β-Catenin induces transcriptional expression of PD-L1 to promote glioblastoma immune evasion

The interaction between programmed cell death 1 (PD-1) on T cells and programmed cell death ligand 1 (PD-L1; also known as B7-H1) on tumor cells inhibits the activation, expansion, and effector functions of antigen-specific CD8⁺ T cells and promotes T cell apoptosis and the induction of regulatory T cells (T reg or suppressor T cells), allowing cancer cells to evade immune destruction (Boussiotis, 2016; Sharma et al., 2017).

Therapeutic targeting of immune checkpoints has resulted in tumor regression in melanoma and non–small cell lung cancer, and clinical trials are ongoing in glioblastoma (GBM) and several other types of cancer (Sharma and Allison, 2015). However, only a portion of patients experience a response to this immunotherapy (Sharma and Allison, 2015). The response to PD-1/PD-L1 blockade is reported to be correlated with PD-L1 expression levels in tumor cells (Herbst et al., 2014; Iwai et al., 2002). Although it has been reported that PD-L1 expression can be up-regulated at transcriptional (Casey et al., 2016; Dorand et al., 2016), translational (Coelho et al., 2017; Parsa et al., 2007), and posttranslational levels via inhibition of its protein ubiquitylation (Li et al., 2016; Lim et al., 2016; Zhang et al., 2018), it remains unclear whether the tumor microenvironment that activates growth factor receptor plays a role in the regulation of PD-L1 transcription and tumor immune evasion.

Aberrant activation of Wnt/β-catenin signaling has been detected in various types of cancer, such as breast, prostate, lung, and hematopoietic malignancies and GBM, and nuclear accumulation of β-catenin has been shown to be positively correlated with poor clinical outcomes, such as cancer progression, invasion, metastasis, and recurrence, resulting in low survival rates (Anastas and Moon, 2013; Kim et al., 2017; Nusse and Clevers, 2017). The binding of Wnt ligand to the low-density lipoprotein receptor-related protein-5/6 and Frizzled receptors results in the inhibition of GSK-3β–mediated phosphorylation of β-catenin. Stabilized β-catenin migrates to the nucleus, interacting with T cell–specific factor (TCF)/lymphoid enhancer–binding factor (LEF) and coactivators to activate Wnt target gene expression (MacDonald et al., 2009). Notably, Wnt-independent β-catenin transactivation has been implicated in tumor development (Lu and Hunter, 2004). Epidermal growth factor (EGF) receptor (EGFR) activation, resulting from the mutation or overexpression of EGFR in many types of human cancer, activates β-catenin (Ji et al., 2009; Lu et al., 2003; Yang et al., 2011). In addition to stabilizing β-catenin by phosphorylating GSK-3β, AKT, which is activated by growth factor receptors and phosphoinositide 3-kinase (PI3K) activation, phosphatase and tensin homolog (PTEN) mutation, or Wnt3A through mTORC2 activation (Esen et al., 2013), directly phosphorylates β-catenin at Ser552 to release its binding to adherens proteins, thereby promoting the nuclear translocation and activation of β-catenin (Fang et al., 2007; Ji et al., 2009; Lu et al., 2003).

In this study, we demonstrated that β-catenin activation induced by both Wnt ligand and activated EGFR results in the binding of the β-catenin/TCF/LEF complex to the promoter region of the CD274 gene to induce PD-L1 expression. Inhibition of AKT or β-catenin depletion reduced PD-L1 expression in tumor cells, enhanced CD8⁺ T cell activation and infiltration into tumors, and reduced tumor growth.

To determine the relationship between β-catenin activation and immune checkpoints, we performed immunohistochemical (IHC) analyses of tumor tissues from 20 patients with low-grade diffuse astrocytoma (World Health Organization grade II), 20 patients with anaplastic astrocytoma (grade III), and 50 patients with GBM (grade IV). We found that the levels of activated nuclear β-catenin, which were positively correlated with PD-L1 expression levels and glioma grades, were inversely correlated with CD8⁺ T cell infiltration (Fig. 1 A and,Fig. S1 A). Treatment of U251, EGFR-overexpressing U87 (U87/EGFR), and LN18 human GBM cells; glioma stem cell (GSC) 6–27 and GSC 7–11 human primary glioma-initiating cells; CT-2A and GL261 mouse glioma cells; DU145 human prostate cancer cells; and PC14 human lung adenocarcinoma cells with Wnt3A stabilized β-catenin protein and increased PD-L1 expression (Fig. 1 B). Flow cytometry analyses revealed that Wnt3A enhanced the PD-L1 protein expression level on the surface of U87/EGFR cells (Fig. 1 C).

Pretreatment of GBM cells with actinomycin D, a transcription inhibitor, reduced Wnt3A-enhanced PD-L1 expression (Fig. 1 D and Fig. S1 B), suggesting that gene transcription regulation of PD-L1 is involved in the response to Wnt3A treatment. Consistently, quantitative PCR analyses showed that Wnt3A enhanced mRNA levels of the CD274 gene (encoding PD-L1; Fig. 1 E) without obviously affecting CD274 mRNA stability (Fig. S1 C) and Ccl4 chemokine gene expression involved in T cell exclusion (Spranger et al., 2015; Fig. S1 D); this increase, together with the increased protein expression of PD-L1 and β-catenin, was attenuated by β-catenin shRNA expression (Fig. 1 F and Fig. S1 E). In addition, expression of the constitutively active β-catenin (CA β-catenin) mutant (Guo et al., 2012) increased CD274 mRNA expression (Fig. 1 G, top panel) and PD-L1 total expression (Fig. 1 G, bottom panel) and surface protein expression (Fig. 1 H). These results indicate that Wnt-induced β-catenin activation results in transcriptional up-regulation of PD-L1 in tumor cells.

EGFR activation enhances β-catenin activity (Ji et al., 2009; Lu et al., 2003; Yang et al., 2011). As expected, EGF treatment of GBM cells (Fig. 2 A, left panel) or expression of the active EGFRvIII mutant (Fig. 2 A, right panel), which lacks 267 amino acids in its extracellular domain and is commonly found in GBM as well as breast, ovarian, prostate, and lung carcinomas (Kuan et al., 2001; Yang et al., 2012a), increased CD274 mRNA and PD-L1 protein expression.

Pretreatment of U87/EGFR (Fig. 2 B) and U251 cells (Fig. S2 A) with actinomycin D greatly inhibited EGF-enhanced PD-L1 expression, and depletion of β-catenin reduced CD274 mRNA and PD-L1 protein expression levels in U87/EGFRvIII cells (Fig. 2 C) and U251 cells treated with EGF (Fig. S2 B). In addition, in the presence of actinomycin D, β-catenin depletion failed to obviously alter CD274 mRNA levels (Fig. S2 C), suggesting that β-catenin primarily regulates PD-L1 expression through transcriptional regulation of CD274 mRNA expression.

Notably, reconstituted expression of RNAi-resistant (r) WT β-catenin, but not the rβ-catenin S552A mutant, which is resistant to phosphorylation and activation by AKT (Fang et al., 2007), largely restored the reduced CD274 mRNA and PD-L1 protein expression in U87/EGFRvIII cells (Fig. 2 D), suggesting that AKT-dependent β-catenin activation is instrumental for EGFR activation-increased PD-L1 expression. Consistent with this result, expression of a constitutively active, myristoylation signal–attached AKT1 (myr-AKT1), which phosphorylates β-catenin at S552, enhanced CD274 mRNA and PD-L1 expression levels (Fig. 2 E). In contrast, treatment with AKT inhibitor MK2206 or PI3K inhibitor LY294002, which reduced β-catenin S552 phosphorylation, led to reduced CD274 mRNA and PD-L1 expression levels in U87/EGFRvIII cells (Fig. 2 F) and U251 cells treated with EGF (Fig. S2 D). Similarly, MK2206 treatment inhibited EGF-induced PD-L1 expression in GSC 6–27 and GSC 7–11 cells (Fig. S2 E).

Notably, the inhibitory effect of MK2206 on CD274 mRNA and PD-L1 protein expression in U87/EGFRvIII cells was abrogated by the expression of CA β-catenin (Fig. 2 G). In addition, overexpression of PTEN, which reduced the phosphorylation levels of AKT T308 and S473 and β-catenin S552, reduced CD274 mRNA and PD-L1 protein expression in EGF-treated PTEN-deficient U251 and U87 cells (Lee et al., 2017; Fig. 2 H). These results indicate that AKT activated by the EGFR-PI3K cascade phosphorylates β-catenin S552 and induces β-catenin–dependent up-regulation of CD274 transcription and PD-L1 expression in tumor cells.

AKT can also be activated by Wnt3A through mTORC2 activation (Esen et al., 2013). As expected, MK2206 treatment inhibited Wnt3A-induced AKT S473 phosphorylation, β-catenin S552 phosphorylation, and CD274 mRNA and PD-L1 expression in U87/EGFR cells (Fig. 2, I and J). These results indicate that AKT activation is also instrumental for Wnt3A-induced and β-catenin activation-mediated up-regulation of CD274 transcription and PD-L1 protein expression levels in tumor cells. The partial inhibition of Wnt3A-induced CD274 transcription and PD-L1 protein expression by MK2206 treatment is likely due to AKT-independent and Wnt-direct activation of β-catenin.

To identify the binding sequence of TCF/LEF in the promoter region of CD274, we analyzed the CD274 promoter and found that (-665)-CTT​TGA​GG-(-658) is a consensus sequence recognized by TCF/LEF (Cadigan and Waterman, 2012; Fig. 3 A). We constructed a luciferase reporter driven by the CD274 promoter that contains either the WT or a mutated potential TCF/LEF–binding sequence. As shown in Fig. 3 B, treatment of U87/EGFR cells with Wnt3A, EGF, or expression of CA β-catenin significantly increased the activity of the WT but not the mutated promoter. As expected, treatment with Wnt3A or EGF increased the binding of β-catenin to LEF-1 (Fig. S3 A), and depletion of LEF-1 (Fig. S3 B) reduced Wnt3A- or EGF-induced luciferase activity (Fig. S3 C). A chromatin immunoprecipitation (ChIP) assay with an anti–β-catenin or an anti–LEF-1 antibody showed that treatment of U87/EGFR cells with Wnt3A, EGF, or expression of CA β-catenin increased the binding of β-catenin and LEF-1 to the promoter region of CD274 (Fig. 3 C). MK2206 treatment reduced Wnt3A-, EGF-, or EGFRvIII-induced luciferase activity driven by the CD274 promoter (Fig. 3 D and Fig. S3 D) and also reduced the binding of β-catenin to the promoter region of CD274 in U87/EGFR (Fig. 3 E) and U87/EGFRvIII (Fig. S3 E) cells. A similar inhibitory effect was also observed via overexpression of PTEN in EGF-treated U251 and U87 cells (Fig. S3, F and G). In addition, restored expression of the β-catenin S552A mutant reduced Wnt3A- and EGF-induced and CD274 promoter-mediated luciferase activity and the binding of β-catenin to this promoter region in U87/EGFR cells (Fig. 3, F and G). These results indicate that the β-catenin/TCF/LEF complex binds to the CD274 promoter region in response to EGFR activation–, PTEN loss–, and Wnt3A-induced AKT activation and enhances CD274 transcription.

To determine the effect of β-catenin–regulated PD-L1 expression in tumor cells on T cell activation, we co-cultured Wnt3A-treated GL261 cells with PMA- and ionomycin-activated mouse primary CD8⁺ T cells and showed that GL261 cell co-culture inhibited IL-2 (Fig. 4 A) and IFNγ (Fig. S4 A) mRNA expression levels in mouse primary CD8⁺ T cells. This inhibition was significantly further enhanced by the expression of PTEN shRNA or CA β-catenin expression in GL261 cells (Fig. 4 A and Fig. S4 A). Similarly, Wnt3A-treated U251 cell co-culture inhibited IL-2 mRNA expression in PMA- and ionomycin-activated Jurkat T cells (Fig. S4 B), and this inhibition was alleviated by the expression of PTEN or β-catenin shRNA in U251 cells (Fig. S4 B).

GL261 cells predominantly express MHC class I (MHC-I) molecules (Garg et al., 2016). We co-cultured MHC-I–depleted GL261 cells with activated CD4⁺ or CD8⁺ T cells. ELISpot analyses showed that depletion of MHC-I by expressing an shRNA of H2-K^D, a gene encoding a subunit of the MHC-I complex, in GL261 cells reduced IL-2 and IFNγ production by CD8⁺ T cells but not CD4⁺ T cells (Fig. S4, C and D). Notably, this reduction was exacerbated by PTEN depletion. These results suggest that CD8⁺ T cell activation can be regulated by both PD-L1 and MHC-I protein expression in tumor cells.

We next examined the role of β-catenin–regulated PD-L1 expression in tumorigenesis and tumor immune evasion in mice. We intracranially injected luciferase-expressing GL261 mouse glioma cells with or without PTEN depletion, CA β-catenin expression, or β-catenin depletion into syngeneic C57BL/6 mice. β-Catenin depletion greatly reduced brain tumor growth (Fig. 4, B and C; and Fig. S4 E) and prolonged mouse survival (Fig. 4 D), whereas PTEN depletion and CA β-catenin expression promoted tumor growth (Fig. 4, B and C; and Fig. S4 E) and shortened mouse survival (Fig. 4 D). IHC staining of tumor tissue showed that PTEN depletion enhanced AKT S473 phosphorylation and β-catenin S552 phosphorylation (Fig. S4 F). In addition, PTEN depletion or CA β-catenin expression increased PD-L1 expression, reduced CD8⁺ T cell infiltration into the tumor tissue (Fig. 4 E), and decreased granzyme B expression, indicating reduced cytotoxic T cell activity (Fig. S4 F). In contrast, β-catenin depletion decreased PD-L1 expression, accompanied by increased CD8⁺ T cell accumulation in the tumor tissue (Fig. 4 E) and increased granzyme B expression (Fig. S4 F). Notably, overexpression of Flag-tagged PD-L1 in β-catenin–depleted GL261 cells largely restored tumor growth and decreased CD8⁺ T cell infiltration (Fig. S4 G). Similar to the effect on tumor growth derived from GL261 cells, β-catenin depletion in CT-2A mouse glioma cells also greatly reduced brain tumor growth (Fig. S4 H, top panel of the mouse images and left panel of the graph) with decreased PD-L1 expression and increased CD8⁺ T cell infiltration (Fig. S4 I).

AKT activation is instrumental for β-catenin–regulated PD-L1 expression, suggesting that AKT inhibition can overcome immune evasion by tumor cells. Intraperitoneal injection of MK2206 or intracranial and intratumoral injection of an anti–PD-1 antibody inhibited GL261 cell–derived tumor growth (Fig. 4, F and G; and Fig. S4 J) and prolonged mouse survival (Fig. 4 H). Similar results were also obtained for CT-2A cell–derived tumor growth (Fig. S4 H, top panel of the mouse images and left panel of the graph) and mouse survival (Fig. S4 K). In addition, an additive effect on tumor growth inhibition and improved mouse survival was observed by combined treatment with MK2206 and the anti–PD-1 antibody (Fig. 4, F–H). Notably, systemic depletion of CD8⁺ T cells using an anti-CD8 antibody treatment (Fig. S4 L) resulted in a reduction in the inhibitory effect of MK2206, anti–PD-1 antibody treatment, and β-catenin depletion on tumor growth (Fig. S4 H, bottom panel of the mouse images and right panel of the graph). Intriguingly, MK2206 treatment-induced inhibition of tumor growth, PD-L1 expression, and CD8⁺ T cell infiltration was abrogated by the expression of active β-catenin in CT-2A cells (Fig. S4 M).

IHC staining of tumor tissue showed that MK2206 treatment inhibited AKT S473 phosphorylation and β-catenin S552 phosphorylation (Fig. S4 N). In addition, MK2206 or an anti–PD-1 antibody treatment reduced PD-L1 expression accompanied by more CD8⁺ T cell infiltration into the tumor tissue (Fig. 4 I) with increased granzyme B expression (Fig. S4 N). These effects were further enhanced by combined treatment with MK2206 and the anti–PD-1 antibody. Together, these results suggest that AKT activates β-catenin, resulting in up-regulated PD-L1 expression and subsequent promotion of tumor cell immune evasion and brain tumor growth.

To determine the clinical relevance of β-catenin–regulated PD-L1 expression, we analyzed The Cancer Genome Atlas (TCGA) data and revealed that GBMs with PTEN mutations, which account for ∼30% of GBMs (Cancer Genome Atlas Research Network, 2008; Riddick and Fine, 2011), had higher levels of CD274 mRNA expression than those with WT PTEN (Fig. 5 A). In addition, the expression levels of the β-catenin downstream genes LEF-1 and PPARδ (Hupe et al., 2017) were positively correlated with CD274 mRNA expression levels (Fig. S5 A). Consistent with these results, immunoblotting analyses of 10 human GBM specimens showed that AKT S473 phosphorylation, β-catenin S552 phosphorylation, and PD-L1 expression levels were inversely correlated with PTEN expression levels (Fig. 5 B). We also analyzed an additional 39 human GBM specimens with WT PTEN expression with immunoblotting assays and revealed that the levels of EGFR Y1068 phosphorylation, AKT S473 phosphorylation, β-catenin S552 phosphorylation, and PD-L1 expression were positively correlated with each other (Fig. S5 B). These results suggest that AKT activation resulting from mutation of PTEN or EGFR activation correlates with the levels of AKT-mediated β-catenin S552 phosphorylation and PD-L1 expression in human GBM.

We next performed IHC analyses of 50 human GBM tissues; the levels of β-catenin S552 phosphorylation were positively correlated with PD-L1 expression and inversely correlated with CD8⁺ T cell infiltration (Fig. 5, C and D). In addition, immunofluorescent analyses showed that active β-catenin and high PD-L1 expression occurred in the same cells of GBM specimens (Fig. S5 C). These results support a role of β-catenin–regulated PD-L1 expression in tumor cell immune evasion in human GBM.

Treatments that target immune checkpoints mediated by PD-1 and PD-L1 have been approved for human cancer, with durable clinical benefits (Cabodevilla et al., 2013; Sharma and Allison, 2015). However, the majority of cancer patients are resistant to anti–PD-1/PD-L1 treatment, and the underlying mechanisms of this resistance remain unclear. We demonstrated here that EGFR activation and active Wnt signaling, which occur in a large variety of cancers, induce transcriptional up-regulation of PD-L1 that is directly mediated by the β-catenin/TCF/LEF transcriptional complex. Importantly, AKT activation, which occurs in most cancer cells and is regulated by growth factors, including Wnt ligands, PI3K activation, and PTEN loss, promotes β-catenin–dependent PD-L1 expression and tumor immune evasion (Fig. 5 E). Active β-catenin signaling in melanoma cells promotes T cell exclusion through decreased CCL4 production in tumor cells (Spranger et al., 2015). However, the Wnt–β-catenin pathway does not affect Ccl4 gene expression in GBM cells, suggesting that CCL4 is not involved in β-catenin–promoted T cell exclusion in the GBM microenvironment.

Inhibition of AKT with MK2206, which is currently in clinical trials for the treatment of cancer (Konopleva et al., 2014), or β-catenin depletion reduces PD-L1 expression levels in tumor cells, enhances CD8⁺ T cell infiltration, and inhibits tumor growth. Alleviation of these effects by the systemic depletion of CD8⁺ T cells further supports a critical role of AKT- and β-catenin transactivation–regulated PD-L1 expression in tumor immunogenicity. As expected, MK2206 combined with an anti–PD-1 antibody, which causes more profound abrogation of immune checkpoint blockade, significantly enhanced CD8⁺ T cell infiltration, and blocked tumor growth. Hence, our findings elucidate a novel mechanism underlying PD-L1 up-regulation that is controlled by widely activated Wnt signaling, EGFR, and AKT and by frequently mutated PTEN; our results also provide a molecular basis for improving the clinical response rate and efficacy of targeted therapy and PD-1/PD-L1 blockade therapy in cancer patients.

Normal rabbit immunoglobulin (sc-2027), normal mouse immunoglobulin (sc-2025), anti–H2-K^D antibody (sc-53852; 1:200 for immunoblotting), and anti–β-catenin antibody (sc-7963; 1:200 for immunoblotting and 1:100 for immunoprecipitation) were purchased from Santa Cruz Biotechnology. Rabbit polyclonal antibodies that recognize human β-catenin (9566; 1:1,000 for immunoblotting), active β-catenin (8814; 1:800 for IHC analysis), human PD-L1 (E1L3N, 13684; 1:1,000 for immunoblotting), AKT (pT308, 4056; 1:1,000 for immunoblotting), AKT (pS473, 4060; 1:1,000 for immunoblotting), AKT (9272; 1:1,000 for immunoblotting), PTEN (9559; 1:1,000 for immunoblotting), LEF-1 (76010; 1:1,000 for immunoblotting and 1:100 for immunoprecipitation), and mouse CD8α (98941; 1:400 for IHC analysis) were purchased from Cell Signaling Technology. Mouse monoclonal antibodies for FLAG (F3165, clone M2; 1:5,000 for immunoblotting), HA (H6908; 1:5,000 for immunoblotting), tubulin (T6074, clone B-5-1-2; 1:5,000 for immunoblotting), PMA (P8139), ionomycin (I3909), and actinomycin D (A1410) were purchased from Sigma-Aldrich. Monoclonal antibody for human CD8 (144B, ab17147; 1:50 for IHC analysis) and polyclonal antibody for Granzyme B were obtained from Abcam. Rat anti-mouse PD-1 (BE0146) and anti-mouse PD-L1 (BE0101) were obtained from Bio X Cell. PD-L1 PE-conjugated antibody (MIH1, 12–5983-42; 1:20 for flow cytometry analysis) was purchased from Thermo Fisher Scientific. Human recombinant EGF (01–407) was obtained from EMD Millipore. Puromycin (540222) and G418 (345810) were purchased from EMD Biosciences. Wnt3A (5036-WN-010) was purchased from R&D Systems. D-luciferin (14681) was obtained from Cayman. HyFect transfection reagents (E2650) were obtained from Denville Scientific. LY294002 (L-7988) was purchased from LC Laboratories. MK-2206 (S1078) was purchased from Selleck Chemicals.

DU145, PC14, and GBM cells, including U251, U87, LN18, EGFR-overexpressing U87 (U87/EGFR), EGFRvIII-overexpressing U87 (U87/EGFRvIII), CT-2A, GL261, and firefly luciferase (FL)–expressing CT-2A and GL261 (CT-2A–FL or GL261-FL) cells were maintained in DMEM or DMEM/F12 supplemented with 10% bovine calf serum (HyClone); these cells are routinely tested for mycoplasma. Human primary GBM cells were maintained in DMEM/F-12 50/50 supplemented with B27, EGF (10 ng/ml), and basic fibroblast growth factor (10 ng/ml). U87 and U251 cells were authenticated using short tandem repeat profiling at The University of Texas MD Anderson Cancer Center Characterized Cell Line Core Facility. Cells were plated at a density of 4 × 10⁵ per 60-mm dish or 10⁵ per well of a 6-well plate 18 h before transfection. Transfection was performed using HyFect transfection reagent (Denville Scientific) according to the manufacturer’s instructions.

PCR-amplified human β-catenin and PD-L1 were cloned into pcDNA3.1/hygro(+)–Flag. The primers for expression plasmid construction were as follows: human PD-L1, 5′-CGG​GAT​CCA​TGA​GGA​TAT​TTG​CTG​TCT​T-3′ (forward) and 5′-CCG​CTC​GAG​GCC​GTC​TCC​TCC​AAA​TGT​GTA​T-3′ (reverse) and human β-catenin, 5′-CGG​GAT​CCA​TGG​CTA​CTC​AAG​CTG​ATT​TGA​TG-3′ (forward) and 5′-CCG​CTC​GAG​GCC​AGG​TCA​GTA​TCA​AAC​CAG​GCC​A-3′ (reverse). Human PTEN was cloned by using the following primers: 5′-CGG​GAT​CCA​TGA​CAG​CCA​TCA​TCA​AAG​AGA​T-3′ (forward) and 5′-GAC​TTT​TGT​AAT​TTG​TGT​ATG​CTG-3′ (reverse). Then, the PTEN coding sequence was inserted into the PCDH-CMV-MCS-EF1-Puro-SFB vector. pLV/β-catenin deltaN90 (CA β-catenin) and pECE-Myr-HA-AKT1 (delta4-129) were purchased from Addgene. pcDNA3.1/hygro(+)–Flag β-catenin S552A was created using the QuikChange site-directed mutagenesis kit (Stratagene).

The following pGIPZ shRNAs were used: control shRNA oligonucleotide, 5′-GCT​TCT​AAC​ACC​GGA​GGT​CTT-3′; mouse PTEN shRNA oligonucleotide, 5′-TAT​AGG​TCA​AGT​CTA​AGT​C-3′; human β-catenin shRNA#1 oligonucleotide, 5′-TTA​CCA​CTC​AGA​GAA​GGA​G-3′, and human β-catenin shRNA#2 oligonucleotide, 5′-TGT​GAT​CCA​TTC​TTG​TGC​A-3′; human LEF-1 shRNA oligonucleotide, 5′-GCT​GGT​CTG​CAA​GAG​ACA​ATT-3′; and mouse H2-K^D shRNA oligonucleotide, 5′-AAG​AGC​GAT​GAG​CAG​TGG​TTC-3′.

Total RNA isolation, RT, and real-time PCR were conducted as described previously (Yang et al., 2012b). The following primer pairs were used: human CD274, 5′-CTG​CAC​TTT​TAG​GAG​ATT​AGA​TC-3′ (forward) and 5′-CTA​CAC​CAA​GGC​ATA​ATA​AGA​TG-3′ (reverse); human IL-2, 5′-GAA​TGG​AAT​TAA​TAA​TTA​CAA​GAA​TCC​C-3′ (forward) and 5′-TGT​TTC​AGA​TCC​CTT​TAG​TTC​CAG-3′ (reverse); human Ccl4, 5′-CTC​TCA​GCA​CCA​ATG​GGC​TC-3′ (forward) and 5′-TCT​TTT​GGT​TTG​GAA​TAC​CAC​AG-3′ (reverse); mouse CD274, 5′-GTG​AAA​CCC​TGA​GTC​TTA​TCC-3′ (forward) and 5′-GAC​CAT​TCT​GAG​ACA​ATT​CC-3′ (reverse); mouse Ccl4, 5′-GTT​CTC​AGC​ACC​AAT​GGG​CTC-3′ (forward) and 5′-TCT​TTT​GGT​CAG​GAA​TAC​CAC​AG-3′ (reverse); mouse IL-2, 5′-GCT​GTT​GAT​GGA​CCT​ACA​GGA-3′ (forward) and 5′-ATC​CTG​GGG​AGT​TTC​AGG​TT-3′ (reverse); β-actin, 5′-ATG​GAT​GAC​GAT​ATC​GCT​GCG​C-3′ (forward) and 5′-GCA​GCA​CAG​GGT​GCT​CCT​CA-3′ (reverse); and mouse IFNγ, 5′-GCA​TCT​TGG​CTT​TGC​AGC​T-3′ (forward) and 5′-CCT​TTT​TCG​CCT​TGC​TGT​TG-3′ (reverse).

Extraction of proteins from cultured cells was performed using lysis buffer (50 mM Tris-HCl [pH 7.5], 0.1% SDS, 1% Triton X-100, 150 mM NaCl, 1 mM dithiothreitol, 0.5 mM EDTA, 100 µM PMSF, 100 µM leupeptin, 1 µM aprotinin, 100 µM sodium orthovanadate, 100 µM sodium pyrophosphate, and 1 mM sodium fluoride). Cell extracts were clarified via centrifugation at 13,400 g, and the supernatants (2 mg protein/ml) were subjected to immunoblot analysis with the corresponding antibodies. The band intensity was quantified using the Image Lab software program (Bio-Rad). Each experiment was repeated at least three times.

Cells were washed in PBS and stained for cell surface PD-L1 on ice for 30 min in FACS blocking buffer (0.5% BSA and 2% FBS in PBS). PD-L1 PE-conjugated antibody was added to the cells on ice for 30 min. After incubation, the cells were washed with FACS buffer (0.5% BSA in PBS) three times and detected by a BD FACSCelesta (BD Biosciences) cytometer. The results were analyzed by FlowJo v10.

The human CD274 promoter was cloned from the genomic DNA of U87 cells. The following primers were used: 5′-CGG​GGT​ACC​ATG​TAG​CTC​GGG​ATG​GGA​AGT-3′ (forward) and 5′-CCG​CTC​GAG​CTC​TGC​CCA​AGG​CAG​CAA​ATC-3′ (reverse). The amplified clone corresponding to the human CD274 promoter (600 bp upstream of the transcription start site) containing the TCF/LEF sequence was inserted into the KpnI/XhoI sites of the pGL3-Basic vector (Promega). Site-directed mutagenesis was performed using 5′-GTC​GAG​GAA​CTG​GAA​GGA​AGT​CAC​AG-3′ (forward) and 5′-CTG​TGA​CTT​CCT​TCC​AGT​TCC​TCG​AC-3′ (reverse) primers and verified by sequencing. The tumor cells were cotransfected with 1 µg of pGL3 empty vector, pGL3 WT TCF/LEF, or pGL3 TCF/LEF mutant vectors and 0.5 µg of pRL-TK vector (as an inner control that contains Renilla luciferase sequences; Promega) by Lipofectamine 2000 (Invitrogen) and grown under different experimental conditions. After incubation, firefly and Renilla luciferase activities were measured using the dual-luciferase reporter assay (Promega), and the ratio of firefly/Renilla luciferase was determined.

A ChIP assay was performed using the SimpleChIP Enzymatic Chromatin IP kit (9003; Cell Signaling Technology). Chromatin prepared from cells (in a 10-cm dish) was used to determine the total DNA input and was incubated overnight with specific antibodies or with normal rabbit or mouse IgG. The immunoprecipitated chromatin was analyzed using a CFX96 real-time PCR detection system (Bio-Rad). The PCR primer sequences for the CD274 promoter were 5′-ATG​TAG​CTC​GGG​ATG​GGA​AGT-3′ (forward) and 5′-TGT​GTG​TGT​GTG​TAT​GGG​TGT​A-3′ (reverse).

GL261 or U251 cells, with or without expression of PTEN shRNA, S protein tag/Flag tag/Streptavidin binding peptide (SFB)–PTEN, CA β-catenin, or β-catenin shRNA, were treated with Wnt3A (20 ng/ml) for 12 h before being co-cultured with mouse primary CD8⁺ T cells or Jurkat T cells that had been activated with PMA (20 ng/ml) and ionomycin (500 ng/ml) for 12 h at a ratio of 4:1 (T cells:tumor cells) for 24 h. The T cells were then collected to analyze IL-2 mRNA expression.

The 10⁸ GL261 cells were collected and resuspended in DMEM. These cells were then frozen in a −80°C freezer for 1 h and thawed at 37°C for three cycles to lyse the cells. The solution was centrifuged at 300 rpm for 30 min, and the supernatants were collected and stored at 4°C for further use as tumor-specific antigens.

The mice were euthanized, and the spleens were collected. The spleens were disassociated by squeezing, and all released and suspended cells were collected. The suspended cells were treated with red cell lysis buffer (Beyotime), and the splenocytes were centrifuged at 400 g for 3 min. The splenocytes were cultured with GL261 cell lysates (1:3 for the cell numbers) for 48 h. The lymphocytes were isolated from cultured splenocytes using gradient centrifugation on histopaque-1077. CD4⁺ T and CD8⁺ T cells were isolated from lymphocytes according to the instructions of CD4⁺ or CD8⁺ T cell isolation kits (Miltenyi Biotec).

GL261 cells expressing a control shRNA or H2-K^D shRNA (MHC-I subunit) were infected with lentivirus expressing PTEN shRNA. CD4⁺ or CD8⁺ T cells were co-cultured with GL261 cells at a ratio of 3:1 for 24 h, and CD4⁺ or CD8⁺ T cells were then collected and subjected to ELISpot analysis with an ELISpot kit (R&D Systems) containing plates coated with an anti–IFNγ or an anti–IL-2 antibody.

We intracranially injected 10⁵ GL261-FL cells (in 5 µl of DMEM/F12 per mouse), with or without expression of PTEN shRNA, CA β-catenin, or β-catenin shRNA, into 4-wk-old male C57BL/6 mice. We divided the mice into two groups. One group consisted of seven mice in each subgroup and was used to monitor tumor growth and perform IHC analyses. The mice were euthanized 15 d after the GL261-FL cells were injected. The brain of each mouse was harvested, fixed in 4% formaldehyde, and embedded in paraffin. Tumor formation and phenotype were determined by bioluminescence imaging and histological analysis of hematoxylin and eosin–stained sections.

The other groups, which consisted of nine mice in each subgroup, were monitored for survival. The results were analyzed using STATISTICA software and presented in Kaplan-Meier plots. Humane endpoints included weight loss of 20–25%, weakness that prevented them from obtaining food or water, loss of appetite (anorexia for 24 h), a moribund state, and an inability to participate in normal activities because of tumor growth. All mice were euthanized under anesthesia after two or more of these humane endpoints had been observed.

For treatment with MK2206 and antibody, 100 mg of anti–PD-1 antibody (RMP1-14; Bio X Cell) or rat IgG (Bio X Cell) as a control was injected intratumorally on days 3, 7, 11, and 15 after GL261-FL cell inoculation, and MK2206 (120 mg/kg) was injected intraperitoneally on days 3, 5, 7, 9, 11, 13, and 15.

All of the mice were housed in the MD Anderson Cancer Center animal facility (Houston, TX); all experiments were performed in accordance with relevant institutional and national guidelines and regulations and were approved by the Institutional Animal Care and Use Committee at the MD Anderson Cancer Center.

C57BL/6 male mice were intraperitoneally injected with 200 mg anti-CD8 antibody (Clone 2.43) or rat IgG isotype. To determine the efficiency of depletion, spleens were collected from the mice at day 3 after injection, and then the lymphocytes were isolated from the spleen cells by using gradient centrifugation on histopaque-1077 (Sigma-Aldrich). The lymphocytes were incubated with anti-CD3e antibody (145-C11; Thermo Fisher Scientific), anti-CD4 antibody (GK1.5; Thermo Fisher Scientific), and anti-CD8 antibody (53–6.7; Thermo Fisher Scientific), and then the amounts of subpopulation were quantified by flow cytometry.

C57BL/6 mice were intraperitoneally injected with 150 mg/kg D-luciferin in PBS 10 min before bioluminescence imaging. Imaging was conducted using a charge-coupled device camera (IVIS 100, Xenogen; exposure time of 1 min, binning of 8, field of view of 15 cm, f/stop of 1, and no filter). Mice were anesthetized with isoflurane (2% vaporized in O₂) and shaved to minimize attenuation of the signal by pigmented black hair. For analysis, total photon flux (photons per second) was measured from a fixed region of interest using Living Image software (Xenogen). Bioluminescent signals within the fixed region of interest were normalized to the background luminescence and were obtained over the same region of interest from animals that had not been injected with D-luciferin.

IHC staining was performed using the VECTASTAIN ABC kit (Vector Laboratories) according to the manufacturer’s instructions. Sections of paraffin-embedded xenograft tumors were stained with antibodies against PD-L1 (mouse), AKT (pS473), β-catenin (pS552), CD8α (mouse), Granzyme B, and nonspecific IgG as a negative control. The paraffin-embedded human GBM tumor sections were stained with antibodies against β-catenin (pS552), human PD-L1, active β-catenin, and CD8 (human) with nonspecific IgG as a negative control. We quantitatively scored the tissue sections according to the percentage of positive cells and staining intensity. We assigned the following proportion scores: 0 if 0% of the tumor cells showed positive staining, 0.1–1.0 if 0.1–1% of cells showed positive staining, 1.1–2.0 if 1.1–10% showed positive staining, 2.1–3.0 if 11–30% showed positive staining, 3.1–4.0 if 31–70% showed positive staining, and 4.1–5.0 if 71–100% showed positive staining. We rated the intensity of staining on a scale of 0 to 3: 0, negative; 1, weak; 2, moderate; and 3, strong. We then combined the proportion and intensity scores to obtain a total score (range, 0–8). The total number of intratumoral CD8⁺ T cells was counted in six low-power fields and presented as an average number of CD8⁺ T cells per low-power field. All GBM patients from whom the tumor samples were obtained had undergone standard radiotherapy after surgery followed by treatment with an alkylating agent (temozolomide in most cases). Consent for collection of patient specimens, the use of human brain tumor specimens, and the database was approved by the MD Anderson Institutional Review Board.

The mRNA expression of CD274, LEF-1, and PPARδ and somatic mutations in PTEN in GBM specimens were downloaded from cBioPortal (https://www.cbioportal.org/; Cerami et al., 2012). The mRNA expression of CD274, LEF-1, and PPARδ was normalized to the Z-score, and the correlative expression of these genes was analyzed.

According to somatic mutations in PTEN, GBM samples were classified into two groups: samples with and without nonsynonymous mutations in PTEN. Nonsynonymous mutations include missense mutations, nonsense mutations, splicing mutations, in-frame insertions and deletions, and frameshift insertions and deletions. The Wilcoxon rank-sum test was used to compare the differences in CD274 mRNA expression between the two groups.

All quantitative data are presented as the mean ± SD of at least three independent experiments. A two-group comparison was conducted using a two-sided, two-sample Student’s t test. A simultaneous comparison of more than two groups was conducted using one-way ANOVA (SPSS statistical package, version 12; SPSS Inc.). Values of P < 0.05 were considered statistically significant.

Fig. S1 shows that Wnt-induced β-catenin activation results in PD-L1 up-regulation in tumor cells. Fig. S2 shows that AKT activation induced by EGFR-dependent PI3K activation, PTEN loss, and Wnt3A enhances PD-L1 expression in a β-catenin–dependent manner. Fig. S3 shows that the β-catenin/TCF/LEF complex binds to the CD274 promoter region in response to AKT activation and enhances CD274 transcription. Fig. S4 shows that β-catenin activation promotes immune evasion of tumor cells and brain tumor growth. Fig. S5 shows that PD-L1 expression is positively correlated with the levels of β-catenin S552 phosphorylation in human GBM.

This work was supported by the Zhejiang University Research Fund (188020*194221901/029; to Z. Lu); the Leading Innovative and Entrepreneur Team Introduction Program of Zhejiang (2019R01001; to Z. Lu); the National Research Foundation of Korea grant funded by the Korean government (MIST; 2020R1C1C1011350); the Bio&Medical Technology Development Program funded by the Ministry of Science, ICT and Future Planning, Republic of Korea (NRF-2017M3A9E4047243); National Natural Science Foundation of China (81902151; to L. Du); and Zhejiang Natural Science Foundation grant LQ19H160010 (to L. Du). Z. Lu is a Kuancheng Wang Distinguished Chair.

Author contributions: Z. Lu, J. Lyu, L. Du, and J.-H. Lee conceived and designed the study. L. Du, J.-H. Lee, H. Jiang, C. Wang, S. Wang, D. Xu, J. Li, X. Qian, and X. Li performed experiments. Z. Zheng and F. Shao performed TCGA database analysis. Y. Xia, Y. Zheng, H.-R. Kim, D. Xing, and P. Liu provided reagents and technical assistance. L. Du, J.-H. Lee, J. Lyu, and Z. Lu wrote the paper with comments from all authors. Z. Lu and J. Lyu supervised the entire project.