Immune responses to infection are optimally designed to generate large numbers of effector T cells while simultaneously minimizing the collateral damage of their potentially lethal actions and generating memory T cells to protect against subsequent encounter with pathogens. Much remains to be discovered about how these equally essential processes are balanced to enhance health and longevity and, more specifically, what factors control effector T cell expansion, differentiation, and memory cell formation. The innate immune system plays a prominent role in the delicate balance of these decisions. Insights into these questions from recent work in the area of effector CD8 T cell differentiation will be discussed.
The generation of long-term immunity and efficacious vaccines against many viral and bacterial agents is dependent on the formation of large numbers of long-lived memory CD8 T cells. Memory CD8 T cells have special features that make them well suited to respond quickly and effectively to reinfection. Compared with naive CD8 T cells, Ag-specific memory CD8 T cells are present in greater number and have an increased ability to survey nonlymphoid (peripheral) sites for the presence of infection (1). Memory CD8 T cells appear poised for a rapid response to secondary infection because they persist in a “progrowth” state with low levels of p27kip and high levels of cyclin-dependent kinase 6 (CDK6)-cyclin D3 complexes (2, 3) and maintain mRNA expression of several cytotoxic proteins, antiviral cytokines, and chemokines (1, 3, 4, 5, 6, 7). These features allow the memory CD8 T cells to expand and develop effector functions faster than naive CD8 T cells. Memory CD8 T cells can also survive long term in the absence of Ag (>2 years in mice and >50 years in humans) (8, 9) and are maintained through a process of self-renewal driven by IL-15 and IL-7 (also called homeostatic proliferation or turnover; Ref. 10). Given their long lifespan and ability to provide protection against recurrent infections, increasing the quantity and quality of memory CD8 T cells is paramount to improving the efficacy of most vaccines.
Acquiring a terminally differentiated effector CD8 T cell state
In general, terminal differentiation of specialized cell types begins with immature precursor cells that progressively acquire (often in a stepwise manner) a distinguished gene expression profile, a specialized set of functions, and an altered morphology. As terminal differentiation proceeds, cells typically display a loss of multipotency (i.e., the ability to become another cell type), proliferative capacity, and telomerase activity (11). The lifespan of terminally differentiated cells can vary quite dramatically from relatively long-lived fates, such as neurons or Ab-secreting plasma cells that live for many years, to short-lived fates, such as neutrophils, intestinal villi, or epidermal skin cells that live for a few hours, days, or weeks, respectively. What properties distinguish terminally differentiated CD8 T cells and how this process influences memory CD8 T cell development are important questions to understand.
In some ways, a mature CD8+ thymic emigrant might be considered a terminally differentiated cell type because it has acquired a distinct and stable naive T cell phenotype, can no longer differentiate into other types of lymphocytes, and does not express telomerase (12, 13). However, the naive stage is not necessarily a terminal endpoint, because when activated by Ag this cell proliferates profoundly and differentiates into an effector CD8 T cell that acquires a different pattern of gene expression, a more specialized set of functions (such as cytotoxicity and antiviral cytokine production), and telomerase activity (4, 14). However, after several rounds of cell division the proliferative potential of effector CD8 T cells declines and they become highly sensitive to cell death (15). For these latter reasons, effector CD8 T cell could be considered terminally differentiated, but if this is the case then how do long-lived memory CD8 T cells arise from this population?
Compared with effector T cells, memory CD8 T cells exhibit less terminally differentiated phenotypes because they are multipotent (they can remain resting memory cells or redifferentiate into cytotoxic effector cells), can self-renew, have a high proliferative potential and increased longevity (1, 16). In addition to these “stem cell-like” properties, there are many notable differences between the memory and effector CD8 T cell gene expression profiles (4, 5). Yet, the memory cell gene expression pattern also bears a strong resemblance to terminally differentiated effector cells (such as the sustained expression of many “effector-like” mRNAs such as IFN-γ, granzymes, perforin, chemokines, and chemokine receptors (3, 4, 17) and the continued expression of telomerase (14). Thus, memory CD8 T cells constitute a unique cell population whose members appear less differentiated than their terminally differentiated effector cell predecessors with regard to proliferative potential and multipotency, yet are imprinted with several key effector cell traits. This review will discuss potential mechanisms for how some effector T cells may be preserved in a less differentiated state during an immune response.
The heterogeneity of effector and memory CD8 T cells
One prominent aspect that has come to light over the past few years is the heterogeneity of effector and memory T cells, and understanding the ontogeny of these different cell types will lend insight into memory T cell development. During many infections and immunizations, diverse effector CD8 T cell populations form that can sometimes be separated into cell subsets that differ in their anatomical locations, effector functions, proliferative capacity, and long-term fates. Several markers have become useful in this regard such as CD62L, CCR7, CD103, and α4β7 (anatomical localization), IL-2/IFN-γ/TNF-α, perforin, granzymes A/B/C/K, and programmed death-1 (effector functions), and Bcl-2, IL-7R, CD122, CD28, CD57, CD27, killer cell lectin-like receptor G1 (KLRG1),3 CXCR3, CD43, and CD62L (survival and/or proliferative capacity). Other markers such as Ly6C, NKG2A, and 2B4 can show heterogeneous expression among effector CD8 T cells but have not yet been attributed to a cellular process (3, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33). Various factors have been recognized to affect the expression of these markers such as the type or duration of infection, certain common γ-chain and inflammatory cytokines, Ag specificity, naive T cell precursor frequency, and their location within the body (see Refs. 7, 19, 20, 21, 22, 26, 27, 28, 29, 30, 31 , and 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46).
With respect to memory CD8 T cell development, our ability to distinguish effector CD8 T cells that are destined to survive and become long-lived memory CD8 T cells (referred to here as memory precursor effector cells; MPECs) from those that are not (referred to as short-lived effector cells; SLECs) has markedly improved in several experimental systems using a combination of functional criteria and surface marker expression. Some studies from mice and nonhuman primates have noted that CD8 and CD4 T cells capable of producing IL-2 or simultaneously producing IFN-γ, TNF-α, and IL-2 (triple producers) preferentially survive and provide greater protection against rechallenge compared with those that only produce one cytokine (21, 30, 47, 48). Other work has found that during acute lymphocytic choriomeningitis virus (LCMV) and Listeria monocytogenes infections in mice, the expression of the receptors IL-7R and KLRG1 are very valuable for detecting MPECs and SLECs (23, 30, 49). For the most part, IL-7R and KLRG1 are inversely expressed on MPECs and SLECs, with SLECs being KLRG1highIL-7Rlow and MPECs being KLRG1lowIL-7Rhigh (30). However, it is important to note that the combination of KLRG1 and IL-7R alone are still not exact indicators of SLECs and MPECs because not all IL-7Rhigh effector CD8 T cells become memory cells, some KLRG1high IL-7Rlow cells persist for some time following primary and/or secondary infections, and some effector and memory CD8 T cells express both KLRG1 and IL-7R (23, 30). Moreover, noninfectious methods of T cell activation (e.g., dendritic cell vaccination) can generate nearly uniform IL-7Rhigh MPEC-like CD8 T cell populations, but many of these cells do not become long-lived memory CD8 T cells (41, 42, 50, 51). In this review, the IL-6 receptor may also help to distinguish memory precursor cells (51). These results suggest that the effector CD8 T cells may reside in a range of differentiated states spanning the SLEC and MPEC states, and more attributes are necessary to refine the identity of these important cell types.
The diversity in effector CD8 T cells is likely to be related to that found in the memory CD8 T cell population, but it has been difficult to track the developmental lineage of CD8 T cells between these two stages because the memory T cell population appears to continuously evolve over time following infections (19, 29, 30, 52, 53). Multiple subsets of memory T cells have been characterized, most notably, the central (TCM) and effector (TEM) memory T cells (based on CD62L and CCR7 expression) (18). Compared with TCM cells, TEM cells have an decreased ability to traffic to lymphoid tissues and appear more terminally differentiated because, in many cases but not all, they have a lower capacity to proliferate to Ag and homeostatic cytokines (IL-15 and IL-7) and shortened telomere length (18, 19, 54). However, even within the TCM and TEM categories there seems to be additional heterogeneity. For example, studies from Sendai virus infection found that the TCM cells persisting a year after infection had a higher proliferative capacity than TCM cells found one month after infection (55). This data is somewhat akin to that found following LCMV infection (4). When these Sendai-specific memory CD8 T cells were interrogated further, the cells expressing heightened levels of CD27 and CXCR3 demonstrated the greatest proliferative potential, yet they did not fall into the traditional TCM or TEM subsets (29).
Understanding the derivation of the memory CD8 T cell subsets is complicated because although several cell surface phenotypes are stable, some are not. For example, following acute LCMV infection the expression of KLRG1 and IL-7R appears quite stable and little conversion is observed between the SLEC and MPEC subsets (23, 30). However, the expression of CD62L and CD27 can vary according to time and/or tissue residence (19, 36, 44, 45). LCMV-specific KLRG1highIL-7Rlow SLECs that persist into “memory” time points retain a TEM cell phenotype (CD62Llow, CD27low, IL-2low, bcl-2low; low proliferative capacity), but the memory cells generated from MPECs are often a mixture of TEM and TCM (23, 30, 32, 33, 41, 45) (N.S.J. and S.M.K., unpublished data). Furthermore, it has been noted by several groups that following some acute infections the memory CD8 T cell population evolves progressively over time into one that is enriched for cells of a higher proliferative capacity (to Ag and homeostatic cytokines) and greater longevity that can be characterized as IL-7Rhigh, KLRG1low, CD27high, CD62Lhigh, CXCR3high, IL-2high, and Bcl-2high (19, 22, 23, 25, 29, 30, 52). This trend can be altered in latent or chronic infections, and the stability of certain memory CD8 T cell subsets may be tailored by the type of viral infection (1, 46, 53). In addition to better understanding the origin of different memory T cell subsets, it is also important to determine their relevance and roles in providing long-term protection.
Developmental lineages of memory CD8 T cells
How is effector CD8 T cell differentiation balanced to permit the formation of effector cell properties in the MPECs and yet prevent them from acquiring a terminal SLEC state? Data from many murine model systems of infection indicate that the precursors to memory CD8 T cells acquire a wide range of potent effector cell qualities and therefore do not “bypass” the effector cell stage. Evidence for this point is illustrated by the memory CD8 T cell gene expression patterns discussed above and several lineage-tracing experiments that showed that at some point memory CD8 T cells expressed effector molecules (4, 5, 56, 57, 58). Moreover, LCMV-specific KLRG1lowIL-7Rhigh MPECs perform most effector functions as well as SLECs (23, 30). This point should be kept in mind when discussing activated T cell fate decisions, because other frequently used terms such as “memory” or “effector” fates, may be misleading as they imply that the “memory” fated CD8 T cells do not acquire effector T cell properties. Indeed, other scenarios may exist during some types of noninfectious immunizations (with weaker inflammatory or antigenic stimulations) where the memory CD8 T cell precursor population does not develop a full-fledged set of effector cell properties. Under these conditions though, factors that enhance the differentiation of effector functions often promote memory CD8 T cell formation (15, 58, 59, 60). Nonetheless, based on the studies made during infection it is clear that the expression of effector molecules does not preclude memory CD8 T cell formation.
Certain developmental models may help to explain how some effector CD8 T cells gain or maintain memory cell potential whereas others do not (Fig. 1). One model may be that all CD8 T cells reach a terminally differentiated effector stage but then some are capable of dedifferentiating into cells that gain longevity and a high proliferative potential (Fig. 1 A). Experimental evidence supports a dedifferentiation model to a certain degree because the surviving IL-7Rhigh MPECs “functionally mature” during the effector to memory (E→M) transition period and gradually acquire an increased capacity to proliferate and produce IL-2, Bcl-2, and CD62L (4, 19, 23, 25, 61). What instructs this functional maturation is unclear, but IL-2 exposure during infection, the presence of CD4 T cells, and maintaining lower expression of T-bet appear to be important for establishing a memory CD8 T cell population with a higher proliferative capacity (61, 62, 63).
A second model for generating CD8 T cell diversity is that an activated CD8 T cell differentiates in a stepwise manner and, in so doing, progressively acquires a more terminally differentiated phenotype; the cells that become MPECs do not progress as far as the SLECs (Fig. 1 B). In this model, often referred to as the “decreasing potential model,” the progression of differentiation may be controlled by successive stimulations with Ag or other signals (64). Thus, the differentiation state of an effector CD8 T cell is reflective of the cumulative history of signals that were encountered during infection. Such a linear model provides a nice mechanism for generating a spectrum of different types of effector and memory CD8 T cells (36, 64, 65, 66). One requisite of the linear model is that all activated CD8 T cells transition through an MPEC stage as they terminally differentiate; if true, then controlling the progression of terminal differentiation might enable the generation of more memory CD8 T cells. Some evidence for this model may be found in studies in which the duration of infection is truncated and activated CD8 T cells develop into protective memory CD8 T cells more efficiently and rapidly, but to date these approaches have not resulted in a larger number of memory CD8 T cells (30, 41, 42, 43, 67, 68, 69). This is likely because a major regulator of clonal expansion is the load or duration of antigenic stimulation (70, 71, 72). Therefore, to generate more memory CD8 T cells one may need to find a way to confine terminal differentiation without minimizing clonal expansion.
A third working model for generating CD8 T cell diversity incorporates divergent developmental pathways and proposes that soon after activation daughter cells are instructed to generate SLEC or MPEC fates (Fig. 1 C). The divergent model need not be completely asymmetric to only produce two cell types but could also incorporate a range of differentiated states according to the overall strength of signal a T cells sees at or near the time of priming (43, 73, 74). Experimental support for this model may be found in a recent study examining the first cell division of an activated CD8 T cell where it appeared to divide in an asymmetric manner producing daughter cells that have increased or decreased memory CD8 T cell potential (75). Other recent work tracking the formation of MPECs and SLECs during LCMV and Listeria infections shows that these two lineages begin to diverge from a common pool of KLRG1low effector cells after >7–10 cell divisions and 4 days of infection (30, 33). In line with this model, several studies indicate that early and brief exposure to Ag and inflammation are sufficient to direct the formation of long-lived and short-lived effector CD8 T cell fates (30, 42, 43, 70).
Clarifying the precise mechanism(s) at play will require elucidation of the numerous genetic pathways that control the balance between terminal differentiation and memory CD8 T cell developmental potential, and in so doing, it is likely that some aspects of all of the above models will be incorporated in this process depending on the cellular phenotype being analyzed. Moreover, the inductive signals that control these cell fate decisions are likely to differ and/or be produced with varying kinetics depending on the infectious pathogen and its tropism. A case in point is that the proportion of IL-7Rlow Ag-specific effector CD8 T cells formed varies across multiple types of infections (LCMV, Listeria, influenza, and vaccinia virus) and between tissues within a given infection (15, 29, 30, 43, 76, 77, 78).
Cytokine control of effector and memory CD8 T cell development
Perhaps our knowledge of effector CD8 T cell differentiation can be enhanced by that of their kindred lymphocytes, CD4 T cells. In many ways one can consider CD4 T cell differentiation as divergent, because under different priming conditions the same naive CD4 T cell can adopt one of several effector cell lineages according to the type of lineage-determining cytokines it was exposed to during activation (e.g., Th1, Th2, Th17, and regulatory T cell specifications are directed by IFN-γ/IL-12, IL-4, IL-6/TGFβ/IL-1β, and TGFβ exposure, respectively) (79).
Although the repertoire of effector CD8 T cells may not be as diverse as that of CD4 T cells, a handful of inflammatory cytokines have been found to influence CD8 T cell differentiation (80). As with Th1 cells, the inflammatory cytokines IL-12, IFN-γ, and IFNαβ potently enhance effector CD8 T cell expansion, cytotoxicity, and production of antiviral cytokines, particularly when CD8 T cells are activated by a weak stimuli or cross-presented Ag (15, 79, 81, 82, 83, 84, 85, 86). Interestingly, the reliance of CD8 T cell expansion on particular inflammatory cytokines (IFNαβ, IFN-γ, and IL-12) or IL-2 may depend on the type of infectious pathogen or the tissues they reside in, respectively (83, 85, 87, 88). Additionally, inflammatory cytokines, particularly IL-12, induce the expression of Bcl-3, which can enhance activated CD8 T cell expansion (89, 90). Some inflammatory cytokines can also modulate the expression of key transcription factors that regulate effector T cell differentiation. T-bet is a major regulator of Th1 effector cell differentiation, whereas T-bet and eomesodermin (eomes), another T-box family member, appear to coordinately regulate the formation of effector CD8 T cell functions (91, 92). IFN-γ is critical for T-bet induction and IL-12 acts to augment its expression in CD4 T cells (79); but in CD8 T cells, IL-12 augments T-bet expression and diminishes eomes expression (30, 86).
The above data emphasize the beneficial role of inflammatory cytokines on maximizing effector T cell expansion and differentiation, but these cytokines may also act as a double-edged sword; on one hand they stimulate effector CD8 T cell function and expansion, but on the other they appear to drive terminal maturation and limit memory cell potential. Recent work has shown that inflammation, most notably by IFN-γ, promotes effector CD8 T cell contraction and down-regulation of IL-7R (84). However, another recent report showed that IFN-γR was needed in a CD8 T cell-autonomous manner for normal memory CD8 T cell formation, so the mechanism of IFN-γ action in different types of immune responses remains to be clarified (93). In Il12−/− mice, a higher frequency of IL-7Rhigh MPECs and memory CD8 T cells form following Listeria infection (85, 86), suggesting that IL-12 plays a critical role in the effector CD8 T cell fate decisions. Our recent work has shown that IL-12 can induce T-bet expression in a dose-dependent manner in activated CD8 T cells and that T-bet may act like a rheostat to determine SLEC vs MPEC fate decisions, with high T-bet expression favoring the formation of KLRG1highIL-7Rlow SLECs and lower expression favoring KLRG1lowIL-7Rhigh MPECs (30). A similar pathway using IL-12 (and IL-18) to drive T-bet expression may be required for the terminal differentiation of short-lived KLRG1high NK cells during murine CMV infection (94, 95, 96, 97). These data demonstrate multiple modes of T-bet function. In both CD4 and CD8 T cells T-bet acts in a divergent manner to specify different cell fates, but in NK cells it may promote terminal maturation in a linear manner. Moreover, in CD4 T cells T-bet operates fairly asymmetrically to specify Th1 vs Th2, Th17, and regulatory T cell fates, whereas in CD8 T cells it operates according to an expression gradient to specify SLEC vs MPEC fates.
Linking the inflammatory cytokines and lineage-determining transcription factors (such as T-bet) to both the development of effector functions and longevity provides a unique way for the innate immune response to regulate effector T cell homeostasis. Additional support for this model may be found in recent studies where specific TLR agonists differentially influenced the generation of long-lived IFN-γ/TNF-α/IL-2 triple-producer CD4 and CD8 T cells. Because T-bet repress IL-2, it will be interesting to determine whether the “memory-enhancing” adjuvants modulated T-bet expression levels in a manner that promoted both IL-2 production and memory cell generation (21, 48, 91). It is most certain that other cytokines aside from IL-12 and IFN-γ will similarly regulate memory T cell potential in activated CD8 T cells, because not all infections and adjuvants elicit the same cytokine profiles. Also, the role of tolerogenic cytokines, such as IL-10, which can regulate IL-12 production, needs to be considered in this process. For example, in Il10−/− mice fewer memory CD8 T cells form following Listeria infection, but this is likely not due to direct effects of IL-10 on the CD8 T cells (98, 99). These data suggest that the balance between IL-12 and IL-10 expression may represent a potential regulatory axis in SLEC and MPEC development during certain types of infections.
The role of other cell intrinsic factors in effector CD8 T cell terminal differentiation
Many other important cell-intrinsic factors regulate the memory potential of developing effector CD8 T cells. Similar to T-bet, the transcriptional regulator ID2 (inhibitor of differentiation 2) plays a critical role in the development of NK and CD8 T cells (100), and ID2-deficient CD8 T cells have a more IL-7RhighCD27high“MPEC-like” phenotype (101). Another transcription factor, Blimp-1 (prdm1), critical for the terminal differentiation of B cells into plasma cells, is expressed at high levels by IL-7Rlow SLECs (63, 102). CD8 T cells in Blimp1-deficient mice have an activated and highly proliferative phenotype, suggesting that Blimp1 expression is antiproliferative (103, 104). Moreover, potential antagonists to Blimp-1, Bcl-6, and its homologue, Bcl-6b, promote memory CD8 T cell development and increase proliferative responses (105, 106). Further evidence suggests that the reduced proliferative potential of KLRG1high CD8 T cells is regulated by their increased expression of the cell cycle inhibitor p27kip (52) and their reduced ability to express Bmi-1, a transcriptional repressor that promotes T cell proliferation (107). Undoubtedly, the list of factors involved in this process will continue to grow as more genes are examined, and gene expression profiling between LCMV-specific MPECs and SLECs has revealed a handful of interesting candidates (30). It will be interesting to discover in the future whether more extensive overlap exists in the genes that regulate differentiation of effector CD8, CD4, NK T cells, and NK cells.
Regulating CD8 T cell longevity and maintenance
One aspect associated with effector CD8 T cell terminal differentiation is its shortened lifespan, but it is not clear how this process is regulated. IL-7 and IL-15 are critical for the long-term survival and homeostatic turnover of memory CD8 T cells (10), and prolonged deprivation of these cytokines has considerable consequences on the formation and or maintenance of memory CD8 T cells (22, 23, 108, 109, 110). Our work indicates that KLRG1highIL-7Rlow SLECs depend on IL-15, but this alone is not sufficient to maintain these cells long term (30). Surprisingly, IL-7R overexpression was unable to save these cells from death after acute infection (52, 111), suggesting that the down-regulation of IL-7R is symptomatic of, but not causal to, their death. Moreover, increasing the expression of Bcl-2 and Bcl-XL or blocking their actions in effector CD8 T cells does not greatly affect the normal rate of effector cell death following infection (10, 22, 52, 109, 112, 113). Currently, key molecules that have been found to promote and prevent effector CD8 T cell contraction are the proapoptotic molecule Bim and the serine protease inhibitor 2A (Spi2A), respectively (114, 115, 116, 117, 118). Together these data suggest that MPECs need to see IL-7 and IL-15 to become long-lived memory CD8 T cells, but SLECs die for reasons other than deprivation of these cytokines.
Implications of effector CD8 T cell terminal differentiation on vaccine design
Understanding the factors that regulate the terminal differentiation of effector CD8 T cells is critical for designing vaccines that generate long-lasting CD8 T cell immunity. Much of the recent work in the vaccine field has focused on generating larger expansions of CD8 T cell subsets because of the direct relationship between effector cell expansion and memory formation (119). A number of adjuvants, being potent inducers of effector CD8 T cell expansion and effector function, have logically been included in a variety of vaccine formulations. However, in light of the additional role that some adjuvants may play in restricting memory CD8 T cell potential, it will be critical to find the proper adjuvants and their respective doses that balance the effector T cell expansion and terminal differentiation best for vaccines and other immunotherapies. Moreover, it is important to remember that studying the phenotype of effector CD8 T cells generated in response to vaccines may be as important as studying the number of CD8 T cells generated in response to vaccination, as demonstrated already by some experimental and clinical data (21, 53, 65).
The authors have no financial conflict of interest.
This work was supported by the National Institutes of Health Grants AI066232 and T32 AI055403, the Burroughs-Wellcome Fund Grant 1004313, the Edward Mallinckrodt, Jr. Foundation, the Cancer Research Institute, and the Richard K. Gershon Predoctoral Fellowship.
Abbreviations used in this paper: KLRG1, killer cell lectin-like receptor; LCMV, lymphocytic choriomeningitis virus; MPEC, memory precursor effector cell; SLEC, short-lived effector cell; TCM, central memory T cell; TEM, effector memory T cell.