Different gene expression profiles between microsatellite instability-high and microsatellite stable colorectal carcinomas

Abstract

Recent molecular genetic studies have revealed that two major types of genomic instabilities, chromosomal instability (CIN) and microsatellite instability (MSI), exist in colorectal carcinomas. In order to clarify the molecular signature related to the CIN and MSI in colorectal carcinomas, we performed transcriptomic expression analysis on eight microsatellite instability-high (MSI-H) colorectal carcinomas and compared the results obtained with that of nine microsatellite stable (MSS) colorectal carcinomas using oligonucleotide microarrays containing 17 334 known genes and 1331 unknown genes or expression sequence tags (ESTs). Unsupervised two-way hierarchical clustering with 5724 genes successfully classified tumors from normal mucosa, and displayed a distinctive MSI-H carcinomas subgroup. Based on intensive filtering, 57 known genes and eight ESTs were found to be highly relevant to the differentiation of MSI-H and MSS colorectal carcinomas. These genes successfully distinguish the new test set of six MSI-H and five MSS colorectal carcinomas. Many up- and downregulated genes in MSI-H colorectal carcinomas were related to the previously reported phenotypic characteristics; increased mucin production and intense peritumoral immune response in MSI-H carcinomas. Some of these differences were confirmed by semiquantitative reverse transcription-PCR and immunohistochemical analysis. Our findings indicate that there are many different genetic and transcriptomic characteristics between MSI-H and MSS colorectal carcinomas, and some of these differently expressed genes can be used as diagnostic or prognostic markers.

Full text

ORIGINAL PAPER
Different gene expression profiles between microsatellite instability-high
and microsatellite stable colorectal carcinomas
Hyunki Kim
1,2,8
, Suk Woo Nam
3,6,8
, Hwanseok Rhee
2
, Long Shan Li
1,2
, Hyun Ju Kang
1,2
, Kwi Hye
Koh
1,2
, Nam Kyu Kim
4
, Jaehwi Song
3,7
, Edison Tak-Bun Liu
5
and Hoguen Kim*
,1,3
1
Department of Pathology, Yonsei University College of Medicine, Seoul 120-752, Korea;
2
Brain Korea 21 Projects for Medical
Sciences, Yonsei University College of Medicine, Seoul 120-752, Korea;
3
Cancer Metastasis Research Center, Yonsei University
College of Medicine, Seoul 120-752, Korea;
4
Department of Surgery, Yonsei University College of Medicine, Seoul 120-752, Korea;
5
Genome Institute of Singapore, 138672, Singapore;
6
Department of Pathology, College of Medicine, The Catholic University of
Korea, Seoul 137-710, Korea;
7
Department of Genetic Engineering, College of Life Science and Natural Resources, Sungkyunkwan
University, Suwon 440-746, Korea
Recent molecular genetic studies have revealed that two
major types of genomic instabilities, chromosomal instabil-
ity (CIN) and microsatellite instability (MSI), exist in
colorectal carcinomas. In order to clarify the molecular
signature related to the CIN and MSI in colorectal
carcinomas, we performed transcriptomic expression ana-
lysis on eight microsatellite instability-high (MSI-H)
colorectal carcinomas and compared the results obtained
with that of nine microsatellite stable (MSS) colorectal
carcinomas using oligonucleotide microarrays containing
17 334 known genes and 1331 unknown genes or expression
sequence tags (ESTs). Unsupervised two-way hierarchical
clustering with 5724 genes successfully classified tumors
from normal mucosa, and displayed a distinctive MSI-H
carcinomas subgroup. Based on intensive filtering, 57
known genes and eight ESTs were found to be highly
relevant to the differentiation of MSI-H and MSS
colorectal carcinomas. These genes successfully distinguish
the new test set of six MSI-H and five MSS colorectal
carcinomas. Many up- and downregulated genes in MSI-H
colorectal carcinomas were related to the previously
reported phenotypic characteristics; increased mucin pro-
duction and intense peritumoral immune response in MSI-
H carcinomas. Some of these differences were confirmed by
semiquantitative reverse transcription–PCR and immuno-
histochemical analysis. Our findings indicate that there are
many different genetic and transcriptomic characteristics
between MSI-H and MSS colorectal carcinomas, and some
of these differently expressed genes can be used as
diagnostic or prognostic markers.
Oncogene advance online publication, 21 June 2004;
doi:10.1038/sj.onc.1207853
Keywords: colorectal carcinomas; microsatellite in-
stability; oligonucleotide microarray; gene expression
profile; molecular classification
Introduction
The molecular genetics of colorectal carcinomas are
among the best understood of common human cancers.
Recent molecular genetic studies have revealed that two
major types of genomic instabilities, chromosomal
instability (CIN) and microsatellite instability (MSI),
exist in colorectal carcinomas (Lengauer et al., 1998).
The majority of colorectal carcinomas are assigned to
the CIN pathway, which is characterized by a high
frequency of allelic losses, deletions and/or mutations of
tumor suppressor genes, and an abnormal tumor DNA
content (Kinzler and Vogelstein, 1996). The mechanism
of tumorigenesis in CIN tumors involves the activation
of oncogenes and the inactivation of tumor suppressor
genes. The loss of one allele and the inactivation of the
other allele by mutation or promoter methylation are
accepted as a general mechanism of tumor suppressor
gene inactivation.
The other pathway, the MSI pathway, begins with the
inactivation of one of a group of genes responsible for
DNA nucleotide mismatch repair, which leads to
extensive mutations in both repetitive and nonrepetitive
DNA sequences with low frequencies of allelic losses
and rare alterations of tumor DNA content (Ionov et al.,
1993; Thibodeau et al., 1993). The mechanism of
tumorigenesis in high microsatellite instability (MSI-H)
tumors is thought to involve frameshift mutations of
microsatellite repeats within coding regions of the
affected target genes, and the inactivation of these
target genes is believed to directly contribute to tumor
development and progression (Duval and Hamelin,
2002).
In addition to different genetic changes during the
course of tumor development, different tumor progres-
sion patterns have been reported in CIN and MSI-H
colorectal carcinomas. CIN colorectal carcinomas
usually follow the classical adenoma–carcinoma se-
quence model and belong to the microsatellite stable
(MSS) carcinomas. In contrast, adenoma formation in
MSI-H colorectal carcinomas is infrequent, and the
carcinomas that follow the MSI-H pathway reveal
Received 2 February 2004; revised 9 April 2004; accepted 28 April 2004
*Correspondence: H Kim, Department of Pathology, Yonsei Uni-
versity College of Medicine, CPO Box 8044, Seoul, Korea; E-mail:
hkyonsei@yumc.yonsei.ac.kr
8
Contributed equally to this study
Oncogene (2004), 1–8
&
2004 Nature Publishing Group
All rights reserved 0950-9232/04 $30.00
www.nature.com/onc

distinct clinicopathologic patterns (Kim et al., 1994;
Alexander et al., 2001). This is a complex and dynamic
process, which is expected to involve genomic changes in
many genes and altered gene expression profiles. The
identification of genes and of gene expression profiles
that contribute to these two different genetic pathways
should significantly improve both tumor classification
and therapy. To address this issue, we undertook a
comprehensive genomic approach in eight MSI-H and
nine MSS colorectal carcinomas by using oligonucleo-
tide microarray. The molecular dissection of genes
differently expressed in colorectal carcinomas was
provoked by the knowledge that MSI-H colorectal
carcinomas are distinguished from MSS carcinomas.
Here, we demonstrate a molecular signature responsible
for MSI in colorectal carcinoma.
Results
Hierarchical clustering analysis identifies two tumor
subsets characterizing MSI in colorectal carcinoma
We analysed the gene expression of colorectal carcino-
mas and matched normal mucosae by using oligonu-
cleotide microarray. We selected eight MSI-H (cases 1–
8) and 10 MSS colorectal carcinomas (cases 9–18). The
relative expression of each gene in tumor and normal
tissue was measured by comparing its expression ratio to
that of Universal Human Reference RNA (Stratagene).
Adequate expression data could not be obtained from
one MSS carcinoma (case 18) and three matched normal
mucosae (cases 8, 12, and 17) due to artefacts in the
hybridized arrays, or poor quality RNA. Thus, these
were excluded from the analysis. We initially tried
molecular pattern analysis to determine whether our
spotted-oligoarray system was able to identify normal
mucosae and colorectal carcinoma by molecular profil-
ing. With a relevant set of prefiltered 5724 genes (see
‘Materials and methods’), we conducted a complete-
linkage hierarchical clustering analysis of 32 arrays (17
carcinomas and 15 normal mucosae). The results
obtained from these 5724 genes in terms of the greatest
expressional differences between the 17 colorectal
carcinomas and the 15 normal mucosae samples are
displayed in Figure 1. This result showed that normal
mucosae and tumors were grouped together on a cluster
dendrogram, implying that thousands of genes either
contribute or are affected by colonic tumorigenesis. We
next tried to identify a robust set of tumor-related genes
by supervised rank-sum analysis by using the Mann–
Whitney rank-sum test. Genes that were differently
regulated in colorectal carcinoma were identified using
stringent selection criteria (See ‘Material and methods’).
Using these selection criteria, 520 genes were selected as
being upregulated (285 genes) or downregulated (235
genes) compared to noncancerous mucosae. Supplemen-
tary Table 1 lists genes differentially expressed in normal
mucosae and carcinomas.
Furthermore, we found that the carcinomas were
clustered into two unique subgroups and that the MSI-
H carcinomas were differentiated from the other
carcinomas (Figure 1), thus indicating that the MSI-H
carcinomas can be distinguished by their intrinsic
molecular signatures. It is interesting that two distinctive
subgroups in cluster analysis of tumors were correlated
with MSI-H and MSS. Therefore, we next explored the
molecular signature contributing to MSI-H in colorectal
carcinoma.
Identification of differentially expressed genes in MSI-H
and MSS carcinomas
We compared the relative expressions of each gene in the
tumors in two ways. First, we compared the relative
Figure 1 Unsupervised hierarchical clustering analysis of 17 colorectal carcinomas and 15 normal mucosae according to the gene
expression. (a) Genes that passed the filtering criteria were used (gene expression values present in more than 75% in all arrays were
taken, and genes having standard deviations of less than 0.35 were discarded). A total of 5724 genes were selected and applied to
complete linkage hierarchical clustering analysis using the uncentered correlation similarity metric method. Red and green indicate
transcript level above and below the median sample: Universal Human Reference RNA expression ratio for all each gene across all
sample, respectively. (b) Separation of normal mucosal tissues and tumors, and separation of MSI-H and MSS tumors are evident
Gene expression profile in MSI-H colorectal carcinomas
HKimet al
2
Oncogene

Table 1 List of upregulated genes in MSI-H carcinomas
UGCluster Symbol Gene name MSI-H/MSS
a
MSI-H/N
b
MSS/N
c
P-value
d
Hs.76422 PLA2G2A Phospholipase A2, group IIA (platelets, synovial fluid) 7.28 1.82 0.25 0.003
Hs.105806 GNLY Granulysin 6.16 7.10 1.15 0.001
Hs.89603 MUC1 mucin 1, transmembrane 3.35 2.10 0.63 0.009
Hs.380933 mRNA; cDNA DKFZp586O1224 (from clone DKFZp586O1224), mRNA sequence 3.25 5.48 1.69 0.001
Hs.91011 AGR2 Anterior gradient 2 homolog (Xenepus laevis) 3.14 2.85 0.91 0.009
Hs.103707 MUC5AC Mucin 5, subtypes A and C, tracheobronchial/gastric 2.96 2.74 0.93 0.007
Hs.119140 EIF5A Eucaryotic translation initiation factor 5A 2.92 4.34 1.49 0.001
Hs.15114 ARHD ras homolog gene family, member D 2.85 1.70 0.60 0.005
Hs.18760 KIAA1389 protein (Homo sapiens), mRNA sequence 2.73 1.86 0.68 0.003
Hs.343628 SIAT4B Sialyltransferase 4B (beta-galactoside alpha-2,3-sialytransferase) 2.71 1.39 0.51 0.003
Hs.118786 MT2A Metallothionein 2A 2.70 0.49 0.18 0.001
Hs.59413 CTSL Cathepsin L 2.65 2.13 0.81 0.009
Hs.14623 IFI30 Interferon, gamma-inducible protein 30 2.62 3.45 1.32 0.003
Hs.173043 MTA1L1 Metastasis-associated 1-like 1 2.62 2.91 1.11 0.001
cDNA FLJ12124 fis, clone MAMMA1000139 2.58 2.23 0.87 0.005
Hs.8986 C1QB Complement component 1, q subcomponent, beta polypeptide 2.56 1.88 0.73 0.001
Hs.110796 SAR1 SAR1 protein 2.56 2.61 1.02 0.002
Hs.270737 TNFSF13B Tumor necrosis factor (ligand) superfamily, member 13b 2.46 1.93 0.78 0.005
Hs.337461 TRIM15 Tripartite motif-containing 15 2.45 1.09 0.45 0.006
Hs.169401 APOE Apolipoprotein E 2.39 2.26 0.94 0.003
Hs.433300 FCER1G Fc fragment of IgE, high-affinity I, receptor for gamma polypeptide 2.35 2.64 1.12 0.003
Hs.2730 HNRPL Heterogeneous nuclear ribonucleoprotein L 2.34 2.96 1.26 0.001
Hs.95655 SECTM1 Secreted and transmembrane 1 2.33 0.63 0.27 0.001
Hs.151734 NUTF2 Nuclear transport factor 2 2.33 0.27 0.11 0.007
Hs.184390 LOC57168 Similar to aspartate beta hydroxylase (ASPH) 2.30 2.05 0.89 0.009
Hs.13015 DNAJC1 DnaJ (Hsp40) homolog, subfamily C, member 1 2.30 0.95 0.41 0.002
Hs.104741 TOPK T-LAK cell-originated protein kinase 2.28 3.48 1.52 0.001
Hs.380778 MT1L Metallothionein 1L 2.23 0.53 0.24 0.004
Hs.75367 SLA Src-like adaptor 2.19 1.25 0.57 0.002
Hs.325978 IL18BP Interleukin 18-binding protein 2.12 2.00 0.94 0.004
Hs.55968 GALNT5 UDP-N-acetyl-alpha-
D-galactosamine:polypeptide N-acetylgalactosaminyltransferase 5 (GalNAc-T5) 2.09 2.03 0.97 0.005
Hs.3164 NUCB2 Nucleobindin 2 2.08 1.78 0.86 0.003
Hs.198248 B4GALT1 UDP-Gal:betaGlcNAc beta 1,4-galactosyltransferase, polypeptide 1 2.07 1.00 0.48 0.001
Hs.91448 DUSP14 Dual-specificity phosphatase 14 2.06 1.81 0.88 0.005
Hs.1695 MMP12 Matrix metalloproteinase 12 (macrophage elastase) 2.05 1.94 0.95 0.007
Hs.237856 PHT2 Peptide transporter 3 2.05 2.10 1.02 0.005
Hs.73010 IFNW1 Interferon, omega 1 2.04 2.75 1.35 0.001
Hs.75703 CCL4 Chemokine (C-C motif) ligand 4 2.03 1.72 0.85 0.001
Hs.75627 CD14 CD14 antigen 2.01 1.32 0.66 0.006
a
MSI-H/MSS, average ratio of intensity of eight MSI-H carcinomas/average ratio of intensity of nine MSS carcinomas.
b
MSI-H/N, average ratio of intensity of eight MSI-H carcinomas/average ratio
of intensity of 15 normal mucosae.
c
MSS/N, average ratio of intensity of nine MSS carcinomas/average ratio of intensity of 15 normal mucosae.
d
P-value, between eight MSI-H carcinomas and nine
MSS carcinomas from Mann–Whitney rank-um test
Gene expression profile in MSI-H colorectal carcinomas
HKimet al
3
Oncogene

gene expressions of carcinomas with respect to Uni-
versal Human Reference RNA, and second we com-
pared the gene expressions of each tumor by analysing
the differences in the relative gene expression of each
gene with respect to the average expressions in 15
matched normal mucosal tissues. Both methods pro-
duced the same result. Based on the level of gene
expression, colorectal carcinomas were separated into
two major groups. The seven MSI-H and eight MSS
carcinomas showed strong tendency to form separate
clusters on two distinct dendrogram branches, and one
MSI-H and one MSS carcinomas formed a minor
branch (Figure 1). Comparison with the cluster dendro-
gram showed several groups of genes common or
specific to the genetic instability type. Supervised
analysis identified 39 genes showing at least twofold
higher mRNA expressions (Table 1), and 26 genes
showing at least 0.5-fold lower expressions in MSI-H
carcinomas compared with MSS carcinomas (Table 2)
(Figure 2a).
Validation test of 65 genes as molecular classifiers for
MSI-H colorectal carcinomas
In order to determine whether these 65 outlier genes
can be used as unique molecular signatures discriminat-
ing the MSI-H colorectal carcinomas, we tested
for expression profiling of another set of 12 colorectal
carcinomas including six MSI-H and six MSS carcino-
mas using new batch of spotted-oligonucleotide
microarrays. Note that these arrays were manufactured
at the DNA microarray core facility of the Department
of Pathology, College of medicine, The Catholic
University of Korea, but containing the same genetic
elements as the previous ones. In this test, we noted
that two genes out of a total of 65 outlier genes
were missing during primary data extraction probably
due to spotting error or stringent selection criteria. We
then performed hierarchical cluster analysis of the 11
colorectal carcinomas by using selected 63 outlier genes
(one array was eliminated due to poor quality of
hybridization result). As expected, the classification of
63 outlier genes resulted in two distinct subgroups on
cluster dendrogram, exhibiting a clear separation
between all six MSI-H and five MSS carcinomas. This
implies that, at least, our 63 outlier genes could be
feasible genetic elements for the molecular classification
of MSS and MSI colorectal carcinomas (Supplementary
Figure 1).
Validation of differentially expressed genes
In order to examine the reliability of microarray data,
we selected two upregulated genes, mucin 1 (MUC1) and
Table 2 List of downregulated genes in MSI-H carcinomas
UGCluster Symbol Gene name MSI-H/MSS
a
MSI-H/N
b
MSS/N
c
P value
d
Hs.24395 CXCL14 Chemokine (C-X-C motif) ligand 14 0.23 0.29 1.23 0.004
Hs.87296 cDNA FLJ20269 fis, clone HEP01293, mRNA sequence 0.26 1.15 4.37 0.005
Hs.9029 HAIK1 Type I intermediate filament cytokeratin 0.29 1.08 3.80 0.002
Hs.22785 GABRE Gamma-aminobutyric acid (GABA) A receptor, epsilon 0.32 0.73 2.24 0.001
Hs.282975 CES2 Carboxylesterase 2 (intestine, liver) 0.32 0.16 0.51 0.004
Hs.127337 AXIN2 Axin 2 (conductin, axil) 0.34 1.25 3.65 0.007
Hs.179704 MEP1A Meprin A, alpha (PABA peptide hydrolase) 0.36 0.10 0.29 0.005
Hs.166705 GPR49 G protein-coupled receptor 49 0.39 1.19 3.03 0.006
Hs.275775 SEPP1 Selenoprotein P, plasma, 1 0.44 0.12 0.27 0.009
Hs.101850 RBP1 Retinol-binding protein 1, cellular 0.44 0.82 1.87 0.005
Hs.38738 CLDN15 Claudin 15 0.44 0.75 1.72 0.004
Hs.278997 CESR Carboxylesterase-related protein 0.44 0.55 1.26 0.001
Hs.272245 cDNA FLJ11170 fis, clone PLACE1007301,
mRNA sequence
0.45 0.29 0.65 0.010
Hs.248019 POU4F3 POU domain, class 4, transcription factor 3 0.45 0.67 1.51 0.007
Hs.49476 cDNA FLJ12815 fis, clone NT2RP2002546,
mRNA sequence
0.46 0.54 1.15 0.009
Hs.1298 MME Membrane metallo-endopeptidase (neutral endopeptidase,
enkephalinase, CALLA, CD10)
0.46 0.84 1.82 0.002
Hs.86327 HOXB9 Homeo box B9 0.46 0.76 1.66 0.004
Hs.18457 FLJ20315 Hypothetical protein FLJ20315 0.47 1.28 2.71 0.002
Hs.151469 CASK Calcium/calmodulin-dependent serine protein kinase
(MAGUK family)
0.47 0.96 2.06 0.002
Hs.130714 HSPC323 (Homo sapiens), mRNA sequence 0.47 0.62 1.32 0.007
Hs.192927 PPP1R14D Protein phosphatase 1, regulatory (inhibitor) subunit 14D 0.47 0.31 0.66 0.004
Hs.36927 HSP105B Heat shock 105 kDa 0.48 2.50 5.18 0.005
Hs.274351 ZDHHC9 Zinc-finger, DHHC domain containing 9 0.48 1.00 2.08 0.001
Hs.151301 CADPS Ca2+-dependent activator protein for secretion 0.48 1.07 2.22 0.002
Hs.380831 FOXO3A Forkhead box O3A 0.48 0.54 1.13 0.001
Hs.85835 cDNA: FLJ22841 fis, clone KAIA4844, mRNA sequence 0.49 0.31 0.63 0.001
a
MSI-H/MSS, average ratio of intensity of eight MSI-H carcinomas/average ratio of intensity of nine MSS carcinomas.
b
MSI-H/N, average ratio of
intensity of eight MSI-H carcinomas/average ratio of intensity of 15 normal mucosae.
c
MSS/N, average ratio of intensity of nine MSS carcinomas/
average ratio of intensity of 15 normal mucosae.
d
P-value, between eight MSI-H carcinomas and nine MSS carcinomas from Mann–Whitney rank-
sum test
Gene expression profile in MSI-H colorectal carcinomas
HKimet al
4
Oncogene

phospholipase A2 group IIA (PLA2G2A) and one
downregulated gene, chemokine (C-X-C motif) ligand
14 (CXCL14) in MSI-H carcinomas, and one upregu-
lated gene AXIN2 in MSS carcinomas. The expressions
of these genes were analysed by semiquantitative reverse
transcription (RT)–PCR using the RNA samples used
for the microarray analysis. The results of semiquanti-
tative RT–PCR were consistent for all the four genes
(Figure 2b). We then performed an immunohistochem-
ical analysis of MUC1 and mucin 5, subtypes A and C
(MUC5AC) by using tissue array containing 57 MSI-H
and 109 MSS carcinomas and their matched normal
mucosae, then compared the results to that of micro-
array data. Both genes showed similar levels of mRNA
and protein expression. The difference in immunohisto-
chemical expression patterns between MSI-H and MSS
colorectal carcinomas were consistently demonstrated in
the 166 colorectal carcinomas. Expressions of MUC1
and MUC5AC were increased in MSI-H colorectal
carcinomas (Figure 2c). MUC1 expression was found in
21 of 57 MSI-H and 23 of 109 MSS carcinomas (Po0.029),
and MUC5AC expression was found in 11 of 57 MSI-H
and five of 109 MSS carcinomas (Po0.002).
Discussion
In this study, we identified distinct gene expression
profiles in MSI-H and MSS colorectal carcinomas by
oligonucleotide microarray analysis. We identified a
large number of genes that were differentially expressed
in normal colonic mucosae and colorectal carcinomas.
We also found two unique subgroups within carcino-
mas, namely, the MSI-H and MSS carcinomas,
which showed different hierarchical cluster patterns.
These data provide an insight into the extent of
gene expression differences underlying different genetic
instabilities in colorectal carcinomas (MSI-H vs
MSS).
We identified 520 (2.8% of those detected) genes or
expression sequence tag (ESTs) that appear to be
differentially expressed (more than twofold and less
than 0.5-fold) in carcinomas vs normal appearing
mucosa. Many of these dysregulated genes at the
transcript level were in accordance with previous reports
(Alon et al., 1999; Kitahara et al., 2001; Notterman
et al., 2001; Birkenkamp-Demtroder et al., 2002,
Bertucci et al., 2004). For example, CA4, CNN1,and
Figure 2 Expression analysis of differently expressed genes between MSI-H and MSS carcinomas. (a) Transcriptional expression
profile of 65 differently expressed genes between MSI-H and MSS carcinomas. Red and green indicate transcript level above and below
the median tumor: Universal Human Reference RNA expression ratio for 65 genes, respectively. (b) Expression analysis of hMLH1,
PLA2G2A, MUC1, AXIN2, and CXCL14 by semiquantitative RT–PCR. Downregulation of hMLH1 in MSI-H carcinomas is evident.
Upregulations of PLA2G2A and MUC1, and downregulations of CXCL14 are found in MSI-H carcinomas. The downregulation of
PLA2G2A and the upregulation of AXIN2 in MSS carcinomas are also shown. (c) Tumor histology and immunohistochemical analysis
of MUC1 and MUC5AC in one MSI-H colorectal carcinoma. Increased expressions of MUC1 and MUC5AC are evident. Summary
of MUC1 and MUC5AC expressions in the 57 MSI-H and 109 MSS colorectal carcinomas. In the MSI-H carcinomas, frequent
expressions of MUC1 and MUC5AC are evident
Gene expression profile in MSI-H colorectal carcinomas
HKimet al
5
Oncogene

FCGBP were downregulated, whereas CCNB2, CSE1L,
NME1, RHEB2, and UBE2C were upregulated in
carcinomas. The functional categories of dysregulated
genes in carcinomas revealed that the genes involved
in nucleic acid metabolism and cell proliferation
are overexpressed in carcinomas. The 285 upregulated
genes in carcinomas belonged to the functional
categories (http://www.geneontology.org) of nucleic
acid metabolism (10%), cell proliferation (10%),
signal transduction (7%), and biosynthesis (6%).
The 235 downregulated genes in carcinomas belonged
to transport (11%), signal transduction (8%),
response to biotic stimulus (6%), and lipid metabolism
(4%).
We successfully identified differential gene expression
in two distinct subgroups of colorectal carcinomas,
MSI-H and MSS carcinomas, by genome wide
gene expression analysis. MSI-H carcinomas arise with
the inactivation of one of a group of genes responsible
for DNA nucleotide mismatch repair, which leads to
extensive mutations in both repetitive and nonrepetitive
DNA sequences with low frequencies of allelic
losses and rare alterations of tumor DNA content
(Ionov et al., 1993; Thibodeau et al., 1993). In contrast,
the MSS carcinomas are characterized by a high
frequency of allelic losses, deletions and/or mutations
of tumor suppressor genes, and abnormal tumor
DNA content (Kinzler and Vogelstein, 1996). Several
clinicopathologic findings are also reported to be
different in MSI-H and MSS carcinomas. MSI-H
colorectal carcinomas show a preponderance for the
right side, poor differentiation, mucin formation,
and peritumoral lymphocytic infiltration (Kim et al.,
1994; Alexander et al., 2001). However, these
clinicopathologic features did not account for the
separation of MSI-H and MSS colorectal carcinomas
(Alexander et al., 2001; Ward et al., 2001). Based on
these distinct molecular and clinical natures, MSI-H
and MSS tumors are expected to be distinguished
by an expression profiling. Also, it is possible to identify
novel differentially expressed genes that may be related
to the MSI-H or CIN molecular pathway. Currently,
two studies are available for the gene expression
differences between MSI-H and MSS tumors. One
study used colorectal cancer cell lines (Dunican et al.,
2002) and the other study used tumor tissues (Mori et al.,
2003). Both studies reported a number of differentially
expressed genes between MSI-H and MSS tumors.
However, the number of reported genes was small
and the usefulness of these genes as molecular
classifiers is not defined statistically. In this study,
we found that MSI-H and MSS carcinomas can be
divided by unsupervised hierarchical clustering analysis,
and identified 65 genes that completely differentiate
MSI-H and MSS carcinomas by supervised analysis.
These genes successfully differentiated the other test set
of 11 colorectal carcinomas as MSI-H and MSS
carcinomas. These findings indicate that MSI-H and
MSS carcinomas are distinct types of colorectal
carcinomas, and the 65 genes can be used as molecular
classifiers.
Many of the dysregulated genes in MSI-H and MSS
carcinomas reported herein make a good deal of sense
even though some of them are unsuspected in colorectal
carcinogenesis. For example, MUC5AC and MUC1 are
involved in mucin production, thus their increased
expression is in good agreement with the high frequency
of mucin production in MSI-H carcinomas (Kim et al.,
1994). Moreover, MUC1 expression in colorectal
carcinoma has been reported to be implicated in tumor
progression and invasion in colorectal carcinoma
(Hiraga et al., 1998). An association between MUC5AC
overexpression and MSI-H colorectal carcinomas had
been reported (Biemer-Huttmann et al., 2000). These
finding suggest that the proportions of the mucin
component are different between MSI-H and MSS
carcinomas. Increased granulysin, a protein present in
the cytotoxic granules of cytotoxic T lymphocytes and
natural killer cells (Stenger et al., 1999), in MSI-H
tumors might represent infiltrating intratmoral T cell in
our MSI-H carcinomas. Intratumoral and peritumoral
lymphocytic infiltration occurs frequently in MSI-H
carcinomas, and a large proportion of cytotoxic T cells
among infiltrating lymphocytes had been reported
(Guidoboni et al., 2001). These findings indicate that
the molecular clusters that are involved in the
differentiation toward MSI-H carcinomas are closely
related with the clinicopathological characteristics of
MSI-H carcinomas. The large numbers of molecular
types related to the phenotypic alterations of MSI-H
carcinomas are believed to enable MSI-H and MSS
carcinoma differentiation at the gene level, while the
morphologic classification of these tumors using clin-
icopathologic parameters is difficult (Alexander et al.,
2001).
Several genes were found that were related to the
APC/b-catenin pathway. PLA2G2A was downregulated
and AXIN2 upregulated in MSS carcinomas. These
altered expressions are interesting because these two
molecules may be directly related with the abnormal
APC/b-catenin pathway in MSS carcinomas. In APC
Min/
þ
mouse, a mouse model of familial adenomatous
polyposis, PLA2G2A was proposed to be a major
modifier, which suppress the severity of intestinal
neoplasia (Cormier et al., 1997; MacPhee et al., 1995).
This finding suggests that the abnormal APC/b-catenin
pathway can be enhanced by the reduced expression of
PLA2G2A in MSS colorectal carcinomas. It has been
demonstrated that AXIN1 and/or AXIN2 form a
multiprotein complex with APC, glycogen synthase
kinase 3b, and b-catenin (Ikeda et al., 1998). Moreover,
AXIN2 induction by b-catenin activation has been
reported (Leung et al., 2002). Therefore, the specific
overexpression of AXIN2 in our MSS carcinomas
reflects the selected activation of the APC/b-catenin
pathway in MSS carcinomas.
In summary, we identified several candidate colorectal
carcinoma markers and molecular clusters capable of
differentiating MSI-H and MSS colorectal carcinomas.
Moreover, genes like these, which help to define specific
genomic instability, may have potential values in
differential diagnosis and prognosis evaluation.
Gene expression profile in MSI-H colorectal carcinomas
HKimet al
6
Oncogene

Materials and methods
Case selection
Eight cases (cases 1–8), confirmed as MSI-H colorectal
carcinoma, and 10 cases (cases 9–18) of MSS colorectal
carcinoma were included in this study. In each case, grossly
normal mucosa remote from the tumor was included as a
control. All cases were selected from consecutively identified
cases at the Gastrointestinal Tumor Working Group Tissue
Bank at Yonsei University Medical Center (Seoul, Korea)
between December 1996 and November 1999. We also selected
six MSI-H and six MSS carcinomas for the test set. These cases
were also selected from the Gastrointestinal Tumor Working
Group Tissue Bank at Yonsei University Medical Center
between December 2000 and November 2003. RNAs, proteins,
and DNAs were extracted from fresh frozen tissues. Tumor
specimens were microdissected on a cryostat and fractionated
to enrich the tumor cell population. Genomic DNA prepara-
tion and the determination of the MSI status of eight MSI-H
carcinomas have been previously reported (Kim et al., 2001;
Kim et al., 2002).
Microarray formulation
High-density spotted-oligonucleotide microarrays were manu-
factured at the array core facility at the Genome Institute of
Singapore. The human Oligolibraryt was purchased from
Compugen/Sigma-Genosys. It consisted of 18 861 oligonucleo-
tides, which represent 18 664 LEADSt clusters plus 197
controls glyceraldehydes-3-phosphate dehydrogenase. A total
of 60 mers of synthesized oligos were robotically printed and
processed. For analysis of the test set, oligonucleotide arrays
containing the same genetic elements were manufactured at the
core facility of Department of Pathology, College of Medicine,
The Catholic University of Korea.
RNA preparation and hybridization
Total RNA was extracted from 100–200 mg of microdissected
frozen tissues using a RNeasy Mini kit (Qiagen, Valencia, CA,
USA) according to the manufacturer’s instructions. Total
RNA (20 mg) was used as input for cDNA targets synthesis as
previously described (DeRisi et al., 1997). The targets and
Universal Human Reference RNA (Stratagene, La Jolla, CA,
USA) were hybridized to an oligonucleotide microarray
containing 18 664 probe sets representing 18 664 unique
(LEADSt) genes, and the array was scanned using GenePix
scanners. Expression values for each gene were calculated by
using GenePix Pro 4.0 analysis software. Some of the
hybridizations were carried out in duplicate with flouro-
reversal to compensate for the different chemical properties of
the fluorescent dye molecules and for potential biases
associated with normalization.
Of 18 colorectal carcinomas from 18 patients samples
processed, 17 colorectal carcinomas and 15 corresponding
normal mucosae were used for subsequent data analysis. The
remaining one tumor and three normal samples either failed
quality control restrictions concerning the amount and quality
of RNA, as assessed by agarose gel electrophoresis, or yielded
poor-quality scans.
Classification of colorectal carcinoma by molecular pattern
analysis
Unsupervised hierarchical clustering analysis was used to
analyse the classifications of 17 colorectal carcinomas and of
15 noncancerous mucosae according to gene expression. We
used a data set of genes that satisfied the filtering criteria (genes
having more than 75% of log-transformed ratio values
presenting in across all arrays were chosen and genes having
less than 0.35 standard deviations of log-transformed ratio
were discarded). The selected gene data set was then applied to
complete-linkage hierarchical clustering analysis using the
uncentered correlation similarity metric method in Cluster
version 2.20, and the resulting expression map was visualized
with Treeview version 1.60 (http://rana.lbl.gov/EisenSoftwar-
e.htm).
Identification of genes responsible for MSI in colorectal
carcinoma
To determine whether colorectal carcinomas can be classified
as MIS-H or MSS according to their transcriptomic patterns,
17 arrays (eight arrays of MSI-H and nine of MSS) of
colorectal carcinomas were assessed using unsupervised
hierarchical clustering analysis. To detect differentially ex-
pressed genes in a given subclass, we ranked the genes using
the Mann–Whitney rank-sum test. Outlier genes responsible
for MSI-H and MSS classifications were selected by Po0.01.
In addition, significant outlier subset genes were further
narrowed by filtering genes showing greater than 72-fold
expression changes in colorectal carcinoma compared to the
average values from normal mucosae.
Semiquantitative RT–PCR
First-strand cDNA was synthesized from 1 mg of total RNA
using random hexamer primers (Qiagen) and M-MLV Reverse
Transcriptase (Invitrogen, San Diego, CA, USA) according to
the manufacturer’s instructions. cDNA (20 ng) from each
sample was used in each reaction. All RT–PCR primers were
designed to contain an exon–exon junction. Reaction was
performed with primer for the specific genes and for the b-actin
in duplex reaction. The range of linear amplification for each
gene and b-actin was examined with serial PCR cycles, and
optimal PCR cycles were determined. For each gene, Gene
Bank accession number, the sequences of the forward primer
and reverse primer, respectively, are as follows: human MutL
homolog 1 (hMLH1), NM_000249, 5
0
-AGATCACGGTG
GAGGACCTT-3
0
,5
0
-CCACATTCTGGGGACTGATT-3
0
;
PLA2G2A, NM_000300, 5
0
-CCTGGGGATACAACTCTG
GA-3
0
,5
0
-TTGCACAGGTGATTCTGCTC-3
0
; MUC1, NM_
002456, 5
0
-AGTTCAGGCCAGGATCTGTG-3
0
,5
0
-CCCCT
ACAAGTTGGCAGAAG-3
0
; Axin2, NM_004655, 5
0
-GTGT
GAGGTCCACGGAAACT-3
0
,5
0
-TTCATCCTCTCGGATC
TGCT-3
0
; CXCL14 , NM_004887, 5
0
-TGTGGACGGGTC
CAAATG-3
0
,5
0
-CTGCGCTTCTCGTTCCAG-3
0
; b-actin,
NM_001101, 5
0
-TGCTATCCCTGTACGCCTCT-3
0
,5
0
-GTA
CTTGCGCTCAGGAGGAG-3
0
. b-Actin was used as an
internal control. After RT–PCR, 5 ml aliquots of the products
were subjected to 2% agarose gel electrophoresis and stained
with ethidium bromide.
Immunohistochemical analysis
Colorectal carcinoma tissue arrays containing 57 MSI-H
carcinomas and 109 MSS carcinomas were constructed from
formal formalin-fixed and paraffin-embedded tissues by
Petagen Inc. (Seoul, Korea) and these arrayed slides were
used for the immunostaining of MUC1and MUC5AC. All
eight MSI-H carcinomas and nine MSS carcinomas used in
the oligonucleotide microarray analysis were included in the
tissue array. Deparaffinization and rehydration were per-
formed using xylene and alcohol. The sections were incubated
for 1 h at room temperature with antibodies against MUC1
Gene expression profile in MSI-H colorectal carcinomas
HKimet al
7
Oncogene

(Novocastra Laboratories Ltd, Newcastle, UK) and MU-
C5AC (Novocastra Laboratories Ltd). Avidin–biotin complex
methodology was employed. The chromogen was diamino-
benzidine and counterstaining was carried out with hematox-
ylin. The expression of the two gene products was categorized
as expressed and negative. In the evaluation of MUC5AC,
cases with definite cytoplasmic staining in more than 10% of
the tumor cells were categorized as expressed, and cases with
definite cytoplasmic staining in less than 10% of the tumor
cells were categorized as negative. In the evaluation of MUC1,
cases with definite membrane and/or cytoplasmic staining in
more than 10% of the tumor cells were categorized as
expressed, and less than 10% of the tumor cells or complete
absence of staining were as negative.
Acknowledgements
This study was supported by a grant of the Korea Health 21
R&D Project, Ministry of Health & Welfare, Republic of
Korea (03-PJ10-PG6-GP01-0002) and the Cancer Metastasis
Research Center at Yonsei University.
References
Alexander J, Watanabe T, Wu TT, Rashid A, Li S and
Hamilton SR. (2001). Am. J. Pathol., 158, 527–535.
Alon U, Barkai N, Notterman DA, Gish K, Ybarra S, Mack
D and Levine AJ. (1999). Proc. Natl. Acad. Sci. USA, 96,
6745–6750.
Bertucci F, Salas S, Eysteries S, Nasser V, Finetti P, Ginestier
C, Charafe-Jauffret E, Loriod B, Bachelart L, Montfort J,
Victorero G, Viret F, Ollendorff V, Fert V, Giovaninni M,
Delpero JR, Nguyen C, Viens P, Monges G, Birnbaum D
and Houlgatte R. (2004). Oncogene, 23, 1377–1391.
Biemer-Huttmann AE, Walsh MD, McGuckin MA, Simms
LA, Young J Leggett BA and Jass JR. (2000). Clin. Cancer
Res., 6, 1909–1916.
Birkenkamp-Demtroder K, Christensen LL, Olesen SH,
Frederiksen CM, Laiho P, Aaltonen LA, Laurberg S,
Sorensen FB, Hagemann R and Orntoft TF. (2002). Cancer
Res., 62, 4352–4363.
Cormier RT, Hong KH, Halberg RB, Hawkins TL, Richard-
son P, Mulherkar R, Dove WF and Lander E. (1997). Nat.
Genet., 17, 88–91.
DeRisi JL, Iyer VR and Brown PO. (1997). Science, 278, 680–
686.
Dunican DS, McWilliam P, Tighe O, Parle-McDermott A and
Croke DT. (2002). Oncogene, 21, 3253–3257.
Duval A and Hamelin R. (2002). Cancer Res., 62, 2447–2454.
Guidoboni M, Gafa R, Viel A, Doglioni C, Russo A, Santini
A, Del Tin L, Macri E, Lanza G, Boiocchi M and Dolcetti
R. (2001). Am. J. Pathol., 159, 297–304.
Hiraga Y, Tanaka S, Haruma K, Yoshihara M, Sumii K,
Kajiyama G, Shimamoto F and Kohno N. (1998). Oncology,
55, 307–319.
Ikeda S, Kishida S, Yamamoto H, Murai H, Koyama S and
Kikuchi A. (1998). EMBO J., 17, 1371–1384.
Ionov Y, Peinado MA, Malkhosyan S, Shibata D and Perucho
M. (1993). Nature, 363, 558–561.
Kim H, Jen J, Vogelstein B and Hamilton SR. (1994). Am. J.
Pathol., 145, 148–156.
Kim N-G., Choi YR, Baek MJ, Kim YH, Kang H,
Kim NK, Min JS and Kim H. (2001). Cancer Res., 61,
36–38.
Kim N-G, Rhee H, Li LS, Kim H, Lee JS, Kim JH, Kim NK
and Kim H. (2002). Oncogene, 21, 5081–5087.
Kinzler KW and Vogelstein B. (1996). Cell, 87, 159–170.
Kitahara O, Furukawa Y, Tanaka T, Kihara C, Ono K,
Yanagawa R, Nita M E, Takagi T, Nakamura Y and
Tsunoda T. (2001). Cancer Res., 61, 3544–3549.
Lengauer C, Kinzler KW and Vogelstein B. (1998). Nature,
396, 643–649.
Leung JY, Kolligs FT, Wu R, Zhai Y, Kuick R, Hanash S,
Cho KR and Fearon ER. (2002). J. Biol. Chem., 277, 21657–
21665.
MacPhee M, Chepenik KP, Liddell RA, Nelson KK,
Siracusa LD and Buchberg AM. (1995). Cell, 81,
957–966.
Mori Y, Selaru FM, Sato F, Yin J, Simms LA, Xu Y, Olaru A,
Deacu E, Wang S, Taylor JM, Young J, Leggett B, Jass JR,
Abraham JM, Shibata D and Meltzer SJ. (2003). Cancer
Res., 63, 4577–4582.
Notterman DA, Alon U, Sierk AJ and Levine AJ. (2001).
Cancer Res., 61, 3124–3130.
Stenger S, Rosat JP, Bloom BR, Krensky AM and Modlin
RL. (1999). Immunol. Today, 20, 390–394.
Thibodeau SN, Bren G and Schaid D. (1993). Science, 260,
816–819.
Ward R, Meagher A, Tomlinson I, O’Connor T, Norrie M,
Wu R and Hawkins N. (2001). Gut, 48, 821–829.
Supplementary Information accompanies the paper on Oncogene website (http://www.nature.com/onc).
Gene expression profile in MSI-H colorectal carcinomas
HKimet al
8
Oncogene