Enhanced killing of chordoma cells by antibody-dependent cell-mediated cytotoxicity employing the novel anti-PD-L1 antibody avelumab

Chordoma, a rare bone tumor derived from the notochord, has been shown to be resistant to conventional therapies. Checkpoint inhibition has shown great promise in immune-mediated therapy of diverse cancers. The anti-PD-L1 mAb avelumab is unique among checkpoint inhibitors in that it is a fully human IgG1 capable of mediating antibody-dependent cell-mediated cytotoxicity (ADCC) of PD-L1-expressing tumor cells. Here, we investigated avelumab as a potential therapy for chordoma. We examined 4 chordoma cell lines, first for expression of PD-L1, and in vitro for ADCC killing using NK cells and avelumab. PD-L1 expression was markedly upregulated by IFN-γ in all 4 chordoma cell lines, which significantly increased sensitivity to ADCC. Brachyury is a transcription factor that is uniformly expressed in chordoma. Clinical trials are ongoing in which chordoma patients are treated with brachyury-specific vaccines. Co-incubating chordoma cells with brachyury-specific CD8+ T cells resulted in significant upregulation of PD-L1 on the tumor cells, mediated by the CD8+ T cells' IFN-γ production, and increased sensitivity of chordoma cells to avelumab-mediated ADCC. Residential cancer stem cell subpopulations of chordoma cells were also killed by avelumab-mediated ADCC to the same degree as non-cancer stem cell populations. These findings suggest that as a monotherapy for chordoma, avelumab may enable endogenous NK cells, while in combination with T-cell immunotherapy, such as a vaccine, avelumab may enhance NK-cell killing of chordoma cells via ADCC.


IntroductIon
Chordoma is a rare bone cancer thought to arise from remnants of the embryonic notochord. Approximately 300 new cases per year are diagnosed in the United States, accounting for 20% of primary spine tumors and 1%-4% of all malignant bone tumors [1,2]. Reported 5-and 10-year survival rates are about 70% and 40%, respectively, a reflection of the slow-growing nature of the disease. Surgery followed by radiotherapy is the standard of care for primary tumors. However, based on anatomic location and tumor size on presentation, a wide curative excision is rarely feasible [2]. Hence, incidence of disease recurrence is common and metastases have been reported in up to 40% of cases. Once metastases develop, median survival is about 1 year [1]. No treatment for advanced chordoma has been approved by the U.S. Food and Drug Administration (FDA), since chordoma is largely resistant to conventional chemotherapy [3]. Thus, there is an urgent need for novel therapeutic modalities for this disease.
Immunotherapy has become an important treatment option for chemotherapy-resistant cancers. Brachyury, a transcription factor uniformly expressed in chordoma [4], appears to be an oncogenic driver for this tumor type [5,6]. In addition to being a diagnostic marker for chordoma, brachyury may be a potential target for treatment [7,8]. Clinical trials of brachyury-specific vaccines are ongoing in chordoma patients. In addition, drugs that inhibit the immune checkpoints programmed cell death protein 1 (PD-1) and its major ligand, programmed death-ligand 1 (PD-L1), have shown clinical activity in diverse cancer types [9][10][11]. The FDA first approved a PD-1 inhibitor for melanoma; more recently, a PD-1 inhibitor was approved for lung cancer. Numerous clinical trials of several drugs targeting the PD-1/PD-L1 axis are ongoing in a range of cancers. Most of these PD-L1 antibodies are the IgG4 isotype and the Fc-modified IgG1 isotype, both of which inhibit the interaction of PD-1 on immune cells with PD-L1 on tumor cells [12]. Avelumab, a fully human IgG1 anti-PD-L1 monoclonal antibody (mAb), is the only anti-PD-L1 mAb that both induces antibody-dependent cellmediated cytotoxicity (ADCC) and blocks the PD-1/PD-L1 pathway. Previously, our group reported that avelumab enhanced ADCC on several cancer cell lines expressing PD-L1 [13]. Other studies have shown that PD-L1 is expressed in chordoma cell lines and chordoma tissue samples [14,15].
Here, for the first time, we demonstrate the potential of anti-PD-L1 antibody therapy for chordoma and report that (a) PD-L1 expression induced by IFN-γ increased the sensitivity of chordoma cells to lysis by natural killer (NK) cells via avelumab-mediated ADCC; (b) tumor antigenspecific CD8 + T cells indirectly induced PD-L1 expression on chordoma cells; (c) upregulated PD-L1 expression on chordoma cells indirectly induced by brachyury-specific CD8 + T cells increased the sensitivity of chordoma cells to avelumab-mediatedADCC; and (d) residential cancer stem cell (CSC) populations in chordoma cells were killed by avelumab-mediated ADCC to the same degree as non-CSC populations within the cells. Our findings suggest that while chordoma responds poorly to conventional therapies such as surgery, radiotherapy, and chemotherapy, immune-mediated therapy may have clinical benefit for some chordoma patients.

Expression profiles of IFN-γ-induced genes in uM-chor1 cells
To further examine the molecular consequences of treating chordoma cells with IFN-γ, we assessed IFN-γ-induced gene expression profiles of UM-Chor1 cells by microarray analysis (Supplemental Figure 1A). IFN-γ treatment upregulated genes in UM-Chor1 cells > 1.5-fold relative to untreated controls (P < 0.05). The highest upregulation was seen in gene TP53INP2 (tumor protein p53 inducible nuclear protein 2), which regulates transcription and enhances starvation-induced autophagy [22]. The second highest upregulation was seen in gene CEBPD (CCAAT/enhancer binding protein [C/EBP] δ), which regulates proinflammatory gene expression [23,24]. IFN-γ treatment downregulated some genes in UM-Chor1 cells > 1.5-fold relative to untreated controls (P < 0.05; (Supplemental Figure 1B). The most downregulated gene, CLDN2, has been identified as a tight junction-specific integral membrane protein [25] whose expression is affected by cytokines [26]. The second most downregulated gene, PHACTR4I, is a tumor suppressor gene that is mutated or downregulated in several cancers [27]. Supplemental Figure 1C shows the predicted pathway of IFN-γ-induced PD-L1 expression, as deduced from the results of microarray analysis. The transcription factor CEBPD is induced by IFN-γ, leading to inhibition of MYC and activation of TLR9, IL10, and TNF, and culminating in upregulation of PD-L1 (CD274) expression. Taken together, these results suggest that CEBPD is potentially involved in the pathway of IFN-γinduced PD-L1 expression in chordoma cells.

Phenotypic signature of a residential CSC population in chordoma cell lines
CSCs have been recognized in recent years as important players in the development of solid tumors. Cells with CSC characteristics are resistant to current treatment modalities including radiation and chemotherapy and are associated with poor treatment response rates and disease recurrence. Certain established tumor cell lines have been reported to harbor residential CSC populations. Previous studies have defined the CSC population in a single chordoma cell line, U-CH1, as expressing CD15 and CD133 [31]. We investigated the relative expression levels (mean fluorescence intensity; MFI) of CD24, CD133, CD15, and ALDH in the CD24 high /CD133 high group in 4 chordoma cell lines ( Figure 4A). Both CD15 and ALDH were markedly increased in the CD24 high /CD133 high group, defined as the residential CSC population, compared to the non-CSC population ( Figure 4A). As an example, the non-CSC population in JHC7 cells were CD24 (8633), CD133 (41), CD15 (233), and ALDH (145). These MFIs were markedly less than the MFIs observed in the CSC population. We observed similar MFIs for the non-CSC populations in the other chordoma cell lines. A residential CSC population was detectable in 4 of 4 cell lines, ranging from 6%-18% of the total population, as determined by CD24 and CD133 co-expression ( Figure 4B). These data suggest that chordoma cells have a CSC subpopulation that can be identified by the stem cell markers CD24, CD133, CD15, and ALDH.

Treating chordoma cells with IFN-γ increases NKcell killing of both CSC and non-CSC populations via Adcc
To determine whether avelumab-mediated ADCC could increase CSC subpopulation killing, we stained UM-Chor1 cells with the CSC markers CD24 and CD133 and treated them with or without IFN-γ. Treatment with IFN-γ did not change the frequency of the CSC subpopulation ( Figure 5A, 5B). Flow cytometric staining analyses of these cells showed a 5-fold increase in PD-L1 expression following IFN-γ treatment ( Figure 5B, inset). UM-Chor1 cells treated with IFN-γ were then subjected to an ADCC assay with avelumab, and the degree of cell death was determined by viability stain exclusion ( Figure 5C). In this assay, we used untreated UM-Chor1 cells as a baseline for comparison to IFN-γ-treated UM-Chor1 cells that underwent ADCC. In the non-CSC group, ADCCmediated cell death was 1.7-fold higher ( Figure 5C; P < 0.001) than baseline cell death. Notably, ADCC-mediated cell death in the CSC group also had a significant increase (1.7-fold; P < 0.001) compared to cell death in the baseline CSC group. These data suggest that IFN-γ increases PD-L1 expression in the CSC subpopulation of chordoma cells, and that avelumab effectively increases ADCC of both the non-CSC and CSC subpopulations to the same degree.

dIscussIon
Immunotherapy has become a standard treatment for patients with certain cancers. The PD-1 inhibitors nivolumab and pembrolizumab are FDA-approved for melanoma and lung cancer. Currently, many clinical trials of agents that block the PD-1/PD-L1 pathway are ongoing in a range of cancers. The finding of a significant correlation between PD-L1 expression levels in tumor tissue and responsiveness to PD-1 pathway blockade [10,32] has led to the investigation of PD-L1 expression in a variety of cancers, with the goal of developing further applications of PD-1 pathway blockade [33][34][35][36]. Feng et al. reported that PD-L1 is expressed in chordoma tissue samples, especially metastatic tumors [14]. It has also been reported that PD-L1 expression can be upregulated by IFN-γ in several cancers [12,37]. Previous studies showed that IFN-γ could induce PD-L1 expression in the chordoma cell lines U-CH1, U-CH2, CH22, and JHC7 [14,15]. Here, we confirmed previous reports that PD-  Oncotarget 33506 www.impactjournals.com/oncotarget L1 expression was upregulated by IFN-γ in U-CH2 and JHC7 (Figure 1) [14,15]. We extended these observations to include the chordoma cell lines UM-Chor1 and MUG-Chor1, and confirmed that they also expressed PD-L1, which was upregulated by IFN-γ (Figures 1, 3, and 5).
The finding of PD-L1 expression in chordoma ( Figure 1) suggested potential clinical benefit from PD-1/ PD-L1 pathway blockade. Several anti-PD-L1 antibodies have been generated and are being evaluated in ongoing phase II/III clinical trials. Of these agents, avelumab appears to be the only one that induces ADCC in NK cells. Previously, our group reported that avelumab enhanced ADCC in various cancer cell lines that express PD-L1 [13]. However, the potential of anti-PD-L1 antibody therapy for ADCC in chordoma has not previously been shown. It should be noted that chordoma cells expressed a relatively high baseline expression of PD-L1 which could be further increased with IFN-γ ( Figure 1). It has been reported that the overall cell-surface surface density of PD-L1 (as determined by, MFI) rather than the percentage of positive cells may be greater predictor or sensitivity to ADCC. Boyerinas et al., conducted flow cytometric analysis of a panel of 18 human tumor cell lines encompassing five different tumor types and showed that human carcinoma cell lines express a broad range of PD-L1 % positive cells and PD-L1 cell surface densities. Moreover, those cell lines with the highest cell surface expression were the most sensitive to ADCC mediated by avelumab [13].
Here, we show that avelumab significantly increased NK-cell lysis via ADCC in 4 of 4 chordoma cell lines ( Figure 2). Moreover, avelumab's efficacy is further enhanced in chordoma cells where treatment with IFN-γ has induced overexpression of PD-L1 ( Figure 2). It has also been reported that IFN-γ produced by activated CD8 + T cells can upregulate PD-L1 expression on tumor cells [38,39]. A potential strategy for activating tumorrecognizing T cells is a vaccine encoding a tumor-specific antigen. Since chordoma expresses the transcription factor brachuyury [30], a rational strategy for treating chordoma may be a vaccine that encodes brachyury.
Historically, transcription factors have been considered undruggable targets [40] for standard cytotoxic agents, but other potential means of targeting brachyury have been proposed, including small inhibitory RNA, epigenetic modulation, and immune-based therapy [41,42]. To date, the only attempts to target brachyury have been with 2 therapeutic cancer vaccines. Our group recently completed a phase I trial of GI-6301 (recombinant yeast-brachyury vaccine) that enrolled 11 patients with advanced chordoma [13]. An ongoing phase II clinical trial in chordoma patients is evaluating a brachyury vaccine in combination with radiation therapy. One patient had a confirmed radiographic partial response by RECIST, another had a mixed response, and the median progressionfree survival in the chordoma group was 8.3 months. The vaccine induced brachyury-specific T-cell responses in the majority of all patients enrolled and in the subset of patients enrolled with chordoma. There were no significant vaccine-related toxicities [43]. Notably, the 2 patients with evidence of tumor shrinkage both had radiotherapy within 3.5 months of enrolling on study. Given the low likelihood of radiographic response with radiotherapy alone in advanced-stage disease, our hypothesis was that radiotherapy had an immunomodulatory effect on chordoma cells, making them more amenable to T cellmediated killing.
To model a patient receiving a brachyury vaccine ( Figure 3A), we co-incubated chordoma cells with brachyury-specific CD8 + T cells. The brachyury-specific CD8 + T cells recognized chordoma cells and increased PD-L1 expression on chordoma cells through the production of IFN-γ ( Figure 3B and 3C). This increased PD-L1 expression significantly increased the chordoma cells' sensitivity to avelumab-mediated ADCC of NK cells ( Figure 3D). In a previous study we observed that radiation did not modulate PD-L1 expression on tumor cell lines [44]. Similarly, chordoma cell lines (JHC7, UM-Chor1, U-CH2, and MUG-Chor1) exposed to 8 Gy radiation showed no increase in PD-L1 expression after 72 h (Supplemental Table 1). Radiation therapy has been reported to upregulate PD-L1 on tumors in-vivo, likely indirectly from the radiation induced inflammatory response. Future studies will focus on how to exploit PD-L1 modulation in response to radiation. These observations suggest a rationale for treating additional cohorts of the ongoing phase II trial (NCT02383498) with avelumab.
It has been shown that MHC-I expression is increased by IFN-γ and upregulated in cancer tissue [16,17]. Our study showed that IFN-γ also increased expression of HLA-ABC in chordoma cells ( Figure  1C). While some have suggested that increased MHC-I expression may induce resistance to NK-cell lysis, as NK cells discriminate between self and non-self by monitoring the expression of MHC-I molecules [45], increased expression of MHC-I has actually been shown to enhance sensitivity to cytotoxic T lymphocytes (CTLs) by upregulating antigen processing and presentation on tumor cells [17]. Our results indicate that chordoma cells treated with vaccine may have increased sensitivity not only to ADCC, but also to CTLs.
We assessed the molecular consequences of IFN-γinduced gene expression by microarray analysis, focusing on UM-Chor1, which showed the greatest increase of IFN-γ-induced PD-L1 expression among the 4 chordoma cell lines ( Figure 1C). The gene with the second highest upregulation, CEBPD, is a transcription factor that modulates many biological processes, including cell differentiation, motility, growth arrest, proliferation, and cell death. Though it was reported that CEBPD was induced by IFN-γ in certain cancers [24,46], CEBPD functions both as a tumor suppressor and a tumor www.impactjournals.com/oncotarget promoter [47]. Here we showed the predicted pathway of PD-L1 expression induced by IFN-γ, deduced from the results of microarray analysis. CEBPD is induced by IFN-γ, which leads to inhibition of MYC and activation of TLR9, IL10, and TNF, and culminates in upregulation of PD-L1 (CD274) expression (Supplemental Figure 1C). This suggests that CEBPD is potentially involved in the pathway of IFN-γ-induced PD-L1 expression in chordoma cells. This would block the innate immune response and result in tumor progression. Further investigation is needed to confirm the function of CEBPD in chordoma cells.
Recently, CSCs have been recognized as critical to the development of solid tumors. CSCs have been reported to be largely resistant to treatments such as radiation and chemotherapy [48]. There have been concerted efforts to determine the markers that define CSCs in several cancers. It is noteworthy that the expression of CSC surface markers is tissue type-specific. For example, the CSC surface markers CD44 + and CD24were defined in breast cancer, CD20 + and ABCB5 + in melanoma, and EpCAM + , CD44 + , and CD166 + in colon cancer [49]. In an effort to define the phenotypic signature of chordoma CSCs, we observed that CD15 and ALDH were also highly expressed in CD24 high /CD133 high ( Figure 4A). ALDH has been widely used as a CSC marker in various cancer types [50]. Thus, we found that CD24 high /CD133 high cells could be defined as a CSC subpopulation in chordoma cells. Moreover, CD24 high /CD133 high cells showed high expression of CD15. These data confirm and extend previous findings that CD15 and CD133 were candidates for CSC markers in chordoma, using a U-CH1 cell line [31]. Our data suggest that, as the stem markers, CD24, CD133, CD15, and ALDH could identify a residential CSC subpopulation in chordoma. However, the potential of immunotherapy for CSCs remains controversial [51]. Previous studies showed that breast CSCs had resistance to NK killing due to reduced expression of the ligands for NKG2D, the stimulatory NK cell receptor [52]. On the other hand, Tallerico et al. reported that colorectal CSCs express higher levels of ligand for the natural cytotoxicity receptors that mediate NK-cell killing [53]. Thus, we investigated the possibility of treating chordoma CSCs with immunotherapy. IFN-γ treatment upregulated PD-L1 expression in the CSC population of chordoma ( Figure 5A, 5B), suggesting PD-L1 blockade as a potential treatment for chordoma CSCs. Furthermore, the chordoma CSCs were killed by avelumab-mediated ADCC to the same degree as the non-CSCs ( Figure 5C).
Our study is the first to show that PD-L1 expression induced by IFN-γ increases chordoma cells' sensitivity to NK-cell lysis via avelumab-mediated ADCC. Moreover, in a model of a patient receiving a tumor antigen-specific vaccine, brachyury-specific CD8 + T cells increased PD-L1 expression on chordoma cells through the production of IFN-γ, increasing the sensitivity of chordoma cells to avelumab-mediated ADCC. We also identified the residential CSC population in chordoma cells and showed that they were killed by avelumab-mediated ADCC. Our findings indicate the potential of avelumab to enable endogenous NK cells to kill chordoma cells via ADCC, as well as the potential of combination therapy, such as a T-cell vaccine and avelumab, to enhance NK-cell killing of chordoma cells via ADCC. Our findings suggest that while chordoma is resistant to conventional therapies such as radiotherapy and chemotherapy, immune-mediated therapy may have clinical benefit for patients with chordoma.
The anti-PD-L1 mAb avelumab and matching IgG1 isotype control were obtained from EMD Serono as part of a Cooperative Research and Development Agreement with the Laboratory of Tumor Immunology and Biology, National Cancer Institute.

Flow cytometry
To assess the effect of IFN-γ on the cell-surface Oncotarget 33508 www.impactjournals.com/oncotarget dye (Thermo Fisher, Waltham, MA). Cells were incubated with the antibodies for 30 min at 4°C, acquired on a FACSCalibur flow cytometer or FACSVerse (Becton Dickinson, Franklin Lakes, NJ), and analyzed using FlowJo software (TreeStar, Inc., Ashland, OR). Isotype control staining was < 5% for all samples analyzed.

Microarray analysis and statistical analysis
UM-Chor1 cells were left untreated or treated with 50 ng/mL of IFN-γ for 24 h. Cells were then harvested and total RNA was isolated using the RNAeasy Plus minikit (Qiagen, Valencia, CA). 100 ng of RNA was reverse transcribed and amplified using a WT expression kit (Ambion, Austin, TX). Sense strand cDNA was fragmented and labeled using a WT terminal labeling kit (Affymetrix, Santa Clara, CA). Three replicates of each group were hybridized to the Human Gene ST 2.0 GeneChip (Affymetrix) and scanned on the GeneChip scanner 3000 (Affymetrix). Data were collected using Affymetrix AGCC software.
Microarray data are available in the GEO under accession number GSE77732. Statistical and clustering analysis for the microarray experiment was performed with Partek Genomics Suite software (St. Louis, MO) using an RMA normalization algorithm. Differentially expressed genes were identified by ANOVA. Genes that were up-or downregulated > 1.5-fold with a P < 0.05 were considered significant. Significant genes were analyzed for enrichment of pathways using Ingenuity Pathway Analysis software (Qiagen, Redwood City, CA).

Antibody-dependent cellular cytotoxicity assay
The ADCC assay was performed as previously reported [13] with indicated modifications. Cells were left untreated or treated with 50 ng/mL of IFN-γ for 24 h. Cells were then harvested and labeled with 111 In. Cells were plated as targets at 2,000 cells/well in 96-well roundbottom culture plates and incubated with 2 μg/mL of avelumab or control isotype antibody at room temperature for 30 min. NK cells were added at 100,000 cells/well at an effector-to-target (E:T) ratio of 50:1. After 4 h, supernatants were harvested and analyzed for the presence of 111 In using a WIZARD2 Automatic Gamma Counter (PerkinElmer, Waltham, MA). Spontaneous release was determined by incubating target cells without effector cells, and complete lysis was determined by incubation with 0.05% Triton X-100. Experiments were carried out in triplicate. Specific ADCC lysis was determined using the following equation: Percent lysis = [(experimental cpmspontaneous cpm) / (complete cpm -spontaneous cpm)] x 100.
To verify that CD16 (FcγRIII) on NK cells engage avelumab-mediated ADCC, CD16 mAb was used to block CD16. NK cells were incubated with 2 μg/mL of CD16 mAb (clone B73.1; eBioscience, San Diego, CA) for 2 h before being added to target cells.
To examine the relationship between a CSC subpopulation and ADCC activity, UM-Chor1 cells were left untreated or treated with 50 ng/mL of IFN-γ for 24 h. Cells were then plated as targets at 50,000 cells/well in 6-well round-bottom culture plates and incubated with 2 μg/mL of avelumab at room temperature for 30 min. NK cells were added at 2500,000 cells/well at an E:T ratio of 50:1. After 4 h, tumor cells were harvested and stained with antibodies for flow cytometry.
UM-Chor1 cells were co-cultured in 12-well plates with Tp2A or normal donor CD8 + T cells at a tumor cell/ CD8 + T cell ratio of 2:1. As a positive or negative control, UM-Chor1 cells were left untreated or treated with 50 ng/mL of IFN-γ for 24 h. After a 24-h co-culture, the supernatant fluid was harvested and the concentration of IFN-γ was measured using a multiplex cytokine/ chemokine kit (Meso Scale Discovery, Gaithersburg, MD). Tumor cells were harvested and used as a target for the ADCC assay, as described above, and stained with PD-L1 antibody for flow cytometry.

Western blot analysis
The Western blot was performed as previously described [55] with indicated modifications. Protein lysate was extracted from UM-Chor1 cells. The primary antibodies used were monoclonal rabbit antibody (mAb 54-1, 1 μg/mL) against human brachyury [29] and GAPDH (Cell Signaling Technology, Danvers, MA).

statistical analysis
Significant differences in the distribution of data acquired by flow cytometry analysis were determined by the Kolmogorov-Smirnov test using FlowJo software (TreeStar, Inc.). Significant differences in the distribution of data acquired by ADCC assays were determined by paired Student's t test with a 2-tailed distribution and reported as P values, using Prism 6.0f software (GraphPad Software Inc., La Jolla, CA).