Imprinting of Skin/Inflammation Homing in CD4⁺ T Cells Is Controlled by DNA Methylation within the Fucosyltransferase 7 Gene

T cell recruitment to peripheral organs is controlled by adhesion molecules acting as address codes to the respective organ (1, 2). During first Ag encounter in tissue-draining lymph nodes, T cells upregulate the respective homing receptors that promote their re-entry into the priming organ. Activation in skin-draining lymph nodes leads to upregulation of E- and P-selectin ligands (E- and P-ligs) as skin-specific homing receptors, whereas during priming of T cells within the gut the integrin complex α₄β₇ is induced, enabling entry into mucosal sites (3).

To keep adaptive T cell responses efficient, effector memory T cells (TEM) memorize their site of priming and, as we described previously, CD4⁺ TEM and a fraction of in vitro–activated T effector cells express selectin ligands for extended periods of time (4). In humans, CD4⁺ T cells specific for skin Ags, such as HSV, express cutaneous leukocyte Ag (CLA)—the corresponding E- and P-ligs in humans—in the absence of overt viral reactivation (5). Similarly, Rota virus–specific T cells have been found to express α₄β₇ following immunization or infection (6). Although a long-term topographical memory of effector T cells against pathogens is thought to be beneficial, chronic organ-specific inflammatory reactions, as for instance in psoriasis or inflammatory bowel disease, might be perpetuated by such a memory. However, the mechanisms by which peripheral homing receptors, such as the skin-specific selectin ligands, are switched on permanently in CD4⁺ TEM are poorly defined.

E- and P-ligs are glycosylated epitopes generated by posttranslational modification of carrier proteins. In murine T cells, synthesis of functional E- and P-ligs depends on the induction of at least two glycosyltransferases, that is, fucosyltransferase VII, encoded by the gene fut7, and core 2 β1,6-N-acetyl-glycosaminyltransferase I, encoded by glycosaminyltransferase 1 (gcnt1) (7, 8). fut7 deficiency results in impaired E- and P-lig synthesis in CD4⁺ T cells whereas gcnt1 deficiency abrogates P-lig but not E-lig expression (7–9). E-ligs predominantly facilitate T cell homing into the skin, whereas P-ligs also promote recruitment into acute inflammation. Hence, fucosyltransferase VII is crucial for both skin- and inflammation-specific homing, whereas core 2 β1,6-N-acetyl-glycosaminyltransferase I is predominately involved in inflammation-specific homing. In accordance with their first-line defense function, granulocytes and monocytes constitutively express E- and P-ligs, ensuring their rapid recruitment during inflammation.

The inducible expression of fut7 and gcnt1 in T cells is controlled by cytokines and environmental factors, in particular vitamins. Thus, fut7 is induced by IL-2 (9) and TGF-β (10), whereas gcnt1 is induced by IL-12 (11). IL-4 and retinoic acid (RA), with the latter being produced by local dendritic cells within intestinal sites, are suppressors of E- and P-lig induction (12). Interestingly, also adult bone marrow progenitor cells tested as immunomodulatory therapeutics were found to downregulate fut7 (13).

Functional commitment of cells and development of stable lineages requires permanent transcriptional programs. This is often governed by epigenetic mechanisms involving DNA methylation, histone modification, and noncoding RNAs, all regulating the accessibility of genes to the transcriptional machinery (14). Previously, we showed that the induction of selectin ligands in CD4⁺ T cells requires cell division, which is considered a window for epigenetic changes. Moreover, artificial demethylation with 5-aza-deoxycytidine in T cell cultures increased P-lig expression, suggesting that DNA methylation controls selectin ligand expression in CD4⁺ T cells (15).

In this study, we report DNA demethylation within an alternative/minimal promoter region of the fut7 gene corresponding with high fut7 expression in skin/inflammation-seeking murine CD4⁺ T cells. RA abrogated the activation-induced demethylation in this region, which started at a CpG within a cAMP response element (CRE). Mutation or deletion of the CRE binding site or methylation of the minimal promoter abolished the transcriptional activity in reporter assays and resulted in impaired binding of CREB. By analysis of human T cell subsets we provide evidence for the conservation of the methylation-dependent control of fut7 across species. Furthermore, analysis of monocytes, expressing FUT7 constitutively, and hepatocytes, devoid of E- and P-lig expression, points to an additional level of methylation-dependent repression of FUT7 in these cells.

Naive CD4⁺ T cells were purified from spleens, peripheral lymph nodes, and mesenteric lymph nodes of BALB/c mice (Charles River Laboratories, Sulzfeld, Germany) by MACS as described (16).

Skin/inflammation- and gut-homing CD4⁺ TEM were isolated from at least 9-mo-old ex-breeder BALB/c mice by first depleting CD8⁺ T cells, macrophages, and B cells by MACS and subsequent FACS sorting (16). Cells from mice were acquired in accordance with institutional, state, and federal guidelines.

PBMCs were isolated from leukocyte filters of healthy female donors by density gradient centrifugation using lymphocyte separation medium (LSM 1077; PAA Laboratories). After erythrocyte lysis, CD4⁺ T lymphocytes were enriched using human CD4 MicroBeads (Miltenyi Biotec) and autoMACS sorting. The enriched population was stained for CD3 (UCHT1), CD4 (OKT4), CD127 (A019D5), CD25 (M-A251), CCR7 (GO43H7; all from BioLegend), and CD45RA (2H4LDH11LDB9; Beckmann Coulter) before sorting on a FACSAria (BD Bioscience) into naive T cell (TN; CD3⁺CD4⁺CD45RA⁺CCR7⁺) and TEM (CD3⁺CD4⁺CD45RA⁻CCR7⁻) populations, while gating out regulatory CD3⁺CD4⁺CD25^highCD127^low T cells. Purity of the sorted populations was >95%.

Monocytes were isolated from the blood of blood donors as described (17). In brief, blood cells were collected by leukapheresis in a Spectra cell separator (Gambro BCT, Lakewood, CO) followed by counterflow elutriation in the following order: platelets, lymphocytes, monocytes, and granulocytes.

Primary human hepatocytes were isolated from patients who underwent hepatectomy because of liver metastasis from healthy liver tissue at least 1 cm distant of the margin of the metastasis by a two-step EDTA/collagenase perfusion technique as described (18, 19).

All studies in humans were approved by the local Ethics Committee (Berlin, Regensburg, Dortmund) and all donors gave consent to the study.

TN were activated in complete RPMI 1640 (cRPMI) containing 10% FCS, IL-12 (5 ng/ml), IFN-γ (20 ng/ml), IL-2 (10 ng/ml), and anti–IL-4 mAb (5 μg/ml) at 1 × 10⁶ cells/ml on plate-bound anti-CD3 (145-2C11)/anti-CD28 (37.51) mAbs. All-trans RA (10 nM; Sigma-Aldrich) or 0.25 μM LE540, an RAR receptor antagonist (Wako Chemicals), was added from the start of culture. E- and P-lig expression was determined by FACS as described previously (16).

PCR-amplified constructs were cloned into the CG-free pCpGL basic plasmid (20) by restriction digest with SpeI and BglII (New England Biolabs, Frankfurt am Main, Germany). The extended −861/−163 conserved noncoding sequence (CNS) enhancer region (16), henceforth referred to as CNS, was cloned into the pCpGL-basic plasmid after inserting a new annealed oligonucleotide containing a new multiple cloning site with BamHI and SwaI RE sites, at the SpeI restriction site. The consensus CRE binding site was mutated by substituting two bases abolishing the CpG site (gacg → gTGg) (21) or deleted entirely by overlap-extension PCR. All constructs were verified by sequencing. pCpGL plasmid (1.5 pmol) and 50 ng of the pGL4.75 (hRluc/CMV) as internal control was transfected (Neon transfection system; Life Technologies) into 1.5 × 10⁶ T cells, activated under Th1 conditions for 2 d. Transfected cells were stimulated on anti-CD3/anti-CD28 for 24 h, removed from stimulus to rest, and expanded for a further 24 h. Then the reporter assay was conducted using the Dual-Luciferase assay kit (Promega) and an Orion L microplate luminometer (Berthold Technologies). Cloning and mutagenesis primer sequences are given in Supplemental Table I.

Plasmids were methylated using M.SssI CpG methyltransferase (New England Biolabs) as described (20). After in vitro methylation, plasmid DNA was extracted using NucleoSpin gel and a PCR clean-up kit (Macherey-Nagel). The efficiency of methylation was verified by digesting both methylated and unmethylated plasmid with the methylation-sensitive restriction enzyme AvaI. To generate plasmids with selective methylation of the minimal promoter, plasmids were methylated using M.SssI CpG methyltransferase or mock methylated, after which the methylated and mock-treated +561/+1523 (minimal promoter) insert was removed using the restriction enzymes BamHI and BglII. The insert was then ligated into an unmethylated plasmid already containing the CNS upstream of the luciferase gene. Primer sequences are given in Supplemental Table I.

Single-stranded DNA oligonucleotides of the wild-type (wt) or mutated CRE site of the murine minimal fut7 promoter and of the regulatory T cell–specific demethylated region (TSDR) from the foxp3 locus, with the latter used as control (22), were synthesized and biotinylated at the 5′ end (BioTeZ). The sequences were as follows: CRE wt, 5′-CAGGGCAAGTGCTGACGCTCCATCAGACTG-3′; CRE mutated, 5′-CAGGGCAAGTGCTGTGGCTCCATCAGACTG-3′; TSDR, 5′-TGCATCCGGCCGCCATGACGTCAATGGCAGAAAAATC-3′; and the respective antisense sequences. For methylated oligonucleotides, all cytosines in CpG motifs were substituted by 5-methyl-2-deoxycytidin during synthesis. After annealing, 200 pmol of oligonucleotides were coupled to 25 μl of prewashed streptavidin agarose beads (Sigma-Aldrich) for 1 h in 400 μl of binding buffer (60 mM KCl, 12 mM HEPES [pH 7.9], 4 mM Tris-HCl [pH 8], 0.5 mM EDTA, 5% glycerol) (23). Fifty micrograms of nuclear extract (nuclear extract kit; Active Motif) of 5 h PMA/ionomycin–stimulated RLM-11-1 cells was preincubated for 1 h at 4°C in 400 μl of binding buffer supplemented with 5 μg of salmon sperm DNA (Invitrogen), 1 mM DTT, and cOmplete protease inhibitors (Roche). Beads were washed twice with 1 ml of binding buffer for removal of unbound oligonucleotides and resuspended with the nuclear extract. After 3 h of incubation at 4°C, beads were washed four times with 1 ml of binding buffer and eluted in 25 μl of 2× SDS sample buffer at 95°C for 5 min. Eluates were analyzed by SDS-PAGE and subsequent immunodetection with the Odyssey Western blot system (LI-COR Biosciences). For detection of proteins, anti-Creb1 (48H2; Cell Signaling Technology) and anti-Ets1 (C-20; Santa Cruz Biotechnology) were used.

Total RNA isolation and quantitative RT-PCR of murine T cells were performed as described (16). Human samples were subjected to RNA sequencing. TruSeq stranded mRNA sequencing libraries were prepared with 500 ng of total RNA according to the manufacturer’s protocol (Illumina). The libraries were sequenced in 2 × 100 nt manner on the HiSeq 2000 platform (Illumina). Sequence reads were mapped to human genome (version hs37d5) (24) with TopHat2 (25) with very sensitive setting for Bowtie2 (26) and gencode.v19 annotation (27).

Standard bisulfite sequencing of all amplicons was performed in cooperation with Epiontis as described (28) and for murine “amplicon4/5” and human samples in cooperation with Prof. J. Walter and Dr. S. Tierling according to Gries et al. (29). Briefly, 300 ng of genomic DNA was bisulfite treated using the EZ DNA Methylation kit (Zymo Research). PCRs were performed using primers with a specific 3′ portion (Supplemental Table I) and a universal 5′ portion according to Illumina’s specifications. Amplicons were purified using Agencourt AMPure beads (Beckman Coulter), diluted, and pooled.

Deep sequencing of bisulfite-treated human DNA samples was performed on the Illumina MiSeq according to the manufacturer’s protocols aiming at 10,000 reads per amplicon. Reads were processed and aligned using the BiQ Analyzer HT software (30) setting the “maximal fraction of unrecognized sites” filter at 0.1.

Raw reads were quality and adapter trimmed with the Trim Galore! wrapper of cutadapt (31), quality controlled with FastQC (http://www.bioinformatics.bbsrc.ac.uk/projects/fastqc/), and mapped to the reference with GSNAP (32), after which a slightly extended Bis–single nucleotide polymorphism (SNP) pipeline (33) was applied. The reads were locally realigned with respect to SNPs and indels from dbSNP (v.138) (http://www.ncbi.nlm.nih.gov/SNP/), duplicates were marked with Picard tools (http://broadinstitute.github.io/picard/), overlapping regions between paired reads were removed with bamUtil (http://genome.sph.umich.edu/wiki/BamUtil), and the quality values were recalibrated before calling the methylation levels.

Raw data are publicly available under the guidelines of the International Human Epigenome Consortium (http://ihec-epigenomes.org/about/policies-and-guidelines/). High-level analysis data can be accessed via the German epigenome program DEEP data portal (http://deep.dkfz.de/#/home) and/or the International Human Epigenome Consortium data portal (http://epigenomesportal.ca/ihec/). Genome raw data have been deposited at the European Genome-phenome Archive (http://www.ebi.ac.uk/ega/), which is hosted at the European Bioinformatics Institute, under accession no. EGAS00001001567. To receive access to these controlled data (bisulfite sequencing data: naive T cells: Hf03_BlTN_Ct_WGBS.MCSv3.20150414.cpg.filtered.CG.bw;TEM:Hf03_BlEM_Ct_WGBS.MCSv3.20150414.cpg.filtered.CG.bw,hepatocytes:Hf01_LiHe_Ct1_WGBS_S.MCSv0.20140122.filtered.CG.bw,monocytes:Hm01_BIMo_Ct_WGBS_E.MCSv0.20140204.filtered.CG.bw; mRNA sequencing data: naive T cells: Hf03_BlTN_Ct_tRNA_M_1.tophat2.20150123.bai; Hf03_BlTN_Ct_tRNA_M_1.tophat2.20150123.bam;TEM:Hf03_BlEM_Ct_tRNA_M_1.tophat2.20150123.bai,Hf03_BlEM_Ct_tRNA_M_1.tophat2.20150123.bam,Hepatocytes:41_Hf01_LiHe_Ct_tRNA_M_1.tophat2.20140507.bai,Hf01_LiHe_Ct_tRNA_M_1.tophat2.20140507.bam,monocytes:Hm01_BlMo_Ct_tRNA_M_1.tophat2.20140508.bai,Hm01_BlMo_Ct_tRNA_M_1.tophat2.20140508.bam), applications can be addressed to the DEEP Data Access Committee (http://www.deutsches-epigenom-programm.de/data-access/).

The coordinates of mouse amplicons were transferred to human by local alignment to −10 kb upstream until the transcription-end-site region of human Fut7 using EMBOSS Water (PMID:10827456). The Fut7 locus was defined as the region covering the gene body and the amplicons. The homolog CRE site position was identified by the consensus sequence (34). Methylation levels of CpG for the negative sense strand were extracted and visualized as methylation boxes. Differentially methylated regions (DMRs) were identified by manual inspection of the methylation box plot, considering the methylation levels and statistical significance of the DMR (unpaired and paired Student t test).

Prism software (GraphPad Software, San Diego, CA) was used for statistical analysis.

To determine the impact of DNA methylation on fut7 expression in murine T cells, we compared DNA methylation across the fut7 locus in CD4⁺ TEM. Skin/inflammation-homing P-lig⁺α₄β₇⁻ and gut-homing P-lig⁻α₄β₇⁺ CD4⁺ TEM were sorted from at least 9-mo-old ex-breeder mice as shown in Fig. 1A. Corresponding to P-lig expression, skin/inflammation-specific CD4⁺ T cells expressed higher levels of fut7 mRNA compared with gut-homing CD4⁺ T cells (Fig. 1B). Bisulfite sequencing of seven amplicons covering the fut7 gene locus (Fig. 1C) showed differential methylation in an intragenic region covered by amplicons 3 and 4/5. Skin/inflammation-homing CD4⁺ TEM showed significantly less DNA methylation across amplicon 3 and amplicon 4/5 compared with gut-homing TEM in a subsequent analysis of independent biological samples (Fig. 1D). Similarly, we found a correlation between fut7 mRNA expression and CpG methylation within amplicon 3 in the T cell lines αβWT, 16.2.11, EL-4, and RLM11 (data not shown).

The region covered by amplicon 4/5 possesses minimal promoter activity and drives transcription of a short fut7 transcript that starts at exon 4 in murine CD4⁺ T cells (16). In contrast, the region covered by amplicon 1 serves as an enhancer, but it may also promote (low-level) transcription of long transcripts starting at exon 1 (16). To assess the impact of CpG methylation in different regions, in particular the minimal promoter region, on the transcriptional activity of fut7, we cloned the entire region containing the minimal promoter and the enhancer of fut7 (−1625/+1523) and two truncated variants of this region into the CG-free pCpGL plasmid (20). The truncated constructs −1625/+573 and +561/+1523 either contained the enhancer comprising a CNS or the minimal promoter, respectively. These constructs were in vitro methylated with M.SssI, a CpG methyltransferase, or mock methylated. Effective methylation was confirmed by protection from restriction digest by Aval, which cuts DNA in a methylation-sensitive manner. Subsequently, methylated and mock-methylated constructs were transfected into primary Th1 cells. The −1625/+1523 construct elicited the greatest reporter activity, which severely dropped when the construct was methylated (Fig. 2A). In contrast, the −1625/+573 construct elicited low reporter activity, which is compatible with low promoter activity of the CNS in Th1 cells, which was also affected by methylation. The construct containing the minimal promoter (+561/+1523) had very low activity in the absence of an enhancer, and methylation seemed to further impair this activity. To decipher specifically the impact of methylation within the minimal promoter in the presence of a functional enhancer, we cloned the +561/+1523 construct after in vitro methylation or mock methylation into the pCpGL plasmid already containing the (nonmethylated) CNS in an enhancer position. The construct containing the methylated minimal promoter elicited lower reporter activity compared with the mock-methylated construct (Fig. 2B), showing that methylation in this region indeed plays a role in controlling fut7 transcription.

As mentioned earlier, RA inhibits the induction of E- and P-ligs (12). Even the minute amounts of RA present within FCS-containing cRPMI reduce the induction of fut7, as addition of the RAR antagonists LE540 or LE135 to Th1 differentiation cultures resulted in enhanced induction of fut7 (data not shown) and suppression of α₄β₇ (12). To study the most divergent conditions resembling the RA-containing mesenteric lymph node and the RA-deprived environment of a peripheral lymph node, we added 10 nM RA or the RAR antagonist LE540 to Th1 polarizing cultures set up in cRPMI. Addition of RA completely abrogated the induction of fut7 mRNA expression in Th1 polarized cells on day 4 and in repetitively stimulated Th1 cells on day 9 of culture. In contrast, Th1 cells cultured in the presence of LE540 showed strong induction of fut7 mRNA expression (Fig. 3A). This corresponds to the more pronounced effect of RA on fut7 expression than on gcnt1 expression under Th1 conditions (35). Consequently, E-lig expression was more affected than P-lig expression by RA (Fig. 3A).

The presence of RA during activation strongly affected the DNA methylation pattern, in particular the region covered by amplicon 4/5 (Fig. 3B). In T cells activated in the presence of LE540, CpGs at position +844 and +940 (relative to transcription start site [TSS]), and to a lower extent CpGs at positions 1082 and 1127, within amplicon 4/5 showed significantly lower methylation already on day 4 after activation compared with TN. In contrast, methylation did not change significantly in T cells activated in the presence of RA. The demethylation in LE540-treated cells seemed to progress from day 4 to day 9 and spread to amplicon 3. TN were heavily methylated in amplicon 4/5 and less methylated in amplicon 3 (Fig. 3B). Comparing the overall mean percentage of methylation of the amplicons in TN and Th1 cells, activated either in the presence of RA or in the presence of LE540, showed that activation induces progressive demethylation within amplicon 4/5, which is prevented by RA (Fig. 3C).

As the early and drastic demethylation of amplicon 4/5 in activated T cells suggested a crucial role of this region in the induction and control of fut7 expression, we performed an in silico search for transcription factor binding sites in this region. Using MatInspector (36), we identified a CRE site spanning the CpG at position 844. Interestingly, in human T cells Tax-dependent transactivation of FUT7 by human T cell leukemia virus-1 was shown to involve a CRE site located upstream of the human FUT7 gene (34). To analyze the impact of the CRE site spanning the CpG 844 for transcriptional activation of the murine fut7 gene, we mutated the CRE site or deleted it entirely in our luciferase reporter vectors. Both mutant constructs showed significantly reduced activity (Fig. 4A), suggesting that binding of CREB, a transcription factor that has been shown to bind DNA in a methylation-dependent manner (37), might be indeed important for the activation of the murine fut7 gene. To analyze the capacity of this region for binding CREB, we performed a pull-down assay using oligonucleotides containing the CRE site of the minimal fut7 promotor. As shown in Fig. 4B, CREB, but not ETS, bound to the CRE site within the minimal fut7 promoter. Binding of CREB (and ETS) to the TSDR served as a binding control (38, 39). Mutation of the CRE site prevented binding of CREB to the CRE site within the minimal fut7 promoter (Fig. 4B). Moreover, when CpGs within the fut7 wt oligonucleotide were methylated, binding of CREB was also inhibited (Fig. 4B), showing the importance of methylation within this region for binding of CREB.

To determine whether methylation-dependent regulation of the fut7 gene is conserved in murine and human T cells, we analyzed the DNA methylation across the FUT7 gene in datasets generated from human TN (CD3⁺CD4⁺CD25^lowCD45RA^highCCR7^high) and CD4⁺ TEM (CD3⁺CD4⁺CD25^lowCD45RA^lowCCR7^low). In line with the enrichment of CLA-expressing cells among TEM, this population expressed higher levels of FUT7 mRNA compared with TN (Fig. 5A). This correlated with reduced DNA methylation within three regions designated as DMRs 1, 2, and 3 on the human FUT7 locus (Fig. 5). Compared to the mouse gene, DMR1 corresponds in part to the amplicon 3 region, which showed selective demethylation in skin/inflammation-homing murine T cells. DMR2 is located partially in the first intron and spans a region that also contains a CRE site and that was reported to exert promoter activity in human Jurkat cells (34). This corresponds to the minimal promoter activity of the respective mouse region (amplicon 4/5). The CRE site also showed a 25% reduction of methylation levels in TEM (TN, 80%; TEM, 60%; coverage, 10), although the levels were higher in comparison with the mean methylation levels of the complete DMR2. To further verify the hypothesis that DNA methylation controls FUT7 expression, we analyzed the methylation of the FUT7 locus in other human cells types that either express FUT7 and selectin ligands constitutively, that is, ex vivo isolated blood monocytes, or that lack expression entirely, that is, primary human hepatocytes (Fig. 5A). Supporting our hypothesis, monocytes showed a demethylated state in most of the noncoding and upstream regions of FUT7 whereas the locus was completely methylated in hepatocytes. Interestingly, even the region corresponding to the mouse CNS (amplicons 1 and 2) was completely methylated in hepatocytes, suggesting that repression of transcription is fixed in these cells, whereas TN, which are not methylated within this region, are poised for inducible expression. Moreover, monocytes seem to express a longer transcript of FUT7 starting upstream of the known human exon 1 (Fig. 5A).

In this study, we show that the establishment of the skin/inflammation-seeking phenotype in murine CD4⁺ TEM depends on DNA demethylation of a minimal promoter of the fut7 gene. The region being heavily methylated in CD4⁺ TN becomes demethylated upon TCR-induced activation and proliferation. Moreover, RA, known to inhibit fut7 induction (12), prevented demethylation of the fut7 minimal promoter. The CpG undergoing strongest demethylation is part of a CRE binding site. We showed binding of CREB to this region in vitro and proved its crucial role for transactivation of the minimal promoter in reporter assays.

Previous studies suggested that binding of CREB to CRE is regulated by DNA methylation (37). For instance, methylation of the central CpG motif in a CRE palindrome promotes silencing of the EBV genome, and similar effects are reported for the regulation of the Pdha-2 gene in the testis (40, 41). In line with these studies, we found that the substitution of cytosine by 5-methylcytosine (5mC) within the CRE site of the fut7 minimal promoter impaired binding of CREB. Furthermore, using selective in vitro methylation, we show that the activity of the minimal fut7 promoter is impaired by methylation. Further evidence for the critical role of DNA methylation in control of selectin ligand expression comes from our previous findings that showed 5-aza-deoxycytidine, a DNA methylation inhibitor, enhancing the expression of P-lig in in vitro cultures (15). However, it remains to be elucidated whether demethylation in the fut7 locus in T cells relies on inhibition of methylation on the newly generated DNA strain during cell division mediated by DNA methyltransferases or on active demethylation involving ten–eleven translocation proteins, which convert 5mC to 5-hydroxymethylcytosine. Both mechanisms are known to be involved in thymic and/or peripheral T cell differentiation (42, 43). However, the bisulfite method used in our study does not distinguish between 5-hydroxymethylcytosine and 5mC (44); therefore, further studies are required to solve the question as to which mechanism mediates TCR activation-induced demethylation within the fut7 minimal promoter.

Regulation of fut7 transcription by CREB seems to be conserved between mice and humans in CD4⁺ T cells. In humans, malignant T cells in adult T cell leukemia, a fatal malignancy of Th cells caused by human T cell leukemia virus infection, express high levels of CLA (45). In transformed T cells, expression of FUT7 correlates with expression of the viral protein Tax, which is able to bind to CREB (34, 45). Binding of Tax to CREB enhances the binding activity of CREB to CRE (34). Later, it was shown that Tax-dependent activation of CREB promotes its binding to a CRE located about 1 kb upstream of the known TSS of the human FUT7 gene (34). The identification of the CRE site in the murine minimal promoter and confirmation of its crucial impact on transcriptional activity indicates a strong similarity between species in fut7 regulation in T cells. The profound effect of RA, that is, the prevention of demethylation and subsequent suppression of fut7 expression, may explain some of the efficacy of RA treatment in patients with adult T cell leukemias with cutaneous involvement reported by Maeda et al. (46). The conserved regulation of fut7 gene activity is further supported by our data analyzing the CpG methylation across the human FUT7 gene. In the present study, we also found selective demethylation of a region adjacent to the CRE site in TEM compared with TN.

This means that not only effector functions are fixed by epigenetic mechanisms (47), but also the homing commitment for different organs. This seems particularly important for CD4⁺ TEM, which constitute a migratory pool of cells, whereas CD8⁺ memory T cells rather build up a pool of tissue-resident cells, for instance in the skin (48), suggesting cooperation of both subsets in secondary responses by different means. The expression of selectin ligands, which is particularly stable on in vivo–generated TEM (4), can support the proinflammatory function of CD4⁺ TEM in cutaneous inflammation (49). Alternatively, fut7 expression is also crucial in skin-specific T regulatory cells as they maintain skin tolerance and therefore require selectin-dependent entry into skin sites for in vivo suppression of peripheral inflammation (50–52). Thus, not only providing immune surveillance but also immune tolerance within the skin, Ag-experienced CD4⁺ effector cells and regulatory T cells must express selectin ligands even in the absence of continuous TCR stimulation. This seems to be controlled by DNA demethylation of the minimal promoter region within the fut7 gene in skin/inflammation-homing T cells, resulting in a long-term topographic commitment.

Interestingly, methylation-dependent control of fut7 apparently operates at multiple levels. Human hepatocytes lacking CLA and FUT7 mRNA expression showed a completely methylated FUT7 locus including heavy methylation at a site corresponding to the murine amplicon 1/2 (Fig. 5), which, in the mouse T cells, constitutes a region with enhancer activity (16). In that previous study, where we also analyzed the histone modification across the murine fut7 locus, we found occupation of this region by a repressive H3K27me3 mark in murine lung fibroblasts. In hematopoietic lineages, including the FUT7 nonexpressing TN, this region was demethylated and, according to our data on the histone modification pattern, decorated with an active histone mark (16). Taken together, this may indicate that a poised state is required for induction of FUT7 expression in TN and hematopoietic cells in general. Induced expression of FUT7 is then stabilized by progressive demethylation of selected additional regulatory elements (including the minimal promoter region), as observed in TEM showing partial demethylation corresponding to the known subfraction of TEM-expressing FUT7. Ex vivo monocytes showed the most widespread demethylation within the FUT7 gene correlating to a constitutive expression of CLA (53). Interestingly, monocytes seem to start transcription further upstream, suggesting that the CNS, which acts as an enhancer in T cells, may drive expression of a monocyte-specific transcript, possibly even independently of the minimal promoter.

In conclusion, our data provide a mechanistic understanding of the skin- and inflammation-seeking homing phenotype and its establishment by epigenetic mechanisms in CD4⁺ memory T cells, and conceivably other migratory cells such as monocytes.

The CpG-free pCpGL plasmid was provided by Dr. Michael Rehli (University Hospital Regensburg). We thank the FACS sorting unit of the German Rheumatism Research Center for assistance in cell sorting. We thank Dr. Georg Damm, Dr. Andreas Nüssler, Dr. Thomas Weiss, Dr. Wolfgang Thasler, and Dr. Jan G. Hengstler for support with primary hepatocytes, and Dr. Nina and Dr. Gilles Gasparoni for help of bisulfite sequencing analyses of human cells.

S.O. is an employee of Epiontis, a company providing epigenetic analyses. The other authors have no financial conflicts of interest.