Supplementary Data

advertisement
Supplementary Material
A hypoxic niche regulates glioblastoma stem cells through hypoxia inducible
factor 2
Supplementary Materials and Methods
Isolation and cultivation of primary glioblastoma cell lines
Primary glioblastoma cell lines (designated GBMxxx) were obtained from glioblastoma
samples from 7 patients, which included 2 males and 5 females aged between 56 and 75
and diagnosed with glioblastoma multiforme; the GBM015 line was derived from a patient
with glioblastoma multiforme relapse. Histological diagnosis was confirmed by a
neuropathologist (TA). Tumor specimens were dissociated using a 0.01% Papain
(Worthington Biochemical), 0.01% DNase DMEM-F12 solution and filtered by 70µm cell
strainer to remove tissue pieces. The isolated tumor cells were propagated as tumor spheres
in serum-free DMEM-F12 supplemented with B27 (Gibco) and 20 ng/ml bFGF and EGF. For
co-culture experiments glioma and HUVEC cells were incubated on a ThinCert 6-well TC
Inserts (0.4 µm pore size, Greiner Bio-One, Solingen, Germany) separated by a membrane
for the indicated time points and harvested for RNA isolation or secondary sphere formation.
Flow cytometry and side population analysis
Identification and isolation of SP in glioblastomas was done as previously described (Goodell
et al., 1996). Briefly, single cell suspensions were obtained from tumor spheres by trypsin
(Gibco) incubation. Following dissociation, viable cells were counted with trypan blue and
incubated with the fluorescent dye Hoechst 33342 (2.5-5 µg/ml final concentration) at a
concentration of 1x106 cells/ml in prewarmed DMEM for 90 min at 37 °C with intermittent
mixing. Addition of verapamil (50 µM), an inhibitor of some ABC transporters, to a small cell
aliquot served to control for the specificity of the SP. Cells were counterstained with
1
propidium iodide (2 µg/ml) to identify dead cells. Cell analysis and sorting was performed on
a triple-laser sorter (FACSVantage, Becton Dickinson). After excitation of the Hoechst dye at
350 nm the fluorescence emission was collected in dual wavelength analysis with a 450/20
nm BP filter (Hoechst blue) and with a 675 nm LP filter (Hoechst red) using a 610 nm DMSP
filter to separate the emission wavelengths. Propidium iodide positive cells were excluded
from the analysis. The SP was defined as described in (Goodell et al., 1996). Murine bone
marrow was used a positive control. In experiments where sorted cells were used for further
in vitro or in vivo analysis, cells were washed in DMEM to remove excess Hoechst dye.
Tumor specimens were disaggregated and dissociated using a 0.01% Papain (Worthington
Biochemical), 0.01% DNase solution as described (Schänzer et al., 2004). For CD133 FACS
analysis the CD133/2 (293C)-PE conjugated antibody (Miltenyi) was used.
Primary glioma tissues
Frozen tissue specimens from gliomas of 115 patients were selected from the tumor tissue
collection of the Department of Neuropathology, Heinrich Heine University, Düsseldorf,
Germany, and investigated in an anonymized manner as approved by the institutional review
board. All tumors were classified according to the criteria of the World Health Organization
(WHO) classification of tumors of the nervous system of 2000, which in case of diffuse
astrocytic gliomas are retained in the revised WHO classification of 2007 (Louis et al., 2007).
The tumor series included 73 primary glioblastomas (WHO grade IV), 10 secondary
glioblastomas (WHO grade IV), 13 anaplastic astrocytomas (WHO grade III) and 19 diffuse
astrocytomas (WHO grade II). Parts of each tumor were snap-frozen immediately after
operation and stored at -80°C. Only tissue specimens with a histologically estimated tumor
cell content of 80% or more were used for molecular analysis. Clinical follow-up data
concerning overall survival were available for 68 primary glioblastoma patients. Commercially
available RNA samples from adult brain (3 different samples) and fetal brain (1 sample) were
obtained from Clontech (Mountain View, CA), Stratagene (La Jolla, CA) and BioChain
(Hayward, CA), respectively.
2
Intracranial tumor xenografts and analysis
Xenograft transplantations were performed in athymic BALB/c nu/nu mice. Animals were
obtained from Charles River and maintained in standard conditions according to the
institutional guidelines. SP and non-SP from bFGF/LIF treated cell cultures were isolated by
FACS sorting. 1000 SP or non-SP cells, respectively, in a volume of 2 µl were
stereotactically implanted intracranially into the left striatum. After the onset of neurological
symptoms all mice for the respective experiment were sacrificed at the same time point.
Brains of euthanized mice were collected, fixed in 4% paraformaldehyde, dehydrated in
increasing concentrations of sucrose and rapidly frozen on dry ice for sectioning with a
cryotome. Free floating sections were stained with hematoxylin and eosin. Tumor volume
was determined using stereological quantification of series of every sixth section (240 µm
intervals) throughout the brain tracing the tumor area with a semi-automatic stereology
system (MicroBrightField).
cDNA sample preparation
For gene expression profiling studies total RNA was extracted from the sorted cell
populations using the RNeasy mini spin columns (Qiagen) with RNase inhibitor (Invitrogen)
supplemented to the final eluate (for overview of study design see Fig. 2A). Two cycles of
linear T7-based in vitro transcription (IVT) (RiboAmp, Arcturus) using a 3-hour incubation at
both IVT rounds was used to amplify the RNA. Labeling of the amplified RNA (aRNA) was
carried out using the standard operating procedures available on-line at www.ktharray.se and
through ArrayExpress microarray data repository (Parkinson et al., 2005). Briefly, for each
labeling 1000 ng of aRNA was reverse transcribed to cDNA for 2 hours at 46 °C using 400 U
of Superscript III (Invitrogen), 5 µg random hexamers and nucleotides containing a 4:1 ratio
of aminoallyl-modified dUTP:unmodified dTTPs. After hydrolysis of the RNA strand,
monofunctional Cy3 and Cy5 esters were coupled to the aminoallyl groups of the purified
cDNA (MinElute purification system, Qiagen). Quantification of labeled cDNA was carried out
3
and equal amounts (in pmoles) of Cy3 labeled reference sample and Cy5 labeled cDNA
were used for each hybridization.
qPCR. Total RNA from cultured cells was harvested using an RNeasy mini kit (Qiagen,
Hilden, Germany). 1 µg of total RNA was transcribed into cDNA, using RevertAid H Minus MMuLV Reverse Transcriptase (MBI Fermentas, St. Leon-Rot, Germany) and diluted to 100
ng/µl for further use in qPCR. From primary glioma tissue, total RNA was extracted by
ultracentrifugation over caesium chloride and transcribed into cDNA as described (van den
Boom et al., 2003). qPCR reactions were performed using ABsolute™ QPCR SYBR Green
Mix (ABgene, Epsom, UK) and the iCycler iQ System (Bio-Rad, Munich, Germany).
Reactions were performed in triplicates of 25 µl total volume, containing 1 µl of 100 ng/µl
cDNA (cell lines) or 2 µl of cDNA (tumor samples). The amount of target mRNA was
determined using the comparative cycle threshold method and normalized relative to the
amount of HPRT mRNA. To compensate for variability between the qPCR runs with the
tumor samples, a cDNA from LN229 spheres (grown with 20 ng/ml bFGF and LIF,
PeproTech, Hamburg, Germany) was used as a reference in every run and all the sample
data was normalized to the mean of the LN229 Ct values.
Microarray hybridization
All probes used for the microarray were annotated according to UniGene. The Unigene
database is updated regularly, which is always associated with some changes for the ESTto-Unigene assignments. This is more pronounced for poorly known ESTs, which frequently
change Unigene cluster ID assignment between Unigene builds. Therefore, we have listed
the Genbank accession number and UniGene annotation for each probe on our array to
facilitate future analysis and comparison with our data, (Table S1). Also, R-based tools for
the re-annotation of the probes have been made available in the KTH-package. A total of 49
hybridizations were carried out using a common-reference design (Fig. 2A, S4B). Each
sample was amplified once and, to allow for comparison between cell lines and growth factor
4
combinations, hybridized against a universal reference RNA sample (Stratagene) in the Cy3
channel. To estimate the reproducibility of the experimental platform and to reduce the
technical variability, most samples were analyzed using a pair of replicated hybridizations,
depicted by double arrows in Fig. 2A or S4B.
Microarray data analysis
Data analysis was carried out in the R environment for statistical computing using the
Bioconductor, aroma and KTH-packages (Gentleman et al., 2004). Median intensity values
for each array feature were used to derive a gene expression measurement for each probe.
A print-tip lowess normalization (Smyth and Speed, 2003) was carried out using the
complete unfiltered data for each slide. Differentially expressed (DE) genes were identified
using an empirical Bayes moderated t-test in the Limma package (Smyth, 2004). Genes that
were DE in at least one of the cell lines and growth factor combinations (contrasts) were
identified using a moderated F-test and a false-discovery rate (fdr) adjustment of the pvalues. Genes with an F-test associated and fdr-adjusted p-value < 0.05 were considered
significantly differently expressed and included in the further analysis. Positive M-values
always indicate for genes more highly expressed in the SP.
Hierarchical clustering was carried out using correlation as distance measurement and
average linkage. Hierarchical clustering analysis of the 265 probes DE in at least two of the
cell lines showed a distinct separation of all samples into two groups, SP and non-SP, each
further separated according to each cell line (Fig. S5). The Pearson’s correlation coefficient
was used for calculation of correlations between hybridizations. Theme enrichment analysis
was carried out for the ‘Biological processes’ branch of the Gene Ontology (Ashburner et al.,
2000), Biocarta, and KEGG (Kanehisa and Goto, 2000) pathway databases using Fisher’s
Exact tests available in the Bioconductor and kth-packages. For Biocarta databases the
analysis was carried out using the genes that were found as DE in any of the three cell lines.
For the KEGG and Gene Ontology analysis the list of genes was restricted to those found as
DE in at least two of the cell lines. Themes and pathways with at least four (Gene Ontology
5
analysis) and three (Biocarta and KEGG) genes in the list of genes more highly expressed in
the SP were included in the analysis. The nominal p-values were adjusted for multiple
hypotheses testing using the fdr-approach.
Statistical analysis.
Results are presented as mean  standard error of the mean (SEM). Statistical comparisons
of values were made using the Student’s t-test or Mann-Whitney U-test followed by a posttest for linear trend. Statistical significance was assumed at p < 0.05. For microarray data
analysis see Supplemental Materials and Methods. Survival correlations were performed by
univariate analysis using Kaplan-Meier survival curve estimation and log-rank tests.
6
Supplementary Figure Legends
Figure S1: Glioblastoma cell lines grow similarly to NSCs in culture and express
progenitor and multilineage markers
A: The glioblastoma cell lines G55, G142 and LN229 form adherent cell layers in the
presence of 10% FCS, but round up and form tumor spheres in serum-free medium,
supplemented with B27, similarly to NSCs. In neurosphere differentiating conditions (addition
of 1% FCS), the tumor spheres from all cells lines adhere and spread out, with individual
cells forming cell protrusions; B: Adherent tumor spheres reveal heterogeneous expression
of various stem cell and differentiation markers (stem cell: Nestin; glial: GFAP, S100β;
neural: βIII-Tubulin (Tuj1), Doublecortin (Dcx)). Interestingly, despite the astrocytic
differentiation of glioblastomas some tumor cells show co-expression of GFAP with Tuj1,
likely reflecting an aberrant differentiation program in tumor cells. Scale bar: 100 µm.
Figure S2: All glioblastoma cell lines tested contain an SP with stem cell
characteristics
A: Eight additional glioblastoma cell lines (G142, G55, U118, U87, U373, U251, U251-A and
LN229) were incubated with Hoechst 33342 and analyzed by dual wavelength FACS
(HoechstRed on the x axis, HoechstBlue on the y axis). All cell lines show a small
subpopulation of low fluorescence (left lower quadrant, green) with the SP phenotype,
analogously to the U343 line shown in Fig. 1C. This SP is depleted after treatment with
verapamil. B: SP and non-SP were isolated by FACS sorting from the G142 and G55
glioblastoma cell lines, grown separately for two weeks in culture and reassessed for their
ability to form SP and non-SP by flow cytometry, analogously to the U343 line shown in Fig.
1D. The SP displays the capacity to re-generate SP and non-SP, whereas the non-SP
demonstrates no (G142) or a greatly reduced (G55) capacity to form SP.
7
Figure S3: Design and analysis of the gene expression profiling study
A: Overview of the target preparation approach. B: Sample hybridization scheme. The
number of FACS purifications carried out for each condition and cell line, respectively, varied
from three (G55FL and G55PFL) to four (LN229FL and U343FL). The cDNA of sorted cell
samples was labeled in Cy5 (head of arrow), and compared to a common reference sample
labeled in Cy3 (tail of arrow). Each arrow represents one hybridization. (1k = 1000 cells:
number of cells used for RNA isolation). C: On the global gene expression level the 49
hybridizations form three distinct groups. In the analysis based on the complete data set and
Pearson’s correlation between sample-to-reference ratio values all hybridizations with
samples originating from the G55 cell line show a higher correlation with each other than with
samples from either the LN229 or the U343 cell line. The same trend is observed for the
other two cell lines. D: Hierarchical clustering, using the correlation as distance measure and
average linkage, generates a similar grouping, with replicated hybridizations clustering tightly
together. Black and red squares correspond to SP and non-SP samples, respectively; yellow
and blue squares represent FL and PFL growth factor combinations, respectively.
Figure S4: Gene Ontology diacyclic graph of themes enriched in SP
Genes with higher expression in the SP were categorized in 37 themes using the Gene
Ontology enrichment test (Table S7). The theme networks are visualized in a diacyclic graph
structure generating four distinct branches that are related to various biological processes.
Functional themes that are overrepresented in the SP are shown in cyan/blue color. Themes
in purple color do not reach statistical significance.
Figure S5: SP signature expressing tumor cells reside within distinct niches
A: Immunohistochemical analysis demonstrates that tumor cells expressing SP signature
genes (AHSA1, GLCCI1, LAMC1) are located in perivasular and perinecrotic areas (see
insets for higher magnifications). V= Vessel, N= Necrosis. Scale bars: 50 µm and 5 µm
(insets). B: Endothelial cells are an important component of the vascular niche. Co8
incubation of glioblastoma cells G55 and ED010 with HUVECs in transwell inserts, allowing
free diffusion of secreted factors without direct cell-to-cell contact, increases the expression
of SP signature genes and CD133, as determined by qPCR analysis (n=3).
Figure S6: siRNA knockdown of HIF-1 and HIF-2
Transfection of siRNA against HIF-1 or HIF-2a in GBM010 cells under hypoxia (1% O2) led
to reduction of the respective HIF isoform, as assessed by qPCR.
9
Supplementary Table Legends
Table S1: List of DE genes in all three cell lines
The list contains all DE genes that are shared by all three cell lines (G55, U343, LN229)
including both growth factor combinations (G55FL, G55PFL).
Table S2: List of SP genes that overlap with previously published NSC or SP
signatures
We performed a meta-analysis comparing our SP signature genes with NSC or SP signature
genes previously identified in other studies. The stem cell populations in the respective
papers that were used for the meta-analysis were derived from the following cell types:
Benchaouir et al. (Benchaouir et al., 2004) – SP from the mouse myogenic C2C12 cell line;
Decraene et al. (Decraene et al., 2005) – SP from the C2C12 cell line; Fortunel et al.
(Fortunel et al., 2003) – neural stem cells (NSCs); Larderet et al. (Larderet et al., 2006) – SP
from primary human breast skin keratinocytes; (Liadaki et al., 2005) – SPs from primary
mouse bone marrow (BM) and skeletal muscle (M) cells; Rochon et al. (Rochon et al., 2006)
– common SP genes between primary mouse bone marrow cells, primary adult germinal
cells, primary muscle cultures and the stromal cell line BMC9. MP – main (non-SP)
population.
Table S3: Pathway and biological process enrichment analysis
Using Fisher’s Exact test we identified pathways that were significantly enriched in genes
more highly expressed in the SP, based on the Biocarta database (http://www.biocarta.com)
or
the
KEGG
database
(Kyoto
Encyclopedia
of
Genes
and
Genomes;
http://www.genome.jp/kegg/). Biological processes involving genes with higher expression in
10
the
SP
were
identified
based
on
the
Gene
Ontology
database
(http://www.geneontology.org/).
11
Supplementary References
Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, et al. Gene ontology: tool
for the unification of biology. The Gene Ontology Consortium. Nat Genet. 2000
May;25(1):25-9.
Beck H, Acker T, Wiessner C, Allegrini PR, Plate KH. Expression of angiopoietin-1,
angiopoietin-2, and tie receptors after middle cerebral artery occlusion in the rat. Am J
Pathol. 2000 Nov;157(5):1473-83.
Benchaouir R, Rameau P, Decraene C, Dreyfus P, Israeli D, Pietu G, et al. Evidence for a
resident subset of cells with SP phenotype in the C2C12 myogenic line: a tool to explore
muscle stem cell biology. Exp Cell Res. 2004 Mar 10;294(1):254-68.
Decraene C, Benchaouir R, Dillies MA, Israeli D, Bortoli S, Rochon C, et al. Global
transcriptional characterization of SP and MP cells from the myogenic C2C12 cell line:
effect of FGF6. Physiol Genomics. 2005 Oct 17;23(2):132-49.
Fortunel NO, Otu HH, Ng HH, Chen J, Mu X, Chevassut T, et al. Comment on " 'Stemness':
transcriptional profiling of embryonic and adult stem cells" and "a stem cell molecular
signature". Science. 2003 Oct 17;302(5644):393; author reply
Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S, et al. Bioconductor:
open software development for computational biology and bioinformatics. Genome Biol.
2004;5(10):R80.
Goodell MA, Brose K, Paradis G, Conner AS, Mulligan RC. Isolation and functional
properties of murine hematopoietic stem cells that are replicating in vivo. J Exp Med. 1996
Apr 1;183(4):1797-806.
Kanehisa M, Goto S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res.
2000 Jan 1;28(1):27-30.
Larderet G, Fortunel NO, Vaigot P, Cegalerba M, Maltere P, Zobiri O, et al. Human side
population keratinocytes exhibit long-term proliferative potential and a specific gene
12
expression profile and can form a pluristratified epidermis. Stem Cells. 2006
Apr;24(4):965-74.
Liadaki K, Kho AT, Sanoudou D, Schienda J, Flint A, Beggs AH, et al. Side population cells
isolated from different tissues share transcriptome signatures and express tissue-specific
markers. Exp Cell Res. 2005 Feb 15;303(2):360-74.
Louis DN, Ohgaki H, Wiestler OD, Cavenee WK. Astrocytic Tumors. WHO Classification of
Tumours of the Central Nervous System. 4th ed. ed: Lyon: IARC Press; 2007. p. 25-49.
Parkinson H, Sarkans U, Shojatalab M, Abeygunawardena N, Contrino S, Coulson R, et al.
ArrayExpress--a public repository for microarray gene expression data at the EBI. Nucleic
Acids Res. 2005 Jan 1;33(Database issue):D553-5.
Rochon C, Frouin V, Bortoli S, Giraud-Triboult K, Duverger V, Vaigot P, et al. Comparison of
gene expression pattern in SP cell populations from four tissues to define common
"stemness functions". Exp Cell Res. 2006 Jul 1;312(11):2074-82.
Schänzer A, Wachs FP, Wilhelm D, Acker T, Cooper-Kuhn C, Beck H, et al. Direct
stimulation of adult neural stem cells in vitro and neurogenesis in vivo by vascular
endothelial growth factor. Brain Pathol. 2004 Jul;14(3):237-48.
Smyth GK, Speed T. Normalization of cDNA microarray data. Methods. 2003 Dec;31(4):26573.
Smyth GK. Linear models and empirical bayes methods for assessing differential expression
in microarray experiments. Stat Appl Genet Mol Biol. 2004;3:Article3.
van den Boom J, Wolter M, Kuick R, Misek DE, Youkilis AS, Wechsler DS, et al.
Characterization of gene expression profiles associated with glioma progression using
oligonucleotide-based microarray analysis and real-time reverse transcription-polymerase
chain reaction. Am J Pathol. 2003 Sep;163(3):1033-43.
13
Download