ALDH1-High Ovarian Cancer Stem-Like Cells Can Be Isolated from Serous and Clear Cell Adenocarcinoma Cells, and ALDH1 High Expression Is Associated with Poor Prognosis

Cancer stem-like cells (CSCs)/cancer-initiating cells (CICs) are defined as a small population of cancer cells that have high tumorigenicity. Furthermore, CSCs/CICs are resistant to several cancer therapies, and CSCs/CICs are therefore thought to be responsible for cancer recurrence after treatment and distant metastasis. In epithelial ovarian cancer (EOC) cases, disease recurrence after chemotherapy is frequently observed, suggesting ovarian CSCs/CICs are involved. There are four major histological subtypes in EOC, and serous adenocarcinoma and clear cell adenocarcinoma are high-grade malignancies. We therefore analyzed ovarian CSCs/CICs from ovarian carcinoma cell lines (serous adenocarcinoma and clear cell adenocarcinoma) and primary ovarian cancer cells in this study. We isolated ovarian CSCs/CICs as an aldehyde dehydrogenase 1 high (ALDH1high) population from 6 EOC cell lines (3 serous adenocarcinomas and 3 clear cell adenocarcinomas) by the ALDEFLUOR assay. ALDH1high cells showed greater sphere-forming ability, higher tumorigenicity and greater invasive capability, indicating that ovarian CSCs/CICs are enriched in ALDH1high cells. ALDH1high cells could also be isolated from 8 of 11 primary ovarian carcinoma samples. Immunohistochemical staining revealed that higher ALDH1 expression levels in ovary cancer cases are related to poorer prognosis in both serous adenocarcinoma cases and clear cell adenocarcinoma cases. Taken together, the results indicate that ALDH1 is a marker for ovarian CSCs/CICs and that the expression level of ALDH1 might be a novel biomarker for prediction of poor prognosis.


Introduction
Ovarian cancer is a malignant disease with high mortality and is the fourth most common cause of cancer-related death in women worldwide. [1,2] Obscure and unclear symptoms make detection of ovarian cancer in the early stage difficult. [3] Generally, ovarian cancer is relatively sensitive to first-line chemotherapy based on platinum/taxane. [4] Clinical complete response (CR) can often be achieved by cytoreductive surgery and chemotherapy in advanced ovarian cancer patients; however, the majority of patients with advanced stage have disease recurrence which is the reason for the high mortality of this disease. [5] Some therapeutic candidates of molecular target drugs for ovarian cancer had been tested, but notable improvements in prognosis were not achieved. [2,6].
Recent progress in cancer research has revealed that cancers are composed of a heterogeneous population of cells and that only a small population of cells called cancer stem-like cells (CSCs)/ cancer-initiating cells (CICs) have high tumor-initiating potential (cancer stem cell hypothesis). CSCs/CICs are defined as a small population of cells that have (1) high tumorigenicity, (2) multiple differentiation ability and (3) self-renewal capability. [7][8][9] Results of recent study have also shown that CSCs/CICs are related to cancer recurrence and resistance to radiation or chemotherapy. [10,11] Therefore, CSCs/CICs are thought to be responsible for cancer recurrence and distant metastasis, and elimination of CSCs/CICs is therefore indispensable for curing cancer.
The function of intracellular ALDH is to catalyze the oxidation of aldehyde, and ALDH therefore plays an important role in cellular homeostasis. Recent studies have revealed that both normal and cancer cells with high levels of ALDH1 activity have the potential to function as stem cells and potentials for selfrenewal capability and stress-resistant properties. [20][21][22].
Also in epithelial ovarian cancer (EOC), some researches have suggested the usefulness of ALDH1 activity to identify CSCs/ CICs. The existence of cells with high ALDH activity (ALDH1-high cells, compared with ALDH1low cells) has also been shown in EOC cell lines and in clinical specimens. [23][24][25] The correlation between ALDH1 activity and prognosis of patients, however, is still controversial in EOC. [26] At the same time, the relevance to ALDH1 expression for each histological subtype of EOC has not been clarified yet.
In this study, we identified and evaluated the stemness of ALDH1 high cells in serous adenocarcinoma and clear cell adenocarcinoma of the ovary, not only from established cell lines but also from primary ovarian cancer cells. We also statistically analyzed the association between ALDH1 expression and clinical outcome for ovarian cancer patients.

Ethics Statement
Mice were maintained and experimented on in accordance with the guidelines of and after approval by the Committee of Sapporo Medical University School of Medicine, Animal Experimentation Center under permit number 08-006. Any animal found unhealthy or sick was promptly euthanized. All studies were approved by Institutional Review Boards (IRB) of Sapporo Medical University Hospital and the IRB of Hakodate Goryokaku Hospital. Written Informed consent was obtained from all patients according to the guidelines of the Declaration of Helsinki.

ALDEFLUOR Assay and Isolation by Fluorescenceactivated Cell Sorting (FACS)
We used an ALDEFLUOR assay kit (Stem Cell Technologies TM , Vancouver, BC, Canada) to determine ALDH1 activity of cells according to the manufacturer's protocol. Cells were suspended in ALDEFLUOR assay buffer containing 1 mmol/l per 1x10 6 cells of the ALDH substrate, boron-dipyrrometheneaminoacetaldehyde (BAAA), and incubated for 50 min at 37uC. Each sample was treated with 50 mmol/l of an ALDH-specific inhibitor, diethylaminobenzaldehyde (DEAB), as a negative control. Stained cells were analyzed by BD FACSAria TM II (BD Biosciences, San Jose, CA, USA). Cells were stained with 1 mg/ml of propidium iodide to evaluate their viability prior to analysis.
For isolation of epithelial cells from solid ovarian cancer specimens, we sorted epithelial cancer cells using anti-CD326 (Ep-CAM) APC antibody (BD Biosciences). We used anti-CD326 microbeads (BD Biosciences) and an Automacs system (Miltenyi Biotec, Bergisch Gladbach, Germany) to enrich epithelial cells from malignant ascites of ovarian cancer cases before the ALDEFLUOR assay.
Cell Cycle Analysis and in vitro Cell Grow Analysis ALDH1 high and ALDH1 low cells were directly fixed with 70% ethanol after sorting. Then they were resuspended in PBS containing 250 mg/ml RNase A (Sigma-Aldrich) for 30 minutes at 37uC and stained with 50 mg/ml propidium iodide for 10 minutes at 4uC in the dark. Stained cells were analyzed with a FACSCalibur (BD Biosciences), and the data were analyzed using the Mod-Fit cell cycle analysis program.
For in vitro cell growth assay, ALDH1 high cells and ALDH1 low cells isolated from AMOC-2 and RMG-1 cells were seeded into a 6-well plate at 5 610 4 cells per well. After incubation for 48 and 96 hours, the cells were removed by trypsin and viable cell numbers were determined using CountessH (Life Technologies).
Sphere-forming Assay/single Cell Sphere-forming Assay ALDH1 high and ALDH1 low cells were isolated by FACS and then cultured in 6-well ultra-low attachment surface dishes (Corning, Tewksbury, MA, USA) at 1000 cells per well. For the single cell sphere-forming assay, both ALDH1 high and ALDH1 low were sorted into 96-well ultra-low attachment dishes (Corning) at a single cell per well. The cells were cultured in stem-cell medium, serum-free DMEM/F12 (Life Technologies) supplemented with N-2 supplement (Life Technologies), 20 mg/ml recombinant human epithelial growth factor (Life Technologies), 10 mg/ml human basic fibroblast growth factor (Sigma-Aldrich), 4 mg/ml heparin (Sigma-Aldrich), 4 mg/ml bovine serum albumin (Life Technologies), 20 g/ml human insulin, zinc solution (Life technologies), and 2.9 mg/ml glucose (Sigma-Aldrich). [29] Morphological change was observed daily under a light microscope for 14 days.

In vivo Tumorigenicity
Sorted ALDH1 high and ALDH1 low cells were resuspended at 1.0610 2 , 1.0610 3 and 1.0610 4 cells in 100 ml PBS and Matrigel (BD Biosciences) mixture (1:1). Then each mixture was injected subcutaneously into the right/left middle back areas of 6-week-old female non-obese diabetic/severe combined immune-deficiency (NOD/SCID) mice (NOD.CB17-Prdkcscid/J, Charles River Laboratory, Yokohama, Japan) under inhalation anesthesia by isoflurane. Tumor initiation and progression were observed weekly until the mice were sacrificed at 7 weeks after injection. External tumor volume was calculated as 0.56 Dmax6 (Dmin) 2 [mm 3 ] (Dmax : long axis, Dmin : short axis of mass).

Immunohistochemical Staining for Ovarian Cancer Tissue
Immunohistochemical staining of ALDH1 and Ki-67 was performed with formalin-fixed, paraffin-embedded sections of 123 epithelial ovarian cancer tissues (62 serous adenocarcinomas, 37 clear cell adenocarcinomas, 18 endometrioid adenocarcinomas and 6 mucinous adenocarcinomas) as described previously. [30] [31] Antigen retrieval was done by boiling sections in 120uC for 5 min in a microwave oven in preheated 0.01 mol/l sodium citrate (pH 6.0). Endogenous peroxidase activity was blocked by 3% hydrogen peroxide in ethanol for 10 min. After blocking with 1% non-fat dry milk in PBS (pH 7.4), the sections were reacted  Table 1. Tumor-initiation incidence of ALDH1 high/low cells derived from AMOC-2 and RMG-1 cells. with mouse anti-ALDH1 monoclonal antibody (1:250, Sigma-Aldrich) or mouse anti-Ki-67 monoclonal antibody (1:100, DAKO, Glostrup, Denmark) for 1 hour followed by incubation with biotinylated anti-mouse IgG (Nichirei) for 30 min. Subsequently, the sections were stained with setreptavidin-biotin complex (Nichirei Biosciences, Tokyo, Japan), followed by incubation with 3,39-diaminobenzidine used as a chromogen and counter-staining with hematoxylin. Cytoplasmic staining was regarded as positive for ALDH1, and nuclei staining were regarded as positive for Ki-67. For evaluation of ALDH1 staining, the cases were divided into two groups (ALDH1 high group and ALDH1 low group) by medians (medians for serous adenocarcinoma the median and clear cell adenocarcinoma being 20% and 15%, respectively).

Matrigel Invasion Assay
The invasive capability of cells was evaluated using matrigel invasion chambers (BD Biosciences). Isolated ALDH1 high and ALDH1 low cells (1.0610 4 ) were plated in each upper chamber in serum-free DMEM. The outer chambers were filled with DMEM including 10% FBS as a chemoattractant. Cells were incubated for 48 hours, and invasive cells were stained with Hematoxylin, mounted on slides, and counted at 400-fold upper field by light microscopy.

Statistical Analysis
Statistical analysis, data fitting and graphics were performed using SPSS software package ver.19 (SPSS, Chicago, IL, USA). Data are shown as mean 6 SD of at least 3 independent experiments, and Student's t-test was used to assess statistically significant differences (p,0.05). Overall survival (OS), defined as interval from the date of first diagnosis to the date of death from disease progression, and progression-free survival (PFS), defined as the interval from the date of first diagnosis to the date of disease progression, were estimated using the Kaplan-Meier method and compared using the log-rank test. Associations of ALDH1 expression with clinical stage, lymph node metastases and dissemination were analyzed by Fisher's test.

Isolation of ALDH1 high Cells from EOC Cell Lines
Several methods to isolate CSCs/CICs have already been described [12], and an aldehyde dehydrogenase 1 high population (ALDH1 high ) identified by the ALDEFLUOR assay was described to be enriched with CSCs/CICs. [20] We therefore analyzed ovarian carcinoma cell lines by the ALDEFLUOR assay to isolate ovarian CSCs/CICs. We investigated 3 human ovarian serous adenocarcinoma cell lines (AMOC-2, HUOA and OVCAR-3) and 3 human clear cell adenocarcinoma cell lines (ES-2, RMG-1 and TOV-21G) (Figure 1). ALDH1 high population was identified in all ovarian carcinoma line cells, and the ratio of ALDH1 high cells ranged from 0.7% for ES-2 cells to 7.9% for TOV-21G cells. We could isolate ALDH1 high cells stably from 4 cell lines (AMOC-2, ES-2, RMG-1 and TOV-21G), and we therefore used these cell lines for further analysis.

Characterization of ALDH1 high Cells
Since CSCs/CICs are known to form spheres in floating culture conditions [29], we analyzed ALDH1 high cells by a sphere-forming assay. A total of 10 3 cells per well were sorted and incubated into a 6-well plate in an anchorage-independent environment. At day 10, ALDH1 high cells derived from AMOC-2 cells showed greater sphere-forming ability than that of ALDH1 low cells. Similar results were obtained from ES-2 cells, RMG-1 cells and TOV-21G cells (Figure 2A). Since the sphere-forming assay is not suitable for quantification of the ratios of CSCs/CICs in ALDH1 high cells, we performed a single cell sphere-forming assay. Sphere formation was observed in 4.86% of the wells of ALDH1 high cells derived from AMOC-2 cells and in 3.13% of the wells of ALDH1 high cells derived from ES-2 cells ( Figure 2B). On the other hand, wells of ALDH1 low cells derived from both AMOC-2 and ES-2 cells did not show any sphere formation.
We then investigated the invasion ability of ALDH1 high and ALDH1 low cells by the matrigel invasion assay. ALDH1 high cells derived from all four line cells showed greater matrigel invading ability than did ALDH1 low cells ( Figure 2C).
The cell cycle status of ALDH1 high cells and that of ALDH1 low cells were analyzed. The ratios of ALDH1 high cells in S phase and G2/M phase were greater than ratio of ALDH1 low cells in S phase and G2/M phase ( Figure 3A). Cell grow analysis revealed that ALDH1 high cells derived from AMOC-2 and RMG-1 cells had higher grow rates than those of ALDH1 low cells ( Figure 3B).

Higher Tumorigenicity of ALDH1 high Cells than that of ALDH1 low Cells
To address the in vivo tumorigenicity of ALDH1 high and ALDH1 low cells from EOC cell lines, xenograft transplantation into NOD/SCID mice by ALDH1 high and ALDH1 low cells derived from AMOC-2 and RMG-1 cells was performed ( Figure 4A, 4B and 4C). As shown in Table 1, tumors were observed in 5 of 5 mice (10 4 cells injection), 4 of 5 mice (10 3 cells injection) and 2 of 5 mice (10 2 cells injection) in which xenotransplantation of ALDH1 high cells derived from AMOC-2 cells was performed. On the other hand, tumors were observed in only 3 of 5 mice (10 4 cells injection) and 2 of 5 mice (10 3 cells injection) in which xenotransplantation of ALDH1 low cells was performed. Furthermore, growth rate of tumors derived from AMOC-2 ALDH1 high cells was significantly faster than that of tumors derived from AMOC-2 ALDH1 low cells ( Figure 4C). Similar results were obtained for RMG-1 cells (Figure 4B and 4C). Histological aspects of tumors derived from ALDH1 high cells and ALDH1 low cells that were isolated from AMOC-2 and RMG-1 cells were investigated ( Figure 4A and 4B). Tumors derived from AMOC-2 ALDH1 high cells and AMOC-2 ALDH1 low cells showed poorly differentiated adenocarcinoma and there was no significant difference. Similarly, tumors derived from RMG-1 ALDH1 high cells and RMG-1 ALDH1 low cells showed poorly differentiated clear cell adenocarcinoma. Immunohistochemical staining revealed that there was no significant difference in ALDH1-positive rate between tumors derived from AMOC-2 ALDH1 high cells and tumors derived from AMOC-2 ALDH1 low cells, whereas the tumor derived from RMG-1 ALDH1 high cells showed significant higher ALDH1 positive rates than that of tumor derived from RMG-1 ALDH1 low cells ( Figure 4D). Furthermore, tumors derived from AMOC-2 ALDH1 high cells and RMG-1 ALDH1 high cells showed significantly higher positive rates for Ki-67 (MIB-1) than those of tumors derived from AMOC-2 ALDH1 low cells and RMG-1 ALDH1 low cells ( Figure 4D).

Identification of ALDH1 high Cells in Primary EOC Samples
To detect ALDH1 high cells in primary ovarian cancer, we analyzed eleven clinical materials using the ALDEFLUOR assay (5 cases of solid ovarian cancer cells and 6 cases of malignant ascites of ovarian cancer cases, summarized in Table 2). CD326positive epithelial cells were identified in solid ovarian cancer tissues, and the positive rates ranged from 8.1% to 81.0% ( Figure 5A, upper panels). ALDH1 high cells were detected in 4 of the 5 cases, and positive ALDH1 high cell rates ranged from 0.9% to 12.0% ( Figure 5A, lower panels). Furthermore, ALDH1 high cells were detected in CD326-positive cells derived from ovarian cancer ascites in 4 of the 6 cases, and the positive rates of ALDH1 high cells ranged from 1.7% to 4.2% ( Figure 5B). Therefore, ALDH1 activity determined by using the ALDEFLUOR assay might be a useful approach for isolation of CSCs/CICs from primary ovarian cancer cells.

Association of High Expression Level of ALDH1 is with Poor Prognosis
A total of 123 epithelial ovarian cancer tissues were immunohistochemically stained with anti-ALDH1 antibody (Table 3, Figure 6A). The medians of ALDH1-positive rates in serous adenocarcinoma cases and clear cell adenocarcinoma cases were 20% and 15%, respectively. We therefore divided the cases into  two groups, ALDH1 high group and ALDH1 low group, according to the medians. As shown in Table 3, there was no significant correlation of the expression level of ALDH1 with age or with each FIGO clinical stage. High expression level of ALDH1 showed no significant correlation with advanced stages (stage III+IV), T factor or lymph node metastasis in both serous adenocarcinoma cases and clear cell adenocarcinoma cases. The log-rank test revealed that higher expression level of ALDH1 is associated with poorer prognosis with a significant difference in OS of patients with serous adenocarcinoma (P = 0.006) and OS of patients with clear cell adenocarcinoma (P = 0.047) than those of lower expression level of ALDH1 ( Figure 6B). Higher expression level of ALDH1 showed a tendency for shorter PFS than that of lower expression level of ALDH1, but the differences were not significant (serous adenocarcinoma: P = 0.062; clear cell adenocarcinoma: P = 0.058).

Discussion
In this study, we isolated ovary CSCs/CICs with high tumorigenicity by the ALDEFLURO assay. An ALDH1 high population could be isolated not only from ovarian cancer cell lines but also from primary ovarian cancer samples. Several methods for isolation of CSCs/CICs have been reported. Indeed, ovarian CSCs/CICs have been successfully isolated by using various methods including the ALDEFLUOR assay [25], side population (SP) analysis [32,33], and use of CD133 + [34], CD44 + CD24 - [35] and CD24 + cells. [36] However, there is a controversial report showing that a CSC/CIC marker does not work in some types of cells. [37,38] Therefore, it is essential to validate the cell population isolated by CSC/CIC markers by several types of analysis. In this study, we confirmed that the ALDH1 high population had higher tumorigenicity and greater sphere-forming ability. These observations indicate that the ALDH1 high cells we used in this study are enriched with CSCs/ CICs. There have been few reports on successful isolation of CSCs/CICs from primly ovarian cancer cells. We isolated ALDH1 high cells from 8 of 11 primary ovarian cancer cases. We could not analyze ALDH1 high cells from primary ovarian cancers because of the limitation of cell numbers; however, our approach is a possible and promising method for isolating CSCs/CICs from primary ovarian cancers.
Major histological subtypes of EOC are serous adenocarcinoma, clear cell adenocarcinoma, endometrioid adenocarcinoma and mucinous adenocarcinoma, and these subtypes are known to have different characteristics in risk factor of carcinogenesis, molecular biological aspects and so on. More information about individual subtypes is needed to improve the survival of patients. Serous adenocarcinoma is the histology subtype with worst prognosis; however, serous adenocarcinoma cases are often diagnosed at advanced stage. Clear cell adenocarcinoma is gradually increasing up to nearly 12% of EOC in Asian countries, and comparison of the cases in the same stage of 3 histological subtypes (clear cell adenocarcinoma, endometrioid adenocarcinoma and mucinous adenocarcinoma) revealed that clear cell adenocarcinoma is the poorest prognosis in every stages. Therefore, clear cell adenocarcinoma is thought to be the highest grade EOC, recently. [39,40] Serous adenocarcinoma cases were mainly analyzed in previous studies, and there have been few studies in which clear cell adenocarcinoma cases were analyzed. [23][24][25] In this study, we analyzed not only serous adenocarcinoma cases but also clear cell adenocarcinoma cases. Therefore, these information will bring insight into not only for serous adenocarcinomas, but also most highest grade clear cell adenocarcinomas.
We found that 4 EOC cell lines including 3 clear cell adenocarcinoma lines were positive for sphere formation with 1000 cells/well. On the other hand, only two cell lines (AMOC-2 and ES-2) showed sphere formation in single cell sphere analysis. These results indicate that CSCs/CICs, which have the ability to form a sphere from only one cell, are enriched in ALDH1 high cells; however the ratios of CSCs/CICs are not so high even in ALDH1 high populations. This result indicated that the ALDE-FLUOR assay is just a surrogate marker for CSCs/CICs and that the combination of ALDEFLUOR assay with other markers might be a better approach to isolate CSCs/CICs.
In previous studies on immunohistochemical staining, contradictory results regarding the association of ALDH1 expression with prognosis in EOC were obtained. Chang et al. reported that ALDH1 expression correlates with favorable prognosis in serous or non-serous ovarian carcinoma [26], while Deng et al. and Wang et al. reported opposite results showing that high ALDH1 activity is linked to poor prognosis in serous adenocarcinoma. [23,25] It should be noted that conditions for immunohistochemical staining and positivity for ALDH1 differed in those studies. In the present study, we set the cut-off lines at 20% for serous adenocarcinoma and 15% for clear cell adenocarcinoma, and we investigated 2 individual histological types. We obtained results showing that high expression level of ALDH1 was associated with poor prognosis in serous and clear cell adenocarcinomas of the ovary. This is the first report showing a relationship between ALDH1 staining and poor prognosis of clear cell adenocarcinomas. On the other hand, ALDH1 expression had no relevance to clinical stage, lymph node metastases or dissemination. These results indicate that ALDH1-positive cells in epithelial ovarian cancer might be responsible for resistance to anti-cancer therapies rather than promotion of diseases. Therefore, ALDH1 might be a novel biomarker for the prediction of prognosis of not only serous adenocarcinoma cases but also clear cell adenocarcinoma cases. Further investigations on molecular aspects of ALDH1 high cells in ovarian cancers are needed.
In summary, we successfully isolated an ALDH1 high cell population with high tumorigenicity not only from serous adenocarcinoma but also from clear cell adenocarcinoma. ALDH1 high cells have higher tumorigenicity and greater sphereforming ability. ALDH1-positive immunohistochemical staining is related to poor prognosis in both serous adenocarcinoma and clear cell adenocarcinoma. These findings indicate that ovarian CSCs/ CICs are positive for ALDH1, and further analysis of ALDH1positive ovarian CSCs/CICs will lead to the establishment of novel strategies for treating ovarian CSCs/CICs.