Dynamic alterations of circulating T lymphocytes and the clinical response in patients with head and neck squamous cell carcinoma treated with nivolumab

Abstract

Cancer immunotherapy using immune checkpoint inhibitors (ICIs) has been recognized as a novel therapeutic option for head and neck squamous cell carcinoma (HNSCC). However, only approximately 20–30% of patients with recurrent/metastatic (R/M) HNSCC benefit. Moreover, the mechanisms underlying the response to ICIs remain unclear. We investigated the proportion, activation status, and expression level of immune checkpoint molecules in circulating T cell subsets in R/M HNSCC patients treated with nivolumab using flow cytometry and mass cytometry, and then determined whether treatment response was associated with these values. We also assessed the changes in the frequency of tumor-associated antigens, MAGE-A4 and p53, -specific T cells prior to and after nivolumab treatment using the IFN-γ ELISPOT assay. The proportion of activated CD4+ and CD8+ TEMRA cells significantly increased in the disease-controlled patients but not in disease-progressed patients. As expected, the expression of PD-1 in T cells markedly decreased regardless of the therapeutic response. Meanwhile, T cell immunoglobulin mucin-3 expression on CD8+ T cells was significantly higher in patients with disease progression than in disease-controlled patients after treatment. The frequency of the tumor-associated antigens, MAGE-A4- and p53-specific T cells, was not correlated with clinical responses; however, in the disease-controlled patients, the frequency of MAGE-A4-specific T cells was significantly augmented. We concluded that in R/M HNSCC patients treated with nivolumab, circulating T cells show dynamic alterations depending on treatment efficacy. An analysis of the immunokinetics of circulating T cells could thus provide new insights into rational therapeutic strategies in cancer immunotherapy for HNSCC.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00262-021-03042-y.

Introduction

In cancer patients, T cells possessing antitumor activity become exhausted in circulation as well as the tumor microenvironment, and cannot exert antitumor effector function through the upregulation of immune checkpoint molecules, including programmed cell death-1 (PD-1), lymphocyte activation gene 3 (Lag-3), and T cell immunoglobulin mucin-3 (Tim-3) [1, 2]. PD-1 is a major regulator of T cell exhaustion, and blockade of the PD-1/ programmed cell death-1 ligand 1 (PD-L1) axis reinvigorates exhausted T cells, leading to tumor regression [3, 4]. Cancer immunotherapy using immune checkpoint inhibitors (ICIs) has been recognized as a novel therapeutic option for various malignancies, including head and neck squamous cell carcinoma (HNSCC) [5–7]. However, only approximately 20–30% of patients with recurrent/metastatic (R/M) HNSCC benefit from ICI treatment. Furthermore, the mechanisms underlying the response to ICIs remain unclear. Tissue-based biomarkers, including PD-L1 expression on tumor cells and tumor-infiltrating immune cells, tumor mutation burden, CD8 T cell infiltration, and interferon (IFN)-γ gene signatures, have been extensively investigated [8–10]. However, tumor tissue sampling has several problems that are invasive, sometimes unrepeatable and inaccessible, and do not reflect comprehensive tumor profiling. When ICIs exert therapeutic effects, tumor-infiltrating tumor-specific T cells are activated and kill tumor cells, but also initiate or boost antitumor immune responses, implying a self-sustaining cancer immunity cycle [11]. Under these circumstances, circulating immune cells in peripheral blood can also serve as surrogate markers to evaluate antitumor immune responses in the tumor microenvironment; this is because the effector cells activated in lymph nodes enter the circulatory system, finally reaching the tumor sites and performing their function [12–14]. Moreover, peripheral blood samples are less invasive, easily accessible, and repeatable for analysis compared to tumor tissue. Kamphorst et al. revealed an increase in Ki-67+ PD-1+ CD8 T cells following PD-1-targeted therapy in 70% of lung cancer patients. Further, 80% of patients with clinical benefit exhibited PD-1+ CD8 T cell responses [15]. Similarly, high central memory T cell-to-effector memory ratios in the blood were found to correlate with longer progression-free survival in lung cancer patients receiving nivolumab [16]. In HNSCC, the percentage of PD-1+ CD8+ effector T cells was significantly lower in responders to nivolumab [17], indicating that the co-expression of immune checkpoint molecules on circulating T cells could be a predictor for the treatment efficacy of ICI therapy. Thus, evaluation of the distributional and functional alterations of peripheral immune cell subsets, particularly T cell subsets, may be useful as biomarker candidates for predicting ICI efficacy.

In the present study, we first investigated the proportion, activation status, and expression level of immune checkpoint molecules in circulating T cell subsets in R/M HNSCC patients treated with nivolumab. Thereafter, we determined whether the response to nivolumab treatment was associated with these values. Second, to gain a deeper understanding of longitudinal phenotypic alterations at the single-cell level, cytometry by time-of-flight (CyTOF) analysis was performed using representative samples obtained from eight patients receiving nivolumab treatment. Third, to evaluate dynamic changes in tumor-specific T cell responses, two tumor-associated antigens (TAAs) preferentially expressed in HNSCC [18, 19], melanoma-associated antigen-A4 (MAGE-A4), and p53 were selected, and the changes in frequency of TAA-specific T cells were analyzed. Elucidation of the relationship between dynamic changes in immune responses following ICI administration and treatment responses could lead to the identification of potential biomarkers in patients with R/M HNSCC as well as clarify the immune mechanisms underlying ICI response or resistance.

Materials and methods

Patients and blood collection

Patients with histologically confirmed R/M HNSCC who were treated with nivolumab were eligible for inclusion in this study. Treatment response was evaluated according to RECIST version 1.1. during weeks 8–10. If the patients had an unacceptable level of drug-related toxic effects or rapid disease progression before evaluation, the therapeutic effects were evaluated at that point.

Peripheral blood samples were obtained from 36 patients with R/M HNSCC treated with nivolumab, and peripheral blood mononuclear cells (PBMCs) were isolated by density gradient centrifugation and cryopreserved until used. If possible, blood samples were collected at two time points: before treatment initiation and after four infusions of nivolumab. This study was approved by the Ethical Committee of Gunma University Hospital (HS2017-152). Written informed consent was obtained from each patient.

Flow cytometry

Cryopreserved PBMCs were thawed, washed, and resuspended in phosphate-buffered saline (PBS) containing 1% fetal calf serum (FCS) and 0.1% NaN₃. Non-specific binding of antibodies to Fc receptors on PBMCs was blocked using BD Fc Block (BD Bioscience, San Jose, CA, USA) in accordance with the manufacturer's instructions. Thereafter, PBMCs were stained with antibodies specific for CD3, CD4, CD8, CD25, CD38, CD45RA, CCR7, cytotoxic T-lymphocyte-associated protein 4 (CTLA-4), PD-1, lymphocyte activation gene 3 (Lag-3), Tim-3, forkhead box P3 (Foxp3), and Ki-67. The antibodies used in this study are listed in Supplementary Table 1. For intracellular staining of Foxp3 and Ki-67, the Foxp3 transcription factor fixation/permeabilization concentrate and diluent solutions (eBioscience, Santa Clara, CA) were used according to the manufacturer's instructions. As a negative control, the cells were stained with mouse IgG isotype control (BD Biosciences). Flow cytometry was conducted using a FACSVerse flow cytometer (BD Bioscience), and data analysis was performed using FlowJo software (TreeStar, Ashland, OR).

Mass cytometry

Mass cytometry was performed using the thawed PBMCs. Nonspecific binding of antibodies to the FC receptors was blocked using BD Fc Block (BD Bioscience). Surface antigen staining was performed using the Maxpar Basic Human T Cell Immuno-Oncology Panel Set (Fluidigm, South San Francisco, CA), which contains 24 metal isotope-conjugated antibodies (Supplementary Table 2), according to the manufacturer’s protocol. Stained cells were fixed overnight with 1% PFA. Fixed cells were stained with 250 nM Cell-ID Intercalator-Ir (Fluidigm) in permeabilization buffer for cell identification, and then washed with Maxpar Cell Acquisition Solution (Fluidigm). Thereafter, cells were resuspended in Maxpar Cell Acquisition Solution with a 1:10 dilution of EQ Four Element Calibration beads (Fluidigm). Samples were acquired on a Helios CyTOF mass cytometer (Fluidigm) at an event rate of 500 events/second or less.

Data analysis of mass cytometry results

FCS data files were bead-normalized using CyTOF software version 7.0.8493 (Fluidigm). Cell debris, dead cells, and doublets were excluded for the identification of live cells using FlowJo software (TreeStar, Ashland, OR, USA). CD45+ CD3+ cells were gated as T cells using FlowJo software. Further analysis was performed using R (The R Foundation for Statistical Computing, Vienna, Austria) in combination with R studio version 1.3.1093 (R studio, Boston, MA, USA) based on a previously described workflow [20]. In brief, the intensity of each marker was transformed using the arcsinh scale with cofactor 5. A multi-dimensional scaling (MDS) plot was constructed to demonstrate the similarities between samples using the plotMDS function from the limma package. Cell clustering was performed using the SOM function from the FlowSOM package and the ConsensusClusterPlus function from the ConsensusClusterPlus package. Uniform manifold approximation and projection (UMAP) for dimension reduction was used to visualize the data in a 2D map. A total of 5000 cells were randomly selected from each sample, and UMAP projection was performed using the umap package. Visualization was subsequently performed using the ggplot package.

Immunohistochemistry (IHC) and evaluation

Immunohistochemical detection of MAGE-A4 and p53 in tumor specimens was performed as previously described [21]. Briefly, formalin-fixed paraffin-embedded specimens were sectioned (3 µm thick) and deparaffinized. Antigen retrieval was achieved by boiling the samples at 98 °C for 30 min in citrate buffer (pH 6.0) and 98 °C for 30 min in 20% zinc sulfate solution for MAGE-A4 and p53 staining, respectively. After blocking, slides were incubated at room temperature for 2 h with primary antibodies (anti-MAGE-A4 antibody [1:2000], clone 57 B, MERCK; anti-p53 antibody [1:600], NCL-L-p53-DO7, NOVOCASTRA) and then incubated overnight at 4 °C. Subsequently, slides were incubated with a secondary antibody (Histofine Simple Stain MAX-PO (MULTI), Nichirei) and the reaction products were detected with 3,3’-diaminobenzidine (DAB, DOJINDO, Kumamoto, Japan). Sections were counterstained with Mayer’s hematoxylin.

Slides were assessed by two independent investigators (S.I. and K.C.) in a blinded fashion. For MAGE-A4, each specimen was positive if specific staining was present. For p53, a microscopic examination of nuclear staining was performed and specific staining in > 10% of the tumor cells was defined as positive expression, as described in a previous report [21].

In vitro sensitization and enzyme-linked immunosorbent spot assays

To evaluate the specific T cell responses of two TAAs, MAGE-A4 and p53, IFN-γ ELISPOT assays were performed as previously described [21]. Briefly, thawed PBMCs were cultured in the presence of recombinant protein (10 µg/mL of MAGE-A4 or p53) in a final volume of 1 mL AIM-V medium supplemented with 10 IU/mL interleukin (IL)-2 and 5 ng/mL IL-7 in a 24-well tissue culture plate. Recombinant MAGE-A4 and p53 proteins were purified using a nickel affinity column from transgenic silkworm silk glands expressing MAGE-A4 and TP53 constructs [21, 22]. After 4 days, 1 mL of AIM-V medium containing 10 IU/mL IL-2 was added to each well. Following 3 days of incubation, PBMCs were harvested, washed, and tested using IFN-γ ELISPOT assays.

ELISPOT assays were performed using a Human IFN-γ single-color ELISPOT kit (Cellular Technology Ltd., Cleveland, OH, USA) according to the manufacturer’s protocol. Effector cells (1–5 × 10⁴ cells/well) were co-cultured with PBMCs (1 × 10⁵ cells/well) in the presence of MAGE-A4 or p53 protein (10 µg/mL each) in a 96-well plate precoated with the IFN-γ capture antibody. PBMCs stimulated with 10 µg/mL phytohemagglutinin (PHA) were used as a positive control. The plates were incubated at 37 °C for 24 h. After incubation, the plates were washed and developed with anti-human IFN-γ (biotinylated) and streptavidin–alkaline phosphatase. The number of spot-forming cells (SFCs) in each well was automatically counted using a CTL-ImmunoSpot Analyzer (Cellular Technology Ltd.). The mean number of spots in control wells (no protein) was subtracted from the mean number of spots in the experimental wells, and the results were expressed as SFC per 5 × 10⁴ cells.

Statistical analysis

GraphPad Prism version 8.0 (GraphPad Software, San Diego, CA, USA) was used to perform statistical analyses. The Mann–Whitney U test, Wilcoxon signed-rank test, and Fisher’s exact test were used to determine significant differences in the continuous and categorical variables. Two-sided P values < 0.05 were considered to be statistically significant.

Results

Patient characteristics

We analyzed 36 patients with R/M HNSCC who received the anti-PD-1 blocking antibody, nivolumab. A summary of the patient characteristics is shown in Table 1. The objective response and disease control rates were 22.2% and 41.6%, respectively. Ten patients discontinued 4 courses of nivolumab due to rapid disease progression or deteriorating conditions (n = 8) and intolerable immune-related adverse events (n = 2). Peripheral blood sample was collected from 24 (66.7%) of the 36 patients at two time points: before treatment initiation and after four infusions of nivolumab.

Table 1.

Demographic and clinical characteristics of HNSCC patients in this study

Clinical variable			Total (%)
			n = 36
Age (years): median (range)			66 (46–86)
Sex	Male		31 (86.1)
	Female		5 (13.9)
Primary site	Paranasal sinus		7 (19.4)
	Oral cavity		3 (8.3)
	Nasopharynx		2 (5.6)
	Oropharynx	HPV-positive	3 (8.3)
		HPV-negative	3 (8.3)
	Hypopharynx		14 (38.9)
	Larynx		2 (5.6)
	Parotid gland		1 (2.8)
	External auditory canal		1 (2.8)
Local recurrence	(−)		19 (52.8)
	(+)		17 (47.2)
Lymph node metastasis	(−)		16 (44.4)
	(+)		20 (55.6)
Distant metastasis	(−)		15 (41.7)
	(+)		21 (58.3)
Nivolumab treatment	< 4 times		10 (27.8)
	≥ 4 times		26 (72.2)
Treatment response	CR		0 (0.0)
	PR		8 (22.2)
	SD		7 (19.4)
	PD		21 (58.3)
TAA expression	MAGE-A4	(−)	15 (41.7)
		(+)	21 (58.3)
	p53	(−)	12 (33.3)
		(+)	24 (66.7)

HNSCC: Head and neck squamous cell carcinoma, CR: complete response, PR: partial response, SD: stable disease, PD: progressive disease, TAA: tumor-associated antigen

Characteristics of the T cell subsets and clinical response to nivolumab

First, we used flow cytometry to determine the proportion of T cell subsets (naïve [TN]; CD45RA+ CCR7+, central memory [TCM]; CD45RA−CCR7+, effector memory [TEM]; CD45RA−CCR7−, and terminally differentiated effector memory (TEMRA); CD45RA+ CCR7−), activation status (CD38+ and Ki-67+), and expression of immune checkpoint molecules (CTLA-4, PD-1, Lag-3, and Tim-3) for CD4+ and CD8+ T cells in peripheral blood collected pre-treatment from R/M HNSCC patients administered nivolumab. The flow cytometry gating strategy used for the experiment is shown in Supplementary Fig. 1. Unfortunately, the proportions of each T cell subset, activated T cells, and immune checkpoint molecule-expressing T cells were not significantly different between the two groups (i.e., disease-controlled and disease-progressed patients) (Supplementary Fig. 2). Likewise, in post-treatment peripheral blood, the proportions of each T cell subset and activated T cells were not significantly different between the two groups, except for the proportion of CD4+ TEM cells (p = 0.04) (Fig. 1a and b). Meanwhile, the expression of two immune checkpoint molecules, Lag-3 and Tim-3, on CD8+ T cells in patients with disease progression was significantly higher than that in disease-controlled patients after nivolumab treatment, as shown in Fig. 1c. The expression of Tim-3 in CD4+ T cells also showed a similar tendency; however, the difference was not statistically significant (CD4+ T cells, Lag-3; p = 0.12, Tim-3; p = 0.05, CD8 + T cells, Lag-3; p = 0.04, Tim-3; p = 0.04).

Fig. 1.

Comparison of the proportions of T cell subset, activated T cells, and immune checkpoint molecule-expressing T cells in peripheral blood post-treatment based on the clinical response to nivolumab and analysis using flow cytometry. Peripheral blood samples of 24 R/M HNSCC patients treated with nivolumab were collected after 4 infusions of nivolumab. a The proportion of T cell subsets for each CD4+ and CD8+ T cells. b The proportion of CD38+ and Ki-67+ cells for each CD4+ and CD8+ T cells. c The proportion of the immune checkpoint molecules, CTLA-4, PD-1, Lag-3, and Tim-3 expressing T cells, for each CD4+ and CD8+ T cells

Alterations of the T cell subsets and clinical response to nivolumab

We examined the alterations in the proportion of T cell subsets, activated T cells, and immune checkpoint molecules expressing T cells for CD4+ and CD8+ T cells in the peripheral blood before and after nivolumab treatment. As shown in Fig. 2a, the proportion of CD4+ TEMRA was significantly increased in disease-controlled patients (p = 0.01), but not in patients with disease progression (p = 0.60). Meanwhile, the proportion of CD8+ TEMRA tended to increase in patients with disease control (p = 0.08). Of note, the proportion of activated CD4+ and CD8+ TEMRA T cells identified by CD38 expression was more clearly increased in disease-controlled patients (Fig. 2b, CD4+ TEMRA; p = 0.03, CD8+ TEMRA; p = 0.01). Moreover, the proportion of activated status was significantly increased in all CD8+ T cell subsets, except TEM in disease-controlled patients (CD8+ T cell; p = 0.02, CD8+ TN, p = 0.003, CD8+ TCM; p = 0.01). Regarding immune checkpoint molecules on CD4+ and CD8+ T cells, PD-1 expression on both T cells was markedly decreased regardless of treatment responses (Fig. 2C, CD4+ T cells in disease-controlled patients; p = 0.001, CD4+ T cells in disease-progressed patients; p = 0.0002, CD8+ T cells in disease-controlled patients; p = 0.001, CD8+ T cells in disease-progressed patients; p = 0.0005).

Fig. 2.

Alterations in the proportion of T cell subsets, activated T cells, and immune checkpoint molecule-expressing T cells in the peripheral blood of R/M HNSCC patients treated with nivolumab. The peripheral blood samples collected at two time points, pre-treatment and post-treatment, from 24 R/M HNSCC patients treated with nivolumab were analyzed and compared. a The proportion of T cell subsets for each CD4+ and CD8+ T cells. b The proportion of CD38+ and Ki-67+ cells for each CD4+ and CD8+ T cells. c The proportion of the immune checkpoint molecules, CTLA-4, PD-1, Lag-3, and Tim-3 expressing T cells, for each CD4+ and CD8+ T cells

Mass cytometry analysis

For a further detailed analysis of the expression patterns of immune checkpoint molecules and activation markers on T cell subsets, the blood samples from eight patients, four patients with PR/SD and four patients with PD, were analyzed using mass cytometry. Figure 3a shows the similarities before and after nivolumab treatment using the MDS plot. The distance between samples was calculated based on the expression of the molecules tested. T cells were clustered into eight T cell subsets according to the expression of CD4, CD8, CD45RO, CD45RA, CD27, and CCR7 (Supplementary Fig. 3a). Visualization was then carried out using UMAP (Fig. 3b). Representative marker expression in T cells is shown in Fig. 3c. The log₂-fold change in each cluster between before and after nivolumab treatment was examined (Supplementary Fig. 3b). Nivolumab treatment resulted in an increase in the proportion of CD8+ T cells, CD4+ TEMRA, and CD8+ TEMRA subsets in disease-controlled patients (Supplementary Fig. 3c and d). Moreover, the expression levels of immune checkpoint molecules on T cells and their changes of each T cell subset before and after nivolumab treatment were examined by treatment efficacy, respectively (Fig. 3d–h). As shown in Fig. 3d and e, PD-1 expression in both CD4+ and CD8+ T cells was decreased after nivolumab treatment regardless of treatment response, and Tim-3 expression in CD8+ T cells after nivolumab treatment was higher in patients with disease progression. These results were consistent with those of flow cytometry analysis. The change in OX-40 expression in both CD4+ and CD8+ T cells was decreased in patients with disease progression, while that of CTLA-4 expression in CD4+ T cells was decreased in disease-controlled patients (Fig. 3g and h).

Fig. 3.

Mass cytometry analysis of R/M HNSCC patients treated with nivolumab. a Multi-dimensional scaffold plot representing 8 patients at two time points: before treatment initiation and after 4 infusions of nivolumab. b Cluster identification visualized using UMAP. c Expression level of the four representing lineage markers, CD4, CD8, CCR7, and CD45RA. d–h Expression level of immune checkpoint molecules on T cells. UMAPs show the normalized expression levels of checkpoint molecules. Bar graphs demonstrate the comparison of the expression levels between before and after nivolumab treatment across treatment efficacy. Heatmaps represent the change in the expression levels after nivolumab treatment compared to baseline (pre-treatment) in each T cell subset across treatment efficacy. The values in the heatmap were calculated by subtracting baseline. UMAP, uniform manifold approximation and projection. The asterisk (*) indicates p value less than 0.05

TAA-specific T cell responses and clinical responses to nivolumab

Flow cytometry and mass cytometry analysis revealed that nivolumab treatment resulted in the activation of effector T cells. We investigated the frequency of TAA-specific T cells before and after nivolumab treatment. For this purpose, two TAAs preferentially expressed in HNSCC, MAGE-A4, and p53, were selected. We first used IHC to investigate the expression of MAGE-A4 and p53 in tumor tissues obtained from 36 patients treated with nivolumab. Representative immunohistochemical staining for MAGE-A4 and p53 is shown in Fig. 4a. MAGE-A4 and p53 expression was detected in 21 (58.3%) and 24 (66.7%) patients, respectively (Table 1). Unfortunately, there was no correlation between TAA expression and treatment response (Supplementary Table 3). We proceeded to investigate the frequency of TAA-specific T cells in the PBMCs of TAA-positive patients before and after nivolumab treatment using the ELISPOT assay (Fig. 4b). Although the frequency of MAGE-A4-and p53 specific T cells was not significantly different between disease-controlled and disease-progressed patients before or after nivolumab treatment (Fig. 4c and d), the frequency of MAGE-A4-specific T cells was significantly increased in disease-controlled patients but not in disease-progressed patients after nivolumab treatment (Fig. 4e, disease-controlled patients; p = 0.04, disease-progressed patients; p = 0.81). There was no significant difference in the frequency of p53-specific T cells (Fig. 4e, disease-controlled patients; p = 0.22, disease-progressed patients; p = 0.55); however, 4 (57.1%) of the 7 patients with PR/SD showed a significant increase in the frequency of p53-specific T cells after nivolumab treatment.

Fig. 4.

TAA-specific response and clinical responses to nivolumab. The expression level of two TAAs, MAGE-A4 and p53, in tumor tissues was investigated, and the frequency of TAA-specific T cells in the PBMCs of TAA-positive patients collected before and after nivolumab treatment was measured. a Representative hematoxylin and eosin staining and immunohistochemistry for MAGE-A4 and p53. Patient 4: MAGE-A4-positive and p53-negative case, patient 31: MAGE-A4-negative and p53-positive case (×100 magnification). b Representative well imaging of ELISPOT assays to detect p53-specific T cell responses before and after treatment. c Comparison of the frequency of TAA-specific T cells between patients with disease progression and disease progression before treatment. d Comparison of the frequency of TAA-specific T cells between patients with disease control and disease progression post-treatment. e Alterations in the frequencies of TAA-specific T cells pre- and post-treatment based on treatment efficacy

Discussion

Here, we analyzed peripheral blood samples collected from R/M HNSCC patients treated with nivolumab. Unfortunately, the parameters investigated in this study, including the proportion of T cell subsets, activated T cells, and immune checkpoint molecule-expressing T cells, and the frequency of TAA-specific T cells, are unlikely to be predictive markers for treatment efficacy. Nonetheless, we found several novel points to understand the immunokinetics caused by ICI treatment. As expected, in disease-controlled patients, the proportion of TEMRA cells, particularly CD4+ T cells, increased after nivolumab treatment. This finding was confirmed by mass cytometry analysis. Moreover, the proportion of activated (CD38+) CD4+ and CD8+ TEMRA cells was significantly increased in the disease-controlled patients. Both CD4+ and CD8+ TEMRA cells are known to be more differentiated in terms of effector function, display potent cytotoxicity, and produce IFN-γ upon stimulation [23–26]. In non-small cell lung cancer patients, Kunert et al. [27] demonstrated that CD8+ T cell populations in partial response patients showed enhanced frequencies of TEMRA cells. Regarding CD4+ T cells, Pirozyan et al. [28] revealed increased IFN-γ production by CD4+ TEMRA cells in baseline samples from the responder group compared with the acquired resistance group. These findings suggest that the immune function of TEMRA cells can be used to predict responders to ICIs. Although our study failed to show any correlation between the proportion of CD4+ and CD8+ TEMRA cells or activated CD4+ and CD8+ TEMRA in baseline samples and treatment efficacy, at least in disease-controlled patients, activated CD4+ and CD8+ TEMRA would play pivotal roles in exerting antitumor effects.

Several studies on T cell activation have measured the expression of Ki-67, a nuclear protein associated with cellular proliferation [15, 29, 30]. In the present study, the proportion of Ki-67+ CD4+ and CD8+ T cells failed to demonstrate any significance in clinical outcomes in patients treated with nivolumab. However, another activation marker, CD38-expressing T cells, particularly CD8+ T cells, showed remarkable changes during nivolumab treatment in disease-controlled patients. Wu et al. reported that CD38+ CD8+ T cells expressed higher levels of cytotoxic molecules, cytokines, and PD-1 than CD38-CD8+ T cells in lung cancer [31]. Moreover, CD38+ CD8+ T cells isolated from tumors can be reinvigorated by anti-PD-1 in vitro. Clarke et al. [32] also carried out single-cell transcriptome analysis of cytotoxic T lymphocytes from anti-PD-1 responders, which revealed a significant enrichment of markers linked to activation (CD38) in post-treatment samples compared with pre-treatment samples. Thus, CD38-expressing CD8+ T cells in peripheral blood as well as tumors of patients treated with nivolumab would be reinvigorated and would function as antitumor effector cells.

Among the immune checkpoints tested, PD-1 expression on T cells was markedly decreased in all T cell subsets, regardless of treatment efficacy. As the anti-PD-1 antibody (MIH4) used for flow cytometry binds to an epitope that is distinct from the epitope bound by nivolumab, the decrease in PD-1 expression on T cells after nivolumab treatment could not be explained by PD-1 epitope saturation by nivolumab. Several studies have demonstrated that PD-1 expression on T cell subpopulations was decreased by in vitro PD-1 blockade with nivolumab using flow cytometry and RT-PCR [33, 34]. Likewise, our results revealed that nivolumab treatment reduced PD-1 expression in all T cells in vivo. Although the mechanisms underlying the downregulation of PD-1 on T cells remain largely unclear, one possibility is that PD-1-expressing T cells reinvigorated by nivolumab exhibit their effector function and finally undergo apoptosis. In addition, non-exhausted T cells that have not yet expressed PD-1 may emerge. As this downregulation was observed in both disease-controlled and disease-progressed patients, this phenomenon could not be limited to antitumor T cell populations. Interestingly, Lag-3 and Tim-3 expression on CD8+ T cells in patients with disease progression was significantly higher than that in disease-controlled patients after treatment. The findings of mass cytometry analysis of Tim-3 on CD8+ T cells also support this result. Koyama et al. showed that Tim-3 is upregulated owing to adaptive resistance to anti-PD-1therapy in both mouse models and patients [35]. In HNSCC, Shayan et al. revealed that the PD-1 blockade of tumor-infiltrating lymphocytes (TILs) led to further Tim-3 upregulation in a PI3K/Akt-dependent manner. In a murine HNC tumor model that is partially responsive to anti-PD-1 therapy, Tim-3 was upregulated in TILs from persistently growing tumors [36]. Our results showed similar Tim-3 upregulation in PBMCs at treatment failure. Currently, clinical trials of anti-Tim-3 antibody combined with anti-PD-1 antibody have already reported that this combination treatment was well tolerated and showed preliminary signs of antitumor activity [37]. Additionally, the altered expression of OX40 on CD8+ T cells and CTLA-4 on CD4+ T cells showed different behaviors between the disease-controlled and progressed patients. Thus, the expression of immune checkpoint molecules, besides PD-1, on T cells, particularly CD8+ T cells, may be a key mechanism of resistance to ICI treatments.

Although CD4+ or CD8+ TEMRA T cells were found to proliferate and were reinvigorated by nivolumab treatment, it is unknown whether effector T cells can recognize and kill tumor cells. To address this issue, we investigated the frequency of TAA-specific T cell responses in patients treated with nivolumab. The frequency of MAGE-A4- or p53-specific T cells before and after nivolumab treatment was not related to treatment efficacy. These findings suggest that the presence or frequency of MAGE-A4- and p53-specific T cells would have no direct effect on treatment efficacy. However, in disease-controlled but not disease-progressed patients, the frequency of MAGE-A4-specific T cells significantly increased after nivolumab treatment. Such findings suggest that some populations of effector T cells are activated by nivolumab and lyse tumor cells, resulting in the release of tumor antigens and subsequent induction of effector T cells, ultimately maintaining the cancer immunity cycle in disease-controlled patients. To date, only few studies have analyzed TAA-specific T cell responses in patients responding to nivolumab. Merhi et al. revealed that the NY-ESO-1-specific T cell response in a patient with recurrent HNSCC increased after nivolumab treatment [38]. In contrast, Veatch et al. reported that CD4+ and CD8+ T cell clones specific for lineage-specific antigens and cancer testis antigens did not expand in frequency in the blood during successful treatment in a patient with melanoma [39]. Thus, the kinetics of TAA-specific T cell response may differ depending on the type of cancer, the type of TAAs, expression level of TAAs, and immune status. Although accumulating evidence indicates that the treatment efficacy of ICIs is likely attributable to neoantigen-specific T cells [40–42], the immunokinetics of TAA-specific T cells may also indirectly reflect treatment efficacy.

In conclusion, in R/M HNSCC patients treated with nivolumab, circulating T cells showed dynamic alterations depending on treatment efficacy. Although further studies with a larger number of HNSCC patients treated with nivolumab will be necessary, an analysis of the immunokinetics of circulating T cells would provide new insights into rational therapeutic strategies in cancer immunotherapy for HNSCC.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

CTLA-4

Cytotoxic T-lymphocyte-associated protein 4

CyTOF

Cytometry by time-of-flight

DAB

3,3’-Diaminobenzidine

ELISPOT

Enzyme-linked immunosorbent spot

FCS

Fetal calf serum

Foxp3

Forkhead box P3

HNSCC

Head and neck squamous cell carcinoma

IFN

Interferon

IHC

Immunohistochemistry

ICI

Immune checkpoint inhibitor

IL

Interleukin

Lag-3

Lymphocyte activation gene 3

MAGE-A4

Melanoma-associated antigen-A4

MDS

Multi-dimensional scaling

PBMC

Peripheral blood mononuclear cell

PD-1

Programmed cell death-1

PD-L1

Programmed cell death-1 ligand 1

PHA

Phytohemagglutinin

R/M

Recurrent/metastatic

SFC

Spot-forming-cells

TAA

Tumor-associated antigen

TCM

Central memory T cells

TEM

Effector memory T cells

TEMRA

Terminally differentiated effector memory T cells

Tim-3

T cell immunoglobulin mucin-3

TN

Naïve T cells

UMAP

Uniform manifold approximation and projection

Author contributions

Hideyuki T. and KC were involved in conception and design. Hiroe T., Hideyuki T., KY, KM, YN, MU, MS, SI, IM, TM, KT and HS were involved in sample preparation and data acquisition. Hiroe T., Hideyuki T., TO, ST and KC were involved in analysis and interpretation of data. Hiroe T. and KC were involved in writing of the manuscript. All authors have read and approved the final manuscript.

Funding

This work was supported in part by a Grant-in-Aid for Scientific Research (Grant No. 20K18243 to Hideyuki T., No. 20K09747 to M.S., No. 19K18794 to S.I., No 19K18758 to I.M., and No. 20H03834 to K.C.) from the Ministry of Education, Culture, Sports, Science and Technology; a grant from the Ministry of Agriculture, Forestry and Fisheries of Japan (Research Project for Sericultural Revolution), Japan; and Financial Support for the Promotion of University-Industry Collaborative Research Based on Regulatory Sciences at Gunma University.

Availability of data and material

The datasets used and/or analyzed in this study are available from the corresponding author upon reasonable request.

Declarations

Conflict of interest

The authors declare that they have no conflict interest.

Ethical approval

This study was approved by the Ethical Committee of Gunma University Hospital (HS2017-152).

Informed consent

Written informed consent was obtained from each patient.

Consent for publication

Not applicable.