Single-CpG resolution mapping of 5-hydroxymethylcytosine by chemical labeling and exonuclease digestion identifies evolutionarily unconserved CpGs as TET targets
© Sérandour et al. 2016
Received: 8 February 2016
Accepted: 9 March 2016
Published: 29 March 2016
Conventional techniques for single-base resolution mapping of epigenetic modifications of DNA such as 5-hydroxymethylcytosine (5hmC) rely on the sequencing of bisulfite-modified DNA. Here we present an alternative approach called SCL-exo which combines selective chemical labeling (SCL) of 5hmC in genomic DNA with exonuclease (exo) digestion of the bead-trapped modified DNA molecules. Associated with a straightforward bioinformatic analysis, this new procedure provides an unbiased and fast method for mapping this epigenetic mark at high resolution. Implemented on mouse genomic DNA from in vitro-differentiated neural precursor cells, SCL-exo sheds light on an intrinsic lack of conservation of hydroxymethylated CpGs across vertebrates.
The recently discovered epigenetic mark 5-hydroxymethylcytosine (5hmC) results from an active DNA demethylation process which involves iterative oxidation of 5-methylcytosine (5mC) driven by Ten-eleven translocation (TET) enzymes and leads to the replacement of 5mC by an unmodified base [1–4]. However, 5hmC is stable enough to be readily detected in DNA, suggesting that, in addition to being an intermediate of DNA demethylation, it may have signaling potential by itself . Hence, genome-wide mapping studies are valuable to understand the function of 5hmC and its importance in gene regulation. Pioneer studies have shown that 5hmC is found almost exclusively (99.89 %) in a CpG dinucleotide context in embryonic stem (ES) cells and pointed to a positive role of 5mC oxidation to 5hmC in the regulation of transcriptional enhancers as well as gene expression [5–11]. Indeed, high 5hmC levels correlate with active chromatin features at enhancers (i.e. H3K4me1 and H3K27ac) and with expression levels in gene bodies. Depending on the expected resolution, several different strategies can be used to map 5hmC. Low resolution (200–300 bp) methods employ hydroxymethylated DNA capture, either with antibodies (hydroxymethylated DNA immunoprecipitation (hMeDIP))  or with streptavidin beads after 5hmC glucosylation and biotinylation (selective chemical labeling (SCL)) . Such methods are sufficient to describe the presence of 5hmC in short genomic regions but, since resolution depends on the size of the DNA fragments, do not allow a precise mapping of the modified base. Whenever single-base resolution is required, for instance to analyze 5hmC distribution with respect to transcription factor binding sites (TFBSs), two methods based on bisulfite (BS) modification of DNA can be used [11, 12]. The first one uses 5hmC protection by glucosylation coupled to 5mC oxidation by recombinant TET followed by BS modification and sequencing (TAB-seq) . The second procedure requires a chemically-induced oxidative deprotection of 5hmC followed by BS modification and sequencing (oxBS-seq) . In the latter, results need to be compared to data obtained with an unmofidied BS-seq procedure which does not discriminate 5mCs from 5hmCs . Although often defined as gold standards, BS-seq-based methods suffer from several drawbacks: (1) efficiency of TAB-seq relies on the use of a highly active recombinant TET enzyme; (2) harsh oxBS conditions lead to a substantial loss of DNA (99.5 % ) and a fairly good correlation between two biological replicates was achieved only after pooling CpG hydroxymethylation scores in given CpG islands ; (3) the current elevated cost of a full genome coverage can be prohibitive; and (4) they require complex bioinformatics . Alternatively, 5hmC can be mapped at single-base resolution through two rounds of MspI digestion of DNA separated by a 5hmC glucosylation step, before size selection and sequencing (RRHP) . Although highly reproducible, this procedure does not cover all CpGs in the genome since MspI requires a CCGG context for DNA cleavage (i.e. 15 % of all CpGs). In addition, restriction enzymes from the PvuRts1I family like AbaSI have been shown to cleave glucosylated 5hmC-containing DNA and to be suitable for genome-wide mapping of the modified base . However, due to their specific sequence requirement and restriction characteristics, theoretically only 58 % of all cytosines can be covered , and ambiguity might exist in 13 % of the cleaved molecules in the Aba-seq assay .
In an effort to develop an alternative approach for single-CpG resolution mapping of 5hmC genome-wide, we adapted a strategy first employed to increase the resolution of chromatin immunoprecipitation (ChIP) through the use of an exonuclease (exo) to trim DNA cross-linked to proteins up to close vicinity of intermolecular bounds (ChIP-exo [18, 19]). This new procedure, called SCL-exo, is shown here to be suited to obtain single-CpG resolution data. Using this approach, we uncovered that, although being included in highly conserved regulatory regions of the mouse genome, a majority of hydroxymethylated cytosines are not conserved in other vertebrate species, suggesting that they might affect chromatin structure rather than directly regulate transcription factor binding.
Results and discussion
In order to evaluate the reproducibility of the SCL-exo procedure Pearson’s correlation coefficient was determined for id CpG.wig files from two technical replicates of SCL-exo (Fig. 2c). Signals from SCL-exo id CpG replicates showed a high correlation (r = 0.72), indicating that SCL-exo is suited for a reproducible identification of hydroxymethylated CpGs. Notably, non-overlapping SCL-exo id CpGs between two replicates had a lower coverage than overlapping id CpGs (Additional file 1d). Hence, increasing sequencing depth might enhance the reproducibility of the method. Considering that the mean signal of overlapping id CpGs was 1.6-fold higher than the mean signal of non-overlapping id CpGs from two replicates with 48 million reads, increasing sequencing depth up to 1.6 × 48 million reads (≈80 million reads) per replicate could allow higher confidence in the identification of hydroxymethylated CpGs. Finally, SCL-exo id CpG signal showed a fairly good correlation with hMeDIP (r = 0.51, Fig. 2d) and 57.02 % of the SCL-exo id CpGs with at least 20× coverage were included in hMeDIP peaks (Fig. 2e). As a possible readout of exonuclase undigested DNA fragments, we selected unique CpGs contained in the first 10 bases of reads obtained by SCL-seq without exonuclease digestion to build a SCL-seq id CpG.wig file. This SCL-seq id CpG signal did not correlate with hMeDIP (r = 0.07, Fig. 2f) and SCL-exo (r = 0.01, Fig. 2g). The mean signal (number of reads) at SCL-seq id CpGs was 2.03 and could thus be considered as a threshold for false identification of hydroxymethylated CpGs. Hence, for the subsequent analysis, only CpGs identified in at least two out of three SCL-exo replicates (consensus id CpGs) at a threshold arbitrarily set to 8× coverage were considered (178,218 id CpGs). The status of hydroxymethylation of 27 selected CpGs from this set (consensus id CpGs) and included in a MspI CCGG restriction site was next verified by using the EpiMark 5-hmC and 5-mC Analysis Kit (New England Biolabs) which allows a quantitative determination of the percentage of hydroxymethylation of CpGs thanks to the insensitivity of glucosylated ChmCGG sites to MspI cleavage. A strong correlation (r = 0.79) between EpiMark and SCL-exo data was observed (Fig. 2h and Additional file 1e). This is in the range of what has been observed when Aba-seq and EpiMark data were compared (r = 0.72) for hydroxymethylated CpGs in ES cells . Interestingly, the hydroxymethylated status of SCL-exo id CpGs not included in hMeDIP peaks was systematically validated with the EpiMark kit, thus indicating that those were not false positive (Additional file 1e). Comparison between the two techniques suggested however that below a threshold of 8 % of hydroxymethylation (as assessed by EpiMark), CpGs were not efficiently identified by SCL-exo (Fig. 2h and Additional file 1e). Here again, increasing sequencing depth might increase the identification rate of these poorly hydroxymethylated CpGs. This correlation study allowed us to estimate that CpGs showing 20 % hydroxymethylation by EpiMark should have an approximate 16× coverage by SCL-exo. Using this threshold of coverage, the calculated overlap between id CpGs from two replicates of SCL-exo was 53.6 %. The genome-wide distribution of SCL-exo id CpGs and probable CpGs was next interrogated with the CEAS annotation tool . As already described for 5hmC-enriched regions from P19 cells recovered by immunoprecipitation , SCL-exo id CpGs were particularly enriched in introns (p = 9.7e−256) and promoters (p = 1.1e−49), although exons might be slightly under-represented due to the fact that “probable CpGs,” which are found in exons for 6.7 % of them, were not included in the analysis (Additional file 1f). In addition, inclusion of id CpGs in enhancers (H3K4me1 positive regions), either active (positive for H3K27ac) or primed (negative for H3K27ac), was proportional to the depth of coverage, suggesting that SCL-exo-identified CpGs with high coverage are likely to be included in functional enhancers.
Collectively, our data indicate that 5hmC can be mapped at single-CpG resolution by SLC-exo. Contrary to methods based on the use of restriction enzymes, SCL-exo has the advantage of being unbiased in terms of sequence context requirement. Using SCL-exo, we demonstrate here that TET enzymes mainly target unconserved CpGs, suggesting that cytosine hydroxymethylation at enhancers might serve a structural role at the level of chromatin rather than having a direct effect on transcription factor binding to DNA.
Cell culture and genomic DNA preparation
P19.6 embryonal carcinoma cells were culture as described  in high glucose Dulbecco’s Modified Eagle Medium supplemented with 10 % fetal calf serum (GIBCO, USA). Neural progenitor cell differentiation was triggered by 10−6 M all-trans retinoic acid (RA) in 10-cm diameter culture dishes. Cells were scraped in phosphate buffered saline 48 h after RA addition and were pelleted at 100 g before genomic DNA extraction using a DNeasy Blood and Tissue kit (Qiagen, France). Mouse ES cells (E14) were grown in high glucose Dulbecco’s Modified Eagle Medium supplemented with 15 % fetal calf serum, 0.1 mM β-mercaptoethanol, 1X non-essential amino acids, and LIF (1000 U/mL). Genomic DNA from E14 cells was extracted with the DNeasy Blood and Tissue kit (Qiagen).
SCL-seq and SCL-exo procedures
A total of 20 μg of DNA in 300 μL of TE buffer (Tris 10 mM, EDTA 0.1 mM, pH 8.0) were sonicated with a Bioruptor (Diagenode, Belgium) to yield 200–500 bp fragments. Each glucosylation and biotinylation reaction was run using reagents from the Hydroxymethyl Collector kit (ref. 55013, Active Motif, Belgium) and 500 ng of sonicated DNA. For SCL-seq, three technical replicates each with 2.5 μg biotinylated genomic DNA were captured on streptavidin-coated magnetic beads (ref. 11205D, Invitrogen). After five washes and elution from the beads according to the manufacturer’s protocol, the captured DNA was purified, precipitated, and pooled for sequencing library preparation using the TruSeq ChIP Sample Prep Kit (Illumina, ref. IP-202-1012). For SCL-exo, three technical replicates each which 2.5 μg of biotinylated genomic DNA were captured on streptavidin-coated magnetic beads. After six washes in RIPA buffer (50 mM HEPES pH 7.6; 1 mM EDTA; 0.7 % Na-Deoxycholate; 1 % NP-40; 0.5 M LiCl) and two washes in Tris 10 mM pH 8, the DNA-beads complexes were processed as previously described : end polishing, ligation of the P7 exo-adapter, nick repair, lambda and RecJf exonuclease digestion, elution, P7 primer extension, ligation of the P5 exo-adapter, PCR amplification, and finally gel-size selection. The exonuclease-digested DNA was eluted from the beads by incubation in 100 μL of Elution Buffer (95 % formamide, 10 mM EDTA) at 90 °C for 5 min, followed by DNA precipitation and resuspension in 20 μL of water. The SCL-seq and SCL-exo libraries were quantified using the KAPA library quantification kit for Illumina sequencing platforms (KAPA Biosystems, KK4824) and 50 bp single-end sequenced as a pool in a single lane of a HiSeq 2000 (Illumina) for RA-treated P19 cells or in four lanes of a HiSeq 2500 (Illumina) for E14 mESCs, following the manufacturer’s protocol. Sequencing data are available at the NCBI GEO database under reference GSE70635.
After 48 h of all-trans retinoic acid treatment, P19.6 cells were cross-linked in 10 mL PBS 1 % formaldehyde for 10 min at room temperature. The reaction was stopped by adding 1 mL of 1 M Glycine. Cells were washed twice in cold PBS, scrapped, and pelleted at 100 g. The ChIPs were performed as described previously  with chromatin from 50.106 cells, using 10 μg of anti-H3K4me1 (Abcam, ref. ab8895) and anti-H3K27ac (Abcam, ref. ab4729) antibodies. The ChIP-seq libraries were prepared using the TruSeq ChIP Sample Prep Kit (Illumina, ref. IP-202-1012) and sequenced on HiSeq 2000. Mapping to the mouse mm8 genome and peak calling were run as described previously .
SCL-exo fastq files were filtered using SolexaQA  to retain high-quality reads only (Q = 20, l = 17) before being mapped to mm8 (P19 cells) or mm9 (E14 cells), forward and reverse strands separately, using Bowtie  with parameters l = 32 bp, n = 1, m = 1, strata, best, and Samtools . The bam files were then processed to generate.wig files using MACS 1.4.0 . Resulting.wig files were filtered to remove UCSC blacklisted regions as well as few regions showing a very high signal and not included in the blacklists. Reads for which a single CpG was found within 10 bases from their 5’ end were selected to build a single-CpG resolution.wig file in which signal at id CpGs corresponds to the sum of the reads covering both Cs (two strands) at each given CpG. High confidence SCL-exo id CpGs (178,218 CpGs) were called when covered more than 8× and found in at least two out of three replicates. For the analysis of the reproducibility of SCL-exo identification of CpGs as a fonction of read density (Additional file 1d), 5000 CpGs identified in two replicates and 5000 CpGs identified in only one replicate out of two were randomly selected. TFBS motif search was run with the SeqPos motif tool from Cistrome  (which does not accept datasets with more than 5000 regions) within 100 bp windows centered either on 3682 id CpGs with a SCL-exo signal above 45 reads and overlapping with a hMeDIP peak or on 3682 randomly selected CpGs (Fig. 5). A SeqPos search for motifs according to the conservation of SCL-exo id CpGs was run on a pool of 6163 id CpGs from mouse chr11, covered at least 20×, and sorted into three groups according to their PhastCons scores: High conservation (1072 id CpGs with a PhastCons score between 0.75 and 1), Intermediate conservation (869 id CpGs with a PhastCons score between 0.075 and 0.75), and No conservation (4222 id CpGs with a PhastCons score between 0 and 0.075). Analysis of the conservation of CpGs included in Meis1-bound enhancers was run on 4959 high confidence SCL-exo id CpGs with more than 30× coverage and found within a Meis1 ChIP-seq peak (Fig. 6). Conservation of TAB-seq id CpGs was analyzed for 90,977 CpGs showing more than 20 % of hydroxymethylation in two replicates of TAB-seq (Additional file 3). In addition, 3887 Meis1-bound TGACAG motifs from RA-treated P19 cells were identified and assigned PhastCons scores. Similarly, 152 c-Myc bound CACGTG sites and 313 N-Myc bound CACGTG sites in mouse ES cells were included in this analysis. Finally, 245 CACGTG sites from RA-treated P19 cells and overlapping with SCL-exo id CpGs with more than 30× coverage were selected. All conservation graphs were generated with Cistrome.
Datasets used in this study
TET1 ChIP-seq and hMeDIP-seq data from RA-treated P19 cells were from the Gene Expression Omnibus repository (GEO - http://www.ncbi.nlm.nih.gov/geo) datasets GSM941681 and GSM941665, respectively. MEIS1 chip-seq data from RA-treated P19 cells were from GSM819083. Mouse ESC CpG hydroxymethylation data were extracted from datasets of two technical replicates of TAB-seq (GSM1180306 and GSM1180307) and another biological replicate (GSM118308). Aba-seq and hMeDIP-seq data from E14 mouse ESCs were from GSE42898 and GSM1087009, respectively. N-Myc and c-Myc bound CACGTG sites in mouse ES cells were extracted from GSM288356 and GSM288356 GEO datasets, respectively.
Availability of supporting data
The datasets supporting the results of this article (SCL-exo, Input-seq, SCL-seq, H3K4me1, and H3K27ac ChIP-seq) are available in the GEO repository under accession number GSE70635.
We are grateful to M. Bizot, G. Palierne, F. Percevault, and E. Jullion for technical assistance. We thank R. Métivier and C. Le Péron for critical reading of the manuscript. GS acknowledges support from the Centre National de la Recherche Scientifique, the University de Rennes 1, The Ligue Contre le Cancer, and Cancéropole Grand Ouest. MW is supported by the EpiGeneSys network of excellence, the INCa, and the European Research Council (ERC Consolidator grant n°615371). EM was supported by a PhD fellowship from the Ministère de l’Enseignement Supérieur et de la Recherche.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Tahiliani M, Koh KP, Shen Y, Pastor WA, Bandukwala H, Brudno Y, et al. Conversion of 5-methylcytosine to 5-hydroxymethylcytosine in mammalian DNA by MLL partner TET1. Science. 2009;324:930–5.View ArticlePubMedPubMed CentralGoogle Scholar
- Kriaucionis S, Heintz N. The nuclear DNA base 5-hydroxymethylcytosine is present in Purkinje neurons and the brain. Science. 2009;324:929–30.View ArticlePubMedPubMed CentralGoogle Scholar
- Ito S, Shen L, Dai Q, Wu SC, Collins LB, Swenberg JA, et al. Tet proteins can convert 5-methylcytosine to 5-formylcytosine and 5-carboxylcytosine. Science. 2011;333:1300–3.View ArticlePubMedPubMed CentralGoogle Scholar
- He YF, Li BZ, Li Z, Liu P, Wang Y, Tang Q, et al. Tet-mediated formation of 5-carboxylcytosine and its excision by TDG in mammalian DNA. Science. 2011;333:1303–7.View ArticlePubMedPubMed CentralGoogle Scholar
- Sérandour AA, Avner S, Oger F, Bizot M, Percevault F, Lucchetti-Miganeh C, et al. Dynamic hydroxymethylation of deoxyribonucleic acid marks differentiation-associated enhancers. Nucleic Acids Res. 2012;40:8255–65.View ArticlePubMedPubMed CentralGoogle Scholar
- Stroud H, Feng S, Morey Kinney S, Pradhan S, Jacobsen SE. 5-hydroxymethylcytosine is associated with enhancers and gene bodies in human embryonic stem cells. Genome Biol. 2011;12:R54.View ArticlePubMedPubMed CentralGoogle Scholar
- Szulwach KE, Li X, Li Y, Song CX, Han JW, Kim S, et al. Integrating 5-hydroxymethylcytosine into the epigenomic landscape of human embryonic stem cells. PLoS Genet. 2011;7:e1002154.View ArticlePubMedPubMed CentralGoogle Scholar
- Wu H, D’Alessio AC, Ito S, Wang Z, Cui K, Zhao K, et al. Genome-wide analysis of 5-hydroxymethylcytosine distribution reveals its dual function in transcriptional regulation of mouse embryonic stem cells. Genes Dev. 2011;25:679–84.View ArticlePubMedPubMed CentralGoogle Scholar
- Song CX, Szulwach KE, Fu Y, Dai Q, Yi C, Li X, et al. Selective chemical labeling reveals the genome-wide distribution of 5-hydroxymethylcytosine. Nat Biotechnol. 2011;29:68–72.View ArticlePubMedPubMed CentralGoogle Scholar
- Pastor WA, Pape UJ, Huang Y, Henderson HR, Lister R, Ko M, et al. Genome-wide mapping of 5-hydroxymethylcytosine in embryonic stem cells. Nature. 2011;473:394–7.View ArticlePubMedPubMed CentralGoogle Scholar
- Yu M, Hon GC, Szulwach KE, Song CX, Zhang L, Kim A, et al. Base-resolution analysis of 5-hydroxymethylcytosine in the mammalian genome. Cell. 2012;149:1368–80.View ArticlePubMedPubMed CentralGoogle Scholar
- Booth MJ, Branco MR, Ficz G, Oxley D, Krueger F, Reik W, et al. Quantitative sequencing of 5-methylcytosine and 5-hydroxymethylcytosine at single-base resolution. Science. 2012;336:934–7.View ArticlePubMedGoogle Scholar
- Huang Y, Pastor WA, Shen Y, Tahiliani M, Liu DR, Rao A. The behaviour of 5-hydroxymethylcytosine in bisulfite sequencing. PLoS One. 2010;5:e8888.View ArticlePubMedPubMed CentralGoogle Scholar
- Robinson MD, Statham AL, Speed TP, Clarck SJ. Protocol matters: which methylome are you actually studying? Epigenomics. 2010;2:587–98.View ArticlePubMedPubMed CentralGoogle Scholar
- Petterson A, Chung TH, Tan D, Sun X, Jia X-Y. RRHP: a tag-based approach for 5-hydroxymethylcytosine mapping at single-site resolution. Genome Biol. 2014;15:456.View ArticlePubMedPubMed CentralGoogle Scholar
- Sun Z, Terragni J, Borgaro JG, Liu Y, Guan S, Wang H, et al. High resolution enzymatic mapping of genomic 5-hydroxymethylcytosine in mouse embryonic stem cells. Cell Rep. 2013;3:567–76.View ArticlePubMedPubMed CentralGoogle Scholar
- Wang H, Guan S, Quimby A, Cohen-Karni D, Pradhan S, Wilson G, et al. Comparative characterization of the PvuRts1l family of restriction enzymes and their application in mapping genomic 5-hydroxymethylcytosine. Nucleic Acids Res. 2011;39:9294–305.View ArticlePubMedPubMed CentralGoogle Scholar
- Rhee HS, Pugh BF. Comprehensive genome-wide protein-DNA interactions detected at single-nucleotide resolution. Cell. 2011;147:1408–19.View ArticlePubMedPubMed CentralGoogle Scholar
- Sérandour AA, Brown GD, Cohen JD, Carroll JS. Development of an illumina-based ChIP-exonuclease method provides insight into FoxA1-DNA binding properties. Genome Biol. 2013;14:R147.View ArticlePubMedPubMed CentralGoogle Scholar
- Shin H, Liu T, Manrai AK, Liu XS. CEAS: cis-regulatory element annotation system. Bioinformatics. 2009;25:2605–6.View ArticlePubMedGoogle Scholar
- Hon GC, Song CX, Du T, Jin F, Selvaraj S, Lee AY, et al. 5mC oxidation by Tet2 modulates enhancer activity and timing of transcriptome reprogramming during differentiation. Mol Cell. 2014;56:286–97.View ArticlePubMedPubMed CentralGoogle Scholar
- Guo H, Zhu P, Wu X, Li X, Wen L, Tang F. Single-cell methylome landscapes of mouse embryonic stem cells and early embryos analyzed using reduced representation bisulfite sequencing. Genome Res. 2013;23:2126–35.View ArticlePubMedPubMed CentralGoogle Scholar
- Smallwood SA, Lee HJ, Angermueller C, Krueger F, Saadeh H, Peat J, et al. Single-cell genome-wide bisulfite sequencing for assessing epigenetic heterogeneity. Nat Methods. 2014;11:817–20.View ArticlePubMedPubMed CentralGoogle Scholar
- Stelzer Y, Shivalila CS, Soldner F, Markoulaki S, Jaenisch R. Tracing dynamic changes of DNA methylation at single-cell resolution. Cell. 2015;163:218–29.View ArticlePubMedGoogle Scholar
- Liu T, Ortiz JA, Taing L, Meyer CA, Lee B, Zhang Y, et al. Cistrome: an integrative platform for transcriptional regulation studies. Genome Biol. 2011;12:R83.View ArticlePubMedPubMed CentralGoogle Scholar
- Swarnalatha M, Singh AK, Kumar V. The epigenetic control of E-box and Myc-dependent chromatin modifications regulate the licensing of lamin B2 origin during cell cycle. Nucleic Acids Res. 2012;40:9021–35.View ArticlePubMedPubMed CentralGoogle Scholar
- Campanero MR, Armstrong MI, Flemington EK. CpG methylation as a mechanism for the regulation of E2F activity. Proc Natl Acd Sci USA. 2000;97:6481–6.View ArticleGoogle Scholar
- Pierard V, Guiguen A, Coin L, Wijmeersch G, Vanhulle C, Van Driessche B, et al. DNA cytosine methylation in the bovine leukemia virus promoter is associated with latency in a lymphoma-derived B -cell line: potential involvement of direct inhibition of cAMP-repsonsive element (CRE)-binding protein/CREmodulator/activation transcription factor binding. J Biol Chem. 2010;285:19434–49.View ArticlePubMedPubMed CentralGoogle Scholar
- Handa V, Jeltsch A. Profound flanking preference of Dnmt3a and Dnmt3b mammalian DNA methyltransferases shape the human epigenome. J Mol Biol. 2005;348:1103–12.View ArticlePubMedGoogle Scholar
- Hu L, Li Z, Cheng J, Rao Q, Gong W, Liu M, et al. Crystal structure of TET2-DNA complex: insight into TET-mediated 5mC oxidation. Cell. 2013;155:1545–55.View ArticlePubMedGoogle Scholar
- Cheng Y, Ma Z, Kim B-H, Wu W, Cayting P, Boyle AP, et al. Principles of regulatory information conservation between mouse and human. Nature. 2014;515:371–5.View ArticlePubMedPubMed CentralGoogle Scholar
- Schmidt D, Wilson MD, Spyrou C, Brown GD, Hadfield J, Odom DT. ChIP-seq: using high-throuput sequencing to discover protein-DNA interactions. Methods. 2009;48:240–8.View ArticlePubMedPubMed CentralGoogle Scholar
- Cox MP, Peterson DA, Biggs PJ. SolexaQA: At-a-glance quality assessment of Illumina second-generation sequencing data. BMC Bioinformatics. 2010;11:485.View ArticlePubMedPubMed CentralGoogle Scholar
- Langmead B, Trapnell C, Pop M, Salzberg SL. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol. 2009;10:R25.View ArticlePubMedPubMed CentralGoogle Scholar
- Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, et al. The Sequence alignment/map format and SAMtools. Bioinformatics. 2009;25:2078–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Zhang Y, Liu T, Meyer CA, Eeckhoute J, Johnson DS, Bernstein BE, et al. Model-based Analysis of ChIP-Seq (MACS). Genome Biol. 2008;9:R137.View ArticlePubMedPubMed CentralGoogle Scholar