Gamma-delta T cells seed tissues perinatally and can protect against murine poxvirus infection (22, 23). In the skin, Vγ5⁺ dendritic epidermal T cells (DETCs) and Vγ4⁺/Vγ6⁺ γδ T cells occupy the epidermis and dermis, respectively (24, 25). We first analyzed granzyme C expression in both populations using flow cytometry of single-cell suspensions of dissociated skin from wild-type C57BL/6 mice ( Figures 1A–D ). Prior to tissue harvest, we injected a CD45.2-specific antibody (Ab) intravenously (IV) to identify and exclude cells circulating in the vasculature from analysis, as previously described (26). We first gated on CD45.2 IV^-, CD45⁺, CD3⁺ lymphocytes and used Vγ5 TCR staining to identify TCR γδ⁺ DETCs (with dermal γδ T cells being Vγ5 TCR⁻) ( Figure 1A ). At steady-state in naïve, specific pathogen-free C57BL/6 mice, from 15 to 35% of DETCs expressed granzyme C ( Figure 1B ). Conversely, we did not observe significant frequencies or numbers of granzyme C-expressing dermal γδ T cells ( Figures 1B, C ). Likewise, DETCs expressed significantly higher levels of granzyme C per cell than the few positive dermal γδ T cells, determined by the mean fluorescent intensity (MFI) of intracellular granzyme C staining ( Figure 1D ). As tissue-resident cells can be perturbed by the enzymatic digestion needed to liberate them for flow cytometry (27), we also examined γδ T cells using confocal microscopy of frozen cross-sections of the ear skin ( Figure 1E ). Confocal images corroborated our flow cytometry findings, with a subset of DETCs expressing granzyme C in the epidermis. Interestingly, we often detected granzyme C localized to the dendritic cellular extensions between epidermal keratinocytes ( Figure 1E ). Thus, even naïve, specific-pathogen free mice have granzyme C-expressing skin-resident lymphocytes.

Gamma-delta T cells upregulate IFN-γ, granzymes A and B, and perforin during viral infection or in response to PAMPs (28–30). To understand the kinetics of granzyme C expression by DETCs and dermal γδ T cells during VACV infection, we first characterized the γδ T cell response in the skin. We infected C57BL/6 mice with VACV-SIINFEKL (expressing a minigene containing residues 257-264 of ovalbumin) in the ear pinna using a bifurcated needle as previously described (31–33). We used VACV as a viral infection model because this virus infects cells in both the epidermis and dermis where DETCs and dermal γδ T cells reside, respectively ( Figure S1 ). We analyzed the single-cell suspensions generated from VACV-infected ears at 0-, 1-, 3-, and 5- days post-infection (dpi) ( Figures S1C–G ). Flow cytometry results revealed that the frequency of DETCs and dermal γδ T cells amongst total leukocytes in the skin decreased significantly at 5 dpi ( Figures S1D–E ). Nevertheless, the total number of DETCs and dermal γδ T cells was significantly higher at 5 dpi ( Figures S1F–G ). Thus, both DETCs and dermal γδ T cells in the skin expand in number during VACV infection but constitute a smaller frequency of CD45⁺ CD3⁺ lymphocytes as new T cells are recruited into the skin.

We next analyzed granzyme C expression in both γδ T cell populations after infection ( Figures 2A–F ). We observed the highest average frequency of granzyme C⁺ DETCs on 5 dpi (~36% compared to ~27% in naïve tissue) ( Figure 2C ). In contrast, the highest frequency of granzyme C⁺ dermal γδ T cells was observed on 1 dpi (5.4 ± 0.81% compared to 3.05 ± 0.71% in naïve tissue) ( Figure 2D ). DETCs expressed more granzyme C per cell than dermal γδ T cells, with the highest granzyme C MFI on 5 dpi ( Figure 2E ). Dermal γδ T cells did not increase granzyme C expression levels by MFI during VACV infection ( Figure 2F ). These data show that epidermal DETCs are the only cutaneous γδ T cell population to upregulate granzyme C during local VACV infection.

We also analyzed the frequencies and numbers of endogenous TCR αβ CD8⁺ T cells that had entered the skin at the same timepoints post-VACV infection ( Figures S2A–F ). At 5 dpi a significant number of CD8⁺ T cells, defined as CD45.2 IV^-, CD45⁺, CD3⁺, Vγ5⁻, TCR γδ−, CD8β⁺, TCRβ⁺, were recruited to the infected skin ( Figures S2A–C ). Few CD8⁺ T cells expressed granzyme C at this timepoint and the frequency of expression was not increased by infection ( Figures S2D–F ). Together, these data show that epidermal DETCs specifically respond to cutaneous poxvirus infection with enhanced granzyme C production during the first 5 days post-infection.

We next analyzed CD8⁺ T_RM in the skin, another epidermal lymphocyte population with notable antiviral activity (34, 35). We infected mice with VACV-SIINFEKL and allowed for the endogenous polyclonal VACV-specific CD8⁺ T cell response to develop and T_RM to form (36). On 28 dpi or greater, we harvested skin, generated single-cell suspensions via enzymatic digestion, and analyzed the frequency of granzyme C expressing T cells using flow cytometry ( Figures 3A–F ). We first gated on CD45⁺, CD45.2 IV^-, CD8β⁺ cells. Both CD103 and CD69 are commonly used as tissue residency markers to identify T_RM in the skin (37). Therefore, we classified CD8⁺ T cells in the skin based on CD103 and CD69 expression and analyzed granzyme C expression in each population ( Figures 3B–F ). CD69⁺ CD103⁺ T_RM had the highest frequency of granzyme C expression at approximately 54% of the population. Cells that did not express either CD69 or CD103 had the lowest frequency of granzyme C expression at 11.8 ± 1.5%. We also analyzed granzyme C expression in CD62L⁺ CD44⁺ central memory T cells in the cervical lymph node and spleen. These non-tissue-resident memory CD8⁺ T cells scantly expressed granzyme C ( Figures S3A–C ). Furthermore, circulating CD62L⁺ CD44^- naïve, CD62L⁺ CD44⁺ central memory, and CD62L^- CD44⁺ effector memory CD8⁺ T cell subsets expressed little granzyme C ( Figures S3D–F ).

As before, we verified granzyme C expression using confocal imaging. For these experiments, we first transferred 1 x 10⁴ dsRed-expressing OT-I TCR transgenic CD8⁺ T cells (recognizing K_b-SIINFEKL) into Cd8a^-/- mice (deficient in CD8⁺ αβ T cells) to allow easy microscopic visualization of T_RM cells in the skin. In contrast to DETCs, we detected OT-I CD8⁺ T cells that expressed granzyme C in both the dermis and epidermis at 28 dpi ( Figure 3G ). We next quantified granzyme C expression in dermal and epidermal OT-I T_RM using confocal microscopy ( Figure S4A ). Both dermal and epidermal cells expressed similar levels of granzyme C based on the quantified intensity of granzyme C fluorescence per cell ( Figures S4B–C ).

To complement our confocal and flow cytometric analyses of granzyme C protein expression, we performed single-cell RNAseq on OT-I CD8⁺ T_RM isolated from the skin ( Figure S5 ). At the mRNA level, approximately 12% of Cd3e ⁺ cells also expressed detectable message for granzyme C. OT-I CD8⁺ T cells expressing Gzmc also co-expressed Cd69 and Itgae (CD103), consistent with our flow cytometry data. Other additional transcripts that were highly co-expressed with Gzmc included Gzmb, Il2rb, Ifng, and Prf1 (perforin).

Together these data show that granzyme C is expressed by resting CD8⁺ T_RM in mouse skin.

We next examined whether VACV-specific T_RM would, like DETCs, upregulate granzyme C during VACV infection, or whether this was a specific feature of γδ T cells. For these experiments, we continued analyses of CD103⁺ CD69⁺ VACV-specific CD8⁺ T_RM in the skin (37). On 28 dpi or greater, we reinfected the ear pinna of mice with the same VACV that was used for initial infection and analyzed T cells on day 2 post-reinfection ( Figures 4A–C ). Secondary infection increased both the frequency and expression level (MFI) of granzyme C in T_RM compared to T_RM from mice only infected once ( Figures 4D–F ). Secondary infection also increased the frequency and expression level of granzyme B- and IFN-γ -producing T_RM ( Figures 4G–J ). Together, these data demonstrate that CD8⁺ T_RM upregulate granzyme C (along with known effector molecules) during re-exposure to VACV.

Although our data thus far suggested that granzyme C can be upregulated in some antiviral T cells during viral infection, it was unclear whether this was a response to recognition of cognate Ag or inflammation induced by infection. To test whether Ag sensing in the tissue was needed for increased granzyme C expression, we again transferred 1 x 10⁴ naïve OT-I CD8⁺ T cells into Cd8a ^-/- mice. We then infected one ear with VACV-NP-S-eGFP (expressing a fusion protein consisting of the nucleoprotein from influenza virus, the SIINFEKL OT-I CD8⁺ T cell determinant, and eGFP) and the other ear with VACV-NP-eGFP (an identical virus that lacks SIINFEKL) as previously described (32) ( Figure 5A ). On 7-, 14-, 21-, and 28- dpi, we removed each ear (keeping them separate), created single-cell suspensions, and analyzed cells via flow cytometry ( Figures 5B–F ). The number of OT-I CD8⁺ T cells per ear were similar between ears infected with virus expressing or lacking cognate Ag on days 7 and 14 dpi ( Figure 5C ). On 21 and 28 dpi, we noted a significant increase in the number of OT-I CD8⁺ T cells in the ears infected with VACV-NP-S-eGFP compared to the ears infected with VACV lacking cognate Ag, consistent with previous reports (38) ( Figure 5C ). Granzyme C expression was detectable in OT-I CD8⁺ T cells in the skin by 7 dpi and occurred in T cells present in ears lacking cognate Ag expression ( Figures 5D–F ). Over time, granzyme C expression was maintained in ears lacking cognate Ag expression, although frequencies of granzyme C⁺ OT-I CD8⁺ T cells were slightly higher in ears containing cognate Ag. Accordingly, there were higher numbers of OT-I CD8⁺ T cells expressing granzyme C in the ears containing cognate Ag at 21 and 28 dpi ( Figure 5F ). These data show that CD8⁺ T cells do not require cognate Ag expression in the skin to produce granzyme C; however, cognate Ag may bolster expression.

Although cognate Ag was not needed for continued granzyme C expression, we next queried whether TCR engagement during secondary infection could upregulate granzyme C. As before, we transferred 1 x 10⁴ naïve OT-I CD8⁺ T cells into Cd8a ^-/- mice and infected both ears with VACV-NP-S-eGFP (containing cognate Ag) to establish OT-I CD8⁺ T_RM in both ears under the same conditions. Beyond 28 dpi, we reinfected one ear with VACV-NP-S-eGFP and the other ear with VACV-NP-eGFP (lacking cognate Ag) ( Figure 6A ). Flow cytometric analysis revealed no statistical difference in the frequency of granzyme B and C expressing CD8⁺ T_RM in the skin during secondary infection in the presence or absence of cognate Ag ( Figures 6B–E ). Granzyme A expression, however, trended toward increased expression in the absence of cognate Ag ( Figures 6B, C ). Thus, viral infection alone can drive the upregulation of granzyme C in CD8⁺ T_RM.

In the converse experiment, we assessed whether cognate Ag alone could drive upregulation of granzyme C (without virus-induced inflammation). After the establishment of OT-I CD8⁺ T_RM, we injected SIINFEKL peptide intravenously and harvested the ear pinna 6 hours post-injection ( Figure 6F ). Flow cytometric analyses revealed significant upregulation of granzyme C⁺ OT-I CD8⁺ T_RM after peptide injection ( Figures 6G–J ). The frequency of granzyme B⁺ T_RM also increased even more dramatically, while granzyme A remained relatively unchanged. Together, these data suggest granzyme C can be upregulated by both virally induced inflammation or TCR stimulation in the absence of other inflammatory stimuli. Furthermore, they reveal differential regulation of specific granzyme expression in response to different stimulation.

We next explored whether local cytokine changes could promote granzyme C expression in the absence of cognate Ag recognition. The cytokine IL-15 is induced during many acute viral infections and is important for tissue-resident lymphocyte maintenance (39, 40). In vitro, IL-15 upregulates granzyme C expression in isolated liver ILC1s (13). We therefore explored whether in vivo treatment with IL-15 alone could induce granzyme C expression. We first assessed the effect of IL-15 administration on DETCs. We intraperitoneally (IP) injected naïve wild-type C57BL/6 mice with exogenous IL-15 every 48 hr (three treatments) and harvested the ear pinna on day 7 ( Figure 7A ). Mice from the IL-15 treatment group had a higher number but not frequency of DETCs compared to the vehicle control ( Figures 7B–D ). DETCs in mice treated with IL-15 had an increased frequency of granzyme C expression and MFI compared to DETCs in mice receiving vehicle control ( Figures 7E, F ). IL-15 treatment also increased the expression level (MFI) and frequency of dermal γδ T cells expressing granzyme C, though this remained low compared to DETCs ( Figure S6 ). Confocal imaging also revealed numerous granzyme C⁺ DETCs in IL-15-treated mice ( Figure 7G ).

Having demonstrated exogenous IL-15 administration could upregulate granzyme C expression in skin-resident γδ T cells, we examined whether IL-15 would have the same effect in CD103⁺ CD69⁺ CD8⁺ T_RM. As before, we infected C57BL/6 mice with VACV-SIINFEKL and allowed for the endogenous polyclonal VACV-specific CD8⁺ T cell response to develop (36). On 28 dpi or greater, we administered IL-15 via IP injection as before, generated single-cell suspensions via enzymatic digestion from the ear pinna and analyzed the frequency of granzyme C-expressing T cells using flow cytometry ( Figures 7H–K ). Like the DETCs, IL-15 increased T_RM expression of granzyme C (in both frequency and MFI) compared to vehicle controls ( Figures 7I–K ). Thus, granzyme C expression is also cytokine responsive in vivo.

Specific pathogen-free C57BL/6N mice were obtained from Taconic Farms. dsRed (Stock Tg(CAG-DsRed*MST)1Nagy/J, #5441); Cd8a^-/- (B6.129S2-Cd8a^tm1Mak/J, #2665); and albino C57BL/6 (B6(Cg)-Tyr^c-2J /J, #58) mice were obtained from Jackson Laboratories. Rag1^-/- (Line 146); Rag2^-/-Il2rg^-/- (Line 111); T-betZsGreen (Line 8419); and OT-I TCR transgenic (C57BL/6NAi-[Tg]TCR OT-1-[KO]RAG1, Line 175) mice were obtained from the NIAID Intramural Research Repository at Taconic Farms. dsRed mice were crossed with OT-I TCR transgenic mice to create dsRed OT-I mice. Rag1^-/- mice were crossed with T-betZsGreen mice and bred to homozygosity to create Rag1^-/- T-betZsGreen mice. Cd8a^-/- mice were crossed with B6 albino mice and bred to homozygosity to create albino Cd8a^-/- mice. 6- to 20-week-old male and female mice were used in experiments. All mice were maintained on standard rodent chow and water supplied as necessary. All animal studies were approved by and performed in accordance with the Animal Care and Use Committee of NIAID.

Viruses used for this study included VACV-NP-S-eGFP (expressing a fusion protein consisting of influenza nucleoprotein, the SIINFEKL T cell determinant, and eGFP); VACV-NP-eGFP (an identical virus to VACV-NP-S-eGFP that lacks SIINFEKL); VACV-SIINFEKL (expressing residues 257-264 of ovalbumin). Recombinant VACV viruses were generated as TK^- viruses using the Western Reserve strain of VACV and have been previously described (32, 61).

Prior to blood collection mice were IV injected with 200 µls of saline containing brefeldin A solution (Biolegend) at a 1:100 dilution. Mice were immediately placed under isoflurane anesthesia and blood was collected through terminal retro-orbital eye bleeds. BioWhittaker Lymphocyte Separation Medium (LSM) (Lonza) was then used to isolate lymphocytes.

To distinguish IV⁺ cells, mice were injected with 3 µgs of pacific blue-conjugated CD45.2 (clone 104.2) intravenously as previously described 3 minutes prior to tissue isolation (63). Suspensions were filtered through 70 µm nylon cell strainers. Cells were stained with a combination of the following antibodies: CD45 (clone 30-F11), CD45.2 (clone 104), CD3 (clone 17A2), CD8β (clone H35-17.2 or YTS156.7.7), CD8α (clone 53-6.7), CD69 (clone H1.2F3), CD103 (clone 2E7), Vα2 (clone B20.1), TCRβ (clone H57-597), TCR γδ (clone GL3), Vγ5 (Tonegawa’s nomenclature (clone 536)), granzyme A (GzA-3G8.5), granzyme B (clone 16G6 or NGZB), granzyme C (clone SFC1D8), IFN-γ (clone XMG1.2), Armenian Hamster IgG Isotype Control (clone HTK888), CD62L (clone MEL-14), CD44 (clone IM7) and fixable viability dyes (Zombie Aqua) from Biolegend, eBiosciences, Invitrogen, or BD Biosciences diluted in PBS and brefeldin A solution (Biolegend) at a 1:1000 dilution. Cells were fixed with 3.2% paraformaldehyde for 15 minutes and intracellular staining was done using 0.5% saponin in Hanks Balanced Salt Solution (HBSS) + 0.1% Bovine Serum Albumin (BSA) for 1 hr at room temperature. dsRed⁺ cells were identified based upon fluorescent protein expression. Cells were analyzed on a Fortessa flow cytometer (BD Biosciences) or 5L 16UV-16V-14B-10YG-8R Aurora (Cytek) and resultant data analyzed using FlowJo software (Treestar).

Recombinant murine IL-15 (PeproTech) was reconstituted in water and diluted in sterile saline solution prior to intraperitoneal (IP) injection of 5 µg/mouse every 48 hr for a total of 3 treatments. Ears were harvested 7 days after the start of the first treatment. SIINFEKL peptide (GenScript) was reconstituted in dimethyl sulfoxide (DMSO) and diluted in sterile saline solution prior to one time IV injection of 200 µg of peptide/mouse. Ears were harvested 6 hours after peptide injection.

CD8⁺ T cells were purified from spleens and lymph nodes using an EasySep Mouse CD8⁺ T cell Isolation Kit (Negative Selection) according to the manufacturer’s instructions (Stem Cell Technologies). Cells were naïve (CD69^-) (antibody information clone: H1.2F3) and > 90% pure by flow cytometry prior to IV transfer. Unless otherwise noted, mice received a standard dose of 1 x 10⁴ OT-I CD8⁺ T cells prior to infection.

Ears were removed on the indicated dpi, fixed in periodate-lysine-paraformaldehyde (PLP) for 24 hr, and moved to 30% sucrose/PBS solution for 24 hr. Ears were embedded in optimal-cutting-temperature (OCT) medium (Electron Microscopy Sciences) in cross-section orientation and frozen in dry-ice-cooled 2-methylbutane. 16-µm sections were cut on a Leica cryostat (Leica Microsystems), blocked with HBSS, 0.1% BSA, 10% bovine and donkey serum, 0.05% Triton X. Tissues were stained with a combination of the following antibodies: purified granzyme C (clone SFC1D8), Alexa Fluor 647 AffiniPure Goat Anti-Armenian Hamster IgG (H+L) or Alexa Fluor^® 594 AffiniPure Goat Anti-Armenian Hamster IgG (H+L), purified Cytokeratin 6 (clone SP87), Alexa Fluor 647 AffiniPure Donkey Anti-Rabit IgG (H+L), conjugated Alexa Fluor 488 anti-mouse TCR γ/δ (clone GL3) antibody, and nuclei stained using DAPI from Biolegend, ThermoFisher, or Jackson ImmunoResearch. Antibodies were diluted in HBSS, 0.1% BSA, 10% bovine and donkey serum, 0.05% Triton X. Images were acquired on a Leica SP8 confocal microscope equipped with hybrid detectors or a Leica Stellaris 8.

Rag1^-/- T-betZsGreen mice received 1 x 10⁴ dsRed OT-I CD8⁺ T cells prior to infection with VACV-SIINFEKL. At the indicated time post infection, mice were injected with 3 µgs of pacific blue-conjugated CD45.2 (clone 104.2) IV as previously described 3 minutes prior to tissue isolation to distinguish IV⁺ cells (63). Ears were harvested, separated into dorsal and ventral sides, diced, and digested in RPMI containing 7.5% fetal bovine serum (FBS), collagenase I (Worthington), and DNAse (Worthington) for 1 hr at 37°C. Suspensions were filtered through 70 µm nylon cell strainers. Cells were stained with a combination of the following antibodies: CD45 (clone 30-F11), CD8α (clone 53-6.7) along with Zombie Aqua viability dye. Using a BD FACS Aria III Cell Sorter, OT-I CD8⁺ T cells and Tbet-ZsGreen group I ILCs were sorted as Viability Dye^- CD45.2 IV^- CD45⁺ dsRed⁺ CD8α⁺ or ZsGreen⁺ cells, respectively, into RPMI containing 10% FBS and HEPES (25 mM). Cells were centrifuged and resuspended in RPMI containing 10% FBS and HEPES (25 mM) at a 1000 cells/μl concentration.

Two technical replicates of the Rag1^-/- T-betZsGreen ear samples were collected for single-cell RNA sequencing using Chromium Next GEM Single Cell 3’ v3.1 standard kit (10X Genomics, Pleasanton, CA). Approximately 3000 cells were targeted in each single-cell preparation. For the preparation of the cDNA and sequencing library generation, libraries were prepared as described in the Chromium Next GEM Single Cell 3’ Reagent Kit v3.1 (Dual Index) user guide to produce barcoded cDNA and perform Illumina sequencing library preparation.

The Illumina library quality was assessed by TapeStation D1000 high sensitivity reagent kit (Agilent, Santa Clara, CA), and DNA concentration was measured by Qubit High Sensitivity reagent kit (Thermo Fisher, Waltham, MA). Samples were diluted for sequencing and pooled equimolarly according to Illumina sequencing protocol to a final concentration of approximately 650 pM. Sequencing was performed NextSeq2000 instrument using two P3 200 cycle kits (Illumina, San Diego, CA) to target approximately 50,000 reads per cell.

After sequencing, the FastQ files were submitted to Cell Ranger ‘count’ and ‘aggregate’ functions. Sample Rag1^-/- T-betZsGreen Ear_1 had 2,744 passing cells with 1,986 median genes per cell and Rag1^-/- T-betZsGreen Ear _2 had 1,910 passing cells with 2,180 median genes per cell. Sequencing saturation for both libraries were 64.8% and 76.8%, respectively. After all quality control of the replicates, the expression of granzyme C and other genes were analyzed in the Loupe browser (10X Genomics, Pleasanton, CA).

Significances were assessed using Prism software (GraphPad) using a Kruskal-Wallis Test (3 or more groups) or unpaired two-tail Mann-Whitney test (2 groups) as indicated in the figure legends. Exact P values are shown throughout, statistical significance was set at P ≤ 0.05.