You are here

Article Text

Article menu
Basic tumor immunology
Original research
Comparative analysis of immune infiltrates in head and neck cancers across anatomical sites
  1. Tara Muijlwijk1,2,3,
  2. Dennis N L M Nijenhuis1,2,3,
  3. Sonja H Ganzevles1,2,3,
  4. Arjen Brink1,2,
  5. Changlin Ke1,2,
  6. Joseph N Fass4,
  7. Venkatesh Rajamanickam4,
  8. C René Leemans1,2,
  9. Yoshinobu Koguchi4,
  10. Bernard A Fox4,
  11. Jos B Poell1,2,
  12. Ruud H Brakenhoff1,2 and
  13. Rieneke van de Ven1,2,3
  1. 1Department of Otolaryngology/Head and Neck Surgery, Amsterdam UMC Location VUmc, Amsterdam, The Netherlands
  2. 2Cancer Biology and Immunology, Cancer Center Amsterdam, Amsterdam, The Netherlands
  3. 3Cancer Immunology, Amsterdam Institute for Infection and Immunity, Amsterdam, The Netherlands
  4. 4Providence Cancer Institute, Earle A Chiles Research Institute, Portland, Oregon, USA
  1. Correspondence to Dr Rieneke van de Ven; r.vandeven{at}amsterdamumc.nl

Abstract

Background The response rate to immune checkpoint inhibitors targeting programmed cell death 1 (PD-1) receptor is 13%–18% for patients with recurrent or metastatic head and neck squamous cell carcinoma (HNSCC). Detailed understanding of the tumor immune microenvironment (TIME) is crucial in order to explain and improve this response rate. HNSCCs arise at various anatomical locations including the oral cavity, hypopharynx, larynx and oropharynx. Studies directly comparing immune infiltration between anatomical sites are scarce. Since the distinct locations could drive deviating microenvironments, we questioned whether the immune composition varies across these HNSCC sites.

Methods Here, we characterized the TIME of 76 fresh tumor specimens using flow cytometry and performed single-cell RNA-sequencing on nine head and neck tumor samples.

Results We found major differences in the composition of the TIME between patients. When comparing anatomical sites: tumors originating from the oral cavity had higher T cell infiltrates than tumors from other anatomical sites. The percentage of tumor-infiltrating T-lymphocytes positive for the immune checkpoint PD-1 varied considerably between patients, with the highest fraction of PD-1+ T cells found in larynx squamous cell carcinomas (SCCs). While we had hypothesized that the anatomical sites of tumor origin would drive sample clustering, our data showed that the type of TIME was more dominant and was particularly driven by the fraction of T cells positive for PD-1. Moreover, a high proportion of PD-1+ CD8+ T cells associated with an improved overall survival. Using single-cell RNA-sequencing, we observed that PD-1 expression was highest in the CD8-ENTPD1 tissue resident memory T cell/exhausted T cell and CD4-CXCL13 type 1 T helper cell clusters.

Conclusions We found that oral cavity SCCs had the highest frequencies of T cells. We also observed considerable interpatient heterogeneity for PD-1 on T cells, with noticeably higher frequencies of PD-1+ CD4+ T helper cells in larynx SCCs. Within the entire cohort, a higher fraction of CD8+ T cells positive for PD-1 was linked to improved overall survival. Whether the fraction of PD-1+ T cells within the TIME enables immune checkpoint inhibitor response prediction for patients with head and neck cancer remains to be determined.

  • Head and Neck Neoplasms
  • Tumor Microenvironment
  • Lymphocytes, Tumor-Infiltrating
  • T-Lymphocytes
  • Programmed Cell Death 1 Receptor

Data availability statement

Data are available upon reasonable request. Data are available upon request by contacting the corresponding author.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.

https://doi.org/10.1136/jitc-2023-007573

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    WHAT IS ALREADY KNOWN ON THIS TOPIC

    • While clear distinctions between the tumor immune microenvironment (TIME) from human papillomavirus (HPV)-negative and HPV-positive head and neck cancers have been reported, a detailed head-to-head comparison of the immune composition of head and neck squamous cell carcinomas (HNSCCs) from distinct anatomical sites was lacking.

    WHAT THIS STUDY ADDS

    • This is the largest dataset performing comparative flow cytometric immune profiling on HNSCC originating at different anatomical locations. We show that the TIME differs across HNSCC anatomical sites; namely, tumors originating from the oral cavity appeared to have the highest T cell infiltrate compared with all other anatomical sites and larynx squamous cell carcinomas have the highest frequency of T cells positive for programmed cell death 1 (PD-1). In addition, our data revealed that the type of TIME, led by the fraction of PD-1+ T cells, determined the clustering of patient samples and that some anatomical sites were more represented in one of the clusters. A high fraction of PD-1+ CD8+ T cells correlated with improved overall survival.

    HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

    • This study raises the question whether profiling the fraction of PD-1+ T cells within the TIME has clinical value in terms of anti-PD-1 response prediction for HNSCCs.

    Background

    Response to immune checkpoint inhibitors (ICI), such as those directed against the programmed cell death 1 (PD-1) receptor, is associated with the composition of the tumor immune microenvironment (TIME).1 2 ICI targeting PD-1 belong to the treatment arsenal for recurrent and metastatic head and neck squamous cell carcinoma (HNSCC) and are being tested in clinical trials combined with standard of care, and in the locally advanced setting as neoadjuvant treatment.3–8 Unfortunately, durable clinical responses are only achieved in 13%–18% of the patients with recurrent/metastatic disease.9 10 While for other solid tumors, such as melanoma, it has been evident what predictive role the TIME has,1 2 this has been documented in less detail for HNSCC. There are, however, a few reports that have linked components of the TIME to response to ICI in HNSCC.5 11 Besides its association with ICI response, the immunological orchestra within the TIME is also related to tumor progression and patient outcome. For instance, tumor-infiltrating CD8 counts have been associated with improved overall survival for patients with HNSCC in several studies.12 13 Furthermore, myeloid signatures and aggregates of infiltrating B cells were found to predict worse or more favorable prognosis, respectively.14 15 Together these findings emphasize the importance of characterizing the HNSCC TIME at the different anatomical sites of origin both for prognosis and for ICI treatment response.

    HNSCC is a heterogeneous group of tumors. Not only at the molecular level,16 but also associated with the different etiologies as well as with the various locations within the mucosal linings of the head and neck region in which these tumors originate. In Western countries, HNSCCs are typically located in the oral cavity, hypopharynx, larynx and oropharynx.17 Tumors in the oral cavity, hypopharynx and larynx are predominately carcinogen-induced while oropharyngeal SCCs (OPSCC) can be divided into tumors as result of persistent transforming infection with the human papillomavirus (HPV-positive OPSCC) and carcinogen-induced HPV-negative OPSCC. These two groups are considered as separate disease entities due to divergent etiology, molecular characteristics and clinical presentation. HPV-positive OPSCCs have a more favorable prognosis than HPV-negative tumors.16 18 19

    It has become apparent that tumors induced by persistent HPV infection have more immunogenicity compared with those of HPV-unrelated tumors.20 Specifically, HPV-positive OPSCCs have larger numbers of cytotoxic T cells21–25 as well as B cells23–26 compared with HPV-negative OPSCCs. Also, the ratio of macrophage type 1 (M1) to macrophage type 2 (M2) is higher in HPV-related tumors.21 23 24 While clear distinctions between the TIME from HPV-negative and HPV-positive OPSCCs have been reported, a detailed comparison of the immune composition of HNSCCs from distinct anatomical sites is lacking. Since HNSCCs arise at various sites with deviating microenvironments, we hypothesized that this may cause differences in their immunological landscape. By way of illustration, the microbiota, which consists of more than 700 different bacterial species, may enhance or suppress antitumor immune responses, depending on the strain.27 Also, the distance to lymphoid structures, such as the Waldeyer ring, which is an arrangement of tonsils in the oropharynx, differs per site and might influence the TIME.

    Here, we investigated and compared the TIME of treatment-naive HNSCCs from the oral cavity, larynx, hypopharynx and oropharynx using multiparametric flow cytometry to study whether the anatomical site of origin dictates the immune composition.

    Methods

    Patients and specimens

    Primary tumor tissue was obtained from patients with HNSCC who underwent either a diagnostic biopsy or surgical removal of the tumor between 2019 and 2022 in Amsterdam UMC, location VUmc. None of the patients underwent chemotherapy or radiation prior to treatment.

    Tumor digestion

    Head and neck cancer biopsies were digested within 24 hours of removal. Tumor tissue was collected in Dulbecco’s Modified Eagle Medium (Gibco) and minced into small pieces using a surgical blade (size no. 22; Swann Morton). The cells were dissolved in Roswell Park Memorial Institute (RPMI) 1640 medium (Lonza) containing 0.2 mg/mL DNase-1 (Roche), 1 mg/mL collagenase A (Roche), 5% heat inactivated fetal bovine serum (Biological Industries) and 1% penicillin and streptomycin (Lonza). The suspension (5 mL) was incubated in a 50 mL tube, on a magnet stirrer in a 37°C water bath for three times 45 min with refreshing the medium after each round. Subsequently, the cell suspension was run through a 70 µm cell strainer (BD Falcon). Red blood cells were lysed by incubating the suspension in shock medium (water supplemented with ammonium chloride (NH4Cl, Merck), potassium bicarbonate (KHCO3, Merck) and EDTA (Riedel-deHaën)) for a maximum of 5 min at 4°C. Viable cells were counted by trypan blue exclusion. Dissociated cells were aliquoted for flow cytometry analysis and for single cell RNA-sequencing (scRNA-seq).

    Phenotypic analysis by flow cytometry

    After tumor digestion, the single-cell suspension was rinsed with phosphate-buffered saline (PBS) with 0.1% bovine serum albumin (BSA) and 0.02% sodium azide (NaN3). After 15 min of incubation with Fc receptor blocking agent (1:25, Miltenyi), the cells were stained with antibodies against CD45, epidermal growth factor receptor (EGFR), CD3, CD4, CD8, CD11c, CD19, CD45RA APC-H7, CD25, lymphocyte-activation gene 3 (LAG-3), PD-1, CD80 and programmed death-ligand 1 (PD-L1) by incubating them for 30 min at 4°C. Details on the fluorochrome-conjugated monoclonal antibodies can be found in the online supplemental methods. For intracellular staining, the eBioscience transcription factor staining buffer set (Invitrogen) was used to fix and permeabilize the cells. Next, the cells were stained with antibodies against forkhead box P3 (FOXP3), glycoprotein A repetitions predominant (GARP) or IgG2b by 30 min incubation at 4°C. A BD LSR Fortessa X20 flow cytometer was used for data acquisition and analyses were performed using FlowJo software V.10.8.1. Detailed descriptions of the flow cytometry analyses are provided in the online supplemental methods.

    Supplemental material

    Single-cell RNA-sequencing

    scRNA-seq experiments were performed for nine tumor specimens (fresh or fresh-frozen) after tumor digestion as detailed above. Fluorescence-activated cell sorting (FACS) was performed using the BD FACSAria Fusion sorter for all samples to enrich for viable cells and to separate immune (CD45+) and non-immune (CD45−) cells. For this, the single-cell suspension was stained with CD45 AF700 (1:200, Biolegend), incubated for 30 min at 4°C. Next, the cells were washed and resuspended in PBS containing 5 mM EDTA and 5% BSA. 10 min before sorting, viability dye 7-AAD (1:40, Sigma) was added to the cells. Viable CD45+ and CD45− cells were collected in RPMI supplemented with 0.1% BSA. After sorting, the cells were kept on ice, washed, resuspended in PBS containing 0.01% BSA and counted for viable cells using trypan blue exclusion. The CD45+ and CD45− fractions were merged again, aiming for a 1:1 ratio for the scRNA-seq experiments.

    Five fresh tumor specimens were processed for scRNA-seq using the Dolomite Bio Nadia drop-seq platform. The scRNA-seq on the Nadia instrument protocol version 1.8 was followed. Four tumor specimens (three thawed fresh-frozen and one fresh sample) were processed for scRNA-seq using the 10× Genomics Chromium Single Cell Controller platform. The Chromium Next GEM Single Cell 3ʹ Reagent Kit V.3.1 was used and manufacturer’s instructions were followed. Detailed descriptions of the scRNA-seq runs and analyses are provided in the online supplemental methods.

    External scRNA-seq dataset

    To confirm our findings and to perform differentially expressed gene (DEG) analysis between PD-1+ and PD-1− T cells, we downloaded scRNA-seq data from 18 HPV-negative HNSCCs from Cillo et al28; see online supplemental methods for details.

    The cancer genome atlas dataset and deconvolution analysis

    Clinical and bulk RNA-seq data were downloaded from the cancer genome atlas (TCGA) HNSCC cohort. Deconvolution was performed with the use of scRNA-seq data from Cillo et al28 and Puram et al29 to determine the immune composition from 354 patients with head and neck cancer; see online supplemental table 2 for clinical characteristics and online supplemental methods for deconvolution analysis.

    Supplemental material

    Statistical analysis

    Statistical analyses were performed using GraphPad Prism V.9.3.1 software or R V.4.2.3. Two-tailed paired non-parametric Wilcoxon rank-sum test and unpaired non-parametric Mann-Whitney tests were used to compare two paired and unpaired groups, respectively. One-way unpaired nonparametric Kruskal-Wallis tests were used to compare multiple groups. P<0.05 was considered a statistically significant difference.

    Results

    Clinical characteristics

    Primary tumor specimens from 76 patients with head and neck cancer were collected for the current study. The clinical characteristics of the patient cohort are summarized in table 1 (online supplemental table 1 for a detailed list). In total, 13 patients with HPV-positive oropharyngeal tumors were included. The other 63 tumors were HPV-negative, from which 13 were OPSCC, 23 OCSCC, 14 HSCC and 13 LSCC. The median age of the patients was 67 with significantly younger patients in the HPV-positive OPSCC group (60 years) compared with the HPV-negative HNSCC sites (69 years, p=0.002), as expected. The majority of the patient cohort was male (69.7%). This was true for every site except for patients with OCSCC of which 47.8% was male. In addition, all patients with LSCC were male in this study. From 25 patients matched tumor-adjacent (non-malignant) mucosal tissue was collected, see online supplemental table 1.

    Supplemental material

    View this table:
    Table 1

    Clinical characteristics of the HNSCC cohort by primary site

    Activated regulatory T cells sparse in tumor-adjacent mucosa while present in tumor tissue

    Flow cytometry was performed to identify the phenotype and frequencies of tumor cells, non-immune CD45- cells and main leukocyte populations (for gating strategy, see online supplemental figure 1). First, immune infiltration within tumor tissue and paired tumor-adjacent mucosal tissue was compared for 25 patients with head and neck cancer (figure 1). We found a larger immune cell frequency in tumor compared with mucosal tissue (figure 1A). Within the immune cell fraction, no differences between tumor and mucosa were observed in frequencies of CD19+ B cells, CD11c+ myeloid cells, total CD3+ T cells and CD4+ T helper cells (figure 1B–D). However, interestingly, the fraction CD8+ T cells was higher in tumor-adjacent mucosa compared with tumor whereas activated regulatory T cells (aTreg) were almost absent in mucosa and predominantly present in the tumor (figure 1F–H). Moreover, also when looking at the proportion of cells out of the total cell fraction analyzed, aTregs were found to be more abundant in tumor samples compared with the adjacent mucosa (figure 1I).

    Supplemental material

    Figure 1
    Figure 1

    Frequencies of immune subsets in tumor tissue and paired tumor-adjacent mucosal tissue from patients with head and neck cancer (n=25). (A) Frequency of CD45+ immune cells out of total cell count. (B–G) Fraction of indicated immune subsets out of total (CD45+) immune cells. (H) Ratio of CD8+ T cells to activated regulatory T cells (aTregs). Data are presented as mean and error bars represent SD. P values were obtained using a paired non-parametric Wilcoxon test. (I) Pie charts of average cell type fractions out of total cells for tumor-adjacent mucosal tissue and paired tumor tissue (n=16) with significant higher fractions of CD19+ B cells (*), CD11c+ myeloid cells (**), total CD3+ T cells (**), CD4+ T helper (Th) cells (*), aTregs (****) and other immune cells (**) in tumor tissue compared with adjecent mucosal tissue. Paired non-parametric Wilcoxon tests were performed with *0.05<p < 0.01, **p<0.01, ***p<0.001, ****p<0.0001.

    Higher T cell frequencies in oral cavity SCCs compared with other sites

    In order to further characterize the TIME of HNSCCs, we quantified immune subsets per patient (figure 2). The immune composition of 41 head and neck cancers are displayed per patient as well as on average per subsite in figure 2A,B, respectively. Only a part of the cohort is shown since for some samples the number of single cells after tumor digestion was too low for the full flow cytometry panel; therefore, only for 41 out of 76 patients CD11c frequencies were assessed (online supplemental table 1). Fractions of T cells, B cells and myeloid cells varied considerably across patients, emphasizing the interpatient heterogeneity of head and neck cancers (figure 2A). Nevertheless, the average immune cell type fractions differed considerably across anatomical sites (figure 2B). To quantify these differences in the TIME across HNSCC sites, immune cell frequencies from the total 76 tumor specimens were plotted and compared between sites (figure 2C–I). Tumors originating from the oral cavity appeared to have the highest T cell infiltrate compared with all other HPV-negative tumors from different anatomical sites (figure 2F). Apparently, this could not be explained by a higher CD8+ cytotoxic T cell fraction (figure 2G). Yet, CD4+ T helper cell fractions were larger in OCSCCs when compared with LSCCs and OPSCCs (figure 2H). Moreover, aTreg fractions were larger in OCSCC compared with OPSCC (figure 2I). To investigate whether these higher T cell frequencies in OCSCC were also perceived in non-tumor tissue, we quantified T cell fractions in available tumor-adjacent mucosal tissue data (online supplemental figure 2). While T cell infiltration was higher in oral cavity SCC, this was not the case in mucosa from the oral cavity compared with mucosa from hypopharynx or larynx. Next, the higher T cell infiltration in OCSCCs was confirmed when analyzing the immune composition of HPV-negative HNSCCs in the TCGA dataset by deconvolution of bulk RNA-seq data (online supplemental figure 3, online supplemental table 3). In line with our data, tumors originating from the oral cavity had higher CD4+ T helper cell frequencies compared with other anatomical sites (online supplemental figure 3B, D).

    Supplemental material

    Figure 2
    Figure 2

    Immune composition in tumor tissue from patients with head and neck cancer per site. Phenotype and frequencies were obtained using flow cytometry. Frequencies are plotted on the y-axis and head and neck squamous cell carcinoma (HNSCC) sites on the x-axis. (A) Fraction of CD8+ T cells, CD4+ T helper (Th) cells, activated regulatory T cells (aTreg), CD11c+ myeloid cells, CD19+ B cells and other immune cells are indicated (A) per patient (n=41), grouped per HNSCC site as well as (B) on average presented in pie charts, per HNSCC site. (C) Frequency of CD45+ immune cells out of total cell count. Fraction of (D) CD19+ B cells, (E) CD11c+ myeloid cells, (F) CD3+ T cells, (G) CD8+ cytotoxic T cells, (H) CD4+ Th cells and (I) aTregs out of total immune cells. All graphs represent n=76 patients except for 3A and 3D (n=41). Data are presented as mean and error bars represent SD. HNSCC sites specified with a colored bar, see legend. To compare anatomical HNSCC sites, p values were calculated using one-way analysis of variance (ANOVA) Kruskal-Wallis test comparing oral cavity SCC (OCSCC), HSCC, larynx SCC (LSCC) and human papillomavirus (HPV)-negative oropharyngeal SCCs (OPSCC). To compare HPV-negative and HPV-positive OPSCC, p values were obtained using an unpaired non-parametric Mann-Whitney test. *P<0.05, **p<0.01 and (J) Immune subtypes plotted against anatomical site, HPV status, T stage, age and sex. P values obtained by one-way ANOVA Kruskal-Wallis test or an unpaired non-parametric Mann-Whitney test with dark blue p<0.01, blue p<0.05, light blue p<0.10 and white p>0.10.

    As expected, HPV-positive OPSCC harbored higher B cell and T cell infiltrates compared with HPV-negative tumors (figure 2D and F). Correspondingly, we observed significantly higher CD4+ T helper frequencies in HPV-positive OPSCCs (figure 2G). Next, HPV-negative tumors were grouped according to tumor size (T stage), age (HPV-negative cohort stratified for median age of 69 years) and sex (online supplemental figure 4). No differences were found in immune composition between smaller (T1 and T2) and larger tumors (T3, T4a and T4b). Interestingly, patients younger than 70 years of age had a significantly lower CD8+ T cell to aTreg ratio compared with older patients (p=0.047, online supplemental figure 4P). Furthermore, women had greater T cell infiltrates compared with men (p=0.043, online supplemental figure 4T).

    Summarizing, the TIME of HNSCCs is rather heterogeneous and varied when comparing primary tumor site, HPV status and sex (figure 2J). Specifically, tumors originating from the oral cavity displayed higher CD4+ T helper cell and aTreg frequencies compared with other anatomical sites.

    Variability of PD-1+ tumor-infiltrating T cells across anatomical sites

    Next, we investigated inhibitory and activation molecules in the TIME of HNSCC. We compared the expression of PD-1, LAG-3 and GARP among T cell subsets from all patients (figure 3A, online supplemental figures 5, 6A and 7). Of these, clearly the checkpoint molecule PD-1 was most abundant on CD8+ T cells, CD4+ T helper cells and aTregs. Subsequently, the presence of PD-1 on T cells across HNSCC sites was explored. Interestingly, a clear separation of tumors could be observed based on T cells positive for PD-1. Namely, one group exhibited high percentages of PD-1+ T cells opposed to tumors with barely any PD-1+ T cells. This separation was observed across all HNSCC sites. Nevertheless, the percentage of CD4+ T helper cells positive for PD-1 was significantly higher in LSCC compared with OCSCC and OPSCC (figure 3B–D). The level of PD-1 expression on T cells was also assessed (online supplemental figure 6). Most T cells exhibited intermediate PD-1 expression (PD-1_int) rather than high expression of PD-1 (PD-1_high), but further quantitative distinction of PD-1 did not yield additional insight in our analysis (online supplemental figure 6B–E).

    Figure 3
    Figure 3

    Expression of programmed cell death 1 (PD-1) receptor, lymphocyte-activation gene 3 (LAG-3) and glycoprotein A repetitions predominant (GARP) on T cells. (A) Percentage of CD8+ T cells, CD4+ T helper (Th) cells and activated regulatory T cells (aTregs) positive for PD-1, LAG-3 and GARP across 73 head and neck squamous cell carcinoma (HNSCC) samples. Expression of PD-1 on (B) CD8+ T cells, (C) CD4+ Th cells and (D) aTregs separated per HNSCC site, indicated by colored violins with dotted lines representing median and quartiles. To compare anatomical HNSCC sites, p values were obtained using one-way analysis of variance Kruskal-Wallis test comparing oral cavity SCC (OCSCC), HSCC, larynx SCC (LSCC) and human papillomavirus (HPV)-negative oropharyngeal SCCs (OPSCC). To compare HPV-negative and HPV-positive OPSCC, p values were obtained using an unpaired non-parametric Mann-Whitney test. *P<0.05, **p<0.01.

    Expression of PD-L1 by EGFR+ CD45– tumor cells as well as CD11c+ myeloid cells was also studied (online supplemental figure 8). Overall, tumor cells from LSCC exhibited higher frequencies of PD-L1+ cells compared with tumor cells from OCSCC. In addition, tumor cells from HPV-positive OPSCC had higher fraction of PD-L1+ cells compared with HPV-negative OPSCC. Furthermore, presence of PD-L1 on tumor and myeloid cells, as well as PD-1 on T cells were plotted against T stage, age and sex; however, no differences were found (online supplemental figure 9D).

    Taken together, PD-1 was found to be the most abundant molecule on T cells from HNSCC from the checkpoint molecules investigated. Noteworthy, tumors located in the larynx appeared to have the highest frequencies of PD-1+ CD4+ T helper cells compared with other anatomical sites. Furthermore, a remarkable separation in tumors presenting high fractions of T cells positive for PD-1 and tumors with limited frequencies of PD-1+ T cells could be observed across anatomical sites. The fraction of PD-1+ T cells did not differ on HPV status, T stage, age or sex (online supplemental figure 10).

    Unsupervised hierarchical clustering of HNSCC TIMEs

    Taking the TIME differences across HNSCC sites into account, we aimed to investigate all parameters in an unbiased manner using an unsupervised hierarchical clustering on 73 tumor specimens based on their immune subset frequencies and marker expression obtained by flow cytometry (figure 4A). Three clusters could be discerned, stratified for the fraction of T cells present, as well as their positivity for PD-1. Cluster 1 (left) displayed moderate-to-high percentages of CD3+ T cells infiltrating and low fractions of PD-1+ CD8+ T cells, CD4+ T helper cells and aTregs, whereas the other two clusters (middle and right: cluster 2 and 3, respectively) displayed relatively high percentages of PD-1+ T cells. Of note, clusters 2 and 3, while both showing moderate-to-high fractions of PD-1+ T cells, were different with regards to T cell frequencies, which were high for cluster 2 and low for cluster 3. Interestingly, the majority of LSCCs (8/12; 67%) were found within cluster 3. Similarly, HPV-negative OPSCCs were also predominately (9/13; 70%) found in cluster 3; however, the majority of HPV-positive OPSCCs (9/12; 75%) were in cluster 2. OCSCCs and HSCCs were found divided across all three clusters.

    Figure 4
    Figure 4

    Unsupervised hierarchical clustering, overall survival and correlation matrix based on immune subset frequencies and checkpoint expression on immune cells of head and neck squamous cell carcinoma (HNSCC) samples. (A) Ward’s hierarchical agglomerative clustering was performed on percentages derived from flow cytometry data, presented as rows. Columns represent 73 patients with HNSCC with cohort number specified. Primary site of cancer, human papilloma virus (HPV) status, sex, age (grouped by median age of 67), T stage and disease stage (according to TNM classification 8th edition, 2017) are specified with colors. Three patients were excluded since no programmed cell death 1 (PD-1) data were available (see online supplemental table 1). (B) Overall survival in 61 HPV-negative HNSCCs with respect to the percentage of CD8+ T cells expressing PD-1. Kaplan-Meier curve for 15 patients with low percentages and 46 patients with high percentages of PD-1-positive T cells with p value and HR displayed. Groups are based on the 25th lowest percentile of PD-1+ CD8+ T cells in HPV-negative HNSCCs. (C) Correlation between cell frequencies and marker expression on tumor and immune cells using all 76 HNSCC samples. Spearman correlation of percentages obtained using flow cytometry, rows with missing data were excluded. Correlation matrix with correlation coefficient from −1 (blue) to 1 (red). P<0.05 considered as statistically significant correlations, marked with *. (D–F) Spearman correlation percentage of CD8+ T cells positive for PD-1 and (D) B cells positive for PD-1, (E) activated regulatory T cells (aTregs) positive for PD-1 and (F) CD4+ T helper (Th) cells positive for PD-1.

    We evaluated the clinical relevance of the three observed clusters by performing a survival analysis but did not find significant differences (online supplemental figure 11). Since the fraction of PD-1+ T cells clearly separated the patient cohort in two main clusters, we also compared survival data of patients with low and high proportions of PD-1+ CD8+ T cells (figure 4B). After examining the distribution of PD-1+ CD8+ T cells across HPV-negative HNSCCs, we divided HPV-negative HNSCC specimens in two groups based on the 25th lowest percentile of the fraction PD-1+CD8+ T cells determined by flow cytometry (online supplemental figure 12). Patients with higher fractions (> 33.2%, n=46) PD-1+ CD8+ T cells had an improved overall survival compared with patients with lower fractions (≤ 33.2 %, n=15, p=0.015; HR 0.32).

    The fraction of PD-1+ CD8+ T cells correlates with PD-1 on other immune subsets, indicative of an activated TIME

    Correlations of immune cell prevalence and checkpoint expression on immune cells were explored using a correlation matrix (figure 4C). In this matrix positive (in red) as well as negative (in blue) correlations within the TIME from 76 head and neck tumor samples are displayed from different HNSCC sites together. When only HPV-negative tumor samples were taken along, a similar correlation matrix was observed (data not shown). Notable are the positive correlations between positivity of PD-1 on CD8+ T cells, CD4+ T helper cells, aTregs and B cells (figure 4D–F), suggestive of a general activation of different tumor-infiltrating lymphocyte (TIL) subsets within the TIME of those patients. Similarly, the presence of the checkpoint molecule LAG-3 on CD8+ T cells, CD4+ T helper cells and aTregs correlated with LAG-3 on other T cells as well as PD-1 on CD4+ T helper cells. Furthermore, tumors with high CD8+ T cell frequencies also had high CD4+ T helper cell and B cell infiltrates. However, infiltrated tumors, with high percentages of immune cells, did not necessarily have a high fraction of T cells positive for PD-1 (online supplemental figure 13A–E). Positivity of PD-L1 on tumor cells did not correlate with PD-1 on T cells (online supplemental figure 12F).

    PD-1 is transcriptionally expressed at highest level in CD8-ENTPD1 tissue resident memory and CD4-CXCL13 type 1 T helper cell clusters

    We further explored the TIME of head and neck cancers using scRNA-seq for nine tumor specimens (figure 5). The Dolomite Bio Nadia drop-seq platform was used to produce single cell RNA-sequencing libraries for five HNSCC samples, while 10× Genomics technology was used for four samples (online supplemental table 1). After preprocessing, we continued with 14 781 cells from 9 samples. For cell-type identification, we used expression of known markers: T cells (CD2, CD3, CD6), B cells (MS4A1/CD20, CD19, BANK1), plasma cells (IGHG, IGKC), myeloid cells (ITGAX/CD11c, LYZ, CD14) granulocytes (TPSAB1, KIT, MS4A2), fibroblasts (DCN, THY1/CD90, COL), tumor/epithelial cells (EGFR, KRT, EPCAM) and endothelial cells (SELE, ENG, PECAM1) (figure 5B,C, online supplemental table 4 for differential expression (DE) features per cluster). Samples from the two different platforms merged well (online supplemental figure 14A). However, when isolating T cells for subclustering, we encountered the problem of low unique gene counts per cell in samples analyzed by the Nadia Dolomite platform (online supplemental figure 14B). This was not a problem when annotating the major subsets in figure 5B. However, when subclustering T cells, differences between clusters are more subtle and therefore technical differences may become more problematic. Therefore, only the four samples obtained with the 10× Genomics platform were used for T cell subclustering analyses.

    Supplemental material

    Figure 5
    Figure 5

    Single-cell RNA-sequencing (scRNA-seq) for nine head and neck squamous cell carcinoma (HNSCC) samples. (A) scRNA-seq workflow. From primary tumor tissue, a single-cell suspension was made. Using fluorescence-activated cell sorting (FACS), viable cells were sorted. The Nadia Dolomite platform or 10× Genomics was used to generate gel bead-in-emulsion (GEMs). After preparing the libraries, they were sequenced using the NextSeq500 (Nadia samples) or NovaSeq6000 (10× samples) (B) Uniform Manifold Approximation and Projection (UMAP) of nine HNSCC samples together. Major cells are annotated. (C) Average expression (from −1 in blue to 2 in red) in major cell clusters (x-axis) of various signature genes used for cell-type annotation (y-axis).

    We isolated 5483 T cells from four HNSCC specimens and were able to annotate γδ T cells, CD4+ T cells and CD8+ T cells (figure 6A, online supplemental figure 14C–E). Next, 2859 CD8+ T cells and 1171 CD4+ T cells were isolated for subclustering. Three evident CD8+ T cell subclusters were distinguished (figures 6B, 5D and F, online supplemental figure 15A–E, online supplemental table 5 for DE features per cluster). First of all, we identified a CD8-CCR7 T naive (Tn)-like cluster with expression of among others naive markers LEF1, TCF7 and SELL as well as IL7R and S1PR1. Additionally, a CD8-GZMK T effector-memory (Tem) cluster was observed with expression of killer cell lectin-like receptor subfamily G member 1 (KLRG1), CXCR3, CXCR4 and CXCR5. Lastly, a CD8-ENTPD1 hybrid tissue-resident memory T cell (Trm)/exhausted T cell (Tex) cluster was identified with high expression of CXCL13 as well as tissue resident markers ITGAE/CD103 and ENTPD1/CD39 and transcription factors ZNF683, ID1, TOX and HOPX. Cells in this cluster expressed effector molecules (GZMH, IFNG and PRF1) as well as exhaustion marker LAYN and inhibitory molecules (LAG-3, TIGIT and HAVCR2/TIM-3). The gene PDCD1 (encoding PD-1) was highest expressed in CD8-ENTPD1 Trm/Tex cells and only moderately in the CD8-GZMK Tem and CD8-CCR7 Tn clusters (figure 6D and F).

    Supplemental material

    Figure 6
    Figure 6

    T cell subclustering from four single-cell RNA-sequencing (scRNA-seq) libraries. (A) Uniform Manifold Approximation and Projection (UMAP) of 5483 T cells isolated from four head and neck squamous cell carcinoma (HNSCC) samples processed with 10× Genomics platform. CD4+ T cell, CD8+ T cell and γδ T cell clusters were identified. The CD4+ T cell and CD8+ T cell clusters (encircled with dotted line) were bioinformatically isolated for further subclustering. (B–C) UMAP of (B) 2859 CD8+ T cells and (C) 1171 CD4+ T (D–E) UMAP of (D) CD8+ T cells and (E) CD4+ T cells with expression (from 0 in gray to 3 in red) of PDCD1/PD-1. (F–G) Average expression (from −1.0 in blue to 1.0 in red) in (F) CD8+ T cell and (G) CD4+ T cell clusters (x-axis) of various features (y-axis).

    Four CD4+ T cell subclusters were identified (figure 6C, E and G, online supplemental figure 15F–J, online supplemental table 6 for DE features per cluster). First, a clear CD4-FOXP3 Treg cluster was distinguished with expression of Treg markers IL2RA/CD25, IKZF2 and BATF as well as inhibitory molecules such as TIGIT, LAYN and LRRC32/GARP. We also observed a CD4-ITGAE Trm cluster with expression of tissue resident markers ITGAE/CD103 and ENTPD1/CD39, effector molecules such as GNLY and inhibitory molecules HAVCR2/TIM-3 and LAG-3. This cluster also expressed high levels of proliferation markers MKI67, TYMS and MCM5. Furthermore, a small CD4-TCF7 Tn-like cluster with expression of among others naive markers LEF1 and CCR7 was notable. Lastly, a CD4-CXCL13 type 1 T helper (Th1) cluster was found with potent CXCL13 and PDCD1 expression and modest CXCR5 expression in only a minority of the cells within the cluster. The highest PDCD1 expression was observed in CD4-CXCL13 Th1 whereas moderate expression was seen in CD4-ITGAE Trm and inferior expression was seen in CD4-TCF7 Tn-like and CD4-FOXP3 Tregs (figure 6E and G).

    Supplemental material

    Summarizing, with scRNA-seq, we confirmed the presence of major cell types in the TIME of head and neck cancer specimens. In addition, we observed highest PDCD1/PD-1 expression in the CD8-ENTPD1 Trm/Tex and CD4-CXCL13 Th1 clusters.

    CXCL13 is most differentially expressed in PD-1-positive CD8+ T cells and CD4+ T helper cells

    To validate our findings and to increase the power for additional analyses on PD-1+ T cells, we used an external 10× scRNA-seq dataset of Cillo et al (online supplemental figure 16).28 After filtering, we ended up with 39 820 immune cells from 18 HPV-negative head and neck tumor samples. We isolated the T cells and performed subclustering on CD8+ T cells and CD4+ T helper cells (online supplemental figure 15A–D).

    In line with the analyses on our scRNA-seq dataset, we could discriminate a CD8-GZMK Tem cluster with expression of KLRG1, CXCR3 and CXCR4 as well as a CD8-ENTPD1 Trm/Tex cluster (online supplemental figures 15E and 16c). Also, a CD8-MKI proliferative cluster was present with comparable expression as CD8-ENTPD1 Trm/Tex cluster but in addition potent expression of cell cycle genes such as TYMS, STMN1, NUSAP1 and USP1. PDCD1/PD-1 expression was most prominent in the CD8-ENTPD1 Trm/Tex and CD8-MKI67 proliferative clusters and inferior in CD8-GZMK Tem, as found in our scRNA-seq dataset (online supplemental figures 16E and 17A, B). To further characterize the PD-1 positive T cells, we performed a mRNA DEG analysis between PD-1+ and PD-1− CD8+ T cells (online supplemental figure 17C, online supplemental table 7). We found 216 DEGs and among the upregulated features in the PD-1 positive group were cytokines such as CXCL13, CCL3, CCL4, inhibitory markers TIGIT, HAVCR2/TIM-3, LAG-3, cytotoxic markers GZMB, IFNG, PRF1 and the tumor reactive marker ENTPD1/CD39. Among the DEGs in PD-1 negative T cells were signature genes for effector memory T cells GZMK and CXCR4.

    Supplemental material

    Regarding CD4+ T helper cells, the CD4-CXCL13 Th1 cluster again exhibited highest PDCD1/PD-1 expression (online supplemental figure 16A, B and F). Also, only modest PDCD1/PD-1 expression was perceived in the CD4-GZMK Tem, CD4-LEF1 Tn-like and CD4-ISG15 interferon T helper cell clusters. When performing a DEG analysis between PD-1+ and PD-1− CD4+ T helper cells, we found 870 DEGs with again CXCL13 and among others CCL4, CCL3, IFNG, LAG3, TIGIT, GZMB and HAVCR2/TIM-3 being potently expressed in PD-1 expressing T cells (online supplemental figure 17C, online supplemental table 8).

    Supplemental material

    Together, using an external scRNA-seq dataset, we confirmed that the CD8-ENTPD1 Trm/Tex and CD4-CXCL13 Th1 clusters exhibited highest PDCD1/PD-1 expression. Moreover, when comparing PD-1 positive with PD-1 negative T cells, cytokines such as CXCL13 as well as inhibitory, cytotoxic and tumor-reactive markers were most discriminative both for the CD8+ T cells and CD4+ T helper cells.

    Discussion

    We hypothesized that the different anatomical sites of origin might dictate the type of TIME within HNSCC, and in this study we therefore characterized the TIME of HNSCC stratified for anatomical site and HPV using multiparametric flow cytometry on human primary tumor material. To our knowledge, this is the first study investigating the immune composition between HNSCC sites using fresh tumor tissue of such an extended patient cohort, covering all major anatomical sites. Moreover, our analysis included a comparison between matched tumor-adjacent mucosa and tumor for twenty five patient samples.

    Comparing tumor-adjacent mucosa to matched tumor tissue, more leukocytes were present within the tumor specimens, with the most noticeable differences being reduced frequencies of CD8+ T cells and a potent increase in aTreg frequencies in tumor compared with adjacent mucosa. This could not be explained by a specific anatomical site showing divergent frequencies of T cell subsets in tumor-adjacent mucosa. While others also reported higher Treg frequency in cancer tissue compared with non-cancerous tissue,30 31 we are the first reporting a matched comparison in this magnitude. Comparing HNSCC anatomical sites, our data showed increased T cell infiltration in tumors originating from the oral cavity compared with oropharyngeal, hypopharyngeal and laryngeal sites. This observation raises the question whether the oral microbiome influenced the higher T cell frequency at this site, since it is reported that microbes residing in the oral cavity could affect the TIME by enhancing the influx of T cells.32 However, when comparing the immune composition of tumor-adjacent mucosal tissue originating from oral cavity to mucosa from the larynx and hypopharynx, a difference was not detected. This suggests that the oral microbiome, at least in normal tissue, did not affect the abundance of T cells at this anatomical site. We are the first to report these elevated levels of T cells in patients with OCSCC. Specter et al studied TILs by disease site using tissue microarray and found an increased immune infiltrate in oropharyngeal tumors compared with those located in the oral cavity or larynx.12 However, it should be emphasized that the comparison was not stratified according to HPV status. Hence, the difference in TILs can presumably be explained by higher infiltration in HPV-positive OPSCC. This emphasizes the importance of stratifying for HPV status before comparing HNSCC anatomical sites. In line with previous studies, we observed increased B and T lymphocytes in the TIME of HPV-positive compared with HPV-negative OPSCCs.20–26 The differences in immune landscape between HPV-related and HPV-unrelated HNSCCs are relatively well-studied. Of note, most studies used tissue microarray,21 immunohistochemistry (IHC)25 or RNA-sequencing data22–24 for this comparison. Extensive flow cytometry studies like the current study, comparing HPV-negative OPSCC versus HPV-positive OPSCC, are scarce. Lechner et al performed flow cytometry but compared HPV-positive OPSCC with HPV-negative tumors from distinct anatomical sites.26

    While several differences in the immune composition were observed between anatomical sites, additional noteworthy findings, unrelated to anatomical site, emerged that we had not anticipated to find. In our dataset, we identified a group of HNSCCs that exhibit high fractions of T cells positive for PD-1. This may reflect lymphocytes upregulating PD-1 as response to previous activation and therefore be indicative of a more inflamed TIME. We speculate that targeting the PD-1/PD-L1 axis might be beneficial especially for these tumors as this would release the break from the PD-1+ T cells present and facilitate their cytotoxic activity. Supporting this hypothesis, in the study of Hanna et al, immune phenotypes were investigated from baseline HNSCC biopsies of anti-PD-1/PD-L1 responders (n=8) and non-responders (n=34). The responder group had a slightly higher percentage of CD8+ T cells positive for PD-1 (21.6% vs 18.9%, not significant).11 This is in line with findings of Prat et al in patients with melanoma, lung cancer and HNSCC. Bulk RNA sequencing prior to anti-PD-1 therapy showed that, although not significant, PD-1 mRNA expression was linked to response. Moreover, PD-1 was significantly higher in patients with non-progressive disease compared with progressive disease (PD).33 Also, Luoma et al found that circulating PD-1+ killer cell lectin like receptor G1 (KLRG1)− CD8+ T cell fractions from patients with OCSCC prior to ICI treatment were significantly higher in patients with high pathological response versus low pathological response.34 Of note, no flow cytometry data were available of tumor biopsies to see whether this predictive value holds true for tumor-infiltrating CD8+ T cells positive for PD-1. While it remains to be determined for HNSCC whether the abundance of PD-1+ T cells has predictive value, for other solid tumors it has been more evident that anti-PD-1/PD-L1 responders have higher fractions of PD-1 on TILs at baseline when compared with non-responders.1 35 36 Kumagai et al found higher percentages PD-1+ CD8+ TILs in patients with gastric cancer, melanoma and non-small cell lung cancer (NSCLC) who responded to ICI compared with non-responders.36 Taken together, it is tempting to speculate that PD-1+ TILs are reinvigorated when targeting the PD-1/PD-L1 axis and so can predict response to PD-1/PD-L1-targeted ICI in HNSCCs.

    We found that the fraction of PD-1+ CD8+ T cells is associated with an improved prognosis in HPV-unrelated HNSCC. PD-1+ TILs might reflect an initiated antitumor response which could explain why their presence is favorable. While we have to take our limited follow-up period into account, our results are in line with previous reports.37 38 Of note, aforementioned papers evaluated PD-1 expression by immunofluorescence (IF) or (multiplex) IHC whereas we are the first to report this improved survival using flow cytometry. Wu et al showed a strong correlation between PD-1+ CD8+ T cell density obtained by IF and the fraction PD-1+ CD8+ T cells by flow cytometry in NSCLC (R2=0.74, p<0.001).39 In addition, Thommen et al presented a correlation between IHC and flow cytometry data in NSCLC (R2=0.76, p<0.001).40 Based on these studies on lung cancer, it seems likely that also for HNSCC, PD-1 expression assessed by different techniques may correlate, but this should be confirmed in matching IHC/flow cytometry datasets. Altogether, while it is evident that frequencies of PD-1+ T cells is associated with an improved prognosis, it remains to be determined whether it could be of added value in anti-PD-1 response prediction for HNSCCs.

    Besides the percentage of T cells positive for PD-1, the level of PD-1 expression could also be of importance. Namely, Thommen et al showed in lung cancer that the fraction of PD-1_high CD8+ T cells was higher in ICI responders compared with non-responders and was associated with improved OS.40 Emphasizing the importance of stratifying PD-1+ TILs into PD-1_high and PD-1_int. In our study, the fraction of PD-1_int CD8+ T cells was the largest and only few tumors had high numbers of PD-1_high CD8+ TILs. This raises the question whether the low number of PD-1_high TILs hampers responsiveness to ICI for patients with HNSCC. Contrary, others report that, while PD-1_high CD8+ T cells are most cytotoxic, they are less proliferative on PD-1 blockade. CD8+ T cells with intermediate PD-1 expression appeared to be most proliferative in response to PD-1 blockade.41

    We noted on transcriptomic level, using scRNA-seq, that the CD8-ENTPD1 Trm/Tex and CD4-CXCL13 Th1 clusters exhibited the highest PDCD1/PD-1 expression, which is in line with previous findings.42–45 Within the CD8-ENTPD1 Trm/Tex cluster, expression of among others ITGAE/CD103 and ENTPD1/CD39 was observed. ITGAE allows interaction with E-cadherin on tumor cells and at the same time retains TILs at epithelial surface so that they do not circulate.46 ITGAE in combination with ENTPD1 expression is indicative for tumor-reactive T cells.42 Trm cells reside in tumor tissue and are responsible for the first adaptive immune response. They are able to react faster than circulating memory cells. They are linked to an improved prognosis in several solid tumors including HNSCC.42 While we observed expression of inhibitory molecules such as LAYN, LAG-3, TIGIT and HAVCR2/TIM-3 in this CD8-ENTPD1 Trm/Tex cluster, these cells also expressed effector molecules GZMH, IFNG and PRF1, which emphasizes their intact functionality. When comparing PD-1 positive with PD-1 negative CD8+ T cells, CXCL13 was most potently expressed in PD-1+ T cells. CXCL13 is known to be important for B cell migration and tertiary lymphoid structure (TLS) formation which are thought to play a favorable role in the antitumor immune response in solid tumors.2 While CXCL13 secretion by CD8+ T cells is not well described in HNSCC, Thommen et al showed in NSCLC that PD-1_high CD8+ T cells, which were predictive for ICI response, produce CXCL13.40 Also, neoantigen-reactive CD8+ T cells in NSCLC express CXCL13.47 48 Furthermore, in breast cancer CD8+ T cells with CXCL13, HAVCR2/TIM-3, LAG-3, PRF1 and GZMB expression expand on anti-PD-1 treatment.49 Goswami et al showed in a bladder cancer murine model that CXCL13-/- mice were resistant for anti-PD-1 therapy.50 While in a pan-cancer cohort of 1008 checkpoint inhibitor treated patients CXCL13 was upregulated in responders versus non-responders,48 the role of CXCL13 expressing PD-1+ CD8+ T cells and their potential role in ICI response still has to be elucidated for HNSCCs. Within the CD4+ T cell clusters, PDCD1/PD-1 mRNA expression was highest in the CD4-CXCL13 Th1 cluster. Moreover, CXCL13 was again explicitly highest expressed in the PD-1 expressing T cells. CD4+ T cells with high expression of CXCL13 and PDCD1 but modest CXCR5 expression were recently described to drive TLS formation in nasopharyngeal SCC.45

    In conclusion, in this study, we show that of the tested anatomical sites, oral cavity SCCs have the highest frequencies of T cells within freshly monitored tumor biopsies. Additionally, tumors that arose in the larynx had the highest frequencies of PD-1+ CD4+ T helper cells. We observed considerable interpatient heterogeneity regarding the immune composition across all tumors. Interestingly, we also observed a high variability in the fraction of T cells positive for PD-1. Namely, part of the patients exhibited high fractions of PD-1+ T cells while others only had few or even lacked PD-1 expression. The fraction of PD-1+ CD8+ T cells was associated with an improved overall survival. Whether the fraction of T cells positive for PD-1 has clinical value in terms of ICI response prediction has to be unraveled in future studies.

    Supplemental material

    Data availability statement

    Data are available upon reasonable request. Data are available upon request by contacting the corresponding author.

    Ethics statements

    Patient consent for publication

    Ethics approval

    In accordance with the Declaration of Helsinki, written informed consent was obtained from all involved individuals. This study was approved by the Institutional Review Board of the VU medical center (ID: 2008.071 / A2016.035, NL22230.029.08). Participants gave informed consent to participate in the study before taking part.

    Acknowledgments

    The authors wish to thank all the patients who participated in this study, Niels Wondergem and Annouk Pierik for patient inclusion and support with tissue collection, Irene Nauta for support with patient database, Cora Chadick and Tanja Konijn from the Microscopy and Cytometry Core Facility, Amsterdam UMC for support with fluorescence-activated cell sorting prior to the single cell RNA-sequencing experiments, Jasper Koning and Klaas Mulder for advice concerning sequencing and analyses, Core Facility Genomics, Amsterdam UMC for running the 10X, clinicians and nurses from the department of Otolaryngology-head and neck surgery from Amsterdam UMC for support with tissue collection.

    References

    1. ↵
      1. Tumeh PC,
      2. Harview CL,
      3. Yearley JH, et al
      . PD-1 blockade induces responses by inhibiting adaptive immune resistance. Nature 2014;515:568–71. doi:10.1038/nature13954
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      1. Helmink BA,
      2. Reddy SM,
      3. Gao J, et al
      . B cells and tertiary lymphoid structures promote immunotherapy response. Nature 2020;577:549–55. doi:10.1038/s41586-019-1922-8
      OpenUrlCrossRefPubMed
    3. ↵
      1. Burtness B,
      2. Harrington KJ,
      3. Greil R, et al
      . Pembrolizumab alone or with chemotherapy versus cetuximab with chemotherapy for recurrent or metastatic squamous cell carcinoma of the head and neck (KEYNOTE-048): a randomised, open-label, phase 3 study. Lancet 2019;394:1915–28. doi:10.1016/S0140-6736(19)32591-7
      OpenUrlCrossRefPubMed
    4. ↵
      1. Cohen EEW,
      2. Soulières D,
      3. Le Tourneau C, et al
      . Pembrolizumab versus methotrexate, docetaxel, or cetuximab for recurrent or metastatic head-and-neck squamous cell carcinoma (KEYNOTE-040): a randomised, open-label, phase 3 study. Lancet 2019;393:156–67. doi:10.1016/S0140-6736(18)31999-8
      OpenUrlCrossRefPubMed
    5. ↵
      1. Vos JL,
      2. Elbers JBW,
      3. Krijgsman O, et al
      . Neoadjuvant immunotherapy with nivolumab and Ipilimumab induces major pathological responses in patients with head and neck squamous cell carcinoma. Nat Commun 2021;12:7348. doi:10.1038/s41467-021-26472-9
    6. ↵
      1. Schoenfeld JD,
      2. Hanna GJ,
      3. Jo VY, et al
      . Neoadjuvant nivolumab or nivolumab plus ipilimumab in untreated oral cavity squamous cell carcinoma: a phase 2 open-label randomized clinical trial. JAMA Oncol 2020;6:1563–70. doi:10.1001/jamaoncol.2020.2955
      OpenUrl
    7. ↵
      1. Leidner R,
      2. Crittenden M,
      3. Young K, et al
      . Neoadjuvant immunoradiotherapy results in high rate of complete pathological response and clinical to pathological downstaging in locally advanced head and neck squamous cell carcinoma. J Immunother Cancer 2021;9:e002485. doi:10.1136/jitc-2021-002485
    8. ↵
      1. Wise-Draper TM,
      2. Gulati S,
      3. Palackdharry S, et al
      . Phase II clinical trial of neoadjuvant and adjuvant pembrolizumab in resectable local-regionally advanced head and neck squamous cell carcinoma. Clin Cancer Res 2022;28:1345–52. doi:10.1158/1078-0432.CCR-21-3351
      OpenUrl
    9. ↵
      1. Ferris RL,
      2. Blumenschein G,
      3. Fayette J, et al
      . Nivolumab for recurrent squamous-cell carcinoma of the head and neck. N Engl J Med 2016;375:1856–67. doi:10.1056/NEJMoa1602252
      OpenUrlCrossRefPubMed
    10. ↵
      1. Mehra R,
      2. Seiwert TY,
      3. Gupta S, et al
      . Efficacy and safety of pembrolizumab in recurrent/metastatic head and neck squamous cell carcinoma: pooled analyses after long-term follow-up in KEYNOTE-012. Br J Cancer 2018;119:153–9. doi:10.1038/s41416-018-0131-9
      OpenUrlCrossRefPubMed
    11. ↵
      1. Hanna GJ,
      2. Lizotte P,
      3. Cavanaugh M, et al
      . Frameshift events predict anti-PD-1/L1 response in head and neck cancer. JCI Insight 2018;3:e98811. doi:10.1172/jci.insight.98811
    12. ↵
      1. Spector ME,
      2. Bellile E,
      3. Amlani L, et al
      . Prognostic value of tumor-infiltrating lymphocytes in head and neck squamous cell carcinoma. JAMA Otolaryngol Head Neck Surg 2019;145:1012–9. doi:10.1001/jamaoto.2019.2427
      OpenUrl
    13. ↵
      1. Feng Z,
      2. Bethmann D,
      3. Kappler M, et al
      . Multiparametric immune profiling in HPV- oral squamous cell cancer. JCI Insight 2017;2:14. doi:10.1172/jci.insight.93652
      OpenUrl
    14. ↵
      1. Tsujikawa T,
      2. Kumar S,
      3. Borkar RN, et al
      . Quantitative multiplex immunohistochemistry reveals myeloid-inflamed tumor-immune complexity associated with poor prognosis. Cell Rep 2017;19:203–17. doi:10.1016/j.celrep.2017.03.037
      OpenUrlPubMed
    15. ↵
      1. Phanthunane C,
      2. Wijers R,
      3. de Herdt M, et al
      . B-cell clusters at the invasive margin associate with longer survival in early-stage oral-tongue cancer patients. Oncoimmunology 2021;10:1882743. doi:10.1080/2162402X.2021.1882743
    16. ↵
      1. Leemans CR,
      2. Snijders PJF,
      3. Brakenhoff RH
      . The molecular landscape of head and neck cancer. Nat Rev Cancer 2018;18:269–82. doi:10.1038/nrc.2018.11
      OpenUrlCrossRefPubMed
    17. ↵
      1. Cramer JD,
      2. Burtness B,
      3. Le QT, et al
      . The changing therapeutic landscape of head and neck cancer. Nat Rev Clin Oncol 2019;16:669–83. doi:10.1038/s41571-019-0227-z
      OpenUrlCrossRefPubMed
    18. ↵
      1. Ang KK,
      2. Harris J,
      3. Wheeler R, et al
      . Human papillomavirus and survival of patients with oropharyngeal cancer. N Engl J Med 2010;363:24–35. doi:10.1056/NEJMoa0912217
      OpenUrlCrossRefPubMedWeb of Science
    19. ↵
      1. Machiels J-P,
      2. René Leemans C,
      3. Golusinski W, et al
      . Squamous cell carcinoma of the oral cavity, larynx, oropharynx and hypopharynx: EHNS-ESMO-ESTRO clinical practice guidelines for diagnosis, treatment and follow-up. Ann Oncol 2020;31:1462–75. doi:10.1016/j.annonc.2020.07.011
      OpenUrlCrossRefPubMed
    20. ↵
      1. Wondergem NE,
      2. Nauta IH,
      3. Muijlwijk T, et al
      . The immune microenvironment in head and neck squamous cell carcinoma: on subsets and subsites. Curr Oncol Rep 2020;22:81. doi:10.1007/s11912-020-00938-3
    21. ↵
      1. Nguyen N,
      2. Bellile E,
      3. Thomas D, et al
      . Tumor infiltrating lymphocytes and survival in patients with head and neck squamous cell carcinoma. Head Neck 2016;38:1074–84. doi:10.1002/hed.24406
      OpenUrl
    22. ↵
      1. Mandal R,
      2. Şenbabaoğlu Y,
      3. Desrichard A, et al
      . The head and neck cancer immune landscape and its immunotherapeutic implications. JCI Insight 2016;1:e89829. doi:10.1172/jci.insight.89829
    23. ↵
      1. Chen X,
      2. Yan B,
      3. Lou H, et al
      . Immunological network analysis in HPV associated head and neck squamous cancer and implications for disease prognosis. Mol Immunol 2018;96:28–36. doi:10.1016/j.molimm.2018.02.005
      OpenUrlPubMed
    24. ↵
      1. Gameiro SF,
      2. Ghasemi F,
      3. Barrett JW, et al
      . Treatment-naive HPV+ head and neck cancers display a T-cell-inflamed phenotype distinct from their HPV- counterparts that has implications for immunotherapy. Oncoimmunology 2018;7:e1498439. doi:10.1080/2162402X.2018.1498439
    25. ↵
      1. Schneider K,
      2. Marbaix E,
      3. Bouzin C, et al
      . Immune cell infiltration in head and neck squamous cell carcinoma and patient outcome: a retrospective study. Acta Oncol 2018;57:1165–72. doi:10.1080/0284186X.2018.1445287
      OpenUrl
    26. ↵
      1. Lechner A,
      2. Schlößer HA,
      3. Thelen M, et al
      . Tumor-associated B cells and humoral immune response in head and neck squamous cell carcinoma. OncoImmunology 2019;8:1535293. doi:10.1080/2162402X.2018.1535293
      OpenUrlPubMed
    27. ↵
      1. Chattopadhyay I,
      2. Verma M,
      3. Panda M
      . Role of oral microbiome signatures in diagnosis and prognosis of oral cancer. Technol Cancer Res Treat 2019;18:1533033819867354. doi:10.1177/1533033819867354
    28. ↵
      1. Cillo AR,
      2. Kürten CHL,
      3. Tabib T, et al
      . Immune landscape of viral- and carcinogen-driven head and neck cancer. Immunity 2020;52:183–99. doi:10.1016/j.immuni.2019.11.014
      OpenUrlCrossRefPubMed
    29. ↵
      1. Puram SV,
      2. Tirosh I,
      3. Parikh AS, et al
      . Single-cell transcriptomic analysis of primary and metastatic tumor ecosystems in head and neck cancer. Cell 2017;171:1611–24. doi:10.1016/j.cell.2017.10.044
      OpenUrlCrossRefPubMed
    30. ↵
      1. Fraga M,
      2. Yáñez M,
      3. Sherman M, et al
      . Immunomodulation of T helper cells by tumor microenvironment in oral cancer is associated with CCR8 expression and rapid membrane vitamin D signaling pathway. Front Immunol 2021;12:643298. doi:10.3389/fimmu.2021.643298
    31. ↵
      1. Lechner A,
      2. Schlößer H,
      3. Rothschild SI, et al
      . Characterization of tumor-associated T-lymphocyte subsets and immune checkpoint molecules in head and neck squamous cell carcinoma. Oncotarget 2017;8:44418–33. doi:10.18632/oncotarget.17901
      OpenUrlPubMed
    32. ↵
      1. Wang K-K,
      2. He K-Y,
      3. Yang J-Y, et al
      . Lactobacillus suppresses tumorigenesis of oropharyngeal cancer via enhancing anti-tumor immune response. Front Cell Dev Biol 2022;10:842153. doi:10.3389/fcell.2022.842153
      OpenUrl
    33. ↵
      1. Prat A,
      2. Navarro A,
      3. Paré L, et al
      . Immune-related gene expression profiling after PD-1 blockade in non-small cell lung carcinoma, head and neck squamous cell carcinoma, and melanoma. Cancer Res 2017;77:3540–50. doi:10.1158/0008-5472.CAN-16-3556
      OpenUrlAbstract/FREE Full Text
    34. ↵
      1. Luoma AM,
      2. Suo S,
      3. Wang Y, et al
      . Tissue-resident memory and circulating T cells are early responders to pre-surgical cancer immunotherapy. Cell 2022;185:2918–35. doi:10.1016/j.cell.2022.06.018
      OpenUrlCrossRef
    35. ↵
      1. Daud AI,
      2. Loo K,
      3. Pauli ML, et al
      . Tumor immune profiling predicts response to anti-PD-1 therapy in human melanoma. J Clin Invest 2016;126:3447–52. doi:10.1172/JCI87324
      OpenUrlCrossRefPubMed
    36. ↵
      1. Kumagai S,
      2. Togashi Y,
      3. Kamada T, et al
      . The PD-1 expression balance between effector and regulatory T cells predicts the clinical efficacy of PD-1 blockade therapies. Nat Immunol 2020;21:1346–58. doi:10.1038/s41590-020-0769-3
      OpenUrlPubMed
    37. ↵
      1. Badoual C,
      2. Hans S,
      3. Merillon N, et al
      . PD-1-expressing tumor-infiltrating T cells are a favorable prognostic biomarker in HPV-associated head and neck cancer. Cancer Res 2013;73:128–38. doi:10.1158/0008-5472.CAN-12-2606
      OpenUrlAbstract/FREE Full Text
    38. ↵
      1. Kikuchi M,
      2. Yamashita D,
      3. Hara S, et al
      . Clinical significance of tumor-associated immune cells in patients with oral squamous cell carcinoma. Head Neck 2021;43:534–43. doi:10.1002/hed.26498
      OpenUrl
    39. ↵
      1. Wu S-P,
      2. Liao R-Q,
      3. Tu H-Y, et al
      . Stromal PD-L1-positive regulatory T cells and PD-1-positive CD8-positive T cells define the response of different subsets of non-small cell lung cancer to PD-1/PD-L1 blockade immunotherapy. J Thorac Oncol 2018;13:521–32. doi:10.1016/j.jtho.2017.11.132
      OpenUrlCrossRefPubMed
    40. ↵
      1. Thommen DS,
      2. Koelzer VH,
      3. Herzig P, et al
      . A transcriptionally and functionally distinct PD-1(+) Cd8(+) T cell pool with predictive potential in non-small-cell lung cancer treated with PD-1 blockade. Nat Med 2018;24:994–1004. doi:10.1038/s41591-018-0057-z
      OpenUrlCrossRefPubMed
    41. ↵
      1. Pauken KE,
      2. Torchia JA,
      3. Chaudhri A, et al
      . Emerging concepts in PD-1 checkpoint biology. Semin Immunol 2021;52:101480. doi:10.1016/j.smim.2021.101480
      OpenUrl
    42. ↵
      1. Duhen T,
      2. Duhen R,
      3. Montler R, et al
      . Co-expression of Cd39 and CD103 identifies tumor-reactive CD8 T cells in human solid tumors. Nat Commun 2018;9:2724. doi:10.1038/s41467-018-05072-0
    43. ↵
      1. Duhen R,
      2. Fesneau O,
      3. Samson KA, et al
      . PD-1 and ICOS coexpression identifies tumor-reactive CD4+ T cells in human solid tumors. J Clin Invest 2022;132:12. doi:10.1172/JCI156821
      OpenUrl
    44. ↵
      1. Ren X,
      2. Zhang L,
      3. Zhang Y, et al
      . Insights gained from single-cell analysis of immune cells in the tumor microenvironment. Annu Rev Immunol 2021;39:583–609. doi:10.1146/annurev-immunol-110519-071134
      OpenUrlCrossRef
    45. ↵
      1. Li JP,
      2. Wu CY,
      3. Chen MY, et al
      . PD-1(+)CXCR5(-)CD4(+) Th-CXCL13 cell subset drives B cells into tertiary lymphoid structures of nasopharyngeal carcinoma. J Immunother Cancer 2021;9. doi:10.1136/jitc-2020-002101
    46. ↵
      1. Le Floc’h A,
      2. Jalil A,
      3. Vergnon I, et al
      . Alpha E beta 7 integrin interaction with E-Cadherin promotes antitumor CTL activity by triggering lytic granule polarization and exocytosis. J Exp Med 2007;204:559–70. doi:10.1084/jem.20061524
      OpenUrlAbstract/FREE Full Text
    47. ↵
      1. Hanada K-I,
      2. Zhao C,
      3. Gil-Hoyos R, et al
      . A phenotypic signature that identifies neoantigen-reactive T cells in fresh human lung cancers. Cancer Cell 2022;40:479–93. doi:10.1016/j.ccell.2022.03.012
      OpenUrl
    48. ↵
      1. Litchfield K,
      2. Reading JL,
      3. Puttick C, et al
      . Meta-analysis of tumor- and T cell-intrinsic mechanisms of sensitization to checkpoint inhibition. Cell 2021;184:596–614. doi:10.1016/j.cell.2021.01.002
      OpenUrlCrossRefPubMed
    49. ↵
      1. Bassez A,
      2. Vos H,
      3. Van Dyck L, et al
      . A single-cell map of Intratumoral changes during anti-PD1 treatment of patients with breast cancer. Nat Med 2021;27:820–32. doi:10.1038/s41591-021-01323-8
      OpenUrlCrossRefPubMed
    50. ↵
      1. Goswami S,
      2. Chen Y,
      3. Anandhan S, et al
      . ARID1A mutation plus CXCL13 expression act as combinatorial biomarkers to predict responses to immune checkpoint therapy in mUCC. Sci Transl Med 2020;12:548. doi:10.1126/scitranslmed.abc4220
      OpenUrl

    Supplementary materials

    Footnotes

    • Twitter @MuijlwijkTara, @bernardafox, @DeRieneke

    • Contributors Data collection: TM. Patient database: SHG. Data analyses: TM, DNLMN, AB, CK, JNF, VR, JBP, YK and RvdV. Supervision: JBP, CRL, YK, BAF, RHB and RvdV. Funding acquisition: CRL, RHB and RvdV. Writing of the original draft: TM. Writing review and editing: JBP, RHB and RvdV. Read and approved the final version of the manuscript: all authors. Guarantor: RvdV.

    • Funding This work is financially supported, in whole or in part, by Cancer Center Amsterdam (CCA, PV 19/02), Amsterdam UMC Young Talent fund.

    • Competing interests None declared.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.