RRHP: a tag-based approach for 5-hydroxymethylcytosine mapping at single-site resolution
© Petterson et al.; licensee BioMed Central Ltd. 2014
Received: 1 April 2014
Accepted: 28 August 2014
Published: 24 September 2014
Current methods for genomic mapping of 5-hydroxymethylcytosine (5hmC) have been limited by either costly sequencing depth, high DNA input, or lack of single-base resolution. We present an approach called Reduced Representation 5-Hydroxymethylcytosine Profiling (RRHP) to map 5hmC sites at single-base resolution by exploiting the use of beta-glucosyltransferase to inhibit enzymatic digestion at the junction where adapters are ligated to a genomic library. Therefore, only library fragments presenting glucosylated 5hmC residues at the junction are sequenced. RRHP can detect sites with low 5hmC abundance, and when combined with RRBS data, 5-methylcytosine and 5-hydroxymethylcytosine can be compared at a specific site.
Since 2009, one of the most rapidly developing subdisciplines in molecular genetics has proven to be the identification and characterization of 5-hydroxymethylcytosine (5hmC). Like its close relative, 5-methylcytosine (5mC), 5hmC is one of the covalent modifications observed in prokaryotic and eukaryotic genomes ,, which constitute an important class of epigenetic modifications. At present, the precise role of 5hmC in the genomic context is under close study from a myriad of angles. One paradigm implicates 5hmC in the oxidative demethylation of cytosine , which has been bolstered by subsequent characterizations of 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC) in the genome . Aside from its mechanistic characterization, 5hmC localization and tissue distributions have also been extensively studied, resulting in clear demonstration of elevated abundance in tissues within the central nervous system (CNS) . Pathologically, a profound depletion of 5hmC is observed across several malignant carcinomas .
As a result of the increased study of this modification, more sensitive tools are required for detection, quantitation, and ultimate mapping of the marker across the genome. While methods such as LC-MS/MS-MRM are useful for sensitive detection and quantitation of 5hmC and other modified nucleosides, most genetic applications require the ability to pin down the mark to a tight region, locus, or specific junction within a locus. Several technologies have become available that utilize old methodologies, such as immunoprecipitation or qPCR, as well as new methodologies, including chemical labeling, single-molecule kinetic monitoring, nanopore conductivity, and more. A recent review highlights the advantages and pitfalls of these techniques .
For many applications, genome-wide approaches, including hMeDIP and bio-orthogonal labeling with glucosylation, provide robust enrichment pools for Sanger sequencing as well as massively parallel (next-generation) sequencing. Despite good coverage of the genome and high specificities, these methods are often limited by input requirements which typically are in the neighborhoods of several micrograms. These amounts of DNA are often not feasible for investigation of precious samples such as stem cells or selectively isolated cellular subpopulations (that is, diverse neuronal cells from a whole brain sample). Importantly, these enrichment-based methodologies, even in highly optimized protocols, lack single-base resolution, and identified hydroxymethylated sites will fall within the range of several hundred to several thousand bases. Depending on how well a particular region is annotated, such resolution is often insufficient to describe activity in transcriptionally relevant sites with confidence. Findings from such studies require subsequent validations with locus-specific assays, such as glucMS-qPCR, to enhance the 5hmC positions.
Recently, two approaches which enable quantitative, single-base resolution mapping of 5hmC have been reported. Oxidative bisulfite sequencing (oxBS-Seq)  takes advantage of selective chemical oxidation via organometallic catalysis to yield 5fC from 5hmC, which is then susceptible to traditional bisulfite conversion and results in a different sequencing signal from the 5mC sibling. The 5hmC level is inferred by comparing the methylation values between the modified and traditional bisulfite sequencing. Although this process allows interrogating of 5hmC at single-base resolution, the oxidation step leads to significant DNA degradation (approximately 0.5% of original DNA fragments are retained through the process, according to the authors), which again restricts its application to very rare samples. In addition to this approach, great strides have been reported with the Tet-assisted Bisulfite Sequencing (TAB-Seq)  approach. In this methodology, 5hmC positions are initially protected by glucosylation and then treated with the Tet enzyme to selectively oxidize naked 5mC positions to 5hmC and then 5fC and 5caC. These 5fC or 5caC positions are susceptible to bisulfite conversion and deamination, so the only remaining cytosine positions are those originating from 5hmC. While the method avoids harsh organometallic treatment for oxidation, it extensively depends upon the Tet enzyme, which is known to present low efficiency (the authors suggested an efficiency of ≥90%, which can render at least 10% of methylated residues unconverted) . Unconverted positions would, therefore, be falsely identified as 5hmC sites and contribute to a higher background signal for the assay.
As an alternative to these methods, we present a novel approach, known as reduced representation 5-hydroxymethylcytosine profiling (RRHP), that avoids harsh chemical conversion processes and affords sequence-level resolution of 5hmC positions. The method features a rapid workflow (<24 total h), allows for starting inputs as low as 100 ng, and offers strand-specific information about 5hmC distribution. The absence of chemical conversions also allows for sequencing of native DNA sequences, which enhances sequencing quality and resulting mapping ratios. Most importantly, the method proves to be a highly reproducible, positive display method, allowing for higher confidence when interrogating positions with low 5hmC content. When combined with existing reduced representation bisulfite sequencing (RRBS) data for the same sample, RRHP allows for both high resolution and accurate quantitation of 5mC and 5hmC positions across the genome simultaneously.
Preparatory scheme and initial evaluation
Statistical analysis of sequencing reads from in silico simulation and RRHP experiment
(- βgt control)
Tagged reads (n)
Tagged reads (%)
5hmc sites (n)
Calculated false calling rates of RRHP for various read counts
Read cutoff no.
Annotative characteristics and unique features
Cross-platform validations and correlation
5hmC profiling in breast and liver cancer samples by RRHP
RRHP features a rapid workflow, avoids harsh chemical modification, and allows processing of DNA inputs as low as 100 ng. Also, RRHP is a positive display that eliminates the need for parallel subtractive sequencing as required by oxBS-Seq and does not need high sequencing depth to detect low 5hmC abundances as required by TAB-Seq, which needs an average of 26.5× read coverage to resolve a single 5hmC site with 20% abundance at a false discovery rate (FDR) <5%. Under the same FDR, we found that RRHP was able to confidently detect approximately one million 5hmC sites in human brain tissue with only 20 to 30 million reads. Since brain tissues have the highest 5hmC content compared to other tissues examined thus far, such sequencing depth should be sufficient for other tissues. However, it would still be helpful if a pilot library for an unknown sample was sequenced with a higher depth and then analyzed as a function of the total number of reads by down-sampling the data to determine the minimal sequencing depth required. This will enable the user to adjust the sequencing depth accordingly and make sequencing runs more cost-efficient. It is also worth noting that samples which need comparative analysis should be sequenced with the same depth or normalization is required. Typically, data for the same sample from runs with different sequencing depth can be normalized by the total number of reads. However, due to the positive display nature of RRHP, the total number of sequencing reads is not only associated with sequencing depth but also related to 5hmC abundance. Therefore, normalization by the total number of reads is inappropriate when comparing different samples since the number of sequencing reads is an indication of 5hmC abundance. Thus, 5hmC sites with housekeeping characteristics are needed to serve as an internal control to normalize samples. Alternatively, a spike-in control with various 5hmC levels would be helpful for normalization. Currently, we maintain the relative 5hmC abundance between samples by multiplexing samples with equal volume, rather than equal mass, for libraries prepared in parallel (that is, same DNA input, same purification, same amplification, and so on). For example, we multiplexed sample RRHP-MspI-2 with half the volume that of RRHP-MspI-1, and as expected, the total number of reads for RRHP-MspI-2 was half of RRHP-MspI-1 (Table 1). Without data normalization, the correlation between the two samples was 0.864, indicating little variation in the sample preparation (Figure 3b). Also, from our comparison of the paired tumor samples, our results were in agreement with previously published data that showed tumors have less 5hmC abundance.
The high sensitivity and low background associated with RRHP allows for both qualitative and relative quantitative descriptions of 5hmC at genomic loci, resulting in high correlation with previously described interrogative methods such as hMeDIP-Seq and glucMS-qPCR. The method detects 5hmC in a strand-specific fashion and can couple analyses of epigenetic modifications with genomic variation, such as SNP detection. When combined with RRBS data, the method allows for high resolution and direct correlation of 5mC and 5hmC positions. This system can be adapted for any platform (such as ion torrent PGM and so on) that utilizes adapterized libraries, and by using other glucose-sensitive restriction enzymes for fragmentation and library digestion, we can profile 5hmC sites in alternative CpG motifs as well as non-CpG contexts. This principle can also be applied in mapping other epigenetic modifications, such as 5fC, 5CaC and N6-methyladenine (6 mA) by using alternative restriction enzymes. In addition, three other enzyme-based methods were also recently developed for genome-wide 5hmC profiling including Aba-Seq, HELP-GT assay and HMST-Seq -. Aba-Seq utilizes a DNA-modification dependent restriction endonuclease, AbaSI, coupled with sequencing. AbaSI recognizes glucosylated 5hmC with high specificity and generates a double strand break 11-13 bp downstream of the recognition site. However, this enzyme prefers sites with two cytosines positioned symmetrically around the cleavage site, and the cleavage efficiency is lower when only one of the two cytosines is a glucosyl-5hmC. Putative 5hmC sites are indirectly deduced by checking for the presence of a cytosine at the expected distances from either side of the mapped cleavage sites. There are two major limits for this method: first, certain 5hmC sites may not be detected due to the low cleavage efficiency caused by the absence of symmetric pattern of the recognition site. Second, it causes assignment ambiguity to the exact cytosine in categories which has 2CGs or 2CHs at the symmetric recognition site, and these sites account for 13% of all identified cleavage sites, according to the authors. In addition, Aba-Seq requires a much higher sequencing depth; over 200 million reads is needed for an Aba-Seq library, making it not cost competitive to RRHP. The other two assays, HELP-GT and HMST-Seq, are more similar to RRHP in terms of the restriction enzyme used and the genomic coverage, but both of them are negative display methods and require subtractive sequencing. In other words, two libraries for each sample have to be sequenced in order to infer the 5hmC status for a CpG site, and it is challenging to normalize the data for subtraction given a variety of factors which may affect the read counts and distribution between the two libraries. Lastly, both HELP-GT and HMST-Seq assays have a very complicated workflow which requires multiple enzymatic digestions, sequential adapterization, bead capture or in vitro transcription, making it not ideal for samples with low DNA input.
Here we present a novel approach, RRHP, for genome-wide profiling of 5hmC, which exploits β-glucosyltransferase (β-GT) to inhibit restriction digestion at adapters ligated to a genomic library, such that only fragments presenting glucosylated 5hmC residues at adapter junctions will be amplified and sequenced. This assay profiles 5hmC sites with single-base resolution in a strand-specific fashion. When combined with existing RRBS data, it allows for simultaneous comparison of 5mC and 5hmC at a specific site. We find that this assay is a robust and cost-efficient tool for profiling 5hmC across the genome.
Adapter design and construction
P5CCG and P7CG adapter pairs were constructed in a manner that allowed for 5'-CG overhangs instead of the standard 5'-T overhangs of the Illumina TruSeq P5 and P7 adapters. In the P5CCG adapter pair, CCGG is retained at the junction following ligation to a library fragment while in the P7CG adapter pair, the CCGG junction becomes TCGG at ligation and is no longer sensitive to HpaII or MspI restriction. For both adapter pairs, two long oligos were hybridized with their respective complementary short oligos at 50 μM with a slow ramp-down (0.1 C/s) from 95 C to 12 C in oligo hybridization buffer (50 mM NaCl, 1 mM Tris-HCl pH 8.0, 100 μM EDTA). The P5 adapter pair was prepared from HPLC-purified oligos (IDT): 5'-ACACTCTTTCCCTACACGACGCTCTTCCGATCTC-3' (long) and 5'-CGGAGATCGGAAGAG-3ddC -3' (short). The P7 adapter pair was prepared from the oligos: 5'-GTGACTGGAGTTCAGACGTGTGCTCTTCCGATCT-3' (long) and 5'- CGAGATCGGAAGAG-3ddC -3' (short).
RRHP library construction and sequencing
Genomic DNA from human male cerebellum or tumor and adjacent normal tissue was purified via phenol:chloroform extraction and digested for 8 h at 37°C with 20 U MspI (NEB). Following digestion, the enzyme was inactivated at 65°C for 15 min and fragmented gDNA was purified using the DNA Clean and Concentrator kit (Zymo Research). For adapterization, 100 or 500 ng of fragmented DNA was ligated overnight at 16°C with 400 U T4 DNA Ligase (NEB) and 500 nM modified P5 and P7 adapters (IDT), such that the CCGG junction was retained at the P5 adapter and destroyed in the P7 adapter (see ‘Adapter construction’ for sequence detail). After overnight incubation, adapters were extended with 2 U GoTaq (Promega) and 500 μM dNTPs (Zymo Research) at 72°C for 30 min. Following extension, adapterized libraries were purified with the DNA Clean and Concentrator kit. Adapterized libraries were glucosylated with 10 U β-GT (Zymo Research) and 100 nM UDPG at 37°C for 4 h. For negative control reactions, β-GT was omitted from the incubation. To all of the reactions, 20 U of MspI or HpaII were then added and incubated at 37°C overnight. After digestion, an additional 20 U of MspI or HpaII were added and allowed to incubate for an additional 1 h. Enzymes were heat inactivated at 65°C and libraries were purified with the DNA Clean and Concentrator kit. Purified libraries were then loaded in a 2.5% (w/v) 50:50 NuSieve:agarose gel and electrophoresed. Size-selected libraries were cut from 110 to 500 bp and purified with the Zymoclean Gel DNA Recovery Kit (Zymo Research). Finalized libraries were then amplified with 500 nM P5/P7 barcoding primers (IDT) in OneTaq 2X Master Mix (NEB) with the thermal profile: 94°C for 30 s, 58°C for 30 s, and 68°C for 30 s, repeated for 10 cycles. Amplifications were sampled for visualization on a 2% agarose and purified with the DNA Clean and Concentrator kit. For sequencing, equal volumes of each amplified library were pooled and diluted to 8 pM for 50 bp singleton reads on the Illumina HiSeq 2000 (Illumina).
Bioinformatic processing and statistical analyses
Sequencing reads from the RRHP assay were first processed to trim off low quality bases and the P7CG adapter at the 3' end of the reads and then aligned to the hg18 build of the human genome using Bowtie0.12.8 and its default parameters with --best. Aligned reads with CCGG tag at 5' end were counted. The correlation analysis between the different RRHP libraries were performed by comparing the presence of the tagged reads at each profiled MspI site, and the Pearson’s coefficient was calculated accordingly. The reads for RRBS library were processed as previously reported. Gene ontology analysis was performed using the Genomic Regions Enrichment of Annotations Tool (GREAT) .
JBP-1-mediated enrichment sequencing library preparation and analysis
The enrichment sequencing libraries were prepared from 1 ug gDNA fragmented with dsDNA Shearase (Zymo Research). Fragments were then A-tailed with Klenow exo- fragment (NEB) and ligated to adapters per standard Illumina library preparation protocols. Libraries were glucosylated with β-GT and enriched via incubation with immobilized JBP-1 using the Quest 5hmC DNA Enrichment Kit (Zymo Research) per the manufacturer’s protocol. Libraries were subjected to limited amplification, purified, and sequenced on the Genome Analyzer IIX platform (Illumina). Resulting reads were trimmed for adapters, aligned to hg18 with Bowtie, and analyzed for enrichment peak calling in MACS.
glucMS-qPCR validation of de novo 5hmC discovery loci
A total of 100 ng of gDNA from the same human male cerebellum sample was glucosylated with 10 U β-GT (Zymo Research) and 100 nM UDPG (Zymo Research) or mock-treated without enzyme at 37°C for 2 h. To the same reactions, 20 U of MspI (NEB) were added and incubated for an additional 2 h. Following heat inactivation at 65°C for 15 min, reactions were purified with the DNA Clean and Concentrator kit (Zymo Research) and quantified. 10 ng of each treatment group was utilized for qPCR in triplicate with QuestTaq qPCR Master Mix (Zymo Research) and 200 nM primers (IDT). Reactions were amplified on a CFX96 cycler (Bio-Rad) with the thermal profile: 95°C for 3 min, 40 cycles of 95°C for 30 s, 60°C for 20 s, 72°C for 20 s, and then a final extension at 72°C for 1 min before a 4°C hold. All amplifications were then subjected to melt curve analysis to ensure specific amplification and identity. Cp values were averaged from three technical replicates for each treatment and 5hmC% was calculated using the equation ((Digested -)-(Digested +) / (Digested -) - (Intact)) × 100%.
Primers for glucMS-qPCR Validation:
Locus 1 Chr1 : 153450946 - 153451176
Fwd: 5' CTTCAGCCCACTTCCCAGAC
Rev: 5' GTGGGTGGGCGACTTCTTAG
Locus 2 Chr17: 7529452 - 7529624
Fwd: 5' AAGGACAGAAGCCCGACAAA
Rev: 5' CAGCTATTCGGGAGGGTGAG
The RRHP sequencing data from this study have been submitted to NCBI’s Gene Express Omnibus (GEO) under accession number GSE49546.
We thank M. Van-Eden, K. Booher, J. Claypool, and P. Shi for helpful comments on the manuscript.
- Kriaucionis S, Heintz N: The nuclear DNA base 5-hydroxymethylcytosine is present in purkinje neurons and the brain. Science. 2009, 324: 929-930. 10.1126/science.1169786.PubMedPubMed CentralView ArticleGoogle Scholar
- Tahiliani M, Koh KP, Shen Y, Pastor WA, Bandukwala H, Brudno Y, Agarwal S, Lyer LM, Liu DR, Aravind L, Rao A: Conversion of 5-methylcytosine to 5-hydroxymethylcytosine in mammalian DNA by MLL partner TET1. Science. 2009, 324: 930-935. 10.1126/science.1170116.PubMedPubMed CentralView ArticleGoogle Scholar
- Guo JU, Su Y, Zhong C, Ming GL, Song H: Hydroxylation of 5-methylcytosine by TET1 promotes active DNA demethylation in the adult brain. Cell. 2011, 145: 423-434. 10.1016/j.cell.2011.03.022.PubMedPubMed CentralView ArticleGoogle Scholar
- He YF, Li BZ, Li Z, Liu P, Wang Y, Tang Q, Ding J, Jia Y, Chen Z, Li L, Sun Y, Li X, Dai Q, Song CX, Zhang K, He C, Xu GL: Tet-mediated formation of 5-carboxylcytosine and its excision by TDG in mammalian DNA. Science. 2011, 333: 1303-1307. 10.1126/science.1210944.PubMedPubMed CentralView ArticleGoogle Scholar
- Globisch D, Münzel M, Müller M, Michalakis S, Wagner M, Koch S, Brückl T, Biel M, Carell T: Tissue distribution of 5-hydroxymethylcytosine and search for active demethylation intermediates. PLoS ONE. 2010, 5: e15367-10.1371/journal.pone.0015367.PubMedPubMed CentralView ArticleGoogle Scholar
- Jin SG, Jiang Y, Qiu R, Rauch TA, Wang Y, Schackert G, Krex D, Lu Q, Pfeifer GP: 5-Hydroxymethylcytosine is strongly depleted in human cancers but its levels do not correlate with IDH1 mutations. Cancer Res. 2011, 71: 7360-7365. 10.1158/0008-5472.CAN-11-2023.PubMedPubMed CentralView ArticleGoogle Scholar
- Fu Y, He C: Nucleic acid modifications with epigenetic significance. Curr Opin Chem Biol. 2012, 16: 516-524. 10.1016/j.cbpa.2012.10.002.PubMedPubMed CentralView ArticleGoogle Scholar
- Booth MJ, Branco MR, Ficz G, Oxley D, Krueger F, Reik W, Balasubramanian S: Quantitative sequencing of 5-methylcytosine and 5-hydroxymethylcytosine at single-base resolution. Science. 2012, 336: 934-937. 10.1126/science.1220671.PubMedView ArticleGoogle Scholar
- Yu M, Hon GC, Szulwach KE, Song C, Zhang L, Kim A, Li X, Dai Q, Shen Y, Park B, Min J, Jin P, Ren B, He C: Base-resolution analysis of 5-hydroxymethylcytosine in the mammalian genome. Cell. 2013, 149: 1368-1380. 10.1016/j.cell.2012.04.027.View ArticleGoogle Scholar
- Gu H, Smith ZD, Bock C, Boyle P, Gnirke A, Meissner A: Preparation of reduced representation bisulfite sequencing libraries for genome-scale DNA methylation profiling. Nat Protoc. 2011, 6: 468-481. 10.1038/nprot.2010.190.PubMedView ArticleGoogle Scholar
- McLean CY, Bristor D, Hiller M, Clarke SL, Schaar BT, Lowe CB, Wenger AM, Bejerano G: GREAT improves functional interpretation of cis-regulatory regions. Nat Biotechnol. 2010, 28: 495-501. 10.1038/nbt.1630.PubMedView ArticleGoogle Scholar
- Robertson AB, Dahl JA, Ougland R, Klungland A: Pull-down of 5-hydroxymethylcytosine DNA using JBP1-coated magnetic beads. Nat Protoc. 2012, 7: 340-350. 10.1038/nprot.2011.443.PubMedView ArticleGoogle Scholar
- Doege CA, Inoue K, Yamashita T, Rhee DB, Travis S, Fujita R, Guarnieri P, Bhagat G, Vanti WB, Shih A, Levine RL, Nik S, Chen EI, Abeliovich A: Early-stage epigenetic modification during somatic cell reprogramming by Parp1 and Tet2. Nature. 2012, 488: 652-655. 10.1038/nature11333.PubMedView ArticleGoogle Scholar
- Zhang Y, Liu T, Meyer CA, Eeckhoute J, Johnson DS, Bernstein BE, Nusbaum C, Myers RM, Brown M, Li W, Liu XS: Model-based analysis of ChIP-Seq (MACS). Genome Biol. 2008, 9: R137-10.1186/gb-2008-9-9-r137.PubMedPubMed CentralView ArticleGoogle Scholar
- Mariani CJ, Madzo J, Moen EL, Yesilkanal A, Godley LA: Alterations of 5-hydroxymethylcytosine in human cancers. Cancers (Basel). 2013, 25: 786-814. 10.3390/cancers5030786.View ArticleGoogle Scholar
- Yang H, Liu Y, Bai F, Zhang JY, Ma SH, Liu J, Xu ZD, Zhu HG, Ling ZQ, Ye D, Guan KL, Xiong Y: Tumor development is associated with decrease of TET gene expression and 5-methylcytosine hydroxylation. Oncogene. 2013, 32: 663-669. 10.1038/onc.2012.67.PubMedPubMed CentralView ArticleGoogle Scholar
- Chen ML, Shen F, Huang W, Qi JH, Wang Y, Feng YQ, Liu SM, Yuan BF: Quantification of 5-methylcytosine and 5-hydroxymethylcytosine in genomic DNA from hepatocellular carcinoma tissues by capillary hydrophilic-Interaction liquid chromatography/ quadrupole TOF mass spectrometry. Clin Chem. 2013, 59: 824-832. 10.1373/clinchem.2012.193938.PubMedPubMed CentralView ArticleGoogle Scholar
- Mendoza-Villanueva D, Zeef L, Shore P: Metastatic breast cancer cells inhibit osteoblast differentiation through the Runx2/CBFβ-dependent expression of the Wnt antagonist, sclerostin. Breast Cancer Res. 2011, 13: R106-10.1186/bcr3048.PubMedPubMed CentralView ArticleGoogle Scholar
- Golshan M, Kuten A, William J, Richardson A, Modarressi A, Matulonis U: Metaplastic carcinoma of the breast with neuroglial differentiation. Breast. 2006, 15: 545-549. 10.1016/j.breast.2005.09.003.PubMedView ArticleGoogle Scholar
- Feng S, Cai M, Liu P, Wei L, Wang J, Qi J, Deng L: Atp6v1c1 may regulate filament actin arrangement in breast cancer cells. PLoS One. 2014, 9: 84833-10.1371/journal.pone.0084833.View ArticleGoogle Scholar
- Wielscher M, Liou W, Pulverer W, Singer CF, Rappaport-Fuerhauser C, Kandioler D, Egger G, Weinhäusel A: Cytosine 5-hydroxymethylation of the LZTS1 gene is reduced in breast cancer. Transl Oncol. 2013, 6: 715-721. 10.1593/tlo.13523.PubMedPubMed CentralView ArticleGoogle Scholar
- Zender L, Xue W, Zuber J, Semighini CP, Krasnitz A, Ma B, Zender P, Kubicka S, Luk JM, Schirmacher P, McCombie WR, Wigler M, Hicks J, Hannon GJ, Powers S, Lowe SW: An oncogenomics-based in vivo RNAi screen identifies tumor suppressors in liver cancer. Cell. 2008, 135: 852-864. 10.1016/j.cell.2008.09.061.PubMedPubMed CentralView ArticleGoogle Scholar
- Sun Z, Terragni J, Borgaro JG, Liu Y, Yu L, Guan S, Wang H, Sun D, Cheng X, Zhu Z, Pradhan S, Zheng Y: High-resolution enzymatic mapping of genomic 5-hydroxymethylcytosine in mouse embryonic stem cells. Cell Rep. 2013, 3: 567-576. 10.1016/j.celrep.2013.01.001.PubMedPubMed CentralView ArticleGoogle Scholar
- Gao F, Xia Y, Wang J, Luo H, Gao Z, Han X, Zhang J, Huang X, Yao Y, Lu H, Yi N, Zhou B, Lin Z, Wen B, Zhang X, Yang H, Wang J: Integrated detection of both 5-mC and 5-hmC by high-throughput tag sequencing technology highlights methylation reprogramming of bivalent genes during cellular differentiation. Epigenetics. 2013, 8: 421-429. 10.4161/epi.24280.PubMedPubMed CentralView ArticleGoogle Scholar
- Bhattacharyya S, Yu Y, Suzuki M, Campbell N, Mazdo J, Vasanthakumar A, Bhagat TD, Nischal S, Christopeit M, Parekh S, Steidl U, Godley L, Maitra A, Greally JM, Verma A: Genome-wide hydroxymethylation tested using the HELP-GT assay shows redistribution in cancer. Nucleic Acids Res. 2013, 41: e157-10.1093/nar/gkt601.PubMedPubMed CentralView ArticleGoogle Scholar
- Weber M, Hellmann I, Stadler MB, Ramos L, Pääbo S, Rebhan M, Schübeler D: Distribution, silencing potential and evolutionary impact of promoter DNA methylation in the human genome. Nat Genet. 2007, 39: 457-466. 10.1038/ng1990.PubMedView ArticleGoogle Scholar
- King DC, Taylor J, Elnitski L, Chiaromonte F, Miller W, Hardison RC: Evaluation of regulatory potential and conservation scores for detecting cis-regulatory modules in aligned mammalian genome sequences. Genome Res. 2005, 15: 1051-1060. 10.1101/gr.3642605.PubMedPubMed CentralView ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. 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.