PD-1 is requisite for skin TRM cell formation and specification by TGFβ

Devi, K. Sanjana P.; Wang, Eric; Jaiswal, Abhinav; Konieczny, Piotr; Kim, Tae-Gyun; Nirschl, Christopher J.; Verma, Akanksha; Liu, Yong; Milczanowski, Julia; Christo, Susan N.; Gandolfo, Luke C.; Haitz, Karyn; Vardam, Trupti D.; Wu, Pinru; King, Sandra L.; Tse, Sze-Wah; Pradhan, Komal; Jiang, Xiaodong; Tian, Tian; Fuhlbrigge, Robert C.; Schmults, Chrysalyne D.; Clark, Rachael A.; Kupper, Thomas S.; Freeman, Gordon J.; Mackay, Laura K.; Naik, Shruti; Newell, Evan W.; Elemento, Olivier; Suarez-Farinas, Mayte; Anandasabapathy, Niroshana

doi:10.1038/s41590-025-02228-1

Download PDF

Article
Open access
Published: 29 July 2025

PD-1 is requisite for skin T_RM cell formation and specification by TGFβ

Nature Immunology volume 26, pages 1339–1351 (2025)Cite this article

18k Accesses
15 Citations
67 Altmetric
Metrics details

Subjects

Abstract

Tissue-resident memory T (T_RM) cells provide infectious, cancer and vaccine-trained immunity across barrier sites. T_RM cells are implicated in autoimmunity, successful response to immune checkpoint blockade in the tumor microenvironment and toxicities that occur after immune checkpoint blockade in peripheral tissues. Here, we identified that signaling through the immune checkpoint programmed death receptor 1 (PD-1) strongly impacts the early specification of CD8⁺ T_RM cells in the skin. PD-1 is expressed broadly across mouse and human skin T_RM cells, in the absence of persistent infection, and is retained on skin T_RM cells in aged mice. PD-1 supports early T_RM cell colonization, skin-specific programming and silencing of other differentiation programs and promotes TGFβ responsivity and skin engraftment. Thus, PD-1 signaling mediates skin T_RM cell specification during immune initiation. These findings may inform therapeutic PD-1 agonist and antagonist use to modulate successful peripheral memory.

Transient regulatory-T-cell interruption promotes skin-resident memory T cells mediated tumor protection

Article Open access 05 July 2023

PD-1 maintains CD8 T cell tolerance towards cutaneous neoantigens

Article 21 June 2023

The precursors of CD8⁺ tissue resident memory T cells: from lymphoid organs to infected tissues

Article 03 September 2021

Main

Generating and maintaining immune memory is essential to protecting pathogen- and injury-exposed barrier sites. Tissue-resident memory T (T_RM) cells are immune sentinels generated after infection (for example, herpes simplex virus (HSV)¹, vaccinia virus² and lymphocytic choriomeningitis virus³) and inflammation⁴ in barrier tissues, including the skin^5,6, lung⁷ and mucosa⁸. T_RM cells provide long-lived protection against pathogen re-encounter. T_RM cells contribute to antitumoral immunity⁹ and cancer–immune equilibrium^9,10 and may limit metastases¹¹. For continuous tissue surveillance, T_RM cells home, engraft, persist and renew locally. Governed T_RM cell formation may enforce homeostasis and limit autoinflammation. Understanding instructive pathways regulating T_RM cell formation, repertoire, niche selection, maintenance and recall is requisite to therapeutically modulating T_RM cell activity, for example, by vaccination or adoptive cell therapy. This would enable strategies to improve cancer, vaccine and antiviral immunity or limit T_RM cell recall-driven tissue pathology.

Programmed death receptor 1 (PD-1) restrains T cell effector activity and is expressed during exhaustion/dysfunction due to chronic T cell antigen receptor (TCR) engagement¹². Less appreciated roles include shaping stemness, metabolism¹³ and regulatory T cell activity by blocking the AKT–mTOR pathway¹⁴. T_RM cells express PD-1 (ref. ¹⁵), but how PD-1 signaling shapes T_RM cell biology is unknown. Drugs inhibiting the immune checkpoint receptor PD-1 and its ligands PD-L1 (CD274) and PD-L2 (CD273) have transformed cancer care. In individuals with cancer, CD3⁺ tumor-infiltrating lymphocytes (TILs) with RNA transcriptomes similar to late T_RM cell and circulating memory T cell programs expand after therapy in metastatic melanoma sites responsive to inhibitory anti-PD-1 therapy. This suggests that anti-PD-1 efficacy depends on successful memory formation and expansion¹⁶. High T_RM cell signature tumor scoring predicts melanoma survival in 15-year cohorts naive to immune checkpoint blockade and 5-year survival in recipients of anti-PD-1 (ref. ¹⁶). How PD-1 signaling impacts memory-like TIL formation, why memory differentiation and PD-1 associate with improved outcomes and how and why anti-PD-1 treatment expands these populations remains unknown. Knowledge of PD-1 biology during T_RM cell and memory differentiation could inform strategies to improve immunity in nonresponders, chimeric antigen receptor T cells and adoptive T cell strategies. PD-1 modulation could also improve immunity from personalized vaccines.

Although inhibitory anti-PD-1 boosts antitumoral immunity, disrupted peripheral tolerance also occurs. Off-target inflammatory immune-related adverse events (irAEs) occur in the tissues of recipients of anti-PD-1 therapy, which limit cancer care. irAEs occur more frequently after anti-PD-1 treatment than in the general population¹⁷. T_RM cells drive the same autoimmune/inflammatory diseases in the general population^4,18. irAEs manifest in T_RM cell-rich tissues, raising the possibility that T_RM cells instigate these reactions. In individuals with solid tumors, the skin is the earliest and most common tissue site affected in roughly 40% of individuals with melanoma receiving anti-PD-1 and 70% when combined with anti-CTLA-4 (refs. ^19,20). Moreover, recipients of anti-PD-1 with rash have improved clinical outcomes²¹. This suggests a relationship between peripheral tissue reactivity and tumor response that is yet undefined.

Defining PD-1 function on T_RM cells may illuminate why anti-PD-1 targets these cells. Here, we observed a critical role for PD-1 in establishing skin CD8⁺ T_RM cells early after infection, whereas effects on circulating memory CD8⁺ T cells occurred later after infection. Using pharmacologic inhibition and genetic loss, we observed that PD-1 enforced early viral skin CD8⁺ T_RM cell specification, lineage fate commitment, tissue niche engraftment and transforming growth factor-β (TGFβ) responsivity.

Results

PD-1 is expressed in mouse and human skin T_RM cells

To identify anti-PD-1-responsive populations, we performed high-resolution mass cytometry (CyTOF) on skin from individuals undergoing abdominoplasty or mastectomy (n = 5, 2 men and 3 women; age = 26–52 years) and peripheral blood mononuclear cells (PBMCs) from healthy individuals (n = 4; 2 men; age = 38 and 39, 2 unknown)²². T_RM cells express CD69, which enables tissue residence by suppressing the functions of the lymphocyte trafficking regulator S1P1 (ref. ²³) and CD103, which binds E-cadherin²⁴. We observed PD-1 cell-surface enrichment on human skin CD4⁺ and CD8⁺CD69⁺ (skin-resident), CLA⁺ (skin homing) and CD45RO⁺ (antigen-experienced) T_RM cells from the epidermis (CD103⁺) and dermis (CD103⁻; Fig. 1a–c and Extended Data Fig. 1a,b). Skin PD-1⁺ T_RM cells expressed CD49a¹⁸, the chemokine receptors CXCR3 and CXCR6 (ref. ²⁵) and low levels of CD62L and CX3CR1 (Fig. 1d and Extended Data Fig. 1c), supporting a resident memory phenotype. Across human tissues, we compared PD-1 CyTOF intensity in CD45RO⁺CD45RA⁻CD69⁺CD49a⁺CD8⁺ T_RM cells or all antigen-experienced CD45RO⁺CD45RA⁻CD8⁺ T cells against PBMCs. Expression of PD-1 was higher on human skin, lung and tonsil CD8⁺ T_RM cells than on PBMCs (Fig. 1e).

Fig. 1: PD-1 is retained on mouse and human CD8+ TRM cells in the absence of persistent antigen. — **Fig. 1: PD-1 is retained on mouse and human CD8⁺ T_RM cells in the absence of persistent antigen.**

PD-1 restrains early T cell activation²⁶. To determine when PD-1 was expressed during CD8⁺ T_RM cell specification, we immunized C57BL/6J mice by skin scarification using vaccinia virus expressing ovalbumin (OVA) protein (VACV-OVAss) 1 day after intravenous transfer of 200,000 antigen-specific OVA-reactive CD8⁺ OT-I cells. PD-1 was rapidly expressed on activated OT-I cells (Extended Data Fig. 1d,e)²⁶ and retained on skin T_RM cells at week 6 after infection (late memory, after virus clearance; Fig. 1f–h and Extended Data Fig. 1f,g). The fraction of PD-1-expressing KLRG1⁺ OT-I effector T (T_eff) cells underwent rapid contraction in lymph nodes (LN) and spleen (Fig. 1i). Unlike the rapid decline of PD-1-expressing KLRG1⁻CD62L⁺ circulating memory (T_CM) and KLRG1⁻CD62L⁻ effector memory (T_EM) OT-I cells at week 1 after infection in LNs and spleen (Fig. 1i and Extended Data Fig. 1h), we observed a gradual contraction between weeks 1 and 6 of PD-1⁺CD69⁺CD103^+/−CD8⁺ T_RM cells in the skin (Fig. 1i, Extended Data Fig. 1h and Supplementary Fig. 1a,b). Both epidermal CD69⁺CD103⁺ and dermal CD69⁺CD103⁻ CD8⁺ T_RM cells expressed PD-1 (Fig. 1i). Over 80% of skin OT-I cells were KLRG1⁻ at week 3 (Supplementary Fig. 1c,d), consistent with CD8⁺ T_RM cells deriving from KLRG1⁻ progenitors²⁵. Skin KLRG1⁻ OT-I cells retained PD-1 at week 6 (Supplementary Fig. 1c–e). We observed that PD-1 was expressed on epidermal and dermal CD8⁺ T_RM cells in both virally immunized sites and distal, nonimmunized skin sites (Extended Data Fig. 1i,j), reflecting broad CD8⁺ T_RM cell seeding. PD-1 was also expressed on epidermal and dermal CD8⁺ T_RM cells at day 657 after infection in five of six immunized aged mice (Fig. 1j,k). Collectively these data suggest that PD-1 is broadly expressed in human skin CD4⁺ and CD8⁺ T_RM cells and on mouse T_RM cells long after infection is cleared and at sites that were never infected.

PD-1 supports T_RM cell formation in skin

To test if PD-1 signaling impacted the formation of skin CD8⁺ T_RM cells, we transferred an equal mix of 100,000 wild-type Pdcd1^+/+ (hereafter PD-1^+/+) CD45.2⁺Thy1.2⁺ and Pdcd1^−/− (hereafter PD-1^−/−) CD45.2⁺Thy1.1⁺ OT-I cells into CD45.1 Thy1.2 C57BL/6J mice intravenously, followed by VACV-OVA_SS 1 day later (Fig. 2a and Supplementary Fig. 2a). At day 10 after infection, the number of PD-1^+/+ OT-I cells was 2.5-fold higher than PD-1^−/− OT-1 cells in infected skin (Fig. 2a). At day 21 after infection, PD-1^+/+ OT-I cells contributed preferentially to T_RM cell numbers (5.8-fold more than PD-1^−/− OT-I T_RM cells; Fig. 2a), indicating that PD-1 expression offered a competitive advantage in the skin. After viral clearance, at day 42 after infection, 90% of total skin CD69⁺ OT-I cells were derived from the PD-1^+/+ donor (mean total of 680 PD-1^+/+ versus 60 PD-1^−/− OT-I cells) with 6.8-fold more donor T_RM cells deriving from the PD-1^+/+ donor than the PD-1^−/− donor (Fig. 2a). This impacted both dermal and epidermal T_RM cells (86% of CD69⁺ skin OT-I cells from the PD-1^+/+ donor rather than the PD-1^−/− donor, mean total of 260 PD-1^+/+ versus 38 PD-1^−/− CD69⁺ OT-I T_RM cells; Fig. 2a–c and Supplementary Fig. 2b–d). We also observed higher numbers of PD-1^+/+ than PD-1^−/− KLRG1⁻CD62L⁺CD69⁻CD103⁺ OT-I T_CM cells in the skin, which are known to be TGFβ dependent²⁷, at days 21 and 42 after infection (Supplementary Fig. 2d). To control for any minor major histocompatibility complex mismatch driving this effect, we also co-transferred 100,000 PD-1^+/+ CD45.2⁺Thy1.2⁺ OT-I cells with 100,000 PD-1^+/+ CD45.2⁺Thy1.1⁺ OT-I cells derived as littermates of the PD-1^−/− OT-I donor mice into CD45.1 Thy1.2 C57BL/6J mice, followed by VACV-OVAss (Fig. 2a). No differences in quantitation of mixed PD-1^+/+ and PD-1^+/+ CD69⁺ OT-I cells were seen at day 10, 21 or 42 after infection (Fig. 2a). However, the number of PD-1^+/+ OT-I cells at day 42 was twice as high when mixed with PD-1^−/− OT-I cells than when mixed with PD-1^+/+ littermate OT-I cells (mean ± s.e.m.: 680 ± 150 cells versus 215 ± 62 cells; Fig. 2a, and data not shown), suggesting a strong selection bias toward PD-1 sufficiency. There was no difference in the number of circulating naive PD-1^+/+ and PD-1^−/− CD69⁻CD103⁻CD8⁺ T cells at days 10 and 21 after infection, with slightly higher numbers of PD-1^+/+ CD69⁻CD103⁻CD8⁺ naive T cells noted at day 42 than PD-1^−/− cells, when the number of circulating T cells in skin was low (Supplementary Fig. 2d). Analysis of fractional differences of PD-1^+/+ or PD-1^−/− OT-I cells in epidermal and dermal T_RM cells and circulating memory and naive CD8⁺ T cells indicated variance in individual mice, as expected during a live viral infection, but no major distribution differences were seen between groups (Supplementary Fig. 2e,f). The numbers of PD-1^−/− circulating total OT-I cells and KLRG1⁻CD62L⁺ T_CM cells were higher than the numbers of PD-1^+/+ counterparts at day 10 after infection in the LNs, but PD-1^−/− OT-I cells underwent rapid contraction to below PD-1^+/+ OT-I cell numbers in the spleen for total OT-I, KLRG1⁻CD62L⁺ T_CM and KLRG1⁻CD62L⁻ T_EM cells by day 42 after infection (Extended Data Fig. 2a–c). Thus, at week 6, we observed higher numbers of PD-1^+/+ OT-I cell-derived KLRG1⁻CD62L⁺CD8⁺ T_CM and KLRG1⁻CD62L⁻CD8⁺ T_EM cells in the LN and spleen (Extended Data Fig. 2a–d).These data suggest that spleen and LN CD8⁺ T_CM cells are also PD-1 dependent, but their formation is impacted much later after infection than skin T_RM cells.

Fig. 2: PD-1 is required for epidermal and dermal TRM cell formation and specifies TRM cell fate. — **Fig. 2: PD-1 is required for epidermal and dermal T_RM cell formation and specifies T_RM cell fate.**

To investigate whether the low number of PD-1^−/− OT-I-derived skin T_RM cells was caused by differences in cell proliferation or death in the draining LNs (dLNs) at days 3, 5 and 7, we mixed PD-1^+/+ and PD-1^−/− OT-I cells and transferred them intravenously into wild-type C57BL/6J mice that were immunized by VACV-OVAss 1 day later and injected intravenously with BrdU at days 0–2, 0–4 or 0–6 after infection. At days 3, 5 and 7 after infection, we assessed BrdU DNA incorporation, which replaces thymidine during S phase and Ki67 staining of G1, S, G2 and M cell cycle stages, to assess cell division and live/dead blue uptake to measure cell death. Ki67 counts were higher than BrdU counts, as it encompassed more cell cycle stages (Extended Data Fig. 3a–c). A ratio of roughly 60% PD-1^−/− to 40% PD-1^+/+ OT-I cells was observed in dLNs at days 3, 5 and 7 after infection, with a small but significant increase in PD-1^−/− OT-I cell numbers on days 5 and 7 (Extended Data Fig. 3a). BrdU incorporation was slightly higher in PD-1^+/+ OT-I cells than in PD-1^−/− OT-I cells at day 3 after infection, and live/dead counts were slightly higher in PD-1^−/− OT-I cells than in PD-1^+/+ OT-I cells at day 7 in the dLNs (Extended Data Fig. 3a–d). However, we did not detect major differences in BrdU uptake, Ki67 staining or live/dead blue staining between PD-1^+/+ and PD-1^−/− OT-I cells in the dLNs at 3, 5 and 7 days after infection (Extended Data Fig. 3b–d).

Next, we tested if tissue engraftment or persistence differences drove the higher number of PD-1^+/+ versus PD-1^−/− OT-I cells observed in the skin at days 10, 21 and 42. Skin PD-1^−/− OT-I T_RM cells showed 75% lower BrdU incorporation than PD-1^+/+ OT-I cells at days 5–10 after infection (91% in PD-1^+/+ OT-I cells versus 23% in PD-1^−/− OT-I cells; Fig. 2d), the window of T_RM cell engraftment and local expansion. To test epidermal occupancy, isolate PD-1 activity during T_RM cell specification and visualize where T_RM cell loss occurred in the absence of PD-1 signaling, we used inhibitory anti-PD-1 and ex vivo imaging. Anti-PD-1 inhibited signaling in cells expressing intermediate, but not high, levels of PD-1, and pharmacologic blockade is less complete than genetic loss²⁸. We transferred 200,000 tdTomato-labeled OT-I cells into CD11c–eYFP mice, in which enhanced yellow fluorescent protein (eYFP) marks dendritic cells (DCs), immunized them by VACV-OVAss 1 day later and injected them intraperitoneally with anti-PD-1 or isotype control at days 0–9 after infection. Cross-sectional skin imaging at days 10–14 after infection indicated loss of tdTomato⁺ OT-I cells near eYFP⁺ DCs in the epidermis and dermis of anti-PD-1-treated mice (Fig. 2e). Epidermal sheets, which allow a robust quantification of epidermal T_RM cells by imaging, showed a 32% drop in mean numbers of epidermal tdTomato⁺ OT-I cells in anti-PD-1-treated mice compared to isotype control-treated mice (Fig. 2f and Supplementary Fig. 2g), indicating that PD-1 promotes the formation of T_RM cells at days 0–9 after infection.

To test the effect of early inhibition of PD-1 on T_RM cell accumulation in skin, we immunized C57BL/6J mice transferred with OT-I cells with VACV-OVAss and administered anti-PD-1 or isotype control by intraperitoneal injection at days 0–9 after infection. Mice treated with anti-PD-1 showed a significant (81%) reduction in the number of epidermal T_RM cells and a 56–57% reduction in the mean number of dermal T_RM cells at week 7 after infection compared to isotype-treated controls (Fig. 2g, Extended Data Fig. 3e and Supplementary Fig. 2h,i). To test if early PD-1 inhibition affected late T_RM cell recall, we treated both groups with the S1P-modifying drug fingolimod (FTY720)²³, which leads to S1PR1 internalization and blocks T cell circulation, starting at week 5 after infection. One week after FTY720 initiation, the right and left flanks were shaved and treated with PBS (left) or SIINFEKL (right) to induce skin T_RM cell activation. Based on flow cytometry analysis, SIINFEKL treatment drove an increase in dermal, but not epidermal, T_RM cells in all groups 1 week after SIINFEKL recall (Fig. 2g). Fewer total dermal T_RM cells expanded 1 week post-SIINFEKL versus PBS challenge on the flank of mice that had received anti-PD-1 than isotype antibody during formation (Fig. 2g). However, the relative SIINFEKL-induced expansion of dermal OT-I T_RM cells was equivalent in isotype- and anti-PD-1-treated groups at day 49 (96% versus 90% expansion, respectively; Fig. 2g). Thus, PD-1 blockade during days 0–9 after infection drove a reduction in the number of T_RM cells at day 49 in PBS-challenged (81% reduction in epidermal T_RM cells and 56% reduction in dermal T_RM cells; Fig. 2g) and SIINFEKL-challenged (81% reduction in epidermal T_RM cells and 57% reduction in dermal T_RM cells; Fig. 2g) flank skin. However, the recall potential of skin OT-I T_RM cells remained intact (90% versus 96% expansion). Likely due to drainage from skin, KLRG1⁻CD62L⁺ OT-I T_CM cells from SIINFEKL dLNs showed a minor recall expansion (Extended Data Fig. 3e), suggesting that antigen recall is largely restricted to skin. Collectively, genetic and pharmacologic data indicate a role for PD-1 signaling in promoting the formation of peripheral skin CD8⁺ T_RM cells.

PD-1 promotes CD8⁺ T_RM cell skin-specific programming

Next, we adoptively transferred a mix of 100,000 PD-1^+/+ and PD-1^−/− OT-1 cells derived from littermates into C57BL/6J mice and used RNA sequencing (RNA-seq) to assess the transcriptomes of PD-1^+/+ and PD-1^−/− CD69⁺CD103^+/- OT-I cells sorted from the skin of recipient mice at day 14 after infection with VACV-OVAss (Fig. 3a, Extended Data Fig. 4a and Supplementary Fig. 3a,b). Differential gene expression analysis (false discovery rate (FDR) = 0.05, | log (fold change) | ≥ 1.5) revealed enrichment of extracellular matrix (ECM)-related transcripts (Col4a1, Col1a1 and Vwf) in PD-1^+/+ OT-I T cells (Fig. 3a). The skin T_RM cell ECM signature enrichment was also detected in other HSV- and VACV-generated skin T_RM cell transcriptomes (Extended Data Fig. 4b,c). PD-1^−/− T cells expressed higher levels of pyrogens (IL1b and Mefv (encoding pyrin)), apoptosis-related proteases (Casp3 and Casp4), actin and cytoskeletal binding proteins (Acta2, Tagln and Vcl), ECM degradation (Ecm1) and urokinase plasminogen activator (Plaur) transcripts than PD-1^+/+ (Fig. 3a and Supplementary Tables 1 and 2). PD-1^−/− OT-I cells showed higher expression of Tox, which is associated with terminal differentiation, interleukin-15 (IL-15)-resistant programming and exhaustion^29,30 (Fig. 3a), suggesting that PD-1 expression may enforce stemness^31,32.

Fig. 3: PD-1-dependent programming is skin specific and highly enforced during skin TRM cell formation. — **Fig. 3: PD-1-dependent programming is skin specific and highly enforced during skin T_RM cell formation.**

To test whether the PD-1-dependent programs were skin specific, we compared differentially expressed genes (DEGs) between skin PD-1^+/+ OT-I cells and PD-1^−/− CD69⁺CD103^+/− T_RM cells to DEGs from spleen PD-1^+/+ versus PD-1^−/− CD8⁺CD44^hiCD62L^hi T_CM cells obtained from published datasets (E-MTAB-1569)³³. The transcriptional signature of skin PD-1^+/+ OT-I cells did not associate with that of spleen PD-1^+/+ DEGs (Fig. 3b). By contrast, skin PD-1^−/− OT-I transcripts were enriched for DEGs also found in spleen PD-1^−/− T cells, with 53% shared genes (Fig. 3b and Extended Data Fig. 4d). These data suggest that PD-1-enforced transcripts are likely skin specific. To define how PD-1 influences early T_RM cell specification, we subtracted naive OT-I transcripts to yield Pdcd1-wild-type (Pdcd1-WT; day 14 PD-1^+/+ minus naive OT-I) or Pdcd1-knockout (Pdcd1-KO) programs (day 14 PD-1^−/− minus naive OT-I; Fig. 3c). We compared unique and shared transcripts in each program by Venn overlap (Fig. 3c). This analysis yielded a PD-1-dependent, 226-transcript program specific to PD-1^+/+ OT-1 cells, which we refer to hereafter as Pdcd1 WT_specific (Fig. 3c, Supplementary Fig. 3c and Supplementary Tables 3 and 4), 296 transcripts that were specific to PD-1^−/− OT-I cells, which we refer to hereafter as Pdcd1 KO_specific (Fig. 3c, Supplementary Fig. 3d and Supplementary Tables 5 and 6), and 291 transcripts that were expressed in both PD-1^+/+ and PD-1^−/− OT-I cells (hereafter ‘shared’; Fig. 3c). We tested whether these signatures were expressed during OT-I skin T_RM cell differentiation during days 0–90 after infection with VACV-OVA using a public dataset (GSE79805)³⁴. Expression of the Pdcd1 WT_specific program increased 153% between days 5 and 30 of skin T_RM cell development (Fig. 3d), suggesting that PD-1-enforced transcripts are highly expressed during skin T_RM cell specification. Expression of the Pdcd1 KO_specific program increased slightly (3%) between days 0 and 5, but not further (Fig. 3d), whereas expression of shared genes increased 25% between days 5 and 30 of T_RM cell development (Fig. 3d). Shared transcripts overlapped with core T_RM cell genes, including G-protein-coupled receptors (GPCRs) Cxcr3 and Cxcr6, and type 1 helper T cell-programming genes (Gzmb, Gzmk and Ifng; Fig. 3e and Supplementary Fig. 3e,f¹⁶). Skin PD-1^+/+ and PD-1^−/− T_RM cells expressed 53 of 121 transcripts, respectively, that were shared by T_RM cells isolated from multiple tissues (Fig. 3e and Supplementary Fig. 3e,f¹⁶). Shared transcripts included transcripts specific to T_RM cells (compared to transcripts from naive T cells and T_CM cells), such as immune checkpoint (Tigit, Ctla4, 2b4 (CD244), Icos (CD278)) and integrin (Itga1 (CD49a)) transcripts, and transcripts distinct from exhaustion, activation and circulating memory cells, such as Nr4a3, Fgl2, Crem, p21/Cdkn1A, Fosl2 and Cish (Fig. 3e). Collectively, these data suggest that PD-1 signaling enforces unique skin-specific programs and that skin PD-1^−/− OT-I T_RM cells retain a sizeable fraction of T_RM cell core program transcripts.

PD-1 modulates terminal CD8⁺ T_eff cell differentiation²⁶, preventing terminal differentiation after immediate antigen re-encounter³⁵. Therefore, we tested whether PD-1 promotes longevity for T_RM cells and exhausted T (T_ex) cells during antigen re-encounter. We detected increased protein expression of the transcription factors T-bet and eomesodermin in LN and spleen PD-1^−/− OT-I cells compared to PD-1^+/+ OT-I cells at day 10 after infection (Extended Data Fig. 4e), consistent with differentiation toward CD8⁺ T_eff cells. However, we did not detect differences in the expression of T-bet, eomesodermin, CD8⁺ T_eff cell-specific KLRG1 or memory-specific CD127 in skin PD-1^−/− versus PD-1^+/+ OT-I cells at day 10 after infection (Fig. 3f and Extended Data Fig. 4e).

To better define how PD-1 signaling impacts T_RM cell lineage commitment, we compared PD-1-dependent and PD-1-independent programs to state-specific signatures from early and late activated CD8⁺ T_eff cells, CD8⁺ T memory (T_M) cells, CD8⁺ T_ex cells and CD8⁺ T_RM cells^16,25,36,37. Shared, PD-1-independent programs most highly overlapped with T_RM cell signatures (Fig. 3g), suggesting that loss of PD-1 did not abrogate T_RM cell specification, whereas Pdcd1 KO_specific programs included transcripts from other T cell differentiation states, which were not present in the Pdcd1 WT_specific programs from PD-1-sufficient skin T_RM cells (Fisher exact test P ≤ 0.05; Fig. 3g and Extended Data Fig. 4f,g). These data suggest that PD-1 signaling both reinforces the skin T_RM cell transcriptional program and represses transcripts associated with other CD8⁺ T cell activation or differentiation states.

PD-1-deficient T cells have impaired TGFβ responsivity

To identify upstream programmatic drivers of PD-1-dependent and PD-1-independent signatures, we performed pathway analysis on Pdcd1 WT_specific, shared and Pdcd1 KO_specific signatures (Fig. 4a). The PD-1-independent shared program included cytokine, chemokine and GPCR signaling (Fig. 4a), consistent with the importance of pertussis-sensitive GPCRs in early T_RM cell formation³⁸. Some Pdcd1 KO_specific pathways (for example, TCR, CD28 co-stimulation and fatty acid metabolism; Fig. 4a) were consistent with known PD-1-dependent pathways¹³. Pdcd1 KO_specific programs also included loss of SMAD2/SMAD3 signaling (Fig. 4a), suggesting deficient upstream TGFβ signaling. In parallel, Pdcd1 WT_specific PD-1-enforced pathways included integrins and ECM (Fig. 4a), which typify a TGFβ signature because ECM sequesters TGFβ³⁹ through small ECM-abundant proteoglycans. Pdcd1 WT_specific pathways included early cell cycle pathways (mitotic G1–G1/S phases; Fig. 4a).

Fig. 4: Pdcd1-KO-specific loss of TGFβ pathway programs. — **Fig. 4: *Pdcd1-*KO-specific loss of TGFβ pathway programs.**

TGFβ is a T_RM cell differentiation cue^25,27,40,41 that preconditions naive T cells to form T_RM cells⁴², increase CD103 expression^25,40 and remain in the epidermis⁴³. To examine if PD-1 promotes TGFβ signaling, we scored Pdcd1 WT_specific, shared and Pdcd1 KO_specific against those from HSV-specific CD8⁺ gBT splenocytes activated in vitro with IL-2, TGFβ or both (GSE125471 (ref. ⁴⁴)). CD8⁺ T cells activated with IL-2 alone induced upregulation of the expression of transcripts from the Pdcd1 KO_specific program (Pdcd1 KO_{specific up}; Fig. 4b), whereas those activated with TGFβ alone dampened the expression of Pdcd1 KO_{specific up} transcripts (Fig. 4c), suggesting that Pdcd1 KO_specific programs might reflect T cell activation and IL-2-induced transcripts opposed by TGFβ. HSV-specific gBT CD8⁺ T cells activated with TGFβ and IL-2 induced the PD-1-independent, upregulated transcripts from the shared program (shared_up) individually or synergistically (Extended Data Fig. 5a). PD-1-dependent upregulated transcripts from the WT_specific program (WT_{specific up}) transcripts were induced only by TGFβ + IL-2 (Extended Data Fig. 5a), suggesting that PD-1 supported their synergy.

Next, we performed comparative genomic analysis between skin PD-1^+/+ and PD-1^−/− CD69⁺CD103^+/− OT-I T_RM cells isolated at day 14 after infection with VACV-OVAss and a publicly available dataset (GSE178769)⁴⁵ of skin Tgfbr2^+/+ or Tgfbr2^−/− T_RM cells, which expressed or lacked expression of the TGFβ receptor 2 (TGFβR2), isolated at day 14 after infection with HSV modified to express OVA (HSV-OVA; Fig. 4d and Supplementary Table 7). Genes enriched in skin PD-1^−/− OT-I T_RM cells compared to skin PD-1^+/+ OT-I T_RM cells, which were highly enriched for transcripts from Tgfbr2^−/− T_RM cells (Fig. 4d,e). A high proportion (96 of 297, 32%) of Tgfbr2^−/− leading-edge genes drove enrichment of PD-1^−/− programs (Fig. 4d,e). These data suggest that PD-1^−/− OT-I cells lack sensitivity to TGFβ and express transcripts detected in T cells activated in vitro without TGFβ or in T_RM cells induced in vivo without TGFβR2.

To validate these transcriptome observations, we assessed the protein expression of PD-1-independent (CXCR3, CXCR6) or dependent (PSMAD2/PSMAD3 pathway) transcripts in OT-I cells isolated at days 14–15 after infection from the spleens and LNs of mice infected with VACVss and treated with anti-PD-1 or isotype control at days 14–15 after infection. CXCR3 and CXCR6 protein expression in dLN or spleen OT-I cells was equivalent in anti-PD-1-treated or isotype control-treated mice (Fig. 4f,g), although expression of pSMAD2 in LN and spleen OT-I cells was lower in anti-PD-1-treated mice than in isotype-treated mice (Fig. 4f,g). These data suggest lower TGFβ signaling in anti-PD-1-treated T cells and that PD-1 sensitizes T cells to TGFβ utilization.

TGFβ signaling compensates for anti-PD-1-induced loss of T_RM cells

To test whether TGFβ signaling compensates for anti-PD-1-driven T_RM cell loss, we assessed T_RM cell formation in anti-PD-1- or isotype-treated mice with or without TGFβ addition. Anti-PD-1 treatment between days 0 and 9 after infection with VACVss drove a 32% reduction in the number of skin CD8⁺ T_RM cells compared to isotype control treatment (Fig. 5a), which impacted both epidermal and dermal CD8⁺ T_RM cells, without loss of OT-I cells in the LN or spleen (Extended Data Fig. 6a and Supplementary Fig. 4a). Anti-PD-1 treatment between days 6 and 12 after infection did not reduce skin CD8⁺ T_RM cell numbers compared to isotype control-treated mice (Extended Data Fig. 6b), indicating that T_RM cell dependency on PD-1 signaling likely occurs between days 0 and 6 after infection. Skin OT-I cells in anti-PD-1-treated, but not isotype-treated, mice exhibited lower PD-1 labeling (Extended Data Fig. 6c), suggesting that anti-PD-1 directly targets skin CD8⁺ T cells. There was no difference in the expression of TGFβR1 and TGFβR2 on skin OT-I T_RM cells between anti-PD-1- and isotype-treated mice (Extended Data Fig. 6c), indicating that TGFβR expression did not account for the differences in T_RM cell formation in anti-PD-1-treated mice.

Fig. 5: Constitutive TGFβR signaling rescues the anti-PD-1-dependent inhibition of TRM cell formation and engraftment in a cell-autonomous manner. — **Fig. 5: Constitutive TGFβR signaling rescues the anti-PD-1-dependent inhibition of T_RM cell formation and engraftment in a cell-autonomous manner.**

In the experimental setting in which we transferred OT-I cells that expressed a tamoxifen-inducible constitutively active form of TGFβR1 (ref. ⁴³) into C57BL/6J mice, immunized with VACV-OVAss 1 day later and administered tamoxifen for the first 5 days after infection and anti-PD-1 or isotype control between days 0 and 9 after infection, we no longer observed a specification defect in dermal and epidermal T_RM cells in the anti-PD-1-treated mice at days 10–12 (Fig. 5b and Supplementary Fig. 4b). These data suggest that constitutive, T cell-intrinsic TGFβ receptor signaling largely rescued or compensated the anti-PD-1 effect on early T_RM cell formation. No differences were seen in the number of LN or spleen OT-I cells, despite evidence of anti-PD-1 targeting to these populations by a decrease in cell-surface staining (Extended Data Fig. 6d,e). There was a significant difference between the anti-PD-1 effect on T_RM cell numbers without (Fig. 5a) or with (Fig. 5b) enforced TGFβR1 activity (P < 0.05).

TGFβ is required for skin T_RM cell engraftment^25,40, and absence of TGFβ drives the loss of epidermal T_RM cell retention^34,40,41,43. To assess whether early PD-1 effects on T_RM cell formation were due to differences in TGFβ utilization that could shape T_RM cell engraftment or retention and the specific effect of PD-1 on T_RM cell engraftment, we used a ‘prime and pull’ assay, in which T cells are primed in vitro and then ‘pulled’ into inflamed skin in vivo to generate T_RM cells (based on the observation that although local cognate antigen re-encounter by T cells facilitates optimal CD8⁺ T_RM cell development, nonspecific tissue inflammation is sufficient for T_RM cell engraftment^46,47). OT-I cells were activated in vitro with cognate SIINFEKL peptide and IL-2 for 5 days and transferred intravenously into C57BL/6J mice treated topically on the flank on day 0 after transfer with the chemical hapten 2,4-dinitrofluorobenzene (DNFB⁴⁶; Fig. 5c and Supplementary Fig. 5a,b). Mice were administered either anti-PD-1 or isotype control by intraperitoneal injection every 3 days between days 0 and 9 after transfer. Anti-PD-1-treated mice showed a 71% reduction in the accumulation of epidermal and dermal OT-I cell engraftment in inflamed skin relative to isotype control-treated mice at day 10 after transfer (Fig. 5c and Supplementary Fig. 5c), suggesting heightened anti-PD-1 effects during the engraftment period. Live/dead blue uptake and LN KLRG1⁻CD62L⁺ T_CM cell numbers were similar between anti-PD-1- and isotype control-treated mice (Extended Data Fig. 7a,b). A minor reduction in spleen KLRG1⁻CD62L⁺ T_CM cells and KLRG1⁻CD62L⁻ T_EM cells was observed in the anti-PD-1-treated mice (Extended Data Fig. 7a,b). Systemic administration of recombinant TGFβ1 between days 0 and 9 after transfer almost entirely rescued the reduction in epidermal and dermal T_RM cell engraftment but not spleen T_CM and T_EM cell numbers (Fig. 5c, Extended Data Fig. 7b and Supplementary Fig. 5c), suggesting a skin-specific effect for TGFβ. When OT-I cells that expressed a tamoxifen-inducible constitutively active TGFβR1 (refs. ^27,43) were transferred into C57BL/6J recipients, followed by tamoxifen treatment between days 0 and 5 after transfer to induce the expression of the constitutively active TGFβR1, the number of T_RM cells in anti-PD-1-treated mice at day 10 was similar to isotype control-treated mice (Fig. 5d, Extended Data Fig. 7c and Supplementary Fig. 5d). These data indicate that anti-PD-1 disrupted a signal critical to T_RM cell engraftment, whereas TGFβ1 or T cell-intrinsic constitutive TGFβR1 activity restored this signal. Collectively, these data reveal an unappreciated role for PD-1 in promoting early T_RM cell competitive niche formation by promoting engraftment and response to TGFβ (Extended Data Fig. 7d).

Discussion

Here, we showed that PD-1 signaling governs early T_RM cell specification. PD-1 promotes early CD8⁺ T cell memory by incorporating local cues, like TGFβ, favoring T_RM cell niche engraftment and limiting cell promiscuity toward other fates.

Given the global impact of PD-1 signaling on all memory T cells, these events may take place at the level of a common memory progenitor. However, PD-1-dependent sensitization to TGFβ may render T_RM cells particularly susceptible to failed retention in skin because T_RM cells use TGFβ to engraft in epithelial tissues and form a niche⁴⁵. PD-1-enforced ECM and GPCR pathways are known to support T_RM cell competitive fitness (for example, engraftment, adhesion and residency) when competing for tissue niche occupancy⁴⁸. TGFβ utilization impacts T_RM cell formation in a site-specific manner⁴⁵. PD-1-enforced utilization of TGFβ may promote T_RM cell formation in TGFβ-rich epithelial sites such as the skin and lungs, where anti-PD-1 treatment-induced irAEs occur frequently and where we observed human CD8⁺ T_RM cells expressing PD-1. Therefore, PD-1 signaling may favor site-specific T_RM cell formation by using TGFβ and may help control metastatic spread but could also make T_RM cells vulnerable to anti-PD-1 toxicity at these TGFβ-rich sites. Therapeutic approaches to optimize TGFβ activation and transient PD-1 agonism may improve T_RM cell specification in tissues and tumors alike. Maintaining PD-1 expression after T_RM cell formation may prevent T_RM cell-driven autoreactivity in tissues by regulating the threshold for local recall across different barrier tissues. These effects remain to be tested. The observation that PD-1 signaling might oppose IL-2 signaling also requires further investigation.

To protect the barrier tissues, high-affinity T_RM cells must remain permissive to rapid, sensitive antigen-driven recall. Although T_RM cells cooperate with circulating memory T cells, T_RM cells are sufficient to independently mediate protective recall^10,49,50. T_RM cell recall-triggered pathogen alarm leads to interferon-γ-dependent recruitment of circulating memory T cells and innate natural killer (NK) cell amplification of immune responses^51,52. One limitation of our study is that we did not directly test the clearance of recall VACV after anti-PD-1-impaired T_RM cell formation. However, the number of T_RM cells has been shown extensively to directly correlate with their recall capacity (for example, viral clearance after rechallenge^{5,46,51,53,54} or tumor protection⁵⁵). Inhibitory anti-PD-1 reduced the number of T_RM cells in the skin and lowered the absolute number of T_RM cells before peptide stimulation, resulting in lower numbers of expanded T_RM cells after recall. Future work will interrogate how PD-1 blockade during T_RM cell formation impacts viral clearance or autoinflammatory disease induction. Also, the cellular partners that express PD-L1 or PD-L2 during the various stages of the T_RM cell life cycle remain to be determined.

Ctla4^−/− mice show lymphoproliferation accompanied by tissue immune cell infiltration and death due to organ failure at 2–3 weeks of age^56,57. Autoinflammation in PD-1^−/− mice is less severe, often requiring a trigger⁵⁷. Less severe abnormalities in the differentiation trajectory of T_eff cells were observed in PD-1^−/− cells than in Ctla4^−/− T cells⁵⁸. PD-1 signaling may instead balance peripheral and central memory. Notably, early T_RM cell selection by PD-1 signaling contrasts the inhibition of CD8⁺ T_CM cells by PD-1, which may shape the early distribution of CD8⁺ T_RM and T_M cells across tissue compartments and favor T_RM cell formation and seeding when local site-specific antigen is encountered. Future work is needed to distinguish PD-1 effects on T_RM cell formation from PD-1 effects during T_RM cell maintenance and recall. This distinction may provide mechanistic insights into T cell repertoire selection, memory formation, protection, antitumoral immunity and recall and pathways regulating autoinflammation and irAE in tissues.

Methods

Mice, cell lines, virus preparation and mass cytometry

All mice were bred and/or housed in a specific pathogen-free facility at Harvard University and Weill Cornell Medicine. The Harvard University Animal Care and Use Committee and Weill Cornell Medicine Institutional Animal Care and Use Committee approved all protocols. C57BL/6 mice and CD45.1 mice were purchased from Jackson or Taconic Labs, whereas Cd274/Pdcd1lg2^−/− (PD-L1/PD-L2^−/−) mice and PD-1^−/− OT-I mice were generously provided by the laboratory of A. Sharpe (Harvard University). Mice used to generate the TGFβ loss dataset (C57BL/6, OT-I.CD45.1 and OT-I.Tgfbr2^fl/fl.dLck-cre.CD45.1 (OT-I Tgfbr2^−/−)) were bred in the University of Melbourne, Department of Microbiology and Immunology. E8i-CreERT2 Tgfbrca^fl/fl Thy1.1/Thy1.2 OT-I mice were bred at the University of Pittsburgh. Freshly isolated LNs and spleens were provided by the laboratory of D. H. Kaplan on behalf of D. Vignali (University of Pittsburgh School of Medicine) and L. Bartholin (INSERM, Lyon), who generated this strain^59,60,61 in some experiments or were generated locally after receipt of this line. Female mice were used at 6–12 weeks of age. All animal experiments were approved by The University of Melbourne Animal Ethics Committee. Sex- and age-matched mice between 6 and 12 weeks of age were used. B16-OVA cells were received from C. Drake (Columbia University) and A. Sharpe (Harvard University) and were grown in DMEM (Gibco) with 10% fetal calf serum (Gibco) and penicillin/streptomycin (Gibco) for no more than two passages before intradermal implantation. VACV-OVA was generously provided by J. Yewdell (National Institutes of Health) and was expanded and titered by standard protocols⁶². Mass cytometry of human skin and blood from donors was performed as previously described²².

Flow cytometric staining and antibodies

All flow cytometric staining was performed on ice as previously described⁶³. Cells were stained using a Live/Dead dye in PBS, followed by extracellular staining in PBS containing 2% fetal calf serum (FACS Buffer) at 4 °C and subsequent fixation/permeabilization before staining for intracellular proteins. The following mouse antibodies were obtained from BD, eBioscience or Biolegend: CD3 (eBio500A2, 17A2), CD4 (RM4-5), CD8 (53-6.7), CD45 (30-F11), CD45.1 (A20), CD45.2 (104), Thy1.1 (HIS51), Thy1.2 (53-2.1, 30-H12), Va2 (B20.1), Vb5 (MR9-4), CD69 (H1.2F3), CD103 (2E7), KLRG1 (2F1), CD44 (IM7), CD62L (MEL-14), CD127 (A7R34), PD-1 (29F.1A12, RMP1-30), PD-L1 (10F.9G2), PD-L2 (TY25), IA/IE (M5/114.15.2), CD11c (N418), langerin (4C7), CD11b (M1/70), T-bet (4B10), Eomes (Dan11mag), TGFβR1 and TGFβR2 (R&D Systems) and Ki67 (16A8). BrdU staining was performed using an APC BrdU flow kit according to manufacturer’s instructions (BD Pharmingen). For pSMAD2 staining, cells were first intranuclearly stained with anti-pSMAD2 (Cell Signaling Technology, E8F3R) and then stained with anti-rabbit IgG (Biolegend, Poly4064).

The following human antibodies were obtained from BD, eBioscience or BioLegend: CD3 (UCHT1, SK7, OKT3), CD8 (SK1), CD45 (HI30), CD69 (FN50), CD103 (B-Ly7), HLA-DR (L243), PD-1 (EH12.2H7), TIM3 (F38-2E2), BDCA1 (L161), BDCA3 (AD5), CD20 (2H7), CD66b (G10F5), CD56 (HCD56), CD11c (Bly6), CD14 (RM052), PD-L1 (29E.2A3), TIM3 (F38-2E2), CD45RO (UCHL1), GRZB (GB11, intracellular staining) and CD127 (A019D5).

HSV-generated T_RM cell single-cell suspensions were stained with fluorescently conjugated anti-CD45.1 (A20), anti-CD45.2 (104) or anti-Vα2 (B20.1) from BD Bioscience and anti-CD3ε (145-2C11) from Biolegend at 4 °C for 60 min in PBS containing 0.5% fetal calf serum and 0.05 M EDTA.

Skin scarification

In experiments using VACV-OVA, age- and sex-matched control and experimental mice were taken, and skin scarification was performed. Briefly, for all priming and maintenance experiments (unless specifically stated), 1 × 10⁶ viral p.f.u. was applied to each ear and 2 × 10⁶ viral p.f.u. was applied to the tail before scarification with a 25-gauge needle. In tumor protection experiments, 2 × 10⁶ viral p.f.u. was applied to the tail before scarification.

Mouse tissue digestion

Isolation of single-cell suspensions from mouse skin was previously described⁶⁴. Briefly, ear skin was cut into small pieces and incubated in RPMI 1640 medium/HBSS (Invitrogen) containing 5% fetal bovine serum (Gibco), 10 mM HEPES (Gibco), 2.5 mg ml⁻¹ collagenase IV/collagenase D (Roche) and 1 mg ml⁻¹ DNase I (Roche) for 90 min at 37 °C. Digested skin was passed through a 70-μm cell strainer (BD), and the remaining skin pieces were further mechanically disrupted using a 3-ml syringe plunger accompanied by continuous washing with RPMI/HBSS^−/− to generate a single-cell suspension. Mouse flank skin was isolated by initially shaving with electric clippers and cutting an ~2 × 2 cm portion of the back skin, whereas the tail skin was removed by peeling the skin off the tail bone using forceps in one continuous motion. Flank or tail skin was cut into small pieces and incubated in 3 ml of digestion cocktail containing 2.5 mg ml⁻¹ collagenase XI (Sigma), 0.5 mg ml⁻¹ hyaluronidase (Sigma) and 0.1 mg ml⁻¹ DNase I in C10 medium (RPMI 1640 supplemented with 10% fetal bovine serum (Gibco), 1 mM sodium pyruvate, 1% HEPES and 1× nonessential amino acids). Skin pieces were digested for 90 min at 37 °C in a shaker at 255 rpm. Twenty milliliters of supplemented RPM1 1640 (or C10 medium) was further added to the digested skin solution and vortexed for 15 s. Digested skin was then passed through a 70-μm cell strainer (BD), and the remaining flank or tail skin pieces were further mechanically disrupted using a 3-ml syringe plunger along with continuous washing with RPMI/HBSS^−/− to generate single-cell suspensions.

Adoptive transfer and T_RM cell formation

For adoptive transfer experiments comparing PD-1^+/+ and PD-1^−/− T cells, naive congenically marked Thy1.1⁺/Thy1.2⁺ PD-1^+/+ or PD-1^−/− CD8⁺ OT-I T cells were isolated using a magnetic bead negative selection kit following the manufacturer’s protocol (Miltenyi). To test PD-1 expression in T_RM versus T_CM and T_EM cells, OT-I Rag1^−/− T cells were used for adoptive transfer from mice bred in-house. To ensure that the expression of PD-1 was not related to the use of Rag1^−/− mice, Thy1.2⁺ OT-I T cells isolated from C57BL/6J mice were also used for adoptive transfer to Thy1.1⁺ recipients, followed by VACV-OVA skin scarification the next day. PD-1 expression was checked at different time points in T_RM versus T_CM and T_EM cells. To test the role of WT or PD-1^−/− T cells in T_RM versus T_CM and T_EM cell formation, PD-1^+/+ OT-I Thy1.2⁺ T cells were isolated from mice purchased from Jackson Laboratories, whereas PD-1^−/− OT-I T cells with Thy1.1 congenic markers were isolated from mice bred in-house. For all adoptive transfer experiments with a single donor, 2 × 10⁵ OT-I T cells were transferred by intravenous injection on day −1. Recipient mice were vaccinated by VACV-OVAss on each ear with 1 × 10⁶ viral p.f.u. and on the base of the tail with 2 × 10⁶ viral p.f.u. on day 0. For all mixing experiments, 1 × 10⁵ PD-1^+/+ C57BL/6J OT-I T cells were mixed with 1 × 10⁵ PD-1^−/− OT-I or 1 × 10⁵ PD-1^+/+ OT-I T cells from littermates (to control for genetic mismatch) and were coadoptively transferred to C57BL/6J recipients, followed by VACV-OVAss on day 0. At day 10, 3 weeks, and 6 weeks postinfection, T cells were isolated from the skin, LNs and spleen, and quantitation was performed. T_RM cells were classified based on the expression of CD69⁺ and CD103^+/−, whereas T_CM and T_EM cells were identified in LNs and spleen based on the expression of CD44⁺/CD127⁺ and CD62L⁺ or CD62L⁻, respectively. In some mixed transfer experiments, controls receiving PD-1^+/+ C57BL/6J OT-I cells were mixed 1:1 with PD-1^+/+ OT-I⁺ littermates to PD-1^−/− siblings, and T_CM and T_EM cells were also identified by KLRG1⁻CD62L⁺ and KLRG1⁻CD62L⁻ gating, respectively. BrdU studies were performed by injecting BrdU intraperitoneally into experimental mice between days 5 and 10 after VACV-OVA infection as a measure of proliferation.

For experiments in which anti-PD-1 was administered during early T_RM cell formation, 2 × 10⁵ OT-I T cells from either PD-1^+/+ OT-I or PD-1^+/+ OT-I E8i-CreERT2 Tgfbrca^fl/+ Thy1.1⁺/Thy1.2⁺ mice were adoptively transferred to recipient mice on day −1. VACV-OVAss was administered on the ear and tail on day 0, and anti-PD-1 was administered intraperitoneally on days 0, 3, 6 and 9. Transferred OT-I cells were quantitated between days 10 and 12. Mice receiving E8i-CreERT2 Tgfbrca^fl/+ OT-I cells were treated with tamoxifen on days 0, 1, 2, 3 and 4 for tamoxifen-induced Cre expression of Tgfbrca^fl/+(constitutively active TGFβR1). T_RM cells were classified based on the expression of CD69⁺ and CD103^+/−. Total CD8⁺ T cells were gated for LNs and spleen.

Tamoxifen treatment

Tamoxifen (T5648, Sigma-Aldrich) was dissolved in corn oil (Sigma-Aldrich) at a final concentration of 10 mg ml⁻¹ by shaking overnight at 37 °C and then stored at −20 °C. Mice were administered tamoxifen for five to six consecutive days starting on day 0 by intraperitoneal injection at 0.05 mg per g (body weight).

Vaccinia viral load determination by quantitative PCR

To evaluate vaccinia viral load, the tails of mice that received PD-1^+/+ or PD-1^−/− OT-I⁺ T cells on day −1 and were immunized on the ear and tail at day 0 with 4 × 10⁶ p.f.u. VACV-OVA were isolated at the time points indicated. Briefly, DNA was purified from the tail using a DNeasy Mini kit (Qiagen) according to the manufacturer’s instructions. Taqman probe and primers used in the assay are specific for the ribonucleotide reductase Vvl4L of vaccinia virus. Forward and reverse primers with the sequences 5′-GACACTCTGGCAGCCGAAAT-3′ and 5′-CTGGCGGCTAGAATGGCATA-3′ were synthesized by Invitrogen, while a Taqman probe with sequence 5′-AGCAGCCACTTGTACTACACAACATCCGGA-3′, 5′-labeled FAM and 3′-labeled TAMRA was synthesized by Applied Biosystems. PCR amplification was performed in a 20-μl volume containing 2× TaqMan Master Mix, 500 nM forward and reverse primer, 750 nM probe and template DNA using a StepOnePlus Real-Time PCR System (Applied Biosystems). Thermal cycling conditions were 50 °C for 2 min (UNG activation) and 95 °C for 10 min (AmpliTaq Gold Activation), followed by 45 cycles of PCR amplification at 95 °C for 15 s and 60 °C for 1 min. Viral load was calculated from a standard curve plotted using DNA from known viral stock p.f.u. PCR-based limits of viral DNA detection were identified as 10¹ copies per ng of viral DNA using a standard curve generated from viral dilution (p.f.u. per μl) on CV-1 cells.

Immunofluorescence and image processing

Skin tissues were obtained from recipient reporter mice expressing CD11c–eYFP⁶⁵ that received adoptive transfer of tdTomato⁺ OT-I cells 1 day before VACV-OVAss. These mice were also treated with isotype control or anti-PD-1 by intraperitoneal injection on the day of VACV-OVAss (day 0) and 3, 6 and 9 days after infection. Ear skin was isolated at day 11 and fixed in fresh 4% paraformaldehyde in PBS for 1 h at 4 °C. Tissues were then washed with PBS three times and incubated in 30% sucrose overnight at 4 °C, followed by subsequent washing with PBS the next day. Embedding of the tissue was performed in OCT (Tissue Tek) and frozen at −80 °C for long-term storage. OCT blocks were cryosectioned (10 μm), and sections were additionally stained with DAPI (Sigma) for nuclei visualization before using mounting medium to mount the slides. Images were acquired with a Zeiss Axioplan2 using a Plan-Apochromat ×20/0.8-NA air objective, and data were analyzed using Fiji (ImageJ) software.

Epidermal sheet and immunostaining

The dorsal and ventral halves of the ear were physically separated. Tissues were incubated in dispase (5 U ml⁻¹, STEMCELL Technologies) for 60 min at 37 °C, allowing the epidermis to be physically separated from the dermis. After separation, the epidermal sheets were washed twice in PBS, fixed for 60 min in 4% paraformaldehyde at room temperature and washed twice in PBS before mounting. Epidermal sheets were mounted in Fluoromount Mounting Medium (Southern Biotech). Images were acquired with a Leica Stellaris confocal microscope using a Plan-Apochromat ×20/0.75-NA CS2 air objective. The number of tdTomato⁺ cells was quantified using ImageJ software and was normalized per mm².

In vitro activation and adoptive transfer of OT-I T cells (prime and pull)

Total splenocytes and cells from LNs isolated from OT-I mice were cultured for 3.5 days together with SIINFEKL (OVA 257-264, InvivoGen) for in vitro activation. First, 20 × 10⁶ cells combined from the spleen and LNs were cultured in 20 ml of Hyclone supplemented RPMI medium with 1 μg ml⁻¹ SIINFEKL peptide in culture flasks on day 0. Cells were split 1:2 with the addition of medium and IL-2 (5 ng ml⁻¹; Peprotech) on days 2 and 3 to maintain T cell proliferation. At day 3.5, primed and activated CD8⁺ OT-I T cells were then collected from the cultures, and 10 × 10⁶ cells were transferred intravenously into recipient mice. On the same day of transfer, 0.5% DNFB was applied topically on shaved flank skin to pull or engraft the activated OT-I cells at the site of inflammation. Anti-PD-1 was administered intraperitoneally on day 0 of inflammation and on days 3, 6 and 9 after DNFB application. In some experiments, additional TGFβ1 cytokine was administered during engraftment (days 1–9) to recipient mice. TGFβ1 (recombinant mouse TGFβ1 (carrier-free); Biolegend) was dissolved in endotoxin-free water at a final concentration of 5 μg ml⁻¹ and stored at −80 °C. Mice were injected intraperitoneally with 0.5 μg of TGFβ1 cytokine for 9 consecutive days starting on day 1 of the engraftment period. For prime and pull with TGFβRCA donor OT-I cells, tamoxifen was administered for five to six consecutive days starting on day 0 of the engraftment period. Transferred OT-I cells quantitated on day 10 were classified as T_RM cells based on the expression of CD69⁺ and CD103^+/−, whereas T_CM and T_EM cells were identified based on the markers CD44⁺ and CD62L⁺ (T_CM) or CD62L⁻ (T_EM), respectively.

HSV infection and cell sorting

HSV infection was performed by scarification using 1 × 10⁶ p.f.u. of the KOS strain modified to express OVA protein (HSV-OVA)²⁵. Transgenic OT-I T cells were activated in culture for 4–5 days with OVA_257–264 (SIINFEKL) peptide-pulsed splenocytes, respectively, in the presence of recombinant human IL-2 (25 U ml⁻¹; Peprotech) at 37 °C with 5% CO₂. Cells (5 × 10⁶) were injected intravenously into mice infected with HSV-OVA 2 days prior. On day 14 after infection, epidermal skin T cells were obtained by removing shaved and depilated skin (1–3 cm²) and incubating in dispase solution (2.5 mg ml⁻¹; Roche) for 90 min at 37 °C. The epidermal layer was manually separated, placed in collagenase III solution (3 mg ml⁻¹; Worthington) containing DNase I (2.5 µg ml⁻¹), chopped into fine pieces and incubated for an additional 30 min. Cells were purified by sorting based on the expression of CD3⁺, Vα2⁺ and CD45.1⁺ using a BD FACSAria.

VACV-OVA infection and cell sorting

In total, 1 × 10⁵ PD-1^+/+ CD8⁺ OT-I T cells were mixed with 1 × 10⁵ PD-1^−/− CD8⁺ OT-I T cells (isolated using negative MACS bead selection from littermates) and were coadoptively transferred to C57BL/6J recipients 1 day before infection. Recipient mice were vaccinated by VACV-OVAss on each ear with 1 × 10⁶ viral p.f.u. and on the base of the tail with 2 × 10⁶ viral p.f.u. At day 14, PD-1^+/+ or PD-1^−/− donor CD8⁺ T cells were isolated from ear skin using the skin digestion protocol, directly sorted into TRIzol and stored and frozen at −80 °C. Naive CD8⁺ OT-I T cells were isolated from the LNs and spleen of naive mice by negative selection on MACS beads and stored in TRIzol before RNA extraction. All PD-1^+/+ and PD-1^−/− donor OT-I cells used for sorting and sequencing were isolated from mice that were littermates.

RNA-seq alignment and gene expression analysis

RNA from PD-1^+/+ or PD-1^−/− OT-I T cells from day 14 skin was extracted from 5,000 cells sorted by FACS using an RNeasy mini kit (Qiagen) as per the instructions provided by the manufacturer. Day 0 naive OT-I T cells were purified using negative MACS bead selection, and RNA was isolated from 5,000 cells using the same RNeasy mini kit protocol. cDNA libraries were constructed using a SMART-Seq v4 (Clontech) ultralow input RNA kit for sequencing. Samples were barcoded and sequenced on an Illumina HiSeq 4000 as paired-end runs with a length of 50 bp. For day 14 skin PD-1^+/+ or PD-1^−/− OT-I T cells versus day 0 naive OT-I T cells, raw sequenced reads were aligned to the mouse reference genome (version mm10 from University of California, Santa Cruz) using STAR (version 2.4.2) aligner. Aligned reads were quantified against the reference annotation (mm10 from Gencode) to obtain fragments per kilobase per million (FPKM) and raw counts using CuffLinks (version 2.2.1) and HTSeq, respectively. For PD-1^+/+ or PD-1^−/− OT-I T cell skin versus naive OT-I T cell samples, genes with a minimum expression of counts per million > 3 in at least two or more samples in a given comparison were included for further analysis. Normalized raw counts were voom transformed, obtaining expression profiles (in log₂ scale) that can be analyzed using linear models. Because biological replicate-specific pairing was present wherein both PD-1^+/+ and PD-1^−/− T cells were isolated from the same mice, differential expression analysis was performed using the R package limma, including both genotype and the replicate as factors. P values of moderated paired t-tests were adjusted for multiple hypothesis testing using the Benjamini–Hochberg approach, which controls the FDR. Genes with an absolute log₂ (fold change)of ≥1.5 and FDR of ≤0.05 were considered significantly differentially expressed.

For Tgfbr2^+/+ versus Tgfbr2^−/− day 14 OT-I T cells, cDNA libraries were prepared from 500 cells sorted by FACS using the SMART-Seq v2 protocol⁶⁶ with the following modifications: (1) 1 mg ml⁻¹ bovine serum albumin lysis buffer (Ambion Thermo Fisher Scientific), (2) the addition of 20 µM TSO and (3) the use of 250 pg of cDNA with one-fifth reaction of an Illumina Nextera XT kit (Illumina). The length distribution of the cDNA libraries was monitored using a DNA High Sensitivity Reagent kit on a PerkinElmer Labchip (PerkinElmer). Two biological replicates were generated with two technical replicates each, and samples were subjected to an indexed paired-end sequencing run of 2 × 51 cycles on an Illumina HiSeq 2000 system (Illumina; 16 samples per lane).

Sequence reads were mapped to the Mus musculus genome (GRCm38/mm10) using the align function from Rsubread⁶⁷, and annotated genes for GRCm38 were counted using featureCounts⁶⁸. Counts from technical replicates were summed by gene to produce two combined biological replicate samples for each group. Combined samples were then upper-quartile normalized, followed by application of RUVs⁶⁹ using a single factor of unwanted variation (k = 1) and mouse housekeeping genes⁷⁰ as ‘negative controls’. The edgeR package⁷¹ was used to calculate upper-quartile normalization factors and to fit gene-wise negative binomial generalized linear models for the experimental design using the result from RUVs as an additional model covariate, where ‘trended’ negative binomial dispersions and a prior count of 2 were used. Likelihood ratio tests were used to test for differential expression. Data are deposited in the Gene Expression Omnibus (GEO) under accession code GSE178769 (ref. ⁴⁵).

Hierarchical clustering of RNA expression and principal component analysis

For the PD-1^+/+ or PD-1^−/− OT-I T cells at day 14 after VACV-OVAss skin versus day 0 naive OT-I T cells comparison, gene expression profiles were represented in a two-dimensional space using principal component analysis and clustered based on expression profiles via unsupervised hierarchical clustering (with Pearson’s correlation as the distance metric and the ward.D agglomeration method). Both methods were performed on the log₂-transformed FPKM expression values in R statistical software.

GSEA of RNA-seq expression data

Experiments comparing either PD-1^+/+ or PD-1^−/− OT-I T cells from day 14 skin to day 0 naive OT-I T cells applied a supervised gene set over-representation-based pathway analysis executed with the online webtool ConcensusPathDB⁷² (cpdb), where pathways with an FDR of ≤0.01 were considered significantly enriched. Raw data are available in GEO under accession number GSE130207.

Similarly, in Fig. 3g and Extended Data Fig. 4f,g, over-representation analysis of the nonexclusive or exclusive T cell differentiation state signatures was performed using a Fisher’s exact test. GSVA was used to test enrichment of PD-1-dependent signatures on the following two publicly available datasets: GSE125471 (downloaded and processed using methods described above to obtain FPKM data) and GSE79805 (downloaded using the package GEOquery in R). Per-sample enrichment was calculated using the single-sample enrichment method -ssGSEA from the GSVA package⁷³ and further analyzed as described in the legend.

Formation-dependent recall

In total, 2 × 10⁵ naive congenically marked PD-1^+/+ Thy1.1⁺ or tdTomato CD8⁺ OT-I cells were isolated using a magnetic bead negative selection kit following the manufacturer’s protocol (Miltenyi) and adoptively transferred to congenically mismatched Thy1.2⁺ C57BL/6J recipients 1 day before infection. Recipient mice were vaccinated with VACV-OVAss on each ear with 1 × 10⁶ p.f.u. and on the base of the tail with 2 × 0⁶ p.f.u. Two hundred micrograms of anti-PD-1 or matched isotype (rat IgG2a, κ) was administered via the intraperitoneal route on days 0, 3, 6 and 9 after vaccination. Mice were started on FTY720 (25 μg per mouse) treatment 1 week (5 weeks after immunization) before skin peptide challenge and continued every other day throughout the rest of the experiment. On week 6 after infection, lateral flanks of each mouse were depilated before peptide challenge. Stock SIINFEKL (OVA_257–264; InvivoGen) was prepared to a 10 μg μl⁻¹ concentration dissolved in DPBS. A mixture of SIINFEKL, acetone and olive oil was prepared in a 1:7.2:1.8 ratio. An equivalent mixture of DBPS, acetone and olive oil was prepared as a control. Twenty-five micrograms of the SIINFEKL mixture was topically applied on the right shaved flank, whereas 25 μg of the control mixture was topically applied on the left shaved flank. Transferred OT-I cells quantitated on day 49 were classified as T_RM cells based on the expression of CD69⁺ and CD103^+/−, whereas T_CM and T_EM cells were identified based on the markers KLRG1⁻ and CD62L⁺ (T_CM) or CD62L⁻ (T_EM), respectively.

Statistics

Data are presented as mean ± s.e.m. or mean ± s.d. Group sizes were determined based on the results of preliminary experiments without predetermination of sample size. Preliminary experiments were performed to determine requirements for sample size, considering resources available and ethical, reductionist animal use. Mice were randomly assigned to groups. Mouse studies were not performed in a blinded fashion. Individual mice are reported as a symbol, with filled versus open symbols indicating replicate experiments. Statistical analyses using two-tailed, unpaired Student’s t-tests were performed to compare different groups within each time point when appropriate using Prism software (GraphPad) with significance reported for P ≤ 0.05.

For the indicated experiments, when repeated measures for the same mice were obtained for different time points, or different organs, or when complex correlation structures were detected within experiments, we opted for a MEM approach to maximize power. MEMs account for the correlation structure between samples and within experiments, modeling all sources of variation. In general, a random intercept was considered for each experimental unit and fixed factors, including treatment, time or group and its interactions. Models were fitted using the lme function from the nmle package in R. Models including heterogeneity of variance and random effect for other factors (for example, anatomic site) were also fitted, and the model with the minimum Akaike information criteria was chosen as the optimal model. Marginal means were estimated from the optimal model using the emmeans package, and hypotheses of interest were tested. Model assumptions were checked, and when departures from normality of the residuals were detected, log transformations were applied. All statistical tests were two sided.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

All statistical code is available at GitHub at https://github.com/maytesuarezfarinas/Anandasabapathy_NatureImmunology_2025/.

All RNA-seq code is available at Zenodo at https://zenodo.org/records/14213189 (ref. ⁷⁴).

References

Gebhardt, T. & Mackay, L. K. Skin-resident memory T cells keep herpes simplex virus at bay. Immunol. Cell Biol. 91, 441–442 (2013).
Article CAS PubMed Google Scholar
Liu, L. et al. Epidermal injury and infection during poxvirus immunization is crucial for the generation of highly protective T cell-mediated immunity. Nat. Med. 16, 224–227 (2010).
Article CAS PubMed PubMed Central Google Scholar
Masopust, D. et al. Dynamic T cell migration program provides resident memory within intestinal epithelium. J. Exp. Med. 207, 553–564 (2010).
Article CAS PubMed PubMed Central Google Scholar
Gaide, O. et al. Common clonal origin of central and resident memory T cells following skin immunization. Nat. Med. 21, 647–653 (2015).
Article CAS PubMed PubMed Central Google Scholar
Jiang, X. et al. Skin infection generates non-migratory memory CD8⁺ T_RM cells providing global skin immunity. Nature 483, 227–231 (2012).
Article CAS PubMed PubMed Central Google Scholar
Beura, L. K. & Masopust, D. SnapShot: resident memory T cells. Cell 157, 1488 (2014).
Article CAS PubMed Google Scholar
Teijaro, J. R. et al. Cutting edge: Tissue-retentive lung memory CD4 T cells mediate optimal protection to respiratory virus infection. J. Immunol. 187, 5510–5514 (2011).
Article CAS PubMed Google Scholar
Iijima, N. & Iwasaki, A. T cell memory. A local macrophage chemokine network sustains protective tissue-resident memory CD4 T cells. Science 346, 93–98 (2014).
Article CAS PubMed PubMed Central Google Scholar
Park, S. L., Gebhardt, T. & Mackay, L. K. Tissue-resident memory T cells in cancer immunosurveillance. Trends Immunol. 40, 735–747 (2019).
Article CAS PubMed Google Scholar
Park, S. L. et al. Tissue-resident memory CD8⁺ T cells promote melanoma–immune equilibrium in skin. Nature 565, 366–371 (2019).
Article CAS PubMed Google Scholar
Han, J. et al. Resident and circulating memory T cells persist for years in melanoma patients with durable responses to immunotherapy. Nat. Cancer 2, 300–311 (2021).
Article CAS PubMed PubMed Central Google Scholar
Wherry, E. J. & Kurachi, M. Molecular and cellular insights into T cell exhaustion. Nat. Rev. Immunol. 15, 486–499 (2015).
Article CAS PubMed PubMed Central Google Scholar
Sharpe, A. H. & Pauken, K. E. The diverse functions of the PD1 inhibitory pathway. Nat. Rev. Immunol. 18, 153–167 (2018).
Article CAS PubMed Google Scholar
Francisco, L. M. et al. PD-L1 regulates the development, maintenance, and function of induced regulatory T cells. J. Exp. Med. 206, 3015–3029 (2019).
Kumar, B. V. et al. Human tissue-resident memory T cells are defined by core transcriptional and functional signatures in lymphoid and mucosal sites. Cell Rep. 20, 2921–2934 (2017).
Article CAS PubMed PubMed Central Google Scholar
Jaiswal, A. et al. An activation to memory differentiation trajectory of tumor-infiltrating lymphocytes informs metastatic melanoma outcomes. Cancer Cell 40, 524–544 (2022).
Article CAS PubMed PubMed Central Google Scholar
Goldinger, S. M. et al. Cytotoxic cutaneous adverse drug reactions during anti-PD-1 therapy. Clin. Cancer Res. 22, 4023–4029 (2016).
Article CAS PubMed Google Scholar
Cheuk, S. et al. CD49a expression defines tissue-resident CD8⁺ T cells poised for cytotoxic function in human skin. Immunity 46, 287–300 (2017).
Article CAS PubMed PubMed Central Google Scholar
Larkin, J. et al. Combined nivolumab and ipilimumab or monotherapy in untreated melanoma. N. Engl. J. Med. 373, 23–34 (2015).
Article PubMed PubMed Central Google Scholar
Wang, D. Y., Johnson, D. B. & Davis, E. J. Toxicities associated with PD-1/PD-L1 blockade. Cancer J. 24, 36–40 (2018).
Article CAS PubMed PubMed Central Google Scholar
Freeman-Keller, M. et al. Nivolumab in resected and unresectable metastatic melanoma: characteristics of immune-related adverse events and association with outcomes. Clin. Cancer Res. 22, 886–894 (2016).
Article CAS PubMed Google Scholar
Wong, M. T. et al. A high-dimensional atlas of human T cell diversity reveals tissue-specific trafficking and cytokine signatures. Immunity 45, 442–456 (2016).
Article CAS PubMed Google Scholar
Matloubian, M. et al. Lymphocyte egress from thymus and peripheral lymphoid organs is dependent on S1P receptor 1. Nature 427, 355–360 (2004).
Article CAS PubMed Google Scholar
Cepek, K. L. et al. Adhesion between epithelial cells and T lymphocytes mediated by E-cadherin and the α^Eβ₇ integrin. Nature 372, 190–193 (1994).
Article CAS PubMed Google Scholar
Mackay, L. K. et al. The developmental pathway for CD103⁺CD8⁺ tissue-resident memory T cells of skin. Nat. Immunol. 14, 1294–1301 (2013).
Article CAS PubMed Google Scholar
Ahn, E. et al. Role of PD-1 during effector CD8 T cell differentiation. Proc. Natl Acad. Sci. USA 115, 4749–4754 (2018).
Article CAS PubMed PubMed Central Google Scholar
Hirai, T. et al. Keratinocyte-mediated activation of the cytokine TGF-β maintains skin recirculating memory CD8⁺ T cells. Immunity 50, 1249–1261 (2019).
Article CAS PubMed PubMed Central Google Scholar
Miller, B. C. et al. Subsets of exhausted CD8⁺ T cells differentially mediate tumor control and respond to checkpoint blockade. Nat. Immunol. 20, 326–336 (2019).
Article CAS PubMed PubMed Central Google Scholar
Doedens, A. L. et al. Molecular programming of tumor-infiltrating CD8⁺ T cells and IL15 resistance. Cancer Immunol. Res. 4, 799–811 (2016).
Article CAS PubMed PubMed Central Google Scholar
Mann, T. H. & Kaech, S. M. Tick-TOX, it’s time for T cell exhaustion. Nat. Immunol. 20, 1092–1094 (2019).
Article CAS PubMed Google Scholar
Gattinoni, L. et al. A human memory T cell subset with stem cell-like properties. Nat. Med. 17, 1290–1297 (2011).
Article CAS PubMed PubMed Central Google Scholar
Im, S. J. et al. Defining CD8⁺ T cells that provide the proliferative burst after PD-1 therapy. Nature 537, 417–421 (2016).
Article CAS PubMed PubMed Central Google Scholar
Charlton, J. J. et al. Programmed death-1 shapes memory phenotype CD8 T cell subsets in a cell-intrinsic manner. J. Immunol. 190, 6104–6114 (2013).
Article CAS PubMed Google Scholar
Pan, Y. et al. Survival of tissue-resident memory T cells requires exogenous lipid uptake and metabolism. Nature 543, 252–256 (2017).
Article CAS PubMed PubMed Central Google Scholar
Odorizzi, P. M., Pauken, K. E., Paley, M. A., Sharpe, A. & Wherry, E. J. Genetic absence of PD-1 promotes accumulation of terminally differentiated exhausted CD8⁺ T cells. J. Exp. Med 212, 1125–1137 (2015).
Article CAS PubMed PubMed Central Google Scholar
Sarkar, S. et al. Functional and genomic profiling of effector CD8 T cell subsets with distinct memory fates. J. Exp. Med 205, 625–640 (2008).
Article CAS PubMed PubMed Central Google Scholar
Doering, T. A. et al. Network analysis reveals centrally connected genes and pathways involved in CD8⁺ T cell exhaustion versus memory. Immunity 37, 1130–1144 (2012).
Article CAS PubMed PubMed Central Google Scholar
Zaid, A. et al. Chemokine receptor-dependent control of skin tissue-resident memory T cell formation. J. Immunol. 199, 2451–2459 (2017).
Article CAS PubMed Google Scholar
Horiguchi, M., Ota, M. & Rifkin, D. B. Matrix control of transforming growth factor-β function. J. Biochem. 152, 321–329 (2012).
Article CAS PubMed PubMed Central Google Scholar
Mohammed, J. et al. Stromal cells control the epithelial residence of DCs and memory T cells by regulated activation of TGF-β. Nat. Immunol. 17, 414–421 (2016).
Article CAS PubMed PubMed Central Google Scholar
Mackay, L. K. et al. T-box transcription factors combine with the cytokines TGF-β and IL-15 to control tissue-resident memory T cell fate. Immunity 43, 1101–1111 (2015).
Article CAS PubMed Google Scholar
Mani, V. et al. Migratory DCs activate TGF-β to precondition naive CD8⁺ T cells for tissue-resident memory fate. Science 366, eaav5728 (2019).
Article CAS PubMed PubMed Central Google Scholar
Hirai, T. et al. Competition for active TGFβ cytokine allows for selective retention of antigen-specific tissue-resident memory T cells in the epidermal niche. Immunity 54, 84–98 (2021).
Article CAS PubMed Google Scholar
Nath, A. P. et al. Comparative analysis reveals a role for TGF-β in shaping the residency-related transcriptional signature in tissue-resident memory CD8⁺ T cells. PLoS ONE 14, e0210495 (2019).
Article CAS PubMed PubMed Central Google Scholar
Christo, S. N. et al. Discrete tissue microenvironments instruct diversity in resident memory T cell function and plasticity. Nat. Immunol. 22, 1140–1151 (2021).
Article CAS PubMed Google Scholar
Mackay, L. K. et al. Long-lived epithelial immunity by tissue-resident memory T (T_RM) cells in the absence of persisting local antigen presentation. Proc. Natl Acad. Sci. USA 109, 7037–7042 (2012).
Article CAS PubMed PubMed Central Google Scholar
Khan, T. N., Mooster, J. L., Kilgore, A. M., Osborn, J. F. & Nolz, J. C. Local antigen in nonlymphoid tissue promotes resident memory CD8⁺ T cell formation during viral infection. J. Exp. Med. 213, 951–966 (2016).
Article CAS PubMed PubMed Central Google Scholar
Zaid, A. et al. Persistence of skin-resident memory T cells within an epidermal niche. Proc. Natl Acad. Sci. USA 111, 5307–5312 (2014).
Article CAS PubMed PubMed Central Google Scholar
Enamorado, M. et al. Enhanced anti-tumour immunity requires the interplay between resident and circulating memory CD8⁺ T cells. Nat. Commun. 8, 16073 (2017).
Article CAS PubMed PubMed Central Google Scholar
Park, C. O. & Kupper, T. S. The emerging role of resident memory T cells in protective immunity and inflammatory disease. Nat. Med. 21, 688–697 (2015).
Article CAS PubMed PubMed Central Google Scholar
Schenkel, J. M. et al. T cell memory. Resident memory CD8 T cells trigger protective innate and adaptive immune responses. Science 346, 98–101 (2014).
Article CAS PubMed PubMed Central Google Scholar
Ariotti, S. et al. T cell memory. Skin-resident memory CD8⁺ T cells trigger a state of tissue-wide pathogen alert. Science 346, 101–105 (2014).
Article CAS PubMed Google Scholar
Park, S. L. et al. Local proliferation maintains a stable pool of tissue-resident memory T cells after antiviral recall responses. Nat. Immunol. 19, 183–191 (2018).
Article CAS PubMed Google Scholar
Van Braeckel-Budimir, N., Varga, S. M., Badovinac, V. P. & Harty, J. T. Repeated antigen exposure extends the durability of influenza-specific lung-resident memory CD8⁺ T cells and heterosubtypic immunity. Cell Rep. 24, 3374–3382 (2018).
Article PubMed Google Scholar
Nizard, M. et al. Induction of resident memory T cells enhances the efficacy of cancer vaccine. Nat. Commun. 8, 15221 (2017).
Article CAS PubMed PubMed Central Google Scholar
Chambers, C. A., Sullivan, T. J. & Allison, J. P. Lymphoproliferation in CTLA-4-deficient mice is mediated by costimulation-dependent activation of CD41 T cells. Immunity 7, 885–895 (1997).
Article CAS PubMed Google Scholar
Mahoney, K. M., Freeman, G. J. & McDermott, D. F. The next immune-checkpoint inhibitors: PD-1/PD-L1 blockade in melanoma. Clin. Ther. 37, 764–782 (2015).
Article CAS PubMed PubMed Central Google Scholar
Fife, B. T. & Bluestone, J. A. Control of peripheral T-cell tolerance and autoimmunity via the CTLA-4 and PD-1 pathways. Immunol. Rev. 224, 166–182 (2008).
Article CAS PubMed Google Scholar
Bartholin, L. et al. Generation of mice with conditionally activated transforming growth factor β signaling through the TβRI/ALK5 receptor. Genesis 46, 724–731 (2008).
Article CAS PubMed Google Scholar
Vincent, D. F. et al. A rapid strategy to detect the recombined allele in LSL–TβRICA transgenic mice. Genesis 48, 559–562 (2010).
Article CAS PubMed PubMed Central Google Scholar
Andrews, L. A. et al. A Cre-driven allele-conditioning line to interrogate CD4⁺ conventional T cells. Immunity 54, 2209–2217 (2021).
Article CAS PubMed Google Scholar
Cotter, C.A., Earl, P. L., Wyatt, L. S. & Moss, B. Preparation of cell cultures and vaccinia virus stocks. Curr. Protoc. Protein Sci. 89, 5.12.1–5.12.18 (2017).
Article PubMed Google Scholar
Anandasabapathy, N. et al. Classical FLT3L-dependent dendritic cells control immunity to protein vaccine. J. Exp. Med. 211, 1875–1891 (2014).
Article CAS PubMed PubMed Central Google Scholar
Kim, T. G. et al. CCCTC-binding factor controls the homeostatic maintenance and migration of Langerhans cells. J. Allergy Clin. Immunol. 136, 713–724 (2015).
Article CAS PubMed Google Scholar
Lindquist, R. L. et al. Visualizing dendritic cell networks in vivo. Nat. Immunol. 5, 1243–1250 (2004).
Article CAS PubMed Google Scholar
Picelli, S. et al. Full-length RNA-seq from single cells using Smart-seq2. Nat. Protoc. 9, 171–181 (2014).
Article CAS PubMed Google Scholar
Liao, Y., Smyth, G. K. & Shi, W. The Subread aligner: fast, accurate and scalable read mapping by seed-and-vote. Nucleic Acids Res. 41, e108 (2013).
Article PubMed PubMed Central Google Scholar
Liao, Y., Smyth, G. K. & Shi, W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics 30, 923–930 (2014).
Article CAS PubMed Google Scholar
Risso, D., Ngai, J., Speed, T. P. & Dudoit, S. Normalization of RNA-seq data using factor analysis of control genes or samples. Nat. Biotechnol. 32, 896–902 (2014).
Article CAS PubMed PubMed Central Google Scholar
Lin, Y. et al. Evaluating stably expressed genes in single cells. Gigascience 8, giz106 (2019).
Article PubMed PubMed Central Google Scholar
Robinson, M. D., McCarthy, D. J. & Smyth, G. K. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 26, 139–140 (2010).
Article CAS PubMed Google Scholar
Cavill, R. et al. Consensus-phenotype integration of transcriptomic and metabolomic data implies a role for metabolism in the chemosensitivity of tumour cells. PLoS Comput. Biol. 7, e1001113 (2011).
Article CAS PubMed PubMed Central Google Scholar
Hanzelmann, S., Castelo, R. & Guinney, J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics 14, 7 (2013).
Article PubMed PubMed Central Google Scholar
Sanjana P. Devi, K. et al. Programmed death-1 is requisite for skin T_RM formation and specification by TGFβ. Zenodo https://doi.org/10.5281/zenodo.14213189 (2024).

Download references

Acknowledgements

We are grateful to the Sharpe and Drake laboratories for providing B16-OVA and EL4-OVA tumor lines and to the laboratory of A. Sharpe for PD-1^−/− Thy1.1 OT-I and PD-L1/PD-L2^−/− mice and to D. Vignali and L. Bartholin who generated tamoxifen-inducible E8i-CreERT2-Tgfbrca^fl/fl Thy1.1/Thy1.2 mice. We thank C. Cabalin and other members of the laboratory of N.A. for critical reading of the manuscript. S.D. was supported in part by a fellowship from SunPharma. C.J.N. was supported by 5T32AR007098, a Dermatology Training Grant. T.S.K. is supported by R01AI041707 (to T.S.K.) and TR01AI097128 (to T.S.K.). R.A.C. is supported by NIH R01CA203721 NIH/NCI (to R.A.C.) and R01AR063962 NIH/NIAMS (to R.A.C.). S.N. is supported by the Burrows Welcome Path forward award, the NIH (1K22AI135099-01, 1DP2AR079173-01 and R01-AI168462) and Neuroimmune Allen Discovery Center and is an NYSCF Robertson Stem Cell Investigator and a Packard Fellow. P.K. is supported by NIH 7K99AR083536-02. G.J.F. is supported by P01AI056299. N.A. and Y.L. were supported by a Melanoma Research Alliance-Brigham and Women’s Hospital combined grant (to N.A.). Additional support was provided by the National Institute of Arthritis and Musculoskeletal and Skin Disease (R01 AR080436 and AR083208; to N.A.) and Weill Cornell Medicine.

Author information

These authors contributed equally: K. Sanjana P. Devi, Eric Wang.

Authors and Affiliations

Department of Dermatology, Meyer Cancer Center, Weill Cornell Medicine, New York, NY, USA
K. Sanjana P. Devi, Eric Wang, Abhinav Jaiswal, Yong Liu, Julia Milczanowski & Komal Pradhan
Department of Immunology and Immunotherapy, Icahn School of Medicine at Mt. Sinai, New York, NY, USA
Piotr Konieczny & Shruti Naik
Department of Microbiology and Immunology, Yonsei University College of Medicine, Seoul, South Korea
Tae-Gyun Kim & Karyn Haitz
Department of Dermatology, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USA
Christopher J. Nirschl, Pinru Wu, Sandra L. King, Sze-Wah Tse, Xiaodong Jiang, Tian Tian, Robert C. Fuhlbrigge & Chrysalyne D. Schmults
Institute for Computational Biomedicine, Caryl and Israel Englander Institute for Precision Medicine, Weill Cornell Medicine, New York, NY, USA
Akanksha Verma & Olivier Elemento
Department of Microbiology and Immunology, The University of Melbourne at The Peter Doherty Institute for Infection and Immunity, Melbourne, Victoria, Australia
Susan N. Christo, Luke C. Gandolfo & Laura K. Mackay
School of Mathematics and Statistics, The University of Melbourne, Melbourne, Victoria, Australia
Luke C. Gandolfo
Walter and Eliza Hall Institute for Medical Research, Parkville, Victoria, Australia
Luke C. Gandolfo
Department of Immunology, Mayo Clinic, Scottsdale, AZ, USA
Trupti D. Vardam
Department of Dermatology, Brigham and Women’s Hospital, Harvard Medical School, Dana-Farber/Brigham and Women’s Cancer Center, Boston, MA, USA
Rachael A. Clark & Thomas S. Kupper
Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, MA, USA
Gordon J. Freeman
Department of Dermatology, Icahn School of Medicine at Mount Sinai, New York, NY, USA
Shruti Naik
Fred Hutchinson Cancer Research Center, Vaccine and Infectious Disease Division, Seattle, WA, USA
Evan W. Newell
Department of Genetics and Genomic Science, Icahn School of Medicine at Mount Sinai, New York, NY, USA
Mayte Suarez-Farinas
Department of Population Health Science and Policy, Icahn School of Medicine at Mount Sinai, New York, NY, USA
Mayte Suarez-Farinas
Department of Dermatology, Englander Institute for Precision Medicine at Weill Cornell Medicine, New York, NY, USA
Niroshana Anandasabapathy
Parker Institute for Cancer Immunotherapy and Sandra and Edward Meyer Cancer Center, Weill Cornell Medicine, New York, NY, USA
Niroshana Anandasabapathy
Department of Microbiology and Immunology, Englander Institute for Precision Medicine at Weill Cornell Medicine, New York, NY, USA
Niroshana Anandasabapathy
Immunology and Microbial Pathogenesis Program, Weill Cornell Medical College, New York, NY, USA
Niroshana Anandasabapathy

Authors

K. Sanjana P. Devi
View author publications
Search author on:PubMed Google Scholar
Eric Wang
View author publications
Search author on:PubMed Google Scholar
Abhinav Jaiswal
View author publications
Search author on:PubMed Google Scholar
Piotr Konieczny
View author publications
Search author on:PubMed Google Scholar
Tae-Gyun Kim
View author publications
Search author on:PubMed Google Scholar
Christopher J. Nirschl
View author publications
Search author on:PubMed Google Scholar
Akanksha Verma
View author publications
Search author on:PubMed Google Scholar
Yong Liu
View author publications
Search author on:PubMed Google Scholar
Julia Milczanowski
View author publications
Search author on:PubMed Google Scholar
Susan N. Christo
View author publications
Search author on:PubMed Google Scholar
Luke C. Gandolfo
View author publications
Search author on:PubMed Google Scholar
Karyn Haitz
View author publications
Search author on:PubMed Google Scholar
Trupti D. Vardam
View author publications
Search author on:PubMed Google Scholar
Pinru Wu
View author publications
Search author on:PubMed Google Scholar
Sandra L. King
View author publications
Search author on:PubMed Google Scholar
Sze-Wah Tse
View author publications
Search author on:PubMed Google Scholar
Komal Pradhan
View author publications
Search author on:PubMed Google Scholar
Xiaodong Jiang
View author publications
Search author on:PubMed Google Scholar
Tian Tian
View author publications
Search author on:PubMed Google Scholar
Robert C. Fuhlbrigge
View author publications
Search author on:PubMed Google Scholar
Chrysalyne D. Schmults
View author publications
Search author on:PubMed Google Scholar
Rachael A. Clark
View author publications
Search author on:PubMed Google Scholar
Thomas S. Kupper
View author publications
Search author on:PubMed Google Scholar
Gordon J. Freeman
View author publications
Search author on:PubMed Google Scholar
Laura K. Mackay
View author publications
Search author on:PubMed Google Scholar
Shruti Naik
View author publications
Search author on:PubMed Google Scholar
Evan W. Newell
View author publications
Search author on:PubMed Google Scholar
Olivier Elemento
View author publications
Search author on:PubMed Google Scholar
Mayte Suarez-Farinas
View author publications
Search author on:PubMed Google Scholar
Niroshana Anandasabapathy
View author publications
Search author on:PubMed Google Scholar

Contributions

N.A., K.S.P.D. and E.W. conceptualized the experiments, performed the data analyses and interpretation and wrote the manuscript. K.S.P.D. and E.W., with help from T.-G.K., C.J.N., Y.L., J.M., P.W., S.-W.T. and K.P., performed all mouse flow cytometry, viral load and imaging experiments, collected data and performed data analysis. E.W.N. planned and executed CyTOF human experiments and analyses. P.K. and S.N. contributed to planning, data collection and analysis of mouse images. A.J., A.V., N.A., O.E. and M.S.-F. performed planning and execution of the genomics analyses. S.N.C., L.C.G. and L.K.M. planned and executed TGFβR2 HSV experiments and provided genomics datasets, and L.K.M. contributed to the analysis from these sets with A.J. and N.A. S.L.K., X.J., R.C.F., T.S.K., T.D.V. and T.T. provided experimental planning, viral training and methodologic support for VACV growth and titering, which was performed by K.S.P.D. K.H., R.A.C. and C.D.S. provided support for data acquistion of human skin. G.J.F. provided anti-PD-1 and paired isotype control antibodies and experimental design planning related to their use. M.S.-F. provided statistical analysis and statistical code across the entire manuscript.

Corresponding author

Correspondence to Niroshana Anandasabapathy.

Ethics declarations

Competing interests

G.J.F. has patents/pending royalties on the PD-1/PD-L1 pathway from Roche, Merck MSD, Bristol-Myers-Squibb, Merck KGA, Boehringer-Ingelheim, AstraZeneca, Dako, Leica, Mayo Clinic and Novartis. G.J.F. has served on advisory boards for Roche, Bristol-Myers-Squibb, Xios, Origimed, Triursus, iTeos, NextPoint, IgM, Jubilant, Trillium and GV20. G.J.F. has equity in Nextpoint, Triursus, Xios, iTeos, IgM and GV20. S.N. is cofounder of Stara Biosciences and serves on the scientific advisory board of Seed. E.W.N. is a cofounder, advisor and shareholder of ImmunoScape and is an advisor for Neo-gene Therapeutics. C.J.N. is an employee and shareholder of Werewolf Therapeutics. M.S.-F. has been a consultant for Symbrio, SpectralMD and DBV. R.A.C. is on the scientific advisory board for AAAS, Almirall and Sedec Therapeutics. N.A. serves on the scientific advisory board of Shennon Biosciences and Panther Life Sciences and is a consultant or lecturer for Johnson and Johnson, Immunitas, 3T, 23andme, Cellino, Kumquat, Verrica, Lytix, Network Bio and Genmab. The other authors have no conflicts to declare.

Peer review

Peer review information

Nature Immunology thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editor: Ioana Staicu, in collaboration with the Nature Immunology team. Peer reviewer reports are available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 PD-1 is rapidly expressed on activated T cells but retained on TRM after VACV-OVAss.

a. Human skin and b. PBMCs showing the relative distribution of NK, CD8+ and CD4 + T cell clusters by CyTOF analysis. c. Quantification of cell percent derived from skin phenograph or louvain clusters. Green lines represent 80% skin specific clusters while blue lines are clusters that are less than 80% skin specific. d. Schematic showing Thy1.1 C57BL/6J OT-1 cells were transferred into Thy1.2 C57BL/6J mice at day −1 followed by VACV-OVAss immunization on day 0 and early harvest kinetics (top). PD-1 expression on OT-I cells isolated from the draining LN (dLN), non-draining LNs, and spleen at 0, 48, 72 and 120 hours after mice received 2 × 106 viral pfu on ear by VACV-OVAss. Flow cytometry of data from one individual mouse is shown per time point (bottom). e. Representative flow cytometry of Vα2 + Vβ5 + CD8+ populations 0, 48, 72 and 120 hours post-infection. f. Day 42 representative flow cytometry gating showing percent of cells with PD-1 cell surface expression using anti-PD-1 clone: RMP1-30 or g. using anti-PD-1 clone: 29 F.1A12. Quantification of percent fraction of PD-1 expressing skin TRM relative to LN and spleen TCM or TEM populations at 6 weeks using anti-PD-1 clone 29 F.1A12 (left). Each dot represents individual mice. h. Schematic showing Thy1.2 + OT-I cell were transferred into Thy1.1 + C57BL/6J mice at day −1, followed by VACV-OVAss (4 × 106 viral pfu on ear and tail) on day 0, and skin, LN and spleen harvest at 1 to 6 weeks post-infection (left). PD-1 geometric mean fluorescence intensity (MFI) on skin, spleen, and LN CD8 + OT-I cell subsets between 1 to 6 weeks post infection (middle). CD69 and CD103 cell-surface expression was used to gate four quadrants of CD8 + OT-I cells in skin. TRM were gated from Va2 + CD8+Thy1.2+ donor cells, and defined as CD69 + CD103+ (epidermal) or CD69 + CD103- (dermal). TCM and TEM were identified by KLRG1-/CD62L+ and KLRG1-/CD62L- gating respectively, while TEffector were defined by KLRG1 + CD62L +/− in LN and spleen (right). i. Schematic of OT-I Thy1.1 transfer into Thy1.2 recipients with immunization as in h. but with harvest of flank, ear, and tail skin at day 10–12 post-infection (top). PD-1 representative histograms of epidermal CD69 + CD103+ or dermal CD69 + CD103- skin OT-I TRM of ear, flank, and tail skin were concatenated from 5 mice per site (bottom). j. % PD-1 positive cells in flank, ear, and tail dermal and epidermal TRM, with mean fluorescence intensity (MFI) of PD-1+ populations. An unpaired T-test was performed for g. Bars and error bars show mean +/- SEM. +p ≤ 0.1 * p ≤ 0.05; ** p ≤ 0.01; *** p ≤ 0.001; **** p ≤ 0.0001.

Extended Data Fig. 2 Comparative analysis of PD-1−/− and PD-1 +/+ skin T cells post infection, and naïve T cells.

a. Schematic showing mixed adoptive transfer of either 1 × 105 each of PD-1 +/+ (red) with PD1−/− (blue) OT-I cells, or PD-1 +/+ OT-I (dark gray) with PD-1 +/+ OT-I (light gray) littermate gender-matched controls into C57BL/6J recipients at day −1, followed by VACV-OVAss immunization at day 0 (top). Donor OT-I quantification was performed from LN and spleen on days 10, 21 and 42 post-infection (bottom). Mixed pairs were distinguished using CD45.1, CD45.2, Thy1.2 and/or Thy1.1 congenic markers. Data were pooled from 3 independent experiments (n = 14) at day 10, 6 independent experiments (n = 30) at day 21, and 4 independent experiments (n = 20) at day 42. Control C57BL/6J mice were pooled from 2 independent experiments (n = 9) at day 10; 1 independent experiment (n = 4) at day 21 and 3 independent experiment (n = 12) at day 42. Statistical significance between the number of PD-1 +/+ OT-I (red box) and PD-1−/− OT-I cells (blue box). *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001 Total number (log) of CD3 + CD8 + Va2 + CD62L + TCM in PD-1 +/+ OT-I (red) and PD-1−/− OT-I cells (blue) isolated from LN. CD8 + TCM were identified in the LN based on the expression of CD44 + /CD127+ and CD62L+ gating. b. Representative flow cytometry of LN OT-I cells. PD-1 +/+ OT-I (red) and PD-1−/− OT-I cells (blue). c. Total number (log) of CD3 + CD8 + Va2 + , CD3 + CD8 + Va2 + CD62L + TCM and CD3 + CD8 + Va2 + CD62L- TEM in PD-1 +/+ OT-I (red) and PD-1−/− OT-I cells (blue) isolated from LN and spleen of hosts, after mixed transfers. Quantification of OT-I cells from mixed transfer controls receiving PD1 +/+ C57BL/6J OT-I cells (dark gray lines) mixed 1:1 with PD-1 +/+ OT-I+ littermates to the PD-1−/− (light gray lines) was included to rule out genetic mismatch. TCM and TEM were identified in spleen based on the expression of CD44 + /CD127+ and CD62L+ or CD62L- respectively on CD8 + T cells. In some littermate control analyses, TCM and TEM were identified by KLRG1-/CD62L+ and KLRG1-/CD62L- gating respectively. Significance of the difference between PD-1 +/+ OT-I (red) and PD-1−/− OT-I cells (blue) OT-I cells isolated after mixed transfer. d. Representative flow cytometry of spleen OT-I cells 10, 21, and 42 days post-infection in mice receiving mixed transfers of PD-1 +/+ OT-I (red box) and PD-1−/− OT-I cells (blue box) OT-I cells. Statistical analysis was estimated using a linear mixed-effects model with group by time interaction, and replication as fixed effects and a random intercept for each mouse. Bars and error bars show mean ± SEM. +p ≤ 0.1 * p ≤ 0.05; ** p ≤ 0.01; *** p ≤ 0.001; **** p ≤ 0.0001.

Extended Data Fig. 3 LN early BrdU, Ki67, and live/dead comparisons; early anti-PD1 effects on circulating memory in spleen & LN.

a. Schematic showing mixed adoptive transfer of 1 × 105 each of CD45.1 + /Thy1.1+ or Thy1.1 + WT and Thy1.1+ or CD45.1 + /Thy1.1 + KO OT-I cells into Thy1.2 + C57BL/6J recipients at day −1, followed by VACV-OVAss immunization on ear and tail at day 0 (top). BrdU was given to mice starting on day 0 and provided on consecutive days between days 0–2, 0–4 or 0–6 until harvest on day 3, 5, and 7 respectively. Data were pooled from 2 independent experiments each for day 3 (n = 12), 5 (n = 12), and 7 (n = 12). Mixed pairs were distinguished using CD45.1, CD45.2, Thy 1.2 and/or Thy 1.1 congenic markers. Percent (left) and total counts (right) of PD-1 +/+ OT-I and PD-1−/− donor OT-I cells are given with a line to connect the paired samples by mouse. b. Percent (left) and total counts (right) of BrdU+ cells from the PD-1 +/+ OT-I and PD-1−/− donor OT-I. c. Percent (left) and total counts (right) of Ki67+ cells from PD-1 +/+ OT-I and PD-1−/− donor OT-I. d. Percent (left) and total counts (right) of Live/Dead Blue+ cells from PD-1 +/+ OT-I and PD-1−/− donor OT-I. e. Schematic showing tdTomato+ or Thy1.1 + OT-I cells were transferred into Thy1.2 + C57BL/6J recipients at day −1, followed by VACV-OVAss immunization at day 0, and treatment with either PD-1 Ab or isotype administered on days 0, 3, 6, and 9 post-infection (top). FTY720 was given every 2 days starting at day 35. At day 42, contralateral R and L depilated flank sites of the same mouse were challenged with topical PBS or SIINFEKL. dLN and spleen were harvested at day 49. Quantification of total KLRG1-CD62L + OT-I TCM numbers (log) from the dLN of the PBS-challenged skin site versus those from the dLN of the SIINFEKL-challenged skin site (bottom, left). Quantification of spleen KLRG1-CD62L + TCM (bottom, middle) and KLRG1-CD62L- TEM (bottom, right). A paired t-test was performed for a-d. Elsewhere, statistics were estimated using a linear mixed-effects model with group and by treatment challenge, as fixed effects and a random intercept for each mouse. Comparisons of KLRG1-CD62L + TCM numbers in PBS versus SIINFEKL-dLN of the same mouse, are shown in red or blue. Data were pooled from 4 independent experiments (Isotype, n = 13; anti-PD-1, n = 18). Bars and error bars show mean ± SEM. +p ≤ 0.1 * p ≤ 0.05; ** p ≤ 0.01; *** p ≤ 0.001; **** p ≤ 0.0001.

Extended Data Fig. 4 Transcriptomics supporting selection of PD-1 sufficient T cell in skin.

a. Schematic of mixed adoptive transfer of PD-1 +/+ with PD-1−/− OT-I T cells, VACV-OVAss (4 × 106 viral p.f.u. on ear and tail), and d14 isolation from skin (top). ATAC sequencing of the PD-1 gene locus shows the lack of PD-1 accessibility in naïve T cells and confirms the presence or absence of exons 2 and 3 in PD-1−/− OT-I T cells (bottom). b. Relative enrichment of ECM/GPCR pathway scores comparing donor skin gB-T CD8 + CD103 + TRM isolated from the skin of C57/BL6 mice 30 days post-infection with HSV, versus naive gB-T cells (GSE47045). Pathway enrichment scores were generated using the GSVA ssGSEA algorithm. c. Relative enrichment of ECM/GPCR pathway scores during VACV-SIINFEKL skin OT-I TRM differentiation (GSE79805). Pathway enrichment scores were generated using the GSVA ssGSEA algorithm and normalized by subtracting naïve T cell scores from all samples. d. Scatter plot depicting expression differences of transcripts in PD-1 +/+ vs PD-1−/− T cells in skin and spleen with DEGs in skin (logFC≥1.5, FDR ≤ 0.05, green), spleen (FDR ≤ 0.05, p ≤ 0.05, yellow) or both (purple), and % of all DEGs (n = 328) within each quadrant. Genes that were not differentially expressed by either skin or spleen T cells are depicted as a contour plot (gray). e. Expression of transcription factors (T-bet, Eomes) in PD-1 +/+ OT-I and PD-1−/− OT-I T cells isolated from LN and spleen at day 10, and expression levels of cell surface markers (CD127, KLRG1) on PD-1 +/+ OT-I and PD-1−/− donor OT-I T cells isolated from ear skin of mice on day 21 or 42 post VACV-OVAss. f. Fisher exact test of exclusive T cell differentiation state signatures in Shared, PD-1−/−specific, or WTspecific programs. g. Fisher exact test of non-exclusive (global) CD8 + T cell differentiation state signatures in shared, PD-1−/−specific, or WTspecific programs. Up and down triangles indicate enrichment driven by transcripts either upregulated or downregulated respectively in day 14 skin OT-I relative to naïve controls. Both size and color indicate the level of enrichment.

Extended Data Fig. 5 WTspecific and Shared program scoring to T cells activated +/- TGFβ and IL-2.

a. Gene Set Variation Analysis score of Shared (purple) and WTspecific (red) transcripts upregulated in skin OT-I vs naïve controls (Sharedup, WTspecific Up, Y-axis) across the transcriptomes from TGFβ/IL2/TGFβ + IL2 stimulated OT-I (GSE125471). Gray boxes represent relative activity of an equal number of random probes. The scaled GSVA score was modelled using MEM (see Methods) with treatment as a fixed effect and a random intercept for each experiment.

Extended Data Fig. 6 Anti-PD-1 administration during priming does not impact T cell accumulation in LN and spleen.

a. Schematic showing Thy1.1 OT-I cell transfer at day −1 into C57Bl/6J Thy.2 recipients with VACV-OVAss immunization at day 0, and treatment with either isotype or anti-PD-1 0, 3, 6, 9 days post-infection (top). LN and spleen were harvested between days 10–12. Quantification of total donor CD8 + OT-I cell number (log). Data were pooled from 6 independent experiments (n = 29 mice per group). Individual mice are shown (middle) and mean per experiment (bottom) b. Schematic showing identical setup as in a. but either isotype or anti-PD-1 was administered 6, 9, and 12 days post-infection (top). Ear skin, LN, and spleen were harvested at day 13. Quantification of total donor skin CD69 + CD103 + /-CD8 + OT-I cells and LN (log), spleen (log) donor CD8 + OT-I cells. Data were pooled from 2 independent experiments pooled (Isotype, n = 9; anti-PD-1, n = 10). Individual animals are shown (middle) and mean per experiment (bottom) c. Schematic showing identical setup as in a. Representative histogram of cell-surface PD-1, TGFβRI and TGFβRII in CD69 + CD103 + /-CD8 + OT-I TRM cells isolated from ear. Histogram of cell-surface PD-1 in OT-I cells isolated from LN and spleen compared to isotype stain. d. Schematic showing identical setup as in a. but with E8i-CreERT2 TGFβRCAfl/+ Thy 1.1/1.2 OT-I cells (top). Tamoxifen was additionally given for 5 consecutive days from days 0–4. Quantification of donor CD8 + TGFβRCAfl/+ Thy 1.1/1.2 OT-I cells (log) from LN and spleen. Data were pooled from 4 independent experiments (Isotype, n = 24; anti-PD-1, n = 22). Individual animals are shown (middle) and mean per experiment (bottom) e. Representative cell-surface PD-1 levels in LN and spleen TGFβRCA fl/+ OT-I cells as in d. to show anti-PD1 targeting effect compared to isotype staining. The difference between anti-PD-1 and isotype treated groups (a and d) were modelled using log transformed data (see statistics section). Statistics were estimated using a linear mixed-effects model with group by treatment interaction, anatomic site and phenotype when relevant, as fixed effects and a random intercept for each mouse or replicate. Bars and error bars show mean ± SEM. +p ≤ 0.1 * p ≤ 0.05; ** p ≤ 0.01; *** p ≤ 0.001; **** p ≤ 0.0001.

Extended Data Fig. 7 In vitro activation of OT-I and engraftment.

a. Schematic of ‘Prime and Pull’ to isolate TRM engraftment (left). Prime: Thy1.1 or TdTomato+ OT-I splenocytes were activated in vitro by culture with SIINFEKL peptide and IL-2. Pull: At day 3.5, activated OT-I cells were adoptively transferred in vivo into Thy1.2 C57BL/6J recipients by intravenous (i.v.) tail vein injection and ‘pulled’ into skin without antigen (engraftment) using 0.5% DNFB (a hapten) applied topically on depilated flank skin to induce inflammation. Mice were treated with either anti-PD-1 or isotype 0, 3, 6, and 9 days after DNFB application. In some groups, TGF-β1 cytokine (0.5 µg) was administered daily throughout the engraftment period (d1-9). Quantification of percent Live/Dead Aqua+ of donor OT-I cells isolated from flank skin on day 10 from groups without TGF-β1 treatment. Each dot represents individual mice (middle) and mean per experiment (right). Data were pooled from 6 independent experiments (Isotype, n = 21; anti-PD-1, n = 23). b. Quantification of donor OT-I CD44 + CD62L + TCM and CD44 + CD62L− TEM from LN and spleen of recipient mice as in a. Individual animals are shown (top) and mean per experiment (bottom). Data were pooled from 6 independent experiments in the groups without TGF-β1 treatment (Isotype, n = 21; anti-PD-1, n = 23) and 3 independent experiments for the TGF-β1 treated groups (Isotype, n = 17; anti-PD-1, n = 17). The difference between plus and minus TGFβ groups (b) were modelled using log transformed data (see statistics section). c. Schematic showing identical setup as in a. but with activated E8i-CreERT2 TGFβRCAfl/+ Thy 1.1 + /1.2 + OT-I cells transferred into Thy1.2 + C57BL/6J recipients and no soluble TGF-β treatment was given (top). Tamoxifen was additionally administered on 6 consecutive days between days 0–5 of the engraftment period. LN and spleen OT-I cells were harvested on day 10 post-infection. Quantification of LN and spleen donor OT-I cells (log), LN OT-I CD44 + KLRG1-CD62L + TCM (log), spleen OT-I CD44 + KLRG1-CD62L + TCM (log) and CD44 + KLRG1-CD62L- TEM (log). Individual animals are shown (middle) and mean per experiment (bottom). Data were pooled from 3 independent experiments (Isotype, n = 13; anti-PD-1, n = 13). Statistics were estimated using a linear mixed-effects model with group by treatment interaction, anatomic site and phenotype when relevant, as fixed effects and a random intercept for each mouse or replicate. Bars and error bars show mean ± SEM. * p ≤ 0.05; ** p ≤ 0.01; *** p ≤ 0.001; **** p ≤ 0.0001. d. Summary showing role of PD-1 on specifying TRM formation.

Supplementary information

Supplementary Information (download PDF )

Supplementary Figs. 1–5 and figure legends.

Reporting Summary (download PDF )

Peer Review File (download PDF )

Supplementary Table 1 (download XLSX )

Supplementary Table 1.

Supplementary Table 2 (download XLSX )

Supplementary Table 2.

Supplementary Table 3 (download XLSX )

Supplementary Table 3.

Supplementary Table 4 (download XLSX )

Supplementary Table 4.

Supplementary Table 5 (download XLSX )

Supplementary Table 5.

Supplementary Table 6 (download XLSX )

Supplementary Table 6.

Supplementary Table 7 (download XLS )

Supplementary Table 7.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Devi, K.S.P., Wang, E., Jaiswal, A. et al. PD-1 is requisite for skin T_RM cell formation and specification by TGFβ. Nat Immunol 26, 1339–1351 (2025). https://doi.org/10.1038/s41590-025-02228-1

Download citation

Received: 12 April 2023
Accepted: 16 June 2025
Published: 29 July 2025
Version of record: 29 July 2025
Issue date: August 2025
DOI: https://doi.org/10.1038/s41590-025-02228-1

This article is cited by

IL-4–STAT6 signaling delays protective CD8+ T cell bystander activation by antagonizing IL-18 sensing
- Nicholas J. Maurice
- Talia S. Dalzell
- Stephen C. Jameson
Nature Immunology (2026)
Six degrees of TGFβ
- Quinn Peters
- Martin Prlic
Nature Immunology (2025)