Prospective identification of tumorigenic osteosarcoma cancer stem cells in OS99-1 cells based on high aldehyde dehydrogenase activity

Prospective identification of tumorigenic osteosarcoma cancerstem cells in OS99-1 cells based on high aldehydedehydrogenase activity

Lin Wang1, Paul Park1, Huina Zhang1,2, Frank La Marca1 and Chia-Ying Lin1,2

1 Department of Neurosurgery, Spine Research Laboratory, University of Michigan Medical School, Ann Arbor, MI2 Department of Biomedical Engineering, University of Michigan, Ann Arbor, MI

High aldehyde dehydrogenase (ALDH) activity has recently been used to identify tumorigenic cell fractions in many cancer

types. Herein we hypothesized that a subpopulation of cells with cancer stem cells (CSCs) properties could be identified in

established human osteosarcoma cell lines based on high ALDH activity. We previously showed that a subpopulation of cells

with high ALDH activity were present in 4 selected human osteosarcoma cell lines, of which a significantly higher ALDH

activity was present in the OS99-1 cell line that was originally derived from a highly aggressive primary human osteosarcoma.

Using a xenograft model in which OS99-1 cells were grown in NOD/SCID mice, we identified a highly tumorigenic

subpopulation of osteosarcoma cells based on their high ALDH activity. Cells with high ALDH activity (ALDHbr cells) from the

OS99-1 xenografts were much less frequent, averaging 3% of the entire tumor population, compared to those isolated directly

from the OS99-1 cell line. ALDHbr cells from the xenograft were enriched with greater tumorigenicity compared to their

counterparts with low ALDH activity (ALDHlo cells), generating new tumors with as few as 100 cells in vivo. The highly

tumorigenic ALDHbr cells illustrated the stem cell characteristics of self-renewal, the ability to produce differentiated progeny

and increased expression of stem cell marker genes OCT3/4A, Nanog and Sox-2. The isolation of osteosarcoma CSCs by their

high ALDH activity may provide new insight into the study of osteosarcoma-initiating cells and may potentially have

therapeutic implications for human osteosarcoma.

Over the past decade, increasing evidence has supported thenotion that tumors are organized by a hierarchy of heteroge-neous cell populations with different proliferation potentialsin which the capability to initiate tumor formation and pro-mote tumor growth exclusively resides in a small subpopula-tion of tumor cells termed cancer stem cells (CSCs) or tu-mor-initiating cells.1,2 According to the CSC model, CSCsmay originate from stem cells or progenitor cells in trans-formed tissues through deregulated self-renewal. Like normalstem cells, CSCs have self-renewal ability, which drivestumorigenicity. Moreover, these cells have the capability todifferentiate, albeit aberrantly, giving rise to a heterogeneouspopulation of differentiated cells that are non-tumorigenic,lack the ability to self-renew, have limited proliferation

potential and constitute the bulk of the tumor. The stem cell-like phenotype of CSCs and their rare number within the tu-mor may account for their ability to escape from conven-tional therapies, thus leading to tumor recurrence and even-tually metastasis, even when the primary lesion has beeneradicated.3,4 Although the specific markers may differ fromone tumor to another, CSCs are primarily characterized bythe ability to form new tumors through serial transplantationin immunodeficient nonobese diabetic (NOD)/severe com-bined immunodeficient (SCID) mice, as well as by thecapacity to display stem/progenitor cell properties such asself-renewal and the ability to reestablish tumor heterogene-ity.5 The identification of CSC populations has significantlyincreased understanding of tumor biology and has importantimplications for new cancer therapies.

Osteosarcoma is the most common primary tumor of bone,with peak incidence in the teenage years. Approximately 400new pediatric cases occur per year in the United States.6

Recently, the existence of stem-like cells in primary osteosarco-mas and cell lines derived from human osteosaracoma wasdemonstrated in a subpoplation of cells capable of self-renewal.These cells have been detected in spherical clones under an-chorage-independent, serum-starved culture conditions, as sidepopulation (SP) cells based on efflux of Hoechst 33342 dye oras CD133þ cells sorted using cancer stem cell marker CD133.7–9 Our previous study also confirmed the presence of a stem-like cell population in 4 separate human osteosarcoma cell

Key words: aldehyde dehydrogenase, cancer stem cells, OS99-1,

osteosarcoma, tumorigenesis

Grant sponsor: Department of Neurosurgery, University of

Michigan

DOI: 10.1002/ijc.25331

History: Received 30 Nov 2009; Accepted 24 Feb 2010; Online 22

Mar 2010

Correspondence to: Chia-Ying Lin, Department of Neurosurgery,

University of Michigan Medical School, Biomedical Science Research

Building, Room 5007, 109 Zina Pitcher Place, Ann Arbor, MI

48109-2200, USA, Tel.: 734-615-0371, Fax: 734-763-7322, E-mail:

[email protected]

Can

cerCellBiology

Int. J. Cancer: 128, 294–303 (2011) VC 2010 UICC

International Journal of Cancer

IJC

lines, as characterized by sphere-forming capacity and byexpression of the stem cell markers.10 Most recently, Levings etal. reported expression of an exogenous human Oct-4 promoterthat was used to identify tumor-initiating cells in osteosar-coma.11 However, no study has reported the identification ofcancer stem-like cells in human osteosarcoma cell lines basedon high aldehyde dehydrogenases (ALDH) activity.

Cytosolic ALDHs are a group of enzymes involved in oxi-dizing a wide variety of intracellular aldehydes into their corre-sponding carboxylic acids.12 Among these enzymes, ALDH1 isthought to have an important role in oxidation of alcohol andvitamin A and in cyclophosphamide chemoresistance.13 Ele-vated levels of ALDH activity (ALDHbr cells) have been foundin murine and human hematopoietic and neural stem and pro-genitor cells compared to other cells.14–16 More recently, highALDH activity has been used to define stem cell populations inmany cancer types including breast cancer,17 liver cancer,18 co-lon cancer19 and acute myeloid leukemia.20 In this study, weinvestigated whether high ALDH activity can be used to iden-tify CSCs in human osteosarcoma cell lines.

Material and MethodsCell culture

Human osteosarcoma cell lines MG63 and Saos-2 were pur-chased from American Type Culture Collection (Manassas,VA). Human osteosarcoma cell line Hu09 was purchased fromHealth Science Research Resources Bank (Osaka, Japan).Human osteosarcoma cell line OS99-1 was obtained from Dr.Sheila M. Nielsen-Preiss (Montana State University). It shouldbe noted that the OS99-1 cell line was originally derived froma highly aggressive primary human osteosarcoma.21 All cellswere routinely cultured in Dulbecco’s Modified Eagle Medium(DMEM)/F12 medium (Gibco, Carlsbad, CA) supplementedwith 10% fetal bovine serum (FBS; Gibco) in a humidifiedatmosphere of 5% CO2 in air at 37�C. For consistent results,cells for ALDEFLUOR cell analysis were used within 2 to 3passages of each cell line and repeated at least 3 times asdescribed previously.22 This was performed to verify the con-sistency of the percentage of ALDHbr cells in each cell lineand confirmed that no significant changes between P2 and P3occurred in each cell line.

ALDEFLUOR cell analysis and flow cytometry sorting

To characterize the population that expresses high levels ofALDH (ALDHbr ) in 4 cell lines, the ALDEFLUOR kit(Stemcell Technologies, Vancouver, BC, Canada) was used toisolate ALDHbr cells, as described previously.17 Incubation ofcells with ALDEFLUOR in the presence of the specificALDH-inhibitor dimethylaminobenzaldehyde (DEAB), whichresults in decreased fluorescence, was used as a negativestaining control for the assay.14–17

Implantation of sorted cells into NOD/SCID mice

To assess the difference in tumorigenicity of ALDHbr andALDHlo cells sorted from the OS99-1 cell line, freshly sorted

cells were washed and suspended in 200 ll of serum-freeHBSS/Matrigel (BD Biosciences, San Jose, CA) mixture (1:1volume) and then injected subcutaneously into the right andleft lower abdominal area of NOD/SCID mice. Our previousexperience indicated that a high degree of cellular dispersionwould be produced when sorted cells were subcutaneouslyinjected into NOD/SCID mice without stromal support;therefore, we used Matrigel to act as a physical barrier toprevent cellular dispersion and the subsequent loss of injectedcells. Coinjection with Matrigel has been shown to increasetumor formation and yield optimal tumor growth.23 Tumorgrowth was monitored weekly for 32 weeks. Tumors formedwere removed and a portion of each tumor was processed forhistological analysis. All animal studies were performedaccording to protocol approved by the Institutional AnimalCare and Use Committee of the University of Michigan.

Xenograft enrichment of human osteosarcoma

OS99-1 cells

It has been shown that cancer stem-like cells can be enrichedwith their tumorigenic capability in the reconstituted, xeno-grafted tumor. To assess whether tumorigenicity of cells iso-lated from xenografts can be enhanced in xenograft comparedto those directly isolated from the cell line, unsorted OS99-1cells were injected into NOD/SCID mice to establish xenografttumors. Tumors formed were removed and harvested for prep-aration of single cell suspensions and histology.

Injection of single cell suspensions of sorted tumor cells

and serial transplantation

Xenografted tumors formed by injection of either unsortedor sorted OS99-1 cells were minced with scissors, mixed with1 mg/ml collagenase Type II (Sigma-Aldrich Co., St. Louis,MO), incubated for 3–4 hr, passed through a 70-lm cellstrainer and then washed twice with DMEM/F12/10% FBSmedium. Cells were resorted into ALDHbr and ALDHlo frac-tions as described above. During flow cytometry, mouse H-2Kb antibody (BD Biosciences Pharmingen, San Jose, CA)was used to eliminate cells of mouse origin from the xeno-graft tumors. Sorted cells were reinjected into the animals asthe primary transplantation. The entire procedure wasrepeated with the formed tumors to conduct the secondaryand tertiary transplantations.

Immunohistochemistry/immunofluorescence

For immunohistochemistry, formalin-fixed paraffin sectionsof tumor samples were stained with goat anti-ALDH1 anti-body (1:50; Santa Cruz Biotechnology, Inc., Santa Cruz, CA)using standard procedures. Immunofluorescent staining wasperformed as we previously reported.10

Cell proliferation assay

The freshly sorted ALDHbr and ALDHlo cells from xenograftswere cultured at a density of 2 � 103 cells per well in a 96-well plate triplicates with 100 ll of culture medium and

Can

cerCellBiology

Wang et al. 295


allowed to grow for 10 days. Cell proliferation was assessedby using Celltiter96 AQueous One Solution reagent asdescribed previously.24 Briefly, at specific time point duringcultivation, the medium was discarded and replaced with 100ll of PBS and then 20 ll of Celltiter96 AQueous One Solutionreagent (Promega, Madison, WI) was added to each well andincubated at 37�C for 2 hr. The absorbance at 490 nm wasrecorded using Bio-TEK ELx 800 Plate Reader (Bio-TekInstruments, Inc., Wilrijk, Belgium).

Soft agar assay

Freshly sorted ALDHbr and ALDHlo cells from xenograftswere seeded in 6-well plates coated with a 1% agarose bottomlayer in culture medium containing 10% FBS, a middle layerof 0.6% agarose including 5 � 103 cells and a top layer ofculture medium only.25 The plates were incubated at 37�Cfor 4 weeks and fixed with formalin for 20 min. Plates werethen stained with 0.005% crystal violet for 1 hr, and colonieswere visualized by trans-UV illumination and counted usinganalysis software QuantityOne (Bio-Rad, Hercules, CA).

Quantitative reverse transcription polymerase chain

reaction (qRT-PCR) analysis of Oct3/4A,

Nanog and Sox-2 mRNA

To assess mRNA expression levels of embryonic stem cellmarkers Oct3/4A, Nanog and Sox-2 in the ALDHbr andALDHlo cells from xenografts, qRT-PCR assay was performedas we previously reported.10 The Eppendorf MastercyclerRealplex Detection System (Eppendorf, Germany) was usedfor conducting qRT-PCR.

Statistical analysis

Data were expressed as mean 6 SD. Statistically significant dif-ferences were determined by 1-way ANOVA analysis, Student t-test and v2 analysis, where appropriate, and defined as p < 0.05.

ResultsHuman osteosarcoma cell lines contain ALDHbr population

Using the marker ALDH that was previously used to identifyCSCs in breast and colon cancer,17,19 we identified a similarpopulation in human osteosarcoma by exposing 4 human os-teosarcoma cell lines to ALDEFLUOR and then analyzing byflow cytometry. In each experiment, a sample of cells wasstained simultaneously with ALDEFLUOR and DEAB, a spe-cific inhibitor of ALDH used to identify the ALDHbr popula-tion. As shown in Figure 1a, all 4 cell lines tested demon-strated ALDH enzymatic activity, although the levels ofactivity were different. The OS99-1 cell line contained 45.07%6 3.75% ALDHbr cells, whereas the Hu09 cell line contained1.84% 6 0.31% ALDHbr cells, which was close to that foundin the Saos-2 cell line (1.56% 6 0.34%), and much higherthan that found in the MG63 cell line (0.59% 6 0.16%). Thelevel of ALDH in OS99-1 was significantly higher comparedto the other 3 cell lines (p < 0.001), whereas no statisticallysignificant differences were detected among the other 3 cell

lines (Fig. 1b). As ALDHbr cells were more abundant inOS99-1 cells, attention was focused on OS99-1 cells.

Tumor-initiating cells are rare in the ALDHbr population

sorted directly from the OS99-1 cell line

To determine whether ALDHbr cells sorted from the OS99-1cell line are more tumorigenic than their ALDHlo counterpartsin vivo, we injected both types of cells separately into NOD/SCID mice, in addition to injection of unsorted cells alone.The sorted ALDHbr and ALDHlo cells, as well as unsortedcells, were all maintained subcutaneously in mice, and tumorgrowth was monitored weekly for 32 weeks, when animalswere sacrificed. The absence or presence of tumor was con-firmed by histological examination. Tumor formation withunsorted OS99-1 cells was highly dependent on the number ofinjected cells. No tumors developed in NOD/SCID mice untilat least 1 � 105 unsorted OS99-1 cells were injected (Table 1).For the sorted cells from the OS99-1 cell line, surprisingly, nostatistically significant difference in tumor formation wasnoticed between injections with ALDHbr and ALDHlo cells (p> 0.05; Table 1). Tumor formation could occasionally befound with injection of 1 � 104 and 1 � 103 ALDHbr cells in1 of 6 mice, respectively, whereas no tumors were observedwith 5 � 104 ALDHbr cells in all 6 mice (Table 1). The size ofthe tumors grew moderately so that the injection of 1 � 104

and 1 � 103 ALDHbr cells developed palpable tumors after 11weeks and 18 weeks, respectively. In comparison, tumor wasformed in 1 of 6 mice 16 weeks after injection with ALDHlo

cells when the injected cell number was 1 � 104.The results revealed that tumorigenic cells that exist in the

ALDHbr population sorted directly from the OS99-1 cell lineare rare and these few cells are not able to consistently giverise to tumors in NOD/SCID mice.

Decreased ALDHbr cells in OS99-1 xenografts

To further validate whether tumorigenicity of cells in thereconstituted tumors differs from that of cells in the cell line,we next isolated ALDHbr cells from xenograft tumors basedon previous findings that xenografts retained many featuresof the primary tumor on multiple passages,26 and using axenograft model to isolate CSCs has been validated in severalcancers.27–32 Surprisingly, we found that the percentage ofALDHbr cells derived from OS99-1 xenografts was 3.13% 60.53% (Fig. 2a), which was around a 15-fold reduction (Fig.2b) compared to that of cells directly isolated from theOS99-1 cell line (Fig. 1a), as described earlier. Immunostain-ing results further confirmed that ALDH-positive cells dra-matically decreased in OS99-1 xenografts compared to thatof the OS99-1 cell line (Figs. 2c and 2d).

Enhanced cell growth rate and colony-forming ability on

soft agar with ALDHbr cells sorted from OS99-1 xenografts

To determine whether ALDHbr cells are more proliferativeand clonogenic than their ALDHlo counterparts after xeno-graft enrichment, we collected ALDHbr and ALDHlo cells

Can

cerCellBiology

296 Identification of osteosarcoma cancer stem cells


from the OS99-1 xenografts and compared their proliferativeand clonogenic abilities using cell proliferation assay and softagar colony formation assay, respectively. After sorting by

flow cytometry, ALDHbr and ALDHlo cells were immediatelycultured in DMEM/F12 medium containing 10% FBS to per-form the proliferation assay. After 6 to 10 days, ALDHbr cells

Figure 1. (a) Representative flow cytometric analysis of cells with high ALDH activity in 4 human osteosarcoma cell lines: OS99-1, 45.07%

6 3.75%; Hu09, 1.84% 6 0.31%; Saos-2, 1.56% 6 0.34% and MG63, 0.59% 6 0.16%. Cells incubated with ALDEFLUOR substrate and

specific inhibitor of ALDH, DEAB, were used to establish the baseline fluorescence of these cells (R0) and to define the ALDHbr population

(R1). (b) Evaluation of ALDHbr cell percentage indicated OS99-1 cells highly expressed ALDH statistically, as compared to other cell lines

(*p < 0.001). Data represent mean 6 SD and are representative of 3 independent experiments.

Table 1. Ability for tumor formation in sorted cells from human osteosarcoma OS99-1 cell line based on ALDH activity

Cell type

Number of cells per injection

1 3 107 1 3 106 1 3 105 5 3 104 1 3 104 5 3 103 1 3 103 5 3 102

Unsorted 5/5 4/5 2/3 0/3 – – – 0/3

ALDHbr – – – 0/6 1/6 0/3 1/3 0/3

ALDHlo (p-value) – – – 0/6 1/6 0/3 0/3 (0.5) 0/3

Note: OS99-1 cells were isolated by flow cytometry as described in Figure 1a based on high ALDH activity and assayed for the ability to formtumors after injection into subcutaneum of NOD/SCID mice. Mice were examined for tumor formation by palpation and subsequent autopsy. Dataare expressed as number of tumors formed/number of injections. Statistically significant difference in tumor formation between ALDHbr cells andALDHlo cells was determined by v2 analysis.

Can

cerCellBiology

Wang et al. 297


reached a logarithmic growth phase, whereas the ALDHlo

cells still grew slowly. The cell proliferation of ALDHbr cellsin cell culture was evaluated to be 4- to 8-fold faster thanthat of ALDHlo cells (p < 0.001; Fig. 3a).

Clone formation assays were also performed after cell sorting.Four weeks after culture, mean clone formation was 25% 62.5% and 3%6 1.2% in ALDHbr and ALDHlo cells, respectively.Statistical analysis showed significant differences in clone forma-tion between the 2 populations (p < 0.001; Figs. 3b and 3c).

Increased gene expression of stem cell markers in ALDHbr

cells sorted from OS99-1 xenografts

Oct3/4, Nanog and Sox-2 are essential transcription factorsthat play a critical role in maintenance of self-renewal and plu-ripotency of embryonic stem cells.33 To determine whether

ALDHbr cells of OS99-1 enriched in xenograft would displayproperties of stem cells and preferentially express those genes,total RNA was extracted from freshly sorted ALDHbr andALDHlo cells, and qRT-PCR was performed to assess geneexpression by the 3 stem cell markers. As shown in Figure 3d,Oct3/4, Nanog and Sox-2 in ALDHbr cells were all consistentlyhigher than those in ALDHlo cells (p < 0.05; p < 0.001).

Enriched propensity of tumor initiation in

ALDHbr OS99-1 cells

To determine whether ALDHbr cells sorted from xenografts aremore tumorigenic than their ALDHlo counterparts in vivo, wetransplanted 100 to 10,000 freshly sorted ALDHbr and ALDHlo

cells from OS99-1 xenograft subcutaneously into NOD/SCIDmice. Experiments were performed in triplicate, as previously

Figure 2. Decreased number of ALDHbr cells shown in xenografts formed by injection of OS99-1 cells. (a) Representative flow cytometric

analysis of ALDHbr cells in OS99-1 xenografts; percentage of ALDHbr cells was 3.13% 6 0.53% in xenograft tumors. Inset shows negative

controls when cells were treated with ALDH-inhibitor DEAB. Data represent mean 6 SD and are representative of 3 independent

experiments. (b) Percentage of ALDHbr cells in OS99-1 xenografts had around 15-fold reduction compared with ALDHbr cells isolated from

OS99-1 cell line, as described in Figure 1a. (c) Immunofluorescent staining shows ALDH expression clearly detected in the cytoplasm of

OS99-1 cells (red areas), and nuclei counterstained with 40, 6-diamidino-2-phenylindole (DAPI) (blue areas). (d) Immunohistochemical

staining indicated sparse ALDH-positive cells detected in OS99-1 xenografts using horseradish peroxidase, with brown staining indicating

ALDH-positive cells. Bar ¼ 50 lm. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Can

cerCellBiology



described.17 Only the injection of ALDHbr cells gave rise to visibletumors, whereas no tumors were observed with ALDHlo cells,even when as many as 10,000 cells were injected (Table 2). Thesize and latency of tumor formation with ALDHbr cells highlycorrelated with the number of injected cells, where the injectionof as few as 100 ALDHbr cells resulted in tumor growth in 2 of 3mice (Fig. 4a). Injection with cell numbers greater than 1,000 wascapable of developing tumors in all experimental animals. Hema-toxylin and eosin staining of the tumor sections confirmed thattumors formed by ALDHbr cells contained malignant cells (Fig.4b), whereas only residual Matrigel and apoptotic cells were seenat the sites of injection with ALDHlo cells (Fig. 4c).

The ability for self-renewal and recapitulation in diverse

phenotype of ALDHbr cells in OS99-1 enriched xenograft

Normal stem cells are defined by their ability to both self-renew and generate phenotypically diverse progeny. To test if

our highly tumorigenic cells also exhibited these properties,we performed serial transplantation with ALDHbr cellsderived from primary tumors that grew from an initial injec-tion of 10,000 ALDHbr cells. ALDHbr cells isolated from theprimary OS99-1 xenografts were injected into mice to formsecondary tumors, followed by the development of tertiarytumors with the secondary-derived ALDHbr cells. The result-ant tumors were reanalyzed for ALDH-marked cells. Thepercentage of ALDHbr cells in the primary, secondary andtertiary tumors were 3.35% 6 0.79%, 3.45% 6 1.21% and3.39% 6 0.82%, respectively, which were all close to the orig-inal xenograft from which they were derived (Fig. 5a–c).

The highly tumorigenic ALDHbr cells produced additionalALDHbr cells as well as phenotypically diverse nontumori-genic cells, recapitulating the same phenotypic complexity ofthe OS99-1 xenografted tumors from which the tumorigeniccells were derived. The histological appearance of the serially

Figure 3. Increased cell growth rate, clone formation ability and stem cell marker gene expression in vitro in ALDHbr cells sorted from OS99-1

xenografts. (a) Freshly isolated ALDHbr and ALDHlo cells were cultured in 96-well plates at a density of 2,000 cells per well and allowed to

grow for 10 days, at specific time point during cultivation, the medium was discarded and replaced with 100 ll of PBS, and then 20 ll of

Celltiter96 AQueous One Solution reagent was added to each well and incubated at 37�C for 2 hr. The absorbance at 490 nm was recorded.

ALDHbr cells grew faster than ALDHlo cells (p < 0.001). (b, c) Freshly isolated ALDHbr and ALDHlo cells were cultured in 6-well plates at a

density of 5,000 cells per well and allowed to grow for 4 weeks. More cell colonies could be found in ALDHbr cells than in ALDHlo cells (p <

0.001). (d) Relative quantitative mRNA expression of Oct3/4A, Nanog and Sox-2 genes in ALDHbr and ALDHlo cells freshly isolated from OS99-1

xenografts. Gene expression levels were normalized to b-actin. Enriched ALDHbr cells expressed significantly higher Oct3/4A, Nanog and Sox-2

compared to their ALDHlo counterparts from the xenografts (*p < 0.05; **p < 0.001). Each experiment was performed 3 times; representative

examples are shown. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Can

cerCellBiology

Wang et al. 299


transplanted tumors was identical, and immunohistochemicalstaining with ALDH1 antibody further confirmed that distri-butions of ALDH expressed in sections of the serially trans-planted tumors were also very similar (Fig. 5d). These datasuggest that ALDHbr cells in OS99-1 or, even more generally,osteosarcoma, represent CSCs that are capable of undergoingself-renewal to maintain the ALDHbr population while pro-ducing differentiated progeny that are nontumorigenic andsimilar to ALDHlo cells.

DiscussionOur investigation indicates that in osteosarcoma, high ALDHactivity can be used to identify a subpopulation of cells thatare highly tumorigenic. Specifically, results from the highlyaggressive human osteosarcoma OS99-1 cell line suggest thata subpopulation of cells play a role in tumor malignancy andexpansion. Although the ALDHbr cells derived from OS99-1xenografts represented a small fraction of the tumor popula-tion, the cells were able to initiate new tumors in NOD/SCIDmice. In addition, ALDHbr cells displayed several featurestypically seen in CSCs, including the ability for self-renewal,generation of differentiated progeny, recapitulation of the tu-mor phenotype from which they were derived, and increasedexpression of stem cell marker genes.

The concept of a CSC was originally introduced based onthe observation that when cancer cells of many different typeswere assayed for their proliferative potential in various assaysin vitro and in vivo, only a minority of cells showed extensiveproliferation.1 Although CSCs have been identified in a varietyof malignancies,27–32 the identification of CSCs in human os-teosarcoma has been more difficult than in tumors originatingfrom other types of tissues. Because of differences in mesen-chymal origin, the markers that have been identified and devel-oped for hematologic, neural and epithelial cancers are notnecessarily applicable for isolation of CSCs from osteosarcoma.To date, the existence of such a stem-like cell population inhuman primary osteosarcomas and osteosarcoma cell lines hasbeen mostly detected by the expression of stem cell markergenes as well as their ability to form spheroids in vitro,7,10 butcells with these stem cell attributes have not been characterizedby their ability to form tumor in vivo. Recently, a separate

study has isolated SP cells from human primary osteosarcomasusing efflux of Hoechst 33342 dye and has demonstrated thatSP cells could initiate tumors in immunodeficient mice.8 How-ever, it has been suggested that identification of the CSCs can-not solely rely on SP sorting, as the possessed SP phenotype isnot universally presented in all CSCs and there may exist otherdefensive mechanisms for CSCs to evade drug therapies thatcannot be identified by Hoechst dye staining.34 Therefore, theSP cells identified by Hoechst dye do not fully represent allCSCs. Seeking a reliable marker to screen tumor-initiating cellswhile preserving most of their functionality and viability thusbecomes extremely critical.

We chose the marker ALDH as a starting point based onprior work on breast CSCs, in which cells with high ALDH ac-tivity were identified as putative CSCs.17 Our results show thatall 4 cell lines contained different percentage of ALDHbr cells,but OS99-1 cell line contained the highest percentage ofALDHbr cells compared to the other 3 cell lines, where Saos-2and MG63 cells were reported as non-tumorigenic cells,35–37

which implies that ALDH activity may be related to theaggressiveness of tumors, as the OS99-1 cell line was originallyestablished from a high-grade human osteosarcoma.21

It is interesting to note that the retained ALDHbr cellswere dramatically decreased when OS99-1 cells grew inreconstituted xenografts. This result is consistent with a pre-vious report showing that ALDH activity in a lung cancercell line had 3-fold reduction occurring in vivo duringgrowth of the tumor.38 A possible explanation for the

Table 2. Ability for tumor formation in sorted cells from humanosteosarcoma OS99-1 xenografts based on ALDH activity

Cell type

Number of cells per injection

1 3 104 5 3 103 1 3 103 5 3 102 1 3 102

ALDHbr 3/3 3/3 3/3 2/3 2/3

ALDHlo 0/3 0/3 0/3 0/3 0/3

Note: Cells from OS99-1 xenografts were isolated by flow cytometrybased on high ALDH activity as described in Figure 2a and assayed forthe ability to form tumors after injection into the subcutis of the NOD/SCID mice. Mice were examined for tumor formation by palpation andsubsequent autopsy. Data were expressed as number of tumorsformed/number of injections.

Figure 4. Tumor formation in NOD/SCID mice injected with highly

tumorigenic cells from OS99-1 xenografts. (a) Representative tumor

growth in a NOD/SCID mouse at the injection site, even with as

few as 100 ALDHbr cells. No tumor formation seen at the injection

site of 100 ALDHlo counterpart. (b) H&E staining of tumor

generated from ALDHbr cells reveals presence of malignant cells.

(c) Injection site of ALDHlo cells contained only residual Matrigel

and apoptotic cells. Bar ¼ 50 lm. [Color figure can be viewed in

the online issue, which is available at wileyonlinelibrary.com.]

Can

cerCellBiology



variation in ALDH activity may be due to different growthconditions. Changes in phenotypical characteristics and amultitude of genetic aberrations could take place within can-cer cell lines passaged in vitro.39 Thus, it is probably morereliable to use an in vivo model that provides a physiologicenvironment to develop candidate CSCs that can be isolatedin human osteosarcoma.

Using freshly sorted ALDHbr and ALDHlo cells fromOS99-1 xenografts, our in vitro experiments revealed thatALDHbr cells grew faster and had a greater ability to formcell colonies than ALDHlo cells, supporting the theory thatALDHbr cells possess stem cell properties of self-renewal andhigh proliferation potential. Using xenograft to derive theALDHbr cells, we further discovered that these cells hadmuch greater tumorigenicity. The subpopulation of ALDHbr

cells that remained (an average of 3%) gave rise to newtumors in animals consistently with the injection of 1,000ALDHbr cells, and even with as few as 100 ALDHbr cellsinjected the incidence of tumor formation remained 2 of 3,which reoccurred through serial transplantation. In addition,based upon analysis of reformed tumors arising from the

transplantation of isolated ALDHbr cells, both ALDHbr andALDHlo cells were present concurrently, and distribution ofthe cells remained closely similar through serial transplanta-tions. These results suggest that only ALDHbr cells, but notALDHlo cells, have the ability to reinitiate tumors and recon-stitute the heterogenicity, reflecting attributes of self-renewaland multipotency.

Further evidence that ALDHbr cells can recapitulate tumordiversity was evident in the immunohistochemistry assay onsections of xenografted tumors. Our results demonstrate thatALDH1-positive cells distributed sparsely over sections oftumors developed by injection of pure ALDHbr cells, in whichALDHlo cells appeared concurrently, indicates a developmentalhierarchy was formed in the xenografted human osteosarcoma.

Analysis of stem cell marker genes showed that thetumorigenic population of ALDHbr cells preferentiallyexpressed stem cell markers Oct3/4A, Nanog and Sox-2. Ithas been suggested that self-renewal and pluripotency of un-differentiated embryonic stem cells are maintained by acrosstalk of these transcription factors.33 Oct3/4A, Nanogand Sox-2 have also been implicated in tumorigenesis,

Figure 5. The highly tumorigenic osteosarcoma cells display CSC properties. (a–c) Phenotypic diversity in tumors arising from ALDHbr cells,

confirmed by flow cytometric analysis of cells with high ALDH and by immunostaining of ALDH expression (d). (a–c) Tumorigenic ALDHbr

cells from OS99-1 xenografts were then isolated and re-injected into NOD/SCID mice. Percentages of ALDHbr cells in 3 rounds of passages

were similar to that observed in the tumor from which they were derived. Insets in a–c show negative controls when cells are treated with

the ALDH-inhibitor DEAB. Tumorigenic cells formed tumors that contained phenotypically diverse cells similar to those seen in the original

xenografted tumor. (d) Expression of ALDH in tumors derived from ALDHbr cells. Tissues were examined for the presence of ALDH in the 3

rounds of tumor passages. Antibody localization was done using horseradish peroxidase, with brown staining indicating the presence of

ALDH antigen. Bar ¼ 50 lm. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Can

cerCellBiology

Wang et al. 301


primarily in germ cell tumors.33 However, they are alsoexpressed in several other types of human cancers, includingbreast cancer and bladder cancer.40 The finding that Oct3/4A,Nanog and Sox-2 are differentially expressed in the tumori-genic population suggests a potential functional role for thesetranscriptions in human osteosarcoma, a possibility thatremains to be investigated.

The capacity for tumor formation between the ALDHbr

and ALDHlo populations sorted directly from the cell linedid not reach statistical significance, unlike cells sorted fromxenografted tumors. It is not clear why the low percentage ofALDHbr cells from the OS99-1 xenografts was more tumori-genic than the high percentage of ALDHbr cells sorteddirectly from the OS99-1 cell line. Further investigationsincluding genetic analysis and microarray analysis are war-ranted to define the cause of the difference.

The finding that osteosarcoma-initiating cells can be identi-fied by high ALDH activity offers several potential advantagesover other methods. First, ALDEFLUOR assay has been devel-oped for the assessment of ALDH activity in viable cells by flowcytometry and has been made commercially available in a kitformat. Second, the detection of ALDH activity is a simpler andmore reliable method than surface marker selection-based meth-ods that require examination of clusters of several surfacemarkers. The identification of a specific surface marker typicallyearns certain credits to target cells of interest, but it is also truethat sometimes the surface phenotype of stem cells may remainstable despite a decline in functional activity as observed in theprogenitors, so that the existence of qualified stem cells maytherefore be overestimated.41 Third, ALDH substrate will notenter the nucleus or bind DNA, and as a consequence, this assaytends to be safer and less toxic compared to other methods that

require UV excitation or stain cell dyes that bind to nucleic acidssuch as in SP analysis.42 Finally, since only viable cells with func-tional enzymatic activity and an intact membrane can retain theALDH substrate, apoptotic and necrotic cells with leaky mem-branes will not be counted in the assessment. The identificationof CSCs based on high ALDH activity represents a very usefulmethod for detection and isolation of CSCs.

Of note, however, high ALDH activity might not be as auniversal marker for stem cells, so ALDEFLUOR assay is lim-ited for identifying cells with the properties expected ofCSCs. It has been demonstrated isolated ALDHbr cellsshowed lower ability to proliferate and migrate as comparedto ALDHlo cells in adipose tissue and endothelial tissue.43,44

Thus, it is critical to carefully characterize the cells isolatedbased on high ALDH activity.

Although our data show that a subpopulation of osteosar-coma cells exists within the OS99-1 cell line with markedlyenhanced tumorigenic potential and stem cell properties,there are limitations to our study. Specifically, an establishedcell line and not primary tumor was investigated. Furtherresearch using primary tumor is therefore necessary to con-firm the findings of this study.

In conclusion, this study provides further support to theCSC hypothesis, with evidence that a subpopulation of cellswith high ALDH activity residing in human osteosarcomacell line OS99-1 is able to initiate and mediate new tumor.

AcknowledgementsThe authors would also like to thank Dr. Sheila M. Nielsen-Preiss fromMontana State University for the generous gift of human osteosarcoma cellline OS99-1. The authors also thank Mrs. Holly Wagner for assistance in thepreparation of the manuscript andMr. Martin J. White for flow cytometry.

References

1. Reya T, Morrison SJ, Clarke MF,Weissman IL. Stem cells, cancer, andcancer stem cells. Nature 2001;414:105–11.

2. Pardal R, Clarke MF, Morrison SJ.Applying the principles of stem-cellbiology to cancer. Nat Rev Cancer 2003;3:895–902.

3. Dean M, Fojo T, Bates S. Tumour stemcells and drug resistance. Nat Rev Cancer2005;5:275–84.

4. Passegue E, Jamieson CH, Ailles LE,Weissman IL. Normal and leukemichematopoiesis: are leukemias a stem celldisorder or a reacquisition of stem cellcharacteristics? Proc Natl Acad Sci USA2003;100 Suppl 1:11842–9.

5. Clarke MF, Dick JE, Dirks PB, Eaves CJ,Jamieson CH, Jones DL, Visvader J,Weissman IL, Wahl GM. Cancer stem cells—perspectives on current status and futuredirections: AACRWorkshop on cancerstem cells. Cancer Res 2006;66:9339–44.

6. Klein MJ, Siegal GP. Osteosarcoma:anatomic and histologic variants. Am JClin Pathol 2006;125:555–81.

7. Gibbs CP, Kukekov VG, Reith JD,Tchigrinova O, Suslov ON, Scott EW,Ghivizzani SC, Ignatova TN, Steindler DA.Stem-like cells in bone sarcomas:implications for tumorigenesis. Neoplasia2005;7:967–76.

8. Wu C, Alman BA. Side population cells inhuman cancers. Cancer Lett 2008;268:1–9.

9. Tirino V, Desiderio V, d’Aquino R, DeFrancesco F, Pirozzi G, Graziano A,Galderisi U, Cavaliere C, De Rosa A,Papaccio G, Giordano A. Detection andcharacterization of CD133þ cancer stemcells in human solid tumours. PLoS One2008;3:e3469.

10. Wang L, Park P, Lin CY. Characterizationof stem cell attributes in humanosteosarcoma cell lines. Cancer Biol Ther2009;8:543–552.

11. Levings PP, McGarry SV, Currie TP,Nickerson DM, McClellan S, Ghivizzani

SC, Steindler DA, Gibbs CP. Expression ofan exogenous human Oct-4 promoteridentifies tumor-initiating cells inosteosarcoma. Cancer Res 2009;69:5648–55.

12. Vasiliou V, Pappa A, Petersen DR. Role ofaldehyde dehydrogenases in endogenousand xenobiotic metabolism. Chem BiolInteract 2000;129:1–19.

13. Duester G. Families of retinoiddehydrogenases regulating vitamin Afunction: production of visual pigment andretinoic acid. Eur J Biochem 2000;267:4315–24.

14. Armstrong L, Stojkovic M, Dimmick I,Ahmad S, Stojkovic P, Hole N, Lako M.Phenotypic characterization of murineprimitive hematopoietic progenitor cellsisolated on basis of aldehydedehydrogenase activity. Stem Cells 2004;22:1142–51.

15. Corti S, Locatelli F, Papadimitriou D,Donadoni C, Salani S, Del Bo R, StrazzerS, Bresolin N, Comi GP. Identification of aprimitive brain-derived neural stem cell

Can

cerCellBiology



population based on aldehydedehydrogenase activity. Stem Cells 2006;24:975–85.

16. Hess DA, Wirthlin L, Craft TP, HerrbrichPE, Hohm SA, Lahey R, Eades WC, CreerMH, Nolta JA. Selection based on CD133and high aldehyde dehydrogenase activityisolates long-term reconstituting humanhematopoietic stem cells. Blood 2006;107:2162–9.

17. Ginestier C, Hur MH, Charafe-Jauffret E,Monville F, Dutcher J, Brown M,Jacquemier J, Viens P, Kleer CG, Liu S,Schott A, Hayes D, et al. ALDH1 is amarker of normal and malignant humanmammary stem cells and a predictor ofpoor clinical outcome. Cell Stem Cell 2007;1:555–67.

18. Ma S, Chan KW, Lee TK, Tang KH, WoJY, Zheng BJ, Guan XY. Aldehydedehydrogenase discriminates the CD133liver cancer stem cell populations. MolCancer Res 2008;6:1146–53.

19. Huang EH, Hynes MJ, Zhang T, GinestierC, Dontu G, Appelman H, Fields JZ,Wicha MS, Boman BM. Aldehydedehydrogenase 1 is a marker for normaland malignant human colonic stem cells(SC) and tracks SC overpopulation duringcolon tumorigenesis. Cancer Res 2009;69:3382–9.

20. Cheung AM, Wan TS, Leung JC, Chan LY,Huang H, Kwong YL, Liang R, Leung AY.Aldehyde dehydrogenase activity inleukemic blasts defines a subgroup of acutemyeloid leukemia with adverse prognosisand superior NOD/SCID engraftingpotential. Leukemia 2007;21:1423–30.

21. Gillette JM, Gibbs CP, Nielsen-Preiss SM.Establishment and characterization of OS99–1, a cell line derived from a highlyaggressive primary human osteosarcoma.In Vitro Cell Dev Biol Anim 2008;44:87–95.

22. Moreb JS, Zucali JR, Ostmark B, BensonNA. Heterogeneity of aldehydedehydrogenase expression in lung cancercell lines is revealed by Aldefluor flowcytometry-based assay. Cytometry B: ClinCytom 2007;72:281–9.

23. Quintana E, Shackleton M, Sabel MS,Fullen DR, Johnson TM, Morrison SJ.Efficient tumour formation by singlehuman melanoma cells. Nature 2008;456:593–8.

24. Chen YC, Hsu HS, Chen YW, Tsai TH,How CK, Wang CY, Hung SC, Chang YL,Tsai ML, Lee YY, Ku HH, Chiou SH. Oct-4 expression maintained cancer stem-likeproperties in lung cancer-derived CD133-positive cells. PLoS One 2008;3:e2637.

25. Chu P, Clanton DJ, Snipas TS, Lee J,Mitchell E, Nguyen ML, Hare E, Peach RJ.Characterization of a subpopulation ofcolon cancer cells with stem cell-likeproperties. Int J Cancer 2009;124:1312–21.

26. Hahn SA, Seymour AB, Hoque AT,Schutte M, da Costa LT, Redston MS,Caldas C, Weinstein CL, Fischer A, YeoCJ, et al. Allelotype of pancreaticadenocarcinoma using xenograftenrichment. Cancer Res 1995;55:4670–5.

27. Al-Hajj M, Wicha MS, Benito-HernandezA, Morrison SJ, Clarke MF. Prospectiveidentification of tumorigenic breast cancercells. Proc Natl Acad Sci USA 2003;100:3983–8.

28. Hemmati HD, Nakano I, Lazareff JA,Masterman-Smith M, Geschwind DH,Bronner-Fraser M, Kornblum HI.Cancerous stem cells can arise frompediatric brain tumors. Proc Natl Acad SciUSA 2003;100:15178–83.

29. Lapidot T, Sirard C, Vormoor J, MurdochB, Hoang T, Caceres-Cortes J, Minden M,Paterson B, Caligiuri MA, Dick JE. A cellinitiating human acute myeloid leukaemiaafter transplantation into SCID mice.Nature 1994;367:645–8.

30. Li C, Heidt DG, Dalerba P, Burant CF,Zhang L, Adsay V, Wicha M, Clarke MF,Simeone DM. Identification of pancreaticcancer stem cells. Cancer Res 2007;67:1030–7.

31. O’Brien CA, Pollett A, Gallinger S, DickJE. A human colon cancer cell capable ofinitiating tumour growth inimmunodeficient mice. Nature 2007;445:106–10.

32. Schatton T, Murphy GF, Frank NY,Yamaura K, Waaga-Gasser AM, Gasser M,Zhan Q, Jordan S, Duncan LM, WeishauptC, Fuhlbrigge RC, Kupper TS, et al.Identification of cells initiating humanmelanomas. Nature 2008;451:345–9.

33. Santagata S, Ligon KL, Hornick JL.Embryonic stem cell transcription factorsignatures in the diagnosis of primary andmetastatic germ cell tumors. Am J SurgPathol 2007;31:836–45.

34. Ma S, Chan KW, Hu L, Lee TK, Wo JY,Ng IO, Zheng BJ, Guan XY. Identificationand characterization of tumorigenic livercancer stem/progenitor cells.Gastroenterology 2007;132:2542–56.

35. Kimura K, Nakano T, Park YB, Tani M,Tsuda H, Beppu Y, Moriya H, Yokota J.Establishment of human osteosarcoma celllines with high metastatic potential tolungs and their utilities for therapeutic

studies on metastatic osteosarcoma. ClinExp Metast 2002;19:477–85.

36. Luo XJ, Chen J, Song WX, Tang N, LuoJY, Deng ZL, Sharff KA, He G, Bi Y, HeBC, Bennett E, Huang JY, et al. OsteogenicBMPs promote tumor growth ofhuman osteosarcomas that harbordifferentiation defects. Lab Invest 2008;88:1264–77.

37. Rodan SB, Imai Y, Thiede MA,Wesolowski G, Thompson D, Barshavit Z,Shull S, Mann K, Rodan GA.Characterization of a human osteosarcomacell-line (Saos-2) with osteoblasticproperties. Cancer Res 1987;47:4961–6.

38. Ucar D, Cogle CR, Zucali JR, Ostmark B,Scott EW, Zori R, Gray BA, Moreb JS.Aldehyde dehydrogenase activity as afunctional marker for lung cancer. ChemBiol Interact 2009;178:48–55.

39. Lee J, Kotliarova S, Kotliarov Y, Li A, SuQ, Donin NM, Pastorino S, Purow BW,Christopher N, Zhang W, Park JK, FineHA. Tumor stem cells derived fromglioblastomas cultured in bFGF and EGFmore closely mirror the phenotype andgenotype of primary tumors than doserum-cultured cell lines. Cancer Cell 2006;9:391–403.

40. Ben-Porath I, Thomson MW, Carey VJ, Ge R,Bell GW, Regev A, Weinberg RA. Anembryonic stem cell-like gene expressionsignature in poorly differentiated aggressivehuman tumors. Nat Genet 2008;40:499–507.

41. Spangrude GJ, Brooks DM, Tumas DB.Long-term repopulation of irradiated micewith limiting numbers of purifiedhematopoietic stem cells: in vivo expansionof stem cell phenotype but not function.Blood 1995;85:1006–16.

42. Pearce DJ, Taussig D, Simpson C, Allen K,Rohatiner AZ, Lister TA, Bonnet D.Characterization of cells with a highaldehyde dehydrogenase activity from cordblood and acute myeloid leukemia samples.Stem Cells 2005;23:752–60.

43. Estes RT, Wu AW, Storms RW, Guilak F.Extended passaging, but not aldehydedehydrogenase activity, increases thechondrogenic potential of human adipose-derived adult stem cells. J Cell Physiol2006;209:987–95.

44. Nagano M, Yamashita T, Hamada H,Ohneda K, Kimura KI, Nakagawa T,Shibuya M, Yoshikawa H, Ohneda O.Identification of functional endothelialprogenitor cells suitable for thetreatment of ischemic tissue using humanumbilical cord blood. Blood 2007;110:151–60.

Can

cerCellBiology

Wang et al. 303


Documents

Prospective identification of tumorigenic osteosarcoma cancer stem cells in OS99-1 cells based on high aldehyde dehydrogenase activity