Content uploaded by Edmund Ui-Hang Sim
Author content
All content in this area was uploaded by Edmund Ui-Hang Sim on Sep 14, 2016
Content may be subject to copyright.
Jurnal Biosains, 17(1), 19–37, 2006
A PRELIMINARY STUDY OF DIFFERENTIALLY EXPRESSED GENES IN
MALAYSIAN COLORECTAL CARCINOMA CASES
1Sim E U H*, 2Bong I P N, 2Balraj P, 2Tan S K, 3Jamal R, 3Sagap I, 3Nadeson S, 3Rose I M
and 4Lim P K M
1Human Molecular Genetics Laboratory, Resource Biotechnology Programme, Faculty of
Resource Science and Technology, Universiti Malaysia Sarawak, 94300 Kota Samarahan,
Sarawak, Malaysia
2Institute for Medical Research (IMR), 50588 Kuala Lumpur, Malaysia
3Faculty of Medicine, Universiti Kebangsaan Malaysia, 56000 Kuala Lumpur, Malaysia
4Malaysian Bio-Diagnostics Research Sdn. Bhd. (MBDr), Block Intron-Ekson UKM-MTDC
Smart Technology Centre, 43650 Bangi, Selangor Darul Ehsan, Malaysia
Abstrak: Setakat ini, mekanisme genetik yang lengkap tentang kemajuan adenoma
kepada karsinoma dalam karsinoma usus besar masih belum diketahui. Demi
memperoleh maklumat genetik tentang lintasan karsinogenesis ini, kami menggunakan
kaedah pencarian gen yang diekspres secara berbeza dalam tumor karsinoma usus
besar. Profil pengekspresan gen daripada kes karsinoma usus besar dikaji menggunakan
sistem DNA Microarray. Kami melaporkan pengawalaturan tinggi dan rendah untuk 819
dan 98 gen masing-masing dalam tumor relatif kepada kawalan normal berkenaan. Corak
pengekspresan berbeza dalam 121 gen adalah kekal dalam semua tumor yang dikaji.
Tiga puluh tiga daripada gen-gen ini adalah gen protein ribosom (RP). Perbandingan data
dengan pangkalan data pengekspresan gen domain awam (Cancer Gene Expression
Database, CGED) menunjukkan 47 gen pengekspresan berbeza adalah konsisten. Dua
puluh dua daripada gen-gen ini adalah gen RP. Antara semua gen RP yang dikenal pasti
dalam kajian ini, corak pengekspresan untuk enam gen adalah selaras dengan sorotan
kajian. Pengekspresan tinggi gen RP L32 dilaporkan buat pertama kali dalam kajian ini
dan telah disahkan melalui penganalisaan RT-PCR. Gen-gen yang tidak berkaitan dengan
RP tetapi penting untuk dipertimbang adalah gen tumour susceptibility (TSG101) dan 20-
kDa myosin light chain (MLC-2). Walaupun penganalisaan microarray kami berdasarkan
saiz sampel kecil (n = 2), kajian awal ini mengenal pasti banyak gen yang dikaitkan
dengan konteks pembentukan dan kemajuan karsinoma usus besar buat pertama kali.
Maka, penemuan kami membekalkan maklumat baru untuk kejadian genetik dalam proses
karsinoma usus besar.
Abstract: Presently, the complete genetic mechanisms for the progression of adenoma to
carcinoma for colorectal carcinoma (CRC) remain largely unclear. In order to obtain
genetic information of this cancer pathway, we searched for differentially expressed genes
in tumours of CRC. Gene expression profiles from CRC cases were assessed via the DNA
microarray system. We report up-regulation and down-regulation of 819 and 98 genes
respectively, in the tumours relative to their normal controls. The differential expression
patterns of 121 genes were persistent in all tumours. Thirty three of these are ribosomal
proteins (RPs) genes. Comparison of the 121 genes with a public domain gene expression
database, the Cancer Gene Expression Database (CGEP), revealed 47 genes to be
consistently differentially expressed in colorectal tumours. Among these, 22 are RP genes.
Among all RP genes identified in this study the over-expression pattern for six of them is
consistent with literature. The up-regulation of RP L32 in CRC tumours was demonstrated
for the first time in this study and it was also verified via reverse transcription-polymerase
19
*Corresponding author: uhsim@frst.unimas.my
Sim E U H et al.
chain reaction (RT-PCR) analysis. Non-RP genes worth noting are the tumour
susceptibility gene (TSG101) and the 20-kDa myosin light chain (MLC-2). Albeit small
sample size (n = 2 for microarray analysis), our preliminary studies revealed many genes
that are brought into the context of CRC tumourigenesis for the first time, thus providing
new clues to the genetic events during colorectal carcinogenesis.
Keywords: Colorectal Carcinoma, DNA Microarray, Ribosomal Proteins (RP), TSG101,
MLC-2.
INTRODUCTION
Colorectal carcinoma (CRC) is one of the human cancer models where the
progressive histopathological stages correlate with gradual but sequential
perturbation of specific genes. These basic and sequential events that correlate
with the adenoma-carcinoma progression have been reviewed by Fodde et al.
(2001). Generally, in the case of a type of inheritable CRC – the familial
adenomatous polyposis (FAP), initiation of tumour formation and clonal evolution
of tumourigenic cells can be triggered by germline inactivating mutations in the
adenomatous polyposis coli (APC) gene. In the context of Wnt-signaling pathway,
the improper function of mutant APC leads to the stabilization of
β
-catenin, and
hence the formation of the
β
-catenin/TCF-LEF (T-cell factor-lymphoid enchancer
factor) complex. This results in the ectopic activation of oncogenes, one of which
is k-ras. This in turn leads to a proportion of precancerous colorectal cells
becoming adenomatous polyps. Further genetic disruption specifically in the form
of Loss of Heterozygosity (LOH) at chromosomes 18q and 17p involving
SMAD2/SMAD4 and TP53 genes respectively, causes the malignant
transformation of benign adenomas to invasive carcinoma of the colorectum. This
basic sequence of genetic events in colorectal cancer evolution may also be true
for sporadic CRC, as APC mutations have been verified in early stages of
sporadic CRC tumours (Powell et al. 1992). Contrary to the phenomenon of APC-
initiated tumourigenesis, genetic defects (causing tumour progression) in the
second familial forms of CRC, the Hereditary Nonpolyposis Colorectal Cancer
(HNPCC), occur in the DNA mismatch repair (MMR) genes (Kinzler & Vogelstein
1996). In the HNPCC cases, MMR deficiencies concomitantly cause adenomas
to acquire higher rate of mutation relative to normal colorectal cells. The resultant
accumulation of mutations in oncogenes and tumor suppressor genes will lead to
malignant transformation of adenomas (Kinzler & Vogelstein 1996).
The variety of genetic pathways that pertain to the adenoma-carcinoma
sequence between FAP (and some sporadic CRC) and HNPCC reflects the
complexities of the molecular mechanisms underlying tumor initiation to
malignant progression of colorectal carcinoma. Presently, the understanding of
these mechanisms remains basic if not partially elucidated. Of late, several
groups have attempted to study the molecular events of colorectal carcinoma
using approaches of gene expression analysis. These include the Serial Analysis
of Gene Expression (SAGE) analysis (Zhang et al. 1997), the Affymetrix Human
GeneChipTM (6500 and 6800) Set oligonucleotide arrays (Notterman et al. 2001),
the 19,200-Element Complementary DNA microarray (Hedge et al. 2001), the
20
Differentially expressed genes in colorectal carcinoma cases
Suppressive Subtractive Hybridization (SSH) techniques (Hufton et al. 1999; Luo
& Lai 2001) and a combination of SSH with cDNA library array technology
(Swearingen et al. 2003). The findings of these studies produced a large
repertoire of differentially expressed genes that are yet to be comprehensively
and accurately incorporated into the existing molecular pathway.
In this study we report the assessment of gene expression profile of CRC
cases via DNA microarray strategy. This has allowed us to identify and verify a
number (121) of persistently differentially expressed genes. Our findings will
inevitably provide important information for a comprehensive delineation of
molecular pathway of CRC.
MATERIALS AND METHODS
Total RNA
Commercially available normal colon (cat. no. 64065-1) and colon tumour total
RNAs (cat. no. 64014-1) were purchased from BD BioSciences (Clontech
Laboratories, Inc.). For our study, these normal colon and colon tumour total RNA
samples are designated as CN and CT respectively. The total RNAs from
colorectal carcinoma and their paired normal tissue biopsies from local patients
are designated as 019T and 056T, and 019N and 056N respectively. These were
extracted using the Trizol method. Basically, frozen CRC and paired normal
tissues were cut into smaller pieces with sterile surgical blade and then
homogenized in 1 ml Trizol reagent (Invitrogen) using a polytron homogenizer.
Following incubation at room temperature for 10 minutes, the homogenate was
mixed with 200 µl chloroform, incubated at room temperature for 3 minutes and
then centrifuged at 12,000 g for 15 minutes at 4°C. Total RNA present in the
colourless aqueous layer (top layer) was precipitated using 0.5 ml isopropanol.
The resulting total RNA pellet was washed with 75% ethanol (1 ml ethanol/ 1 ml
Trizol ratio) and rinsed with 1 ml absolute ethanol. The final and purified dry total
RNA pellet was then dissolved with 70 µl warm elution buffer. Estimation of
concentration and purity of total RNA was performed via spectrophotometric
assay at A260 and A280. For this study, colorectal tumour samples are designated
as 019T and 056T, and their paired normal tissues as 019N and 056N,
respectively.
DNA Microarray
MICROMAXTM Human cDNA I Array slide (cat. no. MPS621) were from
AlphaGene Inc. and PerkinElmer Life Science Inc. (USA). Each microarray slide
contains 2382 elements comprising known human genes, control genes and
housekeeping genes. All genes were spotted in duplicate.
Probe Preparation and Hybridization of Array
Labelling of probes with fluorescent dyes was performed using MICROMAXTM
Direct Labeling Kit (AlphaGene Inc. & PerkinElmer Life Science Inc.; cat no.
MPS502) and according to the manufacturer's protocol. Tumour samples were
labeled with Cyanine 5 (Cy5) and normal samples were labeled with Cyanine 3
21
Sim E U H et al.
(Cy3) in this study. Purification of labeled cDNA probes was conducted using the
isopropanol precipitation method. An equal concentration of Cy3 and Cy5-labeled
cDNA probes were mixed, dried and dissolved in 20 µl hybridizaton buffer, prior
to hybridization on the array slides for overnight at 65°C in a hybridization
incubator. Washing of microarray slides was carried out in 0.5X SSC, 0.01%
SDS; 0.06X SSC, 0.01% SDS; and 0.06X SSC – each for 15 minutes at room
temperature with gentle agitation.
Microarray Data Analysis
Hybridization signals were detected using a flourescence scanner (Typhoon 8600
variable mode imager, Amersham Pharmacia Biotech) and documented using the
ImageQuant software. The data was then processed and analyzed using
GenePix Array Ver. 4.1 program (Axon Instruments Inc., Canada).
Reverse Transcriptase - PCR (RT-PCR)
One to two microlitres of total RNA was used as template for RT-PCR assay in a
20 µl reaction volume. For each assay, generation of first strand cDNAs was
catalyzed by 200U Moloney murine leukemia virus (MMLV-RT), using oligo-dT18
primers. PCR amplification of the first strand cDNAs was carried out using the
parameters of 95°C for 15 minutes (hot start); and then 21 cycles of 95°C × 30
seconds (denature), 59–64°C × 2 minutes (annealing) and 72°C × 2 minutes
(extension); and a final extension step of 72°C for 10 minutes. Primer pairs of
selected genes were designed to amplify regions within the coding region of each
gene. Results of RT-PCR assays were assessed on agarose via gel
electrophoretic analysis.
RESULTS
cDNA Microarray Analysis
Genes that are prominently differentially expressed in both sets are listed in Table
1 and 2. Comparative assessment of CT versus CN revealed 625 and 94 genes
up-regulated and down-regulated in the tumour case, respectively. The highest
level of up-regulation in CT was demonstrated by the osteoblast specific factor 2
(OSF-2p1) gene (82 folds), whereas the highest level of down-regulation in CT
are found in the glutathione peroxidase (12 folds) and TYL genes (12 folds). In
the case of the colorectal tumour sample (019T) and its paired normal control
tissue (019N), 315 genes showed up-regulation in tumour whereas 5 genes were
found to be down-regulated. The MPV17 gene showed the highest level of up-
regulation (98 folds), and the MT-1l gene showed the highest level of down-
regulation (5 folds).
In total, of the 2382 genes (MICROMAXTM Human cDNA I Array system)
analyzed, 917 genes appeared to have differential expression patterns between
normal and tumour cases. Of these, mRNA levels for 819 genes are up-regulated
and 98 genes are down-regulated in tumour cases. Relative expression
difference ranges from 2 to 92 folds, with 12.3% (113/917) of the genes showing
10 folds difference or higher. Generally, the CT/CN set exhibited higher number
of differentially expressed genes (719) compared to the 019T/019N set (320
22
Differentially expressed genes in colorectal carcinoma cases
genes). The ratio of up-regulated to down-regulated genes in the CT/CN set is
approximately 6.67:1 (625:94), in contrast to 63:1 (315:5) for the 019T/019N set.
Table 1: Prominently differentially expressed genes/proteins from microarray analysis of
CN versus CT via the MICROMAXTM Human cDNA I Array system.
Acc. No. Transcripts up-regulated
in CT
Fold Acc. No. Transcripts down-
regulated in CT
Fold
D13665 OSF-2p1 82 D00632 Glutathione peroxidase
(Glu-Ox)
12
L22587 Immunoglobulin heavy
chain, V region (IGH@)
64 X99688 TYL 12
Y14737 Immunoglobulin lambda
heavy chain (Igλ-H)
48 X68485 A1 adenosine receptor 10
X83703 Cytokine inducible nuclear
protein
32 L10335 Neuroendocrine-specific
protein C (NSP)
8
J04765 Osteopontin 32 D29808 T-cell acute
lymphoblastic
7
M80927 Glycoprotein 32 J03483 Leukemia associated
antigen 1 (TAL L A -1)
7
M77844 Oculorhombin 20 L33404 Chromogranin A 7
X13694 Osteopontin 20 U05598 Stratum corneum
chymotryptic enzyme
6
L42611 Keratin 6 isoform K6e
(KRT6E)
16 Y11588 Dihydrodiol
dehydrogenase
6
U62962 Int-6 16 X52426 Apoptosis specific
protein
6
Table 2: Prominently differentially expressed genes from microarray analysis of 019N
versus 019T via the MICROMAXTM Human cDNA I Array system.
Acc. No. Transcripts up-regulated
in 019T
Fold Acc. No. Transcripts down-
regulated in 019T
Fold
X76538 MPV17 98 X76717 MT-1l 5
M86917 Oxysterol-binding protein
(OSBP)
72 M95787 22-kDa smooth muscle
protein (SM22)
4
Z68204 Succinyl CoA synthethase 56 J02854 20-kDa myosin light chain
(MLC-2)
2
M58018 Beta-myosin heavy chain
(MYH7)
39 X13839 Vascular smooth muscle
alpha-actin
2
M11354 H3.3 histone, class B 36 L25798 3-hydroxy-3-
methylglutaryl
coenzyme A synthase
2
S82470 BB1, tumor progression-
enhanced factor gene
36
D88378 Proteasome inhibitor
hPl31 subunit
33
M36340 ADP-ribosylation factor 1
(ARF1)
32
U54558 Translation initiation factor
elF3p66 subunit
32
U75283 Sigma receptor 29
23
Sim E U H et al.
Persistently Differentially Expressed Genes
Comparison of results of differentially expressed genes results between the two
colorectal carcinoma sets revealed 121 genes that are persistently differentially
expressed in both sets (Table 3). Persistent differential expression refers to
expression behaviour of genes that showed consistent differential expression in
both sets of sample analyzed. Except for MLC-2, all of these genes are up-
regulated in tumours. Only five genes (RP L26, chondroitin sulfate proteoglycan,
vasopressin activated calcium mobilizing receptor-like protein, KIAA0428 and
vimentin) showed consistency in the differential expression pattern, where equal
relative fold-difference in the two sets (019N/019T and CN/CT) was observed. A
majority of persistently differentially expressed genes (107 of 121) in the
019N/019T set showed higher level of fold-difference when compared to that of
the CN/CT set. Comparing our results with the Cancer Gene Expression
Database (CGED) (Kato et al. 2005), available at http://cged.hgc.jp, reveals 47
genes that are consistently differentially expressed (Table 3). These are genes
that are differentially expressed in both tumour sets studied, and also listed in the
Cancer Gene Expression Database (CGED) as differentially expressed in
colorectal tumours. Of the 47 genes, 11 show down-regulation in tumours in
contrast to our results of their up-regulation in tumours studied. As such, 29.8%
(36/121) of our microarray results on persistently differentially expressed genes is
validated against an independent set of data (CGED).
Table 3: Persistent differentially expressed genes obtained from comparison among
differentially expressed genes of CN/CT, 019N/019T and CGEP (Kato et al. 2005) sets.
Acc. No. Approximate fold difference
Protein/Gene CT/CN 019T/019N CGED
X13584 mRNA for gamma-aminobutyric-
acidreceptor alpha-subunit (GABA-A
receptor alpha subunit)
2 18 -
U46751 Phosphotyrosine independent ligand p62 for
the Lck SH2 domain
3 18 -
U83857 Aac11 (aac11) 3 17 -
D42054 mRNA for KIAA0092 gene 3 15 -
X14420 mRNA for pro-alpha-1 type 3 collagen 4 15 -
U96915 sin3 associated polypeptide p18 (SAP18) 2 14 3.4
U10248 Ribosomal protein L29 (humrpl29) 3 14 1.4
M81757 S19 ribosomal protein 3 14 1.5
L11566 Ribosomal protein L18 (RPL18) 2 13 -
M14648 Cell adhesion protein (vitronectin) receptor
alpha subunit
4 13 -
D45887 mRNA for calmodulin 2 12 -1.6
Z47087 mRNA for RNA polymerase II elongation
factor-like protein
2 12 9
(continue on next page)
24
Tab l e 3 (continued)
Approximate fold difference
Acc. No. Protein/Gene CT/CN 019T/019N CGED
J02939 Membrane glycoprotein 4F2 antigen heavy
chain mRNA
2 12 -
AF010187 FGF-1 intracellular binding protein (FIBP) 3 12 -
L25899 Ribosomal protein L10 2 12 1.5
U14968 Ribosomal protein L27a 3 12 1.4
X76534 NMB mRNA 2 11 -
Y00387 mRNA for glutamine synthetase (E.C.
6.3.1.2)
2 11 -2.4
X06323 MRL3 mRNA for ribosomal protein L3
homologue (MRL3, mammalian ribosome
L3)
2 11 3
X56999 UbA52 placental mRNA for ubiquitin-52
amino acid fusion protein
3 11 1
J03068 DNF1552 (lung) 2 11 1.4
U51678 Small acidic protein 5 11 -
L38961 Putative transmembrane protein precursor
(B5)
2 11 -
D37991 SSR2 mRNA for beta-signal sequence
receptor
2 11 -
S42658 S3 ribosomal protein 2 11 1.3
D00759 mRNA for proteasome subunit HC2 2 10 NI
U14971 Ribosomal protein S9 2 10 -
D50419 mRNA for OTK18 3 9 -
M15661 Ribosomal protein 5 9 -
D21239 mRNA for C3G protein 3 9 -
X15998 mRNA for the chondroitin sulphate
proteoglycan versican, V1 splice-variant;
precursor peptide
6 9 -
U32944 Cytoplasmic dynein light chain 1 (hdlc1) 2 9 2.6
J05500 Beta-spectrin (SPTB) 3 9 -
M55067 47-kD autosomal chronic granulomatous
disease protein
4 9 -
J04543 Synexin 2 9 -
L06499 Ribosomal protein L37a (RPL37A) 3 9 1.3
J03040 SPARC/osteonectin 10 8 -5.1
X02152 mRNA for lactate dehydrogenase-A (LDH-A,
EC 1.1.1.27)
3 8 -
S54761 Beta 2- mu, beta 2-microglobulin 2 8 -3.2
L06505 Ribosomal protein L12 2 8 9.0
M36072 Ribosomal protein L7a (surf 3) large subunit 4 8 -
L36645 Receptor protein-tyrosine kinase (HEK8) 3 8 -
M58458 Ribosomal protein S4 (RPS4X) isoform 3 8 1.9
(continue on next page)
25
Tab l e 3 (continued)
Approximate fold difference
Acc. No. Protein/Gene CT/CN 019T/019N CGED
M17733 Thymosin beta-4 3 8 -
U69645 Zinc finger protein 2 8 -
M17886 Acidic ribosomal phosphoprotein P1 2 8 1.5
D14530 Homolog of yeast ribosomal protein S28 5 8 1.5
D89729 mRNA for CRM1 protein 6 7 -1.0
D87735 mRNA for ribosomal protein L14 3 7 2.6
Y00711 mRNA for lactate dehydrogenase B (LDH-B) 2 7 1.9
X16064 mRNA for translationally controlled tumor
protein
3 7 -
L13740 TR3 orphan receptor 2 7 -
X03342 mRNA for ribosomal protein L32 2 7 -
M77804 Tryptophanyl tRNA synthetase (IFNWRS) 2 7 -
X91257 mRNA for seryl-tRNA synthetase 2 7 -
D89289 mRNA for N-Acetyl-beta-D-glucosaminide 3 7 -
M21300 Small proline rich protein (sprI) mRNA, clone
15B
4 7 -
U09510 glycyl-tRNA synthetase 3 7 -
U87460 Putative endothelin receptor type B-like
protein
2 7 -
L25085 Sec61-complex beta-subunit 2 6 1.8
M74002 Arginine-rich nuclear protein 2 7 -
X52966 mRNA for ribosomal protein L35a 4 7 -
Y10275 mRNA for L-3-phosphoserine phosphatase 3 7 -
X80909 Alpha NAC mRNA 2 7 -1.3
X63237 Uba80 mRNA for ubiquitin 3 7 1.4
X53505 mRNA for ribosomal protein S12 2 6 -
D88674 mRNA for antizyme inhibitor 2 6 -
M20020 Ribosomal protein S6 8 6 -
U35622 EWS-E1A-F chimeric protein 2 6 -
AF026844 Ribosomal protein L41 2 6 1.7
AF083255 RNA helicase-related protein 2 6 -
M81635 Erythrocyte membrane protein 3 6 -
Z26876 Gene for ribosomal protein L38 4 6 -
AF013168 Hamartin (TSC1) 3 6 -
U26173 bZIP protein NF-IL3A (IL3BP1) 3 6 -
L06498 Ribosomal protein S20 (RPS20) 3 6 -5.1
U14973 Ribosomal protein S29 2 6 2.1
M61866 Krueppel-related DNA-binding protein (PF4)
mRNA, 5' end
3 6 -
U10550 Gem GTPase (gem) 2 6 -
U51432 Nuclear protein Skip 4 6 -
L10413 Farnesyltransferase alpha-subunit 3 6 -
X69392 mRNA for ribosomal protein L26 6 6 -
(continue on next page)
Tab l e 3 (continued)
Approximate fold difference
Acc. No. Protein/Gene CT/CN 019T/019N CGED
X55954 mRNA for HL23 ribosomal protei homologue 4 6 1.4
X64707 BBC1 mRNA 2 5 2.4
L28010 HnRNP F protein 2 5 -
X62691 mRNA for ribosomal protein (homologuous
to yeast S24)
4 5 1.2
S75725 Prostacyclin-stimulating factor 4 5 -
D50372 mRNA for myosin regulatory light chain 4 5 1.6
D80009 mRNA for KIAA0187 gene 5 5 -
M13932 Ribosomal protein S17 2 5 2.2
M76979 Pigment epithelium-differentiation factor
(PEDF)
2 5 -
M62831 Transcription factor ETR101 2 5 -
L19527 Ribosomal protein L27 (RPL27) 3 5 -
M14219 Chondroitin/dermatan sulfate proteoglycan
(PG40) core protein
4 4 -
M34671 Lymphocytic antigen CD59/MEM43 3 4 3.6
L39060 Transcription factor SL1 2 4 -
X81882 mRNA for for vasopressin activated calcium
mobilizing receptor-like protein
4 4 -
U12404 Csa-19 2 3 1.5
M99701 pp21 3 4 -
X12451 mRNA for pro-cathepsin L (major excreted
protein MEP)
3 4 -
D13665 mRNA for osteoblast specific factor 2 (OSF-
2p1)
82 4 -
D90402 mRNA for endothelin receptor (ETR) 3 4 -
AB007888 KIAA0428 4 4 4.6
M64716 Ribosomal protein S25 3 4 -
J02947 Extracellular-superoxide dismutase (SOD3) 3 4 2.3
U14967 Ribosomal protein L21 5 4 -1.0
U47742 Monocytic leukaemia zinc finger protein
(MOZ)
3 4 -
U37230 Ribosomal protein L23a 3 4 -
M94314 Ribosomal protein L30 3 4 -1.3
U82130 Tumor susceptiblity protein (TSG101) 2 4 9
X89401 mRNA for large subunit of ribosomal protein
L21
6 4 -1.0
X52022 RNA for type VI collagen alpha3 chain 4 3 -1.9
AF010313 Pig8 (PIG8) 2 3 -
L03555 Ig rearranged kappa-chain mRNA variable
region, joining region, constant region
2 3 -
M28212 GTP-binding protein (RAB6 4 3 -
U09953 Ribosomal protein L9 2 3 9.0
(continue on next page)
Sim E U H et al.
Tab l e 3 (continued)
Approximate fold difference
Acc. No. Protein/Gene CT/CN 019T/019N CGED
X16478 mRNA 5'-fragment for vimentin N-terminal
fragment
3 3 -
U27655 RGP3 2 3 -
L12535 RSU-1/RSP-1 2 3 -
U76992 Tat - S F 1 3 19 -
U75283 Sigma receptor (hSigmaR1) 2 29 -
J02854 20-kDa myosin light chain (MLC-2) -3 -2 NI
Note: Fold difference of differential expression from CGED is expressed as the ratio of frequency of
ESTs in cancer libraries to frequency of all expressed seqnencetag (ESTs), or vice versa. Positive and
negative values indicate up-regulation and down-regulation in tumours, respectively. The
abbreviation, NI, means no information for frequency of EST is available although expression in
cancer is reported.
RT-PCR Verification
For the purpose of verifying our microarray findings, RT-PCR assays were
performed for a selected number of differentially expressed genes reported.
Genes targeted for analysis were either randomly selected (for 019T/019N set
and persistently expressed genes) or based on prominence in their fold-
difference of differential expression (for the CT/CN set).
In the CT/CN set – OSF-2p1, RT L7a (surf-3), immunoglobulin lambda
heavy chain (Ig
λ
-H) and immunoglobulin heavy chain V region (IGH@) genes
were demonstrated to be up-regulated in the tumour, while the glutathione
peroxidase (Glu-Ox), TYL and the 20-kDa myosin light chain (MLC-2) genes
were shown to be down-regulated in the tumour [Fig. 1 (A)]. Analysis of the
019N/019T set revealed up-regulation of BB1, MPV17 and RP L32 genes in the
tumour [(Fig. 1 (B)]. The up-regulation of TSG101 gene in tumour cases was
verified from the analysis of a second local CRC case, the 056N/056T set [(Fig. 1
(C)]. Analysis of the 056N/056T set [Fig. 1 (C)] also revealed reproducibility of
results for the RP L7a (surf-3) gene [comparison with CN/CT set; [Fig. 1(A)], and
the RP L32 gene [comparison with 019N/019T set; [Fig. 1 (B)]. The GAPDH
control [Fig.1 (A & C)] affirmed the equimolar concentration of starting total RNA
used in the study. On the whole, the RT-PCR assays confirmed the authenticity
of the differentially expressed genes procured from our microarray analysis.
28
Differentially expressed genes in colorectal carcinoma cases
29
400
200
100
300
300
400
300
400
300
400
400
300
(A) bp
GAPDH (82 bp)
MLC-2 (251 bp)
TYL (310 bp)
Glu-Ox (293 bp)
RP L7a (363 bp)
Ig
λ
-H (293 bp)
IGH@ (262 bp)
OSF-2p1 (302 bp)
CT
CN
Figure 1: RT-PCR verification of a subset of differentially expressed genes identified in
the microarray analysis. (A) are results from analysis of CN/CT (commercially-available)
system; and (B) and (C) are results from analysis of local CRC tumour and its paired
normal samples – 019N/019T and 056N/056T, respectively. Fragment sizes indicated in
parentheses are the expected RT-PCR product sizes. The DNA size reference used was
the GeneRulerTM 100 bp DNA ladder (MBI Fermentas). (continue on next page)
Sim E U H et al.
30
300
019T 019N
(B) bp
RP L32
MPV17 (270 bp)
BBI (295 bp)
400
400
300
(C) bp 019N 019T
GAPDH
500
RP L32
RP L7a
TSG101 (261 bp)
500
300
100
500
Figure 1: (continued)
DISCUSSION
The results of this study, like several of others using similar approaches, yielded
a large repertoire of differentially expressed markers in tumours of CRC. The
comparison of their in vivo expression profiles, and hence identification of
common differentially expressed genes has allowed greater accuracy in targeting
genes associated with the genetic events of CRC development. In the CN/CT
system, the tumour sample was not compared to a paired normal. Although this
may suggests incompatible comparison to a certain extent, such approach is still
considered valid, as similar strategy was employed in the published work by
Differentially expressed genes in colorectal carcinoma cases
Swearingen et al. (2003). In their case, primary tumours from different individuals
were compared to the normal colon tissue from another set of different individuals
that were neither afflicted with CRC nor related to those where the tumour
specimens were procured. Furthermore, the occurrence of many persistently
differentially expressed markers in both sets studied (CN/CT and 019N/019T)
provides validation to our approach. To this end, we have identified 121
persistently differentially expressed genes – 47 of which were also listed as
differentially expressed in colorectal tumours according to the public domain gene
expression database, CGED (Kato et al. 2005; http://cged.hgc.jp).
Of the 47 consistent differentially expressed genes, 22 encode RP or
ribosomal protein-related genes. Four of these have been reported by others to
be over-expressed in CRC cases. These are the RP genes of S3 (Pogue-Geile et
al. 1991), S19 (Kondoh et al. 1992), L7a (Wang et al. 2000) and human
homologue of yeast RP S28 (Otsuka et al. 2001). Besides these, two other RP
genes not listed in CGED but have been reported by others to be differentially
expressed are the S6 and S12 RP genes. The S3, S6 and S12 RP genes have
been demonstrated to be over-expressed in adenocarcinoma of the colon relative
to normal colonic mucosa, and also more abundant in adenomatous polyps
(Pogue-Geile et al. 1991). In fact, we showed up-regulation of L26 and L35 RP
genes in our tumour sets, despite failure by Pogue-Geile's group (1991) to detect
their transcripts in either normal or malignant colon using Northern analysis. The
S19 RP genes has been shown to be up-regulated in colon carcinoma tissue and
has increased level of expression that correlates with tumour progression in colon
cancer cell lines (Kondoh et al. 1992). In the case of L7a (surf 3), its up-regulation
in colorectal cancer was demonstrated by Wang et al. (2000). The human
homologue of yeast RP S28 was shown by Otsuka's group (2001) to be up-
regulated in metastatic-tumour-derived cells of colorectal carcinoma compared
with primary-tumour-derived cells.
The fact that a majority (22 of 47) of consistent differentially expressed
genes represent RP genes suggests strong correlation between high levels of RP
mRNAs with neoplasia of the colorectum. It would seem logical that the increased
level of RP mRNAs can be simply due to the presence of higher percentage of
proliferating cells in neoplasia situation. However, Pogue-Geile et al. (1991)
explained that the proliferation rate of colorectal carcinomas cells is not
significantly higher than that of normal colonic mucosa – hence a possibility that
higher mRNAs levels may be due to a decreased mRNA degradation rather than
an increased transcription. This remains to be proven. It is also unclear whether
the increased level of RP transcripts can be causative of carcinogenesis. Indeed
this seems plausible through several findings that include the suggestion of S6
RP's tumour suppressive function in Drosophila hematopoietic system (Watson et
al. 1992), formation of protein complex between human L5 RP with Mdm2 and
p53 (Marechal et al. 1994), mutation of human S19 RP genes in sporadic and
familial cases of Diamond-Blackfan anemia – a syndrome with increased risk of
developing leukemia (Draptchinskaia et al. 1999) and heterozygous mutations in
11 different RP genes that predispose zebrafish to cancer (Amsterdam et al.
2004). Although the general indication from these findings favours the notion that
a reduced level of RP genes would lead to carcinogenesis, studies by others and
31
Sim E U H et al.
ours on human CRC cases suggested otherwise. Thus, the role(s) of RPs in
cancer and the mechanism(s) by which increased RP gene transcripts leads to
carcinogenesis in CRC remains to be studied.
The tumour susceptibility gene, TSG101 was consistently up-regulated in
all tumour samples studied (CT, 019T and 056T) and also listed in the CGED as
up-regulated in colorectal tumours. The TSG101 gene is located at chromosome
11p15.1–15.2 – a region known to contain tumour suppressor genes and was
initially shown to have large intragenic deletions in human breast cancers (Li et
al. 1997). Abnormal/truncated or aberrantly spliced TSG101 transcripts have
been commonly reported in the human breast cancers – amongst other forms of
cancer (Lee & Feinberg 1997; Carney et al. 1998; Turpin et al. 1999; Balz et al.
2002). Interestingly, for cases of CRC, the presence of aberrant transcripts were
considered to be consequence of PCR artifacts as they were also found in the
normal controls to the colon tumours studied (Hampl et al. 1998; Lin et al. 1998).
In its suspected involvement with oncoprotein network, specifically the p53-
MDM2 circuitry, Li et al. (2001) explained that interaction between TSG101 and
MDM2 elevate the level of MDM2, and consequently promote decay of p53.
Coupled with the findings that the over-expression of MDM2 promote loss of
TSG101 (Li et al. 2001), it appears that TSG101 may play the role of regulator
and target of the p53-MDM2 pathway during the control of cellular apoptosis and
proliferation. The correlation between TSG101 and p53 decay was further
strengthened through the findings of homozygous TSG101 -/- mouse embryos
that showed significant accumulation of p53 protein (Ruland et al. 2001).
However, unlike a typical tumour suppressor, studies have shown that under-
expression of TSG101 did not cause cellular over-proliferation. Despite the
suggestive role(s) of the TSG101-MDM2-P53 complex in mediating
carcinogenesis, there is still no data or findings that support predispositional
inactivating intragenic TSG101 mutations in CRC or any other forms of cancers.
One thing remains consistent is the up-regulation of TSG101 in cancers. Our
results from all CRC tumours samples studied (CT, 019T and 056T), the records
from CGED (Kato et al. 2005) and findings by Koon et al. (2004) in malignant
gastrointestinal stromal tumours are evidence to this consistency. Perhaps this
suggests that TSG101 has greater roles in tumour progression of CRC rather
than neoplastic predisposition or tumour suppression. Indeed the deficiency of
TSG101 expression has been directly correlated to cellular growth arrest,
specifically at the growth 1/synthesis (G1/S) phase of the cell cycle checkpoint,
leading to prohibition of cellular growth and proliferation. Evidence to this were
revealed in studies that include the quenching of TSG101 protein in cells via the
introduction of TSG101-specific antibodies (Zhong et al. 1998), observation of
primary mouse embryonic fibroblast derived from TSG101 conditional knockout
mice (Krempler et al. 2002) and RNA interference experiment on prostate cancer
(PC3) and breast cancer (MDA-MB-231) cell lines (Zhu et al. 2004). The actual
influence of TSG101 expression on the control of proliferation of colorectal or
colorectal-derived cells will require further investigation.
In this study, persistent down-regulation in CRC was demonstrated for
the 20-kDa regulatory myosin light chain gene, MLC-2. Although expression of
human MLC-2 has been reported in normal human colon tissue (Kumar et al.
32
Differentially expressed genes in colorectal carcinoma cases
1989), its down-regulated expression in colorectal tumours is reported for the first
time in the study. Products of the MLC-2 gene is important for regulation of
smooth muscle and non-muscle cell contractile activity via phosphorylated MLC-
2-mediated increase of actin-activated myosin ATPase activity (Kumar et al.
1989). Despite using a small number of samples, we suspect that the expression
behaviour observed attributes to the malfunctioning of colorectal smooth muscle
tissues. In fact, when neonatal human prostate epithelial cells were subjected to
multiple X-ray exposures, the derived cell lines showed reduced cell size with
poorly organized actin stress fibres and demonstrated progressive loss of MLC-2
expression (Prasad et al. 1997). Similar expression behaviour of MLC-2 was
observed in human osteosarcoma derived cell clonal cells that have undergone
transformation following infection by Kirsten murine sarcoma virus or by chemical
carcinogen (Kumar & Chang 1992). Interestingly, the oncoprotein-mediated
signally pathway affecting MLC-2 repression was proven by an in vitro study –
where the repression of chicken MLC-2 promoter by the proto-oncogene, fos,
was via a fos-responsive element (FRE) at –1130 to – 1200 bp upstream of the
MLC-2 transcription initiation site (Goswami et al. 1992). In addition, the complete
repression of MLC-2 expression was reported in human osteosarcoma derived
clonal cells transformed by Ha-ras oncogene (Kumar & Chang 1992). Taken
together, these reports and our results suggest that oncogene-mediated
repression of MLC-2 in CRC probably caused both neoplastic development and
malfunction of affected colorectal smooth muscle cells.
Besides differentially expressed genes common for both sets analyzed, a
few of those that are differentially expressed in either set have been shown in
literature to be associated with CRC, or other cancers. Amongst these, are the
oncogenes of K-ras, dek and set. As reviewed in Fodde et al. (2001), expression
of K-ras and other oncogenes is crucial for transformation of aberrant crypt foci to
adenoma. In studies by Hedge et al. (2001), the set and dek oncogenes have
been demonstrated to be over-expressed in liver metastatic cell line derived from
CRC. Similarly, we found these three genes to be up-regulated in one of the
cancer sets studied (CN/CT). The level of up-regulation of the three oncogenes
(K-ras, dek and set) between studies of Hedge's group and ours is comparable
(Table 4). This consistency among findings suggests that the set and dek
oncogenes work in association with K-ras in tumorigenic (adenomatous)
transformation of colorectal mucosal cells, and that the maintenance of their high
expression beyond adenomatous tissue (metastatic and advanced stages)
suggests that their roles may persist throughout early intermediate stages of
carcinogenesis into invasive carcinoma malignancy.
Table 4: Consistent differentially expressed genes from this study and in comparison to
findings from literature.
Fold of differential expression in tumour cases Genes
Hedge et al. (2001) Notterman et al. (2001) This study (CT/CN)
K-ras 2.4 – 3.7 – 2
dek 2.5 – 3
set 2 – 4 – 3
CAS – 4.7 6
33
Sim E U H et al.
Finally, despite the large database of differentially expressed genes from
our findings that could be implicated as CRC-associated genes, a majority of
these genes have never been brought into the context of CRC tumourigenesis.
Therefore, much has yet to be done to fully understand their physiological role(s)
in colorectal organogenesis and carcinogenesis. More importantly, the consistent
expression behaviours of these genes in most forms of CRC have to be
established, as well as their place in the existing gene-gene network and pathway
in CRC development. Two areas of investigation will be crucial for future work.
Firstly, since our findings (DNA microarray data) are based on small sample size
(n = 2), there is a need to establish reproducibility of the differential expression
signatures of these genes via expression profile analysis of more CRC cases.
Secondly, the increasing number of newly identified differentially expressed
genes in CRC must be fully utilize for more comprehensive functional and gene-
gene interaction studies. Ultimately these efforts will provide accurate information
on molecular events occurring during onset of tumourigenesis to progression of
malignancy in cancer of the colorectum.
ACKNOWLEDGEMENTS
The facilities for scanning of microarray slides and software for microarray data
analysis (GenePix Array Ver 4.1) were provided by the Institute of Health and
Community Medicine (IHCM), Universiti Malaysia Sarawak. This work was
supported by a grant from the Malaysian National Biotechnology Directorate –
Medical Biotechnology Cooperative Centre research initiatives (Project No.: 06-
05-01-003 BTK/ER/018).
REFERENCES
Amsterdam A, Sadler K C, Lai K, Farrington S, Bronson R T, Lees J A and Hopkins N.
(2004). Many ribosomal protein genes are cancer genes in zebrafish. PLoS Biology
2(5): 690–698.
Balz V, Prisack H B, Bier H and Bojar H. (2002). Analysis of BRCA1, TP53 and TSG101
germline mutations in German breast and/or ovarian cancer families. Cancer Genet
and Cytogenet 138(2): 120–127.
Carney M E, Maxwell G L, Lancaster J M, Gumbs C, Marks J, Berchuck A and Futreal P
A. (1998). Aberrant splicing of the TSG101 tumor suppressor gene in human breast
and ovarian cancers. J. Soc. Gynecol. Investig. 5(5): 28–85.
Draptchinskaia N, Gustavsson P, Anderson B, Pettersson M, Willig T N, Dianzani I, Ball S,
Tchernia G, Klar J, Matsson H, Tentler D, Mohandas N, Carlsson B and Dahl N.
(1999). The gene encoding ribosomal protein S19 is mutated in Diamond-Blackfan
anemia. Nat. Genet. 21: 169–175.
Fearnhead N S, Britton M P and Bodmer W F. (2001). The ABC of APC. Human Molecular
Genetics 10(7): 721–733.
34
Differentially expressed genes in colorectal carcinoma cases
Fodde R, Smits R and Clevers H. (2001). APC, signal transduction and genetic instability
in colorectal cancer. Nature Reviews 1: 55–67.
Goswami S K, Zarraga A M, Martin M E, Morgenstern D and Siddiqui M A. (1992). fos-
mediated repression of cardiac myosin light chain-2 gene transcription. Cell. Mol.
Biol. 38(1): 49–58.
Hampl M, Hampl J, Plaschke J, Fitze G, Schackert G, Saeger H D and Schackert H K.
(1998). Evidence that TSG101 aberrant transcripts are PCR artifacts. Biochem.
Biophys. Res. Commun. 248(3): 753–760.
Hedge P, Rong Q, Gaspard R, Abernathy K, Dharap S, Earle-Hughes J, Gay C,
Nwokekeh N U, Chen T, Saeed A, Sharov V, Lee N H, Yeatman T J and
Quackenbush J. (2001). Identification of tumour markers in sets of human colorectal
cancer using a 19,200-element complementary DNA microarray. Cancer Res. 61:
7792–7797.
Hufton S E, Moerkerk P T, Brandwijk R, de Bruine A P, Arends J W and Hoogenboom H
R. (1999). A profile of differentially expressed genes in primary colorectal cancer
using suppression subtractive hybridization. FEBS Lett. 463(1–2): 77–82.
Kato K, Yamashita R, Matoba R, Monden M, Shinzaburo N, Takagi T and Nakai K. (2005).
Cancer gene expression database (CGED): A database for gene expression profiling
with accompanying clinical information of human cancer tissues. Nucleic Acids
Research 33: Database issue D533–D536.
Katoh M. (2001). Frequent up-regulation of WNT2 in primary gastric and colorectal cancer.
International Journal of Oncology 19: 1003–1007.
Kinzler K W and Vogelstein B. (1996). Lessons from hereditary colorectal cancer. Cell 87:
159–170.
Kondoh N, Schweinfest C W, Henderson K W and Papas T S. (1992). Differential
expression of S19 ribosomal protein laminin-binding protein and human lymphocyte
antigen class-I messenger RNAs associated with colon carcinoma progression and
differentiation. Cancer Res. 52(4): 791–796.
Koon N, Schneider-Stock R, Sarlomo-Rikala M, Lasota J, Smolkin M, Petroni G, Zaika A,
Boltze C, Meyer F, Andersson L, Knuutila S, Miettinen M and El-Rifai W. (2004).
Molecular targets for tumour progression in gastrointestinal stromal tumours. Gut
53(2): 235–240.
Krempler A, Henry M D, Triplett A A and Wagner K U. (2002). Targeted deletion of the
TSG101 gene results in cell cycle arrest at G1/S and p53-independent cell death. J.
Biol. Chem. 277(45): 43216–43223.
Kumar C C, Mohan S R, Zavodny P J, Narula S K and Leibowitz P J. (1989).
Characterization and differential expression of human vascular smooth muscle
myosin light chain-2 isoform in non-muscle cells. Biochemistry 28(9): 4027–4035.
35
Sim E U H et al.
Kumar C C and Chang C. (1992). Human smooth muscle myosin light chain-2 gene
expression is repressed in ras transformed fibroblast cells. Cell Growth Differ. 3(1):
1–10.
Lee M P and Feinberg A P. (1997). Aberrant splicing but not mutations of TSG101 in
human breast cancer. Cancer Res. 57(15): 3131–3134.
Li L, Li X, Francke U and Cohen S N. (1997). The TSG101 tumor susceptibility gene is
located in chromosome 11 band p15 and is mutated in human breast cancer. Cell
88(1): 143–154.
Li L, Liao J, Ruland J, Mak T W and Cohen S N. (2001). A TSG101/MDM2 regulatory loop
modulates MDM2 degradation and MDM2/p53 feedback control. Proc. Natl. Acad.
Sci. USA 98(4): 1619–1624.
Lin P M, Liu TC, Chang J G, Chen T P and Lin S F. (1998). Aberrant TSG101 transcripts
in acute myeloid leukaemia. Br. J. Haematol. 102(3): 753–758.
Luo M J and Lai M D. (2001). Identification of differentially expressed genes in normal
mucosa, adenoma and adenocarcinoma of colon by SSH. World J. Gastroenterol.
7(5): 726–731.
Marechal V, Elenbaas B, Piette J, Nicolas J C and Levine A J. (1994). The ribosomal L5
protein is associated with mdm-2 and mdm-2-p53 complexes. Mol. Cell Biol. 14:
7414–7420.
Notterman D A, Alon U, Sierk A J and Levine A J. (2001). Transcriptional gene expression
profiles of colorectal adenoma, adenocarcinoma and normal tissue examined by
oligonucleotide arrays. Cancer Res. 61: 3124–3130.
Otsuka M, Kato M, Yoshikawa T, Chen H, Brown E J, Masuho Y, Omata M and Seki N.
(2001). Differential expression of the L-plastin gene in human colorectal cancer
progression and metastasis. Biochem. Biophys. Res. Commun. 289(4): 876–881.
Pogue-Geile K, Geiser J R, Shu M, Miller C, Wool I G, Meisler A I and Pipas J M. (1991).
Ribosomal protein genes are overexpressed in colorectal cancer: Isolation of a
cDNA clone encoding the human S3 ribosomal protein. Mol. Cell Biol. 11(8): 3842–
3849.
Powell S M, Zilz N, Beazer-Barclay Y, Bryan T M, Hamilton S R, Thibodeau S N,
Vogelstein B and Kinzler K W. (1992). APC mutations occur early during colorectal
tumorigenesis. Nature 339: 235–237.
Prasad S C, Thraves P J, Dritschilo A and Kuettel M R. (1997). Protein expression
changes associated with radiation-induced neoplastic progression of human prostate
epithelial cells. Electrophoresis 18(3–4): 629–637.
Ruland J, Sirard C, Elia A, MacPherson D, Wakeham A, Li L, de la Pompa J L, Cohen S N
and Mak T W. (2001). p53 accumulation, defective cell proliferation and early
embryonic lethality in mice lacking TSG101. Proc. Natl. Acad. Sci. USA 98(4): 1859–
1864.
36
Differentially expressed genes in colorectal carcinoma cases
37
Sparks A B, Morin P J, Vogelstein B and Kinzler K W. (1998). Mutational analysis of the
APC/beta-catenin/Tcf pathway in colorectal cancer. Cancer Res. 58: 1130–1134.
Swearingen M L, Sun D, Bourner M and Weinstein E J. (2003). Detection of differentially
expressed HES-6 gene in metastatic colon carcinoma by combination of suppression
subtractive hybridization and cDNA library array. Cancer Letters 198: 229–239.
Turpin E, Dalle B, de Roquancourt A, Plassa L F, Marty M, Janin A, Beuzard Y and de The
H. (1999). Stress-induced aberrant splicing of TSG101: Association to high tumor
grade and p53 status in breast cancers. Oncogene 18(54): 7834–7837.
Wang Y, Cheong D, Chan S and Hooi S C. (2000). Ribosomal protein L7 a gene is up-
regulated but not fused to the tyrosine kinase receptor as chimeric trk oncogene in
human colorectal carcinoma. Int. J. Oncol. 16(4): 757–762.
Watson K L, Konrad K D, Woods D F and Bryant P J. (1992). Drosophila homolog of the
human S6 ribosomal protein is required for tumor suppression in the hematopoietic
system. Proc. Natl. Acad. Sci. USA 89: 11302–11306.
Zhang L, Zhou W, Velculescu V E, Kern S E, Hruban R H, Hamilton S R, Vogelstein B and
Kinzler K W. (1997). Gene expression profiles in normal and cancer cells. Science
276: 1268–1272.
Zhong Q, Chen Y, Jones D and Lee W H. (1998). Perturbation of TSG101 protein affects
cell cycle progression. Cancer Res. 58(13): 2699–2702.
Zhu G, Gilchrist R, Borley N, Chng H W, Morgan M, Marshall J F, Camplejohn R S, Muir G
H and Hart I R. (2004). Reduction of TSG101 protein has a negative impact on tumor
cell growth. Int. J. Cancer 109(4): 541–547.
