Identification of stem cell transcriptional programs normally expressed in embryonic and neural stem cells in alloreactive CD8⁺ T cells mediating graft-versus-host disease

Abstract

A hallmark of graft-versus-host-disease (GVHD), a life-threatening complication after allogeneic hematopoietic stem cell transplantation, is the cytopathic injury of host tissues mediated by persistent alloreactive effector T cells (T_E). However, the mechanisms that regulate the persistence of alloreactive T_E during GVHD remain largely unknown. Using mouse GVHD models, we demonstrate that alloreactive CD8⁺ T_E rapidly diminished in vivo when adoptively transferred into irradiated secondary congenic recipient mice. In contrast, although alloreactive CD8⁺ T_E underwent massive apoptosis upon chronic exposure to alloantigens, they proliferated in vivo in secondary allogeneic recipients, persisted and caused severe GVHD. Thus, the continuous proliferation of alloreactive CD8⁺ T_E, which is mediated by alloantigenic stimuli rather than homeostatic factors, is critical to maintaining their persistence. Gene expression profile analysis revealed that while alloreactive CD8⁺ T_E increased the expression of genes associated with cell death, they activated a group of stem cell genes normally expressed in embryonic and neural stem cells. Most of these stem cell genes are associated with cell cycle regulation, DNA replication, chromatin modification and transcription. One of these genes, Ezh2, which encodes a chromatin modifying enzyme, was abundantly expressed in CD8⁺ T_E. Silencing Ezh2 significantly reduced the proliferation of alloantigen-activated CD8⁺ T cells. Thus, these findings identify that a group of stem cell genes could play important roles in sustaining terminally differentiated alloreactive CD8⁺ T_E and may be therapeutic targets for controlling GVHD.

Introduction

Upon antigen-presenting cell (APC) activation, T cells are “programmed” to undergo clonal expansion, generating large numbers of effector T cells (T_E) while contracting to minimize their potentially lethal activity (1-6). Consequently, the majority of CD8⁺ T_E (∼95%) may die after clearance of the antigen, with some memory T cells surviving contraction (4, 6-8). However, chronically activated T_E can be continually generated during chronic inflammatory conditions, such as responses to chronic infections, autoantigens and alloantigens. A unique clinical example is graft-versus-host disease (GVHD), a life-threatening complication after allogeneic hematopoietic stem cell transplantation (HSCT) (9-13). A hallmark of GVHD is the cytopathic injury mediated by persistent alloreactive T_E, which can occur within weeks and persist for years after transplantation (10-15). GVHD therapy which typically targets T_E have disappointing response rates(∼40%) (16). However, the molecular mechanisms that regulate the persistence of alloreactive T cells during GVHD remain largely unknown.

Emerging evidence indicates that a group of stem cell signals may play important roles in antigen-experienced memory T cells. CD8⁺ memory T cells have the ability to self-renew to survive the lifetime of an individual and can rapidly generate protective T_E upon antigenic rechallenge (1-5). Gene expression profile analysis reveals that CD8⁺ memory T cells and long-term hematopoietic stem cells (HSCs) share a self-renewal transcriptional program (17). Furthermore, antigen-stimulated CD8⁺ T cells undergo an asymmetrical division to regulate the generation of long-term memory T cells (18). Thus, memory T cells are considered to be stem cell-like cells (1, 3-4, 19). Interestingly, Wnt/β-catenin signaling, which is essential for proliferation and self-renewal of adult stem cells (20), has been shown to regulate the generation of CD44^loCD62L^hiCD122^hiBcl-2^hiSca-1^hi CD8⁺ T memory stem cells (T_MSC) (21). These CD8⁺ T_MSC have greater ability than either CD44^hiCD62L^hi central memory (T_CM) or CD44^hiCD62L^lo effector memory T cells (T_EM) to proliferate and generate T_E, thereby destroying tumors (21). This supports our previous observation that CD8⁺ T_MSC are important for sustaining alloreactive T_E mediating GVHD (15). However, these data do not explain why alloreactive CD8⁺ T_E can persist and cause severe GVHD in secondary recipients (14-15). Given that T_E and memory T cells are developmentally linked to each other (1-6, 22), we asked whether alloreactive T_E exposure to chronic alloantigens proliferate and persist through reactivation of distinct families of stem cell genes.

Using mouse models of human GVHD directed against minor histocompatibility antigens (miHAs), we demonstrate that alloantigenic stimuli rather than homeostatic factors are critical to sustaining continuous proliferation of alloreactive CD8⁺ T_E to counteract their massive apoptotic death. We found that a group of stem cell genes normally expressed in embryonic stem cells (ESCs) and neural stem cells (NSCs) was activated in these proliferating alloreactive CD8⁺ T_E upon chronic exposure to alloantigens. Most of these stem cell genes are associated with DNA replication, cell cycle regulation, chromatin modification and transcription. Silencing one of these genes, Ezh2, which encodes an enzyme with methyltransferase activity, inhibited the proliferation of alloantigen-activated T cells. Thus, these stem cell genes could be important therapeutic targets for modulating allogeneic T cell responses and GVHD.

Materials and Methods

Mice

We purchased C57BL/6 (B6; H-2D^b, CD45.2⁺), B6.SJL-Ptprc^a (B6/SJL, H-2D^b, CD45.1⁺), C3H.SW (H-2D^b, CD45.2⁺ and Ly9.1⁺) mice, BALB/b (H-2D^b, CD45.2⁺), B6.β₂ microglobulin gene-deficient mice (B6.B₂M^-/-) and BALB/c (H-2D^d, CD45.2⁺) from Jackson Laboratory (Maine, USA). We supplied transplant recipients with drinking water containing neomycin sulfate and polymyxin B (Sigma) as previously described (23). The Institutional Animal Care and Use Committee of the University of Michigan approved all mouse protocols.

Antibodies, cell lines, cytokines and flow cytometry analysis

All antibodies (Abs) used for immunofluorescence staining were obtained from BD Bioscience Pharmingen. Microbead-conjugated Abs and streptavidin were purchased from Miltenyi-Biotech, and all recombinant cytokines including IL-2, IL-4, IL-15, granulocyte-monocyte colony-stimulating factor (GM-CSF), stem cell factor (SCF) and tumor necrosis factor-α (TNF-α) were from R&D Systems. miHA peptide H60 / MHC-I dimmers were prepared by conjugating H60 peptide to MHC-I dimmers as instructed by the manufacturers (BD Bioscience). We performed immunofluorescence analyses of cell surface phenotypes and intracellular cytokines using FACScan and Canto cytometer (Becton Dickinson) as previously described (23).

For 5-bromo-2′-deoxyuridine (BrdU) incorporation experiments, mice were given sterile drinking water containing 0.8 mg/ml BrdU (Sigma) for 3 days. BrdU labeling was performed as previously described (24). In brief, after surface staining, cells were resuspended in cold 0.15 NaCl, fixed by addition of cold 95% ethanol, incubated for 30 minutes on ice, and washed with PBS. The cells were then fixed using fixation solution from BD Cytofix/Cytoperm™ Kit (BD Bioscience) for 30 minutes, pelleted, and then incubated at 37°C for 30 minutes with 50KU of DNase I (Sigma) in 0.15 NaCl and 4.2 mM MgCl₂, pH5. After washing, cells were stained with FITC-conjugated anti-BrdU for 60 minutes at room temperature, followed by flow cytometry analysis.

Cell preparations

T cell-depleted bone marrow (TCD BM) was prepared by incubating donor BM with microbead-conjugated anti-CD4 Ab and anti-CD8 Ab as previously described (23). CD8⁺ T cells were magnetically isolated from spleens and lymph nodes of mice using microbead-conjugated anti-CD8 Ab (MiniMACS, Miltenyi Biotech). CD8⁺ T cell subsets were further separated using fluorescence activated cell sorter (MoFlo, Beckman Coulter). The purity of each sorted T cell subset was consistently more than 92%. Donor CD8⁺ T cells were labeled with fluorescent dye carboxyfluorescein diacetate succinimidyl ester (CFSE) as described (23). We prepared mature DCs from B6 BM as described (25).

GVHD induction

Mice underwent allogeneic BMT as previously described (23). Briefly, for the C3H.SW anti-B6 mouse GVHD model, we irradiated B6/SJL recipients using a split-dose totaling 10.0Gys from a ¹³⁷Cs source. We mixed donor C3H.SW TCD BM (5 × 10⁶) with or without C3H.SW CD44^loCD62L^hi CD8⁺ naïve T cells (T_N) (2.0 × 10⁶) and transplanted into lethally irradiated B6/SJL recipients (4 to 8 mice per group per experiment). In some experiments, donor C3H.SW CD44^loCD62L^hi CD8⁺T_N (2.0 × 10⁶) were transplanted together with B6/SJL TCD BM (0.5 × 10⁶) into lethally irradiated B6/SJL recipients. In the B6/SJL anti-BALB/b mouse GVHD model, we mixed B6 TCD BM (5 × 10⁶) with or without CFSE-labeled B6/SJL CD44^loCD62L^hi CD4⁺ and CD8⁺ T_N (2.5 × 10⁶ T cells for each) and transplanted into lethally irradiated BALB/b recipients.

Array based mRNA assays

In three repeated experiments, donor alloreactive day-14 CD8⁺ T_MSC and T_E were highly purified, respectively, from 4 B6/SJL mice receiving donor CD44^loCD62L^hi CD8⁺ T_N derived from 6 C3H.SW mice. Donor CD44^loCD62L^hi CD8⁺ T_N were highly purified from pooled CD8⁺ T cells of 2 C3H.SW mice in three separate experiments. Total RNA was prepared from these T cell subsets using TRIzol (Invitrogen Life Technologies, Carlsbad, CA). Biotinylated cDNA was prepared for each sample from 600 ng total RNA using two rounds of reverse-transcription and T7 promoter-based in vitro transcription following hybridization to the arrays. Hybridization, scanning and image analysis of the arrays were performed according to the manufacturer's protocol (Affymetrix, Santa Clara, CA). Mouse Genome 430A 2.0 Arrays containing 22690 probe sets (Affymetrix, Santa Clara, CA) were used to broadly compare the transcription profile of CD8⁺ T_E and T_MSC to that of T_N. The array data are available from Gene Expression Omnibus, accession GSE13743. Using publicly available software (26), we computed trimmed averages of PM-MM differences for each probe set and quantile-normalized these after scaling the arrays to give average probe set values of 1500 units. We then log-transformed using log (max(x+50;0) + 50). Using a 1-way ANOVA model we selected transcripts that gave p<0.01 for comparing pairs of groups that also gave at least a 1.5-fold difference from the means for the paired groups, computed based on differences in the means of log-transformed data. We collapsed probe sets to 13142 distinct genes using Entrez gene IDs. Data can be viewed from the following website: http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=tjunpugmcoqomxc&acc=GSE13743.

Western blot

Samples were separated on a 12% SDS-PAGE gel and transferred onto a nitrocellulose membrane (Millipore Corporation, MA). After an overnight incubation at 4°C with primary Abs, membranes were washed five times and probed with HRP-conjugated secondary Ab (Vector Laboratories, CA) for one hour at room temperature, and immunoreactivity was detected by chemiluminescence (GE Healthcare, NJ).

Real-time reverse transcription polymerase chain reaction (RT-PCR)

The total RNA was extracted from the sorted CD8⁺ T cell subsets using TRIzol (Invitrogen Life Technologies, Carlsbad, CA). Real-time RT-PCR was performed using a SYBR green PCR mix (ABI Biosystem, CA) in the Realplex² Eppendorf Real-time PCR instrument (Eppendorf AG, Westbury, NY). Gene expression levels were calculated relative to the 18S gene. The primer sequences used for real-time RT-PCR include: 18S (5′-GCTGCTGGCACCAGACTT and 3′- CGGCTACCACATCCAAGG), Ifng (5′- ATGAACGCTACACACTGCATC and 3′- CCATCCTTTTGCCAGTTCCTC), Granzyme B (5′- CCACTCTCGACCCTACATGG and 3′- GGCCCCCAAAGTGACATTTATT), Ezh2 (5′- TGCCTCCTGAATGTACTCCAA and 3′- AGGGATGTAGGAAGCAGTCATAC), Tacc3(5′- GAGATGGGGAAGTCCGTTGATG and 3′- CTCTGCTTGGGCCTTGCTGTGT), Birc5 (5′- AACTACCGCATCGCCACCTTC and 3′- TTCTTCCATCTGCTTCTTGACA), Hells (5′- GGGGAGTACCTGGACCTTTTCTTG and 3′- CTGCAGTGTCCCTTGTCTTTTGTG), Pd1(5′- ACCCTGGTCATTCACTTGGG and 3′-CATTTGCTCCCTCTGACACTG), p18 (5′- GTAAACGTCAACGCTCAAAATGG and 3′- GAACCTGGCCAAGTCGAAGG), Casp4 (5′- ACAAACACCCTGACAAACCAC and 3′-CACTGCGTTCAGCATTGTTAAA), and Bcl2 (5′- GTCGCTACCGTCGTGACTTC and 3′-CAGACATGCACCTACCCAGC).

Lentiviral vector construction and viral production

Doxyclycline (Dox) regulated lentiviral vector pLVPT-rtTRKRAB2SM2 (pLVPT^off) was obtained from Addgene (Cambridge, MA)(27). We cloned short-hairpin RNA duplex that specifically targets Ezh2 (Ezh2-shRNA, 5′CGCGTCCCCAAGAGGTTCAGAAGAGCTGTTCAAGAGACAGCTCTTCTGACTGAACCTCTTTTTTTGGAAAT3′) (28) into this pLVPT^off, in which Ezh2-shRNA and GFP are separately driven by H1 promoter and phosphoglycerate kinase (PGK) promoter (named Ezh2-shRNA/GFP-pLVPT^off), as previously described (27). Lentiviral vector encoding scrambled shRNA and GFP was generated as control (named Con-shRNA/GFP-pLVPT^off). In the absence of Dox, Ezh2-shRNA and GFP will be induced, whereas addition of Dox will repress the expression the transcription of both shEzh2 and GFP (27). Production of lentiviruses was done in 293T cells as described (27).

In vivo reconstitution of T cells with inducible knockdown of Ezh2

C-kit⁺ hematopoietic cells were magnetically isolated from B6 mice and infected with Ezh2-shRNA-pLVPT^off in vitro as previously described (29), followed by transplantation into lethally irradiated B6 mice. To repress the expression of Ezh2-shRNA in HSCs during their hematopoietic and thymic reconstitution, all of these recipient mice were given sterile water containing Dox (2mg/ml) from day -2 to 12 weeks after transplantation. HSCs infected with Control-shRNA-pLVPT^off were transplanted as control. Twelve weeks after transplantation, Dox was removed from these mice to induce the expression of Ezh2-shRNA. Seven days later, CD8⁺ T cells were isolated from the spleens and lymph nodes of these mice. GFP⁺CD8⁺ T cells expressing Ezh2-shRNA (named Ezh2-shRNA GFP⁺CD8⁺ T cells) or Control shRNA (named Control-shRNA GFP⁺CD8⁺ T cells) were sorted using the BD FACSAria II cell sorter (BD, Bioscience).

Ex vivo stimulation of CD8⁺ T cells

Sorted Ezh2-shRNA GFP⁺CD8⁺ T_N and Control-shRNA GFP⁺CD8⁺ T_N were stimulated with anti-CD3 Ab and anti-CD28 Ab (2.5μg/ml for each) in 96 well plate as previously described (14-15, 30). In some experiments, unfractionated splenic mononuclear cells that contained Ezh2-shRNA GFP⁺CD8⁺ T cells were cultured in the presence of allogeneic DCs or IL-7. The recovery number of GFP⁺CD8⁺ T cells was assessed by flow cytometry analysis.

Statistical analysis

Survival in different groups was compared by using the log-rank analysis. Comparison of two means was analyzed using the two-tailed unpaired Student t test. Statistical analysis from the gene array is indicated in the text.

Results

Alloantigen-induced continuous proliferation is essential to maintaining chronically activated alloreactive CD8⁺ T_E

Since CD8⁺ T_E are known to be “terminally differentiated” and short-lived cells (1-6), we first determined under what conditions alloreactive CD8⁺ T_E were able to persist in vivo and cause GVHD. We transplanted CFSE-labeled donor CD8⁺ T_N from normal C3H.SW mice together with B6/SJL TCD BM into lethally irradiated B6/SJL recipients. This allowed us to strictly track the fate of infused donor mature T cells while inducing GVHD (15). By day 14 after transplantation, donor alloreactive CD8⁺ T_E became the dominant population (>80%) (named day-14 CD8⁺ T_E, Fig.1A). They had undergone extensive division and produced significantly higher levels of IFN-γ and Granzyme B (Gzmb) (Fig.1A and B). When later adoptively transferred into lethally irradiated secondary allogeneic B6/SJL mice, all the day-14 CD8⁺ T_E had extensively divided in vivo, expressed high levels of proliferating antigen Ki67 seven days after transplantation (Fig.1C). BrdU incorporation analysis showed that about 22% of day 14 CD8⁺ T_E derived from primary GVHD recipient mice and 76% of alloreactive CD8⁺ T_E recovered from secondary recipients of day 14 CD8⁺ T_E had incorporated BrdU (Fig. 1D). These results suggest that a substantial proportion of day 14 CD8⁺ T_E are dividing during GVH reaction.

Fig.1.

Alloreactive CD8⁺ T_E are highly replicating cells causing GVHD. (A) CFSE-labeled donor CD8⁺ T_N (2×10⁶) derived from normal C3H.SW mice were transplanted with B6/SJL TCD BM (0.5×10⁶) into lethally irradiated B6/SJL mice. Donor CD8⁺ T cells were recovered at day 14 after transplantation from the spleen and lymph node of 3 recipients, numerated and stained for flow cytometric analysis. Histograms show the cell division based on CFSE dilution and production of IFN-γ by CD8⁺ T_MSC, T_CM and T_E. Data are representative of three independent experiments. (B) Real-time RT-PCR analysis shows the relative mRNA expression of selected genes in donor day-14 CD8⁺ T_E, T_CM and T_MSC. Data are representative of three independent experiments. (C and D) Day-14 CD8⁺ T_E were re-labeled with CFSE and adoptively transferred into lethally irradiated secondary B6/SJL mice. CFSE-labeled C3H.SW CD8⁺ T_N were transferred as controls. Donor T cells were recovered at day 7 after adoptive transfer, numerated and analyzed for flow cytometry. Histogram shows the cell division of indicated cells. Non-stimulated CFSE-labeled CD8⁺ T_N were used as non-dividing cell control. Contour plots show the expression of Ki67 (C). For BrdU labeling, secondary recipient mice were given drinking water containing BrdU for 3 days prior to the end of the experiments. At day 7, donor cells were recovered, stained with BrdU, and analyzed by flow cytometry. Dot plots show the labeling of BrdU in gated CD8⁺ T_E population (Mean ± SD) (D). (E and F) As described above in (C), the number of donor T cells recovered from these secondary recipients (n=3 for each group) was calculated (E) and the percentage of apoptotic cell was shown (F). Data are representative of two independent experiments. (G) Donor C3H.SW TCD BM (5×10⁶) were transplanted alone, or with donor day-14 CD8⁺ T_E (0.5×10⁶) and donor CD8⁺ T_N (1.0×10⁶) into lethally irradiated secondary B6/SJL mice. Survival of animals was analyzed with the Kaplan-Meier method. *, P<0.05, significant difference.

Interestingly, 4-fold fewer donor T cells were recovered from secondary recipients of day-14 CD8⁺ T_E than that of donor CD8⁺ T_N (Fig.1E). This was associated with significantly increased apoptotic death of donor T cells in day-14 CD8⁺ T_E recipients as compared to CD8⁺ T_N recipients (Fig.1F). Furthermore, adoptive transfer of these day-14 CD8⁺ T_E caused GVHD in secondary recipients, with 70% of them dying from the disease by day 75 after transplantation (Fig.1G). Thus, upon chronic exposure to alloantigens, alloreactive CD8⁺ T_E continuously proliferated to persist while undergoing increased apoptotic death. However, all alloreactive CD8⁺ T_E diminished in vivo without causing GVHD when adoptively transferred into secondary B6.B₂M^-/- mice (data not shown). These results suggest that allogeneic stimuli could be essential to sustaining alloreactive CD8⁺ T_E.

In lethally irradiated allogeneic recipient mice, both alloantigens and lymphopenia-related homeostatic factors (e.g., self-antigen peptides and cytokines IL7 and IL-15) (31) could be responsible for the proliferation of alloreactive CD8⁺ T_E. To assess the possible impact of alloantigens versus homeostatic factors in regulating the persistence of alloreactive T_E, we used the CD4⁺ T cell-dependent B6 anti-BALB/b GVHD model, in which miHA H60-specific (H60⁺) CD8⁺ T_E could be tracked using miHA peptide/MHC class-I dimmer staining (Fig.2A)(14, 32-33). We highly purified H60⁺CD8⁺ T_E at day 12 post-transplantation from GVHD BALB/b mice receiving donor B6/SJL T_N (CD45.1), labeled with CFSE, and adoptively transferred together with B6/SJL CD4⁺ T_E and donor B6 TCD BM into lethally irradiated congenic B6 (CD45.2⁺) mice and allogeneic BALB/b mice. Lethal irradiation of congenic B6 mice creates a lymphopenic environment that might induce homeostatic proliferation and survival of adoptively transferred day 12 B6/SJL CD8⁺ T cells in the absence of miHA H60, whereas lethally irradiated allogeneic BALB/b mice could provide persistent alloantigen-stimulation in addition to homeostatic factors.

Fig.2.

Proliferation and persistence of alloantigen-specific CD8⁺ T_E depend on the presence of chronic alloantigen. (A) CFSE-labeled donor T cells (CD4⁺ and CD8⁺ T cells, 3×10⁶ for each) from normal B6/SJL mice (CD45.1) were transplanted, together with B6 TCD BM (CD45.2, 5×10⁶), into lethally irradiated (10Gys) primary BALB/b recipients (CD45.2). Dot plots show percent of host miHA H60-specific CD8⁺ T_E that were recovered at day 12 after transplantation from spleens and livers of primary GVHD BALB/b recipients (n=4) of B6/SJLTCD BM and B6/SJL T cells. Data are representative of three independent experiments. (B, C and D) Donor alloreactive B6/SJL H60⁺CD8⁺ T_E (5×10⁴) and CD4⁺ T cells (3×10⁵) were isolated 12 days after transplantation, respectively, from these primary GVHD BALB/b recipients (CD45.2) of B6/JL T cells, re-labeled with CFSE, and adoptively transferred together with B6 TCD BM (5×10⁶) into lethally irradiated secondary BALB/b recipients and congenic B6 recipients (CD45.2), respectively. The percentage of donor alloreactive H60⁺CD8⁺ T_E in the peripheral blood of these secondary recipients at day 14, 28 and 43 after adoptive transfer (B). Dot plots show the fraction of H60⁺CD8⁺ T_E in the peripheral blood (C). Donor alloreactive H60⁺CD8⁺ T_E recovered at day 43 from secondary recipients were calculated (D). Data are presented as mean ± SD. * P<0.01. (E) Histograms show the surface markers of donor H60⁺CD8⁺ T_E isolated from the secondary allogeneic BALB/b mice (left), and dot plots show the production of IFN-γ and GZMB (right). Data shown in B, C, D and E are representative of two independent experiments. (F) Survival rate of animals was analyzed by the Kaplan-Meier Method. *, P<0.05, significant difference.

We found that donor H60⁺CD8⁺ T_E were readily detected in the peripheral blood of secondary allogeneic BALB/b recipients (43.3±19%) by day 14 after transplantation, but not in that of secondary congenic B6 recipients (0.1±0.3%) (Fig.2B and C). Forty-three days after transfer, allogeneic BALB/b recipients showed about 80-fold more proliferating H60⁺CD8⁺ T cells than B6 congenic recipients (Fig.2D). These donor H60⁺CD8⁺ T_E from secondary allogeneic BALB/b recipients expressed high levels of CD69, suggesting a recent antigenic stimulation (Fig.2E). Furthermore, they produced high levels of IFN-γ and Granzyme B (Fig.2E) and caused GVHD in these secondary allogeneic BALB/b mice (Fig.2F). In separate experiments, we observed that adoptive transfer of donor alloreactive CD8⁺ T_E alone also caused GVHD in secondary allogeneic BALB/b mice (Fig.2F), suggesting that CD4-help is not essential to already differentiated CD8⁺ T_E to mediate GVHD. None of these congenic B6 mice receiving donor alloreactive CD8⁺ T_E developed clinical signs of GVHD (data not shown). Thus, alloantigenic stimuli rather than homeostatic factors are critical to the continual proliferation and persistence of alloreactive CD8⁺ T_E during GVH reaction.

Gene expression profiles of alloreactive CD8⁺ T_E

To understand the molecular mechanisms by which alloreactive CD8⁺ T_E acquire the ability to continually proliferate upon chronic exposure to alloantigens, we used Affymetrix Mouse Genome A430A 2.0 Array to broadly compare the gene expression profiles of day 14 CD8⁺ T_E to that of CD8⁺ T_N and CD8⁺ T_MSC (Fig.3A). Compared to T_N, a total of 2744 distinct genes were up-regulated (1359) or down-regulated (1385) in CD8⁺ T_E by a 1.5 fold that gave a p<0.01 value for comparing pairs of groups. As predicted by our experimental data (14-15) and others (12, 34), relative to CD8⁺ T_N and T_MSC, CD8⁺ T_E expressed significantly higher levels of genes known to be important for effector functions, including effector molecules, chemokines and chemokine receptors (Table 1). As expected, many other genes engaged in TCR signaling pathway, glycolysis, MAPK pathway and cell adhesion were also altered in CD8⁺ T_E (Table 1).

Fig.3.

Array based mRNA arrays of alloreactive CD8⁺ T cells. (A) Donor CD44^loCD62L^hi CD8⁺ T_N (2×10⁶) from C3H.SW mice (CD45.2) were transplanted with B6/SJL TCD BM(0.5×10⁶) into lethally irradiated B6/SJL mice (CD45.1). Donor CD8⁺ T cells were recovered at day 14 after transplantation from the spleens and lymph nodes of these recipients, magnetically purified and stained with indicated Abs for flow cytometry cell sorting. Affymetrix Mouse Genome 430A 2.0 Array was used to broadly compare the transcription profiles of these CD8⁺ T_MSC and T_E to that of T_N cells. (B) Real-time RT-PCT analysis shows the relative expression of mRNA in each T cell subset. Data are representative of three independent preparations of alloreactive T cells.

Table 1. Altered transcripts in T_MSC and T_E as compared to T_N.

Gene symbols	Gene ID#	Fold Change		Gene symbols	Gene ID#	Fold Change		Gene symbols	Gene ID#	Fold Change
		TMSC/TN	TE/TN			TMSC/TN	TE/TN			TMSC/TN	TE/TN
Effector function				Glycolysis / Gluconeogenesis				mTOR pathway
Gzmb	14939	38.3	306.4	Hk2	15277	2	7.1	Rps6kb2	58988		2.7
Ifng	15978	4.2	211	Tpi1	21991	1.9	4.1	Akt1	11651		2.1
Fasl	14103		66.3	Eno3	13808		4	Eif4e2	26987		2
Klrk1	27007	2.9	3.3	Gapdh	14433	1.5	3.8	Rps6ka1	20111		1.7
Prf1	18646		3	Acyp2	75572	1.8	2.4	Eif4b	75705		-1.9
Tnfsf10	22035		2.7	Pfkp	56421	1.5	1.9	Tsc1	64930		-2
Klra3	16634		2.3	Bpgm	12183		1.6	Rps6ka2	20112		-2.9
Chemokine-chemokine receptor interactions				MAPK signaling pathway				Adhesion and trans-endothelial migration
Ccl5	20304	15.9	453.5	Stmn1	16765	5	19.3	Cd44	12505		14.2
Ccl4	20303		69.6	Dusp1	19252	1.8	4.9	Hmmr	15366		7.2
Ccl3	20302		42	Map3k8	26410	7.6	3.4	Itgb1	16412		5.6
Ccr5	12774	4.7	32.8	Gadd45b	17873		3.2	Spp1	20750		4.5
Ccr2	12772		10.8	Gadd45g	23882	15.3	2.7	Itga4	16401		2.7
Cxcr3	12766	4.1	8.3	Dusp3	72349	1.8	2.5	Sdc4	20971		2.5
Cx3cr1	13051		4.8	Mapkapk2	17164		1.8	Itgav	16410		2.4
Xcl1	16963		4.7	Mapk3	26417		1.8	Lamc1	226519		2
Cxcl10	15945		3.1	Map2k3	26397		1.7	Itga3	16400		2
Cxcl10	15945		3.1	Mapk8	26419	-1.6		Npnt	114249		1.9
Ccr6	12458		2.9	Mapk14	26416	-1.6		Itgb2	16414		1.9
Ccl6	20305		2.7	Map2k6	26399	-1.6		Vcl	22330		1.7
Ccr7	12775		-29.1	Dusp2	13537	-1.6		Tnc	21923		1.6
Cytokine-cytokine receptor interactions				Jak-STAT signaling pathway				Pxn	19303		1.5
Il10ra	16154	1.7	9.9	Pim1	18712	2.3	2.3	Apoptosis
Csf1	12977	-2	6.4	Jak3	16453	1.8	1.9	Casp3	12367		5.4
Il2rb	16185	1.7	6.3	Socs5	56468	-1.7	-1.7	Capn2	12334		2.3
Il10	16153		6.1	Stat6	20852		-1.7	Bid	12122		2.2
Csf2	12981		4.7	Stat1	20846	2	-1.8	Apaf1	11783		2.2
Il15ra	16169	2	2.5	Jak1	16451	-1.5	-1.8	Casp7	12369		2.1
Il12rb1	16161	1.9	2.5	Socs3	12702	2.6	-2.4	Bax	12028		1.8
Il4ra	16190		-2.8	Socs1	12703	1.9		Bcl2l1	12048		1.6
Il6st	16195	-1.6	-4.3	Myc	17869	1.5		Bad	12015		1.5
Il7r	16197	-1.9	-7.3	Toll-like receptor signaling pathway				Bcl2	12043	2	-1.5
Il6ra	16194		-18.6	Tlr7	170743	2.6		Trp53	22059	1.9
Il10rb	16155	1.6		Tlr6	21899	1.5
TCR signaling				TCR signaling				TCR signaling
Tbx21	57765	8.8	21.5	Cdc42	12540		1.6	Ikbkb	16150		-2.5
Map3k8	26410	7.6	3.4	Hras1	15461	1.5	1.5	Pak1	18479		-5.8
Nfatc1	18018		2.6	Ikbkg	16151		-1.5	Pdk1	228026		-6.3
Fyn	14360		2.1	Itk	16428	-1.5	-2	Jun	16476	6.1
Ppp3cc	19057	1.6	1.7	Map3k14	53859	-1.5	-2.2	Fos	14281	4
Zap70	22637		1.7	Tec	21682		-2.5	Lcp2	16822	-1.6

To better define a specific gene signature of alloreactive T cells, we further tested each list of distinct genes for over-representation in 507 lists of Gene Ontology (GO) terms for which there were at least 10 genes present on the array (Affymetrix web site), as well as for 190 lists of pathways from the Kyoto Encyclopedia of Genes and Genomes (KEGG, http://www.genome.jp/kegg/). Compared to both CD8⁺ T_N and T_MSC, CD8⁺ T_E demonstrated a signature expression of genes related to cell cycle, mitosis, apoptosis, and transcription and translation (Table 2).

Table 2. Most significantly enriched GO terms and KEGG pathways.

A) Tests of 507 Geneontology (GO) biological process terms with at least 10 genes present on the arrays.
Comparison	direction	number of probe sets selected (of 22690)	number of distinct genes selected (of 13142)	GO term title	number of genes with GO term on the array	number of those genes selected	p-value
Tmsc vs Tn	up in Tmsc	672	543	cell cycle	335	41	4.1E-10
				immune response	263	29	1.4E-06
				DNA replication	96	16	1.8E-06
				mitosis	109	17	2.2E-06
				cell division	173	22	2.6E-06
				regulation of progression through cell cycle	185	20	8.1E-05
Tmsc vs Tn	up in Tn	545	448	transcription	1125	72	1.1E-07
				regulation of transcription, DNA-dependent	1486	86	4.6E-07
Te vs Tmsc	up in Te	1486	1098	cell cycle	335	102	1.5E-32
				mitosis	109	53	1.1E-28
				cell division	173	66	1.1E-27
				DNA replication	96	40	6.7E-19
				apoptosis	296	54	3.1E-08
				DNA replication initiation	13	9	1.0E-07
Te vs Tmsc	up in Tmsc	1372	1036	ribosome biogenesis and assembly	47	21	9.6E-12
				translation	259	51	7.5E-10
Te vs Tn	up in Te	1817	1359	cell cycle	335	109	2.6E-29
				mitosis	109	52	2.9E-23
				cell division	173	67	3.5E-23
				DNA replication	96	41	1.8E-16
				DNA replication initiation	13	10	2.9E-08
				apoptosis	296	59	5.1E-07
				DNA recombination	35	13	2.4E-05
				regulation of progression through cell cycle	185	38	2.6E-05
Te vs Tn	up in Tn	1837	1359	ribosome biogenesis and assembly	47	18	5.3E-07
				translation	259	49	3.3E-05
				transcription	1125	158	6.9E-05

B) Tests of 190 KEGG pathways
Comparison	direction	number of probe sets selected (of 22690)	number of distinct genes selected (of 13142)	Pathway Title	number of genes in pathway on the array	number of those genes selected	p-value
Tmsc vs Tn	up in Tmsc	672	543	Cell cycle	102	16	4.1E-06
				Pyrimidine metabolism	80	12	1.0E-04
				Jak-STAT signaling pathway	122	14	5.0E-04
Tmsc vs Tn	up in Tn	545	448	Wnt signaling pathway	133	11	5.8E-03

We fit a 1-way ANOVA model to the data for each probe set and used it to compare each pair of groups. We selected those probe sets that gave p-values of p<.01 and average fold differences of at least 1.5-fold as being differentially expressed. We estimated the false discovery rate (FDR) by performing an identical analysis of all data sets with permuted sample labels that did not recapitulate the actual groupings of the 3 groups of samples, and dividing the average number of probe sets selected in these permuted data sets by the actual number of probe sets selected for the original data, and obtained estimated FDRs of 7.5% for T_MSC vs. T_N, 3.1% for T_E vs. T_MSC, and 2.5% for T_E vs. T_N. The number of probe sets selected as being increased or decreased is shown, as well as the number of distinct genes these represented, determined using Entrez Gene identifiers. The mapping of probe sets to genes, as well as the Gene Ontology terms for each gene, were obtained from Affymetrix web site, and were dated July 11, 2007. We tested each list of distinct genes for over-representation in 507 lists of Gene Ontology terms for which there were at least 10 genes present on the array, as well as for 190 lists of pathways from the Kyoto Encyclopedia of Genes and Genomes, obtained Jun 12, 2007 (http://www.genome.jp/kegg/). We present only the most significant lists obtained in each category, and of the most significant lists present all of them (that is, none are hidden). The total number of genes in each list that were on the array is given as well as the number of those that were selected as altered in each pair-wise comparison of groups of samples. The p-values shown are from using these counts to perform one-sided Fisher's Exact tests, however, since many lists were tested, we also give the average number of lists that obtained a p-value this smaller or smaller in 100 analyses of the same data in which the gene labels were randomly permuted.

Detailed analysis revealed that alloreactive CD8⁺ T_E showed decreased expression of anti-apoptotic gene Bcl2, but had increased genes related to apoptosis, including Pdcd1 (also named Pd1), Klrg1, Casp1, Casp3, Caps4, Casp7, Bax, and Bad (Table 1). In contrast, CD8⁺ T_MSC expressed higher levels of Bcl2 than both T_N and T_E, while only minimally changing in the expression of other pro-apoptotic genes (Table 1 and Fig.3B). When compared to CD8⁺ T_N, CD8⁺ T_E show significantly decreased expression of genes related to ribosome biogenesis and assembly, ribosome, translation and transcription (Table 2). Real time RT-PCR validated the expression some of these genes (Fig.3B).

Notably, alloreactive CD8⁺ T_E also had 150-fold more expression of p18^Ink4c (encoded by Cdkn2c) than CD8⁺ T_N and 15-fold than CD8⁺ T_MSC (Fig.3B). Previous studies have shown that p18^Ink4c plays a critical role in negatively regulating cell proliferation and survival (35-38). The loss of p18^Ink4c in T cells causes their hyperproliferation response and proliferation disorder (38-39).

Altogether, this transcriptional signature confirmed that alloreactive CD8⁺ T_E are replicating cells, but are more likely susceptible than CD8⁺ T_MSC to apoptotic death and senescence, consistent with our previous observations (15).

Alloreactive CD8⁺ T_E activate stem cell transcriptional programs

Next we tested 1687 curated gene lists from Molecular Signature Database v2 (MSigDB v2)(40) to inquire which transcriptional program(s) could be associated with the continuous proliferation property of alloreactive CD8⁺ T_E. Based on one-sided Fisher's exact test, we identified that transcripts increased in CD8⁺ T_E were significantly enriched in lists of genes increased in NSCs (345 of 1636 genes, p=7.3 × 10^-44) and ESCs (227 of 1193 genes, p=1.7 × 10^-21) (Fig.4A), which were identified by Ramalho-Santos et al. (41). These alloreactive CD8⁺ T_E-related ESC genes and NSC genes are listed in Table 3 and Table 4. Among them, 171 genes were shared by ESCs and NSCs, 56 appeared in ESCs, and 174 were found only in NSCs. Thus, 401 out of 1369 (29%) of transcripts that were increased in alloreactive CD8⁺ T_E were enriched for ESCs and/or NSCs (Fig.4A). In contrast, transcripts decreased in CD8⁺ T_E were over-represented among HSC-related genes (253 of 1279, 19.8%, p=2.2×10^-25) (Fig.4A).

Fig.4.

Acquisition of stem cell transcriptional programs in alloreactive CD8⁺ T cells. (A) Genes distinctively expressed by CD8⁺ T_MSC and T_E relative to T_N were analyzed for functional set enrichment analysis using curated gene lists from the Ramalho-Santos's array data. (B) Acquisition of stem cell transcriptional programs in P14 LCMV-gp33-specific CD8⁺ T cells. Genes differentially expressed by CD8⁺ T_E-KLRG1^Hi, CD8⁺ T_E-KLRG1^Hint and CD8⁺ memory T cells relative to T_N were analyzed for enrichment for the same stem cell gene lists. Sizes of stem cell gene lists differ slightly from that in Fig.4A since the Sarkar's study used Affymetrix Mouse_430_2, which also had slightly different probe-set annotation than our arrays.

Table 3.

ESC-related genes that are activated in CD8⁺ T_E

Gene symbol	Gene ID#	Fold Change		Gene symbol	Gene ID#	Fold Change		Gene symbol	Gene ID#	Fold Change		Gene symbol	Gene ID#	Fold Change		Gene symbol	Gene ID#	Fold Change
		TMSC/TN	TE/TN			TMSC/TN	TE/TN			TMSC/TN	TE/TN			TMSC/TN	TE/TN			TMSC/TN	TE/TN
Cell cycle				Ubiquitin/Proteolysis/Proteasome				Transcription/Transcription regulation				Receptor/Signal transduction				Microtubule-based process
Ccnb1	268697	2.28	16.85	Cdc20	107995		5.27	Mdfic	16543	2.77	18.58	Spp1	20750		4.52	Prc1	233406	1.86	12.05
Sgol1	72415		11.04	Capn2	12334		2.34	Prdm1	12142		10.62	Aurka	20878		4.47	Cks2	66197	4.85	11.65
Chek1	12649	2.6	10.34	Xpnpep1	170750		2.32	Tacc3	21335		6.06	Exo1	26909		4.02	Anln	68743		4.82
Cks1b	54124	2.58	9.51	Ide	15925		2.28	Csda	56449		5.52	Hif1a	15251		3.71	Dstn	56431		1.74
Cdkn1a	12575		9.17	Coq3	230027	1.67	2.12	Cdkn2b	12579		4.89	Csrp2	13008		3.01	Cytokinesis
Plk1	18817		7.54	Psma1	26440		1.99	Plagl1	22634		3.25	Sap30	60406		2.5	Igfbp7	29817		2.68
Aspm	12316		7.53	Psmb2	26445	1.55	1.78	Sub1	20024		2.62	Dusp16	70686		2.3	Wisp1	22402		2.26
Ect2	13605		6.86	Psmc2	19181		1.63	Myef2	17876		2.09	Rap2a	76108		2.27	Miscellaneous
Itgb1	16412		5.58	Psmb5	19173		1.62	Gtf2e2	68153		1.99	Bax	12028		1.84	Fignl1	60530	3.2	15.73
Ccne1	12447		5.31	Psmd6	66413		1.57	Tceb1	67923		1.95	Pex7	18634	1.6	1.74	Trip13	69716	1.88	8.69
Syce2	71846		5.16	Psma3	19167		1.55	Polr3k	67005		1.77	Bdnf	12064		1.72	Pcbp4	59092		6.39
Cdc25c	12532		4.72	Psma5	26442		1.55	Bzw1	66882		1.68	Gnb1	14688		1.69	Tmem49	75909	2.26	6.27
Ccnf	12449		3.96	Metabolism				Cnot7	18983		1.63	Hnrpab	15384		1.65	Tacstd1	17075		5.17
Dlg7	218977		2.72	Gldc	104174		23.94	Klf9	16601		1.6	Ddx1	104721		1.65	Smoc2	64074		4.05
Cdkn2a	12578		2.55	Gsto1	14873	2.29	7.34	Mtdh	67154		1.57	Inhbb	16324		1.64	Synpo	104027	1.75	3.68
Mybl2	17865		2.43	Bcat1	12035		5.54	Nfyb	18045		1.52	Nasp	50927		1.61	Tipin	66131		3.5
Ran	19384		1.89	Myo5a	17918	1.55	4.68	RNA processing				Sfrs1	110809		1.61	Ccdc99	70385		2.81
Ccng1	12450		1.84	Asns	27053	2.95	3.87	Syncrip	56403	1.79	2.88	Snrpd1	20641		1.55	Lactb2	212442	1.64	2.68
Erh	13877	1.67	1.74	Gcat	26912		2.71	Hnrpll	72692		2.73	Hrb	15463		1.55	Gmfb	63985	1.67	2.64
Psmd1	70247		1.69	Shmt1	20425	1.53	2.55	Ell2	192657		2.42	Rab1	19324		1.53	Wbp5	22381		2.42
Cops5	26754		1.64	Uxs1	67883		2.48	Ncbp2	68092	1.9	2.01	Vdac3	22335		1.52	Errfi1	74155	1.54	2.4
DNA replication/Repair				Hmgcr	15357		1.96	Snrpa1	68981		1.91	Transport (Protein/Ion)				Ccdc80	67896		2.34
Rrm2	20135	9.19	60.07	Pfkfb1	18639		1.8	Cstf3	228410		1.84	Kif20a	19348		4.26	Nucks1	98415		2.3
Cdc6	23834	5.09	21.54	Lypla1	18777		1.79	Sfrs3	20383		1.71	Arl5a	75423		3.8	Mpdu1	24070	1.85	2.29
Il1rl1	17082		19.15	Ywhah	22629		1.74	Cops2	12848		1.7	Snx10	71982	1.97	2.74	Ipp	16351		2.25
Dtl	76843	4.19	18.32	Acot9	56360		1.66	Lsm5	66373		1.7	Clic4	29876	2.31	2.56	Pls3	102866		2.16
Rad51	19361	2.92	15.71	Soat1	20652		1.65	Sfrs2	20382		1.65	Slc35b1	110172		2.15	Phlda3	27280		2.1
Pttg1	30939		8.1	Nup93	71805		1.62	Translation				Arl1	104303		2.07	Dsp	109620		1.96
Dna2l	327762	1.68	7.48	Kpna3	16648		1.56	Farsb	23874		4.68	Lman1	70361		1.98	Prdx1	18477		1.91
Brca1	12189	2.04	5.87	Electron transport				Mrpl27	94064	1.84	2.41	Vps54	245944		1.94	Tmem41b	233724		1.83
Smc2	14211		3.55	Nqo2	18105		3.3	Eif4e2	26987	1.52	2.04	Mtch2	56428		1.86	Rcn1	19672	1.6	1.77
Mcm2	17216	1.56	3.4	Etfb	110826		3.11	Eif1ay	66235		1.86	Snx5	69178	1.77	1.85	Esd	13885		1.73
Rfc5	72151		2.7	Pdia6	71853	1.6	2.42	Mrpl18	67681		1.69	Clcn5	12728		1.85	Psmc3ip	19183		1.71
Mcm3	17215		2.7	Gsr	14782		2.06	Eif3s1	78655		1.63	Slc2a3	20527		1.83	Ranbp1	19385		1.67
Mcm4	17217		2.59	Cyp51	13121		1.94	Mrpl16	94063		1.52	Nup62	18226		1.78	Nup85	445007		1.62
Nudt1	17766	1.66	2.53	Uqcrb	67530		1.92	Protein phosphorylation				Slc33a1	11416		1.72	Ppp1r7	66385		1.62
Orc6l	56452		2.44	Txnl1	53382		1.81	Melk	17279	4.05	17.84	Golt1b	66964		1.7	Psmd7	17463	1.84	1.54
Rfc4	106344		2.4	Sqle	20775		1.69	Bub1	12235		7.69	Slc4a7	218756		1.69	Cox4nb	18117		1.54
Rpa3	68240		2.22	Txndc4	76299		1.6	Wee1	22390	2.07	6.06	Bcap29	12033	1.61	1.65	Asah1	11886		1.53
Dtymk	21915	1.65	2.21	Txnl5	52700		1.6	Ryk	20187		5.21	Slc12a2	20496		1.57	Uck2	80914	1.74	1.52
Hat1	107435		2.15	Chromatin modification/Assembly				Plk4	20873	1.54	4.75	Inhibitorysignal/Apoptosis				Tiprl	226591		1.51
Brca2	12190		2.01	Nusap1	108907	1.86	9.22	Ttk	22137		3.62	Ckap2	80986		5.44
Blm	12144		1.97	Mad2l1	56150	1.61	6.8	Mapk6	50772		2.07	Gas2	14453		4.73
Topbp1	235559		1.96	Tmem38b	52076		2.7	Chek2	50883		2.07	Cell adhesion/Communication
Ranbp5	70572		1.72	Nek2	18005		2.45	Ppap2a	19012		2.05	Adam19	11492		4.89
Dbf4	27214		1.69	Suv39h2	64707		2.04	Pigf	18701		1.61	Adam9	11502		2.82
Ube2n	93765		1.64	Cbx1	12412		1.94	Acp1	11431		1.56	Tnc	21923	1.94	1.56
Nudt5	53893		1.52	Nap1l1	53605		1.65	Ptprk	19272		1.54	Vcl	22330		1.66
				Smarcad1	13990		1.56

Table 4. NSC-related genes that are activated in CD8⁺ T_E.

Gene symbol	Gene ID#	Fold Change		Gene symbol	Gene ID#	Fold Change		Gene symbol	Gene ID#	Fold Change		Gene symbol	Gene ID#	Fold Change		Gene symbol	Gene ID#	Fold Change
		TMSC/TN	TE/TN			TMSC/TN	TE/TN			TMSC/TN	TE/TN			TMSC/TN	TE/TN			TMSC/TN	TE/TN
Cell cycle				DNA replication/Repair				Receptor/Signal transduction				Transport (Protein/Ion)				Transcription/Transcription regulation
Ccnb2	12442	3.64	29.94	Rrm2	20135	9.19	60.07	Epas1	13819		7.54	Sypl	19027		4.33	Mdfic	16543	2.77	18.58
Ccna2	12428	3.31	21.13	Cdc6	23834	5.09	21.54	Rhoc	11853		6.85	Kif20a	19348		4.26	Uhrf1	18140	2.89	9.26
Cdca5	67849	4.46	19.44	Dtl	76843	4.19	18.32	Rhoq	104215	1.69	5.32	Syt11	229521	2.9	3.88	Hells	15201	1.91	7
Ccnb1	268697	2.28	16.85	Uhrf1	18140	2.89	9.26	Arl5a	75423		3.8	Snx10	71982	1.97	2.74	Ezh2	14056		6.95
Sgol1	72415		11.04	Dna2l	327762	1.68	7.48	Hif1a	15251		3.71	Slc35b2	73836		2.61	Tacc3	21335		6.06
E2f8	108961		10.39	Lig1	16881		5.73	Nudt1	17766	1.66	2.53	Clic4	29876	2.31	2.56	Eomes	13813	1.61	5.7
Chek1	12649	2.6	10.34	Rpa2	19891		4.52	Sap30	60406		2.5	Pdzd11	72621	1.66	2.25	Csda	56449		5.52
Cks1b	54124	2.58	9.51	Rfc3	69263	1.6	4.27	Dusp16	70686		2.3	Slc35b1	110172		2.15	Cenpk	60411		4.51
Cdkn1a	12575		9.17	Rrm1	20133		4.18	Dtymk	21915	1.65	2.21	Dbi	13167		2.11	Nsbp1	50887		3.85
Plk1	18817		7.54	Mcm5	17218		4.07	Gpr56	14766		2.08	Lman1	70361		1.98	Klf10	21847		2.98
Aspm	12316		7.53	Smc2	14211		3.55	Akt1	11651		2.07	Vps54	245944		1.94	Wdhd1	218973		2.85
Ect2	13605		6.86	Fen1	14156		3.43	Arl1	104303		2.07	Gipc1	67903	1.55	1.93	Sub1	20024		2.62
Cdkn2c	12580	2.13	6.27	Prim1	19075		3.42	Arhgap21	71435		2	Arf4	11843		1.87	Pmf1	67037		2.45
Itgb1	16412		5.58	Mcm2	17216	1.56	3.4	Tank	21353		1.95	Mtch2	56428		1.86	Myef2	17876		2.09
Gmnn	57441	2.1	5.33	Hmgn2	15331		3.37	Fut8	53618		1.91	Snx5	69178	1.77	1.85	Rab8b	235442		2.09
Cdc45l	12544	1.68	5.13	Mcm7	17220		2.94	Rbl1	19650		1.83	Clcn5	12728		1.85	Gtf2e2	68153		1.99
Cdkn2b	12579		4.89	Prim2	19076		2.76	Tmpo	21917		1.82	Vps29	56433		1.83	Inppl1	16332		1.98
Cdc25c	12532		4.72	Rfc5	72151		2.7	Ptprj	19271		1.74	Rrbp1	81910		1.8	Cdca7l	217946	1.8	1.97
Pola1	18968	1.59	4.55	Mcm3	17215		2.7	Csnk2b	13001		1.73	Trappc1	245828		1.79	Tceb1	67923		1.95
Aurka	20878		4.47	Mcm4	17217		2.59	Ranbp5	70572		1.72	Ssr2	66256		1.73	Hnrpdl	50926	1.72	1.78
Ccnf	12449		3.96	Orc6l	56452		2.44	Gnb1	14688		1.69	Slc33a1	11416		1.72	Polr3k	67005		1.77
Gsg2	14841	1.75	3.11	Rfc4	106344		2.4	Dck	13178		1.66	Vps45	22365		1.72	Bzw1	66882		1.68
Dlg7	218977		2.72	Recql	19691	1.71	2.28	Ddx1	104721		1.65	Atp6v0a2	21871		1.71	Cnot7	18983		1.63
Nfatc1	18018		2.51	Rpa3	68240		2.22	Inhbb	16324		1.64	Slc4a7	218756		1.69	Sin3b	20467		1.63
Ran	19384		1.89	Smc6	67241		2.15	Tyms	22171		1.63	Slc39a6	106957		1.65	Mtdh	67154		1.57
Ccng1	12450		1.84	Hat1	107435		2.15	Aaas	223921		1.59	Bcap29	12033	1.61	1.65	Th1l	57314		1.53
Erh	13877	1.67	1.74	Blm	12144		1.97	Smap1	98366		1.58	Nup93	71805		1.62	Microtubule-based process
Rb1	19645		1.72	Topbp1	235559		1.96	Rab1	19324		1.53	Sfxn1	14057	1.7	1.57	Kif2c	73804	3.23	20.24
Psmd1	70247		1.69	Hmgb1	15289		1.82	Asna1	56495		1.53	Timm17b	21855		1.57	Kif11	16551	1.82	7.9
Cops5	26754		1.64	Ube2n	93765		1.64	Nudt5	53893		1.52	Kpna3	16648		1.56	Kif22	110033	2.33	7.45
Mapre1	13589		1.6	Smarcad1	13990		1.57	Cytokinesis				Slc25a10	27376		1.56	Kif4	16571		3.58
Rap1a	109905		1.53					Prc1	233406	1.86	12.05	Kdelr2	66913		1.56	Mid1ip1	68041		2.07
								Cks2	66197	4.85	11.65	Sec14l1	74136		1.54	Nde1	67203		2.04
								Incenp	16319		4.68					Kif23	71819		1.68
Metabolism				Ubiquitin/Proteolysis/Proteasome				Skap2	54353	1.76	2.09	Chromatin modification/Assembly				Pnkd	56695		2.29
Gldc	104174		23.94	Cdc20	107995		5.27	Protein phosphorylation				Nusap1	108907	1.86		Mpdu1	24070	1.85	2.29
Brca1	12189	2.04	5.87	Capn2	12334		2.34	Melk	17279	4.05	17.84	Mad2l1	56150	1.61		Ipp	16351		2.25
Gpd2	14571		4.17	Xpnpep1	170750		2.32	Bub1	12235		7.69	Asf1b	66929			Tmem97	69071		2.24
Lycat	225010		3.06	Ide	15925		2.28	Wee1	22390	2.07	6.06	Nek2	18005			Pls3	102866		2.16
Gcat	26912		2.71	Psma1	26440		1.99	Ddr1	12305	2.26	5.82	Cbx5	12419			Dek	110052		2.16
Azin1	54375		2.29	Psmb1	19170		1.88	Stk39	53416	1.51	5.32	Pvt1	19296	1.57		Lancl2	71835		2.15
Pkm2	18746		2.19	Ube2l3	22195		1.82	Ryk	20187		5.21	Suv39h2	64707			Pqlc3	217430		2.14
Idh3a	67834		2	Blmh	104184		1.71	Plk4	20873	1.54	4.75	Cbx1	12412			Phlda3	27280		2.1
Gyg	27357		1.98	Ube2t	67196		1.64	Ttk	22137		3.62	Nap1l1	53605			Pih1d1	68845	1.59	2.03
Hmgcr	15357		1.96	Psmb4	19172		1.64	Mapk6	50772		2.07	Translation				Anp32e	66471		1.96
Rpe	66646		1.86	Psmc2	19181		1.63	Cdk2	12566		1.87	Farsb	23874		4.68	Tmem30a	69981		1.92
Sdhb	67680		1.8	Ube2e3	22193		1.63	Commd1	17846		1.81	Lamp2	16784	2.63	3.48	Prdx1	18477		1.91
Acsl5	433256		1.79	Psmb5	19173		1.62	Ptp4a2	19244		1.74	Mrpl27	94064	1.84	2.41	Mlf2	30853		1.88
Lypla1	18777		1.79	Psmd6	66413		1.57	Rps6ka1	20111		1.73	Eif4e2	26987	1.52	2.04	Hccs	15159		1.85
Decr1	67460		1.79	Psma3	19167		1.55	Ptpn9	56294		1.59	Eif1ay	66235		1.86	Actb	11461		1.84
Gstt1	14871		1.78	Psma5	26442		1.55	Ppp5c	19060	1.53	1.58	Mrpl13	68537		1.71	Tex9	21778		1.8
Ywhah	22629		1.74	RNA processing				Acp1	11431		1.56	Mrpl18	67681		1.69	Rcn1	19672	1.6	1.77
Gusb	110006		1.72	Syncrip	56403	1.79	2.88	Inhibitory signal/Apoptosis				Eif3s1	78655		1.63	Nubp1	26425	1.59	1.74
Soat1	20652		1.65	Slbp	20492		2.76	Lgals1	16852	3.91	27.4	Mrpl16	94063		1.52	Psmc3ip	19183		1.71
Aytl2	210992		1.65	Hnrpll	72692		2.73	Birc5	11799	5.33	21.6	Miscellaneous				Pdap1	231887		1.69
Aco1	11428		1.58	Thoc4	21681		2.21	Casp3	12367		5.37	Fignl1	60530	3.2	15.73	Arpc5	67771		1.67
Pcyt1a	13026		1.57	Cpsf2	51786		2.02	Gas2	14453		4.73	Trip13	69716	1.88	8.69	Ranbp1	19385		1.67
Electron transport				Ncbp2	68092	1.9	2.01	Dap	223453		2.37	Pcbp4	59092		6.39	Tmem77	67171		1.65
Acadl	11363		5.27	Snrpa1	68981		1.91	Bcl7c	12055		2.23	Sh3bgrl	56726		4.77	Pon2	330260		1.65
Pdia6	71853	1.6	2.42	Cstf3	228410		1.84	Api5	11800		2.16	Lxn	17035		4.32	Nup85	445007		1.62
Gsr	14782		2.06	Rod1	230257		1.79	Casp7	12369		2.14	Capg	12332	1.95	4.17	Ppp1r7	66385		1.62
Cyp51	13121		1.94	Sfrs3	20383		1.71	Phlda1	21664		1.94	Ltb4dh	67103		4.01	Sephs2	20768		1.61
Uqcrb	67530		1.92	Cops2	12848		1.7	Bax	12028		1.84	Gtse1	29870		3.59	Zdhhc6	66980		1.59
Glrx	93692		1.87	Lsm5	66373		1.7	Bnip2	12175		1.51	Tipin	66131		3.5	Pdia3	14827		1.58
Txnl1	53382		1.81	Sfrs2	20382		1.65	Cell adhesion/Communication				Ccdc99	70385		2.81	Mgat2	217664		1.56
Cyb5b	66427		1.7	Sfrs1	110809		1.61	Itgb1	16412		5.58	Mtm1	17772		2.71	Psmd7	17463	1.84	1.54
Sqle	20775		1.69	Regulation of cell growth/Proliferation				Adam19	11492		4.89	Wbp5	22381		2.42	Psmc1	19179		1.54
Txndc4	76299		1.6	Scin	20259		3.06	Adam9	11502		2.82	Acyp2	75572	1.75	2.41	Asah1	11886		1.53
Cyb5r3	109754		1.57	Chpt1	212862		2.36	Tnc	21923	1.94	1.56	Nucks1	98415		2.3	Peci	23986		1.53

Like T_E, the 543 transcripts increased in T_MSC versus T_N showed significant over-representation of ESC- and NSC-related genes (p=4.8 × 10^-5 and 3.4 × 10^-6, respectively) (Fig.4A). Notably, 72% of ESC- and NSC-related genes selected as increased in T_MSC were also increased in T_E (87 of 121 genes, Tables 3 and 4). Thus, despite their different proliferation capability, both alloreactive CD8⁺ T_E and T_MSC share some common stem cell transcriptional programs.

To validate our observations using an unsupervised approach, we obtained the Sarkar's dataset (GSE10239) (42), fit one-way ANOVA models, and performed probe-set selections identical to those of our array data. In their experiments, both KLRG1^Int and KLRG1^High P14 CD8⁺ T_E specific to lymphocytic choriomenigitis (LCMV)-gp33 showed probe set differences compared to T_N that had significantly overlaps with probe sets that we found to differ between alloreactive CD8⁺ T_E and T_N in our GVHD model (p<10^-200, by Fisher's Exact test in both cases). Most importantly, as with our alloreactive CD8⁺ T_E, P14 CD8⁺ T_E highly enriched for ESC- and NSC-related genes (Fig.4B). In contrast, P14 CD8⁺ memory cells decreased the expression of ESC- and NSC-related genes while activating genes enriched for HSCs (Fig.4B). These data independently confirmed our gene microarray data.

Characterization of stem cell genes activated in alloreactive CD8⁺ T_E

These observations were surprising to us because CD8⁺ T_E have been believed to be terminally differentiated cells (1, 43). To understand what these CD8⁺ T_E-related stem cell genes were, we assigned these genes to several functional categories by filtering for keywords in the GO terms as previously described by Ramalho-Santos et al. (41). As shown in Table 5, we found that CD8⁺ T_E shared the similarity with ESCs and NSCs in the expression of genes associated with: 1) regulation of cell cycle, 2) resistance to stress, 3) chromatin modification and transcription / translation regulation, 4) cell survival and death, 5) signaling, and 6) others. In cell cycle category, CD8⁺ T_E increased both negative (e.g., p18^Ink4c and p21^Cip1) and positive cell cycle regulators (e.g., Ccna and Ccnb). Interestingly, after removing all these cell cycle genes from our array data and Ramalho-Santos's stem cell data set (41), we found that the rest of transcripts increased in alloreactive CD8⁺ T_E remained significant over-representation of ESC- and NSC-related genes (p<5.0^-15, by Fisher's Exact test in both cases). Our further evaluation of Sarkar's (42) and Ramalho-Santos's (41) data sets without these cell cycle genes also revealed significant enrichments in P14 CD8⁺ T_E for ESC- and NSC-related transcripts (p<4.2^-82 in both cases). Thus, cell cycle genes are not the only attributes to the similarity of gene expression between CD8⁺ T_E and embryonic and neural stem cells.

Table 5. Functional categories of 401 CD8⁺ T_E-related stem cell genes (See gene symbols and identification# in Table 3 and Table 4).

Category		Gene Symbols
1. Cell cycle regulation
Cell cycle	(35)	Ccna2, Cdca5, Ccnb1, Sgol1, E2f8, Chek1, Cks1b, Cdkn1a, Ncaph, Plk1, Aspm, Ect2, Cdkn2c, Gmnn, Ccne1, Syce2, Cdc45l, Cdkn2b, Cdc25c, Pola1, Aurka, Ccnf, Gsg2, Dlg7, Cdkn2a, Nfatc1, Mybl2, Ran, Ccng1, Erh, Rb1, Psmd1, Cops5, Mapre1, Rap1a
2. Resistance to stress
DNA replication / repair	(34)	Rrm2, Cdc6, Il1rl1, Dtl, Rad51, Uhrf1, Pttg1, Dna2l, Lig1, Rpa2, Rfc3, Rrm1, Mcm5, Smc2, Fen1, Prim1, Mcm2, Hmgn2, Mcm7, Prim2, Mcm3, Rfc5, Mcm4, Orc6l, Rfc4, Recql, Rpa3, Smc6, Brca2, Blm, Topbp1, Hmgb1, Dbf4, Ube2n
Ubiquitin / Proteolysis / Proteasome	(20)	Cdc20, Coq7, Nedd4, Capn2, Xpnpep1, Ide, Coq3, Psma1, Psmb1, Ube2l3, Psmb2, Blmh, Ube2t, Psmb4, Psmc2, Psmb5, Psmd6, Psma3, Psma5, Ube2e3
Metabolism	(27)	Gldc, Gsto1, Brca1, Bcat1, Myo5a, Gpd2, Lycat, Gcat, Shmt1, Uxs1, Azin1, Pkm2, Idh3a, Gyg, Hmgcr, Rpe, Pfkfb1, Sdhb, Decr1, Acsl5, Lypla1, Gstt1, Gusb, Acot9, Aytl2, Soat1, Aco1
Electron transport	(14)	Acadl, Nqo2, Etfb, Pdia6, Gsr, Cyp51, Uqcrb, Glrx, Txnl1, Cyb5b, Sqle, Txndc4, Txnl5, Cyb5r3
3. Chromatin modification, and transcription / translation regulation
Chromatin modification/Assembly	(12)	Nusap1, Mad2l1, Asf1b, Tmem38b, Nek2, Cbx5, Pvt1, Hat1, Suv39h2, Cbx1, Nap1l1, Smarcad1
Transcription/Transcription regulation	(33)	Mdfic, Prdm1, Uhrf1, Hells, Ezh2, Tacc3, Eomes, Csda, Cenpk, Nsbp1, Klf10, Wdhd1, Sub1, Sap30, Pmf1, Myef2, Rab8b, Gtf2e2, Inppl1, Cdca7l, Tceb1, Rbl1, Tmpo, Hnrpdl, Polr3k, Cops2, Bzw1, Sin3b, Cnot7, Klf9, Mtdh, Th1l, Nfyb
RNA processing	(16)	Syncrip, Slbp, Hnrpll, Ell2, Thoc4, Cpsf2, Ncbp2, Snrpa1, Cstf3, Rod1, Sfrs3, Lsm5, Sfrs2, Ddx1, Sfrs1, Snrpd1
Translation	(11)	Farsb, Lamp2, Mrpl27, Eif4e2, Lamc1, Ap3s1, Mrpl13, Mrpl18, Eif3s1, Dnajc9, Mrpl16
4. Survival / apoptosis
Survival/Apoptosis	(16)	Lgals1, Birc5, Ckap2, Casp3, Gas2, Scin, Grn, Dap, Bcl7c, Api5, Casp7, Skap2, Phlda1, Bax, Nasp, Bnip2
5. Signaling
Protein phosphorylation	(20)	Melk, Bub1, Wee1, Ddr1, Stk39, Ryk, Plk4, Ttk, Mapk6, Chek2, Ppap2a, Cdk2, Commd1, Ptp4a2, Rps6ka1, Pigf, Ptpn9, Ppp5c, Acp1, Ptprk
Receptor/Signal transduction	(24)	Epas1, Rhoc, Rhoq, Hif1a, Grn, Dusp16, Rap2a, Gpr56, Akt1, Arhgap21, Tank, Gipc1, Fut8, Rrbp1, Csnk2b, Ssr2, Ranbp5, Gnb1, Nup93, Nasp, Smap1, Timm17b, Kpna3, Rab1
6. Others
Transport (Protein/Ion)	(33)	Sypl, Kif20a, Syt11, Arl5a, Snx10, Slc35b2, Clic4, Slc35b1, Dbi, Arl1, Lman1, Vps54, Arf4, Mtch2, Clcn5, Snx5, Vps29, Slc2a3, Trappc1, Nup62, Slc33a1, Vps45, Atp6v0a2, Golt1b, Slc4a7, Bcap29, Slc39a6, Aaas, Sfxn1, Slc12a2, Slc25a10, Sec14l1, Asna1
Cell adhesion/Communication	(5)	Itgb1, Adam19, Adam9, Vcl, Tnc
Microtubule-based process / cytokinesis	(12)	Kif2c, Kif11, Kif22, Kif4, Mid1ip1, Nde1, Kif23, Prc1, Cks2, Anln, Incenp, Dstn
Miscellaneous	(89)	Wisp1, Igfbp7, Chpt1, Fignl1, Trip13, Pcbp4, Tmem49, Tacstd1, Sh3bgrl, Spp1, Lxn, Capg, Smoc2, Exo1, Ltb4dh, Asns, Synpo, Gtse1, Tipin, Plagl1, Ccdc99, Mtm1, Lactb2, Gmfb, Nudt1, Wbp5, Acyp2, Errfi1, Htatip2, Ccdc80, Nucks1, Pnkd, Mpdu1, Ipp, Tmem97, Dtymk, Apaf1, Pls3, Dek, Lancl2, Pqlc3, Fyn, Phlda3, Pih1d1, Anp32e, Dsp, Tmem30a, Prdx1, Mlf2, Hccs, Actb, Tmem41b, Tex9, Rcn1, Ptprj, Nubp1, Pex7, Ywhah, Esd, Bdnf, Psmc3ip, Nab1, Pdap1, Arpc5, Ranbp1, Dck, Hnrpab, Tmem77, Pon2, Inhbb, Tyms, Nup85, Ppp1r7, Sephs2, Zdhhc6, Pdia3, Pcyt1a, Mgat2, Kdelr2, Hrb, Psmd7, Psmc1, Cox4nb, Asah1, Peci, Uck2, Nudt5, Vdac3, Tiprl

Genes engaged in resistance to stress represented were a large group of stem cell transcripts activated in CD8⁺ T_E. This group of genes included transcripts of DNA replication and repair, ubiquitin/proteasome, metabolism and electron transport (Table 5). These genes are considered to be related to the stress condition of stem cells (41). For example, genes including Coq7, Ube2l3, Nedd4 and Psma play important role in modifying abnormal or short-lived proteins (44-45). DNA repair gene Rad51 is critical to maintaining chromosomal stability and preventing genetic mutation potentially occurring during cell division (41, 46). Genes associated with metabolisms, such as Gldc and Gsto1, are critical to amino acid and antioxidant metabolism (47-48). The functions of these genes in T cells and stem cells remain largely unknown.

Another important finding was that CD8⁺ T_E activated many genes engaged in DNA methylation, chromatin modification, transcription and survival in ESCs and NSCs. For example, Uhrf1 protein forms complexes with DNA methyltransferase Dnmt1, which may result in an inheritable DNA methylation (49). Hells (also known as Lsh) protein associates with Dnmt3a and Dnmt3b in embryonic cells for DNA methylation and transcription (50). Tacc3 protein can activate gene transcription even prior to demethylation (51). Birc5 (also known as Survivin), is an apoptosis inhibitor in both normal and malignant cells (52). Ezh2, which encodes a chromatin modifying enzyme with methyltransferase activity, orchestrates gene expression in both embryonic and adult stem cells (53-55). The representative gene expression of these chromatin modifiers and transcriptional regulators were shown in Fig.5A, and validated by real-time RT-PCR (Fig.5B). Thus, genes in this category have multiple roles in controlling cell fate, self-renewal, differentiation, survival and memory function.

Fig.5.

Characterization of stem cell genes in alloreactive CD8⁺ T cells. (A) The relative expression of stem cell genes in alloreactive CD8⁺ T_E and CD8⁺ T_MSC are shown with genes identified by functional set enrichment analysis using curated gene lists from MSigDBv2 (Fig.4). (B) Real-time RT-PCR analysis shows the relative mRNA expression of selected genes in each T cell subset. Data are representative of three independent preparations of alloreactive T cells. (C) Western blot analysis shows the expression of EZH2 protein in purified day-14 CD8⁺ T_E, T_MSC and control T_N. (D) donor CD8⁺ T_N and day-14 CD8⁺ T cells were stained with anti-Ezh2 Ab using intracellular staining and analyzed by flow cytometry. Dot plots show the expression of Ezh2 in cell subset of donor T cells. Mean fluorescent intensity shows the amount of testing antigen.

Finally, we noted that CD8⁺ T_E did not activate those genes associated with pluripotency of ESCs, such as Oct4, Sox2, Klf4, Nanog, and c-Myc(56) (Fig. 5A, other data not shown). Furthermore, genes associated with HSC self-renewal were decreased in CD8⁺ T_E. For example, CD8⁺ T_E markedly down-regulated the expression of many genes related to receptor, signaling and transcription that are normally expressed in HSCs, such Il6ra, Il6st, Smad4, Smad7, and c-Myc (Table 6, Fig.5A). Among them, Smad4 and Smad7 have been shown to be required for self-renewal and quiescence of HSCs (57).

Table 6. HSC-enriched genes that are decreased in CD8⁺ T_E.

Gene symbol	Gene ID#	Fold Change		Gene symbol	Gene ID#	Fold Change		Gene symbol	Gene ID#	Fold Change		Gene symbol	Gene ID#	Fold Change		Gene symbol	Gene ID#	Fold Change
		TMSC/TN	TE/TN			TMSC/TN	TE/TN			TMSC/TN	TE/TN			TMSC/TN	TE/TN			TMSC/TN	TE/TN
Cell cycle				Translation				Receptor/Signal transduction				Regulation of cell growth/Proliferation				Camsap1	227634		-1.75
Lats2	50523		-1.6	Mknk2	17347		-1.62	Aplp2	11804	-1.75	-1.5	Ppan	235036		-1.6	Las1l	76130		-1.76
Rassf2	215653		-1.64	Mrpl24	67707		-1.75	Prpf6	68879		-1.5	Creg1	433375		-2.03	Pcf11	74737		-1.78
Flcn	216805		-2.1	Eif4g2	13690	-1.55	-1.8	Mettl3	56335		-1.51	Pctk1	18555		-2.09	Serpini1	20713		-1.79
Sesn1	140742	-2.71	-5.69	Rpl37a	19981		-2.18	Notch1	18128		-1.51	Epc1	13831	-2.03	-2.28	Fnbp4	55935		-1.81
DNA replication/Repair				Rpl13	270106		-2.57	Ikbkg	16151		-1.52	Socs3	12702		-2.41	Ccdc53	67282		-1.85
Polg	18975		-1.55	Rpl22	19934		-2.59	Chd8	67772		-1.52	Negative regulation of transcription				Phf1	21652		-1.86
E4f1	13560		-1.63	Protein phosphorylation				Rasa3	19414		-1.54	from RNA polymerase II promoter				Gtf2ird2	114674		-1.87
Sfpq	71514	-1.54	-1.75	Tesk1	21754		-1.59	Mapk14	26416	-1.55	-1.54	Cutl1	13047		-1.52	Numa1	101706	-1.57	-1.88
Xpc	22591		-2.52	Mast3	546071		-1.59	Mfng	17305		-1.56	Id3	15903		-1.77	Brd1	223770		-1.9
Nfix	18032		-6.12	Prkcq	18761		-1.64	Frap1	56717		-1.57	Bcl6	12053		-1.85	Plekha1	101476		-1.91
Ubiquitin/Proteolysis/Proteasome				Jak1	16451	-1.52	-1.75	Itpkb	320404	-1.81	-1.59	Tcf25	66855		-2.02	Frmd6	319710	-1.93	-1.98
Usp3	235441		-1.54	Csnk1g2	103236		-1.75	Xab2	67439		-1.59	Tle1	21885		-3.63	Akap8l	54194		-1.98
Fbxo38	107035		-1.57	Ptpn21	24000	-1.68	-1.87	Dtx2	74198	-1.77	-1.6	Foxp1	108655	-1.67	-4.93	Grwd1	101612		-1.99
Ubl4	27643		-1.62	Prkce	18754		-1.96	Dicer1	192119	-1.77	-1.63	Transport (Protein/Ion)				Otud5	54644		-1.99
Usp9x	22284		-1.64	Map4k4	26921		-2.44	Srpk2	20817	-1.53	-1.68	Mcfd2	193813		-1.52	Etnk1	75320		-2.02
Fbxo9	71538		-1.72	Tec	21682		-2.47	Mapk8	26419	-1.58	-1.73	Bet1l	54399		-1.59	Abcf3	27406		-2.03
Usp19	71472	-1.66	-1.85	Ikbke	56489		-2.52	Stat6	20852		-1.74	Abcb8	74610		-1.61	Acyp1	66204		-2.04
Rnpepl1	108657	-1.56	-2.04	Ptpn23	104831		-2.87	Ppp2r5e	26932		-1.76	Acox1	11430		-1.65	Nsun4	72181		-2.06
Ube2d2	56550	-1.77	-2.06	Prkd2	101540		-2.98	Phc2	54383		-1.78	Ecgf1	72962		-1.65	Tmem141	51875		-2.07
Rbbp6	19647	-1.5	-2.1	Transcription/Transcription regulation				Rrad	56437		-1.84	Pitpnm1	18739		-1.76	Snord22	83673		-2.07
Zfp292	30046		-2.31	Zzz3	108946		-1.51	Stub1	56424		-1.84	Acadm	11364		-1.82	Lrrc8a	241296	-1.6	-2.1
Rffl	67338		-2.35	Zmynd11	66505		-1.51	Arrb1	109689		-1.87	Mybbp1a	18432		-1.89	Paqr7	71904	-1.68	-2.1
Klk8	259277		-4.21	Zkscan6	52712		-1.53	Azi2	27215	-1.74	-1.91	Pitpnc1	71795	-1.53	-1.97	Cdc42se2	72729	-1.56	-2.12
Metabolism				Nfe2l1	18023		-1.53	Rai1	19377	-1.61	-1.91	Scamp1	107767	-1.59	-1.99	Bola2	66162		-2.16
Tk2	57813		-1.5	Ezh1	14055		-1.55	Mfhas1	52065		-1.98	Hcn3	15168		-2.12	Crlf3	54394	-1.55	-2.21
Glb1l	74577		-1.51	Scmh1	29871		-1.56	Foxo3a	56484		-1.99	P2rx4	18438		-2.44	Zdhhc9	208884		-2.26
Idua	15932		-1.52	Zkscan3	72739		-1.6	Spred2	114716		-2.07	Ramp3	56089		-2.54	Gas5	14455		-2.3
Glul	14645		-1.54	Zkscan1	74570		-1.69	Gaa	14387		-2.13	Laptm4b	114128	-1.88	-2.88	Tmem66	67887		-2.33
Elovl6	170439		-1.55	Rarg	19411	-2.56	-1.73	Dvl1	13542		-2.15	Kcnn4	16534		-3.1	Rbm38	56190		-2.39
Supt6h	20926		-1.61	Zfx	22764		-1.73	Il16	16170	-1.64	-2.26	Rab3ip	216363		-3.12	Jarid1b	75605		-2.4
Pcyt2	68671		-1.69	Mef2d	17261		-1.77	Spsb1	74646		-2.26	Slco3a1	108116		-3.18	Marveld1	277010		-2.47
Mthfr	17769		-1.71	Ncor2	20602	-1.6	-1.81	Gbp2	14469		-2.29	Slc12a7	20499		-4.35	Ctdsp2	52468	-1.94	-2.56
Ndufa6	67130		-1.85	Zfml	18139	-1.79	-2.05	Crebbp	12914	-1.64	-2.41	Abca1	11303		-4.42	Armcx2	67416		-2.68
Pgs1	74451		-2.1	Zfp96	22758		-2.06	Macf1	11426	-1.61	-2.57	Ramp1	51801	-1.68	-17.13	Ipo4	75751		-2.82
Ihpk1	27399		-2.26	Zeb1	21417		-2.11	Il4ra	16190		-2.8	Miscellaneous				Rabac1	14470		-3.3
Cbr1	12408		-2.55	Taf1a	21339		-2.28	Ltb	16994		-2.81	Tnip1	57783	-1.69	-1.51	Plekho1	67220		-3.54
Lpin1	14245		-2.57	Bach1	12013	-1.55	-2.34	Eng	13805		-3.43	Nphp1	53885		-1.52	Peli1	67245	-1.94	-4.03
Man2c1	73744		-2.88	Klf3	16599		-2.52	Smad4	17128	-1.53	-3.53	Isca1	69046		-1.53	Btbd14a	67991		-4.63
Gstk1	76263		-2.99	Jmjd1a	104263		-2.54	Il6st	16195	-1.61	-4.31	Msl2l1	77853		-1.54	Rreb1	68750		-7.28
Adcy6	11512		-3.08	Basp1	70350	-2.71	-2.64	Smad1	17125		-5.01	Nol1	110109		-1.56	Cxxc5	67393	-2.79	-7.61
Dph5	69740		-3.81	Ash1l	192195	-1.83	-2.84	Mettl4	76781	-1.58	-5.5	Dctn1	13191		-1.56
Ldhb	16832	-1.97	-4.83	Zfp1	22640		-2.91	Notch2	18129	-1.58	-5.62	Klhl7	52323		-1.62
Pdk1	228026		-6.29	Zbtb20	56490	-2.61	-3.08	Thra	21833		-5.8	Sh3gl1	20405		-1.63
RNA processing				Dbp	13170	-1.96	-3.63	Inadl	12695		-6.08	Mgea6	217615		-1.65
Rpusd4	71989		-1.51	Smad7	17131	-1.87	-3.68	Il6ra	16194		-18.59	Bcl7b	12054		-1.66
Imp3	102462		-1.54	Irf6	54139	-2.29	-4.63	Cell adhesion/Communication				Gltscr2	68077		-1.7
Mphosph10	67973		-1.62	Ssbp2	66970		-6.88	Nisch	64652	-1.74	-2.16	Serf2	378702		-1.7
Prpf38b	66921		-1.65	Chromatin modification/Assembly				Itga6	16403	-1.67	-3.24	Dhx30	72831		-1.71
Nola3	66181		-1.8	Chd1	12648		-1.59					Cep350	74081	-1.57	-1.74
Utp20	70683		-2.32	Smarca2	67155		-2.44					Cramp1l	57354	-1.78	-1.74

Role of chromatin modifying enzyme Ezh2 in CD8⁺ T cells

Data from previous studies demonstrate that the loss of Ezh2 in mature T cells impairs their proliferative response to anti-CD3 Ab (58). We observed that Ezh2 mRNA and protein were significantly increased in alloreactive CD8⁺ T_E (Fig.5 B and C). Flow cytometry analysis showed at the single cell level that all day 14 CD8⁺ T_E expressed higher levels of Ezh2 protein than T_N (Fig.5D). Further tests using MSigDBv2 demonstrated that alloreactive CD8⁺ T_E activated 23 of 30 Ezh2 target or partner genes previously identified by others(54) (Fig.6A). Ex vivo culture confirmed that purified Ezh2-shRNA GFP⁺CD8⁺ T_N had reduced expression of Ezh2 protein (Fig.6B) and decreased their expansion by approximate 4-folds in response to anti-CD3 and anti-CD28 Abs as compared to Control-shRNA GFP⁺CD8⁺ T cells (Fig.6C). Thus, Ezh2 may play important roles in antigen-activated CD8⁺ T cells.

Fig.6.

Identification of stem cell gene Ezh2 in alloreactive CD8⁺ T cells. (A) The relative expression of 23 Ezh2 target or partner genes that were also selected as increased in alloreactive CD8⁺ T_E relative to T_N. (B) Western blot shows Ezh2 protein in CD8⁺ T_N expressing with Ezh2-shRNA or Control-shRNA. These Ezh2-shRNA GFP⁺CD8⁺ T_N were derived from B6 mice reconstituted with HSCs infected by Ezh2-shRNA/GFP-pLVPT^off. GFP⁺CD8⁺ T_N from B6 mice reconstituted with Con-shRNA/GFP-pLVPT^off were isolated as controls. (C) Sorted Ezh2-shRNA GFP⁺CD8⁺ T_N and Control-shRNA GFP⁺CD8⁺ T_N (1×10⁵ / well, 96 well plate) were stimulated with anti-CD3Ab and anti-CD28 (2.5 μg/ml for each). Five days later, cells were recovered and analyzed with flow cyotmetry for measuring GFP⁺CD8⁺ T cells. (D) Unfractionated Ezh2-shRNA GFP⁺CD8⁺ T_N were stimulated with BABL/c mouse-derived DCs, or with IL-7 alone. Five days later, cells were recovered and analyzed using flow cytometry for measuring GFP⁺CD8⁺ T cells. Fold change of GFP⁺CD8⁺ T cells was calculated based on the output number of GFP⁺CD8⁺ T cells after culture divided by the input number of GFP⁺CD8⁺ T cells before culture. Data shown in (B, C and D) are representative of two independent experiments.

We further asked whether Ezh2 inhibition had differential effects on alloantigen stimulated versus homeostatic cytokine IL-7 mediated CD8⁺ T cell proliferation. Since our previous studies suggest that purified mouse T cells rapidly diminish in cultures in the absence of DCs (30), we stimulated unfractionated Ezh2-shRNA GFP⁺CD8⁺ T_N with allogeneic DCs or with IL-7. Five days later, cells were recovered from the culture and analyzed for the expansion GFP⁺CD8⁺ T cells using flow cytometry. Interestingly, as compared to Control-shRNA GFP⁺CD8⁺ T cells, the expansion of Ezh2-shRNA GFP⁺CD8⁺ T cells was decreased in the culture supplemented with allogeneic DCs but not with IL-7 (Fig.6D). This suggests that Ezh2 may be required for antigen-driven T cell responses rather than homeostatic T cell proliferation.

Discussion

These studies identify a group of stem cell genes that are normally expressed in ESCs and NSCs in alloreactive CD8⁺ T cells. Most of these stem cell genes are found to be essential to cell cycle regulation, DNA replication and repair, stress resistance, chromatin modification and transcription regulation. One of these genes, Ezh2, emerges as an important regulator for the proliferation of antigen-activated CD8⁺ T cells. On the other hand, alloreactive CD8⁺ T_E increase the expression of many genes that mediate cell apoptosis and growth arrest. We demonstrate that these alloreactive CD8⁺ T_E were rapidly diminished in vivo in lethally irradiated secondary congenic recipients, suggesting that homeostatic factors alone are not sufficient to sustaining alloreactive CD8⁺ T_E. However, upon chronic exposure to alloantigens alloreactive CD8⁺ T_E proliferate to persist in vivo despite their massive apoptotic death. These data indicated that although alloreactive CD8⁺ T_E were “terminally differentiated” with drastically increased susceptibility to apoptotic death, they had the ability to survive and persist via the mechanisms of continuous replication in the presence of alloantigens. Thus, these newly identified stem cell genes could be important targets for understanding and modulating allogeneic T cell responses and GVHD.

T cells are known to be stem cell-like cells (1, 3-4). Gene expression profile analysis reveals that memory T cells may trigger HSC-related transcriptional programs to regulate their self-renewal in the absence of antigens (17). We found that alloreactive CD8⁺ T_E activated stem cell transcriptional programs which are operational in both ESCs and NSCs, while genes decreased in CD8⁺ T_E were over-represented in HSC-enriched genes. Our observations were further validated by our re-analyzing the gene expression profile of Sarkar et al.(42), which revealed that ESC- and NSC-related genes were activated in LCMV gp33-specific CD8⁺ T_E, but decreased in CD8⁺ memory T cells. It appears that functional T cell subsets can be defined by different stem cell transcriptional programs which regulate their unique properties of T_E versus memory T cells. It is likely that alloreactive CD8⁺ T_E require ESC- and NSC-related transcriptional programs to generate a large number of functionally active effectors sufficient to eliminate the target antigen.

Interestingly, reactivation of stem cell transcriptional programs occurs in both alloantigen-specific CD8⁺ T_E generated during GVHD and viral antigen-reactive CD8⁺ T_E derived from acute infection. We identified these CD8⁺ T_E-related stem cell transcriptional programs using curated gene lists from MSigDBv2 (40). Although it has been reported that stem cell transcriptional profiles identified by different groups do not correlate well to each other (59), we found that approximately 30% of the genes increased in our alloreactive CD8⁺ T_E are present on ESC and/or NSC gene lists of Ramalho-Santos et al.(41). Other studies also clearly showed a significant correlation of stem cell gene expression profiles identified by different groups (41, 60). Furthermore, we independently validated that the gene expression profiles of our alloreactive CD8⁺ T_E were correlated significantly with those P14 CD8⁺ T_E of Sarkar et al.(42). Thus, these CD8⁺ T_E-related stem cell transcriptional programs may have significant implications in regulating T cells across various types of immune responses.

Overlapped gene expression profiles between CD8⁺ T_E and embryonic and neural stem cells suggest that these cells could share some common properties. ESCs and NSCs characterized in Rammalho-Santos' study are highly proliferating stem cells, whereas HSCs are quiescent cells (41). In their study, ESCs were derived from the inner cell mass of the blastocyst stage of embryos, whereas NSCs were isolated from brain-derived neurospheres. Both ESCs and NSCs were highly purified after ex vivo cultures for gene expression profile analyses. In contrast, HSCs were freshly isolated from BM of normal B6 mice based on dual-dye efflux and HSC markers (41). We found that many of these CD8⁺ T_E-related stem cell genes were associated with cell cycle regulation, DNA replication and repair, and stress resistance. This was in agreement with our findings that a proportion of alloreactive CD8⁺ T_E continually proliferated upon chronic exposure to alloantigens. Thus, genes controlling proliferation of ESCs and NSCs are an important component of the similarity between CD8⁺ T_E and embryonic and neural stem cells.

However, CD8⁺ T_E did not increase the expression of genes associated with pluripotency of ESCs. In contrast, they activated many other stem cell genes that are found to be important for controlling cell fate, differentiation, survival, self-renewal and memory function in ESCs and NSCs, such as Uhrf1,Tacc3, Hells, Birc5, and Ezh2 (37, 49-54, 61). For example, Ezh2 binds to chromatin and DNA during cell dividing, thereby preserving transcriptional programs and cell identity established during earlier response phase (62-63). We found that CD8⁺ T_E had the ability to sustain their effector functions over numerous rounds of cell dividing upon chronic exposure to allogeneic stimuli. Previous reports have also suggested that CD8⁺ T_E can become self-renewing memory T cells upon clearance of the target antigen (4-6, 8, 22). These observations suggest that CD8⁺ T_E share some common properties with ESCs and NSCs in the expression of stem cell transcriptional programs that are engaged in cell fate decision, self-renewal, survival, differentiation and memory function.

Several lines of evidence suggest that Ezh2 may be important for antigen-driven T cell responses. We found that Ezh2 was abundantly expressed in antigen-activated CD8⁺ T cells but not in CD8⁺ T_N. Silencing Ezh2 inhibited CD8⁺ T cell proliferation activated by TCD/CD28 costimulation and allogeneic DCs, which is consistent with a previous report of others (58). Interestingly, knockdown of Ezh2 did not affect mature T cells to proliferate in response to homeostatic cytokine IL-7 alone. Thus, it is unlikely that inhibition of Ezh2 in alloreactive T_E can globally affect donor T cell immunity after allogeneic HSCT. However, further studies are necessary to investigate the impact of Ezh2 inhibition in antigen-activated T cell responses and GVHD.

Our results suggest that APCs may play an important role in regulating stem cell transcriptional programs in CD8⁺ T cells. We found that alloreactive CD8⁺ T_E continuously replicated in secondary allogeneic recipients and caused severe GVHD, but rapidly diminished in congenic recipients where alloantigens were absent. Thus, allogeneic stimuli rather than homeostatic factors are critical to the continuous replication in vivo of alloreactive CD8⁺ T_E. This may explain why APCs are important for alloreactive T cell-mediated GVHD at both the induction and effector phase (13, 23, 64-66). Other studies suggest that antigenic stimulation is also necessary for protective immunity during chronic infection (67-69). It is likely that antigen stimulation sustains the replication of T_E through the activation of stem cell transcriptional programs. However, other non-antigenic stimuli, such as inflammatory cytokines and co-stimulatory signals, could also be important for regulating stem cell transcriptional programs in CD8⁺ T_E. For example, CD4⁺ T cells are found to be important for in vivo development of long- lasting CD8⁺ memory T cells and are required for mediating chronic GVHD (70-72). It is possible that signals derived from CD4⁺ help T cells might affect the expression of these stem cell genes in antigen-activated CD8⁺ T cells.

Stem cell transcriptional programs may also play an important role in alloreactive CD8⁺ T_MSC. Gene expression profile analysis showed that these CD8⁺ T_MSC were less differentiated as they did not produce cytotoxic molecules and inflammatory cytokines. A recent study suggests that activation of Wnt-signaling arrests the effector differentiation while enhancing the generation of CD8⁺ T_MSC with greater ability than mature memory T cells to proliferate and generate tumor-reactive T_E in vivo (21). This further supports that CD8⁺ T_MSC themselves do not mediate tissue injury, but can further differentiate into functional CD8⁺ T_E (15). Notably, 70% of stem cell genes activated in alloreactive CD8⁺ T_MSC remained increased in CD8⁺ T_E. It will be interesting to determine how these ESC- and NSC-related genes affect the proliferation and differentiation of CD8⁺ T_MSC.

In summary, we have identified that “terminally differentiated” alloreactive CD8⁺ T cells activate stem cell transcriptional programs that are normally expressed in ESCs and NSCs. This group of stem cell genes may play important roles in regulating the proliferation and persistence of alloreactive T cells upon chronic exposure to alloantigens in GVHD. Further exploring the specific roles of ESC- and NSC-related stem cell genes in chronically activated T cells could have significant impact on understanding and modulating pathogenic T cell responses in many other inflammatory conditions, such as chronic infections, autoimmune diseases and rejection of grafted solid organs.