Anti–CTLA-4 Activates Intratumoral NK Cells and Combined with IL15/IL15Rα Complexes Enhances Tumor Control

Combination therapy with immune checkpoint inhibitors targeting PD-1 (i.e., nivolumab) and CTLA-4 (i.e., ipilimumab) for advanced melanoma is dramatically successful with up to a 60% response rate (1), and combination therapy with nivolumab and ipilimumab leads to higher response rates than nivolumab alone in patients with brain metastases (2). Likewise, clinical trials have shown efficacy of combination therapy with nivolumab and ipilimumab in a subset of patients with renal cell carcinoma and non–small cell lung cancer (NSCLC; refs. 3, 4). However, for a majority of patients across tumor types, immunotherapies are ineffective. Thus, a major goal in tumor immunology is to further understand the immunomodulation induced by these therapies. This will enable the development of novel therapeutic strategies to enhance responses and foster development of biomarkers to identify patients most likely to respond to specific therapies.

Immune checkpoint inhibitors are designed to prevent the interaction of checkpoint molecules with their ligands, and thus prevent immune inhibition. However, in addition to blocking activity, some antibodies can mediate effector functions when their Fc domain interacts with Fc receptors (FcR). In murine models, anti–CTLA-4 can lead to the death of cells expressing CTLA-4 on their cell surface in an Fc/FcR-dependent manner (5–10). Although CTLA-4 is largely intracellular, it is expressed on the surface of regulatory T cells (Treg) infiltrating murine tumor models (5–7), and anti–CTLA-4 has been shown to be able to mediate depletion of these intratumoral Tregs. In these models anti–CTLA-4–mediated depletion of intratumoral Tregs correlates with antitumor activity of anti–CTLA-4 (5–7). In B16 tumor–bearing mice treated with a GM-CSF–secreting vaccine, Treg depletion has been linked to FcRs on macrophages and nonclassical monocytes (5). However, GM-CSF may induce intratumoral accumulation and activation of myeloid-derived suppressor cells in addition to macrophages and nonclassical monocytes. Natural killer (NK) cells are also able to mediate the depletion of antibody opsonized cells in an FcR-dependent manner. Thus, additional studies are necessary to determine the effect of anti–CTLA-4 on intratumoral NK cells. Such studies may suggest combination strategies likely to enhance the efficacy of anti–CTLA-4 therapies.

In part, because Tregs are not reduced in the blood of patients receiving anti–CTLA-4 antibodies (11–13), it has been thought that antibody-dependent cell-mediated cytotoxicity (ADCC)-mediated depletion of Tregs is not a mechanism of action of anti–CTLA-4 in patients. However, in murine models, intratumoral Tregs have higher surface CTLA-4 (sCTLA-4) expression compared with peripheral Tregs, and anti–CTLA-4 specifically depletes intratumoral Tregs but not peripheral Tregs (5, 7, 13). It has been found that ipilimumab, a human IgG1 antibody, can induce ADCC of Tregs, and reduced intratumoral Tregs following ipilimumab treatment may correlate with efficacy in patients (12, 14). In addition, in patients with tumors with a high mutation burden, the response to ipilimumab is associated with the presence of the high-affinity CD16-V158F germline SNP (8). Despite these data the ability of ipilimumab to deplete intratumoral Tregs in patients remains controversial. Therefore, further investigations are required to determine whether Tregs infiltrating tumors in patients also express higher amounts of sCTLA-4, which would provide a rationale for how anti–CTLA-4 therapies specifically deplete intratumoral Tregs in patients.

Here we show that Tregs infiltrating patient-derived NSCLC tissue expressed more sCTLA-4 than Tregs from matched patient blood or intratumoral effector T cells. Similarly, in a majority of patient-derived melanoma tissue, intratumoral Tregs express more sCTLA-4 than effector T cells. Using the CT26 murine tumor model we found that anti–CTLA-4 therapies induced NK-cell activation and degranulation specifically within the tumor microenvironment. This NK-cell activation coincided with depletion of intratumoral Tregs. Furthermore, combination of anti–CTLA-4 plus IL15/IL15Rα complexes enhanced tumor control in comparison with either monotherapy. Consistently, patients who benefited from ipilimumab had higher intratumoral expression of the NK-cell marker, CD56.

Patient studies were conducted in accordance with ethical guidelines including the Declaration of Helsinki, The Belmont Report, and the U.S. Common Rule. Patient-derived tumor tissue specimens were collected in accordance to the Institutional Review Boards at HFGCC of Christiana Health Care System (NSCLC, Newark, DE) or The University of Pennsylvania (melanoma, Philadelphia, PA). Informed consent was received from participants prior to inclusion in the study.

Murine experiments were approved by the Institutional Animal Care and Use Committees of The Wistar Institute (Philadelphia, PA) or Dartmouth University (Hanover, NH). C57BL/6J mice were obtained from The Jackson Laboratories and BALB/C mice were obtained from Taconic, and kept under specific pathogen-free condition in The Wistar Institute's Animal Facility (Philadelphia, PA).

Unless otherwise indicated, 5 × 10⁵ CT26 tumor cells in 200 μL of PBS were subcutaneously transplanted into the right flank of mice using 27g needles. A total of 2.5 × 10⁵–5 × 10⁵ YUMM1.7 cells or 1 × 10⁶ B16-F1 cells were transplanted subcutaneously on day 0 where indicated. In experiments in which mice were treated with anti–CTLA-4, unless otherwise noted, mice were treated with 300 μg of the indicated anti–CTLA-4 clone or appropriate isotype (both from BioXCell) diluted in PBS, and mice were euthanized 5 days post-anti–CTLA-4 administration for analysis of intratumoral T cells. Where indicated, mice were injected intravenously with 200 μg of anti-FcγRIV (clone 9E9) 1 hour prior to injection of anti–CTLA-4. Where indicated, mice were treated intraperitoneally on days 6 and 8 with 200 μL IL15/IL15Rα complexes, containing 0.5 μg human IL15 (NCI biorepository) and 2.33 μg of soluble murine IL15Rα-Fc (R&D Systems/Thermo Fisher Scientific) per dose, and on days 9, 12, 15, and 17 with 200 μg anti–CTLA-4 (clone 9D9). After treatment initiation, tumor size was measured every 2 to 4 days and mice were euthanized when tumors reached 12 × 12 mm². In experiments in which tumor-bearing mice were euthanized at a set timepoint, tumors were weighed upon euthanasia.

Tyr::CreER⁺Braf^CA/+Pten^fl/fl [Braf^tm1Mmcm Pten^tm1Hwu Tg(Tyr-cre/ERT2)13Bos/BosJ] mice were kindly provided by Marcus Bosenberg (Yale University, New Haven, CT), and bred at Dartmouth University onto a C57BL/6 background, with >98% purity confirmed by congenic testing (DartMouse). For induction of autochthonous tumors, 80 μg of 4-hydroxy-tamoxifen (Sigma) in DMSO was intradermally injected in the right flanks of 4-week-old mice. Tumors and skin samples from age-matched mice were harvested 5 weeks later.

Tumors were extracted, minced, and digested in 5 mL HBSS with 0.5% FBS and 1% HEPES containing 1 mg/mL collagenase type-I from Clostridium histolyticum (Sigma-Aldrich) for 30 minutes at 37°C. Cells were then washed and filtered through 70-μm cell strainers.

Autochthonous tumors and murine skin samples were harvested, minced, and digested for 45 minutes at 37°C in 2 mL HBSS containing 7 mg/mL collagenase D (Roche) and 200 μg/mL DNase I (Roche), using magnetic bar stirring at 300 RPM. Tissue fragments were mechanically dissociated through a 40-μm nylon mesh filter. Samples were washed in RPMI1640 media containing 10% FBS and 2 mmol/L EDTA.

For sCTLA-4 staining, cells were processed as before, stained with antibodies to surface antigens, including CTLA-4 or isotype control, and then two rounds of FASER-PE (Miltenyi Biotec) staining were performed according to the manufacturer's instructions. For mouse FcR staining, cell suspensions were incubated with anti-mouse FcγRIV (CD16.2, clone 9E9, BioLegend) for 20 minutes at 4°C, washed twice, incubated with anti-mouse CD16/32 (clone 2.4G2, BD Bioscience) for 10 minutes at 4°C, then stained with all the other surface antibodies. Fixation and intracellular staining was accomplished with eBioscience FOXP3 Transcription Factor Staining Buffer Set.

Patient-derived tumor tissues were minced into pieces approximately 2 × 2 mm and processed with human Tumor Dissociation Kit (Miltenyi Biotec) according to the manufacturer's instructions. Red blood cell lysis buffer (eBioscience) was used to remove red blood cells from blood specimens. FcRs were blocked with FcR Blocking Reagent (Miltenyi Biotec), and then cells were stained similarly to murine cells. Patient-derived tissues were processed for flow cytometry the day of excision.

Samples were analyzed using LSR IIs (BD Bioscience) equipped with four lasers, a CELESTA (BD Bioscience) flow cytometer equipped with three lasers, or a Gallios Flow Cytometer (Beckman Coulter). Mean fluorescence intensity (MFI) indicates geometric mean of the indicated fluorescence. Data were analyzed using FlowJo versions 9 or 10 (TreeStar).

IL15/IL15Rα complexes were generated by combining 0.5 μg human IL15 (NCI biorepository) with 2.33 μg of soluble murine IL15Rα-Fc (R&D Systems/Thermo Fisher Scientific) per dose for 30 minutes at 37°C. PBS was used to quantum satis (q.s.) IL15 complexes to 200 μL per dose; doses were aliquoted and stored at −80°C. Batches of IL15 complexes were validated once by intraperitoneally injecting a mouse with IL15 complexes containing 0.5 μg human IL15 (NCI biorepository) and 2.33 μg of soluble murine IL15Rα-Fc (R&D Systems/Thermo Fisher Scientific) per dose on days 1, 3, and 5. On day 7, spleens from mice treated with IL15 complexes were then analyzed for size, NK-cell frequency, and frequency of activated CD8⁺ T cells compared with control spleen.

Relative expression of the NK-cell genes NCAM1 (CD56), FCGR3A, NCR1, NCR2, and KLRK1, and the T cell and myeloid genes CD3D, CD8A, CD4, CD14, CD74, CD163, and CX3CR1 in melanoma tissue from patients who received benefit or little to no benefit from anti–CTLA-4 published by Snyder and colleagues (2014) or Van Allen and colleagues (2015) were compared. NCAM1 expression in melanoma tissue from patients who received benefit or little to no benefit from anti–CTLA-4 as published by Snyder and colleagues (2014) was chosen for further analysis. Statistical difference in expression of NCAM1 was only found when analysis was restricted to samples from cutaneous melanoma.

Prism 7 (GraphPad) was used for statistical data analysis. To determine statistical significance, paired or two-group Student t test was used as indicated. For multiple groups of data analysis, ANOVA with Tukey post-hoc multiple comparison was used to determine statistical significance between two groups. For comparisons of multiple groups to a single group, Dunnett post-hoc multiple comparisons test was used to determine statistical significance. For data that did not follow normal distribution, Wilcoxon rank-sum test was used to determine statistical significance. To determine if indicated treatments induced significant difference in tumor progression, the velocities of tumor growth were compared between groups using a linear mixed effects model with the random effect at mouse level. A likelihood ratio testing nested models (with versus without the interaction term of treatment with follow-up days) was used to examine if tumor growth slopes are significantly different between treatments in overall. P < 0.05 was considered statistically significant.

As expected, administration of anti–CTLA-4 clone 9H10, a Syrian hamster IgG, to mice bearing CT26 tumors, an anti–CTLA4–responsive murine colorectal tumor model, led to depletion of Tregs specifically within the tumor microenvironment (Fig. 1A). Anti–CTLA-4 clone 9H10 has been shown to bind to FcγRIV, and mediate depletion of Tregs infiltrating B16 tumors in an FcγRIV-dependent manner (5). This and other data suggest that anti–CTLA-4–induced depletion of intratumoral Tregs may be mediated by nonclassical monocytes (5, 14). Furthermore, we found that pretreatment with an antibody to FcγRIV (clone 9E9), which has been reported to be a blocking antibody, prevented anti–CTLA-4–mediated depletion of Tregs infiltrating CT26 tumors (Fig. 1B). We note, however, that to our knowledge, the ability of anti-FcγRIV clone 9E9 to deplete FcγRIV-expressing cells has not been investigated, and thus a possibility is that administration of FcγRIV depleted FcγRIV⁺ cells, which may also express other FcRs. Nonetheless, our data further support the now prevailing evidence that anti–CTLA-4 clone 9H10 depletes intratumoral Tregs in an FcR-dependent manner.

NK cells also express FcRs and are able to mediate FcR-dependent functions of antibodies. Thus, we investigated the effect of anti–CTLA-4 treatment on NK cells. We found that, 5 days postadministration of a single dose of anti–CTLA-4 clone 9H10, an increased frequency of intratumoral NK cells were activated, as measured by NK-cell expression of CD69 and SCA1 (Fig. 1C and D). However, NK-cell activation appeared to be limited to the tumor microenvironment, as the frequency of activated NK cells was not increased outside of the tumor microenvironment (Fig. 1C). Thus, anti–CTLA-4 led to NK-cell activation specifically within the tumor microenvironment.

As the binding of the Fc region of antibodies to activating FcRs on NK cells can lead to NK-cell activation, it is possible that anti–CTLA-4 bound to intratumoral Tregs activates NK cells to kill Tregs opsonized by anti–CTLA-4. In agreement with this we found that 3 days postadministration of a single dose of anti–CTLA-4 clone 9H10, a timepoint that correlates with Treg depletion (Fig. 1A; refs. 5, 6), an increased frequency of intratumoral NK cells displayed CD107a on their cell surface directly ex vivo (Fig. 1E; Supplementary Fig. S1A). Surface CD107a expression indicates that these NK cells have recently degranulated.

Although expression of FcγRIV on NK cells in the periphery is rare, a fraction of intratumoral NK cells in B6 mice do indeed express FcγRIV (5). Similarly, we found that although peripheral NK cells in BALB/c mice expressed little to no FcγRIV, approximately 20% of NK cells infiltrating CT26 tumors in BALB/c mice expressed FcγRIV (Fig. 1F and G). Furthermore, we found that FcγRIV⁺ NK cells were present within tumors at a similar frequency to FcγRIV⁺ neutrophils and monocytes (Fig. 1H). However, we note that the per cell expression of FcγRIV was significantly higher on intratumoral FcγRIV⁺ neutrophils and monocytes than FcγRIV⁺ NK cells (Fig. 1I). Nonetheless, the expression of FcγRIV⁺ was increased on intratumoral FcγRIV⁺ NK cells compared with splenic NK cells (Fig. 1J). Furthermore, consistent with the dependence of anti–CTLA-4 clone 9H10 on FcγRIV for intratumoral Treg depletion, surface CD107a⁺ NK cells were largely FcγRIV⁺ (Fig. 1K).

Anti–CTLA-4 clone 9H10 is a hamster anti-mouse IgG and thus may bind to FcRs differently than mouse anti-mouse IgG. Thus, to determine whether NK-cell activation following anti–CTLA-4 administration is specific to anti–CTLA-4 clone 9H10 or may be broadly applicable across anti–CTLA-4 therapies able to mediate ADCC, we investigated NK-cell phenotype following administration of anti–CTLA-4 clone 9D9, a murine IgG2b antibody. Murine antibodies of the IgG2b isotype are known to be able to induce NK-cell–mediated ADCC. As with anti–CTLA-4 clone 9H10, anti–CTLA-4 clone 9D9 led to NK-cell degranulation (Fig. 1L). Consistent with anti–CTLA-4–inducing Treg depletion specifically within the tumor microenvironment, the increase in NK-cell degranulation was specific to the tumor microenvironment (Fig. 1L).

NK cells are able to express CTLA-4 when cultured in the presence of IL2 (15). Thus, a possibility was that anti–CTLA-4 activated NK cells by blocking CTLA-4 expressed by NK cells. To investigate this possibility, we used flow cytometry to assess surface and intracellular CTLA-4 expression by NK cells infiltrating CT26 tumors on day 18. As expected, we found that intratumoral Tregs expressed sCTLA-4; however, little to no CTLA-4 expression by NK cells was detected (Supplementary Fig. S1B). From these results we concluded it is unlikely that anti–CTLA-4 activated NK cells by relieving a “brake” imposed on NK cells by their own expression of CTLA-4.

The ability of anti–CTLA-4 to specifically mediate intratumoral Treg depletion (Fig. 1A) is consistent with intratumoral Tregs expressing more sCTLA-4 than peripheral Tregs (Supplementary Fig. S2A), and this has previously been shown of Tregs infiltrating B16-BL6 and CT26 tumor models (5, 6). In agreement with this, anti–CTLA-4 therapy leads to depletion of Tregs with the highest CTLA-4 expression (5, 6). However, activated effector T cells can also have CTLA-4 expression, and it has previously been found that CD4⁺ effector T cells responding to chronic stimulation can further upregulate CTLA-4 (16). Furthermore, CTLA-4 surface retention can be enhanced by stimulation (17). In agreement with this, we found that sCTLA-4⁺ Tregs and conventional CD4⁺ T cells infiltrating well-established YUMM1.7 tumors, a transplanted Braf^V600E/Pten^−/−/Cdkn2a^−/− melanoma model (18), expressed similar amounts of sCTLA-4 (Supplementary Fig. S2B). However, Tregs expressed significantly greater amounts of sCTLA-4 compared with other T-cell subsets infiltrating autochthonous melanomas in a conditionally induced Braf^V600E/Pten^−/− melanoma mouse model (Supplementary Fig. S2C). Using this model we also found that melanoma infiltrating Tregs expressed more sCTLA-4 than peripheral or skin Tregs (Supplementary Fig. S2D). Nonetheless, these data demonstrate that in some tumor microenvironments, a portion of effector T cells may express sCTLA-4 at amounts similar to Tregs. Thus, as anti–CTLA-4 therapies that mediate ADCC-induced depletion of cells with the highest sCTLA-4 expression (5, 6), it was relevant to determine the relative sCTLA-4 expression by T-cell subsets infiltrating patient-derived tumor tissue.

Tregs infiltrating previously untreated patient–derived NSCLC tissue expressed higher amounts of intracellular CTLA-4 (IC CTLA-4) than other intratumoral T-cell subsets (Fig. 2A and B; ref. 8). The majority of CTLA-4 is intracellular (17, 19), and little is known about the expression of sCTLA-4 on T-cell subsets infiltrating patient-derived tumor tissue directly ex vivo. Using flow cytometry, we found that although nearly all Tregs infiltrating patient-derived NSCLC tissue expressed IC CTLA-4 (Fig. 2A), only a fraction (∼10%–65%) of intratumoral Tregs expressed sCTLA-4 (Fig. 2C). sCTLA-4⁺ Tregs expressed more IC CTLA-4 and FOXP3 compared with sCTLA-4⁻ Tregs (Fig. 2D and E). Nonetheless, a higher frequency of Tregs than CD8⁺ or CD4⁺ effector T cells infiltrating patient-derived NSCLC tissue expressed sCTLA-4 (Fig. 2C). In nearly all NSCLC tissues analyzed, sCTLA-4⁺ Tregs expressed more sCTLA-4 on a per cell basis than the sCTLA-4⁺ intratumoral CD8⁺ or CD4⁺ effector T cells from the same tumor tissue (Fig. 2F).

Tregs infiltrating patient-derived NSCLC tissue expressed more intracellular CTLA-4 than Tregs from matched patient blood (Fig. 3A and B; ref. 8). However, even in intratumoral Tregs the majority of CTLA-4 was intracellular. Thus, we compared sCTLA-4 expression on intratumoral and peripheral Tregs to better understand whether, similar to what has been found in murine models, Tregs infiltrating patient-derived tumor tissue express more sCTLA-4 than peripheral Tregs. These results could thus suggest, whether anti–CTLA-4 is likely to specifically target intratumoral Tregs in patients. Comparing the data from Fig. 2 to costained circulating Tregs from matched patient blood, we found that both the frequency of sCTLA-4⁺ Tregs and the intensity of sCTLA-4 expression were higher on Tregs infiltrating patient-derived NSCLC tissue compared with circulating Tregs from matched patient blood (Fig. 3C and D).

To begin to understand factors that contribute to higher sCTLA-4 on intratumoral compared with circulating Tregs, we evaluated IC CTLA-4 and FOXP3 in sCTLA-4⁺ and sCTLA-4⁻ Tregs infiltrating the tumor and in circulation. Although sCTLA-4⁺–circulating Tregs expressed more IC CTLA-4 than sCTLA-4⁻ circulating Tregs (Fig. 3E), they expressed less IC CTLA-4 compared with intratumoral sCTLA-4⁺ Tregs (Fig. 3F). Likewise, whereas sCTLA-4⁺–circulating Tregs expressed more FOXP3 than sCTLA-4⁻ circulating Tregs (Fig. 3G), these Tregs had lower FOXP3 than intratumoral sCTLA-4⁺ Tregs (Fig. 3H). These data suggest that the amount of sCTLA-4 on sCTLA-4⁺ Tregs reflects the cell's microenvironment, the total expression of CTLA-4, and also associates with the amount of FOXP3 expression.

Using flow cytometry, we found that in a majority of patient-derived melanoma tissue a greater frequency of Tregs expressed sCTLA-4 than conventional CD4⁺ T cells from the same tissue (Fig. 4A). Furthermore, as with Tregs infiltrating patient-derived NSCLC tissue, sCTLA-4⁺ Tregs infiltrating patient-derived melanoma tissue expressed significantly more sCTLA-4 on a per cell basis compared with sCTLA-4⁺ CD8⁺ or CD4⁺ effector T cells from the same tumor tissue (Fig. 4B). As for Tregs infiltrating NSCLC, sCTLA-4⁺ Tregs infiltrating patient-derived melanoma tissue expressed more FOXP3 and IC CTLA-4 than sCTLA-4⁻ Tregs from the same samples (Fig. 4C and D).

Productive depletion of intratumoral Tregs may enhance the efficacy of anti–CTLA-4 in patients. In a murine model we found that depletion of intratumoral Tregs following administration of anti–CTLA-4 coincided with activation and degranulation of NK cells. Thus, a possibility was that patients with higher frequencies of intratumoral NK cells may be more likely to benefit from anti–CTLA-4 therapies. In support of this model, analysis of available RNA-seq data (20) revealed that higher intratumoral CD56 (NCAM1) expression was associated with benefit from ipilimumab in patients with cutaneous melanoma (Fig. 4E).

Anti–CTLA-4 activated intratumoral NK cells coincident to intratumoral Treg depletion in the CT26 model (Fig. 1A–E), and intratumoral expression of the NK-cell marker, CD56, was increased in patients with cutaneous melanoma who benefited from ipilimumab. Thus, a possibility was that therapies that enhance intratumoral NK cells may synergize with anti–CTLA4 therapy. To test this possibility, mice bearing CT26 tumors were treated with anti–CTLA-4 (clone 9D9), IL15/IL15Rα complexes (containing 0.5 μg human IL15 and 2.33 μg of soluble murine IL15Rα-Fc per dose; ref. 21), which have been shown to increase intratumoral NK cells (22), or combination therapy. As expected, anti–CTLA-4 monotherapy moderately limited tumor progression, but, in this model, monotherapy with IL15/IL15Rα complexes was ineffective. However, combination therapy with anti–CTLA-4 plus IL15/IL15Rα complexes enhanced tumor control compared with either monotherapy (Fig. 4F).

We showed that Tregs infiltrating murine tumors, as well as patient-derived tumors preferentially expressed sCTLA-4. Consistently, anti–CTLA-4 specifically depleted intratumoral Tregs in an FcR-dependent manner. We further found that anti–CTLA-4 therapies induced NK-cell activation and degranulation particularly in the tumor microenvironment, and this corresponded with anti–CTLA-4–mediated depletion of intratumoral Tregs. Moreover, increased intratumoral CD56 expression was associated with benefit from ipilimumab, and in a murine model anti–CTLA-4 plus IL15/IL15Rα complexes enhanced tumor control compared with either monotherapy. Thus enhanced intratumoral NK-cell activity may contribute to the efficacy of anti–CTLA-4 therapy.

Our data suggest that anti–CTLA-4 opsonizing intratumoral Tregs leads to activation of FcR⁺ cells, including NK cells, which can then mediate ADCC of these anti–CTLA-4 opsonized Tregs. Thus, NK cells may contribute to anti–CTLA-4–mediated depletion of intratumoral Tregs (Supplementary Fig. S3). In agreement with this, NK cells have been shown to be able to mediate ADCC of CTLA-4⁺ Tregs in the presence of a human CTLA-4 antibody (23). Alternatively, intratumoral Tregs may inhibit NK cells, and anti–CTLA-4–mediated depletion of intratumoral Tregs may indirectly allow for NK-cell activation and degranulation when these NK cells target tumor cells. This activation of NK cells following Treg depletion may contribute to the durable responses induced by anti–CTLA-4.

The cytokine IL15 supports NK-cell proliferation, survival, and cytolytic activity, and administration of IL15/IL15Rα complexes has been shown to enhance intratumoral NK-cell frequencies (22). Thus, although we do not rule out the likelihood that the effect of IL15/IL15Ra complexes on CD8⁺ T cells contributes to the increased tumor control seen in mice cotreated with IL15 super agonists and anti–CTLA-4, our data suggest that increased intratumoral NK-cell frequencies and activity may be important for the synergy between IL15 complexes and anti–CTLA-4 in controlling tumor progression. Consistently, increased intratumoral amounts of the NK-cell marker CD56 was associated with benefit from ipilimumab in patients with cutaneous melanoma, further suggesting that increased intratumoral NK cells enhance anti–CTLA-4 efficacy.

The ability of ipilimumab to deplete intratumoral Tregs in patients remains controversial. Although a decrease in intratumoral Tregs following ipilimumab treatment may correlate with efficacy in patients (12, 14), circulating Tregs are not decreased following administration of anti–CTLA-4 (11–13). We found that surface CTLA-4 expression was enhanced on Tregs infiltrating patient-derived tumor tissue compared with circulating Tregs from matched patient blood. These data provide a potential rationale for why anti–CTLA-4 therapy has been found to specifically reduce intratumoral but not circulating Tregs in patients. Thus, anti–CTLA-4 therapies may preferentially deplete intratumoral Tregs in patients, and specifically intratumoral Tregs with enhanced expression of not only CTLA-4 but also FOXP3.

In conclusion, we showed that anti–CTLA-4 therapy induced activation and degranulation of intratumoral NK cells and that combination therapy with IL15/IL15Rα complexes enhanced tumor control. In addition, increased intratumoral amounts of the NK-cell marker CD56 was associated with benefit from ipilimumab in patients with advanced cutaneous melanoma. Nonetheless, it remains to be determined whether NK cells contribute to anti–CTLA-4–mediated Treg depletion and the enhanced tumor control induced by combination therapy with anti–CTLA-4 plus IL15/IL15Rα complexes. Thus, future research should investigate the role of NK cells in anti–CTLA-4 efficacy and whether increased intratumoral NK cell numbers enhance anti–CTLA-4 efficacy.

R.K. Amaravadi has ownership interest (including stock, patents, etc.) in Pinpoint Therapeutics and Immunaccel and is a consultant/advisory board member for Sprint Biosciences. M.J. Turk has ownership interest (including stock, patents, etc.) in Celdara Medical, LLC. L.M. Schuchter has provided expert testimony for DLA piper. T.C. Mitchell is a consultant/advisory board member for Merck, BMS, Incyte, Aduro, and Regeneron. E.L. Stone is director of immune-oncology at and has ownership interest (including stock, patents, etc.) in GigaGen, Inc. M.P. Rubinstein reports receiving commercial research funding from Altor Bioscience and XOMA, and has ownership interest (including stock, patents, etc.) in XOMA. No potential conflicts of interest were disclosed by the other authors.

Conception and design: E. Sanseviero, M.P. Rubinstein, T.C. Mitchell, E.L. Stone

Development of methodology: E. Sanseviero, E.M. O'Brien, M.P. Rubinstein, T.C. Mitchell

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Sanseviero, E.M. O'Brien, J.R. Karras, T.B. Shabaneh, W. Xu, C. Zheng, X. Xu, G.C. Karakousis, R.K. Amaravadi, B. Nam, M.J. Turk, L.M. Schuchter, T.C. Mitchell, E.L. Stone

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Sanseviero, E.M. O'Brien, T.B. Shabaneh, B.A. Aksoy, X. Yin, R.K. Amaravadi, M.P. Rubinstein, Q. Liu, E.L. Stone

Writing, review, and/or revision of the manuscript: E. Sanseviero, E.M. O'Brien, J.R. Karras, T.B. Shabaneh, W. Xu, X. Xu, G.C. Karakousis, R.K. Amaravadi, B. Nam, M.J. Turk, T.C. Mitchell, Q. Liu, E.L. Stone

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.B. Shabaneh, W. Xu, C. Zheng, B. Nam, J. Hammerbacher, M.P. Rubinstein

Study supervision: E.L. Stone

The authors are grateful to Sophia Simon, Autumn Powell, Keira McAfee, and Claire Cohen (Wistar) for genotyping and mouse line maintenance assistance. The authors thank Patricia Swanson and Lori Huelsenbeck-Dill (HFGCC-Christiana Health Care System) for assistance with patient recruitment. The authors are especially grateful for scientific discussions with the Stone Lab and members of The Wistar Institute. This research was sponsored by The Wistar Institute Cancer Center and an American Cancer Society-IRG grant. The research was funded by a Bristol-Myers Squibb-MRA Young Investigator Award (to E.L. Stone), a Wistar/UPenn Specialized Program of Research Excellence grant in Skin Cancers Developmental Research Award (to E.L. Stone), and a grant from BNY Mellon Wealth Management through The Martha W. Rogers Charitable Trust (to E.L. Stone). E.L. Stone was funded in part by the Emerson Collective, 1K01DK095008 (NIDDK), and a PA CURE grant. J.R. Karras was funded in part by a PENNPort IRACDA postdoctoral fellowship. E. Sanseviero was funded in part by a Fellowship from Istituto Pasteur Italia–Fondazione Cenci Bolognetti. Core Facilities were supported by Cancer Center Support Grant CA010815 to The Wistar Institute. B.A. Aksoy and J. Hammerbacher are supported in part by MUSC. M.J. Turk and T.B. Shabaneh are supported by NIH grants R21CA209375-01, R01CA225028, and R01CA205965. The authors are grateful for support from The Tara Miller Melanoma Foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.