Gut memories do not fade: epigenetic regulation of lasting gut homing receptor expression in CD4⁺ memory T cells

Abstract

The concept of a “topographical memory” in lymphocytes implies a stable expression of homing receptors mediating trafficking of lymphocytes back to the tissue of initial activation. However, a significant plasticity of the gut-homing receptor α₄β₇ was found in CD8⁺ T cells, questioning the concept. We now demonstrate that α₄β₇ expression in murine CD4⁺ memory T cells is, in contrast, imprinted and remains stable in the absence of the inducing factor retinoic acid (RA) or other stimuli from mucosal environments. Repetitive rounds of RA treatment enhanced the stability of de novo induced α₄β₇. A novel enhancer element in the murine Itga4 locus was identified that showed, correlating to stability, selective DNA demethylation in mucosa-seeking memory cells and methylation-dependent transcriptional activity in a reporter gene assay. This implies that epigenetic mechanisms contribute to the stabilization of α₄β₇ expression. Analogous DNA methylation patterns could be observed in the human ITGA4 locus, suggesting that its epigenetic regulation is conserved between mice and men. These data prove that mucosa-specific homing mediated by α₄β₇ is imprinted in CD4⁺ memory T cells, reinstating the validity of the concept of “topographical memory” for mucosal tissues, and imply a critical role of epigenetic mechanisms.

Introduction

Long-term immunity relies on memory cells providing protection against recurring infections. The shaping of functional features of memory cells during primary reactions includes the imprinting of migration properties that secure their optimal targeting to potentially affected compartments. This topographical memory, described as organ-specific homing more than 50 years ago,¹ was a well-accepted paradigm that appeared to be supported by a number of studies on gut- and skin-specific homing of T memory cells (T_mem) generated in site-specific immune responses.^{2, 3, 4}

Homing of T cells into distinct tissues depends on the expression of adhesion molecules and chemokine receptors. Naive CD4⁺ T cells (T_N) uniformly recirculate through secondary lymphoid organs via L-selectin and CCR7, a behavior preserved by early stages of T_mem. However, upon further differentiation, T cells home preferentially into peripheral tissues such as the skin and gut or into inflamed sites. Eventually, part of the cells will become tissue-resident cells that stop circulating.^{5, 6} In this study, we investigated the stability and molecular regulation of integrin α₄β₇-dependent mucosa-specific homing of CD4⁺ T_mem from lymphoid tissues that largely belong to the recirculating pool of CD4⁺ T_mem.

T cells homing to skin express E- and P-selectin ligands and the chemokine receptor CCR4 and/or CCR8. By contrast, gut-homing T_mem selectively express the integrin α₄β₇ and partially express the chemokine receptor CCR9, enabling their entry into the intestine and associated lymphoid tissues by interacting with the mucosal addressin cell adhesion molecule-1 (MAdCAM-1) and the chemokine CCL25, respectively.^{7, 8, 9} Expression of these receptors is de novo induced or upregulated upon antigen-induced activation within mesenteric lymph nodes (mLNs) or Peyer’s patches.¹⁰ The vitamin A metabolite retinoic acid (RA), provided by CD103⁺ dendritic cells (DCs) and/or stromal cells in mLNs,^{11, 12} has been identified as a crucial factor together with T-cell activation for induction of α₄β₇ and CCR9.¹³

Notably, the assumption of the homing paradigm that T_mem cells acquire a stable homing phenotype by imprinted expression of a given set of adhesion receptors in CD4⁺ T_mem only has been proven experimentally for selectin ligands so far.^{14, 15} In fact, recent studies in CD8⁺ T cells have rather demonstrated that the mucosal homing receptor α₄β₇ is only transiently expressed on CD8⁺ T cells and that α₄β₇⁺ gut-homing CD8⁺ T cells might convert into a skin-homing phenotype.^{16, 17, 18} This implied plasticity of migratory phenotypes and the possibility that certain homing phenotypes require continuous stimulation by environmental signals to be maintained.

In the first part of this study, we therefore investigated whether CD4⁺ T_mem retain expression of the gut-homing receptor α₄β₇ in vivo or upon restimulation in vitro, or whether continuous instruction by tissue-localized factors is required for the maintenance of gut-homing properties. Indeed, ex vivo isolated α₄β₇⁺ CD4⁺ T_mem maintained their initial α₄β₇ expression level independent of the presence of RA or other environmental factors encountered during recirculation through gut or gut-associated lymphoid tissues (GALTs). To acquire a stable phenotype, repeated stimulation of T_N cells in the presence of RA was required, suggesting a stepwise acquisition of an imprinted state.

In a second part, we analyzed the molecular basis for the imprinted phenotype. The regulation of α₄-integrins is not well understood, despite significant medical relevance of α₄-integrins as a target for antibody therapies in chronic inflammatory bowel disease and multiple sclerosis.

Here we focused on the role of epigenetic mechanisms, as a growing list of evidence suggests that DNA methylation or histone modifications might be the key for a stably imprinted expression pattern in polarized T-helper (Th) subsets, stable lineages such as regulatory T cells, expression of signature cytokines, or master transcription factors.^{19, 20, 21, 22} Albeit it appears suggestive, homing receptors have hardly been analyzed in this respect so far.

While some histone modifications appear to be solely associated with current transcriptional activity, notably DNA methylation and other repressive mechanisms are key candidates for the regulation of a long-term, heritable phenotype.²³ We therefore analyzed the DNA methylation patterns of the Itga4 locus—coding for the integrin α₄ chain—in murine and human T cell subsets. Indeed, differential DNA methylation was observed in distinct regulatory regions and appears to control the stability of Itga4 expression.

Results

Stable expression of α₄β₇ on murine CD4⁺ memory T cells in vivo

First, the stability of α₄β₇ expression on murine CD4⁺ T_mem was analyzed in vivo. For this, we labeled ex vivo isolated, highly pure α₄β₇⁺ and α₄β₇⁻ CD45RB^low CD4⁺ T_mem subsets (Supplementary Figure 1A online) with CFSE, transferred them into recipient mice and reanalyzed α₄β₇ expression on transferred cells 15 days (Figure 1a and b) or 6 weeks later (Figure 1c). The majority of α₄β₇⁺ CD4⁺ T_mem maintained high α₄β₇ expression for at least 6 weeks, whereas transferred α₄β₇⁻ CD4⁺ T_mem remained negative for α₄β₇. These data show that the expression of α₄β₇ in CD4⁺ T_mem is stable in vivo. Interestingly, some α₄β₇⁺ CD4⁺ T_mem were also recovered from the subcutaneous peripheral lymph nodes (pLNs), albeit with reduced levels of retained α₄β₇ compared with the spleen or mLN. This might be due to the heterogeneity of the transferred CD45RB^low CD4⁺ population, which contains, besides α₄β₇⁺ effector/memory cells, central memory T cells (Supplementary Figure 1A and B), which coexpress other homing receptors such as L-selectin (CD62L). L-selectin controls pLN immigration and is likely to cause enrichment of the CD62L⁺ fraction at this site.

Figure 1

Stable expression of integrin α₄β₇ on ex vivo isolated CD4⁺ T_mem. (a and b) α₄β₇ Expression on T_mem 15 days after adoptive transfer. Ex vivo isolated α₄β₇⁺P-lig⁻CD4⁺CD45RB^low and α₄β₇⁻P-lig⁻CD4⁺CD45RB^low T cells were labeled with carboxyfluorescein succinimidyl ester (CFSE) and transferred into recipient mice. Fifteen days after transfer, cells from peripheral lymph node (pLN), mesenteric lymph node (mLN), and spleen (SPL) were reanalyzed for α₄β₇ expression. Representative FACS (fluorescence-activated cell sorter) plots from three independent experiments are shown in (a). Control: staining control FMO-α₄β₇ (fluorescence minus one). Numbers indicate frequencies of α₄β₇⁺ cells in transferred CFSE⁺ cells. Accumulated values from three independent experiments (α₄β₇⁻, n=6; α₄β₇⁺, n=9) are shown in (b) as mean±s.d. of pooled geometric mean fluorescence intensity (GMFI) of CFSE+ cells. (c) α₄β₇ Expression 6 weeks after transfer. Mean±s.d. of pooled GMFI data from three or four mice in three experiments. Student’s t-test with ***P<0.001.

PowerPoint slide

Stable expression of α₄β₇ on murine CD4⁺ T_mem depends neither on stabilizing factors within the GALT nor on the presence of RA

The observed in vivo stability could be mediated either by a cell-intrinsic imprinting of α₄β₇ expression or by continual encounter with environmental signals required for its maintenance. As α₄β₇ is induced by the vitamin A metabolite RA,¹³ we asked whether the maintained expression of α₄β₇ in vivo depends on the presence of RA.

To investigate this question, we generated vitamin A-deficient mice by pre- and postnatal feeding of a vitamin A-deficient diet for at least 14 weeks and used them as recipients for adoptive transfer of α₄β₇⁺ and α₄β₇⁻ CD4⁺ T_mem. As shown in Figure 2a, the expression of α₄β₇ on transferred CD4⁺ T_mem was not affected by vitamin A deficiency. By contrast, endogenous T cells displayed low levels of α₄β₇ under these conditions (Figure 2a), and naive OVA-specific CD4⁺ T cells transferred into vitamin A-deficient mice failed to upregulate α₄β₇ and CCR9 upon oral application of OVA protein and cholera toxin (Figure 2b). This proved the effective removal of RA from the system. The data indicate that while RA is critical for the de novo induction of α₄β₇, it is not required for the maintenance of α₄β₇ on CD4⁺ T_mem.

Figure 2

Stable expression of integrin α₄β₇ on ex vivo isolated T_mem after transfer into vitamin A-deficient mice or exclusion from gut-associated lymphoid tissues (GALT). (a) Expression is retained in the absence of vitamin A. Ex vivo isolated α₄β₇⁺P-lig⁻CD4⁺CD45RB^low and α₄β₇⁻P-lig⁻CD4⁺CD45RB^low T cells were labeled with carboxyfluorescein succinimidyl ester (CFSE) and transferred into control or vitamin A-depleted (VAD) mice. After 15 days, transferred cells were reanalyzed for α₄β₇ expression. At 15 days after transfer, cells from peripheral lymph node (pLN), mesenteric lymph node (mLN) and spleen (SPL) were isolated and reanalyzed for α₄β₇ expression. Representative FACS (fluorescence-activated cell sorter) plots from three independent experiments with eight mice; the frequency of α₄β₇⁺ cells (bold) and geometric mean fluorescence index (GMFI) (italic) among transferred cells is given. (b) Lack of α₄β₇ and CCR9 upregulation verifies depletion of vitamin A. CFSE-labeled Ova-specific CD4⁺CD62L^high T cells from DO11.10 mice were injected into control or VAD mice. At 24 h after transfer, mice were immunized with Ova protein and cholera toxin by oral gavage. On day 4 after immunization, analysis of α₄β₇ expression on OVA-specific (KJ71-16⁺) cells was performed by flow cytometry. Plots are gated on transferred KJ⁺ cells. (c) Expression of α₄β₇ is retained upon exclusion from GALT. Ex vivo FACS-sorted α₄β₇⁺P-lig⁻CD4⁺CD45RB^low and α₄β₇⁻P-lig⁻CD4⁺CD45RB^low T cells were labeled with CFSE and transferred into recipient mice. To exclude α₄β₇⁺ T_mem from GALT recipient mice received anti-MAdCAM-1 (Meca367; 500 μg per mouse) and anti-VCAM-1 Abs (6C7.1; 200 μg per mouse) or, as control, rat IgG intravenously at the day of cell transfer and then every 3 days until the day of analysis. Representative FACS plot from three independent experiments with four mice; the frequency of α₄β₇⁺ cells and GMFI among transfered CFSE⁺ cells is given.

PowerPoint slide

T_mem residing in secondary lymphoid tissues are known to recirculate to some extent between blood, lymph, and tissues.^{24, 25} We therefore asked whether local factors in the gut or GALT were required for the stability of α₄β₇ expression in CD4⁺ T_mem. To prevent access of CD4⁺ T_mem to mucosal sites, we blocked the adhesion molecules MAdCAM-1 and vascular cell adhesion molecule-1 (VCAM-1) involved in the entry of lymphocytes into this compartment by injection of anti-MAdCAM-1 and anti-VCAM-1 antibodies. This treatment reduced the entry of CD4⁺ T_mem into both gut and GALT by 85–95% (Supplementary Figure 2). To examine whether the maintenance of α₄β₇ expression depends on gut-derived instructive signals, we transferred α₄β₇⁺ CD4⁺ T_mem into recipient mice together with injection of either anti-MAdCAM-1 plus anti-VCAM-1 antibodies or a rat-immunoglobulin G (IgG) control antibody. Antibody treatment was repeated every 3 days until analysis on day 15, resulting in the exclusion of transferred CD4⁺ T_mem from the gut and GALT and their accumulation in the spleen (Figure 2c and Supplementary Figure 2). These conditions did not result in any reduction of α₄β₇ expression on transferred α₄β₇⁺ CD4⁺ T_mem (Figure 2c), suggesting that the maintenance of α₄β₇ expression, in contrast to its induction on T_N cells, is independent of signals provided within the gut and GALT, but relies on a cell-intrinsic mechanism.

α₄β₇, but not CCR9, displays stability on murine T_mem during in vitro culture

We tested ex vivo sorted α₄β₇⁺ CD4⁺ T_mem for their phenotypic stability in in vitro cultures and observed that they largely retained α₄β₇ expression upon restimulation (Figure 3a), confirming the imprinted state of α₄β₇ in memory cells. Notably, expression was retained not only in the absence of RA but also in the presence of interleukin-12 (IL-12), which is known to upregulate the competing homing receptors P- and E-selectin ligands (“skin-homing phenotype”).^{26, 27} Selectin ligands are expressed in T_mem, but, for ex vivo isolated cells, in a largely mutually exclusive way to α₄β₇ (see Figure 6 and Supplementary Figure 1).

Figure 3

Integrin α₄β₇, but not CCR9, expression is stably imprinted within ex vivo isolated T_mem. (a) Ex vivo isolated α₄β₇⁺P-lig⁻CD4⁺CD45RB^low T cells were stimulated with plate-bound anti-CD3/28 antibodies and interleukin-2 (IL-2) in the presence or absence of retinoic acid (RA) (1 nM) and IL-12 (5 ng ml⁻¹). On day 5, cells were harvested and analyzed for homing receptor expression by flow cytometry. Some unstained cells (lower left quadrant) were added in the staining control to facilitate compensation. (b) Ex vivo isolated CCR9⁻ α₄β₇⁺ and CCR9⁺ α₄β₇^+/− T cells (P-lig⁻CD4⁺CD45RB^low) were stimulated with plate-bound anti-CD3/28 antibodies and IL-2 in the presence or absence of RA (0.1 and 1 nM). On day 5, cells were analyzed for α₄β₇ and CCR9 expression. FACS (fluorescence-activated cell sorter) plots are representative of three independent experiments each performed in (a) triplicates or (b) duplicates.

PowerPoint slide

Besides α₄β₇, the chemokine receptor CCR9 is also induced by RA. CCR9 is thought to have a role in the immigration of cells to the small intestine; however, in accordance with previous reports,¹⁶ only a small fraction of—mostly α₄β₇⁺—CD4⁺ T_mem from mLNs and spleen express this receptor. When ex vivo-sorted CCR9⁺ T_mem were stimulated in vitro, CCR9 was almost completely lost in the absence but retained in the presence of RA (Figure 3b), suggesting that CCR9 is not imprinted in CD4⁺ T_mem but requires continual RA signals from the environment to remain expressed. These data add further evidence to previous findings, suggesting that the regulation of the gut-homing receptors α₄β₇ and CCR9 differs.^{28, 29}

Repetitive, but not single, stimulation in the presence of RA stabilizes α₄β₇ expression

α₄β₇ Expression can be induced in CD4⁺ T cells by RA, but it is unclear whether this is sufficient to achieve stable α₄β₇ expression as found in CD4⁺ T_mem. Therefore, we induced α₄β₇ expression on murine CD25⁻CD4⁺CD62L⁺ T_N cells during in vitro activation in the presence of RA and anti-IL-4 antibody (IL-4 has been reported to influence negatively α₄β₇ induction in vitro).¹⁶ On days 5 and 11 of culture, the cells were restimulated in the absence or presence of RA. α₄β₇ Expression was analyzed at days 5, 11, and 17. As expected, T-cell activation in the presence of RA resulted in the efficient upregulation of α₄β₇ (Figure 4a and b). However, in vitro-generated α₄β₇⁺ T cells rapidly lost α₄β₇ expression when restimulated for 5 days in the absence of RA. Only α₄β₇⁺ T cells that had been stimulated two times in the presence of RA maintained α₄β₇ expression upon restimulation in the absence of RA (Figure 4a and b). Thus, repetitive stimulation with RA, in contrast to single treatment, was sufficient to stabilize α₄β₇ expression in vitro, suggesting that imprinting of α₄β₇ expression in T cells requires a repeated or prolonged encounter with instructive signals.

Figure 4

Repetitive, but not single, stimulation in the presence of retinoic acid (RA) stabilizes α₄β₇ expression on CD4⁺ T_N cells in vitro. (a and b) CD4⁺ T_N cells were TCR stimulated (1st round) in vitro with plate-bound anti-CD3/CD28 antibodies under neutral (“N”=no further supplements) or polarizing conditions with RA (1 nM) and anti-IL-4 antibody (“RA”). TCR stimulation with one of the two conditions was repeated on day 5 (2nd round) and on day 11 (3rd round). Cells were analyzed for α₄β₇ expression before restimulation on day 5, 11, and 17. (a) Representative FACS (fluorescence-activated cell sorter) plots and (b) accumulated data from three replicates (mean±s.d. of geometric mean fluorescence index (GMFI)) of one representative out of three independent experiments are shown. (c) Stability of α₄β₇ expression in vivo. CD4⁺ T_N cells were stimulated two times with plate-bound anti-CD3/CD28 antibodies with or without RA and anti-IL-4 antibody as above. On day 11, cells were labeled with carboxyfluorescein succinimidyl ester (CFSE) and transferred into recipient mice. Isolation and analysis were carried out on day 15 after transfer. Mean±s.d. of pooled GMFI data from two out of three independent experiments with n=4 or 5 mice in total. Student’s t-test with **P<0.01 and ***P<0.001.

PowerPoint slide

Consistent with the in vitro stability, repeatedly RA-treated T cells retained a higher expression of α₄β₇ compared with control cells on day 15 after transfer into recipient mice (Figure 4c), albeit the expression level was lower than at the time of transfer. This indicates that repetitive rounds of stimulation in the presence of RA resulted in at least a partial stabilization of α₄β₇ expression in T cells.

Two regulatory elements in the murine Itga4 locus show epigenetic modifications correlating with the stability of α₄β₇ expression

Cell-intrinsic imprinting of a given gene activity state often relies on mechanisms of epigenetic regulation such as DNA methylation of critical genomic elements. We therefore performed a pilot MeDIP-on-Chip experiment to search for potential differentially methylated regions in the Itga4 and Itgb7 loci in ex vivo α₄β₇⁺ vs. α₄β₇⁻ CD4⁺ T_mem subsets (data not shown). Differential methylation was apparent in two regions in the Itga4 locus: the Itga4 promoter upstream of the transcription start site containing a CpG-island (CpG-I), and a region in the second intron, which we termed differentially methylated region 1 (dmr1), both containing several CpG motifs (Figure 5a). No dmr were observed in the Itgb7 locus. In line with previous reports,³⁰ we found only Itga4, but not Itgb7, to be regulated by repetitive RA treatment (Supplementary Figure 3). For these reasons, we focused our efforts on clarifiying the mechanism for the stability of α₄β₇ expression on the Itga4 locus.

Figure 5

The promoter and an intronic enhancer in the murine Itga4 locus show transcriptional activity and differential DNA methylation patterns. (a) The genomic organization of the murine Itga4 locus is shown with the position of the analyzed promoter (pro) and enhancer (differentially methylated region 1 (dmr1)) region. The position of CpG motifs within the promoter and in the dmr1 region used in (b and c) are shown as needle heads. CpG motifs shown in black were analyzed for methylation in (d and e). (b and c) The Itga4 pro and the dmr1 enhancer show promoter or methylation-sensitive enhancer activity, respectively. (b) The Itga4 promoter and the dmr1 region were cloned into the promoter-free pGL3 luciferase reporter plasmids (pGL3-basic) and tested for transcriptional promoter activity in murine CD4⁺ Th1 cells in a luciferase reporter assay. In addition, dmr1 was tested in a CpG-free luciferase reporter vector (pCpGL-hEF1) containing the hEF1 promoter. This vector allows application of in vitro methylation (and mock methylation as control), as this affects only the cloned insert. (c) With this approach, the methylation sensitivity of the enhancer function of dmr1 was assessed in primary murine CD4⁺ T cells. (d and e) The degree of DNA methylation of selected CpG motifs (shown in a as black needle heads) in the (d) Itga4 promoter and (e) dmr1 was analyzed by bisulfite sequencing of T_N (n=2) and ex vivo isolated α₄β₇⁺ or α₄β₇⁻ P-lig⁺ T_mem (n=3 each), as well as of cultured T_N after two rounds of retinoic acid (RA) or LE540 treatment followed by a final round in the absence of both additives (N) (n=3 each). Mean (±s.d., if n>2) methylation values for each CpG motif are given. Statistical significance of differences between all analyzed CpG motifs of the respective cell types in a given region was tested as described in the Method section. Significant differences are indicated: ***P<0.001 and NS, not significant. TSS, transcription start site.

PowerPoint slide

We tested both regions in the Itga4 gene for their regulatory activity on transcription using a classic luciferase reporter assay in murine CD4⁺ T cells (Figure 5b and c). As expected, the Itga4 promoter (but not dmr1; Figure 5b) was able to induce expression of the luciferase reporter, whereas the dmr1 element displayed an enhancer function when tested in conjunction with a promoter (Figure 5c).

Next, we designed amplicons on both regions to analyze the methylation status in CD4⁺ T cells. As the promoter contained a large CpG-I, which was largely demethylated (data not shown), we concentrated our analysis on two CpG motifs in the CpG-I shore. Shores of CpG-I have been reported to contain most of the dynamic methylation-dependent regulatory capacity, whereas CpG-Is are usually uniformly demethylated in most promoters.³¹ Indeed, we found significant demethylation of the two CpG-I shore CpGs in α₄β₇⁺ CD4⁺ T_mem as well as in T cells repeatedly stimulated in the presence of RA (Figure 5d). Their α₄β₇⁻ counterparts (α₄β₇⁻P-lig⁺ T_mem, expressing the inflammation/skin-related homing receptor P-selectin ligand (P-lig), as well as T cells stimulated in the presence of the RA antagonist LE540), displayed high methylation levels. Interestingly, ex vivo-isolated T_N cells, which have not yet undergone imprinting for a certain homing phenotype, showed an equally low degree of methylation, suggesting that these cells are permissive for the expression of α₄β₇ in the presence of appropriate conditions, whereas stimulation in the absence of RA signaling or differentiation into skin-homing T_mem promotes acquisition of an inhibitory epigenetic mark at the promoter.

Analysis of seven CpGs in dmr1 showed a significant demethylation in α₄β₇⁺ vs. α₄β₇⁻ CD4⁺ T_mem, whereas RA-induced α₄β₇⁺ T cells displayed only small differences from their counterparts treated with the RA antagonist LE540 or uncommited T_N cells (Figure 5e). These data indicate that the methylation level in dmr1 qualifies as a stabilizer for Itga4 expression in fully imprinted α₄β₇⁺ CD4⁺ T_mem, whereas repetitive stimulation in vitro in the presence of RA might not be sufficient to achieve full epigenetic fixation at this regulatory element. As further support for this conclusion, we found increased stability of α₄β₇ expression in murine T_N, which were treated with the DNA-demethylating drug 5-azacytidine during the RA-mediated induction phase (Supplementary Figure 4).

To confirm the putative methylation-dependent enhancer/stabilizer function of dmr1, we cloned the dmr1 into a CpG-free luciferase reporter³² —which allows selective in vitro methylation of inserted CpG-containing sequences— and performed reporter gene assays in primary murine T cells. Indeed, the enhancer activity of dmr1 was strongly compromised upon methylation (Figure 5c), supporting a functional link between DNA methylation and transcriptional activity.

The human ITGA4 gene also shows epigenetic modification in α₄β₇⁺ vs. α₄β₇⁻ CD4⁺ T cells

The murine Itga4 and human ITGA4 loci show a high degree of structural similarity (Figure 6a) even if the sequence homology is variable. Therefore, similar regulatory mechanisms might apply. Using whole-genome bisulfite sequencing data of human CD4⁺ T cells generated in the context of the German Epigenome Programme “DEEP”,³³ we analyzed the ITGA4 promoter and searched for regions that were differentially methylated between naive and memory cells and might correspond to the murine dmr1. Two DMR were identified for which amplicons for the detailed analysis of DNA methylation were designed (Figure 6a).

Figure 6

Differential DNA methylation in regulatory regions of the ITGA4 locus in ex vivo human T-cell subsets. (a) Structural similarities of the murine Itga4 (above) and the human ITGA4 (below) locus are shown as well as the positions of the promoters (pro/PRO) and the analyzed differentially methylated regions (differentially methylated region 1 (dmr1)/DMR1/DMR2). (b) Human CD4⁺ T-cell subsets were sorted according to the gating shown (left). T_N and T_mem differ in the level of α₄β₇ expressed using the Act-1 antibody. (c–f) The degree of DNA methylation (mean±s.d., n=3) of selected CpG motifs (shown as black needle heads) in the (c) ITGA4 promoter, (d) DMR1 and (e) DMR2 were analyzed by bisulfite sequencing in the sorted CD4⁺ T cell subsets shown in (b). (f) Statistical analyses of the DNA methylation differences over all CpGs analyzed in the indicated region. Significant differences are indicated: *P≤0.05, **P<0.01, and ***P<0.001; NS, not significant. TSS, transcription start site.

PowerPoint slide

Similar to murine T cells, α₄β₇⁺ CD4⁺ T cells are found in the memory T-cell fraction in human peripheral blood and α₄β₇ expression is mutually exclusive to expression of the skin-homing receptor and E-selectin ligand CLA (Figure 6b). In the naive fraction of human CD4⁺ cells, both α₄β₇⁺ and α₄β₇⁻ cells can be detected using the very bright Act-1-PE antibody (Figure 6b), a finding of unknown biological significance that was not reported before and is not observed in murine cells. Compared with T_mem, α₄β₇⁺ T_N cells express detectable, but intermediate, levels of α₄β₇ (Figure 6b). We analyzed the degree of DNA methylation for six CpGs in the CpG-I shore of the promoter as well as for DMR1 and DMR2.

Methylation of the analyzed CpGs in the CpG-I shore of the promoter was generally at a low level, but lowest in the α₄β₇-expressing populations of both T_N and T_mem cells (Figure 6c and f). As observed in murine cells, the highest methylation was found in α₄β₇⁻ (and CLA⁺) memory cells, suggesting an epigenetic silencing of ITGA4 during differentiation into skin-homing T_mem. α₄β₇⁻-naive T cells displayed an intermediate methylation state, suggesting that they are predisposed to ITGA4 expression upon appropriate signaling.

For DMR1, fully commited α₄β₇⁺ CLA⁻ T_mem were the only cell type displaying consistently low levels of methylation among the CpGs of this region, indicating that DMR1 might serve as a stabilizer element (Figure 6d and f). The α₄β₇⁺ fraction of T_N showed a tendency towards decreased methylation compared with the α₄β₇⁻ cell types, which, however, lacked statistical significance.

Interestingly, the methylation pattern in DMR2 discriminated T_mem from T_N, but was not correlated to α₄β₇ expression (Figure 6e and f). This suggests that epigenetic opening of this putative regulatory region might already occur with acquisition of a memory state and might allow for expression of ITGA4 also under conditions, where the α₄-chain, but not the mucosa-specific α₄β₇-integrin is expressed. This would contrast to the role of dmr1/DMR1, which seems to regulate the α₄-chain only in the mucosa-seeking subset of T_mem. In addition, decreased methylation in three regions correlating with upregulated expression in T_mem cells was also found for ITGB1, coding for the β1-integrin chain (Supplementary Figure 6).

Discussion

The differentiation of T-lymphocytes into memory cells is accompanied by functional specialization, including the acquisition of organ-specific homing capabilities. The present study was designed to answer the question whether, and if so which, cell-intrinsic mechanisms allow T cells primed under appropriate conditions to develop a stable topographical memory for the gut, independent of environmental signals. Our present data demonstrate that indeed an “imprinting for gut homing” is observed for the α₄β₇-integrin, which requires repetitive instruction by RA signals. In contrast, stable expression is not inducible for CCR9, at least under the conditions tested here. At the molecular level, this imprinting is based on epigenetic mechanisms, notably regulation of DNA (CpG−) methylation at specific regulatory regions of the Itga4 locus.

The mucosal homing receptor α₄β₇ integrin is expressed at a low, but functional, level on murine naive lymphocytes and enables their recirculation through gut-associated lymphoid tissues.³⁴ Upon activation and differentiation into effector/memory cells, the population splits up into T_mem that are either completely negative or highly positive for α₄β₇, the latter being considered as a stable mucosa-specific memory population.^{10, 35} In contrast, Mora et al.¹⁶ studied the expression of α₄β₇ in CD8⁺ T cells and reported a high degree of plasticity, as freshly generated α₄β₇⁺ CD8⁺ T cells could be converted into α₄β₇⁻ cells by restimulation in the presence of subcutanous instead of mucosal DCs, in line with other reports ¹⁸. Moreover, these authors noted that ex vivo-isolated CD8⁺ T_mem (CCR9⁺α₄β₇^+/−CD44⁺) were also found to change their α₄β₇ expression pattern upon restimulation with the opposite type of DC, suggesting a lack of α₄β₇ stability in CD8⁺ T_mem. More recently, Masopust et al.¹⁷ observed in a comprehensive study a similar instability of α₄β₇ expression on CD8⁺ T cells in vivo. In addition, they found that α₄β₇ was efficiently but only transiently upregulated upon in vivo activation, regardless of the site of antigen contact.

With these data, the original paradigm of organ-specific homing^{1, 36} was questioned. We now demonstrate that CD4⁺ T cells, in contrast, can acquire long-term stability of α₄β₇ expression and hence mucosa-specific homing, restoring the validity of the homing concept for this subset. In fact, differences in the migratory program of CD4⁺ and CD8⁺ T_mem including higher stability of selectin ligands on CD4⁺ T_mem have already been reported¹⁵ and are now further corroborated by our present data.

RA, produced from the precursor vitamin A by stromal cells,^{11, 37} CD103⁺ DCs^{13, 38} in gut-associated lymphoid tissues, or by liver sinusoidal cells,³⁹ is the key inducer of α₄β₇ by regulating the α₄-chain.³⁰ In contrast, basic levels of the β₇-chain are constitutively transcribed in CD4⁺ T cells, outcompeted by high levels of the β₁-chain and regulated by tumor growth factor-β.^{30, 40} We have shown here that RA signals are sufficient to induce α₄β₇ expression, but the stability of expression (in the absence of RA) is not immediately achieved. Two rounds of in vitro activation in the presence of RA induced a certain degree of stability, yet complete stability with sustained high expression of α₄β₇ was found only in ex vivo-isolated T_mem.

Stability in vivo may rely on cell-intrinsic mechanisms, but, alternatively, may be caused by continuous exposure to RA according to three different scenarios: (i) exposure to low levels of RA within the circulation;⁴¹ (ii) contact with RA-generating antigen-presenting cells during passage through the liver;³⁹ (iii) by recirculation through mucosal tissues harboring dendritic cells and/or stromal cells providing RA and/or other unknown instructive signals. However, α₄β₇ expression was neither lost upon transfer into RA-depleted animals nor reduced when recirculation of transferred cells through mucosal tissues was blocked. Thus, the findings of this study strongly argue for cell-intrinsic mechanisms being responsible for the imprinting of the mucosal homing phenotype.

How could stability be achieved at a molecular level? A growing body of evidence assigns a key role to the epigenetic machinery in the fixation of phenotypic profiles, as shown for various processes of cell differentiation and lineage commitment.^{20, 33, 42} In T cells, effector cytokines, master transcription factors and other genes are subject to epigenetic fixation^{21, 22} and some indirect hints from our previous work also pointed towards a role in the regulation of homing.^{14, 43} In particular, DNA methylation at promoters or enhancers appears to be an attractive mechanism for the regulation of cellular memory by control of chromatin accessibility and transcription factor binding, particularly with respect to the heritability of DNA methylation patterns during mitosis.⁴⁴

The results reported here provide first evidence that stable expression of α₄β₇ in the absence of the initial inducer is indeed due to epigenetic regulation at the level of DNA methylation. In both murine and human CD4⁺ T cells, the Itga4/ITGA4 locus was found to display distinct differentially methylated regions (dmr/DMR) whose degree of methylation correlated with α₄β₇ expression, commitment, and/or memory differentiation.

First, several CpGs upstream of the promoter-associated demethylated CpG-I showed distinct differences in methylation. In murine T_N, which express a low but detectable level of α₄β₇,³⁵ the two CpG upstream of the CpG-I were completely demethylated similar to the α₄β₇⁺ fraction of T_mem. Although cells repeatedly stimulated in the presence of RA kept this status, stimulation in the absence of RA as well as in vivo differentiation into α₄β₇⁻ T_mem resulted in a pronounced methylation. Thus, we conclude that the open conformation of this region enables transcription in both T_N and T_mem, which, however, seems to be modulated by further regulatory regions, and can be silencend upon differentiation into non-mucosal subsets of T_mem. It has to be mentioned that the degree of methylation was not uniform among the CpGs analyzed. Whether this relates to specific functions of single CpGs, e.g., by being part of a distinct transcription factor-binding motif, remains to be investigated. In human cells, for which an α₄β₇⁺ as well as an α₄β₇⁻ population can be discriminated among T_N, the latter show moderate methylation, whereas the positive fraction is significantly less methylated. Methylation levels in α₄β₇⁻ T_mem are much higher than in the α₄β₇⁺ T_mem fraction. Thus, the suggested regulatory role of CpG methylation in the promotor-associated region seems to be conserved between mouse and man and to extend to subfractions of naive cells.

Second, differential methylation helped us to identify a novel regulatory region (dmr1) in the murine Itga4 locus that acts as an enhancer. In contrast to the promotor-associated region, the CpG motifs within dmr1 are almost completely methylated in T_N, and there is only a minor, nonsignificant decrease of methylation upon repeated stimulation in the presence of RA, or in α₄β₇⁺ vs. α₄β₇⁻ T_N. However, the methylation level was strongly reduced in α₄β₇⁺ T_mem as compared with α₄β₇⁻ T_mem or T_N. The enhancer activity of dmr1 was completely abolished by artificial methylation, supporting the functional relevance of DNA methylation in controlling this region. A further hint at the functional role of DNA methylation on regulation of α₄β₇⁺ expression is the stabilizing effect of 5-azacytidine, which blocks DNA methylation; however, indirect effects of this genome-wide-acting inhibitor cannot be excluded.

A related site designated as DMR1 was identified in human CD4⁺ T cells, which showed a corresponding demethylation signature correlating to α₄β₇⁺ expression. As an additional indication for a conserved mechanism of regulation via dmr1/DMR1, we found the activating histone modification mark H3K4me1 at the human DMR1, indicating that indeed it may serve as an enhancer element (Supplementary Figure 5).

Recently, the transcription factor BATF has been reported to regulate the expression of the α₄-chain in CD4⁺ cells.⁴⁵ According to the ENCODE data,⁴⁶ BATF binds in a lymphoid cell line at the promoter region; further binding sites for BATF are predicted at the ITGA4 locus including at the DMR1 (using tRap and the jaspar database; data not shown), suggesting BATF as one possible transcription factor involved in α₄β₇ stability.

Whole-genome bisulfite sequencing data of the human ITGA4 locus uncovered an additional DMR (DMR2). Similar to dmr1/DMR1, CpGs 1–5 were almost completely methylated in both α₄β₇⁺- and α₄β₇⁻-naive cell subsets, but a highly significant drop in methylation was observed in both memory subsets. Here the differences between α₄β₇⁺ and α₄β₇⁻ were less prominent and did not reach statistical significance. Hence, this region might be required for expression of the α₄-chain in T_mem in general (e.g. in α₄β₁⁺, but α₄β₇⁻ cells), but additional signals (e.g., from the dmr1/DMR1 enhancer) might regulate it in the mucosa-specific subset.

A subset of effector/memory T cells expresses the related integrin dimer α₄β₁ (“VLA-4”). α₄β₁ but not α₄β₇ has a key role in multiple sclerosis as this integrin is essential for pathogenic T cells to cross the blood–brain barrier^{47, 48} and can be expressed on effector T cells migrating to further tissues such as the skin and the respiratory track as well. Importantly, expression of α₄β₁ and α₄β₇ on memory cells is largely mutually exclusive (data not shown). Our findings on the imprinting of the gut-homing phenotype via epigenetic regulation at DMR1 of the ITGA4 gene apply to the α₄β₇⁺ subset; α₄β₇⁻ P-lig⁺ cells, which partially express α₄β₁, do not show a demethylated DMR1. However, DMR2, which is demethylated in all memory cells could be a candidate for an epigenetically controlled region regulating α₄β₁ in α₄β₇⁻ T_mem cells. Moreover, first evidence that the ITGB1 locus for his part is also regulated by epigenetic mechanisms can be deduced from the differential methylation in several regions of the human ITGB1 gene correlating with differential expression in T_N vs. T_mem (Supplementary Figure 6).

In conclusion, the data of this study confirm the validity of the concept of a topographical memory in CD4⁺ T_mem and provide first evidence that the imprinting of a mucosa-specific homing phenotype involves epigenetic regulation of expression of the α₄-integrin chain.

Methods

Mice C57Bl/6 and congenic Thy1.1 C57Bl/6 mice were bred at the Bundesinstitut fuer Risikobewertung (Berlin, Germany) and were used at 8–12 weeks of age. BALB/c mice and 7–9-month-old exbreeders were from Charles River WIGA GmbH (Sulzfeld, Germany). All animal experiments were performed under specific pathogen-free conditions and in accordance with institutional, state, and federal guidelines as approved by Landesamt für Gesundheit und Soziales (LaGeSo) Berlin.

Human blood samples Buffy coats were from DRK-Blutspendedienst Nord-Ost and were used for isolation of human T-cell populations as approved by the local ethics committee (Ethikkommission, Ethikausschuss 1, CCM; application no. EA1/095/13).

Antibodies and cell culture reagents Antibodies and fusion proteins used were as follows: anti-mCD4 (RM4-5), anti-mCD25 (PC6.1), anti-mCD45RB (16A), anti-mThy1.1 (OX-7), anti-mα₄β₇ (DATK32) and streptavidin PE-Cy7 (all BD Pharmingen, Franklin Lakes, NJ), anti-human IgG (Jackson Immuno Research, West Grove, PA), anti-mCCR9 (242503; R&D Systems, Minneapolis, MN), biotinylated anti-mα₄β₇ (DATK32, anti-mFcγR II/III (2.4G2), anti-mB220 (RA3.6B2), anti-mMac-1 (M1/70.15.11), anti-mCD8 (53-6.7), anti-mCD3 (145.2C11), anti-mCD28 (37.51), anti-OVA-specific mTCR (KJ1.26), anti-mMAdCAM-1 (Meca367), anti-mVCAM-1 (6C7.1), and anti-mIL4 (all produced in house), anti-hCD3 (UCHT1), anti-hCD4 (OKT4), anti-hCD45RO (UCHL1), anti-hCLA (HECA-452) (from Biolegend, San Diego, CA), and anti-hCD45RA (2H4LDH11LDB9; Beckman-Coulter, Brea, CA). Recombinant murine P-selectin-huIgG fusion protein used for labeling P-selectin ligand (P-lig) was kindly provided by D Vestweber (Münster, Germany). The anti-human α₄β₇ antibody Act-1 was provided by Millenium at Takeda Pharmaceutical Company (Cambridge, MA) and PE-conjugated in-house. Microbeads were from Miltenyi Biotec (Bergisch Gladbach, Germany). Mouse recombinant IL-2 (rmIL-2), rmIL-12, and rmIFNγ (interferon-γ) were from R&D Systems, all-trans RA and ovalbumin and cholera toxin were from Sigma-Aldrich, St. Louis, MO, pan-RA receptor antagonist LE540 was from Wako Chemicals (Neuss, Germany). Cell culture was performed in RPMI-1640 (Gibco, Carlsbad, CA and PAA (Cölbe, Germany)) supplemented with 10% fetal calf serum (Sigma).

Flow cytometry For flow cytometry an LSRII (BD Biosciences, Franklin Lakes, NJ) and FlowJo software were used. Gates were set according to FMO (fluorescence minus one) stainings with or without isotype controls. Cell sorting was performed on FACSAria I or II Cell Sorter BD, Franklin Lakes, NJ.

Isolation of ex vivo murine and human T cells For in vitro culture of murine CD4⁺ T_N, pooled erythrocyte-depleted spleen and lymph node cells from BALB/c, C57Bl/6, or Thy1.1 congenic C57Bl/6 mice were stained with anti-mCD4-FITC and anti-mCD25-APC antibodies and depleted of CD25⁺ cells using anti-APC microbeads and the AutoMACS magnetic separation system (Miltenyi Biotec). Subsequently, CD25⁻ cells were positively enriched for CD4⁺ T cells using anti-FITC multisort beads. Release reagent and anti-CD62L microbeads were used to obtain CD25⁻CD4⁺CD62L^high T cells.

For the isolation of murine CD4⁺ T_mem, pooled spleen, peripheral, and mesenteric lymph node cells from 20 to 40 BALB/c exbreeder mice were negatively selected with anti-mCD8, anti-mMac-1, and anti-mB220 antibodies and anti-rat-IgG microbeads. Enriched CD4⁺ T cells were labeled with anti-CD45RB FITC and depleted of CD45RB^hi cells using anti-FITC microbeads. Subsequently, α₄β₇⁺P-lig⁻CD45RB^lowCD4⁺ and α₄β₇⁻P-lig⁻CD45RB^lowCD4⁺ T cells were sorted by flow cytometry. For methylation analyses, the α₄β₇⁻P-lig⁺ subset (“inflammation/skin-specific subset”) was sorted as the α₄β₇⁻ subset.

For isolation of human T cells, peripheral blood mononuclear cells were isolated by density gradient centrifugation using Lymphocyte Separation Medium LSM 1077 (PAA, Cölbe, Germany) from buffy coats. Erythrocytes were lysed in erythrocyte lysis buffer (Buffer EL; Qiagen, Venlo, Niederlande). CD4⁺ T-lymphocytes were enriched using CD4 MicroBeads (Miltenyi Biotec). The enriched population was stained and sorted by flow cytometry into four populations: T_N Act-1⁻=CD3⁺CD4⁺CD45RA⁺CD45RO⁻Act-1⁻CLA⁻; T_N Act-1⁺=CD3⁺CD4⁺CD45RA⁺CD45RO⁻Act-1⁺CLA⁻; T_mem Act-1⁻ CLA⁺=CD3⁺CD4⁺CD45RA⁻CD45RO⁺Act-1⁻CLA⁺; T_mem Act-1⁺=CD3⁺CD4⁺CD45RA⁻CD45RO⁺Act-1⁺CLA⁻.

Purity of the sorted populations was confirmed by flow cytometry and >95%.

In vitro cell cultures Murine naive CD25⁻CD4⁺CD62L^high T cells were stimulated on plate-bound anti-mCD3 (1–4 μg ml⁻¹) and anti-mCD28 (4–8 μg ml⁻¹) antibodies or via irradiated APCs with soluble anti-mCD3 and anti-mCD28 in the presence of rmIL-2 (10 ng ml⁻¹) either under neutral (no further supplements) or under Th1 polarizing conditions (rmIL-12 (5 ng ml⁻¹), rmIFNγ (20 ng ml⁻¹), anti-mIL-4 (5 μg ml⁻¹)), with or without the addition of RA (1 nM) or LE540 (0.25 μM, to block medium-derived RA). On day 3 of culture, T cells were removed from the stimulus and rested for another 1–3 days. Cells were analyzed for α₄β₇ expression on the indicated days and restimulated for prolonged culture periods on days 5 and 11. In contrast to T_N, ex vivo isolated T_mem were stimulated with reduced levels of plate-bound anti-mCD3 (1 μg ml⁻¹) and anti-mCD28 (2 μg ml⁻¹) for 12–16 h and subsequently rested until restimulation.

Analysis of homing receptor expression on adoptively transferred T cells A total of 1.5–3 × 10⁶ cultured CD4⁺ T cells (day 11 of culture) or ex vivo isolated, CFSE-labeled or Thy1.1/2 congenic T_mem were injected into the tail vein of recipients. On day 15 or 6 weeks after T-cell transfer, recipient mice were killed and cells from spleen, pLN, and mLNs were stained for CD4, Thy1.1, and α₄β₇. At these time points, between 0.12 and 0.05% of the CD4⁺ T cells in lymphoid organs were donor-derived.

Generation of vitamin A-deficient mice Female mice received a vitamin A-deficient (ssniff Spezialdiäten GmbH, Soest, Germany) or control diet 10 days after initial mating (embryonic days 7–10). Offspring was weaned at 3 weeks of age and on diet until 19 weeks of age before transfer of T_mem. To verify depletion of vitamin A, CFSE-labeled OVA-specific CD4⁺CD62L^high T cells from DO11.10 mice were transferred to mice fed vitamin A-reduced or control diet. On day 1 after transfer, mice were immunized with 100 mg OVA protein and 10 μg cholera toxin (both Sigma-Aldrich) by oral gavage. Four days after immunization, α₄β₇ and CCR9 expression were analyzed on OVA-specific (KJ⁺) mLN cells.

Exclusion of α₄β₇⁺ T_mem from GALT To exclude α₄β₇⁺ T_mem from GALT, recipient mice received either rat IgG or anti-MAdCAM-1 (Meca367; 500 μg per mouse) plus anti-VCAM-1 (6C7.1; 200 μg per mouse) antibodies intravenously every 3 days. This treatment started on the day of transfer and was maintained until the analysis of α₄β₇ expression on day 15.

Generation of luciferase reporter vectors The Itgα4 inserts were amplified by PCR using mouse genomic DNA and the following primers (Eurofins, Hamburg, Germany): Itga4 promoter_fwd: 5′-CTGGTGGTAGGTATGTCCTGGGGT-3′ and Itga4 promoter_rev: 5′-CGCTCTTGGTGGAGAACATT-3′; dmr1_fwd: 5′-TCAGATTTTGCTAGCCATCCT-3′ and dmr1_rev: 5′-TGCTTCCCACAATTCTAAAACA-3′. PCR fragments for the Itga4 promoter and dmr1 were cloned into the pGL3-basic luciferase reporter vector (Promega, Madison, WI, USA). For the generation of a reporter vector for dmr1 in a CpG-free background,³² the dmr1 region was inserted into the pCpGL-EF1 vector.⁴⁹ Endotoxin-free plasmid DNA for transfection was purified using NucleoBond Xtra Maxi EF (Macherey-Nagel, Düren, Germany).

Luciferase reporter assay Murine-naive CD25⁻CD4⁺CD62L^high T cells were stimulated for two days using plate-bound antibodies and polarized towards Th1 as described above. A total of 1 × 10⁶ in vitro-induced C57Bl/6 Th1 cells were transfected with equimolar amounts (3.12 pmol) of the reporter plasmids using the Neon transfection device (Life Technologies, Carlsbad, CA) applying two pulses (voltage: 1,350 V; width: 20 ms). Fifty nanograms of the pRL-CMV vector (Promega) containing the Renilla luciferase cDNA were co-transfected and used as an internal control for transfection efficiency. Cells were cultured for another 24 h with and another 24 h without plate-bound stimulation antibodies. Cells were harvested and luciferase activity was measured using the Dual-Luciferase Reporter Assay System (Promega). After normalization of firefly activity to Renilla activity, relative firefly activity was normalized to activity of the empty vector.

In vitro methylation of luciferase reporter plasmids Thirty micrograms of plasmid DNA were incubated with M.SssI (2.5 U μg⁻¹ DNA) in the presence of 160 μM S-adenosylmethionine for 4 h at 37 °C, with another 160 μM S-adenosylmethionine added after the first 2 h of incubation. Mock-methylated plasmids were treated identically but without the enzyme and S-adenosylmethionine. Plasmid DNA was purified using the NucleoSpinExtract II Kit (Macherey-Nagel). Successful methylation was verified with the methylation-sensitive restriction enzyme HpaII.

DNA methylation analysis by bisulfite sequencing Bisulfite sequencing for murine and human samples was performed according to Gries et al.⁵⁰ (primers in Supplementary Table 1). Briefly, 300 ng genomic DNA was bisulfite treated using the EZ DNA Methylation Kit (Zymo Research, Irvine, CA, USA). PCRs were performed using primers with a specific 3′-portion and a universal 5′-portion according to Illumina’s specifications (specific portion of primer sequences are listed in Supplementary Table 1). Amplicons were purified using AGENCOURT Ampurebeads (Beckman/Coulter), diluted, and pooled. Deep sequencing was performed on the Illumina (San Diego, CA) 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 (http://www.mpi-inf.mpg.de)⁵¹ setting the maximal fraction of unrecognized sites’ filter at 0.1.

Statistics If not indicated otherwise, data are presented as mean±s.d. or mean±s.e.m.). Significance was determined with Student’s t-test, after testing of normal distribution by the Kolmogorov–Smirnov test. Differences were considered statistically significant with *P≤0.05, **P<0.01, and ***P<0.001 Differences in DNA methylation were analyzed by computing differences of permuted methylation levels on each CpG site. Statistical significance was determined by applying one-sample, one-tailed Student’s t-test on the differences in all CpG in the respective region, with the alternative hypothesis of true methylation difference assumed as >10% or <−10%. Significance levels (*P≤0.05, **P<0.01, and ***P<0.001) were in accordance with the nonparametric, one-sample Wilcoxon’s signed-rank test with similar settings.