Analysis of an alternative human CD133 promoter reveals the implication of Ras/ERK pathway in tumor stem-like hallmarks

The human colon carcinoma cell line Caco-2 and synovial sarcoma cell line Fuji were cultured in DMEM supplemented with 20% and 10% fetal bovine serum (FBS), respectively. The immortalized normal human astrocytes NHA/TS and Ras-transformed NHA/TSR cells were previously established [27]. Cells were treated with the MEK inhibitor U0126 (Cell Signaling Technologies, Beverly, MA, USA) at 20 μM for 72 hours with renewal of media every 24 hours. Fugene HD transfection reagent (Roche, Indianapolis, USA) was used for transfection. For primary neurosphere formation, NHA/TS and NHA/TSR cells (1000 cells/3.5 cm dish) were cultured in neural stem cell (NSC) medium (DMEM/F12 containing 20 ng/ml EGF, 20 ng/ml bFGF, 1× B27, and 1% penicillin and streptomycin) for 2 weeks as described previously with minor modification [28]. After 2 weeks of culture, TSR spheres were collected, enzymatically dissociated by Trypsin-EDTA and reseeded at a density of 1000 cells/3.5 cm dish) for secondary sphere formation. For clonal culture, single NHA/TS or NHA/TSR cell was transferred into 48-well plates containing NSC medium using cell sorter. After 2 weeks, each well was manually screened for a colony using phase contrast microscopy. Thereafter, each individual primary TSR sphere was dissociated and reseeded (1000 cells/3.5cm dish) again for secondary sphere formation. For reactivation of CD133 expression, cells were treated with 5 μM 5-Aza-2-deoxycitidine (5-Aza-dC; Sigma, St. Louis, MO, USA) for the initial 48 hours, and then in combination with 500nM Trichostatine A (TSA; Sigma) for an additional 24 hours.

pCR3.1-Uni-CD133 expression plasmid was kindly provided by Dr. Denis Corbeil [29]. pGL3 enh-P1, P2, and P3 were previously established [24]. To generate pGL3enh-P4 and P5 reporter gene constructs, P4 and P5 regions of human CD133 gene were amplified by PCR using genomic DNA isolated from human normal peripheral blood monocytes. Oligonucleotide primers were designed on the basis of DNA sequences {GenBank: AY438641 for P4 and GenBank: AY438640 for P5}. The PCR products were gel-purified, cloned into pCR-Blunt II-TOPO vector (Invitrogen, Carlsbad, CA, USA) verifying sequences, and subcloned into the pGL3 enhancer luciferase reporter plasmid (pGL3enh) (Promega, Madison, WI, USA). The reporter plasmids driven by P5 or its deleted forms (pGL3enh-P5-1068, -768, -368, -98, -25) and mutant plasmids of P5 for ETS binding core sequence (GGAA) such as pGL3enh-P5-98mEBS#1, pGL3enh-P5-98mEBS #2, and pGL3enh-P5-98mEBS #1#2, were constructed by PCR-based methods. The cDNA encoding the dominant negative forms of Ets2 (corresponding to amino acid residues 332-469) and Elk1 (corresponding to amino acid residues 1-168), which lack the transcriptional activation domain, were amplified from human brain cDNA and inserted into XhoI and NotI sites of the pCXN2-Flag expression plasmid [30]. For retrovirus-mediated transfer of Ets2 and CD133 genes into NHA/TS cells, the each cDNA amplified from human brain cDNA and pCR3.1-Uni-CD133 vector was subcloned into pCX4pur retroviral vector kindly provided by Dr. Tsuyoshi Akagi [31].

RNA was isolated with TRI Reagent (Sigma) according to the manufacturer's instructions. After oligo(dT)-primed reverse transcription of 4 μg of total RNA was conducted, the resulting single-stranded cDNA was amplified by PCR using KOD-Plus DNA polymerase (TOYOBO, Osaka, Japan). The specific primers are listed in Additional file 1, Figure S1.

Immunoblotting was performed as described previously [32]. Briefly, cells were lysed and protein concentration of lysates was determined by Protein Assay reagent (Bio-Rad, Richmond, CA, USA). Thirty μg of proteins were separated by SDS-PAGE and transferred to a polyvinylidene difluoride filter, which was subsequently incubated with Tris-HCl-based buffer (pH 7.4) containing 5% dried nonfat milk for 1 hour at room temperature (RT). Filters were probed with mouse monoclonal antibodies to CD133 (1: 200; Miltenyi Biotec, Auburn, CA, USA), Flag M2 (1: 1000; Sigma), and Actin (1: 2500; Chemicon, Temecula, CA, USA), or rabbit polyclonal antibodies to ERK1 (1: 200; Santa Cruz Biotechnology, Santa Cruz, CA, USA) and phospho-p44/p42 MAPK (1: 1000; Cell Signaling). Bound antibodies were detected with peroxidase-labeled goat antibody to mouse IgG or goat antibody to rabbit IgG and visualized by enhanced chemiluminescence reagents (Amersham Pharmacia Biotech, Freiburg, Germany).

Dual-luciferase reporter assay was performed as described previously [33]. Briefly, cells plated on 24-well plates were transiently transfected with both of CD133 promoter-firefly luciferase plasmids and internal control plasmid as pRL-TK (Promega) encoding Renilla luciferase for monitoring transfection efficiency. After 48 hours, luciferase activity was measured, and the ratio of firefly activity normalized with the Renilla activity was calculated for each transfection. In the experiments using Ets dominant negative constructs and U0126, single reporter assay system was conducted to exclude the possibility that pRL-TK control vector could be affected by the change of Ets expression, by normalizing firefly activity to protein concentration of cell lysate for each transfection or treatment.

Nuclear extracts were prepared as described previously [33]. All synthetic oligonucleotides were annealed and end-labeled by T4 polynucleotide kinase (Promega) with [γ-³²P]ATP (Amersham). The following 25 bp oligonucleotides (double strand DNA) were used for experiments: EBS1, 5'-ATCAGGCAGGAA GGGTAGAATGCTG-3' (position -62 to -38 of CD133 P5 promoter); mEBS1, 5'-ATCAGGCATTAA GGGTAGAATGCTG-3' (position -62 to -38 of CD133 P5 promoter with mutated Ets site); EBS2, 5'-TGCTGGGACAGGAA GTAGCTTGGAG-3' (position -42 to -18 of CD133 P5 promoter); mEBS2, 5'-TGCTGGGACATTAA GTAGCTTGGAG-3' (position -42 to -18 of CD133 P5 promoter with mutated Ets site); Cons. EBS, 5'-GGGCTGCTTGAGGAA GTATAAGAAT-3' (Ets responsive element of the human erbB2 promoter [34]); and -25/-1, 5'-GCTTGGAGGTGGGCCTTAGGCTGGT-3' (position -25 to -10 of the CD133 P5 promoter without Ets site). Nuclear extracts containing 10 μg of proteins were incubated for 20 minutes at RT with ~ 100,000 cpm of the labeled probe in a final volume of 20 μl of 10 mM Tris-HCl (pH 7.5) buffer containing 50mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, 5% glycerol, and 1 μg poly(dI-dC)poly(dI-dC). For competition experiments, 50-fold molar excess of the cold oligonucleotide was added to the nuclear extracts 10 minutes prior to addition of the labeled probe. The reacted probes were subjected to electrophoresis (250 V for 90 minutes at 4°C) on a non-denaturing, 5% polyacrylamide gel containing 1xTGE (25 mM Tris-HCl, 190 mM glycine, 1 mM EGTA) buffer.

For enrichment of CD133-expressing cells, Fuji cells were subjected to immunomagnetic separation using a magnetic activated cell sorting (MACS) CD133 Cell Isolation Kit (Miltenyi Biotec), according to the manufacturer's protocol. Briefly, dissociated cells incubated for 30 minutes at 37°C were labeled with CD133/1 Micro Beads. After washing, labeled cells were loaded onto a column installed in a magnetic field. Trapped cells were collected as CD133high fraction after the column was removed from the magnet. The collected cells were applied to a second column and the purification step was repeated. The flow-through of the MACS column is used as CD133low fraction. To check the expression of CD133, cells were stained with phycoerythrin (PE)-conjugated monoclonal antibodies for human CD133 (CD133/2-PE, Miltenyi Biotec) or isotype control antibody IgG2b-PE (Miltenyi Biotec) (1:10 diluted 4°C for 10 minutes in the dark). After washing, the labeled cells were analyzed by a BD FACS Aria flow cytometer (BD Biosciences, San Jose, CA, USA).

To measure the proportion of SP cells, cells were stained with Hoechst 33342 dye (Molecular Probes, Eugene, OR, USA) as previously described [35]. Briefly, cells (1 × 10⁶ cells/ml) were resuspended in pre-warmed DMEM with 2% FBS. Hoechst 33342 was added at a final concentration of 5 μg/ml in the presence or absence of 50 μM verapamil (Sigma) and the cells were incubated at 37°C for 90 minutes. After incubation, cells were washed and resuspended in cold DMEM with 2% FBS. Propidium iodide (Molecular Probes) at a final concentration of 1 μg/ml was added to discriminate dead cells. Analyses were performed on a BD FACS Aria SORP flow cytometer (BD Biosciences). The Hoechst dye was excited with the UV laser and its fluorescence was measured with a 405/20 nm band pass filter (Hoechst blue) and a 670 nm long pass filter (Hoechst red). Cells were pre-gated through FSC-W vs. FSC-H, and SSC-W vs. SSC-H dot plot to exclude doublet cells.

Colony formation assay was performed as described previously [33]. Briefly, cells (4 × 10⁴ cells per 6cm-dish) were suspended in 0.36% noble agar, and plated on 0.6% bacto agar. Colonies were stained with dimethylthiazol-diphenyltetrazolium (MTT), and photographed by Fuji LAS-1000 imaging system. Colony numbers were calculated by using Multigauge Colony Count Software.

Animal procedures were performed according to a protocol approved by the institutional Animal Care and Use Committee at Hokkaido University Graduate School of Medicine. 5 × 10⁶ cells were mixed with matrigel (BD Biosciences) and subcutaneously injected into female nonobese diabetic/severe combined immunodeficiency (NOD/SCID) mice at 6-8 weeks of age. Tumor formation was assessed at 6 weeks after inoculation.

Comparison between experimental groups was made by Student's t-test. p < 0.05 was considered significant. The TFSEARCH program http://​www.​cbrc.​jp/​research/​db/​TFSEARCHJ.​html was used for analysis of possible binding sites of transcription factors.

To analyze the transcriptional machinery of CD133 gene, we used a human colon carcinoma cell line Caco-2 and a human synovial sarcoma cell line Fuji, because both cell lines express abundant CD133 transcripts and proteins, and also CD133⁺ Caco-2 cells are reported to be able to generate tumors following transplantation in mice, whereas CD133^- cells are not [36]. Our previous studies showed that Caco-2 expresses the transcripts containing all variants of exon 1, and Fuji expresses only exons 1A, 1B, 1C, and 1E (Figure 1D). Immunoblotting and RT-PCR analysis confirmed that the expression levels of CD133 mRNA and protein in Caco-2 cells are higher than those in Fuji cells (Figure 1A, Figure 1B). FACS analysis using anti-CD133 antibody revealed that both cell lines exhibit the cellular heterogeneity for CD133 expression (Figure 1C), as the CD133 positive fraction is 88.9% and 18.2% in Caco2 and Fuji cells, respectively. Luciferase assay demonstrated a remarkable activation of P5 (pGL3enh-P5-1368) and a modest activation of P1 (pGL3enh-P1-1182), and no significant enhancement of P2 (pGL3enh-P2-253) and P3 (pGL3enh-P3-199) was detected in both cells (Figure 1E). A modest activation of P4 (pGL3enh-P4-1451) was observed only in Caco-2 cells, which probably contributes to the expression of exon 1D-containing transcripts.

To determine the minimal region essential for P5 promoter activity, we generated a series of deletion mutants of P5 promoter (pGL3enh-P5-1068, -768, -368, -98, and -25) and found that deletion to -98 holds a similar luciferase activity to the full-length promoter (pGL3enh-P5), but the further deletion to -25 leads to a significant reduction of activity (Figure 2A), suggesting that the region between -98 and -25 contains minimal elements required for CD133 transcription through P5 promoter.

The region between -98 to -25 contains two consensus binding motifs of Ets family proteins (G/C)(A/C)GGAAG(G/T) (Additional file 1, Figure S2A) [37]. To determine whether these Ets motifs are necessary for P5 activity, we introduced a substitutional mutation altering two nucleotides of Ets core sequence GGAA to TTAA, into each or both of the Ets binding sites, designated as pGL3enh-P5-98mEBS#1, #2, and #1#2. Introduction of the mutation at EBS#1 moderately decreased P5 promoter activity and the mutation at EBS#2 remarkably decreased the activity (Figure 2B). These results indicated that the two Ets binding motifs are required for the P5 activity, and that in particular, mutation of EBS#2 resulted in the greatest reduction of P5 activity.

Next, to investigate the specific transcription factor which binds to EBS between -98 to -25 of P5, EMSA was performed using nuclear proteins extracted from Caco-2 and Fuji cells. A 25-bp double-stranded oligoDNA containing putative Ets binding sites was synthesized as probes or as unlabeled competitors (Additional file 1, Figure S2B). As shown in Figure 2C, an EBS2 probe, which was designed to span the -42/-18 region containing one proximal Ets motif (EBS#2), showed two major shifted bands following incubation with the nuclear extract from Caco-2 (upper panel, bands a and b), but the EBS1 probe did not. In addition, these signals were eliminated by the use of mutant probes (mEBS2) and by competition using ETS consensus oligonucleotides (Cons. EBS) [34], as well as by 50-fold molar excess amounts of cold wild-type EBS2 probe. -25/-1 oligonucleotides without Ets binding sequence were not able to affect them. These data indicate that the EBS#2, and not EBS#1, is critical for the complex formation on the CD133 P5 region. Interestingly, four complexes were obtained using the nuclear extracts from Fuji under the same conditions (lower panel, bands a-d), in which complexes a and b appeared to have similar mobility with those in Caco-2. These results demonstrated that both tumor cell lines express any factors that can bind to the EBS in CD133 P5 promoter, although there is a subtle difference in their preferred binding sequence.

Ets family proteins share a conserved DNA-binding domain (ETS domain) that recognizes a GGAA sequence. To determine the implication of Ets factors to regulate CD133 transcription, the expression levels of 21 human Ets genes were examined by semi-quantitative RT-PCR (Figure 3A and Additional file 1, Figure S3), in Caco-2 and Fuji which was separated by the expression amount of CD133 by the MACS system. Ets members were found to be expressed in both tumor lines, but the expression of ETS1, ELK3, ER81, ERG, and FLI1 genes were not detected in Caco-2 cells. Comparing the CD133high fraction with CD133low in Fuji cells, it was revealed that ETS1, ETS2, ELF2, ESE1, ELF4, ERG, FLI1, and FEV genes were highly expressed in the CD133-expressing population.

To test whether Ets proteins could specifically mediate P5 transcription, we used a transient transfection approach, employing dominant-negative Ets2 (Ets2DN) and Elk1 (Elk1DN) lacking a transactivation domain. Expression of Ets2DN and Elk1DN resulted in significant inhibition of P5 activity in both cell lines (Figure 3B, Figure 3C). Since the use of dominant negative Ets constructs can broadly interfere with the function of multiple Ets factors [38], studies utilizing dominant negative forms of Ets cannot identify which members of the Ets family are actually important. Therefore, these results indicate that any Ets factors with ETS domain are required for P5 activity. In addition, the partial decrease of P5 activity by EtsDN constructs might reflect the presence of any redundant factors to achieve P5 transcription.

As several Ets factors, including Ets2 and Elk1, have been shown to be activated through phosphorylation by ERK1/2 (also known as p44/42 MAPK), we examined the contribution of the ERK pathway to CD133 gene expression. ERK1/2 were constitutively activated without the external stimuli in both tumor cell lines, and this phosphorylation of ERK1/2 was diminished by the treatment of U0126 (Figure 4A), a potent and specific inhibitor of MEK1/2, which binds to MEK1 and MEK2 regardless of its activation state to inhibit ERK1/2 phosphorylation noncompetitively. Furthermore, to investigate the involvement of MEK/ERK signaling in P5 activity, reporter analysis was performed in the presence or absence of U0126. Treatment of Fuji cells with U0126 led to the marked inhibition of P5 activity (Figure 4B). No cell toxicity was observed concerning morphology and growth of both cell lines under the conditions of this experiment (data not shown). Consistently, U0126 also markedly decreased the expression of CD133 protein in Fuji cells (Figure 4C), indicating that the MEK/ERK signaling is implicated in P5-mediated CD133 expression. In contrast, the same concentration of U0126 could not affect P5 activity and expression of CD133 in Caco-2 cells, suggesting that another pathway could regulate Ets-mediated P5 transcription in Caco-2 cells. However, it remains possible that the ERK pathway might simultaneously regulate other promoters (P1 to P4). In fact, ERK inhibition decreased the expression of exon 1B- and 1D- and 1E-containing CD133 mRNA, but increase exon 1C-containing one (Additional file 1, Figure S4). These data indicate cell type-specific regulation of CD133 gene expression by the ERK pathway.

To assure the relevance of the ERK pathway to stem-like characteristics, the effect of U0126 on the amount of SP was assessed by flow cytometric analysis. The SP fraction in tumor cells has been known to define the population containing stem-like cells, which highly express ATP-binding cassette (ABC) transporters to efflux both Hoechst dye and chemotherapeutic agents, and to have a high capacity to form tumor xenografts in mice [39]. In our experiments, the side population represented approximately 2% in the Caco-2 cell line (Figure 4D, Figure 4E). Treatment with verapamil, an inhibitor of the ABC transporters, completely ablated this population. In contrast, no distinct SP was visible in the Fuji cell line (data not shown). Treatment of Caco-2 with U0126 dramatically reduced the SP frequency to 0.15% (Figure 4D, Figure 4E). This result emphasized our conclusion that ERK is a key molecule in the signal transduction to maintain stem-like features in tumor cells.

To examine whether Ets factor could increase CD133 expression and confer tumorigenicity in normal cells, we established the immortalized human astrocytes overexpressing Ets2 (NHA/TSE2). The increase of CD133 mRNA expression was detected in NHA/TSE2 comparing to NHA/TS cells and its effect was strongly enhanced by the treatment with demethylating agent 5-Aza-dC and histone deacetyltransferase inhibitor TSA (Figure 5A). 5Aza-dC/TSA treatment has been shown to open the chromosomal region to increase accessibility for transcription factor complexes to assemble at the promoter and drive gene transcription. However, the elevated protein level of CD133 could not be detected by FACS analysis (Additional file 1, Figure S5). Instead, approximately 0.04% of side population diminished by verapamil was observed in the NHA/TSE2 cells, whereas substantial SP was not visible in NHA/TS cells (Additional file 1, Figure S6). To assess the potential of Ets2 on tumorigenicity, NHA/TSE2 cells were subcutaneously implanted into NOD/SCID mice. All positive control mice implanted with 5 × 10⁶ NHA/TSR cells, which is a transformed astrocytes [27], developed tumors at 4-6 weeks with 100% incidence (4/4), whereas any of the mice injected with NHA/TS or NHA/TSE2 cells could not grew as tumors at least by 6 weeks (0/4) (Figure 5B). These data indicate that Ets2 is not sufficient to fully transform normal human astrocytes, and thus the acquisition of tumor stem-like characters may be accomplished by synergistic actions of multiple factors.

NHA/TSR cells form malignant tumors with features of human anaplastic astrocytoma, the WHO classification Grade III in mice [27] and malignant astrocytoma has been well-known to display a stem cell signature [2]. To confirm whether Ras-mediated transformation in human astrocytes could confer stem-like features in parallel with the acquisition of tumorigenicity, we examined the expression level of CD133 and the activity of neurosphere formation in NHA/TSR cells. In the absence of 5Aza-dC/TSA, the more substantial level of CD133 mRNA was detected in NHA/TSR cells comparing to NHA/TSE2 cells (Figure 5A), suggesting that Ras might play an additional role to modulate chromatin structure of CD133 promoter region. Actually, CD133 expression is regulated by promoter DNA methylation [24]. However, the protein level of CD133 could not be increased (Additional file 1, Figures S5 and S7), suggesting that Ras activation is also not sufficient to completely restore CD133 expression in human astrocytes. The generation of neurospheres is considered as an important symbol of in vitro neural stem cell culture, as well as an exhibition of self-renewal capacity [40]. After culturing in NSC medium containing bFGF and EGF for 2 weeks, NHA/TSR cells generated significantly greater neurosphere-like colonies (diameter is ≥ 50 μm) than NHA/TS cells, and this capacity was maintained to generate secondary neurospheres (Figure 5C). To except the possibility of cellular aggregates, single NHA/TSR cell was also sorted using flow cytometry and deposited into 48-well plates containing NSC medium. NHA/TSR had a forming efficiency of 12.7 ± 1.5 (26.4%) single spheres per 48-well plate, as opposed to zero for NHA/TS (Figure 5D). A single sphere could also maintain the capacity to form secondary spheres, as well as differentiate into adherent cells morphologically identical to original NHA/TSR cells in 10% FBS-containing medium (Figure 5E), indicating that constitutive activation of the Ras pathway can trigger the development of stem-like population with self-renewing ability to retain stemness, and possibly that human astrocytes could dedifferentiate to premature state following to the Ras-mediated transformation and/or its sequential combination with immortalization. Side population size was not increased in NHA/TSR cells (Additional file 1, Figure S6). These results indicate that the expression of CD133 protein and appearance of side population are not necessarily required for other tumor stem-like behaviors, at least for the capacity to form neurospheres in vitro and tumors in vivo. Finally, NHA/TS cells retrovirally overexpressed CD133 gene (NHA/TSC) could form a larger number of colonies comparing to NHA/TS cells in soft-agar colony formation assay (Additional file 1, Figure S7). However, their number and size are much less than those of NHA/TSR even after they were cultured for 3 weeks. In addition, mice subcutaneously implanted with 5 × 10⁶ NHA/TSC cells could not develop any tumors at least by 6 weeks (Figure 5B), supporting the view that CD133 is just a concomitant marker for tumorigenic process.