CXCR6 promotes dermal CD8⁺ T cell survival and transition to long-term tissue residence

Abstract

Tissue resident memory T cells (T_RM) provide protection against local re-infection, and yet the interstitial signals that govern their formation and persistence remain poorly defined. Here, we show that antigen-dependent induction of the chemokine receptor CXCR6, is a conserved adaptation to peripheral tissue infiltration that promotes T_RM formation after viral infection. Deficient T_RM formation in the absence of CXCR6 was not explained by trafficking as CXCR6 was not required for tissue entry, was dispensable for the early accumulation of antigen-specific CD8⁺ T cells in skin, and did not restrain their exit. Single cell sequencing indicated that Cxcr6^−/− CD8⁺ T cells were competent to acquire a transcriptional program of residence and T_RM that formed were equally functional compared to their WT counterparts when reactivated greater than 100 days post primary infection. The reduction in Cxcr6^−/− CD8⁺ T cells at memory time points, was associated with impaired redox homeostasis, antioxidant capacity, and increased rates of apoptosis in the dermis during the transition from effector to resident memory. Thus, CXCR6 promotes the adaptation of T cells as they engage antigen in tissue to increase the probability of survival, memory differentiation, and long-term residence.

Introduction

Resident memory T cells (T_RM) provide rapid, localized protection due to their position in barrier tissues. T_RM formation depends first upon the recruitment of effector T cells to sites of inflammation where local cues then direct the acquisition of a transcriptional program required for tissue adaptation and persistence. One key feature of this program is the downregulation of a transcriptional module associated with tissue exit via the lymphatic vasculature, including Klf2, Ccr7, and S1pr1, which presumably helps fix T_RM in a sedentary state. Consistent with this idea, Ccr7 deficiency prevents CD4⁺ T cell egress from the skin via dermal lymphatic vessels following viral infection¹ and boosts CD8⁺ T_RM formation.² Similarly, constitutive expression of S1pr1 limits CD8⁺ T_RM formation in skin³ and the conserved T_RM marker, CD69, is thought to promote residency by sequestering S1PR1 and inhibiting T cell egress.⁴ While these mechanisms may limit the exit of T_RM progenitors from inflamed tissues, the signals that instruct T cell position within tissue along the T_RM differentiation trajectory, and the interstitial cues that switch between these migratory modules, remain incompletely understood.

Interestingly, one of the highly expressed and conserved chemokine receptors in both CD4⁺ and CD8⁺ T_RM across mice and humans is CXCR6.^5–7 The sole ligand for CXCR6, CXCL16,⁸ is expressed by multiple cell types including macrophages, dendritic cells, keratinocytes, and endothelial cells.^9–12 Induced by tumor necrosis factor (TNF)-α and interferon (IFN)-γ, CXCL16 is expressed in a transmembrane form and acts in trans at the cell-cell interface, but may also be cleaved from the cell surface to signal as a classical chemoattractant.¹³ CXCR6 loss impairs T_RM formation or maintenance in lung,¹⁴ skin,¹⁵ liver,¹⁶ and brain;¹⁷ however, the specific mechanism by which CXCR6 guides CD8⁺ T_RM formation remains incompletely understood. In a mouse model of influenza infection, CXCR6 is dispensable for CD8⁺ T cell trafficking to the lung parenchyma but needed for localization to the CXCL16-expressing lung epithelium.¹⁴ In the liver, however, CXCR6 has been shown to be necessary for tissue infiltration during a graft vs host response¹⁸ but also dispensable for early accumulation following malaria infection.¹⁶ Therefore, across peripheral, non-lymphoid tissues, when, exactly where, and how CXCR6 promotes CD8⁺ T cell accumulation remains unclear.

Here we used cutaneous, murine viral infection to define the kinetics of CXCR6 expression and its impact on the transition of effector CD8⁺ T cells to long-term tissue residence. We find that circulating CD8⁺ T cells within non-lymphoid peripheral tissues are distinguished from neighboring CD8⁺ T_RM in part by CXCR6 expression. CXCR6 expression was conserved across anatomical compartments where it was necessary for optimal CD8⁺ T_RM formation. In skin, rather than facilitating early CD8⁺ T cell accumulation or directly preventing egress, we demonstrate that local antigen encounter upregulates CXCR6 to promote CD8⁺ T cell survival allowing for more efficient seeding of the epidermis at memory. In the absence of CXCR6, increased rates of apoptosis were associated with reduced transcriptional programs of redox homeostasis, and the accumulation of mitochondrial reactive oxygen species (ROS) by CD8⁺ T cells in the dermis. These data add to a model for how CD8⁺ T cell positioning in tissues helps orchestrate T_RM formation and implicates CXCR6 as a conserved feature of survival and tissue adaptation.

Materials and methods

Mice

C57BL/6J, B6.SJL-PtprcaPepcb/BoyJ (CD45.1), C57BL/6-Tg(TcraTcrb)1100Mjb/J (OT-1), B6.129P2-Cxcr6tm1Litt/J (Cxcr6^−/−),¹⁹ B6.PL-Thy1a/CyJ (CD90.1), C57BL/6-Tg(Nr4a1-EGFP/cre)820Khog/J (Nur77-GFP) mice were purchased from Jackson Laboratory and bred in specific pathogen-free conditions at Oregon Health and Sciences University or New York University. B6.Cg-Tg(CAG-tdKaede)-15Utr (Kaede-Tg)²⁰ were obtained via D.J. Fowell in agreement with RIKEN BioResource Research Center. K14-VEGFR3-Ig mice²¹ were a generous gift from Dr Kari Alitalo and obtained from Dr Melody A. Swartz. Mice were age and sex matched, and female mice were used unless otherwise stated. OT-1 chimeras were created by adoptively transferring 15,000 CD8⁺ T cells from OT-1 mice into naïve mice. For co-transfer experiments, 15,000 wildtype OT-1 T cells and 15,000 Cxcr6^−/− OT-1 T cells were co-transferred into naïve mice. For experiments with K14-VEGFR3-Ig mice, both male and female recipients were used and 30,000 wild-type OT-1 T cells and 30,000 Cxcr6^−/− OT-1 T cells were co-transferred into naïve mice. Mice were used between 8 and 14 wk of age. All animal procedures were approved and performed in accordance with the Institutional Animal Care and Use Committees at OHSU and NYU Langone Health.

Pathogens and infections

Mice were infected the following day after adoptive transfer, with vaccinia virus (VACV), VACV expressing SIINFEKL (VACV-OVA) or VSV expressing full length ovalbumin (VSV-OVA). VACV, and VACV-OVA were propagated in BSC-40 cells as per standard protocols. Mice were infected by administering 1–5 × 10⁶ PFU of VACV, or VACV-OVA in 10 μl of PBS to the ventral side of the ear pinna followed by 25 pokes with a 29-G needle (skin scarification). For VSV-OVA infections, mice were injected intravenously with 1 × 10⁷ PFU. For reactivation of T_RM with peptide, VV-OVA immune mice were injected RO with 300 μl of 0.5 mg/ml BFA in PBS. Thirty minutes later, cells in blood were labeled with RO injection of 3 μg of anti-CD45 APC. Immediately after IV labeling, 25 μg of SIINFEKL in 10ul of 4:1 Acetone: DMSO was placed on both sides of ear skin and poked 10–15 times with a needle. Mice were taken down 5 h later.

To generate effector CD8⁺ T cells for in vitro studies, ActA deficient Listeria monocytogenes expressing SIINFEKL (LM-OVA) was grown in tryptic soy broth supplemented with 50 μg/ml streptomycin at 37 °C and until it reached 1 × 10⁸ CFU/ml (OD600 = 0.1). Next, 1 × 10⁷ CFU was administered intravenously and spleens harvested 7 d later.

In vivo antibody treatment

Mice were treated with antibodies targeting CXCL16 (R&D Systems, clone: 142417, 25 µg) or rat IgG control (Sigma) on day 12 and 13 prior to sacrifice on day 14. Antibodies were administered via intraperitoneal injection.

In vitro restimulation

Splenocytes were cultured in vitro for 24 h in 1 nm SIINFEKL peptide or with 10 ng/ml of TGF-β in RPMI containing 10% FBS.

In vivo T cell egress

To label Kaede expressing T cells, ear skin was photoconverted 6 d after infection using 405-nm light (Steele 2019) for 2 min (1 min each side) at 10mW. Mice were euthanized 24 h later.

Immunofluorescent microscopy

After euthanizing mice, ears were frozen in OCT on an isopentane bath. The 7 μm sections were cut and fixed in acetone for 15 min. Slides were blocked for 10 min with (1% BSA in PBS) and stained with antibodies in 1%BSA in PBS for 1 to 2 h at room temperature. Antibodies were obtained from Biolegend, Invitrogen, BD Biosciences, Bioss, and Tonbo and included CXCL16 (pAb Bioss), K14 (POLY9060), CD8 (53-6.7), CD45.1 (A20), CD90.1 (OX-7). Nuclei were detected and cover slips were attached using 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI) containing Prolong Diamond Anti-Fade (Life Technologies). Slides were imaged on a Keyence BX-X810 microscope.

Leukocyte isolation

Ears of mice were removed, and the dorsal and ventral sides separated and incubated for 45 min at 37 C in 1 ml of HBSS (Hyclone) containing CaCl₂ and MgCl₂ supplemented with 125 U/ml of collagenase D (Invitrogen) and 60 U/ml of Dnase-I (Sigma-Aldrich). Tissue was smashed on a scored plate and poured through a 70 μm filter. For some experiments, leukocytes were purified by resuspending pellets in 35% Percoll (GE Healthcare) and HBSS followed by room temperature centrifugation at 1800 RPM. Single cell suspensions of leukocytes from lymph nodes and spleens were made by smashing the tissue through a 70 μm filter. Spleens were resuspended for 2 min in 2 ml of ammonium-chloride-potassium lysis buffer. Lymphocyte isolation from all other non-lymphoid tissues was performed as previously described²² using 35% Percoll and HBSS followed by room temperature centrifugation at 1800 RPM.

Flow cytometry

For intravenous (IV) labeling, mice were injected intravenous with 2–3 μg of anti-CD45 antibody and euthanized 3 minutes later. For surface stains, single-cell suspensions were stained with antibodies in FACS buffer (1%BSA in PBS) for 30 min at 4C. Antibodies were obtained from Biolegend, Invitrogen, BD Biosciences, BD Pharmingen, Bioss, and Tonbo and included CXCL16 (12-81), BCL2 (BCL/10C4 or 3F11), CD8 (53-6.7), CD45.1 (A20), CD90.1 (OX-7), CD90.2 (30-H12), CXCR6 (SA051D1), CD69 (H1.2F3), CD103 (2E7), CD44 (IM7), CD45 (30-F11), and CD16/32 (2.4G2). Ghost dye 780 (ZombieRed) was used to measure cell viability. For some flow cytometry analysis, Cxcr6^−/− CD90.1⁺ OT-1 T cells were identified via GFP expression and AF488-CD90.1 antibody. Mitochondrial staining was performed as per manufacturer (Invitrogen) instructions. Staining of intracellular BCL2 was performed with True-Nuclear™ Transcription Factor Buffer Set (Biolegend) as per manufacturer protocol. Samples were analyzed on a BD LSR II, or BD FACSymphony. FlowJo 10.8.1. was used to analyze flow cytometry data.

ELISA

Ears from naïve or previously infected mice were weighed and snap frozen in liquid nitrogen. Ears were homogenized and CXCL16 concentration was quantified using Mouse CXCL16 DuoSet ELISA (R&D Systems) as per the manufacturers protocol.

Sorting and single-cell RNA-seq

Ears of C57BL/6J mice were infected with VACV-OVA the day after receiving 15,000 WT OT-1 T cells and 15,000 Cxcr6^−/− OT-1 T cells. At 21 d post infection, lymphocytes were isolated from skin after intravenous injection of anti-CD45-PE antibody as previously described. WT OT-1 T cells (CD8⁺CD90.2⁺CD44⁺CD45 IV^-CD45.1⁺CD90.1⁻) and Cxcr6^−/− OT-1 T cells (CD8⁺CD90.2⁺CD44⁺CD45 IV^-CD45.1⁻CD90.1⁺) were sorted. The sorted cellular suspensions were loaded on a 10× Genomics Chromium instrument to generate single-cell gel beads in emulsion. Libraries were prepared using Single cell 3’Reagent kits v3.1 (Chromium Next GEM Single Cell 3’ GEM, Library & Gel Bead Kit v3.1, 16 rxns PN-1000121; 10× Genomics) and were sequenced using Illumina Novaseq 6000.

Bioinformatics and pathway analysis

Quality control was performed on sequenced cells to calculate the number of genes, UMIs and the proportion of mitochondrial genes for each cell. Cells with low number of covered genes (gene-count < 200) and high mitochondrial counts (mt-genes > 0.08) were filtered out. There were 8634 cells in the Cxcr6^−/− sample and 7409 cells in the WT sample after filtering. The matrix was normalized based on their library sizes. A general statistical test was performed to calculate gene dispersion, base mean and cell coverage to use to build a gene model for performing Principal Component Analysis (PCA). Genes with high coverage (top 500) and high dispersion (dispersion > 1.5) were chosen to perform PCA and batch alignment using iCellR R package (v1.6.5). T-distributed Stochastic Neighbor Embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP) were performed on the top 10 PCs. PhenoGraph²³ clustering was then performed on the UMAP results. Marker genes were found for each cluster and visualized on heatmaps, bar plots, and box plots. The marker genes were used to determine cell types. Proportion (percentage) of the cell communities in each condition were calculated. The Tirosh scoring method²⁴ was used to calculate gene signature scores for circulating and resident memory.²⁵ RNA velocity results were performed by using velocyto (v0.17)²⁶ (https://doi.org/10.1038/s41586-018-0414-6) and scVelo (v0.2.3) (https://scvelo.readthedocs.io/). Pathway analysis was performed using marker genes in Metascape v.3.5 (http://metascape.org).

Statistics

Statistical analysis was performed using GraphPad Prism 9 software. Comparisons between 2 groups were conducted using paired or unpaired Student’s t-tests. Comparisons between 3 or more unpaired groups were conducted using ANOVA with Tukeys multiple comparison test. Comparisons between 3 or more paired groups were conducted using repeated measures ANOVA with Greenhouse-Geisser correction. Proper sample size was based off of prior experience. Statistical analysis of sequencing data was performed with pairwise Wilcoxon rank test and Bonferroni correction in R. P values are reported as follows: * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001. Error bars show the mean ±SD.

Results

CXCR6 optimizes CD8⁺ T_RM formation across peripheral, non-lymphoid tissues

In mice and humans Cxcr6 is a part of the core cluster of genes associated with tissue residence,^5–7 however, the mechanisms that regulate Cxcr6 expression and its specific role in mediating peripheral tissue CD8⁺ T cell behavior remain unclear. To track CXCR6 expression and determine its functional contribution to T_RM formation we infected C57BL/6J mice with vaccinia virus (VACV) by scarification on the ear pinna (10⁶ PFU). Infection with VACV by scarification peaks day 3–5 post infection, is cleared by day 15, and no viable virus is found at memory time points (Fig. S1A).^27,28 Furthermore, CD8⁺ T cells do not engage antigen in the skin 20 d after infection.²⁹ As such, VACV infection by scarification is used to model the generation of anti-viral, long-lived skin T_RM after an acute viral infection.^28–30 Tracking anti-viral T cells in this model, we found that CXCR6 was highly expressed on the surface of endogenous VACV-specific CD8⁺ T cells (H2-K^b-B8R_20-27) in skin 30 d post infection (d.p.i) but was relatively low in VACV-specific and naïve CD8⁺ T cells in the spleen at matched time points (Fig. 1A, B). While CD103 labels a subset (38 ± 7% of CD69⁺) of antigen-specific T_RM in skin at 28 d.p.i., CXCR6 was highly expressed on a larger percentage of CD69⁺ cells (85 ± 4%) (Fig. 1A). Furthermore, CD69⁺ cells, which are considered to be bona fide T_RM,²⁸ had significantly higher levels of CXCR6 compared to CD69⁻ T cells in skin at the same time point (Fig. 1C), all consistent with a role for CXCR6 in the formation of skin T_RM. We therefore asked if CXCR6 promotes the generation of T_RM post infection with VACV. We co-transferred equal numbers of congenically mismatched, WT and Cxcr6^−/− (Cxcr6^GFP/GFP)¹⁹ OT-1 T cells, specific for H2K^b-OVA_257-264 (SIINFEKL), into mice and infected ear skin with VACV expressing the immunodominant CD8⁺ T cell epitope of ovalbumin (OVA_257-264; VACV-OVA) the following day (Fig. 1D and Fig. S1B). We compared the ratio of WT and Cxcr6^−/− OT-1 T cells more than 45 d.p.i. and observed significant enrichment for WT over Cxcr6^−/− OT-1 T cells in the skin relative to the spleen (Fig. 1E, F). Interestingly, the advantage provided by CXCR6 expression was restricted to the resident compartment. Analysis of both VACV-specific CD69^- circulating memory and CD69⁺ T_RM at late memory time points (>100 d.p.i.) revealed that only T_RM and not circulating memory, either in spleen or circulating through skin, were outcompeted by wildtype (Fig. 1G). Of the OT-1 T cells within the skin, WT cells exhibited more robust CD69 and CD103 expression than Cxcr6^−/− OT-1 T cells (Fig. S1C and D) and were less abundant in both the epidermis and dermis (Fig. S1E, F). These data indicated that CXCR6 was required for optimal CD8⁺ T_RM formation in the skin following VACV infection.

Figure 1.

CXCR6 expression is a conserved feature of resident CD8⁺ T cells after viral infection and necessary for optimal seeding. (A) Representative flow of CD44⁺ vaccinia virus (VACV) specific (H2-K^bB8R_20-27) CD8⁺ T cells 28 days post infection (d.p.i.) (VACV, scarification) and expression of CD69, CD103, and CXCR6. (B, C) Quantification of CXCR6 geometric mean fluorescence intensity (GMFI) on CD44⁺CD8⁺ B8R-specific T cells from (A), as a function of location (B) and CD69 expression (C). (D) Experimental design where 15,000 wild-type (WT) OT-1 T cells (CD45.1⁺CD90.1^-) and 15,000 Cxcr6^−/− OT-1 T cells (CD45.1^-CD90.1⁺) were co-transferred into naïve mice (CD45.1^-CD90.1^-) and the following day, ear skin was infected with VACV expressing OVA_257-264 (VACV-OVA, scarification). (E) Representative flow plots at least 49 d post infection (d.p.i.). Gated on CD8⁺CD44⁺IV^- in the skin. (F) Quantification of the ratio of WT and Cxcr6^−/− OT-1 T cells isolated from the spleen and ear skin from (E). (E, F) Data are cumulative from 3 experiments for a total of n = 12. (G) Quantification of the ratio of WT and Cxcr6^−/− 100–115 d.p.i. with VACV-OVA infection stratified by CD69 expression. (H) CXCR6 expression on CD45.1⁺ OT-1 T cells in non-lymphoid tissues 31 d.p.i. with vesicular stomatitis virus expressing OVA (VSV-OVA) (gated on CD8⁺CD45.1⁺CD44⁺CD69⁺, and IV^- in all tissues except liver and spleen) (SG = salivary gland and FRT = Female reproductive tract). (I) Representative flow plot and (J) quantification of CXCR6 GMFI on IV^-CD45.1⁺ OT-1 T cells from kidney as a function of CD69 expression. (K) Quantification of the ratio of WT and Cxcr6^−/− OT-1 T cells isolated 31 d.p.i. with VSV-OVA. Gated on CD8⁺CD44⁺IV^- in kidney, FRT, brain and salivary gland; CD8⁺CD44⁺ in liver and spleen. Data are cumulative of two experiments for a total of n = 8. Bars represent average + SD. (A–G) Data are representative of at least 2 experiments with n = 4 each. Statistical significance determined with one-way ANOVA (B, G, H, and K), and paired t test (C, F, and J). * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001.

T_RM phenotype is highly dependent upon pathogenic insult and anatomical location.³¹ We asked whether high CXCR6 expression was restricted to CD8⁺ T_RM formed following cutaneous VACV infection by infecting mice with vesicular stomatitis virus (VSV), which establishes CD8⁺ T_RM in diverse non-lymphoid tissues when delivered by retroorbital injection (10⁶ PFU).³² We first transferred naïve CD45.1⁺ OT-1 T cells into mice and infected them with VSV-OVA. As we saw in skin with VACV, in all tissues tested (liver, brain, kidney, salivary gland, and female reproductive tract), CXCR6 expression was higher on tissue CD8⁺ T_RM (at least 30 d.p.i.) than antigen-specific CD8⁺ T cells in the spleen of the same mouse (Fig. 1H). Furthermore, CD69⁺ OT-1 T_RM expressed significantly more CXCR6 compared to CD69^- T cells in the same tissue (Fig. 1I, J), suggesting that CXCR6 expression is a conserved feature of anti-viral, CD8⁺ T_RM. Importantly, the accumulation of T_RM in other peripheral non-lymphoid tissues, including the brain, kidney and female reproductive tract, depended on CXCR6 expression (Fig. 1K), with WT T cells outcompeting Cxcr6^−/− in all tissues examined, though the extent to which CD69 and CD103 were differentially expressed was tissue specific (Fig. S1G–K). These data together indicate that CXCR6 optimizes de novo CD8⁺ T_RM formation in disparate anatomical locations and in response to diverse viral infections.

CXCR6 expression is an early event along the T_RM differentiation trajectory

These data indicated that even with a fixed TCR and at a static time point, CXCR6 seemed to distinguish a pool of circulating and resident CD8⁺ T cells within the tissue interstitium. To resolve the CD8⁺ T cell states where CXCR6 is turned on in the T_RM differentiation trajectory, we performed single-cell RNA sequencing (scRNAseq) on interstitial CD8⁺ T cells. Congenically distinct, OT-1 T cells were transferred intravenously into C57BL/6J mice. The following day, mice were infected with VACV-OVA. Live, extravascular OT-1 T cells (negative for intravascular stain; IV⁻) were sorted from infected skin 21 d.p.i. and submitted for scRNAseq to capture cells at the beginning of memory formation.²⁸ Transcriptional analysis and visualization using Uniform Manifold Approximation Projection (UMAP) identified six clusters (Fig. 2A, Fig. S2A and B, and Table S1). Cluster 1 expressed high levels of Tcf7, Klf2, and S1pr1 and scored for a core circulating gene signature (T_CIRC, Fig. S2C, D), while clusters 2–5 expressed high levels of Itgae and Cd69 and enriched for a transcriptional program associated with residence (T_RM, Fig. S2E and F). Cluster 6 expressed high levels of Zeb2, a marker of terminal differentiation that is absent from long-lived T_RM following lymphocytic choriomeningitis virus (LCMV) infection.³³ Cluster 6 also differentially expressed genes associated with RNA splicing and apoptotic cleavage of cellular proteins, perhaps indicating an increased rate in cell death (Fig. S2B). The heterogeneity observed within the T_RM clusters suggested that even at a single time point and with a fixed TCR, CD8⁺ T cells could be captured in distinct transcriptional states that might reflect either functional or positional heterogeneity within the tissue. We performed scVelo analysis to define a pseudotime trajectory between the clusters, which indicated a directional differentiation path from the circulating cluster 1 towards the T_RM clusters (Fig. 2B–D). Consistent with this trajectory, the T_RM subclusters^2–5 varied in their expression of effector molecule transcripts and several transcription factors related to T cell activation and differentiation (Fig. S2A). Cells in cluster 3 exhibited an intermediate state consistent with their position in pseudotime at the transition from circulation to residence. They did not express transcripts high enough to meet our criteria for marker genes (fold change of 2, relative to average expression in other clusters) but expressed the highest levels of Klf2, Tcf7 and S1pr1 of all the T_RM clusters. Cluster 2 cells were defined by high expression of chemokines Ccl3, Ccl4 and transcription factors Fos and Jun, which may be indicative of recent activation.³⁴ Cluster 5 cells expressed Lmna and Anxa1. Cluster 4 was defined by the highest expression of cytotoxic molecules including Gzma and Gzmf, as well as Krt83, a keratin primarily found in the hair follicle. Cxcr6 expression was elevated in all T_RM subclusters relative to the circulating cluster 1, with high expression already in the intermediate cluster 3, pointing toward the induction of Cxcr6 expression as an early event in T_RM differentiation (Fig. 2E). As previously reported, T_RM CD8⁺ T cells exhibited lower expression of S1pr1 relative to T_CIRC, consistent with its proposed role in tissue egress.³ Of note, little Ccr7 expression was seen in any subset, instead T_CIRC cells expressed Cxcr4 which facilitates egress from melanoma through CXCL12-producing dermal lymphatic vessels.³⁵ In addition to Cxcr6, several receptors including, S1pr4, Ccr2, Ccr5, and Cxcr3 were enriched in T_RM relative to T_CIRC (Fig. S1G, H). As all cells analyzed were negative for an intravascular stain, these data begin to indicate that the high surface CXCR6 expression observed across T_RM likely depended upon a transcriptional event that occurred after tissue entry.

Figure 2.

CXCR6 is required for optimal formation of T_RM across non-lymphoid peripheral tissues. (A) UMAP representation of single-cell RNA-seq data from extravascular (IV^-) OT-1 CD8⁺ T cells isolated out of skin 21 d post infection (d.p.i.) with vaccinia virus expressing OVA_257-264 (VACV-OVA). Circulating (T_circ) and resident (T_RM) memory T cells. Data represent a total of 7409 cells across all clusters. (B) Projected RNA velocity (scVelo) laid over UMAP. Individual iCellR clusters scored for a published (C) circulating memory, (D) resident memory, and (E) normalized Cxcr6 expression. (F) 15,000 wild-type (WT) and 15,000 Cxcr6^−/− OT-1 T cells were co-transferred into naïve mice and the following day, skin was infected with VACV-OVA. Quantification of the ratio of WT and Cxcr6^−/− OT-1 T cells in spleen and skin and 5 and 14 d.p.i. Data are cumulative from 2 experiments with n = 4. Statistical significance determined using paired Student’s t-test (F) or pairwise Wilcoxon rank test relative to cluster 1 (C–E). **** P < 0.0001. CXCR6 (G) Expression of CXCR6 by WT OT-1 T cells in skin relative to spleen over time. Data are representative of at least 2 experiments with n = 3 or 4. (H) Experimental design where OT-1 T cells were transferred to naïve mice, which were subsequently infected on the left ear VACV-OVA and the right with VACV (scarification). (I) Representative histogram and quantification (J) of CXCR6 by OT-1 T cells in spleen or ear skin 14 d.p.i. Geometric mean fluorescence intensity (GMFI). Data are representative of at least 2 experiments with n = 3 or 4. (K and L) Nurr77-GFP OT-1 T cells were transferred to naïve mice and the skin was infected with VACV-OVA the next day. (K) Representative histograms and (L) quantification of CXCR6 by Nurr77-GFP⁺ and Nurr77-GFP^- OT-1 T cells in skin 7 d.p.i. Data are representative of two experiments with n = 5. (M) Representative histograms and (N) quantification of Nur77-GFP in the presence of CXCL16 blockade (aCXCL16) or isotype control 14 d.p.i.. Cells in skin (F) are gated on IV-. Statistical significance determined using paired Student’s t-test (F and L), 1-way ANOVA (G and J), pairwise Wilcoxon rank test relative to cluster 1 (C–E), or unpaired Student’s t-test (N). * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001.

Peripheral tissue antigen encounter upregulates CXCR6 expression

The low levels of CXCR6 expression on circulating (T_CIRC) anti-viral CD8⁺ T cells within the skin interstitium (Figs. 1B and 2E) suggested that CXCR6 expression might not be strictly necessary for infiltration into inflamed skin. To directly test whether CXCR6 is required for tissue infiltration and the early accumulation of anti-viral effector CD8⁺ T cells, we quantified the ratio of co-transferred WT and Cxcr6^−/− OT-1 T cells at 5 and 14 d.p.i. with VACV-OVA (Fig. 2F). Importantly, there was no bias toward WT T cells 5 d.p.i. in the draining LN, indicating that CXCR6 deficiency does not impair the magnitude of priming and expansion (Fig. S3A). Additionally, both WT and Cxcr6^−/− OT-1 T cells infiltrated infected skin and were present at a similar ratio to that found in spleen at early (day 5) and late effector (day 14) time points (Fig. 2F). Consistent with the lack of requirement for tissue entry, CXCR6 expression by WT OT-1 T cells in skin at 5 d.p.i. was almost indistinguishable from expression on WT OT-1 T cells found in the spleen (Fig. 2G), indicating that CXCR6 was likely upregulated after entry into the skin. Indeed, CXCR6 surface expression increased in skin relative to spleen over time, with established CD8⁺ T_RM (day 49) expressing almost 6 times as much surface CXCR6 as their circulating memory counterparts (Fig. 2G). Consistent with cutaneous VACV infection, CXCR6 was also dispensable for early T cell accumulation in the skin following systemic VSV (Fig. S3B), and dispensable for accumulation in the small intestine epithelium, and brain 7 d.p.i (Fig. S3C). WT OT-1 T cells, however, outcompeted Cxcr6^−/− T cells in the liver, salivary gland, female reproductive tract, and kidney at these early time points (Fig. S3C), which may indicate either that low levels of CXCR6 on circulating CD8⁺ T cells can support tissue infiltration in a tissue-specific manner, or that CD8⁺ T cells entering these unique tissue environments are more sensitive to loss of CXCR6 expression.

These data indicated that the induction of CXCR6 within skin was a key event for T_RM formation. Antigen encounter in peripheral non-lymphoid tissue induces CD69 expression and CD8⁺ T_RM formation in both skin^28,30 and lung^36,37 following infection. Given the concordance between CXCR6 expression and CD69, we hypothesized that CXCR6 enrichment in skin was also antigen dependent. To test this, we transferred naïve OT-1 CD8⁺ T cells into mice and the next day infected one ear with VACV and the contralateral ear with VACV-OVA. VACV-OVA primed OT-I T cells infiltrate both infected sites independent of antigen expression,²⁸ allowing us to evaluate CXCR6 levels on a fixed CD8⁺ T cell population in the presence or absence of cognate antigen (Fig. 2H). We observed significantly higher levels of CXCR6 on OT-1 T cells in the presence of cognate antigen (VACV-OVA) compared to VACV infected ears or the spleen 14 d.p.i. (Fig. 2I, J). Interestingly, while cognate antigen clearly boosted surface CXCR6 expression on anti-viral CD8⁺ T cells, there was a significant increase in CXCR6 on bystander OT-1 T cells in VACV-infected skin relative to spleen, consistent with a role for inflammatory cytokines in driving CXCR6 expression in skin (eg, interleukin [IL]-2 and IL-15).¹⁹ In order to further confirm a role for antigen encounter and account for potential cytokine differences in VACV and VACV-OVA infected skin, we used mice that report TCR stimulation via transgenic expression of green fluorescent protein (GFP) under control of the Nr4a1(Nur77) promoter (Nurr77-GFP).³⁸ In VACV-OVA infected skin 7 d.p.i., surface CXCR6 was enriched on Nurr77-GFP expressing OT-1 T cells compared to Nurr77-GFP negative OT-1 T cells (Fig. 2K, L), indicating that recent antigen recognition in skin is associated with increased CXCR6 levels.

These data indicated that CXCR6 induction in response to antigen encounter was required for efficient transition to residence. Given that antigen encounter itself is an important cue for T_RM formation, we sought to further confirm that CXCR6 was a result of TCR stimulation in the tissue rather than a necessary homing signal for initial antigen presenting cell engagement. Following transfer of WT Nur77-GFP OT-I T cells we treated mice with two doses of CXCL16 blocking antibody¹⁷ or isotype control on days 12 and 13 prior to sacrifice 14 d.p.i.. CXCL16 blockade did not significantly reduce the percentage of Nur77⁺ T cells, thereby demonstrating that CXCR6 is not required for interstitial T cell homing to antigen bearing cells (Fig. 2M, N). Further supporting the idea that antigen encounter reinforces CXCR6 expression, ex vivo activation of CD8⁺ effector T cells (7 d.p.i. listeria monocytogenes-OVA) with cognate SIINFEKL resulted in CXCR6 upregulation relative to unstimulated controls over 24 h (Fig. S3D). TGF-β signaling is required for the transition of circulating CD8⁺ T cells toward residence, in part by downregulating S1PR1 and upregulating CD103;^39–41 however, the addition of TGF-β did not affect CXCR6 expression in the presence of antigen stimulation ex vivo (Fig. S3D). These data, together with the observation that CD103⁺ T_RM are a subset of CD69⁺CXCR6⁺ T_RM (Fig. 1), indicate that antigen-dependent upregulation of CXCR6 likely precedes TGF-β driven changes in the epidermis,⁴² and is therefore consistent with CXCR6 as an early feature of CD8⁺ T cell adaptation.

CXCR6/CXCL16 interactions in the dermis do not act to restrain T cell exit

To define the anatomic niche within which CXCR6 acts on CD8⁺ T cell differentiation, we analyzed expression of its ligand. CXCL16 was present in skin at all time points analyzed but appeared to peak in expression 7 d.p.i., returning to basal-like levels at late memory time points (Fig. 3A). Expression in the skin 7 d.p.i. seemed to largely be driven by myeloid cells, including macrophages and type 1 and 2 conventional dendritic cells (Fig. 3B). Consistent with the low expression on CD45^- cells at effector time points (14 d.p.i.), CXCL16 was absent from the epidermis (Fig. 3C), in contrast to epithelial expression observed in the lung.¹⁴ Instead, clusters of CXCL16-expressing cells were found within the dermis of infected skin (Fig. 3C). Given that effector CD8⁺ T cells first enter the dermis through activated capillaries and then transition with time to the epidermal compartment (Fig. 3D), the enrichment of CXCL16 in the dermis might suggest a role for CXCR6 in the transition to epidermal residence.

Figure 3.

Dermal CXCL16 does not restrain T cell egress out of skin. (A) Quantification of CXCL16 in vaccinia virus (VACV) infected skin at various days post infection (d.p.i) by scarification; n = 3–5 per group. (B) Quantification of CXCL16 surface expression (GMFI) 12 d.p.i. on dermal and splenic myeloid cells including macrophages (Mac., CD11b⁺F480⁺), conventional dendritic cells type 1 (cDC1, CD11c⁺MHC-II^HIXCR1⁺) and type 2 (cDC2, CD11c⁺MHC-II^HISIRPA⁺). (C) Representative immunofluorescence images of skin 14 d.p.i. with VACV. Scale bar = 100 μm. (C) The percentage of OT-1 T cells in the epidermis after infection with OVA_257-264 -expressing VACV (VACV-OVA) calculated from immunofluorescent images. Data are cumulative of at least two experiments with n = 5 (day 14) or n = 6 (day 20 and 42–45). (E) Experimental design where 15,000 WT and 15,000 Cxcr6^−/− OT-1 T cells were transferred to K14-VEGFR3-Ig mice and the following day infected with VACV-OVA via skin scarification. (F) Quantification of the ratio of WT and Cxcr6^−/− OT-1 T cells in the spleen and skin of K14 VEGFR3-Ig mice 49 d.p.i. with VACV-OVA. Data are representative from at least 2 experiments with n = 4. (G) Experimental design where congenically distinct 15,000 WT Kaede⁺ OT-1 T cells and 15,000 Cxcr6^−/− Kaede⁺ OT-1 T cells were co-transferred to C57BL/6J mice and infected the following day with VACV-OVA via skin scarification. Kaede expressing cells in skin were photoconverted 6 d.p.i. and harvested 1 day later from skin and draining lymph node (dLN). (H) Representative flow of skin and dLN. (I) The ratio of photoconverted WT Kaede red⁺ to photoconverted Cxcr6^−/− Kaede red⁺ in skin and dLN following photoconversion. Data are representative from at least 2 experiments with n = 5. Statistical significance determined using 1-way ANOVA (A, D) or paired Student’s t-test (F, I). Error bars represent SD. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001.

One hypothesis was that a failure to retain CD8⁺ T cells in the dermis in a CXCL16/CXCR6-dependent manner could lead to increased rates of exit through the proximal dermal lymphatic vessels. Dermal lymphatic vessels facilitate CD4⁺ and CD8⁺ T cell egress out of skin,⁴³ and the egress of CD8⁺ T cells limits the number of CD8⁺ T cells that transition to residence.^3,29 We therefore asked if we could rescue Cxcr6^−/− T_RM in the absence of dermal lymphatic vessels, where a physical loss of vessels would prevent T cell exit. Mice in which a VEGFR3-Immunogloubulin fusion protein is expressed under control of the K14 promoter (K14-VEGFR3-Ig) fail to form mature dermal lymphatic vessels,²¹ restricting both fluid and cellular efflux out of skin.²⁷ K14-VEGFR3-Ig mice exhibit delayed lymphocyte responses to VACV infection but are capable of T cell-mediated viral clearance.^27,29 We co-transferred equal numbers of congenically distinct WT and Cxcr6^−/− OT-1 T cells into naïve K14-VEGFR3-Ig mice and WT controls (Fig. 3E). At early time points post infection, T cells readily infiltrated the skin of both WT and K14-VEGFR3-Ig mice with no enrichment for WT T cells as seen previously (Fig. S4A). At memory time points, we expected that if CXCR6 was a retention signal, the removal of the egress route would be sufficient to rescue the accumulation of Cxcr6^−/− T cells to WT levels in skin. However, the lack of dermal lymphatic vasculature in K14-VEGFR3-Ig mice failed to normalize the ratio of WT and Cxcr6^−/− OT-1 T cells in the skin (Fig. 3F). In line with this finding, Cxcr6^−/− OT-1 T cells were as likely to egress through functional infection-associated lymphatic vessels as their WT counterparts when directly tracked using a photoconvertible transgenic mouse (Kaede-tg).^43,44 Kaede-tg OT-1 CD8⁺ T cells and Kaede-tg Cxcr6^−/− OT-1 CD8⁺ T cells were transferred into mice one day prior to infection with VACV-OVA. At 6 d.p.i., the skin was photoconverted and the number of Kaede-red⁺CD8⁺ OT-I T cells were quantified in draining LNs 24 h later (Fig. 3G, Fig. S4B, C). At this time point, dermal-infiltrating T cells are actively engaging target cells and expressing CXCR6 (Fig. 2G, K, and L). We found no difference in the ratio of photoconverted WT to Cxcr6^−/− CD8⁺ T cells in the skin compared to those that had egressed to the draining LN (Fig. 3H, I). These data indicated that despite being a marker of residence, CXCR6 did not appear to be acting as a physical retention cue. Our findings pointed to a dermally-restricted, CXCR6-dependent mechanism of CD8⁺ T cell accumulation in infected skin.

CXCR6 supports metabolic adaptations during the transition from effector to residence

Our data indicated that CXCR6 likely acted in the dermis to facilitate the transition from effector T cell to T_RM over time, though exactly how this chemokine receptor might do this remained unclear. At 21 d.p.i. Cxcr6 is expressed in all WT OT-1 T cells that bear a transcriptional program consistent with residency (Fig. 2), indicating its expression is an early event in T_RM differentiation. We reasoned that CXCR6 might influence the differentiation potential of effector CD8⁺ T cells in skin. We therefore integrated our WT OT-1 scRNAseq with Cxcr6^−/− OT-1 T cells extracted from the same skin at 21 d.p.i., when WT and Cxcr6^−/− OT-1 CD8⁺ T cells are still at roughly equal ratios (Fig. S5A). Integrated analysis revealed a similar clustering pattern (Fig. 2A) and pseudotime trajectory (Fig. S5B) as seen for the WT cells alone, again with a circulating cluster (T_CIRC: cluster 1), 4 clusters of T_RM (T_RM: clusters 2–5) (Fig. 4A), and a sixth cluster expressing high amounts of Mcm9 (Mini chromosome maintenance 9) and the transcription factor Zeb2, associated with terminal differentiation.⁴⁵ While both WT and Cxcr6^−/− OT-1 T cells were present in all clusters, the distribution of cells within the T_RM clusters was distinct, indicating that the Cxcr6^−/− T cells were biased in their differentiation trajectory (Fig. 4B). Consistent with this observation, while the T_CIRC cluster 1 scored equally for a published circulating memory score between WT and Cxcr6^−/−, the T_RM super cluster demonstrated a reduced capacity to acquire the core T_RM transcriptional program in the absence of CXCR6 (Fig. 4C).²⁵ We therefore performed differential gene expression analysis on the T_RM supercluster and identified several downregulated genes (eg Icos, Il21r, and Nr4a3) (Fig. 4D and Table S2) that are functionally implicated in T_RM formation within various tissue compartments.^25,46,47 Pathway analysis of the differentially expressed genes revealed that WT cells expressed higher levels of genes involved in leukocyte activation, cytokine signaling and regulation of apoptosis (Fig. 4E). Genes enriched in these pathways included negative regulators of TCR and NF-κB signaling (Grap, Tnfaip3, Traf1, Nfkbia, Nfkbie)⁴⁸ and components of the immunoproteasome (Psmb8, Psmb9, Psma6) important for regulating the transition to memory and cell survival.⁴⁹ Referencing the TRRUST database confirmed transcriptional regulation of Nfkb1 (p = 8.232e−8) and Nfkbia (p = 3.476e−5). In contrast, the only pathway significantly upregulated in Cxcr6^−/− cells was oxidative phosphorylation (Fig. 4E), largely driven by genes associated with mitochondrial electron transport complex (Cox6c, Cox7a2, Cox8a, Ndufa4, Ndufb11). Further analysis of the downregulated genes using HumanBase⁵⁰ demonstrated four key modules including NF-κB signaling and cellular detoxification, which was driven by reduced expression of Gstp1, Park7, Prdx1, and Prdx6. The peroxiredoxin family of antioxidant enzymes reduce hydrogen peroxide and alkyl hydroperoxides and thereby play a protective role in cells. Given that our data predicted that Cxcr6^−/− T cells exhibited elevated oxidative phosphorylation and an impaired ability to detoxify reactive oxygen species, we hypothesized this could lead to a survival disadvantage within the dermis. Indeed, T cell memory is dependent upon mitochondrial biogenesis, which increases spare respiratory capacity to promote survival and rapid response to recall.⁵¹ To validate these transcriptional findings, we directly quantified mitochondrial biomass, membrane potential, and mitochondrial reactive oxygen species in our in vivo co-transfer system. We noted that Cxcr6^−/− OT-1 T cells exhibited reduced mitochondrial mass (Mito Deep Red, Fig. S5D), reduced membrane potential (CMX-Ros, Fig. S5E), and elevated mitochondrial reactive oxygen species (MitoSoxRed, Fig. 4G) 14 d.p.i. relative to WT OT-1 T cells in the same tissue. Importantly, differential MitoSoxRed staining was only observed in the CD69⁺ population of interstitial T cells, consistent with a critical role for CXCR6 post TCR engagement. In contrast, no differences in MitoSox Red were observed in the spleen (Fig. 4H).

Figure 4.

Cxcr6^−/− T cells fail to fully acquire a residence transcriptional program and display signs of metabolic dysfunction and apoptosis. (A) UMAP showing merged wild-type (WT) and Cxcr6^−/− OT-1 T cells isolated from skin 21 d post infection (d.p.i.) with vaccinia virus expressing OVA_257-264 (VACV-OVA). Resident (T_RM) and circulating (T_CIRC) clusters defined as in Fig. 2. (B) Cell density plots for WT and Cxcr6^−/− OT-I T cells overlaid on the merged UMAP. (C) Cells in T_CIRC cluster and T_RM clusters scored for published circulating memory (left) or resident memory (right) gene signatures. (D) Volcano plot of differentially expressed genes (DEG) between WT and Cxcr6^−/− cells in combined T_RM clusters.^2–5 (E) Pathway analysis of DEG between WT and Cxcr6^−/− cells in T_RM clusters. P value calculated by hypergeometric distribution test. (F) Human Base analysis of gene modules downregulated in Cxcr6^−/− T_RM relative to WT. (G, H) MitoSox Red GMFI in WT and Cxcr6^−/− OT-I T cells 14 d.p.i in the skin, stratified by CD69 expression (G), and in the spleen (H). Statistical significance determined using paired Student’s t-test (G–I) or pairwise Wilcoxon rank test (C). * P < 0.05, *** P < 0.001, **** P < 0.0001.

In addition to the shifts in transcriptional state, Cxcr6^−/− T cells were also more represented in Cluster 6 (Fig. 4B) which exhibited transcriptional programs consistent with apoptosis (Fig. S2B). We reasoned that the dysregulation of reactive oxygen species and mitochondrial fitness could shunt effector T cells off the T_RM differentiation pathway resulting in poor survival. Cluster 6 exhibited signs of terminal differentiation, expressing high levels of Zeb2 and was positioned at the end of the predicted pseudotime trajectory defined by RNA velocity (Fig. S5B).⁵² Following LCMV infection, a Zeb2-expressing CD8⁺ T_RM cluster did not persist to late memory time points (60–90 d.p.i.),³³ in contrast to a cluster expressing Id3, Jun, Fos, Ccl3, and Ccl4, genes enriched in our T_RM cluster 2, from which cluster 6 arises. Cluster 6 also scored highest for an apoptotic gene signature (R-MMU-109581) (Fig. S5F), with elevated expression of pro-apoptosis genes such as Casp8, Bcl211, and Acin1 (Fig. S5G). Even within the T_RM clusters, we found that transcripts for the anti-apoptotic proteins, Bcl2a1b and Bcl2a1d, were significantly reduced in Cxcr6^−/− relative to WT T cells (Fig. S5H), perhaps indicating these cells were poised for apoptosis. Bcl2a1 is an NF-κB target gene with four isoforms that promote lymphocyte survival and differentiation.⁵³ BCL2 protein was indeed lower in Cxcr6^−/− OT-1 T cells in the skin 21 d.p.i (Fig. S5I and J). These data indicate that while Cxcr6^−/− CD8⁺ T cells could adopt a T_RM-like transcriptional state, they exhibited a deficiency in multiple transcriptional programs, including mitochondrial fitness that may reduce overall survival and transition to long-term residence.

CXCR6 promotes dermal T cell survival during memory formation in infected skin

Our findings indicated that loss of Cxcr6^−/− T cells was a function of poor survival, which is likely driven by a combination of effects including reduced cytokine signaling, metabolic dysfunction, and a failure to fully activate a resident memory transcriptional program. Indeed, analysis of Cxcr6^−/− T cells infiltrating skin 21 d.p.i. exhibited a significant loss of viability relative to their WT counterparts. This change in viability was notably specific to the CD69⁺ population in skin, indicating that CXCR6 provides a specific advantage in the context of local antigen encounter (Fig. 5A, B). Given the high expression of CXCL16 in the dermis, we asked whether CXCR6 preferentially impacted dermal CD8⁺ T cell viability and transition to the epidermis. At 14 d.p.i., Cxcr6^−/− T cells exhibited similar viability in the epidermis but reduced viability in the dermis (Fig. 5C–E), with a consistent drop in numbers in the dermis 14 d.p.i. (Fig. 5F, G). This suggested that WT cells were outcompeting Cxcr6^−/− in the dermis at effector time points, prior to acquiring epidermal residence. These data indicate that Cxcr6^−/− CD8⁺ T cells form T_RM at suboptimal levels due to reduced survival in the dermis during a critical window of T_RM differentiation. Our working model predicts that CXCR6 optimizes survival following TCR engagement in non-lymphoid tissues. Even at very late time points (> 100 d.p.i.) the ratio between WT and Cxcr6^−/− T cells remained approximately 2 to 1 (Fig. 1G), indicating that CXCR6 might be dispensable for long-term maintenance despite its persistent expression. We therefore asked whether the T_RM that form in the absence of CXCR6 are functionally distinct from WT neighbors. Restimulation of established T_RM with cognate peptide induced equal amounts of IFN-γ per cell in both WT and Cxcr6^−/− OT-I T cells (Fig. 5H–J), further supporting the interpretation that CXCR6 promotes the efficiency of T_RM formation, but is not required for their long-term maintenance or function. Our data therefore position CXCL16/CXCR6 as a TCR-dependent signal necessary for optimal interstitial survival during the transition to residence but dispensable for long-term maintenance and effector function (Fig. 5K).

Figure 5.

CXCR6 promotes dermal T cell survival at effector time points in infected skin. (A) Ex vivo uptake of Zombie red viability dye 21 d post infection (d.p.i) with vaccinia virus expressing OVA_257-264 (VACV-OVA) in skin wild-type (WT) and Cxcr6^−/− (KO), stratified by CD69 expression, and in the (B) spleen. Data are cumulative of 2 experiments, each point is a mouse. (C) Representative flow plots and quantification of ex vivo uptake of ZombieRed viability dye in WT and KO OT-1 T cells in the (D) dermis and (E) epidermis 14 d.p.i. with VACV-OVA. Data are representative of at least 2 experiments with n = 4. (F) Representative immunofluorescence image and (G) quantification of WT and Cxcr6^−/− (KO) OT-1 T cells in skin at 14 d.p.i. with VACV-OVA. Data are cumulative of 2 experiments with n = 5 (F, G) and scale bar = 100 μm. (H) Representative flow plots gated on live CD44⁺IV^- WT and Cxcr6^−/− OT-I T cells in VACV-OVA immune mice (>100 d) restimulated with cognate peptide (SIINFEKL; 5 h). (I) Percent and (J) number of IFN-γ producing cells from H. (K) Graphical schematic of proposed mechanism. Statistical significance determined using paired Student’s t-test. * P < 0.05, **P < 0.01, *** P < 0.001.

Discussion

The formation of T_RM depends on mechanisms of tissue adaptation, cellular positioning, and differentiation. While the transcriptional programs that are enriched during transition from effector to memory are well described, the functional implications of key markers remain unclear. Transcriptional analyses of CD8⁺ T cells in diverse anatomical locations of mice and humans identified the chemokine receptor CXCR6 as a highly expressed marker of T_RM.^5–7 In this study, we investigated the functional significance of CXCR6 for T_RM formation. Using competitive co-transfer experiments of Cxcr6^−/− and WT controls, we find that CXCR6 expression is needed for optimal T_RM formation in peripheral non-lymphoid tissues after viral infection. Surprisingly, we found that CXCR6 has no effect on expansion in the draining lymph node, early accumulation of CD8⁺ T cells in the skin, nor does it prevent egress out of skin after infection. Instead, CXCR6 deficiency diminished T_RM quantity by reducing CD8⁺ T cell viability in inflamed skin, specifically within the dermis, leading to a failure to transition to epidermal residence.

Using adoptive transfer experiments and multiple viral models we show that CXCR6-deficiency diminished the expected quantity of T_RM in diverse tissues including brain, kidney and skin. While chemokines and their receptors are known to play critical roles in shaping the spatial distribution of T cells between and within tissues, here we identify a migration-independent mechanism of action. Across various non-lymphoid peripheral tissues and two infections, CXCR6 was not absolutely required for tissue infiltration. This is similar to observations made in mouse models of influenza infection, where CXCR6 is not needed for T cell entry into the lung parenchyma,¹⁴ and HSV infection of skin, where Cxcr6^−/− CD8⁺ T cells still form suboptimal T_RM numbers even after direct intradermal injection.¹⁵ Furthermore, CXCR6 did not appear to help anchor T cells in tissue, as deficient T cells did not exhibit increased rates of lymphatic egress and loss of the lymphatic vasculature was not sufficient to boost T_RM formation. Further, we find CXCL16 to be most highly expressed in the dermis at effector time points after infection, which pointed towards a functional contribution prior to acquisition of epidermal residence. This observation was in contrast to the lung where CXCL16 is expressed by the airway epithelium and thought to promote antigen-specific CD8⁺ T cell homing to that tissue compartment.¹⁴ In VACV-infected skin, CXCR3-deficient T cells fail to properly migrate towards and kill infected cells in the epidermis,⁵⁴ which may hint at tissue and/or viral-specific mechanisms of target cell homing.

Our findings therefore point to a model where CD8⁺ T cell antigen encounter in the dermis and concomitant CXCR6 upregulation precedes migration to the epidermis. Surface expression of CXCR6 by CD8⁺ T cells in skin is upregulated between 5 and 7 d.p.i. and partially depends on local antigen recognition. This is consistent with recent work that demonstrates that TCR signaling strength regulates Blimp1 expression, which thereby tunes the chemotactic features of effector CD8⁺ T cells during the transition to residence.³⁰ Similarly, we recently demonstrated in tumors that local antigen encounter downregulates CXCR4, which acts as a mechanism of tumor egress,³⁵ and in concordance Cxcr4 was transcriptionally downregulated in cells transitioning to residence following infection. In contrast, and as expected based on data presented here, CXCR6 was enriched on tumor-retained CD8⁺ T cells that exhibited signs of chronic antigen stimulation.^10,35 Interestingly, tissue entry and initial antigen presentation may be tightly linked in infected skin presenting the possibility that Cxcr6 upregulation occurs rapidly in a dermal, perivascular niche,⁵⁵ where CXCL16-expressing dendritic cells reside.¹⁰ We found that Cxcr6 was expressed across transcriptional states in resident memory precursors and followed a similar kinetic as antigen-dependent induction of CD69.²⁸ While these findings indicate that antigen encounter can significantly boost CXCR6 expression, we also observe evidence of increased expression upon tissue infiltration in the absence of antigen and do not rule out the impact of inflammatory cytokines on the enforcement of either CXCR6 expression or residence formation. Further consistent with our model, CXCR6 expression on restimulated effector CD8⁺ T cells was not affected by the presence of TGF-β and CXCR6 upregulation precedes TGF-β-dependent CD103 upregulation by several days. CD103 permits adhesion to E-cadherin expressed in epidermal keratinocytes and TGF-β is activated at the epidermal interface.^42,56 While TGF-β may facilitate residence in part through its role in downregulating S1pr1, S1pr5 and Cxcr4,⁴¹ our data suggests CXCR6 function precedes TGF-β dependent mechanisms of residency. Taken together, these findings indicate that CXCR6 upregulation likely takes place in the dermis, early after tissue entry and before CD8⁺ T cell transition to the epidermis.

Our data reveal that CXCR6 plays an important role for CD8⁺ T cells in the transition from circulating effector to resident memory but CXCR6 deficiency does not place an absolute block on T_RM differentiation. However, while Cxcr6^−/− OT-I T cells at an early memory time points adopt the same transcriptional states seen in WT OT-1 T cells taken from skin, we did find that CXCR6-deficient T_RM in the skin score lower for a resident memory gene signature. The differences in this resident memory gene signature, while modest, were driven by reduced expression for over a dozen genes including several genes with known roles in T_RM formation including, Il21r, Nr4a3 and Icos.^25,46,47 Decreased expression of Il21r is particularly interesting as IL-21 promotes memory T cell survival and proliferation, in a similar manner to IL-15.⁵⁷ In addition to these known T_RM-related transcripts, Cxcr6^−/− T cells appear to exhibit dysregulation of NFκ-B signaling, the immunoproteasome, oxidative phosphorylation, and apoptosis, all of which may indicate a cell not activating the appropriate cues to efficiently transition to memory. Consistent with the elevations in oxidative phosphorylation but reduced antioxidant capacity, Cxcr6−/− T cells accumulated more mitochondrial reactive oxygen species when compared to WT OT-I T cells from the same infected skin. It seems likely that these observed changes converge on multiple survival and activation pathways that decrease the probability of T cell survival and therefore long-term residence. These observations altogether may indicate that CXCR6 functions to retain activating CD8⁺ T cells in transient niches that optimize their receipt of survival cues to increase the efficiency of T_RM formation.

Previous studies evaluating the importance of CXCR6 expression by CD8⁺ T cells in anti-tumor responses shed further light on why CXCR6 is crucial for T_RM development. In mouse models of melanoma and pancreatic cancer, CXCR6 expression is necessary for tumor control by CD8⁺ T cells.^10,58 In these systems, CXCR6 optimizes interactions with CXCL16 and IL-15 expressing DCs in perivascular niches of the tumor stroma.¹⁰ T_RM dependency on IL-15 varies by anatomical location⁵⁹ but IL-15 is essential for the maintenance of CD103⁺ T_RM in skin.² IL-15 regulates oxidative phosphorylation for CD8⁺ T cells by inducing mitochondrial biogenesis and also increases BCL-2 expression.⁵¹ Our data associated loss of BCL-2 with a reduction in mitochondrial biogenesis and membrane potential, which could lead to the increased production of reactive oxygen species that we see and ultimately cell death. Consistently we see early cell death in the dermis but not the epidermis, indicating that balancing these signals first in the dermis is crucial for long-term epidermal outcome. These data are consistent with a hypothesis that reduced exposure to IL-15 in the dermis impairs the metabolic reprogramming necessary for survival and transition to memory. We cannot exclude the possibility that CXCR6 might directly activate NF-κB through Akt,⁶⁰ thereby improving CD8⁺ memory T cell fitness in part through direct regulation of BCL2.⁶¹ Further mechanistic studies to dissect the causal relationships between these diverse transcriptional programs and CXCR6-dependent residence are still needed.

In peripheral tissues such as the kidney, where T cells experience elevated levels of hypoxia,⁶² these stress-adaptation mechanisms may have more significant effects on interstitial persistence. While Cxcr6^−/− T cells were equally competent to reside and accumulate in skin at early time points (7 d.p.i.), they were already significantly impaired in the kidney, salivary gland, liver, and female reproductive tract at this same time point. While additional work would be needed to rule out a direct effect of CXCR6 on tissue homing and transendothelial migration, the relatively low levels of expression on circulating CD8⁺ T cells continue to support the hypothesis that CXCR6 helps meet the demands of the interstitial tissue microenvironment. The highly conserved expression of CXCR6 on resident NK and unconventional T cells,^5,6,63–65 suggests it facilitates a critical mechanism of cellular adaptation adopted across peripheral non-lymphoid tissues and inflammatory states.

Our findings demonstrate that CXCR6 instigates mechanisms of cellular adaptation to tissue and promotes local survival by facilitating signals that boost mitochondrial biogenesis and protect cells from death. Here, we have demonstrated that CXCR6 contributes to the process of T_RM formation by facilitating CD8⁺ T cell survival and transition to residency during phases of immune resolution 14–30 d post infection. It will be interesting to determine if CXCR6 remains required to maintain T_RM survival in the epidermis or if it is only needed during these inflammatory periods when CXCL16 expression is elevated, dermal effector cells are transitioning to epidermal memory, and survival cytokines might be limiting. Experimental models allowing for spatial and temporal control of CXCR6 or its ligand, CXCL16, will be needed to determine their role in maintenance and potential to regulate established T_RM populations. Since T_RM are potent sources of localized immunity that can mediate tissue-specific autoimmune responses, mechanisms of cellular adaptation to tissue stressors that are required for T_RM formation and maintenance may represent targets to limit T cell-induced injury across tissue sites.⁶² In sum, our data reveal key parameters that guide CD8⁺T cell residency and support the emerging concept that controlling T cell positioning could be a powerful strategy for tissue-specific immunotherapy across diverse diseases.

Author contributions

Conceptualization: T.A.H., A.W.L. Methodology: T.A.H., Z.L., O.I. Investigation: T.A.H., Z.L., A.C.S., M.M.S., T.M., O.I. Funding acquisition: T.A.H., A.W.L. Supervision: A.W.L. Writing—original draft: T.A.H., A.W.L. Writing—review & editing: T.A.H., O.I., Z.L., A.C.S., M.M.S., T.M., A.W.L.

Supplementary material

Supplementary material is available at The Journal of Immunology online.

Funding

National Institutes of Health grant R01CA238163 (A.W.L.). National Institutes of Health grant T32AI100853 (T.A.H.). National Institutes of Health grant T32CA106195 (M.M.S.). Cancer Research Institute, Lloyd J. Old STAR Award (A.W.L.). American Cancer Society, RSG-18-169-01-LIB (A.W.L.). National Institutes of Health grant P30-CA016087 (Laura and Isaac Perlmutter Cancer Center supporting the Flow Cytometry and Cell Sorting Core and the Genome Technology Center (RRID: SCR_017929). National Institutes of Health grant P30CA069533 (OHSU Knight Cancer Center supporting the Flow Cytometry and Cell Sorting Core).

Conflicts of interest

A.W.L. reports consulting services for AGS Therapeutics. All other authors declare that they have no competing interests.

Data availability

scRNAseq data are deposited in NCBI GEO, GSE223727 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE223727).