CSCs are described as a small population of tumour cells possessing stem-like properties, such as the ability to self-renew, as well as to differentiate into more mature cells that make up the bulk of the tumour, which usually to some extent resembles normal tissue. These cells are also referred to as tumour initiating [1].

The CSCs are in many aspects similar to normal stem cells, and are thought to arise either when normal stem cells gain oncogenic mutations, which confer enhanced proliferation and lack of homeostatic control mechanisms, or alternatively when a progenitor or differentiated cell acquires mutations conferring de-differentiation to a malignant stem-like cell [2]. Since the integrity of stem cells is of critical importance for the organism, several mechanisms that ensure the survival of stem cells have evolved. These mechanisms include enhanced activity of membrane pumps which remove toxic substances [3], and enhanced activity of enzymes such as aldehyde dehydrogenase (ALDH), which confer resistance to toxic agents [4, 5]. ALDH1 was also found to be implicated in regulating the stem cell fate in hematopoietic stem cells (HSCs) [6]. Properties and functions of normal stem cells can also be employed to enrich CSCs. In this respect, the Aldefluor assay, originally optimised to detect ALDH1 expression in HSCs [7] has been used to successfully enrich CSCs from breast cancer [8], leukemia [9], prostate cancer [10], colon cancer [11], bladder cancer [12] and liver cancer [13]. Because the Aldefluor substrate probably is not specific for this isoform [14], we refer only to ALDH-activity. ALDH-activity has also been associated with increased tumourigenicity in osteosarcoma [15]. Furthermore, several groups have reported that expression of ALDH is associated with high grade and poor prognosis in lung cancer [16], leukemia [9], ovarian cancer [17], breast cancer [8, 18], colon cancer [11], prostate cancer [10], bladder cancer [12] and head and neck cancer [19]. ALDH expression has also been correlated with resistance to chemotherapy [19, 20].

The surface molecule CD133, also known as AC133 and prominin-1, is expressed on normal stem cells [21] and on CSCs identified in a range of cancers [22], including cancer of the brain [23, 24], colon [25, 26], pancreas [27] and liver [28]. The majority of research concerning CD133 has been focused on epithelial cancers, but CD133 expressing-cells have also been observed in mesenchymal tumours. Recently, Tirino et al., reported that CD133 is expressed in all of 21 primary bone sarcoma samples analysed (0, 21-7, 85%). Interestingly, the CD133⁺ cells displayed CSC characteristics, such as increased ability to generate tumours in vivo and form spheres in vitro. The CD133⁺ cells were also able to repopulate the culture with CD133^- cells, and were able to undergo differentiation [29]. Others have also reported that a subset of Ewing sarcoma primary tumours [30, 31] and synovial sarcoma primary tumours [32] harbour CD133-expressing cells. In addition, several osteosarcoma cell lines contain subpopulations of cells (typically 3-5%), which are positive for CD133 [33].

Since the markers which are commonly used to isolate CSC populations do not uniquely identify CSCs, CSC enrichment can be improved by combining several markers. For instance, the enrichment of CSC populations from liver cancer cell lines using only CD133 was doubled when CD133 was used in combination with ALDH [13, 28]. Similarly, Ginestier et al demonstrated that breast CSCs could be better enriched by combining Aldefluor with the markers CD44⁺ CD24^- lin^-, originally used by Al-Hajj and co-workers [34].

In this article we confirm that ALDH is expressed in liposarcoma primary material. Using a liposarcoma xenograft model system we show that ALDH is also expressed in this system, and that the combined use of Aldefluor and CD133 enables enrichment of a small cell population by flow cytometry. The Aldefluor^high CD133^high cells have CSC characteristics, such as increased ability to form spheroids in soft agar, and increased tumourigenicity in vivo.

In this study, we initially chose to focus on Aldefluor as a CSC marker for several reasons. Firstly, the Aldefluor assay has been used to successfully isolate CSCs from several malignancies [8–13, 15]. Secondly, we found ALDH1 a clinically relevant marker, identifying subpopulations of cancer cells in all liposarcoma patient samples analyzed. ALDH expression has proven a useful marker for cancers of several tissues [8–12, 16–19, 42]. Thirdly, the Aldefluor assay is less cytotoxic compared to other CSC isolation methods (e.g. side population assay), and since an intact cell membrane is required, only viable cells are isolated. Although the analyses of these phenotypes require separation of individual cells and short term in vitro culturing, we chose to use a xenograft-derived cell model to better mimic the 3D growth conditions and stroma interactions of in vivo human tumors. Furthermore, the continuous passaging of the xenograft ensured the presence of tumour-initiating cells. Moreover, in vitro conditions are not necessarily favorable for maintaining stem-ness, and we therefore compared the effects of two different culturing medium. Morphological observations and Aldefluor analyses of the SW872 xenograft-derived cells maintained in SC or RPMI medium indicated that the SC-medium was the more favourable for maintaining/inducing the CSC phenotype in vitro. The cells displayed an adherent cellular morphology when maintained in RPMI, while the cells grew as detached, round "spheroid"-aggregates when the cells were maintained in SC-medium, a growth-pattern that has been associated with stem-ness [23, 43]. Furthermore, the fact that the percentage of cells which displayed ALDH activity was significantly higher when the cells were maintained in SC-medium also indicated that the SC-medium is favorable for enrichment of CSCs. Moreover, the observed increase in number of cells displaying high Aldefluor activity following a change of medium from RPMI to SC, indicates that a subpopulation of the bulk cells have a potential to become more "stem-like" in response to certain stimuli. It is likely that the 3D cell-cell contacts, as well as the mixture of growth factors in the SC-medium maintain and induce CSC self-renewal. Since a large percentage of the SW872 cells express FGFR1, and the percentage of cells expressing FGFR1 is further increased following culturing in SC-medium (containing bFGF), it is possible that CSCs are enriched through FGFR activation.

A large percentage of the SW872 liposarcoma xenograft-derived cells were Aldefluor positive, making it unlikely that ALDH as a single marker could be used to identify a pure CSC population. Others have shown that the use of Aldefluor in combination with other stem cell markers improves the enrichment of CSCs [8, 13, 42]. A likely cell of origin for the sarcoma-CSC is an MSC-like stem or progenitor cell. However, since no markers are known to uniquely identify MSCs, we investigated a range of markers expressed on MSCs. We also included the stem cell and CSC marker CD133 [22–28, 31]. Although several of the Aldefluor^high surface marker^high subpopulations identified in this screen might enrich for CSCs, the Aldefluor^high CD133^high cells seemed particularly promising. This small subpopulation was only observed in the 3D spheroid culture (SC-medium), indicating that the phenotype was either selectively induced by factors in the SC-medium, or was dependent on the growth pattern.

The functional analysis of the sorted subpopulations of SW872 cells demonstrated that the Aldefluor^high CD133^high cells had a highly increased ability to form spheroids in soft agar, indicating that these cells have an increased ability to self-renew compared to the other sorted cell populations. Interestingly, the Aldefluor^high CD133^high cells had higher capacity to differentiate into adipocytes. Whether the Aldefluor^high CD133^high cells have multi-lineage potential was not tested. However, since the Aldefluor^high CD133^high CSC is likely to originate from a MSC, it would be interesting to investigate the ability of these cells to differentiate into other mesenchymal cell types, such as osteoblasts and chondrocytes. Our in vivo tumourigenicity assay showed that the Aldefluor^high CD133^high subpopulation overall generated tumors more efficiently compared to the other subpopulations when injected subcutaneously into nude mice, in particular at low cell numbers. However, at higher cell numbers tumors were also generated by some of the other subpopulations. Re-analyses of each isolated subpopulation was done by a second round of flow cytometry to determine the purity of the isolated fractions. As demonstrated in Figure 5, the Aldefluor^high CD133^high subpopulation was only enriched to 33% purity, with a large percentage of tumour cells from the other subpopulations "diluting" the CSC population. The Aldefluor^highCD133^low flow sorted subpopulations was clearly "contaminated" with a few Aldefluor^high CD133^high cells, which likely contributed to tumour formation at high cell numbers. The purity of the flow sorting may be compromised by variability in expression and staining, but also by inherent "noise" in the flow sorter. The fact that the Aldefluor^high CD133^high cell population is only enriched also partly explains why tumors are not formed in all Aldefluor^high CD133^high injections. Furthermore, when separating the cells into subpopulations, the CSCs may lack the support of cells that are required to make up a "niche" in vivo.

ALDH1 was expressed in all the liposarcoma patient samples analyzed by IHC. Although the level of expression varied from less than 10% of the tumor cells expressing ALDH1 to more than 50% of the tumor cells expressing ALDH1, we were not able to correlate the differences in level of expression with any particular factors; neither sub-type, tumor location, patient age or tumor grade. Furthermore, we were unable to confirm CD133 expression in the same panel (data not shown). There are several problems associated with CD133 immunohistochemical expression analysis [41]. Several groups have reported that the antibodies binding CD133 detect only the glycosylated epitopes [44]. However, Kemper et al demonstrated that bacterially expressed CD133 or CD133 glycosylation mutants were indeed recognized by the CD133 antibody AC133 used here. Instead the authors concluded that the accessibility of the AC133 epitope varied [45]. Although we cannot confirm CD133 expression in our primary material, CD133 might still be present on the surface, but undetectable by the AC133 antibody due to epitope masking. Alternatively, expression of CD133 may only be present in very few cells or at a frequency below the detection level of immunohistochemistry. This is consistent with Suva et al and Tirino et al who both show that CD133 positive cells are extremely rare in sarcoma patient material [29, 31].

In conclusion, we have demonstrated that ALDH1 is expressed in liposarcoma patient samples, although we were unable to confirm CD133 expression in the same material. We have performed extensive phenotypic analyses of liposarcoma xenograft-derived cells using Aldefluor and surface markers, and as a result identified a CSC-like subpopulation of cells expressing both ALDH and CD133 when cultured as spheroids in SC-medium. Furthermore, we have demonstrated that this phenotype is associated with stem-like abilities, such as increased ability to self-renew and to form tumours in immunodeficient mice. Although it remains to be validated whether Aldefluor and CD133 in combination can be used to isolate CSCs from liposarcomas and sarcomas in general, these markers have proven useful for isolating CSCs across tumor types [13], and may be used as targets for novel CSC-specific therapies. Ongoing work includes specifically targeting and killing the CSC population in our model system.