Histone methyltransferase G9a and H3K9 dimethylation inhibit the self-renewal of glioma cancer stem cells

Abstract Epigenetic modification is crucial to keep the self-renewal and the ‘‘stemness’’ states of stem cells, not letting them to differentiate. The actual roles of Histone 3 Lysine 9 dimethylation (H3K9me2) and its methyltrans- ferase G9a in this process are still unclear, especially in cancer stem cells. In our study, we found an interesting observation that most CD133-positive cells were H3K9me2 negative, both in glioma tissues and in cultured cells, although most cancer cells were detected to be H3K9me2 immunopositive. This implied that the G9a- dependent H3K9me2 was one of the crucial barriers of cancer stem cell self-renewal. To test the hypothesis, we examined the loss-of-function and gain-of-function of G9a. We found that bix01294, the selective inhibitor of G9a, can stimulate the sphere formation rate of glioma cancer stem cells, together with increasing Sox2 and CD133
expressions. The increase of CD133-active stem cells was confirmed by flow cytometry. On the other aspect, over- expression of G9a increased the H3K9me2 and decreased the sphere formation rate as well as the CD133 and Sox2 expressions. Since H3K9me2 modification is the major repressive switch, we predict that the repressive H3K9me2 modification may happen at the CD133 promoter regions. By chromatin precipitation assay, we confirmed that the CD133 and Sox2 promoter regions were modified by the H3K9me2. Therefore, we concluded that the G9a-depen- dent H3K9me2 repression on CD133 and Sox2 was one of the main switches of the self-renewal in glioma cancer stem cells.

Keywords Cancer stem cells · H3K9me2 · G9a ·Self-renewal

Introduction

It requires maintaining the stem cells in their ‘‘stemness’’ state and in their proliferation instead of differentiation. Increasing evidence suggests that the epigenetic mecha- nism is crucial to the maintenance of the ‘‘stemness’’ state. Histone modification appears to be a great barrier between the stem cell and the differentiated cells [1]. Histone modifications can lead to either gene activation or repres- sion, depending upon which residues are modified and the types of modification. H3K9 and H3K27 methylations constitute the two main repressive mechanisms in mam- malian cells. H3K27 has gained significant attention in embryonic stem (ES) cells [2], but the role of H3K9 methylation (especially H3K9me2) in maintaining the ‘‘stemness’’ of cancer stem cells is still unclear.

G9a (histone-lysine N-methyltransferase, H3 lysine-9 specific 3) mediates the histone H3K9 monomethylation and dimethylation (H3K9me1/2) and targets euchromatic loci [3], which are essential for early embryogenesis [4]. A recent genome-wide study of histone H3K9me2 demon- strated that large K9-modified regions were found in dif- ferentiated tissues but little in undifferentiated ES cells. These modifications require G9a [5]. Reduced H3K9me2 of chromatin modifications was crucial to increase stem- like cells in breast cancer, through epithelial-to-mesench- ymal transition process [6].
Our study starts from an interesting observation that most CD133-positive cells were H3K9me2 negative, both in gli- oma tissues and in cultured cells, although most cancer cells were detected to be H3K9me2 immunopositive. This implied that the G9a-dependent H3K9me2 modification was one of the crucial switches for the self-renewal of cancer stem cells, similar to the ES cells. To further test this hypothesis, we examined both the loss-of-function and gain-of-function aspects of G9a, using the G9a inhibitor bix01294 and the G9a overexpression vectors in glioma cancer stem cells.

Materials and methods

Culture of glioma tumor spheres

Glioma cancer stem cells are characterized by the expres- sion of neural stem cell antigens and the ability to grow as non-adherent three-dimensional aggregates, forming ‘‘tumor spheres’’ when cultured in the presence of epider- mal growth factor (EGF) and fibroblast growth factor (FGF) under serum-free conditions [7]. The glioma cells from U251 and U87 cell lines were cultured under the ‘‘stem-like’’ conditions, to let spheres formation. The composition of serum-free medium is as follows: modified Eagle’s medium/F12 (Gibco, Grand Island, NY, http:// www.invitrogen.com) with L-glutamine, B-27 supplement (Gibco, Grand Island, NY, http://www.invitrogen.com), 20 ng/ml bFGF (Invitrogen, Carlsbad, CA, http://www. invitrogen.com), and 20 ng/ml EGF (Invitrogen).

Glioma sphere formation assays and soft agar colony formation assays

Self-renewal of the cancer stem cells was examined by the tumor sphere formation assays. For tumor sphere formation assays, cells were seeded at the density of 100 cells per well into a 96-well plate with 100 ll of serum-free medium described above. Bix01294, (tri-hydrochloride hydrate [2-(hexahydro-4- methyl-1H-1,4-diazepin-1-yl)-6,7-di-methoxy-N-[1-(phenyl- methyl)-4-piperidinyl]-4-quinazolinamine tri-hydrochloride hydrate] (1 mM/ml)), (Sigma-Aldrich, http://www.sigma aldrich.com/china-mainland.html), was added at the time of seeding. The number of newly formed tumor spheres per well was counted microscopically 3–5 days later. The experiments were repeated with 4–6 wells for each time point.

Proliferation assay

The proliferative activity was determined by the MTT method [3-(4,5)-dimethylthiahiazo(-z-y1)-3,5-di-phenyte- trazoliumromide] according to the manufacturer’s instruc- tions. Cells were seeded into 96-well tissue culture plates at the density of 10,000 cells per well in triplicate for 24 h with different concentrations of inhibitors. The optical density was measured at 490 nm after 4-h incubation with MTT reagent.

Plasmid and transfection

Adherent cells were cultured routinely. The G9a overex- pression plasmid pLenti6-MK1-EHMT2-V5 (addgene, http://www.addgene.org/) was transfected with lipofect- amine 2000 as directed.

Chromatin immunoprecipitation assay (ChIP)

ChIP assay was performed using antibodies against the histone H3K9 dimethylation (H3K9me2) (Anti-Histone H3 dimethyl K9, Ab-cam Inc. USA) and the EZ-ChIP kit (Upstate, www.millepore.com). DNA fragments that were modified by H3K9me2 were expected to be pulled down with anti-H3K9me2 antibody. With the precipitated geno- mic DNA, PCR and quantitative PCR were performed to detect the targeted promoter sequences. Primers of the promoter regions of CD133, Sox2, and Sox2 distal enhancer were chosen according to the previous reports [8– 10]. Total genome DNA was used as the input. All the samples were tested in triplicate, and the results from each sample were normalized relative to the input.

The primer sets are as follows: CD133 promoter for ChIP, Forward: 50-CCAGAAGCC GGGTCATAAATAAT; Reverse: 50-AGCGAACCCGTCCACTCCTCACT; Sox2 promoter for ChIP, Forward: 50-AGTTGGACAGGGAGATGGC-30; Reverse: 50-AACCTTCCTTGCTT CCACG-30; and Sox2 distal enhancer for ChIP, Forward: 50-GCTGGATTGAAATAGAG TGGGC-30; Reverse: 50-AAAGAAGATGGTGATA GCAGGAAC-30.

Glioma human brain tissues

Frozen surgically removed human glioma tissues (n = 5) and paraffin-preserved specimens (n = 9) were obtained from the General Hospital of Ningxia Medical university (Yinchuan, China). These specimens were collected from the patients registered at the above-mentioned hospital, and written informed consent was obtained from the patients. This study was approved by the ethics committee of the hospital.

Immunohistochemistry and immunofluorescence

Histopathological, immunohistochemical, and immunoflu- orescent analyses were performed routinely with clinical glioma samples, fresh or paraffin-preserved. Primary anti- bodies for CD133 (1:100), Sox2 (1:100), and H3K9me2 (1:300; Protein Tech Group, Inc, Chicago, IL, http://www. ptglab.com) were used.

Western blot

For Western blotting, the cell spheres were collected by centrifugation and the total protein was isolated using the KEYGEN total protein extraction kit (KEYGEN, Nanjing, China, http://www.keygentec.com.cn). Antibodies used were as follows: CD133 (1:1,000), G9a (1:1,000), Sox2 (1:1,000), OCT4 (1:1,000), H3K9me2 (1:1,000), H3 (1:1,000) (Protein Tech Group, Inc, Chicago, IL, http:// www.ptglab.com), and b-actin (1:5,000) (BIOSS, Beijing, China, http://www.bioss.com.cn). Peroxidase-conjugated goat anti-rabbit IgG was used as the secondary antibody (ZSGB-BIO, Beijing, China, http://www.zsbio.com).

Flow cytometry

Tumor spheres were collected and pipetted into single cells. Single-cell suspensions were stained with antibodies against human CD133 (1:10, PE-conjugated, Miltenyi Biotec) as directed. Typically 1 million cells were stained in an Eppendorf tube. Antibodies were added and incu- bated in the dark for 15–60 min at room temperature. Cells were then centrifuged and washed three times with PBS. Cell suspension was filtered through a 400-mesh screen before analysis. Flow cytometry analysis was carried out on an FACS Caliber (BD, http://www.bd.com/). Matched isotype antibodies were applied in parallel as negative controls. Dead cells were excluded using forward and side scatter gating.

Results

Most CD133-positive cancer stem cells were less H3K9 dimethylated in glioma tissues

H3K9me2 is one of the crucial repressive mechanisms in differentiated cells. So, we first examined the H3K9me2 expression in glioma tissues of WHO III–IV grade patients.

As expected, most samples (8/9) were H3K9me2 immu- nopositive. The positive rate ranged from 35 to 48 %, and the average rate was 40 %. However, there were no sig- nificant differences between the primary and recurrent tissues in terms of the immunopositive rates of H3K9 dimethylation (patient information in the supplementary data).

Unexpectedly, we found that most of these glioma tissue samples were CD133 negative (1/9) (Fig. 1a). Considering the above results that most of these tissues were highly H3K9me2 methylated, we predicted that CD133 may not coexist with the H3K9me2. To test this hypothesis, we analyzed the freshly obtained glioma tissue samples of III– IV grade astrocytoma, by Western blot and double immunostaining of CD133 and H3K9me2.

We used the CD133-positive samples to perform the double staining by Immunofluorescence. We found that most CD133-positive cancer stem cells were
H3K9me2 negative. On the other aspect, most cancer cells were H3K9me2 positive but CD133 negative (Fig. 1b).

As shown by the Western blot in Fig. 1c, most tissue samples showed high H3K9me2 expression but with low or no CD133 expression. The CD133-positive samples showed low H3K9me2 expression, while the G9a expres- sion was lower, too. The U251 cell line was CD133 posi- tive and showed low H3K9me2 and G9a expressions, as expected. We then repeated the double immunostaining of CD133 and H3K9me2 in cultured glioma cell lines and discovered similar results. In adherent cultured U251 cells, only few CD133-positive cells were H3K9me2 negative.

This interesting observation suggested that G9a and H3K9me2 played the key role in blocking the self-renewal of cancer stem cells. Inhibition of G9a and H3K9me2 may be very important for maintaining the self-renewal of cancer stem cells and making them stay at the ‘‘stemness’’ state.

G9a inhibited the self-renewal of glioma stem cells

We chose the specific inhibitor of G9a, bix01294, to test the loss-of-function effects of G9a and H3K9me2. First, we tested the toxicity effect of bix01294 by the MTT prolif- eration assay. We found that bix01294 had no apparent cell toxicity effects when its final concentration was lower than 10 lM (supplementary data). So, we chose 10 lM as the working concentration in the following experiments.

To analyze the self-renewal of glioma stem cells, we used the sphere formation test in serum-free medium in 96 culture plates, by treating cells with bix01294 to inhibit the G9a. As shown in Fig. 2a, b, we found that bix01294 promoted the sphere formation rate of U251 (as well as U87) cells cultured in stem cell media. In bix01294-treated cells, spheres formed very well, with a round, larger, and most energetic appearance under microscope (Fig. 2a).

Fig. 1 Results of glioma tissues. a Immunohistochemistry results of CD133 and H3K9me2 using paraffin-preserved specimens from two patients, a CD133 immunopositive and a CD133 negative. b Immunofluorescent double staining of fresh glioma tissues. Green for CD133; red for H3K9me2. Results of two patients are shown here: a CD133 immunopositive and a CD133 negative. Arrows for the CD133-positive but H3K9me2- negative cells. c Western blot results of fresh glioma tissues from five patients. (Color figure online).

The sphere formation effect of bix01294 was confirmed to be true in large-scale culture systems (with 76-cm2 culture bottle or 10-cm culture dish). Bix01294 can pro- mote the sphere formation with a very stable repeatability both in U251 and U87 cell lines.The gain-of-function of G9a was tested by over- expressing the G9a in these cells. As expected, we found that overexpression of G9a caused the almost complete blocking of sphere formation. Few spheres were found after transfection of G9a plasmid (Fig. 2a, b).

G9a decreased the CD133 and Sox2 expressions

In order to test the elevated amount of cancer stem cells in the spheres formed above, we examined the expression of cancer stem cell marker CD133 and the neural stem cell markers Sox2 and Oct4. As expected, the expression of CD133 was up-regulated in the bix01294-treated spheres. Bix01294 showed an increasing effect on the expressions of CD133 and Sox2. Transfection of G9a overexpression plasmid showed a decreasing effect, both on CD133 and Sox2 (Fig. 2c).

It was surprising that another core transcription factor Oct4 did not change in all the treatments. In contrast to the previous reports that bix01294 could be used as a substitute of Oct4 in induced pluripotent stem cell [11], neither inhibition nor overexpression of G9a changed the expres- sion of Oct4 (Fig. 2c). This result implied that Oct4 may not play crucial roles in the glioma cancer stem cell self- renewal regulation.

Fig. 2 Influence of G9a inhibition or overexpression on the self-renewal of cancer stem cells and on the expressions of CD133 and Sox2. Self-renewal was tested by the tumorsphere formation assays. Cells were cultured in serum-free medium with EGF and bFGF. Bix01294 (bix) was added into the media at the beginning. G9a plasmid was at first transfected into adherent cells. Then these cells were seeded into 96-well plates, with five wells at each point for parallel. After 3–5 days, the number of spheres formed in each well was counted, and the sphere formation rate was calculated. The expressions of G9a and H3K9me2, as well as the cancer stem cell markers, CD133, Sox2, and Oct4, were further confirmed by Western blot. The rescue experiment was performed in the U87 cells. G9a was overexpressed by its expressive plasmid transfection (the top panel G9a). After 24 h transfection, bix01294 (10 lM) was administered into the media of the rescue group (the top panel G9a?bix). Empty plasmid was transfected as control. 48 h later, cells were harvested for the Western blot. H3K9me2, G9a, CD133, Sox2, Oct4, and b- actin were tested as shown here. The expressions of CD133 and Sox2 were inhibited by the G9a overexpression. After the administration of bix01294, the inhibitor of G9a, the expressions of CD133 and Sox2 were partly rescued back. a Photos of the spheres formed in each group. b Sphere formation rate counted. c Western blot results of the G9a and bix01294 treatment. d Rescue experiment results of the G9a effects by bix01294

A rescue experiment was performed to test the specifi- cation of the bix01294. G9a was overexpressed by the plasmid transfection. Bix01294 was then added into the medium after 24 h transfection. Cells were harvested for Western blot analysis. As shown in Fig. 2d, both the CD133 and Sox2 inhibitions by G9a were partly rescued back.

To further confirm the increased number of CD133- active glioma stem cells, we counted the CD133-immu- nopositive cells by flow cytometry with PE-conjugated CD133 antibody (Fig. 3). The number of CD133-positive stem cells was increased after treatment by bix01294, compared with the control (from 31 to 51 %, 29 % increase). We had not perform this test with G9a overex- pression because of the difficulty of sphere formation in this group.

Fig. 3 Flow cytometry results. Cells were cultured into spheres and then were pipetted into single cells for the flow cytometry. The number of CD133-active cancer stem cells was counted by flow cytometry with PE-conjugated CD133 antibodies. The percent of CD133-active cells was increased from 31 % of control to 51 % of bix01294 treatment.

The promoters of CD133 and Sox2 were repressed by H3K9me2 modification

We speculated that the promoters of both CD133 and Sox2 were modified by H3K9me2. Overexpression of G9a would result in the direct repression of CD133 and Sox2 tran- scription at their promoter regions by H3K9me2 modifi- cation. The self-renewal-increasing effect of bix01294 may be directly through the blocking or removal of H3K9me2 inhibition on the CD133 and Sox2 promoters. So, we designed the chromatin immunoprecipitation (ChIP) experiment to further test this opinion.

First, we found that primers for the promoter regions of CD133/Sox2/Sox2-enhancer showed a positive band in PCR results, which means that all three of them were modified by H3K9me2 modification.Second, we found that all three of the amplified bands of CD133, Sox2 promoter, and Sox2 enhancer decreased in the bix01294-treated group, of CD133, Sox2-promoter, and Sox2-enhancer, compared with the control (Fig. 4). This was because bix01294 treatment could decrease the H3K9me2 modification. The anti-H3K9me2 antibody would pull down less DNA fragments. Then the PCR band of amplified products would be less, compared with the untreated control.Third, we found that the decreasing rate of the Sox2 promoter and enhancer was stronger than that of CD133, coincident with the gentle up-regulation of proteins shown in the above Western blot analysis. The results of q-PCR were then further confirmed by PCR and agarose gel (Fig. 4b).

Discussion

Cancer stem cells may have similar regulatory mechanisms on the maintenance of self-renewal and ‘‘stemness.’’ CD133-positive cancer stem cells may thus have similar histone modification characteristics with the ES cells or neural stem cells. ES cells possess ‘‘bivalent domains’’ that contain coexisting active (H3K4me3) and repressive (H3K27me3) marks at promoters of developmentally important genes. Differentiated cells lose this bivalency and acquire a more rigid chromatin structure, the repressive H3K9me2 mark, termed ‘‘LOCKs’’ (large organized chromatin K9 modifications). Recent report shows that in adult human hematopoietic stem and progenitor cells, H3K9me2 chromatin territories are absent in primitive cells and are formed de novo during lineage commitment [12]. In our studies, we found that G9a and H3K9me2 act as the major barriers of the self-renewal regulation in cancer stem cells. On other aspects, G9a and H3K9me2 stop the self-renewal process and then let the cancer stem cells step out of the ‘‘stemness’’ state, similar to the normal stem cells.

Fig. 4 ChIP (Chromatin immunoprecipitation) results show that Bix01294 inhibits the H3K9 dimethylation at the promoter regions of CD133 and Sox2 and at the Sox2 distal enhancer regions. Cells were collected for immunoprecipitation using the anti-H3K9me2 antibody. The precipitated DNA was then used as the model for PCR and Q-PCR, with specific primers for the promoter regions of CD133 and Sox2, as well as the Sox2 distal enhancer regions. Unprecipitated DNA was used as the positive control (input) and the normal IgG- precipitated DNA was the negative control (IgG). Data were counted as the fold of input.

G9a-dependent H3K9me2 modification on the CD133 promoter may be one of the direct repressive switches for the self-renewal of cancer stem cells. Transcription of CD133 was known to be controlled separately by five promoters (P1, P2, P3, P4, and P5) [13]. Many researchers reported that the methylation status of CpG sites residing in P1 and P2 regions was inversely correlated with the expression level of CD133 mRNA in human glioma tissues [14]. Our ChIP results showed that the CD133 P1 promoter was inhibited by the H3K9 dimethylation. Data from our clinical samples supported the hypothesis, too. According to our double-staining experiments, most CD133- and H3K9me2-immunopositive cells were not double positive. Sox2 is one of the core transcription factors in stem cells and the major marker in neural-lineage progenitor cells, which plays a critical role in maintaining the self-renewal of stem cells [15]. Previous reports showed that Sox2 inhibition could decrease the self-renewal of glioma stem cells [16] (Gangemi RM, 2009). The continuous expression of Sox2 increased the sphere formation of glioma cancer stem cells (supplementary data). Self-renewal of breast cancer stem cells was observed depending on the activation of distal enhancer of Sox2 [10]. Our ChIP data demon- strated that the promoter and distal enhancer regions of Sox2 were both repressed by the H3K9me2 modification. Overexpression of G9a enhanced the H3K9me2 repression both on the promoter and on the enhancer of Sox2. It made Sox2 to be silenced by the H3K9me2 modification and then the self-renewal was stopped.

In normal stem cells, G9a-dependent H3K9me2 modi- fication and the DNMT-dependent methylation worked together to repress most of the self-renewal-related genes [17]. Recent studies showed that the acetylating modifying reagents trichostatin A (TSA) and 5Az induced the expression of CD133 in neural-sphere cells [18]. We have compared the sphere formation effects of three inhibitors: bix01294; 5Az; and the HDAC inhibitor, sodium butyrate. We found that only bix01294 treatment had the strongest and most stable effects on promoting the sphere formation (supplementary data). The inhibitor of DNMTs 5Az also enhanced the CD133 expression but not of the sphere formation, partly due to its cell toxicity effects. Bix01294 has a mild cell toxicity effect only at high doses (supple- mentary data). All these evidences supported that G9a- dependent H3K9me2 repression was one of the major switch for the self-renewal in glioma cancer stem cells.

Conclusions

Histone-transferase G9a and H3K9 dimethylation play as the major switch to regulate the self-renewal of glioma stem cells, through the direct repression at BIX 01294 the promoters of CD133 and Sox2.