Blocking Wnt/β-catenin Signal Amplifies Anti-PD-1 Therapeutic Efficacy by Inhibiting Tumor Growth, Migration, and Promoting Immune Infiltration in Glioblastomas

This article is featured in Highlights of This Issue, p. 1221

Gliomas are the most common primary central nervous system tumors (CNST) including astrocytoma, oligodendroglioma, and oligoastrocytoma (1). World Health Organization (WHO) classified them into grades I–IV according to their anaplasia degree (2). Among them, grade IV corresponds to the highly malignant tumors represented by glioblastoma (GBM), which accounting for 80% of primary CNST and 45.2% of all malignant CNST. With an incidence of 5 cases per 100,000 persons, GBM caused 225,000 deaths worldwide (3). At present, surgery combined with systemic temozolomide and radiotherapy is the standard treatment of GBM, which can significantly improve the 2-year survival rate of patients, but has little effect on their overall prognosis. Besides, traditional therapy is associated with significant side effects, such as epilepsy and the destruction of the blood–brain barrier (BBB; refs. 4, 5). With the development of molecular targeted therapy and immunotherapy, researchers are looking for more efficient treatments for patients with GBM.

Tumor cells can exploit programmed cell death 1/programmed cell death ligand 1 (PD-1/PD-L1) axis to inhibit T-cell function and evade immune surveillance (6). PD-1/PD-L1 immune checkpoint blockade (ICB) therapy has achieved remarkable results in different solid tumors, especially lung cancer, kidney cancer, and melanoma in recent years (7). However, the response to anti-PD1/PDL1 monotherapy depends largely on the recruitment of tumor-infiltrating lymphocytes (8). Therefore, its efficacy was demonstrated to be limited in the “immune desert” GBM because of the immune privilege and BBB of CNS (9). To solve this problem, a variety of combined therapies have been studied to improve the response of GBM to anti-PD1/PDL1 immunotherapy.

Wnt protein is a cysteine-rich glycoprotein which participates in the regulation of embryonic development in higher animals. The abnormal activation of classical Wnt/β-catenin signaling pathway affects cell proliferation, migration, and invasion, which is closely related to the occurrence and development of several tumors (10–12). In addition, it also affects immunotherapy by regulating the differentiation and development of dendritic cells (DC), B cells, and T cells (13). For example, specifically knocking out the Wnt canonical signaling pathway receptor LRP5/6 on DCs can effectively enhance the immunotherapeutic effect of tumor-bearing mice (14). More importantly, studies have confirmed that upregulation of Wnt pathway may lead to resistance to immunotherapy by initiating non-T cell–inflamed tumor microenvironment (TME). An article published in Nature has revealed that upregulation of endogenous β-catenin inhibits the production of CCL4, making it impossible to recruit BATF3 DCs into TME thereby preventing T-cell activation (15). Recently, a multi-omic analysis of colorectal cancers discovered the reason that non-Microsatellite Instability (MSI)-high subtype (accounting for 85% of cases) could not benefit from ICB therapy may be attributed to the upregulation of Wnt pathway, which was negatively associated with the immune infiltration in TME (16). Given these research progress, we speculate that the inhibition of Wnt pathway may enhance the immunotherapy effect of PD-1 antibodies in GBM by inhibiting cell growth, and migration, as well as improving T-cell infiltration.

Here, we focus on the effect of GBM-intrinsic Wnt/β-catenin signal on T-cell activity and infiltration. First, we performed The Cancer Genome Atlas (TCGA) analysis and IHC staining of clinical samples to uncover the existence of negative association between Wnt/β-catenin signal and T-cell infiltration in gliomas. Then, we studied Wnt/β-catenin signal's influence on the growth, proliferation and PD-L1 expression of tumor cells, as well as T-cell activation and apoptosis in a co-culture system in vitro comprising GBM cells and Jurkat cells. Finally, the efficacy and mechanism of combination therapy based on targeting Wnt pathway and PD-1 antibody was investigated in a homogeneous GL261 transplantation mouse model. The results of this study will provide experimental basis for promoting the efficacy of PD-1/PD-L1 blockade therapy on GBM through modulating Wnt/β-catenin pathway.

U118, U87, U251 (GBM human cell line), C6, GL261 (GBM murine cell line), and Jurkat cells were purchased from Cell Bank of Chinese Academy of Sciences (Shanghai, P.R. China). They were authenticated by short tandem repeat profile and tested negative for Mycoplasma contamination. The passage number of them in this experiment was between 4 and 6. Cell culture medium and related reagents: RPMI1640 medium and DMEM medium (Corning Life Sciences), penicillin (Beyotime Biotechnology), l-Glutamine, and FBS were obtained from Gibco. The cell incubator maintained 5% CO₂ and 37°C.

Tumor cell suspension (100 μL) was added to the 96-well plate with the concentration of 5 × 10⁴/mL. After treating with XAV939 (Seleck Chemicals, S1180) for a certain time, 0.5 mg/mL of methylthiazolyldiphenyl-tetrazolium bromide (MTT) reagent was added with 10 μL each pore. MTT could react with succinate dehydrogenase in living cells to produce the blue-purple product formazan. DMSO was used to thoroughly dissolve formazan after 4 hours of incubation at 37°C. The optical density was detected by microplate reader at 570 nm.

Cytotoxicity was measured by using CytoTox96Non-Radioactive Cytotoxicity Assay (Promega, G1780). U118 cells (5 × 10³) as target cells (T) and various numbers of Jurkat cells as effector cells (E) were added to “experimental” wells in a 96-well culture plate with or without 25 μmol/L XAV939 according to different T:E ratios. The “Target Maximum” and “Target Spontaneous” wells were seeded with 5 × 10³ U118 cells with or without 10 μL of the Lysis Solution (10×), and “Effector Spontaneous” wells were seeded with Jurkat cells only. The lactate dehydrogenase (LDH) release was detected following manufacturer's instruction after the cells were administrated as stated above for 48 hours. The percent cytotoxicity for each T: E cell ratio was calculated as follows:

Cells or tumor tissues were lysed in RIPA Cell Lysis Buffer (Beyotime, #P0013B) and the total protein concentration was measured by BCA quantification kit (Beyotime, #P0012). Next, same amount of protein (15 μg) was separated by SDS-PAGE and transferred to polyvinylidene fluoride membrane. Primary antibodies against N-cadherin (#13116), E-cadherin (#3195), Slug (#9585), β-Actin (#8457), and β-Tubulin (#2128) were purchased from Cell Signaling Technology; β-catenin (ab32572), PD-L1 (ab238697) were purchased from Abcam. Finally, it was probed with peroxidase-conjugated secondary antibody and washed with TBST. ChemiDoc software (Bio-Rad) was used for the immunoblot detection after bands were treated with ECL substrate (Pierce).

Nonspecific scrambled siRNA (used as control) and prevalidated β-catenin siRNA that targeted TNKS1 or TNKS2 were obtained from Guangzhou RiboBio Co. (siRNA1-GCCACAAGATTACAAGAAA; siRNA2-GACTACCAGTTGTGGTTAA; siRNA3-GATGGACAGTATGCAATGA). It was added in 6-well plates and delivered with Lipofectamine 3000 Transfection Reagent (Thermo Fisher Scientific) when the cell density reached 70%–90%.

After the U118 cells were cultured into confluent monolayers in a 6-well cell culture cluster, a 10 μL sterile tip was used to draw multiple lines on the cell layer and different concentrations of XAV939 were added. The average level of cell migration was analyzed after obtaining an image at different timepoints with an inverted microscope.

Jurkat cells were starved overnight in RPMI1640 medium containing 1% FBS, and then cultured in a T25 culture flask pre-immobilized 2 μg/mL anti-CD3 (OKT-3; BioLegend) for 2 hours incubation. The cells were then added in 24-well plates and incubated with 4 μg/mL anti-CD28 (clone CD28.2; BioLegend) for 48 hours. Then cells were collected and incubated with PD-1 mAb (Cell Signaling Technology, catalog no. 86163) for 1 hour and stained with secondary Alexa Fluor 488 AffiniPure Goat Anti-Rabbit IgG (H+L; YEASEN, catalog no. 33106ES60) for half an hour. Finally, the stained cells were washed by PBS twice and detected in a BD FACS Aria II Flow Cytometer (Becton Dickinson).

GBM cells (U118 or U87) with or without XAV939 and preactivated Jurkat cells were co-cultured at a ratio of 1: 5 in the 6-well plate. After 48 hours interaction, both the adherent tumor cells and suspended Jurkat cells were obtained separately and treated with Annexin V-FITC/PI Detection Kit (KeyGEN Biotech, catalog no. KGA108) as manufacturer's instructions and then assessed by flow cytometry (Becton-Dickinson). Viable cells display FITC Annexin V and PI negative (FITC⁻/PI⁻); cells in early apoptosis are FITC Annexin V positive and PI negative (FITC⁺/PI⁻); cells in late apoptosis are FITC Annexin V and PI positive (FITC⁺/PI⁺) and dead cells are FITC Annexin V negative and PI positive (FITC⁻/PI⁺).

TRIzol reagent (TIANGEN, DP405) was used to extract the total mRNA from cells. HiScript II Q RT SuperMix for qPCR (Vazyme, R123-01) and a MJ Mini Thermal Cycler System (Bio-Rad Laboratories) was used to remove the genomic DNA and perform a reverse transcriptase reaction according to instruction. Then the obtained cDNA products were used to perform RT-PCR or qRT-PCR reactions, respectively. The primers were listed in Supplementary Table S1.

2×Taq Plus Master Mix II (Vazyme, #P213-01) and a MJ Mini Thermal Cycler System (Bio-Rad Laboratories) were used for RT-PCR reaction. Then, the same amount of PCR product (5 μL per well) was separated by 2% agarose. Then photos were taken with ChemiDoc software (Bio-Rad). ChamQTM Universal SYBR qPCR Master Mix (Vazyme #Q711) was used to perform qRT-PCR reactions on a fluorescent quantitative PCR instrument (Bio-Rad). After the reaction, the amplification curve and melting curve of real-time PCR were obtained.

All animal experiments on this subject have been reviewed and approved by the Animal Ethical Committee of School of Pharmacy at Fudan University (Shanghai, P.R. China). The 6-week-old male C57BL/6 mice (16—20 g) were purchased from Shanghai SLAC and injected subcutaneously with 100 μL of GL261 cells suspension (5 × 10⁷/mL) on the back. The experimental animals were kept at 20°C–25°C and received 12 hours of light each day.

When the average tumor volume reached 100–150 mm³, mice were randomly divided into four groups (n = 5 each) and respectively treated with PBS (control group), 2 mg/kg of XAV939 (Selleck Chemicals, S1180) every 3 days from day 0 (XAV939 group and XAV939+α-PD1 group), and 10 mg/kg of anti-PD1 antibody (BioXCell, clone 29F.1A12) on day 1 (α-PD1 group and XAV939+α-PD1 group). XAV939 was dissolved in DMSO at 12 mg/mL (38.42 mmol/L) and diluted with PBS, and the vehicle DMSO showed no influence on tumor growth (Supplementary Fig. S1). The body weight, tumor diameter (a), and short diameter (b) of the mice were measured every 3 days during the administration (tumor volume |$= 0.5\times ab^2$|⁠). Eighteen days after the administration, mice were sacrificed and solid tumors were excised and weighed.

Clinical samples of patients with glioma were obtained from Department of pathology of Minhang Hospital, Fudan University (Shanghai, P.R. China) after obtaining approval from the ethics committee and signing informed consent with the patients. Detailed clinicopathologic characteristics of them have been added in Supplementary Table S2.

Tumor tissue samples were fixed with 4% paraformaldehyde, then sectioned and stained with hematoxylin and eosin (H&E) for histopathologic analysis with an optical microscope. For IHC analysis, anti-IFNγ Rabbit pAb (GB11107), anti-CD4 Rabbit pAb (GB13064-2), anti-PDL1 Mouse pAb (GB13339), anti-CD8 alpha Rabbit pAb (GB11068), horseradish peroxidase (HRP)-conjugated Goat Anti-Rabbit IgG (GB23303), and HRP-conjugated Goat Anti-Mouse IgG (GB23301) were purchased from Servicebio. After paraffin section of tumor tissue was dewaxed and hydrated, the antigen was repaired using Tris/EDTA (pH 9.0). The samples were soaked in 3% H₂O₂ in the dark for 25 minutes, and then blocked with 3% BSA for 30 minutes. After incubating with the relevant primary antibodies at 4°C overnight, sections were treated with secondary antibodies (HRP-labeled) for 50 minutes at room temperature. Then tissue sections were counterstained with hematoxylin and fixed after colored with Diaminobenzidine (DAB). Finally, the positive expression of the samples was analyzed by an inverted microscope (Nikon). Immunopositive areas were evaluated by IHC profiler plug-in component of ImageJ software for three times, and then the data were analyzed by GraphPad Prism 8.

R package TCGAbiolinks was used to obtain the Fragments Per Kilobase Million (FPKM) normalized gene expression data and clinical data for TCGA-glioma samples from UCSC Xena browser (GDC hub: https://gdc.xenahubs.net). To cluster the samples into two groups (cold tumor and hot tumor), FPKM data were converted to Transcripts Per Kilobase Million (TPM) data and genes that are not expressed in more than 10% of the samples were removed, leaving 21,575 genes for analysis. The relative immune infiltration of 28 immune cell types (including B cells, T cells, natural killer cells, DCs, and regulatory T cells) in TME was quantified by ssGSEA. Gene sets used for this quantification were composed of 782 feature genes of 28 immune cell types as reported previously (17, 18). R package ConsensusClusterPlus was used to do hierarchical clustering of samples based on resampling and then samples were divided into low/high T cell–inflamed groups. R package limma was performed to detect the differentially expressed genes in these two groups (abs (log₂FC) > 1 & P < 0.05), and the core of it was to estimate the mean and variance of gene expression in different groups through linear models, so as to perform difference analysis. All heatmaps and Box plots were generated by the R package pheatmap and R package ggplot2, respectively.

Data of this work were showed as means ± SD. The statistical significance between groups was calculated by GraphPad Prism 8 using one-way ANOVAs or two-tailed Student t test. It is considered that P < 0.05 indicates statistical significance.

To study the correlation between immune infiltration and Wnt/β-catenin signal in gliomas, we collected 529 tumor samples from TCGA database and categorized them into low-immune inflamed group (336 samples) and high-immune inflamed group (193 samples) according to the signature gene expression of 28 immune cell types (Fig. 1A). Comparison of gene expression profiles showed that there were 1,548 genes expressed differently in two groups, including 23 Wnt/β-catenin signaling target genes (abs(log₂FC) > 1 & P < 0.05; Supplementary Table S3). Among the 23 differentially expressed Wnt/β-catenin signaling targets, 19 genes were highly expressed in the low-immune inflamed group. Moreover, according to previous researches on the correlation between Wnt/β-catenin signal and immune infiltration (15, 16, 19, 20), we further analyzed nine classic Wnt/β-catenin signal target genes (LRP5, TCF12, SOX2, GSK3B, EFNB3, APC2, AXIN2, WNT7B, and LGR5), and found they were highly expressed in the low-immune inflamed cohort (P < 0.001; Fig. 1B). In addition, data analysis showed that the expression of CTNNB1, which encodes β-catenin, was related to the progression of gliomas and highly expressed in malignant gliomas like GBM (Fig. 1C). Further studies showed that CD8⁺ T-cell infiltration was positively correlated with PD-L1 expression, but negatively correlated with β-catenin target genes across the two immune profile groups (Fig. 1D; Supplementary Table S4).

To further investigate the effect of Wnt/β-catenin pathway on T-cell infiltration in gliomas, we collected three representative glioma clinical samples (GBM, IDH R132H IHC negative) based on the β-catenin expression (high, medium, and low expression), and performed IHC staining to evaluate their CD8⁺ T-cell infiltration along with PD-L1 expression. We observed less CD8⁺ T-cell infiltration and low expression of PD-L1 in sample with high β-catenin expression. In contrast, the sample with less expression of β-catenin displayed obvious CD8⁺ T-cell infiltration and PD-L1 expression (Fig. 2A). Subsequently, we conducted cell experiments to further verify the above analysis results of clinical samples.

We selected multiple GBM cells, including human U87, U118, and U251 cell lines, rat C6 and mouse GL261 cell lines. The expression of β-catenin and PD-L1 were detected by Western blot analysis and RT-PCR at protein and RNA levels, respectively (Fig. 2B–F). The Western blot analysis showed significant expression of β-catenin along with low expression of PD-L1 in U118, C6, and GL261 cells. In contrast, PD-L1 was highly expressed in U251 and U87 cells accompanying with less β-catenin. The RT-PCR results of C6 and GL261 cells were consistent with the Western blot assay, both showing a negative correlation between the expression of PD-L1 and β-catenin. For three human GBM cell lines, PD-L1 expression levels from PCR test were concordant with that from Western blot test, but β-catenin levels were similar in these three cell lines. Thus, PD-L1 expression of GBM cells was negatively correlated with β-catenin expression in the protein level.

Overall, clinical sample analysis and the cell experiments further verify that β-catenin may negatively regulate T-cell infiltration and PD-L1 expression in gliomas. Then, we select U118 and GL261 cells which have high expression of β-catenin and low expression of PD-L1 for subsequent in vitro and in vivo homograft mouse model experiments, respectively.

XAV939 is an effective tankyrase inhibitor targeting Wnt/β-catenin pathway (21). Studies have shown that it has significant antitumor effects on colon (22), breast (23), and lung cancers (24). In this work, Wnt signaling was blocked by XAV939 or β-catenin siRNA in U118 cell line. From Fig. 3A and B, we observed successful inhibition of β-catenin along with increased expression of PD-L1 in U118 cells after administrated with XAV939 (50 μmol/L) or siRNA3. XAV939 treatment also upregulated the PD-L1 expression in GL261 cells (Supplementary Fig. S2). Previous studies have shown that PD-L1 expression is an important biomarker for predicting the therapeutic effect of anti-PD1/PDL1, thus the upregulation of PD-L1 may indicate a better antitumor immune response (25). The subsequent MTT tests displayed that the Wnt blockade by both XAV939 and siRNA could inhibit U118 cell proliferation to some extent time and dose dependently (Fig. 3C–F).

It is reported that overexpression or mutation of GSK3β, β-catenin, APC or Axin genes can induce excessive activation of Wnt/β-catenin signal, thereby regulating E-cadherin and transcription factor SNAI1 and SNAI2 which eventually lead to tumor metastasis and invasion (26, 27). In the cell scratch test, 25 and 50 μmol/L of XAV939 were added to the scratch-treated 6-well plates, and cells was observed at certain times. As shown in Fig. 3G, the cell migration ability of the XAV939 treatment group was obviously lower than that of control group with time and concentration increased, and the statistical graph of migration distance at 24 hours was shown in Fig. 3H.

Above data elucidate that blocking Wnt/β-catenin pathway in U118 cells can inhibit cell growth and migration.

Jurkat is a human T lymphocyte cell line which has been thoroughly characterized and widely used as a stable cell model to study T-cell function and signal pathway (28). Co-culture techniques of Jurkat and cancer cell have been well established and used in many previous studies (29, 30). PD-1 is an inducible receptor which plays an important role in regulating metabolic reorganization and immune balance, and usually overexpressed on activated T cells (31). To study the apoptosis of Jurkat cells and tumor cells in co-culture system, we preactivated Jurkat cells with CD3/CD28 antibodies. The enhanced expression of PD-1 after treatment with CD3/CD28 antibodies as evidenced by flow cytometry and Western blot analysis confirmed that Jurkat cells were effectively activated (Fig. 4A and B). Then the preactivated Jurkat cells were co-cultured with U118 (high β-catenin) or U87 (low β-catenin) cells for 48 hours. Elevated apoptosis was displayed in Jurkat cells co-cultured with U118 cells as compared with Jurkat alone or co-cultured with U87 cells, and this increase could be reversed by XAV939 (Fig. 4C; Supplementary Fig. S3). In the meantime, the apoptosis of U118 cells in co-culture system was increased after administrated with XAV939 (Fig. 4D).

CTLs mediate specific cellular immunity and play a major part in the process of antitumor immunity (32). After migrating to the tumor microenvironment and being activated, CTLs exert the cytotoxic function by secreting various effector molecules, including immune-promoting cytokines like IL2, TNFα, and IFNγ (33). Therefore, we detected the secretion of IL2, TNFα, and IFNγ in the co-culture system by using qRT-PCR to verify the relationship between T-cell cytotoxic function and Wnt/β-catenin signal. The results in Fig. 4E showed that Wnt inhibition by XAV939 obviously improved the secretion of IFNγ by Jurkat cells. Besides, LDH release assay also confirmed that XAV939 could effectively enhance the cytotoxicity of Jurkat cells (Fig. 4F).

In brief, above data indicate that Wnt/β-catenin inhibition can decrease the apoptosis of Jurkat cells and increase apoptosis of U118 cells in co-culture systems. In addition, Wnt/β-catenin inhibitor XAV939 enhance the cytotoxicity of Jurkat cells on GBM cells in co-culture system.

Given all above studies, we speculated that Wnt inhibition may improve the sensitivity of GBM to PD-1 antibody. To test it in vivo, we constructed a homogeneous transplantation model by subcutaneously implanting mouse GL261 cells into C57BL/6 mice. Then tumor-bearing mice were separated into four groups and treated with PBS, XAV939, anti-PD1, and XAV939 combined anti-PD1, respectively. The body weight of each group maintained a steady increase throughout the experiment, indicating a safe administration dosage of XAV939 and PD-1 (Fig. 5A). In the end, the average tumor volume of the XAV939 group was 1219.30 mm³, and that of the PD-1 group was 587.83 mm³ (vs. control group, P < 0.0001). While the tumor volume of the combination treatment group was only 105.64 mm³, indicating a best control of tumor progression (Fig. 5B). Meanwhile, the tumor weight of the combined treatment group was obviously lighter than that of other treatment groups (Fig. 5C and D). H&E staining of tumor tissues showed massive necrosis with increased nucleus dissolution and inflammatory cell infiltration in combined treatment group (Fig. 5E). It was noteworthy that Wnt inhibition by XAV939 elevated the PD-L1 expression in tumor tissues (Fig. 5F), which was consistent with the results of in vitro experiments. The increased PD-L1 expression might be an important reason for improved response of combined treatment group to PD-1 antibody therapy.

Collectively, above data reveal the enhanced anti-GBM effect of PD-1 antibody and upregulated PD-L1 expression triggered by Wnt inhibition in vivo.

To further explore the mechanism behind the enhanced antitumor effect of anti-PD-1 combined with XAV939, we analyzed the biomarkers of tumor proliferation, migration, and T-cell infiltration in tumor tissues. As shown in Fig. 6A, the proliferation biomarker ki67 in the combined treatment group and XAV939 treatment group showed significant downregulated expression compared with anti-PD-1 treatment group (****, P < 0.0001 and ***, P < 0.001, respectively). To confirm the anti-migration activity observed from in vitro wound-healing assay, we detected epithelial–mesenchymal transition (EMT), an important characteristic of tumor invasion and metastasis, in tumor tissues (34). In the EMT process, E-cadherin is one of adherens junction components critical for the epithelial state and Slug is the key transcription factor, whereas N-cadherin is associated with the mesenchymal state. The Western blot analysis exhibited decreased expression of N-cadherin and Slug along with higher expression of E-cadherin in XAV939 and PD-1 antibody combination group (Fig. 6B). Thus, in vitro and in vivo experiments together proved the migration inhibition of GBM cells by downregulation of Wnt/β-catenin signaling pathway. This result was in accord with earlier report that Wnt inhibition reduced cell migration and invasion in glioma cells (35). In addition, IHC assay displayed remarkably improved infiltration of CD4⁺ and CD8⁺ T cells, and raised expression of IFNγ in XAV939 and anti-PD-1 combination therapy, indicating the enhanced antitumor immune response induced by Wnt inhibition (Fig. 6C).

Overall, these results demonstrate that Wnt inhibition can successfully inhibit the proliferation and migration, and enhance the IFNγ expression and T-cell infiltration in GBM.

Immunotherapy targeting PD-1/PD-L1 has achieved remarkable results in the clinical treatment of various solid tumors (36). But its efficacy is closely related to the degree of immune cell infiltration in TME, thus exhibits limited benefit in immunologically cold tumors like glioma, especially GBM (37, 38). As early as 2003, PD-L1 was found to be expressed in glioma cells and tissues, and then it was studied as one of the therapeutic targets and biomarkers of gliomas (39). Analysis of 94 cases of GBM showed that PD-L1 expressed 1% or more in more than half of the samples (40). It is reported that PD-L1 expression as well as T-cell infiltration in TME are all favorable biomarkers for the responses of tumors to PD-1/PD-L1 blockade therapy (41). The brain was originally thought to be an immune exempt organ, but recent studies have confirmed that gliomas could be identified and infiltrated by innate and adaptive immune cells because of the imperfect BBB caused by cancer (42). It is glioma-intrinsic immune suppression that restricts these immune interactions. Therefore, further understanding the mechanisms underlying endogenous immune suppression and insufficient immune infiltration can provide new strategies for promoting glioma anti-PD-1 therapy (43, 44).

In the current research, through the analysis of TCGA data and clinical glioma samples, we learned that Wnt/β-catenin signal was overactivated in gliomas especially high-grade gliomas like GBM, and inversely related to PD-L1 expression and the CD8⁺ T-cell infiltration. This correlation was also found in the tumor tissues of GL261 homograft mouse model, which showed the enhanced CD4⁺/CD8⁺ T-cell infiltration and elevated PD-L1 expression under the inhibition of Wnt/β-catenin signal. The relationship between Wnt/β-catenin signal and T-cell infiltration is consistent with the recent integrative analysis of TCGA across different human tumor types indicating that non-T cell–inflamed tumors are enriched with the activation of tumor-intrinsic Wnt/β-catenin signal (19). The negative correlation between Wnt/β-catenin and PD-L1 expression was also found in different GBM cells. However, this finding is at odds with the recent publication by Du and colleagues, which showed that AKT-mediated β-catenin S552 phosphorylation upregulated the transcriptional expression of PD-L1 (45). This reminds us a complex relationship between β-catenin activation and PD-L1 expression. Intervention strategies based on different Wnt mediators and targets may have diverse effects on the expression of tumor PD-L1. Therefore, further in-depth study of the relationship between Wnt/β-catenin signal and PD-L1 expression is needed.

A previous study observed the methylation of gene promoter which encoding Wnt pathway inhibitor induced abnormal activation of Wnt pathway in malignant astrocytic gliomas, particularly in GBM (46). Another study identified an oncogene PLAGL2 which promoted GBM growth and self-renewing because of Wnt signaling activation (47). All these reports suggest a strong link between Wnt signaling and glioma development. Consistently, our in vitro and in vivo experiment results manifested that the blockade of Wnt signal inhibited GBM growth and migration, and promoted the therapeutic efficacy of PD-1 antibody in GL261 homograft model. This promoting effect was closely related to enhanced T-cell infiltration, T-cell cytotoxic function, and effector IFNγ secretion triggered by Wnt inhibition. Moreover, the upregulated expression of PD-L1 in tumor tissues from GL261 homograft model initiated by Wnt inhibition might improve the response to PD-1 blockade therapy.

It has been reported that some anti-PD-1/L1 responders developed acquired resistance or relapse under the pressure of PD-1/PD-L1 blockade therapy (48). In the current study, the GL261 model initially displayed good response to PD-1 antibody treatment. However, the tumors began to grow faster especially on 15 days after single administration of PD-1 antibody, which suggested that the model developed acquired resistance probably because of the defective immune infiltration and T-cell exhaustion. In contrast, the combination of Wnt inhibition sustained the durability of the PD-1 blockade therapy due to the enhanced anti-migration activity, T-cell infiltration and effector cytokine IFNγ release in tumors.

Above results support our speculation that glioma-intrinsic Wnt/β-catenin signal mediates immune escape and immune therapy resistance, and its blockade can inflame cold tumor and augment the effectiveness of anti-PD-1 by improving immune infiltration and PD-L1 expression. However, our study has some limitations. First, the current study focuses on the effect of glioma-intrinsic Wnt/β-catenin signal on T-cell function and infiltration. Therefore, the flank tumor model was used to discover this effect. To further confirm the results, an intracranial GBM model should be used to evaluate comprehensively the influence of BBB and TME in follow-up studies. Second, only one clinical sample was collected for each of high-, medium-, and low expression of β-catenin. More samples should be analyzed to generalize the finding about negative correlation between Wnt/β-catenin and PD-L1 expression as well as CD8⁺ T-cell infiltration.

In summary, our research first revealed the relationship between Wnt/β-catenin signal and immune infiltration as well as PD-L1 expression in gliomas through the analysis of TCGA data, clinical samples and cell experiments. Subsequently, in the GBM cell and Jurkat cell co-culture system, we confirmed that Wnt inhibition could promote Jurkat cells' cytotoxic function. Furthermore, the in vivo experiments revealed the blockade of Wnt/β-catenin not only restricted the proliferation and migration of GBM, but also promoted T-cell infiltration and effector cytokine IFNγ release, thereby enhancing the therapeutic efficacy of PD-1 antibody (Supplementary Fig. S4). Altogether, the current study provides the evidence that Wnt/β-catenin signal may serve as a prospective target to enhance the sensitivity of GBM to PD-1 blockade therapy.

No disclosures were reported.

H. Zhang: Resources, data curation, software, investigation, methodology, writing–original draft, writing–review and editing. Y. Bi: Data curation, funding acquisition, methodology. Y. Wei: Data curation, software, methodology. J. Liu: Methodology. K. Kuerban: Methodology. L. Ye: Supervision, funding acquisition, validation, methodology, writing–original draft, project administration, writing–review and editing.

This work was supported by Scientific and Innovative Action Plan of Shanghai (no. 20S11901600 and 18431902800), Shanghai Natural Science Foundation Project (no. 19ZR1446100), and MHHFDU-SPFDU Joint Research Fund (RO-MY201803 and RO-MY201714).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.