Area under Immunosurveillance: Dedicated Roles of Memory CD8 T-Cell Subsets

Abstract

Immunological memory, defined as the ability to respond in an enhanced manner upon secondary encounter with the same pathogen, can provide substantial protection against infectious disease. The improved protection is mediated in part by different populations of memory CD8 T cells that are retained after primary infection. Memory cells persist in the absence of pathogen-derived antigens and enable secondary CD8 T-cell responses with accelerated kinetics and of larger magnitude after reencounter with the same pathogen. At least three subsets of memory T cells have been defined that are referred to as central memory CD8 T cells (Tcm), effector memory CD8 T cells (Tem), and tissue-resident memory CD8 T cells (Trm). Tcm and Tem are circulating memory T cells that mediate bodywide immune surveillance in search of invading pathogens. In contrast, Trm permanently reside in peripheral barrier tissues, where they form a stationary defensive line of sentinels that alert the immune system upon pathogen reencounter. The characterization of these different subsets has been instrumental in our understanding of the strategies that memory T cells employ to counter invading pathogens. It is clear that memory T cells not only have a numerical advantage over naive T cells resulting in improved protection in secondary responses, but also acquire distinct sets of competencies that assist in pathogen clearance. Nevertheless, inherent challenges are associated with the allocation of memory T cells to a limited number of subsets. The classification of memory T cells into Tcm, Tem, and Trm may not take into account the full extent of the heterogeneity that is observed in the memory population. Therefore, in this review, we will revisit the current classification of memory subsets, elaborate on functional and migratory properties attributed to Tcm, Tem, and Trm, and discuss how potential heterogeneity within these populations arises and persists.

Immunological memory allows our immune system to respond quicker and more effectively against pathogen reencounter. Three main memory CD8 T-cell subsets have been described based on their trafficking and functional characteristics. Central memory CD8 T cells (Tcm) and effector memory CD8 T cells (Tem) circulate and provide bodywide immune surveillance while tissue-resident memory CD8 T cells (Trm) are confined within peripheral tissues where they exert border patrol.

CLASSIFICATION OF MEMORY CD8 T CELLS

Tcm retain migration properties of naive T cells. Similar to naive T cells, Tcm are characterized by the expression of CCR7 and CD62L that facilitate homing to the lymph nodes (LNs). CCR7 attracts Tcm to its ligands CCL19 and CCL21. These chemokines accumulate on the luminal surface of high endothelial venules (HEVs) in the LNs after local production in HEVs or through transcytosis after production in the LN stroma (Gunn et al. 1998; Baekkevold et al. 2001; Carlsen et al. 2005). CD62L mediates the rolling of these lymphocytes on glycan ligands expressed on HEVs. In this manner, these molecules assist in the extravasation of Tcm into the LNs (Ley and Kansas 2004; Förster et al. 2008; Nolz et al. 2011). Tcm are migratory T cells that only transiently remain in the LNs. Egress from the LNs is an active process controlled by sphingosine 1-phosphate (S1P), which serves as a ligand for the Tcm-expressed S1P receptor, S1PR1 (Lee et al. 1998). S1PR1 directs Tcm across a gradient of increasing S1P concentrations between LNs and lymph, resulting in exit from the LNs (Matloubian et al. 2004; Lo et al. 2005). Eventually, Tcm may return to the LNs via the efferent lymphatics, thoracic duct, and the blood (Gowans and Knight 1964; Nolz et al. 2011). Therefore, Tcm, similar to naive T cells, are constantly recirculating secondary lymphoid organs (SLOs) to patrol for pathogen-derived antigens draining from the peripheral tissues (Sallusto et al. 1999). In contrast to naive T cells, Tcm are equipped with trafficking molecules that optimize their migratory responses during infection. Constitutive expression of CXCR3 enables Tcm to migrate into inflamed LNs, where the interferon (IFN)-induced CXCR3 ligands CXCL9 and CXCL10 have been released. This allows Tcm to strategically position themselves at peripheral areas of inflamed LNs, where lymph-borne pathogens are more likely to enter (Sung et al. 2012; Kastenmüller et al. 2013). Tcm probably profit from the proinflammatory LN environment, where high numbers of mature dendritic cells (DCs) are present that provide optimal stimulatory conditions for T-cell responses. Indeed, Tcm undergo massive proliferation upon reencounter with pathogens vastly outperforming other memory T cells in the generation of secondary T-cell responses (Wherry et al. 2003). Tcm produce the T-cell growth factor interleukin 2 (IL-2) in an autocrine fashion, which may support such robust proliferation (Feau et al. 2011). After restimulation, Tcm differentiate into potent effector T cells that up-regulate effector cytokines such as IFN-γ and tumor necrosis factor α (TNF-α) and cytolytic molecules including granzymes and perforin to eradicate pathogens (Wherry et al. 2003). Thus, restimulated Tcm generate a spectrum of effector and memory subsets including secondary Tcm that may ensure robust T-cell responses in subsequent pathogen encounters.

Tcm may not absolutely require the presence of LNs and spleen to generate secondary responses. Tertiary lymphoid structures are sufficient to support memory T-cell responses after influenza infection (Moyron-Quiroz et al. 2006). Moreover, Tcm have been observed to traffic into inflamed peripheral tissues independently of LN surveillance (Osborn et al. 2019). Interestingly, Tcm can rapidly infiltrate the skin after vaccinia infection and provide protective immunity against the virus without trafficking through the LNs (Osborn et al. 2019). Cellular migration of Tcm into nonlymphoid tissues appears to be driven by their capacity to synthesize a unique set of glycans (Osborn et al. 2017). Tcm, in contrast to Tem, express core 2 O-glycans, which generate functional ligands for E- and P-selectins expressed by activated endothelial cells, and endow Tcm with the ability to directly infiltrate nonlymphoid tissues in response to inflammation (Osborn et al. 2017). Therefore, migration into the LNs may not be strictly necessary for Tcm to mount secondary responses after reinfection.

In humans, stem cell memory T cells (Tscm) have been described. Similar to Tcm, Tscm present a naive-like phenotype characterized by the expression of CCR7 and CD62L in combination with memory T-cell-associated molecules including LFA-1, CD95, and IL-15Rb (Gattinoni et al. 2017). Expression of CD45RA distinguishes Tscm from the CD45RO⁺ Tcm subset (Gattinoni et al. 2017). Tscm form a minor fraction of the memory compartment representing only 2%–3% of total circulating T lymphocytes, but display superior capacity for self-renewal, persistence, and expansion after restimulation when compared with other memory subsets including Tcm (Gattinoni et al. 2009, 2011). In addition, Tscm have the multipotent capacity to form all memory T-cell subsets including secondary Tscm after restimulation (Gattinoni et al. 2011). Nevertheless, Tcm also maintain remarkable potential of self-renewal and reexpansion, given that a single Tcm is sufficient to repeatedly form effector and memory responses after consecutive pathogen rechallenges (Graef et al. 2014). Therefore, despite quantitative differences in expansion and self-renewal capacity, both Tcm and Tscm appear to retain remarkable stem-cell capacity.

Tem constitute the other major subset of circulating memory T cells. Tem lack the homing molecules CCR7 and CD62L and consequently have limited access to LNs. In contrast to Tcm, Tem are mainly present in nonlymphoid peripheral tissues, red pulp of the spleen, and in the circulation (Sallusto et al. 1999; Jung et al. 2010). In comparison to Tcm, Tem have poor proliferative capacity and consequently mount secondary effector responses of smaller magnitude upon restimulation (Sallusto et al. 1999). Both Tcm and Tem express transcripts of IFN-γ and TNF-α enabling them to produce these effector cytokines with accelerated kinetics compared to naive T cells after antigenic stimulation (Wherry et al. 2003; Wolint et al. 2004). However, Tem display superior cytotoxic potential compared to Tcm (Wolint et al. 2004; Appay et al. 2008). In humans, storage of cytotoxic effector molecules such as granzyme B and perforin in lytic granules endow Tem with immediate cytotoxic function after degranulation (Appay et al. 2008; van Aalderen et al. 2017). In contrast, in mice, primary Tem do not maintain protein expression of these cytolytic molecules. However, Tem that undergo two or more episodes of antigenic stimulation do retain granzyme B at the protein level (Masopust et al. 2006a). Therefore, Tem can mount faster cytotoxic responses than Tcm, which may relate to superior protection against vaccinia virus and Listeria monocytogenes (Bachmann et al. 2005; Huster et al. 2006). In contrast, Tcm are better than Tem at resolving infection with LCMV, which targets LNs and spleen, suggesting that Tcm may benefit from their direct access to infection sites within the SLOs to counter LCMV infection (Bachmann et al. 2005). Other factors such as the kinetics of viral replication or the spread of the infection may also influence the relative importance of Tem and Tcm in mediating protective immunity.

The characterization of memory T cells as Tem mainly relies on the lack of expression of the homing molecules CCR7 and CD62L, which may have obscured heterogeneity within this memory subset. Differential expression of the fractalkine receptor CX3CR1 may distinguish two separate subsets within the murine Tem population (Gerlach et al. 2016). Importantly, CX3CR1^hi Tem cells were not recovered from tissue-draining afferent lymphatics, which appears inconsistent with the bodywide immune surveillance of Tem (Gerlach et al. 2016). Therefore, CX3CR1^hi Tem appear to restrict their migration to spleen and blood. In contrast, CX3CR1^int Tem were present in the afferent lymphatics, suggesting that these memory T cells survey the peripheral tissues. Therefore, CX3CR1^int Tem were renamed peripheral memory T cells (Tpm). Besides their migratory properties, CX3CR1^hi Tem and CX3CR1^int Tem also differ in their homeostatic characteristics. CX3CR1^int Tem possess higher self-renewal capacity than CX3CR1^hi Tem under steady state and can reacquire CD62L expression, suggesting that these Tem transit into Tcm (Gerlach et al. 2016). Thus, a substantial fraction of Tem may not continuously migrate in and out of the tissues, but instead remain within the bloodstream to prevent systemic spread of infection.

Trm constitute the third main subset of memory T cells. Trm do not recirculate and permanently locate at barrier sites, such as the skin and mucosa of the lungs, intestine, and female reproductive tract, where they form the dominant memory population (Masopust et al. 2010; Jiang et al. 2012; Mackay et al. 2013; Mueller and Mackay 2016). More recently, Trm populations have also been reported in SLOs and in internal organs including brain, liver, and kidneys (Wakim et al. 2012; Schenkel et al. 2014b; Steinert et al. 2015; Fernandez-Ruiz et al. 2016; Smolders et al. 2018). Trm express tissue-retention molecules that prevent exit of these memory T cells from the tissues and adhesion molecules that facilitate interactions with surrounding cells and the extracellular matrix. The C-type lectin, CD69 is essential for the permanent residence of Trm in the tissues (Mackay et al. 2015a). CD69 is one of the first molecules to be up-regulated on activated effector T cells. Thereafter, CD69 is constitutively maintained on Trm under steady-state conditions, but not on other memory T cells. CD69 mediates the internalization and degradation of the tissue exit receptor S1PR1, resulting in complete removal of S1PR1 from the cell surface, which abrogates the tissue egress capacity of Trm (Matloubian et al. 2004; Shiow et al. 2006; Skon et al. 2013; Mackay et al. 2015a). Trm also express the integrins CD103, CD49a, and LFA-1, which are important adhesion molecules that through the binding of their ligands, E-cadherin, collagens, and ICAM-1, respectively, establish essential interactions of Trm with the surroundings (Hofmann and Pircher 2011; Mackay et al. 2013; Cheuk et al. 2017; Kumar et al. 2017; McNamara et al. 2017). It is important to note that the expression of these Trm-associated molecules is not universal; the majority of Trm in the internal organs do not express CD103 and minor Trm populations lacking CD69 expression have been reported (Steinert et al. 2015; Beura et al. 2018b).

Similar to their circulating counterparts, Trm maintain transcripts of effector molecules including IFN-γ and TNF-α at a deployment-ready mode (Hombrink et al. 2016) to ensure the rapid production and release of these proinflammatory mediators. The privileged location of Trm in the epithelial tissues at hotspots of pathogen entry may empower them to mount immediate responses against reinfection (Gebhardt et al. 2009; Jiang et al. 2012). Indeed, activated Trm can rapidly and exponentially increase the magnitude of immune responses through the recruitment of other immune cells including circulating memory cells (Schenkel et al. 2013, 2014a) and through the establishment of a tissue-wide state of alert (Ariotti et al. 2014). Trm populations in brain, small intestine, and liver, but not those in skin also retain protein expression of granzyme B indicating direct cytotoxic function upon encounter of infected cells (Masopust et al. 2006b; Mintern et al. 2007; Steinbach et al. 2016; Cheuk et al. 2017; Kragten et al. 2018). Trm populations may also take advantage of their long-term persistence in the peripheral tissues to fully adapt to the local conditions of their environment. Tissue-specific adaptations of Trm may support custom-made responses to ensure optimal protection upon pathogen encounter (Kadoki et al. 2017). For instance, lung and liver Trm release IL-22 and GM-CSF, respectively, after challenge of vaccinated mice with the poxvirus, resulting in the triggering of IL-22 versus GM-CSF-driven immune responses (Kadoki et al. 2017). Thus, Trm populations in different tissues are heterogeneous, given that they express unique cytokines for the induction of tissue-tailored immune responses (Wong et al. 2016; Cheuk et al. 2017; Kumar et al. 2017).

Taken together, memory CD8 T cells display substantial heterogeneity in terms of their phenotypical, migratory, or functional features, which appears larger than the diversity captured by the classification of CD8 T-cell memory into the three main subsets.

DIFFERENTIATION OF MEMORY CD8 T CELLS

CD8 T-cell responses are initiated with the priming of naive CD8 T cells by DCs in SLOs after pathogen encounter. Naive CD8 T cells with specificity for pathogen-derived peptides are selectively recruited into CD8 T-cell responses. The large repertoire of T-cell specificities only generates a naive T-cell population with any given specificity in the order of 100–1000 cells (Obar et al. 2008). However, these few antigen-specific naive T cells are able to vastly proliferate and differentiate into effector CD8 T cells that migrate into infected tissues to clear the infection (Butz and Bevan 1998; Williams and Bevan 2007). Effector CD8 T cells clear viral infection through the production of proinflammatory cytokines such as IFN-γ and the release of cytotoxic mediators like perforin and granzyme B. Once the pathogen is cleared, the majority of the effector CD8 T cells are removed by apoptosis (Williams and Bevan 2007). Only a minor fraction of ∼5%–10% of effector CD8 T cells survive contraction and give rise to different memory CD8 T-cell subsets that include Tcm, Tem, and Trm. Here, we will discuss how the diversity in memory CD8 T cells arises during their differentiation process from naive T cells.

Diversification of Effector T Cells into Memory T-Cell Subsets

It has been proposed that memory T cells develop along a linear differentiation path from naive T cells that first differentiate into effector T cells and then into memory T cells. Experiments with lineage tracer mice take advantage of the transient expression of granzyme B (Jacob and Baltimore 1999; Bannard et al. 2009) or IFN-γ (Harrington et al. 2008) during the effector stage and show that memory T cells derive of precursors that have at least acquired expression of these effector molecules early after infection. Studies displaying the epigenetic landscape of effector and memory T cells are in line with these findings. The loci of IFN-γ, granzyme B, and of many other genes of effector T cells remain open in memory T cells despite the down-regulation in expression of these effector molecules (Youngblood et al. 2017). Therefore, memory T cells appear to passage through a mandatory effector stage and then down-regulate expression, but not accessibility, of effector functions and phenotypes (Youngblood et al. 2017). The question arises whether this effector phase is required for all memory subsets and in particular for Tscm, which resemble naive T cells regarding their phenotype and transcriptional profile (Gattinoni et al. 2011; Restifo and Gattinoni 2013; Crompton et al. 2016). Recent studies have shown that Tscm retain epigenetic marks of effector molecules (Akondy et al. 2017). In fact, the epigenetic landscape of Tscm closely resembles that of effector T cells rather than that of naive T cells, suggesting that Tscm, similar to other memory subsets, passage through an effector stage (Abdelsamed et al. 2017; Akondy et al. 2017). Thus, all memory subsets appear to derive from cells that have gone through an effector phase.

It has been shown that distinct types of effector T cells with markedly different memory potential appear early in the CD8 T-cell response. Effector T cells can be separated in memory precursors that retain the capacity to develop into memory T cells and in terminal effectors that do not contribute to the formation of memory T cells. The surface molecules KLRG1 and CD127 (IL-7Rα) distinguish between these short-lived effector cells (SLECs, CD127^loKLRG1^hi) and memory precursor effector cells (MPECs, CD127^hiKLRG1^lo) (Kaech et al. 2003). Recently, the offspring of KLRG1⁺ effector T cells has been followed into the memory phase after Listeria infection using Klrg1 lineage reporter mice (Herndler-Brandstetter et al. 2018). The fate-mapping studies showed that part of the KLRG1⁺ effector T cells was able to down-regulate KLRG1 expression and differentiate into memory T cells (Herndler-Brandstetter et al. 2018). These findings suggest that besides MPECs a fraction of KLRG1⁺ effector cells retains memory potential. Importantly, adoptive transfer studies of MPECs have shown that this population develops into both circulating and resident memory CD8 T cells (Joshi et al. 2007; Mackay et al. 2013; Herndler-Brandstetter et al. 2018). Thus, MPECs in contrast to SLECs retain the potential to develop into Tcm, Tem, and Trm. However, from these studies it remains unclear how MPECs differentiate into different memory T-cell subsets.

The diversification between memory precursors and terminal effectors may already take place at the first cell division. Asymmetric division of a naive T cell into daughter cells that inherit unequal amounts of fate-determining proteins may drive separation of effector and memory T-cell lineages (Chang et al. 2007; Arsenio et al. 2014; Pollizzi et al. 2016; Kakaradov et al. 2017). Transcriptional analysis of individual T cells that had undergone their first cell division revealed two main cell clusters that, based on their transcriptional profiles, resembled terminal effector T cells and memory T cells (Kakaradov et al. 2017). These findings underline the early separation between terminal effectors and memory precursors, but do not reveal heterogeneity in memory precursors, suggesting a common memory progenitor of Tcm, Tem, and Trm at this stage (Kakaradov et al. 2017). Therefore, fate divergence among memory subsets may require more than a single cell division.

Indeed, the visualization of individual fates of naive T cells to form effector and memory T cells suggests a higher degree of complexity in CD8 T-cell responses. These conclusions are based on transfer studies, in which the progeny of single naive T cells was traced, and on barcoding studies, in which individual naive T cells were followed in bulk using unique genetic markers. The contribution of naive T cells to effector and memory responses in these studies appeared highly heterogeneous. In particular, the clonal expansion and differentiation potential varied markedly between individual naive T cells (Stemberger et al. 2007; Buchholz et al. 2013; Gerlach et al. 2013). Interestingly, large clones contained few Tem and Tcm precursors and were dominated by terminal effector cells, whereas minor clones contained much stronger representation of Tem and Tcm precursors (Buchholz et al. 2013). This variation in the progeny of naive T cells may be partially explained by multiple rounds of asymmetric division, translating into a myriad of outcomes regarding the development of effector and memory T cells.

The lineage choices of memory CD8 T cells are effectuated by different sets of transcription factors that characterize the three main subsets of memory T cells. The circulating Tcm and Tem populations up-regulate Klf-2, which induces the expression of S1PR1 to facilitate trafficking throughout the LNs and periphery by enabling these memory T cells to exit the tissues and LNs (Matloubian et al. 2004; Bai et al. 2007; Takada et al. 2011). Tcm specifically express the transcription factor Tcf-1, which ensures access to the LNs through up-regulation of CCR7 and CD62L expression (Zhou et al. 2010). In contrast, Trm do not express Klf-2 or Tcf-1, but instead up-regulate Hobit that together with the related transcription factor Blimp-1 suppresses S1PR1 and CCR7, thereby locking these memory T cells within the peripheral tissues (Mackay et al. 2016). Other transcription factors including Runx3, Tbet, and Notch have been implicated in the development and/or maintenance of Trm (Mackay et al. 2015b; Hombrink et al. 2016; Milner et al. 2017). Interestingly, these Trm-associated transcription factors are also important in driving terminal differentiation of effector T cells and in the acquisition of important effector functions including the production of IFN-γ and/or cytotoxicity (Joshi et al. 2007; Cruz-Guilloty et al. 2009; Kallies et al. 2009; Rutishauser et al. 2009; Backer et al. 2014). Therefore, these transcription factors may instruct the rapid up-regulation of effector functions in Trm upon pathogen reencounter. Thus, Tcm, Tem, and Trm each express unique sets of transcription factors that regulate important aspects of these memory subsets.

Factors Dictating Memory T-Cell Differentiation

T-cell responses are initiated when pathogen-specific naive T cells are primed by antigen-presenting cells (APCs) in the LNs. The affinity of T-cell receptors (TCRs) for antigen bound to MHC molecules on the surface of APCs, costimulatory signals, inflammatory cytokines, and the presence of accessory cells during priming may impact on CD8 T-cell fate. Therefore, the “decision” to differentiate into either Tcm, Tem, or Trm may already take place during priming within the LNs. After priming, effector CD8 T cells migrate from the SLOs to infection sites in the periphery to combat the invading pathogens. Factors such as the presence of local antigen and inflammation may influence the “decision” of effector T cells to join the pool of circulating memory T cells or to settle permanently in the peripheral tissues and develop into Trm. Here, we will discuss how external signals in the LNs and at peripheral infection sites instruct the differentiation program of effector T cells into distinct memory subsets.

In the LNs

Antigen dose as well as antigen affinity have a major impact on CD8 T-cell differentiation into effector and memory T cells. CD8 T cells require DCs to present relevant antigens above a certain threshold to establish stable contacts for development into effector and memory T cells (Henrickson et al. 2013). Although low affinity TCR-ligand interactions are still sufficient for naive T cells to mount complete effector and memory responses, the amplitude of these responses is strongly reduced (Zehn et al. 2009). The strength of TCR stimulation may not only determine the magnitude of T-cell responses, but may also impact on memory T subset differentiation. Several lines of evidence suggest that the development of Tcm requires lower thresholds for TCR signaling than the development of Tem. Priming with attenuated pathogens preferentially results in Tcm formation (van Faassen et al. 2005). Limitation of antigen availability using antigen-depleting antibodies increases CD62L expression on developing memory populations (Obar and Lefrançois 2010). Furthermore, elevation of the precursor frequency of naive T cells, which likely increases competition for available antigen and other resources, induces skewing toward Tcm development (Marzo et al. 2005; Badovinac et al. 2007). Supporting these studies, Tcm carry lower affinity TCRs and display higher clonal diversity than Tem and effector T cells (Kedzierska et al. 2006; Knudson et al. 2013). Tcm may also have lower affinity than Trm. Trm express high levels of the MHC class I coreceptor CD8, which facilitates binding to MHC class I peptide complexes, resulting in higher overall avidity of Trm compared to other memory compartments including Tcm (Frost et al. 2015). However, naive CD8 T cells of low affinity appear to have an advantage in forming Trm in the lung after influenza infection, compared to high-affinity naive T cells that preferentially formed terminal effector cells (Fiege et al. 2019). These findings suggest that Tcm have less stringent antigen requirements for development compared to Tem and Trm, which in turn have reduced antigen requirements compared to terminal effector cells.

TCR stimulation results in up-regulation of the expression of inflammatory receptors such as the high-affinity IL-2 receptors IL-2Rα and IL-12Rβ on CD8 T cells, suggesting that inflammatory signals are tied to antigenic signals in the regulation of CD8 T-cell responses (Presky et al. 1996; Kalia et al. 2010; Pipkin et al. 2010). Indeed, proinflammatory cytokines such as IL-2, IL-12, and IFN-γ may specifically regulate effector CD8 T-cell differentiation, as mice with defects in their signaling pathways display compromised effector responses but normal formation of memory T cells (Badovinac et al. 2004; Takemoto et al. 2006; Joshi et al. 2007; Pearce and Shen 2007; Kalia et al. 2010; Pipkin et al. 2010). Furthermore, truncation of Listeria infection using antibiotic treatment, which reduces both inflammation and antigen load, results in impairment of effector responses but in acceleration of memory formation (Williams and Bevan 2004; Badovinac and Harty 2007). These findings indicate that proinflammatory cytokines favor effector over memory formation. Inflammatory cytokines also appear to differentially instruct the development of memory subsets. Whereas formation of Tcm appears diminished under inflammatory stimuli such as IFN-γ and IL-12 (Stoycheva et al. 2015; Danilo et al. 2018), the generation of Trm may benefit from a proinflammatory environment. Interestingly, optimal generation of Trm, but not circulating memory T cells, requires cross-priming by type 1 classical DCs (cDC1s) in the LNs. These cDC1s provide inflammatory signals including IL-12, IL-15, and CD24, which are relevant for the development of Trm (Iborra et al. 2016). The actions of these cDC1s may induce commitment of effector cells to the Trm lineage already during priming in the LNs (Iborra et al. 2016). Therefore, Trm differentiation during early stages of infection appears to require higher levels of inflammation when compared to Tcm and Tem.

In the Periphery

Antigen and inflammation may also play a role in the differentiation of Trm at later developmental stages, which include migration and the establishment of residency within the peripheral tissues (Fig. 1). Inflammatory stimuli are important to induce recruitment of Trm precursors toward the periphery (Mackay et al. 2013). Effector cells that develop into Trm in skin, vaginal mucosa, and lamina propria of the small intestine require the inflammatory chemokine receptors CXCR3 and CCR5 to access these tissues upon inflammation (Mackay et al. 2013; Iijima and Iwasaki 2014; Bergsbaken and Bevan 2015). One source of inflammation may derive from CD4 T cells within the peripheral tissues, which provide inflammatory stimuli such as IFN-γ. The CD4-derived cytokines trigger the up-regulation of the local production of chemokines including CXCL9 and CXCL10, which serve as ligands of CXCR3, thereby attracting effector T cells into the tissues (Nakanishi et al. 2009). In the absence of CD4 T cells, CD8 T cells have impaired the ability to localize to the airway epithelium (Laidlaw et al. 2014) and the vaginal mucosa (Nakanishi et al. 2009), resulting in decreased Trm formation at these sites. However, CD4 T-cell help may not always be essential for Trm development, given that Trm in skin and the female reproductive tract develop independently from CD4 T cells in the vaccinia and LCMV infection model (Jiang et al. 2012; Beura et al. 2018a). Local inflammatory cues may also contribute to Trm differentiation within the lamina propria of the small intestine. At these sites, macrophages producing the inflammatory cytokines IFN-β and IL-12 have been reported to support the formation of CD103⁻ CD69⁺ Trm (Bergsbaken and Bevan 2015; Bergsbaken et al. 2017). Effector T cells do not require secondary antigen encounter in the periphery for development into Trm in skin, salivary glands, and the upper respiratory tract (Mackay et al. 2013; Thom et al. 2015; Pizzolla et al. 2017). However, antigen presentation at peripheral sites such as in the skin (Iborra et al. 2016; Khan et al. 2016) and brain (Schøller et al. 2019) substantially enhances Trm formation and differentiation, suggesting that Trm expansion may benefit from the presence of local antigen. Therefore, antigen and inflammatory cues within the tissue environment are important drivers for memory differentiation of CD8 T cells in peripheral tissues and favor the development of Trm.

Figure 1.

Differentiation of tissue-resident memory CD8 T cells (Trm). The scheme represents a compilation of factors dictating Trm formation in lymph nodes (LNs), bloodstream, and peripheral tissues. (A) After infection, Trm precursors are primed by cDC1 cells, but not cDC2, in the LNs. Trm precursors exit the LNs and enter the circulation following an increasing gradient of S1P. (B) The inflammatory chemokines CXCL9/10 and CCL5 attract Trm precursors, expressing the respective chemokine receptors CXCR3 and CCR5, toward the periphery allowing Trm precursors to leave the blood and infiltrate peripheral tissues at inflamed sites. (C) In the peripheral tissues, local antigen presentation (by DCs or other cell types) may contribute to Trm formation. In tissues, such as the small intestine, TGF-β dependent (CD103⁺) and independent (CD103⁻) Trm are formed. In the epithelium, integrins on epithelial cells remove the latency-associated protein (LAP), the inhibitory component from TGF-β, thereby enabling activity of TGF-β. In response to TGF-β, epithelial Trm up-regulate CD103. In the lamina propria, CD103⁻ Trm cluster with CX3CR1⁺ macrophages that secrete IL-12 and IFN-α/β. (HEVs) high endothelial venules, (cDC1) classical type 1 dendritic cell, (cDC2) classical type 2 dendritic cell, (S1PR1) sphingosine 1-phosphate receptor, (CXCR3) C-X-C chemokine receptor type 3, (CCR5) C-C chemokine receptor type 5, (CXCL9/10) C-X-C chemokine ligand 9/10, (CCL5) C-C chemokine ligand 5, (TGF-β) transforming growth factor β, (IFN-α/β) interferon α/β, (IL-12) interleukin 12.

MAINTENANCE OF MEMORY SUBSETS

Homeostasis during Quiescence

A cardinal feature of memory T cells is their ability to persist after pathogen clearance. The long-term maintenance of memory T cells does not require repeated encounters with a pathogen. In experimental settings in mice, memory CD8 T cells are retained in stable numbers in the complete absence of MHC class I molecules presenting pathogen-derived antigens (Hou et al. 1994; Lau et al. 1994; Murali-Krishna et al. 1999). In humans, volunteers inoculated with live-attenuated viruses such as the vaccinia or yellow fever vaccines maintain virus-specific memory T-cell populations for decades without ever reencountering the viruses during their lifetime after the vaccination (Hammarlund et al. 2003; Akondy et al. 2017). Individual memory T cells have finite life spans, although they can be remarkably long-lived with estimates in the range of several years (Akondy et al. 2017). Therefore, the maintenance of stable memory populations does not only require long-term survival, but also homeostatic proliferation of memory T cells. The estimates for the turnover of memory T-cell populations range between once per 45 days for mice (Choo et al. 2010) and once per 450 days for humans (Akondy et al. 2017). The wide range in doubling times of memory T cells may be accounted for by differences in host species and by the type of infection, but also by differences in the maintenance of memory subsets. Longitudinal follow-up of virus-specific CD8 T cells in yellow fever vaccinees shows development of a repertoire of memory T-cell subsets including Tscm, Tcm, Tem, and Temra. Over time, all of these memory populations gradually decline in numbers with the exception of Tscm (Fuertes Marraco et al. 2015). Memory T-cell populations that persist for more than a decade after the vaccination are nearly exclusively of the Tscm phenotype (Fuertes Marraco et al. 2015; Akondy et al. 2017). These findings suggest that at least after yellow fever vaccination, the persistence of Tcm, Tem, and Temra ultimately has limitations. These limitations do not appear to be shared with Tscm, which are maintained as a stable population in the absence of antigenic challenges. Under experimentally controlled conditions in mice, direct comparisons have been made between the maintenance of Tcm and Tem using an adoptive transfer approach (Wherry et al. 2003). Obviously, the life span of mice limits the follow-up of memory populations over extended stretches of time. Nevertheless, it was observed that over a time span of ∼30 days, Tcm outperformed Tem in terms of maintenance of the population after transfer in naive recipients. The improved stability of Tcm was attributed to their enhanced potential of homeostatic proliferation compared to Tem. Remarkably, in contrast to Tcm, Tem were not stably maintained and a fraction of these memory cells acquired a Tcm phenotype (Wherry et al. 2003). The ability of Tem to reexpress CD62L appears to reside in the Tpm fraction of Tem. In contrast, CX3CR1^hi Tem are stably maintained and do not convert to Tcm (Gerlach et al. 2016). Thus, the circulating memory T-cell populations are largely maintained independent of each other and their persistence displays a hierarchy in descending order from Tscm to Tcm and Tem and from Tcm to Tem.

Similar to circulating memory T cells, Trm populations in skin, liver, small intestine, and the female reproductive tract are stably maintained (Gebhardt et al. 2009; Jiang et al. 2012; Mackay et al. 2012; Bergsbaken and Bevan 2015; Fernandez-Ruiz et al. 2016). Experiments in an adoptive transfer setting to assess whether Trm populations can independently maintain themselves are challenging due to the difficulty to relocate donor Trm into the peripheral tissues. Adoptive transfer is feasible for liver Trm, probably related to their unique location attached to the luminal side of the endothelium of liver sinusoids, but these populations have not been monitored for more than a few days after transfer (Fernandez-Ruiz et al. 2016). Adoptively transferred Tcm or Tem do not contribute to the formation of Trm in the intestine, suggesting that circulating memory T cells lack the capacity to form Trm under steady-state conditions and implicating independent maintenance of intestinal Trm (Masopust et al. 2010). In contrast, the murine lung environment appears unsuitable to support the long-term persistence of Trm. The slow decline in Trm numbers over time may result from the disappearance of scar tissue associated with prior infection sites, where lung Trm preferentially locate. Replenishment of the declining Trm population has been suggested to occur through recruitment from circulating Tem (Slütter et al. 2017). However, it is unclear whether their contribution to Trm maintenance in the lung constitutes a dominant pathway (Takamura and Kohlmeier 2019). More recently, it has been shown that interstitial Trm in the lung can maintain themselves and repopulate Trm in the airways independent of circulating memory T cells (Takamura et al. 2019; Wein et al. 2019). Therefore, Trm throughout tissues, with the notable exception of airway Trm in the lungs, appear to form self-sustaining memory populations.

Homeostatic Cytokines

Memory CD8 T cells constitutively require cues from homeostatic cytokines for their maintenance during quiescence. In particular, IL-7 and IL-15 are important for the maintenance of memory CD8 T cells under steady-state conditions (Surh and Sprent 2008). IL-7 regulates long-term survival and IL-15 induces homeostatic proliferation of memory CD8 T cells (Schluns et al. 2000; Becker et al. 2002; Goldrath et al. 2002; Tan et al. 2002). Memory CD8 T cells acquire these homeostatic cytokines from IL-7-producing stromal cells located in LNs, spleen, and bone marrow (Link et al. 2007) and from IL-15-producing APCs that trans-present the cytokine on IL-15Ra (Mortier et al. 2009). To respond to these homeostatic cytokines, memory CD8 T cells maintain expression of the IL-7Ra (Kaech et al. 2003) and up-regulate expression of IL-15Rb (Zhang et al. 1998; Cho et al. 1999). IL-7 is essential for circulating memory T cells in spleen and LNs (Schluns et al. 2000) and, independently of IL-7, IL-15 is required for the maintenance of Tcm and Tem (Mortier et al. 2009). Interestingly, Tcm acquire essential IL-15 signals mainly through DCs, whereas Tem have access to IL-15 via both DCs and macrophages (Mortier et al. 2009). IL-7 and IL-15 are also required for Trm maintenance in skin (Mackay et al. 2013; Adachi et al. 2015). In fact, Trm appear to preferentially localize around the hair follicles in skin, where keratinocytes of the epithelial sheath form an important source of IL-7 and IL-15 (Adachi et al. 2015). IL-15 is an essential homeostatic cytokine for other Trm populations in salivary glands, kidney, spleen, and LNs, but not for Trm populations in small intestine and the female reproductive tract (Schenkel et al. 2016). Thus, the homeostatic cytokines IL-7 and IL-15 are important for the maintenance of the majority of, but not all, memory CD8 T-cell populations.

The maintenance of Trm within the epithelia of the skin and the small intestine essentially depends on transforming growth factor (TGF)-β. This cytokine is released in complex with the latency-associated protein (LAP), which inactivates TGF-β signaling by preventing access to the TGF-β receptor. Binding of LAP to αvβ6 and αvβ8 integrins on epithelial cells results in removal of the inhibitory component from TGF-β, thereby enabling activity of TGF-β in close proximity of the epithelium (Fig. 1; Mohammed et al. 2016). In response to TGF-β signaling, epithelial Trm up-regulate the CD103 integrin that allows them to strongly bind to the epithelial cell layer via interactions with E-cadherin (Mackay et al. 2013; Zhang and Bevan 2013; Sheridan et al. 2014). A subset of Trm residing in the lamina propria of the small intestine just underneath the epithelium develops independently of TGF-β and do not express CD103, underlining the local action of TGF-β (Fig. 1; Bergsbaken and Bevan 2015). Trm at other internal sites, such as the liver and kidney, also largely lack expression of CD103, suggesting TGF-β independent maintenance at these sites (Steinert et al. 2015; Fernandez-Ruiz et al. 2016). Circulating memory T cells may also temporarily require TGF-β signaling for passage through the skin. The compartment of circulating memory T cells in spleen and at other locations gradually declines in the absence of active TGF-β through deletion of αvβ6 and αvβ8 integrins (Hirai et al. 2019). Blocking access of circulating memory T cells to the peripheral tissues releases the requirement for TGF-β signaling, suggesting that TGF-β is essential only for local survival during transit of circulating memory T cells through the skin. Tcm and CX3CR1^int Tem, but not CX3CR1^hi Tem, depend on TGF-β signals, which is consistent with previous findings, suggesting that only these memory T cells have the capacity to recirculate the peripheral tissues (Gerlach et al. 2016). Thus, TGF-β is an important cytokine for the maintenance of the epithelial Trm and circulating memory T cells with access to peripheral sites.

Memory Maintenance during Recurrent Pathogen Encounters

In our daily lives, we are most likely recurrently exposed to pathogens, suggesting that memory maintenance involves the replenishment of memory T-cell populations after repeated infections. In experimental settings in mice, the impact of reexposure to pathogens on memory differentiation has been studied using sequential infections with distinct pathogens containing a common T-cell epitope. Repetitive stimulation of T cells in vivo not only increases the size of the memory population, but also induces qualitative differences in secondary memory T cells compared to primary memory T cells (Wirth et al. 2010). Secondary memory T cells are superior to primary memory T cells in providing protection against acute infection (Nolz and Harty 2011), but the underlying principles of how secondary memory T cells achieve improved protection are not completely understood. The frequency of Tcm decreases upon repeated stimulation, suggesting essential contributions from more differentiated memory subsets (Nolz and Harty 2011). Indeed, booster immunizations result in a stronger representation of Tem in the memory compartment (Vezys et al. 2009). Secondary and tertiary Tem display signs of terminal differentiation such as up-regulation of KLRG1 expression. These Tem populations also retain more granzyme B expression compared to primary Tem, suggesting enhanced cytotoxic potential (Jabbari and Harty 2006; Masopust et al. 2006a). Therefore, repeated infections appear to reshape CD8 T-cell memory resulting in the presence of a larger contingent of Tem that displays immediate effector functions.

Memory subsets including Tcm, Tem, and Trm are stably maintained as independent populations under steady-state conditions. However, restimulation enables memory T cells to expand and differentiate into secondary effector and memory T cells. Studies using adoptive transfer of isolated memory populations have shown differential potential in memory subsets to establish secondary memory cells upon rechallenge. Tcm retain the potential to reform Tcm, Tem, and effector cells, whereas Tem can reform themselves and effector cells, but lack the capacity to regenerate Tcm (Huster et al. 2006; Graef et al. 2014). Tcm vastly outcompete Tem in the formation of secondary memory T cells, suggesting that the reconstitution of these populations largely resides within this memory subset (Graef et al. 2014). Tcm also retain the potential to reform Trm, although this capacity may be reduced compared to that of naive T cells (Enamorado et al. 2017; Park et al. 2018). Trm are clearly not terminally differentiated cells, as demonstrated by the potential of Trm to generate local effector responses and secondary Trm populations within the peripheral tissues and the draining LNs (Beura et al. 2018a, 2018b; Park et al. 2018). However, it remains unclear how substantial the contribution of Trm is in the reestablishment of secondary Trm populations. Thus, in particular, Tcm retain stem cell–like properties including self-renewal capacity and multipotency that allow them to repopulate the other memory subsets upon pathogen reencounter. Therefore, recurrent infections appear to be relevant for the long-term maintenance of Tem and Trm populations.

CONCLUDING REMARKS

Our immune system has coevolved with the pathogens that target us and, therefore, we can assume that evolution has endowed us with optimal immune responses to combat infections. Protective immunity relies on the complementary action of different memory CD8 T-cell populations (Fig. 2). Three main subsets have been described: Tcm, Tem, and Trm. These memory T cells show different trafficking properties, which directly relate to their specific functions to build an efficient immune response. Tcm exhibit an extraordinary proliferative capacity and, in steady-state conditions, Tcm mainly recirculate throughout the LNs where they can mount robust responses (Sallusto et al. 2004). Consequently, Tcm may primarily protect from infections where high numbers of effector cells are needed such as in systemic infections. Tem expressing high levels of CX3CR1 do not appear to leave the bloodstream under steady-state conditions (Gerlach et al. 2016), and they may permanently patrol the endothelial lining of blood vessels to contain the spread of pathogens. Trm permanently located within the tissues have a restricted scanning area, but if positioned at the site of pathogen entry, they may rapidly neutralize pathogens before establishment of fulminant infection (Mueller and Mackay 2016; Rosato et al. 2017). CX3CR1^int Tem also known as Tpm have been shown to recirculate throughout the peripheral tissues in steady state (Gerlach et al. 2016), suggesting that Tpm exert immunosurveillance at locations not covered by Trm. Thus, the memory populations appear to have dedicated roles in immunosurveillance. However, memory T cells may display a substantial degree of plasticity to efficiently contribute to pathogen eradication. For example, Tcm may not always require the LNs to mount secondary responses and appear to have superior ability to rapidly infiltrate the peripheral tissues (Osborn et al. 2017, 2019). Moreover, Trm can become dislodged and establish residence within draining LNs (Beura et al. 2018b). Thus, immunosurveillance by memory CD8 T cells appears to be a dynamic process that employs the unique characteristics and plasticity of memory CD8 T cells to provide optimal responses against any given pathogen.

Figure 2.

Migration of memory CD8 T cells. (A) In steady state, central memory CD8 T cells (Tcm) enter the lymph nodes (LNs) through high endothelial venules (HEVs) and exit the LNs through efferent lymphatics. Effector memory CD8 T cells (Tem) (CX3CR1^hi) continuously recirculate and do not exit the bloodstream. Tissue-resident memory CD8 T cells (Trm) are locked within peripheral tissues and they scan a defined area of their surroundings. Tem (CX3CR1^int; peripheral memory T cells [Tpm]) survey the periphery and LNs through afferent lymphatics. (B) After infection, Tcm robustly proliferate in the LNs and can enter the periphery to respond against the pathogen. Tem and Tpm are likely recruited to the site of infection to assist in combat of the infection. Trm immediately respond when the pathogen reaches the Trm scanning area. Trm proliferate locally and a proportion of Trm may become dislodged and travel through afferent lymphatics into the draining LNs. Colored arrows indicate the direction of migration; black arrows indicate extravasation.