Pmel-1 T-cell receptor (TCR) transgenic and C57BL/6J mice were purchased from The Jackson Laboratory.

The GL261 cell line was provided by Dr. Robert M. Prins (University of California, Los Angeles). GL261-Quad cells expressing human gp100_25–33 (hgp100_25-33), chicken OVA_257–264 and OVA_323–339, and mouse I-Eα_52–68 were generated7 and provided by Dr. John R. Ohlfest (University of Minnesota).

Detailed information on antibodies and relevant reagents is available in online supplemental table 1.

Pmel-1 CD8⁺CD44^- T-cells were sorted using CD8α magnetic MicroBeads system (Miltenyi Biotec) and then purified by MoFlo (Beckman Coulter). These cells were stimulated by mitomycin-C-treated splenocytes loaded with 2 µg/mL hgp100_25-33 peptide in the presence of the following cytokines and antibodies from days 0 to 3: Tc1 cells, 100 U/mL IL-2, 2 ng/mL IL-12 and 5 µg/mL anti-IL4 mAb; and Tc17 and Tc17-1 cells, 10 ng/mL TGF-β1, 10 ng/mL IL-6, 10 µg/mL anti-IFN-γ mAb and 5 µg/mL anti-IL4 mAb. After day 4, T-cells were cultured with 100 U/mL IL-2 for Tc1, 10 ng/mL IL-23 for Tc17, and 1 µg/mL IL-12 and 10 ng/mL IL-23 for Tc17-1 cell differentiation.

Intracellular cytokine staining Cytofix/Cytoperm Kit was purchased from BD. The data were analyzed using BD Accuri C6 flow cytometer and software.

Total RNA was extracted and subjected to quantitative real time-PCR analysis as described previously.8 Relative expression of mRNAs compared with control samples was calculated by the ddCt method.

In vitro cytotoxicity was conducted using a 6-hour ⁵¹Cr-release assay as described previously.9 In brief, Tc1, Tc17, and Tc17-1 cells were incubated with GL261 cells loaded with or without hgp100_25-33. In some experiments, Tc1 and Tc17-1 cells were sorted from cervical lymph nodes (CLNs) by MoFlo.

The amount of IFN-γ and tumor necrosis factor (TNF)-α in the supernatant was measured by BD OptEIA ELISA sets.

C57BL/6J mice were intracranially inoculated with GL261-Quad (1×10⁵) cells as described previously.8 Tumor-bearing mice received temozolomide (TMZ; 330 µg/mouse) intraperitoneally on day 9, then intravenously (IV) infusion of Tc1 or Tc17-1 (5×10⁶) cell on day 10 followed by intramuscular injections of polyinosinic-polycytidylic acid stabilized with poly-L-lysine carboxymethylcellulose (poly-ICLC) (2.5 mg/kg) with the peptide vaccine (100 µg/mouse) on day 11 after tumor inoculation. Mice were sacrificed when they showed any of the following signs: hunchback, seizures, hemiparesis, or weight loss of greater than 20%.

Log-rank test by GraphPad Prism software (V.8.4.3) was used to determine significant differences in the survival of mice on Kaplan-Meier plots among the groups. Mean values between the two groups were compared using Student’s t-test.

Using Pmel-1 mouse-derived CD8⁺CD44^- T-cells, we first generated IL-17- and IFN-γ-producing CD8⁺ T (Tc17-1) cells and compared their characteristics with those of IFN-γ-producing CD8⁺ T (Tc1) and IL-17-producing CD8⁺ T (Tc17) cells. Tc1 and Tc17 cells predominantly produced IFN-γ or IL-17, respectively, while Tc17-1 cells represent heterogeneous cell populations, with the majority of cells producing IL-17 and approximately 10% producing both IL-17 and IFN-γ (figure 1A).Consistent with flow cytometry data, Tc17-1 cells expressed lower levels of Ifng mRNA than Tc1 cells (figure 1B). However, Tbx21 mRNA (encoding T-bet) levels were comparable between Tc1 and Tc17-1 cells (figure 1B). On the other hand, mRNA levels of Il17a and Rorc (encoding RORγt) expressed in Tc17-1 cells were at least as high as those in Tc17 cells (figure 1C). Tc17-1 cells expressed CCR6 at higher levels than Tc1 but slightly lower than Tc17, while Tc17-1 and Tc17 cells expressed lower levels of CXCR3 than Tc1 cells (figure 1D and online supplemental figure 1), suggesting that the CCR6 and CXCR3 expression profiles in Tc17-1 cells were closer to those of Tc17 cells than to Tc1 cells.

To investigate the functional properties of Tc1, Tc17, and Tc17-1 cells, we stimulated them with GL261 glioma cells in the absence or presence of exogenous hgp100_25-33 peptide (figure 1E,F). Tc1 cells demonstrated the highest level of IFN-γ production and antigen-specific cytotoxicity compared with those by Tc17-1 and Tc17. While Tc17-1 cells exhibited low but significant antigen-specific IFN-γ production and cytotoxicity, Tc17 showed no cytotoxicity above background levels. On the other hand, Tc17-1 cells produced the highest levels of TNF-α among the three-cell types. Taken together, Tc17-1 cells showed transcriptional characters of both Tc1 and Tc17 cells, while their cytokine and chemokine receptor expression profiles, as well as cytotoxic activity, appear to be closer to those of Tc17 cells in vitro.

We next assessed the therapeutic efficacy of intravenously transferred Pmel-1-derived Tc17-1 in mice bearing GL261-Quad gliomas which express the Pmel-1 epitope hgp100_25-33.7 Because Tc17 cells showed very low cytotoxic activity compared with Tc17-1 cells (figure 1F), we compared the therapeutic effects of Tc17-1 cells with those of Tc1 but not Tc17 cells (figure 2A). We used TMZ because TMZ is a part of the standard-of-care for GBM therapy,10 and TMZ effectively induced lymphodepletion and expansion of intravenously infused T-cells (online supplemental figure 2). Contrary to our hypothesis, Tc17-1 cells did not significantly prolong the survival of mice compared with the control mice (p=0.1879). In contrast, Tc1 cells showed a modest but significant prolongation of survival compared with TMZ treatment alone (p=0.0180, median survival time: 57 days with TMZ, 70.5 days with TMZ +Tc1 and 60.5 days with TMZ +Tc17-1; n=10 mice/group).

To investigate possible mechanisms underlying the lack of therapeutic efficacy with Tc17-1 cells, 7 days after adoptive transfer of T-cells we isolated the transferred T-cells from glioma-draining CLNs11 and measured cytokine production and cytotoxicity (figure 2B,C). The vast majority of both Tc1 and Tc17-1 cells produced IFN-γ, while IL-17 production in Tc17-1 cells largely diminished (figure 2B). These observations are consistent with a previous report that IL-17-producing CD8⁺ T-cells convert to IFN-γ-producing phenotype in vivo.12 However, despite the robust IFN-γ production, the antigen-specific cytotoxicity of Tc17-1 cells remained lower than that of Tc1 cells (figure 2C).

Very late activation antigen (VLA)-4 expression on T-cells mediates T-cell homing into the CNS,11 13 and we have reported that poly-ICLC enhances VLA-4 expression and glioma-homing of vaccine-reactive T-cells, thereby extending the survival of GL261 glioma-bearing mice.14 We, therefore, evaluated the addition of poly-ICLC and the hgp100_25-33 peptide vaccine on adoptive transfer of Pmel-1 Tc1 or Tc17-1 cells in the GL261-Quad glioma model. While the combination of poly-ICLC and peptide vaccine did not prolong the survival, mice receiving Tc17-1 or Tc1 cells showed equally prolonged survival when they also received poly-ICLC and the peptide vaccine (median survival time: 57.5 days with TMZ, 61 days with TMZ+poly+vaccine, 92.5 days with TMZ+poly+vaccine+Tc1 and 91 days with TMZ+poly+vaccine+Tc17-1) (figure 3A). There was no significant difference in median survival time between mice receiving Tc17-1 or Tc1 cells (p=0.6751). We then evaluated the effects of poly-ICLC and the peptide vaccine on Pmel-1-derived (CD90.1⁺) T-cells. Poly-ICLC+vaccine increased the percentage of VLA-4⁺ and IFN-γ⁺ cells in mice that received Tc17-1 cells (figure 3B,C), although the frequency of VLA-4⁺ cells was lower in Tc17-1 than in Tc1 cells even after the poly-ICLC+vaccine (figure 3B). The percentage of VLA-4⁺ cells among Tc1 cells slightly decreased with the poly-ICLC+vaccine. In the spleen on day 3 following the T-cell transfer, Tc17-1 cells showed a trend for higher frequency among all CD8⁺ cells compared with Tc1 cells regardless of the additional treatment (figure 3D). The migration of transferred T-cells in brain tumor tissues was detectable at variable levels on day 25 (figure 3E) in all treatment groups but not on day 3 (data not shown).

While we observed a trend toward better persistence of Tc17-1 cells over Tc1 cells in the spleen (figure 3D), this character of Tc17-1 cells did not translate to improved therapeutic effects in mice bearing intracranial glioma (figures 2A and 3A). To address the underlying mechanisms of these observations, we further evaluated the phenotype and pharmacodynamics of adoptively transferred Tc17-1 and Tc1 cells. As shown in figure 4A, before infusion, the majority of both Tc1 and Tc17-1 cells were CD44⁺ including CD62L⁺ central memory and CD62L⁻ effector memory subsets. However, on day 7 following intravenously infusion, a majority of the Tc17-1 cells in CLNs acquired a CD44^-CD62L⁺ stem-like phenotype while Tc1 cells showed similar profiles as observed before the intravenously infusion (figure 4A). Furthermore, on day 50 following the intravenously transfer (figure 4B), more Tc17-1 cells were present in peripheral blood compared with Tc1 cells. In vitro, Tc17-1 cells showed higher levels of Rorc (figure 1C), Gata3, and Bcl2 (online supplemental figure 3) but lower CXCR3 (figure 1D) when compared with Tc1 cells. These are consistent with a stem-like phenotype as reported recently.15–17

Finally, to investigate mechanisms underlying the recurrence of the GL261-Quad tumors following the adoptive transfer therapy, we established cell lines from each of the recurrent GL261-Quad tumors in the following treatment arms: TMZ, poly-ICLC, and hgp100_25-33 peptide vaccine (referred as Combo); Combo and Tc1 cells; Combo and Tc17-1 cells. We used in vitro-cultured GL261 and GL261-Quad cells as negative and positive controls, respectively. Pmel-1 CD8⁺ T-cells produced IFN-γ in response to GL261-Quad-Combo, but significantly lower levels to GL261-Quad-Combo+Tc1 or GL261-Quad-Combo+Tc17-1 cells (figure 4C; left). The IFN-γ production was restored by exogenous hgp100 peptide (figure 4C; right), indicating that both Tc1 and Tc17-1 cells induced immune-editing in GL261-Quad cells in vivo.