and Immunity (2010) 11, 554–560 & 2010 Macmillan Publishers Limited All rights reserved 1466-4879/10 www.nature.com/gene

ORIGINAL ARTICLE DNA methylome in human CD4 þ T cells identifies transcriptionally repressive and non-repressive methylation peaks

T Hughes1,4, R Webb1,4, Y Fei1,2,JDWren1 and AH Sawalha1,2,3 1Arthritis & Immunology Program, Oklahoma Medical Research Foundation, Oklahoma City, OK, USA; 2Department of Medicine, University of Oklahoma Health Sciences Center, Oklahoma City, OK, USA and 3US Department of Veterans Affairs Medical Center, Oklahoma City, OK, USA

DNA methylation is an epigenetic mark that is critical in determining chromatin accessibility and regulating expression. This epigenetic mechanism has an important role in T-cell function. We used genome-wide methylation profiling to characterize the DNA methylome in primary human CD4 þ T cells. We found that only 5% of CpG islands are methylated in CD4 þ T cells, and that DNA methylation peak density is increased in subtelomeric chromosomal regions. We also found an inverse relationship between methylation peak density and chromosomal length. Our data indicate that DNA methylation in gene promoter regions is not always a repressive epigenetic mark. Indeed, about 27% of methylated genes are actively expressed in CD4 þ T cells. We demonstrate that repressive methylation peaks are located closer to the transcription start site (TSS) compared with functionally non-repressive peaks (À893±110 bp versus À1342±218 bp (mean±s.e.m.), P-value o0.05). We also show that both a larger number and an increased CpG island density in promoter sequences predict transcriptional permissiveness of DNA methylation. TSS in the majority of genes with permissive DNA methylation peaks is in DNase I hypersensitive sites, indicating a failure of DNA methylation to induce chromatin inaccessibility in these loci. Genes and Immunity (2010) 11, 554–560; doi:10.1038/gene.2010.24; published online 13 May 2010

Keywords: CD4 þ T cell; DNA methylation; CpG islands; promoter methylation; methylome

Introduction regulation of interferon-g, and interleukin (IL)-4 produc- tion, key cytokines produced by Th1 and Th2 cells, DNA methylation is an epigenetic mark that is critical respectively. DNA demethylation of the FOXP3 locus is in determining chromatin accessibility and regulating pivotal for regulatory T-cell differentiation, and demethyla- . DNA methylation, which refers to the tion of the IL-2 locus is associated with IL-2 production addition of a methyl group to the fifth carbon in cytosine on T-cell activation.8 Defective T-cell DNA methylation residues within CG dinucleotides, is involved in cell results in T-cell autoreactivity and has an important differentiation, imprinting, X- inactivation, pathogenic role in both drug-induced and idiopathic and suppression of transcriptional noise and ‘parasitic’ lupus, both in human disease and in animal models.7,9,10 DNA.1–4 Abnormalities in the DNA methylation path- Herein, we characterize the DNA methylome in way are associated with pathological consequences. primary human CD4 þ T cells. We map DNA methyla- For example, mutations in the de novo DNA methyl- tion peaks across the genome, and identify genes with transferase DNMT3B result in a syndrome of Immuno- promoter region methylation in CD4 þ T cells using five deficiency, Centromeric instability and Facial anomalies biological replicates. We further identify distinguishing (ICF syndrome).5 A complete deficiency of the DNA features between transcriptionally repressive and non- methyltransferase DNMT1 is incompatible with life. repressive DNA methylation in CD4 þ T cells. Furthermore, acquired abnormalities in DNA methyla- tion are associated with disease conditions, including and autoimmunity.6,7 Results DNA methylation has a critical role in normal T-cell function such as T helper cell differentiation and the We determined genome-wide DNA methylation peaks in primary human CD4 þ T cells using DNA immuno- precipitation (IP) with an anti-5-methylcytidine antibody Correspondence: Dr AH Sawalha, Arthritis & Immunology Pro- coupled with array hybridization. Both input and IP gram, Oklahoma Medical Research Foundation, 825 NE 13th Street, DNA were labeled and co-hybridized to microarray MS#24, Oklahoma City, OK 73104, USA. chips that included B385 000 probes covering all UCSC- E-mail: [email protected] 4These authors contributed equally to this work. annotated CpG islands and promoter regions for all Received 6 January 2010; revised and accepted 2 March 2010; RefSeq genes (NimbleGen, Reykjavı´k, Iceland). The published online 13 May 2010 experiments were carried out using five biological Human CD4 þ T cell DNA methylome T Hughes et al 555 replicates from five normal healthy female donors (age We identified 2902±187 (mean±s.e.m., n ¼ 5) methy- range from 31 to 48 years). Signal intensity data were lation peaks in CD4 þ T-cell DNA. Further, we identified extracted from the scanned images of each array. Scaled 388 genes that have at least one DNA methylation peak log2-ratios of the IP/input DNA were determined from that appears in the À5to þ 1 kb region relative to the signal intensities, and P-values for methylation enrichment transcription start site (TSS) with its center located within were computed using the one-sided Kolmogorov–Smirnov the À5.5 and þ 1.5 kb region in all the five biological test. Methylation peaks were determined. They represent replicates tested. This stringent requirement that all regions with at least two probes with Àlog10 P-values of at genes should be identified in every sample tested has least 2 within a 500 bp window, and a methylation score of the advantage of adding confidence to the target genes at least 2. The methylation score for each peak is the identified near the methylation peaks. average Àlog10 P-values from probes within that peak. We used gene expression data in normal human Several known methylated genetic loci were included CD4 þ T cells (10 biological replicates) available from in our array for quality control. These included the Gene Expression Omnibus, to determine whether there is regions in the HOXA gene cluster, H19/IGF2/ any correlation between gene expression and methyla- KCNQ1 gene cluster and the IGF2R locus. All were tion status. Expression data were available for 202 genes methylated in all five biological replicates used in this with a methylated promoter region in CD4 þ T cells. study. Figure 1a shows the methylation status of the Only 55 genes (27.2%) had at least one transcript H19/IGF2/KCNQ1 gene cluster in our samples. The expressed in normal human CD4 þ T cells. The majority HOXA gene cluster serves both as a positive and of the methylated genes (72.8%) were not expressed. negative control region, as it contains both known Comparatively, when all annotated genes included in the methylated and hypomethylated regions.11 Our data expression array experiment were analyzed, we found confirm this methylation pattern in all five CD4 þ T-cell that 43.7% of genes (9094 out of 20 828 genes examined) DNA samples (Figure 1b). We further validated the were expressed in normal human CD4 þ T cells (w2 ¼ methylation array data in an independent set of 22.0, Po0.0001) (Figure 2a). These findings are consistent samples from another five normal healthy women (age with DNA methylation being largely a repressive range from 22 to 57 years) using bisulfite DNA epigenetic mark in human CD4 þ T cells. However, sequencing of both methylated and hypomethylated 27.2% of methylated genes are transcriptionally active, regions (Figure 1). indicating that DNA methylation is not always asso-

Chromosome 11 1,610,0001,680,000 1,750,000 1,820,000 1,890,000 1,960,000 2,030,000 2,100,0002,170,000 2,240,000 2,310,000 2,380,000 2,450,000 2,520,000 2,590,000 2,660,000 2,730,000 2,800,000 2,870,000 2,940,000 3,010,000 3,080,000 3,150,000 3,220,000

Genes HCCA2 LSP1 TNNT3 H19 IGF2 TSPAN32 CD81 TRPM5 KCNQ1 SLC22A18AS NAP1L4 CARS OSBPL5

8.000 6.400 4.000 Sample 1 2.400 0.000

8.000 6.400 Sample 2 4.000 2.400 0.000

8.000 6.400 4.000 Sample 3 2.400 0.000

8.000 6.400 Sample 4 4.000 2.400 0.000

8.000 6.400 Sample 5 4.000 2.400 0.000

1.000 CpG Islands 0.000

80 85 99 80 84 91 94 87 93 91 91 78 83 91 94 79 83 92 85 79 Chr11: 2,411,956-2,412,208 Methylation (%) 0 255075100

Figure 1 Methylation enrichment signals represented as Àlog10 P-value scores in CD4 þ T-cell DNA from five normal healthy participants in (a) the H19/IGF2/KCNQ1 genetic locus on , and (b) the HOXA gene cluster on chromosome 7. Bisulfite DNA sequencing was used to validate the methylation status in methylated regions within the HOXA3 gene and the KCNQ1/TRPM5 promoter region, and hypomethylated regions in the HOXA1 and HOXA13 promoter regions in an independent set of samples.

Genes and Immunity Human CD4 þ T cell DNA methylome T Hughes et al 556 Chromosome 7 27,072,000 27,080,000 27,088,000 27,096,000 27,104,000 27,112,000 27,120,000 27,128,000 27,136,000 27,144,000 27,152,000 27,160,000 27,168,000 27,176,000 27,184,000 27,192,000 27,200,000 27,208,000 27,216,000 27,224,000 27,232,000 27,240,000 27,248,000 27,256,000

HOX Genes A1 A2A3 A4 A5 A6 A7 A9 A10 A11 A13 EVX1

5.500 3.600 Sample 1 2.400 1.200 0.000

5.500 3.600 Sample 2 2.400 1.200 0.000

5.500 3.600 Sample 3 2.400 1.200 0.000

5.500 3.600 Sample 4 2.400 1.200 0.000

5.500 3.600 Sample 5 2.400 1.200 0.000

1.000 CpG Islands 0.000

1002010012211100 79 97 90 78 89 30 88 90 79 90 89 91 14 3 2 13 0 0 17 17 0 2 0 0000000000000000 88 99 95 95 97 85 95 96 90 93 97 92 13 4 10 16 0 15 3 3 0 0 0 0000000000000000 82 94 97 93 95 57 88 94 92 93 95 82 18 8 5 0 15 11 0 9 0 0 0 30000000000000000 85 95 98 84 93 57 88 91 94 93 94 93 504841404800 0000000000000000 74 87 94 86 84 46 91 93 84 94 94 95 14 5 12 18 3 0 9 0 0 0 0 Chr7: 27,101,779-27,102,082 Chr7: 27,129,376-27,129,708 Chr7: 27,206,717-27,207,948

Methylation (%) 0 25 50 75 100 Figure 1 Continued.

ciated with gene silencing in CD4 þ T cells. There was a CG observed/expected ratio within CpG islands in significant difference in the average distance between the promoter regions of methylated and expressed genes center of methylation peaks and TSS of methylated genes was 0.95±0.03 compared with 0.88±0.02 in methylated that are expressed compared with non-expressed genes. and non-expressed genes (mean±s.e.m., P ¼ 0.01). The center of methylation peaks was on an average We carried out functional analysis of genes that are 449 bp further upstream from the TSS in expressed genes methylated but expressed in CD4 þ T cells and in genes as compared with non-expressed genes (À1342±218 bp that are methylated and non-expressed using Ingenuity versus À893±110 bp (mean±s.e.m.), P-value o0.05). Pathway Analysis software (Ingenuity Systems, Red- These data suggest that a chromatin distance of three wood City, CA, USA). Interestingly, the top functional (449 bp divided by 147 bp/) is networks identified in genes that are expressed showed important in determining whether DNA methylation is involvement in basic cell functions, such as cellular transcriptionally repressive or permissive in a given signaling and interaction, and cellular growth and genetic locus. There was no difference in methylation proliferation (Figure 3a). The non-expressed genes

intensities (as measured by Àlog10 P-value methylation showed functional association with antigen presentation, scores) between transcriptionally repressive and permis- cell-mediated immune response and humoral immune sive methylation peaks (2.97±0.04 versus 2.98±0.03 response (Figure 3b). This suggests that the methylation (mean±s.e.m.), P-value ¼ 0.85). We determined the in non-expressed genes is functionally relevant and that number of CpG islands and the maximum CG dinucleo- those genes that are involved in immune functions are tide density in CpG islands within the À5.5 to þ 1.5 kb non-expressed in primary CD4 þ T cells and can region from the TSS of genes that are methylated and demethylate and become transcriptionally active once T expressed in CD4 þ T cells and genes that are methylated cells are activated and differentiated. but non-expressed. Promoter regions of expressed genes We next analyzed DNase I hypersensitive (HS) sites in were more likely to have a CpG island compared with primary human CD4 þ T cells, as a marker for chromatin non-expressed genes (83.3 versus 64.7%, odds ratio ¼ accessibility. This revealed that TSS is located within a 2.7, w2 ¼ 6.38, P-value ¼ 0.012). There was a significant DNase I HS site in the majority of methylated genes that difference in the mean CpG island CG dinucleotide are expressed but not in non-expressed genes (85.5 densities (as measured by maximum CG observed/ versus 27.9%, w2 ¼ 52.45, P-value o0.0001). This further expected ratio) in promoter regions of expressed com- indicates that DNA methylation is functionally relevant pared with non-expressed genes. The mean maximum in inducing chromatin inaccessibility at the TSS and

Genes and Immunity Human CD4 þ T cell DNA methylome T Hughes et al 557 We found a higher density of methylation peaks in the subtelomeric regions compared with the rest of the in CD4 þ T cells. The mean methylation peak density in the tiled region within the subtelomeric regions (7 Mb regions from the telomeres) was signifi- cantly higher compared with the non-subtelomeric regions in all chromosomes combined (Po0.0001). This enrichment toward the telomeres is not explained by tiling as methylation peaks densities were normalized for the size of the region tiled, and this phenomenon has been recently reported by others.11,14 Enrichment of methylation peaks in the subtelomeric regions was evident and statistically significant in all chromosomes except for chromosome 19 (Table 1). Unexpectedly, we also found a negative correlation between chromosomal lengths and the density of methylation peaks obser- ved within the tiled regions in our samples (r2 ¼ 0.34, P ¼ 0.0035) (Figure 2b). This is not explained by gene density as there is a positive correlation between the number of genes on each chromosome and chromosomal length (r2 ¼ 0.35, P ¼ 0.003).

Discussion DNA methylation is largely a transcriptionally repressive epigenetic mark that induces gene silencing and chro- matin inaccessibility.15,16 DNA methylation induces chromatin inaccessibility and transcriptional repression by several mechanisms. These include the recruitment of members of the methylcytosine binding domain- containing , such as MECP2, which is turn recruit deacetylases that result in chromatin condensation.15,16 In addition, the bulky methyl group on methylcytosine residues can prevent the binding of transcription factors to the promoter sequences of methylated genes.17 Using DNA methylation profiling Figure 2 (a) DNA methylation is significantly associated with in primary human CD4 þ T cells, we shed light on transcriptional repression in CD4 þ T cells (Po0.0001). About 27% transcriptionally permissive DNA methylation marks of genes with promoter region methylation escape transcriptional and find that B27% of methylated genes in primary repression and are actively expressed. (b) Methylation peak density human CD4 þ T cells are expressed. We find that most normalized to tiled regions (peak/Mb) negatively correlates with þ chromosomal length (r2 ¼ 0.34). CpG islands are not methylated in CD4 T cells, consistent with published work in a number of genetic regions in multiple cell types.18,19 transcriptional repression in the latter but not former We report two characteristics that distinguish trans- group of methylated genes. criptionally repressive from permissive DNA methyla- Out of all the 27 458 CpG islands examined within tion peaks in promoter gene sequences. A distance the 22 autosomal chromosomes and the X chromosome, between TSS and the center of methylation peaks of an we found that only 1375±65 (mean±s.e.m.) CpG average of 449 bp further upstream from the TSS islands (5% of islands) include methylation peaks in prevents DNA methylation from inducing chromatin our samples. This indicates that the vast majority of CpG inaccessibility and transcriptional silencing. Our data islands are unmethylated in CD4 þ T cells. This is suggest that methylation peaks that are transcriptionally consistent with recent studies in other cell types.11,12 repressive are on average about six nucleosomes away The regions in the HOXA gene cluster (Chr7: from the TSS. Methylation peaks located on average 26 924 046–27 424 045), the H19/IGF2/KCNQ1 cluster about nine nucleosomes away were not functionally (Chr11: 1 699 992–3 143 916) and the IGF2R gene region repressive. In addition, CpG islands in the promoter (Chr6: 160 309 320–160 447 571) were entirely tiled in our regions of genes that harbor a permissive DNA methyla- arrays. Data from these regions (B2 Mb titled region) tion peak are characterized by an increased maximal CG allowed unbiased examination of the DNA methylation dinucleotide density compared with repressive peaks pattern both in CpG islands and in surrounding genetic (Figure 4). Furthermore, we demonstrate that the regions included within these sequences. We found that majority of genes with a TSS located downstream of most methylation peaks are located outside the CpG permissive DNA methylation peaks are sensitive to islands. Indeed, in genetic loci close to CpG islands, DNase I digestion at the TSS, as indicated by the methylation peaks tend to occur at CpG ‘shores’, just presence of a DNase I HS site, indicating that chromatin outside the boundaries of CpG islands. This observation is accessible and available for transcription machinery has been recently reported in other human tissues.13 binding.

Genes and Immunity Human CD4 þ T cell DNA methylome T Hughes et al 558

Figure 3 Functional network analysis of methylated genes that are expressed in primary human CD4 þ T cells identified involvement in basic cell functions, including cellular signaling and interaction, and cellular growth and proliferation (a). Genes that are methylated and non- expressed are functionally associated with antigen presentation, cell-mediated immune response and humoral immune response (b). Gene key: solid lines indicate direct interaction, dotted lines indicate indirect interaction, an arrow from a to b indicates that a acts on b, a line without an arrowhead indicates binding only and a line with a small vertical line at the end from a to b indicates a inhibits b. Gray indicates genes that are methylated, white indicates genes that are not user specified but incorporated into the network through relationships with other genes. Node shapes are: square, cytokine; diamond (vertical), enzyme; diamond (horizontal), peptidase; dotted rectangular (vertical), ion channel; solid rectangular (vertical), G--coupled receptor; triangle, kinase; oval (horizontal), transcription regulator; oval (vertical), transmembrane receptor; trapezoid, transporter; circle, other.

Table 1 Methylation peak density (peak/Mb) in subtelomeric (up We also observe that transcriptionally permissive to 7 Mb from each telomere) and non-subtelomeric regions in each DNA methylation peaks correspond to genes involved chromosome in cell signaling, growth and proliferation, whereas repressive peaks are more closely associated with genes Chromosome Methylation peak density (peak/Mb) P-value associated with immune response and T-cell differentia- tion. Modulation of repressive DNA methylation Subtelomeric Non-subtelomeric patterns at key regulatory loci has a critical role in lineage differentiation of certain T helper subsets. For Mean s.e.m. Mean s.e.m. example, hypomethylation of a single region in the FOXP3 locus reveals Treg cell lineage commitment and 1 101.5 6.2 44.3 3.4 1.75E-06 is closely related to the longevity of suppressor 2 175.1 8.3 45.6 3.4 6.83E-09 function.20,21 3 62.1 9.0 31.4 1.6 0.002106 Our data indicate that proximity of methylation is 4 142.3 4.4 25.3 2.7 8.37E-11 directly related to the potential for methylation- 5 120.4 5.8 28.8 2.5 6.89E-09 dependent nucleosome remodeling and repression. 6 146.3 10.0 36.1 4.1 2.04E-07 7 188.9 9.2 46.7 4.0 8.62E-09 We suggest a model in which failure of DNA methyla- 8 167.1 9.1 35.5 0.8 6.84E-09 tion to induce inaccessible chromatin configuration is 9 101.2 7.1 63.9 5.1 0.000393 related to both the distance of the methylation peaks 10 189.9 10.0 50.2 5.4 3.39E-08 from TSS and perhaps reduced efficiency to recruit 11 121.6 10.2 39.2 2.9 2.55E-06 transcription repressor complexes, such as MECP2- 12 116.7 7.6 35.5 2.4 2.06E-07 SIN3A-HDAC, as a result of high CpG island density, 13 270.6 15.2 39.2 3.7 5.51E-09 14 87.1 6.5 38.6 2.8 7.93E-06 leading to an open chromatin configuration and avail- 15 65.9 4.4 38.2 2.4 4.86E-05 ability for transcription. 16 149.0 6.6 92.0 6.8 2.47E-05 17 113.1 10.3 55.6 3.3 6.66E-05 18 127.3 11.8 52.0 5.3 3.10E-05 19 111.7 8.7 89.8 9.7 0.066 Materials and methods 20 106.1 7.4 43.0 4.1 3.81E-06 CD4 þ T cell isolation and DNA extraction 21 164.6 6.5 74.6 10.0 3.35E-06 22 160.5 12.5 76.4 4.1 1.45E-05 Peripheral blood mononuclear cells were isolated X 132.8 9.9 69.0 7.3 8.47E-05 from normal healthy donor blood samples by density gradient centrifugation (Amersham Biosciences, Uppsala,

Genes and Immunity Human CD4 þ T cell DNA methylome T Hughes et al 559

HDAC SIN3A

MECP2

Closed chromatin No DNase I HS site No target gene expression

HDAC SIN3A

MECP2

Ac Ac Ac

Ac Open chromatin DNase I HS site Target gene expression Figure 4 A schematic representation showing distinguishing features between repressive and permissive DNA methylation. (a) Repressive methylation peaks are on average 893±110 bp upstream of the transcription start site of target genes, which are characterized by a relatively lower maximum CpG island density. Chromatin is closed at the transcription start site as indicated by the absence of DNase I HS sites. (b) Permissive methylation peaks are further upstream from transcription start site compared with repressive peaks, and are characterized by higher maximum CpG island densities within promoter sequences of target genes. These methylation peaks fail to maintain a closed chromatin configuration. This results in accessible chromatin at the transcription start sites of target genes as evidenced by the presence of DNase I HS sites and gene expression. It is possible that these methylation peaks fail to efficiently recruit transcriptional repressor complexes, such as the MECP2–SIN3A–HDAC complex. MECP2, methyl-CpG-binding protein 2; SIN3A, SIN3 homolog A; HDAC, histone deacetylase.

Sweden). CD4 þ T cells were then isolated through precipitation. We then performed whole genome ampli- magnetic bead separation using direct labeling (Miltenyi fication (WGA2 kit, Sigma-Aldrich, St Louis, MO, USA) Biotec, Auburn, CA, USA) following the manufacturer’s of input and IP DNA. Input and IP DNA from each protocol. DNA was extracted using the DNeasy Kit participant were then labeled with Cy3 and Cy5, (Qiagen, Valencia, CA, USA). The DNA concentration respectively, pooled, denatured and then co-hybridized was then determined using a NanoDrop spectropho- to 385K methylation arrays with tiling that covers all tometer (Thermo Sci., Wilmington, DE, USA). Our UCSC-annotated CpG islands and promoter regions for studies were carried out using ten biological replicates all RefSeq genes (NimbleGen). from ten normal healthy women. All participants signed an informed consent. Our studies are approved by our Bisulfite DNA sequencing institutional review boards. CD4 þ T-cell DNA from an independent set of normal healthy controls was isolated and treated with sodium DNA IP and array hybridization bisulfite using the EZ DNA Methylation-Gold kit (Zymo Genomic CD4 þ T-cell DNA was digested with MseI and Research, Orange, CA, USA). Sodium bisulfite treatment purified using Qiagen Quick PCR Purification kit. Next, will convert unmethylated cytosine residues to thymine, we diluted 1.25 mgofMseI-digested DNA to a final whereas methylated cytosine residues will remain as volume of 300 ml in TE buffer (pH7.5), heat denatured for cytosines. Sodium bisulfite treated DNA was amplified 10 min at 95 1C and then immediately cooled on ice for and directly sequenced (primer sequences are available 5 min. An aliquot of 60 ml was then removed and stored upon request). Percent methylation on each CG site was at À20 1C as the input DNA. IP buffer (60 mlat5Â ) was quantified using the Epigenetic Sequencing Methylation then added to the remaining 240 ml of the digested and analysis software (ESME).22 denatured DNA samples. Anti-5-methylcytidine anti- body (Abcam, Cambridge, MA, USA) was added and the Statistical analysis and methylation peaks identification samples were incubated overnight with agitation at 4 1C. Signal intensity data were obtained and analyzed from Pre-washed protein A agarose beads were then added each scanned array by NimbleGen. The ratio of IP versus and the samples were incubated at 4 1C with agitation for input DNA signals in each probe from each co-

3.5–4 h to obtain DNA–antibody–bead complex. This hybridized sample is determined and the log2-ratio is reaction mix was then spun down at 6000 r.p.m. for 2 min computed and scaled to center the ratio data around at 4 1C and the pellet washed three times with 1 ml 1 Â IP zero. Scaling is carried out by subtracting the bi-weight buffer with 5 min incubations with agitation in between mean for the log2-ratio values for all features on the array centrifugation steps. The beads were then resuspended from each log2-ratio value (complete scaled log2-ratio in 250 ml of digestion buffer and 7 ml of proteinase K data for all probes and all samples are available in online À1 (10 mg ml ) was added and incubated overnight at 55 1C Supplementary file 1). The Àlog10 P-values are calculated with agitation. IP DNA was then purified using ethanol for each probe by placing a fixed-length 750 bp window

Genes and Immunity Human CD4 þ T cell DNA methylome T Hughes et al 560 around each consecutive probe, and the one-sided 3 Mohandas T, Sparkes RS, Shapiro LJ. Reactivation of an Kolmogorov–Smirnov test is used to determine whether inactive human X chromosome: evidence for X inactivation by the probes are drawn from a significantly more positive DNA methylation. Science (New York, NY) 1981; 211: 393–396.

distribution of intensity log2-ratios compared to those in 4 Bird AP. Functions for DNA methylation in vertebrates. Cold the rest of the array. The resulting score for each probe is Spring Harb Symp Quant Biol 1993; 58: 281–285. 5 Hansen RS, Wijmenga C, Luo P, Stanek AM, Canfield TK, the Àlog10 P-value from the windowed Kolmogorov– Smirnov test around that probe. Methylation peaks are Weemaes CM et al. The DNMT3B DNA methyltransferase gene is mutated in the ICF immunodeficiency syndrome. Proc defined as regions with at least two probes with Àlog10 Natl Acad Sci USA 1999; 96: 14412–14417. P-values of at least 2 within 500 bp window, and a 6 Ballestar E, Esteller M. SnapShot: the human DNA methylome methylation score of at least 2. Peaks that are within in health and disease. Cell 2008; 135: 1144–1144 e1141. 500 bp apart are merged. The methylation score for each 7 Pan Y, Sawalha AH. Epigenetic regulation and the pathogenesis

peak represents the average Àlog10 P-value for the probes of systemic lupus erythematosus. Transl Res 2009; 153:4–10. within that methylation peak. Complete methylation peaks 8 Sawalha AH. Epigenetics and T-cell immunity. Autoimmunity data for the autosomal chromosomes and chromosome X 2008; 41: 245–252. are presented in online Supplementary file 2. 9 Sawalha AH, Jeffries M, Webb R, Lu Q, Gorelik G, Ray D et al. Defective T-cell ERK signaling induces interferon-regulated Microarray CD4 þ T-cell expression and DNase I gene expression and overexpression of methylation-sensitive hypersensitivity data genes similar to lupus patients. Genes Immun 2008; 9: 368–378. 10 Sawalha AH, Jeffries M. Defective DNA methylation and Gene expression profile for human CD4 þ T cells CD70 overexpression in CD4+ T cells in MRL/lpr lupus-prone obtained from 10 normal healthy participants was mice. Eur J Immunol 2007; 37: 1407–1413. extracted from Gene Expression Omnibus (GEO acces- 11 Rauch TA, Wu X, Zhong X, Riggs AD, Pfeifer GP. A human B sion: GSE4588). Expression profiling was performed cell methylome at 100- resolution. Proc Natl Acad Sci using the U133 Plus 2.0 Array (Affyme- USA 2009; 106: 671–678. trix, Santa Clara, CA, USA). Genes were considered 12 Shen L, Kondo Y, Guo Y, Zhang J, Zhang L, Ahmed S et al. expressed if the mean normalized signal value in all 10 Genome-wide profiling of DNA methylation reveals a class of samples for at least 1 transcript was above 64. DNase I normally methylated CpG island promoters. PLoS Genet 2007; HS sites in normal human CD4 þ T cells were extracted 3: 2023–2036. 23 13 Irizarry RA, Ladd-Acosta C, Wen B, Wu Z, Montano C, from the UCSC Genome Browser and Tables. Onyango P et al. The human colon cancer methylome shows similar hypo- and hypermethylation at conserved tissue- Bioinformatic analysis specific CpG island shores. Nat Genet 2009; 41: 178–186. The number and density of CpG islands were deter- 14 Lister R, Pelizzola M, Dowen RH, Hawkins RD, Hon G, mined algorithmically using the NCBI Build 36.1 of the Tonti-Filippini J et al. Human DNA methylomes at base human genome (HG18). CpG islands were defined as a resolution show widespread epigenomic differences. Nature stretch of DNA of at least 200 bp with a C þ G content of 2009; 462: 315–322. at least 50% and an observed/expected CG dinucleotide 15 Nan X, Ng HH, Johnson CA, Laherty CD, Turner BM, frequency of at least 0.6. Eisenman RN et al. Transcriptional repression by the methyl- CpG-binding protein MeCP2 involves a histone deacetylase complex. Nature 1998; 393: 386–389. Conflict of interest 16 Jones PL, Veenstra GJ, Wade PA, Vermaak D, Kass SU, Landsberger N et al. Methylated DNA and MeCP2 recruit The authors declare no conflict of interest. histone deacetylase to repress transcription. Nat Genet 1998; 19: 187–191. 17 Bird AP, Wolffe AP. Methylation-induced repression—belts, braces, and chromatin. Cell 1999; 99: 451–454. Acknowledgements 18 Eckhardt F, Lewin J, Cortese R, Rakyan VK, Attwood J, Burger This work was supported by the National Institute of M et al. DNA methylation profiling of human chromosomes 6, 20 and 22. Nat Genet 2006; 38: 1378–1385. Health Grant number P20-RR015577 from the National 19 Weber M, Hellmann I, Stadler MB, Ramos L, Paabo S, Center for Research Resources, Grant number Rebhan M et al. Distribution, silencing potential and evolu- R03AI076729 from the National Institute of Allergy and tionary impact of promoter DNA methylation in the human Infectious Diseases, the Oklahoma Rheumatic Disease genome. Nat Genet 2007; 39: 457–466. Research Core Centers; the Department of Veterans 20 Floess S, Freyer J, Siewert C, Baron U, Olek S, Polansky J et al. Affairs; the University of Oklahoma Health Sciences Epigenetic control of the foxp3 locus in regulatory T cells. PLoS Center; the Oklahoma Medical Research Foundation. The Biol 2007; 5: e38. authors thank Dr J Donald Capra, MD, for his critical 21 Janson PC, Winerdal ME, Marits P, Thorn M, Ohlsson R, review of this paper. Winqvist O. FOXP3 promoter demethylation reveals the committed Treg population in humans. PLoS One 2008; 3: e1612. 22 Lewin J, Schmitt AO, Adorjan P, Hildmann T, Piepenbrock C. References Quantitative DNA methylation analysis based on four-dye trace data from direct sequencing of PCR amplificates. 1 Reik W, Dean W, Walter J. Epigenetic reprogramming in Bioinformatics 2004; 20: 3005–3012. mammalian development. Science (New York, NY) 2001; 293: 23 Boyle AP, Davis S, Shulha HP, Meltzer P, Margulies EH, 1089–1093. Weng Z et al. High-resolution mapping and characterization 2 Li E, Beard C, Jaenisch R. Role for DNA methylation in of open chromatin across the genome. Cell 2008; 132: genomic imprinting. Nature 1993; 366: 362–365. 311–322.

Supplementary Information accompanies the paper on Genes and Immunity website (http://www.nature.com/gene)

Genes and Immunity