate, 50 mM citrate at pH 5.6, and equilibrating the mixture against 1 ml of the latter solution at 16oC. The S13 crystals formed within days in the same trigonal space group as in the previously reported ternary structures (11 -13 ). One S13 complex was present per asymmetric unit. Three weeks after formation of the crystals, the X-ray diffraction intensities were measured at a wavelength of 1.15 Å on a Mar Research imaging-plate diffractometer at beamline X12C of the National Synchrotron Light Source, Brookhaven National Laboratory (67.5o total rotation with 1.5o increment and 90 s exposure at ~16oC). Data acquisition, similar to that described previously (12 ,13 ), yielded 32 880 unique reflections. R-merge = [Sigma]J<I> - IJ/[Sigma]<I> = 0.078 and <I/[sigma]> = 12.1, where I is the measured intensity of each reflection, <I> is the intensity averaged from multiple observations of symmetry-related reflections, and [sigma] is the root mean square deviation from the mean.
The structure was solved by the difference Fourier method using the previously solved HM13 ternary structure (M.HhaI with AdoHcy and a hemi-methylated DNA identical in sequence to the one used here but containing unmodified cytosine at the target) as the initial model (13 ). The model was optimized by rigid-body refinement from 20 to 3 Å and was further refined against diffraction data measured from 5.0 to 2.05 Å using X-PLOR (30 ). After 11 rounds of least square refinement, manual model building, and placement of well-ordered solvent molecules (interpreted as water) by examination of difference electron density, the crystallographic R-factor and R-free ([Sigma]JFo - FcJ/[Sigma]JFoJ, where Fo and Fc are the observed and calculated structure factor amplitudes, respectively) were reduced to 0.186 and 0.231 for ~90 and 10% of the data, respectively. We used the difference Fourier synthesis (Fo - Fc, [alpha]c) to confirm and locate the positions of the sulfur atom of 4"-thio-2"-deoxycytidine and the transferable methyl group of AdoMet.
The key step in the synthesis of the 5"-protected phosphoramidite of N4-benzoyl-4"-thio-2"-deoxycytidine requires either the condensation of a cytosine base with a suitably protected 4-thiosugar compound, or the conversion of the uridine analogue to the corresponding cytidine analogue. Secrist and co-workers have previously reported the synthesis of 4"-thio-2"-deoxycytidine (31 ). Initially this group attempted to prepare 4"-thio-2"-deoxycytidine via direct condensation of cytosine with a suitably protected 4-thiosugar. However, it was not possible to separate the [alpha]/[beta]-anomers at either the protected or deprotected stage and therefore 4"-thio-2"-deoxycytidine was prepared via a uridine to cytidine conversion on the [beta]-anomer of 4"-thio-2"-deoxyuridine. Here we report an improved procedure for the preparation of 4"-thio-2"-deoxycytidine, using N4-benzoylcytosine in a direct condensation approach.
N4-Benzoylcytosine was synthesized, using the procedure of Brown et al. (32 ), in a yield of 93%. This base was then condensed with benzyl 3,5-di-O-benzyl-2-deoxy-1,4-dithio-d-erythro-pentofuranoside, synthesized by the method of Dyson et al. (33 ), to give an anomeric mixture ([alpha]/[beta] ratio 3:2) of the nucleoside derivative N4-benzoyl-3",5"-di-O-benzyl-4"-thio-2"-deoxycytidine. Separation of the anomers either by crystallization of the [beta]-anomer or by silica gel column chromatography proved to be facile, due to the presence of the N4-benzoyl protecting group. Removal of the 3",5"-di-O-benzyl protecting groups was achieved using boron trichloride at low temperature, although the yield was lower than expected due to partial debenzoylation of the N4-benzoyl protecting group. Dimethoxytritylation of the 5"-hydroxyl group followed by phosphitylation of the 3"-hydroxyl group gave the desired phosphoramidite. The overall yield of the phosphoramidite from the 4-thiosugar was 13%; the poor [alpha]/[beta] ratio (3:2) at the condensation step results in a maximum theoretical yield of only 40%.
The 5"-protected phosphoramidite of N4-benzoyl-4"-thio-2"- deoxycytidine proved to be completely compatible with standard solid phase synthesis of oligonucleotides via the phosphoramidite approach.
4"-Thiothymidine is known to inhibit methylation by M.EcoRV when present adjacent to the target adenine that becomes methylated (24 ). We examined the effects of 4"-thio-2"-deoxycytidine as a target nucleotide on catalytic function in a DNA cytosine methyltransferase, M.HhaI. M.HhaI methylates the inner cytosine of the sequence GCGC to yield G5mCGC. We constructed a 37 base double-stranded DNA oligonucleotide containing a centrally located HhaI site (GCGC) in which the inner cytidine was replaced by the thio-modified residue on one strand and by 5-methyl-2"-deoxycytidine in the other strand. Incubation of M.HhaI with the hemi-methylated 37mer 4-thio oligonucleotide resulted in very poor transfer of 3H-methyl groups from the cofactor AdoMet relative to a hemi-methylated 37mer substrate identical in sequence but containing normal cytosine (Fig. 3 ). The level of incorporation of 3H-methyl was comparable to that observed with a non-methylated 37mer containing the non-cognate target sequence GGCC. The turnover rate constant Kcat for methyl transfer was >100-fold lower in the presence of the thio-substituted hemi-methylated substrate (0.020 min-1) relative to unmodified hemi-methylated DNA (2.3 min-1). Kinetic analysis of the inhibition suggests that DNA target sites containing 4"-thio-2"-deoxycytidine interact competitively with respect to unmodified target sites (apparent Ki ~2-5 nM) (Fig. 4 ). This value is comparable to the Km of 2.3 nM reported for a poly(dGdC) substrate (19 ).
One obvious explanation for the inhibition of methylation by 4"-thio-2"-deoxycytidine was the potential disruption of the binding of the methyltransferase to the target sequence. To test this, equilibrium binding assays were performed on the same hemi- methylated oligonucleotides that were refractory to methylation. As shown in Figure 5 , the 4"-thio modified oligonucleotide bound as well as the unmodified oligonucleotide. Equilibrium dissociation constants for unmodified and 4"-thio-2"-deoxycytidine-containing oligonucleotides were determined by Scatchard analysis to be 18 and 21 nM, respectively. Addition of AdoHcy to the binding reaction increased binding of both oligonucleotides as expected from previous work (13 ,22 ,27 ,34 -36 ) (data not shown).
While the equilibrium binding of the 4"-thio modified oligonucleotide was essentially identical to the unmodified control DNA, the off-rates for the two species were considerably different, as shown in Figure 6 . Incorporation of 4"-thio-2"-deoxycytidine into the target sequence substantially increased the off-rate of the DNA relative to unmodified DNA. Half-lives of the 4"-thio and unmodified DNA complexes under the binding conditions used were 60 and 760 min, respectively. In combination with the equilibrium binding studies, this result indicates that the stability of the DNA-methyltransferase complex is substantially altered by the presence of the 4"-thio modification.
Oligonucleotides containing 4"-thiothymidine were previously shown to be resistant to cleavage by the endonuclease EcoRV when the thio-modification was incorporated within the recognition sequence of the enzyme (24 ). The conformation of EcoRV does not change significantly when complexed with thio-modified DNA as opposed to normal DNA (Kostrewa,D., Hancox,E.L., Walker, R.T., Connolly,B.A. and Winkler,F.K., unpublished results). However, the Mg2+ ion required for catalysis was not present in crystals of EcoRV bound to a thio-modified substrate, despite being crystallized out of Mg2+-containing solution. In the current study, we constructed non-methylated versions of the oligonucleotides used in the methylation assays and subjected them to digestion by the cognate restriction endonuclease R.HhaI and its isoschizomer R.HinP1I. After incubation with R.HhaI which cleaves 5" to the outer cytosine (GCG/C), complete digestion was observed (Fig. 7 ). To determine if the proximity of the thio- modified cytidine to the actual site of cleavage was a factor in the digestion, the reaction was repeated using R.HinP1I which cleaves 5" to the inner cytosine (G/CGC), immediately adjacent to the modified residue. As seen in Figure 7 , the R.HinP1I digestion also proceeded to completion. The same results were obtained when the inner 2"-deoxycytidine in both strands of the DNA was replaced with the thio-modified analog (data not shown). Unlike 4"-thiothymidine, incorporation of 4"-thio-2"-deoxycytidine did not reduce susceptibility to cleavage by restriction endonucleases. Whether this difference is due to structural variations induced by the thionucleoside or differences in the details of the cleavage mechanisms remains to be determined. R.EcoRV leaves blunt ends upon cutting, whereas R.HhaI and R.HinP1I leave 5" and 3" tails, respectively.
We have now determined by X-ray crystallography the structure of M.HhaI in a ternary complex with AdoMet and a 13mer non-palindromic DNA duplex containing 4"-thio-2"-deoxycytidine as the target nucleotide on one strand, and 5-methyl-2"- deoxycytidine on the complementary strand. The existence of the sulfur atom of 4"-thio-2"-deoxycytidine and the methyl group of the 5-methyl-2"-deoxycytidine on the complementary strand were confirmed by (Fo - Fc, [alpha]c) electron density maps in which the sulfur atom and methyl group were omitted from the structure factor calculation, respectively (Fig. 8 ). The peak height for the sulfur atom was comparable to that for the sulfur atom of Cys-81 and for the sulfur atom of AdoMet. The target 4"-thio-2"- deoxycytidine was flipped out of the DNA helix, while the 5-methyl-2"-deoxycytidine on the complementary strand remained stacked within the DNA helix.
Oligonucleotides bearing 4"-thio modifications display biological properties that differ from their unmodified counterparts, even though their structure appears to be only subtly altered. The structure of DNA containing 4"-thiothymidine has previously been characterized by X-ray crystallography (23 ) and that of free 4"-thiothymidine by NMR and X-ray crystallography (39 ). In these studies, the conformational changes induced by the thio substitution have been minor, short-range, and limited to the sugar-phosphate backbone in the immediate vicinity of the modified nucleotide. However, incorporation of 4"-thiothymidine adjacent to the target adenine was sufficient to interfere with methyl transfer by M.EcoRV (24 ). In the present work, 4"-thio-2"-deoxycytidine was used to replace the target 2"-deoxycytidine for methylation by M.HhaI. As with M.EcoRV, in solution the methylation reaction was strongly inhibited. This result is somewhat surprising since the presence of sulfur at the 4" position of the sugar is not obviously incompatible chemically with the known mechanism for C5-cytosine methylation. The strong inhibition observed in the solution studies can be overcome over a long period of time in the crystal environment, as evidenced by the substantial, albeit partial, methylation of the cytosine C5 position in the crystal structure of the ternary complex of M.HhaI, 4"-thio-substituted DNA, and AdoMet. These results attest to a greatly-slowed reaction, even in the crystal environment where packing forces stabilize the complex and position the substrate in an optimal orientation for the reaction to proceed.
Why the reaction is slowed is difficult to discern, even from the high resolution crystal structure. It is formally possible that in the absence of crystal packing forces the minor distortions evident in the structure might be amplified sufficiently to dislocate the target cytosine such that it is no longer optimally oriented, leading to the inhibition of methylation seen in solution. However, given the overall similarity in the S13 and HM13 structures, we believe that analysis of the current structure is relevant to understanding this inhibition (especially that seen in the crystal environment). In principle, there are three points at which the methylation reaction could be inhibited: recognition of the target sequence and flipping out the target base, formation of the initial covalent bond at C6, and methyl transfer to C5. In DNA binding studies, the thio modification did not significantly reduce the level of binding as measured under equilibrium conditions, but did significantly increase the off-rate of the bound oligonucleotide, suggesting that the complex forms but fails to become fully stabilized. Although 4"-thio substitution might introduce backbone distortions that raise the thermodynamic barrier to base-flipping, no substantial structural deformations that would be incompatible with base extrusion were observed in the S13 crystal structure. At present, we are unable to assess definitively whether the rate of base flipping is reduced by 4"-thio-2"-deoxycytidine in solution as we lack a reliable assay for this process in the absence of methylation. However, given the equilibrium binding and structural data, as well as the observation that methylation is slowed in the crystal where flipping has definitely taken place, it is unlikely that base flipping is the key step affected. The observed differences in complex stability, as measured by the off-rate of DNA from the complex, occurred in the presence of AdoHcy and not AdoMet, which suggests that the primary effects of 4"-thio substitution occur prior to methyl transfer. An indication that the initial nucleophilic attack may be unfavorable comes from the inhibition seen in solution and the van der Waals length contacts observed between Cys81 and C6 in the crystal. The instability of the DNA-protein complex may also be related to the inability to form the initial transient covalent bond. For these reasons, the step most likely inhibited in solution by the introduction of a sulfur atom at the 4"-deoxyribose position is the formation of the covalent bond between the catalytic cysteine and cytosine C6.
Given the close similarity between the S13 structure and previous M.HhaI structures and the lack of a culpable change in conformation around the reactive C5 or C6 atoms, we propose that the observed inhibition is the result of perturbations in the electronic environment of the cytosine ring. The lower electronegativity of sulfur relative to oxygen (2.5 versus 3.5) and the separation of the 4" atom from the ring atoms lowers the likelihood of direct inductive or resonance effects by sulfur on the electron distribution within the pyrimidine ring. However, it is worth noting that the related compounds 5-fluoro-4"-thio-2"-deoxycytosine and 5-fluoro-4"- thio-2"-deoxyuridine have absorption profiles that differ from the analogous compounds bearing a 4"-oxygen (R.T.W., unpublished results), so some influence may occur. The 4"-sulfur atom may indirectly affect the electronic environment of the pyrimidine ring by perturbing the charge interactions of other groups with the ring atoms. By virtue of its larger van der Waals radius (1.85 versus 1.40 Å for O) sulfur appears to `bump' the nearby Arg165 residue and thereby increase the separation between the terminal amino groups of Arg165 and the nearby charged phosphate group which normally participate in a charge-charge interaction. The separation between the charged functional groups of Arg165 and the O2 carbonyl of cytosine is also slightly increased. It is the latter ring attached atom that may be the focus of the electronic disturbance.
The influence of the pyrimidine ring's delocalized electronic structure on the rate of methylation has been noted previously. Baker et al. (40 ) have suggested that cytosines which are protonated at N3 should be preferentially methylated because protonation substantially increases the charge on C6 (40 ,41 ), predisposing the ring to nucleophilic attack at that position (42 ). Liu and Santi (43 ) have shown that protonation of N3 is required to initiate catalysis of dCMP methylation by the N229D mutant of thymidylate synthase which has a catalytic mechanism similar to that of C5-Mtases. Since the [pi] molecular orbital system of cytosine encompasses the O2 carbonyl, the charge on C6 will be affected by the delocalized charge distribution on O2 as well as N3. Gabbara et al. (21 ) have suggested that delocalized charge may be stabilized on the O2 carbonyl by interaction with the nearby arginine (Arg165) as an alternative to protonation at N3. Their model presupposes that nucleophilic attack has already taken place; however, the influence of Arg165 on the O2 carbonyl that they propose is still of significance. Changes in the electronic environment around the O2 carbonyl may also destabilize the charge distribution on this atom, feeding excess electrons back into the ring. Decreasing the charge on C6 would reduce the propensity for nucleophilic attack on this atom. Alternatively, excessive polarization or protonation of the O2 oxygen may sequester delocalized electrons and inhibit later stages of the reaction. Changes in the electron distribution around the ring may also perturb the pKa of the N3 atom and alter its protonation. The precise nature of the electronic effects occurring in the S13 structure are too complicated to predict accurately based on the current structure, but the importance of the O2 carbonyl in influencing the rate of the methylation reaction is worth noting.
In conclusion, our work suggests that 4"-thio-2"-deoxycytidine may be an excellent compound for inhibiting DNA cytosine methylation. This result may help to explain why incorporation of 4"-thio-2"-deoxynucleotides into DNA results in toxicity to the organism. Thus, 4"-thiothymidine which is a substrate for cellular and viral kinases is toxic to both cells and herpes virus (44 ), whereas 4"-thio-5-ethyl-2"-deoxyuridine which is only a substrate for herpes virus kinase only shows overt toxicity to the virus (45 ). It is probable that the inhibition of methylation caused by the presence of 4"-thio-2"-deoxycytidine in the DNA, seen here, is only one example of many necessary metabolic steps which require precise nucleic acid-protein recognition and which are disrupted by the modification. In the case of cytosine methylation by M.HhaI in solution, the modification probably acts by disrupting the electronic configuration of cytosine required for methyl transfer. The thio-modification does not appear to distort the structure of the DNA grossly nor hinder its ability to be recognized by DNA-binding enzymes. The work also suggests that the currently accepted reaction pathway may require modification in order to include the influence of Arg165 on the reactivity of the cytosine base and a more significant role for the O2 carbonyl in determining the reaction rate. Clearly, the reaction mechanism of cytosine methylation is more sensitive to subtle changes in its environment than once believed and merits further study.
We thank K. McCloy for protein purification and initial crystallization studies of the DNA-protein complex. We are grateful to R. Knott, J. Buswell and the organic synthesis division of NEB for synthesis and purification of oligonucleotides containing 4-thio and 5-fluoro modified nucleosides. We also thank M. O'Gara, A. Mackerel, D. Landry, C. Noren and A. Kumar for helpful discussions and critical comments on the manuscript. The authors wish to thank the Leverhulme Trust and Amgen Inc. for financial support. This work was also supported by grants from the National Institutes of Health (GM49245, GM/OD52117 to X.C. and GM46127 to R.J.R.).
*To whom correspondence should be addressed. Tel: +1 508 927 5054; Fax: +1 508 921 1527; Email: kumar@neb.com
-d-erythro-pentofuranoside (4.23 g, 9.69 mmol) in acetonitrile (25 ml). After 10 min, a solution of N-iodosuccinimide (2.40 g, 10.66 mmol) in acetonitrile (25 ml) was added, upon which the reaction mixture changed color to dark brown. After 5 h, crystals of the [beta]-anomer of the product had formed in the reaction mixture, which was filtered and the crystals washed with acetonitrile. The crystals were then dissolved in dichloromethane and filtered, to separate them from the crushed molecular sieves. The dichloromethane was removed in vacuo to yield the pure [beta]-anomer of the product (0.86 g, 17%) as a fluffy white solid. The filtrate of the reaction mixture had the solvent removed in vacuo and the resulting solid was re-dissolved in dichloromethane. The organic layer was washed with a saturated aqueous solution of sodium thiosulfate (2 * 200 ml) and water (200 ml), dried (MgSO4), filtered and the solvent removed in vacuo. The crude product was purified, and the [alpha]- and [beta]-anomers separated, by flash column chromatography, using an eluent of ethyl acetate/n-hexane (3:1). The appropriate fractions were combined to obtain the product as the pure [beta]-anomer (0.82 g, 16%; overall yield of [beta]-anomer: 1.28 g, 33%) and pure [alpha]-anomer (2.26 g, 45%). The isolated [alpha]/[beta]-anomers of the product (3.54 g, 78%) gave the [alpha]/[beta] ratio as 3:2.
-phosphinoyloxy]-4"-thio-2"- deoxycytidine. To a stirred solution of N4-benzoyl-5"-O-(4,4"- dimethoxytriphenylmethyl)-4"-thio-2"-deoxycytidine (1.611g, 2.48 mmol) in tetrahydrofuran (150 ml), under a nitrogen atmosphere at 0oC, was added N,N-diisopropylethylamine (1.60 g, 2.16 ml, 4.96 mmol) followed by cyanoethyl-N,N-diisopropylchlorophosphoramidite (1.17 g, 1.10 ml, 4.96 mmol). The reaction mixture was then allowed to warm to room temperature and was left stirring for 6 h. The solvent was removed in vacuo and the resultant white solid was dissolved in ethyl acetate (150 ml). The organic layer was washed with a saturated aqueous solution of sodium hydrogen carbonate (2 * 150 ml) and a saturated aqueous solution of sodium chloride (1 * 150 ml). The organic layer was dried (MgSO4), filtered and the solvent removed in vacuo to yield a yellow/brown oil. The product was purified by flash column chromatography, using an initial eluent of dichloromethane (+0.5% TEA), followed by an eluent of dichloromethane/ethyl acetate (7:3). Combination of the appropriate fractions yielded the product (1.53 g, 73%) as a cream foam. Rf [0.17, 0.22 DCM/ EtOAc (7:3) +0.5% TEA]. [delta] (1H-NMR; 400 MHz, CDCl3): 8.95-8.50 (1H, v.br.s, N
REFERENCES