RORγ Agonists Enhance the Sustained Antitumor Activity through Intrinsic Tc17 Cytotoxicity and Tc1 Recruitment

Retinoic acid receptor–related orphan receptor γ (RORγ) is a master transcription factor that controls the differentiation and maintenance of the CD4⁺ and CD8⁺ type 17 effector T-cell subsets, Th17 and Tc17, respectively (1). Type 17 T cells participate in regulation of bacterial, viral, and fungal infections; autoimmune disease; and antitumor immunity (2). Targeting the type 17 signature cytokine IL17 or the generation of type 17 T cells through IL23 has been clinically validated in several autoimmune diseases (3–5). On the other hand, type 17 T cells can mediate potent and durable tumor growth inhibition when transferred to tumor-bearing animals (6–8). Although the prognosis can be poor in some situations where IL17 induces angiogenesis, inflammation, and promotes the initiation and early growth of some tumors (9, 10), the presence of RORγ⁺ type 17 T cells or their hallmark cytokines such as IL17A and GM-CSF is associated with improved antitumor effects in many cancers (11–17). We have previously reported that only a fraction of human T cells coexpress both RORγ and IL17A, suggesting that RORγ and IL17A may play distinct roles in Tc17-mediated antitumor immunity (18).

The antitumor efficacy of type 17 T cells may be mediated by several mechanisms. These include the conversion of Tc17 to Tc1 effectors, the durability of Tc17 cells and formation of memory cells, direct cytotoxic activity, and the ability of type 17 T cells to recruit other immune cells (19). The plasticity of Tc17 cells to transform to cytolytic, IFNγ-secreting Tc1 cells or IFNγ⁺IL17⁺ effector cells has been reported in a number of in vitro and in vivo settings, including in tumor-bearing mice (20, 21). Thus, Tc17 cells serve as a reservoir for Tc1 cell development (22). Type 17 cells in cancer patients produce high concentrations of cytokines, such as IL17A and IL17F, GM-CSF, TNFα, IL2, and IFNγ (19), which contribute to the formation of a proimmune microenvironment. IL17A and IL17F bind to IL17 receptors on endothelial cells, stromal cells, and tumor cells (23) and induce the production of various chemokines such as CCL20, CXCL9, and CXCL10, which recruit monocytes, neutrophils, natural killer (NK) cells, and CD8⁺ T cells into the tumor microenvironment (24, 25).

Tumor antigen–specific type 17 cells eradicate murine tumors better than type 1–polarized clones (7, 8), and the antitumor effect is long-lasting via the enhanced formation of memory cells capable of responding to tumor cell rechallenge (26). However, in vitro type 17 cells had little or no cytotoxic activity (27–29). In contrast to the well-known killing mechanism for Tc1 cells, little is known for Tc17 cells. Type 17 cells express minimal granzyme B, FasL, or TRAIL (27, 28), but increase their expression of TRAIL and FasL after transfer into mice, which is important for their antiviral responses (27, 30). Although human type 17 cells can produce both IFNγ and IL17 in vivo, murine type 17 cells associated with tumor secrete minimal amount of IFNγ (31). Together, these observations challenge the idea of intrinsic cytotoxicity and a direct antitumor effect of type 17 cells.

Here, we showed that Tc17 cells lysed antigen-specific target cells directly. Tc1 cells utilized granzyme B and, to a less extent, TNF superfamily members such as Fas and TRAIL to lysis target cells. Tc17 cells, on the other hand, expressed high amounts of granzyme A and depended on perforin to kill target cells. Tc17 cells had lesser lytic activity compared with Tc1 at initial antigen encounter; however, they could survive longer and maintained their cytotoxicity over a longer time, particularly upon target cell rechallenge. In addition to cell-intrinsic effects, the presence of Tc17 cells increased the survival of cocultured Tc1 cells. In tumors, Tc17 cells infiltrated the tumor microenvironment earlier than Tc1 cells, and the presence of Tc17 cells correlated with the number of CD8⁺ cells recruited into the tumor. An RORγ agonist further enhanced Tc17 survival, increased their cytotoxicity in vitro and in vivo, and recruited host CD8⁺ T cells to the tumor, suggesting the utility of RORγ agonists as an anticancer therapy.

C57BL/6 and BALB/c mice were purchased from Taconic. OT-I (C57BL/6-Tg(TcraTcrb)1100Mjb/J), Thy1.1 (B6.PL-Thy1^a/CyJ), Prf1 KO (C57BL/6-Prf1^tm1Sdz/J), and RORγt KO (B6.129P2(Cg)-Rorc^tm2Litt/J) mice were purchased from The Jackson Laboratory. All animal experiments were conducted according to institutional animal care and safety guidelines and with IACUC approval at the University of Michigan (PRO00007524) and Vet Pathology and IND Services (VPI-001-02).

All chemicals were purchased from Sigma, Avanti Polar Lipids, Tocris (R&D Systems), or Enzo Life Sciences. RORγ agonists LYC-54143 were described (18) and synthesized by Lycera. E.G7-OVA and MC38 cells were purchased from ATCC, Mycoplasma-free and used less than passage 10. MC38 cells were stably transfected with full-length OVA (pGen-OVA, GenScript) to generate MC38.OVA.

Splenocytes from OT-I mice were activated for 4 to 5 days with OVA-derived peptide SIINFEKL (50 ng/mL). For polarizing to Tc17 cells, we added cytokines TGFβ (2.5 ng/mL) and IL6 (10 ng/mL), and for Tc1 cells we added IL12 (10 ng/mL). Alternatively, CD8⁺ T cells were purified from splenocytes from C57BL/6 or Prf1 knockout (KO) mice using an EasySep Mouse CD8⁺ T-cell isolation kit from STEMCELL Technologies, and were activated with plate-bound anti-CD3 (2.5 mg/mL), soluble CD28 (2.5 mg/mL), and differentiated into Tc17 cells with TGFβ (2.5 ng/mL) and IL6 (10 ng/mL), or Tc1 cells with IL12 (10 ng/mL).

Cell-surface expression of coinhibitory receptors was analyzed by flow cytometry. Effector T-cell differentiation was assessed by flow cytometer after 5-hour stimulation with PMA (100 ng/mL), ionomycin (1 mg/mL) in the presence of brefeldin A (eBioscience). Intracellular staining was performed using labeled antibodies after fixation and permeabilization Buffer Set (eBioscience). Antibodies used for flow cytometry were CD8 (PE, clone REA601; Miltenyi), CD73 (eFluor450, clone eBioTY/11.8; eBioscience), IL17a (FITC, clone TC11-18H10.1; BioLegend), IFNγ (PE, clone XMG1.2; eBioscience), Thy1.1 (eFluor450, clone HIS51; eBioscience), FasL (PE, clone MFL3; BioLegend), and TRAIL (APC, clone N2B2; eBioscience). Data were acquired using a BD FACSCanto II flow cytometer and analyzed by FACSDIVA software.

For qPCR analysis, RNA was isolated using an RNeasy mini kit (Qiagen), and mRNA expression was analyzed in StepOnePlus (Life Technologies) real-time PCR instrument using housekeeping gene cyclophilin as internal standards. RNA was first reverse transcribed into cDNA using a High Capacity RNA-to-cDNA kit (Applied Bioscience). qPCR was performed using Power SYBR Green PCR Master Mix (Applied Bioscience) and two-cycle amplification (95°C for 15 seconds and 60°C for 60 seconds) for 40 cycles followed by melting curve. The sequence of primers is listed in Supplementary Table S1.

Splenocytes from OT-I mice were activated with peptides and polarizing cytokines toward Tc1 or Tc17 cells for 3 days. Effector T cells were nucleofected with granzyme A siRNA oligos (Dharmacon) using a P4 Nucleofector Kit (Lonza) and the Nucleofector program CM137. Transfected cells were resuspended in their original conditioned media for 2 days before analysis.

Tc1 and Tc17 cells were differentiated in vitro as indicated above. E.G7-OVA cells and MC38.OVA target cells were labeled with CFSE (10 μmol/L, Thermo Fisher). Various numbers of Tc17 effector (E) cells were mixed with a fixed number of target (T) cells in a 96-round bottom plate, to achieve effector:target (E:T) ratios of 100:1, 30:1, 10:1, 3:1, and 1:1. Mixed cells were incubated for indicated times at 37°C with 5% CO₂. Specific lysis was calculated by % of specific lysis = 100×(1 − experimental (CFSE^low/(CFSE^high + CFSE^low)%)/input (CFSE^low/(CFSE^high + CFSE^low)%)) (32).

To examine the contribution of FasL/Fas and TRAIL/TRAILR2 pathways to the cytotoxicity of Tc1 and Tc17 cells, blocking antibodies for FasL (clone Kay-10, BioLegend), TRAIL (clone N2B2, BioLegend), or isotype control (clone RTK4530 and clone RTK2758, BioLegend) were added at the beginning of in vitro cytotoxicity assays.

To measure the cytotoxicity of Tc17 cells generated from wild-type (WT) or Prf1 KO mice, allogeneic splenocytes from BALB/c mice were incubated with LPS (10 μg/mL; ref. 33) or ConA (2 μg/mL; refs. 34, 35) for 2 days and used as target cells.

To measure the cytotoxicity of antigen-encountered Tc1, Tc17, or Tc1 cocultured with Tc17 cells, Tc1 cells were differentiated from OT-I splenocytes and Tc17 from Thy1.1 OT-I splenocytes. A combination of Thy1.1-positive/negative selection and CD8⁺ T cells selection (STEMCELL Technologies) was used to purify antigen-encountered Tc1 and Tc17 cells for recall response.

E.G7-OVA tumor cells (ATCC) were implanted subcutaneously into the flank of C57/BL6 mice and allowed to grow. In parallel, splenocytes from Thy1.1 OT-I mice were isolated and differentiated into Tc17 cells in vitro in the presence/absence of RORγ agonists for 5 days. Once the tumor was measurable (normally between days 7 and 10 after implant), the expanded T cells were washed out of RORγ agonists, and injected intraperitoneally. Antitumor responses were measured 2 to 3 × per week by caliper measurement. Tumor volume calculation was done as follows: 0.5×(length× (width)²). Mice were taken down after tumor volume reached ethical endpoint of 2,000 mm³.

In vitro experiments were repeated ≥ 3 × unless otherwise noted in figure legends. Most data presented in figures are mean ± standard deviation (SD) of biological replicates indicated in the figure legends. Statistical significance for in vitro data was determined using an unpaired, two-tailed t test (assume Gaussian distribution, with 95% confidence level). Statistics for in vivo data were done using multiple t tests, Mann–Whitney test (fewer assumptions, desired FDR (Q)–1%) in GraphPad Prism 7. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; and n.s., not significant.

Type 17 T cells are described as polyfunctional effector cells that can mediate potent antitumor activity in murine syngeneic models. However, in other models and in vitro, type 17 T cells have been shown to have modest, if any, cytolytic activity. These seemingly inconsistent data may be due to differences in the in vivo microenvironments or different polarization conditions used to generate type 17 cells in vitro. In the literature, a variety of different conditions have been used to generate type 17 cells. Here, we differentiated Tc17 from OT-I cells with a minimal cytokine cocktail of TGFβ and IL6 and compared them with Tc1 cells. We confirmed (18, 36) that Tc17 cells expressed more CD73 and IL17 and less IFNγ than Tc1 cells (Fig. 1A).

We measured the direct cytotoxicity of Tc17 effector cells generated from OT-I mice toward antigen-specific E.G7-OVA lymphoma cells in vitro. After 4 hours of coculture, Tc1 cells killed E.G7-OVA target cells in a dose-dependent manner. However, Tc17 cells were ineffective at any E:T ratio tested (Fig. 1B).

We then measured the cytotoxicity of Tc17 cells in vivo using an adoptive cell transfer (ACT) model. Consistent with the in vitro data, when equal numbers of effector T cells (0.2 × 10⁶ per mouse) were transferred into tumor-bearing mice, Tc1 cells decreased the tumor volume, whereas Tc17 cells were not effective (Fig. 1C). However, increasing the number of Tc17 cells (5 × 10⁶ per mouse) transferred displayed good antitumor activity (Fig. 1D), suggesting that these cells are inefficient cytotoxic T lymphocytes (CTL) but not intrinsically incapable of tumor lysis.

One key difference between in vitro CTL assays and in vivo tumor eradication is the time over which lytic activity is assessed. To address this, we measured in vitro target cell lysis at a later time point. When we incubate effector cells with target cells for 16 hours, Tc17 cells started to show substantial CTL activity (Fig. 1E) and in a dose-dependent manner when higher E:T ratio was utilized (Supplementary Fig. S1A), demonstrating that Tc17 cells have a slow onset lytic mechanism that might be different from Tc1 cells, and the timing used to evaluate Tc17 lytic activity, indeed, influenced their cytotoxicity.

Multiple mechanisms are used by cytotoxic lymphocytes to recognize and lyse target cells, including perforin/granzyme-mediated cleavage of specific intracellular substrates following antigen/MHC or NKG2D recognition, caspase-dependent apoptosis induced by TNF superfamily members such as Fas and TRAIL, as well as antibody-dependent cellular cytotoxicity and antibody-dependent cellular phagocytosis. The slow onset of Tc17 cell cytotoxicity in vitro prompted us to examine the involvement of death ligand/death receptor pathways. Compared with the exocytosis of perforin/granzyme granules, FasL and TRAIL are usually expressed on the cell surface at later time points upon CTL activation and sometimes involve a second wave of de novo synthesis (37, 38).

To test the contribution of FasL/Fas and TRAIL/TRAILR2 pathways to the cytotoxicity of Tc17 cells in vitro, blocking antibodies were used in a 16-hour CTL assay against MC38.OVA target cells which expressed both Fas and TRAILR2 on their surface. Tc17 cells expressed less FasL and TRAIL compared with Tc1 cells (Fig. 2A). We found both Tc1 and Tc17 cells lysed target cells (Fig. 2B), as indicated by fewer target cells remaining after incubation with either effector type (isotype vs. input). Blocking of FasL or TRAIL in Tc1, but not in Tc17 cells, slightly increased the number of remaining target cells (from 19296 ±1,285 to 24,022 ± 3,967 or 28,971 ± 1,501), indicating a minor involvement of FasL and TRAIL pathways for Tc1 cytotoxicity.

We then revisited the potential contributions of the perforin/granzyme pathway toward the Tc17 lytic activity. Compared with Tc1 cells, fewer perforin⁺ cells were found in Tc17 cells, and they expressed less perforin mRNA (Fig. 2C; Supplementary Fig. S1B). As reported in the literature (27, 28), we also observed a decreased percentage of granzyme B⁺ cells and decreased mRNA expression in Tc17 cells (Fig. 2C; Supplementary Fig. S1B). There are 11 different granzymes, and mice deficient in any single granzyme such as granzyme A, B, or M generally display mild cell killing deficiency, highlighting the functional redundancy within the family (39). In contrast, there is only one perforin, and perforin-deficient mice are severely immunodeficient with compromised ability to defend against viral infections and tumors (40). Thus, we utilized perforin-deficient mice to investigate whether the perforin/granzyme pathway is involved in the cytotoxicity of Tc17 cells. WT or perforin KO cells were differentiated into Tc1 or Tc17 cells and tested for their cytotoxicity toward allogeneic ConA- or LPS-activated splenocyte blasts (Fig. 2D; Supplementary Fig. S1C). Loss of perforin completely abolished the killing of Tc17 cells in vitro. These results demonstrated that although their perforin levels are low, this pathway is indispensable for the cytotoxicity of Tc17 cells. Dependence on the perforin/granzyme pathway for Tc17 CTL activity and the low expression of granzyme B suggest that other granzymes are important for Tc17 CTLs.

We next investigated which granzymes are expressed in Tc17 cells, and compared expression with Tc1 cells. The major differentially expressed granzyme mRNA in Tc17 compared with Tc1 cells is granzyme A (Fig. 3A), which was confirmed by assessing protein with flow cytometry (Supplementary Fig. S2A). Tc1 and Tc17 also express granzyme C, F, and G at comparable, albeit low levels (Fig. 3A).

To examine whether granzyme A is involved for Tc17 cytotoxicity, we used granzyme A siRNA to inhibit its production. Nucleofection with granzyme A siRNA did not change the viability or cytokine production profile of Tc17 or Tc1 cells (Supplementary Fig. S2B). We observed reduction of granzyme A mRNA (Supplementary Fig. S2C) and protein (Fig. 3B) in Tc17 (88% and 60% reduction of mRNA and protein, respectively), as well as Tc1 cells (94% and 86% reduction of mRNA and protein, respectively); and no alterations for granzyme B or other granzymes after granzyme A siRNA knockdown (Supplementary Fig. S2D). We then measured the lytic activity of the Tc1 and Tc17 cells with granzyme A knockdown toward target cells (Fig. 3C). Knockdown of granzyme A in Tc17 cells reduced their cytotoxicity, and the effect is greater when a lower E:T ratio of Tc17 cells was used in the CTL assay (22%, 28%, 44%, and 100% reduction at decreasing E:T ratios). There is no difference for Tc1 cells treated with granzyme A siRNA (Supplementary Fig. S2E). These data support the importance of granzyme A for Tc17 cytotoxicity and suggest it is redundant for Tc1 cytotoxicity.

During the course of these experiments, we noted that the number of effector cells remaining at the end of a 16-hour CTL was routinely higher for Tc17 than for Tc1 cells. To determine if this was the result of the coculture with antigen-expressing target cells or intrinsic to the effector T cells, we restimulated Tc1 and Tc17 effector cells with anti-CD3 for 72 hours and found that Tc1 have significantly reduced viability (Fig. 4A). This is consistent with the literature describing a differential sensitivity to activation-induced cell death between Th1 and Th17 cells (41, 42).

To assess how improved effector survival affects lytic activity, we measured kinetics of target cell killing during a prolonged, 72-hour CTL assay. As we observed before, Tc1 cells displayed higher CTL activity at early time points, as fewer target cells were left after 4 hours (40% less) or 24 hours (47% less) of culture with Tc1 cells compared with Tc17 cells (Fig. 4B, middle). However, a similar amount of target cells remained at 48 hours or 72 hours of culture (Fig. 4B, middle), indicating that an equivalent sum of total cytotoxicity was achieved with Tc1 and Tc17 cells over 72 hours. When we measured the number of effector cells during the prolonged CTL assay, we found Tc17 had improved viability over time compared with Tc1 cells (Fig. 4B, right). As a result, Tc17 cells maintained higher effector to target ratios at later time points (Supplementary Fig. S3A). These data demonstrate that improved effector survival of Tc17 cells compensates for their slow onset of lytic activity and contributes to the accumulated lysis of target cells over time and are consistent with the in vivo antitumor activity of Tc17 cells.

After synthesis and maturation, granzymes and perforin are packaged and stored within granules. Upon interacting with antigen-expressing target cells, CTLs release stored granzymes within minutes. To maintain their cytotoxicity in a target-rich environment, CTL cells need to replenish their intracellular pool for continuous granule exocytosis. Given the difference between Tc1 and Tc17 lytic activity over time, we examined the expression kinetics of the major granzymes for each subset. A larger fraction of Tc1 cells expressed granzyme B initially, but this decreased after 48 hours (Fig. 4C). Tc17 cells, on the other hand, maintained the same percentage of granzyme A⁺ cells for as long as we measured (Fig. 4C).

The prolonged cytotoxicity assay in Fig. 4B measured the cumulative effect of target killing over time. We next compared the ability of Tc1 and Tc17 CTLs to serially lyse target cells by determining whether Tc1 or Tc17 effector cells maintained their cytotoxicity after the initial antigen encounter. After 48-hour incubation with target cells, viable effector cells were purified and equal numbers of experienced CTLs or new effector cells were tested for their lytic activity with fresh target cells. Antigen-experienced Tc17 cells lysed target cells in the recall response at a rate comparable with the newly generated Tc17 effector cells, whereas Tc1 cells started to lose their cytotoxicity (20% less; Fig. 4D, middle). Tc17 effector cells had superior viability over Tc1 cells in both the initial antigen encounter (49% increase) and the recall response (309% increase; Supplementary Fig. S3B), and thus maintained a higher E:T ratio in both the initial antigen encounter (52% increase) and the recall response (445% increase; Fig. 4D, right), compared with Tc1 cells. These data indicate that Tc17 effectors have superior survival after antigen encounter and maintain their ability to lyse target cells in serial encounters, which supports their effectiveness in antitumor immune responses.

We first confirmed our previous findings (18) that the RORγ agonist LYC-54143 enhances the direct tumor killing activity of Tc17 cells in vitro and improves the survival of these effectors after encountering antigen-specific target cells. As expected, RORγ agonist added during Tc17 differentiation increased the percentage of IL17⁺ and RORγ⁺ cells and decreased CD73⁺ cells. RORγ agonist treatment increased the lytic activity for Tc17 cells (192% increase; Fig. 5A) and had minimal impact on Tc1 cells (21% increase). Analysis of the effector cells remaining after a 16-hour CTL assay (Fig. 5B) revealed agonist treatment during differentiation further improved the survival of Tc17 effector cells and augments the advantage of Tc17 CTLs over Tc1 cells.

We then investigated in perforin-deficient mice as to whether the perforin/granzyme pathway is involved in the efficacy of RORγ agonists on Tc17 cytotoxicity. RORγ agonists increased the cytotoxicity of WT Tc17 cells toward allogeneic ConA-activated splenocyte blasts as expected, but not in Prf1KO Tc17 cells (Supplementary Fig. S4), suggesting the efficacy of RORγ agonists on Tc17 cytotoxicity is dependent on the perforin/granzyme pathway.

The enhanced CTL activity observed after RORγ agonist treatment was assessed for relevance in an in vivo setting using adoptive transfer of equal numbers of Tc17 cells differentiated in vitro with or without LYC-54143 into E.G7-OVA tumor–bearing mice. We observed tumor shrinkage with Tc17 cells, and the agonist further increased their antitumor activity (Fig. 5C), confirming the physiologic relevance of our in vitro findings (Fig. 1D) and our previous in vivo data in a different adoptive transfer system (18).

In vivo, it is likely that mixed populations of Tc1 and Tc17 effectors coexist in tumor microenvironments and other tissues. After adoptive transfer into tumor-bearing animals, Tc17 cells infiltrated tumors as early as 2 days after transfer, but Tc1 cells were not detected (Fig. 6A, left), suggesting that Tc17 cells are better able to traffic to tumors, whereas Tc1 cells may need to be recruited. We observed a concurrent increased percentage of host CD8⁺ T cells in the tumor-infiltrating lymphocyte (TIL) population the after transfer of Tc17 cells (Fig. 6A, right). This recruitment of host cells by transferred Tc17 cells likely continued over the course of the study, as the percentage of donor cells and host CD8⁺ T cells 21 days after transfer correlated well (r² = 0.93; Fig. 6B), indicating Tc17 cells control the tumor growth not only through their direct cytotoxicity, but also through recruiting other effector cells to the tumor microenvironment. The inclusion of the RORγ agonist LYC-54143 during Tc17 differentiation resulted in the generation of Tc17 cells, which survive better in vivo and are more effective at recruiting host effector T cells to the tumor microenvironment (Fig. 6C).

Tc1 and Tc17 cells express chemokine receptors such as CCR5 or CCR6, which guide these cells to regions of high CCL3/4/5, CXCL9/10/11, or CCL20, respectively (6). Tc17 cells also express chemokines such as CCL3/4/5 and CXCL9/10 themselves, and blocking antibodies to individual chemokine partially decreased their chemotactic activity (6). In our hands, Tc17 cells expressed more chemokine receptors CCR6, CCR7, and CXCR3 compared with Tc1 cells, whereas Tc1 cells expressed more CCR5 (Supplementary Fig. S5A), consistent with literature findings. We also found Tc17 cells expressed high levels of CCL3/4/5, and low levels of CCL20/22, CXCL2/3/9/10. And the levels of CXCL9 and CXCL10 are significantly higher compared with Tc1 cells. RORγ agonists increased the expression of CCR6 and CCL3 in Tc17 cells (Supplementary Fig. S5B), which might contribute to the better migration of Tc17 cells into the tumor mass and recruitment of Tc1 cells.

To investigate whether the persistent Tc17 cells with their sustained lytic activity can benefit the surrounding Tc1 cells directly, we returned to our in vitro system and mixed Tc17 cells with Tc1 cells during the initial target encounter, and then purified Tc1 cells. Tc1 cocultured with Tc17 during the initial target encounter survived better, which improved the E:T ratio of Tc1 cells during the secondary challenge (Fig. 6D). These results suggest that Tc17 cells provide a transferrable factor(s) that enhances the serial killing ability of Tc1 cells.

Type 17 cells, including both Th17 and Tc17, can eradicate large, established murine tumors after adoptive transfer and provide robust protection against repeated tumor challenges (6–8). Despite these observations, little is known about the mechanisms by which Tc17 cells mediate antitumor immunity (43–51). Herein, we demonstrate that Tc17 cells can lyse target cells directly, utilizing four different types of target cells: E.G7-OVA, MC38.OVA or allogeneic ConA- or LPS-activated splenocyte blasts.

Among the nine granzymes we examined, Tc1 expressed more granzyme B, whereas Tc17 expressed more granzyme A. Knockdown of granzyme A specifically decreased the lytic activity of Tc17 cells, but not Tc1 cells, emphasizing the importance of granzyme A to Tc17 cell cytotoxicity. Tc17 cells also expressed granzyme B, C, F, and G at low levels, which might contribute to the residual cytotoxicity in granzyme A knockdown Tc17 cells.

The kinetics and degree to which the individual granzymes and perforin are expressed vary in different T-cell clones in vitro and in vivo, and depend on how they are activated (39). Granzyme B is expressed earlier than granzyme A during initial T-cell activation (52), but the percentage of granzyme B⁺ cells decreases after 3 days (53). Granzyme A and B induce independent cell death pathways, and cells that overexpress Bcl-2 family antiapoptotic proteins, caspase, or granzyme B protease inhibitors are still sensitive to granzyme A (54). FasL/Fas and granzyme B are involved in AICD of Tc1 and Th1 cells. The low levels of granzyme B, in addition to low FasL and high c-FLIP (41, 42), in Tc17 cells might contribute to their AICD resistance. Consistent with this, granzyme B–deficient mice clear both allogeneic and syngeneic tumor cell lines more efficiently than do WT mice (55). Extracellular granzyme A and granzyme K can also directly release proinflammatory cytokines from monocytes, macrophages, and fibroblasts (56). Thus, both Tc1 and Tc17 cells are CTLs with antitumor activities. Tc1 cells may be better suited for acute release of massive amounts of cytotoxic molecules to control initial tumor expansion with the caveat of collateral damage to nearby T cells and nontarget tissue, whereas Tc17 are better for more measured release of granular contents in a sustained way, and involve more effector cells for the antitumor immunity (57–59).

Once inside the tumor, Tc17 cells can induce chemokine production from tissue-resident cells (60,61) and recruit monocytes, neutrophils, NK cells, and CD8⁺ T cells into the tumor microenvironment (24, 25). Using RORγt KO mice, Nunez and colleagues found Th17 cells promoted the recruitment of Th1 cells to the tumor and contributed to antitumor immunity (62). Therefore, it is likely chemokines expressed directly by Tc17 cells and indirectly induced through IL17A/F cytokines, paired with chemokine receptors expressed on Tc1 cells, contribute to the recruitment of Tc1 cells. We observed a concurrent increase of host CD8 T cells in TILs 2 days after transfer of Tc17 cells, and the recruitment of host cells was sustained over time. In addition, our data showed that Tc17 could benefit the surrounding Tc1 cells directly, as the in vitro mixture of Tc1 and Tc17 cells during the initial target encounter improved the survival of Tc1 cells during the secondary challenge. Whether direct cell–cell interaction was absolutely needed or which soluble factor(s) mediated this transferrable beneficial effect is currently under investigation.

In summary, we have demonstrated that: (i) Tc17 can be CTLs, and the conditions under which they were differentiated and evaluated can influence their cytotoxicity, (ii) Tc17 CTLs depend on perforin and granzyme A for their lytic activity, (iii) Tc17 CTLs refill/maintain expression of key cytotoxic molecules and survive better, which allows them to kill more targets over time despite lower expression of cytotoxic molecules, (iv) Tc17 cells recruit Tc1 cells and can improve their ability to serially kill target cells. Our characterization of Tc17 cells adds to our understanding of how these cells participate in antitumor immune responses and provides a rationale for the discovery and development of synthetic RORγ agonist drugs which further potentiate Tc17 CTL activity, persistence, and antitumor efficacy. Currently, our RORγ agonist LYC-55716 is being tested in the clinic for locally advanced or metastatic cancer in a phase I/IIa trial (NCT03396497), and in combination with pembrolizumab for non–small cell lung cancer in a phase Ib trial (NCT02929862, www.clincialtrials.gov). Therapeutic strategies that enhance both Tc1 and Tc17 cells may improve the efficacy observed with current immunotherapies.

W. Zou has ownership interest (including stock, patents, etc.) in Lycera Corp. and is a consultant/advisory board member for the same. X. Hu is Executive Director of Lycera Corp. and has ownership interest (including stock, patents, etc.) in the same. L.L. Carter has ownership interest (including stock, patents, etc.) in Lycera Corp. No potential conflicts of interest were disclosed by the other authors.

Conception and design: X. Liu, W. Zou, X. Hu, L.L. Carter

Development of methodology: X. Liu, E.M. Zawidzka, H. Li

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Liu, E.M. Zawidzka, C.A. Lesch, J. Dunbar, D. Bousley

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Liu, E.M. Zawidzka, H. Li, L.L. Carter

Writing, review, and/or revision of the manuscript: X. Liu, E.M. Zawidzka, X. Hu, L.L. Carter

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Liu

Study supervision: X. Liu, X. Hu, L.L. Carter

We thank Peter L. Toogood, Bruce A. Goldsmith, H. Jeffrey Wilkins, and other ECL members for critically reading the manuscript and providing constructive suggestions for revision. We also thank Timothy J. Ley from Washington University and Xuefang Cao from the University of Maryland for the suggestions on the granzyme-deficient mice.

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.