Epigenetic regulation of CD133/PROM1 expression in glioma stem cells by Sp1/myc and promoter methylation

Abstract

Tumor stem cells, postulated to be the source cells for malignancies, have been identified in several cancers using cell-surface expression of markers including CD133, a pentaspan membrane protein. CD133+ve cells form neurospheres, exhibit self-renewal and differentiation, and are tumorigenic. However, despite its association with stem cells, a causal relationship of CD133 to tumorigenesis remains to be defined. Hypothesizing that specific epigenetic and transcription factors implicated in driving the stem cell state may concurrently regulate CD133 expression in stem cells, we analyzed the structure and regulation of CD133 promoter in glioma stem cells and glioma cell lines. Initially, a minimal promoter region was identified by analyzing the activity of CD133 promoter-driven luciferase-expressing 5’-and 3’-deletion-constructs upstream of the transcription start site. This region contained a CpG island that was hypermethylated in CD133−ve glioma stem cells (GSC) and glioma cells but unmethylated in CD133+ve ones. Of several predicted TF-binding sites in this region, the role of tandem Sp1 (−242 and −221) and two Myc (−541 and −25)-binding sites were examined. Overexpression of Sp1 or Myc increased CD133 minimal promoter-driven luciferase activity and CD133 levels in GSC and in glioma cell line. Mithramycin, a Sp1 inhibitor, decreased minimal promoter activity and downregulated CD133 levels in GSC. Gel-shift assays demonstrated direct binding of Sp1 to their predicted sites that was competitively inhibited by oligonucleotide-binding-site sequences and supershifted by anti-Sp1 confirming the interaction. Sp1 and Myc-antibody chromatin immunoprecipitation (ChIP) analysis in GSC showed enrichment of regions with Sp1 and Myc-binding sites. In CD133−ve cells, ChIP analysis showed binding of the methyl-DNA-binding proteins, MBD1, MBD2 and MeCP2 to the methylated CpG island and repression of transcription. These results demonstrate that Sp1 and Myc regulate CD133 transcription in GSC and that promoter methylation and methyl-DNA-binding proteins cause repression of CD133 by excluding transcription-factor binding.

INTRODUCTION

The cancer stem cell hypothesis postulates the existence of a subset of tumor cells that exhibit properties similar to normal stem cells including self-renewal and differentiation, and generate progeny that form the bulk of the tumor.¹ Isolation of such cells has depended on specific markers several of which also identify normal stem cells. Among the earliest of such markers identified is CD133, which was used to isolate primitive stem and progenitor cells in hematopoietic and non-hematopoietic tissues.² A structural homolog of murine prominin-1 that localizes to the microvilli on the surface of neuroepithelial stem cells,³ it was originally identified as a target for the monoclonal antibody AC133 in CD34+ve hematopoietic cells and was determined to be a glycosylated protein with an apparent molecular weight of 120 kD.⁴ CD133+ve stem and progenitor cells are found in the liver, muscle, kidney, prostate and neural tissue and have been shown to undergo differentiation to neural cells, hepatocytes, myocytes and osteoblasts.^5–7

CD133 has been used as a marker to isolate cancer stem cells in diverse tumors including glioblastoma, pediatric medulloblastoma, ependymoma, colorectal and pancreatic carcinomas.⁸ When isolated from high-grade gliomas, the CD133+ve subset of cells exhibit characteristics of neural stem cells including the ability to generate neurospheres, exhibit self-renewal, undergo differentiation and proliferation; in addition, they also exhibit tumorigenicity.⁹ Although CD133+ve cells seem to select for glioma stem cells (GSC), a subset of CD133−ve cells can also be tumorigenic, suggesting that it is an imperfect marker for GSC.¹⁰ Whether this is due to poor selection of CD133−ve cells based on variable levels of surface expression of the glycosylated protein or expression of different isoforms of the protein, or because both CD133+ve and −ve cells may cooperate in a cellular hierarchy,¹¹ remains unclear. Additionally, CD133 expression in gliomas correlates with poor progression-free and overall survival.¹² CD133+ve cells are also significantly resistant to chemotherapeutic agents compared with autologous CD133−ve cells¹³ and to radiation therapy.¹⁴

As yet, a functional role for CD133 in regulating stem cell biology remains to be established; hence, its association with stem cells suggests that it may be a marker that is co-regulated with other more relevant factors that help determine stem cell characteristics. Promoter methylation is one such mechanism that is associated with downregulation of CD133 in glioblastoma and colon cancer cells.^15,16 However, the regulation of CD133 expression in cancer stem cells is poorly characterized. We hypothesized that understanding the epigenetic regulators of CD133 expression may yield novel insights into regulation of stem-like characteristics of GSC. The regulatory regions of CD133 are complex and control tissue-specific expression of several alternatively spliced 5’-UTR isoforms expressed via at least 5 alternative promoters.¹⁷ A role for promoter methylation in regulating CD133 expression was indicated by the presence of a CpG island encompassing the first exons 1A, 1B and 1C. Relevant to the current study, normal brain tissue was noted to express only the exon 1B containing transcript that could potentially be regulated by promoter methylation.

In the current report, we examined transcriptional regulation of CD133 in GSC that expressed the marker (CD133+ve) versus those that did not (CD133−ve), and also compared these findings with conventional glioma cell lines (CD133−ve). We determined the relevance of promoter methylation and the role of specific transcription factors in regulating CD133 expression. Our results also indicate a new role for Sp1 and Myc in activating CD133 transcription in GSC and show that Myc requires functional Sp1 binding sites to activate CD133 expression. Further, we show that promoter methylation represses CD133 expression in the CD133−ve subpopulation isolated from GSC and that methyl DNA-binding proteins MBD1, MBD2 and MeCP2 bind to the methylated CpG island in the CD133−ve glioma cell lines indicating a role for these proteins in the DNA methylation-induced repression of CD133. Regulation of CD133 may provide insights into the co-regulation of other stem cells markers that are more directly relevant to the stem cell state of GSC and other tumor stem cells.

RESULTS

Characterization of CD133 expression in glioma cells and glioma stem cells

We examined the expression of CD133 at the transcript level in several glioma cell lines and GSC. In addition, we also examined the cell surface expression of the protein in these cell lines using flowcytometric analysis of nonpermeabilized cells. CD133 transcript was seen to be expressed in the GSC but not in glioma cell lines. In addition, cell surface expression of the protein, as determined by flowcytometric analysis of nonpermeabilized cells, was also seen only in and not in glioma cell lines (Figure 1a).

Figure 1.

(a) RT–PCR analysis for CD133 expression in GSC and glioblastoma tumor cell lines (left panel). Flowcytometric analysis of cell surface expression in CD133 in nonpermeabilized GSC and glioma tumor cell lines (right panel) (b) RT–PCR analysis of CD133 transcript variants in GSC11, and GSC23 cells. (c) Relative luciferase activity of CD133 promoter constructs in GSC23 and U251HF cells. 2×10⁴ cells were plated in 12-well plates 24 h before transfection following which the cells were transfected with 0.5 µg of promoter plasmid along with 50 ng of Beta actin-Renilla luciferase plasmid as control. The histograms represent the mean of the values of relative luciferase activity from triplicate transfections and the error bars denote s.d. observed in the given set of experiments.

CD133 promoter-CpG island is involved in regulation of the transcriptional activity of the promoter

In vitro methylation of CpG island in the CD133 promoter represses transcription with loss of protein expression indicating a direct regulatory role for promoter methylation in CD133 expression. To understand the role of this CpG island in transcriptional regulation of CD133 in GSC, we examined the activity of the promoter in GSC (CD133+ve) and in glioma cell lines (CD133−ve). Expression analysis for transcript variants of CD133 using two GSC showed the presence of all three variants in GSC11 but only variants 1B and 1C in GSC23 (Figure 1b); subsequent experiments were conducted using GSC23 (sequences numbered with reference to start site of transcript variant 1C). The CpG island (as determined using EMBOSS CpGPlot with reference sequence GenBank: AY275524) is located between nucleotides −630 and −145 (length 485 bp) upstream of the transcription start site of variant 1C. Deletion constructs of the 5’-sequences upstream of this start site were cloned into PGl4.6 vector and analyzed for their relative luciferase activity relative to the full-length construct P0. The constructs showed promoter activity similar to that of the full-length construct (P3) or a slight decrease (P1 and P2) or increase (in P4). In the P5 construct, luciferase activity levels reached background levels in the cell lines tested. Subsequent analysis of luciferase activity of 3′-deletion constructs revealed decreased promoter activity in constructs P6 and P7. The intervening region between P7 and P5 constructs, which encompasses the CpG island, constitutes the P8 construct that had 12.7- and 5.6-fold transcriptional activity in GSC23 and U251HF cells, respectively, compared with the empty PGl4.6 vector control (Figure 1c).

Transcription factor Sp1 activates CD133 transcription in glioma stem cells

The robust promoter activity displayed by the CpG island sequence suggests a role for this region in regulating CD133 expression. We analyzed this region for transcription factor-binding sites that could regulate CD133 transcription, using the software program, Mat-Inspector, and identified several putative transcription factor-binding sites. Given their relevance to stem cells, we chose to study two tandem GC boxes (Sp1 sites) located (−246 to −232) and (−211 to −225), and two canonical E-box site (Myc sites) located within the CpG island (−545 to −533) and further upstream (−28 to −16) (Figure 2a). Ectopic overexpression of Sp1 co-transfected with promoter construct P8 in U251HF and GSC23 cells resulted in a twofold increase in promoter activity compared with pCMV-gutless control (Figure 2b). To further study the role of Sp1, GSC23 cells were treated with mithramycin, a Sp1 inhibitor, at various concentrations added to the stem cell medium. Immunoblot analysis showed marked repression of CD133 levels (Figure 2c); additionally, mithramycin treatment resulted in downregulation of CD133 promoter activity in GSC23, SNB19 and U251HF cells, (Figure 2d) indicating a role for Sp1 in transcriptional activation of CD133. To confirm binding of Sp1 to the predicted sites, electrophoretic mobility shift assay (EMSA) analysis was performed using the biotin-labeled oligonucleotides corresponding to the sequence of the putative Sp1-binding sites. Incubation in presence of nuclear extracts prepared from GSC23 cells showed a shift in the mobility of the oligonucleotides with a corresponding increase in the complex formation, as the amount of nuclear extract was increased (Figure 3a, lanes 1 and 2). Specificity of Sp1 binding to the oligonucleotide was determined by incubation with unlabeled oligonucleotides comprising of wild-type consensus Sp1-binding sites. The wild-type oligonucleotides were able to compete and thereby diminish DNA-protein complex indicating that Sp1 is involved in the complex formation (Figure 3a, lanes 4–6). Subsequent incubation of the binding reaction in the presence of antirabbit IgG antibody or Sp1 antibody showed a complex formation in the presence of IgG whereas there was absence of complex formation in the presence of anti-SP1 antibody confirming the requirement of Sp1 for complex formation (Figure 3a, lanes 8–10). In-vivo binding of Sp1 protein to the promoter was determined using ChIP analysis of CD133 promoter. Primer sets ChP2 (−579/−340) and ChP3 (−339/−100) were designed to amplify regions encompassing the promoter region containing the putative Sp1 binding sites including a single Sp1 site at (−332/−324) and the tandem Sp1 sites (Figure 3b). Chromatin immunoprecipitation with anti-Sp1 antibody showed an enrichment of CD133 promoter compared with the IgG control (Figure 3c) indicating the occupation of the CD133 promoter CpG island by Sp1. These results show that transcription factor Sp1 binds to the predicted Sp1-binding sites in the CD133 promoter and transcriptionally activates CD133 expression.

Figure 2.

(a) Location of predicted Sp1 and Myc sites in the C133 promoter CpG island. The sequences depicted in bold and underlined denote the predicted binding sites for Sp1 and Myc and the CpG island DNA sequence represented by (bracketed) portion of the sequence. The numbering of the bases is with reference to transcription start of variant exon 1C represented by arrow head, (b) Overexpression of Sp1 increases the transcription driven by the CD133 promoter CpG island in GSC23 cells and U251HF cells. The cells were co-transfected with Sp1 expressing pCMV-Sp1 plasmid (0.5 µg), the promoter construct P8 (0.5 µg) and a beta-actin Renilla luciferase construct (50 ng). Correspondingly, as a control, the cells were transfected with empty pCMV vector (0.5 µg) and the P8 construct in a similar manner to determine the native level of promoter activity. The cells were analyzed for luciferase activity 48 h after transfection. The histograms represent mean of fold change in of relative luciferase activity from triplicate transfections and the error bars represent s.d. of mean. (c) Immunoblot analysis of CD133 expression in GSC23 cells untreated or treated with Mithramycin 50 nm, 100 nm and 150 nm (upper panel). Quantification of the CD133 expression using Image J software (NIH, Bethesda, MD, USA) (lower panel). The histograms represent the fold in expression of CD133 in samples relative to vinculin expression. (d) Effect of Mithramycin treatment on CD133 promoter CpG island-driven luciferase activity in U251HF and SNB19 cells. Cells were transfected with P8 construct along with control beta actin-Renilla luciferase 24 h. After transfection, the transfection medium was removed and replaced with medium supplemented with 0 nm, 50nm, 100 nm, and 200 nm Mithramycin, and the cells were further incubated for further 24 h and subsequently analyzed for luciferase activity. The histograms represent the mean of luciferase activity from three independent transfections and the error bars represent s.d. of mean.

Figure 3.

(a) Sp1 interacts with the predicted Sp1-binding sites in the CD133 promoter CpG island. EMSA analysis was performed using nuclear extracts isolated form GSC23 cells in the presence of 5′-biotinylated oligonucleotides representing the predicted Sp1 site as described in the Materials and methods section. (b) Location of the ChIP primers in the CD133 promoter CpG island. (c) Sp1 chromatin immunoprecipitates contain CD133 promoter DNA corresponding to the Sp1-binding site including the single upstream and tandem downstream sites. PCR reactions using the primer sets indicated by their relative positions in the CD133 promoter along with amplimers from the inputs of the ChIP reactions (IgG).

Myc is involved in the transcriptional regulation of CD133 gene in glioma stem cells

The presence of canonical E-box sites that bind to factors such as Myc and Max by basic helix-loop-helix leucine zipper domain (bHLH-Zip) indicated a putative role for Myc in transcriptional regulation of CD133. Ectopic overexpression of Myc co-transfected with promoter construct P8 showed a twofold increase in transcriptional activity of the luciferase gene compared with empty vector control in both GSC23 cells and U251 cells (Figure 4a). Corresponding analysis of the GSC23 cell lysates showed increased protein levels indicating increase in the promoter activity (Figure 4b). ChIP assays performed to confirm the in-vivo binding of Myc protein to the predicted E-box sites showed amplimers for primer pairs ChP2 (−579/−340) and ChP4 (−99/+21) that span E-box sites whereas primer pair ChP3 (−339/−100) spanning a region not comprising E-box site failed to show enrichment for the CD133 promoter indicating that Myc binds to the predicted E-box in the CD133 promoter (Figure 4c). Supporting the specificity of Myc-induced regulation of CD133 promoter activity, overexpression of the transcription factors, AP2 and ZF5 did not result in increased CD133 promoter-driven luciferase activity (Figure 4d).

Figure 4.

Myc activates transcription of CD133. (a) Effect of over expression of Myc on CD133 promoter CpG island-driven luciferase gene in GSC23 cells and U251HF cells. GSC23 cells were co-transfected with pcDNA-Myc or empty pcDNA along with 0.5 µg of P8 plasmid construct and 50 ng of beta-actin promoter-driven Renilla luciferase (transfection control). Analysis for luciferase activity was performed as described in Materials and methods section. The histograms represent the fold change in relative luciferase activity after normalization with pcDNA experiment values. The error bars represent s.d. of mean of values. (b) Immunoblot analysis of Myc overexpression on the protein levels of CD133 in GSC23 cells. (c) Cartoon depicting the position of the ChIP primers used to amplify the CD133 promoter region spanning the Myc-binding sites (upper panel) and PCR amplification of the ChIP DNA derived from ChIP reaction performed using Myc and IgG antibodies (lower panel). (d) Effect of overexpression of transcription factors Myc, AP2 and ZF5 on a CD133-P8 promoter construct-mediated luciferase activity compared with vector control.

Myc requires functional Sp1 binding sites for transcriptional regulation of CD133 promoter

To determine whether the two identified Sp1 sites have a role in the transcriptional activation, substitution mutations of the individual sites (Sp1MT1, Sp1MT2) and in combination (Sp1MT1-2) were made (Figure 5a). Abrogation of individual sites repressed transcriptional activity by 60 and 90% in the double Sp1 site-mutated promoter when compared with the wild-type promoter in both GSC23 and U251HF cells (Figure 5b). To ascertain the role of upstream E-box site located in the CpG island on the transcriptional activity of the promoter, the E-box was mutated by substituting the conserved bases at the site (Figure 5a). Subsequent analysis for luciferase activity showed repression of the Myc-mutant CD133 promoter compared with the wild type promoter in both GSC23 cells and U251HF cells (Figure 5b). Using the mutant constructs, we further studied the interplay between the Myc and Sp1-binding sites in the CD133 promoter. When Myc was overexpressed in the presence of mutated Sp1 binding site (SP1 MT1-2), luciferase activity of the promoter revealed no activation of the mutant Sp1 promoter compared with empty pcDNA vector control (Figure 5c). This result suggested that downstream functional Sp1-binding sites are required for transcriptional activation by Myc. On the contrary, Sp1 overexpression in the presence of mutated Myc-binding site increased the promoter activity and the extent of activation was similar to that of the wild-type promoter relative to the empty pCMV control vector (Figure 5d), indicating that mutation of the Myc-binding site did not affect transcriptional activation by Sp1.

Figure 5.

Myc requires functional Sp1-binding sites to activate transcription of CD133 promoter CpG island-driven luciferase activity. (a) Cartoon representing the various mutant constructs used for the study. (b) The effect of mutation of Sp1 site one mutant (Sp1 MT1), Sp1 site two mutant (Sp1 MT2), double Sp1 site mutant (Sp1 MT1-2), and E-box site mutant (Myc-MT) on the native activity of the promoter in comparison to wild-type promoter in GSC23 and U251HF cells. A weight of 0.5 µg of the mutant and wild-type constructs were used for analysis along with 50 ng of beta-actin promoter Renilla luciferase (transfection control). The histograms represent the fold change in the relative luciferase activity of the constructs compared with the relative luciferase activity of the wild-type construct. (c) The effect of Myc overexpression on the Sp1 MT1-2 promoter construct: 0.5 µg mutant construct Sp1 MT1-2 was co-transfected in presence of pcDNA or pcDNA-Myc and 50 ng of beta-actin Renilla luciferase construct. Assay for luciferase activity was performed after 48 h after transfection as previously described. The histograms represent the fold change in the relative luciferase activity between pcDNA and pcDNA-Myc experimental values. (d) Effect of Sp1 overexpression on wild-type and Myc-MT promoter. pCMV-Sp1 was co-transfected using either the wild-type promoter or Myc-MT and to determine the native activity of the wild-type promoter WT construct was co-transfected with empty pCMV alone. All transfections were performed along with 50 ng of beta actin Renilla luciferase (transfection control). The histograms represent the fold change in relative luciferase activity of WT, and Sp1 MT1-2 in comparison to pCMV WT values. The error bars in b–d represent s.d. in the mean of values obtained from the three independent transfections.

Downregulation of CD133 in CD133−ve glioblastoma cells involves DNA methylation

Using DNA methylation analysis by bisulfite sequencing PCR (BSP), the methylation status of 30 individual cytosine residues in the CpG island was analyzed. Comparison of the profiles showed dense methylation of the CpG island in all the clones of CD133−ve tumor cell lines (U251HF, SNB19, and D54) analyzed, whereas CD133+ve (GSC2, GSC23 and GSC11) cells showed patchy or no methylation in various clones (Figure 6a). Interestingly, we found that CpG dinucleotides situated in the 2 tandem GC boxes that were determined to bind Sp1 were methylated in most clones analyzed in the CD133−ve cell lines strongly suggesting that DNA methylation in the GC box could hamper Sp1 binding. Flowcytometric analysis of GSC23 cells for CD133 expression showed that only 70–80% cells expressed CD133 that indicated the presence of a subpopulation of CD133−ve cells. Using fluorescence-activated cell sorting analysis, CD133+ve cells were sorted from the CD133−ve fraction in GSC cells (Figure 6b). The validity of this sorting was confirmed by RT–PCR analysis for CD133 messenger RNA; CD133−ve fraction showed low level of expression compared with CD133+ve fraction (Figure 6c). Genomic DNA isolated simultaneously from these cells was subject to BSP analysis. Clones derived from the CD133−ve fraction of GSC2, GSC11 and GSC23 had levels of methylation comparable to that of non-expressing glioma tumor cells, whereas CD133+ve fraction showed minimal levels of methylation (Figure 6c). The presence of dense methylation in the CD133−ve fraction of GSC suggests that DNA methylation could govern spontaneous conversion from CD133+ve to CD133−ve phenotype. Treatment of CD133−ve glioma cells (D54, SNB19 and U251HF cell) with the demethylating agent, 5-aza-2′-deoxycytidine, resulted in re-expression of CD133 in a dose-dependent manner (Figure 6d upper panels) confirming that the promoter was regulated by methylation; conversely, in vitro methylation of CD133 promoter by SS1 methylase represses the transcription of reporter luciferase gene in GSC23 (Figure 6d lower panel). To further understand the role of DNA methylation-induced repression of CD133, we used ChIP assays to study the binding of methyl DNA-binding proteins to the methylated promoter using primers designed to span the entire CpG island. ChIP reactions were performed using anti- MBD1, MBD2, MeCP2 antibodies with antimouse IgG to determine the background precipitation levels. The ChIP experiments indicated that MBD1, MBD2 and MeCP2 bind to the methylated promoter in the chromatin extracts prepared from the U251HF and SNB19 cells relative to IgG control whereas GSC11 and GSC23 cells showed relatively less notable binding to these regions compared with IgG controls except for MBD1 (Figure 7). Interestingly, the methyl-binding protein occupancy in U251HF and SNB19 was most notable in the −339 to −100 region of the promoter that contains with the Sp1-binding sites previously observed in promoter of CD133+ve cells. Our results indicate that the CpG island is densely methylated in CD133−ve cells, and in addition, the methylated region is occupied by methyl-binding proteins, further confirming the role of DNA methylation in repression of CD133 gene and its interaction with Sp1 binding.

Figure 6.

DNA methylation in CD133 promoter CpG island. (a) BSP analysis for DNA methylation of cytosine residues in CD133 promoter CpG island in glioblastoma tumor cell lines U251HF, D54, SNB19 and glioma stem cells, GSC2, GSC11 and GSC23. The solid circle represents methylated cytosine residues, and empty circles represent unmethylated cytosine residues. (b) Results of the flowcytometric sorting used to separate the CD133+ve population from the CD133−ve cells in GSC23 cells (upper panel) and extreme selection of CD133+ve and−ve cells (lower panel), using IgG as negative control. (c) BSP analysis for cytosine methylation in the CD133 promoter CpG island in extreme sorted CD133+ve and CD133−ve fractions of GSC23, GSC2 and GSC11 cells. RT–PCR analysis of CD133 levels in the sorted CD133+ve and CD133−ve GSC subpopulations are shown on the right side panels. (d) Analysis of CD133 mRNA re-expression by the effect of demethylating agent, 5′deoxy-azacytidine, in CD133-nonexpressing D54, SNB19 and U251HF glioma cell lines (upper panel); CD133-expressing GSC23 glioma stem cell line is shown as control and GAPDH represents the loading control. Results of in vitro methylation of the CD133 promoter using a SS1methylase overexpression plasmid or control vector (lower panel).

Figure 7.

ChIP analysis for the binding of methyl DNA-binding proteins to CD133 CpG island promoter. (a) Cartoon depicts the position of the ChIP primers used to amplify the CD133 promoter. (b) Ethidium bromide stained agarose gel with the amplification products of the ChIP reaction. Panels from the uppermost to the lower one show the results of the amplification of the using primers ChP1, ChP2 and ChP3, respectively, for the cell lines indicated.

DISCUSSION

Malignant gliomas were among the first of tumors in which tumor stem cells, the subpopulation of cells with enhanced tumorigenic capacity were described.⁹ Strategies to deplete these cells, which have been shown to be resistant to conventional treatments, are expected to dramatically improve clinical outcome. However, a better understanding of the regulation of GSC is a crucial prerequisite developing strategies to target these cells.

CD133 is a pentaspan transmembrane protein that in its glycosylated form was originally described as a marker for stem and progenitor cells in hematopoietic system.⁴ It was subsequently used to identify putative tumor stem cells from various malignancies including glioblastomas,⁹ breast cancer,^18,19 and colon cancer.²⁰ However, GSC identified using antibodies against CD133 have inconsistent tumorigenic potential, suggesting that it is not a perfect stem cell marker.²¹ Despite these limitations, CD133 has been extensively utilized to enrich for tumor stem-like cells from human malignancies. Although a functional role for CD133 in governing tumor stem cell biology remains to be defined,²² its association with the GSC led to our hypothesizing that CD133 regulation in GSC could shed light on other co-regulated factors that have a more definitive role in controlling the stem-cell state.

Epigenetic factors can provide a flexible means of adaptation for GSC and have been reported to regulate stem cell characteristics. Aberrant promoter methylation of CD133 gene has been reported in glioblastomas and colorectal cancer whereas CD133 promoter in the normal tissue is essentially unmethylated.¹⁶ In the present study, we demonstrate a role for the promoter CpG island in critically regulating CD133 expression in GSC. In addition, we show that established glioma cells lines lose CD133 expression in association with promoter methylation, which is reversible by demethylating agents demonstrating a direct role for epigenetic silencing of the CD133 in these cells. In contrast, GSC expressed CD133 and did not exhibit CD133 promoter methylation. The methylation seen in the established cell lines does not appear to be an artifact of serial culture given that the GSC used in this study were also serially passaged for several years. In addition, when the GSC cells were sorted into CD133+ve and−ve fractions, promoter methylation was seen in CD133−ve fraction of GSC; this was reversible by methylation agents and reproduced by in vitro methylation strongly supporting a functional role for methylation in the repression of CD133. Whether this finding is relevant to the tumorigenicity of the CD133-based fractions remains to be established and is currently under investigation.

Analysis for predicted transcription factor-binding sites revealed several putative sites within the various promoter regions; of particular interest, Sp1- and Myc-binding sites were predicted in the promoter of exon 1B-containing transcript that is predominantly expressed in brain. Sp1 was identified as a DNA-binding factor that regulated transcription by binding to specific sequences on promoters of its target genes and has been postulated to have an important role in stem cell regulation.^23–27 The ubiquitous expression of Sp1 and the presence of putative binding sites in the regulatory regions of several genes have been interpreted as suggesting a non-specific role for this transcription factor in regulating a large number of genes. However, in Sp1−/− mice, only a limited number of genes were found to be expressed at a lower level challenging the notion of Sp1 as a promiscuously interacting protein and indicating a more specific role for Sp1-mediated transcriptional regulation; supporting this concept, subsequent studies have demonstrated that Sp1 may function in a cell type and gene-specific role than previously postulated.^28–30 Myc functions as a proto-oncogene and a weak transcription factor that cooperates with other factors to mediate gene regulation.^31–33 It has also been reported to have a regulatory role in glioma stem cells; GSC that have been subjected to short hairpin RNA-mediated knockdown of Myc expression failed to form neurospheres and lost tumorigenicity.³⁴ In addition, the inactivation of PTEN and p53 has been reported to promote Myc activation that in turn enhances the stem cell state of GSC and prevents differentiation.³⁵ These data suggest a role for Sp1 and Myc in regulating GSC that we examined in the context of CD133 expression.

The presence of a CpG island in the CD133 promoter and its methylation has been previously described;^15,17 however, the role of regulatory factors that control gene expression and their relationship with the promoter have not been characterized. The interaction between transcription factors and CpG island methylation is known to be critical in regulating gene expression. DNA methylation can prevent binding of transcription factors to their cognate-binding sites, repressing transcription.^36,37 For instance, methylation-mediated interference of the binding of Sp1 to its binding sites on promoters is known to inhibit expression of several of its target genes.^38–40 The presence of tandem Sp1 sites within the CpG island region of the CD133 promoter^41,42 and the known role of Sp1 in regulating specific stemness genes^26,27 led us to examine the role of this transcription factor in regulating CD133 expression. Our results demonstrate the direct binding of Sp1 to its predicted binding sites in the CD133 promoter and consequent activation of transcription. To our knowledge, this is the first report to demonstrate an interaction between Sp1 and the CD133 promoter and its role in regulating CD133 promoter activity.

The identification of an E-box that forms a putative Myc-binding site in the CpG island of CD133 raised the possibility of a role for Myc in regulating CD133 expression. As a potent proto-oncogene, deregulated expression of Myc is known to drive proliferation and tumorigenesis.³¹ However, Myc is a relatively weak activator of transcription and functions in partnership with several other factors to influence transcription.³² Cooperation between the Sp1 and Myc sites has been reported in the context of regulation of several genes.⁴³ Dual Sp1 and Myc sites have also been reported as being enriched in the promoter regions of evolutionarily conserved genes suggesting a functional relevance to the cooperation between these transcription factors.³³ In gliomas cells, we identified that Myc overexpression results in CD133 promoter activation in glioma cells, but this was abrogated when the tandem Sp1-binding sites on the promoter were mutated, suggesting that Myc requires functional Sp1 sites to regulate CD133 expression. Thus, in the context of CD133 expression, Sp1 functioned independently to regulate the promoter whereas Myc mediated regulation was dependent on Sp1 function.

In addition to interference with the binding of transcription factors, promoter methylation can indirectly repress transcription by recruitment of methyl-binding proteins.^44–46 MBD1−/− adult neural stem cells showed defects in differentiation; in contrast, embryonal neurogenesis is not affected by the loss of MBD1. On the other hand, MeCP2, which is extensively expressed in the brain, is known to function in maturation but not differentiation. Whether MECP2 or MBD1 and 2 are associated with methylation in tumor stem cells has not been studied. In the context of GSC, we examined the role of several methyl-binding proteins including MeCP2, MBD1 and MBD2 in regulating CD133 expression, given their known role in transcriptional repression associated with methylation.^47–49 ChIP assays demonstrated that MBD1 and MBD2 bind to the methylated promoter in CD133−ve U251HF glioma tumor cells whereas MECP2 was not seen to occupy the promoter. This is in keeping with the known role for MBD1 and 2 in neural differentiation and the lack of such a function for MECP2. Additional studies are being conducted to fully characterize the effect of these methyl-binding proteins in regulation of CD133 promoter in tumor cells.

The results of this study provide new insights into the regulation of CD133 expression through promoter methylation and the recruitment of specific transcription factors. We have identified Sp1 as a critical regulator of CD133 promoter activity and demonstrate that Myc cooperates with Sp1 in controlling CD133 expression. Previous reports of a strong role for Myc in regulating stem cells,^34,35 when taken in the context of our findings of interaction between Sp1 and Myc, also raise the possibility of a similar interaction between these transcription factors in regulating stemness of GSC that are being further explored. Of particular note, our results demonstrate an inverse relationship between promoter methylation mediation repression and Sp1 or Myc-mediated activation of the CD133 promoter; this strongly suggests that DNA methylation can restrict access of Sp1 and Myc to the CD133 promoter and that these two processes may reciprocally control CD133 expression. Given that the role of CD133 in controlling stemness of GSC is uncertain, it is likely regulated along with a set of co-expressed genes that may share epigenetic regulatory mechanisms. Factors such as Sp1 and Myc could hence be used as ‘baits’ to identify the interaction with promoters of true stem cell regulatory genes that are co-regulated with CD133 and could lead to identification of novel genes that control stemness in GSC. In addition, differential methylation profile of the gene promoters in the CD133+ve and CD133−ve cell subpopulations could reveal new insights into the role of DNA methylation as an epigenetic mechanism for transition of CD133+ve to CD133−ve phenotype. We are currently conducting studies to identify novel genes that are epigenetically co-regulated with CD133 in GSC based on the findings of this study. Such genes could in turn be highly relevant as potential targets for therapeutic strategies against GSC.

MATERIALS AND METHODS

Cell lines and culture

GSC23, GSC11 and GSC2 are established cell lines derived from high-grade gliomas, described previously,⁵⁰ which are cultured as neuro-spheres in a stem cell-permissive medium comprising of Dulbecco’s modified Eagle’s medium (DMEM)–F12 supplemented with B27 (1:50; Life Technologies, Carlsbad, CA, USA), 20 ng/ml each of basic fibroblast growth factor and epidermal growth factor (EGF). Glioma tumor cell lines SNB19 and U251 were commercially obtained from ATCC, and D54 was a kind gift of Dr Darrel Bigner, Duke University. These cells were cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% FBS and Penicillin, Streptomycin in 100 mm dishes in a humidified atmosphere containing 5% CO₂ at 37 °C.

Plasmids, transient transfection, and site-directed mutagenesis

Cells were centrifuged at 194 g for 5 min and genomic DNA was isolated using Qiagen DNA isolation kit (Qiagen, Valencia, CA, USA). To generate various 5′-deletion constructs of CD133 promoter upstream of transcription start site, sequences were amplified from genomic DNA isolated from GSC23 cells using appropriate primers as listed in Table 1 (details in Supplementary data). For transient transfection experiments, Sp1 and Myc expression plasmids (pCMVSp1 and pcDNA-Myc) were obtained from a non-profit entity (Addgene Inc., Cambridge, MA, USA) and experiments conducted as outlined in Supplementary data. A site-directed mutagenesis kit (Stratagene, Santa Clara, CA, USA) was used as per the manufacturer’s protocol to generate Sp1- and Myc-binding site mutant constructs; the oligonucleotides used for the mutagenesis reactions are listed in Table 2. The CD133 promoter CpG island was determined using EMBOSS CpGPlot by analyzing genomic nucleotide sequences listed with accession number in GenBank: AY275524.

Table 1.

Primers used for amplification of CD133 promoter region

Construct	Forward primer	Reverse primer
P0 (−1649/+186)	5′-TGAGGTACCTGAGAACAGCAAGGGAGAAATC-3′	5′-CTTGAAGCTTCCCAGTGGATGGAAAGAAGA-3′
P1 (−1345/+186)	5′-TGAGGTACCACTTAATGTTCCTGTCTATCA-3′	5′-CTTGAAGCTTCCCAGTGGATGGAAAGAAGA-3′
P2 (−1054/+186)	5′-TGAGGTACCGTTGAGTTGTAACTCACAGA-3′	5′-CTTGAAGCTTCCCAGTGGATGGAAAGAAGA-3′
P3 (−753/+186)	5′-TGAGGTACCCATTCTAAGTAAGGGACTCT-3′	5′-CTTGAAGCTTCCCAGTGGATGGAAAGAAGA-3′
P4 (−448/+186)	5′-TGAGGTACCCCGTCCGGGACAGAGGAA-3′	5′-CTTGAAGCTTCCCAGTGGATGGAAAGAAGA-3′
P5 (−169/+186)	5′-TGAGGTACCGAGAGGCATCTGCTGAC-3′	5′-CTTGAAGCTTCCCAGTGGATGGAAAGAAGA-3′
P6 (−1694/−448)	5′-TGAGGTACCTGAGAACAGCAAGGGAGAAATC-3′	5′- AGTAAGCTTGGCTTCCTCTGTCCCGGAC-3′
P7 (−1694/−753)	5′-TGAGGTACCTGAGAACAGCAAGGGAGAAATC-3′	5′- AGTAAGCTTAGAGTCCCTTACTTAGAATG-3′
P8 (−753/−158)	5′-TGAGGTACCCATTCTAAGTAAGGGACTCT-3′	5′-AGTAAGCTTCTGGTCAGCAGATGCCTCT-3′

Table 2.

Mutagenesis primers

Construct	Forward Primer	Reverse Primer
Myc MT	5′-GTGACTGAGGCAGATCCCTAA TCTGCACCTGGCCATGCTCTC-3′	5′-AGAGCATGGCCAGGTGCAGATTAGGGATCTGCCTCAGTCAC-3′
Sp1 MT1	5′-GCCAGGGTCTGGCGAGCTAAGGGAA TATTATCAGCAGCGGTGACTAGGGCGG-3′	5′-CCGCCCTAGTCACCGCTGCTGATAATA TTCCCTTAGCTCGCCAGACCCTGGC-3′
Sp1 MT2	5′-GGGCGGCGGCAGCGGTGACTATA TTATAAGCAGGAGCGGGAGCCGG-3′	5′-CCGGCTCCCGCTCCTGCTTATAATATAGTCACCGCTGCCGCC GCCC-3′

Quantitative PCR and semi-quantitative PCR

Experiments were performed as detailed in Supplementary data using priers specific for the CD133 transcript.

Bisulfite genomic sequencing

A DNA methylation kit (Qiagen) was used to detect cytosine methylation with minor modifications. Sodium bisulfite treatment was performed using 2 cg of genomic DNA in the presence of 5 µg carrier RNA and purified as per manufacturer’s protocol. The bisulfate-treated DNA was eluted using 50 µl of water, and desulfonation was performed by adding 3 m sodium hydroxide followed by incubation at room temperature for 5 min. The DNA was precipitated using 3 m sodium acetate and ice-cold ethanol followed by incubation in dry ice for 30 min. The precipitate was centrifuged at 12 000 r.p.m. at 4 °C, reconstituted in nuclease-free water and used immediately for PCR, or stored at −70 °C. For PCR amplification, 2 µl of the bisulfite-modified DNA was added in a final vsolume of 25 µl containing 1 X PCR buffer, dNTPs (1.25 mm each), primers (10 pmoleach), and 2.5 units of Promega Taq (Promega, Madison, WI, USA). The PCR products were resolved on a 1.5% agarose gel, gel-purified using the Qiaquick gel purification kit (Qiagen) and cloned using TOPO TA cloning system (Invitrogen, Grand Island, NY, USA). Ten colonies from each ligation were randomly selected and the PCR products submitted for sequencing. The primers used to amplify CD133 promoter CpG island region in bisulfate-treated DNA were: forward primer 5′-TATTTGGTTATGTTTTTAGTTTTTT-3′ and reverse primer 5′-CCTAATCAACAAATACCTCTCTC-3′.

Immunoblotting and FACS

Cells were dissociated using Accutase (Sigma-Aldrich, St Louis, MO, USA) and resuspended in PBS containing 0.5% bovine serum albumin and 2 mmol/l EDTA. Cells were stained with CD133/2-PE (Miltenyi Biotech, Auburn, CA, USA) at a dilution of 1:25 or isotype control antibody (mouse IgG PE, Miltenyi Biotech) also at a dilution of 1:25 and sorted on a BD FACSCalibur (BD Biosciences, San Jose, CA, USA). Sorted cells were collected in stem cell permissive medium and processed immediately for RNA isolation and subsequently DNA isolation using TRIZOL reagent (Invitrogen).

EMSA

EMSA was performed using a nonradioactivebased assay (Thermo Scientific Life Science, Rockford, IL, USA) (details in Supplementary data).

ChIP assay

ChIP assay was performed using EZ-ChIP chromatin immunoprecipitation kit (Millipore, Billerica, MA, USA) using GSC23 and U251HF cells and the primers indicated in Table 3 (See Supplementary data).

Table 3.

ChIP Primers

Name	Region	Forward primer	Reverse primer
ChP1	−669/−580	5′-GGAGTGCAGGGGGTTGAGCA-3′	5′-GCCCTGGACGCACTCTGATT-3′
ChP2	−579/−340	5′-TCGGGTTTCGCGATCTTTAA-3′	5′-CCGCGGAGGGGCTCAGGCGG-3′
ChP3	−339/−100	5′-CCGGCAGTGGGAGGCGGGCT-3′	5′-CACCCCCAGTACAGTGGAAG-3′
ChP4	−99/+21	5′-TACAGTGAGGAGTGGACGGG-3′	5′-GCAGTTCCTCTGGCCsCCCAG-3′