Novel trispecific killer engager targeting B7-H3 enhances natural killer cell antitumor activity against head and neck cancer

Abstract

Background

Patients with head and neck squamous cell carcinoma (HNSCC), particularly the human papillomavirus negative (HPV⁻) subset, have a dismal prognosis. Furthermore, patients with Fanconi anemia (FA) have a genetic predisposition with a 500-fold to 700-fold higher incidence of HNSCC. Thus, novel and more efficacious therapies are needed. As current immunotherapies often fail due to suppressive elements in the tumor microenvironment (TME), we developed a trispecific killer engager (TriKE) to direct multiple signals to natural killer (NK) cells to overcome the hypoxic TME. This TriKE is comprised of a camelid nanobody that binds to CD16 on NK cells, an interleukin (IL)-15 moiety, and another novel camelid nanobody that binds to the B7-H3 antigen, which is highly and specifically expressed on the tumor cell surface.

Methods

The B7H3 TriKE was generated using a mammalian expression system. Its functionality was evaluated using flow cytometry-based NK cell degranulation, cytokine production, proliferation and live cell imaging cytotoxicity assays. Models of acute and prolonged hypoxia (1% oxygen) were carried out to assess tumor killing. Tumor progression, NK cell persistence, and survival differences between IL-15-treated and TriKE-treated mice were studied using NOD-scidIL2Rg^null (NSG) mice engrafted with human HNSCC.

Results

High B7-H3 expression was found in HPV⁻ HNSCC cell lines, even when the FA gene was knocked out, and The Cancer Genome Atlas patient data showed that high B7-H3 expression predicted poor survival in patients with HPV⁻ HNSCC. Similar to the NK cell activity seen with healthy donors, the B7H3 TriKE enhanced activation, expansion and cytotoxicity of NK cells from patients with HPV⁻ HNSCC, a target population for this therapeutic. Additionally, the B7H3 TriKE improved NK cell cytotoxicity in a three-dimensional spheroid model of HNSCC. In both acute and prolonged hypoxia (1% oxygen), the B7H3 TriKE mediated enhanced tumor killing, mitigating impairment of NK cell cytotoxicity in hypoxia. In vivo, the B7H3 TriKE-treated mice demonstrated substantial antitumor activity and prolonged survival.

Conclusions

The B7H3 TriKE is a novel immunotherapeutic approach that can overcome hypoxic suppression of NK cells in the HNSCC TME. These highly translational studies present an innovative therapy for patients with HNSCC and will be developed further for clinical application.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Natural killer (NK) cells can mediate antibody-dependent cellular cytotoxicity through their Fc receptor (CD16) but encounter challenges in the solid tumor microenvironment where NK cell function is impaired.

WHAT THIS STUDY ADDS

• This study demonstrates that a novel dual camelid B7H3 trispecific killer engager (TriKE) delivers a potent CD16 agonistic signal as well as interleukin-15 stimulation to effectively direct NK cell-mediated killing of head and neck cancer in vitro and in vivo.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The promising preclinical efficacy of the B7H3 TriKE suggests its potential to advance NK cell immunotherapy for solid tumors, and the B7H3 TriKE is planned for further development to treat patients with head and neck cancer.

Background

Head and neck squamous cell carcinoma (HNSCC) accounts for approximately 90% of all head and neck cancers, arising from the mucosal epithelium of the oral or nasal cavity, salivary glands, pharynx and larynx.^{1 2} HNSCC is the seventh most prevalent cancer in the world with 660,000 new cases annually.^{1 3 4} The mortality rate remains high, especially in those with recurrent/metastatic HNSCC or a genetic predisposition, such as Fanconi anemia (FA).4,7 The survival of patients with HNSCC has shown modest improvement, from 5-year survival of 55% in the 1990s to 66% in the 2000s,⁸ but this improvement could be attributed to an increase in the proportion of patients with human papillomavirus (HPV)-associated HNSCC.¹ Analysis from the Surveillance, Epidemiology and End Results registry found worse survival outcomes in patients with HPV-negative (HPV⁻) HNSCC.⁹ Therefore, it is important to make the distinction between HPV-associated HNSCC and non-HPV-associated HNSCC, because they exhibit different clinical behaviors,¹⁰ prognoses,¹¹ responses to treatment¹² and survival.¹³

HPV⁺ HNSCC arises more commonly in the oropharyngeal region, due to oncogenic transformations of the tonsillar crypts by HPV-16, HPV-18, HPV-31 to HPV-33 or HPV-52^{1 14} whereas HPV⁻ HNSCC arises primarily due to tobacco and alcohol consumption, or genetic disorders such as FA.^{1 15} FA is an inherited disease caused by mutations in any of the 23 known genes that have a role in the DNA repair pathway.¹⁶ FA is manifested clinically by physical abnormalities, bone marrow failure, a predilection to leukemia and a roughly 700-fold increased risk of developing HNSCC,⁷ as well as other cancers.¹⁷ While the prognosis for the HPV⁺ subset of patients is relatively favorable with 5-year survival rates of 75–80%, similarly staged HPV⁻ patients have a significantly worse prognosis, with 5-year survival rates of less than 50%.^{18 19} Furthermore, HPV⁺ patients respond better to conventional treatments of HNSCC such as surgery, chemotherapy and radiation therapy in comparison to HPV⁻ patients.^{11 13 20} Treatment options for patients with FA-derived HNSCC are even more limited due to their sensitivity to DNA damaging agents, resulting in low tolerance to highly toxic chemotherapy and radiation.⁷ Surgery is also a less viable option because of the aggressiveness of the solid tumor, as HNSCC is usually diagnosed at an advanced stage in these patients.⁷

In recent years, cancer immunotherapy has led to important breakthroughs in the treatment of hematological cancers²¹ and some solid tumors.²² Checkpoint blockade approaches²³ and chimeric antigen receptor T cell therapies have garnered a lot of interest,²⁴ and other antibody-based therapies have been in the clinic for decades. However, antibody-based therapies for HNSCC have shown limited effectiveness. For example, cetuximab, a monoclonal antibody (mAb) targeting epidermal growth factor receptors, and pembrolizumab, a mAb blocking programmed cell death protein-1 (PD-1), both showed response rates below 20% in metastatic and advanced cases.^{25 26}

Favorable prognosis in HNSCC has been associated with natural killer (NK) cell tumor infiltration,^{27 28} indicating the potential for NK cell immunotherapy in this setting. To increase NK cell specificity, we have developed a B7-H3-targeting trispecific killer engager (TriKE) that can drive NK cell cytolytic responses against B7-H3-expressing tumor cells. B7-H3, a checkpoint member of the B7 family, is highly expressed on HNSCC and its expression correlates with poor prognosis and higher proportions of immunosuppressive cells in the tumor microenvironment (TME).²⁹ Comprehensive multiomics data analyses have found substantial upregulation of B7-H3 gene and protein expression in HNSCC tissues compared with healthy tissues,²⁹ demonstrating a potential for minimizing on-target/off-tumor effects. In this study, we report the preclinical efficacy of the B7H3 TriKE in HNSCC.

Methods

Blood from healthy donors and patients with HPV⁻ HNSCC, and data from The Cancer Genome Atlas

Blood was analyzed from deidentified healthy donors (Memorial Blood Centers, Minneapolis, Minnesota, USA) and patients with HPV⁻ HNSCC (IRB Biobank protocol: STUDY00003639) at MHealth Fairview University of Minnesota Medical Center (Minneapolis, Minnesota, USA) in compliance with guidelines by the Committee on the Use of Human Subjects in Research, and in accordance with the Declaration of Helsinki. HNSCC clinical information from 523 patients was acquired from The Cancer Genome Atlas (TCGA) PanCancer database, through cBioPortal (https://www.cbioportal.org).

Cell lines and cell culture

Raji, DU145 and C4-2 were obtained from American Type Culture Collection (ATCC) and cultured in RPMI-1640 (Gibco Cat. No. 2240–089). HPV⁻ HNSCC cell lines, wild-type (WT) JHU-SCC-FaDu and FANCA knockout JHU-SCC-FaDu, obtained from FA cell line repository through Oregon Health and Science University were cultured in MEM (Corning Cat. No. 10–010-CV) whereas WT Cal27, WT Cal33 and their respective FANCA knockout counterparts (kindly provided by Ramon Garcia-Escudero, Centro se Investigaciones Energéticas Medioambientales y Tecnológicas, Madrid, Spain),³⁰ were cultured in DMEM (Corning Cat. No. MT10013CV). BT12 and BT16, as well as their B7-H3 knockout counterparts (kindly provided by Crystal Macka, Stanford University, California, USA), were cultured in DMEM with 5% GlutaMAX supplement (Gibco Cat. No. 35050061). 2H-11 obtained from ATCC and murine cancer-associated fibroblasts derived from primary mammary tumors in Polyoma Middle T mice (generated in the Provenzano laboratory, University of Minnesota, Minnesota, USA) were cultured in DMEM. All media were supplemented with 10% fetal bovine serum (FBS) (Gibco Cat. No. 26140079) and 100 U/mL Penicillin and Streptomycin (Pen/Strep) (Gibco Cat. No. 15140122). WT NK92 cells (kindly provided by Bruce Walcheck, University of Minnesota, Minnesota, USA), were cultured in Alpha MEM (Gibco Cat. No. 12571), supplemented with 12.5% FBS, 100 U/mL Pen/Strep, 12.5% horse serum (Gibco, Cat. No. 26050088), 200 U/mL recombinant human interleukin (IL)-2 (Prometheus, Cat. No. NDC 65483-116-07) and 0.1 mM $\beta$-mercaptoethanol (Sigma Aldrich Cat. No. M7522). All human cell lines were sent for short tandem repeat genotyping at the University of Arizona Genetics Core for authenticity verification.

B7H3 TriKE plasmid design and protein production

The B7-H3 camelid nanobody sequence was obtained in a custom screen from Creative Biolabs in which clones against B7-H3 were screened for highest affinity and specificity using a phage display single domain antibody library, CaV_HHL-3 (https://www.creative-biolabs.com). The identified sequence was inserted into the backbone of a TriKE plasmid as described previously.³¹ The TriKE plasmids were transfected into Expi293F cells (Gibco Cat. No. A14527) and the protein was purified by immobilized metal ion affinity column using the ÄKTA avant 150 chromatography system. The 10X His-tag at the C-terminus of the TriKE molecule allows for protein purification His-Trap excel columns (Cytiva, Cat. No. 17371206) and downstream detection. Desalting steps were carried out using the HiPrep 26/10 Desalting kit (Cytiva Cat. No. 17508702). The purity of the TriKE protein was validated using SDS PAGE gel stained with One-Step Blue (Biotium Cat. No. 21003–1 L), and protein was quantified using Qubit Protein Assay kit (Invitrogen Cat. No. Q33211).

ELISA

B7H3 TriKE protein was evaluated by custom ELISA using the DuoSet ELISA ancillary reagent kit (R&D Cat. No. DY008B) containing supplementary supplies, whereas the capture and detection proteins were produced in-house using the Expi293 mammalian system. Recombinant human CD16A protein was used as the capture antigen, and detection was via recombinant human B7-H3 extracellular domain protein conjugated to horseradish peroxidase (HRP) using a Lightning-Link HRP conjugation kit (Abcam Cat. No. ab102890).

Matrix-assisted laser desorption ionization-time of flight

Ammonium acetate (Thermo Fisher Scientific Cat. No. AM9070G) buffer exchange for intact protein analysis was carried out, by first diluting protein of interest in ammonium acetate, then passing the diluent through Amicon Ultra centrifugal filter (Millipore Sigma Cat. No. UFC505096). Pierce Bovine Serum Albumin (Thermo Fisher Scientific Cat. No. 23209) was used as the standard, whereas sinapinic acid (Bruker Cat No. 8201345) was used as the matrix for protein analysis. Protein samples were plated onto stainless steel plates (Bruker Cat. No. 8280784) and read using matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) instruments (Bruker autoflex maX).

Evaluation of B7-H3 expression and B7-H3 ligand binding

For phenotyping, cells were stained with APC anti-human B7-H3 (Invitrogen Cat. No. 17-2769-42) at 4°C for 15 min. For binding, cells were incubated with 3 nM B7H3 single domain or B7H3 TriKE at 37°C for 30 min, then stained with Live/Dead Fixable Near-IR (Invitrogen Cat. No. L34976) and PE anti-His tag (BioLegend Cat. No. 362603) or APC anti-His tag (BioLegend Cat. No. 362605) to detect for the 10X His-tag on the TriKE molecule.

Peripheral blood mononuclear cells and NK cell isolation

Peripheral blood mononuclear cells (PBMC) were isolated from Leukopak samples using density-gradient centrifugation with Ficoll-Paque Premium (GE Healthcare, Cat. No. GE17-5442-03). NK cells were enriched from PBMCs using negative selection magnetic beads from EasySep Human NK Cell Enrichment kit (STEMCELL Technologies, Cat. No. 19055). Cryopreserved PBMCs were thawed and rested overnight at 37°C whereas enriched NK cells were used directly in assays.

IL-15 bioavailability assay

NK92 cells were cultured in IL-2-free media for 24 hours, then plated in clear bottom 96-well plate (Corning, Cat. No. 3603) with NCI IL-15 or B7H3 TriKE serially diluted 1:3 eight times, starting from 120 nM. After 48 hours of drug treatment, resazurin (R&D Systems, Cat. No. AR002) was added to the samples, and fluorescence readings were measured using a Tecan Infinite 200 PRO plate reader after 3 hours. Dose-response curves were generated to determine the EC₅₀ values for the respective drugs. From the average of three technical replicates, the IL-15 moiety in the TriKE was determined to be 15 times less potent compared with National Cancer Institute IL-15. Hence, the “equifunctional” dose of IL-15 for 3 nM of TriKE is 0.2 nM, as those amounts elicit an equivalent NK92 stimulation.

Functional assay

PBMCs were co-cultured with target cells at 2:1 effector to target (E:T) ratio, and treated with equifunctional IL-15 or 3 nM TriKE, then stained with FITC anti-human CD107a (BioLegend, Cat. No. 328606). An hour into incubation, GolgiStop (BD Biosciences, Cat. No. 554724) and GolgiPlug (BD Biosciences, Cat. No. 555029) were added 1:100 and 1:150 respectively. After another 4 hours, the cells were stained with Live/Dead Fixable Near-IR, and surface marker antibodies PE-CF594 anti-human CD3 (BD Biosciences, Cat. No. 562280) and PE-Cy7 anti-human CD56 (BioLegend, Cat. No. 318318). The cells were fixed, permeabilized and stained intracellularly using BV650 anti-human interferon-gamma (IFN-$\mathrm{\gamma }$) (BioLegend, Cat. No. 502538).

Real-time tumor killing assay

To carry out cytolytic assays in two-dimensional and three-dimensional formats at low E:T ratios, we used the IncuCyte platform. Nuclight red-labeled (NLR) Cal27 and Cal33 cell lines were generated by transfection with IncuCyte Nuclight Red Lentivirus (Sartorius Cat. No. 4625) and selected for, using 1.5 µg/mL puromycin. For two-dimensional killing assays, 4,000 NLR Cal27 and Cal33 cells per well were plated on 96-well flat-bottomed plates (Corning, Cat. No. 353072) a day before the addition of enriched NK cells at 5:1 E:T, with equifunctional IL-15 or 3 nM TriKE. Live cell imaging was carried out every hour in an IncuCyte ZOOM machine. Red object counts were taken as live cells and normalized to targets alone and t=0. For three-dimensional (spheroid) killing assays, 2.0×10⁴ NLR Cal27 and Cal33 cells per well were added to ultra-low adhesion 96-well plates (Corning, Cat. No. 7007). Spheroids were allowed to form for 3 days before the addition of enriched NK cells at 5:1 E:T, with equifunctional IL-15 or 3 nM TriKE. Live cell imaging was carried out every hour in an IncuCyte S3 machine. The total red object area was taken as tumor spheroid size and normalized to targets alone and t=0 in Microsoft Excel.

Tumor killing in acute hypoxia using AVATAR advanced impedance assay

For assays necessitating evaluation of cytolytic activity in hypoxia, we used the Avatar advanced impedance (AI). 2.5×10⁴ Cal27 and Cal33 cells per well were plated onto PET 96-well plates (Applied Biophysics, Cat. No. 96W20idf) a day before the addition of enriched NK cells at 5:1 E:T, with equifunctional IL-15 or 3 nM TriKE. Impedance readings were measured using AVATAR AI machines, one set to 20% O₂ and another to 1% O₂ and normalized to targets alone and t=0 in Microsoft Excel.

Prolonged hypoxia modeling using AVATAR (Xcell Biosciences) incubators

Enriched NK cells were cultured in 0.08 nM IL-15 or equifunctional TriKE for a week in either a standard incubator (20% O₂) or specialized AVATAR incubators (1% O₂). Following incubation, cells were harvested to run a two-dimensional killing assay as previously described using the IncuCyte ZOOM.

Time-of-flight mass cytometry

PBMCs from healthy and patient donors were barcoded using the Cell-ID 20-Plex Pd Barcoding kit (Fluidigm, Cat. No. 201060) and stained with surface marker antibody cocktail. Then, samples were permeabilized, stained with intracellular antibody cocktail and fixed. The stained cells were incubated overnight with Cell-ID Intercalator (Fluidigm, Cat. No. 201 192A), then washed and run on time-of-flight mass cytometry (CyTOF) cytometer (Standard Biotools, Helios mass cytometer). CyTOF analysis was carried out using Astrolabe Diagnostics and FlowJo V.10.8 Software (BD Life Sciences).

Proliferation assay

PBMCs were labeled with CellTrace Violet dye (Invitrogen, Cat. No. C34557) and incubated with equifunctional IL-15 or 3 nM TriKE at 37°C for a week. Following incubation, cells were harvested and stained for L/D Fixable Near-IR, PE-CF594 anti-human CD3 and PE-Cy7 anti-human CD56.

Tumor killing using xCELLigence impedance assay

For assays requiring high E:T ratios, which are necessary while using cryopreserved PBMCs, we used the xCELLigence platform. 1.25×10⁵ Cal33 cells per well were plated onto PET 96-well E-plates (Agilent, Cat. No. H063028) a day prior to the addition of PBMCs at an E:T of 20:1 and 3 nM TriKE or equifunctional control treatments. Impedance readings were measured using the xCELLigence machine (Agilent, Real-Time Cell Analysis Multiple Plate), and normalized to targets alone and t=0 in Microsoft Excel.

Cal33 xenograft mouse model

Mouse studies were carried out in accordance with guidelines from Institutional Animal Care and Use Committee (IACUC, No. 2207A40255) at the University of Minnesota. NOD-scidIL2Rg^null (NSG) mice (The Jackson Laboratory, #005557) were used, with a sample size of 14 per group that was previously described to provide adequate statistical power.³² 8–10 week-old male and female mice were subcutaneously administered 2.5×10⁵ Cal33 cells and Matrigel at a 1:1 ratio on their left flanks, then injected with 5.0×10⁶ feeder-expanded NK cells³³ intravenously 3 days later. The mice were administered equifunctional IL-15 or 50 µg TriKE subcutaneously into the loose skin over their necks between the scapula three times per week and calipered every week. Mice that had ulceration or reached tumor volume exceeding 2 cm³ were sacrificed and counted as an event for survival studies. On days 14 and 28, the mice were bled, and the blood was stained with BV785 anti-mouse CD45 (BioLegend, Cat. No. 103149), BV605 anti-human CD45 (BioLegend, Cat. No. 368524), PE-CF594 anti-human CD3, PE-Cy7 anti-human CD56, FITC anti-human CD16 (BioLegend, Cat. No. 302006) and APC anti-human B7-H3.

Mouse xenograft tumor infiltration study and histology

8–10 week-old male mice (n=6 per group) were subcutaneously administered 1×10⁶ Cal33 cells and Matrigel at a 1:1 ratio on their left flanks, then allowed to reach tumor volumes of approximately 1000 mm³, before 2.5 million CD3/CD19 depleted PBMCs were administered intravenously. The mice were dosed with equifunctional IL-15 or 50 µg/dose B7H3 TriKE subcutaneously three times over the course of a week. Tumors were harvested, halved and weighed. One half of the tumor was dissociated using the gentleMACS Octo Dissociator (Miltenyi Biotec) and the tumor dissociation kit (Miltenyi Biotec, Cat. No. 130-095-929) before being strained and stained for flow cytometry. The second half was formalin-fixed and embedded in paraffin for histology. For evaluation of necrosis, digital whole slide images were prepared from H&E stained sections with a slide scanner (EasyScan, Motic) and evaluated using the HALO quantitative image analysis platform (Indica Labs). Individual scanned slides were evaluated by a board-certified veterinary pathologist (Doctor of Medical Science (DMS)) and regions of viable tumor and necrosis were manually annotated. Percent necrosis for each tumor was calculated using Microsoft Excel. For identification of NK cells, multiplex spatial histology was performed on five-micron histological sections with a Phenocycler Fusion (Akoya Biosciences). The sections were deparaffinized and rehydrated, followed by antigen retrieval with EDTA and citrate. Anti-human antibodies against CD45 (Cat. No. 4550121), Granzyme B (Cat. No. 4250055), Ki67 (Cat. No. 4450096) and Pan cytokeratin (Cat. No. 4450093) were purchased from Akoya Biosciences. Anti-human CD7 (Abcam, Cat. No. ab230834) was conjugated to a DNA barcode using an antibody conjugation kit (Akoya Biosciences, Cat. No. 7000009). Complementary barcodes conjugated to Atto550, Alexa Fluor 647 and Alexa Fluor 750 were used in cyclical hybridization, buffer exchange, imaging and stripping cycles on the Phenocycler Fusion platform. Sequential images were aligned with 4',6-diamidino-2-phenylindole (DAPI) staining in each cycle to create a six-channel image of each tissue. Cells were analyzed in QuPATH software.³⁴

Flow cytometry

All stained samples were run by flow cytometry using the BD FACSymphony A3 instrument and analyzed using FlowJo V.10.8 Software (BD Life Sciences).

Statistical analyses

Parametric tests such as one-way analysis of variance with multiple comparison adjustments were used to compare control and treatment groups for in vitro assays. Unpaired t-tests were also carried out for CyTOF data to compare healthy and patient samples. Log-rank (Mantel-Cox) test was carried out on Kaplan-Meier survival curve, whereas Mann-Whitney tests were used to compare groups for in vivo assays. The choice of statistical test used was determined based on the distribution characteristics of the dataset, and the Shapiro-Wilk test was used to assess normality. GraphPad Prism was used to analyze data and generate aggregate data graphs with error bars showing mean±SEM.

Results

A novel TriKE molecule effectively targets B7-H3-expressing HNSCC

To investigate the therapeutic potential of targeting B7-H3 in HNSCC, we first verified the expression of B7-H3 on HNSCC. The prognostic significance of B7H3 on HPV⁻ HNSCC was evaluated using the TCGA pancancer dataset, which revealed that high B7-H3 gene expression results in poor overall survival (figure 1A). Next, two native HPV⁻ HNSCC cell lines, Cal27 and Cal33, were analyzed for surface B7-H3 expression. Both WT HPV⁻ HNSCC lines highly express B7-H3 similar to the prostate cancer DU145 line used as a positive control, compared with the Burkitt Lymphoma Raji line used as a negative control³⁵ (figure 1B). The TriKE consists of three arms; a humanized camelid nanobody against CD16, another camelid nanobody against B7-H3 on tumor cells, as well as a WT IL-15 moiety sandwiched between the two camelid nanobodies (figure 1C). This biologic serves as a cytolytic bridge between NK cells and HNSCC, allowing a targeted NK cell effector response against the B7-H3-expressing tumor cells while providing IL-15 cytokine to the NK cells for their survival, expansion and activation. The novel and specific B7-H3-targeted engager was identified using a phage display single domain library (CaV_HHL-3). This dual camelid TriKE molecule differs from the previously described B7H3 TriKE, which has a single chain variable fragment (scFv) targeting component instead of a camelid nanobody. This new construct enhanced NK cell activation, particularly at lower doses, compared with the scFv construct (online supplemental figure S1A,B). Production of the B7H3 TriKE using Expi293 mammalian system was confirmed using an SDS-PAGE gel which showed glycosylated bands of 55 kDa consistent with its predicted molecular weight of 46 kDa (figure 1D). Analysis of the gel bands showed consistency in purity across multiple batches of protein production (online supplemental figure S1C). When tested against B7-H3 WT and B7-H3 KO tumor cell lines, binding of the B7H3 TriKE was only observed against the B7H3 WT cells (figure 1E). To further validate the ability of the B7H3 TriKE molecule to bind to its two target antigens simultaneously, an ELISA assay was performed. The ELISA plate was coated with recombinant human CD16A for capture of the TriKE, and HRP-conjugated recombinant human B7-H3 was used for detection, demonstrating concentration-dependent detection of the molecule (figure 1F). MALDI-TOF characterization of the protein revealed that approximately 75% of the protein exists as monomers, 12% as dimers, 10% as trimers and 3% as tetramers (online supplemental figure S1D). Next, the HNSCC cell lines were also assessed for their binding to the TriKE. Data indicate that all WT lines bind to the camelid nanobody (figure 1G). B7-H3 expression and binding were also investigated in a model of FA using FANCA gene knockout cell lines, Cal27 FANCA-KO and Cal33 FANCA-KO, that recapitulate the FA genetic background³⁰ as more than 60% of patients with FA have biallelic mutations in the FANCA gene (online supplemental figure S1E,F). Based on minimal differences in B7-H3 expression and binding to the B7H3 camelid nanobody between the WT and FANCA-KO cell lines, we proceeded with using the WT cell lines for subsequent assays.

Figure 1. B7-H3-expressing HPV⁻ HNSCC cell lines can be targeted by the B7H3 TriKE. (A) Kaplan-Meier survival curves were generated from TCGA head and neck cancer dataset comparing overall survival of B7-H3-high versus B7-H3-low patients with HPV⁻ HNSCC. (B) HPV⁻ HNSCC cell lines; Cal27 and Cal33, as well as their FANCA knockout counterparts were evaluated for B7-H3 expression. The Burkitt Lymphoma cell line, Raji, which minimally expresses B7H3, was used as a negative control whereas the prostate cancer cell line, DU145, which highly expresses B7-H3, was used as a positive control. (C) B7H3 TriKE molecule consists of a humanized camelid nanobody binding CD16 on NK cells, an interleukin (IL)-15 moiety and another camelid nanobody binding B7-H3 ligand on tumor cells. Created in BioRender. Felices, M (2025) https://BioRender.com/d6ws4qa (D) Three production batches of B7H3 TriKE can be detected using SDS-PAGE gel. (E) Atypical teratoid rhabdoid tumor cell lines, BT12 and BT16, as well as their B7-H3 knockout counterparts were assessed for B7H3 expression and binding to the B7H3 camelid nanobody of the TriKE molecule to assess for specificity. (F) ELISA assay using recombinant human CD16A capture protein and B7-H3 horseradish peroxidase detection protein was carried out to detect the B7H3 TriKE. (G) B7-H3-expressing HPV⁻ HNSCC cells were assessed for binding to the B7-H3 domain of the TriKE molecule. HNSCC, head and neck squamous cell carcinoma; HPV, human papillomavirus; HRP, horseradish peroxidase; NK, natural killer; TCGA, The Cancer Genome Atlas; TriKE, trispecific killer engager; WT, wild-type.

The B7H3 TriKE induces targeted NK cell activation against HNSCC cells

To evaluate the activity of the B7H3 TriKE, NK cell degranulation and cytokine production were evaluated in the presence of HNSCC targets. Functional testing of the B7H3 TriKE was controlled with “equifunctional” concentrations of IL-15 as the IL-15 moiety in the B7H3 TriKE is 15 times less potent than recombinant human IL-15 (online supplemental figure S2A,B). In these assays, NK cells cultured with Cal27 or Cal33 targets induced very low degranulation and IFN-$\mathrm{\gamma }$ production without additional treatment or with the IL-15 control. However, when B7H3 TriKE was added, the proportion of NK cells degranulating and producing IFN-$\mathrm{\gamma }$ markedly increased (figure 2). The low levels of background NK cell degranulation and cytokine production with TriKE treatment in the absence of targets could be explained by B7-H3 expression on dendritic cell and monocyte subsets within the PBMCs as an on-target off-tumor effect (online supplemental figure S2C). To differentiate the relative contributions of the targeting component versus the IL-15, we compared the B7H3 TriKE with a B7H3 BiKE (CD16-B7H3, without the IL-15 component). We found that the B7H3 BiKE induced slightly higher NK cell degranulation compared with the TriKE at low doses. In contrast, the TriKE induced higher IFN-$\mathrm{\gamma }$compared with the BiKE, as IFN-$\mathrm{\gamma }$ production is enhanced by the IL-15 moiety in the TriKE molecule, which primes the NK cells (online supplemental figure S2D,E). The B7H3 single domain alone did not induce NK cell activation beyond baseline levels (online supplemental figure S2F,G). The B7H3 TriKE induced similar amounts of NK cell activation against FANCA KO cell lines as seen in the WT cells (online supplemental figure S2D,E), supporting the applicability of the B7H3 TriKE in targeting FANCA-deficient HNSCC, as well as HPV⁻ HNSCC in general.

Figure 2. B7H3 TriKE triggers NK cell activity against HPV⁻ HNSCC. (A–C) Thawed PBMCs (n=8) from healthy donors were incubated for 5 hours with Cal27 and Cal33 under different conditions: no treatment, equifunctional IL-15 or 3 nM B7H3 TriKE. Flow cytometry dot plots show CD107a as a marker for degranulation and intracellular IFN-γ production for cultures of effectors and Cal33 cells treated under different conditions. (D–E) Aggregate bar graphs show percentage of CD107a+ and IFN-γ+ population gated on live NK cells. One-way ANOVA with post hoc Tukey’s multiple comparisons test was carried out to determine statistical significance. Every mean was compared with the mean of B7H3 TriKE treatment group. Error bars indicate SEM, and statistical significance was determined as *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001. ANOVA, analysis of variance; FITC, fluorescein isothiocyanate; HNSCC, head and neck squamous cell carcinoma; HPV, human papillomavirus; IFN-γ, interferon-gamma; IL, interleukin; NK, natural killer; PBMC, peripheral blood mononuclear cell; TriKE, trispecific killer engager.

The B7H3 TriKE elicits strong NK cell cytotoxicity against HNSCC

To evaluate whether increased NK cell activation translates to increased NK cell killing of HNSCC targets, enriched NK cells from healthy donors were incubated with NLR-labeled Cal27 and Cal33 cells, in the presence of equifunctional IL-15 or 3 nM B7H3 TriKE, and tumor killing was tracked for 48 hours in the IncuCyte platform. The data showed a significant increase in tumor target killing elicited by the B7H3 TriKE (figure 3A,B). To recapitulate a more biologically relevant solid tumor structure, Cal27 and Cal33 spheroids were formed over 72 hours and cytotoxicity was assessed in this system. Data from this three-dimensional assay demonstrated enhanced HNSCC spheroid killing with NK cells and the B7H3 TriKE when compared with other treatments (figure 3C–F).

Figure 3. B7H3 TriKE induces NK cell killing against HPV⁻ HNSCC in real-time imaging assays. (A–B) Enriched NK cells (n=4) from PBMCs of healthy donors were incubated with Nuclight red-labeled Cal27 or Cal33 cells under different conditions: no treatment, equifunctional IL-15 or 3 nM B7H3 TriKE for 48 hours in an IncuCyte ZOOM imager. Quantifications of percent live cells were calculated by normalizing hourly counts of red cells to targets alone at t=0. (C–D) Spheroids of Nuclight red-labeled Cal27 and Cal33 were formed for 72 hours before incubating with enriched NK cells (n=4) under different conditions: no treatment, equifunctional IL-15 or 3 nM B7H3 TriKE for 96 hours in an IncuCyte S3 imager. Representative images of Cal27 or Cal33 spheroids show tumor cell killing over time. (E–F) Aggregate graphs for Cal27 and Cal33 show quantifications of percent total red object area (live cells) by normalizing hourly counts of total red object area to targets alone at t=0. One-way ANOVA with post hoc Tukey’s multiple comparisons test was carried on quantifications of area under the curve to determine statistical significance. Every mean was compared with the mean of B7H3 TriKE treatment group. Error bars indicate SEM, and statistical significance was determined as ****p<0.0001. ANOVA, analysis of variance; 2-D, two-dimensional; 3-D, three-dimensional; HNSCC, head and neck squamous cell carcinoma; HPV, human papillomavirus; IL, interleukin; NK, natural killer; PBMC, peripheral blood mononuclear cell; TriKE, trispecific killer engager.

The B7H3 TriKE can bypass hypoxia-induced suppression of NK cells

As HNSCC is a solid tumor, it is important to consider whether the B7H3 TriKE is as effective in inducing cytotoxicity against HNSCC targets in the hypoxic (1% oxygen) solid TME as it is in the standard (20% oxygen) incubator conditions. To evaluate this, both acute and chronic exposure was evaluated. In the acute scenario, to simulate conditions of circulating NK cells that just entered the hypoxic TME, enriched NK cells were placed in normoxia or hypoxia for 48 hours while cytolytic activity was measured in real time in the presence of equifunctional IL-15 or B7H3 TriKE using a specialized built-in impedance platform (figure 4A). At both oxygen conditions, the B7H3 TriKE was able to robustly boost killing of the tumor targets compared with other conditions (figure 4B–E). In the chronic scenario, a model of prolonged hypoxia was generated by culturing enriched NK cells in low dose IL-15 (0.08 nM) or equifunctional B7H3 TriKE, both of which minimally induce proliferation in hypoxia (1% oxygen) for 7 days (figure 4F).³⁶ Following incubation, the NK cells were treated with equifunctional IL-15 or 3 nM B7H3 TriKE, and cytolytic activity against HPV⁻ HNSCC targets was assessed. Long-term incubation of IL-15 at 20% oxygen showed higher cytotoxicity compared with the IL-15 treatment in the two-dimensional killing assay (without 7-day incubation period) due to priming of the cells (figure 4G,H). After exposure to prolonged hypoxia, IL-15 treatment displayed a strong reduction in killing while the B7H3 TriKE treatment retained robust cytolytic function against HNSCC cells (figure 4I,J). Altogether, these data strongly suggest that the B7H3 TriKE can bypass time-dependent hypoxic suppression of NK cell function.

Figure 4. B7H3 TriKE enhances killing of HPV⁻ HNSCC by NK cells in both acute and prolonged hypoxia models. (A) Enriched NK cells (n=5) from PBMCs of healthy donors were incubated with Cal27 or Cal33 cells under different conditions: no treatment, equifunctional IL-15 or 3 nM B7H3 TriKE for 48 hours in 20% (standard) or 1% (hypoxia) oxygen-controlled impedance-based Avatar AI units. Created in BioRender. Felices, M (2025) https://BioRender.com/rxqaiwy (B–E) Aggregate graphs show killing of tumor targets by NK cells in standard or hypoxic oxygen conditions. (F) Enriched NK cells (n=4) were incubated in 0.08 nM IL-15 or equifunctional B7H3 TriKE over a 7-day period in an incubator with 20% oxygen (standard) or a specialized incubator with 1% oxygen (hypoxia). Cells were briefly taken out of incubators on Day 4 for media and drug change. On Day 7, cells were harvested and plated onto Cal27 or Cal33 targets, treated with equifunctional IL-15 or 3 nM B7H3 TriKE, then placed into the IncuCyte ZOOM imager for 48 hours. (G–J) Aggregate graphs show killing of tumor targets by NK cells cultured in 20% oxygen, or 1% oxygen conditions. Quantifications of percent live cells were done by normalizing hourly counts of red cells to targets alone at t=0. (F) One-way ANOVA with post hoc Tukey’s multiple comparisons test was carried on quantifications of area under the curve to determine statistical significance. Every mean was compared with the mean of B7H3 TriKE treatment group. Error bars indicate SEM, and statistical significance was determined as ****p<0.0001. AI, advanced impedance; ANOVA, analysis of variance; HNSCC, head and neck squamous cell carcinoma; HPV, human papillomavirus; IL, interleukin; NK, natural killer; PBMCs, peripheral blood mononuclear cells; TriKE, trispecific killer engager.

The B7H3 TriKE induces activation and proliferation of NK cells from patients with HPV⁻ HNSCC

As circulating NK cells from patients with cancer may become dysfunctional due to constant exposure to tumor antigens,³⁷ we sought to investigate the characteristics of NK cells from patients with HPV⁻ HNSCC using CyTOF analysis. Six of the seven patients evaluated were treatment-naïve, and one had surgical resection before recurrence of the disease (online supplemental table S1). The levels of NK cell cytolytic effector molecules (granzyme B and perforin), an activating receptor (CD16), an activation marker (CD69) and an immune checkpoint receptor (PD-1) were higher in patients with HPV⁻ HNSCC when compared with age and sex-matched healthy donors, though only granzyme B and PD-1 reached statistical significance (figure 5A–E). Evaluation of NK cell function in the patient samples showed robust B7H3 TriKE-induced increases in degranulation and IFN-$\mathrm{\gamma }$ production against tumor targets (figure 5F–G), with levels comparable to those observed in age and sex-matched healthy donors (online supplemental figure S3B–D). To evaluate tumor killing directly, an impedance-based cytolytic assay was carried out using viably frozen patient PBMCs. There was a consistent increase in tumor killing, noted by a decrease in tumor impedance, when patient PBMCs were treated with the B7H3 TriKE (figure 5H). Induction of cytotoxicity in this experiment was less than that seen in figure 3, which can be explained by the use of cryopreserved PBMCs used here (vs fresh NK cells in figure 3).³⁸ The B7H3 TriKE also induced robust ex vivo proliferation of patient NK cells, similar to what was seen with IL-15 treatment (figure 5I,J), and at levels consistent with those observed in age and sex-matched healthy donors (online supplemental figure S3E–G). These studies using HPV⁻ HNSCC patient-derived blood samples, a target population for this therapy, demonstrate the function of the B7H3 TriKE in a clinically relevant HNSCC patient cohort. There was minimal T cell proliferation showing that the IL-15 in the TriKE mainly targets NK cells (online supplemental figure S3H). We also compared the functionality of the B7H3 TriKE and the B7H3 BiKE (CD16-B7H3, without the IL-15 component) and found that IL-15 in the B7H3 TriKE molecule induces potent NK cell proliferation whereas CD16 ligation alone in the BiKE does not (online supplemental figure S3I,J).

Figure 5. B7H3 TriKE induces activation, proliferation and cytolytic ability of NK cells from patients with HPV⁻ HNSCC. (A–E) Time-of-flight mass cytometry (CyTOF) was carried out using frozen PBMCs (n=7) from HPV⁻ HNSCC patient samples to investigate protein expression levels of granzyme B, perforin, CD16, CD69 and PD-1, that were gated on CD56+ NK cells. Unpaired t-tests were carried out to determine statistical significance. (F–G) Frozen PBMCs (n=7) from HPV⁻ HNSCC patient samples were incubated for 5 hours with Cal33 under different treatments: no treatment, equifunctional IL-15 or 3 nM B7H3 TriKE to evaluate CD107a as a marker for degranulation and intracellular IFN-γ production. Aggregate data shows percent CD107a and IFN-γ induced by NK cells against effectors alone and Cal33. (H) Frozen PBMCs (n=5) from HPV⁻ HNSCC patient samples were incubated with Cal33 cells at an E:T of 20:1 under different treatments: no treatment, equifunctional IL-15 or 3 nM B7H3 TriKE in an impedance-based cytotoxicity assay over a period of 12 hours to evaluate for tumor killing. One-way ANOVA with post hoc Tukey’s multiple comparisons test was carried on quantifications of area under the curve to determine statistical significance. (I–J) CellTrace Violet-labeled frozen PBMCs (n=5) from HPV⁻ HNSCC patient samples were incubated for 7 days under different treatments: no treatment, equifunctional IL-15 or 3 nM B7H3 TriKE to evaluate for NK cell proliferation. Aggregate data show the percentage of NK cells that have undergone more than one division, or three divisions. One-way ANOVA with post-hoc Tukey’s multiple comparisons test was carried out to determine statistical significance. Every mean was compared with the mean of B7H3 TriKE treatment group. Error bars indicate SEM, and statistical significance was determined as *p<0.05, **p<0.01, ****p<0.0001, whereas p<0.1 was indicated numerically, and p>0.1 was not annotated. ANOVA, analysis of variance; E:T, effector to target; HNSCC, head and neck squamous cell carcinoma; HPV, human papillomavirus; IFN-γ, interferon-gamma; IL, interleukin; MMI, mean metal intensity; NK, natural killer; PBMC, peripheral blood mononuclear cell; PD-1, programmed cell death protein-1; TriKE, trispecific killer engager.

The B7H3 TriKE slows tumor growth and prolongs survival of HPV⁻ HNSCC-engrafted mice

The in vivo efficacy of the B7H3 TriKE was tested next in a xenogeneic model using immunocompromised NSG mice that have been engrafted with Cal33 cells. These mice were separated into two groups; (1) tumor + NK + equifunctional IL-15 and (2) tumor + NK + 50 µg/dose B7H3 TriKE (3×/week), with 14 mice in each group (figure 6A). A single NK cell infusion was used to focus on the effects of TriKE-mediated NK cell cytotoxicity and proliferation in vivo. The B7H3 TriKE was dosed subcutaneously in this study to best mimic the route of administration planned for clinical study to enhance the pharmacokinetic profile, as seen with rhIL-15.³⁹ The Kaplan-Meier survival curve showed prolonged survival of the TriKE group (median survival of 42 days) compared with the IL-15 group (median survival of 35 days), supporting in vitro observations of increased cytotoxic capability of NK cells when administered with the B7H3 TriKE molecule (figure 6B). Tumor volumes, measured weekly, indicated delayed tumor growth in the B7H3 TriKE-treated mice when compared with IL-15-treated group (figure 6C,D). TriKE activity was NK cell dependent as B7H3 TriKE alone did not control tumor (online supplemental figure S4A,D). Flow cytometric assessment of mouse blood indicated that mice treated with IL-15 had greater NK cell counts compared with mice treated with the B7H3 TriKE at Day 14, but there were no differences in NK cell counts at Day 28 (figure 6E). While the proportion of CD16+ NK cells decreased by approximately 15% between Day 14 and Day 28, a majority of NK cells at both time points retained CD16 to mediate B7H3 TriKE activity (figure 6F). We also assessed the NK cells for B7-H3 expression, which remained low at Day 14 and Day 28, minimizing any risk of fratricide (online supplemental figure S4C). To further ensure that the tumor control seen in this model was due to TriKE-mediated targeting of NK cells without contribution from murine cells, binding of the B7H3 TriKE to primary murine cancer-associated fibroblasts and murine endothelial cells was ascertained. Minimal binding to both murine cell types was observed (online supplemental figure S4D,E).

Figure 6. B7H3 TriKE slows tumor growth and prolongs survival of Cal33-engrafted mice. (A) 8–10 week old NSG mice (n=14) were injected subcutaneously with 250,000 Cal33 cells/ mouse with Matrigel at a 1:1 E:T. 3 days post-tumor engraftment, treatment groups of mice were administered 5 million expanded NK cells intravenously, and given equifunctional IL-15 or 50 ug/ dose B7H3 TriKE subcutaneously. Throughout the study, the IL-15-treated and B7H3 TriKE-treated mice were given respective drugs 3 days a week. All mice were calipered weekly, and mouse bleeds were conducted on Days 14 and 28 post-tumor engraftment. Created in BioRender. Felices, M (2025) https://BioRender.com/9g4gyu3 (B) Kaplan-Meier survival curve showed probability of survival of mice in different treatment groups over time. Log-rank (Mantel-Cox) test was carried out on Kaplan-Meier survival curve to determine statistical difference between the IL-15 and B7H3 TriKE-treated group. (C) Spaghetti plots of tumor volumes for each individual mouse in different treatment groups were plotted against time. Each line represents a different mouse. (D) Tumor volumes were plotted on Days 14, 21 and 28. (E–F) 100 uL of blood was drawn from all mice, stained for NK cells, and checked for percentage of CD16+ cells gated on NK cells. Mann-Whitney tests were carried out on tumor volumes and mice bleed data to determine statistical difference in tumor volumes, NK cell counts and %CD16+ NK cells between the IL-15 and B7H3 TriKE-treated groups. Error bars indicate SEM, and statistical significance was determined as *p<0.05 and **p<0.01, whereas p>0.1 was not annotated. E:T, effector to target; IL, interleukin; NK, natural killer; TriKE, trispecific killer engager.

The B7H3 TriKE promotes tumor necrosis in HPV⁻ HNSCC-engrafted mice

To investigate whether tumor control in the in vivo study was driven by NK cell infiltration and activation within the TME, a tumor infiltration study was carried out (figure 7A). Cal33 cells were administered subcutaneously, and the tumors were allowed to grow to approximately 1000 mm³ before the administration of NK cells. The mice were dosed with IL-15 or B7H3 TriKE for a week and tumors were harvested and processed for flow cytometric analysis, H&E staining, or multiplex spatial histology. Flow cytometry analysis showed NK cells were present in tumor samples of both treatment groups. NK cell counts, and expression of CD16, CD69 and Ki67 were similar in the IL-15-treated tumors and the B7H3 TriKE-treated tumors (figure 7B–F). In marked contrast, necrosis scoring from the H&E images indicated more necrotic areas in tumors from the B7H3 TriKE-treated mice compared with IL-15-treated mice (figure 7G,H). Multiplex immunofluorescence staining using markers (CD45 and CD7) and features (Granzyme B and Ki67) of NK cells identified NK cells present within the tumors, as seen in flow cytometry data (figure 7I,K). When assessed for the proximity of NK cells to each other within the tumor, the heat density maps (where dark colors indicated low density and bright colors indicated high density) showed NK cells at the core and periphery of the tumors with both treatments (figure 7J,L). These findings confirm that NK cells can infiltrate tumors following TriKE treatment, which contributes to tumor necrosis and superior tumor control.

Figure 7. The B7H3 TriKE promotes tumor necrosis in Cal33-engrafted mice. (A) Subcutaneous Cal33 tumors were allowed to grow in mice (n=6) to 1000 mm³ before a single dose of CD3/CD19-depleted cells was administered intravenously. The mice were dosed with equifunctional IL-15 or 50 ug/ dose B7H3 TriKE subcutaneously three times per week. A week after NK cell administration, the tumors were harvested and halved, with one half being dissociated and examined by flow cytometry for (B-C) human CD45+ and NK cell counts and surface expression of (D-F) CD16, CD69 and Ki67. The other half of the tumors were fixed, paraffin-embedded and examined through (G–H) H&E for tumor necrosis scoring and multiplex spatial histology using antibodies against human proteins to (I, K) identify features (Granzyme B and Ki67) of NK cells (CD45 and CD7) within tumors (Pan cytokeratin) and to (J, L) assess the proximity of NK cells to one another, where dark colors indicated low density and bright colors indicated high density. Unpaired t-tests were carried out to determine statistical significance of flow cytometry data, and quantifications of necrosis. Error bars indicate SEM, and statistical significance was determined as ***p<0.001 whereas p>0.1 was not annotated. The study workflow was created in BioRender. Kennedy, P (2025) https://BioRender.com/mq034t0. FSC, forward scatter; IL, interleukin; NK, natural killer; MFI, median fluorescence intensity’ SSC, side scatter; TriKE, trispecific killer engager.

Discussion

Here, we present a novel therapeutic strategy using a B7H3 TriKE molecule to target HNSCC, a cancer characterized by 66.1%⁴⁰ to 86%⁴¹ B7-H3 expression. We chose to focus our studies mostly on the HPV⁻ population due to poor outcomes in this subset, but our findings apply to the other subsets as well. Our TCGA data analysis highlights the opportunity to target this antigen in HPV⁻ patients. The B7H3 TriKE demonstrated high specificity by binding exclusively to B7-H3-expressing tumor cells, which could minimize on-target off-tumor effects to ensure both efficacy and safety. Besides acting as a cytolytic bridge between NK cells and the tumor cells, we previously showed that TriKE molecules can specifically deliver IL-15 to NK cells instead of T cells.^{31 42} We also show in this study that the B7H3 TriKE does not induce proliferation of T cells from patients with HNSCC. This targeted delivery could not only enhance NK cell activation and proliferation, but also minimize T-cell-mediated toxicity. Indeed, patient-derived NK cells showed enhanced activation and trends of higher proliferation with the B7H3 TriKE compared with IL-15 alone. CyTOF analysis further supported these findings, showing increased levels of CD16, CD69 and PD-1 on NK cells in patients with HPV⁻ HNSCC compared with healthy donors. This suggests that patient NK cells possess the necessary receptors for TriKE-mediated function, but there is a potential for combination of this approach with checkpoint therapy.⁴³

As hypothesized, enhanced tumor killing of HNSCC was observed not only in vitro, but also in an in vivo model. NK cells within patient-derived PBMCs were activated by the B7H3 TriKE and were able to better control the tumor. In a xenograft mouse model, the B7H3 TriKE treatment induced better control of the tumor by expanded NK cells, which could be attributed to Granzyme B+ and Ki67+ NK cell infiltration and localization within the tumor, resulting in higher tumor necrosis when an antibody-dependent cellular cytotoxicity (ADCC) mediating molecule is present. NK cells were still detected in circulation of the B7H3 TriKE-treated mice at Day 28 despite only a single dose of NK cells at the beginning of the study. This finding emphasizes the role of the IL-15 moiety within the B7H3 TriKE to promote NK cell persistence, a limitation that needs to be overcome,⁴⁴ especially in the hypoxic TME setting. Given hypoxia’s role in suppressing immune responses, particularly on NK cells,^{36 45} it is an important consideration when evaluating the efficacy of the B7H3 TriKE. Our studies showed that cytolytic activity against HNSCC in the context of hypoxia was greatly improved by treatment with the B7H3 TriKE, compared with IL-15 alone, indicating that a molecule incorporating ADCC and IL-15 in the same scaffold could help bypass cytolytic limitations driven by hypoxia. While the levels of oxygen deprivation vary across cancers, there is evidence that found HNSCC to be among the most hypoxic tumor types.⁴⁶ Thus, having a therapeutic intervention that can mitigate this immunosuppressive barrier could ensure its efficacy.

One of the key novel aspects of this molecule is in its dual camelid design. The B7H3 TriKE described here differs from the previously published B7H3 TriKE that has an anti-B7H3 scFv.³⁵ The modification of the former scFv tumor-targeting arm to the current camelid nanobody version in the B7H3 TriKE described in this study could be advantageous for several reasons. The smaller molecular weight of the camelid nanobody (15 kDa) compared with the heavier scFv (30 kDa) facilitates biodistribution, which is vital for efficient penetration of solid tumors.⁴⁷ The smaller size also makes it easier to engineer.⁴⁷ The size of TriKE molecules may necessitate more frequent dosing in the clinical setting due to its short half-life determined by rapid clearance through the kidneys.⁴⁷ Camelid nanobodies are also more thermally stable,⁴⁷ making them ideal for therapeutic applications where stability is crucial. Incorporating two camelid nanobodies into the design of the TriKE molecule may reduce steric hindrance, enhance target specificity and binding affinity, and lead to more effective tumor cell engagement and improved therapeutic outcomes. Indeed, functional assays demonstrated that the dual camelid B7H3 TriKE enhanced NK cell activation compared with the scFv version of the molecule.³¹ A potential limitation of the TriKEs is that the individual antigen engagers are monovalent, precluding avidity towards tumor antigens and NK cell receptors. Bivalent approaches to increase avidity are currently being explored.

A consideration for the B7H3 TriKE therapy that remains unexplored in this study is the expression of B7-H3 on stromal cells⁴⁸ and myeloid-derived suppressor cells⁴⁹ within the TME. Even though there might be on-target off-tumor effects, leading to these cells acting as a “sink” for the B7H3 TriKE, targeting these TME cells might be beneficial as the B7H3 TriKE could disrupt supportive functions these cells provide to the tumor, potentially weakening the tumor’s defenses and enhancing immune cell infiltration. This dual effect of targeting both tumor and TME cells offers a promising but complex approach that warrants further investigation to optimize efficacy while minimizing off-target impacts.

Altogether, this study highlights the promising translation of this therapeutic approach to the clinic and suggests that patients with HNSCC, particularly the HPV⁻ populations, including HPV⁻ HNSCC in patients with FA, may benefit from the B7H3 TriKE, as B7-H3 expression was consistent across the subgroups. The efficacy of Food and Drug Administration-approved cetuximab and pembrolizumab is suboptimal in the HNSCC setting and may result in severe side effects in these patients. Unlike current ADCC-centric therapeutic antibodies, the TriKE approach uniquely combines the synergy of coupling ADCC with cytokine signaling to effectively drive NK cells towards HNSCC cells with enhanced cytotoxic potential even in the hypoxic TME. Therefore, the novel B7H3 TriKE, as shown in this study, offers a targeted approach that addresses these limitations in current therapies, and could be a safer, more effective treatment option for patients who have exhausted conventional treatments. Given the consistent B7-H3 expression in patients with HNSCC regardless of HPV status, the B7H3 TriKE could be a versatile therapy for a wider range of HNSCC patient subsets. More extensively, this TriKE immunotherapy could also be used in a broader range of hematologic and solid cancers characterized by high B7-H3 expression. These findings indicate the potential for the B7H3 TriKE treatment in the setting of HNSCC, and merit further clinical investigation.

Data availability statement

Data are available upon reasonable request.