In order to determine the precise role of DNA methylation in the repression of γ-globin gene expression, we devised a biological approach for controlling local methylation. For this purpose we used a (Yeast Artificial Chromosome) YAC transgene carrying the entire human β-globin locus that has been shown to be properly regulated in the mouse [13], [14]. During early development, at the time of implantation, CpG island regions are recognized and protected from the process of de novo methylation [15]. This is accomplished through the involvement of cis acting sequences [16], [17], and one such element (IE—Island Element) identified in the Aprt locus, has been shown to work dominantly to prevent methylation of transgenes in vivo [3]. Preliminary experiments indicated that the insertion of two tandem copies of the IE near the γ-globin promoter could generate a relatively large (1 kb) unmethylated region that was then maintained in all somatic cells of the mouse (see Materials and Methods). On the basis of this observation we engineered a duplicated IE sequence flanked with LoxP sites 40 bp upstream to the γ^A -globin gene transcription start site in a 150 kb YAC which carries the entire human β-globin locus [18], and this construct was used to generate transgenic mice (Figure 1A). Four founder animals with intact YACs were obtained, and Southern blot analysis showed that 3 of them (47, 64, 113) were completely unmethylated at one selective restriction site in the γ^A gene promoter, indicating that this strategy was successful (Figure 1B). In one transgenic animal (66), the promoter remained partially methylated.

In order to manipulate DNA methylation and thereby test its effect on γ^A transcription, transgenic mice were crossed with two types of cre-expressing animals. The first line of cre mice (Cre) contain an expression vector which is constitutively active in all cell types from very early stages in development [19]. In F1 animals generated from this cross, the IE would be removed prior to implantation and, as a result, the γ^A gene becomes methylated (see Figure 1). The Mx-Cre mouse, in contrast, contains an interferon-inducible cre gene [20]. In F1 animals from this cross, the γ^A gene should be protected from de novo methylation at the time of implantation. In the adult, the IE can be removed by treatment with interferon, but the gene should still remain unmodified. Thus, using this strategy, it should be possible to generate matched transgenic mice with a fixed integration site whose only difference is the DNA methylation state at the γ^A promoter throughout the organism [21].

Using the globin YAC transgene, it was now possible to determine the effect of DNA methylation on transcription of the γ^A gene in both non-erythroid and erythroid cells. As non-erythroid cells, we used mouse embryonal fibroblasts (MEFs) generated from Cre or Mx-Cre transgenic (line 64) crosses. The bisulfite sequencing technique was used to measure the methylation state at the γ^A and γ^G promoters. In cells in which the IE was removed pre-implantation (Cre), the γ^A gene was found to be completely methylated at 4 upstream CpG sites. In contrast, cells that retained the IE during implantation (Mx-Cre) were highly unmethylated in this same region (Figure 1C). The γ^G promoter provided an important control that remained methylated in all MEFs from transgenic mice.

Semi-quantitative RT-PCR was used to determine the level of expression of the γ as well as β-globin genes. We took advantage of a unique restriction enzyme site within the γ^A gene to distinguish between γ^A and γ^G transcripts. We found that the unmethylated γ^A gene is transcribed at much higher levels (about 20 fold) than the methylated copy, while both the γ^G and β-globin genes are expressed at similar levels (Figure 2). These results are consistent with previous studies showing that DNA methylation of γ gene promoters inhibits transcription in transfected fibroblasts [22]. It should be noted that these steady state levels of RNA in MEFs are about 8–9 orders of magnitude less than that in fetal liver or adult erythroblasts. Most of this inhibition is clearly independent of DNA methylation.

In order to examine the role of DNA methylation in the control of γ-globin transcription in erythroid cells, we isolated adult spleens from transgenic animals (lines 47, 64, 113) and separated out erythroblasts using a cell-type specific antibody affinity column. In these cells, the β-globin gene is expressed at very high levels (Figure 2B). While the γ-globin sequences are transcribed at a low level compared to β-globin, this is still about 6 orders of magnitude higher than that measured in non erythroid cells. This increased expression is presumably due to the fact that this locus is in a more open overall chromatin conformation in these cells. It should be noted, however, that this amount of γ^A-globin RNA is still considerably lower than the levels produced in fetal liver. Undermethylation of the γ^A gene promoter in spleen erythroblasts brings about a striking 20-fold average increase in the activity of this gene as compared to its methylated counterpart in three independent transgenic founders. In contrast, the constitutively methylated γ^G promoter remained repressed, and this serves as an internal negative control for the induction seen with γ^A.

Although the normal percentage of γ globin expression in human erythroblasts is about 5–10% of the total (γ+β) (http://www.genecards.org/index.shtml; http://symatlas.gnf.org/SymAtlas), this ratio appears much lower (1%) in our mice, and this is, in fact, consistent with previous data from other transgenic animals carrying similar constructs (0.3–3%) [14], [23]. In the absence of DNA methylation on the γ^A promoter, γ globin expression ratio increases (to a level of 3–6%), but if we assume that the γ^G gene is affected by demethylation to the same extent, total γ/γ+β would actually be expected to reach levels of 10–25%, which is approximately the same order of magnitude as that seen in fetal liver (Figure 2 and or cases of transgenic HPFH (30–65%)where both the γ and β genes are active [24]. It thus appears that DNA methylation is serving as a local mechanism to repress the activity of γ-globin genes in adult erythroid cells.

One of the mechanisms by which DNA methylation inhibits gene expression is by bringing about the deacetylation of local histones [3]. We therefore examined the histone modification pattern of nucleosomes positioned in the promoter region of the γ^A and γ^G genes in vivo. This was carried out by isolating mononucleosomes from fibroblast or erythroblast cells and subjecting them to Chromatin Immunoprecipitation (ChIP) analysis using antibodies against Ac-H4. The active mouse β-actin and inactive α actin genes were used as matched controls for verifying the efficacy of immunoprecipitation.

As expected, the methylated γ^A gene promoter in non-erythroid cells is packaged in nucleosomes containing deacetylated histone H4, in a manner similar to other repressed genes. Strikingly, in the absence of DNA methylation, this region becomes acetylated, reaching the same maximum levels probably as a result of methyl binding proteins [12] associated with transcriptionally active genes such as β-actin or β-globin in erythroid cells (Figure 3A). Identical findings were observed for methylation of H3(K4) (data not shown), which is also known to be correlated with gene activity. This change in the histone modification pattern is surprising in light of the fact that γ-globin is expressed at such low levels in non-erythroid cells, suggesting that as opposed to other repression mechanisms, it is DNA methylation that plays a dominant role in setting up or maintaining histone acetylation at this locus.

Histone acetylation was also examined in adult erythroblasts purified from spleen. In these cells, the γ genes are expressed at levels 6 orders of magnitude greater than that seen in fibroblasts. Despite this, both the γ^G and γ^A promoters are packaged with nucleosomes containing deacetylated histone H4 (Figure 3B). This is most likely due to the presence of DNA methylation, since removal of these methyl moieties brings about a dramatic increase in both the acetylation of H4 and methylation of H3(K4) (Figure 3B and data not shown). This suggests that in erythroid cells, it is basically DNA methylation which keeps the γ genes packaged in their deacetylated form, and this probably accounts for their suboptimal level of expression.

One of the major effectors of globin expression in erythroid cells is the presence of highly specific transcription factors such as GATA1, CP1/NF-Y and NFE3 that recognize defined sequence motifs present at many sites within the globin locus, including the γ^A promoter [25]–[28]. In order to test whether DNA methylation plays a role in these interactions, we carried out in vivo footprint analysis in both adult erythroid cells and lymphocytes derived from the spleen. As expected, no significant footprints were observed in the lymphoid cells (data not shown). Similarly, footprints were also absent in normal adult erythroid cells in which the γ^A promoter is methylated. In contrast, when these methyl groups are absent, strong footprints over both the GATA-1 and distal CP1/NF-Y (CCAAT box) binding sites were observed (Figure 3C). It should be noted that these elements themselves do not contain any CpG sites, so it is unlikely that binding is inhibited directly by methyl groups. Rather, it appears that these specific protein-DNA interactions are prevented at the nucleosome level by local methylation-mediated histone deacetylation.

In patients with Hereditary Persistence of Fetal Hemoglobin (HPFH) the γ genes are expressed at unusually high levels in adult erythroid cells [29]. Our results on the role of DNA methylation at this locus strongly suggest that whatever the molecular mechanism at the source of this disease, activation of the γ-globin genes should only be possible if the promoter is demethylated. In order to test this hypothesis, we prepared erythroblast cultures [30] from HPFH patients with different genetic lesions, and then isolated DNA for bisulfite analysis of the γ^A and γ^G genes. In normal pro-erythroblasts, all 5 CpG sites in the γ promoter region are fully methylated. In contrast, we observed under-methylation in both the γ^A (4/7) and γ^G (5/34) promoters in a patient heterozygous for HPFH1 [31], [32] (Figure 4). This skewed pattern is in keeping with measurements showing that these patients also have a higher than normal ratio of γ^A to γ^G expression [32].

Similar bisulfite results were obtained for pro-erythroblast DNA from a patient with β° Thalassemia (IVS-II-I, G⇑A) [32] who expresses high levels of γ^G -globin, and in this case we detected massive under-methylation almost exclusively at the γ^G promoter (Figure 4). These findings are consistent with earlier experiments showing under-methylation at a single restriction site in the γ promoter regions from HPFH patients [33], [34]. Taken together, these results strongly suggest that whatever the mechanism [35], [36] involved in enhancing γ gene transcription, each case of HPFH is probably associated with molecular changes that can bring about local demethylation or perhaps histone acetylation in adult erythroid cells. In further support of this, we find that DNA methylation actually affects protein binding specifically at the exact same sites in the γ promoter that are frequently involved in HPFH [28].

The 120 bp IE from the hamster Aprt gene [21] was inserted as a dimer bounded by two lox elements into an engineered PacI restriction site 40 bp upstream to the human γ^A-globin gene transcription start site (Figure 5). The HindIII/XhaI fragment was then cloned into the yeast integrative plasmid pRS306 that was then inserted by homologous recombination into the human β-globin YAC-harboring yeast cells A201F4.3 [18]. Colonies were subsequently screened for growth on plates lacking uracil. Correct integration was verified by Southern blot analysis. Plasmid sequences were removed by induction of “looping out” through growth on 5FOA and correct excision was verified by Southern blot analysis.

YAC DNA was purified from pulse-field gels according to standard protocols [44] and injected into fertilized mouse eggs that were transferred to foster mothers. Littermates were screened for the β-globin YAC transgene by PCR, and positive samples verified by Southern blot. All three transgenes (lines 47, 64, 113) used for analysis were found to be completely intact and present as a single copy. Founders were crossed with Cre [19] or Mx-Cre mice [20]. To remove the IE sequences after birth Mx-Cre X β-globin YAC F1 mice were injected 4 times with polyI-polyC (Sigma) on alternate days. MEFs (passage number 3–4) carrying a methylated γ^A gene were prepared from Cre embryos. MEFs carrying an unmethylated γ^A gene were prepared from Mx Cre mice and then transfected with a type 5-cytomegalovirus Cre vector [45] to remove the IE element. Methylation patterns in non erythroid cells were analyzed either by Southern blot or bisulfite methylation analysis as described [46]. Following treatment, the DNA was amplified using the PCR primers forward -1 (5′-TTGTTTGAAAGGGTTTTTGGTTAAATTTTATTT-3′) and reverse-1 (5′-TCCTAATCACCAAAACCTACCTTCCCAAAA-3′) and subsequently with the nested primers forward -2 (5′-TTTTGGTTAAATTTTATTTATGGGTTGGTTAGTT-3′) and reverse-2 (5′-AACTTATAATAATAACCTTATCCTCCTCTATAAAATA-3′). The resulting products were cloned using the pGEM-T Easy Vector System I kit (Promega) and individual colonies then sequenced. Methylation patterns in purified erythroid cells were analyzed by digestion with methylation sensitive restriction enzyme followed by amplification by PCR.

Erythroblasts from fetal liver and adult spleen of Cre or Mx-cre crossed transgenic mice were isolated by MACS (Magnetic Activated Cell Sorter) with biotinilated α-ter119 antibodies (PharMingen). Mouse embryonic fibroblasts (MEFs) were derived from Cre or Mx-cre crossed 12.5 dpc transgenic embryos. RNA was extracted with TriPure Isolation reagent (Roche), treated (250 ng) with DNaseI (Promega) and converted to cDNA using the Molony murine leukemia virus reverse transcriptase (Promega) and random hexanucleotide pd(N)₆ primers (Pharmacia) in a reaction volume of 50 µl under conditions recommended by the manufacturer. 1 or 3 µl of the cDNA was used for PCR reactions (95°C for 1 min, 58°C for 1 min, 72°C for 1 min) in the presence of ³²PαdCTP (Amersham) using appropriate primer pairs [47]. These primers were designed to span intron-exon junctions in order to distinguish between cDNA and genomic DNA. RT-PCR fragments were separated on 5% or 7% polyacrylamide gels and exposed for autoradiography. The relative amount of γ^A and γ^G cDNA was defined by digestion with the restriction enzyme PstI that cuts only the γ^A PCR product. The level of γ and β-globin transcription was measured by phosphor-imager analysis of diluted samples after normalizing for Aprt.

Nuclei were prepared from cell culture or purified mouse spleen erythroblasts, treated with an HDAC inhibitor (PMSF) and digested by MNase (micrococcal nuclease) to generate mono-nucleosomes that were separated on a sucrose gradient [48]. Immunoprecipitation (IP) was carried out using anti-acetylated histone H4 or anti me-H3(Lys4) antibodies (Upstate Biotechnologies), and the bound fraction purified by protein A-Sepharose chromatography (Sigma). Input and Bound DNA fractions were analyzed by semiquantitative PCR using two concentrations. Primer sequences are available upon request.

In vivo Dimethyl Sulfate (DMS) footprinting and ligation-mediated PCR (LMPCR) was performed as described earlier [49], [50], except that the in vivo DMS treatment was done on erythroid cells in cell suspension. In a 50 ml conical tube, 10–25 million cells were suspended in 10 ml of culture medium without bovine serum. DMS was added to a final concentration of 0.2%, and the cells were incubated at room temperature for 5 min. 40 ml of ice-cold PBS was added to stop the DMS reaction and the cells were centrifuged quickly (250 g, 3 min), washed two times with 30 ml of ice-cold PBS before DNA preparation. For the LMPCR experiment we designed the first primers in such way that they distinguished between the γ^G and lox-P inserted γ^A promoters. Primer extension was then carried out with either L1n: 5′-GCGTCTGGACTAGGAGCTTTTAAT-3′ or L1: 5′-AAGTAACGCATTTGCT GGAAG-3′ (which includes the IE element). Primer, L2: 5′-TTAATCTCAGACGTTCCAGAAGCGAGTGTG-3′ and LP25 linker primer were used for amplification. The run off hybridization probe was made on PCR fragment L3: 5′-CAGAAGCGAGTGTGTGGAACTGCTGAA-3′ – U: 5′-AAAAAAATTAAGCAGCAGTATCCTCTTGGG-3′ with the L3 primer.