An IL-2 Paradox; Blocking CD25 on T Cells Induces IL-2-driven Activation of CD56^bright NK Cells¹

Patient samples

To conduct these experiments, we used cryopreserved PBMC samples isolated from lymphocytapheresis taken from patients in two separate trials: one as described in (17) where patients who were unresponsive to IFN-β therapy at baseline, were treated with daclizumab and IFN-β combination therapy and taken off IFN-β at late time points, and another trial where daclizumab was used as monotherapy (20). No significant differences were found between the two cohorts unless otherwise noted.

Ex vivo lymphocyte proliferation and FoxP3 staining

Cryopreserved PBMCs were thawed and 1×10⁶ cells were stained for FoxP3 and Ki67 (FoxP3 staining buffer set; eBioscience; San Diego, CA). Samples were acquired by flow cytometry (LSR II, with HTS; BD Biosciences; San Jose, CA) and analyzed with FACS Diva software (BD Biosciences). Antibodies (and clones) used include: Ki-67 (B56), CD127 (HIL-7R-M21), CD56 (B159), CD8 (RPA-T8), CD4 (SK3), CD3 (SK7; all BD Biosciences), FoxP3 (236A/E7) and CD25 (BC96; eBioscience). Staining was performed in duplicates and gating was set on appropriate isotype controls.

Ex vivo signaling assays

Cryopreserved PBMCs were thawed and plated in X-vivo media (Lonza, Walkersville, MD) at 1×10⁵ cells/well in a 96-well plate. They were rested for one hour at 37°C and pulsed with 50–100 IU/ml of IL-2 (NCI RBR Preclinical Repository; Frederick, MD, USA; all patients were tested with 100 IU/ml, a subgroup was tested also with 50 IU/ml) or 10ng/ml IL-7 (PeproTech; Rocky Hill, NJ) for 10 minutes; following which, they were fixed, permeabilized (Cytofix buffer and PhosFlow Perm buffer II, BD Biosciences) and stained with antibodies acceptable for phospho-specific staining. Antibodies used include: CD56 (MY31), CD3 (UCHT1), pSTAT5 (47, pY694-STAT5; all BD Biosciences) and CD4 (RPA-T4; eBioscience). Previous studies have determined the validity of the PhosFlow signaling methodology applied to cryopreserved PBMCs (22).

IL-2 consumption assay

Cryopreserved PBMCs were thawed and NK cells were depleted by CD56 iMag microbeads (BD Biosciences). Alternatively, T cells were isolated from PBMCs from lymphocytaphereses by negative selection (T cell isolation kit II, Miltenyi Biotech). 1×10⁶/ml of T cells or NK-depleted PBMCs were plated in IMDM media (Lonza) containing 10% human serum and either M-A251 control antibody against CD25 which does not block the IL-2-binding tac epitope (BD Biosciences) or daclizumab (Zenapax, Hoffman-La Roche Inc., Nutley, NJ; 10ug/ml each). After incubating with the antibodies at 37°C for one hour, T cells were polyclonally stimulated with CD3/CD28 Dynabeads (Invitrogen; Carlsbad, CA) at a 0.3:1 bead:cell ratio. After IL-2 production tapered off (72 hours later), cells were washed thoroughly, recounted and reseeded at 1×10⁶/ml in 10% human serum media with 20IU/ml IL-2 (NCI RBR Preclinical Repository; Frederick, MD), maintaining original Ab concentrations. After 48 hours, supernatants were collected and measured for IL-2 by ELISA (IL-2 Ready-Set-Go ELISA kit; eBioscience).

In vitro NK cell proliferation assays

Cryopreserved PBMCs from MS patients that have continued on long term daclizumab therapy were thawed and NK cells and T cells were isolated by negative selection (MACS Human NK Cell Isolation Kit and MACS Pan T cell Isolation Kit II, Miltenyi Biotec; Auburn, CA). These populations were determined to be of 89.1% and 96.5% purity respectively by FACS. T cells and NK cells were rested overnight in IMDM media (Lonza) with 10% AB human serum (Gemini Bio-Products; West Sacramento, CA). The NK cells were then CFSE stained (carboxyfluorescein diacetate succinimidyl diester; Molecular Probes, Invitrogen) as described previously (5) and co-cultured with T cells. The T cells were seeded at 5×10⁴ cells/condition (200μl 10% serum in IMDM and preincubated with daclizumab for 1 hour to prevent T cell consumption of IL-2, washed twice and stimulated with CD3/CD28 Dynabeads (Invitrogen) before co-culture with increasing numbers (1×10⁴, 2.5×10⁴ or 5×10⁴) of NK cells. In transwell experiments, PBMCs were CFSE stained and then NK cells were separated/depleted by CD56 microbeads (Miltenyi Biotec). The NK-depleted PBMCs were polyclonally activated with plate-bound anti-CD3 (20ng/ml) and anti-CD28 (10μg/ml) Abs (PeproTech) and the NK cells were added into transwells (3-μm pore size, Polycarbonated Membrane; Corning Costar, Lowell, MA) at a 1:10 NK:T cell ratio. Alternatively, the transwell experiment was repeated with purified, negatively-selected T cells and NK cells. T cells were polyclonally activated for 24–72 hours, extensively washed and re-seeded to the lower compartment of transwells at 1×10⁶ activated T cells in media containing different blocking agents as indicated. Autologous NK cells were isolated from fresh or cryopreserved apheresis samples at the day of co-culture with previously activated T cells. In order to differentiate between NK cells added to the upper compartment and activated T cells added to the lower compartment of transwell, only purified NK cells were CFSE stained before co-culture. Proliferation assays were analyzed by FACS at day 3–5 for CFSE dilution. Absolute numbers of proliferating cells were evaluated as a ratio of CFSE diluted NK cells to FITC beads that were added in equal number to each well before acquisition.

GFP transduction of K562 cells and ex vivo NK cytotoxity assay

To fluorescently tag K562 MHC-I deficient tumor cells, they were transduced with GFP. Briefly, pMSGV1-eGFP, a retroviral vector carrying the GFP gene, was co-transfected with pRD114, providing ENV for efficient packaging into the packaging cell line, 293-GP. Pseudotype virus expressing eGFP was harvested and used for K562 transduction.

To make the GFP viral supernatant, 293GP cells (Clontech; Mountain View, CA) were seeded at 6×10⁶ cells/well in a 10cm plate coated with Poly-D-Lysine (BD Biosciences). 9μg pMSGV1-GFP and 4.5μg pRD114 were transfected into 293-GP using Lipofectamine 2000 (Invitrogen) following the manufacturer’s protocol. 24–48 hr-post transfection, supernatants were collected and spun at 1000 ×g, at room temperature for 10 min to remove cell debris. For the transduction of the GFP gene into K562 cells, 4×10⁶ K562 cells/well, pre-washed with PBS, were mixed with 3 ml viral supernatant, 1 ml fresh media, 10 μg/ml protamine sulfate (APP Pharmaceuticals; Schaumburg, IL) and plated in a 6-well plate. The plate was centrifuged at 1000 ×g, at 32°C for two hours and then cultured in a 37°C, 5% CO₂ incubator. GFP expression was monitored using fluorescence microscopy and FACS.

Single K562 GFP clones were generated following a 0.3 cell/well limiting dilution of the eGFP transduced K562 cell mixture. Cells in eGFP positive wells were expanded and the purity of the clone and intensity of eGFP expression were confirmed by FACS analysis.

Natural killer cell cytotoxicity was measured in a similar fashion to a protocol previously described (23). Cryopreserved PBMCs were thawed and plated at 1×10⁶ cells/well and rested for 30 minutes at 37°C before adding 1×10⁵ GFP-transduced MHC-I deficient K562 target cells and CD107a antibody (H4A3; BD Biosciences). Control wells included effectors (NK cells only with CD107a Ab) without targets and targets (K562 cells only) without effectors. Cultures were incubated overnight, and subsequently stained for surface expression of CD3 (UCHT1), CD8 (RPA-T8), and CD56 (MY31; BD Biosciences). They were acquired by flow cytometry, additionally assessing for the presence of GFP and CD107a. Live GFP⁺ cells were gated on forward and side scatter and gating for CD107a (i.e. degranulation) was set from control wells of effectors without targets. Cells were proportionally enumerated between different conditions using a reference number of fluorescent beads that were added in equal numbers to all conditions.

In vitro NK cytotoxity assay

Fresh PBMCs from normal donors were isolated by ficoll separation (Lymphocyte Separation Medium; Lonza) and NK cells were isolated by negative selection (MACS Human NK cell Isolation Kit; Miltenyi). NK cells were then incubated overnight at 37°C and 5% CO₂ in IMDM media (Lonza) containing 10% human serum (Gemini Bio-Products) with either no additional cytokine, 10 or 100 IU/ml IL-2, or 2 or 20 ng/ml IL-15 (PeproTech). They were then washed thoroughly and some aliquots were taken for intracellular staining of perforin as described (5). Following washing, NK cells were seeded in X-vivo media at 5×10⁴ cells with 5×10⁴ GFP⁺ K562 cells with 4 μl of CD107a PE-Cy5 Ab per pretreatment condition. Control wells were identical to those used in the ex vivo cytotoxicity assay described above. After 4 hours, the coculture assay was stained for CD107a incorporation and intracellular perforin and analyzed for death of GFP+ K562 cells. Antibodies used included CD56 (B159), CD3 (UCHT1), perforin (δG9), granzyme A (CB9) and CD107a (H4A3) (all BD Biosciences). Gating for perforin was set on the appropriate isotype control and gating for CD107a (i.e. degranulation) was set from control wells of effectors without targets. Perforin expression in cultures with targets was measured from CD107a⁺CD56^bright NK cells that had released their granules in culture. Absolute numbers were again acquired using fluorescent beads for normalization.

Statistics

To compare therapy samples to baseline samples and in vitro proliferation data, we analyzed our data using one way RM ANOVA unless otherwise indicated. We used the Friedman one way RM ANOVA on Ranks to analyze in vitro cytotoxicity data using increasing cytokine levels and paired t-test for the IL-2 consumption assay. All differences listed in the text are changes in the median values. In figures, p<0.05 is annotated as *, p<0.01 is marked with ** and p<0.001 is marked with ***. Statistical analyses were performed with SigmaStat software version 3.5.

Daclizumab therapy causes increased proliferation of CD56^bright NK cells

As previously reported (5, 17), CD56^bright NK cells expand upon daclizumab therapy. Our results again illustrate the significant in vivo expansion of CD56^bright NK cells that continues between early and late therapy time-points (B→T1: 115.6%; p<0.01 and T1→T2: 78.9%; p<0.001) (Fig. 1A). This expansion was at least partially driven by enhanced in vivo proliferation of CD56^bright NK cells as demonstrated by an early increase in cells positive for Ki67, a chromosomal marker of cells undergoing mitosis (B→T1: 34.1%; p<0.01). However, the proliferation of CD56^bright NK cells plateau as CD56^bright NK cells expanded, as the proportion of proliferating CD56^bright NK cells decreased slightly between T1 and T2 (T1→T2: −8.8%; Fig. 1B). Nevertheless, the absolute numbers of proliferating CD56^bright NK cells remained significantly higher than at baseline during the entire daclizumab treatment (B→T1: +303.1%; B→T2: +591.5%; p<0.05; Fig. 1B right panel).

In this case we observed differences between the patients in the daclizumab monotherapy trial versus those that took daclizumab as an add-on therapy to IFN-β during the first 6 months of treatment (i.e. combination therapy trial; please see methods for a detailed explanation of the differences between the two trials). At baseline, patients from the combination therapy trial had significantly higher numbers of CD56^bright NK cells (55.4%, p<0.05: Mann-Whitney Rank Sum test) than patients from the monotherapy trial. Accordingly, the decrease in the proportion of Ki67⁺ proliferating CD56^bright NK cells from T1 to T2 was almost entirely driven by patients from the combination therapy trial (−46.0% in combination therapy patients versus −5.7% in monotherapy patients; p<0.05; t-test). Daclizumab treatment did not the affect number of CD56^dim NK cells (−6.6% from B to T2; ns).

FoxP3⁺ T regulatory cell populations decrease after therapy and proliferate at lower rates

We hypothesized that the observed expansion and enhanced proliferation of CD56^bright NK cells was due to higher in vivo availability of IL-2, because daclizumab inhibited IL-2 consumption by T cells. However, in vivo IL-2 production generally occurs at sites of immune activation, such as in lymphoid organs or inflamed tissues, which are not amenable to ex vivo testing. Therefore, we utilized FoxP3⁺ Tregs as a surrogate marker for in vivo IL-2 consumption by T cells, because FoxP3⁺ Tregs are known to constitutively express CD25 and are dependent on IL-2 for their proliferation and survival. Intranuclear staining for transcription factor FoxP3 (example of raw data in Supplementary Fig. 1) revealed that there was a statistically significant decrease in Treg populations induced by daclizumab therapy (B→T1: −39.1%; B→T2: −40.4%; p<0.001; Fig. 2A). Staining for Ki67 identified a proportion of proliferating Tregs, which decreased between B and T1 (B→T1: −24%; p<0.05), but apparently returned to baseline at T2 (B→T2: −1.5%; ns; Fig. 2B). However, when analyzing the number of proliferating Tregs per 1000 lymphocytes (Fig. 2C), it became clear that the absolute numbers of proliferating Tregs remained significantly below baseline values for both T1 and T2 (B→T1: −62.2%; B→T2: −58.4%; p<0.001; Fig. 2C).

The establishment of a new balance between absolute numbers of Tregs and their in vivo cycling indicated some type of functional adaptation of Tregs to the lack of high affinity IL-2 signaling induced by the blockade of CD25 by daclizumab (Fig. 2D). Because Tregs can utilize other γ_c-signaling cytokines (especially IL-7) for their development and survival (24), we investigated the changes in IL-7Rα expression (CD127) on Tregs. We observed a progressive increase in the level of expression of CD127 induced by daclizumab treatment (B→T1: +16%; p<0.01 and B→T2: +32.9%; p<0.01; Fig. 2E). The increased expression of CD127 in FoxP3⁺ CD4⁺ T cells may be seen in recently activated T cells, which lack suppressive function (25). However, it is unlikely that our gating included activated T cells, because the level of CD127 expression on the FoxP3⁺ CD4⁺ T cells remained significantly below the levels of CD127 expression on FoxP3− CD4⁺ T cells (MFI 198.5 vs. 803.8; p<0.001; Rank Sum Test) during the entire treatment period. Because another in vivo marker of the lack of IL-2 signaling on Tregs is their decreased expression of FoxP3 (24), we analyzed the level of FoxP3 expression on Tregs under daclizumab therapy (Fig. 2F) and observed that daclizumab induced a sustained decrease in FoxP3 MFI (B→T1: −12.6%; B→T2: −14.5%; p<0.001).

Signaling dynamics following the blocking of CD25

In order to examine how daclizumab therapy altered IL-2 signaling, we analyzed the phosphorylation of STAT5 on T cell and NK cell subsets in PBMCs by flow cytometry in response to exogenously added IL-2 (50–100 IU/ml).

IL-2-induced STAT5 phosphorylation on T cells (CD4⁺ T cells are depicted as a representative T cell subset; Fig. 3) was significantly inhibited by daclizumab therapy in comparison to baseline samples (B→T1: −40.2%; p<0.001, B→T2: −28%; p<0.001; Fig. 3A). We observed no changes in total STAT5 protein induced by daclizumab therapy (Supplementary Fig 2). There was additionally an early decrease by CD8⁺ T cells in STAT5 phosphorylation to IL-2 (B→T1: −36.8%; p<0.001; data not shown), but phosphorylation rebounded at the later timepoint (B→T2: 5.3%; p<0.05; data not shown). If cells were further incubated in vitro with daclizumab (10μg/ml) prior to IL-2 pulsing, signaling was completely abrogated at all time points (data not shown). As a signaling control we selected IL-7, because it also phosphorylates STAT5 and because we observed an upregulation of CD127 (the α-chain of the IL-7 receptor) on FoxP3⁺ Tregs. Surprisingly, we observed an increase in signaling to IL-7 by CD4⁺ T cells (B→T2: 35.6%; p<0.05; Fig. 3B) and CD8⁺ T cells (B→T2: 36.7%; p<0.001; data not shown).

In contrast to the effect of daclizumab therapy on IL-2 signaling on T cells, we observed no inhibition of IL-2 signaling on CD56^bright NK cells. In fact, there was an upward, non-significant trend in the proportion of CD56^bright NK cells that phosphorylated STAT5 in response to IL-2 throughout the treatment course (Fig 3C). Due to daclizumab-induced expansion of CD56^bright NK cells, this resulted in a highly significant increase in the absolute numbers of IL-2-induced pSTAT5+ CD56^bright NK cells during daclizumab therapy (B→T1: +165.7%, B→T2: +407.0%; p<0.001; Fig 3C; right panel). We observed no significant changes in IL-7 induced STAT5 phosphorylation on CD56^bright NK cells (data not shown).

In vitro model of IL-2 driven proliferation of CD56^bright NK cells

So far our data indicate that daclizumab therapy inhibits IL-2 driven STAT5 phosphorylation by T cells, but does not affect IL-2 signaling on NK cells and that CD56^bright NK cells proliferate more vigorously in vivo after initiation of daclizumab treatment. Activated T cells are the main producers of IL-2 in vivo and we have previously reported (5) that daclizumab treatment does not inhibit IL-2 secretion by activated T cells. Therefore, we wanted to examine if daclizumab limits the consumption of IL-2 by T cells and as a consequence, if the resultant excess of T cell-derived IL-2 can drive proliferation of CD56^bright NK cells.

To this end, we performed an IL-2 consumption assay (Fig. 4A) where T cells were polyclonally activated for 3 days, then extensively washed and re-seeded in the presence of daclizumab or M-A251 control anti-CD25 Ab (10μg/ml each) and exogenously added IL-2 (20 IU/ml). Because the IL-2 production by T cells lasts 24–48h post-stimulation, and the CD25 expression peaks 72h post-stimulation (data not shown), the 72h time point was selected as optimal to measure IL-2 consumption. As demonstrated in Fig. 4A, activated T cells treated with M-A251 had almost completely consumed all exogenously added IL-2 (90.7%; p<0.05; paired t-test) in 24h. However daclizumab-treated activated T cells had consumed only 17% (ns; paired t-test) of exogenously added IL-2.

Next, we evaluated the proliferation of CD56^bright NK cells in co-culture with increasing numbers of T cells polyclonally activated in the presence of daclizumab (Fig. 4B). We observed that CD56^bright NK cells in culture with larger numbers of T cells proliferated more vigorously compared to CD56^bright NK cells cultured with fewer T cells (1:1→2.5:1: 66%; p<0.001, 2.5:1→5:1: 31.4%; p<0.001and 1:1→5:1: 116.6%; p<0.001).

To investigate whether T cell to NK cell contact is necessary for the observed enhanced proliferation we utilized transwells to separate CFSE stained NK cells from polyclonally activated T cells (Fig. 4C & D). We observed significantly enhanced survival and proliferation of CD56^bright NK cells in transwell inserts if T cells were activated in the presence of daclizumab in comparison to control Ab M-A251 (Fig. 4C; compare first and second panels). Furthermore, proliferation of CD56^bright NK cells was almost completely abrogated in the presence of IL-2 neutralizing Ab (Fig. 4C & D). Addition of a low dose of IL-2 (20IU/ml) enhanced proliferation of CD56^bright NK cells above the M-A251 condition, but did not reach levels observed in the daclizumab condition. We also observed that activated NK cells were able to migrate through 3 μm pores of the transwell to the lower compartment, likely due to a strong chemotactic gradient produced by activated T cells. Our set up purity checks demonstrated >95% NK cell purity in the upper compartment and >95% T cell purity in the lower compartment (Supplementary Fig. 3) at the beginning of transwell co-culture. Yet, consistently in every experiment we observed transmigration of activated NK cells through 3 μm semipermeable membrane and in fact significant enrichment of proliferating CD56^bright NK cells in the lower compartment. Therefore, we provide both raw and analyzed data depicting CFSE-stained NK cells and CFSE-negative T cells in both compartments.

Finally, to confirm that activated T cells were driving the proliferation of CD56^bright NK cells through soluble IL-2, we performed supernatant transfer experiments, where we collected supernatants from activated T cells cultured for 24–48h in the presence of M-A251 control Ab or daclizumab in the presence or absence of IL-2 blocking Ab. We isolated NK cells by negative selection from apheresis samples; CFSE stained them and cultured them in supernatants collected from activated T cells for 3 days (Fig. 4E & F). Again, we observed that supernatants from activated T cells cultured in the presence of daclizumab induced proliferation of CD56^bright NK cells that was at least 2–3 fold higher than proliferation of CD56^bright NK cells cultured with the media from M-A251 cultures. Blockade of IL-2 resulted in complete abrogation of survival and proliferation of CD56^bright NK cells.

CD56^bright NK cells exhibit greater cytotoxicity following daclizumab therapy

In addition to describing the expansion of the CD56^bright cell population, we sought to define functional changes of CD56^bright NK cells induced by daclizumab therapy. In order to do this, we used a flow cytometric killing assay with GFP transduced MHC-I deficient cancer cell line, K562, as a target cell and utilized the detection of transient surface expression of LAMP protein CD107a, normally expressed only intracellularly in cytotoxic granules, as identification of effector cells in complex PBMC cultures. While both CD56^dim and CD56^bright NK cells (but not CD8⁺ or CD4⁺ T cells) are cytotoxic in this assay, the proportion of CD56^dim NK cells in PBMC cultures and their activation status was not affected by daclizumab therapy (5); therefore, we expected that the observed changes could be directly attributed to changes in CD56^bright NK cell subsets. We confirmed validity of this assumption by evaluating changes in the degranulation of both NK cell subsets.

In baseline samples we observed on average 14.5% killing of K562 cells (Fig 5A). This baseline cytotoxicity increased by 139.4% (p<0.01) to 34.7% killing at T1 and further by 55.1% (p<0.001) between T1 and T2 (T2: 53.9% killing). This enhanced cytotoxicity was paralleled by a significant increase in the proportion of degranulating CD56^bright NK cells as visualized by CD107a incorporation (B→T1 +44.9%; p<0.001, B→T2 +50.4%; p=0.001; Fig. 5B). Analysis of the absolute numbers of degranulating NK cell subsets demonstrated a highly significant increase in the degranulating CD56^bright NK cells (B→T1 +207.9%; p<0.001, B→T2 +530.2%; p<0.001; Fig. 5C; left panel) which correlated with the observed enhanced killing (R_Spearman = 0.38, p=0.001; Fig. 5D). No significant change in the numbers of degranulating CD56^dim NK cells was observed (Fig. 5C; right panel).

CD56^bright NK cells respond to increases in IL-2 concentration by enhancing their cytotoxicity

While our data point to IL-2 as a central factor in the expansion and activation of CD56^bright NK cells observed during daclizumab therapy, they do not rule out contributions of other factors that may be important in vivo, such as IL-15. Therefore, we next investigated whether IL-2 or IL-15 could induce changes in the function of CD56^bright NK cells in vitro analogous to those observed after daclizumab therapy in vivo. Fresh NK cells were isolated by negative selection from healthy donor apheresis. These NK cells were stimulated overnight with varying amounts of IL-2 and IL-15 and subsequently examined for their ability to kill MHC-I negative targets. Intracellular cytokine staining following overnight activation revealed that all CD56^bright NK cells expressed perforin, but increasing concentrations of IL-2 (Fig. 6A) and IL-15 (Fig. 6D) induced higher perforin expression (MFI) in NK cells that were cultured without targets. IL-2 was more effective in this regard, especially at lower concentrations (10IU/ml of IL-2 = 6.1 ng/ml increased perforin MFI by 61.3%; p=0.001, in comparison to 2ng/ml of IL-15, which increased perforin MFI by 26.9%; p<0.05). When NK cells were co-cultured with targets, they degranulated vigorously (Fig 6B and E), losing their intracellularly pre-formed perforin (Fig 6A and D; NK cells +target) as they killed GFP-tagged K562 cells (Fig. 6C and F). While both degranulation and K562 killing was again enhanced by IL-2 and IL-15; IL-2, especially at low doses, was significantly more effective in functional activation of CD56^bright NK cells. Low dose IL-2 increased degranulation of CD56^bright NK cells by 111.63% (Fig. 6B) in comparison to 35.94% induced by low dose IL-15 (Fig. 6D). Similarly, killing efficiency was enhanced by 32.8% (p<0.05; Fig. 6C) by low dose IL-2 and only by 2.36% (ns; Fig. 6F) by low dose IL-15. The loss of perforin MFI in CD56^bright NK cells upon co-culture with targets (Δ perforin MFI −/+ target) was significantly greater when NK cells were pre-treated with IL-2 (+184.9% with IL-2 10IU/ml and +209.3% with IL-2 100IU/ml; p<0.05 for both; Fig 6A, right panel) as compared to IL-15 (+134.9% with IL-15 2ng/ml and +136.9% with IL-15 20ng/ml, both ns; Fig 6D, right panel). This cytokine-enhanced loss of perforin by CD56^bright NK cells upon their co-culture with targets correlated strongly with enhanced efficacy of K562 killing (R_Spearman = 0.718; p=0.00141; Supplementary Fig 4).