Research ArticleT CELLS

Organ-specific isoform selection of fatty acid–binding proteins in tissue-resident lymphocytes

See allHide authors and affiliations

Science Immunology  03 Apr 2020:
Vol. 5, Issue 46, eaay9283
DOI: 10.1126/sciimmunol.aay9283

The skinny on getting fat

When T cells take up residence in a tissue, adaption to the tissue is key for their survival. Here, Frizzell et al. have studied metabolic adaptation of tissue-resident memory T (TRM) cells at three different sites: skin, liver, and the small intestine. They report that TRM cells in each of tissues rely on distinct members of the fatty acid–binding protein (FABP) family of proteins for uptake of fatty acids. By transferring liver-resident TRM cells into naïve mice, they found that FABP expression of these TRM cells is reprogrammed by the tissue they end up seeding in the recipient mice. The studies add to the growing appreciation of immune cells as integral components of tissues they reside in.

Abstract

Tissue-resident memory T (TRM) cells exist throughout the body, where they are poised to mediate local immune responses. Although studies have defined a common mechanism of residency independent of location, there is likely to be a level of specialization that adapts TRM cells to their given tissue of lodgment. It has been shown that TRM cells in the skin rely on the uptake of exogenous fatty acids for their survival and up-regulate fatty acid–binding protein 4 (FABP4) and FABP5 as part of their transcriptional program. However, FABPs exist as a larger family of isoforms, with different members selected in a tissue-specific fashion that is optimized for local fatty acid availability. Here, we show that although TRM cells in a range of tissue widely express FABPs, they are not restricted to FABP4 and FABP5. Instead, TRM cells show varying patterns of isoform usage that are determined by tissue-derived factors. These patterns are malleable because TRM cells relocated to different organs modify their FABP expression in line with their new location. As a consequence, these results argue for tissue-specific overlays to the TRM cell residency program, including FABP expression that is tailored to the particular tissue of TRM cell lodgment.

INTRODUCTION

Memory T cells are generated by infection and provide enhanced immune responses on pathogen reencounter. Whereas memory T cells that circulate via the blood play a key role in the secondary response (1), it has become increasingly clear that T cells found in nonlymphoid tissues are critical for early infection control especially for pathogens that enter the body using these compartments (26). Some of the peripheral memory T cells permanently reside within the tissues, and these are now known as tissue-resident memory T (TRM) cells. TRM cell development involves a cascade of transcriptional changes, which are well defined for the CD8+ subset. These changes involve the engagement of a common gene expression program driven by master regulators such as members of the T-box and positive regulatory domain (PRDM) families and cytokines including transforming growth factor–β (TGF-β) (716). This transcriptional program results in changes critical to T cell residency, such as the down-regulation of various molecules that promote tissue exit including S1P1 and CCR7, as well as up-regulation of adhesion molecules critical for T cell retention and survival (8, 13, 17). Despite some redundancies and variation, this broad transcriptional program is common to TRM cells from a variety of different locations and forms the foundation for tissue residency.

Recently, it has been reported that TRM cells have altered metabolic capabilities compared with circulating T cells (1821). The latter is consistent with their long-term survival in what are often anatomically sequestered and thus potentially metabolically restrictive compartments throughout the body. In particular, Pan et al. (18) have shown that TRM cells in the skin up-regulate fatty acid–binding protein 4 (FABP4) and FABP5 and this alters their fatty acid uptake capabilities and long-term survival. However, FABPs exist as a family of different isoforms that have different fatty acid specificities. Moreover, FABPs exhibit unique patterns of tissue distribution, with FABP4 and FABP5 expressed abundantly in the skin (18) and differential FABP isoform expression in other tissue sites reflecting difference in local fatty acid distribution (22, 23). Thus, at the start of this investigation, it remained unclear whether CD8+ TRM cells in tissues other than skin expressed FABPs and, if so, whether expression was restricted to FABP4 and FABP5 or followed the tissue-specific pattern of isoform selection as seen for parenchymal cells in those locations. Here, we show that FABPs are widely expressed by CD8+ TRM cells found in diverse organs and not by circulating memory T cells. We found that CD8+ TRM cells exhibited differential FABP isoform expression across tissues and that local environmental cues are critical to such FABP selection not only on TRM cells but also more broadly on other populations of tissue-resident immune cells.

RESULTS

Differential expression of FABPs by TRM cells in nonlymphoid organs

It has been shown that skin CD8+ TRM cells generated in response to vaccinia virus infection express high levels of Fabp4 and Fabp5 (18). To determine whether this pattern of FABP expression extended to skin TRM cells generated in response to a different virus, we used a model of herpes simplex virus (HSV) skin scarification (2). To this end, we transferred congenically labeled naïve CD45.1+ gBT-I transgenic CD8+ T cells that are specific for the immunodominant determinant from HSV (gB498–505) into C57BL/6 mice, which were infected with HSV-1 KOS on the skin flank. By day 30 post-infection, the vast majority of CD45.1+ gBT-I T cells in the skin coexpressed the surface molecules CD69 and CD103 that define skin CD8+ TRM cells in mice (Fig. 1A). gBT-I TRM cells were sorted from the skin by flow cytometry, and gene expression was analyzed by quantitative polymerase chain reaction (qPCR). Akin to those skin CD8+ TRM cells induced by vaccinia virus (18), TRM cells generated in response to HSV expressed Fabp4 and Fabp5 but essentially lacked expression of other FABP isoforms (Fig. 1B).

Fig. 1 Organ-specific expression of FABP isoforms in CD8+ TRM cells.

(A and B) Mice received naïve gBT-I.CD45.1 T cells and were subjected to HSV infection. gBT-I T cells were isolated from the skin 30 days post-infection and tracked on the basis of expression of Vα2 and CD45.1. (A) Analysis of CD103 and CD69 expression by gBT-I T cells in skin. Data are representative of three experiments, with n = 3 to 5 mice. (B) Fabp1-9 expression in gBT-I T cells sorted from the skin, as analyzed by qPCR. Data are representative of two independent experiments, with n = 3 to 5 mice each. (C and D) Mice received P14.CD45.1 T cells and were subjected to LCMV infection and treated with DNFB. P14 T cells isolated at 30 days post-infection from the skin, SI-IEL, and liver were tracked on the basis of expression of Vα2 and CD45.1. (C) Analysis of CD103, CD69, and CXCR6 expression by P14 T cells in skin, SI-IEL, and liver. Data are representative of two experiments, with n = 3 to 5 mice each. (D) Fabp1-9 expression of memory P14 T cells sorted from the spleen, skin, SI-IEL, and liver or the spleen of naïve mice for naïve P14 T cells, as analyzed by qPCR. (E) Heatmap of Fabp1-9 expression in gBT-I T cells sorted from the skin and P14 T cells sorted from the skin, SI-IEL, and liver. Heatmaps represent the relative expression of each Fabp isoform, with expression normalized within each sample row. In (A) and (B), numbers represent percentage of cells in the gate. In (B) and (D), bars represent means ± SEM and symbols represent individual mice. *P < 0.05 and **P < 0.01, ANOVA with Dunn’s post-test.

Given that distinct FABP isoforms are expressed by parenchymal cells in different organs (22, 23), we wanted to determine whether this paradigm extended to TRM cells lodged in diverse tissues. To do this, we used a model of lymphocytic choriomeningitis virus (LCMV) infection that results in disseminated CD8+ TRM cell lodgment (4, 24). Naïve congenically labeled CD45.1+ P14 transgenic CD8+ T cells that are specific for LCMV (gp33–41) were transferred into C57BL/6 mice, and mice were subjected to infection with LCMV Armstrong. Given the limited formation of skin TRM cells after systemic LCMV infection, we enhanced skin T cell lodgment via application of the contact sensitizer 2,4-dinitrofluorobenzene (DNFB) to the skin flank 2 days post-infection (25). Gating strategies used for flow cytometric analyses are shown in fig. S1. By day 30 post-infection, CD45.1+ P14 TRM cells that expressed CD69 and CD103 could be detected in the skin (skin TRM) and epithelium of the small intestine (SI-IEL TRM) and P14 TRM cells in the liver that lack CD103 could be identified by dual expression of CD69 and CXCR6 (Fig. 1C), as described (26). In this setting, FABP expression by TRM cells in multiple barrier tissues could be analyzed concurrently in the same infection model. Skin P14 TRM cells expressed Fabp4 and Fabp5, mirroring those established by HSV infection (Fig. 1, D and E). When TRM cells were isolated from the SI-IEL or liver, a different spectrum of FABP expression was observed. Of the nine known FABP isoforms investigated, liver TRM cells highly expressed Fabp1, associated with some expression of Fabp4 but not Fabp5, whereas SI-IEL TRM cells expressed relatively high levels of Fabp1, Fabp2, and Fabp6, with negligible levels of Fabp4 and Fabp5 that are expressed by TRM cells in the skin (Fig. 1, D and E). These FABP gene expression profiles from CD8+ T cells residing in various organs are in line with known general tissue FABP expression patterns (22, 23). In the spleen, antigen-specific CD8+ naïve (TN), central memory (TCM), and effector memory (TEM) cells were found not to express any FABP family members, Fabp1 to Fabp9. In addition, CD69+ antigen–specific memory CD8+ T cells in the spleen (spleen TRM) (27) did not express any FABP genes (Fig. 1D). Combined, the results show that although FABP usage is a general feature of the TRM cell transcriptional program, the FABP isoform selection is dictated by the local tissue environment.

Kinetics of FABP induction on T cells in nonlymphoid tissues

TRM cell differentiation occurs subsequent to T cell entry into nonlymphoid tissues. We have previously shown that, after HSV infection, skin-infiltrating gBT-I T cells up-regulate CD69 early after skin entry, followed by expression of CD103 (8). We wanted to assess the kinetics of FABP up-regulation on tissue T cells during acquisition of the tissue-resident phenotype. To do this, we followed the development of gBT-I TRM cells in the skin after HSV infection and P14 TRM cells in the SI-IEL and liver after LCMV infection. The proportion of T cells expressing a tissue-resident phenotype increased over time, and by day 30 post-infection, >90% of antigen-specific CD8+ T cells in the skin and SI-IEL showed dual expression of CD69 and CD103. Similarly, antigen-specific CD8+ T cells in the liver progressively increased expression of CD69 and CXCR6 after tissue entry (Fig. 2, A and B). We next tracked changes in the expression of FABP genes as T cells progressed along the TRM cell maturation pathway in the skin, SI-IEL, and liver. During the effector phase of infection (7 days post-infection), we did not observe any FABP expression in CD8+ T cells in the spleen (Fig. 2C). We found that tissue-specific FABPs were up-regulated in gBT-I T cells in the skin and P14 T cells in the SI-IEL and liver, with expression following the acquisition of a tissue-resident phenotype and being highest in TRM cell populations at memory time points (30 days post-infection) (Fig. 2, C and D, and fig. S2). In addition, expression kinetics were not uniform among all FABPs expressed by TRM cells in a single given tissue. In SI-IEL TRM cells, we observed a more rapid up-regulation of Fabp1 and Fabp2, whereas Fabp6 expression gradually increased over time (Fig. 2, C and D). Note that the heatmaps in Fig. 2D depict the relative expression of each FABP isoform per row rather than global expression. These data argue that tailored FABP expression does not occur before tissue entry of antigen-specific CD8+ T cells but rather occurs in situ during resident memory T cell development after T cell lodgment in their tissue of residency.

Fig. 2 Kinetics of FABP gene expression during TRM cell development.

Mice received naïve P14.CD45.1 or gBT-I.CD45.1 T cells and were subjected to LCMV or HSV infection, respectively. P14 and gBT-I T cells were gated on Vα2+CD45.1+ cells. (A) Analysis of CD103 and CD69 expression by P14 T cells in spleen and SI-IEL and gBT-I T cells in skin, and CD69 and CXCR6 expression in P14 T cells in liver at 8, 14, and 30 days post-infection. Data are representative of two experiments, with n = 5 mice. (B) Percentage of CD69+CD103+ gBT-I T cells isolated from the skin and P14 cells isolated from the SI-IEL and CD69+CXCR6+ P14 T cells isolated from the liver at 8, 14, and 30 days post-infection. (C) qPCR analysis of FABP gene expression in CD69CD103CXCR6 P14 T cells sorted from the spleen, CD69+CD103 and CD69+CD103+ from SI-IEL, CD69CXCR6 and CD69+CXCR6+ from liver, and CD69+CD103 and CD69+CD103+ gBT-I T cells sorted from total skin or epidermis at 0, 7, 8, 14, and 30 days post-infection (dpi). Data are pooled from four independent experiments, with n = 5 mice each. (D) Heatmap of FABP gene expression for gBT-I T cells isolated from the skin (CD69+CD103 and CD69+CD103+) and P14 T cells isolated from SI-IEL (CD69+CD103 and CD69+CD103+) and liver (CD69CXCR6 and CD69+CXCR6+) at 8, 14, or 30 days post-infection, as shown in (C). Heatmaps represent relative expression of Fabp isoforms within each sample. Symbols indicate means ± SEM.

Loss of FABP1 impairs the establishment of TRM cells in the liver

It has been shown that T cells lacking both Fabp4 and Fabp5 are greatly impaired in their ability to form long-lasting TRM cells in skin (18). Given the spectrum of FABP expression between different tissues, we wanted to determine whether the loss of an alternative FABP would affect TRM cells in a similar fashion. TRM cells in the SI-IEL and skin express multiple FABP isoforms, and thus, the impact of losing expression of one FABP may be compensated by other coexpressed isoforms. To this point, Pan et al. observed a loss of skin T cell survival in the absence of both Fabp4 and Fabp5 but not cells lacking one of these isoforms (18). In contrast, FABP1 is the dominant FABP expressed by liver TRM cells to such an extent that T cells lacking this isoform might be impaired to form TRM cells in tissue. Using mixed bone marrow (BM) chimeras of wild-type (WT) and FABP1-deficient (Fabp1−/−) cells that were infected with LCMV (Fig. 3A), we found such a defect of LCMV-specific (gp33+) TRM cells in the liver but not in populations recovered from the spleen or SI-IEL after 14 days (Fig. 3B). In separate experiments, WT and Fabp1−/− CD8+ T cells were in vitro activated and transferred into mice. Such effector cell transfers intrinsically result in liver TRM cell formation (28), and to lodge skin TRM cells in the same mice, we treated recipients on the flank with DNFB (Fig. 3C). Consistent with our observations after LCMV infection, we found a selective loss of TRM cells in the liver but not in the spleen or skin (Fig. 3, D and E). This loss of FABP1-deficient liver TRM cells was not absolute potentially due to sustained FABP4 expression in these cells (fig. S3). FABP4 may play a compensatory role in the absence of FABP1 due to functional overlap between FABP isoforms. Last, this reduction of FABP1-deficient cells could be restored upon reexpression of FABP1 (Fig. 3F). In these experiments, WT and Fabp1−/− effector CD8+ T cells were transduced with either “empty” (EV) or FABP1-expressing (FABP1+) retrovirus and cells transferred into mice. Forced expression of FABP1 in FABP1-deficient cells restored the loss of TRM cells in the liver but had no effect on splenic T cell populations (Fig. 3F). Combined, these results demonstrate the selective reliance of TRM cells on FABP expression that mirrors the dominant tissue-specific isoform usage.

Fig. 3 FABP1-deficient T cells show impaired TRM cell generation in the liver.

(A and B) Naïve CD45.1+ mice were irradiated and received BM cells from naïve WT (CD45.1.2+) and Fabp1−/− (CD45.2+) at a 1:1 ratio. Recipient mice were infected with LCMV, and tissues were harvested 14 days post-infection. LCMV-specific cells were tracked using tetramer. (A) Experimental schematic. (B) Quantification of WT and Fabp1−/− CD8+ gp33+ T cells in the spleen, and gp33+ TRM cells from the spleen (CD69+), liver (CD69+CXCR6+), and SI-IEL (CD69+CD103+). Data are pooled from three independent experiments, with n = 3 to 5 mice per group. (C to E) Naïve CD45.1+ mice received in vitro activated WT and Fabp1−/− CD8+ T cells and were treated with DNFB on the skin flank. Cells were isolated 14 days post-transfer. (C) Experimental schematic. (D) Quantification of transferred WT and Fabp1−/− CD8+ T cells in the spleen, and TRM cells from the spleen (CD69+), liver (CD69+CXCR6+), and skin (CD69+CD103+). (E) Ratio of transferred WT and Fabp1−/− CD8+ T cells in the spleen and liver 14 days post-transfer. Data are pooled from three independent experiments, with n = 3 to 5 mice per group. (F) In vitro activated CD8+ T cells from WT (CD45.1.2+) and Fabp1−/− (CD45.2+) mice were transduced with either Fabp1-Ametrine (Fabp1+) or Ametrine-empty (EV) retrovirus. WT and Fabp1−/− cells transduced with EV were mixed at a 1:1 ratio, or WT and Fabp1−/− cells transduced with Fabp1+ were mixed at a 1:1 ratio. Cells were transferred into naïve CD45.1+ mice. The ratio of transferred WT and Fabp1−/− cells in the spleen and liver 14 days post-transfer is shown. The ratio of Fabp1−/− to WT CD8+ or TRM cell numbers in the spleen and liver was calculated for each individual mouse. n = 10 mice. In (B) and (D) to (F), bars represent the mean and symbols represent individual mice. *P < 0.05, **P < 0.01, and ****P < 0.0001, Student’s t test. ns, not significant.

FABP expression profiles are shared by diverse immune cells in the same microenvironment

We next asked whether other immune cells within a given tissue also display a similar selective pattern of FABP expression to those we observed for CD8+ TRM cells. Tissue residency is not just a feature of CD8+ T cells but has also been described for innate lymphocytes including natural killer T (NKT) cells and innate lymphoid cells (ILCs). As a group, these tissue-resident lymphocytes share common transcriptional features that promote their long-term residence (7, 29, 30). We examined FABP gene expression in endogenous immune cell populations, both tissue-resident and migratory, from the skin, SI-IEL, and liver of naïve mice (fig. S1, F to H). In the skin, type 1 ILC (ILC1), dendritic epidermal T cells (DETCs), and Langerhans cells (LCs) expressed high levels of Fabp4 and showed some expression of Fabp5 when compared with naïve T cells from the spleen. In the SI-IEL, all analyzed subsets of lymphocytes isolated from the intraepithelial layer (CD8ααTCRβ, CD8αβTCRβ, CD8ααTCRγδ, and CD8αβTCRγδ) and ILC1 expressed Fabp1, Fabp2, and Fabp6 analogous to SI-IEL TRM cells (Fig. 4, A and B). In the liver, ILC1 expressed Fabp1 and some Fabp4, and NKT cells expressed Fabp1 and Fabp5 (Fig. 4, A and B). Tissue-specific protein expression of FABP isoforms was also verified by Western blot. We observed protein expression of FABP1 in liver-resident lymphocytes (TRM, NK, and NKT cells); expression of FABP4 and FABP5 by skin-resident cells (DETC); and protein expression of FABP1, FABP2, and FABP6 by CD8+ T cells from the SI-IEL (fig. S4). Overall, we found that TRM cells exhibited a similar pattern of FABP expression to that of other immune populations from the same tissue.

Fig. 4 FABP expression by TRM cells mirrors that of endogenous immune cells residing in the same organ.

(A) Fabp1-9 expression in ILC1, LC, and DETC from the skin; ILC1, CD8ααTCRβ, CD8αβTCRβ, CD8ααTCRγδ, and CD8αβTCRγδ cells from the SI-IEL; and ILC1, cNK, and NKT cells from the liver sorted from naïve mice, as analyzed by qPCR. Data are from four independent experiments, with five mice pooled each. (B) Heatmap of Fabp1-9 expression in endogenous populations of naïve mice. Heatmaps represent relative expression of Fabp isoforms within each sample. Data are from three independent experiments, with five mice pooled each. Bars represent means ± SEM. *P < 0.05 and **P < 0.01, ANOVA with Dunn’s post-test.

Tissue-specific cues drive FABP gene expression

Our findings that specific complements of FABP expression are induced subsequent to tissue entry and are shared by other lymphocytes within the same tissue indicated that FABP expression is likely controlled by factors present in the local microenvironment. To test this, we examined induction of FABPs on effector CD8+ T cells (which demonstrated negligible FABP expression) after culture with supernatant obtained from various tissues. To generate these supernatants, we harvested the skin, liver, and spleens from naïve mice and processed these tissues into single-cell suspensions that were then cultured overnight. The following day, the supernatant from each organ culture was incubated with in vitro activated OT-I T cells for 24 hours, after which these effector CD8+ T cells were assessed for changes in FABP gene expression (Fig. 5A). The supernatant harvested from liver tissue homogenates was sufficient to induce the expression of Fabp1 in effector CD8+ T cells, matching the expression pattern observed in CD8+ T cells residing in this organ (Fig. 5B). Moreover, the supernatant obtained from skin tissue induced Fabp4 expression in effector CD8+ T cells (Fig. 5B). In contrast, FABP up-regulation was not observed for CD8+ T cells cultured with supernatant from the spleen or in control samples where T cells were not exposed to any tissue supernatant (Fig. 5B). Thus, tissue-derived factors are sufficient to induce FABP expression in CD8+ T cells, and they can dictate the FABP isoform usage in a tissue-specific fashion.

Fig. 5 Organ supernatant induces tissue-specific FABP expression in CD8+ T cells.

Liver, spleen, and skin from naïve C57BL/6 mice were harvested and processed to single-cell suspensions. Cells from each tissue were cultured overnight in complete RPMI media. Supernatant was collected and cultured with in vitro activated OT-I T cells for 24 hours, which were then collected for qPCR. (A) Experimental schematic. (B) Fabp1-9 expression in OT-I T cells cultured with RPMI, spleen, liver, or skin supernatant for 24 hours. Data are from four independent experiments, with two replicates each. Bars represent means ± SEM, and symbols represent individual replicates. ****P < 0.0001, ANOVA with Dunn’s post-test.

Dynamic expression of FABP genes by CD8+ TRM cells

Tissue-resident cells show little evidence of turnover or recirculation under steady-state conditions (3133). Yet, there is emerging evidence that, after particular local challenges, skin TRM cells can demonstrate functional plasticity and reprogramming (34) or can exit the tissue and give rise to new TRM cell populations in downstream secondary lymphoid organs (35). This infers that although TRM cells adapt to their site of residence, they may exhibit transcriptional flexibility under changing environmental circumstances. We sought to determine whether established TRM cell populations could reprogram their FABP expression profiles if forced to adapt to a different tissue milieu. To address this, we isolated mature populations of TRM cells from the liver and transferred these cells into new recipient animals. CD69+CXCR6+ P14 TRM cells were sorted from the liver of LCMV immune mice (>30 days post-infection), and liver TRM cells were transferred intravenously into naïve animals that were subsequently infected with LCMV. Recipient mice were treated with the contact sensitizer DNFB on the skin flank to nonspecifically recruit LCMV-primed P14 T cells into the skin (Fig. 6A) (25). We found that the progeny of transferred liver TRM cells (ex-liver TRM cells) could be recovered from tissues including the spleen, skin, and SI-IEL 7 days after TRM cell transfer (Fig. 6B). Upon migration to new microenvironments, ex-liver TRM cells changed their FABP expression profile consistent with their new tissue of lodgment. Ex-liver TRM cells recovered from the skin 7 and 30 days post-transfer had down-regulated Fabp1, which was expressed by isolated liver TRM cells before transfer, and up-regulated the FABP associated with skin residency, namely, Fabp4 (Fig. 6C). Similar site-specific changes in FABP expression were also observed in the progeny of ex-liver TRM cells that migrated to the intestinal compartment, as P14 T cells isolated from the SI-IEL showed enhanced expression of Fabp1, Fabp2, and Fabp6 30 days post-transfer (Fig. 6C). These results indicated that TRM cells can exhibit plasticity and adapt their FABP expression to suit new microenvironments.

Fig. 6 TRM cells modulate their FABP expression upon relocation to new microenvironments.

(A to C) Naïve mice received P14.CD45.1 naïve T cells and were infected with LCMV. At 30 days post-infection, liver TRM cells (CD69+CXCR6+) were sorted and transferred intravenously (i.v.) into naïve recipients, followed by LCMV infection and treatment with DNFB in the skin. At 7 and 30 days post-transfer, ex-liver TRM cell P14 progeny was sorted from the spleen, skin, SI-IEL, and liver for FABP expression assessment by qPCR. (A) Experimental schematic. (B) Ex-liver TRM cell progeny was tracked on the basis of expression of Vα2 and CD45.1 at 7 days post-infection from spleen, skin, SI-IEL, and liver. Data are representative of three independent experiments, with n = 5 mice. (C) Fabp1, Fabp2, Fabp4, and Fabp6 expression analyzed by qPCR in ex-liver TRM cells isolated from spleen, skin, SI-IEL, and liver at 7 and 30 days post-infection of recipient mice. Data are pooled from three independent experiments, with five mice each. (D and E) Effector OT-I T cells were transferred i.v. into recipient mice for TRM cell seeding. At 30 days post-transfer, liver OT-I TRM cells were sorted and placed in the skin epidermis of naïve mice by epicutaneous transfer (skin tx). Seven days post-transfer, ex-liver TRM cells were sorted from the skin and FABP gene expression was assessed by qPCR. (D) Experimental schematic. (E) Fabp1-9 expression of transferred liver TRM cells or ex-liver TRM cells sorted from the skin 7 days post-transfer, as analyzed by qPCR. Data are from three independent experiments, with five mice pooled each. Bars represent means ± SEM. Symbols indicate mean. *P < 0.05 and ***P < 0.001, two-way ANOVA with Bonferroni’s post-test.

Given that these experiments were performed in the context of infection with potential antigen restimulation of transferred P14 T cells, whether TRM cell reactivation was required to adapt FABP expression was left unclear. To directly assess whether microenvironmental signals were sufficient for resting TRM cells to change their FABP expression, we performed experiments whereby we directly transferred TRM cells from one environment to another in the absence of any overt in vivo antigen stimulation. In vitro activated OT-I T cells were transferred into mice to generate liver TRM cells (28), and 30 days post-transfer, CD69+CXCR6+ OT-I TRM cells were sorted from the liver and directly placed into the skin epidermis of naïve animals by epicutaneous transfer. In this case, we used a suspension of liver TRM cells in Matrigel Basement Membrane Matrix solution that was administered directly to abraded skin, which permits T cell entry directly into the underlying epidermis (36). Seven days post-transfer, the skin was harvested and ex-liver OT-I TRM cells were sorted for gene expression analysis by qPCR (Fig. 6D). We found that ex-liver OT-I TRM cells placed in the skin mirrored the FABP expression of skin-resident memory T cells by down-regulating Fabp1 (expressed by liver OT-I TRM cells before transfer) and up-regulating Fabp4 and to some extent Fabp5 (Fig. 6E). These results demonstrate that TRM cells can adapt FABP gene expression upon organ repositioning and emphasize that the local microenvironment shapes FABP isoform selection within tissues.

DISCUSSION

Memory T cells have long been described in terms of their ability to migrate throughout the body as a surveillance mechanism against infection. However, TRM cells exist as a sessile population that is particularly effective in control of tumors or pathogens in nonlymphoid environments (37). These are often sequestered locations that are distinct and highly diverse in structure, function, and general accessibility. Thus, whereas certain elements would be common to all tissue-resident lymphocytes, such as the shutdown of migration-promoting mechanisms, other elements would be expressed or modified in a way that optimizes cell lodgment, function, and survival in a particular location.

It is known that the different FABP isoforms have unique lipid specificity (38) and their expression is adapted to the available resources found in distinct locations throughout the body (22, 23). Although it was previously shown that FABP4 and FABP5 are important for TRM cell persistence in the skin (18), it remained unclear whether this was a general property of TRM cells distributed throughout the body. Our results show that although FABP expression is a general feature of TRM cells in a variety of different tissues, such expression is not limited to FABP4 and FABP5. Moreover, our results using FABP1 deletion in liver TRM cells, when combined with those of Pan et al. (18) examining the effects of FABP4 and FABP5 elimination in the skin, argue that selective FABP expression is necessary for TRM cell survival in the nonlymphoid compartment.

Diverse patterns of FABP expression occur in TRM cells, which is linked to location of T cell lodgment and determined by factors present in the tissue. Furthermore, selective FABP isoform expression is seen in other tissue-resident leukocytes such as ILC and tissue-resident myeloid cells, although some variation is seen such as the codominant expression of FABP5 only in liver NKT cells. Consequently, it appears that although FABP expression represents a common element in TRM cell development, the specific form of its expression is inextricably tied to anatomical location within the body. Such specialization makes sense because FABP selection is likely optimized to the milieu in which cells reside, regardless of whether they are lymphoid, myeloid, or parenchymal in origin.

The transcriptional program responsible for tissue residency has been described in much detail over recent years. This includes expression of master transcription factors such as Blimp-1 or its homolog Hobit that together modulate elements such as S1P1 and CCR7, thus inhibiting T cell egress from the tissues (7). Here, we find that factors associated with the tissue microenvironment overlay this fundamental control of tissue residency, which, in this case, fine-tune FABP usage to best fit the given site of T cell lodgment. It may be that the general tissue residency program initiates FABP expression as TRM cells mature in nonlymphoid tissues because, as we show here, this family of molecules is expressed by TRM cells in a variety of different locations. Alternatively, because our results also show that factors unique to each tissue appear to induce the appropriate pattern of FABP selection, their expression may be dependent on mediators within the local environment rather than the residency program per se. Regardless, our results show that TRM cells and other noncirculating lymphocytes are adapted to their particular tissue of residence, which is proposed to optimize uptake of fatty acids. This adaption likely extends to other tissue-specific elements including some critical to TRM cell retention or survival in a manner that is specific for their location of residence.

MATERIALS AND METHODS

Mice, adoptive transfer, and infection

Six- to 10-week-old female C57BL/6, B6.SJL-PtprcaPep3b/BoyJ (CD45.1), C57BL/6xB6.SJL-PtprcaPep3b/BoyJ (CD45.1.2), gBT-I.CD45.1, OT-I.CD45.1, and P14.CD45.1 mice were bred in the Department of Microbiology and Immunology, University of Melbourne. All animal experiments were approved by the University of Melbourne Animal Ethics Committee. Naïve gBT-I T cells isolated from lymph nodes were transferred intravenously at 5 × 104 cells per recipient mouse followed by infection by skin scarification using 1 × 106 plaque-forming units (PFU) of HSV-1 KOS, as described (2). Naïve P14 T cells were transferred intravenously at 5 × 104 cells per recipient followed by intraperitoneal infection with 2 × 105 PFU LCMV Armstrong. In some experiments, mice were shaved and depilated before treatment with 10 μl of DNFB (Sigma-Aldrich) diluted at 0.25% in acetone:oil (4:1) on the skin 2 days after LCMV infection (25). For generation of mixed BM chimeric mice, CD45.1+ mice were irradiated (2 × 5.5 Gy) and reconstituted by intravenous transfer of 2 × 107 BM cells from WT (CD45.1.2+) or FABP1−/− (CD45.2+) mice at the ratio of 1:1. Ratio between donor compartments was assessed within the blood of recipients 8 to 12 weeks after reconstitution before infection. In adoptive transfer experiments of in vitro activated effector CD8+ WT (CD45.1.2+) and FABP1−/− (CD45.1.2+) cells, recipient CD45.1+ mice received 1 × 106 cells intravenously followed by skin DNFB treatment.

Organ processing, flow cytometry, and antibodies

Spleens were processed through metal meshes into single-cell suspensions. Skin tissue was incubated at 37°C for 90 min in dispase (2.5 mg/ml; Roche) followed by separation of the dermis and epidermis. Dermis and epidermis were chopped into small fragments and incubated at 37°C for 30 min in collagenase III (3 mg/ml; Worthington). Livers were processed through 70-μm meshes, and pellets were resuspended in 35% isotonic Percoll (Sigma-Aldrich) followed by centrifugation at 500g for 20 min at room temperature before red blood cell lysis was performed. Small intestines were removed of luminal contents and Peyer’s patches, cut into 1-cm fragments, and incubated at 37°C for 30 min with lateral rotation (230 rpm) in dithioerythritol (15 mg/ml; Sigma-Aldrich) in 10% Hanks’ balanced salt solution/Hepes. Cells were further separated using a 44/67% Percoll density gradient. Isolated cells were stained with the following antibodies and/or tetramers for flow cytometry or cell sorting: anti-CD45.1 (A20), anti-CD8α (53-6.7), anti-CD8β (YTS1 56.7.7), anti-CD3 (500A2), anti-Vα2 (B20.1), anti-CD44 (IM7), anti-CD45 (30-F11), anti-CD62L (MEL-14), anti-CD69 (H1.2F3), anti-CD103 (2E7), anti-CXCR6 (SA051D1), anti-Vγ3 (536), anti-CD11b (M1/70), anti-MHCII (M5.114.15.2), anti-EpCAM (G8.8), anti-NK1.1 (PK136), anti-NKp46 (29A1.4), anti-CD49a (HMa1), anti-CD49b (DX5), anti-TCRβ (H57-597), anti-TCRγδ (B1), anti-CX3CR1 (SA011F11), and anti-KLRG1 (2F1) from BD Biosciences or eBioscience. PBS57-loaded CD1d and H-2Db/GP33 MHC I tetramers were obtained from the tetramer core facility of the National Institutes of Health (NIH). Flow cytometry was performed on an LSRFortessa and analyzed with FlowJo software (TreeStar).

Cell sorting and quantitative real-time qPCR

Lymphocytes were isolated from tissues and sorted into the following subsets using a FACSAria III (BD Biosciences): spleen TN (CD45.1+CD8α+Vα2+CD44CD62L+CD69), spleen TCM (CD45.1+CD8α+Vα2+CD44+CD62L+CD69), spleen TEM (CD45.1+CD8α+Vα2+CD44+CD62LCD69), spleen TRM (CD45.1+CD8α+Vα2+CD44+CD62LCD69+), skin TRM (CD45.1+CD3+Vα2+CD69+CD103+), skin ILC1 (CD3NK1.1+NKp46+CD49a+), skin DETC (CD3+Vγ3+), skin LC (CD3EpCAM+MHCII+CD11b+), liver TRM (CD45.1+CD8α+Vα2+CD69+CXCR6+CD62L), liver ILC1 (CD3NK1.1+NKp46+CD49a+), liver conventional NK (cNK) (CD3NK1.1+NKp46+CD49aCD49b+), liver NKT (CD3+CD1d-PBS57tetramer+), SI-IEL TRM (P14 CD45.1+CD8α+Vα2+CD69+CD103+), SI-IEL ILC1 (CD3NK1.1+NKp46+CD49a+), SI-IEL TCRγδCD8αβ (CD69+CD103+TCRγδ+TCRβCD8α+CD8β+), SI-IEL TCRγδCD8αα (CD69+CD103+TCRγδ+ TCRβCD8α+CD8β), SI-IEL TCRβCD8αα (CD69+CD103+TCRγδTCRβ+CD8α+CD8β), and SI-IEL TCRβCD8αβ (CD69+CD103+TCRγδTCRβ+CD8α+CD8β+). RNA was isolated from sorted samples using an RNeasy Plus Micro Kit (Qiagen) and converted to complementary DNA (cDNA) using the High Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions using an MJ Mini thermal cycler (Bio-Rad). Genes of interest were preamplified from cDNA using TaqMan PreAmp Master Mix (Thermo Fisher Scientific) on an MJ Mini thermal cycler (Bio-Rad), and samples were analyzed by real-time qPCR using the StepOnePlus Real-Time PCR System (Thermo Fisher Scientific). TaqMan probes (Thermo Fisher Scientific) were used for all genes: Tbp Mm00446973_m1, Hprt Mm00446968_m1, Fabp1 Mm00444340_m1, Fabp2 Mm00433188_m1, Fabp3 Mm02342495_m1, Fabp4 Mm00445878_m1, Fabp5 Mm00783731_s1, Fabp6 Mm00434315_m1, Fabp7 Mm00445225_m1, Pmp2 Mm03015239_m1, and Fabp9 Mm0196433_s16. Cycle-threshold values were determined for genes individually, and gene expression was normalized to the housekeeping genes Tbp or HprtCt) and presented as 2−ΔCt [arbitrary units (AU)]. In heatmaps, Fabp isoform relative expression was normalized per row.

Transduction of CD8+ T cells

For transfections, 293T cells were seeded into 96-mm dishes at a density of 5 × 106 cells 1 day before transfection. Cells were transfected with empty- or FABP1-pMSCV-Ametrine–encoding vector using the CalPhos Mammalian Transfection Kit (Takara). Viral supernatant was harvested after 48 hours by filtration (0.22 μm; Millipore), and 0.5 ml of fresh filtered virus supernatant was transferred to 24-well plates coated with RetroNectin (20 μg/ml; Takara). Naïve CD8+ T cells from spleen and lymph nodes from naïve C57BL/6 (CD45.1.2+) or Fabp1−/− (CD45.2+) mice were enriched, and 1 × 106 cells were plated in 24-well plates (Thermo Fisher Scientific) precoated with anti-CD3 (clone 2C11) and anti-CD28 (37.51) (5 μg/ml; eBioscience). In vitro activated CD8+ T cells were transferred to 24-well plates containing viral supernatant and expanded for a further 3 days in the presence of IL-2 (25 U/ml; PeproTech). Transduction efficacy was determined by Ametrine expression, and 2.5 × 105 transduced WT and Fabp1−/− cells were mixed 1:1 and transferred intravenously into C57BL/6 (CD45.1+) mice.

In vitro T cell activation and cell culture with organ supernatant

Naïve OT-I T cells isolated from lymph nodes were in vitro activated by OVA257–264–pulsed target cells in the presence of IL-2 (25 U/ml; PeproTech), as described (28). Effector cells were transferred into mice (1 × 106 cells injected intravenously), or OT-I T cells were plated at 1 × 106 per well in 24-well plates and cultured with tissue supernatant. Tissue supernatants were generated from cell suspensions of spleen, skin, or liver of naïve mice that were cultured in complete RPMI overnight at 37°C, 5% CO2. The supernatant was then collected and spun down at 2000 rpm for 10 min to remove cells. Supernatant was added to in vitro activated OT-I T cells and incubated for 24 hours. Cells were then collected for qPCR analysis.

Intravenous and epicutaneous liver TRM cell transfer

Established populations of liver TRM cells were sorted from LCMV-infected mice (30 days post-infection), and 1 × 105 cells were transferred intravenously into naïve recipient mice. Recipient mice were infected with LCMV, and each side of the ear was treated with 10 μl of DNFB (0.25%) diluted in acetone/oil (4:1) to recruit transferred cells into the skin. Seven or 30 days later, transferred cells were sorted from various organs for qPCR analysis. For epicutaneous transfer experiments, liver TRM cells from mice transferred with effector OT-I cells were sorted and transferred epicutaneously into naïve recipient mice, as described (36). Briefly, recipients were anesthetized, and the skin flank was shaved and depilated followed by abrasion with a power tool with grindstone attachment (Dremel) for 15 s with constant rotation. Liver TRM cells were suspended in Matrigel basement membrane matrix (Corning) at 4 × 105 in 10 μl and applied to the scarified region. Mice were rested for 10 min to allow Matrigel solidification. The area was covered with Opsite Flexigrid (Smith and Nephew), and mice were bandaged for 4 days.

Statistical analysis

Statistical analyses were performed by one- or two-way analysis of variance (ANOVA) test followed by Bonferroni’s post-test, by nonparametric ANOVA with Dunn’s post-test, or by Student’s t test when indicated in Prism 7 (GraphPad). P values were represented by *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Results represent means ± SEM.

SUPPLEMENTARY MATERIALS

immunology.sciencemag.org/cgi/content/full/5/46/eaay9283/DC1

Materials and Methods

Fig. S1. Representative flow cytometric cell sorting gating strategies.

Fig. S2. Kinetics of FABP gene expression during TRM cell development.

Fig. S3. FABP gene expression in FABP1-deficient CD8+ TRM cells.

Fig. S4. FABP protein expression in endogenous populations and liver TRM cells.

Table S1. Raw data.

REFERENCES AND NOTES

Funding: This work was supported by a Howard Hughes Medical Institute and Bill & Melinda Gates International Research Scholarship (OPP1175796) to L.K.M. and National Health and Medical Research Council (NHMRC; APP1129711) to L.K.M. H.F. is supported by an NSF Graduate Research Fellowship and was supported by an NSF Graduate Research Opportunities Worldwide Fellowship. N.G.Z. is supported by FAPESP BEPE Scholarship (2019/12431-2). S.L.P. is supported by a Cancer Council Victoria Postdoctoral Fellowship. H.M. is supported by an Australian Research Council (ARC) Discovery Early Career Researcher Award (DE170100575). J.A.V. is supported by an NHMRC Principal Research Fellowship (1154502) and Program Grant (1113293). L.K.M. is a Senior Medical Research Fellow supported by the Sylvia and Charles Viertel Charitable Foundation. Author contributions: H.F., R.F., D.F., S.N.C., M.E., N.G.Z., B.v.S., S.C.-G., and S.L.P. performed experiments. H.F., R.F., F.R.C., and L.K.M. designed experiments. H.F. and R.F. analyzed data. J.A.V. and H.E.G.M. provided supervision. H.F., R.F., F.R.C., and L.K.M. prepared the manuscript. L.K.M. provided funding and led the research program. Competing interests: The authors declare that they have no competing financial interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Immunology

Navigate This Article