Identification and characterization of cancer stem cells in ovarian yolk sac tumors

Abstract

Recent evidence supports the cancer stem cell theory, that is, that malignant tumors arise from cells termed cancer stem cells or tumor‐initiating cells that have the ability to self‐renew and are responsible for maintaining the tumor. Cells with marked tumor‐initiating capacity have recently been identified in a number of solid tumors. CD133 (PROM1, human prominin‐1) has been used as a marker to detect stem cells (progenitor cells) and cancer stem cells (tumor‐initiating cells) in various tissues. Ovarian yolk sac tumors (YSTs) are rare and highly malignant. The present study was designed to evaluate the tumor‐forming ability of CD133⁺ cells in ovarian YST cell lines and to examine the characteristics of CD133⁺ cells, such as cell growth and invasiveness. Our data suggest ovarian YST to be maintained by a rare fraction of cancer stem‐like cells that express the cell surface marker CD133. (Cancer Sci 2010)

Malignant ovarian germ cell tumors (MOGCTs) constitute 3–5% of all ovarian malignancies.^{( 1 )} They occur mostly in adolescent and young women. The relative frequency of MOGCTs is higher in Japan than in North America or western Europe.^{( 2 )} Ovarian yolk sac tumors (YSTs) account for approximately 20% of MOGCTs, and are the second most frequent histological subtype, after ovarian dysgerminoma.^{( 3 )} Microscopically, YST present various histological patterns and has multiple features mimicking primitive yolk sac mesenchyma.^{( 4 )} The level of alpha‐fetoprotein (AFP) in serum is a useful marker for the diagnosis and management of YST, because it is elevated in all patients with tumors containing a YST component.^{( 5 )} Cisplatin‐based chemotherapy has dramatically improved the prognosis of YST, but most authors agree that the prognosis is still worse than for other subtypes of MOGCT.^{( 6} , ^{7 )} Several prognostic factors have been reported for YST and two studies addressed that staging and optimal cytoreductive surgery strongly affected the prognosis of YST.^{( 1} , ⁸ , ^{9 )} Therefore, 5‐year survival rates for cases involving disseminated tumors remain poor. Peritoneal dissemination is the main metastatic process of YST besides direct extension of the tumor into adjacent tissues.

Until a few years ago, all neoplastic cells within a tumor were suggested to contain tumorigenic growth capacity, but recent evidence hints to the possibility that this ability is confined to a small subset of cancer‐initiating cells, so‐called cancer stem cells (CSCs).^{( 10} , ^{11 )} The current consensus definition describes a CSC as a cell within a tumor that is able to self‐renew and to produce the heterogeneous lineages of cancer cells that comprise the tumor.^{( 12 )} Biologically distinct and relatively rare populations of CSCs have been detected with markers and several methods in a variety of cancers. Numerous markers have been used to isolate and characterize CSCs.

Although its biological function remains unknown,^{( 13 )} CD133 (PROM1, human prominin‐1) is recognized as a stem cell marker for normal and cancerous tissues. CD133 is a 5‐transmembrane cell surface glycoprotein of 865 amino acids with a total molecular weight of 120 kDa. Indeed, CD133 alone or in combination with other markers is currently used for the identification and isolation of CSCs from many malignant tumors. For example, CD133 has been found on CSCs in cancers such as brain tumors, including medulloblastomas^{( 14} , ¹⁵ , ^{16 )} and glioblastomas,^{( 15} , ^{17 )} leukemia,^{( 18 )} colon cancer,^{( 19} , ^{20 )} pancreatic cancer,^{( 21 )} lung cancer,^{( 22 )} and liver cancer.^{( 23 )} Prostate cancer stem cells have been characterized as CD44⁺/α₂β₁ ^hi/CD133⁺.^{( 24 )}

A lack of human ovarian YST cell lines and the fact that YST is rare has made this tumor more difficult to study than most other solid tumors. However, the establishment of the human ovarian YST cell lines NOY1 and NOY2 in 2008^{( 25 )} and, moreover, the identification and characterization of cancer stem cells of its cells, may be useful for the development of a novel therapy for ovarian YST.

In the current study, we examined CD133 expression in these cell lines and carried out experiments in vitro and in vivo with isolated CD133⁺ and CD133⁻ YST cells to further elucidate the potential role of CD133⁺ cells in ovarian YST.

Materials and Methods

Cell culture.  Two human ovarian YST cell lines (NOY1 and NOY2), as well as SKOV‐3 cells (a human epithelial ovarian carcinoma cell line) and Hep3B cells (a human hepatocellular carcinoma cell line) were used in this study. They were conserved in our laboratory. The NOY1 and NOY2 cell lines were previously established at our institute.^{( 25 )} The cells were maintained in RPMI‐1640 medium (Sigma, St. Louis, MO, USA) supplemented with 10% FCS, penicillin (100 U/mL), and streptomycin (100 μg/mL) at 37°C in a humidified atmosphere containing 5% CO₂.

Human peritoneal mesothelial cells (HPMCs) were isolated from surgical specimens of omentum with the consent of each patient, as described previously.^{( 26 )} Briefly, small pieces of omentum were surgically resected under sterile conditions and trypsinized at 37°C for 30 min. The suspension was then passed through a 200‐μm pore nylon mesh to remove undigested fragments, and centrifuged at 760g for 5 min. The cells collected were cultured in RPMI‐1640 medium supplemented with 10% FCS. In subsequent experiments, cells were used during the second or third passage after primary culture. The HPMCs were identified by immunostaining using mouse monoclonal antibodies against cytokeratin 19 and vimentin (Dako, Glostrup, Denmark). A single isolated NOY1‐CD133⁺ cell was cultured in serum‐free medium supplemented with 10 ng/mL basic fibroblast growth factor (bFGF; R&D Systems, Minneapolis, MN, USA) and 20 ng/mL epidermal growth factor (EGF; R&D Systems) as a spheroid culture.

Immunohistochemical staining.  Immunohistochemical analyses of CD133 were carried out on human ovarian YST tissues and s.c. transplanted tumors in mice. Five ovarian YST tissues were obtained from five patients who underwent surgical treatment at Nagoya University Hospital (Nagoya, Japan) after obtaining informed consent. All samples were fixed in 10% formalin and embedded in paraffin. Sections were cut at a thickness of 4 μm. For heat‐induced epitope retrieval, deparaffinized sections in 0.01 M citrate buffer were treated three times at 90°C at 750 W for 5 min using a microwave oven. Immunohistochemical staining for CD133 was carried out using the avidin–biotin immunoperoxidase technique (Histofine SAB‐PO kit; Nichirei, Tokyo, Japan) according to the manufacturer′s instructions. As the primary antibody for staining CD133, a rabbit polyclonal CD133/1 (AC133) antibody (Abgent, San Diego, CA, USA) was used at a dilution of 1:50. All sections were counterstained with hematoxylin.

Immunofluorescence staining.  NOY1 and NOY2 cells were grown on chamber slides (Nalge Nunc International, Tokyo, Japan). They were fixed for 30 min with 4% paraformaldehyde and washed several times with PBS. Coverslips were incubated in blocking solution containing 1% BSA in PBS for 1 h, and incubated with the CD133/1 (AC133) antibody (Abgent) for 1 h at room temperature. For nuclear staining, samples were incubated with Hoechst 33342 (Dojindo, Kumamoto, Japan) at room temperature. After incubation with the appropriate secondary antibody (CD133, goat anti‐rabbit rhodamine‐conjugated IgG; Santa Cruz Biotechnology, Santa Cruz, CA, USA), fluorescence was visualized with a fluorescence microscope (IX71; Olympus, Tokyo, Japan).

Flow cytometric analysis.  Flow cytometry was carried out to quantify the expression of CD133 on the surface of the cells. Data were acquired using a FACSCalibur (Becton Dickinson, San Jose, CA, USA), and analyzed using CellQuest Pro software (Becton Dickinson). Scatter plots were used to exclude cell aggregates. At least 1.0 × 10⁴ events were acquired for each analysis. Gating was implemented on the basis of negative‐control staining profiles.

Western blot analysis.  The Western blot analysis of CD133, AFP, CD44, CD29 (integrin β1), Nanog, Oct3/4, SOX2, and CXCR4 was carried out as described previously.^{( 27 )} Briefly, 30 μg of total cell lysate was electrophoresed on a 10–15% SDS‐polyacrylamide gel and transferred electrophoretically to Immobilon membranes (Millipore, Bedford, MA, USA). After treatment with blocking solution (5% nonfat dry milk/0.1% Tween‐20/PBS), the membranes were incubated overnight with a recommended dilution of primary antibodies. We used the following antibodies: anti‐CD133/1 antibody (Abgent); anti‐AFP antibody (Dako, Glostrup, Denmark); anti‐CD44 and anti‐CD29 antibody (Pharmingen, San Diego, CA, USA); anti‐Nanog and anti‐Oct3/4 antibodies (R&D Systems); anti‐SOX2 antibody (Santa Cruz Biotechnology); anti‐CXCR4 antibody (Chemicon, Temecula, CA, USA), and anti‐β‐actin antibody (Sigma). The membranes were washed three times with Tween/PBS for 15 min each time, then incubated with the appropriate secondary antibody for 1 h. After washing with Tween/PBS, proteins were visualized using enhanced chemiluminescence reagent (Amersham Pharmacia Biotech, Tokyo, Japan) followed by exposure to X‐ray film.

Magnetic cell sorting.  Magnetic cell sorting (MACS) was carried out using a Mini & MidiMACS Separator and Indirect CD133 MicroBead Kit (Miltenyi Biotec, Auburn, CA, USA) to isolate CD133⁺ and CD133⁻ cells from NOY1 and NOY2. All the procedures were carried out according to the manufacturer’s instructions. Cells were labeled with primary mouse CD133/1 antibody, subsequently magnetically labeled with rat anti‐mouse IgG1 MicroBeads, and separated on a MACS LS or MS column. To further improve the purity of the negative fraction, we used LD columns developed for the gentle depletion of MicroBead‐labeled cells. These columns were used to obtain >95% pure CD133⁻ populations. The purity of sorted cells was evaluated by flow cytometry, immunofluorescence staining, and Western blotting. As NOY1 cells separated more efficiently than NOY2 cells, they were used in subsequent experiments. NOY1‐CD133⁺ and CD133⁻ cells were examined in terms of their ability to form tumors in vivo, and their proliferation, colony formation, migration, and invasive ability in vitro.

In vitro cell proliferation assay.  Cells were plated in hexaplicate at a density of 1500 cells in a 200‐μL volume in 96‐well plates, and cultured for 1–5 days. Cell viability was assayed using a modified tetrazolium salt MTT assay carried out using a CellTiter 96 Aqueous One Solution Cell Proliferation Assay kit (Promega, Madison, WI, USA) according to the manufacturer’s instructions. Absorbance was measured at 492 nm using a microplate reader (Labsystems; Multiskan Bichromatic, Vienna, VA, USA).

In vitro invasion assay.  Cell invasion was evaluated using 24‐well Matrigel invasion chambers (Becton Dickinson Labware, Franklin Lakes, NJ, USA).

Human peritoneal mesothelial cells were suspended in the lower chamber at a concentration of 1.0 × 10⁴/mL in 700 μL RPMI‐1640 medium supplemented with 10% FCS. After the cells had reached confluency, the culture medium in the lower chamber was replaced with fresh medium containing 10% FCS. After 48 h incubation, NOY1‐CD133⁺ and CD133⁻ cells were suspended in the upper chamber at a final density of 5.0 × 10⁴/mL in 200 μL serum‐free RPMI‐1640 medium. After 9 h incubation, the remaining tumor cells on the upper surface of the filters were removed by wiping with cotton swabs, and the invading cells on the lower surface were stained with May–Grünwald Giemsa. The number of cells on the lower surface of the filters was counted under a microscope (magnification, ×100), and we carried out four individual experiments in the invasion assay in triplicate.

In vitro migration assay.  Cell migration was assayed in 24‐well Transwell cell culture chambers (Costar, Corning Inc., Lowell, MA, USA). Cells were suspended in the upper chamber at a final concentration of 5.0 × 10⁴/mL in 200 μL RPMI‐1640 medium. The subsequent procedures were the same as those used for the invasion assay.

Colony forming assay.  Cells were grown in 60‐mm culture dishes.

A single cell suspension was obtained by filtering the supernatant through 100‐μm and 40‐μm cell strainers (BD Biosciences, Franklin Lakes, NJ, USA).

The dissociated cells were plated at a density of 200 cells in 60‐mm culture dishes (Asahi Glass, Chiba, Japan) and subjected to May‐Grünwald Giemsa staining on day 12. The culture medium was changed every 3 days. Colonies with a diameter larger than 500 μm were then counted as one positive colony. The number of colonies was counted per 2.25 cm², and five individual experiments were carried out.

In vivo studies.  Five‐week old female BALB/c nude mice were obtained from Japan SLC (Nagoya, Japan). The treatment protocol followed the guidelines for animal experimentation adopted by Nagoya University.

A total of 1.0 × 10⁶ and 5.0 × 10⁵ isolated NOY1‐CD133⁺ and CD133⁻ cells were inoculated s.c. in 0.2 mL serum‐free RPMI‐1640 medium through a 24‐gauge needle into the lower flank of 6‐week‐old female BALB/c nude mice for a tumorigenesis test. Six‐week‐old BALB/c nude mice were also intra‐abdominally injected with 5.0 × 10⁶, 1.0 × 10⁶, and 1.0 × 10⁵ isolated NOY1‐CD133⁺ or CD133⁻ cells in 0.5 mL serum‐free RPMI‐1640 medium to cause peritoneal metastasis. To examine the metastatic potential of the cells, mice were killed 8 weeks after the injection, and i.p. dissemination was evaluated (n = 9). In addition, s.c. inoculated mice were killed 18 weeks after the injection (n = 3–4).

Densitometric analysis.  The photographs of membranes were scanned with a densitometer (Shimazu CS‐930; Shimazu, Kyoto, Japan).

Statistical analysis.  For results of experiments in vivo and in vitro, statistical comparisons between groups were carried out using the non‐paired Student’s t‐test or anova with Bonferroni corrections. Differences between groups were considered statistically significant at P < 0.05. Data are expressed as the mean ± SD.

Results

CD133 expression in primary ovarian YST and cell lines derived from ovarian YST patients.  To determine whether CD133⁺ cells are present in ovarian YST patients, we first carried out immunohistochemical staining of CD133 in five human ovarian YST tissue samples. The CD133⁺ fraction in the ovarian YST specimens was <3%. In all cases, only a small percentage of CD133‐expressing cells could be identified in ovarian YST samples (Fig. 1A). Subsequently, we detected CD133 expression in ovarian YST cell lines (NOY1, NOY2) using flow cytometry, Western blotting, and immunofluorescence staining. The mean fluorescence intensity in NOY1 and NOY2 cells was 47.3 and 26.2, respectively. The proliferation of NOY1 and NOY2 cells positive for CD133 in the flow cytometric analysis was 45.43 ± 4.3% (mean ± SD) and 40.94 ± 3.3%, respectively (Fig. 1B). Likewise, these cell lines tested positive for CD133 expression on Western blotting and immunofluorescence staining (Fig. 1C,D). Hep3B cells were used for a positive control in the Western blot analysis.^{( 28 )} SKOV‐3 cells were similarly used as a negative control.^{( 29 )} Data were obtained from three individual experiments, respectively.

Figure 1.

 Detection of CD133⁺ cells in primary ovarian yolk sac tumor (YST) cell lines (NOY1 and NOY2) derived from YST patients. (A) Immunohistological staining of CD133 in surgically resected ovarian YST tissue samples using a CD133‐specific polyclonal antibody. Magnification: left, ×40; middle, ×100; right, ×400. (B–D) CD133 expression in ovarian YST cell lines NOY1 and NOY2. (B) Representative results of flow cytometric analysis are shown using CD133/2‐PE stain. The M1 gates were defined as the intrinsic PE fluorescence of unlabelled control cells. M2 gates were used to define CD133⁺ cells. Purple‐filled area, cells stained with CD133/2⁻PE; white‐filled area, unlabelled control cells. NOY1‐CD133⁺ = 45.43%; NOY2‐CD133⁺ = 40.94%. (C) Western blot analyses of CD133 expression in Hep3B hepatocellular carcinoma, SKOV3 ovarian adenocarcinoma, and NOY1 and NOY2 cells. Hep3B, positive control; SKOV3, negative control. (D) Immunofluorescence staining of CD133 in NOY1 and NOY2 cells. CD133, Rhodamine Red; nuclei, Hoechst 33342 stain.

Magnetic bead cell sorting.  NOY1‐CD133⁺ and NOY1‐CD133⁻ cells were sorted with MACS. The purity of the CD133⁺ and CD133⁻ cell populations was 94.67 ± 1.87% and 2.96 ± 1.93%, respectively (Fig. 2A). The mean CD133 fluorescence intensity of NOY1‐CD133⁺ and CD133⁻ cells was 88.5 ± 0.6 and 9.9 ± 2.2, respectively. The purity was consistent with the data obtained by Western blotting and immunofluorescence staining (Fig. 2B,C). The relative levels of CD133 and AFP expression in NOY1‐CD133⁺ cells are shown in Figure 2(B).

Figure 2.

 NOY1‐CD133⁺ and NOY1‐CD133⁻ cells were sorted with magnetic cell sorting. (A) Flow cytometric analysis of CD133 expression in ovarian yolk sac tumor NOY1 cells using CD133/2‐PE stain. Green line, NOY1 wild type; orange‐filled area, NOY1‐CD133⁺; purple‐filled area, NOY1‐CD133⁻. The sorting purity of the freshly sorted NOY1‐CD133⁺ and CD133⁻ cells was 94.67 ± 1.87% and 2.96 ± 1.93%, respectively. (B) Western blot analysis of CD133 and alpha‐fetoprotein (AFP) expression in sorted NOY1‐CD133⁺ and CD133⁻ cells. The purity was also confirmed by Western blotting. In contrast to CD133 expression, AFP expression was slightly lower in NOY1‐CD133⁺ cells than CD133⁻ cells. (C) Immunofluorescence staining of CD133 expression in NOY1‐CD133⁺ and CD133⁻ cells. Similarly, the purity was consistent with data obtained by immunofluorescence staining. Bars represent 200 μm. (D) Western blot analysis of stem cell markers (Nanog, Oct3/4, and Sox2) in NOY1 wild type, CD133⁺, and CD133⁻ cells isolated from NOY1 cells. Levels of stem cell markers were higher in NOY1‐CD133⁺ than in NOY1‐CD133⁻ cells.

Expression of “stemness”‐related molecules in NOY1‐CD133⁺ cells.  To determine whether NOY1‐CD133⁺ cells possess certain stem cell‐like properties, Western blotting was carried out with lysate from the NOY1 wild type, CD133⁺, and CD133⁻ cells (Fig. 2D). The expression of stem cell markers is important for self‐renewal, and differentiation of stem cells such as Nanog, Oct3/4, and SOX2 was increased in NOY1‐CD133⁺ cells. In contrast, the expression of AFP was decreased in NOY1‐CD133⁺ cells (Fig. 2B). The relative levels of CD133, Nanog, Oct3/4, and SOX2 expression in NOY1‐CD133⁺ cells are shown in Figure 2(D).

Previous studies have shown normal and cancer stem cells to organize into a spherical structure. In an attempt to show other characteristic features of cancer stem cells, we cultured NOY1‐CD133⁺ and CD133⁻ cells in serum‐free medium containing EGF and bFGF. In the NOY1‐CD133⁺ cell culture, floating spherical structures were observed. After 1 day of culture without EGF and bFGF in the presence of 10% FCS, floating sphere‐like aggregates attached to the plastic dish, gradually migrating from tumor spheres and differentiating into large and adherent cells.

Colony formation and proliferation assay.  To document the ability of NOY1‐CD133⁺ cells to form colonies, single‐cell suspensions were plated at a density of 200 cells in 60‐mm culture dishes in the limiting dilution experiment. As shown in Figure 3(A), colonies (cellular aggregates larger than 500 μm/2.25 cm²) were formed much more efficiently by NOY1‐CD133⁺ cells, which gave rise to a threefold larger number of colonies than that documented for the CD133⁻ population (P < 0.0001).

Figure 3.

 (A) Colony formation assay after 12 days incubation of 200 separated NOY1 ovarian yolk sac tumor cells. Left, photograph of a dish with colonies; right, number of colonies per dish. Data represent the mean ± SD. W/T, NOY1 wild type. *P < 0.001; **P < 0.05. (B) Cell proliferation assay of NOY1 cells. Data represent the mean ± SD. W/T, NOY1 wild type. Migration (C) and invasion (D) assays of NOY1‐CD133⁺ and CD133⁻ cells were examined. (C) Data represent the mean ± SD. *P = 0.0003. (D) The assay was carried out after 9 h of culture. Data represent the mean ± SD. *P = 0.0417.

Figure 3(B) shows 5‐day proliferation curves of NOY1 wild type, CD133⁺ and CD133⁻ cells under the same conditions. There were no apparent differences in cell proliferation between the two populations using a modified 3‐(4, 5‐dimethythiazol‐2‐yl)‐5‐(3‐carboxymethoxyphen 1)‐2‐(4‐sulfophenyl)‐2H‐tetrazolium salt (MTS) assay with the Cell Titer 96 Aqueous One Solution Celt Proliferation Assay kit (Promega, Madison, WI, USA).

Motility and invasive capabilities of NOY1‐CD133⁺/CD133⁻ cells.  We assessed differences in migration and invasion. As shown in Figure 3(C), the number of migrating cells was significantly reduced among NOY1‐CD133⁻ cells compared with CD133⁺ cells (P < 0.0003). As shown in Figure 3(D), NOY1‐CD133⁺ cells were more invasive than CD133⁻ cells in the Matrigel invasion assay (P < 0.0417).

Expression of CXCR4 and stem/progenitor markers CD44 and CD29 in isolated CD133⁺ and CD133⁻ cells.  To better characterize the ovarian YST CSC population, we further analyzed the expression of several cell surface markers in NOY1‐CD133⁺ and CD133⁻ cells. Markers studied included CXCR4 (chemokine receptor), CD44 (cell adhesion molecule), and CD29 (integrin β1). Because stromal cell‐derived factor 1 is an important mediator in cell migration, we investigated the expression of its specific receptor CXCR4 in NOY1‐CD133⁺ cells. CD44 expression distinguishes a number of CSCs, including those associated with head and neck squamous cell, breast, pancreatic, prostate, and colon cancers. CD29 was used to identify CSCs in prostate cancer. By Western blotting, CXCR4 levels were found to be markedly lower in NOY1‐CD133⁻ cells than NOY1‐CD133⁺ cells (Fig. 4A). NOY1‐CD133⁺ cells expressed slightly higher levels of CD44 compared with their CD133⁻ counterparts (Fig. 4B). CD29 was not preferentially expressed by either of the CD133 subpopulations (Fig. 4B). The relative levels of CXCR‐4, CD44, and CD29 expression in NOY1‐CD133⁺ cells are shown in Figure 4(A,B). In prostate cancer, a CSC population with the CD44⁺/integrinα2β1^hi/CD133⁺ phenotype has been identified as showing extensive proliferation, self‐renewal, differentiation, and invasion.^{( 24 )} CD29 (integrin β1) may not be associated with CSCs in ovarian YST. Cancer stem cells might use CXCR4 to locate a permissive microenvironment and, once there, the interaction of SDF‐1α from the peritoneum with CD44/hyaluronic acid at the cell surface could facilitate adhesion to the ECM at the peritoneum in ovarian YST.

Figure 4.

 Western blot analyses of CXCR4 (A) and the stem/progenitor markers CD44 and CD29 (B) in isolated NOY1 wild type, CD133⁺, and CD133⁻ cells. CXCR4 levels were notably higher in NOY1‐CD133⁺ cells than in NOY1‐CD133⁻ cells (A). CD44 levels were slightly higher in NOY1‐CD133⁺ than in CD133⁻ cells (B). CD29 was not preferentially expressed by either of the CD133 subpopulations (B).

CD133⁺ cells are more tumorigenic than CD133⁻ cells in vivo.  To determine whether NOY1‐CD133⁺ cells are more tumorigenic than their CD133⁻ counterparts in vivo, we carried out tumor development experiments using NOY1‐CD133⁺ and CD133⁻ cells. Purified cells were maintained either subcutaneously or intra‐abdominally in nude mice. Representative examples of tumor nodules formed by the injection of purified CD133⁺ and CD133⁻ cells are shown in Figure 5(A,C). A significant difference in tumor incidence was observed between the CD133⁺ and CD133⁻ populations (Fig. 5D, Table 1). As few as 1 × 10⁵ CD133⁺ cells were sufficient for the consistent development of tumors in nude mice, whereas at least 50 times as many CD133⁻ cells were necessary to consistently generate tumors in the same model (Table 1). The injection of fewer CD133⁻ cells resulted in no tumors. Tumor nodules that formed in the mice were confirmed by pathologic examination. Tumors from nude mice injected s.c. were stained with CD133 antibody (Fig. 5B). Interestingly, only a small percentage of CD133⁺ cells could be identified in xenograft tumor samples, the same as in primary ovarian YST samples.

Figure 5.

 NOY1‐CD133⁺ cells are more tumorigenic than CD133⁻ cells in vivo. (A) Representative results of experiments in vivo. We inoculated BALB/c nude mice s.c. with 5 × 10⁵ NOY1‐CD133⁺ cells and the same number of CD133⁻ cells. The development of tumors was observed 17 weeks after the s.c. injection. (B) Immunohistochemical staining for the expression of CD133 was carried out with s.c. xenograft tumor tissues (B). Only a small percentage of CD133⁺ cells could be identified in xenograft tumor samples, the same as in primary ovarian yolk sac tumor samples. (C) NOY1‐CD133⁺ or CD133⁻ cells (5 × 10⁶) were i.p. transplanted into nude mice. Photographs of mice taken 8 weeks after the i.p. injection. (D) The xenograft tumors weighed significantly less in the mice injected with NOY1‐CD133⁻ cells than CD133⁺ cells at 8 weeks after the i.p. injection(n = 9). Data represent the mean ± SD. *P < 0.05.

Table 1.

 Tumorigenecity of NOY1‐CD133⁺ and CD133⁻ cells in nude mice. There was a significant difference in the incidence of tumor between the two populations

Cell type	Cell numbers injected per mouse
	s.c. 1 × 106	s.c. 5 × 105	i.p. 5 × 106	i.p. 1 × 106	i.p. 1 × 105
NOY1‐CD133+	3/3	4/4	9/9	4/4	2/4
NOY1‐CD133−	0/3	0/4	1/9	0/4	0/4

Tumor incidence: number of tumors detected/number of injections.

Discussion

Cancer stem cells may derive from transforming mutations that occur in untransformed stem cells, progenitor cells, mature cells, and cancer cells. Considerable efforts have been directed toward the identification of CSC markers in malignant ovarian tumors.

Ovarian YSTs are rare and highly malignant tumors of utmost importance, occurring in children and young adults. Experiments using primary tumor specimens are difficult for rare tumors and, until recently, no YST cell line was available. Therefore, research into YSTs has not proceeded as far as research into other solid tumors. We were the first to establish human ovarian YST cell lines (NOY‐1 and NOY2).^{( 25 )} In the present study, we attempted to isolate CD133⁺ cells from these cell lines and identify cancer stem‐like cells. The CD133⁺ fraction in the ovarian YST specimens was small. But the percentage of CD133⁺ cells in ovarian YST cell lines was higher than that in clinical ovarian YST samples. It is possible that more tumor‐initiating cells or CSCs remain than in clinical samples, after the establishment of cell lines from tumors. Isolated NOY1‐CD133⁺ cells showed more tumorigenic potential than CD133⁻ cells in the in vivo tumor formation assay. Furthermore, it was shown that CD133⁺ cells have greater colony‐forming ability, motility, and invasive ability than CD133⁻ cells in vitro. With regard to the tumorigenicity of CSCs in vivo, Zang et al. recently isolated and characterized CSCs in primary ovarian tumors using a spheroid culture method, and reported that the i.p. injection of only 5.0 × 10³ ovarian tumor sphere‐forming cells (CD44⁺ CD117⁺ cells) was enough to propagate the original tumor. By contrast, the results of the present study indicate that a minimum of 1.0 × 10⁵ CD133⁺ cells was necessary for the consistent development of tumor in nude mice. Unfortunately, NOY1 cells have relatively low tumorigenic potential, therefore a large number is needed for both subcutaneous and intra‐abdominal injections into BALB/c nude mice.

Cancer stem cells may not only be associated with the initiation and growth of tumors but also play a crucial role in metastasis. The peritoneal mesothelium is the main metastatic site in cases of YST besides direct extension of the tumor into neighboring tissue. The prognosis in disseminated cases remains poor. Interestingly, the present findings indicated that NOY1‐CD133⁺ cells are more tumorigenic than CD133⁻ cells in the peritoneum of nude mice, suggesting a possible link between CD133 expression and peritoneal dissemination. Moreover in this study, we showed that NOY1‐CD133⁺ cells expressed higher levels of CXCR4 than CD133⁻ cells. The SDF‐1/CXCR4 axis was originally detected in leukocytes, being responsible for the trafficking and homing of hematopoietic progenitors.^{( 30 )} Recent reports suggest CXCR4 to be a key regulator of tumor invasiveness leading to local progression and tumor metastasis.^{( 31 )} This SDF‐1/CXCR4 system also appears to regulate the metastasis of other cancer types, including cancers of the lung, colon, and breast.^{( 32} , ³³ , ^{34 )} Moreover, we previously reported SDF‐1 to have various roles in the progression of epithelial ovarian cancer, including the stimulation of cell proliferation, cell motility, attachment to human peritoneal mesothelial cells, and the in vivo development of peritoneal metastasis through CXCR4.^{( 35 )}

Stem cells of various tissues exist within protective niches that are composed of a number of differentiated cell types.^{( 36} , ^{37 )} These mature cells provide direct cell contact and secreted factors that maintain stem cells primarily in a quiescent state. For example, normal hematopoietic stem cells (HSCs) exist in vascular niches, into which endothelial cells secrete factors that regulate HSC function. Moreover SDF‐1a secretion from niche microenvironments is critical for the maintenance of HSCs.^{( 38} , ³⁹ , ^{40 )} Another report showed that CD44 on AML leukemic stem cells (LSCs) was necessary for survival in niche environments in an immune‐deficient mice xenograft model.^{( 41 )} We showed that CD44 expression was slightly higher in NOY1‐CD133⁺ cells than CD133⁻ cells. Although the detailed mechanism is still under investigation, it may be possible that SDF‐1a from the mesothelium and CD44 in CD133⁺ cells are important for the maintenance of NOY1 stem‐like cells. We would like to examine these underlying regulatory mechanisms in our next study.

Another report showed that Oct3/4 identifies cells with pluripotent potential in human germ cell tumors.^{( 42 )} Oct3/4, also known as otf3 or pou5f1, is a member of the POU family of transcription factors, which is expressed in pluripotent mouse and human embryonic stem and germ cells.^{( 43} , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷ , ^{48 )} Expression of its gene is downregulated during differentiation.^{( 49 )} Pou5f1 is critical for the self‐renewal of embryonic stem cells.^{( 50 )} Interestingly, we showed that stem cell markers such as Oct3/4, Nanog, and Sox2 are expressed at higher levels in NOY1‐CD133⁺ cells than CD133⁻ cells. Therefore, these reports support the possibility that cancer stem‐like cells express the cell surface marker CD133.

A consequent question is whether all CD133⁺ cells are CSCs, or whether CD133⁺ is a feature of a fraction of cells that include stem cells and more differentiated progenitor cells. O’Brien et al. ^{( 19 )} observed that not all the CD133⁺ cells in a tumor had colon cancer–initiating potential. In our pilot studies, more than 1 × 10⁷ unsorted NOY1 cells were needed to generate tumors by i.p. inoculation in BALB/c nude mice (data not shown). The injection of 1 × 10⁵ CD133⁺ cells resulted in tumor growth in two of four mice (Table 1). Thus only a fraction of NOY1‐CD133⁺ cells represent cancer stem cells, although the frequency of CSCs may be approximately 100‐fold higher among CD133⁺ cells than unsorted cells. Cancer stem cells in other cancer types have also been characterized by more than one surface marker, for example, CD44⁺/integrinα2β1^hi/CD133⁺ in prostate cancer. For this reason, the identification of other surface markers (such as CD44, CD117, and CXCR4) or use of a sphere culture method, as a way to better characterize ovarian YST CSCs, is worthwhile. Such work is currently under further investigation in our laboratory.

In conclusion, this study provides a possible link between CD133 expression and ovarian YST CSCs. It is important that agents directed against CSCs be able to distinguish between cancer stem cells and normal stem cells. Further elucidation of the relationship between CD133 and stemness should make it possible to develop new therapeutic approaches for the treatment of CSCs in the future.

Identification of the ovarian YST stem cell would be a critical step in advancing the development of novel therapeutic strategies in the management of ovarian YST.