The genomic profile of human malignant glioma is altered early in primary cell culture and preserved in spheroids

Abstract

Screening of therapeutics relies on representative cancer models. The representation of human glioblastoma by in vitro cell culture models is questionable. We obtained genomic profiles by array comparative genomic hybridization of both short- and long-term primary cell and spheroid cultures, derived from seven glioblastomas and one anaplastic oligodendroglioma. Chromosomal copy numbers were compared between cell cultures and spheroids and related to the parental gliomas using unsupervised hierarchical clustering and correlation coefficient. In seven out of eight short-term cell cultures, the genomic profiles clustered further apart from their parental tumors than spheroid cultures. In four out of eight samples, the genetic changes in cell culture were substantial. The average correlation coefficient between parental tumors and spheroid profiles was 0.89 (range: 0.79-0.97), whereas that between parental tumors and cell cultures was 0.62 (range: 0.10-0.96). In two out of three long-term cell cultures progressive genetic changes had developed, whereas the spheroid cultures were genetically stable. It is concluded that genomic profiles of primary cell cultures from glioblastoma are frequently deviant from parental tumor profiles, whereas spheroids are genetically more representative of the glioblastoma. This implies that glioma cell culture data have to be handled with the highest caution.

Full text

SHORT COMMUNICATION
The genomic profile of human malignant glioma is altered early in primary
cell culture and preserved in spheroids
PC De Witt Hamer
1
, AAG Van Tilborg
1,2
, PP Eijk
3
, P Sminia
4
, D Troost
2
, CJF Van Noorden
5
,
B Ylstra
3
and S Leenstra
1
1
Department of Neurosurgery, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands;
2
Department of
Neuropathology, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands;
3
Section Micro Array Facility,
Department of Pathology, VU University Medical Center, Amsterdam, The Netherlands;
4
Department of Radiation Oncology, VU
University Medical Center, Amsterdam, The Netherlands and
5
Department of Cell Biology and Histology, Academic Medical Center,
University of Amsterdam, Amsterdam, The Netherlands
Screening of therapeutics relies on representative cancer
models. The representation of human glioblastoma by
in vitro cell culture models is questionable. We obtained
genomic profiles by array comparative genomic hybridiza-
tion of both short- and long-term primary cell and spheroid
cultures, derived from seven glioblastomas and one anaplas-
tic oligodendroglioma. Chromosomal copy numbers were
compared between cell cultures and spheroids and related to
the parental gliomas using unsupervised hierarchical
clustering and correlation coefficient. In seven out of eight
short-term cell cultures, the genomic profiles clustered
further apart from their parental tumors than spheroid
cultures. In four out of eight samples, the genetic changes in
cell culture were substantial. The average correlation
coefficient between parental tumors and spheroid profiles
was 0.89 (range: 0.79–0.97), whereas that between parental
tumors and cell cultures was 0.62 (range: 0.10–0.96). In two
out of three long-term cell cultures progressive genetic
changes had developed, whereas the spheroid cultures were
genetically stable. It is concluded that genomic profiles of
primary cell cultures from glioblastoma are frequently
deviant from parental tumor profiles, whereas spheroids are
genetically more representative of the glioblastoma. This
implies that glioma cell culture data have to be handled
with the highest caution.
Oncogene (2008) 27, 2091–2096; doi:10.1038/sj.onc.1210850;
published online 15 October 2007
Keywords: glioma; primary cell culture; spheroid; com-
parative genomic hybridization; microarray
Glioblastoma has a poor survival, despite maximal
neurosurgical resection and chemoirradiation (Stupp
et al., 2005). In the development of new treatment
modalities, biologically representative cancer models are
essential. In vivo models appear to retain genetic
hallmarks of human glioblastoma (Pandita et al.,
2004). However, therapeutic screening is limited to a
low-throughput basis, whereas in vitro cancer models are
suited for high-throughput screening of novel therapeu-
tics. In vitro models include established cell line cultures,
short-term primary cell cultures and organotypic spher-
oids. Often, the in vitro sensitivity of cancer cells does not
correlate well with in vivo response. Especially, cell
cultures have been long questioned to represent glioma
biology due to changes under standard culture condi-
tions (Wolff et al., 1999; Voskoglou-Nomikos et al.,
2003; Lee et al., 2006). Therefore, short-term primary cell
cultures (with less than 10 trypsinization passages) and
particularly three-dimensional spheroid models were
considered to reflect the tumor biology better (Suther-
land, 1988; Bjerkvig et al., 1990). We hypothesized that
spheroids more closely mirror the genetics of the primary
tumors than primary cell cultures. Therefore, we
compared the genomic profiles of short-term primary
cell cultures, spheroid cultures and parental tumors of
which cultures were derived. The genetic alterations in
glioblastoma involved in gliomagenesis and progression
are well characterized (Koschny et al., 2002; Kotliarov
et al., 2006). Notably, specific alterations correlate with
response to therapy and prognosis in particular for
oligodendroglioma (Cairncross et al., 1998; Burton et al.,
2002). To compare the genetic stability of primary cell
cultures to those of spheroid cultures, copy number
abnormalities were determined by genome-wide array
comparative genomic hybridization (array CGH) (Snij-
ders et al., 2001). We found that profiles from a
substantial number of short-term primary cell cultures
had changed considerably compared to the parental
tumor and that changes were progressive over weeks,
whereas spheroid profiles were genetically stable.
Genomic profiles during short-term primary cell and
spheroid culture
As a first approach to determine similarities and
differences in genetics, unsupervised clustering analysis
was performed with 24 genomic profiles collected from
Received 10 April 2007; revised 28 August 2007; accepted 17 September
2007; published online 15 October 2007
Correspondence: Dr P De Witt Hamer, Department of Neurosurgery,
Academic Medical Center, University of Amsterdam, Room H2-238,
PO Box 22660, 1100 DD Amsterdam, The Netherlands.
E-mail: P.C.DeWittHamer@amc.nl
Oncogene (2008) 27, 2091–2096
&
2008 Nature Publishing Group
All rights reserved 0950-9232/08 $30.00
www.nature.com/onc

eight parental tumor samples, their primary cell and
spheroid cultures (Figure 1a). Profiles derived from the
same tissue material tended to cluster together, specifi-
cally those of samples 33, 32, 153, 60 and 58. In seven
out of eight samples, the spheroid culture profile
clustered closer to the parental tumor profile than the
primary cell culture profile, with the exception of sample
58, where parental tumor, primary cell culture and
spheroid culture had nearly identical profiles. The
primary cell culture profiles of samples 3446, 160, 32
and 55 had altered substantially compared to the
parental tumor profile. Notably, the primary cell
cultures from samples 3446, 160 and 55 had reverted
genome-wide to a near diploid status. This diploidy
suggests contamination of primary cultures with normal
cells, but the glial origin of cultured cells was confirmed
by their GFAP positivity (primary antibody from Dako,
Haverlee, Belgium; data not shown), indicating that
glioma cells and not normal cells were cultured.
The copy numbers that changed in the primary cell
culture profiles included genomic regions that are
associated with therapeutic response and prognosis
(Cairncross et al., 1998; Burton et al., 2002). For instance,
1p loss as observed in the parental tumor of samples 3446
and 55 was reconstituted in the corresponding primary
cell culture profiles. Furthermore, loss of chromosome 10
as observed in the parental tumor of samples 3446, 160
and 55 was reconstituted in the primary cell culture
profiles. Amplification on 7p, the EGFR gene locus, as
observed in the parental tumor samples 33 and 160, was
lost in the short-term primary cell cultures.
Genomic profiles were in general well-preserved in
short-term spheroid cultures (Figure 1a), which is in
contrast with profiles of short-term primary cell
cultures. The average correlation coefficient between
the whole-genomic primary cell culture and parental
tumor profiles was 0.62 (range: 0.11–0.96) and between
the whole-genomic spheroid culture and parental tumor
profiles was 0.89 (range: 0.79–0.97) (Figure 1b).
Genomic changes occur progressively in primary cell
cultures, whereas spheroid cultures are genetically stable
To study the genetic stability of cell and spheroid cultures
during an expanded period of time, we compared array
CGH profiles from three patients after 2, 6 and 12 weeks
in culture with the parental tumor samples as a reference
(Figures 2a–c). Primary cell culture profiles were preserved
initially in samples 58 and 60. However, progressive
changes occurred, most markedly at 12 weeks of culture
(Figure 2d). For instance, an initial 9p loss in the parental
tumor of sample 55 became a 9p gain in primary cell
culture at 12 weeks (Figure 2a). A normal copy number of
2p and 11p in the parental tumor of sample 58 became a
2p and 11p gain in primary cell culture at 6 weeks and loss
of the entire chromosome 2 and 11p occurred at 12 weeks
(Figure 2b). A gain in chromosome 3 in the parental tumor
of sample 60 was initially preserved in primary cell culture
at 2 weeks and was lost at 12 weeks. Finally, an initial loss
of 19q in the parental tumor of sample 60 was
reconstituted at 12 weeks of primary cell culture
(Figure 2c).
The genomic profiles of spheroid cultures were stable
up to 12 weeks of culture, resembling those of their
parental tumors (Figure 2).
Array CGH confirms genetic alterations characteristic for
glioblastoma
To validate the robustness of the array CGH technique,
known genetic alterations were identified in the parental
Figure 1 Comparison of genomic profiles of short-term primary cell and spheroid cultures derived from seven glioblastomas and one
anaplastic oligodendroglioma. The clinical, histological and culture characteristics of the samples are shown in Supplementary Table
S1. (a) Genomic profiles from the parental tumor (T, white), 2-week primary cell culture (C, yellow) and 2-week spheroid culture
(S, blue) are displayed in cluster tree order. Data were obtained using an array comparative genomic hybridization (aCGH) with
custom arrays containing 6.4K BAC clones, prepared as described (Smeets et al., 2006). Therefore, BACs were spotted in triplicate
onto CodeLink slides (Amersham Biosciences, Buckinghamshire, UK), using the OmniGrid 100 Microarrayer (Genomic Solutions,
Ann Arbor, MI, USA) equipped with SMP3 pins (TeleChem, Sunnyvale, CA, USA). DNA was extracted from samples and peripheral
blood from 18 male blood donors as reference using standard SDS-proteinase K methods and collected by ethanol precipitation.
Labeling of extracted DNA with Cy3/Cy5 and hybridization to the arrays were performed as described (Smeets et al., 2006). The array
slides were scanned with Microarray Scanner G2505B (Agilent Technologies, Palo Alto, CA, USA), and the spot intensities were
measured with BlueFuse (release 3.2; BlueGnome, Cambridge, UK; http://www.bluegnome.co.uk/). The data set was deposited at
GEO (accession GSE6042) (Barrett et al., 2005). For further analysis, spots were excluded with quality flag o1 or confidence value
o0.85. Triplicates of spot ratios (Cy3/Cy5 intensity signals) were fused within BlueFuse in a confidence weighted fashion. The resulting
log
2
transformed ratios were normalized to their mode value and shown as black dots. The gray reference lines represent 1 and þ 1
log
2
ratio values. Data are truncated at 1.4 and þ 1.4 (B0.1% of data points). The chromosomal positions of the BAC clones were
retrieved from the UCSC genome browser (h16 freeze, 07/2003, NCBI built 34; http://genome.ucsc.edu/). Chromosomal numbering is
printed above the data. After ordering the clones by chromosomal position, gains and losses for genomic regions were called with CGH
plotter using default settings (release 7.1.2005; moving average filtering of 5 and a default constant value of 6) (Autio et al., 2003).
Outlined in red and green are estimated copy number gains and losses, respectively, of genomic regions according to the CGH-plotter
method. The ordered whole-genome profiles with data points that retain the amplitude of genomic changes as called by CGH-plotter
were hierarchically clustered using Pearson’s uncentered correlation coefficient as metric with average linking in Cluster (release 3.0),
based on an algorithm as described (Eisen et al., 1998) and plotted in TreeView (release 1.0.12). Qualitatively similar results were
obtained when the correlation coefficients were calculated by Kendall’s t or Spearman’s r statistics (data not shown). The
chromosomes X and Y were omitted from the cluster analysis, because the intensity ratios were considered noninformative regarding
cancer-related copy number abnormalities. Nodes in the cluster tree to the right increase in similarity of genomic profiles from right to
left. (b) Bar chart comparing similarity of genomic profiles for primary cell culture and parental tumor (yellow), and spheroid culture
and parental tumor (blue) of each of the eight samples, expressed as Pearson’s correlation coefficients using the R package (release
2.2.1; http://cran.r-project.org/).
Genomic profiles of glioblastoma in culture
PC De Witt Hamer et al
2092
Oncogene

Genomic profiles of glioblastoma in culture
PC De Witt Hamer et al
2093
Oncogene

glioma samples. Several specific genetic alterations have
been observed in human glioblastoma, including losses
of 9p21 (CDKN2A), 10q23 (PTEN) and 13q14 (Rb), and
gains of 1q32 (MDM4), 7p12 (EGFR), 8q24 (c-MYC)
and 12q13–15 (CDK4); and in oligodendroglioma, loss
of 1p and 19q (Koschny et al., 2002). The frequency of
involvement of copy number changes in the present
analysis with relatively few samples was consistent with
this. In our study, loss of 9p21 was observed in five out
of eight samples, loss of chromosome 10 in all samples,
four of which included apparent homozygote losses
(log
2
ratioo0.8) of PTEN at 10q23 and loss of 13q14
in three out of eight samples. Furthermore, gain of 1q32
was observed in two samples, gain of chromosome 7 in
four samples, two of which were apparent EGFR
amplifications (log
2
ratio>3) at 7p12, gain of 8q24 in
one sample and gain of 12q13–15 in one sample.
Experimental modeling of glioblastoma
Evident advantages of two-dimensional monolayer cell
systems for experimental studies are ease of culture, low
Figure 2 Comparison of genomic profiles in time of primary cell and spheroid cultures for three glioblastoma samples. (a–c) Genomic
profiles of the parental tumor, primary cell and spheroid culture ordered by culture time in weeks (w). The short-term culture profiles at
2 weeks are similar to those shown in Figure 1a, and same scaling, numbering and color coding were used. (d) Bar chart comparing the
Pearson’s correlation coefficients of genomic profiles of primary cell and spheroid cultures with the parental tumor. Numbers 2, 6 and
12 below the bars refer to the culture time in weeks.
Genomic profiles of glioblastoma in culture
PC De Witt Hamer et al
2094
Oncogene

expense and prompt and reproducible results on a large
scale. Crucial, however, for studies involving therapeutic
screening is whether tumor biology is adequately
portrayed by this model system. Recently, significant
genomic rearrangements in early passages (up to p10) of
primary cell culture were reported based on single
nucleotide polymorphism analysis in one glioblastoma
sample (Lee et al., 2006). In the present study, we
demonstrate genetic discordance between the parental
tumor and derived primary cell cultures to be frequent
and to appear early, already after 2 weeks of culture.
Therefore, primary cell cultures inappropriately reflect
glioblastoma genetically. Although we provide no direct
evidence that the genetic discordance results in altered
therapeutic responses, it emphasizes the invalid repre-
sentation of tumor biology by primary cell cultures. This
was also shown by alterations in ploidy status of glioma
cell cultures at later passages (Bigner et al., 1987). In
contrast, several studies report concordant cytogenetics
in cell culture and parental tissue (Izumoto et al., 1995;
Hartmann et al., 1999). We suspect that the discrepancy
with our current study is a result of either the low
resolution of techniques such as ploidy analysis and
karyotyping or the limited number of target genes. The
organotypic spheroid model appears to be more suitable
for therapeutic screening on a large scale, since genetic
alterations are better preserved.
There are several potential explanations for the
preservation of genomic profiles in spheroids and the
progressive genetic changes in primary cell cultures.
First, subpopulations may have been selected during
primary cell culture, due to culture conditions and
technical procedures, such as trypsinization, the serum
type or the affinity for plastic. The existence of cell
subpopulations is supported by considerable genetic
heterogeneity observed within glioblastomas and glioma
cell cultures (Hartmann et al., 1999; Steilen-Gimbel
et al., 1999; Loeper et al., 2001). Second, the divergent
clonal evolution may have been accelerated during
primary cell culture, because cell proliferation is more
pronounced in monolayer cell cultures than in spheroid
cultures. Third, the clonal evolution may be more
divergent in primary cell culture, because the progenitor
cells or tumor stem cells that presumably drive the cancer
cell population were possibly lost early in primary cell
culture and maintained in spheroid culture. This
assumption is supported by preliminary data showing
stem cell marker (CD133)-positive immunostaining in
frozen sections of spheroids after long-term culture (data
not shown). However, the continuing debate on cancer
stem cell detection in glioma precludes convincing proof
of this assumption. Selection of subpopulations at
initiation of culture due to microheterogeneity in the
original tumor sample is not a very likely explanation,
because fragments of tumors used for primary cell and
spheroid cultures were carefully homogenized. The exact
nature of the genetic instability of cells in culture is not
known, but it casts doubt on the representation of
glioblastoma biology by cultured glioma cells.
In summary, the present study demonstrates genomic
profiles of primary cell cultures that are deviant from
their parental tumors within the limitations of array
CGH analysis. Characteristic genetic abnormalities are
preserved in organotypic spheroid cultures until at least
12 weeks. Therefore, spheroids are genetically a more
representative model for human glioblastoma. These
observations imply that the relevance of primary cell
cultures for therapeutic screening studies, even after
short-term culture, should be seriously questioned.
Acknowledgements
Mr RH Wessel, Mr T Krugers, Mr A Mrs
ˇ
ic¸, Mrs M Tijssen
and Mrs W Tigchelaar are kindly acknowledged for their
skillful technical assistance, and the Departments of Neuro-
surgery of the Academic Medical Center from the University
of Amsterdam and of the VU University Medical Center,
for providing tumor material. We thank Mrs TMS Pierik for
her excellent assistance with preparation of the manuscript
format.
References
Autio R, Hautaniemi S, Kauraniemi P, Yli-Harja O, Astola J, Wolf M
et al. (2003). CGH-Plotter: MATLAB toolbox for CGH-data
analysis. Bioinformatics 19: 1714–1715.
Barrett T, Suzek TO, Troup DB, Wilhite SE, Ngau WC, Ledoux P
et al. (2005). NCBI GEO: mining millions of expression profiles—
database and tools. Nucleic Acids Res 33: D562–D566.
Bigner SH, Mark J, Bigner DD. (1987). Chromosomal progression of
malignant human gliomas from biopsy to establishment as
permanent lines in vitro. Cancer Genet Cytogenet 24: 163–176.
Bjerkvig R, Tonnesen A, Laerum OD, Backlund EO. (1990). Multi-
cellular tumor spheroids from human gliomas maintained in organ
culture. J Neurosurg 72: 463–475.
Burton EC, Lamborn KR, Feuerstein BG, Prados M, Scott J, Forsyth
P et al. (2002). Genetic aberrations defined by comparative genomic
hybridization distinguish long-term from typical survivors of
glioblastoma. Cancer Res 62: 6205–6210.
Cairncross JG, Ueki K, Zlatescu MC, Lisle DK, Finkelstein DM,
Hammond RR et al. (1998). Specific genetic predictors of
chemotherapeutic response and survival in patients with anaplastic
oligodendrogliomas. J Natl Cancer Inst 90: 1473–1479.
Eisen MB, Spellman PT, Brown PO, Botstein D. (1998). Cluster
analysis and display of genome-wide expression patterns. Proc Natl
Acad Sci USA 95: 14863–14868.
Hartmann C, Kluwe L, Lucke M, Westphal M. (1999). The rate of
homozygous CDKN2A/p16 deletions in glioma cell lines and in
primary tumors. Int J Oncol 15: 975–982.
Izumoto S, Arita N, Ohnishi T, Hiraga S, Taki T, Hayakawa T. (1995).
Homozygous deletions of p16INK4A/MTS1 and p15INK4B/MTS2
genes in glioma cells and primary glioma tissues. Cancer Lett 97:
241–247.
Koschny R, Koschny T, Froster UG, Krupp W, Zuber MA. (2002).
Comparative genomic hybridization in glioma: a meta-analysis of
509 cases. Cancer Genet Cytogenet 135: 147–159.
Kotliarov Y, Steed ME, Christopher N, Walling J, Su Q, Center A
et al. (2006). High-resolution global genomic survey of 178 gliomas
reveals novel regions of copy number alteration and allelic
imbalances. Cancer Res 66: 9428–9436.
Lee J, Kotliarova S, Kotliarov Y, Li A, Su Q, Donin NM et al. (2006).
Tumor stem cells derived from glioblastomas cultured in bFGF
and EGF more closely mirror the phenotype and genotype of
Genomic profiles of glioblastoma in culture
PC De Witt Hamer et al
2095
Oncogene

primary tumors than do serum-cultured cell lines. Cancer Cell 9:
391–403.
Loeper S, Romeike BF, Heckmann N, Jung V, Henn W, Feiden W
et al. (2001). Frequent mitotic errors in tumor cells of genetically
micro-heterogeneous glioblastomas. Cytogenet Cell Genet 94: 1–8.
Pandita A, Aldape KD, Zadeh G, Guha A, James CD. (2004).
Contrasting in vivo and in vitro fates of glioblastoma cell
subpopulations with amplified EGFR. Genes Chromosomes Cancer
39: 29–36.
Smeets SJ, Braakhuis BJ, Abbas S, Snijders PJ, Ylstra B, van de Wiel MA
et al. (2006). Genome-wide DNA copy number alterations in head and
neck squamous cell carcinomas with or without oncogene-expressing
human papillomavirus. Oncogene 25: 2558–2564.
Snijders AM, Nowak N, Segraves R, Blackwood S, Brown N, Conroy J
et al. (2001). Assembly of microarrays for genome-wide measure-
ment of DNA copy number. Nat Genet 29: 263–264.
Steilen-Gimbel H, Steudel WI, Feiden W, Moringlane JR, Henn W,
Zang KD. (1999). Genetic heterogeneity in human astrocytomas:
spatial distribution of P16 and TP53 deletions in biopsies. Cancer
Genet Cytogenet 113: 115–119.
Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJ
et al. (2005). Radiotherapy plus concomitant and adjuvant temozolo-
mide for glioblastoma. N Engl J Med 352: 987–996.
Sutherland RM. (1988). Cell and environment interactions in tumor
microregions: the multicell spheroid model. Science 240: 177–184.
Voskoglou-Nomikos T, Pater JL, Seymour L. (2003). Clinical
predictive value of the in vitro cell line, human xenograft,
and mouse allograft preclinical cancer models. Clin Cancer Res 9:
4227–4239.
Wolff JE, Trilling T, Molenkamp G, Egeler RM, Jurgens H. (1999).
Chemosensitivity of glioma cells in vitro: a meta analysis. J Cancer
Res Clin Oncol 125: 481–486.
Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc).
Genomic profiles of glioblastoma in culture
PC De Witt Hamer et al
2096
Oncogene