Combinatorial Natural Killer Cell–based Immunotherapy Approaches Selectively Target Chordoma Cancer Stem Cells

Chordoma is a rare tumor derived from notochord remnants that has a propensity to recur and metastasize despite conventional multimodal treatment. Cancer stem cells (CSC) are implicated in chordoma's resistant and recurrent behavior; thus, strategies that target CSCs are of particular interest. Using in vitro cytotoxicity models, we demonstrated that anti-programmed death ligand 1 (anti–PD-L1; N-601) and anti-EGFR (cetuximab) antibodies enhanced lysis of chordoma cells by healthy donor and chordoma patient NK cells through antibody-dependent cellular cytotoxicity (ADCC). Treatment of NK cells with an IL15 superagonist complex (N-803) increased their cytotoxicity against chordoma cells, which was further enhanced by treatment with N-601 and/or cetuximab. PD-L1–targeted chimeric antigen receptor NK cells (PD-L1 t-haNKs) were also effective against chordoma cells. CSCs were preferentially vulnerable to NK-cell killing in the presence of N-601 and N-803. Flow cytometric analysis of a chordoma CSC population showed that CSCs expressed significantly more NK-activating ligand B7-H6 and PD-L1 than non-CSCs, thus explaining a potential mechanism of selective targeting. These data suggest that chordoma may be effectively targeted by combinatorial NK cell–mediated immunotherapeutic approaches and that the efficacy of these approaches in chordoma and other CSC-driven tumor types should be investigated further in clinical studies. Significance: Combinatory immunotherapy using NK-mediated approaches demonstrates robust antitumor activity in preclinical models of chordoma and selectively targets chordoma CSCs.


Introduction
Chordoma is a rare, slow growing, locally invasive primary bone tumor thought to develop from embryonic notochord remnants, resulting in its predilection for the axial skeleton. The annual incidence of chordoma in the United presentation or specificity for their activity, and coordinate immune responses with antibodies and T cells.
Prior work has demonstrated the efficacy of avelumab (anti-programmed death ligand 1; PD-L1), the only immune checkpoint inhibitor that also promotes antibody-dependent cellular cytotoxicity (ADCC) with NK cells, in preclinical models of chordoma (8). N-601 is an anti-PD-L1 mAb and structural analog of avelumab that has not been previously investigated. N-803 (formerly ALT-803) is a clinical grade IL15 superagonist complex known to induce proliferation and activation of NK-and T-cell immune compartments (9,10).
Cancer stem cells (CSC), or tumor-initiating cells (TIC), are a subpopulation of malignant cells that can drive tumorigenesis and disease relapse (17)(18)(19)(20). This cell type is implicated in tumor heterogeneity and resistance to chemotherapy and radiation. CSCs have been identified in myeloid malignancies, glioblastoma, and cancers of the breast, colon, pancreas, and skin (21). The presence of a CSC population may be in part responsible for chordoma's recurrent nature and resistance to standard therapeutic approaches (8,22,23). Novel therapeutic options should be evaluated in their activity against chordoma CSCs, but the efficacy of immunotherapy against chordoma CSCs remains largely unstudied. Prior work has demonstrated that residential chordoma CSCs can be identified by coexpression of surface markers CD15, CD24, and CD133 (8,22). Herein, we aim to further characterize this cellular subpopulation and investigate the potential of combinatorial NK cell-mediated approaches to selectively target chordoma CSCs. This study also intends to provide preclinical evidence to support future clinical investigation of combinatory NK cell-based immunotherapy approaches in chordoma that implement ADCCmediating antibodies, NK-cell augmentation through cytokine agonism, and targeted adoptive NK-cell therapies.

Cell Lines and Culture
UM-Chor1 (clival) and JHC7 (sacral) were kindly provided by the Chordoma Foundation on approximately April 2, 2015 with short tandem repeat (STR) validation. U-CH1 (sacral), U-CH2 (sacral), UM-Chor5 (clival), and MUG-CC1 (clival) were obtained from ATCC on approximately July 20, 2020 with STR validation. All cell lines were passaged twice per week, used at a low (<30) passage number, and serially verified to be free of Mycoplasma infection using a two-step luminescence assay (Lonza Bioscience).

Study Approval
Healthy donor peripheral blood mononuclear cells (PBMC) were obtained from the NIH Clinical Center Blood Bank (NCT00001846). Whole blood from a treatment-naïve chordoma patient was collected through a tissue procurement protocol (NCT03429036) between the NIH and Johns Hopkins University Hospital (Baltimore, MD). Whole blood from patients with chordoma receiving immuno-oncology therapy (NCT02383498, NCT03173950,   NCT03349983, NCT01519817, NCT01772004, NCT02179515, NCT02536469) was obtained from collaborators at the NIH (Bethesda, MD). Each study was approved by the Institutional Review Board and informed written consent was obtained for enrollment on the respective institutional review board protocol.
These studies were conducted in accordance with U.S. Common Rule.

Immune Reagents and NK Cells
An anti-PD-L1 antibody (N-601), an IL15/IL15r superagonist (N-803), and the PD-L1-specific chimeric antigen receptor-engineered high-affinity NK cells (PD-L1 t-haNKs) were provided by ImmunityBio under a Cooperative Research and Development Agreement (CRADA) with the NCI of the NIH (Bethesda, MD). Cetuximab (Erbitux) was obtained from Lilly. Human IgG1 isotype control antibody and human anti-CD16 antibody were obtained from BioLegend.
PBMCs were isolated from chordoma patient whole blood using BD Bioscience Vacutainer CPT tubes according to the manufacturer's protocol. NK cells were isolated from PBMCs using the Miltenyi Biotec Human NK Cell Isolation (negative selection) Kit according to the manufacturer's protocol with >90% purity.
NK cells were rested overnight in RPMI1640 media before being used as effector cells in assays.
For experiments using chordoma patient NK cells as effector cells, treatments undergone by patients prior to providing blood for this study are as follows: One patient, CP1(n), had a newly diagnosed clival chordoma and provided blood on the day of primary surgical resection. This patient had not received any other therapies (i.e., radiation, immunotherapy) prior to providing blood. A second patient, CP2(io), had a recurrent sacral chordoma and was previously treated with surgical resection, radiation, yeast brachyury vaccine (NCT02383498), anti-PD-1 mAb (NCT03173950), and modified vaccinia Ankara -Fowlpox brachyury vaccine (NCT03349983). A third patient, CP3(io), had a recurrent thoracic spine chordoma and was previously treated with surgical resection, radiation, yeast brachyury vaccine (NCT01519817), anti-PD-L1 mAb (NCT01772004), modified vaccinia Ankara-based vaccine modified to express brachyury and T-cell costimulatory signals (NCT02179515), anti-IL8 mAb (NCT02536469), and modified vaccinia Ankara -Fowlpox brachyury vaccine (NCT03349983).

Indium-111 NK-Cell Killing Assay
NK cells (effector cells) were isolated from healthy donor (HD) PBMCs or chordoma patient PBMCs as previously described. NK cells isolated from PBMCs were cultured overnight in the NK-cell media described above with or without 50 ng/mL of N-803. All NK cells were washed in the above-described media after overnight incubation. Cryopreserved PD-L1 t-haNK cells were thawed on the day of the experiment and washed two times in the previously described NK-cell media before being adjusted to the desired concentration and plated.
Chordoma cells (target cells) were labeled with 111 In (10 μL/100,000 cells). NK cells (healthy donor, chordoma patient, or PD-L1 t-haNKs) were added to wells at various effector-to-target (E:T) ratios, depending on the experiment. Specific E:T ratios are indicated in the figure legends. All E:T ratios are 20:1 unless indicated otherwise. After 20 hours, assay plates were centrifuged at 1,500 rpm and supernatants were quantified for the presence of 111 In using a PerkinElmer WIZARD2 Automatic Gamma Counter. Spontaneous 111 In release was determined by incubating target cells without effector cells, and complete lysis was determined by incubating target cells with 0.05% Triton X-100. Experimental lysis was standardized using the following equation: Percent lysis = [(experimental cpm − spontaneous cpm)/(complete cpm − spontaneous cpm)] × 100. Negative control values in each bar graph represent spontaneous lysis of target cells without effector cells. All experiments were carried out in technical triplicate, with each individual experiment being repeated at least three separate times unless otherwise specified.

Flow-based NK-Cell Killing Assay
UM-Chor1 cells were labeled with CellTrace Violet Proliferation dye (Thermo Fisher Scientific) and cocultured for 4 hours with healthy donor-derived NK cells that had been treated with N-803 overnight. Afterwards, the cocultures were stained with Live/Dead Fixable Aqua stain (Thermo Fisher Scientific) and the CSC marker antibodies described above. Flow cytometry was performed on BD LSRFortessa (BD Biosciences) and analyzed using FlowJo V.10.7.1 (TreeStar). The CSC and non-CSC populations were downsampled using FlowJo such that each cellular subtype would have similar numbers of cells in the histogram.

Statistical Analysis
Experiments were carried out in technical triplicate with each experiment being repeated at least three times unless otherwise specified due to limitations by the NK-cell donor. Comparisons in NK-cell killing models of two independent groups were performed using a Student t test with a two-tailed distribution at a confidence level of 95%. Comparison of more than two independent groups were performed using one-way ANOVA with Tukey multiple comparisons test. All error bars indicate SEM. Test results are reported as P values with a significance cutoff set at P < 0.05. All analyses were performed using GraphPad 9.0 software (GraphPad Software Inc.).

Data Availability
The data generated in this study are available upon request from the corresponding author.

Chordoma Cell Lines from Distinct Anatomic Locations are Susceptible to Lysis by Healthy Donor NK Cells
Preclinical and clinical evidence indicate that tumor immune microenvironments are shaped by tissue of origin and anatomic location, thus impacting disease progression and response to immunotherapies (24). Given the anatomically distinct locations along the axial skeleton where chordoma arises and previously reported differences in survival between cranial chordomas and other locations (25), we first sought to determine whether there is a difference in target surface marker expression between three chordoma cell lines derived from clival chordomas (UM-Chor1, MUG-CC1, UM-Chor5) and three lines from sacral chordomas (JHC7, U-CH1, U-CH2; refs. [26][27][28][29]30). Chordoma tissue samples and cell lines have been shown to express PD-L1 (8,(31)(32)(33)34) and EGFR (35)(36)(37), both of which are surface antigens with clinically available mAbs that mediate ADCC with NK cells (38,39). All six chordoma cell lines examined expressed HLA-A, B, C/MHC-1, PD-L1, and EGFR as assessed by percent positive cells and MFI (Table 1). There were no statistically    Next, we interrogated the susceptibility of chordoma cell lines to lysis by healthy donor (HD) NK cells and determined whether the site of anatomic derivation had any significant impact on this susceptibility. Consistent with prior studies using in vitro NK-cell killing assays (8,39), the average NK cell-mediated lysis of chordoma cell lines was significantly higher than spontaneous lysis nontreated (NT) controls but remained low overall (average: 10.7%; SEM:

HLA-A,B,C/MHC-I PD-L1 EGFR
1.5%), highlighting the potential for augmentation in NK cell-mediated lysis of chordoma cells (Fig. 1). There was no significant difference between NK cellmediated lysis of clival and sacral chordoma cell lines (P = 0.09), indicating that anatomic derivation of chordoma cell lines does not alter sensitivity to lysis by NK cells.
We used a recently developed anti-PD-L1 antibody, N-601, to perform similar in vitro ADCC killing assays with healthy donor NK cells as effector cells and chordoma cell lines as target cells. N-601 significantly increased NK cell lysis of 6 of 6 chordoma cell lines compared with isotype control antibody ( Fig. 2A). The average increase in NK cell-mediated tumor cell lysis by N-601 in >20 independent experiments was 6.2-fold (P ≤ 0.0001) compared with isotype control antibody.
Given that N-601 is a structural homolog of avelumab and maintains the human Fc element, we hypothesized ADCC to be the primary mechanism through which N-601 enhanced NK cell-mediated killing of chordoma cells. Blocking CD16 on NK cells with a neutralizing antibody before coculture with chordoma cells entirely abrogated the NK-cell killing of tumor cells (P ≤ 0.0001; Fig. 2A and B), indicating that N-601 enhanced NK-cell lysis of chordoma cells through an ADCC-dependent mechanism.

IL15/IL15r Superagonist (N-803) Enhances NK-Cell Killing of Chordoma Cells, and Combinatorial Treatment with Anti-PD-L1 Antibody (N-601) and N-803 Further Enhances Cytotoxicity
N-803 is a humanized IL15 superagonist that stimulates and expands NK cells and T cells (9,10). While this immunostimulatory complex has demonstrated antitumor efficacy as a monotherapy and in combination against multiple cancer types in clinical studies (10, 15) NCT03054909, NCT02099539, NCT01946789, NCT03853317, NCT03022825, NCT03127098), it has yet to be explored in chordoma. NK cells treated with N-803 were significantly more cytolytic against chordoma cell lines compared with untreated NK cells (Fig. 2B).
The average increase in tumor cell lysis by N-803 treatment of NK cells in >20 independent experiments was 13.6-fold (P ≤ 0.0001) compared with untreated controls. We

Doublet Treatment of Chordoma Cells with Anti-PD-L1 (N-601) and Anti-EGFR (Cetuximab) Antibodies Further Enhances Sensitivity to Lysis by IL15 Superagonist (N-803)-activated NK Cells
Flow cytometric analysis of 6 chordoma cell lines revealed higher EGFR expression than PD-L1 expression (Table 1; Supplementary Fig. S1A). Furthermore, PD-L1 + cells also tended to be EGFR + , as evidenced by the robust cytotoxicity of cetuximab-mediated ADCC (Fig. 3). However, we identified small subpopulations of cells that appeared to be PD-L1 + /EGFR − (Supplementary Fig. S1A), thus theoretically posing additional antitumor activity by combining therapeutic antibodies targeting both surface antigens.
To model this doublet ADCC-promoting antibody treatment strategy with N-803-enhanced effector cells ( Supplementary Fig. S1B), two chordoma cell lines were treated with N-601 and cetuximab before being targeted by N-803activated NK cells. This combination approach proved to be marginally more effective than the next best therapeutic combination of cetuximab and N-803 (UM-Chor1 P = 0.003; JHC7 P = 0.03; Fig. 4).

Chordoma Patients' NK Cells Mediate Significant Lysis of Chordoma Cells and are Enhanced with Anti-PD-L1 Antibody (N-601), Anti-EGFR Antibody (Cetuximab), and IL15/IL15r Superagonist (N-803)
Patients with cancer have alterations in immune cell function (40), thus we next sought to determine whether NK cells from patients with chordoma are responsive to ADCC-mediating antibodies and N-803 exposure. We isolated NK cells from three patients with chordoma. Similar to NK cells from healthy donors, NK cells from 3 of 3 patients with chordoma demonstrated enhanced cytotoxicity after exposure to N-601 (Fig. 5A), cetuximab (Fig. 5B), N-803 ( Fig. 5A and B), and combination therapy (P ≤ 0.05; Fig. 5A and B). This finding indicates that NK cells in patients with chordoma are not inherently impaired in their ability to lyse chordoma cells and be enhanced by ADCC and IL15 superagonism.
Interestingly, NK cells from 2 of 3 patients with chordoma exhibited significantly increased cytotoxic activity at baseline compared with healthy donor controls, and this persisted after treatment with N-601, cetuximab, N-803, and combination therapy (P ≤ 0.0001). These patients, labeled CP(io), had recurrent chordoma and had undergone various treatment modalities (refer to Materials and Methods) prior to providing blood for this study (Fig. 5A and B; Supplementary Fig. S2). The patient with NK-cell activity similar to that of healthy donors, labeled CP1(n), had a primary chordoma and had not been treated prior to providing blood for the study (Fig. 5A and B).

PD-L1 t-haNK Cells Induce Lysis of Chordoma Cells
As a potential alternative clinical option for PD-L1-positive tumors, we interrogated the sensitivity of chordoma cell lines to lysis by human NK cells that have been genetically engineered to express the high-affinity CD16 receptor, endoplasmic reticulum-retained IL2, and a CAR-specific against PD-L1 (41, 42). The in vitro antitumor efficacy of these PD-L1-targeted high-affinity natural killer (PD-L1 t-haNK) cells against human cancer cell lines is dependent on PD-L1 expression levels on tumor cells (20). Here, we hypothesized that treatment of chordoma cell lines with IFNγ would upregulate PD-L1 expression, thus increasing their sensitivity to lysis by PD-L1 t-haNK cells. Our study showed that IFNγ increased PD-L1 expression (both % positive cells and MFI) on 2 of 2 chordoma cell lines (Fig. 6A) and in turn, enhanced the susceptibility of these cell lines to lysis by PD-L1 t-haNK cells (P ≤ 0.0001; Fig. 6B).
Despite containing the high-affinity CD16 receptor, it has been shown that tumor cell lysis by PD-L1 t-haNK cells is not enhanced by therapeutic antibodies that induce ADCC, such as avelumab or cetuximab (20). Previous interrogation of this finding revealed that PD-L1 t-haNK killing mediated through the CAR occurs more readily than killing through ADCC (20). Similarly, treatment of PD-L1 t-haNK cells with N-803 does not enhance their cytotoxic function, as this engineered NK-cell line does not express the IL15 receptor (20). Here, we confirm that PD-L1 t-haNK lysis of chordoma cell lines is not enhanced by cotreatment with N-601, cetuximab, or N-803 ( Supplementary Fig. S3).

Anti-PD-L1 Antibody (N-601)-mediated ADCC and IL15/IL15r Superagonist (N-803) Activation of NK Cells Enhance Lysis of Chordoma CSCs
CSCs are a subpopulation of cells within certain tumor types that play a significant role in tumorigenesis, metastasis, and resistance to therapy (18,19,43) and should therefore be a primary target of anticancer therapy. We have previously reported that exposure of chordoma cells to IFNγ increases PD-L1 expression on the cell surface, augmenting anti-PD-L1 antibody-mediated ADCC (8).
Here, we investigated the susceptibility of chordoma CSCs by N-803-activated NK cells and anti-PD-L1-driven ADCC. Flow cytometry was utilized to distinguish the CSCs from the non-CSCs and to determine the degree of cell death in the two compartments using a viability exclusion dye. As defined in previous studies (8,22), we designated the UM-Chor1 CSCs as cells that coexpress the surface markers CD15, CD24, and CD133 (Fig. 7A). Similar to our data in Figs

Discussion
Chordoma is a devastating malignancy with resistance to traditional anticancer treatment modalities such as radiation and chemotherapy, thus necessitating the interrogation of novel treatment approaches. Prior foundational work has characterized the expression of PD-L1 (31)(32)(33)34) and EGFR (35,36,37) in chordoma tumor specimens, thus setting the stage for novel targeted immunotherapeutic approaches to be interrogated. Numerous mAbs targeting PD-L1 and EGFR have been generated and approved to treat cancer. However, only one such antibody for each target molecule (PD-L1, avelumab; EGFR, cetuximab) retains the native human Fc region, enabling it to inhibit the desired signaling pathway and promote ADCC with NK cells. N-601 is a novel anti-PD-L1 structural analog of avelumab that also maintains the native human Fc region for ADCC coordination. Here, we capitalize on this unique and dynamic antibody feature and demonstrate that both N-601 and cetuximab enhance chordoma cell sensitivity to lysis by NK cells through an ADCC-dependent mechanism in 6 of 6 chordoma cell lines. (Figs. 2A and 3A). with N-803 activation of NK cells (Fig. 4). This is the first chordoma study demonstrating the efficacy of combinatorial immunotherapy with ADCCmediating antibodies and an IL15 superagonist complex in an NK cell-based treatment paradigm. These data provide the preclinical rationale for the clinical investigation of immunotherapeutic combinations with NK-cell effectors in the treatment of chordoma, and also support future studies exploring the translation of such combinations to mediate T-cell activation against chordoma.
The failure of conventional cancer therapies, such as chemotherapy and radiation, in the chordoma setting may be partially due to CSCs (45,46). While non-CSCs are susceptible to anticancer therapeutics, treatment-resistant CSCs can self-renew and maintain tumor growth. Harnessing NK-cell activity is a promising anti-CSC approach because NK cells are not affected by the downmodulation of antigen processing and presentation machinery molecules observed in CSCs (47). In addition, studies have shown that CSCs from various cancer types are susceptible to IL15-activated NK cells and this targeting was associated with increased expression of NK ligands on the surface of CSCs (48,49). Here, we show that lysis of UM-Chor1 CSCs and non-CSCs by NK cells is improved in the presence of N-601 (Fig. 7C). When NK cells were activated by N-803, lysis of both tumor cell populations was further enhanced. This finding may be partially attributed to the MICA and MICB (a ligand for NKactivating receptor NKG2D) expression in both CSCs and non-CSCs, and to the expression of B7-H6 (ligand for the NK-activating receptor NKp30) in the CSC subpopulation (Fig. 7B). Combinatory treatment with N-601 and N-803 was effective at further enhancing lysis of the CSC population, which may be attributed to upregulation of PD-L1 in the CSCs compared with non-CSCs ( Fig. 7B). Although CSCs make up only a minority subset within chordoma cell lines (8,22), our study illustrates a preferential killing of CSCs compared with non-CSC in the described combinatory treatment strategy. A more comprehensive interrogation of chordoma CSC and non-CSC NK ligand expression is needed to better understand the driving mechanisms behind the observed differential target lysis. Moreover, differential NK-cell killing of the two cellular subpopulations may be further elucidated by sorting for CSCs and non-CSCs and coincubating equal numbers of cells with NK cells at increasing E:T ratios (20). This immunotherapy approach may be applicable to other tumor types with identifiable CSC populations and warrants further investigation.
Cancer induces a systemic immunosuppressive environment through altering the function of cellular and noncellular immunity (40). To support the translation of this proposed immunotherapy framework into the clinical setting, we included NK cells from three patients with chordoma and compared their cytotoxic function to healthy donor NK cells both at baseline and in the presence of ADCC-mediating antibodies and N-803. Chordoma patients' NK cells were reflective of healthy donor NK cells throughout this study in their positive response to treatment with N-601, cetuximab, and/or N-803 ( Fig. 5A-C). This result, though limited in sample size, suggests that chordomas do not impair NK-cell cytotoxic function or response to therapeutic enhancement with ADCC-mediating antibodies or IL15 superagonism. Identification of two patients with chordoma with enhanced NK-cell cytotoxicity ( Fig. 5B and C) prompted interrogation of these patients' clinical histories. These two patients had undergone surgical resections, multiple rounds of radiation, and participated in various previous immunotherapy protocols (refer to Materials and Methods) prior to providing blood for this study, while the patient with nonenhanced NK-cell activity had not received any prior therapy before providing study blood. Interestingly, neither of the patients with enhanced NK-cell function had received any therapy in at least 3 months before providing blood for this study. These results may suggest that prior therapy in patients with chordoma, likely one of the immunomodulating agents listed previously, can impart a sustained impact on the cytotoxic activity of NK cells without compromising their ability to respond to treatment with ADCC-mediating antibodies and N-803. Such sustained NK-cell responses to anticancer therapeutics have yet to be defined in the medical literature and the mechanism remains unknown. Conclusions drawn from this interesting finding are limited by sample size and lack of patient NK cells before and after treatment with immunotherapy, but the replicated response in two separate patients with chordoma warrants further investigation in a controlled clinical trial setting.
While this study focuses on ADCC-mediating antibodies and NK-cell antitumor activity, the agents investigated here are uniquely dynamic due to their ability to also stimulate T-cell responses. First, T cells comprise a much larger percentage of the lymphocyte population than do NK cells. N-601 is an immune checkpoint inhibitor, which has the established primary function  Limitations to this study include the in vitro nature of our preclinical models, which is related to poor NK-cell engraftment in PBMC humanized triple knockout NSG mice as well as other barriers to experimentation with murine chordoma xenograft models. As previously discussed, mechanistic drivers of the enhanced cytotoxic activity of NK cells from patients with recurrent chordoma with prior treatments remains unknown and should be investigated further in a controlled clinical trial setting. Also previously mentioned, the dynamic ability of this combinatory model to stimulate T-cell responses through immune checkpoint blockade and IL15 superagonism are not explored here. Future preclinical and clinical studies using more comprehensive biologic models will be critical in evaluating the antitumor efficacy of the above proposed treatment framework.