Generation of effector CD8⁺ T cells and their conversion to memory T cells

Cardinal features of memory T cells

The formation of immunological memory following acute infections or immunization is a hallmark of adaptive immunity, whereby an organism ‘remembers’ the original pathogen encountered and mounts more robust humoral and cellular responses to rapidly control reinfection and reduce the severity of disease (1–6). Immunological memory mainly consists of long-lived plasma cells that constitutively secrete high-affinity neutralizing antibodies, and memory B and T cells (5). Although the concepts and principles of immunological memory have governed the design and application of vaccines for centuries, the underlying mechanisms that regulate memory B- and T-cell formation remain poorly defined.

During viral infection, circulating virus-specific naive T cells engage with antigen-presenting dendritic cells (DCs) via the formation of immunological synapse (7–9). After this first 18- to 24-h initiation stage, activated T cells start to rapidly proliferate (10, 11). Pathogen-specific T cells can expand approximately 10,000 fold at a rate of 4–6 h per cell cycle and complete up to 20 cell divisions in a week’s time, during which they also differentiate and acquire effector functions and the ability to migrate to the sites of infection (3–5). These functional cytotoxic T lymphocytes (CTLs) then elaborate cytokines and granzymes to kill pathogen-infected cells and control the infection (3, 6). Following pathogen clearance, effector CD8⁺ T cells undergo equally precipitous contraction, wherein the majority of pathogen-specific effector CD8⁺ T cells die via apoptosis, but approximately 5–10% of them survive to further mature into long-lived protective memory CD8⁺ T cells (4–6). Memory CD8⁺ T cells can persist for very long periods of time in the absence of antigen (Ag) and provide life-long protection in mice and humans (12, 13). Typically, the number of memory CD8⁺ T cells remains quite stable overtime, largely through interactions with interleukin-7 (IL-7) and IL-15 that mediate survival and self-renewal (referred to as homeostatic proliferation or turnover) of memory CD8⁺ T cells (3, 6, 14).

As memory CD8⁺ T cells differentiate, a number of interesting stem cell-like properties are bestowed onto these cells, such as longevity, telomerase expression, and the ability to self-renew themselves through homeostatic turnover (4, 6). In addition, memory CD8⁺ T cells exist in a poised multipotent state because they persist as resting memory CD8⁺ T cells, but, upon reinfection and antigenic stimulation, they have the capacity to rapidly proliferate and differentiate into secondary effector CD8⁺ T cells (3, 6, 14). The combination of these functional attributes, with the sheer increase in the relative abundance of memory T cells compared with an unimmunized individual, constitute the basis of long-term T-cell-mediated immunity. In this review, we aim to summarize the latest progress made in the field with a focus on CD8⁺ T-cell memory ontogeny and maintenance.

Generating effector CD8⁺ T-cell heterogeneity and memory cell potential

It is now evident that the populations of antigen-specific CD8⁺ T cells are quite heterogeneous, and not all effector CD8⁺ T cells, and memory CD8⁺ T cells for that matter, acquire these quintessential memory cell properties equivalently (6, 15). As effector CD8⁺ T cells expand and differentiate during a primary response most cells terminally differentiate into end-stage effectors that have a shortened lifespan and die following infection, whereas a smaller subset of cells differentiates into memory cell precursors that have increased memory cell potential (6, 15–17). An elegant study utilizing the transfer of single naive CD8⁺ T cells demonstrated that an individual precursor cell has the capacity to give rise to these differently fated effector cell populations and memory CD8⁺ T cells, indicating that naive CD8⁺ T cells are not pre-programmed to develop into effector or memory cells and that the specification of these fates occurred after T-cell activation (18). However, the diversity of the effector cell population is not limited to just these two cell subsets because additional types of effector CD8⁺ T cells exist that appear to display a mix of terminally differentiated and memory precursor cell properties and have intermediate lifespans and proliferative responses to secondary infections (19, 20, N.S. Joshi and S.M. Kaech, unpublished data). Thus, perhaps a more accurate description of effector CD8⁺ T-cell differentiation is that it occurs along a continuum with cells harboring greater memory cell potential and longevity on the one end and terminally differentiated effector cells on the other end, but in between these two extremes lies effector cells that exist in intermediate differentiation states (Fig. 1). Such a model helps to account for the numerous subsets of effector CD8⁺ T cells seen during immune responses as well as provides plasticity within an effector ‘lineage’, permitting cells to slide between differentiation states according to signal input. Additionally, it provides for the multipotency of memory cells to further differentiate along this continuum into terminal effector CD8⁺ T cells upon restimulation.

During many acute infections, a handful of markers have become helpful to dissect the effector population into subpopulations that have more or less memory cell potential (6, 15, 16, 21–23). Increased expression of IL-7Rα was the first marker found to distinguish memory precursor cells from terminally differentiated, shorter lived effector cells, and IL-7Rα is functionally required for their long-term survival (21, 23). Other proteins that co-segregate with increased IL-7Rα expression on antigen-specific CD8⁺ T cells include Bcl-2, CD27, CXCR3, and CD28 (6, 16, 22). Although the IL-7Rα^hi effector CD8⁺ T cells preferentially survive following infection compared with IL-7Rα^lo cells, there is some death in this population as the memory CD8⁺ T cells form (6, 15). However, the memory cells that descend from the IL-7Rα^hi cells display the trademark memory cell properties most distinctly, such as increased proliferative responses to antigen, IL-2 production, and the ability to self-renew. Additionally, the IL-7Rα^hi memory CD8⁺ T cells have the greatest capacity to develop into ‘central’ memory CD8⁺ T cells (T_CM) that express lymph node-homing receptors and reside in lymphoid tissues (6) (Fig. 1).

By contrast, increased expression of CD57 and KLRG1 and decreased expression of IL-7Rα, CXCR3, CD27, and CD28 are associated with effector or memory CD8⁺ T cells that are cytotoxic and produce interferon (IFN)-γ but are relatively senescent (6, 15, 16, 24, 25). In addition, these types of CD8⁺ T cells tend to have the lower expression of IL-2 and telomerase, shortened telomeres, heightened levels of the cell cycle in inhibitor p27^kip, and decreased in AKT signaling (26–28, T.W. Hand and S.M. Kaech, unpublished data). Moreover, in several mouse models of infection, antigen-specific KLRG1-expressing effector CD8⁺ T cells that do not express IL-7Rα have reduced longevity and recall response compared with cells that are KLRG1^loIL-7R^hi (16, 21, 23, 29–33). In such settings, we have referred to KLRG1^hiIL-7Rα^lo CD8⁺ T cells as short-lived effector cells (SLECs) and KLRG1^loIL-7Rα^hi cells as to memory precursor effector cells (MPECs) (16). Lastly, some effector CD8⁺ T cells express markers of terminal differentiation (such as KLRG1), yet appear to survive the effector-to-memory transition phase. These cells tend to maintain higher expression of cytotoxic molecules longer after pathogen clearance compared with KLRG1^loIL-7R^hi memory cells, are more prevalent in peripheral tissues like the liver and lung, and can be considered as effector memory (T_EM) cells (6, 22, 34–36).

The diversity of effector and memory CD8⁺ T cells is considerably more expansive than that described above and cells that display ‘hybrid’ phenotypes are also found throughout immune responses. For instance, double-negative KLRG1^loIL-7Rα^lo effector CD8⁺ T cells are present during infection and may represent cells of an early effector state that have not reached terminal differentiation (19, 20). Moreover, double-positive KLRG1^hiIL-7Rα ^hi cells are common in resting memory CD8⁺ T-cell populations. These cells have a longer half-life than KLRG1^hiIL-7Rα ^lo cells, but it is reduced relative to KLRG1^loIL-7R^hi memory cells. Moreover, compared to KLRG1^loIL-7Ra^hi memory cells, KLRG1^hiIL-7Rα ^hi memory cells tend to proliferate less to antigen and homeostatic cytokines and produce less IL-2. Some of these types of memory CD8⁺ T cells may represent memory CD8⁺ T cells that have experienced multiple infections as these cells accumulate T_EM phenotypes, yet remain longer lived (37, 38). Although the categorization of effector and memory CD8⁺ T cells is helpful to dissect differences among subsets, further studies are required to illustrate the stabilities of each of these subsets based on the aforementioned surface makers. Furthermore, the inclusion of additional markers amplifies the heterogeneity and the spectrum of differentiation states that can be acquired by effector and memory CD8⁺ T cells.

Role of antigenic stimulation in effector expansion and differentiation

CD8⁺ T cells specific to any peptide only consist of a small fraction of naive repertoire; therefore, robust proliferation is required to sufficiently combat rapidly replicating pathogens (39). It was thought that repeated antigenic stimulation may be required for the sustained effector T-cell expansion. However, recent studies demonstrated that a brief antigen exposure could set forth an ‘autopilot’ CD8⁺ T-cell response without the continued need for antigen stimulation (40–42). Activated CD8⁺ T cells then go through a number of divisions, acquire effector function, and give rise to memory cells (40–42).

During an infection, the extent of T-cell proliferation is governed by the amount of antigen available in vivo and the strength of TCR signal. By infecting mice with recombinant vaccinia virus strains that produced either high or low quantities of an ovalbumin (OVA) epitope, one group showed that the magnitude of the responding CTL population was proportional to epitope abundance (43). The strength of antigenic stimulation for CD8⁺ T-cell activation is determined by the binding affinity of TCR to MHC I–peptide complex as well as the duration of their interaction. A recent study elegantly demonstrated that T-cell receptor (TCR) transgenic CD8⁺ T cells were able to sufficiently respond to altered peptide antigens with a broad range of affinity during Listeria infection. Although strong TCR stimulation was favored for sustained T-cell expansion, it was not a prerequisite for complete activation. Even very low affinity antigens induced detectable effector and memory CD8⁺ T-cell generation. Nevertheless, other studies have shown that a prolonged time was required to commit naive T cells to proliferate under lower antigen doses than higher antigen doses in the presence of costimulation (44–46). Taken together, these studies suggested that naive T cells can incorporate the strength and duration of TCR signals to reach a threshold of full activation, resulting in antigen-independent effector development. Importantly, this initial TCR activation also instigates the developmental program that sufficiently drives long-lived memory cell formation.

Role of costimulation in memory development and maintenance

Costimulatory molecules (CD28, CD27, 4-1BB, and OX40) expressed by CD8⁺ T cells are important for T-cell activation, expansion, survival, and memory formation (4, 47). Antigenic stimulation in the absence of costimulatory signals was thought to induce tolerance or clonal depletion (48, 49). The roles of costimulatory molecules in generating successful effector CD8⁺ T-cell response are more critical when T cells are primed weakly or of short duration. Furthermore, costimulatory molecules have been shown to be involved in memory T-cell development. For example, a recent study has indicated that OX-40-deficient mice had reduced memory CD8⁺ T-cell formation after Listeria infection (50). Aside from OX-40, other costimulatory molecules, including CD28, CD27, and 4-1BB, have also been shown to contribute to memory CD8⁺ T-cell generation and/or longevity in variety of infection models (4, 51–53).

Costimulatory signals are also required for maximal maintenance of T-cell memory. IL-7 and IL-15, cytokines that are required for memory T-cell homeostatic turnover, have been reported to upregulate the costimulatory tumor necrosis factor receptor (TNFR) family members OX40 and 4-1BB, respectively, and this may provide additional survival signals to memory CD8⁺ T cells (53, 54). Several TNFR family members, including 4-1BB, recruit TNFR-associate factor-1 (Traf-1) for downstream signaling. Like 4-1BB and OX-40 deficiency, Traf-1 deficiency has a modest but significant effect upon memory CD8⁺ T-cell formation and survival (55, 56). Finally, 4-1BB and OX-40 may positively feedback on the expression of the IL-2Rα and IL-7Rα, thus further promoting T-cell survival (57). Thus, IL-7 and IL-15 provided in certain ‘niches’ may affect the interactions between cytokines, cytokine receptors, and costimulatory receptors to sustain T-cell survival and their long-term maintenance (54).

Role of inflammation in the effector and memory cell fate decisions

It has become increasingly clear that cytokine milieu initiated through pathogen-associated molecular patterns (PAMPs) upon pathogen encounter have immediate effects on activated CD8⁺ T cells and determine the clonal burst size, acquisition of effector functions, and modulate effector versus memory cell fate decisions (3, 48, 58). Early studies suggest that signal 3 provided by proinflammatory cytokines, mainly IL-12, IFN-γ, and IFN-αβ, is required to break the tolerance when naive T cells were activated with weak TCR stimulation both in vitro and in vivo (48, 59, 60). In addition, signal 3 also enhances T-cell expansion via promoting proliferative effector CD8⁺ T cells survival during various infections (60–63). Although the underlying mechanisms are incompletely clear, some studies suggested that IL-12 might enhance activated CD8⁺ T-cell survival by upregulating Bcl3 and/or inhibiting caspase-3 catalytic function (64, 65). Another important role of signal 3 is to promote activated CD8⁺ T cells to acquire effector function by means of producing cytotoxic cytokines, such as IFN-γ and TNF-α (48, 60, 66). Additionally, it is interesting to note that the responsiveness of CD8⁺ T cells to signal 3 for their optimal expansion and survival is largely dependent on the nature of pathogens. For instance, the expansion of pathogen-specific CD8⁺ T cells that are deficient of IFN-αβ receptor was greatly reduced during lymphocytic choriomeningitis virus (LCMV) infection, but the same CD8⁺ T cells expand normally during Listeria infection. On the contrary, CD8⁺ T-cell expansion during Listeria infection relies mainly on IL-12 signal, but IFN-αβ is not necessary (3, 6, 58, 67).

Accumulating evidence supports the concept that inflammation also plays a key role in effector and memory cell fate decisions (3, 6, 15, 16, 68–72). Original studies demonstrated that truncating Listeria infection using antibiotics blunted contraction of antigen-specific CD8⁺ T cells (70, 71). This observation was further confirmed in the setting of immunization with peptide-pulsed, mature DCs, in which delivery of antigen and costimulatory signals (signals 1 and 2) sufficed to generate memory CD8⁺ T cells, but the presence of adjuvants (such as CpG) during immunization accentuated not only the expansion of activated CD8⁺ T-cell population but also its subsequent contraction (33, 69, 72). Collectively, these data suggested that certain pro-inflammatory cytokines, primarily IL-12 and IFN-γ, intensified antigen-specific CD8⁺ T-cell expansion and differentiation but also induced terminal differentiation and a shortened lifespan (3, 6, 15, 16, 71–74) (Fig. 1). In addition to IL-12 and IFN-γ, other candidate cytokines that may have a similar role in effector and memory cell fate decisions are type I IFNs, other IL-12 family members, and IL-2 (75, 76).

Owing to the early production of IFN-γ and IL-12p70 during infection, one would presume that their effects on activated CD8⁺ T-cell proliferation, differentiation, and cell fate decision appear very early during Listeria infection and certain immunizations (77, 78, W. Cui and S.M. Kaech, unpublished data). Interestingly, pathogen-specific CD8⁺ T cells become refractory to IFN-γ but remain responsive to IL-12 soon after their activation (78, W. Cui and S.M. Kaech, unpublished observation). This observation suggests that IFN-γ and IL-12 may regulate effector and memory CD8⁺ T-cell differentiation in a coordinated manner. (77, 79). IFN-γ is produced mainly by NK cells during the initial stage of innate response to Listeria infection (72, 80) and plays an important role in the first line of defense by directly inhibiting viral or intracellular bacterial replication, upregulating major histocompatibility complex class II (MHC II) and macrophage function (81, 82). IFN-γ also enhances IL-12p70 production from macrophages and DCs by inducing IL-12p35 expression (72), and this creates a positive circuit in which IL-12 feeds back to further enhance IFN-γ production by natural killer (NK) cells (72, 77). IL-12 then directly acts on activated CD8⁺ T cells to augment their proliferation, differentiation, and formation of terminally differentiated CTLs (16, 32, 66, 72–74).

Timing and coupling of signals 1, 2, and 3

The above findings demonstrate the close relationship between CD8⁺ T cells and antigen-presenting cells and NK cells (and probably other innate cells), and the cytokines they produce, such as IFN-γ and IL-12, during infection or immunization. Moreover, early studies have shown clear synergy between signals 1, 2 and 3 and that their coordinated exposure renders complete T-cell activation, proliferation, and development of cytotoxic activity (3, 6, 48). Interestingly, the coupling of these signals also modulates memory and effector cell fate decisions. Recent studies from our laboratory and others have shown that antigen stimulation and inflammatory cytokines, such as IL-12, need to be delivered simultaneously to induce maximal effector CD8⁺ T-cell differentiation and loss of memory cell properties (30, 33, 72). Likewise, shortening the duration of infection impedes the formation of KLRG1^hiIL-7R^loCD27^loIL-2^lo and effector cells and hastens the formation of memory cell properties (16, 33, 72, 83). The coupling of these signals during infection probably also explains why ‘latecomer’ cells (i.e. naive CD8⁺ T cells that are activated 2–3 days following the start of infection) preferentially generate memory-like CD8⁺ T cells; in this scenario, the initial peak of pro-inflammatory cytokines is in decline when the CD8⁺ T cells are activated (83, 84).

Although the precise details for how signals 1 and 3 integrate to influence effector and memory cell potential are unknown, one could postulate that TCR signals open up key cell fate-determining loci that are targets of IL-12/signal transducer and activator of transcription 4 (STAT4) signals. An example of such transcriptional regulation is IL-12Rβ2 itself. IL-12Rβ2 is silenced in naive T cells due to the inaccessible nature of its chromatin structure, but TCR activation is necessary and sufficient to open IL-12Rβ2 regulatory elements, through BAF complex-mediated modeling leading to early IL-12Rβ2 expression (85). Subsequent IL-12 signaling and STAT4 activation further augments IL-12Rβ2 expression, directly leading to enhanced IL-12 signaling and terminal differentiation (85).

Transcriptional regulation in effector and memory cell fate decisions

An activated CD8⁺ T cell is exposed to a myriad array of signals, such as cytokines, chemokines, growth factors, nutrients, and environmental cues. How these signals are transmitted into cells and translated into gene expression patterns that promote effector differentiation yet also preserve a long-lived and multipotent pool of cells that can self-renew is an important and complicated question. Here, we summarize the role of a few transcriptional regulators identified in this process during acute infections in mice (Fig. 1).

T-bet (encoded by Tbx21) was originally identified in CD4⁺ T cells as a T-box transcription factor responsible for T-helper 1 (Th1) lineage commitment, but, more recently, T-bet has been found to play an important role in effector and memory cell fate decisions (16, 73, 86–89). T-bet expression is induced initially by TCR signaling and augmented by IL-12 signals in activated CD8⁺ T cells (16, 73). When T-bet expression was examined in virus-specific CD8⁺ T cells, it was found to be elevated in the KLRG1^hiIL-7R^lo shorter lived effector CD8⁺ T cells relative to the KLRG1^loIL-7R^hi memory precursor effector CD8⁺ T cells (16, 90). This finding suggested that an expression gradient of T-bet acted like a rheostat to control the balance between terminal effector CD8⁺ T-cell differentiation and memory cell potential in effector CD8⁺ T cells. Higher amounts of T-bet instructed KLRG1^hiIL-7R^lo terminal effector cell formation, but lower amounts appeared to permit normal memory cell formation. For example, T-bet-deficient memory CD8⁺ T cells expressed less IL-15Rβ chain (CD122), were defective in the expression of other MPEC-signature genes, and did not proliferate as well as wildtype cells to secondary infection (6, 15, 16, N.S. Joshi and S.M. Kaech, unpublished data). Eomesodermin (Eomes), another T-box factor expressed in activated CD8⁺ T cells, is also important for CD122 and perforin expression in CD8⁺ T cells (16, 76, 90). Eomes has an interesting relationship with T-bet because on the one hand, T-bet and Eomes appear to cooperate in CTL function and memory T-cell homeostasis. T-bet and Eomes coordinate the expression of CD122 in memory CD8⁺ T cells (16, 90), and CD8⁺ T cells that are doubly deficient in both genes are incapable of generating CTLs during LCMV infection. Instead, Tbx21^−/− Eomes^−/− CD8⁺ T cells abnormally differentiated into IL-17-producing CD8⁺ T cells that caused excessive neutrophil infiltration and a lethal inflammatory syndrome (86, 87). On the other hand, they also seem to work in opposition at some level because in contrast to T-bet, IL-12 paradoxically suppresses Eomes expression and Eomes expression preferentially increases relative to T-bet as memory CD8⁺ T cells form and mature (73). However, the mechanisms by which T-bet and Eomes cooperate or possibly antagonize each other at different loci remains to be elucidated.

Another pair of transcriptional regulators that function in memory CD8⁺ T-cell development is Blimp-1 and BCL-6. These are transcriptional repressors that mutually antagonize each other’s expression and are best known for regulating the development of germinal center B cells and long-lived plasma cells (91, 92). Blimp-1 is necessary for the formation of terminally differentiated, senescent, and long-lived plasma cells that reside in the bone marrow (93, 94). In T cells, genetic ablation of Blimp-1 causes a fatal lymphoproliferative autoimmune disorder, demonstrating a role for Blimp-1 in T-cell homeostasis (95, 96). Recently, the function of Blimp-1 in antigen-specific CD8⁺ T-cell development has been examined in greater detail (17, 97, 98). During acute LCMV infection, which is a systemic viral infection, Blimp-1 was expressed to a higher degree in terminally differentiated KLRG1^hiIL-7R^lo effector CD8⁺ T cells relative to KLRG1^loIL-7R^hi cells, and Blimp-1-deficient effector CD8⁺ T cells primarily developed into memory precursor cells and acquired central memory cell characteristics, such as CD62L expression, IL-2 production, and enhanced proliferative responses, more rapidly than their wildtype counterparts (17). However, Blimp-1 is important for maximizing the formation of certain effector functions in CD8⁺ T cells and the control of viral infections. For instance, Blimp-1 is essential in CD8⁺ T cells for the control of influenza infection, a localized respiratory infection, and defective viral control correlated with impaired effector cell migration to the lungs (97). In addition, Blimp-1 is critical to normal perforin, granzyme B and K expression, which is important for cytotoxic CD8⁺ T-cell activity (17, 97). During chronic LCMV infection, Blimp-1 was needed both for optimal effector functions as well as for expression of inhibitory receptors (such as PD-1, LAG3, CD160, and 2B4), which consequently decreases virus-specific CTL functions (i.e. induces CD8⁺ T-cell ‘exhaustion’) to prevent fatal immunopathology (98). Interestingly, rendering the virus-specific CD8⁺ T cells haplo-insufficient in Blimp-1 led to better viral control because this treatment provided a healthier balance between effector cell function and exhaustion. Limiting Blimp-1 by one-half was sufficient to sustain a level of CTL function but was insufficient for maximal expression of the inhibitory receptors.

In B cells, Blimp-1 and BCL-6 reciprocally antagonize each other to regulate the balance between germinal center B cells and plasma cells (92). Evidence suggests that this set of transcription factors acts similarly as a genetic switch in the regulation of effector and memory T-cell fate decisions. In the absence of Blimp-1, BCL-6 expression is increased, and this probably contributes the phenotypes of Blimp-1-deficient CD8⁺ T cells because BCL-6-overexpression in CD8⁺ T cells also recapitulates several of the phenotypes exhibited by Blimp-1-deficient cells such as increased formation memory CD8⁺ T cells (preferentially T_CM cells) and increased proliferative responses to secondary immunization and homeostatic cytokines (99–101). Conversely, a BCL-6-deficiency in CD8⁺ T cells causes several of the opposite phenotypes (99–101). Additionally, BCL-6b, a homolog of BCL-6, is also important for generating memory CD8⁺ T cells that can respond robustly to secondary infection (102).

Inhibitor of DNA binding 2 (Id2) is an inhibitor of the E protein family of transcription factors, and this family of proteins regulates many aspects of lymphocyte development and maintenance (103). Recently, a role for Id2 in memory CD8⁺ T-cell generation in the murine model of Listeria infection has been described (104). Id2 mRNA is upregulated in antigen-specific CD8⁺ T cells at the peak of expansion and persists into the memory phase. Id2-deficient CD8⁺ T cells initially expanded well but then started to die off as a result of increased expression of several pro-apoptotic molecules. Consequently, this led to significantly smaller numbers of effector and memory CD8⁺ T cells after infection, but, interestingly, the frequency of CD62L^hiCD8⁺ T cells was increased considerably in the absence of Id2 (104).

Given memory CD8⁺ T cells resemble hematopoietic stem cell (HSC) in some ways, such as their ability to maintain a multipotent state and self-renew themselves, some transcription factors involved in stem cell longevity and maintenance have been proposed to be involved in memory T-cell development. Indeed, comparison of genome-wide transcription profiles between HSCs and memory T and B cells identified overlap in a small set of genes, suggesting, to some degree, the possibility of a shared genetic program between these different long-lived cell types (105).

Bmi-1, a member of the polycomb repressive complex, is rapidly expressed upon TCR ligation in primary CD8⁺ T cells and enhances their proliferative responses. Knockdown of Bmi-1 by RNAi impaired CD8⁺ T-cell proliferation and effector cytotoxicity, and conversely, Bmi-1 overexpression increased their proliferative responses to homeostatic cytokines. Interestingly, upon restimulation, Bmi-1 is induced in KLRG1^loIL-7R^hi memory precursor cells but not in KLRG1^hiIL-7R^lo effector cells, suggesting that Bmi-1 is important for the long-term maintenance and protection of memory CD8⁺ T cells (106). Another study has shown Wnt3a or inhibitors of GSK3β-induced Wnt signaling ceased activated CD8⁺ T-cell differentiation and promoted the formation of a subset of CD44^loCD62^hiSca-1^hi self-renewing multipotent CD8⁺ T cells that have been referred to as memory stem cells (107). This Wnt-like signaling is associated with increases in Tcf7 and Lef1 expression, and both of these transcription factors are found in T cells with the increased potential to form memory cells in vivo (107). In the future, as we elucidate the transcriptional networks that regulate effector and memory CD8⁺ T-cell development, it will be interesting to delve deeper into the potential overlap in genetic programs between stem cells and memory T cells as well as to identify how these transcriptional pathways intersect and cross-regulate each other to ultimately determine the differentiation state and cell fate of effector T cells.

Metabolic switch during effector-to-memory transition

After explosive proliferation, CD8⁺ effector T cells halt their division and gradually enter a resting quiescent stage, whereby they periodically turnover or homeostatically proliferate (4, 6). Between these two distinct phases, pathogen-specific CD8⁺ T cells undergo a dramatic metabolic switch. In resting cells (including naive T cells), cellular homeostasis is maintained largely by chemical energy through oxidative phosphorylation in mitochondria (108). Upon activation, effector CD8⁺ T cells increase glucose uptake and switch to anabolic metabolism, which is characterized by heightened mammalian target of rapamycin (mTOR) activity and glycolysis (108). Presumably, this anabolic state is necessary for activated CD8⁺ T cells to undergo robust proliferation and increase protein and membrane synthesis during viral infection because perturbation of mTOR function, such as with a high dose rapamycin treatment, greatly blunts effector cell expansion (109). However, after pathogen clearance, the effector CD8⁺ T cells appear to ‘reset’ back to a catabolic state to survive mitogen and growth factor withdrawal (110).

Recent studies have shed some light on how these two metabolic states are regulated in CD8⁺ T cells during and after infection. One study showed that inhibiting mTOR activity by rapamycin treatment enhances the magnitude, quality, and the rate of memory formation during LCMV infection (109). Likewise, another report suggested that the activation of an inhibitor of mTOR, 5′-adenosine monophosphate-activated protein kinase (AMPK), which can be induced by cellular stress and adenosine triphosphate (ATP) deprivation, promotes the transition from an anabolic effector cell to a catabolic memory cell via induction of fatty acid oxidation (FAO). This model was supported by experiments examining TNF receptor-associated factor 6 (TRAF6)-deficient CD8⁺ T cells, which were defective in FAO and generated very few memory CD8⁺ T cells after Listeria infection. Metformin, an anti-diabetic drug and activator of AMPK, can partially bypass the need of TRAF-6 and restore memory cell generation in TRAF6-deficient CD8⁺ T cells (110). Collectively, these results suggest that two opposing metabolic pathways – PI3K/Akt/mTOR-mediated cellular growth and AMPK/FAO-mediated cellular homeostasis – have to be tightly controlled to ensure the proper effector differentiation and memory development (Fig. 1).

Although the precise molecular mechanisms of how these metabolic pathways control cell fate decisions remain unclear, one could postulate that mTOR kinase may modulate specification of effector or memory cell fates via gene expression. Consistent with this notion, earlier studies have found that mTOR is required for Th1, Th2, and Th17 CD4⁺ effector T-cell development, whereas it represses Treg development. Specifically, mTOR increases the activity of STATs and the expression of their downstream lineage-determining transcription factors (such as T-bet, GATA3, and RORγt). mTOR also suppresses Foxp3 expression in Tregs (111). A recent study further extended this paradigm to CD8⁺ T cells by showing that IL-12/STAT4-mediated mTOR activity sustains T-bet and simultaneously suppresses Eomes expression. Therefore, mTOR integrates environmental cues and acts as a rheostat to control T-cell fate decisions via modulation of T-bet and Eomes expression (112). Given that increased T-bet expression drives formation of short-lived, terminally differentiated KLRG1^hiIL-7R^lo effector CD8⁺ T cells during viral infection, this finding helps to explain why inhibition of mTOR via rapamycin or metformin promotes memory cell formation in the above studies. Thus, these drugs offer promising therapeutic approaches for future vaccine development against infection and tumors.

Apoptosis versus survival during contraction phase

Apoptosis or programmed cell death can occur through two major pathways: the death receptor pathway (an extrinsic pathway) and the mitochondrial pathway (an intrinsic pathway) (127, 128). Typically, the death receptor pathway initiates via interaction of TNFR family members with their extracellular ligands such as TNF-α, Fas ligand, and TRAIL (127–129). The mitochondrial pathway is initiated by disruption of the mitochondrial membrane from cellular stress, which leads to the release of cytochrome c into cytoplasm and the formation of apoptosome. Both pathways end with the downstream ‘effector’ caspase-3 and -7 activation that results in apoptosis (127, 130).

Recent findings suggest that both pathways are involved in effector T-cell death during the contraction phase. Following acute infection, viral specific CD8⁺ T-cell contraction is profoundly attenuated in TNFRI and TNFRII double-deficient mice due to their overlapping function (131). Fas, however, does not appear to play a critically important role in acute effector CD8⁺ T-cell contraction on its own (132), although it seems to account for the reduced numbers of antigen-specific T cells during chronic infections or autoimmune settings in the presence of persistent antigen (133–135). The Bim-induced intrinsic pathway has been found to play a central role in regulating CD8⁺ T-cell contraction, whereby Bim-deficient effector CD8⁺ T cells have prolonged survival after pathogen removal (134–138) (Fig. 1). However, in the context of Bim deficiency, the function of Fas in activated CD8⁺ T-cell death can be seen following acute infection because a double deficiency in Bim and Fas increases memory T-cell formation more than a Bim deficiency alone (134, 135). Despite an incomplete understanding of the molecular mechanism, Bim activity is probably controlled by the relative level of anti-apoptotic molecules Bcl-2, Bcl-XL, A1, and MCL-1 (139–141). Both Bcl-2 and MCL-1 have been shown to be required for sufficient memory T-cell survival (142, 143) (Fig. 1). Although Bim-deficient T cells do not contract initially at the same rate, they do gradually decay and reach the similar numbers as wildtype memory T cells (132, 134, 135, 137). This observation suggests that there are alternative signals in limitation that ultimately govern the number of memory T cells that can persist.

Common γ chain cytokines IL-7, IL-15, and IL-21 in memory maintenance

IL-7 and IL-15 are critical cytokines that regulate the transition from effector-to-memory CD8⁺ T cells, as both of these cytokines are critical to the survival and homeostasis of memory T cells (6, 14, 47). In general, it is thought that IL-7 primarily supports survival, whereas IL-15 induces basal level of homeostatic turnover, but both factors cooperate to support both cell processes (14, 19, 21, 47, 144–146). For instance, antigen-specific CD8⁺ T cells require IL-7R to form normal numbers of memory CD8⁺ T cells, and in the absence of both IL-15 and IL-7, essentially no memory CD8⁺ T cells form after infection. Likewise, if IL-7 or IL-15 levels are elevated via transgenic overexpression or administration of exogenous cytokines, greater numbers of memory CD8⁺ T cells persist under these conditions (14, 21, 27, 47, 147–150).

As described above, the expression of IL-7R is dynamically regulated in the antigen-specific CD8⁺ T cells during an infection. Naive CD8⁺ T cells express IL-7R, but IL-7R expression precipitously falls on activated CD8⁺ T cells as they expand and differentiate during infection (21, 23, 29, 151). At the peak of expansion, most effector CD8⁺ T cells are IL-7R^lo; however, a smaller subset of IL-7R^hi cells exist that express greater amounts of Bcl-2 and serine protease inhibitor 2A (Spi2a) and preferentially survive and generate long-lived memory CD8⁺ T cells that self-renew (21, 23, 27, 29, 146, 152, 153).

Given the IL-7Rα expression pattern on effector CD8⁺ T cells, it was initially predicted that IL-7R repression was the underlying cause for effector CD8⁺ T-cell death and contraction following infection. However, this model proved to be incorrect because enforced IL-7R expression, via an IL-7R transgene, on all virus-specific CD8⁺ T cells did not save the terminally differentiated KLRG1^hi effector CD8⁺ T cells that are naturally IL-7R^lo (27, 154). These results indicated that IL-7R repression is not causal to effector cell death and that expression and IL-7 signaling is not an instructive signal for memory T-cell formation (27, 47, 154).

The essential role of IL-15 in memory T-cell maintenance and homeostasis was first identified because it could induce memory T-cell proliferation in vitro (155, 156) and fewer polyclonal memory CD8⁺ T cells were found in IL-15^−/− and IL-15rα^−/− mice (157, 158). Subsequently, the effects of IL-15 on effector and memory CD8⁺ T-cell development during acute viral and bacterial infections were examined and these studies showed that IL-15 was needed for memory CD8⁺ T-cell homeostasis and drove the basal turnover of the memory CD8⁺ T cells. In the absence of IL-15, the antigen-specific memory CD8⁺ T-cell population slowly eroded over time. Additionally, in some infections, but not all, IL-15 is needed for maximal effector cell expansion and this may be due in part to necessity of IL-15 for the survival of KLRG1^hiIL-7R^lo effector CD8⁺ T cells (16, 19, 159, 160).

IL-15 is induced by type I IFNs, and macrophages and DCs are the primary producers and presenters of IL-15 to CD8⁺ T cells (14, 159, 161). Interestingly, a recent paper elegantly illustrated that the cell type most important for presenting IL-15 to memory CD8⁺ T cells varied according to the type of memory T cell (T_CM versus T_EM) and the tissue the memory CD8⁺ T cells resided in. For instance, DC-presented IL-15 selectively supported T_CM, whereas macrophage-presented IL-15 supported both T_CM and T_EM cells and was the most important cell type to maintain memory CD8⁺ T cells in the bone marrow (162).

IL-21, the most recently identified common γ chain cytokine, is closely related to IL-2 and is mainly produced by CD4⁺ T cells and NKT cells (163, 164). IL-21 is an important inducer of Blimp-1 and BCL-6, and it regulates B-cell differentiation into plasma cells, T-follicular helper function, and Th17 development (164–170). IL-21 can also synergize with IL-2 and IL-15 to promote cytotoxic CD8⁺ T-cell differentiation (163, 164). For instance, combinations of IL-15 and IL-21 augment effector CD8⁺ T-cell proliferation, survival, and cytotoxic activity. Microarray analyses further revealed that this was in part through the upregulation of granzyme B and c-Jun (171). In line with this finding, IL-21r^−/− mice have reduced effector CD8⁺ T-cell expansion and cytotoxicity after immunization (171, 172). For these reasons, it was surprising that a recent study also found that IL-21 suppresses the antigen-induced expression of CD44, IFN-γ, granzyme B, and Eomes. These IL-21-treated effector cells rendered better anti-tumor immunity when adoptively transferred into tumor-bearing mice (172). IL-21 can also synergize with IL-7 and IL-15 to induce memory CD8⁺ T-cell homeostatic proliferation, but IL-21 has little effect on this process alone (171). The role of IL-21 in CD8⁺ T-cell expansion, differentiation, and memory formation during acute LCMV infection is largely dispensable, but during chronic LCMV infection, it is critical to sustain the function of virus-specific CTLs (173–175).