Liposarcoma cells with aldefluor and CD133 activity have a cancer stem cell potential

The use of surplus patient material for cancer research is based on general written information and consent from the patients, combined with approval from the Regional Ethics Committee of Southern Norway for each project (Permit S-06133). All procedures involving animals were performed according to protocols approved by the National Animal Research Authority in compliance with the European Convention for the Protection of Vertebrates Used for Scientific Purposes (approval ID 1499, http://​www.​fdu.​no).

Ten formalin-fixed and paraffin-embedded liposarcoma patient samples were obtained from the Department of Pathology at Oslo University Hospital (The Norwegian Radium Hospital). More specifically, the samples included 3 well-differentiated liposarcomas (grade 1-2), 3 de-differentiated liposarcomas (grade 4), 2 myxoid and round cell liposarcomas (grade 3-4) and 2 pleomorphic liposarcomas (grade 4). Four μm thick sections were made and processed for immunohistochemistry using the Dako EnVision™ Flex+ System (K8012, Dako Corporation) and Dakoautostainer. Sections were deparaffinized and epitopes unmasked using PT-Link (Dako) and EnVision™ Flex target retrieval solution, low pH. After blocking endogenous peroxidase with 0.03% hydrogen peroxide (H₂O₂) for 5 minutes, the sections were incubated with monoclonal mouse antibodies ALDH (1:3000, BD Transduction Laboratories™) and CD133/1 (AC133) (1:25, Miltenyi Biotec Inc.) over night at 4°C. Subsequently, the slides were incubated with EnVision™ Flex+ Mouse linker (15 min) and EnVision™ Flex/HRP enzyme (30 min). Tissue was stained for 10 minutes with 3'3-diaminobenzidine tetrahydrochloride (DAB) and then counterstained with haematoxylin, dehydrated and mounted in Diatex. Normal liver and the CaCO2 cell line (American Type Culture Collection No. HTB37 (Rockville, MD)) have been included as positive controls for ALDH and CD133, respectively. Negative controls included replacement of monoclonal antibodies with mouse myeloma protein of the same subclass and concentration as monoclonal antibodies. The immunoreactivity was evaluated according to the number of positively stained tumour cells (0 = none; 1 < 10%; 2 = 10 - 50%; 3 > 50%).

The ATCC liposarcoma cell line SW872 (HTB92) (originally generated from a surgical specimen with histopathology of undifferentiated malignant liposarcoma.) was utilized to establish a xenograft in locally bred athymic NCR nu/nu mice (nude mice). The xenograft was then passaged to a new mouse before the tumour reached maximum 2 cm³. In order to extract cells from the xenografts, typically 6 - 8 tumors were minced in Hank's buffered saline solution (Invitrogen). The tissue-pieces were then incubated in 5 U/ml collagenase 4 (Worthington's) in DMEM:F12 (Gibco) for 45 minutes to 1 hour at 37°C. Cells were collected by passing the mixture through a 70 μm filter. The cells were subsequently maintained in either standard RPMI (Lonza) containing 10% fetal bovine serum (PAA laboratories Gmbh), 1× glutamax (Gibco) and 1 μg/ml penicillin/streptomycin (Lonza) or in stem cell (SC)-medium (70% mouse embryonic fibroblast conditioned medium (R&D systems) mixed with 30% of human embryonic stem cell medium (containing 20% "knock-out" serum replacement (Invitrogen), 1% non essential amino acids (Gibco), 4 ng/ml bFGF (Invitrogen), 0, 1 mM β-mercaptoethanol (Sigma), 1× glutamax (Gibco) in DMEM:F12 (Gibco))). The cells were maintained in culture for 10-14 days before analyses were performed. Adherent cells were dissociated when sub-confluent using TrypLE (Invitrogen).

Spheroid-shaped aggregates were dissociated by 45 minutes incubation in TrypLE (Invitrogen) at 37°C. Adherent cells were detached by a shorter incubation in TrypLE. Aldefluor staining (Stem Cell Technology) was performed at the concentration of 1 × 10⁶ cells/ml Aldefluor assay buffer, according to the protocol recommended by the manufacturer. On all occasions the monoclonal mouse antibody TRA-1-85-APC (1:20, R&D systems), which recognizes an epitope found on all human cells, was included. On some occasions the cells were subsequently labeled with one of the following monoclonal mouse antibodies CD44-PE (1:10), CD90-PE (1:20), CD73-PE (1:10) (All from BD Pharmingen), CD105-PE (1:20, eBioscience), CD133/2(293C)-PE (1:10, Miltenyie Biotec. Inc), STRO-1-PE (1:20, Santa Cruz Biotec) or fibroblast growth factor receptor (FGFR)1 (M19B2) (1:100, Abcam). Cells stained with FGFR1 antibody were subsequently labeled with Alexa Fluor 647 donkey anti-mouse IgG (H+L) (1 μg/million cells, Invitrogen-Molecular Probes). The cells were incubated on ice for 40 minutes. The cells were then washed and filtered through a 40 μm filter, and subsequently analyzed or sorted by flow cytometry. Analyses were performed using a FACS ARIA-2 (Becton Dickenson). Viable singlets which were TRA-1-85⁺ were sorted into the following four fractions: Aldefluor^high CD133^high, Aldefluor^high CD133^low, Aldefluor^low CD133^low and Aldefluor^low CD133^high. The flow cytometry sorted cells were subject to viability analysis by trypan blue staining, before subsequent experiments were performed.

One thousand cells from each flow cytometry sorted subpopulation were plated in 0, 3% soft agar (Difco) in SC-medium in 35 mm non-adhesive dishes. Two hundred and fifty μl SC-medium was added once a week. Uniform spheroids of minimum 50 μm were counted approximately four weeks post plating.

Cells were grown in standard RPMI (Lonza) containing 10% fetal bovine serum (PAA laboratories Gmbh), 1× glutamax (Gibco) and 1 μg/ml penicillin/streptomycin (Lonza), supplemented with an adipocytic differentiation cocktail (50 μM Indomethacin, 1 μM Dexamethason, 0, 5 mM isobutyl-methyl-xanthine (IBMX)). Following 21 days in culture, the cells were fixed in 70% ethanol and subsequently stained in 0, 3% oil red O, and analyzed in a fluorescence microscope (Olympus IX81). Lipid droplets in mature adipocytes appeared red.

Serial dilutions (100 - 25 000 cells) of each sorted subpopulation were injected subcutaneously into the flanks of locally bred athymic NCR nu/nu mice (nude mice). TRA-1-85⁺ (human specific epitope) cells were injected as unselected controls. The cells were diluted in a final volume of 100 μl DMEM:F12 (Gibco). Viability of the injected cells was confirmed by trypan blue (Sigma) staining prior to injection.

Immunohistochemical analyses of ALDH1 expression in liposarcoma patient samples confirmed that 10 out of 10 samples expressed ALDH1. More specifically, 8 out of 10 samples expressed ALDH1 in more than 50% of the tumour cells. One patient sample displayed ALDH1 expression in 10 - 50% of the tumour cells, and for one patient sample, less than 10% of the tumour cells were ALDH1 positive (Figure 1, Table 1). The samples represented a range of liposarcoma sub-types (well-differentiated, de-differentiated, myxoid/round celled and pleomorphic liposarcoma). We were not able to find any correlations between particular liposarcoma subtypes and the level of ALDH1 expression in this small and diverse panel.

Having confirmed that ALDH1 is indeed expressed in human liposarcomas, we wanted to investigate whether liposarcoma ALDH-positive cells could be associated with CSC activity. We preferred to use a xenograft model, because the passing of the xenograft from mouse to mouse ensures that the growth conditions are physiological and that tumour initiating cells are present. Aldefluor analysis of cells extracted from the SW872 liposarcoma xenograft showed that the SW872 xenograft cells, like the liposarcoma patient samples, displayed ALDH activity (11% of the cells were Aldefluor^high: Figure 2B), making xenograft-derived SW872 cells a suitable model for further analyses of ALDH-positive cells.

In order to maintain the extracted cells in a culturing medium best suited for enriching CSCs, we first investigated the effect of different culturing media on the expression of ALDH and other stem cell markers. The extracted xenograft cells were therefore maintained for 10-14 days in either standard RPMI medium containing fetal bovine serum (RPMI) or stem cell medium (SC-medium) containing "knock-out" serum replacement, mouse embryonic fibroblast (MEF) conditioned medium and basic fibroblast growth factor (bFGF), commonly included in embryonic stem cell medium to prevent differentiation [35]. The cellular morphology was highly dependent on the culturing medium. Cells maintained in RPMI exhibited an adherent morphology and cells maintained in SC-medium attached to each other and grew as large aggregated spheroids in 3D suspension (Figure 2A). Cellular growth as spheroids in suspension has previously been associated with stem-ness and tumor initiating activity [36, 37]. Interestingly, when the cells had been maintained in SC-medium, a larger percentage of the cells displayed ALDH activity (average 35% Aldefluor positive cells), compared to the average 11% observed when the cells were maintained in RPMI. The ALDH inhibitor diethylamino-benzaldehyde (DEAB) could block this activity (Figure 2B). Furthermore, when cells were initially incubated in RPMI for 6 days and then transferred into SC-medium for the remaining period, the percentage of cells displaying Aldefluor activity increased (data not shown). These findings indicate that the cells comprise a degree of plasticity, and that cells which have the capacity to become more stem-like may do so in the presence of growth factors in the SC-medium. For instance, FGF signaling is implicated in regulation of self-renewal and differentiation. Since bFGF binds to and activates FGF Receptor 1 (FGFR1) [38], we decided to investigate FGFR1 membrane expression in SW872. Interestingly, we found that FGFR1 was highly expressed in the SW872 cell line. Furthermore, expression of FGFR1 was induced during culturing of xenograft-derived SW872 cells in SC-medium (86, 8%) compared to culturing in RPMI (62, 8%) (Table 2), indicating that activation of FGFR1 may result in expansion of CSCs. According to the CSC theory, the CSC population represent a small sub-population within the tumor [2]. In keeping with this theory, others have shown that a smaller, better enriched CSC population is isolated by flow cytometric cell sorting when combining the Aldefluor assay with antibody staining of CSC surface antigens [8, 13]. Thus, we would expect the large Aldefluor^high cell population observed after culturing the cells in SC-medium to be heterogeneous, and the CSCs to represent a smaller population within the Aldefluor^high cell population.

In the case of liposarcoma, a likely cell of origin for the CSC would be a mesenchymal progenitor or stem cell (MSC). To our knowledge, no surface marker is known to uniquely identify MSCs, so we first tested the cell surface expression of the following markers, which are known to be expressed on MSCs: CD44, CD73, CD105, CD90 and STRO-1 [39, 40]. We also included the stem cell and CSC marker CD133 in our screen [41]. In addition we performed phenotypic analyses of the original SW872 cell line (Table 2). With the aim to identify a small Aldefluor^high surface marker^high (double-positive) cell population, we performed the Aldefluor assay in combination with antibody staining against each surface marker. When testing Aldefluor in combination with CD90, CD44 or CD105 staining, we found that dual expression was observed in a small percentage of the cells following culturing in RPMI. The percentage of double-positive cells increased dramatically to approximately 40% due to an increasing number of cells expressing ALDH when the cells were maintained in SC-medium (Table 2). Next we tested Aldefluor in combination with STRO-1 or CD73 staining, and found that a relatively small percentage of cells were double-positive, independent of medium. Finally, we tested Aldefluor in combination with CD133 and found that no cells were double-positive when the cells were incubated in RPMI. However, interestingly we found that 0, 1% of the cells displayed an Aldefluor^high CD133^high phenotype when maintained in SC-medium. Because CSCs are expected to represent a small fraction of the tumour cells, using CD90, CD44 or CD105 in combination with Aldefluor would not be likely to result in sufficient enrichment of CSCs. On the contrary, CD73, STRO-1 and CD133 might be suitable as CSC-markers, since these markers, when combined with Aldefluor, identified a small population of SW872 xenograft-derived cells. The Aldefluor^high CD133^high phenotype was consistently observed in a small population (0, 1 - 1, 7%, n = 9) of cells cultured in SC-medium. The Aldefluor^high CD133^high subpopulation disappeared when cells were cultured in RPMI, indicating that the combined expression of these two stem cell markers had been induced by factors in the stem cell media. Subsequently, we were interested in evaluating whether cells with an Aldefluor^high CD133^high phenotype comprised a CSC-potential. We therefore decided to perform further characterization of this subpopulation with respect to CSC abilities.

Using flow cytometry, we isolated 4 subpopulations based on ALDH and CD133 expression. In order to investigate the different cell population's stem-like ability to self-renew, we performed spheroid assays in soft agar. The Aldefluor^high CD133^high cell population generated well-defined, round spheroids of approximately 50 μm in size (Figure 3A), at a frequency of up to 1 out of 4 cells. All the other three subpopulations generated spheroids at a significantly lower frequency (Figure 3B).

According to the theory, a CSC has the ability to both generate more CSCs through self-renewal, and to undergo partial differentiation generating heterogeneous cancer cells, which make up the bulk of the tumour. Liposarcomas are in part composed of adipocytes and a potential liposarcoma CSC should therefore have the capacity to differentiate into adipocytes. When culturing the sorted cell populations in the presence of an adipocytic differentiation cocktail, we found that the Aldefluor^high CD133^high cells were able to differentiate into mature adipocytes more efficiently than the other sorted cell populations (Figure 4).

One of the hallmarks of CSCs is the increased ability to form tumors in vivo. Following flow cytometry, serial dilutions (100, 1000, 5000 and 25 000 cells) of the four sorted subpopulations were injected into immunodeficient nude mice. The Aldefluor^high CD133^high cells produced tumors more efficiently in nude mice compared to the other sorted cell populations (Table 3). As few as 100 Aldefluor^high CD133^high cells were sufficient to generate tumors in 14% of the mice, whilst no tumors were formed when the other subpopulations were injected at this cell dilution. When injecting 5000 cells of the Aldefluor^high CD133^high subpopulation, the majority of the injections (66%) resulted in tumour formation. We were unable to obtain sufficient number of cells to inject 25 000 Aldefluor^high CD133^high cells.