Monocyte-Derived Cells in Tissue-Resident Memory T Cell Formation

Tissue-resident memory T cells (Trm) are a recently identified subset of memory T cells capable of stably residing in lymphoid and nonlymphoid organs without accessing the blood (1–4). Trm can be detected in various nonlymphoid tissues such as skin, female reproductive tract (FRT), small intestine, lung, salivary gland, brain, and liver as well as in secondary lymphoid organs (5, 6). Trm are highly protective against different viral, bacterial, and fungal infections (1, 7–11, 12, 13), and in the control of tumors (14–17). Their localization within peripheral tissues in which infections frequently occur confers an advantage in responding rapidly to invading pathogens. Upon re-encounter with Ag, Trm quickly produce proinflammatory cytokines such as IFN-γ, which not only has direct antiviral properties but also facilitates the recruitment of circulating memory T cells, B cells, and innate leukocytes to the site of infection (18–21). Trm from several tissues express high levels of granzyme B and are therefore poised to rapidly eliminate infected cells (12, 22, 23).

The development of Trm involves the integration of multiple signals from the tissue microenvironment, leading to the upregulation of Trm-specific gene programs and the suppression of genes involved in tissue egress (4, 24–27). CD8⁺ Trm typically express CD69 with or without the integrin CD103 (22, 28), whereas CD4⁺ Trm often express CD69 and the integrin CD11a (7). CD69 contributes to tissue retention by blocking responsiveness to sphingosine-1-phosphate (S1P) gradients through downregulation of the S1P receptor 1 (29, 30). TGF-β promotes tissue residency by inducing CD103 expression through downregulation of T-bet (31, 32). CD103 on CD8⁺ Trm binds to E-cadherin expressed by epithelial cells, allowing the retention of Trm within the epithelium (22, 28, 33, 34). Evidence that CD69 and CD103 contribute to tissue retention of Trm comes from studies showing that CD69 or CD103 deficiency results in defective Trm development or accumulation (28). The gold standards to confirm tissue residency are parabiosis and tissue transplant experiments (1, 7, 35). Additionally, the labeling of vascular cells by i.v. infusion of T cell reactive Abs, along with staining for CD69 ± CD103 for CD8⁺ T cells, is widely used to identify cells that are tissue resident at a particular point in time (7, 36).

A key question in the field is the nature of the cells that produce the signals for Trm localization and formation. Monocytes, dendritic cells (DCs), and macrophages play important roles in regulating T cell fate by presenting Ags, secreting cytokines, and providing costimulatory signals. Recent fate-mapping studies and high-dimensional single-cell analyses have given fresh insight into the ontogenic diversity of these immune cells, resulting in a growing interest in understanding the distinct function of monocyte-derived APC (MoAPC) as opposed to conventional DCs (cDC) or embryonic-derived macrophages in health and disease contexts [for review, see (37–39)]. In this review, we will discuss recent evidence that monocytes and MoAPC critically contribute to Trm formation in the context of infection.

Monocytes are mononuclear leukocytes that constitute, along with cDC and macrophages, the mononuclear phagocyte system. One of the most important new insights that has emerged over the last few years has been that monocytes are not the definitive precursors of most steady-state DCs and macrophages. Indeed, we now appreciate that most, but not all, tissue-resident macrophages in healthy lung, heart, liver, brain, and skin are in fact derived from embryonic precursors [reviewed in Ref. 37]. In contrast, cDC are derived from a common DC precursor in the bone marrow (BM) that is distinct from the common monocyte precursor [reviewed in Ref. 38, 39]. Nonetheless, a proportion of DC and macrophages in tissues at steady-state are MoAPC. Importantly, under inflammatory conditions, a large number of circulating Ly6C^hi monocytes infiltrate tissues where they differentiate in situ in response to local signals (40). MoAPC that arise under inflammatory conditions are often referred to as inflammatory APC (infAPC). In mice, infAPC can be identified by their high expression of CD11b, F4/80, and the Fc receptors CD64 and FcεRI. In contrast, cDC lack CD64 and FcεRI and embryonically derived tissue macrophages lack FcεRI (41). Although FcεR1 is well known for its role in the IgE response of mast cells, its cross-linking on DC and monocytes has been shown to regulate both pro- and anti-inflammatory cytokine production, depending on context (42). In humans, however, FcεR1 is also expressed on cDC2 (43), and thus this marker is not useful to define infAPC in humans. Mouse MoAPC are sometimes subdivided into monocyte-derived DC and monocyte-derived macrophages depending on whether they adopt cDC or macrophage properties, such as MHC class II (MHC II), CD11c, or CD64 expression and the ability to activate T cells or to mediate phagocytosis. Whether this diversity reflects heterogeneity or plasticity is unclear (38). Although MoAPC can adopt cDC and macrophage properties, there is increasing evidence indicating that MoAPC play nonredundant roles with those of cDC and tissue macrophages during infection, cancer, and many other pathological conditions (44).

Although cDC capable of cross-presentation are critical for the priming of CD8⁺ Trm during viral infection (45), there is accumulating evidence for contributions of MoAPC in the positioning, formation, or maintenance of Trm. In the steady-state, there are minimal numbers of MoAPC in the tissues; however, their rapid accumulation within the first few days of infection (40) in tissues where Trm differentiation takes place, puts them in an ideal location to influence Trm formation. The egress of monocyte precursors from the BM into the circulation and their entry into the tissues requires expression of the chemokine receptor CCR2 (40). Thus, CCR2-deficient mice or mice expressing the human diphtheria toxin receptor (DTR) under control of the CCR2 promoter have been used to eliminate monocytes and MoAPC to study their contribution to Trm formation.

In a study of respiratory poxvirus infection in mice, MoAPC accumulated in the lung, peaking at day 4–8 postinfection (p.i.), and persisted at lower levels through day 32 p.i., dependent on CCR2. In contrast, other lung immune subsets, such as neutrophils and cDC, were unaffected by CCR2 deficiency (46). Using adoptive transfer of wild-type CD8⁺ TCR–transgenic cells into CCR2-deficient mice, Desai et al. (46) showed that the CCR2-dependent MoAPC were dispensable for Trm formation out to day 20 p.i. but controlled Trm numbers by day 50, suggesting a role for inflammatory MoAPC in the maintenance of vaccinia-specific CD8⁺ Trm.

During respiratory influenza infection, pulmonary Ag encounter is necessary for the establishment of Trm (47). Interestingly, depletion of DCs after initial priming did not affect the number of NP-specific T cells in the lung, whereas inhibition of monocyte recruitment to the lung, using CCR2^−/− mice, significantly reduced the number of lung extravascular CD8⁺ T cells and Trm (48). Thus, MoAPC are important late in the primary response for Ag presentation and lung Trm formation (46, 48).

CD69⁺CD103⁺CD8⁺ Trm largely reside in epithelial layers of skin, respiratory tract, and other mucosal tissues, whereas CD69⁺CD103⁻ CD8⁺ Trm are usually found within the tissue parenchyma or within the lamina propria (6, 20, 49–51). Following respiratory influenza infection, CD8⁺ Trm cells localize within specific niches in contrast to circulating effector memory T cells, which are more sparsely distributed (52, 53). Of interest, the niche for CD69-independent retention of influenza-specific CD8⁺ Trm in the mouse lung was found to be located at sites of tissue regeneration after injury, which the authors termed repair-associated memory depots (53). To our knowledge, the APC in the CD8 Trm niche have not been well characterized.

CCR2-dependent MoAPC are found in association with CD4⁺ Trm in the skin following HSV-1 infection (54). CD4⁺ Trm are typically found within clusters of cells below the epithelial layers (20, 55). These clusters, referred to as memory lymphocyte clusters (20), have been identified in many tissues (including the FRT, skin, lung, and intestine) and are distinct from tertiary lymphoid structures or nasal associated lymphoid tissues, as they lack B cells and lymphatics (6, 20, 52, 55). As will be discussed in the next section, CD11b⁺ macrophages are important in the maintenance of these memory lymphocyte clusters in the FRT (20), although whether these are CCR2-dependent MoAPC remains to be determined.

Several studies have demonstrated a role for MoAPC in providing chemokines or cytokines required for the appropriate localization or retention of Trm. For example, oral infection of mice with the Gram-negative bacterium Yersinia pseudotuberculosis gives rise to two subsets of CD69⁺CD8⁺ Trm in the intestine, the TGF-β–dependent CD103-positive subset and the TGF-β–independent CD103-negative subset (50), each with unique gene expression profiles and distinct localization (56). The CD103⁺CD8⁺ population is scattered throughout the tissue, whereas the CD8⁺CD103⁻CD69⁺ Trm subset differentiates in immune cell clusters in the areas of bacterial infection. The persistence of the CD8⁺CD103⁻ Trm subset is dependent on IFN-β and IL-12, which is produced by inflammatory MoAPC recruited to the intestine during Y. pseudotuberculosis infection (50). To demonstrate the requirement for MoAPC in IL-12 production for CD103⁻ CD8⁺ Trm formation, Bergsbaken et al. (56) generated mixed radiation chimeras using BM from IL-12p35–deficient and CCR2-DTR mice, allowing deletion of IL-12 only from CCR2-positive cells. The results showed that IL-12 from CCR2-dependent cells drives the differentiation of the CD103⁻CD8⁺ Trm population. The importance of the CD103⁻CD8⁺ Trm is inferred from their location at the site of infection and from the finding that mice lacking CXCR3 lack the CD103⁻ Trm and fail to control Y. pseudotuberculosis (50).

CD4 T cells are important in control of HSV-2 in the FRT (20). During HSV-2 infection, CD4⁺ and CD8⁺ T cells (including those with a Trm phenotype) are found in memory lymphocyte clusters in the FRT along with macrophages and DCs. The CD4⁺ T cells produce IFN-γ in response to Ag stimulation (20). In turn, IFN-γ induces CCL5 and CXCL9, largely from CD11b⁺ macrophages. The depletion of these macrophages by diphtheria toxin treatment of CD11b-DTR mice demonstrated the importance of the CD11b⁺ macrophages in maintaining the memory lymphocyte clusters and providing CCL5 for CD4⁺ and CD8⁺ Trm retention. Thus, this study defined a feedback loop whereby IFN-γ produced locally by CD4⁺ Trm in the memory lymphocyte clusters induced CCL5 to retain the CD4⁺ Trm (20). Whether these tissue-resident macrophages are derived from recruited inflammatory monocytes is not clear from the markers used, but the finding that the number of CCL5⁺ macrophages increased after immunization is not inconsistent with this hypothesis. Moreover, in the skin, CD4⁺ T cells are also recruited by CCL5 produced by CD11b⁺ cells (55). Following HSV-1 infection of the skin, a mixture of resident and circulating CD4⁺ memory cells persist around the hair follicles (55), together with a population of DC and macrophages. The DC and macrophages remain elevated around the site of resolved infection in wild-type mice but were substantially reduced in CCR2-deficient mice, indicating their monocyte origin (54). Taken together, these studies suggest that MoAPC are important in CCL5 release to retain circulating and resident CD4⁺ T cells. As these cells are retained in the skin after resolution of infection (54), this raises the question of how MoAPC contribute to the long-term maintenance of Trm.

In the context of a vaccine strategy, i.v. immunization of rhesus macaques with peptide Ag and an adjuvant combining an agonistic anti-CD40 Ab plus the TLR3 agonist polyinosinic:polycytidylic acid (Poly IC) led to high levels of CD103⁺CD8⁺ Trm cells in the lung (57). The accumulation of the lung Trm correlated with early IL-10 production by blood monocytes, whereas other blood immune subsets did not produce significant amounts of IL-10 in response to these stimuli. Treatment of mice with the same immunization regimen led to similar results. Interestingly, s.c. immunization was less effective in inducing the systemic production of IL-10 by monocytes and was less effective in inducing a CD103^hi population of Trm in the lung. The authors identified an autocrine signaling loop whereby anti-CD40 and Poly IC treatment of human or rhesus monocytes, purified from blood but not skin, responded by secreting IL-10, which in turn induced TGF-β production by the monocytes, leading to CD103 expression on naive T cells (57). Interestingly, this study suggested that circulating monocytes were important in inducing CD103 expression on naive T cells ex vivo; however, it remains to be determined whether the key role of these cells in vivo is in circulation or once they enter the tissue.

Taken together, the studies mentioned above point to the importance of CCR2-dependent MoAPC in circulation as well as in tissues, such as the skin, intestine, lung and potentially the FRT, in providing key signals for localization, differentiation, and survival of Trm (Fig. 1).

The size of the T cell memory pool is widely thought to be dependent on the size of the effector precursor pool (58). Evidence that effector T cells give rise to Trm comes from fate-mapping studies, which show that cells that previously expressed the effector T cell marker KLRG1 can give rise to all types of memory T cells, including Trm (59). As TNFR family members are well known to provide signals that sustain effector cell survival (60, 61), it is perhaps not surprising that several TNF/TNFR family members contribute to Trm formation in the tissues during viral infection (62–65). In this section, we first discuss the evidence that MoAPC are the major cell type that express TNF family ligands during viral infection and then we discuss the role of this signal, which we call signal 4, in Trm formation.

During infection with lymphocytic choriomeningitis virus clone (LCMV) 13, inflammatory MoAPC, characterized by expression of F4/80, CD64, and FcεR1, were the main cell types expressing the TNF family ligands GITRL, 4-1BBL, OX40L, and CD70 in the spleen, LN, liver, and lung, with similar expression on the CD11b^hi MERTK⁺ monocyte-derived macrophage subset and the CD11c^hi MERTK⁻ monocyte-derived DC subset. As both MoAPC types could acquire Ag in vivo and present it to naive CD4⁺ T cells ex vivo, they are referred to as infAPC. In contrast to the infAPC, classical CD11c^hi MHC II^hi DC expressed minimal levels of these 4 TNF family proteins on their surface throughout LCMV13 infection and plasmacytoid DC (pDC) also were negligible for surface expression of these ligands (66, 67). The infAPC were also distinct from cDC in lacking the signature transcription factor Zbtb46 and in their higher expression of CCR2 compared with cDC (66). The induction of TNF family ligands on infAPC was dependent on type I IFN signaling both in vitro and in vivo (66). As type I IFN induces a different gene signature in cDC, pDC, and infAPC, it appears that cell type–specific effects of type I IFN are responsible for this differential regulation of TNF family ligands on APC subsets (66). The preferential induction of 4-1BBL and GITRL on infAPC over cDC and pDC was also observed in the lung and draining LN (dLN) during influenza A virus infection (62, 66). An earlier study had also reported that CD70 expression during influenza infection is limited to the CD11b⁺ migratory DC, although whether these were MoAPC or cDC was not clear from the markers used (68).

Splenic infAPC isolated at day 2 after LCMV13 infection have ∼10-fold lower MHC II than cDC and are inferior to cDC in CD4⁺ T cell priming (66). Moreover, TNFR family members such as GITR and 4-1BB are upregulated by early TCR signaling. Thus, it seemed likely that TNF family signaling was a postpriming event. To test this hypothesis, Chang et al. (66) transferred a 1:1 mixture of GITR^+/+ and GITR^−/− TCR–transgenic LCMV-specific CD4⁺ T cells into B6 mice and monitored their upregulation of early activation markers and intracellular signaling intermediates following LCMV13 infection. By 12 h p.i., the majority of transferred CD4⁺ T cells in the spleen had upregulated CD69 and 4-1BB, suggesting that they had been exposed to Ag. However, at this time point, there was no difference in expression of activation markers between GITR^+/+ and GITR^−/− Ag–specific T cells within the same mice, arguing against a role for GITR in initial T cell activation. However, by 24 h and continuing for the 72 h of the experiment, GITR^+/+ T cells had higher levels of intracellular p-NF-κB p65 as well as higher levels of the downstream mammalian target of rapamycin (mTOR) target phospho-ribosomal protein S6 (p-S6), than GITR^−/− T cells within the same mice. These studies suggest that GITR signaling takes place postpriming, coincident with peak expression of GITRL on infAPC. Using mice in which exon 2 of GITRL (an 11 aa segment) was deleted, leading to a hypofunctional molecule, the authors showed that the GITR-dependent signals require GITRL on Lyz2-positive cells. As neutrophils do not express TNF family ligands, the use of the Lyz2-positive cells suggests that MoAPC are key for the postpriming GITRL signal. We refer to this TNF-dependent postpriming signal as signal 4 (Fig. 2A). GITR-dependent signal 4 in T cells led to modulation of cytokine and TNF receptors associated with cell survival including upregulation of CD25, CD127, and OX40, as well as downregulation of herpesvirus entry mediator (66), a ligand for the inhibitory receptor BTLA (69).

GITR is also intrinsically important on both CD4⁺ and CD8⁺ effector T cells for their optimal accumulation and the subsequent accumulation of Trm in the lungs after influenza infection (62, 70). Moreover, GITR is important in rescuing low-affinity NP-specific effector T cells, an effect that requires GITRL on Lyz2-positive cells, consistent with a role for MoAPC in provision of the GITRL signal. As GITR on the CD4⁺ and CD8⁺ T cells had a similar effect on the numbers of effector T cells at day 10 and the numbers of Trm at day 30, the authors concluded that GITR likely impacts Trm formation by affecting the size of the Trm precursor pool. GITR had a greater effect on effector T cell accumulation in the lung than in the dLN, raising the possibility that GITR signaling in the lung tissue could be important. To test this idea, Chu et al. used mixed adoptive transfer of GITR^+/+ and GITR^−/− TCR–transgenic CD4⁺ or CD8⁺ T cells and monitored intracellular signaling intermediates during influenza A virus infection. GITR^+/+ T cells within the same mouse had higher levels of p-S6 than GITR^−/− T cells in the lung tissue at day 5 p.i., the time of maximal accumulation of GITRL⁺ infAPC in the lung. GITR^+/+ cells also showed higher levels of p-S6 than GITR^−/− cells in the dLN at day 3 (62). This leads to the model that GITRL on MoAPC first provides survival signals to primed T cells in the dLN and then again when cells reach the lung tissue, providing local signals to allow effector T cells to survive and give rise to Trm (Fig. 2B).

Radiation chimeras, in which B6 mice were reconstituted with a 1:1 mix of BM from 4-1BB^+/+ and 4-1BB^−/− mice, have similarly shown that 4-1BB on CD8⁺ T cells is critical in CD8⁺ effector T cell accumulation and Trm formation following influenza A virus infection (63, 64). As 4-1BBL is mainly expressed on infAPC in both the dLN and lung after viral infection (66, 67), the assumption is that 4-1BBL on the infAPC is required for Trm formation, although this remains to be formally tested. Zhou et al. (63, 64) also showed that delivery of exogenous 4-1BBL could enhance Trm formation and greatly increase the duration of protection to influenza A virus. 4-1BB expression is higher on reactivated memory T cells as compared with naive T cells (71). Therefore, Zhou et al. delivered either influenza NP alone or influenza NP with 4-1BBL in a replication-defective adenovirus at the boost phase of a prime boost regimen. Inclusion of 4-1BBL in an NP-containing adenovirus vector, delivered 30 d following primary influenza A virus infection, dramatically increased the circulating effector pool as well as the population of Trm observed 7 mo later compared with a vector containing NP alone (64). The effects of 4-1BBL on increasing Trm formation were dependent on codelivery of Ag and costimulation intranasally rather than systemically, required 4-1BB and TRAF1 expression by the CD8⁺ T cells, and were inhibited by rapamycin, implicating mTOR signaling in Trm formation (63, 64).

Anti-OX40 Abs delivered systemically during vaccinia virus infection also greatly increased the number of memory T cells observed in the lung tissue a year later (72), demonstrating that OX40 plays an important role in generating long-lived memory CD8⁺ T cells in the lung (72). Although intravascular staining with an anti–T cell Ab was not used in this study, it is likely that at least a significant proportion of this long-lived vaccinia virus-specific T cell population is a tissue-resident population.

Taken together, the above studies suggest that OX40, 4-1BB, and GITR signaling on effector T cells intrinsically enhances the size of the long-lived tissue-resident memory population, likely through recognition of their ligands on infAPC, both in the LN and in the tissues. Although a role for CD70 in Trm formation has not been formally demonstrated, it is important in CD8⁺ T cell accumulation during viral infection and likely also contributes to Trm formation (68, 73).

Based on the above discussion, we define signal 4 in Fig. 2. TCR and CD28-dependent signals provide signal 1 and 2, cytokines provide signal 3. TNFRs on primed T cells bind TNF family ligands on MoAPC, which provides signal 4 for T cell accumulation. The justification for calling this signal 4 is that using GITRL as proof of principle, we showed that signal 4 takes place postpriming and on Mo-APC rather than the priming cDC (66). Signal 4 is particularly important for T cells to accumulate in the tissues such that they can give rise to Trm (62–64).

Human CD69⁺ Trm have been studied in healthy tissues of organ donors and in patients undergoing lung resection (74, 75). Human Trm show a core signature that is shared with mouse Trm, including transcripts for key adhesion molecules and chemokine receptors associated with Trm localization, expression of genes for inhibitory receptors, and downregulation of transcripts associated with tissue egress (74, 75). Evidence suggesting a role for 4-1BB in human tumor-associated Trm came from a study comparing the transcriptome of CD8⁺ T cells in the tumor, so-called tumor infiltrating lymphocytes (TIL), with those outside the tumor (non-TIL) in lung cancer patients (76). CD8⁺ TIL were enriched in cells with a Trm signature (high for CD103 and CXCR6, low in KLF2 and S1P receptor 1). Moreover, this Trm signature was associated with better survival. Of interest, these TILs also expressed higher levels of 4-1BB and PD1 than the activated non-TIL and were enriched in signaling pathways downstream of 4-1BB (76), including TRAF1, JNK, and p38, signals previously shown to be critical in 4-1BB signaling (77–79). Similarly, Hombrink et al. (75) found that Trm from the lungs of patients undergoing lung resection for cancer or transplantation for end-stage pulmonary disease had a strong NF-κB signature, including expression of TRAF1, a gene induced by TNFR signaling and also important for signaling downstream of 4-1BB (80, 81). Thus, it seems likely that human lung and lung tumor–associated Trm also use TNFRs such as 4-1BB for survival signaling during their formation or maintenance. Taken together, circumstantial evidence suggests that TNFRs such as 4-1BB may play a role in human Trm formation, although this remains to be formally demonstrated.

In this review, we have discussed the evidence that inflammatory MoAPC that are recruited to tissues during infection and inflammation play a key role in Trm formation. As discussed above, there is accumulating evidence that MoAPC play an important role in providing chemokines and cytokines that facilitate the localization or retention of Trm and their precursors such that they can receive key differentiation and survival signals locally (Fig. 1). The accumulation of MoAPC in tissues at sites of infection fits with a role for these cells in sustaining survival of effector T cells as they enter the tissues and differentiate into Trm. Type I IFN–dependent induction of TNF family ligands occurs almost exclusively on MoAPC during viral infection, suggesting that these cells are the key providers of TNF family ligands for Trm formation during viral infection, a signal we refer to as signal 4 (Fig. 2). Signal 4 has largely been defined using viral infection models and GITRL as proof of principle. Further work is required to look for the effect of deletion of TNF family ligands specifically from MoAPC on Trm formation in different contexts.

A more complete understanding of the specific features of MoAPC that drive Trm formation and maintenance will be important for optimizing vaccines and immunotherapies. To date, evidence for the role of MoAPC in Trm formation comes largely from infectious disease models. However, Trm are also associated with improved outcome in human cancer (76), and understanding the role of MoAPC in Trm formation in the tumor microenvironment (TME) is an important area for future investigation. Much of the research into myeloid cells in the TME has provided evidence that myeloid-derived suppressor cells support tumor growth and prevent control by the immune system. However, immunotherapies have the potential to redirect the function of MoAPC toward promotion of antitumor responses. For example, expression of PDL1 on tumor infiltrating MoAPC was suggested to account for the efficacy of PDL1 blockade (82, 83). Determining conditions that induce MoAPC with a phenotype that supports Trm formation and tumor control is an important therapeutic goal.

Key questions remain. With respect to signal 4, is Ag presentation required for TNF family signals to T cells? As TNFRs are generally induced on activated T cells by TCR signaling, it seems likely that MHC-dependent signals would go hand in hand with signal 4, but this remains to be tested. In this regard, it is of interest that a monocyte-derived DC2-like population capable of cross-presentation was recently identified in the TME (84). What about after Ag clearance? Do TNF family ligands contribute to long-term Trm maintenance? If yes, how are the receptors and ligands induced or maintained after the infection and inflammation subsides? Much less is known about MoAPC in the formation of human Trm. Single-cell transcriptomics suggests that the myeloid compartment of human and mouse is somewhat conserved between mouse and human in the lung, but less so in the blood (85). However, to date, the expression of TNF family ligands in humans has not been subjected to a detailed enough analysis to confirm whether MoAPC are key in TNF family costimulatory ligand expression in humans. As the TNF family ligands 4-1BBL, OX40L, CD70, and GITRL are not highly expressed, they also fail to show up in many of the single-cell transcriptomics analyses.

Questions also remain about what happens to the MoAPC after resolution of infection and their role in the niches supporting Trm formation and maintenance. Collins et al. (54) have documented the sustained accumulation of MoAPC in the skin after resolution of HSV-1 infection. Do MoAPC or their derivatives persist at other tissue sites after resolution of infection, how can they be identified, and what signals do they provide to the Trm p.i.? Fate-mapping studies could be useful to follow the MoAPC after resolution of infection. Trm are found in clusters in specific niches in the tissues. How do MoAPC contribute to the spatial organization of these niches, and do they show different gene signatures in different tissues and microenvironments? It will be of interest to determine if MoAPC are found in contact with human Trm in the tissues. With the advent of multiplex microscopy techniques, the identification of unique markers for MoAPC subsets from single-cell transcriptomics, and the accessibility of human tissues for analysis (86), it should be possible to address some of these questions in humans as well as in animal models.

The authors have no financial conflicts of interest.