|Home | About | Journals | Submit | Contact Us | Français|
Memory lymphocytes are characterized by their ability to exhibit a rapid response to the recall antigen, in which differential transcription plays a significant role, yet the underlying mechanism is not understood. We report here a genome-wide analysis of histone methylation on two histone H3 lysine residues (H3K4me3 and H3K27me3) and gene expression profiles in naïve and memory CD8 T cells. We found that a general correlation exists between the levels of gene expression and the levels of H3K4me3 (positive correlation) and H3K27me3 (negative correlation) across the gene body. These correlations display four distinct modes: repressive, active, poised, and bivalent, reflecting different functions of these genes. Furthermore, a permissive chromatin state of each gene is established by a combination of different histone modifications. Our findings reveal a complex regulation by histone methylation in differential gene expression and suggest that histone methylation may be responsible for memory CD8 T cell function.
Immunological memory provides a fundamental basis of vaccination, yet it remains unresolved how the immune system achieves this long lasting enhanced function to re-encounter the same pathogen. At the center of immunological memory are the memory lymphocytes that are capable of mounting a rapid and robust cellular response and have the stem cell like ability of self-renewal (Dutton et al., 1998; Antia et al., 2005). Memory T cells are heterogeneous populations that can be divided into two functional subsets: central memory (TCM) and effector memory (TEM) T cells, based on their different homing capacities and effector functions (Lanzavecchia and Sallusto, 2005). The functional properties of memory T cells are acquired after activation of naïve cells in which transcriptional regulation of specific genes plays a central role in the process of memory cell generation and subsequent maintenance.
Antigenic stimulation results in serial transcriptional changes in naïve CD8 T cells and leads to their differentiation to effector CD8 T cells capable of eliminating infected target cells (Williams and Bevan, 2007). A few transcription factors that are essential for this differentiation process have been identified including TBX21 (T-bet), EOMES, and PRDM1 (BLIMP1) (Sullivan et al., 2003; Pearce et al., 2003; Martins et al., 2006). The activation of these transcription factors in turn induces the expression of effector function-related genes including interferon gamma, perforin and granzyme B. After the clearance of infected cells and pathogens, memory cells emerge and express a profile of genes that are different from both naïve and effector CD8 cells (Kaech et al., 2002). Within the memory CD8 T cell subsets, TCM expresses some homing and anti-apoptosis genes while TEM expresses genes associated with effector functions, pro-apoptotic signaling and certain chemokines (Willinger et al., 2005; Chtanova et al., 2001; Klebanoff et al., 2005). The identification of these differentially expressed genes in memory CD8 T cell subsets provides a transcriptional basis of memory CD8 T cell function. However, the mechanism by which this differential gene expression is established and maintained in memory T cells is currently unknown.
Euchromatin states facilitate gene expression while heterochromatin states are associated with gene silencing (Li et al., 2007). Accumulating evidences suggest that the covalent modifications of histone N-terminal tails such as acetylation, methylation, and phosphorylation can act to regulate chromatin states (Kouzarides, 2007; Berger, 2007). Among the different histone modifications, histone acetylation such as acetylation of histone 3 lysine 9 (H3K9ac) and H3K14ac is generally associated with euchromatin and active gene transcription (Roh et al., 2006; Wang et al., 2008) while histone methylation displays a more complex relationship with chromatin states (Barski et al., 2007; Guenther et al., 2007). The methylation of histone H3 lysines 4, 36 and 79 (H3K4, H3K36, and H3K79) is associated with euchromatin whereas the methylation of H3K9, H3K27, and H4K20 is associated with heterochromatin and gene silencing (Barski et al., 2007; Guenther et al., 2007). There are three forms of methylation of histone tail residue, corresponding to the addition of mono-, di-, and tri-methyl groups and each form of methylation may have different consequence in shaping of the chromatin states (Klose and Zhang, 2007; Barski et al., 2007).
Histone hyperacetylation is found in cytokine genes (Il4, Il7r, and Ifng) in Th1 and Th2 CD4 T cells (Yamashita et al., 2004; Fields et al., 2002) and in CD8 T cell subsets (Northrop et al., 2006; Chandele et al., 2008). Furthermore, hyperacetylation has been found in the loci of many differentially expressed genes of memory CD8 T cells (Fann et al., 2006), particularly in cytokine and effector genes (Northrop et al., 2008; Araki et al., 2008). More importantly, induced hyper- or hypoacetylation in these gene loci results in increased or decreased gene expression in CD8 T cells. Together, these findings indicate that histone acetylation regulates differential gene expression and thereby the function of memory CD8 T cells. Compared to histone acetylation, the methylation of a limited number of histone tail residues has been found in the same cytokine gene loci (Il4 and Ifng) in CD8 T cells (Makar et al., 2003; Shnyreva et al., 2004). However, the role of histone methylation in the regulation of differential gene expression during the differentiation from naïve to memory CD8 T cells has not previously been examined.
To address whether there is an association of histone methylation and the differential gene expression and function of memory CD8 T cells, we performed a global analysis of histone methylations in parallel with genome-wide gene expression profiling in naïve and memory CD8 T cell subsets. We found a general relationship between gene expression levels and the distribution of histone methylation, i.e. a positive correlation of gene expression with the levels of H3K4me3 and a negative correlation with the levels of H3K27me3, in CD8 T cell subsets. Furthermore, we have identified four distinct forms of the relationship between gene expression and the histone methylation in memory CD8 T cells: repressed, active, poised, and bivalent. These different modes of association provide novel insights into the complex regulation of gene expression and function in memory CD8 T cells. Finally, we demonstrated that actively expressed genes and permissive chromatin can be associated with different histone modifications (H3K4me3 and H3K9ac) and that reduced H3K9ac levels are associated with a decrease in mRNA levels in memory cells.
To determine the global histone methylation status and its relationship to gene expression in CD8 T cell differentiation, we isolated naïve and memory CD8 T cell subsets (TCM and TEM) from peripheral blood of healthy adults by cell sorting and analyzed the genome-wide distribution of H3K4me3 and H3K27me3 by ChIP-Seq and the global gene expression profile by Agilent human whole genome chip of these CD8 T cell subsets (Figure 1). We generated native chromatin containing primarily mononucleosomes by micrococcal nuclease (MNase) digestion and performed ChIP assay with anti-H3K4me3 and anti-H3K27me3 antibodies. The ChIP DNA fragments were sequenced using Illumina-Solexa 1G Genome Analyzer as described (Barski et al., 2007; Wang et al., 2008). A total of ~28 million sequence tags (~3 million tags of H3K4me3 and ~6 million tags of H3K27me3 for each of CD8 subsets) were generated for naïve and memory (TCM and TEM) cells and mapped to the human genome. In parallel, global gene expression profiles of the same subsets of CD8 T cells at rest and after activation with anti-CD3 plus anti-CD28 antibodies (anti-CD3/CD28) for 16 hours was determined. Genes whose expression levels were significantly changed among subsets or after activation were identified by a combination of ANOVA (FDR≤0.01), intensity changes (≥2 fold), and minimum expression intensity (≥1.5) (the complete data set can be found at the NCBI Gene Expression Omnibus with accession number GSE14422). We confirmed the correlation between the levels of H3K4me3 or H3K27me3 with mRNA by quantitative PCR and RT-PCR.
To assess the relationship of H3K4me3- and H3K27me3-associated chromatin states and gene expression, we analyzed 16,314 annotated genes by quantitating the levels of H3K4me3 and H3K27me3 in the gene body (defined as 1 kb upstream of the transcription start site (TSS) to the end of the gene divided by the total number of sequence tags) and the mRNA levels (the complete data set can be found at NCBI Gene Expression Omnibus with accession number GSE12616). The level of H3K4me3 displayed a general positive correlation with mRNA levels in naïve and memory (TCM and TEM) CD8 T cells (Figure 2A). Memory cells (TCM and TEM) had higher regression coefficient values (R2=0.870 for both TCM and TEM) than did naïve cells (R2=0.511). Furthermore, memory cells (TCM and TEM) had more genes with higher levels of H3K4me3 (>2×10−5 tags per gene locus) than did naive cells: TEM, 4878 (30%); TCM, 3290 (20%); and naïve, 2064 (13%) (Figure 2C). In contrast, the level of H3K27me3 showed a general negative correlation with the mRNA levels of their corresponding genes (Figure 2B). Similar to H3K4me3, memory (TCM and TEM) cells also had higher regression coefficient values (TCM: R2=0.834 and TEM: R2=0.806) than did naïve cells (R2=0.697). However, unlike H3K4me3, there was no obvious difference in the number of the genes with high levels of H3K27me3 (>1×10−5 tags per gene locus) in naïve and memory cell subset: TEM, 1662 (10%); TCM, 1058 (8.1%); and naïve, 1372 (8.4%) (>1×10−5 tags per gene locus) (Figure 2C). These findings show a general correlation between the levels of these histone modifications and the level of gene expression from naïve to memory CD8 T cell subsets.
H3K27me3 is mainly catalyzed by Polycomb Repressive Complex 2 (PRC2) and associated with gene silencing (Cao et al., 2002; Schubert et al., 2006). When we scanned the genome for large-scale H3K27me3-enriched domains in each of the CD8 T cell subsets, we found that several regions were differentially enriched with H3K27me3 in naïve or memory CD8 T cell subsets (Table S1). Ten genomic regions, spanning from 0.2 to 1 Mb in size, with higher H3K27me3 levels in naïve cells than in memory subsets were identified (Table S1, three of them shown in Figure 3A). Each of these regions contains multiple genes with functions associated with cell signaling (RALGPS1 and GARNL3), transcription (ZIM2, PEG3, ZIM3, DUXA, and ZNF264), metabolism (USP29 and B3GALT5), transport (SLC2A8), cell junction (IGSF5), and cell adhesion (DSCAM). One representative H3K27me3 domain on chromosome 9q had consistently higher levels of H3K27me3 in naïve cells than in memory cells, especially in TEM (Figure 3B). In parallel, we also identified ten genomic domains with high level of H3K27me3 in memory CD8 T cells (region spanning from 0.2–0.45 Mb) (Table S1) and the details of three domains are shown in Figure 3C. Multiple genes are in these regions including genes involved in type I interferon (IFNA13, IFNA2, IFNA8, IFNA1, and IFNE1), apoptosis (BCL2L14), cell adhesion (C21orf29), signaling (LRP6), transport (TRPM2) and cytoskeleton (KRTAP10-1, 2, 3, 4, 5, 6, and 7). The human type I IFN gene cluster on chromosome 9p was marked with high levels of H3K27me3 in TEM cells (Figure 3D). Considering the role of interferon in immune function, it is of great interest to understand the role of these clustered high levels of H3K27me3 in regulation of naïve and memory CD8 T cell function.
Gene expression analysis identified three differentially gene expression patterns in memory CD8 T cells as compared to naïve CD8 T cells: 1) repressed gene (significantly lower mRNA levels in resting memory cells than in resting naïve cells), 2) active gene (significantly higher mRNA levels in resting memory cells than in resenting naïve cells), and 3) poised gene (no significantly difference between resting memory and naïve CD8 T cells but significantly induced in activated memory cells than in activated naïve cells). To further characterize their relationship between histone modifications and gene expression, we identified four distinct modes of association of gene expression with histone methylation (H3K4me3 and H3K27me3) in memory CD8 T cells (either TCM or TEM or both) as compared to the corresponding naïve CD8 T cells: 1) Repressed genes were associated with lower H3K4me3 and higher levels of H3K27me3 at their corresponding loci in resting memory CD8 T cells; 2) Active genes were associated with relatively high levels of H3K4me3 but low levels of H3K27me3 at the corresponding locus in resting memory CD8 T cells; 3) Poised genes had similar patterns of H3K4me3 and H3K27me3 to the active genes in resting memory CD8 T cells; and 4) Bivalent genes were associated with high levels of H3K4me3 and H3K27me3 in resting memory CD8 T cells.
271 genes were highly expressed in naïve CD8 T cells but were significantly lower in memory (TCM and/or TEM) CD8 T cells. Among these genes, 105 were repressed in both TCM and TEM (memory common repressed genes), 150 were repressed in TCM alone, and 16 were repressed in TEM alone. The functions of these genes were diverse, ranging from cell adhesion, to signaling to transcription (Table S2). Parallel to their repressed mRNA expression status, the corresponding H3K4me3 levels were significantly lower in memory subsets than in naive CD8 T cells (Figure 4A). Two representative repressed genes, SH3 domain binding glutamic acid-rich protein like 2 (SH3BGRL2) and Synaptotagmin III (SYT3), were analyzed in detail (Figure 4B). SH3BGRL2 is involved in the control of redox dependent processes and interact with PKCθ resulting in the inhibition of c-Jun, AP-1 and NF-κB (Mazzocco et al., 2002). SYT3 is essential for CXCR4 cycling and plays a role in T cell migration (Masztalerz et al., 2007). Their differential expression patterns suggest that they may participate in the differentiation and/or maintenance of naïve and memory CD8 T cells.
434 genes were identified. Among these genes, 150 were active in both TCM and TEM (memory common active genes), 55 active in TCM alone, and 229 active in TEM alone. These highly expressed genes in memory CD8 T cells also displayed diverse function and many of them were related to memory CD8 T cell function (Table S3). In a general agreement with their differential expression, the H3K4me3 levels in active gene loci were significantly higher in memory CD8 T cell subsets than in naive CD8 T cells (Figure 4C). We selected six representative active genes for further investigation (Figure 4D and Figure S1). PR domain containing 1, with ZNF domain (PRDM1, also as BLIMP1), a memory common active gene, has been shown to play an important role in the differentiation of CD8 T cells (Martins et al., 2006). Killer cell lectin-like receptor subfamily G, member 1 (KLRG1), another memory common active gene, is expressed in effector and memory CD8 T cells and has recently been shown to be a marker to distinguish between short-lived effector cells and long-lived memory precursor effector cells (Beyersdorf et al., 2001; Joshi et al., 2007) (Figure 4D). Chemokine (C-C motif) receptor 4 (CCR4), an active gene in TCM, is a receptor for CCL17 and CCL22 and plays a role in T cell homing to skin and T cell migration to inflamed tissues (Sallusto et al., 2000). Granzyme A (GZMA) and perforin 1 (PRF1), active genes in TEM, play a central role in CD8 T cell-mediated cytotoxicity along with granzyme B(GZMB) (Russell and Ley, 2002; Voskoboinik et al., 2006). Eomesodermin homolog (EOMES), an active gene in TEM, is crucial for the effector function of memory CD8 T cells (Pearce et al., 2003) (Figure S1). Our analyses suggested that these actively expressed genes in memory CD8 T cell subsets (either TCM and/or TEM) correlate well with the function of these memory cells and thus provide the signature genes for memory CD8 T cells.
Fifty-two genes were identified (Table S4). Among these genes, forty-six were expressed in TCM alone, six were expressed in TEM alone, and none was found to be expressed in both TCM and TEM. We found that the H3K4me3 levels in these poised gene loci were significantly higher in TCM or TEM than in naive CD8 T cells (Figure 4E). We analyzed four representative poised genes for more details (Figure 4F and Figure S2). Inhibitor of DNA binding 2, dominant negative helix-loop-helix protein (ID2), a poised gene in TCM, regulates genes that influence survival and the magnitude of effector responses in CD8 T cells (Cannarile et al., 2006). Pleiomorphic adenoma gene-like 1 (PLAGL1), a poised gene in TEM, induces apoptotic cell death and cell cycle G1 arrest (Varrault et al., 1998) (Figure 4F). Src-like-adaptor (SLA) is involved in the TCR signaling during thymocyte development (Myers et al., 2005) and Nuclear factor, interleukin 3 regulated (NFIL3) is a Ca2+ regulated gene that plays a critical role in signaling of cell survival (Nishimura and Tanaka, 2001). Both of SLA and NFIL3 were poised genes in TCM (Figure S2). Thus, these TCM poised genes (ID2, SLA, and NFIL3) may facilitate a rapid transition of memory cells from resting to activated state.
It has been suggested that genes with bivalent histone modifications play an important role in embryonic stem cell development (Bernstein et al., 2006). Here we identified 25 bivalent genes with high levels of both H3K4me3 and H3K27me3 in their gene loci that were not expressed in resting memory CD8 T cells but were significantly induced after activation (defined by 1) increase of mRNA levels in activated over resting memory cells by at least 2 fold, 2) increase of mRNA levels in activated memory over activated naïve by at least 2 fold, and FDR<0.01). Among these genes, three were found in both TCM and TEM; nine were bivalent in TCM alone; and thirteen were bivalent in TEM alone (Table S5). The functions of these bivalent genes in memory CD8 T cells were not limited to developmental regulation (Figure 5A). However, like stem cells, memory T cells have the capacities of self-renewal and further differentiation. KIAA1804 mixed lineage kinase 4 (MLK4), a member of MAPK kinase kinases (MAPKKKs) that control apoptosis (Gallo and Johnson, 2002) and Mohawk homeobox (MKX) were bivalent genes in both TCM and TEM (Figure 5B and Figure S3) while Gap junction protein beta 7 (GJB7, also known as connexin25) was a bivalent gene in TEM (Figure S3). To determine whether increased mRNA level after activation was associated with an increase of H3K4me3 levels, we examined the levels of mRNA and H3K4me3 in memory CD8 T cells and found a parallel increase of MLK4 mRNA and H3K4me3 in the MLK4 locus in both TCM and TEM (Figure 5C). This finding confirms that the switch from a bivalent state to an open chromatin after activation is associated with an increased expression.
The functional differences between TCM and TEM have been reported but the molecular basis for their functional differences is not fully understood. We hypothesized that differentially expressed genes account for their functional difference and that different histone methylation in these gene loci is responsible for their differential gene expression. We have identified 137 genes differentially expressed in TCM and TEM (71 highly expressed in TCM and 66 in TEM, see Experimental Procedure for details) from microarray analysis and confirmed by quantitative RT-PCR. The functions of these genes are summarized in Figure 6A (the complete list of genes can be found in Table S6). In agreement with the genome wide gene analysis (Figure 2), comparing the ratios of mRNA level (TCM/TEM) and of H3K4me3 level (the number of H3K4me3 tags in TCM/the number of H3K4me3 tags in TEM) of these differentially expressed genes reveals a positive correlation (n=137, R2=0.39) between the levels of gene expression and the levels of H3K4me3 (Figure 6B). This suggests that H3K4me3 may play an important role in the differential gene expression of memory CD8 T cell subsets. However, not all of the differentially expressed genes exhibited this correlation between mRNA and H4K3me3.
Although the permissive chromatin-related histone modifications, such as H3K4me3 and H3K9ac, are generally found in the same location and appear to act cooperatively (Wang et al., 2008), it is currently not clear whether different modifications play different roles in various gene loci. Our analysis of H3K4me3 and mRNA profiles suggests that some genes have lower levels of H3K4me3 but higher levels of mRNA. For example, tumor necrosis factor receptor superfamily, member 9 (TNFRSF9 or 4-1BB) is an important co-stimulatory receptor for CD8 T cell activation (Shuford et al., 1997). TNFRSF9 was highly expressed and yet had low levels of H3K4me3 at its gene locus in memory CD8 T cells. To determine whether other active histone modifications contribute to the determination of mRNA level, we analyzed H3K9ac in this locus and found that the level of H3K9ac was significantly higher in memory cell subsets than in naïve cells while no difference exists in the levels of H3K4me3 among the subsets (Figure 7A). This finding suggests that H3K9ac but not H3K4me3 is associated with active gene expression of TNFRSF9. However, most actively expressed genes had high levels of both H3K4me3 and H3K9ac in their gene loci. For example, signaling lymphocytic activation molecule family member 1 (SLAMF1 also known as SLAM), a self-ligand receptor of activated T cells and dendritic cells, was highly expressed in memory cells with high levels of both H3K4me3 and H3K9ac at its locus (Figure 7B).
To further determine the role of higher levels of H3K9ac in the expression of these two genes, we treated memory CD8 T cell subsets with curcumin, a histone acetyltransferase (HAT) inhibitor that inhibits p300/CBP (Balasubramanyam et al., 2004; Araki et al., 2008). Expectedly, curcumin reduced the levels of H3K9ac in these two gene loci but had no obvious effect on the levels of H3K4me3 in TCM and TEM under the condition we tested (Figure 7C). The level of mRNA, H3K9ac and H3K4me3 in the ACTB gene did not change with curcumin treatment. In parallel, the mRNA levels of TNFRSF9 and SLAMF1 were significantly decreased in curcumin-treated cells. However, the reduction of TNFRSF9 mRNA level (~7 fold) was greater than that of SLAMF1 (~2 fold). Together, these results suggest that both H3K9ac and H3K4me3 contribute to the gene expression but the relative significance may differ from gene to gene.
The fundamental transcription changes (activation or repression) of specific genes during the differentiation of naïve CD8 T cells to memory (TCM and TEM) CD8 T cells upon antigenic stimulation have begun to be identified (Kaech et al., 2002; Willinger et al., 2005; Chtanova et al., 2001; Klebanoff et al., 2005) yet the genome-wide relationship between gene expression changes and chromatin states has not been addressed. The findings presented here reveal a histone methylation-mediated chromatin basis for differential gene expression in memory (TCM and TEM) CD8 T cells and provide new insight into the role of histone methylation in memory CD8 T cell differentiation and function. These findings are significant at least in two aspects.
Firstly, they demonstrate that transcriptional control at the chromatin level is tightly regulated and specifically tailored for the function of memory CD8 T cells (TCM or TEM or both). For example, genes encoding signaling molecules (ID2, SLA, and NFIL3) are expressed at similar levels in resting TCM and naïve CD8 T cells, but TCM have permissive chromatin state (poised) whereas naïve cells have a repressive chromatin state at these gene loci. Upon antigenic stimulation, TCM more rapidly transcribe ID2, SLA, and NFIL3 for enhanced signaling (Williams and Bevan, 2007) than do naïve cells. Thus, a fast and robust induction of activation is ensured in memory CD8 T cells (TCM) only after activation while the unwanted activation signals are kept in tight control at the resting state. Furthermore, the bivalent chromatin states provide another layer of flexibility in the rapid increase of gene expression. MLK4 and MKX can provide an enhanced signaling that facilitates a rapid transition of gene transcription from resting to activated states of memory CD8 T cells.
Second, the findings presented here also provide direct evidence that differential gene expression in memory CD8 T cells is associated with epigenetic changes in histone modifications. These epigenetic changes are acquired during the differentiation of naïve cells to memory CD8 T cells and facilitate the transcriptional programs that are essential for the function of memory CD8 T cells. Thus, histone modifications may serve a chromatin basis for the rapid and robust induction of effector functions and long life of memory CD8 T cells. Further studies will be necessary to directly test the roles of histone modifications in regulation of differential gene expression and function in memory CD8 T cells.
The complexity of histone modifications suggests that the different types of modifications at different residues of the N-terminal tails of histones are likely to play different roles in regulation of gene expression (Strahl and Allis, 2000). In the past decades, many histone modifications have been identified and their association with either a permissive chromatin or a heterochromatin is becoming clear (Kouzarides, 2007). Previous studies using CD4 T cells show a general association of certain histone modifications (e.g. H3K9ac, H3K4me3, and etc.) with permissive chromatin and other modifications (e.g. H3K27me3 and H3K9me3) with heterochromatin (Roh et al., 2005; Barski et al., 2007). Furthermore, H3K4me3 may act downstream of transcription initiation or perhaps during re-initiation (Sims, III et al., 2007). Recently, an analysis of 39 different histone modifications at a genome-scale in CD4 T cells suggests that numerous modifications may act cooperatively to prepare chromatin for transcriptional regulation (Wang et al., 2008). However, it is unclear how many of active modifications are needed for maintaining active transcription and whether different types of modifications are interchangeable for maintaining the open chromatin state. In addition, it has not been determined if H3K9ac serves a mark before or during the active transcription while H3K4me3 acts as a memory of the past transcription.
The findings presented here demonstrate that some actively transcribed genes are not associated with high levels of H3K4me3 but rather with high levels of H3K9ac in their gene promoters, suggesting that different histone modifications can independently establish the permissive chromatin state of a gene locus. Furthermore, reducing H3K9ac levels leads to the down-regulation of the genes that were differentially highly expressed in memory cells. The diverse degree of reduction of TNFRSF9 and SLAMF1 mRNA levels suggest that H3K4me3 and H3K9ac are both involved in the regulation of gene expression. However, their relative significance in the regulation of gene expression may depend on their levels in that specific gene locus. Therefore, it is of great interest to understand how these different modifications are established and maintained in a specific gene locus and what minimum number or combination of different histone modifications is required to establish an open chromatin state of a gene locus.
In summary, we have demonstrated a general correlation between the levels of gene expression and histone methylations (positive correlation with the H3K4me3 and negative correlation with H3K27me3) in naïve and memory CD8 T cells (TCM and TEM) at a genome scale. Based on the relative levels of H3K4me3, H3K27me3, and of mRNA, we identified four distinct relationships (repressed, active, poised and bivalent). These distinct associations provide a means for differential regulation of gene expression and explain how the rapid and robust memory T cell response is controlled at the transcription level. Extended analysis of memory T cell subsets further confirmed differential histone methylation as a potential chromatin basis for differential gene expression and function between TCM and TEM cells. Finally, we demonstrated that active gene transcription associated open chromatin can be achieved by different histone modifications such as H3K4me3 or H3K9ac and that their relative significance may be gene-specific. Together, these findings suggest that histone methylation may serve as a chromatin basis for the regulation of gene expression and the function of memory CD8 T cells. Further studies will undoubtedly shed new light on the mechanisms underlying the unique features of memory T cells as well as provide new avenues for evaluating a memory response and for improving the efficiency of vaccines.
The antibodies used in this study are as follows: Fluorescein isothiocyanate (FITC)-conjugated anti-CD45RA and Phycoerythin (PE)-conjugated anti-CD62L were purchased from eBioscience (San Diego, CA); Tri color (TC)-conjugated anti-CD8 from Invitrogen (Carlsbad, CA); Anti-histone H3 (tri methyl K4, ab8580), anti-histone H3 (tri methyl K27, ab6002), and anti-histone H3 (acetyl K9, ab4441) were from Abcam (Cambridge, MA); Purified rabbit IgG from Upstate Biotech (Billerica, MA). Micrococcal nuclease (MNase) was purchased from Sigma-Aldrich (St Louis, MO).
Peripheral blood was obtained from healthy adults via the NIA Clinical Core Facility (IRB-approved protocol MRI2003-054). The procedure for isolation and stimulation of naïve and memory CD8 T cells was previously described (Fann et al., 2006). In brief, peripheral blood mononuclear cells (PBMC) were isolated by Ficoll (GE Healthcare, Chicago, IL) gradient centrifugation. Naïve and memory CD8 T cells were then enriched by removing other types of cells through incubation with a panel of mouse mAbs against CD4, CD19, CD11b, CD14, CD16, MHC class II, erythrocytes, platelets and CD45RO (for the enrichment of naïve cells) or CD45RA (for that of memory cells). Ab-bound cells were subsequently removed by incubation with anti-mouse IgG-conjugated magnetic beads (Qiagen, Valencia, CA). These enriched naïve and memory CD8 T cells were further purified into CD8+CD45RA+CD62L+ naïve T cells, CD8+CD45RA−CD62L+ TCM and CD8+CD45RA−CD62L− TEM by a cell sorter (MoFlo; Dako Cytomation, Carpentaria, CA). Sorted naïve, TCM and TEM CD8 T cells were over 98% pure and were either used right away or incubated with anti-CD3/CD28 coupled magnetic beads (Invitrogen) at the cell:bead ratio of 1:1 for 16 hours in RPMI-1640 with 10% Fetal bovine serum and penicillin (10 U/ml)/streptomycin (10 μg/ml) (Invitrogen). The freshly isolated and stimulated cells were used for ChIP-Seq and gene expression microarray analyses.
The procedure of ChIP-Seq was previously described (Barski et al., 2007). In brief, 2 × 107 freshly sorted naïve, TCM and TEM CD8 cells were digested with MNase (2 units) for 10 minutes at 37 °C water bath to generate native chromatin template consisting primarily of mononucleosomes. Native chromatin templates were then incubated with anti-histone H3K4me3 or H3K27me3 antibodies and antibody-bound DNA fragments were extracted. After the specificity of immunoprecipitation was confirmed with known genes by real-time PCR, the DNA fragments were modified to construct a sequencing library according the manufacturer’s manual (Illumina/Solexa, San Diego, CA) with a final 18-cycle PCR amplification. DNA fragments around 200 bp (mononucleosomal DNA + adaptor) were selectively recovered from 2% agarose gel for further cluster generation and sequencing using Solexa/Illumina 1G Genome Analyzer. Sequenced tags were exported and mapped to the human genome following the Illumina Analysis Pipeline. The UCSC genome browser was then used to present the data by converting the output of analysis pipeline into browser extensible data (BED) file.
Large-scale H3K27me3-enriched genomic regions were identified using the “island” method to separate signal from noise and to take into account the unequal numbers of tags for different libraries. This method identifies significant clustering of H3K27me3 enrichment signals that are unlikely to appear by chance. Briefly, all eligible summary windows of 50kb were first identified. An eligible window was defined as meeting a required enrichment p-value of 0.2 based on a Poisson background model. Consecutive eligible windows merge to form an island, which is scored according to its probability under a random background model. The score threshold for significant islands is determined by a requirement that the expect number of such islands in the background model is 1. A H3K27me3 domain is cell-type specific if it does not overlap with domains in other cell-types. We further selected the cell-specific enriched regions with a criterion that the tag differences between naïve and memory (TCM and/or TEM) should be greater than 1.4 fold.
Calculation of tags in a gene locus was carried out by the following procedure. The tags were binned into non-overlapping windows of 200 bp. In order to filter noise, only windows with the tag number above a threshold were selected for further analysis. The threshold is determined by a p-value of 10−3 based on a Poisson background model. The number of normalized sequence tags of a gene locus was calculated by the number of tags in the selected windows from the 1 kb upstream of the transcription starting site to the end of the gene divided by the total number of sequence tags of the whole genome. The ratio of tags was compared between naïve and memory (TCM and TEM) CD8 T cells. In general, the ratio of H3K4me3 greater than 1.6 of a gene locus between naïve and memory (TCM and TEM) cells were used to define repressed or active or poised genes. For bivalent genes, the normalized tag numbers of H3K4me3 on each gene body were ≥ 5 × 10−6 and of H3K27me3 ≥ 2 × 10−6.
The procedure of DNA microarray analysis was previously described (Fann et al., 2005). Total RNA was extracted from freshly isolated and stimulated (16 hours) naïve, TCM and TEM CD8 T cells from 9 different healthy adult donors to form 3 RNA pools. The quality and quantity of total RNA were analyzed by an Agilent Biolanalyzer (Agilent Technologies, Palo Alto, CA). Three pools of RNA were used in the microarray experiment. We used Agilent Whole Human Genome 44K Oligo Chip (Agilent Technologies) for gene expression analysis according to the manufacturer’s procedure. Total RNA (400 ng) was used to make probe through one round of amplification and was followed by cRNA synthesis in the presence of 0.3 mM cyanine 3-CTP (Cy3). In parallel, cyanine 5-CTP (Cy5) was used to label universal human reference RNA (Stratagene, La Jolla, CA). The quantity of amplified cRNA was measured by NanoDrop (Agilent Technologies), and 750 ng Cy3-labeled cRNA and 750ng Cy5-labeled reference cRNA were mixed in a volume of 500 μl hybridization solution with the gene chip at 60 °C for 17 hours. After a standard washing step, the slide was air-dried by nitrogen gas and scanned by the Agilent microarray scanner. The output file consisted of processed signal intensities from Cy3 and Cy5 fluorescent channels of 40,943 targets (approximately 16,314 annotated genes with defined genomic location) per gene chip using Feature Extraction software (Agilent Technologies). Three independent microarray experiments were performed for each naïve cells, and TCM and TEM before and after stimulation. A modified ANOVA analysis was used on log-transformed data and the statistical significance was determined using the false discovery rate (FDR) using a web based analysis software (NIA array analysis, http://lgsun.grc.nia.nih.gov/ANOVA/).
Genes that met the three criteria, (1) the FDR less than 0.01, (2) the intensity difference greater than 2, and (3) the intensity of higher expressed gene greater than 1.5, were defined as differentially expressed. The criteria for memory cell repressed genes were 1) the expression intensity difference of resting naïve cells (N0) over resting memory cells (M0) ≥ 2 fold, 2) FDR ≤ 0.01, and 3) the expression intensity of the gene in N0 ≥ 1.5. The criteria for memory cell active genes were 1) the expression intensity difference of M0 over N0 ≥ 2 fold, 2) FDR ≤ 0.01, and 3) the expression intensity of the gene in M0 ≥ 1.5. The criteria for memory cell poised genes were 1) the expression intensity difference between N0 and M0 was within ± 1.5 fold, 2) the expression intensity difference of activated memory cells (MS) over activated naive cells (NS) ≥ 2 fold, 3) FDR ≤ 0.01, and 4) the expression intensity of the gene in MS ≥ 1.5.
The procedure was standard as previously described (Fann et al., 2006). In brief, total RNA was extracted from fresh and stimulated naïve and memory CD8 T cells by Stat 60 (Tel-Test, Friendswood, TX) of 12 normal donors to form four pooled RNA. The quantity of total RNA was measured by NanoDrop (Agilent Technologies) and 500 ng of total RNA was used to synthesize cDNA by reverse transcriptase (SuperScript II, Invitrogen). PCR was carried out in 20 μl of total volume with 0.4 μM primers using a SyBr Green kit (Applied Biosystems, Foster City, CA) for 40 cycles on ABI Prism 7400 (Applied Biosystems). The specific amplification of RT-PCR products was confirmed by agarose (1.5%) gel electrophoresis. Primers of PCR were designed by using the Primer Express software (Applied Biosystems) and made by Integrated DNA Technologies (Coraville, IA). The primer sequences of the genes for the confirmation can be found in Table S7.
Curcumin (Sigma), a HAT inhibitor, was dissolved in ethanol at 10 mM. Memory CD8 T cells were cultured in the presence or absence of 20 μM curcumin for 2 hours, and followed by stimulation with anti-CD3/28 antibodies for 6 hours (Araki et al., 2008). Cells were then harvested for the analyses of mRNA and regular ChIP assay.
Two-tailed Student’s t test was used in the analysis of the levels of mRNA and of histone methylation. The significance was defined by a P value ≤ 0.05.
We thank Drs. Richard Hodes and Mike Pazin for critical reading the manuscript. We thank Francis J. Chrest and Cuong Nguyen of the Flow Cytometry Unit for cell sort, Alexei Sharov for helping the microarray analysis, and Karen Madara and Sandy Alcorta at the NIA Apheresis Unit for collecting blood samples. This research was supported by the Intramural Research Programs of the National Institute on Aging and National Heart, Lung, and Blood Institute, National Institutes of Health (NIH).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.