Tethering IL2 to Its Receptor IL2Rβ Enhances Antitumor Activity and Expansion of Natural Killer NK92 Cells

Natural killer (NK) cells are lymphocytes endowed with the innate ability to attack malignant and virus infected cells (1–3). Several interleukins, and in particular IL2, activate and expand T cells and NK cells (4). Systemic IL2 supplementation could therefore enhance immunity in cancer and viral infection. However, tumor cells and their microenvironment (TME) often repress NK cells antitumor activity by orchestrating a multitude of escape mechanisms (5).

Clinical trials using high-dose IL2 infusions have met limited success due to severe side effects that mimic sepsis (6–8), while low-dose IL2 efficacy is limited by the short half-life (less than 10 minutes) of IL2 in vivo (9), and due to depletion of low IL2 doses by Tregs and other lymphoid cells (10). Several strategies based on IL2 have aimed to enhance NK cytotoxicity while reducing toxicity in patients, with limited efficacy. Cultured ex vivo autologous NK cells activated and induced to proliferate by IL2 display less antitumor efficacy (11) than allogeneic NK cells (12), because self-class I HLA signaling suppresses NK cytotoxicity and cytokine release (13). However, in order for allogeneic NK cells to be effective, pretransfer lymphodepletion is required to reduce competition for growth factors and cytokines (14, 15). Moreover, IL2 is needed to sustain NK cytotoxicity after in vivo transfer, exposing patients to systemic side effects.

Past efforts to express endogenous IL2 in NK cells (16) or to express membrane-bound endogenous IL2 (17) showed limited success with micrometastatic models and were not as efficacious as NK cells stimulated with IL2 ex vivo. The limited success of strategies using NK cells could be explained by the failure of activated NK cells to outcompete Tregs for cytokines in the host and the immunosuppressive effect of the TME, which includes myeloid-derived suppressor cells (MDSC). Both MDSCs and Tregs suppress NK-cell functions either by direct contact or by secretion of TGFβ1 (18, 19).

To selectively activate and expand NK cells without exogenous IL2, while maintaining NK cytotoxicity and proliferation both in vitro and in vivo, circumvent the requirement of IL2Rα and its lack of expression in NK cells, thus avoiding IL2 off-target effects, cytokine competition, and activation of downregulating lymphoid cells like Tregs. We created a novel chimeric IL2-IL2Rβ (CIRB), consisting of IL2 tethered via a linker to IL2Rβ, which functions like constitutively activated IL2Rβ. NK92 cells expressing CIRB (NK92^CIRB) produce anticancer effects in vitro that are equivalent to or better than NK92^IL2. Importantly, compared with IL2-stimulated NK92 and NK92^IL2, the anticancer activity and growth of NK92^CIRB cells were resistant to the immunosuppressive cytokine TGFβ1 and dexamethasone. Moreover, the in vivo anticancer activity of NK92^CIRB was significantly superior to that of NK92^IL2. NK92^CIRB cells were also significantly more resistant to radiation and showed longer survival in tumor-bearing animals. Surprisingly, NK92^CIRB and to a lesser extent NK92^IL2 express CD16, while it was not detected in IL2-stimulated NK92. CD16 expression synergized with trastuzumab to exert substantial antibody-dependent cellular cytotoxicity (ADCC). In addition, NK92^CIRB cells have higher expression of NKP30, NKP44, and perforin-1 than NK92^IL2 and in contrast to NK92^IL2 increased their production of granzyme-B, TNFα, and IFNγ upon contact with cancer cells. In conclusion, the novel chimera CIRB endows NK92 cells with very useful attributes that could improve immune therapy of cancer and potentially viral infections.

Dexamethasone, chloroquine, Matrigel (catalog no. 126-2.5), and human glycosylated IL2 were purchased from Sigma-Aldrich Co; horse serum (HS), DMEM/F12, Lipofectamine 2000, and TRIzol from Life Technologies; FBS from Atlanta Biologicals; RPMI1640 from Lonza; Smartscribe and Blueprint Onestep RT-PCR Takara kit from Clontech Laboratories; platinum SYBR Green qPCR from Invitrogen; PfuUltra DNA polymerase from Stratagene; and human TGFβ1 from Antigenix America Inc. IL2 was from MGH-DF/HCC Recombinant Protein Core; human IL4 from Shenandoah Biotechnology Inc; and anti-HER2 (trastuzumab), humanized antibody (BioVision Inc.), and PNGase F from New England Biolabs.

HEK293T, NK92, NK92-MI, PC-3, HepG2, MDA-MB-231, Panc-1, BT474, and U266 cells were from ATCC, and U251GM was a gift from Dr. Samuel Rabkin (Massachusetts General Hospital, Boston MA). U266-GFP-Luc cells were generated by lentiviral transduction using CSCW-GFP lentiviral vectors. Tumor cell lines PC-3, U251GM, U266, Panc-1, BT474, and MDA-MB-231 were cultured in complete RPMI1640 medium, HEK293T and HepG2 in complete DMEM/F12, and NK92 and derived cell lines in RPMI1640 as described previously (20) with 100 IU/mL IL2. All cell lines and assay cultures were maintained at 37°C and 5% CO₂. All cell lines were obtained between 2015 and 2016 and were used until a maximum passage of 20 when they were replaced with a fresh passage. Monitoring for mycoplasma contamination was done using MycoFluor mycoplasma detection kit (Molecular Probes).

IL2 cDNA was amplified from human brain total RNA by RT-PCR using forward primer 5′-TGCAGGATCCACTCACAGTAACCTCAACTCC-3′ and reverse primer 5′-TGCACTCGAGAGTGAAACCATTTTAGAGCC-3′ and cloned in BamHI-XhoI in pCDNA4-TO. To build the CIRB chimera, we first fused IL2 and the extracellular domain of its receptor IL2Rα, which was amplified by RT-PCR from NK92 total RNA using forward oligonucleotide 5′-GGATTACCTTTTGTCAAAGCATCATCTCAACACTGACTGAGCAGAAGCTCATTTCGGAAGAAGACCTTGAAATGGAGACCAGTCAGTTTCCAGG-3′, bridging IL2 C-terminal (12 amino acids before the stop codon), and contains the cMyc Tag, the sequence between amino acids 187–194 of IL2Rα as well as and the noncoding 3′ sequence of IL2 plasmid. This primer was used with reverse oligonucleotide 5′-CCTGATATGTTTTAAGTGGGAAGCACTTAATTATCAGATTGTTCTTCTACTC TTCCTCTGTCTCC -3′. The amplified fragment was used as an oligonucleotide to mutagenize IL2 wild-type, resulting in an IL2-IL2Rα chimera. To build CIRB final chimera construct, the IL2-IL2Rα chimera was used to amplify IL2 with a C-terminal cMyc tag followed by only the extracellular domain of IL2Rα and then followed by the N-terminal fragment of IL2Rβ using forward 5′-TGCAGGATCCACTCACAGTAACCTCAACTCC-3′ and reverse 5′-GGGAAGTGCCATTCACCGCGCAGGAAGTCTCACTCTCAGGA-3′. The product was then reamplified using the same forward primer and the reverse 5′-GGCTCTCGAGTTGTAGAAGCATGTGAACTGGGAAGTGCC ATTCACCGC-3′. An XbaI site in IL2 was first removed by mutagenesis using primers forward 5′-CATCTTCAGTGCCTAGAAGAAGAACTC-3′ and reverse 5′-GAGTTCTTCTTCTAGGCACTGAAGATG-3′. IL2Rβ was then amplified using forward 5′-TTCCCAGTTCACATGCTTCTACAAGTCGACAGCCAACATCTCCTG-3′ and reverse 5′-AGCTTCTAGACTCGAGTTATCACACCAAGTGAGTTGGGTCCTGACCCTGG -3′. Next, the fragment IL2-cMyc-IL2Rα was open Xho-XbaI and IL2Rβ was added as SalI-XbaI fragment to form the chimera CIRB. Both IL2 and CIRB were transferred from pcDNA4-TO using SpeI (blunt end) and XhoI to CSCW-mCherry digested with BamHI (blunt end) and XhoI. All constructs were sequenced.

HEK293T cells were transfected using Lipofectamine 2000 with 2.4 μg DNA of pVSV, (Clontech Laboratories), pCMVdr8.2dvrp Addgene (plasmid # 8455), and the lentiviral construct CSCW-GFP or CSCW-mCherry vectors (MGH Vector Core, Boston, MA) to express either CIRB or IL2 constructs, using the ratios 1:0.4:1, respectively, with 25 μmol/L chloroquine. Six hours posttransfection, media were changed and lentiviral supernatant was collected 3 days later, filtered through a 0.45-μm syringe. NK92 cells were infected by spinoculation at 1,800 × g for 45 minutes at an optimal multiplicity of infection (MOI) of 46 lentiviral particles per cell in a 2 mL Eppendorf tube containing 2 × 10⁵ cells. Infected cells were plated with IL2 (100 IU/mL) for two days, then weaned of exogenous IL2.

Phosphorylated Stat 5 was detected using rabbit anti-STAT5 phospho (Tyr694) antibody (BioLegend) while the chimera CIRB was detected by mouse anti-human IL2 antibody (PeproTech), in 40 μg cell lysates obtained using RIPA lysis buffer (Santa Cruz Biotechnology). Secondary donkey anti-mouse IRDye 800CW and goat anti-rabbit IRDye 680RD antibodies were from LI-COR Biosciences.

NK-cell markers' expression was verified using mouse anti-human antibodies to CD45-APC-CY7, CD25-FITC, CD16-PE, CD3-PE-CY7, CD56-PAC BLUE, and CD122-PE, were from BD Biosciences. Anti-human antibodies to NKG2D-APC, MHC-1 HLA-A2-APC, NKP30-PE, NKP44-PE-CY7, NKP46-FITC, Granzyme B-FITC, perforin1-PE, IFNγ-APC, TNF-α1- APC-CY7, were from BioLegend, DAPI from Invitrogen, mouse anti-cMyc: sureLight APC was from Columbia Biosciences. Cells were sorted at MGH Flow Cytometry Core facility using a BD 5 laser SORP FACS Vantage SE Diva system (BD Biosciences). FACS data and ∑Median statistics were analyzed using FlowJo Software (Tree Star, Inc.). Human primary NK (hNK) cells were extracted from peripheral blood of healthy donors using the Rosettesep human enrichment kit (StemCell Technologies).

A total of 8 × 10³ U266GFP cells (selected for firm adherence) were plated in triplicate in 96-well plates. Twenty-four hours later NK92^IL2, NK92^CIRB, and NK92 (prestimulated for 24 hours with IL2, 100 IU/mL) were added at effector:target cells ratios (E:T) of 1:8, 1:4, 1:2, 1:1, and 2:1. After two days of coculture, NK cells were suspended to allow further killing of U266GFP cells. After a total 4 days of coculture, survival of U266GFP cells was quantified by GFP-emitted fluorescence.

In other experiments, we evaluated NK-cell lines' anticancer effect against U251GM, PC-3, Panc-1, MDA-MB-231, and HepG2 cells. A total of 3.2 × 10⁴ cancer cells were first plated in a 24-well plate for either 24 hours prior to adding NK92 (prestimulated with IL2, 100 IU/mL), NK92^IL2, or NK92^CIRB at E:T ratio of 2:1 target cancer cell, or only 5 hours before adding NK92^IL2 or NK92^CIRB at E:T ratios of 0:1, 1:1, 2:1, and 3:1 for each cancer cell line. Cocultured cells were then incubated for 4 days. Viability of cancer cells was determined using a 0.1% Crystal violet in a10% alcohol solution followed by extraction using 70% ethanol and reading absorbance at 595 nm.

BT474 (8 × 10³ cells) were plated in 96-well plates. Twenty-four hours later, cells were incubated for 20 minutes at room temperature with 1 μg/mL trastuzumab before the addition of NK cells at an E:T ratio of 2:1. After 3 days of incubation, viability of cancer cells was determined using Crystal violet staining.

NK92, NK92^IL2, and NK92^CIRB (6.4 × 10⁴ cells), were plated and exposed for 24 hours to TGFβ1 (20 ng/mL), or dexamethasone (0.5 μmol/L). During this time, NK92 cells were incubated with IL2 at 20 IU/mL. U251GM, PC-3, Panc-1, MDA-MB-231, and HepG2 cells (3.2 × 10⁴ cells) were then added to NK cells at an E:T ratio of 2:1. Cocultured cells were incubated for 4 days and cell viability of cancer cells was determined using Crystal violet staining.

To determine the impact of immunosuppressors TGFβ1, IL4, and dexamethasone on the growth of NK92^IL2 and NK92^CIRB (3.0 × 10⁴ cells), were plated in triplicate in a 12-well plate, grown under TGFβ1 (10 ng/mL), IL4 (10 ng/mL), or dexamethasone (1 μmol/L) for 3 days, and then refreshed for another 3 days of growth under the same conditions for a total of 6 days. The viability and growth of cells were determined with Trypan blue exclusion using a Bio-Rad TC20 automated cell counter.

All experiments involving animals were approved by the Institutional Animal Care and Use Committee at Massachusetts General Hospital (Boston, MA). U251GM or PC-3 cells were suspended in serum-free RPMI1640 containing 20% Matrigel and injected subcutaneously as 4 × 10⁶ cells for PC-3 or 3 × 10⁶ cells for U251GM in a volume of 0.5 mL using a 0.5-inch 29-gauge needle and a 1-mL insulin syringe in 1-week-old (24–25 g) male NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ immunodeficient mice (Jackson Laboratory). Tumor areas (length × width) were measured twice a week using Vernier calipers (Manostat Corp.) and tumor volumes were calculated on the basis of: volume = π/6 (length × width)³/2. Treatment with NK92^CIRB or NK92^IL2 was initiated when the average tumor volume reached approximately 200 mm³ for PC-3 or approximately 160 mm³ for U251. For animals bearing PC-3 tumors, freshly prepared NK cells were suspended in PBS irradiated with 500 cGy and administered as 4 weekly injections (15 × 10⁶ cells in 200 μL per mouse), via the tail vein. For animals bearing U251 cells, NK92^CIRB or NK92^IL2 were not irradiated.

U251MG tumor cells were grown subcutaneously in Nod/scid mice to approximately 160 mm³. Nonirradiated NK92^IL2 and NK92^CIRB (10⁷ cells), were injected, via the tail vein. A second injection of nonirradiated NK cells (5 × 10⁶ cells) was carried out 4 days later. Seventeen days later, animals were killed and cardiac blood was collected, processed, and analyzed by flow cytometry using human-specific anti-CD45 and mCherry fluorescent protein.

After irradiation at 10 Gy (0.83 Gy for 12 minutes), NK92^CIRB and NK92^IL2 were cultured and their survival was determined, using Trypan blue exclusion every 24 hours for 3 days.

Natural cytotoxicity receptors NKP30, NKP44, NKP46, cytolytic enzymes perforin-1 and granzyme-B, and cytokines TNFα and INFγ were quantified by qRT-PCR using the primers listed in Supplementary Table S1. Protein expression profiles were also determined in NK cells by flow cytometry under TGFβ1 and dexamethasone treatments as well as under activation by contact with PC-3 cells

Statistical significance of differences was determined by two-tailed Student test, one-way ANOVA, paired Tukey multiple comparison test. All tests included comparisons to untreated samples or as indicated in the text. Statistical significance is indicated by *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.001. Analyses were performed using Prism software version 6 (GraphPad Software).

The quaternary crystal structure of IL2 receptor complex (21) shows that the C-terminal end of IL2 and the N-terminal residue of IL2Rβ are separated by 41Å. For a linker between IL2 and the N-terminus of IL2Rβ, we choose the extracellular domain of IL2Rα. A cMyc tag was added between IL2 and the linker. The linker fold was predicted computationally to be a helix-dominated structure (Supplementary Fig. S1). Linker flexibility was assessed using the computational method of Karplus and Shultz method (22), which indicates better than average flexibility (1 or greater on a 0–2 scale) at all the peptide linkages. The mature receptor IL2Rβ protein without signal peptide was placed after the linker to yield the chimera CIRB. CIRB and IL2 were separately cloned in lentiviral vector coexpressing mCherry (Fig. 1A).

NK92^IL2 and NK92^CIRB cell lines established with lentiviral infection at an identical MOI of 46 acquire IL2 independence and proliferate indefinitely. Both cell lines showed similar growth during a 6-day period, faster than NK92-MI, another IL2-independent cell line (Supplementary Fig. S2), with robust survival after subjection to multiple freezing and plating cycles in culture, comparatively to NK92 and NK92-MI. Using an anti-cMyc mAb, we next examined the expression of CIRB at the cell surface of transiently transfected HEK293 (Fig. 1B) and stable NK92^CIRB (Fig. 1C). We found clear evidence of the surface expression of cMyc in NK92^CIRB, but not in NK92^IL2. CIRB expression was further confirmed using an anti-CD122 mAb, which recognizes the native IL2Rβ as well as the chimera. Figure 1D shows the endogenous CD122 present in NK92^IL2 cells, as expected, but at levels lower than in NK92^CIRB, which express both IL2Rβ and CIRB. The expression of the full-length chimera CIRB was further detected by Western blot analysis using monoclonal anti-human IL2. Figure 1E shows a full-length size of 95 kDa, which is higher than the predicted size of 80 kDa due to posttranslational glycosylation. Indeed upon deglycosylation treatment with PNGase F, the chimera molecular weight was reduced to approximately 80 kDa. Although IL2 produced by NK92^IL2 could be detected by RT-PCR (Fig. 1F) and its secretion in the media could support the growth of bystander NK92 (Fig. 1G) it could not, however, be detected by Western blot analysis, as the limit of detection for anti-IL2 antibody is 3 ng/lane. In fact, the IL2-producing NK92-MI was reported to express 19.4 pg/mL, an amount that could be detected only by ELISA (20).

We first compared cytotoxicity of NK92, NK92^IL2, and NK92^CIRB cells against an adherent multiple myeloma cell line U266GFP. Figure 2A shows that NK92 cell line, although prestimulated with 100 IU/mL of IL2, was far less cytotoxic than NK92^IL2 or NK92^CIRB. NK92^IL2 and NK92^CIRB cells showed equivalent cytotoxicity toward U266GFP, suggesting comparable levels of activation in vitro. NK92 cell line uniquely lacks killer immunoglobulin receptor (KIR). As a result, the cell surface expression of MHC-I in tumor cell lines should not account for a better or worse susceptibility to NK-cell–mediated killing. We sought to confirm this for NK92^CIRB by first determining the levels of the dominant MHC-1 antigen HLA-A2 in a panel of five human cancer cell lines. Figure 2B shows that, except for PC-3, the tumor cell lines we used in this study all express high HLA-A2 levels.

We then compared anticancer activity of NK-cell lines against five human cancer cell lines. In one experiment (Fig. 2C) cells were plated 24 hours prior to adding NK-cell lines. At an E:T ratio of 2:1, the cytotoxicity of NK92^IL2 and NK92^CIRB were generally equivalent with a slight edge to NK92^CIRB and superior to NK92. In other experiments, cancer cells were plated only 5 hours prior to adding NK cells (Fig. 2D). Under these conditions, the cytotoxicity of all NK-cell lines was greater than in cells plated for 24 hours. However, NK92^CIRB showed more cytotoxicity than NK92^IL2, at most E:T ratios. This difference was more evident with the most resistant cancer cell lines U251GM and Panc-1. As expected, MHC-1 antigen HLA-A2 expression is irrelevant to NK92^CIRB-mediated cytotoxicity.

TGFβ1 is an immunosuppressor overexpressed in the TME, known to inhibit NK cells' functions by destabilizing several activation signals (23). The glucocorticoid dexamethasone impairs the function of lymphocytes in part by suppressing IL2 production from CD4⁺ T cells (24). IL4 was reported to inhibit the proliferation of NK cells (25). We tested the effects of these immunosuppressors on NK92^IL2 and NK92^CIRB by culturing cells for 6 days in the presence of TGFβ1 (10 ng/mL), IL4 (10 ng/mL), or dexamethasone (1 μmol/L). Figure 3A shows that NK92^IL2 did not survive the exposure to dexamethasone. Their proliferation was inhibited strongly by TGFβ1 and to some extent by IL4. In contrast, NK92^CIRB proliferation was not significantly affected by TGFβ1 or IL4 and was weakly inhibited by dexamethasone (Fig. 3B).

The effects of TGFβ1 (20 ng/mL), and dexamethasone (0.5 μmol/L) pretreatments for 24 hours on NK-cell cytotoxicity were then evaluated against five cancer cell lines at an E:T ratio of 2:1 (Fig. 3C). NK cells affected more killing in these experimental conditions as cancer cells were added to already plated NK92 cells and are more vulnerable if not already attached. In addition, NK92 cells were plated with IL2 in the media (20 IU/mL) and are therefore more active than in other experiments. NK92^CIRB cytotoxicity against MDA-MB-231, PC-3, and HepG2 was not affected. Dexamethasone severely reduced cytotoxicity of NK92^IL2 toward all cancer cell lines. Similarly, TGFβ1 significantly reduced NK92^IL2 cytotoxicity against most cancer cell lines except for MDA-MB-231 and HepG2. Surprisingly, NK92 cells also showed resistance to dexamethasone inhibition in MDA-MB-231 and HepG2, probably due to IL2 presence in the media. Overall immunosuppression of NK92^CIRB was weaker than in the NK92 and NK92^IL2 lines and was also dependent on the target cancer cell line. Flow cytometry analysis shows both dexamethasone (1 μmol/L) and TGFβ1 (10 ng/mL) affected NKP30 expression dramatically in NK92^IL2 (Fig. 3D), whereas NK92^CIRB resisted dexamethasone and experienced a marginal effect under TGFβ1. NKP44 expression was equally but marginally reduced, whereas NKP46 was not affected in both cell lines. Annexin V staining shows that only dexamethasone induced moderate apoptosis in NK92^IL2 (11.5% cell death), whereas NK92^CIRB apoptotic cells represented approximately 5%–6%. Therefore, in NK92^IL2, the reduced cell growth under TGFβ1 and dexamethasone is probably caused by slow cell growth to which NK92^CIRB cells are resistant.

In accordance with their original characterization (26), NK92 cells were found CD56⁺, CD3⁻, CD16⁻, CD25⁺, CD45⁺, and NKg2D⁻ (Fig. 4A). NK92^CIRB cells expressed higher amount of CD16 and lower amounts of CD25 than both IL2-stimulated NK92 and NK92^IL2. We also examined NK92-MI, and in accordance with a previous report (20), did not find any expression of CD16 (Fig. 4B). Similarly, CD16 was undetectable in NK92 treated with either glycosylated or nonglycosylated IL2 (Fig. 4C). Freshly isolated human NK (hNK) cells displayed different patterns of marker expression. Unlike NK92, the cell lines NK92^IL2, NK92^CIRB, and hNK cells are all CD16⁺, with expression levels of hNK > NK92^CIRB > NK92^IL2 (Fig. 4D). hNK cells express much higher levels of CD16 and NKG2D than all NK92 cell lines and are virtually CD25⁻. Of the NK92 cell lines, NK92^CIRB expression for CD25 and CD16 were most similar to hNK.

We next examined the impact of CD16 expression on ADCC using trastuzumab against an HER2-positive breast cancer cell line BT474. Figure 4E shows that in the absence trastuzumab and when using an E:T ratio of 2:1, BT474 cells were not affected by the direct cytotoxicity of NK cells. However, in the presence of 1 μg/mL trastuzumab, both NK92^CIRB and NK92^IL2 exerted substantial cytotoxicity of about 60% and 50%, respectively. Trastuzumab alone at 1 μg/mL did not affect BT474 survival significantly. Similarly, NK92 did not provoke any significant cytotoxicity against BT474 in the presence of trastuzumab.

Stat 5 phosphorylation, a downstream signaling associated with IL2 receptor activation, was nearly identical between IL2-stimulated NK92 (100 IU/mL) and NK92^CIRB and modestly higher by 17% in NK92^IL2, suggesting an equivalency in activation signaling (Fig. 5A). However, qRT-PCR analysis (Fig. 5B) of NK92^CIRB revealed differences in the expression of natural cytotoxicity receptors NKP30 (1.7-fold), NKP44 (9-fold), and NKP46 (1.4-fold) compared with NK92^IL2 and NK92 stimulated with IL2 for 48 hours. In NK92^IL2, NKP44 expression also increased (3.3-fold). While perforin-1 expression was similar in all cell lines, granzyme-B expression declined and TNFα increased in NK92^IL2 and both declined marginally while IFNγ increased in NK92^CIRB. We sought to validate these data by investigating the protein expression of NK receptors (Fig. 5C) and effectors (Fig. 5D) by flow cytometry. A robust FlowJo statistic module revealed striking correlations between the RNA levels seen in Fig. 5B and the histograms' ∑median protein expression calculated in Fig. 5E. To further delineate the major contributor(s) to NK-cell lines' cytotoxicity against cancer cells, we determined the excess ratio of each protein between cell line pairs. Figure 5F shows that NKP44 and NKP30 are the main differentiators between NK92^IL2 and NK92 with an excess of 202% and 24% respectively. In NK92^CIRB, this excess over NK92 is 303% and 61%. However, an excess of perforin-1 (21%) makes NK92^CIRB far more potent than NK92^IL2. Indeed, a direct comparison of NK92^IL2 and NK92^CIRB shows the latter to contain about 30% more of NKP30, 34% of NKP44, and 28% of perforin-1. We also note that NKP30 expression was downregulated by dexamethasone and TGFβ1 in NK92^IL2 rendering these cells less potent against cancer cells (Fig. 3D). Therefore, we deduce that NKP30, NKP44, and perforin-1 are distinguishing features of NK92^CIRB from NK92 and NK92^IL2.

The survival and systemic circulation of nonirradiated NK92^IL2 and NK92^CIRBin vivo were evaluated in the context of U251MG tumor-bearing animals. Figure 6A shows that within 24 hours of the first injection of live NK92^CIRB cells, rapid tumor volume regression of 46% was observed, while NK92^IL2 cells caused 35% reduction. In contrast, tumors continued to grow in untreated animals to reach a maximal limit size nearing 200 mm³ before regressing. This size-dependent limited growth is due to the poor angiogenesis of these tumors, which can be improved by VEGF expression (27). Tumor regression for NK92^CIRB-treated group continued after the second injection while the tumors in NK92^IL2-treated animals resumed growth and did not respond until day 18. Three weeks later, the untreated and NK92^IL2 groups showed a similar tumor size. In comparison, the NK92^CIRB-treated group displayed a significant tumor volume reduction of 86%. Seventeen days after NK-cell injections, blood from mice was analyzed by flow cytometry for cells expressing mCherry and human CD45. Figure 6B shows that only NK cells expressing CIRB, but not IL2, can persist in tumor-bearing animals.

FDA requires NK92 cell irradiation between 5 and 10 Gy prior to infusion to prevent proliferation. The irradiated NK92 cells' viability declines dramatically within 2 days. NK92^IL2 and NK92^CIRBwere irradiated at 10 Gy (0.83 Gy for 12 minutes) and then plated in complete NK92 media to determine their survival using Trypan blue every 24 hours. Figure 6C shows that 24 hours postirradiation, 57% of NK92^CIRB and 45% of NK92^IL2 cells survive. The survival advantage of NK92^CIRB was statistically significant at days 1 and 2 (*, P < 0.05).

Androgen receptor and PSA-negative PC-3 cells (28) form very aggressive tumors in Nod/Scid mice. When tumor volumes reached approximately 200 mm³ (day 28), irradiated NK cells (500 cGy) were administered as 4 weekly injections via the tail vein. Figure 6D shows that the growth of PC-3 tumors in the NK92^CIRB-treated group was slowed after the first injection. After the last NK92^CIRB injection, a significant tumor growth delay of about 17 days was recorded in the period between first and fourth NK92^CIRB injections (**, P < 0.01), comparatively to the untreated group. In contrast, the NK92^IL2-treated group tumors produced only a tumor delay of 7 days from the untreated tumor group (*, P < 0.05).

To better understand the events occurring during NK92^CIRB andNK92^IL2 contact with cancer cells, NK cells were cocultured with PC-3 cells for 3 hours, after which, their expression profiles for NKP30, NKP44, NKP46 (Fig. 7A), perforin-1, granzyme-B, TNFα and IFNγ (Fig. 7B), were compared with those in NK cells grown alone. A robust statistical analysis of histograms ∑Median expression using FlowJo software revealed striking differences in activation between NK92^IL2 andNK92^CIRB (Fig. 7C; Supplementary Table S2). While both cell lines have within 3 hours discharged 17.5% of their perforin-1 content into PC-3 cells, granzyme-B, TNFα and IFNγ, were reduced in NK92^IL2, suggesting release, while they increased in NK92^CIRB by 36%, 21%, and 16%, respectively. As NK92^CIRB released perforin-1 and because granzyme-B is codelivered with perforin-1 (29), this increase suggests replenishment of these effectors, which are usually released within minutes of contact with cancer cells. NKP46 remained unchanged in NK92^CIRB, whereas surface expression of NKP30 and NKP44 were dramatically reduced by 23% and 18%, respectively, but only by 7% and 11% in NK92^IL2. This reduction in surface markers upon contact with cancer cells suggests receptor internalization that occurs upon their activation by ligands on cancer cells. The overexpression of NKP30 and NKP44 on NK92^CIRB compared with NK92^IL2 could drive a stronger activation that would explain granzyme-B, TNFα, and IFNγ high levels compared with NK92^IL2. After 3 hours of contact with cancer cells, NK92^CIRB cells still harbored all receptors and effectors in excess over NK92^IL2 (Supplementary Table S2).

IL2 can bind to receptor IL2Rα (CD25) with low affinity (30), or with an intermediary affinity to IL2Rβ (CD122) when associated with the common IL2Rγ chain (CD132; refs. 31, 32). However, this binding affinity is higher when all three receptors are combined (33). Adult NK cells may express only IL2Rβ and IL2Rγ subunits (34) and are, therefore, relatively insensitive to low doses of IL2, but acquire sensitivity upon IL2Rα expression (35). A recently developed IL2 “superkine” (36) that bypasses IL2Rα by binding directly and with high affinity to IL2Rβ produced better antitumor effects than wild-type IL2 in mice. However, it still causes some form of pulmonary edema.

We created a novel chimera CIRB made of IL2 and its receptor IL2Rβ, joined by a peptide linker derived from the extracellular domain of IL2Rα. The linker was computationally determined as reasonably flexible, without adversely affecting the chimera stability, which is generally inversely correlated to flexibility (37). CIRB induces indefinite cell expansion and conferred an in vitro cytotoxicity similar or higher than that elicited by IL2 expression. In vivo, the anticancer activity of NK92^CIRB against solid tumors was superior to that elicited by NK92^IL2. In addition, CIRB confers, in contrast to IL2, substantial resilience to TGFβ1, dexamethasone, and IL4. This advantage could be crucial in the TME where TGFβ1 is secreted by a variety of cells including cancer-associated fibroblasts (38), and exists in a membrane-bound form on Tregs to induce NK cells' anergy (39), or by MDSCs to inhibit NKG2D expression, and IFNγ production in NK cells (40). Cancer cells also regularly shed tumor-derived exosomes containing membrane-bound TGFβ1 that downregulate NKG2D (41), and inhibit IL2 signaling (42). TGFβ1 mediates NK inhibition by an induced miR-183 that represses the coactivator/adapter DAP12 expression, thus destabilizing several activation signals in NK cells (23). CIRB expression in NK92^CIRB cells also provides resistance to dexamethasone, whereas NK92^IL2 cells' growth was halted. Dexamethasone impairs the function of lymphocytes in part by suppressing IL2 production from CD4⁺ T cells, and reducing the activation receptors NKG2D and Nkp46 in NK cells (24). Glucocorticoid hormones can interfere with macrophage activation and antigen presentation, and repress the transcription of several proinflammatory cytokines, chemokines, cell adhesion molecules, and enzymes involved in the inflammatory response (43). The extreme sensitivity of NK92^IL2 to dexamethasone, could be explained by the reported destabilization of IL2 RNA (44), which could potentially occur in NK92^IL2 but not when IL2 is fused with IL2Rβ RNA as in NK92^CIRB.

CIRB, and to a lesser degree the stable expression of IL2, allowed substantial CD16 expression in NK92. However, exogenous recombinant IL2 was not able to mediate such expression. Similarly, NK92-MI, which produces and secretes IL2, was found deficient in CD16, as reported previously (20). When combined with trastuzumab CD16 expression further enhanced NK92^CIRB andNK92^IL2 cytotoxicity by ADCC. Interestingly, CD25 expression declined dramatically in NK92^CIRB (Fig. 4A). This unique phenotype of substantial CD16 and low CD25 expression can synergize with the ADCC-mediated action of approved therapeutic antibodies. CIRB induced substantial RNA expression of NKP44 (9-fold), NKP46 (1.4-fold), and NKP30 (1.7-fold) as well as a modest but significant increase in IFNγ. As expected, CIRB induced protein expression increase for NKP44 (4-fold) and NKP30 (1.6-fold). Finally, perforin-1 expression increased slightly by 21% in NK92^CIRB.

Furthermore, upon contact with PC-3 cancer cells, NK92^CIRB cells experienced an increase in granzyme-B, TNFα, and IFNγ, whereas NKP30, NKP44, and perforin-1 contents decreased, suggesting activation and internalization of these receptors with release of perforin-1. These features might explain the superior anticancer effect of NK92^CIRBin vivo. We note that NKP30 was downregulated in NK92^IL2 treated with dexamethasone and TGFβ1 (Fig. 3D), which led to a reduced anticancer activity (Fig. 3C). This could suggest a prominent role for NKP30 in NK92^CIRB where it is overexpressed and might offer resistance against the action of dexamethasone and TGFβ1.

Genetic modifications introducing CD16 in NK cells were shown to increase NK-cell–mediated ADCC against multiple myeloma when combined with elotuzumab (45). CD16 induction only in NK92^CIRB and NK92^IL2, but not in NK92-MI or NK92 stimulated with IL2, could be explained by the persistent IL2 signaling that somehow translates into stronger activation and growth of NK92^CIRB and NK92^IL2. In fact, the growth rates of both NK92^CIRB and NK92^IL2 were 2-fold that of NK92-MI (Supplementary Fig. S2). Another indication of higher activation of NK92^CIRB and NK92^IL2 is the dramatic induction of NKP44, compared with NK92 stimulated with IL2 for 48 hours. In addition, NK92^CIRB can proliferate in vivo far longer and also have a better survival after irradiation than NK92^IL2 cells. They also surpass that of NK92-MI when exposed to similar conditions (20). In vitro, NK92^IL2 cells secrete sufficient IL2 to sustain their activation and proliferation. In vivo, however, they may not be able to produce enough IL2 extracellular concentrations to sustain proliferation. This could be compounded by the competition for IL2 by Tregs and other immune cells in an immunocompetent animal.

Many encouraging advances have been achieved in NK-cell–directed immunotherapy (46). However, the increasing demand for NK cells' expansion ex vivo requires both highly activated cells and reduced costs of cell expansion. Cellular immunotherapy using donor NK cells is an emerging field that could achieve significant anticancer effects, safely and without the risk of inducing graft-versus-host disease. This safety feature as well as the off-tumor/on-target toxicity are currently hindering the success of CAR-T technology (47). Several NK-cell lines (Khyg-1, NKL, NKG, NK-YS, YT, YTS, and HANK-1 cells) are currently used in preclinical studies. However, only NK92 was extensively evaluated for its safety and efficacy in clinical settings (48, 49). Unlike primary hNK cells, NK92 cells constitute a stable and homogenous population, amenable to genetic modification by lentiviruses, a gene transfer platform that has shown a good safety profile for lymphocytes (50). Our novel strategy of fusing interleukins to their receptors is a novel platform that can achieve a far better cytokine activation, with specificity, and without systemic toxicity or competition by other cellular components of the immune system. Self-activation of NK cells provides several distinguishing features, such as resilience to TGFβ1 or glucocorticoid hormones, substantial expression of CD16, higher survival after irradiation, and a superior antitumor activity in vivo. This new cytokine–cytokine receptor platform may be further improved to enhance NK cells' anticancer and possibly antiviral activity.

No potential conflicts of interest were disclosed.

Conception and design: Y. Jounaidi, J.F. Cotton, K.W. Miller, S.A. Forman

Development of methodology: Y. Jounaidi, J.F. Cotton, S.A. Forman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Jounaidi, J.F. Cotten

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Jounaidi, S.A. Forman

Writing, review, and/or revision of the manuscript: Y. Jounaidi, J.F. Cotton, K.W. Miller, S.A. Forman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Jounaidi

Study supervision: Y. Jounaidi

The authors wish to thank Dr. Rooma Desai of this laboratory for helpful comments and suggestions and Prof. Raje Nopoor from the Center for Multiple Myeloma, Massachusetts General Hospital Cancer Center for her very helpful advice and suggestions.

This work was supported in part by a DAPPCM Innovation Grant 226026 (to Y. Jounaidi), NIH grant GM089745 (to S.A. Forman), and NIH grants GM058448 (to K.W. Miller) and HL117871 (to J.F. Cotten).

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.