The Ugly Duckling Turned to Swan: A Change in Perception of Bystander-Activated Memory CD8 T Cells

A hallmark feature of adaptive immunity is the development of immunologic memory. CD8 memory T cells (T_mem) respond rapidly upon reencounter with cognate Ag by acquiring effector function and initiating cell division (1). Bystander activation of CD8 T_mem is driven by proinflammatory cytokines such as type I IFNs, IL-12, IL-15, and IL-18 and occurs in the absence of agonist TCR signals. This has been demonstrated in mouse models using TCR-transgenic T cells as well as with polyclonal CD8 T_mem using Nur77-GFP reporter mice. Briefly, GFP expression is transiently increased in T cells of these reporter mice after the cells receive a TCR signal, including those mediated by very weak agonists (2). Bystander-activated CD8 T_mem did not show increased GFP expression, demonstrating that bystander activation occurs independently of (measurable) agonist TCR signals (3). Thus, bystander activation of CD8 T_mem is a distinct phenomenon from TCR cross-reactivity, which can occur even in unrelated infections but is still TCR-mediated (4). The direct functional outcome of TCR-mediated versus bystander-mediated T_mem activation appears remarkably similar and can in both instances include T cell proliferation (5), cytokine expression (6–9), and direct target cytolysis (3, 10–12). The Ag-specific T cell response is stringently regulated and requires two signals (TCR + costimulation) to allow for the initial activation of a naive T cell, and even a third signal (such as type I IFN or IL-12) for a CD8 T cell to acquire effector function (13). Furthermore, the duration of effector function is inherently limited, as T cells acquire expression of inhibitory proteins during the effector stage including expression of PD-1 and CTLA-4 (14). Prolonged activation of T cells leads to T cell exhaustion (also referred to as T cell dysfunction), which is characterized by a TCR signaling–dependent loss of the ability to proliferate or produce effector molecules, such as IFN-γ, in response to stimuli (15).

Given that numerous regulatory mechanisms are in place to tightly control the TCR-driven effector T cell response, it raises the question why T_mem are seemingly easily activated by inflammatory signals. One could argue that control mechanisms to regulate bystander activation of T_mem may not be critical if it only occurs in very rare and specific scenarios such as systemic viral infections, which is how bystander activation was initially discovered (5). Indeed, it had been proposed that bystander activation is not of major biological consequence (16). A lack of relevance could be possible if bystander activation of T_mem was a vestigial feature that stems from the gradual development of the adaptive immune system from the innate immune system (17). In this review, we discuss the evidence that suggests that bystander activation is neither a rare occurrence nor a vestigial feature but a critical part of the early immune response to infection as well as the ongoing immune response during chronic infections and other conditions of chronic inflammation.

The first indication that CD8 T_mem become activated during an infection even in the absence of Ag was made by Tough and colleagues (5) when they reported that a large fraction of CD8 T_mem proliferated in response to systemic infection with lymphocytic choriomeningitis virus, vaccinia virus, and vesicular stomatitis virus. Based on the large fraction of CD8 T_mem that proliferated, Tough et al. hypothesized that proliferation was driven by type I IFN rather than Ag. To test this hypothesis, they induced a type I IFN response by injecting mice with poly I:C and again observed that a substantial part of the T_mem population proliferated and referred to it as bystander proliferation. Importantly, bystander proliferation was much more limited compared with Ag-driven proliferation, as it appeared to only induce one round of cell division (5, 18). Whereas Ag-specific T cell expansion results in a substantial temporary increase of the T cell compartment during acute infections, bystander proliferation is seemingly providing little to no contribution to this increase in T cell numbers and also appears to have a negligible effect on the overall size of the T_mem compartment (18). Studies by Welsh and colleagues (19, 20) suggested that bystander activation could even lead to a net loss of T_mem during heterologous infections. Although these initial observations of bystander activation focused primarily on proliferation, subsequent studies found that bystander-activated CD8 T_mem also gained effector function, including the ability to secrete IFN-γ and express granzyme B (GzmB) (3, 7, 8, 21), which led to the shift in terminology from bystander proliferation to bystander activation. Importantly, bystander activation of CD8 T_mem that are not Ag specific has also been demonstrated during the course of systemic viral infections in human cohorts including primary HIV (22, 23), primary EBV (24), and (to varying degrees) during acute dengue virus infections (25–27). In these latter human studies, peptide/MHC tetramers were used to distinguish Ag-specific from bystander-activated T_mem. Of note, using human T cells specific for chronic viral infections such as CMV and EBV to assess bystander activation must be done carefully because it is challenging to distinguish between bystander activation and activation as a result of local viral reactivation.

Bystander activation of CD8 T_mem is not limited to systemic viral infections. Bacterial motifs can similarly elicit inflammation to bystander activate CD8 T_mem (28–30), as do systemic Listeria monocytogenes (3, 6–9, 31–33) and Yersinia pseudotuberculosis (33) infections. Importantly, recent studies demonstrate that bystander activation of CD8 T_mem also occurs during acute, localized infections. The initial splenic phase of a low-dose L. monocytogenes infection (6), lung infection with Staphylococcus aureus (34), or influenza A virus (IAV) (21, 35) all result in bystander activation of CD8 T_mem at the site of infection.

In context of a low-dose L. monocytogenes infection that was delivered i.v., bystander activation of CD8 T_mem was observed within the first 24 h of infection. These bystander-activated T cells were specifically enriched in splenic white pulps with L. monocytogenes lesions and expressed GzmB but did not express Ki-67, suggesting that they were not proliferating (6). The size of the bystander-activated CD8 T_mem population increased in the subsequent days, but then declined substantially between days 3 and 5 postinfection. Similarly, experiments that challenged mice intranasally with LPS or heat-killed or live bacteria revealed that tissue resident memory T cells (T_RM) located in the lung were bystander activated within 3 h of challenge (34). Together, these studies highlight that bystander activation occurs rapidly in the early stages of an immune response. Although studying acute localized infections is easily accomplished in the mouse model system, studying a defined localized primary infection is much more challenging in human cohorts. A recent set of studies used liver biopsies from patients diagnosed with acute hepatitis A (AHA) and found evidence of bystander activation of CD8 T_mem in these patients (10). Of note, hepatitis A virus (HAV) does not cause chronic liver disease, but symptoms and inflammation can last for weeks and up to several months, which raises the question if a state of bystander activation is maintained in settings of prolonged, chronic inflammation, or chronic infections.

Animal models demonstrate that chronic Leishmania major (11, 12) and Borrelia burgdorferi (36) infections result in prolonged bystander activation of CD8 T_mem at sites of infection. Similarly, bystander activation of CD8 T_mem was observed in patients infected with hepatitis B virus (25, 37). Two recent reviews provide a more detailed overview of bystander activation in context of chronic infections (38, 39). Although bystander activation of T_mem is observed in settings of chronic inflammation, it is noteworthy that it is less clear how long the state of bystander activation can last for a CD8 T_mem on a single-cell level. To our knowledge, there are no studies that have addressed whether a CD8 T_mem can remain bystander activated for prolonged periods of time. Other possibilities include that bystander-activated CD8 T_mem could change expression of chemokine receptors and leave the site of inflammation, they could become nonresponsive to proinflammatory cytokines or have other regulatory mechanisms that would allow a return to steady state. Bystander-activated CD8 T_mem could also undergo apoptosis or related forms of programmed cell death. This has been observed in some mouse models of heterologous viral infections (19), in which chronic STAT1 signaling, mediated by type I IFNs (20) or IL-6 (40), drives bystander CD8 T_mem loss.

Although bystander activation has been predominantly studied in the context of various infections, there is also evidence for bystander activation of CD8 T_mem by sterile inflammation, vaccination, and the tumor microenvironment (TME). Evidence of bystander activation of CD8 T_mem has been observed in rheumatoid arthritis (RA) (41) and celiac disease (42). A recent study by Simoni and colleagues revealed the presence of bystander CD8 T_mem, including IAV-, CMV-, and EBV-specific T cells, in human solid tumors. It is still unclear if these CD8 T_mem are merely bystanders in the sense that they are not tumor-specific, or if they are actually bystander activated and secrete IFN-γ or other effector molecules (43). Finally, recent evidence suggests that i.m. vaccination with a modified vaccinia Ankara–based vector elicits an inflammatory response that is sufficient to bystander activate CD8 T_mem in a human cohort and can be measured 3 d postimmunization, using a preimmunization blood draw as a baseline (6). Another study examined bystander activation following yellow fever virus (YFV) and Dryvax vaccination and did not find evidence of bystander activation (44). However, the earliest time points examined in in this latter study were days 9 and 11 following Dryvax and YFV administration, respectively. YFV immunization results in transient viremia that can be measured between days 3 and 7 (44). Given that bystander activation of T_mem is transient and depends on the presence of inflammatory cues, the lack of bystander activation at time points past viremia is in line with mouse models of acute infection, in which bystander activation is limited to the first few days of infection.

Together, these studies indicate that bystander activation of CD8 T_mem occurs over a wide range of inflammatory conditions, including situations when inflammation is localized and not pathogen induced.

In contrast to T_mem cells, there is no evidence that naive T cells undergo bystander activation. Instead, naive CD8 T cells become briefly activated in context of a systemic viral infection or L. monocytogenes infection, but may undergo apoptosis without a subsequent TCR signal (20, 45). Type I IFN was the first signal identified as sufficient to induce bystander proliferation of T_mem, but it can also lead to bystander activation (21). In addition to type I IFN, at least two of the following cytokines in combination are sufficient to trigger bystander activation: IL-12, IL-15, IL-18 (6–8, 30, 46–48), or TLR2 signaling (36) (Fig. 1A). Freeman and colleagues (46) tested over 43 murine cytokines in over 1800 different cytokine combinations to induce TCR-independent IFN-γ production in a rather heroic effort of identifying additional cytokine combinations that can elicit or diminish bystander activation of CD8 T_mem. The resulting data were complex and highlighted the importance of defining the cytokine composition of an inflammatory environment to understand why a T_mem cell did or did not acquire a bystander-activated phenotype.

Measuring IFN-γ secretion and GzmB expression is certainly useful to assess bystander activation, but it is important to consider that other functional properties, cell fate, etc. may change as well, but these potential changes have not been extensively investigated to date. Similarly, how proinflammatory signals affect different T_mem subsets is poorly understood. Although this brief review is focused on bystander activation of CD8 T_mem, other reviews provide an overview of how CD4 T_mem (36) and mucosal-associated invariant T cells (49, 50) can become bystander activated. Proinflammatory cytokines also activate invariant NKT cells (51, 52) and γδ T cells (53) in the absence of cognate Ag. The notion that cytokine signals may elicit different functional programs is particularly relevant for CD4 T_mem given their functional breadth (54). Importantly, bystander activation is not necessarily homogenous within or across CD8 T_mem subsets. The phenotype of a CD8 T_mem is shaped by its initial Ag-driven encounter (55), variable expression of cytokine receptors across T_mem subsets, which is further affected by previous Ag experience (56), and T cell responses to proinflammatory cytokines such as IL-12 can also vary by biologic sex (57). Following infection with L. monocytogenes, a significant fraction of bystander-activated (i.e., GzmB-expressing) CD8 T_mem had an effector memory phenotype, but bystander activation was also observed in CD8 T_mem with a central memory (CD62L⁺) phenotype (6). In this L. monocytogenes model, splenic CD8 T_RM showed only limited bystander activation, but splenic CD8 T_RM may not be representative of CD8 T_RM in other tissues (34). Together, these data demonstrate that within the CD8 T_mem compartment, there is memory subset and tissue compartment heterogeneity in bystander activation.

CD8 T cells with a memory phenotype are not necessarily generated by previous exposure to Ag. Virtual CD8 T_mem cells acquire a memory phenotype because of homeostatic proliferation (58). Virtual CD8 T_mem cells also become bystander activated during heterologous infection in an IL-15–dependent manner (32), suggesting that bystander effector programs are not exclusive to CD8 T_mem that have been expanded by Ag. White and colleagues provided intriguing evidence that a virtual memory T cell–like population also exists in humans and accumulates with age (32). Given the age-associated increase of memory T cells in humans over time (32), determining how different T_mem subsets contribute to and affect immune responses in the elderly in a TCR-independent manner is of great clinical interest.

Together, these studies highlight that most CD8 T_mem subsets are capable of becoming bystander activated and that this is in part dictated by cytokine receptor expression levels and may be further shaped by the tissue microenvironment.

Professional APCs and other innate cells are thought to have a key role within the first 24 h of infection as the initial source of IFN-γ (59). T_mem have, so far, not been commonly associated with contributing to the early stages of an immune response, which may be related to the fact that mice are kept in specific pathogen-free environments and typically used for experiments at 6–10 wk of age, when they still have a rather limited T_mem compartment. This may have led to underestimating the role of bystander-activated T_mem in this early phase of an immune response. Studies using “dirty mice” will help mimic a more human-like environmental exposure while maintaining the other advantages of the mouse model system (60).

Bystander-activated CD8 T_mem are observed during the earliest stages of infection, well before Ag-specific T cell responses arise (3, 6, 7, 31). Some CD8 T_mem are spatially positioned within lymph nodes to facilitate interactions with pathogen-sensing phagocytes (61), and CD8 T_mem can be bystander activated in tissues in situ (34) (Fig. 2A). Moreover, bystander CD8 T_mem can also relocate to sites of early infection (6, 35, 62, 63). A recent study demonstrated that bystander CD8 T_mem are recruited to early sites of L. monocytogenes replication in a CXCR3-dependent manner (Fig. 2B) (6). This CXCR3-mediated relocation resulted in dense clusters of bystander-activated CD8 T_mem that circumscribed L. monocytogenes lesions as assessed by immunofluorescence (6). Remarkably, this CXCR3-dependent recruitment mechanism mirrors that of Ag-specific T cells (64), and bystander-activated T_mem and Ag-specific T cells can overlap spatially and temporally (6). In context of L. monocytogenes infection, the frequency of bystander-activated CD8 T_mem wanes as Ag-specific T cells start to accumulate in areas of infection at ∼5 d postinfection (6). Similarly, the frequency of bystander-activated CD8 T_mem declines as Ag-specific cells infiltrate IAV-infected lungs (62), suggesting that bystander-activated CD8 T_mem begin to stand down as Ag-specific cytotoxic T cells take over. Because bystander-activated CD8 T_mem secrete IFN-γ and express GzmB, they inevitably affect the early immune response, which, in turn, will shape the subsequent adaptive immune response. The impact of bystander-activated CD8 T_mem on a developing Ag-specific T cell response has not been assessed, but will likely depend on several factors including the site of inflammation, the extent of inflammation, and the availability of Ag.

In general, IFN-γ has been shown to activate microbicidal effector programs in macrophages and other APCs, including production of reactive oxygen species, increased phagocytosis, and upregulation of Ag presentation/costimulatory molecules to generate Ag-specific T cell responses (31, 34, 65) and is thus a central cytokine of the host’s immune response. Mouse model studies with L. monocytogenes have been leveraged to dissect the importance of direct and indirect IFN-γ effects. IFN-γ responses are critical to coordinate primary immune responses against L. monocytogenes (66). During the innate phase of L. monocytogenes infection, IFN-γ–deficient mice suffer from a markedly higher splenic L. monocytogenes burden; however, the adoptive transfer of a wild-type bystander CD8 T_mem population dramatically lowers L. monocytogenes burden at this timepoint (7). This phenomenon is a result of IFN-γ, derived from transferred bystander CD8 T_mem that became activated, orchestrating effector responses in nearby APCs (31) (Fig. 1). IFN-γ signaling leads to an increase in phagocytosis and production of reactive oxygen species by APCs (31), which can directly limit L. monocytogenes replication (67–69). APCs also increase Ag presentation and expression of costimulatory molecules (31), which are critical signals for priming Ag-specific T cells. Ge and colleagues (34) recently reported that the early burst of IFN-γ production by bystander CD8 T_RM recruited neutrophils, which, in turn, limited S. aureus growth during the first 3 d of bacterial pneumonia. Of note, CD8 T cell–derived IFN-γ that is available in the first 24 h of an infection can restrict Ag-specific effector CD8 T cell differentiation in a paracrine manner resulting in an altered effector to memory balance (70). Furthermore, IFN-γ could also act in an autocrine manner on CD8 T_mem (Fig. 1). Together, this suggests that IFN-γ derived from bystander-activated CD8 T_mem interacts upstream and downstream of the innate immune system to control early pathogen replication.

GzmB codelivery with perforin to a target cell results in its apoptotic death, but this cytotoxic payload poses risk to nearby cells (71). Ag-specific CD8 T cells secrete GzmB at the immunosynapse formed by their TCRs and cognate Ag/MHC class I on targets, minimizing the chance of GzmB uptake by unintended targets (71, 72). So how can GzmB-expressing bystander CD8 T_mem kill or identify target cells in the absence of cognate Ag? This remained unclear until it was shown that bystander-activated CD8 T_mem can identify and kill target cells in an NKG2D-dependent manner (Fig. 1C, 1E) (3). The immunoreceptor NKG2D engages a suite of stress-induced NKG2D ligands (NKG2DLs), which serve as generalized signals of infection, stress, or transformation (73–75). NKG2D is used by NK cells to survey for and eliminate NKG2DL-expressing cells. In vivo blockade of NKG2D-NKG2DL interactions led to increased bacterial loads early after infection even in the absence of NK cells, indicating that bystander-activated CD8 T_mem are needed to eliminate L. monocytogenes–infected APCs expressing NKG2DLs (3).

Numerous studies have documented the benefit of effector responses by bystander-activated CD8 T_mem in multiple animal models of infection, including L. monocytogenes (3, 7, 32, 33), IAV (35, 76), Y. pseudotuberculosis (33), murine gammaherpesvirus 4 (77), and S. aureus pneumonia (34). Together, these studies suggest that the role of bystander-activated CD8 T_mem during an acute infection is to help minimize pathogen spreading. Given the timing of how bystander activation appears to resolve when the Ag-specific T cell response takes over, it seems that this mechanism could help minimize the risk that the pathogen outruns host immunity before the Ag-specific adaptive immune response kicks in.

Although NKG2D-dependent killing can help control early pathogen spread during an acute infection, this mechanism was subsequently identified to cause immunopathology in context of chronic infections (11, 12) as well as human AHA (10). During AHA, both HAV-infected and -uninfected hepatocytes upregulate NKG2DLs (10). Bystander-activated CD8 T_mem maintain high NKG2D expression, especially in the presence of IL-15, and can target NKG2DL-expressing hepatocytes, whereas TCR stimulation downregulates NKG2D expression in HAV Ag–specific T cells (10). Bystander-mediated killing of NKG2DL-expressing cells in vitro strongly correlated with measures of liver damage in AHA patients, highlighting the probable role of bystander-activated CD8 T_mem in off-target damage (10). However, it is unclear whether these innate-like killing mechanisms, although capable of off-target damage, contribute to HAV clearance. During chronic cutaneous leishmaniasis, bystander CD8 T_mem are recruited to infected tissues, express GzmB, but fail to upregulate IFN-γ (11). Stromal cells within L. major–infected ears uniformly upregulated NKG2DLs, which rendered them susceptible to NKG2D-mediated killing (11, 12). NK cell depletion did not alter pathology, whereas CD8 T cell depletion and/or NKG2D blockade dramatically reduced tissue pathology (11, 12). The notion that bystander-activated CD8 T_mem use NKG2D to kill target cells in a manner similar to NK cells somewhat blurs the adaptive-innate dichotomy, but NK cells and bystander-activated CD8 T_mem appear to have distinct roles because bystander-activated CD8 T_mem were the main driver of NKG2D-dependent pathology in context of chronic infection (11) and responsible for NKG2D-dependent early pathogen control following acute infection (3).

Similar to chronic infections, bystander-activated CD8 T_mem are also implicated in contributing to pathology in autoimmune diseases. The signals that drive bystander activation may be different in these scenarios. Pathologically high levels of IL-15 are found in affected tissues from patients with celiac disease (42, 78) and RA (79, 80), among other autoimmune disorders [reviewed by Jabri and Abadie (81)]. This IL-15 exposure dually enhances the direct cytotoxicity of bystander-activated CD8 T_mem by upregulating GzmB (82) as well as the NKG2D immunoreceptor (83), which is needed to eliminate NKG2DL-expressing targets. During high IL-15 exposure, the enhanced cytotoxic potential of bystander-activated CD8 T_mem, paired with aberrant expression of NKG2DL in affected tissues, biases for bystander-mediated cytopathies in celiac disease (42) and RA (41). Although bystander-activated CD8 T_mem can kill NKG2DL-expressing targets and propagate autoimmune damage when exposed to pathologically high levels of inflammation, these proinflammatory cues must be maintained to sustain bystander activation. Withdrawal of bystander-activating cytokines can abrogate the cytolytic potential of bystander-activated CD8 T_mem in vitro (84), highlighting the critical role for IL-15 in this context.

TMEs with a lymphocyte infiltrate are also a type of chronically inflamed tissue. Bystander CD8 T_mem infiltrate the TME, in which they can even outnumber tumor-specific T cells and may be distinguished from tumor Ag–specific cells via low CD39 expression (43, 85, 86). Although murine tumor Ag–specific T cells within the TME are typically refractory to stimulation with cognate Ag and cannot be rescued by PD-1 blockade, bystander CD8 T_mem within the tumor maintain responsiveness, spared by their lack of tumor Ag specificity (87).

NKG2D-mediated killing by NK cells and IFN-γ secretion by either tumor Ag–specific T cells and NK cells can contribute to solid tumor clearance (88, 89). Because bystander-activated CD8 T_mem can also employ these mechanisms, they could play a role in antitumor immunity. Indeed, recent mouse model studies demonstrate that bystander CD8 T_mem can contribute to tumor clearance once appropriately activated. Intratumoral injection of the cognate peptide Ag for bystander CD8 T_mem can elicit IFN-γ production and limit tumor growth in vivo (90). Other studies demonstrate that bystander-activated CD8 T_mem can kill tumor cells in an Ag-independent manner in vitro (91–93). Bystander CD8⁺ virtual T_mem become activated and upregulate GzmB upon exposure to tumor cells pretreated with chemotherapeutics that target DNA replication (94) and directly killed MHC class I–deficient targets in a GzmB-dependent manner in vitro (94). Of note, there is strong evidence that recognition of self-antigen/MHC is still required for bystander activation (95). Whether self-antigen/MHC is only required for activation or also to sustain effector function remains to be determined, but is particularly relevant for understanding the antitumor potential of bystander-activated T cells. Although the ability to leverage bystander-activated CD8 T_mem as a therapeutic modality may hinge on the subset and history of these cells (96), these data highlight the potential therapeutic promise of activated bystander CD8 T_mem. Furthermore, this underscores the importance of understanding the yet to be identified cues that permit and regulate bystander-mediated effector functions.

CD8 T cell exhaustion/dysfunction has been proposed to be a functional adaptation to allow for mediating some pathogen control during chronic infections while minimizing tissue damage (97). Similarly, bystander-mediated pathologies during chronic infections may be acceptable collateral damage to help limit pathogen spread. Although mechanisms that control and turn off Ag-specific T cell responses have been studied extensively, relatively little is known in regards to mechanisms that curb bystander activation. Bystander-activated CD8 T_mem appear to be more susceptible to restraint by anti-inflammatory cytokines, including IL-10, than TCR-activated T cells (98). Notably, this IL-10 pathway is disrupted in a model in which TLR2-mediated bystander activation contributes to Lyme arthritis (36). Thus, although TLR2-agonizing pathogen-associated molecular patterns persist in joints affected by Lyme arthritis (99), immunoregulatory cytokines may attenuate improper bystander activation and downstream pathologies.

TCR-mediated activation of T cells also elicits the expression of proteins with inhibitory function such as PD-1 and CTLA-4 and thus provide a cell-intrinsic negative feedback signal to help turn off the T cell response. Importantly, IL-15, among other common γ-chain cytokines, can induce PD-1 expression on CD8 T_mem in vitro (100). Whether this can serve as a cell-intrinsic negative signal in the absence of a TCR signal is unclear. A combination of IL-12, IL-15, and IL-18 was sufficient to induce surface CTLA-4 expression on mucosal-associated invariant T cells in the absence of a TCR signal but did not have the same effect on CD8 T_mem (101). These data highlight that bystander activation has T cell subset–specific effects that require further studies. Together, these data suggest cell-intrinsic and -extrinsic mechanisms may exist to limit bystander activation.

CD8 T_mem populations that can be bystander activated by inflammation are aplenty: conventional T_mem, virtual T_mem, and T_RM all demonstrate the propensity to mount effector responses in an Ag-independent, inflammation-dependent manner. During localized acute infection, bystander-activated T_mem help orchestrate the early immune response via secretion of IFN-γ while simultaneously limiting pathogen spread via direct target cytolysis. These bystander responses are mounted in the first hours of infection and are highly spatially coordinated. Thus, bystander-activated CD8 T_mem can be considered a “first responder,” amplifying the early immune response and preceding appearance of Ag-specific T cells. On the flipside, continuous bystander activation of CD8 T_mem exacerbates tissue damage, causing immunopathologies. Only very few studies have addressed how effector functions of bystander-activated CD8 T_mem are turned off, but a better understanding of these mechanisms will help to leverage bystander CD8 T_mem therapeutically, whether the goal is to enhance bystander activation to improve antitumor immunity or to diminish their function in context of autoimmunity.

The authors have no financial conflicts of interest.