|Home | About | Journals | Submit | Contact Us | Français|
The differentiation state of CD8+ T cells has emerged as a crucial determinant of their ability to respond to tumor and infection. Signals from T-cell receptors, co-stimulatory molecules and cytokine receptors direct the differentiation process. These signals ‘program’ sustained and heritable gene expression patterns that govern progressive differentiation and lineage commitment. The epigenetic mechanisms by which T cells are programmed are just beginning to be elucidated. Understanding the mechanisms that control CD8+ T-cell differentiation is important in the development of novel immunotherapy strategies.
Antigen-presenting cells (APCs) and other immune and non-immune cells direct CD8+ T-cell differentiation by engaging receptors for a host of membrane-bound and soluble molecules [1••,2,3]. Signals from these receptors induce epigenetic changes that ‘program’ sustained but mutable gene expression patterns that govern progressive differentiation and lineage commitment decisions [4,5]. Emerging evidence indicates that the differentiation state crucially determines CD8+ T-cell effectiveness in responding to infection and tumor [6,7••,8,9••,10,11]. What are the differentiation states that characterize effective CD8+ T cells? How can CD8+ T cells be programmed to differentiate into optimal effector cells? Here we discuss the concepts of T-cell programming and differentiation and their implications for the development of potent new CD8+ T cell-based immunotherapy.
Dynamic CD8+ T cell–APC interactions drive CD8+ T-cell proliferation, differentiation and lineage commitment. This process results in the generation of cells that have diverse phenotypic and functional characteristics [3,12,13]. The nomenclature, definitions, characteristics and differentiation pathways for CD8+ T cells are controversial, but two broad categories are generally acknowledged and have been named for their apparent function [12,14,15]. ‘Effector’ cells (TEFF) are highly cytolytic in vitro and express high levels of molecules required for cell killing, such as perforin, granzymes, interferon (IFN)-γ, tumor necrosis factor (TNF) and FAS ligand (FASL) [13,14]. ‘Memory’ cells are less cytolytic in in vitro assays, but exhibit increased survival, show the capacity for antigen-independent self-renewal, and respond vigorously to secondary antigen challenge [8,12,14,16].
Two subsets of memory cells, ‘effector memory’ (TEM) and ‘central memory’ (TCM), were originally identified on the basis of tissue homing molecules and effector function . TCM were described as CD62L+CCR7+ cells that home to lymph nodes and have relatively low immediate effector function. TEM were defined as CD62L−CCR7− cells that preferentially home to peripheral tissues and inflammatory sites and possess relatively high immediate effector function. More recently, the distinction between TEM and TCM based on effector function has been questioned [18,19]. Generalizations about these two memory subsets have been further confused by the discovery of CCR7− antigen-experienced T cells in lymph nodes [20,21]. Another subset, ‘memory stem cells’, was recently identified in a murine model of graft-versus-host disease. These post-mitotic CD44loCD62LhiCD8+ T cells were characterized by the expression of Sca-1, CD122 and BCL-2 and were capable of self-renewal and could generate all subsets of memory and effector CD8+ T cells [22••].
Conflicting models of CD8+ T-cell differentiation have been proposed to explain the generation of memory subsets. A model suggesting that TCM and TEM arise from distinct lineages was based on the finding that T-cell receptor (TCR) repertoires of TCM and TEM in the peripheral blood of healthy individuals are largely distinct . However, subsequent studies showing that TCM and TEM clones can derive from a common naïve precursor have cast doubt on the separate lineage hypothesis .
A linear pathway of differentiation has instead gained acceptance; however, contention remains about whether memory cells arise from TEFF or vice versa. A progressive sequence of naïve → TEFF → TEM → TCM differentiation has been proposed based on the finding that CD62L+ cells emerged after the adoptive transfer of CD62L− enriched memory cells [6,25•]. Whether these results reflect a true TEM → TCM conversion or a selective survival and/or proliferation advantage of TCM over TEM has not been convincingly demonstrated.
Instead, mounting data, including ex vivo phenotypic analyses of virus-specific CD8+ T cells in acute and chronic viral disease, measures of telomere length, and in vitro differentiation studies, support a linear sequential progression (naïve → TCM → TEM → TEFF) model [7••, 9••,12,26]. The recent finding that the gene expression signature of TCM lies between that of naïve cells and TEM further supports the naïve → TCM → TEM → TEFF sequence .
Expanding on this model, memory stem cells might represent the earliest antigen-experienced cell population to emerge in the CD8+ T-cell differentiation pathway [22••]. Paralleling B-cell differentiation, memory CD8+ T cells might represent effector cells in an arrested differentiation state [15,28]. Thus, CD8+ T-cell differentiation could be regarded in terms of a continuum from early to late effectors rather than the movement of a T cell between subsets descriptively named memory stem cell, TCM, TEM and TEFF subsets .
Where on this differentiation continuum are CD8+ T cells most capable of eradicating infection or tumor? Late effectors, owing to their high in vitro cytotoxicity, were initially thought to be the most effective cells. However, increasing evidence indicates that progressive differentiation leads to decreased ability to eliminate infection or tumor. For example, impaired function of late effector cells has been observed in patients that have progressive human immunodeficiency virus infection. CD8+ T cells in these patients displayed a late effector (CD27−CD28−) phenotype and decreased proliferative capacity, deemed ‘replicative senescence’ [9••]. Furthermore, in the absence of early effectors, late effectors were unable to control cytomegalovirus replication in patients co-infected with human immunodeficiency virus . In clinical trials of adoptive cell transfer therapy for cancer, clonal populations of CD8+ T cells that had been multiply stimulated proved ineffective [30,31]. These cells appear to represent late effectors that have poor survival capability as they have a late effector phenotype and do not engraft or persist after adoptive transfer [30,31]. In contrast, treatment with less-expanded tumor infiltrating lymphocytes caused objective responses in about 50% of treated patients in a recent clinical trial . In patients treated with tumor infiltrating lymphocytes, tumor regression and T-cell persistence correlate with increased telomere length [10,33], and cells that persist express an early effector (CD27+CD28+) phenotype . These data suggest that an early differentiation state is important for T-cell efficacy.
Studies in murine models have confirmed the superior function of less-differentiated CD8+ T cells and have offered insight into the mechanisms that confer a functional advantage to these cells in vivo. In models of viral and intracellular bacterial infection, early effectors provide greater protective immunity, eliminate virus more efficiently, and display greater replicative capacity than late effectors [6,11,13,35]. Similarly, when adoptively transferred, early effectors induce better tumor regression [7••,8], display higher engraftment efficiency, and mediate more severe graft-versus-host disease than more highly differentiated cells [36,22••]. The mechanisms that underlie the loss of function associated with progressive differentiation are complex and include decreased survival and proliferation capacity, reduced responsiveness to homeostatic cytokines, decreased capacity for self-renewal, inability to differentiate into diverse cell types, and impairment of lymphoid tissue homing (Figure 1) [6,7••,14,15].
How can CD8+ T-cell differentiation be directed to generate early effectors that are programmed for optimal immune function? Stimulation of CD8+ T cells through TCRs, co-stimulatory molecules and cytokines program changes in gene expression that can be heritable owing to epigenetic modifications in gene transcription [3–5,37]. Although the precise mechanisms are largely unknown, it is likely that these changes include DNA methylation, methyl-CpG-binding proteins, and histone modifications that affect the accessibility to regulatory regions of transcription factors that serve as ‘master regulators’ [4,5].
Advances have been made in identifying the master regulator transcription factors that govern T-cell differentiation [4,5,38–40]. GATA-binding protein 3 (GATA-3) induces the uncommitted CD4+ T cells that emerge from the thymus to differentiate into IL-4-, IL-5- and IL-13-releasing T helper (Th)2 cells, whereas T-bet (encoded by Tbx21−/−) promotes differentiation into IFN-γ -releasing Th1 cells. Expression of these master regulators, and therefore lineage commitment, is stabilized by amplification loops and by repression of alternative pathway genes . Nevertheless, the functionality of cells polarized to the Th1 lineage by IFN-γ or the Th2 lineage by IL-4 can be partially reversed by switching to the opposite polarizing cytokine; however, this plasticity decreases with progressive differentiation [38,41].
Less is known about lineage commitment in CD8+ T cells. Commitment to the CD8+ effector cell linage is redundantly determined by T-bet and the T-bet paralog eomesodermin (Eomes). Expression of these transcription factors is also important for maintenance of committed cells. Eomes+/−Tbx21−/− mice are deficient in IL-15-dependent lymphocyte lineages including CD8+ memory cells [42••]. Paralleling memory B-cell development, maintenance of T-cell memory might also require transcriptional repressors that arrest the differentiation process. The transcriptional repressor B-cell CLL/lymphoma 6 (Bcl-6) represses B lymphocyte-induced maturation protein-1 (Blimp-1), a transcriptional activator responsible for plasma-cell differentiation. This repression arrests the differentiation of germinal center B cells, thus enabling the generation and maintenance of B-cell memory . Recent studies in Bcl6−/− and Bcl6 transgenic mice have also revealed a role for Bcl-6 in CD8+ T-cell memory formation [43,44]. Furthermore, a Bcl-6 homologue, Bcl-6b, was recently reported to be important in maintaining memory CD8+ T-cell replication potential . Transcription-repressing isoforms of lymphoid enhancer-binding factor 1 (Lef1) and transcription factor 7 (Tcf7), which maintain hematopoietic stem cells in an undifferentiated pluripotent state, might also be required to arrest differentiation and to maintain memory CD8+ T cells . These molecules are highly expressed in naïve and TCM CD8+ T cells, and their expression decreases with progressive differentiation . Furthermore, T cells from TCF7−/− mice spontaneously differentiate more rapidly than cells from TCF7+/− littermates, supporting the hypothesis that TCF7 is important in preventing differentiation . New data derived from the analysis of global gene expression are consistent with the hypothesis that hematopoietic stem cells and memory T and B cells — the only cells of the hematopoietic system able to undergo self-renewal for the lifetime of the organism — might share a common pattern of gene expression [49•]. We are just beginning to elucidate how transcription factors determine T-cell fate. Understanding how these master regulators guide T-cell differentiation is crucial to our efforts to generate optimal effector cells.
Membrane-bound and soluble factors direct programmed changes in CD8+ T-cell development. Increases in duration, magnitude and frequency of the TCR stimulus drive progressive differentiation [2,3,50,51]. The TCR signal is integrated with signals from diverse costimulatory, inhibitory and cytokine receptors. The program imparted by the TCR is sustained by demethylation of the IL-2 promoter, and, depending on the strength of the antigen stimulus, one or both IL-2 alleles can be activated [52,53]. IL-2 expression is also regulated by the costimulatory molecule CD28, which induces histone acetylation and loss of cytosine methylation at the IL-2 promoter/enhancer . Costimulation through CD28 also induces preferential differentiation of T cells into TCM, revealing some of the complexity of the differentiation process .
The activities of diverse soluble factors help to shape the complex differentiation process of T-cell differentiation. The impact of distinct cytokines on lineage commitment decisions is better established in CD4+ than CD8+ T cells. IFN-γ, acting through T-bet, induces naïve CD4+ T cells to become IFN-γ-releasing Th1 cells, whereas IL-4, acting through GATA-3, programs differentiation into IL-4-, IL-5- and IL-13-releasing Th2 cells. Evidence for additional CD4+ lineages generated under the influence of diverse cytokines is now emerging: in the absence of IFN-γ and IL-4, IL-23 can induce naïve CD4+ T cells to differentiate into IL-17-releasing Th-17 cells . Alternatively, Th-17 cells might be induced by the exposure to TGF-β and IL-6. Transcriptional factors responsible for the generation of Th-17 cells have not been yet identified, but it appears that neither GATA-3 nor T-bet play a role . Furthermore, TGF-β and IL-10 might influence CD4+ T cells to acquire the regulatory attributes of Th3 and Tr1 cells, respectively [57,58].
Lineage commitment by CD8+ T cells is also affected by the cytokine milieu. CD8+ T cells can be induced to differentiate into cytotoxic T cells secreting Th1-like (Tc1) or Th2-like (Tc2) cytokines by Th1-and Th2-polarizing cytokines respectively [59,60]. However, just as it is becoming clearer that CD4+ T cells differentiate along manifold lineages, it seems likely that CD8+ T-cell differentiation occurs along multiple lines (Figure 2). Many cytokines, including the common gamma-chain (γC) cytokines, are integral to CD8+ T-cell differentiation . For example, IL-15 directs CD8+ T cells to preferentially differentiate into TCM whereas IL-2 promotes them to differentiate into TEM. These phenotypic and functional differences are reflected in the distinct gene expression patterns of IL-15- and IL-2-programmed cells [7••,8,62,63].
The most recently discovered of the γC cytokines, IL-21, also confers particular features to CD8+ T cells. IL-21 generates CD8+ T cells that have a distinct CD45RO+CD28hi stable phenotype. In contrast to cells grown in IL-2, IL-21-programmed CD8+ T cells retain their capacity to produce IL-2 after antigen exposure [64•,65•]. The pattern of global gene expression by IL-21-programmed CD8+ T cells is unique compared with IL-2-, IL-7- and IL-15-programmed cells, and is characterized by simultaneously increased transcription of genes that encode lymphoid homing molecules and cytolytic effector molecules (CSH, unpublished). This surprising finding challenges the paradigm of linear differentiation for CD8+ T cells, which is characterized by an inverse relationship between the expression of genes for lymphoid homing and for effector function. Thus, differentiation of T cells appears to be multidimensional and is influenced by a host of soluble and cell-bound factors that we have only begun to explore [66–69].
A deeper understanding of programming and differentiation of CD8+ T cells might be valuable in the development of adoptive cell transfer-based immunotherapies for treatment of cancer and chronic infectious disease. It now seems clear that cells at early stages of differentiation have enhanced therapeutic efficacy, but attempts to generate cells that have desired programs have only just begun.
A significant problem in the translation of these ideas into human immunotherapies is that T-cell populations specific for tumor-associated antigens and antigens expressed by chronic pathogens are generally terminally differentiated (exhausted). Recent work suggests it might be possible to reprogram exhausted cells by blockade of negative regulatory receptors, such as programmed death-1 (PD-1) [70••]. Alternatively, antigen-specific cells that have optimal differentiation potential might be generated through gene transfer technology. Although this technology is still evolving, high-efficiency transfer of genetic sequences for specific TCRs into the genome of naïve or less-differentiated CD8+ T cells might soon be possible. In the future, one might be able to attenuate or even reverse progressive differentiation of T cells through manipulation of transcriptional master regulators. The reprogramming of B cells could now be a reality [71••]. As our understanding of programming and differentiation signals improves we might be able to generate a greater diversity of cells and apply them to the treatment of cancer and chronic infectious diseases.
This work was supported by the Intramural Research Program of the Center for Cancer Research, NCI, NIH. The authors thank Alan Hoofring for his excellent artistic contributions to the illustrations in this manuscript. The authors also thank Steven A Rosenberg, James C Yang, Richard M Sherry and Mark E Dudley for helpful discussions.
This review comes from a themed issue on Lymphocyte effector functions Edited by Stephen Schoenberger
Papers of particular interest, published within the annual period of review, have been highlighted as:
• of special interest
•• of outstanding interest