PD-L1-specific helper T-cells exhibit effective antitumor responses: new strategy of cancer immunotherapy targeting PD-L1 in head and neck squamous cell carcinoma

Background Head and neck squamous cell carcinoma (HNSCC) originates from squamous epithelium of the upper aerodigestive tract and is the most common malignancy in the head and neck region. Among HNSCCs, oropharynx squamous cell carcinoma (OSCC) has a unique profile and is associated with human papillomavirus infection. Recently, anti-programmed cell death-1 monoclonal antibody has yielded good clinical responses in recurrent and/or metastatic HNSCC patients. Therefore, programmed death-ligand 1 (PD-L1) may be a favorable target molecule for cancer immunotherapy. Although PD-L1-expressing malignant cells could be targeted by PD-L1-specific CD8+ cytotoxic T lymphocytes, it remains unclear whether CD4+ helper T lymphocytes (HTLs) recognize and kill tumor cells in a PD-L1-specific manner. Methods The expression levels of PD-L1 and HLA-DR were evaluated using immunohistochemical analyses. MHC class II-binding peptides for PD-L1 were designed based on computer algorithm analyses and added into in vitro culture of HTLs with antigen-presenting cells to evaluate their stimulatory activity. Results We found that seven of 24 cases of OSCC showed positive for both PD-L1 and HLA-DR and that PD-L1241-265 peptide efficiently activates HTLs, which showed not only cytokine production but also cytotoxicity against tumor cells in a PD-L1-dependent manner. Also, an adoptive transfer of the PD-L1-specific HTLs significantly inhibited growth of PD-L1-expressing human tumor cell lines in an immunodeficient mouse model. Importantly, T cell responses specific for the PD-L1241-265 peptide were detected in the HNSCC patients. Conclusions The cancer immunotherapy targeting PD-L1 as a helper T-cell antigen would be a rational strategy for HNSCC patients.


Background
Head and neck squamous cell carcinoma (HNSCC) originates from squamous epithelium of the upper aerodigestive tract, which includes the nasal and oral cavity, pharynx, and larynx, and is the most common malignancy in the head and neck region with over 600,000 new cases diagnosed each year [1,2]. Although smoking and alcohol consumption are major risk factors for the development of most HNSCCs, oropharynx squamous cell carcinoma (OSCC) has a unique profile and is associated with human papillomavirus (HPV) infection [3,4]. Interestingly, patients with HPV-positive oropharyngeal cancer had better 3-year overall survival (OS) and progression-free survival (PFS) rates than those with HPV-negative cancer after treatment with fractionated radiotherapy [5].
Cancer immunotherapy with immune checkpoint inhibitors has been the focus of many studies since the efficacy of immunotherapy targeting the immune checkpoint molecule programmed cell death-1 (PD-1) and its ligand PD-L1 was demonstrated [6][7][8][9][10][11]. PD-L1 plays an important role in immune regulation by binding to PD-1 expressed on effector T-cells to induce apoptosis or anergy in order to prevent autoimmune disease [12,13]. Furthermore, tumor cells also take advantage of PD-L1 to escape from antitumor immune responses. Indeed, high PD-L1 expression is frequently found in tumor tissues and correlates with poor prognosis [14][15][16][17].
Therefore, blockade of the PD-1/PD-L1 signaling pathway by using specific antibodies to PD-1, such as nivolumab, yielded remarkable clinical responses in metastatic melanoma [9], non-small cell lung cell cancer [18], and renal cell carcinoma [19]. The efficacy of immunotherapy, particularly blockade of the PD-1/PD-L1 pathway, in HNSCC patients was recently demonstrated [20], although HNSCC was initially recognized as an immunosuppressive tumor from the perspective of lower lymphocyte count, spontaneous apoptosis of cytotoxic T lymphocytes (CTLs), and poor antigen-presenting function in patient blood samples [21]. Furthermore, 6-month OS and PFS rates of recurrent and/or metastatic HNSCC patients treated with pembrolizumab, an anti-PD-1 monoclonal antibody, were 23% and 59%, respectively, showing a favorable response similar to single-drug cetuximab [22,23].
Based on this evidence, PD-1/PD-L1 signaling plays a critical role in suppressing immune responses against HNSCC as well, suggesting that immunotherapy targeting PD-L1-expressing HNSCC cells by acquired immunity would be a rational antitumor strategy. Indeed, PD-L1 is a favorable target molecule for cancer immunotherapy and PD-L1-expressing malignant cells were killed by PD-L1-specific CD8 + CTLs in a PD-L1-dependent manner [24,25]. However, there are no reports about PD-L1-specific CD4 + helper T lymphocytes (HTLs).
In cancer immunotherapy, HTLs not only support CTLs by promoting effector functions and long-term survival but also have direct cytotoxicity against cancer cells via effector cytokines [26]. Thus, we hypothesized that PD-L1-specific HTLs are also required for enhancing effective antitumor immunotherapy.
In the current study, we defined the helper epitope peptide in PD-L1 for inducing PD-L1-specific HTLs from peripheral blood of healthy donors for the first time. PD-L1-specific HTLs produced effector cytokines and demonstrated cytotoxicity against PD-L1-expressing tumor cells. Remarkably, PD-L1-specific HTLs adoptively transferred into immunodeficient mice significantly inhibited growth of PD-L1-positive human lung carcinoma. Also, specific T-cells to the peptide were observed in the HNSCC patients. These findings suggest that PD-L1 could be a promising antitumor target and immunotherapy using PD-L1-specific HTLs would be a rational approach for patients with HNSCC.

Flow cytometry
HLA-DR and PD-L1 expression on tumor cell surfaces were evaluated by flow cytometry using anti-HLA-DR monoclonal antibody (mAb) (G46-6) conjugated with fluorescein isothiocyanate (BD Pharmingen) and anti-PD-L1 mAb (MIH) conjugated with phycoerythrin (eBioscience, Thermo Fisher Scientific). Mouse IgG2a antibody (MOPC-173) and mouse IgG1 antibody (MOPC-21) were purchased from BioLegend and used as isotype controls for HLA-DR and PD-L1, respectively. All tumor cell lines were treated with or without 500 IU/ ml interferon gamma (IFN-γ) for 48 h before analysis. The samples were analyzed using the BD Accuri C6 flow cytometer and software (BD Biosciences).

Western blotting
Tumor cell lines (1 × 10 6 ) were washed in phosphatebuffered saline (PBS) and lysed in LDS sample buffer (Invitrogen, Thermo Fisher Scientific). The cell lysate was subjected to electrophoresis in a 4-12% NuPAGE Bis-Tris SDS-PAGE gel (Invitrogen, Thermo Fisher Scientific) under reducing condition and transferred to an Immobilon-P membrane (Merck Millipore). The membrane was blocked in PBS containing 0.01% Tween 20 and 5% non-fat dry milk for 1 h at room temperature and incubated with polyclonal rabbit anti-human PD-L1 (E1L3N, Cell Signaling Technology) diluted 1:1000 in blocking buffer for overnight at 4 °C, or anti-β-actin mAb (C4, Santa Cruz Biotechnology) diluted 1:2000 in blocking buffer as the control for 1 h at room temperature. After washing, the membrane was incubated with horseradish peroxidase-labeled sheep anti-rabbit or anti-mouse IgG and visualized using the Amersham ECL Prime Western Blotting Detection System (GE Healthcare Life Sciences).

Knockdown of PD-L1 using siRNA
Tumor cell lines Sa-3, HSC-4, SAS, and Lu65 were transfected with PD-L1 siRNA using Lipofectamine RNAi MAX Reagent (Invitrogen, Thermo Fisher Scientific). We used a mixture of three types of siRNA, which were (5′ to 3′) GAG GAA GAC CUG AAG GUU CAG CAU A, CCU ACU GGC AUU UGC UGA ACG CAU U, and UGA UAC ACA UUU GGA GGA GAC GUA A. Stealth siRNA duplex for targeting PD-L1 and recommended Stealth siRNA negative control duplex for medium GC content (catalog number: 12935112) were purchased from Invitrogen. Transfected tumor cells were used for assay after 96-h incubation of transfection.

In vitro generation of PD-L1-specific CD4 + HTLs
The procedure for the induction of tumor-specific CD4 + HTLs has been described in detail previously [29]. Briefly, monocytes and CD4 + T-cells were purified from peripheral blood mononuclear cells (PBMCs) using MACS microbeads for CD14 and CD4, respectively (Miltenyi Biotech). To prevent the antibodies from binding non-specifically, we used the FcR blocking reagent (Miltenyi Biotech). Monocytes were differentiated into dendritic cells (DCs) using granulocyte macrophage colony-stimulating factor (GM-CSF) (50 ng/ml) and interleukin (IL)-4 (1000 IU/ml). PD-L1 peptide-pulsed DCs (3 μg/ml for 3 h at room temperature) were co-cultured with autologous CD4 + T-cells in 96-well flat-bottomed culture plates. Seven days later, the CD4 + T-cells were restimulated in individual microcultures with PD-L1 peptide-pulsed γ-irradiated autologous PBMCs (3 μg/ ml), and 2 days later, recombinant human IL-2 (10 IU/ml) was added. After the two cycles of peptide stimulation, PD-L1-specific T-cell lines were expanded by weekly restimulation with cognate peptides (3 μg/ml)-pulsed irradiated autologous PBMCs. Production levels of IFN-γ (BD Pharmingen) and Granzyme B (MABTECH) in culture supernatants were determined by ELISA kits according to the manufactures' instructions. We measured the absorption at 450 nm by GloMax Discover Microplate Reader (Promega). Three percent of human male AB serum (Innovative Research)-supplemented AIM-V medium (Invitrogen, Carlsbad, CA) was used as complete culture medium for all experiments. All blood materials were acquired after informed consent was appropriately obtained.

H&E staining and immunohistochemistry
Formalin-fixed tissue sections were subjected to H&E staining according to a standard protocol. Immunohistochemistry was performed using the Envision ™ HRP System (Agilent Technologies Dako) as described previously [31]. Formalin-fixed, paraffin-embedded tumor tissue sections were acquired from oropharynx cancer patients. Ventana PD-L1 (SP263) rabbit mAb (1:1 dilution) (Roche) or Dako HLA-DR Antigen, Alpha-Chain (TAL.1B5) mouse mAb (1:80 dilution) (Agilent) were used as primary antibody against PD-L1 and HLA-DR, respectively. The institutional ethics committee approved this study, and written informed consent was obtained from all patients who provided tissue samples. Distribution of HLA-DR staining was graded by percentage of tumor cells that were positive and then divided into quartiles as: 0-9%; negative, 10-25%; weak, 26-45%; moderate, and 51-100%; strong as previously reported [32].

Statistical analysis
Values shown are the means of triplicate determinations (Figs. 2, 3, 4, 5) or six mice (Fig. 6). All data were analyzed by Student's t-test, one-way ANOVA with the Holm post hoc test, or unpaired t test. P values < 0.05 were considered statistically significant.

Various expression levels of PD-L1 and HLA-DR in oropharynx squamous cell carcinoma
Immunohistochemical analyses were performed to assess PD-L1 and HLA-DR expression in oropharynx cancer tissues from 24 patients treated at Otolaryngology, Head and Neck Surgery, Asahikawa Medical University. Each specimen was blindly checked by three pathologists and classified into four types according to immunostaining intensity for PD-L1 or HLA-DR as follows: negative, weak, moderate, or strong staining. PD-L1 expression in specimens with 5% tumor cell membranous staining and HLA-DR expression in specimens with 10% tumor cell membranous staining were considered "positive", as previously reported [33][34][35]. As shown in Fig. 1a For HLA-DR immunostaining, 16 (66.7%), 2 (8.3%), 3 (12.5%), and 3 (12.5%) cases of 24 cases were negatively, weakly, moderately, and strongly stained with anti-HLA-DR antibody, respectively (Fig. 1e-h). Seven of 24 cases (29.2%) were positive for both PD-L1 and HLA-DR. Data are summarized in Table 1. There were not any differences between the expression patterns and staging of the patients (data not shown). H&E staining was also performed as shown in Fig. 1.   were detected. Each PD-L1 241-265 -specific CD4 + T-cell line released IFN-γ in a dose-dependent manner (Fig. 2a).
To define their HLA-DR restriction, we evaluated the reactivity of PD-L1 241-265 -specific CD4 + T-cells to autologous PBMCs in the presence of PD-L1 241-265 peptide by using anti-HLA-DR or anti-HLA class I mAbs. The IFN-γ production of both PD-L1 241-265 -specific CD4 + T-cell lines were inhibited by anti-HLA-DR mAbs, but not by anti-HLA class I mAbs, suggesting that peptide recognition of both PD-L1 241-265 -specific CD4 + T-cell lines was restricted to HLA-DR (Fig. 2b). Furthermore, we assessed the reactivity of PD-L1 241-265 -specific CD4 + T-cell lines using mouse fibroblasts (L-cells) transfected with HLA-DR allele gene as APCs. The T-cell line G1 responded to L-DR4 cells and the T-cell line G2 responded to L-DR9 cells, indicating that these T-cell lines G1 and G2 were restricted to HLA-DR4 and HLA-DR9, respectively (Fig. 2c).

Recognition of PD-L1 peptides by PBMCs from HNSCC patients
It is important to confirm whether the PD-L1 241-265 peptide also shows antigenic activity in patients with HNSCC for clinical applications because HNSCC has been reported to dysregulate immune cells [21]. Thus, we performed a short-term culture of PBMC derived from 5 patients with HNSCC and 2 healthy donors in the presence of the PD-L1 241-265 peptide because the volumes of blood obtained from these patients were small. As shown in Fig. 4d, substantial T-cell responses to PD-L1 241-265 peptides were observed not only in healthy donors but also in HNSCC patients (4/5 tested). This means that the precursor of PD-L1 241-265 -specific CD4 + T-cells surely exists in HNSCC patients. tumor cell lines HSC-4 and Lu65, but not against HLA-DR-unmatched tumor cell line Sa-3, pretreatment of tumor cell lines with IFN-γ was required for upregulating both HLA-DR and PD-L1 (Fig. 5a). Also, PD-L1 241-265 -specific CD4 + T-cell line G1 produced granzyme B against HLA-DR-matched tumor cell lines pretreated with IFN-γ (HSC-4 and Lu65), but not against Sa-3 (Fig. 5b).

Discussion
In the current study, we newly identified a PD-L1-derived helper epitope peptide (PD-L1 241-265 ) and demonstrated the potential use of PD-L1 as a tumorassociated antigen (TAA). PD-L1 241-265 efficiently stimulated and expanded PD-L1-specific HTLs from peripheral blood of healthy donors and patients with HNSCC. PD-L1 241-265 -specific HTLs showed not only cytokine production but also cytotoxicity against tumor cells in a PD-L1-specific manner. Surprisingly, adoptively transfer of PD-L1 241-265 -specific HTLs into immunodeficient mice significantly inhibited growth of PD-L1-expressing human lung carcinoma. Because 7 of 24 cases (29.2%) of OSCC were positive for both PD-L1 and HLA-DR expressions, PD-L1-targeted immunotherapy using PD-L1 241-265 peptide would be applicable to almost 30% of OSCC patients. PD-L1 is expressed in many malignancies including HNSCCs, in which PD-L1 expression ranges from 46 to 91% [34,37,38]. Although it has been reported that there was a strong correlation between PD-L1 expression and OS in HNSCC [39], our immunohistological analyses showed no correlation between them. This discrepancy may be caused by our small sample size (24 cases).
OSCC is distinguished from other HNSCCs in terms of etiological cause, which includes smoking, alcohol consumption, and especially HPV infection. Indeed, in a worldwide systematic review, the percentage of HPV-positive tumors was significantly higher (35.6%) in OSCC than in oral and laryngeal SCCs (23.5% and 24.0%, respectively) [40]. Moreover, OS of patient with OSCC was significantly longer in HPV-positive cases than in HPV-negative cases in contrast to patients with other HNSCCs (oral cavity, hypopharyngeal, and laryngeal cancer) [41]. In our immunohistological analyses, we also found significantly higher PD-L1 expression levels in the p16-positive group compared with the p16-negative group [12/15 (80 [35]. However, the relation between PD-L1 expression and HPV status remains unclear because some reports have shown no such association [34,37,38]. Therefore, this association must be further elucidated. Among HPVpositive OSCC, PD-L1-positive cases also showed significantly longer PFS than PD-L1-negative cases [37]. This finding suggests that while PD-L1 expression on tumor cells is generally considered to indicate poor prognosis, HPV and PD-L1 status might be good markers for prognosis in patients with OSCC. Although PD-L1 is regarded as a functional molecule that sends a suppressive signal to effector T-cells, it could also be a target antigen for acquired immunity. For example, Minami et al. showed by using PD-L1 [11][12][13][14][15][16][17][18][19] and PD-L1 [41][42][43][44][45][46][47][48][49][50] peptides that PD-L1-specific CTLs were induced in patients with renal cell cancer to kill PD-L1-expressing tumor cells in an HLA-A24-restricted manner [25]. Additionally, Munir et al. demonstrated PD-L1 15-23 peptide could induce HLA-A2-restricted CTLs specific for PD-L1 [24]. Furthermore, they tried using PD-L1 242-264 peptide to stimulate HLA-A2-restricted CD8 + T-cells and found it impotent. Interestingly, we found helper epitopes in PD-L1 242-264 peptide, which is an almost the same sequence as the peptide using this study (PD-L1 241-265 ). Our defined peptide efficiently induced PD-L1 241-265 -specific HTLs from several healthy donors in an HLA-DR-restricted manner, HLA-DR4 or HLA-DR9, suggesting that PD-L1 241-265 peptide is promiscuous. We also immunohistologically detected HLA-DR in 8/24 cases (33.3%) of OSCC, indicating that these tumors could be directly targeted by Th1 cells. Therefore, PD-L1-specific HTLs could play a role not only as a helper cell for CTLs but also as a direct killer cell against tumor cells in patients with OSCC, indicating that antitumor immunotherapy targeting PD-L1 would be a promising strategy for OSCC. Moreover, it has been reported that cytotoxicity of HTLs against cancer cells is mediated through the release of effector cytokines such as IFN-γ, perforin, and granzyme B [26]. We indeed found that PD-L1-specific HTLs showed cytotoxicity against tumor cells in vitro with producing granzyme B and inhibited tumor growth in vivo. To our knowledge, this is the first detailed report about PD-L1-specific HTLs.
Since PD-L1 is expressed not only on tumor cells but also on immune cells such as DCs, autoimmune disease caused by PD-L1-specific T-cells must be considered. Shamaila et al. showed PD-L1-specific CTLs had cytotoxicity to autologous DCs expressing PD-L1 [24]. However, in the present study, PD-L1 241-265 -specific HTLs produced less IFN-γ against autologous DCs compared with tumor cells. This discrepancy may be due to the difference in effector functions between HTLs and CTLs. HTLs mainly function in support of CTLs, while CTLs mainly play the role of killer cell. Therefore, the reaction to PD-L1-expressing normal cells including immune cells must be closely monitored in the clinical setting although we did not see high PD-L1 expression in specimens from non-cancerous patients.
PD-L1 is also expressed on myeloid-derived suppressor cells (MDSCs) to regulate auto-immunity [42]. However, in cancer immunotherapy, the existence of MDSCs indicates poor prognosis and treatment efficacy [43] because of their suppressive function against antitumor immune cells. We, therefore, expect that PD-L1-specific HTLs could eliminate MDSCs in a PD-L1-dependent manner, thereby improving the tumor microenvironment. However, we found that PD-L1-specific HTLs had low killing activity against DCs that expressed PD-L1. Therefore, the interaction between PD-L1-specific HTLs and MDSCs should be further investigated.
Although vaccination with TAA has been designed and implemented for many years, its clinical efficacy has not yet been demonstrated despite success in increasing tumor-specific T-cells in treated patients [44,45]. The remarkable clinical efficacy of immune checkpoint inhibitors demonstrates that the tumor microenvironment is under a more highly suppressive condition than we expected and PD-L1 is one of the key molecules involved in immune suppression. Desired immune responses, therefore, would not be achieved by only vaccinating cancer patients with TAA peptides without blocking immune suppressive signaling, even if immunoadjuvants are simultaneously used. In the current study, we intratumorally injected IFN-β but not IFN-γ in vivo analysis because immunoadjuvants such as poly(I:C) and CpG are often used in clinical settings and mainly induce type I IFNs to enhance immune activity in cancer patients. However, type I IFNs also upregulate PD-L1 molecules in immune cells and tumor cells [46]. This "double-edged sword" effect of type I IFNs is a challenge to be overcome. However, it is a great advantage for a vaccine therapy targeting PD-L1. Indeed, PD-L1-specific HTLs showed higher cytotoxic activity against IFN-γ-pretreated tumor cell lines than untreated tumor cell lines in vitro and intratumoral treatment of IFN-β did not negatively affect the function of PD-L1-specific HTLs for inhibiting tumor growth in vivo. Therefore, although efficacy of the PD-L1 peptide vaccine remains to be evaluated for future, the combination therapy of the PD-L1 peptide with a type I IFN-inducing adjuvant would be a rational strategy for patients with HNSCC.
(See figure on previous page.) Fig. 5 The cytotoxicity of PD-L1 241-265 -specific CD4 + T-cell lines against HLA-matched tumor cell lines expressing PD-L1. a PD-L1 241-265 -specific CD4 + T-cell lines (G1; HLA-DR4-restricted) were cocultured with CFSE-labeled tumor cell lines HSC-4 and Lu65 expressing PD-L1 pre-treated with or without IFN-γ (500 U/ml). HLA-DR-unmatched cell line Sa-3 was used as a negative control. After 6 h of coculture, the cells were collected to evaluate percentages of dead cells by using 7-AAD with flow cytometry. E:T (Effector: Target cells) ratio was 0:1, 10:1, and 30:1. Left panels show representative data of flow cytometry analysis. Right panels show the averages of cytotoxicity of the G1 cell lines against each tumor cell lines. Each result is representative of two separate experiments. b PD-L1 241-265 -specific CD4 + T-cell lines (G1; HLA-DR4-restricted) were cocultured with tumor cell lines HSC-4 and Lu65 pre-treated with or without IFN-γ (500 U/ml). HLA-DR-unmatched cell line Sa-3 was used as a negative control. Supernatants were collected and analyzed by ELISA for Granzyme-B release after 24 h of coculture. Bars and error bars indicate the mean and SD of triplicate determinations, respectively. Each result is representative of two separate experiments