IL-7 and IL-21 are superior to IL-2 and IL-15 in promoting human T cell–mediated rejection of systemic lymphoma in immunodeficient mice

The immune system can recognize and eliminate some cancers in mice and human subjects.¹  Within the tumor microenvironment, however, tumor-infiltrating lymphocytes may be rendered incapable of rejecting tumor cells and/or remain insufficient in number. Several tumor escape mechanisms, including down-regulation of immunogenic tumor antigen, major histocompatibility complex class I, or antigen-processing machinery, as well as expression of a variety of immunosuppressive factors, such as interleukin-10 (IL-10), tumor growth factor-β, and programmed death ligand-1 (PD-L1), have been documented.²  Furthermore, suppressive immune cells including arginase- and indoleamine-2,3-deoxygenase–expressing myeloid suppressor cells³  and CD4⁺CD25⁺Foxp3⁺ regulatory T cells⁴  can be recruited to, or induced by, the tumor microenvironment. Molecules that sustain or enhance the function of tumor-reactive T cells, or that counteract inhibitory factors, are therefore likely to be beneficial for T cell–mediated tumor rejection.

Numerous cytokines profoundly affect T-cell development, differentiation, and homeostasis. IL-2, IL-7, IL-15, and IL-21 are members of a cytokine family whose heteromeric receptors share the common γ chain (γ_c). Each cytokine has been described as a T-cell growth factor⁵  and each has been used to augment the T-cell antitumor immune response,^6-10  most notably IL-2.^11,12  At a finer level, however, each cytokine possesses nonredundant functions that differentially shape T-cell responses: IL-2 plays a crucial role in the development and maintenance of regulatory T cells, a function not shared among other γ_c-cytokines¹³ ; IL-7 mediates homeostasis of naive and memory CD4⁺ and CD8⁺ T cells; and IL-15 is essential for maintenance of the CD8⁺ memory T-cell subset.¹⁴  The role of IL-21 in T cell–mediated tumor immunity is less defined, with reports demonstrating its antitumor efficacy as a single agent,⁹  or only in synergistic combination with IL-15.¹⁰ 

Despite extensive preclinical studies, and growing clinical data, a direct, multiparameter investigational comparison of the in vivo contributions of γ_c-cytokines to antitumor human T-cell biology has not been reported. Adoptive transfer approaches are well suited to compare the function and potency of a constant number of T cells exposed to different γ_c-cytokines.

In the following studies, we have taken a genetic approach to analyze and compare the antitumor properties of IL-2, IL-7, IL-15, and IL-21. Using bicistronic retroviral vectors, we engineered human primary T cells to recognize the CD19 antigen and overexpress a γ_c-cytokine, and compared their ability to eradicate disseminated lymphomas in nonobese diabetic/severe combined immunodeficient (NOD/SCID)/γ_c^null mice. We analyzed the contribution of each cytokine to T-cell proliferation, survival, effector function, and phenotype. Our studies reveal that all γ_c-cytokines augment tumor rejection, but through singularly divergent mechanisms.

Plasmids encoding the SFG oncoretroviral vector were prepared using standard molecular biology techniques. Synthesis of 19z1, Pz1, and ΔLNGFR have been described.^7,15,16  Human IL-2, IL-15, and IL-21 cDNA was prepared from phytohemagglutinin (PHA)–stimulated human peripheral blood leukocytes (PBLs); human IL-7 cDNA was obtained from Invivogen. To increase IL-15 secretion, the N-terminal signal peptide of IL-15 was replaced with the human IL-2 signal peptide and a FLAG tag was added to the C-terminus.^17,18  The P2A bicistronic element was prepared from purified oligodeoxynucleotides (Sigma-Genosys).¹⁹  PG13 oncoretroviral vector producer cell lines were prepared from plasmid-transfected H29 cell supernatant.

Construction of Raji-eGFP/Firefly Luciferase (Raji-GL), EL4-CD19, EL4-PSMA, and the artificial antigen-presenting cells (AAPCs) NIH3T3-CD19-CD80 and NIH3T3-PSMA-CD80 have been described.^7,15 

Peripheral blood was obtained from healthy donors after informed consent under a protocol approved by the Memorial Sloan-Kettering Cancer Center (MSKCC) institutional review board in accordance with the Declaration of Helsinki. PBLs were isolated by density gradient centrifugation and activated with PHA for 48 hours. Activated T cells were then isolated by negative selection using a pan T-cell isolation kit (Miltenyi Biotec) and were transduced on 2 consecutive days by centrifugation in retronectin-coated (Takara), oncoretroviral vector–bound plates. Seven days after PHA stimulation, T-cell cultures were exposed to irradiated AAPCs. Exogenous cytokine of any type was not administered at any time unless explicitly stated.

Flow cytometry was performed on a BD LSRII. Anti–human granzyme A (CB9), interferon-γ (IFN-γ; B27), tumor necrosis factor-α (TNF-α; MAb11), IL-2 (MQ1-17H12), CD27 (L128), Bcl-2 (6C8), CCR7 (3D12), CCR7 isotype (R35-95), and CD19 (SJ25C1) were obtained from BD; anti–human CD3 (S4.1), CD4 (S3.5), CD8 (3B5), CD28 (10F3), CD45RA (MEM-56), LIVE/DEAD violet, 4′,6-diamidino-2-phenylindole, and goat anti–mouse immunoglobulin G (used for detection of 19z1 and Pz1) were obtained from Caltag/Invitrogen; anti–human CD45RO (UCHL1) was obtained from Beckman Coulter; anti–human CD8 (RPA-T8), CD62L (DREG-56), Foxp3 (236A/E7), and Foxp3 isotype were obtained from eBioscience.

Mice were cared for in accordance with the institutional guidelines of MSKCC. Male 6- to 12-week-old NOD/SCID/γ_c^null mice (The Jackson Laboratory) were inoculated intravenously with 5 × 10⁵ Raji-GL. Six days later, 2 × 10⁵ 19z1⁺CD8⁺ T cells were infused intravenously; cell dose was based on the percentage CD3h⁺19z1⁺CD8⁺CD4⁻. T cells were injected 6 days after AAPC stimulation. To quantify tumor burden (measured in photons/s/cm²/steradian), a constant region of interest was made that encompassed the entire mouse except for the tail. Living Image software (Xenogen) was used to acquire and quantify bioluminescent imaging (BLI) datasets as described.²⁰  Mice were considered at the experimental end point if they developed hind-limb paralysis, were otherwise moribund, or had a tumor burden greater than 2 × 10⁷ photons/s/cm²/sr.

Enzyme-linked immunosorbent assay (ELISA) kits were used to quantify cytokine production: IL-2 and IL-7 (R&D), IL-15 (BD), and IL-21 (eBioscience). For blocking assays, anti–human IL-2Rα (20 μg/mL, clone 22722; R&D) and anti–human IL-7Rα (10 μg/mL, clone R34-34; IMGENEX) were added to cultures. For accumulation experiments involving cytokine titration, 19z1-ΔLNGFR cells were treated with IL-2 (Chiron), IL-7 (BD), IL-15 (R&D), or IL-21 (Biosource) as indicated. Fold increase in T cells was calculated using viable cell number (Guava) and percentage of 19z1⁺ T cells. Chromium-release assays were performed as described.¹⁵  For proliferation and effector cytokine production assays, 19z1⁺ T cells 7 days after AAPC stimulation were purified by positive selection with the anti-19z1 monoclonal antibody 19E3-phycoerythrin,²¹  a generous gift from Dr I. Rivière (MSKCC), and anti-phycoerythrin microbeads (Miltenyi Biotec). After purification, T cells were more than 99% 19z1⁺ (data not shown). For proliferation assays, purified 19z1⁺ T cells were labeled with carboxyfluorescein succinimidyl ester (CFSE; Molecular Probes) prior to stimulation with OKT3 (eBioscience) and anti-CD28 (BD). For effector cytokine production assays, purified 19z1⁺ T cells were likewise stimulated, but in the presence of brefeldin A (GolgiPlug; BD) and monensin (GolgiStop; BD).

The paired t test was used to analyze normalized Bcl-2 data. Tumor burden at specific time points was analyzed using the Wilcoxon rank sum test. Survival data were analyzed using the log-rank test. The unpaired t test was used to analyze all other data.

To establish conditions whereby human primary T cells would be continuously exposed to γ_c-cytokines in vitro and in vivo, we took a genetic approach based on the transduction of human peripheral blood T lymphocytes. Rapid generation of antitumor cytokine-secreting CD19-reactive human primary T cells was accomplished using bicistronic γ-retroviral vectors (Figure 1A) encoding 19z1, an major histocompatibility complex–independent chimeric antigen receptor (CAR) specific for CD19,⁷  and either IL-2, IL-7, IL-15, or IL-21. A doubly mutated human low-affinity nerve growth factor receptor (termed ΔLNGFR) served as control. Gene transfer into human primary T cells with the cytokine-encoding vectors was efficient (> 60%) and comparable between vectors (Figure 1B and Table 1). Significantly, the mean fluorescence intensity (MFI) of 19z1 in transduced cells was also similar between groups (Figure 1B), and T-cell cultures were more than 99% CD3⁺ (Table 1). Furthermore, although there was a trend in the IL-2– and IL-15–transduced groups toward an increase in the CD8⁺/CD4⁺ ratio, the difference was not significant, nor was it significantly different in the IL-7– and IL-21–transduced groups (Table 1).

To quantify vector-encoded cytokine output, supernatants of cytokine-transduced T cells were analyzed by ELISA (Figure 1C). IL-2–, IL-7–, IL-15–, and IL-21–transduced T cells accumulated their respective cytokine to 68, 37, 11, and 16 pg/10⁶ transduced cells/12 hours, respectively. The higher amount of vector-encoded IL-2, in comparison with IL-15, has been previously noted.²²  Addition of anti–IL-2Rα to IL-2–transduced cultures and anti–IL-7Rα to IL-7–transduced cultures increased cytokine accumulation (277 and 184 pg/10⁶ transduced cells/12 hours, respectively, Figure 1C), suggesting the presence of receptor-mediated cytokine uptake. Although blocking antibodies for IL-15Rα and IL-21R were not commercially available, it is possible that a similar phenomenon occurred in IL-15– and IL-21–transduced cultures, accounting for the smaller amount of cytokine observed. Knowing that it was not technically feasible to equalize the molar output for every cytokine, we sought to compare the effects of ectopic overexpression of each cytokine expressed from the same promoter.

To compare vector-encoded cytokine expression with exogenous cytokine treatment, 19z1-ΔLNGFR–transduced T cells were treated with titrated amounts of human IL-2, IL-7, IL-15, and IL-21, and the fold increase in 19z1⁺ T cells in response to weekly AAPC stimulation was monitored (Figure 1D). IL-2–transduced T cells accumulated to a level similar to ΔLNGFR-transduced T cells treated with 2 to 20 U/mL IL-2. IL-7–transduced T cells accumulated to a similar extent as ΔLNGFR-transduced T cells treated with either 2 ng/mL or 20 ng/mL IL-7, suggesting that in our system a threshold had been reached by both exogenous cytokine doses and IL-7 overexpression by which the accumulation capacity had been met. IL-15–transduced T cells accumulated to a similar extent as ΔLNGFR-transduced T cells treated with 0.1 ng/mL IL-15. Finally, IL-21–transduced T cells accumulated to a lesser extent than ΔLNGFR-transduced T cells treated with 1 ng/mL IL-21.

We compared the antitumor potency of γ_c-cytokine–transduced T cells in a xenogeneic adoptive transfer model in immunodeficient mice bearing systemic, established Raji human B-cell lymphoma. Previous animal studies from our laboratory used the SCID/bg mouse,^7,23,24  a strain devoid of T cells and B cells but retaining modest NK-cell activity. Because of the potential for T cell–derived γ_c-cytokines, particularly IL-15, to augment host NK cell function, we developed a mouse model using the more severely immunodeficient NOD/SCID/γ_c^null mouse. In a direct comparison, both Raji cells and 19z1⁺ T cells persisted to a greater extent in NOD/SCID/γ_c^null mice than in SCID/bg mice (supplemental Figure 1A, available on the Blood Web site; see the Supplemental Materials link at the top of the online article). In NOD/SCID/γ_c^null mice, the Raji tumor cells extensively infiltrated the bone marrow, lymph nodes, liver, spinal cord, and pituitary gland (supplemental Figure 1B-F). Importantly, treatment of Raji cells with exogenous cytokine had no effect on tumor growth kinetics (supplemental Figure 2A).

To determine the in vivo efficacy of γ_c-cytokine–transduced T cells, NOD/SCID/γ_c^null mice bearing established, systemic Raji tumors received 2 × 10⁵ 19z1⁺CD8⁺ T cells of each vector type. Tumor burden in individual mice was monitored biweekly using in vivo bioluminescent imaging (BLI) in the control groups and 5 treatment groups (Figure 2A). Control mice were treated with T cells transduced with analogous cytokine-encoded vectors, but with a CAR specific for prostate-specific membrane antigen (PSMA), termed Pz1 (supplemental Figure 3). In mice receiving no treatment or any PSMA-targeted T-cell treatment, all mice succumbed to disease with a median survival of 30 and 28 days, respectively (Figure 2B). In contrast, treatment with CD19-specific T cells coexpressing any or none of the 4 γ_c-cytokines significantly delayed tumor progression and induced more frequent long-term remissions. 19z1-IL7 and 19z1-IL21 mice had an overall superior survival benefit compared with 19z1-ΔLNGFR-mice, whereas 19z1-IL2– and 19z1-IL15–treated mice did not.

However, the imaging studies revealed dramatically distinct patterns in tumor progression. Mice receiving no treatment or unspecific treatment died with large liver and bone marrow tumor burden, as indicated by bioluminescent signal in the abdomen and lower limbs and confirmed by pathology (supplemental Figure 1B-F). In contrast, mice receiving CD19-specific T cells of any type became negative for bioluminescent signal in the limbs, suggesting successful tumor eradication in the bone marrow. Mice that went on to fail CD19-specific T-cell therapy almost uniformly died with progressively accumulating abdominal tumor burden due to failure to eradicate liver disease (Figure 2A and data not shown). Of note, outgrowth of hepatic lymphoma in treated mice was not due to loss of CD19 tumor antigen because tumors removed from mice at end point documented persisting CD19 expression (supplemental Figure 2B). Most importantly, we soon noticed that a substantial number of mice died with a low, near-background bioluminescent signal (< 10⁴ photons/s/cm²/sr), which was confirmed to reflect absence of residual tumor in all mice subjected to necropsy (n = 5, data not shown). This occurred in 10 (43%) of 23 19z1-IL2 mice, 13 (54%) of 24 19z1-IL7 mice, and 6 (26%) of 23 19z1-IL15 mice. In contrast, only 2 (9%) of 23 19z1-ΔLNGFR mice and 1 (4%) of 23 19z1-IL21 mice died with a low tumor burden. Death of these mice with low tumor burden was due to cytokine-exacerbated xenogeneic graft-versus-host disease (XGVHD), as confirmed by clinical and histologic examination (J.C.M. and M.S., unpublished observations).

Because most mice succumb from either massive tumor burden or from XGVHD without evidence of tumor, and because we could differentiate these modes of death by BLI, the antitumor activity of the different T-cell groups is best analyzed by plotting tumor-free survival rather than overall survival. This method assumed that mice dying with a tumor burden of less than 10⁴ photons/s/cm²/sr died of XGVHD, but after successful tumor eradication. As shown in Figure 2C, mice that received any of the 4 γ_c-cytokine–transduced T cells survived significantly longer than mice receiving 19z1-ΔLNGFR cells, with 19z1-IL7 and 19z1-IL21 inducing greater responses than 19z1-IL2 and 19z1-IL15.

To elucidate the mechanism underlying the differential in vivo potency of cytokine-transduced T cells, we investigated their preinfusion and postinfusion function and phenotype. The response of transduced T cells to antigenic stimulation was assessed in vitro by monitoring absolute T-cell expansion and accumulation after repeated exposure to CD19⁺ AAPCs in the absence of added cytokine (Figure 1D). After the first round of stimulation, all groups expanded, more so in the IL-2 and IL-7 groups than the ΔLNGFR, IL-15, and IL-21 groups (19z1-ΔLNGFR, 11-fold expansion; 19z1-IL2, 21-fold; 19z1-IL7, 18-fold; 19z1-IL15, 13-fold; 19z1-IL21, 12-fold). After the second round of CD19⁺ AAPC stimulation, IL-7–transduced T cells had accumulated to the greatest extent, followed by IL-2 (19z1-ΔLNGFR, 7-fold; 19z1-IL2, 32-fold; 19z1-IL7, 96-fold; 19z1-IL15, 10-fold; 19z1-IL21, 12-fold). This trend continued after the third AAPC stimulation except that 19z1-IL21–transduced T cells became less abundant than the 19z1-ΔLNGFR–transduced T cells, indicating that constitutive IL-21 overexpression may have an antiproliferative or proapoptotic effect on human primary T cells. Overall, these results indicate that IL-2, IL-7, and to a lesser extent, IL-15, but not IL-21, can augment antigen-induced accumulation of specific T cells. The accumulation pattern of CD4⁺ and CD8⁺ T cells within these groups paralleled the growth of CD3⁺ T cells (supplemental Figure 4).

We therefore measured the proliferative capacity of cytokine-transduced T cells. 19z1⁺ T cells were CFSE-labeled and activated by anti-CD3/anti-CD28 antibodies (Figure 3A). IL-7– and IL-21–transduced CD4⁺ and CD8⁺ T cells responded quickest as measured by CFSE dilution on day 3. Near complete CFSE dilution was achieved by day 8 for the IL-7–transduced CD4⁺ and CD8⁺ T cells and IL-21–transduced CD8⁺ T cells. In comparison, the CD4⁺ subset of the IL-2–, IL-15–, and IL-21–transduced T cells and the CD8⁺ subset of the IL-2– and IL-15–transduced T cells proliferated less, albeit more than the control 19z1-ΔLNGFR–transduced T cells. These results indicate that each cytokine supported mitogen-induced proliferation, but to different extents.

Because IL-21–transduced T cells exhibited marked early proliferation in response to stimulation, but accumulated very poorly upon repeated antigenic stimulation (Figure 1D), we examined expression of Bcl-2, a key negative regulator of lymphocyte apoptosis. Overexpression of IL-2, IL-7, and IL-15 in CD4⁺ T cells resulted in a 144%, 167%, and 130% increase in Bcl-2 MFI compared with ΔLNGFR-transduced CD4⁺ T cells (Figure 3B). A parallel but more extensive Bcl-2 up-regulation was observed for CD8⁺ T cells transduced with IL-2 (209%), IL-7 (249%), and IL-15 (227%) compared with ΔLNGFR-transduced CD8⁺ T cells. In contrast, Bcl-2 expression was lower in IL-21–transduced CD4⁺ T cells (79%); the same trend was observed for CD8⁺ T cells (77%), although this decrease did not reach statistical significance (P = .06). These data confirm the role of these γ_c-cytokines in modulating Bcl-2 expression by human T cells.^25-27 

To investigate the influence of ectopic γ_c-cytokine overexpression on T-cell effector function, we first quantified the cytolytic activity of cytokine-transduced T cells in chromium-release assays (Figure 4A). IL-2– and IL-15–transduced T cells lysed relevant targets significantly more efficiently than ΔLNGFR-transduced T cells at 1 and 3 E/T ratios, respectively, in agreement with a published report.²²  This enhancement was not observed in other cytokine-transduced groups.

We next determined whether expression of the cytolytic molecule granzyme A was affected by overexpression of γ_c-cytokines (Figure 4B). Overexpression of each cytokine significantly increased the percentage of granzyme A⁺ CD8⁺ T cells over ΔLNGFR-transduced CD8⁺ T cells. Likewise, IL-2 and IL-15 overexpression significantly increased the percentage of granzyme A⁺ CD4⁺ T cells over ΔLNGFR-transduced CD4⁺ T cells, whereas overexpression of IL-7 and IL-21 did not. The higher percentage of granzyme A⁺ IL-2– and IL-15–transduced T cells thus paralleled the functional cytolytic activity.

We also assessed the IFN-γ and TNF-α response of transgenic T cells to mitogenic restimulation (Figure 4C). In the CD4⁺ subset, IFN-γ was expressed in a significantly higher percentage of IL-2–, IL-7–, and IL-15–transduced T cells, but not IL-21–transduced T cells, compared with ΔLNGFR-transduced CD4⁺ T cells. In the CD8⁺ subset, a significantly higher percentage of IFN-γ–expressing T cells was found in the IL-7– and IL-15–transduced T cells compared with ΔLNGFR-transduced T cells, but not IL-2– or IL-21–overexpressing groups. The same trend was observed for TNF-α expression in both CD4⁺ and CD8⁺ T cells. These data indicate that IL-7 and IL-15 overexpression can significantly up-regulate in vitro IFN-γ and TNF-α expression in both CD4⁺ and CD8⁺ tumor-specific T cells.

The γ_c-cytokines influence the differentiation state of tumor-specific memory-phenotype (CD45RA⁻CD45RO⁺) T cells and their central memory (CCR7⁺CD62L⁺, T_CM) or effector memory (CCR7⁻CD62L^+/−, T_EM) subtypes.^8,28,29  To determine whether cytokine overexpression likewise affected T-cell differentiation, we first assayed for CD45RA and CD45RO expression. Seven days after AAPC stimulation, nearly all transduced T cells were CD45RA⁻CD45RO⁺ (supplemental Figure 5), indicating that cytokine transduction did not significantly affect in vitro formation of memory-phenotype T cells.

We next assessed the short-term generation of CD8⁺ T_CM and T_EM subsets (Figure 5). CCR7 expression was low in CD45RA⁻CD8⁺ ΔLNGFR-, IL-7–, IL-15–, and IL-21–transduced groups and negative in the IL-2–transduced group (Figure 5A). Interestingly, CCR7 was more highly expressed in the cocultured, untransduced (19z1⁻) subset of each group, indicating that the antigen-specific AAPC stimulation was at least partially responsible for CCR7 down-regulation. Compared with ΔLNGFR-transduced T cells, IL-15–transduced memory-phenotype CD8⁺ T cells lost expression of CD62L, whereas the IL-7–transduced group largely retained CD62L; no significant difference was observed between ΔLNGFR-transduced and IL-2– or IL-21–transduced groups (Figure 5B). Our data indicate that γ_c-cytokine overexpression did not promote in vitro formation of CD8⁺ T_CM cells; cytokine-transduced CD8⁺ T cells were phenotypically more similar to T_EM cells.

T_EM cells are not major producers of IL-2 compared with their T_CM counterparts.²⁸  To further verify that the cytokine-transduced T cells in our system are of the T_EM phenotype, we assessed their potential to express IL-2 (Figure 4C). Expectedly, most IL-2–transduced CD4⁺ and CD8⁺ T cells expressed IL-2 constitutively. Of the remaining groups, only the CD4⁺ IL-7–transduced T cells contained a higher percentage of IL-2⁺ T cells compared with ΔLNGFR-transduced T cells. These data help confirm the T_EM phenotype of γ_c-cytokine–overexpressing T cells.

Human CD8⁺ T_EM cells can be further subdivided according to their cell surface expression of the costimulatory molecules CD27 and CD28.^30,31  The subsets thus defined (Figure 5C) differ widely in their function, with T_EM1 (CD27⁺CD28⁺), T_EM2 (CD27⁺CD28⁻), and T_EM3 (CD27⁻CD28⁻) cells representing memory-like, intermediate effectors, and late effectors, respectively.³¹ 

We thus delineated the T_EM subsets of cytokine-transduced memory-phenotype CD8⁺ T cells by measuring expression of CD27 and CD28 (Figure 5C). In all groups, the majority of 19z1⁺CD8⁺ T cells were of the T_EM1 subtype. However, a significantly smaller percentage of T_EM1 cells existed in IL-2–, IL-7–, and IL-15–transduced groups compared with the ΔLNGFR-transduced group, but not in the IL-21–transduced group. The reduced percentage of T_EM1 cells translated into higher percentages of T_EM2 and T_EM3 populations, indicating a progressive differentiation of these T cells to a late effector phenotype. These data indicate that IL-21 overexpression preserved a limited differentiation CD27⁺CD28⁺ T_EM1 phenotype.

Of note, overexpression of any γ_c-cytokine did not influence the generation of CD4⁺Foxp3⁺ T cells (supplemental Figure 6).

Because in vitro data suggested that we had created T cells with the ability to proliferate and/or survive (Figure 3), we determined whether long-term surviving mice still harbored human T cells and whether their presence and phenotype correlated with efficacy. Surviving mice were killed at or after day 106 after tumor challenge (day 100 after adoptive transfer) and spleens were analyzed for the presence of human T cells (Figure 6). A definitive population of human T cells was detected in the spleens of 1 (25%) of 4 19z1-ΔLNGFR mice, 0 (0%) of 7 19z1-IL7 mice, 3 (75%) of 4 19z1-IL15 mice, and 4 (40%) of 10 19z1-IL21 mice (Figure 6A). Unfortunately, no 19z1-IL2 survivors were available for analysis. Human T cells were detected in the bone marrow in a similar proportion of mice (data not shown). T cells were predominantly or exclusively CD8⁺ except in the 19z1-IL21 mice in which prominent CD4⁺ populations existed in both spleen and bone marrow (Figure 6A and data not shown).

Splenocytes from mice with prominent human T-cell populations were further characterized. Human CD4⁺ and CD8⁺ T cells from mice in the 19z1-ΔLNGFR and 19z1-IL21 groups were predominantly CD45RA⁻ (Figure 6B), consistent with their phenotype at infusion (supplemental Figure 5). By contrast, CD8⁺ T cells from 19z1-IL15 mice were predominantly CD45RA⁺CCR7⁻CD62L^+/−, resembling a previously described population of CD45RA⁺ effector memory T cells (T_EMRA).³²  In addition, CD8⁺ T cells recovered from 19z1-IL15 mice contained a CCR7⁺CD62L⁺ population, implicating IL-15 in the formation of a CD45RA⁺ antigen-experienced cell type that possesses cell surface molecules resembling T_CM cells (T_CMRA).^32,33  Long-term persisting CD8⁺ T cells remained functional, as demonstrated by their ability to produce IFN-γ upon stimulation (data not shown).

γ_c-cytokines exert profound effects on tumor-specific T-cell proliferation, differentiation, and survival, and as such have considerable therapeutic potential. Although much is known about their respective functions in mice, their effect on human cells is less well known, especially in vivo, and comparative studies are lacking. We have therefore taken a genetic approach to study the influence of IL-2, IL-7, IL-15, and IL-21 on tumor-targeted human primary T cells in a xenogeneic tumor model in which a constant dose of CD8⁺ CD19-targeted T cells was administered to mice with established, systemic lymphoma. In our model, we demonstrate that each of these cytokines potently enhanced tumor eradication, with IL-7 and IL-21 demonstrating superiority to IL-2 and IL-15. Our combined in vitro and in vivo findings, summarized in Table 2, demonstrate distinct mechanisms of action.

Constitutive expression of IL-7 increased in vitro T-cell accumulation in response to repeated exposure to antigen through a combined increase in proliferation and T-cell survival. However, 19z1-IL7 T cells did not exhibit increased tumor lysis, suggesting that the enhanced antitumor activity was mediated primarily by a quantitative increase in tumor-reactive T cells, consistent with the known ability of IL-7 to maintain T-cell numbers.³⁴  Surprisingly however, no residual T cells were found in long-term survivors, which suggests that IL-7 is unable to sustain memory T cells in this model. This may be explained by the down-regulation of IL-7Rα expression due to chronic IL-7 exposure,³⁵  thus depriving memory T cells of this critical homeostatic signaling pathway.³⁴ 

IL-21 appeared to exert few if any effects on T-cell expansion in vitro, but displayed unexpected potency in vivo. Similar to IL-7, constitutive expression of IL-21 initially supported extensive proliferation and did not increase the cytolytic potential of T cells. The overall accumulation of cultured 19z1-IL21 T cells, however, was markedly less than observed for 19z1-IL7 T cells, consistent with decreased Bcl-2 expression. Yet, in vivo, 19z1-IL21 T cells were highly efficacious in promoting tumor eradication. In contrast to 19z1-IL7–treated mice, large numbers of T cells with a T_EM1 phenotype persisted in the bone marrow and spleen of 19z1-IL21–treated long-term survivors. In aggregate, these findings suggest that enhanced tumor rejection resulted from improved T-cell expansion rather than the up-regulation of effector function.

In contrast to IL-7 and IL-21, constitutive IL-15 expression increased effector functions (eg, higher CTL activity and IFN-γ expression), but did not substantially increase T-cell accumulation in vitro. However, this cytokine sustained a very significant population of long-term surviving T cells with T_CM phenotypic properties, which may contribute to their enhanced efficacy in addition to the increased effector function. Interestingly, these CD8⁺ T cells, which displayed a CD45RA⁻CCR7⁻CD62L^+/− phenotype at the time of adoptive transfer, reverted to a CD45RA⁺ phenotype, resembling T_EMRA cells, an antigen-experienced CD45RA⁺CCR7⁻CD62L⁻ population thought to derive from CD8⁺ T_CM cells.³²  A population of CD8⁺CD45RA⁺CCR7⁺CD62L⁺ T cells was also observed exclusively in 19z1-IL15 mice, which may represent T_CMRA cells, a cell type of yet unclear significance.^32,33  The in vivo transition of cells with a T_EM phenotype to cells resembling T_CM-related subtypes parallels a recent study demonstrating that T_CM-derived clones, despite displaying a T_EM phenotype at the time of transfer, persist and revert to a T_CM phenotype in vivo.³⁶  Our data suggest that IL-15 is sufficient to mediate this transition.

Although sharing a common ability to persist in long-term survivors, IL-15– and IL-21–transduced T cells were phenotypically dissimilar, establishing that (1) these closely related cytokines influence the persistence and differentiation of distinct antitumor T-cell subsets and (2) there is more than one T-cell phenotype capable of mediating effective antitumor responses. Whereas these 2 cytokines induced persisting T cells, IL-7 did not, and we were unable to ascertain the effect of IL-2 because of poor overall survival of 19z1-IL2 mice.

Recent data suggest that adoptive transfer of T cells with a less differentiated phenotype correlates with superior antitumor immunity.^37-40  Our ability to rapidly redirect polyclonal peripheral blood T cells by retroviral-mediated gene transfer of a CAR allowed us to minimize the culture period (13 days) prior to adoptive transfer and reduce the generation of terminally differentiated, exhausted T cells. This is in contrast to expanded tumor-infiltrating lymphocytes, which, although capable of inducing tumor regression in melanoma patients,⁴¹  may display reduced therapeutic potency because of shortened telomeres and loss of CD27 and CD28 expression.⁴⁰ 

A proposed mechanism for IL-21 efficacy is the preservation of less differentiated CD27⁺CD28⁺ T_EM1 cells, a population that has been compared with T_CM cells.³¹  Importantly, T_EM1 cells are the least differentiated, have the highest telomerase activity, and have the shortest replicative history,³¹  which may make them the preferred effector memory subset for adoptive transfer.³⁷  The inability of IL-21 to induce in vitro differentiation of T_EM cells is consistent with its previously demonstrated ability to preserve the naive phenotype of CD4⁺ and CD8⁺ T cells.^38,42  Our data are in agreement with the demonstrated ability of IL-2, IL-7, and IL-15, but not IL-21, to down-regulate CD28 expression on human CD8⁺ T cells.^43,44  Our findings are also in agreement with reports describing the ability of IL-2 and IL-15 to down-regulate CD27 expression to a greater extent than IL-7.^25,45,46  IL-21, in the context of IL-2 and IL-7 cotreatment, was recently described to maintain CD27 expression on CD8⁺ T cells.⁴⁷  The mechanism by which adoptively transferred T cells that retain a more naive phenotype may provide superior antitumor immunity is ill defined. More differentiated T cells may be less efficacious because of loss of costimulatory receptor expression and tissue homing capacity.³⁷ 

Our results provide novel insights into the merits of xenogeneic adoptive transfer models to analyze human T-cell memory phenotypes. Although such models are clearly limited in their ability to recapitulate many important facets of human T-cell biology, it is noteworthy that the phenotypes of T cells persisting 100 days after adoptive transfer in NOD/SCID/γ_c^null mice are consistent with observations made with genetically marked, persisting cells introduced in nonhuman primates³⁶  and consistent with the phenotype of IL-15 and IL-21 transgenic mice, in which memory-phenotype CD8⁺ T cells selectively accumulate.^18,48  Although the role of CD4⁺ T cells in syngeneic tumor rejection is well established, the role of this population in xenogeneic systems is uncertain.⁷ 

Finally, our findings have important implications for the investigation of γ_c-cytokines in cancer treatment. IL-2 is currently the only Food and Drug Administration–approved γ_c-cytokine, and comparative clinical studies would be difficult to implement. Our results point to the great promise of IL-7 to support the potency of adoptively transferred T cells. The effect was limited in our model, but restoration of IL-7Rα expression via retroviral-mediated gene transfer may be an option to prolong responsiveness to IL-7 treatment.⁴⁹  IL-15 and IL-21 seem better suited to promote long-term T-cell persistence under the present conditions, albeit through different mechanisms.

Our studies did not aim to validate the adoptive transfer of genetically modified T cells that constitutively express cytokines for direct clinical application. However, T cell–mediated local cytokine delivery may avoid, or at least reduce, the severe toxicity that often limits systemic cytokine administration by limiting cytokine delivery to the tumor microenvironment, draining lymph nodes, and other immunologically relevant locations. The clinical use of γ_c-cytokine overexpression may necessitate a vector-encoded suicide gene, as these T-cell growth factors may increase the risk of T-cell transformation either by themselves or by potentiating insertional mutagenesis.^18,25  As an alternative to constitutive expression, transgene-encoded cytokine expression could be made dependent on antigen-specific T-cell activation to restrict cytokine delivery to sites of T-cell activation.⁵⁰ 

We thank Dr Gabrielle A. Rizzuto for technical assistance and critical reading of the paper; Alberto del Rosario for technical assistance; and the MSKCC Laboratory of Comparative Pathology.

This work was supported by National Institutes of Health (NIH) grant CA59350, Mr William Goodwin and Mrs Alice Goodwin and the Commonwealth Foundation for Cancer Research and the Experimental Therapeutic Center of MSKCC, NIH MSTP grant GM07739, and a Cancer Research Institute Predoctoral Fellowship (J.C.M.).

National Institutes of Health

Contribution: J.C.M. designed experiments, performed research, analyzed data, and wrote the paper; and M.S. designed experiments, analyzed data, and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Michel Sadelain, Box 182, Memorial Sloan-Kettering Cancer Center, New York, NY 10065; e-mail: m-sadelain@ski.mskcc.org.