GLIS3 drives epithelial-mesenchymal transition and cancer stem-like traits in stomach adenocarcinoma via TGFBR3-Hedgehog signaling.

BACKGROUND

GLIS family zinc finger 3 (GLIS3) is a transcription factor implicated in multiple malignancies, but its role in stomach adenocarcinoma (STAD) and its downstream effector axis remain unclear. We investigated whether GLIS3 coordinates epithelial-mesenchymal transition (EMT) and cancer stem cell (CSC)-like programs in STAD through a signaling cascade.

METHODS

GLIS3 expression and prognostic associations were analyzed using transcriptomic datasets from The Cancer Genome Atlas, Genotype-Tissue Expression, and Gene Expression Omnibus. GLIS3 protein levels were evaluated by immunohistochemistry in 133 paired STAD tissues. Survival was assessed by Kaplan-Meier analysis together with multivariable regression models. Gain- and loss-of-function studies were performed in AGS and HGC-27 cells to evaluate proliferation, migration, EMT and CSC-like phenotypes. Mechanistic experiments interrogated a GLIS3-transforming growth factor beta receptor 3 (TGFBR3)-Hedgehog axis using ChIP-qPCR, dual-luciferase assays, co-immunoprecipitation, GLI reporter assays, and vismodegib treatment. Tumor growth was assessed in xenograft models.

RESULTS

GLIS3 was consistently upregulated in STAD across public datasets and in the clinical cohort (median H-score: 112 vs 39, P < 0.001) and was associated with adverse clinicopathological features and poor survival. In the TCGA cohort, higher GLIS3 expression was associated with greater nodal burden, more advanced nodal and metastatic status, and worse vital status. In multivariable Cox models restricted to cases with defined stage classifications, GLIS3 remained independently associated with worse overall survival in the primary model (hazard ratio = 1.45, 95% confidence interval: 1.04-2.02; P = 0.031), with a consistent result in the sensitivity model. Functionally, GLIS3 enhanced proliferation and migration and promoted EMT marker switching and CSC-like traits. Mechanistically, GLIS3 transcriptionally activated TGFBR3 and increased Hedgehog pathway activity; bidirectional rescue experiments showed that TGFBR3 overexpression partially restored, whereas TGFBR3 silencing attenuated, GLIS3-driven EMT/CSC-like phenotypes and growth-related functions. Xenograft experiments supported an in vivo growth-promoting role of the GLIS3-TGFBR3-Hedgehog axis.

CONCLUSIONS

GLIS3 identifies an aggressive STAD phenotype and engages a TGFBR3-Hedgehog program linked to EMT and CSC-like features. These findings support further evaluation of GLIS3 as a biomarker for risk stratification and provide a rationale for biomarker-guided targeting of the GLIS3-TGFBR3-Hedgehog axis in STAD.