T cell immunity is often defined in terms of memory lymphocytes that use the blood to access a range of organs. T cells are involved in two patterns of recirculation. In one, the cells shuttle back and forth between blood and secondary lymphoid organs, whereas in the second, memory cells recirculate between blood and nonlymphoid tissues. The latter is a means by which blood T cells control peripheral infection. It is now clear that there exists a distinct memory T cell subset that is absent from blood but found within nonlymphoid tissues. These nonrecirculating tissue-resident memory T (TRM) cells develop within peripheral compartments and never spread beyond their point of lodgement. This review examines fixed immune surveillance by TRM cells, highlighting features that make them potent controllers of infection in nonlymphoid tissues. These features provide clues about TRM cell specialization, such as their ability to deal with sequestered, persisting infections confined to peripheral compartments.
One of the hallmarks of immunological memory is its systemic character, such that a given Ag generates an enhanced secondary response anywhere in the body, regardless of its initial point of encounter. Systemic immunity involves Abs and B and T cells that combine for effective protection against a range of pathogenic agents. Early examinations of lymphocyte migration established that systemic immunity was dependent on circulating lymphocytes, in combination with their passage through the regional lymph nodes (1–3). The latter structures play a pivotal role in adaptive immunity because of their involvement in lymphocyte priming, particularly postinfection at remote sites, such as the skin or mucosal surfaces. T cells have two means by which they access the regional lymph nodes (4, 5). The first is by passing between the large polygonal cells that line the high endothelial venules. Separately, T cells also can enter through afferent lymphatic vessels that drain peripheral tissues. Regardless of how they enter, T cells exit lymph nodes via the efferent lymphatics, which ultimately return their content to the blood by means of the thoracic duct.
Elegant studies examining lymphocyte migration through intestinal lymph nodes led Cahill and colleagues to propose that the different recirculating pathways involved distinct T cell populations (6, 7). They argued that T cells in the central blood-and-nodal pattern of migration were not the same as those that recirculated through peripheral tissues. It was ultimately demonstrated that blood-to-tissue recirculation primarily involved Ag-experienced or “memory” T cells, whereas naive T cells entered the lymph nodes directly from the blood via the high endothelial venule (8). In a landmark article, Sallusto et al. (9) added an extra layer to this concept by proposing that memory cells consisted of two subsets, each with a distinct pattern of recirculation through the body. They separated memory cells in human blood into two subsets that did or did not express CCR7. Although the demarcation was experimentally correlated with the ability or inability to exert immediate effector function, the major implication of this binary categorization was the distinct migratory potential for each of the memory subsets (10). Critically, the CCR7+ subset was expected to have a blood-and-lymph node pattern of recirculation, much like naive T cells. These were termed central memory T (TCM) cells. The CCR7− effector memory T (TEM) cell subset, in contrast, was proposed to consist of cells capable of recirculating through nonlymphoid tissues.
There is a surprising lack of evidence supporting the proposal that TEM cells dominate tissue recirculation. Although CCR7− memory cells can be found in afferent lymph, they are in the minority, and T cells emerging from the tissues do not appear enriched for this particular subset (11). Indeed, CCR7 was proposed to be necessary for T cell exit from peripheral compartments (11–13). In some circumstances, molecules other than the classical TEM cell/TCM cell discriminators CCR7 and CD62L are better indicators of peripheral infection control, such as the activation-linked markers CD27 and CD43 (14, 15). Regardless, the underlying concept that memory T cells preferentially recirculate through peripheral tissues remains unchallenged. Such a mechanism can be defined as recirculating peripheral immune surveillance, and it is a key contributor to systemic immunity.
Peripheral immunity and fixed immune surveillance
Around the time that the TEM cell/TCM cell paradigm was proposed, various groups described the dissemination of memory T cells to organs such as the gut, liver, lung, and skin (16–18). Because TEM cells were expected to recirculate through the periphery, and the tissue cells showed enhanced effector function (16), memory cells in the nonlymphoid organs were generally believed to belong to the TEM cell subset. Although many of these T cells were indeed recirculating, evidence began to emerge that at least some remained permanently in the periphery (19–21). In a classical study, Klonowski et al. (20) used parabiosis, a procedure that involves joining the blood supply of two animals, to examine memory T cell recirculation through peripheral tissues. Certain organs, such as lung and liver, appeared to permit T cell recirculation, whereas others, notably small intestine and brain, showed far more limited mixing with the blood. As a consequence, it was proposed that some tissues, but not others, restricted lymphocyte entry during the memory phase of the response. By extension, this also suggested that the memory cells within the restricted tissues formed a long-surviving resident population.
Further examination of tissue residence was carried out in experiments involving transplantation of T cell–replete tissue fragments (22–25). It was shown that CD8+ T cells in HSV-infected sensory ganglia could mount a recall proliferative response completely in situ (23), marking them as tissue resident with a limited ability to enter the circulation. Transplantation of intestinal and skin fragments demonstrated that a subset of the CD8+ T cells residing in these tissues could be retained long-term in isolation from the blood supply and that these were true memory cells with protective capabilities (24, 25). The memory T cells in these diverse nonlymphoid compartments expressed a common set of surface markers, with the combination of CD69 and CD103 being the most striking difference between the fixed tissue cells and memory cells found in the circulation (24, 26). As a consequence, it was proposed that these were a distinct memory population separate from the circulating TEM cells and TCM cells (24). These fixed tissue memory cells were simply called tissue-resident memory T (TRM) cells to fit in with the existing nomenclature.
Formal differentiation from the circulating populations was provided by microarray analysis of CD8+CD103+ TRM cells, which defined a core transcriptional signature that set TRM cells apart from the recirculating TEM and TCM cell populations (27, 28). The CD8+ TRM cells can be lodged in a highly localized fashion by infection or inflammation, forming a focused area of long-lived Ag-specific T cells (29, 30). It should be noted that TRM cells also show some level of disseminated lodgement, especially after repeated Ag exposure (31). But biases still remain, as demonstrated by Gaide et al. (32), who showed enhanced challenge responses at regions of skin subjected to repeated contact sensitization, mirroring the observations of increased immunity in skin having cleared prior infection (24, 31).
This form of enhanced local reactions in skin can be partly explained by the migration pattern for CD8+ TRM cells, which exhibit restricted lateral movement within the epidermis (29, 33, 34). Although these constraints limit T cell spread (34), there is enough mobility to permit effective scanning of the immediate environment for the presence of Ag (33). Overall, the emerging data showed that, in contrast to the memory T cells that are involved in circulating immune surveillance of peripheral tissues, TRM cells are spatially and anatomically restricted. As a consequence, their role in peripheral compartments could best be defined as contributing to fixed immune surveillance. This highly localized form of immunity contrasts with the systemic immunity that is typically associated with the circulating memory population.
TRM cell lodgement in diverse lymphoid and nonlymphoid tissues
At face value, the parabiosis data from Klonowski et al. (20) suggested that memory cell sequestration from the blood was tissue dependent, because some organs showed limited lymphocyte replacement, whereas others did not. It now appears that much of that recirculation can be attributed to blood contamination of the permissive organs (35). Residency is far more widespread than previously envisaged, and its basis is inherent to the TRM cells rather than the tissue in which they lodge (27, 36, 37). CD8+ TRM cells are found throughout the body, with their detection reported in lung, kidney, brain, and salivary gland (29, 38–40). In addition, some memory cells in primary and secondary lymphoid organs remain fixed and fail to equilibrate with the wider circulation (41–43). Single-cell TCR sequence examination of a range of tissues from organ donors showed that CCR7− memory T cells in human lymphoid organs are distinct from each other, defining them as noncirculating (42). In mice, parabiosis was used to show that a CD69+CD62L− subset of memory CD8+ T cells remains permanently lodged in the secondary lymphoid organs after lymphocytic choriomeningitis virus infection (43). These memory cells localized to the splenic marginal zones and red pulp, as well as the subcapsular regions of lymph nodes, positioning them at sites of Ag entry.
Differences between CD4 and CD8 TRM cells
Although the best-defined peripheral TRM cells belong to the CD8+ memory subset, CD4+ T cells that appear resident in nonlymphoid compartments also have been described. CD4+ memory cells are often the dominant population in peripheral tissues and, at least in human skin, most appear to have a resident phenotype (44, 45). Tissue-resident CD4+ T cells were formally described more than a decade ago in lung (19), although the wider implication of this observation was left unexplored. More recently, parabiosis studies showed that CD4+ T cells in the lung and female reproductive tract can remain in disequilibrium from the circulation (46, 47). In the case of the latter, residency was found to involve a complex interplay among macrophages, Ag recognition, and the concomitant production of IFN-γ and downstream chemokines that maintained all of the components in what was termed a “memory lymphocyte cluster.” This mechanism of CD4+ TRM cell retention was in sharp contrast to CD8+ TRM cells in this same tissue, because no such cluster formation or Ag persistence was required for CD8+ TRM cell retention (30, 48). Fundamental differences in microanatomical localization were noted for memory CD4+ and CD8+ T cells in mouse skin, where they segregated into respective dermal and epidermal regions (29). Similar separation was seen in human HSV infection on resolution of genital herpetic disease (49–51) and on experimental induction of psoriasis in human skin fragments (22). Although both CD4+ and CD8+ TRM cells clearly exist, there appear to be potential differences in terms of their localization, retention, and development.
The CD103-expressing TRM cell subset
CD103 is found on at least a subset of CD8+ TRM cells in a range of organs, including gut, kidney, brain, skin, female reproductive tract, and thymus (24, 38, 40, 41, 52). CD103 is the α-chain of the integrin αEβ7, which binds E-cadherin expressed on epithelial cells in skin and gut (27, 39, 53–55). E-cadherin is also expressed by TRM cells in a range of organs (28, 40). CD103 functions in the survival or retention of TRM cells once they have entered the tissues (27, 38, 40), although there is some suggestion that it may also act in tissue localization (56, 57). In the small intestine, it was argued that CD103 is not required for tissue entry after lymphocytic choriomeningitis virus infection, but it is essential for survival or retention of CD8 TRM cells in the epithelium (38). In skin, CD8+ TRM cell precursors upregulate CD103 after they reach the epithelium, where its expression promotes the survival of the intraepithelial memory cells (27).
Despite the strong association of this marker with the TRM cell subset, there is mounting evidence that not all TRM cells express CD103. The CD4+ TRM cells found in the leukocyte aggregates in the female reproductive tract lack CD103 expression (47), as do CD8+ TRM cells in secondary lymphoid organs (43) and in clusters within the gut wall after Yersinia pseudotuberculosis infection (58). Strikingly, the scattered CD8+ TRM cells found between the latter clusters are predominantly CD103+, as are the dispersed intraepithelial CD8+ TRM cells in the small intestine, skin, and female reproductive tract (24, 29, 48, 56). There are clear differences between CD103− and CD103+ CD8+ TRM cells that extend beyond the expression of this one molecule, with the latter showing poor proliferative potential on adoptive transfer (40). Moreover, CD103+ TRM cells have much lower expression of the key tissue egress molecule S1PR1, as well as the upstream transcription factor KLF2 (28, 58). Thus, some CD103− tissue T cells may be slowly recirculating TEM-like cells or precursors transitioning to a CD103+ mature form. However, CD103− T cells can also persist in tissues for prolonged periods (43, 58), meaning that these are bone fide TRM cells. For CD4+ TRM cells, it would appear that CD103 is not an absolute marker for residency, because at least some tissue-egressing CD4+ T cells express this molecule (13). However, of the CD8+ T cells found in peripheral tissues, those with surface CD103 proved to be resident or heavily implicated as possessing this characteristic (24, 25, 38, 59). It is reasonable to assert that CD103 identifies those CD8+ memory T cells in nonlymphoid tissues that are TRM cells. However, not all TRM cells in the body necessarily express CD103.
The existence of human TRM cells
There is considerable evidence for TRM cells in human tissues, with these cells being generated either by infection or as a consequence of inflammatory disease (60). One of the earliest experiments implying the existence of TRM cells involved T cell–containing human skin fragments from psoriasis patients. Boyman et al. (22) showed that prepsoriatic skin transplanted onto immune-deficient mice developed disease in a tissue-retained T cell–dependent manner. TRM-like cells have been proposed to exist upon resolution of skin infection with HSV (49, 50), and the presence of these cells correlates with enhanced virus control (61). TRM cells also were found in human lung (59), with influenza-specific CD8+ T cells localizing to the CD103+ intraepithelial lymphocyte fraction (62). Fixed drug eruption in skin also was linked with intraepithelial CD8+CD103+ T cells (63). This allergic response manifests as cutaneous lesions at fixed regions of the body, which appear only on drug ingestion. The remitting and recurring nature of this condition at fixed areas of the body strongly implicates TRM cell involvement. Probably the most definitive proof for human TRM cells came from compelling studies by Clark and colleagues (45, 64) on patients with cutaneous T-cell lymphoma. These individuals were treated with low doses of anti-CD52 that eliminates T cells via Ab-dependent cytotoxicity, a mechanism that operates in the blood but not peripheral tissues. As a consequence, this approach removes all circulating T cells from the body while sparing T cells in the skin. Critically, systemic Ab treatment of patients left a surviving population of T cells in the skin, defining them as skin TRM cells. Interestingly, skin CD62L+CCR7+ T cells also were eliminated, implicating these as a likely circulating TCM cell. Overall, although the lymphocyte content of human tissues, especially skin, can differ dramatically from what is found in the mouse (65), there is clear evidence for the common existence of TRM cells, often with overlapping phenotypes.
TRM precursor cells and factors that control their development and survival in peripheral tissues
CD8+ TRM cells develop from common memory cell precursors (27, 32) that are present early postinfection and then quickly disappear from the circulation (25). Circulating memory cell precursors can be identified by their expression of the α-chain of IL-7R and their lack of expression of the differentiation marker KLRG-1 (66–68). They give rise to long-lived circulating memory cells, typically belonging to the TCM cell subset. Akin to long-lived TCM cells, TRM precursor cells also are derived from the KLRG-1− effector-like population (27, 57). In detailed experiments that tracked TRM cell development in skin, it was shown that memory cell conversion to a mature TRM cell phenotype occurred after the precursor cells migrated into the outer epidermis, which is their ultimate site of persistence (27). This maturation resulted in the upregulation of the prototypic TRM cell surface markers CD69 and CD103, as well as the acquisition of a unique transcriptional profile that defines this memory subset. The maturation included shutdown of molecules associated with tissue egress. Skon et al. (37) showed that TRM cell maturation resulted in the downregulation of the transcription factor KLF2, which otherwise drove the expression of tissue-egress molecules CCR7 and S1PR1. Shutdown of tissue egress is critical to TRM cell development because forced expression of S1PR1 inhibits TRM cell formation (37), whereas loss of CCR7 has the opposite effect (27). In a related manner, elimination of CD69, which functionally counteracts S1PR1, also limits CD8+ TRM cell development (69).
One key factor that has repeatedly been linked to TRM cell maturation is TGF-β. This cytokine is a driver of CD103 upregulation (70, 71), so its involvement is likely confined to the CD103+ TRM cell subset. CD8+ T cells unable to signal through the TGF-βR pathway fail to form mature CD69+CD103+ TRM cells in a variety of tissues, including the intestine, skin, and lung (27, 36, 72, 73). In lung, TGF-β–driven TRM cell maturation involves a Smad4-independent signaling pathway (73) and is influenced by proximity to the airways, the likely site of TGF-β production (36). Other soluble factors implicated in TRM cell development include TNF-α and IL-33, which combine with TGF-β to downregulate KLF2 transcription (37, 38). In addition, TRM cell development in skin is IL-15 dependent (27) and influenced by unknown ligands for the aryl hydrocarbon receptor (34).
Finally, Ag plays an intriguing and variable role in TRM cell formation. Ag is not required for the development of some TRM cells, especially intraepithelial TRM cells in skin, female reproductive tract, and intestine (30, 38, 48). In contrast, for CD8+ memory cells in brain, sensory ganglia, and, possibly, lung, Ag recognition is mandatory for full maturation to the CD103+ TRM cell phenotype (30, 40, 74, 75). It is unclear what factors determine the Ag requirement, although persisting presentation is probably not necessary to maintain the TRM cell phenotype (40). However, even when Ag is needed for TRM cell formation, as it is for CD8+CD103+ memory cells in the sensory ganglia, the TGF-β dependence remains unchanged (L.K. Mackay and F.R. Carbone, unpublished observations).
Mechanistic analysis of TRM cell function
TRM cells lodged in peripheral tissues are broadly functional (40, 76) and capable of protecting against infection (24, 39, 48, 57). Indeed, in cases in which comparison between TRM cells and circulating memory cells was possible, TRM cell involvement proved pivotal in infection control (30, 31, 77). It is implicit that enhanced protection, at least at body surfaces, is directly linked to the immediate proximity of TRM cells to the site of pathogen entry. Presumably, this occurs, in part, by circumventing the time delay associated with recruitment of memory cells from the circulation. However, TRM cells are also capable of recruiting other immune components to sites of infection. Masopust and colleagues (77, 78) showed that CD8+ TRM cell–derived IFN-γ indirectly promotes recruitment of circulating memory T cells and B cells to sites of infection. Although this takes time, enhanced leukocyte recruitment accelerates pathogen clearance. In addition, surface-lodged TRM cells promote NK cell and dendritic cell activation on challenge, which would have more immediate effects (78). Impressively, Ariotti et al. (79) showed that Ag stimulation of skin-embedded TRM cells resulted in rapid, broad upregulation of innate components in the skin, including the almost immediate expression of the IFN-induced antiviral product IFITM3. TRM cell recruitment of innate components was shown to have the added advantage of broadening protection, resulting in bystander control of unrelated pathogens (78, 79). Thus, although activation of TRM cells is Ag specific, the protection resulting from their activation does not need to be.
TRM cells are specialized for the control of sequestered, persisting infection
It is easy to appreciate the advantages of systemic immune surveillance, a phenomenon that underpins modern vaccination. For example, inoculation of a small patch of skin with vaccinia virus renders an individual protected against subsequent smallpox infection via the respiratory tract. However fixed immune surveillance appears counterintuitive, especially within the context of protection against de novo infection. The paradox is starkly highlighted with one of the more striking examples of fixed immune surveillance involving the lodgement of CD8+ TRM cells within a few millimeters of the site of prior infection with HSV (29, 34). At first glance, it is hard to comprehend the advantage of such extremely focused immune protection, when all of the skin and a range of mucosal surfaces are potential sites of de novo infection with this virus. However, the benefits are obvious; the TRM cells are lodged at exactly the surfaces where HSV periodically emerges during bouts of recrudescent disease. To underscore this, Corey and colleagues (49, 50, 61) showed that the presence of CD8+ T cells in the skin directly correlates with attenuation of disease severity and virus replication during bouts of reactivation. In addition, HSV-specific CD8+ TRM cells can control reactivation in the sensory ganglia and maintain virus in a quiescent or latent state (80–83). Indeed, CD8+ TRM cells in ganglia can undergo local proliferative responses on repeated rounds of HSV reactivation (76), despite being subjected to chronic stimulation (81, 84).
The preceding section highlights the clear advantage of fixed immune surveillance and hints at potential TRM cell specialization. Specifically, TRM cells offer superior functionality in settings of persisting infection, especially during periods of latency when inflammation and recruitment from the blood has been extinguished. Such dissociation from the recirculating memory pool is seen during latent HSV infection (85), which is also likely to feature Ag expression that is too low for the activation and recruitment of circulating T cells (86). In addition, this specialization probably extends to the final stages of cleared infection, when pathogen load has declined to otherwise undetectable levels (40, 58). The combination of local immune surveillance in low-inflammatory settings brings into play other scenarios, such as infection with commensal microorganisms. Naik et al. (87) showed that skin infection with selected strains of the commensal bacteria Staphylococcus epididymis resulted in accumulation of IL-17–producing, intraepithelial CD8+CD103+ T cells in the absence of overt inflammation. Although the experiments did not formally identify these as CD8+ TRM cells, their intraepithelial localization and CD103 expression heavily implied that they were indeed nonmigratory memory cells sequestered from the circulation. Therefore, these CD8+ T cells would act in local immune control of what is, in effect, a silent ongoing infection. In this way, they behave much like CD8+CD103+ TRM cells in sensory ganglia that maintain a state of latency in the case of HSV infection. Strikingly, it was shown that CD8+ T cells in S. epididymis infection promoted nonspecific innate immunity in colonized skin that protected against bystander infection, much like Ariotti et al. (79) found for TRM cells lodged by local vaccination. Thus, by interacting with the surface microbiota, TRM cells embedded in epithelial surfaces may maintain the immune integrity of barrier surfaces.
Adaptive immunity at body surfaces is often described in terms of recirculating lymphocytes involved in surveillance of peripheral tissues against possible infection. Fixed immune surveillance and TRM cells need to be viewed in a different context. In certain situations, such as TRM cell control of latency, the term “memory” seems inappropriate, because the body is dealing with a current, rather than a past, infection. Yet, like a true memory population, TRM cells can persist indefinitely after infection has resolved and provide enhanced protection on subsequent pathogen encounter (24, 30). Another obvious contrast from the traditional view of immunity is the highly focused pattern of lodgement and protection. This sometimes extreme anatomical restriction immediately raises the question of whether TRM cells can provide broader immunity, as would be achieved with recirculating immune surveillance. However, TRM precursor cells show some level of dissemination, especially after repeated Ag encounter (31). Moreover, in certain cases, TRM cells can be lodged in such a way as to provide complete organ protection while maintaining anatomical restrictions. This was seen for the female reproductive tract, in which TRM cell lodgement by either localized inflammation or exogenous chemokine administration protected against vaginal HSV challenge (30, 48) and controlled surface replication, thus limiting downstream latent infection (30). Although these are encouraging results, TRM cells remain relatively unexplored in comparison with their circulating counterparts, as does the contribution of fixed immune surveillance to overall immunity.
I thank Bill Heath, Laura Mackay, Thomas Gebhardt, and Scott Mueller for comments and suggestions.
The author has no financial conflicts of interest.