Epigenetic regulation of cancer stem cell marker CD133 by transforming growth factor-β


  • Hanning You,

    1. Department of Pediatrics and Pharmacology, Pennsylvania State University College of Medicine, Penn State Children's Hospital, Hershey, PA
    Search for more papers by this author
  • Wei Ding,

    1. Department of Pediatrics and Pharmacology, Pennsylvania State University College of Medicine, Penn State Children's Hospital, Hershey, PA
    Search for more papers by this author
  • C. Bart Rountree

    Corresponding author
    1. Department of Pediatrics and Pharmacology, Pennsylvania State University College of Medicine, Penn State Children's Hospital, Hershey, PA
    • Department of Pediatrics and Pharmacology, Pennsylvania State University College of Medicine, 500 University Drive, H085, Hershey, PA 17033
    Search for more papers by this author
    • fax: 717-531-0653

  • Potential conflict of interest: Nothing to report.


Hepatocellular carcinoma (HCC) is the third leading cause of cancer mortality worldwide. CD133, a transmembrane glycoprotein, is an important cell surface marker for both stem cells and cancer stem cells in various tissues including liver. CD133 expression has been recently linked to poor prognosis in HCC patients. CD133+ liver cancer cells are characterized by resistance to chemotherapy, self-renewal, multilineage potential, increased colony formation, and in vivo cancer initiation at limited dilution. Recent studies demonstrate that CD133 expression is regulated by DNA methylation. In this study, we explored the role of transforming growth factor β (TGFβ), a multifunctional cytokine that plays a critical role in chronic liver injury, in the regulation of CD133 expression. TGFβ1 is capable of up-regulating CD133 expression specifically within the Huh7 HCC cell line in a time- and dose-dependent manner. Most important, TGFβ1-induced CD133+ Huh7 cells demonstrate increased tumor initiation in vivo. Forced expression of inhibitory Smads, including Smad6 and Smad7, attenuated TGFβ1-induced CD133 expression. Within CD133− Huh7 cells, TGFβ1 stimulation inhibited the expression of DNA methyltransferases (DNMT) 1 and DNMT3β, which are critical in the maintenance of regional DNA methylation, and global DNMT activity in CD133− Huh7 cells was inhibited by TGFβ1. DNMT3β inhibition by TGFβ1 was partially rescued with overexpression of inhibitory Smads. Lastly, TGFβ1 treatment led to significant demethylation in CD133 promoter-1 in CD133− Huh7 cells. Conclusion: TGFβ1 is able to regulate CD133 expression through inhibition of DNMT1 and DNMT3β expression and subsequent demethylation of promoter-1. TGFβ1-induced CD133+ Huh7 cells are tumorigenic. The mechanism by which TGFβ induces CD133 expression is partially dependent on the Smads pathway. HEPATOLOGY 2010

CD133 (AC133 or prominin 1) is a pentaspan, transmembrane protein that was first identified in mouse neuroepithelial stem cells1 and later described in human hematopoietic stem cells.2 Although its exact biological function remains unclear, CD133 is considered a putative stem cell marker in diverse hematopoietic and nonhematopoietic tissues and cancers.3, 4 A series of recent publications demonstrated that CD133+ cancer cells possess many stem cell characteristics, including those associated with liver,3, 4 pancreas,5 colon,6 ovary,7 brain,8 and skin.9 We have recently demonstrated that CD133 reliably identifies liver cancer stem cells (CSCs) in two independent murine models of chronic injury.10-12 In the human HCC cell line Huh7, CD133+ cells demonstrated many stem cell-like properties including colony formation, self-renewal and differentiation ability, as well as a greater ability to initiate tumors in vivo compared to CD133− cells.3

Despite the high volume of recent publications related to CD133, little is known about the regulation of this important stem cell and CSC marker. In glioblastoma cells, CD133 expression is regulated by cellular stress and hypoxia.13 In terms of treatment applications, down-regulating CD133 expression impairs cell proliferation and metastasis in melanoma.14

With regard to liver cancer, HCC is the third leading cause of cancer mortality in the world.15 Current curative treatments such as surgical resection and transplant are limited to the early disease stage. Chemotherapy has generally not improved overall mortality in HCC except for a recent report using sorafenib, which improved advance stage mortality by less than 3 months.16

During chronic liver injury, transforming growth factor β (TGFβ) plays an important role in fibrosis progression. TGFβ is a pluripotent cytokine that is capable of exerting its biological effects on tissue and organ development, cellular proliferation, differentiation, survival, apoptosis, and fibrosis. In the liver, TGFβ is hypothesized to serve as an important link between chronic injury, cirrhosis, and HCC.17 Although TGFβ is able to initiate and drive fibrosis by inducing extracellular matrix synthesis in chronic liver diseases, the exact role of TGFβ in liver cancer initiation and progression is still unclear. Previous reports indicate that TGFβ expression is decreased in early-stage HCC and increased in late-stage HCC.18, 19 A more recent report indicated that dysregulation of the TGFβ pathway leads to HCC through disruption of normal liver stem cell development.20

Aberrant DNA methylation is an event that is common to many human cancers.21, 22 In the liver, there is currently no defined relationship between DNA methylation patterns and etiologic agents such as hepatitis B and C virus (HBV, HCV). In colon cancer, de novo CpG island hypomethylation has been linked to down-regulated DNA methyltransferase (DNMT1 and DNMT3β).23 In our investigations of murine liver injury and CD133+ CSCs, we have previously noted that in a liver-specific hypomethylation model (methionine adenosyltransferase 1A-deficient mice) the level of CD133+ oval cells is higher compared to other models of liver injury.11, 12, 24

Based on the potential role of TGFβ in liver cancer progression and the importance of CD133 expression in liver CSC populations, the goal of this study was to explore the mechanisms by which TGFβ may regulate CD133 expression. Using Huh7 HCC cells we demonstrated that CD133 expression was up-regulated by TGFβ1 stimulation in a time- and dose-dependent manner. Furthermore, TGFβ1-induced CD133 expression was attenuated by enforced expression of inhibitory Smads. In addition, both DNMT1 and DNMT3β expression were inhibited by TGFβ1, and TGFβ1 stimulation resulted in significant demethylation of the CD133 promoter-1. Most important, TGFβ1-induced CD133+ Huh7 cells demonstrated a significant increase in tumor initiation capacity compared to CD133− cells in vivo. Taken together, our novel findings proposed a new mechanism by which TGFβ regulates expression of CD133 by way of epigenetic events.


CSC, cancer stem cells; DAC, 5-aza-2'-deoxycytidine; DNMT, DNA methyltransferase; FACS, flow cytometry; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; mRNA, messenger RNA; PCR, polymerase chain reaction; qPCR, quantitative PCR; RT-PCR, reverse-transcription polymerase chain reaction; TβR, TGFβ receptor; TGF, transforming growth factor; TSA, trichostatin A.

Materials and Methods

Cell Culture.

Huh7 cells were kindly provided by Dr. Jianming Hu (Penn State College of Medicine) and cultured as described.25

Animal Care.

Nude mice (Jackson Laboratory, Bar Harbor, ME) were fed ad libitum a standard diet (Harlan Teklad irradiated mouse diet 7912, Madison, WI) and housed in a temperature-controlled animal facility with a 12/12-hour light/dark cycle. All procedures were in compliance with our institution's guidelines for the use of laboratory animals and approved by the Institutional Animal Care and Use Committee.

Flow Cytometry (FACS) Analysis.

FACS experiments were performed as described.12 Briefly, one million Huh7 cells were incubated with mouse antihuman CD133/2-PE (Miltenyi Biotec, Auburn, CA). Analysis was performed using a FACS Calibur (BD Biosciences, Falcon Lakes, NJ). Analysis was done using the Flow-Jo program (Tree Star, Ashland, OR). Positive and negative gates were determined using immunoglobulin G (IgG)-stained and unstained controls.

DNA Constructs.

pCS2-Smad6 (Plasmid 14960), pCMV5-Smad7-HA (Plasmid 11733) were provided by Addgene (Cambridge, MA). Human CD133 promoter-1 driven luciferase reporter vectors were generated according to the published procedure.26 Briefly, human CD133 promoter-1 (−1100/+10) DNA fragments were amplified through polymerase chain reaction (PCR) and subcloned into pGL3-firefly enhancer luciferase vector (Promega, Madison, WI). The vectors were amplified in competitive cells, purified by Wizard Plus SV Minipreps DNA Purification System (Promega), and verified by DNA sequencing.

CD133+ and CD133− Huh7 Cell Isolation.

The Miltenyi MACS system was used per the manufacturer's protocol as described.10


Cell lysates were harvested and analyzed as described.10


Trizol reagent (Invitrogen, Carlsbad, CA) was used to isolate total RNA from cells according to the user's manual provided by the manufacturer as described.10 Standard reverse-transcription PCR (RT-PCR) was performed using primers and conditions listed in the Supporting Information Table. Quantitative PCR (qPCR) experiments were performed using an ABI-Prism 7700 Thermal Cycler and Taqman Universal PCR Master Mix (Applied Biosystems, Foster City, CA). Human gene CD133, DNMT1, DNMT3α, and DNMT3β was measured using primer/probe sets (Applied Biosystems), respectively, and relative gene expression levels were calculated by normalization to human GAPDH. Quantitation was performed with SDS (Cary, NC) 2.2.2 software using the 2(−ΔΔCt) equation.

Tumorigenicity Assay.

Cells were counted with trypan blue exclusion and then resuspended in 1× phosphate-buffered saline (PBS) for transplant at a concentration of 3 × 105 cells/100 μL mixed with Matrigel at a ratio of 1:1; 3 × 105 cells were inoculated into 6-week-old nude mice subcutaneously. Caliper measurements were used to determine tumor volume.

Transient DNA Plasmid Transfection.

Cells were cultured to 60% confluence and transfected with DNA plasmids using Fugene 6 reagents (Roche, Indianapolis, IN) per the manufacturer's instructions.

Nuclear DNMT Activity Assay.

CD133− Huh7 cells were stimulated with 10 ng/mL TGFβ1 for 48 hours. Nuclear protein was extracted using a nuclear extraction kit (Epigentek, Brooklyn, NY); 5 μg nuclear protein were applied for DNMT activity assay which was performed using a EpiQuik DNA methyltransferase activity assay kit (Epigentek) per the manufacturer's protocol.


Genomic DNA was isolated from cells using a Wizard SV Genomic DNA purification System (Promega) and quantified using a ND-1000 spectrophotometer. Bisulfite modification was conducted using an EZ DNA methylation Kit (Zymo Research, Orange, CA) per the manufacturer's protocol. Briefly, 500 μg genomic DNA was incubated with CT conversion reagent for 16 hours at 50°C in the dark, followed by incubation with binding buffer and bound to Zymo-Spin IC column matrix. The DNA was washed and desulfonated and the bisulfite modified DNA was eluted with 10 μL elution buffer. DNA fragments of CD133 promoter-1 were amplified using primers that were designed using PSQ Assay Design Software version 1.06 (Biotage, Charlottesville, VA). Biotinylated P1 forward and reverse primers and conditions are presented in the Supporting Information Table, with initial amplification using 2 μL bisulfate modified DNA as template. The PCR product was purified using avidin-conjugated beads, purified single-strand DNA was subjected to pyrosequencing in PyroMark Q24 system (Biotage) using specific sequencing primers, P1 Seq-1 or P1 Seq-2, as listed, respectively. P1 Seq-1: 5′ AAATCTACCTCAATCACTTA 3′; P1 Seq-2: 5′ TATAAAAATACCTACTCAAC 3′. The data were analyzed using PyroMark Q24 software v. 1.09 (Biotage).

Statistic Analysis.

The paired two-tailed Student's t test was used when comparing two groups. A P value less than 0.05 was considered statistically significant. Analysis of variance was used for comparison of multiple groups, followed by pairwise multiple comparison procedures (Systat Software, Richmond, CA).


TGFβ1 Up-regulates CD133 Expression in Huh7 Cells.

Recent reports indicate that CD133 expression is controlled by microenvironment changes within the CSC niche.13, 27 We hypothesized that CD133 expression is regulated by known growth factors, such as TGFβ, that are highly expressed in cirrhotic liver. To test our hypothesis, Huh-7 cells were treated using 10 ng/mL TGFβ1 and analyzed using FACS, real-time PCR, and immunoblot. The number of CD133-expressing cells increased from 50% ± 4% to 75% ± 8% after 48 hours TGFβ1 treatment (Fig. 1A, P < 0.05). Huh-7 cells were then separated into CD133+ and CD133− cells. CD133+ and CD133− cells were treated with 10 ng/mL TGFβ1 for defined time intervals. Figure 1B,C shows that CD133 expression was induced by TGFβ1 treatment at both the messenger RNA (mRNA) and protein level.

Figure 1.

TGFβ1 up-regulates CD133 expression in Huh7 cells. (A) Huh7 cells were stimulated by 10 ng/mL TGFβ1 for 48 hours and analyzed by FACS. (B) Huh7 cells were sorted into CD133+ (closed bar) and CD133− (open bar) populations. The cells were stimulated with 10 ng/mL TGFβ1 for defined intervals and qPCR was performed (**P < 0.01). Bars represent mean ± SD after normalization to endogenous control (n = 3). (C) Huh7 cells were sorted into CD133+ and CD133− populations, stimulated with 10 ng/mL TGFβ1 for defined intervals. Immunoblot and densitometry analysis was performed, with fold change of CD133 expression after normalization to individual loading control β-actin (n = 3).

As recent reports indicated that cellular stress induces CD133 expression in glioblastoma,13, 27 and as our protocol for stimulation requires a serum-free period, CD133+ and CD133− Huh7 cells were challenged by serum deprivation for 48 hours combined with 10 ng/mL TGFβ1 stimulation. As shown in Fig. 2, TGFβ1, and not starvation, significantly induced CD133 expression. In addition, we performed dose- and time-dependent experiments on the effect of TGFβ1 on CD133 expression. CD133− Huh7 cells were stimulated with up to 20 ng/mL TGFβ1 for 48 hours. As depicted in Fig. 3A, CD133 expression was induced by TGFβ1 in a dose-dependent manner up to 2.5 ng/mL, and dosages between 2.5 and 10 ng/mL had similar effects on CD133 expression induction. CD133− cells were then treated with 5 ng/mL TGFβ1 for up to 48 hours, followed by repeat treatment with 0 to 10 ng/mL TGFβ1 for an additional 24 hours. As shown in Fig. 3B, TGFβ1-induced CD133 expression was in a time-dependent fashion, and once CD133 expression was induced, the expression remained elevated even after TGFβ1 stimulation was removed.

Figure 2.

Starvation is not involved in CD133 expression regulation. Huh7 cells were sorted into CD133+ and CD133− populations. The cells were cultured in serum-deprived medium for defined intervals without and with 10 ng/mL TGFβ1. Immunoblot was used to determine CD133 expression levels (upper panel), densitometry analysis results of relative CD133 expression presented in the lower panel after normalized to loading control β-actin (CD133+ closed bars, CD133− open bars; *P < 0.05; **P < 0.01). Bars represent mean ± SD (n = 3).

Figure 3.

TGFβ1 induces CD133 expression in a time- and dose-dependent manner. (A) CD133− Huh7 cells were incubated with TGFβ1 at indicated concentrations. Immunoblot was used to determine CD133 expression, and densitometry analysis was performed to show relative units of CD133 expression normalized to loading control GAPDH. (B) CD133− Huh7 cells were stimulated with different concentrations of TGFβ1 as indicated, and after 48 hours of incubation cells were retreated with TGFβ1 at a defined dosage for additional 24 hours. Immunoblot analysis was used to determine CD133 expression, and densitometry analysis of relative CD133 expression levels after normalization to loading control GAPDH is presented.

TGFβ1-Induced CD133+ Huh7 Cells Are Tumorigenic in Nude Mice.

As CD133 is a CSC marker in Huh7 cells, we questioned if the TGFβ1-induced CD133+ Huh7 cells have the property of tumor initiation in vivo, comparable to native CD133+ Huh7 cells. Freshly isolated, untreated CD133+ and CD133− Huh7 cells were used as controls. Thirty days after inoculation all of the mice transplanted with native CD133+ cells were sacrificed because the tumor size reached the endpoint according to our protocol (>3,500 mm3). As demonstrated in Fig. 4A, 6 and 12 hours of TGFβ1 stimulation increased CD133 expression in CD133− cells; 35 days after inoculation in nude mice, TGFβ1-induced CD133+ cells were significantly more tumorigenic compared to native CD133− cells (Fig. 4B,C).

Figure 4.

TGFβ1-induced CD133+ Huh7 cells are tumor-initiative. (A) CD133− Huh7 cells were treated with 2 ng/mL TGFβ1 for indicated intervals and thereafter incubated in TGFβ1-free medium for up to 48 hours. Immunoblot was conducted to verify CD133 expression levels and GAPDH was used as a loading control. (B) TGFβ1-induced CD133+ Huh7 cells were subcutaneously inoculated into nude mice (n = 6/group). The mice were sacrificed after 35 days and a representative photograph demonstrates tumor formation. (C) The tumorigenesis potential of TGFβ1-induced CD133+ cells were compared with native CD133− Huh7 cells in nude mice (*P < 0.05 compared with untreated cells). Tumor size presented as mean ± SD (n = 6).

TGFβ Induces CD133 Expression Partially Through the Smad Pathway.

Following activation of TGFβ receptors, Smad2 and Smad3 are phosphorylated and form a heterocomplex, Smad2/3/4, which translocates to the nucleus to regulate responsive gene transcription.28 In order to test whether TGFβ induces CD133 expression through Smad-dependent pathways, we used inhibitory Smads, Smad6 and Smad7, which are able to block heterocomplex formation. Huh7 cells were transfected with Smad629 and Smad730 vectors, and 48 hours after transfection cells were stimulated with 5 ng/mL TGFβ1. In qPCR analysis, elevated CD133 mRNA induced by TGFβ1 was significantly attenuated by inhibitory Smads (Fig. 5A). This expression pattern was confirmed at the protein level (Fig. 5B).

Figure 5.

TGFβ1 induces CD133 expression through a Smad-dependent pathway. Huh7 cells were cultured to 60% confluence, followed by transfection with inhibitory Smad DNA constructs. Then 48 hours after initial transfection, cells were stimulated with 5 ng/mL TGFβ1 for 48 hours, and empty vector (EV) transfected cells without TGFβ1 stimulation were used as negative control. (A) qPCR was conducted to determine the relative level of CD133 mRNA. (B) Immunoblot was used to determine CD133 expression and densitometry analysis was performed to calculate the relative CD133 expression units (**P < 0.01). Bars represent mean ± SD (n = 3).

CD133 Promoter-1 Methylation Down-Regulates Transcription.

In colon cancer cells CD133 expression is regulated by promoter methylation. Compared with the parental HCT116 cell line, a double knockout line with disruption of DNA methyltransferases DNMT1 and DNMT3β demonstrates increased CD133 expression.8 To test if similar epigenetic regulation is involved in CD133 expression in liver cancer, a DNMT inhibitor (5-aza-2′-deoxycytidine, DAC) and a histone deacetylase inhibitor (trichostatin A, TSA) were introduced. As shown in Supporting Information Fig. 1, CD133 expression in CD133− Huh7 cells was up-regulated by DNMT inhibitor in a time- and dose-dependent manner.

The CD133 gene transcription is controlled by five alternative promoters, with promoter-1 and -2 active in liver tissues.26 The 5′ proximal gene region of CD133 containing promoter-1 and -2 are located in a CpG island, indicating that they may be sensitive to DNA methylation. Given that inhibition of DNMTs induced CD133 expression, we investigated whether TGFβ1 regulated CD133 transcription through CD133 promoter demethylation. Using human CD133 promoter-1 driving luciferase vectors, we performed in vitro methylation using M. SssI followed by restriction enzyme HpaII cleavage to verify methylation status (Supporting Information Fig. 2). The luciferase activity of CD133 promoter-1 was significantly reduced by 53-fold following in vitro methylation, indicating that CD133 transcription was silenced by promoter methylation (Supporting Information Fig. 3).

TGFβ Inhibits DNMT1 and DNMT3β Expression.

To examine if DNMTs regulate CD133 expression through promoter methylation, we examined DNMT1, DNMT3α, and DNMT3β expression in CD133+ and CD133− cells. As demonstrated, there was no significant difference in DNMT1 expression between CD133+ and CD133− cells at baseline, and TGFβ1 stimulation reduced DNMT1 expression (Fig. 6A, upper panel). DNMT3α expression was significantly higher in CD133− cells compared to CD133+ cells, and TGFβ1 was not able to significantly alter DNMT3α expression (Fig. 6A, middle panel). DNMT3β expression is elevated in CD133− cells compared to CD133+ cells, and, DNMT3β expression was significantly suppressed by TGFβ1 stimulation in CD133− cells (Fig. 6A, lower panel).

Figure 6.

TGFβ1 inhibits DNMT1 and DNMT3β expression. (A) CD133+ and CD133− Huh7 cells were stimulated with 5 ng/mL TGFβ1 for 48 hours. qPCR was used to determine relative mRNA levels of DNMT1 (upper panel), DNMT3α (middle panel), and DNMT3β (lower panel), respectively. (B) Huh7 cells were transfected with indicated DNA constructs of inhibitory Smads and 48 hours after transfection the cells were stimulated with 5 ng/mL TGFβ1 for an additional 48 hours. Standard RT-PCR was conducted to analyze DNMT3β expression. (C) qPCR was performed to quantify relative DNMT3β expression (*P < 0.05). Bars represent mean ± SD (n = 3).

Given that DNMT1 and DNMT3β expression appeared to be regulated by TGFβ, we used inhibitory Smads to examine if this effect was promoted by a Smad-dependent pathway. Huh7 cells were transfected with Smad6 or Smad7 vectors followed by 5 ng/mL TGFβ1 stimulation. TGFβ1 suppression of DNMT1 expression was not rescued by inhibitory Smads (data not shown). However, inhibitory Smads were capable of attenuating the effect of TGFβ1 on DNMT3β expression (Fig. 6B). This expression pattern was confirmed using qPCR (Fig. 6C). These results indicate that TGFβ1 is involved in the regulation of DNMT expression. Next, we examined nuclear DNA methyltransferase activity in CD133− cells. After TGFβ stimulation, global DNMT activity in CD133− cell nuclei was significantly reduced, from 7.8 ± 2.5 units/hr/mg in untreated cells to 4.4 ± 0.8 units/hour/mg (n = 3, mean ± SD, P < 0.05).

TGFβ1 Induces Demethylation of CD133 Promoter-1.

Because TGFβ1 suppresses DNMT activity and increases CD133 expression, we investigated if TGFβ1 induces CD133 expression through promoter demethylation. CD133− cells were stimulated with 5 ng/mL TGFβ1 for 48 hours and genomic DNA was extracted and subjected to pyrosequencing analysis. As shown in Fig. 7A, two DNA fragments containing seven potential CpG methylation sites within CD133 promoter-1 were analyzed. TGFβ1 was able to significantly reduce the methylation percentage across multiple CpG sites after 48 hours incubation (Fig. 7B, Supporting Information Fig. 4). Interestingly, the more distal the CpG methylation site was from exon 1A, the more heavily methylated it was, which is a pattern that has been described.31

Figure 7.

TGFβ1 induces demethylation in CD133 promoter-1 in CD133− Huh7 cells. (A) Schematic illustration of P1 forward and P1 reverse primers were used to amplify DNA fragment of CD133 promoter-1. Putative CpG sites are shown as vertical bars. The thick vertical bars represent those CpG sites that were covered by sequencing primers, P1 Seq-1 and P1 Seq-2. (B) CD133− Huh7 cells were stimulated with 5 ng/mL TGFβ1 for 48 hours. Bisulfite modified DNA was amplified using template followed by pyrosequencing. The data represent the percentage of methylation at corresponding CpG sites, with CpG site number corresponding to those sites identified in the schematic above. (*P < 0.05 and **P < 0.01). Bars represent mean ± SD (n = 3).

Proposed Mechanism of TGFβ1 Induces CD133 Expression.

Based on our findings, we proposed a novel mechanism by which TGFβ1 induces CD133 expression, as shown in Fig. 8. After TGFβ1 binds to TβRII, TβRI is phosphorylated, and thereafter activated receptor complexes propagate TGFβ signaling through phosphorylating receptor-associated Smads. After Smad2 and Smad3 phosphorylation, Smad4 is recruited as a co-Smad, then the activated Smad2/3/4 heterocomplexes translocate to nucleus in which it regulates responsive gene transcription including DNMT1 and DNMT3β. Decreased DNMT1 and DNMT3β expression may result in demethylation in responsive gene promoters, such as CD133 promoter-1, which leads to enhanced gene transcription.

Figure 8.

Schematic of proposed mechanism by which TGFβ1 regulates CD133 expression. After TGFβ binds to receptor, Smad heterocomplex is formed, which then acts to suppress the expression of DNMTs. Lower levels of DNMTs lead to hypomethylation of CD133 promoter-1, which in turn result in increased CD133 transcription.


We and others have previously demonstrated that CD133 is a promising liver CSC surface marker.10–12, 24 CD133+ liver CSCs are resistant to chemotherapy and apoptosis.10 Given that TGFβ is a key cytokine that may link chronic liver injury to CSCs,32 the goal of this study was to understand the association between TGFβ and CD133 expression. As clearly demonstrated, CD133 expression was up-regulated by TGFβ1 stimulation through epigenetic regulation of CD133 promoter methylation. Furthermore, TGFβ1-induced CD133+ cells demonstrated increased tumorigenicity compared to CD133− cells.

CD133 is a pentaspan, transmembrane glycoprotein. In murine models of chronic liver injury CD133 expression steadily increases as injury progresses to HCC.10–12, 24 During these investigations we noted that a murine model associated with a liver-specific hypomethylation state (MAT1A−/−) had significantly more CD133+ oval cells compared to other murine models.10–12, 24 In terms of stem cells giving rise to human HCC, Sell and Dunsford33 originally proposed this concept. This initial hypothesis has been supported by numerous murine models and human cell line investigations, but definitive proof that human HCC is derived from CSCs is still lacking.34 A number of recent publications demonstrated that various solid tumors, such as colon, brain, ovarian, thyroid, and prostate cancers are derived from CD133+ CSCs.7, 35, 36 Specifically within colon cancer, CD133 expression is an independent prognostic marker for poor survival.36 In the liver, two independent groups demonstrated that CD133+ liver CSCs display significant in vivo tumorigenesis and stem cell-like properties.3, 4 Furthermore, increased CD133 expression has been directly linked to poor prognosis in human patients with HCC.34 Although no treatment specifically using CD133 has been published in liver cancer, enforced down-regulation of CD133 expression impaired cell proliferation, motility, and metastasis in melanoma.14 Given all of these findings, we postulate that CD133 is not only an important prognostic marker of HCC progression, and CSCs specifically, but a potential therapeutic target as well.

The vast majority of HCC develops after prolonged, chronic liver injury, particularly from HCV or HBV infections.15 It was originally proposed that liver stem cells, or oval cells, have the capacity to regenerate the liver during these times of chronic liver injury.37, 38 Recently, Amin and Mishra20 proposed a refinement of this oval cell model, where activated liver stem cells acquire resistance to TGFβ-induced cell growth inhibition and differentiation, and thus are able to escape the normal cell cycle control and undergo malignant transformation.

During chronic liver diseases, TGFβ is secreted by nonparenchymal cells such as hepatic stellate cells and acts as a stimulator of extracellular matrix production, resulting in fibrosis and cirrhosis.17 Nearly 95% of HCC is developed in a cirrhotic liver in chronic hepatitis C infection and almost 60% during chronic hepatitis B.17 As a profibrotic growth factor in liver, TGFβ acts as an inhibitor of hepatocyte proliferation; however, the exact role of TGFβ in HCC initiation and progression remains controversial. It is believed that TGFβ inhibits carcinogenesis at the early stage and acts as a promoter of cancer progression at the later stage disease.39 Increased hepatocarcinogenesis from stem cells through disruption of TGFβ and IL-6 signaling provided additional evidence of the association between TGFβ and HCC.32

Some studies revealed that cellular stress, hypoxia, for example, and the mTOR signal pathway, are able to induce CD133 expression in cancer cells.27, 40 This finding indicated that CD133 expression in liver cells may be in response to microenvironmental alterations. For example, in a cirrhotic liver the cells are exposed to a microenvironment abundant in TGFβ. Therefore, we postulated that elevated TGFβ might be able to trigger CD133 transcription. In our current findings we demonstrated that TGFβ1 was capable of inducing CD133 expression in CD133− Huh7 cells. Although TGFβ1-induced CD133+ Huh7 cells were less tumorigenic than that of native CD133+ Huh7 cells, the induced CD133+ cells were characterized as significantly more tumorigenic than native CD133− cells in vivo. These findings might serve as an additional link between TGFβ and malignant transformation in chronic liver injury.

A previous publication demonstrated that CD133 expression is controlled by five alternative promoters, with promoter-1 and -2 active in liver tissue.26 CD133 promoter-1 and -2 are located in a CpG island, indicating that CD133 transcription in the liver may be regulated by epigenetic modification through promoter methylation status. Promoter methylation is an important mechanism that leads to gene transcriptional silencing. CpG hypermethylation in promoter regions of tumor suppressors has been linked to tumorigenesis.21, 22 For example, Ras and downstream Ras effectors were activated due to epigenetic silencing of inhibitors of the Ras pathway in HCC.41 In terms of CD133, recent reports indicate that expression is inversely correlated with promoter methylation.7 In colorectal and ovarian cancer cells, demethylating reagents enhance CD133+ cell numbers, and a high frequency of hypermethylation in CD133 promoter in CD133− colorectal cancer and glioblastoma cell lines was found using methylation sensitive PCR.8 However, significant CD133 promoter methylation was absent in normal colon and brain tissues, which highlights the complexities of CD133 promoter methylation in diverse tissues and cells.8 Furthermore, within the CD133 promoter-1 region the degree of methylation changes is based on the separation of the individual CpG site from exon1.31 Our current results demonstrated that CD133 expression was enhanced by inhibiting DNMT activity and in vitro methylation silenced promoter-1.

DNA methylation status is regulated directly by DNMTs, which possess de novo methylation activity.21 Here we demonstrated that DNMT3α and DNMT3β expression was significantly higher in CD133− cells compared with CD133+ cells. These results support our hypothesis that CD133 expression in CD133− cells was silenced by promoter CpG methylation. Furthermore, we demonstrated that DNMT1 and DNMT3β expression was regulated by TGFβ stimulation. Our data are consistent with the results of enhanced CD133 expression from colon cancer cells, in which both DNMT1 and DNMT3β were deleted.8, 23 In addition, we demonstrated that TGFβ stimulation effectively reduced total nuclear DNMT activity. We conclude that DNMT1 and DNMT3β are critical enzymes in the mechanism of TGFβ1-induced CD133 expression.

Given that CD133 promoter methylation has specific patterns in diverse tissues, we chose pyrosequencing as a means to quantify the promoter methylation degree within multiple CpG sites. Our data demonstrated that TGFβ1 is capable of significantly reducing CD133 promoter-1 methylation by 10% to 40% in five out of seven CpG sites analyzed. Although the effect of TGFβ1 on methylation in individual CpG sites is relatively small, the overall effect of accumulated demethylation induced by TGFβ1 in multiple CpG sites likely has the significant influence on CD133 transcription that we observed. Therefore, we propose that TGFβ1-induced demethylation in CD133 promoter might act as a rheostat to regulate CD133 transcription.

Although multiple publications demonstrated that CD133 is a marker of CSCs with tumorigenic properties from diverse tissues, a recent study indicated that both CD133+ and CD133− metastatic colon cancer cells were capable of initiating tumor formation.42 This finding indicated that CD133 by itself might not be critical for tumor initiation. We propose that further investigations are required before the role of CD133 in liver cancer initiation and progression is fully elucidated. Our results do provide a link between CD133 expression regulation and TGFβ within this evolving field.

In summary, this work describes a mechanism by which TGFβ regulates CD133 expression through demethylation of promoter-1. This epigenetic regulation might act as a fine-tune control of CD133 transcription rather than as an on/off switch. Our current findings provide a novel therapeutic strategy to interfere with HCC initiation through modifying the CD133 promoter methylation status by potentially targeting TGFβ/Smads or DNMTs.


We thank Drs. Vrana and Freeman of the Functional Genomics Core at the Penn State College of Medicine. Important Penn State Functional Genomics Core Facility instrument purchases were made possible through Tobacco Settlement Funds and through the Penn State Cancer Institute contract with the Department of the Navy. We thank Dr. Laura Carrel and Sarah Arnold-Croop from the Penn State College of Medicine for their insight and assistance in pyrosequencing methods.