Recent studies have described a subpopulation of cancer cells within tumors termed 'cancer stem cells' (CSCs), which have stem-like properties such as self-renewal and the ability to differentiate into multiple cancer cell types [
1‐
7]. The CSC theory suggests that such CSCs persist in tumors as a distinct population and cause relapse and metastasis by giving rise to new tumors [
8‐
10]. Although CSCs make up only a small fraction of a tumor, they possess the unique capability to regenerate a tumor whereas most tumor cells lack this regenerative capability [
11,
12]. By means of a non-obese diabetic/severe combined immunodeficiency disease (NOD/SCID) xenotransplant assay in combination with specific cell surface markers (CD44
+CD24
-/low), CSCs were enriched from metastatic and primary breast tumors and were shown to have the ability to reestablish tumor heterogeneity after transplantation [
1]. Since then, additional CSC markers have been proposed and studied to isolate putative tumor stem cell populations. However, as demonstrated by a recent report from Stuelten and colleagues [
13], the complexity of CSC markers continues to pose challenges for identifying and isolating the putative tumor stem cell populations by the cell-sorting approach. In addition to initiating tumors, CSCs are thought to be capable of initiating metastasis. The link between CSCs and metastasis has been suggested by several studies. First, breast CSCs were shown to invade through Matrigel, a basement membrane matrix used routinely as an indicator of metastatic potential of cancer cells [
14]. Second, a recent study demonstrated that there is a link between epithelial-mesenchymal transition (EMT) and breast CSCs [
15]. Furthermore, the prevalence of CD44
+CD24
- cells in breast cancer patients indicates a link between high numbers of stem-like cancer cells and metastasis [
16]. However, only a few studies have directly tested the metastatic capability of putative CSCs
in vivo. Collective evidence from a few studies that directly tested the
in vivo metastasis using sorted CSCs suggests that the CSC phenotype alone may exhibit invasive property
in vitro but is inadequate to determine or predict
in vivo metastasis. For example, in pancreatic cancer, CSCs (CD133
+ cells) were not able to metastasize when injected orthotopically at low numbers [
17]. In mammary carcinomas, CD44
+CD24
low cells were invasive
in vitro but the phenotype was not sufficient for metastasis when cells were injected intracardiacally
in vivo [
14]. Therefore, we set out to investigate alternative mechanisms that could enrich for breast CSCs with tumor-initiating and metastatic capabilities.
To identify factors that distinguish the malignant subpopulation within breast tumors, we began to explore environmental influences known to associate with aggressively metastatic breast tumors. It has been postulated that hypoxia contributes directly to the development of more aggressive cancers by exerting selective pressure on the tumor cell population to favor cells that can survive decreased O
2 and nutrients [
18‐
20]. During tumor development, rapid expansion of cancer cells creates a hypoxic microenvironment that is followed by periods of reoxygenation to promote tumor progression. These two aspects of tumor progression (hypoxia and reoxygenation) cooperate to provide growth advantages essential for the progressive development of aggressive tumors [
21]. Although extensive efforts have been devoted to understanding the effect of hypoxia on tumor progression, two areas of tumor biology remain unclear. What is the effect of fluctuating oxygen tension on tumor progression? How does hypoxia drive an irreversible phenotype without genetic manipulation? It is known that both hypoxia and consecutive hypoxia/reoxygenation can exert a variety of effects on tumor cell biology, including activation of pro-survival signal transduction pathways, aberrant genetic and epigenetic alterations, and increased tumor angiogenesis. In some studies, hypoxia/reoxygenation was shown to drive expression of proteins associated with poor prognosis [
22]. Furthermore, susceptibility of genomic instability can vary after acute or chronic exposure to hypoxia followed by reoxygenation [
23]. Therefore, we hypothesize that hypoxia/reoxygenation cycles may provide the driving force to select for a highly metastatic breast CSC subpopulation.
To study the effect of hypoxia/reoxygenation cycles on breast cancer, we exposed two metastatic human breast cancer cell lines (MDA-MB 231 and BCM2) to cycles of chronic hypoxia and nutrient deprivation. After one cycle of hypoxia and reoxygenation, we observed a small cell population that survived hypoxia as spherical clusters under hypoxic conditions and that resumed proliferation after reoxygenation. We then isolated and exposed this novel subpopulation to additional cycles of hypoxia/reoxygenation and established a distinct subpopulation of cells from these two breast cancer cell lines. As expected, this novel subpopulation showed increasing viability under hypoxia and the ability to proliferate as either an adherent monolayer or substrate-independent tumor spheres. Interestingly, an increased fraction of the cell population was found to express CD44+/CD24-/ESA+ cell surface markers. However, this novel subpopulation was distinguished by being highly tumorigenic and metastatic and showed upregulation of EMT markers. These findings strongly suggest that we have succeeded in isolating a unique metastatic CSC population by exposing breast cancer cells to repetitive cycles of hypoxia and reoxygenation.