SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Increasing evidence has revealed the importance of cancer stem cells (CSCs) in carcinogenesis. Although liver CSCs have been identified in hepatocellular carcinoma (HCC) cell lines, no data have shown the presence of these cells in human settings. The present study was designed to delineate CSCs serially from HCC cell lines, human liver cancer specimens to blood samples, using CD90 as a potential marker. The number of CD90+ cells increased with the tumorigenicity of HCC cell lines. CD45CD90+ cells were detected in all the tumor specimens, but not in the normal, cirrhotic, and parallel nontumorous livers. In addition, CD45CD90+ cells were detectable in 90% of blood samples from liver cancer patients, but none in normal subjects or patients with cirrhosis. A significant positive correlation between the number of CD45CD90+ cells in the tumor tissues and the number of CD45CD90+ cells in the blood samples was identified. CD90+ cells sorted from cell lines and CD45CD90+ cells from the tumor tissues and blood samples of liver cancer patients generated tumor nodules in immunodeficient mice. Serial transplantation of CD90+ cells from tumor xenografts generated tumor nodules in a second and subsequently third batch of immunodeficient mice. Treatment of CD90+ CSCs with anti-human CD44 antibody induced cell apoptosis in a dose-dependent manner. Conclusion: Identification of CD45CD90+ CSCs in both tumor tissues and circulation suggests that CD45CD90+ could be used as a marker for human liver cancer and as a target for the diagnosis and therapy of this malignancy. (HEPATOLOGY 2008.)

Hepatocellular carcinoma (HCC), the fifth most common cancer in the world, comprises more than 90% of human liver cancers. The incidence of HCC is increasing due to hepatitis B and C viral infection.1–3 Hepatic resection and liver transplantation are the mainstays that may cure HCC, but the 5-year survival rate is mainly dependent on tumor staging at the time of diagnosis.4–6 The majority of HCC patients present with an advanced stage for which chemotherapy and radiotherapy have limited efficacy.7, 8 Although the mechanism of hepatocarcinogenesis has been studied for many years, none of the identified genes or molecules is universally expressed by tumor cells, leading to the suspicion that current studies might have focused only on the heterogenic “end products”—adult tumor cells—but not the “root” cancer stem cells (CSCs).9–14

Early diagnosis and treatment of HCC remain challenging due to lack of highly specific and sensitive markers. The highest sensitivity and specificity of reported genes and molecules for the diagnosis of HCC could only reach 60%-70%.15 If liver CSCs could be identified, it would be reasonable to postulate that the molecules expressed by CSCs would become highly specific and sensitive markers for the diagnosis of liver malignancy. However, there has been no study demonstrating the presence of functional CSCs in tumor specimens, though some reports have shown the presence of these cells in HCC cell lines.16–19 These studies used a marker of hematopoietic stem cells, CD133, to identify CSCs in HCC cell lines. Although CD133+ cells could be detected in some of the tumor tissues by immunohistochemistry,18, 19 no report showed the functional aspects of these cells. In addition, our previous study has reported that CD133 was a marker of circulating endothelial progenitor cells in HCC patients,20 leading to the hypothesis that CD133 might not be sensitive and specific to represent CSCs in HCC.

In this study, we used CD90, which is expressed by hepatic stem/progenitor cells (HSPCs) during liver development, but rarely in the adult liver,21, 22 to identify CSCs in HCC cell lines, tumor specimens, and blood samples from liver cancer patients. By characterizing tumorigenic potentials of CD90+ cells from different sources, we may understand the cellular basis which mediates hepatocarcinogenesis and aggressive growth pattern of human liver cancer.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Cell Lines.

The human cell lines MIHA (an immortalized human hepatocyte cell line),23 HepG2, Hep3B, PLC, Huh7 (ATCC, Manassas, VA), MHCC97L and MHCC97H (HCC cell lines with low and high metastatic properties, respectively24), and a chemically transformed mouse hepatocyte cell line (BNL, ATCC) were maintained as monolayer culture in high glucose Dulbecco's modified Eagle's medium with 10% fetal bovine serum and 1% penicillin (Life Technologies, Carlsbad, CA) at 37°C in a humidified atmosphere of 5% CO2 in air.

Patients and Sample Collection.

Thirty-four patients with liver tumors who had undergone laparotomy (26/34 had curative hepatic resection and 8/34 had liver biopsy) were included in the study. Pathological diagnosis was made according to the histology of tumor specimens or biopsy, and examinations were performed by experienced pathologists. Nineteen patients with cirrhosis (10/19 received liver transplantation) and 19 normal subjects (10/19 were liver donors) were used as controls. The diagnosis of cirrhosis was made via histological findings in liver explants or CT/MRI scans. The majority of HCC (28/34) and cirrhosis patients (16/19) was serum hepatitis B surface antigen–positive. All the tissue and blood samples were obtained from consenting patients, and the study was approved by the Institutional Review Board of the University of Hong Kong.

Tumorous, parallel nontumorous, cirrhotic, and normal liver tissues were harvested at the time of operation and placed in Dulbecco's modified Eagle's medium. The procedure of cell isolation from liver tissues was performed as described previously25 with some modifications. In brief, after digestion with type IV collagenase (100 U/mL) (Sigma-Aldrich, St. Louis, MO) at 37°C for 15 minutes, tissues were squashed and cell suspension was passed through 40 μM nylon mesh. After lysis of red blood cells, cells were counted and analyzed via flow cytometry or cell sorting.

Ten milliliters of ethylene diamine tetraacetic acid blood were collected before operation. Mononuclear cells were isolated from the ethylene diamine tetraacetic acid blood using Ficoll-Paque PLUS (Amersham Bioscience, Buckinghamshire, England) density gradient centrifugation before proceeding to flow cytometry analysis or cell sorting.

Flow Cytometry.

Isolated cells from HCC cell lines, tumor tissues, and blood were labeled with the anti-human antibodies FITC-ESA, PE-CD24, PE-C-kit, PE-KDR, purified and PE-CD90, APC-CD45, APC-CD34 (BD Biosciences Pharmingen, San Diego, CA), PE-Cy5-CD44, PE-Cy5-CXCR4 (eBioscience, San Diego, CA), and PE-CD133 (Miltenyi Biotech Inc., Bergisch Gladbach, Germany) and detected in a FACS Calibur (Becton Dickinson Immunocytometry Systems, San Jose, CA). Appropriate isotypes of irrelevant antibodies were used as controls.

Cell Sorting.

A cocktail phycoerythrin-conjugated mouse anti-human or rabbit anti-mouse CD90 antibody was made according to the manufacturer's instruction (StemCell Technologies Inc., Miami, FL). Cells from PLC, MHCC97L, or BNL cell lines were suspended in phosphate-buffered saline with 2% fetal bovine serum and 0.5 μM ethylene diamine tetraacetic acid, labeled with anti-CD90 antibody cocktail and mixed with magnetic microbeads. The CD90+ and CD90 cells were then separated by a magnet. The purity of sorted cells was evaluated via flow cytometry. To characterize the tumorigenicity of CD45CD90+ cells from the tumor tissues or blood, a 2-step separation was performed, including CD45 depletion and CD90 selection (Miltenyi Biotech Inc.).

Blockade of CD44 Activity.

The sorted CD90+ cells from the MHCC97L cell line were treated with different doses (0, 3, 6, 12, and 24 μg/mL) of anti-human CD44 antibody (International Blood Group Reference Laboratory, Bristol, England) for 24 hours. The cells were then labeled with annexin-V and propidium iodide (BD Biosciences Pharmingen) and detected in a FACS Calibur (Becton Dickinson Immunocytometry Systems).

HCC Xenografts in Immunodeficient Mice and HCC Isografts in BALB/c Mice.

Male BALB/c nude mice, severe-combined immunodeficient (SCID)/Beige mice, and BALB/c mice (4-5 weeks old) were purchased from the Animal Laboratory Unit of the University of Hong Kong. They were maintained under standard conditions, and cared for according to the institutional guidelines for animal care. All experiments were approved by the Committee on the Use of Live Animals in Teaching and Research at the University of Hong Kong. As PLC and MHCC97L cells could generate tumor nodules at the back of nude mice,26 CD90+ and CD90 cells were injected subcutaneously at the 2 sides of a same mouse for easy visualization and comparison. On the other hand, to favor the growth of cells isolated from human tumor tissues and blood, cells were injected orthotopically into the left lobe of livers of SCID/Beige mice as described previously.27 Cells from BNL were also injected orthotopically into the liver of BALB/c mice to generate syngenic HCC model. Animals were sacrificed at the indicated time intervals when tumor nodules were identified on the body surface of the nude mice, or in the liver of SCID/Beige mice or BALB/c mice by laparotomy, otherwise they were kept monitored till the second or third time points.

Histological Studies.

The embedded tissues were cut into 5-μm-thick sections for histological studies by hematoxylin-eosin staining, and immunohistochemistry (IHC) of CD90 and human hepatoyte (Dako, Glostrup, Denmark).28 Sections were incubated with sequential primary and secondary antibodies for 1 hour at room temperature. Nucleus counterstaining was performed using hematoxylin.

Results

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

CD90+ Cells from HCC Cell Lines Displayed Tumorigenic Properties.

Flow cytometry was used to determine CD90 expression in 6 human HCC cell lines, with MIHA as a control. The 6 HCC cell lines harbor different properties of tumorigenicity, with the lowest in HepG2 and the highest in MHCC97H cells. Tumorigenicity was defined as the ability of a certain number of cells (4 × 102 to 1 × 106) to form tumor nodules in immunodeficient mice within a certain time interval (2, 3, and 4 months for cell lines and 3, 4, and 5 months for cells isolated from the tumor tissues or blood samples, respectively). CD90+ cells were detected in all the HCC cell lines, but not in MIHA. A positive correlation between the number of CD90+ cells and tumorigenicity of cell lines was identified (Fig. 1A). The presence of CD90+ cells in both PLC and MHCC97L cell lines was further confirmed by immunofluorescent staining (Fig. 1B).

thumbnail image

Figure 1. (A) Distribution of CD90+ cells in human HCC cell lines, detected by flow cytometry. An increasing number of CD90+ cells were detected in the HepG2, Hep3B, PLC, Huh7, MHCC97L, and MHCC97H cell lines. In contrast, no CD90+ cells were detected in MIHA. (B) Expression of CD90 in PLC and MHCC97L cell lines was confirmed via immunofluorescent staining. Arrows point to CD90+ cells. (Original magnification ×200.) (C) CD90+ cells were sorted by magnetic microbeads, with purity ranging from 80.1% to 83.2%. (D) Two months after cell injection, visible tumor nodules were found at the site with CD90+ cell injection in the nude mice, but not at the site with CD90 cell injection. The histology of generated tumor nodules demonstrated similar structure to that of primary tumor. Immunohistochemistry (IHC) identified scattered and clustered distribution of CD90+ cells in the tumor tissues. Arrows point to CD90+ cells. Magnifications of ×200 and ×400 were used for hematoxylin-eosin staining and IHC, respectively.

Download figure to PowerPoint

CD90+ cells were then sorted from PLC and MHCC97L cell lines, respectively, using magnetic microbeads (Fig. 1C), and injected subcutaneously into the nude mice, with CD90 cells as a control. To serve a positive control of tumorigenicity in PLC and MHCC97L cell lines, 1 × 106 total tumor cells were also injected subcutaneously into the nude mice. Starting from 2 months after cell injection, CD90+ cells from both cell lines generated tumor nodules in the nude mice, with the lowest cell number of 2000, and the formed tumor nodules displayed similar histology to that of the original tumor. In contrast, CD90 cells did not form tumor nodules even with the cell number of 1 × 105 at 4 months. By IHC, diffused and clustered distribution of CD90+ cells was identified in the tumor tissues, with both membrane and cytoplasmic patterns of CD90 expression in the cells (Fig. 1D and Table 1).

Table 1. Tumorigenic Potential of CD90+ Cells from HCC Cell Lines
 No. of Injected CellsNo. of Mice with Tumor Formation/Total No. of Mice with Cell Injection[Range of Tumor Size of the Largest Tumor Nodule (mm)]
2 Months3 Months4 Months
  1. Animals were sacrificed when tumor nodules were identified in the nude mice; otherwise, they were continuously monitored until the second or third time points.

MHCC97L    
 CD90+4 × 1020/50/50/5
 2 × 1033/5 (2–4)2/2 (2–3) 
 1 × 1045/5 (3–7)  
 CD904 × 1020/50/50/5
 2 × 1030/50/50/5
 1 × 1040/50/50/5
 5 × 1040/50/50/5
 Total tumor cells1 × 1065/5 (3–6)  
PLC    
 CD90+8 × 1020/50/50/5
 4 × 1032/5 (2–3)1/3 (2)2/2 (2–3)
 2 × 1045/5 (2–5)  
 CD908 × 1020/50/50/5
 4 × 1030/50/50/5
 2 × 1040/50/50/5
 1 × 1050/50/50/5
 Total tumor cells1 × 1065/5 (3–6)  

Serial transplantation was performed by sorting CD90+ cells from tumor xenografts, and injected subcutaneously into a second batch and subsequently a third batch of nude mice. At the same observation time (2 or 3 months), tumor nodules were also developed by CD90+ cells, and the number of mice with tumor formation was comparable to those with tumor nodules developed by CD90+ cells sorted from cell lines (Table 2).

Table 2. Tumorigenic Potentials of CD90+ Cells from HCC Cell Lines or CD45CD90+ Cells from Tumor Specimens or Blood Samples in Serial Transplantation
 Primary CellsPassage 1Passage 2
2 Months3 Months2 Months3 Months2 Months3 Months
  1. Animals were sacrificed when tumor nodules were identified in the nude mice or in the livers of SCID mice at the indicated time points; otherwise, animals were continuously monitored until the second time point. The number of mice in each passage at the indicated time points represents the number of mice with tumor nodules generated by CD90+ cells sorted from different tumor xenografts of a previous passage versus total number of mice with cell injection.

MHCC97L (CD90+)      
 2 × 1033/52/23/52/23/51/2
 1 × 1045/55/54/51/1
PLC (CD90+)   
 4 × 1032/51/33/50/23/50/2
 2 × 1045/54/50/13/51/2
 Primary CellsPassage 1Passage 2
3 Months4 Months3 Months4 Months3 Months4 Months
Tumor tissues (CD45CD90+)      
 4 × 1036/123/62/51/32/51/3
 8 × 10310/122/24/51/14/51/1
Blood (CD45CD90+)      
 4 × 1030/105/103/51/23/51/2
 8 × 1030/1010/104/51/14/51/1

CD45CD90+ Cells from Tumor Specimens Exhibited Tumorigenic Properties.

Based on the findings in cell lines, CD90 was then used as a marker to characterize CSCs in human liver cancer specimens, including 23 HCCs and 3 intrahepatic cholangiocarcinomas (Cholangio Ca). Because CD90 was also expressed by some lymphocytes, a combination of CD45CD90+ was used to define nonlymphatic CD90+ cells in the tumor tissues. Flow cytometry detected a very low number (0%-0.05%) of CD45CD90+ cells in the normal, cirrhotic, and parallel nontumorous livers. In contrast, CD45CD90+ cells were detected in all the tumor tissues, but not in the normal, cirrhotic, and parallel nontumorous livers (Fig. 2A and Table 3). By IHC, CD90+ cells were detected in the tumor tissues, with both scattered and clustered patterns (Fig. 2B). Similar to our findings, the existence of CD90+ cells has been reported in approximately 0.01% in the normal adult liver,29 and is highest in the neoplastic tissues.30 Further multimarker analysis demonstrated that certain proportions of CD45CD90+ cells in the tumor tissues concomitantly expressed stem cell markers such as CD133, ESA, CXCR4, CD24, KDR, and CD44 (Supplementary Table 1).

thumbnail image

Figure 2. (A) Flow cytometry detected small numbers (0%-0.05%) of CD45CD90+ cells in the normal, cirrhotic, and nontumorous tissues, whereas a distinct population of CD45CD90+ cells was detected only in the tumor tissues. The dot plots represent 1 sample in each group, respectively. (B) Localization of CD90+ cells in the tumor tissues was further confirmed via immunohistochemistry (IHC). Arrows point to CD90+ cells. (Original magnification ×400.) (C) CD45CD90+ cells were sorted from the tumor tissues, with purity ranging from 78.9% to 84.8%. The injected CD45CD90+ cells from the tumor tissues generated tumor nodules in the liver of SCID/Beige mice. The histology of tumor tissue showed a moderately differentiated HCC, which was similar to that of primary tumor. The white arrow points to a tumor nodule. IHC detected extensive positive signals of human hepatocyte antigen, and scattered and clustered distribution of CD90+ cells in the generated tumor tissue. Black arrows point to hepatocyte antigen-positive and CD90+ cells. Magnifications of ×200 and ×400 were used for hematoxylin-eosin staining and IHC, respectively. (D) CD90+ cells were sorted from the MHCC97L cell line and treated with different doses (0-24 μg/mL) of anti-human CD44 antibody for 24 hours. Blockade of CD44 activity induced CD90+ cell apoptosis in a dose-dependent manner.

Download figure to PowerPoint

Table 3. Distribution of CD45CD90+ and CD45CD133+ Cells in Tissue and Blood Samples from Normal Subjects and Patients with Cirrhosis and Liver Cancer
GroupsType of SamplesNo. of CasesAge (Years)Sex (M:F)Tumor Size (cm)No. of Positive Cells (% in Gated Cells)
CD45CD90+CD45CD133+
  1. The number of positive cells was expressed as median percentage (range).

NormalTissue1050 (14–62)7:3 0.05 (0–0.3)0.2 (0–0.76)
CirrhosisTissue1049 (47–64)7:3 0.08 (0–0.41)0.4 (0–1.13)
Liver cancers (HCC and Cholangio Ca)Nontumor1650 (47–69)13:3 0.06 (0–0.66)0.56 (0–1.57)
 Tumor2651.5 (25–75)19:75.4 (1.3–16)2.51 (0.42–8.57)1.13 (0–2.41)
      CD45CD90+CD45CD133+
NormalBlood1949 (14–72)13:6 00.01 (0–0.02)
CirrhosisBlood1955 (33–73)13:6 00.01 (0–0.03)
Liver cancers (HCC and Cholangio Ca)Blood3450 (25–75)26:8 0.3 (0–1.2)0.2 (0–0.75)

CD45CD90+ cells were then sorted from the tumor tissues using magnetic microbeads and were injected orthotopically into the livers of SCID/Beige mice. Three months after cell injection, 6 of 12 mice with CD45CD90+ injection formed tumor nodules at the cell number of 4000. When the observation time was extended to 4 months, 3 more mice developed tumor nodules at the cell number of 4000, while 1 mouse developed a tumor nodule at the cell number of 1000. However, further extension of observation time to 5 months did not increase the number of tumor-bearing mice. When the cell number was increased to 8000, 100% tumor development in SCID/Beige mice was observed at 4 months. The generated tumor nodules demonstrated similar histology to that of the primary tumor. On the other hand, the sorted CD45CD90+ cells from normal, cirrhotic, and nontumorous tissues did not generate tumor nodules in SCID/Beige mice. Moreover, CD45CD90 cells from the tumor tissues also did not develop tumor nodules, even with the cell number of 100,000 at 5 months (Table 4). By IHC using anti-human hepatocyte antibody, the majority of tumor cells were stained positive. In addition, IHC detected both scattered and clustered distribution of CD90+ cells in the generated tumor tissues (Fig. 2C).

Table 4. Tumorigenic Potential of CD45CD90+ Cells from Human Liver Cancer Specimens and Blood Samples
Tumor tissueNo. of Injected CellsNo. of Mice with Tumor Formation/Total No. of Mice Injected with Cells from Different Patient Samples [Range of Tumor Size of the Largest Tumor Nodule (mm)]
3 Months4 Months5 Months
  1. Animals were sacrificed when tumor nodules were identified in the mouse liver at the indicated time points by laparotomy; otherwise, they were continuously monitored until the second or third time points.

 CD45CD90+1 × 1030/121/12 (2)0/11
 4 × 1036/12 (1.8–7)3/6 (1.5–2)0/3
 8 × 10310/12 (2.5–8)2/2 (2.4–3) 
 CD45CD901 × 1030/120/120/12
 5 × 1030/120/120/12
 1 × 1040/120/120/12
 1 × 1050/120/120/12
 Total CD45 cells (1 × 106)10/12 (3–9)2/2 (1.5–2) 
Blood    
 CD45CD90+1 × 1030/100/100/10
 5 × 1030/105/10 (2–2.5)0/5
 1 × 1040/1010/10 (1.5–3.5) 
 CD45CD901 × 1030/100/100/10
 5 × 1030/100/100/10
 1 × 1040/100/100/10
 1 × 1050/100/100/10
 Total CD45 cells (1 × 106)0/1010/10 (2–4) 

CD90+ cells were then isolated from the tumor xenografts and injected into the livers of a second batch and subsequently a third batch of SCID/Beige mice. At the same observation time (3 or 4 months), these cells displayed similar tumorigenicity to the CD45CD90+ cells sorted from tumor specimens (Table 2).

Comparable to the findings in tumor tissues, the majority of CD90+ cells in HCC cell lines also concomitantly expressed CD44 (Supplementary Table 2). To further explore the role of CD44 in cell survival, CD90+ cells were sorted from MHCC97L cell line and treated with different doses of anti-human CD44 antibody. Blockade of CD44 activity significantly induced death of CD90+ cells in a dose-dependent manner (Fig. 2D).

CD45CD90+ Cells Were Detectable in the Circulation of Liver Cancer Patients.

Because tumor recurrence rate is high in HCC, even after curative hepatic resection or liver transplantation, it is reasonable to postulate that CSCs might be present in the circulation. To test this hypothesis, we assessed the distribution of CD45 CD90+ cells in blood samples via flow cytometry. A distinct population of CD45CD90+ cells was identified in 31 out of 34 blood samples from liver cancer patients, but not in normal controls or patients with cirrhosis (Fig. 3A and Table 3). In addition, a significant positive correlation between the number of CD45 CD90+ cells in the tumor tissues and CD45CD90+ cells in blood samples was identified (γ = 0.519, P = 0.007) (Fig. 3B).

thumbnail image

Figure 3. (A) Flow cytometry identified CD45CD90+ cells only in blood samples from liver cancer patients. The dot plots represent 1 sample in each group. (B) The number of CD45CD90+ cells in the tumor tissues was positively correlated with the number of CD45CD90+ cells in the circulation (γ = 0.519, P = 0.007; linear regression analysis). (C) CD45CD90+ cells were sorted from blood samples of liver cancer patients, with purity ranging from 81.9% to 84.5%. (D) The sorted CD45CD90+ cells, but not CD45CD90 cells, formed tumor nodules in the liver of SCID/Beige mice. The histology of tumor showed a poorly differentiated HCC, with structure similar to that of the primary tumor. Immunohistochemistry (IHC) detected strong positive signals of human hepatocyte antigen and CD90+ cells in the tumor xenograft. Arrows point to hepatocyte antigen-positive and CD90+ cells. Magnifications of ×200 and ×400 were used for hematoxylin-eosin staining and IHC, respectively.

Download figure to PowerPoint

CD45CD90+ cells were then sorted from the blood samples (Fig. 3C), and injected into the liver of SCID/Beige mice. Four months after cell injection, tumor formation was detected in 5 out of 10 mice with 5000 CD45CD90+ cell injection. When the cell number was increased to 10,000, 10 of 10 mice developed tumors in the liver at 4 months (Table 4). Further serial transplantation of CD90+ cells from tumor xenografts could generate tumors in a second and subsequently a third batch of immunodeficient mice. By IHC, human hepatocyte antigen and scattered CD90+ cells were detected in the tumor xenograft (Fig. 3D).

Discussion

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

By screening the expression of HSPC markers in the 6 human HCC cell lines (Supplementary Table 2), we found that CD90 expression was positively associated with the tumorigenicity of tumor cells, providing a clue that CD90 might be a marker for CSCs. In addition, the sorted CD90+ cells, but not CD90 cells, from HCC cell lines generated tumor nodules in the nude mice, further confirming that CD90+ cells bore tumor initiating activities.

The above data were then validated in human tumor specimens. Nonlymphatic CD90+ cells were detected in all the tumor specimens but not in the normal, cirrhotic, or nontumorous livers, and these cells could generate tumor nodules in the livers of SCID/Beige mice. In addition, CD90+ cells from tumor xenografts could propagate tumorigenicity to a second and subsequently a third set of SCID/Beige mice, suggesting their stem cell-like properties. When 2 HSPC markers (CD90 and CD44) were used together, all the tumor tissues were found to bear CD45CD90+CD44+ cells. Therefore, we hypothesized that liver CSCs might concomitantly express 2 or more HSPC markers, and these additional stem cell markers might represent different subpopulations of CSCs. Such multimarker hypothesis has been suggested in identifying cancer stem cells in breast and pancreatic cancers.9, 14 Because CD44 is an adhesion molecule that helps tumor cell invasion and migration,31 the CD90+CD44+ phenotype of CSCs might explain the aggressive growth pattern of HCC.

The finding that CD45CD90+ cells in the circulation could generate tumor nodules suggested the presence of circulating CSCs in human liver cancers. However, the reason it took longer for the circulating CD45CD90+ cells to develop tumors than cells from the tumor tissues remains to be explored. One possibility is that circulating CSCs need time to migrate and settle in the favorable organs before differentiating into adult phenotypes. Another reason might be that circulating CSCs could harbor different characteristics from tissue CSCs, because we observed that the majority of tissue CSCs was CD90hi (high level of fluorescent intensity by flow cytometry) cells, whereas the majority of circulating CSCs were CD90lo (low level of fluorescent intensity) cells. Because hepatitis viruses infect every type of cell in the human body, including circulating stem/progenitor cells, there is a possibility that mutations occur when these cells differentiate into adult phenotypes during the process of repair after liver injury.32 However, our findings could not answer whether the circulating CD45CD90+ cells were a source of liver cancers, or they were detached from the primary tumors, though we did detect a significant positive correlation between the number of CD45CD90+ cells in the tumor tissues and CD45CD90+ cells in the blood.

CD133 has been shown to be a marker for CSCs in some types of cancers, including cell line–induced HCC. However, our findings did not support CD133 as a sensitive and specific marker for human liver cancer. Using CD45 to exclude lymphocytes, CD45CD133+ cells were only detectable in approximately 70% of tumor tissues. In addition, only a proportion of CD45CD90+ cells concomitantly expressed CD133, whereas all the CD45CD133+ cells concomitantly expressed CD90, suggesting that CD45CD90+ was more sensitive than CD45CD133+ to represent CSCs in human liver cancers. Moreover, CD45CD133+ cells were also detectable in some tissues and blood samples from normal controls and patients with cirrhosis, indicating that CD45CD133+ was also not specific for liver malignancy.

Kelly et al.33 recently challenged the animal models that have been used to study human malignancy. To confirm whether CD90+ cells did drive liver cancer formation, we sorted CD90+ cells from a mouse chemically transformed hepatocyte cell line to determine their tumorigenicity in immune competent BALB/c mice and confirmed the tumor-initiating properties of these CD90+ cells (Supplementary Table 3). Therefore, we considered that Kelly et al.'s study did not actually separate CSCs from non-CSCs, but only identified 2 subpopulations of CSCs. Because the expression of stem cell markers varies from cell lines and tumor specimens, a convincing conclusion could only be drawn via selection of an appropriate marker to identify true CSCs.

This is the first study to demonstrate the presence of CSCs in human liver cancer, both locally and systemically. Due to small number of cases, statistical analysis could not be performed to evaluate the correlation between the number of CSCs and clinico-pathological features of liver cancer patients. However, universal presence of CD45CD90+ cells in the tumor tissues and CD45CD90+ cells in more than 90% of blood samples suggests that CD45CD90+ could be a highly specific and sensitive circulating marker for the diagnosis of human liver cancers. Moreover, the detectable level of circulating CD45CD90+ cells in patients with small tumors (≤5 cm) provides a hint that this combination might be used as a marker for the diagnosis of early stage tumors. Finally, the finding that CD44 blockade could induce death of CD90+ cells suggests the possible therapeutic strategies by targeting CD45CD90+CD44+ CSCs in the future.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

The immortalized human hepatocyte cell line (MIHA) was a kind gift from J. Roy-Chowdhury of the Albert Einstein College of Medicine, New York, NY. The MHCC97L and MHCC97H cell lines were kindly provided by Zhao-You Tang of the Liver Cancer Institute, Zhongshan Hospital, Fudan University, Shanghai, P.R. China.

References

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Supplementary material for this article can be found on the H EPATOLOGY website ( http://interscience.wiley.com/jpages/0270-9139/suppmat/index.html ).

FilenameFormatSizeDescription
hep22082-Supplementarytables.doc60KSupporting Information file hep22082-Supplementarytables.doc

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.