An Adversarial DNA N⁶-Methyladenine-Sensor Network Preserves Polycomb Silencing

Summary

Adenine N6 methylation in DNA (6mA) is widespread among bacteria and phage and is detected in mammalian genomes, where its function is largely unexplored. Here we show that 6mA deposition and removal are catalyzed by the Mettl4 methyltransferase and Alkbh4 dioxygenase, respectively, and that 6mA accumulation in genic elements corresponds with transcriptional silencing. Inactivation of murine Mettl4 depletes 6mA and causes sublethality and craniofacial dysmorphism in incross progeny. We identify distinct 6mA sensor domains of prokaryotic origin within the MPND deubiquitinase and ASXL1, a component of the Polycomb repressive deubiquitinase (PR-DUB) complex, both of which act to remove monoubiquitin from histone H2A (H2A-K119Ub), a repressive mark. Deposition of 6mA by Mettl4 triggers the proteolytic destruction of both sensor proteins, preserving genome-wide H2A-K119Ub levels. Expression of the bacterial 6mA methyltransferase Dam, in contrast, fails to destroy either sensor. These findings uncover a native, adversarial 6mA network architecture that preserves Polycomb silencing.

Graphical Abstract

Highlights

• •6mA deposition and erasure by mammalian Mettl4 and Alkbh4, respectively
• •Mettl4-deficient mice display craniofacial dysmorphism
• •6mA triggers proteolysis of its cognate sensor proteins ASXL1 and MPND
• •Adversarial 6mA network architecture preserves Polycomb silencing

Kweon et al. reveal the molecular infrastructure of a mammalian DNA N6-methyladenine (6mA) methylation, erasure, and sensing system. Deposition of 6mA by the Mettl4 methyltransferase triggers the proteolytic destruction of its cognate sensor proteins, the histone H2A deubiquitinases ASXL1 and MPND, thus preserving Polycomb gene silencing.

Introduction

DNA base modification systems evolved under the intense selective pressure of bacterial predation by prokaryotic viruses (phage) to discriminate self from invasive genomes and were, in turn, co-opted by phage as a counter-restriction strategy (Bryson et al., 2015, Iyer et al., 2009, Schlagman and Hattman, 1983, Weigele and Raleigh, 2016). In bacteria, enzymatically modified DNA bases also perform essential roles in DNA replication and repair (Abeles et al., 1993, Au et al., 1992). A multitude of DNA base modifications are found in prokaryotes, yet relatively few have been detected in higher eukaryotes, with 5-methylcytosine (5mC) and its enzymatically oxidized derivatives representing well-characterized exceptions (Schübeler, 2015, Tahiliani et al., 2009).

Methylation of adenine at the N6 position in DNA (6mA), among the most prevalent DNA modifications in prokaryotes, was recently detected in metazoan genomic DNA, including in vertebrates (Greer et al., 2015, Koziol et al., 2016, Wu et al., 2016, Zhang et al., 2015, Zhu et al., 2018), suggesting a previously unrecognized form of gene control. Deposition of 6mA in the nematode worm C. elegans is catalyzed by DAMT-1, a member of the Ime4-like/MT-A70 clade of methyltransferases (Greer et al., 2015, Iyer et al., 2016), while N6AMT1, a member of the HemK family of protein N-methyltransferases, may perform a similar role in human cells (Kusevic et al., 2016, Xiao et al., 2018). Conversely, the enzymatic removal of 6mA has been attributed to the Alkbh1 and Alkbh4 dioxygenases in mammalian cells and invertebrates, respectively (Greer et al., 2015, Wu et al., 2016, Xiao et al., 2018, Xie et al., 2018, Zhang et al., 2015). The function of mammalian orthologs of Alkbh4 in the oxidative demethylation of 6mA, if any, has not been addressed experimentally.

6mA has been implicated in gene transactivation in the unicellular eukaryote Chlamydomonas and in the invertebrates Drosophila and C. elegans (Fu et al., 2015, Greer et al., 2015, Zhang et al., 2015). In C. elegans, mutually reinforcing deposition of the activating histone mark H3K4me2 and 6mA defines a gene regulatory circuit (Greer et al., 2015). In contrast, 6mA is associated with epigenetic silencing in mammalian cells (Wu et al., 2016, Xie et al., 2018). In metazoans, silenced chromatin domains are established principally through the action of Polycomb repressive complexes (PRC1 and PRC2), which catalyze conjugation of monoubiquitin to histone H2A (H2A-K119Ub) and trimethylation of histone H3 at lysine 27 (H3K27me3), respectively (Gao et al., 2012, Kirmizis et al., 2004, Schuettengruber et al., 2017, Wang et al., 2004). Both of these marks are reversible: H3K27 demethylation is catalyzed by Kdm6 enzymes (Bosselut, 2016, Manna et al., 2015), while MYSM1/MPND deubiquitinases (DUBs) and the Polycomb repressive deubiquitinase (PR-DUB) complex deconjugate monoubiquitin from H2A-K119Ub (Scheuermann et al., 2010, Zhu et al., 2007). High-sensitivity mass spectrometry studies have established that PR-DUB contains the core catalytic subunit BAP1, a ubiquitin hydrolase, and the regulatory components ASXL1 or ASXL2, which regulate BAP1 (Hauri et al., 2016, Kloet et al., 2016). ASXL paralogs contain a central ASXH domain that stimulates BAP1 deubiquitinase activity in vitro (Sahtoe et al., 2016). How PR-DUB composition and catalytic activity are controlled in mammalian cells is largely unknown.

In this work, we uncover the components of a 6mA deposition, erasure, and sensor network in mammalian cells. We identify structurally distinct 6mA sensor domains within MPND and ASXL1 and demonstrate that native deposition of 6mA stimulates the proteolytic destruction of both sensor proteins. These observations reveal how 6mA functions antagonistically against its cognate sensors to preserve and sustain Polycomb-mediated gene silencing.

Results

Mammalian Mettl4 and Alkbh4 Enzymes Catalyze 6mA Deposition and Erasure, Respectively

Mettl4 belongs to a subclade of MT-A70 adenine methyltransferases (Bujnicki et al., 2002) that is distinct from RNA-directed methyltransferases and includes both C. elegans DAMT-1 and the bacterial DNA 6mA methyltransferase M.MunI (Iyer et al., 2016) (Figures 1A and 1B). Expression of Flag-tagged E. coli 6mA methyltransferase Dam (Flag-Dam) or Mettl4 (Flag-Mettl4), but not a Mettl4 catalytic site mutant (DPPW->DRRW), in HEK293T cells promoted the accumulation of 6mA in genomic DNA, as detected by dot blot using 6mA antisera and by liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis of DNA hydrolysates (Figure 1C). To evaluate the methyltransferase activity of Mettl4 in vivo, we applied CRISPR/Cas9 gene targeting in mouse zygotes to generate inactivating mutations in Mettl4 and obtained pups carrying a heterozygous deletion of exon 6, which creates a translational frameshift that ablates the catalytic domain, for a predicted methyltransferase knockout (KO) allele (Figure S1). Quantitative LC-MS/MS analysis of genomic DNA hydrolysates prepared from strain-matched wild-type (WT) and Mettl4 KO embryonic stem cells (ESCs) revealed depletion of 6mA in the KO samples to undetectable levels from a mean of 8.6 ppm (relative to dA) in WT ESCs (Figure 1D, left panel). In agreement with previous reports of a nonredundant, essential role for the Mettl3/Mettl14 complex in mRNA adenine N6-methylation (m6A) (Dominissini et al., 2012, Geula et al., 2015, Liu et al., 2014), inactivation of Mettl4 did not significantly decrease the level of m6A in mRNA (Figure 1D, right panel). Inactivation of Mettl4 also diminished 6mA levels in spleen genomic DNA (Figure 1E), suggesting a role for Mettl4 in 6mA deposition during development and in adult tissues.

Figure 1.

Mettl4 and Alkbh4 Catalyze Deposition and Erasure, Respectively, of 6mA

(A) Schematic presentation of Mettl4 protein structure and linear arrangement of conserved motifs in methyltransferase domain (amino acids 257–471), showing detailed alignment to motif IV sequence logo in catalytic site of MT-A70 adenine N6-methyltransferases.

(B) Clustering of proteins within the MT-A70 family. RNA-specific methyltransferases form a tight cluster of closely related sequences, while Mettl4 and the DNA adenine N6-methyltransferases DAMT-1 (C. elegans) and M. MunI (Mycoplasma) (red circles) are positioned apart.

(C) HEK293T cells transfected with empty vector (EV) or expressing Flag-Dam or the wild-type or DPPW catalytic-site mutant (PPmut) variants of Flag-Mettl4 were harvested and proteins in whole-cell extracts resolved by SDS-PAGE and immunoblotting with Flag antisera. Actin, loading control. 6mA and 5mC in the same samples were detected by dot blot of genomic DNA (n = 3 experiments), and mean 6mA levels relative to dA were quantified by LC-MS/MS analysis of genomic DNA hydrolysates (n = 2 experiments).

(D) Levels of 6mA relative to total dA in genomic DNA (left panel) and m6A relative to total adenosine (A) in mRNA (right panel), purified from WT and Mettl4 KO ESCs, as determined by quantitative LC-MS/MS. Dashed line represents limit of detection. n.d., not detected. Data are plotted as mean with SD (n = 2 experiments).

(E) Overlaid extracted LC-MS chromatograms of dA and 6mA in genomic DNA hydrolysates prepared from WT and Mettl4 KO spleens.

(F) In vitro 6mA demethylation assay. Recombinant His-Alkbh4 was purified from bacterial cell extracts (left panel). Irrelevant lanes were omitted from the gel. Human genomic DNA containing 6mA was incubated in buffer supplemented with Fe²⁺ and 2-oxoglutarate cofactors in the absence or presence of His-Alkbh4 or vitamin C (VitC) for the indicated times. 6mA in each sample was detected by dot blot analysis (n = 3 experiments).

(G) Alkbh4 demethylates 6mA in double-stranded (ds) DNA. Single-stranded (ss) or double-stranded DNA oligonucleotide containing unmodified adenine or 6mA was incubated in the absence or presence of His-Alkbh4 for the indicated times. Following the reactions, 6mA in 10 pmol of each DNA sample was detected by dot blot analysis (n = 3 experiments).

See also Figures S1 and S2.

We then examined the activity of murine Alkbh4, which is orthologous to 6mA demethylases in Drosophila and C. elegans (Greer et al., 2015, Zhang et al., 2015). Purified, His-tagged Alkbh4 (His-Alkbh4) catalyzed the demethylation of 6mA in genomic DNA substrate in the presence of Fe²⁺ and 2-oxoglutarate cofactors and was not measurably stimulated by the addition of ascorbic acid (Vitamin C) (Figure 1F). His-Alkbh4 also efficiently demethylated DNA duplex oligonucleotide containing 6mA base modifications within a consensus motif (5′-AGAAGAGGA-3′) identified through methylated DNA immunoprecipitation sequencing (MeDIP-seq) of genomic DNA from mouse ESCs and embryonic day (E) 9.5 embryos (strain C57BL/6x129Sv) using 6mA-specific antisera (Figures 1G and S2). No effect of His-Alkbh4 was detected on single-stranded DNA substrate, even after prolonged incubations (Figure 1G). We conclude that Mettl4 and Alkbh4 catalyze the deposition and oxidative demethylation, respectively, of 6mA in DNA. These enzymes may act complementarily with N6AMT1 and Alkbh1 to control the genomic distribution of 6mA in different cell types (Wu et al., 2016, Xiao et al., 2018, Xie et al., 2018, Zhu et al., 2018).

Craniofacial Dysmorphism and Embryonic Sublethality in Mettl4 Knockout Progeny

Mettl4 and Alkbh4 transcripts are expressed in mouse ESCs, as detected by reverse-transcription-quantitative PCR (RT-qPCR), with increased levels in E9.5 embryos and in some adult tissues (Figure S3A). In ESCs, Mettl4 transcripts are significantly more abundant than those encoding N6AMT1 (Shen et al., 2012) (Figure S3B), implicating Mettl4 in the regulation of 6mA in ESCs. To investigate the physiological role of Mettl4 in development, we intercrossed Mettl4+/− heterozygous mice and recovered Mettl4 KO mice at the expected Mendelian ratio (Figure 2A). Mettl4 KO mice were fertile at 6 weeks of age; however, Mettl4 KO incrosses produced significantly smaller litters than those obtained from strain-matched WT controls (Figure 2B), and ∼6.4% of incross progeny displayed gross anatomical defects, including anophthalmia and craniofacial dysmorphism (n = 5/78 pups from 14 litters) (Figures 2B and 2C). Adult (10–12 weeks old) Mettl4 KO mice were moribund and exhibited splenomegaly (Figures 2D and 2E) with anemia and severe leukopenia (Figure 2F), indicative of aberrant hematopoiesis. These observations implicate Mettl4 in the control of hematopoietic lineage specification and craniofacial patterning.

Figure 2.

Embryonic Sublethality and Craniofacial Dysmorphism in Mettl4 KO Incross Progeny

(A) Genotypic analysis of progeny from intercrossed Mettl4 ± heterozygous (Het) mice, showing expected and observed frequencies for each genotype. n.s., not significant, Chi-square test.

(B) Mating scores of strain-matched wild-type (WT) controls, intercrossed Mettl4 ± heterozygotes, and incrossed Mettl4 KO mice. Black dots indicate the numbers of pups in each litter at the day of birth or recovered by Caesarian section at E18.5. Red lines indicate the mean litter size. p value was calculated by two-tailed t test. ^∗∗p < 0.01. The percentages of progeny that displayed craniofacial or limb dysmorphism are shown below each plot.

(C) Neonatal Mettl4 KO pups from a single litter of incrossed Mettl4 KO parents. White arrow indicates mandibular malformation. Scale bar, 10 mm.

(D) Representative image of spleens recovered from individual adult (10–12 weeks old) WT or Mettl4 KO mice. Rulers show scale in cm. Separate images for WT and KO spleens were juxtaposed at the dashed line.

(E) Bar graph representation of spleen mass divided by total body mass for adult WT and Mettl4 KO mice (n = 6 mice per group). Error bars indicate SD ^∗∗p < 0.01, two-tailed t test.

(F) (Left panel) Peripheral blood count of white cells (WBC) and lymphocytes (left-hand scale) and red blood cells (RBC, right-hand scale) in adult WT and Mettl4 KO mice. (Right panel) Hematocrit (HCT) analysis. Each point corresponds to values from an individual mouse. Brackets indicate normal range and mean values in WT mice. p values were calculated using unpaired, two-tailed t test. ^∗p < 0.05.

See also Figures S1 and S3.

6mA Deposition Triggers Proteolysis of the Sensor Proteins ASXL1 and MPND

To elucidate how 6mA is linked to gene control, we sought to identify proteins that selectively recognize this modified base. Candidate 6mA-interacting proteins in bacteria have been proposed on the basis of their recurrent association in gene operons with, or fusion to, DNA adenine N6-methyltransferases (Aravind and Iyer, 2012). One such protein domain, the HB1, ASXL, restriction endonuclease HTH domain (HARE-HTH), is highly conserved and contained within the N terminus of most metazoan ASXL orthologs, including human ASXL1 (Figure 3A). A second candidate 6mA sensor of prokaryotic origin, the restriction enzyme-adenine methylase-associated (RAMA) domain, is found in the related mammalian DUBs MYSM1 and MPND (Iyer et al., 2016) (Figure 3A). In DNA pull-down assays, the standalone purified HARE-HTH and MPND RAMA domains, but not the MYSM1 RAMA domain, bound selectively to agarose resin coated with duplex DNA oligonucleotide containing 6mA in a native sequence context (5′-AACAGAAGAGG-3′), but not to DNA containing the 5mC or 5-carboxylcytosine (5caC) modifications or to 6mA within a non-native sequence context (5′-GATC-3′) specified by E. coli Dam (Figures 3B and S4). The absence of selective binding to 6mA by the MYSM1 RAMA domain suggests functional divergence of this domain across mammalian paralogs.

Figure 3.

6mA Deposition Triggers Proteolysis of the Sensor Proteins ASXL1 and MPND

(A) Domain structures of ASXL1, showing the positions of the HARE-HTH and PHD domains and the central proline-rich region (PRR), and the MYSM1 and MPND deubiquitinases, with the percentage amino acid identity and similarity between RAMA domains. JAMM/DUB denotes the deubiquitinase catalytic domain.

(B) In vitro DNA pull-down assays. Purified recombinant HARE-HTH and RAMA domains were incubated in the presence of untreated resin or resin coated with duplex DNA containing unmodified adenine (A) or 6mA. After extensive washing, bound protein was resolved by SDS-PAGE and detected by immunoblotting using the indicated antisera (n = 3 experiments).

(C) HEK293T cells expressing Flag-ASXL1.591 and either empty vector, Flag-Mettl4, or Flag-Dam were harvested and proteins in whole-cell extracts resolved by SDS-PAGE and immunoblotting with the indicated antisera (upper panels). Dot blot analysis shows relative 6mA levels in genomic DNA recovered from the same samples (n = 3 experiments).

(D) Cells expressing Flag-RAMA, Flag-Mettl4, or Flag-Dam were harvested and proteins in whole-cell extracts resolved by SDS-PAGE and immunoblotting (upper panels). Dot blot analysis shows relative 6mA levels in genomic DNA recovered from the same samples (n = 2 experiments).

(E) Accumulation of Asxl1 and reduction of H2A-K119Ub and H3K27me3 in Mettl4 KO cells. Whole-cell extracts (upper panels) and purified histones (lower panels) were prepared from spleens isolated from individual WT or Mettl4 KO mice and proteins resolved by SDS-PAGE and immunoblotting (n = 2 experiments).

(F) Accumulation of Mpnd in Mettl4 KO cells. Protein extracts were prepared from spleens isolated from individual WT or Mettl4 KO mice and resolved by SDS-PAGE and immunoblotting (n = 2 experiments).

See also Figures S4 and S5.

We then co-expressed in HEK293T cells Flag-Mettl4 together with a truncated form of ASXL1 (Y591X; hereafter, ASXL1.591), which is readily resolved by SDS-PAGE and is fully competent for activation of PR-DUB (Balasubramani et al., 2015), reasoning that increased deposition of 6mA might enhance the fraction of chromatin-associated ASXL1. Unexpectedly, however, expression of Flag-Mettl4 sharply diminished Flag-ASXL1.591 levels, both in whole-cell extracts and in a chromatin-enriched subcellular fraction (Figures 3C and S5A). Exposure of cells to the proteasome inhibitor MG132 or co-expression of Flag-Alkbh4 rescued Flag-ASXL1.591 in the presence of Flag-Mettl4 (Figures S5B and S5C), while expression of a Mettl4 catalytic site mutant failed to destabilize Flag-ASXL1.591 (Figure S5D). These observations together implicate 6mA deposition in ASXL1 proteolysis. Notably, Flag-ASXL1.591 largely escaped proteolysis in HEK293T cells co-expressing Flag-Dam, despite greater overall levels of 6mA in the genomic DNA of these cells (Figure 3C). The standalone MPND RAMA domain was similarly diminished by co-expression of Flag-Mettl4, but not of Flag-Dam (Figure 3D).

We found that a synthetic reporter protein containing the N-terminal region of ASXL1 (amino acids 1–115) fused upstream of firefly luciferase was also destabilized in the presence of co-expressed Flag-Mettl4 (Figure S5E). Mammalian ASXL paralogs contain a short, lysine-rich segment (amino acids 2–9), adjacent to the HARE-HTH domain, that is contained in the Mettl4-sensitive reporter protein (Figure S5F, top). Deletion of this segment (ΔN9) resulted in a gross accumulation of Flag-ASXL1.591 in HEK293T cells and blocked Mettl4-dependent proteolysis (Figure S5F) but had no detectable effect on the association of ASXL1 with chromatin (Figure S5G). The ΔN9 variant of ASXL1 also strongly stabilized co-expressed Flag-HA-BAP1 (Sowa et al., 2009) (Figure S5H). Thus, the HARE-HTH domain, in conjunction with the adjacent, lysine-rich N terminus of ASXL1, acts as a 6mA-dependent proteolytic targeting module that regulates the stability of ASXL1 and PR-DUB catalytic subunit BAP1.

Prompted by these findings, we examined the impact of Mettl4 on Polycomb repressive histone marks. Expression of Flag-Mettl4 in HEK293T cells stimulated a marked increase in H2A-K119Ub, with a more modest effect on H3K27me3, as determined by immunoblot analysis of purified histones (Figure S5I). Conversely, both H2A-K119Ub and H3K27me3 were depleted in histones purified from in Mettl4 KO spleens, concomitant with increased levels of Asxl1, Bap1 and Mpnd in the same tissue samples (Figures 3E and 3F). Together, these observations implicate 6mA deposition-coupled proteolysis of Asxl1 and Mpnd in the preservation of Polycomb repressive histone marks.

6mA Stimulates Proteolysis of ASXL1 through the E3 Ubiquitin Ligase TRIP12

We reasoned that the enzymes responsible for proteolysis of ASXL1 might be detected in association with the PR-DUB complex. High-sensitivity mass spectrometry studies of BAP1 immunopurified from mammalian cell extracts identified the HECT domain ubiquitin E3 ligase TRIP12/TRIPC as a candidate-interacting protein (Hauri et al., 2016, Kloet et al., 2016). In humans, germline-inactivating mutations in TRIP12 are associated with limb deformation and craniofacial dysmorphism, including a depressed nasal bridge, hypoplastic alae nasi, Cupid’s bow of the upper lip, a small lower jaw, and large ear lobes (Bramswig et al., 2017, Zhang et al., 2017); these features strikingly overlap with those of Bohring-Opitz syndrome patients carrying germline mutant alleles of ASXL1 (Hoischen et al., 2011, Shashi et al., 2016), suggesting overlapping functions of ASXL1 and TRIP12 in craniofacial patterning.

Given these similarities, we examined the interaction of TRIP12 with ASXL1. Flag-ASXL1.591 co-immunoprecipitated with endogenous TRIP12 in HEK293T cell extracts (Figure 4A). Importantly, expression of Flag-Mettl4 stimulated engagement of Flag-ASXL1.591 with TRIP12 while concomitantly reducing the total amount of Flag-ASXL1.591 in the lysate (Figure 4A, lanes 1–3). Expression of GFP-tagged TRIP12 (Gudjonsson et al., 2012) also resulted in diminished levels of Flag-ASXL1.591 (Figure 4A, lanes 4 and 5). Conversely, depletion of TRIP12 using two independent short hairpin RNAs (shRNAs) strongly promoted accumulation of Flag-ASXL1.591 (Figure 4B) and rendered Flag-ASXL1.591 refractory to proteolysis in the presence of co-expressed Flag-Mettl4 (Figure 4C). Together, these observations suggest that deposition of 6mA stimulates the engagement of ASXL1/PR-DUB with TRIP12, promoting proteolysis of ASXL1 (Figure 4D).

Figure 4.

The E3 Ubiquitin Ligase TRIP12 Mediates Proteolysis of ASXL1

(A) Lysates prepared from HEK293T cells expressing the indicated combinations of Flag-ASXL1.591, Flag-Mettl4, or GFP-TRIP12 were immunoprecipitated using TRIP12 antisera or isotype-matched control IgG. Inputs and immunoprecipitated proteins were resolved by SDS-PAGE and immunoblotting with Flag or TRIP12 antisera (n = 2 experiments).

(B) HEK293T cells expressing Flag-ASXL1.591 and either a control shRNA targeting GFP or independent shRNAs (sh1 or sh2) targeting TRIP12 were lysed and protein extracts resolved by SDS-PAGE and immunoblotting. Tubulin, loading control (n = 3 experiments).

(C) Protein extracts prepared from cells expressing Flag-ASXL1.591, in the absence or presence of Flag-Mettl4 or sh1-TRIP12, were resolved by SDS-PAGE and immunoblotting with the indicated antisera (n = 3 experiments).

(D) Model depicting 6mA deposition by Mettl4, recruitment of ASXL1/PR-DUB to 6mA, and engagement of PR-DUB with the E3 ubiquitin ligase TRIP12. These steps are proposed to stimulate ASXL1 proteolysis, inactivating PR-DUB and thereby preserving the Polycomb repressive mark H2A-K119Ub in chromatin.

Mettl4-Dependent Regulation of Asxl1 and H2A-K119Ub at Defined Genic Elements

To better understand the relationship between 6mA, its cognate sensor proteins, and Polycomb repressive marks, we profiled the genomic distribution of 6mA in mESCs using MeDIP-seq. Results from this approach identified 4,922 unique 6mA peaks, which overlapped significantly with 6mA base modifications identified through single-molecule real-time (SMRT) sequencing (Wu et al., 2016) (Figures 5A and 5B; Table S1), supporting the specificity of the 6mA peaks identified through the orthogonal MeDIP approach. A disproportionate fraction of 6mA peaks localized in intergenic regions, in line with previous observations (Figure 5A) (Wu et al., 2016). In genic regions, 6mA tag density decreased approaching the transcription start site in gene promoters, while increasing progressively toward the transcription end site (Figure 5C). In gene bodies and flanking the TSS, the highest 6mA density was detected in genes expressed at relatively low levels in WT ESCs (Figure 5C).

Figure 5.

Ectopic Asxl1 and Mpnd Correspond with Loss of Polycomb Silencing in Mettl4 KO ESCs

(A) Pie chart illustrating the distribution of called 6mA peaks (n = 4,922) across promoter (−2 kb to TSS), TSS downstream (0–2 kb downstream of TSS), 5′ and 3′ UTR, coding exon, intron, and intergenic regions. Red numbers indicate the fold enrichment or depletion of 6mA at each feature relative to a random distribution.

(B) Venn diagram showing overlap of 6mA peaks identified by MeDIP and called 6mA bases identified by PacBio SMRT sequencing analysis (Wu et al., 2016). p value, Fisher’s exact test.

(C) Normalized 6mA tag density plotted 1 kb upstream of the TSS, across the first 3 kb of a metagene, and 1 kb downstream of the TES for all genes (black), the top 10% of highly expressed genes (blue), and the bottom 10% of least expressed genes in WT ESCs (red).

(D) Venn diagrams showing the number and relative distribution of Asxl1, Bap1, O-GlcNAc, Mpnd, and H2A-K119Ub (H2A-Ub) peaks in WT and Mettl4 KO ESCs.

(E) Normalized tag density of ectopic Asxl1 peaks induced in Mettl4 KO ESCs (left) with heatmap representation of peaks ranked-ordered by the mean signal (right), each plotted across a window centered on the TSS.

(F) Genome browser view showing profiles of 6mA or isotype-matched control IgG in WT ESCs (top) and Asxl1, Mpnd, H2A-Ub, Bap1, and O-GlcNAc in WT and Mettl4 KO ESCs at two representative loci. Shaded vertical bars highlight regions containing ectopic Asxl1 or Mpnd and depletion of H2A-Ub in Mettl4 KO ESCs. RefSeq exon structures (blue) for each annotated gene are shown at the bottom.

(G) ChIP-qPCR analysis of the Rpl13 and Dvl3 genes in WT and Mettl4 KO ESCs. Antisera specific for Asxl1, H2A-K119Ub002C or Bap1 were used for chromatin immunoprecipitation. The mean fold enrichments normalized to isotype-matched IgG control are shown for each condition. Error bars indicate SEM (n = 2 experiments).

(H) RT-qPCR analysis of Rpl13 and Dvl3 transcript levels in WT and Mettl4 KO ESCs. The mean value of WT control samples is set as 1. Error bars indicate SEM (n = 2 experiments).

(I) Volcano plot presentation of transcript levels for genes expressed in WT and Mettl4 KO ESCs as determined by RNA-seq. Genes strongly up- or downregulated (FC > 2.0, FDR-adjusted p < 0.05) in Mettl4 KO cells are identified and indicated in red or green, respectively.

(J) Gene ontology analysis of differentially expressed genes induced in Mettl4 KO cells showing involvement in embryonic development and tissue patterning. The yellow vertical line indicates the threshold for significance.

See also Tables S1 and S2.

Inactivation of Mettl4 generated a ∼5-fold increase in the number of called Asxl1 peaks (Figures 5D and 5E; Table S2) and a 41.3% increase in Bap1 peaks (Figure 5D) identified using stringent cutoff criteria (FDR adjusted p < 0.05), in agreement with Asxl1-mediated stabilization of Bap1. In contrast, the number of called peaks and the distribution of O-linked N-acetylglucosamine (O-GlcNAc), which is attached to chromatin-associated target proteins by PR-DUB component Ogt1 (Hauri et al., 2016, Kloet et al., 2016), remained largely unchanged (Figure 5D). Thus, Ogt1 may be recruited to DNA independently of 6mA, potentially through other DNA-binding components of PR-DUB (Hauri et al., 2016, Kloet et al., 2016). Inactivation of Mettl4 also produced a ∼80% increase in the number of called Mpnd peaks relative to WT (Figure 5D), further supporting a role for 6mA in destabilization of both classes of H2A DUBs. Accumulation of ectopic Asxl1 and Mpnd in Mettl4 KO ESCs corresponded with a 51.6% reduction in H2A-K119Ub peaks, as determined by ChIP-seq analysis (Figures 5D and 5F) with ChIP-qPCR validation at the Rpl13 and Dvl3 loci (Figure 5G), and expression of both genes was accordingly induced in Mettl4 KO ESCs, as determined by RT-qPCR analysis (Figure 5H).

To assess the global impact of Mettl4 inactivation on gene expression, we performed RNA-seq transcriptome profiling of WT and Mettl4 KO ESCs and identified 198 genes that displayed significantly altered expression, with 135 upregulated and 63 downregulated genes (FC > 2.0, FDR-adjusted p < 0.05) (Figure 5I). Gene ontology analysis of differentially expressed genes revealed significant involvement in tissue specification and organ morphogenesis (Figure 5J), supporting a role for 6mA-dependent gene control in the regulation of embryonic tissue patterning.

Discussion

Here we reveal the molecular infrastructure of a mammalian-DNA 6mA methylation, erasure, and sensing system and elucidate its function in gene control. The system minimally encompasses the Mettl4 methyltransferase and Alkbh4 dioxygenase, which catalyze 6mA deposition and erasure, respectively, and two structurally distinct classes of 6mA sensor domains attached to unrelated deubiquitinases, which nonetheless remove monoubiquitin from the same substrate: H2A-K119Ub, a Polycomb repressive histone mark. Strikingly, deposition of 6mA by Mettl4 triggers the proteolytic destruction of each sensor protein, ASXL1 and MPND, thereby preserving H2A-K119Ub levels. In contrast, both sensors are largely refractory to 6mA deposited in a non-native sequence context through heterologous expression of the bacterial 6mA methyltransferase Dam in mammalian cells. Altogether, these observations reveal a native system architecture that operates through a 6mA-deposition-coupled sensor destruction mechanism to preserve Polycomb gene silencing.

The methylome defined by 6mA shares several important similarities with 5mC, including reversibility through oxidative demethylation and a system output that impacts gene expression through the modulation of histone marks (He et al., 2011, Kweon et al., 2017, Schübeler, 2015, Xiao et al., 2018, Xie et al., 2018). 5mC differentially repels or tethers methylation-selective transcription factors, effector proteins, and associated chromatin-modifying enzymes such as histone deacetylases to DNA, silencing gene transcription (Schübeler, 2015, Yin et al., 2017). 6mA, in contrast, operates through detection-triggered proteolysis and depletion of its cognate sensors. Iterative cycles of sensor engagement and destruction may greatly amplify the 6mA output space, rationalizing the significant impact of this base modification on gene control notwithstanding an exponentially (∼1,000-fold) lower abundance relative to 5mC (Schübeler, 2015, Smith et al., 2012). Importantly, 6mA may regulate sensor protein expression or stability through additional transcriptional or post-transcriptional mechanisms. It will also be important to determine whether 6mA can regulate chromatin states and gene expression independently of its destabilizing effects on ASXL1 and MPND.

Global depletion of 6mA has been reported in aggressive tumors (Xiao et al., 2018), raising the possibility that accumulation of ASXL1 and hyperactivation of PR-DUB drive deregulated gene expression, oncogenesis, and metastatic progression. Heterozygous nonsense alleles of ASXL1, which have been proposed to encode truncated gain-of-function proteins (Asada et al., 2018, Balasubramani et al., 2015, Guo et al., 2018), represent a prevalent class of ASXL1 mutations in leukemic cells (Balasubramani et al., 2015, Boultwood et al., 2010, Dey et al., 2012). Against this backdrop, it will be interesting to determine whether inactivation of ALKBH1 or ALKBH4 can elevate genomic 6mA levels and thus drive the proteolytic destruction of oncogenic ASXL1 truncations. Future studies may address the feasibility and impact of 6mA modulation as an anticancer therapeutic strategy.

STAR★Methods

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse monoclonal anti FLAG M2	Sigma-Aldrich	Cat#F3165; RRID: AB_259529
Rabbit polyclonal anti Histone H3K27me3	Active Motif	Cat#39155; RRID: AB_2561020
Rabbit polyclonal anti 6mA	Active Motif	Cat#61495
Rabbit polyclonal anti 5mC	Active Motif	Cat#61255
Mouse monoclonal anti luciferase	Santa Cruz Biotechnology	Cat#sc-74548; RRID: AB_1125118
Mouse monoclonal anti ASXL1	Santa Cruz Biotechnology	Cat#sc-293204;
Mouse monoclonal anti HA	Santa Cruz Biotechnology	Cat#sc-57592; RRID: AB_629568
Rabbit monoclonal anti Histone H2A	ThermoFisher	Cat#700158; RRID: AB_2532288
Rabbit monoclonal anti Histone H2A-Ub	Cell Signaling Technology	Cat#8240; RRID: AB_10891618
Rabbit polyclonal anti TRIP12	ThermoFisher	Cat#PA5-38016; RRID: AB_2554620
Mouse monoclonal anti O-GlcNAc	Santa Cruz Biotechnology	Cat#sc-59623; RRID: AB_784962
Mouse monoclonal anti S-Tag	MilliporeSigma	Cat#71549-3; RRID: AB_10806301
Rabbit polyclonal anti beta-tubulin	Abcam	Cat#ab6046; RRID: AB_2210370
Mouse monoclonal anti beta-actin	Santa Cruz Biotechnology	Cat#sc-47778; RRID: AB_2714189
Mouse monoclonal anti His tag	ThermoFisher	Cat# MA1-135; RRID: AB_2536841
Mouse monoclonal anti GAPDH	Cell Signaling Technology	Cat#97166; RRID: AB_2756824
Goat anti mouse IgG-HRP	Santa Cruz Biotechnology	Cat# sc-2005; RRID: AB_631736
Goat anti rabbit IgG-HRP	Santa Cruz Biotechnology	Cat#sc-2004; RRID: AB_631746
Bacterial and Virus Strains
Rosetta 2(DE3)pLysS Competent Cells	MilliporeSigma	Cat#71403
Chemicals, Peptides, and Recombinant Proteins
MG132	MilliporeSigma	Cat#M8699-1MG
Dam methyltransferase	New England Biolabs	Cat#M0222S
Power SYBR Green PCR Master Mix	ThermoFisher	Cat# 4367659
DNA Degradase Plus	Zymo Research	Cat#E2020
Mouse LIF	MilliporeSigma	Cat#ESG1106
RNase A	ThermoFisher	Cat#12091021
RNase H	New England Biolabs	Cat#M0297S
Mpnd-RAMA-His	This study	N/A
MYSM1-RAMA-His	This study	N/A
HARE-HTH-His	This study	N/A
Alkbh4-His	This study	N/A
Nuclease P1	MilliporeSigma	Cat#N8630-1VL
BioT	Bioland Scientific	Cat#B01S
Ampure XP resin	BeckmanCoulter	Cat#A63880
Protein A/G agarose	Santa Cruz Biotechnology	Cat#sc-2003; RRID: AB_10201400
Critical Commercial Assays
Quick Genotyping DNA kit	Bioland Scientific	Cat#GT01-01
LookOut Mycoplasma PCR Detection kit	MilliporeSigma	Cat#MP0035-1KT
DNEasy Blood and Tissue kit	QIAGEN	Cat#69506
Quick-RNA miniprep kit	Zymo Research	Cat#R1054
Maxima First Strand cDNA Synthesis kit	ThermoFisher	Cat#K1642
NEBNExt rRNA Depletion kit (Human/Mouse/Rat)	New England Biolabs	Cat#E6310S
MAGnify ChIP kit	ThermoFisher	Cat#492024
KAPA HyperPrep kit	Roche	Cat#07962312001
KAPA Stranded mRNA Sequencing kit	Roche	Cat#07962282001
Deposited Data
6mA SMRT sequencing	Wu et al., 2016	GEO: GSE71940
mESC RNA-seq	Shen et al., 2012	GEO: GSM723776
ChIP, MeDIP, and RNA sequencing	This study	GEO: GSE105006
Mendeley dataset	This study	http://doi.org/10.17632/rfmf3tb5z8.1
Experimental Models: Cell Lines
Mouse Mettl4 KO ESCs	This study	N/A
Human HEK293T	ATCC	Cat#CRL-11268; RRID: CVCL_1926
Experimental Models: Organisms/Strains
Mouse: Mettl4em6Dfel: 129/Sv ∗ C57BL/6N	This study	N/A
Oligonucleotides
Primers for qPCR	Table S3	N/A
6mA consensus Forward: 5′-GATGCAAGCATCAGC AAC6mAGAAG6mAGGATCTCAGGTGCAGCGC-3′	This study	N/A
Unmethylated consensus Forward: 5′-GATGCAAGCATC AGCAACAGAAGAGGATCTCAGGTGCAGCGC-3′	This study	N/A
Consensus Reverse: 5′-Biotin-GCGCTGCACCTGAGAT CCTCTTCTGTTGCTGATGCTTGCATC-3′	This study	N/A
Dam-methylated substrate Forward: 5′-ATAACTACACT ATCATGCGCTGACCCACAACATCCGAAGGGATCCAG GTTAAGTAATATAATTGCGCCTTAT-3′	This study	N/A
Dam-methylated substrate Reverse: 5′-Biotin-ATAAGG CGCAATTATATTACTTAACCTGGATCCCTTCGGATGTT GTGGGTCAGCGCATGATAGTGTAGTTAT-3′	This study	N/A
5mC control for pull-down: 5′-biotin-GGA CCACCG GTG GTC C-3′	This study	N/A
5caC control for pull-down Forward: 5′-TACACTATCAT GXGCTGACCCACAACATCCGA-3′, where X is 5caC	This study	N/A
5caC control for pull-down Reverse: 5′-biotin-TCGGATG TTGTGGGTCAGCGCATGATAGTGTA-3′	This study	N/A
Recombinant DNA
Plasmid: pcDNA3.1-HARE-HTH-luciferase	This study	Addgene plasmid 106427
Plasmid: p3xFlag-Alkbh4	This study	Addgene plasmid 105240
Plasmid: p3xFlag-Mettl4	This study	Addgene plasmid 86665
Plasmid: pET30a-HARE-HTH	This study	Addgene plasmid 107164
Plasmid: pET30a-Mpnd-RAMA	This study	Addgene plasmid 112226
Plasmid: pET30a-MYSM1-RAMA	This study	N/A
Plasmid: p3XFlag-Dam	This study	Addgene plasmid 121909
Plasmid: p3xFlag-RAMA	This study	N/A
Plasmid: pCMV6-XL4 ASXL1 (p.Y591X) 3x Flag	Balasubramani et al., 2015	Addgene plasmid 74261
Plasmid: Flag-HA-BAP1	Sowa et al., 2009	Addgene plasmid 22539
Plasmid: pX330-U6-Chimeric_BB-CBh-hSpCas9	Cong et al., 2013	Addgene plasmid 42230
Plasmid: pAc-GFP-TRIP12	Lukas lab; Gudjonsson et al., 2012	N/A
Plasmid: pLKO.1-sh1-Trip12	MilliporeSigma	Cat#TRCN0000273207
Plasmid: pLKO.1-sh2-Trip12	MilliporeSigma	Cat#TRCN0000022376
Plasmid: pLKO.1-shGFP	MilliporeSigma	Cat#SHC002
Plasmid: Mettl4 CRISPR/Cas9 KO	Santa Cruz Biotechnology	Cat#sc-429415
Plasmid: Mettl4 HDR	Santa Cruz Biotechnology	Cat#sc-429415-HDR
Software and Algorithms
MACS (v2.1.1)	Zhang et al., 2008	http://liulab.dfci.harvard.edu/MACS/
Trim Galore (v0.3.5)	Babraham Bioinformatics	https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/
Bowtie (v1.2.2)	Langmead et al., 2009	http://bowtie-bio.sourceforge.net/index.shtml
Cistrome CEAS	Shin et al., 2009	https://anaconda.org/bioconda/cistrome-ceas
NGSPLOT (v2.63)	Shen et al., 2014	https://github.com/shenlab-sinai/ngsplot
Bedtools (v2.27.0)	Quinlan and Hall, 2010	https://bedtools.readthedocs.io/en/latest/
Cufflinks (v2.2.1)	Trapnell et al., 2010	http://cole-trapnell-lab.github.io/cufflinks/
DAVID	Huang da et al., 2009	https://david.ncifcrf.gov/

Contact for Reagent and Resource Sharing

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Douglas E. Feldman (defeldma@usc.edu).

Experimental Model and Subject Details

Generation of Mettl4 KO mice

Founder mice were generated at Applied StemCell, Inc. (Milpitas, CA, USA) at an AAALAC International accredited facility (Protocol 000587). All animals were treated in accordance with guidelines established by the National Institutes of Health using a protocol approved by the USC Institutional Animal Care and Use Committee (Protocol 20520).

To generate the CRISPR/Cas9 guide RNA (gRNA) expression vectors targeting murine Mettl4, gRNA candidates were selected based on proximity to regions flanking exon 6 and favorable gRNA off-target profiles. Duplex oligonucleotides were ligated into BbsI-digested, phosphatase-treated pSpCas9(BB) (pX330; Addgene plasmid 42230) (Cong et al., 2013) encoding human codon-optimized SpCas9 and a chimeric single guide (sg)RNA. All constructs were confirmed by DNA sequencing.

For gene targeting, a mixture of Cas9 mRNA and sgRNAs targeting regions flanking exon 6 of murine Mettl4 was generated by in vitro transcription of both pX330-Mettl4 sgRNA constructs and microinjected into the cytoplasm of zygote-stage embryos, which were harvested from 129/Sv females mated with C57BL/6N males. The injected embryos were then implanted in CD-1 foster mice. Pups were screened for deletions in Mettl4 using PCR primers flanking the targeted region. All genomic deletions were confirmed by TOPO-TA cloning of the PCR product and sequencing analysis.

From a single round of microinjection, five mice were born and four contained deletions of exon 6. Three founder mice were selected for mouse line expansion and backcrossed to C57BL/6N at least three times prior to experimental analysis. For genotyping, ear tissues from founders and offspring mice were collected and DNA extraction was performed individually using the Quick Genotyping DNA kit (Bioland Scientific). PCR reactions were performed in 25 μL total volume using the following parameters: 95°C, 2 min; 35 cycles of [95°C, 15 s; 60°C, 15 s; 72°C, 30 s], 72°C, 4 min; 4°C, hold.

Cell line generation

To generate Mettl4 KO ESCs, mice heterozygous for the exon 6 deletion allele were intercrossed and blastocysts were harvested at E3.5, transferred onto MEF-coated 96-well plates and cultured with 2i/LIF. Additional KO lines were generated by transfecting Mettl4 heterozygous ESC with Mettl4 CRISPR/Cas9 KO Plasmid and HDR vector (Santa Cruz Biotechnology) containing a puromycin resistance cassette. Established ESC clones were genotyped by PCR and confirmed by immunoblot. Male wild-type or Mettl4 KO cells were then expanded in cell culture for subsequent studies. HEK293T cells, a human female embryonic kidney cell line stably expressing the SV40 large T antigen, were purchased from ATCC (CRL-3216). Both HEK293T and ESC lines were routinely authenticated by morphology check with microscope.

Cell Culture

Cells were passaged every 2-3 days and maintained at 37°C and 5% CO₂ in a humidified incubator. Mettl4 KO and matched wild-type ESCs were cultured on gelatin-coated plates in high-glucose DMEM (ThermoFisher Scientific) containing 15% FBS (Clontech), 1x non-essential amino acids, 2 mM L-glutamine, 100 μM beta-mercaptoethanol, 1000U/mL leukemia inhibitory factor (MilliporeSigma) and 1x antibiotic-antimycotic (ThermoFisher).

HEK293T cells were cultured in high-glucose DMEM supplemented with 10% FBS, 1x non-essential amino acids, and 2 mM L-glutamine, and 1x antibiotic-antimycotic. All cells were tested for absence of mycoplasma contamination using the LookOut Mycoplasma PCR Detection kit (MilliporeSigma).

Sample collection

Whole mouse spleens were collected from adult mice (10-12 weeks old) by manual dissection and immediately flash-frozen in liquid nitrogen. Mouse whole blood from was collected n EDTA tubes from 3-4 randomly selected adult female mice of each genotype.

Method Details

Transfection

Cells cultured in 10 cm dishes or 6-well plates were transfected with purified plasmid DNA (8 μg or 1.5 μg, respectively) using BioT transfection regent (Bioland Scientific) according to the manufacturer’s protocol.

Dot blot analysis

Human genomic DNA was isolated from cultured HEK293T cells using the DNEasy Blood and Tissue kit (QIAGEN). Importantly, to avoid contamination with rRNA that contains m6A, samples treated overnight with RNase A (20 μg/mL) and RNase H (5 U). DNA was denatured in 0.4 M NaOH, 10mM EDTA at 95°C for 10 min, then neutralized by the addition of an equal volume of cold 2.0 M ammonium acetate (pH 7.0). Denatured DNA samples were serially diluted in TE buffer and a total of 3 μL of the diluted samples were spotted onto a nitrocellulose membrane (Biad) followed by drying under UV crosslinking for 5 min. Membranes were blocked in 2% milk and incubated with a 1:1000 dilution of 6mA antisera at 4°C overnight. After washing three times with TBST (20 mM Tris pH 7.5, 150 mM NaCl, 0.1% Tween-20), the membrane was incubated with the HRP-conjugated anti-rabbit IgG secondary antibody. The membrane was washed with TBST three times prior to incubation in SuperSignal West Pico (ThermoFisher) chemiluminescent substrate and imaging.

Purification of recombinant proteins

Plasmids encoding the His-Alkbh4, His-HARE-HTH and His-RAMA domains were transformed into E. coli strain Rosetta 2 (DE3) (MilliporeSigma). Bacterial cultures were grown to an A₆₀₀ of 0.4-0.5, then induced with 0.1 mM IPTG for 4 h at 30°C. Bacterial pellets were resuspended in ice-cold PBS supplemented with 100mM NaCl, 1mM Imidazole, 0.2% NP-40, complete protease inhibitor cocktail, 5 mM sodium fluoride, 1 mM sodium orthovanadate, lysozyme (0.2 mg/mL) and DNase I (10 μg/mL), then sonicated in 5-6 bursts of 15 s with 1 min cooling in between bursts. Soluble proteins (5 ml) were incubated with 250 μL Profinity Nickel-charged resin (Bio-Rad) for 1 h at 4°C with gentle rotation. Resin was washed twice in lysis buffer and eluted for 30 min in elution buffer (1XPBS, 100mM NaCl, 250mM Imidazole, 5% glycerol), then pooled and dialyzed overnight at 4°C into 1XTBS containing 5% glycerol. Samples were aliquoted and stored at −80°C.

DNA demethylation assays

Genomic DNA prepared from HEK293T cells (1.0 μg) or oligonucleotide DNA substrate (25 μM) was incubated with 2.5 μg of His-Alkbh4 in demethylation buffer (50 mM Tris pH 7.6, 150 mM NaCl, 5% glycerol, 0.1 mM DTT, 1 mM 2-oxoglutarate, 200 μM (NH)₄Fe(SO)₄), in the absence or presence of 1 mM ascorbic acid (Vitamin C), for the indicated times at 37°C.

Oligonucleotide DNA pull-down assays

HPLC-purified DNA oligonucleotides

(5′- GATGCAAGCATCAGCAACAGAAGAGGATCTCAGGTGCAGCGC-3′, where A is 6mA or A) containing a 5′ biotin moiety attached by a 15-atom tetra-ethylene glycol (TEG) spacer arm were purchased from TriLink and IDT, respectively, dissolved to 0.15 mg/mL in annealing buffer (20 mM Tris pH 7.5, 50 mM NaCl, 1mM EDTA), mixed at a 1:1 molar ratio with unmodified reverse complement oligo, then denatured at 95°C for 5 min and allowed to cool gradually to room temperature. To generate Dam-methylated oligonucleotides, HPLC-purified oligonucleotide (5′-ATAACTACACTATCATGCGCTGACCCACAACATCCGAAGGGATCCAGGTTAAGTAATATAATTGCGCCTTAT-3′) containing a 5′ biotin moiety was dissolved to 0.15 mg/mL in annealing buffer (20 mM Tris pH 7.5, 50 mM NaCl, 1mM EDTA), mixed at a 1:1 molar ratio with unmodified reverse complement oligo, then incubated at 95°C for 5 min and allowed to cool gradually to room temperature. Duplex DNA was incubated in the presence 5 U of recombinant Dam (New England Biolabs) for 18 h at 37°C, then purified by spin column.

For 5mC-modified control DNA duplexes, HPLC-purified palindromic DNA oligonucleotide (5′-GGA CCACCG GTG GTC C −3′) containing a 5′ biotin moiety attached by a

TEG spacer arm (IDT) was dissolved to 0.15 mg/mL in annealing buffer, incubated at 95°C for 5 min and allowed to cool gradually to room temperature.

For 5caC- modified DNA duplexes, HPLC-purified 32-mer oligonucleotides (5′-

TACACTATCATGXGCTGACCCACAACATCCGA-3′, where X is 5caC) (TriLink Technologies) was mixed with an equimolar amount of complementary oligonucleotide containing

a 5′ biotin-TEG moiety, then dissolved to 0.15 mg/mL in annealing buffer, incubated at 95°C for 5 min and allowed to cool gradually to room temperature.

Coupling of DNA to resin was performed essentially as described (Spruijt et al., 2013). Briefly, 1.5 μg (10 μl) of duplex oligo was coupled to 100 μL of streptavidin-agarose resin (Gold Biotechnology) in 100 μL DNA binding buffer (10mM Tris-HCl pH 8.0, 1M NaCl, 1mM EDTA pH 8, 0.05% NP40) for 4 h at 25°C with gentle rocking. Coupling of the DNA to the beads was verified by agarose gel electrophoresis. Resin was washed twice in 10 mM HEPES pH 7.5, 100 mM NaCl, 2.5% glycerol, and aliquots were stored at −20°C for subsequent use. Purified His-RAMA or His-HARE-HTH (1.0 μg) was incubated with 30 μL of uncoated control resin or DNA-coated resin in 100 μL protein binding (PB) buffer (50 mM Tris pH 8.0, 150 mM NaCl, 1.0 mM DTT, 0.25% NP-40) in the presence of 4 μg poly-dI:dC for 2 h at 4°C. Resin was recovered by centrifugation at 7,000 g for 1 min and washed three times with PB buffer, then boiled for 5 min at 95°C in 1X Laemmle Sample Buffer and resolved by SDS-PAGE for immunoblot analysis.

RT-qPCR

Total RNA was isolated from cultured ES cells or homogenized embryos using the Quick-RNA miniprep kit (Zymo Research) and cDNA prepared using the Maxima First Strand cDNA Synthesis kit (ThermoFisher) according to the manufacturer’s protocol. qPCR reactions were prepared with the Power SYBR Green PCR Master Mix (ThermoFisher) and cycling performed using a ViiA7 Real-Time PCR system (Applied Biosystems). Primers used for qPCR are listed in Table S3. Expression levels were normalized to Gapdh or β-actin in the same samples.

LC-MS/MS analysis

Genomic DNA (250 ng) purified from isogenic wild-type and Mettl4 KO ESCs or adult spleens was treated with DNA Degradase Plus (Zymo Research) according to the manufacturer’s protocol. For m6A quantitation, total RNA was extracted using the Quick-RNA miniprep kit according to the manufacturer’s instructions. rRNA was removed using the NEBNExt rRNA Depletion kit (New England Biolabs). 200 ng of purified RNA was digested by P1 nuclease (2 units) in 25 μl buffer containing 25 mM NaCl and 1.0 mM of ZnCl2 for 2 h at 37 °C. Samples (20 μl) were diluted 2-fold with nuclease-free water and mixed with aqueous formic acid to 0.1% v/v, final concentration of nucleosides (50 ng/μl), and aliquots injected into a reverse phase UPLC column (Eclipse C18 2.1 × 50 mm, 1.8 um particle size, Agilent) equilibrated with buffer A (0.1% formic acid) and eluted (200 μl/min) with an increasing concentration of buffer B (methanol) with mass spectrometry detection using an Agilent 6460 QQQ triple-quadrupole tandem mass spectrometer (MS/MS) set to multiple reaction monitoring (MRM) in positive electrospray ionization mode. Replicate samples were analyzed using an Agilent Accurate Mass 6230 TOF coupled with an Agilent 1290 Liquid Chromatography system. To measure 6mA, the areas of each LC-MS measurement were determined by extraction of the accurate parent-daughter ion mass transitions of 266.1-150.0 for 6mA and 252.1-136.0 for dA from the total ion current. Quantification of the ratio 6mdA/dA was performed using the calibration curves obtained from nucleoside standards and the linear fits of the determined area ratios over the known amount were used to calculate the nucleoside contents. Quantitation of m6A was performed based on nucleoside-to-base ion transitions (282.1-150.0 for m6A and 268.1-136.0 for A) and a standard curve of pure adenine nucleoside.

ChIP, MeDIP and RNA library preparation

For ChIP, strain-matched wild-type or Mettl4 KO mouse ES cells (C57BL/6N x 129/Sv background) were fixed in 1% paraformaldehyde for 10 min at 37°C, followed by quenching with 125 mM glycine for 5 min. Cells were lysed and processed using the MAGnify ChIP kit ThermoFisher). Lysates were sonicated using a sonic dismembrator (FisherScientific) at power setting 5 in 10 s pulses followed by 1 min on ice, for a total of 3 min sonication. Nucleoprotein complexes were immunoprecipitated the indicated antisera (2.5 μg) equivalent amounts of isotype-matched control IgG per sample. For MeDIP experiments, genomic DNA was isolated from ES cells using the DNeasy Blood & Tissue Kit and sonicated for a total of 3 min as described above for ChIP samples. Samples were processed using the MAGnify ChiP kit according to the manufacturer’s protocol. Library preparations were performed using the KAPA HyperPrep kit (Roche) with 12 cycles of PCR amplification. For RNA-seq, total RNA was extracted from mouse ES cells using the Zymo Quick-RNA miniprep kit and rRNA was subtracted using the NEBNext rRNA Depletion kit. Libraries were prepared using the KAPA Stranded mRNA Sequencing kit. Amplified libraries were pooled, purified twice with 0.9X Ampure XP beads (BeckmanCoulter) and analyzed on an Illumina HiSeq2000 or NextSeq500.

Sequencing data processing

For ChIP-seq, MeDIP-seq and RNA-seq, raw reads were trimmed to remove adapters and low-quality reads using Trim Galore (v0.3.5). Bowtie (v1.2.2) (Gudjonsson et al., 2012, Langmead et al., 2009) was used to obtain reads mapped uniquely to the mouse genome (mm10). ChIP and DIP peak candidates were identified with MACS (v2.1.1) (Zhang et al., 2008) using input as the control dataset. To remove nonspecific signals, IgG samples were processed similarly and their normalized read density (RPKM) values were subtracted from modification-specific peaks. For MeDIP, the empirical FDR was estimated by exchanging the MeDIP and IgG control samples and identifying peaks in the control sample using the same set of parameters used for the MeDIP sample. Called peaks were annotated with Cistrome CEAS (Shin et al., 2009). Normalized read density plots centered on annotated elements were generated using NGSPLOT (v2.63) (Shen et al., 2014). Peak overlap statistics were calculated using Bedtools (v2.27.0) (Quinlan and Hall, 2010). For RNA-seq, Cufflinks (v2.2.1) (Trapnell et al., 2010) was used to assemble transcripts, estimate their abundances and test for differential transcript expression. All GO analyses used DAVID (Huang da et al., 2009).

MT-A70 sequence analysis

Sequences were clustered using CLANS (Frickey and Lupas, 2004) (P-value cutoff 0.001) with sequences from PDB and UniProt. For UniProt, a subset of protein IDs (EC:2.1.1.62 (RNA MTases) and EC:2.1.1.72 (DNA MTases)) were downloaded from ExPASy (https://enzyme.expasy.org/). For PDB, sequences were chosen based on keyword search. For Motif identification, human METTL4 sequence was tested against pfam05063: MT-A70 using the NCBI PSSM Viewer.

Quantification and Statistical Analysis

No statistical methods or criteria were used to estimate sample size or to include/exclude samples. Microsoft Excel was used to process data. Statistical details such as N and error calculations are provided in the Legends.

Data and Software Availability

The Gene Expression Omnibus (GEO) accession number for the ChIP-seq, MeDIP-seq and RNA-seq data reported in this paper is GSE105006.

Published: April 11, 2019