Seven- to eight-week old female congenic C57BL/6 mice with allele Ly5.1+ (T cell donor mice) or allele Ly5.2+ (T cell recipient mice) were purchased from Taconic (Hudson, NY) and maintained in sterile housing. All mouse work was reviewed and approved by the University of Wisconsin and William S. Middleton Memorial VA Hospital and VA Ann Arbor Healthcare Animal Care and Use Committees.

Viral infections and tumor inoculations were performed per our previously published protocols [7, 9, 10]. In brief, Ly5.1+/C57BL/6 mice were infected with 2 × 10⁵ PFU of Armstrong strain lymphocytic choriomeningitis virus (LCMV; gift from Marulasiddappa Suresh, University of Wisconsin) via intraperitoneal injection to generate CD8+ T cells specific for GP33, a class I MHC-restricted LCMV surface glycoprotein. Splenocytes were harvested 8 days after infection (at which time LCMV peptide-specific CD8+ T cells are of effector differentiation) to obtain bona fide T_eff (as evidenced by CD127^low/KLRG1^high expression) and 50–80 days after infection (at which time LCMV peptide-specific CD8+ T cells are of memory differentiation) to obtain bona fide T_mem (as evidenced by CD127^high/KLRG1^low expression) [7, 8]. The yield of GP33-specific memory CD8+ T cells was increased by adoptively transferring Ly5.1+/C57BL/6 mice with 10³ Ly5.1+/CD8+ T cells derived from splenocytes of Ly5.1+/P14 TCR transgenic mice (C57BL/6 background mice with CD8+ T cell specificity for GP33 [7]. The spontaneous mouse melanoma cell line B16F10 was transfected with a plasmid encoding GP33, and poorly immunogenic clones with low levels of GP33 expression were selected to create a new B16GP33 cell line, which we have previously demonstrated to be of comparably low immunogenicity as parental B16F10 [9, 10]. This line was cultured in RPMI-1640 medium (Mediatech, Herndon, VA) with 10% fetal bovine serum, 2 mM L-glutamine, 100 μg/mL streptomycin, and 100 U/mL penicillin (Life Technologies, Inc., Grand Island, NY). Ly5.2+/C57BL/6 mice received subcutaneous injections of 1 × 10⁶ B16GP33 cells suspended in serum-free RPMI-1640 to generate flank melanoma tumors.

Adoptive cell transfer was performed per our previously published protocol [7, 8, 10]. As previously observed, GP33-specific CD8+ T cells typically comprised approximately 8–10% of splenocytes harvested from effector and memory T cell donor mice [7, 8]. After obtaining splenocytes from mice infected with LCMV, magnetic bead separation columns (Miltenyi, Auburn, CA) were used to isolate CD8+ T cells. Flow cytometry was performed to quantify the percentage of tumor-specific GP33-specific CD8+ T cells. 10⁵ GP33-specific CD8+ T_eff, T_mem, or 5 × 10⁴ T_eff + 5 × 10⁴ T_mem were adoptively transferred via retro-orbital intravenous injections into B16F10- or B16GP33 melanoma-bearing mice at specific time points after tumor inoculation.

For multiple timepoints, a total number of 2 × 10⁴ GP33-specific CD8+ T_eff and T_mem or both were separately co-cultured with 2 × 10⁴ B16GP33 melanoma cells at an overall effector:target ratio of 1:1 in 48-well plates in the presence of 10 ng/mL recombinant human IL-2 (BD Biosciences, San Jose, CA) for MTT assays as previously described [7]. T cells were characterized and isolated using the methodology described for adoptive cell transfer. Viable B16GP33 melanoma cells were quantified using the Celltiter 96 non-radioactive MTT assay kit (Promega, Madison, WI) following the manufacturer’s instructions.

For 5 h, GP33-specific CD8+ T_eff and T_mem were cultured separately or together in RPMI-1640 with 10 ng/mL IL-2 and 0.1 μg/mL GP33 peptide (gift from Marulasiddappa Suresh, University of Wisconsin). Conditioned media was collected by centrifuging wells and collecting the supernatant. For 12 h, GP33-specific CD8+ T_eff and T_mem were then co-cultured with B16GP33 melanoma cells with or without conditioned media and anti-IL-2 (BD Biosciences, San Jose, CA) for MTT assays as described above.

Flow cytometric analysis was performed per our previously published protocol [7–10]. Once harvested, tumor samples were processed into single-cell suspensions with mechanical fractionation on a fine wire mesh followed by lymphocyte isolation using Histopaque (Sigma Aldrich, St. Louis, MO). Splenocytes before ACT and tumor-infiltrating lymphocytes (TIL) 4 to 14 days after ACT were stained with APC-labeled MHC class I tetramers loaded with GP33 peptide (NIH Tetramer Facility Core, Atlanta, GA), PerCP-labeled anti-CD8 mAb, PE-labeled anti-Ly5.1 mAb, and FITC-labeled anti-CD44 mAb (BD Biosciences, San Jose, CA). Intracellular staining required stimulation of lymphocytes with media alone or with GP33 peptide (1 μg/mL), brefeldin A, and recombinant human IL-2 (10 ng/mL) at 37°C for 5 h. A BD LSRFortessa flow cytometer (BD Biosciences, San Jose, CA) was used and data were analyzed using FlowJo software (Tree Star Inc., Ashland, OR).

Experimental data was analyzed using IBM SPSS statistical software version 23 (Armonk, NY). T-test was used to compare groups and a repeated-measures analysis of variance (ANOVA) with pair-wise comparisons performed using Fisher’s protected least significant difference tests were used to compare multiple groups. Significance was defined as p < 0.05, and error bars represent standard errors of the mean.

To verify the peptide specificity of our ACT model, C57BL/6 mice were treated with 10⁵ GP33-specific T_eff one day after subcutaneous inoculation with B16GP33 melanoma or parental B16F10 melanoma (which does not express GP33). As shown in Fig. 1a-b, early ACT strongly suppressed the growth of B16GP33 melanoma but not B16F10 melanoma, suggesting that the effect of ACT was peptide-specific. Moreover, the kinetics of B16GP33 tumor growth in immunocompetent mice was comparable to those of parental B16F10 tumor. To compare the efficacy of ACT with different T cell subsets, Ly5.2+/C57BL/6 mice received ACT of 10⁵ GP33-specific CD8+ T_eff, T_mem, or T_eff + T_mem on day 7 after subcutaneous inoculation with 10⁶ B16GP33 cells. As shown in Fig. 1c-d, memory ACT resulted in significantly greater inhibition of tumor growth compared to effector ACT as previously described.⁷ Combinatorial ACT with T_eff + T_mem had significantly greater inhibition of tumor growth compared to ACT with T_eff or T_mem separately, making this combination the most effective strategy for ACT.Fig1Fig. 1

ACT with a combination of T_eff + T_mem resulted in optimal control of melanoma growth. Mice bearing B16GP33 (a) or B16F10 (b) melanoma were untreated (control) or treated with ACT of 10⁵ GP33-specific CD8+ T cells one day after tumor inoculation. Serial tumor measurements confirmed that ACT resulted in a GP33-specific inhibition of melanoma tumor growth. (c) Mice bearing B16GP33 melanoma were treated 7 days after tumor inoculation with no treatment (control) or 10⁵ T_eff, 10⁵ T_mem, or 5 × 10⁴ T_eff + 5 × 10⁴ T_mem. Comparison of treated groups (d) demonstrates optimal control of melanoma growth with combinatorial ACT. This experiment was repeated three times with similar results, 4–5 mice per group. (+ p < 0.05 compared with control; * p < 0.05 compared with effector ACT; Ɨ p < 0.05 compared with memory ACT)

To determine if the efficacy of combinatorial ACT was associated with stronger intratumoral T cell infiltration, mice were euthanized 4 and 14 days after ACT to allow for tumor harvesting and flow cytometric analysis of TIL populations. Representative data shown in Fig. 2a demonstrate that CD8+ T cells comprised a higher percentage of TIL in mice treated with ACT in comparison with untreated controls. As shown in Fig. 2b, total numbers of intratumoral CD8+ were transiently elevated on day 4 in mice treated with effector ACT, but this effect was short-lived; in contrast, intratumoral CD8+ T cells remain elevated on day 14 in mice treated with memory and combinatorial ACT.Fig2Fig. 2

Combinatorial ACT results in a stronger intratumoral T cell response. Seven days after tumor inoculation, Ly5.2+ mice were treated with 10⁵ T_eff, 10⁵ T_mem, or 5 × 10⁴ T_eff + 5 × 10⁴ T_mem derived from Ly5.1+ mice. TIL were harvested from melanoma tumors resected 4 or 14 days after ACT and stained for flow cytometric analysis. (a) Representative TIL data (gated on lymphocyte populations) demonstrate that CD8+ T cells comprised the highest percentage of TIL after effector ACT on day 4 and after combinatorial ACT on day 14. (b) Overall comparison of total numbers of intratumoral CD8+ T cells on days 4 and 14 demonstrated a transient increase after effector ACT and a sustained increase after memory ACT and combinatorial ACT. This experiment was repeated twice with similar results, 4–5 mice per group. (* p < 0.05 compared with control condition; † p < 0.05 compared with day 4)

To characterize the increase in intratumoral CD8+ T cell infiltration seen following ACT, flow cytometric analysis was used to quantify the presence of Ly5.1+ adoptively transferred CD8+ T cells and Ly5.2+ endogenous CD8+ T cells. Representative flow cytometric data are shown as percentages in Fig. 3a; overall data are shown as total cell numbers in Fig. 3b-c. As shown in Fig. 3b, Ly5.1+ adoptively transferred CD8+ T cells comprised a small and short-lived fraction of TIL after effector ACT but were present in substantial and stable numbers after memory ACT; combinatorial ACT, which used half the number of T_mem used in memory ACT, resulted in smaller numbers of intratumoral Ly5.1+/CD8+ TIL. In contrast, as shown in Fig. 3c, combinatorial ACT resulted in the strongest intratumoral infiltration of endogenous CD8+ T cells.Fig3Fig. 3

The stronger local CD8+ T cell response seen with combinatorial ACT is largely the result of endogenous T cell recruitment. Seven days after tumor inoculation, Ly5.2+ mice were treated with 10⁵ T_eff, 10⁵ T_mem, or 5 × 10⁴ T_eff + 5 × 10⁴ T_mem derived from Ly5.1+ mice. Melanoma tumors were resected 4 or 14 days after ACT and TIL were harvested for flow cytometric analysis. (a) Representative data demonstrate higher percentages of Ly5.1+/CD8+ T cells in tumors after memory and combinatorial ACT. (b) Analysis of total numbers of Ly5.1+/CD8+ T cell populations showed that significant and durable infiltration of adoptively transferred cells was observed after memory ACT; combinatorial ACT, which used half the number of adoptively transferred T_mem as memory ACT, resulted in smaller numbers of intratumoral Ly5.1+/CD8+ T cells. (c) Analysis of total numbers of Ly5.2+/CD8+ T cell populations showed that memory ACT resulted in a significant recruitment of endogenously-derived CD8+ T cells that persisted on day 14; despite using half the number of T_mem, combinatorial ACT also resulted in durable infiltration of endogenously-derived CD8+ T cells. This experiment was repeated twice with similar results, 4–5 mice per group. (* p < 0.05 compared with control condition; † p < 0.05 compared with day 4)

To determine if combinatorial ACT improved systemic antitumor responsiveness in addition to local antitumor activity, splenocytes harvested 14 days after ACT were stimulated with or without GP33 peptide in the presence of IL-2 and brefeldin A, and stained to assess intracellular levels of IFNγ expression. Representative flow cytometric data are shown as percentages in Fig. 4a; overall data are shown as total cell numbers in Fig. 4b. As shown in Fig. 4b, memory ACT resulted in stronger splenocyte reactivity to GP33 than effector ACT, as we have previously described.⁷ However, combinatorial ACT produced significantly greater IFNγ expression among splenocytes than memory ACT, suggesting that combinatorial ACT may induce a much stronger level of CD8+ T cell responsiveness to tumor antigen.Fig4Fig. 4

Combinatorial ACT results in the strongest level of systemic CD8+ T cell responsiveness to tumor antigen. Seven days after tumor inoculation, Ly5.2+ mice were treated with 10⁵ T_eff, 10⁵ T_mem, or 5 × 10⁴ T_eff + 5 × 10⁴ T_mem derived from Ly5.1+ mice. Spleens were resected 14 days after ACT and splenocytes were incubated with media alone (control) or with tumor antigen (GP33). (a) Representative data demonstrate negligible levels of IFNγ expression by splenocytes in response to GP33 stimulation after no ACT or effector ACT, with increasing levels seen after memory ACT and combinatorial ACT, respectively. (b) The highest levels of IFNγ expression by splenocytes in response to GP33 stimulation were seen after combinatorial ACT. This experiment was repeated with similar results, 4 mice per group. (* p < 0.05 compared with control condition; † p < 0.05 compared with effector ACT, ‡ p < 0.05 compared with memory ACT)

To further explore the improved efficacy of combinatorial ACT, T_eff and T_mem were co-cultured in vitro with B16GP33 melanoma cells in the presence of IL-2 for 72 h. As shown in Fig. 5, serial MTT assays demonstrated that T_eff inhibited melanoma more potently than T_mem at 6 h. However, whereas the ability of T_eff to inhibit melanoma was lost at later timepoints, T_mem continued to inhibit melanoma for the duration of time points tested (up to 72 h), suggesting that the enhanced therapeutic efficacy of combinatorial ACT could be due to the temporal combination of early melanoma inhibition mediated by T_eff plus delayed but durable melanoma inhibition mediated by T_mem.Fig5Fig. 5

T_mem and T_eff have different temporal patterns of melanoma inhibition in vitro. T_eff and T_mem were co-cultured with B16GP33 melanoma cells in the presence of 10 ng/mL recombinant human IL-2 for 72 h. Control wells contained no T cells. Serial MTT assays were performed every 6 h. Over approximately the first 8 h, T_eff produced significantly greater inhibition of B16GP33. At all later time points, T_mem demonstrated significantly greater suppression of tumor growth. This experiment was repeated twice with similar results, 3 samples per group. († p < 0.05 compared with T_eff)

To determine if the improved efficacy of combinatorial ACT was attributable to local interactions between cell types, T_eff, T_mem, and T_eff + T_mem were co-cultured with B16GP33 cells in the presence of IL-2 for 8 h. As shown in Fig. 6a, whereas T_eff and T_mem resulted in comparable levels of melanoma inhibition at this time point, the combination of T_eff + T_mem resulted in a stronger level of melanoma inhibition. To determine if this enhancement was attributable to paracrine effects, T_eff and T_mem were separately stimulated with GP33 peptide for 5 h, after which cells were pelleted by centrifugation and conditioned media were collected. As shown in Fig. 6b, the addition of conditioned media from T_mem enhanced the ability of T_eff to inhibit melanoma to a comparable level seen with T_eff + T_mem. The addition of conditioned media from T_eff did not enhance the ability of T_mem to inhibit melanoma (data not shown). When anti-IL-2 blocking antibody was added to conditioned media from T_mem, this effect was lost, suggesting that paracrine release of IL-2 from T_mem may potentiate tumor cell death caused by T_eff. The addition of anti-IL-2 blocking antibody alone did not affect the ability of T_eff to inhibit melanoma.Fig6Fig. 6

Paracrine release of IL-2 from T_mem may potentiate melanoma inhibition by T_eff. The local interaction of T_eff and T_mem was examined by performing in vitro melanoma inhibition assays. (a) T_eff and T_mem and T_eff + T_mem were co-cultured with B16GP33 for 8 h. Although melanoma inhibition was comparable between T_eff and T_mem, significantly stronger inhibition of melanoma was observed with the combination of T_eff + T_mem. (b) Conditioned media (CM) was collected from T_mem stimulated with GP33 for 5 h. The addition of CM enhanced the efficacy of T_eff to a similar degree as the addition of T_mem. The effect of CM was negated by the addition of anti-IL-2 blocking antibody; the addition of anti-IL-2 blocking antibody alone did not affect the ability of T_eff to inhibit melanoma. The addition of anti-IL-2 blocking antibody blunted the ability of T_eff + T_mem to inhibit melanoma, but it could not be determined if this effect was exerted on T_eff or T_mem (data not shown). This experiment was repeated twice with similar results, 3 samples per group. (p-values as shown on horizontal bars)