Chemokine Receptor–Dependent Control of Skin Tissue–Resident Memory T Cell Formation

T cells are important for protection from pathogens and tumors. Following the clearance of a pathogen, populations of memory T cells develop with different functional, phenotypic, and migratory properties (1). Subsets of memory T cells, referred to as central and effector memory T cells, recirculate via the blood through secondary lymphoid organs, returning to the circulation via the lymphatics. Effector memory T cells can also migrate through nonlymphoid tissues before returning to lymph nodes (LN) via lymphatics. Notably, nonlymphoid tissues also contain a substantial proportion of nonrecirculating, tissue-resident memory T (T_RM) cells (2). These cells can persist for long periods in tissues without exiting into the bloodstream. T_RM cells have been observed in most tissues and organs in mice, and in many tissues in humans.

CD8 and CD4 memory T cell populations are found in tissues, often with clear differences in the localization of these two subsets within the microarchitecture of tissues such as the skin, intestines, lungs, and reproductive tract (1, 3). T_RM cells are defined on the basis of expression of the markers CD69 and the integrin α_E chain, CD103. Additionally, T_RM cells have been shown to express a transcriptional profile distinct from that of recirculating memory T cells and related to that of populations of tissue-resident innate lymphoid cells (4, 5). Notably, T_RM cells in different nonlymphoid tissues possess unique gene expression signatures, indicating that local signals shape the development of memory T cells resident in different tissues.

Upon re-exposure to a pathogen, T_RM cells can contribute to early protection by releasing cytokines that stimulate local innate immunity and help to recruit T cells from the blood (6, 7). Integral to this ability to respond rapidly upon reinfection, T_RM cells within tissues perform key local immunosurveillance functions. These cells do so by migrating constantly within the nonlymphoid tissue environment, increasing the probability of contact with infected cells or APCs (8, 9). In the skin, CD8⁺ T_RM cells are restricted to the epidermis and hair follicle epithelium (10). In contrast to T cells within the dermis that have an amoeboid shape and relatively fast speed of migration (∼6 μm/min), T_RM cells within the epidermis adapt a striking dendritic morphology and slow mode of migration (1–2 μm/min). This slow random migration promotes the long-term persistence of T_RM cells at sites of prior infection in the skin and may enhance local immunosurveillance by enhancing the ability of cells to scan the environment (8, 9).

Although the epidermal environment appears to enforce a constrained mode of migration in skin T_RM cells, what signals control the morphology and motility of these cells are not known. The range of signals that influence T_RM cell formation and persistence within tissues, including the skin, are just beginning to be determined. Signaling through the TGF-β receptor and aryl hydrocarbon receptor, IL-15 signals, as well as CD103 and CD69 expression, are each required for skin T_RM cell development or persistence (5, 11, 12). Furthermore, downregulation of key molecules required for T cell trafficking, including CCR7, S1PR1, and KLF2, as well as regulation of T-bet and Hobit are central to tissue residence by enforcing a gene expression program that prevents migration of T_RM cells out of the tissues (4, 5, 11, 13). Whether other chemokine or cytokine receptors also contribute to skin T_RM cell formation, migration, or persistence needs to be determined.

We have examined signals controlling T_RM cell morphology and motility in the skin of mice, using intravital two-photon microscopy to track T cells in real time. Via the topical application of various inhibitors to the skin, we could directly alter T_RM cell behavior. This revealed key roles for actin polymerization and myosin-II, as well as the Rho-associated coiled-coil kinases (ROCK) for T_RM cell dendritic morphology, and a critical role for microtubules for maintaining T_RM cell shape. Skin T_RM cells required pertussis toxin (PTx)–sensitive signals for motility, yet the chemokine receptors CXCR3, CCR8, CXCR6, and CCR10 were not required. However, CD8⁺ T cells lacking expression of CXCR6 or CCR10 formed reduced numbers of skin T_RM cells, and lack of CCR10 also impaired numbers of circulating memory T cells, demonstrating a role for CCR10 for T cell priming and memory formation after skin infection.

C57BL/6J, B6.SJL-Ptprc^a Pep3^b/BoyJ (B6.CD45.1), gBT-I.CD45.1, gBT-I.EGFP, gBT-I.dsRed, gBT-I.EGFP Itgae^−/− (CD103^−/−), gBT-I.CCR10^−/−, gBT-I.EGFP CCR8^−/−, OT-I.EGFP CXCR3^−/−, and CXCR6^−/− mice were used in this study. gBT-I mice encode transgenes expressing TCR recognizing the HSV glycoprotein B–derived epitope gB_498–505. CCR10-deficient (EGFP knock-in) mice were from Lu and colleagues and C. Gerard (14). Animal experiments were approved by The University of Melbourne Animal Ethics Committee.

Epicutaneous infection with HSV-1 KOS was performed as described elsewhere (15). Briefly, mice were anesthetized with a 1:1 mixture of ketamine and Xylazil (10 μl/g body weight) by i.p. injection. Hair was removed from the left skin flank and depilated (Veet; Reckitt Benckiser). Scarification was performed by lightly abrading the skin using a grindstone tip of a rotary tool (Dremel) and infected with 10⁶ PFUs of HSV.

Formulations of inhibitors were prepared for transcutaneous delivery. Cytochalasin D (20 μg), blebbistatin (50 μg), or Y27632 (50 μg) were mixed with acetone (in an 80 μl vol). Colchicine (50 μg) or PTx (100 ng) was added to 100 mg of Sorbolene cream and mixed thoroughly. Compounds mixed in acetone were applied to the skin using a 200-μl micropipette. Compounds mixed in Sorbolene were applied and the mixture was contained on the skin using a bandage consisting of an adhesive film (OpSite Flexigrid; Smith & Nephew) and surgical tape (Micropore tape, 3M). The mixture was allowed to absorb through the skin for 3 h prior to removal of the bandage. Mice were imaged 8–10 h after application.

Mice were euthanized by CO₂. Mice were perfused with PBS and 1 cm² skin regions were excised. Skin samples were placed into dispase (2.5 mg/ml in PBS) and incubated at 37°C for 90 min, then placed into collagenase type III (3 mg/ml in RPMI 1640 plus 10% FCS and 5 μg/ml DNase I), finely chopped and further incubated at 37°C for 30 min. Samples were transferred into RPMI 1640 plus 10% FCS, and vigorously mixed. Cells were successively filtered through 70- and 30-μm nylon meshes and resuspended in FACS buffer (PBS with 2% FCS, 5 mM EDTA, 0.1% NaN₃), prior to Ab staining. LN and/or spleens were harvested and pushed through a 70-μm sieve to obtain single-cell suspensions and resuspended in FACS buffer for Ab staining. Propidium iodide was added to samples prior to acquisition by flow cytometry (BD FACSCanto or LSRFortessa). Data were analyzed using FlowJo software (Tree Star). The following Abs were purchased: BD Pharmingen: Vα2-PE-Cy7 (B20.1) and CD69-PE ( 1H.2F3); eBioscience: CD3e-PerCP-Cy5.5 (145-2C11), CD44–Alexa Fluor 700 (IM7), CD8α–allophycocyanin–eFluor 780 (53-6.7), CD62L-PerCP-Cy5.5 (MEL-14), CD45.1-FITC (A20), CD45.1-PE (A20), and CD45.2–Alexa Fluor 700 (104); and BioLegend: CD103-allophycocyanin (2E7).

gBT-I T cells were incubated with 5 μM CellTrace Violet at 37°C for 10 min and washed in complete media. T cells were then adoptively transferred i.v. into recipient mice prior to infection as described above. For in vitro–generated effector CD8⁺ T cells, splenocyte samples enriched for CD8⁺ T cells were activated by anti-CD3ε (5 μg/ml; clone 145-2C11; eBioscience) and anti-CD28 (5 μg/ml; clone 37.51). Activated CD8⁺ T cells (5 × 10⁶) were transferred at a 1:1 ratio of wild-type (WT) and CXCR3^−/− or CXCR6^−/− cells into recipients by intradermal injection (five 20-μl injections over an area of skin 1 × 1.5 cm²) with a 30-gauge needle. Mice were shaved and depilated before the application of 15 μl 2,4-dinitrofluorobenzene (DNFB) in acetone/oil (4:1) to a 1-1.5 cm² area of skin 24 h prior to T cell transfer.

Intravital imaging of the skin flank was performed as described elsewhere. Briefly, mice were anesthetized with isoflurane (Cenvet; 2.5% for induction, 1–1.5% for maintenance, vaporized at 80:20 mixture of O₂ and air), shaved on the left flank, and hair was depilated. Two incisions (∼15 mm apart) were made longitudinally along the left flank to contain the herpes lesion site. Subcutaneous connective tissue below the dermis was carefully cut away from the dermis and an 18-mm-wide × 1-mm-thick stainless steel platform was inserted under the exposed dermis, upon which the underside of the dermis was attached using Vetbond tissue adhesive (3M). The edges of the skin were lined with vacuum grease (Dow Corning), warm PBS was placed on top of the skin, and a glass coverslip was carefully positioned. Imaging was performed with an upright LSM 710 NLO multiphoton microscope (Carl Zeiss) with a ×20/1.0 numerical aperture water immersion objective enclosed in an environmental chamber (Precision Plastics) maintained at 35°C with heated air. Fluorescence excitation was provided by a Chameleon Vision II Ti:Sapphire laser (Coherent) with dispersion correction and fluorescence emission detected using external nondescanned photomultiplier tubes. EGFP and second harmonic generation (SHG) were excited at 920 nm, whereas samples with dsRed and EGFP were excited at 900 nm. For four-dimensional data sets, three-dimensional z-stacks (40 μm thick) were acquired every 60 s for 45–60 min. Raw imaging data were processed using Imaris (Bitplane). Cells were surface-rendered using Imaris surface module to generate sphericity data. Videos generated in Imaris were composed in After Effects (Adobe).

Comparison of data sets was performed using Kruskal–Wallis tests with a Dunn posttest or Mann–Whitney U test where appropriate.

T cells that enter the epidermis acquire a dendritic morphology and low migrational velocity that become more pronounced after the resolution of infection (8–10). The shape of skin CD8⁺ T_RM cells is highly dynamic during the process of migration within the constrained epidermal layer, and it appears unique from the mode of migration used by T cells in other tissues (16). We performed intravital two-photon microscopy on the flank skin of anesthetized mice to visualize HSV-specific gBT-I CD8⁺ T_RM cells in the epidermis. We observed elongated cell bodies and persistent extension and retraction of dendritic protrusions from the T_RM cells as they navigated the epidermis (Fig. 1A). Many of the smaller protrusions appeared to resemble lamellipodia, suggesting that the cells probed their environment to move and make directional decisions.

To best understand the molecular mechanisms controlling the behavior of skin CD8⁺ T_RM cells, in situ analysis of cell movement was required. To influence this behavior, we applied various compounds topically to the skin of mice. We developed a protocol to apply nonaqueous compounds dissolved in acetone topically to the shaved skin of mice. Skin T_RM cells treated with vehicle alone (acetone) displayed no change in their morphology or motility (Fig. 1B) compared with untreated T_RM cells (8). Because polymerization of actin is crucial for cell migration and morphology, in particular to control cell shape and cytoplasmic projections (17), we first examined the effect of disrupting the actin cytoskeleton of skin T_RM cells. Examination of the skin of the mice after treatment did not reveal obvious changes or disruption of the epidermis or underlying SHG⁺ dermis. Compared to mice treated with vehicle alone, treatment with the actin depolymerization compound cytochalasin D induced marked rounding of the T_RM cells and a significant reduction in cell motility (Fig. 1B–E, Supplemental Video 1). Because T cells also require myosin-II for motility, at least for amoeboid modes of migration (18), we treated the skin of mice with the compound blebbistatin to inhibit myosin-II function. T_RM cells rounded up and reduced motility after blebbistatin treatment, although some cells retained the ability to extend some small dendritic projections, albeit with a reduced motility (Fig. 1B–E). We then examined a role for the Rho GTPase ROCK for T_RM cell motility. ROCK has been shown to play an important role in T cell migration within constrained environments by activating myosin-IIA–mediated actomyosin contraction (19). In dendritic cells ROCK is also important for dendrite formation (20). Treatment of skin with the ROCK inhibitor Y27632 induced reduced formation of dendritic projections by the T_RM cells and inhibited motility (Fig. 1). We also treated mice with a Rac1 inhibitor, NSC1, to determine a role for branched actin formation and membrane protrusions; however, no difference was observed between treated and untreated mice (data not shown). This could be due to compensation by Rac2, which is also required for T cell migration (21).

Next, we examined a role for microtubules for T_RM cell migration. Microtubules are important for organization of cytoskeletal structures to stabilize cell uropods and lamellopodia, and microtubule retraction may help cells to migrate through confined spaces (22, 23). Treatment of T_RM cells in the skin with the microtubule destabilization agent colchicine resulted in a marked change in the morphology and behavior of the cells (Fig. 1C–E, Supplemental Video 1). We observed marked elongation of T_RM cells, some extending dendrites for 30–40 μm, yet retaining a rounded and sessile cell body. We also observed rounded cells and substantial cell debris that was indicative of cell death. Although some cells appeared to migrate normally after treatment, we were unable to confirm that the compound evenly penetrated across the skin. Notably, we saw similar effects on skin T_RM cell morphology and migration after treatment with another microtubule-destabilizing agent nocodazole (Supplemental Fig. 1A). These data show that skin T_RM cells use actin-driven migration machinery, including the Rho GTPase ROCK and myosin-II, to adopt a unique dendritic morphology and migrate within the epidermis. Notably, skin T_RM cells also require an intact microtubule network to migrate and maintain cellular integrity within the epidermis.

T cells resident in the epidermis after infection or inflammation express high levels of the integrin α_E chain, CD103 (24), which binds to the ligand E-cadherin. Although CD8⁺ T cells that lack expression of CD103 can initially enter the epidermis and develop into T_RM cells, persistence of these cells over time is impaired (5). It was previously demonstrated that skin dendritic epidermal γδT cells lacking CD103 had fewer dendrites in vivo and showed slower motility in vitro (25). To investigate the role for CD103 expression by CD8⁺ T_RM cells, we cotransferred CD103^−/− and WT HSV-specific TCR transgenic CD8⁺ T cells from gBT-I mice into WT recipients. Following infection with HSV, we imaged both WT and CD103^−/− epidermal T_RM cells. CD103^−/− T_RM cell morphology was equivalent to that of WT cells (Fig. 2A, Supplemental Video 2). Tracking of the cells revealed greater motility by CD103^−/− T_RM cells compared with WT cells (Fig. 2B). These data suggest that CD103 expression may contribute to T_RM cell adhesion in the epidermis and thereby influence local retention. However, the unique cell morphology of skin T_RM cells was not influenced by expression of CD103.

We next sought to determine other signals capable of influencing T_RM cell formation and behavior in the skin. We treated the flank skin of mice with PTx dissolved in Sorbolene cream to inhibit G_αi protein–coupled receptors. In contrast to vehicle treatment, PTx induced rounding of skin CD8⁺ T_RM cells and a substantial loss of dendritic projections (Fig. 2C, 2D, Supplemental Video 3). Some T_RM cells retained the ability to extend cytoplasmic projections and “wiggle” their cell bodies back and forth. However, the overall behavior of the cells was markedly altered, indicating that G protein–coupled receptors are required for T_RM cell morphology and motility.

Because PTx can inhibit chemokine receptor signaling, which may contribute to T_RM cell behavior, we examined the roles of chemokine receptors involved in CD8⁺ T cell responses and homing to the skin. We recently demonstrated that expression of CXCR3 on effector T cells was important for entry of T_RM cell precursors into the epidermis, and fewer T_RM cells formed from CXCR3^−/− T cells (5). We sought to determine whether the migration of skin T_RM cells or their dendritic morphology required expression of CXCR3. We transferred equal numbers of in vitro–activated WT and CXCR3^−/− T cells into WT mice and then treated the flank skin with the contact-sensitizing agent DNFB to recruit T cells into the skin where they formed populations of CD69⁺CD103⁺ T_RM cells (Supplemental Fig. 1B) (26). Intravital imaging revealed that T_RM cells lacking CXCR3 displayed equivalent cell morphologies and motility compared with WT cells (Fig. 3A, 3B, Supplemental Video 4), suggesting that CXCR3 signals are not required for T_RM cell migration in the epidermis.

We next examined the role of CCR8, because we found that this chemokine receptor was highly expressed specifically by skin T_RM cells (5). We examined whether the migration or morphology of skin T_RM cells was altered in the absence of CCR8. However, we observed no difference in the morphology of the two populations, and both displayed a similar rate of motility in the skin (Fig. 3C, 3D, Supplemental Video 4). We then cotransferred WT and CCR8^−/− gBT-I CD8⁺ T cells into mice prior to skin HSV infection and examined T cell responses in the draining brachial LN (bLN), spleen, and skin. Equivalent numbers of WT and CCR8^−/− CD8⁺ T cells were found in all tissues 10 d after skin HSV infection (Fig. 3E), indicating that expression of CCR8 was not required for T cell responses or migration to the skin. Likewise, numbers of skin T_RM cells that formed after infection were equivalent between populations of WT and CCR8^−/− T cells. Therefore, despite increased expression of CCR8 by skin T_RM cells, we did not find a role for expression of this chemokine receptor for formation, persistence, or migration of these memory T cells.

T cells resident in several tissues, including the skin, express the chemokine receptor CXCR6 (4). The ligand for CXCR6, CXCL16, can be expressed by keratinocytes as well as dendritic cells in various tissues. T_RM cells in the liver express high levels of CXCR6, and expression of this receptor contributes to maintenance of memory cells in this site (27, 28). We examined a role for this chemokine receptor using homozygous CXCR6^GFP/GFP knock-in mice (CXCR6^−/−), comparing T cells from these mice with WT cells. T_RM cells generated following DNFB treatment were imaged in the skin, revealing that cells lacking CXCR6 showed more rounded cell morphologies, whereas migration in the epidermis was unaffected (Fig. 4A, 4B, Supplemental Video 4). We also observed a marked reduction in numbers of CXCR6^−/− T_RM cells in the epidermis (WT, 42.3 ± +7.5 cells/mm²; knockout [KO], 6.8 ± 1.6 cells/mm² skin from six mice).

To examine the influence of CXCR6 expression on memory T cell formation, we cotransferred activated WT and CXCR6^−/− CD8⁺ T cells into mice prior to DNFB treatment and examined T cell responses. Equivalent numbers of WT and KO CD8⁺ T cells were found in the spleen (Fig. 4C). However, numbers of CXCR6^−/− T cells were reduced in the skin by day 10 and significantly fewer CD69⁺CD103⁺ T_RM cells formed in the skin. Intradermal injection of in vitro–activated CXCR6^−/− CD8⁺ T cells to distinguish potential defects in recruitment versus retention of these cells in the skin revealed decreased numbers of T_RM cells compared with WT T cells (Fig. 4D). The CXCR6^−/− and WT T cells were present in similar numbers in the spleen, indicating that poor survival of the CXCR6^−/− T cells may have contributed to the reduced retention and formation of T_RM populations in the skin. These data identify an important role for the chemokine receptor CXCR6 for the formation of T_RM cells in the skin.

We next examined a role for CCR10 in T_RM cell formation and migration in the skin. CCR10 is a chemokine receptor implicated in the homing of T cells to the skin (29), and CD8⁺ T cells have been shown to be reduced in the skin of CCR10-deficient mice (30). The ligand for CCR10, CCL27, is highly expressed in the epidermis (29). To examine CCR10 in the context of skin infection, WT and CCR10^−/− gBT-I T cells were cotransferred into CCR10^+/− mice prior to HSV infection and 30 d later, the memory T cell populations that formed in the skin were imaged by intravital two-photon microscopy. The morphology and motility of skin CCR10^−/− and WT T_RM cells was indistinguishable (Fig. 5A, 5B, Supplemental Video 4). These results indicated that expression of CCR10 by skin T_RM cells was not required for migration. However, we observed that numbers of CCR10^−/− T_RM cells were slightly reduced (WT, 71 ± 7.5 cells/mm²; KO, 24.7 ± 3.8 cells/mm² skin from seven mice).

To directly determine whether CCR10 contributed to memory T cell formation in the skin after HSV infection, CCR10^+/− and CCR10^−/− gBT-I CD8⁺ T cells were cotransferred in equal numbers into CCR10^+/− mice and T cell responses were examined in the draining bLN, spleen, and skin. Ten days after infection, numbers of CCR10^−/− and CCR10^+/− gBT-I T cells were not significantly different in the spleen and skin, although based on mean values 2- to 3-fold more CCR10^+/− gBT-I T cells were present in these tissues compared with CCR10-deficient cells (Fig. 5C, 5D). Notably, by day 30, numbers of CCR10^−/− memory gBT-I T cells were significantly reduced in the spleen and skin, and this was most pronounced among T_RM cells in the skin, which contained ∼12-fold more CCR10^+/− gBT-I T cells. This was not due to a delay in the activation or expansion of the CCR10^−/− T cells, as proliferation was similar between WT and KO CD8⁺ T cells in the draining bLN 3 d postinfection (Fig. 5E). Importantly, CD8⁺ T cells responding to skin infection upregulated CCR10 expression by the first division, with expression among virus-specific T cells highest at the peak of the response (day 7) (Fig. 5F, 5G). CCR10 expression was downregulated on most memory T cells in lymphoid organs and skin >30 d postinfection, indicating that CCR10 expression by CD8⁺ T cells controls effector and memory cell responses during skin virus infection, possibly by promoting their migration and survival.

T_RM cells that form in the skin after infection play an important role in rapid protection from reinfection (2). Integral to this role, T_RM cells must accomplish effective immunosurveillance of their tissue microenvironment in the search for invading pathogens. The epidermis presents a highly constrained environment for the T cells to navigate, and we have shown that T_RM cells adopt a striking morphology that is accompanied by a slow mode of migration (8). Interestingly, epidermal γδT cells and Langerhans cells also adopt dendritic cell morphologies (31), although both of these cell subsets extend thinner and more defined dendrites from largely immotile cell bodies. In contrast, epidermal T_RM cells remain motile and extend and retract many cellular protrusions as they move.

To facilitate dissection of the molecular mechanisms of T_RM cell morphology, we treated the skin of mice with selected compounds to determine the effects on T cell behavior in situ. The actin cytoskeleton was integral to skin T_RM cell migration and for the dynamic cell morphology that T_RM cells presumably use to probe their environment and generate adhesions to help propel the cells forward. We do not rule out potential effects of topical inhibitor treatment on surrounding keratinocytes, although we predict this would be secondary to the T cell–intrinsic roles of the cytoskeletal components. Similar to T cells migrating in less constrained tissues such as LN, myosin-II was required for T_RM cell migration in the epidermis. Myosin-II might therefore help T_RM cells to navigate the epidermis by preventing prolonged adhesion to the environment, much like it does in LN (18). Thus, T cell migration within the confined epidermis requires the actomyosin cytoskeleton to facilitate the probing dendritic morphology that is required for navigation in this constrained microenvironment.

Our experiments demonstrate a key requirement for microtubules for T_RM cell migration in the skin. Memory T cells treated with a microtubule-destabilizing drug continued to stretch and project dendrites, but were observed to often overextend and to disintegrate or lose pieces of cytoplasm. These stretched T_RM cells maintained a sessile cell body, suggesting that microtubules were required to propel the cells forward. We recently found that the guanine nucleotide exchange factor DOCK8, through CDC42 and p21-activated kinase, is required to coordinate cytoskeletal structures (32). T cells lacking DOCK8 stretched and died (cytothripsis) within constrained environments and were unable to form T_RM cell populations in the skin. This disturbed cytoskeletal behavior was accompanied by abnormal positioning of microtubules. We also show in the present study that T_RM cells required ROCK for migration in the epidermis. It has been shown that ROCK can influence actomyosin contractility as well as microtubule stability (22). Microtubules have also been proposed to contribute to the deformability of T cells in confined spaces (23). Together with our data in the present study, this suggests that microtubules are critical for the proper coordination of T cell migration in tight cellular environments such as the epidermis. Whether T cells also require microtubules for migration in other tissues is less clear, although we would predict that this would be highly dependent on the tissue microenvironment.

We found that although required for long-term retention of skin T_RM cells, CD103 expression contributed to cell adhesion but did not control cell morphology. This is in contrast to skin dendritic epidermal γδT cells that were shown to have an altered morphology in the absence of CD103 (25). We also found that PTx-sensitive signals are required for this cellular behavior. Contrary to our prediction that chemokine receptors expressed by skin T_RM cells (CCR8, CXCR3, and CXCR6) would influence the motility or morphology of the cells, we observed no difference in responses by T_RM cells lacking CCR8 or CXCR3, and only a minor influence on cell morphology by CXCR6. The role that increased CCR8 expression on skin T_RM cells versus circulating memory T cell subsets plays is unclear. Potentially, this receptor is important for recall responses or other functions of T_RM cells upon reinfection, similar to the function of CCR8 in regulatory T cells in tissues (33).

We observed a marked reduction in numbers of skin T_RM cells lacking expression of CXCR6. This suggests that CXCR6 is required for the formation or maintenance of T_RM cells in the skin, as previously shown for memory CD8⁺ T cells in the liver (28). Intradermal injection of CXCR6^−/− T cells revealed reduced numbers of skin T_RM cells, but similar numbers in the spleen, which may indicate a role for this chemokine receptor for local survival. In contrast, no role for the expression of CCR8 by skin T_RM cells could be determined in our experiments. Additionally, CCR10 was important for the priming of effector and memory T cell populations after skin HSV infection. CCR10 is implicated in the migration of T cells into the skin during inflammation (29), and early during inflammation the CCR10 ligand CCL27 can be detected in skin-draining LN (34). However, a role for CCR10 in CD8⁺ T cell responses to virus infection has not been shown. In humans, the vitamin D3 metabolite 1,25(OH)₂D₃ is processed by dendritic cells and induces CCR10 expression in T cells (35). In mice, this pathway may involve further signals to induce CCR10 on T cells in skin-draining LN. Our results demonstrate CCR10 upregulation on virus-specific CD8⁺ T cells during the early phase of the immune response. Responses by CCR10-deficient gBT-I CD8⁺ T cells were impaired, with accumulation of CCR10-deficient T_RM cells substantially impaired in the skin, suggesting reduced retention or survival of these cells. However, neither circulating memory T cells nor skin T_RM cells expressed CCR10, indicating a likely role for CCR10 during the effector phase in response to skin infection. Whether CCR10 signals are important to expand or sustain CD8⁺ T cells after other skin infections is unclear and will require further investigation.

In summary, we have shown that skin T_RM cells adapt to the constrained epidermal environment and migrate using pronounced cellular projections that require functional actomyosin contractility and support from microtubules. Skin T_RM cells also required signals from PTx-sensitive receptors for migration and cellular morphology. Although the integrin CD103 and chemokine receptors CCR8, CXCR3, CXCR6, and CCR10 were not critical for epidermal T_RM cell migration, formation of skin memory T cells was impaired when CD8⁺ T cells lacked expression of CXCR6 or CCR10. Further work will be needed to define the roles of these chemokine receptors in T cell migration, survival, and immune responses in peripheral tissues.

We thank C. Jones, G. Davey, and M. Damtsis for technical assistance.

The authors have no financial conflicts of interest.