Insights into phenotypic and functional CD8⁺ T_RM heterogeneity

Summary

Cytotoxic CD8⁺ T cells recognize and eliminate infected or cancerous cells. A subset of CD8⁺ memory T cells called tissue-resident memory T cells (T_RM) resides in peripheral tissues, monitors the periphery for pathogen invasion, and offers a rapid and potent first line of defense at potential sites of re-infection. T_RM cells are found in almost all tissues and are transcriptionally and epigenetically distinct from circulating memory populations, which shows their ability to acclimate to the tissue environment to allow for long-term survival. Recent work and the broader availability of single-cell profiling has highlighted T_RM heterogeneity among different tissues, as well as identified specialized subsets within individual tissues, that are time and infection dependent. T_RM cell phenotypic and transcriptional heterogeneity has implications for understanding T_RM function and longevity. This review aims to summarize and discuss the latest findings on CD8⁺ T_RM heterogeneity using single-cell molecular profiling and explore the potential implications for immune protection and the design of immune therapies.

Introduction

The immune system plays a crucial role in protecting the body from pathogens and cancerous cells. Upon antigen encounter, CD8⁺ T cells proliferate rapidly and differentiate into a heterogeneous population of short-lived terminal effector cells (TE) and memory-precursor cells (MP) that mediate pathogen clearance. Following the resolution of infection, the population of effector T cells contracts. However, a small proportion of CD8⁺ T cells persists and forms memory cells, which provide long-lived protection. CD8⁺ memory cells have classically been studied in the blood and lymphoid organs. These long-lived circulating CD8⁺ memory cells can be subdivided into central memory cells (T_CM) and effector memory cells (T_EM) based on their surface marker expression, trafficking patterns, and functional capabilities upon reinfection¹. While T_CM cells are predominately found in the blood and in secondary lymphoid tissues and are characterized by a high proliferative capacity, T_EM cells patrol the blood and can transiently enter tissues during acute infections.

However, during acute infection, a subset of CD8⁺ T cells migrates into tissues and gives rise to tissue-resident memory CD8⁺ T cells (T_RM)^2–4 that provide local protection. T_RM constitute a large portion of the CD8⁺ T cell memory population and can remain lodged in tissues for the lifetime of the organism, without recirculating^5,6. T_RM cells scan tissues for pathogens and offer a rapid and potent first line of defense due to their enhanced effector functions and proximity to sites of reinfection⁵.

Defining tissue-residency

The strongest evidence that T_RM reside in tissues without re-entering circulation and are a distinct memory T cell population arose from studies using parabiotic surgery in which the circulatory systems of two mice are joined. This procedure results in an equilibrium of circulatory T cells from the two animals. However, tissue-resident T cells did not reach homeostasis and remained tissue-specific in each mouse^2,7–10. Studies in humans have exploited human leukocyte antigen (HLA)-mismatched allografts that have different HLA alleles between the organ donor and recipient to longitudinally follow donor-derived T_RM cells in transplants. Using this approach, it has also been proven in humans that T_RM cells can reside in organs for extended periods of time without draining to the lymph nodes or entering circulation^11–13. Considering the experimental burden of performing parabiosis experiments in mice, many studies instead use an intravascular staining approach to mark T cells in the vasculature (thus inferred to be recirculating) with a fluorochrome-labelled antibody. T cells residing in tissues are not marked by the intravenous injection of the antibody, which has shown to faithfully reflect the residency of most antigen-specific T_RM observed via parabiosis¹⁴. This strategy efficiently labels cells in the vasculature, but cannot discriminate between cells that reside in the tissue permanently and cells that only transiently enter the tissue. Nonetheless, for the purposes of this review, tissue-residency will be inferred via the absence of intravenous labelling.

Common adaptations of memory T cells to tissue-residency

Although T_RM cells share some transcriptional features with T_CM and T_EM cells, they are transcriptionally and epigenetically distinct from circulating memory populations^15–18. A transcriptomic analysis of T_RM across many tissues compared to their recirculating counterparts allowed the definition of “core” tissue-residency and circulatory gene expression signatures, which have helped define the tissue residency state^15–17,19. Tissue residency is achieved by upregulation of adhesion and retention molecules and chemokine receptors (CD103, CD69, CCR9) and by reduced expression of lymphoid homing molecules (S1PR1, CCR7, CD62L) that help T cells to leave non-lymphoid tissues.

Multiple transcription factors that orchestrate the tissue-adaptation process have been identified. For example, CCR7, CD62L, and S1PR1 are controlled by the transcription factor KLF2, the expression of which is downregulated during T_RM differentiation³. S1PR1 binds to sphingosine-1-phosphate (S1P), which is highly abundant in the blood and promotes tissue egress. Downregulation of S1PR1 is, hence, a necessity for tissue-residency. CD69 is a c-type lectin protein that can form a complex with S1PR1 and inhibit binding of S1P to S1PR1, thus hindering egress from tissues of T_RM cells³. A recent study showed that downregulation of S1PR5, another receptor for S1P, is essential for T_RM formation in the skin^20,21, where S1PR5-deficiency leads to enhanced T_RM formation. Interestingly, S1PR5 does not interact with CD69²¹. In contrast to S1PR1, which is directly controlled by KLF2, S1PR5 is regulated via T-bet and Zeb2²⁰, demonstrating how tissue-residency is the result of multiple coordinated adaptation processes by T cells. Downregulation of S1P-receptors is accompanied by the downregulation of the T-box transcription factors T-bet and Eomes, which subsequently renders T_RM cells responsive to TGF-β^22,23. TGF-β signaling induces the expression of CD103, thus promoting T_RM cell retention, especially in epithelial tissues^24,25. Furthermore, transcription factors that facilitate expression of the tissue-residency signature, while suppressing circulatory-associated genes, include Blimp1¹⁵, Hobit¹⁵, and Runx3¹⁶. In summary, while there is not a single T_RM-specific “marker,” T_RM cells are commonly characterized by expression of two surface molecules: CD103 and CD69. However, the expression of these two canonical T_RM cell-surface proteins in tissues is neither uniform within a tissue, nor uniform across different tissue-resident T cells¹⁷.

However, T_RM cells do not constitute a homogenous memory T cell population. In fact, recent work has highlighted T_RM heterogeneity within tissues and among tissues with differences in surface marker expression, transcriptional changes, functionality, and longevity^17,26. T_RM cells have been identified in almost every human and murine tissue and infiltration and long-term maintenance of these cells in non-lymphoid tissues requires T cells to acclimate to the specific environments that may differ in a broad range of ways, including the availability of nutrients, metabolite composition, cytokine milieu, cell composition, and matrix proteins. It is therefore not surprising that, although T_RM cells share a common residency gene-expression signature, they also require tissue-specific acclimatization to persist in and patrol these unique and specialized environments. In addition to heterogeneity of T_RM cells observed across different tissues, heterogeneity is additionally apparent within a single tissue, where T_RM subsets resembling those observed in circulatory memory populations exist and change in their relative abundance over the course of an infection²⁷. Furthermore, distinct infections can result in formation of different T_RM populations; for example, T_RM cells in the intestine that persist after Yersinia pseudotuberculosis or intravenous Listeria monocytogenes infection show different expression of cell surface proteins when compared to T_RM cells that arise after acute lymphocytic choriomeningitis virus (LCMV) infection^28–30.

Single-cell RNA sequencing (scRNA-seq) is a powerful approach that can reveal heterogeneity within cell populations and has been used extensively in recent years to probe the dynamic gene expression patterns of a wide range of immune cell types in health and disease^1,31–33. scRNA-seq combined with reporter- and fate-mapping approaches has led to important advances in the understanding of T_RM intra- and inter-tissue heterogeneity with implications for T_RM cytotoxicity, function, longevity, and plasticity. In this review, we will summarize and discuss the latest findings on CD8⁺ T_RM heterogeneity using single-cell molecular profiling and highlight the different levels of T_RM heterogeneity as well as the potential implications for immune protection and the design of immune therapies.

Intertissue heterogeneity

Although most reports studying T_RM cells focus on epithelial tissues like the intestine or the skin, T_RM cells can be found in almost every organ. Besides the small intestine and the skin, T_RM cells have, for example, been described in the kidney^17,34, liver^12,15, salivary gland (SG)^17,35,36, adipose tissue¹⁷, pancreas², stomach², female reproductive tract⁶, lung^37,38, and colon^2,26. Although, T_RM cells in these organs vary in their durability⁶ – ranging from a half-life of 82 days in the uterus to no decay in the salivary gland – they lodge in these tissues for long periods of time, highlighting the necessity of recently migrating T cells to specifically acclimate to their new environment as they become long-lived T_RM³⁹. Much work has focused on understanding how T_RM cells become resident. Common adaptations include the downregulation of tissue egress molecules, such as S1PR1, and the upregulation of retention signals. However, considering that all organs fulfill distinct physiological roles, provide specialized tissue architectures, and have a unique composition of cells and structural tissue, it seems very likely that, in addition to common adaptations to tissue retention in general, T_RM cells also need to make tissue-specific adjustments in gene expression, metabolic state, and homeostatic dependencies. In addition to a lineage CD8⁺ T cell-defining signature and memory T cell signature, CD8⁺ T_RM cells need to acquire a “common” residency signature as well as tissue-specific adaptations that mediate long term survival and function in that environment (Figure 1).

Figure 1.

In addition to a shared lineage CD8⁺ T cell-defining signature and a memory T cell signature, T_RM cells need to acquire a transcriptional residency signature. Long term survival and function within the tissue are mediated by tissue-specific adaptations of T_RM cells that allow them to optimally acclimatize to their tissue of residence. These tissue-specific adaptations can be shared across multiple cell types.

Tissue-specific adaptations of other immune cell types that establish residency, such as tissue-resident macrophages, are well established and provide a basis for considering such changes for CD8⁺ T_RM^40–42. Specialized macrophages with distinct functions can be found in the spleen, skin (Langerhans cells), serosal tissues, lung, liver (Kupffer cells), gastro-intestinal tract, bones (osteoclasts), and the central nervous system (microglia)⁴⁰. These tissue-specific adaptations are, by necessity, not reflected in the common T_RM residency signatures observed when identifying the common adaptations of T_RM originating from many tissues^15,16. A careful comparison of gene expression across tissues in the same infection setting and the development of single-cell RNA sequencing (scRNA-seq) approaches has broadened our understanding of the differences in T_RM populations, allowing differentiation between ubiquitous tissue-specific changes to the T_RM cell population and changes in the abundance of shared, heterogeneous T_RM cell populations¹⁷. How these specific adaptations are regulated is not completely understood, but may be initiated by distinct signals from cytokines, chemokines, cell-cell interactions, or metabolites, and could be reflected by unique patterns of transcription factor expression and lead to distinct expression patterns of surface proteins.

Cytokines required for Trm differentiation

In accordance with different transcriptional acclimatization of T_RM cells to tissues, T_RM cells require distinct cytokines to both start and sustain their differentiation process into tissue-resident memory cells. For example, TGF-β, a pleiotropic cytokine, has been well established in its importance for T_RM formation in the skin, intraepithelial lymphocyte (IEL), and salivary gland, but it is not required for liver or kidney T_RM formation^{17,22,23,25,36,43}. Furthermore, it has been shown, that IL-15, a common gamma chain cytokine that promotes homeostatic proliferation and survival of memory cells^44,45, is required for T_RM survival in the skin^22,46, liver⁴⁷, salivary gland, and kidney⁴⁸ but is dispensable for T_RM homeostasis in the pancreas, female reproductive tract, and the small intestine in mice^48,49. However, even if some T_RM populations do not require IL-15 for maintenance, these cells still proliferate in response to IL-15, possibly to transiently amplify memory populations⁵⁰. Not only do the different requirements for T_RM development highlight the necessity for tissue-specific adaptation, but they might also indicate the existence of distinct T_RM subsets that are driven and maintained by distinct cytokines and cell interactions.

T_RM cells in different organs are dependent on distinct transcription factors

In line with the idea of tissue-specific adaptations of T_RM cells, the transcription factors required for T_RM differentiation differ among tissues. While some transcriptional adaptations (e.g., downregulation of KLF2) are shared among T_RM populations, differential transcription factor dependence has been observed. For example, the transcription factor Hobit is required for skin, but not lung T_RM generation¹⁵. Another example for tissue-specific transcription factors is the transcriptional repressor hypermethylated in cancer 1 (Hic1). Hic1 is induced during human iTreg differentiation⁵¹ and was shown to regulate homeostasis of intestinal lymphocyte populations in mice, thereby preventing the development of intestinal inflammation⁵². In LCMV infection, Hic1 expression by T cells is largely restricted to the small intestine (Figure 2), and knockdown of Hic1 specifically reduces T_RM cells in the small intestine, while other organs remain unaffected¹⁷. In line with that, overexpression of Hic1 leads to a specific accumulation of T_RM cells in the small intestine¹⁷. Mechanistically, this can partially be explained by the finding that Hic1 regulates expression of P2RX7¹⁷, an extracellular ATP receptor that enhances metabolic fitness⁵³ and contributes to small intestinal T_RM survival^53,54.

Figure 2.

Expression of T_RM molecules, lineage defining transcription factors, cytokines and integrins. Although T_RM share common adaptation signatures, expression of T_RM associated genes is heterogeneous across tissues and infections. Data from Crowl et al., 2022 (CD8⁺ P14 T cells isolated from spleen and blood (T_MEM) and IV⁻ CD8⁺ P14 T cells from IEL, kidney, liver, SG, and fat (T_RM) at day 30 post infection with LCMV Armstrong, T_MEM: all P14 cells); Mackay et al., 2016 (gp33 and np396 tetramer LCMV-specific CD8 T cell populations day 40 post LCMV Armstrong infection, and gBT-I cells 40 days post HSV infection); and Poon et al., 2023 (scRNA-seq of CD3⁺ T cells of two donors, aggregated to pseudo-bulk samples per donor). Mean expression values are plotted.

For generation of the heatmaps, data was downloaded from GEO (GSE182274, GSE70813, and GSE206507). For bulk RNA Seq data (GSE182274 and GSE70813), fastq files were processed using the nextflow pipeline (nf-core/rnaseq: 3.10.1⁹⁹). Variance stabilizing transformation (VST) was performed using DESeq2¹⁰⁰. Mean VST values for selected genes were plotted using the R package ComplexHeatmap. For GSE206507 raw counts were downloaded from GEO and pseudobulk samples were created using the aggregateAcrossCells function from scuttle¹⁰¹ in R with the groups tissue and donor. Pseudobulk samples were then processed in DESeq2 and visualized with ComplexHeatmap as described above.

In addition to tissue-specific transcription factors, the dependency of T_RM on T_RM-defining transcription factors varies among tissues. Deletion of Runx3 reduces T_RM cell numbers in multiple non-lymphoid organs, including in kidney and salivary gland, but its deletion leads to the strongest reduction of T_RM cells in the small intestinal epithelium and affects CD103⁺ CD69⁺ TRM cells more compared to CD103⁻ TRM cells¹⁶. Similarly, Blimp1 deletion impairs T_RM formation in the IEL and SG more than the kidney¹⁷. Hence, tissue-specificity is not only achieved by the expression of tissue-defining transcription factors, but also by modulation of the expression of T_RM-inducing transcription factors. It is likely that an interplay of both mechanisms leads to the optimal adaptation to tissues that T_RM require for persistence. Therefore, further understanding and identification of tissue-specific transcriptional networks will be the basis for engineering tissue-programmed T cells.

T_RM in different tissue sites are phenotypically distinct

Phenotypic comparisons of T_RM in distinct tissues reveal tissue-specific patterns of expression of cell-surface receptors by T_RM cells. While CD103 and CD69 expression, as well as loss of IL-18R expression, are commonly associated with and used for identification of T_RM cells, substantial differences in expression of these molecules can be observed among different T_RM populations. CD103 is an alpha integrin that, upon heterodimerization with a beta integrin, can bind to E-cadherin and facilitate retention in epithelial tissues. Thus, CD103 expression is mainly restricted to T_RM cells associated with epithelial barrier tissues, including skin, small intestine, and salivary gland^17,22,26,36. However, the integrin superfamily contains numerous subunits (with 18 alpha and 8 beta subunits)⁵⁵, and a recent study shows that expression of the beta integrins differs substantially among tissues (Figure 2). For example, there are multiple differences in alpha- and beta-integrins between the small and large intestine: αE integrin expression is higher in the small intestine compared to the colon, whereas expression of α4β1 is higher in T_RM cells from the colon²⁶. These findings indicate that tissue-retention modules are specifically adapted to distinct regions of the intestine. CD69, which antagonizes S1PR1 and thereby hinders tissue egress, is widely used to identify T_RM cells. However, the expression levels of CD69 in T_RM cells and the functional relevance of CD69 for formation or maintenance of T_RM cells varies vastly among tissues. In murine acute LCMV infection, CD69 expression is not needed for T_RM formation in the small intestine, but necessary for kidney T_RM cells, and forced expression of CD69 increases T_RM formation in the kidney⁵⁶. Downregulation of IL-18R is associated with the establishment of T_RM cells in the kidney³⁴. However, similar to CD103 and CD69, its expression varies among T_RM cells from different tissues: IL-18R expression is lost in IV-negative T_RM cells from the IEL after LCMV infection in mice but only partially downregulated in T_RM cells obtained from kidney and SG, and it is still expressed in visceral-adipose tissue and liver T_RM cells¹⁷. Ly6C expression in reduced in T_RM cells compared to T_CM cells^49,57,58, but the expression of Ly6C also greatly depends on the studied tissue^17,57. These findings demonstrate that the functional relevance of surface proteins commonly used for T_RM identification largely depends on the tissue.

T_RM cell in different tissues exhibit specific metabolic adaptations

Another level of specialization of T_RM cells to their given tissue environment is reflected in their usage of fatty acid binding proteins (FABP)⁵⁹ (Figure 2). T_RM cells rely on fatty acid uptake for their survival⁶⁰. However, the FABP family exists as a large family with many different isoforms that are expressed in a tissue-specific manner. Interestingly, T_RM in different tissues express different isoforms of FABP, highlighting their specific adaptation to the tissue microenvironment⁵⁹. Upon relocation, T_RM cells adapt their FABP expression profile to the new location. Thus, isoform usage is determined by tissue-derived factors, and T_RM cells actively and continuously sense their environment and adapt to it⁵⁹.

T_RM heterogeneity among organs in humans

Most studies of T_RM heterogeneity among tissues have used murine models. Further, most studies rely on the adoptive transfer of T cell receptor (TCR)-transgenic antigen-specific T cells, and thus do not assess the contribution of a polyclonal endogenous T cell response and the relevance of the TCR clonotypes to T_RM cells. A recent report, however, addresses these limitations and extends our knowledge of human tissue-resident T cells^61,62. In this study, lymphocytes from the lung, jejunum, abdominal skin, their draining lymph nodes (LN) (pulmonary LN, mesenteric LN, and inguinal LN), and from blood and spleen were collected and used for a comprehensive profiling with CyToF, TCR-sequencing, and single-cell RNA sequencing. In accordance with findings from the murine T_RM models, the authors found a unique composition of CD4⁺ and CD8⁺ T_RM cells in the assessed barrier tissues: T cells lodged in the skin had high expression of CCR4, CCR10 and CXCR4, and the Th2 lineage-defining transcription factor GATA3; T cells that reside in the intestine exhibited a high expression of integrins (ITGAE, ITGA1, ICAM1), and T cells isolated from the lung displayed a site-specific expression pattern of CTLA4, IL10, and PDCD1⁶¹ (Figure 2). In addition to location-specific expression patterns, the authors identified common adaptations among all barrier tissues and expression patterns that are shared among only some tissues. For example, T cells in the jejunum and lung were characterized by high expression of CXCR6 and Th17 signatures genes (CCL20, RORA, RORC, and IL17A)⁶¹, and T_RM cells in the lung and skin showed increased expression of genes encoding matrix- and adhesion-associated molecules as well as PRDM1 encoding for BLIMP1, a transcription factor associated with tissue-residency and effector functions. These transcriptomic changes may indicate the specific needs of T cells to maintain immunity against various types of pathogens in threatened tissues throughout human life. Importantly, many of these site-specific adaptations are by human and mouse T cells; for example, expression of ITGAE, CCR9 or HIC1 and downregulation of KLF2 in the intestine or high expression of AHR in the skin. These findings indicate that some of these transcriptional profiles are conserved across species (Figure 2). However, expression of some genes is not conserved; for example, T cells in the jejunum exhibit elevated expression of IL17A in humans, but not in mice. These differences might arise from differential exposure of humans to pathogens (in contrast to mice in a specific-pathogen free environment), but also might reflect variance due to polyclonal CD3⁺ T cells for human versus TCR transgenic CD8⁺ T cells for mice or additional differences. Considering an individual’s age and history of infection also adds a new layer of complexity to understanding T_RM phenotype, function, differentiation, and maintenance in humans. Insight into these topics is essential to improve targeted and tissue-adapted immune responses at barrier tissues in humans.

Clonal relationships between circulatory and tissue-resident memory T cells

How T cell clones that contribute to circulatory memory populations are proportional to tissue-resident T cell populations and whether or not the TCR clones found in T_RM population are unique to the T_RM populations or even the specific tissue are all topics of great interest. Using TCR sequencing, a recent report described that certain TCR clones expand more in barrier sites compared to circulatory memory populations, indicating that the TCR pool in circulation is not representative of that seen in tissues⁶¹ — an observation that has important implications in numerous areas, such as vaccine design and boosting strategies⁶³. This finding is corroborated by a report using clonal tracing of T cell clones that showed that graft versus host disease (GvHD) in a murine model is maintained by expansion of a tissue-residing TCF1⁺ subpopulation and not by recruitment of T cells from secondary lymphoid organs or the blood⁶⁴. Furthermore, in humans, TCR clones for CD4⁺ T_RM cells segregated between the barrier tissues with only minimal overlap between the skin, lung, and jejunum, where CD4⁺ T cell clones are less disseminated and more site-specific compared to CD8⁺ T cell clones⁶¹. It is possible that this clonal segregation is a result of different infections or pathogens that T cells encounter at these three distinct barrier sites, but the possibility that certain TCR clones preferentially give rise to T_RM cells in a specific organ cannot yet be ruled out. However, even within a naive T cell population that expresses the same TCR, a clonal population that possesses a heightened potential to form T_RM cells was shown to exist⁶⁵. In this study, barcode labelled naive OT-I T cells were transferred into recipient mice, and T_RM cells were induced by vaccination or herpes simplex virus (HSV) infection. While T cell clones contributed proportionally to circulatory and skin effector T cells, a disparity in T cell clone distribution was seen at memory timepoints⁶⁵, suggesting that distinct T cell clones do preferentially give rise to T_RM cells. Interestingly, there is evidence that these T cell clones might become preconditioned towards their T_RM fate even before entering the tissue^65,66.

Intratissue heterogeneity

The heterogeneity of cell surface receptor expression and overall gene expression indicated by scRNA-seq within a population of T_RM cells from a given organ varies greatly. For example, T_RM cells in the dermis are heterogeneous in their expression of CD69, with only approximately 70% of the dermal T_RM cells expressing CD69, whereas almost all T_RM cells in the epidermis express CD69²². In contrast, the majority of T_RM cells in the dermis do not express CD103, but the expression pattern of CD103 on epidermal T_RM is heterogeneous²². The heterogeneous expression of surface proteins of T_RM within a tissue is not limited to the skin and has also been observed in other tissues. T_RM in the salivary gland show a heterogeneous expression for both CD69 and CD103^17,36, and kidney T_RM are characterized by an intermediate expression of CD69 but almost no expression of CD103^17,34 (Figure 3a). Similarly, liver-resident T_RM exhibit low to intermediate expression of CD69 and no expression of CD103^17,36. Similar findings are observed in human studies—whereas most of the intestinal T cells are positive for CD103 and CD69, expression of these two tissue-residency promoting molecules is heterogeneously expressed by T cells resident in the lung or skin and also varies between CD8⁺ and CD8⁻ T cells (Figure 3b). While most small intestinal intraepithelial T_RM cells in mice express CD103 and CD69 after acute LCMV infection, only about 60% of small intestinal T_RM cells from the lamina propria, and approximately 30% of colonic T_RM cells, express CD103 and CD69²⁶. In addition to CD69 and CD103, other T_RM markers and transcription factors including Il18r, Cd49a, Ly6c, Tcf1, T-bet, and Eomes have been shown to exhibit variable expression within the tissue as well as among tissues^{17,23,26,34,36}. The expression levels of these canonical T_RM molecules are further subject to changes over time: e.g., expression of CD103 and CD69 in small intestinal T_RM cells increases from the effector phase to memory timepoints, whereas expression of IL-18R decreases. Hence, the time after infection adds a new dimension to T_RM heterogeneity. This idea is further supported by the finding that the importance of T_RM-driving factors can vary depending on the time after infection. For example, the inducible costimulator (ICOS) receptor is essential for T_RM establishment, but not maintenance⁶⁷, and the transcription factor EOMES is not required for formation of small intestinal T_RM, but is needed for their long-term maintenance²⁶. In summary, these findings support the hypothesis that different subtypes of T_RM exist and that the composition of the T_RM pool might vary over time.

Figure 3.

scRNA-seq reveals intertissue and intratissue heterogeneity. T_RM cells within a single tissue display varying expression of genes associated with effector and memory potential, showing that distinct subsets of T_RM cells exist. Scales for each gene are consistent across tissues to allow for comparison within and among tissues. Data from (a) Crowl et al. 2022 (scRNA-seq of murine IV⁻ CD8⁺ P14 T cells from IEL, kidney, SG, and liver at day 32 post infection with LCMV Armstrong) and (b) Poon et al., 2023 (scRNA-seq of human CD3⁺ T cells of two donors). Donor D551: Jejunum and Lung; Donor D492: Skin.

For visualizing the intra-tissue heterogeneity, each tissue dataset (downloaded from GEO GSE182275 and GSE206507) was normalized separately using sctransform with method = “glmGamPoi” in Seurat¹⁰². Dimensional reduction was performed using the RunPCA function in Seurat following by RunUMAP using the first 20 PCAs. In addition, data imputation was performed using MAGIC¹⁰³ using the default settings and the exact solver. Manually selected genes to highlight inter- and intratissue heterogeneity were plotted. Color scale was limited to the 98^th expression percentile.

Parallel subsets between circulatory memory and tissue-resident memory T cells

The circulatory memory T cell compartment is phenotypically and functionally heterogeneous, with effector and memory T cell populations consisting of phenotypic and functional subsets that evolve over time. Terminally differentiated, short-lived effector cells (CD127⁻KLRG1⁺) and memory-precursor cells (CD127⁺KLRG1⁻) are predominantly found during the effector phase of infection^68,69. These cells continue to differentiate over time and form circulatory memory T cells that can be broadly divided into T_CM and T_EM, which derive from memory precursor cells and t-T_EM cells that come from KLRG1^hi effector cells. Recent data show that intratissue heterogeneity of T_RM populations reflects T_RM subsets that functionally resemble those observed in circulation^27,31,70.

Two distinct subsets of T_RM cells that mirror those seen in circulatory memory populations have been identified²⁷. While small intestinal T cells showed an enrichment for effector gene-signatures early after an infection, gene-expression at later timepoints was enriched for memory signatures, highlighting a continuous differentiation of T_RM cells over time. Among the transcriptional regulators with differential expression over the course of infection were Blimp1 (Prdm1) and Id3²⁷. Id3 is an inhibitor of E protein transcription factors. It regulates long-lived circulatory memory formation in CD8⁺ T cells⁷¹, and is most highly expressed in small intestinal T_RM cells at late time points after infection. In contrast, Blimp1, which marks terminally-differentiated T cells, displayed an inverse expression pattern in studies, peaking at day 4 of infection and subsequently declining in small intestinal T_RM cells²⁷ and skin T_RM cells¹⁵. Using Blimp1 and Id3 reporter mice, two distinct subsets of small intestinal T_RM cells were observed: Blimp1^hiId3^lo T_RM cells expressed canonical effector genes (i.e., Cx3cr1, Zeb2, Klrg1, Gzma and Gzmb), and, in contrast, Id3^hiBlimp1^lo T_RM cells expressed genes reminiscent of memory-precursor cells (i.e., Bach2, Tcf7, and Cd27)²⁷. Furthermore, Id3^hi T_RM cells have shown elevated degranulation capacity and cytokine production of IFNγ, TNF, and IL-2 and yield a greater frequency of both circulating and resident secondary effector T cell populations after rechallenge compared to the Id3^lo subsets²⁷. Thus, Id3^hi small intestinal T_RM cells exhibit multifunctionality and memory potential, which is consistent with the greater proliferation capacity of Id3^hi circulatory memory T cells.

Analysis of T_RM differentiation in tissues has largely relied on bulk RNA sequencing and a small number of phenotypic markers, limiting the ability to discern functional differences and intermediate states of differentiation that might arise within the T_RM population^15,16. scRNA-seq overcomes this limitation and has identified clear functional and phenotypical subsets among T_RM cells within the same tissues and highlighted similarities to the memory differentiation of memory CD8⁺ T cells in blood and secondary lymphoid organs. A longitudinal scRNA-Seq study³¹ of the T cell response to the Armstrong strain of LCMV (LCMV Armstrong) analyzed small intestinal T_RM in comparison to circulatory memory cells and found a transcriptional memory signature that is used by both circulatory and tissue-resident subtypes of memory CD8⁺ T cells throughout their differentiation³¹. In addition to this shared signature, intestinal T_RM cells exhibited a distinct signature, which may contribute to their adaptation to the intestinal environment (Figure 1, tissue-specific adaptations). Importantly, by focusing the analysis on T_RM cells, the study also showed that the T_RM population is heterogenous throughout the course of an infection and exists in multiple clusters, many of which were present at several time points. These findings suggest that multiple subtypes of T_RM cells exist. For example, at day 60 of infection, two distinct types of clusters existed within the small intestinal T_RM population: one with higher expression of memory-associated transcription factors, such as Id3, Jun, Fos, and Klf2, and higher expression of Il7r (encoding for CD127), and another with expressed transcription factors, such as Zeb2, that might indicate a more terminally differentiated cell state and have been shown to promote terminal differentiation of circulatory effector and memory T cells⁷² and lower expression of Il7r³¹. The difference between these two T_RM subsets was also highlighted by higher cytokine production after restimulation with a cognate antigen in the CD127^high T_RM cells compared to CD127^low T_RM cells. Therefore, these subsets seem to phenotypically and functionally mirror the Id3^hi and Blimp1^hi subsets that were identified using reporter mice and confirm the presence of functionally relevant T_RM subsets that change in their relative abundance over time.

A recent study using flow-cytometry and scRNA-seq to analyze donor-derived T_RM cells from intestinal transplant recipients reveals substantial heterogeneity within the T_RM compartment in a single tissue can be observed in humans as well¹³. Remarkably, by exploiting HLA-mismatches between the organ donor and recipient, the authors prove that donor-derived T_RM can survive in the organ graft for at least 5 years following intestinal transplant¹³. Much like the mouse T_RM compartment, two distinct populations were observed among the donor-derived intestinal T_RM cells: a first population expressed higher levels of ITGAE, IL7R, KLRB1 and CCR6, indicating a more memory-like cell state, and a second population expressed higher levels of KLRG1, cytotoxic granules and the transcription factor ZEB2, thus resembling more terminally differentiated effector T cells¹³. Extending the characterization of intestinal T_RM to the disease state, a different study analyzing intestinal T cells from patients with ulcerative colitis and healthy controls revealed enrichment of T_RM-like cells with inflammatory properties (higher expression of ZEB2, TBX21, and PRF1) compared to controls⁷⁰. These data suggest, that T_RM cells normally exist in an equilibrium of multiple distinct differentiation states and that an imbalance of these T_RM differentiation states might be associated with, or even causative for, autoimmune-mediated diseases such as ulcerative colitis.

A recent complementary study highlights heterogeneity within the human skin T_RM population⁶¹. Using scRNA-seq of skin-resident T cells from two individuals, the authors were able to identify clusters of skin T_RM cells that were transcriptionally poised towards Th1-like (TBX21), Th17-like (RORA), and cytotoxic (GZMA, GZMK, NKG7, PRF1, IFNG) responses⁶¹. The concept of skin T_RM heterogeneity is further corroborated by previous studies that identified functionally and phenotypically discrete populations of resident and recirculating memory T cells in the skin^73,74. CD103⁻ T_RM cells mainly localized in the dermis and a higher proliferative capacity compared to CD103⁺ T_RM cells, which were enriched in the epidermis and had higher effector functions⁷³.

In summary, these studies support the idea that multiple subsets within the T_RM population with distinct functional capacities and transcriptomic programs exist in both mice and humans. This heterogeneity observed within the T_RM population is also likely reflected in the T_RM-mediated protection from secondary infection. More cytotoxic T_RM cells might preferentially act on infected cells, and their reduced proliferation capacity might limit immunopathology caused by activated T_RM cells. In contrast, T_RM cells with higher stemness or “memory potential” might replenish and sustain the T_RM pool over longer time periods. In line with this idea, recent studies have identified that CD103⁻ T_RM cells have the capacity to replenish and augment the CD103⁺ T_RM pool upon a secondary infection^75–77. Given the strong evidence for distinct subsets within intestinal T_RM cells, it will be interesting to see which factors drive and maintain differentiation of these subsets. It is tempting to speculate that these cells might be localized in different areas of the tissue and that the tissue microenvironment, the level of inflammation, the abundance of antigens, the presence of cytokines, chemokines, and metabolites, and the interactions of T cell partners might govern the T_RM cell adaptation to the tissue to allow for optimal tissue surveillance.

Infection-mediated variation in tissue-resident memory T cells

Infection type dictates the T_RM phenotype

In addition to the environmental influences derived from different tissue that may regulate T_RM formation and maintenance within a tissue^5,49, the type of infection also shapes the outcome of T_RM differentiation. For example, LCMV Armstrong causes an acute and systemic infection that leads to formation of T_RM populations in many tissues. In this infection setting, the T_RM population within the small intestine almost entirely expresses CD69 and CD103. In contrast, following Yersinia pseudotuberculosis infection, two distinct T_RM populations form in the small intestine²⁸: a CD103⁺ T_RM population and a CD103⁻ T_RM population. The CD103⁺ subset constitutes the majority of T_RM cells in the intraepithelial layer, whereas both populations are equally abundant in the lamina propria. Unlike CD103⁻ T_RM cells that develop in the absence of TGF-β, cluster around areas of bacterial infection, and are enriched in the lamina propria, CD103⁺ cells require TGF-β for their development and are scattered throughout the intestine²⁸. Importantly, CD103⁻ T_RM cells in this setting are not precursor cells of CD103⁺ T_RM but represent a distinct population that is stable over time. It has been shown that a single infection can give rise to two independent CD8⁺ T_RM populations and that the phenotypic composition of the T_RM population induced by different pathogens may reflect the unique signals associated with the infection type.

Route of infection alters the T_RM phenotype

The phenotypic and transcriptional features of T_RM cells are not only a result of the infection model used, but also depend on how the pathogen is administered. For example, oral infection with a mouse-adapted Listeria monocytogenes strain that contains a mutation in the internalin A protein to facilitate invasion of murine epithelial cells efficiently induced T_RM cells in the small intestine³⁰. However, mice infected with the same bacterial strain via an intranasal method, had fewer intestinal T_RM cells, and these cells did not fully convert to a memory phenotype⁷⁸. Similarly, infection of mice intranasally with an influenza virus expressing ovalbumin did not induce memory cells in the intestine³⁰. Thus, the route of infection, and thereby the location of T cell priming, dictates where and to what extent T_RM cells form in peripheral tissues.

Another study comparing how T cell priming in the spleen versus T cell priming in the mesenteric lymph node contributes to T cell differentiation into CD103⁺ intestinal T_RM cells^29,79 provides further evidence that the location of T cell priming regulates T_RM formation. Comparing intravenous infection, where T cells were primed in the spleen, to foodborne infection with Listeria monocytogenes, where T cells were primed in the mesenteric lymph node, revealed that only mesenteric lymph-node-primed T cells give rise to CD103⁺ intestinal T_RM cells²⁹. This difference in T_RM licensing is attributed to the presence of retinoic acid in the mesenteric lymph node, which regulates expression of intestinal homing receptors, such as CCR9 and α4β7-integrin, as well as genes that are part of the T_RM gene-expression signature (Hic1, Xcl1, Itgae, and P2rx7)²⁹.

These findings are of direct therapeutic relevance. While intramuscular, intravenous, or intranasal administration of an mRNA vaccine targeting the influenza A virus nucleoprotein were all capable of establishing memory cells in the lung and draining lymph nodes, the combination of intramuscular priming with an additional intranasal boost archived the highest frequencies of lung T_RM cells³⁸. This “prime-and-spike” approach has also been successfully used in mouse models of COVID-19 infection⁶³, but, surprisingly, it was not able to induce an effective immune response in a human phase I trial⁸⁰. It is not entirely clear why these differences occur, but it might be partially explained by different dosages of the vaccine or the way the vaccine is administered⁸⁰. However, these results highlight the need for a better understanding of how T_RM can be induced at specific tissue sites for therapeutic purposes. Thus, both tissue microenvironment and type of infection, and, therefore, inflammatory microenvironments, significantly impact T_RM development in response to acute infection and have direct implications for vaccination strategies to archive infection experience and induce optimal tissue protection.

Common adaptations of immune cells to tissue residency in specific tissues

In addition to CD8⁺ T_RM cells, many other immune cells, including CD4⁺ T cells, innate lymphoid cells (ILCs), macrophages, and natural killer (NK) cells, form tissue-resident populations similar to those described for CD8⁺ T cells^{15,16,40,41,81–84}. Interestingly, some of the transcriptional programs inducing tissue residency and inhibiting tissue egress seem to be shared among the different tissue-resident immune cell types. For example, Blimp1 and Hobit have well-established roles in both CD8⁺ T_RM and NK and NKT tissue-resident memory cells¹⁵, and suppression of Klf2 and S1PR1 and upregulation of Hobit is shared between CD4⁺ and CD8⁺ T_RM cells⁸⁴. These findings suggest that mechanisms for immune residency are shared between different immune cells and could reflect the adaptation to the unique environment of a specific tissue.

Runx3, which drives T_RM formation in CD8⁺ T_RM cells through a TGF-β-dependent transcriptional mechanism, is not required for tissue residency of CD4⁺ T_RM cells in the dermis, which instead relies on Runx1^16,85. This discrepancy might, in part, be explained by different localization and expression profiles of CD8⁺ versus CD4⁺ T_RM cells in the skin. CD4⁺ T_RM express less CD103 compared to CD8⁺ T_RM cells, preferentially reside in the dermal layer of the skin, and display a higher dynamic movement compared to CD8⁺ T_RM cells⁸⁶. Notably, overexpression of Runx3 in CD4⁺ T_RM cells changes the localization of CD4⁺ T_RM in the skin with a preference to the epithelium, similar to CD8⁺ T_RM cells, and enhances effector functionallity⁸⁵. These findings show that ectopic expression of Runx3 in CD4 T cells induces a residency program similar to that seen in CD8 T_RM cells, which is usually not acquired by CD4 T_RM cells due to the lack of Runx3 expression.

In addition to a core residency program shared among tissues, organ-specific adaptations seen in T_RM cells are also mirrored by other immune cell types (Figure 1). For example, the transcription factor Hic1, which was first identified as a tumor suppressor in human cancers^87,88, is relatively specifically expressed in intestinal T_RM cells¹⁷. Interestingly, the intestine-specific expression pattern of this transcription factor is also seen in other immune populations, including innate lymphoid cells (ILCs), CD4⁺, T_reg, and macrophage resident cells^17,52,89,90: Loss of Hic1 reduced accumulation of T cells and ILC3 in the lamina propria and intraepithelial lymphocyte compartments^52,89. These findings suggest, that Hic1 could serve a broad role in establishing or maintaining tissue-residency in the small intestine and that its function is shared among many cell types. More generally, these findings support the idea that tissue-specific adaptations can be shared across multiple types of resident immune cells (Figure 1).

In a similar example of common adaptations of immune cells to a tissue, the tissue-specific isoforms of fatty acid binding proteins (FABP) of CD8⁺ T_RM cells mimic the expression pattern seen in other resident immune cells for skin, small intestine, and liver⁵⁹. Thus, even though resident immune cells exhibit distinct differentiation trajectories and are derived from distinct precursor cells, they can share common adaptation methods to the tissue they reside in. Intriguingly, understanding tissue-specific adaptations in one resident population might be extrapolated to other immune cells. Considering that an effective immune response requires coordination of many different immune cells, each with a specific function, the ability to boost tissue-acclimatization of multiple immune cells simultaneously within a given tissue could be a promising therapeutic concept. In contrast, inhibiting tissue influx of immune cells by inhibiting common adaptation processes or, conversely, directing regulatory T cells to a tissue, could be an effective therapy for autoimmune and immune-mediated diseases, highlighting the need to further understand how immune cells acclimatize to tissues.

T_RM heterogeneity is reflected in T cell function

The pathognomonic function of memory cells is to protect from reinfection, where the memory cells re-encounter their cognate antigen. Among the memory cell population, T_RM cells provide the first line of defense at barrier tissues. Upon activation, T_RM cells not only produce cytokines, such as IFNγ or TNF, but also, and maybe even more importantly, they can alert the neighborhood and recruit circulating lymphoid and non-lymphoid cells to the tissue to aid in the response^49,78,91. Importantly, T_RM cells can rapidly proliferate, traffic to regional lymph nodes, and even contribute to secondary circulatory memory populations^{26,36,92–94}. Thus, T_RM cells exhibit features of effector cells as well as features that are reminiscent of T_CM cells.

By comparing T_RM cells from the small intestine to the colon after acute LCMV infection in mice, a recent study identified higher expression of granzyme A and B in small intestinal T_RM cells compared to T_RM cells residing in the colon²⁶. However, the highest expression of IFNγ was observed in colon T_RM cells and for both the small intestine and the colon, IFNγ was higher in intraepithelial T_RM cells compared to T_RM cells located in the lamina propria²⁶. Importantly, this functional heterogeneity among intestinal T_RM cells did not appear to be primarily driven by differences in CD69 and CD103 expression²⁶ but reflected the specific functional adaptation of cells to their residing tissue.

The functional capacity of T_RM cells is diverse among tissues and even among subsets within a distinct tissue. This diversity adds another layer of complexity to understanding how T_RM adapt to barrier tissues and optimally respond to different types of antigens to sustain a long-lived memory population.

T_RM in secondary infections

Memory T cells are functionally specialized to provide enhanced protection against subsequent infection. T_RM cells locally surveil non-lymphoid tissues and provide the first line of defense upon reinfection at barrier sites. T_RM can undergo in situ proliferation^48,92–95 independent of help from CD4⁺ T cells⁹³ and can autonomously amplify local immune surveillance and memory in various barrier sites in response to a range of stimuli⁹³.

Secondary systemic adaptive immune responses are commonly attributed to circulatory memory cells. Recent reports suggest, however, that T_RM cells themselves can undergo retrograde migration and substantially contribute to secondary circulatory memory cells, thus forming an “outside-in” immune response^36,92,94. For example, in one experiment, congenically distinct OT-I T_RM engrafted skin was transplanted to infection-matched mice, and T_RM cells were then reactivated locally with their cognate peptide. After 2–3 weeks, ex-T_RM cells were detectable in the draining lymph node and gave rise to circulating T_CM and T_EM cells. Interestingly, this egress from the non-lymphoid tissue could be reduced by treatment with FTY720, a S1P-receptor modulator, suggesting that the tissue egress is at least partially mediated by S1P⁹². In contrast, another report studying the recall response of skin T_RM cells did not find a contribution of T_RM cells to the circulatory memory response⁹⁵. In this study, HSV-specific T cells were injected into mice, and skin-resident T_RM cells were induced with DNFB, which acts as a non-specific stimulus to recruit T cells into the skin. Thereafter, these mice were rechallenged with local HSV infection⁹⁵. In this experimental setting, local proliferation of T_RM cells that provided localized protection from HSV-infection could be observed. However, skin T_RM cells were constrained to their location and did not migrate towards the lymph node after HSV infection⁹⁵. It is unclear, if these differences were a result of the different infections used or how T_RM cells were induced, but these findings highlight the diversity of site-specific immunity.

Most studies focusing on T_RM “stemness,” or ability to contribute to secondary responses and memory formation, rely on adoptive transfer experiments of T_RM cells into new recipient mice. Using the transfer T_RM has revealed that small intestinal intraepithelial T_RM cells can lose the expression of their hallmark molecules CD69 and CD103 and transdifferentiate to secondary circulatory memory populations^{26,36,49,92,94}. Phenotypically, these cells mostly give rise to T_EM-like memory cells (as identified by CD127⁺CD62L⁻), but few ex-T_RM cells also upregulate L-selectin (CD62L), a marker that is classically expressed by T_CM cells^49,92. After transfer of equal numbers of T_RM, T_EM, and T_CM cells into freshly infected recipient mice, cell numbers of secondary circulatory memory cells derived from ex-T_RM cells were higher than those derived from ex-T_EM, but lower than those derived from ex-T_CM cells. Thus, T_RM cells can display “developmental plasticity” and are not terminally-differentiated⁹². This finding is also supported by a machine-learning approach trained on a whole-genome bisulfite sequencing dataset using naive T cells as a reference for the highest stemness and exhausted T cells as a reference for the lowest plasticity⁹². The authors found that T_RM cells had an intermediate plasticity score between T_EM and T_CM cells. However, despite their ability to contribute to secondary memory populations, these ex-small-intestinal T_RM cells continued to exhibit phenotypic traces of their tissue origin and had a higher expression of Ly6C and CCR9 compared to ex-T_CM and ex-T_EM cells. In line with this, and despite their phenotypic plasticity, small intestinal ex-T_RM cells retain a bias to their tissue of origin⁹².

Additional studies comparing the capacity of T_RM cells from different tissues to contribute to secondary memory responses found substantial differences between skin ex-T_RM cells and liver ex-T_RM cells: skin ex-T_RM cells were unable to expand and repopulate the splenic or hepatic T cell pool to the same extent as ex-liver T_RM cells³⁶. This difference is partially explained by the finding that T_RM cells from the skin depend on TGF-β, whereas liver T_RM can form independently of TGF-β. Furthermore, using a tamoxifen-inducible knockout of TGFBRII in T_RM cells after acute LCMV infection, has shown that T_RM cells in the small intestine and salivary gland require ongoing TGF-β signaling, whereas T_RM cells in the kidney, liver, and visceral adipose tissue can persist without TGF-β¹⁷. This finding suggests that TGF-β imprinting may be one underlying mechanism repressing T_RM memory potential. This is further supported by the finding that TGF-β-dependent CD103⁺ salivary gland ex-T_RM cells are disadvantaged in repopulating the splenic memory population compared to their CD103⁻ counterparts, which do not require TGF-β³⁶, and also by a report showing that TGF-β-dependent T_RM formation limits the response to tumor vaccines⁹⁶. Reflecting the intra-tissue heterogeneity of T_RM cells, it has been shown that the Id3^high T_RM populations, which exhibit increased multifunctionality, yield a greater frequency of circulatory resident populations after rechallenge compared to their Blimp1^high counterpart²⁷. This study showed that the observed heterogeneity within the T_RM populations is functionally relevant in primary and secondary immune responses and that different T_RM subsets may be differently poised to contribute to secondary infection.

In conclusion, T_RM cells need distinct signals, cytokines, and transcriptional programs to adapt to their specific niche in their tissue of residence. This leads to specialized T_RM cells that exhibit different phenotypic traits, distinct functions, and also different potential to contribute to secondary infection response. Hence, by both studying the differences and commonalities between T_RM from a broad range of tissues, we can infer what makes a T_RM cell responsive to secondary infections. It will be interesting to see how tissue-specific adaptations by primary T_RM cells influence their ability to respond to secondary infection and contribute to secondary memory and whether tissue-specific adaptations require a trade-off with future memory potential.

However, one main limitation of analyzing the contribution of T_RM cells to secondary infections by transferring sorted populations into new hosts is that T_RM cells are subjected to harsh methods of isolation that might alter their biology due to the necessity of isolating them from the tissue. Further insights were gained by studies focusing on secondary T_RM responses in situ. Using intravital mucosal imaging, T_RM cells were found to proliferate within their residing tissue upon antigen restimulation, and it was found that the pool of secondary T_RM cells was mostly derived from primary T_RM cells independently of T_CM cells and proliferation in lymphoid tissues⁹³. Furthermore, using reporter and fate-mapping mouse models has brought many new insights into how T_RM cells within barrier sites respond to reinfection. These reporter mice models rely on genes that are uniquely expressed in T_RM cells and are hence used to mark and track T_RM cells during secondary infection without the need to isolate the memory cells, disrupt them from their residing environment, and transfer them into a new host.

For example, the transcription factor Hobit can be expressed in T_RM cells in the liver, the small intestinal IEL, and LPL T_RM cells, but not in circulatory or mesenteric lymph node memory T cells¹⁵. Using a Hobit-reporter mouse (tdTomato integrated into the Hobit locus) and a Hobit fate-mapping mouse (Hobit-Cre R26-LSL-eYFP), it has been confirmed that reinfection causes local expansion of T_RM cells. These cells can also drain into lymph nodes and contribute to circulatory memory where they resemble T_EM cells expressing KLRG1 and CX3CR1, and only a small proportion expresses the T_CM marker CD62L⁹⁴. These secondary ex-T_RM cells downregulate the expression of Hobit, which is in line with the finding that IEL ex-T_RM cells can lose expression of CD103 and CD69 after transfer and reinfection. These results underscore that at least certain parts of the T_RM adaptation program are not permanently imprinted and that T_RM cells can adapt to new environments and transdifferentiate into other T cell memory subsets.

Hobit expression is neither restricted to T_RM cells from a specific organ nor specifically expressed in a functionally distinct T_RM subset. Two new complementary studies addressed this by developing CD103⁺ fate-mapping mice that specifically and permanently mark CD103⁺ T_RM cells, thereby allowing dissection of the differences between CD103⁺ and CD103⁻ T_RM populations^75,76. Using a CD103-CreER^T2 mouse crossed to a flox-STOP-flox fluorescent reporter mouse, both groups could selectively mark cells that express CD103 at a specific time by treatment with tamoxifen at memory timepoints, thus excluding cells that might only temporally express CD103 during early activation or the effector phase. Despite using different infection models (VSV, Listeria monocytogenes, LCMV, and Yersinia pseudotuberculosis) and different TCR transgenic mice, both studies found, that CD103⁺ T_RM cells residing in the intestinal epithelium were poorly reactivated upon reinfection. Surprisingly, reporter positive CD103⁺ cells did not retreat to the draining mesenteric lymph nodes. Instead, the CD103⁻ T_RM population responded to secondary infection, showing an increase in TCR signaling (as measured by Nur77) and in proliferation compared to their CD103⁺ T_RM counterparts^75–77. Since CD8⁺ CD103⁺ T_RM cells do not show an increase in TCR signaling activity and seem to respond in a TCR-independent manner after antigen reencounter, it will be interesting to determine what signals activate these cells. A recent report found that CD4⁺ T_RM cells in the lung that formed after Bordetella pertussis infection in mice, were able to respond to non-cognate immune challenges^97,98. These findings show that the TCR-independent activation of T_RM can occur in both CD4⁺ and CD8⁺ T_RM cells and suggest that T_RM might acquire innate-like features to respond to secondary infection at barrier sites. The lack of proliferative capacity of CD103⁺ T_RM cells might be a physiological adaptation to limit immunopathology.

Supporting the idea that CD103⁻ T_RM cells possess a higher proliferative capacity, these studies found that the percentage of reporter-positive cells within the CD103⁺ T_RM population decreased after reinfection, suggesting that the pool of CD103⁺ cells is re-seeded by CD103⁺ T_RM cells. To exclude a competition for antigen and space limitations, CD103-Cre-ER^T2 mice were crossed to DTR^floxed mice, thus, after treatment with tamoxifen, all CD103⁻ cells could be depleted with diphtheria toxin to efficiently eliminate competition between CD103⁺ and CD103⁻ T_RM cells, as well as circulatory memory cells during reinfection. Interestingly, even in these artificial conditions, CD103⁺ T_RM cells only marginally contributed to circulatory memory populations, showing that the limited expansion capacity is an intrinsic feature of CD103⁺ T_RM cells⁷⁶. In summary, these studies suggest, that CD103⁻ T_RM cells in the small intestine might represent a precursor T_RM cell population with the ability to replenish and sustain the CD103⁺ T_RM pool in the tissue.

In previous reports, T_RM cells were shown to locally proliferate upon antigen re-encounter^92,93,95, and it is still unclear what causes these divergent results. This might be explained in part by different infection models, different induction of T_RM, or different organs studied. However, these differences make clear that the heterogeneity within a T_RM population in a single tissue results in different functional capacities and that CD103⁺ and CD103⁻ T_RM subsets need to be assessed separately. In terms of cytokine production, CD103⁺ cells predominately produced Granzyme A, whereas CD103⁻ T_RM cells secreted higher levels of IFNγ, TNF, and IL-2, which is in line with the higher production of these cytokines previously seen in Id3^hi cells, which also exhibited a greater plasticity upon transfer and reinfection^27,75. However, in contrast, these Id3^hi cells exhibited more expression of CD103⁺, indicating that one marker alone might not be sufficient to delineate among the functionally distinct subsets of T_RM cells in the small intestine.

Outlook

Localized tissue-immunity is an integral component of our body’s immune defenses. Understanding, how immune cells enter and persist within a tissue has direct therapeutic potential. Recent studies highlight that T_RM cells need to acquire tissue-specific changes to allow for optimal acclimatization to these tissues. Currently, most vaccination strategies focus on generating pathogen-specific memory cells or antibodies. However, recent studies suggest that by directing these memory cells to the exposed tissue, protection against infection can be enhanced. A better understanding of how memory cells adapt to the tissue microenvironment will hopefully allow for tissue-tailored vaccination strategies. Additionally, releasing T_RM cells from tissues could boost secondary systemic immune responses. Besides prevention of diseases, T cell therapies provide a novel treatment strategy for cancer and autoimmune diseases. Directing T cells to solid tumors remains a challenge, and mechanisms learned from tissue-resident cells could be extrapolated to improve tumor-infiltration and survival of T cells. Alternatively, allowing tissue egress of exhausted tumor-infiltrating T cells could render them more susceptible to checkpoint therapy, since they would be removed from the immunosuppressive tumor environment. Similarly, selectively targeting resident immune cells in autoimmune-diseases or after organ transplantation provides an exciting new treatment approach where alloreactive T cells could be depleted or functionally impaired^16,17.

Single-cell RNA sequencing has helped to dissect T_RM heterogeneity, but we are only beginning to understand the complexity of T_RM cell populations and the underlying mechanisms resulting in these specialized T cell subsets. By studying common mechanisms across multiple T_RM populations and dissecting individual adaptations of T_RM from a broad range of tissues, we can infer what signal, cytokines, cell-cell interaction, or environmental sensing mechanisms are required to initiate T_RM differentiation. Recent technological advances that allow the study of cellular transcriptomes in intact tissues, combined with computational methods creating cell-cell communication networks, will deepen our understanding of the signals driving T cell residency. This knowledge on how T_RM differentiation is embedded in a cellular and environmental network will bring us one step closer to tissue-directed immune-cell therapies.