Global DNA Methylation Remodeling Accompanies CD8 T Cell Effector Function Christopher D. Scharer, Benjamin G. Barwick, Benjamin A. Youngblood, Rafi Ahmed and Jeremy M. Boss This information is current as of October 1, 2021. J Immunol 2013; 191:3419-3429; Prepublished online 16 August 2013; doi: 10.4049/jimmunol.1301395 http://www.jimmunol.org/content/191/6/3419 Downloaded from

Supplementary http://www.jimmunol.org/content/suppl/2013/08/20/jimmunol.130139 Material 5.DC1 References This article cites 81 articles, 25 of which you can access for free at: http://www.jimmunol.org/content/191/6/3419.full#ref-list-1 http://www.jimmunol.org/

Why The JI? Submit online.

• Rapid Reviews! 30 days* from submission to initial decision

• No Triage! Every submission reviewed by practicing scientists by guest on October 1, 2021 • Fast Publication! 4 weeks from acceptance to publication

*average

Subscription Information about subscribing to The Journal of Immunology is online at: http://jimmunol.org/subscription Permissions Submit copyright permission requests at: http://www.aai.org/About/Publications/JI/copyright.html Email Alerts Receive free email-alerts when new articles cite this article. Sign up at: http://jimmunol.org/alerts

The Journal of Immunology is published twice each month by The American Association of Immunologists, Inc., 1451 Rockville Pike, Suite 650, Rockville, MD 20852 Copyright © 2013 by The American Association of Immunologists, Inc. All rights reserved. Print ISSN: 0022-1767 Online ISSN: 1550-6606. The Journal of Immunology

Global DNA Methylation Remodeling Accompanies CD8 T Cell Effector Function

Christopher D. Scharer,* Benjamin G. Barwick,* Benjamin A. Youngblood,*,† Rafi Ahmed,*,† and Jeremy M. Boss*,†

The differentiation of CD8 T cells in response to acute infection results in the acquisition of hallmark phenotypic effector functions; however, the epigenetic mechanisms that program this differentiation process on a genome-wide scale are largely unknown. In this article, we report the DNA methylomes of Ag-specific naive and day-8 effector CD8 T cells following acute lymphocytic chorio- meningitis virus infection. During effector CD8 T cell differentiation, DNA methylation was remodeled such that changes in DNA methylation at gene promoter regions correlated negatively with gene expression. Importantly, differentially methylated regions were enriched at cis-elements, including enhancers active in naive T cells. Differentially methylated regions were associated with cell type–specific transcription factor binding sites, and these transcription factors clustered into modules that define networks targeted Downloaded from by epigenetic regulation and control of effector CD8 T cell function. Changes in the DNA methylation profile following CD8 T cell activation revealed numerous cellular processes, cis-elements, and transcription factor networks targeted by DNA methylation. Together, the results demonstrated that DNA methylation remodeling accompanies the acquisition of the CD8 T cell effector phenotype and repression of the naive cell state. Therefore, these data provide the framework for an epigenetic mechanism that is required for effector CD8 T cell differentiation and adaptive immune responses. The Journal of Immunology, 2013, 191: 3419–3429.

n response to acute infection, naive CD8 T cells differentiate Epigenetic mechanisms ensure the maintenance and inheritance http://www.jimmunol.org/ into effector cells capable of killing infected cells and clearing of gene-expression programs through cell division and include DNA I the infection. Effector CD8 T cell function is characterized by methylation and histone modifications (8, 9). Mammalian DNA the induction of a specific transcriptional program that drives rapid methylation primarily involves the methylation of CpG dinucleo- proliferation, expression of key cytokines and effector proteins tides and is associated with a repressed epigenetic state when found necessary for cell killing, and the capacity to migrate into infected in gene promoters (8, 10–12). DNA methylation is maintained or tissue (1–4). Upon Ag clearance, 90% of effector cells undergo deposited de novo by one of three DNA methyltransferases (DNMT1, apoptosis, whereas the remaining cells complete their differenti- DNMT3A, or DNMT3B). Methylated CpG DNA is recognized/ interpreted by a family of methyl-CpG–binding proteins (10, 13,

ation into a pool of memory CD8 T cells (5). A number of the by guest on October 1, 2021 critical transcription factors that drive this differentiation program, 14). DNA methylation readers and writers are vital components of such as Blimp-1 (Prdm1), Tbet (Tbx21), and Eomesodermin (Eomes) the adaptive immune response. Deletion of the maintenance meth- have been identified and characterized (2, 6, 7). However, little is yltransferase DNMT1 during T cell development resulted in normal known about the epigenetic programs that enforce the induced gene lineage formation but led to homeostatic defects and the inability expression changes and permit faithful inheritance of effector to silence lineage-specific genes in CD4 T cell differentiation (15). function during the proliferative phase of infection. Similarly, deletion of the de novo methyltransferase DNMT3A in CD4 T cells did not affect lineage specification but permitted ectopic cytokine expression and increased lineage plasticity (16). *Department of Microbiology and Immunology, Emory University, Atlanta, GA 30322; Interestingly, conditional deletion of DNMT1 at the time of CD8 † and Emory Vaccine Center, Emory University, Atlanta, GA 30322 T cell activation resulted in a diminished effector pool, fewer Received for publication May 24, 2013. Accepted for publication July 22, 2013. memory CD8 T cells, and a reduced ability to clear Ag (17). In This work was supported by National Institutes of Health Grants PO1AI 080192-05 contrast, deletion of the DNA methylation reader MBD2 had no and U19 AI05726-08 to (J.M.B. and R.A.), as well as by an American Cancer Society effect on proliferation but inhibited the formation of a functional Postdoctoral Fellowship (PF-09-134-01-MPC to B.A.Y.). memory compartment following viral challenge (18). These studies The sequences presented in this article have been submitted to the Gene Expression Omnibus database (http://www.ncbi.nlm.nih.gov/geo/) under accession number demonstrate the importance of DNA methylation in maintaining GSE44638. phenotypic programs during the rapid proliferation of effector CD8 Address correspondence and reprint requests to Dr. Jeremy M. Boss, Emory Univer- T cells and indicate that correct interpretation of methylated se- sity, 1510 Clifton Road, Room 3001, Atlanta, GA 30322. E-mail address: jmboss@ quences is functionally important for the adaptive immune response. emory.edu The dynamic nature of DNA methylation in hematopoietic cell The online version of this article contains supplemental material. types has been cataloged during early differentiation of the myeloid Abbreviations used in this article: cBS, clonal bisulfite sequencing; ChIP-seq, chro- and lymphoid lineages (19), mouse erythropoiesis (20), and between matin immunoprecipitation sequencing; D8, day 8; DMR, differentially methylated region; ENCODE, Encyclopedia of DNA Elements; ES, embryonic stem; FDR, human regulatory T cells and naive CD4 cells (21). However, none false discovery rate; GO, Gene Ontology; H3K27ac, histone H3 lysine 27 acetylated; of these studies profiled a clonal or Ag-specific population of cells. H3K4me, histone H3 lysine 4 methyl; H3K27me, histone H3 lysine 27 methyl; LCMV, lymphocytic choriomeningitis virus; MeDIP, methyl DNA immunoprecipitation; A recent study in CD4 T cells demonstrated that developmental MeDIP-seq, methyl DNA immunoprecipitation sequencing; PDE, putative distal en- TCR-specific signaling can establish a pre-existing methylation hancer; RefSeq, NCBI Reference Sequence Database; rpm, reads per million; TSS, profile, such that only CD4 T cells that upregulated Foxp3 and transcription start site. exhibited a specific methylation epitype were able to differentiate Copyright Ó 2013 by The American Association of Immunologists, Inc. 0022-1767/13/$16.00 into the regulatory T cell lineage (22). These data suggest epi- www.jimmunol.org/cgi/doi/10.4049/jimmunol.1301395 3420 EFFECTOR CD8 T CELL EPIGENETICS genetic heterogeneity in the naive CD4 T cell pool and highlight to the manufacturer’s recommended protocol (Illumina). Following agarose the importance and value of profiling the linear differentiation of gel size selection, naive and D8 effector CD8 T cell samples were im- m cells that possess a single TCR. munoprecipitated with 1 g of an anti-5-methylcytosine Ab (Eurogentec) at 4˚C overnight. Methylated DNA was purified, and one third of the Dynamic DNA methylation in CD8 T cells has been studied only sample was amplified for 12 cycles by PCR along with the input fraction at the single gene level. The effector cytokine Ifng and inhibitory to generate sequencing libraries. Libraries were quantitated by Agilent receptor Pdcd1 genes are methylated and silenced in naive CD8 BioAnalyzer, and each library was sequenced using a single-end, 50-bp T cells, demethylated and expressed at the effector stage, and protocol on a single lane of a HiSeq 2000 instrument at the Southern California Genotyping Consortium. All sequencing data are available at remethylated when expression is silenced in memory CD8 T cells the Gene Expression Omnibus database (http://www.ncbi.nlm.nih.gov/geo/ (23, 24). Methylation changes at the Pdcd1 locus occurred in key ) under accession number GSE44638. regulatory regions (23, 25). Changes in DNA methylation at cell MeDIP-seq analysis type–specific enhancer regions for Ifng were described during CD4 T cell lineage specification but only at the promoter in CD8 Raw sequencing reads were mapped to the mouse genome (mm9) using T cells (24, 26). These studies show that DNA methylation is Bowtie (31), and only those uniquely mappable and nonredundant reads dynamically regulated during differentiation of CD8 T cells and were used in subsequent analyses. Manipulation, annotation, and analysis of sequence data were performed using HOMER software (32) and custom suggestthatitistargetedtocis-regulatory elements. Following R and PERL scripts, which are available upon request. To identify DMRs, differentiation cues in embryonic stem (ES) cells, DNA-binding we used the MEDIPS package that makes use of a sliding-window factors direct sequence-specific modifications to the epigenome, approach (33). Briefly, we extended each read to 300 bp, such that influencing local epigenetic states, including DNA methylation they represent the average fragment size of immunoprecipitated DNA as determined by the Agilent BioAnalyzer trace. We compared the average (27). Therefore, mapping the sites of differential DNA methylation

enrichment measured as the mean reads per million (rpm) in 10-bp bins, Downloaded from during differentiation could identify novel regulatory elements and for 200-bp sliding windows, and incremented the window 100 bp at a time. the sequences bound by factors that drive epigenetic remodeling. Bin, window, and increment parameters that produced the fewest number Moreover, such maps provide a foundation for how CD8 T cell of DMRs not containing a CpG were chosen after comparison of a range differentiation programs are imprinted within the genome. of values. Regions in which naive and D8 effector CD8 T cell samples did not contain enrichment . 90th percentile of the input (rpm) were elimi- Infection of C57BL/6 mice with the lymphocytic choriome- nated because of lack of enrichment. Regions that contained input, naive, ningitis virus (LCMV) Armstrong strain results in an acute in- or D8 effector enrichment . 99.999th percentile were eliminated as se- fection that typically is cleared within 8 d. At this time point, Ag- quencing artifacts. The p values calculated by MEDIPS were corrected for http://www.jimmunol.org/ specific CD8 T cell numbers have peaked, and the cells have fully multiple hypotheses testing using the Benjamini–Yekutieli correction. A false discovery rate (FDR) , 0.01 was considered significant. Signifi- differentiated into effector cells. To derive an understanding of the cance of DMR enrichment at promoters and enhancers was determined epigenetic mechanisms controlling this differentiation, we mapped by permutation testing. The locations of the promoters or enhancers were genome-wide DNA methylation in naive and day 8 (D8) effector shuffled 1000 times, and the number of DMRs that overlapped the per- CD8 T cells following LCMV Armstrong infection. DNA meth- muted set of promoters or enhancers was recalculated and compared with ylation was globally remodeled during effector cell differentiation, the actual number of overlaps. The p value was equal to the number of times the permuted data had a larger overlap than the actual data, divided and promoter DNA methylation changes correlated inversely with by the permutations. gene expression. We examined active enhancers and observed that by guest on October 1, 2021 an overall gain in methylation occurred at these regions in D8 ef- Bisulfite sequencing fector CD8 T cells, suggesting that an epigenetic silencing mech- Ten nanograms of genomic DNA was treated with sodium bisulfite to anism is functioning at these sites during their differentiation. convert unmethylated cytosines to uracils with the EpiTect Bisulfite Kit, Importantly, we found that differentially methylated regions (DMRs) according to the manufacturer’s protocol (QIAGEN). Bisulfite-converted DNA was amplified by PCR, and the amplicons were cloned using TOPO were enriched for transcription factor motifs that coordinated ef- TA vectors (Invitrogen). Positive colonies were purified and sequenced fector CD8 T cell differentiation and function in response to external (Beckman Coulter Genomics). Bisulfite sequences were aligned to their stimuli. Together, these data identify novel regulatory regions and respective in silico bisulfite-converted genomic sequences using the Bio- transcription factor networks, which suggest that DNA methylation conductor Biostrings R package and custom R scripts, which are available is an important and dynamic epigenetic mechanism contributing to upon request. Only sequences that contained data for all CpGs and with a bisulfite conversion rate . 95% (as determined by conversion of non-CpG the formation and/or maintenance of the effector CD8 T cell state. cytosines to uracil and subsequent amplification as thymine) were included in the analysis. Significance was determined by a one-sided Fisher exact Materials and Methods test. All bisulfite PCR primers are listed in Supplemental Table I. Isolation of Ag-specific CD8 T cells Gene expression and histone data analysis LCMV-specific splenic CD8 T cells were obtained from naive transgenic Differential gene expression between naive and D8 effector CD8 T cells was P14 mice that have a transgenic TCR recognizing the H-2 Db GP33-41 described previously (4). Microarray data were downloaded from the Gene epitope of LCMV (28). Chimeric mice were generated by i.v. adoptive Expression Omnibus database, accession number GSE9650, and genes that transfer of 105 congenically labeled Thy1.1+ naive LCMV-specific CD8 changed .1.5-fold with an FDR , 0.05 were identified by Significance T cells into Thy1.2+ C57BL/6 recipients. Twenty-four hours after adoptive Analysis of Microarrays (34). Histone modification profiles present in the transfer, chimeric mice were infected with 2 3 105 PFU LCMV Arm- thymus were identified by chromatin immunoprecipitation sequencing strong. LCMV-specific effector CD8 T cells were obtained from the spleen (ChIP-seq) as part of the Encyclopedia of DNA elements (ENCODE) 8 d postinfection (D8 effectors). CD8 T cells were purified by FACS using project (35). All ENCODE data sets used in this study are listed in Sup- fluorescently labeled CD90.1 (Thy1.1) and CD8 Abs, as previously de- plemental Table I. Raw ChIP-seq reads were mapped to the mouse genome scribed (1, 5, 18, 24). Naive P14 cells were FACS purified and used as Ag- (mm9) using Bowtie (31), and significantly enriched peaks relative to input specific naive CD8 T cells. All animal experiments were approved by the control were identified by HOMER software (32) using the “histone” Emory University Institutional Animal Care and Use Committee. setting. Overlapping and unique peaks were determined using Bedtools (36). Methyl-DNA immunoprecipitation sequencing Results Methyl-DNA immunoprecipitation sequencing (MeDIP-seq) was per- DNA methylation of naive and D8 effector P14 CD8 T cells formed, as described previously (29, 30). Briefly, 2 mg pooled genomic DNA from three independent isolations of naive and D8 effector CD8 DNA methylation at Ifng and Pdcd1 gene promoters in CD8 T cells and input control were sonicated to an average size of 300 bp, end T cells changed during their differentiation and correlated nega- repaired, and A-tailed, and sequencing adaptors were ligated according tively with transcription of these genes (23, 24), suggesting that The Journal of Immunology 3421 other genes or perhaps many genes within CD8 T cells use DNA positive regulatory elements that become activated or repressed methylation to mediate control of gene expression. To investigate epigenetically. the global role for DNA methylation in CD8 T cells during an DMRs occurred at biologically relevant gene promoters that are acute viral infection, the LCMV Armstrong strain was used as an necessary for effector CD8 T cell function. For example, genes with infection model. P14-transgenic mice harbor a knock-in TCR demethylated promoter DMRs included Gzmb and Zbtb32 (Fig. specific for the H2-Db–restricted GP33-41 epitope of LCMV (28). 1F). The Gzmb gene encodes the serine protease granzyme B, The use of P14-transgenic mice ensure that TCR affinity for which is important for effector CD8 T cell cytotoxic function (42). peptide–MHC I molecules is equal among clonal P14 CD8 T cells In D8 effectors compared with naive cells, Gzmb was strongly during development and an immune response (22). Adoptive demethylated in the gene body and almost completely lacked transfer of naive P14 CD8 T cells into wild-type mice, followed by a methyl-DNA signal at its promoter region. Zbtb32, a transcrip- infection with LCMV Armstrong leads to a robust adaptive im- tion factor induced in activated lymphocytes (43, 44), demon- mune response and viral clearance (1, 5, 37). Thus, naive splenic strated significant demethylation at the transcription start site (CD44lowCD8+) P14 CD8 T cells were purified and adoptively (TSS), gene body, and upstream elements. Cxcr2 and Tcf7 sig- transferred into wild-type mice, which were subsequently infected nificantly gained methylation in D8 effectors compared with naive with LCMV Armstrong. Eight days postinfection, at the peak of cells (Fig. 1G). Cxcr2 encodes a chemokine receptor implicated in the adaptive immune response (1), Ag-specific P14 effector CD8 cellular senescence (45) and contained several methylated DMRs T cells (D8 effector) were isolated based on expression of the that were concentrated at the promoter region. The transcription congenic marker Thy1.1 (Fig. 1A). Naive and D8 effector CD8 factor Tcf7, which is essential for lymphocyte development and T cells were purified in triplicate and were found to exhibit surface differentiation (46), was heavily methylated in the gene body, but expression of key phenotypic markers typical of their respective it did not contain a DMR in the promoter due to the presence of a Downloaded from developmental stage (Supplemental Fig. 1) (1, 4, 37). CpG island, which is likely constitutively unmethylated (Fig. 1E). DNA from naive and D8 effector samples was isolated, and the In hematopoietic development, sequences immediately proximal respective samples were pooled and subjected to methyl-DNA to CpG islands, termed CpG island shores, were shown to exhibit immunoprecipitation (MeDIP) using a 5-methylcytosine specific highly dynamic DNA methylation levels compared with CpG is- Ab. The precipitated DNA and input control were analyzed by deep lands (19). Similarly, we identified DMRs at CpG island shores at sequencing. Sequencing was performed to a depth that generated both Zbtb32 and Tcf7. Overall, the MeDIP-seq data suggested that http://www.jimmunol.org/ roughly 10-fold coverage for half of the mouse genomic CpGs (29, DNA methylation was globally reprogrammed and possibly tar- 30, 38). Analysis of the MeDIP-seq data indicated a robust linear geted to specific regulatory regions that may be important for enrichment at low-density CpG loci that was highly similar be- controlling the expression of genes central to the adaptive immune tween naive and D8 effectors and greater than that observed at response and CD8 T cell function. CpG-dense regions or the input control (Fig. 1B). This is con- sistent with previous observations that areas of low CpG density Bisulfite sequencing validates MeDIP-seq DMRs are typically methylated compared with those regions of high CpG To verify the MeDIP-seq data and provide a single-CpG resolution density, such as CpG islands, which remain unmethylated (27, 33, map of select regions, we validated a number of DMRs by clonal by guest on October 1, 2021 38, 39). Analysis of all mouse CpG islands indicated that the bisulfite sequencing (cBS) from two or three independent isolations majority are unmethylated in naive and D8 effectors (Fig. 1C). of naive and D8 effector CD8 T cells. DMRs in biologically rel- The small percentage of heavily methylated CpG islands primarily evant and statistically representative TSSs were chosen for vali- was located outside of promoter regions, which was noted previ- dation. As indicated by the MeDIP-seq and cBS, genes encoding ously in humans and mice (40, 41). the serine proteases important for CD8 T cell cytotoxic activity, Gzmb and Gzmk, contained promoter DMRs that became almost Global DNA methylation dynamics during differentiation of completely demethylated in D8 effectors (Fig. 2A). cBS of the naive to effector CD8 T cells integrin Itgae gene demonstrated increased methylation directly at To identify DMRs between naive and D8 effectors, we applied a the TSS, whereas the chemokine receptor Ccr7 showed significant genome-wide sliding-window approach to identify methyl-enrichment gains in methylation upstream of the TSS (Fig. 2B). Together, changes . 1.5-fold with an FDR , 0.01 that were enriched above these data validated the methylation differences identified by background (33). This approach identified 296,007 demethylated MeDIP-seq and provided a single-CpG resolution of the dynamics DMRs and 350,666 methylated DMRs in D8 effectors compared at promoter regions for genes important in the immune response. with naive CD8 T cells (Fig. 1D). The changes in promoter meth- ylation for all 25,387 murine Reference Sequence database (RefSeq) DNA methylation correlates inversely with gene expression promoters were analyzed, because promoter-proximal sequences DNA methylation is a repressive epigenetic modification, and it represent focal control points for transcriptional regulation. RefSeq was shown to negatively influence transcription (12, 14). The promoters were annotated for the presence of a DMR, and their CD8 T cell response to LCMV results in global transcriptional change in methylation was plotted with respect to the average remodeling during differentiation, with hundreds of genes differ- methylation enrichment in naive and D8 effectors (Fig. 1E). This entially expressed in D8 effectors compared with the naive state revealed that 54% of RefSeq promoters contained at least one (4). To investigate the relationship between DNA methylation and DMR (29% methylated, 14% demethylated, and 11% both). The gene expression in CD8 T cells, we correlated the change in overlap of both methylated and demethylated DMRs with promoters promoter methylation in D8 effectors with known transcriptional was greater than that expected by chance (p , 0.001). Promoters changes in this system. DNA methylation demonstrated a signifi- that contained a DMR displayed various levels of methylation cant negative correlation with gene expression (Spearman rank change but overall trended in the same direction as the DMR correlation r = 20.41, p , 0.001) (Fig. 3A). Genes that increased that they overlapped. The 11% of promoters that included both in expression largely were demethylated at their promoters (upper methylated and demethylated DMRs displayed the largest left quadrant), whereas genes that became transcriptionally si- variation in methylation changes. The presence of both DMRs lenced gained methylation in D8 effectors (lower right quadrant). suggested that these promoters might contain both negative and For example, the chemokine receptor Ccr7 was significantly 3422 EFFECTOR CD8 T CELL EPIGENETICS Downloaded from http://www.jimmunol.org/ by guest on October 1, 2021

FIGURE 1. DNA methylation is globally dynamic during effector CD8 T cell differentiation in response to LCMV Armstrong infection. (A) Repre- sentative plots of naive CD8 T cells (CD8+CD44low) isolated by FACS from P14-transgenic mice and adoptively transferred into wild-type hosts and infected with LCMV Armstrong. Eight days postinfection, Ag-specific effector CD8 T cells (D8 effectors; CD8+Thy1.1+) were purified by FACS. (B) The density of CpGs in 300-bp windows across the genome was calculated, and MeDIP-seq data in each window were plotted as rpm. MeDIP-seq signal was consistently enriched at low-density CpG regions in both naive and D8 effectors. No enrichment was observed for input control. (C) Heat map showing MeDIP-seq read density surrounding all 16,206 murine CpG islands. Each row represents 5 kb surrounding one CpG island, with read density normalized to rpm. Rows are clustered hierarchically, and the right annotation bar denotes CpG islands that overlap a promoter. (D) Heat map of MeDIP-seq read density at 646,673 DMRs identified between naive and D8 effector CD8 T cells; 296,007 DMRs became demethylated and 350,666 became methylated .1.5-fold during D8 effector differentiation. Each row represents 5 kb surrounding one DMR with read density normalized to rpm. Rows are sorted by decreasing change in methyl-DNA enrichment. (E) Scatter plot of methylation difference (D8 effector 2 naive) versus average MeDIP-seq tag densities at all 25,387 murine RefSeq promoters. Promoter regions were defined as the upstream 2000 bp and downstream 500 bp from the TSS of each gene. RefSeq promoters that overlapped a DMR are color coded, and the percentage of total DMRs that overlap promoters is portrayed in a pie chart. (F) The genes for the CD8 T cell effector protease Gzmb and the transcription factor Zbtb32 demonstrating significant demethylation changes are plotted depicting the location of the DMRs (blue), CpGs (gray), CpG islands (CpGi – yellow), gene structure, and read density for naive (orange) and D8 effectors (green). (G) The chemokine receptor Cxcr2 and T cell developmental transcription factor Tcf7 are significantly methylated in D8 effectors. Each gene is plotted as above, showing methylated DMRs (red). methylated and repressed following CD8 T cell differentiation, naive and D8 effectors were grouped according to functional a change that we validated by cBS (Fig. 2B). The Ifng promoter classes, as previously described (4), and the association of pro- was one of the strongest demethylated and induced genes, a result moter methylation change with gene expression was plotted. DNA that supported previous observations in CD8 and CD4 T cells (24, methylation at promoters correlated inversely with gene expres- 26). Interestingly, the chemokine receptor Ccr5 and its ligand sion within functional groups, suggesting that these genes may be Ccl5 were both robustly upregulated and demethylated. regulated by coordinated epigenetic mechanisms (Fig. 3B). For The above results suggest that functional families of genes may example, three inhibitory receptor genes (Klrg1, Klrc1, and Ctla4) be coordinately regulated by DNA methylation. To determine whether were upregulated and demethylated in D8 effectors. Genes im- this was the case, all genes that were differentially expressed between portant for homing and migration displayed the most significant The Journal of Immunology 3423

FIGURE 2. Bisulfite sequencing–validated MeDIP-seq methylation changes. A representative set of methylated and demethylated DMRs were validated by bisulfite sequencing. Two loci that contained demethylated DMRs Gzmk and Gzmb (A) and two loci Itgae and Ccr7 that exhibited methylated DMRs (B) Downloaded from were chosen. For each gene, the location of bisulfite primers, DMRs, and CpGs are indicated in relation to the TSS. Each line of bisulfite sequencing data represents an individual clone, with black circles denoting methylated CpGs and open circles representing unmethylated CpGs. The location of the DMR analyzed is diagramed above each bisulfite data set. All loci were validated from two or three independent samples of naive and D8 effector time points and demonstrated statistically significant changes according to a one-sided Fisher exact test.

trend, because the majority of induced genes were demethylated. inverse correlation between changes in promoter DNA methyla- http://www.jimmunol.org/ Ccr7 and Ccr9, two cytokine receptor genes important for naive tion and gene-expression levels. The relationship between DNA T cell development and homeostasis, were both downregulated methylation and gene expression implicates a key role for epi- and methylated. Additionally, genes involved in apoptosis, such as genetics in establishing the D8 effector transcriptional program. Casp1, and the calcium-signaling proteins Anxa1 and Anxa2 were upregulated and demethylated and may play a role in promoting Thymic enhancers gain DNA methylation in effector CD8 cell death of effector CD8 T cells at later stages of the process. T cells Transcription factors represent essential components of T cell The ENCODE project mapped histone modifications in whole effector function that dynamically integrate TCR signals and ex- thymus, an organ that harbors the developing T cell lineage ternal stimuli to drive cellular differentiation. The Wnt pathway is (Supplemental Table I) (35, 54). We hypothesized that histone by guest on October 1, 2021 necessary for T cell development and homeostasis and is repressed patterns in the thymus would provide insight into loci important at the effector stage (46–48). Three members of the Wnt-signaling for T cell development and/or maintenance of a naive T cell. The pathway (Tcf7, Lef1, and Cxxc5) were downregulated and meth- histone H3 lysine 4 trimethylation modification (H3K4me3) marks ylated. Interestingly, Batf, which modulates energy metabolism the TSS of genes and is positively associated with transcription and mediates T cell differentiation and exhaustion (49, 50), was (55) and negatively associated with DNA methylation (56, 57). upregulated and demethylated along with Bhlhe40, a factor in- Conversely, histone H3 lysine 27 trimethylation (H3K27me3)is volved in cellular differentiation and circadian rhythm (51–53). a repressive epigenetic modification and has been mechanistically DNA methylation did not completely correlate with gene ex- linked to DNA methylation at a number of loci (58). We charac- pression in all instances. For example, the inhibitory receptor terized the average MeDIP-seq signal at regions of H3K4me3 or Gp49a was upregulated but gained DNA methylation in its pro- H3K27me3 in the thymus. DNA methylation was significantly moter (Fig. 3B, upper left panel). These changes could be due to enriched at loci containing repressive histone marks in both naive the methylation of inhibitory sequences or a temporal delay in the and D8 effectors compared with regions that contain H3K4me3 acquisition of repressive epigenetic marks and effects on mRNA (Fig. 4A). Additionally, D8 effectors contained significantly more levels. Nevertheless, the majority of genes demonstrated a significant DNA methylation at regions of H3K4me3 than did naive cells,

FIGURE 3. Promoter DNA methylation correlates inversely with gene expression. (A) Scatter plot showing the mean change in promoter methyl-DNA enrichment (rpm) relative to gene expression fold change for all genes differentially expressed between naive and D8 effector CD8 T cells. Promoter regions were defined as the upstream 2000 bp and downstream 500 bp from the TSS of each gene. Gene expression correlated negatively with promoter methylation (Spearman rank correlation r = 20.41, p , 0.001). (B) Scatter plots of methyl- DNA enrichment difference and gene expression fold change for four functional groups of genes important to effector CD8 T cell function. 3424 EFFECTOR CD8 T CELL EPIGENETICS Downloaded from http://www.jimmunol.org/

FIGURE 4. Enhancers active in the thymus become methylated in D8 effector CD8 T cells. (A) Histone maps generated from the thymus were me3 downloaded from the ENCODE project (35). Regions containing the repressive histone modification H3K27 were significantly enriched for DNA by guest on October 1, 2021 methylation compared with regions with the active histone marker H3K4me3.(B) Heat maps of H3K4me3, H3K4me1, and H3K27ac read density (rpm) at 19,365 regions in the adult mouse thymus. Each region is annotated for proximity (within 1 kb) to a TSS with a black bar. Proximal promoter elements (top) contain H3K4me3, whereas PDEs (bottom) lack H3K4me3 and are marked by high levels of H3K4me1 and H3K27ac. Rows for both proximal promoter elements and PDEs were hierarchically clustered separately across all three histone modifications. (C) Methylation changes at all 5471 active thymic PDEs showed a trend toward becoming methylated in D8 effectors. Methylation change is calculated as average MeDIP rpm in D8 effectors minus naive for each PDE. Gene annotations were predicted based on mapping each element to the closest TSS. (D) PDEs primarily overlap methylated DMRs. The overlap of DMRs and PDEs was computed, and percentages are indicated in a pie chart. Asterisk indicates significant overlap of methylated DMRs with PDE regions greater than random chance (p , 0.001). (E) Box plot of absolute methylation difference in PDEs that overlap methylated and demethylated DMRs. PDEs that overlapped methylated DMRs had significantly greater changes in methylation compared with those that overlapped demethylated DMRs. Significance was calculated using the Student t test. (F) Density plot summarizing the distribution of PDEs located within the gene body or surrounding 10 kb of genes differentially expressed between naive and D8 effector CD8 T cells. Location of PDEs within gene bodies was calculated as the distance from the TSS, and the position was normalized to a percentage of the gene body length. (G) Scatter plot displaying the mean change in PDE DNA methylation relative to gene expression fold change for all genes differentially expressed and annotated with a PDE. Select PDEs that map to CD8-relevant genes are highlighted in the key. *p , 10217,**p , 10232, Student t test. n.s., Not significant; TTS, transcription termination sequence. consistent with observations that more TSSs gained DNA meth- enriched for both H3K4me1 and H3K27ac modifications. The regions ylation in D8 effectors (Fig. 1E). DNA methylation at regions were then stratified by the presence or absence of H3K4me3.This enriched for H3K27me3 in the thymus were globally unchanged, approach established 5670 PDEs, 99% of which were $2kbaway suggesting that these regions are permanently repressed once they from the nearest TSS (Fig. 4B). To investigate the role for DNA acquire H3K27me3. methylation at thymic enhancers, we analyzed the average change in Transcriptional enhancers provide critical control points for cell DNA methylation from naive to D8 effectors. Although PDEs be- type–specific gene expression and were shown to undergo epige- came both methylated and demethylated, there was a trend toward netic changes, including DNA methylation, during cellular differ- more enhancers gaining DNA methylation in D8 effectors (Fig. 4C). entiation (27, 57, 59). For example, during ES cell differentiation, PDEs methylated in D8 effectors were identified around genes that DNA methylation was found to be more dynamic at distal regula- became repressed, such as Tcf7 and Ccr7. In addition, genes tory regions outside of core promoters (57), suggesting that similar that became activated in D8 effectors were associated with PDEs events may also occur in differentiating T cells. The presence of that became demethylated, such as Gzmb. We annotated the PDEs both H3K4me1 and H3K27 acetylated (H3K27ac) histone modi- for overlap with DMRs and found that 36.3% of the PDEs active fications in the absence of H3K4me3 separates active distal en- in the thymus overlapped with a methylated DMR, whereas only hancers from promoter-proximal TSS elements (60). Putative distal 8% overlapped with a demethylated DMR (Fig. 4D). The overlap enhancers (PDEs) in thymic cells were identified by mapping regions of methylated DMRs was more than expected by random chance The Journal of Immunology 3425

(p , 0.001), whereas there was not a significant enrichment of demethylated DMRs in thymic enhancers. Additionally, the change in MeDIP-seq signal was significantly greater at PDEs that over- lapped a methylated DMR than a demethylated DMR (Fig. 4E). These data indicate that enhancers active in the thymus robustly gain DNA methylation in D8 effectors compared with naive cells. To assess the relationship between the DNA methylation status of the PDEs and the expression of the genes that were activated or repressed in this system, PDEs were mapped to gene bodies and to the sequences 10 kb upstream of the TSS or downstream of the transcription termination sequence. A total of 164 PDEs was identified within the 10-kb limit for 76 differentially expressed genes. PDEs were enriched inside the gene body and were depleted around transcription start and termination sites (Fig. 4F). Upstream and downstream PDEs demonstrated a biphasic distribution, suggesting that both short-range (,5 kb) and long-range (.5 kb) enhancers exist for these genes. Next, changes in PDE DNA methylation were correlated with changes in gene expression (Fig. 4G). Inverse correlations between PDE DNA methylation and gene expression were observed for many of the genes, with several Downloaded from genes having multiple PDEs that changed in a similar manner (Lef1, Rgs10, Ccr7, and Sell). The annotation of these genes with PDEs potentially identified an additional layer of epigenetic reg- ulation outside of promoter regions and suggests that DNA meth- ylation may play an important role in regulating enhancer activity FIGURE 5. DMRs are enriched for transcription factor (TF) binding during effector CD8 T cell differentiation. sites important for naive and effector function. Transcription factor motifs http://www.jimmunol.org/ significantly enriched in either methylated DMRs (A) or demethylated DMRs are enriched for functional transcription factor motifs DMRs (B) were identified. The binding motif, transcription factor, and We demonstrated that DNA methylation dynamically changed at DNA-binding domain family are indicated for the most significant site for each motif. Enriched motifs in methylated DMRs are mainly develop- both distal enhancers and promoters for genes differentially ex- mental transcription factors that may be important for naive T cell matu- pressed in CD8 T cells in response to LCMV infection. These ration or homeostasis. Demethylated DMRs were enriched for motifs of data predicted that DMRs contain cis-elements and could be transcription factors that signal from the TCR and respond to external bound by transcription factors important in CD8 T cell function, stimuli. bZIP, Basic leucine zipper; ETS, E-twenty six; HLH, helix-loop- as well as that demethylated and methylated DMRs may have helix; HMG, high-mobility group; MH1, MAD homology 1; NR, nuclear by guest on October 1, 2021 unique functions. Using consensus DNA-binding sequences termed receptor; RHD, Rel homology domain; ZF, zinc finger. “logos,” which were generated from published ChIP-seq experi- ments, we searched for transcription factor motifs that were signifi- cantly enriched in the methylated DMRs versus the demethylated To identify transcription factor modules that cooperate to control DMRs. DMRs methylated in D8 effectors contained motifs for effector CD8 T cell function, we identified motifs that co-occurred developmental transcription factors, such as the ETS (GABPA), within the same DMRs by correlating the number of motifs that High Mobility Group (SOX3 and TCF4), Forkhead (FOXA1), and occur in each DMR between any two transcription factors. Tran- Zinc Finger (ZFX and BCL6) families (Fig. 5A, Supplemental scription factor motif co-occurrence, measured by Spearman rank Table II). Methylation of these sites may be a repressive epige- correlation r, was used to hierarchically cluster transcription netic mechanism to silence these transcriptional networks that are factors, revealing modules commonly bound at the same DMRs active in the developing or naive T cell. In contrast, DMRs deme- (Fig. 6A). This analysis identified eight modules enriched in the thylated in D8 effectors were highly enriched for transcription demethylated DMRs (Fig. 6A, Table I). Six of the eight modules factor families known to play roles in the effector response, such as correlated along the axis and contained motifs for the calcineurin the bZIP (c-JUN and CEBPB), Rel Homology Domain (NFATc1 pathway factor NFAT, AP-1 transcription factors c-JUN and and NFKB), and IRF (IRF4) families (Fig. 5B, Supplemental JUND, nuclear receptors, homeobox factors, and helix-loop-helix Table II). Loss of methylation at these DMRs may lead to an DNA-binding proteins. Additionally, GATA3 interacted with the accessible and open chromatin environment that facilitates binding AP-1 factors in Module 3, and HIF1A and NRF1 were predicted of these factors and promote effector differentiation and function in to interact with the helix-loop-helix factors in Module 8. Tran- response to external stimuli. DNA methylation was reported to scription factor motifs enriched in methylated DMRs also cluster prevent DNA binding for a number of proteins (61–65). Separate into five modules designated as A through E (Supplemental Fig. from correlating with a repressive local chromatin environment, 2A). These modules included factors belonging to the ETS and DNA methylation may physically restrict DNA binding for those STAT family of transcription factors (Supplemental Fig. 2B). factors that contain a CpG in their recognition motif, providing To determine the biological function of each module, the sig- multiple epigenetic mechanisms to regulate transcription factor nificantly enriched Gene Ontology (GO) terms present in the genes activity. regulated by each module were identified. GO terms were clustered to identify shared processes between transcription factor modules Transcription factor modules in D8 effector function (Fig. 6B). For the modules identified in the methylated DMRs, no Transcription factors cooperatively bind cis-elements to regulate significant GO terms could be found to assign specific functions to transcription in multifactor modules, thereby integrating multiple those modules. In contrast, with the exception of Module 7, all of signaling pathways to regulate a biological process (32, 66, 67). the modules in demethylated DMRs shared broad overlapping 3426 EFFECTOR CD8 T CELL EPIGENETICS

chemotaxis was uncovered. Module 1 was involved in diverse cel- lular processes, such as differentiation, leukocyte activation, pro- grammed cell death, and leukocyte proliferation. Module 1 integrated signals from the cAMP pathway through ATF3 and E-box binding factors MYC, MAX, USF1, and BHLHE40. Although roles in the immune response have been described for the other factors in module 1 (69–72), no functional role has been described for BHLHE40 in CD8 T cells. Interestingly, Modules 2 and 3 shared the AP-1 factors; however, Module 2 uniquely contained CREB, whereas Module 3 contained NFE2. Distinct functions for Module 2 included leukocyte and cell activation, whereas unique processes for Module 3 included cell growth, leukocyte proliferation, and signal transduction. These results suggest separate roles for the CREB-AP1moduleinpromotingTcellactivation,whereasthe MAPK-AP1 pathways are predicted to regulate T cell expansion and proliferation.

Discussion The differentiation from naive to effector CD8 T cells following acute infection is a critical step in the adaptive immune response. Downloaded from The gene-expression changes that occur following differentiation cues have been studied (2, 4); however, little is known about how the epigenome is reprogrammed to permit the acquisition of new cellular functions. In this article, we report the genome-wide DNA methylation profile of naive and D8 effector CD8 T cells using MeDIP-seq. Consistent with the scale of epigenetic changes pre- http://www.jimmunol.org/ viously reported during ES cell differentiation (57), we find ex- tensive global reprogramming of DNA methylation during effector CD8 T cell differentiation. Gene expression and promoter DNA methylation were negatively correlated, demonstrating that pre- vious findings at the Ifng (24) and Pdcd1 (23) loci defined a broad FIGURE 6. Transcription factor modules cooperate to promote effector epigenetic mechanism for gene regulation in CD8 T cells. Meth- CD8 T cell function. (A) Transcription factor co-occurrence at demethy- ylated and demethylated DMRs were enriched in 54% of murine RefSeq promoters, which greatly exceeds the known transcrip- lated DMRs was clustered hierarchically to identify factors that potentially by guest on October 1, 2021 cooperate in modules at the same DMR. Co-occurrence is calculated by tional reprogramming events during CD8 T cell differentiation. Spearman rank correlation of the number of transcription factor binding Changes in DNA methylation may result in the fine-tuning of gene sites in each DMR between any 2 of the 35 significant transcription factor expression that is not detectable by microarray technology, a shift motifs. Eight modules used in subsequent analyses are outlined in red. (B) to a new set of cis-elements that drive/inhibit gene expression in DMRs containing each module were mapped to the nearest gene, and GO D8 effectors and not naive cells, and/or define chromatin structural analysis was performed to identify biological processes that each module changes that serve to maintain the D8 effector cell fate program. affected. Significant GO terms enriched in any of the eight modules were In this vein, we identified 76 genes with expression changes that clustered and shared, and unique processes are depicted in a heat map. Each row represents the presence or absence of the indicated GO term, and could be annotated to one or more PDEs whose DNA methylation each column is a transcription factor module. status changed between naive and D8 effector CD8 T cells. Thus, the correlation of gene expression, promoter DNA methylation, and PDE DNA methylation, implies that the epigenome is actively functions with roles in processes, such as cellular differentiation, reprogrammed as CD8 T cells differentiate to facilitate the emer- metabolism, and cell motility, and were analyzed further. Six gence of new gene-expression patterns and effector CD8 T cell modules were enriched for the regulation of gene expression on- function. tology, suggesting that they play a role in integrating signals that The finding that some DMRs were enriched for functional fine-tune downstream transcriptional networks. Module 6, con- transcription factor motifs provided an additional layer of epige- sisting of nuclear receptors, was enriched for known functions in netic regulation for DNA methylation in CD8 T cell differentiation. apoptosis (68); however, interestingly, a putative novel role in Overall, consensus motifs that were found in demethylated DMRs in D8 effectors were correlated with transcription factors that Table I. Transcription factor modules present in demethylated DMRs contribute to the effector function, such as NFATc1 and c-JUN. Conversely, motifs in DMRs that were methylated in D8 effec- Module Transcription Factor Motifs tors were mostly associated with transcription factors that function 1 ATF3, USF1, BHLHE40, c-MYC, MAX in developmental/differentiation pathways. Thus, the changes in 2 CRE, JUND, c-JUN DNA methylation appear to coincide with a cell fate program that 3 NFE2, c-JUN, AP-1 has shifted from the ability to differentiate cells to one that is 4 NANOG, HOXB4, PDX1 5 NFAT-AP-1, NFATC1 focused on effector outcome and function. 6 NR4A1, RXR The unique transcription factor motifs enriched in demethylated 7 GATA3, JUND, c-JUN and methylated DMRs indicated that changes in DNA methylation 8 HIF1A, NRF1, ATF3, USF1, BHLHE40, c-MYC, MAX might restrict the accessibility to DNA for transcription factors. CRE, cAMP responsive element. In fact, the physical binding of c-JUN, JUND (65), c-MYC (64), The Journal of Immunology 3427

CREB/ATF (62), CTCF (61), and ETS1 (63) to DNA was dem- The H3K4me1/2 histone demethylase LSD1 inactivates enhancers onstrated previously to be methylation sensitive. Interestingly, during ES cell differentiation (75) and may be active in this sys- these factors are ubiquitously expressed during CD8 T cell dif- tem to prepare these sites for de novo methylation. Additionally, ferentiation (2, 4). Therefore, as CD8 T cells differentiate into the transcriptional repressor E2F6 can recruit DNMT3B directly effector cells and the DNA methylation landscape is remodeled, so to DNA (76), suggesting that alternative chromatin-independent too are the accessible binding sites for these factors. This predicts mechanisms exist for the acquisition of novel DNA methylation that a precise order of events occurs for inactive (methylated) patterns. elements in naive CD8 T cells as they transition into an active The loss of DNA methylation in D8 effectors could have oc- state (demethylated) in D8 effectors. For example, in CD8 T cells, curred via two general processes: passive and active. Following the Pdcd1 (PD-1) locus undergoes such a process. In naive CD8 activation, CD8 T cells expand exponentially, allowing simple T cells, Pdcd1 is extensively methylated and silent (23, 25). Pdcd1 failure of DNMT1 to remethylate CpGs as a possible mechanism. is induced following the binding of NFATc1 to its cis-element Passive demethylation can also occur by the recruitment of factors (25), suggesting that NFATc1, which contains no CpG in its DNA that block DNMT1 binding, such as GATA3 (77). Evidence for binding site, can access the Pdcd1 DNA and initiate expression active demethylation also exists, because demethylation of the Il-2 of Pdcd1. Upon NFATc1 binding, the chromatin structure of the promoter in CD4 T cells and the Ifng promoter in memory CD8 region adopts an active conformation (i.e., histone H3 and H4 T cells was observed during experimental conditions of cell cycle acetylation), and during the initial stages of the effector response arrest, suggesting an active process at these sites (24, 78). The the region loses its DNA methylation (23, 25). Likewise, many of TET family of enzymes can hydroxylate 5-methylcytosine to the enriched transcription factor motifs do not contain a CpG in begin the process of demethylation that uses the base-excision their binding motif. Therefore, there may be a temporal sequence repair pathway and culminates in an unmodified cytosine (79– Downloaded from of transcription factor binding events that first initiates an open 81). The recent discovery and characterization of these enzymes chromatin conformation and demethylation of the DNA. This provide credence to the notion that active DNA demethylation would be followed by the binding of DNA methylation–sensitive can occur. Although NANOG can recruit TET1 and TET2 to pro- transcription factors and new gene-expression patterns. moters during induced pluripotent stem cell reprogramming (82), Transcription factor networks integrate external signaling path- no targeting mechanism has been identified for locus-specific active ways to modulate gene expression and initiate cellular differentia- demethylation in CD8 T cells. A cellular demand for rapid gene- http://www.jimmunol.org/ tion. Originally termed enhanceosomes (66), cis-acting regulatory expression kinetics may require one method over another, but both elements frequently contain multiple transcription factor binding active and passive mechanisms are likely occurring at the deme- sites that form a module to integrate signaling pathways, a process thylated DMRs in CD8 T cells. that has been best studied in ES cells (67) and between macro- The data presented in this article revealed the global acquisition phages and B cells (32). Our DNA sequence–centric approach of novel DNA methylation patterns in effector CD8 T cells that differs from previous transcription-focused module analyses in CD8 facilitated the expression of the effector phenotype while repressing T cells that analyzed gene-expression correlations (2). Although our the naive transcriptional program. Importantly, the sites of DNA prediction of DNA-binding motifs relied on high-quality ChIP-seq methylation remodeling identified both trans- and cis-factors im- by guest on October 1, 2021 data that are not available for every transcription factor, we iden- portant for naive and effector CD8 T cell identity and function. tified transcription factor modules of both known and novel function Therefore, these data have drawn a blueprint for the foundation of in effector CD8 T cells. For example, Module 5 consisted of the one of the epigenetic programs associated with CD8 T cell dif- known NFAT–AP-1 interaction (73). In addition, this approach ferentiation. Further understanding of the mechanisms that write, allowed us to dissect out two roles for the AP-1 family factors in erase, and interpret this blueprint will ultimately be important for modulating calcium- and MAPK-signaling pathways. Many mod- manipulating the epigenome of these cells for the creation of ules participated in overlapping functions, suggesting that a com- novel vaccines and therapeutic treatments involving CD8 T cells. plex integration of signaling networks may be required to maintain effector function during clonal expansion. Moreover, our results Acknowledgments predict broader roles in CD8 T cell function for a nuclear receptor We thank members of the Boss and Ahmed laboratories for helpful com- module consisting of NR4A1 and RXR. The nuclear receptors may ments and critiques, as well as the Southern California Genotyping Consor- regulate integral CD8 T cell processes, such as activation, prolif- tium for sequencing and the Emory University Flow Core for cell isolation. eration, metabolism, and chemotaxis. These findings highlight the complex interaction of signaling pathways that is required to main- Disclosures tain a highly dynamic and metabolically active cell population. The authors have no financial conflicts of interest. Our experiments revealed that both gain and loss of DNA methylation occurred during CD8 T cell differentiation. The ac- References quisition of new DNA methylation implies that the de novo DNA 1. Blattman, J. N., R. Antia, D. J. Sourdive, X. Wang, S. M. Kaech, K. Murali- methyltransferases DNMT3A and 3B were active during this Krishna, J. D. Altman, and R. Ahmed. 2002. Estimating the precursor frequency process. Indeed, gene-expression analyses for the DNMTs showed of naive antigen-specific CD8 T cells. J. Exp. Med. 195: 657–664. 2. Doering, T. A., A. Crawford, J. M. Angelosanto, M. A. Paley, C. G. Ziegler, and that, during effector differentiation, all components of the meth- E. J. Wherry. 2012. Network analysis reveals centrally connected genes and ylation machinery were expressed (2, 4, 23). DNMT3A and pathways involved in CD8+ T cell exhaustion versus memory. Immunity 37: 1130– DNMT3B can be recruited to DNA by interactions with the his- 1144. 3. Parish, I. A., and S. M. Kaech. 2009. Diversity in CD8(+) T cell differentiation. tone methyltransferase G9a (74) or via interaction with DNMT3L Curr. Opin. Immunol. 21: 291–297. (56). The DNMT3A–DNMT3L complex binds chromatin con- 4. Wherry, E. J., S. J. Ha, S. M. Kaech, W. N. Haining, S. Sarkar, V. Kalia, taining unmethylated histone H3K4 (H3K4me0) (56). Active TSSs S. Subramaniam, J. N. Blattman, D. L. Barber, and R. Ahmed. 2007. Molecular me3 signature of CD8+ T cell exhaustion during chronic viral infection. Immunity 27: and PDEs are surrounded by nucleosomes that are H3K4 and 670–684. H3K4me1 modified, respectively, which implies that there is a 5. Murali-Krishna, K., J. D. Altman, M. Suresh, D. J. Sourdive, A. J. Zajac, J. D. Miller, J. Slansky, and R. Ahmed. 1998. Counting antigen-specific CD8 stepwise process that must remove the H3K4 methylation before T cells: a reevaluation of bystander activation during viral infection. Immunity 8: the region can become a strong substrate for de novo methylation. 177–187. 3428 EFFECTOR CD8 T CELL EPIGENETICS

6. Kallies, A., A. Xin, G. T. Belz, and S. L. Nutt. 2009. Blimp-1 transcription factor 34. Tusher, V. G., R. Tibshirani, and G. Chu. 2001. Significance analysis of micro- is required for the differentiation of effector CD8(+) T cells and memory arrays applied to the ionizing radiation response. Proc. Natl. Acad. Sci. USA 98: responses. Immunity 31: 283–295. 5116–5121. 7. Intlekofer, A. M., A. Banerjee, N. Takemoto, S. M. Gordon, C. S. Dejong, 35. ENCODE Project Consortium. 2011. A user’s guide to the encyclopedia of DNA H. Shin, C. A. Hunter, E. J. Wherry, T. Lindsten, and S. L. Reiner. 2008. elements (ENCODE). PLoS Biol. 9: e1001046. Anomalous type 17 response to viral infection by CD8+ T cells lacking T-bet 36. Quinlan, A. R., and I. M. Hall. 2010. BEDTools: a flexible suite of utilities for and eomesodermin. Science 321: 408–411. comparing genomic features. Bioinformatics 26: 841–842. 8. Bird, A. 2002. DNA methylation patterns and epigenetic memory. Genes Dev. 37. Wherry, E. J., J. N. Blattman, K. Murali-Krishna, R. van der Most, and 16: 6–21. R. Ahmed. 2003. Viral persistence alters CD8 T-cell immunodominance and 9. Moazed, D. 2011. Mechanisms for the inheritance of chromatin states. Cell 146: tissue distribution and results in distinct stages of functional impairment. J. Virol. 510–518. 77: 4911–4927. 10. Goll, M. G., and T. H. Bestor. 2005. Eukaryotic cytosine methyltransferases. 38. Down, T. A., V. K. Rakyan, D. J. Turner, P. Flicek, H. Li, E. Kulesha, S. Gra¨f, Annu. Rev. Biochem. 74: 481–514. N. Johnson, J. Herrero, E. M. Tomazou, et al. 2008. A Bayesian deconvolution 11. De Carvalho, D. D., J. S. You, and P. A. Jones. 2010. DNA methylation and strategy for immunoprecipitation-based DNA methylome analysis. Nat. Bio- cellular reprogramming. Trends Cell Biol. 20: 609–617. technol. 26: 779–785. 12. Smith, Z. D., and A. Meissner. 2013. DNA methylation: roles in mammalian 39. Weber, M., I. Hellmann, M. B. Stadler, L. Ramos, S. Pa¨a¨bo, M. Rebhan, and development. Nat. Rev. Genet. 14: 204–220. D. Schu¨beler. 2007. Distribution, silencing potential and evolutionary impact of 13. Wade, P. A. 2001. Methyl CpG-binding proteins and transcriptional repression. promoter DNA methylation in the human genome. Nat. Genet. 39: 457–466. Bioessays 23: 1131–1137. 40. Hughes, T., R. Webb, Y. Fei, J. D. Wren, and A. H. Sawalha. 2010. DNA 14. Bird, A. P., and A. P. Wolffe. 1999. Methylation-induced repression—belts, methylome in human CD4+ T cells identifies transcriptionally repressive and braces, and chromatin. Cell 99: 451–454. non-repressive methylation peaks. Genes Immun. 11: 554–560. 15. Lee, P. P., D. R. Fitzpatrick, C. Beard, H. K. Jessup, S. Lehar, K. W. Makar, 41. Illingworth, R. S., U. Gruenewald-Schneider, S. Webb, A. R. Kerr, K. D. James, M. Pe´rez-Melgosa, M. T. Sweetser, M. S. Schlissel, S. Nguyen, et al. 2001. A D. J. Turner, C. Smith, D. J. Harrison, R. Andrews, and A. P. Bird. 2010. Orphan critical role for Dnmt1 and DNA methylation in T cell development, function, CpG islands identify numerous conserved promoters in the mammalian genome. and survival. Immunity 15: 763–774. PLoS Genet. 6: e1001134. 16. Thomas, R. M., C. J. Gamper, B. H. Ladle, J. D. Powell, and A. D. Wells. 2012. 42. Hatcher, V. B., M. S. Oberman, G. S. Lazarus, and A. I. Grayzel. 1978. A cy- Downloaded from De novo DNA methylation is required to restrict T helper lineage plasticity. J. totoxic proteinase isolated from human lymphocytes. J. Immunol. 120: 665–670. Biol. Chem. 287: 22900–22909. 43. Miaw, S. C., A. Choi, E. Yu, H. Kishikawa, and I. C. Ho. 2000. ROG, repressor 17. Chappell, C., C. Beard, J. Altman, R. Jaenisch, and J. Jacob. 2006. DNA of GATA, regulates the expression of cytokine genes. Immunity 12: 323–333. methylation by DNA methyltransferase 1 is critical for effector CD8 T cell 44. Yoon, H. S., C. D. Scharer, P. Majumder, C. W. Davis, R. Butler, W. Zinzow- expansion. J. Immunol. 176: 4562–4572. Kramer, I. Skountzou, D. G. Koutsonanos, R. Ahmed, and J. M. Boss. 2012. 18. Kersh, E. N. 2006. Impaired memory CD8 T cell development in the absence of ZBTB32 is an early repressor of the CIITA and MHC class II gene expression methyl-CpG-binding domain protein 2. J. Immunol. 177: 3821–3826. during B cell differentiation to plasma cells. J. Immunol. 189: 2393–2403. 19. Ji, H., L. I. Ehrlich, J. Seita, P. Murakami, A. Doi, P. Lindau, H. Lee, 45. Acosta, J. C., A. O’Loghlen, A. Banito, M. V. Guijarro, A. Augert, S. Raguz, M. J. Aryee, R. A. Irizarry, K. Kim, et al. 2010. Comprehensive methylome map M. Fumagalli, M. Da Costa, C. Brown, N. Popov, et al. 2008. Chemokine sig- http://www.jimmunol.org/ of lineage commitment from haematopoietic progenitors. Nature 467: 338–342. naling via the CXCR2 receptor reinforces senescence. Cell 133: 1006–1018. 20. Shearstone, J. R., R. Pop, C. Bock, P. Boyle, A. Meissner, and M. Socolovsky. 46. Verbeek, S., D. Izon, F. Hofhuis, E. Robanus-Maandag, H. te Riele, M. van de 2011. Global DNA demethylation during mouse erythropoiesis in vivo. Science Wetering, M. Oosterwegel, A. Wilson, H. R. MacDonald, and H. Clevers. 1995. 334: 799–802. An HMG-box-containing T-cell factor required for thymocyte differentiation. 21. Schmidl, C., M. Klug, T. J. Boeld, R. Andreesen, P. Hoffmann, M. Edinger, and Nature 374: 70–74. M. Rehli. 2009. Lineage-specific DNA methylation in T cells correlates with 47. Willinger, T., T. Freeman, M. Herbert, H. Hasegawa, A. J. McMichael, and histone methylation and enhancer activity. Genome Res. 19: 1165–1174. M. F. Callan. 2006. Human naive CD8 T cells down-regulate expression of the 22. Ohkura, N., M. Hamaguchi, H. Morikawa, K. Sugimura, A. Tanaka, Y. Ito, WNT pathway transcription factors lymphoid enhancer binding factor 1 and M. Osaki, Y. Tanaka, R. Yamashita, N. Nakano, et al. 2012. T cell receptor transcription factor 7 (T cell factor-1) following antigen encounter in vitro and stimulation-induced epigenetic changes and Foxp3 expression are independent and in vivo. J. Immunol. 176: 1439–1446. complementary events required for Treg cell development. Immunity 37: 785–799. 48. Yu, S., X. Zhou, F. C. Steinke, C. Liu, S. C. Chen, O. Zagorodna, X. Jing, 23. Youngblood, B., K. J. Oestreich, S. J. Ha, J. Duraiswamy, R. S. Akondy, Y. Yokota, D. K. Meyerholz, C. G. Mullighan, et al. 2012. The TCF-1 and LEF-1 by guest on October 1, 2021 E. E. West, Z. Wei, P. Lu, J. W. Austin, J. L. Riley, et al. 2011. Chronic virus transcription factors have cooperative and opposing roles in T cell development infection enforces demethylation of the locus that encodes PD-1 in antigen- and malignancy. Immunity 37: 813–826. specific CD8(+) T cells. Immunity 35: 400–412. 49. Quigley, M., F. Pereyra, B. Nilsson, F. Porichis, C. Fonseca, Q. Eichbaum, 24. Kersh, E. N., D. R. Fitzpatrick, K. Murali-Krishna, J. Shires, S. H. Speck, B. Julg, J. L. Jesneck, K. Brosnahan, S. Imam, et al. 2010. Transcriptional J. M. Boss, and R. Ahmed. 2006. Rapid demethylation of the IFN-gamma gene analysis of HIV-specific CD8+ T cells shows that PD-1 inhibits T cell function occurs in memory but not naive CD8 T cells. J. Immunol. 176: 4083–4093. by upregulating BATF. Nat. Med. 16: 1147–1151. 25. Oestreich, K. J., H. Yoon, R. Ahmed, and J. M. Boss. 2008. NFATc1 regulates 50. Kuroda, S., M. Yamazaki, M. Abe, K. Sakimura, H. Takayanagi, and Y. Iwai. PD-1 expression upon T cell activation. J. Immunol. 181: 4832–4839. 2011. Basic leucine zipper transcription factor, ATF-like (BATF) regulates epi- 26. Schoenborn, J. R., M. O. Dorschner, M. Sekimata, D. M. Santer, M. Shnyreva, genetically and energetically effector CD8 T-cell differentiation via Sirt1 ex- D. R. Fitzpatrick, J. A. Stamatoyannopoulos, and C. B. Wilson. 2007. Com- pression. Proc. Natl. Acad. Sci. USA 108: 14885–14889. prehensive epigenetic profiling identifies multiple distal regulatory elements 51. Lecomte, V., E. Meugnier, V. Euthine, C. Durand, D. Freyssenet, G. Nemoz, directing transcription of the gene encoding interferon-gamma. [Published er- S. Rome, H. Vidal, and E. Lefai. 2010. A new role for sterol regulatory element ratum appears in 2007 Nat. Immunol. 8: 893 and 2008 Nat. Immunol. 9: 105.] binding protein 1 transcription factors in the regulation of muscle mass and Nat. Immunol. 8: 732–742. muscle cell differentiation. Mol. Cell. Biol. 30: 1182–1198. 27. Stadler, M. B., R. Murr, L. Burger, R. Ivanek, F. Lienert, A. Scho¨ler, E. van 52. Honma, S., T. Kawamoto, Y. Takagi, K. Fujimoto, F. Sato, M. Noshiro, Y. Kato, Nimwegen, C. Wirbelauer, E. J. Oakeley, D. Gaidatzis, et al. 2011. DNA-binding and K. Honma. 2002. Dec1 and Dec2 are regulators of the mammalian molecular factors shape the mouse methylome at distal regulatory regions. [Published er- clock. Nature 419: 841–844. ratum appears in 2011 Nature 484: 550.] Nature 480: 490–495. 53. Shen, M., E. Yoshida, W. Yan, T. Kawamoto, K. Suardita, Y. Koyano, 28. Pircher, H., K. Bu¨rki, R. Lang, H. Hengartner, and R. M. Zinkernagel. 1989. K. Fujimoto, M. Noshiro, and Y. Kato. 2002. Basic helix-loop-helix protein Tolerance induction in double specific T-cell receptor transgenic mice varies DEC1 promotes chondrocyte differentiation at the early and terminal stages. J. with antigen. Nature 342: 559–561. Biol. Chem. 277: 50112–50120. 29. Harris, R. A., T. Wang, C. Coarfa, R. P. Nagarajan, C. Hong, S. L. Downey, 54. Dunham, I., A. Kundaje, S. F. Aldred, P. J. Collins, C. A. Davis, F. Doyle, B. E. Johnson, S. D. Fouse, A. Delaney, Y. Zhao, et al. 2010. Comparison of C. B. Epstein, S. Frietze, J. Harrow, R. Kaul, et al; ENCODE Project Consor- sequencing-based methods to profile DNA methylation and identification of tium. 2012. An integrated encyclopedia of DNA elements in the human genome. monoallelic epigenetic modifications. Nat. Biotechnol. 28: 1097–1105. Nature 489: 57–74. 30. Weber, M., J. J. Davies, D. Wittig, E. J. Oakeley, M. Haase, W. L. Lam, and 55. Barski, A., S. Cuddapah, K. Cui, T. Y. Roh, D. E. Schones, Z. Wang, G. Wei, D. Schu¨beler. 2005. Chromosome-wide and promoter-specific analyses identify I. Chepelev, and K. Zhao. 2007. High-resolution profiling of histone methyl- sites of differential DNA methylation in normal and transformed human cells. ations in the human genome. Cell 129: 823–837. Nat. Genet. 37: 853–862. 56. Ooi, S. K., C. Qiu, E. Bernstein, K. Li, D. Jia, Z. Yang, H. Erdjument-Bromage, 31. Langmead, B., C. Trapnell, M. Pop, and S. L. Salzberg. 2009. Ultrafast and P. Tempst, S. P. Lin, C. D. Allis, et al. 2007. DNMT3L connects unmethylated memory-efficient alignment of short DNA sequences to the human genome. lysine 4 of histone H3 to de novo methylation of DNA. Nature 448: 714–717. Genome Biol. 10: R25. 57. Meissner, A., T. S. Mikkelsen, H. Gu, M. Wernig, J. Hanna, A. Sivachenko, 32. Heinz, S., C. Benner, N. Spann, E. Bertolino, Y. C. Lin, P. Laslo, J. X. Cheng, X. Zhang, B. E. Bernstein, C. Nusbaum, D. B. Jaffe, et al. 2008. Genome-scale C. Murre, H. Singh, and C. K. Glass. 2010. Simple combinations of lineage- DNA methylation maps of pluripotent and differentiated cells. Nature 454: 766– determining transcription factors prime cis-regulatory elements required for 770. macrophage and B cell identities. Mol. Cell 38: 576–589. 58. Vire´, E., C. Brenner, R. Deplus, L. Blanchon, M. Fraga, C. Didelot, L. Morey, 33. Chavez, L., J. Jozefczuk, C. Grimm, J. Dietrich, B. Timmermann, H. Lehrach, A. Van Eynde, D. Bernard, J. M. Vanderwinden, et al. 2006. The Polycomb R. Herwig, and J. Adjaye. 2010. Computational analysis of genome-wide DNA group protein EZH2 directly controls DNA methylation. Nature 439: 871–874. methylation during the differentiation of human embryonic stem cells along the 59. Mikkelsen, T. S., M. Ku, D. B. Jaffe, B. Issac, E. Lieberman, G. Giannoukos, endodermal lineage. Genome Res. 20: 1441–1450. P. Alvarez, W. Brockman, T. K. Kim, R. P. Koche, et al. 2007. Genome-wide The Journal of Immunology 3429

maps of chromatin state in pluripotent and lineage-committed cells. Nature 448: 72. Douglas, N. C., H. Jacobs, A. L. Bothwell, and A. C. Hayday. 2001. Defining the 553–560. specific physiological requirements for c-Myc in T cell development. Nat. 60. Creyghton, M. P., A. W. Cheng, G. G. Welstead, T. Kooistra, B. W. Carey, Immunol. 2: 307–315. E. J. Steine, J. Hanna, M. A. Lodato, G. M. Frampton, P. A. Sharp, et al. 2010. 73. Jain, J., P. G. McCaffrey, V. E. Valge-Archer, and A. Rao. 1992. Nuclear factor of Histone H3K27ac separates active from poised enhancers and predicts devel- activated T cells contains Fos and Jun. Nature 356: 801–804. opmental state. Proc. Natl. Acad. Sci. USA 107: 21931–21936. 74. Epsztejn-Litman, S., N. Feldman, M. Abu-Remaileh, Y. Shufaro, A. Gerson, 61. Wang, H., M. T. Maurano, H. Qu, K. E. Varley, J. Gertz, F. Pauli, K. Lee, J. Ueda, R. Deplus, F. Fuks, Y. Shinkai, H. Cedar, and Y. Bergman. 2008. De T. Canfield, M. Weaver, R. Sandstrom, et al. 2012. Widespread plasticity in novo DNA methylation promoted by G9a prevents reprogramming of embry- CTCF occupancy linked to DNA methylation. Genome Res. 22: 1680–1688. onically silenced genes. Nat. Struct. Mol. Biol. 15: 1176–1183. 62. Kim, H. P., and W. J. Leonard. 2007. CREB/ATF-dependent T cell receptor- 75. Whyte, W. A., S. Bilodeau, D. A. Orlando, H. A. Hoke, G. M. Frampton, induced FoxP3 gene expression: a role for DNA methylation. J. Exp. Med. 204: C. T. Foster, S. M. Cowley, and R. A. Young. 2012. Enhancer decommissioning 1543–1551. by LSD1 during embryonic stem cell differentiation. Nature 482: 221–225. 63. Polansky, J. K., L. Schreiber, C. Thelemann, L. Ludwig, M. Kru¨ger, 76. Velasco, G., F. Hube´, J. Rollin, D. Neuillet, C. Philippe, H. Bouzinba-Segard, R. Baumgrass, S. Cording, S. Floess, A. Hamann, and J. Huehn. 2010. Meth- A. Galvani, E. Viegas-Pe´quignot, and C. Francastel. 2010. Dnmt3b recruitment ylation matters: binding of Ets-1 to the demethylated Foxp3 gene contributes to through E2F6 transcriptional repressor mediates germ-line gene silencing in the stabilization of Foxp3 expression in regulatory T cells. J. Mol. Med. (Berl.) murine somatic tissues. Proc. Natl. Acad. Sci. USA 107: 9281–9286. 88: 1029–1040. 77. Makar, K. W., M. Pe´rez-Melgosa, M. Shnyreva, W. M. Weaver, D. R. Fitzpatrick, 64. Prendergast, G. C., and E. B. Ziff. 1991. Methylation-sensitive sequence-specific and C. B. Wilson. 2003. Active recruitment of DNA methyltransferases regulates DNA binding by the c-Myc basic region. Science 251: 186–189. interleukin 4 in thymocytes and T cells. Nat. Immunol. 4: 1183–1190. 65. Spruijt, C. G., F. Gnerlich, A. H. Smits, T. Pfaffeneder, P. W. Jansen, C. Bauer, 78. Bruniquel, D., and R. H. Schwartz. 2003. Selective, stable demethylation of the M. Munzel, M. Wagner, M. Muller, F. Khan, et al. 2013. Dynamic readers for 5- interleukin-2 gene enhances transcription by an active process. Nat. Immunol. 4: (hydroxy)methylcytosine and its oxidized derivatives. Cell 152: 1146–1159. 235–240. 66. Thanos, D., and T. Maniatis. 1995. Virus induction of human IFN beta 79. Ito, S., A. C. D’Alessio, O. V. Taranova, K. Hong, L. C. Sowers, and Y. Zhang. gene expression requires the assembly of an enhanceosome. Cell 83: 1091– 2010. Role of Tet proteins in 5mC to 5hmC conversion, ES-cell self-renewal and 1100. inner cell mass specification. Nature 466: 1129–1133. 67. Chen, X., H. Xu, P. Yuan, F. Fang, M. Huss, V. B. Vega, E. Wong, Y. L. Orlov, 80. Tahiliani, M., K. P. Koh, Y. Shen, W. A. Pastor, H. Bandukwala, Y. Brudno, W. Zhang, J. Jiang, et al. 2008. Integration of external signaling pathways with S. Agarwal, L. M. Iyer, D. R. Liu, L. Aravind, and A. Rao. 2009. Conversion of Downloaded from the core transcriptional network in embryonic stem cells. Cell 133: 1106–1117. 5-methylcytosine to 5-hydroxymethylcytosine in mammalian DNA by MLL 68. Calnan, B. J., S. Szychowski, F. K. Chan, D. Cado, and A. Winoto. 1995. A role partner TET1. Science 324: 930–935. for the orphan steroid receptor Nur77 in apoptosis accompanying antigen- 81. Cortellino, S., J. Xu, M. Sannai, R. Moore, E. Caretti, A. Cigliano, M. Le Coz, induced negative selection. Immunity 3: 273–282. K. Devarajan, A. Wessels, D. Soprano, et al. 2011. Thymine DNA glycosylase is 69. Corre, S., and M. D. Galibert. 2005. Upstream stimulating factors: highly ver- essential for active DNA demethylation by linked deamination-base excision satile stress-responsive transcription factors. Pigment Cell Res. 18: 337–348. repair. Cell 146: 67–79. 70. Thompson, M. R., D. Xu, and B. R. Williams. 2009. ATF3 transcription factor 82. Costa, Y., J. Ding, T. W. Theunissen, F. Faiola, T. A. Hore, P. V. Shliaha,

and its emerging roles in immunity and cancer. J. Mol. Med. (Berl.) 87: 1053– M. Fidalgo, A. Saunders, M. Lawrence, S. Dietmann, et al. 2013. NANOG- http://www.jimmunol.org/ 1060. dependent function of TET1 and TET2 in establishment of pluripotency. Na- 71. Dang, C. V. 2012. MYC on the path to cancer. Cell 149: 22–35. ture 495: 370–374. by guest on October 1, 2021 Scharer et al. Effector CD8 T cell epigenetics

Supplemental Figure1: CD8 naïve and D8 effector Sorting and Phenotype

Characterization. (A) Three independent biological replicates of naïve P14 CD8 T cells

(CD8+CD44low) were purified by FACS according to surface expression of CD8 and

CD44. (B) Each isolation of naïve P14 CD8 T cells was characterized for surface expression of known phenotypic markers by flow cytometry. The majority of transgenic

CD8 T cells were specific for the GP33 epitope of LCMV and demonstrated positive surface expression of Thy1.1, CD62L, and CD127 and were negative for Klrg1 and PD-1 receptors. (C) Three independent biological isolations of D8 effector CD8 T cells

(CD8+Thy1.1+) were isolated to high purity by FACS according to surface expression of

CD8 and the congenic marker Thy1.1. (D) Surface expression of phenotypic markers for

D8 effector CD8 T cells (blue) were characterized by flow cytometry. Endogenous naïve

CD8 T cells (red) are shown as a comparison to highlight the change in expression of each marker. D8 effectors displayed high expression of Klrg1, moderately increased expression of PD-1 and CD25, and reduced expression of CD62L and CD127 compared to naïve CD8 T cells.

Supplemental Figure 2: Transcription factor co-occurrence present in methylated

DMRs. (A) Transcription factor motif co-occurrence at methylated DMRs was hierarchically clustered to identify factors that could bind a sequence within the same

DMR. Co-occurrence between any two transcription factor motifs enriched in the same methylated DMR was calculated by Spearman's rank correlation. Five potential modules

(A-E) are highlighted in blue. (B) Table of transcription factor motifs present in methylated DMR modules. Some of the transcription factors within the groupings could bind to the identical or similar sequences (e.g., the STATs in module C).

Scharer et al. Effector CD8 T cell epigenetics

Supplemental Table 1: PCR primers for clonal bisulfite amplification and ENCODE accession numbers for thymus histone data.

Supplemental Table 2: Complete list of transcription factor motifs enriched in the

DMRs.

Supplemental Table 1

Bisulfite Sequencing Primers

Primer Sequence (5'-3') BS.GzmB.Fwd TTGTTAGTAAAGGTTAGATGTGGGT BS.GzmB.Rev TCTACCCTCCTACTACTAATAACAAC BS.GzmK.Fwd GGGGTATGGGATATTTATATTTTGAAG BS.GzmK.Rev CTATCCTTCTCCATCACAAATCCATCC BS.Itgae.Fwd AATATGGTTGGGGATTGATTTTATGAG BS.Itgae.Rev CCACTTCATAACTAAAACAAATACAAC BS.Ccr7.Fwd GAGGTGGTTTAGTAGTTAATAGAGA BS.Ccr7.Rev AAACATATACCACCATACCTAACCT

ENCODE Data

Tissue Antibody DCC_Accession Resource Provider Lab Institution Thymus H3K4me1 wgEncodeEM002475 Ren, B. LICR-m Thymus H3K4me3 wgEncodeEM002476 Ren, B. LICR-m Thymus H3K27ac wgEncodeEM002474 Ren, B. LICR-m Thymus H3K27me3 wgEncodeEM002724 Ren, B. LICR-m Thymus Input wgEncodeEM002477 Ren, B. LICR-m Supplemental Table 2

Motifs Enriched in Demethylated DMRs Motif Name Consensus P-value FDR Jun-AP1(bZIP)/K562-cJun GATGASTCATCN 1E-251 0 AP-1(bZIP)/ThioMac-PU.1 VTGACTCATC 1E-249 0 c-Myc(HLH)/LNCAP-cMyc VCCACGTG 1E-68 0 NF-E2(bZIP)/K562-NFE2 GATGACTCAGCA 1E-28 0 bHLHE40(HLH)/HepG2-BHLHE40 SGKCACGTGM 1E-23 0 USF1(HLH)/GM12878-Usf1 SGTCACGTGR 1E-15 0 EGR(Zf)/K562-EGR1 TCCGCCCACGCA 1E-13 0 JunD(bZIP)/K562-JunD ATGACGTCATCN 1E-12 0 NFAT:AP1/Jurkat-NFATC1 SARTGGAAAAWRTGAGTCAB 1E-12 0 NFAT(RHD)/Jurkat-NFATC1 ATTTTCCATT 1E-10 0 c-Jun-CRE(bZIP)/K562-cJun ATGACGTCATCY 1E-10 0 Max(HLH)/K562-Max RCCACGTGGYYN 1E-08 0 Pdx1(Homeobox)/Islet-Pdx1 YCATYAATCA 1E-08 0 NFkB-p65-Rel(RHD)/LPS-exp GGAAATTCCC 1E-07 0 Hoxb4/ES-Hoxb4(GSE34014) TGATTRATGGCY 1E-07 0 AARE(HLH)/mES-cMyc GATTGCATCA 1E-07 0 Nanog(Homeobox)/mES-Nanog RGCCATTAAC 1E-06 0 IRF4(IRF)/GM12878-IRF4 ACTGAAACCA 1E-06 0 NRF1/Promoter STGCGCATGCGC 1E-06 0 Egr2/Thymocytes-Egr2(GSE34254) NGCGTGGGCGGR 1E-06 0 Sp1(Zf)/Promoter GGCCCCGCCCCC 1E-05 0 c-Myc/mES-cMyc VVCCACGTGG 1E-05 0 ATF3(bZIP)/K562-ATF3 SGGTCACGTGAC 1E-05 0.0001 HIF-1a(HLH)/MCF7-HIF1a TACGTGCV 1E-04 0.0002 T1ISRE(IRF)/Ifnb-Exp ACTTTCGTTTCT 1E-03 0.0017 Nur77(NR)/K562-NR4A1(GSE31363) TGACCTTTNCNT 1E-03 0.003 MYB(HTH)/ERMYB-Myb(GSE22095) GGCVGTTR 1E-03 0.0035 CEBP:AP1/ThioMac-CEBPb DRTGTTGCAA 1E-02 0.01 HOXA2(Homeobox)/mES-Hoxa2 GYCATCMATCAT 1E-02 0.0104 CRE(bZIP)/Promoter CSGTGACGTCAC 1E-02 0.0184 X-box(HTH)/NPC-H3K4me1 GGTTGCCATGGCAA 1E-02 0.0202 GATA-IR3(Zf)/iTreg-Gata3(GSE20898) NNNNNBAGATAWYATCTVHN 1E-02 0.0244 ISRE(IRF)/ThioMac-LPS-exp AGTTTCASTTTC 1E-02 0.025 GRE/RAW264.7-GRE VAGRACAKWCTGTYC 1E-02 0.0463 RXR(NR/DR1)/3T3L1-RXR TAGGGCAAAGGTCA 1E-02 0.0465

Motifs Enriched in Methylated DMRs Motif Name Consensus P-value FDR ZFX(Zf)/mES-Zfx AGGCCTRG 1E-283 0 ZNF711(Zf)/SH-SY5Y-ZNF711 AGGCCTAG 1E-209 0 EBF1(EBF)/Near-E2A GTCCCCWGGGGA 1E-137 0 SPDEF(ETS)/VCaP-SPDEF ASWTCCTGBT 1E-135 0 Nkx2.5(Homeobx)/HL1-Nkx2.5.biotin RRSCACTYAA 1E-118 0 Bcl6(Zf)/Liver-Bcl6(GSE31578) NNNCTTTCCAGGAAA 1E-114 0 ETV1(ETS)/GIST48-ETV1 AACCGGAAGT 1E-106 0 AP-2alpha(AP2)/Hela-AP2alpha ATGCCCTGAGGC 1E-97 0 TATA-Box(TBP)/Promoter CCTTTTAWAGSC 1E-94 0 Erra(NR)/HepG2-Erra CAAAGGTCAG 1E-90 0 EWS:FLI1-fusion(ETS)/SK_N_MC-EWS:FLI1 VACAGGAAAT 1E-87 0 GABPA(ETS)/Jurkat-GABPa RACCGGAAGT 1E-83 0 ZNF143|STAF(Zf)/CUTLL-ZNF143(GSE29600) ATTTCCCAGVAKSCY 1E-82 0 ELF1(ETS)/Jurkat-ELF1 AVCCGGAAGT 1E-78 0 AP2gamma(AP2)/MCF7-TFAP2c HHTGSCCTSAGGSCA 1E-78 0 E2A-nearPU.1(HLH)/Bcell-PU.1 NVCACCTGBN 1E-77 0 ERG(ETS)/VCaP-ERG ACAGGAAGTG 1E-71 0 FOXA1:AR/LNCAP-AR AGTAAACAAAAAAGAACAND 1E-70 0 PRDM14(Zf)/H1-PRDM14 RGGTCTCTAACY 1E-64 0 HRE(HSF)/HepG2-HSF1 BSTTCTRGAABVTTCYAGAA 1E-62 0 FOXA1(Forkhead)/MCF7-FOXA1 WAAGTAAACA 1E-62 0 NF1(CTF)/LNCAP-NF1 CYTGGCABNSTGCCAR 1E-61 0 E2A(HLH)/proBcell-E2A DNRCAGCTGY 1E-61 0 Tcf4(HMG)/Hct116-Tcf4 ASATCAAAGGVA 1E-60 0 EWS:ERG-fusion(ETS)/CADO_ES1-EWS:ERG ATTTCCTGTN 1E-54 0 Supplemental Table 2

Nr5a2(NR)/mES-Nr5a2 BTCAAGGTCA 1E-54 0 Tcf3(HMG)/mES-Tcf3 ASWTCAAAGG 1E-53 0 Unknown/Homeobox/Limb-p300 SSCMATWAAA 1E-52 0 BORIS(Zf)/K562-CTCFL CNNBRGCGCCCCCTGSTGGC 1E-52 0 FOXA1(Forkhead)/LNCAP-FOXA1 WAAGTAAACA 1E-51 0 HNF4a(NR/DR1)/HepG2-HNF4a CARRGKBCAAAGTYCA 1E-48 0 ETS(ETS)/Promoter AACCGGAAGT 1E-47 0 Smad3(MAD)/NPC-Smad3(GSE36673) TWGTCTGV 1E-46 0 ETS1(ETS)/Jurkat-ETS1 ACAGGAAGTG 1E-45 0 Ets1-distal(ETS)/CD4+-PolII MACAGGAAGT 1E-44 0 AR-halfsite(NR)/LNCaP-AR CCAGGAACAG 1E-41 0 Nr5a2(NR)/Pancreas-LRH1(GSE34295) BTCAAGGTCA 1E-37 0 CRX(Homeobox)/Retina-Crx GCTAATCC 1E-36 0 Tcf12(HLH)/GM12878-Tcf12 VCAGCTGYTG 1E-33 0 SEP3(MADS)/Arabidoposis-Flower-Sep3 CCAAAAAGGG 1E-33 0 RARg(NR)/ES-RARg(GSE30538) AGGTCAAGGTCA 1E-30 0 EBF(EBF)/proBcell-EBF DGTCCCYRGGGA 1E-29 0 CTCF(Zf)/CD4+-CTCF AYAGTGCCMYCTRGTGGCCA 1E-29 0 CHR/Cell-Cycle-Exp SRGTTTCAAA 1E-27 0 Stat3+il23(Stat)/CD4-Stat3 SVYTTCCNGGAARB 1E-26 0 Gfi1b(Zf)/HPC7-Gfi1b MAATCACTGC 1E-24 0 PU.1(ETS)/ThioMac-PU.1 AGAGGAAGTG 1E-24 0 HEB?/mES-Nanog CACAGCAGGGGG 1E-24 0 Atoh1(bHLH)/Cerebellum-Atoh1 VNRVCAGCTGGY 1E-21 0 Stat3(Stat)/mES-Stat3 CTTCCGGGAA 1E-20 0 Sox3(HMG)/NPC-Sox3(GSE33059) CCWTTGTY 1E-19 0 STAT6(Stat)/CD4-Stat6 ABTTCYYRRGAA 1E-18 0 NeuroD1(bHLH)/Islet-NeuroD1(GSE30298) GCCATCTGTT 1E-17 0 GFY-Staf/Promoters RACTACAATTCCCAGAAKGC 1E-17 0 Mef2a(MADS)/HL1-Mef2a.biotin/ CYAAAAATAG 1E-16 0 NFkB-p65(RHD)/GM12787-p65 WGGGGATTTCCC 1E-16 0 STAT1(Stat)/HelaS3-STAT1 NATTTCCNGGAAAT 1E-16 0 TCFL2(HMG)/K562-TCF7L2(GSE29196) ACWTCAAAGG 1E-16 0 Olig2(bHLH)/Neuron-Olig2(GSE30882) RCCATMTGTT 1E-15 0 MyoD(HLH)/Myotube-MyoD RRCAGCTGYTSY 1E-15 0 Sox2(HMG)/mES-Sox2 BCCATTGTTC 1E-14 0 ETS:RUNX/Jurkat-RUNX1 RCAGGATGTGGT 1E-13 0 FOXP1(Forkhead)/H9-FOXP1(GSE31006) NYYTGTTTACHN 1E-13 0 STAT4(Stat)/CD4-Stat4 NYTTCCWGGAAR 1E-13 0 Foxo1(Forkhead)/RAW-Foxo1 CTGTTTAC 1E-13 0 Tlx?/NPC-H3K4me1 CTGGCAGSCTGCCA 1E-12 0 STAT6/Macrophage-Stat6 TTCCKNAGAA 1E-12 0 Hoxc9/Ainv15-Hoxc9 GGCCATAAATCA 1E-11 0 NF1-halfsite(CTF)/LNCaP-NF1 YTGCCAAG 1E-10 0 STAT5(Stat)/mCD4+-Stat5a|b RTTTCTNAGAAA 1E-09 0 ETS:E-box/HPC7-Scl AGGAARCAGCTG 1E-09 0 TEAD4(TEA)/Tropoblast-Tead4(GSE37350) CCWGGAATGY 1E-08 0 VDR(NR/DR3)/GM10855-VDR+vitD ARAGGTCANWGAGTTCANNN 1E-08 0 GATA3(Zf)/iTreg-Gata3(GSE20898) AGATAASR 1E-08 0 NRF1(NRF)/MCF7-NRF1 CTGCGCATGCGC 1E-08 0 TR4(NR/DR1)/Hela-TR4 GAGGTCAAAGGTCA 1E-08 0 GATA-DR8(Zf)/iTreg-Gata3(GSE20898) AGATSTNDNNDSAGATAASN 1E-07 0 HOXA9/HSC-Hoxa9(GSE33509) GGCCATAAATCA 1E-07 0 Esrrb(NR)/mES-Esrrb KTGACCTTGA 1E-07 0 TEAD(TEA)/Fibroblast-PU.1 YCWGGAATGY 1E-07 0 Cdx2(Homeobox)/mES-Cdx2 GYMATAAAAH 1E-07 0 RUNX1(Runt)/Jurkat-RUNX1 AAACCACARM 1E-07 0 Oct4(POU/Homeobox)/mES-Oct4 ATTTGCATAW 1E-07 0 EKLF(Zf)/Erythrocyte-Klf1(GSE20478) NWGGGTGTGGCY 1E-06 0 NF1:FOXA1/LNCAP-FOXA1 WNTGTTTRYTTTGGCA 1E-06 0 Lhx3(Homeobox)/Forebrain-p300 CTAATTAGCH 1E-05 0 Supplemental Table 2

Sox6(HMG)/Myotubes-Sox6(GSE32627) CCATTGTTNY 1E-05 0 FXR(NR/IR1)/Liver-FXR AGGTCANTGACCTB 1E-05 0 ARE(NR)/LNCAP-AR RGRACASNSTGTYCYB 1E-05 0 OCT4-SOX2-TCF-NANOG((POU/Homeobox/HMG)/mES ATTTGCATAACAATG 1E-05 0 Gata2(Zf)/K562-GATA2 BBCTTATCTS 1E-05 0 Foxa2(Forkhead)/Liver-Foxa2 CYTGTTTACWYW 1E-04 0 Tbx5(T-box)/HL1-Tbx5.biotin AGGTGTCA 1E-04 0 E2F(E2F)/Cell-Cycle-Exp TTSGCGCGAAAA 1E-04 0 Klf4(Zf)/mES-Klf4 GCCACACCCA 1E-04 0 Tcfcp2l1(CP2)/mES-Tcfcp2l1 NRAACCRGTTYRAACCRGYT 1E-04 0.0001 ZBTB33/GM12878-ZBTB33 GGVTCTCGCGAGAAC 1E-04 0.0001 E2F/Hela-E2F1/Hoemr CWGGCGGGAA 1E-03 0.0003 GLI3(Zf)/GLI3-ChIP-Chip CGTGGGTGGTCC 1E-03 0.0005 GATA-DR4(Zf)/iTreg-Gata3(GSE20898) AGATGKDGAGATAAG 1E-03 0.0013 Gata1(Zf)/K562-GATA1 SAGATAAGRV 1E-02 0.0019 BMYB(HTH)/Hela-BMYB(GSE27030) NHAACBGYYV 1E-02 0.002 Oct2(POU/Homeobox)/Bcell-Oct2 ATATGCAAAT 1E-02 0.0021 REST-NRSF(Zf)/Jurkat-NRSF GGMGCTGTCCATGGTGCTGA 1E-02 0.0028 HNF6(Homeobox)/Liver-Hnf6(ERP000394) NTATYGATCH 1E-02 0.003 GATA:SCL/Ter119-SCL CRGCTGBNGNSNNSAGATAA 1E-02 0.0078 Hnf1(Homeobox)/Liver-Foxa2 GGTTAAWCATTAA 1E-02 0.0084 Tbx20(T-box)/Heart-Tbx20(GSE29636) GGTGYTGACAGS 1E-02 0.0086 SCL/HPC7-Scl AVCAGCTG 1E-02 0.0107 RUNX2(Runt)/PCa-RUNX2(GSE33889) NWAACCACADNN 1E-02 0.012 PAX3:FKHR-fusion(Paired/Homeobox)/Rh4-PAX3:FKHR ACCRTGACTAATTNN 1E-02 0.0134 PU.1-IRF/Bcell-PU.1 MGGAAGTGAAAC 1E-01 0.018 RFX(HTH)/K562-RFX3 CGGTTGCCATGGCAAC 1E-01 0.0272 RUNX(Runt)/HPC7-Runx1 SAAACCACAG 1E-01 0.031 RUNX-AML(Runt)/CD4+-PolII GCTGTGGTTW 1E-01 0.031 GFX(?)/Promoter ATTCTCGCGAGA 1E-01 0.0311 Pax7-long(Paired/Homeobox)/Myoblast-Pax7(GSE25064) TAATCHGATTAC 1E-01 0.0339 Znf263(Zf)/K562-Znf263 CVGTSCTCCC 1E-01 0.0349 GFY(?)/Promoter ACTACAATTCCC 1E-01 0.0382 CArG(MADS)/PUER-Srf CCATATATGGNM 1E-01 0.0382 Reverb(NR/DR2)/BLRP(RAW)-Reverba GTRGGTCASTGGGTCA 1E-01 0.0418