ZNF117 regulates glioblastoma stem cell differentiation towards oligodendroglial lineage

Genome-wide RNAi screen for identification of GSC differentiation regulators

To identify genes that regulate GSC differentiation and viability, we performed a genome-wide RNAi screen using the Dharmacon siGenome library. This library consists of siRNAs targeting 18,119 genes, with four siRNAs specific to each gene (Supplementary Data 1). The primary screen was performed in triplicate using GS5 GSCs, which have been well-characterized by our group and others^6,14,15, through a reverse transfection procedure (Fig. 1a). Five days post-transfection, the cells were fixed and stained with Hoechst 33342, which identifies all cells, and an anti-Nestin antibody, which identifies Nestin⁺ GSCs. By using a NES (Nestin) siRNA as a positive control and a RISC-free siRNA as a negative control, we were able to confirm up to 93% knockdown of the targeted genes (Fig. 1b, Supplementary Fig. 1). Results were then sorted based on a statistically significant reduction of the Nestin⁺ cell population, as well as the total cell population. The top 100 genes, which demonstrated the greatest effect in reducing the number of total cells and/or the number of Nestin+ cells, were selected (Supplementary Table 1). These genes underwent a secondary validation screen, where transfections were repeated using a single siRNA per well rather than pooled siRNAs targeting a specific gene. Transfection and staining were performed in an identical manner to the primary screen in triplicate but with four siRNAs for each gene. Control groups were treated with RISC-free siRNA. In the end, all genes were ranked based on their ability to regulate the percentage of total cells (cytotoxicity group) or the percentage of Nestin⁺ cells (differentiation group) (Supplementary Table 1). As expected, most genes in the cytotoxicity group were involved in cell cycle regulation, such as CCNB1 and WEE1, or are components in the ubiquitin–proteasome system (UPS), such as UBC and UBB, which are known to be active in cancer cells (Fig. 1c). The top five genes identified to regulate the population of Nestin⁺ cells included ZNF117, MRLC2, CASS4, DUB3, and ARRB1 (Fig. 1d). Among them, ZNF117 demonstrated the greatest inhibitory effect. Each of the four siRNAs targeting ZNF117 consistently reduced the percentage of Nestin⁺ cells (Fig. 1d).

a Schematic diagram of genome-wide RNAi screening through reverse transfection. Single GS5 cells were plated on polyornithine coating 384-well plates, in which siRNA complexes were pre-assembled. Five days later, cells were stained with Hoechst 33342 (blue, nuclei staining) and anti-Nestin antibody⁵⁰, followed by imaging using an Opera high-content screening system. b Representative images of cells treated with RISC-free control siRNA, siNestin, or siUBC. Experiments were performed in triplicate. Scale bar, 20 μm. c The top 5 genes which demonstrated the greatest inhibitory effects on survival of total cells. d The top five genes which demonstrated the greatest effects on reducing Nestin⁺ cell population. For c and d, each gene was targeted by four individual siRNAs, instead of a pool of four siRNAs. AB(AB00477)-xxx represents the ID number of the specific siRNA coded by Dharmacon. Black line represent mean, dots indicate values. Data are presented as mean ± SD (n = 3). Statistical differences were determined by two-tailed student’s t-test.

scRNA-seq suggests that ZNF117 regulates GSC differentiation towards oligodendroglial lineage

To determine which genes regulate GSC differentiation into terminal cell types, we performed scRNA-seq on GS5 cells¹⁵. To minimize batch-to-batch variability, we prepared four batches of cDNA libraries. Each batch contained between 471 and 1824 cells, 1811–2729 median genes per cell, 5258–9207 median unique molecular identifiers (UMIs), and 20,023–23,000 total genes. In all, we obtained 4592 single-cell transcriptomes at a depth of at least 10,000 reads per cell. After quality-control filtering (see Methods), 3412 cells with expression levels for 21,404 genes were used for downstream analysis (Supplementary Table 2).

To detect specific cell types within the tumor, we visualized GS5 cells with UMAP and grouped them using unbiased, graph-based clustering (Fig. 2a, Supplementary Fig. 2a). Gene expression levels for ZNF117, NES (GSCs), SNAP25 (neurons), GFAP (astrocytes), and PLP1 (oligodendrocytes) were visualized across GS5 cells (Fig. 2b). To identify specific cell populations, we used cell type markers (Supplementary Fig. 2b, c), which had been reported in previous studies to be related to specific cell types^{4,14,16,17,18,19,20,21,22,23,24,25,26,27}. As expected, the same clusters were enriched for the same cell type. In total, we identified 426 GSCs, 802 neuron-like cells, 353 astrocyte-like cells, and 800 oligodendrocyte-like cells from the dataset (Fig. 2c).

a UMAP of scRNA-seq data from GS5 cells. Unsupervised clustering based on top 1000 variable genes reveals cell–cell heterogeneity within the GS5 tumor population. b Expression of cell type-specific markers and ZNF117 in UMAP. Expression of NES (GSC), GFAP (astrocytes), PLP1 (oligodendrocytes), and SNAP25 (neurons) is shown. c Distribution of cell types identified from GS5 scRNA-seq data. Neurons and oligodendrocyte each compose of around 33% of the tumor, while GSCs and astrocytes each compose of 18% and 15% of the tumor, respectively. d Single-cell trajectories of cell types identified in c. GSCs are in the center of the graph, while each of the differentiated cell types form their own distinct branch. e Venn diagram of significant genes that regulate oligodendrocyte differentiation from single-cell trajectory analysis and Nestin expression from the RNAi screen. ZNF117 appears in both screens. f Heatmap of significant genes that regulate GSC to oligodendrocyte differentiation. ZNF117 decreases in expression as GSCs differentiate into oligodendrocytes. Color bar represents normalized expression. Genes are hierarchically clustered according to their expression pattern. q-value < 5e-02.

Next, we constructed single-cell trajectories from GSCs and differentiated cells to determine which genes mediate GSC differentiation into specific lineages. When aligning the cells in pseudotime, we observed that GSCs settled into the center of the single-cell trajectories, while neuron-like cells, astrocyte-like cells, and oligodendrocyte-like cells were each isolated on their own distinct branches (Fig. 2d). To identify genes that regulate GSC differentiation into each cell type, we determined the differentially expressed genes on each branch, with GSC to neuron-, astrocyte-, and oligodendrocyte-like cells corresponding to states 1, 4, and 3, respectively in Supplementary Fig. 2d.

We compared genes that inhibit GSC differentiation with the 100 genes selected for validation in the RNAi screen and found that 15, 10, and 9 genes overlap as potential candidates which inhibit GSC differentiation towards oligodendroglial (Fig. 2e), neuronal (Supplementary Fig. 2g), and astroglial (Supplementary Fig. 2h) lineages, respectively (Supplementary Table 3). Among the overlapping genes, ZNF117 was significantly differentially expressed in GSC to oligodendroglial differentiation (Fig. 2f), but not other lineages (Supplementary Fig. 2e, f). This finding suggests that ZNF117 may regulate GSC differentiation towards oligodendroglial lineage.

Experimental validation of ZNF117 as GSC differentiation regulator

To confirm that ZNF117 regulates GSC differentiation, we differentiated GS5 cells by culturing them in medium containing 10% FBS and examined the changes in expression of ZNF117. Western Blot (WB) analysis showed that the expression level of ZNF117 significantly decreased with culture time; by the end of 3 weeks, when GS5 cells were fully differentiated based on flow cytometry analysis (Supplementary Fig. 3), the expression of ZNF117 was barely detectable (Fig. 3a).

a WB analysis of the expression of ZNF117 in GS5 cells cultured in serum-free or cultured in serum-containing medium for 2 or 3 weeks (wk). Experiments were performed in triplicate. b WB analysis of the expression of ZNF117 in GS5 parental cells and cells treated with Cas9 and the indicated sgRNAs. c Proliferation of GS5 cells treated with Cas9 and the indicated sgRNAs (n = 3 biologically independent samples). d Stem cell frequency of GS5 cells treated with Cas9 and the indicated sgRNAs as determined by limiting dilution assay. e–g Representative flow cytometry images (e) and corresponding quantification (f) (n = 3 biologically independent samples, black line represents mean, dots indicate values), and immunostaining (g) of Nestin⁺ and GalC⁺ cell populations in GS5 cells after treatment with Cas9 and the indicated sgRNAs. h Kaplan–Meier survival analysis of mice received inoculation of the indicated cells. i H&E staining of residual tumors isolated from mice receiving inoculation of the indicated cells. Data are presented as mean ± SD (n = 3). Statistical differences were determined by two-tailed student’s t-test. *P-value < 0.05; **P-value < 0.01; ***P-value < 0.001. Scale bar, 50 μm.

To further investigate the biological functions of ZNF117 as a differentiation regulator, we employed CRISPR to downregulate the expression of ZNF117 in GS5 cells. Unlike siRNA that targets mRNA, Cas9/sgRNAs induces genetic knockout at the genomic DNA level, which allows for the generation of stable cell lines with persistently low expression of the targeted genes. We established GS5 cells with stable overexpression of Cas9 and sgRNAs targeting different regions of ZNF117 (Supplementary Fig. 4a). Control cells were prepared through the same procedures but with an sgRNA targeting GFP. The resulting cells were designated as GS5^sgGFP, GS5^sgZNF117-1, and GS5^sgZNF117-2. WB analysis showed that the selected sgRNAs efficiently reduced the expression of ZNF117 in GS5 cells (Fig. 3b). We analyzed the impact of ZNF117 downregulation on cell proliferation, stemness, and differentiation. Results in Fig. 3c, d show that downregulation of ZNF117 significantly reduced both cell proliferation rate and stem cell frequency. Further analyses by flow cytometry and immunostaining found that downregulation of ZNF117 reduced the population of Nestin⁺ cells, while notably increasing the GalC⁺, but not GFAP⁺, or Tuj-1⁺, population within GS5 cultures (Fig. 3e–g, Supplementary Fig. 4b). The increase in oligodendroglial population was further confirmed by the expression of OLIG1 (Supplementary Fig. 4c). Next, we overexpressed ZNF117 gene in GS5^sgZNF117-1 cells and found the phenotypic changes associated with downregulation of ZNF117 were reversed (Supplementary Fig. 5). These findings, which are consistent with the observations from both the RNAi screen (Fig. 1d) and scRNA-seq analysis (Fig. 2c–e), suggest that ZNF117 regulates GSC differentiation towards the oligodendroglial lineage.

We determined the effect of ZNF117-induced differentiation on tumor development in vivo through inoculation of engineered cells into the brains of nude mice. Control mice received injection of the same amount of GS5^sgGFP cells. Afterward, the mice were monitored for survival. We found that downregulation of ZNF117 significantly prolonged the survival of tumor-bearing mice. Mice inoculated with GS5^sgZNF117-1 cells, which have the lowest expression level of ZNF117 among all the three cells, survived for >100 days post-inoculation. In contrast, the median survival for mice bearing tumors derived from GS5^sgZNF117-2 cells was 49 days, compared to 37 days for mice bearing GS5^sgGFP tumors (Fig. 3h). The observed difference in inhibition of tumor formation between the two sgRNAs could be attributed to the fact that, compared to sgZNF117-2, sgZNF117-1 is more efficient in down-regulating ZNF117 expression (Fig. 3b) and in inducing differentiation (Fig. 3e–g). To exclude the possibility that the marked tumor-inhibitory ability of sgZNF117-1 is caused by off-target effects, we identified genes that have mismatches with sgZNF117-1 and determined whether these genes have detectable disruption after sgZNF117-1-was delivered. No genes had zero or one mismatch with sgZNF117-1, and six genes had two to four mismatches in exons. We sequenced the six genes and did not detect significant cutting activity (Supplementary Fig. 6). H&E staining showed that loss of sgZNF117 correlated with dramatically decreased tumor size (Fig. 3i). Ki-67 staining revealed that residual tumors derived from GS5^sgZNF117-1 cells had reduced proliferation rates (Supplementary Fig. 4d).

Collectively, these results suggest that ZNF117 regulates GS5 cell differentiation towards oligodendroglial lineage, and downregulation of ZNF117 inhibits the proliferation, stemness, and tumor development of GSCs. Compared to sgZNF117-2, sgZNF117-1 reduced the expression of ZNF117 at higher efficiency and led to greater differentiation effects. Therefore, sgZNF117-1 was selected for further studies and is designated sgZNF117 in the remainder of the manuscript.

Validation in additional GSC cell cultures and xenografts

To exclude the possibility that the observed differentiation effects of ZNF117 is limited to GS5 cells, we applied the same approach to characterize the role ZNF117 in PS30 and PS24 cells, two additional well-characterized GSC cultures^6,15. We found that the expression of ZNF117 in PS30 cells decreased with continuous culture in FBS-containing medium (Fig. 4a). Through transduction with lentiviruses containing sgZNF117, we generated PS30 cells with downregulation of ZNF117 (Fig. 4a). The resulting PS30^sgZNF117 cells were characterized for cell proliferation, stemness, differentiation, and tumorigenicity. Cells treated with sgGFP (PS30^sgGFP cells) were included as a control. We found that downregulation of ZNF117 effectively reduced the rate of cell proliferation (Fig. 4b) and frequency of stem cells (Fig. 4c), induced differentiation preferentially towards oligodendroglial lineage (Fig. 4d–f), and inhibited tumor development in mice (Fig. 4g, P = 0.0007). The same trends were also observed in PS24 cells in vitro and in vivo (Supplementary Fig. 7). Therefore, our findings in both GSC cultures and mouse xenografts are consistent with those identified in GS5 cells, suggesting that the biological effects of ZNF117 are not unique to GSCs of specific origin.

a WB analysis of the expression of ZNF117 in PS30 cells cultured in serum-containing medium or with constitutive expression of Cas9 and the indicated sgRNAs. Experiments were performed in duplicate. b, c Proliferation (b) and stem cell frequency (c) of PS30 cells treated with Cas9 and the indicated sgRNAs (n = 3 biologically independent samples). d–f Representative flow cytometry images (d) and corresponding quantification (n = 3 biologically independent samples, black line represents mean, dots indicate values). e and immunostaining (f) of Nestin⁺ and GalC⁺ cell populations in PS30 cells after treatment with Cas9 and the indicated sgRNAs. Scale bar, 50 μm. g Kaplan–Meier survival analysis of the mice received inoculation of the indicated cells (n = 7). Data are presented as mean ± SD. Statistical differences were determined by two-tailed student’s t-test. *P-value < 0.05; **P-value < 0.01.

ZNF117 regulates Notch signaling through interaction with JAG2

The ZNF117 gene encodes a protein containing multiple C2H2-type zinc finger motifs and is predicted to have DNA-binding transcription factor activity based on gene ontology annotations. To determine the transcriptional activity of ZNF117, we performed a whole-transcript expression analysis using Affymetrix HuGene-2.0 arrays and found GS5^sgZNF117 and control GS5^sgGFP cells to be markedly different at the transcriptional level (Fig. 5a). In total, 3,642 transcripts were differentially expressed by over 2-fold (Supplementary Data 2), including 1781 that were up-regulated and 1861 that were downregulated in GS5^sgZNF117-1 cells. Pathway analysis of both the up-regulated and downregulated genes reveals that the Notch pathway is among the major pathways which are potentially regulated by ZNF117 (Supplementary Fig. 8a). Further analysis revealed that major Notch-related genes, including NOTCH1, NOTCH2, NOTCH3, DLL3, DTX1, DTX3, LFNG, MAML2, and NOTCH2N, are in the downregulated cohort and, NOTCH1, MET, and PRDX1, are in the up-regulated cohort (Fig. 5b). Down-regulation of NOTCH1, NOTCH2, and NOTCH3 was confirmed by qRT-PCR analysis (Supplementary Fig. 8b). We further quantified the activity of the Notch signaling using Cignal™ pathway reporters and found that downregulation of ZNF117 significantly inhibited Notch activity (Fig. 5c). No significant differences in activity of Wnt signaling were detected between GS5^sgZNF117 and control GS5^sgGFP cells (Supplementary Fig. 8c). These results suggest that ZNF117 exerts its biological functions through regulation of the Notch pathway.

a Heatmap of genome-wide transcription profiles of GS5^sgGFP and GS5^sgZNF117-1. Colors represent normalized expression. b Volcano plot shows that ZNF117 regulates Notch signaling. c Pathway reporter assay suggests that Notch activity is downregulated in ZNF117 knockout cells. d Genes identified by CHIP-seq and their distance to transcription start site (TSS). e Genome browser view of the distribution of ZNF117-binding regions identified by ChIP-seq. f ChIP-PCR validation of JAG2 as a ZNF117-binding candidate. g Expression of JAG2 in GS5^sgGFP and GS5^sgZNF117 as determined by qRT-PCR. h Schematic diagram of JAG2 reporter construction. i Characterization of the effect of ZNF117 downregulation on transcription activity of the indicated sequences within JAG2 based on luciferase reporter assay. Data are presented as mean ± SD. Black line represents mean, dots indicate values (c, f, g, i). Statistical differences were determined by two-tailed student’s t-test. *P-value < 0.05; **P-value < 0.01; ***P-value < 0.001.

To identify specific genes which are regulated by ZNF117, we performed ChIP-seq. In total, DNA fragments of 4,616 genes were enriched. Among them, 8%, 11%, and 81% were located in introgenic, proximal, and intergenic regions, respectively (Supplementary Fig. 8d). For 327 genes, including JAG2, the DNA fragments were located in the proximal promoter region (Fig. 5d, e, Supplementary Data 3). We examined three genes, including IL6ST, CDH4, and JAG2, which are known to be relevant to GBM, by ChIP-PCR, and confirmed that all of them were enriched (Fig. 5f, Supplementary Fig. 8e). Because the cDNA array analysis found that Notch signaling is regulated by ZNF117, we focused on JAG2, a Notch ligand known to regulate the Notch pathway^28,29. We found that downregulation of ZNF117 reduces the expression of JAG2 (Fig. 5g). To determine how ZNF117 directly regulates JAG2, we cloned sequences of 500, 1500, and 7150 flanking the predicted ZNF117-binding site within JAG2 into PGL4.2-TATA luciferase construct (Fig. 5h) and determined the expression of luciferase in cells that were engineered to express either Cas9/sgGFP or Cas9/sgZNF117 after transfection with the reporter constructs. We found that, compared to those in control cells, the luciferase signal in Cas9/sgZNF117 expressing cells was significantly lower (Fig. 5i). To further provide evidence that ZNF117 regulates GSC differentiation through JAG2, we overexpressed JAG2 under a CMV promoter in GS5^sgZNF117-1 cells and found that overexpression reversed ZNF117 knockout-mediated oligodendroglial lineage differentiation (Supplementary Fig. 5). Taken together, these results suggest that ZNF117 regulates Notch signaling through interaction with JAG2.

ZNF117 as a therapeutic target for GBM differentiation therapy

Analysis of RNA-seq database by The Cancer Genome Atlas (TCGA) shows that ZNF117 is expressed at a higher level in GBM than normal brain tissues (Supplementary Fig. 9a). Further analysis of several GBM public databases, including TCGA RNA-seq database, The Repository of Molecular Brain Neoplasia Data (REMBRANDT) database, and Chinese Glioma Genome Atlas (CGGA), found that the expression of ZNF117 is negatively correlated with patient survival (Supplementary Fig. 9b–d). Collectively, these analyses suggest ZNF117 as a potential therapeutic target for GBM differentiation therapy.

We assessed if ZNF117 can be targeted for GBM treatment experimentally in tumor-bearing mice. To enable downregulation of ZNF117 in vivo, we employed liposome-templated hydrogel nanoparticles (LHNPs), which we recently developed for targeted delivery of CRISPR/Cas9 to brain tumors³⁰. LHNPs were synthesized with encapsulation of ribonucleoprotein (RNP) complexes of Cas9 protein and sgRNA targeting either ZNF117 or GFP. The NPs were further engineered through surface conjugation of iRGD, a peptide with high affinity for α_vβ₃/α_vβ₅ integrins^31,32, and internal encapsulation of Lexiscan, a small molecule that can transiently open the blood-brain barrier (BBB)^33,34 (Fig. 6a). As a result, LHNPs can penetrate the BBB and accumulate preferentially in tumors through an autocatalytic, brain tumor-targeting mechanism^35,36,37. The resulting LHNPs are spherical in shape, have a diameter of ~90 nm (Fig. 6b), and were confirmed to selectively transfect brain tumor cells after intravenous administration (Supplementary Fig. 10). We characterized LHNPs in GS5 and PS30 cells. We found that the expression of ZNF117 in both PS30 and GS5 cells treated with LHNPs loaded with ZNF117-targeted RNP was significantly decreased, compared to control cells treated with LHNPs loaded with GFP-targeted RNP (Fig. 6c, Supplementary Fig. 11a). Because ZNF117 expression decreased more significantly in PS30 cells compared to GS5 cells after treatment, we chose to determine the therapeutic benefit of ZNF117-targeted therapy in PS30-derived mouse xenografts. Mouse xenografts were established through intracranial inoculation of PS30 cells that were engineered to express luciferase. Seven days later, the mice were randomly grouped and treated with LHNPs loaded with ZNF117-targeted RNP, PBS, or control LHNPs loaded with GFP-targeted RNP. Treatments were given at a dose of 1 mg NPs (sgRNA equivalent dose of 0.2 ug) per injection intravenously through tail veins three times a week for three consecutive weeks. The mice were monitored for survival and tumor development, which was determined based on luciferase imaging by an In Vivo Imaging System (IVIS). Results in Fig. 6d, e show that treatment with LHNPs loaded with ZNF117-targeted RNP significantly inhibited tumor growth and improved the survival of tumor-bearing mice. The median survival of the treatment group was 48 days, compared to 33 days for mice treated with control LHNPs (P = 0.007) and 32 days for mice treated with PBS. At the end of the study, the mice were euthanized, and their brains were isolated and examined. H&E staining confirmed that residual tumors in control groups were significantly larger than those in the treatment group; in contrast to tumors in the control groups, the residual tumors in the treatment group showed a benign phenotype with reduced nuclear-to-cytoplasm ratios (Fig. 6f). Further WB analysis showed that compared to that in the control group, the expression of ZNF117 in residual tumors isolated from the treatment group was significantly lower (Fig. 6g). Consistent with the findings in cell culture (Fig. 4d, e), immunostaining revealed that treatment with LHNPs loaded with ZNF117-targeted RNP significantly reduced Nestin⁺ while enriching GalC⁺ cells (Fig. 6h). These findings suggest that the observed inhibition of tumor development in the treatment group resulted from ZNF117-mediated differentiation.

a Schematics of LHNPs. b A representative image of LHNPs as captured by transmission electron microscopy (TEM). Experiments were performed in triplicate. Scale bar: 500 nm. c WB analysis of ZNF117 expression in PS30 cells treated with LHNPs loaded with Cas9 and sgGFP or sgZNF117. Experiments were performed in triplicate. d Representative images of tumors in the brain based on IVIS imaging. e Kaplan–Meier survival analysis of mice receiving the indicated treatments. f Representative H&E images of the brain isolated from mice receiving the indicated treatments. Experiments were performed in triplicate. Scale bar, 50 μm. g WB analysis of ZNF117 expression in residual tumors isolated from mice received the indicated treatments. Experiments were performed in triplicate. h Representative immunostaining images of residual tumors isolated from mice received the indicated treatments for expression of Nestin and GalC. Scale bar, 20 μm. i Inhibition of the proliferation of PS30 cells treated with Cas9 with the indicated sgRNA by TMZ (n = 3 biologically independent samples). Data are presented as mean ± SD. Black line represents mean, dots indicate values. Statistical differences were determined by two-tailed student’s t-test. *P-value < 0.05.

Previous studies suggest that GSCs, but not differentiated cells, are resistant to TMZ, and improved treatment of GBM can be achieved through combination of TMZ chemotherapy and differentiation therapy^3,38. Consistently, we found that downregulation of ZNF117 sensitized both PS30 and GS5 cells to TMZ treatment in vitro (Fig. 6i, Supplementary Fig. 11b). We determined the efficacy of TMZ chemotherapy in combination with ZNF117-mediated differentiation therapy in tumor-bearing mice. Seven days after inoculation of PS30 cells, mice were treated with TMZ alone or TMZ in combination with LHNPs loaded with ZNF117-targeted RNP. The same treatment regimen as described above was used. TMZ was given at 0.1 mg/kg intraperitoneally after intravenous administration of LHNPs. We found that treatment with LHNPs loaded with ZNF117-targeted RNP significantly prolonged the survival of mice receiving treatment with LHNPs loaded with GFP-targeted RNP (41 days vs. 33 days, P = 0.007). We found that the combination therapy resulted in greater therapeutic benefits than either TMZ or LHNPs alone (Fig. 6d–f). The median survival of mice in the combination group was 48 days; in comparison, the median survival for the TMZ treatment group and the group treated with LHNPs loaded with ZNF117-targeted RNP were 37 days (P = 0.004) and 42 days (P = 0.02), respectively. The results suggest that ZNF117-mediated GSC differentiation therapy sensitizes GBM to TMZ chemotherapy.