CD133⁺ Liver Cancer Stem Cells from Methionine Adenosyl Transferase 1A–Deficient Mice Demonstrate Resistance to Transforming Growth Factor (TGF)-β–Induced Apoptosis

Abstract

Methionine adenosyltransferase (MAT) is an essential enzyme required for S-adenosylmethionine biosynthesis. Hepatic MAT activity falls during chronic liver injury, and mice lacking Mat1a develop spontaneous hepatocellular carcinoma by 18 months. We have previously demonstrated that CD133⁺CD45⁻ oval cells isolated from 16-month-old Mat1a^−/− mice represent a liver cancer stem cell population. The transforming growth factor β (TGF-β) pathway constitutes a central signaling network in proliferation, apoptosis, and tumorigenesis. In this study, we tested the response of tumorigenic liver stem cells to TGF-β. CD133⁺CD45⁻ oval cells were isolated from premalignant 16-month-old Mat1a^−/− mice by flow cytometry and expanded as five clone lines derived from a single cell. All clone lines demonstrated expression of both hepatocyte and cholangiocyte markers and maintained a small population (0.5% to 2%) of CD133⁺ cells in vitro, and three of five clone lines produced tumors. Although TGF-β1 inhibited cell growth equally in CD133⁻ and CD133⁺ cells from each clone line, the CD133⁺ population demonstrated significant resistance to TGF-β–induced apoptosis compared with CD133⁺ cells. Furthermore, CD133⁺ cells demonstrated a substantial increase in mitogen-activated protein kinase (MAPK) pathway activation, as demonstrated by phosphorylated extra-cellular signal-regulated kinase levels before and after TGF-β stimulation. MAPK inhibition using mitogen-activated protein kinase kinase 1 (MEK1) inhibitor PD98059 led to a significant increase in TGF-β–induced apoptosis in CD133⁺ cells. Conversely, a constitutively active form of MEK1 blocked the apoptotic effects of TGF-β in CD133⁻ cells.

Conclusion

CD133⁺ liver cancer stem cells exhibit relative resistance to TGF-β–induced apoptosis. One mechanism of resistance to TGF-β–induced apoptosis in CD133⁺ cancer stem cells is an activated mitogen-activated protein kinase/extracellular signal-regulated kinase pathway.

Hepatocellular carcinoma (HCC) is the third leading cause of cancer-related mortality worldwide.¹ The prognosis of HCC depends on the cancer stages at the time of diagnosis. Although surgical therapies have led to an improvement in the 5-year survival of select patients, most patients with HCC gain no significant benefit from traditional chemotherapy.²

Recently, a number of studies have demonstrated that solid tumors such as colon,³ pancreatic,⁴ brain,⁵ and prostate cancers⁶ are initiated from cancer stem cells (CSCs). CSCs are resistant to injury and may account for the relative failure of traditional chemotherapy.⁷ Several studies have linked CD133 expression to liver CSCs, demonstrating that CD133⁺ cells from established HCC cell lines displayed significant tumorigenic capacity.^8,9 In patients with HCC, a hepatoblast phenotype correlates with a significantly worse prognosis.¹⁰

In a previous study, we defined a CD133⁺ CSC population isolated from methionine adenosyltransferase 1a (Mat1a)-deficient mice during premalignant liver injury.¹¹ Methionine adenosyltransferase (MAT) is an essential enzyme responsible for the synthesis of S-adenosylmethionine, the principal methyl donor required for glutathione biosynthesis.¹² The relationship between hepatic MAT activity and chronic liver disease in human patients is well established.^12–14 Mat1a knockout mice demonstrated hepatosteatosis, oxidative liver injury, and spontaneous development of HCC by 18 months.¹⁴ What was unknown from our initial research was a specific mechanism of survival of the CD133⁺ cell population.

Resistance to transforming growth factor β (TGF-β) has been postulated to be an early event in HCC development. ^15,16 TGF-β is the prototype of a large family of structurally related growth and differentiation factors that initiates its signals from a receptor complex,¹⁷ and intermediary activated Smads translocate into the nucleus, where they induce or suppress transcription of defined genes.¹⁸ In hepatocytes, TGF-β acts as a principle growth inhibitor,¹⁹ mediated by inducing expression of the CDK inhibitors p21 and p15, and down-regulating c-myc, cyclin D, and cyclin E.²⁰ In addition, TGF-β induces apoptosis in several established human liver cell lines, including HepG2 hepatoma and HepG3 HCC cells.²¹ However, the precise role of TGF-β in HCC progression remains complex and depends on the stage of the tumor.¹⁵

In order to understand the role of TGF-β in the homeostasis of CD133⁺ liver CSCs, we tested both the cell growth inhibitory and apoptotic effects of TGF-β on CD133⁺ CSCs with Mat1a deficiency. Although there is no significant difference in the cell growth inhibition in CD133⁺ and CD133⁻ cells in response to TGF-β, CD133⁺ cells did exhibit relative resistance to the apoptotic effects of TGF-β as compared with CD133⁻ cells. Our results indicate that one mechanism for the resistance to TGF-β–induced apoptosis in CD133⁺ CSCs is an activated mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (Erk) pathway.

Materials and Methods

Reagents

See Supplementary Fig. 1 for detailed list of all reagents.

Cell Culture

Mat1a^−/− cells expanded from a single CD133⁺CD45⁻ cell were cultured in 1:1 Dulbecco’s modified Eagle’s medium/F12 (Sigma) containing 10% fetal bovine serum as described.^11,22 CD45 depletion was conducted using Miltenyi CD45 microbead antibodies (Miltenyi Biotec Inc, Auburn, CA) per the manufacturer’s protocol, followed by CD133⁺CD45⁻ cell flow cytometry (FACS) isolation for single cell expansion. Unless otherwise specified, 2 × 10⁴ cells/cm² were plated.

Animal Care

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

Tumor Formation Assay

Cells were counted with trypan blue exclusion, as described,^11,22 and were resuspended in phosphate-buffered saline (PBS) for transplantation at a concentration of 2 × 10⁶ live cells/200 µL (1:1 Matrigel/PBS). Six-week-old female nude mice (Jackson Laboratory, Bar Harbor, ME) or syngeneic wild-type mice were used for tumor formation analysis using subcutaneous inoculation, and tumors were isolated after 15 weeks.

Real-Time Polymerase Chain Reaction

Total RNA was extracted from 2 × 10⁴ cells/cm² in six-well plates using Trizol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer’s protocol. RNA was quantified using an ND-1000 spectrophotometer (Nanodrop Technologies, Wilmington, DE) and complementary DNA constructed as described.²² Real-time polymerase chain reaction (PCR) experiments were conducted using an ABI-Prism 7700 Thermal Cycler and TaqMan Universal PCR Master Mix (Applied Biosystems, Foster City, CA). The housekeeping gene β-Actin was used for all ΔΔCt calculations. Relative expression was calculated for the genes p15, p21, cyclin D1, and c-myc and was assessed using primer/probe sets (Applied Biosystems). Standard RT-PCR was conducted using the primers listed in Supplementary Fig. 1. Primers were selected in separate exons, with PCR conditions as described.¹¹

Anchorage-Independent Growth Assays

After 0.8% bottom soft agar gel was plated in six-well culture plates, 20,000 cells were prepared in 0.3% agar gel in each well and loaded to the top of the bottom agar when it was completely solidified. The plates were placed in a 37°C humidified incubator with 5% CO₂. One hundred microliters of fresh medium was added to the top of agar every other day. After 21 days, cells were stained with 0.005% crystal violet.

Apoptosis Assays

For DNA laddering assays, after 1 hour in serum-free medium, cells were treated with 5 ng/mL of TGF-β₁ for 24 hours and collected for DNA isolation using a Suicide Track DNA Ladder Isolation Kit (Calbiochem, La Jolla, CA). DNA fragments were separated using 1.5% agarose gel and visualized using ethidium bromide staining. For caspase-3 assays, cells were treated with serum-free medium for 1 hour, then with 5 ng/mL of TGF-β₁ for the indicated duration. Cells were trypsinized, fixed, permeabilized, and stained with activated caspase-3–phycoerythrin antibody (BD Biosciences, San Diego, CA). For annexin V/propridium iodine (PI) staining, cells were pretreated with TGF-β as described above, trypsinized, and stained using the Annexin V/PI Apoptosis Kit (BioVision, Mountain View, CA) according to the manufacturer’s protocol. FACS analysis was conducted on a BD FACS-Calibur. Analysis was conducted using FlowJo (Tree Star, Ashland, OR). For apoptosis assays, 2 × 10⁴ cells/cm² cells were plated on 60-mm dishes.

Western Blot Analysis

For Western blot analysis, cell lysates were harvested by the addition of lysis buffer (40 mM Tris [pH 7.4], 150 mM NaCl, 10 mM ethylene diamine tetraacetic acid, 10% glycerol, 1% Triton X-100, 10 mM glycerophosphate, 1 mM Na₃VO₄, 1 mM phenylmethylsulfonyl fluoride) supplemented with protease inhibitor cocktail tablets (Roche, Indianapolis, IN). Forty micrograms of protein lysates were separated on a Nu-PAGE 4% to 12% Bis-Tris Gel (Invitrogen, Carlsbad, CA) and transferred to a polyvinylidene difluoride membrane (Invitrogen), as described.²³ Signals were detected using the enhanced chemiluminescence solutions (Thermo Fisher Scientific, Rockford, IL). Densitometry was analyzed using ImageJ 1.40 software (National Institutes of Health) and normalized with pan-Erk1/2 expression levels.

CD133⁺ Cell Isolation

CD133⁺ cell isolation was performed using Miltenyi MACS systems according to the manufacturer’s protocol as described.²⁴ Cells were trypsinized and suspended in 500 µL of 1 × PBS/2 mM ethylene diamine tetraacetic acid/0.5% bovine serum albumin buffer and incubated with magnetic microbeads conjugated with anti–Prominin-1 antibody (catalog #130-092-333) prior to separation using Miltenyi LS column.

Tritiated Thymidine Incorporation Assay

2 × 10⁴ cells/cm² were plated in triplicate in six-well plates. Cells were pulsed with 1 µCi/mL tritiated thymidine for 2 hours, then washed with 1 × PBS, and precipitated with 10% trichloroacetic acid for 10 minutes, and solubilized with 0.2N sodium hydroxide/salmon DNA buffer before quantitation with a scintillation counter.

Cell Viability Assay

Cell viability was performed using the XTT [2,3-bis(2-methoxy-4-nitro-5- sulfophenyl)-2H-tetrazolium-5-carboxanilide] kit (Trevigen, catalog #4891-025-K) according to the manufacturer’s protocol. 1 × 10⁴ cells/well were plated in 96-well plates. Twenty-four hours after either β-galactosidase (β-Gal) or a constitutively active form of mitogen-activated protein kinase kinase 1 (CA-MEK1) adenoviral infection, cells were treated in serum-free medium for 1 hour, then were incubated in the presence or absence of 5 ng/mL of TGF-β for an additional 12 hours prior to analysis.

Complementary DNA Microarray

Complementary DNA from CD133⁺ and CD133⁻ cells were hybridized to Illumina Mouse ref. 8 gene chip (Illumina, San Diego, CA) according to the manufacturer’s standard protocol. Housekeeping genes were used as standards to generate expression levels, and data analysis was conducted using a 1.4-fold or greater change in expression, with P < 0.01 considered significant.

Adenovirus Infection

All recombinant adenoviruses were expanded, purified, and titrated using BD Clontech’s Adeno-X Rapid Titer Kit per manufacturer’s protocol (Mountain View, CA) in a HEK293 monolayer of cells. Using either CA-MEK1 or β-Gal adenoviral constructs above, cells were infected using the indicated virus multiplicity of infection (MOI) as described.²⁵

Statistical Analyses

The paired 2-tailed Student t test was used when comparing 2 groups. A p value less than 0.05 was considered significant.

Results

Mat1a^−/− Clone Lines Maintain Bipotent Capacity of Liver Oval Cells

Five clone lines were expanded from single C133⁺CD45⁻ non-parenchymal cells isolated from 16-month-old Mat1a^−/− mice, as depicted in Fig. 1A.¹¹ As shown in Fig. 1B, all five clone lines express both hepatocyte and cholangiocyte markers, such as Albumin and Ck-19. In addition, each clone line expresses a number of oval cell associated genes: Hnf3β, Hnf1α, and αFP.^22,26 Western blot confirms αFP and CD133 protein expression in all clone lines (Fig. 1C). There was no difference in gene expression between early (passages 3–6) and later (passages 30–35) cells (data not shown). Within each clone line, the number of CD133⁺ cells remained relatively stable at 0.5% to 2% across multiple passages.

Fig. 1.

Single-cell isolation for clonal expansion. (A) Schematic representation of single-cell FACS isolation of CD133⁺CD45⁻ cells from 16-month-old Mat1a^−/− mice. On day 1, single cells were plated into individual wells of a 96-well plate. After 21 days, 10/96 wells showed colonies with growth over 50% of the well bottom. Five of these colonies were selected for expansion (CD133⁺ clone lines 1–5). (B) RT-PCR. In each clone line 1–5, coexpression of Albumin and CK-19 indicated bipotential stem cells, with strong expression of oval cell–associated transcription factors hepatocyte nuclear factor 1α and 3β (HNF1α and HNF3β) and ABCG2 transmembrane pump. Strong expression of growth factor receptors c-Met and epithelial growth factor receptor (EGFr) were also detected by RT-PCR. (C) Western blot. High levels of prominin-1 (CD133) and αFP protein were detected in each clone line (+ control = positive control for Western blot).

Anchorage-Independent Growth and Xenograft Tumor Formation from Mat1a^−/− CD133⁺ Clone Lines

Our previous study demonstrated that bulk culture of CD133⁺ cells isolated from Mat1a^−/− mice produced tumors in 40% of immune-deficient mice.¹¹ As shown in Fig. 2A, all five clone lines grew in an anchorage-independent manner. In order to assess the tumor-forming ability of CD133⁺ cell-derived clone lines in vivo, a tumor model with immune-deficient mice was used. Two million cells isolated from each clone line were subcutaneously inoculated into immune-deficient mice (n = 4 per cell line, five lines total, passage 5). Of the five lines expanded from single CD133⁺ cell (clone lines 1–5), three lines formed tumors in nude mice at passage 5 (line 1 2/4, line 2 ¼, line 3 4/4, line 4 0/4, and line 5 0/4). There was no tumor formation in the control mice injected with Matrigel and PBS carrier. Tumor histology revealed hepatoma-like cells with mixed epithelial cell morphology and columnar/cuboidal cells, and the average tumor size was 200 ± 80 mg (Fig. 2B). Two million cells from tumorigenic line 3 were also transplanted into syngeneic wild-type mice. The small tumors in 25% of transplantations demonstrated hepatoma-like cells on hematoxylin-eosin staining (2/8, tumor size 50 and 100 mg) (Supplementary Fig. 2A–B). Subsequent analyses focus on tumorigenic lines: CSC clone lines 1, 2, and 3.

Fig. 2.

CD133⁺ Cancer stem cells. (A) All five CSC clone lines demonstrated anchorage-independent growth in soft agar at 21 days (4× and 20× objective), and (B) three of the five lines formed subcutaneous tumors in nude mice (2 × 10⁶ cells injected, passage 5). Hematoxylin-eosin staining demonstrates tumor histology with mixed hepatoma-like populations of liver epithelial tumors (4× and 40× objectives).

Cell Growth Inhibition in Response to TGF-β

Given the growth inhibition effects of TGF-β, we tested the proliferation response of CSC clone lines after TGF-β stimulation. In serum-free conditions, 14 ± 5% of CSC clone line 1–3 cells enter S phase of cell cycle (1 hour bromodeoxyuridine pulse). After 5 ng/mL of TGF-β₁ stimulation (24 hours), the number of cells entering S phase was significantly decreased by nearly 90% compared with serum-free controls, indicating that Mat1a^−/− CSCs are sensitive to growth inhibition by TGF-β (Fig. 3A). This level of inhibition was observed in all three CSC clone lines (Fig. 3B). When the CSC clone lines were separated based on CD133 expression, CD133⁻ and CD133⁺ cells were equally sensitive to growth inhibition by TGF-β₁, showing 85% inhibition of [3H]-thymidine incorporation (Fig. 3C). For this and future analyses, isolation of CD133⁺ and CD133⁻ cells was conducted using cells from the same culture plate.

Fig. 3.

CSC clone lines demonstrate growth inhibition in response to TGF-β stimulation. (A) Compared with serum-free baseline conditions, 24-hour TGF-β stimulation results in a significant decrease in cell cycling using bromodeoxyuridine 1-hour pulse (n = 3 lines, each in triplicate, + standard deviation). *P < 0.05. (B) TGF-β stimulation results in similar decrease in cell cycling in all three CSC cell lines. (C) No difference in cell cycle inhibition in response to TGF-β stimulation between C133⁺ and CD133⁻ cells isolated from each CSC clone line using [³H]-thymidine incorporation. (B, C) Twenty-four-hour TGF-β stimulation, 2-hour [³H]-thymidine pulse, percent inhibition after TGF-β stimulation compared with serum-free controls, each line in triplicate, + standard deviation (P > 0.05 for each line CD133⁺ versus CD133⁻).

No Difference in TGF-β Signal Proteins

In order to examine if TGF-β signal pathway elements were differentially expressed in CD133⁺ and CD133⁻ cells, we used standard immunoblot assays to measure the protein levels of TGF-β receptor, Smad2/3, and Smad4, as well as the inhibitory Smad6/7 proteins. There was no substantial difference in the protein expression of either TGF-β receptor or Smad proteins between CD133⁺ and CD133⁻ cells with and without TGF-β stimulation (Supplementary Fig. 3). We could not detect Smad6/7 proteins in either CD133⁻ or CD133⁺ cells.

Analysis of Expression of TGF-β Inducible Genes Related to Cell Cycle Regulation

To further characterize any difference in the responsiveness after TGF-β–induced G1 phase cell cycle arrest, we used TaqMan real-time RT-PCR to examine TGF-β–regulated genes. p15^INK4b(p15) and p21^WAF1/CIP1(p21), the potent inhibitors of cyclin-dependent kinases, function as cell cycle inhibitors by blocking cyclin D and cyclin E. As shown in Fig. 4A, in both CD133⁻ and CD133⁺ cells, the expression of p21 was up-regulated, whereas cyclin D1 and c-myc were down-regulated 4 hours after TGF-β stimulation. The expression of c-myc and cyclin D1 remained at a suppressed level 12 hours after TGF-β treatment (Fig. 4B). There was no significant difference between CD133⁺ and CD133⁻ cells in the fold changes of p15, p21, c-myc, and cyclin D1 messenger RNA (mRNA) levels after TGF-β stimulation.

Fig. 4.

Real-time PCR analysis of CD133⁺ and CD133⁻ cells in response to TGF-β stimulation. Real-time PCR analysis of TGF-β down-stream targets, p15 and p21, and cell cycle regulators, c-myc and cyclin D1, (A) 4 hours and (B) 12 hours after TGF-β stimulation (5 ng/mL). No significant difference between CD133⁺ and CD133⁻ cells was detected (2 × 10⁴ cells/cm² plated, data presented as fold change in expression, n = 3 lines each in triplicate + standard error).

CD133⁺ Cells Demonstrate Resistance to TGF-β–Induced Apoptosis

TGF-β can function by inhibition of cell cycle and induction of apoptosis in murine primary hepatocytes and hepatocytic cell lines,²¹ as well as several HCC cell lines.^27,28 Apoptosis was determined using DNA laddering, activated caspase-3 labeling, and annexin V/PI staining. When CSC clone lines were exposed to TGF-β1 for 24 hours, DNA laddering was detectable in both the detached and the attached fractions but not in control serum-free cells (Supplementary Fig. 4). Using activated caspase-3 FACS analysis, the number of apoptotic cells increased in both CD133⁺ and CD133⁻ cell fractions with increased time of TGF-β stimulation (Fig. 5A). For all future experiments, we chose a 12-hour time point of TGF-β incubation. When we tested CD133⁺ and CD133⁻ cells, obtained from the same culture plate of the CSC clone lines, the CD133⁺ cells (lines 1–3) demonstrated a significant resistance to TGF-β–induced apoptosis compared with CD133⁻ cells, displaying a 1.5-to 3-fold reduction in the number of apoptotic cells stained with annexin V/PI on FACS analysis (Fig. 5B).

Fig. 5.

CD133⁺ cells demonstrate resistance to TGF-β–induced apoptosis. (A) Timing of TGF-β–induced apoptosis in CD133⁺ and CD133⁻ cells with activated caspase-3 levels increasing from 4 to 16 hours using intracellular active caspase-3 FACS analysis (2 × 10⁴ cells/cm² plated from CSC clone line 1, each time point in triplicate + standard deviation). *P < 0.05. (B) CD133⁺ cells demonstrate less apoptosis compared with CD133⁻ cells from the same culture plates from each of the CSC clone lines 1–3. Apoptosis determined using annexin V/PI staining and FACS analysis (2 × 10⁴ cells/cm² plated, each line in triplicate, 50,000 events counted/replicate + standard error). *P < 0.05.

MAPK/Erk Was Constitutively Activated in Mat1a^−/− Clone Lines

In the mRNA microarray analysis, the Ras/MAPK/Erk signal pathway components (RAB5C RAS oncogene, MEK1, and p14) are all up-regulated in CD133⁺ cells compared with CD133⁻ cells (Supplementary Fig. 5A). Among these genes, MEK1 lies upstream of MAP/Erk, and MEK1 stimulates the enzymatic activity of MAPKs. To test the hypothesis that the Ras/MAPK/Erk pathway may execute an antiapoptotic role in Mat1a^−/− CD133⁺ CSCs, we isolated CD133⁻ and CD133⁺ cells from CSC clone lines to determine the activated Erk levels. As shown in Fig. 6A,B, Erk was constitutively phosphorylated in both CD133⁻ and CD133⁺ cells, with an overall 1.8-fold increase in phosphorylated Erk1/2 (p-Erk1/2) level in CD133⁺ cells compared with CD133⁻ cells when signals were normalized with pan-Erk1/2. In addition, as shown in Fig. 6C, TGF-β suppressed p-Erk1/2 after a short period of TGF-β exposure in both populations.

Fig. 6.

Elevated p-Erk in CD133⁺ cells. (A) Western blot analysis demonstrates elevated p-Erk in CD133⁺ cells compared with CD133⁻ cells in each of the CSC clone lines 1–3. pan-Erk was used as a loading control. (B) Densitometry analysis of all three CSC lines revealed a significant increase in p-Erk levels in CD133⁺ cells (2 × 10⁴ cells/cm² plated, n = 3 lines + standard error). *P < 0.05. (C) Western blot analysis demonstrated that TGF-β stimulation results in a decrease in p-Erk levels, with a smaller decrease in CD133⁺ cells.

Specific MEK1 Chemical Inhibitor Increased Sensitivity to TGF-β–Induced Apoptosis in CD133⁺ Cells

In order to verify whether blockade of the MAPK pathway is capable of reversing the resistance of Mat1a^−/− CD133⁺ CSCs to TGF-β–induced apoptosis, we used PD98059, an inhibitor that blocks MEK1, the upstream kinase of Erk1/2. As shown in Fig. 7A, p-Erk1/2 levels were reduced in a dose-dependent manner by PD98059. At 25 µM of PD98059, p-Erk1/2 was inhibited 80% to 90% in CSC clone lines 1 and 3 (Fig. 7B). CD133⁺ and CD133⁻ cells from the CSC clone lines were treated with 25 µM of PD98059 for 1 hour, cultured in serum-free medium for 1 hour, and stimulated with 5 ng/mL of TGF-β₁ for 12 hours. As shown in Fig. 7C, 5.4 ± 0.2% of CD133⁺ cells underwent apoptosis upon TGF-β stimulation after DMSO pre-treatment. After pretreatment with PD98059, TGF-β stimulation significantly increased the number of CD133⁺ cells undergoing apoptosis to 17.8 ± 0.4% (P < 0.05 versus DMSO group), demonstrating that the survival advantage of CD133⁺ cells is reversed with MEK1 inhibition. A similar increase in apoptosis was also observed in CD133⁻ cells after PD98059/TGF-β treatment. Pretreatment with either DMSO or PD98059 without TGF-β stimulation did not result in a significant change in the number of apoptotic cells.

Fig. 7.

Mek1 inhibition reverses survival advantage of CD133⁺ cells by decreasing p-Erk levels. The MEK1 inhibitor PD98059 was used to block Erk activation. (A) Erk phosphorylation decreased with increasing dose of PD98059. pan-Erk was used as a loading control. (B) Graph of densitometry comparing p-Erk with pan-Erk in CSC cell lines 1 and 3. (C) PD98059 reverses the survival advantages of CD133⁺ cells. TGF-β stimulation with PD98059 pretreatment resulted in a significant increase in apoptosis in both CD133⁺ and CD133⁻ cells. Apoptosis measured using annexin V/PI staining and FACS analysis (2 × 10⁴ cells/cm² plated, n = 3, 50,000 events/replicate + standard error). #1,#2,#3P < 0.05.

CA-MEK1 Protects CD133⁻ Cells from TGF-β–Induced Apoptosis

In order to determine if superactivated MAPK signals are capable of antagonizing the apoptosis induced by TGF-β in CD133⁻ cells, we employed an adenoviral construct that expresses CA-MEK1. CA-MEK1 contains S218E/S222E mutations and is activated without ligand binding.²⁹ To determine appropriate adenovirus concentrations, Mat1a^−/− CSC clone lines were infected with adenovirus-expressing β-Galactosidase (β-Gal) with an MOI of 0, 5, 10, 25, 50, and 100. Twenty-four hours after adenoviral infection, over 95% of cells were positively stained with X-Gal at MOI of 100 adenovirus and 80% of cells contain positive staining at MOI 50 (Supplementary Fig 6). When we infected CSC clone lines with CA-MEK1 adenovirus, Erk1/2 was phosphorylated in a dose-dependent manner (Fig. 8A,B).

Fig. 8.

Forced increase in p-Erk provides a survival advantage in CD133⁻ cells. Adenovirus vectors with β-Gal or CA-MEK1 were used at different MOI. (A) Western blot of p-Erk and pan-Erk demonstrated a marked increase in p-Erk levels in cells infected with CA-MEK1 adenovirus compared with β-Gal adenovirus, with increased levels at higher MOI. (B) Graph of densitometry analysis of p-Erk/pan-Erk ratio demonstrates a marked increase in the p-Erk/pan-Erk ratio with increasing MOI of CA-MEK1 adenovirus. (C) CA-MEK1 confers a survival advantage in CD133⁻ cells after TGF-β stimulation (2 × 10⁴ cells/cm² plated annexin V/PI FACS analysis, n = 3, 50,000 events/replicate + standard error). #1,#2,#3P < 0.05.

To confirm that CA-MEK1 is capable of antagonizing TGF-β–induced apoptosis in CD133⁻ cells from CSC clone lines, we used adenoviral infection with MOI 50. Both CD133⁻ and CD133⁺ cells were infected with either β-Gal or CA-MEK1 adenovirus for 24 hours. As shown in Fig. 8C, the number of apoptotic cells was significantly increased in CD133⁻ cells infected with β-Gal adenovirus 12 hours after TGF-β stimulation compared with CD133⁻ cells infected with CA-MEK1 adenovirus (β-Gal 15.3 ± 1.6% versus CA-MEK1 2.2 ± 0.2% [P< 0.001]). CD133⁺ cells infected with either β-Gal or CA-MEK1 demonstrated a relative resistance to TGF-β-induced apoptosis compared with CD133⁻ cells in either the β-Gal or CA-MEK1 group. To further test the CD133⁻ cell survival after CA-MEK1 adenoviral infection, we tested the cell viability using the XTT assay. CD133⁻ cell viability was significantly reduced in cells infected with β-Gal adenovirus after TGF-β treatment, and CD133⁻ cell viability remained at pretreatment levels in the CA-MEK1 adenovirus-infected cells (Supplementary Fig. 7). These results further indicated that superactivated MAPK pathway signaling in CD133⁺ cancer stem cells provides a protective role against TGF-β–induced apoptosis.

Induction of CD133 Expression in Cells with Superactivated Erk

Our final series of experiments tested the potential relationship between CD133 expression and the superactivated MAPK pathway. The CA-MEK1 adenoviral construct was able to induce a five-fold increase in CD133 expression compared with the β-Gal adenovirus, as measured by FACS CD133 cell surface staining (β-Gal 0.5 ± 0.3% versus CA-MEK1 2.3 ± 0.3% [P < 0.05]) (Supplementary Fig. 8). Inhibition of MEK-1 with PD98059 had no significant effect on CD133 expression (data not shown).

Discussion

Our previous study demonstrated that CD133⁺ cells represent a liver CSC population in Mat1a^−/− mice.¹¹ Given this work, our primary goal was to determine a mechanism of CD133⁺ CSC survival during chronic injury.

Liver stem cells proliferate during chronic liver injury.²² The majority of HCC develops on this background of chronic injury, such as during chronic hepatitis B or C infection.^1,30 During chronic injury due to viral infection, TGF-β is produced by non-parenchymal cells and acts as a negative regulator of hepatocyte proliferation.³¹ Under this circumstance, liver stem cells with an ability to antagonize the cell growth inhibitory or apoptotic effects of TGF-β are potentially able to repopulate the damaged liver. Although deregulated TGF-β has been studied in HCC progression,¹⁶ the exact role of TGF-β in the homeostasis of liver progenitor cells remains largely unknown.

In some studies, hepatic progenitor cells display resistance to proapoptotic and antiproliferative effects of endogenous TGF-β compared with the well-differentiated mature hepatocytes.³² In fetal hepatocytes, Sanchez et al.³³ observed that 50% of the cells survive despite increasing concentration of TGF-β. These surviving fetal liver cells were less differentiated with respect to liver-specific transcription factor activity, were still able to undergo growth arrest in response to TGF-β, and appeared completely unresponsive to TGF-β–induced apoptosis.

In terms of progression from chronic injury to HCC, several studies have indicated that a significant subset of HCC originates from liver CSCs. In studies of established HCC cell lines such as Huh7, only cells expressing CD133 are capable of expanding and forming tumors in vivo.³⁴ Given that CD133 is a marker of oval cells²² as well as liver CSCs, we postulate that these tumorigenic CD133⁺ CSCs, isolated from Mat1a^−/− mice, are derived from liver stem cells.¹¹

In our current study, we used Mat1a^−/− mice to demonstrate that tumorigenic CD133⁺ liver progenitor cells have acquired a survival advantage against TGF-β–induced apoptosis. Compared with CD133⁻ cells, we did not see a significant difference in the cell growth inhibition by TGF-β in CD133⁺ cells. In addition, when comparing CD133⁺ to CD133⁻ cells, we also did not observe a significant change in mRNA levels for the cell cycle proteins p15, p21, cyclin D1, and c-myc.

Furthermore, in both CD133⁻ and CD133⁺ cells, the inhibitory proteins Smad6/7 are not detectable; and there was a very low level of Smad6/7 mRNA expression. In one study, rat oval cells were less sensitive to TGF-β–induced cell growth inhibition due to the up-regulated Smad6.¹⁹ This study suggests that inhibitory Smad6 plays a critical role in the regulation of cell proliferation in oval cells. In our study, the very low levels of Smad6/7 mRNA and undetectable protein in Mat1a^−/− CSC clone lines may explain why both CD133⁻ and CD133⁺ cells are equally sensitive to TGF-β growth arrest. Furthermore, it has been reported that TGF-β–mediated apoptosis is not dependent on the Smad pathway,³⁵ indicating that the cell growth inhibitory and apoptotic effects of TGF-β are mediated by distinct signaling pathways.

In this study, up-regulated MAP kinase signaling was associated with C133⁺ cell survival against TGF-β–induced apoptosis. Up-regulated MAPK signaling has been well documented in HCC,³⁶ indicating that Erk activation is important for liver cancer cell proliferation and survival. In chronic viral hepatitis, hepatitis C virus core protein and hepatitis B × gene protein can activate the Ras/MAPK/Erk pathway and play critical roles in the initiation and development of HCC.^37,38 Alterations in the MAPK pathway with elevated Erk levels have been described in Mat1a deletion mice, which develop HCC spontaneously by 18 months.³⁹ Moreover, the specific inhibitors of MEK1/2, PD98059, and U0126 and Erk1/2 antisense oligonucleotide can inhibit HCC cellular proliferation in a dose-dependent manner.⁴⁰ However, the dysregulation of Ras/MAPK/Erk signals in the initiation and maintenance of liver CSCs remains largely unknown. Interestingly, a recent report indicates that mitogen-activated protein kinase 2, a member of the MAPK/Erk pathway, was up-regulated in prostate progenitor cells expressing CD133.⁴¹

We previously demonstrated increased k-Ras expression within specific populations of tumorigenic stem cells isolated from Mat1a-deleted mice.¹¹ We now demonstrate that activated MAPK signaling appears to confer a relative resistance to TGF-β–induced apoptosis in CD133⁺ cells compared with CD133⁻ cells. For the first time, our results demonstrate that increased MAPK signaling results in an overactivated Erk1/2 specifically in CD133⁺ CSCs. Our working hypothesis is that CD133⁺ CSCs play a critical role in the tumorigenesis through an acquired survival advantage: resistance to TGF-β–mediated apoptosis. Furthermore, our observations suggest that aberrant MAPK/Erk pathway in liver cancer stem cells may play a pivotal role in the initiation and development of HCC.

The molecular mechanism of TGF-β and MAP kinase signals in the homeostasis of liver stem cells still needs to be elucidated. Our future work will focus on the mechanism underlying the transcriptional/translational regulation of CD133 in liver stem cells and to define molecular therapeutic targets of liver CSCs.

Abbreviations

CA-MEK1

constitutively active form of MEK1

CSC

cancer stem cell

Erk

extracellular signal-regulated kinase

FACS

flow cytometry; β-Gal, β-galactosidase

HCC

hepatocellular carcinoma

MAPK

mitogen-activated protein kinase

MAT

methionine adenosyltransferase

MEK1

mitogen-activated protein kinase kinase 1

MOI

multiplicity of infection

mRNA

messenger RNA

PBS

phosphate-buffered saline

PCR

polymerase chain reaction

p-Erk

phosphorylated extracellular signal-regulated kinase

PI

propridium iodine

RT-PCR

reverse-transcription polymerase chain reaction

TGF-β

transforming growth factor β

XTT

2,3-bis(2-methoxy-4-nitro-5-sulfophenyl)-2H-tetrazolium-5-carboxanilide.