key: cord-103517-1itisiup
authors: Kwesi-Maliepaard, Eliza Mari; Aslam, Muhammad Assad; Alemdehy, Mir Farshid; van den Brand, Teun; McLean, Chelsea; Vlaming, Hanneke; van Welsem, Tibor; Korthout, Tessy; Lancini, Cesare; Hendriks, Sjoerd; Ahrends, Tomasz; van Dinther, Dieke; den Haan, Joke M.M.; Borst, Jannie; de Wit, Elzo; van Leeuwen, Fred; Jacobs, Heinz
title: The histone methyltransferase DOT1L prevents antigen-independent differentiation and safeguards epigenetic identity of CD8+ T cells
date: 2019-11-18
journal: bioRxiv
DOI: 10.1101/826255
sha: 
doc_id: 103517
cord_uid: 1itisiup

Cytotoxic T-cell differentiation is guided by epigenome adaptations but how epigenetic mechanisms control lymphocyte development has not been well defined. Here we show that the histone methyltransferase DOT1L, which marks the nucleosome core on active genes, safeguards normal differentiation of CD8+ T cells. T-cell specific ablation of Dot1L resulted in loss of naïve CD8+ T cells and premature differentiation towards a memory-like state, independent of antigen exposure and in a cell-intrinsic manner. Without DOT1L, the memory-like CD8+ cells fail to acquire full effector functions in vitro and in vivo. Mechanistically, DOT1L controlled T-cell differentiation and function by ensuring normal T-cell receptor density and signaling, and by maintaining epigenetic identity, in part by indirectly supporting the repression of developmentally-regulated genes. Through our study DOT1L is emerging as a central player in physiology of CD8+ T cells, acting as a barrier to prevent premature differentiation and supporting the licensing of the full effector potential of cytotoxic T cells.

Lymphocyte development and differentiation is tightly regulated and provides the basis for a functional adaptive immune system. Development of mature T cells initiates in the thymus with progenitor T cells that have to pass two key checkpoints: T cell receptor (TCR) β selection and positive selection, both of which are controlled by intricate signaling pathways involving the pre-TCR/CD3 and αβTCR/CD3 complexes, respectively 1 . Upon positive selection, mature thymocytes are licensed to emigrate and populate peripheral lymphatic organs as naïve T cells (T N ). Further differentiation of naïve T cells into effector or memory T cells normally depends on TCR-mediated antigen recognition and stimulation.

However, it has become evident that a substantial fraction of mature CD8 + T cells acquires memory-like features independent of exposure to foreign antigen. The origin and functionality of these unconventional memory cells in mice and humans, also referred to as innate or virtual memory cells, are only just being uncovered [2] [3] [4] .

The dynamic changes during development and differentiation of CD8 + T cells are governed by transcriptional and epigenetic changes, including histone modifications that are controlled by chromatin modifiers. Well-established histone marks are mono-and tri-methylation of histone H3K4 at enhancers (H3K4me1) and promoters (H3K4me3), H3K27me3 at repressed promoters, and H3K9me2/3 in heterochromatin [5] [6] [7] [8] [9] [10] . Although epigenetic 'programming' is known to play a key role in T cell development and differentiation, the causal role of epigenetic modulators in T cell differentiation is still poorly understood, especially for chromatin modifiers associated with active chromatin 5 .

One of the histone modifications positively associated with gene activity is mono-, di-and trimethylation of histone H3K79 mediated by DOT1L. This evolutionarily conserved histone methyltransferase methylates H3K79 in transcribed promoter-proximal regions of active genes 11, 12 .

Although the association with gene activity is strong, how H3K79 methylation affects transcription is still unclear and repressive functions have also been proposed [13] [14] [15] [16] [17] [18] . DOT1L has been linked to several critical cellular functions, including embryonic development, DNA damage response, and meiotic checkpoint control 19, 20 and DOT1L has also been shown to function as a barrier for cellular reprogramming in generating induced pluripotent stem cells 21 . DOT1L gained wide attention as a specific drug target in the treatment of MLL-rearranged leukemia, where MLL fusion proteins aberrantly recruit DOT1L to MLL target genes leading to their enhanced expression [22] [23] [24] . A similar dependency on DOT1L activity and sensitivity to DOT1L inhibitors was recently observed in thymic lymphoma 25 . Interestingly, inhibition of DOT1L activity in human T cells attenuates graft-versus-host disease in adoptive cell transfer models 26 and it regulates CD4 + T cell differentiation 27 .

Given the emerging role of DOT1L in epigenetic reprogramming and T-cell malignancies, we investigated the role of DOT1L in normal T cell physiology using a mouse model in which Dot1L was selectively deleted in the T cell lineage. Our results suggest a model in which DOT1L plays a central role in CD8 + T cell differentiation, acting as a barrier to prevent premature antigen-independent differentiation, and promoting the licensing of the full effector potential by maintaining epigenetic integrity.

Given the essential role of DOT1L in embryonic development 28 , we determined the role of DOT1L in T cell development and differentiation by employing a conditional knock-out mouse model in which Dot1L is deleted in the T-cell lineage by combining floxed Dot1L with a Cre-recombinase under the control of the Lck promoter, which leads to deletion of exon 2 of Dot1L during early thymocyte development (Sup Fig. 1a) 25 . The observed global loss of H3K79me2 in Lck-Cre +/-;Dot1L fl/fl mice, as confirmed by immunohistochemistry on fixed thymus tissue (Sup Fig. 1b) , agreed with the notion that DOT1L is the sole methyltransferase for H3K79 11, 25, 28, 29 .

To validate the efficacy of Dot1L deletion at the single-cell level, we developed an intracellular staining protocol for H3K79me2. Histone dilution by replication-dependent and -independent means has been suggested to be the main mechanisms of losing methylated H3K79 15, [30] [31] [32] . Flow-cytometric analyses of thymocyte subsets from Lck-Cre +/-;Dot1L fl/fl mice (hereafter, KO) revealed that double-negative (DN, CD4 -CD8 -) thymocytes started losing H3K79me2. From the subsequent immature single positive state (ISP) onward all the thymocytes had lost DOT1L mediated H3K79me (Sup Fig. 1c ). This confirmed that upon early deletion of Dot1L, successive rounds of replication in the thymus allowed for loss of methylated H3K79.

No changes in H3K79 methylation levels were found in T-lineage cells of Lck-Cre +/-;Dot1L wt/wt control mice (hereafter, WT). Ablation of Dot1L did not significantly affect the overall thymic cellularity, however it led to a reduction in the number of mature single positive (SP) CD4 + and CD8 + thymocytes (both 2.5-fold) while the ISP and CD4 + CD8 + double positive (DP) subsets were not significantly affected, suggesting a role of DOT1L in controlling intrathymic T cell selection (Sup Fig. 1d-e) .

In the spleen, overall cellularity was also not affected but within the T cell compartment, CD4 + T cells were drastically reduced (3.2-fold), whereas CD8 + T cells were increased (1.7-fold) in number (Sup Fig.   1e -f). However, while flow cytometry of H3K79me2-stained splenic T cells confirmed the lack of DOT1L activity in CD8 + T cells and CD44 -CD62L + CD4 + T cells, CD44 + CD62L -CD4 + cells showed a partial loss of H3K79me2 and CD4 + CD25 + regulatory T cells (Treg) remained H3K79me2 positive (Sup Fig. 1g ). Since earlier in development, CD4-expressing cells in the thymus were mostly H3K79me2 negative, this suggests that a strong selection occurred for the maintenance of DOT1L for the development of Tregs in this mouse model. Indeed, partial deletion of Dot1L in CD4 + cells was confirmed by PCR analysis (Sup Fig.   1h ). Here, we focused our study on defining the role of DOT1L in the cytotoxic T cell compartment in which efficient deletion of Dot1L and loss of H3K79me2 was found in both the thymus and the periphery.

CD8 + T cell differentiation was strongly affected by the absence of DOT1L. Analysis of CD8 + T cell subsets in the spleen revealed that Dot1L-KO mice showed a severe loss of naïve (CD44 -CD62L + ) CD8 + T (T N ) cells which was linked to a massive gain of the CD44 + CD62L + phenotype, a feature of central memory T cells (T CM ; Fig. 1A -B). Lck-Cre +/-;Dot1L fl/wt heterozygous knock-out mice (Het) did not show any phenotypic differences of CD8 + T cells (Fig. 1A-B) . The lack of haplo-insufficiency was further confirmed by principal component analysis of RNA-Seq data indicating that WT and Het CD8 + T cells clustered together but were separated from the KO CD8 + T cells, excluding gene-dosage effects (Sup Fig. 1i ). Therefore, we restricted our further studies to the comparison of the KO and WT mice. The strong shift towards a CD8 + memory-phenotype in Dot1L KO was unexpected because Dot1L-KO mice were housed under the same conditions as their WT controls and the mice had not been specifically immunologically challenged.

To unravel the molecular identity of the CD44 + CD62L + Dot1L-KO cells in more detail we performed RNA-Seq analysis on sorted CD44 -CD62L + (T N ) and CD44 + CD62L + (T CM ) CD8 + T cell subsets from WT and KO mice. Based on differential gene expression between T N and T CM CD8 + cells from WT mice, naïve and memory gene signatures were defined. Interestingly, overlay of these signatures on WT CD44 -CD62L + (T N ) and KO CD44 + CD62L + (T CM ) cells showed that differential expression between T N and T CM cells was mostly conserved when Dot1L was ablated. (Fig. 1c) , although there was misregulation of other genes as well (see below). These data suggest that in the absence of DOT1L, CD8 + T cells cannot retain the naïve identity but rather acquire, prematurely and in the absence of any overt immunological challenge, a transcriptome of memory-like CD8 + T cells.

Although WT and KO mice were exposed to the same environment it cannot be excluded that KO mice responded differentially to antigens in the environment. If this is the case one expects skewing in the clonality of the TCRβ gene usage. In order to investigate this possibility, we examined the TCRβ repertoire. Tcrb-sequencing revealed no difference in productive clonality scores between WT and KO CD8 + T cells (Sup Fig. 2a) . Also, CDR3 length as well as Tcrb-V and Tcrb-J gene usage were unaffected (Sup Fig. 2b-d) . These data, together with the nearly complete loss of naïve CD8 + T cells, argued against any antigen-mediated bias in the selection for CD8 + T cells and indicated that CD44 + CD62L + memory-like CD8 + T cells in KO mice were polyclonal and arose by antigen-independent differentiation of T N cells.

Antigen-independently differentiated memory-like CD8 + T cells have already been described in the literature and their origins and functions are subject of ongoing studies [2] [3] [4] 33 . Depending on their origin and cytokine dependency they are referred to as 'virtual' or 'innate' memory cells. Virtual memory CD8 + T cells have been suggested to arise in the periphery from cells that are CD5 high , related to high TCR affinity, and require IL-15 4, 34 . In contrast, innate memory CD8 + T cells develop in the thymus and their generation and survival are generally considered to be dependent on IL-4 signaling 3 . We here collectively refer to them as Antigen-Independent Memory-like CD8 + T cells (T AIM ). A common feature of these T AIM cells is reduced expression of CD49d 35 , a marker that is normally upregulated after antigen exposure. In addition, they express high levels of T-bet and EOMES, encoding two memory/effector transcription factors 36, 37 . In both Dot1L-KO and WT, the majority of the CD44 + CD62L + (T CM ) cells were CD49d negative. As a control, CD44 + CD62Leffector T cells (T EFF ) from WT mice challenged with Listeria monocytogenes were mostly CD49d positive (Fig. 1d) . This further indicates that the generation of CD44 + CD62L + memory-like T cells in KO mice is independent of antigen exposure. Of note, the percentage of CD49d-negative CD44 + CD62L + cells that we observed in WT mice corresponds to the percentage of T AIM cells reported in WT B6 mice 38 . Regarding the expression of T-bet and EOMES, most of the KO CD8 + T cells co-expressed T-bet and EOMES. Furthermore, EOMES was expressed at a higher level in KO CD8 + T cells as compared to their WT counterpart ( Fig. 1e-f ). Together these characteristics are all in agreement with antigen-independent differentiation of naïve CD8 + -T cells in the absence of DOT1L.

To determine whether peripheral T AIM cells observed in the Dot1L-KO setting originate intrathymically, as reported previously for IL-4 dependent innate memory T cells 3 , we compared RNA-Seq data from SP CD8 + thymocytes from KO and WT mice. Analyzing the relative distribution of memory and naïve signature genes revealed that memory genes were among the genes upregulated in KO SP CD8 + thymocytes (Fig. 2a) . Importantly, like in peripheral CD8 + T cells the expression of T-bet and Eomes was highly upregulated in KO SP CD8 + thymocytes. This transcriptional upregulation was corroborated by flow cytometric analysis of protein expression. Intracellular staining for the transcription factors showed that a small but substantial subset of the SP CD8 + KO thymocytes expressed T-bet and EOMES at the protein level (Fig. 2b-c) . Together with the unperturbed TCRβ repertoire, this further supports the notion that differentiation of Dot1L-KO CD8 + T cells towards memory-like cells initiates intrathymically in an antigen-independent manner. Innate memory cells have been suggested to arise in the thymus in response to an increase in IL-4 producing PLZF high invariant NKT (iNKT) cells or γδ T cells 3 . However, iNKT cells were nearly absent in the thymus of Dot1L-KO mice ( Fig. 2d-e) . Furthermore, the number of γδ T cells did not differ significantly between WT and KO mice (Fig. 2f) . Previous studies on innate memory T cells from different mouse models demonstrated that introduction of a transgenic TCR, inhibiting the generation of iNKT and γδ T cells, prevents the development of innate memory cells 39, 40 . However, introduction of the transgenic OT-I TCR, a condition under which the number of iNKT and γδ T cells is strongly reduced 41, 42 , did not affect the memory-phenotype of Dot1L-KO CD8 + T cells (Fig. 2g) . Together, these findings indicate that the intrathymic differentiation of T AIM CD8 + cells in the absence of DOT1L did not depend on an excess of IL4-producing cells in the thymic microenvironment as reported for innate memory T cells. Rather the formation of T AIM cells in Dot1L-KO mice likely relates to a cell-intrinsic mechanism.

One of the cell-intrinsic mechanisms reported to be involved in the formation of T AIM cells is aberrant TCR signaling 2, 34, 43, 44 . Furthermore, treatment of human T cells with DOT1L inhibitor impaired TCR sensitivity and attenuated low avidity T cell responses 26 . This led us to investigate the expression of genes encoding TCR signaling components in the absence of DOT1L. RNA-Seq analyses confirmed that many TCR signaling genes were differentially expressed between WT and KO SP CD8 + thymocytes (Fig.   3a ). Importantly, CD3ζ (CD247), a critical rate limiting factor in controlling the transport of fully assembled TCR/CD3 complexes to the cell surface, was downregulated in Dot1L-KO T cells 1, [45] [46] [47] [48] . In addition, other components of the TCR/CD3 complex like CD3e and its associated co-receptor CD8a/b were also downregulated in KO T cells. H3K79me2 ChIP-Seq showed that these genes had H3K79me2 in WT mice and might therefore be directly regulated by DOT1L (Sup Fig. 3a-c) . As a consequence, one expects TCR/CD3 and CD8αβ to be reduced at the cell surface of T AIM cells, which we confirmed by flow cytometry (Fig. 3b ). In addition to the CD3/TCR complex, we observed downregulation of Itk, a key TCR signaling molecule reported to be involved in innate memory CD8 + T cell formation 49 . To exclude that the impaired TCR signaling in KO T cells could be compensated for by the selection of thymocytes expressing TCRs with altered affinity, we kept the TCR affinity identical by crossing the OT-I TCR transgene into our system. Thymocytes expressing OT-I are positively selected in the presence of MHC class I (H-2K b ), mainly generating CD8 + SP T cells expressing the exogenous OT-I TCR, with concomitant reduction of the CD4 + lineage 50 . If DOT1Ldeficiency impairs TCR surface density and signaling, positive selection of conventional OT-I CD8 thymocytes is expected to be compromised. In Dot1L-KO mice expressing OT-I, the number of SP CD8 + thymocytes was decreased (3.8-fold) compared to WT mice expressing OT-I (Fig. 3c ). This revealed that like with endogenous TCR, early intrathymic ablation of Dot1L in the T cell lineage prohibits positive section of conventional OT-I CD8 + T cells, but yet supports the generation and selection of T AIM cells (Fig. 2g ) of which the vast majority expressed both exogenous TCR chains (TCRαV2 and TCRβV5) (Fig. 3d) . Consistent with the low surface expression of TCR/CD3 and CD8 in the KO condition, the surface expression of OT-I TCR was also lower, as determined by SIINFEKL/H-2K b tetramer staining (Sup Fig. 3d ). This was further validated by staining with antibodies for the transgenic TCR chains TCRαV2 and TCRβV5 which indicated two-fold reduction in KO ( Fig. 3b and d) .

The TCRαV2 element of the OT-I TCR was under the control of an exogenous promoter, suggesting that the reduced surface expression of OT-I TCR on KO T cells did not relate to transcriptional silencing of native TCR gene promoters. Instead, the observed lower TCR surface levels in KO T cells likely relate to the reduced CD3ζ expression 48 . In conclusion, DOT1L regulates the levels of TCR complex and signaling molecules independent of the selected TCR. In the absence of these regulatory mechanisms, the identity of naïve CD8 + T cells cannot be maintained, contributing to premature T cell differentiation.

A hallmark of memory T cells is that they proliferate faster and produce IFNγ rapidly upon stimulation compared to naïve T cells 51 . Likewise, innate and virtual memory T cells rapidly produce IFNγ upon TCR stimulation 38, 44 . To determine the proliferative potential of Dot1L-KO T AIM cells, we activated B-cell depleted, CFSE-labeled splenocytes in vitro with anti-CD3 and anti-CD28 antibodies. Proliferation was determined by CFSE dilution. Dot1L-KO CD8 + T cells proliferated faster than WT cells (Fig. 4a ). In addition, they rapidly became activated CD44 + CD62L -(Sup Fig. 4a ). However, in contrast to WT, KO CD8 + T cells failed to produce IFNγ ( Fig. 4b and Sup Fig. 4b) , and expressed lower levels of CD69, a marker of activation (Sup Fig. 4c ). This indicates a functional impairment of Dot1L-KO T cells. To test their intrinsic competence to produce IFNγ, Dot1L-KO T cells were stimulated in a TCR-independent manner with PMA and Ionomycin. Intracellular staining for IFNγ revealed that the percentage of stimulated CD8 + T cells producing IFNγ was 1.9-fold higher in KO as compared to WT (Sup Fig. 4d ). This suggests that Dot1L-KO CD8 + T cells only partially respond to TCR stimulation, but that they have the intrinsic capacity to produce IFNγ. The responders in the WT population likely include the naturally existing virtual memory population ( Fig. 1 ). Taken together, Dot1L KO T AIM cells displayed some memory-associated features but simultaneously are functionally impaired.

To determine the in vivo immune responsiveness of Dot1L-KO T cells, mice were challenged with a sublethal dose of Listeria monocytogenes. The immune response was monitored in the blood at day -3, 4 and 7 post injection (Fig. 4c ). Already at day 4 most CD8 + T cells in the blood had differentiated towards an effector-phenotype (CD44 + CD62L -) in KO mice (Fig. 4d) , which is in line with the in vitro stimulation data. However, in accordance with the peak of a normal immune response initiated from naïve T cells 52 , the peak of activated effectors in WT mice was reached only after seven days. At day 3 and day 7 post injection, mice were sacrificed and spleen and liver were used to determine clearance of the Listeria by counting colony-forming units. Despite the fact that in KO at day 3 most T AIM cells had differentiated into activated T cells, the Listeria infection was not cleared. At day 7 complete clearance was observed in the spleen from WT-mice, which is in accordance with literature 52 ; however, the Dot1L KO mice failed to clear Listeria (Fig. 4e-f ). In order to investigate the reason for the compromised clearance in KO, despite the rapid T EFF formation in the blood upon infection, we examined the splenic CD8 + T cells subsets in more detail. In WT mice the peak of activated (CD44 + CD62L -) CD8 + T cells was at day 7, whereas in Dot1L KO mice this was at day 3 ( Fig. 4g -h and Sup Fig. 4e ). However, the absolute numbers of total CD8 + T cells in KO on day 3 were not higher than in WT. Furthermore, in WT the number of CD8 + T cells at day 7 had increased 8.6-fold, whereas in KO this was only 4.9-fold, suggesting a lack of expansion despite the acquisition of an effector phenotype (Fig. 4i) . Indeed, staining for Ki67, a marker of proliferation showed that at day 3, the frequency of Ki67 + CD44 + CD62L -CD8 + T cells was not increased in KO cells compared to WT (Fig. 4j) . Besides the reduced proliferative capacity in vivo, even up to day 7, the KO CD44 + CD62L -CD8 + T cells showed impaired induction of the effector hallmarks KLRG1, IFNγ and CD49d ( Fig. 4k -l, Sup Fig. 5b ). This demonstrates that in normal CD8 + T cells DOT1L-mediated H3K79 methylation generally marks expressed genes, but only a subset of the methylated genes needs H3K79 methylation for maintaining full expression levels. Thus, in normal T cells DOT1L does not act as a transcriptional switch but rather seems to be required for transcription maintenance of a subset of methylated genes that are already on and it indirectly promotes repression of genes.

Amongst the differentially expressed genes marked with H3K79me2, we searched for candidate direct target genes that could mediate the role of DOT1L in terminal T cell differentiation. We were especially interested in factors that could explain the prominent gene derepression in Dot1L-KO cells. To this end we selected genes that were significantly downregulated in KO and had harbored H3K79me2 at the 5' end of the gene, both in T N and T CM . We further narrowed the list down to genes that were annotated as "negative regulator of transcription by RNA polymerase II", resulting in 14 transcriptional regulators.

Among those, Ezh2 emerged as a potentially relevant target of DOT1L that could explain part of the derepression of genes in Dot1L-KO cells (Table 1 ). EZH2 is part of the Polycomb-repressive complex 2 (PRC2), which deposits H3K27me3 54-59 , a mark involved in repression of developmentally-regulated genes and in switching off naïve and memory genes during terminal differentiation of effector CD8 + T cells 40, 60 . Although the change in Ezh2 mRNA expression was modest between WT and KO (Sup Fig. 5c ), the Ezh2 gene was H3K79me2 methylated (Sup Fig. 5d) , and Dot1L-KO mice share several T-cell phenotypes with Ezh2-KO mice 40, 60 , although this is dependent on the mouse model used [61] [62] [63] . To further investigate the idea that misregulation of PRC2 targets could be one of the downstream consequences of loss of DOT1L in CD8 + T cells, we compared the gene expression changes in an Ezh2-KO model 61 with those seen in Dot1L-KO CD8 + T cells. This revealed substantial overlap between the derepressed genes in the two models (Fig. 5b) , suggesting a functional connection between two seemingly opposing epigenetic pathways. Furthermore, we determined H3K27me3 scores based on previous ChIP-Seq studies 64 and compared them with the gene expression in WT and Dot1L-KO SP CD8 + thymocytes and peripheral CD8 + T cells. This analysis showed that the genes that were upregulated in Dot1L-KO and lack H3K79me2 in WT were strongly enriched for H3K27me3 in WT memory precursor CD8 + T cells (Fig. 5c) . As a control, expression matched non-differentially expressed genes were not enriched for H3K27me3 (Fig. 5c) . Taken together, these findings suggest that one of the consequences of loss of DOT1L-mediated H3K79me2 is derepression of a subset of PRC2 targets that are normally actively repressed. This, together with the other transcriptional changes likely contributes to the perturbation of the epigenetic identity of CD8 + T AIM cells, thereby compromising their ability to terminally differentiate in vivo.

The histone methyltransferase DOT1L has emerged as a druggable target in MLL-rearranged leukemia and additional roles in cancer have been suggested [22] [23] [24] [65] [66] [67] [68] . This in combination with the availability of highly-specific DOT1L inhibitors make DOT1L a potential target for cancer therapy. However, the role of In CD8 + T cells, loss of DOT1L resulted in a massive gain of CD44 + CD62L + memory-like cells associated with a simultaneous loss of the naïve compartment. These cells start to acquire memory features intrathymically, express a diverse un-skewed TCRβ repertoire, and lack expression of CD49d. Taken together, this suggests that these memory-like CD8 + T cells arose independently of foreign antigens, leading us to designate these cells as antigen-independent memory-like T AIM CD8 + cells. Importantly, such unconventional memory-phenotype cells constitute a substantial (15-25%) fraction of the peripheral CD8 + T cell compartment in WT mice and humans 4, 35, 44, 78, 79 . The fraction of unconventional memory-phenotype T cells further increases with age 80, 81 . The biological role of antigen-inexperienced memory CD8 + T cells is still not completely understood, but these cells have been observed to respond more rapidly to TCR activation than T N cells and they have been suggested to provide by-stander protection against infection in an antigen-independent way 34, 44 . However, in older mice, virtual memory cells have been reported to lose their proliferative potential and acquire characteristics of senescence 82 .

The mechanism of unconventional memory-phenotype differentiation is also poorly understood 2 . In some mouse models CD8 + T cells differentiate upon excess production of IL-4 by iNKT or γδ T cells 3, 39, [83] [84] [85] [86] . In contrast, other mouse models report that antigen inexperienced memory CD8 + T cell differentiation is cell intrinsic 87, 88 . Here we show that DOT1L plays an important role in T AIM CD8 + cell differentiation by cell-intrinsic mechanisms.

It has been established in several independent mouse models that the quality of TCR signaling closely relates to the formation of T AIM cells 34, 43, 44 . Our results show that loss of DOT1L leads to reduced surface expression of the CD3/TCR complex and co-receptors. This phenotype is likely related to the reduced expression of CD3ζ (Cd247), a target of DOT1L and a rate limiting molecule for assembly and transport of TCR/CD3 complexes to the cell surface 1, [45] [46] [47] [48] . The failure to upregulate the TCR/CD3 complex upon positive selection likely prohibits differentiation of conventional T N and supports formation of T AIM cells.

In addition, Dot1L ablation perturbed expression of TCR signaling genes, including Itk (IL2-inducible T cell kinase; a member of the Tec kinase family). Disruption of ITK signaling has also been reported to lead to antigen-independent T-cell differentiation 49 . Together, this suggests that one of the key functions of DOT1L is to ensure adequate TCR surface expression and signaling to maintain naivety and prevent T AIM cell differentiation. The discovery of DOT1L as a key player in preventing premature antigen-independent differentiation towards memory-type cells warrants further investigation and will aid in further uncovering the origin and regulation of this emerging and intriguing subset of the immune system.

Functionally, Dot1L-KO CD8 + T AIM cells partially displayed features of memory cells in vitro but in vivo showed an impaired immune response against Listeria monocytogenes. The rapid differentiation towards effector cells suggests that Dot1L-KO T AIM cells might execute a partial by-stander response, but DOT1L appears to be required to accomplish full effector function. Interestingly, similar features have been observed for virtual memory cells 44 . The reduced TCR density and affected TCR signaling network in Dot1L-KO T cells likely further contribute to the impaired immune response.

How does DOT1L affect T cell differentiation at the chromatin level? Inspection of the transcriptome and epigenome provided evidence that DOT1L methylates transcriptionally active genes in T cells and positively affects gene expression. However, only a subset of the targets required DOT1L for maintenance of normal expression levels, which is in agreement with previous observations 89, 90 . Why some genes depend on DOT1L/H3K79 methylation and others not is not known yet although a recent study indicates that in MLL-rearranged leukemic cell lines some genes harbor a 3' enhancer located in the H3K79me2/3 marked genic region, which can make them more sensitive to loss of DOT1L 89 . One of the genes that was H3K79 methylated and dependent on DOT1L in normal peripheral CD8 + T cells was Ezh2 and EZH2/PRC2 targets were derepressed in the absence of DOT1L. Importantly, analysis of data from Kagoya et al. 26 showed that Ezh2 expression is also reduced in human T cells in which DOT1L was inactivated not by deletion but by treatment with a DOT1L inhibitor (Sup Fig. 5e ). This suggests that the epigenetic crosstalk that we uncovered in mice is evolutionarily conserved. Interestingly, recruitment of DOT1L to nucleosomes by its interaction partner AF10 has been shown to be negatively affected by methylation of H3K27 91 . Therefore, the crosstalk between DOT1L and PRC2 might involve mutual interactions. Restricting DOT1L to nucleosomes unmethylated at H3K27 might be one of the mechanisms by which H3K79me and H3K27me3 are directed to non-overlapping sites, as we also observed in T cells (Fig. 5C) .

Derepression of some of the targets of PRC2 is just one of the consequences of loss of DOT1L. Besides Ezh2, DOT1L affects the expression of other genes, including several additional candidate transcription regulators ( Table 1 ). In the future, it will be important to determine other mechanisms by which DOT1L affects the CD8 + T cell transcriptome to fully understand its central role in CD8 + T cell biology. Finally, we cannot exclude that DOT1L has additional methylation targets besides H3K79 that contribute to role of DOT1L in safeguarding T cell differentiation and effector functions. Although understanding the mechanisms in more detail will require further studies, the role of DOT1L in preventing premature differentiation and safeguarding the epigenetic identity is conserved in other lymphocyte subsets. In an accompanying study (Aslam et al) , we observed in an independent mouse model that loss of Dot1L in B cells also led to premature differentiation, perturbed repression of PRC2 targets, and a compromised humoral immune response. Therefore, DOT1L is emerging as a central epigenetic regulator of lymphocyte differentiation and functionality.

In conclusion, we identify H3K79 methylation by DOT1L as an activating epigenetic mark critical for CD8 + T cell differentiation and maintenance of epigenetic identity. Further investigation into the central role of the druggable epigenetic writer DOT1L in lymphocytes is likely to provide novel strategies for immune modulations and disease intervention [92] [93] [94] [95] .

Lck-Cre;Dot1L fl/fl mice have been described elsewhere previously 25 (Table 2) . Of note, for OT-I tetramer stains the CD8 antibody clone 53-6.7 was used 98 . For intracellular staining, cells were fixed and permeabilized using the Transcription Factor Buffer kit (Benton Dickinson). Antibodies for intracellular staining were diluted 1:200 in Perm/Wash buffer (Table   3) . For H3K79me2 staining, cells were first stained with surface markers and fixed and permeabilized as 

RNA-Seq reads were mapped to mm10 (Ensembl GRCm38) using TopHat with the arguments `--prefiltermultihits -no-coverage-search -bowtie1 -library-type fr-firststrand` using a transcriptome index. Counts per gene were obtained using htseq-count with the options `-m union -s no` and Ensembl GRCm38.90 gene models. Analysis was restricted to genes that have least 20 counts in at least 4 samples, and at least 5 counts in 4 samples in specific contrasts, to exclude very low abundance genes. Differential expression analysis was performed on only relevant samples using DESeq2 and default arguments with the design set to either Dot1LKO status or cell type. Adaptive effect size shrinkage was performed with the ashr R package to control for the lower information content in low abundance transcripts. Genes were considered to be differentially expressed when the p-value of the negative binomial Wald test was below 0.01 after the Benjamini-Hochberg multiple testing correction. Sets of differentially expressed genes in indicated conditions were called 'gene signatures'. An exception was made for the Ezh2 RNA-Seq data from He et al., 61 wherein the dispersion was estimated with a local fit, using `estimateDispersion` function in DESeq2 with the argument `fitType = "local"` and using an adjusted pvalue cutoff of 0.05, to increase the number of detected differential genes. Principal component analysis was performed using the `prcomp` function on variance stabilizing transformed data of all samples with the `vst` function from the DESeq2 package on all samples using default arguments. For analyses where we performed expression matching, we chose genes with an absolute log2 fold changes less than 0.1 and false discovery rate corrected p-values above 0.05 that were closest in mean expression to each of the genes being matched without replacement.

Sorted cells were centrifuged at 500 rcf. The pellet was resuspended in IMDM containing 2% FCS and formaldehyde (Sigma) was added to a final concentration of 1%. After 10 min incubation at RT glycine (final concentration 125 mM) was added and incubated for 5 min. Cells were washed twice with ice-cold PBS containing Complete, EDTA free, protein inhibitor cocktail (PIC) (Roche). Cross-linked cell pellets were stored at -80 o C. Pellets were resuspended in cold Nuclei lysis buffer (50mM Tris-HCl pH 8.0, 10mM EDTA pH8.0, 1%SDS) + PIC and incubated for at least 10 min. Cells were sonicated with PICO to an average length of 200-500bp using 30s on/ 30s off for 3 min. After centrifugation at high speed debris was removed and 9x volume of ChIP dilution buffer (50mM Tris-HCl pH8, 0.167M NaCl, 1.1% Triton X-100, 0.11% sodium deoxycholate) + PIC and 5x volume of RIPA-150(50mM Tris-HCl pH8, 0.15M NaCl, 1mM EDTA pH8, 0.1% SDS, 1% Triton X-100, 0.1% sodium deoxycholate) + PIC was added. Shearing efficiency was confirmed by reverse crosslinking the chromatin and checking the size on agarose gel.

Chromatin was pre-cleared by adding ProteinG Dynabeads (Life Technologies) and rotation for 1 hour at 4 o C. After the beads were removed 2μl H3K79me1, 2μl H3K79me2 (NL59, Merck Millipore) and 1μl

H3K4me3 (ab8580, Abcam) were added and incubated overnight at 4 o C. ProteinG Dynabeads were added to the IP and incubated for 3 hours at 4 o C. Beads with bound immune complexes were subsequently washed with RIPA-150, 2 times RIPA-500 (50mM Tris-HCl pH8, 0.5M NaCl, 1mM EDTA pH8, 0.1% SDS, 1% Triton X-100, 0.1% sodium deoxycholate), 2 times RIPA-LiCl (50mM Tris-HCl pH8, 1mM EDTA pH8, 1% Nonidet P-40, 0.7% sodium deoxycholate, 0.5M LiCl2) and TE. Beads were resuspended in 150 μl Direct elution buffer (10mM Tris-HCl pH8, 0.3M NaCl, 5mM EDTA pH8, 0.5%SDS) and incubated overnight at 65 o C and input samples were included. Supernatant was transferred to a new tub and 1μl

RNAse A (Sigma) and 3 μl ProtK (Sigma) was added per sample and incubated at 55 o C for 1 hour. DNA was purified using Qiagen purification columns.

Library preparation was done using KAPA LTP Library preparation kit using the manufacturer's protocol with slight modifications. Briefly, after end-repair and A-tailing adaptor were ligated followed by Solid Phase Reverisble Immobilization (SPRI) clean-up. Libraries were amplified by PCR and fragments between 250-450 bp were selected using AMPure XP beads (Beckman Coultier). The libraries were analyzed for size and quantity of DNAs on a 2100 Bioanalyzer using a High Sensitivity DNA Kit (Agilent), diluted and pooled in multiplex sequencing pools. The libraries were sequenced as 65 base single reads on a HiSeq2500 (Illumina).

ChIP-Seq samples were mapped to mm10 (Ensembl GRCm38) using BWA-MEM with the option '-M' and filtered for reads with a mapping quality higher than 37. Duplicate reads were removed using MarkDuplicates from the Picard toolset with `VALIDATION_STRINGENCY=LENIENT` and `REMOVE_DUPLICATES=false` as argument. Low quality reads with any FLAG bit set to 1804 were removed. Bigwig tracks were generated by using bamCoverage from deepTools using the following arguments: `-of bigwig -binsize 25 -normalizeUsing RPGC -ignoreForNormalization chrM -effectiveGenomeSize 2652783500`. For visualization of heatmaps and genomic tracks, bigwig files were loaded into R using the `import.bw()` function from the rtracklayer R package. TSSs for heatmaps were taken from Ensembl GRCm38.90 gene models by taking the first base pair of the 5' UTR of transcripts.

When such annotation was missing, the most 5' position of the first exon was taken.

Genes were selected that are down in KO with adjusted p-value < 0.01, with H3K79me2 normalized reads > 20 and the GO annotation "negative regulator of transcription by RNA polymerase II".

Between 100,000 and 1,100,000 sorted cells were used for TCRβ-Seq. DNA was extracted using Qiagen DNA isolation kit (Qiagen) and collected in 50 μl TE buffer. The concentration was determined with Nanodrop. Samples were sequenced and processed by Adaptive Biotechnologies using immunoSEQ mm TCRB Service (Survey).

Statistical analyses were performed using Prism 7 (Graphpad). Data are presented as means ±SD unless otherwise indicated in the figure legends. The unpaired Student's t-test with two-tailed distributions was used for statistical analyses. A p-value <0.05 was considered statistically significant. * p < 0.05, ** p < 0.01, *** p < 0.001.

We thank the NKI animal pathology facility for histology and immunohistochemistry, as well as advice, the NKI genomics core facility for library preparations and sequencing, the NKI Flow Cytometry facility for assistance, the NKI Animal Laboratory Intervention Unit for performing immunization experiments, and the caretakers of the NKI Animal Laboratory facility for assistance and excellent animal care. The following reagent(s) was/were obtained through the NIH Tetramer Core Facility: CD1d-PBS57 and CD1d- Percentage of CD44 CD62L subsets in CD8 + T cells from spleen at day 3 and h) day 7. i) absolute number of CD3 + CD8 + T cells in spleen at day 3 and day 7. j) Intracellular Ki67 positive cells in CD8 + splenocytes. k) Percentage of KLRG1 + in CD8 + T cells from the spleen. l) Percentage of IFNγ-producing cells in CD8 + T cells from spleen after four hours stimulation with PMA/Ionomycin. g-l) Data is represented as mean +SD. 

Composition and function of T-cell receptor and B-cell receptor complexes on precursor lymphocytes

Similar but different: Virtual memory CD8 T cells as a memory-like cell population

Innate memory T cells

Antigen-inexperienced memory CD8(+) T cells: where they come from and why we need them

Epigenetic control of CD8(+) T cell differentiation

The interface between transcriptional and epigenetic control of effector and memory CD8 + T-cell differentiation

Remembering to remember: T cell memory maintenance and plasticity

Epigenetic Maintenance of Acquired Gene Expression Programs during Memory CD8 T Cell Homeostasis

The molecular basis of the memory T cell response: differential gene expression and its epigenetic regulation

Metabolic and Epigenetic Coordination of T Cell and Macrophage Immunity

The upstreams and downstreams of H3K79 methylation by DOT1L

DOT1L/KMT4 recruitment and H3K79 methylation are ubiquitously coupled with gene transcription in mammalian cells

Dot1l deficiency leads to increased intercalated cells and upregulation of V-ATPase B1 in mice

The ZFP-1(AF10)/DOT-1 complex opposes H2B ubiquitination to reduce Pol II transcription

Nucleosome Turnover Regulates Histone Methylation Patterns over the Genome

The histone modification pattern of active genes revealed through genome-wide chromatin analysis of a higher eukaryote

DOT1L and H3K79 Methylation in Transcription and Genomic Stability

Dot1a-AF9 complex mediates histone H3 Lys-79 hypermethylation and repression of ENaCalpha in an aldosterone-sensitive manner

The emerging roles of DOT1L in leukemia and normal development

Methylation of Histone H3K79 by Dot1L Requires Multiple Contacts with the Ubiquitinated Nucleosome

Chromatin-modifying enzymes as modulators of reprogramming

MLL-rearranged leukemia is dependent on aberrant H3K79 methylation by DOT1L

The DOT1L inhibitor pinometostat reduces H3K79 methylation and has modest clinical activity in adult acute leukemia

Leukaemic transformation by CALM-AF10 involves upregulation of Hoxa5 by hDOT1L

Conserved crosstalk between histone deacetylation and H3K79 methylation generates DOT1L-dose dependency in HDAC1-deficient thymic lymphoma

DOT1L inhibition attenuates graft-versus-host disease by allogeneic T cells in adoptive immunotherapy models

A chemical biology toolbox to study protein methyltransferases and epigenetic signaling

The histone H3K79 methyltransferase Dot1L is essential for mammalian development and heterochromatin structure

Early mammalian erythropoiesis requires the Dot1L methyltransferase

Progressive methylation of ageing histones by Dot1 functions as a timer

Direct screening for chromatin status on DNA barcodes in yeast delineates the regulome of H3K79 methylation by Dot1

Patterns and mechanisms of ancestral histone protein inheritance in budding yeast

The virtuous self-tolerance of virtual memory T cells

Strong homeostatic TCR signals induce formation of self-tolerant virtual memory CD8 T cells

The antigen-specific CD8 + T cell repertoire in unimmunized mice includes memory phenotype cells bearing markers of homeostatic expansion

Type i interferons regulate eomesodermin expression and the development of unconventional memory CD8(+) T cells

Effector and memory CD8+ T cell fate coupled by T-bet and eomesodermin

Virtual memory CD8T cells display unique functional properties

The Tec Family Tyrosine Kinases Itk and Rlk Regulate the Development of Conventional CD8+ T Cells

Early transcriptional and epigenetic regulation of CD8 + T cell differentiation revealed by singlecell RNA sequencing

Immediate antigen-specific effector functions by TCR-transgenic CD8+ NKT cells

Isotypic exclusion of γδ T cell receptors in transgenic mice bearing a rearranged β-chain gene

Two separate defects affecting true naive or virtual memory T cell precursors combine to reduce naive T cell responses with aging

Virtual memory T cells develop and mediate bystander protective immunity in an IL-15-dependent manner

The CD3 ζ Subunit Contains a Phosphoinositide-Binding Motif That Is Required for the Stable Accumulation of TCR-CD3 Complex at the Immunological Synapse

The T Cell Receptor: Critical Role of the Membrane Environment in Receptor Assembly and Function

TCR ζ-chain downregulation: Curtailing an excessive inflammatory immune response

Genetic and Mutational Analysis of the T-Cell Antigen Receptor

TCR signaling via Tec kinase ITK and interferon regulatory factor 4 (IRF4) regulates CD8+ T-cell differentiation

T cell receptor antagonist peptides induce positive selection

Effector and memory T-cell differentiation: implications for vaccine development

T-cell activation, proliferation and apoptosis in primary Listeria monocytogenes infection

Reciprocal intronic and exonic histone modification regions in humans

The Polycomb complex PRC2 and its mark in life

PRC2 is high maintenance

The Complexity of PRC2 Subcomplexes

Mechanisms Regulating PRC2 Recruitment and Enzymatic Activity

PRC2.1 and PRC2.2 Synergize to Coordinate H3K27 Trimethylation

Structure, mechanism, and regulation of polycomb-repressive complex 2

Polycomb Repressive Complex 2-Mediated Chromatin Repression Guides Effector CD8+ T Cell Terminal Differentiation and Loss of Multipotency

Ezh2 phosphorylation state determines its capacity to maintain CD8+ T memory precursors for antitumor immunity

miR-155 harnesses Phf19 to potentiate cancer immunotherapy through epigenetic reprogramming of CD8 + T cell fate

Ezh2 Regulates Activation-Induced CD8+ T Cell Cycle Progression via Repressing Cdkn2a and Cdkn1c Expression

Epigenetic landscapes reveal transcription factors that regulate CD8+T cell differentiation

The protective role of DOT1L in UV-induced melanomagenesis

A novel germline variant in the DOT1L gene co-segregating in a Dutch family with a history of melanoma

Prognostic and therapeutic value of disruptor of telomeric silencing-1-like (DOT1L) expression in patients with ovarian cancer

The Histone Methyltransferase DOT1L Promotes Neuroblastoma by Regulating Gene Transcription

The epigenetic control of stemness in CD8+ T cell fate commitment

Acetylation of the Cd8 Locus by KAT6A Determines Memory T

Histone Deacetylase 7 Regulates Cell Survival and TCR Signaling in CD4/CD8 Double-Positive Thymocytes

Histone Deacetylase 7 Functions as a Key Regulator of Genes Involved in both Positive and Negative Selection of Thymocytes

HDAC7, a thymus-specific class II histone deacetylase, regulates Nur77 transcription and TCRmediated apoptosis

Histone Deacetylase 3 is Required for Efficient T-cell Development

HDAC3 restrains CD8-lineage genes to maintain a bi-potential state in CD4+CD8+ thymocytes for CD4-lineage commitment

Primary cutaneous CD4+ small/medium pleomorphic T-cell lymphoproliferative disorder: Where do we stand? A systematic review

Derivation and Maintenance of Virtual Memory CD8 T Cells

Phenotype of NK-like CD8(+) T cells with innate features in humans and their relevance in cancer diseases

Cutting edge: Central memory CD8 T cells in aged mice are virtual memory cells

Virtual memory cells make a major contribution to the response of aged influenza-naïve mice to influenza virus infection

Age-Related Decline in Primary CD8+ T Cell Responses Is Associated with the Development of Senescence in Virtual Memory CD8+ T Cells

Tec kinase Itk in γδT cells is pivotal for controlling IgE production in vivo

A non-canonical function of Ezh2 preserves immune homeostasis

Coupling of T cell receptor specificity to natural killer T cell development by bivalent histone H3 methylation

Histone Deacetylase 7 mediates tissue-specific autoimmunity via control of innate effector function in invariant

Bcl11b prevents the intrathymic development of innate CD8 T cells in a cell intrinsic manner

The NF-κB1 transcription factor prevents the intrathymic development of CD8 T cells with memory properties

DOT1L inhibition reveals a distinct subset of enhancers dependent on H3K79 methylation

MLL-AF4 Spreading Identifies Binding Sites that Are Distinct from Super-Enhancers and that Govern Sensitivity to DOT1L Inhibition in Leukemia

The PZP Domain of AF10 Senses Unmodified H3K27 to Regulate DOT1L-Mediated Methylation of H3K79

Chromatin regulatory mechanisms and therapeutic opportunities in cancer

Epigenetic modulators, modifiers and mediators in cancer aetiology and progression

Advances in epigenetics link genetics to the environment and disease

Clinical epigenetics: seizing opportunities for translation

A conditional knockout resource for the genome-wide study of mouse gene function

lck-Driven Cre Expression Alters T Cell Development in the Thymus and the Frequencies and Functions of Peripheral T Cell Subsets

Critical role for CD8 in T cell receptor binding and activation by peptide/major histocompatibility complex multimers

Nonprocessive methylation by Dot1 leads to functional redundancy of histone H3K79 methylation states