The Interaction between DNMT1 and High‐Mannose CD133 Maintains the Slow‐Cycling State and Tumorigenic Potential of Glioma Stem Cell

Abstract The quiescent/slow‐cycling state preserves the self‐renewal capacity of cancer stem cells (CSCs) and leads to the therapy resistance of CSCs. The mechanisms maintaining CSCs quiescence remain largely unknown. Here, it is demonstrated that lower expression of MAN1A1 in glioma stem cell (GSC) resulted in the formation of high‐mannose type N‐glycan on CD133. Furthermore, the high‐mannose type N‐glycan of CD133 is necessary for its interaction with DNMT1. Activation of p21 and p27 by the CD133–DNMT1 interaction maintains the slow‐cycling state of GSC, and promotes chemotherapy resistance and tumorigenesis of GSCs. Elimination of the CD133–DNMT1 interaction by a cell‐penetrating peptide or MAN1A1 overexpression inhibits the tumorigenesis of GSCs and increases the sensitivity of GSCs to temozolomide. Analysis of glioma samples reveals that the levels of high‐mannose type N‐glycan are correlated with glioma recurrence. Collectively, the high mannose CD133–DNMT1 interaction maintains the slow‐cycling state and tumorigenic potential of GSC, providing a potential strategy to eliminate quiescent GSCs.


Introduction
Cancer stem cells (CSCs) are thought to be responsible for tumor growth. [1] For example, CD133+ glioma cells form neurospheres, display multilineage differentiation capabilities in vitro, and are highly tumorigenic in the brains of immunocompromised migration. For example, CD133 can recruit HDAC6 to deacetylate -catenin to activate -catenin signaling. [15] Our previous studies have shown that the interaction between CD133 and the PI3K regulatory subunit p85 can activate the PI3K/Akt pathway to promote tumorigenesis of GSCs. [16] In addition, activation of FAK by the interaction between CD133 and Src promoted the migration of tumor cells. [17] Collectively, CD133 is a functional marker of CSCs. Therefore, clarifying CD133-interacting proteins might help to understand the mechanisms of CSCs maintaining quiescence.
Although CD133 has been a potential target for cancer treatment, [18] the structural ambiguity of N-glycan of CD133 limits its application in the isolation and elimination of CSCs. CD133 is a highly glycosylated membrane glycoprotein and contained nine N-linked glycosylation sites. [19] AC133 antibody, which is used to isolate stem cell, is widely reported to bind the glycosylated epitopes on CD133. [20,21] More importantly, during CSC differentiation, the N-glycan structure of CD133, but not CD133 protein or mRNA, was changed. [22] Therefore, the glycosylation status of CD133 is closely related to the cell differentiation. However, the structure of N-linked glycan of CD133 in CSCs remains uncovered. Here, we found that the structure of N-glycan of CD133 in GSC is high-mannose type. The high-mannose CD133 maintains the slow-cycling state and tumorigenic potential of GSCs through inhibiting the nuclear translocation of DNA methyltransferase 1 (DNMT1). DNMT1, a member of the DNA methyltransferase family, is responsible for maintaining methylation patterns located in CG dinucleotide-rich regions. [23] DNMT1 regulates chromatin organization, DNA repair, cell cycle regulation, and apoptosis. [24] The contributions of DNMT1 to CSCs have been extensively studied. Dnmt1 is essential for the maintenance of leukemia stem cells and mammary and CSC. Conversely, DNMT1 knockdown induces EMT and cancer stem-like phenotypes in prostate cancer. [25] Down-regulation of DNMT1 increases self-renewal potential of hepatoma cells. [26] Our finding proves that the high mannose CD133-DNMT1 interaction maintains the slow-cycling state and tumorigenic potential of GSC, providing a potential strategy to eliminate quiescent GSCs.

CD133 Interacts with DNMT1 In Vitro and In Vivo
We used a yeast two-hybrid screen in an attempt to identify CD133-interacting proteins. The C-terminal cytoplasmic domain of CD133 (residues 813-865) was used as the bait (Figure 1A). The cDNA encoding C-terminal cytoplasmic domain of CD133 (residues 813-865) was cloned into pGBKT7 vector and was used as the bait to screen pACT2-human cDNA libraries (human fetal brain). We isolated 6 positive clones from 1 × 10 6 clones of a human fetal brain library. Among the positive clones, we identified 3 encoding partial sequences of DNMT1 (112-235) ( Table  S1, Supporting Information). The interaction of CD133 (residues 813-865) with DNMT1 (112-235) in vitro was confirmed by GST pull-down assay ( Figure 1B). To validate the physical interaction between CD133 and DNMT1 in vivo, we isolated CD133+ and CD133-cells from human glioblastoma samples (T21286, T12752, and T08492; pathological data see Table S2, Supporting Information) as previously described ( Figure S1A, Supporting Information). [2,27] CD133+ tumor cells showed characteristics consistent with CSCs: namely, neurosphere formation (Figure S1B, Supporting Information); multilineage differentiation with markers for astrocytes (GFAP), neurons (MAP2) or oligodendrocytes (O4) ( Figure S1C, Supporting Information), and expression of stem cell markers Nestin and Sox2 ( Figure S1C,D, Supporting Information). CD133+ tumor cells were highly tumorigenic in the brains of immunocompromised mice, and CD133-cells did not form detectable tumor even when implanted at 5 × 10 5 cells per mouse, except for occasional small tumor from a single xenograft source ( Figure S1E-G, Supporting Information). Endogenous DNMT1 interacted with endogenous CD133, as shown by reciprocal co-IP assays ( Figure 1C and Figure S1H, Supporting Information). However, CD133 did not interact with other DNMT family members, including DNMT3A or DNMT2 ( Figure 1D).
Next, we searched the region responsible for the interaction between CD133 and DNMT1. By strep pull-down assay, CD133 Cterminal segment (amino acids 848-865) interacted with DNMT1 in vitro ( Figure 1E). A co-IP assay in CD133+ glioma cells expressing CD133 shRNA and either shRNA-resistant wild-type CD133 or its deletion mutant, further showed that the CD133(1-862) mutant lacking a region between residues 863 and 865 could not interact with DNMT1 ( Figure S1I, Supporting Information). In a yeast two-hybrid system, amino acids 112-235 of DNMT1 interacted with the CD133 C-terminus (Table S1, Supporting Information). Co-IP assays in CD133+ cells expressing FLAG-tagged DNMT1 deletion mutants showed that the deletion of a region between DNMT1 residues 155 and 163 reduced the interaction between DNMT1 and CD133 ( Figure 1F and Figure S1J, Supporting Information). To examine the interaction between fulllength CD133 and DNMT1 in vitro, strep-tagged CD133 protein and its mutant Del(836-865) was purified by Strep-Tactin affinity. By Coomassie blue staining, the purity of CD133 purified protein was over 90% ( Figure S1K, Supporting Information). By Strep pull-down assay, deletion of CD133 amino acids 848-865 reduced the interaction between CD133 and DNMT1 in vitro (Figure S1L, Supporting Information). By enzyme-linked immunosorbent assay (ELISA), CD133 bound to DNMT1 with Kd = ∼150 nM (Figure S1M, Supporting Information). Together, CD133 interacts with DNMT1 in GSCs depending on its C-terminal cytoplasmic domain.

The Interaction Between CD133 and DNMT1 Inhibits the Nuclear Translocation of DNMT1
CD133 is mainly located on the cell surface. [28] DNMT1 is usually located in the nucleus. [24] We presumed that CD133 might regulate the nuclear localization of DNMT1. By immunofluorescence analysis, exogenous CD133-GFP co-localized with DNMT1-dsRed in the cytoplasm in CD133+ cells ( Figure S2A, Supporting Information). Immunofluorescence analysis showed the colocalization between endogenous CD133 and DNMT1 in the cytoplasm in CD133+ cells (Figure 2A), and in GBM tissues ( Figure  S2B, Supporting Information). Cytoplasmic CD133 is partly located in endosome. [29] Consistent with this finding, cytoplasmic CD133 in CD133+ glioma cells colocalized with the endosome marker EEA1 ( Figure 2B), not with the Golig marker GM130 ( Figure S2C, Supporting Information). By immunoprecipitation . This protein is modeled as having an extracellular N terminus, a cytoplasmic C terminus, two small cytoplasmic loops, and two large extracellular loops. C-terminal cytoplasmic domain of CD133 (residues 813-865) (indicated by square frame) is used as the bait for yeast two-hybrid screen. B) In vitro interaction between CD133 and DNMT1. GST or GST-CD133(813-865) proteins are incubated with purified His-DNMT1 (112-235) protein. The GST pull-down products are blotted with anti-GST and anti-His antibodies. C) CD133 interacts with DNMT1 in vivo. The lysates of CD133+ cells and CD133cells isolated from glioblastoma samples are subjected to IP using anti-CD133 (Clone W6B3C1) or anti-DNMT1 antibodies, followed by immunoblotting (IB) with anti-CD133 or anti-DNMT1 antibodies. Whole-cell lysates are analyzed by IB with anti-CD133 or anti-DNMT1 antibodies as input. D) The interaction between CD133 and the members of DNMT is examined by Co-IP assay. Lysates of CD133+ cells are subjected to IP using anti-CD133 antibody (Clone W6B3C1), followed by IB with anti-CD133 (Clone W6B3C1), anti-DNMT1, anti-DNMT2, or anti-DNMT3A antibodies. Whole-cell lysates are analyzed by IB with anti-CD133, anti-DNMT1, anti-DNMT2, or anti-DNMT3A antibodies as input. E) Strep peptide or strep-tagged CD133 c-terminal deletion mutant are incubated with purified DNMT1 protein. The Strep pull-down products are blotted with anti-DNMT1 antibody. F) The lysates of CD133+ cells expressing FLAG or DNMT1-FLAG or DNMT1(Del(155-163)) are subjected to IP using anti-FLAG antibody, followed by IB with anti-FLAG, or anti-CD133 antibodies (Clone W6B3C1). Whole-cell lysates are analyzed by IB with anti-CD133 (Clone W6B3C1), anti-FLAG, or anti-GAPDH antibodies as input.
CD133 knockdown did not change the expression level of DNMT1 mRNA or DNMT1 protein ( Figure S2D,E, Supporting Information). CD133 knockdown increased the nuclear translocation of DNMT1 in CD133+ glioma cells without obviously changing the nuclear translocation of DNMT3a or DNMT3b (Figure 2D and Figure S2F, Supporting Information). The nuclear translocation of DNMT1 was decreased in CD133+ glioma cells compared to CD133-glioma cells ( Figure 2E,F). Thus, CD133 inhibits DNMT1 nuclear translocation. To explore the significance of the CD133-DNMT1 interaction in CD133-reduced DNMT1 nuclear translocation, CD133+ glioma cells were expressed CD133 shRNA and either shRNA-resistant wild-type CD133 or shRNA-resistant CD133(1-862) mutant. The nuclear translocation of CD133-binding deficient DNMT1 mutant (Del(155-163) was markedly enhanced, compared to wild type DNMT1 ( Figure  2H). Ectopic expression of shRNA-resistant wild-type CD133, but not expression of the CD133(1-862) mutant, restored the effect of CD133 knockdown on the nuclear translocation of DNMT1 ( Figure 2G and Figure S2I, Supporting Information). Deletion of the DNMT1 region between aa 155 and 163, which is essential for the interaction between CD133 and DNMT1, significantly increased the nuclear localization of DNMT1 in CD133+ cells ( Figure 2H). Thus, the CD133-DNMT1 interaction inhibits the nuclear translocation of DNMT1.
Next, a series of experiments were performed to examine the contribution of the CD133-DNMT1 interaction to high level of p21 and p27 in GSCs. First, CD133 knockdown increased the level of 5-methylcytosine in CD133+ glioma cells ( Figure  S3D). Ectopic expression of shRNA-resistant wild-type CD133, but not that of the CD133(1-862) mutant, restored the effect of CD133 knockdown on the level of 5-methylcytosine ( Figure S3E, Supporting Information). Second, ectopic expression of shRNAresistant wild-type CD133, but not that of the CD133(1-862) mutant, restored the inhibitory effect of CD133 knockdown on the promoter methylation of p21 and p27 ( Figure S3F, Supporting Information). Third, ectopic expression of shRNA-resistant wildtype CD133, but not that of the CD133(1-862) mutant, eliminated the effects of CD133 knockdown on the expression of p21 and p27 ( Figure S3G, Supporting Information). Finally, downregulation of DNMT1 by shRNA ( Figure S3H), restored the effect of CD133 knockdown on the promoter methylation of p21 and p27 ( Figure  S3I, Supporting Information). Together, CD133 upregulates p21 and p27 depending on its interaction with DNMT1.

Nuclear localization of DNMT1 Inhibits the Self-Renewal Capacity and Tumorigenesis of GSCs
P21 and p27 maintain the quiescence and self-renewal capacity of stem cells. [31][32][33] P21 and p27 promote the association of CDK4 with the D-type cyclins, inhibit the kinase activity of CDK4 i which could phosphorylate Rb protein, and subsequently prevent cell cycle progression. [34] Downregulation of p21 or p27 in CD133+ cells ( Figure S4A,B, Supporting Information), increased the activity of CDK4 ( Figure S4C,D, Supporting and matched CD133-cells is determined by IB. Histone H3 is used as the nuclear marker, and -tubulin is used as the cytosolic marker. E. The figures are presented out of three independent experiments. F) The relative densities of DNMT1 to Histone H3 are quantified using densitometry. Values are normalized to that of CD133+ cells. Results are expressed as mean ± SD from three independent experiments; t test, ***p < 0.001, Student's t-test. G) The level of nuclear DNMT1 in CD133+ cells expressing control shRNA, CD133 shRNA1, CD133 shRNA1 + shRNA-resistant wild-type CD133, or CD133 shRNA1 + shRNA-resistant CD133(1-862) mutant is determined by IB. Histone H3 is used as the nuclear marker, and -tubulin is used as the cytosolic marker. The relative densities of DNMT1 to Histone H3 are quantified using densitometry. Values are normalized to that of cells expressing control shRNA. Results are expressed as mean ± SD from three independent experiments; ***p < 0.001, Student's t-test. H) The level of nuclear DNMT1-FLAG in CD133+ cells expressing FLAG-DNMT1 or DNMT1(Del(155-163)) is determined by IB. Histone H3 is used as the nuclear marker, and -tubulin is used as the cytosolic marker.  (155-163)). GAPDH is used as a loading control. B) By Infinium MethylationEPIC BeadChip arrays, the methylation of 680 annotated genes is increased in CD133+ cells expressing DNMT1 (Del(155-163)) compared to control cells (p < 0.001, Δ ≥ 0.15). Gene ontology results (top five, according to p value) for 680 genes in which methylation is upregulated in CD133+ cells expressing FLAG-DNMT1(Del(155-163)) are shown. C) qRT-PCR quantification of the indicated gene mRNA levels in T21286 CD133+ cells expressing FLAG or FLAG-DNMT1(Del (155-163)). Data is shown as mean ± SD from three independent experiments; ***p < 0.001, **p < 0.01, *p < 0.05, #p > 0.05, Student's t-test. D-E) The methylation rate of p21 and p27 promoters in CD133+ cells and matched CD133-cells from T21286 (D) and T12752 (E) are analyzed by bisulfite sequencing. Methylation levels are determined by the ratio of converted C nucleotides to total C nucleotides following bisulfite treatment under CpG island. Results are expressed as mean ± SD from three independent experiments; ***p < 0.001, Student's t-test. F) Western blot analysis of p21 and p27 expression in T21286 CD133+ cells and CD133-cells. GAPDH is used as a loading control. G) Western blot analysis of p21 and p27 expression in T21286 CD133+ cells treated with 2% FBS for 7 days. GAPDH is used as a loading control. H) Chromatin immunoprecipitation (ChIP) assay is performed in CD133+ cells and CD133-cells using a DNMT1 specific antibody, followed by PCR amplification of p21 and p27 promoter regions between +250 to position −100. Chromatin (defined as input) and GAPDH products immunoprecipitated by DNMT1 Ab are used as positive and negative control. Information). Consistent with this finding, the kinase activity of CDK4 was lower in CD133+ cells than in CD133-cells ( Figure  S4E, Supporting Information). By FCS analysis, downregulation of p21 or p27 in CD133+ cells promoted G1/S transition ( Figure  S4F,G, Supporting Information), and increased the ratio of EdUpositive cells ( Figure S4H, Supporting Information). Quiescence of stem cells acts to limit the accumulation of DNA damage in normal and CSCs. [35] -H2AX foci are widely used as a marker of DNA damage. [36] Downregulation of p21 or p27 increased the ratio of -H2AX-positive cells ( Figure S4I, Supporting Information). HSCs remain in quiescence to sustain their long-term selfrenewal potential. [37] The single-cell neurosphere formation assay is the conventional method to measure the self-renewal capacity of GSCs. [11,27] Down-regulation of p21 or p27 increased the diameter of spheres at passage 1 ( Figure S4J, Supporting Information). However, the number of spheres was significantly decreased at passage 2 ( Figure S4J,K, Supporting Information). By the in vivo limiting dilution assay, downregulation of p21 or p27 inhibited the tumorigenesis of CD133+ cells ( Figure S4L, Supporting Information). Thus, down-regulation of p21 or p27 promotes the proliferation and reduces the self-renewal potential of GSCs.
The catalytic cysteine C1226 at DNMT1 (human) is necessary for its DNA methyltransferase activity. [38,39] Due to the FLAG tag and the deletion of amino acids, catalytic cysteine at position 1226 was shifted to position 1229 in the FLAG-DNMT1 Del(155-163) protein. To evaluate the effect of CD133-inhibited DNMT1 nuclear localization in GSCs, CD133+ cells were constructed to express FLAG-tagged DNMT1 Del(155-163) or its C1229S mutant ( Figure 4A). Ectopic expression of DNMT1 Del(155-163), but not its C1229S mutant, significantly increased the level of 5-methylcytosine in CD133+ cells ( Figure 4B) and the ratio of EdU-positive cells ( Figure 4C). And, ectopic expression of DNMT1 Del(155-163), but not its C1229S mutant reduced the sphere formation of CD133+ cells at passage 2 (Figure 4D,E), inhibited the stem cell activity by limited dilution assay ( Figure 4F,G), and inhibited the tumor-initiating capacity of CD133+ cells ( Figure 4H). Importantly, overexpression of DNMT1 Del(155-163), but not its C1229S mutant, increased the survival of tumor-bearing mice ( Figure 4I-K). We further evaluated whether DNMT1 regulates the quiescence state of CD133+ cells through inhibition of p21 and p27 expression. Ectopic expression of DNMT1 Del(155-163), but not its C1229S mutant increased the G1/S transition in CD133+ cells ( Figure S4M,N, Supporting Information). The effect of DNMT1 Del(155-163) on the sphere formation and cell proliferation of CD133+ cells could be partially rescued by ectopic expression of p21 or p27 ( Figure  S4O,P, Supporting Information). Together, the nuclear localization of DNMT1 inhibits GSC self-renewal and tumorigenesis, mainly depending on its DNA methylation activity.

CD133 Maintains the Self-Renewal and Tumorigenesis of GSCs Through Its Interaction with DNMT1
CD133 knockdown impaired the self-renewal of GSCs. [16,40] Supporting this notion, CD133 knockdown inhibited the expression of p21 and p27 ( Figure S5A, Supporting Information) and increased pRb phosphorylation ( Figure S5B, Supporting Informa-tion). And, CD133 knockdown reduced the sphere formation of CD133+ cells ( Figure S5C,D, Supporting Information), inhibited the stem cell activity by limited dilution assay ( Figure S5E,F, Supporting Information), and reduced the tumor-initiating capacity of CD133+ cells ( Figure S5G, Supporting Information). Next, the self-renewal and tumorigenic abilities of CD133+ glioma cells expressing CD133 shRNA and either shRNA-resistant wild-type CD133 or shRNA-resistant CD133(1-862) mutant were evaluated. The number of spheres was significantly decreased in passage 2-3 with knockdown of CD133. The inhibitory effect of CD133 knockdown on neurosphere formation was fully rescued by the expression of shRNA-resistant wild-type CD133, but not by the shRNA-resistant CD133(1-862) mutant ( Figure 5A,B and Figure S5H, Supporting Information). And, by limited dilution assay, the inhibitory effect of CD133 knockdown on stem cell activity of GSC was fully rescued by the expression of shRNA-resistant wild-type CD133, but not by the shRNA-resistant CD133(1-862) mutant ( Figure S5I,J, Supporting Information). Furthermore, ectopic expression of wild-type CD133, but not the CD133(1-862) mutant, fully rescued inhibitory effect of CD133 depletion on the tumor-initiating capacity of GSCs ( Figure 5C). Importantly, CD133 knockdown increased the survival of tumor-bearing mice, which was fully restored by expression of shRNA-resistant wildtype CD133, but not by the shRNA-resistant CD133(1-862) mutant ( Figure 5D-G). By histochemical staining in xenografts, a reduction in Nestin, CD133, and cytoplasmic DNMT1 in xenografts formed by CD133+ cells expressing CD133 shRNA could be rescued by ectopic expression of wild-type CD133, not by expression of the CD133(1-862) mutant ( Figure S5K, Supporting Information). Thus, the CD133-DNMT1 interaction is critical for CD133 sustaining GSC self-renewal and tumorigenesis.

The CD133-DNMT1 Interaction Maintains GSC Slow-Cycling State
The contribution of the CD133-DNMT1 interaction to GSC quiescence was next investigated. By EdU incorporation assay, CD133 knockdown exhibited a 3-to 4-fold increase in EdU incorporation compared to the control cells (Figure 6A,B). CD133glioma cells exhibited a 4-to 5-fold increase in EdU incorporation compared to CD133+ glioma cells ( Figure 6C). FACS analysis further showed that downregulation of CD133 promoted the G1/S transition ( Figure S6A, Supporting Information). We next performed CIdU and IdU incorporation assays to confirm the quiescence state of GSCs in vivo. CD133+ glioma cells were transplanted into mouse brain and CldU and IdU were injected 7 days and 28 days, respectively. The mice were killed 24 h after injection of IdU ( Figure 6D). CD133+ glioma cells possessed higher CIdU incorporation and lower IdU incorporation than CD133-glioma cell ( Figure 6E). CD133 knockdown increased H2AX foci formation in GSCs ( Figure S6B,C, Supporting Information). Thus, CD133 downregulation enhances the proliferation and DNA damage of GSCs.
We next performed a series of experiments to examine the contribution of the CD133-DNMT1 interaction to GSC quiescence. First, forced expression of shRNA-resistant wild-type CD133, but not of CD133(1-862), rescued the effect of CD133 knockdown on EdU incorporation and G1/S transition ( Figure 6F  Third, forced expression of p21 or p27 rescued the effect of CD133 knockdown on the EdU incorporation and H2AX foci formation of CD133+ cells ( Figure 6G,H). Thus, upregulation of p21 and p27 by the CD133-DNMT1 interaction maintains GSC slow-cycling state.

The CD133-DNMT1 Interaction Promotes the Resistance of GSCs to the Chemotherapeutic Agent Temozolomide
Quiescent CSCs are resistant to conventional chemotherapy and radiation. [4,43,44] GSCs are resistant to conventional chemotherapy and radiation. [45,46] Next, the contribution of the CD133-DNMT1 interaction to GSC resistance to temozolomide was evaluated. CD133+ glioma cells were transplanted into mouse brain. After 4 weeks, tumor-bearing mice were orally administered temozolomide ( Figure 7A). After temozolomide chemotherapy, the expression of the DNA damage marker -H2AX and the apop-tosis marker cleaved-caspase-3 was obviously increased, indicating the effectiveness of temozolomide treatment ( Figure 7B and Figure S7A, Supporting Information). By immunohistochemical (IHC) staining, the level of cytoplasmic DNMT1 was obviously increased after temozolomide treatment ( Figure S7B, Supporting Information).
By FACS analysis of CD133 expression in xenografts, temozolomide treatment increased the ratio of CD133+ cells (Figure 7C), indicating that CD133+ cells are resistant to temozolomide. Supporting this point, the ratio of apoptotic cells induced by temozolomide was significantly reduced in CD133+ cells compared to CD133-cells ( Figure S7C, Supporting Information).
Next, the contribution of the CD133-DNMT1 interaction to GSC resistance to temozolomide was examined. First, CD133 knockdown increased the percentage of apoptotic cells induced by temozolomide ( Figure S7D, Supporting Information). Ectopic expression of p21 or p27 rescued the effect of CD133 depletion on the resistance of CD133+ cells to temozolomide ( Figure 7D). Second, the inhibitory effect of CD133 knockdown on the resistance of CD133+ cells to temozolomide was fully rescued by the expression of shRNA-resistant wild-type CD133, but not by the shRNA-resistant CD133(1-862) mutant ( Figure 7E). Third, TMZ reduced the tumor-initiating capacity of shRNACD133-1/CD133+ cell expressing shRNA-resistant CD133(1-862) mutant, but not shRNA-resistant wild-type CD133 ( Figure 7F). The TAT-strep-CD133(848-865) peptide increased the ratio of apoptotic CD133+ cell induced by temozolomide ( Figure S7E, Supporting Information). Resistance to chemotherapeutic agents promotes cancer recurrence. [47,48] By IHC staining on paraffinembedded sections from paired primary and recurrent tissues, the level of cytoplasmic DNMT1 in recurrent tissues was significantly higher than in primary tissues ( Figure 7G,H). Together, the CD133-DNMT1 interaction promotes GSC resistance to the chemotherapeutic agent temozolomide.

The High-Mannose N-Glycan of CD133 Promotes Its Interaction with DNMT1
After the differentiation of GSCs, the levels of CD133 protein and mRNA expression were decreased (Figure 8A-C). Interestingly, the molecular weight of CD133 was obviously increased after GSCs differentiation ( Figure 8A  dotted line). The change of protein molecular weight frequently results from posttranslational modification. CD133 is a heavily N-glycosylated protein. [49] The glycan structure of glycoproteins can be recognized by plant lectins ( Figure 8D). [50] CD133 immunoprecipitated from GSCs could be recognized by ConA lectin (recognizing high mannose), but not by PHA-L lectin (recognizing -1,6 branched N-acetylglucosamine). However, CD133 immunoprecipitated from differentiated cells could be recognized by PHA-L lectins, but not by Con A lectin ( Figure 8E-G). Thus, during the differentiation of GSC, the structure of CD133 N-glycan is converted from the high-mannose type to the complex type.
N-Glycosylation regulates protein stability and protein delivery to the cell membrane. [53,54] Knockdown of MAN1A1 in differentiated cells ( Figure S8B, Supporting Information), did not change the delivery of CD133 to the cell surface ( Figure S8C, Supporting Information). Knockdown of MAN1A1 in differentiated cells increased the interaction between CD133 and DNMT1 ( Figure 8J). Ectopic expression of MAN1A1 in CD133+ cells decreased the interaction between CD133 and DNMT1 ( Figure  S8D, Supporting Information). Furthermore, MAN1A1 overexpression increased the nuclear location of DNMT1, which could be blocked by CD133 knockdown ( Figure S8E, Supporting Information). MAN1A1 overexpression induced the level of 5methylcytosine, which could be blocked by CD133 knockdown or CD133-binding deficient DNMT1 mutant ( Figure 8K,L). Col-lectively, high-mannose N-glycan inhibits DNA 5-methylation by maintaining the CD133-DNMT1 interaction.
Next, the contribution of lower MAN1A1 expression to the characteristics of GSCs was evaluated. Ectopic expression of MAN1A1 increased TMZ-induced CD133+ cells apoptosis, which was obviously blocked by CD133 knockdown (Figure S8F, Supporting Information). MAN1A1 overexpression inhibited the self-renewal and tumorigenesis of CD133+ cells, which was obviously blocked by CD133 knockdown (Figure S8G,H, Supporting Information). Knockdown of MAN1A1 inhibited serum-induced differentiation of CD133+ cells ( Figure S8I-K, Supporting Information). We isolated cells with high mannose N-glycan (HM) from human glioblastoma samples (T21278, T22456) through magnetic cell sorting using Con A lectin ( Figure S8L, Supporting Information). HM+ tumor cells showed characteristics consistent with CSCs, neurosphere formation ( Figure S8M, Supporting Information), and expression of the stem cell markers Sox2 ( Figure S8N, Supporting Information). HM+ tumor cells were highly tumorigenic in the brains of immunocompromised mice, and HM-cells did not form detectable tumor even when implanted at 10 5 cells ( Figure S8O,P, Supporting Information). The ratio of apoptotic cells induced by temozolomide was significantly reduced in HM+ cells compared to HM-cells ( Figure S8Q, Supporting Information). By IHC staining on paraffin-embedded sections from paired primary and recurrent tissues, the level of high mannose N-glycan in recurrent tissues was significantly higher than in primary tissues ( Figure S8R,S, Supporting Information). Together, high-mannose type N-glycan is an enrichment marker for CSCs in human glioblastoma.

Discussion
We present evidence that the lower expression of MAN1A1 results in the formation of high-mannose type N-glycan of CD133 in GSCs. The interaction between high-mannose CD133 and DNMT1 blocks the nuclear translocation of DNMT1. Activation of p21 and p27 expression by the CD133-DNMT1 interaction maintains GSC quiescence, self-renewal, chemotherapy resistance, and tumorigenesis ( Figure S8T, Supporting Information). Increasing evidence has shown that the quiescent state of CSCs protects them from DNA damage. The relatively quiescent state is necessary for preserving the self-renewal of CSCs. The Results are expressed as mean ± SD from three independent experiments; ***p < 0.001, **p < 0.01, Student's t-test. Scale bar represents 10 μM. C) Wild type CD133, but not CD133(1-862) mutant, rescues the effect of CD133 knockdown on the tumor-initiating capacity of CD133+ cells. An intracranial limiting dilution tumor formation assay (employing 10 000, 5000, 1000, and 500 cells per mouse) is performed using CD133+ cells infected with the indicated lentivirus. The table displays the number of mice developing tumors. D-G) CD133+ cells from glioblastoma specimen T21286 (D,F) or T12752 (E,G) expressing Control shRNA, CD133 shRNA1, CD133 shRNA1+shRNA-resistant wild type CD133, or CD133 shRNA1+shRNA-resistant CD133(1-862) are implanted into immunocompromised mice brain (5000 cells per mouse). Mice are sacrificed when they are moribund or 120 days after implantation. D,E) Survival of mice (n = 6) is evaluated by Kaplan-Meier analysis (**p < 0.01, log rank test). F,G). H&E staining of mouse brain shows tumors formation by CD133+ cells expressing Control shRNA, CD133 shRNA1, CD133 shRNA1+shRNA-resistant wild type CD133, or CD133 shRNA1+shRNA-resistant CD133(1-862). Scale bar, 1 cm. shR, shRNA-resistant. Scale bar, 1 cM. H) T12752 CD133+ cells treated with the indicated peptides are i implanted into immunocompromised mice brain (5000 cells per mouse). Mice are sacrificed when they are moribund or 120 days after implantation. Survival of mice (n = 6) is evaluated by Kaplan-Meier analysis (**p < 0.01, log rank test). I,J) T12752 CD133+ cells treated with the indicated peptides cells are subcutaneously injected into immunodeficient mice. I) The images of the xenograft of GSC treated with control or peptides. Scale bar, 1 cm. J) Tumor volumes are measured after tumor cell inoculation every three days. Results are expressed as mean ± SD (n = 6 mice; **p < 0.01). Student's t-test. Results are expressed as mean ± SD from six independent experiments; ***p < 0.001, Student's t-test. D,E) In vivo CIdU and IdU incorporation assay is used to examine the proliferation of GSCs. D. CD133+ cells are intracranially implanted into immunocompromised mice brain. 7 days later, mice are intraperitoneally injected with CIdU. After 4 weeks, mice are intraperitoneally injected with IdU. 24 h later, mice are sacrificed and perfused. E) Glioblastoma orthotropic xenograft is assessed by immunofluorescence staining of CD133 (red), IdU (green), and CIdU (purple). White arrow indicates the CD133+ cells. Scale bar represents 10 μM. F) Immunofluorescence analysis the percentage of EdU-positive cells in CD133+ cells expressing control shRNA, CD133 shRNA1, CD133 shRNA1+shR CD133, or CD133 shRNA1+shR CD133(1-862). Results are expressed as mean ± SD from six independent experiments; ***p < 0.001, Student's t-test. shR, shRNAresistant. G) Analysis of the percentage of EdU-positive cells in CD133+ cells expressing control shRNA or CD133 shRNA1 and control or p21 or p27. Results are expressed as mean ± SD from six independent experiments; **p < 0.01, *p < 0.05, Student's t-test. H) Immunofluorescence analysis of H2AX foci formation in CD133+ cells expressing control shRNA or CD133 shRNA1 and control or p21 or p27. The number of H2AX foci-positive cells is measured. Results are expressed as mean ± SD from three independent experiments; #p > 0.05, ***p < 0.001, Student's t-test. identification of the CD133/DNMT1 signaling axis provides a new insight into regulating the quiescence of GSC and a therapeutic target for the abrogation of quiescent CSCs.

The High Mannose Form of CD133 is Required for Its Interaction with DNMT1
The glycosylation status of CD133 is closely related to cell differentiation. [22] We find that differentiation provokes CD133 glycosylation structure from high mannose type to complex type. The high-mannose N-glycan promotes the self-renewal and tumorigenesis of GSCs by increasing the interaction between CD133 and DNMT1. High mannose N-glycan promotes the metastasis, migration, and growth in vivo of cancer cells in various tissues. [52,55] Our findings provide a new role of highmannose N-glycan in cancer progression.
The mechanisms by which the high-mannose N-glycan of CD133 regulates its interaction with DNMT1 remain unclear. Ectopic expression of MAN1A1 did not influence CD133 protein expression or delivery to the cell surface. Previous studies have shown that inhibition of the CD133 complex N-glycosylation influences the recognition of the AC133 epitope without changing CD133 protein expression. [56] N-glycosylation regulates the structure of membrane protein. [57] Thus, we presume that the Nglycosylation of CD133 regulates its interaction with DNMT1 by influencing the CD133 structure on the cell surface.

High Mannose-Type N-Glycan is an Enrichment Marker for GSCs
Another important finding is that glioma cells with high mannose N-glycan on cell surface showed characteristics consistent with CSCs. Actually, it has been reported that high mannose glycan on cell surface is associated with stem cell characteristic(s). For example, human embryonic stem cells were found to have high levels of high mannose glycan on cell surface. [58] Glycosylation features associated with MSCs rather than differentiated cells included high-mannose type N-glycans. [59] Accordingly, high mannose-type N-glycan is an enrichment marker for GSCs. Supporting this finding, high-mannose glycans were increased according to HCC dedifferentiation. [60] Growing evidence suggests that two distinct types of CSC lead to the formation of GBM. Type I CSC lines display "proneural" signature genes and are CD133 positive. Type II CSC lines show "mesenchymal" transcriptional profiles and lack CD133 expression. [61,62] Collectively, we presume that high-mannose type N-glycan might be an enrichment marker for proneural GSC.

Nuclear Localization of DNMT1 Inhibits the Slow-Cycling State and Tumorigenesis of GSCs
The contributions of DNMT1 to CSCs have been extensively studied. In pancreatic ductal adenocarcinoma, pharmacologic or genetic targeting of DNMT1 in CSCs reduces their selfrenewal and in vivo tumorigenic potential. [63] In breast cancer cells, DNMT1 deletion inhibited the self-renewal and proliferation of cancer-initiating cells. [64] However, silencing DNMT1 promoted the induction of the CSC phenotype in prostate cancer cells. [25] DNMT1 knockdown increases self-renewal potential and tumorigenesis of hepatoma cells. [26] In glioma, the level of 5-mC is negatively related to the grade of glioma. [65] Lower methionine inhibits the expression of DNMT1 and promotes the self-renewal and tumorigenesis of GSCs. [66] We found that nuclear DNMT1 promoted the proliferation of GSCs. The quiescent state of stem cells acts to limit the accumulation of DNA damage in normal and CSCs. [35] Thus, nuclear DNMT1 inhibited the long-term self-renewal potential and tumorigenesis of GSCs. Collectively, DNMT1 has opposite effects on CSCs from different tumor sources. Glioblastoma is a rapidly evolving highgrade astrocytoma by the presence of necrosis and microvascular hyperplasia. [67] Quiescence or slow-dying helps to maintain GSC in niche. Proneural GSCs expressed CD133 and mesenchymal GSCs lacked CD133 expression. [62] Therefore, the CD133 mainly interacts with DNMT1 in the proneural GSC. DNMT1 knockdown promotes the tumorigenesis of hepatoma stem cells through up-regulation of BEX1. [68] Here, we provided evidence that up-regulation of p21 and p27 by CD133-DNMT1 interaction promotes the GSC quiescence. The quiescent or slowgrowing state of CSCs protects them from DNA damage. Accumulation of DNA damage results in the cell exhaustion. [8] Thus, the quiescent or slow-growing state of CSCs is necessary for the self-renewal and tumorigenesis of CSC. [7] Thus, we presume that DNA damage induced by the inhibition of CD133-DNMT1 interaction inhibited the self-renewal and tumorigenesis of GSC. Our finding provides a new mechanism of GSC quiescence. However, CSCs display significant phenotypic and functional heterogeneity. For example, breast CSCs display plasticity transitioning between quiescent mesenchymal-like (M) and proliferative epithelial-like (E) states. [69] The contribution of the CD133-DNMT1 interaction in the CSCs transition between quiescent and proliferative states needs further examination. In summary, our results uncover CD133 as a crucial regulator of quiescence, self-renewal, and tumorigenesis of GSCs. More importantly, elimination of the interaction between CD133 and DNMT1 by a cell-penetrating peptide inhibits the self-renewal and tumorigenesis of GSCs and increases the sensitivity of GSCs to temozolomide. Thus, targeting the interaction between CD133 and DNMT1 might help to abrogate quiescent CSCs. Our findings not only provide an improved understanding of the fundamental role of high-mannose N-glycan of CD133 in the tumorigenesis of GSCs, but also suggest an additional target for the abrogation of quiescent CSCs.

Experimental Section
Isolation of CD133+ and CD133-Cells: CD133+ cells were isolated from primary surgical GBM biopsy specimens in accordance with protocols approved by the Fudan University Institutional Review Broads. All patients have been informed and consented to involve in this study. CD133+ and CD133-cells were isolated through magnetic cell sorting with CD133 cell isolation Kit (Miltenyi Biotec, cat#130-100-857) as previously described. [2] Isolation of Con A+ and Con A-Cells: Con A+ cells and Con A-were isolated from primary surgical GBM biopsy specimens. Fresh tissues were minced and treated with 0.2% collagenase (Sigma) and 1% Dispase II (Sigma) at 37°C for 1-2 h. The resulting single-cell suspensions were filtered through Cell Strainer (Corning). Cells were re-suspended in MACS buffer containing biotinylated Con A antibody (Sigma) and incubated on ice for 30 min. After washing, the cells were incubated with Streptavidinconjugated MicroBeads for 15 min on ice. Positive cells were re-suspended in MACS buffer containing biotinylated PHA-L antibody and biotinylated DSL antibody (Sigma) and incubated on ice for 30 min. After washing, the cells were incubated with Streptavidin MicroBeads for 15 min on ice. PHA-L and DSL-positive cells were excluded. The ratio of high mannose N-glycan in Con A+ cells and Con A-cells were then analyzed by FCS.
Tumor Formation Assay: Intracranial transplantation of tumor cells into 6 to 8-week old immunodeficient mice was performed in accordance with a Fudan University Institutional Animal Care and Use Committeeapproved protocol concurrent with national regulatory standards. Mice were maintained for up to 180 days or until the development of neurologic signs that significantly inhibited their quality-of-life (e.g., ataxia, lethargy, seizures, inability to feed, etc.).
To examine the effect of peptides on the growth of GSCs, cells treated with the peptides were subcutaneously injected into immunodeficient mice. Tumor size was measured and their volumes were calculated using the equation of L (length) × W (width) 2 /2.
Neurosphere Formation Assay: For single-cell neurosphere formation assay, 48 h after treatment with the indicated lentivirus, cells were trypsinized and single-cell suspensions were cultured in 24-well plates containing supplemented DMEM/F12 medium with lower concentration growth factor (2 ng mL −1 EGF). After 5 days, the number of neurospheres/well was quantified (passage 1). For secondary/tertiary neurospheres formation assay (passages 2 and 3), the established neurospheres were dissociated into single cells and were cultured in 24-well plates. The number of spheres with secondary neurospheres was counted after 5 days.
For the limiting dilution assay of GSCs, 0, 5, 10, 15, 20, and 25 cells were seeded into a 24 well plate each. After 2 weeks, the number of wells which had neurospheres was counted (n = 10), ***p < 0.001 by ELDA analysis. [70] Analysis of GBM Subtype: Briefly, RNA from patient samples was extracted and profiled on Affymetrix Human Exon 1.0 ST Gene Chips according to the manufacturer's protocol. Proneural, neural, classical, and mesenchymal GBM subtypes were determined by clustering of expression data from the Affymetrix HuEx array platform using the previously published gene marker. [71] Cell Cultures: The sorted CD133+ cells were cultured in the DMEM/F12 media supplemented with B27 lacking vitamin A (Invitrogen), 2 μg mL −1 heparin (Sigma), 20 ng mL −1 EGF (Chemicon), and 20 ng mL −1 FGF-2 (Chemicon) for a short period before treatment and analysis. CD133 − tumor cells were plated in DMEM with 10% fetal bovine serum for at least 12 h to permit cell survival. Prior to performing experiments with CD133cells, DMEM with 10% fetal bovine serum was replaced with supplemented DMEM/F12 media in order for experiments to be performed in identical media.
For ectopic expression of p27 or p21, the LV-P21-FLAG or LV-P27-FLAG plasmid was constructed by inserting full-length human p21 cDNA or p27 cDNA into the LV-FLAG lentivirus vector between BamHI and AgeI site.
DNA methylation analysis: Cytosine methylation was determined using bisulfite sequencing. Briefly, genomic DNA was subjected to bisulfite conversion with the EpiTect bisulfite kit (Qiagen) according to the manufacturer's instructions. Modified DNA was purified with a Qiagen Gel Extraction Kit. The fragments, which encompasses the CpG sequences in this region of p21 and p27 promoters was amplified by PCR using modified DNA as templates. The PCR products were sub-cloned into a TA-cloning vector. 50 clones for each sample was sequenced. The amount of mC relative to global cytidine (5 mC + dC) can be calculated for each sample, and this can be compared between the experimental and control samples.
Global DNA methylation quantification: The global DNA methylation (5-mC) was quantified by MethylFlash Methylated DNA Quantification Kit (Epigentek). Briefly, 50 ng of DNA from CD133+ cells and CD133-cells was used for incubation with both capture and detection antibodies. After washed with PBS for at least three times, the absorbance of the sample was measured in a microplate spectrophotometer at 450 nm (BioTek Instruments, USA). The percentage of the whole genome 5-methylcytosine (5-mC) was calculated according to manufacturer's instructions.
Cell cycle analysis: CD133+ cells expressing the indicated plasmids were rinsed with phosphate buffered saline (pH 7.4) and then collected by centrifugation at 4°C. Pellets were re-suspended in ice cold 70% ethanol and rinsed in PBS for three times. Then, cells were re-suspended in PBS containing 20 μg mL −1 propidium iodide (PI). After washed for three times, Fluorescence was measured using a flow cytometer.
DNA methylation microarray: DNA methylation microarray analysis of human cells was analyzed using the Infinium® MethylationEPIC Bead-Chip(Illumina, San Diego, CA)according to Illumina's instruction. DNA methylation levels ( values) of individual genomic blocks were evaluated using the mean values of all the probes within individual genomic blocks. Initially, probes containing single nucleotide polymorphisms was excluded. Next, the K-nearest neighbor method was used and the mixture quantile amplification method for imputation and normalization, respectively. Subsequently, the threshold value 0.1 of average in each group was filtered, and the P value based on the false discovery rate was 0.001.
Chromatin Immunoprecipitation: Analysis the binding of DNMT1 to p21 promoter region was performed using chromatin immunoprecipitation (ChIP) assay kit (Upstate, 17-295). Briefly, cells were cross-linked by addition of 1% formaldehyde for 15 min at 37°C. After washed with cold PBS, cells were lysed in an SDS lysis buffer (1% SDS, 50 mM Tris at pH 8, 20 mM EDTA). The lysates were sonicated to shear DNA to lengths between 150 and 700 base pairs (bp). After tenfold dilution in ChIP dilution buffer (16.7 mM Tris, 0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 167 mM NaCl), IPs were carried out overnight at 4°C with 2 μg of DNMT1 (Abcam, cat# ab13537) or 2 μg of normal mouse IgG as a negative control. Fifty microliters of protein G beads were added to each sample for 4 h, and the beads were then washed as per the Upstate Biotechnology ChIP protocol. DNA was eluted twice with 100 μL of TE with 1% SDS for 10 min at 65°C. The cross-links were reversed overnight at 65°C. Proteinase K was added for 1 h at 65°C, and then DNA was recovered by phenol extraction and ethanol precipitation. Immunoprecipitated DNA was analyzed for the presence of the p21 promoter by PCR.
Lectin Blot: The immunoprecipitates were subjected to western blot analysis according to the standard procedures. For lectin staining, the PVDF membrane was blocked in 5% BSA in TBS for 4 h at RT. The membrane was washed twice for 10 min with TBS, then once with lectin vehicle (1 mM MgCl 2 , 1 mM MnCl 2 , and 1 mM CaCl 2 in TBS) before 1 h incubation with biotinylated lectin (1:2000, Vector Laboratories). The membrane was washed three times for 10 min each in TBST (1% Tween-20 in TBS), and were then incubated 45 min with Streptavidin-HRP (1:2000, Southern Biotech).
Yeast Two-Hybrid Analysis: The cDNA encoding C-terminal cytoplasmic domain of CD133 (residues 813-865) was cloned into pGBKT7 vector and was used as the bait to screen the pACT2-human cDNA libraries (human fetal brain). Positive interactions were verified by -galactosidase assay.
Purification of Strep-Tagged CD133 Protein: HEK293T cells expressing CD133 and its mutant were treated with N-glycosylation inhibitor Kifunensine. Cells were lysed at 4°C for 2 h using lysis buffer (150 mM NaCl, 100 mM Tris (pH 8.0), 0.5% Triton X-100, 1 mM EDTA, protease inhibitor mixture). The supernatants of cell lysates were incubated with Strep-Tactin agarose at 4°C for 14-18 h. After incubation, the agarose was washed three times in lysis buffer containing 2 M NaCl to eliminate nonspecific proteins. Desthiobiotin (2.5 mM) was used to elute Strep-CD133 proteins. The elution of CD133 protein was concentrated using an ultrafiltration tube. The purified effect of CD133 protein was determined by Coomassie Blue staining.
Measurement of DNMT1 binding to CD133 and its mutants by ELISA. ELISA was performed as previously described. [73] ELISA plate wells were coated with a CD133 antibody (Miltenyi Biotec, Cat # 130-092-395) by incubating 1 μg/100 μL of the antibody per well at 4°C for 12-14 h. After wells were washed with PBST (PBS with 0.05% Tween 20), wells were blocked by incubation with PBS containing 1% bovine serum albumin (BSA) for 2 h at RT. Next, each well was added with serially diluted recombinant human DNMT1-his (Active motif) (final concentration: 0.01, 0.1, 1, 10, 40, 80, 160, 640, 1280, and 2560 nM). Purified strep-tagged human CD133 or CD133 mutant (final concentration: 250 nM) were added to each DNMT1-containing well. After 2 h at 37°C, each well was washed with PBST and incubated with anti-DNMT1 antibody (abcam) for 2 h at RT. After being washed with PBST for 3 times, the wells were examined by horseradish peroxidase-based detection systems. After adequate color development, 100 μl per well of STOP solution was added, followed by absorbance reading at 450 nm by the Microplate Reader from BioTek.
Statistical Analysis: In general, significance was tested by unpaired two-tailed Student's t test using GraphPad InStat 5.0 software. For animals' studies, Kaplan Meier curves and log-rank analysis were performed. P values < 0.05 were considered statistically significant, with *p < 0.05, **p < 0.01, ***p < 0.001, respectively. Results are expressed as the mean ± standard deviation (SD) from at least 3 independent experiments.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.