Abstract
In recent years, tissue-resident memory T cells (TRM) have emerged as essential components of immunological memory. Following antigenic challenge, TRM remain in nonlymphoid tissues and defend against re-exposure. Although accumulating evidence suggests important roles for TRM in mediating protective immunity, fundamental aspects of the population biology of TRM remain poorly understood. In this article, we discuss how results from different systems shed light on the ecological dynamics of TRM in mice and humans. We highlight the importance of dissecting processes contributing to TRM maintenance, and how these might vary across phenotypically and spatially heterogeneous subsets. We also discuss how the diversity of TRM communities within specific tissues may evolve under competition and in response to antigenic perturbation. Throughout, we illustrate how mathematical models can clarify inferences obtained from experimental data and help elucidate the homeostatic mechanisms underpinning the ecology of TRM populations.
Introduction
Following primary infection and resolution of the associated cellular immune response, memory T cells are generated to protect against secondary exposures. It is well established that memory T cells comprise heterogeneous subsets. Whereas central memory T cells and effector memory T cells circulate between the blood and secondary lymphoid organs (SLOs) or peripheral tissues, respectively, tissue-resident memory T cells (TRM) are found in diverse tissue sites and provide crucial defenses against previously encountered pathogens (1–13). TRM are widely dispersed throughout the body and have been associated with protection against viral and bacterial infections, antitumor immunity, and the pathology of allergic and autoimmune diseases (3, 7, 14–21).
TRM are phenotypically and transcriptionally distinct from effector memory and central memory T cells, and the factors involved in their differentiation and maintenance have been studied extensively (reviewed in Refs. 22–28). As shown in parabiosis experiments involving the conjoining of two mice, TRM persist within a wide variety of nonlymphoid tissues (NLT) (17, 29, 30). TRM have also been identified in draining SLOs following local infection in the female reproductive tract (FRT) and skin (31). Evidence that TRM persist in the presence of agents that deplete circulating peripheral T cells (e.g., FTY720 or anti–T cell Abs) further indicates that TRM are noncirculating and largely maintained independently of circulating populations (14, 32, 33). Whereas TRM are phenotypically heterogeneous, and there are no perfect markers of tissue residency, the most commonly associated marker in mice is CD69. Although CD69 is often attributed to recent antigenic stimulation, it may also function to retain TRM within NLT (22), and parabiosis experiments have confirmed its utility in distinguishing tissue-resident from circulating cells (11, 17, 25). Another frequently used marker is CD103, which is predominantly associated with CD8+ TRM (34), although CD103+ CD4+ TRM are found in skin and FRT (3, 34–37). Other markers, such as CD11a and CXCR3, are associated with the TRM phenotype in certain tissues (11, 32, 34, 38–40).
Despite this extensive characterization, many fundamental aspects of TRM biology remain poorly understood. For example, what determines the longevity of memory encoded by TRM? How does the presence of TRM impact the recruitment of new Ag-specific T cells upon exposure to related or unrelated pathogens? Is there competition between pre-existing and newly generated TRM, and if so, what factors mediate this competition? To address these questions requires formulating a quantitative ecology of tissue-resident memory, so that we may understand how continual exposure to environmental and infectious Ags impacts the distribution, diversity, and persistence of TRM at different sites across the body.
By pairing quantitative models with experimental observations, one can test different hypotheses regarding cellular turnover and interactions, estimate quantities that may not be directly measured, and generate quantitative, testable predictions. Notably, mathematical models have elucidated mechanisms underlying the generation and resolution of effector T cell responses and the maintenance of naive and circulating memory T cell populations (reviewed in Ref. 41; see also, for example, Refs. 42–52). Further afield, there is a rich literature employing models to understand the ecological processes sustaining communities of plants, animals, and infectious agents (e.g., see Refs. 53–56). In this article we collate information regarding the ecological dynamics of TRM, including current quantitative estimates of growth and loss rates in different tissues. We first consider studies performed in mice, as these make up the bulk of the work to date, and then discuss our understanding of TRM ecology in humans. Throughout this review we illustrate how mathematical tools can be harnessed to refine and enhance current experimental insights and highlight open questions and areas for future work.
Identifying ontogenic pathways
The ontogeny of TRM across different tissues has not been fully characterized. Although, in general, it seems that TRM are enriched for relatively long-lived, quiescent cells in an early differentiated state (57, 58), a definitive differentiation pathway remains elusive. One barrier to consensus in this area is that T cell responses are heterogeneous and move through high-dimensional phenotypic trajectories, of which any given experiment only views a projection. However, by allowing us to frame mechanistic descriptions of the dynamics of proliferation, loss, and differentiation, mathematical models can be used to explore competing hypotheses regarding patterns of T cell differentiation (e.g., see Refs. 45, 47, 51, 59).
Quantifying the rate of TRM generation from precursor populations, and whether this rate changes over both short timescales (such as during the contraction phase of an immune response) and throughout life, is important for understanding how TRM populations are established and maintained. Recent work suggests that host age modulates the rate of TRM generation, as activated T cells in infants show reduced capacity to differentiate into lung TRM compared to cells in adults (60). Such hypotheses can be explored using models that test the impact of host and/or cell age on population kinetics (49, 52, 61). For instance, a comparison of models describing different mechanisms of naive T cell homeostasis identified increasing cell survival with postthymic age as the most likely determinant of population stability throughout life (49). Such methods could be used to dissect the roles of host and cell age in modulating TRM formation.
Discriminating between models of T cell responses in vivo, and of TRM ontogeny in particular, requires frequent sampling of the phenotypic profile of T cells during the acute, resolution and memory phases, both in the tissue and draining lymph nodes. Such data are typically cross-sectional, but even data from single-cell fate-mapping experiments leave uncertainty regarding lineage relationships (45, 59, 62, 63). Nevertheless, if candidate mechanisms or pathways can be expressed as mathematical models, there exist statistical methods to compare the support for each (e.g., see Ref. 64). Such methods assess the ability of different models to describe observations while penalizing those that are too complex and tend to “over-fit”—that is, those models that are overly tailored to the particulars of a given (inevitably noisy) dataset and do not generalize well to data from other experiments. The most parsimonious models can thus be identified from a set of candidates and used to infer biological parameters and underlying mechanisms. Additional data may then be used to test model predictions and further assess their biological relevance.
Quantifying and dissecting contributions to TRM longevity
Perhaps the key motivation for understanding mechanisms of TRM development and maintenance is to optimize immunization protocols and maximize the longevity of tissue-resident immunity. Important questions include: does the longevity of TRM clones depend on Ag specificity and/or tissue location? Does longevity map directly to the phenotypic heterogeneity of TRM? And to what extent does competition between TRM, either at the level of the whole population or among Ag-specific populations, influence longevity? To tackle these questions quantitatively, we first note that the longevity of a TRM population is determined by the balance between self-renewal, loss, and recruitment from circulating precursor populations (Box 1). In particular, if there is proliferative self-renewal of TRM, the lifetime of individual TRM cells may be shorter than the lifetime of the population as a whole. In what follows, then, we use the term “longevity” to define the persistence of a population, combining the net effects of influx, division, and loss.
Box 1: Modeling TRM homeostasis
Consider a single homogeneous population of TRM, T. A simple model of homeostasis assumes that newly recruited TRM cells enter the population from a precursor subset at a constant rate, r. They then divide in situ at rate p and are lost at some rate d, which reflects both cell death and egress from the tissue (Fig. 1). The average residence time of an individual cell in the tissue is 
. These processes can be expressed mathematically as
(1)where 
is the net loss rate from the population, and represents the balance of loss and self-renewal through division.
If there is no recruitment from precursor populations, then 
and Eq. 1 can be solved to give
(2)where 
is the initial number of TRM. Thus, assuming cells are lost at a faster rate than they divide (i.e., 
), the population will decline exponentially at net loss rate 
. This quantity, rather than the residence time of any particular cell, is most relevant for quantifying the duration of immunity (i.e., the persistence of a T cell population that undergoes both loss and self-renewal). It follows that the average time for this population to halve in size (t1/2) is 
. One can also use the net loss rate to estimate, for example, the time taken for the population size to fall below a defined threshold of protection.
Finally, if 
, then some cell loss will be masked by recruitment of new TRM. Models that fail to account for this recruitment will underestimate the net loss rate, 
, and thus overestimate the corresponding population t1/2. If r is constant and 
, the population will eventually reach an equilibrium, 
, where influx and renewal are balanced by loss. This steady-state population size can be found by setting 
, giving 
.
Schematic of TRM homeostasis model. Cells are recruited at a constant rate r, self-renew through division at rate p, and are lost at rate d, leading to exponentially distributed residence times within the tissue. Extensions of this model include different residence time distributions or decomposing the decay rate into distinct terms for cell death and tissue egress (107).
Measuring longevity is hampered by the censoring of experimental information—estimates can be dependent on the duration of sampling. Nevertheless, broad patterns have emerged that indicate potential tissue specificity in TRM longevity: populations in the lungs and liver persist for weeks to months (7, 11, 32, 65–68), whereas those in tissues such as the brain and skin may persist for months to years (Table I) (4–6, 8, 12, 17, 35). Although the longevity of TRM may also be phenotype specific, as it is in circulating memory subsets (41), little is known in this regard. Despite evidence that CD8+ memory cell populations persist longer than CD4+ memory cells in tissues such as the skin and FRT (5, 35, 69, 70), there have been no direct comparisons of the longevity of CD4+ and CD8+ TRM. Similarly, although some evidence suggests that regulatory CD4+ T cells, another distinct subset of resident memory, can persist for over 50 d in the skin and lungs, their population dynamics are, in general, more poorly defined than those of conventional CD4+ and CD8+ TRM (71, 72).
Although experimental studies have provided valuable insights into longevity in different tissues, the majority measure TRM abundance over time and cannot untangle the contributions of recruitment, self-renewal, and loss. For instance, the rapid loss of influenza-specific TRM in the lung has been attributed to all three components of population longevity: high levels of apoptosis-driven loss, low levels of proliferation, and decreasing recruitment of circulating precursors over time (65, 67, 73). However, direct quantitative comparisons of the relative roles of each are currently lacking. Ultimately, to develop a mechanistic understanding of TRM homeostasis, we need to quantify the relative contributions of proliferation, recruitment, and loss. These factors are discussed in detail below.
The extent of TRM recruitment at steady-state (in the absence of infection) varies across NLT, cell phenotypes, and possibly infections. Recently-generated CD4+ TRM in the lung, and CD8+ TRM in the brain and skin, remain stable in numbers in the absence of input from the circulation (32, 35, 74). In contrast, CD8+ TRM in the lung are at least partly replenished by circulating memory precursors (67, 75), and there may also be low-level seeding in the FRT (17). Given the sensitivity of the above analyses to sampling times, and the potential for changes in the rate of recruitment from precursors over time (67), further work is needed to measure the long-term contribution of recruitment to TRM maintenance.
Although TRM populations expand efficiently in response to antigenic restimulation (12, 58, 76), the extent to which they self-renew at steady-state is unclear. BrdU-labeling experiments have indicated that low-level proliferation occurs in the lung, skin, and brain, with 0–5% BrdU uptake over 7 d (3, 67, 74). However, it has also been shown that a more substantial fraction (around 9%) of established TRM in the brain express Ki67, a marker of recent division (12). Evidence that these Ki67hi cells are established TRM, rather than recently-divided immigrants, implies a more important role for self-renewal in certain tissues (12, 74).
Finally, loss of TRM can result from cell death (turnover) or tissue egress. In some systems, such as the lungs, TRM express lower levels of the antiapoptotic protein Bcl-2, and increased levels of the proapoptotic caspase 3/7, than their circulating counterparts, suggesting impaired cell survival (67). However, this impairment is not apparent for TRM in the skin, illustrating that loss rates likely vary across different tissues and cell phenotypes.
Combining these experimental observations with quantitative models will enable 1) characterization of homeostatic dynamics as cells die, divide, and/or differentiate and 2) rates of recruitment, self-renewal, and loss to be inferred (41, 46, 48). Although some quantitative methods have been employed to infer population half-lives in the liver and lungs (Table I; see Refs. 11, 68, 77), these may be overestimates, as cell recruitment was not taken into account (Box 1). To discriminate self-renewal of existing TRM from the recruitment of new cells, one can use fate-mapping systems or adoptive transfer of congenically labeled T cells prior to secondary challenge. As an example, the former has been employed to track the differentiation of congenically labeled cells that develop normally in the thymus and are released into a lymphoreplete periphery (47). In combination with BrdU labeling, models were used to quantify heterogeneity in the dynamics of effector and memory CD4+ T cells and the rate of naive cell recruitment into these populations. Applying similar approaches to TRM could disentangle the processes of recruitment and renewal and identify heterogeneity in population dynamics. In particular, heterogeneity is suggestive of TRM subpopulations with distinct homeostatic niches (Box 2), and the combination of modeling and fate mapping may reveal whether these coexist independently or whether there is interconversion between them. These methods also allow projection of TRM population dynamics beyond typical experimental time courses, to develop a more complete picture of TRM homeostasis and persistence (Box 3).
Box 2: Heterogeneity in TRM populations
The homogeneous model of TRM homeostasis (Eq. 1) assumes the rates of input, loss, and division are constant over time and across all cells in the population. However, these assumptions may not always hold. For instance, input rates may wane over time, as has been found in the lung (67), and phenotypic heterogeneity may lead to different rates of division and/or loss, as has been found in other T cell subsets (43, 47). Such mechanisms can lead to heterogeneous population dynamics. For example, differences in net loss rates among phenotypic subsets can manifest as multiple phases of exponential decline, rather than the single phase of a homogeneous population (Fig. 2) (46). By modifying Eq. 1 to reflect these scenarios, heterogeneity in TRM ontogeny and homeostasis can be identified and compared (although the ability to detect such heterogeneity requires sufficient data, such as detailed time courses from fate-mapping and DNA-labeling experiments) (47, 108, 109).
Impact of heterogeneity on population dynamics. (A) Assuming precursor input is negligible, homogeneity in the net loss rate across the population leads to a single exponential decay function (Eq. 2). (B) Conversely, if there is heterogeneity in net loss among cell subsets (e.g., because of different death rates), then the observed decay (dashed line) reflects the sum of the decay kinetics of each subset (red and blue lines).
Box 3: Mathematical analysis of published data
To illustrate the insights that can be gained from mathematical modeling, we collated data from two independent studies of CD8+ TRM persistence in mice. One describes the number of OT-I TRM in the lungs and nasal tissues (NT) from 20–120 d following infection with the recombinant X31-OVA strain of influenza virus (66). The other follows the number of P14 TRM in the skin and lungs 20–200 d postinfection with either Vaccinia-GP33 (skin) or PR8-GP33 (lungs) (67). The latter study also follows the endogenous TRM response to the NP366 protein following infection with the PR8 strain of influenza.
Because of sparse information on the extent of TRM recruitment in each tissue, we make the simplifying assumption that 
(i.e., influx is negligible over the experimental time frame). Although this assumption may be applicable to the NT and skin, it is unlikely to hold strongly in the lung (35, 66, 67). We also assume cells within each population are homogeneous with respect to their rates of division (p) and loss (d). TRM decline can then be modeled using Eq. 2 to estimate the t1/2 of the population (Fig. 3). If in situ proliferation is minimal, then 
and the net loss rate of the population (
) is equivalent to the loss rate of individual cells (d); this may be the case for lung and NT TRM (66, 67). More generally, however, 
(12), and distinguishing the contributions of p and d to the net loss rate requires further data regarding TRM self-renewal (e.g., Ki67 expression), death, or egress. Note also that the net loss rate will be underestimated if there is significant TRM recruitment (Box 1).
Importantly, mathematical models can be used to project population dynamics beyond experimental time frames. For example, our model predicts that OT-I (Fig. 3A) and NP366 (Fig. 3B) TRM cells will become undetectable in ∼220 and 210 d, respectively. These estimates are in agreement with the observation that lung TRM are undetectable 7 mo (∼210 d) postinfection (7).
Mathematical analysis of published TRM data. Eq. 2 fitted to data from (66) (A) and (67) (B). Solid lines are the model predictions, points are the data, and gray shaded regions are prediction envelopes obtained by sampling parameters from their 95% confidence intervals. Least-squares optimization was performed on the log-scale and residuals were normally distributed (using the Shapiro–Wilk test), except those for the nasal tissues (NT). The NT fits are thus included for approximate comparison only. Note that we use the above approach to illustrate how a simple model can be applied to experimental data. In a more thorough analysis, these predictions would then be compared with those from alternative models to identify the most parsimonious description of the data. Code for the above analysis can be found at https://github.com/SineadMorris/TRMecology.
TRM competition and community composition
Community composition refers to the diversity and spatial distribution of cells coexisting within a tissue. This composition may be altered if TRM compete with other cells (including other TRM subsets) within their local niche. In the following section we consider competition and its subsequent impact on cell diversity within NLT, first by focusing on competition arising during homeostasis or following a single inflammatory response, and then by considering how populations change across multiple exposures.
First, it is important to understand the extent to which TRM compete with one another for homeostatic resources. In particular, competition implies the existence of a “carrying capacity” that may limit the amount of TRM a tissue can support. Such a restriction on population size may be important in the skin following HSV challenge, as TRM numbers eventually plateau with increasing numbers of transferred effector cells, despite a continued increase in numbers in the spleen (58). Although saturation in TRM formation must be carefully distinguished from saturation of other upstream immune responses (78), identifying potential carrying capacities is important as the density of Ag-specific TRM within a tissue can influence the efficacy of protection (11, 58). Models can provide further insights into these systems by exploring the density of TRM required to provide effective protection and predicting when populations will fall below this threshold (77, 79).
One mechanism that may drive cell competition is cytokine-mediated signaling. For example, TRM maintenance and homeostatic division in the skin, salivary glands, liver, and kidney may be IL-15 dependent, with IL-15 deficiency leading to significant reductions in cell numbers (34, 68, 80–82). However, these data do not address whether IL-15 is limiting at steady-state and thus are not direct evidence for cytokine-mediated competition under normal conditions. Another mechanism through which TRM may compete is Ag availability. If TRM formation is Ag dependent, immunodominance hierarchies can arise among the newly generated population: more abundant epitopes stimulate greater numbers of the corresponding clone. Such hierarchies have been observed in the lungs, skin, and FRT (3, 8, 66, 83, 84). The clonal composition of TRM at a tissue site may be further shaped by influx from memory precursors (17, 67) and subsequent Ag exposure (76).
One caveat to the above discussion is that most studies of TRM maintenance use specific pathogen-free (SPF) or germ-free mice, which are typically isolated from exposure to environmental Ags. As such, the density and diversity of TRM in tissues may be relatively low. In contrast, greater insights may be gained from so-called “dirty mice,” which are exposed to more diverse environmental Ags than those in traditional SPF facilities, and in which TRM are typically found at greater densities (31). The populated tissues of these mice may more closely reflect normal conditions, and thus provide a better system to study competition between cells.
With respect to sequential exposures, multiple antigenic stimuli may change the diversity of the TRM community as newly recruited and pre-existing cells compete for resources within a particular niche (85). In the skin, proliferation of pre-existing TRM following HSV rechallenge is accompanied by increased apoptosis, suggesting prior population expansion may be controlled during recall responses (58). Despite this, a stable population of original cells can coexist with new recruits following either homologous or heterologous infection (58). In the FRT, TRM numbers increase following consecutive homologous challenge, primarily due to in situ proliferation of pre-existing cells (76). Finally, in the liver, pre-existing and newly established TRM coexist following consecutive transfers of three heterologous populations (68). Together, these results suggest heterogeneous TRM populations can persist in a number of NLT following multiple exposure events. However, the extent to which this applies over longer timescales, across other sites, and against more biologically relevant exposure histories (e.g., SPF versus dirty mice) has yet to be determined.
An additional challenge in understanding community composition is to identify the specific rules of replacement—is TRM homeostasis democratic, in the sense that all specificities or phenotypes compete equally for space and/or homeostatic resources in the absence of Ag? Do pre-existing TRM tend to exclude new immigrants, or do newer cells have a competitive advantage? Does competition take place on a global level (independent of Ag specificity), or are there Ag-specific niches, created by either persistent Ag or cell congregation at previous sites of infection? These questions can be addressed by further measuring the persistence of epitope-specific TRM after multiple exposures to related and heterologous pathogens (68). Such insights will be of critical importance in understanding how host immunity changes in response to sequential infections.
The ecology of tissue-resident memory in humans
TRM in humans are less well studied, primarily because of difficulties in sampling NLT. Although in vivo challenge experiments have been carried out with healthy volunteers (10, 86), such studies are generally rare. Nonetheless, recent insights have been gained from a variety of valuable sources, including tissues sampled from organ donors and patients undergoing surgical procedures (84, 87–92). These snapshots in time reveal the diversity of TRM populations that accumulate throughout life in response to multiple pathogen exposures. In addition, they complement studies of germ-free and SPF mice that do not consider dynamics in the context of pre-existing immune landscapes.
In general, the phenotypes of human TRM are relatively well characterized. Despite differences in the transcription factors regulating their differentiation (93–97), human TRM share core transcriptional and phenotypic profiles with their murine counterparts, including high expression of CD69 and enrichment for CD103 in CD8+ populations (84, 88, 94, 98–102). Additional surface markers, including CD49a, CXCR6, and CD101, are also enriched at multiple sites (94, 103). Another notable finding from human studies is that TRM are more widely distributed than in laboratory mice: TRM have been identified in human lymphoid tissues (e.g., spleen, lymph nodes, and tonsils) in addition to NLT and are the most abundant T cell subset at these sites (88, 94). This greater dissemination may reflect increased exposure to environmental Ags (101) and is consistent with recent findings that TRM accumulate in the draining lymph nodes of mice following repeated antigenic stimulation (31). Overall, these studies highlight the importance of considering environment and host ecology when developing a conceptual framework of TRM dynamics.
Determining the dynamics of TRM maintenance in human tissues is particularly challenging given the difficulty of tracking cell populations over time. Nevertheless, cross-sectional analyses of mucosal and lymphoid tissues has revealed that TRM are maintained at remarkably stable frequencies from childhood well into old age, at levels that are tissue specific (104, 105). Important insights have also been gained from longitudinal studies. First, TRM clones specific for the allergen diphenylcyclopropenone (DPCP) persist in healthy human skin for at least 4 mo (10). Second, donor TRM persist in lung and intestine transplant recipients for over 1 y following transplantation (91, 92). Collectively, these findings suggest stable maintenance of human TRM, although quantitative estimates from other tissues and Ags will be required before broad conclusions can be drawn.
The relative contributions of recruitment, self-renewal, and loss to overall population maintenance are also not well known. However, the gradual formation of recipient TRM following lung transplantation suggests continual recruitment can occur (92). In addition, cross-sectional data from donor tissues indicate homeostatic proliferation of resident CD8+ and CD4+ cells is detectible, but low, across mucosal sites (89). In particular, few resident skin and intestine cells express Ki67 (91, 106), and expression levels in the lung are <8% (94). Although these observations suggest relatively low average rates of homeostatic self-renewal, recent partitioning of TRM based on their ability to efflux fluorescent dye has revealed underlying differences in Ki67 expression: cells capable of efflux (efflux+) show reduced rates of proliferation compared with efflux− cells (104). Moreover, the efflux+ subset exhibited increased CD127 expression and enhanced IL-7 signaling following ex vivo stimulation, suggesting greater survival capacity. Together, these findings suggest important heterogeneity in cell quiescence and survival that modulates TRM longevity at the population level.
Finally, with respect to competition and community composition, humans are exposed to a greater number and diversity of environmental Ags than germ-free or SPF mice. Thus, one important question is how the accumulation and composition of different TRM communities is impacted by these more frequent and multifaceted challenges. Recent sampling of donor tissue suggests that a diverse range of influenza-specific TRM clonotypes can be maintained in the adult human lung (84). A high degree of TCR diversity has also been observed in healthy human skin and is largely maintained in vitro during IL-2– and IL-15–induced expansion (87). These findings hint that human NLT can support an array of coexisting TRM communities. Mapping the dynamics of such communities, in the face of differential individual exposure history and genetic background, will be a key challenge for the future. In the absence of longitudinal sampling, one approach may be to use quantitative imaging of human tissue sections, including markers of proliferation and apoptosis, to establish the spatial organization of TRM in different organs and how self-renewal and turnover reflect tissue location and local TRM densities.
Conclusions
TRM are a crucial component of the adaptive immune system and provide frontline defenses against invading pathogens. Understanding the mechanisms underlying population persistence and community composition is therefore crucial to enhancing immune memory and improving vaccine-mediated protection. Although recent experimental studies have provided many insights into the molecular factors governing cell differentiation and maintenance, a broader picture of TRM ecology is still lacking. Mathematical modeling can help bridge this gap by providing a quantitative framework for exploring multiple different aspects of TRM ecology. In particular, integrating models with data will allow us to disentangle the contributions of recruitment, proliferation, and loss to TRM longevity, and project population dynamics beyond current experimental limits. Success in this endeavor will require longitudinal data from NLT and SLOs—a particularly challenging prospect for human studies.
Disclosures
The authors have no financial conflicts of interest.
Footnotes
This work was supported by the National Institutes of Health (R01 AI093870 to A.J.Y.).
Abbreviations used in this article:
- FRT
- female reproductive tract
- NLT
- nonlymphoid tissue
- SLO
- secondary lymphoid organ
- SPF
- specific pathogen-free
- TRM
- tissue-resident memory T cell.
- Received July 5, 2019.
- Accepted September 16, 2019.
- Copyright © 2019 by The American Association of Immunologists, Inc.
References
In this issue
