Determinants of successful CD8⁺ T cell adoptive immunotherapy for large established tumors in mice

Abstract

Purpose

Adoptive cell transfer (ACT) of tumor-infiltrating or genetically engineered T cells can cause durable responses in patients with metastatic cancer. Multiple clinically-modifiable parameters can comprise this therapy, including cell dose and phenotype, in vivo antigen re-stimulation, and common gamma-chain (γ_c) cytokine support. However, the relative contributions of each these individual components to the magnitude of the anti-tumor response have yet to be quantified.

Experimental design

To systematically and quantitatively appraise each of these variables, we employed the Pmel-1 mouse model treating large, established B16 melanoma tumors. In addition to cell dose and magnitude of in vivo antigen re-stimulation, we also evaluated the relative efficacy of central memory (T_CM), effector memory (T_EM), and stem cell memory (T_SCM) subsets on the strength of tumor regression as well as the dose and type of clinically available γ_c cytokines, including IL-2, IL-7, IL-15, and IL-21.

Results

We found that cell dose, T cell differentiation status, and viral vaccine titer each were correlated strongly and significantly with the magnitude of tumor regression. Surprisingly, although the total number of IL-2 doses was correlated with tumor regression, no significant benefit to prolonged (≥6 doses) administration was observed. Moreover, the specific type and dose of γ_c cytokine only moderately correlated with response.

Conclusions

Collectively, these findings elucidate some of the key determinants of successful ACT immunotherapy for the treatment of cancer in mice and further demonstrate that γ_c cytokines offer a similar ability to effectively drive anti-tumor T cell function in vivo.

Introduction

Adoptive cell transfer (ACT), the infusion of ex vivo expanded anti-tumor or antiviral lymphocytes to patients, represents a potent therapy for certain advanced solid tumors (1–3), relapsed hematologic malignancies (4), and post transplant viral-induced lymphoproliferative disorders (5). Once the purview of highly specialized research centers, multiple technical and conceptual innovations in recent years have allowed this promising approach to cancer therapy to extend to a wider range of institutions and patients. For example, the methodology used to generate tumor infiltrating lymphocytes (TIL) has been greatly simplified by using unselected and minimally cultured T cells as the infusion product (6;7). Importantly, the ability of TIL to mediate objective cancer regression in roughly 50% of metastatic melanoma patients when combined with prior host conditioning using lymphocyte-depleting chemotherapy (8) has now been demonstrated at multiple institutions outside of the intra-mural National Institutes of Health (6;9). Moreover, with the recent advent of clinical grade retro- and lenti-viral vectors capable of genetically redirecting the reactivity of autologous peripheral lymphocytes towards tumor or viral-associated antigens by introduction of exogenous T cell receptors (TCR) (10–13) or chimeric antigen receptors (14–18), adoptive T cell immunotherapy can potentially be deployed more widely and easily. Thus, there has been a notable increase in the number of phase I and II clinical trials using TIL or genetically re-directed T cells initiated throughout the world in recent years (3).

ACT allows for the precise manipulation of multiple variables that may impact treatment efficacy. Previous work in humans has correlated in vivo persistence of transferred T cells (18;19), telomere length (19), the re-expression of CD27 after IL-2 withdrawal (19), and lympho-depleting host conditioning (20;21) with cancer regression. In addition to host pre-conditioning before cell transfer (22–24), pre-clinical experiments in mice have identified additional parameters associated with treatment outcome, including the dose (23;25) and differentiation status of transferred T cells (25–28), reactivation of transferred T cells at the time of cell infusion (29–31), and the efficiency of TCR transduction for gene-modified T cells (23). While each of these factors may incrementally improve T cell therapies, knowledge of which factors contribute most greatly to tumor regression is essential for prioritizing future clinical development. It is for this reason that we sought to systematically and quantitatively assess the impact of multiple variables potentially amenable to clinical manipulation to improve future adoptive T cell-based immunotherapy protocols.

Specifically, we evaluated the impact of the strength of in vivo Ag re-stimulation with a recombinant viral vaccine, T cell dose and differentiation status, and schedule, dose, and type of exogenous common gamma chain (γ_c) cytokine support on the magnitude of tumor regression. As the donor source of T cells, we used cells derived from the Pmel-1 TCR transgenic mouse whose CD8⁺ T cells have specificity for a D^b-restricted epitope derived from the shared self/tumor antigen (Ag) gp100 (29). Our findings demonstrated that not all parameters evaluated contributed equally to the magnitude of the anti-tumor response mediated by adoptively transferred CD8⁺ T cells. Indeed, intensity of in vivo Ag re-stimulation, T cell differentiation status, and the absolute number of infused T cells were each strongly correlated with tumor regression. By contrast, the number, dose, and type of γ_c cytokine administered to support transferred T cells were each less strongly associated with treatment outcomes. Taken together, these results provide for the rationale prioritization, design, and execution of future T cell based immunotherapy protocols in humans.

Materials and Methods

Mice and tumor lines

The Pmel-1 (The Jackson Laboratory) TCR transgenic mouse has been previously described (29). Female C57BL/6 mice (The Jackson Laboratory) were used at 6 to 12 weeks of age as recipient mice in all cell transfer experiments. B16 (H-2^b), a spontaneous gp100⁺ murine melanoma cell line, was obtained from the National Cancer Institute tumor repository and maintained in complete media (CM). All animal experiments were conducted with the approval of the National Cancer Institute(NCI) Animal Use and Care Committee.

Generation of CD8⁺ T cells for adoptive transfer

To generate Ag-specific CD8⁺ T cells for cell transfer, Pmel-1 splenocytes were isolated, pulsed with 1 μM of human gp100_25–33, and expanded for 1 week in the presence of CM supplemented with 4 ng/mL of recombinant human IL-2 (rhIL-2; Novartis). In some experiments, CD8⁺ T memory stem cells (T_SCM), CD8⁺ T central memory cells (T_CM), or CD8⁺ T effector memory cell (T_EM) were generated as described previously (28). Briefly, naïve Pmel-1 CD8⁺ T cells were isolated from a single cell suspension of splenocytes after ACK lysis using a MACS negative selection kit (Miltenyi Biotec) and cells were activated with 2μg/ml of plate bound anti-CD3, 1 μg/ml of soluble anti-CD28 (both from BD Biosciences), and expanded in 10 ng/mL of IL-2. T_SCM and T_CM cells were generated in the presence of 7 μM of the GSK-3βinhibitor TWS119 (EMD Biosciences) dissolved in DMSO (Sigma-Aldrich) while T_EM cells were generated from separate cultures derived from an aliquot of the same single cell suspension in the presence of vehicle only. After 5d in culture, cells were fractionated into CD44^lowCD62L^high (T_SCM), CD44^highCD62L^high (T_CM), and CD44^highCD62L^low (T_EM) populations to a purity of >98% using a FACSAria cell sorter(BD Biosciences).

Adoptive cell transfer, poxvirus vaccination, and γ_c cytokine administration

Unless otherwise indicated, mice (n≥5 per group) were injected s.c. with 2 × 10⁵ to 5 × 10⁵ B16 melanoma cells. Ten to 14 days later, treated mice received intravenous (i.v.) injections of CD8⁺ T cells and indicated doses of one of two previously described (29) recombinant poxviruses encoding hgp100 (rPVhgp100): recombinant fowlpox virus expressing human gp100 ( rFPhgp100) or recombinant vaccinia virus expressing hgp100 (rVVhgp100). Prior work has demonstrated that both poxviruses provide similar efficacy when used in tumor treatment experiments (29). Where indicated, treated mice received 500 cGy of sublethal irradiation prior to ACT to mimic the lymphodepletion performed in our current clinical protocols (22). In addition to vaccination, all treated mice also received indicated doses of exogenous γ_c cytokine support with either rhIL-2 (Novartis), IL-7 (Peprotech), IL-15 (Peprotech), or IL-21 (R&D Systems) by twice daily intra-peritoneal (i.p.) injections. Except for rhIL-2, where dosing was varied between 1 to 14 total doses, all treated mice received 6 total doses of a γ_c cytokine. All tumor measurements were performed in a blinded fashion.

Statistics

The products of perpendicular tumor diameters were plotted as the mean ± SEM for each data point and tumor treatment graphs were compared using the Wilcoxon rank sum test. Regression analysis of the slope of tumor regression as a function of different variables related to ACT was performed as described previously (25;32;33) using a minimum of 3 mice for each parameter analyzed. Individual mice must have been evaluable for at least 3 separate time points in order for a slope to be calculated. Differences in animal survival were assessed using the log-rank test. All statistics were calculated with Prism 5 GraphPad software (GraphPad Software Inc).

Results

Dose of adoptively transferred CD8⁺ T cells strongly correlates with tumor regression

In some clinical ACT trials, the absolute number of infused CD8⁺ T cells has retrospectively been correlated with the likelihood of obtaining an objective response consistent with standardized oncologic response criteria (6). This finding has not been uniformly found in all ACT trials, however (8;20). Moreover, the generation and expansion of large numbers of T cells for infusion requires a significant investment in limiting resources such as allogeneic feeder cells and soluble OKT-3, laboratory and incubator space in GMP facilities, and technical personnel to care for these cells. For these reasons, we sought to determine the impact of the infusion dose of Ag-specific CD8⁺ T cells on the magnitude of tumor regression. We therefore adoptively transferred titrated doses of Pmel-1 CD8⁺ T cells into mice bearing established B16 melanoma tumors. Similar to current human cell transfer protocols (6–8;20;34), mice received ex vivo expanded self/tumor-reactive CD8⁺ T cells grown in IL-2. In addition to ACT, all treated mice received an identical, intermediate titer (10⁷ pfu) of rFPhpg100 vaccination and exogenous IL-2 as well as pre-conditioning with sublethal irradiation.

We found that infusion with as few as 10⁴ CD8⁺ T cells caused a significant delay in tumor growth compared with untreated control mice (Figure 1A; P < 0.05 for all treatment groups compared with untreated control). In order to gain further insight into the dose-response relationship between infusion cell dose and the magnitude of the antitumor response, we plotted the slope of tumor regression of treated mice as a function of cell dose on a semi-log plot. We found a significant linear correlation between the dose of infused anti-tumor CD8⁺ T cells and strength of tumor regression (Figure 1B; R² = 0.71, P < 0.03). Finally, to assess whether these improvements in tumor regression with increased cell dose translated into extended survival, we plotted the survival of treated mice (Figure 1C). We found a significant, step-wise improvement in animal survival as the total number of infused CD8⁺ T cells was increased (P < 0.05). A similar dose-response relationship between the number of infused anti-tumor CD8⁺ T cells and the strength of tumor regression and improvement in animal survival was also observed in mice that did not receive prior lymphodepletion (Supplementary figure 1A–C). These data demonstrate that the dose of infused tumor-reactive CD8⁺ T cells infused is strongly correlated with the magnitude of tumor regression.

Figure 1.

Dose of adoptively transferred CD8⁺ T cells strongly correlates with in vivo antitumor treatment efficacy. (A) Sublethally irradiated B6 mice bearing 13 day established s.c. B16 tumors were left untreated as controls (×) or received 10⁷ pfu of rFPhgp100 viral vaccine and exogenous rhIL-2 (36 μg/dose x 6 doses) in combination with 10⁴ (△), 10⁵ (◆), 10⁶ (○), or 10⁷ (■) in vitro-activated Pmel-1 CD8⁺ T cells. Results are presented as the mean measurements from 5 mice per group (± SEM) and are representative of 3 independently performed experiments. (B) Antigen-specific CD8⁺ cell dose correlates with the magnitude of regression for established subcutaneous B16 melanoma tumors on a semi-log plot. The slope of tumor regression (mm²/day) for individually treated mice is shown as a function of in vitro-activated Pmel-1 CD8⁺ T cell dose. All mice received identical host pre-conditioning, rFPhgp100 vaccination, and exogenous rhIL-2. Log regression analysis was performed to determine the best line of fit and the Correlation Coefficient (R²) and the significance of correlation (P) were calculated. R² = 0.71, P < 0.03. (C) Survival curves of untreated mice and mice receiving, exogenous IL-2, rFPhgp100 vaccine, and titrated doses of in vitro-activated adoptively transferred Pmel-1 CD8⁺ T cells. * = P < 0.05, ** = P < 0.01 using Log-rank statistics. Survival data shown was derived from 2 pooled independently performed experiments.

CD8⁺ T cell differentiation status is inversely correlated with in vivo anti-tumor treatment efficacy

Presently, it has been difficult to uncouple the processes of T cell proliferation and expansion from differentiation and senescence such that prolonged ex vivo culture results in therapeutically less effective T cells in both mice (25) and humans (35–37). As the variables of cell yield and differentiation status are at apparent odds with one another, we next aimed to quantify the extent to which T cell differentiation status may impact in vivo treatment efficacy. Therefore, we next assessed the anti-tumor response mediated by adoptive transfer of a minimal number (4 × 10⁴) of Ag-specific CD8⁺ T cells possessing a defined differentiation status based on the pattern of expression for the surface molecules CD62L and CD44. Using these markers, mouse CD8⁺ T cells may be subdivided into defined antigen-experienced T cell populations termed T effector memory (T_EM; CD62L^loCD44^hi), T central memory (T_CM; CD62L^hiCD44^hi), and a recently described subset termed T stem cell memory (T_SCM; CD62L^hiCD44^lo)(28), each of which has orthologues among human T cells (38). T cell cultures were generated either in the presence or absence of the glycogen synthase kinase-3β (GSK-3β) inhibitor TWS119, a pharmacologic agonist of the canonical Wnt/β-catenin pathway, and T cell subsets were FACS-sorted to a purity of >98% prior to cell transfer. All treated mice received sublethal irradiation, vaccination with 10⁷ pfu of rVVhgp100, and exogenous IL-2 support.

At this near limiting dose of adoptively transferred T cells, we observed no significant delay in tumor growth in untreated mice compared with mice receiving T_EM CD8⁺ T cells (Figure 2A; P > 0.05). By contrast, adoptive transfer of either T_CM or T_SCM CD8⁺ T cells caused a significant delay in tumor growth compared with both untreated controls as well as mice receiving T_EM cells (Figure 2A; P < 0.01 and < 0.05, respectively). Further, T_SCM cells caused a more sustained reduction in tumor growth compared with T_CM cells (P < 0.05). When the slope of tumor regression was plotted as a function of T cell differentiation status, we observed a significant linear correlation between these two parameters (Figure 3B; R² = 0.72, P < 0.0001). This improvement in the control of tumor growth with transfer of progressively less-differentiated CD8⁺ T cells also translated into a significant enhancement in overall survival of treated animals (Figure 2C; P < 0.02). These findings demonstrate that beyond the absolute number of adoptively transferred CD8⁺ T cells, T cell differentiation status can also significantly impact the outcome of ACT therapy, even when limiting numbers of T cells are infused.

Figure 2.

CD8⁺ T cell differentiation status is inversely correlated with in vivo anti-tumor treatment efficacy. (A) Sublethally irradiated B6 mice bearing 10 day established s.c. B16 tumors were left untreated as controls (×) or received 4 × 10⁴ of FACS-sorted Pmel-1 CD8⁺ T cells with the phenotype of T stem cell memory (T_SCM) (■), T central memory (T_CM) (○), or T effector memory (T_EM) (▲) populations generated by in vitro priming of Pmel-1 splenocytes with 1 μM hgp100_25-33 in the presence of 7 μM of TWS119 and 10 ng/mL of rhIL-2. T_SCM = CD44^lowCD62L^high; T_CM = CD44^highCD62L^high; T_EM = CD44^highCD62L^low. All treated mice received rVVhgp100 vaccination (10⁷ pfu) and exogenous rhIL-2 (36 μg/dose × 6 doses). Results are presented as the mean measurements from 5 mice per group (± SEM) and are representative of 3 independently performed experiments. (B) CD8⁺ cell differentiation status inversely is correlated with the magnitude of regression for established subcutaneous B16 melanoma tumors. The slope of tumor regression (mm²/day) for individually treated mice is shown as a function of differentiation status of adoptively transferred CD8⁺ T cells, as determined by classification into T_SCM, T_CM, or T_EM subsets. All mice received identical host pre-conditioning, rVVhgp100 vaccination, and exogenous rhIL-2. Linear regression analysis was performed to determine the best line of fit and the Correlation Coefficient (R²) and the significance of correlation (P) were calculated. R² = 0.72, P < 0.0001. (C) Survival curves of untreated mice and mice receiving, exogenous IL-2, rVVhgp100 vaccine, and adoptively transferred T_SCM, T_CM, or T_EM Pmel-1 CD8⁺ T cells. * = P < 0.02, ** = P <0.0001 using Log-rank statistics. Data shown was derived from 2 pooled independently performed experiments.

Figure 3.

Intensity of in vivo antigen re-stimulation using a recombinant viral vaccine is a critical determinant of the in vivo anti-tumor efficacy of adoptively transferred CD8⁺ T cells. (A) Sublethally irradiated B6 mice bearing 12 day established s.c. B16 tumors were left untreated as controls (×) or received 10⁶ adoptively transferred Pmel-1 CD8⁺ T cells in combination with either 10⁵ (△), 10⁶ (◆), 10⁷ (○), or 10⁸ (■) pfu of rFPhgp100 vaccine. In addition to in vivo antigen re-stimulation with a recombinant viral vaccine, all treated mice also received exogenous rhIL-2 (36 μg/dose × 6 doses). Results are presented as the mean measurements from 5 mice per group (± SEM) and are representative of 2 independently performed experiments. (B) Viral vaccine titer correlates with the slope of tumor regression for individual mice on a semi-log plot. The slope of tumor regression (mm²/day) for individual treated mice is shown plotted as a function of rFPhgp100 vaccine titer. Log regression analysis was performed to determine the best line of fit and the Correlation Coefficient (R²) and the significance of correlation (P) were calculated. R² = 0.78, P < 0.001. (C) Survival curves of untreated mice and mice receiving adoptively transferred Pmel-1 CD8⁺ T cells, exogenous IL-2, and titrated doses of a rFPhgp100 vaccine. * = P < 0.01 using Log-rank statistics. Data shown are representative of two independently performed experiments. (D) Percent change in weight of tumor-bearing mice irradiated and left untreated (×) or who received 10⁶ Pmel-1 CD8⁺ T cells, 10⁸ pfu of rVVhgp100, and exogenous IL-2 (■). Results are the mean percent change in body weight (± SEM) of 5 mice per group relative to a DO measurement. * = P < 0.05, ** = P < 0.01 by Student’s T test.

Intensity of in vivo antigen re-stimulation is a critical determinant of the anti-tumor efficacy of adoptively transferred CD8⁺ T cells

Antigen re-stimulation, either immediately prior to or directly following cell transfer, has been shown to significantly enhance the in vivo anti-tumor efficacy of adoptively transferred T cells in both pre-clinical (29–31;39;40) and early phase clinical studies (18;41). We therefore next sought to test the hypothesis that increasing the strength of in vivo Ag re-stimulation could improve the anti-tumor function of adoptively transferred CD8⁺ T cells, we infused mice bearing established B16 tumors with a dose (10⁶) of IL-2 expanded Pmel-1 CD8⁺ T cells previously established to be non-curative in our model (25). In order to systematically modulate the intensity of Ag re-stimulation, treated mice were given titrated doses of a recombinant viral vaccine expressing an anchor-modified variant of gp100 (rFPhgp100) over a 4-log dose range. Additionally, all treated mice received lymphodepletion with 500 cGy of irradiation prior to cell transfer and exogenous IL-2 to mimic current cell transfer protocols in patients (6;7;20).

We found that at or above a threshold vaccine titer of 10⁶ pfu, the in vivo antitumor function of transferred CD8⁺ T cells was significantly enhanced compared with non-treated control mice (Figure 3A; P < 0.02 for all groups). By contrast, there were no significant differences in the rate of tumor growth observed between control mice and mice that received ACT and vaccination with a rFPhgp100 titer of 10⁶ pfu or less (P > 0.05).

In order to gain further insight into the dose-response relationship between vaccine titer and the magnitude of the antitumor response, we next evaluated the slope of tumor regression for individual treated mice as a function of rFPhgp100 titer. We found that the titer of vaccine was strongly correlated with the slope of tumor regression for individual mice on a semi-log plot (Figure 3B; R² = 0.78, P < 0.001). To determine whether the magnitude of vaccine-enhanced tumor regression mediated by adoptively transferred CD8⁺ T cells correlated with improved animal survival, we plotted the survival of mice receiving titrated doses of vaccine in combination with ACT. Similar to the pattern of tumor regression, we saw a significant and progressive improvement in animal survival as the intensity of vaccination was increased (Figure 3C; P < 0.01). A similar, although less prominent, trend of improved tumor regression and animal survival with a progressive increase in vaccine titer was also observed in mice that did not receive lymphodepletion prior to cell transfer (Supplementary Figures 2A–C).

In a recent report, lethal toxicity from an uncontrolled cytokine storm was observed in mice receiving ACT of self/tumor-reactive CD8⁺ T cells followed by repeated vaccination with a synthetic long peptide (42) combined with IL-2 and imiquimod, a toll-like receptor (TLR) 7 agonist (43). We therefore sought to determine whether mice treated with a single injection of the highest dose (10⁸ pfu) of a recombinant poxvirus vaccine we tested experienced similar toxicities. Treated mice displayed a moderate (<10%) but significant loss of body weight in the first week compared with irradiated but otherwise untreated controls, but this weight loss entirely abated by the second week (Figure 3D). Importantly, no treated mice at any of dose levels evaluated acutely expired following treatment (Figure 3C and Supplementary Figure 2C).

Taken together, these data indicated that the strength of in vivo Ag re-stimulation provided by a recombinant viral vaccine is strongly and significantly correlated with the magnitude of anti-tumor treatment efficacy mediated by adoptively transferred CD8⁺ T cells in both lympho-depleted as well as lympho-replete hosts.

Number of IL-2 doses rapidly reaches an asymptotic dose-response relationship

Adoptively transferred T cells are supported with exogenous IL-2 following cell infusion in many current clinical cell transfer protocols (6;20;44). When administered as a single bolus agent, the absolute number of IL-2 doses received during the first cycle of therapy has been correlated with response to treatment in patients with metastatic melanoma (45). However, IL-2 can cause a number of significant dose-limiting toxicities, including hypotension, pulmonary edema, and neuro-cognitive changes. Moreover, exogenous IL-2 may also potentiate the suppressive activity and homeostasis of CD4⁺FOXP3⁺ T regulatory (Treg) cells by the nature of their constitutive expression of a component of the high affinity IL-2 receptor, CD25 (46;47). Thus, we wished to establish whether there are an optimal number of IL-2 doses to administer following ACT in our model.

To address this question, we treated sub-lethally irradiated mice bearing established B16 melanoma tumors with adoptive transfer of 10⁶ IL-2 expanded Pmel-1 CD8⁺ T cells, 10⁷ pfu rFPhgp100 vaccine, and between 0 to 14 total doses of rhIL-2 administered as a 36 μg intra-peritoneal injection twice daily. We found a threshold effect for the total number of IL-2 doses such that the administration of 2 or more doses was associated with a significant delay in tumor growth compared with untreated controls (Figure 4A; P < 0.02). By contrast, ACT and vaccination alone or ACT, vaccination and a single dose of IL-2 was insufficient to impact the rate of tumor growth (P > 0.05). Overall, we found a moderate but significant correlation between the total number of IL-2 doses administered and the magnitude of tumor regression (Figure 4B; R² = 0.51, P < 0.0001 for the overall assessment of 0 to 14 doses). Surprisingly, however, we found no significant benefit to protracted IL-2 administration beyond 6 doses (Figure 4B; R² = 0.06, P = 0.29 for the assessment of 6 to 14 doses). These data demonstrate that the benefit of exogenous IL-2 to support adoptively transferred T cells rapidly reaches asymptotic dose-response relationship.

Figure 4.

The impact of exogenous IL-2 on the antitumor treatment efficacy of adoptively transferred CD8⁺ T cells rapidly reaches an asymptotic dose-response relationship. (A) Sublethally irradiated B6 mice bearing 12 day established s.c. B16 tumors were left untreated as controls (×) or received 10⁶ in vitro-activated Pmel-1 CD8⁺ T cells and 10⁷ pfu of rFPhgp100 vaccine either alone (□) or in combination with 1 (●), 2(△), 6 (◆), 10 (○), or 14 (■) doses of rhIL-2 (36μg/dose). Results are presented as the mean measurements from 5 mice per group (± SEM) and are representative of 2 independently performed experiments. (B) The total number of IL-2 doses administered overall correlates with the magnitude of tumor regression mediated by adoptively transferred Pmel-1 CD8⁺ T cells, but not beyond 6 doses. The slope of tumor regression (mm²/day) for individually treated mice is shown as a function of total number of IL-2 doses administered. All mice received identical host pre-conditioning, dose of Pmel-1 CD8⁺ T cells, and dose of rFPhgp100 vaccine. Linear regression analysis was performed to determine the best line of fit, Correlation Coefficient (R²), and the significance of correlation (P) for the overall trend of 0 through 14 doses and the subset of ≥6 doses. R² = 0.51 and P < 0.0001 for the overall slope (—); R² = 0.06 and P < 0.29 for the slope of ≥6 doses of IL-2 (…).

IL-2, IL-7, IL-15, and IL-21 cause dose-dependent augmentation of in vivo antitumor treatment mediated by adoptively transferred CD8⁺ T cells

Having established that 6 doses of exogenous IL-2 was optimal for antitumor treatment in the Pmel-1 model, we next sought to determine the relative impact of dose titrating 6 total doses of alternative γ_c-signaling cytokines. Specifically, we compared escalated doses of IL-7 (48;49), IL-15 (50), and IL-21 (51;52) to IL-2 as each of these cytokines are known to activate, expand, or promote the survival of T cells and all are either commercially available for human use or have entered into human clinical trials. While each of the γ_c cytokines have overlapping functions by nature of their shared ability to activate common signal transducer and activator of transcription (Stat) family members, including Stat5, individual cytokines may possess unique or selective activities (53).

We therefore adoptively transferred a non-curative dose (10⁶) of in vitro-activated, IL-2 expanded Pmel-1 CD8⁺ T cells into sub-lethally irradiated mice bearing established B16 melanoma tumors in combination with 10⁷ pfu of rFPhgp100 vaccine. Additionally, treated mice received exogenous γ_c cytokine support at a dose range in which other investigators observed immunologic activity (54;55). Similar to what we observed with respect to the total number of IL-2 injections, we found only a moderate correlation between the dose of any given γ_c cytokine and the magnitude of the anti-tumor response (Figure 5; R² = 0.31 to 0.51, P < 0.001). IL-15 had the weakest linear dose-response correlation (R² = 0.31) while IL-21 had the strongest (R² = 051). However, when the dose-response relationship was projected as a decay exponential curve, IL-15 was found to most rapidly reach a plateau in response compared with the other cytokines, followed by IL-2, suggesting that some cytokines can reach their maximal in vivo activity more rapidly than others. No overt evidence of cytokine-related toxicity was observed with any of the cytokines tested as the vast majority of mice treated at the highest dose levels remained alive beyond 2 weeks after the final cytokine dose (data not shown). Taken together, these data demonstrate that all γ_c cytokines are similarly capable of augmenting the in vivo anti-tumor function of adoptively transferred CD8⁺ T cells.

Figure 5.

IL-2, IL-7, IL-15, and IL-21 cause dose-dependent augmentation of in vivo antitumor treatment mediated by adoptively transferred CD8⁺ T cells. Dose of exogenously administered γ_c-signaling cytokines correlates with the slope of tumor regression mediated by adoptively transferred CD8⁺ T cells. The slope of tumor regression (mm²/day) for individually treated mice is shown as a function of dose of γ_c cytokine administered. Linear regression (solid black line) analysis was performed to determine the best line of fit for tumor treatment associated with each γ_c cytokine and the Correlation Coefficient (R²) and the significance of correlation (P) were calculated. IL-2, R² = 0.39 and P < 0.0001; IL-7, R² = 0.48 and P < 0.001; IL-15, R² = 0.31 and P < 0.0001; IL-21, R² = 0.51 and P < 0.001. Grey dashed line = exponential decay curve.

Discussion

In the present study, we found that the parameters of CD8⁺ T cell dose, the differentiation status of the input population of T cells, and the intensity of in vivo Ag re-stimulation were each independently strongly (R² > 0.7) and significantly (P < 0.03) correlated with the magnitude of tumor regression in our model. By contrast, the total number of IL-2 doses administered and the specific type and dose of γ_c cytokine used to support adoptively transferred T cells were only moderately (R² = 0.31 to 0.51) correlated with the anti-tumor response. Our conclusions were based on data obtained using CD8⁺ T cells derived from a single TCR transgenic mouse (Pmel-1) that targets the shared melanocyte/melanoma differentiation antigen gp100, an Ag naturally expressed on the surface of the transplantable but genetically un-manipulated melanoma cell line B16. Despite these caveats, our findings suggest multiple parameters that may be optimized to improve the efficacy of future T cell-based adoptive immunotherapy clinical trials.

Recently, Rizzuto et al. proposed that a parabolic dose-response relationship exists between the number of adoptively transferred tumor-reactive CD8⁺ T cells and the magnitude of the anti-tumor response (56). This conclusion was based on experiments in which above a relatively small threshold number (10⁵) of infused naïve self/tumor-reactive CD8⁺ T cells, diminished on-target therapeutic and autoimmune effects were observed. Using Ag-experienced rather than naïve T cells and a similar dynamic range of cell numbers, we found neither a plateau nor a decrease in the dose-response curve. On the contrary, we found a progressive improvement in the strength of tumor regression as the total number of adoptively T cells was increased. These findings held true regardless of whether or not the host received pre-conditioning with radiation. This apparent discrepancy between the behavior of naïve and Ag-experienced T cells targeting the same Ag may in part be explained by a greater sensitivity to Ag, differential requirement for co-stimulation, and faster recall kinetics of memory relative to naïve CD8⁺ T cells (38).

While these data, when viewed in isolation, would appear to suggest that infusion of the maximum possible number of T cells should be a goal of future T cell transfer protocols, this approach is not without important potential tradeoffs. Thus far, it has been difficult to uncouple the processes of cell expansion and differentiation status such that the longer cells are expanded ex vivo, the more senescent they become (25;36). Thus, we also assessed the impact of T cell differentiation status on treatment outcomes while holding the number infused cells constant. We found a significant linear correlation between the differentiation status of infused T cells and the strength of tumor regression in the order T_SCM >T_CM > T_EM. Importantly, when T_EM cells were transferred at near limiting numbers (10⁴ cells), we found no statistical difference in anti-tumor treatment efficacy compared with untreated control mice. Although not specifically evaluated in our current experiments, previous work in mice (25;26) and humans (57) has demonstrated that less differentiated T cells possess a number of desirable attributes relative to their more highly differentiated counterparts, including a greater proliferative capacity and ability to persist following cell transfer. In humans, the ability to persist has been positively correlated with response to ACT treatments (18;19;57). These findings, taken together with our observations regarding the impact of cell dose on treatment outcomes, demonstrate that the quality of transferred T cells, here assessed by surface phenotypes corresponding to different memory T cells subsets, is as important to the efficacy of T cell transfer therapy as is the absolute number of cells infused. For these reasons, methods to generate large numbers of relatively less differentiated T cells for adoptive immunotherapy remains a key clinical priority (58–60).

Our present findings again emphasize the potential benefit that providing Ag stimulation or a mimetic for Ag stimulation to adoptively transferred T cells may provide to inducing maximal tumor regression (61). While Ag-specific vaccination in combination with ACT or infusion of tumor-reactive T cells co-expressing receptors for a chronic viral infection has been associated with augmented tumor regression in some patients (18;41), a consistent benefit to adding a mono-specific vaccine with cell transfer therapy beyond cell transfer alone has yet to be observed (8). Several factors may account for this phenomenon, including the use of relatively weak immunogens, such as peptide vaccines with oil-in-water emulsions as adjuvant (8). For this reason, clinical protocols using recombinant viral vaccines encoding tumor-associated Ags in combination with adoptive transfer of TIL or TCR gene-engineered T cells are actively being pursued by our group. Alternatively, the use of high-intensity conditioning regimens prior to ACT (24;33) or the “programming” of T cells at the time of cell infusion may (39) also provide stimulatory ques that mimic vaccination but in an Ag-independent manner.

IL-2 is often used to support the persistence, expansion, and activation of adoptively transferred T cells in many current clinical studies. However, as IL-2 possesses a number of limiting toxicities requiring close patient monitoring in an in-patient setting, and as this cytokine has the potential to expand Tregs as well as activate effector cells (46;47), minimizing the amount of IL-2 administered would be a desirable goal if it did not negatively impact treatment outcomes. We therefore assessed the impact of varying the total number of IL-2 doses administered to support transferred T cells on tumor regression. We found that there was an overall moderate (R² > 0.5) correlation between the number of IL-2 doses administered and the anti-tumor response. Surprisingly, however, we found that the benefit of IL-2 administration reached an asymptotic dose-response relationship such that beyond 6 doses, there was no significant improvement in the antitumor response (R² = 0.06). These findings may offer some reassurance to providers caring for patients on study as the total number of IL-2 doses an individual patient receives is often dictated by physiology rather than a pre-determined total number of doses.

Finally, we compared the impact of supporting adoptively transferred cells with different clinically available γ_c cytokines, including IL-7, IL-15, and IL-21, relative to IL-2. Although each of these cytokines signal in part through Stat5 based on their shared use of the γ_c receptor, each may in principle also deliver unique signals to responding T cells (53). We found that each γ_c cytokine was capable of augmenting tumor regression mediated by adoptively transferred T cells in a dose-dependent manner. Overall, no individual cytokine appeared to be capable of vastly improved strength of tumor regression compared with another. However, IL-15, and to a lesser extent IL-2, appeared to more rapidly reach a plateau in their dose-response curve compared with the other cytokines, such as IL-21, which maintained a much more linear relationship. This phenomenon may in part be due to the ability of these cytokines to be presented in trans to neighboring cells (62;63). Trans presentation allows for cytokines, such as IL-15, to maintain a higher serum half life and greater in vivo activity than would otherwise occur in the absence of this phenomenon (64). Indeed, a recent study evaluating the impact of daily injections of IL-15 to non-human primates found persistently elevated levels of plasma IL-15 and prolonged biologic activity (65).

In summary, we found that multiple parameters, including cell dose and differentiation status, strength of in vivo Ag re-stimulation, and the duration and type of γ_c cytokine support may each be adjusted to improve tumor treatment outcomes using adoptively transferred tumor reactive CD8⁺ T cells. Although in this study we systematically and quantitatively evaluated the impact of adjusting individual variables at a time, multiple variables may be adjusted concurrently in future ACT clinical trials in an effort to improve outcomes. For this reason, these findings provide several rational means of improving future T cell adoptive immunotherapy protocols.

Supplementary Material

1. Supplemental Figure 1.

Dose of adoptively transferred CD8⁺ T cells correlates with in vivo antitumor treatment efficacy in the non-ablated setting. (A) B6 mice bearing 13 day established s.c. B16 tumors were left untreated as controls (×) or received 10⁷ pfu of rFPhgp100 viral vaccine and exogenous rhIL-2 (36 μg/dose x 6 doses) in combination with 10⁴ (△), 10⁵ (◆), or 10⁶ (○) in vitro-activated Pmel-1 CD8⁺ T cells. Results are presented as the mean measurements from 5 mice per group (± SEM) and are representative of 2 independently performed experiments. (B) Antigen-specific CD8⁺ cell dose correlates with the magnitude of regression for established subcutaneous B16 melanoma tumors on a semi-log plot. The slope of tumor regression (mm²/day) for individually treated mice is shown as a function of in vitro-activated Pmel-1 CD8⁺ T cell dose. All mice received identical rFPhgp100 vaccination and exogenous rhIL-2. Log regression analysis was performed to determine the best line of fit and the Correlation Coefficient (R²) and the significance of correlation (P) were calculated. R² = 0.64, P < 0.03. (C) Survival curves of untreated mice and mice receiving, exogenous IL-2, rFPhgp100 vaccine, and titrated doses of in vitro-activated adoptively transferred Pmel-1 CD8⁺ T cells. * = < 0.01 using Log-rank statistics. Data shown is representative of 2 independently performed experiments.

2. Supplemental Figure 2.

Intensity of in vivo antigen re-stimulation using a recombinant viral vaccine is a determinant of the in vivo anti-tumor efficacy of adoptively transferred CD8⁺ T cells in the non-ablated setting. (A) B6 mice bearing 12 day established s.c. B16 tumors were left untreated as controls (×) or received 10⁶ adoptively transferred Pmel-1 CD8⁺ T cells in combination with either 10⁵ (△), 10⁶ (◆), 10⁷ (○), or 10⁸ (■) pfu of rFPhgp100 vaccine. In addition to in vivo antigen re-stimulation with a recombinant viral vaccine, all treated mice also received exogenous rhIL-2 (36 μg/dose × 6 doses). Results are presented as the mean measurements from 5 mice per group (± SEM) and are representative of 2 independently performed experiments. (B) Viral vaccine titer correlates with the slope of tumor regression for individual mice on a semi-log plot. The slope of tumor regression (mm²/day) for individual treated mice is shown plotted as a function of rFPhgp100 vaccine titer. Log regression analysis was performed to determine the best line of fit and the Correlation Coefficient (R²) and the significance of correlation (P) were calculated. R² = 0.60, P < 0.01. (C) Survival curves of untreated mice and mice receiving titrated viral vaccine doses. * = P < 0.01. Data shown are representative of two independently performed experiments.

Statement of translational relevance.

Simplification of the methods used to generate tumor-infiltrating lymphocytes combined with the ability to genetically redirect T cell-reactivity towards tumor-associated antigens has led to an increased number of phase I/II adoptive cell transfer (ACT) trials in recent years. Optimizing the design and execution of future ACT protocols will require a detailed understanding of how different parameters related to cellular therapies might impact treatment outcomes. Using the Pmel-1 CD8⁺ T cell receptor transengic model, a realistic murine model for the treatment of established melanomas targeting the shared self/tumor antigen gp100, we systematically and quantitatively assessed the relative contributions of multiple parameters potentially amenable to clinical manipulation towards treatment efficacy. Specifically, we evaluated the impact of intensifying in vivoantigenic re-stimulation with a viral vaccine, cell dose and differentiation status, and duration and type of cytokine support following cell transfer. Collectively, these findings provide for the rationale design of ACT-based immunotherapy trials.