U N
C O
R R
E C
T E
D
P R
O O
F
1
2Whole genome DNA methylation analysis bad on high throughput 3quencing technology
4Ning Li a,b,c,1,Mingzhi Ye a,1,Yingrui Li a,1,Zhixiang Yan a ,Lee M.Butcher e ,Jihua Sun a ,Xu Han a ,5Quan Chen a ,Xiuqing zhang a ,Jun Wang a,d,*
6a
Beijing Genomics Institute at Shenzhen,Shenzhen 518000,China
7b
The Graduate University of Chine Academy of Sciences,Beijing 100062,China 8c
Beijing Institute of Genomics,Chine Academy of Sciences,Beijing 100029,China 9d
Department of Biology,University of Copenhagen,Copenhagen,Denmark 10
e
UCL Cancer Institute,Paul O’Gorman Building,University College London,72Huntley Street,London WC1E 6BT,UK
1113a r t i c l e i n f o 14Article history:
15Accepted 15April 201016Available online xxxx 17Keywords:
18DNA methylation 19Bisulfite-quencing 20MeDIP-quencing 21MBD-quencing
22
23a b s t r a c t
24There are numerous approaches to decipher a whole genome DNA methylation profile (‘‘methylome”),25each varying in cost,throughput and resolution.The gold standard of the methods,whole genome bisul-26fite-quencing (BS-q),involves treatment of DNA with sodium bisulfite combined with subquent 27high throughput quencing.Using BS-q,we generated a single-ba-resolution methylome in human 28peripheral blood mononuclear cells (in press).This BS-q map was then ud as the reference methylome 29to compare two alternative quencing-bad methylome assays (performed on the same donor of 30PBMCs):methylated DNA immunoprecipitation (MeDIP-q)and methyl-binding protein (MBD-q).In 31our analysis,we found that MeDIP-q and MBD-q are complementary strategies,with MeDIP-q more 32nsitive to highly methylated,high-CpG densities and MDB-q more nsitive to highly methylated,33moderate-CpG densities.Taking into account the size of a mammalian genome and th
e current expen 34of quencing,we feel 3gigabas (Gbp)45bp paired-end MeDIP-q or MBD-q uniquely mapped reads 35is the minimum requirement and cost-effective strategy for methylome pattern analysis.
36Ó2010Elvier Inc.All rights rerved.
37
3839 1.Introduction
40Cytosine DNA methylation is a major epigenetic system that 41modifies the DNA molecule itlf.In eukaryotes,DNA methylation 42is involved in diver process,including embryogenesis,genomic 43imprinting,X-inactivation and tumorigenesis in mammals,and 44transposon silencing in plants [1–4].In plant cells,methylated 45cytosine (5mC)occurs in CG-(CpG),CHG-and CHH-contexts [5].46In mammalian somatic tissues,DNA methylation occurs almost 47exclusively (99.98%)at CpG dinucleotides [6,7];recent evidence 48in human embryonic stem cells,however,has shown that nearly 49one-quarter of all 5mC occurs at non-CpGs,particularly within 50gene bodies [8].
51A detailed exploration of the role of DNA methylation in diver 52cellular process requires accu
rate and nsitive methods,of 53which veral genome-wide and whole genome assays exist.Gen-54ome-wide methods include combining enrichment-bad tech-55
niques such as methylated DNA immunoprecipitation (MeDIP)
56and Methyl-Binding Domain (MBD)protein with microarrays 57(chips).MeDIP us an antibody that recognizes 5mC to pull down 58the methylated fraction of a genome,while MBD exploits MECP2’s 59affinity for CpGs.MeDIP-chip and MBD-chip have been ud inde-60pendently by veral groups to study DNA methylation at a gen-61ome-wide scale in both plant [9]and human cells [10,11].
62With the advent of next generation quencing,many estab-63lished methods of genome interrogation were upscaled.Logically,64MeDIP was promoted from microarrays to next generation 65quencing:the result,MeDIP-q,continues to enrich for -66quences containing 5mC but is no longer limited to elucidating 67methylation at loci tiled by oligonucleotides on a microarray,and 68has been ud in plant genomics [12],as well as charting the first 69mammalian methylome [13].Similarly,MBD has been combined 70with next generation quencing and recently,three isogenic can-71cer cell lines were studied using what the authors term,MBD-iso-72lated Genome Sequencing [MiGS;14].
73Another method,reduced reprentation bisulfite-quencing 74[15],us enzymatic digestion to reduce genome complexity fol-75lowed by subquent bisulfite treatment and next generation 76quencing.Becau treatment of DNA with sodium bisulfite con-77verts all unmethylated cytosines to uracils,but leaves methylated 78
cytosines unchanged (albeit without the methyl group)[16],
1046-2023/$-e front matter Ó2010Elvier Inc.All rights rerved.doi:10.h.2010.04.009
*Corresponding author at:Beijing Genomics Institute at Shenzhen,Shenzhen 518000,China.Fax:+8675525273796.
E-mail address:wangj@genomics (J.Wang).1
The authors
contributed equally to this work.Methods xxx (2010)xxx–xxx
Contents lists available at ScienceDirect
Methods
journal homepage:www.el v i e r.c o m /
l o c a t e /y m e t h
Plea cite this article in press as:N.Li et al.,Methods (2010),doi:10.h.2010.04.009
U N
C O
R R
E C
T E
D
P R O O
F
79single-ba-resolution methylation maps can be achieved by map-80ping quence reads to in silco converted reference quences.This 81method has the advantage of low starting amounts of template 82(30ng)but interrogates approximately 10%of the genome.
83The gold standard however,for determining the DNA methylo-84me remains bisulfite-bad.In esnce,whole genome bisulfite-85quencing (BS-q)is whole genome requencing,preceded by
86treatment of genomic DNA with sodium bisulfite.Excluding repet-87itive regions,this technique is capable of determining the state of 88virtually all cytosines in the genome.But given the bioinformatic 89challenges of aligning bisulfite converted quences to in silico con-90verted reference genomes,single-ba-resolution methylome 91maps of Arabidopsis [17]and human [8]–although formidable 92achievements –have only emerged recently.
93Using BS-q,we deciphered an esntially complete whole 94genome human DNA methylome of peripheral blood mononuclear 95cells (PBMCs)(in press).We ud this methylome as a reference to 96evaluate two alternative whole methylome assays:MeDIP-q and 97MBD-q.Becau all assays were performed using the same do-98nor’s PBMCs,this is the first study to asss the true whole genome 99accuracy and nsitivity of MeDIP-q and MBD-q assays.Here 100we propo suggestions for MeDIP-q and MBD-q protocols 101
and compare the advantages of each.
102 2.Materials and methods 103 2.1.Reagents and equipment
104Reagents and equipment ud for immunoprecipitation,bisul-105fite conversion and library preparation are listed in Tables 1and 1062.IP buffer and digestion buffer were lf-prepared solutions
with 107
a receipt on their method introduction.
108 2.2.Establishing a standard reference DNA methylome in PBMCs
109The methylome data ud as a standard reference was gener-110ated as part of the YanHuang (YH)DNA methylome project (in 111press).YH is an Asian male who whole genome has been re-112quenced previously using genetic material from peripheral blood 113mononuclear cells (PBMCs)[18].The same sample of PBMCs ud 114in the whole genome requencing project was ud to create 115the reference methylome using BS-q,as well as the MeDIP-q 116and MBD-q experiments detailed in this paper.In total,117103.5Gbp of paired-end quence data were generated using BS-118q.Of the 18,962,679CpGs prent in the unique haploid part 119(2.21Gbp)of the YH reference genome quence,approximately 12099.86%were covered by at least one unambiguously mapped read 121of quality score >14on either strand and 92.62%were unambigu-122ously covered on both strands.Bad on read alignment to in silico 123converted non-CpG cytosines,the bisulfite conversion rate was 12499.8%,ensuring reliable ascertainment of CpG methylcytosines 125with a fal positive rate of <0.5%.The rate of unconverted 126non-CpG cytosin
es was 0.2%,indicating very low methylation lev-127els of the bas.BS-q of YH excluded non-CpG methylation and 128this single-ba-resolution methylome is the standard reference to 129
脸上有色斑estimate the accuracy and nsitivity of MeDIP-q and MBD-q.
130 2.3.DNA bisulfite protocol
担当
131In our experience,we recommend starting BS-q with 5l g 132DNA.Prior to library construction,DNA was fragmented using a 133Covarias sonication system with the parameters listed in Table 3;134the resulting DNA smear should be enriched between 100and 135
500bp.
136Following fragmentation,libraries were constructed using the 137Illumina Paired-End protocol consisting of end repair,<A>ba 138addition and methylated-adaptor ligation.Ligated DNA was bisul-139fite converted using the EZ DNA Methylation-Gold kit (ZYMO).140DNA libraries of different inrt size were excid from the same 141lane of a 2%TAE agaro gel.Normally,three bands are excid cor-142responding to DNA inrt sizes of 80–100bp,100–120bp to 120–143150bp.Products were purified by using QIAquick Gel Extraction 144kit and amplified by PCR.PCR was carried out in a final reaction 145volume of 50l l consisting of 20l l purified DNA,4l l 2.5mM 146
dNTP,5l l 10Âbuffer,0.5l l JumpStart™Taq DNA polymera,
Table 1Reagents.Reagents
Catalog No.Supplier 10ÂPolynucleotide kina buffer B904Enzymatics
dNTP solution t N201L T 4DNA polymera P708L Klenow enzyme照葫芦画瓢
P706L T4polynucleotide kina Y904L 10Âblue buffer B0111mM dATP
Klenow (30–50exo-)
P701-LC-L 2ÂRapid ligation buffer B101
T4DNA liga (rapid)L603-HC-L
PE adapter oligo mix Takara/IDT/Illumina oligo kit
Index PE adapter oligo mix Takara/IDT/Illumina oligo kit Phusion DNA polymera F-531L
NEB
PCR primer PE 1.0Takara/IDT/Illumina oligo kit
PCR primer PE 2.0
Takara/IDT/Illumina oligo kit
Index PCR primer PE
Takara/IDT/Illumina oligo kit 50bp ladder marker MD108-01TIANGEN DL2000marker MD114-02TIANGEN k -HindIII marker D3403A TaKaRa 50bp DNA ladder
N3236L NEB EZ DNA Methylation-Gold kit™D5005ZYMO
Magnetic Methylated DNA Immunoprecipitation kit
mc-magme-048
Diagenode
MethylMiner™Methylated DNA Enrichment kit
ME10025Invitrogen Anti-5-methylcytosine mou mAb (16233D3)
NA81Merk Immunoprecipitation kit–Dynabeads Ò
Protein A
100.6D 100.6D Immunoprecipitation kit –Dynabeads ÒProtein G 100.7D 100.7D Pfx
马灯C11708-021Invitrogen JumpStart™Taq DNA polymera D9307Sigma
Quant-iT™dsDNA HS Assay kit Q32851Invitrogen QIAquick Gel Extraction kit 28706Qiagen Proteina K
PB0451–50mg
Sangon
Table 2
日历用英语怎么说Equipments.Equipments
Catalog No.
Supplier
PCR thermal cycler Veriti thermal cycler ABI Agilent 21002100Bioanalyzer Agilent
NanoDrop 1000
Spectrophotometer Thermo Fisher Scientific Covaris sonication system S-2
Covaris Thermomixer Thermomixer comfort Eppendorf Centrifuge 5417R
Eppendorf Qubit
Invitrogen
2
N.Li et al./Methods xxx (2010)xxx–xxx
Plea cite this article in press as:N.Li et al.,Methods (2010),doi:10.h.2010.04.009
U N
C O
R R
E C
阻力线
T E
D
P R
O O
F
1472l l PCR primers and 37.5l l UltraPureTM Water and the following 148thermal cycling program:94°C 30s,10cycles of 94°C 30s,60°C 14930s,72°C 30s then prolong with 1min at 72°C and products 150could be hold at 4°C.PCR products were quenced with an Illu-151
mina GAII.
152 2.4.Methylated DNA immunoprecipitation (MeDIP)quencing 153protocol
154There are veral versions of MeDIP,each with slight variations 155of antibody to DNA ratio and IP incubation time.In this paper,we 156ud two methods and we evaluate their efficiency and data qual-157ity:(1)a lf-prepared method,developed by scientists and (2)a 158
commercial kit-bad method,also capable of being automated.
159 2.4.1.MeDIP using lf-prepared method
堆雪人图片160A library was prepared from 10l g original gDNA;DNA was 161fragmented with the same parameters as the DNA bisulfite proto-162col.End repair,<A>ba addition and adaptor ligation steps were 163performed using Illumina’s Paired-End DNA Sample Prep kit fol-164lowing the manufacturer’s instructions.Immunoprecipitation was 165performed after adaptor ligation.
166Methylated DNA was pulled-down with Weber et al.’s method 167[10,11,19]with a few revisions.Ten microgram of mono-clonal 1685mC antibodies were incubated for 2h with 4l g DNA at 4°C with 169overhead shaking.One-hundred microliter of Dynabeads A and G 170were pre-washed three times with 800l l PBS–BSA 0.1%,respec-171tively.Beads were resuspended in 50l l 1ÂIP buffer which con-172tained 100mM Na-phosphate pH 7.0,0.14M NaCl and 0.05%173Triton X-100,added to the DNA–antibody mixtures,and incubated 174for 2h at 4°C with overhead shaking.The Bead–DNA–
antibody 175mixture was washed three times with 700l l 1ÂIP buffer with 176magnetic-bead capture in between;washed supernatant was kept 177in clean tube for qPCR control.Beads were resuspended in 250l l 178digestion buffer which contained 50mM Tris pH 8.0,10mM EDTA 179and 0.5%SDS and 7l l proteina K (10mg/ml;Sangon),then incu-180bated for 3h at 50°C with shaking to prevent the beads from 181ttling.
182The methylated fraction was purified on a single ZYMO DNA 183Clean &Concentrator-5column following the manufacture’s 184instructions.DNA was eluted in 27l l buffer EB and its concentra-185tion measured.MeDIP DNA was amplified by adaptor-mediated 186PCR in a final reaction volume of 50l l consisting of 23l l purified 187DNA,25l l Phusion DNA polymera mix and 2l l PCR primers .188Amplification consisted of 94°C 30s,16cycles of 94°C 30s,18960°C 30s,72°C 30s then prolong with 1min at 72°C and prod-190ucts could be hold at 4°C.Amplifications quality and quantity were 191evaluated by Agilent 2100Analyzer purified by 2%agaro gel and 192eluted in 15l l buffer EB.The recovery efficiency was evaluated 193with qPCR and data showed on Table 4.
1942.4.2.MeDIP using kit method
195A library was prepared from 5l g original gDNA and DNA was 196fragmented with the same para
meters as the DNA bisulfite proto-197col .End repair,<A>ba addition and adaptor ligation steps were 198performed using Illumina’s Paired-End DNA Sample Prep kit fol-199lowing the manufacturer’s instructions.Adaptor-ligated DNA was 200immunoprecipitated by anti-5mC,and MeDIP products were vali-201dated by qPCR using SYBR green mastermix (Applied Biosystems)202and primers for positive and negative control regions supplied in 203the MeDIP kit (Diagenode).qPCR validation procedures consist of 20495°C 5min,followed by 40cycles 95°C 15s and 60°C 1min.205MeDIP DNA was purified with ZYMO DNA Clean &Concentra-206tor-5column following the manufacture’s instructions and ampli-207fied by adaptor-mediated PCR in a final reaction volume of 50l l 208which consisting the same reagents with lf-prepared method .209After excising amplified DNA between 220and 320bp on a 2%aga-210ro gel,amplification quality and quantity were evaluated using 211an Agilent 2100Analyzer and DNA 1000chips.The recovery effi-212ciency was evaluated with qPCR and data showed on Table 4.
2132.5.MBD (Methylated DNA Binding Domain)quencing protocol 214A library was prepared from 10l g original gDNA and DNA was 215fragmented with the same parameters as the DNA bisulfite proto-216col then purified with QIAquick PCR purification kit.End repair,217<A>ba addition and adaptor ligation steps were performed using 218Illumina’s Paired-End DNA Sample Prep kit followin
g the manufac-219turer’s instructions.Adaptor-ligated DNA was incubated with the 220MBD proteins and Dynabeads for 1h at room temperature with 221overhead shaking.Bead–DNA–MBD mixtures were eluted in a 222stepwi gradient of increasing NaCl concentration with 200,223350,450,550,600,700,800,1000and 2000mM.The recovered 224DNA in each fraction was largely double-stranded.Libraries were 225quality controlled by PCR and Sanger quencing before Illumina 226GAII single-end quencing.However,DNAs eluting from 600227and 700mM NaCl concentrations exhibited higher duplication 228than 550mM NaCl concentration.To remedy this we quenced li-229braries belonging to two parately constructed batches:(1)the 230initial 550mM NaCl elution and (2)a parately prepared t of li-231braries from 550,600,700and 800mM NaCl elutions.Single-end 232quencing was ud to quence both first and cond batches.233A comparison of qPCR values and summary quencing statistics 234are shown in Tables 5and 6,respectively.2352.6.Experiences and troubleshooting
2362.6.1.DNA quantity,fragment size and amplification factors
237For BS-q,5l g DNA was needed for library preparation.For 238
paired-end methylated library quencing,we prefer to exci
Table 4
MeDIP real-time quantitative PCR results.Recovery (%)=%(meDNA-IP/Total input)=2[(Ct(10%input)À3.32)ÀCt(meDNA ÀIP)]Â100%Specificity (%)=1À(Enrichment ummeth/Enrichment meth)Â100.Method Primers
Primer quence
Recovery (%)Specificity (%)
Self-prepared method
P4994
F:GGGAATATAAGGAGCGCACA 44.098.3
R:TCGGTTAAAACGGTCAGGTC
P8804F:CGAGGCGTGAGTTATTCCTG 1.0–R:CTCTTGTGGCTGAGCTCCTT
MeDIP-kit
Diagenode +
–
57.298.8Diagenode À
–
0.7
–
Table 3
Sonication parameters to enrich the DNA to between 100and 500bp.Treatment 1
Duty/cycle (%)10Intensity 5Cycle/burst 200Time (s)960Treatment 2Time (s)0Treatment 3Time (s)0Treatment 4Time (s)
0Cycle
16
N.Li et al./Methods xxx (2010)xxx–xxx
3
Plea cite this article in press as:N.Li et al.,Methods (2010),doi:10.h.2010.04.009
U N
C O
R R
E C
T E
D
P R O O
F
239
methyl-adaptor-ligated DNA from the agaro gel below 150bp,as 240longer DNA library sizes have lower conversion rates after sodium 241bisulfite treatment.
242Our data and experience and suggest that three different size 243ranges from 80to 150bp should be gel purified,then subject to 244two independent adapter-mediated PCR reactions with between 2454–6cycles,depending on the amount of starting DNA:we strongly 246recommend 6cycles when starting with 5l g gDNA and 4cycles 247when starting with 10l g.For MeDIP and MBD libraries however,248we recommend between 12and 14cycles due to DNA loss in-249curred during the methylation enrichment steps.
250In total,all three technologies focud on 150bp or smaller in-251rt DNA fragment size and they need at least 5l g starting DNA 252amount with our quencing conditions.As part of our ongoing 253work we are investigating the effect of using much longer inrt 254sizes using lower starting DNA material with the aim of increasing 255output.
256 2.6.2.Antibody and protein quantity
257Anti-5mC and MBD proteins play key roles in binding methyl-258ated DNA.However,the ratio of proteins and DNA was controver-259
sial;using the kit-bad MeDIP method,our data showed that
260150ng anti-5mC was sufficient to successfully enrich 1l g adap-261ter-ligated DNA for its methylated fraction,while 10l g anti-5mC 262successfully enriched 4l g adapter-ligated DNA using the lf-pre-263
pared method.
2643.Results
2653.1.MeDIP-quencing data analysis by lf-prepared MeDIP method 266In total,7.9Gbp 45bp pair-ends data were generated by an Illu-267mina Genome Analyzer II for MeDIP-quencing,of which 4.9Gbp 268(62.4%)were uniquely mapped onto the reference genome.Using 269MACS [20]583,142peaks were identified in 7.9Gbp MeDIP-270quencing data (Table 7).To examine the effect of incread -271quence data on patterns of enrichment,MeDIP-q data were ran-272domly extracted from the total 7.9Gbp.The number of CpG Islands 273(CGIs)captured by peaks
was calculated for cumulative data bins 274reprenting 10–100%of total MeDIP data.Using the reference 275methylome,the number of MeDIP-peaks that overlapped with CGIs 276suggests that MeDIP data were largely enriched for highly methyl-277ated,high-CpG den regions.Fig.1for example,shows a law of 278diminishing returns with just 40%of the total MeDIP data enough 279to capture 89.9%of methylated CGIs captured using all 7.9Gbp 280MeDIP data.Conquently,and considering current quencing 281costs,we recommend that 4Gbp MeDIP-q data are sufficient 282for a CGI-centric approach to study methylation patterns.
283We also found that by increasing MeDIP-q data,the number 284of CGIs captured by MeDIP-peaks is much more than the number 285of MeDIP-peaks captured by CGIs.In other words,for every addi-286tional MeDIP-peak identified,multiple distinct CGIs are captured.287However,MeDIP-peaks identified with 2Gbp quencing data gen-288erally overlapped only one CGI,suggesting that it is preferable to 289generate clor to 3Gbp quencing data.
290We next sought to quantitate the methylation level and CG den-291sity of the peaks that were enriched for using MeDIP-q.Using BS-292q data,DNA methylation levels were calculated in 43,242,227293300bp windows that satisfied four criteria:(1)ba quality of -294
quenced data P 15;(2)CpG duplication 61.5;(3)at least one CpG
Table 6
MBD libraries evaluated by Sanger quencing.Batch Library Total reads Mapped reads Map rate (%)Duplicate rate (%)1st
library t
550mM (200–300bp)
42421000.0550mM (300–400bp)4040100 5.1
600mM-13737100 5.6700mM-1393897 2.7600mM-2676710013.6700mM-279779715.82nd
library t
550mM 30301000.0600mM 45451000.0700mM 4139950.0800mM
45
44
98
0.0
Table 5
Real-time quantitative PCR results on MBD.Method Primers Primer quence NaCl eluting
concentration (mM)Methyl control DNA qPCR Ct value (1st library t)Methyl control DNA qPCR Ct value (2nd library t)MBD
control
Methylated DNA control
F:ACA GGG CGT GTT AAC GAT ATA A
200
25.6
–350
24.2
–450
22.3
–
乞伏炽磐550
19.2
21.3R:CGC TGG TAG GAA CGA GAG TC
60017.320.270022.919.80022.122.6100024.9–2000
25.8
25.2
Non-methylated DNA control
F:GTC GCC ACA CCA ATT CGT TAC TCA
20031.7–35027.9–45029.6–55030.823.5R:AGA TCC CTG ACT GAA TTG CGA CGA
60032.125.870032.827.480032.527.3100032.7–
2000
32.4
27.3
4
N.Li et al./Methods xxx (2010)xxx–xxx
Plea cite this article in press as:N.Li et al.,Methods (2010),doi:10.h.2010.04.009
U N
C O
R
D
P R
O O
F
295was quenced in 300bp window and (4)the total depth of all 296CpGs in each 300bp window covered by unique mapped tag 297P 20.We term windows matching the criteria,‘‘eligible win-298dows”.We then filtered eligible windows to include only tho 299that overlapped with MeDIP-peaks identified using MACS and 300found the resulting t of eligible windows were characterized by 301high-CpG density and high DNA methylation level.
302We then examined the effect of incread quencing data on 303the CpG density and methylation level of the eligible windows 304overlapped with MeDIP-peaks.From the entire 7.9Gbp quencing 305data,enrichment maps were simulated using randomly extracted 306data amounting to 1.8,3.6,5.4and 7.9Gbp.The main effect of 307increasing quence data from MeDIP-q experiments is the in-308cread capture of intermediate-CpG density MeDIP-peaks.For 309example,the heat maps illustrated in Fig.2a–d show that as more 310quence data are incorporated,MeDIP enrichment migrates more 311obviously from right (high-CpG density)to left (relatively modest 312CpG density)than it does from top (high methylation level)to bot-313tom (low methylation),with the most profound change occurring 314between 1.8and 3.6Gbp.
315Eligible windows were next filtered to focus on regulatory re-316gions [2Kbp upstream of TSS,50-untranslated region (UTR)and 317coding quence (CDS)].Unsurprisingly,Fig.4illustrates th
at MeD-318IP-q data enriches for high-CpG densities and high methylation 319
levels in regulatory regions but that the patterns do not reflect
320BS-q profiles.Similarly,this relationship also holds for genomic 321repeat features Fig.S-1.When analyzed this way (i.e.,untrans-322formed enrichment),it is clear that MeDIP profiles differ dramati-323cally from bisulfite-quencing profiles,which tend to exhibit a 324narrower range of methylation levels across the continuum of 325CpG density.Converly,when enrichment is ud to index abso-326lute methylation,MeDIP profiles show a more pronounced spread 327of methylation levels.Therefore,MeDIP enrichment profiles are 328best ud to distinguish whether high-CpG density regions are 329highly methylated or not,but not recommended to decipher a 330methylome on the basis of enrichment data only.We note,how-331ever,that the relationship between enrichment and CpG density 332is not new,and methods are available [13,21]to account for this 333relationship,allowing correction in silico to better estimate abso-334
lute methylation (e Section 4).
3353.2.The MBD-quencing data analysis of various elution 336concentration
337Becau the MBD protocol offers differential fractionation of 338captured DNA bad on CpG methylation density,the various salt 339concentrations of 600,700and 800mM NaCl were chon to elute 340highly methylated CpG-rich regions.Approximately 750Mbp 34135bp single-end reads were generated for each salt concentration 342library,and MACS identified peaks associated with enrichment.343Fig.3shows the effect of increasing NaCl concentration on the 344DNA methylation level and CpG density of eligible windows over-345lapped with MBD-peaks.As NaCl concentration increas from 600346to 800mM,the enriched DNA profiles were compresd to retain 347higher CpG density regions.Fig.3D shows that 800mM salt con-348centration removed many of the highly methylated DNA fragments 349and the enrichment result of 600and 700mM is similar.Fig.3E de-350scribes the effect of NaCl concentration on MBD-q’s nsitivity to 351CpG density,800mM shows highest affinity for high-CpG density,352whereas 600mM enriched for lower CpG density DNA fragments.353We feel that 700mM is the most preferable salt concentration to 354elute in as it strikes a good balance between CpG density and 355
DNA methylation level.
3563.3.Comparison of MeDIP-quencing and MBD-quencing
357To compare MeDIP-q and MBD-q we generated $4.4Gbp 358paired-end reads from the 700mM NaCl concentration.Table 8359summarizes the captured data of $7.9Gbp MeDIP-q total data 360and $4.4Gbp MBD8-q total data.To compare affinity bias 361associated with MeDIP-q and MBD-q,we ud 50%of the 3627.9Gbp MeDIP-q data and 100%of 4.4Gbp MBD-q data.We 363ud BS-q data to asss the relative nsitivity of MeDIP-q 364
and MBD-q for enrichment of regions varying in CpG density
Table 7
In total,$7.9Gbp MeDIP-q data were generated.MeDIP-q data (10–100%)were randomly extracted from the total 7.9Gbp to reveal diminishing returns with respect to incread quence data.#of total
MeDIP data (%)Read length
#of peaks
Peaks coverage (bp)
#of CpG in peaks
#of CpG
(methylation rate P 60%)in peaks #of peaks overlapped with CGIs UCSC)#of CGIs overlapped with peaks (UCSC)1044PE 134,16142,598,3312934,1471374,736510551562044PE 261,12686,249,4995401,4892441,493657266873044PE 342,502124,512,4607261,1823202,459718173554044PE 411,075161,828,8768910,0043855,392752177585044PE 447,598192,341,04610,107,4124316,090767779526044PE 485,791218,703,61711,121,7284714,922788182007044PE 526,063237,496,87011,868,0705014,973807184078044PE 542,093261,669,78412,674,9225320,906816585259044PE 584,333270,371,59013,079,0015491,49083428704100
44PE 583,142295,484,76013,813,231
5760,598
8357
8760
N.Li et al./Methods xxx (2010)xxx–xxx
5
Plea cite this article in press as:N.Li et al.,Methods (2010),doi:10.h.2010.04.009