Several recent studies have implicated N6-mA in epigenetic silencing, especially at long interspersed element 1 (LINE-1) endogenous transposons in mouse embryonic stem (mES) cells, in brain tissue under environmental stress, and in human lymphoblastoid cells and tumorigenesis1,7,8,9. Although these findings underscore the importance of N6-mA in mammalian biology and human disease, the underlying mechanisms of N6-mA-mediated gene silencing have thus far remained unclear.

N6-mA is enriched at AT-rich regions in mammalian genomes1, especially in areas that interact with the nuclear architecture, such as matrix/scaffold attachment regions (M/SARs). M/SARs are liable to undergo DNA unpairing under torsional stress (known as SIDD or as base unpairing regions (BURs))2,3,4 during replication, transcription, and recombination3,5,6. A study that used genome-wide permanganate and S1 endonuclease mapping (single-stranded (ss)DNA-seq10) demonstrated a strong correlation between SIDD regions identified experimentally and those predicted by computational approaches11. SIDD regions in M/SARs are crucial for organizing large chromatin domains, such as establishing and maintaining heterochromatin–euchromatin boundaries12,13,14 and facilitating long-range interactions15.

SATB1, a well-known SIDD-regulating protein, is mainly expressed in developing T cells16, epidermis17, and trophoblast stem cells (TSCs)18,19. SATB1 binds to SIDD and then stabilizes the DNA double helix2,20, thereby establishing and maintaining large-scale euchromatin and heterochromatin domains12,13,14. For example, SATB1 binds directly to key enhancers in pro- and pre-T cells, thereby activating gene expression in these cells while repressing the mature T cell fate21. During early embryogenesis, SATB1 and SATB2 form an intricate network that regulates ES cell differentiation22. In addition, there is emerging evidence that Satb1 is markedly upregulated during extraembryonic tissue development and promotes extraembryonic fates such as trophectoderm and primitive endoderm lineages18,19,23.

In this work, we show that N6-mA contributes to TSC development and differentiation by antagonizing SATB1 function at SIDD. Moreover, ALKBH1, the DNA demethylase of N6-mA, highly prefers SIDD sequences as substrates24, which further strengthens the connection between N6-mA and DNA secondary structures during cell fate transitions.

N 6-mA at SIDD during TSC development

As N6-mA is present at low levels (6–7 parts per million (ppm)) in mES cells1, we searched for conditions under which N6-mA levels were upregulated. Notably, N6-mA levels are positively correlated with the in vivo developmental potential (pluripotency) of these conditions as determined by tetraploid complementation (4N). N6-mA is greatly diminished under traditional 2i conditions (cultures with ERK and GSK3b inhibitors, which are 4N negative) but retained under some alternative 2i conditions (4N positive) (Extended Data Fig. 1a), which may explain discrepancies in the literature25.

As previous studies showed that N6-mA demethylase Alkbh1-deficient mice developed trophectodermal defects26, we next investigated the role of N6-mA in the development and differentiation of TSCs. We leveraged a Cdx2-inducible expression system27 (iCdx2) to model TSC development in cell culture, which manifests a well-synchronized and efficient cell fate transition process (Fig. 1a). Consistent with previous work28, our RNA sequencing (RNA-seq) analysis showed that iCdx2 cells go through first a transition state, then a TSC-like cell (TSC-LC) state, before undergoing trophectoderm lineage differentiation (Extended Data Fig. 1b). The N6-mA level increased transiently at a time window that coincides with the emergence of TSC-LCs, and then tapered off during trophectoderm lineage differentiation (Fig. 1a). Accordingly, the expression of Alkbh1 inversely correlated with N6-mA levels during this cell fate transition (Extended Data Fig. 1b).

Fig. 1: N6-mA is upregulated at SIDD regions during TSC development.
figure 1

a, Top, schematic of the iCdx2 ES cell-to-TSC fate transition system. Bottom, DNA dot blotting of N6-mA at different time points during differentiation. D, day. Experiments were repeated independently three times with similar results. For blot source data, see Supplementary Fig. 1. b, Average N6-mA reads per genomic content (RPGC) at different time points, centred at differentially increased N6-mA peaks (day 5 versus day 0). c, Mass spectrometry (MS) analysis of N6-mA. Left, schematic of matrix attachment region (MAR) DNA extraction. Middle, N6-mA levels from different chromatin fractions and mock control (buffer, nucleases, and proteinase K). One-way ANOVA (P < 0.0001) followed by Tukey’s multiple comparisons test. Right, MAR DNA N6-mA compared to total DNA at time points during differentiation. Two-way ANOVA (day, P = 0.0005; fraction, P < 0.0001) followed by Tukey’s test. Mean ± s.e.m. of three biological replicates. dA, deoxyadenosine; ND, not detected. d, Schematic of ssDNA-seq protocol. e, Top, Venn diagram showing the percentage of N6-mA DIP-seq differentially increased peaks (day 5 versus day 0) that intersect with ssDNA-seq peaks. Number of peaks per dataset shown; empirical P value computed against genome random (see Methods). Bottom, aggregation of ssDNA-seq signal over N6-mA DIP-seq differentially increased peaks or the peaks shuffled to random genomic positions. f, ssDNA-seq and N6-mA DIP-seq signals at a representative genomic locus. Input signal is overlaid in grey. Genes, peaks, and predicted SIDD regions are as labelled.

We then used DNA immunoprecipitation followed by high-throughput sequencing (DIP-seq) to investigate N6-mA distribution at different stages during cell fate transition. To ensure specificity towards N6-mA in DNA, samples were first treated with extensive RNase digestion (see Methods). This protocol did not pull down any significant signal from RNA–DNA hybrids (P = 0.999) above background (unmodified DNA controls) (Extended Data Fig. 1c). Most of the sequencing reads were mapped (95.55%, with unique mapping) to the mouse genome, with non-significant contributions from mitochondrial DNA and essentially no contributions from microbes. Peaks were called with high confidence using uniquely mapping reads against various controls, such as IgG or whole-genome amplification (WGA) (Extended Data Fig. 1d; see Methods). Consistently, the DIP-seq approach also detected more N6-mA peaks at the transition state than at the other stages (Extended Data Fig. 1e) and the highest aggregate signal over peaks at the transition (Fig. 1b). Bioinformatic analysis revealed that N6-mA is mainly found in intergenic regions with a very high AT content (61.6% AT; Extended Data Fig. 1f), such as LINE-1s, but not other classes of transposons (Extended Data Fig. 1g, h, Supplementary Table 1). Notably, an established algorithm11 predicted that 60–66% of N6-mA peaks are SIDD (Extended Data Fig. 1i). To confirm these results, we extracted M/SAR DNA using a well-established protocol29 and quantified N6-mA levels by mass spectrometry, which can distinguish between methylated and unmethylated dA as well as RNA N6-methyladenosine (m6A, not detected in the samples) (Fig. 1c, Extended Data Fig. 1j). This approach confirmed a tenfold enrichment of N6-mA in M/SAR DNA and its upregulation during the cell fate transition. We next used ssDNA-seq10,11 to interrogate the SIDD regions in our experimental system (Fig. 1d). In self-renewing mES cells (TT2 wild-type), ssDNA-seq reads were enriched at SINE elements, and more signal was observed at the gene bodies and upstream of promoters of expressed versus non-expressed genes, similar to a previous study in B cells11 (Extended Data Fig. 1k, l). Notably, ssDNA-seq peaks in iCdx2 cells become enriched at LINE-1 elements, where N6-mA deposition is increased (Extended Data Fig. 1l–n). Uniquely mapped N6-mA peaks significantly overlapped with ssDNA regions (Fig. 1e). Similar overlap was also observed when all sequencing reads, including unique and non-unique, were included in the analysis24. A representative trophoblast gene locus showed N6-mA directly overlapping with ssDNA peaks (Fig. 1f). Collectively, our genomic and biochemical data demonstrate that N6-mA is enriched at SIDD regions during the transition from ES cells to TSCs.

N 6-mA abolishes SATB1–DNA binding in vitro

We further investigated whether the presence of N6-mA affects the binding of regulators of SIDD, such as SATB1. First, we quantified the in vitro binding of SATB1 to N6-mA-modified oligodeoxynucleotides or unmodified controls. The structure of the CUT1 domain, a DNA-binding domain of SATB120,30, shows that the α3-helix of SATB1 inserted into the major groove of double-stranded DNA (dsDNA)30. Q390 and Q402 formed hydrogen bonds with the seventh deoxyadenosine of the substrates, while the tenth deoxyadenosine (referred to as 7th N6-mA or 10th N6-mA, respectively) is located outside the binding pocket31 (Fig. 2a). We quantitatively determined the binding affinities between SATB1 and SIDD sequences with N6-mA-modified oligos at the seventh and tenth positions using isothermal titration calorimetry (ITC; Fig. 2b). Of note, the binding affinities were determined using a SATB1 monomer, and the in vivo binding of SATB1 can be augmented by multimerization2,20. The results showed that unmodified dsDNA binds to SATB1 with similar affinities to previous reports31 (KD = 5.9 μM). Notably, this interaction was abolished by the 7th N6-mA (KD undetectable), whereas the 10th N6-mA had no obvious effects (KD = 7.0 μM).

Fig. 2: DNA N6-mA modification abolishes binding of SATB1 to DNA in vitro.
figure 2

a, The structure of SATB1 in complex with unmodified dsDNA (PDB code: 2O49)30. The structure of SATB1 is shown in white and the structure of dsDNA is shown in green as cartoon. The seventh and tenth deoxyadenosines are shown in yellow as sticks. b, ITC titration curves and fitting curves of SATB1 titrated into unmodified SIDD dsDNA substrates and SIDD dsDNA substrates modified with N6-mA at the seventh or tenth position. c, The SPRi response curves of unmodified and 7th N6-mA-modified SIDD dsDNA binding to wild-type and mutant SATB1 proteins. Immobilized protein concentration, 1 mM; injected DNA concentration, 2.5 μM. d, Summary of KD values determined by SPRi from unmodified and 7th N6-mA-modified SIDD dsDNA binding to wild-type and mutant SATB1. Mean ± s.d. of three biological replicates. Data shown in bd are representative of three independent experiments with similar results.

We next sought to engineer SATB1 mutants that could tolerate the 7th N6-mA, which would be helpful for elucidating the function of N6-mA in vivo. The surface plasmon resonance imaging (SPRi) results revealed that the Q390RQ402A mutant bound similarly to both unmodified and 7th N6-mA SIDD sequences, whereas the Q402N and Q402T mutants did not bind either sequence, presumably owing to the loss of coordinated interactions (Fig. 2c, d, Extended Data Fig. 2a). Biolayer interferometry (BLI) confirmed these results (Extended Data Fig. 2b, c).

In addition, the DNA-binding properties of SATB1 are also influenced by sequence specificity (Extended Data Fig. 2d, e). A few other known AT-rich DNA-binding proteins tested did not display N6-mA specificity. ARID3A recognized neither the unmodified nor the 7th N6-mA substrates, whereas FOXM1 and FOXD3 bound to both with similar binding affinities (Extended Data Fig. 2f, g).

N 6-mA antagonizes SATB1 binding in vivo

The biochemical results prompted us to investigate the effect of N6-mA on interactions between SATB1 and chromatin. Consistent with previous results18,19, we found marked upregulation of SATB1 during the cell fate transition from ES cells to a TSC-LC fate, especially at maximal N6-mA levels. Paradoxically, despite this global increase, SATB1 chromatin immunoprecipitation followed by high-throughput sequencing (ChIP–seq) demonstrated a 50% reduction in SATB1 peaks at days 5–6, when N6-mA levels are highest (Fig. 3a).

Fig. 3: N6-mA antagonizes binding of SATB1 to the chromatin during TSC development.
figure 3

a, SATB1 binding is inversely correlated with N6-mA levels. Left, western blot of SATB1 expression during cell fate transition at indicated time points. The experiments were repeated independently three times with similar results. Right, number of SATB1 ChIP–seq peaks at indicated days. For blot source data, see Supplementary Fig. 1. b, Aggregation of differential signals of N6-mA DIP and SATB1 ChIP centred at differentially increased N6-mA peaks (day 5 versus day 0). c, Number of SATB1 ChIP–seq peaks in control or Alkbh1-overexpressing (OE) cells. d, Top, Venn diagram showing the intersection between N6-mA differentially decreased peaks and SATB1 ChIP–seq differentially increased peaks in Alkbh1-overexpressing versus control cells. Bottom, overlap between the intersecting regions and ssDNA-seq peaks, not to scale. Number of peaks per dataset given in diagram; empirical P value computed versus genome random. e, Average signal and heat maps of SATB1 ChIP–seq around N6-mA peak centres in control or Alkbh1-overexpressing cells.

Further analysis revealed that binding of SATB1 to chromatin is inversely correlated with N6-mA deposition, and that SATB1 sites are excluded from N6-mA sites (Extended Data Fig. 3a, b). The differentially decreased SATB1 peaks (day 5 versus day 0) overlapped significantly (P < 1 × 10−6) with both the gained N6-mA sites at day 5 and the ssDNA-seq peaks (Extended Data Fig. 3c). Concordantly, aggregation profiling also showed that SATB1 ChIP–seq signals decreased at peak centres where N6-mA signals increased during the cell fate transition (day 5 versus day 0; Fig. 3b).

We further interrogated N6-mA function by overexpressing Alkbh1, which encodes the DNA N6-mA demethylase1,9,24,32,33,34. Consistent with several previous studies1,9,35, overexpressed ALKBH1 was localized primarily to the nucleus, but not the mitochondria, in ES cells and TSC-LCs (Extended Data Fig. 3d). As expected, overexpression of Alkbh1 greatly reduced N6-mA upregulation during cell fate transition (Extended Data Fig. 3e), and a significant proportion of ALKBH1-regulated N6-mA peaks (P < 1 × 10-6) occurred at SIDD identified by ssDNA-seq (Extended Data Fig. 3f). Upon overexpression of Alkbh1, the number of SATB1 ChIP–seq peaks more than doubled (Fig. 3c), whereas overall SATB1 levels remained unchanged (Extended Data Fig. 3g). The increased SATB1 peaks in cells overexpressing Alkbh1 intersected significantly with differentially decreased N6-mA peaks (P < 1 × 10−6, Alkbh1 overexpression versus control); more than 50% of those intersected peaks (SATB1 increased and N6-mA decreased) overlapped with ssDNA peaks (Fig. 3d). There was also a conspicuous increase in the SATB1 ChIP–seq signal at differentially decreased N6-mA sites in Alkbh1-overexpressing cells compared to controls (Fig. 3e). In addition, the increased SATB1 peaks were enriched on young LINE-1s, but not other classes of endogenous transposons, and overlapped with ssDNA regions (Extended Data Fig. 3h, i, Supplementary Table 1). Thus, N6-mA antagonizes the binding of SATB1 to specific SIDD sequences in vitro and with chromatin in vivo.

N 6-mA controls euchromatin boundaries

Given the well-known role of SIDD and SATB1 in facilitating long-range chromatin interactions, we used assay for transposase-accessible chromatin using sequencing (ATAC-seq) to interrogate chromatin accessibility in iCdx2 cells (Fig. 4, Extended Data Fig. 4). Although ATAC-seq and N6-mA peaks are mutually exclusive at each other’s peak centres (Fig. 4a), further analysis revealed a striking ‘peak and valley’ distribution pattern. N6-mA signals are maximal at the boundaries of the ATAC-seq inserts, mostly 0.5–1.0 kb from the centres of the ATAC-seq peaks; similarly, ATAC-seq signals are maximal at the boundaries of the N6-mA peaks (Fig. 4a). Correlation analysis demonstrated that although N6-mA and ATAC-seq peaks do not directly overlap at their peak summits, these peaks are significantly enriched at each other’s boundaries (approximately 0.5–1 kb; Fig. 4b). Moreover, when Alkbh1 was overexpressed, chromatin accessibility increased at the original (control) N6-mA peaks and spread to much larger chromatin domains beyond the boundaries of N6-mA peaks, such as at the Dlk1Dio3 locus (Fig. 4c, d). In summary, N6-mA is enriched at boundaries of euchromatin as defined by ATAC-seq, and thereby prevents euchromatin from spreading into heterochromatin.

Fig. 4: N6-mA is enriched at euchromatin boundaries, thereby preventing the spreading of euchromatin.
figure 4

a, Aggregation profiles of day 5 ATAC-seq inserts at day 5 versus day 0 differentially increased N6-mA peak centres (left) and day 5 N6-mA DIP-seq signal at ATAC-seq day 5 peak centres (right). b, Left, histogram distribution of the distance between ATAC peaks and N6-mA peaks at day 5. Right, bar chart showing the percentage of N6-mA peaks harbouring ATAC peaks within a certain distance at day 5. Empirical P value computed versus genome random. c, Aggregation profiles showing differential ATAC-seq signals in Alkbh1-overexpressing versus control cells at N6-mA differentially decreased (overexpression versus control) peak centres and flanking regions. d, Differential Alkbh1 overexpression versus control ATAC-seq signal, day 5 N6-mA DIP-seq, and N6-mA peaks at the centromeric end of the Dlk1–Dio3 imprinting locus.

N 6-mA is vital in early embryogenesis

Gene set enrichment analysis (GSEA) demonstrated that LINE-1 and genes involved in trophoblast differentiation were significantly upregulated in Alkbh1-overexpressing cells, with specificity for genes that are upregulated in spiral artery trophoblast giant cells (SpA-TGCs)36, but not in the other trophectoderm lineages (Fig. 5a, Extended Data Fig. 5a, b), as confirmed by quantitative PCR with reverse transcription (RT–qPCR) (Extended Data Fig. 5c). Furthermore, the majority of imprinted genes, such as H19, which have critical roles in TSC and placental development37,38, were upregulated when N6-mA was removed by overexpression of Alkbh1; maternally expressed genes appeared to be more severely affected than paternally expressed genes (Extended Data Fig. 5d). We then confirmed the accelerated differentiation of TGC-like cells (TGC-LCs) by using two methods to quantify polyploid (>4N) cells (Extended Data Fig. 5e, f). Thus, the acceleration of TGC formation only by the overexpression of catalytically active Alkbh1 suggests that N6-mA maintains the TSC fate.

Fig. 5: Epigenetic silencing pathway mediated by N6-mA and SATB1 has critical roles in trophectoderm development.
figure 5

a, GSEA of 273 spiral artery trophoblast giant cell (Spa-TGC) genes reveals significant enrichment (P < 0.001) in Alkbh1-overexpressing cells compared to control cells. Enrichment score reflects the degree to which Spa-TGC genes are overrepresented at the top or bottom of a ranked list of genes from RNA-seq data, and significance is determined empirically (see Methods). b, Numbers of developing embryos or resorptions from heterozygous Alkbh11a/1c crosses. n = 25 total embryos from four independent matings. c, Haematoxylin and eosin (H&E) staining of the maternal–fetal interface shows strong diminishing of TGCs in Alkbh11a/1a placentas. Alkbh11a/1a mouse placental tissues at E8.5 are depleted of secondary giant cells (Gi), which normally reside between the decidual (De) layer and the chorio-allantoic plate (Ch, Al), as seen in the Alkbh11c/1c sample. Representative images of three litter-matched E8.5 or E12.5 embryos per genotype from three dams. Scale bars, 100 μm. d, Multi-plane quantification shows depletion of trophoblast giant cells in Alkbh11a/1a placentas (18.8) compared to Alkbh11c/1c controls (53.3). Two-sided paired t-test (t = 5.06, degrees of freedom (df) = 2). Mean ± s.e.m. of five planes through the maternal–fetal interface of three litter-matched mice per genotype, from three dams. e, Left, schematic of Satb1 knockout (KO) rescue experiment in iCdx2 cells overexpressing wild-type (WT) or N6-mA tolerant mutant (Mut) Satb1. Right, RT–qPCR showing that the expression of a TGC-specific marker gene (Prl3b1) is more efficiently rescued by the Satb1 N6-mA tolerant mutant. One-way ANOVA (P < 0.0001) followed by Tukey’s multiple comparisons test on data from day 10. Mean ± s.e.m. of three biological replicates. f, Proposed model of the function of N6-mA, ALKBH1 and SATB1 at euchromatin boundaries.

Source data

To investigate an early developmental phenotype in an Alkbh1-deficient background, which was reported differently by previous studies26,39, we generated a new strain of Alkbh1-knockout mice using the ‘knockout-first’ strategy40 targeting exon 2 (tm1a allele, henceforth called 1a), and its floxed rescue (tm1c allele, henceforth called 1c) (Extended Data Fig. 5g). Although heterozygous mice were born alive at expected Mendelian ratios, no homozygous Alkbh11a/1a mice were observed at birth, indicating a fully penetrant embryonic lethality phenotype (n = 68 pups, 12 litters; Extended Data Fig. 5h). This phenotype was completely rescued by removing the targeting cassette via Flp to create a floxed exon 2 (1c), which rules out concerns for off-target or secondary mutations; Alkbh11c/1c mice were viable, fertile and indistinguishable from wild-type mice (n = 52 pups, 9 litters; Extended Data Fig. 5g, i). Furthermore, we leveraged an immunofluorescence approach9 to interrogate N6-mA levels in iCdx2 cells and embryos at embryonic day 8.5 (E8.5) (Extended Data Fig. 5j, k). Digestion with S1 nuclease, micrococcal nuclease, or DNase I (in addition to RNase A/T1) greatly reduced or completely abolished the signal, whereas treatment with RNase H did not affect the signal, indicating that the signal comes exclusively from DNA. Moreover, examination of the cohort of embryos from a timed mating experiment demonstrated that Alkbh11a/1a embryos developed at expected Mendelian frequencies, but were all deceased and undergoing resorption by E17.5 (n = 25 embryos, 4 matings; Fig. 5b). Histological examination of E8.5 and E12.5 maternal–fetal interfaces, including the chorioallantoic plate and surrounding decidua, demonstrated that deletion of Alkbh1 induced a significant reduction in TGCs (Fig. 5c, d). RNA-seq of the maternal–fetal interface at E10.5 (Extended Data Fig. 5l, Supplementary Table 2) showed that a few key factors involved in trophoblast development were significantly reduced, including Gpc3, which is expressed by differentiating human syncytiotrophoblasts41. Genes involved in the regulation of maternal–fetal blood pressure and the hypoxia response, such as Agt42 and Nppc43, were also aberrantly regulated. Of note, the TGC depletion phenotype and gene expression changes cannot be explained by expression of the gene Nrp, which partially overlaps with exon 1 of Alkbh1, because it is minimally expressed in placental tissue at similar levels across genotypes (Extended Data Fig. 5m). In summary, Alkbh1 deficiency in vivo increases placental N6-mA and inhibits TSC differentiation into TGCs, corroborating the finding that overexpression of Alkbh1 facilitates TSC differentiation in cell culture.

As our results showed that N6-mA antagonizes the interaction of SATB1 with chromatin, we next used a CRISPR–Cas9 approach to knock out Satb1 in the iCdx2 system (Extended Data Fig. 6a). RNA-seq demonstrated that trophectoderm genes were inefficiently induced in Satb1-knockout cells (Extended Data Fig. 6b). Consequently, Satb1 deficiency impairs differentiation of TSC-LCs into TGC-LCs, as assayed by flow cytometry and morphology (Extended Data Fig. 6c, d). Thirty-four per cent of genes activated by Satb1 are also repressed by N6-mA, as identified in Alkbh1-overexpressing cells (Extended Data Fig. 6e).

We next leveraged the aforementioned N6-mA-tolerant SATB1 mutant (Q390RQ402A) (Fig. 2) to further elucidate the epistasis between Satb1 and N6-mA (Fig. 5e). We reconstituted Satb1-knockout cells with wild-type Satb1 or the N6-mA-tolerant Satb1 mutant and showed that reconstitution with the tolerant mutant generated more polyploid TGC-LCs at a faster rate than with the wild-type gene (Extended Data Fig. 6c, d). Concordantly, the tolerant mutant rescued the expression of the downregulated trophectoderm lineage markers in the Satb1-knockout cells more efficiently than did the wild-type Satb1 (Fig. 5e, Extended Data Fig. 6f). These results further support the notion that N6-mA modulates the ES cell-to-TSC fate transition by antagonizing SATB1 function.


In this work, we have shown how N6-mA helps to regulate chromatin structure by antagonizing SATB1 at SIDD during early development (Fig. 5f). N6-mA accumulates at the boundaries between euchromatin and heterochromatin and restricts euchromatin regions from spreading, thereby preventing ES cells from adopting a TSC fate. These findings are corroborated by genetic studies of Alkbh1-knockout mice. Notably, the DNA N6-mA demethylase ALKBH1 also prefers the unpairing SIDD sequences24. These findings together reveal an unexpected mode of epigenetic regulation by the newly identified DNA modification N6-mA via DNA secondary structures during early development.

SATB1, a well-known SIDD-binding factor, is a critical chromatin organizer for orchestrating large-scale chromatin structures12,13,14,20. Our current work reveals an unexpected role for N6-mA as an antagonist of SATB1 function in vitro and in vivo. Strikingly, binding of SATB1 to SIDD sequences is completely abolished in the presence of N6-mA (KD undetectable); this is unusual for epigenetic effector binding proteins, which usually manifest several-fold differences in binding. Consistent with the long-appreciated role of SATB1, N6-mA is specifically enriched at the boundaries between euchromatin and heterochromatin, thereby controlling the spread of euchromatin. The major targets of the N6-mA–SIDD–SATB1 pathway appear to be imprinting genes, such as H19, which is important for TSC differentiation37. In parallel, this pathway may regulate gene expression by controlling the chromatin landscape at long ranges, as indicated by the changes in chromatin accessibility found here; this also warrants further research in the future.

In summary, we have shown that N6-mA contributes to epigenetic regulation by antagonizing the function of SATB1. This study sheds new light on the mechanisms by which epigenetic modification and DNA secondary structures regulate chromatin structure and gene expression in early development.


iCdx2 ES cell differentiation

The Tet-inducible-Cdx2-Flag ES cell line (iCdx2 cells) was obtained from the NIA mES cell bank27. iCdx2 cells were maintained on feeder cells in DMEM supplemented with 15% fetal bovine serum (Gibco), 1% non-essential amino acids, 2 mM l-glutamine, 1,000 units of mLIF (EMD Millipore), 0.1 mM β-mercaptoethanol (Sigma), antibiotics and 2 μg/ml doxycycline (Sigma). This line and all iCdx2-derived lines were tested for mycoplasma contamination prior to experimentation. When starting differentiation44, iCdx2 cells were selectively seeded on gelatin-coated plates after stepwise elimination of the feeder mouse embryonic fibroblasts, then Cdx2 overexpression was induced by removing doxycycline. The cells were split at day 1 after doxycycline withdrawal and cell samples were collected on the following days.

Mouse ES cells in 2i/L culture were cultured in commercially available 2i medium kit (Milipore, SF016-200). a2i/L medium contains a 1:1 mixture of DMEM/F12 supplemented with N2 (Invitrogen) and neurobasal medium with glutamine (Invitrogen) supplemented with B27 (Invitrogen), 1× Pen/Strep (Invitrogen), 1,000 units of mLIF (Milipore), 1.5 μM CGP77675 (Tocris) and 3 μM CHIR99021 (Tocris). PKCi/L medium contains a 1:1 mixture of DMEM/F12 supplemented with N2 (Invitrogen) and neurobasal medium with glutamine (Invitrogen) supplemented with B27 (Invitrogen), PenStrep (Invitrogen), 1,000 units of mLIF (Milipore) and 5 μM Gö6976 (Tocris).

Generation of Satb1-knockout cell line

Two guide RNAs targeted to flank exon 6 were chosen. The guide RNA oligos were annealed and cloned into the lentiCRISPRv1 vector with hygromycin resistance. Then two constructs were transfected into ES cells by jetPRIME transfect reagent (Polyplus 11407). Twenty-four hours after transfection, 400 μg/ml hygromycin was added to the medium for 2 days. ES cells were then seeded at low density to obtain single-cell-derived colonies. Then, 72 ES cell colonies were randomly picked up and screened by PCR genotyping (Extended Data Fig. 6a). PCR screening primer sequences can be found in Supplementary Table 3.

Establishment of Satb1-overexpressing cell line

The wild-type and mutant (Q390RQ402A) human Satb1-myc DNA sequences were inserted into pLV-EF1a-IRES-Hygro (pLV), a lentivirus-based vector with hygromycin resistance. For the Satb1 rescue experiment, wild-type and Q390RQ402A mutant Satb1 pLV constructs were introduced to Satb1 knockout ES cells by lentivirus; the original pLV empty vector was chosen as a control. After viral infections, the cells were selected with hygromycin at 400 μg/ml for 4 days, and then the cells were expanded for the following experiments.

DNA extraction

Cellular genomic DNA was purified using the DNeasy kit (Qiagen 69504) with minor modifications. In brief, the cell lysate was treated with RNase A/T1 mix (Thermo Fisher EN0551) at 37 °C for 30 min. Then Proteinase K was added to the lysate and treated at 56 °C for 3 h or overnight. The remaining steps were performed as described in the kit, and the DNA was eluted with ddH2O. Furthermore, to remove trace RNA contamination, eluted DNA was treated with once more with RNase A/T1 for 2 h and then RNase H (NEB M0297S) overnight, followed by phenol-chloroform extraction and ethanol precipitation.

N 6-mA dot blotting

DNA samples were denatured at 95 °C for 5 min, cooled on ice, and neutralized with 10% vol of 6.6 M ammonium acetate. Samples were spotted on the membrane (Amersham Hybond-N+, GE) and air dried for 5 min, then UV-crosslinked (2 × auto-crosslink, 1800 UV Stratalinker, Stratagene). Membranes were blocked in blocking buffer (5% milk in PBS, 0.1% Tween 20 (PBS-T)) for 1 h at room temperature, then incubated with N6-mA antibody (1:1,000, Synaptic Systems 202-003) at 4 °C overnight. Membranes were washed with PBS-T three times and then incubated with HRP-linked secondary anti-rabbit IgG antibody (1:5,000, Cell Signaling 7074S) for 1 h at room temperature. After three washes with PBS-T, the membrane signal was detected with SuperSignal West Dura Extended Duration Substrate kit (Thermo Fisher 34076).

MAR DNA extraction

The MAR DNA extraction was performed as described29,45 with modifications. In brief, the cell pellets were resuspended with buffer A (10 mM HEPES, PH 7.0, 10 mM KCl, 1.5 mM MgCl2, 0.34 M sucrose, 10% glycerol, and proteinase inhibitor cocktail (Roche)) and treated with 0.1% Triton X-100 for 7 min on ice. After a wash with buffer A, the cells were resuspended with buffer A and MNase buffer (NEB). Five microlitres of MNase was added into the cell suspension and incubated at 37 °C for 5 min, then 0.5 mM EGTA was added to terminate the digestion reaction. After washing with low-salt buffer (buffer A with 400 mM NaCl) for 10 min, the cell pellets were resuspended with high-salt buffer (buffer A with 2M NaCl) and rotated for 30 min at 4 °C. After high-speed centrifugation (14,000 rpm) for 15 min, the cell pellets were resuspended with buffer A and treated with RNase A/T1 for 2 h and then RNase H at 37 °C, then Proteinase K was added and incubated at 56 °C for 2 h. The DNA was finally purified by phenol-chloroform extraction and ethanol precipitation.

Flow cytometry

Cells were collected using trypsin treatment. Cells were washed with ice-cold PBS three times and fixed with cold 70% ethanol overnight at 4 °C. After washing with cold PBS three times, cells were resuspended and stained with 0.5 μg/ml DAPI for 10 min, then washed with cold PBS two more times. Samples were analysed by a BD LSR II.

Liquid chromatography with tandem mass spectrometry analysis of N 6-methyl-2′-deoxyadenosine

We added 0.1 U nuclease P1, 0.25 nmol erythro-9-(2-hydroxy-3-nonyl)-adenine hydrochloride (EHNA), and a 3-μl solution containing 300 mM sodium acetate (pH 5.6) and 10 mM ZnCl2 to 1 μg DNA. EHNA served as an inhibitor for adenine deaminase to minimize the deamination of adenosine. The reaction mixture was incubated at 37 °C for 24 h. Next, 0.1 U alkaline phosphatase, 2.5 × 10−4 U phosphodiesterase 1 and 4 μl 0.5 M Tris-HCl buffer (pH 8.9) were added. After digestion at 37 °C for 2 h, the resulting digestion mixture was neutralized with 3 μl 1.0 M formic acid. Then 200 pmol 15N-labelled 2′-deoxyadenosine and 25 fmol D3-labelled N6-methyl-2′-deoxyadenosine were spiked into the digestion mixture, followed by chloroform extraction to remove the enzymes used for the DNA digestion. The resulting aqueous layer was dried and reconstituted in 10 μl ddH2O, to which 90 μl acetonitrile was added. The solution was subsequently centrifuged, and the supernatant was dried and re-dissolved in 50 μl acetonitrile/ddH2O (95:5, v/v) for liquid chromatography with tandem mass spectrometry (LC–MS/MS) analysis.

LC–MS/MS measurements were conducted on an Agilent 1200 series (Waldbronn, Germany) coupled with LTQ Orbitrap Velos mass spectrometer (Thermo Fisher Scientific, San Jose, CA) equipped with a heated electrospray ionization source. Chromatographic separation was conducted on an Agilent Zorbax HILIC Plus column (2.1 mm × 100 mm, 3.5 μm). The mobile phase was (A) H2O containing 0.2% formic acid and 10 mM ammonium formate; (B) acetonitrile containing 0.2% formic acid, 2 mM ammonium formate and 0.06 mM malic acid. An isocratic mode of 5% A and 95% B was used, and the flow rate was 100 μl/min. The injection volume was 25 μl. High-energy collisional dissociation was used and the transitions of m/z 252.1→136.0 (2′-deoxyadenosine), m/z 257.1→141.0 ([15N5]2′-deoxyadenosine), m/z 266.1→150.0 (N6-methyl-2′-deoxyadenosine) and m/z 269.1→153.0 ([D3]N6-methyl-2′-deoxyadenosine) were monitored for quantifications.

Protein expression and purification

The cDNAs encoding SATB1 (amino acids 368–456), FOXM1 (amino acids 225–326) and FOXD3 (amino acids 140–236) were cloned into the pRSFDuet vector. Point mutations were generated using a site-directed mutagenesis kit (Stratagene). All unmodified and modified DNA oligos were synthesized by GenScript.

Proteins SATB1, FOXM1 and FOXD3 were expressed in Escherichia coli strain BL21 (DE3) in the presence of 0.4 mM IPTG. Overnight-induced cells were collected by centrifugation and re-suspended in lysis buffer: 100 mM NaCl, 20 mM Tris pH 7.5. Then cells were lysed with an Emulsiflex C3 (Avestin) high-pressure homogenizer. After centrifugation at 16,770g, the supernatant was applied to the HisTrap column (GE Healthcare). The resultant protein was further purified by anion-exchange chromatography. The peaks eluted were applied to a Superdex 75 10/300GL (GE Healthcare) gel filtration column. All proteins were stored in 100 mM NaCl, 20 mM Tris pH 7.5 at ~20 mg/ml in a −80 °C freezer. Mutants were expressed and purified using essentially the same procedure as for the wild-type proteins.

Surface plasmon resonance imaging

To measure the interaction between immobilized proteins and flowing nucleic acids, an SPR imaging instrument (Kx5, Plexera, USA) was used to monitor the whole procedure in real time. In brief, a chip with well-prepared biomolecular microarray was assembled with a plastic flow cell for sample loading. The DNA samples were prepared at a series of concentrations (2.5, 1.2, 0.6 and 0.3 μM) in TBS running buffer (20 mM Tris, 100 mM NaCl, pH 7.5) while a 10 mM glycine-HCl buffer (pH 2.0) was used as regeneration buffer. A typical binding curve was obtained by flowing DNA samples at 2 μl/s for 300 s association and then flowing running buffer for 300 s dissociation, followed by 200 s regeneration buffer at 3 μl/s. Binding data were collected and analysed by a commercial SPRi analysis software (Plexera SPR Data Analysis Module, Plexera, USA).

Biolayer interferometry assays

His-tagged wild type and mutant SATB1 proteins were immobilized onto anti-His antibody-coated biosensors (Fortebio, Pall Life Sciences). After being washed with binding buffer (20 mM Tris pH 7.5, 100 mM NaCl), the sensors were dipped into binding buffers containing unmodified and modified double-strand DNA oligos at various concentrations for 5 min for complete association followed by 5 min disassociation in binding buffer without DNA. Kinetics were recorded using Octet RED96 (Fortebio, Pall Life Sciences) at 1,000 rpm vibration at 25 °C.

Isothermal titration calorimetry

For ITC measurement, synthetic DNA oligos and the purified proteins were extensively dialysed against ITC buffer: 20 mM Tris pH 7.5, 100 mM NaCl. The titration was performed using a MicroCal iTC200 system (GE Healthcare) at 25 °C. Each ITC titration consisted of 17 successive injections with 0.4 μl for the first and 2.4 μl for the rest. Usually, proteins at 0.5 mM were titrated into DNA oligos at 0.05 mM. The resultant ITC curves were processed using Origin 7.0 software (OriginLab) according to the ‘One Set of Sites’ fitting model.

N 6-mA DIP sequencing and analysis

N6-mA DIP was performed as described1. In brief, 5 μg extracted genomic DNA was sonicated to 200–500 bp with a Bioruptor. Then, NEBNext adaptors were ligated to genomic DNA fragments after end repair following the NEBNext-UltraII library prep manual. The ligated DNA fragments were denatured at 95 °C for 10 min and chilled on ice to form ssDNA fragments. N6-mA enriched DNA fragments were enriched and purified using N6-mA antibody (5 μg for each reaction, Synaptic Systems 202-003) according to the Active Motif hMeDIP protocol. Immunoprecipitated and input DNA were PCR amplified with NEBNext Multiplex Oligos for Illumina indexing primers. Sequencing was performed with the Illumina HiSeq 4000 platform. Basecalling and adaptor sequence trimming were processed using the standard Illumina workflow. After sequencing and filtering, high-quality raw reads were aligned to the mouse genome (UCSC, mm9) with Bowtie2 (2.3.1, default settings)46. To identify uniquely mapped N6-mA peaks before peak calling, multiple-alignment reads were filtered using Samtools to retain reads with map quality (MapQ) scores above 20 and with properly oriented read mates. N6-mA enriched regions were called against single-stranded input DNA as control with SICER (version 1.1, window size 200, gap 600, false discovery rate (FDR) < 0.01)47. Including IgG pull-down and whole genome wide amplification as controls in peak calling generated very similar results (Extended Data Fig. 1). High overlaps were observed among biological and technical replicates of N6-mA DIP-seq peaks during cell fate transition. On average, 73.4% of total peaks during transition are overlapped in pairwise comparison between replicates; and 59.6% of unique peaks that excluded some of the multiple mapping reads are overlapped. Differential N6-mA deposition regions among two samples of interest were identified with the SICER-df setting. N6-mA genome browser tracks were generated by deepTools v3.1.3 with reads extended to match the fragment size defined by the two read mates and normalized to reads per genomic content (RPGC). Categorical mapping and quantification on endogenous transposons was performed by counting reads on each repeat element based on the UCSC Genome Browser RepeatMaster track of the reference genome mm9. Deposition on each endogenous transposon category was enumerated by BEDTools coverage and normalized by per million mapped reads. LINE-1 transposons were grouped by evolutionary age into young: < 1.5 million years (Myr; L1Md_A, L1MdT, L1Md_Gf); middle: > 1.5 Myr & < 6 Myr (L1Md_F, L1VL); old > 6 Myr (Lx, L1_Mus) as previously described48. Intersections were performed by USeq49. Part of the data analysis was done by in-house customized scripts in R, Python or Perl.

To test the N6-mA antibody binding to DNA and RNA-DNA hybrid in DIP, 1 pmol of different oligos (single N6-mA containing oligo, double N6-mA containing oligos, single unmodified DNA oligo and DNA-RNA hybrid with RNA m6A modification) were added into 5 μg sonicated genomic DNA. The spiked-in oligos were heated at 95 °C for 10 min then slowly cooled down in a PCR machine and tested by agarose gel electrophoresis. Then DIP was performed according to the denaturing DNA IP protocol (Active Motif 55010). After denaturing under 95 °C for 10 min, DNA was quickly transferred on ice for 5 min, then 5 μg N6-mA antibodies or rabbit IgG were added and incubated overnight at 4 °C. Then sheep anti-rabbit IgG dynabeads (Invitrogen 11203D) were added and incubated for 2 h at 4 °C. After washing and elution using the buffer from the Active Motif hMeDIP kit, the enriched DNA fragments were purified by phenol-chloroform extraction followed by ethanol precipitation. Real-time PCR was performed with the iQ SYBR Green Supermix (BioRAD 170-8882) and quantified by a CFX96 or CFX384 system (BioRAD).


ssDNA-seq was performed as previously described with minor modifications11. ES cells (8 × 107) were cultured in feeder-free conditions. Low-salt buffer (15 mM Tris-HCl, pH 7.5, 60 mM KCl, 15 mM NaCl, 5 mM MgCl2, 0.5 mM EGTA, and 300 mM sucrose) at 37 °C was added for 5 min. Cells were then treated with 100 mM KMnO4 for 80 s at 37 °C (‘treated’ sample), and the reaction was quenched by the addition of 50 mM EDTA, 700 mM β-mercaptoethanol, and 1% (w/v) SDS. In parallel, the same number of cells were treated with water (‘blank’ sample) and processed similarly. After overnight proteinase K digestion at 37 °C, DNA was extracted twice with phenol and once more with phenol:chloroform:isoamyl alcohol (25:24:1, v/v/v; PCI), and precipitated with 2 M ammonium acetate in ethanol (PCI extraction with ethanol precipitation).

Genomic DNA was resuspended in 10 mM Tris-HCl, pH 8.0, and 1 mM EDTA buffer (TE buffer) and treated with RNase A/T1 (1:50 dilution, 40 μg/ml and 100 U/ml, respectively) for 1 h at 37 °C, then PCI extracted with ethanol precipitation and resuspended in TE buffer. Free 3′ ends formed due to random DNA breakage during sample preparation were blocked by treatment with 100 μM cordycepin-5′-triphosphate sodium salt (Sigma Aldrich C9137) and 400 U of terminal transferase (NEB M0315; TdT) in 1 × TdT buffer in a reaction volume of 3 ml for 2 h at 37 °C. DNA was PCI extracted with ethanol precipitation and resuspended in TE buffer.

Digestion of ssDNA was carried out by dividing each of the treated and blank samples into separate reactions with 0, 50, 100, or 200 U of S1 nuclease (Thermo Fisher EN0321) in 500 μl of the supplied reaction buffer and incubating for 30 min at 37 °C. DNA was PCI extracted with ethanol precipitation and resuspended in TE buffer. Based on a fragment size distribution of 2–10 kb, the 50 U S1-treated samples (treated and blank) were chosen for further processing. Next, 35 μg of DNA was biotinylated with 300 U TdT, 250 μM dATP, 250 μM dCTP, and 50 μM Biotin-16-dUTP (Roche 11093070910) in a final volume of 300 μl of 1× TdT buffer at 37 °C for 30 min. Reactions were stopped with 15 μl of 0.5 M EDTA. PCI extraction with ethanol precipitation was performed, followed by a second ethanol precipitation to remove free biotin, and samples were resuspended in TE buffer.

Samples were sonicated with a Covaris S220 to generate DNA fragments between 200 and 700 bp and PCI extracted with ethanol precipitation to remove free biotin. DNA was resuspended in TE buffer, and 10% of the treated sample was saved for the input control. With the remaining sample, biotinylated fragments were pulled down using streptavidin-coated beads using the manufacturer’s protocol (Dynabeads kilobase BINDER Kit, Thermo Fisher 60101). Fragments were released by incubating beads with 50 U S1 nuclease in 100 μl reaction buffer for 15 min at 37 °C. Finally, DNA was purified with a MinElute PCR Purification Kit (Qiagen) and eluted in 30 μl of the manufacturer’s elution buffer. Sample concentrations were measured using a Qubit 3.0 fluorometer, and at least a tenfold greater pulldown efficiency for treated versus blank was confirmed.

ssDNA-seq library preparation and high-throughput sequencing data processing and analysis

Treated samples and their input controls were further processed for high-throughput sequencing. Sequencing libraries were made using the NEBNext Ultra II DNA Library Prep kit with 10–100 ng of each sample and input. Libraries were prepared according to the manufacturer’s instructions without size selection. Library concentrations were measured with Qubit and quality control performed with Agilent 2100 Bioanalyzer. Libraries were pooled and sequenced paired-end 2 × 100 bp in one lane of an Illumina HiSeq4000.

Sequencing reads were filtered and pre-analysed with the Illumina standard workflow. After filtering, raw reads (in fastq format) were aligned to the mouse genome (UCSC, mm9) with Bowtie2 (2.3.1) using the default settings. Non-uniquely mapped reads were filtered out using a MapQ > 10. PCR and optical duplicates were then removed with Picard (2.9.0; BigWigs for genome browser tracks were generated using deepTools and normalized to reads per genomic content (RPGC). Aggregation profiles and heatmaps were generated from bigWigs over regions of interest, and in indicated plots, presented as a log2 fold change over input. After alignment and processing, ssDNA enriched regions were called with SICER (version 1.1, FDR <0.01, input DNA as control), with window size 200 bp and gap size 600 bp.

Analysis of ssDNA enrichment in transposable elements

Raw fastq reads were analysed by SalmonTE (version 0.8.2) with default parameters and fold enrichment of samples over input was calculated for each family of transposable elements in the mouse index50. Standalone aggregation profiles were constructed from bigWigs without filtering for uniquely mapped reads over RepeatMasker annotations of full-length (longer than 6 kb) young LINE-1 elements (L1Md_A, Tf, and Gf). Heatmaps used only uniquely mapped data over the same annotated LINE-1 regions.

Chromatin immunoprecipitation

Cells (107) were washed once with DPBS and then fixed in 5 ml DMEM medium with 1% formaldehyde for 10 min at room temperature with rotation. Fixation was terminated with addition of 100 mM glycine followed by 5 min rotation at room temperature. Cells were then scraped and spun down at 4 °C at 1,000 rpm for 5 min, followed by two washes with ice-cold PBS. Cell pellets were resuspended in 1 ml lysis buffer 1 (50 mM Hepes-KOH, pH7.5, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% NP-40, and 0.25% Triton X-100) followed by rotation at 4 °C for 10 min. Cell pellets were spun down at 1,400g for 5 min and resuspended in 1 ml lysis buffer 2 (10 mM Tris-HCl, pH8.0, 200 mM NaCl, 1 mM EDTA, and 0.5 mM EGTA). Following another 10 min incubation at 4 °C, cell pellets were again spun down at 1,400g for 5 min and resuspended in 200–300 μl lysis buffer 3 (10 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.5% Na-deoxycholate and 0.5% N-lauroylsarcosine). Cells were then sonicated to about 500 bp fragments using a Covaris S200 with the following parameters: peak power: 120, duty factor: 2.0, cycle/burst: 200, time: 10 min, and power output: 2.1. The sonicated cell solution was then diluted with lysis buffer 3 plus 1% Triton X-100, incubated for 15 min with rotation at 4 °C and spun down at 20,000g for 15 min. Ten microlitres of anti-SATB1 antibody (Abcam ab109122) was added to the supernatant; an equivalent amount of rabbit IgG antibody was used as a control. After overnight incubation at 4 °C, antibodies were pulled down using anti-rabbit Dynabeads (10 μl per 1 μg antibody). After incubation with beads for 2 h at 4 °C, samples were extensively washed three times with RIPA buffer (50 mM Hepes-KOH, pH 7.6, 500 mM LiCl, 1 mM EDTA, 1% NP40 and 0.7% Na-deoxycholate), rotating at 4 °C for 10 min each time. After the last wash, beads were washed once with 1 × TE, 50 mM NaCl buffer. DNA was then eluted in elution buffer (50 mM Tris-HCl, pH8.0, 10 mM EDTA and 1% SDS) twice at 65 °C for 20 min. Eluted DNA was then incubated at 65 °C overnight to reverse cross-link. The solution was treated with RNase A/T1 mix and incubated at 37 °C for 1 h, followed by the addition of 3 μl proteinase K and incubation for 2 h at 56 °C. The solution was then phenol-chloroform extracted once followed by DNA extraction using the MinElute PCR Purification Kit (Qiagen 28004). DNA was eluted with EB buffer for downstream library preparation using the NEBNext UltraII library preparation kit. Sequencing was performed with Hiseq 2000, and the output sequencing reads were filtered and pre-analysed with Illumina standard workflow. After filtration, the qualified tags (in fastq format) were aligned to the mouse genome (mm9) with Bowtie2 (2.3.1)46. After mapping, multiple-alignment reads were filtered using Samtools to maintain reads with quality score above 20 and properly paired. Then SATB1-enriched regions were called with SICER (window size 200, gap 600, FDR <0.01, input DNA as control). Quantification of transposable elements was performed as described for N6-mA DIP-seq.

Statistics on peak intersections

USeq was used to intersect peaks and determine an empirical P value from two given sets of peaks. Each peak in the second set was shuffled within the same chromosome, and this ‘genome random’ set was intersected with the first peak set. This process was repeated 1 × 106 times, after which the mean intersections of the randomized set and the number of times there were more intersections in the genome random than the original set was determined. A count of zero trials where genome random was greater than the original was reported as P < 1 × 106. It was also verified that the original had > twofold enrichment of intersections compared to the mean of genome random.


ATAC-seq was performed as described51. In brief, 50,000 cells were collected and resuspended in ice-cold ATAC-RSB buffer (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl2, 0.1% NP40, 0.1% Tween-20, and 0.01% digitonin). After incubation on ice for 3 min, the cells were washed with ice-cold ATAC-RSB buffer without NP40. Then cells were resuspended in transposition mix composed of 0.33 × DPBS, 0.01% Tween-20, 1 × TD buffer (Nextera DNA Library Prep Kit, Illumina) and transposase (100 nM final concentration, Illumina). Transposition reaction was incubated at 37 °C for 30 min in a thermomixer with mixing. Then transposed DNA was purified by the MinElute PCR Purification Kit. Eluted DNA was amplified using NEB Next UltraII Q5 Master Mix for five cycles. A real-time qPCR amplification was performed to determine additional cycles. The amplified DNA libraries were size-selected and purified by AMPure beads according to the manufacturer’s instructions. After sequencing and filtering, high-quality raw reads were aligned to the mouse genome (UCSC, mm9) with Bowtie2 (2.3.1, default settings). PCR duplicates were removed by Samtools. ATAC peaks were identified using MACS2 (2.1.1)52 with BAMPE mode. Peaks overlapping with Encode blacklist regions were removed. To generate bigwig files, reads with insertion sizes greater than 120 bp were retained and normalized by reads per genomic content (RPGC). Quantification of intersections was performed by USeq.

Maternal–fetal interface RNA extraction

At 10.5 days post conception, whole maternal–fetal interfaces were collected by excising the decidua and placenta from the embryo and myometrium. Samples were stored at −80 °C until homogenization and RNA extraction. Care was taken to reduce excess myometrial, perimetrium, amnion, and umbilicus tissue. Tissues were lysed using the QIAzol lysis reagent (Qiagen 79306) in conjunction with the Tissuelyser II homogenization system (Qiagen 85300) per the manufacturer’s instructions.

RNA-seq and RT–qPCR approaches

RNA was extracted with miRNeasy kit (Qiagen 217004) and standard RNA protocol. The quality of RNA samples was measured using the Agilent Bioanalyzer. RNA-seq libraries were constructed with Illumina Truseq stranded mRNA library prep kit and sequenced on an Illumina HiSeq 2000. Transcriptome mapping was performed with STAR version 2.5.2b using mm9 Gencode vM1 release exon/splice-junction annotation. Differential expression analysis was performed using cufflinks version 2.2.1 using UCSC gene annotation with first-strand library type and the correction of sequencing bias and multiple reads mapping. GSEA on RNA-seq data was conducted through the GSEA Preranked tool53 based on cuffdiff output and the target gene sets (SpA-TGCs or Prdm1-TGCs36). Statistical significance for GSEA is determined empirically by permuting the phenotype labels 1,000 times to generate a null distribution of the enrichment score, then comparing the actual enrichment score to the null. Gene ontology enrichment analysis was performed using PANTHER version 1454. Part of the data analysis was done by in-house customized scripts in R. Supplementary RNA-seq analysis was performed using HTSEQ -count and DESeq2 using the local alignment paradigm.

For RT–qPCR, cDNA libraries were generated with the iScript cDNA Synthesis Kit (BioRAD 170-8891). Real-time PCR was performed with the iQ SYBR Green Supermix (BioRAD 170-8882) and quantified by a CFX96 or CFX384 system (BioRAD).

Mouse breeding

Animal experiments were performed with approval from Yale University’s Institutional Animal Care and Use Committee and in accordance with all relevant ethical regulations and guidelines of Yale University and the NIH. C57BL/6N-Alkbh1tm1a(EUCOMM)Hmgu/Ieg (Alkbh1tm1a/wt) males and females were purchased from the EMMA repository (derived from a C57BL/6N background ES cell line)55. Heterozygous females were backcrossed once onto C57BL/6J males. Next, females were crossed with FLPe-expressing males (Jackson Laboratory) to generate mice with the Alkbh1tm1c allele. Intra-litter matings of 6–10-week-old Alkbh1tm1a/wt or Alkbh1tm1a/tm1c mice were performed to remove the FLPe allele and then maintained for more than five generations before performing subsequent experiments.

Timed matings were performed by synchronizing the oestrus cycle of 12-week-old female Alkbh1tm1a heterozygotes via exposure to male soiled bedding, and subsequently placing two females per Alkbh1tm1a male heterozygote per cage. Females were checked for vaginal plugs daily in the early morning, and the presence of a plug was standardized as day E0.5. Pregnant animals were euthanized using CO2 per university guidelines, and embryos and placental tissue were dissected out. Sample sizes were not predetermined for mating experiments; 2–3 biological samples with multiple sections were chosen for staining and histology. Blinding was not performed, and randomization was not applicable to this study.

Placental tissue histological and N 6-mA immunofluorescence staining

Placental tissue was immediately fixed in 4% formaldehyde overnight and transferred to 70% ethanol. Placental tissue was paraffin embedded and sectioned by Yale Pathology Tissue Services (YPTS); subsequent H&E staining of placental tissues was performed by YPTS.

N6-mA immunofluorescence of paraffin-embedded tissues was performed via the following protocol: slides were heated to 55 °C for 45 min, and then deparaffinized in xylenes for 15 min. Rehydration of the tissue was performed via submersion in 100%, 90%, 80%, 70% EtOH for 5 min each, followed by 10 min in PBS. Cells were permeabilized by incubation in 0.2% Triton X-100 in PBS for 20 min followed by washes with 0.1% Tween 20 in PBS (PBS-T). DNA was hydrolysed by incubation in 2N HCl for 20 min, followed by neutralization with 0.1 M sodium borate for 30 min. After washes with PBS-T, slides were blocked with 6 μl/ml RNase A/T1 (Thermo Fisher EN0551) with additional 50 μg/ml RNase A (Qiagen 19101), and 50 U/ml RNase H (NEB M0297) as indicated in 10% goat serum PBS-T overnight at 37 °C. Slides were incubated in 1:400 α-N6-methyladenine monoclonal antibody (Cell Signaling Technologies 56593, lot 1) overnight at 4 °C. After rinsing in PBS-T, slides were incubated with 1:1,000 goat-anti-rabbit IgG Alexa Fluor 488 (Invitrogen A11008) at room temperature for 60 min. Slides were rinsed in PBS-T for 5 min, incubated in 1:1,000 DAPI for 3 min and rinsed in PBS three times. After immunofluorescence staining, slides were preserved using Fluoromount-G (Electron Microscopy Sciences 17984-25) and stored at 4 °C before imaging.

Immunofluorescence staining of cell culture samples

The stable ES cells carrying pLV-Flag-HA-Alkbh1 or pLV-Alkbh1-HA constructs were grown on gelatin treated slides for 24 h. Cells were fixed with 1% paraformaldehyde and permeabilized with 0.2% Triton X-100 in PBS. Then the slides were incubated in blocking buffer (2% bovine serum albumin, 0.2% Triton X-100 in PBS) for 1 h, incubated with 1:500 HA-tag antibody (Cell Signaling 3724S) in blocking buffer at 4 °C overnight, and then incubated with the secondary antibody. DNA was stained with DAPI for 5 min. Slides were mounted with Fluormount-G.

N6-mA staining in cultured cells was performed as described9 with minor modifications. Cells were grown on gelatin-treated Millicell EZ SLIDES (Millipore PEZGS0816). Cells were then fixed with 1% paraformaldehyde and permeabilized with 0.2% Triton X-100 in PBS. After PBS wash, the slides were treated with 2N HCl for 20 min, then neutralized with 0.1 M sodium borate for 30 min. After PBS washes, the slides were blocked in blocking buffer (2% bovine serum albumin, 0.2% Triton X-100 in PBS) with RNase A/T1 for 1 h, incubated with 1:400 N6-mA antibody (202-003, Synaptic Systems) in blocking buffer at 4 °C overnight. After PBS washes, the slides were incubated with goat anti-rabbit IgG, Alexa Fluor Plus 594 (Invitrogen A32740) for 1 h. DNA was stained with DAPI for 5 min. Slides were mounted with fluormount-G. Images were acquired with a Leica SP5 confocal laser microscope. As nuclease controls, slides were treated with 20 U/ml DNase I (NEB M0303S) or 20,000 gel units/ml MNase (NEB M0247S) overnight at 37 °C before the 2N HCl DNA denaturing step. 100 U/ml RNase H (NEB M0297S) or 2 U/ml S1 nuclease (Thermo Fisher EN0321) treatment were carried out overnight at 37 °C after DNA denaturing with 2N HCl.


Immunofluorescence microscopy was performed within 7 days of staining using (1) an Olympus IX81 Fluoresence microscope at 20x magnification and equal exposure time for each respective channel for tissue sections and (2) a Leica TCS SP5 Spectral Confocal Microscope at 40x magnification and 2.0x zoom for cultured cells. Images were captured and prepared using (1) MetaMorph imaging software (Molecular Devices LLC, v7.7.0) or (2) LAS AF software suite (Leica Microsystems, v2.6.0). Brightfield microscopy was performed using an Olympus BX51 microscope at 10× and processed using CellSens (Olympus, v1.17).

For H&E TGC quantification, maternal-fetal interfaces were sectioned to 8 μm thick, and 16 sections were collected from each implantation site. Five sections spanning 128 μm were chosen for TGC quantification from representative interfaces, and TGCs between the chorioallantoic plate (E8.5)/labyrinth (E12.5) and decidua were counted manually. Quantitative analysis was performed on three litter-matched sets of embryos from three different litters by individuals blinded to the genotype of the sample.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.