Cisplatin Alters Antitumor Immunity and Synergizes with PD-1/PD-L1 Inhibition in Head and Neck Squamous Cell Carcinoma

Head and neck squamous cell carcinoma (HNSCC), the sixth most common cancer type worldwide, is an aggressive malignancy with little improvement in overall survival during the past few decades (1). Clinical trials have focused on improving survival for patients with HNSCC related to carcinogen exposure, who have a relatively poor prognosis, and reducing treatment-related toxicities in patients with human papillomavirus (HPV)-associated HSNCC, which portends a relatively favorable prognosis (2). Cisplatin chemotherapy has been the standard systemic therapy for HNSCC for decades, but it offers modest increases in survival and substantial toxicity when used at traditional high doses (3, 4). Antibodies targeting the programmed death 1 (PD-1) immune checkpoint pathway have recently been FDA approved based on clinical trials showing sustained responses in patients with recurrent or metastatic HNSCC that were refractory to multiple prior treatments (5, 6). However, not all HNSCC patients respond to anti–PD-1 monotherapy. Combinations of immunotherapy with standard therapies are, thus, an attractive option for head and neck cancer patients.

The combination of anti–PD-1 immunotherapy with cisplatin chemotherapy for HNSCC is currently under exploration in clinical trials. However, the impact of cisplatin alone or in combination with checkpoint inhibitors on antitumor immunity has not been explored in preclinical models of HNSCC. Prior preclinical and clinical studies in other cancer types suggest that cisplatin and other cytotoxic chemotherapy drugs may enhance antitumor immunity by enhancing immunogenic cell death (ICD), increasing expression of MHC class I and other components of the antigen-processing machinery (APM), enhancing numbers and function of immune effector cells, and decreasing numbers of immunosuppressive cells (7–10). However, other preclinical studies suggest that cytotoxic chemotherapy can upregulate tumor cell expression of PD-L1 (11, 12), potentially impairing antitumor immunity by inducing T-cell anergy.

In the current study, we used multiple human and mouse HNSCC cell lines and antigen-specific, engineered mouse T cells to explore the effects of cisplatin on APM, PD-L1 expression, and T-cell function. A syngeneic mouse model of oral cancer was used to mechanistically investigate the effects of cisplatin and PD-L1/PD-1 blockade on tumor growth in vivo. We show that in these preclinical models of HNSCC, cisplatin can have both immune-enhancing and immunosuppressive effects, which may be dose dependent and partially reversible by combination with PD-L1/PD-1 blockade.

UMSCC-1, UMSCC-46, and UMSCC-74A cell lines were obtained from Dr. T. Carey at the University of Michigan and were authenticated by PCR and maintained as previously described (13). UPCI SCC90 cells were obtained from Dr. Robert Ferris and Dr. Susanne Gollin at the University of Pittsburgh and were validated there by PCR, exome sequencing, and fluorescence in situ hybridization as described (14, 15). Human cell lines were maintained in MEM or DMEM with 1% penicillin/streptomycin, 1% glutamine, and 10% FBS. Mouse oral cancer (MOC) 1 and 2 cell lines were obtained from Dr. R. Uppaluri in 2014 and were authenticated by exome sequencing and maintained as previously described (16, 17). All cell lines were stored in liquid nitrogen and cultured for no longer than 6 months or 20 passages before use. All cell lines were tested monthly for mycoplasma.

Recombinant human interferon gamma (IFNγ; catalog no. 570206) and antibodies to human HLA-ABC (catalog no. 311404) and PD-L1 (catalog no. 329706) were obtained from Biolegend. The antibody for intracellular calreticulin was obtained from R&D Systems (catalog no. IC3898R), and the antibodies to TAP1, TAP2, LMP2, and ERp57 were obtained from Abcam (catalog numbers ab83817, ab180611, ab190350, and ab13506, respectively). Pharmaceutical grade cisplatin was obtained from the veterinary pharmacy at NIH. Antibodies for in vivo mouse treatments specific for PD-L1 (clone 10F.9G2), PD-1 (clone RMP1-14), NK1.1 (clone PK136), and CD8 (clone YTS 169.4) were from BioXCell. Fluorescent-conjugated flow cytometry antibodies for mouse tumor experiments were obtained from eBioscience (CD137/41BB, catalog no. 46-1371-80) or Biolegend: CD8 (catalog no. 100712), CD45.2 (catalog no. 109806), CD80 (catalog no. 104721), CD11b (catalog no. 101211), CD11c (catalog no. 117307), CD107a (catalog no. 121619), Ly6G (catalog no. 127623), Ly6C (catalog no. 128017), and H-2Kb/H-2Db (catalog no. 114611).

Cells were plated at 5 × 10⁴ to 1 × 10⁵ cells per well in 6-well plates and allowed to adhere overnight prior to treatment with IFNγ (10 ng/mL) or cisplatin (see doses in Supplementary Table S1). PE Annexin V/7AAD Apoptosis Detection Kit (BD Biosciences) and Cell Proliferation Kit I/XTT (Sigma) cell death assays were used according to the manufacturer's instructions.

Cell lines were harvested with TrypLE Select (Thermofisher) prior to staining for flow cytometry. For intracellular staining, cells were fixed in 2% paraformaldehyde (PFA) and permeabilized with 100% methanol overnight at −20°C. Data were acquired using a BD FACS Canto I cytometer and BD FACS Diva software, and then analyzed using FlowJo software. Cell viability was verified using 7AAD staining. Median fluorescence intensity values for isotype controls were subtracted for each condition, and “fluorescence minus one” controls were tested for each multicolor flow panel.

Wild-type, female C57BL/6 mice at 6 to 8 weeks were obtained from Charles River Laboratories. Mice were injected in the right flank with MOC1 (5 × 10⁶, in matrigel) or MOC2 (1 × 10⁵) cells and allowed to grow for 11 to 14 days, and then randomized into treatment groups. Mice were then treated with cisplatin (cis-diamminedichloroplatinum(II): CDDP; 3 or 5 mg/kg/week) and/or antibodies to PD-1 or PD-L1 (200 μg, twice/week) or vehicle (saline/PBS) by intraperitoneal injection. Cisplatin-treated mice received additional saline supplementation (0.5 mL daily × 2 days following cisplatin by subcutaneous injection). The cisplatin dose chosen was the maximum tolerated weekly dose that did not cause excessive illness or weight loss in preliminary experiments, as we were unable to find pharmacokinetic data in the literature comparing murine doses to clinically relevant doses. Antibodies blocking PD-1 and PD-L1 were used at 200 mcg twice weekly by intraperitoneal injection as established in the literature (18–21). Because previous experiments in our laboratories have shown no effect of isotype controls for anti–PD-1 and anti–PD-L1 antibodies (rat IgG2a and IgG2b, respectively) on tumor growth, control animals were not treated with isotype control antibodies in these experiments.

CD8⁺ T cells and NK cells were depleted in a subset of mice by intraperitoneal injection of depleting antibodies (clones listed above) at 200 mcg twice weekly as previously described and validated (22).

In a subset of mice from each experiment, tumors were harvested prior to completion of treatment (day 20 or day 25 as indicated), digested into single-cell suspensions with mechanical dissociation followed by enzymatic dissociation using Miltenyi's mouse tumor dissociation kit and GentleMACS Dissociator as previously described (22), then analyzed by flow cytometry for tumor cell surface markers and immune cell phenotypes. Mouse hearing tests (auditory brainstem response thresholds and distortion product otoacoustic emission amplitudes) were performed as previously described (23, 24), and kidney histology (H&E staining of paraffin-embedded tissue) was performed using standard methods by Histoserv (Germantown, Maryland). All animal procedures were approved by the NIDCD Animal Care and Use Committee.

To generate SIINFEKL-specific cytotoxic T lymphocytes (CTL), splenocytes from OT-1 mice were cultured in the presence of SIINFEKL (2 μg/mL) with daily 2:1 splitting. For IFNγ capture assays, OT-1 splenocytes were cultured with cisplatin at indicated doses for 24 hours, then stimulated with SIINFEKL peptide. CD8⁺Vα2⁺ T cells were then labeled for IFNγ capture using a Mouse IFN-γ Secretion Assay Detection Kit according to the manufacturer's instructions (Miltenyi) and analyzed by flow cytometry.

For CTL killing assays, 1 × 10⁴ target cells (ovalbumin-expressing MOC1 cells) were plated in a 96-well E-Plate (ACEA Biosciences) and cultured with or without cisplatin at indicated doses for 24 hours prior to adding OT-1 CTL effector cells. Alteration of impedance was acquired using the xCELLigence Real-Time Cell Analysis (RTCA) platform per the manufacturer's instructions. Triton X-100 (0.2%) was added to some wells to verify complete loss of cell index with total cell lysis, and CTLs alone were plated up to 1 × 10⁶/well to verify that they do not contribute to gain of impedance. Percent loss of cell index for a given time point was calculated as: 1 − (experimental cell index/control cell index).

Data from in vitro experiments were analyzed by one- or two-way ANOVA where appropriate. Tumor growth curves were compared by linear regression. Animal survival curves were made using the Kaplan–Meier method with comparison by using log-rank (Mantel–Cox) testing. GraphPad Prism software was used for statistical testing. P < 0.05 was used to determine statistical significance.

To investigate whether cisplatin might enhance antigen presentation and adaptive immunity by mechanisms other than release of antigen from dying cells, several APM components were assessed in HNSCC cell lines treated with cisplatin. First, doses needed to kill 25% (LD₂₅) or 50% (LD₅₀) of cells in 72 hours were determined for each cell line (Supplementary Table S1). Treatment with these doses of cisplatin or IFNγ (positive control) alone or in combination increased cell surface MHC class I/HLA-ABC (Fig. 1A). Intracellular calreticulin, ERp57, and LMP2 also increased with cisplatin, with additive increases in ERp57 and LMP2 noted with combination cisplatin and IFNγ in UMSCC-74A cells (Fig. 1B–D). Cisplatin-induced increases in transporter of antigen proteins (TAP1 and TAP2) were noted only in the UMSCC-74A cells (Fig. 1E and F), which retain expression of wild-type TP53 (25), an established inducer of TAP1 (26).

Although cisplatin-induced increases in APM component expression may enhance antitumor immunity, we hypothesized that it may also impair antitumor immunity in other ways, such as upregulation of immune checkpoints or impairment of T-cell function. HNSCC cell lines treated with the LD₂₅ or LD₅₀ of cisplatin (or IFNγ as a positive control) for 48 hours showed increased surface expression of PD-L1 (Fig. 2A). To investigate T-cell function, antigen-specific IFNγ production by OT-1 T cells was measured after stimulation by pulsing with SIINFEKL peptide in the presence or absence of cisplatin, with doses chosen based on the plasma concentrations of cisplatin (2–6 μg/mL) following a standard-of-care dose of cisplatin (100 mg/m²) in the clinical setting (27). Although the lower doses of cisplatin did not drastically impair IFNγ production, the doses reflecting peak serum concentrations (4–6 μg/mL) significantly reduced IFNγ production by stimulated T cells (Fig. 2B). Tumor cell killing by CTLs was investigated using an antigen-specific, impedance-based assay with MOC1 cells expressing ovalbumin and OT-1 T cells. Low doses of cisplatin or T cells when used alone did not induce tumor cell killing. However, when used together, sublethal doses of cisplatin (determined by prior XTT assays) enhanced tumor cell killing by antigen-specific CTLs (Fig. 2C–E). The T cells in these experiments were cocultured with cisplatin along with the tumor cells, but they were not added to the treatment wells until 24 hours after the tumor cells. To determine whether the cytotoxic T-cell function might be dampened if T cells were treated with cisplatin at the same time as tumor cells, we repeated the experiments but pretreated the T cells with cisplatin prior to adding to the impedance plates with the tumor cells. The results were similar (Supplementary Fig. S1). These results suggest that low doses of cisplatin may upregulate PD-L1 on tumor cells but have only modest effects on T-cell IFNγ production and may enhance antigen-specific T-cell killing of tumor cells.

Taken together, the immune-enhancing effects and increased PD-L1 noted with cisplatin in vitro provide a rationale for combined use of cisplatin and PD-L1/PD-1 blockade in vivo. We explored this possibility in mice bearing MOC1 tumors, which typically induce immune responses and respond to immunotherapies, whereas MOC2 tumors, which are aggressive, do not typically induce immune responses and do not respond well to immunotherapies (16, 17, 22, 28). In MOC1-bearing mice, cisplatin (5 mg/kg/week) and anti–PD-L1 did not delay tumor growth when used alone but provided a significant tumor growth delay and prolonged survival when used concurrently (Fig. 3). The tumor growth delay was less pronounced when anti–PD-L1 was combined with a lower dose of cisplatin (3 mg/kg/week; Supplementary Fig. S2). Flow cytometry of tumors from a subset of animals showed that cisplatin did increase tumor cell expression of PD-L1 but did not alter the numbers of CD8⁺ T cells or other intratumoral immune stimulatory or suppressor cells previously defined in this model (Fig. 4). In contrast, treatment of mice bearing nonimmunogenic MOC2 cells with cisplatin induced only a slight and short-lived tumor growth delay with no improvement in survival, and anti–PD-L1 with or without cisplatin did not significantly delay tumor growth (Supplementary Fig. S3), supporting the idea that an underlying antitumor immune response is required for efficacy of cisplatin and anti–PD-L1.

Although combination cisplatin and anti–PD-L1 provided a delay in tumor growth, only three anti–PD-L1 doses could be used to avoid toxicity, and the tumors eventually resumed growing. To see if long-term treatment would be more effective, we treated MOC1-bearing mice with cisplatin and/or an antibody to PD-1, which was better tolerated for up to several weeks in preliminary experiments. A delay in tumor growth when cisplatin and anti–PD-1 were used concurrently was similar to that seen with cisplatin and anti–PD-L1 (Fig. 5A and B), consistent with the limited expression of PD-L2 seen in MOC1 tumors (29). To determine whether the sequence of drug administration would affect tumor growth, additional groups were treated with cisplatin followed by anti–PD-1 one day later, or vice versa, which was not as effective as concurrent use (Fig. 5C and D). Tumors eventually resumed growth toward the end of the treatment period, suggesting that either the anti–PD-1 of heterologous rat origin may be inactivated after a certain number of doses, or the tumors develop resistance mechanisms to cisplatin and/or anti–PD-1 over time. In contrast to tumors analyzed at four days after the second dose of cisplatin, tumors from these experiments analyzed at two days following cisplatin did not show increased PD-L1 (Fig. 6A). As before, the numbers of CD8⁺ T cells and other intratumoral immune cells did not decrease with cisplatin (Fig. 6). The overall numbers of intratumoral CD8⁺ T cells and CD11b⁺ CD11c⁺ cells (presumed to be dendritic cells) were not decreased in animals treated with cisplatin with or without anti–PD-1 versus controls, but the trend toward increased CD8⁺ and CD11b⁺ CD11c⁺ cells seen with anti–PD-1 alone was lost when cisplatin was added (Fig. 6D and G). Fewer CD8⁺ T cells were positive for CD107a (Fig. 6E), a marker of functional activation, suggesting some degree of impaired antitumor immunity when cisplatin was added to anti–PD-1.

To elucidate the mechanisms by which cisplatin and PD-L1/PD-1 blockade induce a tumor growth delay in MOC1 tumors in vivo, MOC1-bearing mice were treated with concurrent cisplatin and anti–PD-L1 or anti–PD-1, in addition to antibodies to deplete CD8⁺ T cells or NK cells. The delay in tumor growth seen with cisplatin and anti–PD-L1 was not significantly altered by depleting CD8⁺ T cells or NK cells (P = 0.13 for cisplatin plus anti–PD-L1 alone vs. with CD8 depletion; Fig. 7A). When the experiment was repeated using anti–PD-1 instead of PD-L1 antibody, the difference between growth of tumors treated with combination therapy with and without CD8⁺ T-cell depletion was minimal and, again, not statistically significant (P = 0.28 for cisplatin + anti–PD-1 alone vs. with CD8 depletion; Fig. 7B), but remained significantly different from the control group. Taken together, these results suggest that the antitumor efficacy of cisplatin and PD-L1/PD-1 blockade in this preclinical model depends on mechanisms independent of CD8⁺ T cells and NK cells.

Cisplatin is known to induce clinically significant toxicity to the inner ear and kidney in susceptible individuals, and this has been reproduced in preclinical animal models (24, 30). Given that cisplatin-induced ototoxicity and nephrotoxicity may involve inflammatory cytokines (30), the use of immunotherapy could theoretically worsen these toxicities seen frequently with cisplatin. To investigate this possibility, kidney histology following 6 weeks of treatment in MOC1-bearing mice was compared across treatment groups, and none of the kidneys exhibited histologic signs of cisplatin toxicity (Supplementary Fig. S4A). Hearing tests performed on mice treated with 6 weekly doses of cisplatin showed no evidence of hearing impairment compared with untreated mice of the same age (Supplementary Fig. S4B and S4C).

An expanding series of studies suggest that cytotoxic chemotherapy may be immunogenic. It has been suggested that cisplatin may enhance host immunity in several tumor types by upregulating MHC class I and other components of the antigen processing machinery by inducing components of ICD and by increasing the ratio of immune effector to immunosuppressive cells in the tumor microenvironment (7, 8, 31). However, cisplatin is known to reduce numbers of circulating immune cells, suggesting it can be directly toxic to immune cells. Tumor cells treated with cytotoxic drugs may upregulate immune checkpoint ligands as a form of immune escape (11, 12). In the current study, we used preclinical models of head and neck cancer to show that cisplatin may enhance antitumor adaptive immunity by upregulating antigen processing machinery and enhancing tumor cell killing by CD8⁺ T cells without decreasing T-cell numbers. However, consistent with findings in other tumor types treated with cytotoxic drugs, we found that exposure to cisplatin caused tumor cells to upregulate surface levels of PD-L1. This PD-L1 increase occurred in vitro after 48 hours but was only seen in vivo after 4 days; PD-L1 was not increased in tumors from cisplatin-treated animals after two days.

Our data suggest that the balance of immune-enhancing and immunosuppressive effects of cisplatin in HNSCC may be dose dependent. The increase in several APM components examined was seen at sublethal doses in vitro, and these doses are all below the peak serum concentration seen in patients treated with cisplatin at a high dose of 100 mg/m² (27). The APM changes included increases in both HLA-ABC/MHC class I and multiple chaperone proteins, which are critical for antigen processing and presentation to T cells. There were additive increases in APM components in one cell line treated with both IFNγ and cisplatin, suggesting that cisplatin may enhance IFNγ-induced antigen processing in some situations. Increases in TAP1/2 were only noted in the cell line with wild-type TP53, but not in HPV-positive cells or HPV-negative cells with mutated TP53. Dependence of TAP1 expression on functional p53 has been reported (26). Because the majority of tumors harbor TP53 mutations or HPV oncoproteins, which inactivate p53 (32), this finding suggests that the inability to increase TAP1/2 may be a form of immune escape in HNSCC. Cisplatin treatment of HNSCC cells also activates STAT1 (33), which may also be needed for maximized expression of APM components in addition to p53 (26, 34–36). Thus, differences in p53 and STAT1 expression may, in part, explain the variability in APM component expression following cisplatin treatment in different HNSCC tumors or cell lines.

Although increases in PD-L1 were also noted at relatively low doses, substantial impairment of IFNγ production by T cells was seen at higher doses, which are achieved only briefly in the plasma of patients treated with the highest dose used for HNSCC (100 mg/m²; refs. 3, 4, 27). In our mouse experiments, we noted that the maximum tolerated dose of cisplatin (5 mg/kg/week, based on prior experiments) provided a more pronounced tumor growth delay when paired with anti–PD-L1 than a lower dose (3 mg/kg/week). Although the higher dose did not reduce numbers of CD8⁺ T cells compared with untreated animals, tumors from animals treated with cisplatin and anti–PD-1 did not show robust increases in CD8⁺ T cells that we have previously seen animals treated with anti–PD-1 alone. A similar effect on CD107⁺ T cells was seen with combined cisplatin and anti–PD-L1/PD-1, suggesting that T-cell degranulation may be impaired as well. Taken together, these data suggest that there may be an optimal dose of cisplatin that enhances antigen presentation and T-cell activation without inducing untoward effects on T-cell numbers or function. However, our in vivo depletion studies suggest that the effects of the combination are to some degree dependent on mechanisms other than CD8⁺ T cells in this chemically induced HNSCC model. Another study in a syngeneic mouse model of HPV-related HNSCC found a delay in tumor growth with cisplatin that was not seen in Rag-1 mice lacking functional T and B cells (37), suggesting that T cells may be important for the maximal response to cisplatin chemotherapy in that model.

It is possible that cisplatin may enhance antitumor immunity in other ways, such as by inducing ICD or by reducing the number or function of immunosuppressive cells in the tumor microenvironment. Although cisplatin is thought to be a poor inducer of ICD compared with oxaliplatin and other chemotherapy drugs (38), data from our laboratory suggest that cisplatin induces all of the classic markers of ICD in preclinical models of HSNCC (39). A prior study with murine and human data showed that chemotherapy (carboplatin and paclitaxel) reduces numbers of immunosuppressive MDSCs in cervical cancer (40). Although the MOC1 model is largely driven by MDSCs (28), we did not see any differences in MDSC numbers in tumors from cisplatin-treated mice in our study, though we did not look at multiple time points or perform functional studies.

A study published while our experiments were under way explored immune effects of cisplatin in combination with a different checkpoint inhibitor, anti–CTLA-4, in HPV-transformed lung and embryonic cell models that are far more sensitive to cisplatin monotherapy (31). The authors found that the tumor growth delay induced by cisplatin alone was largely attenuated by depleting CD8⁺ T cells, a relationship that was not clearly seen in our study. The authors also noted that the effects of cisplatin were highly dependent on activation of dendritic cells, with cisplatin inducing increases in CD70, CD80, and CD86 on myeloid cells; cisplatin was also less effective in animals lacking these costimulatory molecules (31). In our study, we saw a trend toward increased CD80 on dendritic cells with cisplatin, but this did not reach statistical significance. The authors were able to clear tumors in mice when cisplatin was combined with blockade of CTLA-4. The effects of PD-1/PD-L1 blockade were not investigated in that study (31). CTLA-4 has been noted to play a greater role in blocking development of new T-cell responses, whereas PD-L1 promotes exhaustion of established responses (41, 42). Together, these studies suggest that combinations of cisplatin with PD-L1/PD-1 and CTLA-4 immune checkpoint inhibitors merit further investigation.

The findings in our study are timely and pertinent, due to the FDA approval of anti–PD-1 antibodies and paradigm shifts in dosing of cisplatin in the treatment of HNSCC. The traditional regimen for treatment of both nonsurgical and surgical HNSCC patients was to administer cisplatin doses of 100 mg/m² every 3 weeks for a total of 3 doses with concurrent radiation, based on large clinical trials showing that cisplatin given in this fashion improves survival (3, 4). However, providers have begun to administer weekly doses of 40 mg/m² for a total of 6 to 7 doses, despite the lack of large, prospective clinical trials comparing the efficacy of this regimen to the more traditional regimen (43, 44). Retrospective trials suggest that the low-dose, weekly regimen may be equivalent and better tolerated, and many centers around the world have shifted to this regimen (43, 44). It is unclear which of these cisplatin dosing strategies might be more immunogenic. Our in vivo data suggest that concurrent administration of these agents may be more effective than administering cisplatin prior to checkpoint inhibitors or vice versa. Several clinical trials combining cisplatin with checkpoint inhibitors for HNSCC are under way. Correlative immune studies and toxicity data from these trials could help elucidate the best way to combine these agents in order to maximize antitumor immunity without excess toxicity.

This study has several limitations, including our inability to detect subtle changes in numbers and phenotype of immune cells of the tumor microenvironment by flow cytometry of tumors from a relatively small number of animals. It is also unclear how the doses of cisplatin and anti–PD-1/PD-L1 that we used compare with doses used clinically, as we were unable to find good pharmacokinetic data for cisplatin-treated mice. However, we did perform preliminary experiments to establish a dose of cisplatin that is relatively well tolerated when given weekly over several weeks as is done in HNSCC patients, and we used doses of anti–PD-1/PD-L1 that have been established to dampen PD-1 responses. We also attempted to determine whether known toxicities of cisplatin to the inner ear and kidney would be worsened by PD-1 blockade, which could theoretically exacerbate the inflammatory changes seen in these toxicities. Our cisplatin treatment paradigm, however, did not induce any hearing loss or detectable changes in kidney histology. Experiments are currently under way in a mouse model known to cause cisplatin-induced hearing loss (24), in order to determine whether the inner ear toxicity may be exacerbated by adding twice-weekly anti–PD-1 antibody.

In summary, cisplatin chemotherapy given at an optimal dose may enhance adaptive antitumor immunity with minimal effects on the numbers or function of effector immune cells in the tumor microenvironment or cisplatin-related toxicities. Cisplatin may also cause upregulation of tumor cell PD-L1, which may in part explain the additive antitumor activity of cisplatin and PD-L1/PD-1 blockade. This information will help in interpreting results of ongoing clinical trials combining these agents, and for comparison of potential future therapeutic combinations for the treatment of head and neck cancer.

N.C. Schmitt reports receiving a commercial research grant from Astex Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Tran, C.T. Allen, C. Van Waes, N.C. Schmitt

Development of methodology: L. Tran, C.T. Allen, R. Xiao, N.C. Schmitt

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Tran, C.T. Allen, R. Xiao, E. Moore, R. Davis, K. Spielbauer, N.C. Schmitt

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Tran, C.T. Allen, R. Xiao, E. Moore, R. Davis, K. Spielbauer, N.C. Schmitt

Writing, review, and/or revision of the manuscript: L. Tran, R. Xiao, C. Van Waes, N.C. Schmitt

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Tran, S.-J. Park, N.C. Schmitt

Study supervision: L. Tran, N.C. Schmitt

Supported (in part) by the Intramural Research Program of the NIH, NIDCD, projects ZIA-DC-DC000087, ZIA-DC-DC000090, and ZIC DC000080.

The authors would like to acknowledge Christopher Silvin for flow cytometry assistance; Megan Morisada for technical assistance with figures; and Jeffrey Schlom, James Hodge, and Zhong Chen for feedback on the project and manuscript.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.