A fusion protein complex that combines IL12, IL15, and IL18 signaling to induce memory-like NK cells for cancer immunotherapy

Abstract

NK cells are a promising cellular therapy for cancer, with challenges in the field including persistence, functional activity, and tumor recognition. Briefly, priming blood NK cells with recombinant human (rh)IL12, rhIL15, and rhIL18 (12/15/18) results in memory-like NK cell differentiation and enhanced responses against cancer. However, the lack of available, scalable good manufacturing process (GMP)-grade reagents required to advance this approach beyond early phase clinical trials is limiting. To address this challenge, we developed a novel platform centered upon an inert tissue factor scaffold for production of heteromeric fusion protein complexes (HFPCs). The first use of this platform combined IL12, IL15 and IL18 receptor engagement (HCW9201) and the second adds CD16 engagement (HCW9207). This unique HFPC expression platform was scalable with equivalent protein quality characteristics in small- and GMP-scale production. HCW9201 and HCW9207 stimulated activation and proliferation signals in NK cells, but HCW9207 had decreased IL18 receptor signaling. RNAseq and multidimensional mass cytometry revealed parallels between HCW9201 and 12/15/18. HCW9201 stimulation improved NK cell metabolic fitness and resulted in the DNA methylation remodeling characteristic of memory-like differentiation. HCW9201- and 12/15/18-primed similar increases in short-term and memory-like NK cell cytotoxicity and IFNγ production against leukemia targets, as well as equivalent control of leukemia in NSG mice. Thus, HFPCs represent a protein engineering approach that solves many problems associated with multi-signal receptor engagement on immune cells, and HCW9201-primed NK cells can be advanced as an ideal approach for clinical GMP-grade memory-like NK cell production for cancer therapy.

Introduction

Natural killer (NK) cells are innate lymphoid cells that are a promising cellular immunotherapy for the treatment of cancers; however, the optimal approach to enhancing NK cell effector function and recognition remain key areas of investigation (1–3). Brief activation of NK cells via combined IL12, IL15, and IL18 receptors induces memory-like (ML) NK cell differentiation, and results in subsequent enhanced NK cell persistence in vivo, target recognition, and effector functions (4–6). These findings led to clinical testing of ML NK cells in a first-in-human NK cell study at Washington University, revealing safety and the ability of these cells to induce remissions in patients with hematologic malignancies (7,8). Despite the fact that NK cell therapy has been translated to the clinic and mechanisms important for memory-like NK cell differentiation have been elucidated (8–12), several aspects of human ML NK cell biology still remain unclear, including the epigenetic mechanisms responsible for enhanced function and metabolic attributes (13,14). Currently, advancing this ML NK cell-based therapy is limited by Good Manufacturing Practice (GMP)-grade recombinant cytokines, in particular recombinant human (rh)IL18, to induce the ML NK cell differentiation program. Thus, a critical barrier in applying this approach in large-scale clinical trials, and combining with expansion approaches, is a lack of readily produced GMP-grade cytokine receptor-stimulating agents.

One approach to enhance cytokine signaling is to engineer fusion proteins providing new attributes, most commonly with the Fc portion of antibodies (Abs) (15). For example, N-803 (previously ALT-803), represents a complex between an IL15 variant and the IL15Rα sushi domain fused to IgG Fc, yielding a soluble form of the physiologic trans-presentation cytokine-receptor complex with prolonged in vivo persistence (16). N-803 acts as an IL15Rβγ super agonist in enhancing the activation, proliferation, cytokine secretion, and cytotoxicity of NK cells, resulting in potent killing of tumor and virus-infected cells (16–20). However, there remain challenges in large-scale GMP production of more complex immune-modulating agents that include proper protein folding and efficient purification for larger fusion proteins that include multiple protein domains.

Here, we addressed this challenge by developing a platform to express and purify soluble fusion proteins and protein complexes comprising cytokine, ligand, receptor, and single-chain antibody (scFv) domains. We found that the extracellular domain of human tissue factor (TF) could act as a fusion protein scaffold to allow high-level mammalian cell expression of difficult-to-produce proteins (i.e., IL15). The TF fusions could also be readily purified by anti-TF Ab affinity chromatography. This protein expression strategy was scalable and could be used to generate large amounts of heteromeric fusion protein complexes (HFPCs) under GMP conditions for cancer immunotherapy. The first HFPCs produced comprised IL12, IL15, and IL18 (referred to as HCW9201) and IL12, IL15, IL18, and a CD16 ligation domain (HCW9207). In this report, we evaluated HCW9201 and HCW9207 for their ability to activate NK cells through individual cytokine receptors, and in turn differentiate ML NK cells. We also provide new data on the epigenetic changes in human ML NK cells and metabolic features of cytokine- and HCW9201-activated NK cells.

Materials and Methods

Reagents, mice, and cell lines

Recombinant human (rh) cytokines were obtained from the following: IL12p70 (Biolegend), rhIL18 (InvivoGen or R&D Systems), rhIL15 (Miltenyi or NCI), and rhIL2 (Proleukin, Clinigen). CHO-K1 cells (ATCC, CCL-61) have been validated for GMP production of recombinant proteins as outlined in (21). Daudi cells (ATCC, CCL-213) (18), Raji cells (ATCC, CCL-86) (18), K562 cells (ATCC, CCL-243) (CBReGFP) (7), and 32Dβ cells (ATCC,CRL 11346) transfected with pREP9 (Invitrogen) encoding human IL15Rβ were cultured as described previously (22). All cell lines obtained from ATCC were cultured and expanded per ATCC recommendations, viably cryopreserved, and stored in liquid nitrogen. Once thawed, cultures were maintained for less than two months of continuous culture according to ATCC instructions. All cell lines were verified Mycoplasma-free by the MycoAlert Plus Mycoplasma Detection Kit (Lonza) (performed by the Washington University Tissue Culture Support Service or ATCC Universal Mycoplasma Detection Kit (ATCC, 30–1012K). HEK-Blue IL12 and HEK-Blue IL18 reporter cell lines were from InvivoGen and cultured as recommended and verified mycoplasma free using the ATCC Universal Mycoplasma Detection Kit. Antibodies for flow cytometry and CyTOF are described in Supplemental Tables S1 and S2.

NSG Mouse models

NOD-scid IL2Rgamma^null mice were obtained from Jackson Laborator and maintained under specific pathogen free conditions until 6–8 weeks of age. Trimethoprim and sulfamethoxazole (0.258 mg/mL, Hi Tech Pharmacal) was provided via drinking water on day of irradiation and maintained for 3 weeks of the study. For the tumor rejection model, nine normal human donors, (11–12 mice per group from three separate experiments) or NK cell persistence experiments (4 normal human donors performed in 3 separate experiments, 2 mice per group) were examined. All animal experiments were performed in accordance with our animal protocol approved by the Washington University Animal Studies Committee.

Production of HCW9201 and HCW9207

HCW9201 is a complex of two fusion proteins, one (IL18/TF/IL15) comprising IL18 and IL15 domains linked to the extracellular amino acid domain of human tissue factor (TF) and another (IL12/IL15RαSu) comprising a single-chain form of IL12 linked to the soluble domain of IL15Rα (IL15RαSu). HCW9207 is a similar two-protein IL18/IL15/IL12 complex with the addition of an anti-human CD16 scFv linked to the IL12/IL15RαSu fusion protein (IL12/IL15RαSu/anti-CD16 scFv) (23). The individual protein domains were fused without additional linker sequences (Figure 1A). The corresponding coding DNA sequences were synthesized (Genewiz) (24), cloned into pMSGV-1 modified expression vectors (25), and transfected into CHO.K1 cells. Co-expression of the two polypeptides in CHO cells allows for formation of the protein complex via high-affinity interactions between the IL15 and IL15RαSu domains and secretion of the complexes into the culture media. The HFPCs were then detected with product-specific ELISAs (24).

Figure 1. Biochemical characteristics of heteromeric fusion protein complexes (HFPCs).

A. Cartoon models of HFPC comprised IL18/TF/IL15 and IL12/IL-15RαSu complex (HCW9201, upper) or IL12/IL15RαSu/anti-CD16scFv complexed to IL18/TF/IL15 (HCW9207, lower). B-C. The recombinantly expressed HFPCs are highly glycosylated and therefore difficult to resolve by standard SDS-PAGE. To examine the protein characteristics, the proteins were deglycosylated and run on 4–12% SDS-PAGE Bis-Tis gels under reducing conditions and stained with InstantBlue. Panel (B). Lane 1 - deglycosylated HCW9201; Lane 2 - non-deglycosylated HCW9201. (C). Lane 1 - non-deglycosylated HCW9207; Lane 2 - deglycosylated HCW9207. Lane B in (C) contains a control of the deglycosylation mix in the absence of recombinant protein. D. HPLC-SEC analysis of purified HCW9201 and HCW9207 samples. E. Detailed upstream (blue) and downstream (green) GMP-scale processes are shown. Characteristics of engineering and GMP-scale processes are shown in Supplementary Table S3. A model of the potential structure of HCW9201 is shown in Supplementary Figure S1.

Production cell banks (HCW9201 and HCW9207) and a GMP-compliant master cell bank (HCW9201) were generated from stably transfected clonal cell lines following limited dilution cloning. Subsequent HFPC production was conducted using fed-batch methods with chemically defined media in scalable stir tank bioreactors. (Sartorius). Briefly, 0.4 × 10⁵ cells were seeded in a bioreactor at 80% of the final working volume of the bioreactor in EX-CELL® Advanced CHO Fed-batch Medium (Sigma-Aldrich) supplemented with 4 mM glutamine and 2 mM GlutaMAX (Life Technologies). Culture pH was controlled at pH 7.0 ± 0.2 with a CO₂/sodium carbonate cascade, dissolved oxygen was controlled at 40% using an O₂/CO₂ cascade, and temperature was maintained at 37°C. After 4 days in culture, the vessel temperature was dropped to 28°C, and 20% of the final culture volume of CHO CD EfficientFeed™ B AGT™ Nutrient Supplement (Life Technologies) was added to the bioreactor. The bioreactor was harvested on day 18 or when the culture viability dropped below 60%. The harvest was clarified using D0SP/X0SP depth filters (Millipore) at a 2:1 ratio (0.0054/0.0028 m²/L of harvest) followed by aseptic filtration using an Opticap XL 150 Capsule filter (Millipore) (0.0044 m²/L of harvest). HCW9201 and HCW9207 were purified from clarified bioreactor harvests using immunoaffinity chromatography with anti-TF Ab-conjugated Sepharose resin, resin. Briefly, monoclonal anti-TF (designated as HCW9101) was immobilized on CNBr-activated Sepharose 4 Fast Flow (Cytiva) resin per manufacturer’s instructions. The clarified cell culture harvest was loaded onto the HCW9101 immunoaffinity column equilibrated with 4 column volumes of PBS, followed by wash with 5 column volumes of PBS and 5 column volumes of 0.1 M sodium citrate (pH 5.0). HCW9201 or HCW9207 bound to the immunoaffinity column was eluted with 0.1 M acetic acid (pH 2.9–3.0). The protein peak was collected and then neutralized to pH 7.5 with 1 M Tris base. The pH adjusted sample was then buffer-exchanged into PBS using a 50 mL centrifugal device with molecular weight cutoff of 30 kDa (from Millipore). A GMP-compliant manufacturing process (scaled from 2L to 200L) was developed for HCW9201 consisting of immunoaffinity chromatography, low pH viral inactivation/depth filtration, anion exchange chromatography, nanofiltration, and ultrafiltration/diafiltration steps employing commercially scalable methods (Figure 1E). The purified product was characterized and released using the quality tests shown in Supplementary Table S3.

High-performance liquid chromatography – size exclusion chromatograph (HPLC-SEC)

HCW9201 or HCW9207 sample was diluted to 0.5 mg/mL in PBS mobile phase (0.2 M potassium phosphate, 0.25 M potassium chloride, pH 6.0) and applied onto TSKgel G3000 SWxl, 7*8x300mm analytical SEC column (Tosoh Bioscience) with a mobile phase flow rate of 0.8 mL/minute. Each Injection was run for 20 minutes and detected by ultraviolet light at 280 nm. Molecular weight was calculated using Gel Filtration Standard (BioRad) run under the same conditions.

Protein Deglycosylation

Deglycosylation of HCW9201 or HCW9207 was performed with Protein Deglycosylation Mix II (New England Biolabs) following Manufacturer’s Instructions. Briefly, 100 µg of HCW9201 or HCW9207 was heated at 75°C for 10 minutes in 1x Mix Buffer 2, then mixed with 5 µL of Protein Deglycosylation Mix II enzyme in a total volume of 50 µL. The deglycosylation reaction was first incubated at room temperature for 30 minutes, then transferred to 37°C for 16 hours.

Poly-acrylamide gel electrophoresis

HCW9201 or HCW9207 protein (deglycosylated or non-deglycosylated) was resolved by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) on 4–12% NuPage Bis-Tris gel (ThermoFisher Scientific,) under reducing conditions. After electrophoresis, the gel was stained with InstantBlue (Abcam) for 35 minutes, followed by destaining overnight in purified water.

N-terminal amino acid sequence analysis

HCW9201 protein was deglycosylated and resolved by reduced SDS-PAGE gel as described above. Protein bands on the SDS gel were then electroblotted onto a PVDF membrane (Immobilon PSQ, Millipore). The PVDF membrane was stained with InstantBlue (Abcam) for 35 minutes, followed by destaining overnight in purified water. Each protein band was cut from PVDF membrane and sent to Molecular Structure Facility, University of California, Davis (451 E. Health Sciences Dr. 1414 GBSF, Davis, CA, 95616) for N-terminal sequence analysis using Edman Sequencing method.

Biological activity on HEK cells (IL15, IL18, and IL12)

IL15 activity of the HFPCs was determined based on 32Dβ cell proliferation in a WST-1 assay per manufacturer’s instructions (Fisher Scientific). Briefly, 32Dβ cells (2 x 10⁴ cells/well) were incubated with increasing concentrations of HFPCs for 48 hours at 37°C. Cell proliferation reagent WST-1 (Roche, 11644807001) was added during the last 4 hours of cell growth according to the manufacture’s procedures. Conversion of WST-1 to the colored formazan dye by metabolically active cells was determined through absorbance measurements at 440 nm. IL12 and IL18 activities were determined using HEK-Blue IL12 and HEK-Blue IL18 reporter cell lines as recommended by manufacturer (InvivoGen). Briefly, HEK-Blue IL12 or HEK-Blue IL18 reporter cells (5 x 10⁴ cells/well) were incubated with increasing concentrations of HFPCs for 24 hours at 37°C. Next day, 20 µL of supernatant from the induced HEK-Blue IL12 or HEK Blue IL18 reporter cells was added to 180 µL of QUANTI-Blue (InvivoGen, rep-qb1), a subtrade for SEAP (secreted embryonic alkaline phosphatase), followed by incubation for 3 hours at 37°C. The absorbance was measured at 620 nm to determine IL12 or IL18 activity based on SEAP activity to change QUANTI-blue to a purple-blue color. The EC50 was determined with the dose-response curve generated from the experimental data by nonlinear regression variable slope curve-fitting with Prism8 software (GraphPad).

NK cell isolation and culture

Fifty-six ealthy donor PBMCs were collected from anonymous, healthy, platelet donors through the Mississippi Valley Regional Blood Center or One Blood, Orlando, FL. NK cells (≥ 95% CD56⁺CD3⁻) were isolated using RosetteSep (STEMCELL Technologies) as described (9). ML and control NK cells were generated as described previously (7) except the media used for activation, ML differentiation, phosphorylation, RNASeq, ,CyTOF and NSG experiments was NK media (Miltenyi) supplemented with 5% (v/v) heat-inactivated human AB pooled sera (Millipore Sigma) (CNKM). RPMI-1640 supplemented with 2 mM L-glutamine, antibiotics (penicillin, 100 U/mL; streptomycin, 100 µg/mL; ThermoFisher), and 10% (v/v) fetal bovine serum (Cytiva) was used for the methylation and metabolic studies. For activation, NK cells were incubated with individual cytokines: IL12p70 (10 ng/mL), IL15 (50 ng/mL), and IL18 (50 ng/mL) (12/15/18) (7), or HFPCs at 50 to 100 nM at two million per mL in a low adherence cell culture tray (Costar). After 12–16 hours of activation, NK cells were washed three times in 0.5% (w/v) human serum albumin in HBSS to remove residual cytokines or HFPCs. Cells were resuspended to two million per mL and plated in the presence of rhIL15 (1 ng/mL), with every other day media exchanges supplemented with fresh rhIL15 (1 ng/mL) for 6–7 days in assay-specific media.

Flow cytometry

For labeling of NK cells, cells were suspended in FACS buffer: 0.5%(w/v) BSA, 2 mM EDTA in PBS with no Ca²⁺ and Mg²⁺. All washes consisted of resuspension in the FACS buffer and pelleted at 660 x g for 4 minutes at ambient temperature. Cells were incubated with antibody cocktails to surface molecules as described in the specific assay for 30 minutes at ambient temperature in the dark and washed. For detection of intracellular molecules, cells were processed using Fix/Perm System (BD Biosciences) per manufacture recommendation, blocked with 7% (v/v) heat-inactivated goat serum in Perm/Wash buffer, and incubated with cytokine antibodies for 30 minutes at 4°C in the dark. Cells were washed twice and fixed until collection on a Navios, 3 laser 10 parameter flow cytometer (Beckman Coulter).

NK cell functional and cytotoxicity assays

For human NK cell activation status, a portion of activated or control NK cells were harvested and incubated with anti-CD107a and GolgiPlug/GolgiStop per manufacture recommendation (BD Biosciences) for 1 hour at 37°C. The cells were washed, placed in FACS buffer, and labeled with antibodies for CD45, CD3/CD56/CD16/CD25 for 30 minutes at ambient temperature. Cells were washed and then fixed/permeablized per manufacture recommendation (BD Biosciences) at 4°C. Permeablilzed cells were incubated in 7% (v/v) heat-inactivated goat serum (Sigma) to block non-specific sites and then incubated with antibodies to IFNγ and TNF for 30 minutes at 4°C in the dark. Finally, cells were washed and immediately collected on a Navios3 laser 10 parameter flow cytometer (Beckman Coulter). For detection of intracellular cytokines produced in response to tumor targets or IL12/IL15, cells were incubated at a 5:1 effector:target ratio, and functional assays were performed as described (5), except Zombie Green (Biolegend) was used to detect viable cells in this assay. Data were acquired on Navios Flow Cytometer and analyzed on FlowJo v.10.6.

Cytotoxicity assays (4-hour ⁵¹Cr-release) were performed as described (26) with K562, Daudi, and Raji targets, or using K562-Luc2 (ATCC), and the One-Glo Ex Kit per manufacturer recommendations (Promega) in 96-well white plates (Corning). Luminescence was detected on a Genios luminometer (Tecan). For antibody-dependent cellular cytotoxicity (ADCC), target cells were incubated with Rituxan (monoclonal anti-CD20 at 10 ug/mL; Roche) for 30 minutes at room temperature and washed to remove excess antibody prior to use. For ⁵¹Cr release assays, 4-hour lytic supernatants were collected per manufacturer recommendation and resultant counts per minute (cpm) was assessed on a Wallac 1450 MicroBeta TriLux Liquid Scintillation Counter. Percent specific lysis was calculated: [(cpm_test-cpm_spontaneous)/cpm_max-cpm_spontaneous)]*100.

Assessments of intracellular signaling

Phospho-flow cytometry methods were previously described (27). Briefly, viably cryopreserved NK cells were thawed in complete NK media and allowed to rest at 37°C for 30 minutes. HFPC at 50 and 100 nM or 12/15/18 (10/50/50 ng/mL, respectively) individual cytokines were used to activate NK cells at appropriate time intervals: 2 hours for STAT4, 1 hour for AKT and ERK, and 15 minutes for NFΚB-p65, STAT5, and p38-MAPK. Plate bound anti-FcγRIIIa experiments were performed as previously described (10).

Proliferation

Proliferation of flow sorted CD56^bright and CD56^dim NK cells was performed as described (5), with modifications. Freshly isolated human NK cells were purified by RosetteSep (StemCell Technologies), and labeled with CFSE according to manufacturer recommendations (Invitrogen). After washing, cells were stained with anti-CD45/CD3/CD56/CD16 for 30 minutes at ambient temperature in the dark, washed again, passed through a 70 u nylon mesh, and then immediately were flow-sorted (Sony SY3200 Synergy, Sony Biotechnology) using a 70 u tip into CD45⁺CD3⁻CD56⁺CD16-bright (br) or -dim populations to greater than 75% CD56brCD16dim or greater than 90% CD56dimCD16br purity. Five million sorted cells were seeded into one milliliter of NK media and activated with 12/15/18 (10/50/50 ng/mL) or 100nM HFPC or LD15 control, incubated for 16 hours, washed, and plated for memory-like differentiation in the presence of IL15 (1 ng/mL) with 1:1 media exchanges every other day. On day 5 post-activation, cells were harvested, washed, and labeled with anti-CD45/CD3/CD56/CD16 for 30 minutes, washed, and immediately acquired on Navios Flow Cytometer Data was analyzed using FlowJo v.10.3 software with Proliferation module to determine the percent divided cells in the CD56-bright and CD56-dim populations.

RNA Sequencing and analysis

NK cells from three normal human donors (5x10⁵) from each step of activation to memory-like differentiation were centrifuged, lysed in TRIzol (Invitrogen), and immediately flash frozen and stored at −80°C. After all samples were collected, TRNA was isolated (Zymo). Samples passing RIN analysis (Agilent) were moved to high-quality cDNA preparation with Clontech SMARTer ultra-low input kit according to manufacturer’s protocol. The samples were then indexed, pooled, and sequenced on an Illumina NovaSeq. Basecalls and demultiplexing were performed with Illumina’s RTA version 1.9 and bcl2fastq2 software, with a maximum of one mismatch in the indexing read. RNA-seq reads were then aligned to the Ensembl release 76 primary assembly with STAR version 2.5.1a (28). Gene counts were derived from the number of uniquely aligned unambiguous reads by Subread:featureCount version 1.4.6-p5 (29). Isoform expression of known Ensembl transcripts were estimated with Salmon version 0.8.2 (30). Sequencing performance was assessed for the total number of aligned reads, total number of uniquely aligned reads, and features detected. The ribosomal fraction, known junction saturation, and read distribution over known gene models were quantified with RSeQC version 2.6.2 (31).

All gene counts were then imported into the R/Bioconductor package EdgeR (32), and TMM normalization size factors were calculated to adjust samples for differences in library size. Ribosomal genes and genes not expressed in at least three samples greater than one count-per-million were excluded from further analysis. The TMM size factors and the matrix of counts were then imported into the R/Bioconductor package Limma (33). Weighted likelihoods based on the observed mean-variance relationship of every gene and sample were then calculated for all samples with the voomWithQualityWeights (34) function using a model matrix that included every condition as well as patient specific blocking factors. The performance of all genes was assessed with plots of the residual standard deviation of every gene to their average log-count with a robustly fitted trend line of the residuals. Differential expression analysis was then performed to analyze for differences between conditions, and the results were filtered for only those genes with Benjamini-Hochberg false-discovery rate adjusted p-values less than or equal to 0.05. For each contrast extracted with Limma, global perturbations in known Gene Ontology (GO) terms, MSigDb, and KEGG pathways were detected using the R/Bioconductor package GAGE (35) to test for changes in expression of the reported log 2 fold-changes reported by Limma in each term versus the background log 2 fold-changes of all genes found outside the respective term. The R/Bioconductor package heatmap3 (36) was used to display heatmaps across groups of samples for each GO or MSigDb term with a Benjamini-Hochberg false-discovery rate adjusted p-value less than or equal to 0.05. Perturbed KEGG pathways where the observed log 2 fold-changes of genes within the term were significantly perturbed in a single-direction versus background or in any direction compared to other genes within a given term with p-values less than or equal to 0.05 were rendered as annotated KEGG graphs with the R/Bioconductor package Pathview (37). Data is available on GEO (GEO GSE172100).

qRT-PCR

For qRT-PCR of CISH, total RNA from six normal human donor NK cells, LD 15 control, 12/15/18 and 100 nM HCW9201 activations was isolated using DirectZol Micro RNA Prep (Zymo). 100 ng cDNAwas prepared using Life Technologies High-Capacity cDNA Reverse Transcription Kit with random primers. CISH and ACTB control were amplified using the ABI primer sets Hs00367082_g1 and Hs01060665_g1, respectively in triplicate, using the TaqMan Gene expression system (Applied Biosystems) on an ABI7300 RT-PCR system (20 uL reaction). Fold-change was calculated 2^ΔΔCT.

Metabolic assessments

Human NK cells (2x10⁵) were seeded in Cell-Tak-coated Seahorse Bioanalyzer XFe96 (Agilent) culture plates in Seahorse XF RPMI medium, pH 7.4 supplemented with 2 mM L-glutamine for the glycolysis stress test. For the mitochondrial stress test, cells were seeded in Seahorse XF RPMI medium, pH 7.4 supplemented with 10 mM glucose and 2 mM L-glutamine. Assays were performed following the manufacturer’s instructions. The data was analyzed using Wave software (Agilent). Rapamycin (100 ng/mL) was added for inhibition studies, as indicated.

Pyrosequencing analysis of DNA methylation

The level of DNA methylation at the IFNG distal promoter conserved noncoding sequence 1 (CNS-1) enhancer region containing four characterized CpG sites (located at positions −4360, −4325, −4293, and −4278 relative to the transcription start site (TSS) (38) was determined using pyrosequencing reactions performed at Johns Hopkins University Genetic Resources Core Facility (JHU GRCF). This DNA sequencing method relies on light detection based on a chain reaction when pyrophosphate is released during sequential addition of nucleotides during the synthesis of a complementary strand of DNA as described (39). Genomic DNA extraction: Primary human NK cells were purified from two healthy donor PBMC as described (9). The purified NK cells were phenotyped, and treated with either individual cytokines (12/15/18) as described (7) or HFPCs (50 to 100 nM). After 12–16 hours of activation, NK cells were washed three times and plated in rhIL15 (1 ng/mL) with every other day media exchanges. Following treatment, NK cells (0.2 – 1.0 x 10⁶) were collected and washed with 0.5 mL DPBS (pH 7.4). Wet cell pellets were then either stored at −20°C until processed directly for genomic DNA extraction using the QIAamp DNA Micro Kit (QIAGEN). The DNA concentration was determined using the µDrop with the Multiskan Sky microplate reader (Thermo Scientific). Bisulfite conversion of genomic DNA were performed using the EZ DNA Methylation-Direct Kit (ZYMO). Five hundred nanograms of genomic DNA determined using the µDrop Microplate Reader (Thermo) was used per manufacture recommendations. For DNA methylation controls, the human non-methylated (UNMET) and methylated (MET) genomic DNA (ZYMO) were used. PCR amplification of dsPCR-CNS1 (247 bp) products from bisulfite-converted nDNA using the PyroMark PCR Kit (QIAGEN). For amplification of the IFNG CNS-1 region (4 kb upstream IFNG TSS) containing the four informative CpG sites, the PCR primers IFNG-CNS1F and IFNG-CNS1R-bio were used. Forty nanograms of bi-sulfite converted DNA was subjected to PCR amplification CFX96 thermal cycler (BioRad; 15 min at 95°C, followed by 50 cycles of 35 sec at 95°C, 35 sec at 55°C, and 40 sec at 72°C, with a final extension of 5 minutes at 72°C). PCR products dsPCR-CNS1 (247 bp) were analyzed in a 1.2% TAE agarose gel and stained with SYBR Safe DNA Gel Stain per manufacturer recommendation using a 1 kb Plus DNA Ladder (Thermo) as molecular weight marker. Pyrosequencing analysis of DNA methylation at the CpG sites within the IFNG CNS-1 region: dsPCR-CNS1 CpG PCR products were subjected to pyrosequencing analysis of DNA methylation using the DNA sequence primers C4399-CNS1F CNS1F (for analysis of CpG sites −4399, −4377, −4360, and −4325; DNA sequence: 5’-GGG GAT TTA GAA AAA T-3’) and C4293-CNS1F (for analysis of CpG sites −4293, −4278, and −4227; DNA sequence: 5’-TGT ATG ATG TTA GGA GTT T-3’). The pyrosequencing reactions were performed at JHU GRCF (https://grcf.jhmi.edu/dna-services/methlylation/analysis-via-pyrosequencing/). The pyrosequencing runs and profiles (determination of % DNA methylation) for each IFNG CNS-1 CpG sites (from −4399 to −4278) was provided by JHU GRCF. For analysis, the methylation percentages of the four IFNG CNS-1 informative CpG sites were averaged for each treatment. Unpaired Student t tests with two-tailed p value were used for statistical analysis of DNA methylation data using the GraphPad Software Prism 8 (Version 8.3.0).

Mass Cytometry

Three normal human donor NK cells isolated by RosetteSep and activated as described earlier were used for both RNASeq and CyTOF. Antibodies used for this study were either obtained by Fluidigm or conjugated in-house (Supplementary Table S2). Conjugations, titrations, surface and intracellular staining, and data acquisition and analysis were performed as described (8).

Ribbon Model

Ribbon model of HCW9201 heterodimeric cytokine fusion created in PyMOL2. The crystal structures of IL18 (PBD: 3WO2), tissue factor (PBD: 1BOY), IL15/IL15RA (PBD: 2Z3R), and IL12p70 (PBD: 1F45) were imported into the PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC.. Relevant amino acids were added or deleted to accurately reflect the fusion protein sequence. Molecular domains were then aligned to minimize steric hindrance.

BLI, survival, and persistence

Analysis was performed as described (7). Mice were irradiated at 125cGy one day prior to leukemia or NK cell injection. For leukemia elimination model, one million K562 CBReGFP cells were introduced via tail vein, and tumor burden monitored using on an amiHT optical imaging system (1–60sec exposure, bin8, FOV12cm, open filter) by intraperitoneal injection with D-luciferin (150mg/kg in PBS, Gold Biotechnology) after isoflourane anesthesia (2% vaporized in O₂). Total photon flux (photons/sec) was measured from fixed regions of interest over the entire mouse using Aura Imaging Software v3.2. For leukemia challenge experiments, 3–5 million NK cells were administered retro-orbitally on day 3 post leukemia injection and supported with IL2 (50,000 IU; Proleukin, Clingen) three times per week. Leukemia burden was assessed by weekly imaging and overall survival monitored until day 400.

For NK cell persistence in vivo, irradiated mice were injected with 2–5 million activated or control NK cells, through the tail vein and supported with IL2 (50,000 IU) 3 times per week for one week. Blood and spleen were harvested one week post NK cell injection. Blood was obtained by cardiac puncture, and spleen single-cell suspensions were prepared by mechanical disruption and passage through 70 um nylon mesh. RBCs were lysed, and splenocytes were stained with Zombie NIR (Biolegend) per manufacture recommendations, washed, and then stained with anti-murine CD45/anti-human CD45/CD3/CD56/hNKG2A staining for 30 minutes at 4°C. cell were then washed and immediately collected on a Beckman Coulter Gallios, 3-laser, 10 parameter Flow Cytometer. Data was analyzed using FlowJo v10.7 software.

Statistics

Statistics were performed using GraphPad Prism v.9. All ordinal data were tested for normal distribution (Shapiro-Wilk) and appropriate parametric/non-parametric tests used, as indicated in the figure legends.

Results

Heteromeric fusion protein complex (HFPC) production using a tissue factor (TF) platform

To overcome the challenge with large-scale production of complex multimeric fusion proteins, a novel protein expression platform was developed using a truncated human tissue factor (TF) domain as a fusion scaffold for expressing and purifying soluble fusion proteins and protein complexes comprising multiple functional domains. A pairing domain feature (e.g., IL15 and soluble IL15 receptor α) was further incorporated into the TF-based fusion concept to generate heteromeric, multi-functional protein complexes. To address the unmet need for GMP-grade reagents (7,8) that stimulate NK cells (5), we designed and produced HCW9201 (IL12p70, IL18, and IL15 trans-presented in the IL15Rα sushi domain (IL15RαSu) and HCW9207 (IL12p70, IL18, IL15/IL15RαSu and an anti-CD16 scFv)(Figure 1A, Supplemental Figure S1). After expression of the constructs in CHO-K1 cells and affinity purification of the soluble HFPCs, each polypeptide in the HFPC proteins was resolved by poly-acrylamide gel electrophoresis (Figure 1B–C). Bands for expected IL12/IL15RαSu or IL12/IL15RαSu/anti-CD16scFv, and IL18/TF/IL15 fusion proteins were clearly seen in HCW9201 and HCW9207 products, respectively. A truncated form of IL18/TF/IL15 was also detected in both HFPC preparations (Figure 1B/C). N-terminal amino acid sequence analysis of this protein band indicated that IL18 was cleaved between its C-terminal region and the N-terminal of TF. The partial degradation of the IL18/TF/IL15 fusion protein appeared to occur during the cell culture process. Other bands present in the purified protein lanes correspond to deglycosylation enzymes (Figure 1C, lane B). HPLC-SEC analysis was conducted to evaluate product purity (Figure 1D). The full-length and truncated forms of IL18/TF/IL15 copurified and were not resolvable by HPLC-SEC due to the small difference in their molecular masses (IL18 is approximately 18 kDa). Although the presence of the cleavage product was noted in all batches of the fusion protein (including the GMP large-scale batch), the biological activity of IL18 was ensured by quantitation on the HEK-IL18 dependent reporter cell line (Supplementary Table S3). HPLC-SEC results indicated that both HCW9201 and HCW9207 existed as dimers, with a small percent of high molecular protein aggregates observed in HCW9201 and relatively higher percent of protein aggregates in HCW9207. Based on our experience, and as demonstrated for HCW9201, protein aggregates could be significantly reduced by process development. The advantage of this protein expression platform is the scalability and consistency of producing biologically active, multi-gram quantities of HFPC protein (Figure 1E). Further characterization of identity, purity, and activity were completed for GMP-grade HCW9201 (detailed in Supplemental Table S3), and HCW9201 was qualified for ex vivo use in generating ML NK cells for human clinical trials. Thus, this unique protein manufacturing approach consistently produced the engineered HFPCs.

HCW9201 and HCW9207 induce specific signals downstream of individual cytokine receptors

We first performed experiments to confirm HCW9201 and HCW9207 induced signaling downstream of each individual cytokine receptor. Both HFPCs stimulated individual cytokine-dependent reporter cell lines with EC50 values defined for IL12, IL15, and IL18 (Figure 2A). There were differences in the activities using these cell-based assays; primarily with HCW9207 exhibiting less potent IL15 biological activity compared to HCW9201 and showing similar IL12 and IL18 activities (within assay variation).

Figure 2. HPFCs induce specific signals downstream of IL12, IL15, and IL18 receptors in reporters and primary human NK cells.

A. Dose-response curves of HCW9201 (red) and HCW9207 (purple) showing the concentration of HFPCs required to activate individual cytokine-dependent reporters in 32Dβ (IL15-dependent) and HEK blue (IL12p70- or IL18-dependent) cell lines. Data are from two independent experiments. B. Representative donor NK cell phospho-flow histograms examining the IL12, IL15, and IL18 signaling pathways. Purified NK cells were incubated with 12/15/18, HCW9201, HCW9207 at 50 or 100 nM for 0.25, 1, or 2 hours. depending upon pathway examined. C. Violin plots of the mean fluorescence intensity (MFI) fold-change from baseline. Data are from n=9 donors analyzed in 4 independent experiments. Significance was measured by 2-way ANOVA. *p≤0.05; **p≤0.01, ***p≤0.001. Additional violin plots for pAKT and pERK shown in Supplementary Figure S2.

HCW9201 and HCW9207 signaling in primary human NK cells

NK cell responses via cytokine receptors depend on the signaling cascade and phosphorylation of multiple molecules (40,41). To determine if signaling pathways normally activated in NK cells during cytokine stimulation were being triggered appropriately by HFPCs, we compared the MFI fold-change from baseline (resting) to activation with the combination of IL12p70, IL15, and IL18 (12/15/18) or HFPCs in CD56^bright and CD56^dim subsets (42)(representative flow histograms in Figure 2B, Supplementary Figure S2A); summary data in Figure 2C, Supplementary Figure S2B). IL12R signaling through phosphorylated (p)STAT4 was equivalent in CD56^bright NK for all stimulators, with minor differences in CD56^dim NK cells suggesting that the EC50 value differences observed in Figure 2A were not large enough to impact maximal signaling. For IL15R signaling, pSTAT5, pAKT, and pERK were examined (27). The results demonstrated that both 12/15/18 and HCW9201 induced robust and equivalent STAT5 phosphorylation, whereas HCW9207 had reduced pSTAT5 in CD56^dim NK cells. For IL18R signaling, 12/15/18 and HCW9201 induced similar NFkB-p65 and p38 MAPK phosphorylation over baseline, but HCW9207 had substantially reduced activity. Cross-linking of CD16 concurrent with HCW9201 stimulation did not alter these signals, suggesting that the addition of ITAM-based signals did not result in the signaling change (Supplementary Figure S2C), consistent with less effective IL18R signaling by HCW9207. Collectively, these data confirmed that HCW9201 induced cytokine receptor signaling via IL12R, IL15R, and IL18R, similar to 12/15/18, with reduced IL18R and IL15R signals observed with HCW9207 in primary NK cells.

We next compared the HFPCs to established ability of 12/15/18 to activate and induce proliferation of human NK cells (5). NK cells were labeled with CFSE and flow-sorted into CD56^bright and CD56^dim subsets (Figure 3A) and incubated for 12–16 hours with low dose (LD, 1 ng/mL) IL15 (control), 12/15/18, HCW9201, or HCW9207, washed, and replated in IL15 (1 ng/mL) to support survival for memory-like differentiation as modeled in Supplementary Figure 3A. After 5 days, proliferation was quantified using CFSE dilution (Figure 3B–C). CD56^bright and CD56^dim NK cells proliferated equivalently to 12/15/18 and HCW9201 or HCW9207 and significantly greater than IL15 controls. Activation of human NK cells for 12–16 hours with IL12, IL15, and IL18 synergistically increases expression of IFNγ and CD25, prototypic markers of combined cytokine activation (5,43). Dose-response curves of HFPCs identified maximal activation of CD25 and IFNγ at ≥100 nM (Supplementary Figure S3B), and 50 and 100 nM concentrations were then further tested (Figure 3D–E). The expression of CD25 and IFNγ was equivalently increased over controls following activation with 12/15/18 and HCW9201. However, HCW9207 exhibited reduced CD25 and IFNγ expression, consistent with reduced IL18 signaling (Figure 2). In contrast, surface CD107a was induced with HCW9207 over HCW9201 or 12/15/18, which is consistent with CD16 signaling via ITAMs triggering degranulation (Supplementary Figure S3C). Short-term cytokine activation also enhanced NK cell cytotoxicity, and HCW9201 and 12/15/18 had similar improvements in leukemia cell killing over IL15 only controls (Figure 3F).

Figure 3. HCW9201 and HCW9207 induced proliferation and short-term activation of NK cells.

A. Representative pre- and post-sorted CD56^bright and CD56^dim subset purity. B. Representative donor demonstrating the proliferation measured by CFSE dilution in sorted CD56^bright and CD56^dim subsets. LD, low dose IL15 (1 ng/mL). C. Summary data of proliferation measured by CFSE dilution. 2-way ANOVA. Data are from four independent experiments. D. Representative flow cytometry bivariate plots following 12–16-hour activation with 12/15/18; HCW9201 or HCW9207 effects on activation via CD69, IFNγ (intracellular), CD16, CD25, and TNF (intracellular). E. Summary data of (D). Statistical analysis of percent positive (top row) and mean fluorescence intensity (MFI, bottom row) of activation markers. CD107a, CD16^bright, and TNF are shown in Supplementary Figure S3. Mixed-effects analysis, Dunnet’s multiple comparisons test. Data are from n=6–7 donors analyzed in 4 independent experiments. F. NK cells were activated for 12–16 hours with HCW9201, 12/15/18, or LD IL15 (1 ng/mL). After activation, cells were washed to remove cytokine, and co-incubated with K562-luciferase targets at the indicated effector:target ratio for 24 hours. Killing was then read using Promega Bright-Glo EX. Results are representative of 2 independent experiments. Data were compared using REML Mixed-Effects model. Summarized data show mean and SE. *p≤0.05; **p≤0.01, ***p≤0.001, ****p≤0.0001.

Similar molecular programs are activated by 12/15/18 and HCW9201 and HCW9207

To define the molecular activation programs exhibited by 12/15/18 and HFPCs, bulk RNAseq was performed after 12–16 hours of priming and after subsequent in vitro memory-like NK cell differentiation (Figure 4A). Over three thousand genes were significantly differentially expressed (adjusted p value<0.05) when comparing baseline NK cells, LD IL15, and 12–16 hours of activation with 12/15/18, HCW9201, or HCW9207 (Supplementary Figure S4A, Supplementary Table S4). The top 50 differentially expressed genes (p adj<0.05) are shown in the heatmap (Figure 4B), with IFNG showing the highest induction over control. Linear regression analysis of the log-fold change over control revealed greater similarity between 12/15/18 and HCW9201, compared to 12/15/18 and HCW9207 (Figure 4C). There were some differentially expressed genes in 12/15/18 and HCW9201 that had reduced expression following HCW9207, including CSF2, EBI3, CD274, TNFSF4, CDKN1A, C15orf48, LGALS3, SOD2, and MFSD2A. When comparing the RNAseq profiles of 12/15/18, HCW9201, and HCW9207-primed cells that were differentiated in vitro in IL15, bulk RNAseq revealed few transcript differences, consistent with a return to a more resting state (Supplementary Table S5) and likely partially confounded by an IL15-induced molecular program (Figure 4D, Supplementary Figure 4B–C). Six genes (CCR5, CXCR6, IFNG, CCL5, FAM107B and TCEAL9) were shared by all three activation methods (Supplementary Figure S4C). Significant reduction in CISH transcript was evident in D6 ML NK cells induced by both HFPCs and validated by qRT-PCR (Supplementary Figure S4D). Based on the activity assay comparisons (Figures 2–4), HCW9201 was determined to be the lead HFPC to replace 12/15/18 and was the focus of the remainder of the study.

Figure 4. Transcriptome and IFNγ CNS-1 CpG analysis demonstrates similar molecular program and DNA methylation patterns in NK cells following 12/15/18- or HCW9201-activation and ML NK cell differentiation.

A. RNA-Seq experimental workflow showing approach to activation, ML differentiation, and RNA sequencing. B. Heatmap of the top fifty differentially expressed genes compared to IL15 control NK cells from three normal donors (adjusted p<0.05). C. Linear regression analysis of the log fold-change of all transcripts comparing 12–16-hour activated HCW9201 or HCW9207 to 12/15/18. Annotated is the top overexpressed gene shared between all of the activations (IFNγ). D. ML NK cells differentiated for 6 days in IL15 compared to control IL15 only NK cells. Selected functionally significant genes are annotated in volcano plots of log fold-change of control (IL15 only) to 12/15/18, HCW9201, and HCW9207. A horizontal line demarcates genes with a −logFDR ≥ 1.3. See also Supplementary Figure S4. E. Schematic of CpG sites evaluated for DNA methylation changes within the IFNγ CNS-1 region. F. DNA methylation (%) mean+/-SD of key IFNγ CNS-1 CpG sites following the indicated treatment. Data are from 2 independent experiments. Unpaired Student t tests with two-tailed p value were used for statistical analysis of DNA methylation data (ns, not significant; *p<0.05; **p<0.002). See Supplementary Figure S6 for DNA methylation of individual CpG sites in NK cells for each donor.

CNS-1 is equivalently demethylated in 12/15/18- and HCW9201- activated NK cells

The CNS-1 region of the IFNγ locus has been shown to be important in NK cell differentiation (44) and is regulated epigenetically by DNA methylation after activation with 12/15/18 in mice (11), leading to memory-like NK cell differentiation. To evaluate this mechanism in human NK cells and to compare both 12/15/18 and HCW9201 treatments, the methylation status of CNS-1 CpG sites (−4360, −4325, −4293 and −4278) was interrogated after overnight activation and 14 days of differentiation (Figure 4E). These four CpG sites showed a significant reduction in DNA methylation in NK cells activated with 12/15/18 and HCW9201 (Figure 4F), with the greatest reduction in DNA methylation observed at the CpG sites located at −4360 and −4325 (Supplementary Figure S5). These data demonstrate that the IFNγ CNS-1 locus is demethylated in human ML NK cells after induction with 12/15/18 or HCW9201 agents.

12/15/18 and HCW9201 enhance cellular metabolism

Metabolism is an important factor in NK cell functional responses (45), and IL2, IL15, and IL12 have been shown to impact glycolysis and oxidative phosphorylation in human NK cells (46), yet it is unknown how combined 12/15/18 activation affects measures of metabolism. To define the impact of 12/15/18 and HCW9201 on the metabolism of human NK cells, extracellular flux assays were performed. Both 12/15/18 and HCW9201 increased the rate of glycolysis in human NK cells to a significantly greater magnitude compared to resting or IL15 controls (Figure 5A–B). This included significantly greater glycolytic capacity. 12/15/18 and HCW9201 activation also increased the mitochondrial oxygen consumption rate (OCR) to similar extents (Figure 5C–D) and showed significantly increased OCRs over IL15 controls. The enhanced glycolysis, glycolytic reserve, glycolytic capacity, and basal respiration were inhibited by rapamycin (Figure 5B and 5D, Supplementary Figure S6), suggesting a role for mTOR signaling in these metabolic changes. Collectively, these data showed that signals via the IL12, IL15, and IL18 receptors similarly enhanced key metabolic pathways of glycolysis and oxidative phosphorylation for energy production over resting and IL15 control-stimulated NK cells. The metabolic fitness may be linked to enhanced proliferation and cytotoxic function of NK cells.

Figure 5. NK cell metabolism is similarly enhanced by 12/15/18 and HCW9201 over IL15 controls.

Metabolic parameters were examined for resting, IL15 (control), 12/15/18, and HCW9201-activated NK cells using XFe96 Extracellular Flux Analyzer measurements of extracellular acidification rate (ECAR). Complete ECAR analysis consisted of four stages: basal (without drugs), glycolysis induction (10 mM glucose), maximal glycolysis induction (2 μM oligomycin), and glycolysis inhibition (100 mM 2-deoxy-glucose). All graph error bars show SEM. A. Representative donor glycolysis stress test tracing with measurement of ECAR with the indicated stimulation. B. Summary data from (A) showing glycolysis, glycolytic reserve, and glycolytic capacity, in the presence or absence of rapamycin (right), N=6; two independent experiments performed in triplicate. Ordinary one-way ANOVA, *p<0.05, **p≤0.01, ***p≤0.0001, ****p≤0.0001. C. Representative donor mitochondrial respiration tracing of oxygen consumption rate (OCR). D. Summary data from (C) showing basal respiration, maximal respiration, and spare respiratory capacity in the presence or absence of rapamycin when activated by 12/15/18 and HCW9201. N=6; ordinary one-way ANOVA, *p=0.03, **p≤0.005, ****p≤0.0001. Data shown is summarized from n=6 donors analyzed in 3 independent experiments.

Mass cytometry reveals 12/15/18 or HCW9201-primed and ML NK cells are identical

Multi-dimensional analysis of 12/15/18-primed and resulting memory-like NK show phenotypic differences from conventional NK cells, both in vitro and in vivo (7,8). HCW9201- and 12/15/18-primed and memory-like NK cells were compared using a 37 parameter CyTOF panel (Figure 6). Representative viSNE maps of 12/15/18- and HCW9201-activated human NK cells (Figure 6A–B) revealed similar mapping patterns after the short-term activation (12–16 hours), as well as after ML NK cell differentiation (day 6). Unsupervised FlowSOM analysis defined six metaclusters (Figure 6C), which did not differ in abundance between 12/15/18 and HCW9201 conditions (Figure 6D). The median expression of all 37 markers was compared between baseline (resting) NK cells and HCW9201 and showed a lack of correlation after linear regression (Figure 6E). In contrast, comparisons of HCW9201 and 12/15/18 after 12–16 hours or after subsequent ML NK cell differentiation (day 6) demonstrated a significant concordance. Collectively, these data demonstrate that incubation with 12/15/18 or HCW9201 results in a similar activation and ML NK cell multidimensional phenotype.

Figure 6. HCW9201 and 12/15/18 induce the same activation phenotype and ML NK cell phenotype using multidimensional mass cytometry.

Purified NK cells were incubated overnight with 12/15/18 or HCW9201 (12–16 hours), washed, replated in IL15 (1 ng/mL for survival), and harvested at day 6 to assess ML phenotype. A. Representative viSNE maps of 12/15/18- and HCW9201-activated NK cells. B. Representative viSNE maps of 12/15/18- and HCW9201-induced ML NK cells (day 6). C. FlowSOM was used to identify NK cell metaclusters. Each metacluster was identified on composite viSNE and indicated by color. D. Summary data (mean+/-SEM) of three donors. E. Regression assessing median expression (solid line) and error (dotted lines) of all markers used to define viSNE maps at the indicated timepoint. Data are from 3 independent experiments with one donor each (N=3 donors).

HCW9201- and 12/15/18-induced ML NK cells have potent function on restimulation in vitro and in vivo

Increased IFNγ expression and cytotoxic responses after challenge with the K562 leukemia are hallmarks of 12/15/18-induced ML NK cells (5). 12/15/18- and HCW9201-primed ML NK cells differentiated for 6 days were challenged with K562 cells, and IFNγ was measured after 5 hours. Both 12/15/18- and HCW9201-induced ML NK cells produced significantly more IFNγ than LD IL15 controls, with HCW9201-induced ML NK cells producing the greatest levels (Figure 7A). ML NK cells differentiated after priming by 12/15/18 or HCW9201 had equivalently improved cytotoxicity against K562 targets compared to IL15 controls (Figure 7B). 12/15/18- and HCW9201-primed MLNK cells demonstrated similar improvements in killing of the resting NK-resistant Daudi and Raji tumor targets with and without anti-CD20 to simulate ADCC (Supplementary Figure S7). To compare ML NK cells induced by HCW9201 and 12/15/18 in vivo, cells were injected into NSG mice engrafted with K562 leukemia and compared to IL15 NK cell and no NK cell controls (Figure 7C). Significantly enhanced survival was seen for mice treated with 12/15/18- (P=0.012) and HCW9201- (P=0.037) induced ML NK cells compared to the no NK cell control group (Figure 7D). 12/15/18-induced ML NK cells resulted in improved survival compared to the IL15 only control NK cell group (P=0.048), with HCW9201-induced ML NK cell treatment trending toward survival improvement (P=0.057). Using bioluminescent imaging (BLI), we found that leukemia was significantly and similarly reduced by both 12/15/18- (P=0.02) and HCW9201- (P=0.0086) primed NK cells compared to control NK cells (Figure 7E–F, Supplementary Figure S8A). To address in vivo persistence, we compared the in vivo NSG persistence of control NK cells to 12/15/18- and HCW9201-induced ML NK cells 7 days after transfer. There was a significant increase in human NK cells in the 12/15/18 and HCW9201 groups compared to control group (Figure 7G–H), and increased persistence of NKG2A⁺ NK cells in the blood and spleen at D7 post-injection (Supplementary Figure S8B). These data overall indicate that HCW9201 recapitulates the ML NK cell induction with similar function to 12/15/18-induced ML NK cells.

Figure 7. Functional comparisons of 12/15/18- and HCW9201-primed ML NK cells.

A. ML NK cells were differentiated over 6 days following 12–16-hour activation with indicated priming stimulus, and IFNγ expression from ML-NK cells was assessed after challenge with K562 tumor targets. Data are from 6 donors (paired) analyzed in 3 independent experiments. Statistical differences were assessed by 2-way ANOVA, multiple comparisons, *p<0.01, ****p<0.001. B. ML NK cells primed with the indicated stimuli were used as effectors against K562 leukemia in a 4-hour ⁵¹Cr cytotoxicity assay. Mean specific killing+/-SEM. 2-way ANOVA, 3 normal donors, 2 independent determinations, ****p<0.0001. C. Schema of in vivo model with NSG mice and K562-luciferase leukemia cells. BLI, bioluminescence imaging. D. Survival curves of NSG mice engrafted with K562-luciferase leukemia and treated with NK cells primed by low-dose (LD) IL15 (control), 12/15/18, or HCW9201 (each group, N=11–12 mice/group). Log-rank comparisons: No NK control vs. 12/15/18: p=0.012; No NK control vs. HCW9201: p=0.037; LD IL15 control vs. 12/15/18: p=0.048; LD vs. HCW9201: p=0.057; 12/15/18 vs. HCW9201: p=0.89 (not significant, NS). E-F. BLI assessment. (E) Example BLI from a single mouse for each NK cell priming condition. (F) Summary time course of BLI measuring K562 burden. Statistical differences were assessed by 2-way ANOVA with multiple comparisons. Data are combined from 3 independent experiments (N=11–12 mice/group). G-H. Purified human NK cells from four normal donors, 2 mice per group, three independent experiments were incubated overnight with low-dose IL15, IL12/15/18, or HCW9201, washed, and injected into irradiated (125 cGy) recipient NSG mice i.v. IL2 (50,000 IU) was injected i.p. every 2–3 days to support human NK cell survival, and spleen and blood assessed 7 days later. G. Representative flow plots from the peripheral blood of recipient mice assessing human NK cells (hCD45⁺). Numbers represent the frequency of cells within the indicated gate. H. Summary data from (G) from the indicated tissue. Data are represented as mean and SEM. Data were tested for normal distribution then analyzed using the appropriate comparison (ordinary ANOVA/ Kruskal Walis). *p<0.0159, **p<0.0092. See also Supplementary Figure S8.

Discussion

In this study, we describe a new approach to construct multi-functional heteromeric fusion protein complexes using a TF scaffold platform. The extracellular domain of human TF was selected because it has a rigid elongated structure comprised mainly β-sheets with its N- and C-termini located at distal ends (>70 Å apart) of the polypeptide (47), permitting genetic fusions of other protein domains without anticipated steric interference. The extracellular domain of human TF does not interact with the cell membrane phospholipid bilayer and, as a result, does not exhibit procoagulant activity (48). Human TF is expressed at high levels by most cell types and is not expected to be immunogenic in humans. Consistent with these properties, we found that genetic fusion to the TF domain promoted increased production of difficult-to-express proteins, such as IL15. The TF fusion proteins could be readily purified by affinity chromatography using an anti-TF and low pH elution conditions, similar to those used in Protein A-based affinity purification of Abs. To generate multichain protein complexes, we also incorporated genetic fusions to the human IL15 and IL-15RαSu domains. When co-expressed in CHO cells, the fusion proteins formed a soluble stable heterodimeric complex through high-affinity interactions between IL15 and IL15RαSu domains. This approach offers an alternative to immunoglobulin (Fc) and other engineered protein scaffolds, which typically require introduction of multiple mutations or other non-human sequences or complicated in vitro assemble/purification methods to generate bi- or multi-specific complexes (49,50). Using the TF scaffold platform, we constructed more than 30 fusion complexes comprising various cytokines, ligands, receptors, and scFvs. Our characterization of HCW9201 and HCW9207 described in this study verify high-level production and purification of HFPCs that retain the expected biological activities of their multiple cytokine and scFv domains. We also provide a scalable approach for generating large-scale GMP-grade HFPCs to support clinical applications.

HCW9201 and HCW9207 were designed to coordinately activate the IL12, IL15, and IL18 receptors, with (HCW9207) or without (HCW9201) engagement of CD16. We hypothesized that HCW9201 and HCW9207 would result in changes to NK cell activation and differentiation into ML NK cells, similar to the 12/15/18 cytokine combination (5,7). A comprehensive comparison of these HFPCs to recombinant human IL12, IL15, and IL18 revealed similar cytokine receptor signaling, activation, molecular programs, and functional ML NK cell differentiation with HCW9201 and 12/15/18. However, HCW9207 exhibited reduced IL18 and IL15 signaling and had several differences in induced gene expression in primary NK cells compared to 12/15/18 and HCW9201. We also demonstrated that 12/15/18 and HCW9201 enhanced the metabolism of NK cells and resulted in reduced methylation of key CpG sites within the IFNγ CNS-1 regulatory region. Finally, HCW9201- and 12/15/18-primed NK cells exhibited similar enhanced cytotoxicity and ADCC against leukemia targets in vitro and equivalently controlled leukemia in NSG mice. Based on these results, HCW9201 was found to have essentially identical activity as the combination of individual IL12, IL15, and IL18 cytokines in effectively generating ML NK cells, solving the barrier to GMP-grade reagents to produce ML NK cells for large/late-stage clinical trials.

Memory-like NK cells were first reported following combined IL12, IL15, and IL18 stimulation in mice (4), which was confirmed in human NK cells (5). Additional signals have been reported to contribute to ML NK cell induction, including ligation of the activating receptor CD16, which enhances ML NK cell functionality when combined with cytokines (5). The activity of CD16 and consequent ITAM signaling to promote memory-like NK cell responses was confirmed following activation with the high-affinity anti-CD16a/CD30 bi-specific protein AFM13 (51). HCW9207 was designed to improve on individual 12/15/18 targeted with HCW9201. However, due to reduced cytokine signaling of HCW9207 in NK cells, HCW9201 was nominated as the lead HFPC with subsequent production of a large-scale GMP lot that is sufficient to support future clinical evaluation. Fusion of the anti-CD16 scFv domain did not provide additional activity to the HCW9201 complex; thus, further protein engineering is underway to test new approaches to coordinately engage the IL12, IL15, and IL18 receptors and other activating and co-activating receptors to further improve on ML NK cell generation. We also explored new aspects of human ML NK cell biology, identifying that HCW9201 and 12/15/18 activation resulted in enhanced mTOR-dependent metabolism, which likely contributed to subsequent ML NK cell enhanced proliferation and effector functions. As previously reported following 5 days of IL12/15/18 stimulation (52), we provide data demonstrating that similar to murine ML NK cells (11), HCW9201 and 12/15/18-induced human ML NK cells also had selective demethylation of key CpG regulatory sites within the IFNγ CNS-1 region, providing further evidence of epigenetic reprogramming of DNA methylation as one mechanism responsible for enhanced effector function. The mechanism of epigenetic regulation of ML NK cells is a nascent area in the field, and key questions include the relative contributions of DNA methylation and histone modifications. Finally, new HFPCs are being developed and tested via the TF-based platform to facilitate the controlled expansion and survival of ML NK cells.

We also examined the molecular program of NK cells after HCW9201, HCW9207, and 12/15/18 activation, and the subsequent ML NK cell differentiation. Coordinated ligation of the IL12, IL15, and IL18 receptors resulted in changes in more than 3000 mRNA transcripts, with a high concordance between HCW9201 and 12/15/18. Key hallmarks of NK cell activation were equivalent, including transcription of the IFNG and IL2RA (encoding CD25) genes. Although HCW9207 was also concordant with 12/15/18, the correlation was less prominent and differences were observed. The immune inhibitory checkpoint LAG3 was noted to be induced predominantly after 12/15/18, HCW9201, and HCW9207 activation, identifying a new potential checkpoint for NK cells following activation. The chemokine receptors CCR1 and CCR5 were also upregulated, indicating that ML NK cells may have an enhanced potential to traffic to sites of inflammation. In contrast to acute activation, following ML NK cell differentiation, few transcripts are differentially regulated between IL15 only-treated control NK cells, and those primed with 12/15/18, HCW9201, or HCW9207. This may reflect a high ‘background’ of IL15-stimulated mRNA changes, which was evident when comparing to naïve NK cells. This also likely reflects molecular program heterogeneity because not all NK cells resulting from 12/15/18 receptor priming and differentiation are equal, and a subset of resulting cells may be memory-like with particularly enhanced anti-tumor function. Future studies will address this using single-cell RNAseq. Despite this background, some genes were found to be altered in ML NK cells. One finding was that the suppressor of cytokine signaling CISH was downregulated at the transcript level in HCW9201- and HCW9207-induced NK cells. Because this negative regulator of cytokine signaling was previously implicated in limiting NK cell function (53), this may indicate a mechanism of enhanced ML NK cell functionality and persistence. Indeed, the enhanced metabolism and CISH could be linked, as CISH deletion in iPSC-differentiated NK cells demonstrate improved metabolic activity dependent on mTOR signaling (54). Chemokine receptor and trafficking molecules were also identified, including SELL (L-selectin), CCR5, and CXCR6. Previous work has identified that CD62L protein is increased on ML NK cells in vitro and in vivo (7,8), suggesting that SELL transcription is one mechanism contributing to this phenotype. The gene encoding the serine protease granzyme K (GZMK) was also increased, suggesting an enhanced capacity for cell death and potentially triggering inflammation. CXCR6 was also more abundant in ML NK cells compared to controls, which has been implicated in other forms of NK cell memory (55). These hypothesis-generating findings will require further investigation to clearly define their role and importance for ML NK cell biology.

ML NK cells generated with signals via the IL12, IL15, and IL18 receptors have been translated into the clinic as a cellular therapy for leukemia. These trials were preceded by a comprehensive pre-clinical evaluation of ML NK cells that established their cytokine responsiveness, unique aspects of their biology, and enhanced anti-leukemia activity in vitro and in vivo (5,7,9,10). Here, we compared HCW9201 to 12/15/18 and found overlap with these pre-clinical attributes, supporting the use of HCW9201 as an approach to ligate cytokine receptors to generate ML NK cells for clinical use. In the first-in-human phase 1 study, ML NK cells were safe and did not cause inflammatory adverse events characteristic of CAR T cells, such as cytokine release syndrome (CRS) and immune cell-associated neurotoxicity syndrome (ICANS). Complete remission was induced in nearly half of relapsed/refractory AML patients, identifying promising avenues of further clinical investigation in older, unfit patients and those who require additional therapy as a bridge to potentially curative hematopoietic cell transplantation (7,8). Expanding on this, ML NK cells have also demonstrated prolonged persistence in vivo when used to augment allogeneic HCT (56,57), and induce complete remission in the setting of relapsed AML after and allogeneic HCT (58). ML NK cells have robust CD16 expression and exhibit improved ADCC (10), indicating that they can be directed to targets using therapeutic mAbs or bi/tri-specific NK cell engagers. ML NK cells can also be engineered with chimeric antigen receptors (CAR) and exhibit improved functional activity compared to conventional CAR NK cells (59). To explore these clinical opportunities in the clinic, HCW9201 now provides an ideal GMP agent that can be advanced from early phase to late phase clinical trials. Toward this end, GMP-grade HCW9201-primed NK cells have now been used as an NK cellular therapy, as part of a clinical trial (NCT01898793).

In summary, we report that HCW9201 is a HFPC that recapitulates the IL12, IL15, and IL18 receptor signaling required to enhance NK cell function and results in ML NK cell differentiation equivalent to a combination of individual IL12, IL15, and IL18 cytokines. New aspects of ML NK cells were revealed, including enhanced metabolism, and an improved molecular understanding of the human 12/15/18-induced ML NK cell. Using this novel TF-based platform, GMP-grade HCW9201 and other HFPCs can be generated as new tools for cancer immunotherapy, addressing a major barrier for advancing ML NK cell therapy to larger clinical trials.

Synopsis:

A scalable platform, with GMP application, that can be used to generate a multitude of different heteromeric proteins is presented. The platform’s use to generate memory-like NK cells for immunotherapy is demonstrated and provides insights into their biology.