Embryonic stem cells (ESCs) hold promise for cell and tissue replacement approaches to treating human diseases based on their capacity to differentiate into a wide variety of somatic cells and tissues. However, long-term in vitro culture and manipulations of ESCs may adversely affect their epigenetic integrity, including imprinting. We have recently reported aberrant biallelic expression of IGF2 and H19 in several rhesus monkey ESC lines, whereas SNRPN and NDN were normally imprinted and expressed predominantly from the paternal allele. The dysregulation of IGF2 and H19 that is associated with tumorigenesis in humans may result from improper maintenance of allele-specific methylation patterns at an imprinting center (IC) upstream of H19. To test this possibility, we performed methylation analysis of several monkey ESC lines by genomic bisulfite sequencing. We investigated methylation profiles of CpG islands within the IGF2/H19 IC harboring the CTCF-6 binding site. In addition, the methylation status of the IC within the promoter/exon 1 of SNURF/SNRPN known as the Prader-Willi syndrome IC was examined. Our results demonstrate abnormal hypermethylation within the IGF2/H19 IC in all analyzed ESC lines, whereas the SNURF/SNRPN IC was differentially methylated, consistent with monoallelic expression.
ESCs are derived from the inner cell mass of blastocysts and can be propagated indefinitely under certain in vitro conditions while maintaining pluripotency. Upon induced or spontaneous differentiation, ESCs are capable of differentiating into a wide variety of cells and tissues constituting a whole organism, thus offering an unlimited source of cells for transplantation medicine. However, long-term in vitro culture and manipulation of ESCs may adversely affect their genetic and epigenetic integrity, including karyotype , methylation profiles, and imprinted gene expression [2, 3], raising potentially serious safety concerns associated with their use in cell-based therapy. Moreover, an increased incidence of epigenetic abnormalities has been associated with embryos produced in vitro by various assisted reproductive technologies [4, 5]. Because ESCs are routinely derived from in vitro-produced blastocysts, it is likely that any epigenetic errors extant in embryos will be maintained in the resultant ESCs. Rhesus monkey ESCs share a number of properties with human ESCs, and their derivation and use is not affected by bioethical concerns, thus making them an ideal model for basic research, as well as for preclinical transplantation studies .
Genomic imprinting involves epigenetic modification of a gene or a chromosomal region that results in the absolute or preferential monoallelic expression of a specific parental allele. Imprinted genes tend to cluster in the genome; two such clusters implicated in human disease and intensively studied are located on chromosome 15q11–q13, known as the Prader-Willi syndrome (PWS)/Angelman syndrome (AS) region , and 11p15.5 , known as the Beckwith-Wiedemann syndrome (BWS) region. Imprinting in these regions is controlled in cis by so-called imprinting centers that regulate parent-specific expression of target genes bidirectionally over long distances. Mechanisms involved in the control of imprinted gene expression are complex and poorly understood (reviewed in ). Epigenetic modifications must be reprogrammed during development, involving first erasure of old marks during germ cell development and establishment of new marks in a parent-of-origin-specific manner. Methylation of CpG dinucleotides within imprinting centers (ICs) is likely one of the initial mechanisms differentially marking parental chromosomes in gametes. Once established, locus-specific DNA methylation profiles must be stably maintained in future generations of cells.
The BWS region is organized into two imprinted domains controlled by separate ICs. The adjacent IGF2 and H19 genes are located in one of these domains, and their imprinted expression is reciprocally controlled by a common IC located upstream of H19 harboring a CpG island that is methylated on the paternal chromosome and unmethylated on the maternal chromosome in mice and humans. Imprinting in this domain, according to the insulator model, involves the 11-zinc-finger protein CTCF [10, –12]. Loss of parent-specific methylation at this site correlates with abnormal expression of IGF2 and H19 and has been linked to patients with BWS , Wilms tumors , bladder and colon cancers [14, –16], and osteosarcoma . The human IC contains seven potential CTCF-binding sites , with the sixth site (CTCF-6) demonstrating allele-specific methylation correlated with imprinted IGF2/H19 expression .
An imprinting cluster on human chromosome 15q11–q13 contains several paternally expressed genes, including SNRPN and NDN, associated with the PWS phenotype and a single maternally expressed gene implicated in AS . Genetic evidence based on microdeletions and balanced translocations in PWS patients has mapped the promoter/exon 1 locus of SNURF/SNRPN gene as an IC [18, –20]. This locus appears to be necessary for the erasure, establishment, and maintenance of the paternal imprints in the PWS/AS region. Mouse and human somatic cells show allele-specific CpG methylation of the PWS IC, with complete methylation of the maternal chromosome and no methylation of the paternal chromosome [21, 22].
We have recently demonstrated aberrant biallelic expression of IGF2 and H19 in rhesus monkey ESC lines (ORMES series ), whereas SNRPN and NDN were normally imprinted and expressed from the paternal allele . In contrast, expanded blastocyst-stage embryos, from which these ESCs were derived, exhibited normal paternal expression of IGF2 and maternal expression of H19, suggesting that imprinting marks for these genes were already established at this stage of development. Thus, it is likely that abnormal biallelic expression of IGF2 and H19 in ESCs was acquired during initial establishment or culture. This pattern of expression was retained in differentiated pancreatic and neuronal lineages derived from these ESCs, indicating irreversible loss of imprinting at this locus. This dysregulation of IGF2 and H19 expression is most likely a result of improper maintenance of allele-specific methylation patterns in the corresponding IC. To address this hypothesis, we performed a comprehensive methylation analysis by investigating the region corresponding to the human IC harboring the CTCF-6 binding site. In addition, we examined the methylation status of the PWS-IC within promoter/exon 1 of SNURF/SNRPN. Our results demonstrate an abnormal methylation pattern within the IGF2/H19 imprinting domain in all analyzed ESC lines, whereas the SNURF/SNRPN IC was differentially methylated.
Materials and Methods
Rhesus monkey ESC culture has been described previously [6, 23]. Briefly, ESCs were grown on feeder layers of mitotically inactivated mouse embryonic fibroblasts in medium consisting of Dulbecco's modified Eagle's medium/Ham's F-12 medium (Invitrogen, Carlsbad, CA,http://www.invitrogen.com) supplemented with 15% fetal bovine serum (FBS) (HyClone, Logan, UT, http://www.hyclone.com), 1 mM l-glutamine (Invitrogen), 0.1 mM β-mercaptoethanol, and 1% nonessential amino acids (Invitrogen). Cultures were maintained at 37°C, 5% CO2 and 95% of balance air. Medium was changed daily, and ESC colonies were split every 5–7 days by manual disaggregation, with collected cells replated on fresh feeder layers.
Polymerase Chain Reaction-Based Amplification and Allele-Specific Expression Analysis of Imprinted Genes
Total RNA from ESCs was isolated as we described previously  and treated with DNase I before cDNA preparation using SuperScript III First-Strand Synthesis System for reverse transcription-polymerase chain reaction (RT-PCR) (Invitrogen) according to the manufacturer's instructions. Characteristics of the single-nucleotide polymorphisms (SNPs) used for allele-specific expression analysis, PCR primers, and conditions were described in detail previously [23, 24]. Expressed alleles were determined by direct sequence analysis of the RT-PCR amplicons and/or by restriction fragment length polymorphism . In all RT-PCRs, samples were also analyzed without RT to exclude the possibility of genomic DNA contamination.
Bisulfite Modification and Sequencing of gDNA
Genomic DNA was extracted from isolated ESC colonies, sperm, or muscle biopsy samples using QIAamp DNA Micro Kit or DNeasy Tissue Kit (Qiagen, Valencia, CA, http://www1.qiagen.com) according to the manufacturer's instructions. Approximately 2 μg of gDNA was modified by bisulfite treatment using a CpG Genome Modification Kit (Chemicon, Temecula, CA, http://www.chemicon.com) according to the manufacturer's protocol. Seminested primers for bisulfite sequencing were designed using the online software MethPrimer (http://www.urogene.org/methprimer) as described elsewhere . The sequence, annealing temperature, and cycle number of each primer pair are listed in Table 1.
Table Table 1.. Seminested polymerase chain reaction primers used for bisulfite sequencing
PCRs were carried out in a 20-μl volume containing 2.5 mM MgCl2, 2 mM dNTP mixture, 0.4 μM each primer, 40 ng of template DNA, and 1 U of AccuSure DNA polymerase (Bioline, Randolph, MA, http://www.bioline.com). Amplicons were electrophoresed through 1.6% Tris borate-EDTA agarose gels stained with ethidium bromide and visualized on a UV transilluminator. PCR products were recovered from stained gels (QIAquick Gel Extraction Kit; Qiagen), ligated with plasmid vector (TOPO TA Cloning Kit for Sequencing; Invitrogen), and cloned according to the manufacturer's protocol. Individual bacterial colonies were transferred to LB/Amp medium and cultured overnight with shaking. Cultures were then processed with Qiaprep Spin Mini-prep Kit (Qiagen) according to the manufacturer's protocol, resulting in a single cloned PCR species per plasmid. Individual clones were then sequenced with an ABI 3100 capillary genetic analyzer (Applied BioSystems, Foster City, CA, http://www.appliedbiosystems.com) using BigDye terminator sequencing chemistry . Sequencing results were analyzed using Sequencher software (Gene Codes Corp., Ann Arbor, MI, http://www.genecodes.com).
Methylation Analysis by Southern Blot
For Southern analyses, approximately, 4 μg of gDNA from muscle control and ORMES cell lines was digested with EcoNI and the CpG methylation-blocked enzyme BsaHI. (New England Biolabs, Ipswich, MA, http://www.neb.com). The samples were electrophoresed through a 1% agarose gel and transferred to a nylon membrane. The blot was then hybridized with a probe, whose template was generated by PCR of gDNA from the control tissue, using the following primers: forward, 5′–3′, AGTGCAGGCTCACACATCATAGTC; reverse: 5′–3′, TAGTCTCTGAGCAAGTAGCGCATC. The probe was produced by random priming (Megaprime DNA Labeling Systems; Amersham Biosciences, Piscataway, NJ, http://www.amersham.com) and labeled with 32P-dCTP (PerkinElmer Life and Analytical Sciences, Boston, http://www.perkinelmer.com). The hybridization was carried out at 65°C overnight in Rapid-hyb buffer (Amersham Biosciences). The blot was washed for 30 minutes in 2× SSC/0.1% SDS and then for 20 minutes in 0.1× SSC/0.1% SDS.
Previously, we reported DNA sequence information for the rhesus monkey corresponding to both the human IGF2/H19 IC spanning the CTCF-6 binding site and the SNURF/SNRPN IC . We also demonstrated that these regions were differentially methylated in muscle tissue of juvenile monkeys, in agreement with the monoallelic expression of four imprinted genes: IGF2, H19, SNURPN, and NDN. Our initial evaluations of the methylation status of these domains in control muscle tissue by bisulfite sequencing resulted in the recovery of predominantly methylated clones at an 8:2 ratio, suggesting significant bias, possibly during PCR amplification or cloning steps of bisulfite-treated DNA. Attempts to design several sets of different primers did not improve the outcome. Here, the methylation status of these regions in sperm, control muscle tissue, and monkey ESCs was further analyzed by bisulfite genomic sequencing, with subsequent identification of parental alleles based on existence of sequence polymorphisms. To determine the methylation status of parental alleles, we used four previously defined SNPs within the IGF2/H19 IC and five SNPs in the SNURF/SNRPN IC . We genotyped 17 rhesus monkey ESC lines (ORMES-1 through -17), as well as the parental combinations involved in embryo production and, subsequently, ESC line derivation. The genotypes were determined by PCR amplification of gDNA followed by direct sequencing, as described previously . Informative, heterozygous cell lines were selected for further study; seven were heterozygous for at least one IGF2/H19 IC SNP site, and six were heterozygous for a SNURF/SNRPN IC SNP site (Table 2). Although SNPs for the IGF2/H19 IC at positions 668 (A/G) and 712 (C/A) were useful for allele-specific methylation analysis, the C/T polymorphism at position 494 was not helpful for bisulfite sequencing since unmethylated cytosines are converted to uracils that are detected as thymidines after sequencing, thereby masking the original C alleles. On the other hand, the C/T polymorphic site at position 485 was located within a CpG site; therefore, methylated alleles would remain unchanged, and thus, this SNP was semi-informative in determining methylated parental clones. Interestingly, the ORMES-1 cell line was most informative based on polymorphisms on three SNP sites for IGF2/H19 and two of five SNPs for SNURF/SNRPN ICs.
Table Table 2.. Presence of SNPs in existing monkey ESC lines (ORMES series ) within H19/IGF2 and SNURF/SNRPN ICs
Initially, we confirmed our previously published results and detected abnormal biallelic H19 and IGF2 expression in informative ORMES cell lines, whereas both NDN and SNRPN were expressed from paternal alleles only  (results not shown). We also verified normal maternal H19 expression in ORMES-5 exclusively, whereas IGF2 was biallelic in this cell line. To correlate methylation with expression pattern of these imprinted genes, further methylation studies were conducted on ORMES gDNA isolated from the same passage numbers.
Bisulfite sequencing of IGF2/H19 IC revealed the methylation status of 24–27 individual CpG sites within this region, including five CpG sites within the putative CTCF-6 binding motif (Fig. 1). Both alleles in the sperm sample recovered from the heterozygous male were heavily methylated, suggesting that paternal methylation marks for the IGF2/H19 IC are already established in mature primate gametes. Control skeletal muscle tissue from a juvenile monkey demonstrated the expected methylation of paternal clones, particularly within the CTCF binding site, whereas the four maternal alleles analyzed were completely unmethylated. In contrast, whereas the majority of paternal clones in ESCs were completely methylated, only approximately 50% of the analyzed maternal clones in ORMES-1 and ORMES-3 ESC lines were unmethylated. In the case of ORMES-3, hypomethylation of a few paternal clones was also observed. The ORMES-5 ESC line displayed a methylation pattern closest to the control, with only two of seven maternal clones predominately methylated. There were no unmethylated clones detected in ORMES-6. ORMES-7, -8, and -10 were not heterozygous, precluding separation of parental alleles. However, up to 30 randomly chosen clones were analyzed for these cell lines, and either most or all of the sequenced clones were methylated.
To further validate these bisulfite sequencing data, we performed independent methylation analysis by methylation-sensitive Southern analysis. Southern blot analysis was carried out on the same samples of gDNA isolated from skeletal muscle and ORMES cell lines. The presence of both cut and uncut fragments by the methylation-sensitive BsaHI restriction enzyme was observed in control muscle tissue, suggesting the presence of both methylated and unmethylated alleles (Fig. 2). Similar results were also produced with ORMES-1 and ORMES-5, indicating the same ratio of methylated and unmethylated alleles within the H19/IGF2 IC in each sample. In ORMES-3 and -7, the level of methylation was considerably higher than that of the control. In contrast, no unmethylated alleles were observed in ORMES-6, -8, and -10 (Fig. 2), suggesting complete methylation of both alleles, in agreement with bisulfite sequencing analysis.
Bisulfite sequencing of heterozygous sperm, skeletal muscle tissue, and four ORMES cell lines within the SNURF/SNRPN IC allowed analysis of 25 individual CpG sites (Fig. 3). Both alleles in the sperm DNA recovered from the heterozygous male were completely unmethylated. This finding suggests that methylation marks for the primate SNURF/SNRPN IC have been reset in a parent-of-origin-specific manner in mature male gametes. As expected, in control muscle tissue, paternal clones were unmethylated, and the majority of maternal alleles were predominately methylated. However, several maternal clones showed a distinctive pattern of methylation at some CpG sites (34, 12, 15, 16, 20–22, 24), whereas others were unmethylated. In informative ORMES-1 and -6, maternal clones were heavily methylated, whereas paternal clones were unmethylated, with sporadic methylation particularly at CpG sites 15, 16, 21, and 22. Although ORMES-3 and -5 were not heterozygous and parental clones could not be identified, both methylated and unmethylated clones were present. In the case of ORMES-3, approximately half of the clones were methylated and the other half unmethylated. In ORMES-5, approximately one-third of the analyzed clones were unmethylated. Unmethylated clones in these cell lines also showed periodic methylation at a few specific CpG sites similar to that seen in ORMES-1 and -6.
The presence of SNPs in the monkey provided an excellent opportunity to study parent-specific methylation status within the postulated ICs. Using this polymorphism and highly sensitive genomic bisulfite sequencing analysis, we demonstrated aberrant methylation profiles within the IGF2/H19 IC, potentially accounting for the observed biallelic expression of both IGF2 and H19 in rhesus monkey ESCs . Moreover, independent methylation-sensitive Southern analysis corroborated this conclusion. In contrast, the SNURF/SNRPN IC was differentially methylated, as expected.
According to the chromatin insulation model, if both alleles of the IGF2/H19 IC were unmethylated, H19 would be biallelically expressed, with silenced IGF2. In contrast, if both ICs were methylated, H19 would be silenced, with IGF2 biallelically expressed. Although this pattern of reciprocal regulation of IGF2 and H19 expression has been demonstrated during normal human development  and in patients with Wilms tumors , biallelic expression of one gene and independent normal monoallelic expression of the other has been seen [17, 29]. Hypothetically, hypermethylation of the maternal allele detected in the present study may explain the biallelic expression of IGF2; however, H19 showed loss of imprinting as well, rather than silencing. Interestingly, normal biallelic expression of IGF2 was detected in some parts of brain; however, this biallelic expression occurred despite normal patterns of CTCF binding . The existence of brain-specific enhancers for IGF2 upstream of the chromatin boundary was proposed as a potential explanation for this phenomenon. Another suggestion was that CTCF binding is not by itself sufficient to establish the boundary, and other downstream mechanisms may be involved. Indeed, CTCF has been shown to undergo a post-translational modification: poly(ADP-ribosyl)ation resulting in a significant change in the three-dimensional and electrostatic properties of the protein . Poly(ADP-ribosyl)ation of the CTCF combined with binding to the IC was associated with function of CTCF as a transcriptional regulator of Igf2 expression .
Another possible explanation relates to heterogeneity in the cell populations tested. At least four phenotypes may exist within the sampled cell line: normal differentially methylated, maternally hypermethylated, hypomethylated maternal and paternal alleles, and cells with abnormally methylated maternal and unmethylated paternal alleles. The corresponding expression situations would be as follows: normal monoallelic expression of both IGF2 and H19, biallelic expression of IGF2 with H19 silencing, IGF2 repression with biallelic expression of H19, and abnormal maternal IGF2 and paternal H19 expression. At any given time, all or some epigenetically different populations of cells may coexist within the same cell line. Moreover, epigenetic alterations in population of ESCs may not arise at once but rather during longer in vitro culture periods. Thus, biallelic expression of IGF2 and H19 detected in these ESCs could be a result of simply random sampling of colonies representing different populations of cells.
In mouse ESCs and ESC-derived fetuses, biallelic methylation of the Igf2/H19 IC was associated with biallelic repression of H19 and maternal expression of Igf2 . Similar to monkey ESCs, this aberrant imprinting apparently did not affect mouse ESC phenotype. However, injection of ESCs into blastocysts resulted in embryonic abnormalities and lethality in chimeric fetuses . It is possible that imprinting abnormalities observed in monkey ESCs, specifically biallelic activation of IGF2, may confer selective advantages during extended culture. Such cells could rapidly predominate within the former culture, leading to a population in which the majority or all of cells express the abnormality. Interestingly, recurring karyotypic abnormalities in human ESCs have been described, including trisomy of chromosomes 17q and 12, and these changes were associated with significant proliferative advantages during in vitro propagation [1, 31, 32].
In contrast to mouse and monkey ESCs, normal monoallelic expression of IGF2 and H19 has been reported for differentiated populations derived from human embryonic germ cells . The status of several human heterozygous ESC lines at early passages has also been determined with monoallelic expression of both IGF2 and H19 and normal methylation patterns in the corresponding IC . Interestingly, one human ESC line at passages greater than 66 showed activation of the previously silent allele, leading to biallelic expression of H19. These apparent differences in epigenetic stability between human and rhesus monkey ESC lines could be species-specific. However, it is possible that imprinting is influenced by culture conditions used to isolate and propagate cells. The monkey ESCs analyzed here were cultured in the presence of FBS, whereas most human ESCs are maintained in medium supplemented with knockout serum replacement. Although the molecular basis remains unclear, culture in the presence of serum may significantly alter kinetics of the cell cycle, in turn interfering with normal maintenance of imprints . In particular, gain of maternal methylation at the IGF2/H19 IC has been detected in mouse embryos and ESCs cultured in the presence of fetal bovine serum [2, 36]. Factors present in FBS may also interfere with normal expression of enzymes responsible for establishment and maintenance of DNA methylation.
The establishment of differentially methylated patterns is achieved by mutual action of DNA methyltransferases (DNMTs). Specifically, DNMT3a and DNMT3b are responsible for the de novo methylation of unmodified DNA and establishment of hemimethylated patterns, whereas DNMT1 is crucial for maintenance of hemimethylated CpG sites . Overexpression of DNMT1 results in hypermethylation, leading to biallelic expression of Igf2 . Interestingly, the mouse Igf2/H19 imprinting domain is particularly susceptible to alterations caused by Dnmt1 modulations. In contrast, several other imprinted genes, including Snrpn, are mainly resistant to de novo methylation, even when Dnmt1 is experimentally overexpressed . In mouse ESCs, several stress factors routinely associated with in vitro culture did not interfere with normal methylation of Snurf/Snrpn imprinting cluster . This is consistent with the stable imprinting of SNRPN and NDN that we observed in monkey ESC lines .
Our methylation studies were focused on the region corresponding to the human CTCF-6 binding site. This choice was justified by the evidence implicating this region in the function of the IGF2/H19 IC [8, 13, , , –17]. However, differential methylation at the CTCF-1 site has been reported with a correlation of hypomethylation of this region with colorectal cancers [13, 16]. Furthermore, analysis of microdeletions at various CTCF target sites in patients with BWS and Wilms tumors suggests that sites other than CTCF-6 may also play important roles in generating a functional insulator block and imprinted expression of both IGF2 and H19 [40, 41].
Biallelic expression of IGF2 has been associated with more than 20 different tumor types, including bladder and colon cancers [14, –16] and osteosarcoma . As a potent cell survivor factor, IGF2 may act by stimulating ESC proliferation such that overexpression, predicated by biallelic activation, leads to uncontrolled cell proliferation, overgrowth, and malignant transformation. Recently, a direct causal role of aberrant methylation and loss of imprinting in tumor formation was illustrated in a conditional knockout mouse model .
The conditions that resulted in aberrant hypermethylation within the IGF2/H19 IC in monkey ESCs may also have triggered methylation abnormalities in other imprinted or nonimprinted genes, resulting in reactivation of oncogenes by promoter hypomethylation or silencing of tumor suppressor genes caused by hypermethylation. Hypermethylation of promoter regions for several genes associated with cancer development has been noted for human ESCs . Particularly, hypermethylation of RASSF1, a putative tumor suppressor gene, was detected in several late-passage human ESC lines.
The consequences of transplantation of specific phenotypes derived from primate ESCs with imprinting disbalance into a recipient is unclear. However, cellular overproliferation and tumor formation resulting upon tissue or cell transplantation are potential clinical problems that must be addressed before clinical trials of ESC-based therapy are instigated.
D.W. has acted as a consultant for the NIH within the last 2 years.
We acknowledge the Division of Animal Resources and the Molecular Biology Core at the Oregon National Primate Research Center for assistance and technical services. We are grateful to Drs. John Hennebold and Xuemei Wu for help with Southern analyses; Darlene Pedersen, Cathy Ramsey, Carrie Greenberg, and Michelle Sparman for technical assistance; Julianne White for administrative support; and Joel Ito for help with illustration materials. The ART/ESC core facility assisted by providing semen samples, media, and mouse embryonic fibroblast feeder layers for ESCs. This study was supported by NIH Grants RR00163 to D. Dorsa, HD18185 to R.L. Stouffer, and RR15199 to S.M.M. H.S. is currently affiliated with the Department of Biochemistry, Mahidol University, Bangkok, Thailand. A.F. is currently affiliated with the Department of Obstetrics & Gynecology, University of Tokyo, Hongo, Kunkyosku, Tokyo, Japan.