The differentiation of naive CD4+ T lymphocytes into Th1 and Th2 lineages generates either cellular or humoral immune responses. Th2 cells express the cytokines IL-4, -5, and -13, which are implicated in asthma and atopy. Much has been published about the regulation of murine Th2 cytokine expression, but studies in human primary T cells are less common. We have developed a method for differentiating human CD45RA+ (naive) T cells into Th1 and Th2 populations that display distinct cytokine expression profiles. We examined both CpG methylation, using bisulfite DNA modification and sequencing, and chromatin structure around the IL-4 and IL-13 genes before and after human T cell differentiation and in normal human skin fibroblasts. In naive cells, the DNA was predominantly methylated. After Th2 differentiation, DNase I hypersensitive sites (DHS) appeared at IL-4 and IL-13 and CpG demethylation occurred only around the Th2-specific DHS. Both DHS and CpG demethylation coincided with consensus binding sites for the Th2-specific transcription factor GATA-3. Although fibroblasts, like naive and Th1 cells, did not express IL-4 or IL-13, DHS and unmethylated CpG sites that were distinct from the Th2-specific sites were observed, suggesting that chromatin structure in this cluster not only varies in T cells according to IL-4/IL-13 expression but is also tissue specific.
During CD4+ T lymphocyte differentiation, at least two distinct subsets of cells, Th1 and Th2, are derived from antigenically naive precursors, depending on the nature of the pathogen or allergen encountered (1, 2). These subsets are characterized by the different arrays of cytokines that they secrete on activation. Th1 cells express IFN-γ and TNF-β and regulate cell-mediated immunity against intracellular pathogens. Th2 cells express IL-4, -5, and -13, which promote Ab-mediated (humoral) responses and eosinophil-mediated effects (1) and are also implicated in asthma and atopy (3). The major factor that determines the commitment of naive T cells to either the Th1 or the Th2 lineage is exposure to IL-12 or IL-4, respectively. This process has been extensively investigated in mice (4). In Th1 cells, signals from the IL-12R activate the STAT4 pathway, which, together with signals from the TCR and the transcription factor T-bet, plays a role in activating IFN-γ expression (5, 6). In Th2 cells, signals from the IL-4R activate the STAT6 pathway, leading to the activation of many IL-4-regulated genes (7). Two Th2-specific transcription factors have been identified: c-Maf and GATA-3 (8, 9). c-Maf activates IL-4 expression (10) and GATA-3 regulates IL-4, -5, and -13 expression (11, 12).
The genes encoding the human Th2 cytokines IL-4 and IL-13 are located on the long arm of chromosome 5 between the ubiquitously expressed genes KIF3A and RAD50 (see Fig. 1 A) (13). The organization of the murine IL-4/IL-13/IL-5 locus is similar to that of the human locus, and a 401-bp noncoding sequence that is conserved between humans and mice, termed conserved noncoding sequence-1 (CNS-1),3 is found between IL-4 and IL-13 (14). It has recently been suggested that CNS-1 is a coordinate regulator of IL-4, IL-5, and IL-13 (15). Chromatin studies in murine T cells have shown that Th2 differentiation from naive cells is accompanied by the appearance of Th2-specific DNase I hypersensitive sites (DHS) around the IL-4 and IL-13 genes (16, 17). Because most of these DHS were not seen in naive or Th1 cells, it was proposed by Agarwal and Rao (16) that the chromatin is “closed” in naive cells, and that after Th2 cell differentiation the chromatin structure of the IL-4/IL-13 cluster “opens” to enable the expression of these cytokines.
Chromatin structure is linked to the methylation status of the cytosines in CpG dinucleotides (18). Changes in CpG methylation patterns have been found to correlate with several processes within the murine immune system, including Ig κ-chain rearrangement during B lymphocyte differentiation (19), TCR-β chain rearrangement (20), and IL-3 and IFN-γ transcription during CD8+ T lymphocyte differentiation (21, 22). Limited CpG methylation analysis of the murine IL-4 and IL-5 genes using methylation-sensitive restriction enzymes has led to reports that the DNA in IL-4 and IL-5 is methylated in naive and Th1 cells but becomes demethylated in Th2 cells (16, 23). More recently, T cell-specific deletion of the DNA methyltransferase Dnmt1 resulted in decreased methylation and increased activation-induced expression of several cytokines in naive murine T cells, suggesting that CpG methylation plays an important role in limiting the expression of these genes in naive cells (24).
The chromatin structure and DNA methylation status of IL-4 and IL-13 have not been previously studied in human T cells. In this work we show that Th2-specific DHS appear in the second intron of IL-4, around the IL-13 promoter and near (but not in) CNS-1 after 14 days of human CD4+ T cell differentiation. These DHS coincide with several regions of regulatory interest in murine Th2 cytokine genes. This is also the first in-depth analysis, using bisulfite DNA modification and sequencing, of the CpG methylation status across the IL-4 and IL-13 genes before and after differentiation. In Th2 cells, CpG demethylation was not the locus-wide event reported to occur in the murine Th2 cytokine cluster (25) but occurred only around DHS. We compared DHS and CpG methylation patterns in T cells with those in normal human skin fibroblasts, a nonhemopoietic lineage, and found that the DNA around IL-4 and IL-13 in nonexpressing cells is neither fully methylated nor insensitive to DNase I digestion, in contrast to previous reports (16, 17, 23). Chromatin structure and DNA methylation status differ between hemopoietic (T cells) and nonhemopoietic (fibroblasts) cells, suggesting that they are tissue specific.
Materials and Methods
Cells and tissue culture
Venous blood was taken from nonatopic healthy human male volunteers using heparin as an anticoagulant. Ethical approval for the use of human volunteers in this study was obtained from the institutional ethical review committee. PBMCs were isolated using Lymphoprep (Nycomed, Oslo, Norway) according to the manufacturer’s instructions. CD4+ T cells were isolated from PBMC using a CD4 Positive Isolation kit (Dynal Biotech, Oslo, Norway) according to the manufacturer’s instructions. Naive CD45RA+ cells were purified from CD4+ cells by depletion of CD45RO+ cells using mouse anti-human CD45RO Ab (UCHL1; 0.5 μg/1 × 106 cells; BD PharMingen, San Diego, CA) and rat anti-mouse IgG2a Dynabeads (Dynal Biotech) according to the manufacturer’s instructions. The purity of fractionated cell populations was determined by FACS analysis using FITC-conjugated anti-CD45RA (L48; BD PharMingen), PE-conjugated anti-CD45RO (UCHL1; BD PharMingen), CyChrome-conjugated anti-CD4 (RPA-T4; BD PharMingen); FITC-conjugated anti-CD14 (MφP9; BD PharMingen), and PE-conjugated anti-CD16 (3G8; BD PharMingen). Samples were analyzed on a FACSCalibur (BD Biosciences, Mountain View, CA).
Purified CD45RA+ cells (1 × 106/ml) were cultured in RPMI 1640 (Life Technologies, Rockville, MD) supplemented with 10% FCS, 2 mM l-glutamine (Life Technologies), 100 U/ml penicillin (Life Technologies), and 100 μg/ml streptomycin (Life Technologies). Cells were stimulated with plate-bound anti-CD3 (1 μg/ml; clone OKT3) and anti-CD28 (2 μg/ml; clone 15E8; Central Laboratory of The Netherlands Red Cross Blood Transfusion Service, Amsterdam, The Netherlands), and rIL-2 (50 U/ml; Eurocetus, Amsterdam, The Netherlands). To direct Th1 differentiation, IL-12 (2.5 ng/ml; R&D Systems, Minneapolis, MN) and anti-IL-4 (5 μg/ml; clone MP4-25D2; BD PharMingen) were added. For Th2 differentiation, IL-4 (12.5 ng/ml; NBS Biologicals, Huntingdon, Cambridgeshire, U.K.), anti-IFN-γ (5 μg/ml; clone B-B1; BioSource International, Camarillo, CA), and anti-IL-10 (5 μg/ml; clone JES3-9D7; BioSource International) were added. After 4 days, the cells were expanded under the same conditions in the absence of anti-CD3 or anti-CD28. Cells were then restimulated every 7 days. When required, cells were activated with PMA (5 ng/ml; Sigma-Aldrich, St. Louis, MO) and ionomycin (500 ng/ml; Calbiochem, La Jolla, CA) for 4 h. After differentiation and before DHS analysis, dead cells were removed from the culture using the Dead Cell Removal kit (Miltenyi Biotec, Auburn, CA) according to the manufacturer’s instructions.
Untransformed human fibroblasts from the skin of three normal males were obtained from the European Cell Culture Collection (ECACC; nos. 90011806, 90011807, and 90011810) (26). They were grown in MEM with Earle’s salts (Life Technologies) supplemented with penicillin, streptomycin, and l-glutamine as above, 1× nonessential amino acids (Life Technologies), and 15% FCS. Where necessary, cells were activated with IL-1β, TNF-α, and IFN-γ (10 ng/ml each; R&D Systems) for 24 h.
DNA sequence and computational analysis
The contig shown in Fig. 1 was created from sequences deposited in the GenBank database (accession nos. AC004237, AC004039, AC004041, and AC004042) using MacVector version 6.5 (Oxford Molecular, Oxford, U.K.). Numbering throughout this report is from the start of sequence AC004237. Sequence analysis for restriction enzyme sites, CpG sites, and PCR primer design and sequence alignment was performed using MacVector. All DHS and CpG sites that are described in this paper are found between two EcoRI sites that are located at the 3′ end of IL-4 and upstream of IL-13, respectively.
Isolation of total cellular RNA and DNA was performed using the RNA/DNA Mini kit (Qiagen, Valencia, CA) according to the manufacturer’s instructions and RT-PCR was performed as previously described (27) using 25 ng (T cells) or 500 ng (fibroblasts) of reverse-transcribed RNA. The primers used in RT-PCR for β-actin, KIF3A, IL-4, IL-13, and RAD50 have been described previously (27). RT-PCR primers for IFN-γ and GM-CSF were as follows: IFN-γ sense, 5′-GCAGGTCATTCAGATGTAGCGG; IFN-γ antisense, 5′-TGTCTTCCTTGATGGTCTCCACAC; GM-CSF sense, 5′-GCCAGCCACTACAAGCAGCAC; GM-CSF antisense, 5′-CAAAGGGGATGACAAGCAGAAAG. MWG Biotech (Ebersberg, Germany) supplied all primers.
The final number of PCR cycles was selected so that a clearly visible amplimer could be seen before the PCR reached product saturation (Fig. 1, C and D).
DNase I hypersensitivity analysis
DHS analysis of the T cells was performed using methods adapted from Cockerill (28). Briefly, T cell nuclei were prepared by washing cells in ice-cold PBS and resuspending them in lysis buffer (20 mM Tris (pH 7.5), 15 mM NaCl, 6 mM MgCl2, 0.1% Nonidet P-40, 20% glycerol, 0.1 mM EDTA, 0.1 mM EGTA, and 0.1 mM PMSF). After a 6-min incubation on ice, the nuclei were washed twice with wash buffer (lysis buffer without Nonidet P-40) and once with DNase I digestion buffer (DDB) (wash buffer with 1 mM CaCl2). The nuclei were resuspended in DDB before digesting equal aliquots with 10–26 U of DNase I (Sigma-Aldrich) at 25°C for 10 min. Fibroblast chromatin was digested with DNase I following cell permeabilization as described (29). After DNase I treatment, DNA was isolated using the DNeasy Tissue kit (Qiagen) according to the manufacturer’s instructions.
DNase I-digested DNA was then digested with EcoRI or BclI and precipitated. Five micrograms per lane of restriction-digested DNA were electrophoresed through a 0.7% TAE agarose gel and Southern-blotted as described (30). Fragments to be used as probes in DHS analyses were amplified from genomic DNA using the following primers: IL-4 DHS probe, 5′-CCAATCAGCACCTCTCTTCCAG and 5′-AACCTCAGAATAGACCTACCTTGCC; CNS-1 DHS probe, 5′-CAGTCCTCAGGAGATGTGATTGTG and 5′-GTCAGGAGAGGGGCAGAACAG; IL-13 DHS probe, 5′-GACTCCTGGTGTCCACTGCTTTAG and 5′-TCAAAAATGTCTTGGGTAGGCG. Probes were radiolabeled with [α-32P]dCTP or [α-32P]dATP (3000 Ci/mmol) using the Prime-a-Gene Labeling System (Promega, Madison, WI) and prehybridization and hybridization, using ULTRAhyb (Ambion, Austin, TX), were conducted according to the manufacturer’s instructions. DHS were mapped using at least two restriction enzymes, and their location was confirmed by probing both ends of the same parent fragment.
Bisulfite modification of DNA
CpG methylation was examined in DNA from T cells of two normal male volunteers and in DNA from untransformed fibroblasts of three normal males obtained from ECACC (see Cells and tissue culture) with similar results; all CpG methylation data presented in this report are from one of these individuals, respectively. The method for bisulfite modification of DNA was adapted from Frommer et al. (31). Ten micrograms of genomic DNA were digested with the restriction enzyme AflIII (New England Biolabs, Beverly, MA) to fragment the DNA. For sodium bisulfite modification, DNA was denatured in a final volume of 111 μl 0.3 M NaOH (Sigma-Aldrich) at 37°C for 15 min. To this was added 1.1 ml of bisulfite/hydroquinone solution (pH 5; 9.5 g sodium bisulfite (Sigma-Aldrich), 1.5 ml 3 M NaOH, and 2.5 ml 20 mM hydroquinone (Sigma-Aldrich) in a final volume of 20 ml). The samples were incubated in the dark at 55°C for 5 h. Salt was removed from the modified DNA using the Wizard DNA Clean-Up System (Promega) according to the manufacturer’s instructions, and the DNA was eluted in 100 μl H2O. The DNA was desulfonated in a final concentration of 0.3 M NaOH at 37°C for 15 min. The samples were neutralized by adding 1 volume of 6 M ammonium acetate (pH 7), and the DNA was precipitated with three volumes of 100% ethanol. The precipitated DNA was washed twice with 70% ethanol, dried, and resuspended in 100 μl EB buffer (10 mM Tris (pH 8.5); Qiagen).
PCR and sequencing of BSM DNA
Primers for the amplification of bisulfite-modified (BSM) DNA were designed to be specific for modified DNA only and are as follows: IL-4 promoter BSM-PCR, 5′-GTTTTGTGAGGTTGTTTAAAGTTTTGATG and 5′-CTAATTAACCCCAAATAACTAACAATC; IL-4 DHS I and II BSM-PCR, 5′-GAGAAAATGTATTATTAGTTGTTAAATT and 5′-CATTTTATCTAAAAAACTTCCTATAAC; IL-13 promoter BSM-PCR, 5′-TTGGGTGATGTTGATTAGTTTTTTAATGAG and 5′-CAAATCTTAAAAACTCTACCCTAAACCC. PCR were conducted under the following conditions: 400 ng BSM DNA, 1× PCR buffer II (PE Applied Biosystems, Foster City, CA), 3 mM MgCl2, 0.2 mM dNTPs, 2.5 U AmpliTaq Gold (PE Applied Biosystems), and 1 μM of each primer in a final volume of 100 μl. Cycling conditions were as follows: 10 min at 95°C, followed by five cycles of 2 min at 95°C, 3 min at 60°C, and 3 min at 72°C, then a further 30–35 cycles of 1 min at 95°C, 1 min at 60°C, 1 min at 72°C, and a final 10-min extension step at 72°C. To test the specificity of the primers for modified DNA, the reactions were also performed using 400 ng unmodified, AflIII-digested human genomic DNA (Promega).
Direct sequencing of BSM-PCR products was performed using the Thermo Sequenase Radiolabeled Terminator Cycle Sequencing kit (Amersham Pharmacia Biotech, Piscataway, NJ) according to the manufacturer’s instructions. The sequences were checked against those published, and the intensity of the C vs T bands on the autoradiographs was evaluated visually. PCR fragments were also cloned into the vector pCR 2.1 TOPO using the TOPO TA Cloning kit (Invitrogen, San Diego, CA) according to the manufacturer’s instructions. Plasmid DNA was sequenced with the Big Dye DNA Sequencing kit (PE Applied Biosystems) according to the manufacturer’s instructions, using the M13 forward and reverse primers. The reactions were analyzed on a PE Applied Biosystems 377 sequencing machine.
Restriction enzyme analysis of CpG methylation
Genomic DNA from activated and unactivated Th1 and Th2 cells and from fibroblasts was digested first with EcoRI, then with HpaII or MspI. The DNA was electrophoresed and Southern blotted as described above before hybridizing each region with up to five probes (data not shown) to enable accurate mapping of HpaII site digestion.
Differentiation of human Th1 and Th2 cells in vitro
The surface expression of different isoforms of CD45 can be used to distinguish naive CD4+ T cells (which express CD45RA) from memory T cells (which express CD45RO) (32, 33). Therefore, CD45RA+ T cells were purified from CD4+ T cells to form the naive starting population (Fig. 1,B, top and middle panels). After CD45RO depletion, the T cells were at least 98% CD4+CD45RA+ in all experiments. NK cells (CD16+) or monocytes (CD14+) represented <1% of the population (Fig. 1 B, bottom panels).
The naive CD4+CD45RA+ T cells were differentiated to a Th1 (using IL-12 and anti-IL-4) or a Th2 (using IL-4, anti-IFN-γ, and anti-IL-10) phenotype. Cytokine expression was analyzed by RT-PCR before and after differentiation. CD45RA+ cells stimulated with PMA and ionomycin for 4 h did not express IL-4, IL-13, or IFN-γ mRNA (Fig. 1,C, top panel). Under the same conditions, the in vitro-differentiated Th1 cells expressed IFN-γ mRNA (Fig. 1,C, middle panel) and the Th2 cells expressed IL-4 and IL-13 mRNA (Fig. 1,C, bottom panel), indicating that, after 14 days in culture, the CD4+CD45RA+ cells had differentiated into either Th1 or Th2 populations. Transcription of KIF3A and RAD50 mRNA was also examined. KIF3A and RAD50 mRNAs were expressed constitutively in both naive and differentiated T cell populations (Fig. 1,C), suggesting that these genes are regulated independently of IL-4 and IL-13. KIF3A and RAD50 were not up-regulated upon activation. Intracellular cytokine staining and FACS analyses (data not shown) indicated that <1% of the Th1 population expressed IL-4 and IL-13, while 90.8% of these cells expressed IFN-γ. Of the Th2 population, 19.8% of the cells expressed IL-4, 36.2% expressed IL-13, and 1.4% expressed IFN-γ (data not shown). IL-4 and IL-13 expression was not detected in fibroblasts, even upon activation (Fig. 1 D). GM-CSF expression, which was used as an activation marker in fibroblasts, was detected only in the activated population.
DHS and CpG methylation in the IL-4 gene
Unactivated cells were used in this study so that differentiation-specific (as opposed to activation-specific) chromatin structure changes could be examined in the absence of active IL-4 and IL-13 transcription. DNase I hypersensitivity assays revealed three DHS in the IL-4 gene in Th2 cells (Fig. 2,A, middle panel, and C). The sites designated I, II, and III were ∼1, 1.2, and 2.3 kb downstream of the IL-4 transcription start site. The positions of DHS I and II were confirmed by comparing the position of the DHS bands with that of an EcoRI-Fse I fragment (data not shown). Therefore, the Th2-specific DHS were located in the second intron of IL-4, and DHS I and II coincided with several consensus GATA-3 binding sites (Fig. 2,E). Very faint bands corresponding to DHS II and III are visible in the Southern blot of Th1 cell DNA (Fig. 2 A, left panel). However, these are substantially weaker than the hypersensitive sites visible in Th2 DNA and may be caused by the activation of the IL-4 locus in a small minority of cells in the Th1 population.
Two DHS were observed in fibroblast chromatin (Fig. 2 A, right panel, and C). These DHS were distinct from those observed in Th2 chromatin: DHS FI is ∼0.8 kb from the IL-4 transcription site, i.e., just downstream of the second exon, and DHS FII is ∼3.4 kb from the transcriptional start site. The presence of these fibroblast-specific DHS in IL-4 was confirmed using fibroblasts from a second individual. CD45RA+ naive T cells could not be analyzed by this method due to the large number of cells required. The appearance of DHS in the IL-4 gene of both Th2 cells and fibroblasts was intriguing and prompted us to investigate CpG methylation across this gene in expressing and nonexpressing cell types.
Initially, we examined the IL-4 gene by hybridizing DNA that had been digested with the methylation-sensitive restriction enzyme HpaII to five radiolabeled probes that spanned the entire region (Fig. 2,B, only one Southern blot is shown). Differences in HpaII digestion between Th1 and Th2 cells were seen at two sites between the Th2-specific DHS I and II (Fig. 2, B and D, ∗). The unmethylated CpGs at these sites in Th2 cells thus coincided with the appearance of Th2-specific DHS; these CpG sites were methylated in Th1 cells. Activation did not affect CpG methylation at HpaII sites in either Th1 or Th2 cells (Fig. 2,B). Although fibroblasts do not express IL-4, the IL-4 gene was not completely methylated at all CpG sites. Indeed, HpaII digestion indicated that more CpG sites were unmethylated in these cells than in T cells (Fig. 2, B and D).
HpaII digestion analysis yields information about relatively few CpG sites. Of the 115 CpG sites found in the 8.7-kb IL-4 gene, only 14 fall within a HpaII restriction site. Hence, we investigated CpG methylation status in more detail using bisulfite modification and sequencing. This procedure involves using sodium bisulfite to convert unmethylated, but not methylated, cytosine residues to uracil (31). Subsequent amplification and sequencing of the modified DNA enables one to distinguish the originally methylated cytosines from the unmethylated cytosines, which are amplified as thymines. The regions where HpaII digestion patterns differed between the cell types (Fig. 2 D) were amplified from BSM DNA from naive, Th1, and Th2 cells and fibroblasts, and the PCR products were sequenced both directly and after cloning, to obtain different but complementary sets of information. Direct sequencing of BSM-PCR products reveals the overall methylation status of the total population of input fragments, and whether or not the modification reaction (i.e., conversion of unmethylated cytosine to thymine) has proceeded to completion. Cloning the BSM-PCR amplimers and sequencing the plasmid insert reveals the methylation status of the CpG sites within a single DNA fragment. Sequencing several such clones reveals any variability in CpG methylation patterns within the population as well as any correlations between the methylation status of particular CpG sites.
Seven CpG sites (Fig. 3,A, numbered 1–7) are located in the 318-bp region that encompasses the IL-4 proximal promoter, the transcriptional start site, and the first exon. Direct sequencing of the BSM-PCR amplimers (Fig. 3,C) shows that the CpG sites in the first IL-4 exon of T cells are predominantly methylated (i.e., remain unconverted; Fig. 3,C, lanes C). By contrast, in the Th2 DNA sequence of the IL-13 promoter (see Fig. 7,C) the T and C bands of CpG sites 4 and 5 are of similar intensity, indicating that these cytosines are methylated in only approximately half the DNA fragments. These findings are in good agreement with the sequences of the individual cloned fragments of amplified DNA (Fig. 3 B), which confirms the validity of both methods. Furthermore, the variety of CpG methylation patterns observed in the cloned fragments suggests that a representative population of the modified DNA molecules is amplified.
The seven CpG sites in the IL-4 promoter and first exon are predominantly methylated in all three T cell populations, which suggests that demethylation of CpG sites around the IL-4 promoter is not required for IL-4 expression in T cells. This region was also largely methylated in fibroblasts. We also used BSM PCR and sequencing to analyze the region in intron 2 that coincided with Th2-specific DHS I and II. Nine CpG sites are found in this region (Fig. 4,A), two of which correspond to the HpaII sites highlighted by an asterisk in Fig. 2,D. In contrast to the region around the IL-4 promoter, partial demethylation of CpG sites was seen after Th2, but not Th1, differentiation. Eighty percent of the Th2 DNA fragments had at least one unmethylated CpG site, compared with only 30–40% of the CD45RA+ and Th1 DNA fragments. Furthermore, CpG sites 1 and 9, which are close to the Th2-specific DHS I and II, were both unmethylated in over half the Th2-derived fragments (Fig. 4 C). Most of the fragments that were unmethylated at CpG site 1 were also unmethylated at CpG site 9. The unmethylated CpG sites in the Th1-derived fragments may result from a small minority of cells that might be capable of expressing IL-4, although our findings that CpG methylation patterns can differ between cells in a population are consistent with those of Fitzpatrick et al. (21), which showed that clonal heterogeneity is a prominent feature of cytokine expression in primary T cells.
This entire region was largely unmethylated in fibroblasts. Because the IL-4 CpG methylation findings in fibroblasts were unexpected, we confirmed them by both MspI/HpaII digestion and BSM sequencing in DNA from three individuals, at passage numbers ranging from 7 to 22 (data not shown). We found very few differences in methylation patterns, either among the three individuals or between passage numbers. Rare differences in the degree of methylation occurred at a single CpG site, rather than across an entire region, and were of the order of a 10–50% increase or decrease in methylation status (data not shown). A trend of consistently increased or decreased methylation changes with increasing passage was not observed.
In summary, the IL-4 DNA methylation patterns in all four cell types (Figs. 2 D, 3B, and 4B) show clearly that naive cell and Th1 cell DNA is predominantly methylated in IL-4 and that, after Th2 differentiation, the DNA becomes demethylated only in the second intron at a region coinciding with DHS formation, which suggests that these two events are linked. Meanwhile, despite the unmethylated status of the fibroblast DNA across the 5′ half of IL-4, any correlation between DHS and CpG demethylation is absent, in contrast to the Th2 cells.
DHS and CpG methylation around CNS-1
Because recent reports suggested that CNS-1 might be a coordinate regulator of Th2 cytokine expression (14, 15), we examined this region for chromatin and CpG methylation changes. In differentiated Th2 cells, two DHS were observed between CNS-1 and IL-13 (Fig. 5,A, middle panel, and C, designated I and II); in Th1 cells the corresponding band is again considerably weaker. In fibroblasts, DHS designated FI and FII were also seen (Fig. 5 A, right panel), one of which was located similarly to DHS I in Th2 cells. Unlike in the murine locus (17), DHS did not appear in CNS-1 itself: the Th2 DHS are ∼500 bp away from this conserved region in the direction of IL-13. No DHS were found between IL-4 and this region in differentiated Th1 or Th2 cells or fibroblasts (data not shown).
To investigate CpG methylation status in this region, HpaII digestion was performed (Fig. 5, B and D). Of the 10 HpaII sites within the 6.5-kb EcoRI fragment that includes CNS-1, eight were totally methylated in Th1 and Th2 cells (Fig. 5,D). One HpaII site, which coincided with DHSII, was less methylated in Th2 cells than Th1 cells (Fig. 5, B and D, ∗). Once again, the unmethylated CpG site coincided with DHS formation. As in the IL-4 gene, the fibroblasts were unmethylated at more CpG sites than were the T cells, and unmethylated sites did not correspond with DHS. Bisulfite sequencing analysis of DNA within CNS-1 showed that the CpG sites here were completely methylated in naive and Th1 cells and 90–95% methylated in Th2 cells (data not shown).
DHS and CpG methylation around IL-13
In resting Th2 cells, a DHS was observed at the transcription start site of IL-13 (Fig. 6,A, middle panel, and B, DHSIII). In Th1 cells, a very weak corresponding band was also observed. Again, GATA-3 binding sites are located within this DHS (Fig. 6,C). Two other DHS (I and II) were seen in Th2 cells in the CpG-rich region that lies ∼1 kb upstream of IL-13. DHS I is located within a CpG island and corresponds to a sequence that is conserved between the murine and human sequences (34). In fibroblasts, DHS FII coincided with Th2 DHS I in the CpG island, and another DHS, designated FI, was located 1 kb further upstream of IL-13, between the CpG island and an Alu repeat (Fig. 6 A, right panel, and B).
The 2.5-kb CpG-rich region located ∼1 kb upstream of IL-13 contains both a CpG island (as defined by Gardiner-Garden and Frommer in Ref. 35) and an Alu repeat (Fig. 6 B). Initially, HpaII digestion analysis of the CpG methylation status around IL-13 was attempted, but the many fragments generated by the abundant HpaII sites in the CpG-rich region complicated analysis of the data. Consequently, bisulfite modification and sequencing of the CpG-rich region was performed, which confirmed that, in all four cell types used in this study, all the CpG sites were unmethylated in the CpG island but methylated in the Alu repeat (data not shown). The methylation status of the CpG sites between the CpG island and the Alu repeat was variable.
The discovery of the Th2-specific DHS III prompted us to investigate the methylation status of the CpG sites around the transcriptional start site of IL-13 in T cells and in fibroblasts. The region that was amplified contains seven CpG sites (Fig. 7 A).
DNA from Th2 cells was largely unmethylated, unlike the IL-4 promoter: 87% of Th2-derived fragments were unmethylated in at least one CpG site, compared with 53% of Th1-derived DNA fragments and only 27% of naive cell-derived fragments. Demethylation in the Th2 cells occurred mainly at CpG sites 4–7 (Fig. 7, B and D), which are found in the IL-13 proximal promoter. Moreover, all four sites were demethylated in half the cloned DNA fragments. This finding is analogous to that of Fitzpatrick and colleagues (21, 22), where bisulfite sequencing of the IFN-γ promoter in CD8+ cells showed that demethylation at CpG sites was more likely to occur upstream of the transcriptional start site. CpG demethylation and DHS formation at the transcriptional start site of IL-13 thus coincide with the ability to express this gene in T cells. Direct sequencing of the fibroblast BSM-PCR products revealed that these CpG sites are predominantly methylated in fibroblasts (Fig. 7,C); this was confirmed by sequencing cloned individual DNA fragments (Fig. 7 B).
We used an in vitro differentiation system to investigate the epigenetic and chromatin structure changes around human IL-4 and IL-13 that accompany Th1 and Th2 differentiation from naive precursors. After Th2 differentiation, CpG demethylation coincides with DHS formation at consensus GATA binding sites and is not the locus-wide event reported in the murine locus (25). The presence of DHS and high levels of CpG demethylation in IL-4 and around CNS-1 in nonexpressing fibroblasts is also of interest.
We discovered several Th2-specific DHS around IL-4 and IL-13 in resting human Th2 cells after 14 days of differentiation (Fig. 8A). In IL-4, Th2-specific DHS appeared in the second intron and were absent in the promoter. Several GATA-3 consensus binding sites are present at DHS I and II, and Th2-specific CpG demethylation in the IL-4 gene was observed only in this region (Fig. 2). Although there is a high level of conservation between the human and murine DNA sequences in this region, it is interesting that not all the GATA-3 sites are conserved between the species: two of the consensus binding sites are present only in the human sequence (Fig. 2,E). It is likely that IL-4 DHS I and II define an enhancer in human Th2 cells. Similar Th2-specific DHS have been observed in the second intron of the murine IL-4 gene (Fig. 8 B, HSII) (16) in a region found to be a weak IL-4 enhancer in transgenic mice (36). These DHS and others around IL-13 could be induced by ectopic GATA-3 expression in murine Th1 cells (12, 37). Studies in murine mast cells also identified a mast cell-specific IL-4 DHS with enhancer function (38). This enhancer is activated by the mast cell-specific transcription factors GATA-1, GATA-2, and PU.1 (39) and it plays a role in establishing and maintaining IL-4 CpG demethylation in IL-4-expressing mast cells, but only in the presence of intact GATA binding sites (40). The connection between DHS formation and regulatory elements containing GATA sites was suggested after GATA-1 was shown to break DNA/histone contacts when bound to nucleosomal DNA (41). GATA family members associate with CREB-binding protein, and CREB-binding protein is an acetyltransferase that acetylates not only histones (acetylated histones are a characteristic feature of transcriptionally active chromatin) but also GATA proteins (42, 43, 44, 45). The fact that GATA sites were required to establish and maintain CpG demethylation in murine mast cells (40) may provide a link between the consensus GATA binding sites, DHS formation, and CpG demethylation that we observed in human Th2 cells.
In the human IL-13 gene, the Th2-specific DHS I and III coincided with regions of sequence that are conserved between mice and humans (Fig. 6,C and Ref. 34). DHSIII is located at the IL-13 promoter, where three potential GATA-3 binding sites are located (Fig. 6). In murine Th2 cells, these sequences, which are conserved in humans, were required for cell-specific, GATA-3-driven expression of IL-13 in overexpression studies (34). We also observed that the three CpG sites (Fig. 7, labeled 5, 6 and 7) located near the GATA-3 binding sites were generally unmethylated in the human Th2 cells, but that DHS or CpG demethylation of the IL-13 promoter were not seen in fibroblasts. This suggests that CpG demethylation and DHS formation around the GATA-3 sites at the IL-13 promoter are required for IL-13 expression. Whether CpG demethylation plays a direct role in IL-4/IL-13 transcriptional regulation or is merely a consequence of GATA-3-driven DHS formation requires further functional studies.
We also observed fibroblast-specific DHS that were distinct from Th2-specific DHS. From this we infer that genes that are not expressed in a particular tissue are not necessarily always free of DHS. Indeed, the presence of DHS in chromatin that is otherwise DNase I insensitive has been reported previously (46). We have not yet performed functional studies on any of the DNase I-hypersensitive elements that were identified in either Th2 cells or fibroblasts, but it is possible, given the transcriptional silence of the IL-4 in fibroblasts, that the fibroblast-specific DHS in the second intron of IL-4 may indicate a silencer or an insulator. Many intron-located silencers have been identified (47), including one in the first intron of the CD4 gene, which represses promoter activity in CD4+ T cells but not in CD8+ T cells (48, 49). Alternately, the presence of DHS in fibroblasts may suggest a functionally defined chromatin structure that may be relevant to the maintenance of the expression of the other genes (e.g., RAD50 or KIF3A) in the cluster. The absence of the Th2-specific transcription factor GATA-3 suggests a different mechanism of DHS formation in this nonhemopoietic lineage.
The CpG sites in the 5′ half of the fibroblast IL-4 gene were also largely demethylated (Fig. 3 A). The contrast between these findings and those in murine fibroblast cell lines (16), which showed that the DNA across the IL-4, IL-5, and IL-13 genes was methylated in this cell type, may be because we used nontransformed, nonimmortalized fibroblasts from normal human skin: immortalized cell lines have been shown to be hypermethylated at nonessential genes (50). The presence of unmethylated CpG sites in IL-4 in a differentiated cell type that does not express the gene (i.e., fibroblasts) suggests that another, tissue-specific process maintains its suppression. Alternatively, perhaps CpG methylation in a tissue that lacks Th2-specific factors may not be relevant to the regulation of this gene. It might thus be possible that CpG methylation in IL-4 and IL-13 could be a mechanism of suppression in potentially permissive lineages like naive and Th1 cells.
In conclusion, these data confirm that chromatin structure in this locus differs between human naive and Th1 cells, and Th2 cells. The appearance of Th2-specific DHS coincides with Th2-specific CpG demethylation (Figs. 2, B and D, 5, B and D, and 7, ∗), and both DHS formation and CpG demethylation occur near consensus GATA binding sites in IL-4 and IL-13. CpG sites that are not in the vicinity of DHS remain methylated after differentiation to the Th2 phenotype. It will be interesting to see whether the recruitment of chromatin-modifying activities by GATA-3 is responsible for the DHS formation and CpG demethylation that we observed. Meanwhile, different patterns of DHS and unmethylated CpG sites are seen in the fibroblast IL-4/IL-13 cluster and the link between DHS formation and CpG demethylation is absent, which suggests that chromatin structure and DNA methylation play different roles in IL-4/IL-13 regulation in nonhemopoietic cells from those in lymphocytes.
We are grateful to Ian Kirby and John Newell-Price for technical assistance, to Tak Lee for support and encouragement, to Paul Lavender, Joan Boyes, and Patrick Varga-Weisz for critical review of the manuscript, and to Richard Meehan and Kasia Hawrylowicz for helpful discussions.
This work was supported by Medical Research Council Grant G9815922.
Abbreviations used in this paper: CNS-1, conserved noncoding sequence-1; DHS, DNase I hypersensitive site; BSM, bisulfite-modified.