Epigenetic Networks Regulate the Transcriptional Program in Memory and Terminally Differentiated CD8+ T Cells

This information is current as Ramon M. Rodriguez, Beatriz Suarez-Alvarez, José L. of September 26, 2021. Lavín, David Mosén-Ansorena, Aroa Baragaño Raneros, Leonardo Márquez-Kisinousky, Ana M. Aransay and Carlos Lopez-Larrea J Immunol published online 14 December 2016

http://www.jimmunol.org/content/early/2016/12/14/jimmun Downloaded from ol.1601102

Supplementary http://www.jimmunol.org/content/suppl/2016/12/14/jimmunol.160110 Material 2.DCSupplemental 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 September 26, 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 © 2016 by The American Association of Immunologists, Inc. All rights reserved. Print ISSN: 0022-1767 Online ISSN: 1550-6606. Published December 14, 2016, doi:10.4049/jimmunol.1601102 The Journal of Immunology

Epigenetic Networks Regulate the Transcriptional Program in Memory and Terminally Differentiated CD8+ T Cells

Ramon M. Rodriguez,* Beatriz Suarez-Alvarez,* Jose´ L. Lavı´n,† David Mose´n-Ansorena,‡ Aroa Baragan˜o Raneros,* Leonardo Ma´rquez-Kisinousky,* Ana M. Aransay,† and Carlos Lopez-Larrea*,x

Epigenetic mechanisms play a critical role during differentiation of T cells by contributing to the formation of stable and heritable transcriptional patterns. To better understand the mechanisms of memory maintenance in CD8+ T cells, we performed genome- wide analysis of DNA methylation, histone marking (acetylated lysine 9 in histone H3 and trimethylated lysine 9 in histone), and -expression profiles in naive, effector memory (EM), and terminally differentiated EM (TEMRA) cells. Our results indicate that DNA demethylation and histone acetylation are coordinated to generate the transcriptional program associated with memory cells. Conversely, EM and TEMRA cells share a very similar epigenetic landscape. Nonetheless, the TEMRA transcriptional Downloaded from program predicts an innate immunity phenotype associated with never reported in these cells, including several mediators of NK cell activation (VAV3 and LYN) and a large array of NK receptors (e.g., KIR2DL3, KIR2DL4, KIR2DL1, KIR3DL1, KIR2DS5). In addition, we identified up to 161 genes that encode transcriptional regulators, some of unknown function in CD8+ T cells, and that were differentially expressed in the course of differentiation. Overall, these results provide new insights into the regulatory networks involved in memory CD8+ T cell maintenance and T cell terminal differentiation. The Journal of Immunology, 2017, 198: 000–000. http://www.jimmunol.org/

ytotoxic CD8+ T lymphocytes are essential determinants population that remains quiescent until it encounters a suitable Ag. of the adaptive immunity required for the clearance of Upon Ag activation, naive cells undergo rapid clonal expansion intracellular pathogens and cancer cells. Each individual while acquiring cytolytic functions and cytokine production ca- C + possesses a large repertoire of naive CD8 T cells, a long-lived pacity. Most of these effector cells die soon after pathogen clearance, except for a small fraction of long-term memory cells that are capable of regaining effector functions upon Ag reen- *Department of Immunology, Central University Hospital of Asturias, 33011 Oviedo, Spain; †Genome Analysis Platform, CIC bioGUNE and CIBERehd, Technological counter. Therefore, critical questions arise as to how naive cells by guest on September 26, 2021 Park of Bizkaia, 48160 Derio, Spain; ‡Biostatistics and Computational Biology, maintain homeostasis and cellular phenotype in the quiescent Dana-Farber Cancer Institute and Harvard School of Public Health, Boston, MA state, how they modify their transcriptional program in response to 02215; and xFundacio´n Renal I´n˜igo A´ lvarez de Toledo, 28003 Madrid, Spain Ags, and how memory cells remain quiescent while retaining ORCIDs: 0000-0002-3431-4206 (B.S.-A.); 0000-0003-0914-3211 (J.L.L.); 0000- 0002-5019-5753 (L.M.-K.); 0000-0002-8271-612X (A.M.A.). effector functions in response to stimuli. The thorough molecular characterization of the distinct lymphocyte Received for publication June 24, 2016. Accepted for publication November 13, + 2016. subsets identified various key mediators of CD8 T cell effector This work was supported by the Plan Nacional de I+D+I 2008–2011 and the Euro- and memory functions, including T-box 21 , eomeso- pean Union Fondos Feder, Instituto de Salud Carlos III (Grants PI12/02587 and PI16/ dermin (EOMES), B lymphocyte–induced maturation protein 1 01318), Red Espan˜ola de Investigacio´n Renal (Grants RD12/0021/0018, RD12/0021/ 0021, and RD16/0009/0020), and Plan de Ciencia, Tecnologı´a e Innovacio´n 2013– (BLIMP1), TCF1, and ID (1–5). Nonetheless, this list is 2017 del Principado de Asturias (GRUPIN-14-030). CIC bioGUNE support was unlikely to be complete and does not fully explain how the provided by The Department of Industry, Tourism and Trade of the Government of transcriptional program is dynamically altered during differen- the Autonomous Community of the Basque Country (Etortek Research Programs 2007–2015), the Innovation Technology Department of Bizkaia County, and the tiation or how it is stabilized in quiescent cells. Epigenetic CIBER (Biomedical Research Networking Centre) program at Instituto de Salud mechanisms, such as DNA methylation and histone marks, are Carlos III. essential for generating stable and heritable gene-expression The sequences presented in this article have been submitted to the National Center patterns, which they achieve by altering chromatin transcrip- for Biotechnology Information Omnibus under accession number GSE83159 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE83159). tional potential. Consequently, epigenetic analyses are of in- Address correspondence and reprint requests to Dr. Carlos Lopez-Larrea, Department creasing importance for the valuable insights into the of Immunology, Hospital Universitario Central de Asturias, 33011 Oviedo, Spain. differentiation processes that they provide. Indeed, genome-wide E-mail address: [email protected] analyses demonstrated that some histone marks (dimethylation The online version of this article contains supplemental material. of lysine 4 and trimethylation of lysines 4 and 27 in histone H3) Abbreviations used in this article: BLIMP1, B lymphocyte–induced maturation pro- are highly dynamic in naive and memory CD8+ T cell subsets tein 1; CEBPD, CCAAT/enhancer binding protein d; ChIP, chromatin immunopre- cipitation; DEG, differentially expressed gene; DMR, differentially methylated (6–8). Moreover, genomic analysis of DNA methylation after region; EM, effector memory; EOMES, ; FC, fold change; FDR, false activation of mouse naive CD8+ T cells clearly demonstrated the discovery rate; GO, ; H3K9Ac, acetylated lysine 9 in histone H3; H3K9me3, trimethylated lysine 9 in histone H3; KEGG, Kyoto Encyclopedia of extensive DNA methylation remodeling associated with acqui- Genes and Genomes; PRF1, perforin 1; TEMRA, terminally differentiated EM; sition of effector function (9). These analyses yielded a large TSS, transcription start site; ZSCAN1, zinc finger and SCAN domain containing catalog of novel regulatory regions located in promoters, gene protein 18. body regions, and distant regulatory regions, such as enhancers Copyright Ó 2016 by The American Association of Immunologists, Inc. 0022-1767/16/$30.00 and superenhancers. In any case, a long-standing question is how

www.jimmunol.org/cgi/doi/10.4049/jimmunol.1601102 2 EPIGENETIC NETWORKS DURING CD8+ T CELL DIFFERENTIATION different epigenetic mechanisms, including histone marks and DNA immunoprecipitation (ChIP)-sequencing data sets were generated from methylation, contribute in a coordinated fashion to differentiation two technical replicates obtained from a pool of three samples. After and to the maintenance of the homeostasis of naive and effector immunoprecipitation, ChIP libraries were prepared with a TruSeq ChIP Sample Preparation Kit (catalog number IP-202-1012), following the memory (EM) cells. To better understand how epigenetic mecha- TruSeq ChIP Sample Preparation Guide (catalog number 15023092 Rev. nisms coordinate to regulate transcriptional networks during CD8+ A; both from Illumina). In brief, 15 ng of immunoprecipitated DNA was T cell ontogeny, we compared genomic histone marking (acetylated end-repaired. Then, A-tailing and TruSeq-indexed adaptor ligation were lysine 9 in histone H3 [H3K9Ac] and trimethylated lysine 9 in performed. Inserts between 250 and 500 bp were selected from agarose gels and, finally, libraries were enriched by PCR (30 s at 98˚C; 18 cycles of histone H3 [H3K9me3]), DNA methylation, and gene expression in 10 s at 98˚C, 30 s at 60˚C, 30 s at 72˚C; 5 min at 72˚C; and pause at 10˚C). human naive cells with resting EM cells and terminally differenti- Libraries were visualized on an Agilent 2100 Bioanalyzer using an Agilent ated EM (TEMRA) cells. Our findings demonstrate that these epi- High Sensitivity DNA Kit (catalog number G2938-90320; Agilent Tech- genetic mechanisms are temporally and spatially coordinated to nologies) and quantified using quantitative PCR with a Kapa Library regulate gene-transcription networks in resting CD8+ T cells and Quantification Kit (Master Mix and DNA Standards; catalog number KK4824; Kapa Biosystems) and a Qubit dsDNA HS Assay Kit (catalog reveal a fine-scale map of novel regulatory regions and transcrip- number 32851; Life Technologies). Libraries were pooled at a final con- tional modulators. centration of 9 pM, and sequencing of the pools was carried out using a HiScanSQ platform (Illumina). Raw reads data were converted by Illu- mina’s CASAVA1.8.2 software into FASTQ format. Quality control of Materials and Methods FASTQs was performed using FASTQC software (http://www. Isolation of human CD8+ T cells bioinformatics.babraham.ac.uk). One hundred– paired-end reads were mapped to the Homo sapiens (hg19/GRCh37) reference genome.

PBMCs were isolated from healthy donors using a Lymphoprep density Mapping was performed by Bowtie 2 (10), with a mismatch of n = 2. This Downloaded from gradient. CD8+ T cell subsets were isolated from PBMCs by FACS. Naive + + 2 + + process yielded a SAM alignment file for each sample. Using this method, cells were identified as CD3 CD8 CD4 CCR7 CD45RA , EM cells were we obtained an average of 16 million mapped reads per sample for identified as CD3+CD8+CD42CCR72CD45RA2, and TEMRA cells were H3K9Ac (TNaive = 15,908,471; TEM = 12,567,009; TTEMRA = 21,744,558) identified as CD3+CD8+CD42CCR72CD45RA+. Isolated cell subsets and 21 million mapped reads for H3K9me3 (TNaive = 26,390,129; TEM = were .95% pure for the indicated phenotype upon FACS reanalysis. All 18,874,129, TTEMRA = 17,505,035). Results of the read mapping were Abs were obtained from BioLegend. Cells were sorted with a BD FAC- exported as input for the peak calling step, which was performed by two SAria II Cell Sorter (BD Biosciences). Informed consent was obtained peak calling tools, MACS (11) and SPP (12), a peak calling algorithm from participants in accordance with the Declaration of Helsinki. The implemented as an R package. For visualization, ChIP signals were dis- http://www.jimmunol.org/ study was performed in accordance with the approved guidelines estab- tributed in 100-bp windows and normalized to reads per million. ChIP lished by the Research Ethics Board of the Spanish Research Council. enrichment in specific genomic regions was plotted using SeqMonk tools DNA extraction and whole-genome methylation profiling (Babraham Bioinformatics). Total DNA was extracted with an ATP Genomic DNA Mini Kit (ATP Western blotting Biotech), following the manufacturer’s protocol. DNA was quantified with Naive CD8+ T cells were isolated from PBMCs by magnetic bead sepa- 3 + a Qubit 2.0 Fluorometer (Life Technologies). Integrity was assessed in 1 ration using the Naive CD8 T Cell Isolation Kit (Miltenyi Biotec), fol- TAE agarose gels. Sodium bisulfite conversion of 500 ng of good-quality lowing the manufacturer’s instructions. Cells were cultured in AIM V total DNA was performed with the EZ DNA methylation kit (Zymo Re- medium (Life Technologies) and activated with 1 mg/ml of anti-CD28 and search). Whole-genome methylation was analyzed with Infinium 3 mg/ml of anti-CD3 (both from BioLegend) for 72 h, in the presence or by guest on September 26, 2021 HumanMethylation450 BeadChip Kit (Illumina), according to Illumina’s absence of 5 mM ICG001 inhibitor (Selleck Chemicals). Cells were lysed Infinium HD assay methylation protocol. This array provides genomic with 2% SDS buffer in the presence of protease inhibitor (Complete mini coverage of a total of 26,658 CpG islands (96%), 21,231 UCSC refGenes EDTA-free; Roche Applied Science) and phosphatase inhibitor (sodium (including the 59 and 39 untranslated regions), 80,538 informatically pre- orthovanadate; Sigma-Aldrich) for 10 min on ice. After SDS-PAGE, pro- dicted enhancers, and 59,916 DNase-hypersensitive sites. Methylation teins were detected with Abs against STAT1 (1:1,000; Cell Signaling), datasets were generated from two biological replicates, both obtained from phospho-STAT1 (1:1,000, Tyr701; Cell Signaling), and b-actin (1:10,000; a pool of three samples using identical amounts of DNA per donor and cell Santa Cruz) and the secondary Ab conjugated with HRP (1:2000; Sigma- type. Finally, raw data were decoded with GenomeStudio software (Illu- Aldrich). Western blot images were obtained with a ChemiDoc MP System mina). and quantified with Quantity One analysis software (both from Bio-Rad). RNA extraction and whole-genome gene expression Flow cytometry Total RNA was isolated with an RNAqueous-Micro Kit (Ambion). RNA For cell surface staining, cells were washed in Dulbecco’s PBS (DPBS) and integrity was evaluated with a 2100 Bioanalyzer (Agilent Technologies) and incubated with the primary Ab for 30 min at 4˚C. All Abs against surface quantified with a Qubit 2.0 Fluorometer (Life Technologies). Whole- markers were purchased from BioLegend. For intranuclear staining, naive genome expression was characterized using HumanHT-12 v4 Expression CD8+ T cells were fixed and permeabilized with the NF Fixation and BeadChip Kit (Illumina). cRNA synthesis was performed with a Targe- Permeabilization Buffer Set (BioLegend), following the manufacturer’s tAmp Nano-g Biotin-aRNA Labelling Kit for the Illumina System (Epi- instructions. After incubation with anti-LEF1 Ab (Abcam), cells were centre; catalog number TAN07924). Amplification, labeling, and incubated with a secondary Ab conjugated to PE (BioLegend) for 30 min hybridization were performed according to Illumina’s Whole-Genome at room temperature. Cell proliferation was analyzed by CFSE staining Gene Expression Direct Hybridization protocol. GenomeStudio analysis (BioLegend), following the manufacturer’s instructions. After staining, software (Illumina) was used to convert raw data into a GenomeStudio’s cells were washed and analyzed in a Gallios Flow Cytometer (Beckman Final Report (sample probe profile). Coulter). Chromatin immunoprecipitation and sequencing Small interfering RNA transfection Chromatin of isolated CD8+ T cells was immunoprecipitated with an iDeal ChIP-seq Kit (Diagenode), following the manufacturer’s instructions. Each Naive CD8+ T cells were isolated from PBMCs by magnetic bead sepa- sample was obtained from a pool of three individuals. Abs against ration using a Naive CD8+ T Cell Isolation Kit (Miltenyi Biotec). Purity of H3K9me3 and H3K9Ac were obtained from Active Motif, and the Ab the naive CD8+ T cells was assessed by flow cytometry using Abs against against LEF1 was from Abcam. Normal IgG was used as a negative control human CD8, CD45RA, and CCR7 (BioLegend). A pool of four small (Abcam). Chromatin fragments (250–500-bp) were produced in a water interfering RNA sequences against LEF1 was obtained from Dharmacon bath sonicator (Bioruptor; Diagenode). The fragment size of sheared and transfected with HiPerFect Transfection Reagent (QIAGEN), follow- chromatin was assessed with a 2100 Bioanalyzer System (Agilent). ing the manufacturer’s instructions. Knockdown efficiency was assessed Immunoprecipitated DNA was analyzed in triplicate by real-time PCR. by quantitative real-time PCR using universal SYBR Green PCR Master The primers used, 59-ACTGACCTCCGTGTTGCTAA-39 (sense) and 59- Mix (Applied Biosystems). Tumor protein translationally controlled 1 was AACAAAGGCGTCAACTCTGC-39 (antisense), amplified a region 350 used as a reference gene to standardize data, following the DD cycle bp upstream from the STAT1 transcription start site (TSS). Chromatin threshold method. Primers 59-TCCTGGAGAAAAGTGCTCGT-39 (sense) The Journal of Immunology 3 and 59-TCTTCGCCGAGATCAGTCAT-39 (antisense) were used for LEF1 EM cells remained stable in TEMRA cells, and only 12 CpG sites detection. Primers for tumor protein translationally controlled 1 detection (0.07%) reversed the methylation changes (Fig. 1D). Finally, we were 59-GATCGCGGACGGGTTGT-39 (sense) and 59-TTCAGCGGAGG- observed that most DMRs were located outside CpG islands, in CATTTCC-39 (antisense). LEF1 knockdown was also evaluated by flow cytometry, as previously indicated. gene bodies and 39UTR regions (Fig. 1E). DMRs within the promoter were more frequent in distal regions (2200 to 21500 nt Statistical analysis upstream of TSS) than in proximal regions (21to2200 nt up- Raw data from GenomeStudio were processed with Bioconductor project stream of TSS). packages (13) in the R statistical computing environment. Using the lumi With regard to the biological functions of the differentially package (14), raw methylation data were background corrected, log2 methylated genes, GO and Kyoto Encyclopedia of Genes and transformed, quantile adjusted for color balance, and quantile normalized. Genomes (KEGG) pathway analysis revealed that many critical Expression data were also background corrected and log2 transformed. Probes with detection values .0.01 in all samples were filtered, and the genes of the EM expression program were differentially methylated remaining values were quantile normalized. Transcripts were considered to between naive and memory cells (Supplemental Table I). These be expressed above the background level if they had detection values of p include demethylation of various essential determinants of CD8+ , 0.01. For DNA methylation analysis, sex were excluded, T cell differentiation, such as EOMES, T-box 21 protein, and and annotation of regulation and site location was recorded using Illumi- na’s IlluminaHumanMethylation450k.db package. Pairwise methylation BLIMP1 (encoded by PRDM1). In addition to these key tran- and expression comparisons were estimated between naive and EM cells scription factors, DNA methylation remodeling is associated with and between EM and TEMRA cells. A linear model was fitted to the data, many biological functions typically involved in T differentiation, and empirical Bayes-moderated t-statistics tests were calculated using the such as cytotoxicity (e.g., perforin 1 [PRF1], IFNG, granzyme B limma package (15). False discovery rates (FDRs) were calculated using the Benjamini–Hochberg procedure. In the DNA methylation array only [GZMB], GZMH, GZMK), cytoskeleton organization (e.g., actinin Downloaded from those probes with an associated FDR-adjusted p , 0.05 and an M-dif- a 1, ACTN2, ACTN4, DOCK2, NUAK2, CORO2B, SSH1), cell ference . 1.5 in each pairwise comparison were considered differentially adhesion (e.g., ITGB1, CD58, ITGA6), and chemokine signaling methylated (16). For expression analysis, a gene was considered to be (e.g., CCL5, CXCR1, CX3CR1, CXCR4). A more surprising differentially expressed when one of its associated probes showed an FDR- finding from the CD8+ T cell DNA methylome analysis is that adjusted p , 0.05 and a .1.5-fold absolute change. Gene ontology (GO) enrichment analysis was performed with the DAVID GO Web-based tool. many important genes of the TCR signaling pathway undergo de

Heat maps were generated with the gplots package. The functional protein- novo methylation in EM cells, including the CD3 complex com- http://www.jimmunol.org/ interaction network was built using STRING v10 and Cytoscape software ponent CD247 and two essential mediators of the CD28 costim- (17, 18) and then exported and visualized in Gephi software (19). Func- ulation signal and TCR activation: phospholipase Cg1 and tional interactions within the network included genomic context predic- tions, high-throughput laboratory experiments, coexpression, automated phosphatidylinositol-4,5-bisphosphate 3–kinase catalytic subunit text mining, and curated databases. Only those interactions with a com- D. Other downstream effectors of the TCR, such as p21 protein bined confidence score . 0.4 were included in the network. All raw data (Cdc42/Rac)-activated kinases 2 and 7 (PAK2 and PAK7), which were submitted to the NCBI Gene Expression Omnibus under accession are required for actin cytoskeletal remodeling triggered by TCR, number GSE83159 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi? and B cell CLL/lymphoma 2 were de novo methylated in EM cells acc=GSE83159). as well. We observed an opposite trend with respect to deme- thylation of the negative regulator of T cell response PDCD1 and by guest on September 26, 2021 Results + the Src family tyrosine kinase FYN, whose sustained expression DNA methylation profiling during human CD8 T cell and activation were linked to the negative-feedback regulation of differentiation TCR signaling. Overall, these results suggest that DNA methyl- To investigate DNA methylation dynamics during CD8+ T cell ation contributes to the tuning of TCR signaling in resting EM differentiation, we first isolated six sets of matching samples cells. corresponding to naive T cells (CD3+CD8+CD42CCR7+CD45RA+), In addition to all of the genes involved in the canonical T cell EM T cells (CD3+CD8+CD42CCR72CD45RA2), and TEMRA functions, we noted DNA methylation changes in two major T cells (CD3+CD8+CD42CCR72CD45RA+) from peripheral regulatory axes: the Wnt/b-catenin and the TGFb signaling blood (Fig. 1A). Methylation datasets were generated from two pathways. Binding of TGFb to its is associated with replicates per cell type, and each biological replicate was obtained phosphorylation of SMAD proteins, which, in turn, are associated from a pool of three samples. Two-dimensional principal com- with the suppression of lymphocyte proliferation and activation ponent analysis revealed that naive samples were clearly grouped while promoting survival of memory T cells (20, 21). During the as a distinct population, whereas EM and TEMRA samples differentiation of memory cells, 33 of the 80 genes linked to the appeared tightly clustered, indicating that these cell types have a TGFb signaling pathway (Gene Set #233725; GeneWeaver) un- very similar DNA methylation profile (Fig. 1B). Statistical anal- derwent DNA methylation changes, including five members of the ysis of differentially methylated regions (DMRs; M-difference . SMAD family (SMAD1, SMAD2, SMAD3, SMAD6, and SMAD7). 1.5, adjusted p , 0.05) revealed that 16,949 CpG sites associated Interestingly, DNA demethylation was observed in the SMAD3 with 6188 genes were differentially methylated between naive and gene and in three known cofactors of this protein: tran- EM cells (Supplemental Table I). Most of the annotated DMRs scription factor 4, E2F dimerization partner 1 (TFDP1), and the showed loss of methylation (13,849 CpG sites), whereas ,20% corepressor p107 (RBL1). In response to TGFb signaling, these were de novo methylated (3100 CpG sites) (Fig. 1C). In contrast, four proteins form a complex that translocates to the nucleus, EM and TEMRA populations were almost identical at the DNA repressing c- expression and cell proliferation (22). Thus, methylation level, and only 313 CpG sites were differentially DNA demethylation of these genes could be associated with cell- methylated between the two populations. Similar to what was cycle arrest and survival of resting memory T cells. In addition to observed in naive and EM cells, terminally differentiated cells TGF-b signaling, many genes involved in the Wnt/b-catenin accumulated more demethylation events (203 CpG sites) than de pathway, which plays a critical role during CD8 differentiation, novo methylation events (110 CpG sites). These results indicate showed associated DMRs in memory cells compared with naive that, in general, EM differentiation was primarily associated with cells. In fact, 53 of 139 genes of the Wnt/b-catenin pathway (Gene a global loss of methylation, at least within the array CpG cov- Set #233740; GeneWeaver) were differentially methylated. erage. In addition, most of the annotated DMRs between naive and These include genes encoding Wnt receptors (FZD1 and FZD4), 4 EPIGENETIC NETWORKS DURING CD8+ T CELL DIFFERENTIATION Downloaded from http://www.jimmunol.org/ by guest on September 26, 2021

FIGURE 1. DNA methylation during CD8+ T cell differentiation. (A) Representative plot of FACS-isolated T cell populations. (B) Principal component analysis based on methylation profiles of CD8+ T cell populations. (C) Scatter plots of annotated DMRs during CD8+ T cell differentiation. (D) Heat map showing methylation values of all annotated DMRs between naive and EM populations. These DMRs preserve their methylation values in TEMRA cells. (E) Distribution of DMRs between naive and memory CD8+ T cell populations according to gene locations and CpG content. Bars represent the number of probes differentially methylated (DMRs) within the total number of probes in each type of region in the array, expressed as a percentage. TSS200 is defined as the region from the TSS to 2200 nt upstream of TSS; TSS1500 covers 2200 to 21500 nt upstream of TSS. Shores are the regions 2 kb upstream and downstream of a CpG island; shelves cover the 2-kb region beyond a shore. members of the b-catenin degradation complex, such as AXIN1, datasets from the same samples used for the DNA methylation the b-catenin negative regulator CTNNBIP1, and the down- analysis. Principal component analysis based on gene-expression stream transcription factors LEF1 and TCF7. changes showed a similar sample clustering to that of DNA methylation, although EM and TEMRA cells were more clearly Correlation between DNA methylation and gene-expression + grouped as distinctive populations (Fig. 2A). For gene-expression changes during CD8 T cell differentiation analysis, a gene was considered to be a differentially expressed To investigate the correlation between DNA methylation and gene gene (DEG) if it exhibited a .1.5-fold difference and an adjusted expression in CD8+ T cell subsets, we generated gene-expression p value , 0.05. Using these criteria, we observed that gene- The Journal of Immunology 5 expression changes were much more pronounced between naive TGFb signaling pathway and included upregulated (SMAD3, and EM cells (1161 DEGs) than between EM and TEMRA cells SMAD7, and TGFBR3) and downregulated (TAB1) genes.

(105 DEGs); however, unlike the DNA methylation results, we did + not observe any specific bias toward upregulation or downregu- Histone marking in CD8 T cell memory cells lation (Fig. 2B, Supplemental Table II). Unsurprisingly, functional In addition to DNA methylation, covalent modifications of histone analysis showed that immune-related genes were highly enriched N-terminal tails contribute to the gene-expression changes ob- among annotated DEGs, especially those associated with T cell served during lymphocyte differentiation. To further explore the activation and differentiation, but it also included many genes functional interactions between gene expression and the epige- involved in chemotaxis, proliferation, and cell death (Supplemental netic landscape in CD8+ T cells, we analyzed the genomic dis- Table II). In addition, 161 genes were annotated as being tran- tribution of a histone mark associated with active transcription, scriptional regulators and included well-known T cell tran- H3K9Ac, and a heterochromatin-associated mark, H3K9me3, in scription factors, such as EOMES, RUNX3,andPRDM1,and the naive, EM, and TEMRA subsets. Genomic ChIP-sequencing many others with unknown functions in CD8+ T cells (e.g., analysis of CD8+ T cell samples showed that both histone marks HOPX1, MAFF, zinc finger and SCAN domain containing pro- are widely distributed in many genes throughout the genome tein 18 [ZSCAN18], BCOR), drawing attention to the transcrip- (Supplemental Table III). H3K9Ac had a similar genomic distri- tional complexity that is apparent during CD8+ T cell differentiation. bution in all samples and, as expected, was more commonly as- The top-ranked differentially expressed transcription factors are sociated with expressed genes (Fig. 3A). Consistent with previous shown in Table I. reports, this mark was clearly enriched around TSSs (Fig. 3B)

We next evaluated whether gene-expression changes were as- compared with gene body and 59 distal regions (23). Nonetheless, Downloaded from sociated with DNA methylation during CD8+ T cell differentiation. the region entirely overlapping with the TSS appeared to be de- We did not find any correlation between DNA methylation and void of H3K9Ac, probably as a result of nucleosome displacement gene-expression changes in TEMRA cells, but we did note a close at the active promoters. Overall, we observed many more correlation between naive and EM cells during the transition. In H3K9Ac-enriched genes in EM cells (13,625 genes with 32,623 fact, 274 of the 1161 (24%) annotated genes that were differen- overlapping peaks) and TEMRA cells (14,088 genes with 39,999

tially expressed between naive and EM cells were inversely cor- peaks) than in naive lymphocytes (8023 genes with 12,500 peaks), http://www.jimmunol.org/ related with the degree of DNA methylation (Fig. 2C). DNA suggesting a global increase in the genomic distribution of histone demethylation in EM cells was far more closely correlated with acetylation in memory cells. KEGG pathway analysis of these gene-expression changes than de novo methylation, suggesting a genes showed that this histone mark was most closely associated more important regulatory role for demethylation events. In ad- with immune-related genes, particularly with genes in the TCR dition, we wanted to investigate how DMRs in different gene re- signaling pathway (data not shown). In contrast, H3K9me3 was gions relate to gene-expression changes in CD8+ T cell subsets enriched more frequently in silenced genes, although the differ- (Fig. 2D). We found that DMRs associated with gene-expression ence between active and silenced genes was small, whereas changes were concentrated in the 59 untranslated region and gene KEGG pathway analysis showed this mark to be more common in body regions rather than in promoter regions. With respect to the genes associated with cell adhesion and neural system functions. by guest on September 26, 2021 CpG content, DMRs affecting gene expression were more fre- In addition, H3K9me3 was not significantly augmented around the quent in CpG-poor regions and were mostly absent from CpG TSS in any of the samples, but it was slightly more enriched in the islands. gene body regions of all of them. Similar to histone acetylation, To evaluate the functional relationships between DEGs asso- H3K9me3 was also excluded from the TSS-overlapping area. CpG ciated with DNA methylation changes during the differentiation of islands showed very high levels of H3K9Ac, whereas H3K9me3 naive cells into EM cells, we generated a functional protein- exhibited a converse trend. H3K9Ac was enriched at higher levels interaction network. These genes were grouped according to toward the island edges (shores). Finally, we analyzed histone their biological and molecular functions, highlighting those with marking at experimentally validated enhancer regions (VISTA more interactions within the network (Fig. 2E). Using this method, enhancers). The histone mark H3K9Ac appeared to be enriched to we observed many genes associated with T cell activation within some extent in these regions, whereas H3K9me3 was clearly ex- the grid. Most of these genes (72%) are upregulated and deme- cluded, suggesting a possible role in the regulation of enhancer thylated during the differentiation of naive cells into memory cells activity. and include some essential T cell activation and cytotoxicity Overall, H3K9Ac was enriched in regulatory regions associated genes, such as IFNG, TNF, B cell CLL/lymphoma 2, CD8a with active transcription, such as promoter regions or enhancers in molecule (CD8A), PRF1, GZMB, EOMES, and the transcription CD8+ T cell subsets, which are often subjected to dynamic factor BLIMP1 (PRDM1). We also observed upregulation and modifications linked to gene-expression and DNA-methylation demethylation in some genes associated with cytoskeleton orga- changes. This was clearly observed in some key genes associ- nization and cell adhesion, such as integrin b 2 and actinin a 4. ated with T cell functions. For instance, overexpression of the Interestingly, actinin a 1 was among the most downregulated (fold proinflammatory CCL5 in EM cells correlated with a progressive change [FC] = 213.7) and de novo methylated genes (six anno- accumulation of H3K9Ac around the TSS, spreading over the tated DMRs within the gene body and promoter) in EM cells, second and third exons in TEMRA cells (Fig. 3C), in accordance although the functional implications of the differential expression with the high CCL5 mRNA levels reported in resting CD8+ of actinin variants in lymphocytes are unknown. We also con- memory cells (24). Moreover, histone acetylation gain was ac- firmed that DNA methylation changes in genes involved in the companied by DNA demethylation in the same region, high- Wnt/b-catenin signaling pathway were often associated with lighting the dynamic coordination between various epigenetic gene-expression changes. Most of these genes (LEF1, TCF7, mechanisms at the CCL5 promoter. Nonetheless, epigenetic AXIN2, and SGK1) were downregulated and de novo methylated changes were not always associated with the canonical gene in EM cells, suggesting an epigenetic silencing associated with promoters. In the gene encoding the BLIMP1 this pathway. Finally, DNA methylation changes observed in EM (PRDM1), H3K9Ac levels increased primarily in an exonic region cells were also associated with gene-expression changes in the located ∼13 kb from the TSS, coinciding with the initiation of the 6 EPIGENETIC NETWORKS DURING CD8+ T CELL DIFFERENTIATION Downloaded from http://www.jimmunol.org/ by guest on September 26, 2021

FIGURE 2. Gene-expression changes during CD8+ T cell differentiation. (A) Principal component analysis based on expression profiles of isolated populations. (B) Scatter plots of annotated DEGs during CD8+ T cell differentiation. (C) Correlation between DNA methylation and gene-expression changes. Each bar represents the number of DEGs associated with DNA methylation changes (FC . 1.5). +, de novo methylation or gene-expression upregulation; 2, demethylation or gene-expression downregulation. (D) Distribution of DMRs associated with changes in gene expression, according to CpG content and gene region. The data are represented as the ratio of the frequency of observed DMRs in each region type associated with gene-expression changes/frequency of types of region in the Illumina HumanMethylation450 array. (E) Functional interaction network of DEGs with associated DNA methylation changes during CD8+ T cell differentiation. Genes are grouped according to major molecular and biological functions. The top 10% of genes, ranked by their centrality (number of edges) among the network, are shown as a larger size. Some overlapping biological functions are highlighted with gray boxes. coding sequence of the transcript variant 2. Furthermore, this re- from the naive to the EM phenotype, strongly suggesting the gion contains a reported DNase hypersensitivity cluster (EN- presence of an alternative TSS driving variant 2 expression in CODE) and undergoes DNA demethylation during the transition CD8+ memory cells. Conversely, in other actively transcribed The Journal of Immunology 7

Table I. Top-ranked differentially expressed transcription factors (from highest to lowest magnitude of change)

Naive Versus EM EM Versus TEMRA Up MAF, HOPX, EOMES, CEBPD, ZNF683, ZBTB16, SMAD7, MAFF, RUNX3, ZNF296, FOXP4, ZFPM1, BCOR, ZEB2, and ATF3 Down LEF1, TCEA3, ZNF274, JUNB, ZSCAN18, CEBPD, RORC, SMAD3, CREB1, ERN1, SATB1, BACH2, ZNF518B, ZNF101, MYC, FOXJ2, NR4A3, SAFB2, ZBTB16, and TCF7, ZBTB25, MYB, ZNF671, and ATF6 ZNF577 CREB1, cAMP responsive element binding protein 1; ERN1, endoplasmic reticulum to nucleus signaling protein 1; ZEB2, zinc finger E-box binding 2.

genes in memory cells, such as EOMES, H3K9Ac changes were (e.g., KRLD1, killer cell lectin-like receptor subfamily F mem- not restricted to the promoter region and showed a very high ber 1) NK cell receptors, as well as cytotoxicity-related genes, degree of enrichment throughout the entire gene sequence, even in such as GZMB (Fig. 5A, 5B). KEGG pathway analysis indicated the 39 genomic region. Naive cells showed a very high level of that genes associated with NK cytotoxicity were clearly enriched enrichment of H3K9me3 in the 59 region spanning 3 kb from the among DEGs (Fig. 5C). In addition, the functional protein- TSS, which was quickly eliminated from the memory subsets. interaction network showed that most DEGs were centered Downloaded from Dynamic changes of H3K9me3 in EOMES exemplify that this around NK cell functions (Fig. 5B), including NK cell receptors, histone mark, which is rarely enriched in regulatory regions, can as well as genes involved in NK cell activation (VAV3 and LYN) nevertheless contribute to the epigenetic silencing of some critical and NK cell homing and chemotaxis (e.g., S1P5, CCL20). Finally, genes during CD8+ T cell differentiation. Finally, loss of H3K9Ac we observed downregulation of 10 transcriptional modulators in was observed during the differentiation process. For example, TEMRA cells (Table I), most of which had not previously been

ZSCAN18, which belongs to a family of transcription factors with associated with T cell differentiation and among which we found http://www.jimmunol.org/ no known link to lymphocyte biology, showed epigenetic silenc- at least three transcription factors associated with the TGF-b ing associated with gain of methylation and loss of H3K9Ac at a pathway: SMAD3, cAMP responsive element binding protein 1, DNase-hypersensitivity cluster located in the second coding and CCAAT/enhancer binding protein d (CEBPD). Overall, gene- exon. This region corresponded to an alternative TSS pre- expression changes observed in TEMRA cells were linked to in- dicted by the EPONINE algorithm, indicating that it is prob- nate immunity and were often accompanied by epigenetic alter- ably acting as the main TSS of this gene in naive CD8+ Tcells. ations in associated regulatory regions (Fig. 5D). For instance, the In addition, ZSCAN18 showed low levels of H3K9me3 across killer cell lectin-like receptor subfamily D member 1 showed a the whole locus. Nonetheless, these histone marks do not seem very abrupt enrichment of H3K9Ac in TEMRA cells that started at to play a regulatory role in this gene, because it remained the TSS and extended over four exons. Upregulation of killer cell by guest on September 26, 2021 stable during differentiation, regardless of the observed gene- lectin-like receptor subfamily D member 1 was also coordinated expression changes. with DNA demethylation at the promoter region during differen- Finally, we wanted to evaluate how histone dynamics were tiation of EM and TEMRA cells. Following a similar trend, associated with DNA methylation and gene-expression changes in sphingosine-1-phosphate receptor 5, which is associated with NK naive and memory CD8+ T cells. To explore these epigenetic in- cell migration, showed high levels of H3K9Ac, starting at the TSS teractions within the T cell program, we annotated the loss or gain and spreading over the first coding exon. In others cases, such as of H3K9Ac and H3K9me3 peaks in the functional interaction for killer cell lectin-like receptor subfamily F member 1, H3K9Ac network that we generated previously with the DEGs associated dynamics and gene-expression changes were not associated with with DNA methylation changes (Fig. 4). As expected, most of DNA demethylation. We also observed a 39 region of ∼4 kb with these genes showed H3K9Ac enrichment, which is consistent with high levels of H3K9me3 that decreased in memory cells as tran- the global demethylation trend that we observed within this net- scription increased, suggesting a regulatory function in this region. work (Fig. 2E). This result suggests that DNA demethylation is Conversely, accumulation of H3K9me3 in the promoter and 39 frequently accompanied by H3K9Ac during CD8+ T cell differ- region of the killer cell lectin-like receptor subfamily G member 1 entiation. Nonetheless, H3K9me3 appeared unchanged in most was stable and not associated with the observed gene-expression genes, consistent with the poor association of this histone mark changes. with gene-expression changes in memory cells. LEF1 is epigenetically regulated in CD8+ T cells + Transcriptional landscape in terminally differentiated CD8 The importance of epigenetic dynamics depends on their impact on T cells gene-expression changes, especially for those genes that may play TEMRA cells represent an EM subset associated with chronic Ag key regulatory roles during CD8+ T cell differentiation. Accord- exposure and immune system aging that is characterized by a ingly, we found an inverse correlation between gene expression decreased proliferation potential and conserved strong effector and DNA methylation changes for 40 transcriptional regulators activity (25, 26). As indicated previously, EM and TEMRA cells between naive and memory cells, some of which were not de- were very similar at the epigenetic and gene-expression levels. scribed previously in CD8+ T cells. For instance, we observed DNA methylation changes annotated between the two subsets demethylation and upregulation of the endoplasmic reticulum to were primarily associated with cell adhesion (Supplemental nucleus signaling protein 1, which is an effector of endoplasmic Table I) and were rarely correlated with gene-expression reticulum stress and a positive regulator of proliferation-related changes. In contrast, DEGs were primarily associated with genes. Another interesting example is the zinc finger E-box immune functions. Among these, we observed upregulation of binding homeobox 2, which was associated with effector and many Ig-like (e.g., KIR2DL4, KIR3DL1) and C-type lectin-like memory functions in murine CD8+ T cells (27–29). The fact that 8 EPIGENETIC NETWORKS DURING CD8+ T CELL DIFFERENTIATION Downloaded from http://www.jimmunol.org/ by guest on September 26, 2021

FIGURE 3. Histone marking in CD8+ T cell populations. (A) Correlations between gene expression and histone marking. Genes annotated as expressed and not expressed in CD8+ T cell samples were plotted against the number of overlapping peaks for each gene sequence. Total and partial overlapping peaks were included. Only statistically significant peaks (p , 1025) were used in the analysis. Boxes indicate mean values and error bars represent SD. Significant differences were assessed by the Kruskal–Wallis test. (B) Density plots in 100-bp windows showing the genome-wide distribution of H3K9Ac and H3K9me3 around regulatory regions and genes in memory CD8+ T cells. Density within the gene body, CpG islands, and enhancer regions was normalized to represent the relative enrichment of each feature length. (C) Epigenetic remodeling and gene-expression changes in key memory CD8+ T cells genes. The TSS is indicated by an arrow. H3K9Ac, H3K9me3, and DNase-hypersensitive clusters are represented across the gene length. The y-axis indicates tag counts. Uniform scales are used for all samples within each gene. Black dots represent differentially methylated CpG sites associated with each gene.DNA methylation levels of each CpG site are represented by a dashed line in the graph. this gene is demethylated and upregulated in human EM cells ous transcriptional regulators with unknown functions in CD8+ strongly supports these results. Conversely, we found de novo T cells, such as chromobox homolog 5, which is a heterochromatin- methylation coordinated with expression downregulation in vari- associated protein associated with transcriptional silencing. Irrespective The Journal of Immunology 9 Downloaded from http://www.jimmunol.org/ by guest on September 26, 2021

FIGURE 4. Histone marking association with DEGs associated with DNA methylation changes during CD8+ T cell differentiation. To show the cor- relation among DNA methylation, gene expression, and histone marking, H3K9Ac and H3K9me3 dynamics are indicated within the functional interaction network previously generated with DEGs associated with DNA methylation changes. Histone mark changes are expressed as an increase (red) or loss (blue) of the absolute number of overlapping peaks in each gene in the network. Stable histone marking is indicated in yellow. of this, one of these transcription factors, LEF1, was the most flow cytometry. To explore the role of LEF1 further, we down- highly downregulated gene in EM cells (FC = 219.8). LEF1 regulated this gene in naive cells by small interfering RNA downregulation appeared to be associated with loss of H3K9Ac transfection (Fig. 6C). Unexpectedly, the gene-expression profile and gain of methylation at the promoter region, coinciding with a showed a very moderate effect of LEF1 knockdown on naive predicted DNase-hypersensitivity site (Fig. 6A) and indicating cells (Fig. 6D, Supplemental Table IV). The most significantly that this gene is subjected to robust epigenetic silencing during downregulated gene was STAT1, suggesting a functional link differentiation. This transcription factor is a downstream effector between the Wnt/b-catenin and STAT1 signaling pathways. of the Wnt/b-catenin pathway, which was associated previously Nonetheless, the ChIP experiment did not indicate any signifi- with memory functions in CD8+ T cells (30). Nonetheless, it is cant binding of LEF1 to the STAT1 promoter region, indicating unclear why this factor is so highly expressed in the non- that the effect of LEF1 is likely to be indirect or to occur through stimulated naive subset. A high level of expression of LEF1 in binding to nonproximal promoter regions (Fig. 6E). However, we naive cells was confirmed at the protein level by flow cytometry were able to confirm that blockade of the Wnt/b-catenin sig- (Fig. 6B). LEF1 expression showed progressive downregulation naling pathway with a specific inhibitor of the b-catenin tran- during CD8+ T cell differentiation, achieving maximum levels in scriptional coactivator CBP reduced STAT1 protein levels in naive cells and minimum levels in EM and TEMRA cells. This resting and TCR-stimulated naive CD8+ T cells and almost to- expression pattern was very similar to the distinctive CCR7 tally eliminated STAT1 phosphorylation in stimulated cells (Fig. expression in lymphocytes, allowing a comparable discrimina- 6F). Finally, b-catenin inhibition visibly reduced the expansion tion of the naive, central memory, EM, and TEMRA subsets by of naive CD8+ T cells upon TCR stimulation but did not clearly 10 EPIGENETIC NETWORKS DURING CD8+ T CELL DIFFERENTIATION Downloaded from http://www.jimmunol.org/

FIGURE 5. Transcriptional landscape in terminally differentiated CD8+ T cells (TEMRA). (A) Comparison of gene-expression changes in EM cells (horizontal axis) and TEMRA cells (vertical axis) versus naive cells. (B) Functional interaction network generated with only the annotated DEGs between EM and TEMRA cells. The most relevant functions are highlighted. Network centrality is indicated by a grayscale. (C) KEGG pathway analysis of DEGs between EM and TEMRA cells. (D) Epigenetic marking and gene-expression pattern of NK cell–related genes. Black dots represent differentially by guest on September 26, 2021 methylated CpG sites associated with each gene. DNA methylation levels of each CpG site are represented with a dashed line in the graph. affect the activation markers CD69 and IL-2R (CD25), indicat- activation, such as IFNG or TNF, are also demethylated in ing that this pathway is only involved in some aspects of the memory cells, indicating that DNA methylation contributes, in T cell–activation program (Fig. 6G). part, to the primed state in memory cells, allowing fast gene transcription in response to forthcoming regulatory events. Moreover, DNA demethylation seems to make a more prominent Discussion contribution to this primed state given that de novo methylation is Our results provide new evidence of the coordinated participation much less frequent in memory cells. Our results indicate that at of DNA methylation and histone marking in the transcriptional least those CpG sites within the microarray coverage show a networks associated with EM functions. First, we demonstrated global trend toward demethylation, similar to what other analyses that memory cells present an intense epigenetic remodeling linked revealed about the differentiation of other hematopoietic cells, to gene-expression changes compared with naive cells. Many of the such as B lymphocytes and intrathymic progenitors (31–33). observed DNA methylation changes were associated with effector Histone marking is also very dynamic during T cell differen- functions and were similar to those observed previously after tiation. Trimethylation of lysines 4 and 27 in histone H3 is known to in vivo activation of murine naive cells, including cytotoxicity- be highly dynamic during the activation and differentiation of associated genes (e.g., PRF1, GZMK, GZMB, IFNG), members CD8+ lymphocytes (6–8). We extended these studies by analyzing of the Wnt signaling pathway (TCF7 and LEF1), and those in- an activating and a repressing epigenetic mark associated with a volved in homing and migration (CCL5 and CCR2) and apoptosis specific histone residue: lysine 9 in histone H3 . The histone mark (ANXA1 and ANXA2) (9). These indicate that, to some extent, H3K9Ac is a well studied epigenetic modification that is typically memory cells retain the epigenetic changes acquired during acti- associated with active gene transcription. We observed that active vation, and that these probably facilitate activation upon Ag genes frequently present acetylation enrichment that extends from reencounter. For instance, the proinflammatory CCL5 is deme- the promoter region throughout the entire length of the gene, thylated in effector cells (9) and, according to our findings, these rather than being restricted to promoter regions. These acetylation demethylation changes are maintained in memory cells. This re- dynamics are often observed in potential DNase-hypersensitivity sult is also consistent with the high levels of CCL5 mRNA clusters (predicted by ENCODE) and in coordination with DNA reported in resting memory cells that allow these cells to secrete methylation changes, allowing the identification of relevant reg- CCL5 protein immediately upon TCR triggering (24). In general, ulatory regions and alternative TSS. Another interesting fact is the many important genes typically upregulated upon CD8+ T cells overall increase in the number of genes with H3K9Ac peaks in The Journal of Immunology 11 Downloaded from http://www.jimmunol.org/

+ FIGURE 6. Stage-dependent epigenetic regulation of LEF1 during CD8 T cell differentiation. (A) H3K9Ac, H3K9me3, DNA methylation, and mRNA by guest on September 26, 2021 expression of the LEF1 gene. Black dots represent differentially methylated CpG sites associated with LEF1 during CD8+ T cell differentiation. DNA methylation levels of each CpG site are represented by a gray line in the graph. (B) LEF1 expression in CD8+ T cell populations analyzed by flow cytometry. Representative graphs of LEF1 expression in naive, EM, and TEMRA populations (right panels). (C) LEF1 knockdown in naive CD8+ T cells. LEF1 expression was analyzed by quantitative PCR and flow cytometry. (D) Heat map showing gene-expression changes in naive CD8+ T cells after LEF1 knockdown. (E) Quantitative ChIP analysis of LEF1 occupancy at the STAT1 promoter region. (F) Western blot analysis of STAT1 and p-STAT1 in resting and TCR-stimulated naive CD8+ T cells after 72 h of treatment with the Wnt/b-catenin signaling pathway inhibitor ICG001. (G) Flow cytometry analysis of cell proliferation (CFSE staining) and the activation markers CD25 and CD69 after treatment with ICG001.

resting memory cells compared with naive cells. This suggest an In addition, H3K9Ac is highly enriched in CpG islands, although expansion of the H3K9Ac genomic distribution in memory this result was expected because most gene promoters have an cells, although this was not accompaniedbyanincreaseinthe associated CpG island. It is also highly enriched in experimentally absolute number of transcribed genes. Recently, it was observed validated enhancer regions, which is consistent with the reported that lymphocyte activation is associated with chromatin remodeling, presence of this histone mark in active enhancers in embryonic which is partially maintained in memory cells to preserve these stem cells (23). Therefore, H3K9Ac has a regulatory role beyond cells in a primed state (34). Some transcription factors and his- gene promoters, which highlights the importance of mapping this tone marks, such as dimethylated lysine 4 and acetylated lysine histone mark to identified active regulatory regions in CD8+ 27 in histone H3, seem to contribute to the generation and T cells. maintenance of these chromatin changes. Thus, the combination In contrast, the observed genomic distribution of H3K9me3 was of global DNA demethylation and histone acetylation that we vastly different and not as well correlated with gene expression. observed may reflect a transcriptionally poised state in memory This histone modification is a less well-studied epigenetic mark cells that allows faster response upon Ag re-encounter. Indeed, that was associated previously with heterochromatin formation. histone acetylation preceding gene transcription was observed in However, we only found a modest association between H3K9me3 some genes in memory CD8+ cells (35–37). Our data support and gene silencing, which argues against it having an exclusive role these results at the genomic scale. Overall, the evidence seems as a repressing mark. Indeed, poor association with gene repression to indicate that memory maintenance in memory lymphocytes is was reported in human CD4+ T cells in which H3K9me3 was often dependent on a complex chromatin-remodeling mechanism that observed in actively transcribed genes (38). We also observed this takes place primarily during the activation process in which behavior in expressed genes, such as ZSCAN18 or killer cell histone marking, DNA methylation, and transcription factor lectin-like receptor subfamily G member 1, raising the question of binding intervene. how this histone mark contributes to transcriptional regulation. 12 EPIGENETIC NETWORKS DURING CD8+ T CELL DIFFERENTIATION

These results indicate that more studies are required to determine observed downregulation of STAT1 after LEF1 knockdown, sug- the function of H3K9me3 in active genes. However, it is important gesting that the b-catenin axis could be associated with STAT1 to point out that we observed dynamic changes of this histone transcription and activation. This was demonstrated by the fact mark that were associated with gene silencing in some key genes, that inhibition of b-catenin with a specific inhibitor of the tran- such as in the EOMES promoter; therefore, the regulatory role of scriptional coactivator CBP reduced STAT1 levels in naive cells H3K9me3 during differentiation is probably dependent on the while preventing STAT1 phosphorylation upon activation. STAT1 chromatin context. is required for clonal expansion of activated CD8+ T cells (45), Our findings enable the comparison of EM and TEMRA cells. and our results suggest that b-catenin inhibition has a cytostatic Although the frequency of TEMRA cells has been associated with effect that could be mediated by STAT1, even though the molec- chronic infection and aging (39), the mechanisms underlying the ular mechanism mediating STAT1 transcription through this generation of this cell subset are not well understood. Conse- pathway remains to be elucidated. In contrast, a recent study quently, comprehensive molecular profiling is required to under- showed that LEF1 has intrinsic histone deacetylase activity, stand the phenotypic characteristics of the TEMRA compartment. allowing it to act as a transcriptional repressor (46). Thus, it could Previously, it was observed that progressive epigenetic changes be hypothesized that the histone acetylation dynamics observed are associated with CD8+ T cell differentiation (7). However, one during lymphocyte differentiation are partially dependent on important conclusion of our analysis is the remarkable similarity LEF1 downregulation in memory cells, contributing to the acti- between EM and TEMRA cells at the transcriptional and epige- vation of genes associated with effector and memory functions. netic levels. This suggests that TEMRA cells may not be a truly Further studies are needed to clarify the role of LEF1 histone distinct population and may only represent aged EM cells or deacetylase activity during CD8+ T cell activation and differen- Downloaded from possibly fully functional EM cells in a different activation state. tiation. Regardless of these similarities, we found a small difference that In summary, our study provides a comprehensive resource that highlights some traits typically associated with terminal differ- identifies key epigenetic changes that take place in a coordinated entiation, such as overexpression of GZMB and downregulation of manner during CD8+ T cell differentiation, greatly improving our PD-1 (40, 41). Gene-expression analysis showed differential ex- understanding of how the regulatory networks contribute to the

pression of only 105 genes, in contrast to the 1161 annotated acquisition and maintenance of the transcriptional program. In the http://www.jimmunol.org/ genes that differ between naive and EM cells. Perhaps a more future, more refined epigenetic maps that include epigenetic significant observation is that a strikingly large fraction of these modifications and the molecular signals associated with specific genes are related to innate immunity and NK cytotoxicity. In fact, epigenetic dynamics will pave the way for novel therapeutic the innate phenotype that we observed is more evident than what strategies based on the epigenetic reprogramming of immune was reported previously and includes upregulation of 12 NK re- cells. ceptors, mediators of NK activation and NK homing, which may indicate a unique capability of these cells to access specific mi- Disclosures croenvironments and to respond in the absence of TCR priming. The authors have no financial conflicts of interest. This innate phenotype observed in TEMRA cells is highly rele- by guest on September 26, 2021 vant in the context of aging and inflammation and should be studied further. References Comprehensive molecular studies that include several epigenetic 1. Joshi, N. S., W. Cui, A. Chandele, H. K. Lee, D. R. Urso, J. Hagman, L. Gapin, and S. M. Kaech. 2007. Inflammation directs memory precursor and short-lived layers are required to derive detailed maps of the transcriptional effector CD8(+) T cell fates via the graded expression of T-bet transcription networks associated with the differentiation process and the factor. Immunity 27: 281–295. maintenance of cell identity. Using this approach, we identified 2. Intlekofer, A. M., N. Takemoto, E. J. Wherry, S. A. Longworth, J. T. Northrup, + V. R. Palanivel, A. C. Mullen, C. R. Gasink, S. M. Kaech, J. D. Miller, et al. LEF1 as one of the most critically regulated genes in CD8 sub- 2005. Effector and memory CD8+ T cell fate coupled by T-bet and eomeso- sets. The strong epigenetic silencing observed in EM cells, which dermin. Nat. Immunol. 6: 1236–1244. involves a gain of DNA methylation and a loss of H3K9Ac at the 3. Rutishauser, R. L., G. A. Martins, S. Kalachikov, A. Chandele, I. A. Parish, E. Meffre, J. Jacob, K. Calame, and S. M. Kaech. 2009. Transcriptional repressor promoter and gene body region, and the steep transcriptional Blimp-1 promotes CD8(+) T cell terminal differentiation and represses the ac- dynamics between naive and EM cells strongly suggest a critical quisition of central memory T cell properties. Immunity 31: 296–308. role during CD8+ differentiation. Although LEF1 downregulation 4. Masson, F., M. Minnich, M. Olshansky, I. Bilic, A. M. Mount, A. Kallies, T. P. Speed, M. Busslinger, S. L. Nutt, and G. T. Belz. 2013. Id2-mediated in- in EM cells is well documented, its role during differentiation is hibition of E2A represses memory CD8+ T cell differentiation. J. Immunol. 190: not as clear (42). This transcription factor is a downstream effector 4585–4594. of the Wnt/b-catenin pathway, which was recently associated with 5. Zhou, X., S. Yu, D. M. Zhao, J. T. Harty, V. P. Badovinac, and H. H. Xue. 2010. Differentiation and persistence of memory CD8(+) T cells depend on T cell memory formation. It was reported that overexpression of stabi- factor 1. Immunity 33: 229–240. + lized b-catenin in CD8 cells hampers effector functions while 6. Araki, Y., Z. Wang, C. Zang, W. H. Wood, III, D. Schones, K. Cui, T. Y. Roh, favoring memory T cell formation (43). In contrast, concomitant B. Lhotsky, R. P. Wersto, W. Peng, et al. 2009. Genome-wide analysis of histone deletion of LEF1 and TCF1 greatly diminishes the memory re- methylation reveals chromatin state-based regulation of gene transcription and function of memory CD8+ T cells. Immunity 30: 912–925. sponse, indicating that the effect of b-catenin on memory for- 7. Crompton, J. G., M. Narayanan, S. Cuddapah, R. Roychoudhuri, Y. Ji, W. Yang, mation depends, at least in part, on these two downstream S. J. Patel, M. Sukumar, D. C. Palmer, W. Peng, et al. 2015. Lineage relationship transcription factors (44). LEF1 deficiency alone only moderately of CD8 T cell subsets is revealed by progressive changes in the epigenetic landscape. Cell. Mol. Immunol. 13: 502–513. impairs the memory response; thus, it is not clear why this gene is 8. Russ, B. E., M. Olshanksy, H. S. Smallwood, J. Li, A. E. Denton, J. E. Prier, so strongly expressed in naive cells and is epigenetically silenced A. T. Stock, H. A. Croom, J. G. Cullen, M. L. Nguyen, et al. 2014. Distinct in EM cells. To further explore the transcriptional function of epigenetic signatures delineate transcriptional programs during virus-specific + CD8(+) T cell differentiation. [Published erratum appears in 2014 Immunity LEF1, we performed a knockdown experiment on naive CD8 41: 1064.] Immunity 41: 853–865. T cells, although the observed effect was remarkably modest at the 9. Scharer, C. D., B. G. Barwick, B. A. Youngblood, R. Ahmed, and J. M. Boss. transcriptional level. This result argues in favor of a redundant 2013. Global DNA methylation remodeling accompanies CD8 T cell effector function. J. Immunol. 191: 3419–3429. effect of LEF1 and other related transcription factors involved in 10. Langmead, B., and S. L. Salzberg. 2012. Fast gapped-read alignment with the b-catenin pathway, such as TCF1 and TCF7. Nonetheless, we Bowtie 2. Nat. Methods 9: 357–359. The Journal of Immunology 13

11. Feng, J., T. Liu, B. Qin, Y. Zhang, and X. S. Liu. 2012. Identifying ChIP-seq 30. Thaventhiran, J. E., D. T. Fearon, and L. Gattinoni. 2013. Transcriptional reg- enrichment using MACS. Nat. Protoc. 7: 1728–1740. ulation of effector and memory CD8+ T cell fates. Curr. Opin. Immunol. 25: 12. Kharchenko, P. V., M. Y. Tolstorukov, and P. J. Park. 2008. Design and analysis 321–328. of ChIP-seq experiments for DNA-binding proteins. Nat. Biotechnol. 26: 1351– 31. Lee, S. T., Y. Xiao, M. O. Muench, J. Xiao, M. E. Fomin, J. K. Wiencke, 1359. S. Zheng, X. Dou, A. de Smith, A. Chokkalingam, et al. 2012. A global DNA 13. Gentleman, R. C., V. J. Carey, D. M. Bates, B. Bolstad, M. Dettling, S. Dudoit, methylation and gene expression analysis of early human B-cell development B. Ellis, L. Gautier, Y. Ge, J. Gentry, et al. 2004. Bioconductor: open software reveals a demethylation signature and transcription factor network. Nucleic development for computational biology and bioinformatics. Genome Biol. 5: Acids Res. 40: 11339–11351. R80. 32. Rodriguez, R. M., B. Suarez-Alvarez, D. Mose´n-Ansorena, M. Garcı´a-Peydro´, 14. Du, P., W. A. Kibbe, and S. M. Lin. 2008. lumi: a pipeline for processing Illu- P. Fuentes, M. J. Garcı´a-Leo´n, A. Gonzalez-Lahera, N. Macias-Camara, mina microarray. Bioinformatics 24: 1547–1548. M. L. Toribio, A. M. Aransay, and C. Lopez-Larrea. 2015. Regulation of the 15. Smyth, G. K. 2004. Linear models and empirical bayes methods for assessing transcriptional program by DNA methylation during human ab T-cell devel- differential expression in microarray experiments. Stat. Appl. Genet. Mol. Biol. opment. Nucleic Acids Res. 43: 760–774. 3: Article3. doi:10.2202/1544-6115.1027. 33. Rodriguez, R. M., C. Lopez-Larrea, and B. Suarez-Alvarez. 2015. Epigenetic 16. Du, P., X. Zhang, C. C. Huang, N. Jafari, W. A. Kibbe, L. Hou, and S. M. Lin. dynamics during CD4(+) T cells lineage commitment. Int. J. Biochem. Cell Biol. 2010. Comparison of Beta-value and M-value methods for quantifying 67: 75–85. methylation levels by microarray analysis. BMC Bioinformatics 11: 587. 34. Bevington, S. L., P. Cauchy, J. Piper, E. Bertrand, N. Lalli, R. C. Jarvis, 17. Shannon, P., A. Markiel, O. Ozier, N. S. Baliga, J. T. Wang, D. Ramage, L. N. Gilding, S. Ott, C. Bonifer, and P. N. Cockerill. 2016. Inducible chromatin N. Amin, B. Schwikowski, and T. Ideker. 2003. Cytoscape: a software envi- priming is associated with the establishment of immunological memory in ronment for integrated models of biomolecular interaction networks. Genome T cells. EMBO J. 35: 515–535. Res. 13: 2498–2504. 35. Araki, Y., M. Fann, R. Wersto, and N. P. Weng. 2008. Histone acetylation fa- 18. Szklarczyk, D., A. Franceschini, S. Wyder, K. Forslund, D. Heller, J. Huerta- cilitates rapid and robust memory CD8 T cell response through differential ex- Cepas, M. Simonovic, A. Roth, A. Santos, K. P. Tsafou, et al. 2015. STRING pression of effector molecules (eomesodermin and its targets: perforin and v10: protein-protein interaction networks, integrated over the tree of life. Nucleic granzyme B). J. Immunol. 180: 8102–8108. Acids Res. 43: D447–D452.

36. Dispirito, J. R., and H. Shen. 2010. Histone acetylation at the single-cell level: a Downloaded from 19. Bastian, M., S. Heymann, and M. Jacomy. 2009. Gephi: an open source software marker of memory CD8+ T cell differentiation and functionality. J. Immunol. for exploring and manipulating networks. Available at: https://gephi.org/ 184: 4631–4636. publications/gephi-bastian-feb09.pdf. 37. Fann, M., J. M. Godlove, M. Catalfamo, W. H. Wood, III, F. J. Chrest, N. Chun, 20. Letterio, J. J. 2005. TGF-beta signaling in T cells: roles in lymphoid and epi- L. Granger, R. Wersto, K. Madara, K. Becker, et al. 2006. Histone acetylation thelial neoplasia. Oncogene 24: 5701–5712. is associated with differential gene expression in the rapid and robust memory 21. Filippi, C. M., A. E. Juedes, J. E. Oldham, E. Ling, L. Togher, Y. Peng, CD8(+) T-cell response. Blood 108: 3363–3370. R. A. Flavell, and M. G. von Herrath. 2008. Transforming growth factor-beta 38. Barski, A., S. Cuddapah, K. Cui, T. Y. Roh, D. E. Schones, Z. Wang, G. Wei, suppresses the activation of CD8+ T-cells when naive but promotes their survival I. Chepelev, and K. Zhao. 2007. High-resolution profiling of histone methyl-

and function once antigen experienced: a two-faced impact on autoimmunity. http://www.jimmunol.org/ ations in the . Cell 129: 823–837. Diabetes 57: 2684–2692. € 22. Shi, Y., and J. Massague´. 2003. Mechanisms of TGF-beta signaling from cell 39. Fulo¨p, T., A. Larbi, and G. Pawelec. 2013. Human T cell aging and the impact of membrane to the nucleus. Cell 113: 685–700. persistent viral infections. Front. Immunol. 4: 271. 23. Karmodiya, K., A. R. Krebs, M. Oulad-Abdelghani, H. Kimura, and L. Tora. 40. Appay, V., R. A. van Lier, F. Sallusto, and M. Roederer. 2008. Phenotype and 2012. H3K9 and H3K14 acetylation co-occur at many gene regulatory elements, function of human T lymphocyte subsets: consensus and issues. Cytometry A 73: while H3K14ac marks a subset of inactive inducible promoters in mouse em- 975–983. bryonic stem cells. BMC Genomics 13: 424. 41. Duraiswamy, J., C. C. Ibegbu, D. Masopust, J. D. Miller, K. Araki, G. H. Doho, 24. Marc¸ais, A., C. A. Coupet, T. Walzer, M. Tomkowiak, R. Ghittoni, and P. Tata, S. Gupta, M. J. Zilliox, H. I. Nakaya, et al. 2011. Phenotype, function, J. Marvel. 2006. Cell-autonomous CCL5 transcription by memory CD8 T cells is and gene expression profiles of programmed death-1(hi) CD8 T cells in healthy regulated by IL-4. J. Immunol. 177: 4451–4457. human adults. J. Immunol. 186: 4200–4212. 25. Geginat, J., A. Lanzavecchia, and F. Sallusto. 2003. Proliferation and differen- 42. Willinger, T., T. Freeman, M. Herbert, H. Hasegawa, A. J. McMichael, and M. F. Callan. 2006. Human naive CD8 T cells down-regulate expression of the tiation potential of human CD8+ memory T-cell subsets in response to antigen or by guest on September 26, 2021 homeostatic cytokines. Blood 101: 4260–4266. WNT pathway transcription factors lymphoid enhancer binding factor 1 and 26. Setoguchi, R., Y. Matsui, and K. Mouri. 2015. mTOR signaling promotes a transcription factor 7 (T cell factor-1) following antigen encounter in vitro and robust and continuous production of IFN-g by human memory CD8+ T cells and in vivo. J. Immunol. 176: 1439–1446. their proliferation. Eur. J. Immunol. 45: 893–902. 43. Zhao, D. M., S. Yu, X. Zhou, J. S. Haring, W. Held, V. P. Badovinac, J. T. Harty, 27. Best, J. A., D. A. Blair, J. Knell, E. Yang, V. Mayya, A. Doedens, M. L. Dustin, and H. H. Xue. 2010. Constitutive activation of Wnt signaling favors generation and A. W. Goldrath, Immunological Genome Project Consortium. 2013. Tran- of memory CD8 T cells. J. Immunol. 184: 1191–1199. scriptional insights into the CD8(+) T cell response to infection and memory 44. Zhou, X., and H. H. Xue. 2012. Cutting edge: generation of memory precursors T cell formation. Nat. Immunol. 14: 404–412. and functional memory CD8+ T cells depends on T cell factor-1 and lymphoid 28. Dominguez, C. X., R. A. Amezquita, T. Guan, H. D. Marshall, N. S. Joshi, enhancer-binding factor-1. J. Immunol. 189: 2722–2726. S. H. Kleinstein, and S. M. Kaech. 2015. The transcription factors ZEB2 and T- 45. Quigley, M., X. Huang, and Y. Yang. 2008. STAT1 signaling in CD8 T cells is bet cooperate to program cytotoxic T cell terminal differentiation in response to required for their clonal expansion and memory formation following viral in- LCMV viral infection. J. Exp. Med. 212: 2041–2056. fection in vivo. J. Immunol. 180: 2158–2164. 29. Omilusik, K. D., J. A. Best, B. Yu, S. Goossens, A. Weidemann, J. V. Nguyen, 46. Xing, S., F. Li, Z. Zeng, Y. Zhao, S. Yu, Q. Shan, Y. Li, F. C. Phillips, E. Seuntjens, A. Stryjewska, C. Zweier, R. Roychoudhuri, et al. 2015. Tran- P. K. Maina, H. H. Qi, et al. 2016. Tcf1 and Lef1 transcription factors establish scriptional repressor ZEB2 promotes terminal differentiation of CD8+ effector CD8(+) T cell identity through intrinsic HDAC activity. Nat. Immunol. 17: 695– and memory T cell populations during infection. J. Exp. Med. 212: 2027–2039. 703.