TIGIT and PD-1 Immune Checkpoint Pathways Are Associated With Patient Outcome and Anti-Tumor Immunity in Glioblastoma

The Cancer Genome Atlas (TCGA) database was used to assess survival of patients with GBM in accordance with gene expression levels of immune checkpoint molecules. Survival analysis was performed through the cBioPortal platform using a z-score of 1.0 for all checkpoint receptors and their respective ligands. The correlation of checkpoint gene expression with z score >2.0 was considered upregulated expression. Kaplan-Meier (KM) survival curves were generated to determine overall survival (OS) and disease-free survival (DFS).

The study uses RNA-seq datasets of GBM tissue from The Cancer Genome Atlas (TCGA), and the raw expression files were downloaded from TCGA Genomic Data Commons (GDC) Data Portal. Reads were quantified and mapped to human genome (Ensembl GRCh38 Homo sapiens) Salmon version 0.8.2 (22). Transcript-per-kilobase-million (TPM) were used for gene-correlation and pathway analyses. Pearson’s rank correlation analysis was performed for TIGIT and PDCD1. Genes with statistically significant correlation (p value < 0.05 and false discovery rate (FDR) p < 0.05) were used to determine pathway enrichment using Gene Ontology (GO) (23) for Reactome (version 65 Released 2020-11-17) (24) and PANTHER Overrepresentation Test (Released 20200728) (25) curated pathways. Pathway enrichment cutoff was set for p<0.05 using Fisher’s Exact test and FDR p<0.05 and enrichment scores greater than 1. Immunological network analysis was performed using ClueGo v2.5.7 (26) and Cytoscape 3.8 (27) with the current parameters: GO ImmuneSystemProcess EBI-UniPort, GO term fusion, network specificity was set to medium-detailed, pathways’ p value <0.05 with Benjamini-Hochberg correction. Positively and negatively correlated genes were used to determine positively and negatively associated networks, respectively.

scRNA-seq data were obtained from Wang et al. (28) and processed as described previously (28). Briefly, the neoplastic cells and non-neoplastic cells were separated via copy number variation (CNV). The presence/absence of CNVs was assessed with CONICSmat (29), and the primary cell types of non-neoplastic cells (i.e. monocytes/myeloid) were identified by using ELSA (30). CD11b⁺ monocyte/myeloid cell population was sampled for further analysis using Seurat package on Bioconductor (R) (31). Following Elbow Plot analysis, the number of principal components analysis (PCA) was set up to 3 with 0.2 resolution for UMAP clustering.

GL261 cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM, Gibco) supplemented with 10% fetal bovine serum (FBS, Hyclone), 1x antimycotic-antimycotic solution (Gibco), 1% L-glutamine, ß-mercaptoethanol, 200 µg sodium pyruvate, and 1x NEAA. Cell lines were kept in a 37°C humidified incubator with 5% CO2. Cell number and viability were measured using the trypan exclusion method (0.4% trypan Blue, Gibco).

C57BL/6J mice (Stock No. 000664) and B6.Cg-Thy1 a/Cy Tg(TcraTcrb)8Rest/J (PMEL; Stock No. 005023 (32)) were purchased from the Jackson Laboratory and housed in animal facility of the UPMC Children’s Hospital of Pittsburgh. Animals were kept in the facility for at least one week prior to performing any procedures to minimize stress-related symptoms. 5–6-week-old female were used in the experiments. All experiments were conducted following protocols approved by the University of Pittsburgh Institutional Animal Care and Use Committee (IACUC).

Mice were anesthetized by mask inhalation of 1.5% vaporized isoflurane throughout the surgical procedure. GL261 cells (100,000 cells in 2 μL DPBS) were stereotactically implanted into the caudate nucleus using the following coordinates relative to bregma: x = +2.5 mm (lateral), y = +1.5 mm (anterior), and z = 2-3.0 mm (inferior) (33). MRI was performed 7 days post tumor cell implantation to confirm tumor presence, and again at day 40 to measure tumor size growth in control-treated animals and αTIGIT & αPD1 dual blockade-treated animals. All mice were randomized prior to their separation into treatment groups. IgG1 (clone MOPC-21), IgG2a (clone RTK2758), αPD1 (clone RMP1-14) and αTIGIT (clone 1G9) antibodies were obtained from Bio-X-Cell Antibodies were dosed at 200 μg per animal and administered via intraperitoneal (i.p.) injection, as described previously (34) for both the survival and immunophenotyping studies. Anti-TIGIT and anti-PD1 treatments were given on the same day twice per week starting on day 8, for a total of 7 doses. Mice were euthanized after receiving seven doses of immune-checkpoint inhibitor therapeutic antibodies (αTIGIT/αPD1) to investigate biological endpoint and immune cell phenotype. Mice were monitored for weight loss and morbidity symptoms for survival study. All survival experiments were repeated in triplicate with at 4-6 animals per group.

For the biological endpoint study, mice were euthanized on day 22 (CO2 asphyxiation followed by cervical dislocation) post-tumor inoculation. Brains were dissected and processed for flow cytometry analysis. Brains were homogenized in Collagenase IV Cocktail (3.2 mg/mL collagenase type IV, 1.0 mg/mL deoxyribonuclease I, 2 mg/mL Soybean Trypsin Inhibitor). Samples were centrifuged for 5 minutes at 1500 rpm, followed by red-blood cell (RBC) lysis using ACK lysing buffer (Lonza). Cell viability was measured using the trypan blue exclusion method. Cells were resuspended in FACS Buffer (DPBS with 1% BSA) and centrifuged for 5 minutes at 300 g, after which the pellet was resuspended in FACS buffer. The cells were then stained with appropriate antibodies and acquired on a BD LSR Fortessa flow cytometer.

Patient-derived GBM tissue was dissociated, using Accutase (1:10), to form a single cell suspension (SCS). SCS was centrifuged at 1500rpm for 5 mins. The pellet was resuspended in 5mL of 70% Percoll solution. A Percoll gradient of 5mL of 37 and 5mL of 30% Percoll sequentially, was then overlaid onto the tumor-containing 70% Percoll solution. The tumor gradient solution was centrifuged at 2400 rpm for 20 minutes. Immune cells at the interphase were collected and washed once with PBS. The cells were then stained with appropriate antibodies and acquired on a BD LSR Fortessa flow cytometer.

Peripheral blood samples were collected in preservative-free heparin tubes (10 U/mL) and layered into an equal volume of Ficoll-Hypaque density gradient solution (Amersham Pharmacia Biotech Ltd., Little Chalfont, UK). Samples were then centrifuged at 2250 rpm for 20 minutes. After removal of the top layer (plasma), the mononuclear cells (PBMCs) were collected and washed twice with PBS (Hyclone™, GE Healthcare). Cell viability was determined by trypan blue exclusion and exceeded 95%. The cells were then stained with appropriate antibodies and acquired on a BD LSR Fortessa flow cytometer.

Immunosuppressive myeloid cells were generated as described previously (35). Briefly, tibia and femur-derived BM cell from C57BL/6j mice were cultured in complete DMEM media supplemented with 10 ng/ml each of GM-CSF and IL-4. On day 3, floating cells were removed, and medium was replaced with 1:1 complete DMEM media to GL261 tumor-derived conditioned media (TCM), supplemented with GM-CSF and IL-4 for 3 additional days prior to use.

T cell suppression assay was performed as described previously (21, 36). In brief, hGP100-restricted (B6.Cg-Thy1a/Cy TCR-transgenic) CD8⁺ T cells were isolated from PMEL-mice (32) using magnetic bead separation (Miltenyi Biotec) and labeled with Cell-Trace proliferation dye (Invitrogen. Cat. No C34557↗) according to the manufacture guidelines. Feeder cells (antigen presenting cells) were generated from non-CD8⁺ cell fraction and were treated with 10 μg/ml of mitomycin at 37°C for 1 hour to cease proliferation (37). T cells and feeder cells were co-cultured with BM-derived immunosuppressive myeloid cells in the presence of 100 U/mL hIL-2 (PeproTech), 100 µg/mL hGP10025-33 peptide (antigen), and 10 µg/mL of either IgG2a (RTK2758) – as control, or αPD1 (RPMI14) and αTIGIT (1G9). Cells were collected and analyzed on day 4 by flow cytometry.

Prior to cell surface staining, samples were stained with cell viability dyes (GhostDye or 7AAD) in PBS for 20 minutes in 4°C and then washed with FACS buffer. For mouse immune cell staining, the cell suspensions were blocked with 1% anti-mouse Fc-receptor (CD16/CD32) in FACS buffer for 20 minutes, then washed and stained with fluorescently labeled anti-mouse antibodies for 45 minutes in FACS buffer at 4°C. TILs (n=5) and PBMCs from 2 matched, 3 unmatched and 3 healthy donor (HD) patient samples (n=8) were washed with PBS and stained with cell-surface antibodies for 30 minutes at 4°C per the manufacture guidelines. After staining, cells were washed with FACS buffer and fixed with fixation buffer (BD Cytofix/Cytoperm buffer). The cells were washed with FACS buffer, resuspended in FACS buffer and analyzed by flow cytometry. The antibody clones were purchased from BioLegend or eBioscience and used for flow cytometry as follows: For mouse cell staining: CD4 (GK1.5), CD8 (53-6.7), CD11b (M1/70), CD45 (30-F11), Gr-1 (RB6-8C5), CD3 (17A2, and 145-2C11), Granzyme B (QA16A02). For human cell staining: CD45 (2D1), CD11b (ICRF44), CD3 (C3e/1308), CD8 (OKT-8), PD-1 (EH12.2H7), PD-L1 (MIH2), CD33 (WM53), CD226 (11A8), TIGIT (A15153G), CD155/PVR (SKII.4). GhostDyes (TONOBO) UV450 and Red-780, and 7AAD were used to stain for cell viability (live/dead) according to the manufacture guidelines. Gating was performed on live CD45⁺ cells to designate all immune cells. All samples were analyzed on a BD LSRFortessa (BD Biosciences). Data were analyzed using BD FACSDiva (BD Biosciences) and FlowJo V10 data analysis software (FlowJo LLC).

Kaplan-Meier survival curves were generated to determine survival and then compared using the log-rank Mantel Cox test. One-way analysis of variance (ANOVA) with Kruskal Wallis multiple comparisons test was used to compare assays containing more than two groups. Statistical significance was considered as p <0.05. Normal distribution was assumed unless specified overwise in the text or figure legend. The analyses were performed using GraphPad Prism 8 or Bioconductor (R programing) on RStudio.

To identify putative immunotherapy targets for GBM, we evaluated the expression of IC genes and their ligands in RNA sequencing (RNA-seq) data of 153 GBM tumor samples in the TCGA database (38). We first assessed the correlation of 15 established IC gene expression levels with overall survival (OS) and disease-free survival (DFS) (39, 40). Upregulated expression was defined as expression z score greater than 2. Our data demonstrate that upregulated expression (red lines) of TIGIT and PDCD1 (gene encoding PD1) were associated with poor patient outcome and increased mortality as compared with patients who had no change in TIGIT and PDCD1 RNA expression (green lines) (Figures 1A, B). Upregulated ICOS expression was also associated with reduced OS and DFS, although the data did not reach significance (Figures 1A, B). However, upregulated expression of other IC receptor genes, including CTLA4, LAG3, TIM3 (HVAC1), BTLA4, and CD224 were not associated with changes in OS and DFS. Interestingly, expression of genes for CD155 (PVR), PD-L1, and ICOS-L, the ligands for TIGIT, PD1, and ICOS, respectively, was also significantly associated with decreased OS and DFS, whereas upregulated expression of other IC gene ligands did not affect these parameters in GBM patients (Figures 2A, B). Although our survival analysis assessed patients with elevated expressed based on Z score (i.e. compared to mean expression of that gene), we further examined the absolute expression of each gene to determine the extent of therapeutic utility among all patients. Our data show that a large portion of patients showed to have physiologically relevant expression levels (TPM>1) of the genes encoding to the checkpoint receptors PD-L1 (94%) and CD155 (PVR; 100%) (Supplemental Figure 1). Additionally, PD1 and TIGIT were reported to be expressed by large frequencies of GBM CD4⁺ and CD8⁺ TILs (34, 41). Taken together these data suggest that PD1/TIGIT-targeted therapy may be relevant for many patients with GBM.

Our data revealed that TIGIT/CD155 and PD1/PD-L1 checkpoint genes were significantly associated with GBM clinical outcome, thus we next analyzed RNA-seq data from these patients to identify genes and pathways which may be involved with TIGIT and PD1 expression in GBM. Notably, the expression of TIGIT and PDCD1 were significantly correlated with each other (Figure 3A), suggesting a rationale for dual blockade of these checkpoint molecules in GBM patients. Despite their significant correlation, TIGIT and PDCD1 may be associated with unique gene networks and pathways (42). Therefore, we next interrogated the gene networks associated with the expression of TIGIT and/or PDCD1 in GBM. We identified a total of 6347 genes which correlated with TIGIT and PDCD1 expression with high statistical significance (p<0.05 and FDR<0.05) (Figure 3B). While many genes correlated with both TIGIT and PD1 expression, we also identified a large number of genes and pathways uniquely correlated with either TIGIT or PD1 (Figures 3B, C).

TIGIT/PD1 (shared)-associated pathways included immune related pathways, such as Toll-like receptor (TLR) signaling, interleukins signaling (such as IL-10 and IL-2), TCR signaling and T cell activation, and innate immune system pathways (Figure 3D). Interestingly, TIGIT-associated pathways included Treg development, MHC class I presentation, caspases and death-receptors signaling, control of cell cycle transition, regulation of TLR and Nf-kB signaling, and p53 regulation (Figure 3D). PD1-associated pathways included cell motility, oxidation and phagocytosis, IL-12 mediated Jak-STAT signaling, MHC class II and antigen presentation, and EGF receptor signaling (Figure 3D). Immunological network analysis showed that many immune responses were strongly associated with the expression of TIGIT and PD1, mostly T cell activating and regulation of immunity, but also innate immune functions such as leukocytes degranulation, and functions of macrophages and dendritic cells (Figure 3E).

Together, these data suggest that upregulated expression of TIGIT and PD1 may confer immunosuppression and tumor aggression in GBM patients through both shared and distinct pathways, and therefore targeting both these pathways may be beneficial for improving clinical outcome of GBM patients.

Our data suggest a beneficial outcome for IC blockade of TIGIT and/or PD1 in GBM. To investigate this hypothesis, C57BL/6 mice were intracranially injected with syngeneic GL261 cells, followed by 7 doses of immunotherapy with (1) isotype control antibodies, (2) αPD1, (3) αTIGIT, or (4) a combination of αTIGIT/αPD1 therapeutic antibodies, administered twice per-week starting on day 8 post-tumor injection (Figure 4A). Analysis of immune cells was performed uniformly across groups on day 22 post-tumor implantation (biological endpoint) followed by MRI analysis for tumor size on day 40 for control and dual αTIGIT/αPD1-treatment groups (Figure 4A). Control mice (Isotype; black line) displayed median survival of 33 days (range: 29-51 days) with severe morbidity signs and did not reach long-term survival endpoint (Figure 4B). While αTIGIT monotherapy (green line) moderately improved survival, treatment with αPD1 (blue line) or a combination treatment of αTIGIT/αPD1 (red line) significantly prolonged animal survival as compared with isotype treated animals (Figure 4B). The median survival of αTIGIT treatment was 34 days (range: 32-43 days) while αPD1 monotherapy (green line) was 37 days (range 32-74 days). Notably, αTIGIT/αPD1 dual treatment most significantly prolonged mice survival with median survival of 48 days (range 39-74 days) (Figure 4B). MRI analysis showed that in αTIGIT/αPD1 treated animals the tumor size was significantly smaller than tumors in isotype-treated animals (Figure 4C). These data confirm previous results in which immunotherapy combination of αPD1 with αTIGIT reduced tumor burden and improved survival of mice with glioma (34).

To examine a mechanism by which the combination therapy improved anti-tumor immunity, we explored the effect of treatment on tumor-infiltrating lymphocytes (TILs) and their cytolytic phenotype on day 22 post tumor implantation. Although αPD1 monotherapy did not significantly affected the percentages of CD4⁺ TILs, treatment with either αTIGIT or αTIGIT/αPD1 resulted in a significant increase of CD4⁺ TILs as compared with control animals (Figure 4D). Additionally, we noted a significant increase in percentages of CD8⁺ TILs in tumors in animals treated with αPD1 or αTIGIT/αPD1, but not in αTIGIT-monotherapy treated animals (Figure 4E). Analysis showed that treatment with either αPD1 or αTIGIT monotherapy resulted in a mild increase in the percentages of CD8⁺ granzyme-B⁺ TILs, while this effect was significantly increased in αTIGIT/αPD1 combination treatment (Figure 4F). These data are complementary to previous results showing that αPD1 or αTIGIT immunotherapy enhances the expression of TNFα and IFNγ in TILs from GBM (34) as well as other cancers (43), and suggest that the therapeutic effect of checkpoint blockade with αTIGIT/αPD1 may work through distinct mechanisms to affect CD4⁺ and CD8⁺ TILs and promote anti-glioma immunity.

Previous reports have shown that MDSCs stimulate suppressive mechanism to develop a pre-metastatic niche, promote tumor growth, inhibit anti-tumor function of TILs, and negate immunotherapy which results as resistance to IC blockade (44, 45). Furthermore, MDSCs have been shown to contribute to immunosuppressive microenvironment in gliomas, including GBM (45–47). MDSCs were shown to express PD-L1 (48). Additionally, inhibition of TIGIT was reported to abrogate MDSC immunosuppressive capacity in vitro (49). Together, these data suggest that targeting PD1 and TIGIT pathways may affect MDSCs in GBM. However, the effects of these checkpoint on MDSC infiltration in gliomas are ill defined (45). We, therefore, investigated if MDSCs were affected by immunotherapy in our model on day 22 (biological endpoint; end of immunotherapy). Shown in Figure 5A, glioma infiltrating MDSC subsets were characterized by the expression of Gr1 and CD11b as follows: PMN MDSCs were defined as CD11b⁺Gr1^high cells, whereas Mo MDSCs were defined as CD11b⁺Gr1^{intermediate (int)} cells (18). We evaluated the levels of MDSCs and their subsets following immunotherapy (Figure 5A; lower panel). Our data show, that compared with isotype treatment, both αTIGIT monotherapy and dual blockade of TIGIT & PD1 significantly reduced the frequencies of GL261 glioma infiltrating MDSCs (CD45⁺CD11b⁺Gr1⁺ cells), most strikingly for αTIGIT/αPD1 combination therapy (Figure 5B). Treatment with αPD1 showed a trend of decreasing MDSC percentages, though the results did not achieve statistical significance (Figure 5B). Analysis of MDSC subsets revealed that αTIGIT/αPD1 dual treatment significantly decreased the frequencies of PMN MDSCs (Figure 5C), while Mo MDSCs levels remained mostly unaltered (Figure 5D). Furthermore, we observed a statistically significant increase in ratios of CD8⁺ T cells over total MDSCs in tumors when mice were treated with αTIGIT monotherapy or αTIGIT/αPD1 combination therapy (Figure 5E). Blockade of PD1 alone did not significantly increase the CD8⁺ T cells/MDSCs ratios (Figure 5E). Together, our data reveal a mechanism of TIGIT/PD1 blockade in glioma and suggest distinct roles of these ICs on MDSC subsets and in regulating tumor immunity.

We next evaluated the potential of αTIGIT/PD1-immunotheraphy to impact MDSC-like cell in GBM patients. For that, we first analyzed single-cell (sc)RNA-seq data of CD11b⁺ myeloid cells from GBM patients (28) for the expression of PD1, TIGIT, and their ligands. Myeloid cells were confirmed based on the expression of CD45 (PTPRC), CD14, and CSF1R genes (Figure 6A) (28, 50). Of note, we identified 4 unique clusters of tumor-associated myeloid/macrophage cells (TAMs) in GBM, which had distinct expression profiles (Figure 6A and Supplemental Figure 2A). The expression of CD33, an hematopoietic progenitor cell marker which commonly used to identify pan-MDSCs (51), was distributed throughout the TAM clusters. Nonetheless, the expression of inhibitory and suppressive markers, including genes for IL-4R, IL-10, IL-6, VEGFA, CCL2 and IL-1β (Figure 6A and Supplemental Figure 2A), were mostly expressed by cluster 0, suggesting that these cells had a tumor-promoting and immune-suppressing functions, which resemble MDSC-like cells (52). Interestingly, PDCD1 (PD1), CD274 (PD-L1), and CD226, were also predominantly expressed by cluster 0 (Figure 6B and Supplemental Figure 2B). CD155 (PVR) was also associated and expressed by cluster 0, although less frequent than CD226, PD1, and PD-L1 (Figure 6B and Supplemental Figure 2B). PVRL2, another inhibitory receptor that bind to TIGIT (53), was also expressed by TAMs, with highest expression levels in cluster 0 (Figure 6B and Supplemental Figure 2B). We did not detect TIGIT expression in TAMs by scRNA-seq (Figure 6B). Additionally, we noted high expression and associated of ICOS-L with cluster 0 (Supplemental Figure 2B), which interestingly was also associated with GBM patient OS and DFS (Figures 2A, B). These data suggest that immunosuppressive TAMs, such as MDSCs, express genes for PD1, PD-L1 and TIGIT-ligands. Therefore, evaluated the protein expression of these markers on CD45⁺ immune cells in patient derived GBM tissue and PBMCs, and healthy donor (HD) PBMCs. The frequencies of CD11b⁺ CD33⁺ cells in GBM TILs were on average higher compared to GBM PBMCs, and were significantly higher than of HD PBMCs (Figure 6C and Supplemental Figure 2C), suggesting that MDSCs are present at high levels in GBM and could contribute to the TME immunosuppression (46). Consistent with our scRNA-seq data, CD11b⁺ CD33⁺ cells had higher expression levels of PD1, PD-L1, PVR, and CD226 in TILs, as compared with CD11b⁺ CD33⁺ cells from PBMCs of GBM patients and HDs, most notably for PD-L1 and CD226 expression (Figure 6D). Moreover, as compared to PBMCs samples we noted an increased expression of PD1, PD-L1 and TIGIT on CD8⁺ TILs, while CD4⁺ TILs had mostly upregulated expression of TIGIT (Figure 6E). Together, these data suggest that CD11b⁺ CD33⁺ TAMs may promote immunosuppressive functions at-least in part through expression of PD1/TIGIT-checkpoint ligands. To test this hypothesis, hGP100-restricted naïve T cells isolated from pmel mice (32) were activated in vitro with hGP100_25–33 peptide and feeder cell (antigen presenting cells; APCs) in the presence of bone-marrow derived myeloid cells (putatively MDSC-like cells) cultured in GL261 cell-derived tumor-conditioned media and treated with αTIGIT and αPD1. Our data indicated that glioma-conditioned immunosuppressive myeloid cells significantly inhibited CD8⁺ T cell proliferation, which was restored by the addition of αTIGIT or αPD1 (Figure 6F). In summary, these new data suggest that immunosuppressive myeloid cells, and presumably MDSCs, suppress anti-tumor immunity by inhibiting antigen-specific T cell function in GBM, at least in part via TIGIT and PD1 pathways, which may have major implication to patient treatments by immunotherapy.