Gel-free multiplexed reduced representation bisulfite sequencing for large-scale DNA methylation profiling
- Patrick Boyle†1,
- Kendell Clement†1, 2, 3, 4,
- Hongcang Gu†1,
- Zachary D Smith1, 2, 3,
- Michael Ziller1, 2, 3,
- Jennifer L Fostel1,
- Laurie Holmes1,
- Jim Meldrim1,
- Fontina Kelley1,
- Andreas Gnirke1 and
- Alexander Meissner1, 2, 3Email author
© Boyle et al.; licensee BioMed Central Ltd. 2012
Received: 1 June 2012
Accepted: 3 October 2012
Published: 3 October 2012
Sequencing-based approaches have led to new insights about DNA methylation. While many different techniques for genome-scale mapping of DNA methylation have been employed, throughput has been a key limitation for most. To further facilitate the mapping of DNA methylation, we describe a protocol for gel-free multiplexed reduced representation bisulfite sequencing (mRRBS) that reduces the workload dramatically and enables processing of 96 or more samples per week. mRRBS achieves similar CpG coverage to the original RRBS protocol, while the higher throughput and lower cost make it better suited for large-scale DNA methylation mapping studies, including cohorts of cancer samples.
DNA methylation plays an important role in mammalian development [1, 2] and is frequently altered in diseases, including cancer . It is generally thought that methylation acts in a repressive function within regulatory contexts [4, 5]. DNA methylation in mammalian genomes occurs mostly within the context of the CpG dinucleotide  and is generally seen in CpG-poor regions. In contrast, CpG-rich regions naturally exhibit low methylation states [7–10].
Many techniques have been developed to investigate global DNA methylation patterns . Comparison of next-generation sequencing-based technologies showed that most methods produce similar results [12, 13], but that the optimal sequencing strategy may depend on sample DNA amount, as well as the desired genome coverage and sequencing depth [14, 15]. Whole-genome bisulfite sequencing of randomly sheared genomic DNA is the most comprehensive, but also most costly, method, while more focused approaches such as reduced representation bisulfite sequencing (RRBS) allow larger numbers of samples to be analyzed at reduced costs [8, 15–17].
RRBS utilizes the cutting pattern of MspI (C^CGG) to systematically digest DNA to enrich for CpG dinucleotides. As opposed to whole-genome bisulfite sequencing, every fragment produced by MspI digestion will contain DNA methylation information for at least one CpG dinucleotide . Another benefit of RRBS is that promoters, CpG islands, and other genomic features are disproportionally enriched genomic features because of the frequency of MspI cut sites in these regions [8, 16].
RRBS reduces the complexity of the genome - and thus the sequencing cost - by selecting a subset of MspI fragments based on their size for sequencing. In the standard RRBS protocol, this size selection is done by preparative gel electrophoresis, which is laborious and difficult to automate, thereby limiting the throughput of the method. For example, using our more recently published protocol , which includes a manual 40 to 220 bp size cut on an agarose gel, it is possible to produce around 12 to 24 RRBS libraries within a two-week time period. We reasoned that removing MspI fragments <40 bp by a simple clean-up protocol followed by bisulfite conversion, PCR and cluster amplification on an Illumina flowcell (all of which select against large fragments) could result in a similar size distribution of MspI fragments and comparable reduced representation of the genome as in the traditional, gel-based protocol. Taking advantage of increased sequencing throughput and the ability to barcode sequencing libraries, we have developed a new 'gel-free' multiplexed RRBS protocol, called mRRBS, which allows processing of samples in batches of 96 or more.
In addition to multiplexing and skipping the preparative gel, the mRRBS protocol was simplified and streamlined, eliminating several other steps of the original RRBS protocol. For example, the addition of Klenow fragment (3'→5' exo-) directly to the post-digested MspI/DNA mixture for end repair, and adding the A-tail minimizes clean-up steps and loss of material. The replacement of multiple phenol:chloroform steps described in the original RRBS method [8, 15] with a single solid phase reversible immobilization (SPRI) bead clean-up after adapter ligation also helped improve the ease and efficiency of the library generation process.
Rapid library generation using mRRBS will greatly increase the throughput while notably reducing the cost per sample. As a proof of concept, we show the generation of 96 libraries using the new mRRBS protocol and provide statistics as well as comparative performance measures of this improved method. To facilitate future large-scale studies we also provide detailed reagent lists and the costs (labor and reagents) per sample.
Results and discussion
Streamlining the RRBS protocol
First, enzymatic reactions were processed in a 96-well PCR plate using a 12-channel pipette (Figure S1 in Additional file 1). All 96 samples were quantified, and DNA samples were diluted to an equal concentration (20 ng/μl). DNA (5 μl, 100 ng) from each sample was used for the proof-of-concept experiment. Second, because both MspI digestion and end repair/A-tailing by Klenow fragment (3'→5' exo-) can be carried out in the same buffer, the MspI inactivation and removal steps were eliminated. The DNA purification after A-tailing and adapter ligation used a modified '1 tube with bead' SPRI clean-up method  in which several reaction steps are carried out in a single tube. Third, Illumina TruSeq adapters with unique six-base identifiers were used to tag and pool 12 samples early in the process and sequence them later in a single lane of Illumina Hi-Seq. In theory, many more samples can be tagged and processed as a pool. However, at this time only 24 different methylated TruSeq adapters are commercially available. Fourth, we simplified and streamlined the clean-up process. Traditional phenol extraction followed by ethanol precipitation is tedious and time consuming, but DNA purification before adapter ligation using regular spin columns significantly reduces recovery rates of small (<70 bp) DNA fragments. We eliminated these and also skipped the preparative electrophoresis step, relying on SPRI bead clean-up to remove small MspI fragments and bisulfite-induced DNA fragmentation, and amplification bias to select against unwanted large fragments. Together, these modifications reduce library processing time by about two days (Figure 1; Figure S1 in Additional file 1).
To reduce the occurrence of adapter dimers, we used a lower concentration of adapters (30 nM) than recommended by the manufacturer (see Materials and methods for details). In addition, we extracted the library DNA after the final PCR using two subsequent rounds of SPRI bead clean-ups to minimize primer-dimers in the final libraries (Figure S2 in Additional file 1).
Finally, to overcome density limitations and problems with cluster localization on the Illumina Hi-Seq flowcell due to non-random distribution of bases at the beginning of each read (each read starts with a C or a T, depending on the methylation status of the MspI site, followed by two Gs in a row), we implemented a custom Illumina sequencing protocol called 'dark sequencing'. In this custom protocol, no image is recorded during the first three sequencing cycles and cluster localization is deferred to cycles 4 through 7 (Figure S3 in Additional file 1). After cluster definition, the sequencing primer with seven newly synthesized bases attached is melted off and washed away. Fresh sequencing primer is annealed and the crucial first position that indicates the methylation status of the MspI site is determined at the beginning of a new 29-base read (see Materials and methods for details).
Validation of gel-free mRRBS libraries
We constructed 96 mRRBS libraries from one plate of DNA samples. When assayed on a gel, the size distribution of the final pooled PCR-amplified libraries was similar to that of the original RRBS protocol [8, 15] (Figure S2 in Additional file 1).
Summary of mRRBS performance
1× coverage CpG count
5× coverage CpG count
10× coverage CpG count
84 HQ samples
Summary for 12 RRBS and 12 mRRBS libraries
1× coverage CpG count
5× coverage CpG count
10× coverage CpG count
12 RRBS samples
12 mRRBS samples
Because the mRRBS protocol avoids the gel size selection step, in silico analysis was used to determine coverage rates for different sizes of fragments. The hg19 human genome was digested with MspI in silico, and the resulting fragments were binned by size. In Figure 2c, we measured the percentage of fragments of each size that were covered by at least one read. Fragments with a size range of 60 to 300 bp appear well-covered in most samples, with a slight, PCR-induced bias toward fragments of shorter length. In Figure 2d, coverage depth is shown for corresponding fragment sizes. For each bin of fragments with a certain length, the average coverage of all fragments of that size is reported, though this likely underestimates the actual average CpG coverage because artificially digested fragments that are not covered by any sequencing reads are included in the mean coverage calculation. On average, fragments with a size range of 60 to 300 bp are covered at least 5-fold, which is the recommended coverage threshold [8, 12]. Indeed, CpGs captured with at least 5× coverage correlate highly between sequencing runs of the same sample, whereas correlation between CpGs captured with a lower coverage show a lower correlation (Figure S4 in Additional file 1).
Comparison of genomic coverage
Assessment of PCR-induced chimeras
Barcoding DNA samples early in the process and bisulfite-converting and PCR-amplifying them as a pool contributes significantly to the overall ease and efficiency of the mRRBS protocol. However, pooling prior to PCR carries the risk of cross-sample confusion by chimeric events that cause reads from one sample to be associated with the barcode of another sample.
To assess the magnitude of this potential problem, we prepared a barcoded mRRBS library from in vitro CpG-methylated mouse DNA (95% of mRRBS reads were completely methylated) and a barcoded library from wild-type mouse DNA where many regions are far less methylated (45% of mRRBS reads were completely unmethylated). As shown in Figure S6 in Additional file 1, PCR-amplifying both libraries as a pool did not change the proportion of completely methylated, completely unmethylated and partially methylated reads assigned to each sample based on its respective barcode. This indicates a low or undetectable rate of PCR-induced chimerism in mRRBS libraries that does not affect interpretation or analysis of the generated libraries.
Moreover, when calculated, the rate of chimeric artifacts that join unrelated genomic loci in both RRBS and mRRBS data sets is extremely low. In the twelve mRRBS used to compare genomic coverage above, the average rate of one or more mismatches in read alignment was only 1.4 × 10-5. This indicates that such disruptive chimeras happen very infrequently.
Cost reduction and protocol efficiency
Cost comparison of RRBS and mRRBS
Total (96 samples) enzymes
Other supplies and sequencinga
Total (96 samples) other supplies and sequencing
Total (96 samples) supplies + salary
Total per sample cost
The mRRBS protocol presented here shortens the time required to produce bisulfite-converted libraries from 9 days in our previous RRBS protocol  to around 6 days (Figure 1). Moreover, by performing all initial library assembly steps in 96-well plates and by using unique, per sample, barcoded adapters, many more samples can be processed in parallel, making it possible to generate hundreds of libraries per month. The multiplexing adapters and Illumina Hiseq 2000 technology enable the sequencing of a dozen or more libraries per lane and substantially reduce the per sample cost. Despite these protocol modifications and simplifications, the coverage remains comparable to that described for the original RRBS protocol [8, 15], with a below threshold rate (<5 million aligned reads) in our proof of concept plate that is clearly acceptable given the higher throughput and lower cost when generating and sequencing libraries in parallel. DNA quantity or quality may be responsible for some of the observed variability in individual library performance and remains a critical part of the RRBS protocol that likely has stronger effects when using a multiplexed strategy. In summary, mRRBS allows throughputs comparable to array-based platforms such as the Illumina 450k, at a reduced cost, with better genomic coverage and lower genomic DNA input.
Materials and methods
Genomic DNA purification
Genomic DNA was isolated as previously reported [15, 19]. Purified DNA was quantified using a Quant-iT DNA Broad Range assay kit (Invitrogen, Grand Island, NY, USA, catalogue number Q-33130) and subsequently diluted to 20 ng/μl in low TE (10 mM Tris-HCl, 0.1 mM EDTA, PH 8.0). Equal amounts of DNA samples (100 ng) were added to distinct wells in a 96-well PCR plate (Axygen, Union City, CA, USA, catalogue number PCR-96M2-HS-C). For the chimera experiment, CpG Methylated NIH 3T3 mouse genomic DNA was purchased from New England Biolabs (Ipswich, MA, USA).
Samples of 5 μl genomic DNA were transferred to a new 96-well PCR plate with a 12-channel pipette. The MspI (New England Biolabs, catalogue number R0106L) digestion was conducted in a 30 μl reaction containing 3 μl of 10× NEB buffer 2, 1 μl of MspI (20 U/μl) and 21 μl H2O. To facilitate pipetting, a master mixture for 110 reactions, which compensates for reagent loss, was set up as follows: 330 μl of 10× NEB buffer 2, 110 μl of MspI and 2,310 μl of H2O. Next, 220 μl of the master mixture was added to each of the 12 wells in a row of a 96-well plate. Out of these, 25 μl were then pipetted to the sample/DNA plate using a 12-channel pipette. After carefully sealing the plate with one piece of adhesive tape sheet (Qiagen, Valencia, CA, USA, catalogue number 19570), the plate was then spun down briefly, vortexed to mix and was further spun for 30 s at 2,000 rpm in a PCR plate centrifuge. The plate was then incubated overnight at 37°C in an incubator. A diagnostic gel can be run on select samples at this point to determine MspI digestion efficiency, although this is usually not necessary (Figure S2a in Additional file 1).
Gap filling and A-tailing
Without deactivating MspI and cleaning-up the digestion reactions, DNA end repair and A-tailing were conducted by adding Klenow fragment (3'→5' exo-) (New England Biolabs, catalogue number M0212L) and dNTP mixture containing 10 mM dATP, 1 mM dCTP and 1 mM dGTP (New England Biolabs, catalogue number N0446S) directly into each well of the digestion plate. To simplify pipetting, an excessive amount of master mixture (110×) containing 110 μl of the Klenow fragment (3'→5' exo-) and 110 μl of the dNTP mix was made, and an aliquot of 18 μl was pipetted to each of the 12 wells in a clean row of a 96-well plate; 2 μl of that mix was added to each sample using a 12-channel pipette. Next, the sample plate was sealed and spun briefly to bring down any liquid accumulated on plate walls. The plate was vortexed to mix and spun for 30 s at room temperature using the plate centrifuge. The reaction was performed in a thermocycler (Eppendorf, Mastercycler EP Gradient S) without the heated lid. The program was set to 30°C for 20 minutes, 37°C for 20 minutes then 4°C indefinitely. The two temperatures are necessary for each step, the gap filling and the A-tailing, to facilitate both reactions.
A 2× concentration of SPRI AMPure XP beads (Beckman Coulter, Brea, CA, USA, catalogue number A63881; 64 μl beads for 32 μl sample) were added to each well using an 8-channel pipette. Beads and samples were mixed by pipetting up and down at least five times. Then, the mixtures were incubated at room temperature for 30 minutes. After DNA binding, the 96-well plate was placed onto a DynaMag™-96 Side magnet (Invitrogen, catalogue number 123-31D) for 5 minutes. The supernatant was carefully removed from the side opposite the accumulated beads, and the beads were then washed twice with 100 μl of 70% ethanol. Five minutes after the second wash, the ethanol was removed, and the duplex of the plate and the DynaMag™-96 Side magnet was put into a fume hood to dry the beads for 10 minutes. After drying of the beads, 20 μl of EB buffer (New England Biolabs, catalogue number B1561) was added to each well using an 8-channel pipette. The plate was then covered with a new tape sheet, vortexed to resuspend DNA, and spun down as described previously.
Multiplexed adapter ligation
A 110× ligation master mix was made for 96 reactions as follows: 330 μl of 10× T4 ligation buffer, 110 μl of T4 ligase (New England Biolabs, catalogue number M0202M), and 440 μl of H2O (1× volume: 3 μl of 10× T4 ligation buffer, 1 μl of T4 Ligase, 4 μl of H2O). Master mix (72 μl) was added to each of the 12 wells in a clean row of a 96-well plate. Next, 18 μl of each Illumina TruSeq adapter (Illumina, Dedham, MA, USA, catalogue number PE-940-2001; from a 1:20 diluted 9 μM stock) were added to corresponding wells in the row (Illumina TruSeq adapters contain 5 mC instead of C and can therefore be used for RRBS). After mixing the adapter-ligase mixtures, 10 μl of each was distributed to correlated samples using a 12-channel pipette. This brought the ligation reaction volume of each sample to 30 μl. The plate was placed into a thermocycler and incubated at 16°C overnight without the heated lid- the heated lid could potentially destroy the ligase.
Library pooling and bisulfite conversion
After ligation the plate was removed from the thermocycler and the beads were resuspended. Next, the plate was placed back into the thermocycler, and the enzyme was deactivated at 65°C for 20 minutes. It is important to note that the beads need to be resuspended prior to enzyme deactivation because resuspension is difficult after heating to 65°C. Samples were then pooled into eight 1.5 ml microfuge tubes. To bind the DNA back to the beads, a 2× solution (720 μl) of 20% polyethylene glycol (8,000 g/mol), 2.5 M NaCl was added to each tube. The samples were mixed and incubated at room temperature for 30 minutes to ensure maximum binding. After incubation, the samples were put onto a DynaMag™-2 magnet (Invitrogen, catalogue number 123-21D) and incubated for 5 minutes to allow bead attraction to the magnet. The liquid was removed, and the beads were washed with 1.0 ml of 70% ethanol. After removing the ethanol, the tubes were placed in the fume hood to dry the beads until cracks were observed (taking about 30 to 50 minutes). For eluting DNA from the beads, 25 μl of EB buffer was added to each tube; the tubes were vortexed for 20 s and were then centrifuged briefly. The tubes were placed back onto the magnet and the eluent (about 23 μl) was transferred to a new 1.5 ml microfuge tube. About 2 μl is lost due to adherence to the beads, and 3 μl of each sample was set aside for the ligation efficiency test by PCR as described previously , except that 0.3 μM of TruSeq primers (forward primer, 5'-AATGATACGGCGACCACCGAGAT-3'; reverse primer, 5'-CAAGCAGAAGACGGCATACGA-3'; Integrated DNA Technologies, Coralville, IA, USA) were utilized.
The remaining 20 μl samples were put through two consecutive bisulfite conversions, and bisulfite converted DNA was cleaned up as described in . After determining the optimized PCR cycle number for each sample, a large-scale PCR reaction (200 μl) for each sample was performed as recommended .
Final SPRI bead clean-up
After the PCR was completed, each well was pooled into a 1.5 ml tube. A 1.2× SPRI bead clean-up (240 μl SPRI beads into a 200 μl library pool) as mentioned above was conducted to remove PCR primers and adapter dimers. The DNA was eluted in 40 μl of EB buffer. To minimize adapter dimers, a second round of SPRI bead clean-up was performed at 1.5× (60 μl SPRI beads into a 40 μl library pool). The final library DNA samples were eluted with 40 μl EB buffer. The pooled libraries were quantified using a Qubit fluorometer (Invitrogen catalogue number Q32857) and a Quant-IT dsDNA HS assay kit (Invitrogen catalogue number Q-33120), and the qualities were determined by running a 4 to 20% Criterion precast polyacrylamide TBE gel (Bio-Rad, Waltham, MA, USA, catalogue number 345-0061). An equal quantity of starting genomic DNA prevents a bias toward more concentrated libraries, so accuracy in these measurements is imperative for sequencing success. The samples were sequenced on an Illumina Hiseq 2000 machine at the Broad Institute Sequencing Platform.
The MspI recognition cut site (C^CGG) creates fragments that will make the first three bases of every read non-random. This would result in high apparent cluster density, poor DNA cluster localization, and significant data loss during sequencing on the Illumina HiSeq 2000. To improve performance of these samples and increase coverage obtained, we used a method referred to as 'dark sequencing' in which imaging and cluster localization were delayed until the fourth cycle of sequencing chemistry, beyond the extent of bias from the MspI cut site (Figure S3 in Additional file 1).
To do this, we loaded a HiSeq 2000 with a custom recipe file co-developed with Illumina plus extra reagents to support primer re-hybridization. The custom recipe created a new initial 'template read' in which the first three biased bases were incorporated without imaging, followed by four cycles that were incorporated, imaged, and used by the sequencer for cluster localization. Next, the recipe removed the newly synthesized strand using NaOH and a buffer wash, re-hybridized fresh sequencing primer to the sample, and began read 1 data collection as usual from the first base but using the pre-existing cluster map or 'template' generated by the template read. HiSeq Control Software (HCS) provided by Illumina prevented cluster intensity files from the template read to enter downstream analysis.
As all custom chemistry steps were defined by the recipe, this workflow required very little additional hands-on time compared to a standard HiSeq run setup. The template read took approximately 6 h and consumed seven cycles of sequencing reagents prior to the start of data collection. Additional reagents to support re-hybridization after the template read were loaded at the beginning of the run alongside other read 1 and index read sequencing reagents. The following positions differed from the standard setup for an indexed single read run: Pos 16, 3 ml Read 1 Sequencing primer; Pos 18, 5 ml 0.1 N NaOH, Pos 19, 6 ml Illumina wash buffer.
After the removal of adapters and barcodes, 29 bp reads were aligned to the hg19 genome using MAQ. CpG methylation calling was performed by observing the bisulfite transformation in the read as opposed to the genome sequence.
RRBS data have been deposited at the Gene Expression Omnibus (GEO) under accession [GSE40429].
polymerase chain reaction
reduced representation bisulfite sequencing
short interspersed repetitive element
solid phase reversible immobilization
We would like to thank all members of the Meissner and Gnirke labs, in particular Christoph Bock, for helpful suggestions. We would also like to thank all members of the Broad Sequencing Platform for their continued support. KC was supported by the National Science Foundation Graduate Research Fellowship under Grant No. 112237. This work was funded by a Broad SPARC grant and the NIH Roadmap Epigenomics Program (U01ES017155). AM is supported by the Pew Charitable Trusts.
- Li E, Bestor TH, Jaenisch R: Targeted mutation of the DNA methyltransferase gene results in embryonic lethality. Cell. 1992, 69: 915-926. 10.1016/0092-8674(92)90611-F.PubMedView ArticleGoogle Scholar
- Okano M, Bell DW, Haber DA, Li E: DNA methyltransferases Dnmt3a and Dnmt3b are essential for de novo methylation and mammalian development. Cell. 1999, 99: 247-257. 10.1016/S0092-8674(00)81656-6.PubMedView ArticleGoogle Scholar
- Jones PA, Baylin SB: The epigenomics of cancer. Cell. 2007, 128: 683-692. 10.1016/j.cell.2007.01.029.PubMedPubMed CentralView ArticleGoogle Scholar
- Bird A: DNA methylation patterns and epigenetic memory. Genes Dev. 2002, 16: 6-21. 10.1101/gad.947102.PubMedView ArticleGoogle Scholar
- Meissner A: Epigenetic modifications in pluripotent and differentiated cells. Nat Biotechnol. 2010, 28: 1079-1088. 10.1038/nbt.1684.PubMedView ArticleGoogle Scholar
- Ziller MJ, Muller F, Liao J, Zhang Y, Gu H, Bock C, Boyle P, Epstein CB, Bernstein BE, Lengauer T, Gnirke A, Meissner A: Genomic distribution and inter-sample variation of non-CpG methylation across human cell types. PLoS Genet. 2011, 7: e1002389-10.1371/journal.pgen.1002389.PubMedPubMed CentralView ArticleGoogle Scholar
- Deaton AM, Bird A: CpG islands and the regulation of transcription. Genes Dev. 2011, 25: 1010-1022. 10.1101/gad.2037511.PubMedPubMed CentralView ArticleGoogle Scholar
- Meissner A, Mikkelsen TS, Gu H, Wernig M, Hanna J, Sivachenko A, Zhang X, Bernstein BE, Nusbaum C, Jaffe DB, Gnirke A, Jaenisch R, Lander ES: Genome-scale DNA methylation maps of pluripotent and differentiated cells. Nature. 2008, 454: 766-770.PubMedPubMed CentralGoogle Scholar
- Weber M, Hellmann I, Stadler MB, Ramos L, Paabo S, Rebhan M, Schubeler 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
- Smith ZD, Chan MM, Mikkelsen TS, Gu H, Gnirke A, Regev A, Meissner A: A unique regulatory phase of DNA methylation in the early mammalian embryo. Nature. 2012, 484: 339-344. 10.1038/nature10960.PubMedPubMed CentralView ArticleGoogle Scholar
- Beck S, Rakyan VK: The methylome: approaches for global DNA methylation profiling. Trends Genet. 2008, 24: 231-237. 10.1016/j.tig.2008.01.006.PubMedView ArticleGoogle Scholar
- Bock C, Tomazou EM, Brinkman AB, Muller F, Simmer F, Gu H, Jager N, Gnirke A, Stunnenberg HG, Meissner A: Quantitative comparison of genome-wide DNA methylation mapping technologies. Nat Biotechnol. 2010, 28: 1106-1114. 10.1038/nbt.1681.PubMedPubMed CentralView ArticleGoogle Scholar
- Harris RA, Wang T, Coarfa C, Nagarajan RP, Hong C, Downey SL, Johnson BE, Fouse SD, Delaney A, Zhao Y, Olshen A, Ballinger T, Zhou X, Forsberg KJ, Gu J, Echipare L, O'Geen H, Lister R, Pelizzola M, Xi Y, Epstein CB, Bernstein BE, Hawkins RD, Ren B, Chung WY, Gu H, Bock C, Gnirke A, Zhang MQ, Haussler D, Ecker JR, Li W, Farnham PJ, Waterland RA, Meissner A, Marra MA, Hirst M, Milosavljevic A, Costello JF: Comparison of sequencing-based methods to profile DNA methylation and identification of monoallelic epigenetic modifications. Nat Biotechnol. 2010, 28: 1097-1105. 10.1038/nbt.1682.PubMedPubMed CentralView ArticleGoogle Scholar
- Beck S: Taking the measure of the methylome. Nat Biotechnol. 2010, 28: 1026-1028. 10.1038/nbt1010-1026.PubMedView 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
- Gu H, Bock C, Mikkelsen TS, Jager N, Smith ZD, Tomazou E, Gnirke A, Lander ES, Meissner A: Genome-scale DNA methylation mapping of clinical samples at single-nucleotide resolution. Nat Methods. 2010, 7: 133-136. 10.1038/nmeth.1414.PubMedPubMed CentralView ArticleGoogle Scholar
- Meissner A, Gnirke A, Bell GW, Ramsahoye B, Lander ES, Jaenisch R: Reduced representation bisulfite sequencing for comparative high-resolution DNA methylation analysis. Nucleic Acids Res. 2005, 33: 5868-5877. 10.1093/nar/gki901.PubMedPubMed CentralView ArticleGoogle Scholar
- Fisher S, Barry A, Abreu J, Minie B, Nolan J, Delorey TM, Young G, Fennell TJ, Allen A, Ambrogio L, Berlin AM, Blumenstiel B, Cibulskis K, Friedrich D, Johnson R, Juhn F, Reilly B, Shammas R, Stalker J, Sykes SM, Thompson J, Walsh J, Zimmer A, Zwirko Z, Gabriel S, Nicol R, Nusbaum C: A scalable, fully automated process for construction of sequence-ready human exome targeted capture libraries. Genome Biol. 2011, 12: R1-10.1186/gb-2011-12-1-r1.PubMedPubMed CentralView ArticleGoogle Scholar
- Smith ZD, Gu H, Bock C, Gnirke A, Meissner A: High-throughput bisulfite sequencing in mammalian genomes. Methods. 2009, 48: 226-232. 10.1016/j.ymeth.2009.05.003.PubMedPubMed CentralView ArticleGoogle Scholar
- Heintzman ND, Hon GC, Hawkins RD, Kheradpour P, Stark A, Harp LF, Ye Z, Lee LK, Stuart RK, Ching CW, Ching KA, Antosiewicz-Bourget JE, Liu H, Zhang X, Green RD, Lobanenkov VV, Stewart R, Thomson JA, Crawford GE, Kellis M, Ren B: Histone modifications at human enhancers reflect global cell-type-specific gene expression. Nature. 2009, 459: 108-112. 10.1038/nature07829.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/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.