Article

Tumor stem cells derived = from=20 glioblastomas cultured in bFGF and EGF more closely mirror the = phenotype=20 and genotype of primary tumors than do serum-cultured cell=20 lines

Jeongwu Lee,^¹=20Svetlana Kotliar= ova,^¹,³=20Yuri Kotliarov,<= SPAN=20 class=3Dja50-ce-sup>^¹,³=20Aiguo Li,^¹,³=20Qin Su,^¹=20Nicholas=20 M. Donin,^¹=20Sandra Pastorino= ,^¹=20Benjamin=20 W. Purow,^¹=20Neil Christopher= ,^¹=20Wei Zhang,^¹=20John = K. Park,^²=20and Howard = A. Fine^¹,

1 Neuro-Oncology=20 Branch, National Cancer Institute, National Institute of = Neurological=20 Diseases and Stroke, National Institutes of Health, Bethesda, = Maryland=20 20892
2 Surgical=20 Neurology Branch, National Institute of Neurological Diseases and = Stroke,=20 National Institutes of Health, Bethesda, Maryland 20892

Correspondence:
Howard A. Fine
Ph: 301 402 = 6383;=20 Fax: 301 480 2246
hfine@mail.nih.gov

Article Options
Table=20 of Contents Summary PDF=20 (1404K) Suppleme= ntal=20 Data
Search Medline = for
Jeongw= u=20 Lee Howa= rd=20 A. Fine
Related = Articles
Radial=20 glia cells are candidate stem cells of epen... Stem=20 Cells and Cancer... Early=20 inactivation of p53 tumor suppressor gene = co... More=20 related=20 articles...

Summary

The concept of tumor stem cells = (TSCs)=20 provides a new paradigm for understanding tumor biology, although = it=20 remains unclear whether TSCs will prove to be a more robust model = than=20 traditional cancer cell lines. We demonstrate marked phenotypic = and=20 genotypic differences between primary human tumor-derived TSCs and = their=20 matched glioma cell lines. Unlike the matched, traditionally grown = tumor=20 cell lines, TSCs derived directly from primary glioblastomas = harbor=20 extensive similarities to normal neural stem cells and = recapitulate the=20 genotype, gene expression patterns, and in vivo biology of human=20 glioblastomas. These findings suggest that TSCs may be a more = reliable=20 model than many commonly utilized cancer cell lines for = understanding the=20 biology of primary human tumors.

SIGNIFICANCE

We developed=20 and characterized a model system in which primary human = glioblastoma cells=20 are propagated in vitro under conditions optimal for either normal = NSCs or=20 typical glioma cell lines. We demonstrated that TSCs closely mimic = the=20 genotype, gene expression profile, and biology of their parental = primary=20 tumors. By contrast, tumor cells grown under standard = serum-containing=20 culture conditions result in the loss of TSCs and ultimately lead = to the=20 outgrowth of cells that are vastly different both genetically and=20 biologically from their parental tumors. These observations bring = into=20 question the relevance of standard cancer cell lines for studying = the=20 biology of human cancer and for screening new therapeutic agents. = TSCs may=20 prove to be a more reliable model system.

Introduction

Cancer cell=20 lines have been the historical standard both for exploring the = biology of=20 human tumors and as preclinical models for screening potential = therapeutic=20 agents. It has become increasingly clear, however, that phenotypic = characteristics and the multitude of genetic aberrations found = within=20 repeatedly in vitro passaged cancer cell lines often bear little=20 resemblance to those found within the corresponding primary human = tumor=20 (Deschavanne et al., 1996; Hecht et al., 1995; Pardal et al., 2003; Romer et al., 2005). This not only in part explains why both in vitro = and in=20 vivo cancer cell line-based preclinical therapeutic screening = models have=20 been poorly predictive for identifying clinically useful = therapeutic=20 agents, but may also have led to some important misinterpretations = regarding the relevance of aberrant signaling pathways within cell = lines=20 compared to primary tumors. This realization led us on a search = for a more=20 biologically relevant model system for exploring glioma biology = and for=20 the screening of new therapeutic agents.

Neural stem=20 cells (NSCs) and in situ glioma cells share the characteristics of = continuous self-renewal, extensive brain parenchymal=20 migration/infiltration, and potential for full or partial=20 differentiation=E2=80=94properties not found in glioma cell lines = (Morshead et=20 al., 2004; Sanai et al., 2005). These observations suggest that at least some in = situ=20 glioma cells may harbor features consistent with tumor stem cells = (TSCs)=20 (Pardal et al., 2003; Phillips et al., 2006; Reya et al., 2001; Sanai et al., 2005). Brain tumor stem-like cells have been recently = reported;=20 however, besides the morphological similarities and = differentiation=20 capacity of these cells, little else is understood regarding the = actual=20 similarities between glioma TSCs, normal human NSCs, and the = primary=20 tumors (Galli et al., 2004; Hemmati et al., 2003; Singh et al., 2003; Singh et al., 2004; Taylor et al., 2005; Yuan et al., 2004; Zhu et al., 2005). We therefore undertook a series of studies to = readily=20 identify and better characterize glioma TSCs in an attempt to = understand=20 the relationship between these TSCs, the primary tumor, and = traditional=20 glioma cell lines.

One = obvious=20 difference in the in vitro propagation of NSCs compared to the = commonly=20 used glioma cell lines is the requirement for serum. All = established=20 glioma cell lines (like most other cancer cell lines) are grown in = media=20 containing serum, whereas NSCs are grown in serum-free media, = since serum=20 causes irreversible differentiation of NSCs (Gage et al., 1995; McKay, 1997; Reynolds et al., 1992). To explore how growth under these two different = in vitro=20 conditions affects primary tumor cells, we prepared single cell=20 suspensions of freshly resected and dissociated glioblastoma (GBM) = tissues=20 from patients. These cells were cultured under conditions optimal = for=20 propagation and nondifferentiation of normal NSCs = (=E2=80=9CNBE=E2=80=9D conditions:=20 serum-free Neurobasal media supplemented with basic = FGF and=20 EGF), or conditions optimal for growth of glioma and most = other=20 cancer cell lines (=E2=80=9Cserum=E2=80=9D conditions: DMEM media = containing 10% fetal=20 bovine serum). We have established a number of GBM-derived = NBE- and=20 serum-cultured cells from this protocol, two of which (lines 0308 = and=20 1228) are described in detail here.

Results

In vitro characterization of GBM = cells=20 cultured in NBE and serum conditions

Glioma cells=20 cultured in NBE and serum conditions displayed profound biological = differences in vitro (Figure 1). First, cells cultured = in NBE=20 conditions (0308-NBE and 1228-NBE) readily proliferated both as=20 nonadherent, multicellular spheres in uncoated plates = (=E2=80=9Cneurospheres=E2=80=9D) and=20 as an adherent monolayer in poly-L-lysine/laminin-coated plates, = as is=20 seen with normal NSCs. In contrast to the NBE cells, cells = cultured in=20 serum conditions (0308-serum and 1228-serum) formed a monolayer = that=20 initially had a highly heterogeneous morphology but within a month = became=20 homogeneous with a morphology reminiscent of fibroblasts or = epithelial=20 cells (Figure 1A). Although this dramatic = morphological change of serum-cultured primary glioma cells has = been=20 routinely observed, it has never been understood. Second, cells = cultured=20 in NBE proliferated at a constant rate regardless of passage = number,=20 whereas cells cultured in serum showed initial growth followed by = a=20 plateau phase, only to eventually proliferate at a much greater = rate in=20 later passages (Figure 1B). To determine whether = the=20 differences in the growth kinetics between 0308- and 1228-serum = cells were=20 secondary to the different growth conditions or whether the = differences=20 were secondary to the inadvertent in vitro selection of a clonal=20 population of cells, we evaluated the proliferative kinetics of = cells=20 derived from multiple different single clones. The cells were = initially=20 derived from either NBE or serum conditions and then transferred = to both=20 NBE and serum conditions (Figure 1C and 1D). Upon the = addition of serum,=20 NBE cells showed a decreased rate of proliferation initially, = followed by=20 late exponential growth, recapitulating the growth pattern seen in = the=20 polyclonal serum cell population (Figure 1C). By contrast, cellular=20 proliferation of serum cells nearly ceased following transfer to = NBE=20 conditions (Figure 1D). Thus, it appears that = the initial=20 culture of primary tumor cells under serum conditions leads to = changes=20 that cannot be subsequently reversed following transition to NBE=20 conditions.

Figure 1. In vitro=20 characterization of tumor cells cultured in NBE and serum=20 conditions

A: = Microphotographs of=20 NBE-cultured glioblastoma (GBM) cells at early passage (Aa) and late = passage (Ab) and = serum-cultured GBM=20 cells at early passage (Ac) and late = passage (Ad). Note that = the late=20 passage serum cells are morphologically similar to U251 = cells, a=20 common glioma cell line (Ae).

B: Proliferation = kinetics=20 of NBE-cultured and serum-cultured GBM cells. 0308-NBE cells = (blue=20 lines) and 1228-NBE cells (green lines) at all passages, but = only=20 late passage 0308-serum cells (red lines with filled = triangle and=20 filled circles) and late passage 1228-serum cells (pink = lines with=20 open rectangles and open triangles) proliferated at constant = rate.=20 By contrast, early passage 0308-serum cells (red line with = diamond=20 markers) and early passage 1228-serum cells (pink line with = X=20 markers) proliferated very slowly initially followed by a = plateau=20 phase, before their exponential growth phase.

C and D: = Proliferation=20 kinetics of clonally derived cells grown under NBE condition = (C) and serum = condition (D). Monoclonal = populations=20 of 0308 (triangle, n =3D 3), 1228 (rectangle, n =3D 3), and = polyclonal=20 normal human NSCs (circle, n =3D 3) initially expanded in = NBE=20 conditions (C) = or serum=20 conditions (D) = were=20 further cultured in NBE and serum conditions for a month, = and cell=20 numbers were counted. Error bars represent standard=20 deviation.

View larger version: [In=20 this window] [In new = window]

A = third way in=20 which NBE- versus serum-cultured cells differed was that = NBE-cultured=20 cells but not serum-cultured cells had the potential for = multilineage=20 differentiation and clonogenicity, central features of NSCs (Figure 2). Approximately 90% of = NBE cells were=20 positive by immunohistochemical analysis for the NSC markers = Nestin, Sox2,=20 and SSEA-1 (Brazel et al., 2005; Capela et al., 2002; Lendahl et al., 1990; Park et al., 2003). Western blot analysis confirmed abundant = expression of=20 these NSC markers in NBE cells (Figure 2B) (Brazel et al., 2005; Park et al., 2003; Sakakibara et al., 1996). After removal of growth factors (bFGF and EGF), = or the=20 addition of retinoic acid (RA) and serum, NBE cells began to lose = their=20 NSC markers and developed morphologies and immunohistochemical = staining=20 patterns consistent with cells of glial and neuronal lineages. = Unlike=20 normal NSCs, however, about 40% of cells costained for both glial = (GFAP)=20 and neuronal (TuJ1) markers, suggesting that differentiation = pathways in=20 these NBE cells are not entirely normal (Galli et al., 2004; Hemmati et al., 2003). By contrast, 0308- and 1228-serum cells = expressed few or=20 no NSC markers (Figure 2A, 2B, and 2D). Cells were = only weakly=20 positive for GFAP and TuJ1 and did not respond to differentiation = cues=20 such as RA and CNTF (Figure 2D and data not shown). = Next, we=20 performed single cell neurosphere formation assays in order to = determine=20 the clonogenicity of NBE- and serum-cultured cells. 0308-NBE and = 1228-NBE=20 cells had clonogenic frequencies of 23.4% (=C2=B12.4%) and 17.4% = (=C2=B13.8%),=20 respectively, reminiscent of NSC-derived neurospheres, whereas = 0308- and=20 1228-serum cells failed to form neurosphere-like cells when plated = in NBE=20 conditions. Finally, a fourth way that NBE- and serum-cultured = glioma=20 cells differed was in their telomerase activities, although both = cells=20 retained telomeres as determined by fluorescence in situ = hybridization=20 (FISH) using telomere-specific probes (Figure 2E). 0308-NBE and 1228-NBE = cells had=20 consistent telomerase activity regardless of passage number, = whereas=20 telomerase activity was lost when these cells were cultured in=20 serum-containing media, consistent with what occurs in normal NSCs = (Haik=20 et al., 2000; Ostenfeld et al., 2000). Likewise, early passage 0308-serum and = 1228-serum cells=20 did not have appreciable telomerase activity. Interestingly, = however,=20 serum-cultured cells regained telomerase activity in later = passages,=20 coincident with their late passage exponential growth phase (Yang = et al.,=20 2004) (Figure 2F). Taken together, these = data=20 demonstrate that both 0308- and 1228-NBE cells contain many of the = self-renewal and differentiation characteristics of NSCs, whereas=20 serum-cultured cells do not.

Figure 2. Evaluation = of stem=20 cell-like characteristics of GBM cells cultured in NBE and = serum=20 conditions

A: = Immunohistochemical=20 analysis of NBE and serum GBM cells (passage 10). The = majority of=20 NBE cells cultured under NBE conditions are immunopositive = for=20 Nestin, SSEA-1, and Sox2 (Aa and Ac). When these = cells were=20 cultured in serum condition for 10 days, they lost Nestin = expression=20 and became positive for GFAP and TuJ1 (Ab). About 40% of = these=20 cells were positive for both TuJ1 and GFAP (shown as yellow = in Ab). Serum cells = do not=20 express either Nestin or Sox2 (Ad). DAPI = staining (shown=20 as blue) was performed in Ad to identify = cells.

B: Western blot = analysis of=20 various NSC markers in NBE and serum GBM cells. All cell = lysates=20 were obtained from cells at passage 15. =CE=B2-actin blot = was used as a=20 loading control.

C and D: = Differentiation=20 potential of NBE- and serum-cultured GBM cells in various=20 conditions. Cells (C,=20 0308-NBE cells p15; D,=20 0308-serum cells p15) were cultured in the indicated = conditions for=20 10 days and processed for immunohistochemistry, and then=20 immunopositive cells were counted. N2 condition indicates = the=20 withdrawal of bFGF and EGF from NBE media. RA condition = indicates=20 the addition of all-trans-retinoic = acid (2 =CE=BCM)=20 in N2 media. Double-positive cells for two antigens were = counted=20 positive for each antigen and also represented separately = (e.g.,=20 TuJ1+/GFAP+; double-positive cells for both antigens). Error = bars=20 represent standard deviation.

E: Preserved = telomeres in=20 0308-NBE and 0308-serum cells (passage 15) by = telomere-PNA-FISH=20 analysis; telomeres were labeled as green dots.

F: Quantitative = ELISA for=20 measuring telomerase activities in various cells as = determined by=20 the telomeric repeat amplification protocol assay (Roche). y = axis=20 represents absorbance at 450 nm, which is proportional to = the=20 telomerase activity. When NSCs and NBE cells were cultured = in serum=20 conditions for 10 days (marked as diff.), their telomerase=20 activities decreased to very low levels. The experiments = were=20 performed in triplicates, and error bars represent standard=20 deviation.

View larger version: [In=20 this window] [In new = window]

Tumorigenic potential of GBM cells = cultured in=20 NBE or serum conditions

In = order to=20 evaluate the tumorigenic potential of the NBE cells compared to = the serum=20 cells, we first determined their ability to grow in soft agar (Table S1 in the Supplemental Data available with = this article=20 online). Cells were seeded into two different soft agar layers, = one=20 prepared with NBE media and the other with serum media, and then = colonies=20 were counted 3 weeks later. Both early and late passage NBE cells=20 consistently formed colonies in soft agar without serum, but not = in the=20 agar with serum, suggesting that the clonal expansion of these = cells=20 requires lack of differentiation. Conversely, early passage serum = cells=20 failed to form colonies under either soft agar condition. Late = passage,=20 exponentially growing serum cells, however, readily formed large = colonies=20 in serum-containing agar but not in NBE-agar.

To = test in vivo=20 tumorigenic potential of NBE and serum cells at different = passages, we=20 stereotactically injected the cells into the brains of neonatal=20 immunodeficient SCID mice. Two to three months later, nearly all = mice=20 (16/17) injected with 0308-NBE cells developed tumors, whereas = none of the=20 13 mice injected with 0308-serum cells developed tumors (Table S2). 0308-NBE and 1228-NBE = cells retain=20 a consistent tumorigenic potential independent of passage number = (as low=20 as 1000 cells), whereas serum-cultured cells at early passages are = nontumorigenic. When established NBE cell-derived intracranial = xenograft=20 tumors were dissociated, recultured under NBE conditions, and then = injected back into the brains of new recipient mice, there was no = loss of=20 tumorigenic potential (Table S3). By contrast, when the = same=20 xenograft-derived cells were grown under serum conditions, all = subsequent=20 tumorigenic potential was lost, suggesting that the loss of = tumorigenicity=20 of these glioma cells is due to the serum-culture condition.=20 Interestingly, although early passage serum-cultured cells did not = form=20 tumors, the late passage, exponentially growing, = telomerase-positive=20 0308-serum cells began to produce intracranial tumors at an = increasing=20 rate with progressive passage number. By contrast, no tumors were=20 generated from passage ten 1228-serum cells 6 months after=20 injection.

Phenotypic characterization of = tumors derived=20 from NBE- and late passage serum-cultured GBM cells

There were=20 significant histopathological differences between the intracranial = tumors=20 derived from NBE-cultured cells and those derived from late = passage=20 serum-cultured cells (Figure 3). Intracranial tumors = generated by=20 0308-NBE and 1228-NBE cells demonstrated extensive infiltration = into the=20 surrounding cerebral cortex, a pathognomonic feature of human GBMs = (Galli=20 et al., 2004; Maher et al., 2001; Sanai et al., 2005; Singh et al., 2004). NBE GBM cells showed a predilection for = migrating along=20 white matter tracts such as the corpus callosum, as is = characteristically=20 seen with GBM cells in the brains of patients (Figure 3A, 3D, and 3E). = Additionally, a=20 significant number of cells were seen migrating toward the = olfactory bulb=20 (Figure 3F). By stark contrast, all = of the=20 tumors generated from late passage 0308-serum cells were well = delineated=20 with little tumor cell infiltration into the surrounding normal = brain, a=20 characteristic phenotypically identical to the human tumor = xenografts=20 generated from the standard glioma cell lines (e.g., U251, U87MG = cells)=20 (Figure 3B and 3C). These data = demonstrate that=20 tumors derived from NBE cells, but not serum cells, phenocopy the=20 critically important histopathological features of the original = human GBM=20 tumor.

Figure 3. Phenotypic=20 characterization of tumors derived from NBE- and late = passage=20 serum-cultured GBM cells

A=E2=80=93C: = Representative=20 microphotographs of NBE-cultured cell-derived intracranial = xenograft=20 tumor (0308-NBE p20) (A), = serum-cultured=20 cell-derived tumor (0308-serum p35) (B), and tumor = generated=20 from U87MG glioma cell lines (C). Note the = extensive=20 infiltration into the surrounding cerebral cortex in A only. Tu = indicates=20 central tumor mass.

D=E2=80=93F: = Infiltrative/migratory=20 nature of xenograft tumors derived from NBE-cultured GBM = cells.=20 Numerous NBE GBM cells (stained by brown color) diffusely=20 infiltrated adjacent cortex (D) and migrated = along the=20 corpus callosum (CC, blue-stained by luxol fast blue dye) = (E) and olfactory = bulb (F). Inset in F is a = hematoxylin/eosin=20 (H/E)-stained section depicting normal olfactory bulbs. LV = in D indicates = lateral=20 ventricle.

G=E2=80=93I: = Immunohistochemical=20 staining of NBE cell-derived tumors by using antibodies for = Nestin=20 (dark brown in G), Sox2=20 (red/purple in H), and=20 GFAP (red/purple in I).=20 Scale bar, 100 =CE=BCm.

View larger version: [In=20 this window] [In new = window]

Transcriptome analysis of NBE- and=20 serum-cultured GBM cells

To = further=20 understand the differences between NBE- and serum-cultured GBM = cells, we=20 generated gene expression profiles of NBE cells, serum cells, = their=20 derived xenograft tumors, and the original GBMs taken directly = from the=20 patients (see experimental schema in Figure S1). NBE- and = serum-cultured GBM cells=20 from different passages were also included in the analyses in = order to=20 evaluate the potential effects of in vitro passage number on gene=20 expression.

Unsupervised=20 hierarchical cluster analysis of all in vitro and in vivo = xenograft=20 samples (n =3D 79) revealed two distinct clusters; one cluster = consisted of=20 NBE cells and NSCs, and the other consisted of serum cells and ten = commonly utilized glioma cell lines (with the one exception of = 1228-serum=20 passage 3 cells; see below) (Figure 4A and 4B). Gene expression = profiles of=20 xenograft tumors derived from both 0308- and 1228-NBE cells were = similar=20 to those of NBE cells and NSCs, but distinct from those of = serum-cultured=20 cells and their derivative tumors (0308-serum-p25-derived tumors), = suggesting that the distinct expression profiles of NBE-cultured = and=20 serum-cultured cells reflect intrinsic biological differences = between the=20 cells rather than some transient effect of serum on cellular gene=20 expression (i.e., serum-responsive genes) in the serum-cultured=20 cells.

Figure 4. Global = expression=20 profiling analysis of NBE- and serum-cultured GBM cells

A: Dendrogram of = cluster=20 determined by unsupervised hierarchical cluster analysis of = 79 in=20 vitro and in vivo xenograft samples. Pearson = correlation-based=20 algorithm was applied on the basis of similarity in the = expression=20 pattern over all genes (over 42,000 probe sets). All samples = represent independent biological samples. 1228_S_p8 = indicates the=20 cells at passage 8 (p8) cultured in serum condition (S), = derived=20 from the patient tumor 1228 (1228). IC and SC indicate = xenograft=20 tumors generated by intracranial and subcutaneous injection = of the=20 cells, respectively. Detailed descriptions about the samples = are=20 provided in the Supplemental Data (experimental schema = in Figure S1). Two main clusters were = marked by=20 different colors (blue and red lines). Blue and red bars on = the=20 upper row indicate in vitro samples cultured in NBE and = serum=20 conditions, respectively. Color bars on the lower row = indicate the=20 origin of the corresponding samples. 0308 and 1228 cells are = distinguished by two different tones of light blue. Purple = and=20 orange bars represent commonly used glioma cell lines and = NSCs,=20 respectively. The above labeling schemes are consistently = used in=20 subsequent figures.

B: Principal = component=20 analysis (PCA) of the previous data sets including the = parental=20 primary GBMs. Small and large balls indicate in vitro and in = vivo=20 xenograft samples, respectively. Colors of balls indicate = the origin=20 of samples (see above). Parental patient tumors are marked = as balls=20 with red circle. x, y, and z axes represent three major = principal=20 components (PC). Note two distinct clusters; one cluster = consists of=20 NBE cells and their derivative xenograft tumors, NSCs, and = parental=20 patient tumors, whereas the other cluster consists of serum = cells,=20 ten commonly used glioma cell lines, and their derivative = tumors.=20 Detailed descriptions about the samples are provided in Figure S3.

C: Dendrogram of = cluster=20 determined by unsupervised hierarchical cluster analysis of = in vivo=20 xenograft and primary GBM samples. Gray bars indicate = patient GBM=20 samples (except 0308 and 1228 patient tumors). Blue and red = lines=20 mark two main clusters, respectively.

View larger version: [In=20 this window] [In new = window]

Since=20 NBE-cultured GBM cells reproduced the histopathology of human GBMs = in=20 situ, we next evaluated whether the global expression profiles of = NBE=20 cell-derived tumors resembled those of primary human tumors (Figure 4C). To this end, we = compared the=20 expression profiles of xenograft tumors derived from NBE- and=20 serum-cultured GBM cells to those of 21 randomly chosen GBMs, = including=20 the parental 0308 and 1228 GBMs. Unsupervised hierarchical cluster = analysis revealed that both 0308- and 1228-NBE-derived tumors were = closely=20 related to their parental tumors and other primary GBMs (Figure 4C). Tumors derived from = late passage=20 0308-serum cells, however, had markedly different profiles from = their=20 matched NBE-derived tumor xenografts and from parental 0308 tumors = as well=20 as from other primary GBMs, clustering with xenograft tumors = derived from=20 the U87MG and U251 traditional glioma cell lines (Figure 4C and Figure S2).

Passage-dependent gene expression = profile=20 changes in NBE and serum GBM cells

Our = in vitro=20 and in vivo findings demonstrate that the phenotype of NBE cells = was well=20 maintained regardless of passage number, while that of serum cells = was=20 changed dramatically upon serial passages. Consistent with this, = principle=20 component analysis (PCA) indicated that early passage (p3) = 1228-serum=20 cells were intermediate between the NBE group and the serum group, = whereas=20 later passage 1228-serum cells (p8 and p13) were tightly clustered = with=20 all other traditional glioma cell lines (Figure 4A and 4B). Thus, gene = expression=20 profiles of serum-cultured cells progressively diverge away from = those of=20 primary GBMs and NSCs toward those that resemble established = glioma cell=20 lines with increasing passage (Figure 4B). In order to further = analyze gene=20 expression changes in NBE and serum cells with increasing passage = number,=20 we performed PCA and two-way ANOVA analysis (Figure S3). All 1228-NBE cells, = regardless of=20 passage, clustered tightly, and the numbers of genes = differentially=20 expressed between different passages of cells were small (Table S4). By contrast, = approximately 5000=20 probe sets were differentially regulated in early (p3) versus late = passage=20 1228-serum cells (p8 and p13), raising the possibility that these=20 significant gene expression changes in serially passaged serum = cells may,=20 in part, be responsible for the different phenotype of late = passage serum=20 cells and NBE cells as well as primary GBMs (i.e., does one = progressively=20 lose expression of GBM-defining genes with serial passage in = serum?). To=20 test this hypothesis, we performed a supervised hierarchical = cluster=20 analysis of the entire data sets by using 2029 probe sets that = were=20 expressed more than 2-fold in 1228-serum p3 cells compared to the = later=20 passages (p8 and p13) (Figure 5A). Expression levels of = almost all of=20 these genes were very high in primary GBMs, but not in the later = passage=20 0308- and 1228-serum cells or glioma cell lines. Interestingly, = about 75%=20 of these 2029 probe sets were also highly expressed in NSCs and = NBE cells=20 regardless of passage number. Functional annotation of this gene = set by=20 Ingenuity Pathway Knowledge Base predicted that many of these = genes are=20 involved in CNS development, CNS function, and cell-to-cell = signaling (Table S5). On the other hand, a = quarter of the=20 probe sets were represented only in primary GBMs but not in NBE or = serum=20 groups. These probe sets were mainly immune system-related genes,=20 consistent with the minor contamination of CD45-positive = lymphocytes in=20 early passage glioma cultures. Conversely, we identified 945 probe = sets=20 that were upregulated in later passage (p8 and p13) compared to = early=20 passage (p3) of 1228-serum cells. Supervised analysis using these = probe=20 sets indicated that these genes are also upregulated in 0308-serum = cells=20 and glioma cell lines, but not in NBE cells, NSCs, and primary = GBMs (data=20 not shown). Taken together, these data support the notion that=20 serum-cultured cells rapidly lose characteristics of the primary = GBMs upon=20 passage in vitro, whereas NBE-cultured GBM cells maintain their = parental=20 phenotypes.

Figure 5. Supervised=20 hierarchical analyses of NBE- and serum-cultured GBM = cells

A: Heatmap of = supervised=20 hierarchical cluster analysis using = =E2=80=9Cpassage-dependent gene sets=E2=80=9D as=20 classifiers. See text for the detail. Dendrogram of cluster = is shown=20 on the left of the heatmap. Expression level of a given gene = is=20 indicated by red (high) and green (low) in the heatmap. = Samples used=20 for generating =E2=80=9Cpassage-dependent gene sets=E2=80=9D = are marked by colored=20 boxes on the right.

B: Heatmap of = supervised=20 hierarchical cluster analysis of = =E2=80=9CNBE-specific=E2=80=9D and =E2=80=9Cserum-specific=E2=80=9D=20 probe sets. Samples used for generating these probe sets are = shown=20 in colored boxes on the right of the heatmap.

View larger version: [In=20 this window] [In new = window]

Analysis of NBE- and serum-specific = genes and=20 their expression in primary GBMs

We = next=20 evaluated the genes that were differentially expressed between = NBE- and=20 serum-cultured cells. As determined by two-way ANOVA (with cell = origin and=20 culture condition as the factors), 5795 and 6895 probe sets were=20 differentially expressed in 0308- and 1228-NBE cells compared to = 0308- and=20 1228-serum cells, respectively. Among them, 1864 probe sets were = commonly=20 upregulated in both 0308- and 1228-NBE cells, and 1444 probe sets = were=20 commonly upregulated in serum cells. Supervised hierarchical = cluster=20 analysis using these probe sets revealed that the 1864 = NBE-specific probe=20 sets were highly expressed in primary GBM tumors as well as NSCs, = but not=20 in glioma cell lines (Figure 5B). By contrast, most of = 1444=20 serum-specific probe sets were highly expressed in glioma cell = lines, but=20 not in GBM tumors. Functional annotation of the NBE-specific probe = sets=20 using Ingenuity and gene ontology analysis tool GOSTAT predicted = that a=20 majority of these genes are involved in CNS function and = development,=20 consistent with the previous analysis (Table S6 and data not shown). = Representative=20 gene lists of these probe sets are presented in Table S7. Notably, numerous stem=20 cell-associated genes were found in the NBE-specific sets, but not = in=20 serum-specific sets.

Based on high=20 expression of a number of stem cell-associated genes in the NBE = cells, we=20 performed a stratified analysis of all of the in vitro and in vivo = NBE and=20 serum cells, as well as the primary GBMs, using probe sets highly = enriched=20 for stem cell-related genes. The gene set was selected by first = finding=20 the genes that overlapped between three distinct murine = NSC-enriched gene=20 lists reported in three different studies, and then converting = these genes=20 (approximately 200) to their human orthologs (Fortunel et al., = 2003; Ivanova et al., 2002; Ramalho-Santos et al., 2002). Clustering data showed that these stem = cell-associated=20 genes are very similarly regulated in both NBE cells and NSCs as = well as=20 in primary GBMs in contrast to serum cells and glioma cell lines = (Figure S4 and Table S8).

Taken together,=20 these data demonstrate that NBE cells are remarkably similar to = normal=20 NSCs and that NBE cells and their derivative tumors properly = maintain many=20 biological characteristics of the parental GBMs and other primary = GBMs,=20 whereas traditionally grown, serum-cultured cancer cell lines do=20 not.

Genotypic characterization of NBE- = and=20 serum-cultured GBM cells

In = order to=20 evaluate genomic changes in NBE- and serum-cultured glioma cells = in=20 detail, we performed single nucleotide polymorphism (SNP) analysis = as well=20 as spectral karyotyping (SKY) and Giemsa banding analysis (Figure 6). We identified = homozygous deletion=20 of the INK4a/ARF locus at chromosome 9, loss of chromosome = 10q,=20 trisomy of chromosome 7, and local amplification of EGFR = locus in=20 both 0308 and 1228 surgical samples. These genetic alterations are = common=20 genomic features in a significant percentage of primary GBMs = (Vogelstein=20 et al., 2004; Wiltshire et al., 2000). The detailed genomic characteristics of the 0308 = and 1228=20 cells are summarized in Table S9.

Figure 6. Genomic = analysis of=20 NBE- and serum-cultured GBM cells

A: Heatmap = representing=20 loss-of-heterozygosity (LOH) areas in 0308-NBE and serum = cells at=20 different passages. LOH areas were determined by comparing = SNP=20 allelic calls of genomic DNA from tumor cells to those of = peripheral=20 blood DNA from the same patient. The calculation and = smoothing were=20 performed using dChip software. x axis represents = chromosomes in=20 numerical order. Chromosome X was omitted, since it is not=20 informative in LOH analysis. Black areas represent LOH. Blue = and red=20 bars on the right of the heatmap represent the cells = cultured in NBE=20 and serum conditions, respectively. For clonally derived = cells=20 (labeled with clone number), passage number was not = assigned. The=20 culture time difference between early and late clones is=20 approximately 1 month. Note that LOH of chromosomes 9 and 10 = is=20 evident in all of samples including the parental 0308 = patient tumor.=20 NBE-cultured cells maintain parental genotype regardless of = passage=20 and clones, whereas most of the late passage serum cells = revealed=20 LOH of chromosomes 4 and 17.

B: Genomic = amplification=20 (red) and deletion (green) areas in 0308-NBE and serum cells = at=20 different passages. Copy number values were calculated using = Affymetrix chromosome copy number analysis tool and smoothed = by=20 running average over 15 consecutive SNPs. Note the = amplification of=20 chromosomes 7 and 20 and deletion of chromosomes 9 and 10 in = parental tumors. Late passage serum cells revealed = significant=20 genomic alterations including deletion of chromosomes 4, 17, = and=20 22.

C=E2=80=93F: = Spectral karyotyping=20 (SKY) (C and = E) and Giemsa = banding=20 analysis (D = and F) revealed that = 0308-NBE=20 (passage 40) and 1228-NBE cells (passage 20) are near = diploid,=20 whereas 0308-serum cells (passage 15) and 1228-serum cells = (passage=20 20) are predominantly pseudotriploid or = tetraploid.

View larger version: [In=20 this window] [In new = window]

Given the=20 changes in growth rates, morphology, tumorigenicity, and gene = expression=20 over time in the serum-cultured cells, we evaluated serial genomic = DNA=20 profiles of NBE- and serum-cultured cells at various passage = numbers by=20 SNP analysis (Figure 6A and 6B). Even after 1 = year of=20 maintenance in NBE culture condition (more than 70 passages so = far), both=20 polyclonal and monoclonal populations of 0308-NBE cells largely = maintained=20 their parental tumor genotype. By sharp contrast, 0308-serum cells = underwent significant genomic rearrangements as early as passage = 10 (less=20 than 2 months of culture) represented by pseudotetraploidy, loss = of=20 heterozygosity (LOH), and deletion of the entire chromosome 4 and=20 chromosome 17 (Figure 6). Different subclones of = 0308-serum=20 cells showed other unique changes, including LOH of the entire = chromosome=20 6 (Figure 6A and 6B). None of these = genomic=20 alterations were detected in the 0308 parental tumor, nor in early = passage=20 cells, suggesting that these additional genomic changes occurred = during=20 the in vitro culture (Figure 6A and 6B). Giemsa banding = data=20 revealed that over 95% of late passage 0308-serum cells were=20 pseudotriploid or tetraploid (more than 200 nuclei examined) with = a modal=20 chromosome number of 77 ranging from 63 to 89. By contrast, more = than 90%=20 of 0308-NBE cells at passage 40 were near diploid. Similarly, = about 70% of=20 1228-serum cells at passage 20 were pseudotetraploid, whereas = passage 20=20 1228-NBE cells were near diploid (Figure 6E and 6F).

Intriguingly,=20 LOH in chromosomes 4 and 17 found in most of the late passage = 0308-serum=20 cells corresponds to the chromosomal locations of the hCDC4 = (Fbxw7) and p53 (TP53) genes, respectively = (Figure 7). Coincident with the = onset of=20 increased proliferation, tumorigenicity, and aneuploidy, late = passage=20 0308-serum cells were found to ultimately downregulate the = remaining=20 wild-type hCDC4 gene, essentially eliminating expression of hCDC4 = protein.=20 HCDC4, an E3 ubiquitin ligase, is reported to be involved in = degradation=20 of aurora kinase A (STK15 [AURKA]) (Mao et al., = 2004; Marumoto et al., 2005; Yang et al., 2004).

Figure 7. Genomic = alterations of=20 GBM cells cultured in serum condition in vitro

A and B: LOH of = chromosome=20 4 in later passage 0308-serum cells was confirmed by genomic = PCR=20 sequencing (A) = and=20 allelic-specific PCR (B). Boxed area in = A indicates = genomic=20 sequence of SNP_A-1510430 (hCDC4 locus in chromosome 4). = Note the=20 loss of DNA polymorphism in 0308-serum cells (C only instead = of C/T=20 sequences). B: = DNAs from=20 serum cells, NBE cells, and blood from the 0308 patients = were PCR=20 amplified by using allelic-specific primers. C and T = represent 3=E2=80=B2=20 end sequences of allelic-specific primers (sequences shown = in Experimental Procedures). Note the = absence of DNA=20 amplification in serum cells by allelic-specific PCR. DNA = marker is=20 indicated by M.

C: Western blot = analysis of=20 hCDC4 and STK15 in 0308-NBE and serum cells. =CE=B2-actin = Western blot=20 was used as a loading control.

D: LOH of entire = chromosome=20 17 in later passage 0308-serum cells (p10) determined by SNP = calls.=20 All SNPs located in chromosome 17 are pseudocolored by = heterozygous=20 calls (AB calls; yellow) and homozygous calls (AA and BB = calls; red=20 and blue, respectively). Absence of heterozygous calls = indicates the=20 loss of either maternal or paternal chromosome. Note the = complete=20 loss of heterozygous calls (yellow) in serum cells at = passage=20 10.

E: Genomic = sequence of the=20 p53 gene in 0308-serum cells. Boxed area indicates genomic = DNA=20 sequence difference. DNA from peripheral blood contained = wild-type=20 sequence, whereas parental tumors have both wild-type and = mutant=20 sequences. Note the absence of wild-type sequence in serum=20 cells.

View larger version: [In=20 this window] [In new = window]

Indeed, late=20 passage 0308-serum cells not only lose hCDC4 expression but also = have=20 significant upregulation of STK15 (Figure 7C). This is of potential = importance=20 given the recent description of hCDC4 as a haploinsufficient tumor = suppressor gene, inactivation of which results in aneuploidy (Mao = et al.,=20 2004; Rajagopalan et al., 2004). In addition, late passage 0308-serum cells lose = the entire=20 chromosome 17, which contains the wild-type p53 allele, leaving = only the=20 mutant p53 allele found in both the 0308-NBE cells and the = parental tumor=20 (Figure 7D and 7E). Loss of p53 = function has=20 also been associated with genomic instability (Duensing et al., = 2005; Fujiwara et al., 2005). Thus, the loss of both hCDC4 and p53 function in = 0308-serum cells, with their potentially important biological=20 consequences, further demonstrates the significant differences = between=20 serum cells and their matched parental tumors.

Discussion

The = TSC concept=20 has potentially profound implications both for basic biological = studies=20 and for the development of new therapeutic strategies (Dean et = al., 2005; Huntly et al., 2005). However, the true relatedness of previously = reported TSCs=20 to normal stem cells and their exact relevance to our current in = vitro and=20 in vivo models of cancer, as well as to primary human tumors in = situ,=20 remain unclear (Dontu et al., 2003; Huntly et al., 2005; Kim et al., 2005; Pardal et al., 2003; Reya et al., 2001). Using a model system derived from primary GBMs, = we have=20 demonstrated that NBE-cultured cells derived from primary GBMs = bear=20 remarkable similarity to normal NSCs. These similarities include = the=20 ability of NBE cells to form neurospheres in vitro, potential for=20 indefinite self-renewal, ability for terminal differentiation into = glial=20 and neuronal lineages, the possession of gene expression profiles = similar=20 to NSCs, and genetic stability over serial passage in vitro. = Nevertheless,=20 NBE cells also harbor all of the genetic aberrations found within = the=20 primary tumor, have gene expression profiles similar to the GBMs = they were=20 derived from, and appear to be the principal tumorigenic cell type = that=20 can recapitulate the overall in vivo phenotype of the parental = GBM.=20 Therefore, cells fulfilling the criteria of TSCs are maintained in = NBE=20 conditions (Reya et al., 2001).

By = contrast,=20 cells derived from the same GBM specimens as the NBE cells, but = grown=20 under standard in vitro conditions using serum-containing media, = lose=20 their self-renewing capabilities, have no ability to = differentiate, have=20 gene expression profiles that are similar to neither NSCs nor the = primary=20 GBMs they were derived from, and are neither clonogenic nor = tumorigenic.=20 While 0308 serum-cultured cells did ultimately regain tumorigenic=20 potential in later passages, these late passage cells do not = recapitulate=20 the tumorigenic phenotype of the original tumor and undergo = significant de=20 novo genomic alterations. This occurrence precisely matches the = phenotypic=20 and genotypic pattern found in most of the commonly used glioma = cell lines=20 in that only a small subset of these cell lines are tumorigenic in = vivo,=20 and of those, none demonstrate the tumor invasiveness seen with = the=20 NBE-derived tumors and that characterize human gliomas (Giannini = et al.,=20 2005). Additionally, it has been reported that most = primary=20 gliomas are near diploid, whereas established glioma cell lines = are=20 predominantly polyploid (Bigner et al., 1990; Hecht et al., 1995; Kubota et al., 2001; Wiltshire et al., 2000). Therefore, we propose that the inherent = =E2=80=9CTSC=E2=80=9D population=20 within primary GBMs is quickly lost in typical glioma culture = conditions,=20 and the cells found following prolonged in vitro passage are the = product=20 of an outgrowth of a cell clone(s) that has undergone profound = =E2=80=9Cde novo=E2=80=9D=20 genetic and/or epigenetic changes (Figure 8). In this context, it is = not=20 surprising that the gene expression profiles and the in vitro and = in vivo=20 behavior of serum-cultured cells appear remarkably similar to = other=20 established glioma cells lines but highly different from their = matched=20 NBE-cultured cells, parental tumors, and other primary human GBMs = (Mehrian=20 Shai et al., 2005).

Figure 8. A = hypothetical model=20 depicting the relationship between primary patient tumors, = TSC=20 cultures, and commonly used glioma cell lines

A: Summary of our = findings=20 in 0308 and 1228 GBM cells.

B: NBE culture = conditions=20 maintain the properties of primary tumors by preserving the = putative=20 TSC populations (shown as dark blue). Rather than diverting = TSCs=20 away from NSC-like behavior, it appears that the = GBM-defining=20 genetic perturbations (e.g., INK4a/ARF deletion, loss of = chromosome=20 10q, EGFR amplification; marked as yellow stars) merely = endow=20 oncogenic properties onto a population of cells with = inherent stem=20 cell properties (i.e., NSCs). By contrast, the TSC = population is=20 quickly lost through differentiation following growth in = serum,=20 similar to that seen in normal NSCs. TSC progenies (shown as = light=20 blue) are neither clonogenic nor tumorigenic despite the = fact that=20 they harbor the same genetic changes (marked as yellow = stars) found=20 in the corresponding TSCs and primary tumor. Continuous = cultures of=20 these cells in serum conditions result in the outgrowth of a = subpopulation(s) of cells harboring additional genetic = and/or=20 epigenetic changes (marked as black spots) not found in the = TSCs or=20 primary GBMs. These late passage serum-grown cells (shown as = red)=20 are vastly different both genetically and biologically = compared to=20 the primary tumors from which they are derived, but similar = to=20 commonly used glioma cell lines. Vertical green bar = represents the=20 increasing number of passages in vitro.

View larger version: [In=20 this window] [In new = window]

Several groups=20 have reported that TSC-like cells can be isolated from the = established=20 tumor cell lines by culturing these cells in serum-free media with = selected growth factors such as PDGF, bFGF, and EGF (Kondo et al., = 2004; Patrawala et al., 2005). If a subpopulation of TSCs can persist even = after years of=20 culture in vitro, it would be interesting to understand how these = cells=20 maintain their TSC-like properties under differentiation-inducing=20 conditions. Alternatively, it is possible that cells with = stem-like=20 properties reemerge from some established cell line populations, = through=20 epigenetic reprogramming and/or selection of a subpopulation of = cells with=20 genomic instability as seen in our 0308 serum-cultured cells = (Fujiwara et=20 al., 2005; Kondo et al., 2004; Shin et al., 2003). In either case, further studies evaluating the = relatedness=20 of these established cell line-derived =E2=80=9Cstem = cell-like=E2=80=9D cells to their=20 parental tumors and a thorough phenotypic/genotypic = characterization of=20 these cells would ultimately be required to truly understand how = closely=20 they fulfill the criteria of TSCs.

In = summary,=20 TSCs derived from primary GBMs bear remarkable phenotypic = similarity to=20 human NSCs and appear to be the tumorigenic component of at least = a subset=20 of human GBMs. Since TSCs maintain the ability to terminally=20 differentiate, agents that induce differentiation may have a = therapeutic=20 role in the treatment of the TSC compartment. Thus, NBE cells may = be an=20 optimal model system for the preclinical screening of such agents. = By=20 contrast, tumor cells grown under standard serum-containing cell = culture=20 conditions result in the loss of TSCs and ultimately lead to the = outgrowth=20 of a population of cells that are vastly different both = genetically and=20 biologically from the primary tumors from which they were derived. = These=20 observations bring into question the relevance of standard cancer = cell=20 lines for studying the biology of human cancer.

As = we=20 increasingly come to appreciate the significant heterogeneity = within=20 specific subtypes of solid tumors like GBMs and as we develop more = rationally based, molecularly targeted drugs, we will move = increasingly=20 toward the era of personalized therapy for individual tumors. Such = an=20 individualized therapeutic approach will require a model system = for=20 identifying and understanding the basic genotype, gene expression=20 profiles, and/or in vitro and in vivo biological characteristics = of unique=20 tumors from individual patients. The overarching implication of = our=20 findings is that many traditionally utilized cancer cell lines may = be=20 unreliable model systems for understanding the biology of primary = human=20 tumors, for screening new therapeutic agents, and ultimately for = guiding=20 personalized tumor therapy. TSCs, however, may prove to be a more = reliable=20 model system.

Experimental procedures

Tumor specimens and primary tumor=20 cultures

Following=20 informed consent, tumor samples classified as GBM based on World = Health=20 Organization (WHO) criteria were obtained from patients undergoing = surgical treatment at the National Institutes of Health in = accordance with=20 the appropriate Institutional Review Boards (Kleihues et al., = 2002). Within 1 to 3 hr after surgical removal, tumors = were=20 washed and enzymatically dissociated into single cells. Red blood = cells=20 were removed by differential centrifugation. Tumor cells were = cultured in=20 either NBE media consisting of Neurobasal media (Invitrogen), N2 = and B27=20 supplements (0.5=C3=97 each; Invitrogen), human recombinant bFGF = and EGF (50=20 ng/ml each; R&D Systems), or serum media consisting of DMEM = media=20 (Invitrogen) with 10% fetal bovine serum (Cellgro). For = neurosphere=20 culture of NBE cells, uncoated plastic dishes were used. For = adherent=20 culture of NBE cells, the plates were precoated with = poly-L-lysine/laminin=20 mixture (Invitrogen).

Immunohistochemistry

Immunohistochemistry of = paraffin=20 sections was performed as previously described. The following = antibodies=20 were used as primary antibodies: GFAP (DAKO), Sox2, Nestin = (R&D=20 Systems), and SSEA-1 (Developmental Studies Hybridoma Bank). The = corpus=20 callosum was stained by luxol fast blue (Sigma), according to the=20 manufacturer's recommendation.

Intracranial tumor cell injection = into SCID=20 mice

For = evaluation=20 of tumorigenicity, cells cultured in either NBE and serum = conditions were=20 resuspended in 2 =CE=BCl of HBSS and injected stereotactically = into neonatal=20 SCID mice. For the analysis of tumor histology, both neonatal and = 6- to=20 8-week-old mice were used. Coordinates for stereotactical = injections into=20 the adult mice were 3 mm distal to the midline, 2 mm anterior to = the=20 coronal suture, and 2.5 mm deep from the dura.

FISH analysis and SKY of tumor = cells

For = cytogenetic=20 studies, cells were arrested at the metaphase by colcemid (Sigma), = incubated in hypotonic buffer, and then fixed in methanol/acetic = acid=20 mixture. For Giemsa banding of chromosomes, more than 100 nuclei = were=20 analyzed for each cell population. Telomere FISH analysis was = performed=20 according to the manufacturer's instructions (Dako). EGFR FISH was = performed using centromere 7 and an EGFR probe set = (Vysis).

RNA expression array and data=20 analysis

Total RNAs were=20 isolated using TRIZOL (Invitrogen) and further purified using = RNeasy Mini=20 Kit (Qiagen). Samples were processed and hybridized to Affymetrix = U-133=20 plus2 GeneChip Arrays according to the Affymetrix GeneChip = Expression=20 Analysis Technical Manual (Affymetrix). Briefly, 6 =CE=BCg of = total RNA was=20 converted to cDNA with Superscript reverse transcriptase = (Invitrogen),=20 using T7-linked Oligo (dT) primer. Complementary DNA was = transcribed in=20 vitro using the T7 Bioarray High Yield RNA Transcript Labeling Kit = (ENZO=20 Diagnostics) to generate biotinylated cRNA. Purified cRNA (20 = =CE=BCg) was=20 fragmented and hybridized to GeneChip arrays. All arrays were = confirmed to=20 be within acceptable minimal quality control parameters including = the=20 signal intensity ratio of the 5=E2=80=B2 and 3=E2=80=B2 ends of = the internal control genes=20 of =CE=B2-actin and GAPDH less than 2. The initial gene expression = analysis=20 data files (CEL files) were generated using Affymetrix GeneChip = Operating=20 Software (GCOS) version 1.1.

The = gene=20 expression CEL files were normalized using dChip invariant method = and=20 PM-MM difference model was used to obtain the expression values = (Li et=20 al., 2001). Probe sets with zero variance and only = =E2=80=9Cabsent calls=E2=80=9D=20 across all the samples were removed prior to data analysis. The=20 hierarchical cluster analysis, principal component analysis (PCA), = and=20 two-way analysis of variance (ANOVA) were performed using Partek = software=20 6.0 (Partek Inc.). The Pearson correlation with average linkage = and=20 Euclidean distance with average linkage was used for unsupervised=20 hierarchical clustering and for supervised hierarchical = clustering,=20 respectively. For ANOVA, the false discovery rate of 0.05 and = 2-fold=20 change were applied as thresholds to select the up- or = downregulated gene=20 lists.

SNP array and data analysis

Genomic DNA=20 from patient tumor samples and cultured cells was prepared using = the=20 QiaAmp DNA kit (Qiagen). DNA from patients' blood was prepared by = QiaAmp=20 DNA blood mini kit (Qiagen). Array experiments were performed = according to=20 the manufacturer's recommendations (Affymetrix). Briefly, 250 ng = of=20 genomic DNA was digested with XbaI (New England Biolabs), ligated = to XbaI=20 adaptors, and subsequently amplified by PCR. Fragmented PCR = products were=20 then labeled, denatured, and hybridized to the arrays. After=20 hybridization, the arrays were stained on the fluidics station 450 = and=20 scanned using a high-resolution microarray scanner 3000.

GeneChip Human=20 Mapping 10K arrays (Affymetrix) covers 11,555 single-nucleotide=20 polymorphism (SNP) loci in the human genome with average = resolution of one=20 SNP every 210 kb. SNP calls were determined by GDAS version 2.0.=20 Chromosome copy numbers were estimated using Affymetrix GeneChip=20 Chromosome Copy Number Tool version 1.1.1(Huang et al., 2004).

PCR sequencing and quantitative = real-time=20 PCR

Genomic=20 alterations identified by the SNP analysis were verified by PCR = and=20 subsequent sequencing. The presence of LOH of chromosome 4 in = later=20 passage 0308-serum cells was confirmed by PCR sequencing of = several SNPs=20 including SNP_A-1510430 and SNP_A-1515580 (Affymetrix). The = following=20 primers were used: 5=E2=80=B2-AGTCCCTTGAGGCATGTTGG-3=E2=80=B2 and=20 5=E2=80=B2-GTCAACTTGAGCAAAAAAATCTAGC-3=E2=80=B2 (for = SNP_A-1510430),=20 5=E2=80=B2-GATTCATTGTTGTTGCTAAAAGTG-3=E2=80=B2 and 5=E2=80=B2- = TGCCTCTCTTCTCTCTTCCC-3=E2=80=B2 (for=20 SNP_A-1515580). For allele-specific PCR for the above SNPs, the = following=20 primers were used: 5=E2=80=B2-CGCAACAACAATGAATCTCC-3=E2=80=B2 and=20 5=E2=80=B2-CGCAACAACAATGAATCTCT-3=E2=80=B2 = (SNP_A-1510430),=20 5=E2=80=B2-GAGCCCTTGGCTTCTGATATA-3=E2=80=B2 and = 5=E2=80=B2-GAGCCCTTGGCTTCTGATATG-3=E2=80=B2 (SNP_A-1515580).

Real-time PCR=20 was performed on an ABI Prism 7900 sequence detection system = (Applied=20 Biosystems) according to the manufacturer's instructions. All = samples=20 including no template controls were assayed in triplicates. The = relative=20 amount of targets transcripts was normalized to the number of = human GAPDH=20 transcripts found in the same sample. The human reference RNA = (Qiagen) was=20 used as the calibrator. The relative quantitation of target gene=20 expression was performed with the standard curve or comparative = cycle=20 threshold (C_T) method.

Acknowledgments

This research=20 was supported by the Intramural Research Program of the NIH, = National=20 Cancer Institute, Center for Cancer Research. We thank Dr. David=20 Livingston for critical reading of the manuscript. We are grateful = to Dr.=20 Eugene Major for providing human fetal neural progenitor cells, = Dr. Roscoe=20 Stanyon and NCI-Frederick cytogenetics core facility for SKY = analysis, and=20 Dr. Alexander Vortmeyer for histological evaluation of = tumors.

Supplemental=20 data

Document S1. Four supplemental figures and nine = supplemental=20 tables

Accession=20 numbers

The = microarray=20 data have been submitted to the Gene Expression Omnibus (GEO) = database at=20 http://www.ncbi.nlm.nih.gov/geo/; the accession = number is=20 GSE4536.

References

Bigner, = S.H., and=20 Vogelstein, B. (1990). Cytogenetics and=20 molecular genetics of malignant gliomas and = medulloblastoma.=20 Brain Pathol. 1, 12-18. = [Medline]=20

Brazel, = C.Y.,=20 Limke, T.L., Osborne, J.K., Miura, T., Cai, J., Pevny, L., and = Rao,=20 M.S. (2005). Sox2 expression = defines a=20 heterogeneous population of neurosphere-forming cells in the adult = murine=20 brain. Aging Cell 4, 197-207. = [Medline]=20

Capela, = A., and=20 Temple, S. (2002). LeX/ssea-1 = is=20 expressed by adult mouse CNS stem cells, identifying them as=20 nonependymal. Neuron 35, 865-875. [Medline]=20 [Summary]=20 [Full=20 Text]

Dean, = M., Fojo, T.,=20 and Bates, S. (2005). Tumour = stem cells=20 and drug resistance. Nat. Rev. Cancer 5, 275-284. = [Medline]=20

Deschavanne, P.J.,=20 and Fertil, B. (1996). A review = of human=20 cell radiosensitivity in vitro. Int. J. Radiat. Oncol. Biol. Phys. 34,=20 251-266. [Medline]=20

Dontu, = G.,=20 Abdallah, W.M., Foley, J.M., Jackson, K.W., Clarke, M.F., = Kawamura, M.J.,=20 and Wicha, M.S. (2003). In = vitro=20 propagation and transcriptional profiling of human mammary = stem/progenitor=20 cells. Genes Dev. 17, 1253-1270. = [Medline]=20

Duensing, A., and=20 Duensing, S. (2005). Guilt by=20 association? p53 and the development of aneuploidy in = cancer.=20 Biochem. Biophys. Res. Commun. 331,=20 694-700. [Medline]=20

Fortunel, N.O.,=20 Otu, H.H., Ng, H.H., Chen, J., Mu, X., Chevassut, T., Li, X., = Joseph, M.,=20 Bailey, C., Hatzfeld, J.A. et al. (2003). Comment on = =E2=80=9C=E2=80=98Stemness=E2=80=99: Transcriptional profiling of=20 embryonic and adult stem cells=E2=80=9D and =E2=80=9CA stem cell = molecular=20 signature.=E2=80=9D. Science 302, author reply=20 393, 393. [Medline]=20

Fujiwara, T.,=20 Bandi, M., Nitta, M., Ivanova, E.V., Bronson, R.T., and Pellman, = D.=20 (2005). Cytokinesis failure generating = tetraploids promotes tumorigenesis in p53-null cells. = Nature 437, 1043-1047.=20 [Medline]=20

Gage, = F.H., Ray,=20 J., and Fisher, L.J. (1995). Isolation,=20 characterization, and use of stem cells from the CNS. = Annu. Rev. Neurosci. 18, 159-192. [Medline]=20

Galli, = R., Binda,=20 E., Orfanelli, U., Cipelletti, B., Gritti, A., De Vitis, S., = Fiocco, R.,=20 Foroni, C., Dimeco, F., and Vescovi, A. (2004). Isolation and characterization of = tumorigenic,=20 stem-like neural precursors from human glioblastoma. = Cancer Res. 64,=20 7011-7021. [Medline]=20

Giannini, C.,=20 Sarkaria, J.N., Saito, A., Uhm, J.H., Galanis, E., Carlson, B.L.,=20 Schroeder, M.A., and James, C.D. (2005). Patient tumor EGFR and PDGFRA gene = amplifications=20 retained in an invasive intracranial xenograft model of = glioblastoma=20 multiforme. Neuro-oncol. 7, 164-176. = [Medline]=20

Haik, = S., Gauthier,=20 L.R., Granotier, C., Peyrin, J.M., Lages, C.S., Dormont, D., and = Boussin,=20 F.D. (2000). Fibroblast growth = factor 2=20 up regulates telomerase activity in neural precursor cells. = Oncogene 19, 2957-2966. = [Medline]=20

Hecht, = B.K.,=20 Turc-Carel, C., Chatel, M., Grellier, P., Gioanni, J., Attias, R., = Gaudray, P., and Hecht, F. (1995). Cytogenetics of malignant gliomas: I. The = autosomes=20 with reference to rearrangements. Cancer Genet. Cytogenet. 84, = 1-8.=20 [Medline]=20

Hemmati, H.D.,=20 Nakano, I., Lazareff, J.A., Masterman-Smith, M., Geschwind, D.H.,=20 Bronner-Fraser, M., and Kornblum, H.I. (2003). Cancerous stem cells can arise from = pediatric brain=20 tumors. Proc. Natl. Acad. Sci. USA 100,=20 15178-15183. [Medline]=20

Huang, = J., Wei, W.,=20 Zhang, J., Liu, G., Bignell, G.R., Stratton, M.R., Futreal, P.A., = Wooster,=20 R., Jones, K.W., and Shapero, M.H. (2004). Whole genome DNA copy number changes = identified by=20 high density oligonucleotide arrays. Hum. Genomics 1,=20 287-299. [Medline]=20

Huntly, = B.J., and=20 Gilliland, D.G. (2005). Leukaemia stem=20 cells and the evolution of cancer-stem-cell research. = Nat. Rev. Cancer 5,=20 311-321. [Medline]=20

Ivanova, N.B.,=20 Dimos, J.T., Schaniel, C., Hackney, J.A., Moore, K.A., and = Lemischka,=20 I.R. (2002). A stem cell = molecular=20 signature. Science 298, 601-604. = [Medline]=20

Kim, = C.F., Jackson,=20 E.L., Woolfenden, A.E., Lawrence, S., Babar, I., Vogel, S., = Crowley, D.,=20 Bronson, R.T., and Jacks, T. (2005). Identification of bronchioalveolar stem = cells in=20 normal lung and lung cancer. Cell 121, 823-835. [Medline]=20 [Summary]=20 [Full=20 Text]

Kleihues, P.,=20 Louis, D.N., Scheithauer, B.W., Rorke, L.B., Reifenberger, G., = Burger,=20 P.C., and Cavenee, W.K. (2002). The WHO=20 classification of tumors of the nervous system. = J. Neuropathol. Exp. = Neurol. 61, 215-225. [Medline]=20

Kondo, = T., and=20 Raff, M. (2004). Chromatin = remodeling and=20 histone modification in the conversion of oligodendrocyte = precursors to=20 neural stem cells. Genes Dev. 18, 2963-2972. = [Medline]=20

Kondo, = T.,=20 Setoguchi, T., and Taga, T. (2004). Persistence of a small subpopulation of = cancer=20 stem-like cells in the C6 glioma cell line. Proc. Natl. Acad. Sci. USA = 101, 781-786. [Medline]=20

Kubota, = H.,=20 Nishizaki, T., Harada, K., Harada, K., Oga, A., Ito, H., Suzuki, = M., and=20 Sasaki, K. (2001). Identification of=20 recurrent chromosomal rearrangements and the unique relationship = between=20 low-level amplification and translocation in glioblastoma.=20 Genes Chromosomes Cancer 31, = 125-133.=20 [Medline]=20

Lendahl, U.,=20 Zimmerman, L.B., and McKay, R.D. (1990). CNS stem cells express a new class of = intermediate=20 filament protein. Cell 60, 585-595. [Medline]=20 [Summary]=20

Li, C., = and Wong,=20 W.H. (2001). Model-based = analysis of=20 oligonucleotide arrays: expression index computation and outlier=20 detection. Proc. Natl. Acad. Sci. USA 98, = 31-36.=20 [Medline]=20

Maher, = E.A.,=20 Furnari, F.B., Bachoo, R.M., Rowitch, D.H., Louis, D.N., Cavenee, = W.K.,=20 and DePinho, R.A. (2001). Malignant=20 glioma: genetics and biology of a grave matter. = Genes Dev. 15,=20 1311-1333. [Medline]=20

Mao, = J.H.,=20 Perez-Losada, J., Wu, D., Delrosario, R., Tsunematsu, R., = Nakayama, K.I.,=20 Brown, K., Bryson, S., and Balmain, A. (2004). Fbxw7/Cdc4 is a p53-dependent, = haploinsufficient=20 tumour suppressor gene. Nature 432, 775-779. [Medline]=20

Marumoto, T.,=20 Zhang, D., and Saya, H. (2005). Aurora-A=E2=80=94a guardian of poles. = Nat. Rev. Cancer 5,=20 42-50. [Medline]=20

McKay, = R.=20 (1997). Stem cells in the central = nervous=20 system. Science 276, 66-71. [Medline]=20

Mehrian = Shai, R.,=20 Reichardt, J.K., Ya-Hsuan, H., Kremen, T.J., Liau, L.M., = Cloughesy, T.F.,=20 Mischel, P.S., and Nelson, S.F. (2005). Robustness of gene expression profiling in = glioma=20 specimen samplings and derived cell lines. Brain Res. Mol. Brain Res. = 136, 99-103. [Medline]=20

Morshead, C.M., and=20 van der Kooy, D. (2004). Disguising adult=20 neural stem cells. Curr. Opin. Neurobiol. 14, = 125-131.=20 [Medline]=20

Ostenfeld, T.,=20 Caldwell, M.A., Prowse, K.R., Linskens, M.H., Jauniaux, E., and = Svendsen,=20 C.N. (2000). Human neural = precursor cells=20 express low levels of telomerase in vitro and show diminishing = cell=20 proliferation with extensive axonal outgrowth following=20 transplantation. Exp. Neurol. 164, 215-226. = [Medline]=20

Pardal, = R., Clarke,=20 M.F., and Morrison, S.J. (2003). Applying=20 the principles of stem-cell biology to cancer. Nat. Rev. Cancer 3,=20 895-902. [Medline]=20

Park, = I.K., Qian,=20 D., Kiel, M., Becker, M.W., Pihalja, M., Weissman, I.L., Morrison, = S.J.,=20 and Clarke, M.F. (2003). Bmi-1 = is=20 required for maintenance of adult self-renewing haematopoietic = stem=20 cells. Nature 423, 302-305. [Medline]=20

Patrawala, L.,=20 Calhoun, T., Schneider-Broussard, R., Zhou, J., Claypool, K., and = Tang,=20 D.G. (2005). Side population is = enriched=20 in tumorigenic, stem-like cancer cells, whereas ABCG2+ and = ABCG2=E2=88=92 cancer=20 cells are similarly tumorigenic. Cancer Res. 65, 6207-6219. = [Medline]=20

Phillips, H.S.,=20 Kharbanda, S., Chen, R., Forrest, W.F., Soriano, R.H., Wu, T.D., = Misra,=20 A., Nigro, J.M., Colman, H., Soroceanu, L. et al. (2006). = Molecular subclasses of high-grade glioma = predict=20 prognosis, delineate a pattern of disease progression, and = resemble stages=20 in neurogenesis. Cancer Cell 9, 157-173. = [Medline]=20 [Summary]=20 [Full=20 Text]

Rajagopalan, H.,=20 Jallepalli, P.V., Rago, C., Velculescu, V.E., Kinzler, K.W., = Vogelstein,=20 B., and Lengauer, C. (2004). Inactivation=20 of hCDC4 can cause chromosomal instability. Nature 428, 77-81.=20 [Medline]=20

Ramalho-Santos, M.,=20 Yoon, S., Matsuzaki, Y., Mulligan, R.C., and Melton, D.A. = (2002).=20 =E2=80=9CStemness=E2=80=9D: = transcriptional profiling of=20 embryonic and adult stem cells. Science 298, 597-600. = [Medline]=20

Reya, = T., Morrison,=20 S.J., Clarke, M.F., and Weissman, I.L. (2001). Stem cells, cancer, and cancer stem = cells.=20 Nature 414, 105-111. [Medline]=20

Reynolds, B.A.,=20 Tetzlaff, W., and Weiss, S. (1992). A=20 multipotent EGF-responsive striatal embryonic progenitor cell = produces=20 neurons and astrocytes. J.=20 Neurosci. 12, 4565-4574. [Medline]=20

Romer, = J., and=20 Curran, T. (2005). Targeting=20 medulloblastoma: small-molecule inhibitors of the Sonic Hedgehog = pathway=20 as potential cancer therapeutics. Cancer Res. 65, 4975-4978. = [Medline]=20

Sakakibara, S.,=20 Imai, T., Hamaguchi, K., Okabe, M., Aruga, J., Nakajima, K., = Yasutomi, D.,=20 Nagata, T., Kurihara, Y., Uesugi, S. et al. (1996). Mouse-Musashi-1, a neural RNA-binding = protein highly=20 enriched in the mammalian CNS stem cell. Dev. Biol. 176,=20 230-242. [Medline]=20

Sanai, = N.,=20 Alvarez-Buylla, A., and Berger, M.S. (2005). Neural stem cells and the origin of = gliomas.=20 N.=20 Engl. J. Med. 353, 811-822. [Medline]=20

Shin, = H.J., Baek,=20 K.H., Jeon, A.H., Park, M.T., Lee, S.J., Kang, C.M., Lee, H.S., = Yoo, S.H.,=20 Chung, D.H., Sung, Y.C. et al. (2003). Dual roles of human BubR1, a mitotic = checkpoint=20 kinase, in the monitoring of chromosomal instability. = Cancer Cell 4,=20 483-497. [Medline]=20 [Summary]=20 [Full=20 Text]

Singh, = S.K.,=20 Clarke, I.D., Terasaki, M., Bonn, V.E., Hawkins, C., Squire, J., = and=20 Dirks, P.B. (2003). Identification of a=20 cancer stem cell in human brain tumors. Cancer Res. 63,=20 5821-5828. [Medline]=20

Singh, = S.K.,=20 Hawkins, C., Clarke, I.D., Squire, J.A., Bayani, J., Hide, T., = Henkelman,=20 R.M., Cusimano, M.D., and Dirks, P.B. (2004). Identification of human brain tumour = initiating=20 cells. Nature 432, 396-401. [Medline]=20

Taylor, = M.D.,=20 Poppleton, H., Fuller, C., Su, X., Liu, Y., Jensen, P., Magdaleno, = S.,=20 Dalton, J., Calabrese, C., Board, J. et al. (2005). Radial glia cells are candidate stem cells = of=20 ependymoma. Cancer Cell 8, 323-335. = [Medline]=20 [Summary]=20 [Full=20 Text]

Vogelstein, B., and=20 Kinzler, K.W. (2004). Cancer = genes and=20 the pathways they control. Nat. Med. 10, 789-799. = [Medline]=20

Wiltshire, R.N.,=20 Rasheed, B.K., Friedman, H.S., Friedman, A.H., and Bigner, = S.H.=20 (2000). Comparative genetic patterns = of=20 glioblastoma multiforme: potential diagnostic tool for tumor=20 classification. Neuro-oncol. 2, 164-173. = [Medline]=20

Yang, = H., Ou, C.C.,=20 Feldman, R.I., Nicosia, S.V., Kruk, P.A., and Cheng, J.Q. = (2004).=20 Aurora-A kinase regulates telomerase = activity=20 through c-Myc in human ovarian and breast epithelial cells. = Cancer Res. 64, 463-467. = [Medline]=20

Yuan, = X., Curtin,=20 J., Xiong, Y., Liu, G., Waschsmann-Hogiu, S., Farkas, D.L., Black, = K.L.,=20 and Yu, J.S. (2004). Isolation = of cancer=20 stem cells from adult glioblastoma multiforme. Oncogene 23,=20 9392-9400. [Medline]=20

Zhu, = Y., Guignard,=20 F., Zhao, D., Liu, L., Burns, D.K., Mason, R.P., Messing, A., and = Parada,=20 L.F. (2005). Early inactivation = of p53=20 tumor suppressor gene cooperating with NF1 loss induces malignant=20 astrocytoma. Cancer Cell 8, 119-130. = [Medline]=20 [Summary]=20 [Full=20 Text]

Article = History

Received: December 5, = 2005
Revised:=20 February 2, 2006
Accepted: March 22, 2006
Published: May 15, = 2006

Footnotes

3 These authors contributed equally to this=20 work.

Radial=20 glia cells are candidate stem cells of epen...
Stem=20 Cells and Cancer...
Early=20 inactivation of p53 tumor suppressor gene co...

More=20 related articles...

Article Options

Table=20 of Contents
Summary
PDF=20 (1404K)
Suppleme= ntal=20 Data

Search = Medline=20 for

Jeongw= u=20 Lee
Howa= rd=20 A. Fine

Cell		Immunity		Neuron		Molecular=20 Cell		Structure
Current=20 Biology		Developmental = Cell		Cancer Cell		Chemistry &=20 Biology		Cell=20 Metabolism

Article

Tumor stem cells derived = from=20 glioblastomas cultured in bFGF and EGF more closely mirror the = phenotype=20 and genotype of primary tumors than do serum-cultured cell=20 lines

Jeongwu Lee,1=20 Svetlana Kotliar= ova,1,3=20 Yuri Kotliarov,<= SPAN=20 class=3Dja50-ce-sup>1,3=20 Aiguo Li,1,3=20 Qin Su,1=20 Nicholas=20 M. Donin,1=20 Sandra Pastorino= ,1=20 Benjamin=20 W. Purow,1=20 Neil Christopher= ,1=20 Wei Zhang,1=20 John = K. Park,2=20 and Howard = A. Fine1,

In vitro characterization of GBM = cells=20 cultured in NBE and serum conditions

Tumorigenic potential of GBM cells = cultured in=20 NBE or serum conditions

Phenotypic characterization of = tumors derived=20 from NBE- and late passage serum-cultured GBM cells

Transcriptome analysis of NBE- and=20 serum-cultured GBM cells

Passage-dependent gene expression = profile=20 changes in NBE and serum GBM cells

Analysis of NBE- and serum-specific = genes and=20 their expression in primary GBMs

Genotypic characterization of NBE- = and=20 serum-cultured GBM cells

Tumor specimens and primary tumor=20 cultures

Immunohistochemistry

Intracranial tumor cell injection = into SCID=20 mice

FISH analysis and SKY of tumor = cells

RNA expression array and data=20 analysis

SNP array and data analysis

PCR sequencing and quantitative = real-time=20 PCR