Memory-like Differentiation, Tumor-Targeting mAbs, and Chimeric Antigen Receptors Enhance Natural Killer Cell Responses to Head and Neck Cancer

Head and neck squamous cell carcinoma (HNSCC) remains a clinical challenge due to high morbidity and mortality rates associated with current standard-of-care treatments (1, 2). The introduction of immune checkpoint blockade (ICB) with antibodies against PD-1 and PDL-1 has significantly improved treatment outcomes in patients with metastatic HNSCC. Despite advances with PD-1 checkpoint blockade, response rates and overall survival remain low at 20% and 13 months, respectively (3), highlighting the need to explore alternative and complementary immunotherapies for this disease.

To date most immunotherapies for metastatic HNSCC focus on improving T-cell responses (1, 2). In contrast, strategies to enhance innate immunity, including natural killer (NK) cells, have not been widely studied. NK cells are cytotoxic innate lymphoid cells that display potent effector responses against a wide variety of pathogen-infected and tumor cells (4). The NK cell responses to a target cell are regulated by the balance of signaling through inhibitory receptors that recognize MHC-I, and activating receptors that recognize stress-induced ligands, with functionality tuned by cytokine receptors (5). NK cells mediate cytotoxic functions against target cells via granules containing perforin and granzymes or via death receptor ligands. NK cells can be directly activated via signaling through CD16, a receptor that crosslinks after binding the F_c region of IgG bound to target cells. This triggers antibody-dependent cellular cytotoxicity (ADCC), which can be harnessed using therapeutic mAbs to trigger killing of tumor cells (6). Activated NK cells can also secrete cytokines (e.g., IFNγ, TNF) and chemokines (e.g., MIP1α) that modulate the function and trafficking of other immune cells (4). Thus, mechanisms of NK cell recognition of tumor cells are fundamentally distinct from those of T cells. These features allow NK cells to uniquely circumvent immune evasion mechanisms involving reduced MHC-I expression (7), which occurs in 20% to 70% of HNSCC tumors (8–10).

NK cells are frequently deficient or dysfunctional in patients with cancer (11). A prospective study revealed an association between low NK cell function and increased risk of developing cancer (12). Patients with HNSCC have a lower peripheral blood (PB) NK cell number, compared with healthy donor (HD) controls at all stages of disease (13). Furthermore, low NK cell numbers correlate with unfavorable outcomes at all HNSCC stages, including shorter survival (9). Nevertheless, there have been limited data reported on HNSCC patient NK cell phenotype compared with age- and sex-matched healthy individuals. HNSCC exhibit one of the highest median NK cell infiltrations among any tumor type based on transcriptomic analysis (14). However, small exploratory studies of tumor-associated NK cells from patients with HNSCC have shown decreased expression of the activating receptors NKG2D, DNAM-1, NKp30, CD16, and 2B4, and higher expression of the inhibitory receptor NKG2A, compared with PB NK cells (13). Other preclinical HNSCC models suggest that NK cells differentiate into a hyporesponsive state once they reach the tumor microenvironment (14). Thus, new therapeutic strategies to enhance NK cell responses are needed to improve tumor control.

The paradigm of innate immune memory, also referred to as trained immunity, is a rapidly evolving field. It is now widely recognized that innate immune cells can have enhanced recall responses, which was previously thought to be unique to adaptive T and B lymphocytes. Memory-type NK cells differentiate depending on the initiating stimulus, including viral infections, haptens, and combined cytokines (15). Human NK cells activated briefly through IL12, IL15, and IL18 receptors differentiate into memory-like (ML) NK cells that display enhanced functionality including antitumor responses (16). ML NK cells exhibit a variety of attributes that confer superior ability to recognize and control tumor cells, including upregulation of activating receptors, the ability to ignore inhibitory signaling through inhibitory killer Ig-like receptors (iKIR), and enhanced persistence in vivo (17–19). ML NK cells have been pre-clinically evaluated in melanoma and ovarian cancer (20, 21). In addition, in a phase I study for patients with relapsed/refractory acute myeloid leukemia (AML), ML NK cells safely expand without cytokine release syndrome (CRS) or immune effector cell–associated neurotoxicity syndrome (ICANS), and induced composite complete remission (CR) in 47% of patients (22), making them a promising clinical platform for NK cellular therapy. When combined with hematopoietic cell transplantation, same donor ML NK cells expanded in vivo >1,000-fold, persisted for >3 months, and induced composite CRs in 87% of patients with relapsed/refractory AML (23). Furthermore, opportunities to improve ML NK cell recognition of tumor targets have been preclinically advanced for hematologic malignancies, including enhanced ADCC via CD16a (18, 24, 25), and engineering with chimeric antigen receptors (CAR; ref. 26).

EGFR is expressed on more than 90% of HNSCC (10) and is targeted by the FDA-approved mAb cetuximab (27), with one mechanism of action directing NK cells via ADCC. Another promising drug target is Erythropoietin-producing hepatocellular receptor A2 (EphA2), a surface antigen on solid tumors, and its overexpression causes oncogenesis, epithelial–mesenchymal transition, angiogenesis, and cell growth (28, 29). EphA2 CAR T-cell immunotherapy revealed potent responses against EphA2+ solid tumors in vitro and in vivo in preclinical models (30). Preliminary results of EphA2 CAR T cells in a phase I study in patients with recurrent glioblastoma were recently published, providing proof-of-principle for this CAR target (31).

Here, we evaluate ML NK cell differentiation, ADCC targeting of EGFR via cetuximab, and EphA2-CAR engineering as strategies to improve NK cell responses to HNSCC. We hypothesized that the combination of ML NK differentiation alone, or in combination with targeting of EGFR or EphA2, will overcome ineffective NK cell responses against HNSCC tumors in vitro and in vivo. We tested this hypothesis with IL12, IL15, and IL18 to induce ML NK differentiation, EGFR-targeting with cetuximab, and engineering an anti-EphA2 CAR in preclinical models, revealing new translational strategies for NK cell therapy of HNSCC.

The following recombinant human (rh) cytokines were used: rhIL12 (BioLegend), rhIL18 (Akron), and rhIL15 (Miltenyi Biotec; ref. 16). Human HNSCC cell lines UM-SCC1 (Sigma-Aldrich, SCC070; RRID:CVCL_7707), Cal27 (ATCC, CRL-2095; RRID:CVCL_1107), UM-SCC9 (ATCC, CRL-1629; RRID:CVCL_7793), and UM-SCC47 (Sigma-Aldrich, SCC071; RRID:CVCL_7759) were cultured according to ATCC instructions. Cell line verification was performed on all cell lines used in this article by GRCF DNA services at Johns Hopkins University (Baltimore, MD) using short tandem repeat (STR) profiling following ANSI/ATCC ASN-0002-2011, Authentication of Human Cell Lines: Standardization of STR Profiling guidelines. All cell lines were tested to be Mycoplasma free (MycoAlert Plus Mycoplasma Detection Kit, Lonza Rockland, Inc.). EphA2 knockout (KO) cell lines were generated via CRISPR-Cas9 using the Neon Electroporator (EphA2 single-guide RNA sequence: AGGTGCATCAGAGCCGGCGA; Synthego).

Deidentified HNSCC patient samples were collected on Washington University Institutional Review Board–approved protocol (2021-03013; Supplementary Table S1). Resected tumor specimens were mechanically and enzymatically digested to prepare single-cell suspensions. NSG (NOD-scid IL2rg^null; RRID:IMSR JAX:005557) mice (age 6–8 weeks) were obtained from The Jackson Laboratory, bred, and maintained under specific pathogen-free conditions, and used in accordance with our animal protocol approved by the Washington University Institutional Animal Care and Use Committee.

Leukapheresis chambers were obtained from anonymous healthy platelet donors, and NK cells were purified using RosetteSep (StemCell Technologies; ≥95% CD56⁺CD3⁻) followed by Ficoll centrifugation. ML NK cells were generated by overnight stimulation with rhIL12 (10 ng/mL), rhIL18 (50 ng/mL), and rhIL15 (50 ng/mL); control NK cells were cultured in low-dose IL15 (1 ng/mL) as described previously (16).

NK cells were stimulated at indicated timepoints with HNSCC targets for 6 hours in presence of 1 ng/mL rhIL15. Effector to target (E:T) ratio was 5:1, unless otherwise indicated, with anti-CD107a antibody (BioLegend; RRID:AB_1227509) for 6 hours, with Golgi Plug/Stop present in the last 5 hours. When indicated, NK cells were preincubated, IgG1 Isotype control (5 μg/mL; BioLegend; RRID:AB_2801451), anti-NKG2D (5 μg/mL; BioLegend; RRID:AB_2810480), anti-CD2 (5 μg/mL; BD Biosciences; RRID:AB_395731) or anti-CD226 (5 μg/mL; BioLegend; RRID:AB_1279155) blocking antibodies 30 minutes before incubation with tumor targets. In cetuximab experiments, tumor cells were preincubated with anti-EGFR antibody cetuximab (10 μg/mL; Lilly) prior to incubation with NK cells. Cells were stained for flow cytometry analysis as described previously (19). Degranulation (CD107a), TNF and IFNγ were assessed by flow cytometry (16, 20, 26). Data were acquired on a Gallios flow cytometer (Beckman Coulter) and analyzed using FlowJo software (Tree Star v10.6; RRID:SCR_008520).

Cytotoxicity of NK cells was assessed in standard 4-hour ⁵¹Cr release assays as described previously (32). Targets were incubated with cetuximab or isotype control mAb for 30 minutes, washed, and then cocultured with NK cells. Specific killing by NK cells was also evaluated using the IncuCyte live-cell analysis system. Stably transduced GFP-expressing HNSCC were incubated with cetuximab or isotype controls or blocking mAbs for 30 minutes, washed, plated in a 96-well plate, and imaged 2 hours prior to the addition of NK cells as indicated (20). For EphA2 CAR IncuCyte experiments, NK cells were sorted into GFP-positive (CAR+) or GFP-negative (CAR−) cells prior to culturing with HNSCC cell lines and primary patient samples.

NSG mice were irradiated with 250 cGy and intraperitoneally injected with 2.5 × 10⁵ UM-SCC1 CBR-Luc cells on day 0. On day 3, tumor-bearing mice were intraperitoneally injected with 5 × 10⁶ control or IL12/15/18-activated NK cells (19). The presence of tumor was confirmed by bioluminescent imaging (BLI) before NK cell injection, and the mice were imaged biweekly for the duration of the experiment on the AMI-HT optical imaging system (1–60 seconds exposure, bin8, FOV12 cm, open filter) as described previously (19, 24). rhIL2 (50,000 IU) and rhIL15 (10 ng) per mouse were administered intraperitoneally three times per week to support the survival of transferred NK cells. For the long-term survival monitoring experiment in Fig. 1J, cytokine injections were stopped after day 32. For cetuximab experiments, 1 mg/kg of IVIG was given per mouse on day 2 (to block unoccupied Fc receptors in NSG mice). On day 3, 2.5 mg/kg per mouse of cetuximab or rituximab was intraperitoneally injected 1 hour prior to NK cell injection. A total of 1 × 10⁶ control or IL12/15/18-activated NK cells were intraperitoneally injected in tumor-bearing mice. rhIL2 (50,000 IU) per mouse was administered intraperitoneally three times per week to support the survival of transferred NK cells.

The cassette encoding a single-chain variable fragment targeting CD19 (clone: FMC63) CD8α transmembrane, CD137, and CD3ζ was previously cloned into the MND lentiviral backbone to generate the CD19-CD8a-CD137-CD3ζ (19-CAR) vector (26). EphA2-specific CARs (clone:IgG28) EphA2-CD8a-CD137-CD3ζ were similarly cloned using the same strategy (29). A third-generation packaging system pseudotyped with VSV-G was used to generate lentivirus in 293T cells (RRID:CVCL_0063). The produced lentivirus was used to transduce NK cells by spinfection as described previously (26).

Differences between groups were assessed using unpaired t test or ANOVA as indicated within each figure using Prism v9 (RRID:SCR_002798). Linear mixed models were used for repeatedly measured data, followed by the Tukey post hoc test for multiple comparisons. P value < 0.05 was considered significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Raw data, cell lines, and constructs generated in this study are available upon request from the corresponding author.

We hypothesized that ML NK cells would exhibit enhanced functional responses against HNSCC, compared with conventional NK (cNK) cells. To test this, ML NK cells were generated by overnight IL12, IL15, and IL18 activation (Fig. 1A), allowed to differentiate for 7 days in low-dose IL15 (1 ng/mL), and then stimulated with HNSCC cell lines. cNK cells were maintained in low-dose IL15 to support survival (16). In response to various HNSCC cell lines used in our study that included both human papillomavirus (HPV)-positive (UM-SCC47) and HPV-negative (UM-SCC1, UM-SCC9, Cal 27) lines, ML NK cells had significantly increased IFNγ and TNF production, as well as surface CD107a (a surrogate measure of degranulation), compared with cNK cells from the same donors (Fig. 1B–E; Supplementary Fig. S1A–S1C). ML NK cells demonstrated significantly increased short-term cytotoxicity against HNSCC cell lines compared with cNK cells (Fig. 1F; Supplementary Fig. S1D). The enhanced control of HNSCC cells by ML NK cells was prominent in long-term in vitro tumor killing assays, in which ML NK cell eliminated the tumor targets, while cNK cells only delayed tumor cell outgrowth (Fig. 1G). These results indicate that ML NK cells exhibit an enhanced ability to control HNSCC targets in vitro.

Next, human ML NK cells were evaluated for their ability to control HNSCC targets in vivo in immunodeficient NSG mice (Fig. 1H). NSG mice were intraperitoneally injected with 2.5 × 10⁵ UM-SCC1-CBR-luc cells, and 3 days later received a single intraperitoneal injection of 5 × 10⁶ cNK cells or ML NK cells. rhIL2 and rhIL15 were administered to support human NK cell survival (17, 19, 24). Mice receiving cNK cells failed to control HNSCC tumor burden, compared with control mice not receiving human NK cells (tumor only; Fig. 1I). In contrast, mice treated with a single dose of ML NK cells from the same donors exhibited significantly improved control of HNSCC cells up to day 14 compared with mice that received cNK cells (Fig. 1I; Supplementary Fig. S1E). Furthermore, we observed improved survival of mice treated with ML NK cells compared with mice treated with cNK cells, with median survival of 78 days and 108 days, respectively (Fig. 1J). Similar to the intraperitoneal tumor model, we also observed significantly lower tumor signals from subcutaneous flank tumors that received ML NK cells compared with the tumors that received cNK cells (Supplementary Fig. S1G and S1H). These data indicate that ML NK cells have improved in vivo control of HNSCC targets.

A paucity of information exists as to which NK cell–activating receptor–ligand interactions mediate responses against HNSCC targets. To address this, NK cell receptor ligand expression was assessed on primary HNSCC cells and established HNSCC cell lines. Primary HNSCC tumor cells (HSNCC Pt1, HSNCC Pt2) were obtained from primary HNSCC lesions, and IHC staining of cytokeratin 5/6, and p63 confirmed these cells to be HNSCC (Supplementary Fig. S2A). Across primary HNSCC samples, MICA/B (stress-induced ligand for NKG2D), CD58 (ligand for CD2), as well as CD112 and CD155 (ligands for DNAM-1) were highly expressed (Fig. 2A; Supplementary Fig. S2B). Next, to investigate the mechanistic contribution of these receptors to functional NK cell recognition of HNSCC targets, NKG2D, CD2, and DNAM-1 were blocked using mAbs. In the presence of combined blockade of NKG2D, DNAM-1 and CD2, IFNγ and TNF production from both cNK and ML NK cells significantly diminished (Fig. 2B–E). These differences were not associated with reduced cell viability of cNK or ML NK cells (Supplementary Fig. S3A and S3B). Furthermore, degranulation was markedly reduced in both cNK cells and ML NK cells when treated with the combination blockades (Fig. 2D). Consistent with this, the sustained in vitro killing of primary HNSCC targets by cNK cells and ML NK cells was also significantly impaired following blockade of these activating receptors (Fig. 2F). Individual blockade of NKG2D, DNAM-1, or CD2 did not substantially impact NK cell cytokine, but there was reduced tumor cell killing (Supplementary Fig. S3C–S3H). ML NK cells had greater reduction in IFNγ production and killing compared with cNK cells when treated with the combination antibody blockade, consistent with higher expression of NKG2D, CD2, and DNAM1 by ML NK cells compared with cNK cells (Supplementary Fig. S3I–S3K; refs. 16, 19). Together, these results demonstrate the critical role of NKG2D, DNAM-1, and CD2 in the NK recognition of HNSCC, especially the enhanced response of ML NK cells against HNSCC tumor targets.

Cetuximab is an anti-EGFR mAb used for the treating metastatic HNSCC; however, responses to cetuximab when used in the second-line setting are typically of short durations (33, 34). Previous studies have shown that cetuximab can enhance NK cell functional responses against HNSCC by providing additional recognition via CD16a (35). Thus, we evaluated the ability of cetuximab to enhance ML NK cell responses against HNSCC. UM-SCC1, UM-SCC9, or UM-SCC47 cells were preincubated with cetuximab or IgG1 isotype control prior to incubation with cNK and ML NK cells. Cetuximab significantly increased IFNγ, TNF, and degranulation in cNK cells compared with isotype control as expected (Fig. 3A). Notably, the best responses were observed in ML NK cells treated with cetuximab (Fig.3A), with significant improvements over cNK cells treated with cetuximab. In addition, the combination of NK cells and cetuximab resulted in significantly increased cytotoxicity against HNSCC cells in both short-term and long-term killing assays, with ML NK cells demonstrating significantly better killing compared with cNK cells (Fig. 3B and C). Notably, ML NK cells combined with cetuximab eliminated HNSCC tumor targets at low (e.g., 1:1) E:T ratios (Fig. 3C). These results demonstrate that cetuximab can direct and further enhance the ability of ML NK cells to control HNSCC cell line targets in vitro.

We next asked whether cNK and ML NK cell responses against patient-derived primary HNSCC targets could be further enhanced by cetuximab (Fig. 3D). The addition of cetuximab significantly increased IFNγ, TNF, and degranulation by ML NK cells compared with cNK cells in response to primary HNSCC cells (Fig. 3E). Compared with cNK cells, ML NK cells exhibited enhanced in vitro cytotoxicity against primary HNSCC targets (Fig. 3F). Notably, the improved in vitro control of primary HNSCC by ML NK cells was again evident at the low E:T ratio of 0.5 NK cell to 1 target (Fig. 3G). The ability of ML NK cells to kill primary HNSCC cells was further improved when cetuximab was added to the assay (Fig. 3G). Cetuximab alone did not influence the growth of HNSCC tumor cells in vitro (Supplementary Fig. S4A–S4C). These data support that the combination of cetuximab and ML differentiation significantly improved the ability of NK cells to control primary HNSCC targets in vitro, especially at lower E:T ratios.

Next, the combination of human ML NK cells and cetuximab was evaluated for its ability to control HNSCC targets in vivo using a human xenograft model into NSG mice (Fig. 4A). Rituximab (anti-CD20) was used as a control antibody treatment as UM-SCC1 does not express CD20. Mice receiving cNK cells plus cetuximab had similar tumor burden to mice receiving cetuximab alone (Fig. 4B and C). In contrast, mice treated with ML NK cells plus cetuximab exhibited a significantly increased control of HNSCC cells in vivo as measured by BLI, and this is improved over cetuximab alone and cNK cells plus cetuximab, with low tumor burden over the entire duration of the experiment (Fig. 4B and C). No weight loss was observed in mice treated with cetuximab or rituximab (Supplementary Fig. S4D). These data indicate that when combined with cetuximab, a single injection of ML NK cells provided enhanced control of primary HNSCC cells in vivo, compared with cNK cells.

EphA2 is a tumor-associated antigen expressed on HNSCC and other solid tumors that has demonstrated potential to be targeted using CAR T cells (Supplementary Fig. S5A and S5B; refs. 30, 31). Prior work by our group showed that CD19-CAR ML NK cells exhibit an enhanced response to CD19+ lymphoma, compared with CD19-CAR cNK cells (26). To understand whether an EphA2-CAR expression would improve ML NK cell responses against HNSCC targets, we generated EphA2-CAR ML NK cells, utilizing an anti-EphA2 scFv, CD8α transmembrane domain, and CD137 and CD3ζ signaling motifs (Fig. 5A). We observed higher transduction efficiency in ML NK cells compared with cNK cells, and thus in subsequent analyses we flow gate on GFP-positive (CAR-positive) and GFP-negative (CAR-negative) cells within cNK and ML NK cells to specifically identify and compare CAR-positive or -negative NK cells (Supplementary Fig. S5D). EphA2-CAR ML NK cells (GFP positive) exhibited increased IFNγ production and degranulation against UM-SCC9, UM-SCC47, UM-SCC1, and CAL27, in addition to UD-SCC2 that is otherwise resistant to ML NK cells (Fig. 5B–G; Supplementary Fig. S6) compared with CAR-negative (GFP-negative) ML NK cells. To assess how CAR expression affects ML NK cells’ ability to kill HNSCC cells, we utilized sorted GFP+ and GFP− ML NK cells for IncuCyte assays. EphA2-CAR ML NK cells demonstrated a significantly improved ability to eliminate the UM-SCC9 cell line in sustained in vitro cytotoxicity assays, compared with donor-matched ML NK cells (Fig. 5F). These experiments demonstrate that ML differentiation and CAR engineering can be combined to further enhance effector functions against EphA2+ solid tumor targets.

To determine whether CAR intracellular signaling is required for enhanced antitumor responses, we developed an EphA2-CAR with a truncated cytoplasmic domain that lacks signaling components (Supplementary Fig. S7A). ML NK cells expressing this EphA2-CAR^trunc did not demonstrate the enhanced IFNγ production and degranulation observed in EphA2-CAR ML NK cells against UM-SCC9 (Supplementary Fig. S7B and S7C). Next, the specificity of EphA2-CAR ML NK cells for EphA2 was investigated using wildtype (WT) UM-SCC9 or EphA2 KO UM-SCC9 cells (Supplementary Fig. S7D). EphA2-CAR ML NK cells exhibited a significant increase in IFNγ production and degranulation against WT UM-SCC9 targets, but these enhanced responses were not observed against EphA2 KO UM-SCC9 targets (Fig. 5G and H). This finding was also confirmed with EphA2-negative lymphoma cell line Raji (Supplementary Fig. S7E and S7F). CD19-CAR ML NK cells did not have enhanced responses to UM-SCC9 targets (EphA2+, CD19−) compared with ML NK cells without CAR (Supplementary Fig. S7I and S7J). Furthermore, compared with ML NK cells without CAR, EphA2-CAR ML NK cells but not CD19-CAR ML NK cells exhibited enhanced killing of UM-SCC9, even at low CAR⁺ E:T ratios of 1:1 (Supplementary Fig. S7K). Together, these data demonstrate that the enhanced response by EphA2-CAR ML NK cells were CAR antigen specific and required intracellular CAR signaling.

Next, we investigated the functionality of EphA2-CAR ML NK cells against primary HNSCC cells. The two primary HNSCC samples in our study both had high EphA2 expression (Fig. 6A). EphA2-CAR ML NK cells have significantly increased IFNγ production and degranulation against primary patient HNSCC cells, compared with CAR-negative ML NK cells from the same donor (Fig. 6B and C; Supplementary Fig. S7L). This enhanced response was abrogated in EphA2-CAR^trunc ML NK cells (Fig. 6D). EphA2-CAR ML NK cells have significantly improved killing compared with CAR-negative ML NK cells in sustained in vitro cytotoxicity assay against primary HNSCC cells (Fig. 6E). These EphA2-CAR ML NK cell responses against primary HNSCC also required intact CAR intracellular signaling and are CAR target specific (Supplementary Fig. S7M). Thus, engineering with an EphA2-CAR provides a potent activating signal to ML NK cells to specifically attack primary HNSCC targets.

Here, we discovered that ML NK cells exhibit enhanced in vitro and in vivo effector functions against primary HNSCC targets, compared with cNK cells. NKG2D, DNAM-1, and CD2 were identified as key activating receptors responsible for recognition of HNSCC by both cNK and ML NK cells. Moreover, directing cNK and ML NK cells with cetuximab further improved NK cell functional responses against HNSCC, with the combination of cetuximab and ML NK cells consistently demonstrating the most robust responses. We also engineered ML NK cells to express an anti-EphA2 CAR to enhance recognition of HNSCC. ML differentiation and CAR expression combined (EphA2-CAR ML NK cells) yielded enhanced IFNγ production and cytotoxicity against HNSCC targets compared with EphA2-CAR cNK cells and ML NK cells without CAR expression. Thus, this study reports three highly translational and complementary approaches that are amenable to combinatorial testing: ML differentiation, EGFR targeting with cetuximab, and anti-EphA2 CAR engineering to improve NK cell responses against HNSCC. These approaches warrant clinical investigation.

The findings here are promising approaches to address unmet clinical needs in patients with HNSCC. ICB therapy targeting PD-1 as a monotherapy or combined with chemotherapy has led to improved clinical responses in patients with metastatic HNSCC and is currently first-line standard-of-care treatment for patients with metastatic HNSCC (3). However, over 80% of patients with metastatic HNSCC treated with ICB will not respond to treatment, and almost all patients who have an initial response progress within 1 year of treatment (3). A major ICB resistance mechanism is the downregulation of MHC-I antigen presentation pathways leading to evasion of cytotoxic T-cell antigen-specific immune responses (36). This loss of MHC-I expression along with stress-induced ligands allows NK cells to recognize and exert their cytotoxic and immunomodulatory functions against the tumor cells, and thus NK cells are an alternative approach to T cell–based immunotherapies (37). However, previous studies have demonstrated that NK cell numbers and functionality are decreased in patients with HNSCC in advanced clinical stages and as disease progresses (9). This is due to downregulation of NK cell–activating receptors, upregulation of inhibitory receptors, and inhibition through circulating soluble NKG2D ligands (13, 14). We reasoned that ML NK cells can overcome these barriers as they have increased activating receptor expression (NKG2D, NKp46, DNAM-1; refs. 16, 19), the ability to ignore signals via iKIRs (19), and potent in vitro Fc-triggered responses (18, 24). Previous studies have shown that ML NK cells have improved cytotoxicity and cytokine production against human melanoma (20) and ovarian (21) cancers. Preclinical syngeneic mouse models revealed increased trafficking and persistence within solid tumors by ML NK cells compared with cNK cells (38). Studies have also demonstrated ML NK cells have improved metabolic fitness (24, 39), and emerging data suggest that this translates into function in the challenging solid tumor microenvironments (40). These reports support the translation of ML NK cells as therapy for patients with HNSCC. In this study, we advance the field by demonstrating that ML NK cells have improved responses against HPV+ and HPV− HNSCC cell lines and primary HNSCC cells. We further identified NKG2D, DNAM-1, and CD2 as key activating receptors involved, consistent with the findings that their respective ligands MICA/B, CD112, CD155, and CD58 are upregulated on the majority of HNSCC cell lines and primary patient samples assessed.

The use of mAbs to enhance the cytotoxicity of NK cells through ADCC killing has been investigated clinically across multiple cancers. Cetuximab is a mAb that recognizes EGFR expressed on tumor cells (>90% of HNSCCs are EGFR+; ref. 41) and is used clinically in HNSCC and colorectal cancer (2). Cetuximab has demonstrated modest clinical activity as a monotherapy, with response rates of 13%, and a progression-free survival (PFS) of 2 to 3 months (34). Similarly, cetuximab combined with chemotherapy yields modest response rates of 36% and median PFS of 5.6 months (33). These relatively low responses are likely in part due to compromised Fc-bearing cells (NK cells and monocytes/macrophages) within HNSCC and downregulation of CD16 expression, which has been reported in patients across multiple cancers (42). Prior reports established that ML NK cells have enhanced in vitro IFNγ responses to CD16a triggering, as well as enhanced ADCC in vitro, suggesting that memory-like NK cells will be effective responders to antibody-opsonized targets (18, 24, 25). To circumvent issues with impaired NK cell functionality, we evaluated the ability of allogenic ML NK cells from HD combined with cetuximab to enhance functionality of NK cells against HNSCC cells. We demonstrate that cetuximab improved cytotoxicity and cytokine production of both cNK and ML NK cells against HNSCC targets both in vitro and in vivo. ML NK cells when combined with cetuximab offer superior ADCC against HNSCC tumor-bearing hosts compared with cNK cells, which could serve as an additional mechanism for overcoming the suboptimal ADCC seen in patients with HNSCC.

The solid tumor–associated antigen EphA2 is a receptor tyrosine kinase that is expressed at relatively low levels on healthy tissues but is overexpressed on many solid tumors (43). EphA2 CAR T cells have been tested in preclinical osteosarcoma and glioblastoma models (44), and are currently in early phase clinical development for glioblastoma (31). Early reports in patients with glioblastoma showed that EphA2 CAR T cells administered intravenously expanded and trafficked to the brain and cerebrospinal fluid, but no clinical responses were reported and 2 of 3 patients developed grade 2 CRS (31). CAR-engineered NK cells offer several advantages over CAR T cells. CAR NK cells may abrogate issues with CAR antigen escape often seen in CAR T cell therapy as NK cells can engage tumors through their natural cytotoxicity receptors and are amendable to combination therapy with tumor-targeting mAbs. In addition, CD16 expression on mature NK cells allows for flexible dual-targeting strategies with therapeutic mAbs or bispecific antibodies that ligate CD16 and a tumor-restricted antigen. CAR NK cells and ML NK cells have also demonstrated reduced risks of GVHD, CRS, ICANS and prolonged cytopenias, clinically (22, 23, 45, 46). Our group previously developed a rapid process to engineer CD19-CAR ML NK cells for B-cell malignancies (26). Building on this concept, we broaden this finding to solid tumors with EphA2-targeting CAR ML NK cells, and show that EphA2-CAR ML NK cells have enhanced responses against HNSCC, compared with EphA2-CAR cNK cells and ML NK cells without CAR, which are specific to EphA2+ tumor targets.

Similar to all preclinical models our study has limitations. NSG xenografts test the activity of ML NK cells in relative isolation, and cannot recapitulate the immune cell interactions in the tumor microenvironment. Ideally, these approaches would be complemented by syngeneic mouse models in immunocompetent mice. However, HNSCC syngeneic tumor models are lacking, with current models using carcinogens to induce oral cavity cancers in mice or genetically engineered mouse models (47). The extensive prior work evaluating ML NK cells against other cancer types in vitro and in vivo (16, 18–20, 24–26, 38), and within clinical trials (19, 22, 23, 46, 48, 49), mitigates these limitations as solid tumor clinical studies are designed and implemented.

In summary, ML NK cells exhibit enhanced ability to control HNSCC targets in vivo and in vitro, which is further improved with the addition of cetuximab or engineering with EphA2 CAR. Importantly, we demonstrate multiple approaches to improve NK cell functioning against primary HNSCC targets. This provides a rationale for the use of ML NK cells in combination with cetuximab or EphA2 CAR as an alternative cellular therapy to treat patients with advanced HNSCC. These NK cell therapy approaches warrant clinical testing in early phase clinical trials, with the potential of future combinations including dual targeting with EphA2 CAR ML NK cells and cetuximab.

M.T. Jacobs reports grants from NIH and ASCO Conquer Cancer Foundation during the conduct of the study. J.A. Foltz reports grants from NIGMS during the conduct of the study; personal fees from CPRIT outside the submitted work; and has a patent for WO2019/152387, US 63/018, 108 pending, licensed, and with royalties paid from Kiadis and a patent for anti-canine NKp46 antibody licensed and with royalties paid from EMD Millipore. C.C. Cubitt reports equity in Pionyr Immunotherapeutics. J.P. Zevallos reports other support from Droplet Biosciences and Vine Medical, as well as grants from Merck outside the submitted work. D.R. Adkins reports grants and personal fees from Merck, Cue Biopharma, Blueprint Medicine, Kura Oncology, Vaccinex, Boehringer Ingelheim, and Gilead Sciences; personal fees from Exelixis, Immunitas, TargImmune Therapeutics, twoXAR, Xilio Therapeutics, Eisai Europe, Coherus Biosciences, and Jazz Pharmaceuticals; and grants from Pfizer, Eli Lilly, Celgene/BMS, Novartis, AstraZeneca, Cofactor Genomics, Debiopharm Group, ISA Pharmaceuticals, Beigene, Roche, Immutep, Hookipa Biotech, Epizyme, Adlai Nortye, BioAtla, Calliditas Therapeutics, Genmab, Natco Pharma, Tizona Therapeutics, Seagen, Surface Oncology, Janux, Inhibrx, Takeda, and Alentis outside the submitted work. M.M. Berrien-Elliott reports personal fees and other support from Wugen during the conduct of the study; and has a patent for 15/983,275 pending and licensed to Wugen, a patent for 62/963,971 licensed to Wugen, and a patent for PCT/US2019/060005 pending and licensed to Wugen. T.A. Fehniger reports grants, personal fees, and other support from Affimed and Wugen during the conduct of the study; other support from Indapta and Orca Bio; grants from HCW Biologics; personal fees and other support from AI Proteins and Smart Immune outside the submitted work; and has a patent for 15/983,275 pending and licensed to Wugen, a patent for 62/963,971 pending and licensed, and a patent for PCT/US2019/060005 licensed to Wugen. No disclosures were reported by the other authors.

M.T. Jacobs: Conceptualization, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. P. Wong: Formal analysis, investigation, methodology, writing–review and editing. A.Y. Zhou: Formal analysis, investigation, writing–review and editing. M. Becker-Hapak: Formal analysis, investigation, methodology, writing–review and editing. N.D. Marin: Methodology, writing–review and editing. L. Marsala: Investigation. M. Foster: Investigation. J.A. Foltz: Writing–review and editing. C.C. Cubitt: Writing–review and editing. J. Tran: Writing–review and editing. D.A. Russler-Germain: Writing–review and editing. C. Neal: Investigation. S. Kersting-Schadek: Investigation. L. Chang: Investigation. T. Schappe: Investigation. P. Pence: Investigation. E. McClain: Investigation. J.P. Zevallos: Resources, writing–review and editing. J.T. Rich: Resources, writing–original draft. R.C. Paniello: Resources, writing–review and editing. R.S. Jackson: Resources, writing–review and editing. P. Pipkorn: Resources, writing–review and editing. D.R. Adkins: Methodology, writing–review and editing. C.J. DeSelm: Resources, methodology, writing–review and editing. M.M. Berrien-Elliott: Formal analysis, methodology, writing–review and editing. S.V. Puram: Resources, methodology, writing–review and editing. T.A. Fehniger: Conceptualization, resources, supervision, funding acquisition, writing–original draft, writing–review and editing.

The authors thank our patient volunteers and the head and neck cancer physician, nursing, and research assistant teams who care for them at the Washington University School of Medicine. We acknowledge support from the Siteman Flow Cytometry Core and Siteman Tissue Procurement Core. We thank the Genome Engineering and iPSC Center (GEiC) at the Washington University in St. Louis for their gRNA validation services.

The research reported in this publication was supported by grants from the NIH NCI K12CA167540 (M.T. Jacobs), P50CA171963 (T.A. Fehniger, M.M. Berrien-Elliott), R01CA205239 (T.A. Fehniger), P30CA91842 (T.A. Fehniger), 1K08CA237732 (S.V. Puram). National Institute of General Medical Sciences T32GM139799 (J.A. Foltz), 1F31GM146361-01 (J. Tran). NIH National Institute of Allergy and Infectious Disease F30AI161318 (C.C. Cubitt). NIH National Heart, Lung, and Blood Institute T32 HL007088 (P. Wong). NIH National Institute of Dental and Craniofacial Research 5DP5OD026427 (C.J. DeSelm). Additional support was provided by the Siteman Cancer Center (T.A. Fehniger), the ASCO Young Investigator Award via Conquer Cancer Foundation (M.T. Jacobs, D.A. Russler-Germain), Dean's Scholars award from the Washington University Division of Physician-Scientists, which is funded by a Burroughs Wellcome Fund Physician-Scientist Institutional award (M.T. Jacobs), the Paula C and Rodger O. Riney Blood Cancer Initiative (T.A. Fehniger), the Lymphoma Research Foundation (T.A. Fehniger, D.A. Russler-Germain). Barnes Jewish Hospital Foundation (S.V. Puram) and Doris Duke Fund to Retain Clinician Scientists and Clinician Scientist Development Award (S.V. Puram).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.