Inhibition of ICAM1 diminishes stemness and enhances antitumor immunity in glioblastoma via β-catenin/PD-L1 signaling

Abstract

Glioblastoma (GBM) stem cells (GSCs) are pivotal in tumor initiation, recurrence, and therapeutic resistance, underscoring their critical role in the complex pathology of GBM. Despite their recognized importance, the mechanisms by which GSCs facilitate immune evasion, especially in emerging immunotherapies, remain incompletely understood. Here, we identify intercellular adhesion molecule 1 (ICAM1) as a key regulator of GSC stemness and tumorigenicity, promoting an immunosuppressive microenvironment via β-catenin/PD-L1 signaling. Mechanistically, ICAM1 interacts with ZNRF3, leading to its autoubiquitination and clearance, stabilizing LRP6, and activating β-catenin signaling, which upregulates PD-L1 expression. Combined treatment with anti-ICAM1 and anti-PD-1 antibodies results in the most effective tumor inhibition and significantly extends survival in ICAM1-overexpressing GBM models. CyTOF and flow cytometry analyses reveal that ICAM1 overexpression reduces cytotoxic CD8⁺ T cell populations via PD-L1/PD-1 interactions, reversible by PD-1 blockade. Our findings highlight the co-targeting of ICAM1 and PD-1 as a promising strategy against immune evasion in GBM.

Fig. 1. ICAM1 is associated with poor prognosis of GBM patients.

a A schematic representation illustrating the screening process that identified ICAM1 as a therapeutic target in GBM. b Genetic alteration status of candidate genes across various brain tumor types from the TCGA database. Created in BioRender. Zheng, Y. (2025) BioRender.com/8tkn3ep. c ICAM1 expression in non-tumor versus GBM tissues from the TCGA database. Data are presented as min to max, median, and the 25th and 75th percentiles, with non-tumor (n = 10) and GBM (n = 528) samples as biological replicates. d ICAM1 expression in different brain tumor types from the TCGA database. Data are presented as min to max, median, and the 25th and 75th percentiles, with astrocytoma (n = 253), oligodendroglioma (n = 204) and GBM (n = 152) samples as biological replicates. e Kaplan-Meier survival curves for GBM patients with high (n = 261) or low (n = 264) ICAM1 expression based on the median expression in the TCGA dataset. f Kaplan-Meier survival curves for both GBM and LGG patients with high (n = 305) or low (n = 303) ICAM1 expression based on the median expression in the TCGA dataset. g Kaplan-Meier survival curves for LGG patients with high (n = 137) or low (n = 144) ICAM1 expression based on the median expression in the TCGA dataset. h Correlation analysis for ICAM1 mRNA expression and ImmuneScore calculated using the ESTIMATE algorithm. i Immunoblot (IB) analysis of ICAM1 expression in GSCs. GAPDH was used as a loading control. j Representative flow cytometry analysis of ICAM1⁺ cell proportion in GSCs (n = 3, biological replicates). Statistics: c Two-tailed Student’s t-test. d One-way ANOVA with Tukey’s multiple comparisons test. e–g Log-rank test. h Two-sided Pearson correlation test (n = 152 patients’ samples). Image: i Representative blots (n = 3), GAPDH was used as a loading control. Source data are provided as a Source Data file.

Fig. 2. ICAM1 regulates the functional traits of GSCs.

RT-qPCR (a) and IB (b) analysis of ICAM1, Nestin, and CD44 expression in 83 GSC infected with lentivirus expressing shICAM1 or shCtrl. RT-qPCR (c) and IB (d) analysis of ICAM1, Nestin, and Sox2 expression in 528 GSC infected with shICAM1-1/−2 or shCtrl lentivirus. Limiting dilution assays (LDAs) performed using 83 (e) and 528 GSCs (f) infected with shICAM1-1/−2 or shCtrl lentivirus (n = 2, biological replicates). RT-qPCR (g) and IB(h)analysis of ICAM1, Nestin, and Sox2 expression in X01 GSC infected with lentivirus expressing ICAM1 or vector control. RT-qPCR (i) and IB (j) analysis of ICAM1, Nestin, and CD44 expression in 131 GSC infected with ICAM1 or vector control lentivirus. LDAs performed using X01 (k) and 131 (l) GSCs infected with ICAM1 or vector control lentivirus (n = 2, biological replicates).m,n IB analysis of ICAM1, Nestin, CD44, and Sox2 in 83 and 528 GSCs cultured without serum (day 0) and with 10% serum. Serum-cultured 83 and 528 GSCs were harvested at the indicated times (days 3, 5, and 7). o IB analysis of Icam1-Flag, Nestin, Sox2, and CD44 in serum-free cultured GL261 infected with lentivirus expressing Icam1-Flag or vector control. p LDAs performed using serum-free cultured GL261 infected with Icam1-Flag or vector control lentivirus (n = 2, biological replicates). q Changes in body weight of mice implanted with 83-shCtrl, 83-shICAM1-1/−2 cells (1 × 10⁴ cells/mouse). r Representative H&E staining of the whole brains of mice as in (q). Scale bar: 500 μm. s Kaplan-Meier survival curves of the orthotopic xenograft mouse model as in (q). MST, median survival time. Statistics: a, c, g, i Data are presented as mean ± SD (n = 3, biological replicates), two-tailed Student’s t-test. q Data are presented as mean ± SD (n  =  6 mice/group), day 21, shCtrl vs. shICAM1-1/−2, one-way ANOVA with Tukey’s multiple comparisons test. s, MST, median survival time. Log-rank test (n  =  6 mice/group). Images: b, d, h, j, m–o Representative blots (n = 3, biological replicates). The samples derive from the same experiment but different gels for Nestin and CD44, and another for ICAM1 with GAPDH (b, j, m) were processed in parallel. The samples derive from the same experiment but different gels for Nestin with Sox2, and another for ICAM1 with GAPDH (d, h, n) were processed in parallel. The samples derive from the same experiment but different gels for Nestin with Sox2, another for ICAM1 with GAPDH, and another for CD44 (o) were processed in parallel. GAPDH was used as a loading control. Source data are provided as a Source Data file.

Fig. 3. ICAM1 regulates GSC stemness, proliferation, and tumor progression through the Wnt/β-catenin signaling pathway.

a Enrichment analysis in ICAM1^high vs. ICAM1^low group from CGGA and TCGA datasets for the indicated pathways. The differentiation of ICAM1 expression is based on the extreme 2% thresholds at both ends of its expression distribution. NES: normalized enrichment score. b IB analysis of ICAM1, β-catenin, cyclin D1 and Axin2 expression in 83 GSC infected with shICAM1-1/−2 or shCtrl lentivirus. c IB analysis of β-catenin in fractionated nuclear and cytosolic lysates from 83 GSC infected with shICAM1-1/−2 or shCtrl lentivirus. Lamin B and α-Tubulin were used as markers for the nucleus and cytoplasm, respectively. d Dual-luciferase activity assay showing the effect of varying concentrations of ICAM1 on the reporter genes containing TCF/LEF-binding sequences in 293 T cells, 24 h post-transfection. IB analysis of ICAM1, Nestin, β-catenin, and cyclin D1 expression in X01 (e) and 131 (f) GSCs infected with ICAM1 or vector control lentivirus, and treated with ICG001 (10 µM) or vehicle. LDAs performed using X01 (g) and 131 (h) GSCs infected with ICAM1 or vector control lentivirus, and treated with ICG001 (10 µM) or vehicle (n = 2, biological replicates). Cell proliferation assays using X01 (i) and 131 (j) GSCs infected with ICAM1 or vector control lentivirus, and treated with ICG001 (10 µM) or vehicle. k Luminescence images of mice bearing orthotopic xenografts of X01-luc GSC infected with ICAM1 or vector control lentivirus. Mice in the ICAM1 group were intraperitoneally injected with ICG001 (40 mg/kg) on days 10, 12, 14, 17, and 19 post-tumor implantations (2 × 10⁴ cells/mouse). l Changes in body weight of mice as in (k). m Kaplan-Meier survival curves of mice bearing orthotopic xenografts as in (k). Statistics: d Data are presented as mean ± SD (n = 3, biological replicates), one-way ANOVA with Tukey’s multiple comparisons test. i, j Data are presented as mean ± SD (n = 3, biological replicates), two-tailed Student’s t-test. l Data are presented as mean ± SD (n  =  6 mice/group), day 25, ICAM1 vs. vector and ICAM1 + ICG001, one-way ANOVA with Tukey’s multiple comparisons test. m Log-rank test (n  =  6 mice/group). Images: b, e, f representative blots (n = 3, biological replicates). The samples derive from the same experiment but different gels for β-catenin with cyclin D1, another for ICAM1 with GAPDH, and another for Axin2 (b) were processed in parallel. The samples derive from the same experiment but different gels for β-catenin, LaminB1, and α-Tubulin (c) were processed in parallel. The samples derive from the same experiment but different gels for β-catenin with cyclin D1, another for ICAM1 with GAPDH, and another for Nestin (e, f) were processed in parallel. GAPDH was used as a loading control (b, e, f). Source data are provided as a Source Data file.

Fig. 4. ICAM1 binds to and induces autoubiquitination of ZNRF3, activating the Wnt/β-catenin signaling pathway.

Correlation analysis of ICAM1, FZD family members, and ZNRF3/RNF43 in the CGGA (a), TCGA (b) datasets. IB analysis of ICAM1, ZNRF3, RNF43, LRP6, and FZD4 expression in X01 and 131 GSCs (c) infected with ICAM1 or vector control lentivirus, and in 83 and 528 GSCs (d) infected with shICAM1-1/−2 or shCtrl lentivirus. e IB analysis of ICAM1, ZNRF3, and RNF43 expression in X01 (left) and 131 (right) GSCs infected with ICAM1 or vector control lentivirus and treated with MG132 (10 µM) and TAK-243 (0.5 µM). f Co-immunoprecipitation (co-IP) assay using 293 T cells with exogenous overexpression of Flag-tagged ZNRF3 and HA-tagged ICAM1. HA was used as the precipitating antibody. g Co-IP assay using 293 T cells with exogenous overexpression of Flag-tagged ZNRF3 and HA-tagged ICAM1. Flag was used as the precipitating antibody. h In vitro binding affinity between ICAM1 and ZNRF3 was measured using a microscale thermophoresis (MST) assay. ICAM1 concentration was fixed at 50 nM, while ZNRF3 concentrations ranged from 2.5 µM to 0.07 nM. The binding curve showed a dissociation constant (Kd) of 580 nM. The inset shows the thermophoretic movement of fluorescently labeled proteins. Fnorm = F1/F0 (Fnorm: normalized fluorescence; F1: fluorescence after thermodiffusion; F0: initial fluorescence or fluorescence after T-jump). i Co-IP analysis of in vitro ubiquitination in 293 T cells with exogenous overexpression of Flag-tagged ZNRF3 and HA-tagged ICAM1. The Flag was used as the precipitating antibody. j Schematic illustrating ICAM1-induced autoubiquitination of ZNRF3, which leads to the activation of Wnt/β-catenin signaling pathway. Created in BioRender. Yin, J. (2025) BioRender.com/98gcufu. Statistics: a Two-sided Spearman correlation test. Images: c–g, i Representative blots (n = 3, biological replicates). The samples derive from the same experiment but different gels for RNF43 and ZNRF3, another for LRP6 with FZD4, and another for ICAM1 with GAPDH (c, d) were processed in parallel. The samples derive from the same experiment but different gels for ZNRF3 and RNF43, and another for ICAM1 with GAPDH (e) were processed in parallel. The samples derive from the same experiment but different gels for HA with GAPDH and another for Flag (f, g) were processed in parallel. The samples derive from the same experiment but different gels for Flag and HA, and another for UBI with GAPDH (i) were processed in parallel. GAPDH was used as a loading control. Source data are provided as a Source Data file.

Fig. 5. PD-L1 identified as a downstream effector of ICAM1-Wnt/β-catenin signaling, promoting immune evasion.

a RT-qPCR analysis of Icam1 expression in GL261 cells infected with Icam1 or vector control lentivirus. b Cell proliferation assays using GL261-Icam1 or GL261-vector cells. c Luminescence images of mice bearing orthotopic xenografts of GL261-luc cells infected with Icam1 or vector control lentivirus. d Kaplan-Meier survival curves of mice as in (c) (n = 6, 5 × 10⁴ cells/mouse). e IB analysis of ICAM1, β-catenin, cyclin D1, PD-L1, CD47, CD24, and B2M expression in GSC209, 83, and 528 GSCs infected with shICAM1-1/−2 or shCtrl lentivirus. f IB analysis of ICAM1, β-catenin, cyclin D1, PD-L1, CD47, CD24, and B2M expression in X01 GSC, 131 GSC, and GL261 cells infected with ICAM1 or vector control lentivirus. Correlation analysis of ICAM1 and CD47 or PD-L1 in all glioma (g) and GBM patient samples (h). Data were collected from the CGGA portal. IB (i) and cell proliferation (j) analysis in 528 GSC treated with anti-ICAM1 antibody (1, 2, 4 µg/mL) or isotype antibody. β-catenin, cyclin D1, PD-L1, CD47, CD24, and B2M expression were analyzed in (i). IB (k) and cell proliferation (l) analysis in 83 GSC treated with anti-ICAM1 antibody (4 µg/mL) or isotype antibody. β-catenin, cyclin D1, PD-L1, CD47, CD24, and B2M expression were analyzed in (k). IB analysis of ICAM1, β-catenin, cyclin D1, PD-L1, CD47, CD24, and B2M expression in X01 (m) and 131 GSCs (n) infected with ICAM1 or vector control lentivirus, and treated with ICG001 (10 µM) or vehicle. o Schematic illustration of the proximal region of human pGL3-CD274 promoter (upper). Luciferase reporter assays of TCF/LEF activity in 293 T cells transfected with luciferase reporter vectors containing TCF/LEF sequences of CD274 promoter (lower). p Killing efficiency of anti-ICAM1 (4 µg/mL), ICG001 (10 µM), and anti-PD-L1 (10 µg/mL) in 83-luc GSC infected with CD19 lentivirus and co-cultured with CD19-CAR T cells. q Schematic illustrating the role of ICAM1 in regulating the Wnt/β-catenin/PD-L1 signaling pathway. Created in BioRender. Yin, J. (2025) BioRender.com/vhr5dod. Statistics: a, b, j, l Data are presented as mean ± SD (n = 3, biological replicates), two-tailed Student’s t-test. d Log-rank test (n  =  6 mice/group). g, h Two-sided Pearson correlation test. o Data are presented as mean ± SD (n = 3, biological replicates), one-way ANOVA with Tukey’s multiple comparisons test. p Data are presented as mean ± SD (n = 2, biological replicates), two-tailed Student’s t-test. Image: e, f, i, k, m, n Representative blots (n = 3, biological replicates). The samples derive from the same experiment but different gels for β-catenin with cyclin D1, another for PD-L1 with CD24, another for CD47 with B2M, and another for ICAM1 with GAPDH (e, f, m, n) were processed in parallel. The samples derive from the same experiment but different gels for β-catenin with cyclin D1, another for PD-L1 with CD24, another for CD47 with B2M, and another for GAPDH (i, k) were processed in parallel. GAPDH was used as a loading control. Source data are provided as a Source Data file.

Fig. 6. PD-L1 regulates stemness, proliferation, and tumorigenicity of GSCs.

a IB analysis of PD-L1 expression in 83 (left) and 528 (right) GSCs infected with shPD-L1-1, shPD-L1−2 or shCtrl lentivirus. b Cell proliferation assays using 83 (left) and 528 (right) GSCs infected with shPD-L1-1/−2 or shCtrl lentivirus. c LDAs performed using 83 (left) and 528 (right) GSCs infected with shPD-L1-1/−2 or shCtrl lentivirus. d IB analysis of ICAM1 and PD-L1 expression in 83 (left) and 528 (right) GSCs infected with shICAM1, PD-L1 or shCtrl lentivirus. e Cell proliferation assays using 83 (left) and 528 GSCs (right) infected with shICAM1, PD-L1, or shCtrl lentivirus. f LDAs performed using 83 (left) and 528 (right) GSCs infected with shICAM1, PD-L1 or shCtrl lentivirus. g IB analysis of Icam1, β-catenin, cyclin D1, PD-L1, and LRP6 expression in GL261 infected with Icam1, shPD-L1, or vector control lentivirus. h Cell proliferation assays using GL261 infected with Icam1, shPD-L1, or vector control lentivirus. i Luminescence images of mice bearing orthotopic xenografts of GL261-luc cells infected with Icam1, shPD-L1, or vector control lentivirus (5 × 10⁴ cells/mouse). j Kaplan-Meier survival curves of mice as in (i). Statistics: b, e, h Data are presented as mean ± SD (n = 3, biological replicates), two-tailed Student’s t-test. j Log-rank test (n  =  6 mice/group). Image: a, d, g Representative blots (n = 3, biological replicates). The samples derive from the same experiment but different gels for PD-L1 and GAPDH (a) were processed in parallel. The samples derive from the same experiment but different gels for ICAM1 with GAPDH and another for PD-L1 (d) were processed in parallel. The samples derive from the same experiment but different gels for β-catenin with cyclin D1, another for LRP6 with PD-L1, and another for ICAM1 with GAPDH (g) were processed in parallel. GAPDH was used as a loading control. Source data are provided as a Source Data file.

Fig. 7. ICAM1 alters the tumor immune microenvironment involved in GBM progression.

a Schematic illustration of anti-PD-1 treatment in the GL261 orthotopic GBM mouse model. Mice bearing orthotopic xenografts of GL261-luc cells infected with Icam1 or vector control lentivirus. Icam1-overexpressing mice were intraperitoneally injected with anti-PD-1 or isotype antibodies, while vector control mice received isotype antibody treatment. Created in BioRender. Yin, J. (2025) BioRender.com/pxjp05q. b Luminescence images of mice as in (a). c Kaplan-Meier survival curves of mice as in (a) (5 × 10⁴ cells/mouse). d Representative flow cytometry analysis of GBM tumors as in (a). Three days after the final anti-PD-1 antibody treatment, tumor-infiltrating lymphocytes (TILs) were analyzed for the relative proportions of naïve, effector, and central memory CD8⁺ T cells. e Schematic illustration of anti-PD-1, anti-Icam1, or combination treatments at the indicated timepoints. Mice bearing orthotopic xenografts of GL261-luc cells infected with Icam1 or vector control lentivirus. Icam1-overexpressing mice were intraperitoneally injected with anti-PD-1, anti-Icam1, or a combination of both antibodies at the indicated timepoints. Control mice were treated with isotype antibodies. Created in BioRender. Yin, J. (2025) BioRender.com/930czlq. f Luminescence images of mice as in (e). g Kaplan-Meier survival curves of mice as in (f) (5 × 10⁴ cells/mouse). Statistics: c, g Log-rank test (n  =  6 mice/group). Source data are provided as a Source Data file.