Selective chemical tracking of Dnmt1 catalytic activity in live cells

Summary

Enzymatic methylation of cytosine to 5-methylcytosine in DNA is a fundamental epigenetic mechanism involved in mammalian development and disease. DNA methylation is brought about by collective action of three AdoMet-dependent DNA methyltransferases, whose catalytic interactions and temporal interplay are poorly understood. We used structure-guided engineering of the Dnmt1 methyltransferase to enable catalytic transfer of azide tags onto DNA from a synthetic cofactor analog, Ado-6-azide, in vitro. We then CRISPR-edited the Dnmt1 locus in mouse embryonic stem cells to install the engineered codon, which, following pulse internalization of the Ado-6-azide cofactor by electroporation, permitted selective azide tagging of Dnmt1-specific genomic targets in cellulo. The deposited covalent tags were exploited as “click” handles for reading adjoining sequences and precise genomic mapping of the methylation sites. The proposed approach, Dnmt-TOP-seq, enables high-resolution temporal tracking of the Dnmt1 catalysis in mammalian cells, paving the way to selective studies of other methylation pathways in eukaryotic systems.

Graphical abstract

Highlights

• •A single alanine substitution in Dnmt1 confers catalytic transfer of extended groups
• •Electroporation permits facile delivery of AdoMet analogs into live mammalian cells
• •Engineered Dnmt1 adds trackable azide tags at its native target sites in cellulo
• •Dnmt-TOP-seq enables genome-wide tracking of Dnmt1 activity in live mammalian cells

Stankevičius et al. report engineering of mouse Dnmt1 methyltransferase for catalytic transfer of extended moieties onto DNA from synthetic cofactor analogs. Genomic installation of the engineered Dnmt1 and pulse internalization of a matching cofactor by electroporation enable selective covalent tagging and precise mapping of catalytic Dnmt1 targets in live pluripotent cells.

Introduction

Epigenetic regulation of cell phenotype and fate via reversible changes in DNA modification and chromatin structure underlies the mechanisms of development, aging, and disease. One of the fundamental epigenetic mechanisms is methylation of cytosines in CpG dinucleotides to 5-methylcytosine (m⁵C), which varies across different genetic loci, cells, and tissues. The DNA methylation toolkit in mammals comprises three catalytically active DNA methyltransferases (DNMT or Dnmt for mouse proteins) (Bestor, 2000; Tajima et al., 2016), which use the S-adenosyl-L-methionine (AdoMet) cofactor as the methyl group donor. The first characterized mammalian methylase, DNMT1, preferentially acts on hemimethylated CpG sites (Vilkaitis et al., 2005; Jurkowska and Jeltsch, 2016) (Figure 1A) and is thought to be largely responsible for maintaining pre-existing methylation patterns after DNA replication. The other two major types of mammalian methyltransferases, DNMT3A and DNMT3B, show no such substrate preference and are assigned major roles in methylation of unmodified genomic regions (de novo methylation). Loss of the DNMT1 function is directly linked to tumorigenesis and chromosomal instability (Eden et al., 2003; Gaudet et al., 2003), whereas DNMT3B mutations cause a severe autosomal disease, called ICF syndrome (Hansen et al., 1999; Okano et al., 1999; Xu et al., 1999). Disruption of each individual Dnmt gene in experimental mice leads to a distinct but eventually lethal phenotype, emphasizing the complexity and importance of DNA methylation in mammalian development. On the reverse pathway, DNA methylation can be gradually lost by dilution in the absence of proper maintenance methylation during DNA replication (passive demethylation). DNA methylation can also be actively “erased” via enzymatic oxidation of m⁵C by the TET oxygenases to 5-hydroxymethylcytosine (Kriaucionis and Heintz, 2009; Tahiliani et al., 2009) and then to 5-formylcytosine and 5-carboxylcytosine, which are efficiently excised by TDG glycosylase (Denis et al., 2011).

Figure 1.

Structure-guided engineering of the Dnmt1 methyltransferase for catalytic transfer of extended functionalized groups

(A) Enzymatic modification of hemimethylated CG/m⁵CG sites by the mouse Dnmt1 methyltransferase. Biological methylation using the AdoMet cofactor yields m⁵C, whereas bioorthogonal transfer of an extended side chain with a functional azide group from a synthetic cofactor analog, Ado-6-azide, yields 5-(6-azidohex-2-ynyl)cytosine (N₃-m⁵C) (shown in cyan and magenta, respectively).

(B) Structure of the catalytic/cofactor-binding pocket of WT (PDB: 6w8w) and engineered Dnmt1-R1576/N1580 variants. The methyl-accepting C5 atom of the target 5-fluorocytosine residue (shown in cyan stick presentation) and the methyl-releasing sulfur atom of the bound AdoHcy (green sticks) are connected with a dashed magenta line.

(C) Sequence conservation of the R1576 and N1580 residues in motif X across animal Dnmt1-like proteins and the GCGC-specific bacterial M.HhaI methyltransferase. Identical amino acids are shown in gray boxes.

(D) Enhanced azidoalkylation of DNA by the engineered variants of Dnmt1 methyltransferase. HPLC-MS/MS analysis of nucleosides obtained after hydrolysis of Dnmt1-modified poly(dI-dC)·poly(dI-dC) DNA. Reactions were performed using 6 μM DNA substrate, 100 μM AdoMet or Ado-6-azide, and 400 nM Dnmt1 variant for 40 min at 37 °C.

(E) Methylation activity of Dnmt1 variants toward hemimethylated and nonmethylated 25-mer DNA duplexes. Triplicate reactions containing 6 μM DNA substrate, 5.3 μM [methyl-³H]-AdoMet, and 20 nM WT or 100 nM engineered variant as shown were incubated for 40 min at 37 °C. Error bars denote ±SD.

Although DNA methylation is essential for mammalian development, the precise interplay among the three methylating enzymes remains elusive (Lee et al., 2014). Apart from substantial data supporting the classical division of the DNMTs into the maintenance and de novo functions (Arand et al., 2012), significant “collaboration” and overlap among the three methylation “writers” has been noted in recent studies (Baubec et al., 2015; Liao et al., 2015; Haggerty et al., 2021). Several methodological and technical bottlenecks grossly hinder further advance in determining the exact location, timing, and role of epigenetic marks deposited by each DNMT. First, typically, an aggregate methylation signal, i.e., the methyl groups collectively deposited by all DNMTs, can be measured by the majority of profiling techniques, which complicates dissection of their individual enzymatic contributions. Second, the roles of individual DNMTs in DNA methylation are often inferred from indirect evidence—the level of gene expression and/or physical localization of the proteins. While constituting an important layer of regulation, the presence of DNMT proteins at specific genomic regions does not always lead to their methylation—as the catalytic activity of all DNMTs is allosterically modulated by numerous external factors (Denis et al., 2011; Li et al., 2011; Jeltsch and Jurkowska, 2016), post-translational modifications (Ling et al., 2004; Estève et al., 2011; Qin et al., 2011), and variations of AdoMet/AdoHcy concentrations, some binding events may serve purely noncatalytic functions. Third, DNMT gene knockouts often used for such studies induce massive undermethylation of the genome, leading to significant phenotypic changes of the cells (Tsumura et al., 2006; Baubec et al., 2015; Liao et al., 2015; Haggerty et al., 2021). Such dramatic perturbations unavoidably lead to a significantly distorted view of the cellular behavior of the DNMTs.

To circumvent these limitations, we set out to create an experimental platform that permits high-resolution selective tracking of the catalytic contribution of individual DNMT enzymes in live mammalian cells. Our general strategy was thus to replace one of the three DNMTs with a sterically engineered orthogonal variant that preferentially utilizes a synthetically extended AdoMet analog to permit selective chemical tagging of its natural genomic targets (Figure 1A) upon delivery of the cofactor analog inside the engineered mammalian cells. Using the tethered oligonucleotide-primed sequencing (TOP-seq) technique (Staševskij et al., 2017), the deposited chemical groups were then exploited as chemical handles for reading the adjoining sequences and precise genomic mapping of the tagged methylation sites to reveal the Dnmt1-specific genomic methylation profiles. Altogether, we show that our new approach enables selective high-resolution genome-wide tracking of the Dnmt1 catalysis in mouse embryonic stem cells (mESCs), paving the way to deciphering epigenetic mechanisms in numerous developmental and disease model systems.

Design

The creation of an experimental platform for high-resolution selective tracking of the catalytic activity of an individual DNMT enzyme in live mammalian cells entailed the following four key steps: (1) engineering of a bioorthogonal Dnmt-cofactor pair that can deposit trackable chemical groups on DNA, (2) genome editing of a selected mammalian cell line to replace the engineered codon(s) in the Dnmt1 locus, (3) controlled delivery of the orthogonal cofactor into the live engineered cells, and (4) readout of the genomic positions of the Dnmt-specific chemical marks.

Our previously exploited technology named methyltransferase-directed transfer of activated groups (mTAG) (Dalhoff et al., 2006; Lukinavičius et al., 2007) afforded precise covalent tagging of DNA using appropriate DNA methyltransferases (MTases) as targeting vehicles. During these studies, we found that wild-type (WT) bacterial DNA cytosine-5 MTases were inactive with AdoMet analogs carrying side chains longer than four carbon units, but they could be sterically engineered for acceptance of larger groups (Lukinavičius et al., 2012; Tomkuvienė et al., 2019) by two-three alanine replacements of conserved residues located in the cofactor-binding pocket. Using structure-guided modeling, we selected three potential positions for steric engineering of the cofactor-binding pocket of the mouse Dnmt1 methyltransferase to enable selective deposition of extended chemical moieties containing a terminal azide group from a synthetic cofactor analog, Ado-6-azide. Our selection of a longer, linear propargylic 6-carbon moiety as the tagging unit was motivated by its poor acceptance by most WT MTases (Lukinavičius et al., 2013), which seemed particularly important for confining the transalkylation activity to the engineered MTase in the context of other endogenous AdoMet-dependent MTases

Derivation of a Dnmt1^N1580A knockin (KI) mESC line was carried out using CRISPR-Cas9 genome editing (Okamoto et al., 2019). To cause minimal alterations of the mESCs in which the activity of Dnmt1 is to be tracked, we decided to make only the required single-codon changes in the genomic copies of Dnmt1.

Unfortunately, AdoMet and its analogs are incapable of penetrating the cell membrane and thus are excluded from entering the cells from the medium under standard conditions (McMillan et al., 2005). The permeability of the cell membrane to charged compounds can be increased by installing dedicated membrane transporters (Tucker et al., 2003; Agrimi et al., 2004) or by supplying additional encapsulating delivery compounds (Zawada et al., 2018), which may lead to significant adverse effects. We therefore turned to exploring the suitability of electroporation (Rols, 2006), which uses electric discharge for a temporary enlargement and generation of membrane pores and has proven effective for the introduction of foreign genetic material into mammalian cells.

Finally, genome-wide profiling of the deposited bioorthogonal chemical marks was performed using our previously developed analytical approach, TOP-seq. The approach is based on the “click”-mediated covalent tethering of oligonucleotide primers that permit nonhomologous priming of the DNA polymerase at these internally tagged sites to directly produce adjoining regions for their sequencing and precise genomic mapping (Staševskij et al., 2017)

Results

Engineering Dnmt1 for catalytic transfer of extended moieties from synthetic AdoMet analogs

Guided by available X-ray structures (Song et al., 2012; Adam et al., 2020), we selected three positions for individual and double alanine replacements in Dnmt1 (Figures 1B and 1C). For in vitro studies of the transalkylation activity, the WT and four engineered variants were cloned and expressed as truncated (residues 291–1,620) versions in methylotrophic yeast Pichia pastoris in high yield and in a predominantly soluble form (Figures S1A and S1B).

Initially, the produced Dnmt1 variants were analyzed by comparing their activity with AdoMet and one of its extensively used extended analogs, Ado-6-azide, which contains the 6-azidohex-2-ynyl side chain carrying a terminal azide group (-CH₂-C≡C-(CH₂)₃-N₃) (Figure 1A). For kinetic activity assays, we employed two types of previously proven Dnmt1 substrates: 25-mer oligonucleotide duplexes, containing a centrally positioned hemimethylated or unmethylated CpG site and homopolymeric DNAs poly(dI-dC)·poly(dI-dC), or poly(dG-dC)·poly(dG-dC), containing iterate CpI or CpG target sites. In control experiments, the WT Dnmt1 variant isolated from Pichia pastoris showed identical specific activity on homopolymeric substrates as previously described baculovirus-derived preparations (Vilkaitis et al., 2005) but hardly detectable activity with the Ado-6-azide cofactor (Figures 1D, 1E, and S1C). Thorough kinetic studies of the engineered Dnmt1 variants under steady-state and single-turnover conditions showed their decreased methylation capacity, which pointed to notable alterations of both AdoMet binding and catalysis (Figures S2 and S3; Tables 1 and S1). To our delight, the transalkylation activity with Ado-6-azide was markedly enhanced in the case of Dnmt1-N1580A and R1576A/N1580A variants. The identity of the enzymatically produced “extended” nucleoside, 5-(6-azidohex-2-ynyl)-2′-deoxycytidine (N₃-m⁵C), was confirmed by tandem MS/MS (Figures 1D and S1D) and chromatographic comparison with corresponding products of engineered prokaryotic C5-MTases (not shown). Altogether, the enhanced transalkylation power and reduced affinity toward the natural cofactor conferred an inverse cofactor selectivity by the N1580A and R1576A/N1580A variants (14-fold and 8-fold Ado-6-azide over AdoMet preference under single-turnover conditions, respectively; Table 1) and warranted selection of these designer “alkyltransferases” for further comparative in vitro studies.

Table 1.

Single-turnover kinetic parameters of Dnmt1 variants on 25-mer duplex CG-HM DNA for AdoMet-dependent methylation and Ado-6-azide-dependent azidoalkylation

Dnmt1 variant	kchemAdoMet methylation		kchemAdo-6-azide azidoalkylation		Cofactor selectivity
	min−1	Mut/WT, folda	min−1	Mut/WT, folda	Ado-6-azide/AdoMet, fold
WT	3.0 ± 0.13	1	0.005 ± 0.005	1	1/600
N1580A	0.104 ± 0.002	1/29	1.4 ± 0.13	265	14
R1576A/N1580A	0.035 ± 0.002	1/86	0.27 ± 0.025	53	8

See Figures S2 and S3 for details.

^aValues relative to those of WT.

The capacity of the sterically engineered cofactor pocket was further examined against a selection of extended cofactor analogs (Figure 2A). Remarkably, both selected Dnmt1s were highly efficient in transferring propargylic linkers carrying a variety of functional or reporter groups, including a bulky biotin moiety. The catalytic rate of Dnmt1-N1580A with Ado-6-azide cofactor under single-turnover conditions matched within error that of the WT enzyme in the presence of AdoMet (Figure S3; Table 1) and that nearly complete modification of a short hemimethylated duplex could be achieved by brief incubation at cofactor concentrations as low as 1 μM (Figure 2B). Along with efficient two-step labeling of hemimethylated GCG sites in plasmid DNA (Figure S4), this illustrates an immense potential of this new chemoenzymatic tool for orthogonal derivatization and manipulations of DNA.

Figure 2.

Enhanced catalytic activity of the engineered Dnmt1 variants with AdoMet analogs containing sulfonium-bound propargylic side chains

(A) DNA modification reactions were carried out for 1 h at 37°C using 50 nM hemimethylated 25-mer duplex CG-HM in which the unmethylated strand was 5′-³²P-labeled, 100 nM Dnmt1 and 100 μM AdoMet, or its synthetic analog (partial structures shown in top panel). Resulting modified DNA was reannealed with a 125-fold molar excess of an unmodified complementary strand, digested with the methylation-sensitive HhaI endonuclease, fractionated by denaturing PAGE and 5′-³²P-labeled strands, and visualized by autoradiography (see also Figure S4).

(B) The effect of Ado-6-azide concentration on activity of Dnmt1 variants under single-turnover conditions. Reactions were carried out using 50 nM DNA, 100 nM Dnmt1, and Ado-6-azide as shown.

However, the key factor for achieving the ultimate goal of chemical tracking of Dnmt1 action in living cells is robust transalkylation activity in the presence of the endogenous cofactor. Therefore, we compared the incorporation of azidohexynyl groups from Ado-6-azide in the presence of AdoMet by the selected Dnmt1 variants using different experimental setups. Enzymatic methylation and alkylation of DNA was probed using HPLC-MS/MS analysis of nucleosides (Figures 3A and S5A) or by radiography of electrophoretically separated 5′-³²P-labeled strands (Figure 3B). The latter experiment showed that in the presence of equal amounts of both cofactors, the extended cofactor was preferentially used by both engineered variants (the level of alkylation 65%–67%). In contrast, while some alkylation can be observed with Ado-6-azide alone, the WT Dnmt1 yielded no detectable alkylation products in the presence of competing AdoMet. In mouse cell lysates, the exogenous Dnmt1 proteins again showed an inverse catalytic proficiency order for the methylation (WT >> N1580A, R1576A/N1580A, Figure S5B) as compared with the azidoalkylation (WT << R1576A/N1580A < N1580A, Figure S5C) reactions. These observations reassured that the catalytic transfer of bioorthogonal groups by Dnmt1 could be achieved in the background of endogenous AdoMet and other components present inside the mammalian cells.

Figure 3.

Cofactor selectivity of the wild-type and engineered Dnmt1 variants

(A) Comparative analysis of DNA modification products by the Dnmt1 variants in the presence of AdoMet, Ado-6-azide, and their equimolar mixture under steady-state conditions. DNA modification reactions containing 6 μM poly(dI-dC)·poly(dI-dC) DNA substrate, 400 nM Dnmt1 variants, and 100 μM cofactor were incubated for 40 min at 37°C; modified DNA was hydrolyzed to nucleosides and analyzed using HPLC-MS/MS.

(B) Comparative analysis of DNA modification products by the Dnmt1 variants at varied ratios of the AdoMet and Ado-6-azide cofactors under single-turnover conditions. DNA modification reactions containing 50 nM hemimethylated oligonucleotide, 100 nM Dnmt1, and 100 μM cofactor or their mixture as indicated were incubated for 60 min at 37°C. Reactions were analyzed as described in Figure 2.

Engineering mESCs for Dnmt1-directed deposition of extended azide tags

Derivation of a Dnmt1^N1580A KI mouse embryonic stem E14TG2a cell line was carried out by replacing the AAT codon to GCT in exon 38 using CRISPR-Cas9 genome editing (Okamoto et al., 2019). A 70-nt single-stranded donor template with an additional intronic G to T mutation at the PAM sequence was used to prevent recutting of the KI locus (Figures S6A–S6C). The correctness of the biallelic N1580A substitution was confirmed via restriction analysis and genomic sequencing of adjacent regions. qRT-PCR analysis confirmed that the expression level of Dnmt1-N1580A in KI cells was identical within error with that of Dnmt1-WT in the parental E14TG2a cell line (Figure S6D). The produced KI line, along with the WT and Dnmt1 KO controls, was then examined for global m⁵C levels using HPLC-MS/MS analysis (Figure S6E). The cells were cultured in media with serum and leukemia inhibitory factor (LIF), which exhibit global DNA hypermethylation (Leitch et al., 2013) (Figure 4A). Notably, the observed intermediate genomic m⁵C levels in the KI cells indicated that the endogenously expressed Dnmt1-N1580A retained partial methylation potential in vivo, in line with its methylation activities observed in vitro. Addition of the Ado-6-azide cofactor to the KI mouse ESC lysates conferred efficient azidoalkylation of exogenous pΔL2-14 plasmid DNA containing hemimethylated CGC/Gm⁵CG sites (Gerasimaitė et al., 2009) by the engineered Dnmt1 but not by the WT control (Figure 4B). These observations clearly suggested that selective deposition of extended groups by the endogenously expressed Dnmt1 variant could be achieved, provided that the synthetic cofactor could be made available inside the mouse cells.

Figure 4.

Installation of bioorthogonal functional groups in DNA by endogenous Dnmt1-N1580A in mESC lysates and live cells

(A) Strategy for selective Dnmt1-directed bioorthogonal pulse-tagging of endogenous DNA targets in a living cell. m, methylated cytosine.

(B) Endogenous Dnmt1-N1580A alkyltransferase activity in cell lysate. DNA alkylation reactions were performed for 1 h at 37°C in 20 μL of corresponding mESC lysates supplemented with 100 μM Ado-6-azide and 500 ng of in vivo-hemimethylated pΔL2-14 plasmid DNA (Gerasimaitė et al., 2009), and DNA hydrolysates were analyzed for N₃-m⁵C using HPLC-MS/MS.

(C) Experimental procedure for selective Dnmt1-directed pulse-tagging of endogenous DNA targets in live mESCs.

(D) HPLC-MS/MS analysis of genomic cytosine modification in Dnmt1^WT and Dnmt1^N1580A mESCs after electroporation with 1 mM of Ado-6-azide.

(E) Effects of cofactor concentration and postelectroporation incubation (chase) time on Dnmt1-directed intragenomic incorporation of N₃-m⁵C in mESCs.

(F) Viability of WT and engineered mESCs after electroporation in the presence of 1 mM of Ado-6-azide. Cell survival was determined using an MTT assay 24 h after electroporation. Mean ± SD of at least three independent replicates. ^∗p < 0.05; n.s., not significant. See also Figure S6.

To this end, we examined the suitability of electroporation (Rols, 2006), which uses electric discharge for a temporary enlargement and generation of membrane pores and has proven effective for the introduction of foreign genetic material into mammalian cells. After extensive experimental trials, we established electroporation conditions that permitted a clearly detectable pulse labeling of genomic DNA in the KI cells by exogenous Ado-6-azide (Figure 4C). Typically, serum/LIF medium-grown cells were subjected to electroporation and brief exposure (25 min) to serum-free medium containing Ado-6-azide cofactor, and then reversed to serum medium and incubated for an additional 1–6 h. Intragenic incorporation of N₃-m⁵C in KI mESCs was linearly dose dependent (at least up to 1 mM cofactor) and peaked at ∼3 h incubation time (Figures 4D and 4E); our practically attained N₃-m⁵C levels (with 1 mM Ado-6-azide and 3 h incubation) were about 2 orders of magnitude below the endogenous levels of m⁵C (0.02% and 4%, respectively; see Figures 4E and S6E). A methyltetrazole dye (MTT) cell viability assay performed 24 h after such electroporation-driven pulse labeling showed a very high survival rate of the WT and KI mESCs both in the presence and absence of exogenous cofactor in the medium (Figure 4F), indicating that neither the electroporation shock nor the genomic incorporation of the extended m⁵C analog at this level was detrimental to the cells. Altogether, these results clearly demonstrate that a synthetic cofactor can be delivered to and utilized in live mammalian cells by the engineered Dnmt1 in a temporally controlled manner. This provides a new chemical tool for “recording” the genomic footprint of the Dnmt1 catalytic activity in live mammalian cells in the background of native DNA methylation machinery.

Genome-wide selective tracking of Dnmt1 catalysis in live ESCs

Finally, we assessed the suitability of the engineered mESC line to report on the intracellular Dnmt1-dependent methylation events using the Ado-6-azide-dependent chemical pulse-tagging of genomic DNA. We performed genome-wide profiling of the deposited bioorthogonal chemical marks using our previously developed analytical approach, TOP-seq. The approach is based on the “click”-mediated covalent tethering of oligonucleotide primers that permit nonhomologous priming of the DNA polymerase at these internally tagged sites to directly produce adjoining regions for their sequencing and precise genomic mapping (Staševskij et al., 2017). We prepared triplicate Dnmt-TOP-seq libraries (two technical replicates each) from the serum/LIF-cultivated Dnmt1^N1580A (engineered) and Dnmt1^WT (control) mESCs following their pulse labeling with 1 mM Ado-6-azide as depicted in Figure 4C. The number of sequencing reads obtained from the Ado-6-azide-treated engineered cells (30–60 M per technical replicate) was 2–3 orders of magnitude higher than that in the controls (Figures S7A and S7B), as were the correlations between biological replicates both genome-wide and within individual genomic elements (Figure S7C). Importantly, we found that the start position of the overwhelming majority (>90%) of mapped Dnmt1^N1580A reads occurred within 4 nt from a genomic CpG site (Figures 5A and S7D). This was very reassuring, given that Dnmt1 is a CpG-specific methyltransferase, and therefore such reads were attributed as identified CpGs or Dnmt1-modified CpGs (although it is possible that a small fraction of non-CpG reads represent genuine Dnmt1 methylation targets in mESCs, they were excluded from further analysis to minimize noise). The average coverage of identified CpGs was 6–8× per replicate (Figures S7E and S7F) and showed an even distribution in the DNA strands and across the chromosomes (Figure S7G). In contrast, the fraction of CpG-associated reads was only ∼10% in the control libraries, indicating their highly random character and thus a negligible contribution from the Dnmt1-WT or de novo Dnmt3a or 3b methyltransferases (Figures 5A and S7D).

Figure 5.

Analysis of genomic Dnmt1 modification sites in mESCs using Dnmt-TOP-seq

(A) Distance distribution of read start positions to a nearest CpG site in the Dnmt-TOP-seq libraries prepared from WT and Dnmt1^N1580A mESCs 3 h after electroporation with 1 mM Ado-6-azide.

(B) Dnmt-TOP-seq CpG modification profiles along generalized genomic elements for various gene types. Modified CpG sites were computed in the upstream (4 kb from TSS), gene body (from TSS to TTS normalized by gene length), and downstream (4 kb from TTS) regions. Processed pseudogenes are reverse-transcribed copies of mRNAs that lack introns, whereas unprocessed pseudogenes are produced by gene duplication and may contain introns.

(C) Enrichment analysis of Dnmt-TOP-seq genomic elements and regulatory features. Odds ratio denote the enrichment (>1) or the depletion (<1) of particular genomic element terms in the genome-wide DNA modification profile of Dnmt1^N1580 mESC.

(D) Enriched GO terms of genes containing methylated CGI promoters in Dnmt1^N1580A cells. CGIs bearing at least one modified CpG in all biological replicates were designated for the analysis. q value denotes false discovery rate. See also Figures S7 and S8.

To establish the suitability of the produced Dnmt1^N1580A (further on Dnmt1-TOP-seq) libraries for epigenome studies, we looked at the genome-wide distribution of identified modified CpG sites along a range of genomic features. In protein-coding genes, the CpG modification profile showed a characteristic profile (Jung et al., 2017; Buitrago et al., 2021) with a sharp drop at TSSs followed by a gradual increase toward the 3′ end of genes and a second drop around TTS (Figure 5B). In addition, long regulatory RNAs showed a similar but less pronounced modification profile, in agreement with their gene-like transcriptional status. As a background (untranscribed) reference, we looked at pseudogenes and found that unprocessed pseudogenes indeed maintained a low and uniform level of modification throughout their length. Notably, processed pseudogenes manifested a distinct profile with two DNA modification peaks at TSS and TTS, separated by a valley at gene bodies (Figure 5B). This unexpected observation clearly points at some yet unreported phenomenon that involves epigenetic regulation of these seemingly silent elements (Troskie et al., 2021).

In the next step, we performed functional enrichment analysis (Hansen et al., 2011) of the Dnmt-TOP-seq libraries and identified significant Dnmt1 modification differences between genomic region sets and regulatory elements (Figure 5C). We found strong hypomodification of CpG islands, 5′ UTR/promoter regions, and specific types of noncoding regulatory RNAs, such as miRNAs and snRNAs. In contrast, modified CpGs were enriched in intron/exon coding regions with downstream 3′ UTRs, long noncoding RNAs, and splicing-associated snRNAs. Notably, our analysis revealed contrasting modification of mobile elements (high in SINEs and low in LINEs), in good agreement with previously observed genome-wide methylation of distinct genomic regions in mouse ES-E14TG2a (Zhao et al., 2014). In spite of the generally low methylation levels, the modified positions in CGIs appeared to be well covered in the Dnmt-TOP-seq libraries (Figure S8A). To gain a further system-level understanding of the Dnmt1 activity, we inspected which gene pathways were enriched for Dnmt-TOP-seq modifications. We selected fragments satisfying the criteria of FDR-adjusted p value > 0.05 and absolute log transformed Fisher’s estimate >1. We found the majority of such elements to be depleted in modified CpG sites, except for enriched DNA methylation in gene bodies (Figures S8B and S8C). Gene ontology (GO) analysis of the hypomodified genes found that they were strongly enriched for genes related to multicellular organism development processes and cell fate commitment (Figure S8D). Methylated CGIs within promoters showed an enrichment in nuclear division and in pathways related to meiotic cell cycle, gamete generation, and piRNA metabolic processes (Figure 5D), whereas the most highly enriched GO terms (statistical significances are weaker compared with those obtained with promoter CpGs) among genes with methylated CGIs assigned to intergenic CGIs included categories such as development and cell-cell signaling processes (Figure S8E). The specific enrichment of cell cycle and noncoding RNA regulatory functions among methylated promoter-associated CpGs, but not among abundantly methylated genes per se or intergenic CGIs, proposed regulatory coherence by this type of CpGs.

In an effort to decouple the contribution of Dnmts in the establishment and maintenance of DNA methylation, we compared the Dnmt1-specific modification profile with genome-wide binding profiles of Dnmt3a and Dnmt3b in mESCs obtained by ChIP-seq (Weinberg et al., 2019). Analysis of intergenic, intragenic, and promoter CGIs revealed a quite intricate interplay of the three DNMTs (Figures 6A, S9A, and S9B), suggesting that methylation of intragenic and promoter CGIs could require contributions from Dnmt1 and Dnmt3b. Remarkably, in LINE and LTR retrotransposons, we can see that the patterns of Dnmt3a1 and Dnmt3b localization show no correlation with the Dnmt1 activity, suggesting that the de novo methyltransferases have a weak, if any, impact on the LINE and LTR modification in ESCs (Figure 6B). The latter results are consistent with previous findings implicating Dnmt1 as the key player in silencing of mobile elements and maintaining genome stability (Ran et al., 2013; Min et al., 2020; Haggerty et al., 2021).

Figure 6.

Contribution of Dnmt1 to methylation of genomic CpG sites in mESCs

(A and B) Comparison of Dnmt-TOP-seq CpG modification profiles (top panel in yellow) with Dnmt3a1 and Dnmt3b ChIP-seq normalized profiles (data obtained from Weinberg et al., 2019) in and around CpG islands located in promoters (2 kb upstream of protein-coding genes), intragenic and intergenic regions (A), or LINE and LTR elements (B). Profiles representing 20% slices of the most modified (top), moderately modified (mid), and least modified (bottom) regions were derived from experimental Dnmt-TOP-seq data. Bottom panel: average ChIP-seq read profiles for Dnmt3a (upper) and Dnmt3b (lower) at genomic regions selected above.

(C) Validation of CGI modification profiles in H1fnt and Sfi1 genes. Columns denote read coverage or methylation levels of a particular CpG determined by Dnmt-TOP-seq (upper panel) or bisulfite sequencing (lower panel), respectively. Gaussian kernel-smoothed profiles are shown as dashed lines. Error bars denote ±SD.

Having demonstrated that Dnmt-TOP-seq can produce comprehensive high-resolution genome-wide profiles of the Dnmt1-directed modification, we went on to the level of individual loci to see how well it compares with native m⁵C patterns, which represent collective methylation produced by the three cellular Dnmts. A total of 6 selected CGI regions, including 4 hypomodified loci (CGIs of Tfcp2I1, Cdh7, Gli2, and Ddx18 genes) involved in mESC self-renewal pathways (Das et al., 2013; Ming et al., 2020) or essential for embryonic development (Pieters and van Roy, 2014) and 2 moderately modified regions (CGIs of Sfi1 and H1fnt genes) associated with spermatogenesis (Martianov et al., 2005; Lambrot et al., 2013), were validated using padlock bisulfite sequencing (BS-seq). As expected, BS-seq analysis of the hypomodified regions revealed virtually no CpG methylation, clearly confirming our Dnmt-TOP-seq data and thus excluding the possibility that the lack of Dnmt-TOP-seq reads was due to insufficient sensitivity (Figure S9C). In further agreement, BS-seq analysis revealed 30%–40% methylation coverage in the two moderately modified regions (Figure S9D). Overall, we found a good resemblance between the positional methylation profiles of the latter loci obtained by BS-seq and our method (Figure 6C), except for a notable deviation at the 5′ end of the promoter CGI in Sfi1 gene. We also performed BS-seq analysis of our generated isogenic Dnmt1 KO mESCs (fully inactivated Dnmt1). The differential KI-KO BS-seq signal was expected to roughly reflect the Dnmt1-N1580A methylation activity with AdoMet, which in turn could be compared with the Dnmt-TOP-seq profile. The presented data (Figure 6C) indeed show a good concordance (shape similarity) between the differential BS-seq and Dnmt-TOP-seq profiles. The left shoulder of the BS-seq signal in Sfi1 would be expected to derive from de novo Dnmt3 methylation. This is consistent with previously reported DNMT3a recruitment at CGI shores, although individual isoforms may show different cell-specific genomic binding (Manzo et al., 2017). In contrast, methylation at the central region of intergenic CGI in H1fnt was increased in KO cells, suggesting that Dnmt1 could somehow interfere with de novo methylation activity at this particular locus in Dnmt1^N1580A mESCs. Future detailed studies using Dnmt-TOP-seq are needed to gain a better understanding of the mechanisms underlying the temporal activity of Dnmt1 in mESCs.

Discussion

Previously, AdoMet analog-based profiling of protein methylation targets was described in cell lysates (Sohtome and Sodeoka, 2018), and prototype systems were developed for analysis of histone methylation (Wang et al., 2013) and mRNA adenine-N6 methylation (Hartstock et al., 2018; Shu et al., 2020) in transfected and/or methionine-deprived cells. Here, we present the first engineered mammalian system that enables activity-based chemical tracking of a specific DNA methylating enzyme in live cells under nearly native conditions. Our first step was the successful engineering of a mammalian Dnmt to an alkynyl transferase that preferentially uses synthetic AdoMet analogs for transfer of bioorthogonal chemical moieties carrying a range of functional and reporter groups on to its genuine targets in DNA (Figures 1 and 2). The transalkylation rate by the engineered Dnmt1 from the Ado-6-azide cofactor is nearly on par with the methyltransfer rate of the WT enzyme in the presence of AdoMet (1.4 and 3.0 min⁻¹, respectively) and is 265-fold higher than that of the WT and Ado-6-azide pair (Table 1). The DNMTs contain well-conserved catalytic domains with ten identified sequence motifs shared between eukaryotic and prokaryotic cytosine-5 MTases. Remarkably, although similar engineering of bacterial orthologs typically requires two replacements in conserved motifs IV and X (Lukinavičius et al., 2012; Deen et al., 2018), the largest enhancement in transalkylation activity and cofactor selectivity was achieved by a single replacement of the conserved Asn1580 to Ala in motif X of Dnmt1 (Figures 1B and 1C). The side chain of this residue serves to sterically confine the conformation of the sulfonium-bound methyl group for its in-line attack of the C5 nucleophile; although its removal leads to looser binding and positioning of AdoMet, it creates a wider exit channel for an extended transferable group out of the cofactor pocket (see Figures 1B and 7 in Lukinavičius et al., 2012). Besides obvious advantages at the genomic CRISPR-Cas engineering step (Figure S6), a single-codon replacement also offers best assurances that other biologically important functions of Dnmt1, such as allosteric regulation or interaction with other proteins, will not be affected.

Here, we also demonstrate that electropermeabilization can be used for efficient delivery of exogenous positively charged compounds into mammalian cells with no apparent size limitation. Metabolic in-cell production of AdoMet analogs has been proven for short transferrable groups (3 carbon atoms) and can generally be achieved under conditions of methionine deprivation (Hartstock et al., 2018; Shu et al., 2020). The latter may lead to dramatically altered DNA methylation, DNMT expression, and significant phenotypic changes of the stem cells (Shiraki et al., 2014; Jung et al., 2017). Our optimized pulse-labeling procedure showed negligible effects on the functionality and viability of ESCs (Figure 4). Inside the engineered cells, the internalized Ado-6-azide cofactor is utilized by the engineered Dnmt1 variant to selectively tag its biological methylation sites in vivo, whereas in its absence, the enzyme performs normal methylation functions using endogenous AdoMet (Figure 5). The discharge intensity, cofactor concentration, and exposure time are the major factors determining the temporal parameters and the amplitude of a labeling pulse. Our currently demonstrated density of intracellular chemical tagging was in the order of 1% of the steady-state genomic methylation level, which seemed like a reasonable trade-off between the tracking sensitivity and the preservation of a near-native state of the genome during the chase phase. The azidoalkylation density can be increased further, if necessary, by using higher concentrations of the cofactor analog (Figure 4E, left), which was currently limited by the cost and available synthetic capacity. On the other hand, the produced biallelic KI conferred a 30% reduction in global cytosine methylation (Figure S6E), which could likely be reversed by replacing only one genomic copy of Dnmt1 in mESCs. Altogether, by further optimization of these and other parameters, the system can be tailored to meet a range of experimental demands.

The in vivo-tagged DNA contains CpG target sites sparsely decorated with an unnatural chemical group. Previous studies found that azidoalkylated DNA is immune to the action of restriction endonucleases in vitro and modification-dependent restriction systems, such as McrBC, both in vitro and in vivo; it can transform E. coli cells with high efficiency, indicating a good acceptance of the artificially modified DNA in bacterial cells (Lukinavičius et al., 2012, 2013). The azidoalkylated cytosines are also largely inert to TET oxidation and base excision DNA repair (Tomkuvienė et al., 2020), and they do not hinder the action of DNA polymerases (Kriukienė et al., 2012), so they may be gradually diluted upon replication or removed via the long-patch repair mechanisms. A higher biological “durability” of the bioorthogonal chemical tags, as compared with the TET-vulnerable methyl groups, should enable direct study of the kinetic aspects of DNA methylation at single-CpG resolution, which has not been possible to date (Bachman et al., 2014; Spada et al., 2020).

The in vivo-generated chemical footprint of the Dnmt1 methyltransferase catalysis lends itself for precision mapping in the genomic sequence using the TOP-seq technique (Figures 5 and 6). The Dnmt-TOP-seq maps permit comprehensive high-resolution analysis of the enzyme-specific methylation landscape. This is attested by good general agreement of the inferred local and genome-wide features with BS-seq data and functional/transcription status of mESCs. On the other hand, examples of fine comparisons of the Dnmt1 catalytic footprints and the aggregate DNA methylation signal point at new intriguing facts and potential mechanistic findings. Moreover, the deposited orthogonal groups will surely be amenable to long-read DNA mapping using Oxford nanopore or single-molecule, real-time (Roberts et al., 2013; Logsdon et al., 2020) sequencing technologies, which are capable of sensing smaller epigenetic modifications on DNA. Alternatively, the deposited chemical tags could potentially be appended with fluorescent reporters for spatial nuclear mapping using super-resolution imaging (Wang et al., 2016) or used for covalent cross-linking and genomic mapping of spatially proximal sites (akin to Hi-C technology; Belton et al., 2012) for 3D tracking of the catalytic trajectories. Altogether, an inherently low selectivity of the cell electropermealization and the acceptance of a wide variety of chemical moieties including biotin by the engineered Dnmt1 (Figure 2A) provide unprecedented flexibility in chemical tracking and manipulation of epigenetic processes in live cells. Due to absolute sequence identity of the conserved catalytic motifs of the mammalian DNMT1 proteins (Figure 1C), the established approach will likely be “portable” for studies of human and other vertebrate cells. Similar engineering of other mammalian cell lines will provide powerful tools to directly track the action of individual DNMTs in a range of developmental and disease models. The new type of epigenomic information (Dnmt-selective methylation profile), which was inaccessible using previous technologies, will facilitate resolution of the many puzzles of how genomic methylation is established and maintained during development, senescence, and disease.

Limitations

As a pilot model aimed at demonstrating the general capabilities of the new approach, the described system is not optimized to suit a particular experimental study. At the DNA-tagging step, many factors such as dosage of the engineered Dnmt1, cofactor concentration, cofactor chemistry, electroporation intensity, and labeling duration can be varied to achieve desired labeling parameters and the level of deviation from a native state. To this end, although the WT and KI mESCs genetically differ only by three nucleotides, the engineered Dnmt1 is a weaker methyltransferase, leading to lower general genomic methylation levels in cells grown under LIF conditions. At the genomic mapping step, the chemical groups were shown to be well detectable and mappable using the TOP-seq technology, but other DNA sequencing modalities can also be used if single-molecule long-read analysis is desired.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, peptides, and recombinant proteins
Gelatin	Sigma-Aldrich	Cat#G1890
Trypsin-EDTA	Gibco	Cat#15400-054
Dulbecco‘s modified Eagle‘s medium	Gibco	Cat#11960-044
Embryonic stem-cell FBS	Gibco	Cat#16141-079
Penicillin-Streptomycin	Gibco	Cat#15140-122
Sodium pyruvate	Gibco	Cat#11360-070
2-mercaptoethanol	Gibco	Cat#21985-023
GlutaMAX	Gibco	Cat#35050-061
MEM Non-essential amino acids	Gibco	Cat#11140-050
Mouse leukemia inhibitory factor	EMD Millipore	Cat#ESG1106
CHIR99021	Sigma-Aldrich	Cat#SML1046
PD0325901	Sigma-Aldrich	Cat#PZ0162
Opti-MEM	Gibco	Cat#11058-021
TranscriptAid T7 High Yield Transcription Kit	Thermo Fisher Scientific	Cat#K0441
GeneArt Platinum Cas9 Nuclease	Invitrogen	Cat#B25640
Lipofectamine LTX	Invitrogen	Cat#15335-100
Puromycin	Sigma-Aldrich	Cat#P8833
Phire Tissue Direct PCR Master Mix	Thermo Fisher Scientific	Cat#F170L
S-adenosyl-L-homocysteine	Sigma-Aldrich	Cat#A9384
S-adenosyl-L-methionine	Sigma-Aldrich	Cat#A7007
AdoButen (S-Adenosyl-S-crotyl-L-homocysteine)	Prof. Elmar Weinhold (Stecher et al., 2009)	N/A
AdoButyn (S-Adenosyl-S-but-2-ynyl-L-homocysteine)	Dalhoff et al., 2006	N/A
AdoEnYn (S-Adenosyl-S-pent-4-yn-2-enyl-L-homocysteine)	Peters et al., 2010	N/A
Ado-6-azide (S-Adenosyl-S-(6-azidohex-2-ynyl)- L-homocysteine)	Masevičius et al., 2016	N/A
Ado-6-amine (S-Adenosyl-S-(6-aminohex-2-ynyl)-L-homocysteine)	Masevičius et al., 2016	N/A
Ado-6-ethyne (S-Adenosyl-S-(oct-2,6-diynyl)-L-homocysteine)	Lukinavičius et al., 2013	N/A
Ado-13-biotin	Osipenko et al., 2017	N/A
Poly(dI-dC) • Poly(dI-dC)	Sigma-Aldrich	Cat#P4929
Poly(dG-dC) • Poly(dG-dC)	Sigma-Aldrich	Cat#P9389
[methyl-3H]-S-adenosyl-L-methionine	PerkinElmer	Cat#NET155001MC
[γ-32P]ATP	PerkinElmer	Cat#NEG502Z001MC
T4 PNK	Thermo Fisher Scientific	Cat#EK0032
Geneticin	Sigma-Aldrich	Cat#A1720
PMSF	Sigma-Aldrich	Cat#78830
cOmplete Mini EDTA-free Protease Inhibitor Cocktail	Roche	Cat#4693159001
SatI restriction enzyme	Thermo Fisher Scientific	Cat#ER1641
FastDigest BspTI	Thermo Fisher Scientific	Cat#FD0834
FastDigest NotI	Thermo Fisher Scientific	Cat#FD0596
FastDigest EcoRI	Thermo Fisher Scientific	Cat#FD0274
FastDigest SacI	Thermo Fisher Scientific	Cat#FD1134
FastDigest HhaI	Thermo Fisher Scientific	Cat#FD1854
FastDigest XbaI	Thermo Fisher Scientific	Cat#FD0685
Scintillation cocktail Rotiszint® Eco plus	Carl Roth	Cat#0016.3
Proteinase K, recombinant, PCR grade	Thermo Fisher Scientific	Cat#EO0491
RNase A	Thermo Fisher Scientific	Cat#EN0531
Genomic DNA Clean & Concentrator-10 Kit	Zymo Research	Cat#D4011
Nuclease P1 from Penicillium citrinum	Sigma-Aldrich	Cat#N8630
FastAP Thermosensitive Alkaline Phosphatase	Thermo Fisher Scientific	Cat#EF0654
dCTP	Thermo Fisher Scientific	Cat#R0151
dGTP	Thermo Fisher Scientific	Cat#R0161
dATP	Thermo Fisher Scientific	Cat# R0141
dTTP	Thermo Fisher Scientific	Cat#R0171
5-Methyl-dCTP	Thermo Fisher Scientific	Cat#R0431
Alkyne MegaStokes dye 608	Sigma-Aldrich	Cat#79249
Igepal CA-630	Sigma-Aldrich	Cat#I8896
BCN-amine	Sigma-Aldrich	Cat#745073
MTT	Sigma-Aldrich	Cat#M5655
Fast DNA End Repair Kit	Thermo Fisher Scientific	Cat#K0771
Klenow Fragment, exo-	Thermo Fisher Scientific	Cat#EP0422
T4 DNA Ligase	Thermo Fisher Scientific	Cat#EL0011
CuBr, 99.999%	Sigma-Aldrich	Cat#254185-10G
DMSO	Sigma-Aldrich	Cat#472301
THPTA	Sigma-Aldrich	Cat#762342-500MG
dNTP Mix (2 mM each)	Thermo Fisher Scientific	Cat#R0241
Pfu DNA polymerase (recombinant)	Thermo Fisher Scientific	Cat#EP0502
Platinum SuperFi PCR Master Mix	Thermo Fisher Scientific	Cat#12358010
Dynabeads MyOne C1 Streptavidin magnetic beads	Thermo Fisher Scientific	Cat#65002
GeneJET PCR purification kit	Thermo Fisher Scientific	Cat#K0702
GeneJet NGS Cleanup kit	Thermo Fisher Scientific	Cat#K0851
DNA Clean & Concentrator-5 Kit	Zymo Research	Cat#D4013
MagJET NGS Cleanup and Size Selection Kit	Thermo Fisher Scientific	Cat# K2821
Agilent High Sensitivity DNA Kit	Agilent	Cat#5067-4626
Ion PI™ Hi-Q™ OT2 200 Kit	Thermo Fisher Scientific	Cat# A26434
Ion PI™ Hi-Q™ Sequencing 200	Thermo Fisher Scientific	Cat# A26772
Ion PI™ Chip Kit v3	Thermo Fisher Scientific	Cat# A26771
GeneJet RNA purification kit	Thermo Fisher Scientific	Cat#K0731
dsDNase	Thermo Fisher Scientific	Cat#EN0771
RevertAid™ RT reverse transcriptase	Thermo Fisher Scientific	Cat#EP0441
2x SYBR™ Green PCR Master Mix	Thermo Fisher Scientific	Cat#K253
Phusion U Hot Start polymerase	Thermo Fisher Scientific	Cat#F555S
EZ DNA Methylation-Gold Kit	Zymo Research	Cat#D5005
Deposited data
Dnmt-TOP-seq data	This work	GEO: GSE182445
ChIP-seq data for Dnmt3a1 and Dnmt3b	Weinberg et al., 2019	GEO: GSE118785
Raw gel images	This work	Mendeley Data: 10.17632/kddxk2d7pc.1
Experimental models: Cell lines
Mouse embryonic stem cells, E14TG2a	ATCC	CRL-1821
Mouse embryonic stem cells, E14TG2a Dnmt1N1580A	This work	N/A
Mouse embryonic stem cells, E14TG2a Dnmt1-KO	This work	N/A
P. pastoris GS115	Thermo Fisher Scientific	Cat#C18100
Oligonucleotides
See Table S2 for oligonucleotide sequences	Metabion	N/A
Biotinylated alkyne-containing DNA oligonucleotide, 5’-TXTTTTGTGTGGTTTGGAGACTGACTACCAGATGT AACA-Biotin; X=C8-Alkyne-dT	BaseClick	N/A
Complementary priming strand with custom LNA modifications and phosphorothioate linkages at the 3′ end, 5’-TGTTACATCTGGTAGTCAGTCTCCAAACCACACAA	Exiqon	N/A
Recombinant DNA
pSpCas9(BB)-2A-Puro (PX459) plasmid	Addgene (Rohde et al., 2010)	Cat#62988
pΔL2-14 plasmid	Gerasimaitė et al., 2009	N/A
pUC19 plasmid	Thermo Fisher Scientific	Cat#SD0061
pPIC3.5K	Thermo Fisher Scientific	Cat#V17320
pPIC3.5K-Dnmt1-dN	Vilkaitis et al., 2005	N/A
Software and algorithms
Cutadapt (3.4)	Martin, 2011	https://cutadapt.readthedocs.io
R (3.5)	R project	https://www.r-project.org/
FASTX (0.0.13)	Hannon Lab, CSHL	http://hannonlab.cshl.edu/fastx_toolkit/index.html
BWA (0.7.17)	Li and Durbin, 2009	http://bio-bwa.sourceforge.net/
PyMOL (1.7)	PyMOL Molecular Graphics System by Schrödinger	https://pymol.org/
liftOver	UCSC genome browser store	https://genome-store.ucsc.edu
clusterProfile (3.10.1)	Yu et al., 2012	http://bioconductor.org/packages/release/bioc/html/clusterProfiler.html
Other
Repeat Masker annotation	Karolchik et al., 2004	https://genome.ucsc.edu
CpG island annotation	Karolchik et al., 2004	https://genome.ucsc.edu
Gene annotation	Frankish et al., 2019	https://www.gencodegenes.org
Detailed bench protocol	This work	Methods S1

Resource availability

Lead contact

Further information and reasonable requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Saulius Klimašauskas (saulius.klimasauskas@bti.vu.lt).

Materials availability

Plasmids generated in this study will be made available upon request.

Mouse cell lines generated in this study will be made available upon request.

Experimental model and subject details

Mouse embryonic stem E14TG2a cell line was obtained from the American Type Culture Collection of Authenticated Cell Cultures (ATCC CRL-1821). Cells were grown on 0.15% gelatin-coated dishes in Dulbecco‘s modified Eagle‘s medium (DMEM; Gibco) containing 15% fetal bovine serum (Gibco), 1× penicillin/streptomycin (Gibco), 0.1 mM sodium pyruvate (Gibco), 0.1 mM β-mercaptoethanol (Gibco), 1 mM L-alanyl-L-glutamine (Gibco), 1× non-essential amino acids (NEAA; Gibco), 1×10³ U/ml mouse leukemia inhibitory factor (mLIF; Millipore), 3 μM CHIR99021 (Sigma-Aldrich) and 1 μM PD0325901 (Sigma-Aldrich). Cells were maintained at 37°C in a humidified atmosphere containing 5% CO₂.

Generation of murine embryonic Dnmt1^N1580A knock-in cells using CRISPR-Cas9 editing

An E14TG2a cell line bearing a homozygous N1580A codon substitution in exon 38 of the mouse Dnmt1 locus was generated using Cas9/sgRNA RNP nuclease and a 70-mer ssDNA as a HDR template sgRNA was designed using the Benchling tool (www.benchling.com) and synthesized using a TranscriptAid T7 High Yield Transcription Kit (Thermo Fisher Scientific) as described previously (Modzelewski et al., 2018). Before cell lipofection, Cas9/sgRNA RNP complexes were assembled in 25 μL of Opti-MEM medium (Gibco) using 500 ng of recombinant Cas9 protein (Invitrogen) and 125 ng of in vitro transcribed sgRNA at room temperature for 10 min. Subsequently, 500 ng of ssDNA template (Metabion) were added and the mixture was incubated for additional 5 min. 7.5×10⁴ E14TG2a cells were transfected with the Cas9-sgRNA RNP and ssDNA mixture using Lipofectamine LTX (Invitrogen) according to the manufacturer’s instructions using reverse transfection. 24 h after cell transfection, single cells were isolated into individual gelatin-coated wells in 96-well plates by serial dilution. After next 6 days of culture, single clones were picked and passaged in gelatin-coated 24-well plates. Approximately up to two next weeks during clone expansion, 1×10⁵ cells of each clone were collected and screened by PCR using a Phire Tissue Direct PCR Master Mix and SatI (Thermo Fisher Scientific) restriction analysis according manufacturer’s recommendations. To generate Dnmt1 KO mESCs, E14TG2a cells were transfected with only one PX459 plasmid encoding the sgRNA sequence (500 ng) using Lipofectamine LTX as described above. Cells resistant to 2 μg/mL puromycin were plated for clonal expansion. All alterations in the Dnmt1 coding sequence in selected clones were verified by Sanger sequencing. Primer sequences are listed in Table S2.

Method details

Methyltransferase substrates

AdoMet and AdoHcy were purchased from Sigma-Aldrich. [methyl-³H]-AdoMet was acquired from PerkinElmer. AdoMet cofactor analogues were chemically synthesized from AdoHcy as previously described (Dalhoff et al., 2006; Peters et al., 2010; Lukinavičius et al., 2013; Masevičius et al., 2016; Osipenko et al., 2017) and chromatographically enriched in the enzymatically active S,S-isomer using reversed-phase preparative HPLC. For large scale in vivo labeling experiments, Ado-6-azide was produced as previously described (Masevičius et al., 2016) and used without reversed-phase preparative HPLC purification as a mixture or R,S- and S,S-diastereomers.

Poly(dI-dC)·poly(dI-dC) and poly(dG-dC)·poly(dG-dC) DNAs were purchased from Sigma Aldrich. Hemimethylated or non-methylated 25-mer DNA duplexes were obtained by annealing complementary oligonucleotides (Metabion). To generate radiolabeled DNA duplexes, 5′-ends of 25-mer oligonucleotides were labeled using 0.5 U/μL T4 PNK (ThermoFisher Scientific) and 0.167 μM [γ-³²P]ATP (PerkinElmer) for 30 min at 37°C. Duplexes were prepared by annealing 5’-labeled oligonucleotides with complementary unlabeled strands. Sequences of oligonucleotides are listed in Table S2.

pΔL2-14 plasmid DNA containing hemimethylated Gm⁵CG/CGC sites was prepared as described previously (Gerasimaitė et al., 2009). The plasmid encodes an engineered GCG-specific DNA MTase which confers in vivo hemimethylation of 80 non-palindromic (GCGD, D=not C) and full (double-strand) methylation of 14 palindromic (GCGC) sites on the plasmid when expressed in bacterial cells. Briefly, E. coli ER2267 cells were transformed with pΔL2-14 and grown in 5 mL of LB media supplemented with 100 μg/mL ampicillin overnight at 37°C. 1/5 of cell culture was transferred in 200 mL of LB medium with 100 μg/mL ampicillin and cells were grown to OD₆₀₀ ∼0.6–0.8. 0.4 mM IPTG was added and cells incubated for 2 h at 37°C to induce the expression of the hemimethylase gene. Cells were harvested by centrifugation and plasmid DNA was purified using GeneJET Plasmid Miniprep kit (Thermo Fisher Scientific). The methylation level of the substrate DNA was estimated by digestion with m⁵C sensitive restriction endonucleases R.HhaI (cleaves GCGC sites) and R.Bsh1236I (CGCG) followed by agarose gel electrophoresis..

Dnmt1 mutagenesis, expression and purification

In silico examination and modelling of the catalytic/cofactor binding pocket in the co-crystal structures of mouse Dnmt1-AdoHcy complexes (PDB entries: 6w8v and 6w8w) was carried out using the PyMOL Molecular Graphics System (v. 1.7).

The pPIC3.5K vector containing a His-tagged N-truncated form (residues 291–1602) of the mouse DNA methyltransferase Dnmt1 was constructed as described previously (Vilkaitis et al., 2005). Site-directed mutagenesis of the Dnmt1 variants was performed using a two-step megaprimer method as described previously (Baranauskė et al., 2015). Briefly, PCR fragments containing site-directed Dnmt1 mutations were generated during a two round PCR using primers listed in Table S2 and pPIC3.5K-Dnmt1-dN vector as a template. Obtained fragments were cloned into BsptI and NotI sites (for Q1230A substitution) or EcoRI and NotI sites (for R1576A, N1580A and R1576A/N1580A substitution) of pPIC3.5K-Dnmt1-dN. To obtain P. pastoris transformants, SacI linearized plasmids were electroporated in P. pastoris GS115 (his4) strain. His⁺ transformants were selected on a minimal agar medium (0.67% YNB, 2% glucose) and clones harboring a high copy number of Dnmt1 variant were subsequently selected on YPD-agar plates containing 2.5 mg/mL geneticin (Sigma).

His-tagged Dnmt1 protein variants were purified as described previously (Vilkaitis et al., 2005). Briefly, P. pastoris cells were grown in BMG medium (100 mM potassium phosphate, pH 6, 1.34% YNB, 0.4 ppm biotin, 1% glycerol) at 30°C until OD₅₉₅ of 2–2.5 and then resuspended in BMM medium (100 mM potassium phosphate, pH 6, 1.34% YNB, 0.4 ppm biotin) containing 1% methanol. Following 24 h of methanol induction, cells were harvested and disrupted with a Bead Beater (Roth) in lysis buffer containing 50 mM Na₂HPO₄, 10% sucrose, 250 mM NaCl, 3 mM MgCl₂, 0.1% Triton X-100, 0.1 mM PMSF and cOmplete Mini EDTA-free Protease Inhibitor Cocktail (Roche), pH 7.4. Soluble fraction was collected after centrifugation for 20 min at 45000g and filtered through a 0.45 μm Millex-HV PVDF filter (Millipore). Filtered lysate was loaded onto a HiTrap IMAC HP column (Merk) preloaded with Ni²⁺ and equilibrated with wash buffer (50 mM Na₂HPO₄, 250 mM NaCl, 10% (w/v) sucrose, 3 mM MgCl₂, 0.1% triton X-100, 5 mM 2-mercaptoethanol, pH 6.2). Dnmt1 proteins were eluted with wash buffer supplemented with 1 M imidazole (pH 6.2) and dialyzed twice against buffer A overnight and against storage buffer (50 mM Na₂HPO₄, 250 mM NaCl, 0.1% triton X-100, 5 mM 2-mercaptoethanol and 50% glycerol, pH 7.4). Aliquoted proteins were stored at -20°C

Radiometric ³H incorporation analysis

Dnmt1 catalytic parameters were determined by measuring the incorporation of ³H-methyl group using [methyl-³H]-AdoMet as described previously (Vilkaitis et al., 2005). Briefly, reactions were performed in reaction buffer (20 mM Tris-HCl, pH 7.4, 0.5 mM EDTA, 0.2 mM DTT, 0.1 mg/mL bovine serum albumin) containing 20 nM WT or 100 nM engineered Dnmt1, 5.3 μM [methyl-³H]-AdoMet, 6 μM DNA substrate (poly(dI-dC)·(dI-dC) or poly(dG-dC)·(dG-dC), hemimethylated or non-methylated 25-mer duplex). All reactions were conducted for 40 min at 37°C and quenched by heating for 10 min at 80°C. In separate series substrate concentrations were varied in the range of 0.02–11 μM for cofactor, or 0.002–12 μM (target sites) for DNA. To determine the levels of ³H-methylcytosine, samples were spread on 2.3 cm DE-81 filter papers (Whatman, UK), washed four times with 50 mM NaPO₄ buffer (pH 7.4), twice with H₂O, twice with 96% ethanol, once with acetone and dried under IR lamp. ³H-methyl group incorporation was measured in 2 mL of scintillation cocktail Rotiszint® Eco plus (Carl Roth) using a Hidex 300SL (LabLogic Systems) scintillation spectrometer.

HPLC-MS/MS analysis of modified nucleosides

Typically, 50 ng of modified DNA was digested to nucleotides with 0.01 U/μL of nuclease P1 (Sigma) in 40 μL of P1 reaction buffer (10 mM NaAc, 1 mM ZnAc, pH 5.5) for 4 h at 50°C followed by an overnight incubation with 0.01 U/μL of FastAP alkaline phosphatase (Thermo Fisher Scientific) at 37°C. Reactions were quenched by heating at 80°C for 10 min and centrifugation at 19000g for 30 min at 4°C. Centrifugate samples were loaded on an integrated HPLC/ESI-MS/MS system (Agilent 1290 Infinity/ 6410 Triple Quad LC/MS) equipped with a Supelco Discovery HS C18 column (7.5 cm × 2.1 mm, 3 μm) by elution with a linear gradient of solvents A (0.0075% formic acid in water) and B (0.0075% formic acid in acetonitrile) at a flow of 0.3 mL/min at 30°C as follows: 0–5 min, 0% B; 5–15 min, 10% B; 15–20 min, 100% B. Mass spectrometer was operating in the positive ion MRM mode and intensities of nucleoside-specific ion transitions were recorded: N₃-m⁵dC 349.2→233.1, m⁵dC 242.1→126.1; dC 228.1→112.0, dG 268.1→152.1.

For quantitative analysis, standard curves were generated using FastAP-treated nucleotides (Thermo Fisher Scientific): m⁵dC (linear range of 0.0125–0.8 pmol) and dC, dG, dT, dA (0.38–12 pmol). N₃-m⁵dC calibration was achieved by analysis of 0.2–25 fmol of hydrolysates obtained from hemimethylated 25-mer oligonucleotide duplex that was fully modified with Dnmt1 N1580A protein in the presence of Ado-6-azide, and thus contained equimolar amounts of m⁵dC and N₃-m⁵dC). To compensate for DNA input variations, all signals were normalized to dG. Data were analyzed using Agilent MassHunter software and Microsoft Excel.

Single-turnover kinetic analysis

Single turnover reactions containing 100 nM Dnmt1, 50 nM appropriately 5’-³²P labeled 25-mer oligonucleotide duplex and 100 μM cofactor (AdoMet or Ado-6-azide) in reaction buffer (20 mM Tris-HCl, pH 7.4, 0.5 mM EDTA, 0.2 mM DTT, 0.1 mg/mL bovine serum albumin) were incubated for specified periods of time and quenched by heating for 10 min at 80°C. Control reactions contained no Dnmt1. Resulting modified DNA was reannealed with a 125-molar excess of an unmodified complementary 25-mer strand and digested with the R.HhaI endonuclease (ThermoFisher Scientific) in FastDigest buffer for 15 min at 37°C. Samples were fractionated on a 13% denaturing polyacrylamide gel with 7 M urea under denaturing conditions. Gels were exposed to Phosphor imaging plates and scanned by FLA-5100 Image Reader (Fujifilm) with a red 635 nm laser and IP filter and analyzed using Multi Gauge software (Fujifilm).

Comparative analysis of the Dnmt1catalytic activity with AdoMet and Ado-6-azide

Catalytic reactions were carried out with 100 nM Dnmt1 (WT, N1580A and R1576A/N1580A) 50 nM 5’-³²P labeled hemimethylated duplex, and 100 μM cofactor (AdoMet and/or Ado-6-azide premixed at certain ratios) for 1 h at 37°C and then enzyme was inactivated by heating for 10 min at 80°C. To differentiate the migration of methylated and azidoalkylated strands in polyacrylamide gels, “click” modification of the latter strand was performed by supplementing reactions with 1 mM BCN-amine (Sigma) for 3 h at 37°C. Samples were next proceeded as described above.

Analysis of the Dnmt1 catalytic activity with extended AdoMet analogs

DNA modification reactions were performed using 50 nM 5’-³²P labeled hemimethylated 25-mer oligonucleotide duplex, 100 nM Dnmt1 variant in the presence of 100 μM AdoMet or synthetic cofactors in reaction buffer (20 mM Tris-HCl, pH 7.4, 0.5 mM EDTA, 0.2 mM DTT, 0.1 mg/mL bovine serum albumin) for 1 h at 37°C quenched by heating for 10 min at 80°C. Samples were processed as described above.

Two-step sequence-specific labelling of hemimethylated pΔL2-14 plasmid by copper-free click chemistry

500 ng of pΔL2-14 plasmid DNA was supplemented with 100 nM Dnmt1 and 100 μM Ado-6-azide in reaction buffer (20 mM Tris-HCl, pH 7.4, 0.5 mM EDTA, 0.2 mM DTT, 0.1 mg/mL bovine serum albumin) and incubated for 1 h at 37°C. Reactions were stopped by heating for 10 min at 80°C, supplemented with 1% SDS, treated with 1 μg/mL of Proteinase K for 30 min at 55°C and purified by ethanol precipitation. Modified plasmid DNA was then treated with 5 μM alkyne MegaStokes 608 (Sigma) in 20 mM of Tris-HCl buffer (pH 7.4) for 3 h at 37°C and precipitated with ethanol. 300 ng of labelled DNA was fragmented with EcoRI and XbaI (ThermoFisher Scientific) and analyzed on a 1% agarose gel. Gels were scanned for MegaStokes 608 fluorescence in a FLA-5100 Image Reader using a 473 nm laser. To visualize bulk DNA, gels were stained with EtBr, rescanned and analyzed using Multi Gauge software.

Preparation of cell lysates

10⁶ murine embryonic stem cells were collected and washed once in 1 ml of PBS, once in 500 μL wash buffer (20 mM Tris-HCl, pH 7.4, 20 mM NaCl, 6 mM MgCl₂) and spun down for 2 min at 1000 g at 4°C. Cell pellet was gently suspended by pipetting in 500 μL hypotonic buffer (20 mM Tris-HCl, pH 7.4, 20 mM NaCl, 6 mM MgCl₂, 1 mM DTT, 0.1% Igepal CA-630). After incubation for 10 min in ice, cells were spun down for 5 min at 1000 g at 4°C. Cell pellet was gently resuspended in 100 μL of nuclear extraction buffer (60 mM Tris-HCl, pH 7.4, 1.2 mM EDTA, 0.5% Igepal CA-630, 1 mM DTT, 0.1 mg/mL bovine serum albumin, cOmplete Mini EDTA-free Protease Inhibitor Cocktail) and sonicated 5 times (30 s pulse and 30 s pause) at 4°C in a Bioruptor sonicator (Diagenode). Soluble cell lysates were collected following centrifugation for 15 min at 14000 g at 4°C.

Evaluation of the Dnmt1 activity in cell lysates

In vitro Dnmt1 methylation activity in freshly prepared cell lysates was determined by measuring incorporation of ³H-methyl group from [methyl-³H]-AdoMet into DNA. Methylation reactions were performed in 20 μL of E14TG2a cell lysate containing 100 nM Dnmt1, 5.3 μM [methyl-³H]-AdoMet and 6 μM DNA. In vitro alkylation reactions were performed in 20 μL of E14TG2a cell lysate containing 100 nM Dnmt1, 100 μM Ado-6-azide and 500 ng of pΔL2-14 plasmid DNA. For the evaluation of endogenous Dnmt1 activity, alkylation reactions were performed as above with no exogenous Dnmt1 protein added. All reactions were conducted for 1 h at 37°C and stopped by heating for 10 min at 80°C. Alkylation reactions were then supplemented with 1% SDS and treated with 1 μg/mL Proteinase K for 30 min at 55°C and modified DNA was purified by phenol-chloroform extraction.

Cell electroporation with Ado-6-azide

Cells were plated in 24-well plates at a density of 5x10⁴ cells per well and incubated for 48 h in DMEM media supplemented with fetal bovine serum and mLIF. Each well was washed twice with Opti-MEM medium (ThermoFisher Scientific) and inoculated with freshly prepared 300 μL of 0.5–1 mM Ado-6-azide solution in Opti-MEM medium and electroporated using a NEPA21 electroporator (Nepa Gene) fitted with a cell culture plate electrode CUY900 at a cell poring pulse voltage of 200 V and a pulse duration of 5 ms. After electroporation, cells were incubated for further 25 min at 37°C, washed with Opti-MEM, covered with fresh mESC medium and incubated for further 1–6 hours. Cell pellets from each sample were collected and stored at -20°C.

Cell viability MTT assay

Cells were treated with 500 μL of 0.1 mg/mL 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT, Sigma) in PBS for 3 min at 37°C, washed twice with PBS and lysed with 200 μL neat isopropanol. Formazan absorption in samples was measured at 570 nm using a Synergy H4 (Biotek) plate reader.

gDNA isolation and quantitative HPLC-MS/MS

5x10⁵ of pelleted embryonic stem cells were resuspended in 200 μL of a 1:1 mixture of PBS and Lysis buffer (0.2 M Tris-HCl, 0.25 M NaCl, 0.025 M EDTA, 0.5% SDS, pH 8; 1 mg/mL Proteinase K) and incubated for 40 min at 55°C. Samples were then treated with RNase A (ThermoFisher Scientific) for 10 min at room temperature and genomic DNA isolated using a Genomic DNA Clean & Concentrator-10 Kit (Zymo Research) and stored at -20°C. 500 ng of isolated gDNA was treated with P1 endonuclease and FastAP phosphatase and subjected to MS/MS-HPLC spectrometer as described above.

Preparation of Dnmt-TOP-seq libraries of mESCs

The preparation of Top-seq libraries was carried out as described (Staševskij et al., 2017). Briefly, 900 ng of genomic DNA, isolated from WT or Dnmt1^N1580A E14TG2a cells electroporated with 1 mM Ado-6-azide, was sheared to 200 bp with a Covaris E220 sonicator. Fragmented DNA was end-repaired using a DNA End Repair Kit (Thermo Fisher Scientific) as recommended by the manufacturer and purified using a GeneJet PCR Purification Kit (Thermo Fisher Scientific). Next, DNA fragments were 3’-A-tailed with Klenow exo- polymerase in Klenow buffer (ThermoFisher Scientific) in the presence of 0.5 mM dATP at 37°C for 45 min followed by inactivation at 75°C for 15 min and purified through a DNA Clean & Concentrator-5 column (Zymo Research). Annealed partially complementary A1/A2 adapters (Metabion) were ligated using 15 U of T4 DNA Ligase (ThermoFisher Scientific) in ligase buffer at 22°C overnight following by thermal inactivation at 65°C for 10 min and purification through a DNA Clean & Concentrator-5 column (Zymo Research). Next, DNA was supplemented with 20 μM biotinylated alkyne-containing DNA oligonucleotide (5′-T(alkyneT)TTTTGTGTGGTTTGGAGACTGACTACCAGATGTAACA-(biotin)-3′; Base-click) and 8 mM CuBr: 24 mM THPTA mixture (Sigma) in 50% of DMSO, incubated for 20 min at 45°C, and diluted to ∼1% DMSO before purification by a GeneJet NGS Cleanup kit (ThermoFisher Scientific). Biotinylated DNA was enriched by incubation with 0.1 mg Dynabeads MyOne C1 Streptavidin (ThermoFisher Scientific) in buffer A (10 mM Tris–HCl (pH 8.5), 1 M NaCl) at room temperature for 3 h on a roller. After incubation, DNA-bound beads were washed twice with buffer B (10 mM Tris–HCl (pH 8.5), 3 M NaCl, 0.05% Tween 20); twice with buffer A (supplemented with 0.05% Tween 20); once with 100 mM NaCl, resuspended in water and heated for 5 min at 95°C to elute the enriched DNA fraction. Recovered DNA was subsequently applied to a priming reaction containing 1 U Pfu DNA polymerase (ThermoFisher Scientific), 0.2 mM dNTP and 0.5 μM complementary priming strand 5′-TGTTACATCTGGTAGTCAGTCTCCAAACCACACAA-3′ (with custom LNA modifications and phosphorothioate linkages at the 3′ end; Exiqon). The priming reaction was performed at the following cycling conditions: 95°C 2 min; 5 cycles at 95°C 1 min, 65°C 10 min, 72°C 10 min. Amplification of a primed DNA library was carried out by adding the priming reaction mixture to 50 μL of amplification reaction containing 25 μL of 2 × Platinum SuperFi PCR Master Mix (ThermoFisher Scientific) and barcoded fusion PCR primers A(Ad)-EP-barcode-primer (63 nt) and trP1(Ad)-A2-primer (45 nt) at 0.5 μM each (both primers contained phosphorothioate modifications). Thermocycler conditions were as follows: 94°C 4 min; 15 cycles at 95°C 1 min, 60°C 1 min, 72°C 1 min. Final libraries were size-selected for ∼ 300 bp fragments using a MagJET NGS Cleanup and Size Selection Kit (ThermoFisher Scientific). Library quality and quantity were assessed with a 2100 Bioanalyzer (Agilent). Libraries were subjected to Ion Proton (ThermoFisher Scientific) sequencing. Primer sequences are enlisted in Table S2.

Sanger bisulfite sequencing

400 ng of genomic DNA isolated from ES cells was bisulfite-converted using a Zymo DNA Methylation Gold Kit (Zymo Research) and then subjected to PCR to amplify target regions using Phusion U Hot Start polymerase (Thermo Fisher Scientific). Both procedures were performed according to manufacturer's instructions. Obtained PCR products were purified using a GeneJET PCR purification kit (Thermo Fisher Scientific) and cloned into the pUC19 vector. Sixteen clones of each sample were randomly selected for Sanger sequencing. CpG methylation status of bisulfite sequencing data was evaluated using BISMA tool (Rohde et al., 2010). All primers are listed in Table S2.

RT-qPCR

5x10⁵ ES cells were processed using a GeneJet RNA purification kit (Thermo Fisher Scientific) and then treated with dsDNAse (Thermo Fisher Scientific). cDNA synthesis was performed with 500 ng of total RNA and RevertAid RT reverse transcriptase (Thermo Fisher Scientific). Quantitative real-time PCR analysis was carried out on a Rotor-Gene 6000 (Corbett Life Science) using a SYBR™ Green PCR Master Mix (Thermo Fisher Scientific). All protocols were carried out according to manufacturer’s recommendations. Determined mRNA levels were normalized to Gapdh expression. PCR primers (Metabion) are listed in Table S2.

Quantification and statistical analysis

The biochemical data analysis and statistical tests used are described within the figure legends and methods section describing each experiment.

Processing of Dnmt-TOP-seq data

mESC TOP-seq data was processed as described previously (Gibas et al., 2020). First, reads longer than 80 nt (cutadapt -1 80) and containing 5’ adapter sequences were retained (cutadapt -g ˆGTTACATCTGGTAGTCAGTCTCCAAACCACACAAAAA -e 0.1 -O 10), followed by trimming of 3’ adapter sequence using cutadapt tool (cutadapt -a AGATTGGAAGAGTGTCATGTAGGGAAAGAGTG -e 0.1 -O 10) (Martin, 2011). FASTX quality trimmer was then used to trim 3’ read ends below a phred quality score of 20 (fastq_quality_trimmer -t 20 -l 15) (http://hannonlab.cshl.edu/fastx_toolkit/index.html). Processed read sequences were mapped to the mouse genome (build mm10) using BWA tool and only mappings with quality score 30 or above were retained (Li and Durbin, 2009). Identical reads (the same original length and starting genomic coordinate) were termed PCR duplicates and only one per group was retained. All selected reads were assigned to a particular CpG site with absolute distance between the read start and CpG site between 0 and 4 bp using an in-house script. Reads with identical distance to multiple CpG sites were omitted from analysis. Finally, CpG coverage was defined as a sum of all assigned reads on both strands.

Data analysis pipeline is provided in the following scheme:

Functional enrichment analysis

Modified promoter CpG island Gene Ontology (GO) analysis. All CGIs intersecting with a protein-coding gene upstream region (2 kb) were defined as promoter CGIs. For each biological replicate fraction of CpG sites with at least one read was computed and only CGIs where all biological replicates had identified CpG sites were selected. Next, GO Enrichment Analysis of selected genes was performed using the enrichGO function from the clusterProfile suite (Yu et al., 2012). All protein-coding genes were used as a background set and only GO enrichments with Benjamini-Hochberg adjusted q-value less than 0.001 were selected.

Enrichment of various genomic elements with the modification signal was computed using Fisher's exact test. First, a contingency table was formatted for each CpG site falling within a specific genomic element and containing at least one read. Next, Fisher’s exact test was performed to estimate the odds ratio and significance for each enrichment or depletion. Modification enrichment within protein-coding gene bodies was calculated in a similar fashion. For each gene, a contingency table was computed to test if a CpG site contains at least one read and overlaps a specific gene. Fisher's exact test was used to calculate the odds ratio and p-value. Next, genes that showed the same direction of change (either depletion or enrichment) and FDR adjusted q-value less than 0.05 in all biological replicates were defined as depleted or enriched with DNA modification. Finally, GO enrichment analysis was performed as mentioned above and only genes with Benjamini-Hochberg adjusted q-value less than 0.05 were selected.

Mouse genome sequence in the fasta file format, Repeat element (Repeat Masker annotation, rmsk) and CpG island (cpgIslandExt annotation) genomic locations in the BED file format were downloaded from the UCSC database (Karolchik et al., 2004). Reference gene annotations was obtained from the GENCODE project in the GTF file format (Frankish et al., 2019).

Public Sequencing Datasets

Dnmt-TOP-seq data have been deposited at the NCBI GEO under accession number GSE182445. ChIP-seq data for mouse Dnmt3a1 and Dnmt3b were downloaded through NCBI GEO GSE118785 (Weinberg et al., 2019).

Published: March 3, 2022

Data and code availability

• •Dnmt-TOP-seq sequencing data reported in this paper have been deposited at GEO and are publicly available as of the date of publication. Accession numbers are listed in the key resources table. Raw gel images have been deposited at Mendeley Data and are publicly available as of the date of publication. DOIs are listed in the key resources table.
• •This paper does not report original code. All software used for analysis is listed in the key resources table and is freely available online.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.