Immune-Stimulatory Effects of Rapamycin Are Mediated by Stimulation of Antitumor γδ T Cells

mTOR inhibitors treat cancers, though primarily by inhibiting tumor mTOR signals (1). Interest in mTOR inhibitors was increased by the demonstration that rapamycin prolongs life in mice (2). Cancer prevention is proposed as a mechanism for rapamycin-mediated lifespan extension (3–5), also making it a candidate cancer prevention agent. We showed that rapamycin reduced benign papillomas and completely prevented squamous cell carcinomas in a carcinogen (dimethylbenz(a)anthracene, DMBA) and inflammation (12-O-tetradecanoylphorbol-13-acetate, TPA)-induced mouse skin cancer model (6). Although mTOR is increased in many cancer cells where it regulates cellular growth, metabolism, and other functions (3, 4, 7), mTOR also has significant immune effects that are little reported in cancer or longevity studies. For example, mTOR regulates Th1 cell function and differentiation, and CD8⁺ memory T cells that protect from cancer, and regulatory T cells and myeloid-derived suppressor cells that promote cancer, respectively (8–12). We showed that rapamycin significantly altered many immune cell populations in naïve C57BL/6J mice (13). As mTOR regulates numerous, essential immune functions, we postulated that rapamycin-mediated cancer prevention or treatment also included immune mechanisms. Here we tested immune effects of rapamycin in DMBA/TPA skin carcinogenesis where immune mediators driving or protecting against carcinogenesis are well-defined (14–17). We show that IFNγ and γδ T cells cooperatively facilitate rapamycin-mediated cancer prevention and treatment in DMBA/TPA carcinogenesis. Rapamycin-independent IFNγ promoted local γδ T-cell accumulation whose anticancer function was augmented by rapamycin. Rapamycin also augmented activation and cytotoxicity of human γδ T cells against human squamous cell carcinoma cells in vitro and in vivo, thus supporting rapamycin to boost the efficacy of γδ T-cell–based therapies now in clinical trials (18).

Mice [wild-type (WT), IFNγ^−/−, δ TCR^−/−, CXCR3^−/−, IFNγR1^−/−, Prf1^−/−] were 8- to 12-week old C57BL/6J females (BL6, The Jackson Laboratory). NSG mice were 12-week-old males (Jackson). Genetic identity was confirmed by tail snip PCR for all mice, plus flow cytometry for all mice except NSG. All knockout mice were backcrossed ≥10 times onto the C57BL/6 background, except CXCR3^−/− mice were backcrossed 8 times. All mice were maintained in specific pathogen-free conditions. Mice were given microencapsulated oral rapamycin (eRapa) or Eudragit control chow (Rapamycin Holdings) ad libitum 30 days before DMBA, and throughout study at the known cancer-preventing 14 parts/million rapamycin concentration (3, 4, 6). On the basis of our prior DMBA/TPA data (6), assuming a type I error rate of 0.05 and a two-sided test, the t test of the log-transformed time-to-tumor incidence/log-rank test requires 20 mice per group to achieve 80% power. All animal studies were performed using procedures approved by the University of Texas Health Science Center, San Antonio (UTHSCSA, San Antonio, TX) Animal Care and Use Program. Human subjects gave written, informed consent and their studies were approved by our Institutional Review Board.

The protocol was done as described previously (6). Mice were blindly assessed weekly for tumors, identified as skin lesions >1 mm persisting until study conclusion. Tumors were measured with Vernier calipers and areas calculated as length × width. We used established methods (14) to define papillomas versus carcinomas. For histologic confirmation, tumors were excised, formalin-fixed, paraffin-embedded, hematoxylin and eosin stained, and blindly evaluated by a pathologist. Some experiments employed short course TPA (6 applications) given after DMBA.

Dorsal skins with removed tumors were shaved, excised, cleaned of fat, and placed in PBS, then transferred dermal side down into complete RPMI medium (10% FBS, l-glutamine, penicillin/streptomycin, HEPES, MEM nonessential amino acids, sodium pyruvate) with 2.4 U/mL dispase II (Roche Life Science) for 2 hours, 37°C, with approximately 50 rpm shaking. The epidermal layer was then scraped off with a scalpel for analyses.

To generate single-cell suspensions, epidermal scrapings were resuspended in complete RPMI, vortexed, pipetted to disaggregate cells, and filtered (40-μm strainer). For antigen re-expression, cells were rested overnight at 37°C with 5% CO₂ as described previously (15). Before staining, cells were filtered as above and counted with a Vi-cell XR (Beckman Coulter). We used validated commercial reagents: LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (Life Technologies), anti-mouse CD3 (17A2), CD25 (PC61.5), Granzyme B (GzA-3G8.5), Lamp-1 (eBio1D4B), NK1.1 (PK136), IL17A (XMG1.2), IL22 (1H8PWSR; eBioscience), anti-mouse γδ TCR (GL3), CXCR3 (CXCR3-173), CD69 (H1.2F3), Vγ5 (536), Vγ4 (UC3-10A6), Vγ1 (2.11; Biolegend), anti-mouse CD45 (30-F11; BD Biosciences), on Becton-Dickinson LSR II hardware using FACSDiva software. For cytokine staining, cells were stimulated with Leukocyte Activation Cocktail with Golgiplug (BD Biosciences) for 5 hours, surface stained, fixed, and permeabilized with Foxp3/transcription factor buffer (eBioscience), and then stained.

Epidermal scraping lysates was generated as described previously (6). Mouse 26-plex kits (eBioscience) tested epidermal lysates as per manufacturer's protocol. Data were acquired and analyzed on a Luminex 200 (Millipore).

Skin tumors were minced with a razor blade and digested, 37°C with 5% CO₂ in RPMI plus 1 mg/mL collagenase IV (Sigma) and 300 μg/mL DNase (Sigma). Digests were filtered through a 100-μm cell strainer, washed with RPMI, and assessed using a Dead Cell Apoptosis Kit (Life Technologies). Apoptotic cells were Annexin V⁺ propidium iodide⁻.

Recipient WT BL6 and syngeneic IFNγR1^−/− mice were housed in a pie cage (Braintree Scientific, Inc.) and irradiated (10 Gy) using a Gamma Cell 40 137-Cesium Irradiator. WT or IFNγR1^−/− bone marrow cells (1 × 10⁷) were transferred intravenously into recipients, which recovered for two months. Reconstitution was confirmed by flow cytometry of peripheral blood using anti-mouse CD119 (IFNγR1, BD Biosciences) antibody. Bone marrow chimera mice were fed eRapa/Eudragit as above. At 13 weeks post-DMBA administration, tumor growth was delayed compared with nonirradiated mice as described previously (19), necessitating a second 100-μg DMBA application. Studies were terminated at 30 weeks post-DMBA administration.

Flow cytometry–sorted γδ T cells (2 × 10⁴) from skin draining lymph nodes (dLN; CD3⁺ γδTCR⁺) or epidermis (CD45⁺CD3⁺ γδTCR^mid or γδTCR^hi) of WT or Prf1^−/− mice on Eudragit control or eRapa for 2 weeks were injected in 10-μL PBS into tumors (20 weeks post-DMBA) of δ TCR^−/− mice, respectively, on Eudragit or eRapa. The same number of identically isolated cells were injected subcutaneously underneath the same tumor. Controls were noninjected tumors of the same size on the same mice.

γδ T cells were sorted from skin dLNs (CD3⁺ γδTCR⁺) or epidermis (CD45⁺CD3⁺ γδTCR⁺) of WT mice on Eudragit control or eRapa for 1 month using a Becton-Dickinson FACSAria II. αβ T cells were CD3⁺ γδ TCR⁻. Because of low cell yield, γδ T cells were pooled from 10 mice/group. Cells (1.5 × 10⁶) were lysed in RIPA buffer and 15 μg protein was run on a gel and transferred to polyvinylidene difluoride membranes as described previously (6). Blots were incubated overnight at 4°C with 1:1,000 diluted phospho- and/or total antibodies against indicated proteins (all Cell Signaling Technology) plus anti-mouse β-actin (Santa Cruz Biotechnology) and detected by enhanced chemiluminescence (Pierce). Band quantification was by ImageJ software (NIH, Bethesda, MD).

Blood mononuclear cells from peripheral venipuncture of 4 normal subjects and 2 patients with muscle-invasive bladder cancer were obtained by Ficoll density gradient centrifugation and human γδ T-cell culture was done as described previously (20). Cells were cultured in 6-well plates at 2 × 10⁶ cells/mL in complete RPMI, 15 μmol/L isopentenyl pyrophosphate triammonium salt (Sigma), and 100 U/mL human recombinant IL2 (Miltenyi Biotec) ± 0.5 or 0.01 nmol/L rapamycin. Fresh medium ± rapamycin was replaced every 2–3 days. On day 14, cultures were assessed by flow cytometry using anti-human CD3 (UCHT1), IFNγ (4S.B3; eBioscience), anti-human Vγ9 (B3), Vδ2 (B6), NKG2D (1D11), CD56 (NCD56), CD69 (FN50), Lamp-1 (H4A3), γδ TCR (B1), and CD25 (BC96) antibodies (Biolegend).

Authenticated (short tandem repeats method) human SCC4 squamous cell carcinoma cells (gift from Dr. Cara Gonzales, UTHSCSA, San Antonio, TX) were labeled with 1.2 μmol/L carboxyfluorescein succinimidyl ester (CFSE) and 2 × 10⁴ cells were incubated with 14-day–cultured human γδ T cells at indicated effector:target ratios in U-bottom plates. After 5 hours, cells were stained for viability with Ghost Dye UV 450 (Tonbo Biosciences) before acquiring data on a Becton-Dickinson LSR II. Dead SCC4 cells were positive for CFSE and viability dye. Background was wells with only target cells. For inhibition studies, 10 μg/mL anti-human γδ TCR (B1, Biolegend), anti-NKG2D (1D11, Biolegend), or both were incubated with γδ T cells for 30 minutes before coculturing with SCC4 cells, 25:1 ratio.

Using the classical Winn assay (21), human γδ T cells cultured 14 days in indicated rapamycin concentrations were comixed with SCC4 cells, 1:3 (γδ T-cell:SCC4) ratio in 100-μL PBS just before subcutaneous injection into NOD/SCID/IL2rγ–deficient mice without further treatments. Control tumors contained SCC4 cells only. Tumor volume was calculated by modified ellipsoid formula: (length × width²)/2.

Analyses were done using GraphPad Prism 5.00 with appropriate statistical tests as indicated.

We previously showed that eRapa potently prevented DMBA/TPA skin carcinogenesis in WT mice versus Eudragit (empty encapsulation) control (6). Microencapsulation improves drug stability and delivery versus nonencapsulated rapamycin (2). In cancer (22) and DMBA/TPA carcinogenesis specifically (17), IFNγ inhibits tumorigenesis. We fed BL6 IFNγ^−/− mice eRapa or Eudragit control before DMBA initiation and 24 weeks of TPA promotion. Unlike WT mice, eRapa did not reduce tumor multiplicity in IFNγ^−/− mice (Fig. 1A), suggesting IFNγ was required for eRapa protection from skin tumor development. However, eRapa reduced total tumor burden (Fig. 1B) and individual tumor area (Fig. 1C and D) similarly in WT and IFNγ^−/− mice, suggesting IFNγ was dispensable for eRapa-mediated tumor growth control. As most tumors are benign papillomas, we assessed protection from malignant degeneration into squamous cell carcinoma (SCC). eRapa completely protected WT mice from SCC consistent with our recent report (6), whereas 16.7% of IFNγ^−/− mice on eRapa developed cancers (Fig. 1E and F). Thus, IFNγ is required here for eRapa-mediated cancer prevention but not benign neoplasia prevention.

Systemic levels of IFNγ increased in mice treated with eRapa compared with controls, although no statistically significant differences were observed (Fig. 1G). These data and those showing that IFNγ is required for eRapa-mediated cancer prevention (Fig. 1E) suggest indirect IFNγ effects and/or a requirement for eRapa-independent IFNγ. We thus examined epidermal immune cell changes between WT and IFNγ^−/− mice on DMBA/TPA and found increased frequencies and numbers of potentially detrimental αβ T cells (Fig. 1H) in the absence of IFNγ.

Potentially protective natural killer cells (Fig. 1I) and γδ T-cell receptor (TCR)^hi T cells (Fig. 1J) were not modulated by IFNγ. Strikingly, γδ TCR^mid T-cell frequency and numbers were reduced 2-fold in IFNγ^−/− mice (Fig. 1K), consistent with IFNγ-mediated epidermal γδ TCR^mid T-cell recruitment. eRapa did not alter these results, consistent with indirect IFNγ effects in eRapa-mediated skin cancer protection.

T cells are critical mediators of antitumor immunity and cancer prevention (23), but play contrasting roles in DMBA/TPA carcinogenesis: αβ T cells generally are protumorigenic (14, 15), whereas γδ T cells are protective (14, 16). As we previously showed that epidermal αβ T cells numbers were unaltered by eRapa in this model (6) and now show that epidermal γδ TCR^mid T cells are decreased in IFNγ^−/− mice (Fig. 1K), we analyzed γδ T-cell subsets. Figure 2A shows γδTCR^mid/TCR^hi and αβ TCR⁺ flow cytometry gates with/without DMBA/TPA. We focused on γδ TCR^mid T cells as they required IFNγ for maximum epidermal infiltration (Fig. 1K) and were increased in epidermis with DMBA/TPA–induced inflammation (Fig. 2A and B). γδ TCR^mid T cells were significant IL22 and IL17A producers (Fig. 2C), consistent with reports showing their contributions to psoriatic skin inflammation (24) and antitumor immunity (25). Using defined γ TCR subsets (26), we identified γδ TCR^mid T cells as approximately 30% Vγ4⁺, <5% Vγ1⁺, <1% Vγ5⁺, but the majority were Vγ5⁻Vγ4⁻Vγ1⁻ (Fig. 2D), consistent with a distribution of γδ T cells of dermal origin important in cutaneous immune surveillance (27). In contrast, γδ TCR^hi T cells did not require IFNγ for maximal epidermal infiltration (Fig. 1J), and were relatively decreased with DMBA/TPA–induced inflammation (Fig. 2A and B). Essentially all γδ TCR^hi T cells were Vγ5⁺ (Fig. 2D), consistent with dendritic epidermal T cells (DETC). γδ T cells from skin dLNs of DMBA/TPA–exposed mice resembled γδ TCR^mid T cells: approximately 40% Vγ4⁺ and <1% Vγ5⁺ (Fig. 2E), with one notable difference being approximately 50% of skin dLN γδ T cells were Vγ1⁺. Potentially protumorigenic αβ T cells were increased in epidermis with DMBA/TPA (Fig. 2A and B and Supplementary Fig. S1A) as reported previously (15).

γδ T cells are in clinical trials as cancer therapy based on their cytolytic capacity among other features (18) that is enhanced by rapamycin in vitro (20). To test a role for γδ T cells in eRapa-mediated cancer prevention, BL6 δ TCR^−/− mice (lacking γδ T cells) were fed eRapa or Eudragit control and given DMBA/TPA as for above studies. Strikingly, eRapa was ineffective in reducing tumor multiplicity (Fig. 3A), total tumor burden (Fig. 3B), tumor area (Fig. 3C and D), and malignant degeneration (Fig. 3E and F) in γδ T-cell–deficient mice. Together with IFNγ effects, these results demonstrate for the first time that rapamycin uses immune mechanisms to prevent cancer in specific settings and discloses a previously unknown in vivo γδ T-cell effect of rapamycin. Furthermore, To test relative contributions of γδ T cells versus IFNγ in eRapa-mediated skin cancer prevention, we compared SCC prevalence in IFNγ^−/− mice versus δ TCR^−/− mice on eRapa and found them indistinguishable (16.7% vs. 25.9%, respectively, P = 0.50, χ²). Thus, γδ T-cell and IFNγ contributions to eRapa-mediated skin cancer prevention both appear important and neither predominates based on these data, further supporting their roles in cancer prevention here.

As IFNγ is important for T-cell CXCR3 expression and induction of CXCR3 ligands like CXCL10 (28), IFNγ could recruit epidermal antitumor T cells independent of eRapa. In support, reduced epidermal γδ TCR^mid T cells in DMBA/TPA–challenged IFNγ^−/− mice (Fig. 1K) coincided with 2-fold reduced CXCR3 expression on γδ TCR^mid T cells (Fig. 4A), which was not further altered by eRapa. The CXCR3 chemokine receptor is important for αβ T-cell trafficking to inflamed skin and specifically in TPA-induced inflammation (15). γδ TCR^mid T-cell epidermal trafficking in BL6 CXCR3^−/− mice in response to DMBA/TPA was significantly impaired (Fig. 4B). In contrast, γδ TCR^hi T-cell frequency was unaltered in the absence of CXCR3 (Supplementary Fig. S1B), indicating that γδ T cells subsets can differentially utilize CXCR3 to traffic to inflamed skin. Furthermore, only the Vγ5⁻Vγ4⁻Vγ1⁻ γδ TCR^mid subset used CXCR3 for maximal epidermal recruitment (Fig. 4C) as evidenced by their reduced frequency in the absence of CXCR3. In addition to reduced CXCR3-expressing γδ TCR^mid T cells, DMBA/TPA–challenged IFNγ^−/− mice had almost undetectable epidermal CXCL10 (Fig. 4D), irrespective of eRapa treatment. Other chemokines tested (e.g., MCP-1/2, MIP-1α/β, RANTES) were not reduced in IFNγ^−/− mice (Supplementary Fig. S2). Thus, eRapa-independent IFNγ promotes γδ T-cell migration to inflamed epidermis by inducing CXCR3 on γδ TCR^mid T cells and its ligand, CXCL10, from epidermal cells. Finally, DMBA/TPA–challenged IFNγ^−/− mice had reduced epidermal γδ TCR^mid T cells expressing activation (e.g., CD69, CD25), cytotoxic (e.g., granzyme B, lamp-1), and proinflammatory (e.g., IL17A, IL22) molecules (Fig. 4E), but eRapa did not further alter these. Thus, eRapa-independent IFNγ is needed for maximally activated and potentially cytotoxic epidermal γδ TCR^mid T cells.

To address cellular IFNγ targets in eRapa-mediated skin cancer prevention, bone marrow chimeras generated from WT and syngeneic IFNγR1^−/− mice (lacking IFNγ signaling) were given eRapa, initiated with DMBA, and then promoted with TPA for 30 weeks. WT or IFNγR1^−/− recipients of WT bone marrow had the greatest tumor multiplicity (Fig. 5A) and tumor burden (Fig. 5B), suggesting IFNγ signaling in hematopoietic (immune) cells was needed for maximum benign skin tumorigenesis. WT recipients of WT bone marrow experienced clear eRapa-mediated SCC protection versus IFNγR1^−/− recipients of WT or IFNγR1^−/− bone marrow although the difference was not statistically different versus WT recipients of IFNγR1^−/− bone marrow, nor was the total number of tumors (benign plus malignant; Fig. 5C and D). Thus, IFNγ signaling is needed in both hematopoietic and nonhematopoietic cells for optimal eRapa-mediated neoplasia protection. WT recipients of WT bone marrow had the highest epidermal γδ TCR^mid T-cell frequency (Fig. 5E), suggesting that IFNγ signaling in both hematopoietic and nonhematopoietic cells is needed for optimal γδ T-cell trafficking into inflamed skin. WT recipients of IFNγR1^−/− or WT bone marrow had similar epidermal CXCL10, whereas IFNγR^−/− recipients of either WT or IFNγR^−/− bone marrow had similar CXCL10 that was significantly lower than WT recipients (Fig. 5F) supporting the concept that IFNγ signals to nonhematopoietic cells are required for maximum CXCL10 production for this trafficking, although other factors also appear to be involved.

As γδ T cells are required for rapamycin-mediated cancer protection in this model (Fig. 3), we next assessed in vivo effects of eRapa on γδ T cells. eRapa did not alter epidermal γδ T-cell numbers in WT mice given DMBA/TPA (Fig. 1J and K), so we tested eRapa effects on γδ T-cell antitumor activity by isolating WT γδ T cells from skin dLNs of mice on Eudragit control or eRapa and injecting them into established tumors on δ TCR^−/− mice correspondingly treated with Eudragit control or eRapa (Fig. 6A). Two weeks later, only injected tumors on eRapa-fed δ TCR^−/− recipients underwent significant tumor regression versus noninjected tumors on the same mice (Fig. 6B–D). Injected tumors on control Eudragit-fed recipients did not regress, indicating eRapa was required for γδ T-cell antitumor activity. To identify regression mechanisms, we injected tumors of δ TCR^−/− mice with γδ T cells from skin dLN of BL6 Prf1^−/− mice lacking cytotoxic perforin. Unlike WT γδ T cells, Prf1^−/− γδ T cells did not regress tumors in eRapa-fed mice (Fig. 6E–G). Furthermore, WT γδ T-cell injection induced the greatest tumor apoptosis in eRapa-fed δ TCR^−/− mice (Fig. 6H), supporting the conclusion that eRapa boosts antitumor γδ T-cell cytotoxicity in vivo that contributes to eRapa-mediated cancer prevention.

Initially, we used γδ T cells from skin dLN instead of epidermis in these studies as greater numbers could be isolated and as they are <1% Vγ5⁺ (DETC) and approximately 40% Vγ4⁺ (Fig. 2E), similar to dermal γδ TCR^mid T cells infiltrating DMBA/TPA–inflamed epidermis. Lymphoid-derived Vγ4⁺ γδ T cells mediate perforin-dependent skin cancer protection (29), consistent with our findings (Fig. 6B–G). These data show that transferred skin dLN γδ T cells are similar to γδ T cells of dermal origin (i.e., γδ TCR^mid) and not DETC, and are consistent with γδ T cells of dermal origin as the important antitumor γδ T cells. In confirmation, injection of skin γδ TCR^mid (isolated from DMBA/TPA–treated epidermis) but not γδ TCR^hi T cells (DETC), mediated tumor regression in eRapa-fed δ TCR^−/− recipients (Fig. 6I).

Rapamycin inhibits mTORC1, but we previously showed that eRapa-mediated protection against DMBA/TPA carcinogenesis occurred in the absence of detectable mTORC1 suppression in skin or tumors (6). In agreement with that data, eRapa did not suppress mTORC1 signaling (e.g., p-rpS6, p-4E-BP1) in DMBA/TPA–inflamed epidermal or skin dLN γδ T cells (Supplementary Fig. S3A and S3B). In striking contrast, eRapa reduced mTORC1 signaling in conventional αβ T cells in epidermis of the same mice (Supplementary Fig. S3A and S3B).

γδ T cells kill a variety of cancers with little normal tissue damage (30), making γδ T-cell–based approaches promising cancer immunotherapies. To test rapamycin effects on human γδ T-cell–mediated antitumor activity, human peripheral blood γδ T cells were cultured in vitro with isopentenyl pyrophosphate, IL2 ± 0.01 nmol/L or 0.5 nmol/L rapamycin [approximating rapamycin levels in skin or serum of eRapa-fed mice, respectively, as we reported (6)]. After 14 days in culture, Vγ9⁺ T-cell frequency was increased >50-fold in all culture conditions. Approximately 70% of live cells were Vγ9⁺, with >98% also expressing Vδ2 (Fig. 7A) consistent with reports (20, 31). Rapamycin (0.5 nmol/L) notably increased activation and functional markers (e.g., CD69, Lamp-1, IFNγ), whereas 0.01 nmol/L induced an intermediate activation phenotype (Fig. 7A). In agreement with phenotypic data, γδ T cells cultured in 0.5 nmol/L rapamycin killed SCC4 human squamous cell carcinoma cells in vitro significantly better than nonrapamycin-treated γδ T cells (Fig. 7B), consistent with another report using different human cancer cells (20). In contrast, 0.01 nmol/L rapamycin did not augment SCC4 killing. γδ T cells recognize tumor-associated antigens via TCR or NKG2D receptors (32). Using blocking antibodies, we found that γδ T-cell killing of SCC4 cells occurred primarily through TCR and not NKG2D (Fig. 7C). To investigate rapamycin effects on γδ T-cell–mediated SCC4 cell killing in vivo, cultured γδ T cells were comixed 1:3 with SCC4 cells just before subcutaneous injection into NOD/SCID/IL2rγ-deficient (NSG) mice that received no further rapamycin. γδ T cells alone potently inhibited SCC4 growth; however, γδ T cells that were cultured with 0.5 nmol/L rapamycin mediated significantly improved tumor growth inhibition (Fig. 7D, top). At 24 days postinjection, tumors from NSG mice were digested to determine γδ T-cell recovery among the different conditions. The greatest number of γδ T cells was recovered from the 0.5 nmol/L rapamycin condition, suggesting that rapamycin improved tumor γδ T-cell survival in tumors (Fig. 7D, bottom). To test whether human γδ T cells can be distinguished by TCR expression, we assessed the mean fluorescence intensity of γδ TCR expression in cultured γδ T cells obtained from blood. In contrast to mouse skin γδ T cells, we could not distinguish subsets based on TCR expression (Fig. 7A) and culture in rapamycin did not alter this finding (Supplementary Fig. S4). These data further support the concept that mouse and human γδ T cells have distinct phenotypes, despite functional similarities.`

mTOR inhibition to treat cancer is proposed to work principally through inhibiting tumor-intrinsic mTOR, and is only modestly effective clinically (1). We hypothesized that the significant immune effects of mTOR inhibition could affect clinical outcomes and that understanding such effects would provide critical insights into using mTOR inhibitors more effectively in cancer prevention or treatment and identify relevant clinical settings.

We focused on cancer prevention based on data showing that cancer prevention using mTOR inhibition with eRapa in distinct tumors is potentially more efficacious than mTOR inhibition as cancer treatment (3, 4, 6). Nonencapsulated rapamycin also prevents cancer in various models (33, 34). In this DMBA/TPA skin carcinogenesis model, endogenous immunity affected eRapa-mediated cancer prevention by two complementary mechanisms. First, eRapa-independent IFNγ was required for maximum γδ T-cell recruitment to epidermis, even though eRapa (insignificantly) increased serum IFNγ. These IFNγ signals appeared to act on γδ T cells to promote their trafficking, likely by promoting their CXCR3 expression, and acted on nonhematopoietic cells to generate CXCL10 mediating CXCR3-dependent trafficking. As some γδ T cells still traffic to inflamed epidermis in the absence of IFNγ, there likely are other unidentified γδ T-cell recruitment factors. Nevertheless, we make the novel observation that specific host immune attributes (e.g., sufficient IFNγ) modulate optimal mTOR inhibitor efficacy, an understanding that can now be applied to clinical mTOR inhibition.

Second, eRapa augmented γδ T-cell antitumor effects in vivo, which has not been previously reported. Dendritic epidermal T cells (DETC, identified here as γδ TCR^hi and Vγ5⁺), are the primary γδ T-cell population in naïve mouse epidermis. Because of their interaction with Langerhans cells and major production of IFNγ and granzyme, they are important in cutaneous surveillance against infections (32) and against the mouse PDV squamous cell carcinoma cell line (derived from DMBA-treated primary keratinocytes) in vitro (16). We found that γδ TCR^mid T cells in this model are significant IL17A producers, consistent with known γδ T cells of dermal origin that are important mediators of antitumor immunity against MCA205 mouse sarcomas (25). By injecting γδ T cells into established tumors, we now demonstrate that neither untreated DETC (γδ TCR^hi) nor γδ TCR^mid T cells alone can regress DMBA/TPA tumors in vivo, but eRapa induces effective γδ TCR^mid T-cell–mediated tumor regression in vivo. Thus, in vitro cytotoxicity of γδ TCR^hi cells against a skin cancer cell line (16) might not fully reflect in vivo effects. The γδ TCR^mid T cells recruited to DMBA/TPA–inflamed mouse epidermis were largely Vγ4⁺ or Vγ5⁻Vγ4⁻Vγ1⁻. Vγ4⁺ γδ T cells have been described as trafficking to inflammation via CCR2 (35) or CCR6 (24). We now show that the Vγ5⁻Vγ4⁻Vγ1⁻ population requires CXCR3 to traffic maximally to inflamed epidermis in this model where they can mediate antitumor responses. Winkler and colleagues (15) found that DMBA/TPA tumorigenesis is reduced in BL6 CXCR3^−/− mice not on rapamycin through reduced trafficking of protumorigenic epidermal ab T cells. Our data show that CXCR3 on γδ TCR^mid T cell is possibly protective against DMBA/TPA skin cancer. A recent report emphasized the role of CXCR3 in attracting conventional antitumor T cells to the skin cancer, melanoma (36). Thus, additional studies of the CXCR3 axis in endogenous and rapamycin-facilitated cancer protection are warranted.

Previous work in FVB mice suggests that γδ T cells are critical for protection against DMBA/TPA carcinogenesis (14) and that FVB δ TCR^−/− mice are more susceptible to SCC versus WT. However, we found similar DMBA/TPA carcinogenesis in untreated WT BL6 versus BL6 δ TCR^−/− mice. Thus, in BL6 mice, γδ T cells are not intrinsically protective, but become so with rapamycin treatment, suggesting that rapamycin can prevent cancers where intrinsic γδ T-cell–dependent protection is lacking, expanding the potential clinical utility of this approach.

Small-molecule mTOR inhibitor effects on specific mTOR signaling in cancer prevention and treatment remain incompletely understood. We found that mTORC1 was not suppressed in epidermal or dLN γδ T cells of eRapa-fed mice in skin carcinogenesis although local αβ T cells exhibited mTORC1 suppression. Thus, (unsurprisingly) small-molecule mTOR inhibitor effects will differ in distinct immune cells, as we recently reported (13). Boosting eRapa-mediated γδ T-cell cytotoxicity could be mTORC1-independent. For example, the rapamycin–FKBP12 complex could be influencing calcium flux through ryanodine receptors on γδ T cells to boost function (37). Alternative possibilities include: (i) an mTORC1 component not assessed here could mediate protection, (ii) mTOR effects in acute versus chronic suppression could differ, (iii) mTOR inhibition in another cell population mediates observed γδ T-cell effects, or (iv) mTORC2 effects might predominate, as we found in whole skin lysates and papillomas of DMBA/TPA mice on chronic eRapa (6). Further work on mTOR signaling is warranted, which could help identify additional druggable targets or better available drugs or combinations for cancer treatment or prevention.

In mice, >90% of epidermal T cells are γδ T cells, unlike in humans, whose epidermal T cells are 18%–29% γδ T cells (38). Nonetheless, kidney transplant recipients with skin cancers experienced tumor regression when switched to the mTOR inhibitor everolimus after failing other transplant rejection treatments (39). These data support the concept that mTOR inhibitors treat cancer in human tissues where γδ T-cell populations are less abundant than in mice. The human blood Vγ9⁺Vδ2⁺ T cells that we used to test antitumor effects are primarily IFNγ producing and thus not equivalent to mouse dermal γδ TCR^mid T cells characterized here. In further contrast, human γδ T-cell subsets could not be distinguished by TCR expression in culture with or without rapamycin and from normal or cancer subjects. Nonetheless, we show that IFNγ and γδ T cells appear to contribute distinctly to cancer prevention, and the antitumor properties of human γδ T cells could be enhanced by rapamycin in vitro and in vivo. Thus, further work to understand antitumor effects of rapamycin or rapalogs on human γδ T cells is warranted.

As cancer prevention is generally more cost effective, safer, and tolerable than treatment, a potentially broad-spectrum pharmacologic approach to cancer prevention would be extremely useful. Nonetheless, legitimate safety concerns, particularly for immunosuppression (37), dictate prudence in moving mTOR inhibitors into cancer prevention trials. Our demonstration of significant intestinal neoplasia protection and survival extension in eRapa-fed Apc^Min/+ mice (3), a model for familial adenomatous polyposis in which colon cancer risk is high, suggests that a cancer prevention trial in such a high-risk population could be justified.

As mTOR inhibition improves antigen-specific immunity (40) including in aged humans (41), and immune memory (11), combining rapamycin with antigen-specific approaches could be useful in cancer. For example, combining rapamycin with the immune checkpoint inhibitor, anti-CTLA-4, and a vaccine further boosted the immune response against EL4 lymphoma challenge (42). Everolimus in humans and rapamycin in mouse cancer models improved conventional antitumor T-cell responses (12). These studies establish that rapalogs enhance anticancer immunity including in humans, mediated largely by antigen-specific immunity. Our studies further show that rapamycin also enhances effective MHC-independent γδ T-cell–mediated antitumor immunity. Thus, combining rapalogs to boost human leukocyte antigen–unrestricted γδ T cells functions, such as with FDA-approved bisphosphonates (43), plus antigen-specific approaches could be effective cancer therapies.

Rapamycin is our proof-of-concept agent for cancer prevention, but ultimately might not be the best agent for specific scenarios. Recent work continues to show safety of mTOR inhibitors in promoting beneficial immunity in mice (40) and in elderly humans (41), where influenza immunization responses were improved by rapamycin or everolimus, respectively. In addition, a recent trial testing long-term eRapa in the marmoset monkey found it to be well tolerated with no serious adverse effects (44). Thus, mTOR inhibition alone or in rational combinations with agents boosting γδ T-cell functions or antigen-specific immunity deserve more study in cancer treatments and prevention.

No potential conflicts of interest were disclosed.

Conception and design: V. Dao, T.J. Curiel

Development of methodology: V. Dao, S. Pandeswara, T.J. Curiel

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V. Dao, Y. Liu, R.S. Svatek, A. Liu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V. Dao, R.S. Svatek, J.A.L. Gelfond, T.J. Curiel

Writing, review, and/or revision of the manuscript: V. Dao, J.A.L. Gelfond, V. Hurez, T.J. Curiel

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V. Dao, S. Pandeswara, A. Liu, T.J. Curiel

Study supervision: T.J. Curiel

We thank S. Lao for breeding and genotyping mice and C. Gonzales for providing the human SCC4 cell line.

This work was supported by Jess Hay Chancellor's Fellowship (AG038048, TL1TR001119, F30CA180377; to V. Dao) and grants from The Holly Beach Public Library, The Owens Foundation, The Barker Foundation, and The Skinner endowment (CA170491, to T. Curiel; CA54174, to J. Gelfond and T. Curiel).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.