Cytosine analogues as DNA methyltransferase substrates

Abstract DNA methyltransferases are drug targets for myelodysplastic syndrome (MDS), chronic myelomonocytic leukemia (CMML), acute myelogenous leukemia (AML) and possibly β-hemoglobinopathies. We characterize the interaction of nucleoside analogues in DNA with a prokaryotic CpG-specific DNA methyltransferase (M.MpeI) as a model for mammalian DNMT1 methyltransferases. We tested DNA containing 5-hydroxymethylcytosine (5hmC), 5-hydroxycytosine (5OHC), 5-methyl-2-pyrimidinone (in the ribosylated form known as 5-methylzebularine, 5mZ), 5,6-dihydro-5-azacytosine (dhaC), 5-fluorocytosine (5FC), 5-chlorocytosine (5ClC), 5-bromocytosine (5BrC) and 5-iodocytosine (5IC). Covalent complex formation was by far most efficient for 5FC. Non-covalent complexes were most abundant for dhaC and 5mZ. Surprisingly, we observed methylation of 5IC and 5BrC, and to a lesser extent 5ClC and 5FC, in the presence, but not the absence of small molecule thiol nucleophiles. For 5IC and 5BrC, we demonstrated by mass spectrometry that the reactions were due to methyltransferase driven dehalogenation, followed by methylation. Crystal structures of M.MpeI-DNA complexes capture the ‘in’ conformation of the active site loop for analogues with small or rotatable (5mZ) 5-substituents and its ‘out’ form for bulky 5-substituents. Since very similar ‘in’ and ‘out’ loop conformations were also observed for DNMT1, it is likely that our conclusions generalize to other DNA methyltransferases.


Supplementary Figures
Fig. S3.MS of the total digest of modified 30-mer oligonucleotides.Oligonucleotides were degraded to 2 -deoxynucleoside monophosphates and dephosphorylated.The resulting mixtures were analyzed by LC-MS (MH+ mode).During ionization in the mass spectrometer, glycosidic bonds break, so that the free bases can be detected.Chromatograms are shown in the region where free bases elute: (A) without mass filtering, (B) with filtering for the expected mass of the protonated modified base.The identities of the expected modified bases are given next to the chromatograms.Bromine has two natural isotopes of nearly equal abundance that are separated by 2 Da in mass.For the filtering, the indicated mass of the 5BrC base was chosen.Oligoduplex sequences are indicated in Fig. 1 and Table S1.The chromatogram at the top shows the analysis for the 5mC containing oligo complementary to the one with the modification.All other chromatograms are for the strand carrying the modification.The masses of the unprotonated free bases are given next to the corresponding peaks.M.MpeI at increasing concentrations (0.5, 1, 2 and 4 µM) was incubated with modified oligoduplexes (1 µM) in the presence of SAM and DTT.Samples were loaded onto native 6% polyacrylamide TBE gels either directly after incubation (left, native conditions) or after additional treatment with 1% SDS at 65 °C for 5 min (right, denaturing conditions). 32P labeled DNA was detected by autoradiography.The chromatograms were filtered for the masses of the protonated forms of G, C, T, 5IC, 5mC, SAM, SAH (152, 112, 127, 238, 126, 298 and 385 Da).The sum of the peak areas for G was used for separate normalization of each spectrum.The areas for compounds were subsequently normalized to 100% for all spectra with the average values for C/5mC+DTT (C, T, 5mC), 5IC/5mC+DTT (5IC), M.MpeI+5IC/5mC+DTT+SAM at 0 h (SAM), M.MpeI+C/5mC+SAM at 7 h (SAH).Endpoint assays were performed in duplicate, time courses in triplicate.Results were averaged and standard errors were calculated.Note that the scale for 5mC differs from other series due to the presence of 5mC in the lower strand (lower left axes in green).1) and (B) the 5BrC->5mC containing complex presented in this work and contoured at 1.5 rmsd.(C) Theoretical densities were calculated as described (7) and contoured at either 5.5 (for 5FC and 5mC) or 3 rmsd (5BrC).The comparison of the observed and predicted densities suggests that the complexes correspond to pre-and post-reaction states.Note that due to the loop "out" conformation of M.MpeI in complex with 5BrC, the SAM most likely replaced SAH after the reaction.The red lines represent the contacts with more than 0.4 Å overlap.Such contacts were observed for (B) 5OHC, (C) 5mC (5BrC derived) and (D) 5m-dhaC (dhaC derived) for which the active site loop was present in the "out" conformation in the crystals.M.MpeI in complexes with 5FC and 5mZ DNA had the loop in the "in" conformation.In these cases steric conflicts did not occur, except with the active site Cys135 Sγ (not shown).The models were generated without adjustments of the protein and/or DNA, but some adaptive fit is expected in the actual complexes.S3.The ED maps correspond to the composite omit maps obtained with COMIT (3,4) after 10 cycles of restrained REFMAC refinement ( 6) and contoured at 1.5 rmsd.(B) The cryoEM reconstruction of the DNMT1 -5FC oligoduplex complex (10).The two PDB depositions correspond to the two models obtained for different selections of images from the same data collection.The EP maps were contoured at 7.5 rmsd.(C,D) Theoretically predicted ED and EP maps were calculated as described ( 7) and contoured at either 5.5 (ED) or 4 (EP) rmsd.5fC) and (B) 5-carboxylcytosine (5caC).The assay was performed as in Fig. 3.In principle, the methyl transfer to 5caC could occur by formation of the methyl ester of the carboxylate.However, we consider this as unlikely for structural reasons.Instead, we interpret transfer of radiolabeled methyl to 5caC as methyltransferase catalyzed decarboxylation, followed by methylation.The higher reaction rate for 5caC than for 5fC likely reflects differences in the first step.5fC is known to be largely refractory to methyltransferase driven release of formic acid, whereas 5caC is prone to methyltransferase catalyzed decarboxylation in the absence of SAM (11).In the presence of SAM, the newly formed C can be methylated.The slightly higher reaction rate in the presence of DTT than in its absence is due to a larger fraction of active M.MpeI with a reduced cysteine in the catalytic center.S4).(A) In the case of M.MpeI (top row) the "out" conformation occurs in the presence of 5OHC, dhaC and 5BrC (converted to 5m-dhaC and 5mC, respectively) and the "in" conformation for 5FC and 5mZ.In the case of DNMT1 (second row) the "out" conformation occurs in the autoinhibited, inhibited, and DNA-free forms of the enzyme.The "in" conformation is observed for productive complexes.The state of the catalytic loop observed in the non-productive forms of mammalian DNMT1 methyltransferase is similar to the "out" form of the corresponding loop observed in M.MpeI analogue complexes.The state of the loop observed in the productive complex of DNMT1 resembles the catalytically competent "in" form of the loop in the M.MpeI structures.(B) A similar distribution of "in" and "out" conformations can also be observed for the bacterial methyltransferases M.HhaI and M.HaeIII.(C) For maize ZMET2 methyltransferase (13) and putative methyltransferases from S. flexneri (PDB IDs: 3LX6, 3ME5, unpubl.)and E. coli prophage (PDB ID: 7G7U, unpubl.) the productive complexes are missing and thus the "out" conformations can only be deduced based on structural similarities.The substrate/product base is indicated in magenta, SAM/SAH in orange.Non-productively bound DNA molecules are not shown.

Fig. S2 .
Fig. S2.Verification of the oligonucleotide quality.Oligonucleotide integrity was checked by 20% PAGE in TAE buffer and detected by (A) Cy3 fluorescence, then stained with (B) GelRed.

Fig. S4 .
Fig. S4.Covalent complex formation between M.MpeI and modified oligoduplexes in the presence of S-adenosylmethionine (SAM) and absence of thiol reducing agents.The complex formation was monitored by 10% SDS PAGE after (A) 1 h and (B) overnight incubation.DNA was detected by Cy3 fluorescence (left), protein by Coomassie staining (right).Note that under electrophoresis conditions, DNA migrated out of the gel.ThermoFisher PAGERuler Prestained Protein Ladder was used as the protein size marker in all experiments (the expected protein size is 47.31 kDa).

Fig. S7 .
Fig. S7.Non-covalent versus covalent complex formation between M.MpeI and modified oligoduplexes.M.MpeI at increasing concentrations (0.5, 1, 2 and 4 µM) was incubated with modified oligoduplexes (1 µM) in the presence of SAM and DTT.Samples were loaded onto native 6% polyacrylamide TBE gels either directly after incubation (left, native conditions) or after additional treatment with 1% SDS at 65 °C for 5 min (right, denaturing conditions).32P labeled DNA was detected by autoradiography.

Fig. S11 .
Fig. S11.Dependence of the methyl transfer on the M.MpeI -DNA stoichiometry.M.MpeI catalyzed the transfer of the 3 H labelled CH3 group from SAM to the hemimethylated DNA containing either (A) C or (B) 5FC in the substrate strand.100 pmol of DNA and 20, 100 or 200 pmol of M.MpeI were used.SAM and DNA were separated by filter binding or ethanol precipitation.The transfer of the radioactive CH3 group was monitored using a scintillation counter.The experiment was repeated three times with two technical replicas each.Error bars corresponding to SD were calculated using Prism GraphPad Software.

Fig
Fig. S12.M.MpeI and small molecule thiol nucleophile driven dehalogenation.DNA in the samples was degraded to single nucleotides, dephosphorylated and analyzed by LC-MS.The LC traces shown for C and 5mC are filtered for the expected masses.All reactions were scaled to the signal from a G base.The ordinate values were chosen so that the peak heights for the C or 5mC control in the DTT containing buffer are 100%.The halogen peaks were scaled with the peak height of the modified base in DTT containing buffer.The xaxes correspond to the elution time (min).The experiments were performed in duplicate and the average values were used for scaling.(A) Control experiments for SAM dependent M.MpeI activity show the expected disapearance of the C peak and doubling of the 5mC peak.As predicted the reaction is abolished by the catalytic mutation and the absence of SAM.(B) The controls used for normalization.(C) Experiments with the halogenated oligonucleotides.

Fig
Fig. S15.Electron density maps for co-crystals of 5BrC containing DNA and M.MpeI.(A) 5mC:G pair with G in the substrate strand and 5mC in the complementary strand and (B) the flipped-out base bound in the substrate binding pocket.5-bromocytosine bound as a substrate was converted to 5mC before or during the crystallization process as evidenced by the composite omit density contoured at 1 rmsd (grey), difference density at +/-3 rmsd (green/red), and anomalous density at 3 rmsd (yellow).The B-factors of phosphorous atoms and carbons of the C5 methyl groups are indicated.Data collection was performed at the 0.9116 Å wavelength, where the theoretically predicted f″ values are approximately equal to 3.75 e for Br, 0.157 e for P and 0.003 e for C.

Fig. S16 .
Fig. S16.Experimental electron densities for the complexes of M.MpeI with (A) 5FC and (B) 5BrC (converted to 5mC) containing oligoduplexes and (C) theoretically predicted maps.The electron densities (ED) correspond to the composite omit maps calculated after 10 REFMAC (6) refinement cycles for (A) previously published M.MpeI complex with 5FC containing DNA (1) and (B) the 5BrC->5mC containing complex presented in this work and contoured at 1.5 rmsd.(C) Theoretical densities were calculated as described (7) and contoured at either 5.5 (for 5FC and 5mC) or 3 rmsd (5BrC).The comparison of the observed and predicted densities suggests that the complexes correspond to pre-and post-reaction states.Note that due to the loop "out" conformation of M.MpeI in complex with 5BrC, the SAM most likely replaced SAH after the reaction.

Fig. S19 .
Fig. S19.Different scenarios for the dehalogenation of 5BrC (shown) or 5IC (analogous).(A) Possible mechanism involving formation of a covalent complex, followed by HBr elimination, and finally enzyme regeneration.(B) Alternative reaction mechanisms, based on literature about chemical dehalogenation of 5-halopyrimidines by small molecule thiol nucleophiles (8).

Table S1 .
List of phosphoramidites and oligonucleotides used in this study.(A) Phosphoramidite building blocks for base analogues, (B) DNA sequences and sources, (C) DNA applications.

Table S3 .
Structural studies of C5 DNA methyltransferases in complex with 5FC containing oligoduplexes.The structures included in Fig.S20are indicated in bold.

Table S4 .
Conformations of the active site loop in the C5-methyltransferase structures with the greatest similarity to M.MpeI.