Virus-specific CD8 T cells: activation, differentiation and memory formation


Annette Oxenius, Institute of Microbiology, Swiss Federal Institute of Technology, Eidgenössische Technische Hochschule Zurich, Wolfgang-Pauli-Strasse 10, HCI G401, 8093 Zurich, Switzerland. e-mail:


CD8 T cells are pivotal for the control of many intracellular pathogens, and besides their role in immediate control of infections, CD8 T cells have the capacity to differentiate into long-lived antigen-independent memory CD8 T cells, at least in situations of acute and resolved infections. The population of memory cells is heterogeneous with respect to their phenotype, their anatomical localization and their functional capacities in order to afford optimal protection against secondary infections. In the past years, it has become clear that multiple in vivo parameters are involved in shaping the composition of the memory CD8 T cell population, including antigen load, duration and strength of CD8 T cell stimulation, the level of inflammation, availability of CD4 T cell help and CD8 T cell precursor frequencies. With respect to the timing when CD8 T cells are committed to become memory cells, several models have been proposed. In contrast to acute, resolved infection, the continued in vivo exposure to high levels of antigen during persistent chronic viral infection precludes the development of long-lived antigen-independent memory CD8 T cells and might even result in severe dysfunction of virus-specific CD8 T cells.

Memory is a hallmark of adaptive immunity and develops after primary encounter with a specific immunogen. The notion of ‘memory’ embraces many constituents, including long-lived antibody titers, elevated frequencies of antigen-experienced lymphocytes, strategic anatomical distribution of memory lymphocytes, reduced activation thresholds, enhanced and accelerated effector functions and pronounced recall proliferation potential. The notion of ‘memory’ also implies that secondary infections are dealt with in a different manner in comparison with a primary infection, in most cases leading to accelerated clearance of the secondary infection and an ameliorated disease course. These two aspects of memory, namely its constituents and its biological consequences, have in the past fostered interesting and often controversial discussions about the relevance and nature of memory lymphocytes and their role in mediating protective memory. In this review, we will focus our attention on one constituent of memory, namely CD8 memory T lymphocytes and discuss selected topics about their phenotype and function (summarized in Table 1), parameters involved in their differentiation (summarized in Table 2), their protective role in secondary infections and the impact of persistent infections on the long-term differentiation, function and maintenance of virus-specific CD8 T cells (summarized in Table 3).

Table 1.   Phenotypical and functional markers of naive, effector and memory CD8
T cell population markerExpressionFunction
  1. PD-1, programmed death-1; KLRG1, killer cell lectin-like receptor G1; IL-2, interleukin-2; IFNγ, interferon γ; TNFα, tumor necrosis factor α; LN, lymph node; ECM, extracellular matrix; LFA-1α, lymphocyte function-associated antigen 1α.

CD62LHighLowHighLN homing
CD44IntHighHighAdhesion to ECM
CD11aIntHighHighAdhesion; LFA-1α chain
CCR7HighLowHighLN homing
CD127HighLowHighHomeostatic proliferation
KLRG1LowHigh/lowLowSenescence marker
CD122Int.Int.HighHomeostatic proliferation
CD43LowHighLowLymphocyte adhesion and activation
PD-1LowHighLowNegative regulation
CD69LowHigh/lowLowEarly activation marker; LN retention
CD57LowLowLowSenescence/end-stage differentiation marker
IL-2 productionLow/highLowHigh 
IFNγ productionLow/HighHigh 
TNFα productionLowHighHigh 
Table 2.   Factors that influence the generation of memory CD8 T cells in the setting of acute/resolved infection
Intracellular, CD8 T cell-intrinsic moleculesCD4 T cellsMembrane-bound molecules and soluble factorsOther parameters
  1. APC, antigen-presenting cell; DC, dendritic cell; IL-15, interleukin-15; TCR, T cell receptor; TEM, effector memory T cells.

Longevity of memory CD8 T cells is positively influenced by upregulation of anti-apoptotic molecules such as Bcl-2 and Bcl-XL (165) [they are not instructive for their differentiation (116, 163, 166, 167)]CD4 T helper cells increase generation of memory CD8 T cells (5, 17–21, 92, 93, 129, 131, 133, 134)Memory CD8 T cells primed in the absence of IL-2 exhibit impaired secondary expansion (143–145)Duration of antigen exposure: more rapid differentiation of memory CD8 T cells by reduced time of antigen exposure (10, 12, 15, 16, 39, 40, 61, 107, 122)
Pro-apoptotic Bcl-2 homology 3-only protein Bim increases CD8 T cell contraction (113, 114)T helper cells ‘license’ APCs via CD40/CD40L interaction (125–127)DCs, activated and matured by antigen-specific CD4 T cell help, secrete and bind IL-15, which then is trans-presented to CD8 T cells and thereby enhances CD8 T cell priming and memory formation (146–148)Strength of TCR stimulus (10, 13, 14)
Spi2A protects CD8 T cells from programmed cell death (121). Spi6 protects the differentiating CD8 T cell from self-inflicted death by its own effector molecules (120, 121)CD4 T helper cells interact directly with CD8 T cells via CD40/CD40L (135, 136)CCR5 ligands promote memory CD8 T cell differentiation (148, 149)Precursor frequency of specific CD8 T cells (15, 16); high precursor frequencies favoring transitional TEM development (27)
Deletion of T-bet drives the differentiation of central memory CD8 T cells (154)CD4 T helper cells enhance generation of memory CD8 T cells via OX40 engagement (138, 139) Size of clonal expansion determines frequencies of memory CD8 T cells (37, 38)
Eomesodermin and T-bet have instructive roles for the development of memory cells with long-term self-renewal capacity (157)CD4 T cells enhance generation of memory CD8 T cells via upregulation of CD27 (141) Asymmetric cell division leads to the generation of two distinct types of daughter cells; the APC-proximal cell with effector cell characteristics and the APC-distal cell with memory cell characteristics (72)
  Level of inflammation during priming (101, 39). Reduced levels of inflammation favor development of CD8 memory precursor cells (40, 61, 123)
Table 3.   CD8 T cell ‘memory’ during acute/resolved, persistent latent and chronic productive infection
infection (e.g.
Persistent latent infection (e.g. CMV)Chronic productive
infection (e.g. LCMV
  • *

    At late time points after infection.

  • 1

    1 CD8 T cell frequencies increase during latency.

  • 2

    2 CD8 T cell frequencies do not increase during latency.

  • CMV, cytomegalovirus; LCMV, lymphocytic choriomeningitis virus; PD-1, programmed death-1; KLRG1, killer cell lectin-like receptor G1; IL, interleukin; IFNγ, interferon γ; TNFα, tumor necrosis factor α; LAG-3, lymphocyte-activation gene-3.

Effector functionsHigh proliferative potentialIntermediate levels of IFNγ and TNFα productionPotent effector functions (IFNγ, TNFα, IL-2)Functionally exhausted (no IL-2 and TNFα production, low levels of IFNγ production)
Potent effector functions (IFNγ, TNFα, IL-2)Low/no IL-2 productionHigh proliferative potentialMaintained in vivo cytotoxicity
In vitro cytotoxicity after restimulationIntermediate proliferative capacityMaintained ex vivo and in vivo cytotoxic activityPoor in vitro proliferative capacity
Maintained in vivo cytotoxicityMaintained in vivo cytotoxicity  
Homeostatic turnover via IL-7 and IL-15Maintenance dependent on IL-2Rα signals on CD8 T cellsHomeostatic turnover via IL-7 and IL-15Maintenance dependent on IL-2Rα signals on CD8 T cells
Long-term regulation of CD8 T cell numbers and effector function   Negative regulation of maintenance and of effector functions of CD8 T cells via IL-10, PD-1 and LAG-3


Essentially there is no such cell as ‘the memory CD8 T cell’ that can be unambiguously categorized by virtue of specific markers and/or function. Memory CD8 T cells represent rather a dynamic population of diverse subpopulations of antigen-experienced cells that evolve and differentiate during acute infection and continue to do so after resolution of primary infection. These subsets of memory cells differ with respect to their half-lives, their anatomical localization, their proliferative potential, their immediate effector function, their requirements for maintenance, their phenotype and their protective potential (1–12). Different parameters influence the relative composition and size of the memory CD8 T cell population; these include the relative strength and composition of the three signals that are perceived by a CD8 T cell during its activation process [signal 1: T cell receptor (TCR) interaction with peptide–major histocompatibility complex (MHC) complexes (10, 13, 14); signal 2: interaction with costimulatory molecules; and signal 3: inflammatory cytokines relayed to the CD8 T cell either directly via the priming antigen-presenting cell (APC) or via an alternative bystander source]. Furthermore, the duration of antigen exposure, the level and duration of inflammation, the precursor frequency of antigen-specific CD8 T cells (15, 16), the availability of CD4 T cell help (17–21), the time after primary infection and the clonal burst size of the CD8 T cell response also influence the relative size and composition of the memory CD8 T cell compartment. As one major determinant of the function and phenotype of antigen-experienced CD8 T cells is the level and duration of antigen contact, the nature of the infecting pathogen plays a critical role in shaping the population of ‘memory’ cells. In case of a viral infection, three different scenarios can be distinguished: (a) acute/resolved infection after which the viral pathogen is cleared from the host after a given time period; (b) persistent latent infection during which the viral pathogen is controlled after acute infection but remains present in the infected host mostly in form of latency or at immunopriviledged sites – which is coupled, however, with sporadic reactivation events; and (c) chronic active infection during which continuous viral replication takes place. The term ‘memory’ CD8 cell really only applies for the situation of an acute/resolved infection or immunization. In the latter two situations it is more difficult to distinguish between memory and effector-type cells; we therefore prefer relating to antigen-experienced cells in these situations. These three situations will be specifically addressed in the fourth, fifth and sixth sections. In the first three sections, we will discuss some general features of memory CD8 T cells and their differentiation pathways as they evolve after an acute/resolved type of infection.

During an acute/resolved viral infection four consecutive phases can be distinguished: CD8 T cell activation, expansion, contraction and memory (1, 12, 22, 23). Clonal expansion of antigen-specific CD8 T cells occurs normally during the first week of infection, during which the numbers of antigen-specific CD8 T cells increase up to 104–105-fold (24–26). The precursor frequencies of naïve antigen-specific CD8 T cells range from about 100 to 500 (27) and during the process of clonal expansion they can expand to up to 10 million of specific cells. Depending on the CD8 T cell precursor frequency, on the abundance, the stability and the presentation kinetics of a given antigenic peptide, CD8 T cells with various specificities for a given complex pathogen are induced and dominant or subdominant CD8 T cells responses are distinguished depending on their relative abundance. During this clonal expansion phase, CD8 T cells not only proliferate but also differentiate into effector cells and memory precursor cells. Effector cells migrate to peripheral tissues (28) where they can directly eliminate virally infected target cells via direct cytotoxicity or cytokine production or they can recruit (and activate) other effector leukocytes by virtue of local chemokine secretion. Thus, the main goal of the clonal expansion and differentiation phase is to generate armed effector cells that are instrumental for the immediate control of the infection (12, 29–33), but also to preserve a subset of these cells for heightened specific CD8 T cell immunity in case of a subsequent infection with the same pathogen. In general, 5–20% of the cells present at the peak of the primary expansion survive the contraction phase with a majority of the effector cells undergoing apoptosis (34–36). Normally this contraction phase coincides with control of viral infection, however this relation is not causal (35), as contraction also occurs in case of persistent viral infections. The magnitude of the clonal burst size generally correlates with the level of memory cells that develops after an acute/resolved viral infection (37, 38). An exception to this general rule are situations of weak inflammatory conditions, in which a relative larger proportion of CD8 T cells at the peak of the response can become memory cells (39, 40), likely at the expense of less differentiated effector cells. The 5–20% of cells that survive the clonal burst constitute the pool of memory cells, which changes (as a population) slowly in time with respect to phenotype, anatomical localization and direct effector functions.

In a very useful paradigm antigen-experienced T cells were divided into two major populations, based on their anatomical location: central memory T cells (TCM), residing preferentially in secondary lymphoid organs, and effector memory T cells (TEM), localized in peripheral tissues and in the spleen (41–45). The main markers for the distinction of these two subsets are the lymph node (LN) homing markers CD62L and CCR7, with CD62L+CCR7+ cells representing the TCM cells and the CD62LCCR7 cells representing the TEM cells. In the recent years, however, it has become apparent that this division into two populations of memory cells is perhaps too simplistic and a variety of other phenotypical and functional markers have been added to characterize the spectrum of antigen-experienced CD8 T cells (Table 1).

The different types of memory cells that develop after acute/resolved infection distinguish themselves from naïve cells by their more rapid exertion of effector functions (such as cytotoxicity, cytokine and chemokine secretion), by their reduced costimulatory requirements for recall activation, by their increased functional avidity and by their wide-spread anatomical location (1, 12, 46–49). However, many of these attributes of ‘memory’ cells vary according to the specific subset of memory cell. TEM-like cells exhibit, to various degrees, immediate cytotoxicity, whereas TCM-like cells are poorly cytotoxic ex vivo (8). The cytokine secretion potential between TEM- and TCM-like cells mainly differs with respect to interleukin-2 (IL-2) production (8, 50–52), with the TCM-like cells being able to produce significant amounts of IL-2. The proliferative potential after secondary stimulation is much greater for the TCM-like cells; however, more recent reports have also shown that TEM-like cells can have significant proliferative potential – in particular at mucosal sites and at late time points after primary infection (53). Lastly, the self-renewing capacity seems to be largely confined to a subpopulation of TCM cells; however, definitive markers of these stem-cell-like memory cells remain to be identified.

What is the relationship between all those phenotypically and functionally distinct subsets of memory cells? A definitive answer to this question is rather difficult to address in experimental terms and various studies have come to different conclusions. While some studies show that TCM and TEM cells are fixed populations (16, 54–56), with the TCM-like cells having the possibility to differentiate into TEM-like cells upon antigen re-encounter (56); others demonstrated that TEM can convert to TCM in the absence of antigen (8, 15, 33, 57, 58). As mentioned above, the distinction of memory cells in only two subpopulations of TCM and TEM memory cells is likely to be insufficient to appreciate the heterogeneity of memory cells and might therefore explain, at least partly, the different conclusions about their respective relationship. It has recently been suggested to categorize the TEM cells into a transitional subset (which can revert to TCM cells in the absence of antigen stimulation) and an end-stage TEM subset (which are short-lived and cannot revert to TCM). The factors that influence the relative balance between end-stage TEM cells and transitional TEM cells include the precursor frequency [high precursor frequencies such as in adoptive transfer of TCR transgenic CD8 T cells favoring transitional TEM development (27)], the strength and length of stimulation by antigen and the availability of CD4 T cell help.

A closely related question to the lineage relationship between different subpopulations of memory cells is the question when and how memory cells are generated. There are currently several models that are proposed for the generation of memory cells:

Uniform potential model

This model proposes that CD8 T cells differentiate after activation into effector cells with the potential to become a memory cell thereafter [first a TEM-like memory cell followed by a TCM-like memory cell (59)]. The decision for selective survival of a subset of effector cells is governed, for example, by nutrient availability, access to growth factors and antigen (60). This simple linear model, however, does not account for the heterogeneity among memory cells and gives poor indication about the timing of memory development.

Decreasing potential model

This model proposes that high levels of TCR stimulation (with respect to strength, number of contacts and duration of antigen availability) lead to the differentiation of fully differentiated end-stage TEM-like memory cells with poor potential for long-term survival and self-renewal. Conversely, reduced levels of TCR stimulation favor the development of TCM and transitional TEM cells, which exhibit long-term survival and self-renewal (22). This model accounts for the heterogeneity of memory cells and suggests that ‘late-comers’ into the population of antigen-primed CD8 T cells (when the antigen load is starting to decline) might have increased chances to become long-lived memory cells (61–64). This model also accounts for the presence of mainly short-lived end-stage effector (memory) cells in situations with chronic antigen availability (65–67).

Fixed lineage model

This model proposes that an early decision after initial CD8 T cell activation is made toward the differentiation into effector or memory cells, suggesting that fully differentiated effector cells and memory cells exist concomitantly (however, this is also possible in the decreasing potential model) (11, 55, 68). This early selective decision requires that preformed memory cells might somehow evade effector cell differentiation (if this would be due to a reduced strength of stimulation, then it would be similar to the decreasing potential model) (69). There is good evidence for the coexistence of a small number of memory (precursor) cells and effector cells at the peak of clonal expansion (10, 39, 59, 70) – but this does not necessarily mean that the decision to become a memory cell precursor is a cell-autonomous decision; this might also be a consequence of a particular ‘antigen-APC-experience’ of these cells.

The observation that TCM- and TEM-like cells exhibit different TCR repertoires was considered as evidence that different types of naïve CD8 T cells are recruited into these two memory cells subsets (71). However, using CD8 T cell clonotyping analyses, it was shown that CD8 T cell clonotypes are shared among different memory and effector CD8 T cell subsets (57). In support of an early event deciding about effector vs memory cell differentiation fate, it was recently shown that asymmetric cell division leads to the generation of two distinct types of daughter cells; one cell (proximal to the APC) is equipped with effector cell characteristics whereas the APC-distal cell exhibits more memory cell characteristics (72). However, additional parameters than the asymmetric cell division will have to contribute to the long-term fate of the two distinct daughter cells, otherwise one would have to expect a 50:50 ratio between effector and memory cells at the peak of the response (which is clearly not the case). Furthermore, the decisive role of asymmetric division for effector and memory cell generation is also questioned by a recent study showing that a single naïve antigen-specific cell can give rise to the development of a pool of antigen-experienced cells with the full heterogeneity in phenotypes and functional characteristics (73).

Irrespective of the models discussed above, there is ample experimental evidence for the (early) existence of memory precursor cells (74) and several markers have been proposed to identify these memory precursor cells during the peak expansion phase. These markers include IL-7Rα, killer cell lectin-like receptor G1 (KLRG1) and IL-2 production and will be discussed in more detail in the fourth section.

The definition of memory (and memory precursor cells) often relies on their potential for secondary expansion, on their longevity and self-renewing capacity (12). With respect to the more general, biological aspect of memory, namely increased protection against secondary infection, it is of considerable importance to define the protective capacity of the various memory CD8 T cell subsets. The various conditions in which the protective capacities of different memory CD8 T cell subsets were analyzed suggest that the relative protective capacity of CD8 memory T cell subsets varies depending on the nature of the challenge pathogen (75). As TCM-like cells have a greater proliferative potential, they have been shown to confer protection against systemic infections, local and respiratory challenge infections (9, 53, 59). However, this protection is mostly not operational very early after challenge infection as it relies on secondary expansion, differentiation and migration of reactivated TCM cells (9). In contrast, immediate protection (in particular in peripheral tissues) is better provided by tissue-resident TEM-like cells, as for example shown against Vaccinia virus protection in ovaries (9), protection against Listeria monocytogenes infection (75, 76), protection against Sendai or influenza virus in the lung (77–80) or malaria in the liver (81). The issue about the protective potential of pathogen-specific memory CD8 T cells is of considerable relevance in the context of CD8 T cell-based vaccine development such as for HIV and TB infection. Particularly in HIV infection, it is widely believed that polyfunctional, TCM-like cells are the target population to be induced by a vaccine (82–84) – and this is mainly based on the observation that patients who naturally control HIV replication to a very good extent, exhibit such a population of HIV-specific CD8 TCM cells (85, 86). However, this might very well be a consequence rather than a cause of antiviral control (87). Thus, while a vaccine-induced population of proliferation-competent TCM-like memory cells will be a prerequisite for the rapid expansion and differentiation of secondary effector cells, this process might be too slow for effective early combat of an initially mostly peripheral mucosal infection and it might thus be of advantage to additionally induce a strong population of TEM-like cells that are resident and ready to act at the site of infection.


As the level and the quality of memory CD8 T cell responses is influenced to a great extent by processes that are initiated during the priming phase, we will discuss here various parameters that quantitatively and qualitatively influence CD8 T cell expansion and thereby memory generation as well as parameters that act during CD8 T cell contraction (summarized in Table 2).

During the priming event, a naïve CD8 T cell has to integrate three signals originating from the priming APC, consisting of TCR stimulation, engagement of costimulatory molecules and exposure to a variety of inflammatory cytokines. In several (mostly in vitro priming) experiments it was shown that a brief antigen encounter initiated a full differentiation program in CD8 T cells, which was coined by the term ‘autopilot’ (12, 88–91). However, it has become clear that the situation might be different in vivo with longer antigen exposure being required for effective CD8 T cell activation and additional signals (in particular inflammatory cytokines) acting during the process of CD8 T cell priming in the context of an infection (4). Besides these inflammatory signals, signals from other cell types (e.g. CD4 T helper cells) may be essential for priming and memory CD8 T cell generation under certain circumstances (17–21, 92, 93).

Proinflammatory cytokines act via several ways during the priming event of CD8 T cells: they can directly influence the level of antigen (as they might themselves contribute to the control of infection) and they are (together with pathogen-associated molecular patterns) key for the maturation of dendritic cells (DCs) to become potent activators of naïve CD8 T cells. Matured DCs express increased levels of costimulatory molecules, which interact with the respective ligands on the CD8 T cell (CD28, CD27, CD40, 41BB, ICOS and OX40) (94–96). There is a large body of evidence that these costimulatory interactions are relevant for the efficiency of the CD8 T cell priming process, but it remains unclear how these signals exactly instruct memory development.

Finally, proinflammatory cytokines can act directly on the primed CD8 T cells and thereby modulate proliferation, survival and acquisition of effector functions (97). The cytokines involved in these processes include IL-12 (98), type I interferons (IFNα/β) (99, 100) and type II interferons (IFNγ) (101, 102). The relative importance of these cytokines with respect to CD8 T cell priming seems to depend largely on the nature of the infecting pathogen (99, 100, 103). IL-12 is mainly produced by DCs and acts directly on CD8 T cells to increase their proliferation, survival and differentiation into effector cells with cytolytic potential (98, 104). The timing of exposure to IL-12 seems critical as it conveys maximal effects when present together with signals 1 and 2 (98) and it may act as a survival signal via upregulation of the anti-apoptotic protein Bcl-3 (105) or via inhibition of the pro-apoptotic enzyme caspase-3 (106). Furthermore, IL-12 was also shown to increase the expression levels of the transcription factor T-bet, an effector cell differentiation factor that is expressed at high levels in effector cells and at lower levels in precursors of long-lived memory cells (107). IFNα/β can also function as signal 3 as it can support the survival of primed CD8 T cells, contribute to CD8 T cell expansion and development of lytic effector function in vitro (108) and is essential for the expansion and survival of CD8 T cells after lymphocytic choriomeningitis virus (LCMV) infection in vivo (99, 100). Interestingly, there is only a modest role for IFNα/β as signal 3 in Vaccinia virus or L. monocytogenes infection, supporting the notion that the biology of the infection is critical in determining the signal 3 requirements of CD8 T cells. IFNα/β seems to act subsequent to signals 1 and 2 to promote survival of the activated cells (99). Also IFNγ can directly act on CD8 T cells and promote their expansion in vivo as shown for LCMV infection (101, 102), but this is not a general finding as IFNγ is not involved in promoting the expansion and acquisition of effector functions of CD8 T cells after L. monocytogenes infection (101, 109). IFNγ has to act as a third signal in very close temporal proximity to signals 1 and 2 as CD8 T cells are know to downregulate the IFNγ-R2 already 12 h post-activation (110).

During the contraction phase, 80–95% of all antigen-specific CD8 T cells are eliminated in a regulated process of selective apoptosis. While earlier studies described activation-induced cell death being responsible for the down-sizing of the responses with involvement of the death receptors Fas and tumor necrosis factor receptor I (TNFRI), or cytotoxic T-lymphocyte antigen 4 (CTLA-4) as a critical negative regulator of CD8 T cell activation, more recent studies have challenged their importance in mediating CD8 T cell contraction (111, 112) and have focused more on the regulation of pro- and anti-apoptotic molecules. In particular members of the Bcl-2 family of pro- and anti-apoptotic molecules are regulating CD8 T cell survival or apoptosis. Bim, a pro-apoptotic molecule, seems to play a particularly important role as its deficiency leads to a marked reduction in CD8 T cell contraction (113, 114). Furthermore, IL-7Rα identifies, at least in certain circumstances, early memory CD8 T cell precursors and it was initially proposed that signaling via the IL-7R might deliver specific survival signals. However, it has become clear that IL-7 and also IL-15 (both being important for the homeostatic self-renewal of memory cells) have probably no instructive role in the generation of short- or long-lived memory cells (8, 39, 107, 115–117). Other factors that regulate the contraction involve IFNγ (with IFNγ promoting contraction after LCMV and L. monocytogenes infection (101)) and the level of inflammation during the priming phase (with less inflammation during the priming phase leading to reduced peak expansion but also to reduced contraction (39) and hence a relative increase in memory development). It should be noted, however, that CD8 T cell priming in the absence of inflammation leads to clonal anergy and/or deletion (118, 119).

Other important parameters involved in shaping the extent of peak expansion and contraction of CD8 T cells are the serine protease inhibitors Spi6 and Spi2A (120, 121). Spi2A is upregulated in LCMV-specific memory precursor cells and protects them from programmed cell death (121). Spi6 is upregulated during CD8 T cell activation and protects the differentiating CD8 T cell from self-inflicted death by its own effector molecules (in particular granzyme B). Spi6-deficiency leads to massively reduced peak expansion after LCMV infection but surprisingly not to reduced memory CD8 T cell levels (120), supporting that the process of contraction is not a prerequisite for the generation of memory CD8 T cell populations.

With respect to the rate with which memory (precursor) CD8 T cells are generated, signal 1 seems to play an important role. It was shown that reducing the time of antigen exposure in vivo (e.g. by blunting the infection) led to the differentiation of CD8 T cells expressing relatively increased levels of CD127 and CD62L and lower levels of KLRG1 at the peak of the response and this was associated with more rapid, but not necessarily more pronounced, memory CD8 development (10, 12, 15, 16, 39, 40, 107, 122). Compared with curtailed antigen exposure, reduced levels of inflammation also favor the development of CD127+ memory precursor cells (40, 123) and in certain situations the presence of CD4 T cell help positively influences the frequency of CD127+ memory precursor cells (92, 124). The following chapter will specifically focus on the role of CD4 T helper cells in the generation of memory CD8 T cells.


In many situations, CD4 T cells are required for the effective activation and expansion of antigen-specific CD8 T cells, in particular in the setting of non-inflammatory conditions (e.g. minor antigens, DC immunization or protein immunization). In these situations CD4 T cells contribute essentially to the activation and maturation of professional APCs, thereby ‘licensing’ the APCs to become potent inducers of CD8 T cells. This ‘licensing’ was shown to occur predominantly via CD40–CD40L interactions between activated CD4 T cells and APCs (125–127) and requires that the epitopes presented to CD4 T cells and CD8 T cells have to be displayed on the same APC (128). In contrast, many pathogenic infections (such as influenza, LCMV, VSV, L. monocytogenes or other pathogens) induce strong primary CD8 T cell responses in the absence of CD4 T cell help (129–131). The ability of these infectious agents to directly activate APCs via pattern recognition receptors is thought to circumvent the need for CD4 T cell help (132, 133). However, more recent reports indicated that even in the setting of infections, CD4 T cells play a crucial role in shaping the CD8 T cell response; while priming of CD8 T cells occurs normally in the absence of CD4 T cell help, memory CD8 T cell differentiation, maintenance and the ability to undergo secondary expansions critically depend on the presence of CD4 T cell help in many (but not all) immunization protocols or infection models (5, 129, 131, 133, 134).

There are a number of mechanisms by which CD4 T cells mediate help either for priming of CD8 T cell responses or for CD8 T cell memory differentiation and maintenance. These include direct cell–cell interactions between CD4 T cells and APCs or direct contact between CD4 and CD8 T cells. Membrane molecules that were shown to be involved in such cell–cell interactions include members of the TNFR super-family (CD40, OX40 and CD27), which provide critical costimulatory signals that promote CD8 T cell survival and differentiation. The role of CD40–CD40L interactions was extensively studied over the last years and seems to be mainly relevant for licensing of CD40 expressing DCs via interaction with activated CD4 T cells expressing CD40L. One study suggested a direct interaction between CD8 and CD4 T cells through CD40–CD40L to promote the development of memory CD8 T cells (135) via regulating the expression of multiple genes involved in CD8 memory generation, survival and functionality as well as their sensitivity to immunoregulation (136). However, other groups that studied the significance of CD40 expression on CD8 T cells during the course of primary and secondary responses to different infections (L. monocytogenes, LCMV, influenza) could not confirm a requirement for a direct CD40–CD40L interaction between CD8 and CD4 T cells (21, 134, 137).

OX40, a costimulatory molecule expressed on both activated CD4 and CD8 T cells, influences the generation and survival of the memory CD8 T cells as treatment with an activating anti-OX40 antibody greatly increases memory CD8 T cell numbers in the presence of CD4 T cell help (138, 139). OX40 signaling in CD8 T cells leads to an increased expression of CD25, accounting for enhanced survival mediated through IL-2 signals that upregulate the PI3 kinase/phospho-Akt pathway (140). However, direct anti-OX40 stimulation of CD8 T cells does not overcome the need for CD4 T cells to promote the long-term maintenance of antigen-specific memory CD8 T cells (138, 139).

CD4 T cells can also instruct upregulation of CD27 on LCMV-specific memory CD8 T cells (141) and high CD27 expression correlated with increased proliferative potential and autocrine IL-2 production. Moreover, CD27 signaling on CD4 T cells also contributes to their ability to support the programming of memory CD8 T cells during an intranasal OVA immunization protocol and the protein MS4A4B (membrane-spanning 4 domains, subfamily A, member 4B, also known as Chandra) seems to be a marker of memory CD8 T cells having experienced CD27-proficient CD4 T cell help during priming (142).

Besides direct cell–cell interactions, CD4 T cells can also provide help either by producing soluble factors themselves or by instructing other cells to produce soluble mediators. Recent evidence indicates that CD8 T cells primed in the absence of IL-2 have a profound impairment in their ability to undergo secondary expansion after re-encounter with antigen. Interestingly, IL-2 signals perceived during the priming period seem to be essential for the programming of memory CD8 T cells. However, this CD4 T cell help is likely not to be conferred in an antigen-specific, cognate manner (143–145). Additionally, IL-15, even though not produced by the CD4 T cells themselves, can function as a mediator of CD4 T cell help. DCs, activated and matured by antigen-specific CD4 T cell help, secrete and bind IL-15, which then is trans-presented to CD8 T cells and thereby enhances CD8 T cell priming and memory formation (146, 147). Interestingly, experimentally induced IL-15 overexpression in DCs can circumvent the necessity of CD4 T cell help for efficient priming and memory formation of CD8 T cells (148). Furthermore, the quality of CD8 T cell priming might also be influenced by CCR5 ligands, which are produced either by CD4 T cells or by APCs having engaged with CD4 T cells (149).

“Helpless” memory CD8 T cells generated and maintained in a CD4 T cell-deficient environment can commit to a genetic program in which they undergo activation-induced cell death upon restimulation. For LCMV infection it was shown that tumor necrosis factor-related apoptosis inducing ligand (TRAIL) mRNA was selectively upregulated in helpless memory CD8 T cells following in vitro restimulation, and a blockade of TRAIL could rescue the secondary expansion of helpless memory CD8 T cells (20). However, two other reports using TRAIL knockout mice have challenged this model by showing that TRAIL deficiency is insufficient to overcome the defective functionality of helpless memory CD8 T cells (150, 151). Interestingly, IL-15, if co-delivered with vaccines, induces an upregulation of anti-apoptotic genes and thus prevents TRAIL-mediated apoptosis (148).

There is limited knowledge about the molecular differences between helped and helpless memory CD8 T cells. Recently, it was suggested that differential epigenetic remodeling of genes involved in CD8 T cell functions occurs. Specifically, histone acetylation of the IFNγ promoter region was observed in helped CD8 T cells but not in helpless CD8 T cells, thus promoting IFNγ gene transcription in helped CD8 T cells. Effector functions of helpless CD8 T cells could at least partially be restored by treating the cells with a chemical inhibitor of the histone deacetylase during the priming phase (152, 153). Furthermore, elevated levels of the transcription factor T-bet were identified in unhelped memory CD8 T cells, which is associated with an effector memory phenotype. Deletion of T-bet drives the differentiation of central memory CD8 T cells and more importantly prevents the phenotypic and functional defects in memory CD8 T cells normally seen in the absence of CD4 T cell help (154).

Currently, two models (programming vs maintenance) exist about the time span when CD4 T cell help is necessary. The first model proposes an ‘imprinting’ or programming of memory CD8 T cells by CD4 T cells during the first few days of initial activation (19, 129) and not during the secondary challenge. The second model suggests that CD4 T cell help is required during the entire phase of memory CD8 T cell differentiation and is thus important for the maintenance of memory CD8 T cells (129, 131). However, a recent finding also points to a role of CD4 T cell help during the challenge phase (21). There is no general consensus about the time requirements for CD4 T cell help in different experimental settings and it seems likely that they differ in various experimental models [(134); unpublished observations]. Thus, it will be of importance to dissect and compare the CD4 T cell help requirements in terms of timing and mechanisms individually in the context of various defined experimental models.


Infection with a low dose of either LCMV WE or LCMV Armstrong has been used over the past decades to investigate effector and memory CD8 T cell differentiation upon acute and resolved viral infection (Fig. 1A) (22, 23, 32, 155). This system has proven very helpful in dissecting parameters that influence memory CD8 T cell differentiation and also for the definition of markers that identify early precursor memory cells. Upon primary activation, LCMV-specific memory CD8 T cells expand massively and a majority of these cells differentiate into effector cells, which can be regarded as terminally differentiated cells. This differentiation process is characterized by progressive changes in gene expression, with upregulation of genes related to cytotoxicity, antiviral cytokines, chemokines and telomerase activity (156). Furthermore, the transcription factors eomesodermin and T-bet have instructive roles for the differentiation of cytotoxic effector functions and are essential for the development of memory cells with long-term self-renewal capacity (157). Terminally differentiated effector cells, after several rounds of cell division, loose their proliferative capacity and become sensitive to activation-induced cell death. An elegant global gene expression analysis of naïve, effector and memory populations of LCMV-specific CD8 T cells revealed that besides expression of key effector molecules such as granzymes, perforin and cytokines, many genes involved in signal transduction, cell migration and cell division are differentially expressed in memory vs naïve LCMV-specific CD8 T cells (156). In fact, about 60% of the gene expression changes were comparable in effector and memory cells in relation to naïve cells, indicating that a majority of the gene expression pattern seems stably altered in LCMV-specific memory cells. However, in this study, effector and memory CD8 T cells were defined solely by the time of their isolation after LCMV infection, with the population of effector cells being isolated at day 8 after infection and the memory cells at day >40 after infection. Thus, there was no distinction made between subsets of memory cells present at day 8 or day >40 after infection.

Figure 1.

 Virus load, CD8 T cell dynamics and CD8 T cell composition during acute/resolved infection (A), latent persistent infection (B) and active chronic infection (C). During acute/resolved infection (e.g. low-dose LCMV Armstrong or LCMV WE, A), infectious virus is typically cleared 8–10 days after infection due to the cytotoxic activity of LCMV-specific CD8 T cells. At the peak of the CD8 expansion (day 8) about 95% of the LCMV-specific T cells show an effector phenotype and function and only about 5% of the cells survive the ensuing contraction phase. Early during memory development (day 30), a majority of the LCMV-specific CD8 T cell population exhibits a TEM-like phenotype whereas late (>day 200) after infection, the population of LCMV-specific memory CD8 T cells exhibits a TCM-like phenotype. During latent, persistent infection (e.g. MCMV, B), infectious virus is cleared with different kinetics in various organs, the salivary gland being the last organ in this sequence (about days 50–60). However, MCMV infection is never completely cleared and MCMV persists in the form of a latent infection with occasional sporadic and local reactivation events that are immediately controlled by the immune response. MCMV-specific CD8 T cells peak in most organs 6–8 days after infection and the overall MCMV-specific CD8 T cell population is composed of many individual specificities that behave quite differently during acute and latent infection. Whereas CD8 T cells with certain specificities show similar dynamics as LCMV-specific CD8 T cells in (A), with a peak expansion followed by contraction and maintenance of stable memory frequencies (‘non-inflators’ in red), others show more moderate peak expansions but continue to increase in frequencies and numbers during the period of viral latency (‘inflators’ in pink). The ‘non-inflators’ exhibit a TCM-like phenotype whereas the ‘inflators’ exhibit a TEM-like phenotype. During chronic/active infection (e.g. high-dose LCMV clone13 or LCMV Docile, C), infectious virus may be cleared after 2–3 months of infection or may never be controlled, depending on the dose of the inoculum and the immuno-competence of the host. LCMV-specific CD8 T cells show peak expansions between days 6 and 8 followed by a contraction phase that, depending on the epitope-specificity, either leads to a reduction of LCMV-specific CD8 T cell numbers (red) or even culminates in their complete deletion (pink). The population of LCMV-specific CD8 T cells that is maintained during chronic active infection exhibits a high in vivo turnover and is completely dependent on the presence of cognate antigen for their maintenance as a population. CD8 T cells show an exhausted phenotype (characterized by dysfunctional cytokine production) and express elevated levels of PD-1; PD-1hi expressing cells seem to be terminally exhausted (‘exhausted-term.’) and cannot be reinvigorated by PD-L1 blockade whereas PD-1int expressing cells (‘exhausted-preterm.’) can be functionally restored by PD-L1 blockade. LCMV, lymphocytic choriomeningitis virus; MCMV, murine cytomegalovirus; TEM, effector memory T cell; TCM, central memory T cells; PD-1, prozrammed death-1; PD-L1, programmed death-1 ligand 1.

Concomitant to the presence of terminally differentiated effector cells (at the peak of the response), a small population (about 5%) of antigen-specific cells exists that is believed to represent the memory precursor cells (37, 38). These cells differ from the terminally differentiated effector cells by their ability to retain their proliferative potential and their multipotency, as they survive the contraction phase, are responding to the cytokines IL-7 and IL-15 by homeostatic turnover (i.e. they show self-renewal) (33, 158–161) and can rapidly differentiate into effector cells upon secondary encounter with the antigen (12). Several markers have been characterized that identify the population of these long-lived memory precursor cells. The first of these markers was IL-7Rα, which is expressed on about 5% of the antigen-specific CD8 T cells at the peak of expansion (8, 34, 124). Sorting of LCMV-specific IL-7Rα+ and IL-7Rα cells at the peak of the expansion, followed by adoptive transfer into naïve hosts, showed that the IL-7Rα+ cells were efficiently maintained in the antigen-naïve host, whereas the IL-7Rα cells continuously declined in numbers after transfer (8, 34, 124). More recently, precursors of long-lived memory CD8 T cells were further characterized to express intermediate levels of the KLRG1 and to produce IL-2 (61). These memory precursors were also cytotoxic, expressed granzyme B and produced inflammatory cytokines. Conversely, at the same time point, effector CD8 T cells (non-IL-2 producing, KLRG1hi) exhibited a more differentiated gene expression profile combined with a status of reduced metabolical fitness of short-lived effector cells. The differentiation of the specific population of memory precursor cells was attributed to relatively low levels of in vivo antigen encounter, as a reduction of antigenic stimulation at the later phase of the acute infection promoted the generation of this memory precursor subset (61). Recently, it was also demonstrated that the expression level of the transcription factor T-bet correlates with effector vs memory precursor fate at the peak of the CD8 T cell expansion upon LCMV infection and that the level of T-bet expression relates to the amount of inflammation (in particular IL-12) (107).

It should be noted, however, that high IL-7Rα expression, intermediate KLRG1 and IL-2 production capacity do not unambiguously identify the precursors of long-lived memory cells as it was shown that not all IL-7Rαhi cells differentiate into memory cells (10, 117, 162, 163), some IL-7Rαlow KLRG1hi cells persist into the memory phase and some memory cells express high levels of both IL-7Rα and KLRG1 (61).

As discussed in the introductory chapters, memory cells are composed of a heterogeneous population of antigen-experienced cells and it is difficult to attribute a distinct gene expression profile to these cells as their gene expression pattern is to a large extent reminiscent of the expression pattern of effector cells. Furthermore, direct ex vivo single cell gene expression profiling of effector and memory cells confirmed the large heterogeneity within these populations, with seemingly stochastic gene expression profiles of effector molecules (e.g. perforin, granzymes, etc.) in single memory cells (161, 164). It is conceivable that many in vivo parameters shape the actual gene expression status of a memory cell (for instance the temporal distance to the last antigen encounter or the exposure to inflammatory or homeostatic cytokines). Furthermore, several hallmarks of memory cells (including the proliferative recall potential and the responsiveness to homeostatic cytokines) seem to increase with time after resolution of the acute infection, even though this is not associated with marked changes in the expression of several surface markers that are used to identify memory CD8 T cell subsets. For example, in Sendai virus infection, TCM-like cells had an increased proliferative potential 1 year after infection compared with 1 month after infection (53) and further analysis of the proliferation-competent memory compartment revealed that these cells could be identified by high expression levels of CD27 and CXCR3 (and not by the ‘classical’ TCM markers CCR7 and CD62L).

A hallmark of memory cells that develop during and after an acute resolved infection is their longevity (as a population) and their maintenance in an antigen-independent manner (as evidenced by adoptive transfer of memory cells into naïve hosts). In LCMV infection, the population of LCMV-specific memory cells is maintained at a constant level from about 2 weeks after infection for the rest of the mouse life (up to 3 years) (Fig. 1A and Table 3). This constant number of memory cells is maintained by the action of IL-7 and IL-15, which both allow for a slow antigen-independent turnover of these cells (159). Furthermore, the longevity of memory cells is also positively influenced by upregulation of anti-apoptotic molecules such as Bcl-2 and Bcl-XL (165). However, all of these positive regulators are only operational once the long-lived memory cells have differentiated, i.e. they are not instructive for their differentiation (116, 163, 166, 167).


Much less is known about the development, function and especially the maintenance of CD8 “memory” T cells during persistent latent infection. A prototype of a persistent latent infection is cytomegalovirus (CMV) infection. Although CMVs are strictly species specific, viruses from different hosts are comparable not only due to their genetic homology but also with respect to the type of immune responses they elicit. During primary encounter infectious CMV is detectable in most organs. Although the acute infection is readily controlled by the immune system in most organs after a couple of weeks, productive CMV replication persists in the salivary glands over months before it is eventually controlled and the infection enters into a latent stage. It is believed that during the latent phase of infection low levels of viral gene expression occur sporadically and in a random manner during viral reactivation events (168). However, full assembly and release of infectious virus is suppressed in the immunocompetent host by sequential checkpoints of viral gene transcription (168, 169). Nevertheless, it is assumed that these reactivation events lead to low level of antigen presentation, which shapes the CMV-specific CD8 T cell response in a very peculiar manner. During murine CMV (MCMV) infection the kinetics of the virus-specific CD8 T cell response was extensively studied and can be divided into three parts (170–173): during the acute phase of infection, massive expansion of CMV-specific CD8 T cells is followed by a contraction phase comparable to acute resolved infections. During the following second phase of virus persistence, a subset of MCMV-specific CD8 T cells starts to accumulate, eventually reaching high and stable virus-specific CD8 T cell pools, which are characteristic for the third phase of the response. Similar accumulations of CMV-specific CD8 T cells have been described in seropositive elderly humans and in long-term CMV-infected macaques (174–177).

In mice, a wide variety of MCMV-derived CD8 T cell epitopes are characterized and two distinct subgroups of epitope-specific CD8 T cell responses can be distinguished (170–173, 178, 179): so-called ‘inflationary’ and ‘non-inflationary’ CD8 T cell responses (Fig. 1B and Table 3). Whereas inflationary CD8 T cell responses continuously increase during the first months of viral latency, non-inflationary responses show a pattern reminiscent of CD8 T cell responses in acute resolved infections (i.e. a clonal burst followed by contraction during early stages of infection and establishment of a stable, low-frequency memory pool). Inflationary and non-inflationary CD8 T cell responses cannot only be distinguished by their different kinetics but also by their phenotypical and functional characteristics (172, 173, 180, 181). ‘Non-inflators’ exhibit a classical central memory phenotype with high surface expression of LN homing receptor CD62L as wells as the homeostatic cytokine receptors IL-7Rα and IL-15Rα and the costimulatory molecules CD27 and CD28 (172, 173). In contrast, ‘inflators’ display a more activated phenotype with generally low expression levels CD27 as well as IL-7Rα and IL-15Rβ (172, 181). Cell surface expression of the NK-cell inhibitory molecules NKG2A and KLRG1 is generally low on non-inflators (181).

Functional differences between inflators and non-inflators are less pronounced: CD8 T cells of both subsets can divide extensively ex vivo as well as in vivo after adoptive transfer into acutely infected hosts (181). Further, both CD8 T cell subsets secrete IFNγ and TNFα after ex vivo antigen restimulation and frequencies of IFNγ-positive cells are comparable to tetramer positive cells, arguing against functional exhaustion of MCMV-experienced CD8 T cells (172, 181). Further, both subsets are able to kill peptide loaded target cells in vivo as well as ex vivo to a similar degree (172). The main functional differences between the two populations lies in their capacity to produce IL-2: whereas a small fraction of non-inflationary CD8 T cells secretes IL-2 after ex vivo short-term stimulation, inflators do not produce IL-2 (172, 181).

The mechanisms how CMV-specific CD8 T cells are triggered to accumulate over time are still poorly understood, albeit accumulation seems to depend on IL-2R signaling (144). It is believed that sporadic gene expression during latent infection leads to antigen presentation, which results in continued stimulation of CMV-specific CD8 T cells (182–184). However, whether antigen is presented on MHC class I of infected cells or by bystander cells through cross-presenting is not clear yet. In vitro, CMV-specific CD8 T cells are not stimulated directly by infected cells such as fibroblasts due to CMV-encoded immuno-evasins that downregulate MHC class I expression on actively infected cells [reviewed in (185)]. Infection of mice with a recombinant MCMV virus lacking expression of the major immuno-evasin genes had no effect on the epitope distribution, magnitude and accumulation patterns of the CD8 T cell response (186). This led to the conclusion that CD8 T cells are mainly primed and maintained through cross-presentation and not by directly infected cells during acute and latent MCMV infection. Interestingly, inflation of MCMV-specific CD8 T cells did not occur in MCMV-infected germ-free mice (187). It remains to be shown if the microbiota is responsible for low-level CMV reactivation events and hence low level of antigen presentation, or if the microbiota is required to shape the antigen-responsiveness of the immune system per se, or if cross-reactivity between MCMV-derived and microbiota-derived epitopes is responsible for the accumulation of virus-specific CD8 T cells.

In contrast to the non-inflationary CD8 T cell population, maintenance of the inflationary CD8 T cell pool seems to be antigen-dependent as these CD8 T cells were unable to undergo homeostatic proliferation after adoptive transfer into naïve hosts (172, 181), which is already implied by their low expression levels of IL-15R and IL-7R. Interestingly, even adoptive transfer of CMV-specific splenic CD8 T cells into latently CMV-infected mice led to a decline of transferred cells over time with only sporadic cell divisions, suggesting that priming and recruitment of new thymic emigrants is needed to maintain the constant high numbers of CMV-specific CD8 T cells. However, at least two findings contradict this notion: first, splenic CD8 T cells, isolated during acute infection were maintained in recipients after adoptive transfer during an extended period of time. Second, thymectomized mice show a similar pattern of inflating CD8 T cell populations as their untreated counterparts up to day 200 post-infection (unpublished data by S. M. Walton). Furthermore, it was shown that certain HCMV-specific CD8 T cell clones persist in their respective hosts over an extended period of time (188, 189). Thus, priming of naïve thymic emigrants may occur and may contribute to a certain extent to the population of CMV-specific CD8 T cells during latency, but is most likely not the main mechanism responsible for maintenance and accumulation of CMV-specific CD8 T cells.

Although CMV-specific CD8 T cells clearly show protective capacities against CMV disease in adoptive transfer models after immunosuppressive treatment [reviewed in (190)], their role during a normal course of CMV infection remains somewhat elusive. In a recent study, Simon et al. (169) showed that Ld-restricted IE1 CD8 T cells play a relevant role in inhibiting progression of transcription from IE1 genes to IE3 and even late genes during reactivation events in the latent phase of infection. However, control of viral replication during acute MCMV infection as well as establishment of latency is not influenced by the absence of CD8 T cells (191, 192). Also, CMV-specific CD8 T cells are only partly needed to keep virus in a latent stage as full virus replication is prevented in an hierarchical and redundant manner by several cell subsets of the immune system such as NK cells, B cells and CD4 and CD8 T cells (193).

In summary, in latent/persistent infection with low antigen loads, antigen-experienced CD8 T cells exhibit a spectrum of effector and memory phenotypes, depending on their specificity. CD8 T cells with effector phenotypes are maintained over a long period of time likely due to constant low level of antigen stimulation. These cells are not functionally exhausted as they are able to produce high amounts of proinflammatory cytokines and can protect immunosuppressed hosts in adoptive transfer studies.


The development of long-lived antigen-independent memory CD8 T cells is precluded when the antigen persists at high levels. In case of chronic virus infections, cancers (194) and persistent bacterial infections (195), continued in vivo exposure to high levels of antigen leads to severe dysfunction of CD8 T cells, also designated CD8 T cell exhaustion. Originally CD8 T cell exhaustion was described after chronic infection of mice with LCMV (196), but was also shown to occur after HCV (197) and HIV (198–202) infection of humans, SIV in rhesus macaques (203), polyomavirus (204), adenovirus (205), Friend retrovirus (206) and mouse hepatitis virus (207) infection of mice.

Chronic productive infection is well studied after high-dose infection of mice with LCMV clone13 or LCMV Docile. Infectious virus may be cleared after 2–3 months or may never be controlled, depending on the dose of the inoculum and the immuno-competence of the host. In this setting, the LCMV-specific CD8 T cell pool expands and contracts early after infection. Depending on the epitope specificity, LCMV-specific CD8 T cells are maintained as a population (e.g. LCMV gp33-specific cells) due to continuous stimulation with antigen or are physically deleted (e.g. LCMV np396-specific cells) (196, 208, 209) (Fig. 1C and Table 3). Furthermore, LCMV-specific CD8 T cells gradually lose their effector functions: first their ability to produce IL-2 and to proliferate is lost, then TNFα secretion is stopped and last IFNγ production is forfeited. Degranulation and cytotoxicity can also be impaired (209, 210); however, this is not generally the case (211). In a recent report exhausted LCMV-specific CD8 T cells were found to overexpress inhibitory receptors, to exhibit major changes in TCR and cytokine signaling pathways, to have altered expression of genes involved in chemotaxis, adhesion and migration and to have profound metabolic and bioenergetic deficiencies (212).

In contrast to acute infections, where central and effector memory CD8 T cells develop, the wide majority of LCMV-specific CD8 T cells stays CD62L negative and IL-7Rα chain negative (8). Despite low expression levels of the receptors for the homeostatic cytokines IL-7 and IL-15 (65, 66, 213), LCMV-specific CD8 T cells persist during chronic infection. Exhausted LCMV-specific CD8 T cells are completely dependent on the presence of cognate antigen for their maintenance (66) and they cycle more rapidly than classical memory CD8 T cells (66, 67). Whether the maintenance of the CD8 T cell pool is dependent on recruitment of naïve thymic emigrants is under discussion. It was suggested that the virus-specific CD8 T cell population is constantly replenished by recruitment of naïve thymic output, which leads to continuous priming of new virus-specific CD8 T cells (214). However, a congenically marked donor population remained stable during continuous antigen exposure (66) and the population of virus-specific CD8 T cells was stable in thymectomized mice (215), suggesting that the maintenance of the CD8 T cell pool is independent of recruitment of thymic emigrants.

Maintenance of exhausted CD8 T cells also requires the presence of IL-2 (143, 144, 216). This could be demonstrated by injection of IL-2 during chronic LCMV infection, which sustained the LCMV-specific CD8 T cell response (216). Also chronic infection of mixed bone marrow chimeras with 50% IL-2Rα-deficient and 50% IL-2Rα-competent bone marrow showed that the IL-2Rα-deficient LCMV-specific CD8 T cells were rapidly depleted after initial expansion, while IL-2Rα-competent LCMV-specific CD8 T cells were maintained (144).

Up to now the molecular mechanisms inducing CD8 T cell exhaustion after chronic viral infection are not very well defined. Recently, a selective silencing of signaling pathways was shown to account for the block in cytokine production. Despite intact TCR-triggered Ca2+ flux, intact ERK phosphorylation and NFκB translocation and normal levels of cytoplasmic NFAT2, exhausted CD8 T cells show an impairment of activation-induced nuclear translocation of NFAT2 (67).

Another critical factor controlling the induction of CD8 T cell exhaustion was shown to be TCR signaling inhibitory molecule Cbl-b. Cbl-b is an E3 ligase and downregulates T cell signaling by ubiquitination of downstream effectors like protein kinase C θ and phospholipase C-γ1. Interestingly, downregulation of IFNγ production in gp33-specific CD8 T cells of chronically infected Cbl-b−/− mice was significantly delayed compared with chronically infected wild-type mice and the number of np396-specific CD8 T cells decreased significantly slower after chronic infection of Cbl-b−/− than of wild-type mice (217), pointing to a role of Cbl-b in enhancing CD8 T cell exhaustion.

Two additional pathways downregulating the functional activity of CD8 T cells during chronic viral infection were identified: the programmed death-1/programmed death-1 ligand 1 (PD-1/PD-L1) pathway and the IL-10/IL-10R pathway. Blockade of these pathways not only leads to increased CD8 T cell effector functions, but even to viral clearance (218–220).

PD-1, an inhibitory receptor of the CD28 superfamily (221), was shown to be expressed in elevated levels on exhausted CD8 T cells compared with functional memory CD8 T cells in the LCMV system (218) as well as on HIV (222–227), HCV (228–230) and tumor cell-specific CD8 T cells (231, 232). Blockade of PD-1/PD-L1 interactions after chronic LCMV infection increased the numbers of LCMV-specific CD8 T cells. Strikingly, these were functionally enhanced and could clear the virus (218). This was even the case in the absence of CD4 T cell help, i.e. under more stringent conditions as helpless CD8 T cells show more pronounced functional defects (218). Blocking antibodies directed against PD-L1 were also shown to enhance therapeutic vaccination during chronic LCMV infection (233). Recently, PD-1 expressing CD8 T cells were divided into PD-1hi and PD-1int expressing cells. In contrast to the PD-1int population, which can be functionally restored by PD-L1 blockade, the PD-1hi subset seems to be terminally exhausted and cannot be reinvigorated by PD-L1 blockade (234).

CTLA-4, another CD28 family member, seems to be upregulated specifically by HIV-specific CD4 T cells but not by CD8 T cells. Antibody-mediated blockade of CTLA-4 was shown to reverse HIV-specific CD4 T cell exhaustion in vitro (235). Nevertheless, even though CTLA-4 is also upregulated on CD8 T cells after chronic LCMV infection, in this system CTLA-4 blockade does not seem to increase CD8 T cell functionality (218).

Clearance of a chronic LCMV infection can also be observed after blocking the interaction of the inhibitory cytokine IL-10 with its receptor (219) and in IL-10 knockout mice (220). Lack of IL-10R signaling led to higher frequencies of virus-specific CD4 and CD8 T cells and CD8 T cells were able to produce enhanced amounts of IFNγ, TNF and IL-2 (219, 220). IL-10 blocking was also shown to facilitate vaccine-induced CD8 T cell responses and accelerates viral clearance of persistent LCMV infection (236).

Recently, transient FYT720 treatment was suggested for the treatment of chronic infections (237). FYT720 is generally regarded as an immunosuppressive drug because it induces a profound but transient sequestration of lymphocytes in the LNs. Strikingly, FYT720 treatment not only led to viral clearance if the mice were treated early after infection, but also during established LCMV infection (237). The mechanisms remain to be elucidated. Whether treatment with FYT720 or with antibodies blocking PD-1/PD-L1 or IL-10/IL-10R interactions can be used to enhance therapy of human chronic infections without induction of immunopathology will have to be investigated in detail.


CD8 memory T cells represent a spectrum of cells with different functions, phenotypes and half-lives, dependence on homeostatic cytokines or antigen that is critically shaped by the biology of the infecting pathogen (strength and duration of activating signals, amount of inflammation, persistence). In general, less antigen exposure and less inflammation favor the differentiation of a relative larger proportion of long-lived TCM-like memory cells. Conversely, more inflammation and longer duration of antigen exposure produced relatively greater proportions of short-lived TEM-like cells. With respect to vaccines that should operate at a CD8 T cell level, it seems of advantage to generate both types of memory cells, as TCM-like memory cells are long lived and a source for rapid secondary expansion while TEM allow more rapid effector functions at peripheral sites of infection. In case of a chronic, active viral infection, the continuous in vivo antigen exposure in combination with negative regulatory mechanisms prevents the development of long-lived antigen-independent memory CD8 T cells and even leads to severe dysfunctions within virus-specific CD8 T cells. Specific interference with these negative regulatory mechanisms (such as by IL-10 or PD-1 blocking) might open new intervention strategies to reinvigorate CD8 T cell function in these situations – however, potential immunopathological consequences of such interventions have to be thoroughly evaluated.