Instability of Retroviral DNA Methylation in Embryonic Stem Cells

Authors

  • Shigeru Minoguchi,

    1. Division of Host-Parasite Interaction, Department of Microbiology and Immunology, Institute of Medical Science, University of Tokyo, Tokyo, Japan
    Search for more papers by this author
  • Hideo Iba Ph.D.

    Corresponding author
    1. Division of Host-Parasite Interaction, Department of Microbiology and Immunology, Institute of Medical Science, University of Tokyo, Tokyo, Japan
    • Division of Host-Parasite Interaction, Department of Microbiology and Immunology, Institute of Medical Science, University of Tokyo, 4-6-1 Shirokanedai, Minato-ku Tokyo, 108-8639, Japan. Telephone: 81-3-5449-5730; Fax: 81-3-5449-5449
    Search for more papers by this author

Abstract

The epigenetic status of pluripotent stem cells has been demonstrated to be extremely unstable. In our current study, we have attempted to further investigate the epigenetic dynamics of the stem cell genome by monitoring the expression of the murine stem cell virus (MSCV) retroviral vector in embryonic stem (ES) cells. Although MSCV is progressively silenced by proviral DNA methylation in ES cells, a substantial number of MSCV-transduced ES cell clones do show variegated proviral expression. This expression profile is due in part to the transient and reversible properties of MSCV silencing. However, the spontaneous reactivation rates of the silenced proviruses differ significantly between these variegated clones, indicating that the reversibility of silencing is dependent on the proviral integration site. Our current data suggest that the fidelity of DNA methylation among the genomic sequences that flank the proviral integration sites may be the determinant of this reversibility of MSCV silencing. Given that the adjoining epigenome environment affects the epigenetic regulation of proviral DNA, the reversible MSCV silencing effect is thus likely to reflect a unique and interesting feature of ES cell epigenome regulation that has not previously been revealed.

Disclosure of potential conflicts of interest is found at the end of this article.

Introduction

Embryonic stem (ES) cells can give rise to all cell types during ontogeny, and the pluripotency of these cells is maintained during self-renewal by the epigenetic control of their gene expression. The recent epigenetic profiling of undifferentiated ES cells has revealed the presence of bivalent chromatin, in which both active and repressive histone modifications are juxtaposed, and this conflicting histone code has been postulated to maintain the developmental regulator genes in an expression-ready state [1, 2]. In this regard, the DNA methylation of the promoters of the Nanog and Oct4 genes, which are pivotal regulators of pluripotency, is tightly associated with their repression in differentiated cells [3, [4]–5]. Hence, a deeper understanding of the epigenetic features that distinguish ES cells from differentiated cells will be an important development in stem cell biology.

Although the epigenetic profiling of stem cells has been the subject upon which many studies have focused to date, the dynamic features of the stem cell epigenome have not been well characterized. This is an important issue as although there is an obviously great potential for the use of ES cells as therapeutic tools for degenerative diseases, a number of previous studies have raised concerns about the epigenetic instability of these cells [6, 7]. ES cell lines can differ in their DNA methylation profiles during their derivation, and methylation changes can then further accumulate during prolonged culture [8]. The epigenetic dysregulation of the ES cell genome also seems to occur more frequently than the developmental abnormalities that are observed in the recipient embryos [7]. Hence, the epigenotyping of ES cells should be adopted as a prerequisite safety evaluation before their use in therapeutic applications. In any event, the possibility that epigenetic instability is an intrinsic property of stem cells indicates that an elucidation of the molecular mechanisms underlying this epigenetic drift will be important for our understanding of the ES cell genome.

Undifferentiated ES cells are nonpermissive for the provirus transcription of the Moloney murine leukemia virus (MLV) retroviral vector. Although the murine stem cell virus (MSCV) vector, a variant of MLV, can escape this immediate block and initiate provirus expression in ES cells, substantial populations of the resulting active proviruses progressively lose their expression during long periods of culture [9]. Proviral DNA methylation plays a causative role in this MSCV silencing at later time points of transduction, as this virus does not show such long-term silencing in ES cells that are deficient for the maintenance DNA methyltransferase, Dnmt1 [9]. Consistent with this, silenced proviruses can be reactivated by treatment with 5-aza-2′-deoxycytidine (5-aza-CdR), a potent inhibitor of DNA methylation [9]. Intriguingly, however, MSCV silencing can still occur in ES cells that are homozygous for mutations in the de novo DNA methyltransferases, Dnmt3a and Dnmt3b [10]. Since histone H3 hypoacetylation is also closely associated with MSCV silencing in both wild-type and de novo methyltransferase mutant ES cells, epigenetic chromatin modifications also play a key role in the establishment of these silenced states [11].

In our current study, we have attempted to further address the epigenetic dynamics of the ES cell genome by monitoring MSCV silencing. As shown previously, however, a substantial number of clonal proviral integrants display variegated expression, which is at least in part due to the reversibility of gene silencing [11, 12]. However, we show in our present analyses that such instability of silencing is dependent on the chromosomal loci at which the proviral integration has occurred. We determined the integration sites of several proviral clones that showed different silencing stabilities and examined the possibility that the epigenetic instability of the MSCV proviral genome may in fact reflect the epigenomic regulation of the chromosomal regions around these sites.

Materials and Methods

ES Cell Culture and Retroviral Transduction

The EB5 and 7aabb ES cell lines were used in all experiments. ES cells were maintained in Glasgow minimum essential medium (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) supplemented with 10% fetal calf serum, 2 mM l-glutamine, 1 mM sodium pyruvate, 0.1 mM nonessential amino acids, 100 μM 2-mercaptoethanol, and 100 U/ml leukemia inhibitory factor (LIF). 5-aza-CdR was purchased from Calbiochem (San Diego, http://www.emdbiosciences.com) and dissolved in 50% acetic acid. ES cells were incubated with 150 nM 5-aza-CdR for 3 days and chased for 2 days prior to fluorescence-activated cell sorting (FACS) analysis. Media containing 5-aza-CdR were changed daily during treatment. FK228 (a gift of Astellas pharma Inc., Tokyo, http://www.astellas.com) was added to the culture medium at a concentration of 0.6 ng/ml. The pMCs vector was generated via a substitution of a 3′-long terminal repeat (LTR) region between the NheI and KpnI sites of the pMYs vector for the corresponding restriction fragment of MSCV [13]. The vesicular stomatitis virus G protein-pseudotyped retrovirus of pMCs-internal ribosomal entry site-enhanced green fluorescence protein (IRES-EGFP) (pMCs-IG) was produced with the PLAT packaging cell line as previously reported [14, 15]. For retroviral transduction, ES cells were incubated with retroviral supernatants for 8 hours in the presence of LIF and 5 μg/ml polybrene and cultured for an additional 18 hours after a threefold dilution with fresh ES cell culture medium. Cell sorting analysis and flow cytometry analysis were performed using FACSAria and FACSCalibur (Becton, Dickinson and Company, Franklin Lakes, NJ, http://www.bd.com), respectively. For green fluorescent protein (GFP) fluorescence analysis, dead cells were eliminated on the basis of their forward and side scatter and propidium iodide fluorescence.

RNA Interference Experiments

The pSilencer 3.1-H1 Puro vector (Ambion, Austin, TX, http://www.ambion.com) was used to express short-hairpin RNAs in all RNA interference experiments. The following target sequences for Dnmts were used: Dnmt1, 5′-GCAAAGAGTATGAGCCAATATTTGG-3′; Dnmt3a, 5′-GCACAACAGAGAAACCTAAGGTCAA-3′; Dnmt3b, 5′-ACCCAAGCGCCTCAAGACAAATAGC-3′. As a negative control, a random sequence that does not to target any murine mRNA was used: 5′-CGATTCGCTAGACCGGCTTCATTGC-3′ [16]. Transfections were performed using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, http://www.invitrogen.com) according to the manufacturer's instructions. After 24 hours, transfected cells were selected in puromycin for 2 days and subsequently cultured without this drug before harvesting. The following primary antibodies were used to validate the knockdown effects: anti-Dnmt1 antibodies (Abcam, Cambridge, MA, http://www.abcam.com), an anti-Dnmt3a monoclonal antibody (clone 64B1446; Imgenex, San Diego, http://www.imgenex.com), and an anti-actin monoclonal antibody (clone Ab-5; Transduction Laboratories, Lexington, KY, http://www.bdbiosciences.com/pharmingen). After reaction with primary antibodies, the blots were incubated with horseradish peroxidase-conjugated secondary antibodies (Chemicon, Temecula, CA, http://www.chemicon.com) and developed using an enhanced chemiluminescence plus Western blotting detection system (Amersham Biosciences, Piscataway, NJ, http://www.amersham.com).

Genomic Methylation Analysis

Southern blot hybridization was performed under standard conditions. The pMCs-IG proviral probe encompassing the gag region was generated by excision of the SpeI-BamHI restriction fragment from the pMCs vector. The cDNA probe encoding full-length GFP was generated by polymerase chain reaction (PCR) amplification. Hybridization signals were visualized and quantified using a FLA5100 fluoro image analyzer (Fujifilm, Tokyo, http://www.fujifilm.com). Bisulfite genomic sequencing was carried out as previously described [15]. The following primers were used to amplify the proviral region around the 5′-LTR: 5′-AGAATAGAGAAGTTTAGATTAAGGTTAGG-3′ and 5′-ATCRATAATCCCTAAACTAAAATCTCCA-3′. The following primers were used to amplify the genomic regions near proviral insertion sites 9 and 17: site 9, 5′-TTGTATTTGATGGGTTAGTTGGTTTGTGAG-3′ and 5′-TTCCTCTCTAAATCACCTTATTAACTCCAT-3′; site 17, 5′-AAGAATTTTTGGGTGATAGAGTTTTGGAGA-3′ and 5′-TCTACAATAAAATAAATCCTACAAAATACT-3′. Nearest-neighbor analysis was performed as previously described [17]. Methylated and unmethylated 2′-deoxycytidine 3′-monophosphates were quantified using the FLA5100 analyzer to calculate percentages of CpG methylation.

Genomic Mapping of Proviral Integration Sites

Genomic DNA extracted from MSCV-(pMCs-IG)-transduced ES cell clones was digested with BamHI and BglII and subjected to ligation at 16°C overnight. This facilitated the generation of self-ligated circular DNA carrying both a 3′-terminal proviral fragment and its neighboring genomic sequences. PCR using the following primer set was then performed to amplify DNA fragments encompassing vector-cellular DNA junctions: 5′-GCTGGAGTACAACTACAACAGCCACAACGTCTAT-3′ and 5′-GTCTTGTAGTTGCCGTCGTCCTTGAAGAAGATG-3′. Amplified fragments were subjected to sequencing analyses using the primer 5′-AGTCCTCCGATTGACTGAGTCG-3′.

Results

Expression Profiles of MSCV Proviral Clones in ES Cells

To evaluate MSCV silencing as an experimental model in the investigation of epigenetic dynamics in ES cells, we transduced EB5 cells with an MSCV-based retrovirus vector, pMCs-IG. This construct encodes green fluorescence protein (GFP) under the control of the MSCV-derived 5′-LTR promoter and glutamine tRNA primer-binding site and thus allows us to monitor the epigenetic status of the proviral genome using GFP fluorescence. In spite of inclusion of an IRES element upstream of GFP, this vector gave sufficient expression in ES cells to detect GFP-positive cells (Fig. 1B). For simplicity, pMCs-IG is referred to as MSCV hereafter. To prevent multicopy integrants from affecting our results, we transduced this vector at a low multiplicity of infection in all of these experiments. GFP+ single cells were isolated by FACS at 3 days after transduction and clonally propagated for several weeks to confer silencing. Twenty-four clones were then analyzed for GFP expression by flow cytometry at 21 days after transduction (the progressive decrease in the GFP+ fraction within the MSCV-transduced mixed population generally ceases within 2–3 weeks; data not shown).

Figure Figure 1..

Unstable silencing of murine stem cell virus (MSCV) proviral clones in ES cells. (A): GFP expression patterns for 24 representative integrants (nos. 1–24) at 21 and 31 days after transduction. Single GFP+ cells were isolated by sorting MSCV-transduced EB5 cells at 3 dpi and were then clonally propagated. Cells were then stained with propidium iodide, electrically gated for live cells, and analyzed for GFP expression. Cell numbers in the GFP+ fractions are shown as percentages. (B): Flow cytometry analyses of three representative clones displaying bimodal GFP expression. (C): GFP+ and GFP fractions were isolated from the three indicated clones by fluorescence-activated cell sorting, followed by the propagation and analysis of GFP expression. The data shown are percentages of the GFP+ cell fractions. Abbreviations: dpi, days postinfection; GFP, green fluorescent protein.

We observed that a large fraction of the clonal proviral integrants showed a bimodal variegated expression (Fig. 1A, 1B), as previously reported [12]. Only 1 clone (clone 4) out of 24 showed stable unimodal silencing (Fig. 3A) and contained a functional integrant (supplemental online Fig. 1A). The expression patterns of each clone at 21 days were largely maintained by 31 days after transduction (Fig. 1A). Since the partially silenced phenotypes of clones 9, 11, and 17 would help us investigate the epigenetic dynamics of the proviral genome (Fig. 1B), we scrutinized these three clones in additional experiments and confirmed by Southern blotting that these select clones harbored single-copy integrants (supplemental online Fig. 1B).

To examine the mechanisms underlying the variegated expression in the three selected clones, sustained GFP+ and silenced GFP cell fractions were isolated by flow sorting at 25 days after transduction and were subsequently analyzed for the maintenance of an expressed or silenced state. As shown in Figure 1C, we observed that the GFP+ sorted fractions of these three clones displayed progressive silencing to a similar extent, indicating that de novo silencing still occurs at later time points after transduction. We also observed that the simultaneously obtained GFP fractions from clones 11 and 17 exhibited spontaneous reactivation, albeit to various extents (Fig. 1C). Notably, one of our variegated clones (clone 9) never showed reactivation (Fig. 1C), which indicated that the reversibility of silencing does not always account for variegated expression. Moreover, all of the GFP+ and GFP single cell clones originating from clone 17 were found to show similar variegated expression patterns after clonal propagation (supplemental online Fig. 2A). This suggests that reversible silencing is common to a substantial population of the cells of this clone and thus is probably dependent on the specific genomic locus at which proviral integration had occurred.

Epigenetic Features of Variegated MSCV Clones

We next examined whether the instability in MSCV silencing in ES cells was in fact a reflection of the methylation dynamics of the proviral DNA. For this purpose, we performed Southern blotting analysis of both GFP+ and GFP sorted fractions of the most unstable clone, clone 17 (Fig. 2A–2D). Since the silencing status of clone 17 was found to be very unstable, the sorted cells were propagated for 3 days to obtain the minimum quantity of genomic DNA that is required for Southern analysis. During this propagation period, further silencing and reactivation occurred, resulting in a low purity for each fraction analyzed (Fig. 2A, 2D). Nevertheless, the relative proportions of unmethylated provirus observed at both the BssHII and HpaII sites were in good agreement with the observed percentages of the GFP+ cell populations in each sorted fraction (Fig. 2C, 2D). Hence, proviral methylation closely correlates with its silenced status even in an extremely reversible integrant. Although there are multiple HpaII sites in the proviral genome, partially digested DNA fragments corresponding to provirus that had been methylated in a mosaic fashion were not distinctly detected (Fig. 2C). This suggests that many, if not all, of the CpG sites in the provirus have the same methylation status, which may facilitate MSCV silencing in an all-or-none fashion.

Figure Figure 2..

Epigenetic features of MSCV silencing. (A): Cell preparation for Southern blot methylation analysis. GFP+ and GFP fractions of clone 17 were isolated by fluorescence-activated cell sorting (FACS) and propagated for 3 days prior to genomic DNA preparation. Each cell fraction was then analyzed by flow cytometry on days 0 and 3 postsorting. (B): Schematic diagram of the entire MSCV proviral DNA. HpaII (MspI), EcoRI, and BssHII sites are shown, together with the hybridization probe used for Southern blotting. Two HpaII sites flanking the probe and BssHII sites in the 5′-LTR were analyzed for methylation. If these sites are unmethylated, proviral DNA digested with HpaII or both BssHII and EcoRI will yield a 0.9- or 1.5-kb fragment, respectively. (C): Southern blot analysis of the GFP+ and GFP fractions of clone 17 shown in (A). Genomic DNA was digested with the indicated restriction enzymes and hybridized with the probe described in (B). The locations of the size markers are shown on the left. Arrowheads indicate the expected 0.9- and 1.5-kb proviral DNA fragments if they are unmethylated. Multiple extra signals were also detected in all lanes, which correspond to cross-hybridizations with endogenous C-type retroviruses. To distinguish MSCV proviral signals within this background, DNA from uninfected EB5 cells was analyzed in parallel (lanes 1, 2, and 7). The endogenous retroviral signals enabled us to roughly estimate that equivalent amounts of genomic DNA had been loaded in each lane. As a control for the complete digestion with HpaII, digestion with its methylation-insensitive isoschizomer, MspI, was also performed (lanes 2, 4, 6). (D): Quantitative comparisons of the results of the Southern blotting shown in (C) with the GFP expression patterns shown in (A). The relative intensities of the HpaII-sensitive 0.9-kb band and the BssHII-sensitive 1.5-kb band were quantified by FLA5100 image analyzer (Fujifilm, Tokyo), and these values indicate the levels of unmethylated provirus. The percentages of the GFP+ cell populations in the GFP+ and GFP sorted fractions are shown in parallel. (E): GFP+ and GFP fractions of clone 17 were obtained by FACS. Immediately after sorting, genomic DNA was extracted from the isolated GFP+ and GFP fractions and subjected to sodium bisulfite sequencing. Open and filled ovals indicate unmethylated and methylated CpG sites, respectively. A schematic diagram of the proviral 5′-LTR region is also shown at the bottom in relation to the sequenced region. (F): The reversible clones 11 and 17 were treated with 5azaCdR at 35 days postinfection and analyzed for GFP expression by flow cytometry at 40 days postinfection. A mixed cell population, following sorting of a GFP+ cell fraction by FACS at 3 days after transduction, was also treated with 5azaCdR at 28 days postinfection and is shown in parallel. Abbreviations: 5azaCdR, 5-aza-2′-deoxycytidine; GFP, green fluorescent protein; kb, kilobase(s); LTR, long terminal repeat; MSCV, murine stem cell virus; pMCs-IG, pMCs-IRES-EGFP; R, terminal direct repeat RNA; U3, unique regulatory sequences at the 3′ end; U5, unique regulatory sequences at the 5′ end.

To elaborate proviral methylation patterns in the active and repressive states, we also performed sodium bisulfite genomic sequencing analysis (Fig. 2E). Both GFP+ and GFP fractions were obtained by sorting clone 17 cells, and these were immediately analyzed for their methylation status around the 5′-LTR region. Whereas the active provirus in the GFP+ fractions was found to be mostly unmethylated, the proviral genomes in the GFP sorted fractions were significantly methylated. A few proviral genomes in the GFP fractions nevertheless showed mosaic patterns of partial methylation, which could have originated from the cells expressing GFP below detectable levels immediately after reactivation had occurred.

To clarify the functional relationship between DNA methylation and the unstable silencing of the MSCV provirus, we tested the effects of 5-aza-CdR on the reversible silencing status of clones 11 and 17 together with a mixed cell population that was propagated from a GFP+ fraction obtained by sorting at 3 days after transduction. We found that 5-aza-CdR significantly augments the GFP+ fractions of both clones, presumably by reactivating the transiently silenced proviruses (Fig. 2F).

Since histone H3 hypoacetylation within the proviral genome is known to be quite closely associated with MSCV silencing [11], we next examined whether a histone deacetylase (HDAC) inhibitor, FK228, would reactivate the transient silencing status of reversible clone 17. We found that FK228 treatment indeed induced the reactivation of clone 17 but not of unimodally silenced clone 4 (Fig. 3A). Hence, the interplay between histone modification and DNA methylation is likely to play a role in maintaining the unstable silencing phenomenon that we observed for MSCV in ES cells.

Figure Figure 3..

The histone deacetylase inhibitor induces global and retroviral DNA demethylation in embryonic stem cells. (A): Clones 4 and 17 were treated with 0.6 ng/ml FK228 for 24 hours and subjected to flow cytometry analysis. (B): Upper panel: Schematic diagram showing the AatII and NheI sites in the murine stem cell virus provirus in relation to a cDNA probe that encodes GFP, together with the expected fragment sizes from this proviral DNA when digested with both of these enzymes. Depending on the methylation status of the AatII site, proviral DNA will yield a 2.6-kb fragment if unmethylated, or a 3.3-kb fragment if methylated. Middle panels: Clones 9 and 17 were treated with 0.6 ng/ml FK228. Twenty-four hours after this treatment, genomic DNA was extracted and subjected to Southern blot analyses with the GFP probe. Arrowheads indicate AatII-sensitive 2.6-kb and AatII-insensitive 3.3-kb fragments. Lower panels: To confirm that there were similar effects of FK228 on the global histone acetylation levels of clones 4 and 17, whole cell lysates were prepared in both cases at 24 hours after treatment with this inhibitor and were subjected to immunoblotting for histone H3 protein expression with anti-histone H3 rabbit polyclonal antibodies. Immunoblotting of these lysates was also performed for H3 acetylation with an anti-acetyl-histone H3 rabbit polyclonal antibody. (C): EB5 and 7aabb cells were treated with 0.6 ng/ml FK228 for 2 days and subjected to nearest-neighbor analysis. Also shown are the percentages of CpG methylation, calculated as mdCp/(mdCp + dCp), where mdCp and dCp indicate methylated and unmethylated 2′-deoxycytidine 3′-monophosphates, respectively. Abbreviations: AcH3, acetylated histone H3; FK, FK228; GFP, green fluorescent protein; kb, kilobase(s); Me, methylation; pMCs-IG, pMCs-IRES-EGFP.

HDAC inhibitors had been previously shown to induce global and gene-specific DNA demethylation [18, 19]. We found also in our current experiments that the FK228 treatment of ES cells significantly decreases their global methylation (Fig. 3C). A previous report had proposed that HDAC inhibitors may cause DNA demethylation by downregulating Dnmt3b [20]. However, our current data suggest that FK228 can also induce global demethylation in 7aabb ES cells that are homozygous for both Dnmt3a and Dnmt3b mutations (Fig. 3C). Thus, HDAC inhibitors may affect the global methylation of the ES cell genome by alternative mechanisms such as active demethylation [21]. We therefore examined the possibility that FK228 would induce alterations in the MSCV proviral methylation status, which would then be followed by proviral reactivation. Indeed, we found that FK228 rapidly induces the proviral demethylation of reversible clone 17 but not of clone 4 (Fig. 3B), which is in good agreement with its effects on the maintenance of silencing shown in Figure 3A. Since DNA methylation is always coupled tightly with unstable silencing, as demonstrated in our current experiments, the demethylation of proviral DNA is likely to occur upon spontaneous reactivation of the reversible integrants.

Vulnerability of Methylation Maintenance in an Epigenetically Unstable Provirus

To gain mechanistic insights into the epigenetic instability of MSCV proviral genomes, we determined the regulatory role of the DNA methyltransferases in maintaining the stable and unstable silencing of variegated clones 9 and 17. For this purpose, we used transient knockdown experiments using short-hairpin RNAs that are specific for the Dnmts. We found from these analyses that transient Dnmt1 depletion augments the GFP+ fractions of both clones, albeit to different extents (Fig. 4A), which demonstrated the feasibility of this approach. Notably, this shDnmt1-induced augmentation was retained even after Dnmt1 expression was restored (Fig. 4A, 4B). This prolonged effect suggests that Dnmt1 depletion alters the epigenetic status of proviral DNA, most likely via DNA methylation. On the other hand, single depletions of either Dnmt3a or Dnmt3b were found to slightly augment the GFP+ fractions of clone 17, but only at later time points (Fig. 4A). The combined depletion of both these enzymes showed more apparent effects compared with individual knockdowns. Thus, both Dnmt3a and Dnmt3b may play a role in maintaining reversible silencing, presumably by repairing proviral methylation maintenance errors. In contrast, neither a single nor combined loss of Dnmt3a and Dnmt3b activity was found to affect the robust silencing of clone 9. The probable differences in the fidelity of maintaining methylation patterns between clones 9 and 17 may well account for their differing susceptibilities to the transient depletion of the Dnmt3 family.

Figure Figure 4..

Effects of transient Dnmt depletion on the stable and unstable silencing phenotypes of murine stem cell virus (MSCV)-infected ES cells. (A): Clones 9 and 17 were transiently transfected with shRNA expression plasmids targeting Dnmt1 (shDnmt1), Dnmt3a (shDnmt3a), or Dnmt3b (shDnmt3b) mRNA. For the combined depletion of Dnmt3a and Dnmt3b, the cells were cotransfected with the shRNA vectors shDnmt3a and shDnmt3b. As an NC, an shRNA that had been designed not to target any specific murine mRNAs (shNC) was used. At 1 day post-transfection, the cells were selected in 1 μg/ml puromycin for 2 days, chased, and then harvested for flow cytometry analysis at the indicated time points. RNAi-induced derepression was calculated as 100 × (D − N) / (100 − N), where D and N are percentages of GFP+ fractions in the shDnmt- and shNC-transfected cells, respectively. The averages of three independent experiments are shown, with error bars indicating SDs. The data obtained for clone 9 are indicated by the dashed lines. (B): Re-expression kinetics of the endogenous Dnmts in transient knockdown experiments. EB5 cells were transfected and selected in puromycin as described in (A) and subjected to immunoblotting for Dnmt1, Dnmt3a2, or Dnmt3b1 protein expression. Dnmt3b1 was detected by an anti-Dnmt3a antibody (clone 64B1446) as previously reported [32]. As a control, β-actin protein expression was analyzed in parallel. (C): The AatII site methylation of the variegated 9 and 17 proviruses following Dnmt1 knockdown was assessed by Southern blot analysis as described in Figure 3B. Arrowheads indicate M and U fragments of clones 9 (lanes 2 and 3) and 17 (lanes 5 and 6), transfected with NC (lanes 2 and 5) or Dnmt1 (lanes 3 and 6) shRNA expression plasmids. To identify MSCV proviral signals, DNA from uninfected EB5 cells was also analyzed as a control (lanes 1 and 4). (D): Sodium bisulfite proviral methylation analysis of clones 9 and 17 at 4 days after shRNA administration. (E): Nearest-neighbor analysis of the clone 9 and 17 genomes at 4 days after shRNA administration. The percentage of CpG methylation was calculated as mdCp/(mdCp + dCp). Abbreviations: dAp, 2′-deoxyadenosine 3′-monophosphate; dCp, 2′-deoxycytidine 3′-monophosphate; dGp, 2′-deoxyguanosine 3′-monophosphate; dTp, 2′-deoxythymidine 3′-monophosphate; M, methylated; Me, methylation; NC, negative control; RNAi, RNA interference; sh, short-hairpin; U, unmethylated.

To validate the effects of Dnmt1 depletion on the MSCV proviral DNA methylation status, an AatII site in the gag gene within this proviral genome was analyzed for methylation by Southern blotting at 4 days after short-hairpin RNA administration (Fig. 4C). The Dnmt1 knockdown induced a stronger level of demethylation of clone 17 compared with clone 9 (Fig. 4C), even though it induced genome-wide demethylation of both clones to similar extents (Fig. 4E). Furthermore, we also determined the precise CpG methylation patterns of both clones following the depletion of Dnmt1 (Fig. 4D). The Dnmt1 knockdown resulted in a significant increase of fully or severely unmethylated proviral genome fractions in clone 17. This observation may be attributable to the possibility that Dnmt1 inactivation generates large successive methylation gaps as it acts in a processive manner [22, 23]. De novo silencing processes should be required for the rerepression of such severely demethylated provirus, which might necessarily require long periods of culturing, as shown in Figure 4A.

Relationships Between the Epigenomic Environment and the Epigenetic Instability of the MSCV Provirus Genome

Our early data strongly suggested that the silencing stability of the MSCV provirus is dependent upon its chromosomal location. We thus speculated that if the epigenetic features of proviral DNA could be affected by the epigenomic environment around the integrated site, the instability in proviral silencing may in fact be a direct reflection of the epigenetic regulation of the adjoining genome. Hence, we determined the chromosomal locations of the characterized proviral clones (Fig. 5A). Intriguingly, the extremely unstable clone 17 was mapped into about the middle of a large intergenic region. In contrast, the proviral integrants of the stably silenced clones 4 and 9 were found to be embedded in genes. The site of integration of weakly reversible clone 11 was found in a putative gene that has been predicted to encode a hypothetical protein. These observations led us to speculate that stable silencing might be associated with regions harboring functional genes.

Figure Figure 5..

Genomic mapping and epigenetic characterization of murine stem cell virus proviral integration sites in ES cells. (A): The chromosomal positions of the clonal integrants 4, 9, 11, and 17 were determined as described in Materials and Methods. The positions and directions of each provirus are shown as arrows in relation to the neighboring genomic organizations indicated in the National Center for Biotechnology Information database. Filled and empty boxes indicate coding and noncoding exons of the neighboring genes, respectively. The transcriptional orientation is also depicted by arrows. The two lollipops indicate the genomic regions that were subjected to bisulfite sequencing shown in (B). (B): Parental EB5 cells were transfected with NC or Dnmt1 short-hairpin RNA expression plasmids. The cells were selected in puromycin as described in Figure 4A and subjected to bisulfite methylation analysis at 4 days after transfection. The chromosomal locations of the analyzed regions are shown in (A). Abbreviations: Chr., chromosome; kb, kilobases; NC, negative control; sh, short-hairpin.

The genomic loci for stable and unstable integrants 9 and 17, respectively, were further examined for their methylation stability in parental EB5 cells upon Dnmt1 knockdown (Fig. 5B). The analyzed genomic regions are detailed in supplemental online Figure 3. We found that the genomic region at which clone 17 had been mapped was significantly demethylated by shDnmt1 treatment, which was in clear contrast to clone 9. These observations led us to propose that the epigenetic features of proviral DNA, at least in part, mimic those of the neighboring genomic sequences. Thus, the frequently observed reversibility of MSCV silencing may be suggestive of epigenetic instability in unexpectedly widespread regions of the stem cell genome.

Discussion

In our current study, we used the MSCV retroviral vector to further elucidate the epigenetic dynamics of the stem cell genome. The epigenetic instability of the MSCV provirus shows significant clonal variability that is dependent upon the chromosomal integration sites. DNA methylation was found to be tightly associated with a silenced state, even for extremely unstable proviral clones. Furthermore, we show that both the unstable proviral DNA and its adjacent genomic sequences together display significant instability upon Dnmt1 knockdown in terms of maintenance methylation, which suggests that the epigenetic features of the proviral genome reflect the epigenomic environment around its site of integration. Thus, our study may provide a useful model to explore where and how epigenetic drifts can occur globally within the ES genome.

Our current data provide significant evidence to suggest that the transient and reversible silencing of the MSCV retroviral genome is always associated with proviral DNA methylation. Hence, the simplest explanation for the instability of MSCV silencing is that the methylation of the proviral genome is also transient and reversible. However, this would be somewhat at odds with the commonly held view that DNA methylation is quite stably inherited even after multiple rounds of genomic replication [24]. An alternative explanation may thus be that histone modifications and not DNA methylation are responsible for this reversible silencing effect, since histone H3 hypoacetylation has been reported to be tightly associated with MSCV silencing [11]. Nevertheless, an accumulating body of evidence now suggests that the methylation patterns of the ES cell genome are sometimes poorly maintained.

The genetic inactivation of both the Dnmt3a and Dnmt3b genes has been shown to lead to a progressive loss of genome-wide methylation in mouse ES cells, which suggests that Dnmt3a and Dnmt3b compensate for inaccurate maintenance methylation by Dnmt1 [25, 26]. It also has been reported that many female ES cell lines undergo a loss of global DNA methylation shortly after their derivation due to the reduced expression of Dnmt3a and Dnmt3b [27]. Notably, methylation changes have also been frequently observed even in normal male ES cells under conventional culture conditions [7, 8]. Although epigenetic alterations in ES cells could depend on their genetic background or the culture conditions, our findings may suggest that substantial loci within the ES genome are epigenetically unstable and have frequent occurrences of maintenance methylation errors. These findings thus raise substantial concerns about handling these cells in both research and clinical settings.

From a scientific point of view, this epigenetic instability seems likely to be a feature of cultured ES cells. One might therefore speculate that the epigenetic drift maintains ES cells in a pluripotent state. Mammalian development is accomplished by two major waves of genome-wide demethylation [28]. The germ-cell genome undergoes epigenetic reprogramming by erasure of the somatic methylation patterns through active demethylation mechanisms [29]. Hence, differentiation potential of stem cells could be based on their epigenetic plasticity. The possible physiological relevance of the epigenetic instability in ES cells may be thus an attractive topic for future investigations. In addition, of further interest is the possibility that the epigenetic fluctuation would affect the lineage commitments upon ES cell differentiation. This possibility will be applicable to technical improvement to efficiently and selectively generate differentiated cells from ES cell culture.

From a practical viewpoint, a loss of imprinting by aberrant DNA methylation can arise in cultured ES cells and not be corrected upon differentiation, which might lead to aberrant gene expression. Since a transient loss of global methylation can induce widespread cancer formation even after gross remethylation [30], it might therefore be necessary to survey the epigenetic abnormalities in ES cells prior to their clinical usage. However, since it is extremely difficult to define the extent to which regions other than the imprinted loci are affected in terms of their DNA methylation pattern, it will become important to control the epigenetic instability by in vitro manipulation. In this regard, we found that culturing our reversible clones under low-serum conditions caused a reduction in the size of their GFP+ fractions (supplemental online Fig. 2B), which suggests that there may be a way to suppress epigenetic drift via the maintenance of an undifferentiated state. For example, reactivation could be controlled by specific culture environments or some drug treatments. To this end, our unstable MSCV clones could be a useful tool for the screening of these conditions. Future studies may eventually allow us to develop quality control procedures for human ES and induced pluripotent stem cells.

The mechanisms underlying the demethylation of the reversible MSCV integrant described in this report remain to be elucidated. However, several lines of evidence suggest that the proviral demethylation that occurs upon the spontaneous reactivation of this virus is mediated by a passive demethylation mechanism rather than an active one. This is partly because cell division seems to be coupled with this reactivation. Serum reduction was found to suppress the reactivation of our reversible clones, possibly via growth retardation (supplemental online Fig. 2B). In addition, we found that the G2/M cell cycle population was slightly augmented in the GFP-positive fractions of reversible clone 17 but not in those of irreversible clone 9 (supplemental online Fig. 2C). Hence, we surmise that the replication of the genome is required for the loss of methylation. Furthermore, our present data suggest that the epigenetic instability of proviral genomes is widespread within the stem-cell genome. Since such genome-wide instability via active demethylation has not been previously observed, passive demethylation is most likely to induce spontaneous reactivation of the reversible integrants. It is noteworthy in this regard that FK228 treatments induced the demethylation of reversible integrants. Previous reports have demonstrated that HDAC inhibitors can induce active demethylation under some conditions in human cancer cell lines [21, 31]. The possible interaction of epigenetic instability with an active demethylation pathway thus needs to be investigated in future studies.

To elaborate epigenetic dynamics in stem cells, we have intensively characterized three clones (clones 9, 11, and 17) that show typical variegated expression patterns. We analyzed 24 MSCV clones for their GFP expression, of which we obtained only one clone (clone 23) showing stable expression (Fig. 1A). It is also important to understand the epigenomic environments that prevent this provirus from being de novo methylated, because MSCV may become a promising expression vector in gene therapy experiments using human ES and iPS cells harboring genetic disease backgrounds. The strong dependence of proviral expression patterns on their genomic loci underscores the need for using such minor clones to achieve stable transgene expression in these cells. Alternatively, it might be possible to enrich cellular fractions with stable expression in mixed proviral population by specific culturing methods such as cell cycle synchronization. Further investigation from this perspective may thus be worthwhile for the future practical use of retroviral-mediated gene transfer in regenerative medicine.

Summary

In conclusion, MSCV retroviral DNA methylation is poorly maintained in embryonic stem cells. The epigenome environments around the proviral integration sites may be the determinant of this epigenetic vulnerability of the retroviral genome. These findings will aid in our understanding of epigenetic regulation in ES cell genome, and facilitate our ability to sup-press epigenetic drift of the pluripotent stem cells for their clinical usage.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.

Acknowledgements

We thank Hitoshi Niwa and Masaki Okano for providing the ES cell lines and Toshio Kitamura for providing the retroviral vectors and packaging cell lines. This work was supported by Grants-in-Aid from the Japan Society for the Promotion of Science and the Ministry of Education, Culture, Sports, Science and Technology of Japan.

Ancillary