CD137 (4-1BB) Costimulation Modifies DNA Methylation in CD8⁺ T Cell–Relevant Genes

Costimulation dictates the outcome of antigen recognition by T cells. Immediate changes in signaling, gene expression, and metabolism take place if costimulation is provided (1) but, in addition, subtle long-term changes occur that leave T cells poised to more robustly respond if challenged at a later time point with antigen (2). Several gene-expression control mechanisms may be involved in such long-term regulation. Transcription factors do not adequately explain the long-term effects, which suggests a role for epigenetic modifications (2, 3). Epigenetic control of gene expression acts as a switch to either induce or repress the transcriptional activity of multiple genes implicated in different physiological and pathological conditions. Specific epigenetic mechanisms have been identified as being responsible for regulating the expression of certain immune-related genes (3, 4). One of these epigenetic mechanisms is DNA methylation, which consists of the addition of a methyl group to the 5′ carbon of cytosine within cytosine-guanine dinucleotides (CpG). CpG DNA methylation is considered perhaps the most fundamental molecular phenomenon determining chromatin accessibility to the transcriptional machinery and thus leads to gene expression regulation. Gene-specific DNA methylation is largely dependent on the activity of DNA methyltransferases (DNMT1, DNMT3a, and DNMT3b) that catalyze the transfer of a methyl group from S-adenosyl methionine to DNA (5). These enzymes are drawn onto selective genome locations to methylate cytosine bases in a sequence-specific fashion by poorly understood targeting mechanisms. Such methylation patterns are subsequently inherited by daughter cells following mitoses (6). Alterations in methylation patterns influence the balance of transcripts in cells and contribute to pathological conditions such as cancer and the deregulation of the immune system (7).

Critical loci in T lymphocytes are regulated by gene methylation and chromatin accessibility, including the FOXP3 locus in natural Tregs (8), the PD-1 locus (PDCD1) in exhausted T cells (9), and the differentiation to the IFNγ-producing phenotype under the influence of IL12 (10). How the epigenetic machinery selectively controls these phenomena in a gene-specific manner remains poorly understood. Genome-wide approaches to resolving this issue have seldom been undertaken (3, 9).

In T-cell activation, costimulation via the TNFR family members is key to survival, acquisition of effector functions, and memory differentiation (11–13). CD137 (4-1BB, TNFRSF9) is not an exception (14). It gains surface expression on T cells only following TCR-mediated priming (15), although its levels are augmented by CD28 costimulation (16). Ample experimental evidence shows that when CD137 meets its ligand or agonist monoclonal antibodies (mAb) on CD8⁺ T cells, effector functions (17), survival (18), and memory generation are costimulated (13, 19–21). In this sense, agonist mAbs potentiate curative immune responses in mice bearing tumors through immune mechanisms mediated primarily by CD8⁺ cytotoxic T lymphocytes (14, 22). In this regard, the fully human agonist mAbs to CD137, urelumab (23) and utomilumab (24), are undergoing clinical trials as single agents or in combination with other immunostimulatory mAbs (14). In rodent models, CD137 engagement leads to more robust T-cell responses, even when such mice are rechallenged with cognate antigen months after treatment (25, 26), indicating the need for long-term regulation mechanisms imprinted during the primary response (14).

Hypothesizing that such long-term effects of CD137 ligation could be the consequence of epigenetic changes encompassing chromatin remodeling, we performed experiments with genome-wide high-throughput DNA methylation arrays to identify immune genes on which CD137 would influence cytosine methylation patterns at specific motifs. Using human primary CD8⁺ T cells from series of healthy volunteers, the DNA methylation changes were confirmed and found to result in up- or downregulation of mRNA and protein expression of such genes.

Human CD8⁺ T lymphocytes were isolated from the peripheral blood of healthy donors by Ficoll gradients, following a negative selection with CD8⁺ T Cell Isolation Kit by autoMACS Pro (Miltenyi Biotec). Blood samples were obtained from Navarra Blood and Tissue Bank. Navarrabiomed Biobank, Navarra Health Department. CD8⁺ T lymphocytes were activated in 12-well plates previously coated with anti-CD3ϵ (1 μg/mL, clone OKT3) and anti-CD137 (10 μg/mL, 6B4 or urelumab) or respective isotype-matched control Ab (10 μg/mL) at 1.75 × 10⁶ cells/well in RPMI 1640 medium (Gibco) supplemented with 10% FBS (Sigma-Aldrich), 100 IU/mL penicillin and 100 μg/mL streptomycin (Gibco) for 5 days (activation period). On day 5, T lymphocytes were transferred onto 12-well plates in culture media supplemented with hIL7 (25 ng/μL, Immunotools) for another 5 days (resting period). Restimulation was attained by transferring of T cells to plates coated with anti-CD3ϵ mAb (1 μg/mL, clone OKT3).

DNA from CD8⁺ T lymphocytes was isolated using DNeasy Blood & Tissue kit (Sigma) and quantified by Quant-iT PicoGreen dsDNA Reagent (Invitrogen). The integrity was analyzed in a 1.3% agarose gel. Bisulfite conversion of 600 ng of each DNA sample was performed according to the manufacturer's recommendation for Illumina Infinium Assay. Effective bisulfite conversion was checked for three controls that were converted simultaneously with the samples. Four microliters of bisulfite converted DNA were used to hybridize on Infinium Human Methylation 450 BeadChip, following the Illumina Infinium HD Methylation protocol. Chip analysis was performed using Illumina HiScan SQ fluorescent scanner. The intensities of the images were extracted using GenomeStudio (2010.3) Methylation module (1.8.5) software. The methylation score of each CpG is represented as the beta (β) value.

The 450K DNA methylation array by Illumina is an established, highly reproducible method for DNA methylation detection and has been validated in two independent laboratories (27). The 450K DNA methylation array includes 485,764 cytosine positions of the human genome that were filtered by sex chromosome CpGs (avoiding sex link alterations) and nonvalid CpGs (P < 0.001). The intensities of the images were extracted and normalized using GenomeStudio (2011.1) Methylation module (1.9.0) software.

For determining differentially methylated CpGs an analysis using an absolute difference in beta values of 0.25 and a standard deviation <0.1 were used for selecting the most relevant positions.

Pyrosequencing analyses to determine CpG methylation status were developed as previously described (28). Briefly, a minimum of 500 ng of DNA were converted using the EZ DNA methylation Gold (ZYMO RESEARCH) bisulfite conversion kit following the manufacturer's recommendations. Specific sets of primers for PCR amplification and sequencing were designed using specific software (PyroMark assay design version 2.0.01.15). Primer sequences were designed, when possible, to hybridize with CpG-free sites to ensure methylation-independent amplification (see Supplementary Table S3). PCR was performed under standard conditions with biotinylated primers and the PyroMark Vacuum Prep Tool (Biotage) was used to prepare single-stranded PCR products according to the manufacturer's instructions. PCR products were observed on 2% agarose gels before pyrosequencing. Reactions were performed in a PyroMark Q24 System version 2.0.6 (Qiagen) using appropriate reagents and protocols, and the methylation value was obtained from the average of the CpG dinucleotides included in the sequence analyzed. Controls to assess correct bisulfite conversion of the DNA were included in each run, as well as sequencing controls to ensure the reliability of the measurements. Graphic representation of methylation values shows bars identifying CpG sites that present percentage methylation values.

T-lymphocyte samples were collected at activation, resting phase, and restimulation time points. Total RNA extraction was carried out by using TRIzol (Invitrogen) following reverse transcriptions with M-MLV reverse transcriptase (Invitrogen). Quantitative RT-PCR (qRT-PCR) was performed with iQ SYBR green supermix in a CFX real-time PCR detection system (Biorad). Primer pairs used to detect gene expression are shown in Supplementary Table S3. Mature miR-21 expression was assessed by real-time PCR analysis using Taqman microRNA assays (Life Technologies). Reverse transcription was performed on RNA using the TaqMan MicroRNA Reverse Transcription Kit (cat no 4366596). Quantitative RT-PCR was performed with 2× Taqman Fast Universal PCR Master Mix (cat no 4366072) according to the manufacturer's protocol, with specific TaqMan primers. Gene and (MIR21) expression data were normalized with levels of the housekeeping gene H3 and RNU6B, respectively, and represented according to this formula 2^{ΔCt (CtH3 or RNU6-Ctgene)}, where Ct corresponds to cycle number.

T lymphocytes were prestained with the Zombi NIR Fixable viability kit (Biolegend) as a live/dead marker and pretreated with Beriglobin before staining. Surface staining was performed with the following mAbs purchased from Biolegend: CD8-BV510 (5K1), CD96-PE (NK92.39), CCR5-PerCPC5.5 (HEK/1/85a), CXCR5-BV421 (J252D4), mouse IgG1-PE and IgG1-BV421 (MOPC-21), and rat IgG2a-PerCPC5.5 (RTK2758) as an isotype-matched negative control. The True-Nuclear Transcription Factor Buffer Set (Biolegend) was used for intracellular staining of TCF1 using anti–TCF1-AF647 (7F11A10) and mouse IgG1-AF647 (MOPC-21), both purchased from Biolegend. Cell acquisition was carried out with FACSCanto II and FlowJo (Treestar) software was used for data analysis.

CCL5 protein expression was measured from the supernatants of CD8⁺ T cell cultures using the RANTES (CCL5) human SimpleStep ELISA Kit (Abcam, ref AB174446).

Human CD8⁺ T lymphocytes were immunomagnetically purified from peripheral blood of healthy donors as described in Material and Methods. Subsequently, naïve and antigen-experienced T-cell subsets were FACS-sorted as previously described (29). Briefly, CD8⁺ T cells were pretreated with Beriglobin before staining with the following antibodies for cell surface markers: CD8-APC (RPA-T8), CD27-FITC (o323), CD62-L PE (DREG-56), and CD45RA-PerCPC5.5 (HI100) and sorted in a FACSAria (BD). Each purified subpopulation was stained with either Cell Trace Violet (Invitrogen) or CFSE (BD) or left unstained and following prelabeling populations were remixed. Different combinations of these stainings were performed to have cultures with each population stained with both cell dyes, rendering similar results. The resulting remixed CD8⁺ T cells were activated with plate coated anti-CD3 mAb and costimulatory mAb (Fig. 1) in 96-well plates in identical density (cells/cm²) for 5 days. An identical experimental procedure described in the Materials and Methods section and depicted in Fig. 1 were carried out.

CD14 cells from PBLs of healthy donors were purified with CD14 MicroBeads (Miltenyi) and cultured in AIM V Medium (Gibco) with 1,000 u/mL of rhu-IL4 (R&D) and huGM-CSF (Bayer). After 6 days, cells were collected, washed, and seeded in 8-μm pore polycarbonate Transwell (Costar), at 10⁵ cells per well.

The percent migration of dendritic cells (DC) to CCR5 was assessed toward recombinant human CCL5 (Preprotech; 300 ng/mL) and toward supernatants derived from the 36-hour–restimulated CD8⁺ T cells that had been previously activated in presence of either urelumab or hIgG (from Fig. 4B). Following a 12-hour culture in the Transwell setting, migrated DC of the bottom chambers were recovered and counted. The percentage of migration computes is referred to input DC. To block CCL5-driven cell migration, DCs were incubated for 1 hour with 0.8 μg/mL of human CCL5-blocking antibody (R&D) before the migration assay. All the experiments were performed in triplicate.

Prism software (GraphPad Software, Inc.) was used to analyze statistical differences of absolute methylation level and mRNA and protein expression of target genes by applying the paired Student t test, Mann–Whitney U test or the Wilcoxon paired test. Values of *, P < 0.05; **, P < 0.01; ***, P < 0.001 were considered significant.

Our initial hypothesis was that CD137 signaling in primed T lymphocytes would result in chromatin remodeling through sequence-specific genomic DNA methylation. To study modifications in DNA methylation at the genomic level, we used culture plate-bound CD3ϵ mAb to mimic TCR priming together with urelumab (23) or 6B4 (30) as CD137 agonists (Fig. 1A). CD8⁺ T cells were immunomagnetically isolated from the blood of healthy donor volunteers and seeded onto plates coated with anti-CD3ϵ and anti-CD137. At the indicated points of time, samples were retrieved to isolate genomic DNA, mRNA, or to measure protein expression in the cells and supernatants. In the process of stimulation in culture, samples were transferred on day +5 to antibody-free plates to resemble a return-back to a resting status in the presence of the homeostatic cytokine IL7. On day +10 of culture, T cells were stimulated again with solid-phase bound anti-CD3ϵ as a surrogate of a subsequent antigen encounter. Samples were collected at a series of time points following this secondary stimulation. This experimental approach sought to observe changes caused by costimulation that would become imprinted into the chromatin of CD137-costimulated CD8⁺ T cells. In order to assess nuclear gene DNA methylation in a genome-wide fashion, the 450K DNA-methylation array was used in a series of CD8⁺ T cells from three independent individuals costimulated by urelumab (or control IgG4 mAb) or by 6B4 acting on CD8⁺ T cells from an additional unrelated donor. Differential methylation analyses were performed at day +10, representing T lymphocytes in which the CD137-costimulation epigenetic modifications would be already established, potentially leaving the cells poised to long-term respond in a distinct manner.

At day +10 of CD8⁺ T-cell culture, 1,028 differentially methylated CpGs corresponding to 907 genes (Fig. 1B) were found to be modified by CD137 costimulation in their methylation status at specified CpGs, as determined by an unbiased bioinformatic analyses at consistent loci either after urelumab or 6B4 activation (Supplementary Tables S1 and S2, respectively). As a general tendency, most of the changes in CpGs whose methylation status was modified in response to anti-CD137 mAb consisted of demethylation (87% CpGs for urelumab; 69% CpGs for 6B4) rather than hypermethylation. From the list of 52 genes that were differentially methylated by CD137 costimulation observed both with urelumab and 6B4 (Fig. 1B), several genes (in bold) attracted our attention for their well-described involvement in immune cell performance. This hand-picked list was supported by PubMed publication searches and Gene Ontology (GO) analyses, that showed an enrichment of immune response functions (GO:0006955, FDR = 0.0129). This group includes genes related to T-cell costimulation/coinhibition, inflammation and inflammatory chemotaxis, immune cell transcription factors, and the microRNA miR-21.

To confirm the reproducibility of these findings, we analyzed the identified methylated or demethylated immune-relevant sequences by pyrosequencing in a series of unrelated individuals (n = 19). As we hypothesized, these genes were modified in a consistent fashion with the findings reported by the DNA-methylation microarrays, following both urelumab- or 6B4-elicited costimulation (Fig. 2A and B).

DNA-methylation patterns are known to be different in naïve versus antigen-experienced CD8⁺ T lymphocytes from human peripheral blood. DNA-methylation status of TCF7 and CCL5 loci, which were identified in our genome-wide screenings (Fig. 1) as differentially methylated following CD137 ligation, is indeed known to be differentially methylated in naïve versus memory CD8⁺ T cells (4, 31). In fact, when we FACS-sorted naïve versus memory and effector CD8⁺ T cells from peripheral blood (Supplementary Fig. S1A), we saw that our most relevant identified genes followed similar DNA methylation pattern in peripheral blood naïve versus memory CD8⁺ cells (CD96, HHLA2, MIR21, and CXCR5; Supplementary Fig. S1B).

To address whether the frequencies of originally naïve versus memory cells changed in the resulting cultures, we sorted these subsets by FACS to high purity, labeled them with distinct fluorescent dyes, remixed them, and followed their frequency during culture. We did not observe any significant changes in the composition of the resulting cultures if costimulated with CD137 or control antibody (Supplementary Fig. S1C). These results suggest that changes in their relative abundance do not explain changes in DNA methylation.

We decided to evaluate whether DNA-methylation changes observed in our study altered mRNA transcription. To study this correlation, a series of quantitative RT-PCR analyses were performed on the selected immune-relevant genes. For consistency in our analysis, experiments with an unrelated series of donors were performed under costimulation with urelumab (Fig. 3A) or with 6B4 (Fig. 3B). CD8⁺ T-cell samples from each individual were paired at each time point and a sufficient number of cases were studied to mitigate the intrinsic genetic and epigenetic variability of human populations. Experiments studying protein expression were done when feasible (Fig. 4).

CD96 mRNA was consistently downregulated at all time points, including restimulation (Fig. 3). This surface molecule is considered an important negative regulator of T and NK activation following engagement to its ligand CD155 (32), whose blockade enhances T and NK activation. Hence, we postulated that reduced expression could result in enhanced activation and effector performance. We compared the intensity of surface CD96 expression by FACS, and our results were in agreement with this conclusion in the majority of individual cases (Fig. 4A).

HHLA2 (B7-H5) is another emerging surface checkpoint receptor of the B7 family for T-cell activation, but in this case, mRNA expression is clearly increased upon CD137 costimulation (Fig. 3A and B). The functional role of this surface molecule and its putative ligands remains obscure and controversial (33, 34). The absence of reliable detecting antibody reagents precluded analysis of the protein in this case. However, the data suggest that this moiety is relevant for human T-cell biology even if not conserved in mice (33).

We found that the methylation status of CCR5 and CXCR5 was modified (Fig. 2), and such chemokine receptors were involved in shaping CD8⁺ T-cell attraction to activated myeloid cells and fellow lymphocytes under inflammatory conditions. Reduction of CXCR5 mRNA and protein should mitigate the tendency of such T cells to migrate to germinal centers, potentially leaving them free to accomplish other tasks (35, 36), whereas enhanced expression of CCR5 might make them more likely to meet and interact with myeloid cells such as macrophages and dendritic cells. In this context, the function of CCR5 would also be enhanced by the concomitant upregulation observed with CCL5 (Figs. 3 and 4B), which is one of its ligands. Indeed, augmented CCL5 accumulation in the tissue culture supernatant was readily observed upon restimulation of previously CD137-costimulated CD8⁺ T cells (Fig. 4B). When these culture supernatants were tested for their ability to attract monocyte-derived DCs, we found that CD137-costimulated supernatants were more powerful chemoattractants than their respective hIgG4 isotype control supernatants. This effect could be neutralized by a CCL5-blocking mAb (Fig. 4C).

The transcription factor TCF1 (encoded by TCF7) was drastically reduced (by as much as 85%) at the mRNA and protein levels by CD137 costimulation (Figs. 3 and 4A). This opens up an interesting area of research since the TCF1 transcription factor, in conjunction with BCL6, seems to be critical in the control of memory and stemness of T cells (37). Indeed, TCF1 expression in T cells is associated with the exhausted phenotype (35, 37). The methylation status of the GFI-1 transcriptional repressor is also regulated by CD137 costimulation (Fig. 2A and B), and this factor could mediate broader transcriptional effects on its target genes. Although its role in CD8⁺ T cells remains poorly understood, key effects in Th17 and Th2 biology have been reported (38).

Controlling microRNA by gene methylation also constitutes a mechanism to spread gene regulation to their targeted sequences. In our hands, MIR21 expression was decreased by DNA demethylation at its locus (Fig. 5B and C).

This study shows that T-cell costimulation via CD137 (4-1BB) results in epigenetic changes in the chromatin that affect key genes playing a role in the ensuing immune response. These findings may find application in understanding mechanisms of action of different immunotherapies and might be useful in the development of pharmacodynamics biomarkers. Our approach is based on simplistic culture systems in which primary human CD8⁺ T lymphocytes are costimulated in the presence of agonist antibodies and restimulated again to mimic antigen reexposure following a resting phase. Genome-wide DNA-methylation analysis using microarrays was chosen to explore this subject. We focused on those genes whose attributed functions are predicted to affect T-cell immunity and therefore CD137-based immunotherapy. Genes whose DNA methylation is influenced by CD137 are broader than those on which we have focused; however, our repeated findings conclusively indicated consistent regulation dependence on the CD137 costimulation pathway.

Several studies integrating DNA-methylation profiles and gene expression data have shown that methylation at different genomic regions (promoters, gene bodies, or intergenic regions) is related to gene expression levels (28, 39). Although a correlation between methylation and gene expression has been observed in multiple studies, this scenario can be different depending on the methylation levels and the genomic regions. In this regard, correlation between gene body methylation and expression was observed to be not only positive (40) or negative (41) but also dependent on cell type (42). Hence, we have explored changes in mRNA and, when feasible, protein expression.

CD96 is one of the consistently regulated genes that caught our attention for its immune relevance as a checkpoint on T and NK cell activation. Even if the functional role of the observed changes remains to be determined, it is likely to be involved in rendering T cells prone to activation. This is consistent with the observations that experimental lung metastases that result from the intravenous injection of tumor cells are decreased upon the attenuation of the CD96 coinhibitory pathway (43). As a result, CD96 is now considered a very attractive immune checkpoint for immunotherapy (32).

Key to the performance of T cells is their ability to migrate up chemokine gradients or to produce chemokines that attract other leukocytes, encounters with whom could be functionally important. Methylation changes to the CCR5, CXCR5, and CCL5 genes are very provocative and likely to affect cytotoxic T-lymphocyte (CTL) migration and homing, as well as favoring certain cell-to-cell interactions amidst the tumor microenvironment or lymphoid tissue. Indeed, we observed that CCL5 produced by CD137-costimulated CD8⁺ T cells attracted monocyte-derived DCs.

In addition to chemotaxis, other important functions may have been overlooked in our analyses due to our imperfect knowledge of gene functions on the immune response. For instance, regulation of the inflammasome gene AIM2 (Fig. 2A and B), which was reported in the AAI 2015 meeting (44) as a key regulator for CD4⁺ T-cell memory that is epigenetically regulated by DNA methylation.

For gene-methylation regulation to be more effective, the transcriptional control of other mechanisms that influence gene expression would extend the effect to a broader list of downstream genes. Hence, control over transcription factors or microRNAs by epigenetic mechanisms is considered to be highly influential for gene-expression reprogramming. For this reason, we focused on transcription factors and noncoding RNAs that could influence long-term T-cell behavior. We found TCF1 and miR-21 to stand out. Their influence on secondary target genes remains to be seen, but our findings open a novel layer of gene-expression regulation and a complex field to be explored.

TCF7 during chronic viral infections is known to be epigenetically regulated, resulting in changes in chromatin accessibility (45). TCF1 and CXCR5 are coordinately regulated in CD8⁺ T cells, which results in their differentiation into a subset similar to follicular T helper cells (T_FH like) in mice (35). Our interpretation is that CD137 costimulation leading to DNA methylation changes would downregulate this T_FH-like differentiation pattern. We have previously reported that human T_FH cells are one of the only human lymphocyte subsets expressing baseline CD137 in healthy conditions (14). Modulation of CD137-dependent epigenetic changes in the TCF7 locus, which encodes the TCF1 transcription factor, is of considerable interest, because this pathway orchestrates the stemness of T cells (46), as well as their exhausted phenotype (35, 37, 46). Along this line of reasoning, TCF1 reduction could mitigate undesired functional phenotypes of CD8⁺ T cells, thereby potentially enhancing tumor immunity.

Decreases in miR-21 could result in regulation of a number of secondary genes. MiR-21 is considered an onco-miR, because it promotes tumor progression and turns on immunosuppressive mechanisms in colorectal cancer (46). However, little is known about how miR-21 would affect primary CD8⁺ T-cell differentiation and biology, although it has been reported to affect T-cell activation (47, 48), apoptosis (49), and differentiation. Other microRNAs have been recognized as key factors in the regulation of CTL physiology (50).

The observed epigenetic changes do not only take place during activation, but persist, pointing to their role in shaping ultimate responses upon antigen restimulation. Modifying DNA methylation involves DNA replication and, therefore, we are likely underestimating epigenetic remodeling, because only part of the T-cell cultures may have divided a sufficient number of times under the influence of CD3 stimulation and CD137 costimulation. Because of methodological constraints, we have not subdivided and isolated naïve and central memory peripheral blood CD8⁺ T cells, and the effects could vary depending on the relative abundance of each lymphocyte subset in the peripheral blood corresponding to each individual donor. To address this point, we set up costimulation cocultures with FACS-sorted prelabeled naïve, memory, and effector cells. No significant differences in proportions between CD137-costimulated and control CD8⁺ T lymphocytes at the end of the 10-day cultures were found. However, changes in DNA methylation induced by CD137 costimulation coincided with those that we reported comparing antigen-experienced with antigen-naïve CD8⁺ T cells. These findings reinforce the idea that CD137 costimulation sets a course toward memory differentiation, whose mechanistic underpinnings are the subject of ongoing research.

Chromatin remodeling by epigenetic mechanisms of CD8⁺ T cells modifies the expression of key genes upon T-cell activation and differentiation, for instance, during viral infection (51). CD28 costimulation can also change the methylation of the IL2 promoter (52). Here, we connect the activity of a TNFR-family costimulatory member to DNA methylation, which is a form of long-term control of gene expression. The particular molecular pathways under CD137 control that lead to sequence-specific methylation/demethylation of the target genes remain to be uncovered. Another implication of these data is that drugs modifying DNA methylation, such as DNMT inhibitors, would potentially modify CD137-based immunotherapy. In fact, modification of DNA-methylation affects genes relevant for Th1 biology in vivo (53).

Application of our results and conclusions to CD137-targeted immunotherapy could still be premature. Our experiments were performed on resting peripheral blood CD8⁺ T cells, whereas the main target of anti-CD137 in tumor-bearing hosts is proposed to be dysfunctional (exhausted) intratumoral T lymphocytes.

All in all, our study shows a new functional outcome following CD137 costimulation. This costimulatory function is being exploited for cancer immunotherapy with agonist antibodies (14) or chimeric antigen receptors encompassing the CD137 cytoplasmic tail (54). Our results on epigenetic reprogramming of cytotoxic T lymphocytes by CD137 costimulation imply far-reaching consequences to their functionality. Epigenetic reprogramming is unique in the sense that would leave the T cells poised to respond differentially when reencountering antigen in the future.

I. Melero reports receiving a commercial research grant from Bristol-Myers Squibb, Roche, and Alligator and is a consultant/advisory board member for Bristol-Myers Squibb, Roche, Bayer, Tusk, AstraZeneca, Lilly, Alligator, and Bioncotech. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.A. Aznar, S. Labiano, A. Diaz-Lagares, J. Sandoval, I. Melero

Development of methodology: M.A. Aznar, S. Labiano, A. Diaz-Lagares, C. Molina, M. Esteller, I. Melero

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.A. Aznar, S. Labiano, A. Diaz-Lagares, C. Molina, A. Azpilikueta, I. Etxeberria, A.R. Sanchez-Paulete, M. Esteller, J. Sandoval

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.A. Aznar, S. Labiano, A. Diaz-Lagares, C. Molina, I. Etxeberria, M. Esteller, J. Sandoval, I. Melero

Writing, review, and/or revision of the manuscript: M.A. Aznar, S. Labiano, A. Diaz-Lagares, I. Etxeberria, M. Esteller, J. Sandoval, I. Melero

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Garasa, A. Azpilikueta, A.R. Sanchez-Paulete, A.J. Korman

Study supervision: J. Sandoval, I. Melero

I. Melero is supported by grants from MINECO (SAF2011-22831 and SAF2014-52361-R), Departamento de Salud del Gobierno de Navarra, Redes temáticas de investigación cooperativa RETICC, European Commission VII Framaework and Horizon 2020 programs (IACT and PROCROP), Fundación de la Asociación Española Contra el Cáncer (AECC), Fundación BBVA and Fundación Caja Navarra. S. Labiano is the recipient of predoctoral scholarship from MINECO. A. Diaz-Lagares is funded by Río Hortega Grant CM14/00067 from ISCIII. J. Sandoval is funded by “Miguel Servet Program” from the FEDER, FSE, and ISCIII (CP13/00055) and a contribution from “Corte de Honor and FMV de 2011.”

We are grateful to Drs. Alvaro Teijeira, Ana Rozaut, Jose Luis Perez-Gracia, Miguel Fernandez de Sanmamed, and Juan José Lasarte for helpful scientific discussions. We also appreciate technical support by Diana Garcia, Carles Arribas, and Elixabet Bolaños. Navarrabiomed tissue bank nurses and medical staff are also gratefully acknowledged.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.