Tumor-treating fields (TTFields) are a localized, antitumoral therapy using alternating electric fields, which impair cell proliferation. Combining TTFields with tumor immunotherapy constitutes a rational approach; however, it is currently unknown whether TTFields’ locoregional effects are compatible with T cell functionality. Healthy donor PBMCs and viably dissociated human glioblastoma samples were cultured under either standard or TTFields conditions. Select pivotal T cell functions were measured by multiparametric flow cytometry. Cytotoxicity was evaluated using a chimeric Ag receptor (CAR)–T–based assay. Glioblastoma patient samples were acquired before and after standard chemoradiation or standard chemoradiation + TTFields treatment and examined by immunohistochemistry and by RNA sequencing. TTFields reduced the viability of proliferating T cells, but had little or no effect on the viability of nonproliferating T cells. The functionality of T cells cultured under TTFields was retained: they exhibited similar IFN-γ secretion, cytotoxic degranulation, and PD1 upregulation as controls with similar polyfunctional patterns. Glioblastoma Ag–specific T cells exhibited unaltered viability and functionality under TTFields. CAR-T cells cultured under TTFields exhibited similar cytotoxicity as controls toward their CAR target. Transcriptomic analysis of patients’ glioblastoma samples revealed a significant shift in the TTFields-treated versus the standard-treated samples, from a protumoral to an antitumoral immune signature. Immunohistochemistry of samples before and after TTFields treatment showed no reduction in T cell infiltration. T cells were found to retain key antitumoral functions under TTFields settings. Our data provide a mechanistic insight and a rationale for ongoing and future clinical trials that combine TTFields with immunotherapy.
Introduction
Tumor-treating fields (TTFields) are a locoregional antitumoral treatment that uses alternating electric fields tuned to target proliferating cells. TTFields exert an electric torque on highly polar proteins cardinal to the mitotic phase of cell division, thereby disrupting mitotic spindle formation. These physical interactions result in inhibiting proliferation and inducing cancer cell death (1).
Optune is a medical TTFields delivery device that has demonstrated antimitotic effects in preclinical and clinical research on several solid tumors (2). Optune has been U.S. Food and Drug Administration–approved for treatment of recurrent and newly diagnosed glioblastoma (GBM), in which, when delivered concurrently with standard chemoradiation therapy, it was shown to significantly prolong progression-free and overall survival while maintaining good quality of life (3). It was recently approved also for the treatment of malignant pleural mesothelioma (4), and it is currently being tested in phase III clinical trials for ovarian (NCT03606590), pancreatic (NCT03377491), liver (NCT03606590), gastric (NCT04281576), and non–small-cell lung cancer (NCT02973789). Notwithstanding, even following a survival extension of 4.8 mo, the median survival of newly diagnosed GBM patients remains only 20.8 mo, highlighting the need for more effective multimodal treatments (5).
Cancer immunotherapy is a central arm in battling malignancies (6). Responses to immunotherapy frequently positively correlate with tumor neoantigen load, PDL1 expression and the abundance of tumor-infiltrating T lymphocytes (TILs) as found in immunologically “hot” tumors (7). Unfortunately, gliomas have low neoantigen load and are considered immunologically “cold” (8, 9), with the majority of patients failing to respond to immunotherapy (9, 10).
The immune system may play an important role in mediating the clinical effects of TTFields. TTFields was shown to reduce distant metastatic spread to the lungs in a rabbit renal cancer model and to enhance the peri- and intratumoral infiltration of CD4+, CD8+ T cells (11). It was shown to enhance antitumor efficacy when combined with anti-PD1 in mouse tumor models, plausibly by their induction of immunogenic cell death (ICD) in tumor cells (12–14). In a retrospective study in GBM patients, the overall survival benefit of TTFields was found to correlate with higher blood T cell counts and with low (<4.1 mg) or no administration of immunosuppressive dexamethasone (15). Of note, being a locoregional physical treatment, TTFields is likely to cause minimal or no systemic immunosuppression. Several early phase trials combining TTFields with checkpoint inhibitors (CPIs) (NCT02973789, NCT03405792, NCT03903640, NCT03430791) or other immunotherapeutics (NCT03223103) are underway. However, the direct effects of TTFields on T cells, the main drivers of antitumoral responses are unknown. Understanding the extent to which T cells generate an immune response under the locoregional effects of TTFields is therefore of key mechanistic importance to current and future trials combining immunotherapy with TTFields.
In this study, we employed the inovitro system, enabling application of TTFields on cultured cells, to evaluate the effect of TTFields on T cells. We examined TTFields’ effects on T cell proliferation, viability, and select pivotal antitumoral functions using healthy donor PBMCs and TILs from resected GBM samples. We evaluated direct cytotoxicity using a CAR-T cell–based system. We also performed immunohistochemical (IHC) analysis and comparative transcriptomic evaluation on GBM samples obtained from patients before and after TTFields treatment to assess TTFields’ treatment effects on T cell infiltration and on broad immune gene expression.
Materials and Methods
Cells, tissue culture, and patients’/donors’ samples
PBMC were produced from leukocyte-enriched fractions from healthy blood donations obtained from the Israeli blood bank. All tissue culture was performed in X-VIVO15 serum-free defined medium (Lonza Pharma and Biotech, Basel, Switzerland) supplemented with 1:5000 Benzonase (Novagen, Billerica, MA) to gently dissociate DNA and RNA split from dead cells.
Human GBM viable single-cell suspensions were produced by dissociating tumor samples using neutral protease from Clostridium histolyticum (AMSBio, Abingdon, UK) (16). The dissociated tumor cells were freshly assayed. Healthy donor blood and patients’ GBM samples were received under institutional ethical approval (0408-10-TLV). Samples were received from patients signing an informed consent.
CAG multiple myeloma cells were obtained from the American Type Culture Collection (Manassas, VA) and maintained in RPMI 1640 (Life Technologies, Grand Island, NY), 10% FBS, 1% sodium pyruvate, 1% glutamine and penicillin/streptomycin (Biological Industries, Beit Haemek, Israel). HER2-expressing CAG cells were generated by transfection (jetPRIME; Polyplus-transfection, llkirch-Graffenstaden, France) with human HER2 and G418 resistance genes. HER2-CAG cells were cultured in the CAG medium containing 0.5 mg/ml G418 (Life Technologies). T cells expressing an anti-HER2 CAR were generated by retroviral transduction of healthy donor PBMC samples with a vector encoding the N29 anti-Her2 CAR and GFP [see (17)].
Institutional ethical approval was received for use of blood donations and GBM samples (0408-10-TLV). Samples were received from patients signing an informed consent. All authors all coauthors approve this submission. All data and materials are available upon request.
TTFields application in vitro
Cell culturing under TTFields conditions was conducted using the inovitro TTFields system (18) according to manufacturer's instructions unless described otherwise. Cells were seeded in the inovitro culture dishes and incubated either on baseplates generating an electric field within a heating/cooling incubator (Binder, Tuttlingen, Germany) or in a standard incubator. In vitro application of TTFields accelerated the evaporation of water from the media, thus distilled water was replenished daily in all plates. The electric parameters used in all assays matched those used in the clinic. Field frequency was set to 200 kHz (and also to 150 kHz in Supplemental Fig. 3), and field intensity was set at 2–2.5 V/cm (peak-to-peak), which is within the range found to optimally inhibit GBM cell growth (19) and as used in GBM patients.
Flow cytometry
Multicolor flow cytometric sample analysis was performed using a Canto-II Flow Cytometer (BD Biosciences, Franklin Lakes, NJ). Staining was performed in the dark and at room temperature. Abs used were as follows: CD3, CD4, and CD19 from eBioscience (San Diego, CA), CD107a, CD14, PD1, and IFN-γ from BioLegend (San Diego, CA) and CD8a from BD Biosciences. Data analysis was performed using FlowJo (Ashland, OR).
For CFSE staining, PBMC or GBM cell suspensions (1 × 107 to 5 × 107 cells per milliliter) were stained with CFSE (Thermo Fisher Scientific, Waltham, MA) 2 µM CFSE and 1.25% FCS 10 min, room temperature. The cells were then washed twice with PBS, 2.5% FCS, counted using trypan blue (Sigma-Aldrich, St. Louis, MO) and used for subsequent assays.
Viability assay
CFSE-stained PBMCs (3 × 106 cells per dish) were seeded in inovitro dishes, and either stimulated with 1% PHA (Life Technologies) or left untreated. The cells were incubated for 3.5 d under either TTFields or standard culture conditions. Cells were then harvested, stained for viability using violet fixable LIVE/DEAD amine viability dye (ViViD-L34955; Thermo Fisher Scientific) and then extracellularly for CD8a, CD14, and CD19. CD3 and CD4 were stained intracellularly using the Fix/Perm Kit (BD Biosciences) (20) as T cell activation drives their internalization. Cells were fixed with 1% formaldehyde (Electron Microscopy Sciences, Hatfield, PA) and analyzed by flow cytometry.
To enable inclusion of dead cells in this assay, autofluorescent debris (particles with medium-to-high signals on all channels that can be mistaken for dead cells) were gated out based on a higher CFSE signal and a lower side-scatter compared with T cells. Viability rates were calculated as live/total cells. The cells were categorized as live (ViViDlow) or dead (ViViDhigh), and as proliferating (CFSElow) or nonproliferating (CFSEhigh).
T cell activation and polyfunctionality assays
CFSE-stained PBMC (3 × 106) or dissociated single cells from GBM samples were cultured as described above. Five hours before harvesting, the cultures was supplemented with fluorochrome-conjugated CD107a Ab, 0.07% Golgistop (BD Biosciences) and 0.1% brefeldin A (Sigma-Aldrich). Cells were collected, stained for viability, then stained extracellularly for CD8a, CD14, CD19, and PD1, and intracellularly for CD3, CD4, and IFN-γ.
In vitro cytotoxicity assay
Target cells (CAG-HER2 or CAG) and effector anti-HER2 human CAR-T cells were cocultured for 8 h at varying effector/target ratios in standard or TTFields culture conditions. Dishes were supplemented with 1:5000 Benzonase to reduce split DNA/RNA-related clumping. Cultures were stained with CD3, CD138 (a myeloma marker), and propidium iodide, staining dead cells (Sigma-Aldrich), and then analyzed by flow cytometry. Cell death rates in “target-only” were subtracted from those in matching “effector/target.” The unspecific killing of parental CAG (non-HER2) was subtracted from its corresponding CAG-Her2 sample.
Statistical analysis
In vitro quantitative data are presented as mean ± SE. Statistical significance was determined by paired, two-tailed Student’s t test and noted where significant (*p < 0.05, **p < 0.005). All in vitro experiments were repeated at least three times.
Tissue immunohistochemistry and cell frequency analysis
Tumor samples were obtained from four GBM patients before and after TTFields treatment (patient data in Supplemental Table STI). Tissue slides were immunohistochemically stained by hospital pathology services for CD3, CD4, and CD8 by the DAB-staining procedure. Tissue slides were scanned and then analyzed with the Aperio Imagescope program. Stained (i.e., brown-colored) nucleated cells (i.e., blue center) were counted in 4–6 representative 0.4 mm2 fields in each slide. The cell counts were then averaged and normalized to cells per millimeter squared.
Tissue processing for RNA sequencing and library construction
Paired tumor tissues from 12 newly diagnosed GBM patients were obtained before and after treatment with standard chemoradiation (six controls) or with TTFields + chemoradiation (six TTFields). Processing and RNA extraction were performed with the PerfectPure RNA Tissue Kit (5 prime, Hilden, Germany). Illumina TruSeq RNA Library Prep v2 was used for sample preparation, generating mRNA-based libraries from total RNA input. Indexed samples were sequenced, in single-read mode, using the Illumina HiSeq 2500.
Transcriptomic analysis
RNA-sequencing (RNA-Seq) data were analyzed with DESeq2 software (21). Differential gene expression was compared on a broad list of 712 immune-related genes using the reads per kilobase million values. Statistical analysis was performed using the negative binomial generalized linear model. Differences in expression before and after treatments were calculated per patient, and then averaged as net treatment effect per group. Fold-changes of net treatment effects (average of TTFields group over average of control group) were defined as significant when above 2 or below 0.5, with a Benjamini–Hochberg multiple comparison–corrected p < 0.1. The protumoral/antitumoral/mixed activity of each significantly altered gene was determined using published literature. The list of 712 genes related to immune activity was compiled using the Nanostring “nCounter PanCancer immune profiling panel” and the Thermo Fisher Scientific “Oncomine immune response” gene lists and the general literature. The list surveys all chemokines, cytokines and their receptors, transcription factors, immune checkpoint–related molecules, and other immune-related genes. The significance of the association (contingency) between protumoral to antitumoral genes (defined by the literature) significantly up or downregulated by TTFields- or by standardly treated patients was calculated using the Fisher exact method.
Results
TTFields reduce the viability of proliferating T cells
TTFields exert a cytostatic and cytotoxic effect on proliferating tumor cells. To evaluate T cell viability under TTFields, PBMCs were stained with CFSE and stimulated with PHA to induce proliferation. PHA was selected because of its TCR-dependent activation that simulates physiological activation requiring signal transduction, organelle redistribution, and actin cytoskeletal dynamics (22). Note that PHA stimulation is robust, leading to some activation-induced cell death. This affects both cell numbers and the viability rates. Cell doublets, B cells, monocytes, and autofluorescent debris were gated out and then CD4+ and CD8+ T cells were monitored for their proliferation status and viability (Fig. 1A).
Effects of TTFields on the viability of peripheral blood T cells. CFSE-stained healthy donor PBMC were treated with PHA or left untreated and incubated under TTFields or standard conditions for 3.5 d. (A) Representative flow cytometric gating strategy for PHA-stimulated cells (top three rows) and partial gating for nonstimulated T cells. CD4+8– Th cells or CD8+4– CTL were divided into four groups according to viability (ViViD) and proliferation (CFSE) status. (B) Relative T cell numbers and viability rates in PHA-stimulated and nonstimulated samples, according to proliferative status. Data were normalized by setting the value of live, nonproliferating cells in control samples to one in each repeat. Stacked bars show normalized numbers of live cells. Viability rates (no. live / no. total) appear under each bar. Graphs show the mean of eight independent repeats using PBMC from four different donors tested during four separate experiments (two samples tested per donor in separate experiments). *p < 0.05.
Effects of TTFields on the viability of peripheral blood T cells. CFSE-stained healthy donor PBMC were treated with PHA or left untreated and incubated under TTFields or standard conditions for 3.5 d. (A) Representative flow cytometric gating strategy for PHA-stimulated cells (top three rows) and partial gating for nonstimulated T cells. CD4+8– Th cells or CD8+4– CTL were divided into four groups according to viability (ViViD) and proliferation (CFSE) status. (B) Relative T cell numbers and viability rates in PHA-stimulated and nonstimulated samples, according to proliferative status. Data were normalized by setting the value of live, nonproliferating cells in control samples to one in each repeat. Stacked bars show normalized numbers of live cells. Viability rates (no. live / no. total) appear under each bar. Graphs show the mean of eight independent repeats using PBMC from four different donors tested during four separate experiments (two samples tested per donor in separate experiments). *p < 0.05.
No proliferation was detected in the PHA-nonstimulated samples. The number of cells in the nonstimulated TTFields-treated samples was lower by ∼15% to standardly incubated controls (Fig. 1B, left, p = 0.01) with no change in viability rates between groups (both 98%). This suggests that TTFields have a small to negligible effect on the number and the viability of nonstimulated T cells.
PHA stimulation drove some of the T cells to proliferate and affected both cell numbers and viability. The numbers of PHA-stimulated nonproliferating (CFSEhigh) T cells in the TTFields-cultured samples was ∼35% lower than controls (p = 0.015 and p = 0.03 for CD4+ and CD8+ cells, respectively); however, their viability rates remained unaltered (Fig. 1B, right). In the same samples, the number of PHA-stimulated, proliferating (CFSElow) T cells in the TTFields-cultured samples was lower by ∼75% to controls (p = 0.009 and p = 0.01 for CD4+ and CD8+, respectively) along with a decline in the viability rates, especially for the CD8+ T cells (52% in the controls versus 40% in TTFields). These results align with TTFields’ mechanism of action, i.e., marked inhibition and cytotoxic effect on proliferating cells with low impact on nonproliferating cells (23).
Key antitumoral functions in PHA-stimulated blood T cells are unaffected by TTFields
Many immune-related cellular processes such as secretion, degranulation, and surface molecule mobilization rely on cytoskeletal dynamics or on vesicular transport that may conceivably be perturbed by TTFields. To investigate whether key antitumoral functions are affected by TTFields, we simultaneously monitored proliferation (CFSE), activation/exhaustion (PD1), IFN-γ secretion and degranulation of cytotoxic granules [CD107a (24)] at single-cell resolution.
Gating identified CD4+ and CD8+ T cells, alongside CD3+CD14–CD19–CD4–CD8– cells, the majority of which are γδ-T cells (not shown) (Fig. 2A). We found that TTFields reduced to approximately half (but did not abolish) the fraction of proliferating CD4+ or CD8+ T cells. Importantly all other evaluated antitumoral functions were virtually unchanged between TTFields and standard PBMC cultures (Fig. 2B).
Effects of TTFields on the activation of blood-borne T cells. CFSE-stained donor PBMC were treated with PHA or left untreated and incubated under TTFields or standard conditions for 3.5 d. (A) Representative gating strategy for CD3+8+4– (CTL) evaluated for surface expression of PD1 and CD107a, for CFSE dilution and for IFN-γ production. Similar gating was performed for CD3+4+8– (Th) and CD3+4–8– T cells (the majority of which are γδ-T cells in healthy donor blood) (B–D) Net activation per function was calculated by subtracting the percent of positive cells in nonstimulated samples from the percent in matched PHA-stimulated samples. Graphs show the mean of 9–11 independent repeats using PBMC from six different donors tested over seven separate experiments. (B) Summary of functional responses grouped per single function. (C) Summary of polyfunctional analysis. Cells were divided into 15 mutually exclusive functional groups according to expression of one or more functions. Net percentage of polyfunctional-positive cells is presented. (D) Analysis of T cell viability in activated cells. Total number of live T cells in all samples (I) and in PHA-stimulated samples only (II). Graphs (III) and (IV) Cell numbers of activated nonproliferating cells (negative for proliferation but positive one of more other functions) in CD4+ and CD8+ T cells, respectively. *p < 0.05, **p < 0.005.
Effects of TTFields on the activation of blood-borne T cells. CFSE-stained donor PBMC were treated with PHA or left untreated and incubated under TTFields or standard conditions for 3.5 d. (A) Representative gating strategy for CD3+8+4– (CTL) evaluated for surface expression of PD1 and CD107a, for CFSE dilution and for IFN-γ production. Similar gating was performed for CD3+4+8– (Th) and CD3+4–8– T cells (the majority of which are γδ-T cells in healthy donor blood) (B–D) Net activation per function was calculated by subtracting the percent of positive cells in nonstimulated samples from the percent in matched PHA-stimulated samples. Graphs show the mean of 9–11 independent repeats using PBMC from six different donors tested over seven separate experiments. (B) Summary of functional responses grouped per single function. (C) Summary of polyfunctional analysis. Cells were divided into 15 mutually exclusive functional groups according to expression of one or more functions. Net percentage of polyfunctional-positive cells is presented. (D) Analysis of T cell viability in activated cells. Total number of live T cells in all samples (I) and in PHA-stimulated samples only (II). Graphs (III) and (IV) Cell numbers of activated nonproliferating cells (negative for proliferation but positive one of more other functions) in CD4+ and CD8+ T cells, respectively. *p < 0.05, **p < 0.005.
An key feature associated with effective responses to pathogens or tumors is the ability of individual activated T cells to respond by several concurrent functions, also known as polyfunctionality (25). A polyfunctionality analysis of T cells was performed on the dataset presented in (Fig. 2B showing that, with the exception of proliferation, activated CD4+ and CD8+ T cells exhibited similar polyfunctional patterns under TTFields as under standard culture conditions (Fig. 2C). The data illustrate shifts of T cells from parallel polyfunctional groups that differed only in proliferation (e.g., from proliferating PD1+CD107+ to nonproliferating PD1+CD107+ cells). This may reflect relative shifts in fractions where part of the proliferating cells die, increasing the relative fraction of similarly activated but nonproliferating cells. These findings also demonstrate that proliferation is independent of other key antitumoral T cell functions, as previously shown by others (26). Although CD107a is not a standard Th marker, activated CD4+ T cells have been shown to upregulate CD107a and secrete granzyme B (27). CD4+CD107+ T cells were reported to exhibit enhanced survival compared with their CD4+CD107a– counterparts (28).
The above multiparametric dataset also enabled absolute enumeration of live cells according to cell type (CD4/8) and to proliferative status. Similar to the findings in (Fig. 1, the number of live T cells found in nonstimulated samples cultured under TTFields was 85% that of controls (nonsignificant, p = 0.48) (Fig. 2DI). As in (Fig 1B, the PHA-stimulated cultures contained significantly fewer T cells (p = 0.024) compared with standard-cultured conditions (Fig. 2DII). Importantly, there was no significant difference in the number of cells in these cultures when counting only the T cells which exhibited one or more activation parameters (“activated”) other than proliferation (Fig. 2DIII, 2DIV, CD4+p = 0.85, CD8+p = 0.37). These results indicate that TTFields neither reduce the viability/number nor the functional responses in activated nonproliferating T cells. Note that the group of PHA-stimulated, activated but nonproliferating cells that was evaluated in (Fig. 2D is a subgroup of the PHA-stimulated, nonproliferating cells evaluated in (Fig. 1B.
Key antitumoral functions in GBM TILs are unaffected by TTFields
When compared with blood-borne T cells, TILs frequently distribute differently between memory/effector subsets and exhibit different effector and proliferation capacities (26, 29). The GBM microenvironment modifies T cell phenotypes (6), and TILs may thus respond differently than PBMC to TTFields. To examine the function of GBM TILs under TTFields conditions, viably dissociated (16) human GBM samples were incubated under standard or TTFields conditions (gating in Supplemental Fig. 1A). As with PBMCs, TIL proliferation was inhibited but not fully abolished under TTFields, whereas all other evaluated antitumoral T cell functions were unaffected (Fig. 3A). Upon PHA stimulation, TILs exhibited comparable polyfunctional responses under TTFields as under standard culture conditions (Fig. 3B).
Effect of TTFields on activation of human GBM TILs. Fresh GBM samples were enzymatically dissociated to viable single cells, stained with CFSE, stimulated by PHA, or left untreated, and incubated under TTFields or standard conditions for 3.5 d. Cells were harvested and stained as described in (Fig. 2. See gating in Supplemental Fig. 1A. PBMC samples were treated identically and used as guides for gating of data (not shown). Net activation was calculated by subtracting percent of positive cells in nonstimulated samples from percent in matched PHA-stimulated sample. Graphs show means and SE of three independent repeats with samples from three different GBM patients. (A) Summary of specific functional responses in CD4+ and CD8+ T cells. (B) Summary of polyfunctional analysis. Cells were divided into 15 mutually exclusive functional groups according to expression of one or more functions. Net percent of polyfunctional-positive cells is presented. (C) Relative rates and polyfunctionality of PD1+ TILs (TASTs) in the nonstimulated samples. The PD1+ T cells were divided into eight distinct polyfunctional groups according to their expression of one or more functions. *p < 0.05, **p < 0.005.
Effect of TTFields on activation of human GBM TILs. Fresh GBM samples were enzymatically dissociated to viable single cells, stained with CFSE, stimulated by PHA, or left untreated, and incubated under TTFields or standard conditions for 3.5 d. Cells were harvested and stained as described in (Fig. 2. See gating in Supplemental Fig. 1A. PBMC samples were treated identically and used as guides for gating of data (not shown). Net activation was calculated by subtracting percent of positive cells in nonstimulated samples from percent in matched PHA-stimulated sample. Graphs show means and SE of three independent repeats with samples from three different GBM patients. (A) Summary of specific functional responses in CD4+ and CD8+ T cells. (B) Summary of polyfunctional analysis. Cells were divided into 15 mutually exclusive functional groups according to expression of one or more functions. Net percent of polyfunctional-positive cells is presented. (C) Relative rates and polyfunctionality of PD1+ TILs (TASTs) in the nonstimulated samples. The PD1+ T cells were divided into eight distinct polyfunctional groups according to their expression of one or more functions. *p < 0.05, **p < 0.005.
Because the considerable heterogeneity in TIL frequencies among the different GBM samples, comparative cell enumeration (as in (Fig. 2D) was not applicable. A more limited analysis of the relative fractions of live TILs was performed (Supplemental Fig. 1B) revealing no significant differences between control and TTFields-cultured GBM samples in the fraction of activated TILs responding to PHA stimulation by one or more functions other than proliferation (as in (Fig. 2D). In the current study, stimulation of tumor-derived T cells under TTFields conditions was not performed under in vivo conditions (e.g., using rodent models), but rather under fresh, ex vivo conditions.
Notably, TIL enzymatic dissociation from tumor samples could potentially affect the functionality of TILs. Notwithstanding, the gentle dissociation methodology used in this study (16) together with the strong functional responses observed in the PHA-stimulated conditions suggest that the assayed TILs remained functionally intact.
Tumor Ag–specific T cells exposed to TTFields exhibit unaltered viability and function
TTFields are not expected to negatively affect tumor Ag–specific T cell (TAST) proliferation in tumor-draining lymph nodes located outside the treated field. However, TASTs may also proliferate within tumors (30) where TTFields may have an effect. To date, there are no unequivocal methods to identify which TILs constitute TASTs. However, Ahmadzadeh et al. (31) and Donia et al. (32) independently showed that the overwhelming majority of TASTs expresses PD1 on their surface. We relied on this characterization to evaluate the effect of TTFields on intratumoral TASTs by analyzing PD1+ TILs in GBM cultures incubated without PHA stimulation under standard or TTFields conditions. Note that the rate of PD1+ T cells within GBM TILs was considerably higher than the rate of PD1+ T cells in the matching patients’ PBMC (25% versus 3% in CD8+, and 27% versus 7% in CD4+ T cells, respectively; Supplemental Fig. 1A). This reflects local activation and/or exhaustion of some TILs (31). Importantly, we found that nonstimulated PD1+ TASTs under TTFields or standard culture conditions exhibited similar baseline mono- and polyfunctionally (Fig. 3C).
Cytotoxic capacity of CAR-T cells is unaffected by TTFields
T cell–mediated cytotoxicity depends on the formation of a functional cytotoxic synapse. This requires extensive microtubule and centrosome realignment, which may plausibly be disrupted by TTFields (33). Given the importance cytotoxicity for effective immunotherapy and that our above findings on degranulation (CD107a) relate to a surrogate cytotoxicity marker, we directly assessed cytotoxicity under TTFields conditions using a CAR-T cell–based assay. CAR-T cells rely on the same intracellular machinery and processes as nonengineered T cells, supporting our choice of assay (34). In consideration that CAR-T cells are currently under clinical evaluation for GBM and other solid tumors (35), this assay also provides a general evaluation of possible compatibility of CAR-T cells with TTFields.
Human anti-HER2 CAR-T cells were cocultured with target cells, either parental (unaltered) CAG human myeloma cell line or CAG, which ectopically expresses HER2 (Fig. 4A). Target cell killing by HER2-specific CAR-T cells was unaffected by TTFields (Fig. 4B). Alternative gating of the same dataset examining the effector cells confirmed that TTFields had no effect on the viability of CAR-T effectors during the course of the assay (not shown). Taken together, these results demonstrate that TTFields do not interfere with T cell–mediated cytotoxicity.
Effect of TTFields on direct cytotoxicity of CAR-T cells. Effector human anti-HER2 CAR-T cells and target cells expressing HER2 (CAG-Her2) or not expressing it (CAG) were cocultured under TTFields or control conditions for 8 h. Target cells were also cultured alone to determine background cell death. (A) Top row, Representative gating strategy. Bottom row, Dot plots displaying the viability for all four effector/target ratios tested. (B) Specific killing curves. The dead cell fraction in “target only” samples was subtracted from the concordant samples (lighter versus darker lines, respectively). Net specific killing (black line) was calculated by subtracting the fraction of killed cells in the CAG samples from that in the CAG-HER2 samples. Representative results (CAR transduction efficiency ≈ 35%) from one assay are shown. Similar results were observed in three independent repeats. Averages not presented because of markedly varying killing rates among donors, stemming from differences in and CAR transduction efficiencies and donor cells cytotoxic potency.
Effect of TTFields on direct cytotoxicity of CAR-T cells. Effector human anti-HER2 CAR-T cells and target cells expressing HER2 (CAG-Her2) or not expressing it (CAG) were cocultured under TTFields or control conditions for 8 h. Target cells were also cultured alone to determine background cell death. (A) Top row, Representative gating strategy. Bottom row, Dot plots displaying the viability for all four effector/target ratios tested. (B) Specific killing curves. The dead cell fraction in “target only” samples was subtracted from the concordant samples (lighter versus darker lines, respectively). Net specific killing (black line) was calculated by subtracting the fraction of killed cells in the CAG samples from that in the CAG-HER2 samples. Representative results (CAR transduction efficiency ≈ 35%) from one assay are shown. Similar results were observed in three independent repeats. Averages not presented because of markedly varying killing rates among donors, stemming from differences in and CAR transduction efficiencies and donor cells cytotoxic potency.
Evaluating the clinical compatibility of TTFields with antitumoral immunity
To support our in vitro findings, we sought to examine the effect of TTFields on the immune response in clinical settings. First, we compared T cell infiltration rates in GBM samples resected before and after TTFields treatment from four patients by performing IHC staining for CD3, CD4, or CD8 (patient details in Supplemental Table STI). Although following TTFields treatment one patient’s CD3+ and CD8+ TIL counts were slightly reduced, the infiltration rates in the other monitored patients were either unchanged (one patient) or strongly increased (two patients) (Fig. 5A). The representative images from one patient in (Fig. 5B demonstrate that TTFields may accommodate, in some cases, large increase in TIL density. Obtaining paired before-after tissue blocks of samples was confined to a small number of tissue blocks we were able to recover and was limited only to patients receiving chemoradiation + TTFields, not to those receiving chemoradiation only. Nonetheless, the data demonstrate that TTFields application in patients does not preclude T cells infiltration into tumors and that in some cases, TIL frequencies had significantly increased under treatment.
T cell infiltration rates and immunity-related gene expression in GBM tumors treated with TTFields therapy. (A and B) GBM tumor samples were obtained from four patients before and after treatment with TTFields combined with standard chemoradiation. Samples were embedded in paraffin blocks, sliced, and stained by H&E for nuclear visualization and by IHC for CD3, CD4, and CD8. (A) Cells were counted in 4–6 representative fields from each slide at original magnification ×20 and cell count per millimeter squared was calculated. Graphs show mean cell densities (cells per millimeter squared) of CD3+, CD4+, and CD8+ cells for each patient before and after therapy. (B) Representative IHC images of tissue section from patient 4*. Positive cells are stained brown. (C) GBM samples were obtained before and after a treatment from six patients treated by TTFields + chemoradiation and six patients treated by standard chemoradiation only (controls). Gene expression analysis was performed by RNA-Seq. The negative binomial generalized linear model was used to analyze expression following treatment. Net treatment effect was calculated per patient by subtracting a gene’s expression level at pretreatment from its expression level posttreatment. Differences between treatments were calculated as fold change between average net effect in the TTFields group versus control group. A list of 712 immune activity-related genes was evaluated. A significant effect of TTFields on expression was defined as an average fold change of either >2 or <0.5, with p < 0.1. (C) A list of the genes exhibiting significantly altered expression. Each gene was designated as having primarily pro-/anti-/mixed-tumoral activity based on literature review.
T cell infiltration rates and immunity-related gene expression in GBM tumors treated with TTFields therapy. (A and B) GBM tumor samples were obtained from four patients before and after treatment with TTFields combined with standard chemoradiation. Samples were embedded in paraffin blocks, sliced, and stained by H&E for nuclear visualization and by IHC for CD3, CD4, and CD8. (A) Cells were counted in 4–6 representative fields from each slide at original magnification ×20 and cell count per millimeter squared was calculated. Graphs show mean cell densities (cells per millimeter squared) of CD3+, CD4+, and CD8+ cells for each patient before and after therapy. (B) Representative IHC images of tissue section from patient 4*. Positive cells are stained brown. (C) GBM samples were obtained before and after a treatment from six patients treated by TTFields + chemoradiation and six patients treated by standard chemoradiation only (controls). Gene expression analysis was performed by RNA-Seq. The negative binomial generalized linear model was used to analyze expression following treatment. Net treatment effect was calculated per patient by subtracting a gene’s expression level at pretreatment from its expression level posttreatment. Differences between treatments were calculated as fold change between average net effect in the TTFields group versus control group. A list of 712 immune activity-related genes was evaluated. A significant effect of TTFields on expression was defined as an average fold change of either >2 or <0.5, with p < 0.1. (C) A list of the genes exhibiting significantly altered expression. Each gene was designated as having primarily pro-/anti-/mixed-tumoral activity based on literature review.
We next evaluated how TTFields modulate immune activity within GBM tissues by transcriptomic analysis. GBM samples were obtained before and after treatment of six patients treated by standard chemoradiation protocol (controls) and six patients treated by chemoradiation and TTFields (patient details in Supplemental Table STI). Differential gene expression analyses (Supplemental Fig. 2) demonstrated similar T cell infiltration trends as found in the IHC (Fig. 5). Specifically, the net post- minus pretreatment expression levels of the CD3, CD4, and CD8 genes (representative of T cell infiltration) were slightly higher in the TTFields group than in the control group, both per individual patients and per the averaged group. Although transcriptomic signatures are not as definitive as IHC to evaluate infiltration, it nevertheless supports our IHC data using an additional set of 12 before/after samples.
We next evaluated the effect of TTFields on the expression of a comprehensive set of 712 immune-related genes. The table in (Fig. 5C summarizes all genes within this gene set found to be significantly up- or downregulated in either group. Genes were classified as having protumoral, antitumoral, mixed, or unknown roles based on published literature.
Overall, the data demonstrated a significant shift in the immune-related gene expression from protumoral to antitumoral signature following TTFields + chemoradiation treatment compared with treatment by chemoradiation alone (p = 0.026 Fisher exact). Notably, the antitumoral genes that were upregulated by TTFields’ included granulysin, a key cytotoxic molecule, NKG2D (an NK activator receptor associated with TTFields-induced NK killing of glioma cells) (36) and t-bet (a hallmark Th1 transcription factor). TTFields’ treatment downregulated protumoral genes including IL18 (a proinflammatory cytokine), CXCL14 (a chemokine driving a suppressive tumor microenvironment), PDL2 (related to T cell exhaustion), and IL4 and IL17RB, which polarize Th to ineffective or protumoral functions.
Discussion
Knowing the effect of TTFields on T cells is critical when considering the pivotal role of T cells in antitumoral immune responses. By evaluating key antitumoral T cell functions ex vivo, our findings (Figs. 1, 2, and 3) indicate that TTFields operated at established therapy-relevant conditions do not hinder key T cell functions aside from proliferation. Our findings hold not only to cells derived from healthy donor blood but also to GBM TILs. Cytotoxicity is also not impeded by TTFields as shown by a CAR-T cell–based assay. To our knowledge, our assay also represents the first time that CAR-T cell functionality was examined in the context of TTFields.
A differential response of tumor cell lines to various TTFields frequencies has been previously demonstrated (18). Matching optimal frequencies to specific cancers is a cornerstone of TTFields clinical use, as demonstrated by the recent approval of TTFields treatment for mesothelioma employing 150 kHz frequency rather than 200 kHz as used for GBM. An in-depth study of T cell functionality under the entire clinically relevant frequency spectrum of TTFields is not within the scope of this study. Notwithstanding, we did test how the phenomena observed in 200 kHz apply to additional frequencies in the range of 100–500 kHz (Supplemental Fig. 3A, 3B). The strongest negative effects on proliferating T cells were observed at 200 kHz followed by 300 kHz. TTFields’ frequencies of 100 kHz and 150 kHz showed no negative effects on T cell viability or functionality.
Although there is no evidence of a reduction in T cell counts within tumors or in peripheral blood in numerous TTFields preclinical or clinical studies, the inhibitory impact of TTFields on the capacity of intratumoral T cell to proliferate could be of concern. It is not yet clear if, and to what extent, intratumoral TASTs proliferate within GBM and other solid malignancies. Expanded TCRs clonotypes found among TIL populations in GBM and other tumors indicate that proliferation of TASTs does occur (37) but not where it had occurred. Murine studies have yielded conflicting findings, some having shown that brain-Ag–specific T cells do not proliferate within the brain (38), whereas others demonstrating that the brain microenvironment may support in situ proliferation under some circumstances (39).
Of note is that brain- or tumor-specific T cell priming, and proliferation were shown to occur in the meningeal lymphatics (40) and within the deep cervical lymph nodes (6), found outside the TTFields-affected area. The case of mesothelioma, in which TTFields have been approved, or lung and liver cancers, in which TTFields are currently in advanced clinical phases, might be different. In these cases, some tumor-draining lymph nodes may be found within the treated field. Notwithstanding, the data presented in Supplemental Fig. 3 demonstrate that 150 kHz (the frequency used to treat said tumors) had no negative effects on the number, functionality, or viability of T cells.
Importantly, several studies have indicated that only peripheral T cells, and not intratumoral ones, bear sustained proliferative capabilities, and that T cell activation and proliferation must occur within peripheral lymphoid organs to achieve tumor eradication (30, 41). A recent study provided evidence of clonotypic expansion of effector-like T cells in four different tumor types, both within the tumor as well as in adjacent normal tissue and within peripheral blood (42). Patients exhibiting a wide clonotypic expansion pattern were those most likely to respond to anti-PDL1 therapy. The authors suggested that tumor-Ag–specific TILs are replenished by nonexhausted T cells arriving from outside the tumor. If this applies also to brain tumors, it is reasonable to expect that the effect of TTFields on the proliferation of intratumoral T cells should have minor clinically meaningful implications. The effects that TTFields were found to have on proliferating T cells suggest that appropriate consideration of TTFields parameters (i.e., field frequency, intensity, and administration protocols) should be taken to balance between maximizing tumor cell killing and minimizing the disruption of T cell function.
In recent years there have been a surge of scientific data demonstrating that beyond their documented immunosuppressive effects, both temozolomide (43) and irradiation (44) can exert potent immunostimulatory effects either alone or in combination (43). These effects are mediated by temozolomide through the enhancement of mortal levels of autophagy (43) and by irradiation through the induction of ICD and by the production of type I IFNs, downstream of CGAS signaling driven by cytosolic dsDNA (44).
Whereas the ICD-related effects of TTFields were demonstrated elsewhere (12–14), to our knowledge, our study is the first to report comparison of patients’ RNA-Seq data obtained from pre-postchemoradiation treatment versus prepost-TTFields + chemoradiation treatment. This comparison enables to dissect the relative immune contribution of TTFields. Our transcriptional analysis focused on a large set of immune-related genes and demonstrated a shift from a protumoral Th2 or a Th17-related profile, previously associated with unfavorable prognosis in GBM (45), to a more antitumoral Th1-associated profile (46). Regarding T cells infiltration to tumors, The IHC data from four patients that was complemented by RNA-Seq data from 12 additional patients, demonstrated that TTFields do not preclude nor reduce the accumulation of T cells within GBMs.
The in vitro mechanistic data in this study focused on evaluating potentially deleterious effects of TTFields on the T cell level, whereas the clinical part demonstrated potentially beneficial effects of TTFields on the patient level. TTFields’ potential to clinically synergize with immunotherapy is beyond the scope of this work and is currently addressed by several clinical trials. However, TTFields ability to operate alongside, and in some cases, synergize with antitumoral responses has been demonstrated when it was shown to enhance T cell infiltration in various tumor models (11, 14). Voloshin et al. have recently provided direct evidence of TTFields inducing ICD in multiple murine tumor cell lines. The authors demonstrated synergism in vivo in two murine tumor models between treatment with a PD1 inhibitor and TTFields application (14).
Correlative clinical data from GBM patients revealed that the survival benefit from TTFields correlates with higher blood T cell counts and with low or no administration of dexamethasone (15). Although these data do not reveal cause and effect, they concur with both the mouse in vivo data (14) and with our human data, substantiating a possible contribution of an intact immune compartment to the clinical benefits provided by TTFields.
Given our findings and the nature of TTFields, several approaches for combining this modality with immunotherapy are conceivable. Combining TTFields with CPIs may couple tumor cell death by ICD (14) (TTFields) with attenuation of T cell exhaustion (e.g., PD1), or expanded breadth of responding T cells (CTLA4). This approach bears some similarity to the combination of CPIs with ICD-inducing oncolytic viruses (47) or radiotherapy (48, 49).
The first clinical studies combining treatment with alternating electrical-fields and various immune-modulating treatments are already underway. As preclinical and clinical data demonstrate the significant overall antitumor immunity function of TTFields, the mechanistic and clinical data provided in this study strengthens these clinical studies’ rationale and provides important knowledge required to interpret their results.
Disclosures
Z.R. is serving as a consultant for and holding stock options in Novocure. The other authors have no financial conflicts of interest.
We thank Dr. Moshe Giladi, and Dr. Tali Voloshin for inovitro technical support and for fruitful scientific exchange.
This work was supported by a grant from Novocure. The grant partially (I.V. and L.G.P.) or fully (G.D.) covered salary and travel assistance for presenting the results of this study. R.G. received an unrelated institutional grant from Novocure. Z.R. received an institutional grant from Novocure.
The RNA-Seq data presented in this article have been submitted to the European Genome-Phenome Archive (https://ega-archive.org/studies/EGAS00001005392) under accession number EGAS00001005392.
Conception and design: I.V., Z.R., R.G., G.D., H.S.G., L.G.P., and Z.E.; development of methodology: I.V., G.D., H.S.G., L.G.P., M.R., T.S., A.G.-L., and N.P.; conducting experiments: G.D., H.S.G., L.G.P., O.H., and N.P.; analyzing data: I.V., G.D., H.S.G., L.G.P., and M.R.; writing the manuscript: I.V., G.D., and H.S.G. All authors have read and approved the final manuscript.
The online version of this article contains supplemental material.