Edinburgh Research Explorer

Immortality but not oncogenic transformation of primary humancells leads to epigenetic reprogramming of DNA methylation andgene expression

Citation for published versionGordon K Clouaire T Bao XX Kemp SE Xenophontos M de Las Heras JI amp Stancheva I 2014Immortality but not oncogenic transformation of primary human cells leads to epigenetic reprogramming ofDNA methylation and gene expression Nucleic Acids Research vol 42 no 6 pp 3529-3541httpsdoiorg101093nargkt1351

Digital Object Identifier (DOI)101093nargkt1351

LinkLink to publication record in Edinburgh Research Explorer

Document VersionPublishers PDF also known as Version of record

Published InNucleic Acids Research

Publisher Rights Statementcopy The Author(s) 2013 Published by Oxford University PressThis is an Open Access article distributed under the terms of the Creative Commons Attribution License(httpcreativecommonsorglicensesby30) which permits unrestricted reuse distribution and reproduction inany medium provided the original work is properly cited

General rightsCopyright for the publications made accessible via the Edinburgh Research Explorer is retained by the author(s)and or other copyright owners and it is a condition of accessing these publications that users recognise andabide by the legal requirements associated with these rights

Take down policyThe University of Edinburgh has made every reasonable effort to ensure that Edinburgh Research Explorercontent complies with UK legislation If you believe that the public display of this file breaches copyright pleasecontact openaccessedacuk providing details and we will remove access to the work immediately andinvestigate your claim

Download date 22 May 2021

Immortality but not oncogenic transformationof primary human cells leads to epigeneticreprogramming of DNA methylation andgene expressionKatrina Gordon Thomas Clouaire Xun X Bao Sadie E Kemp Maria Xenophontos

Jose Ignacio de Las Heras and Irina Stancheva

Wellcome Trust Centre for Cell Biology School of Biological Sciences University of Edinburgh Michael SwannBuilding Mayfield Road Edinburgh EH9 3JR UK

Received August 18 2013 Revised December 3 2013 Accepted December 5 2013

ABSTRACT

Tumourigenic transformation of normal cells intocancer typically involves several steps resulting inacquisition of unlimited growth potential evasionof apoptosis and non-responsiveness to growthinhibitory signals Both genetic and epigeneticchanges can contribute to cancer developmentand progression Given the vast genetic heterogen-eity of human cancers and difficulty to monitorcancer-initiating events in vivo the precise relation-ship between acquisition of genetic mutations andthe temporal progression of epigenetic alterationsin transformed cells is largely unclear Here weuse an in vitro model system to investigate the con-tribution of cellular immortality and oncogenictransformation of primary human cells to epigeneticreprogramming of DNA methylation and gene ex-pression Our data demonstrate that extension ofreplicative life span of the cells is sufficient toinduce accumulation of DNA methylation at genepromoters and large-scale changes in gene expres-sion in a time-dependent manner In contrastcontinuous expression of cooperating oncogenesin immortalized cells although essential for anchor-age-independent growth and evasion of apoptosisdoes not affect de novo DNA methylation at pro-moters and induces subtle expression changesTaken together these observations imply thatcellular immortality promotes epigenetic adaptationto highly proliferative state whereas transforming

oncogenes confer additional properties to trans-formed human cells

INTRODUCTION

It is widely recognized that tumours and tumour-derivedcell lines exhibit altered patterns of DNA methylation andgene expression in comparison with normal tissues andprimary cells Gain of DNA methylation at normallyDNA methylation-free gene promoters and extensiveloss of DNA methylation throughout the genome havebeen detected in a variety of tumour types (1ndash4)Aberrant methylation of gene promoters can lead tostable silencing of tumour suppressor genes and consti-tutes an alternative mechanism to genetic loss of genefunction that can be brought about by mutations dele-tions and chromosomal rearrangements (134) Loss ofDNA methylation from repetitive sequences is thoughtto promote genomic instability which often accompaniescancer progression (56)Despite the wealth of data documenting these findings

it is largely unclear when and how the changes in DNAmethylation occur in transformed human cells (3)Tumours usually initiate from a small number of mutantcells and these tumour-initiating cells are difficult todetect isolate and monitor in long-term studies (7)Similar limitations apply to most available mouse cancermodels The vast majority of epigenetic studies on humancancers are carried out either on limited amount of clinicalmaterial isolated from patients when the disease is welladvanced or on cell lines established from tumours andmaintained in culture for extended periods of timeAlthough data indicating strong correlation between

To whom correspondence should be addressed Email istanchevaedacukPresent addressesKatrina Gordon Institute of Immunity and Infection Research University of Edinburgh West Mains Road Edinburgh EH9 3JT UKThomas Clouaire LBCMCP Universite Paul SabatierndashCNRS UMR 5088 31062 Toulouse FranceMaria Xenophontos The EMBL-European Bioinformatics Institute Wellcome Trust Genome Campus Hinxton Cambridge CB10 1SD UK

Published online 26 December 2013 Nucleic Acids Research 2014 Vol 42 No 6 3529ndash3541doi101093nargkt1351

The Author(s) 2013 Published by Oxford University PressThis is an Open Access article distributed under the terms of the Creative Commons Attribution License (httpcreativecommonsorglicensesby30) whichpermits unrestricted reuse distribution and reproduction in any medium provided the original work is properly cited

at Edinburgh U

niversity on October 3 2014

httpnaroxfordjournalsorgD

ownloaded from

accumulated epimutations and tumour gradetype areavailable for colon lung prostate and breast cancer(8ndash11) the precise timing of the initial methylationevents and the progression of epigenetic alterations inhuman cells undergoing tumourogenic transformationhave been difficult to estimate due to the vast genetic het-erogeneity of human cancers In most cases it is extremelychallenging to determine the precise relationship betweengenetic background oncogenic mutations genomicinstability and detected epigenetic changes (12)To circumvent these limitations and generate a cancer

model system amenable to long-term tracking of epigeneticevents and further mechanistic studies we used an estab-lished method to transform human somatic cells in vitrousing a combination of well-defined factors (13) We estab-lished isogenic immortalized and transformed human celllines derived from primary foetal lung fibroblasts (MRC-5)and followed the temporal changes in gene expression andDNAmethylation at gene promoters in these independentbut related to each other cell populations Our analysesshow that MRC-5 cells immortalized by expressionof human telomerase reverse transcriptase (hTERT)catalytic subunit and transformed MRC-5 cells express-ing hTERT SV40 large T-antigen (T-Ag) and constitu-tively active oncogenic H-RASGV12 progressivelyaccumulate extensive changes in gene expression andde novo DNA methylation at gene promoters thatbecome apparent after 50 population doublings (pd) inculture Remarkably de novo DNA methylation at genepromoters occurred at specific loci with similar timingin both the immortalized and transformed cell lines sug-gesting that gain of DNA methylation does not requireexpression of oncogenes The accumulation of DNAmethylation at gene promoters took place predominantlyat genes that were transcriptionally inactive in the parentalcell line but did not correlate with pre-existing Polycomb-dependent H3K27 trimethylation (H3K27me3) previouslyreported to pre-mark promoters for de novo DNA methy-lation (14ndash16) Importantly immortalized and trans-formed cell lines displayed different gene expressionprofiles indicating that the presence of oncogenes modu-lates the properties of immortal cells Our data demon-strate that programmed de novo DNA methylation atspecific loci and adaptation of transcriptional output ofthe genome to a highly proliferative state can occur indiploid human cells without a major input from oncogenicproteins On the other hand transforming oncogenes con-tribute to further modulation of gene expression andpromote evasion of apoptosis and anchorage-independentgrowth which are essential properties of cancer cells

MATERIALS AND METHODS

Cell lines and viral infections

The human male foetal lung fibroblast cell line MRC-5(ATCC number CRL-171) and all MRC-5-derivedcells were cultured in MEM (Life Sciences) supplementedwith 10 foetal calf serum 1mM non-essential aminoacids 1mM sodium pyruvate 100Uml penicillin1mgml of a streptomycin and 2mM L-glutamine The

pBABE-Neo-hTERT pBABE-Hygro-SV40 T-Ag andpBABE-Puro-H-RASV12G plasmids were packaged intoretroviral particles in amphotropic Phoenix A cell lineCulture supernatants were harvested 48 h later and theretroviral titres determined by infection of NIH-3T3mouse fibroblast cells The MRC-5hTERT cell line wasgenerated by infecting 105 MRC-5 cells with pBABE-Neo-hTERT retroviral particles at multiplicity ofinfection (MOI)=1 in the presence of 4 mgml polybreneAfter a 7-day selection with 250 mgml G418 drug-resist-ant colonies were pooled and designated as passage 1 TheMRC-5TSR cell line was generated by infecting the MRC-5hTERT with retroviral particles carrying pBABE-Hygro-T-Ag and subsequently after selection with 150mgmlhygromycin for 7 days with packaged pBABE-Puro-H-RASV12G Selection with 1 mgml puromycin wasapplied and resistant cells were pooled and designated aspassage 1

Growth curves

A total of 5104 cells were initially seeded (cell input= n0)into six-well dishes and the cell yields (n) were recordedat each passage and the population doublings calculatedfrom the formula MPD (mean population doubling)=332 (log10 n- log10 n0)

Soft agar assays

A bottom layer of 10ml 16 agar (BioGene) in MRC-5growth medium was prepared in 10-cm tissue culturedishes and allowed to solidify Each cell line (set up intriplicate) was seeded at a density of 105 cellsdish in a4ml top layer of 08 agar in MRC-5 growth mediumCells were incubated at 37C with weekly overlays with4ml of top layer agar without cells After 6ndash8 weekscolonies were scored blind

Telomere repeat amplification protocol

Cell extracts were prepared and assayed for telomeraseactivity using the TRAPeze kit (Millipore) following themanufacturerrsquos instructions Pilot experiments were ini-tially carried out on a range of protein concentrations(02ndash2 mg) to determine the linear range for each cell lineAssays were routinely carried out using 1 mg of proteinextract for each cell line The polymerase chain reaction(PCR) products were resolved on 10 non-denaturingpolyacrylamide gel visualized by staining with SYBRGreen and scanned at 473 nm on FLA-5100 scanner

Telomere restriction fragment Southern blots

Five microgram of genomic DNA prepared from each ofthe cell lines was digested with Hinf I and Rsa I restrictionenzymes overnight at 37C The digests were resolved in a1 TrisndashAcetatendashEDTA gel and transferred to Zeta-Probe GT membrane (BioRad) with 04M NaOH for16 h The DNA was cross-linked to the membrane usinga ultraviolet cross-linker (Stratagene) set on autocross-linkmode (120 000 mJoules) and the blot hybridized with aradiolabelled oligonucleotide probe [(TTAGGG)3] at42C overnight in buffer containing 1mM ethylenediami-netetraacetic acid (EDTA) 05M NaHPO4 and 7sodium dodecyl sulphate (SDS) The blot was

3530 Nucleic Acids Research 2014 Vol 42 No 6

at Edinburgh U

niversity on October 3 2014

httpnaroxfordjournalsorgD

ownloaded from

subsequently washed with 3 saline-sodium citrate (SSC)01 SDS at 42C and exposed to X-ray film at 70C

Western blots

Total cell extracts were prepared from 1107 cells byresuspending the cell pellet in 2ml of lysis buffer (10mMTrisndashHCl pH 75 1mM MgCl2 1mM EGTA 5mM b-mercaptoethanol 05 Cholamidopropyl- dimethylam-moniopropanesulfonate (CHAPS) (Sigma C-5849) and10 glycerol) and incubated on ice for 30min The cellswere centrifuged at 13 000 rpm for 30min at 4C andsupernatants taken Nuclear extracts were prepared aspreviously described (17) Fifty microgram of either totalcell protein extract or nuclear extract prepared fromindicated cell lines was resolved on either 10 or 15sodium dodecyl sulphatendashpolyacrylamide gel electrophor-esis gels and transferred to polyvinylidene fluoride(PVDF) membrane (Biorad) The blots were probedwith anti-SV40 T-antigen (Santa Cruz sc-148) anti- H-RAS (Santa Cruz sc-520) anti HDAC-1 (Santa Cruz sc-7872) and anti-tubulin (Cancer Research UK) antibodiesin 1Tris-buffered saline (TBS) buffer with 01 Tween20 followed by appropriate secondary anti-mouse IR800and anti-rabbit IR670 antibodies (LiCOR Biosciences)Images were collected on Odyssey scanner (LiCORBiosciences) and quantified with Image Studio software(LiCOR Biosciences)

Immunohistochemistry

A total of 2 105 cells were seeded onto 19-mm coverslipsin six-well dishes and 06mM H202 added for 2 h Cellswere then washed with phosphate-buffered saline (PBS)and allowed to recover overnight in growth media Afterfixation in 3 formaldehyde for 15min the cells wereincubated in blocking solution (PBS with 5 bovineserum albumin and 03 Triton X-100) for 1 h and thenincubated with anti-p21 antibody (Cell Signalling 2947)overnight at 4C The cells were rinsed three times withPBS and incubated for 2 h with Alexa 488 conjugated sec-ondary antibody Finally the cells were washed with PBScounter stained with diamidino-2-phenylidole (DAPI) andmounted in Prolong Gold (Life Sciences) Images weretaken at 20 magnification on Olympus BX61 fluores-cence microscope equipped with ColorViewII cameraand AnalySIS software

Methylated DNA affinity purification

Affinity purification of methylated DNA (MAP) wascarried out essentially as described (18) Briefly genomicDNA from MRC-5 cells MRC-5hTERT (50 and 100 pd)and MRC-5TSR (50 and 100 pd) was digested with MseIand 50 mg of digested DNA was loaded onto 1ml chroma-tography column (Tricorn GE Healthcare) containing50mg of His-tagged methyl-CpG binding domain ofMeCP2 protein bound to nickel-charged ChelatingSepharose Fast Flow (GE Healthcare) The column waswashed with 10 volumes of buffer A [20mM HEPES pH75 100mM NaCl 01 Tween 20 10 glycerol 05mMphenylmethylsulfonyl fluoride (PMSF)] followed bybuffer A with increasing concentration of NaCl

(01ndash07M) Methylated DNA was eluded with buffer Acontaining 1M NaCl Triplicate runs were done for eachof the cell lines and methylated fractions identified byPCR for known methylated regions The methylatedfractions were pooled concentrated and subjected towhole-genome amplification (WGA kit Sigma Aldrich)alongside MseI-digested input genomic DNA Amplifiedsamples were labelled with Cy3 and Cy5 respectively andco-hybridized to H18-RefSeq promoter microarrays(Roche NimbleGen) The data from promoter microarrayexperiments can be assessed at ArrayExpress accessionnumber E-MTAB-2005

Promoter microarray data analyses

The normalization of the microarray data and analyses ofpromoter regions were carried out with custom-designedsoftware lsquoPrometheusrsquo essentially as described previously(17) Briefly raw fluorescent intensity values were Loessnormalized using LIMMA package in R and the log2values for either MAPinput or ChIPinput werecalculated for each individual probe Subsequently forcomparison between microarray experiments the probevalues were scaled to have the same median absolutedeviation The log2 values of all probes located within a1000-bp window around the transcription start site (TSS)(+500 to 500) were aggregated into a single median log2value for each promoter

Chromatin immunoprecipitation

Chromatin immunoprecipitations (ChIPs) were carried outin triplicate as described (19) with antibodies againstacetylated H3 (Millipore 06ndash599) and H3K27me3(Millipore 07ndash449) ChIP and input DNA were amplifiedusing WGA kit (Sigma) labelled with Cy-dyes andhybridized to H18-RefSeq promoter microarrays (RocheNimbleGen) Microarray data were analysed as describedabove The data from ChIP experiments can be assessed atArrayExpress accession number E-MTAB-2004

Sodium bisulphite DNA sequencing

Sodium bisulphite treatment was carried out essentially asdescribed (20) and processed for sequencing as outlined in(21) PCR primers were designed manually or usingMethPrimer software (22) and are available on requestThe PCR products were cloned using the CloneJet PCRCloning Kit (Thermo Scientific) and sequenced usingBigDye Terminator v31 reagents (Applied Biosystems)BiQ Analyzer software (23) was used to analyse themethylation status of sequences

Methylated DNA immunoprecipitation

Methylated DNA immunoprecipitation (MeDIP) wasperformed as described (24) with minor modificationsGenomic DNA was fragmented by sonication to 300ndash1000 bp For each immunoprecipitation 4 mg of denaturedsonicated DNA was incubated for 2 h at 4C with 10 mlof anti-5-methylcytosine monoclonal antibody(Eurogenetec) A total of 40 ml of M280 Dynabeadsconjugated with sheep anti-mouse IgG (Dynal Biotech)were added and incubated for a further 2 h before

Nucleic Acids Research 2014 Vol 42 No 6 3531

at Edinburgh U

niversity on October 3 2014

httpnaroxfordjournalsorgD

ownloaded from

washing with IP buffer (10mM Na2HPO4pH 70 140mMNaCl 005 Triton X-100) Proteinase K digestion wascarried out at 50C overnight and the methylated DNAwas recovered using Pure-Link PCR Purification Kit(Life Sciences)

Quantitative PCR

Quantitative PCR was carried out with SYBR GreenMaster Mix (Roche) according to manufacturerrsquos instruc-tions on a LightCycler 480 (Roche) Quantitative PCR onMeDIP samples was carried out using 2 ml of MeDIPDNA and 20 ng of total input DNA Enrichments in theMeDIP fraction were calculated as a percentage of inputQuantitative reverse transcription PCRs were carriedin triplicate using independent complementary DNA(cDNA) synthesis reactions (Superscript II LifeSciences) as template Three independent RNAs prepar-ations from three different flasks of cells were used forcDNA synthesis Fold changes relative toGlyceraldehyde-3-phosphate dehydrogenease (GAPDH)were calculated using the Pfaffl method (25) Primer se-quences are available on request

Gene expression analyses

Total RNA was purified using Trizol reagent (LifeSciences) and double-stranded cDNA was synthesizedfrom 10 mg of total RNA using SuperScript double-stranded cDNA synthesis kit (Life Sciences) accordingto manufacturerrsquos instructions All samples were labelledwith Cy3 dye and hybridized in triplicate to Human GeneExpression H18 Build 475K expression microarrays(Roche NimbleGen) Raw intensity values were quantilenormalized using the BioConductor package LIMMAThe log2 values of the probes associated with each tran-script were summarized into a single log2 value usingmedian polish procedure A linear model was fit to thedata with LIMMA calculating the expression ratioM=log2 and moderated t-statistics adjusting P-valuesfor multiple testing The false discovery rate wasobtained using the BenjaminindashHochberg method andfalse discovery rate lt005 cut-off applied to all Mvalues The raw data from expression microarray experi-ments can be downloaded from ArrayExpress accessionnumber E-MTAB-2003 Functional gene annotation wasperformed by DAVID (httpdavidabccncifcrfgov) (26)

Analyses of IMR90 histone modifications data

Histone modification data for IMR90 cell line correspond-ing to the set of methylated promoters in MRC-5hTERT

cells at 100 pd were extracted from Ensembl API (version67) using HMoTF package of script written in Perl whichretrieved ChIP-seq data from annotated peaks taking intoaccount the genomic coordinates of the analysed regionsThe data were converted to Z-scores and plotted as heatmaps using the open source TIGR MultiExperimentViewer (MeV) software

RESULTS

Generation of immortalized and transformed cell lines

To follow the epigenetic changes that accompany theimmortalization and transformation of normal diploidhuman somatic cells we generated two isogenic cell lineswith defined characteristics (Figure 1A) First weintroduced by retroviral infection the catalytic componentof human telomerase (hTERT) into foetal lung fibroblastcell line MRC-5 which normally has a limited life span inculture and enters senescence after 20ndash25 pdThis step ensured that the immortalized cells stablymaintain their telomeres and do not become aneuploidafter prolonged culturing (Supplementary Figure S1Aand B) After selection for cells expressing hTERT wesequentially introduced by retroviral infection into asubset of hTERT-expressing cells two oncogenes theSimian Virus 40 large T-Ag and constitutively active onco-genic H-RASG12V (Figure 1A) We will refer to thesetwo cell lines as MRC-5hTERT and MRC-5TSRrespectively

Stable expression of hTERT in MRC-5hTERT cells wassufficient to bypass senescence and extend the proliferativelife span of the parental MRC-5 cell line beyond one yearin culture (Figure 1B) Therefore the MRC-5hTERT cellscan be designated as immortalized as has been previouslyreported (1327) To confirm that the extended life span inculture of MRC-5hTERT cells was due to persistent tel-omerase activity we performed a Telomere repeatamplification protocol (TRAP) assays on cell extractsfrom early (50 pd 3 months in culture) and late (100 pd6 months in culture) passage cells In parallel we alsoexamined the MRC-5TSR cells to ask whether telomeraseactivity is stably maintained in the presence of oncogenesWe detected a characteristic 6-bp laddering in MRC-5hTERT and MRC-5TSR cells at 50 and 100 pd but not inthe parental cell line and this activity was lost upon heatinactivation of the extracts (Figure 1C) The absence oftelomerase activity in MRC-5 cell line is consistent withthe limited life span of these cells in culture (Figure 1B)The level of telomerase activity detected in MRC-5hTERT

and MRC-5TSR cells was sufficient to extend the averagelength of telomeres from 6 togt 10 kb as indicated by theprogressive increase in telomere restriction fragmentlength (Supplementary Figure S1C)

To confirm that the introduced oncogenes are expressedcontinuously in MRC-5TSR cells we performed westernblots on extracts from the parental cell line as well asMRC-5hTERT and MRC-5TSR cells at 50 and 100 pdThese experiments detected robust expression of SV40T-Ag and elevated levels of H-RAS in the MRC-5TSR

cell line (Figure 1D and E) Because the H-RASantibody cannot discriminate between the endogenousand the mutant RAS protein we also cloned andsequenced cDNA from MRC-5TSR cells This confirmedthe presence of mutant H-RASG12V in MRC-5TSR cell lineand detected a 114 ratio of wild-type to mutant H-RASmessenger RNA (mRNA) (Supplementary Figure S1D)Taken together these experiments demonstrate thathTERT and the introduced oncogenes SV40 T-Ag andH-RASG12V are stably expressed in the transduced

3532 Nucleic Acids Research 2014 Vol 42 No 6

at Edinburgh U

niversity on October 3 2014

httpnaroxfordjournalsorgD

ownloaded from

MRC-5 cells and that this expression does not changesignificantly over long periods of time in culture

Characterization of immortalized and transformedcell lines

The SV40 large T-Ag is known to bind both p53 and pRBtumour suppressor proteins and impair their normalfunction in controlling cell cycle checkpoints uponcellular stress and cell cycle progression respectively(28) To determine whether MRC-5TSR cells displayreduced p53 and Rb activity and thus can be consideredtransformed we investigated their response to oxidativeDNA damage and acquisition of anchorage-independentgrowth Upon treatment with hydrogen peroxide (H2O2)we observed accumulation of p53-regulated Cyclin-dependent kinase (CDK) inhibitor protein p21 in thenuclei of MRC-5 and MRC-5hTERT cells which lackSV40 T-Ag and are expected to have normal p53-depend-ent response (Figure 2A and B left and middle panels) Incontrast MRC-5TSR cells expressing SV40 T-Ag did not

accumulate nuclear p21 in response to peroxide treatment(Figure 2A and B right panel) indicating that p53-de-pendent response to DNA damage is abrogated in thesecellsTo examine the acquisition of anchorage-independent

growth by the transformed MRC-5 cells which shouldbe largely dependent on constitutive expression of onco-genic H-RAS (1329) we scored the ability of MRC-5MRC-5hTERT MRC-5TSR cells and a control squamouslung carcinoma NCI H-520 cell line to form coloniesin soft agar Consistent with the stable expression ofH-RASG12V MRC-5TSR cells grown for either 50 or100 pd produced robust colonies in soft agar althoughwith lower frequency than the control H-520 cells(Figure 2C and D) Neither the MRC-5 cells nor theimmortalized MRC-5hTERT cell line formed colonies insoft agar (Figure 2C and D) From these experimentswe conclude that MRC-5hTERT cells although immortaldo not have the characteristic transformed properties ofMRC-5TSR cells

A pBabe-Neo-hTERT pBabe-Hygro-SV40 T pBabe-Puro-H-RAS

B

Time (days)

MRC-5

MRC-5hTERT

MRC-5TSR

G

row

th

(pop

ulat

ion

doub

lings

)

MRC-5 MRC-5hTERT (immortalised) MRC-5TSR (transformed)

MRC-5

+_ +_ +_ +_ +_ _+

hTERT TSR Ctrl50 100 50 100pd

C

E

H-Ras

α-Tubulin

50 100 50 100

hTERT TSRMRC-5

pd

D

0

50

100

150

200

250

50 100 50 100

hTERT TSRMRC-5

pd6

70 -

100 -T-Ag

HDAC1

6

50 100 200150 250 300 350 400

26 -

55 -

6

V12G

HI

Figure 1 Generation of immortal and transformed human cell lines (A) Immortalized (MRC-5hTERT) and transformed (MRC-5TSR) human celllines were generated from embryonic lung fibroblasts MRC-5 by stepwise infection with retroviral particles driving the expression of hTERT SV40T-Ag and H-RASV12G lsquoNeorsquo lsquoHygrorsquo and lsquoPurorsquo indicate drug resistance markers neomycin hygromycin and puromycin respectively carried by theretroviral vectors (B) Growth of MRC-5 MRC-5hTERT and MRC-5TSR cell lines measured as population doublings over 400 days in culture Theparental MRC-5 cell line enters senescence after 20 pd (C) Telomerase activity detected by TRAP in cell extracts from MRC-5 MRC-5hTERT andMRC-5TSR cells lsquoCtrlrsquo is a control telomerase-positive NCI-H520 lung cancer cell line The lsquoplusrsquo and lsquominusrsquo symbols indicate whether the extractshave been subjected to heat inactivation (HI) lsquopdrsquo represents population doublings (D) A Western blot probed with anti-SV40 T-Ag antibodiesshows stable expressed of T-Ag only in MRC-5TSR cell HDAC1 is a loading control (E) MRC-5TSR cells show elevated levels of H-RAS protein dueto expression of exogenous H-RASV12G (see also Supplementary Figure S1D) a-Tubulin is a loading control

Nucleic Acids Research 2014 Vol 42 No 6 3533

at Edinburgh U

niversity on October 3 2014

httpnaroxfordjournalsorgD

ownloaded from

Progressive accumulation of DNA methylationat gene promoters in immortalized and transformedcell lines

We next asked whether DNA methylation patterns remainstable in MRC-5hTERT and MRC-5TSR cells over time andwhether the transformation by oncogenes is required toinduce changes in DNA methylation at gene promotersthat are characteristic of many human tumours Todo so we used methyl-CpG binding domain affinity puri-fication (MAP) of methylated DNA combined withhybridization to microarrays containing probes for24 659 human protein-coding RefSeq gene promoters(Figure 3A) To distinguish significant changes in DNAmethylation close to TSS from more distal DNA methyla-tion patterns we analysed the microarray data separatelyfor 1-kb promoter regions (plusmn500 bp from TSS) and theupstream regions (500 to 1500 bp from TSS) asdescribed previously (17) A minimal cut-off for medianlog2 MAPinput difference between cell lines of 1 (corres-ponding to 2-fold change in DNAmethylation) was used inall analyses We examined DNA methylation at gene pro-moters in MRC-5hTERT and MRC-5TSR cells at 50 and100 pd in culture and compared these values with thosefor the parental MRC-5 cell line (Figure 3B and

Supplementary Table S1) These analyses detected a pro-gressive gain of DNA methylation at gene promoters inimmortalized and transformed cells at 50 and 100 pdcompared with the parental cell line However wedetected no significant differences between MRC-5hTERT

and MRC-5TSR at either 50 or 100 pd when DNA methy-lation patterns in these cell lines were compared with eachother (Figure 3C) Most de novo DNA methylation events(250 promoters) occurred late between 50 and 100 pdrather than early (32ndash70 promoters) (Figure 3B and D)and affected promoters with low intermediate and highCpG density (Figure 4A) Importantly promotersmethylated early (by 50 pd) remained methylated at latepassages (100 pd) suggesting that once DNA methylationwas established at gene promoters it was stably maintainedthrough subsequent cell divisions (Figure 3D) De novoDNA methylation events did not affect preferentiallygenes located close to telomeres but occurred at locidistributed throughout the genome (SupplementaryFigure S2) This suggests that gain of DNA methylationat gene promoters in MRC-5hTERT and MRC-5TSR cellsdid not result from spreading of subtelomeric heterochro-matin from the extended telomeres

B MRC-5 hTERT TSR

DAPI

p21

A C

Cell Line Colonies in soft agar

NCI-H520 2910 +- 645

MRC-5 0

hTERT pd 50 0

TSR pd 50 240 +- 40

TSR pd 100 350 +- 70

D

MRC-5 hTERT

TSR NCI-H520

MRC-5 hTERT TSR

DAPI

T-Ag

01 H2O2

Figure 2 Transformation-induced properties of MRC-5TSR cells (A) Immunostaining of MRC-5 MRC-5hTERT and MRC-5TSR cells with antibodiesagainst SV40 T-Ag (red) The cells were counterstained with DAPI (blue) (B) Immunostaining of H2O2-treated MRC-5 MRC-5hTERT and MRC-5TSR cells with anti-p21 antibodies (green) and DAPI (blue) MRC-5TSR cells do not show high levels of p21 induction upon oxidative stress Thescale bars in (A) and (B) represent 200mm (C) Micrographs of cells grown in soft agar for 4 weeks Only the MRC-5TSR cells and the control cellline NCI-H520 form colonies in soft agar (D) A table showing the number of colonies formed by MRC-5 MRC-5hTERT MRC-5TSR and NCI-H520cells in soft agar The MRC-5TSR cell line was scored at 50 and 100 pd

3534 Nucleic Acids Research 2014 Vol 42 No 6

at Edinburgh U

niversity on October 3 2014

httpnaroxfordjournalsorgD

ownloaded from

We further confirmed the presence and timing of DNAmethylation at gene promoters by MeDIP (24) followedby quantitative PCR at specific early (SHOX2 andC1QTNF) and late (BOLA-1) methylated promoters(Figure 4B) as well as by bisulphite DNA sequencing ofthree selected promoters (SHOX2 RPL37 and BOLA-1)in the parental cell line as well as in the MRC-5hTERT andMRC-5TSR cells at 50 and 100 pd (Figure 4C) In all casesthe MeDIP assays and bisulphite DNA sequencing were inagreement with the microarray data

Taken together these analyses indicate that identicaltime-dependent changes in DNA methylation at gene

promoters occur in two independent cell populationsand that gain of DNA methylation at promoters doesnot require the presence of oncogenes such as SV40 T-Ag and oncogenic H-RAS

De novo DNA methylation occurs predominantly atinactive gene promoters

It has been reported that promoters that carry chromatinmarked by Polycomb Repressive Complex 2 (PRC2)-de-pendent repressive histone H3 lysine 27 trimethylation(H3K27me3) are more susceptible to de novo DNA

B

A

C D

log2 MAPinput MRC-5 log2 MAPinput MRC-5

MRC-5 vs hTERT 50 pd MRC-5 vs hTERT 100 pd hTERT 50 pd vs TSR 50 pd

log2 MAPinput MRC-5

Fragmented DNA

MRC-5

hTERT50 pd

100 pd

TSR50 pd

100 pd

Cell lines

Me DNA

Input

Me DNA

Input

MAP

Me DNA

Input

MAP

MAP

Microarrays

Labelling

Promoter analysis

24659 human promoters

Labelling

Labelling

-500 bp +500 bp

log2 MAPinput MRC-5 log2 MAPinput hTERT

log2 MAPinput hTERT

-4-3-2-101

2

0-1-2-3 1 2-4 0-1-2-3 1 2-4

0-1-2-3 1 2-4 0-1-2-3 1 2-4 0-1-2-3 1 2-4

-4-3-2-101

2

-4-3-2-101

2

-4-3-2

-101

2

-4-3-2-101

2

-4-3-2-101

2

MRC-5 vs TSR 50 pd MRC-5 vs TSR 100 pd hTERT 100 pd vs TSR 100 pd

log 2

MA

Pin

put h

TE

RT

log 2

MA

Pin

put T

SR

log 2

MA

Pin

put h

TE

RT

log 2

MA

Pin

put T

SR

log 2

MA

Pin

put T

SR

log 2

MA

Pin

put T

SRn=70 n=287

n=1

r=0868 r=0658 r=086

r=0666 r=0887

n=2 n=87 n=80

n=76 n=2

n=32 n=301

r=0848 n=24

4-4 0log2 MAPinput

n=14

n=301

0-1-2-3 1 2-4

MRC-5

hTERT 5

0 pd

hTERT 1

00 p

d

TSR 50

pd

TSR 100

pd

Figure 3 Accumulation of DNA methylation at gene promoters in the immortalized and transformed cell lines (A) Detection of methylated genepromoters in MRC-5 cells early (50 pd) and late (100 pd) passage MRC-5hTERT and MRC-5TSR cell lines by Methylated DNA Affinity Purification(MAP) coupled with hybridization to promoter microarrays representing 24 659 human RefSeq genes Regions spanning probes from 500 bp to+500 bp relative to TSS were interrogated (B) Log2 plots show differentially methylated gene promoters in early and late passage MRC-5hTERT andMRC-5TSR cells relative to the parental cell line Promoters displaying 2-fold gain of DNA methylation are marked in red Promoters with 2-foldloss of DNA methylation are marked in blue (C) Log2 plots comparing DNA methylation patterns at gene promoters between MRC-5hTERT andMRC-5TSR cell lines at early (50 pd) and late (100 pd) passages (D) A heat map visualization of de novo methylated gene promoters (n=301) inMRC-5hTERT and MRC-5TSR cell lines at early (50 pd) and late (100 pd) passages in comparison with the parental cell line MRC-5 Promotersmethylated at early passage in MRC-5hTERT and MRC-5TSR cells remain methylated in late passage cells

Nucleic Acids Research 2014 Vol 42 No 6 3535

at Edinburgh U

niversity on October 3 2014

httpnaroxfordjournalsorgD

ownloaded from

B

0

5

10

15

20

25

SHOX2 C1QTNF BOLA-1 XIST no CpG ANKRD42

MRC-5 hTERT50 pd

hTERT100 pd

IN

PU

T

Earlyn=70

Laten=301

0

20

40

60

80

100

T

otal

HCP

LCP

ICP

A

C BOLA-1SHOX2 RPL37

MRC-5

hTERT100 pd

TSR100 pd

CpG

Mse I

1380 bp 1690 bp 1930 bp

85

96

120 23

26

21

+366 +591 +103 +585 -555 +878

hTERT 50 pd

TSR 50 pd

Figure 4 Validation of DNA methylation data obtained from promoter microarray analyses (A) A bar graph representation of low (LCP) inter-mediate (ICP) and high CpG density (HCP) promoters among loci that are methylated either early or late in MRC-5hTERT cell line (B) DNAmethylation levels at early (SHOX2 and C1QTNF) and late (BOLA-1) methylating gene promoters in MRC-5 and MRC-5hTERT cell lines detected byMeDIP Constitutively methylated promoter of the non-coding RNA XIST serves as a positive control ANKRD42 is a promoter that lacksmethylation in all cell lines at any time point lsquono CpGrsquo is a region on chromosome X that lacks CpGs (C) Validation of de novo DNA methylationat SHOX2 RPL37 and BOLA-1 gene promoters in MRC-5hTERT and MRC-5TSR cells at 50 and 100 pd by bisulphite DNA sequencing MethylatedCpGs are shown as black circles unmethylated CpGs as white circles The graphs at the top of the panel show CpG dinucleotides 1 kb promoterregion analysed by microarray data processing the span of the MseI restriction fragment and the region analysed by bisulphite DNA sequencing

3536 Nucleic Acids Research 2014 Vol 42 No 6

at Edinburgh U

niversity on October 3 2014

httpnaroxfordjournalsorgD

ownloaded from

methylation during differentiation of embryonic stem cellsto neurons than active promoters marked by H3K4me3(30) A further correlation between H3K27me3 at genepromoters in non-transformed cells and gain of DNAmethylation at such sites in lung colorectal and breastcancer cell lines has been observed in several independentstudies (14ndash16) This led to the suggestion that in tumoursthere is a frequent switch from the potentially reversiblePolycomb-mediated gene silencing to a more stable long-term repression by DNA methylation (15)

To examine whether de novo DNA methylation inMRC-5hTERT and MRC-5TSR cells occurs preferentiallyat promoters that are either pre-marked by H3K27me3or at those associated with actively transcribed genes weanalysed H3K27me3 and H3 acetylation at gene pro-moters in the parental MRC-5 cells by ChIP combinedwith hybridization to promoter microarrays as describedearlier in the text These experiments showed that 28 ofall promoters that acquire DNA methylation in MRC-5hTERT and MRC-5TSR cells carry H3K27me3 in MRC-5cells and only 14 are enriched in H3 acetylated chroma-tin (Figure 5) Both active promoters enriched inacetylated H3 and Polycomb-silenced loci enriched inH3K27me3 (with few exceptions) displayed a tendencyto be methylated late by 100 but not by 50 pd suggestingthat both modifications delay the appearance of DNAmethylation However 58 of promoters that becomede novo methylated in MRC-5hTERT and MRC-5TSR cellshad neither H3K27me3 nor acetylated H3 in the parentalcell line

Multiple histone modifications have been mapped byhigh-throughput approaches in IMR90 fibroblast cellline which similar to MRC-5 is derived from humanfoetal lung Comparison between the two cell linesrevealed broadly similar patterns of H3K27me3 andhistone acetylation at gene promoters as well as presenceof H3K4me3 at loci carrying acetylated H3 (Figure 5)Given the similarity of chromatin modification betweenIMR90 and MRC-5 cells we sought to determinewhether other histone modifications present at loci thatlack either H3K27me3 or H3 acetylation in MRC-5 cellscould potentiate gain of DNA methylation at gene pro-moters in MRC-5hTERT and MRC-5TSR cells Of all the22 histone modifications examined in the IMR90 cellsonly H3K36me3 normally present within transcribedregions of the genome (31ndash33) was apparent at 19 ofgene promoters that acquire DNA methylation in MRC-5hTERT andMRC-5TSR cells (Figure 5) Interestingly manyof the H3K36me3-marked promoters represent alternativedownstream TSSs which drive the expression of truncatedvariant transcripts (Supplementary Figure S3)Importantly about half of the loci that were methylatedearly (by 50 pd) in the MRC-5hTERT and MRC-5TSR cellsshowed enrichment for H3K36me3 in the primary parentalcell line (Figure 5) Taken together these analyses indicatethat promoters of silenced genes that are either devoid ofknown modifications or enriched for H3K36me3 a modi-fication refractive to initiation of transcription (31) areprone to DNA methylation early in immortalized humancells whereas promoters of either actively transcribed orPolycomb-silenced genes tend to be methylated late by

100 pd However none of the examined chromatin modifi-cations can be considered predictive of whether or not agene promoter will become de novo methylated inimmortalized cells Many loci carrying similar histonemarks did not accumulate DNA methylation in MRC-5hTERT and MRC-5TSR cells at late passage in culture

Immortalized and transformed cells progressivelyaccumulate changes in gene expression

Given that in MRC-5hTERT cells we observed gain ofDNA methylation primarily at promoters of genes that

MRC-5Histone PTMs

DNA methylation

IMR90Histone PTMs

-30 30

19

z-score

n=301

MRC5

H3K9

K14ac

hTERT 5

0 pd

hTERT 1

00 p

d

TSR 100

pd

H3K27

me3

H3K36

me3

H3K27

me3

H3K4m

e3

28

14

39

Figure 5 Histone modifications at gene promoters that undergo denovo DNA methylation in the immortalized cells A heat map repre-sentation of post-translational histone modifications (PTMs) found inthe parental MRC-5 cell line and a related foetal lung fibroblast cellline IMR90 at gene promoters that become methylated in MRC-5hTERT

and MRC-5TSR cell lines Antibodies against histone H3 acetylated atK9 and K14 or trimethylated at K27 were used for ChIP coupled withhybridization to promoter microarrays Publicly available data forhistone PTMs for IMR90 cells was used in these analyses OnlyH3K36me3 H3K27me3 and H3K4me3 data for IMR90 cells areshown Except these three modifications and H3H4 acetylation (notshown) no other modifications were found significantly enriched at thisset of promoters

Nucleic Acids Research 2014 Vol 42 No 6 3537

at Edinburgh U

niversity on October 3 2014

httpnaroxfordjournalsorgD

ownloaded from

were inactive in the parental cell line we asked whethergene expression patterns in hTERT-immortalized cellsremain stable after 50 and 100 pd in culture We alsosought to determine whether expression of SV40 T-Agand constitutively active H-RASG12V in MRC-5TSR cellshad significant role in reprogramming gene expressionprofiles as reported for short-term studies of human cellstransformed by viral oncogenes (3435) To address thesequestions we used microarrays to examine gene expres-sion patterns in MRC-5hTERT and MRC-5TSR cells at 50and 100 pd and compared these to each other and to theparental MRC-5 cell line Surprisingly we found that cellsimmortalized by hTERT progressively accumulate signifi-cant changes in gene expression which were also shared

by the MRC-5TSR cells (Figure 6A) Thus we detected1193 transcripts that were upregulated and 571 transcriptsthat were downregulated by 3-fold or more inimmortalized and transformed cells by 100 pd comparedwith the parental cell line (Supplementary Table S2)Upregulated transcripts could be divided into twodistinct groups genes that were weakly expressed in theMRC-5 cells but upregulated in MRC-5hTERT and MRC-5TSR cell lines (Group 1 upregulated) and genes that wereexpressed in MRC-5hTERT and MRC-5TSR cells but not inthe parental cell line (Group 2 activated) (Figure 6A)Gene ontology and gene set enrichment analyses showedthat transcripts from Group 1 included proteins involvedin cytoskeletal organization and cell migration whereas

A B

-25 25Z score

-25 25Z score

Group 1 n=354 (20)

cytoskeleton organisation (plt10-4)

cell differentiation (plt10-2)

Group 2 n=839 (47)

protein transport (plt10-5)

protein kinase activity (plt10-3)

RNA processing (plt10-2)

- RNA splicing (plt10-2)

- mRNA export (plt10-2)

cancer associated signalling pathways (p=005)

- colorectal cancer (p=005)

- melanoma (plt10-2)

- lung cancer (p=004)

Group 3 n=571 (33)

regul of transcription factor activity (plt10-3)

cell differentiation (plt10-2)

response to extracellular signalling (plt10-2)

regulation of angiogenesis (plt10-2)

Group 4 n=57 (27)

cell cycle (plt10-3)

Group 5 n=45 (21)

regulation of transport (plt10-2)

cell-cell signalling (plt10-2)

Group 6 n=108 (51)

regulation of cell growth (plt10-4)

tissue morphogenesis (plt10-2)

nucleosome assembly (plt10-2)

MRC-5

hTERT 5

0 pd

hTERT 1

00 p

d

TSR 100

pd

MRC-5

hTERT 5

0 pd

hTERT 1

00 p

d

TSR 100

pd

TSR 50

pd

Figure 6 Changes in gene expression in immortalized and transformed cell lines (A) A heat map showing immortality-associated changes in geneexpression in MRC-5hTERT and MRC-5TSR cell lines at 50 and 100 pd Three groups of genes can be clearly distinguished The most significantfunctions of representative up- and downregulated groups of genes identified by gene ontology and gene set enrichment analyses are indicated (B) Aheat map representation of transformation-associated changes in gene expression in MRC-5TSR cell line in comparison with the primary MRC-5 andimmortalized MRC-5hTERT cells The most significantly enriched biological functions attributed to the three groups of genes are indicated

3538 Nucleic Acids Research 2014 Vol 42 No 6

at Edinburgh U

niversity on October 3 2014

httpnaroxfordjournalsorgD

ownloaded from

many of the transcripts from Group 2 encode for proteinsimplicated in cancer-associated signalling pathways regu-lation of MAP kinase cascade protein transport andRNA splicing (Figure 6A) The transcripts downregulatedin MRC-5hTERT and MRC-5TSR cells (Group 3) wereenriched in regulators of cell differentiation modulationof transcription factor activity and proteins involved inresponse to extracellular signalling Interestingly anumber of genes that change their expression inimmortalized cells eg PI3K MDM2 SMAD23 andSTAT1 (Supplementary Figure S4) are implicated in theevasion of apoptosis and acquisition of insensitivity togrowth-inhibiting signals which are characteristicfeatures of tumour cells We validated these expressionchanges by independently performed quantitative re-verse transcription PCRs on several selected transcripts(Supplementary Figure S6A and B)

We detected a much smaller number of up- anddownregulated transcripts (210) that could be attributedto the constitutive expression of oncogenes as the levelsof these mRNAs were different between MRC-5hTERT

and MRC-5TSR cells (Figure 6B and SupplementaryTable S2) Here we also identified three distinct groups oftranscripts (labelled Groups 4 5 and 6) The mRNAs fromGroup 4 were downregulated in late passage MRC-5hTERT

cells but these were highly expressed in MRC-5 as well asMRC-5TSR cells and included genes involved in cell cycleregulation such as Securin CDC25 phosphatase and thekinase Aurora B The transcripts from Group 5 wereenriched for regulators of transport and cellndashcell signallingand were expressed neither in MRC-5 nor in MRC-5hTERT

cells but were progressively upregulated exclusively in thetransformed MRC-5TSR cell line (Figure 6B) FinallyGroup 6 included transcripts that were upregulated in theimmortalized MRC-5hTERT cells but expressed in theMRC-5TSR cell at levels comparable with the parental cellline This group was enriched in regulators of cell growthtissue morphogenesis and nucleosome assembly Asexpected many of the proteins with altered levels of expres-sion in MRC-5TSR cells belong to cancer-associatedsignalling pathways and have roles in promoting cellu-lar proliferation angiogenesis and cell survival(Supplementary Figure S5) Although some of these com-ponents are upregulated already inMRC-5hTERT cells theirlevels of expression are further enhanced upon introductionof oncogenes Independently performed quantitativereverse transcription PCRs on a subset of transcripts werein agreement with the microarray data (SupplementalFigure S6C and D)

Taken together these analyses demonstrate that sus-tained expression of hTERT leads to significant andcomplex large-scale reprogramming of the transcriptionaloutput of the genome which is likely to reflect adaptationto highly proliferative state On the other hand expressionof SV40 T-Ag and oncogenic H-RASV12G in hTERT-immortalized cells induces fewer sustainable changes ingene expression but these might be essential fortumorigenisity and acquisition anchorage-independentgrowth

DISCUSSION

Aberrant DNA methylation at gene promoters has beenreported for many tumours and typically is accompaniedby lack of transcription from the associated geneAlthough there are many specific examples of silencingof tumour suppressor genes by promoter DNA methyla-tion recent high-throughput analyses in breast colorectaland other types of cancer have suggested that the vastmajority of gene promoters methylated in tumours repre-sent developmentally regulated loci which are alreadyrepressed in pre-cancerous tissues (3637) These observa-tions highlight the coexistence of lsquodriverrsquo and lsquopassengerrsquode novomethylation events that occur in tumours implyingthat most changes in DNA methylation at gene promotersare unlikely to contribute to cancer formation (153839)Nevertheless several important questions arise fromthese studies How are the aberrant patterns of DNAmethylation brought into existence What are thedynamics of de novo DNA methylation and the moleculardeterminants of this process Are epigenetic alterationslinked intrinsically to genetic determinants of tumourformationTo address some of these questions we used a model

system which allows defined genetic components to besequentially introduced into primary human cells withnormally finite life in culture The contribution of thesegenetic components to changes in growth characteristicsof modified cells gene expression patterns and promoterDNA methylation could then be examined by high-throughput assays Thus the expression of the catalyticsubunit of telomerase enzyme (hTERT) in MRC-5 foetallung fibroblasts generated an immortal cell line withlife span extended for gt200 cell generations whereasfurther expression of collaborating oncogenes SV40T-Ag and H-RASV12G in hTERT-immortalized cellsproduced an isogenic transformed cell line characterizedby acquisition of anchorage-independent growth Ourdetailed investigation of promoter DNA methylation inthese two isogenic cell lines identified loci that are proneto time-dependent de novoDNA methylation and led us toconclude that the changes in DNA methylation at pro-moters do not require expression of oncogenes Near iden-tical changes in DNA methylation at gene promoters tookplace in the immortalized (MRC-5hTERT) and transformed(MRC-5TSR) cell lines with stable diploid karyotype Thisis somewhat surprising given that constitutively activeK-RAS and H-RAS have been implicated in DNA methy-lation-mediated silencing of specific genes (4041) Incontrast to these findings our data firmly suggest thatcellular immortality conferred by hTERT expression issufficient to promote de novo DNA methylation at genepromoters Whether the immortal and transformed cellsdisplay differences in DNA methylation elsewhere in thegenome is yet to be determinedIn agreement with recent studies (42) the vast majority

of de novo DNA methylation events in MRC-5hTERT andMRC-5TSR cell lines occurred at promoters of genes thatwere already silenced in the parental cell line Some ofthese represent loci carrying repressive H3K27me3 andH3K36me3 histone modifications However it seems

Nucleic Acids Research 2014 Vol 42 No 6 3539

at Edinburgh U

niversity on October 3 2014

httpnaroxfordjournalsorgD

ownloaded from

unlikely that histone modifications determine whether ornot a promoter will become methylated in immortalizedcells About 40 of the loci hypermethylated in MRC-5hTERT and MRC-5TSR cells were devoid of H3K27me3and H3K36me3 in the parental cell line and had no otherdetectable known modifications in the closely relatedIMR90 fibroblasts Taken together these data suggestthat lack of promoter activity and potentially stablybound transcription factors which could protect suchloci against DNA methylation machinery (4344) mayresult in gradual acquisition of DNA methylation overtime Our data also indicate that promoters of activelytranscribed genes marked by H3 acetylation andH3K4me3 tend to be more stably protected Few activepromoters became methylated in the immortal cells andin all cases this occurred at late passage In contrast tosilenced genes methylation of active promoters could rep-resent rare driver methylation events which promote cellproliferation and survival It is plausible that stochasticDNA methylation events take place in immortalizedcells and these patterns are under constant surveillanceand selection Therefore only those methylation eventsthat occur either at weakly protected silenced promotersor genes inactivation of which favours long-term survivalwill be tolerated and stably propagated in the immortalcell populations As immortality and in many cases theexpression of hTERT (45) is a hallmark of all tumoursthis may explain why aberrant DNA methylation is such aprevalent feature in a variety of cancer cell typesAnother essential feature of hTERT-immortalized cells

is time-dependent acquisition of large-scale changes in geneexpression (4246) Given the stable diploid karyotypeof MRC-5hTERT cells these expression patterns must beepigenetic by nature as they cannot be explained by aneu-ploidy or alterations in DNA sequence In contrast to denovoDNA methylation events the changes in gene expres-sion observed inMRC-5hTERT cells are likely to result fromselective pressure to enhance traits that favour long-termsurvival and stable proliferation in culture As the evasionof apoptosis effective repair of DNA damage and robustprogression through the cell cycle are essential propertiesof tumour cells it is probably not surprising that proteinswith known function in cancer-associated signallingpathways show altered expression in the immortal cellsAlthough subsequent introduction of SV40 T-Ag and con-stitutively active H-RAS into hTERT-immortalized cellsresults in fewer high-amplitude changes in gene expressionour data indicate that the presence of cooperating onco-genes promotes subtle alterations in many signallingpathways confers insensitivity to growth signals andacquisition of anchorage-independent growth Takentogether these observations imply that telomerase-induced immortality is sufficient for large-scale repro-gramming of DNA methylation at gene promoters andexpression patterns in diploid human cells to a state thatresembles pre-cancerous lesions Such reprogrammingreflects the intrinsic plasticity of immortal cell genomewhich in combination with oncogene-dependent modula-tion of responses to stress and growth signals may favouradaptation to a variety of cellular and tissue microenviron-ments and ultimately support tumour growth

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online

ACKNOWLEDGEMENTS

We thank Dr Scott Lowe (Memorial Sloan-KetteringCancer Center New York USA) and Dr Robert AWeinberg (Whitehead Institute for Biomedical ResearchCambridge MA USA) for providing plasmids and themembers of Stancheva lab for helpful comments duringthe preparation of this manuscript

FUNDING

This research was supported by Cancer Research UKSenior Fellowship [C7215A8983] and EMBO Long-termfellowship (to TC) The Wellcome Trust Centre for CellBiology is supported by core funding from the WellcomeTrust [092076] Funding for open access charge TheWellcome Trust via University of Edinburgh

Conflict of interest statement None declared

REFERENCES

1 JonesPA and BaylinSB (2002) The fundamental role ofepigenetic events in cancer Nat Rev Genet 3 415ndash428

2 JonesPA and BaylinSB (2007) The epigenomics of cancerCell 128 683ndash692

3 BaylinSB and JonesPA (2011) A decade of exploring thecancer epigenome-biological and translational implicationsNat Rev Cancer 11 726ndash734

4 EstellerM (2008) Epigenetics in cancer N Engl J Med 3581148ndash1159

5 EhrlichM (2009) DNA hypomethylation in cancer cellsEpigenomics 1 239ndash259

6 HonGC HawkinsRD CaballeroOL LoC ListerRPelizzolaM ValsesiaA YeZ KuanS EdsallLE et al (2012)Global DNA hypomethylation coupled to repressive chromatindomain formation and gene silencing in breast cancer GenomeRes 22 246ndash258

7 VisvaderJE and LindemanGJ (2008) Cancer stem cells in solidtumours accumulating evidence and unresolved questions NatRev Cancer 8 755ndash768

8 SelamatSA ChungBS GirardL ZhangW ZhangYCampanM SiegmundKD KossMN HagenJA LamWLet al (2012) Genome-scale analysis of DNA methylation in lungadenocarcinoma and integration with mRNA expression GenomeRes 22 1197ndash1211

9 FacklerMJ UmbrichtCB WilliamsD ArganiP CruzLAMerinoVF TeoWW ZhangZ HuangP VisvananthanKet al (2011) Genome-wide methylation analysis identifies genesspecific to breast cancer hormone receptor status and risk ofrecurrence Cancer Res 71 6195ndash6207

10 HinoueT WeisenbergerDJ LangeCP ShenH ByunHMVan Den BergD MalikS PanF NoushmehrH vanDijkCM et al (2012) Genome-scale analysis of aberrant DNAmethylation in colorectal cancer Genome Res 22 271ndash282

11 KobayashiY AbsherDM GulzarZG YoungSRMcKenneyJK PeehlDM BrooksJD MyersRM andSherlockG (2011) DNA methylation profiling reveals novelbiomarkers and important roles for DNA methyltransferases inprostate cancer Genome Res 21 1017ndash1027

12 BrenaRM and CostelloJF (2007) Genome-epigenomeinteractions in cancer Hum Mol Genet 16 R96ndashR105

13 HahnWC CounterCM LundbergAS BeijersbergenRLBrooksMW and WeinbergRA (1999) Creation of humantumour cells with defined genetic elements Nature 400 464ndash468

3540 Nucleic Acids Research 2014 Vol 42 No 6

at Edinburgh U

niversity on October 3 2014

httpnaroxfordjournalsorgD

ownloaded from

14 SchlesingerY StraussmanR KeshetI FarkashS HechtMZimmermanJ EdenE YakhiniZ Ben-ShushanEReubinoffBE et al (2007) Polycomb-mediated methylation onLys27 of histone H3 pre-marks genes for de novo methylation incancer Nat Genet 39 232ndash236

15 Gal-YamEN EggerG IniguezL HolsterH EinarssonSZhangX LinJC LiangG JonesPA and TanayA (2008)Frequent switching of polycomb repressive marks and DNAhypermethylation in the PC3 prostate cancer cell line Proc NatlAcad Sci USA 105 12979ndash12984

16 WidschwendterM FieglH EgleD Mueller-HolznerESpizzoG MarthC WeisenbergerDJ CampanM YoungJJacobsI et al (2007) Epigenetic stem cell signature in cancerNat Genet 39 157ndash158

17 MyantK TermanisA SundaramAY BoeT LiC MerusiCBurrageJ de Las HerasJI and StanchevaI (2011) LSH andG9aGLP complex are required for developmentally programmedDNA methylation Genome Res 21 83ndash94

18 IllingworthR KerrA DesousaD JorgensenH EllisPStalkerJ JacksonD CleeC PlumbR RogersJ et al (2008)A novel CpG island set identifies tissue-specific methylation atdevelopmental gene loci PLoS Biol 6 e22

19 ClouaireT WebbS SkeneP IllingworthR KerrAAndrewsR LeeJH SkalnikD and BirdA (2012) Cfp1integrates both CpG content and gene activity for accurateH3K4me3 deposition in embryonic stem cells Genes Dev 261714ndash1728

20 FeilR CharltonJ BirdAP WalterJ and ReikW (1994)Methylation analysis on individual chromosomes improvedprotocol for bisulphite genomic sequencing Nucleic Acids Res22 695ndash696

21 SuzukiMM KerrAR De SousaD and BirdA (2007) CpGmethylation is targeted to transcription units in an invertebrategenome Genome Res 17 625ndash631

22 LiLC (2007) Designing PCR primer for DNA methylationmapping Methods Mol Biol 402 371ndash384

23 BockC ReitherS MikeskaT PaulsenM WalterJ andLengauerT (2005) BiQ Analyzer visualization and qualitycontrol for DNA methylation data from bisulfite sequencingBioinformatics 21 4067ndash4068

24 MohnF WeberM SchubelerD and RoloffTC (2009)Methylated DNA immunoprecipitation (MeDIP) Methods MolBiol 507 55ndash64

25 PfafflMW (2001) A new mathematical model for relativequantification in real-time RT-PCR Nucleic Acids Res 29 e45

26 Huang daW ShermanBT and LempickiRA (2009) Systematicand integrative analysis of large gene lists using DAVIDbioinformatics resources Nat Protoc 4 44ndash57

27 TaylorLM JamesA SchullerCE BrceJ LockRB andMackenzieKL (2004) Inactivation of p16INK4a with retentionof pRB and p53p21cip1 function in human MRC5 fibroblaststhat overcome a telomere-independent crisis duringimmortalization J Biol Chem 279 43634ndash43645

28 AhujaD Saenz-RoblesMT and PipasJM (2005) SV40 large Tantigen targets multiple cellular pathways to elicit cellulartransformation Oncogene 24 7729ndash7745

29 ThullbergM GadA Le GuyaderS and StrombladS (2007)Oncogenic H-Ras V12 promotes anchorage-independentcytokinesis in human fibroblasts Proc Natl Acad Sci USA 10420338ndash20343

30 MohnF WeberM RebhanM RoloffTC RichterJStadlerMB BibelM and SchubelerD (2008) Lineage-specificpolycomb targets and de novo DNA methylation definerestriction and potential of neuronal progenitors Mol Cell 30755ndash766

31 CarrozzaMJ LiB FlorensL SuganumaT SwansonSKLeeKK ShiaWJ AndersonS YatesJ WashburnMP et al(2005) Histone H3 methylation by Set2 directs deacetylation ofcoding regions by Rpd3S to suppress spurious intragenictranscription Cell 123 581ndash592

32 MikkelsenTS KuM JaffeDB IssacB LiebermanEGiannoukosG AlvarezP BrockmanW KimTK KocheRPet al (2007) Genome-wide maps of chromatin state in pluripotentand lineage-committed cells Nature 448 553ndash560

33 HawkinsRD HonGC LeeLK NgoQ ListerRPelizzolaM EdsallLE KuanS LuuY KlugmanS et al(2010) Distinct epigenomic landscapes of pluripotent andlineage-committed human cells Cell Stem Cell 6 479ndash491

34 FerrariR PellegriniM HorwitzGA XieW BerkAJ andKurdistaniSK (2008) Epigenetic reprogramming by adenoviruse1a Science 321 1086ndash1088

35 HorwitzGA ZhangK McBrianMA GrunsteinMKurdistaniSK and BerkAJ (2008) Adenovirus small e1a altersglobal patterns of histone modification Science 321 1084ndash1085

36 SproulD NestorC CulleyJ DicksonJH DixonJMHarrisonDJ MeehanRR SimsAH and RamsahoyeBH(2011) Transcriptionally repressed genes become aberrantlymethylated and distinguish tumors of different lineages in breastcancer Proc Natl Acad Sci USA 108 4364ndash4369

37 SproulD KitchenRR NestorCE DixonJM SimsAHHarrisonDJ RamsahoyeBH and MeehanRR (2012) Tissueof origin determines cancer-associated CpG island promoterhypermethylation patterns Genome Biol 13 R84

38 KeshetI SchlesingerY FarkashS RandE HechtMSegalE PikarskiE YoungRA NiveleauA CedarH et al(2006) Evidence for an instructive mechanism of de novomethylation in cancer cells Nat Genet 38 149ndash153

39 De CarvalhoDD SharmaS YouJS SuSF TaberlayPCKellyTK YangX LiangG and JonesPA (2012) DNAmethylation screening identifies driver epigenetic events of cancercell survival Cancer Cell 21 655ndash667

40 GazinC WajapeyeeN GobeilS VirbasiusCM andGreenMR (2007) An elaborate pathway required for Ras-mediated epigenetic silencing Nature 449 1073ndash1077

41 MeiFC YoungTW LiuJ and ChengX (2006) RAS-mediated epigenetic inactivation of OPCML in oncogenictransformation of human ovarian surface epithelial cellsFASEB J 20 497ndash499

42 LandanG CohenNM MukamelZ BarA MolchadskyABroshR Horn-SabanS ZalcensteinDA GoldfingerNZundelevichA et al (2012) Epigenetic polymorphism and thestochastic formation of differentially methylated regions in normaland cancerous tissues Nat Genet 44 1207ndash1214

43 LienertF WirbelauerC SomI DeanA MohnF andSchubelerD (2011) Identification of genetic elements thatautonomously determine DNA methylation states Nat Genet43 1091ndash1097

44 MacleodD CharltonJ MullinsJ and BirdAP (1994) Sp1 sitesin the mouse aprt gene promoter are required to preventmethylation of the CpG island Genes Dev 8 2282ndash2292

45 BlascoMA (2005) Telomeres and human disease ageing cancerand beyond Nat Rev Genet 6 611ndash622

46 MilyavskyM ShatsI ErezN TangX SenderovichSMeersonA TabachY GoldfingerN GinsbergD HarrisCCet al (2003) Prolonged culture of telomerase-immortalized humanfibroblasts leads to a premalignant phenotype Cancer Res 637147ndash7157

Nucleic Acids Research 2014 Vol 42 No 6 3541

at Edinburgh U

niversity on October 3 2014

httpnaroxfordjournalsorgD

ownloaded from