IL15 Infusion of Cancer Patients Expands the Subpopulation of Cytotoxic CD56^bright NK Cells and Increases NK-Cell Cytokine Release Capabilities

Natural killer (NK) cells were discovered by their ability to lyse tumor cells without prior sensitization (1–4). This ability was the basis of attempts to exploit NK cells in the treatment of patients with cancer (5). NK cells activated and expanded in vitro were used for adoptive transfer, or cytokines and activators were used to affect NK cells in vivo. Several problems limited the success of these treatments, such as limited survival of adoptively transferred NK cells and toxicities caused by cytokine treatments (5).

Three subsets are distinguished within human NK cells based on their expressions of CD56 and CD94 (6). CD56^bright and CD56^dim subsets are believed to represent sequential differentiation stages of the same lineage (7, 8). A transitional subset of CD56^dim NK cells expresses CD94 (9). The three subsets differ in their locations with high portions of CD56^dim NK cells present in blood and spleen and CD56^bright NK cells mostly residing in lymph nodes and endometrium (10). CD56^dim NK cells have a higher capacity to recognize and kill target cells, whereas CD56^bright NK cells respond to activation by cytokine production, suggesting an immunoregulatory role (11). This functional distinction can be overcome in vitro with the addition of cytokine cocktails containing IL12, IL15, and IL18 that induce regulatory functions in CD56^dim and cytotoxicity in CD56^bright NK cells (12, 13). Cytokine incubations also induce the expression of activation markers on NK cells (14, 15).

NK cells possess various ways of recognizing target cells and distinguishing them from nontransformed cells (16). The Fc receptor FcRγIII (CD16) is mainly expressed in the CD56^dim subset and mediates the deletion of target cells via antibody-dependent cellular cytotoxicity (ADCC). Both NK subsets recognize stress-receptor ligands such as the MHC class I–related chains MICA and MICB through their expression of NKG2D. NK cells sense MHC class I presence on target cells as it serves as an inhibitor of activation. This ability to respond to MHC class I absence requires, however, a “licensing” step during NK-cell development, and NK cells without KIR expression that are largely found within the CD56^bright subset are unable to kill target cells that lack MHC class I (17, 18). In general, the decision to kill results from a balance of inhibitory and activating signals.

The cytokine that governs most of NK-cell biology is IL15, deficiency of which causes NK-cell absence in mice (19). Mice injected with IL15 show expansions in populations of NK cells (20). IL15 is also necessary for NK-cell activation (21). As a consequence, IL15 was shown to increase the survival of mice that bear NK-cell–sensitive tumors either as single agent or in combination with several antitumor drugs (22, 23). We have previously reported clinical trials' data using recombinant human IL15 given to cancer patients that mainly showed expansions in populations of NK cells (24).

Here, we investigate functions, phenotypes, and population expansion rates of NK-cell subsets in cancer patients before and after IL15 infusions. We show that the main impacts of this treatment were observed within the CD56^bright subpopulation that responded with substantial expansion and augmented abilities to respond with cytotoxicity and cytokine production.

A cohort of 9 patients was enrolled in this study (NCT01572493) that included one with metastatic melanoma, two with metastatic non–small cell lung cancer one with osteogenic sarcoma, one with metastatic prostate cancer, one with metastatic uterine papillary serous clear cell cancer, one with metastatic pancreatic cancer, one with squamous cell carcinoma, and one with colon adenocarcinoma. All patients had received prior therapy. All patients had adequate organ functions. rhIL15 was produced as previously described (24). All patients received 2 μg/kg/d rhIL15 via continuous intravenous infusion for 10 days that was followed by 30 days without IL15 treatments. Five patients received a second 10-day cycle of the same IL15 infusion schedule. No additional therapy was administered.

Blood samples were taken the same day prior to the beginning and 1 day after the completion of infusion cycles. Cytometry analyses were done on fresh cells and repeated on frozen cells. To prepare frozen cell samples, peripheral blood mononuclear cells (PBMCs) were purified via Ficoll centrifugation and frozen at 15 × 10⁶ cells per vial. After thawing, PBMCs were rested overnight at 37°C in RPMI/10% human AB serum (Mediatech). Sortings were done using FACSARIA FUSION after staining PBMCs with CD3, CD56, and CX3CR1 for use in cytotoxicity assays, or with CD3, CD56, and CD94 for proliferation assays. Dead cells were excluded using 7-AAD (BD Biosciences). Blood samples from normal donors were provided by the NIH blood bank and were handled as described for patient samples.

Antibodies used in this study are listed in Supplementary Table S1. PBMCs were blocked with human TruStain Fcx [Biolegend; 15 minutes, room temperature (RT)] and stained with antibodies recognizing surface proteins (45 minutes, 4°C). Erythrocytes in analyses of fresh samples were lysed with Cal-Lyse (Invitrogen). Cells were permeabilized with BD Cytofix/Cytoperm (BD Biosciences) to stain intracellular cytokines and cytolytic proteins, or with Transcription Factor Fixation Permeabilization (eBioscience) to stain Ki67. Stains for intracellular proteins were done with the provided buffers (30 minutes, RT). Analyses were done with FACS Verse and BD Fortessa.

For cytokine production analyses, thawed and rested PBMCs were stimulated with PMA (5 ng/mL; Sigma) plus Ionomycin (500 nmol/L; Sigma), or with human IL12 (10 ng/mL; Peprotech) plus human IL18 (50 ng/mL; R&D Systems). Alternatively, PBMCs were exposed to the target cell lines K562 (ATCC CCL-243) or to anti–CD20-coated Raji (ATCC CCL-86) that had been preincubated with 10 μg/mL of the antibody (30 minutes, 4°C) and washed. Stimulation and coincubations were done for 5 hours at 37°C with Brefeldin A present during the last 4 hours. For proliferation assays, sorted cells were carboxyfluorescein succinimidyl ester (CFSE)-labeled and cultured in IL15 at indicated dosages for 7 days.

We used K562, CD20-coated Raji, and C1R-MICA (kindly provided by Dr. Groh-Spies, Fred Hutchinson Cancer Research Center) as target cells and uncoated Raji as control cells. Target and control cells were labeled at different CFSE concentrations (1 vs. 0.1 μmol/L; Invitrogen). We mixed target and control cells at equal numbers and added various numbers of sorted NK cells. Incubations were done in the presence of human IL15 (2 ng/mL, 18 hours, 37°C). The ratios of surviving propidium iodide–negative target and control cells with or without NK-cell exposures were determined by FACS and used to calculate the percentages of target cell lysis with the formula: 100 − T_NK/C_NK ÷ T_con/C_con × 100 (T = target, C = Control, _NK = NK-cell presence, _CON = NK-cell absence).

We used PRISM7. P values were calculated using the paired t test to compare cytometry analyses of patient samples before and after treatments. Unpaired t tests were used for cytotoxicity assays.

We studied a cohort of 9 patients with metastatic cancer (24). Patients received two 10-day cycles of continuous infusions of recombinant human IL15 at 2 μg/kg/d that were interrupted by 30-day periods without IL15 treatments. We analyzed NK cells from blood samples that were taken immediately prior to the start of the infusion and 1 day after the completions of the infusion cycles.

Prior to infusions, NK-cell numbers and subset ratios were within the normal range (Fig. 1A). IL15 treatments caused an increase in the total number of NK cells (Fig. 1A) as has been reported before (24). The amplitudes of increase were greatest among CD56^bright NK cells. Their numbers increased on average more than 350-fold following both treatment cycles and returned to normal between treatments. Lesser increases were measured for CD56^dim NK cells (Fig. 1A). Successive treatment cycles limited their ability to expand with approximately 20- and 10-fold increases in cell numbers during the first and second cycles, respectively. NK-cell numbers of the CD56^dim subset stayed slightly elevated throughout the resting period. Thus, increased NK-cell numbers in response to IL15 infusions were predominantly caused by expansions of CD56^bright NK cells.

We analyzed Ki67 in ex vivo NK cells as a marker of their proliferation. Both NK-cell subsets responded to IL15 infusions with increased proliferation (Fig. 1B) with the highest portions of Ki67-expressing cells being found within the CD56^bright subset. Approximately 75% of these cells proliferated after IL15 treatments mirroring increases in cell numbers (see above). We observed proliferation of CD56^dim NK cells only after IL15 infusions with nearly half of all cells expressing Ki67 (Fig. 1B). IL15 infusions also caused increased expressions of the IL15 receptor signaling β-chain CD122 in both NK-cell subsets with total amounts higher in CD56^bright NK cells both before and after IL15 treatments. These data suggest that IL15 infusions were associated with upregulation of CD122 particularly within the CD56^bright subset that sensitized NK cells to IL15 signaling and resulted in increased cell proliferation and population expansion.

Some studies suggest the existence of an intermediary population between both NK-cell subsets that can be identified by its expression of CD94 within CD56^dim NK cells (9). We investigated the effect of IL15 infusions on this NK-cell subset. When compared with the two other subsets, the increase in numbers of CD56^dim CD94^high NK cells was lower than the population expansion evident for CD56^bright progenitor cells but higher than that of CD56^dim CD94^low NK cells (Fig. 1C). The rank order of these three NK-cell subsets according to population expansion corresponded to rank order according to the percentage of proliferating cells that were positive for Ki67 (Fig. 1C). Moreover, IL15 infusions increased CD122 expression in CD56^dim CD94^high NK cells to levels lower than in CD56^bright but higher than in CD56^dim CD94^low NK cells (Fig. 1C), further suggesting a direct correlation between CD122 expression and proliferation rates. We also studied whether the three NK-cell subpopulations do require different amounts of IL15 to respond with proliferation in vitro (Supplementary Fig. S1). We sorted the NK-cell populations from blood cells of normal donors and observed that the differentiation from CD56^bright to CD56^dim/CD94^high to CD56^dim/CD94^low NK cells was accompanied by increasing requirements of IL15 to induce their proliferation. Together, these data suggest an inverse correlation between NK-cell differentiation and IL15 responsiveness that resulted in a preferential expansion of the least differentiated CD56^bright subpopulation in response to IL15 infusion.

We analyzed whether IL15 infusions altered the surface expressions of molecules that are involved in cytotoxicity, differentiation, cell adhesion, chemokine response, and mechanisms of inhibition or activation. Normally, CD56^dim NK cells express more CD16 and CX3CR1 than do CD56^bright NK cells. After IL15 infusions, CD56^bright NK cells expressed amounts of CD16 and CX3CR1 similar to those expressed by CD56^dim NK cells (Fig. 2A). Treatment of CD56^bright NK cells with IL15 reduced expression of CCR7, CD27, and CD25. Conversely, CD56^dim NK cells treated with IL15 acquired features normally observed in CD56^bright NK cells in that expression of NKG2D, Trail (Fig. 2A), and IL18R (Fig. 3D) rose, subpopulations that expressed CD62L or CXCR3 appeared, and subpopulations that expressed CD57 disappeared (Fig. 2A). In CD94^low and CD94^high subpopulations, NKG2D and CD57 expression changed in both subpopulations. Increased expression of CD62L, CXCR3, and Trail was only seen in the intermediary CD94^high cells (Fig. 2B). Common to all subsets were decreases in the surface amounts of CD2 and increases of NKp30, NKp46 (Fig. 2A), and CD122 (Fig. 1B). We found no changes in expression of KIRs, NKG2A/C, CD69, DNAM1, and other markers (Supplementary Fig. S2). Thus, IL15 infusions caused phenotypic changes in all subsets of NK cells. Expression of certain receptors associated with cytotoxicity, CD16, NKG2D, NKp30, and NKp46 was induced by IL15 treatments.

NK cells respond to various stimulations with the production of cytokines (12, 25, 26). We investigated whether this ability was affected by IL15 infusions. We exposed total PBMCs from patients before and after IL15 infusions to combinations of PMA/Ionomycin, IL12/IL18, or to K562 or anti–CD20-coated target cells for 5 hours, and then assessed intracellular cytokine levels in gated CD56^dim and CD56^bright NK cells by FACS (Fig. 3). Although IL15 infusions were associated with increased cytokine production for all NK-cell types, the CD56^bright subpopulation was most affected: We observed increases in production of IFNγ, TNFα, and GM-CSF in response to PMA/Ionomycin (Fig. 3A) or IL12/IL18 (Fig. 3B) within the CD56^bright population of cells. In contrast, CD56^dim cells only increased IFNγ production only in response to IL12/IL18. CD56^bright NK cells responded to target cell exposures with the production of IFNγ and TNFα only after IL15 infusions; the same response was not affected by IL15 treatments in the CD56^dim subset (Fig. 3C). Some of the augmented cytokine responses could be related to the increased IL18R expression in CD56^dim NK cells after IL15 treatments, but cytokine expression was constitutively higher from CD56^bright NK cells (Fig. 3). Expression of the IL12R was unchanged (Fig. 3). Together, these data show a sensitization of CD56^bright NK cells to respond with cytokine production.

Immunotherapies target NK cells for their ability to recognize and lyse tumor cells. A main cytotoxic pathway involves degranulation of vesicles that contain granzymes A and B as well as perforin. We analyzed the intracellular amounts of these molecules in NK cells before and after IL15 infusions. CD56^dim NK cells constitutively expressed all three molecules, and IL15 infusions were associated with diminished amounts of granzyme B and perforin (Fig. 4A, top). We observed opposite reactions from CD56^bright NK cells, in which IL15 treatment resulted in increased amounts of all three cytolytic molecules (Fig. 4A, bottom). Thus, IL15 infusions appear to equip CD56^bright NK cells to become cytotoxic.

We assessed the cytotoxic abilities of NK-cell subsets. We used three target cell lines, the killing of which involved different receptors: anti-CD20 (rituximab) antibody-coated RAJI, K562, and C1R-MICA. These cell lines are recognized by NK cells via CD16-mediating ADCC, natural cytotoxicity receptors (Nkp30 and NKp46), and NKG2D, respectively (27, 28). Target cells were mixed with control cells insensitive to NK cells (uncoated RAJI) and exposed to sorted NK cells for 18 hours in vitro (Supplementary Fig. S3). The ratio of surviving target to control cells was used to calculate the percentage of lysed target cells. Figure 4B shows that cytotoxic activities were augmented for both CD56^dim and CD56^bright subsets of NK cells from IL15-treated patients when compared with untreated donors in response to all three target cells. Changes in cytotoxicity were greater after IL15 infusions for CD56^bright NK cells that reached cytotoxic response levels similar to CD56^dim NK cells. Together, these data show that IL15 infusions enabled all subsets of NK cells to respond with increased cytotoxicity.

NK-cell subpopulations have been defined in humans based on their expression of CD56: Cells that express this marker at low intensity are considered the cytotoxic NK-cell subset, whereas CD56^bright cells are believed to be immunoregulatory actors (10, 11). Here, we present studies that characterize the effects of IL15 infusions on various aspects of NK-cell function in cancer patients. We show that IL15 infusions were associated with over 350-fold expansion of the CD56^bright NK-cell subpopulation and a 20-fold expansion of the CD56^dim NK subpopulation. The CD56^bright cells showed increased proliferation that may have been caused by higher CD122 expression. Population expansion rates were more than 10-fold higher in the CD56^bright subset following treatment when compared with other NK-cell subsets resulting in nearly equal numbers of both types of cells present in the peripheral blood after 10-day IL15 infusions. In addition, CD56^bright NK cells acquired the capability to respond to target cell activation by both increased cytokine production and cytotoxicity as a result of the treatment. These data suggest that CD56^bright NK cells generated by IL15 infusions have a potential to respond to tumors sensitive to NK cells.

We analyzed cytokine production of NK-cell subsets. CD56^bright NK cells respond to cytokine or PMA/Ionomycin stimulations with the generation of several cytokines (12, 25, 26, 29). We observed that this activity was even more pronounced after IL15 infusions. The effect could be explained by increased cytokine receptor expressions, but additional intracellular changes are likely to exist as suggested by the improved responses to PMA/Ionomycin that did not require receptor expression. In addition, cytokine generation was only observed after coincubations with NK target cells after IL15 infusions. This suggests that treatment induced expression of receptors, normally absent on CD56^bright NK cells, that could recognize ligands on tumor cells. In the case of antibody-coated target cells, this improved recognition may involve the Fc receptor CD16, expression of which was upregulated after IL15 infusion. IL15 treatment–induced CD56^bright NK cells also responded to K562 cells. The receptors that recognize this cell line appear to be more complex (27, 28) and may involve NKp30, NKp46, and NKG2D, expression of which had also been increased after IL15 infusions. Thus, multiple mechanisms are implicated in increased cytokine responses in IL15 infusion-induced CD56^bright NK cells.

We analyzed surface marker expression changes that were induced by IL15 infusions. We observed that changes in both subpopulations contributed to an increased phenotype similarity between them. We believe that this is most consistent with activities that were induced by higher concentrations of IL15 that had been achieved after infusions when compared with physiological steady-state levels. Such higher cytokine concentrations could have activated a signaling pathway with similar signaling strength in both cell types to result in similar expressions of some surface molecules. Infusions also caused an increase of CD62L⁻ and a decrease of CD57⁻expressing populations among CD56^dim cells as well as a decrease of the CD27⁺ population among CD56^bright cells, suggesting that IL15 treatments induced expansion of less mature NK-cell populations (30–34).

Many phenotypic changes that were induced by IL15 infusions are also found after IL15-induced ex vivo cultures of NK cells (14, 15, 35–37). These include increases of NKp30, NKp46, NKG2D, Trail, and CXCR3, and decreases of CD57, CD27, and CCR7. One exception, however, exists. Cultured NK cells typically upregulate markers of activation including CD69, CD25, NKp44, and DNAM1 (14, 15, 38). We did not observe inductions of these markers after IL15 infusions. Because bona fide activating ligands are unlikely to be induced by culturing, the discrepancy between culture and infusions is most consistent with inhibitory mechanisms in vivo that are lost during culturing resulting in NK-cell activation in vitro only. A lack of NK-cell activation in IL15-infused cancer patients also suggests that no encounters between cancer and the analyzed blood NK cells had occurred that would have caused their activation.

The relationship between CD56^dim and CD56^bright NK cells remains a matter of debate. Most publications suggest CD56^bright NK cells represent a developmentally precursor stage (10). Others describe the ability to derive CD56^bright from CD56^dim via NK-cell activation (39, 40). Along this line, the increased presence of CD56^bright NK cells following IL15 treatments could have been caused by either different expansion rates of both cell subsets or a conversion of CD56^dim into CD56^bright NK cells. Most of our findings support the former: Analyses ex vivo showed increased proliferation rates in the CD56^bright subset (Fig. 1B). In addition, in vitro proliferation assays point to an increased ability of CD56^bright NK cells to proliferate in response to IL15 (Supplementary Fig. S1). Moreover, one would expect that descendants of a recent CD56^dim to CD56^bright conversion would retain some of the phenotypical markers of their precursor cells. Most of our data appear to suggest the opposite in that the majority of surface markers that are distinct between both subsets were expressed at similar levels on CD56^bright NK cells before and after treatments. These include CXCR3, CX3CR1, CD16, NKG2A, and CD158b (Fig. 2; Supplementary Fig. S2). CD25 and CCR7 are exceptions in that their expressions were downregulated on posttreatment CD56^bright NK cells to levels similar of CD56^dim NK cells. In summary, our data suggest that the predominant appearance of CD56^bright NK cells in response to IL15 treatments resulted mainly from a superior ability of this subset to proliferate.

An issue for all NK-cell–based cancer immunotherapy targeting solid tumors is their ability to migrate from blood to tumor sites. We had analyzed the expression of several chemokine receptors whose involvement in tumor-directed migration has been reported (5, 41, 42). No clear pattern emerged. Expression of CXCR3 and CX3CR1 increased, expression of CCR7 decreased, and expression of CXCR1 remained unchanged. Further studies are necessary to delineate migration patterns of IL15 infusion–induced NK cells.

We observed improved cytotoxic activities for both CD56^bright and CD56^dim NK cells against diverse target cell lines with various recognition requirements. Several changes in CD56^bright cells that had been caused by IL15 treatments could account for this increase: Besides improved recognition of target cells, these cells had increased intracellular amounts of the molecules Granzyme A, Granzyme B, and Perforin that are used for killing target cells. This was not the case for CD56^dim NK cells that also showed increased killing activities despite lower amounts of granzymes and perforin. However, increased amounts of membrane-bound death receptor ligand Trail may contribute to stronger cytotoxic responses as may the improved recognition via NKG2D for MICA-expressing targets, NKp30 and NKp46 for K562. Thus, the goal of inducing augmented cytotoxic NK-cell activities by IL15 infusions was achieved that may suggest a potential utility in cancer immunotherapy.

In summary, continuous infusions of IL15 in cancer patients is accompanied by a preferential expansion of populations of phenotypically immature CD56^bright NK cells with increased abilities to recognize tumor cells and react by cytokine production and cytotoxicity. On the basis of the present study, IL15, by increasing NK-cell numbers and their cytotoxic capacity, may increase the efficacy of anticancer therapies. IL15 could be useful in cases of tumors that have deleted Class I MHC where direct NK cytotoxicity may be effective. IL15 may be useful in combination with therapies that are based on antitumor antibodies of which the effects include tumor killing via ADCC. To translate this opportunity in the treatment of leukemia, we have initiated a clinical trial using IL15 with anti-CD52 alemtuzumab (NCT02689453), and we are planning to combine IL15 with anti-CCR4 mogamulizumab in patients with HTLV-1–associated adult T-cell leukemia/lymphoma (43).

No potential conflicts of interest were disclosed.

Conception and design: S. Dubois, J.R. Müller, T.A. Waldmann

Development of methodology: S. Dubois, J.R. Müller

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Dubois, K.C. Conlon, J. Hsu-Albert, T.A. Waldmann

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Dubois, J.R. Müller

Writing, review, and/or revision of the manuscript: S. Dubois, K.C. Conlon, J.R. Müller, T.A. Waldmann

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N. Beltran, B.R. Bryant

Study supervision: T.A. Waldmann

We thank Dr. Groh-Spies, Fred Hutchinson Cancer Research Center, for providing the C1R-MICA cell line and Drs. Sharrow (EIB, NCI, NIH) and Livak (LGI, NCI, NIH) for cytometry and cell sortings. This study was approved by the Institutional Review Board of the NCI, NIH.

This study was supported by the intramural program of the NCI, NIH.

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.