Expansion of CD133‐Expressing Liver Cancer Stem Cells in Liver‐Specific Phosphatase and Tensin Homolog Deleted on Chromosome 10‐Deleted Mice

PTEN (phosphatase and tensin homolog deleted on chromosome 10) is a lipid phosphatase that regulates mitogenic signaling pathways, and deficiency of PTEN results in cell proliferation, survival, and malignancy. Murine liver‐specific Pten deletion models develop liver malignancy by 12 months of age. Using this model, we describe a population of CD133+ liver cancer stem cells isolated during the chronic injury phase of disease progression and before primary carcinoma formation. We performed immunohistochemistry and flow cytometry isolation using livers from 3‐ and 6‐month‐old PtenloxP/loxP; Alb−Cre+ mice (mutants) and controls. CD133+CD45− nonparenchymal (NP) cells were analyzed for gene expression profile and protein levels. Single CD133+CD45− oval cells were isolated for clonal expansion and tumor analysis. Cultured and freshly isolated liver CD133+CD45− and CD133−CD45− NP cells were injected into immune‐deficient and immune‐competent mice. In mutant mice, the NP fraction increased in CD133+CD45− cells in 3‐ and 6‐month‐old Pten‐deleted animals compared with controls. Clone lines expanded from single CD133+CD45− cells demonstrated consistent liver progenitor cell phenotype, with bilineage gene expression of hepatocyte and cholangiocyte markers. CD133+ cells from expanded clone lines formed robust tumors in immune‐deficient and immune‐competent mice. Furthermore, freshly isolated CD133+CD45− NP liver cells from 6‐month‐old mutants formed tumors in vivo, and CD133−CD45− NP cells did not. Consistent with a cancer stem cell phenotype, CD133+ cells demonstrate resistance to chemotherapy agents compared with CD133− cells. CD133+CD45− nonparenchymal cells from chronic injury PtenloxP/loxP; Alb−Cre+ mice represent a bipotent liver progenitor cell population with cancer stem cell phenotype. STEM CELLS 2009;27:290–299


INTRODUCTION
Hepatocellular carcinoma (HCC) is one of the most common solid tumors in the world, rated fifth in incidence and third in mortality [1]. As much as 50% of HCC can be defined as having a progenitor cell phenotype, with markers of both hepatocytes and cholangiocytes [2]. Although these progenitor-based HCCs have a more aggressive phenotype, the exact cellular origin of these progenitor-based tumors is unknown [3].
During chronic liver injury and cirrhosis, two models of carcinoma progression have been proposed: HCC and cholangiocarcinoma (CC) arise from mature hepatocytes and cholangiocytes, respectively, or HCC and CC arise from a common progenitor cell [2][3][4]. Current research describes the liver progenitor cell, or oval cell (OC), as a small cell, within the nonparenchymal (NP) fraction of the liver, that resides near the terminal bile ducts, at the hepatocyte-cholangiocyte interface [5,6]. Although the exact role of oval cells in liver regeneration is unknown, oval cells proliferate during chronic injury [7].
In HCC and CC, deregulation of the epithelial growth factor receptor (EGFR)/phosphoinositide 3Ј-kinase (PI3K)/AKT signal pathway has been clearly demonstrated as a cause of cell cycle progression and cancer formation [8 -11]. Human HCC demonstrates increased expression of EGFR ligands in HCC tissue compared with normal hepatocytes in surrounding liver tissue [12,13]. Recently, chronic hepatitis B virus and hepatitis C virus infection-associated HCC was linked to upregulated EGFR/PI3K/AKT signaling [9]. Furthermore, downregulation Author contributions: C.B.R.: conception and design, financial support, collection and assembly of data, data analysis and integration, manuscript writing, and final approval of the manuscript; W.D.: conception and design, collection and assembly of data, data analysis and integration, and final approval of the manuscript; L.H.: collection and assembly of data, data analysis and integration, and final approval of the manuscript; B.S.: conception and design, financial support, data analysis and integration, manuscript writing, and final approval of the manuscript. of PTEN (phosphatase and tensin homolog deleted on chromosome 10), the primary negative regulator for the EGFR/PI3K/ AKT signaling pathway, is associated with many human HCC patients [11,14].
In human disease, PTEN is a tumor suppressor mutated in a wide range of cancers [15]. PTEN is a phosphoprotein/phospholipid dual-specificity phosphatase that antagonized the activity of PI3K [16]. PTEN acts as a tumor suppressor in most cells, and loss of PTEN leads to constitutive activation of AKT and resistance to apoptosis [17,18]. Specific downregulation of the tumor suppressor PTEN has been demonstrated in human patients and validated with murine models of HCC [14,15]. In mice, liver-specific loss of PTEN (Pten loxP/loxP ; AlbϪCre ϩ ) leads to the development of both HCC and CC by 1 year, indicating the potential expansion of a malignant stem cell population during disease progression [19,20]. In human patients, PTEN downregulation is associated with a poor prognosis [14].
Using the murine model of liver-specific PTEN loss, we intend to define a population of liver cancer stem cells isolated during chronic liver injury and before primary carcinoma formation. On the basis of prior studies, the Pten loxP/loxP ; AlbϪCre ϩ mouse model develops chronic steatosis and injury prior to the onset of liver carcinoma. In a recent study, we demonstrated that CD133ϩ oval cells isolated from premalignant methionine adenosyltransferase 1A (Mat1a)-deficient mice contributed to a tumorigenic phenotype when grafted [21]. These liver cancer stem cells from Mat1aϪ/Ϫ were identified using CD133 expression and isolated before primary hepatocellular carcinoma development. Using human HCC cell lines, other groups have demonstrated that CD133 expression defines an aggressive cancer stem cell phenotype [22,23]. Many tumors can be attributed in part to cancer stem cells, which are resistant to chemotherapy and can account for chemotherapy failure [24].
Here, we tested the hypothesis that deletion of Pten leads to the proliferation and enrichment of liver progenitor cells. Fluorescence-activated cell sorting (FACS) isolation of CD133ϩ CD45Ϫ liver nonparenchymal cells from premalignant 3-and 6-month-old Pten loxP/loxP ; AlbϪCre ϩ mice identified cells with an oval cell phenotype. After clonal expansion of single CD133ϩCD45Ϫ cells, 100% maintained expression of hepatoblast/oval cell-associated genes, confirming the stem cell nature of these CD133ϩ cells. Consistent with cancer stem cells, CD133ϩCD45Ϫ cells demonstrated resistance to chemotherapy compared with CD133ϪCD45Ϫ cells. Furthermore, CD133 expression was highly correlated with tumor formation in both cultured and freshly isolated cells.

Pten loxP/loxP ; Alb؊Cre ؉ Mice and Nude Mice
Mice were fed ad libitum a standard diet (Harlan Teklad irradiated mouse diet 7912; Harlan Laboratories, Madison, WI, http://www. harlan.com) and housed in a temperature-controlled animal facility with 12-hour light-dark cycles. Animals were treated humanely, and all procedures were in compliance with our institutions guidelines for the use of laboratory animals and approved by the Institutional Animal Care and Use Committee. Three-and 6-month-old Pten loxP/loxP ; AlbϪCre ϩ mice (mutants) and wild-type (WT) littermates were used for all experiments, as described [19]. Six-weekold nude mice (Jackson Laboratory, Bar Harbor, ME, http://www. jax.org) and WT mice were used for tumor formation analysis.

FACS Analysis of the Oval Cell-Enriched NP Fraction
One million freshly isolated liver NP cells, RBC-and CD45depleted, were resuspended in PBS. Alternatively, cells in culture were trypsinized for 3 minutes and washed once in culture medium and a second time in PBS. Following Fc blocking, the following combinations of FACS antibodies were added: CD45 fluorescein isothiocyanate (FITC), phycoerythrin (PE), and allophycocyanin (APC); CD34 FITC and PE; Thy 1.2 FITC and PE; c-Kit FITC, Sca-1 FITC, and PE; CD49f PE (BD Pharmingen); and CD133 FITC and PE (eBioscience Inc., San Diego, http://www.ebioscience. com). Cells were then incubated at 4°C for 30 minutes. Cells were washed with PBS prior to analysis using a FACSCalibur (BD Biosciences). Cell isolation was conducted on a FACSVantage (BD Biosciences). Compensation for FITC, PE, and APC was performed using compensation beads (BD Pharmingen). Analysis was done using the Flow-Jo program (Tree Star, Ashland, OR, http://www. treestar.com). Positive and negative gates were determined using IgG-stained and unstained controls.

Real-Time Polymerase Chain Reaction
Cells were pelleted at 200g for 5 minutes, and total RNA was extracted using the RNeasy Kit (Qiagen, Hilden, Valencia, CA, http://www1.qiagen.com) per the manufacturer's protocol. RNA was quantified using an ND-1000 spectrophotometer (NanoDrop, Wilmington, DE, http://www.nanodrop.com). Two hundred nanograms of purified mRNA per 20-l reaction volume was used to construct first-strand cDNA using an oligo(dT) reverse transcriptase kit at 37°C for 60 minutes (Invitrogen, Carlsbad, CA, http://www. invitrogen.com). Real-time experiments were conducted by use of an ABI Prism 7700 Thermal Cycler and TaqMan Universal PCR Master Mix (Applied Biosystems, Foster City, CA, http://www. appliedbiosystems.com). Housekeeping genes included ␤-Actin, Hypoxanthine Phosphoribosyl-transferase 1, and Ubiquitin C, and a geometric mean of housekeeping C T values was used for all ⌬⌬C T calculations [27]. The relative level of expression was calculated for the genes Albumin, Ck19, Hnf4␣, c-Met, Epithelial growth factor receptor (Egfr), Cyclin D1, Kras, Nras, Survivin, c-Myc, Nanog, Oct4, and Abcg2. These genes were assessed using real-time polymerase chain reaction (PCR) primer/probe sets (Applied Biosystems). Amplification efficiency was determined by the ⌬⌬C T method of the amplification plots [28].

Gene Expression of CD133؉CD45؊ Oval Cells
RNA was extracted directly from the culture well using an RNeasy Kit (Qiagen). Per 20-l reaction volume, 200 ng of purified RNA was used in the synthesis of first-strand cDNA as described. All primers were selected in two separate exons to distinguish cDNA from possible contaminating genomic DNA, and PCR conditions were used as described [26]. (Primer sequences are given in supporting information Table 1).

Single-Cell Analysis
For single-cell experiments, 4,6-diamidino-2-phenylindole (DAPI) was used as a marker of cell viability. Single CD133ϩ CD45ϪDAPIϪ OCs (n ϭ 192) were isolated using a FACSVantage set for single cell purity, and single cells were robotically plated directly into a flat-bottomed, 96-well laminin-coated plate (BD Biosciences) with 100 l of medium per well as described [30]. After 24 hours, an additional 100 l of medium was added to each well. Approximately half of the medium was replaced after 1 week. After 3 weeks in culture, single-cell derived colonies that filled Ͼ50% of the well area had RNA extracted for reverse transcription (RT)-PCR analysis (n ϭ 5 for PCR analysis).

Tumor Formation Assay
CD133ϩCD45Ϫ and CD133ϪCD45Ϫ cells were isolated from culture and counted with Trypan blue exclusion to determine numbers of live cells. Cells were resuspended in PBS:Matrigel (1:1) for transplant at a concentration of 1 ϫ 10 5 to 2 ϫ 10 6 live cells per 50 l. Six-week-old immune-deficient nude mice and WT mice were injected subcutaneously (SQ). Mice were sacrificed after defined intervals. For freshly isolated cell transplant, CD133ϩCD45Ϫ and CD133ϪCD45Ϫ cells were isolated from 3-and 6-month-old Pten loxP/loxP ; AlbϪCre ϩ mice. Cells were counted, and 1 ϫ 10 6 cells were immediately transplanted SQ into nude mice.

Statistical Analyses
A paired, two-tailed Student's t test was used when comparing two groups. A p value of less than .05 was considered significant. Analysis of variance was used for comparison of multiple groups, followed by pairwise multiple comparison procedures (Systat Software Inc., Richmond, CA, http://www.systat.com).

Expansion of Oval Cells with Age in Pten loxP/loxP ; Alb؊Cre؉ Mice
Three-and 6-month-old Pten loxP/loxP ; AlbϪCreϪ (control WT) and Pten loxP/loxP ; AlbϪCre ϩ (mutant) mice were analyzed for degree of injury and OC proliferation using hematoxylin and eosin (H&E) staining and pan-CK FIHC as an OC/bile duct marker [33]. In the control animals, H&E staining revealed no injury, steatosis, or OC proliferation (Fig. 1A, 1B). In the Pten loxP/loxP ; AlbϪCre ϩ mice, H&E staining revealed significant injury and steatosis without evidence of gross or microscopic tumor [19,20]. Further analysis demonstrated increased duct and oval cell proliferation associated with chronic injury (Fig. 1D-1H). Immunohistochemistry for pan-CK, a duct and oval cell marker, revealed increased staining in both 3-and 6-month-old mutant mice compared with control mice, consistent with the overall injury (Fig. 1C, 1F, 1I).
Whole liver tissue lysate from 3-and 6-month-old Pten loxP/loxP ; AlbϪCreϩ mice demonstrated loss of PTEN protein and constitutive phosphorylation of AKT compared with agematched controls (Fig. 1J). PCR amplification of mouse tails from Pten loxP/loxP ; AlbϪCre ϩ mice confirmed Pten deletion, as indicated by excision of the Pten locus (Pten ⌬5 ) (data not shown).

Enrichment of CD133؉ Oval Cells
To further characterize their lineage potential, CD133ϩCD45Ϫ OCs were isolated from 6-month-old Pten loxP/loxP ; AlbϪCre ϩ mice and plated in six-well, laminin-coated tissue culture plates using a fetal hepatoblast medium [29]. Laminin-coated plates allow for further selection of liver stem cells [21]. On day 1, small round cells were observed adhering to the plate. After 1 week in culture, clusters of cells were observed growing, with the majority of the cells having the morphology of cuboidal cells in clusters/sheets (Fig. 2B, 2C). After 1 week in culture, the cells were 90% confluent, and they were trypsinized, replated 1:3, and passaged every 3 days thereafter. After 4 months of replating (passage 25), the basic morphology of these bulk culture cells remained unchanged. Protein cell lysate from CD133ϩCD45Ϫ oval cell bulk culture demonstrated complete loss of PTEN, phosphorylation of AKT, and strong expression of ␣-fetoprotein (␣FP) (Fig. 2D). We also observed single cells or small clusters of cells with alkaline phosphatase expression, a potential marker of progenitor cells (supporting information Fig. 1) [34,35]. Costaining of hepatocyte (␣FP) and cholangiocyte (biliary cytokeratin [CK]) markers was conducted to demonstrate an oval cell phenotype within these bulk cultures. Of the 1,000 cells counted, 72 cells were positive for ␣FP, 63 cells were positive for CK, and 59 cells (6% of the total counted) were positive for both markers.

Single-Cell Isolation Defines Bipotency in CD133؉CD45؊ OCs
To isolate this potential progenitor cell population, single cells were isolated by FACS using automated plating robotics. CD133ϩCD45ϪDAPIϪ cells were selected from early passage CD133ϩ bulk cultures (passage 6). Absence of DAPI is a marker of cell viability, as DAPI stains the nuclei of dead cells. Of 192 individual cells plated, 26 (13.5%) demonstrated colonies growing over Ͼ50% of the well bottom after 2 weeks. These colonies were replated into six-well plates, and the basic cell morphology matched that of the original culture, with flat, cuboidal cells. After 2 weeks, 100% of analyzed colonies (n ϭ 5) demonstrated gene expression of both hepatocyte (Albumin) and cholangiocyte (Ck19) markers, using RT-PCR (Fig. 3A). All CD133ϩ derived clone lines demonstrated expression of oval cell/hepatoblast-associated genes (␣Fp, Hnf1␣, Hnf3␤, and Abcg2) (Fig. 3A) [36] and expression of growth factor receptors c-Met and Egfr. Lastly, no expression of hematopoietic (Cd45) or satellite cell (␣-Smooth muscle actin) markers were demonstrated in any CD133ϩ derived clone line (Fig. 3A).
Protein lysate from each CD133ϩ derived clone line was analyzed for PTEN, phospho-AKT, pan-AKT, and ␤-Actin. This functional analysis demonstrated complete loss of PTEN protein in all CD133ϩ derived clone lines with constitutive phosphorylation of AKT (Fig. 3B). Pten deletion was confirmed with PCR amplification, as indicated by excision of the PTEN locus (Pten ⌬5 ) in all five CD133ϩ derived clone lines (data not shown). Bipotent cells (defined as ␣FP and CK coexpressing on

Tumor Formation from CD133؉ Derived Clone Lines
To assess the tumor forming ability of CD133ϩ derived clone lines, a tumor model using immune-deficient mice was initially used. Two million cells isolated from each CD133ϩ derived clone line (lines A-E; passage 6 after single-cell isolation) were injected SQ into immune-deficient mice (n ϭ 6 per cell line; five lines total). Four of the five oval cell lines expanded from single CD133ϩ cells formed tumors to various degrees (clone line A, three of six; clone line B, two of six; clone line C, one of six; clone line E, three of six; Fig. 4A-4C). Tumor histology revealed mixed populations with dysplastic columnar and cuboidal epithelial cells, with similar histology in all tumors (Fig.  4D-4I). Clone line D did not contribute to tumor formation in this xenograft model. FACS analysis of tumor cells demonstrated a high level of CD133 expression (Fig. 5A, 5B). PCR analysis of tumor cells demonstrated strong expression of liverspecific genes Albumin and Ck19 (Fig. 5C), an indication of bipotential differentiation from the grafted CD133ϩCD45Ϫ cells. Western blot of tumor cells confirmed consistent loss of PTEN protein, constitutive phosphorylation of AKT, and high levels of CD133 protein (Fig. 5D).
We further compared the expression of different progenitor and liver cell markers between the tumorigenic clones (clones A, B, C, and E) and the nontumorigenic clone, clone D. Realtime PCR analysis demonstrated that the tumorigenic lines had increased expression for the progenitor cell markers Survivin, cMyc, Egfr, and CD133 and lower expression of the differentiated cell markers CK19 and Hnf4␣, compared with the nontumorigenic clone line (Fig. 6).
CD133Ϫ cells (CD133ϩ, three of six tumors; CD133Ϫ, zero of six tumors; supporting information Table 3). Furthermore, 1 million cultured CD133ϩ cells formed tumors in immune-competent WT mice (two of six; passage 14; Fig. 7B, 7C; supporting information Table 3), and CD133Ϫ cells did not (zero of six; supporting information Table 3).  Given that cells in vitro may acquire additional transformation events with passage, we isolated CD133ϩCD45Ϫ and CD133ϪCD45Ϫ NP liver cells from 3-and 6-month-old Pten loxP/loxP ; AlbϪCre ϩ mice and injected 1 million cells directly (without culture) into nude mice. Only the CD133ϩ fraction from 6-month-old mutants gave rise to tumors (two of six; Fig. 7D, 7E; supporting information Table 3). The CD133Ϫ fraction from the same set of mutant livers did not contribute to tumor formation (zero of eight; supporting information Table 3). None of the CD133ϩ and CD133Ϫ cells from 3-month-old mutants gave rise to tumors (CD133ϩ, zero of four; CD133Ϫ, zero of four; supporting information Table 3). The tumors that arose from cultured and freshly isolated CD133ϩ cells demonstrated both hepatocyte-like and duct-like structures, indicating a potential stem cell-initiated tumor (Fig. 7C, 7E; supporting information Fig. 2).
To determine whether the CD133ϩ cells from each clone line have additional CSC characteristics, we analyzed cell viability after chemotherapy treatment. CD133ϩ and CD133Ϫ cells from the tumorigenic clone lines were exposed to IC90, IC75, and IC50 concentrations of doxorubicin and 5-FU. As demonstrated, CD133ϩ cells have significant survival at each concentration for these two chemotherapy drugs compared with CD133Ϫ cells (supporting information Figs. 3, 4).

DISCUSSION
The ability to identify and isolate adult stem cells during premalignant liver injury is an essential first step in studying liver cancer stem cells [37]. In models of liver damage, with hepatic infiltration of inflammatory cells, separation of hematopoietic cells from liver cells is critical prior to OC analysis [26]. In terms of defining a reliable OC immunophenotype, the surface markers previously attributed to OCs, such as c-kit and CD34, are traditionally assigned to hematopoietic stem cells [38]. Recently, we defined a population of cells that express CD133 but not the hematopoietic marker CD45. The cell surface marker CD133 has been associated with both pluripotent stem cells and cancer stem cells of epithelial origin [39]. Upregulated CD133 expression was detected in rat OCs, isolated by size, as part of a microarray screen [40]. Our group used single-cell gene expression analysis to confirm the bilineage hepatic potential of CD133ϩCD45Ϫ OCs isolated from murine liver injury models [7].
In this study, we used CD133ϩCD45Ϫ as marker to identify a population of cells that are proliferating during premalignant liver injury in the liver Pten deletion mouse model. In this model, we and other have documented chronic liver injury and steatosis prior to the development of primary liver carcinomas [19,20]. Interestingly, analysis of this murine model of Pten deletion reveals a bilineage origin of tumors, with the development of both HCC and CC [41]. Three possible cellular origins of HCC and CC have been proposed previously: (a) the mature hepatocyte causing HCC; (b) the mature cholangiocyte causing CC; and (c) the bipotent cancer stem cell causing HCC, CC, and mixed tumors [3]. Tumors with progenitor or OC phenotypes represent 25%-50% of HCC and have a more aggressive phenotype [2]. Despite the effort of multiple international investigators, a defined, specific population of liver cancer stem cells remains elusive [24,42].
The cell surface marker CD133 has been used to identify populations of liver cancer stem cells. Using HCC cell lines, CD133 expression identifies an aggressive phenotype of cancer stem cell [22,23]. Furthermore, in multiple human gastrointestinal cancer cell lines, CD133ϩ cells are associated with the side-population cancer stem cell phenotype, which has ABCG2 membrane-pump expression and resistance to chemotherapy. Our group has recently described CD133ϩ liver cancer stem cells isolated during premalignant liver injury in methionine adenosyltransferase 1A-deficient mice [21]. Using the same basic techniques described, we isolated the CD133ϩCD45Ϫ oval cells from the liver Pten deletion model. This CD133ϩCD45Ϫ oval cell population is more resistant to doxorubicin and 5-FU treatment in culture than the CD133Ϫ cell population, supporting the concept that CD133 expression identifies a cancer stem cell population. We demonstrated that these cells are bipotential, with markers of both hepatocytes and cholangiocytes. To further determine whether deletion of Pten in this specific CD133ϩCD45Ϫ cell population is sufficient to convert OCs to liver cancer stem cells, we performed multiple cell transplant studies. The results from these experiments demonstrate that expression of CD133 is associated with the tumorigenic cell fraction. The clone line with the lowest CD133 expression (clone D) was unable to form tumors in nude mice, further supporting the notion that CD133 cell surface expression is a cancer stem cell marker in liver.
In liver cancer, which is the third cause of cancer mortality worldwide, upregulation of the EGFR/PI3K/AKT signaling plays a key role in cell growth regulation [43]. The molecular interaction of this signaling pathway with the most common etiological agents of HCC, hepatitis B and C viruses, is also established [44,45]. As the negative regulator of the EGFR/ PI3K/AKT signaling pathway, PTEN loss is unequivocally correlated with human liver cancer.

SUMMARY
In summary, using the liver Pten deletion mouse model, we demonstrated that deletion of Pten leads to the proliferation of the liver OC population defined as CD133ϩCD45Ϫ NP cells. Given that tumors that are formed from CD133ϩ cells demonstrate both hepatocyte-like and duct-like structures, the CD133ϩ progenitor population is a potential cellular origin of the bilineage tumor phenotype (HCC and CC) observed in this Pten deletion mouse model. This data suggest that the association between high malignant potential/poor prognosis and PTEN mutation observed with human liver cancer may have a progenitor cell mechanism. Future studies focusing on ablation of this liver cancer cell population are needed to test the therapeutic potential of targeting this cell population.

DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest. , with H&E staining of tumor (ϫ20 objective). CD133Ϫ fraction from same 6-month-old mutants did not contribute to tumor formation (zero of eight). (Summary is given in supporting information Table 3.)