GSK3 inhibition reduces the viability of glioma stem cells and induces cell death
To identify potential therapeutic targets in glioma stem cells (GSCs), we performed a small-scale screening with small molecules from the signal transduction small molecules in the murine glioma stem cell line HTS6. Among the compounds tested, CHIR99021 (CHIR), a highly specific GSK3 inhibitor, significantly reduced the viability of HTS6 cells as well as other murine and human glioma stem cell lines (Fig. 1A). Importantly, CHIR exhibited no cytotoxic effects on primary astrocytes derived from neonatal C57/B6 mice (Fig. 1B). This selective inhibition of GSCs by CHIR was recapitulated with LiCl, another GSK3 inhibitor, in a concentration-dependent manner (Supplementary Fig. 1).
We further investigated the effects of CHIR on cell death and proliferation in HTS cells. CHIR treatment induced robust cell death, as determined by Annexin V staining (Fig. 1C, D), and significantly decreased cell proliferation, as determined by EdU incorporation assays (Fig. 1E & F). These results indicate that inhibition of GSK3 effectively impairs GSC viability by inducing both cell death and proliferation arrest.
GSK3β, but not GSK3α, is crucial for the growth of glioma stem cells
GSK3 is present in mammalian cells in the form of two paralogs, GSK3α and GSK3β. To determine which paralog is essential for GSC survival, we employed shRNA-mediated knockdown of GSK3α and GSK3β in HTS cells. Knockdown of GSK3β sensitized HTS cells to CHIR-induced growth inhibition, whereas knockdown of GSK3α had no significant effect on cell viability (Fig. 2A & B). Given these results, we focused on GSK3β for subsequent experiments. Knockdown of GSK3β by shRNA delayed the growth of HTS cells by reducing their proliferative capacity (Fig. 2C-E), but had minimal impact on apoptosis (Supplementary Fig. 2). Notably, knockdown of GSK3β significantly diminished sphere formation in HTS cells in terms of both the frequency and sphere size (Fig. 2F-H), indicating that GSK3β is critical for the self-renewal properties of GSCs. In vivo, knockdown of GSK3β in HTS cells prolonged the survival of tumor-bearing mice (Fig. 2I), further supporting its role as a key regulator of GSC-driven tumorigenesis.
Activation of the canonical WNT pathway mediates the effects of GSK3 inhibition
Given that GSK3β is a key component of the β-catenin destruction complex, we hypothesized that the canonical WNT pathway mediates the cytotoxic effects of GSK3 inhibition in GSCs. To test this hypothesis, we generated HTS6 cells overexpressing a constitutively active form of β-catenin (CTN*) or a dominant-negative form of TCF (dTCF), the latter of which blocks β-catenin activity (Fig. 3A, Supplementary Fig. 3A). Activation of the WNT pathway through CTN* overexpression sensitized HTS6 cells to CHIR, whereas blocking the pathway by dTCF rescued cell viability (Fig. 3B). These results suggest that the GSK3β-β-catenin axis plays a central role in mediating the inhibitory effect of CHIR on GSCs.
Overexpression of CTN* in HTS6 cells phenocopied the effects of GSK3β knockdown by reducing cell proliferation and self-renewal capacity in vitro (Fig. 3C-E, Supplementary Fig. 3B-D) and prolonging survival in tumor-bearing mice (Fig. 3F). These findings suggest that canonical WNT signaling is sufficient to affect the viability and tumorigenic potential of GSC, similar to the phenomena observed by Chang et al. in colon cancer cells13.
Activation of the WNT signaling pathway correlates with better prognosis in glioma patients
To investigate the clinical relevance of our findings, we performed a single-sample gene set enrichment analysis (ssGSEA) 14 using RNA-seq data from The Cancer Genome Atlas (TCGA) 15 and the Chinese Glioma Genome Atlas (CGGA) 16. In both datasets, activation of the canonical WNT pathway was associated with a significantly better prognosis in glioma patients (Fig. 4A, B, Supplementary Fig. 4A). Expression of the WNT ligands WNT7A and WNT7B, which are highly expressed in the brain, was inversely correlated with glioma grade, and lower expression of WNT7B was predictive of poorer patient survival (Supplementary Fig. 4B-E).
Furthermore, we observed that the majority of IDH1-mutated tumors with a CpG island methylator phenotype (G-CIMP) 17 and a large proportion of tumors classified as proneural subtype exhibited high canonical WNT activity, whereas mesenchymal gliomas showed low WNT activity (Fig. 4C, Supplementary Fig. 4F, G). This subtype-specific enrichment of WNT activity correlates well with clinical outcomes, as both G-CIMP and proneural gliomas are known to confer a better prognosis2,16.
To investigate whether high levels of canonical WNT signaling may be involved in the maintenance of the subtype of GSCs, we treated two proneural human GSCs in the presence of CHIR with TNFα, which is known to promote the mesenchymal subtype18, in the presence of CHIR. Consistent with previous reports, TNFα treatment increased the expression level of the mesenchymal marker CD44, and this effect can be attenuated by additional CHIR (Fig. 4D), suggesting that activation of WNT signaling helps GSCs to maintain a proneural signature.
WNT activation promotes Wwc1 expression and represses Yap via LATS1 activation
To uncover the downstream mechanisms by which WNT activation inhibits GSC growth, we performed a microarray analysis comparing gene expression in HTS6 cells expressing CTN* or dTCF under different concentrations of CHIR (Fig. 5A). Wwc1 emerged as a key candidate whose expression correlated with CHIR-induced cytotoxicity. Using real-time PCR, we confirmed that the expression of Wwc1 was significantly upregulated in HTS6 cells treated with 10 or 20 µM CHIR, but remained unchanged at 5 µM CHIR, a concentration that had no effect on viability. In contrast, expression of the canonical WNT target gene Axin2 increased continuously with increasing CHIR concentration (Fig. 5B).
Overexpression of CTN* also induced Wwc1 expression, whereas dTCF repressed it, supporting the role of β-catenin in regulating Wwc1 transcription (Fig. 5B). Western blot analysis confirmed that CHIR treatment increased Wwc1 protein levels (Fig. 5C). We identified multiple TCF4 binding sites within the Wwc1 promoter, and deletion of these sites abrogated CTN*-induced Wwc1 expression, confirming that β-catenin directly regulates Wwc1 transcription (Fig. 5D & E). Furthermore, silencing Wwc1 expression by shRNA in HTS cells partially abrogated the cytotoxicity of high concentrations of CHIR (Fig. 5F & G).
In glioma samples, Wwc1 expression correlates inversely with the degree of tumor grades and is significantly more pronounced in patient samples with higher WNT activity (Fig. 5H & I). Similar to activation of canonical WNT signaling, higher Wwc1 expression in gliomas is associated with a significantly better prognosis in both the TCGA and CGGA datasets (Fig. 5J and Supplementary Fig. 5).
Wwc1 is a known regulator of the Hippo pathway, where it modulates LATS1 phosphorylation to repress Yap activity19-21. Consistent with this, CHIR treatment increased LATS1 phosphorylation in HTS6 cells (Fig. 6A), leading to subsequent phosphorylation and degradation of Yap (Fig. 6B & C). Overexpression of CTN* or knockdown of GSK3β also decreased Yap protein levels, whereas overexpression of dTCF reversed this effect (Fig. 6D & E). Silencing of LATS1/2 or overexpression of YapS127A partially rescued CHIR-induced cell death, further confirming the role of LATS1-Yap signaling in mediating the cytotoxic effects of WNT activation (Fig. 6G & H).
To further investigate the relationship between WNT activation and Yap downregulation in a clinical setting, we examined the level of the Yap target gene CTGF in medulloblastoma, as a molecular subtype of medulloblastoma is associated with WNT activation22,23. Consistent with our observation, the WNT subgroup of medulloblastoma has the lowest CTGF level, while the SHH subset, which is reported to be associated with Yap amplification and upregulation, has the highest CTGF level24 (Fig. 6I). A similar trend has been observed in glioma patient samples: CTGF expression was lower in tumors with higher WNT activity; its expression correlated positively with tumor grade and predicted worse outcome (Fig. 6J-L). Histologically, we found that tumor samples showing high expression of WNT7B also showed high expression of the proneural marker OLIG2, but relatively low expression of the mesenchymal marker CD44 and the Yap target gene CTGF (Fig. 6M).
To investigate the role of Yap in GSCs, we knocked down Yap with shRNAs and found decreased HTS cell growth (Fig. 7A-E) and significantly decreased sphere formation ability (Fig. 7F). GSEA analysis revealed that Myc targets and cell cycle related E2F targets were significantly downregulated (Fig. 7G). Yap was reported to directly regulate Myc expression25 and high concentration of CHIR treatment resulted in downregulation of c-myc level in HTS cells (Fig. 7H). Many studies have shown that overexpression of Yap is associated with chemoresistance26,27. Consistent with previous observations, HTS cells in which either Gsk3b or Yap expression was downregulated showed higher sensitivity to the chemotherapeutic agents carboplatin and 5FU (Fig. 7I & J).
Since CHIR requires a relatively high concentration to show cytotoxicity on glioma cells, we screened several known GSK3 inhibitors and found that alsterpaullone (ALP) was 5-10-fold more potent than CHIR (Fig. 8A). Similar to CHIR and other GSK3 inhibitors, ALP can also activate the WNT reporter (Fig. 8B). It can also downregulate Yap protein levels at 1 μM (Fig. 8C). In vivo, ALP treatment delays the growth of tumor allografts compared to the DMSO control group (Fig. 8D, E, Supplementary Fig. 8A & B).