Advertisement

Neurochemical Research

, Volume 40, Issue 2, pp 336–352 | Cite as

Classification of Subpopulations of Cells Within Human Primary Brain Tumors by Single Cell Gene Expression Profiling

  • Elin Möllerström
  • Bertil Rydenhag
  • Daniel Andersson
  • Isabell Lebkuechner
  • Till B. Puschmann
  • Meng Chen
  • Ulrika Wilhelmsson
  • Anders Ståhlberg
  • Kristina Malmgren
  • Milos Pekny
Overview

Abstract

Brain tumors are heterogeneous with respect to genetic and histological properties of cells within the tumor tissue. To study subpopulations of cells, we developed a protocol for obtaining viable single cells from freshly isolated human brain tissue for single cell gene expression profiling. We evaluated this technique for characterization of cell populations within brain tumor and tumor penumbra. Fresh tumor tissue was obtained from one astrocytoma grade IV and one oligodendroglioma grade III tumor as well as the tumor penumbra of the latter tumor. The tissue was dissociated into individual cells and the expression of 36 genes was assessed by reverse transcription quantitative PCR followed by data analysis. We show that tumor cells from both the astrocytoma grade IV and oligodendroglioma grade III tumor constituted cell subpopulations defined by their gene expression profiles. Some cells from the oligodendroglioma grade III tumor proper shared molecular characteristics with the cells from the penumbra of the same tumor suggesting that a subpopulation of cells within the oligodendroglioma grade III tumor consisted of normal brain cells. We conclude that subpopulations of tumor cells can be identified by using single cell gene expression profiling.

Keywords

Astrocytoma grade IV Oligodendroglioma grade III Glioblastoma multiforme Single cell gene expression profiling 

Introduction

Quantitative measurement of mRNA isolated from non-dissociated tissue reflects average mRNA expression levels within the tissue and thereby masks variation in gene expression between individual cells. The gene expression of a specific group of cells, in particular when these constitute a minority subpopulation, can be hidden in the noise of other cells within the tissue [1, 2]. In contrast, single cell gene expression profiling using reverse transcription quantitative PCR (RT-qPCR) is a method to analyze gene expression at the level of individual cells and allows to obtain their true molecular signature.

Until recently, analyses of single cells have usually been performed by using imaging techniques such as immunocytochemistry, immunohistochemistry, FISH, or by flow cytometry. These techniques assess morphological properties, DNA copy numbers or largely qualitative expression of specific proteins [3]. By implementing single cell gene expression profiling, it is possible to detect, measure and correlate the expression of a larger number of genes within individual cells. Single cell gene expression profiling reveals molecular properties of the cells and complements as well as substantially expands histological classifications of cells and tissues. Single cell gene expression profiling starts to be implemented in the characterization of several types of malignant tumors, for example colon cancer [4], circulating melanoma cells, or breast cancer cells [5, 6]. We previously applied this technique on primary astrocyte cultures and neurosphere cells and it allowed us to identify distinct subpopulations of astrocytes and to make interaction maps for the assessed genes [7].

In 2011, 1 349 patients in Sweden were diagnosed with brain tumors or tumors in other parts of the nervous system, which corresponds to 2.5 % of total cancer incidence [8]. The majority of these were gliomas [8]. Gliomas arise from either neural stem/progenitor cells or glial cells within the nervous system and astrocytomas and oligodendrogliomas are the most common subtypes. The survival varies widely depending on the tumor grade [9]. Oligodendrogliomas are generally divided into grade II and grade III, where grade II patients have a median survival time of 11.6 years and grade III patients 3.5 years. When it comes to astrocytomas, 96 % of patients with grade I astrocytomas survive for more than 10 years from diagnosis, whereas median survival time for astrocytomas grade IV (glioblastoma multiforme) patients is only 0.4 years. Unfortunately, astrocytomas grade IV account for 69 % of all astrocytomas and oligodendrogliomas [9], and therefore most glioma patients have a low long-term survival. Individual gliomas are known to be very heterogeneous, both genetically and histologically, and with respect to the expression of specific molecules [10, 11, 12]. The existence of self-renewing cancer stem cells, responsible for the tumor initiation and progression [13, 14] has been postulated and it is possible that many of the cells within a tumor are tumor supporting cells rather than highly proliferative tumor cells. Bonavia et al. [15] compare an astrocytoma grade IV tumor to a community where all different types of cells contribute to create a tumor-beneficial environment within the cancer tissue.

In this study, we developed a protocol for obtaining molecular signature of individual cells from fresh adult human brain tissue, using single cell gene expression analysis of selected genes. This was used to assess the molecular differences and similarities between cell populations within brain tumors and the tumor penumbra tissue.

Methods

Tumor Tissue

Tumor and tumor penumbra tissue was collected during surgery from a male, 49 years old, diagnosed with astrocytoma grade IV (tumor tissue) and a male, 63 years old, diagnosed with oligodendroglioma grade III (tumor tissue and tumor penumbra; Fig. 2).

Ethics Statement

The patients’ written informed consent was obtained and tumor tissue was collected in accordance with the ethical guidelines of the Sahlgrenska Academy University Hospital (approved by the Regional Ethics Board, registration number 179-08, 2008-06-12).

Tissue Dissociation

The tissue specimens were collected at the surgical theatre and kept in ice cold Hibernate A medium (Brain Bits, Springfield, Illinois, USA) for approximately 20 min before enzymatic and mechanic dissociation. To dissociate the tissue into single cells, Brain tumor dissociation kit (Trypsin) from Miltenyi Biotech (Cologne, Germany) was used according to the manufacturer’s protocol with minor modifications (plastic 1 ml tips were used instead of fire-polished Pasteur-pipettes). Briefly, the tissue was cut into small pieces, incubated with enzymes at 37 °C and mechanically dissociated via titration, and then passed through a 70 µm cell strainer. To remove red blood cells and myelin debris the single cell suspension was treated with Pharmlyse (Becton–Dickinson, Franklin Lakes, New Jersey, USA) followed by magnetic associated cell sorting (MACS) using Myelin removal beads II (Miltenyi Biotech) according to the manufacturers´ protocols. Briefly, cells were washed with Pharmlyse, followed by a wash with MACS buffer (PBS containing 0,5 % BSA, autoMACS rinsing solution and MACS BSA stock solution 20:1, Miltenyi Biotech). Then, the cell suspension was incubated with magnetic beads connected to myelin antibodies, passed through a magnetic field where the myelin debris remained and the flow-through was collected, leaving a suspension of viable single cells. The cells were kept in MACS buffer on ice until sorted by FACS.

Collection of RNA from Individual Cells

To place single cells in separate wells of a 96-well plate (Applied Biosystems, Life Technologies, Carlsbad, California, USA), the flow-cytometric cell sorting function of a FACS Aria II (Becton–Dickinson) was used. The cells were sorted into wells containing 5 µl Ultrapure water (Gibco, Life Technologies, Carlsbad, California, USA) with 1 µg/µl bovine serum albumin (BSA; Fermentas, Thermo Fisher Scientific, Waltham, Massachusetts, USA). This immediately lysed the cells and made the RNA available for further experiments. The plate containing lysed single cells was snap frozen on dry ice and stored at −80 °C until reverse transcription was performed.

Reverse Transcription (RT)

RT was carried out using SuperScript™ III Reverse Transcriptase (Invitrogen, Life Technologies, Carlsbad, California, USA). 1.5 µl suspension containing dNTPs (Sigma-Aldrich, Munich, Germany), oligo(dT) 15 primers and random hexamers (both Eurofins MWG Operon, Ebersberg, Germany) were added to the 5 µl of RNA suspension and the mixture was incubated at 65 °C for 5 min. Next, 3.5 µl suspension containing 5× SuperScript buffer, RNaseOut, dithiothreitol (DTT) and SuperScript III enzyme (all Invitrogen) was added and the plates were incubated at 25 °C for 5 min, 50 °C for 60 min, 55 °C for 15 min and 70 °C for 15 min using CFX 96 Real Time System (Bio-Rad, Hercules, California, USA). The final concentrations in a total reaction volume of 10 µl were 0.5 mM dNTPs, 2.5 μM oligo(dT) 15 primers, 2.5 μM random hexamers, 1× SuperScript buffer, 5 mM DTT, 10 U RNaseOut and 50 U Superscript III enzyme.

Preamplification

The primers for the genes of interest (Tables 1, 2) were designed using the Primer 3 software (http://frodo.wi.mit.edu/primer3/input.htm), tested for primer dimers in Netprimer (Premier Biosoft International, Palo Alto, California, USA) and checked for specificity using the BLAST algorithm (NCBI, NIH, Bethesda, Maryland, USA) and obtained from Sigma-Aldrich. They were designed to span introns when possible and expected qPCR product length was confirmed using agarose gelelectrophoresis (Invitrogen). For preamplification, 20 µl IQ supermix (Bio-Rad), 1.6 µl primerpool (containing 1 µM of each primer) and 14.4 µl Ultrapure H2O (Gibco) were added to 4 µl of cDNA for each sample in a total volume of 40 µl. Each 96-well plate contained three positive controls with cDNA extracted from solid tumor tissue and three negative controls containing H2O. The mixture was incubated at 95 °C for three minutes, followed by 14 cycles of 95 °C for 20 s, 55 °C for 3 min and 72 °C for 20 s, in a LightCycler 480 (Roche Diagnostics, Basel, Switzerland). The run was ended with a final step of 72 °C, during which the run was interrupted and the plate put on dry ice. The preamplified cDNA were kept cold while diluted 1:20 in TE-buffer (Invitrogen), before qPCR analysis.
Table 1

List of the assessed genes

Gene

Gene name

Function

AIF1

Allograft inflammatory factor 1

Cell M

ALDH1L1

Aldehyde dehydrogenase 1 family, member L1

Cell A

CCNE1

Cyclin E1

CC

CD63

CD63 molecule

BM

CDK4

Cyclin-dependent kinase 4

CC

CNP

2’,3’-Cyclic nucleotide 3’ phosphodiesterase

Cell O

EDNRB

Endothelin receptor type B

Cell A, BM

EGFR

Epidermal growth factor receptor

Prol

ENO2

Enolase 2 (gamma, neuronal)

Cell N

ERBB2

v-erb-b2 Avian erythroblastic leukemia viral oncogene homolog 2

BM

FN1

Fibronectin 1

Met

GFAP

Glial fibrillary acidic protein

Cell A

GFAPdelta

Glial fibrillary acidic protein delta

Cell A

GLUL

Glutamate-ammonia ligase (glutamine synthetase) 

Cell A

IL8

Interleukin 8

Cell M, Angio

MAPT

Microtubule-associated protein tau

Cell N

MBP

Myelin basic protein

Cell O

MCM2

Minichromosome maintenance complex component 2

Prol

MDM4

Mdm4 p53 binding protein homolog (mouse)

OG

MGMT

O-6-Methylguanine-DNA methyltransferase

TS

MKI67

Marker of proliferation Ki-67

Prol

NES

Nestin

Cell NS

NOTCH1

Notch 1

TS

OLIG2

Oligodendrocyte lineage transcription factor 2

BM

PCNA

Proliferating cell nuclear antigen

Prol

PECAM1

Platelet/endothelial cell adhesion molecule 1

Cell E, Angio

PROM1

Prominin 1 (CD133)

Cell CS, OG

RB1

Retinoblastoma 1

TS, CC

SEC61G

Sec61 gamma subunit 

BM, OG

SOX2

SRY (sex determining region Y)-box 2

Cell NS

SUZ12

SUZ12 polycomb repressive complex 2 subunit

BM

SYNM

Synemin, intermediate filament protein

Cell A, Cell SM, BM

TP53

Tumor protein p53

TS

TUBB3

Tubulin, beta 3 class III

Cell N, BM

VEGFA

Vascular endothelial growth factor A

Angio

VIM

Vimentin

Cell A

The genes were selected as cell type markers (Cell A astrocyte marker; Cell CS cancer stem cell marker; Cell E endothelial cell marker; Cell M microglia marker; Cell N neuronal marker; Cell NS neural progenitor marker; Cell O oligodendrocyte marker; Cell SM smooth muscle marker)

Prol proliferation markers, Angio angiogenesis markers, CC involved in cell cycle control, BM cancer biomarkers, OG proto-oncogenes, TS tumor suppressor genes, Met genes involved in tumor metastasis

Table 2

Primer sequences

Gene

NCBI accession number

Intron spanning

Forward primer

Reverse primer

Length (bp)

AIF1

NM_001623.3

yes

CTGTCTCCCCACCTCTACCA

AAGTTTCTCCAGCATTCGTTTC

265

ALDH1L1

NM_001270364.1

yes

CCTTCCAACCCTCCTGCTAC

CGGCACTCCATCCTTCTC

177

CCNE1

NM_001238.2

yes

CCGGTATATGGCGACACAAG

TACGCAAACTGGTGCAACTT

123

CD63

NM_001780.5

yes

ATCATCGCAGTGGGTGTCT

CGAAGCAGTGTGGTTGTTTT

234

CDK4

NM_000075.3

yes

CAGATGGCACTTACACCCGT

CAGCCCAATCAGGTCAAAGA

195

CNP

NM_033133.4

yes

CGCTCTACTTCGGCTGGTTC

GCCTGGGGGTCTCTTTCC

185

EDNRB

NM_000115.3

yes

AGCAAAAGATTGGTGGCTATTC

CAGAGGGCAAAGACAAGGAC

192

EGFR

NM_005228.3

yes

CCTTCACACATACTCCTCCTCTG

TCACATCTCCATCACTTATCTCCTT

246

ENO2

NM_001975.2

yes

AGGCTGGCTACACGGAAAA

ACATTGGCTGTGAACTTGGA

235

ERBB2

NM_004448.2

yes

ACCTGGAACTCACCTACCTG

TCACTTGGTTGTGAGCGATG

99

FN1

NM_212482.1

yes

GGAGACACCTGGAGCAAGAA

GCATCCCCACAGAGTAGACC

235

GFAP

NM_002055.4

yes

GCAGACCTTCTCCAACCTG

TGCCTCACATCACATCCTTG

161

GFAPdelta

NM_001131019.2

yes

CCGTGCAGACCTTCTCCAA

CGTATTGTGAGGCTTTTGAGATATCT

100

GLUL

NM_002065.5

yes

AGATTGCGGGGACTAATGC

TGGTGCTGAAGTTGGTATGG

210

IL8

NM_000584.3

yes

CCTGATTTCTGCAGCTCT

TTTGGGGTGGAAAGGTTT

102

MAPT

NM_016835.4

yes

CCTGGCGGAGGAAATAAAA

GCTGAGATGCCGTGGAGA

138

MBP

NM_001025081.1

yes

GATGAAAACCCCGTAGTCCA

TCCCTTGAATCCCTTGTGAG

185

MCM2

NM_004526.3

yes

CGGCGTGACAACAATGAG

CCCAAACCAGAATCCCAAG

287

MDM4

NM_002393.4

yes

CCAGAAAAAGAACTACAGAAGACGA

CTACATCCCACTCCTCAAATCC

150

MGMT

NM_002412.3

yes

CTGAATGCCTATTTCCACCA

ATTGCTCCTCCCACTGCTC

206

MKI67

NM_002417.4

yes

GAGAGGAGGGAGAAAGAGTGG

TTGGTTGGAAATGAAGTTGTTG

388

NES

NM_006617.1

no

GCTGGAGTGGAAAATGAGGA

ACCTGTTGTGATTGCCCTTC

236

NOTCH1

NM_017617.3

yes

GCCAGACCAACATCAACGAG

GGGGCAGACACAGGAGAAG

212

OLIG2

NM_005806.3

no

CCGTTCCTCCCTGTCTCTC

TCGGCAGTTTTGGGTTATTC

142

PCNA

NM_002592.2

yes

GTGGAGAACTTGGAAATGGAA

ACCGTTGAAGAGAGTGGAGTG

160

PECAM1

NM_000442.4

yes

TCAGAAGGACAAGGCGATT

GGATGGAGCAGGACAGGTT

233

PROM1

NM_006017.2

yes

CACTTACGGCACTCTTCACC

TCTATTCCACAAGCAGCAAAA

177

RB1

NM_000321.2

yes

AAAGGACCGAGAAGGACCA

AAGGCTGAGGTTGCTTGTGT

180

SEC61G

NM_014302.3

yes

GGGATTCATTGGCTTCTTTG

GAGTTTCTCACACCCTCACACTT

133

SOX2

NM_003106.3

no

ACACCAATCCCATCCACACT

CCTCCCCAGGTTTTCTCTGT

116

SUZ12

NM_015355.2

yes

AGCCATCACCAAACTCAGAAA

GCTTTTTACCTGTGGGAACTTG

126

SYN

NM_145728.2

no

TTCCCAACCCTCCATTCAT

GTTCCTTCCCCAAAACATCC

184

TP53

NM_000546.5

yes

CCCTTCCCAGAAAACCTACC

GCCTCACAACCTCCGTCA

233

TUBB3

NM_006086.3

yes

GTGCGGAAGGAGTGTGAAAA

CGGAAGCAGATGTCGTAGAG

281

VEGFA

NM_001025366.2

yes

CAGAAGGAGGAGGGCAGAA

TCAGGGGCACACAGGATG

153

VIM

NM_003380.3

yes

CAGATGCGTGAAATGGAAGA

TGGAAGAGGCAGAGAAATCC

222

Quantitative Real Time PCR (qPCR)

qPCR analysis was performed in 10 µl reactions, containing 2 µl diluted preamplified cDNA, 1× TATAA SYBR GrandMaster Mix (TATAA Biocenter, Gothenburg, Sweden) and a final primer concentration of 0.4 µM. Same primers were used as for preamplification (Table 2). The temperature was set to 95 °C for three minutes followed by 45 cycles of 95 °C for 5 s, 60 °C for 10 s and 72 °C for 15 s, followed by a melting curve step ranging from 60 °C to 95 °C. The qPCR experiments were performed in 384-well plates (4titude, Wotton, Surrey, UK) in a Viia7 instrument (Applied Biosystems).

Data Analysis and Statistics

The qPCR data was pre-processed using the GenEx Professional version 5.4.4 software (MultiD, Gothenburg, Sweden) according to the single cell data handling steps described by Ståhlberg et al. [16], with few modifications. In brief, elimination of false positives (based on melt curves and gel electrophoresis) was followed by interplate calibration using the mean of the three positive controls, gene expression above Cq 40 was set to Cq 40 and all negatives were set to Cq 41. The relative gene expression for each cell and gene was calculated by assigning one molecule to Cq 41, and thereafter log 2 transforming all the data. Cells expressing fewer than two of the selected genes were excluded from further analysis. The binary data were analysed by using Fisher’s exact test in PASW Statistics 18 (SPSS Inc., IBM, Armonk, New York, USA). Spearman correlations were calculated in the PASW software (IBM) using log scaled quantitative data including zeros for non-expressing cells and the genes with the highest number of correlations were compiled in correlation tables. Excel (Microsoft, Albuquerque, New Mexico, USA) was used to prepare diagrams showing percentages of cells expressing genes. Multivariate analyses and calculation of differences in gene expression were all performed in GenEx. Principal component analyses (PCA) [17], Kohonen Self Organizing Maps (SOMs) [18] were performed using autoscaled data, while hierarchical clusterings [19] with heat maps were performed using mean-centred data. For the SOMs, the following settings were used; 0.40 learning rate, 2 neighbours and 5,000 iterations and 2 groups. The cells separated into identical groups in repeated analyses. To calculate differences in gene expression levels between two groups, non-scaled log 2 transformed data was used and non-parametric Mann–Whitney tests [20] with Bonferroni correction for multiple testing [21] were performed.

Results

Development of a Protocol for Obtaining Viable Single Cells from Freshly Isolated Human Brain Tissue for Single Cell Gene Expression Profiling

We developed a protocol for dissociating human brain tumor tissue as well as the tissue of the adjacent tumor penumbra into single viable cells and analyzed the gene expression of these cells using reverse transcription quantitative real time PCR (RT-qPCR, Fig. 1). The tissue was resected in the surgical theatre, where a neuronavigation system was used to determine the precise location of the resected tissue (Fig. 2). The tumor/tumor penumbra tissue dissociation steps included mechanical dissociation, enzymatic digestion and removal of myelin debris and lysis of erythrocytes. The complete tissue dissociation process and collection of single cells took approximately 3 h and the cells remained viable until lysis and, if placed in cell cultures, they survived and propagated (data not shown). RNA from individual cells in 96-well plates was snap frozen and stored at −80 °C. This was followed by reverse transcription, pre-amplification using the primers for selected genes (Tables 1, 2), qPCR, statistical analysis and classification of individual cells with the use of unsupervised algorithms, specifically Self organizing map (SOM), Principal component analysis (PCA) and hierarchical clustering. It was previously demonstrated that SOM constitutes a useful bioinformatics approach to identify cell subpopulations [7].
Fig. 1

Gene expression profiling of individual cells in human brain tissue—a schematic presentation of the experimental procedure

Fig. 2

Patient data and the location of the resected tissue. a Pre-operative MRI of the astrocytoma grade IV patient with the area of resected tumor tissue indicated in blue. b Pre-operative MRI of the oligodendroglioma grade III patient with neuronavigation coordinates (red dots) at surgery, pointing to the tumor (left) and tumor penumbra (right) tissue that are to be resected. Bottom panel shows photographs of temporal lobe surface at surgery with the area encircled in blue of the tumor proper (left) and the tumor penumbra (right) that are to be resected

The protocol was applied to tumor resections from two patients; one with an oligodendroglioma grade III tumor, the other with an astrocytoma grade IV (glioblastoma multiforme) tumor. From the first patient, we assessed the gene expression in cells from the tumor proper as well as from the tumor penumbra, and for the second patient we only had access to the tumor proper. Our panel of genes included cell type markers, genes expressed by proliferating cells (non-tumor and tumor cells), angiogenesis-related genes, proto-oncogenes and tumor suppressor genes, and other tumor cell markers (Table 1).

Oligodendroglioma Grade III Cells Clustered into Two Distinct Subpopulations

We assessed gene expression in 79 cells from the oligodendroglioma grade III tumor (Table 3) and classified them into subpopulations according to their molecular signature (Fig. 3). SOM analysis was used for separation of the cells into subpopulations. The SOM analysis repeatedly identified two cell clusters, which were also detectable in the PCA plot (Fig. 3a) and further confirmed in the hierarchical clustering (Fig. 3e).
Table 3

Percentage of cells expressing the selected genes in cells of oligodendroglioma grade III tumor (OT), the penumbra of oligodendroglioma grade III tumor (OP) and astrocytoma grade IV tumor (A)

Gene symbol

OT n = 79 (%)

OP n = 38 (%)

A n = 86 (%)

P value positive cells OT versus OP

AIF1

18

50

6

<0.001***

ALDH1L1

4

0

1

0.304

CCNE1

28

CD63

44

45

86

0.560

CDK4

32

13

91

0.024*

CNP

22

8

73

0.053

EDNRB

9

0

6

0.058

EGFR

35

0

57

<0.001***

ENO2

33

3

73

<0.001***

ERBB2

3

0

16

0.454

FN1

1

0

6

0.675

GFAP

22

5

8

0.020*

GFAPdelta

4

0

0.304

GLUL

66

76

83

0.175

IL8

22

26

2

0.361

MAPT

42

3

70

<0.001***

MBP

3

MCM2

3

0

56

0.454

MDM4

38

39

83

0.516

MGMT

1

3

6

0.546

MKI67

49

NES

5

3

10

0.475

NOTCH1

25

5

73

0.006**

OLIG2

49

8

74

<0.001***

PCNA

38

39

81

0.516

PECAM1

4

34

2

<0.001***

PROM1

15

RB1

28

21

66

0.290

SEC61G

41

24

91

0.055

SOX2

16

5

47

0.075

SUZ12

51

42

88

0.253

SYNM

14

5

9

0.138

TP53

8

3

33

0.271

TUBB3

16

5

77

0.075

VEGFA

10

18

42

0.168

VIM

11

13

13

0.500

N number of cells; P values for binary comparisons were obtained by Fisher’s exact test

*** P < 0.001; ** P < 0.01; * P < 0.05

Fig. 3

Oligodendroglioma grade III tumor cells separated into two groups using SOM analysis. a The groups from the SOM analysis visualized by PCA scatter plot, one rectangle represents one cell. b PCA loading plot, indicating the weight of each gene along the PC1 and PC2 separation axes. c Diagram showing percentages of cells in the two subpopulations expressing the selected genes. P values for binary comparisons were obtained by Fisher’s exact test. *** P < 0.001; ** P < 0.01; * P < 0.05. d Box plots of log 2 transformed expression levels of genes which expression differed between the red and green subpopulations as determined by Mann–Whitney test and Bonferroni correction for multiple testing. The boxes indicate the 25th and 75th percentiles, red circles indicate outliers and asterisks extreme outliers. e Heat map with hierarchical clustering of individual cells with the grouping based on SOM analysis visualized by green and red at the bottom. Two groups of cells, similar to those seen in the SOM analysis, appeared in the hierarchical clustering

The first cell cluster consisted of 39 % of all oligodendroglioma grade III tumor cells (Fig. 3, red). Genes that influenced separation of cells in Fig. 3a in the PCA were CD63, CDK4, CNP, EDNRB, EGFR, ENO2, GFAP, MAPT, NOTCH1, OLIG2, SEC61G, SUZ12 and TUBB3 (Fig. 3b, PCA loadings). These genes were also expressed in a higher proportion of cells in the red group compared to the green group, together with SOX2 and RB1 (Fig. 3c). Furthermore, CD63, CDK4, CNP, EGFR, ENO2, MAPT, OLIG2, SEC61G and SUZ12 showed a higher quantitative expression level in the red group compared to the green group (Fig. 3d). Although GFAP is usually expressed in astrocytes [22, 23], CNP in oligodendrocytes [24], ENO2 and TUBB3 [25] in neurons and SOX2 in neural progenitor cells [26] and astrocytes [27], and MAPT (TAU) more known for its involvement in Alzheimer’s disease [28], all these genes have also been implicated in the context of malignant tumors (CD63 [29], CDK4 [30, 31, 32], CNP [33], EDNRB [34], EGFR [10, 35], ENO2 [36], MAPT [28, 37, 38], NOTCH1 [39], OLIG2 [40, 41], RB1 [42, 43, 44], SEC61G [45], SOX2 [46], SUZ12 [47], TUBB3 [48] ). Hence, this subpopulation showed gene expression profiles that corresponded to a more cancer-like phenotype than the cells in the green group and we thereby classified them as tumor cells. In addition, the gene expression of CD63, CDK4, CNP, EGFR, ENO2 NOTCH1, OLIG2, SEC61G, SUZ12 and TUBB3 correlated individually in single cells from this tumor (Table 4), which indicates that those genes are co-regulated in these tumor cells.
Table 4

Spearman correlation coefficients for oligodendroglioma grade III tumor cells, selected genes

 

AIF1

CD63

CDK4

CNP

EGFR

ENO2

MAPT

NOTCH1

OLIG2

SEC61G

SUZ12

TUBB3

n = 79

n = 14

n = 35

n = 25

n = 17

n = 28

n = 26

n = 33

n = 20

n = 39

n = 32

n = 40

n = 13

AIF1

1

           

CD63

−0.11

1

          

CDK4

−0.31**

0.43***

1

         

CNP

−0.21

0.48***

0.50***

1

        

EGFR

−0.33**

0.50***

0.41***

0.47***

1

       

ENO2

−0.32**

0.32**

0.52***

0.41***

0.57***

1

      

MAPT

−0.37***

0.47***

0.50***

0.61***

0.69***

0.63***

1

     

NOTCH1

−0.11

0.38***

0.36***

0.54***

0.38***

0.33**

0.41***

1

    

OLIG2

−0.38***

0.48***

0.46***

0.52***

0.69***

0.54***

0.72***

0.30**

1

   

SEC61G

−0.14

0.20

0.46***

0.41***

0.35***

0.24*

0.43***

0.28*

0.36***

1

  

SUZ12

−0.21

0.24*

0.22

0.26*

0.56***

0.38***

0.45***

0.06

0.51***

0.27*

1

 

TUBB3

−0.09

0.46***

0.35**

0.47***

0.29**

0.15

0.33**

0.39***

0.44***

0.36***

0.15

1

N number of cells expressing the respective gene

*** P < 0.001; ** P < 0.01; * P < 0.05

The second cell cluster consisted of 61 % of all oligodendroglioma grade III cells (Fig. 3, green). Those cells localized to the right in the PCA, a consequence of the expression of the microglia marker AIF1 [49] expressed by 29 %, an inflammation marker IL8 [50] expressed by 33 %, and an endothelial cell marker PECAM1 [51] expressed by 6 % of cells in this cluster (Fig. 3b). Expression of AIF1 and IL8 correlated in the cells of the oligodendroglioma tumor (cc. 0.53, p value <0.001) and it has previously been shown that microglia cells produce IL8 [52]. Furthermore, the cells in this cell cluster expressed fewer genes related to tumor properties than the cells of the first cell cluster (Fig. 3c, d). Thus, these cells show expression profiles corresponding to normal cells present within the tumor, and at least some of these cells are likely to be microglia and cells of the vascular system.

Interestingly, none of the cells from the oligodendroglioma grade III tumor expressed proliferation marker MKI67 [53], and only two cells expressed proliferation marker MCM2 [54] (Table 3), which indicates that relatively low proportion of cells of this tumor were proliferating.

Cells of the Second Cell Cluster of the Oligodendroglioma Tumor Shared Molecular Signature with Cells from the Oligodendroglioma Penumbra

PCA plot of the cells from the oligodendroglioma grade III tumor and its penumbra showed that the penumbra cells (Fig. 4a, black) clustered together with the normal cells of the oligodendroglioma grade III (green). This was also seen in the hierarchical clustering where the penumbra cells were within the second oligodendroglioma tumor cell cluster of normal cells, with only a single penumbra cell clustering with tumor cells (Fig. 4d). There were no differences in gene expression between the cluster of normal cells within oligodendroglioma tumor cells and oligodendroglioma penumbra cells (Mann–Whitney tests with Bonferroni correction for multiple testing). PECAM1 were expressed in a higher proportion of penumbra cells compared to “normal cells” of the tumor (Fig. 4c, black vs. green), but except for this gene, the groups did not differ when it comes to proportion of cells expressing specific genes. This further implies that the cells of the oligodendroglioma grade III second cell cluster were normal cells.
Fig. 4

Oligodendroglioma grade III tumor and tumor penumbra cells. The two groups of tumor cells separated by SOM analysis and described in Fig. 3 (red and green cells) are compared with tumor penumbra cells (black). a PCA scatter plot of the three groups of cells, one rectangle represents one cell. The green population of the tumor cells clusters with the penumbra cells. b PCA loading plot, indicating the weight of each gene along the PC1 and PC2 separation axes. c Diagram showing percentages of cells in the tumor penumbra cells (black) and the two subpopulations of tumor cells (red and green) expressing the selected genes. d Heat map with hierarchical clustering of individual cells of green and red tumor cells together with black tumor penumbra cells shows that the green tumor cell population clusters with the penumbra cells

Astrocytoma Grade IV Cells Clustered into Three Subpopulations

We assessed gene expression in 86 cells (Table 3) from the astrocytoma grade IV tumor and classified them into subpopulations based on their molecular signature (Fig. 5). The PCA pointed to three groups of cells (Fig. 5a), these groups were also present in the hierarchical clustering (Fig. 5g). The group consisting of only three cells (Fig. 5a, grey) was excluded since this allowed a better separation of the remaining cells (Fig. 5c–f). These three cells were defined by the expression of GFAP, VIM, ALDH1L1, EDNRB and FN1 (Fig. 5b). VIM (vimentin) and GFAP encode intermediate filament proteins often co-expressed in astrocytes [55]. These cells expressed FN1 (fibronectin), and two of them EDNRB (endothelin receptor B). Both FN1 and EDNRB have been reported to be upregulated in reactive astrocytes [56, 57], and therefore, these cells are likely to be reactive astrocytes.
Fig. 5

Astrocytoma grade IV tumor cells. a PCA scatter plot suggested the existence of three cell populations (indicated in grey, yellow and blue). b PCA loading plot, indicating the weight of each gene along the PC1 and PC2 separation axes. c PCA scatter plot with the grey cell population removed from the analysis to improve separation of the yellow and blue groups along the PC2 axis. d PCA loading plot for (c), indicating the weight of each gene along the PC1 and PC2 separation axes. e Box plots of log 2 transformed expression levels of genes which expression differed between the yellow and blue subpopulations as determined by Mann–Whitney test and Bonferroni correction for multiple testing. The boxes indicate the 25th and 75th percentiles, red circles indicate outliers and asterisks extreme outliers. f Diagram showing percentages of cells in the yellow and red subpopulations expressing the selected genes. P values for binary comparisons were obtained by Fisher’s exact test. *** P < 0.001; ** P < 0.01; * P < 0.05. g Heat map with hierarchical clustering of individual cells with the grouping based on SOM analysis visualized by grey, yellow and blue at the bottom. Three groups of cells, similar to those seen in the SOM analysis, appeared in the hierarchical clustering

The second cell cluster consisted of 86 % of the astrocytoma grade IV tumor cells (Fig. 5, yellow). This group showed high proportion of cells expressing cancer-related genes: CD63, CDK4, CNP, EGFR, ENO2, MAPT, MCM2, MDM4 [58, 59], MKI67, NOTCH1, OLIG2, PCNA [60], RB1, SEC61G, SUZ12, TP53 [9] and TUBB3 (Fig. 5f). The expression of CD63, CDK4, CNP, ENO2, MAPT, MCM2, MKI67, NOTCH1, OLIG2, PCNA, RB1, SEC61G, SUZ12 and TUBB3 correlated in individual cells of the astrocytoma grade IV tumor (Table 5), which suggests that these genes are co-regulated. The expression of many of these genes also correlated in the cells of the oligodendroglioma (CD63, CDK4, CNP, ENO2, NOTCH1, OLIG2, SEC61G, SUZ12 and TUBB3; Table 4).
Table 5

Spearman correlation coefficients for astrocytoma grade IV tumor cells, selected genes

 

AIF1

CD63

CDK4

CNP

ENO2

MAPT

MCM2

MKI67

NOTCH1

OLIG2

PCNA

RB1

SEC61G

SUZ12

TUBB3

n = 86

n = 5

n = 74

n = 78

n = 63

n = 63

n = 60

n = 71

n = 42

n = 63

n = 64

n = 70

n = 57

n = 78

n = 76

n = 66

AIF1

1

              

CD63

−0.32**

1

             

CDK4

−0.39***

0.38***

1

            

CNP

−0.31**

0.50***

0.61***

1

           

ENO2

−0.31**

0.39***

0.35***

0.32**

1

          

MAPT

−0.30**

0.38***

0.29**

0.25*

0.32**

1

         

MCM2

−0.25*

0.35***

0.52***

0.42***

0.31**

0.18

1

        

MKI67

−0.21

0.28**

0.36***

0.32**

0.32**

0.17

0.46***

1

       

NOTCH1

−0.31**

0.39***

0.41***

0.41***

0.30**

0.34***

0.18

0.27**

1

      

OLIG2

−0.32**

0.34***

0.36***

0.36***

0.29**

0.16

0.37***

0.44***

0.37***

1

     

PCNA

−0.36***

0.45***

0.64***

0.47***

0.50***

0.29**

0.79***

0.67***

0.29**

0.47***

1

    

RB1

−0.23*

0.41***

0.53***

0.42***

0.32**

0.23*

0.47***

0.42***

0.22*

0.34***

0.59***

1

   

SEC61G

−0.38***

0.48***

0.43***

0.51***

0.38***

0.35***

0.26*

0.31**

0.34***

0.47***

0.40***

0.32**

1

  

SUZ12

−0.32**

0.34***

0.63***

0.45***

0.48***

0.54***

0.60***

0.46***

0.37***

0.43***

0.68***

0.47***

0.47***

1

 

TUBB3

−0.22*

0.32**

0.46***

0.36***

0.40***

0.42***

0.26*

0.37***

0.26*

0.42*

0.34**

0.32**

0.26*

0.40***

1

N number of cells expressing the respective gene

*** P < 0.001; ** P < 0.01; * P < 0.05

The third cell cluster consisted of nine astrocytoma grade IV tumor cells (10 %, Fig. 5, blue). This cell cluster can be defined by expression of microglial marker AIF1 (55 % of the cells in the blue cluster expressed AIF1, Fig. 5d). AIF1, IL8 and VIM were all expressed in a higher proportion of cells in the third cluster compared to the second cluster of tumor cells (Fig. 5f). AIF1 expression in individual cells negatively correlated with the expression of oligodendrocyte marker CNP [24], neuronal markers ENO2 and TUBB3, and genes related to tumor cell proliferation/behavior CDK4, MCM2, MKI67, NOTCH1, OLIG2, PCNA, SEC61G, SUZ12 (Table 5). Thus, the third cell cluster consisted of cells that did not express the tumor related genes, some of them being microglia.

Subpopulations of Astrocytoma Grade IV and Oligodendroglioma Grade III Tumor Cells Shared Molecular Signature with Cells from the Oligodendroglioma Penumbra

A PCA plot of all the cells from oligodendroglioma grade III tumor and penumbra, and astrocytoma grade IV tumor (Fig. 6a) showed that subpopulations of these cells (second cell cluster) of oligodendroglioma grade III (green), and third cell cluster of astrocytoma grade IV (blue) grouped together with cells of the oligodendroglioma penumbra (black). This was mainly influenced by the expression of AIF1, PECAM1 and IL8 (Fig. 6b). These three groups of cells showed comparable gene expression. Also the proportions of cells within each group expressing the selected genes were similar (Fig. 6c). This implies that also the cells of the third cell cluster of astrocytoma grade IV (blue) were nonmalignant cells present within the tumor, and that these cells did not differ between these two tumor types.
Fig. 6

Comparison of cell populations in oligodendroglioma grade III tumor and tumor penumbra with astrocytoma grade IV tumor cells. The cell populations from previous SOM analyses were used with the color-coding of individuals cells as in Figs. 3, 4 and 5. a All populations visualized by PCA scatter plot, one rectangle represents one cell. b PCA loading plot, indicating the weight of each gene along the PC1 and PC2 separation axes. c Diagram showing percentages of cells expressing the selected genes within each cell population. d Box plots of log 2 transformed expression levels of genes which expression differed between the yellow and red subpopulations as determined by Mann–Whitney test and Bonferroni correction for multiple testing. The boxes indicate the 25th and 75th percentiles, red circles indicate outliers and asterisks extreme outliers

Tumor Cells of the Astrocytoma Grade IV and Oligodendroglioma Tumor Samples Have Distinct Signatures

The first cell cluster of the oligodendroglioma grade III (tumor cells; red) were separated in PCA (Fig. 6a) from the second cell cluster of astrocytoma grade IV (tumor cells; yellow). These two groups differed in the expression of several genes and the three proliferation markers (MCM2, MKI67 and PCNA) showed higher expression in the astrocytoma tumor cells than in the oligodendroglioma tumor cells. Also CDK4, CNP, and NOTCH1 showed a higher expression in the astrocytoma tumor cells. GFAP, MAPT and OLIG2 had a higher expression in the oligodendroglioma tumor cells. This demonstrates that the cells of these two tumor samples have distinct molecular signatures (Fig. 6a, d).

Discussion

In this study, a protocol for single cell gene expression profiling of freshly isolated brain cells was developed and evaluated on individual cells from one oligodendroglioma grade III and one astrocytoma grade IV tumor. We show that cells of the tumor proper from both tumors cluster into subpopulations with different molecular signatures. Both tumors contained subpopulations of cells which expression profile matched the profile of the tumor penumbra, i.e. cells of relatively normal brain tissue (Fig. 6).

The expression of proliferation markers MCM2, MKI67, and PCNA was higher in the tumor cells of the astrocytoma grade IV tumor compared to the oligodendroglioma grade III tumor (Fig. 6d). Furthermore, MKI67 and MCM2 were expressed in half of the cells of the astrocytoma (49 and 56 %, respectively), but no oligodendroglioma tumor cells expressed MKI67, and only 3 % of these cells expressed MCM2. PCNA was expressed by 81 % of the astrocytoma tumor cells, but only 38 % of the oligodendroglioma tumor cells (Table 3). This indicated that the proliferation was substantially higher in the astrocytoma tumor, and this was also compatible with the clinical history of the patients. The oligodendroglioma patient was diagnosed 29 years ago and received a radiation therapy 10 years ago as the only treatment for his slowly growing tumor. The astrocytoma patient had clinical symptoms of the tumor for only 1 month before surgery, indicating a fast growing tumor. This was also compatible with the histopathological evaluation of the tumor tissue that showed high mitotic activity in the astrocytoma grade IV tumor, but only a limited one in the oligodendroglioma grade III tumor.

Cell subpopulations in the oligodendroglioma grade III tumor and astrocytoma grade IV tumor showed similar expression profiles as cells of the oligodendroglioma penumbra. Subpopulations of cells within the oligodendroglioma and astrocytoma tumors, and tumor penumbra cells of the oligodendroglioma did not differ in the expression of any of the assessed genes, and they clustered together in the PCA of all the examined cells (Fig. 6a). A high proportion of these cells expressed microglia marker AIF1, and PECAM1, a marker of endothelial cells, indicating that some of them are likely to be microglia cells or cells of the vascular system (Fig. 6c).

The cells of oligodendroglioma grade III and astrocytoma grade IV differed in the expression of several genes. A number of genes were expressed in cells in both tumors: CD63, CDK4, CNP, EGFR, ENO2, GFAP, MAPT, NOTCH1, OLIG2, SEC61G, SUZ12 and TUBB3. In the astrocytoma grade IV, additional genes were expressed: MCM2, MDM4 MKI67, PCNA, RB1 and TP53. Likewise, the oligodendroglioma tumor cells showed a higher expression level of GFAP, MAPT and OLIG2 compared to astrocytoma cells (Fig. 6d). Ligon et al. showed that oligodendrogliomas had higher expression of OLIG2 than high grade astrocytomas, although both tumor types expressed this gene [40]. It was reported that oligodendroglioma cells express GFAP [61, 62], and that the number of GFAP expressing cells decrease with increasing grade of astrocytoma [63]. This is in concordance with our results: 90 % of the second cell cluster within the oligodendroglioma tumor expressed GFAP, whereas only three cells within the astrocytoma tumor expressed GFAP. The expression of MAPT (TAU) was not previously reported in astrocytomas or oligodendrogliomas, although it was associated to several other tumor types [37, 38, 64]. NOTCH1 were more highly expressed in the cells of astrocytoma grade IV compared to the oligodendroglioma grade III. This is compatible with the finding that NOTCH1 expression increases with increasing grade of astrocytomas [65].

In conclusion, we have made the initial evaluation of a method for classification of populations of cells based on their molecular characteristics. Single cell gene expression profiling seems to be a highly useful method allowing identification of molecular signatures of individual subpopulations of cells within a given tumor and can be used to compare molecular signatures among individual tumors and between tumor types and grades.

Notes

Acknowledgments

We would like to thank Dr. Yolanda de Pablo for providing mouse brain tissue for development of the cell dissociation protocols. This work was supported by Swedish Medical Research Council (11548 to MP), ALF Gothenburg (11267 to MP, 11392 to BR, 137241 to AS), Söderbergs Foundations, Hjärnfonden, Amlöv’s Foundation, E. Jacobson’s Donation Fund, NanoNet COST Action (BM1002), EU FP 7 Program EduGlia (237956) to MP, EU FP 7 Program TargetBraIn (279017), AFA Research Foundation, Gothenburg Foundation for Neurological Research, FoU (Västra Götalandsregionen), the Swedish Cancer Society, and Wilhelm and Martina Lundgren foundation.

References

  1. 1.
    Liu X, Long F, Peng H, Aerni SJ, Jiang M, Sanchez-Blanco A, Murray JI, Preston E, Mericle B, Batzoglou S, Myers EW, Kim SK (2009) Analysis of cell fate from single-cell gene expression profiles in C. elegans. Cell 139(3):623–633. doi: 10.1016/j.cell.2009.08.044 CrossRefPubMedGoogle Scholar
  2. 2.
    Wills QF, Livak KJ, Tipping AJ, Enver T, Goldson AJ, Sexton DW, Holmes C (2013) Single-cell gene expression analysis reveals genetic associations masked in whole-tissue experiments. Nat Biotechnol 31(8):748–752. doi: 10.1038/nbt.2642 CrossRefPubMedGoogle Scholar
  3. 3.
    Wang D, Bodovitz S (2010) Single cell analysis: the new frontier in ‘omics’. Trends Biotechnol 28(6):281–290. doi: 10.1016/j.tibtech.2010.03.002 CrossRefPubMedCentralPubMedGoogle Scholar
  4. 4.
    Dalerba P, Kalisky T, Sahoo D, Rajendran PS, Rothenberg ME, Leyrat AA, Sim S, Okamoto J, Johnston DM, Qian D, Zabala M, Bueno J, Neff NF, Wang J, Shelton AA, Visser B, Hisamori S, Shimono Y, van de Wetering M, Clevers H, Clarke MF, Quake SR (2011) Single-cell dissection of transcriptional heterogeneity in human colon tumors. Nat Biotechnol 29(12):1120–1127. doi: 10.1038/nbt.2038 CrossRefPubMedCentralPubMedGoogle Scholar
  5. 5.
    Powell AA, Talasaz AH, Zhang H, Coram MA, Reddy A, Deng G, Telli ML, Advani RH, Carlson RW, Mollick JA, Sheth S, Kurian AW, Ford JM, Stockdale FE, Quake SR, Pease RF, Mindrinos MN, Bhanot G, Dairkee SH, Davis RW, Jeffrey SS (2012) Single cell profiling of circulating tumor cells: transcriptional heterogeneity and diversity from breast cancer cell lines. PLoS One 7(5):e33788. doi: 10.1371/journal.pone.0033788 CrossRefPubMedCentralPubMedGoogle Scholar
  6. 6.
    Ramsköld D, Luo S, Wang YC, Li R, Deng Q, Faridani OR, Daniels GA, Khrebtukova I, Loring JF, Laurent LC, Schroth GP, Sandberg R (2012) Full-length mRNA-Seq from single-cell levels of RNA and individual circulating tumor cells. Nat Biotechnol 30(8):777–782. doi: 10.1038/nbt.2282 CrossRefPubMedCentralPubMedGoogle Scholar
  7. 7.
    Ståhlberg A, Andersson D, Aurelius J, Faiz M, Pekna M, Kubista M, Pekny M (2011) Defining cell populations with single-cell gene expression profiling: correlations and identification of astrocyte subpopulations. Nucleic Acids Res 39(4):e24. doi: 10.1093/nar/gkq1182 CrossRefPubMedCentralPubMedGoogle Scholar
  8. 8.
    Socialstyrelsen Cancerfonden (2013) Cancer i siffror 2013, vol 2013. Socialstyrelsen, Cancerfonden, StockholmGoogle Scholar
  9. 9.
    Ohgaki H, Kleihues P (2005) Population-based studies on incidence, survival rates, and genetic alterations in astrocytic and oligodendroglial gliomas. J Neuropathol Exp Neurol 64(6):479–489PubMedGoogle Scholar
  10. 10.
    Nafe R, Glienke W, Burgemeister R, Gangnus R, Haar B, Pries A, Schlote W (2004) Regional heterogeneity of EGFR gene amplification and nuclear morphology in glioblastomas. An investigation using laser microdissection and pressure catapulting. Anal Quant Cytol Histol 26(2):65–76PubMedGoogle Scholar
  11. 11.
    Shapiro JR, Yung WK, Shapiro WR (1981) Isolation, karyotype, and clonal growth of heterogeneous subpopulations of human malignant gliomas. Cancer Res 41(6):2349–2359PubMedGoogle Scholar
  12. 12.
    Walker C, du Plessis DG, Joyce KA, Machell Y, Thomson-Hehir J, Al Haddad SA, Broome JC, Warnke PC (2003) Phenotype versus genotype in gliomas displaying inter- or intratumoral histological heterogeneity. Clin Cancer Res 9(13):4841–4851PubMedGoogle Scholar
  13. 13.
    Singh SK, Clarke ID, Terasaki M, Bonn VE, Hawkins C, Squire J, Dirks PB (2003) Identification of a cancer stem cell in human brain tumors. Cancer Res 63(18):5821–5828PubMedGoogle Scholar
  14. 14.
    Vescovi AL, Galli R, Reynolds BA (2006) Brain tumour stem cells. Nat Rev Cancer 6(6):425–436. doi: 10.1038/nrc1889 CrossRefPubMedGoogle Scholar
  15. 15.
    Bonavia R, Inda MM, Cavenee WK, Furnari FB (2011) Heterogeneity maintenance in glioblastoma: a social network. Cancer Res 71(12):4055–4060. doi: 10.1158/0008-5472.CAN-11-0153 CrossRefPubMedCentralPubMedGoogle Scholar
  16. 16.
    Ståhlberg A, Rusnakova V, Forootan A, Anderova M, Kubista M (2012) RT-qPCR work-flow for single-cell data analysis. Methods. doi: 10.1016/j.ymeth.2012.09.007 Google Scholar
  17. 17.
    Ringner M (2008) What is principal component analysis? Nat Biotechnol 26(3):303–304. doi: 10.1038/nbt0308-303 CrossRefPubMedGoogle Scholar
  18. 18.
    Kohonen T (1982) Self-Organized Formation of Topologically Correct Feature Maps. Biol Cybern 43(1):59–69. doi: 10.1007/Bf00337288 CrossRefGoogle Scholar
  19. 19.
    Ward JH (1963) Hierarchical grouping to optimize an objective function. J Am Stat Assoc 58(301):236–244. doi: 10.2307/2282967 CrossRefGoogle Scholar
  20. 20.
    Mann HB, Whitney DR (1947) On a test of whether one of two random variables is stochastically larger than the other. Ann Math Stat 18(1):50–60CrossRefGoogle Scholar
  21. 21.
    Dunn OJ (1961) Multiple comparisons among means. J Am Stat Assoc 56(293):52–64CrossRefGoogle Scholar
  22. 22.
    Eng LF, Ghirnikar RS, Lee YL (2000) Glial fibrillary acidic protein: GFAP-thirty-one years (1969-2000). Neurochem Res 25(9–10):1439–1451CrossRefPubMedGoogle Scholar
  23. 23.
    Cahoy JD, Emery B, Kaushal A, Foo LC, Zamanian JL, Christopherson KS, Xing Y, Lubischer JL, Krieg PA, Krupenko SA, Thompson WJ, Barres BA (2008) A transcriptome database for astrocytes, neurons, and oligodendrocytes: a new resource for understanding brain development and function. J Neurosci 28(1):264–278. doi: 10.1523/JNEUROSCI.4178-07.2008 CrossRefPubMedGoogle Scholar
  24. 24.
    Brunner C, Lassmann H, Waehneldt TV, Matthieu JM, Linington C (1989) Differential ultrastructural localization of myelin basic protein, myelin/oligodendroglial glycoprotein, and 2′,3′-cyclic nucleotide 3′-phosphodiesterase in the CNS of adult rats. J Neurochem 52(1):296–304CrossRefPubMedGoogle Scholar
  25. 25.
    Marangos PJ, Schmechel DE (1987) Neuron specific enolase, a clinically useful marker for neurons and neuroendocrine cells. Annu Rev Neurosci 10:269–295. doi: 10.1146/annurev.ne.10.030187.001413 CrossRefPubMedGoogle Scholar
  26. 26.
    Schwartz PH, Nethercott H, Kirov II, Ziaeian B, Young MJ, Klassen H (2005) Expression of neurodevelopmental markers by cultured porcine neural precursor cells. Stem Cells 23(9):1286–1294. doi: 10.1634/stemcells.2004-0306 CrossRefPubMedGoogle Scholar
  27. 27.
    Jinno S (2011) Regional and laminar differences in antigen profiles and spatial distributions of astrocytes in the mouse hippocampus, with reference to aging. Neuroscience 180:41–52. doi: 10.1016/j.neuroscience.2011.02.013 CrossRefPubMedGoogle Scholar
  28. 28.
    Souter S, Lee G (2010) Tubulin-independent tau in Alzheimer’s disease and cancer: implications for disease pathogenesis and treatment. Curr Alzheimer Res 7(8):697–707CrossRefPubMedGoogle Scholar
  29. 29.
    Rorive S, Lopez XM, Maris C, Trepant AL, Sauvage S, Sadeghi N, Roland I, Decaestecker C, Salmon I (2010) TIMP-4 and CD63: new prognostic biomarkers in human astrocytomas. Mod Pathol 23(10):1418–1428. doi: 10.1038/modpathol.2010.136 CrossRefPubMedGoogle Scholar
  30. 30.
    Haas K, Staller P, Geisen C, Bartek J, Eilers M, Moroy T (1997) Mutual requirement of CDK4 and Myc in malignant transformation: evidence for cyclin D1/CDK4 and p16INK4A as upstream regulators of Myc. Oncogene 15(2):179–192. doi: 10.1038/sj.onc.1201171 CrossRefPubMedGoogle Scholar
  31. 31.
    Lam PY, Di Tomaso E, Ng HK, Pang JC, Roussel MF, Hjelm NM (2000) Expression of p19INK4d, CDK4, CDK6 in glioblastoma multiforme. Br J Neurosurg 14(1):28–32CrossRefPubMedGoogle Scholar
  32. 32.
    Tsutsui T, Hesabi B, Moons DS, Pandolfi PP, Hansel KS, Koff A, Kiyokawa H (1999) Targeted disruption of CDK4 delays cell cycle entry with enhanced p27(Kip1) activity. Mol Cell Biol 19(10):7011–7019PubMedCentralPubMedGoogle Scholar
  33. 33.
    Zorniak M, Clark PA, Leeper HE, Tipping MD, Francis DM, Kozak KR, Salamat MS, Kuo JS (2012) Differential expression of 2′,3′-cyclic-nucleotide 3′-phosphodiesterase and neural lineage markers correlate with glioblastoma xenograft infiltration and patient survival. Clinical Cancer Res 18(13):3628–3636. doi: 10.1158/1078-0432.CCR-12-0339 CrossRefGoogle Scholar
  34. 34.
    Lahav R (2005) Endothelin receptor B is required for the expansion of melanocyte precursors and malignant melanoma. Int J Dev Biol 49(2–3):173–180Google Scholar
  35. 35.
    Heimberger AB, Suki D, Yang D, Shi W, Aldape K (2005) The natural history of EGFR and EGFRvIII in glioblastoma patients. J Transl Med 3:38. doi: 10.1186/1479-5876-3-38 CrossRefPubMedCentralPubMedGoogle Scholar
  36. 36.
    Soh MA, Garrett SH, Somji S, Dunlevy JR, Zhou XD, Sens MA, Bathula CS, Allen C, Sens DA (2011) Arsenic, cadmium and neuron specific enolase (ENO2, gamma-enolase) expression in breast cancer. Cancer Cell Int 11(1):41. doi: 10.1186/1475-2867-11-41 CrossRefPubMedCentralPubMedGoogle Scholar
  37. 37.
    Rouzier R, Rajan R, Wagner P, Hess KR, Gold DL, Stec J, Ayers M, Ross JS, Zhang P, Buchholz TA, Kuerer H, Green M, Arun B, Hortobagyi GN, Symmans WF, Pusztai L (2005) Microtubule-associated protein tau: a marker of paclitaxel sensitivity in breast cancer. Proc Natl Acad Sci USA 102(23):8315–8320. doi: 10.1073/pnas.0408974102 CrossRefPubMedCentralPubMedGoogle Scholar
  38. 38.
    Souter S, Lee G (2009) Microtubule-associated protein tau in human prostate cancer cells: isoforms, phosphorylation, and interactions. J Cell Biochem 108(3):555–564. doi: 10.1002/jcb.22287 CrossRefPubMedCentralPubMedGoogle Scholar
  39. 39.
    Fassl A, Tagscherer KE, Richter J, Berriel Diaz M, Alcantara Llaguno SR, Campos B, Kopitz J, Herold-Mende C, Herzig S, Schmidt MH, Parada LF, Wiestler OD, Roth W (2012) Notch1 signaling promotes survival of glioblastoma cells via EGFR-mediated induction of anti-apoptotic Mcl-1. Oncogene 31(44):4698–4708. doi: 10.1038/onc.2011.615 CrossRefPubMedGoogle Scholar
  40. 40.
    Ligon KL, Alberta JA, Kho AT, Weiss J, Kwaan MR, Nutt CL, Louis DN, Stiles CD, Rowitch DH (2004) The oligodendroglial lineage marker OLIG2 is universally expressed in diffuse gliomas. J Neuropathol Exp Neurol 63(5):499–509PubMedGoogle Scholar
  41. 41.
    Ligon KL, Huillard E, Mehta S, Kesari S, Liu H, Alberta JA, Bachoo RM, Kane M, Louis DN, Depinho RA, Anderson DJ, Stiles CD, Rowitch DH (2007) Olig2-regulated lineage-restricted pathway controls replication competence in neural stem cells and malignant glioma. Neuron 53(4):503–517. doi: 10.1016/j.neuron.2007.01.009 CrossRefPubMedCentralPubMedGoogle Scholar
  42. 42.
    Cho HS, Hayami S, Toyokawa G, Maejima K, Yamane Y, Suzuki T, Dohmae N, Kogure M, Kang D, Neal DE, Ponder BA, Yamaue H, Nakamura Y, Hamamoto R (2012) RB1 methylation by SMYD2 enhances cell cycle progression through an increase of RB1 phosphorylation. Neoplasia 14(6):476–486PubMedCentralPubMedGoogle Scholar
  43. 43.
    Nakamura M, Yonekawa Y, Kleihues P, Ohgaki H (2001) Promoter hypermethylation of the RB1 gene in glioblastomas. Lab Invest 81(1):77–82CrossRefPubMedGoogle Scholar
  44. 44.
    Murphree AL, Benedict WF (1984) Retinoblastoma: clues to human oncogenesis. Science 223(4640):1028–1033CrossRefPubMedGoogle Scholar
  45. 45.
    Lu Z, Zhou L, Killela P, Rasheed AB, Di C, Poe WE, McLendon RE, Bigner DD, Nicchitta C, Yan H (2009) Glioblastoma proto-oncogene SEC61gamma is required for tumor cell survival and response to endoplasmic reticulum stress. Cancer Res 69(23):9105–9111. doi: 10.1158/0008-5472.CAN-09-2775 CrossRefPubMedCentralPubMedGoogle Scholar
  46. 46.
    Gangemi RM, Griffero F, Marubbi D, Perera M, Capra MC, Malatesta P, Ravetti GL, Zona GL, Daga A, Corte G (2009) SOX2 silencing in glioblastoma tumor-initiating cells causes stop of proliferation and loss of tumorigenicity. Stem Cells 27(1):40–48. doi: 10.1634/stemcells.2008-0493 CrossRefPubMedGoogle Scholar
  47. 47.
    Iliopoulos D, Lindahl-Allen M, Polytarchou C, Hirsch HA, Tsichlis PN, Struhl K (2010) Loss of miR-200 inhibition of Suz12 leads to polycomb-mediated repression required for the formation and maintenance of cancer stem cells. Mol Cell 39(5):761–772. doi: 10.1016/j.molcel.2010.08.013 CrossRefPubMedCentralPubMedGoogle Scholar
  48. 48.
    Levallet G, Bergot E, Antoine M, Creveuil C, Santos AO, Beau-Faller M, de Fraipont F, Brambilla E, Levallet J, Morin F, Westeel V, Wislez M, Quoix E, Debieuvre D, Dubois F, Rouquette I, Pujol JL, Moro-Sibilot D, Camonis J, Zalcman G, Intergroupe Francophone de Cancerologie T (2012) High TUBB3 expression, an independent prognostic marker in patients with early non-small cell lung cancer treated by preoperative chemotherapy, is regulated by K-Ras signaling pathway. Mol Cancer Ther 11(5):1203–1213. doi: 10.1158/1535-7163.MCT-11-0899 CrossRefPubMedGoogle Scholar
  49. 49.
    Ahmed Z, Shaw G, Sharma VP, Yang C, McGowan E, Dickson DW (2007) Actin-binding proteins coronin-1a and IBA-1 are effective microglial markers for immunohistochemistry. J Histochem Cytochem 55(7):687–700. doi: 10.1369/jhc.6A7156.2007 CrossRefPubMedGoogle Scholar
  50. 50.
    Harada A, Sekido N, Akahoshi T, Wada T, Mukaida N, Matsushima K (1994) Essential involvement of interleukin-8 (IL-8) in acute inflammation. J Leukoc Biol 56(5):559–564PubMedGoogle Scholar
  51. 51.
    Woodfin A, Voisin MB, Nourshargh S (2007) PECAM-1: a multi-functional molecule in inflammation and vascular biology. Arterioscler Thromb Vasc Biol 27(12):2514–2523. doi: 10.1161/ATVBAHA.107.151456 CrossRefPubMedGoogle Scholar
  52. 52.
    Ehrlich LC, Hu S, Sheng WS, Sutton RL, Rockswold GL, Peterson PK, Chao CC (1998) Cytokine regulation of human microglial cell IL-8 production. J Immunol 160(4):1944–1948PubMedGoogle Scholar
  53. 53.
    Raghavan R, Steart PV, Weller RO (1990) Cell proliferation patterns in the diagnosis of astrocytomas, anaplastic astrocytomas and glioblastoma multiforme: a Ki-67 study. Neuropathol Appl Neurobiol 16(2):123–133CrossRefPubMedGoogle Scholar
  54. 54.
    Taran K, Sitkiewicz A, Andrzejewska E, Kobos J (2011) Minichromosome maintenance 2 (MCM2) is a new prognostic proliferative marker in Wilms tumour. Pol J Pathol 62(2):84–88PubMedGoogle Scholar
  55. 55.
    Eliasson C, Sahlgren C, Berthold CH, Stakeberg J, Celis JE, Betsholtz C, Eriksson JE, Pekny M (1999) Intermediate filament protein partnership in astrocytes. J Biol Chem 274(34):23996–24006CrossRefPubMedGoogle Scholar
  56. 56.
    Sofroniew MV (2005) Reactive astrocytes in neural repair and protection. Neuroscientist 11(5):400–407. doi: 10.1177/1073858405278321 CrossRefPubMedGoogle Scholar
  57. 57.
    Rogers SD, Peters CM, Pomonis JD, Hagiwara H, Ghilardi JR, Mantyh PW (2003) Endothelin B receptors are expressed by astrocytes and regulate astrocyte hypertrophy in the normal and injured CNS. Glia 41(2):180–190. doi: 10.1002/glia.10173 CrossRefPubMedGoogle Scholar
  58. 58.
    Danovi D, Meulmeester E, Pasini D, Migliorini D, Capra M, Frenk R, de Graaf P, Francoz S, Gasparini P, Gobbi A, Helin K, Pelicci PG, Jochemsen AG, Marine JC (2004) Amplification of Mdmx (or Mdm4) directly contributes to tumor formation by inhibiting p53 tumor suppressor activity. Mol Cell Biol 24(13):5835–5843. doi: 10.1128/MCB.24.13.5835-5843.2004 CrossRefPubMedCentralPubMedGoogle Scholar
  59. 59.
    Francoz S, Froment P, Bogaerts S, De Clercq S, Maetens M, Doumont G, Bellefroid E, Marine JC (2006) Mdm4 and Mdm2 cooperate to inhibit p53 activity in proliferating and quiescent cells in vivo. Proc Natl Acad Sci USA 103(9):3232–3237. doi: 10.1073/pnas.0508476103 CrossRefPubMedCentralPubMedGoogle Scholar
  60. 60.
    Kubben FJ, Peeters-Haesevoets A, Engels LG, Baeten CG, Schutte B, Arends JW, Stockbrugger RW, Blijham GH (1994) Proliferating cell nuclear antigen (PCNA): a new marker to study human colonic cell proliferation. Gut 35(4):530–535CrossRefPubMedCentralPubMedGoogle Scholar
  61. 61.
    Kros JM, Van Eden CG, Stefanko SZ, Waayer-Van Batenburg M, van der Kwast TH (1990) Prognostic implications of glial fibrillary acidic protein containing cell types in oligodendrogliomas. Cancer 66(6):1204–1212CrossRefPubMedGoogle Scholar
  62. 62.
    Nakopoulou L, Kerezoudi E, Thomaides T, Litsios B (1990) An immunocytochemical comparison of glial fibrillary acidic protein, S-100p and vimentin in human glial tumors. J Neurooncol 8(1):33–40PubMedGoogle Scholar
  63. 63.
    van der Meulen JD, Houthoff HJ, Ebels EJ (1978) Glial fibrillary acidic protein in human gliomas. Neuropathol Appl Neurobiol 4(3):177–190CrossRefPubMedGoogle Scholar
  64. 64.
    Pusztai L, Jeong JH, Gong Y, Ross JS, Kim C, Paik S, Rouzier R, Andre F, Hortobagyi GN, Wolmark N, Symmans WF (2009) Evaluation of microtubule-associated protein-Tau expression as a prognostic and predictive marker in the NSABP-B 28 randomized clinical trial. J Clin Oncol 27(26):4287–4292. doi: 10.1200/JCO.2008.21.6887 CrossRefPubMedCentralPubMedGoogle Scholar
  65. 65.
    Xu P, Yu S, Jiang R, Kang C, Wang G, Jiang H, Pu P (2009) Differential expression of Notch family members in astrocytomas and medulloblastomas. Pathol Oncol Res 15(4):703–710. doi: 10.1007/s12253-009-9173-x CrossRefPubMedGoogle Scholar

Copyright information

© Springer Science+Business Media New York 2014

Authors and Affiliations

  • Elin Möllerström
    • 1
  • Bertil Rydenhag
    • 3
  • Daniel Andersson
    • 1
    • 4
  • Isabell Lebkuechner
    • 1
  • Till B. Puschmann
    • 1
  • Meng Chen
    • 1
  • Ulrika Wilhelmsson
    • 1
  • Anders Ståhlberg
    • 1
    • 4
  • Kristina Malmgren
    • 3
  • Milos Pekny
    • 1
    • 2
  1. 1.Center for Brain Repair and Rehabilitation, Department of Clinical Neuroscience and Rehabilitation, Institute of Neuroscience and PhysiologySahlgrenska Academy at the University of GothenburgGöteborgSweden
  2. 2.Florey Institute of Neuroscience and Mental HealthParkvilleAustralia
  3. 3.Department of Clinical Neuroscience and Rehabilitation, Institute of Neuroscience and PhysiologySahlgrenska Academy at the University of GothenburgGöteborgSweden
  4. 4.Sahlgrenska Cancer Center, Department of Pathology, Institute of BiomedicineSahlgrenska Academy at the University of GothenburgGöteborgSweden

Personalised recommendations