Deregulation of RB1 expression by loss of imprinting in human hepatocellular carcinoma

Journal of PathologyJ Pathol 2014; 233: 392–401Published online 16 June 2014 in Wiley Online Library(wileyonlinelibrary.com) DOI: 10.1002/path.4376

ORIGINAL PAPER

Deregulation of RB1 expression by loss of imprinting in humanhepatocellular carcinoma

Sumadi Lukman Anwar,1# Till Krech,1 Britta Hasemeier,1 Elisa Schipper,1 Nora Schweitzer,2 Arndt Vogel,2Hans Kreipe1 and Ulrich Lehmann1*

1 Institute of Pathology, Medizinische Hochschule Hannover, Hannover, Germany2 Department of Gastroenterology, Hepatology and Endocrinology, Medizinische Hochschule Hannover, Hannover, Germany

*Correspondence to: U Lehmann, Institute of Pathology, Medizinische Hochschule Hannover, Carl-Neuberg-Strasse 1, D-30625 Hannover, Germany.E-mail: [email protected]

#Present address: Department of Surgery, Faculty of Medicine, Universitas Gadjah Mada, Yogyakarta, Indonesia.

AbstractThe tumour suppressor gene RB1 is frequently silenced in many different types of human cancer, includinghepatocellular carcinoma (HCC). However, mutations of the RB1 gene are relatively rare in HCC. A systematic screenfor the identification of imprinted genes deregulated in human HCC revealed that RB1 shows imprint abnormalitiesin a high proportion of primary patient samples. Altogether, 40% of the HCC specimens (16/40) showed hyper-or hypomethylation at the CpG island in intron 2 of the RB1 gene. Re-analysis of publicly available genome-wideDNA methylation data confirmed these findings in two independent HCC cohorts. Loss of correct DNA methylationpatterns at the RB1 locus leads to the aberrant expression of an alternative RB1–E2B transcript, as measured byquantitative real-time PCR. Demethylation at the intron 2 CpG island by DNMT1 knock-down or aza-deoxycytidine(DAC) treatment stimulated expression of the RB1–E2B transcript, accompanied by diminished RB1 main transcriptexpression. No aberrant DNA methylation was found at the RB1 locus in hepatocellular adenoma (HCA, n = 10),focal nodular hyperplasia (FNH, n = 5) and their corresponding adjacent liver tissue specimens. Deregulated RB1expression due to hyper- or hypomethylation in intron 2 of the RB1 gene is found in tumours without loss ofheterozygosity and is associated with a decrease in overall survival (p = 0.032) if caused by hypermethylation ofCpG85. This unequivocally demonstrates that loss of imprinting represents an important additional mechanism forRB1 pathway inactivation in human HCC, complementing well-described molecular defects.Copyright © 2014 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Keywords: retinoblastoma gene (RB1); loss of imprinting (LOI); hepatocellular carcinoma; DNA methylation; epigenetics

Received 19 February 2014; Revised 30 April 2014; Accepted 11 May 2014

No conflicts of interest were declared.

Introduction

RB1 is the first identified tumour suppressor gene thatis frequently inactivated in several major cancers (see[1] and references therein). Germline mutations in thisgene are associated with a high risk of developingretinoblastoma [2]. The pRb protein plays a key rolein cell cycle regulation, particularly during the S–G2transition. Unphosphorylated pRb binds the E2F tran-scription factor, thereby down-regulating E2F targetgenes. Phosphorylation of pRb by cyclin-dependentkinases (CDKs) results in the release of E2F proteinsfrom the complex to initiate transcription of genesinvolved in cell-cycle progression [3]. In addition, pRbregulates several downstream pathways involved in celldifferentiation [4], apoptosis [5], DNA damage response[6], senescence [7,8] and embryogenesis [9]. In HCC,mutations in RB1 are rare events, although the pRb pro-tein is frequently down-regulated in HCC cells [10–12].

Several mechanisms have been proposed for the absenceof RB1 expression in HCC, including genetic loss, epi-genetic defects and post-transcriptional degradation[13–17].

A recent study using genome-wide DNA methy-lation analysis in a patient with extensive imprintingdefects discovered that the human RB1 gene is actuallyimprinted [18]. Genomic imprinting is a biologicalphenomenon leading to parent-of-origin-specific geneexpression, which is frequently disturbed in manydiseases, including cancer [19,20]. The differentiallymethylated region (DMR) showing parent-of-origin-specific DNA methylation at the RB1 gene is locatedwithin a CpG island in intron 2 (called ‘CpG85’; seeFigure 1A). In humans, this region is specifically methy-lated on the maternal allele, whereas in the mouse (andrat) genome RB1 is not imprinted, due to the absenceof the retrotransposon-derived differentially methylatedregion in exon 2 of the RB1 gene [21]. Differential

Copyright © 2014 Pathological Society of Great Britain and Ireland. J Pathol 2014; 233: 392–401Published by John Wiley & Sons, Ltd. www.pathsoc.org.uk www.thejournalofpathology.com

Loss of imprinting of RB1 in HCC 393

Figure 1. (A) Schema of the RB1 locus. The paternal and maternalalleles are depicted with the methylation status of the three CpGislands. The three CpG islands, CpG106, CpG42 and CpG85, contain106, 42 and 85 individual CpG dinucleotides, respectively. Thealternative exon 2B, with three different start sites of transcriptionfor the alternative RB1–E2B transcript, has been identified byKanber et al [18]. (B) Sequence of the exon 2B–exon 3 junction ofthe alternative RB1–E2B transcript verified in human HCC cell linesand primary human HCC specimens. The last nucleotide at the 3′

boundary of exon 2B (guanosine) is located at position 48894051 ofchromosome 13 (assembly GRCh37/hg 19). Shown is the sequenceobtained from HLE cDNA. Identical exon 2B–exon3 junctions wereidentified in cDNA from primary human HCC specimens.

methylation in this region regulates expression of thealternative transcript E2B, which is able to distort RB1main transcript expression to the maternal allele [18].Interestingly, the paternal transmission of a specific RB1mutation (IVS6+ 1) is associated with the developmentof more frequent tumours and with a more severe clinicalcourse than the maternal transmission of this mutation,thus indicating a parent-of-origin effect of the RB1 gene[22]. Therefore, we included RB1 in our recently initi-ated systematic screen for imprinting defects in humanHCC [23] and demonstrate unequivocally that loss ofimprinting (LOI) at the RB1 locus can be found in humanHCC and occurs independently of loss of heterozygosity(LOH). Hypomethylation at CpG85 leads to an increaseof alternative RB1–E2B transcript and concomitantdown-regulation of RB1 main transcript, whereashypermethylation is associated with reduced overallsurvival.

Materials and methods

Patient samples and cell linesA total of 104 primary fresh snap-frozen specimens,comprising 40 HCCs, 10 HCAs and five FNHs, togetherwith the corresponding adjacent liver tissues from 34HCCs, eight HCAs, two FNHs and five healthy liversamples, were retrieved from the archives of the Insti-tute of Pathology, Hannover Medical School (Germany)

and analysed anonymously, according to the guidelinesfrom the local Ethics Committee (’Ethik-Kommissionder Medizinischen Hochschule Hannover’, Head, Pro-fessor Dr Tröger). Tumour cell content was confirmedto be at least 70%. Clinicopathological variables aresummarized in Table S1 (see supplementary material).Two HCC cell lines (Huh7 and HepG2) were obtainedfrom the American Tissue Culture Collection (ATCC;Rockville, MD, USA) and cultured following ATCC’srecommendations. Two additional HCC cell lines (HLEand HLF) were available through the HCC researchconsortium, SFB-TRR 77 ’Liver Cancer’ [24]. Identityof all cell lines was validated using short tandem-repeatprofiling, following the protocol of the German Collec-tion of Microorganisms and Cell Cultures (available atwww.dsmz.de). All experiments using cell lines wereperformed in subconfluent density when the cells werein exponential cell growth.

DNA and RNA extractionIsolation of high molecular weight DNA from thefresh-frozen primary specimens was performed byovernight digestion with proteinase K, followed byphenol–chloroform extraction following standard pro-cedures. Total RNA was extracted using TRIZOLTM

(Life Technologies, Karlsruhe, Germany), following theprotocol provided by the manufacturer.

Bisulphite conversion and methylation analysisBisulphite modification of genomic DNAs was per-formed using an EZ DNA Methylation KitTM (ZymoResearch, HiSS Diagnostics, Freiburg, Germany),according to the manufacturer’s protocol. Approxi-mately 25 ng bisulphite-modified DNA was used forPCR amplification prior to pyrosequencing, which wasperformed as described [25]. Primers used for pyrose-quencing are listed in Table S2 (see supplementarymaterial). The methylation value for a given sample wascalculated as the mean of all individual values of allCpG dinucleotides, as determined by the software Pyro-Q-CpG™ (Qiagen, Hilden, Germany), from two inde-pendent pyrosequencing runs. ’Hypermethylated’ and’hypomethylated’ were defined as methylation valuesabove and below the mean of the adjacent liver tissue,± 2× standard deviation (SD; meanadj. ± 2× SD). Forbisulphite sequencing, PCR products were insertedinto plasmid vector using TOPO-TATM Cloning Kit(Life Technologies). Individual clones were purifiedwith QIAprep® Spin Miniprep Kit (Qiagen, Hilden,Germany) and sequenced with GenomeLabTM GeneticAnalysis System (Beckman Coulter, Krefeld, Germany),using vector primers (M13 and/or T7) and aGenomeLabTM DTCS Quick Start kit (Beckman Coul-ter). At least 10 clones were sequenced for every sample.All primer sequences are listed in Table S2 (see supple-mentary material). The selection of primer sequencesensures that the related loci on chromosomes 9 and 22do not influence the outcome of the methylation studies.


394 SL Anwar et al

The primers show several mismatches if aligned to theRB1 intron 2-related sequences on chromosomes 9 and22 [21]. Moreover, seven and eight of nine potentialmethylation sites interrogated by our pyrosequencingassay (see supplementary material, Figure S1) aremutated in the related sequences on chromosomes 9 and22, respectively.

Reverse transcription and quantitative real-timePCR (qRT–PCR)RNA (1 μg) was reverse-transcribed using a HighCapacity cDNA Reverse Transcription Kit (Life Tech-nologies), following the manufacturer’s protocol.Quantification of total RB1 transcripts was performedwith a TaqMan® Gene Expression AssayTM (Hs01078066_m1). GUSB (Hs00939627_m1) and GAPDH(Hs02758991_g1) were used for normalization (LifeTechnologies). qRT–PCR was performed in the ABIPrism 7500 Sequence Detection System, using Taq-Man Universal PCR Master Mix (Life Technologies).Expression in primary tumour samples was displayed asfold change relative to the expression level in the pairedperitumoural liver tissues. Main RB1 and alternativeRB1–E2B transcripts were quantified with real-timePCR in a 25 μl reaction volume with 400 nM primers(see supplementary material, Table S2, for all primersused in this study), 1:250 000 SYBR Green®, 1:100ROX, 200 μM dNTPs, 2.5 mM MgCl2, 1× Platinum-TaqPCR buffer and 1.25 U Platinum® Taq DNA polymerase(Invitrogen, Darmstadt, Germany).

DNMT1 knock-down and DAC treatment in HCC celllinesSiRNA-mediated DNMT1 knock-down in HLE cellsand DAC treatment of HLF, HuH7, and HepG2 cellswas performed essentially as described previously[23,26].

Detection of loss of heterozygosity and CTNNB1mutationFor the detection of LOH at the RB1 locus, first all34 HCC samples with available adjacent normal livertissue were tested with primers amplifying the VNTRmarker in intron 20. Those specimens homozygous inthe normal liver tissue for this marker were further anal-ysed using the marker D13S153. PCR was performed ina final volume of 25 μl, containing 20 ng genomic DNA,0.5 U Taq Polymerase (Platinium Taq, LifeTechnolo-gies), 200 μM dNTPs and 10 pM from each primer; forprimer sequences and PCR cycle conditions, see supple-mentary material, Table S2. Mutations in exon 3 of theCTNNB1 gene were analysed by bidirectional Sangersequencing, using primers described by Huss et al [27].

Western blot analysisProtein lysates (15 μg) were separated on 10% precastSDS–polyacrylamide gels (BioRad, Munich, Germany)and transferred onto Hybrid-P polyvinylidene difluoride(PVDF) membrane (Amersham Biosciences, Freiburg,Germany). The antibodies used were mouse monoclonalanti-DNMT1 antibody (IMG-261A, clone 60B1220.1,1:500 diluted; Imgenex, San Diego, CA, USA), mousemonoclonal anti-β-actin antibody (ab6276 clone AC-15,1:5000 diluted; Abcam, UK) and anti-mouse secondaryantibody HRP (R1253HRP, 1:10 000 diluted; Acris,Herford, Germany).

Statistical analysisStatistical analyses were performed using GraphPadPrism v. 5.01 for Windows (La Jolla, CA, USA). TheMann–Whitney U-test was used to compare continuousvariables (DNA methylation levels, fold changes of RB1main and alternative transcript expression, overall sur-vival) and χ2 test for relationships between categoricalvariables (hypermethylation versus hypomethylation,up-regulation versus down-regulation, LOH versus no

Figure 2. Aberrant DNA methylation at RB1 imprinted locus (CpG85) in HCC. All individual methylation values from 40 primary HCCspecimens and the 34 adjacent liver tissues are displayed. Three regions within the RB1 gene were analysed: (A) CpG106 (promoter);(B) CpG42; and (C) CpG85 (RB1 imprinting control region in intron 2). The pyrosequencing assays interrogated in CpG106 11 CpGs, inCpG42 eight CpGs, and in CpG85 nine CpGs; black triangles, HCCs; grey triangles, peritumoural liver tissues; horizontal dotted lines in rightpanel, the 95% confidence interval. For an overview of the gene structure, see Figure 1A.



LOH). Kaplan–Meier curve and long-rank (Gehan–Breslow–Wilcoxon) tests were used to compare survivalof the HCC patients. p< 0.05 was considered statisti-cally significant.

Results

Aberrant DNA methylation at RB1 imprinted locusin human HCCA systematic screen for imprinted loci deregulated inhuman HCC revealed that the well-known tumour sup-pressor gene RB1 shows distinct hyper- or hypomethy-lation at a newly identified differentially methylatedregion (DMR) in a high number of specimens. Employ-ing quantitative high-resolution pyrosequencing, 16 of40 specimens (40%) displayed DNA methylation abnor-malities in CpG island 85, located within intron 2 ofthe RB1 gene (Figure 2); Figure S2 (see supplementarymaterial) shows the methylation level for tumour and

adjacent liver tissue on a case-by-case basis. This regionwas recently identified by Kanber et al [18] as a DMRdisplaying parent-of-origin-specific DNA methylation,regulating the expression of an alternative RB1 tran-script (RB1–E2B). In contrast to CpG85, CpG islands106 and 42 showed no changes in DNA methylationlevel. CpG106 is located in the promoter region and dis-plays only low-level methylation in tumour and adjacentliver in all samples under study (mean 10.2%), whereasCpG42, located in intron 2 upstream of CpG85, showsconstitutive high DNA methylation (mean 82.6%).

Using publicly available datasets, we re-analysedDNA methylation at these three CpG islands in twoindependent HCC cohorts comprising 63 and 62 cases,respectively [28,29]. Both studies employed theInfinium HumanMethylation27k array, which analysesthree CpG dinucleotides in CpG106, two CpGs inCpG42 and four CpGs in CpG85. Differential DNAmethylation was not observed in CpG106 and CpG42.However, hyper- or hypomethylation at CpG85 werefrequently detected (see supplementary material,

Figure 3. Methylation at the RB1-imprinted locus (CpG85) correlates with RB1 gene expression. (A) Total RB1 expression correlatedsignificantly with the methylation level at the RB1-imprinted locus (Spearman’s correlation, r = 0.47, p = 0.006; the single outlier wasexcluded from the statistical analysis). Expression of total RB1 transcript (RB1 main transcript plus alternative RB1–E2B transcript) wasanalysed by TaqMan assays. (B) Main transcript and alternative transcript RB1–E2B expression levels were measured individually andshowed an inverse correlation. (C) Main RB1 transcript expression also showed a positive correlation with DNA methylation at CpG85(Spearman’s correlation, r= 0.42, p= 0.01: leaving out the three potential outliers changed the correlation only slightly; r= 0.45, p= 0.01).


396 SL Anwar et al

Figure S3). These results confirm our findings inindependent patient cohorts with comparable frequen-cies, using a completely different methodology.

Similar to the situation in morphologically incon-spicuous liver tissue from HCC patients (’adjacent livertissue’) in healthy liver samples (n= 5), CpG106 is notmethylated (mean methylation level 13%), CpG42 isfully methylated (mean methylation level 85.8%) andCpG85 is partially methylated (mean methylation level56%) (see supplementary material, Figure S4).

Bisulphite sequencing revealed very homogeneousDNA methylation patterns of fully methylated and fullyunmethylated alleles in CpG85 in six tumour spec-imens analysed (see supplementary material, FigureS5). Together with ca. 50% methylation level in thisregion in adjacent non-tumourous liver tissue, theseresults strongly support the notion of allele-specificDNA methylation at CpG85 in the RB1 gene, despitethe absence of informative SNPs in the region understudy in our cohort.

Aberrant RB1 and RB1–E2B expression in humanHCCIn order to evaluate the relationship between hyper- orhypomethylation of CpG85 and RB1 expression, wemeasured total RB1 mRNA expression in primary HCCspecimens, using quantitative real-time PCR (see sup-plementary material, Figure S6, for transcript-specificassay design). DNA methylation at CpG85 correlatedpositively with total RB1 mRNA expression (p= 0.006,Figure 3A). The alternative RB1–E2B transcript nega-tively affects RB1 main transcript expression throughtranscriptional interference [18]. We therefore quan-tified expression of the RB1–E2B transcript (usingalternative exon E2B and exon 3 junction overlappingprimers; see supplementary material, Figure S6) andRB1 main transcript (with exon 1–2 junction overlap-ping primers). As expected, expression of RB1–E2Btranscripts correlates inversely with the main tran-script (R2 = –0.41 and p= 0.026; Figure 3B) in primaryhuman HCC specimens. DNA methylation at CpG85also showed positive correlation with RB1 main tran-script expression (R2 = 0.42 and p= 0.01; Figure 3C).Direct sequencing of PCR products generated byRB1–E2B-specific primers from cell-line and tumourcDNA confirmed the transcription of the alternativeexon 2B fused to exon 3 (see Figure 1B), as originallydescribed by Kanber et al [18]. This demonstrates forthe first time the expression of the alternative RB1–E2Btranscript in primary human tumour specimens.

Forced expression of RB1–E2B transcriptdiminishes RB1 main transcript expressionTo elucidate the role of hyper- and hypomethylationat CpG85 in regulating the expression of RB1 mainand alternative transcripts, DNMT1 knock-down wasperformed in the human HCC cell line HLE. AftersiRNA-mediated DNMT1 knock-down (see supplemen-tary material, Figure S7), significant DNA methylation

reduction at CpG85 was observed (Figure 4A, upperpanel). The reduction of DNA methylation was accom-panied by a significant increase in RB1–E2B alternativetranscript expression and a significant decrease in RB1main transcript expression (Figure 4B). Treatment ofthree different HCC cell lines (HLF, HuH7 and HepG2)with the DNMT inhibitor 5-aza-2-deoxycytidine (DAC)produced similar results; a significant reduction inDNA methylation at CpG85 after DAC treatment isaccompanied by increased expression of RB1–E2Balternative transcript and diminished expression of RB1main transcript (Figure 4B).

DNA methylation at RB1 locus in benign livertumoursTo address whether hyper- or hypomethylation atthe RB1-imprinted locus is specific for malignanttransformation in liver cells, benign liver tumours (hep-atocellular adenoma, HCA, n= 10, and focal nodularhyperplasia, FNH, n= 5), together with the corre-sponding adjacent liver tissue, were analysed. Similarto the DNA methylation patterns in healthy liver, allbenign liver tumour specimens showed no or low DNAmethylation at CpG106, high methylation at CpG42and intermediate methylation (∼50%) at CpG85 (seesupplementary material, Figure S8).

Correlation of RB1 LOI with other molecularand clinical parametersSince loss of heterozygosity (LOH) is a classical,well-established and frequent mechanism for tumoursuppressor gene inactivation in general and is welldescribed for the RB1 gene in HCC [14], we analysedthe presence or absence of LOH at the RB1 locus in ourHCC cohort. LOH is clearly detectable in 16 of 34 cases(47%), informative either for the VNTR in intron 20 orD13S153. Imprint aberrations (gain or loss of methy-lation at CpG85) correlate statistically significantlywith LOH (Fisher’s exact test, p= 0.014), indicating acooperative effect of these two mechanisms. From the16 cases with LOH, six display hypermethylation andfive hypomethylation. From the 18 cases without LOH,three display hypermethylation and one hypomethyla-tion, representing cases with unambiguous LOI. FigureS9 (see supplementary material) presents all data fromFigures 2, 3 and S2, separated into three groups (LOH,LOI without LOH, and without LOH or LOI).

A mutation of CTNNB1 (β-catenin) exon 3 leads toconstitutive activation of the Wnt signalling pathwayand is the most frequent oncogene mutation in livertumours (see [30] and references therein). In our series itis found in nine of 34 paired HCC samples (26.5%), butwas correlated with neither the occurrence of LOI northe presence of LOH at the RB1 locus (data not shown).Neither LOI and LOH correlate with the presence orabsence of hepatitis infection in our cohort (data notshown).

To gain first insights into the clinical impact of imprintaberrations at the RB1 locus, the relation between DNA



Figure 4. Demethylation at the RB1-imprinted locus leads to increased expression of alternative RB1–E2B transcript and decreasedexpression of RB1 main transcript. (A) DNA methylation analysis in HLE cells after DNMT1 knock-down and after DAC treatment in HLF,HuH7 and HepG2 cells. (B) Expression of RB1 main transcript and E2B alternative transcript in HLE cells after DNMT1 knock-down andafter DAC treatment in HLF, HuH7 and HepG2 cells: ***p < 0.001; **0.001 < p < 0.01; *0.05 < p < 0.01.


398 SL Anwar et al

methylation abnormalities at CpG85 and overall survivalwas analysed in our HCC cohort. Gain of DNAmethylation at RB1 CpG85 as a marker for imprintinstability is associated with reduced overall survival(Gehan–Breslow–Wilcoxon test, p= 0.032; Figure 5).LOH, together with reduced E2B expression, is alsoassociated with reduced survival (Gehan–Breslow–Wilcoxon test, p= 0.02). If the mean survival is com-pared for various dichotomizations of the patient cohort(eg hyper- or hypomethylation versus proper imprint-ing, reduced LINE methylation versus unaltered LINE1methylation, LOH versus no LOH, down-regulationof RB1 or E2B versus no down-regulation), a strikingfeature is that the largest differences in mean survivalare always observed if DNA methylation parametersare compared and not if genetic defects (LOH) or tran-scriptional alterations (up- or down-regulation of RB1or E2B) are used for dichotomization (data not shown).However, due to the modest sample size, not all of thesedichotomizations reach statistical significance, despitelarge differences in mean survival (due to spreadingof the data). Therefore, these associations have to beconfirmed in larger independent patient cohorts.

Discussion

Erroneous regulation of genomic imprinting has beenassociated with many types of human cancer, not onlyin embryonic tumours but also in colorectal cancer [31],breast cancer [32], ovarian cancer [33], lung cancer [34]and leukaemia [35]. However, only a limited numberof studies thus far have addressed the dysregulationof genomic imprinting in HCC, one of the most com-mon types of cancer worldwide. Studies of genomicimprinting in HCC have mostly focused on a few loci[36], primarily the well-characterized IGF2–H19 locus[37]. We could identify only 12 publications in PubMed(1991–January 2014) analysing allele-specific DNAmethylation and/or expression of imprinted loci in

Figure 5. Aberrant DNA methylation at the RB1 DMR correlateswith poor HCC survival. Median survival rates, hypermethylationand normo/hypomethylation were 34 and 156 weeks, respectively(p = 0.032).

primary human HCC specimens (excluding all animaland cell line studies and all studies not addressingspecifically allele-specific DNA methylation and/orexpression, and evaluating only studies published inEnglish) [23,38–48].

Loss of RB1 expression in HCC results in tumourmultiplicity and chromosomal instability [49]. Liver-specific genetic ablation of RB1 in mice leads to poly-ploidy and chromosomal instability, which are bothclosely linked to liver tumourigenesis [50,51]. Severalmechanisms for RB1 inactivation have been impli-cated in HCC, ie LOH [13,14,52], hypermethylationp16INK4A or p15INK4B [16], amplification of cyclin D1[53], over-expression of gankyrin [15] and complexationby viral proteins HBx, preS2 or NS5B [54,55]. Promoterhypermethylation as a further mechanism of RB1 geneinactivation has been reported in 21% (n= 33) [56] and26% (n= 91) of cases [10]. However, Yu et al [57] andPark et al [58] found no methylation change in HCCtumours (n= 29 and n= 27, respectively). Those fourstudies used methylation-specific PCR (MS–PCR) forDNA methylation analysis. Using a quantitative method,we did not observe hypermethylation at the RB1 pro-moter (CpG106) in our HCC series. Also, in the Illumina27 k methylation array datasets [28,29], no evidence forRB1 promoter methylation could be found (see supple-mentary material, Figure S3). These discrepancies aremost probably due to the different methodologies used.Quantitative methylation techniques, such as pyrose-quencing or the 27 k Illumina methylation array, enablea stringent threshold definition, avoiding the scoring ofbackground signals as ’hypermethylation’. MS–PCR ishighly sensitive and cost-effective for DNA methylationdetection but often gives false-positive results [59].

During the preparation of this manuscript, Rumbajanet al [60] reported imprint defects in a series of 12 hep-atoblastomas, the most frequent primary liver tumour inchildren. Among other loci RB1 was also analysed andshown to harbour gain of DNA methylation in five of12 cases at CpG85 (42%). However, no data concerningthe expression of RB1 main and alternative transcriptwere reported.Our results demonstrate not only hyper-or hypomethylation at RB1 DMR but also dysregu-lation of RB1 main transcript and expression of analternative RB1–E2B transcript. This is the first reportdemonstrating RB1 LOI and concomitant expressionof the alternative RB1–E2B transcript in cultured andprimary human tumour cells (Figure 1B). Re-analysis ofpublicly available methylation data [28,29] confirmeddifferential methylation of CpG85 in intron 2 in twoindependent HCC cohorts (see supplementary material,Figure S3). Unfortunately, survival and mRNA expres-sion data are not available for these cohorts. The DNAmethylation pattern at the RB1 locus in benign livertumours and the peritumoural tissue is undistinguish-able from the pattern in healthy liver specimens (seesupplementary material, Figures S4, S8), indicatingthat epigenetic instability at the RB1-imprinted locusis specific for the process of malignant transforma-tion, and not a by-product of increased proliferation.



The largest differences in mean survival are observedif the cohort is dichotomized along DNA methyla-tion features (and not along genetic or transcriptionalaberrations). Even though these associations haveto be confirmed in independent (and preferentiallylarger) cohorts, they represent a strong argument that’imprint instability’ at the RB1 locus has an impact onsurvival.

A search for open reading frames (ORFs) containedwithin the alternative E2B transcript revealed multiplesmall ORFs in all reading frames on both strands, anda single large ORF in the same orientation and readingframe as the main pRb ORF, encoding a shortenedversion of pRb missing the first 112 amino acids.This deletion affects a large portion of the N-terminaldomain (RbN), which is important for E2F release viainteraction with the pRb pocket domain (see [61] andreferences cited therein). The enormous complexityof the pRb interactome [62] makes it very likely thatadditional functions are affected by this N-terminal112 amino acids deletion. The molecular structure andcellular functions of this putative new member of theRb family have to be addressed in future studies: Ithas to be shown whether this shortened 93 kDa RB1variant is indeed expressed and whether it is identicalto the N-terminally truncated 94 kDa isoform alreadyidentified by Hu et al 20 years ago [63].

Our data confirm, on the one hand, that LOH is thepredominant mechanism for RB1 inactivation in humanHCC but, on the other hand, demonstrate unequivocallythe presence of LOI in the absence of LOH (in four of18 cases). These findings identify a novel mechanismfor RB1 inactivation in HCC through loss of genomicimprinting. These results also amend the currently stillwidely-held simplistic view that loss of methylationalways leads to an increase in mRNA expression (see[64] for further exploration of this point) and confirm theimportance of quantitative methods for the analysis ofDNA methylation patterns in health and disease. Loss ofpRb function in HCC has been associated with increasedploidy, genomic instability and poor prognosis, there-fore epigenetic instability at the RB1 locus resulting inreduced pRb activity might be used in the future as anew biomarker for diagnosis and prognosis in HCC.In addition, all clinical trials evaluating the usefulnessof Cdk inhibitors for the prevention of pRb inactiva-tion have to consider this new mechanism for RB1inactivation.

Acknowledgements

UL and HK received a grant from the German ResearchCouncil, Deutsche Forschungsgemeinschaft (DFG;Grant No. SFB-TRR77, ’Liver cancer’ (Project B1);SLA was supported by a PhD fellowship from theMolecular Medicine programme in Hannover Biomed-ical Research School (HBRS), Hannover MedicalSchool, Germany. The funding bodies did not have

any role in study design, data mining and evaluation orpreparation of the manuscript.

Author contributions

SLA and UL conceived and planned the study withsupport from AV and HK. SLA, BH, and ES carriedout all experiments and prepared all figures, TK andHK selected and contributed the histopathological spec-imens and analysed data, NS and AV collected clinicaldata and analysed data. SLA and UL evaluated all dataand wrote the manuscript with support from TK, AV,and HK. All authors were involved in writing the finalversion of the manuscript and had approved it for sub-mission.

References1. Classon M, Harlow E. The retinoblastoma tumour suppressor in

development and cancer. Nat Rev Cancer 2002; 2: 910–917.2. Lohmann DR. RB1 gene mutations in retinoblastoma. Hum Mutat

1999; 14: 283–288.3. Knudsen ES, Knudsen KE. Retinoblastoma tumor suppressor: where

cancer meets the cell cycle. Exp Biol Med 2006; 231: 1271–1281.4. Calo E, Quintero-Estades JA, Danielian PS, et al. Rb regulates

fate choice and lineage commitment in vivo. Nature 2010; 466:1110–1114.

5. Ciavarra G, Zacksenhaus E. Multiple pathways counteract cell deathinduced by RB1 loss: implications for cancer. Cell Cycle 2011; 10:1533–1539.

6. van Harn T, Foijer F, van Vugt M, et al. Loss of Rb proteins causesgenomic instability in the absence of mitogenic signaling. Genes Dev

2010; 24: 1377–1388.7. Chen T, Xue L, Niu J, et al. The retinoblastoma protein selectively

represses E2F1 targets via a TAAC DNA element during cellularsenescence. J Biol Chem 2012; 287: 37540–37551.

8. Aksoy O, Chicas A, Zeng T, et al. The atypical E2F family memberE2F7 couples the p53 and RB pathways during cellular senescence.Genes Dev 2012; 26: 1546–1557.

9. Sage J. The retinoblastoma tumor suppressor and stem cell biology.Genes Dev 2012; 26: 1409–1420.

10. Edamoto Y, Hara A, Biernat W, et al. Alterations of RB1, p53 andWnt pathways in hepatocellular carcinomas associated with hepatitisC, hepatitis B and alcoholic liver cirrhosis. Int J Can 2003; 106:334–341.

11. Laurent-Puig P, Zucman-Rossi J. Genetics of hepatocellular tumors.Oncogene 2006; 25: 3778–3786.

12. Knudsen ES, Knudsen KE. Tailoring to RB: tumour suppressor statusand therapeutic response. Nat Rev Cancer 2008; 8: 714–724.

13. Zhang X, Xu HJ, Murakami Y, et al. Deletions of chromosome 13q,mutations in Retinoblastoma 1, and retinoblastoma protein state inhuman hepatocellular carcinoma. Cancer Res 1994; 54: 4177–4182.

14. Piao Z, Kim H, Jeon BK, et al. Relationship between loss of het-erozygosity of tumor suppressor genes and histologic differentiationin hepatocellular carcinoma. Cancer 1997; 80: 865–872.

15. Higashitsuji H, Itoh K, Nagao T, et al. Reduced stability ofretinoblastoma protein by gankyrin, an oncogenic ankyrin-repeatprotein overexpressed in hepatomas. Nat Med 2000; 6: 96–99.

16. Qin Y, Liu JY, Li B, et al. Association of low p16INK4a andp15INK4b mRNAs expression with their CpG islands methylationwith human hepatocellular carcinogenesis. World J Gastroenterol

2004; 10: 1276–1280.


400 SL Anwar et al

17. Temming P, Corson TW, Lohmann DR. Retinoblastoma tumorigene-sis: genetic and epigenetic changes walk hand in hand. Future Oncol

2012; 8: 525–528.18. Kanber D, Berulava T, Ammerpohl O, et al. The human retinoblas-

toma gene is imprinted. PLoS Genet 2009; 5: e1000790.19. Robertson KD. DNA methylation and human disease. Nat Rev Genet

2005; 6: 597–610.20. Horsthemke B. In Brief: Genomic imprinting and imprinting dis-

eases. J Pathol 2014; 232:485–487.21. Kanber D, Buiting K, Roos C, et al. The origin of the RB1 imprint.

PLoS One 2013; 8: e81502.22. Klutz M, Brockmann D, Lohmann DR. A parent-of-origin effect

in two families with retinoblastoma is associated with a distinctsplice mutation in the RB1 gene. Am J Hum Genet 2002; 71:174–179.

23. Anwar SL, Krech T, Hasemeier B, et al. Loss of imprinting andallelic switching at the DLK1–MEG3 locus in human hepatocellularcarcinoma. PLoS One 2012; 7: e49462.

24. Buurman R, Gurlevik E, Schaffer V, et al. Histone deacetylases acti-vate hepatocyte growth factor signaling by repressing microRNA-449in hepatocellular carcinoma cells. Gastroenterology 2012; 143:811–820, e811–815.

25. Potapova A, Albat C, Hasemeier B, et al. Systematic cross-validationof 454 sequencing and pyrosequencing for the exact quantificationof DNA methylation patterns with single CpG resolution. BMC

Biotechnol 2011; 11: 6.26. Anwar SL, Albat C, Krech T, et al. Concordant hypermethylation

of intergenic microRNA genes in human hepatocellular carcinomaas new diagnostic and prognostic marker. Int J Can 2013; 133:660–670.

27. Huss S, Nehles J, Binot E, et al. β-Catenin (CTNNB1) mutations andclinicopathological features of mesenteric desmoid-type fibromato-sis. Histopathology 2013; 62: 294–304.

28. Neumann O, Kesselmeier M, Geffers R, et al. Methylome analysisand integrative profiling of human HCCs identify novel protumori-genic factors. Hepatology 2012; 56:1817–1827.

29. Shen J, Wang S, Zhang YJ, et al. Genome-wide DNA methyla-tion profiles in hepatocellular carcinoma. Hepatology 2012; 55:1799–1808.

30. Zucman-Rossi J. Molecular classification of hepatocellular carci-noma. Dig Liver Dis 2010; 42(suppl 3): S235–241.

31. Cui H, Onyango P, Brandenburg S, et al. Loss of imprinting in col-orectal cancer linked to hypomethylation of H19 and IGF2. Cancer

Res 2002; 62: 6442–6446.32. Yuan J, Luo RZ, Fujii S, et al. Aberrant methylation and silencing of

ARHI, an imprinted tumor suppressor gene in which the function islost in breast cancers. Cancer Res 2003; 63: 4174–4180.

33. Huang Z, Wen Y, Shandilya R, et al. High-throughput detection ofM6P/IGF2R intronic hypermethylation and LOH in ovarian cancer.Nucleic Acids Res 2006; 34: 555–563.

34. Kohda M, Hoshiya H, Katoh M, et al. Frequent loss of imprinting ofIGF2 and MEST in lung adenocarcinoma. Mol Carcinog 2001; 31:184–191.

35. Vorwerk P, Wex H, Bessert C, et al. Loss of imprinting of IGF-II

gene in children with acute lymphoblastic leukemia. Leuk Res 2003;27: 807–812.

36. Uribe-Lewis S, Woodfine K, Stojic L, et al. Molecular mechanismsof genomic imprinting and clinical implications for cancer. Expert

Rev Mol Med 2011; 13: e2.37. Chao W, D’Amore PA. IGF2: epigenetic regulation and role in

development and disease. Cytokine Growth Factor Rev 2008; 19:111–120.

38. Takeda S, Kondo M, Kumada T, et al. Allelic-expression imbalanceof the insulin-like growth factor 2 gene in hepatocellular carcinomaand underlying disease. Oncogene 1996; 12: 1589–1592.

39. Kim KS, Lee YI. Biallelic expression of the H19 and IGF2 genes in

hepatocellular carcinoma. Cancer Lett 1997; 119: 143–148.

40. Li X, Nong Z, Ekstrom C, et al. Disrupted IGF2 promoter control by

silencing of promoter P1 in human hepatocellular carcinoma. Cancer

Res 1997; 57: 2048–2054.

41. Uchida K, Kondo M, Takeda S, et al. Altered transcriptional regula-

tion of the insulin-like growth factor 2 gene in human hepatocellular

carcinoma. Mol Carcinog 1997; 18: 193–198.

42. Schwienbacher C, Gramantieri L, Scelfo R, et al. Gain of imprint-

ing at chromosome 11p15: a pathogenetic mechanism identified

in human hepatocarcinomas. Proc Natl Acad Sci USA 2000; 97:5445–5449.

43. Scelfo RA, Schwienbacher C, Veronese A, et al. Loss of methylation

at chromosome 11p15.5 is common in human adult tumors. Onco-

gene 2002; 21: 2564–2572.

44. Poirier K, Chalas C, Tissier F, et al. Loss of parental-specific methy-

lation at the IGF2 locus in human hepatocellular carcinoma. J Pathol

2003; 201: 473–479.

45. Midorikawa Y, Yamamoto S, Ishikawa S, et al. Molecular karyotyp-

ing of human hepatocellular carcinoma using single-nucleotide poly-

morphism arrays. Oncogene 2006; 25: 5581–5590.

46. Huang J, Zhang X, Zhang M, et al. Up-regulation of DLK1 as an

imprinted gene could contribute to human hepatocellular carcinoma.

Carcinogenesis 2007; 28: 1094–1103.

47. Wu J, Qin Y, Li B, et al. Hypomethylated and hypermethylated

profiles of H19DMR are associated with the aberrant imprinting of

IGF2 and H19 in human hepatocellular carcinoma. Genomics 2008;

91: 443–450.

48. Braconi C, Kogure T, Valeri N, et al. microRNA-29 can regulate

expression of the long non-coding RNA gene MEG3 in hepatocellular

cancer. Oncogene 2011; 30: 4750–4756.

49. Srinivasan SV, Mayhew CN, Schwemberger S, et al. RB loss pro-

motes aberrant ploidy by deregulating levels and activity of DNA

replication factors. J Biol Chem 2007; 282: 23867–23877.

50. Mayhew CN, Bosco EE, Fox SR, et al. Liver-specific pRB loss results

in ectopic cell cycle entry and aberrant ploidy. Cancer Res 2005; 65:4568–4577.

51. Mayhew CN, Carter SL, Fox SR, et al. RB loss abrogates cell

cycle control and genome integrity to promote liver tumorigenesis.

Gastroenterology 2007; 133: 976–984.

52. Zondervan PE, Wink J, Alers JC, et al. Molecular cytogenetic eval-

uation of virus-associated and non-viral hepatocellular carcinoma:

analysis of 26 carcinomas and 12 concurrent dysplasias. J Pathol

2000; 192: 207–215.

53. Betticher DC, Heighway J, Hasleton PS, et al. Prognostic signifi-

cance of CCND1 (cyclin D1) overexpression in primary resected

non-small-cell lung cancer. Br J Cancer 1996; 73: 294–300.

54. Kim YJ, Jung JK, Lee SY, et al. Hepatitis B virus X protein

overcomes stress-induced premature senescence by repressing

p16(INK4a) expression via DNA methylation. Cancer Lett 2010;

288: 226–235.

55. McGivern DR, Villanueva RA, Chinnaswamy S, et al. Impaired

replication of hepatitis C virus containing mutations in a con-

served NS5B retinoblastoma protein-binding motif. J Virol 2009; 83:7422–7433.

56. Roncalli M, Bianchi P, Bruni B, et al. Methylation framework of

cell cycle gene inhibitors in cirrhosis and associated hepatocellular

carcinoma. Hepatology 2002; 36: 427–432.

57. Yu J, Ni M, Xu J, et al. Methylation profiling of twenty

promoter-CpG islands of genes which may contribute to hepa-

tocellular carcinogenesis. BMC Cancer 2002; 2: 29.

58. Park HJ, Yu E, Shim YH. DNA methyltransferase expression and

DNA hypermethylation in human hepatocellular carcinoma. Cancer

Lett 2006; 233: 271–278.



59. Kristensen LS, Hansen LL. PCR-based methods for detectingsingle-locus DNA methylation biomarkers in cancer diagnostics,prognostics, and response to treatment. Clin Chem 2009; 55:1471–1483.

60. Rumbajan JM, Maeda T, Souzaki R, et al. Comprehensive analy-ses of imprinted differentially methylated regions reveal epigeneticand genetic characteristics in hepatoblastoma. BMC Cancer 2013;13: 608.

61. Rubin SM. Deciphering the retinoblastoma protein phosphorylationcode. Trends Biochem Sci 2013; 38: 12–19.

62. Hasan MM, Brocca S, Sacco E, et al. A comparative study of Whi5and retinoblastoma proteins: from sequence and structure analysis tointracellular networks. Front Physiol 2014; 4: 315.

63. Xu HJ, Xu K, Zhou Y, et al. Enhanced tumor cell growth suppressionby an N-terminal truncated retinoblastoma protein. Proc Natl AcadSci USA 1994; 91: 9837–9841.

64. Reddington JP, Pennings S, Meehan RR. Non-canonical functionsof the DNA methylome in gene regulation. Biochem J 2013; 451:13–23.

SUPPLEMENTARY MATERIAL ON THE INTERNETThe following supplementary material may be found in the online version of this article:

Figure S1. DNA sequence of the region of CpG85 analysed by bisulphite sequencing and pyrosequencing.

Figure S2. DNA methylation level at CpG85 (measured by pyrosequencing) in HCC and adjacent liver tissue on a case-by-case level.

Figure S3. Re-analysis of genome-wide DNA methylation data in HCC samples at RB1 CpG106, CpG42 and CpG85.

Figure S4. DNA methylation analysis at RB1 CpG106, CpG42, and CpG85 in healthy liver tissues.

Figure S5. Bisulphite sequencing results for individual clones for the RB1-imprinted locus in primary HCC specimens.

Figure S6. Location of primers for RT–PCR enabling specific detection of main RB1 transcript and alternative E2B transcript as well as totalRB1 mRNA.

Figure S7. Reduction of DNMT1 mRNA and protein expression in HLE cells via siRNA knock-down.

Figure S8. DNA methylation analysis at RB1 CpG106, CpG42, and CpG85 in benign liver tumours.

Figure S9. Data from Figures 2, 3 and S2 separated into groups I–III.

Table S1. Clinicopathological parameters of the study cohort.

Table S2. List of all primers used in this study.


Deregulation of RB1 expression by loss of imprinting in human hepatocellular carcinoma

Documents

Transcript of Deregulation of RB1 expression by loss of imprinting in human hepatocellular carcinoma