Hypoxia Enhances Tumor Stemness by Increasing the Invasive and Tumorigenic Side Population Fraction

Transcription

1 CANCER STEM CELLS Hypoxia Enhances Tumor Stemness by Increasing the Invasive and Tumorigenic Side Population Fraction BIKUL DAS, a,b,c RIKA TSUCHIDA, b DAVID MALKIN, b,c,d GIDEON KOREN, c,e SYLVAIN BARUCHEL, b,c HERMAN YEGER a,f,g a Developmental Biology and Stem Cell Program, b Division of Hematology and Oncology, Department of Pediatrics, and f Division of Pathology, Department of Paediatric Laboratory Medicine, The Hospital for Sick Children, Toronto, Ontario, Canada; c Institute of Medical Sciences and Departments of d Medical Biophysics, e Clinical Pharmacology, and g Laboratory Medicine and Pathobiology, University of Toronto, Toronto, Ontario, Canada Key Words. Side population cells Hypoxia Oct-4 expression levels SDF-1 Stemness Tumor stem cell Disclosure of potential conflicts of interest is found at the end of this article. ABSTRACT Although advances have been made in understanding the role of hypoxia in the stem cell niche, almost nothing is known about a potentially similar role of hypoxia in maintaining the tumor stem cell (TSC) niche. Here we show that a highly tumorigenic fraction of side population (SP) cells is localized in the hypoxic zones of solid tumors in vivo. We first identified a highly migratory, invasive, and tumorigenic fraction of posthypoxic side population cells (SPm [hox] fraction) in a diverse group of solid tumor cell lines, including neuroblastoma, rhabdomyosarcoma, and small-cell lung carcinoma. To identify the SPm (hox) fraction, we used an injured conditioned medium derived from bone marrow stromal cells treated with hypoxia and oxidative stress. We found that a highly tumorigenic SP fraction migrates to the injured conditioned medium in a Boyden chamber. We show that as few as 100 SPm (hox) cells form rapidly growing tumors in vivo. In vitro exposure to hypoxia increases the SPm (hox) fraction significantly. Quantitative realtime polymerase chain reaction and immunofluorescence studies showed that SPm (hox) cells expressed Oct-4, a stemness gene having a potential role in TSC maintenance. In nude mice xenografts, SPm (hox) cells were localized to the hypoxic zones, as demonstrated after quantum dot labeling. These results suggest that a highly tumorigenic SP fraction migrates to the area of hypoxia; this migration is similar to the migration of normal bone marrow SP fraction to the area of injury/hypoxia. Furthermore, the hypoxic microenvironment may serve as a niche for the highly tumorigenic fraction of SP cells. STEM CELLS 2008;26: INTRODUCTION Hypoxic microenvironments play an important role in the trafficking of normal stem cells, including bone marrow (BM)-derived side population (SP) cells. The BM-derived SP cells are localized in their hypoxic niche to maintain their quiescent and undifferentiated state [1 3]. During stress, quiescent hematopoietic stem cells (HSCs) are mobilized/expanded in the hypoxic zone [2, 3] and then migrate to the area of injury for repair and regeneration [4]. The trafficking of bone marrow-derived endothelial progenitor cells from their niche in the bone marrow to an area of injury may rely on a hypoxia gradient [5, 6]. Exposure of stromal cells to hypoxia may increase the tissue level of stromal derived factor-1 alpha (SDF-1 ), leading to the migration of BM-derived SP cells, including endothelial progenitor cells [6]. Tumor hypoxia is characterized by zones of chronic and intermittent hypoxia [7]. The latter, an in vivo process of hypoxia-reoxygenation, may mimic injury/stress and therefore may serve as a niche for the highly tumorigenic fraction of SP cells and also the tumor stem cell (TSC) fraction. Recently, SP cells obtained from glioma, breast, prostate, neuroblastoma, hepatocellular carcinoma, ovarian carcinoma, and gastrointestinal tumor cell lines [8 16] have shown TSClike properties, including a high degree of tumorigenicity. Our preliminary study showed that the SP fraction is enriched in TSC-like cells and increases following exposure to hypoxia/ reoxygenation (Das B, Baruchel S, Yeger H. Hypoxia increases the population of stem cell-like side population cells in neuroblastoma. Third Annual Meeting of the International Society for Stem Cells Research. San Francisco, June 23, 2005). Considering that tumor SP fractions have many similarities to their normal counterpart, including high self-renewal capacity, expression of the stemness gene Oct-4 and high tumorigenic (repopulation) capacity [8], the tumor SP fraction may also migrate to the hypoxic/ischemic regions of a tumor (Fig. Author contributions: B.D.: conception and experimental design, collection, analysis and interpretation of data, collaborative efforts, manuscript writing and revisions, journal correspondence; R.T.: collection and assembly of data, critical review of the manuscript; D.M.: financial support, provision of study material, critical review of the manuscript; G.K.: administrative support; S.B.: administrative support, critical review of the manuscript; H.Y.: financial and administrative support, data analysis and interpretation, critical revision of the manuscript. Correspondence: Herman Yeger, Ph.D., Division of Pathology, Department of Pediatric Laboratory Medicine, Hospital for Sick Children, 555 University Avenue, Toronto, Ontario M5G 1X8, Canada. Telephone: ; Fax: ; hermie@sickkids.ca; or Bikul Das, M.B.B.S., Ph.D., Developmental Biology and Stem Cell program, Research Institute, Hospital for Sick Children, 555 University Avenue, Toronto, Ontario M5G 1X8, Canada. Telephone: ; Fax: ; bikul.das@utoronto.ca Received August 30, 2007; accepted for publication April 24, 2008; first published online in STEM CELLS EXPRESS May 8, AlphaMed Press /2008/$30.00/0 doi: /stemcells STEM CELLS 2008;26:

2 Das, Tsuchida, Malkin et al Figure 1. Preparation of BMSC-derived injured conditioned medium. (A): Hypothesis: A highly tumorigenic SP fraction migrates to the hypoxic/necrotic zone. Chemokines, including SDF-1, secreted by stromal cells in the hypoxic zone attract the SP fraction. (B): The SDF-1 level increased significantly in the injured conditioned medium (BMSC treated with hypoxia and hydrogen peroxide) compared with hypoxia treatment alone (n 3). (C): The SDF-1 level increased significantly in the H-146 xenograft-derived stromal cells treated with hypoxia and hydrogen peroxide compared with untreated tumor stromal cells (n 2). Baseline SDF-1 level was significantly higher in tumor-derived stromal cells compared with murine BMSC (n 2). SDF-1 was measured by enzyme-linked immunosorbent assay., p.002;, p.05. Abbreviations: BMSC, bone marrow-derived stromal cells; hrs, hours; SP, side population. 1A). Here we exploited this idea as a possible way to enrich for a highly tumorigenic fraction of SP cells of several diverse tumors, including neuroblastoma, rhabdomyosarcoma, and small-cell lung carcinoma. We first found that a highly migratory side population (SPm) fraction can be collected in a Boyden chamber, where cells migrated toward conditioned medium of bone marrow stromal cells exposed to hypoxia and hydrogen peroxide injured conditioned medium. We then showed that exposure of the SP fraction to hypoxia and reoxygenation increases the size of the tumorigenic SPm fraction and also increases cellular expression of Oct-4. These post-hypoxia SPm (SPm [hox] ) cells showed significantly increased tumorigenic potential in a xenograft mouse model. Furthermore, we found that the highly tumorigenic SPm fraction is localized in vivo to the hypoxic zone of the tumor xenografts. Thus, a unique selection strategy has been developed that allows isolation of a small subpopulation of cells from tumors that have the characteristics of TSC and its tumorigenic potential. MATERIALS AND METHODS Cell Culture and Induction of Hypoxia SK-N-BE(2) (neuroblastoma) and RH-4 (rhabdomyosarcoma) cell lines were maintained in supplemented -minimal essential medium ( -MEM) and Dulbecco s modified Eagle s medium (DMEM) (Wisent Inc., Saint-Jean-Baptiste de Rouville, QC, Canada, [17]. The H-146 (small-cell lung carcinoma) cell line was obtained from American Type Culture Collection (ATCC) (Manassas, VA, and maintained in RPMI-1640 (Wisent) medium as per ATCC instructions. The human embryonic stem cell line human embryonic stem BG10V (ATCC; SCRC-2002) was maintained on mitomycin-c-inactivated mouse embryonic fibroblasts (ATCC; SCRC ) as described [18]. Radiobiological hypoxic conditions ( 0.1% O 2 ) were established in a sealed chamber using the BBL GasPak Plus anaerobic system envelopes with a palladium catalyst (Becton, Dickinson and Company, Cockeysville, MD, as previously described [19]. Fluorescence-Activated Cell Sorting Analysis and Isolation of SP Cells Standard protocols [9, 20] were used to analyze and isolate SP cells using the Hoechst dye (Molecular Probes, Eugene, OR, exclusion method with slight modifications [21]. Details are given in the supplemental online Materials and Methods. Real-Time Quantitative Reverse Transcription-Polymerase Chain Reaction Real-time quantitative reverse transcription-polymerase chain reaction (qpcr) was performed using TaqMan Gene Expression Assays (Applied Biosystems, Foster City, CA, appliedbiosystems.com) at 40 cycles with 100 ng of starting cdna. RNA was quantified with the Ct method as described [21] using

3 1820 Hypoxia Enhances Tumor Stemness SDS software, version (Applied Biosystems). Details are given in the supplemental online Materials and Methods. Bone Marrow Stromal Cell-Derived Injured Conditioned Medium Primary human bone marrow stromal cells (obtained from donors for BM transplantation after obtaining informed consent and proper ethical practice as approved by the Hospital for Sick Children s Research Ethics Board) were maintained in stromal cell medium ( -MEM, 10% horse serum, 10% fetal bovine serum [FBS], 50 M 2-mercaptoethanol, and 1 M hydrocortisone). To obtain the injured conditioned medium, cells grown in T25 tissue culture flasks were washed with phosphate-buffered saline twice and then treated with 0.5 mm H 2 O 2 in 5 ml of serum-free media and exposed to 20 hours of extreme hypoxia followed by 4 hours of normoxia. Subsequently, conditioned media were collected and stored at 20 C. Isolation and Culture of H-146 Xenograft-Derived Tumor Stromal Cells We used a differential sedimentation technique to isolate relatively pure tumor stromal cells from established xenografts as described [22]. Briefly, H-146 xenografts of Balb/c nude/nude mice ( ml) were dissociated with collagenase type I (1 mg/ml; Boehringer Mannheim) at 37 C in DMEM with 10% FBS with agitation for 1 2 hours and then 5 minutes without shaking; the stromal cell-rich supernatant was collected by centrifugation; and cells were resuspended and cultured in DMEM with 10% FBS in T-25 tissue culture flasks. After 2 days of culture, the medium, along with suspended H-146 cells, was removed and replaced with fresh medium containing 2% FBS. One group was exposed to hypoxia and hydrogen peroxide as described in the text. SDF-1 concentration in the supernatant was measured using a commercially available enzyme-linked immunosorbent assay (ELISA) kit (R&D Systems Inc., Minneapolis, Isolation and Culture of Murine Bone Marrow-Derived Stromal Cells Bone marrow cells obtained from normal, healthy 8 10-week-old BALB/c nude/nude mice were cultured in DMEM with 10% FBS. Following 5 days of culture, suspended cells were removed, and cells were trypsinized and seeded at per T25 flask in DMEM containing 2% FBS for 24 hours. SDF-1 concentration in the supernatant was measured using a commercially available ELISA kit (R&D Systems). In Vivo Tumorigenic Assay Tumorigenicity of the SP and non-sp cells was measured by injecting viable cells subcutaneously into female nude mice (BALB/c, nude/nude). Details are given in the supplemental online Materials and Methods. Matrigel Invasion Assay A Boyden chamber invasion assay was performed as previously described [23], with the following modifications. Briefly, 8- m pore size polyvinyl membrane-based chambers (Corning Life Sciences, Lowell, MA, were coated with 100 l of ice-cold Matrigel (7.5 mg/ml; BD Biosciences, San Diego, and incubated at 37 C for 4 hours. Appropriate numbers of cells (following trypsin neutralization) were added to the upper chamber, and the lower chamber was filled with appropriate media as desired. The chamber was incubated at 37 C for 8 24 hours, and invading cells were counted as described after crystal violet staining [23]. When required, the invading and noninvading cells were isolated after brief trypsinization and expanded using the appropriate culture medium. In Vivo Detection of Tumor Hypoxia Tumor hypoxia was assessed by pimonidazole staining as described [24]. Briefly, pimonidazole hydrochloride (Chemicon, Temecula, CA, Millipore, Billerica, MA, was injected i.v. (60 mg/kg) to tumor-bearing mice (xenografts of 2 ml were used for this experiment). Two hours after the injection, mice were sacrificed, and tumor tissue was fixed in 10% paraformaldehyde. Sections of hypoxic tumor cells were stained with a monoclonal pimonidazole antibody (Chemicon; Millipore) or quantified by fluorescence-activated cell sorting (FACS) as described [24]. For the detection of hypoxic cells in dissociated Matrigel, plugs were fixed with 70% ethanol and stained with pimonidazole antibody, and positive cells were enumerated under an epifluorescence microscope. In Vivo Cell Tracking with Quantum Dots We used quantum dot nanocrystals (Q-Tracker 655 Cell Labeling kit; Q25021MP; Invitrogen, Carlsbad, CA, com) to label and track the SPm (hox) cells in vivo in the hypoxic zones of tumor xenograft. These quantum dots (QDs) are peptide-bonded; they enter the cytoplasm of live cells and locate in the periplasmic vesicles [25 27]. Cells were labeled with Q-Tracker 655 (5 nm) for 60 minutes as per the manufacturer s instructions, and labeled SK-N-BE(2) SPm (hox) cells suspended in 100 l of medium were injected in the intracardiac site of the SK-N- BE(2) xenograft (1 1.5 cm 3 )-bearing nude/nude mice (n 4) under isoflurane anesthesia as described [28]. After specific time points (6 hours and 5 days) of injection, mice were sacrificed, and tumor tissues were cryosectioned and stained with anti-pimonidazole antibody to stain the hypoxic zone. The labeled cells were visualized using an epifluorescence microscopy. Cell counting was performed according to Gasparini s criteria [29], where four areas with the high concentration of quantum dot-stained cells were selected under a low-power field ( 100). Then stained cells were counted under 200 magnification using an epifluorescence microscope, and the mean count of four areas was taken. Immunohistochemical, Immunofluorescence, and Confocal Analysis Standard procedures were performed as described [19]. Details are given in the supplemental online Materials and Methods. Statistical Analysis The data are presented as mean SD. The statistical calculations were performed with GraphPad Prism 4.0 (GraphPad Software, Inc., San Diego, using Student s t test for cell survival, SP cell proportion, and QD assays. ELISA and qpcr data were analyzed using one-way analysis of variance with Newman-Keuls post hoc test. The analysis for tumorigenic potential was done by Porter and Berry s maximum likelihood analysis [30]. Statistical significance was set at p.05%. RESULTS Preparation of BM Stromal Cell-Derived Injured Conditioned Medium to Isolate an SPm Fraction We hypothesized that, similar to normal stem cells, the highly tumorigenic fraction of SP cells may also migrate to the area of injury (i.e., hypoxic/necrotic core of the tumor; Fig. 1A). Tumor hypoxia is characterized by zones of chronic and intermittent hypoxia [7], the latter may simulate oxidative-reperfusion injury. Stromal cells and fibroblasts residing in this microenvironment of hypoxia and oxidative stress may liberate potent chemokines, including SDF-1, that may attract the TSC fraction (Fig. 1A). When human BM stromal cells were exposed to hydrogen peroxide and hypoxia, the SDF-1 level increased significantly compared with hypoxia treatment alone (p.0023, Fig. 1B). Similar results were obtained from tumorderived stromal cells (Fig. 1C), suggesting that tumor stromal cells, when exposed to hypoxia and oxidative stress, increase the secretion of the SDF-1 chemokine. We decided to use bone

4 Das, Tsuchida, Malkin et al Figure 2. SPm cells are more tumorigenic than SPn cells. (A): Characteristic fluorescence-activated cell sorting profile of Hoechst dye-stained SK-N-BE(2) cells, presented as a distinct tail on the histogram. Addition of verapamil (50 M) completely inhibited Hoechst efflux from SP cells. To the right, a schematic diagram of a Boyden chamber assay used to isolate SPm cells. (B): SK-N-BE(2) SPm cells adhered to the Boyden chamber membranes after 24 hours of migration. (C): Injured conditioned medium attracted more SK-N-BE(2) SPm cells compared with conditioned medium ( vs ; n 4) or 20 ng/ml recombinant SDF-1 ( vs ; n 4). (D): High tumorigenicity of SPm cells compared with SPn cells. SPm tumors were obtained by injecting SK-N-BE(2) and H-146 SPm cells into nude mice, s.c., and harvested 4 weeks after injection. The RH-4 SPm xenograft shown was obtained 10 weeks following injection of cells. A large number of SPn and non-sp cells ( ) yielded no tumors even after 14 weeks of follow-up., p.0007;, p Abbreviations: FBS, fetal bovine serum; SP, side population; SPm, migratory side population; SPn, nonmigratory side population. marrow-derived stromal cells instead of xenograft-derived stromal cells because of difficulty in isolating pure and relatively sufficient numbers of tumor-derived stromal cells. SPm Cells Are Enriched in TSC Fraction Tissue hypoxia and SDF-1 play important roles in the recruitment of stem cells to the site of injury [6]. Therefore, SPm cells that migrate to a BM stromal cell-derived injured conditioned medium (Fig. 2A) may be enriched in a tumorigenic fraction of SP cells. To investigate this possibility, we used three cell lines SK-N-BE(2) (neuroblastoma), RH-4 (rhabdomyosarcoma), and H-146 (small-cell lung carcinoma)-derived SP cells. These cell lines contain 1% 1.5% SP cells; express breast cancer resistance protein 1, Oct-4, and Nanog [21] (supplemental online Fig. 1); and can be maintained in serum-free media supplemented with

5 1822 Hypoxia Enhances Tumor Stemness Table 1. Tumorigenic SK-N-BE(2) neuroblastoma cells are highly enriched in the SPm (hox) fraction Cell fraction Tumors/injections SK-N-BE(2) 10/10 3/4 3/8 1/6 0/8 SP 4/4 4/4 2/8 Non-SP 0/10 0/6 SPm 2/2 4/8 1/6 0/14 SPm (hox) 8/8 6/6 9/13 7/10 3/10 SPn (hox) 0/10 SPn 0/6 Non-SPm 0/4 Mouse passage 3 a SPm 4/4 SPn 0/6 SPm and SPm (hox) fractions were isolated and injected s.c. as described in the text. The proportion of tumor stem cells (TSC) is calculated by maximum likelihood analysis method [30]. TSC frequency for unsorted cells: 1 per (95% confidence of to ); SPm cells: 1 per (95% confidence of to ); SPm (hox) cells: 1 per 1,160 (95% confidence of 1 per 794 2,180). Considering that this analysis method ignores cell-cell interactions and microenvironmental factors, the calculations may underestimate the proportion of tumorigenic cells [32]. The mice were observed for 4 5 months, or until the tumor reached 2 ml. a SPm (hox) xenograft-derived tumor cells were sorted for SP fraction to isolate SPm cells, which was again injected ( ) to obtain passage 2. Passage 3 cells were used for the tumorigenic study shown here. Abbreviations: SP, side-population; SPm, migratory side-population; SPm (hox), post-hypoxia migratory side-population; SPn, nonmigratory side-population; SPn (hox), post-hypoxia nonmigratory side-population. growth factors as described [21] (supplemental online Materials and Methods). To examine the tumorigenic potential of SPm cells, sorted SP cells were allowed to invade through a Boyden chamber, and the invaded SPm cells were collected by trypsinization and injected subcutaneously into nude mice (Fig. 2A). SK-N-BE(2) SP cells showed a sixfold increase in migration to injured conditioned medium compared with untreated conditioned medium (p.0007, Fig. 2B, 2C). RH-4- and H-146-derived SP fractions showed a similarly higher migration to injured conditioned medium compared with untreated medium (data not shown). When the SPm fraction was injected subcutaneously into nude mice, cells formed tumors in 8 10 weeks, whereas even a large number ( )of nonmigratory side population (SPn) cells did not form tumors (Fig. 2D). Maximum likelihood estimation of TSC frequency revealed a 51-fold enrichment of TSCs in SPm compared with unsorted SK-N-BE(2) cells (Table 1). Similar results were obtained from H-146 and RH-4 cells (supplemental online Table 1). These results suggest that injured conditioned medium may be used to enrich the TSC fraction within the SP cells. We also injected non-sp fractions, which did not form tumors even following the injection of a large number of cells (Table 1). Considering that Hoechst dye is retained in the non-sp fraction, cellular viability may be compromised [31]. However, we found that the viability and proliferation of non-sp cells were similar to those of SP cells (supplemental online Fig. 2). Previous reports suggest that only a small number of TSCs ( 100 cells) are required to form tumors [8, 32, 33]. Here we found that at least SPm cells were required to form tumors (Table 1). Hence, we decided to further enrich for the TSC fraction within the SPm cells as follows. Enrichment of TSC Fraction Within the SPm Cells by Hypoxia/Reoxygenation Treatment: Isolation of SPm (hox) Fraction Hypoxic exposure increases the self-renewal of human ES cells [34] and neural crest stem cells [35]. Similarly, TSCs residing in the hypoxic microenvironment of tumors may self-renew and expand. Earlier, we developed an artificial in vitro system of hypoxia and reoxygenation to study hypoxia-induced drug resistance, where tumor cells were exposed to 24 hours of extreme hypoxia followed by reoxygenation to mimic intermittent hypoxia [19]. Here we further modified the in vitro assay, where SP cells were exposed to 24 hours of hypoxia followed by 1 4 days of reoxygenation. We speculate that hypoxia and reoxygenation may expand the TSCs within the SPm fraction so that a better enrichment of TSC population can be achieved. Exposure to 24 hours of hypoxia and 4 days of reoxygenation increased the percentage of SK-N-BE(2) SP cells by more than 10-fold (p.0042, Fig. 3A). Although it can be argued that SP cells may be selectively increased since non-sp cells may die in the hypoxic zone, SK-N-BE(2) cells are highly resistant to severe hypoxia (0.1% oxygen) [19]. When FACS-sorted SP and non-sp cells were exposed to hypoxia/reoxygenation, the survival of both SP and non-sp cells remained equivalent immediately after hypoxia (Fig. 3B). Following 4 days of reoxygenation the number of SP cells increased by 29% (p.0053), whereas the number of non-sp cells decreased by 9% (p.07; Fig. 3B). Similar increases in SP cells were observed in H-146 and RH-4 cells, whereas the numbers of non-sp cells remained same (data not shown). Most importantly, the post-hypoxia SPm (SPm [hox] ) fraction increased by 1.5-fold (p.0067), whereas the RH-4 and H-146 SPm (hox) fractions increased by four- and threefold, respectively (Fig. 3C). Although it can be argued that SPm (hox) cells may be selectively increased since SPn cells may die in the hypoxic zone, immediately after hypoxia the number of SPm (hox) cells did not increase (Fig. 3C). Instead, it took 4 days of culture under normoxia to increase the SPm (hox) fraction, suggesting that the increase in SPm (hox) may be related to hypoxia/reoxygenation-induced expansion of the SPm (hox) cells. To investigate whether the increase of SPm (hox) cells may also be related to hypoxia-induced expression of the SDF-1 chemokine receptor CXCR4, we examined the expression of this receptor by qpcr analysis. We found that expression of CXCR-4 is threefold higher in SPm (hox) compared with SPm cells (p.0007), and even 2 ng/ml SDF-1 (the amount present in the injured conditioned medium) was able to attract the SPm (hox) fraction. However, immediately after hypoxia, the expression of CXCR4 in SPm (hox) cells and SDF-1 -mediated migration of SP cells did not increase (supplemental online Fig. 3), suggesting that a hypoxia-induced induction of CXCR4 was not involved in the increase of the SPm (hox) fraction. The percentage of BCRP1- expressing cells remained similar between normoxia and post-

6 Das, Tsuchida, Malkin et al Figure 3. Hypoxia/reoxygenation treatment enhances SPm fraction: isolation of SPm (hox) cells. (A): Temporal increase of SP cell fraction following exposure to hypoxia followed by 4 days of normoxia ( vs ; n 8). SKNBE cells grown in 10% serum were exposed to hypoxia for 24 hours followed by normoxia, and SP cells were analyzed by fluorescence-activated cell sorting (FACS). (B): Significant increase in the number of SP cells following exposure to hypoxia/reoxygenation compared with normoxia (n 4). FACSsorted SP and non-sp cells were exposed to hypoxia/reoxygenation. Total viable cell count was done by trypan blue. (C): Significant increase of invasive SP cells following exposure to hypoxia/reoxygenation compared with untreated SP cells ( vs ; n 5). To the right is a histogram showing a similar increase of SPm (hox) cells in RH-4 cell line following exposure to hypoxia/reoxygenation (H-146: vs ; RH-4: vs ; n 4). SP cells were allowed to migrate for 24 hours in a Boyden chamber, and injured conditioned medium was used as a chemokine. (D): Table showing in vivo growth potential of SPm (hox) fractions in liver following tail-vein injection. To the right, photo shows a liver tumor derived from SPm (hox) (10 3 cells injected) cells that took 4 weeks to cause abdominal swelling, whereas a tumor of comparable size derived from SPm cells (10 4 cells injected) took 10 weeks to cause abdominal swelling., p.05;, p.007; #, note 50% take for 10 3 SPm (hox) cells injected compared with 0% take for unsorted cells. Abbreviations: hrs, hours; SKNBE, SK-N-BE(2); SP, side population; SPm, migratory side population; SPm (hox), post-hypoxia migratory side population; SPn, nonmigratory side population. hypoxia SP fractions (data not shown), suggesting that BCRP1 induction itself may not be a major factor in SPm (hox) increase. SPm (hox) Fraction Is Highly Enriched in TSCs The SPm (hox) cells showed increased self-renewal capacity compared with SP cells in a methylcellulose-based assay (data not shown), and only 100 SK-N-BE(2) SPm (hox) cells were required to form xenografts. Furthermore, the proportion of tumorigenic cells increased from 1 per in SPm fraction to 1 per 1,160 cells in SPm (hox) fraction, a 130-fold enrichment of tumorigenic cells (Table 1). In an orthotopic model of rhabdomyosarcoma, 100 RH-4 SPm (hox) cells gave rise to five tumors out of 10 injections (supplemental online Table 1). Maximum likelihood estimation revealed an 2,000-fold enrichment of tumorigenic cells in SK-N-BE(2) SPm (hox) fraction compared with unsorted cells (Table 1). For the RH-4 and H-146 SPm (hox) cells, the enrichment was 4,000-fold and 3,800-fold, respectively (supplemental online Table 1). Furthermore, SPm (hox) cell-derived tumors demonstrated a significantly shorter latency period compared with tumors derived from similar number of SPm cells (supplemental online Fig. 4). Brabletz et al. suggested a potential correlation between proportion of migratory TSCs and aggressive metastatic growth [36]. We found that when injected intravenously into the mouse tail vein, only SPm (hox) cells formed palpable liver growth within 4 5 weeks (50% take; n 6), whereas a similar number of SPm cells failed to form liver growths (n 6). Injection of at least SPm cells and a period of at least 10 weeks were required to form palpable liver growths (50% take; n 2). (Fig. 3D). In Vivo Localization of SPm (hox) Cells in the Tumor Hypoxic Zone The possibility of expansion/selection of TSC fraction in hypoxia and its migration to an injured conditioned medium suggest that within an in vivo xenograft, TSCs may be enriched in the hypoxic zone. To investigate this possibility, hypoxic cells residing within the tumor xenografts were labeled with pimonidazole, a nontoxic dye that specifically labels hypoxic cells in vivo [24], as shown in Figure 4A. The pimonidazole-labeled tumor tissues were excised and dissociated, and SP and non-sp cells were sorted by FACS (Fig. 4B). A portion of the FACS-sorted SP and non-sp fractions were fixed and stained with anti-pimonidazole antibody and quantified by FACS. The SP fraction showed 25% pimonidazole-positive cells compared with 7% in the non-sp fraction (p.0278; Fig. 4C). Furthermore, of the total number of pimonidazole-positive SP cells, the SPm fraction contained 84% (27.5% of a total of 32.5% pimonidazole-positive SP cells; p.0077; Fig. 4D). In the RH-4 and H-146 xenografts, pimonidazole-positive cells were enriched 3- and 12-fold, respectively, in the SPm fraction (p.05; supplemental online Fig. 5).

7 1824 Hypoxia Enhances Tumor Stemness Figure 4. In vivo localization of SPm cells in the tumor hypoxic zone. (A): SK-N-BE(2) xenograft showing CD34 ve blood vessels and the corresponding area of pimonidazole ve hypoxic zone (magnifications: upper panels, 4; lower panels, 40). Arrows indicate blood vessels and corresponding area of hypoxic zone, where a gradient of pimonidazole-labeled cells is present. (B): Fluorescence-activated cell sorting (FACS) profile showing that SK-N-BE(2) xenograft contains a very high SP fraction (37.5% viable SP cells). (C): Sorted SP and non-sp cells were fixed, and pimonidazole ve cells were quantified by FACS. To the right, the histogram shows significantly higher number of pimonidazole ve cells in the SP fraction compared with the non-sp fraction ( vs ; n 3). (D): Image shows immunofluorescence staining of pimonidazole ve SPm cells obtained from xenografts. To the right, the histogram shows that the majority of the pimonidazole-stained cells were contained in the SPm fraction compared with the SPn fraction ( vs ; n 3). For the migration assay, cells were used., p.05;, Abbreviations: ve, positive; SP, side population; SPm, migratory side population; SPn, nonmigratory side population. We found that the in vivo SP fraction is very high in the SK-N- BE(2) xenografts (37.5%; Fig. 4B). The in vivo SP fractions in RH-4 and H-146 were also more than 20% (data not shown). Since we sorted SP cells from the tumor xenografts, there is a possibility of contamination of the tumor SP fraction with mouse SP cells. However, a detailed investigation that includes immunofluorescence-based expression of tumor-specific markers revealed a negligible degree of contamination of SP cells with mouse cells (supplemental online Fig. 6A, 6B). Furthermore, using a TaqMan mouse glyceraldehyde-3-phosphate dehydrogenase primer, which can detect even a minute amount of mouse cdna ( ng), we did not find evidence of SP cell contamination with mouse cells (supplemental online Fig. 6C). TSCs give rise to phenotypically diverse tumorigenic and nontumorigenic populations in vivo and at the same time maintain their tumorigenic capacity [8, 32, 33, 37]. Considering that a majority of SPm cells are localized in the hypoxic zones in vivo, these cells may retain the tumorigenic ability of SPm (hox) cells. We investigated this possibility and found that in vivo-derived SPm cells consistently showed tumor formation, which was similar to in vitro-derived SPm (hox) cells (100% take following cells

8 Das, Tsuchida, Malkin et al Figure 5. In vivo migration of QD-labeled SK-N-BE(2) post-hypoxia migratory side population (SPm [hox] ) cells into the tumor hypoxic zone. (A): Representative epifluorescence images of QD-labeled cells costained with nuclear 4,6-diamidino-2-phenylindole on days 1 and 6. To the right, histogram shows percentage of QD labeling cells. (B): Image shows localization of QD-labeled SPm (hox) cells in the pimonidazole-stained zones of SK-N-BE(2) xenograft 5 days following intracardiac injection of cells To the right, the histogram shows majority of QD-labeled cells in the pimonidazole ve zone following 6 hours ( vs ; n 3) and 120 hours ( vs ; n 4) after injection. (C): Images show Ki67 staining (green, Alexa Fluor 488) of QD-labeled SPm (hox) cells (yellow arrow) in the pimonidazole zone (red arrow, blue cytoplasmic stain, Alexa Fluor 350). To the right, image shows a non-pimonidazole-stained zone, where many tumor cells are Ki67 ve (yellow arrow). Note the absence of blue cytoplasmic stain. To the extreme right, the histogram shows Ki67 ve cell counting done 5 days after the injection of QD-labeled cells ( vs ; n 4). Cell countings were done according to Gasparini s criteria (supplemental online Materials and Methods)., p.05;, p.0006; #, p.018. Abbreviations: ve, positive; -ve, negative; hrs, hours; QD, quantum dot. injected; Table 1). On the other hand, in vivo-derived SPn did not form tumors (Table 1). Thus, the highly tumorigenic SPm (hox) cell fraction can give rise to phenotypically diverse tumorigenic and nontumorigenic populations, with no evidence of decrease tumorigenicity of the in vivo resident SPm fraction. In Vivo Migration and Localization of Quantum Dot-Labeled SPm (hox) Cells in the Tumor Hypoxic Zone Considering that SPm and SPm (hox) fractions were isolated using the functional activity of the migration of stem cells to the area of injury, we investigated the in vivo migration and localization within tumor tissues when injected systemically to tumor-bearing mice. QD labeling of tumor cells is found to be a reliable technology for tracking migratory tumor cells in vivo because of its nontoxic nature, durability of the fluorescence intensity, convenient labeling technique, and easy localization of its high-intensity fluorescence signal in the tissue microenvironment [25 27]. SK-N-BE(2) SPm (hox) cells were loaded with QDs (Q-Tracker 655; Invitrogen). QD aggregates were seen in 90% 95% of cells following loading with 5 nm QDs, and the aggregates were retained for 6 days (Fig. 5A) without any detectable changes in growth and migration (supplemental online Fig. 7). When injected into the intracardiac site in the

9 1826 Hypoxia Enhances Tumor Stemness SK-N-BE(2) xenograft-bearing nude mice (n 4; details in Materials and Methods), the majority of QD-labeled SPm (hox) cells were observed in the pimonidazole-stained zones within 6 hours after injection (p.041), and their number increased by 2.5-fold in the hypoxic zone 5 days after injection (p.018), whereas the number of QD-labeled cells in the normoxic zone (pimonidazole-negative) did not increase significantly (Fig. 5B). When the QD-labeled cells were costained with Ki67 antibody, the 5-day-postinjection group showed 3 of 10 cells proliferating in the hypoxic zone compared with only 0.5 cell per 10 cells proliferating in the normoxic (pimonidazole-negative) zone (p.0006; Fig. 5C), suggesting that the observed enrichment of QD-labeled SPm (hox) cells in the hypoxic zone was due to the active migration as well as expansion of these cells in the hypoxic zone. Similar results of migration and proliferation within the hypoxic zone were observed with the injection of QD-labeled RH-4 SPm (hox) cells ( ) into the orthotopic xenograft-bearing mice (supplemental online Fig. 8). Injection of a similar number of parental RH-4 cells did not show QDlabeled cells in the tumor (data not shown). Molecular Marker of SPm (hox) Fraction: High Oct-4 Expression The above results suggest that highly migratory, TSC-like SPm (hox) cells may localize to the hypoxic zone in vivo, and some of these cells may actively proliferate whereas others remain quiescent. To further investigate the localization and expansion of SPm (hox) cells in the hypoxic zone in vivo, we first searched for a potential surface marker for SPm (hox) cells. However, qpcr analysis of the established markers, such as CD133, CD44, CD49x, CD29, and CD24, did not exclusively differentiate the SPm (hox) fraction from the rest of the SP population (data not shown). Recently, using a TaqMan qpcr primer for human Oct-4 (Hs _g1; Applied Biosystems), which specifically detects only parental Oct-4 (NM_ ), we showed that SK-N-BE(2) and RH-4 SP cells express Oct-4 [21], an embryonic stemness gene [38] found to be expressed in TSCs [8] (supplemental online Fig. 1D). We performed qpcr analysis using the same primer and found a very high level of Oct-4 in SPm (hox) cells compared with other SP fractions (Fig. 6A). Immunofluorescence study showed a dramatically increased expression of Oct-4 in the SPm (hox) fraction, whereas non-sp fraction did not express Oct-4 (Fig. 6B), thus confirming the qpcr results (Fig. 6A). Most importantly, we found a significant increase in Oct-4-expressing cells in the SPm (hox) fraction compared with the SPm fraction following 4 days of culture in SK-N-BE(2) as well as RH-4 cell lines (p.5; Fig. 6B), whereas immediately after hypoxia, the number of Oct-4-expressing cells did not increase (data not shown). Expansion of Oct-4-Expressing SPm (hox) During Hypoxia/Reoxygenation Stress The apparent increase of Oct-4-expressing SPm (hox) cells following hypoxia/reoxygenation may be the outcome of (a) conversion of Oct-4-nonexpressing to Oct-4-expressing cells by the process of dedifferentiation, or (b) expansion of a pre-existing SPm (hox) population with high Oct-4 expression. We did not find evidence for the dedifferentiation of non- Oct-4 cells to Oct-4 cells since Oct-4 expression did not increase in the post-hypoxia nonmigratory side population (SPn [hox] ) fraction (Fig. 6A). SPn (hox) fraction did not show any marked increase of Oct-4 cells (Fig. 6B), which suggests that the expansion of pre-existing Oct-4-expressing cells is more likely than the dedifferentiation of non-oct-4 to Oct-4 cells. Furthermore, SPn (hox) cells remained nontumorigenic even after the injection of a large number of cells (Table 1; supplemental online Table 1). The most likely possibility is the expansion of a pre-existing, quiescent population of Oct- 4-positive SPm (hox) cells following hypoxia/reoxygenationinduced stress. This possibility is supported by our finding of a dramatic increase of Oct-4 expressing cells in the SPm (hox) fraction (Fig. 6B). We estimated the overall increase of Oct-4-positive population in vitro as follows: the Oct-4- positive SK-N-BE(2) SP fraction was originally 0.15% (of 10,000 SP cells, 67 were SPm cells, of which 22% were Oct-4-positive; Figs. 3C, 6B); it then increased to 0.5% following hypoxia/reoxygenation (of 10,000 post-hypoxia SP cells, 93 were SPm (hox) cells, of which 52% were Oct-4- positive; Fig. 3C, 6B). Considering that the SP fraction itself increased by 10-fold following exposure to hypoxia/reoxygenation (Fig. 3A), therefore, the actual increase of Oct-4 positive cells was 30-fold. We then compared the in vitro versus in vivo expansion of Oct-4-expressing SP cells. To do so, we first quantified the number of Oct-4-expressing cells in the tumor hypoxic zone of SP cell-derived xenografts ( cells injected; 1.5 ml size). Xenografts labeled with pimonidazole were dissociated into single cells and then subjected to Hoechst staining in vitro to isolate SP cells. The sorted SP cells were double-stained with pimonidazole and Oct-4. Flow cytometry analysis showed that most of the Oct-4-positive SP cells were also stained for pimonidazole (the Q2 region of Fig. 6Cii). The total fraction of Oct-4-positive SP cells was 6.3% in the xenograft (Fig. 6C, Q1 Q2), compared with 0.12% in vitro (data not shown). This increase of Oct-4-expressing SP cells in vivo (6.3% vs. 0.12%, a 52-fold increase) is higher than in vitro expansion following hypoxia and 4 days of reoxygenation (52-fold vs. 32-fold). This result suggests that the in vivo tumor microenvironment-mediated hypoxia/reoxygenation is much more effective than our laboratory-based, artificially created environment of hypoxia/reoxygenation for the expansion of TSClike cells. This finding of the expansion of a pre-existing, quiescent population of Oct-4-positive SPm (hox) cells following hypoxia/reoxygenation-induced stress is proposed in a schematic diagram as a process of stemness switch (Fig. 6D). DISCUSSION Solid tumor hypoxia is a well-known factor in tumor aggressiveness [19, 39]. Here we show that hypoxia increases a fraction of SP cells having high tumorigenicity and TSC-like characteristics, including Oct-4 expression. This highly tumorigenic fraction is localized in the hypoxic zones of tumors. Our primary objective was to study the migration and localization of the tumor SP cells to their hypoxic and necrotic zones. For such studies, the use of cell line-derived SP cells as opposed to primary tumor-derived SP cells has several advantages, including easy availability. A disadvantage is that established cell line-derived SP cells may not entirely reflect all the SP characteristics of a primary tumor, including the TSC-like character of SP cells [40]. In this regard, it can be mentioned here that the idea and early evidence of TSCs came from analyzing established cell lines [41]. Recent studies on tumor cell line-derived SP fractions can provide further insight into the regulation of primary TSC growth and stemness [14]. We isolated SPm (hox) cells having TSC-like characteristics, including Oct-4 expression and high self-renewal activity. Furthermore, we found that SPm (hox) cells are highly tumorigenic compared with the SPm and SPn (hox) fractions. The idea of TSC was proposed approximately 150 years ago, and recent advances in stem cell biology have given fresh

10 Das, Tsuchida, Malkin et al Figure 6. The SPm (hox) fraction is enriched with Oct-4-expressing cells. (A): Quantitative reverse transcription-polymerase chain reaction data showing the Oct-4 mrna level in the SPm (hox) fraction differed significantly from the SP fraction in SK-N-BE(2) and RH-4 cells. The Ct value of a tumor cell sample was calibrated against the Ct value of an hes cell (BG01V) sample to obtain the Ct value as described [61]. Error bars indicate SEM. (B): Confocal images of SK-N-BE(2) and RH-4 SPm (hox) cell showing the Oct-4 staining cells. Tera-2 SP cells were used as a positive control, whereas RH-4 non-sp cells were used as a -ve control ( 63). On the right, the histogram shows that SPm (hox) cells are highly enriched in Oct-4-positive cells compared with SP and SPm cells. Cells were stained with Oct-4 antibody and counted per 100 4,6-diamidino-2-phenylindolepositive cells in 10 microscopic fields SK-N-BE(2): SPn (hox), 1.5 1; SPm (hox), ; versus SPm, (p.038), and SP, (n 3). RH-4: SP, ; SPn (hox) : 1.0 1; SPm (hox) ; versus SPm, ; p.0314 (n 3). Tera-2 cells were used as a positive control. (Ci): A representative fluorescence-activated cell sorting profile of SK-N-BE(2) xenograft-derived SP cells. (Cii): A small fraction of sorted SP cells showed high Oct-4 and high pimonidazole staining (Q5 region). The sorted SP cells were stained with anti-pimonidazole (chicken anti-mouse IgG Alexa Fluor 488 as secondary antibody) and anti-oct-4 (goat anti-mouse phycoerythrin as secondary antibody). Results shown are representative of a typical experiment out of more than three independent experiments (data from 10,000 single-cell events). (D): A schematic model of the proposed stemness switch showing the expansion of Oct-4-expressing, highly tumorigenic cells from a quiescent state to an actively expanding state (the switch) following hypoxia/reoxygenation stress. The nontumorigenic cells are shown in orange., p.05. Abbreviations: -ve, negative; hes, human embryonic stem; SP, side population; SPm, migratory side population; SPm (hox), post-hypoxia migratory side population; SPn, nonmigratory side population; SPn (hox), post-hypoxia nonmigratory side population. understanding of the cancer stem cell hypothesis. The most important concept of the hypothesis is that a small fraction of tumor cells retains a self-renewal property that drives the tumorigenic process [42]. On the basis of the self-renewal property, TSCs have been identified, including the isolation of a SP fraction from various tumors [43]. Recent advances in stem cell biology suggest that apart from the high self-renewal capacity, stem cells may have another important property: the migration and homing to an area of injury [44]. Imitola et al. showed that injury-related SDF-1 attracts central nervous system stem cells

11 1828 Hypoxia Enhances Tumor Stemness to the area of injury [45]. This migration to the area of injury/ wound may also play an important role in tumor angiogenesis and metastasis. Kucia et al. argued that metastatic cancer cells may have a similar property of migration to areas of injury where SDF-1 may play an important role [46]. Furthermore, it has been shown that bone marrow-derived endothelial progenitor cells migrate to tumors for neoangiogenesis [47]. Here we developed a novel experimental model to isolate a highly tumorigenic fraction of SP cells on the basis of the property of stem cell migration to an area of injury. We used an injured conditioned medium to attract the SPm fraction. The injured conditioned medium was prepared by treating human BM stromal cells with hypoxia and hydrogen peroxide. Such a treatment enhances SDF-1 secretion and may simulate the hypoxic microenvironment of a tumor. We found that a SDF-1 concentration (2 ng/ml) similar to that found in the injured conditioned medium attracted approximately half the SPm (hox) cells compared with the injured conditioned medium (supplemental online Fig. 3), suggesting that additional chemokines present in the medium may assist in recruitment of SPm (hox) cells. We recently found that the CCL-2 chemokine is expressed in the SK-N- BE(2) cells [48]. The potential role played by CCL-2 and other chemokines in the recruitment of SPm (hox) in the hypoxic zones require further investigation. The isolated SPm fraction showed higher tumorigenic activity than the SPn fraction and also showed a higher number of Oct-4-positive cells, suggesting that the SPm fraction is enriched in highly tumorigenic, TSC-like cells. We then developed a pimonidazole-based isolation of SP fraction in the hypoxic zone in vivo and showed that most of the SPm (hox) cells were also stained with pimonidazole. The results suggest that hypoxia may act as zone of injury in a tumor, where highly tumorigenic cells may migrate and/or accumulate. Hypoxia has been found to play an important role in the migration of normal stem cells. Ceradini et al. showed that stem cells, including endothelial progenitor cells, migrate to the area of hypoxic zones mediated by SDF-1 gradient [5, 6]. Normal stem cells have also been found to migrate to the tumor hypoxia zone. In a murine glioblastoma model, hypoxia greatly increased the migration of endothelial progenitor cells [49]. Therefore, our demonstrations that QD-labeled SPm (hox) cells migrate, home, and then localize to the hypoxic niche following intracardiac injection suggests that TSCs may recruit the stem cell s unique ability to home to hypoxic tissues. The mechanism of such migration and homing to the hypoxic zone is not clear. It is possible that a chemokine gradient including SDF-1 may play an important role that requires further investigation. Considering that intracardiac-injected QD-labeled SPm (hox) cells were able to home and then migrate to the area of hypoxia within the tumor, it is possible that a communication channel may exist between distant vessels and the hypoxic niche within the tumor. A physiological process such as vessel mimicry may connect the hypoxic niche with the distant vessels for adequate TSC trafficking (Das B. Idea and evidence for tumor stemness switch. International Workshop on Cancer Stem Cell, European Institute of Oncology, Milan, Italy. November 11, 2005). A key feature of a normal stem cell is its interaction with the immediate microenvironment, thereby forming the stem cell niche [50]. The stem cell-niche interaction plays a key role in the stress-induced expansion and mobilization of stem cells. In the BM microenvironment, during steady state homeostasis, HSCs in their niche remain quiescent, and only a few HSCs are found in the peripheral circulation. Following chemotherapyinduced stress such as cyclophosphamide treatment, SDF-1 is released in the marrow, leading to proliferation of quiescent stem cells and their subsequent mobilization into the circulation for repair/regeneration [51, 52]. Recently, Lévesque et al. showed that following cyclophosphamide-induced stress, the hypoxic zone in the BM niche increases, suggesting a potential role of hypoxic niche in the HSC mobilization process [1]. Hypoxia may be a feature of the HSC niche [4]. Quiescent bone marrow stem cells (Lin CD34 CD 38 ) expand following exposure to hypoxia, leading to increased BM repopulation activity [53]. Similar to the role of hypoxia-induced stress in HSC expansion and mobilization, tumor hypoxia may also expand TSCs. We found that a quiescent population of SP cells (the SPm (hox) cells) expand following hypoxia/reoxygenation stress. This process of switching from a quiescent to an active state of proliferation of TSCs may be described as a process of stemness switch (Fig. 6D) (Das B. Idea and evidence for tumor stemness switch. International Workshop on Cancer Stem Cell, European Institute of Oncology, Milan, Italy. November 11, 2005) and may contribute to expansion of TSCs, leading to tumor aggressiveness. Mackillop et al. speculated that TSC expansion may lead to tumor progression [41]. Singh et al. found that aggressive brain tumors having an increased proportion of CD133-positive cells correlated with in vitro primary sphere formation ability [33], suggesting a potential correlation between number of TSCs and tumor aggressiveness (short latency and progression period). We found that injection of SPm (hox) cells (containing 50% Oct-4-positive cells) was associated with a significantly shorter latency period compared with a similar number of SPm cells (containing 20% Oct-4- positive cells) (supplemental online Fig. 4), suggesting that an increased proportion of TSCs in a given tumor cell population may accelerate the tumor growth. The tumor hypoxic niche may serve as a natural site for the enrichment/expansion of TSCs and subsequent rapid tumor progression, which may explain the close correlation among degree of hypoxia and tumor aggressiveness [39]. Singh et al. found that aggressive glioblastoma contains 20% 28% CD133-positive brain tumor TSCs [33]. Al-Hajj et al. found 11% 35% CD44 CD24 /low Lineage breast cancer stem cells in primary tumors [32]. O Brien et al. found 12% (mean) CD133-positive colon cancer stem cells in primary tumors [37]. Interestingly, the aggressiveness of these solid tumors is correlated with hypoxia [54, 55]. Therefore, a potential correlation between the proportion of TSCs, tumor hypoxia, and tumor aggressiveness should be investigated. It is important to investigate the molecular mechanism that could maintain and expand the SPm (hox) cells in their hypoxic niche. We found that the Oct-4-positive SPm (hox) fraction is increased following exposure to hypoxia. Earlier, it has been shown that hypoxia may enhance Oct-4 expression by upregulating HIF-2 [56]. Recently, we found that vascular endothelial growth factor (VEGF)/Flt1 autocrine signaling may regulate Oct-4 expression in tumor SP cells [21]. Earlier, we reported that an autocrine loop between HIF-1 and VEGF/Flt1 was involved in the survival of a highly drug-resistant fraction of SK-N-BE(2) cells during hypoxia [19]. We are currently investigating the potential link between the HIF-1 /HIF-2 and Oct-4 pathway in the survival and self-renewal of SPm (hox) cells in the hypoxic microenvironment. SUMMARY We isolated a highly enriched TSC-like SPm (hox) cell fraction on the basis of the concept of migration of SP cells to sites of injury. Originally, the concept of TSC was proposed on the basis of the striking similarity in self-renewal activity of both normal stem cells and tumor cells [10, 11, 33, 42, 57 59]. Another, very important activity of normal stem cell is the migration to the site of injury [45, 60]. Therefore, our results indicate that the tumor