Glioma stem–like cells (GSC) with tumor-initiating activity orchestrate the cellular hierarchy in glioblastoma and engender therapeutic resistance. Recent work has divided GSC into two subtypes with a mesenchymal (MES) GSC population as the more malignant subtype. In this study, we identify the FOXD1–ALDH1A3 signaling axis as a determinant of the MES GSC phenotype. The transcription factor FOXD1 is expressed predominantly in patient-derived cultures enriched with MES, but not with the proneural GSC subtype. shRNA-mediated attenuation of FOXD1 in MES GSC ablates their clonogenicity in vitro and in vivo. Mechanistically, FOXD1 regulates the transcriptional activity of ALDH1A3, an established functional marker for MES GSC. Indeed, the functional roles of FOXD1 and ALDH1A3 are likely evolutionally conserved, insofar as RNAi-mediated attenuation of their orthologous genes in Drosophila blocks formation of brain tumors engineered in that species. In clinical specimens of high-grade glioma, the levels of expression of both FOXD1 and ALDH1A3 are inversely correlated with patient prognosis. Finally, a novel small-molecule inhibitor of ALDH we developed, termed GA11, displays potent in vivo efficacy when administered systemically in a murine GSC-derived xenograft model of glioblastoma. Collectively, our findings define a FOXD1–ALDH1A3 pathway in controling the clonogenic and tumorigenic potential of MES GSC in glioblastoma tumors. Cancer Res; 76(24); 7219–30. ©2016 AACR.

Glioblastoma is the most common and fatal primary brain tumor in adults. Glioblastoma tumors are resistant to conventional radiotherapy and chemotherapies, making the current available treatments ineffective (1). Intratumoral cellular heterogeneity in glioblastoma contributes to tumor aggressiveness and therapy resistance, with glioma stem–like cells (GSC) at the apex of the hierarchy (1, 2). GSCs are poorly differentiated tumor cells with stem cell properties including self-renewal, and are responsible for tumor initiation (3–5). Recently, we and others established patient-derived GSC lines from glioblastoma patients, which contain two distinct and mutually exclusive GSC subtypes termed proneural (PN) and mesenchymal (MES; refs. 4, 6). MES GSCs, the more aggressive and radioresistant subtype, express higher levels of ALDH1A3 than PN GSCs in vitro (6), suggesting that ALDH1A3 is a potential marker of MES GSCs. Therefore, understanding the regulatory pathway(s) controling ALDH1A3 expression in this cell type would be expected to identify new and relevant therapeutic targets.

The Forkhead family of transcription factors (TF) regulate a wide variety of cellular functions during development and many are implicated in cancers (7, 8). As a member of this family, FOXD1 is preferentially expressed in human embryonic tissues, including kidney and testis, but not in adult tissues (9). Moreover, FOXD1 regulates organogenesis (10–12), especially the commitment to a mesenchymal lineage during organogenesis (13–15). FOXD1's regulatory role in stemness is further demonstrated by facilitating the reprogramming of mouse embryonic fibroblasts into induced pluripotent stem cells (16). Recent studies have suggested parallels between reprogramming and tumorigenesis, and highlighted the shared transcription factors involved. Similarly, the deregulation of FOXD1 is implicated in the tumorigenesis of cancers of prostate, breast, and clear cell sarcoma of the kidney (17–19). Nonetheless, the physiologic roles of FOXD1 in brain cancers and GSCs remain unknown.

In this study, we demonstrate that FOXD1 is critical for the maintenance of MES GSCs, and thus, the tumorigenicity of this subtype of glioblastoma tumors. By exploring our genome-wide expression profiling data, we identified FOXD1 as the most upregulated Forkhead TF in the patient-derived MES GSC–enriched cultures. We found that FOXD1 promotes clonogenicity and tumorigenicity of MES GSCs both in vitro and in vivo by the direct transcriptional regulation of the key molecule ALDH1A3. Subsequently, we verified that the tumorigenicity of FOXD1 is mediated by ALDH1A3. We also proved that FOXD1 and ALDH1A3 are prognostic factors in glioma clinical samples. Finally, we developed novel anti-ALDH small-molecule inhibitors and demonstrated effectiveness in vitro and in vivo.

Ethics

This study was performed under the supervision of the respective Institutional Animal Care and Use Committees, and the Human Subjects Research protocols were approved by the Institutional Review Board at the University of Alabama at Birmingham (Birmingham, AL), MD Anderson Cancer Center (MDACC, Houston, TX) and/or Ohio State University (OSU, Columbus, OH) as described previously (6, 20).

GSC cultures

The glioma (neuro)spheres used in this study were generated at OSU and MDACC. Neurosphere cultures from clinical samples were established and characterized as described previously (Supplementary Table S1; refs. 6, 20). The detailed source, year of receipt, and culture methods are described in Supplementary Information. The unique identities of each glioma neurosphere line were confirmed by short tandem repeat (STR) analysis as described in Supplementary Table S2 (20).

Mouse intracranial xenograft tumor models

The GSC suspension (1 × 104 cells for MES83 and 2.5 × 105 cells for MES267) was injected into the brains of nude mice (6-week-old) as described previously (20, 21). When neurological symptoms were observed, mice were sacrificed and mouse brains were collected for analysis.

Lentivirus transduction

Two independent lentiviral shRNA constructs for knocking down FOXD1 were purchased from Sigma (TRCN0000013970 and TRCN0000230322). For overexpression, cDNAs of FOXD1 (RC220504, Origene), and FOXG1 (RC207964, Origene) were subcloned into a lentiviral vector (Origene, PS100064) according to the manufacturer's protocol. Lentiviruses were packaged in 293FT cells (from Invitrogen at 2013). The lentivirus transduction was performed as described previously (20).

Neurosphere formation assay

MES83 and MES28 GSCs infected with lentivirus were seeded into 96-well plates at 1, 10, 20, 30, 40, and 50 cells per well. After 7 days for MES83 GSCs and 10 days for MES28 GSCs, the numbers of spheres with diameters greater than 60 μm were counted. Data were analyzed as described previously (http://bioinf.wehi.edu.au/software/elda/; ref. 22).

IHC scoring

German immunohistochemical score (GIS) was used to evaluate the expression of FOXD1 and ALDH1A3 (23, 24). The detailed procedure is provided in the Supplementary Information.

Western blot analysis

The detailed procedure is provided in the Supplementary Information. Original film scans of Western blots are shown in Supplementary Fig. S1.

GEO accession numbers

The accession numbers of GEO datasets used in this study are GSE67089, GSE4290, GSE4536, and GSE2223.

Drosophila stock

The following transgenic Drosophila flies were used: repoGAL4 UAS-GFP/TM3, Sb, UAS-ptenRNAi, UAS-RasV12/Cyo, UAS-aldhRNAi/Cyo, ptcGAL4 UAS-GFP/Cyo, UAS-fd59ARNAi, and UAS-fd59A. The Drosophila flies for all experiments were incubated at 25°C.

Immunofluorescence, imaging, and quantification in Drosophila samples

Immunofluorescence was performed as described previously (25) and the images were captured using the Olympus Fluoview 1000 confocal microscope.

Statistical analysis

Data are presented as mean ± SD. The number of replicates for each experiment is stated in the figure legends. Statistical differences between and among groups were determined by two-tailed t test and one-way ANOVA followed by Dunnett post test, respectively. The statistical significance of Kaplan–Meier survival plot was determined by log-rank analysis. Statistical analysis was performed by Microsoft Excel 2013 and GraphPad Prism 6.0, unless mentioned otherwise in the figure legend. P < 0.05 was considered as statistically significant.

Additional details about the materials and methods are available in the Supplementary Information.

FOXD1 and FOXG1 exhibit inverse expression pattern in GSCs

To identify the TFs critical for the GSC phenotype, we explored our dataset [GSE67089 (6)] with 30 patient-derived glioma sphere cultures and 3 human fetal brain–derived sphere cultures (normal spheres) for the expression levels of all TFs (1,988 genes; ref. 26). Twenty-eight TFs were found upregulated in MES glioma spheres by more than 10-fold compared with normal neural progenitor cells. Ten TFs were at higher levels in MES glioma spheres by more than 11-fold than the levels of these TFs in PN glioma spheres. Among them, seven TFs were overlapped, including FOSL1, FOXD1, PLAGL1, STAT6, ARNTL2, BNC1, and HTAIP2 (Fig. 1A).

Figure 1.

FOXD1 is a key MES GSC transcription factor. A, Genome-wide transcriptome microarray analysis (GSE67089) shows that FOXD1 is one of seven upregulated transcription factors in MES glioma spheres when compared with neural progenitors (NP; >10-fold) and PN glioma spheres (>11-fold). B, mRNA expression levels of the Forkhead TF family members in the GSE67089 dataset reveal that FOXD1 and FOXG1 are the highest expressed genes in MES and PN glioma spheres, respectively. C and D, qRT-PCR analyses of FOXD1 (C; MES vs. PN, P < 0.001, n = 3; MES vs. mixed, P < 0.001, n = 3, one-way ANOVA) and FOXG1 (D; MES vs. PN, P = 0.0025, n = 3; MES vs. mixed, P = 0.2537, n = 3, one-way ANOVA) mRNA in the indicated glioma spheres. E, Western blot analyses of ALDH1A3, FOXD1, FOXG1, CD44, AXL, and Olig2 expression in the indicated glioma spheres. β-Actin served as a loading control. F, Immunocytochemistry analyses of FOXD1 in MES83, PN84, and PN157 glioma spheres. Hoechst was used for nuclear staining. Scale bar, 50 μm. G, qRT-PCR analyses of FOXD1 mRNA in ALDHhigh and ALDHlow cells derived from MES83 glioma spheres. (P < 0.001, n = 3, t test) H, qRT-PCR analyses of FOXG1 mRNA in CD133high and CD133low cells derived from PN157 glioma spheres. (P = 0.0017, n = 3, t test).

Figure 1.

FOXD1 is a key MES GSC transcription factor. A, Genome-wide transcriptome microarray analysis (GSE67089) shows that FOXD1 is one of seven upregulated transcription factors in MES glioma spheres when compared with neural progenitors (NP; >10-fold) and PN glioma spheres (>11-fold). B, mRNA expression levels of the Forkhead TF family members in the GSE67089 dataset reveal that FOXD1 and FOXG1 are the highest expressed genes in MES and PN glioma spheres, respectively. C and D, qRT-PCR analyses of FOXD1 (C; MES vs. PN, P < 0.001, n = 3; MES vs. mixed, P < 0.001, n = 3, one-way ANOVA) and FOXG1 (D; MES vs. PN, P = 0.0025, n = 3; MES vs. mixed, P = 0.2537, n = 3, one-way ANOVA) mRNA in the indicated glioma spheres. E, Western blot analyses of ALDH1A3, FOXD1, FOXG1, CD44, AXL, and Olig2 expression in the indicated glioma spheres. β-Actin served as a loading control. F, Immunocytochemistry analyses of FOXD1 in MES83, PN84, and PN157 glioma spheres. Hoechst was used for nuclear staining. Scale bar, 50 μm. G, qRT-PCR analyses of FOXD1 mRNA in ALDHhigh and ALDHlow cells derived from MES83 glioma spheres. (P < 0.001, n = 3, t test) H, qRT-PCR analyses of FOXG1 mRNA in CD133high and CD133low cells derived from PN157 glioma spheres. (P = 0.0017, n = 3, t test).

Close modal

Forkhead TFs are involved in deciding cell fates during development and tumorigenesis (7, 16). Our previous study demonstrated the requirement of one Forkhead TF, FOXM1, for the survival and proliferation of oncogenic but not normal neural stem cells (27–29). Therefore, we focused on the Forkhead TFs. FOXD1 was the most upregulated MES-associated gene in the Forkhead family, whereas FOXG1 was the most upregulated gene in PN glioma spheres (Fig. 1B; Supplementary Fig. S2A and S2B). This was validated by qRT-PCR (Fig. 1C and D) and Western blot analysis (Fig. 1E) with six patient-derived GSC-enriched cultures. In Western blot assays, ALDH1A3, CD44, and AXL were included as MES markers while OLIG2 as a PN marker. Similarly, immunocytochemistry exhibited that glioma sphere line MES83 has a higher expression of nuclear FOXD1 than PN84 and PN157 (Fig. 1F). We then investigated whether the stem cell fraction in MES and PN glioma spheres express FOXD1 and FOXG1, respectively. Our previous studies identified CD133 as a PN GSC marker while ALDH1A3 as a MES GSC marker (4, 6, 27). As expected, FOXD1 was almost exclusively expressed in the ALDHhigh subfraction of MES83 cells, but not in ALDHlow subpopulations (Fig. 1G and Supplementary Fig. S2C). In contrast, the CD133high subpopulation had significantly higher expression of FOXG1 than CD133low subpopulations in the PN157 glioma spheres (Fig. 1H and Supplementary Fig. S2D). To extend this observation, we performed bioinformatics analysis using another clinical data (IVY GBM Atlas Project database; http://glioblastoma.alleninstitute.org/). Glioblastoma is known to display intratumoral cellular heterogeneity. We investigated RNA-seq data that was collected from different regions of glioblastoma tumor tissues that exhibit distinct cellular subtypes including PN and MES tumor cells. This analysis showed an inverse pattern of the expression levels of FOXG1 and FOXD1 in different regions (Supplementary Fig. S2E and S2F). In contrast, there were no significant differences in either FOXD1 or FOXG1 among the glioblastoma subtypes in the TCGA dataset (Supplementary Fig. S2G). The TCGA dataset measures the average expression levels of genes. This maybe the possible reason why the expression level of either FOXG1 or FOXD1 failed to show a significant difference among the four subtypes of glioblastoma. Collectively, these data suggest that FOXD1 and FOXG1 exhibit inverse expression profiles in vitro: FOXD1 is elevated in MES, whereas FOXG1 is elevated in PN GSCs.

FOXD1 expression is associated with poorer prognosis in glioma patients

We then examined whether FOXD1 expression is associated with the histopathologic grade of glioma and patient prognosis. With immunohistochemistry, 44 glioma tumor tissues with varying grades and nine adjacent normal brain tissues were analyzed. Intense immunostaining, indicating FOXD1 expression, was observed in the nuclei of high-grade glioma tissues (grade II: 2 of 4; grade III: 12 of 16; grade IV: 10 of 24) but rarely in normal brain tissues (1/9; Fig. 2A and B). The survival periods of patients with high levels of FOXD1 were significantly shorter than those with FOXD1 intermediate expression (Fig. 2C, P = 0.0039). Inversely, patients with low FOXD1 expression exhibited better prognosis than those with high expression (Fig. 2C, P = 0.0324). Consistently, glioblastoma-derived spheres expressed higher FOXD1 than normal brain-derived spheres (GSE4536, Fig. 2D; ref. 30). Similar results were obtained from two other datasets (GSE2223, Fig. 2E; ref. 31; and GSE4290, Fig. 2F; ref. 32), and from the Rembrandt database (Fig. 2G). The elevated expression of FOXD1 was also associated with poorer survival in Rembrandt database (Fig. 2H). Altogether, these data suggest that the expression of FOXD1 is elevated in highgrade glioma and is a clinically relevant target in glioblastoma.

Figure 2.

FOXD1 expression is clinically relevant in high-grade gliomas. A and B, Representative immunohistochemical images (A) and analyses (B) of FOXD1 in WHO grade IV (glioblastoma), grade III glioma, grade II glioma, and nontumor brain samples. Scale bar, 50 μm. (nontumor, n = 9; grade II, n = 4; grade III, n = 16; grade IV, n = 24). C, Kaplan–Meier analyses evaluating the correlation between FOXD1 protein expression and survival of 40 high-grade glioma patients (FOXD1 high vs. low, P = 0.0324; FOXD1 high vs. intermediate, P = 0.0039; FOXD1 intermediate vs. low, P = 0.2896, log-rank test). D–F, Analyses of the indicated GEO datasets show a higher expression of FOXD1 in glioma than in NSC samples and nontumor tissues. D, Lee dataset [GSE4536; NSCs, n = 3; glioblastoma (GBM), n = 22; GSCs, n = 20; P < 0.001, one-way ANOVA, probe set 206307_s_at]. E, Bredel dataset (GSE2223; nontumor, n = 4; glioblastoma, n = 29; P < 0.0001, t test, probe set 1876). F, Sun dataset (GSE4290, nontumor, n = 23; grade II, n = 45; grade III, n = 31; glioblastoma, n = 81; P < 0.0001, one-way ANOVA, probe set 206307_s_at). G, Analysis of the Rembrandt data shows a higher FOXD1 mRNA expression in astrocytoma (n = 148), oligodendroglioma (n = 67), mixed groups (n = 11), and in glioblastoma samples (n = 228) than nontumor samples (n = 28; probe set 206307_s_at). H, Analysis of the Rembrandt database indicates the inverse correlation between FOXD1 mRNA expression and post-surgical survival of glioma patients (P = 0.0171, FOXD1 upregulated > 1.3-fold, n = 29 vs. FOXD1 downregulated < −1.3-fold, n = 42; P = 0.0243, downregulated < −1.3-fold, n = 42 vs. intermediate n = 470, probe set: 206307_s_at).

Figure 2.

FOXD1 expression is clinically relevant in high-grade gliomas. A and B, Representative immunohistochemical images (A) and analyses (B) of FOXD1 in WHO grade IV (glioblastoma), grade III glioma, grade II glioma, and nontumor brain samples. Scale bar, 50 μm. (nontumor, n = 9; grade II, n = 4; grade III, n = 16; grade IV, n = 24). C, Kaplan–Meier analyses evaluating the correlation between FOXD1 protein expression and survival of 40 high-grade glioma patients (FOXD1 high vs. low, P = 0.0324; FOXD1 high vs. intermediate, P = 0.0039; FOXD1 intermediate vs. low, P = 0.2896, log-rank test). D–F, Analyses of the indicated GEO datasets show a higher expression of FOXD1 in glioma than in NSC samples and nontumor tissues. D, Lee dataset [GSE4536; NSCs, n = 3; glioblastoma (GBM), n = 22; GSCs, n = 20; P < 0.001, one-way ANOVA, probe set 206307_s_at]. E, Bredel dataset (GSE2223; nontumor, n = 4; glioblastoma, n = 29; P < 0.0001, t test, probe set 1876). F, Sun dataset (GSE4290, nontumor, n = 23; grade II, n = 45; grade III, n = 31; glioblastoma, n = 81; P < 0.0001, one-way ANOVA, probe set 206307_s_at). G, Analysis of the Rembrandt data shows a higher FOXD1 mRNA expression in astrocytoma (n = 148), oligodendroglioma (n = 67), mixed groups (n = 11), and in glioblastoma samples (n = 228) than nontumor samples (n = 28; probe set 206307_s_at). H, Analysis of the Rembrandt database indicates the inverse correlation between FOXD1 mRNA expression and post-surgical survival of glioma patients (P = 0.0171, FOXD1 upregulated > 1.3-fold, n = 29 vs. FOXD1 downregulated < −1.3-fold, n = 42; P = 0.0243, downregulated < −1.3-fold, n = 42 vs. intermediate n = 470, probe set: 206307_s_at).

Close modal

FOXD1 is required for the clonogenicity of MES glioblastoma in vitro and in vivo

To investigate the physiologic role of FOXD1 in MES GSCs, we knocked down the expression of FOXD1 using two different lentiviral shRNA vectors in MES glioma sphere lines (MES83 and MES28; refs. 4, 6); a nontargeting shRNA (shNT) was used as a negative control. Western blotting showed that both shRNA clones were capable of knocking down FOXD1, with clone #1 having higher efficiency (Fig. 3A). As a result, MES glioma spheres with FOXD1 knocked-down (KD) grew significantly slower (33) than control cells (Fig. 3B). Moreover, the sphere-forming capacity of these cells was dramatically decreased (Fig. 3C and D). Of note, the extent to which cell growth and clonogenicity were inhibited was more prominent with shFOXD1#1 than shFOXD1#2, which was consistent with their effects on FOXD1 KD. Our previous study demonstrated that MES, not PN GSCs, depended more on glycolysis (6). As expected, lactic acid levels as a measure of glycolysis were largely reduced in MES83 cells transduced with shFOXD1#1, indicating that FOXD1 may be involved in the regulation of glycolysis of MES GSCs (Supplementary Fig. S3). To determine whether FOXD1 is essential for tumorigenicity in MES glioma spheres in vivo, we used orthotopic xenografts into mouse brains. Injection of MES83 spheres into the striatum of immunocompromised mouse brains resulted in lethal tumors within 30 days (6, 34). Although the cells transduced with the lentiviral shNT construct did not show altered tumorigenesis or subsequent mouse survival, xenografting of the FOXD1-silenced MES glioma spheres diminished the proportion of tumor formation (3/5 mice) and prolonged survival periods (Fig. 3E). Histologically, the tumors in the control mice were highly vascularized glioblastoma-like brain tumors with central necrosis, and developed by day 18 after transplantation (Fig. 3F and G).

Figure 3.

FOXD1 regulates MES GSC growth both in vitro and in vivo. A, Western blot analyses of MES83 and MES28 glioma spheres transduced with shRNA targeting FOXD1 (shFOXD1#1 or shFOXD1#2) or a nontargeting control (shNT). B,In vitro growth assay shows that shRNAs targeting FOXD1 (shFOXD1#1 and shFOXD1#2) inhibit cell proliferation of MES83 and MES28 glioma spheres (P < 0.0001, n = 6, one-way ANOVA). C, Representative images of MES83 and MES28 glioma spheres transduced with shRNA targeting FOXD1. shNT served as a control. Scale bar, 60 μm. D,In vitro clonogenicity assays (limiting dilution neurosphere formation assays) indicate that FOXD1 shRNA decreases clonogenicity of MES83 and MES28 cells (MES83, P < 0.001; and MES28, P < 0.001, ELDA analyses). E, Kaplan–Meier analysis of nude mice harboring intracranial tumors derived from MES83 GSCs transduced with shNT (n = 6) or shFOXD1#1 (n = 5; P = 0.0014, with log-rank test) F and G, Representative images of brains (F) and H&E-stained brain sections (G) of mice after intracranial transplantation of MES83 glioma spheres transduced with shNT or shFOXD1#1. Scale bar, 1 mm (G, top) and 100 μm (G, bottom).

Figure 3.

FOXD1 regulates MES GSC growth both in vitro and in vivo. A, Western blot analyses of MES83 and MES28 glioma spheres transduced with shRNA targeting FOXD1 (shFOXD1#1 or shFOXD1#2) or a nontargeting control (shNT). B,In vitro growth assay shows that shRNAs targeting FOXD1 (shFOXD1#1 and shFOXD1#2) inhibit cell proliferation of MES83 and MES28 glioma spheres (P < 0.0001, n = 6, one-way ANOVA). C, Representative images of MES83 and MES28 glioma spheres transduced with shRNA targeting FOXD1. shNT served as a control. Scale bar, 60 μm. D,In vitro clonogenicity assays (limiting dilution neurosphere formation assays) indicate that FOXD1 shRNA decreases clonogenicity of MES83 and MES28 cells (MES83, P < 0.001; and MES28, P < 0.001, ELDA analyses). E, Kaplan–Meier analysis of nude mice harboring intracranial tumors derived from MES83 GSCs transduced with shNT (n = 6) or shFOXD1#1 (n = 5; P = 0.0014, with log-rank test) F and G, Representative images of brains (F) and H&E-stained brain sections (G) of mice after intracranial transplantation of MES83 glioma spheres transduced with shNT or shFOXD1#1. Scale bar, 1 mm (G, top) and 100 μm (G, bottom).

Close modal

FOXD1 regulates ALDH1A3 transcription in MES glioma spheres

Because FOXD1 is almost exclusively expressed in ALDHhigh MES glioblastoma cells, not in ALDHlow cells (Fig 1G), we examined whether FOXD1 transcriptionally regulates the ALDH1A3 gene, thereby orchestrating the stem cell properties in MES GSCs. Indeed, FOXD1 silencing in the MES83 glioma spheres significantly decreased the ALDH1A3 mRNA and protein levels (Fig. 4A and B). As there is a common Forkhead TF binding motif in the promoter region of the ALDH1A3 gene (Supplementary Fig. S4A), we hypothesized that FOXD1 activated ALDH1A3 transcription. Thus, we performed a luciferase assay to measure ALDH1A3 promoter activity upon exogenous expression of FOXD1 in 293T cells and MES83 spheres (Fig. 4C and D). As expected, ectopically expressed FOXD1 increased ALDH1A3 reporter activity. As FOXD1 and FOXG1 exhibit mutually exclusive expression patterns (Fig. 1B–D), we assumed that FOXG1 may counteract FOXD1 activity in MES GSCs. Intriguingly, coexpression of FOXG1 and FOXD1 in MES83 spheres counteracted ALDH1A3 reporter activities driven by FOXD1 alone (Fig. 4C and D). Furthermore, the transcriptional regulation of ALDH1A3 by FOXD1 was confirmed by chromatin immunoprecipitation PCR, which indicated that FOXD1 directly binds to the promoter region of the ALDH1A3 gene in MES28 glioma spheres (Fig. 4E). In addition, the overexpression of FOXD1 in MES83 increased the expression of ALDH1A3 (Fig. 4F). To investigate the physiologic role of the FOXD1–ALDH1A3 axis in MES GSCs, we evaluated whether ALDH1A3 overexpression rescues the phenotypes of MES glioma spheres induced by FOXD1 KD. The reduced in vitro cell growth and neurosphere formation by FOXD1 KD were partially, yet not completely, restored by the overexpression of ALDH1A3, but not by the control vector (Fig. 4G and H and Supplementary Fig. S4B). On the other hand, FOXD1 overexpression alone did not induce any noticeable changes in the PN spheres, at least in the expressions of representative markers (AXL, CD44, and Olig2; Supplementary Fig. S4C). We also analyzed GSE67089 dataset in our previous publication (6) for the expression levels of FOXD1 with all the members of ALDH family. We consistently found that FOXD1 is coexpressed with ALDH1A3, but not other ALDH members, in MES but not in PN subtype spheres (Supplementary Fig. S4D). Collectively, these data suggest that FOXD1 is required, but not sufficient, for the establishment of the MES phenotype in GSCs.

Figure 4.

ALDH1A3 is a functional MES GSC marker and is transcriptionally regulated by FOXD1. A, qRT-PCR analyses of ALDH1A3 and FOXD1 mRNA in MES83 glioma spheres transduced with shFOXD1#1 or shNT (P < 0.001, n = 3, with t test). B, Western blot analyses of ALDH1A3 in MES83 and MES28 glioma spheres transduced with shFOXD1#1, shFOXD1#2, or shNT. β-Actin served as a loading control. C and D, Luciferase assays with 293T cells (C) and MES83 glioma spheres (D) cotransfected with an ALDH1A3 promoter reporter plasmid together with overexpression vectors for the indicated genes (n = 3). E, ChIP-qPCR assay using Myc antibody or control IgG in MES28 glioma spheres transfected with a FOXD1 (Myc-DDK-tagged) plasmid shows the binding of FOXD1 on the ALDH1A3 promoter (P < 0.001, n = 3, t test). F, Western blot analyses of the indicated proteins in MES83 glioma spheres transduced with the overexpression plasmids encoding FOXD1, FOXG1, or empty vector. G and H,ALDH1A3 overexpression partially restores the in vitro proliferation (G; P < 0.001, one-way ANOVA) and neurosphere formation capacities (H; P < 0.001, ELDA analysis), which are inhibited by shFOXD1#1 or #2 in MES83 glioma spheres.

Figure 4.

ALDH1A3 is a functional MES GSC marker and is transcriptionally regulated by FOXD1. A, qRT-PCR analyses of ALDH1A3 and FOXD1 mRNA in MES83 glioma spheres transduced with shFOXD1#1 or shNT (P < 0.001, n = 3, with t test). B, Western blot analyses of ALDH1A3 in MES83 and MES28 glioma spheres transduced with shFOXD1#1, shFOXD1#2, or shNT. β-Actin served as a loading control. C and D, Luciferase assays with 293T cells (C) and MES83 glioma spheres (D) cotransfected with an ALDH1A3 promoter reporter plasmid together with overexpression vectors for the indicated genes (n = 3). E, ChIP-qPCR assay using Myc antibody or control IgG in MES28 glioma spheres transfected with a FOXD1 (Myc-DDK-tagged) plasmid shows the binding of FOXD1 on the ALDH1A3 promoter (P < 0.001, n = 3, t test). F, Western blot analyses of the indicated proteins in MES83 glioma spheres transduced with the overexpression plasmids encoding FOXD1, FOXG1, or empty vector. G and H,ALDH1A3 overexpression partially restores the in vitro proliferation (G; P < 0.001, one-way ANOVA) and neurosphere formation capacities (H; P < 0.001, ELDA analysis), which are inhibited by shFOXD1#1 or #2 in MES83 glioma spheres.

Close modal

FOXD1 and ALDH1A3 are evolutionarily conserved genes that contribute to glial neoplasms

Previously, we reported that ALDHhigh MES glioma sphere cells have higher clonogenic potential in vitro than ALDHlow cells in MES glioblastoma (6). To compare the in vivo tumorigenic abilities of ALDHhigh and ALDHlow cells in MES glioma spheres, we injected ALDHhigh or ALDHlow MES83 cells into mouse brains at two different cell numbers (100 and 1,000 cells). When 100 ALDHlow or ALDHhigh cells were injected into mouse brains, only 1 of 6 mice injected with ALDHlow cells died by 24th day after transplantation. In contrast, 4 of 6 mice in the ALDHhigh group died due to tumor burden. Similar results were observed when 1,000 tumor cells were injected (Fig. 5A), indicating that the tumor-initiating cells reside in the minor subset of MES glioblastoma cells with elevated ALDH1 activity (approximately 7% of the MES83 cells).

Figure 5.

RNA interference–mediated silencing of FD59A and ALDH attenuates growth of Drosophila glial neoplasia. A, ALDEFLUOR assay (top) in MES83 glioma spheres with or without ALDH1 inhibitor DEAB. Bottom, frequencies of tumor formation of ALDH1high and ALDH1low cell populations of MES83 glioma spheres in mice. B, Expression of Fd59A (Drosophila ortholog of FOXD1; red, gray) in the Drosophila CNS derived from larvae of repoGAL4 UASGFP and repoGAL4 UASGFP UASPTENRNAi UASRasv12. Glial cells are marked by GFP (green). C, Expression levels of ALDH (red, gray) in the larval CNS of repoGAL4 UASGFP (wild-type), repoGAL4 UASGFP UASPTENRNAi, repoGAL4 UASGFP UASRasv12, repoGAL4 UASGFP UASPTENRNAi UASRasv12, and repoGAL4 UASGFP UASPTENRNAi UASRasv12 fd59ARNAi. D, The effects of ALDHRNAi and fd59ARNAi on the growth of glial neoplasms of repoGAL4 UASPTENRNAi UASRasv12 larvae. Both RepoGAL4 UASGFP (wild-type) and repoGAL4 UASPTENRNAi UASRasv12 samples are included for comparison. E, The quantification of brain tumor volume (brain lobe size in pixels) from the indicated larvae (repoGAL4 UASPTENRNAi UASRasv12 ALDHRNAi vs. repoGAL4 UASPTENRNAi UASRasv12, P = 0.008, n = 3, one-way ANOVA; repoGAL4 UASPTENRNAi UASRasv12 fd59ARNAi vs. repoGAL4 UASPTENRNAi UASRasv12, P = 0.007, n = 3, one-way ANOVA). F, The schematic diagram depicts that ALDH and fd59A (dFOXD1) are evolutionarily conserved genes contributing tumorigenesis of glial neoplasms.

Figure 5.

RNA interference–mediated silencing of FD59A and ALDH attenuates growth of Drosophila glial neoplasia. A, ALDEFLUOR assay (top) in MES83 glioma spheres with or without ALDH1 inhibitor DEAB. Bottom, frequencies of tumor formation of ALDH1high and ALDH1low cell populations of MES83 glioma spheres in mice. B, Expression of Fd59A (Drosophila ortholog of FOXD1; red, gray) in the Drosophila CNS derived from larvae of repoGAL4 UASGFP and repoGAL4 UASGFP UASPTENRNAi UASRasv12. Glial cells are marked by GFP (green). C, Expression levels of ALDH (red, gray) in the larval CNS of repoGAL4 UASGFP (wild-type), repoGAL4 UASGFP UASPTENRNAi, repoGAL4 UASGFP UASRasv12, repoGAL4 UASGFP UASPTENRNAi UASRasv12, and repoGAL4 UASGFP UASPTENRNAi UASRasv12 fd59ARNAi. D, The effects of ALDHRNAi and fd59ARNAi on the growth of glial neoplasms of repoGAL4 UASPTENRNAi UASRasv12 larvae. Both RepoGAL4 UASGFP (wild-type) and repoGAL4 UASPTENRNAi UASRasv12 samples are included for comparison. E, The quantification of brain tumor volume (brain lobe size in pixels) from the indicated larvae (repoGAL4 UASPTENRNAi UASRasv12 ALDHRNAi vs. repoGAL4 UASPTENRNAi UASRasv12, P = 0.008, n = 3, one-way ANOVA; repoGAL4 UASPTENRNAi UASRasv12 fd59ARNAi vs. repoGAL4 UASPTENRNAi UASRasv12, P = 0.007, n = 3, one-way ANOVA). F, The schematic diagram depicts that ALDH and fd59A (dFOXD1) are evolutionarily conserved genes contributing tumorigenesis of glial neoplasms.

Close modal

Next, we examined the functional role of FOXD1 and ALDH1A3 in Drosophila. In flies, the orthologs of FOXD1 and ALDH1A3 are fd59A (35) and DmALDH (36), respectively. fd59A is highly expressed in the Drosophila larval (embryonic) central nervous system (CNS; ref. 35). We used recently established Drosophila glioma model involving coactivation of oncogenic Ras and PI3K pathways by the GAL4 UAS system (repoGAL4 UASPTENRNAi UASRasV12), which causes the formation of invasive glial neoplasms that mimic human gliomas (37). In the larval glial neoplasms derived from the repoGAL4 UASPTENRNAi; UASRasV12 larvae, Fd59A, and ALDH levels were substantially upregulated compared with the normal CNS (repoGAL4 UASGFP; Fig. 5B and C). To investigate whether the formation of the glial neoplasms in Drosophila requires ALDH, we transduced UASALDHRNAi into Drosophila glioma tumors. Elimination of ALDH resulted in reduction in glioma growth by 1.7-fold (Fig. 5D and E, P = 0.008). This antitumor effect of ALDH silencing was also observed in the mosaic models of Drosophila eye cancer (Supplementary Fig. S5). Next, we investigated the effects of downregulation of Fd59A in these cancers. As shown in Fig. 5D and E, downregulation of Fd59A resulted in a reduction in glioma growth (P = 0.007) with a concomitant downregulation of ALDH levels (Fig. 5C). Taken together, these data suggest that FOXD1 and ALDH1A3 are evolutionarily conserved genes that contribute to CNS tumorigenesis (Fig. 5F).

ALDH1A3 expression indicates poor prognosis of post-surgical glioma patients

To interrogate whether ALDH1A3 is a clinically relevant biomarker, we analyzed 40 high-grade glioma patient samples (Fig. 6A). The log-rank test showed that the glioma patients with higher ALDH1A3 protein expression has significantly shorter post-surgical survival periods compared with either the intermediate or low ALDH1A3 protein expression groups (Fig. 6B, P = 0.0285, ALDH1A3high vs. ALDH1A3intermediate; and P = 0.0016, ALDH1A3high vs. ALDH1A3low). In addition, the data from Rembrandt database demonstrates a similar pattern (Supplementary Fig. S6). These data indicate the clinical significance of ALDH1A3.

Figure 6.

The novel ALDH small-molecule inhibitor GA11 attenuates MES GSC growth both in vitro and in vivo. A, Representative immunofluroescence images of ALDH1A3 in 40 high-grade glioma samples. DAPI was used for nuclear labeling. Scale bar, 50 μm. B, Kaplan–Meier analysis of ALDH1A3 expression indicates the negative correlation between ALDH1A3 protein expression and survival in high-grade glioma patients. (ALDH1A3 high vs. ALDH1A3 intermediate, P = 0.0285; ALDH1A3 high vs. ALDH1A3 low, P = 0.0016; and ALDH1A3 intermediate vs. ALDH1A3 low, P = 0.1769, with log-rank test). C, The comparison of the essential core structure of the naturally occurring ALDH inhibitor, daidzin, and the structures of synthesized novel imidazo [1,2-a] pyrimidine ALDH inhibitors, GA11 and GA23. D, Log dose–response analysis of GA11 (top) and GA23 (bottom) in yeast. E, Flow cytometry analyses using ALDEFLUOR indicate that both GA11 and GA23 (5 μmol/L, 30 minutes) inhibit ALDH activity in MES83 glioma spheres. F, Log dose–response analyses of the effects of GA11 on the viabilities of MES83, MES267, PN157, PN711, glioma spheres, and NHA cells. G and H, Treatment with GA11 (intraperitoneal injection, 20 mg/kg for 7 days from day 7) prolongs survival periods of mice bearing MES83-derived intracranial tumors (G; P = 0.0096, with log-rank test) and those of mice with MES267-derived intracranial tumors (H; P = 0.0262, with log-rank test).

Figure 6.

The novel ALDH small-molecule inhibitor GA11 attenuates MES GSC growth both in vitro and in vivo. A, Representative immunofluroescence images of ALDH1A3 in 40 high-grade glioma samples. DAPI was used for nuclear labeling. Scale bar, 50 μm. B, Kaplan–Meier analysis of ALDH1A3 expression indicates the negative correlation between ALDH1A3 protein expression and survival in high-grade glioma patients. (ALDH1A3 high vs. ALDH1A3 intermediate, P = 0.0285; ALDH1A3 high vs. ALDH1A3 low, P = 0.0016; and ALDH1A3 intermediate vs. ALDH1A3 low, P = 0.1769, with log-rank test). C, The comparison of the essential core structure of the naturally occurring ALDH inhibitor, daidzin, and the structures of synthesized novel imidazo [1,2-a] pyrimidine ALDH inhibitors, GA11 and GA23. D, Log dose–response analysis of GA11 (top) and GA23 (bottom) in yeast. E, Flow cytometry analyses using ALDEFLUOR indicate that both GA11 and GA23 (5 μmol/L, 30 minutes) inhibit ALDH activity in MES83 glioma spheres. F, Log dose–response analyses of the effects of GA11 on the viabilities of MES83, MES267, PN157, PN711, glioma spheres, and NHA cells. G and H, Treatment with GA11 (intraperitoneal injection, 20 mg/kg for 7 days from day 7) prolongs survival periods of mice bearing MES83-derived intracranial tumors (G; P = 0.0096, with log-rank test) and those of mice with MES267-derived intracranial tumors (H; P = 0.0262, with log-rank test).

Close modal

The novel ALDH1 inhibitor GA11 has anti-glioblastoma effects in vitro and in vivo

On the basis of the inhibitory effects of FOXD1 silencing on the MES GSC–derived mouse brain tumors and of ALDH1A3 silencing on the Drosophila brain cancer model, we sought to design novel, clinically efficacious small-molecule inhibitors selectively targeting the ALDH1 activity for glioblastoma therapies. The natural product daidzin and its congeners have been reported to inhibit ALDH1 (38). To design clinically applicable analogs with better pharmacokinetic properties (39), we identified the imidazo [1,2-a] pyrimidine heterocyclic core as the essential scaffold. We added two planar, aromatic, and lipophilic areas to this scaffold in positions two and six of the nucleus, thereby generating the novel small molecules GA11 and GA23 (Fig. 6C and Supplementary Fig. S7A). The structural assessment confirmed that the replacement of glucopyranose portion of daidzin with an aromatic ring reduces both hydrogen bond forming potential and rotatable bonds, resulting in smaller molecules with a lower polar surface area and limited flexibility. This modification retained the core structure of daidzin essential for inhibiting ALDH1 activity, while establishing a drug-like profile, that is, a more favorable prediction of the blood–brain barrier (BBB) penetration. Indeed, an in silico evaluation confirmed that the physical–chemical properties of these novel compounds were fully consistent with those required for BBB penetration (Supplementary Table S3; ref. 40). In addition, the molecular weights of GA11 and GA23 (close to 310, the mean value of marketed CNS drugs) are significantly lower than that of daidzin. Moreover, the H-bonding potential of these novel compounds is commensurate to that of successful CNS drug candidates. Indeed, the sum of heteroatoms capable of hydrogen bonding, like nitrogen (N) and oxygen (O), is less than five, conferring on the molecules a high probability of entering the CNS. In addition, the rotatable bond count is less than eight, and the limited flexibility of the molecules should promote passive permeation through BBB. Lowering both molecular weight and H-bonding potential have a significant effect on the polar surface area (PSA) of the compounds, which is a key factor in determining the extent of BBB penetration. Unlike daidzin, the novel compounds have PSA values lower than the stringent cutoff 60–70 Å, which characterizes commercial CNS drugs (40).

To validate the inhibitory effects, we first determined the enzymatic activity of yeast ALDH incubated with GA11 and GA23 in an in vitro enzymatic assay (30% of the ALDH sequences in yeast and eukaryotes are conserved; ref. 41). As expected, both GA11 and GA23 inhibited the ALDH enzymatic activity with IC50 values in the micromolar range (Fig. 6D). GA11 was more potent. Furthermore, both GA11 and GA23 inhibited human ALDH1 demonstrated by a marked reduction of ALDHhigh cellular populations in MES83 glioma spheres treated with GA11 and GA23 (Fig. 6E). As a result, both GA11 and GA23 inhibited the growth of MES glioma spheres. In contrast, PN glioma spheres were relatively more resistant to these two compounds (Fig. 6F and Supplementary Fig. S7B–S7D). More importantly, systemic treatment of MES83- and MES267-based mouse brain tumors with GA11 significantly attenuated tumor growth, thereby extending the survival of tumor-bearing mice compared with the vehicle-treated counterparts (Fig. 6G and H and Supplementary Fig. S7E).

Glioblastomas display intratumoral cellular heterogeneity; within the same tumor, different cell populations respond differently to therapies (3, 42, 43). GSCs are at the apex of the cellular hierarchy (3, 42). Thus, understanding how GSCs are maintained may improve the efficacies of current therapeutic strategies. Recent evidence suggests that GSCs are subclassified into two subtypes with MES GSCs being more therapy resistant (4, 6). Therefore, identification of MES GSC–regulatory molecules has the potential to lead to novel and efficacious glioblastoma therapeutics. Our study (Fig. 5A) is the first demonstrating that ALDH1high glioblastoma cells are more tumorigenic in vivo than the ALDH1low counterparts, indicating that ALDH1 is a specific marker for MES GSCs. This is further supported by analyses of clinical glioma samples (Fig. 6B). To develop chemotherapeutics targeting MES GSCs, we synthesized a novel class of imidazo [1,2-a] pyridine derivatives called GA11 and GA23 as ALDH1 inhibitors. These compounds were designed on the basis of the conserved structural traits of the known natural occurring inhibitors including daidzin. In addition to fulfilling pharmacophore needs, these novel compounds possess a good hydrophilic–lipophilic balance, which allows for better penetration of the BBB. In principle, both these features make GA11, in particular, an attractive drug candidate to target GSCs in glioblastoma tumors. This structural prediction was validated in a cell-free kinase assay, which proved the ability of GA11 to inhibit the target enzyme ALDH in yeast (Fig. 6D) and further, an in vitro ALDEFLUOR assay showed the substantial decline of human ALDH1 activity in GA11-treated MES glioma spheres (Fig. 6E). Consistently, GA11 inhibited in vitro glioma sphere proliferation and in vivo xenograft growth in mouse brains (Fig. 6F–H, respectively). Further preclinical evaluation of GA11 and its analogs is currently underway for clinical development of anti-ALDH1 therapeutics for glioblastoma.

Another novel set of findings in this study is that ALDH1A3 is transcriptionally regulated by FOXD1 (Fig. 4E). The ALDH1A3 signaling in cancers is likely evolutionally conserved on the basis of the fact that ALDH1A3 silencing in vivo displayed substantial suppression of the genetically engineered Drosophila brain cancers (Fig. 5D and E). Interestingly, this FOXD1-mediated ALDH1A3 transcriptional activation was counteracted by another Forkhead family member, FOXG1, in the MES GSC–enriched cultures (Fig. 4C and D). These data are supported by the presence of a Forkhead TF consensus sequence in the ALDH1A3 promoter region. These data suggest that MES GSCs hijack the molecular mechanism for normal development (e.g., mouse retina) to promote their tumor growth (44). It is not clear, however, whether these two Forkhead TFs compete for the DNA binding in the ALDH1A3 promoter or whether FOXG1 indirectly influences FOXD1 signaling in MES GSCs. In addition, the restoration of ALDH1A3 by exogenous expression did not fully rescue the defects in MES glioma sphere growth caused by FOXD1 silencing (Fig. 4G), suggesting that additional undetermined oncogenic mechanisms are likely involved in MES glioblastoma. Further studies are needed to fully clarify the mechanisms by which FOXD1 maintains the MES GSCs phenotype and thus, glioblastoma tumorigenesis and therapy resistance.

In conclusion, this study describes the upregulation of ALDH1A3 and FOXD1 in clinical glioma samples and establishes their functional roles in the maintenance of MES GSCs, and therefore MES glioblastoma tumorigenesis. Furthermore, this study provided the first evidence supporting the fact that elevated FOXD1 expression is a negative prognostic factor in glioma, and establishes a role for FOXD1 directly regulating ALDH1A3 transcription in MES GSCs. Taken together, the FOXD1–ALDH1A3 axis is critical for tumor initiation in MES GSCs, therefore providing possible new molecular targets for the treatment of glioblastoma and other ALDH1-activated cancers.

No potential conflicts of interest were disclosed.

Conception and design: P. Cheng, V. Coviello, K.P.L. Bhat, C.L. Motta, M. Kango-Singh, I. Nakano

Development of methodology: P. Cheng, S. Sartini, C.L. Motta

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P. Cheng, I. Waghmare, A. Mohyeldin, M.S. Pavlyukov, M. Minata, C.L.L. Valentim, R.R. Chhipa, B. Dasgupta, C.L. Motta, M. Kango-Singh

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Cheng, J. Wang, I. Waghmare, S. Kim, A. Mohyeldin, B. Dasgupta, C.L. Motta, M. Kango-Singh

Writing, review, and/or revision of the manuscript: P. Cheng, I. Waghmare, S. Kim, A. Mohyeldin, Z. Zhang, C.L. Motta, M. Kango-Singh, I. Nakano

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): P. Cheng, J. Wang, S. Sartini, S. Kim, A. Mohyeldin, I. Nakano

Study supervision: I. Nakano

We thank all the members in the Nakano laboratory for constructive discussion for this study. We also thank Amber K. O'Connor for editing for this study.

This study is supported by grants P01CA163205, R01NS083767, R01NS087913, and R01CA183991 (I. Nakano). P. Cheng was supported by The First Hospital of China Medical University. I. Waghmare was supported in part by the Graduate Program at the University of Dayton. M. Kango-Singh was supported in part by Start-up funds from the University of Dayton.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Wen
PY
,
Kesari
S
. 
Malignant gliomas in adults
.
N Engl J Med
2008
;
359
:
492
507
.
2.
Stupp
R
,
Mason
WP
,
van den Bent
MJ
,
Weller
M
,
Fisher
B
,
Taphoorn
MJ
, et al
Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma
.
N Engl J Med
2005
;
352
:
987
96
.
3.
Venere
M
,
Fine
HA
,
Dirks
PB
,
Rich
JN
. 
Cancer stem cells in gliomas: identifying and understanding the apex cell in cancer's hierarchy
.
Glia
2011
;
59
:
1148
54
.
4.
Bhat
KP
,
Balasubramaniyan
V
,
Vaillant
B
,
Ezhilarasan
R
,
Hummelink
K
,
Hollingsworth
F
, et al
Mesenchymal differentiation mediated by NF-kappaB promotes radiation resistance in glioblastoma
.
Cancer Cell
2013
;
24
:
331
46
.
5.
Waghmare
I
,
Roebke
A
,
Minata
M
,
Kango-Singh
M
,
Nakano
I
. 
Intercellular cooperation and competition in brain cancers: lessons from Drosophila and human studies
.
Stem Cells Translational Med
2014
;
3
:
1262
8
.
6.
Mao
P
,
Joshi
K
,
Li
J
,
Kim
SH
,
Li
P
,
Santana-Santos
L
, et al
Mesenchymal glioma stem cells are maintained by activated glycolytic metabolism involving aldehyde dehydrogenase 1A3
.
Proc Natl Acad Sci U S A
2013
;
110
:
8644
9
.
7.
Hannenhalli
S
,
Kaestner
KH
. 
The evolution of Fox genes and their role in development and disease
.
Nat Rev Genet
2009
;
10
:
233
40
.
8.
Nakano
I
. 
Transcription factors as master regulator for cancer stemness: remove milk from fox?
Expert Rev Anticancer Ther
2014
;
14
:
873
5
.
9.
Pierrou
S
,
Hellqvist
M
,
Samuelsson
L
,
Enerback
S
,
Carlsson
P
. 
Cloning and characterization of seven human forkhead proteins: binding site specificity and DNA bending
.
EMBO J
1994
;
13
:
5002
12
.
10.
Hatini
V
,
Huh
SO
,
Herzlinger
D
,
Soares
VC
,
Lai
E
. 
Essential role of stromal mesenchyme in kidney morphogenesis revealed by targeted disruption of Winged Helix transcription factor BF-2
.
Genes Dev
1996
;
10
:
1467
78
.
11.
Thackray
VG
. 
Fox tales: regulation of gonadotropin gene expression by forkhead transcription factors
.
Mol Cell Endocrinol
2014
;
385
:
62
70
.
12.
Carreres
MI
,
Escalante
A
,
Murillo
B
,
Chauvin
G
,
Gaspar
P
,
Vegar
C
, et al
Transcription factor Foxd1 is required for the specification of the temporal retina in mammals
.
J Neurosci
2011
;
31
:
5673
81
.
13.
Fetting
JL
,
Guay
JA
,
Karolak
MJ
,
Iozzo
RV
,
Adams
DC
,
Maridas
DE
, et al
FOXD1 promotes nephron progenitor differentiation by repressing decorin in the embryonic kidney
.
Development
2014
;
141
:
17
27
.
14.
Hung
C
,
Linn
G
,
Chow
YH
,
Kobayashi
A
,
Mittelsteadt
K
,
Altemeier
WA
, et al
Role of lung pericytes and resident fibroblasts in the pathogenesis of pulmonary fibrosis
.
Am J Respir Crit Care Med
2013
;
188
:
820
30
.
15.
Gumbel
JH
,
Patterson
EM
,
Owusu
SA
,
Kabat
BE
,
Jung
DO
,
Simmons
J
, et al
The forkhead transcription factor, Foxd1, is necessary for pituitary luteinizing hormone expression in mice
.
PLoS one
2012
;
7
:
e52156
.
16.
Koga
M
,
Matsuda
M
,
Kawamura
T
,
Sogo
T
,
Shigeno
A
,
Nishida
E
, et al
Foxd1 is a mediator and indicator of the cell reprogramming process
.
Nat Commun
2014
;
5
:
3197
.
17.
Karlsson
J
,
Holmquist Mengelbier
L
,
Ciornei
CD
,
Naranjo
A
,
O'Sullivan
MJ
,
Gisselsson
D
. 
Clear cell sarcoma of the kidney demonstrates an embryonic signature indicative of a primitive nephrogenic origin
.
Gen Chromosomes Cancer
2014
;
53
:
381
91
.
18.
van der Heul-Nieuwenhuijsen
L
,
Dits
NF
,
Jenster
G
. 
Gene expression of forkhead transcription factors in the normal and diseased human prostate
.
BJU Int
2009
;
103
:
1574
80
.
19.
Zhao
YF
,
Zhao
JY
,
Yue
H
,
Hu
KS
,
Shen
H
,
Guo
ZG
, et al
FOXD1 promotes breast cancer proliferation and chemotherapeutic drug resistance by targeting p27
.
Biochem Biophys Res Commun
2015
;
456
:
232
7
.
20.
Kim
SH
,
Ezhilarasan
R
,
Phillips
E
,
Gallego-Perez
D
,
Sparks
A
,
Taylor
D
, et al
Serine/threonine kinase MLK4 determines mesenchymal identity in glioma stem cells in an NF-kappaB-dependent manner
.
Cancer Cell
2016
;
29
:
201
13
.
21.
Cheng
P
,
Phillips
E
,
Kim
SH
,
Taylor
D
,
Hielscher
T
,
Puccio
L
, et al
Kinome-wide shRNA screen identifies the receptor tyrosine kinase AXL as a key regulator for mesenchymal glioblastoma stem-like cells
.
Stem Cell Rep
2015
;
4
:
899
913
.
22.
Hu
Y
,
Smyth
GK
. 
ELDA: extreme limiting dilution analysis for comparing depleted and enriched populations in stem cell and other assays
.
J Immunol Methods
2009
;
347
:
70
8
.
23.
Remmele
W
,
Schicketanz
KH
. 
Immunohistochemical determination of estrogen and progesterone receptor content in human breast cancer. Computer-assisted image analysis (QIC score) vs. subjective grading (IRS)
.
Pathol Res Pract
1993
;
189
:
862
6
.
24.
Hutterer
M
,
Knyazev
P
,
Abate
A
,
Reschke
M
,
Maier
H
,
Stefanova
N
, et al
Axl and growth arrest-specific gene 6 are frequently overexpressed in human gliomas and predict poor prognosis in patients with glioblastoma multiforme
.
Clin Cancer Res
2008
;
14
:
130
8
.
25.
Verghese
S
,
Bedi
S
,
Kango-Singh
M
. 
Hippo signalling controls Dronc activity to regulate organ size in Drosophila
.
Cell Death Differ
2012
;
19
:
1664
76
.
26.
Ravasi
T
,
Suzuki
H
,
Cannistraci
CV
,
Katayama
S
,
Bajic
VB
,
Tan
K
, et al
An atlas of combinatorial transcriptional regulation in mouse and man
.
Cell
2010
;
140
:
744
52
.
27.
Nakano
I
,
Joshi
K
,
Visnyei
K
,
Hu
B
,
Watanabe
M
,
Lam
D
, et al
Siomycin A targets brain tumor stem cells partially through a MELK-mediated pathway
.
Neuro-oncol
2011
;
13
:
622
34
.
28.
Joshi
K
,
Banasavadi-Siddegowda
Y
,
Mo
X
,
Kim
SH
,
Mao
P
,
Kig
C
, et al
MELK-dependent FOXM1 phosphorylation is essential for proliferation of glioma stem cells
.
Stem Cells
2013
;
31
:
1051
63
.
29.
Minata
M
,
Gu
C
,
Joshi
K
,
Nakano-Okuno
M
,
Hong
C
,
Nguyen
CH
, et al
Multi-kinase inhibitor C1 triggers mitotic catastrophe of glioma stem cells mainly through MELK kinase inhibition
.
PLoS One
2014
;
9
:
e92546
.
30.
Lee
J
,
Kotliarova
S
,
Kotliarov
Y
,
Li
A
,
Su
Q
,
Donin
NM
, et al
Tumor stem cells derived from glioblastomas cultured in bFGF and EGF more closely mirror the phenotype and genotype of primary tumors than do serum-cultured cell lines
.
Cancer Cell
2006
;
9
:
391
403
.
31.
Bredel
M
,
Bredel
C
,
Juric
D
,
Harsh
GR
,
Vogel
H
,
Recht
LD
, et al
Functional network analysis reveals extended gliomagenesis pathway maps and three novel MYC-interacting genes in human gliomas
.
Cancer Res
2005
;
65
:
8679
89
.
32.
Sun
L
,
Hui
AM
,
Su
Q
,
Vortmeyer
A
,
Kotliarov
Y
,
Pastorino
S
, et al
Neuronal and glioma-derived stem cell factor induces angiogenesis within the brain
.
Cancer Cell
2006
;
9
:
287
300
.
33.
Nakano
I
,
Kornblum
HI
. 
Methods for analysis of brain tumor stem cell and neural stem cell self-renewal
.
Methods Mol Biol
2009
;
568
:
37
56
.
34.
Cheng
P
,
Phillips
E
,
Kim
SH
,
Taylor
D
,
Hielscher
T
,
Puccio
L
, et al
Kinome-wide shRNA screen identifies the receptor tyrosine kinase AXL as a key regulator for mesenchymal glioblastoma stem-like cells
.
Stem Cell Rep
2015
;
4
:
899
913
.
35.
Lee
HH
,
Frasch
M
. 
Survey of forkhead domain encoding genes in the Drosophila genome: Classification and embryonic expression patterns
.
Dev Dyn
2004
;
229
:
357
66
.
36.
Chakravorty
S
,
Wajda
MP
,
Vigoreaux
JO
. 
Courtship song analysis of Drosophila muscle mutants
.
Methods
2012
;
56
:
87
94
.
37.
Read
RD
,
Cavenee
WK
,
Furnari
FB
,
Thomas
JB
. 
A drosophila model for EGFR-Ras and PI3K-dependent human glioma
.
PLoS Genet
2009
;
5
:
e1000374
.
38.
Gao
GY
,
Li
DJ
,
Keung
WM
. 
Synthesis of potential antidipsotropic isoflavones: inhibitors of the mitochondrial monoamine oxidase-aldehyde dehydrogenase pathway
.
J Med Chem
2001
;
44
:
3320
8
.
39.
Del Turco
S
,
Sartini
S
,
Sentieri
C
,
Saponaro
C
,
Navarra
T
,
Dario
B
, et al
A novel 2,3-diphenyl-4H-pyrido[1,2-a]pyrimidin-4-one derivative inhibits endothelial cell dysfunction and smooth muscle cell proliferation/activation
.
Eur J Med Chem
2014
;
72
:
102
9
.
40.
Pajouhesh
H
,
Lenz
GR
. 
Medicinal chemical properties of successful central nervous system drugs
.
NeuroRx
2005
;
2
:
541
53
.
41.
Saigal
D
,
Cunningham
SJ
,
Farres
J
,
Weiner
H
. 
Molecular cloning of the mitochondrial aldehyde dehydrogenase gene of Saccharomycesce revisiae by genetic complementation
.
J Bacteriol
1991
;
173
:
3199
208
.
42.
Pan
Q
,
Li
Q
,
Liu
S
,
Ning
N
,
Zhang
X
,
Xu
Y
, et al
Concise review: targeting cancer stem cells using immunologic approaches
.
Stem Cells
2015
;
33
:
2085
92
.
43.
Lathia
JD
,
Mack
SC
,
Mulkearns-Hubert
EE
,
Valentim
CL
,
Rich
JN
. 
Cancer stem cells in glioblastoma
.
Genes Dev
2015
;
29
:
1203
17
.
44.
Tian
NM
,
Pratt
T
,
Price
DJ
. 
Foxg1 regulates retinal axon pathfinding by repressing an ipsilateral program in nasal retina and by causing optic chiasm cells to exert a net axonal growth-promoting activity
.
Development
2008
;
135
:
4081
9
.