Central memory T cells are the most effective precursors of resident memory T cells in human skin

Abstract

The circulating precursor cells that give rise to human resident memory T cells (T_RM) are poorly characterized. We utilized an in vitro differentiation system and human skin-grafted mice to study T_RM generation from circulating human memory T cell subsets. In vitro T_RM differentiation was associated with functional changes, including enhanced IL-17A production and FOXP3 expression in CD4⁺ T cells and granzyme B production in CD8⁺ T cells, changes that mirrored the phenotype of T cells in healthy human skin. Effector memory T cells (T_EM) had the highest conversion rate to T_RM in vitro and in vivo but central memory T cells (T_CM) persisted longer in the circulation, entered the skin in larger numbers and generated increased numbers of T_RM. In summary, T_CM are highly efficient precursors of human skin T_RM, a feature that may underlie their known association with effective long-term immunity.

One sentence summary:

Central memory T cells are the most efficient precursors for human skin resident memory T cells.

Introduction

Resident memory T cells (T_RM) accumulate in epithelial barrier tissues following infection, undergo a distinct differentiation program, have strong effector functions and provide potent and front-line protection against known pathogens (1–3). It is not known which circulating human T cell subsets have the ability to differentiate into T_RM in peripheral tissues and what if any functional differences exist in T_RM generated from distinct memory cell types. A better understanding of human T_RM precursor cells is needed to develop vaccination strategies that optimally populate peripheral tissues with T_RM. We utilized an in vitro differentiation system incorporating human keratinocytes and an in vivo differentiation system using immunodeficient mice grafted with human skin to study the ability of human circulating precursor cells to give rise to T_RM. We report here our findings that CCR7⁺ L-selectin⁺ central memory T cells are the most efficient precursors for human skin T_RM.

Results

Multiple circulating memory T cell subsets generate T_RM in vitro

We tested three major circulating CD45RA⁻ memory T cell subsets for their ability to generate T_RM (Fig. 1A). Central memory T cells (T_CM) co-express L-selectin and CCR7, actively recirculate between the blood and lymph nodes, and a subset of these cells also recirculate through skin and other healthy human peripheral tissues (4, 5). Effector memory T cells (T_EM) lack expression of L-selectin and CCR7, are tropic for peripheral tissues and can differentiate into highly protective effector cells. Migratory memory cells (T_MM) are a subset recently described in humans that express CCR7 but lack L-selectin, have effector functions intermediate between T_CM and T_EM, and recirculate between the blood and the skin but appear to be excluded from central lymph nodes (6).

Figure 1. Multiple human circulating memory T cell subsets can generate T_RM in vitro.

(A) The gating strategy used for sorting of human peripheral blood CD45RA⁻ memory T cells into purified populations of CCR7⁻L-selectin⁻ effector memory T cells (T_EM), CCR7⁺L-selectin⁻ migratory memory T cells (T_MM) and CCR7⁺L-selectin⁺ central memory T cells (T_CM) is shown. (B) Experimental strategy for the study of T_RM generation in vitro. Sorted purified T cell subsets from human blood were stimulated for 24 hours with an anti-CD3/CD2/CD28 beads, cultured on monolayers of keratinocytes, fibroblasts or on tissue culture plastic and analyzed by flow cytometry. (C) Seven days of culture on keratinocyte monolayers generated single positive CD69⁺CD103⁻ and double positive CD69⁺CD103⁺ T cells (left panel), similar in phenotype to those observed in healthy human skin (right panel). (D, E) Time course of CD69 and CD103 acquisition in vitro after culture on keratinocyte monolayers. A representative culture is shown; similar results were obtained in a total of four donors. (F-I) All memory subsets tested gave rise to CD69⁺ T cells, some of which also upregulated CD103. Results for CD4 (top) and CD8 (bottom) T cells are shown after culture on tissue culture plastic (plastic), keratinocyte monolayers (ker) or fibroblast monolayers (fib). The mean and SEM of at least six donors are shown. (J, K) Culture on keratinocyte monolayers generated increased numbers of CD4 (J) and CD8 (K) T cells expressing T_RM markers. The mean and SEM of a minimum of four donors are shown. (L) CD4⁺ and CD8⁺ T cells were equally efficient at up regulating CD69 (left panel) but CD8⁺ T cells were more efficient at generating double positive CD69⁺CD103⁺ T cells. The mean and SEM of at least four donors cultured on keratinocyte monolayers are shown. (M,N) Naïve T cells did not upregulate T_RM markers in vitro. CD4⁺ (M) and CD8⁺ (N) CD45RO⁻CD45RA⁺ naïve T cells were bead stimulated and cultured on keratinocyte monolayers then analyzed by flow cytometry. Similar results were obtained in a total of three naïve donors. Naïve T cells produced significantly fewer CD69⁺ CD103⁺ T cells (right panels); the mean and SEM of three naïve donors and six memory donors are shown. Significance was determined by a one-way ANOVA and Tukey’s post-hoc test for multiple comparisons (F-K, M,N) or by two-tailed T-tests (L).

We stimulated flow sorted human memory T cell subsets for 24 h with anti-CD2/CD3/CD28 beads, then cultured them on human keratinocytes, fibroblasts or tissue culture plastic and subsequently assayed for the expression of the T_RM markers CD69 and CD103 by flow cytometry (Fig. 1B). In humans, CD69 is expressed at high and constant levels by all T_RM and CD103 is additionally expressed by a subset of T_RM enriched for epithelial localization (2, 6). Culture of CD45RA⁻ memory T cells from peripheral blood on keratinocyte monolayers for seven days generated a population of CD69⁺ T cells, a subset of which also expressed CD103, similar to the T_RM populations observed in human skin (Fig. 1C). When purified populations of CD4⁺ T_CM, T_MM and T_EM were bead stimulated and added to keratinocyte monolayers, CD69 was transiently upregulated at 24 hours, consistent with its known role as an early activation marker (Fig. 1D). After declining at 48 hours, CD69 was subsequently stably expressed on a subset of T cells, some of which also expressed CD103 (Fig. 1D,E). T cell expression of CD69 and CD103 remained stable for as long as the feeder keratinocyte monolayer was intact. After one week in culture, the keratinocyte monolayer degenerated and the number of CD69 and CD103-expressing T cells subsequently declined (Fig. 1D,E). All three circulating subsets tested gave rise to populations of CD69⁺ and CD103⁺ expressing T cells in vitro (Fig. 1F–K). CD4⁺ and CD8⁺ T_EM memory subpopulations generated the most CD69⁺ T cells (Fig. 1F,H). CD4⁺ T_EM also gave rise to the largest number of CD69⁺CD103⁺ co-expressing T cells (Fig. 1G), whereas CD8 T_MM and T_EM were equally effective at generating CD69⁺CD103⁺ double positive cells (Fig 1I). Culture on keratinocyte monolayers was the most effective stimulus for inducing CD69 and CD103 expression on T cells (Fig. 1J,K). CD4⁺ and CD8⁺ T cells gave rise to similar numbers of CD69⁺ T cells whereas CD8⁺ T cells generated more CD69 and CD103 double expressing cells (Fig. 1L). In contrast to memory T cell subsets, naïve T cells only transiently upregulated CD69 and did not upregulate CD103 (Fig. 1M,N). Given the short time course, these experiments do not evaluate the ability of effector T cells generated from naïve T cell to develop into T_RM,

CXCR6, a molecule necessary for T_RM generation, was also upregulated on in vitro CD69-expressing T cells and inclusion of the mTOR inhibitor rapamycin blocked induction of CD69 and CD103 in both CD4⁺ and CD8⁺ T cells (Fig. S1). Expression of CCR4 and L-selectin were lost on the majority of T cells that acquired T_RM markers (Fig. S2).

T_RM generated from distinct circulating T cell subsets differ in their functional characteristics

All three memory circulating T cell subsets tested (T_CM, T_MM, T_EM) gave rise in vitro to T cells that expressed T_RM markers. We next used mass cytometry to determine if there were functional differences between T_RM generated in vitro from distinct circulating precursor T cell populations. T cells expressing T_RM markers were generated in vitro from each memory subset; CD69⁺ CD103⁻ and CD103⁺ cells from each subset were then pooled and analyzed by mass cytometry (Fig. 2A). Clustering suggested functional differences existed between T_RM generated from distinct precursors (Fig. 2B, individual tSNE plots shown in Fig. S3, antibodies used included in Table S1). In-depth analyses demonstrated that CD4⁺ CD69⁺ CD103⁻ T cells generated from T_EM expressed significantly more IFNγ, those derived from T_MM expressed significantly more IL-17A and those derived from T_CM had significantly higher numbers of FOXP3⁺ T cells (Fig. 2C–E). Similar differences were observed in the CD4⁺ CD103⁺ populations, although only FOXP3 findings reached statistical significance. There was also a nonsignificant trend towards higher IL-13 production by T_EM -derived CD103⁺ CD4⁺ T_RM (Fig. 2F).

Figure 2. T_RM generated from distinct circulating T cell subsets differ in their functional characteristics.

(A) Mass cytometry was used to compare the functional characteristics of T_RM generated in vitro from circulating precursor T cell populations. T cells expressing T_RM markers were generated in vitro from each memory subset; CD69⁺ CD103⁻ and CD103⁺ cells from each subset were pooled and analyzed together by mass cytometry. (B) Clustering suggested functional differences existed between T_RM generated from distinct precursors. (C-M) The expression of each cytokine in the CD69⁺ CD103⁻⁻ and CD103⁺ T cell populations for both CD4⁺ and CD8⁺ T cells are shown. The mean and SEM of experiments from three separate donors are shown. Individual tSNE plots and a list of antibodies used are included in Fig. S3 and Table SI. Significance was determined by a one-way ANOVA and Tukey’s post-hoc test for multiple comparisons.

Among CD8+ T_RM, there was significantly higher granzyme B production and a trend towards higher perforin production by T_EM-derived T_RM (Fig. 2G,H). There were no significant differences in the production of IL-2, IL-4, IL-10, IL-22 or TNFα (Fig. 2I–M).

Acquisition of T_RM markers is accompanied by changes in T cell function

The distinct functional characteristics of T_RM derived from different circulating T cell subsets could reflect pre-existing differences between the precursor populations or differences induced by T_RM differentiation. We compared expression of each cytokine or polarization marker by the precursor population (input cells) with that of the post culture cells, comparing it to: i) the cells that did not upregulate T_RM markers, ii) CD69⁺ CD103⁻ T_RM and iii) CD103⁺ T_RM (Fig. 3). Surprisingly, IL-17A production and FOXP3 expression were significantly increased, and IL-4 production was significantly decreased in CD4⁺ T cells that acquired T_RM markers, compared to the input population and cells that did not upregulate T_RM markers (Fig. 3A–C). There was also a trend towards higher IL-22 production in T cells that acquired T_RM markers (Fig. 3D). There were no significant differences in the expression of IL-17A, FOXP3, IL-4 or IL-22 among the different input memory cell populations. The observed changes with T_RM marker acquisition occurred in all of the precursor cell types but were most significant in T_MM (for IL-17A) and T_CM (for FOXP3). Interestingly, these cytokine shifts mirrored differences in cytokine production and FOXP3 expression between T cells isolated from blood and healthy human skin T (Fig. 3A–D third panels). CD8⁺ T cells underwent a marked induction of granzyme B expression with T_RM marker acquisition; this change was observed in all precursor populations but was most significant for T_EM-derived T_RM (Fig. 3E). There was also a trend towards enhanced perforin production with T_RM marker acquisition (Fig. 3F). Again, these shifts mirrored the increased levels of granzyme B and perforin expressed by healthy skin T cells as compared to circulating T cells (Fig. 3E,F, third panel). In contrast, there were no significant changes in the production of IL-13, IFNγ, IL-2 or IL-10 with T_RM marker acquisition (Fig. 3G–J). The increased IFNγ production we observed in T_RM derived from T_EM resulted from a higher level of production by T_EM precursor cells that was maintained after T_RM transition (Fig. 3G).

Figure 3. Acquisition of T_RM markers is accompanied by changes in T cell function.

Mass cytometry was utilized to compare production of cytokines and expression of FOXP3 in freshly isolated precursor T cell populations (Prec), vs. double negative (DN) CD69⁻ CD103⁻ T cells, single positive CD69⁺ CD103⁻ (CD69⁺) and CD103⁺ (CD103⁺) T cells isolated after 1 day bead stimulation and 7 days of culture on keratinocyte monolayers. (A-J) Expression of the indicated cytokines is shown in each population, with comparisons to expression in blood CLA⁺ skin tropic T cells and T cells from healthy skin (A-F). The mean and SEM of mass cytometry results from three separate donors are shown. Significance was determined by a one-way ANOVA and Tukey’s post-hoc test for multiple comparisons (mass cytometry) or by two-tailed T-tests (flow cytometry studies of blood vs. skin).

Central memory T cells are the most efficient T_RM precursors in vivo

To evaluate the ability of human T cell subsets to give rise to T_RM in vivo, we grafted NSG mice with human neonatal foreskin, a tissue that lacks T cells (6), and infused them i.v. with allogeneic PMBC or sorted PBMC T cell subsets from a second human donor (Fig. 4A)(6). Human T cells that migrate into the human skin graft proliferate in response to allogeneic antigen-presenting cells, and a subset of these T cells differentiate into T cells expressing CD69 and/or CD103 (6). The two defining characteristics of resident memory T cells are that they persist within peripheral tissues and are nonrecirculating (2). We isolated T cells from the human skin graft of PBMC-injected mice three weeks and six weeks after PBMC infusion and studied the cells by flow cytometry. Both CD4⁺ and CD8⁺ CD69⁺CD103⁻ and CD69⁺CD103⁺ T cells were present within the graft at three weeks and persisted at similar numbers six weeks after PBMC infusion, demonstrating persistence within the skin (Fig. 4B). To demonstrate a lack of recirculation, we treated skin-grafted mice intraperitoneally with alemtuzumab, an anti-CD52 antibody that depletes T cells in the blood but spares tissue resident memory T cells (5, 6). Alemtuzumab treated mice retained CD69 and CD103 expressing T cells within the graft but cells lacking CD69, including CCR7⁺L-selectin⁺ cells, were depleted (Fig. 4C,D). These data confirm that the cells expressing T_RM markers in our model are true T_RM. We then utilized skin-grafted mice to determine which circulating precursor populations were the most effective at generating T_RM in vivo in human skin. Foreskin-grafted mice were injected intravenously with populations of T_CM, T_MM and T_EM purified by flow sorting and T cells were harvested from the skin graft and studied by flow cytometry at three weeks and six weeks after grafting (Fig. 4E–M). At three weeks after grafting, populations of CD69⁺CD103⁻ and CD69⁺CD103⁺ T cells were present within the graft (Fig. 4E). T_EM had the highest percentage of conversion to CD69⁺ T_RM, analogous to our in vitro study finding (Fig. 4F,G). However, there were significantly more T cells in the skin grafts of T_CM- and T_MM-injected mice, as compared to T_EM-injected mice, as measured by quantitative flow cytometry (Fig. 4H) and histologic counts of T cells in immunostained sections (Fig. 4I). Because of this difference in T cell density in skin, 3 and 6 week grafts from T_CM-injected mice contained significantly more CD4 and CD8 T_RM than mice injected with other precursor populations (Fig. 4J–M).

Figure 4. Central memory T cells are the most efficient T_RM precursors in vivo.

(A) An in vivo model using NSG mice grafted with human foreskin and infused with allogeneic human PBMC was utilized to study T_RM formation in vivo in human skin. (B) T_RM generated in this model were persistent. T cells were isolated from skin grafts three and six weeks after allogeneic PBMC infusion. The numbers of single positive CD69⁺ CD103⁻ T_RM (left panel) and double positive CD69⁺ CD103⁺ T_RM (right panel) are shown for total CD3⁺, CD4⁺ and CD8⁺ T cells. The mean and SEM of 8 (3 weeks) and at least 4 (6 weeks) experiments are shown. (C,D) T_RM generated in this model do not recirculate. Skin-grafted mice were treated i.p. with alemtuzumab or control Ig. (C) Example histograms from control Ig treated (cont Ig, left panel) and alemtuzumab treated mice (αCD52, right panel) are shown. Similar results were obtained in a total of four experiments, using four different blood donors. (D) The number of recirculating (CCR7⁺/L-selectin⁺) and T_RM populations in mice treated with control Ig (cont Ig) and alemtuzumab (αCD52) are shown. The mean and SEM of four experiments using four different blood donors are shown. (E-G) All three memory subsets generated T_RM in vivo. Foreskin-grafted mice were injected i.v. (1 mouse per precursor population) with flow purified populations of T_CM, T_MM and T_EM and T cells were harvested from the skin graft and studied by flow cytometry after three weeks. A representative histogram is shown (E); similar results were obtained in a total of eight experiments, using eight different blood donor and a total of 24 mice. T_EM had the highest percentage of conversion to CD69⁺ T_RM (F,G). Cells were quantified as the number of T cells per 10⁵ total viable cells (of all types) isolated from collagenase digested skin grafts. The mean and SEM of eight experiments are shown, using eight blood donors and 24 mice. (H, I) The total number of T cells in the skin grafts of injected mice are shown, as measured by quantitative flow cytometry (H) and histologic counts of T cells in CD3⁺ immunostained sections (I). The mean and SEM of 8 (H) and 6 (I) experiments are shown. (J-M) T_CM generated the most T_RM in vivo. The numbers of CD69⁺ CD103⁻ (J,K) and CD69⁺ CD103⁺ T_RM (L,M) among CD4⁺ (J,K) and CD8⁺(L,M) T cells are shown. The mean and SEM of experiments from eight donors (3 weeks) and at least four donors (6 weeks) are shown. Significance was determined by a one-way ANOVA and Tukey’s post-hoc test for multiple comparisons.

Diversity and functional characteristics of in vivo generated T_RM

Grafts from T_CM-injected mice contained dense infiltrates of T cells localized to both the lower parts of the epidermis and the dermis six weeks after T cell infusion (Fig. 5A). We carried out high-throughput TCR sequencing to track T cell clones within the grafts. Individual T cell clones identified in three week skin grafts were also identifiable at six weeks, demonstrating that T_CM-injected mice generated a diverse population of persistent T_RM in skin (Fig. 5B,C). T_CM-injected mice had an increased number of unique T cell clones in skin, although this was not statistically significant (Fig. 5D). T cell clonality, a measure of the diversity of the overall T cell repertoire, was similar between groups, demonstrating that the increased number of T_RM in T_CM-injected mice was not the result of clonal expansion of a small number of founder T_RM (Fig. 5E). We carried out NanoString-based gene expression profiling of the skin grafts to determine if the functional differences identified by our in vitro studies were also present in in vivo generated T_RM. T cell normalized expression of TBX21 (T-bet) was significantly higher in T_EM-injected mice (Fig. 5F), analogous to the higher production of IFNγ observed in T_RM generated from T_EM in vitro. FOXP3 expression was highest in skin grafts from T_CM-injected mice, and three IL-17 pathway associated genes (IL17A, RORC and IL23R) were most highly expressed in grafts from T_MM-injected mice, although only IL23R was statistically significant (Fig. 5G,H). There were no significant differences in expression of IL4 and TNF (Fig. 5I,J). In summary, the functional TH1/TC1 dominance of T_EM, TH17/TC17 dominance of T_MM and FOXP3/Treg dominance of T_CM observed in in vitro generated T_RM were detectable to some extent in vivo. Expression of CCR4 and L-selectin were lost on the majority of T cells that acquired T_RM markers (Fig. S2).

Figure 5. Diversity and functional characteristics of in vivo generated T_RM.

(A) T cell infiltration into skin as assessed by CD3 immunostaining of grafts harvested six weeks after T cell injection is shown. Representative images from a total of eight experiments using eight different blood donors are shown. Scale bars = 100μm. (B,C) T_CM-injected mice generated a diverse population of persistent T_RM in skin. High-throughput TCR sequencing (HTS) results from 3 week (B) and 6 week (C) post infusion grafts from a single blood donor are shown. T cell clones identified in three weeks skin grafts were also identifiable at six weeks. (D, E) The number of unique T cell clones (D) and T cell clonality (E) in skin 6 weeks after infusion in T_CM-injected mice is shown. The mean of experiments from two different blood donors are shown. (F-J) NanoString-based gene expression profiling was utilized to study functional characteristics of T_RM generated from each memory subset. Results for type I (F), regulatory T cell (G), type 17 (H), type 2 (I) and TNFα (J) associated genes are shown. The mean and SEM from three experiments using three different blood donors, are shown. Significance was determined by a one-way ANOVA and Tukey’s post-hoc test for multiple comparisons.

T_CM are the most efficient T_RM precursors because they persist longer in the circulation, are skin tropic and replenish other memory T cell subsets

We analyzed memory T cell precursor populations by mass cytometry to determine if the T_EM fraction had characteristics that would make them less likely to enter or survive in skin (Fig. 6A–G). T_CM and T_MM had significantly higher expression levels of cutaneous lymphocyte antigen (CLA), the homing addressin that mediates rolling on E-selectin expressing cutaneous endothelial cells, a necessary step for entry into skin (7) (Fig. 6A). CCR4, a second homing addressin necessary for entry into skin, was expressed at similar levels in all subsets (Fig. 6B). Expression of CCR5, a chemokine receptor important for migration into affected organs in both mouse and human graft versus host disease (GVHD), was actually significantly higher on T_EM than on T_CM (8–10). Levels of CXCR3, a receptor associated with migration into inflamed tissues, initial recruitment of T cells into GVHD sites, and required for differentiation into CD103⁺ T_RM in mice were also higher on T_EM and T_MM than on T_CM (11, 12). These findings agree with prior studies showing that T_EM express higher levels of CXCR3 and CCR5 than T_CM (13–15).

Figure 6. T_CM persist longer in the circulation, are skin tropic and replenish other memory T cell subsets.

(A-G) Circulating memory T cell precursor populations were analyzed by mass cytometry for the expression of cutaneous lymphocyte antigen (A), chemokine receptors (B), HLA-DR (C), BCL-2 (D), exhaustion markers PD-1 and Tim-3 (E,F) and Ki-67 (G). The mean and SEM of results from six different blood donors are shown. (H,I) T_CM persist longer in the circulation. T cells from the blood and spleen of injected mice three weeks after infusion were analyzed by flow cytometry. The mean and SEM of six experiments using six different blood donors are shown. (I) T_CM replenished other T cell subsets. The phenotype of T cells in blood three weeks after T cell infusion is shown. The mean and SEM of experiments using three different blood donors are shown. Significance was determined by a one-way ANOVA and Tukey’s post-hoc test for multiple comparisons.

There was also no significant difference between memory T cell populations in activation status (HLA-DR) or apoptotic marker expression (BCL-2) (Fig. 6C,D). T_EM expressed significantly higher levels of the exhaustion marker PD-1 but there were no differences in Tim-3 or in recent proliferation history (Ki-67, Fig. 6E–G). Taken together, these findings demonstrate that T_CM are more skin tropic and less exhausted than T_EM.

Analyses of the blood and spleen at the end of the three-week experiment demonstrated that T_CM and T_MM also persisted longer in the periphery than T_EM (Fig. 6H). There were 19-fold more human T cells persisting in the circulation of T_CM-injected mice three weeks after injection than in T_EM-injected mice. Analysis of the phenotype of circulating T cells blood three weeks after infusion demonstrated that T_CM were capable of replenishing the circulation with all three memory T cell subsets, T_CM, T_MM and T_EM (Fig. 6I). In contrast, T_EM did not replenish other T cell subsets.

Discussion

The circulating precursor cells that give rise to human resident memory T cells (T_RM) are poorly characterized in humans and only partially characterized in mice (11, 16, 17). We utilized an in vitro differentiation system and human skin-grafted mice to study T_RM generation from circulating human memory T cell subsets.

Our in vitro and in vivo studies demonstrate that all circulating memory T cell subsets tested can differentiate into T_RM after entry into the skin. These results suggest that there is not a single unique precursor cell type for T_RM. Instead, the T_RM differentiation program can be induced in any memory cell that has encountered its cognate antigen and has entered a peripheral tissue such as the skin. However, there were functional biases in T_RM derived from different memory T cell subsets; T_EM, T_MM and T_CM were particularly effective at generating IFNγ, IL-17A and FOXP3⁺ T_RM, respectively. We also observed changes in T cells as they underwent T_RM differentiation and our results suggest that peripheral tissues may have a role in tuning these responses. IL-17A and granzyme B production and FOXP3 expression were increased, and IL-4 production was decreased in skin-generated T_RM in our models; these shifts match the differences observed between circulating and skin T cells from healthy individuals.

Our studies in human skin-grafted mice found that T_CM were the most effective precursor cells of skin resident memory T cells in vivo. Although T_EM had a higher T_RM conversion rate, suggesting, they may represent a state closer to T_RM, T_CM generated vastly increased numbers of T_RM in skin. Circulating T_CM were more skin tropic, less exhausted, persisted longer in the circulation and had the ability to differentiate into T_CM, T_MM and T_EM in vivo. Prior studies in mice and macaques have found that T_CM persist in the circulation and can replenish other cell types, confirming this characteristic is not unique to humans (5, 18). Earlier in vitro studies have also shown that human T_CM can differentiate into other memory T cell types (19).

T_CM have been given little historical credit for their ability to protect peripheral tissues. T_CM were first characterized as recirculating only through the blood and lymph nodes and as having poor effector functions (4). However, more recent studies have demonstrated that T_CM are in fact tropic for multiple peripheral tissues, have impressive effector functions and are found in healthy human peripheral tissues including skin, lung, colon and cervix (20). In healthy human skin, T_CM comprise 10–20% of the total T cell population (6). In addition, the protective role of T_CM is now being acknowledged in mouse models of infection. CD8⁺ T_CM entered skin and generated functional CD69⁺CD103⁺ T_RM following vaccinia infection in mice (21). The ability of T_CM to seed peripheral tissues with highly protective T_RM may explain their known association with long-term immunity to pathogens in both humans and animal models (22).

Most studies of T_RM generation in mice have focused on CD8⁺ T cells, likely because the most common experimental infections used, including vaccinia, HSV and LCMV are viral in nature and tend to generate CD8⁺ centric immune responses. However, the majority of T_RM in healthy human skin are CD4⁺ (6). To our knowledge, these are the first studies of human CD4⁺ T_RM generation. We find that CD4⁺ and CD8⁺ T cells have similar abilities to generate T_RM, although CD8⁺ T cells are more likely to co-express CD103. This is consistent with the role of anti-viral CD8⁺ T_RM, such as those recognizing HSV, to reside in or near the epidermis (23).

In summary, our studies demonstrate that multiple human memory T cell subsets can give rise to T_RM and that functional changes accompany T_RM differentiation. T_CM generate more T_RM than other memory T cell subsets because of their ability to persist and populate the skin in large numbers. These studies demonstrate that T_CM can provide long lasting immunity in both the circulating and tissue compartments, and that generation of T_CM should be a goal of vaccination strategies for infections and cancer.

Materials and Methods:

Study design:

The study was designed to evaluate the ability of human memory T cell subsets to differentiate into T_RM. In vitro cultures and injection into NSG mice grafted with human neonatal foreskin was used. Data was analyzed by flow cytometry, mass cytometry, immunohistochemistry, high-throughput TCR sequencing and measurement of cytokine associated genes by NanoString. The studies were unblinded and not randomized.

Human samples:

All studies were performed in accordance with the Declaration of Helsinki. Blood from healthy individuals was obtained after leukapheresis. Neonatal foreskins were obtained from infants undergoing circumcision at the Brigham and Women’s Hospital. All tissues were collected with previous approval from the Institutional Review Board of the Mass General Brigham Human Research Committee (Mass General Brigham Research Management, IRB protocol numbers: 2018P002169 and 2012P000406).

Flow sorting of purified memory T cell subsets from human peripheral blood:

PBMC were stained with antibodies to CD3 (BioLegend, San Diego, CA), CD62L (BD Biosciences, San Jose, CA), CCR7 (R&D Systems, Minneapolis, MN), and CD45RA (BioLegend) then sorted into T_CM, T_MM and T_EM fractions on a FACS Aria cell sorter (BD Biosciences).

In vitro culture on tissue culture plastic, fibroblasts or keratinocytes:

Primary human keratinocytes and fibroblasts were cultured to confluence in 96-well flat-bottomed plates. T cells from the three sorted memory subtypes were stimulated with αCD3/CD2/CD28 T cell activation beads (Miltenyi Biotec, San Diego, CA) at a 1:10 ratio for 24 hr. Stimulated memory T cells from each subtype were added to designated wells of confluent layers of keratinocytes, fibroblasts, or tissue culture wells without keratinocytes or fibroblasts (plastic) in Iscove’s Modified Dulbecco’s Medium supplemented with 10% FBS and L-glutamine. At indicated time points, T cells were collected and immunostained with directly conjugated anti-human monoclonal antibodies from BioLegend and Life Technologies. Acquisition of flow cytometry samples was performed on a FACS Canto instrument and data were analyzed using FACS Diva software V8.0 (BD Biosciences).

mTOR inhibition:

Primary human keratinocytes were cultured to confluence in 96-well flat-bottomed plates. PBMC were stimulated with αCD3/CD2/CD28 T cell activation beads (Miltenyi Biotec, San Diego, CA) at a 1:10 bead:T cell ratio for 24 hours. PBMC were added to wells of keratinocytes in Iscove’s Modified Dulbecco’s Medium supplemented with 10% FBS and L-glutamine (control medium) or in control medium plus rapamycin (Selleck Chemicals, Houston, TX) at indicated concentrations. Some wells cultures were started in control medium and switched to rapamycin containing medium on day 4 or 8. Cultures were fed with control or rapamycin containing medium every other day for 7 or 14 days. At the end of the culture period, T cells were collected and stained with directly conjugated anti-human monoclonal antibodies from BioLegend and Life Technologies.

Human engrafted mouse model:

Human neonatal foreskins were grafted onto the backs of 6- to 8-week old nonobese diabetic /severe combined immunodeficient/IL-2 receptor γ chain null mice (NSG, Jackson Laboratory, Bar Harbor, ME). One week later, 3 × 10⁶ allogeneic PBMCs, T_CM, T_MM or T_EM were injected intravenously. In indicated experiments, alemtuzumab (2.5 μg) or isotype control (human IgG1, BioLegend) was injected i.p. three times per week for 3 weeks. Three weeks after cell injection, blood, spleen and skin grafts were harvested for analysis. All experiments were approved by the Brigham and Women’s Hospital Institutional Animal Care and Use Committee (IACUC protocol # 2016N000591).

Flow cytometric analysis of T cells from blood, spleen, and human skin grafts:

Human skin grafts were removed from mice and T cells were isolated by collagenase digestion as previously described (24). Spleens were manually dissociated. Splenic cell suspensions and blood were subjected to red blood cell lysis (Red Blood Cell Lysing Buffer, Sigma-Aldrich), filtered through a 40-µm cell strainer, then stained with directly conjugated antibodies and analyzed as described above.

Immunohistochemical studies:

Human CD3 was detected in formalin-fixed, paraffin-embedded (FFPE) tissue sections (5 µm) by standard immunohistochemical techniques with anti-human CD3 (rabbit polyclonal, Dako # A0452, Santa Clara, CA) at a 1:250 dilution and 3,3’-diaminobenzidine (DAB) as the substrate.

CyTOF analyses of memory T cells subsets:

Human PBMC were immunostained with rare earth metal-tagged monoclonal antibodies. Available pre-conjugated antibodies were purchased from Fluidigm (San Francisco, CA) and others were conjugated to metal isotopes using the Maxpar antibody conjugation kit (Fluidigm) according to manufacturer’s instructions. Cells were stained as previously described (6) and analyzed on a CyTOF 2 mass cytometer (Fluidigm) at an event rate of approximately 500 cells/second. To normalize CyTOF data over different days, EQ Four Element Calibration Beads (Fluidigm) were added in all samples. Data were analyzed with Cytobank software (Beckman Coulter, Indianapolis, IN). To remove debris and doublets, single cells were gated based on cell length and DNA content as described by Bendall et al. (25). High dimensional single-cell data produced by mass cytometry were interpreted using the Cytobank viSNE package, allowing visualization of high-dimensional cytometry data on a 2-dimensional map at single-cell resolution preserving the nonlinearity (26). The antibodies used for mass cytometry studies are included in Table S1 and tSNE plots for each marker are included in Figure S3.

High-throughput TCR sequencing:

DNA was extracted from human skin grafts using the QIAamp DNA Mini Kit (Qiagen, Valencia, CA) as per manufacturer’s instructions with overnight tissue digestion. TCRβ CDR3 regions were amplified and sequenced by ImmunoSEQ (Adaptive Biotechnologies, Seattle, WA) from 300 ng DNA template as previously described (27).

Statistics:

Information on the statistical tests used for each set of experiments is included in the figure legends. For comparison of two groups, two-tailed T-tests were used. For comparison of more than two groups, significance was determined by a one-way ANOVA and Tukey’s post-hoc test for multiple comparisons. In all comparisons, α= 0.05.

Data and materials availability:

All data are included in the manuscript or the Supplementary Materials.