A more profound understanding of the multifaceted biological functions of DNA methylation in mammalian and other genomes will remain elusive unless we have at hand the complete nucleotide sequences of these genomes including the fifth nucleotide. Researchers interested in the function of m5C have, therefore, been disappointed by the nevertheless admirable results of the human genome project. The human epigenome project has been initiated, and its results, once at least partly completed, will undoubtedly fill a serious gap in the anatomy of the human genome.
In the early 1990s, my laboratory has begun, as a pilot project as it were, to study DNA methylation patterns in various parts of the human genome. A part of these results has been recently summarized [5]. A more complete survey will be presented here. Our studies also had the aim to contribute to the understanding of epigenetic mechanisms and of human disease.
Interindividual concordance in human DNA methylation patterns. We have asked the question of how tightly preserved patterns of DNA methylation actually are in the promoter and 5´-upstream regions of a human gene among several individuals of different ethnic origins. The human, like many other eukaryotic genomes are characterized by the existence of complex patterns of DNA methylation which reflect in an unidentified way states of gene activities and inactivities and, equally important and related, of the chromosomal structure of the (human) genome. The 5´-upstream and promoter regions of the human genes for tumor necrosis factors TNFalpha and TNFbeta were screened with the genomic sequencing technique [9, 10] for the presence of m5C residues [17, 18].
Human DNA was derived from peripheral blood granulocytes, lymphocytes, or from sperm. The results indicated that patterns of DNA methylation at least in these genome segments were interindividually highly conserved. Thus, in the TNFalpha DNA from granulocytes of 15 individuals of African, Caucasian, or Chinese origin, the m5C residues were consistently found in 5´-CG-3´ dinucleotide positions -IX, -X, -XI (upstream of the cap-site) and in position +XVI (downstream of it). Very different distributions of m5C residues were observed in human cell lines HL60, Jurkat, and RPMI 1788. The TNFalpha gene is transcribed in human granulocytes. A very different result emerged for the promoter and upstream regions of the human gene for TNFbeta, which is not transcribed in human granulocytes. All 13 5´-CG-3´ dinucleotides in this segment were methylated, two only hemimethylated. Again, this pattern held true in the granulocytes from nine different individuals. The same sequence was completely unmethylated in human lymphocytes from the same individuals, in sperm and in the human cell lines RPMI 1788 and HL60, but almost completely methylated in cell line Jurkat [17]. These data document that methylation patterns in human DNA can be very different in different cell lines, but can be interindividually highly concordant. Moreover, patterns of DNA methylation in specific genome segments can vary a great deal.
The patterns of DNA methylation in the human TNFalpha and TNFbeta genes in granulocytes, monocytes and in several cases of acute (AML) or chronic myeloid leukemia (CML) were found to be very similar, except for one AML, in which the region in the TNFalpha gene was completely unmethylated, and several leukemia cases in which many sites in the TNFbeta gene were only hemimethylated [18]. In T and B lymphocytes of many individuals and in a number of Hodgkin and non-Hodgkin lymphomas, both the TNFalpha and TNFbeta genes were un- or hypomethylated. The DNA in HeLa cells in culture was completely methylated in the upstream and promoter regions of both genes [18]. If leukemia- or lymphoma-specific patterns should exist, they are very complex and not readily recognizable by this type of analysis. We have also compared methylation patterns by HpaII (5´-CCGG-3´) and HhaI (5´-GCGC-3´) cleavages of human DNA from European and Japanese individuals across about 500 kb of randomly selected DNA sequences in the human genome and found complete interindividual congruence of patterns by this method of admittedly intermediate sensitivity [19].
Methylation patterns in genetically imprinted regions of the human genome. The Prader-Willi/Angelman region on chromosome 15q11-q13 of the human genome is genetically imprinted, i.e. on the maternally and on the paternally inherited chromosome different genes are activated and others silenced. The molecular mechanisms underlying genetic imprinting are not completely understood. However, there is much evidence that differences in the methylation patterns in imprinted regions on the two alleles play an important role in the imprinting phenomenon. As part of a study on DNA methylation patterns in the human genome, we investigated all 5´-CG-3´ dinucleotides in the vicinity of exon 1 of the SNRPN and the D15S63 loci on chromosome 15q. The SNRPN transcripts might be involved in imprint switching during gametogenesis. By using the bisulfite genomic sequencing technique, we looked at individual chromosomal PCR products from normal individuals, from Prader-Willi and from Angelman patients. In this region, non-5´-CG-3´ C-residues were never methylated. Around exon 1 of the SNRPN gene, >96% of the 23 5´-CG-3´-dinucleotides were methylated on the maternal chromosome, as apparent from the genomic sequencing data with DNA from Prader-Willi patients in whom this segment was deleted on the paternal chromosome. In contrast, the same region on the paternal chromosome was completely devoid of methylated 5´-CG-3´ dinucleotides [20]. Angelman syndrome patients carry a deletion of the region on the maternal chromosome. The methylation status in the D15S63 locus, however, was quite different in that only two CfoI/HhaI sites were methylated to >96% on the maternal chromosome. The remaining five 5´-CG-3´ dinucleotides in this segment were methylated only 45-70% on the maternal, and to only 5-14% on the paternal chromosome [20, 103]. In an extension of this study [21], it was demonstrated again by bisulfite genomic sequencing that the 16 5´-CG-3´ dinucleotides in the 1.15 kb AS-SRO region on human chromosome 15q were methylated to 83 to 87% on both the maternal and paternal chromosomes in healthy individuals as well as in Prader-Willi and Angelman syndrome patients. There may be a low degree of mosaicism but there are no parent-of-origin-specific differences in the methylation patterns in this part of the genome [21]. These findings attest to the significant variability of the methylation patterns even in imprinted parts of the human genome.
Patterns of 5´-CG-3´ methylation in the promoter of the FMR1 gene; relevance for the fragile X syndrome. In patients suffering from the fragile X (FRAXA) syndrome, a naturally occurring 5´-(CGG)n-3´ repeat in the promoter and the 5´-untranslated regions (5´-UTR) of the FMR1 gene on human chromosome Xq27.3 is expanded excessively. In normal individuals, the value n ranges between 6 and 40, in premutation females n assumes values between 40 and 200, while in affected individuals the repeat n lies between 200 and >2000. The expanded repeat is hypermethylated, perhaps because such expansions are recognized as foreign DNA and become subject to modification. The ensuing inactivation of the FMR1 gene is the most likely cause for the disturbed embryonal and fetal development, which is the basis for the syndrome [104].
By applying the bisulfite genomic sequencing technique, we determined the methylation profiles in the promoter and 5´-UTR of the FMR1 gene on single chromosomes of healthy individuals and of selected premutation carriers and FRAXA patients [24]. In the DNA from FRAXA patients, there is considerable variability in the lengths of the 5´-(CGG)n-3´ repeats and in the levels of methylation in the repeats and the 5´-UTR regions in that all patients seem to be mosaics with respect to both parameters. In one patient with repeat lengths between n = 15 and >200, shorter repeats (n = 20 to 80) were methylated or unmethylated, longer repeats (n = 100 to 150) were often completely methylated. A particular repeat in this patient with n = 160, proved to be completely devoid of m5C residues. This repeat mosaicism was observed in several FRAXA patients analyzed in our laboratory [24]. As expected for healthy females with one at least partially inactivated X chromosome, hypermethylated repeats and 5´-UTR sequences were found. We also demonstrated that the authentic FMR1 promoter from healthy individuals was sensitive to methylation as shown by comparing the transient activities of constructs carrying the luciferase gene under the control of the unmethylated or the SssI (5´-CG-3´) completely methylated FMR1 promoter in human HeLa or 293 cells [24]. The methylated, inactive FMR1 promoter regions do not bind to specific cellular proteins as determined by footprinting analyses, whereas active, unmethylated promoter regions do bind proteins [23].
The promoter and 5´-upstream region of the RET protooncogene, a gene involved in the causation of Hirschsprung disease. The RET (rearranged during transformation) protooncogene plays a role in the causation of some familial or sporadic cases of Hirschsprung disease which results from an impaired development of the neural crest-derived neurons of the enteric ganglia [105]. We investigated the level of DNA methylation in a DNA segment of about 1000 bp in the promoter and 5´-upstream region of this gene [22]. By again applying the bisulfite genomic sequencing technique, DNA from peripheral white blood cells (PWBCs) from healthy individuals or from Hirschsprung disease patients was used as well as DNA from different human tissues and from a human embryonic kidney cell line. In a DNA section starting about 790 bp upstream from the transcriptional start site, a few m5C residues were found. However, in a 5´-CG-3´ rich 400 bp stretch in the RET gene promoter with 49 such dinucleotide pairs not a single m5C residue was present, although the RET protooncogene was not transcribed in many human tissues. Weak transcriptional activity was detected in many neural crest-derived human tissues. Obviously, the RET gene promoter was not silenced by a long-term signal like promoter methylation, possibly because it had to be expressed occasionally, and its transcription was controlled by factors other than DNA methylation. In in vitro experiments, in which the transcriptional activity of the RET gene promoter was assessed in linkage to an indicator gene after transfection into human cells, the activity of this promoter was decreased by HpaII (5´-CCGG-3´) methylation and abolished by SssI (5´-CG-3´) methylation. Hence the RET protooncogene promoter is sensitive to DNA methylation at least in transfection and transient transcription experiments [22].
Genes for proteins in the human erythrocyte membrane. Alterations in the structure of the erythrocyte membrane can be related to mutations in specific genes for proteins, which are essential elements of this membrane. These structural alterations of the erythrocyte membrane are responsible for hematologic diseases like hereditary elliptocytosis or hereditary spherocytosis [106]. By the bisulfite genomic sequencing procedure we determined patterns of methylation in the promoter and 5´-regions of the following human genes: the protein 4.2 (P4.2) gene (ELB42), the band 3 (B3) gene (EPB3), and the beta-spectrin (beta-Sp) gene (SPTB) [25, 26]. The promoter regions of the EPB3 and ELB42 genes were extensively methylated, whereas the promoters of the SPTB and the ankyrin genes were unmethylated. This finding again points to the interindividual conservation of certain patterns in the distribution of m5C residues in the human genome. The human SPTB promoter conforms to expectations in that it is unmethylated and fully active throughout erythroid development. In contrast, high levels of promoter methylation correlate with promoter activity for the EPB3 and ELB42 genes during their sequential activation in erythrocyte differentiation (Remus et al., submitted). In this respect, the EPB3 and ELB42 genes may resemble the genes of frog virus 3. This analysis was extended to patients with red cell membrane diseases, such as complete P4.2 deficiency due to mutations in the ELB42 gene, with hereditary spherocytosis with EPB3 mutations, and to hereditary elliptocytosis with mutations in the SPTB gene. Patterns of methylation in these patients were in general very similar to those of normal individuals [26].
Promoter and exon 1 of the human gene for the interleukin 2-receptor alpha-chain (IL-2Ralpha). The IL-2Ralpha gene is expressed in stimulated, but not in resting human T lymphocytes and plays an important role in promoting the T cell-mediated immune response. The -300 to +300 promoter/exon 1 region of the IL-2Ralpha gene was analyzed by the genomic sequencing technique for its content of methylated 5´-CG-3´ dinucleotides. In the cell types investigated--sperm, placenta, granulocytes, T-CLL, B-CLL, Jurkat, KB cells--m5C residues were not found in unstimulated or in stimulated lymphatic cells [107]. The 5´-CG-3´ sequence in position +198 was partly methylated. Even in cell types not relevant for the immune response, like in the human KB cell line, this regulatory region was consistently unmethylated. The promoter of a functionally essential human gene would not be long-term silenced by the methylation signal or else it could not be flexibly reactivated upon demand. Obviously, transient mechanisms of gene shut-off would be operationally preferred in these instances and thus remain independent of DNA methylation.
Human Alu sequences associated with specific genes. The human Alu sequences belong to the short interspersed repeat elements (SINE), comprise about 5% of the human genome and are about 300 bp long. The Alu elements might have been derived from reverse transcripts since they carry a 3´-dA-rich track. We analyzed the state of DNA methylation in the human Alu sequences associated with the genes for alpha1-globin, the tissue plasminogen activator (TPA), the adrenocorticotropic hormone (ACTH), and for angiogenin. DNA was investigated from lymphocytes, granulocytes, brain, heart muscle, and sperm as well as from human HeLa and KB cells. Both methylation-sensitive restrictases and genomic sequencing techniques were employed. In primary human cells, these Alu elements were highly methylated, but there were distinct differences in specific Alu sequence elements. In the DNA from haploid spermatozoa, Alu elements were often hypomethylated. The in vitro transcription of Alu elements was inhibited by 5´-CG-3´ methylation of these sequences [91]. The patterns observed in these specific Alu elements were identical in different individuals. The high level of DNA methylation in the Alu sequences associated with specific genes was consistent with their transcriptional silencing.
Promoter of the polymerase I-transcribed human ribosomal genes. The 5´-CG-3´-rich promoter region of the DNA-dependent RNA polymerase I (rDNA) genes was analyzed for the presence of m5C nucleotides by the genomic sequencing technique in DNA from primary human cells, from human tumor cells, and from human cell lines [108]. In none of the primary human cells or tumor cells was the rDNA promoter methylated. In contrast, in the DNA from the human cell lines HeLa (cervical cancer), KB (oral cancer), Jurkat (T cells), or CEM (T cells), the 5´-CG-3´ dinucleotides were methylated between 50% (KB) and 85% (Jurkat). Apparently, in the primary human cells, in granulocytes, T lymphocytes, spermatozoa, as well as in chronic T cell (T-CCL), myeloid (CML), or B cell (B-CLL) leukemia cells, which are all actively dividing, the essential rDNA genes need be transcribed actively and are not methylated [108]. In cell lines, rDNA genes are also actively expressed, and alternate mechanisms of overexpression must exist. Could some of the rDNA genes be over-amplified?
Randomly selected human genes in different Hodgkin's lymphoma and leukemia cell lines and in normal human lymphocytes. Several of the proto-oncogenes, parts of the TNFalpha and TNFbeta genes, the insulin receptor and lamin C genes were investigated by using the methylation-sensitive restriction enzyme HpaII and the control enzyme MspI. There were regions completely devoid of methylation; others were completely or partly methylated. Various lymphoma and leukemia cell lines differed among each other in different regions of the genome and differed again from the patterns observed in normal primary human lymphocytes [109]. Obviously, there is great variability, and no simple rules can be derived for the general characteristics of methylation patterns in leukemic versus normal human white cells.
The promoter of the human 5´-(CGG)n-3´-binding protein (CGGBP1). From the nuclei of human HeLa cells, we isolated a 20 kD protein which binds specifically to 5´-(CGG)n-3´ sequences [110, 111] and which might play a role in the control of promoters rich in CGG sequences like the FMR1 promoter in human DNA [112]. The human gene for this protein, termed CGGBP1, was located to human chromosome 3p, and its promoter was characterized in detail [113]. In several different human cell types, this promoter was unmethylated. The complete in vitro premethylation of all 18 5´-CG-3´ dinucleotides in this promoter led to its inactivation upon transfection into human HeLa cells with the luciferase gene as activity indicator [113].
Towards a complete nucleotide sequence of the human genome [114]. Projects have been initiated to complete the human genomic DNA sequence by including the 5th nucleotide and to initiate a Human Epigenome Project [11, 12]. Obviously, this will be a very important, but at the same time demanding, task. Above, I have summarized our contributions, of course not towards the solution, but towards a more general appreciation of the problem that the mammalian genomes harbor functionally important patterns of DNA methylation. The distribution of m5C residues in the human genome is thought to be essential for the understanding of chromatin structure and of the regulation of human gene expression in the many different cell types and during development.
From the available data, the following list of problems and desirable approaches can be compiled. Of course, we want to be cautious and cannot claim general validity of any of the presently plausible observations and conclusions:
- patterns of m5C distribution across the human genome are highly specific. Each region, each promoter exhibits its own individual pattern. The patterns can be interindividually conserved at least in several regions of the human genome;
- it is likely that each human cell type could have a different pattern in each genome segment;
- long-term promoter inactivity is generally associated with hypermethylation of the promoter. Inactive promoters can, however, also be un- or hypomethylated, particularly when they have to be occasionally reactivated. The state of promoter methylation by itself cannot reveal the activity status of a promoter;
- promoter strength might affect the pattern imposed upon a particular promoter;
- there are distinct differences in the patterns of methylation between normal human lymphocytes and lymphoma or leukemia cell lines in different segments of the genome. It is at present not possible to derive functionally meaningful conclusions from these differences other than that there are extensive alterations. We pursue the possibility that the process of oncogenic transformation of a human cell is associated, possibly causally, with global changes in the genome organization which is also reflected in drastically altered methylation patterns. From our work on Ad12-induced hamster tumors and on Ad12 DNA- or bacteriophage lambda DNA-transgenic hamster cells, we consider it likely that the insertion of foreign DNA is at least partly responsible for these alterations. Of course, it remains to be determined whether any of the global changes in methylation and transcription patterns demonstrated in tumor cells are the cause or the consequence of oncogenic transformation.
Answers to all your Biology Questions
