Peripheral blood samples from de-identified HNSCC patients (n = 32) were obtained from the University of Cincinnati Medical Center. HNSCC patients included in this study were treatment-naïve and had a positive diagnosis of HNSCC by tissue biopsy (See Table 1 for a summary of patient demographics and Supplementary Table S1 for clinical information). Peripheral blood samples of 7 healthy donors (HDs, 4 males and 3 females, age range between 30 and 65 years) were collected from individual donors and from discarded blood units (Hoxworth Blood Center, University of Cincinnati). The demographics of the donors from Hoxworth Blood center were not available. Informed consent was obtained from all HNSCC patients and HDs. The data collected in the study were managed using the Research Electronic Data Capture (REDCap) tools licensed to the University of Cincinnati. Sample collection was approved by the University of Cincinnati Institutional Review Board (IRB no. 2014-4755).

Human serum, l-glutamine, sodium hydroxide, poly-l-lysine, LY294002 HCl, ionomycin, calmodulin, poly-l-lysine, thapsigargin (TG), tetraethylammonium-chloride (TEA-Cl), 1,2-Bis(2-Aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid (BAPTA), MgCl₂ were purchased from Millipore Sigma. Sterile, 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), RPMI-1640 medium, fetal bovine serum, penicillin, streptomycin, Fura-2 and phosphate buffered saline (PBS) were obtained from ThermoFisher. Phosphatidylinositol 3-phosphate diC16 (PI3P) was purchased from Echelon Biosciences. Pembrolizumab (Merck Sharpe and Dohme Corp) samples used in this study were received from the leftover pharmacy supply at Cincinnati Children’s Hospital. Atezolizumab was purchased from Biovision life sciences. PD-L1-Fc chimera was obtained from R&D systems. Stock solutions of LY394002 and TG were prepared in dimethyl sulfoxide and used at 0.1% dilution. Stock solutions of PD-L1-Fc, pembrolizumab and atezolizumab were prepared in sterile PBS.

Peripheral blood mononuclear cells were isolated from whole blood by Ficoll-Paque density gradient centrifugation (Cytiva) as previously described (Chimote et al., 2018). CD8⁺ PBTs were isolated by negative selection using EasySep Human CD8⁺ T cell Enrichment kit (StemCell Technologies). Post isolation, CD8⁺ PBTs were maintained in RPMI-1640 medium supplemented with 10% human serum, 200 U/ml penicillin, 200 mg/ml streptomycin, 1 mM l-glutamine, and 10 mM HEPES. Activation of cells was performed either by adding 40.5 nM of phorbol-12-myristate-13-acetate (PMA, Millipore Sigma) and 1.5 µM of ionomycin (Millipore Sigma) or in cell culture plates coated with 10 μg/ml anti-CD3 and anti-CD28 antibodies (BioLegend) for 72–96 h (h) as previously described (Chimote et al., 2016; Chimote et al., 2018). After isolation, some fresh cells were used for functional studies, the remaining cells were frozen and used later on for flow cytometry experiments.

CD8⁺ PBTs from HNSCC patients were activated for 72–96 h using PMA and ionomycin followed by treatment with the αPD-1 antibody pembrolizumab (10 μg/ml) and/or the αPD-L1 antibody atezolizumab (1 and 10 μg/ml) for 6 h prior to performing functional studies. CD8⁺ PBTs from HDs were plated on cell culture plates coated with PD-L1 (PD-L1-Fc, R&D Systems) (10 μg/ml) and activated with either PMA and ionomycin or (with anti-CD3/CD28 antibodies for 72–96 h followed by treatment with the αPD-1 antibody pembrolizumab (10 μg/ml) for 6 h before performing functional studies. For prolonged treatment with PD-L1, cells were activated for 120 h with plate-bound anti-CD3/CD8 antibodies along with PD-L1-Fc in T cell medium supplemented with 20 IU/ml of IL-2.

Whole cell patch-clamp electrophysiology was used to measure the activity of KCa3.1 and Kv1.3 channels in activated CD8⁺ PBTs cells of HDs and HNSCC patients as described previously (Chimote et al., 2020). The external solution consisted of (in mM): 160 NaCl, 4.5 KCl, 1 MgCl₂, 10 HEPES, pH 7.4. Internal solution consisted of (in mM): 145 K-Aspartate, 2 MgCl₂, 8.5 CaCl₂, 10 EGTA, 10 HEPES, pH 7.2 TRIS buffer (1 μM free Ca²⁺ concentration). Borosilicate glass (World Precision Instruments) pipettes (4–5 MΩ resistance) were fabricated using a P-97 horizontal puller (Sutter Instruments). A voltage-ramp pulse protocol (from −120 to + 50 mV, for 200 ms, holding potential of −70 mV, every 15 s) was used to elicit the currents from CD8⁺ PBTs. Currents were recorded, amplified and digitalized (Axon 200B and Digidata 1320A, Molecular Devices) through a 16-bit A-D/D-A interface. Data acquisition was performed using pClamp 8.0 software (Molecular Devices) and signals were low pass filtered at 2 kHz and digitalized at 100 kHz. The conductance (G) of KCa3.1 channels was calculated as, the ratio of linear fraction of the currents to the slope of ramp voltage stimulus (measured in the voltage range between −100 and −80 mV) after subtraction of leak current. (Chimote et al., 2018). The G of Kv1.3 channels was determined from the same recordings by measuring the peak currents at +50 mV after subtraction of KCa3.1 currents extrapolated by linear regression. This protocol accurately record and separate KCa3.1 and Kv1.3 currents (Newton et al., 2020). LY294002 was delivered extracellular via a manual perfusion system. Calmodulin and PI3P were dissolved in the internal solution and delivered intracellularly via the patch-pipette. The KCa3.1 and Kv1.3 Gs were measured in at least three to five cells for each condition per individual patient.

For divalent free (DVF) currents, cells patched in the whole-cell configuration were pre-incubated with TG (1 µM) in external solution without Ca²⁺ for 10 min followed by perfusion with 20 mM Ca²⁺ for 1 min and, finally, with DVF solution for 2 min (Vaeth et al., 2017). DVF currents were measured to amplify the CRAC currents. The DVF Ringer’s solution contained (in mM): 150 NaCl, 10 HEDTA, 1 EDTA and 10 HEPES (pH 7.4 with NaOH). The 20 mM Ca²⁺ external solution consisted of (in mM): 130 NaCl, 4.5 KCl, 20 CaCl₂, 10 D-glucose and 5 HEPES (pH 7.4 with NaOH). 10 mM TEA-Cl was added to all extracellular solutions to block voltage-gated K⁺ channels. The pipette solution contained (in mM): 135 Cs aspartate, 8 MgCl₂, 8 BAPTA and 1 HEPES (pH 7.2 with CsOH). A ramp (-100 to +100 mV, holding potential of +30 mV) protocol was used for 100 m every 1.5 s to record DVF currents. Analysis of DVF current was performed by measuring the peak current value at the voltage of −100 mV during the ramp step protocol.

Intracellular Ca²⁺ fluxes were measured in activated CD8⁺ PBTs using the ratiometric Ca²⁺ sensitive dye Fura-2. Perfusion with the sarco-endoplasmic pump inhibitor TG allowed us to measure Ca²⁺ fluxes that are independent of TCR stimulation and only rely on downstream signaling events initiated by the release of Ca²⁺ from the endoplasmic reticulum (Robbins et al., 2005). T cells were plated on poly-l-lysine coated coverslips followed by treatment with 1 µM Fura-2 AM (ThermoFisher) at 37°C for ∼30 min. Cells were then washed with RPMI-1640 and maintained at 37°C prior to recordings using InCyt-Im2 Ca²⁺ imaging system (Intracellular Imaging). Coverslips were mounted on the microscope and perfused with a Ca²⁺ free solution for 5 min, followed by perfusion with 1 μMTG in Ca²⁺ free solution for other 5 min to deplete the Ca²⁺ from the intracellular Ca²⁺ stores and open the CRAC channels. Finally, 0.5 mM Ca²⁺ solution was perfused for 10–15 min to allow Ca²⁺ influx through CRAC channels. Ca²⁺ free solution had the following composition (in mM): 155 NaCl, 4.5 KCl, 1 MgCl₂, 10 HEPES, 10 glucose, 2 EGTA, pH 7.4. The 0.5 mM Ca²⁺ solution had the following composition (in mM): 155 NaCl, 4.5 KCl, 2.5 MgCl₂, 10 HEPES, 10 glucose, 0.5 CaCl₂, pH 7.4. A standard calibration curve was used to correlate ratiometric Fura-2 values (340/380 nm ratio) with known Ca²⁺ concentrations as per the protocol provided by the manufacturer. Changes in Ca²⁺ values [Delta (Δ) Ca²⁺], a measure of Ca²⁺ fluxing ability, were determined as the difference between the peak of Ca²⁺ reached after 0.5 mM Ca²⁺ and the baseline Ca²⁺ after the perfusion with 1 μMTG in Ca²⁺ free solution, immediately before the addition of 0.5 mM Ca²⁺. Statistical analysis was performed to detect significant difference in ΔCa²⁺ values post treatment for an individual patient or healthy donor. Only those donors who showed a statistically significant increase in the Ca²⁺ values before and after αPD-1 were included in the positive response (PR) group while those who did not show any statistically significant increase were included in the no-response (NR) group. Individual single cell ΔCa²⁺ values from donors in the PR group and the NR group were then combined to detect significance differences between the two groups. This is because of low sample quantity and variable cell count per patient.

CD8⁺ PBT cells were maintained at a density of 1 × 10⁶ cells/mL and stimulated with PMA/ionomycin for 72 h. Cells were then rinsed with 1x PBS, followed by staining for flow cytometry. The proportion of dead cells were determined using the Zombie UV fixable viability kit (Biolegend). The cells were then fixed with 1% paraformaldehyde (ThermoFisher), washed with 1x PBS and stained overnight with mouse anti-human anti-KCa3.1 biotinylated antibodies (clone 6C1, Alomone). Then, the cells were washed with 1x PBS and incubated for 30 min at room temperature with the following anti-human antibodies (all from Biolegend): anti-CD8-PacificBlue (cloneHIT8A), anti-CD69-APCFire750 (clone FN50), anti-LAG3-BV510 (clone 11C3C56), anti-PD1-BV605 (clone EH12. 2H7), anti-TIM3-BV786 (clone F38-2F2), and anti-streptavidin-PECy7 (clone). Next, the cells were washed with 1x PBS, and permeabilized with BD Cytofix/Cytoperm kit (BD Biosciences) as per the manufacturer’s instructions, and incubated for 30 min at 4°C with mouse anti-human anti-Calmodulin-PerCP (clone 2D1, NOVUS Biologicals), and anti-human anti-Ki67-BV711 (Biolegend).

For determination of ion channel expression, cells were fixed with 4% paraformaldehyde, washed with PBS followed by overnight incubation at 4°C with the following primary antibodies, mouse anti-human KCa3.1 (6C1/ATTO-488), guinea pig anti-human Kv1.3, rabbit anti-human Orai1 (all from Alomone labs) followed by anti-guinea pig (Dy350 goat anti-guinea pig IgG/Thermo Fisher) and anti-rabbit (Alexa Fluor 594 goat anti-rabbit IgG/Thermo Fisher) secondary antibodies. To stain for intracellular ion channel epitopes, cells were permeabilized with BD Cytofix/Cytoperm kit (BD Biosciences) as per manufacturer’s instructions. Cells were then stained for rabbit anti-human STIM1 (Proteintech) primary antibodies followed by secondary antibodies (Alexa Fluor 594 goat anti-rabbit IgG/Thermo Fisher). Specificity of these antibodies was previously reported by our laboratory (Chimote et al., 2018; Newton et al., 2020). All flow cytometry data were collected on LSR Fortessa or LSR II flow cytometer (BD Biosciences), using the FACS Diva software v.6.0. At least 30,000 total events were acquired. Fluorescence minus one controls were also included. The flow cytometry data were analyzed using FlowJo Software version 10.6.1 ((BD Biosciences).

Statistical analyses were performed using Student’s t test (paired or unpaired), Mann-Whitney rank sum test (in experiments where samples failed normality or had unequal variance), and ANOVA or ANOVA on Ranks as indicated. Post hoc testing on ANOVA was done by multiple pairwise comparison procedures using the Holm-Sidak, Dunn’s or Tukey’s methods depending on sample normality and variance. Statistical analysis was performed using SigmaPlot 13.0 (Systat Software Inc.). p value of less than or equal to 0.05 was considered as statistically significant.

Ion channels are fundamental regulators of T cell Ca²⁺ signaling and effector functions (Feske et al., 2015) and we have shown that treatment of HNSCC patients with pembrolizumab increases KCa3.1 and Kv1.3 activity in CD8⁺ PBTs (Newton et al., 2020). Herein, we conducted in vitro experiments to study in detail the effect of pembrolizumab on ion channels in HNSCC T cells. We tested the effect of a short exposure (6 h) to pembrolizumab (αPD-1 antibody) (10 μg/ml) on activated CD8⁺ PBTs. Whole-cell currents were measured before and after αPD-1 treatment to assess KCa3.1 and Kv1.3 activity (Figure 1A). αPD-1 treatment increased the conductance of both KCa3.1 and Kv1.3 channels (Figures 1B,C). We did not detect any effect of αPD-1 on CRAC channel activity (Figures 1D,E). Further control experiments ruled out the possibility that the increase in KCa3.1 and Kv1.3 activity was an artefact due to the extra 6 h the cells treated with αPD-1 were maintained in culture. The KCa3.1 and Kv1.3 conductance in untreated cells at the beginning of the experiment (0 h) and after 6 h (equivalent to the time of exposure to αPD-1) were comparable (Supplementary Figure S1). There was no difference in the capacitance values, an arbitrary measure of cell size and activation state, before and after treatment with pembrolizumab (Supplementary Figure S2). Furthermore, flow cytometry experiments showed no changes in the expression of either ion channels or activation and exhaustion markers in CD8⁺ PBTs from HNSCC patients after αPD-1 treatment (Figure 1F; Supplementary Figures S3–S4). These findings indicate that αPD-1 treatment increases KCa3.1 and Kv1.3, but not CRAC channel, activity in CD8⁺ PBTs of HNSCC patients. Furthermore, αPD-1 treatment did not change the expression of typical markers associated with exhaustion and activation in CD8⁺ PBTs of HNSCC patients.

Since K⁺ channels regulate T cell Ca²⁺ signaling and effector functions, we investigated the functional consequences of the increase in K⁺ channel activity in CD8⁺ PBTs of HNSCC patients induced by αPD-1 treatment. We measured the Ca²⁺ fluxing abilities of CD8⁺ PBTs from HNSCC patients using a TG-based protocol that allows us to assess Ca²⁺ fluxes dependent on ion channel activity while bypassing the TCR (Newton et al., 2020). Activated CD8⁺ PBTs from HNSCC patients were treated with αPD-1 for 6 h (same protocol as Figure 1) and Ca²⁺ fluxes were measured. We observed variable Ca²⁺ responses in these patients’ samples. Therefore, we defined the presence or absence of a response to αPD-1 based on a statistically significant increase in ΔCa²⁺ values before and after in vitro αPD-1 treatment in individual patients. In a subset of patients (n = 3/9), αPD-1 treatment increased the Ca²⁺ response of T cells (Figure 2A). We termed these patients as “positive-response” (PR) patients. Patients whose T cells did not show any increase in the Ca²⁺ response to αPD-1 treatment in vitro were defined as “no-response” (NR) patients (n = 6/9), (Figure 2B). αPD-1 treatment induced a 46% increase in ΔCa²⁺ in the PR group (Figure 2A). Interestingly, we observed that the ΔCa²⁺ at baseline (before αPD-1 treatment) of CD8⁺ PBTs from PR patients was 67% lower than NR patients; even lower than that of HDs CD8⁺ PBTs (Figure 2C). Indeed, we observed a significant negative correlation between the increase in ΔCa²⁺ (fold change for individual patient) induced by αPD-1 versus the baseline ΔCa²⁺ of all HNSCC patients (Figure 2D). We performed this type of analysis because we wanted to assess whether the mean baseline ΔCa²⁺ of a patient could be a predictor of his/her ability to respond to pembrolizumab. This indicates that HNSCC patients whose CD8⁺ PBTs have low baseline Ca²⁺ fluxing abilities were more likely to show a positive response to αPD-1 treatment. Overall, these results indicate that αPD-1 treatment has a rapid effect on ion channel-regulated Ca²⁺ fluxes of HNSCC CD8⁺ PBTs and point to baseline Ca²⁺ fluxing ability of T cells as a possible predictive marker of αPD-1 treatment response. However, questions still remain about the mechanisms by which αPD-1 antibody regulates K⁺ channel activity: is this effect the result of blocking PD-L1/PD-1 binding, and what are the downstream signaling pathways involved?

We tested the corollary of our earlier observation with αPD-1 treatment hypothesizing that prevention of or breaking the interaction of PD-L1 with PD-1 is responsible for the improved K⁺ channel function. While we have no cancer cells in our in vitro setting, CD8⁺ PBTs of HDs and HNSCC patients can be a source of PD-L1 as both express it (Supplementary Figure S5). Others have shown that human CD3⁺ T cells express PD-L1 (Succaria et al., 2020). Therefore, we next tested the effect of an anti-PD-L1 antibody (αPD-L1), atezolizumab, on KCa3.1 and Kv1.3 channels in activated CD8⁺ PBTs of HNSCC patients reasoning that if the effect of pembrolizumab is due to the disruption of PD-L1/PD-1 binding, a similar effect should be produced by αPD-L1. Short-term treatment of HNSCC T cells with atezolizumab (1 and 10 μg/ml for 6 h) increased KCa3.1 function, while Kv1.3 function was increased only at the higher concentration (Figures 3A,B). These results suggest that engagement of PD-L1 by PD-1 may have negative effects on K⁺ channel function in HNSCC T cells and disruption of its binding to the cognate receptor restores the channel’s function. Therefore, we tested the effect of PD-L1 (plate-bound PD-L1-Fc; 10 μg/ml) on KCa3.1 and Kv1.3 in CD8⁺ PBTs of HDs. Cells were incubated with PD-L1 for 72 h followed by treatment with αPD-1 antibody (10 μg/ml) for 6 h. Our data showed that PD-L1 reduced KCa3.1 activity by 65.44%, but, at this concentration, had no effect on Kv1.3 (Figures 3C,D). A similar effect was observed after 5 days exposure to PD-L1 (Supplementary Figure S6). Furthermore, αPD-1 treatment reversed the effect of PD-L1 on KCa3.1 activity (Figure 3C). These results show that, in HDs CD8⁺ PBTs, KCa3.1 channels appear highly sensitive to PD-1 signaling, more than Kv1.3.

We then measured Ca²⁺ fluxes in activated CD8⁺ PBTs cells of HDs exposed to PD-L1 to assess the functional effect of PD-L1. Nine out of twelve samples from HDs showed response to PD-L1 and only the CD8⁺ PBTs samples from donors that showed this response were used in analysis. Cells exposed to PD-L1 showed a small (14%) decrease in ΔCa²⁺ as compared to control cells (Figure 4B). Treatment with pembrolizumab significantly increased ΔCa²⁺ by 39% in PD-L1 exposed cells, and by 25% in 9/12 HDs in the control untreated group (Figures 4A,B). It is noteworthy that treatment with pembrolizumab showed a similar increase of 45% in ΔCa²⁺ in HNSCC CD8⁺ PBTs (Figure 2A). These results indicate that PD-L1 reduced Ca²⁺ fluxing abilities of T cells in HDs thus supporting the notion that high exposure to PD-L1 may decrease the Ca²⁺ fluxing ability of HNSCC CD8⁺ PBTs like it occurs for TILs (Chimote et al., 2017). Furthermore, they suggest that KCa3.1 may be highly sensitive to PD-1 signaling, more than Kv1.3. We thus proceeded to investigate the mechanism through which PD-1 regulates KCa3.1 activity.

Activation and function of KCa3.1 is controlled by a variety of molecules and phosphorylation events (Ohya and Kito, 2018). KCa3.1 function requires binding of the intracellular Ca²⁺ sensor calmodulin (CaM). We recently showed that CD8⁺ PBTs of HNSCC patients have reduced levels of CaM which diminished their KCa3.1 activity (Chimote et al., 2020). PD-L1 is available in the plasma of HNSCC patients both as soluble molecule or carried by exosomes, and the presence of PD-L1⁺ exosomes has been correlated with the increased immune suppressive state of these patients (Theodoraki et al., 2018). These findings raise the possibility that a reduction in CaM by PD-L1 mediates the suppression of KCa3.1 currents. Therefore, we proceeded to determine the signaling pathways involved in PD-L1/PD-1 signaling by exposing HD T cells to PD-L1 for 3 days (72 h; like in the experiments reported above) and 5 days. The latter resembles more the HNSCC patients’ setting where T cells are exposed to PD-L1, both in circulation and in the tumor, for a long time. Exposure to PD-L1 for 3 days did not reduce CaM expression (Figures 5A,B) and, consequently, intracellular supplementation of CaM did not rescue KCa3.1 inhibition (Figures 5C,D). We thus investigated whether the phosphoinositide 3-kinase (PI3K)—phosphatidylinositol-3 phosphatase (PI3P) signaling pathway was instead involved. In T lymphocytes, PI3K favors the production of PI3P from phosphatidylinositol (PI); PI3P activates the nucleoside diphosphate kinase B (NDPK-B) which, ultimately, increases KCa3.1 activity via histidine phosphorylation (Srivastava et al., 2005; Srivastava et al., 2006b). PD-1 signaling in T cells involves the blockade of PI3K and, consequently, the suppression of downstream effector functions (Patsoukis et al., 2013). We performed patch-clamp experiments to determine if the inhibition of KCa3.1 activity by PD-L1 can be mimicked by the PI3K inhibitor LY294002 and rescued by PI3P. Activated CD8⁺ PBTs from HDs were pretreated with LY294002 (10 μM, for 15 min) followed by measurement of KCa3.1 in presence and absence of PI3P (100 nM) delivered intracellularly via patch pipette (Figure 6A). LY294002, similar to PD-L1, reduced KCa3.1 currents while PI3P reversed this inhibition in both LY294002 and PD-L1 treated cells. (Figures 6A–C). These results provide the evidence of involvement of PI3K in the early (3 days) effect of PD-L1 on KCa3.1 activity. Longer exposure to PD-L1 revealed a diminished contribution of PI3K signaling and a role for CaM becomes evident (Figures 6D,E). Electrophysiological experiments showed that PI3P only partially restored the KCa3.1 activity of HDs CD8⁺ PBTs treated with PD-L1 for 5 days (Figure 6D). Flow-cytometry data showed that at this time point there was a significance 40% reduction in CaM expression and not KCa3.1. Interestingly, we observed also a selective reduction in Stim1 (the partner protein of Orai1 that forms CRAC channels) (Figure 6E). Overall, these findings support a role for PI3P signaling and CaM in mediating the effect of PD-L1/PD-1 interaction on KCa3.1 channels in CD8⁺ T cells.