Potent Cytolytic Activity and Specific IL15 Delivery in a Second-Generation Trispecific Killer Engager

Natural killer (NK)-cell–based immunotherapies are quickly gaining traction preclinically and in early-phase clinical trials (1). This is due to the ability of NK cells to lyse tumors with minimal priming, their lack of MHC restriction, their capacity to exert antibody-dependent cellular cytotoxicity (ADCC), and to robustly and rapidly produce inflammatory cytokines (2). A number of NK-cell immunotherapy modalities are being actively explored, including cellular, cytokine, checkpoint inhibition, and ADCC-driven therapies (3). Developing affordable NK-cell–based immunotherapeutic off-the-shelf products will significantly strengthen clinical practice.

Both cytokine- and ADCC-mediated immunotherapies are ideal candidates for off-the-shelf approaches. NK cells express cytokine receptors that modulate NK-cell effector function, development, proliferation, and homeostasis. IL2 and IL15 activate a host of downstream signaling molecules that result in enhanced function, proliferation, and survival (4, 5). Because of reduced toxicities and its NK-cell–stimulating effects without the augmentation of regulatory T cells, IL15 has become a priority cytokine of interest in NK immunotherapy (6, 7). Recombinant human IL15 and an IL15 superagonist complex (N-803, formerly ALT-803) are under investigation in several tumor settings alone and in combination with adoptive transfer in patients with leukemia (8–11). Early clinical data indicate that IL15 can induce NK-cell expansion necessary to enhance immunotherapy (12, 13). IL15 also has the ability to prime CD16 signaling on NK cells (14). Among these approaches, NK-cell–driven function through mAbs and through bi- and trispecific engagers is of significant clinical interest (15–19). Our group previously described a platform, termed tri-specific killer engager (TriKE), which merges the concepts of cytokine signaling and bispecifics into a single molecule to maximize NK-cell immunotherapeutic potential (3).

TriKE molecules are composed of three arms: a single-chain variable fragment (scFv) targeting the activating receptor CD16 on NK cells, an scFv targeting a tumor-associated antigen (TAA), and an IL15 moiety to drive NK-cell expansion, survival, and priming (3, 16). We previously described TriKE molecules against a host of TAAs, including a molecule targeting CD33 meant to enhance NK-cell immunotherapy against myeloid malignancies including acute myeloid leukemia (AML) and myelodysplastic syndromes (MDS), and other CD33-expressing malignancies such as systemic mastocytosis (20–22). This first-generation TriKE molecule, termed 161533 in previous studies, is called scFv16-m15-33 in this context for clarity and contains a mutant IL15 with an N72D substitution (abbreviated m15; ref. 23), and a phase I/II clinical trial is underway for treatment of refractory AML, high-risk MDS, and systemic mastocytosis (NCT03214666). This study describes the effect of substituting a “humanized” anti-CD16 single-domain, camelid antibody to replace the anti-CD16 scFv, which enabled presentation of wild-type (wt) IL15 without use of m15. The resulting molecule, cam16-wt15-33 TriKE, displays stronger IL15 signaling capabilities, better NK-cell activation, and increased NK-cell–mediated tumor control both in vitro and in vivo. Thus, we postulate that the second-generation TriKE molecule may prove more efficacious for clinical development.

Human healthy donor blood was obtained from Memorial Blood Bank, and processed to isolate peripheral blood mononuclear cells (PBMC) using density gradient Ficoll-Paque (GE Healthcare). PBMCs were either cryopreserved in liquid nitrogen with a mix of 90% FBS (Gibco) and 10% DMSO (Sigma), or were processed fresh to enrich NK cells (90%–95%) using the EasySep Human NK Cell Enrichment Kit (catalog no.:19055, Stemcell Technologies), or were CD3/CD19-depleted using Positive-Selection Kits (catalog no.:17851&17854, Stemcell Technologies).

All human samples, including AML patient blasts, were obtained after approval from the Institutional Review Board at the University of Minnesota (Minneapolis, MN) and following written informed consent, received in compliance with guidelines by the Committee on the Use of Human Subjects in Research and in accordance with the Declaration of Helsinki. PBMCs collected 28 days after double umbilical cord blood (UCB) transplant (N = 9; were UCB transplants, but used two units from separate donors to maximize stem cell input) or matched sibling donor (MSD) transplant (N = 9; four bone marrow and five G-CSF–mobilized peripheral blood stem cell grafts) were cryopreserved and stored by Translational Therapy Shared Resource (Masonic Cancer Center, University of Minnesota, Minneapolis, MN). For the patient-derived xenograft (PDX) model, AML blasts were collected between 2012 and 2016 and cryopreserved from an apheresis of a de novo (no prior chemotherapy) patient with AML with normal cytogenetics, which resulted in 87% AML blasts. AML blasts were 98.9% CD34⁺ and 99.2% CD33⁺. Samples were stored by Heme Malignancy Tissue Bank at the University of Minnesota (Minneapolis, MN), also managed by the Translational Therapy Shared Resource.

The native CD33⁺ acute promyelocytic leukemia cell line HL-60 was received from ATCC on April 2017 for use in this study. HL-60luc cells, obtained from Xianzheng Zhou in 2013 at the University of Minnesota (Minneapolis, MN; now at New York Medical College, Valhalla, NY), were authenticated using short tandem repeat DNA profiling, according to the International Cell Line Authentication Committee by the University of Arizona Genetics Core (Tucson, AZ) in March 2017. HL-60 and HL-60luc cells were used for studies outlined here from 2017 to 2020. Cells were cultured in RPMI1640 (Gibco; catalog no.: 224000-089) supplemented with 10% FBS (Gibco catalog no.: 2614-079) at 37°C, 5% CO₂. CTLL-2 cells were received from ATCC in August 2016 and also cultured in RPMI1640 supplemented with 10% FBS and IL2 (Prometheus, 100 U/mL) at 37°C, 5% CO₂. All cell lines used in studies were utilized at an early passage (P1–P3) and were Mycoplasma negative using the Universal Mycoplasma Detection Kit (ATCC) per the manufacturer's protocol.

Mouse experiments were performed after approval from and in accordance with and the guidelines of Institutional Animal Care and Use Committee at the University of Minnesota (Minneapolis, MN). NOD scid gamma (NSG) mice and NSG-SGM3 triple-transgenic mice, expressing human IL3, GM-CSF, and SCF, were obtained from The Jackson Laboratory. Age matched mice, in the 6- to 8-week-old range for the PDX model and 8- to 10-week-old range for the HL-60 model, were used in all studies.

Equimolar concentrations of scFv16-m15-33 and recombinant human IL15 (rhIL15) were serially diluted as noted in RPMI media or heat-inactivated Human AB Serum (Valley Biomedical, Inc.). The Quantikine ELISA Kit (R&D Systems) was used per the manufacturer's protocol to detect IL15. For in vivo detection of IL15, NSG mice were injected intravenously with equimolar amounts of scFv16-m15-33 (25 μg) or rhIL15 (4.85 μg). Facial vein bleeds were carried out to collect serum at 30, 60, 120, and 240 minutes after injection for IL15 quantitation. Serum samples were diluted 1:100 and 1:1,000 for the IL15 ELISA assay. An IL15 standard curve, using the manufacturer provided recombinant IL15 within the ELISA kit, was established to determine the relative concentration. A Tecan Infinite M200Pro plate reader was used to measure absorbance.

Prior to beginning of the assay, IL2 was washed off, and 2 × 10⁴ CTLL-2 cells were incubated in a 96-well plate for 48 hours with indicated concentrations of scFv16-m15-33 and rhIL15. Cells were then incubated for 1 hour at 37°C, 5% CO₂ with Resazurin (catalog no.: AR002, R&D Systems), a blue reagent that is reduced by metabolically active viable cells to form pink resorufin, identifiable at 570 nm absorbance (24). A Tecan Infinite M200Pro plate reader was used to measure absorbance.

Production steps of the second-generation TriKE cam16-wt15-33, including construction, verification, isolation, and purification, were performed as previously described for the first-generation TriKE (20, 25). The final construct of the cam16-wt15-33 NcoI/XhoI gene fragment was encoded from a start codon, a humanized camelid anti-CD16 VHH (cam16), a 20-amino acid flanking sequence (PSGQAGAAASESLFVSNHAY), wt human IL15, a 7-amino acid flanking sequence (EASGGPE), and anti-CD33 scFv to construct the TriKE sequence (patent US 2018/0282386 A1).

An Octet-based label-free binding assay (OctetRED96, ForteBio, Molecular Devices) was performed to measure binding affinity of the analytes scFv16-m15-33 and cam16-wt15-33 to the ligands CD16a and CD16b. The ligands, rhCD16A or rhCD16B with a c-terminal 6x-HIS Tag (R&D Systems, catalog no.: 4325-FC-050 and 1597-FC-050/CF, respectively) were captured on biosensor tips using His tag-specific dip and read biosensors. The biosensor coated with a ligand was dipped into buffer containing an analyte for association and then dipped into assay buffer (PBS). The binding and dissociation were measured using Bio-Layer Interferometry technology. The Octet assays were carried out by Precision Antibody.

The CellTrace Violet Proliferation Kit (catalog no.: C34557, Thermo Fisher Scientific) was used to evaluate proliferation (dye dilution) by flow cytometry. A total of 3 × 10⁵ PBMCs were labeled, treated with indicted proteins (human rhIL15 obtained from R&D Systems, catalog no.: 247-ILB) diluted in RPMI1640 (catalog no.: 224000-089, Gibco) supplemented with 10% FBS (catalog no.: 2614-079, Gibco), harvested 7 days later, and analyzed on an LSRII (BD Biosciences). For the characterization of NK and T cells, LIVE/DEAD NEAR IR (catalog no.: L34976, Thermo Fisher Scientific), PE-CY7–conjugated anti-CD56 (HCD56, BioLegend), and PE-CF594–conjugated anti-CD3 (UCHT1, BD Biosciences) were used. NK cells and T cells were identified as LiveDead⁻CD56⁺CD3⁻ and LiveDead⁻CD56⁻CD3⁺ cells, respectively. Cells were run on an LSRII flow cytometer and analyzed using FlowJo Software (BD Biosciences).

Noted human PBMCs (5 × 10⁵/well) were incubated overnight with noted treatments, harvested, and stained with Live/Dead Fixable Aqua Staining Kit (catalog no.: L-34966, Thermo Fisher Scientific), anti-CD56, anti-CD3, PE-conjugated anti-CD69 (FN50, BioLegend), and APC-conjugated anti-CD25 (M-A251, BioLegend) to evaluate NK-cell activation marker expression with flow cytometry. Cells were run on an LSRII flow cytometer and analyzed using FlowJo Software (BD Biosciences).

Flow cytometric assessment of NK-cell function was carried out as described previously (25). Briefly, following addition of noted treatments [diluted in RPMI1640 (Gibco, catalog no.: 224000-089) supplemented with 10% FBS (Gibco, catalog no.: 2614-079)] and HL-60 tumor targets [at a 2:1 effector:target (E:T) ratio], healthy donor or posttransplant patient PBMCs were stained with FITC-conjugated anti-CD107a (H4A3, BioLegend) at the beginning of a 4-hour incubation. One hour after the addition of anti-CD107a, cells were given Golgi Stop (1:1,500) and Golgi Plug (1:1,000; both from BD Biosciences), and incubated for 3 hours. Cells were then stained with the Live/Dead Fixable Aqua Staining Kit (catalog no.: L-34966, Thermo Fisher Scientific), anti-CD56, and anti-CD3, fixed in 2% paraformaldehyde (catalog no.: 50980487, Thermo Fisher Scientific), and permeabilized (catalog no.: 00-8333-56, eBioscience). Permeabilized cells were stained with BV650-conjugated IFNγ (4S.B3, BioLegend) and were evaluated with flow cytometry. Effectors (normal donor or day 28 posttransplant patient PBMCs) were cocultured with targets in each of the experiments. Cells were run on an LSRII flow cytometer and analyzed using FlowJo Software (BD Biosciences).

A total of 2 × 10⁴ HL-60 cells were labeled using CellTrace Far Red Proliferation Kit (catalog no.: C34564, Thermo Fisher Scientific). Healthy donor enriched NK cells were incubated at a 2:1 E:T ratio with CellTrace Far Red–labeled HL-60 targets with noted treatments (30 nmol/L) and with IncuCyte Caspase-3/7 Green Apoptosis Assay Reagent (1:250, catalog no.: 4440, Essen BioScience) in RPMI1640 (catalog no.: 224000-089, Gibco) supplemented with 10% FBS (catalog no.: 2614-079, Gibco). Plates were placed in an IncuCyte Zoom (Sartorius Inc.) for 24 hours. Readings were taken every 30 minutes. A graph was created representing the percentage of live HL-60 targets (CellTrace Far Red⁺Caspase3/7⁻) normalized to live targets alone at the starting time point (0 hour).

The previously described HL-60luc/human NK-cell xenogeneic NSG mouse model was implemented (20). Briefly, mice were irradiated (275 cGy), intravenously injected with 0.75 × 10⁶ HL-60luc cells, and intravenously injected with 1 × 10⁶ CD3/CD19-depleted human NK cells 3 days later; and treated for 2 weeks intraperitoneally, 5 days a week for the 2 weeks, with different concentrations (90, 30, or 10 μg/injection) of the second-generation cam16-wt15-33 TriKE or the first-generation scFv16-m15-33 TriKE. At days 7, 14, and 21, tumor load was assessed after d-luciferin (GoldBio) intraperitoneal injection (150 mg/kg) using Bioluminescence Imaging (IVIS 100 In Vivo Imaging System). Mice were bled via facial vein on day 21 (100 μL blood/mouse), red blood cells lysed using a quick (45 second) water lysis, then resuspended in PBS, and cells were stained with Brilliant Violet 605–conjugated anti-human CD45 (HI30, BioLegend), anti-CD56, and anti-CD3. Data were acquired on an LSRII flow cytometer for 60 seconds/sample and analyzed using FlowJo Software (BD Biosciences).

An AML PDX model was used to compare the activity of cam16-wt15-33 with rhIL15. Six- to 8-week-old NSG-SGM3 mice were irradiated (125 cGy) and injected intravenously the next day with 2 × 10⁶ primary AML cells. Progress of AML engraftment was monitored with weekly facial vein bleeds stained for anti-CD45 and APC-conjugated anti-CD33 (WM53, BioLegend) to determine the proportion of CD45⁺CD33⁺ blast cells until detection of at least 1% human AML in blood was achieved (about 6 weeks after AML injection). Mice then received 3 × 10⁶ CD3/CD19-depleted NK cells and subsequently received intraperitoneal treatments of either 5 μg rhIL15 three times a week or 30 μg cam16-wt15-33 five times a week for a total of 3 weeks. Mice were euthanized and blood, bone marrow, and spleen were harvested on day 21. Blood and splenocytes were stained with anti-human CD45, anti-CD56, and anti-CD3 to determine CD45⁺CD56⁺CD3⁻ NK-cell expansion. Bone marrow was stained with anti-human CD45, FITC–conjugated anti-CD34 (561, BioLegend), and anti-CD33 to determine CD45⁺CD33⁺ AML blasts.

GraphPad Prism (GraphPad Prism Software, Inc.) was used to calculate statistics via one-way ANOVA or paired t test methods (as indicated), and to generate graphs with error bars showing the mean ± SEM and statistical significance coded as *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001.

We previously reported a TriKE composed of a scFv against CD16 (scFv16), an N72D m15 shown to signal better than wtIL15, and an scFv against CD33 termed scFv16-m15-33 (20, 23). Because versions of this molecule containing a wtIL15 moiety were not functional, the mIL15 version of this first-generation TriKE, scFv-m15-33, was selected. Although this molecule displays cytotoxic activity mediated by NK-cell engagement and proliferative activity by the IL15 domain, we hypothesized that the activity of the IL15 moiety might be compromised in the first-generation TriKE backbone. To test this, IL15 protein detection in the scFv16-m15-33 TriKE was compared with detection of equimolar amounts of rhIL15 using an ELISA assay (Fig. 1A). Data indicated a significant reduction in IL15 detection when comparing the scFv16-m15-33 TriKE with rhIL15. Results were identical when comparing detection in media (RPMI with 10% FBS) or human serum.

NSG mice were injected with equimolar concentrations of rhIL15 or the scFv16-m15-33 TriKE, and serum was analyzed for 4 hours after administration. IL15 at the 30- and 60-minute post-administration time points were significantly greater after rhIL15 injection compared with the scFv16-m15-33 TriKE (Fig. 1B). At later time points (2 and 4 hours), detection was low for both because of rapid clearance based on size (12.8 kDa for rhIL15 vs. 66.05 kDa for the scFv16-m15-33 TriKE). To determine whether functional activity was altered, the scFv16-m15-33 TriKE was compared with rhIL15 over a concentration range using an assay that detects NADP activity on CTLL-2 cells that signal through IL15 but do not contain CD16 or CD33 receptors (Fig. 1C). Together, these results indicated that m15 in the TriKE signals with about 25-fold less activity than rhIL15 when presented with the scFv against CD16.

IL15 detection and signaling data indicate that incorporation of IL15 into the TriKE resulted in reduced activity. Because the scFv16-m15-33 TriKE contains multiple scFvs, each of which contained a variable heavy (V_H) and variable light (V_L) chain, the impact on diminished IL15 activity may be mediated by inefficient folding of the scFvs (Fig. 1D). Previous studies show that molecules containing multiple scFvs have potential for aggregation and nonspecific folding (26–29). The V_H of one scFv aggregates to the V_L of the other scFv and vice versa. To overcome this challenge, a single-domain antibody (sdAb or V_HH) sequence was substituted for the scFv anti-CD16 binding domain. Because the sdAb contained only one V_H domain that could not aggregate with a V_L from an scFv, the TriKE molecule was forced to fold precisely without mispairing (Fig. 1E). To accomplish this, the CDR1, CDR2, and CDR3 regions of a camelid (llama) anti-CD16 sdAb were cloned into a humanized sdAb backbone, termed cam16 within this study (30, 31). The humanized sdAb backbone, also derived from a camelid, contains four natural amino acid changes generated through evolution that prevent binding of the sdAb to V_L fragments. The backbone was humanized by alteration of 12 amino acids to increase the humanness score, as described previously (31). The cam16 was then cloned into a TriKE backbone containing wtIL15 and an scFv anti-CD33 (cam16-wt15-33 TriKE) or a m15 and anti-CD33 (cam16-m15-33). Because no difference was detected in proliferative activity between the cam16-wt15-33 TriKE versus the cam16-m15-33 TriKE, the wt version was selected for development (Supplementary Fig. S1A and S1B).

To compare the binding capabilities of the cam16-wt15-33 and scFv16-m15-33 TriKEs, Octet analysis was performed to measure the affinity of each TriKE against the CD16A and CD16B proteins (Fig. 1F). Addition of the cam16 arm in the cam16-wt15-33 TriKE improved binding affinity by 28-fold against CD16A compared with the scFv16-m15-33 TriKE (KD = 2.32 × 10⁻⁸ vs. 6.56 × 10⁻⁷ mol/L, respectively). The cam16-wt15-33 TriKE had stronger binding against CD16B than the scFv16-m15-33 TriKE (KD = 3.40 × 10⁻⁸ vs. 1.44 × 10⁻⁶ mol/L, respectively). Both molecules bound CD33 with relatively similar affinity (KD = 2.56 × 10⁻⁸ mol/L for cam16-wt15-33 vs. 2.21 × 10⁻⁸ mol/L for scFv16-m15-33; Supplementary Fig. S1C and S1D). Data indicate that the second-generation cam16-wt15-33 TriKE could bind CD16 with greater affinity than the first-generation scFv16-m15-33 TriKE.

To investigate the ability of this new construct to drive IL15 signaling, NK-cell proliferation was evaluated on CellTrace-labeled PBMCs incubated with different concentrations of rhIL15, scFv16-m15-33, or cam16-wt15-33. Total NK-cell proliferation was highest with cam16-wt15-33 at all concentrations (Fig. 2A), and NK-cell proliferation beyond three divisions (Fig. 2B) was highest in the cam16-wt15-33 group at concentrations of 5 nmol/L or greater. Although, we have previously shown that TriKEs can synergize CD16 and IL15 signaling to enhance phospho-STAT5 signaling, this synergy only occurs in the presence of targets needed to cross-link CD16 (21). In the absence of HL-60 targets, IL15 activity increased independently of the CD16 signaling. The IL15 specificity was tested by evaluating T-cell proliferation (Fig. 2C and D). Despite driving the most NK-cell proliferation, the cam16-wt15-33 TriKE drove T-cell proliferation significantly less than rhIL15 and the scFv-m15-33 TriKE at the highest concentration tested. Results indicated that the cam16 engager induced delivery of the wt15 within the TriKE backbone, specifically to NK cells, without cross-stimulation to T cells. Because IL15 signaling can also induce expression of activation markers CD69 and CD25, NK cells were incubated with rhIL15, scFv16-m15-33, or cam16-wt15-33 overnight and expression of activation markers was evaluated. With the exception of the lowest dose, CD69 proportion on NK cells was similar, but the median fluorescence intensity (MFI) for CD69 was higher at all concentrations tested with the cam16-wt15-33 TriKE compared with the scFv16-m15-33 (Fig. 2E and F). The proportion and median florescence of CD25-expressing NK cells was substantially higher with cam16-wt15-33 treatment compared with scFv16-m15-33 (Fig. 2G and H). Compared with rhIL15 at the 30 nmol/L concentration, the cam16-wt15-33 TriKE median CD25 expression was higher. These data indicated that IL15 in the cam16-wt15-33 TriKE backbone was stronger than that on the scFv16-m15-33 TriKE and similar to rhIL15.

We next evaluated the ability of the cam16-wt15-33 TriKE to activate NK-cell effector functions against CD33-expressing targets. To determine the differential in function between the second-generation cam16-wt15-33 TriKE and the first-generation scFv16-m15-33 TriKE, NK cells were incubated with tumor targets and a gradient of TriKE concentrations (61.7 pmol/L to 15 nmol/L). Degranulation and IFNγ expression were determined after 4 hours (Fig. 3A and B). The second-generation cam16-wt15-33 TriKE was functionally superior to the first-generation scFv16-m15-33 TriKE throughout the spectrum of concentrations. Increased functionality was evident at the 30 nmol/L concentration compared with rhIL15 (Fig. 3C and D). To evaluate this function over time, we tested the TriKEs in an IncuCyte tumor killing assay (Fig. 3E and F; Supplementary Fig. S2A). Enriched NK cells were plated with CellTrace Far Red–labeled HL-60 tumor targets, the indicated drugs, and caspase 3/7 green reagent. The number of live tumor cells (CellTrace⁺caspase 3/7⁻) was evaluated every 30 minutes for 24 hours. The cam16-wt15-33 TriKE induced more rapid and robust killing than the scFv16-m15-33 TriKE or controls at all time points tested, and increasing the E:T ratio increased killing of the HL-60 cells in the presence of the cam16-wt15-33 TriKE, as expected (Supplementary Fig. S2B).

We compared short exposure with the TriKE molecules to better understand binding characteristics of the two molecules to effectors or targets, which may be important clinically. Effector function (degranulation and cytokine production) was tested after effector or target short-term exposure (15 minutes) to TriKE molecules or rhIL15 followed by coculture of cells. When PBMCs were incubated with TriKE molecules or rhIL15 for 15 minutes, washed, and then cultured with HL-60 tumor targets for 4 hours, the cam16-wt15-33 TriKE induced significantly higher degranulation (CD107a expression) and IFNγ production on the NK cells. In contrast, the scFv16-m15-33 and rhIL15 exhibited weak activity after short-term exposure (Fig. 4A). A similar pattern was seen when HL-60 cells were incubated for 15 minutes with TriKEs, washed, and then incubated with PBMCs to evaluate NK-cell function (Fig. 4B). Because binding to the HL-60 targets via the same anti-CD33 component should be identical between the two TriKEs, the increased activity in this setting was likely because of enhanced signaling mediated by the cam16-wt15-33 once bound. However, it should be pointed out that the short-term cam16-wt15-33 TriKE exposure to PBMCs yielded stronger activation than short-term exposure to HL-60 targets, perhaps highlighting differences in CD16 binding affinity between the two TriKEs. Other exposure times (5 minutes, 30 minutes, and 1 hour) were tested and yielded similar results (Supplementary Fig. S2C).

To evaluate the impact on proliferation, CellTrace-labeled PBMCs were incubated with TriKEs or rhIL15 for 15 minutes, washed, and then incubated for 7 days (Fig. 4C). The cam16-wt15-33 TriKE induced the greatest proliferation, followed closely by the scFv16-m15-33 TriKE. rhIL15 alone did not induce proliferation. This differential likely stemmed from differences in binding stability mediated through the cam16 or scFv16, as CD16 signaling alone is not active in this setting with no targets present.

To evaluate the impact of TriKE molecules on NK cells posttransplant, patient PBMCs were collected 28 days after transplant, at a time when IFNγ production defects have been shown to occur despite high early engrafting NK-cell numbers (32). Eighteen allogeneic hematopoietic cell transplant (alloHCT) recipients, from UCB and MSD transplants, were studied (Supplementary Table S1). Recipients received a mix of conditioning regimens (myeloablative and reduced intensity), but had high proportions of NK cells [50.70% ± 8.07% (mean/SEM) for UCB vs. 37.57% ± 4.44% for MSD]. CD16 expression was variable from patient to patient and differed on the basis of transplant source (59.39% ± 4.30% in UCB recipients vs. 50.01% ± 6.00% in MSD recipients). AlloHCT samples and targets were treated in a standard 4-hour assay (no priming), or were preincubated with treatments overnight, to determine the priming effect of the IL15 moiety in the TriKEs, and then incubated for 4 hours with tumor targets. Under “no priming” conditions, the cam16-wt15-33 TriKE induced higher alloHCT NK-cell degranulation than all other groups, but little inflammatory cytokine production (Fig. 5A and B; Supplementary Fig. S3A and S3C). Overnight incubation led to strong degranulation against HL-60 targets with all treatments (Fig. 5C; Supplementary Fig. S3B and S3D, top). In contrast, inflammatory cytokine production of overnight-incubated alloHCT NK cells resulted in similar IFNγ production between the rhIL15 and cam16-wt15-33 groups, which induced superior IFNγ production compared with the scFv16-m15-33 TriKE (Fig. 5D; Supplementary Fig. S3B and S3D, bottom). Taken together, the data indicated that short exposure to the cam16-wt15-33 TriKE induced better NK-cell degranulation in alloHCT NK cells than all other treatments, and the IL15 moiety within the cam16-wt15-33 TriKE could induce similar priming as rhIL15 with longer exposure.

Two different models were used to evaluate the efficacy of the cam16-wt15-33 TriKE in vivo. One employed intravenous injection of a luciferase-labeled AML cell line (HL-60luc) into NSG mice to track tumor development using live imaging. Three days after tumor inoculation, NK cells were injected intravenously, and mice were treated 5 days a week for 2 weeks with different concentrations (90, 30, or 10 μg/injection) of the second-generation cam16-wt15-33 TriKE or the first-generation scFv16-m15-33 TriKE (Fig. 6A). The mice were evaluated for tumor burden using bioluminescence over the course of the 3 weeks (Fig. 6B; Supplementary Fig. S4A and S4B). Data showed that at 90 and 30 μg doses, the cam16-wt15-33 TriKE treatment induced significantly better tumor control (130-fold and 12-fold, respectively) than the scFv16-m15-33 TriKE (Fig. 6C). The differential in proliferative activity of the cam16-wt15-33 TriKE with the scFv16-m15-33 TriKE is illustrated by higher NK-cell counts in the blood at day 21 at all doses tested (Fig. 6D).

The second in vivo system used a PDX tumor model in which primary AML blasts from a patient with cancer were injected into NSG-SGM3 mice (blast characteristics found in Supplementary Fig. S4C), and tumors were allowed to grow for about 6 weeks. Once the tumor was established, NK cells were intravenously injected, and mice were treated for 3 weeks with cam16-wt15-33 TriKE or rhIL15 (intraperitoneally) to evaluate therapeutic efficacy (Fig. 6E). Absolute leukemia burden in the bone marrow and NK-cell numbers in the blood were then assessed. cam16-wt15-33 TriKE treatment resulted in less tumors in the bone marrow than with rhIL15 treatment (Fig. 6F), despite lower NK-cell numbers found in the blood (Fig. 6G). NK-cell numbers did not differ in the spleen, indicating that there might be differences in localization/effect of rhIL15 versus cam16-wt15-33 TriKE (Supplementary Fig. S4D). In vivo data indicated that the cam16-wt15-33 TriKE induced better AML tumor control.

The TriKE platform is unique among NK engagers because it incorporates cytokine signaling, specific tumor targeting, and activation of ADCC within a single molecule (16–19). TriKEs not only drive specific tumor killing but also induce NK-cell proliferation and survival via the IL15 moiety. Unlike T cells (e.g., CD3 stimulation in bispecific T-cell engagers), which can be induced to expand through cross-linking of activating receptors, NK cells require a cytokine signal to induce proliferation. Although, activating receptor signals may synergize with cytokines to enhance proliferation, NK-cell–activating receptors, including CD16, do not induce proliferation on their own (33). The first-generation TriKE, scFv16-m15-33 (161533 in previous studies), provides both cytokine and activating signals (20). Our findings indicated that IL15 functionality within this molecule is limiting. The IL15 moiety within the first-generation TriKE showed a 25-fold less function compared with rhIL15. This decrease may be influenced by steric hindrance mediated by mispairing of the two scFvs (34). We, therefore, substituted in a humanized sdAb against CD16 incapable of pairing with the light chain of the anti-CD33 scFv, resulting in improved function of the IL15 moiety within the second-generation cam16-wt15-33 TriKE. An alternative possibility to decreased mispairing is that the smaller size of the sdAb may also reduce steric hindrance to present IL15 more efficiently. Regardless of the mechanism, the IL15 activity was altered by its flanking engaging sequences. This allowed the use of wtIL15 within this platform (23) and no longer required the use of enhanced IL15 mutants, used previously (20). In all aspects, including proliferation, degranulation, inflammatory cytokine secretion, tumor killing, and in vivo function, the second-generation cam16-wt15-33 was more potent than first-generation scFv16-m15-33. This differential in functionality was particularly evident in the HL-60luc xenogeneic mouse experiments. The cam16-wt15-33 TriKE induced increases in tumor control and NK-cell numbers compared with the scFv16-m15-33 TriKE.

Besides delivering a targeted ADCC signal, TriKEs specifically deliver IL15 to NK cells mediated by anti-CD16 engagement (20, 25). The second-generation cam16-wt15-33 TriKE bound to CD16 on NK cells 28-fold better than scFv16-m15-33. This differential in binding was also reflected in the delivery of IL15 to NK cells versus T cells. Despite delivering a stronger IL15 signal to NK cells, the cam16-wt15-33 TriKE, compared with rhIL15 and scFv16-m15-33, induced minimal stimulation of T cells, presumably by specific delivery to CD16-expressing cells. This targeted IL15 delivery may be particularly important for a better safety profile, which will need to be tested clinically. Cytokines have been reported to have a role in the development of leukemias and solid tumors. Targeted delivery would minimize this potential risk (35–38). IL15 may exacerbate autoimmune disorders such as rheumatoid arthritis, systemic lupus erythematous, inflammatory bowel disease, and multiple sclerosis (39). Thus, targeted delivery of IL15 via the cam16-wt15-33 TriKE may limit NK-independent autoimmune effects.

In the hematopoietic transplantation setting, specifically targeting IL15 to NK cells via the cam16-wt15-33 may enhance donor graft function to promote graft-versus-leukemia responses (1, 2, 40–43). Targeted delivery of IL15 could amplify NK-cell expansion and function without off-target stimulation of CD8⁺ T cells that are known to exacerbate potentially lethal graft-versus-host disease (44). The targeted delivery of IL15 to NK cells, coupled with the ability to potently drive ADCC in vitro and in vivo, makes cam16-wt15-33 TriKE an immunotherapeutic candidate for ongoing research. NK-cell engagers have the potential to work alone on endogenous NK cells, in settings were NK cells are diminished or dysfunctional, and in combination with NK-cell adoptive transfer.

M. Felices reports grants from the U.S. Department of Defense, NIH, NCI, and GT BioPharma during the conduct of the study; grants from GT BioPharma outside the submitted work; and a patent for US 2018/0282386 A1 pending, licensed, and with royalties paid from GT BioPharma. T.R. Lenvik reports grants from the U.S. Department of Defense, NIH, NCI, and GT BioPharma during the conduct of the study; grants from GT BioPharma outside the submitted work; and a patent for US 2018/0282386 A1 pending, licensed, and with royalties paid from GT BioPharma. A.J. Lenvik reports grants from the U.S. Department of Defense during the conduct of the study. D.K. Schirm reports grants from the the U.S. Department of Defense and NCI during the conduct of the study. M.F. Kaminski reports grants from the NCI and the U.S. Department of Defense during the conduct of the study. D.A. Vallera reports receiving royalties from GT BioPharma during the conduct of the study, as well as a patent for WO2017062604A1 pending and licensed to GT BioPharma. J.S. Miller reports grants and personal fees from GT BioPharma during the conduct of the study; grants and personal fees from Fate Therapeutics outside the submitted work; and a pending patent for TriKE, owned by the University of Minnesota and licensed to GT BioPharma. No potential conflicts of interest were disclosed by the other authors.

M. Felices: Conceptualization, supervision, funding acquisition, visualization, methodology, writing–original draft, project administration, writing–review and editing. T.R. Lenvik: Conceptualization, investigation, methodology. B. Kodal: Investigation, visualization, methodology. A.J. Lenvik: Investigation, visualization, methodology. P. Hinderlie: Investigation, visualization, methodology. L.E. Bendzick: Investigation, visualization, methodology. D.K. Schirm: Investigation, methodology. M.F. Kaminski: Investigation, visualization, methodology. R.T. McElmurry: Investigation. M.A. Geller: Conceptualization, supervision. C.E. Eckfeldt: Conceptualization, supervision. D.A. Vallera: Conceptualization, supervision, methodology, project administration. J.S. Miller: Conceptualization, supervision, writing–original draft, project administration, writing–review and editing.

The authors acknowledge the Translational Therapy, Flow Cytometry, and Imaging cores at the University of Minnesota for their excellent services. This work was supported, in part, by NCI P30 CA077598, Minnesota Masonic Charities, and the Killebrew-Thompson Memorial Fund. The authors also thank Xianzheng Zhou, at New York Medical College, for use of his HL-60luc cells. This work was supported, in part, by U.S. Department of Defense CA150085 (to M. Felices), NIH U01HL127479 (to D.A. Vallera), NCI P01 CA111412 (to M. Felices and J.S. Miller), P01 CA65493 (to M. Felices and J.S. Miller), R35 CA197292 (to M. Felices and J.S. Miller), and the Randy Shaver Cancer Research and Community Foundation (to D.A. Vallera). TriKE is a trademark of GT Biopharma, Inc., and this work was supported in part by their research funding (to M. Felices, D.A. Vallera, and J.S. Miller).

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.