Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Molecular karyotyping of human hepatocellular carcinoma using single-nucleotide polymorphism arrays


Genomic amplification of oncogenes and inactivation of suppressor genes are critical in the pathogenesis of human cancer. To identify chromosomal alterations associated with hepatocarcinogenesis, we performed allelic gene dosage analysis on 36 hepatocellular carcinomas (HCCs). Data from high-density single-nucleotide polymorphism arrays were analysed using the Genome Imbalance Map (GIM) algorithm, which simultaneously detects DNA copy number alterations and loss of heterozygosity (LOH) events. Genome Imbalance Map analysis identified allelic imbalance regions, including uniparental disomy, and predicted the coexistence of a heterozygous population of cancer cells. We observed that gains of 1q, 5p, 5q, 6p, 7q, 8q, 17q and 20q, and LOH of 1p, 4q, 6q, 8p, 10q, 13q, 16p, 16q and 17p were significantly associated with HCC. On 6q24–25, which contains imprinting gene clusters, we observed reduced levels of PLAGL1 expression owing to loss of the unmethylated allele. Finally, we integrated the copy number data with gene expression intensity, and found that genome dosage is correlated with alteration in gene expression. These observations indicated that high-resolution GIM analysis can accurately determine the localizations of genomic regions with allelic imbalance, and when integrated with epigenetic information, a mechanistic basis for inactivation of a tumor suppressor gene in HCC was elucidated.


Genomic amplification of oncogenes and inactivation of tumor suppressor genes are frequently associated with carcinogenesis. Comparative genomic hybridization (CGH) has been used extensively to detect genome-wide copy number changes in various types of cancers and to determine the localization of expression of many oncogenes and tumor suppressor genes (Kallioniemi et al., 1992). Owing to its low resolution (greater than 20 Mb), CGH has not been useful in identifying micro-deletions or -amplifications. Array-based CGH using genomic DNA or cDNA clones has been developed to achieve higher resolution (Pinkel et al., 1998; Pollack et al., 1999). Furthermore, Lucito et al. (2003) have developed a new methodology, called ROMA (representational oligonucleotide microarray analysis), which achieves this purpose with an average resolution of 30 kb throughout the genome.

Despite the improvement in resolution, array-based CGH and high-density oligonucleotide arrays used for ROMA could only detect changes at the gene level and not at the level of individual alleles. Allelic changes, including hemizygous deletion with a gain of the opposite allele, so-called uniparental disomy (UPD) or trisomy (UPT), are important in elucidating the molecular mechanisms of cancer (Engel, 1980). For example, accurate determination of the copy number of each allele separately in the UPD region, which may contain a region of methylation or null mutation of tumor suppressor genes, may allow the mechanism of carcinogenesis to be determined more clearly and comprehensively (Grundy et al., 1994; Murthy et al., 2002; Raghavan et al., 2005). In addition, accurate measurement of changes in copy number using high-resolution methods will help in the detection of heterogeneous populations among cancer cells (Benetkiewicz et al., 2005; Buckley et al., 2005).

Loss of heterozygosity (LOH) status can be analysed comprehensively by allelotyping of the genomic DNA from cancer cells using hundreds of polymorphic markers from each chromosomal arm (Weissenbach et al., 1992). Polymerase chain reaction (PCR)-based measurement of simple sequence length polymorphisms is a reliable method for detecting LOH, but is both expensive and labor intensive. Recently, global and high-resolution analyses of LOH using single-nucleotide polymorphism (SNP) arrays, which was originally designed for high-throughput SNP analysis (Wang et al., 1998; Kennedy et al., 2003), have been performed in various cancers (Lindblad-Toh et al., 2000; Mei et al., 2000; Lieberfarb et al., 2003; Janne et al., 2004). Comparing this method to PCR-based microsatellite analysis directly, Hoque et al. (2003) validated the accuracy of LOH analysis using SNP arrays. Furthermore, we and other groups have developed novel algorithms for detecting copy number changes at both the gene level and the allelic level (Bignell et al., 2004; Zhao et al., 2004; Ishikawa et al., 2005; Nannya et al., 2005).

In the present study, we applied a newly developed Genome Imbalance Map (GIM; Ishikawa et al., 2005) algorithm to analyse genomic alterations in clinical specimens of hepatocellular carcinoma (HCC). We demonstrated the robustness of GIM analysis based on genotyping arrays in the accurate analysis of copy number alterations and allelic imbalance in the cancer genome in a single experiment. Furthermore, by integrating the gene expression data with genomic alterations, we have demonstrated that genomic alterations are reflected in the gene expression profile (Kano et al., 2003; Midorikawa et al., 2004). Our results confirmed a positive correlation between genome dosage and transcription.


Genome Imbalance Map can detect regions of genome imbalance in hepatocellular carcinoma samples

Gene and allele copy number analyses were performed in all 36 HCC samples by taking the copy numbers of peripheral blood lymphocytes as matched controls (Figure 1). Using a 10K SNP array, the mean call rate and mean heterozygous call rate in HCC and controls were 93.4±2.7 and 27.4±3.2 and 95.9±1.9 and 31.0±0.9%, respectively. The signal intensity ratio and the GIM figure in each sample from the array data are available online (

Figure 1

Genome Imbalance Map (GIM) of a representative hepatocellular carcinoma (HCC) sample. (a) Total gene dosage analysis across the whole genome in patient 21. Annotated copy numbers were estimated from the allelic dosage analysis as described elsewhere. Regions of copy number alteration are indicated by arrows. The inset shows amplification of c-MET on 7q31.2 by fluorescent in situ hybridization (FISH). Orange: RP11-163C9 (c-MET); green: CEP7. (b) Allelic dosage analysis across the whole genome in patient 21. Alleles with higher and lower copy numbers are connected by red and blue lines, respectively. Regions with gain, loss of heterozygosity (LOH), amplification and uniparental disomy (UPD) are indicated with arrows. Gain regions, 1q21.3-1qter, 5p, 5q, 7p, 7qcen-7q31.31, 12p13.2–12p13.3, 13q32.3-13qter, 14q, 20p, 20q; LOH regions, 6q26–6q27, 8p12.2–8p23, 10q22.2-10qter, 12pcen-12p12.3, 12q, 17p; amplified region, 7q31.2; and UPD region, 13qcen-13q32.1.

Gains on chromosome arms of HCC were observed on 1q (72.2%), 5p (25.0%), 5q (30.5%), 6p (33.3%), 7q (22.2%), 8q (61.1%), 17q (25.0%) and 20q (25.0%), and LOH was detected on 1p (22.2%), 4q (27.7%), 6q (27.7%), 8p (55.5%), 10q (33.3%), 13q (47.2%), 16p (25.0%), 16q (36.1%) and 17p (66.7%) (Figure 2). Amplification was also observed on 5p15, 6p24, 7q31, 10p11, 11q14, 11q32 and 17q12 in each case, and the possible genes in these regions are summarized in Table 1.

Figure 2

Overview of genome imbalance. Patient number and chromosome sites are indicated horizontally and vertically, respectively. Chromosome alteration groups: red, gain; green with asterisk, amplification; blue, LOH; yellow, uniparental disomy (UPD); gray, retention; white, not informative. WD, MD and PD indicate well, moderately and poorly differentiated hepatocellular carcinoma (HCC), respectively.

Table 1 Candidate genes in amplified regions

To provide a comparison between our data and previously reported observations of LOH of HCC obtained by comprehensive allelotyping studies, chromosomal arms with LOH in HCC are summarized in Table 2. Most regions with LOH, as determined by GIM, have also been indicated in previous studies using comprehensive allelotyping, with the exception of 1q in which our data indicated chromosomal gains in 25 cases and UPT in two cases, and 9p of which the percentage of LOH was 19.4%. Most regions with chromosomal gains were also consistent with those determined by CGH, as reported previously (data not shown).

Table 2 Chromosomal arms with LOH by GIM and comprehensive allelotyping analysis in HCC

Regions of allelic imbalance in hepatocellular carcinoma

We found UPD and UPT in 13 samples, that is, uniparental remaining alleles duplicated or triplicated where LOH was demonstrated – the regions on 1p, 1q, 2p, 2q, 3p, 6q, 9p, 9q, 10q, 12p and 13q. To confirm UPD in these regions, we performed fluorescent in situ hybridization (FISH) and SNP-based LOH analysis on 13q arm and also showed UPT on 1q arm (Figure 3a and e).

Figure 3

Fluorescence in situ hybridization and loss of heterozygosity (LOH) analysis for validation of allelic imbalance in 1q and 13q arms in patient 19 (ad) and patient 27 (eh). Genome Imbalance Map (GIM) shows LOH in 13q12.11–32.1 and uniparental disomy (UPD) in 13q31.2–34 regions in patient 19 and uniparental trisomy (UPT) in 1q arm in patient 27. (a and e) Genome Imbalance Map of patients 19 and 27, respectively. Black arrow, RP11-40H10 and RP11-14C20 (c); black arrowheads, RP11-118K20 (d); white arrow, RP11-89N3 (g); white arrowhead, RP11-79B11 (h and i). (b and f) Single-nucleotide polymorphism (SNP) analysis using direct sequence at rs1336934 (SNP1) (C/T) in patient 19 and at rs157864 (SNP5) (C/T) in patient 27 (arrowhead) demonstrated LOH on 13q arm and 1q arm in patients 19 and 27, respectively. (c) Fluorescent in situ hybridization (FISH) analysis demonstrated two pairs of distinct signals in liver cancer cells. Orange: RP11-40H10 (13q31.3–32.1); green: RP11-14C20 (13q32.2). (d) One orange signal (13q31.1) and two green signals (13q32.2) were found in liver cancer cells. These data confirmed that patient 19 had UPD in 13q31.2–34 and a breakpoint between 13q31.1 and 13q31.2. In patient 27, GIM showed UPT on 1q arm and fractional copy number on 1p arm. (g) Chromosome 1 centromere (orange) and 1q32.2 (green) in a frozen section of hepatocellular carcinoma (HCC). Three pairs of two distinct signals were observed in liver cancer cells. (h and i) Using FISH analysis, mixed cancer cells with one copy and two copies of 1p arm are shown (h). On the other hand, all the noncancerous liver cells had two signals (i). One signal was observed in 61.2% of all tumor nuclei, whereas the percentage of nuclei with normal copy number was only 16.2% in the control normal liver parenchyma. Orange: 1p31.3.

Single-nucleotide polymorphism analysis at rs1336934 (SNP1) and rs1392000 (SNP2) demonstrated heterozygosity of SNP1 and 2 in the blood sample from patient 19. To further verify LOH in this specimen, we sequenced PCR fragments encompassing SNP1 and 2 and confirmed the LOH at these SNP loci. The LOH on 13q31.2–34 in patient 19 is shown in Figure 3b. On the other hand, FISH analysis using bacterial artificial chromosome (BAC) clones RP11-40H10 and RP11-14C20 verified disomy on 13q32.1-13qter in patient 19 (Figure 3c), indicating that this patient has UPD on 13q32.1-13qter. Similarly, UPD regions on 13q were confirmed in patients 11 and 21 (data not shown). In addition, 13q in patient 19 was composed of a monosomic region (13q14.11–13q31.1) and a UPD region (13q31.2–13q34) consecutively by GIM, which was clearly demonstrated by FISH analysis with BAC clones RP11-118K20 and RP11-14C20 (Figure 3d).

Furthermore, in patients 27 and 29, we observed LOH on 1q arm despite three copies of the remaining allele. Single-nucleotide polymorphism analysis at rs724781 (SNP3), rs1395548 (SNP4) and rs157864 (SNP5), and FISH analysis by RP11-89N3 corroborated the microarray data, that is, co-occurrence of LOH and trisomy on 1q in patient 27 (Figure 3f and g).

Through these results, allelic imbalance regions identified by GIM, UPD or UPT were validated by SNP analysis and FISH analysis.

Fractional copy number is owing to heterogeneity of cancer cells

Genome Imbalance Map analysis showed an intermediate pattern in nine cases, possibly reflecting a mixture of more than two heterozygous populations of liver cancer cells, which suggested chromosomal instability or heterogeneity of the tumor. Among 10 cases in which we examined copy number by FISH at three loci, we observed heterogeneity in two cases at different loci: 1p in one and 1q in the other. Heterogeneity of imbalanced loci is not dependent on chromosomal regions but rather seems to be specific to samples, because heterogeneity was often seen in multiple regions (five of nine cases), and was observed notably in poorly differentiated tumors (four of seven PD cases).

The estimated copy number was approximately 1.5 at 1p arm in patient 27 (Figure 3e). Fluorescent in situ hybridization analysis showed that cancer cells are composed of those with one or two copies at nearly the same ratio, whereas chromosomes of all the corresponding hepatocytes have two copies of 1p arm (Figure 3h and i), indicating that fractional copy number in GIM reflects the heterogeneity of cancer cells.

PLAGL1 gene as a candidate gene responsible for hepatocarcinogenesis

A putative tumor suppressor gene, the PLAGL1 locus on chromosome 6q24–25, was recently identified as an imprinted region (Abdollahi et al., 1997) and allelic loss of the long arm of chromosome 6 has also been reported in many types of human cancer (Taguchi et al., 1993; Thrash-Bingham et al., 1995). We observed LOH in eight of 36 patients, including UPD in two patients, and therefore we analysed the epigenetic status of PLAGL1 and its expression levels.

We examined methylation status of the promoter region of the PLAGL1 gene using methylation-specific PCR (MSP) on 15 specimens, including eight cases with LOH (two of which were UPD) and seven with retention of heterozygosity (ROH), after confirmation of the methylation status of both the background liver and lymphocytes from the patient (Supplementary Figure S1). Seven of the eight LOH (+) tumors were methylated, whereas all of the patients with ROH had both methylated and unmethylated PLAGL1 promoter alleles (Figure 4a). In accordance with methylation status, reduction of PLAGL1 gene expression in HCC was observed by quantitative PCR (qPCR) (Figure 4b). The expression ratio in patient 19, who has allelic loss of 6q arm and unmethylated promoter region in PLAGL1, was similar to those in patients with ROH of 6q. These observations suggested that the imprinted gene, PLAGL1, is often silenced by allelic loss and could be a candidate gene involved in hepatocarcinogenesis.

Figure 4

Epigenetic regulation of 6q24–25. (a) Methylation status of the promoter region of the PLAGL1 gene by methylation-specific PCR. M, methylated; U, unmethylated; T, liver cancer; N, background liver. ΦX174 DNA-HAE III digest was used for a marker. (b) Analysis of PLAGL1 gene expression in liver cancer by quantitative PCR (qPCR). Expression ratio was calculated as fold difference from the background liver.

Gene expression profiles reflect chromosomal alterations

To confirm that alterations in mRNA expression level reflect gain or loss of genomic copy number, we compared the expression intensity with the genome dosage in the same samples (Figure 5).

Figure 5

Comparison of genomic alteration and gene expression status. (a) Total gene dosage and expression analysis across the whole genome of patient 11. Blue dots represent hepatocellular carcinoma (HCC)/liver expression intensity ratio and red continuous lines indicate copy numbers. After normalization as described in Materials and methods, gene expression intensity ratio was computed within a 5 Mb moving average and plotted. In patient 11, gains were observed on 5q, 6p, 7q and 10p, and loss of heterozygosity (LOH) was observed on 2q, 4q, 9p, 10q 15q and 16p. Gene expression levels changed in accordance with genomic alterations. (b) Both the average and standard deviation of expression intensity ratio were calculated for each copy number. Differences at P<0.01 were considered statistically significant.

Based on the expression data of patient 11, the HCC/liver normalized intensity ratio was mapped on the chromosomal regions (Figure 5a). When we compared gene expression levels with gene dosage obtained by GIM analysis in the same sample, increases were observed in levels of expression of genes on 5q, 6p, 7q and 10p, regions in which genomic copy number also showed an increase. On the other hand, expression levels of genes on 2q, 4q, 9p, 10q, 16q and 17p, which were demonstrated to be LOH regions in this sample, were decreased. As shown in Figure 5b, average expression level of each gene increased in accordance with genome dosage (P<0.00001, Student's t-test).

These results demonstrated that gene expression levels tend to change according to the chromosomal copy number and that gene expression profiles reflect genome alteration.


We used GIM analysis with genotyping arrays to generate high-throughput information on not only DNA copy number changes but also allelic imbalance for HCC samples. Our results indicated that GIM can detect the allelic status more accurately than methods used previously (Kusano et al., 1999; Niketeghad et al., 2001). For example, several HCC samples analysed in the present study had UPD or UPT, which may be missed or recognized as gain regions by CGH, as both CGH and array-based CGH can only detect total copy changes (Hashimoto et al., 2004; Katoh et al., 2005; Patil et al., 2005). Uniparental disomy or UPT are exceptional derivations of a pair of offspring chromosomes from one parent only (Engel, 1980) and cause an increased risk of recessive disorders, such as Wiedemann–Beckwith (Henry et al., 1991), Prader–Willi (Nicholls et al., 1989) and Angelman syndromes (Malcolm et al., 1991) owing to reduction to homozygosity (Engel, 1993). Furthermore, UPD regions have been shown to contain genes responsible for carcinogenesis, which have been implicated in Wilms' tumor (Grundy et al., 1994), leukemia (Raghavan et al., 2005) and breast cancer (Murthy et al., 2002), but have never been described in HCC. Our data showed that UPD is frequently observed on chromosome 6q, 10q and 13q, where LOH is observed repeatedly and contains suppressor genes, such as PTEN, DMBT1, BRCA2, RB and DLC2. Therefore, UPD in 6q, 10q and 13q, resulting in duplication of mutated alleles, could be a mechanism for this inactivation.

Among the altered chromosomal regions, 6q was particularly intriguing, because imprinting gene clusters have been reported in this region (Temple et al., 1995). We examined whether simultaneous epigenetic changes, such as methylation, occur in these genes. The PLAGL1 gene, which is located on chromosome 6q24–25, inhibits tumor cell growth through induction of apoptotic cell death and G1 arrest (Varrault et al., 1998). PLAGL1 has been suggested to be a tumor suppressor gene in ovarian cancer (Cvetkovic et al., 2004), and has recently been demonstrated to be regulated in an epigenetic manner (Abdollahi et al., 2003). In the present study, in tumors with LOH or UPD, we found methylation of CpG islands of PLAGL1 of the remaining allele by MSP, and marked reduction of expression intensity in accordance with methylation status using qPCR in cancerous tissues. Thus, if we overlooked UPD in this region and erroneously considered it to be in ROH status, the explanation of epigenetic alteration in this case would be obscure.

Second, GIM analysis provides accurate allelotyping in clinical specimens. Previous allelotyping studies on HCC indicated LOH in 1q, 7p and 8q, in which most samples in the present study showed genomic gain without LOH (Boige et al., 1997; Nagai et al., 1997; Piao et al., 1998; Kondo et al., 2000; Okabe et al., 2000; Laurent-Puig et al., 2001; Jou et al., 2004). This discrepancy may be owing to severe ‘allelic imbalance,’ based on the polyploidy often observed in HCC, despite ROH. As allelotyping studies are often based on PCR assays at microsatellite loci, bias in amplification between alleles and contamination with DNA from normal cells is possible. Therefore, marked differences in signal intensity between two alleles may be explained as LOH, resulting in misunderstanding of allelic imbalance. Genome Imbalance Map showed that seven of 36 cases (19.4%) harbored LOH in 9p, which most researchers have demonstrated as a region of LOH, and therefore the present data were consistent with those of previous studies (Table 2).

Although we identified amplification on 5p15, 6p24, 7q31, 10p11, 11q14, 11q32 and 17q12 in each case, we have not discussed homozygous deletion in this report, because the probe density on the 10K array is not sufficiently dense to allow detection of homozygous deletions, which are usually extremely small. It should be possible to detect more homozygous deletions with a higher density SNP array. Nevertheless, we observed one homozygous deletion at 8p in two cases, which has been validated by PCR with STS markers in this region, and further analysis of these homozygous deletions is now in progress.

Furthermore, GIM often shows fractional allelic copy number, which is caused by mixture of cancer cells with different copy numbers in certain regions. Such patterns involving two levels of heterozygous deletion were previously documented using array-based CGH (Benetkiewicz et al., 2005; Buckley et al., 2005). In this study, we confirmed that fractional allelic copy number is owing to heterogeneity of cancer cells using FISH analysis, and is not owing to contamination by surrounding normal cells, because the GIM algorithm can be applied when the ratio of contamination of the surrounding normal cells is less than 50% (Ishikawa et al., 2005) and the tumor cells occupied more than 90% in the HCC samples used in this study (Supplementary Figure S2). Fluorescent in situ hybridization analysis showed that one allele was lost completely in all the tumor cells, implying that the one or two copies of 1p observed in the tumor cells with two copies were copies of the remaining allele, whereas the cells with one copy did not have such copies, which may duplicate in future resulting in UPD on 1p arm in patient 27.

We and other groups reported previously that genome dosage reflects expression imbalance (Pollack et al., 2002; Midorikawa et al., 2004). Upender et al. (2004) demonstrated that genomic aneuploidy affects gene expression pattern using the chromosome transfer model. To confirm these observations, the effects of genome imbalance on the transcriptome were validated by direct comparison with expression data from the same samples in the present study. Previously, we also demonstrated stepwise chromosomal expression changes in the progression of HCC (Midorikawa et al., 2004), that is, expression gain in 1q and 8p, and loss of 8p, 13q and 17p were observed in well-differentiated liver cancer, whereas expression gains in 12q, 17q and 20q, and reduction in 4q were demonstrated only in advanced HCC. On the other hand, GIM showed frequent allelic gains at 1q, 5q, 6p and 8q, and LOH at 8p and 16q regardless of cancer differentiation grade, and gains at 12q and LOH at 15q were detected specifically in poorly differentiated tumors. In nine positive chromosomal regions identified in the Expression Imbalance Map (EIM), the present data observed by GIM were consistent with our previous observations, that is, poorly differentiated grade tumors contained more chromosomal aberrations in these regions (Supplementary Figure S3). However, gains at 5p, 5q, 6p and 7q, and LOH at 1p, 6q, 16p and 16q, which were detected by GIM, were not identified using EIM, partly because we determined the positive region based on the range of alteration (>3 Mb), which is affected by the density of probes. We used a U95A expression array (Affymetrix, Santa Clara, CA, USA), which includes about only 12 000 probes in EIM analysis in our previous study, and availability of much higher-density arrays will make it possible to identify more altered regions, such as gain at 5p, 5q, 6p and 7q, and LOH at 1p, 6q, 16p and 16q, using EIM.

We applied genotyping array analysis to study genome imbalance in liver cancer using a recently developed method for signal standardization. Comparison of our observations with those of previous studies on chromosomal imbalance by CGH and comprehensive allelotyping localizes the particular chromosomal alterations that conventional methods may overlook or miscomprehend their importance validated by FISH and LOH analysis. This method has the advantage that genome dosage and allelic imbalance, including cancer-specific genomic alterations, can be observed simultaneously, which could lead to the identification of responsible genes with both ease and a high degree of reliability.

Materials and methods

Patients and tissue samples

In all, 36 patients with HCC undergoing hepatectomy in the Hepato-Biliary-Pancreatic Surgery Division, Department of Surgery, Graduate School of Medicine, University of Tokyo were recruited in this study. All subjects gave their informed consent to participation in the study. Among the 36 patients with HCC, 14 were positive for hepatitis B surface antigen (HB), 19 for hepatitis C viral antibody (HC), one for both HB and HC and two for neither HB nor HC. High-density genotyping microarrays (GeneChips) from Affymetrix were used for analysis of primary tumor and peripheral blood samples. Clinical parameters and tumor status based on histological findings of resected specimens are summarized in Supplementary Table S1.

The surgical specimens were immediately cut into small pieces after resection, snap frozen in liquid nitrogen and stored at −80°C.

Genomic DNA extraction and oligonucleotide microarray

Genomic DNA was isolated from tumor tissues or lymphocyte pellets using a QIAamp DNA Mini Kit (Qiagen, Valencia, CA, USA), according to the manufacturer's specifications. Experimental procedures for GeneChip™ were performed according to GeneChip Expression Analysis Technical Manual (Affymetrix, Santa Clara, CA, USA), using a Human Mapping 10K Array XbaI 131 kit (Affymetrix).

Genome Imbalance Map

The GIM algorithm was applied to raw data of HCC and peripheral blood obtained from SNP arrays. Gene locus information was obtained from the websites for Genes On Sequence Map (Homo sapiens build 34). The basic concept of GIM involves normalization of probe-level signals, as described previously (Ishikawa et al., 2005). Briefly, the signal intensity ratio between the raw signal intensity from the cancer and paired normal samples was calculated from the perfect match (PM) probes for each SNP locus by taking the median after omitting the highest and lowest values. Second, we calculated the adjusted ratio, which is the raw ratio divided by the expected ratio. The expected ratio is calculated by adjusting several parameters for each experiment, for example, length of XbaI fragment, percentage of GC of XbaI fragment, local GC content, hybridization free energy of 25-mer probe sequences and genomic mean of signal intensity of PM probes from reference sample.

Fluorescent in situ hybridization and single-nucleotide polymorphism-based loss of heterozygosity analysis

Fluorescent in situ hybridization analysis, and LOH analysis at SNP loci (SNP analysis) were performed to validate the interpretation of the GIM analysis. We obtained nine genomic BACs from a human BAC library (Advanced Geno Techs Co., Tsukuba, Japan), which contained the following regions: RP11-79B11 for 1p31.3 (used for FISH analysis of patient 27), RP11-89N3 for 1q32.2 (patients 27 and 29); RP11-90B18 for 10q23.3 and RP11-166J24 for 10q23.31 (patients 10, 19 and 22); and RP11-93G2 for 13q31.1, RP11-118K20 for 13q31.1, RP11-40H10 for 13q31.3-32.1 and RP11-14C20 for 13q32.2 (patients 11, 19 and 21). The relationships of the BAC clones to the SNP probes are summarized in Table 3 and Supplementary Figure S4. Sections (5 μm thick) were cut from fresh–frozen liver cancer and background liver tissues, which were fixed in 80% ethanol for 24 h, embedded in paraffin and arranged in a tissue microarray, and were used for microwave-assisted FISH analysis (Kitayama et al., 2003). Bacterial artificial chromosomes clones were labeled with Alexa 546 and 488 (Molecular Probes, Eugene, OR, USA) by nick translation. With regard to centromere enumeration probes, Spectrum Green-labeled CEP10 and Spectrum Orange-labeled CEP1 (Vysis Inc., Downers Grove, IL, USA) were used with differently labeled RP11-79B11 and RP11-89N3; RP11-90B18 and RP11-166J24, respectively. Hybridization and evaluation of the results were performed as described previously with minor modifications (Kitayama et al., 2003). DAPI I (4,6-diamidino-2-phenylindol, 1000 ng/ml; Vysis Inc.) was used for nuclear counterstaining. The samples were promptly observed using a fluorescence microscope (BX-51; Olympus, Japan, Tokyo, Japan) equipped with epifluorescence filters and a photometric CCD camera (Sensicam; PCO Company, Kelheim, Germany). The number of FISH signals per cell was counted for a total of more than 100 intact and non-overlapping cell nuclei. Furthermore, to prevent counting of ‘artificial loss’ by missing the part of the nuclei in tangential views, we applied the same method in the background liver and the cutoff value for background was intentionally set at a higher percentage, up to 20% (Sano et al., 2006).

Table 3 Relationship of BAC clones and SNP probes used in FISH analysis

Single-nucleotide polymorphism analysis was applied to assess LOH in UPD regions determined by GIM. A search was performed for SNPs on 1q and 13q arms in the National Center for Biotechnology Information's SNP database. SNPs within 500 bp were amplified by PCR using the following flanking primers and genotyped by direct sequencing:


Forward primer 5′-IndexTermGATTGGTTCCAGGACTAGAG-3′

Reverse primer 5′-IndexTermGAAAGTGCCAAGCTGTACAC-3′


Forward primer 5′-IndexTermTTCACTCACTCTGGGCTATC-3′

Reverse primer 5′-IndexTermTCAACAGCTCCCCTTGATAC-3′


Forward primer 5′-IndexTermCCTACTTTTCCTCCCCTTTG-3′

Reverse primer 5′-IndexTermCCATGCTGGGATCTTGAATG-3′


Forward primer 5′-IndexTermAGCTAAGAGCCAACTCTAGC-3′

Reverse primer 5′-IndexTermCACTCTTTCTCCTCTTGCTC-3′


Forward primer 5′-IndexTermAACGAACCACAGTTCCCAAC-3′

Reverse primer 5′-IndexTermTGGATACTCCTTGGAGCTTC-3′

Disappearance of heterozygosity on informative SNPs in tumors was considered to be LOH.

Bisulfite modification and methylation-specific polymerase chain reaction and quantitative polymerase chain reaction

Aliquots of 2 μg of genomic DNA extracted from cancerous tissues and from the background liver and lymphocytes of the same patients as controls were treated with sodium bisulfite (Sigma, St Louis, MO, USA) and used for MSP, as described previously (Herman et al., 1996). Bisulfite-modified DNA was amplified with PLAGL1 gene-specific primers:





The PCR conditions consisted of denaturation of template at 94°C for 15 min and successive cycles of 94°C for 30 s, 60°C for methylation-specific or 58°C for unmethylated primers for 1 min, and 72°C for 1 min, with a final extension 72°C for 5 min. Reaction products were separated by electrophoresis on 2% agarose gels, stained with ethidium bromide and photographed.

Quantitative polymerase chain reaction for detecting expression of PLAGL1 gene was performed using an iCycler (Bio-Rad, Hercules, CA, USA). The primers and PCR conditions used were as described previously (Cvetkovic et al., 2004). After relative quantification by measuring the ratio between the mean value of the target gene and that of β-actin in each sample, PLAGL1 mRNA expression was compared with the background liver.

RNA extraction and oligonucleotide microarray for gene expression studies

Total RNA was isolated from HCC and the background liver as described previously (Midorikawa et al., 2002). Experimental procedures for CodeLink Human Whole Genome Bioarray were performed according to the manufacturer's specifications (Amersham Biosciences, Piscataway, NJ, USA), using 10 μg of total RNA.

Normalization and filtering of intensity of gene expression

Before further statistical analysis, we normalized and filtered the raw data. A quantile normalization procedure was used for the probe intensity distribution across different chips. The average of expression level intensity, the normalized intensity, was scaled to 1, and each expression data sets where the normalized intensity values were less than 1 were excluded from analysis.

Comparison of gene expression with gene dosage

To parallelize gene dosage and gene expression data, we average both gene copy number by SNP array analysis and expression ratio (HCC/liver) by expression microarray analysis in 100-kb windows. The 100-kb averaged value was smoothed within a 5-Mb moving median. Low probe density regions (less than 1 probe in 5 Mb window) were eliminated from analysis.


  1. Abdollahi A, Godwin AK, Miller PD, Getts LA, Schultz DC, Taguchi T et al. (1997). Cancer Res 57: 2029–2034.

  2. Abdollahi A, Pisarcik D, Roberts D, Weinstein J, Cairns P, Hamilton TC . (2003). J Biol Chem 278: 6041–6049.

  3. Benetkiewicz M, Wang Y, Schaner M, Wang P, Mantripragada KK, Buckley PG et al. (2005). Genes Chromosomes Cancer 42: 228–237.

  4. Bignell GR, Huang J, Greshock J, Watt S, Butler A, West S et al. (2004). Genome Res 14: 287–295.

  5. Boige V, Laurent-Puig P, Fouchet P, Flejou JF, Monges G, Bedossa P et al. (1997). Cancer Res 57: 1986–1990.

  6. Buckley PG, Jarbo C, Menzel U, Mathiesen T, Scott C, Gregory SG et al. (2005). Cancer Res 65: 2653–2661.

  7. Cvetkovic D, Pisarcik D, Lee C, Hamilton TC, Abdollahi A . (2004). Gynecol Oncol 95: 449–455.

  8. Engel E . (1980). Am J Med Genet 6: 137–143.

  9. Engel E . (1993). Am J Med Genet 46: 670–674.

  10. Grundy P, Wilson B, Telzerow P, Zhou W, Paterson MC . (1994). Am J Hum Genet 54: 282–289.

  11. Hashimoto K, Mori N, Tamesa T, Okada T, Kawauchi S, Oga A et al. (2004). Mod Pathol 17: 617–622.

  12. Henry I, Bonaiti-Pellie C, Chehensse V, Beldjord C, Schwartz C, Utermann G et al. (1991). Nature 351: 665–667.

  13. Herman JG, Graff JR, Myohanen S, Nelkin BD, Baylin SB . (1996). Proc Natl Acad Sci USA 93: 9821–9826.

  14. Hoque MO, Lee CC, Cairns P, Schoenberg M, Sidransky D . (2003). Cancer Res 63: 2216–2222.

  15. Ishikawa S, Komura D, Tsuji S, Nishimura K, Yamamoto S, Panda B et al. (2005). Biochem Biophys Res Commun 333: 1309–1314.

  16. Janne PA, Li C, Zhao X, Girard L, Chen TH, Minna J et al. (2004). Oncogene 23: 2716–2726.

  17. Jou YS, Lee CS, Chang YH, Hsiao CF, Chen CF, Chao CC et al. (2004). Cancer Res 64: 3030–3036.

  18. Kallioniemi A, Kallioniemi OP, Sudar D, Rutovitz D, Gray JW, Waldman F et al. (1992). Science 258: 818–821.

  19. Kano M, Nishimura K, Ishikawa S, Tsutsumi S, Hirota K, Hirose M et al. (2003). Physiol Genomics 13: 31–46.

  20. Katoh H, Shibata T, Kokubu A, Ojima H, Loukopoulos P, Kanai Y et al. (2005). J Hepatol 43: 863–874.

  21. Kennedy GC, Matsuzaki H, Dong S, Liu WM, Huang J, Liu G et al. (2003). Nat Biotechnol 21: 1233–1237.

  22. Kitayama Y, Igarashi H, Watanabe F, Maruyama Y, Kanamori M, Sugimura H . (2003). Lab Invest 83: 1311–1320.

  23. Kondo Y, Kanai Y, Sakamoto M, Mizokami M, Ueda R, Hirohashi S . (2000). Hepatology 32: 970–979.

  24. Kusano N, Shiraishi K, Kubo K, Oga A, Okita K, Sasaki K . (1999). Hepatology 29: 1858–1862.

  25. Laurent-Puig P, Legoix P, Bluteau O, Belghiti J, Franco D, Binot F et al. (2001). Gastroenterology 120: 1763–1773.

  26. Lieberfarb ME, Lin M, Lechpammer M, Li C, Tanenbaum DM, Febbo PG et al. (2003). Cancer Res 63: 4781–4785.

  27. Lindblad-Toh K, Tanenbaum DM, Daly MJ, Winchester E, Lui WO, Villapakkam A et al. (2000). Nat Biotechnol 18: 1001–1005.

  28. Lucito R, Healy J, Alexander J, Reiner A, Esposito D, Chi M et al. (2003). Genome Res 13: 2291–2305.

  29. Malcolm S, Clayton-Smith J, Nichols M, Robb S, Webb T, Armour JA et al. (1991). Lancet 337: 694–697.

  30. Mei R, Galipeau PC, Prass C, Berno A, Ghandour G, Patil N et al. (2000). Genome Res 10: 1126–1137.

  31. Midorikawa Y, Tsutsumi S, Nishimura K, Kamimura N, Kano M, Sakamoto H et al. (2004). Cancer Res 64: 7263–7270.

  32. Midorikawa Y, Tsutsumi S, Taniguchi H, Ishii M, Kobune Y, Kodama T et al. (2002). Jpn J Cancer Res 93: 636–643.

  33. Murthy SK, DiFrancesco LM, Ogilvie RT, Demetrick DJ . (2002). Mod Pathol 15: 1241–1250.

  34. Nagai H, Pineau P, Tiollais P, Buendia MA, Dejean A . (1997). Oncogene 14: 2927–2933.

  35. Nannya Y, Sanada M, Nakazaki K, Hosoya N, Wang L, Hangaishi A et al. (2005). Cancer Res 65: 6071–6079.

  36. Nicholls RD, Knoll JH, Butler MG, Karam S, Lalande M . (1989). Nature 342: 281–285.

  37. Niketeghad F, Decker HJ, Caselmann WH, Lund P, Geissler F, Dienes HP et al. (2001). Br J Cancer 85: 697–704.

  38. Okabe H, Ikai I, Matsuo K, Satoh S, Momoi H, Kamikawa T et al. (2000). Hepatology 31: 1073–1079.

  39. Patil MA, Gutgemann I, Zhang J, Ho C, Cheung ST, Ginzinger D et al. (2005). Carcinogenesis 26: 2050–2057.

  40. Piao Z, Park C, Park JH, Kim H . (1998). Int J Cancer 75: 29–33.

  41. Pinkel D, Segraves R, Sudar D, Clark S, Poole I, Kowbel D et al. (1998). Nat Genet 20: 207–211.

  42. Pollack JR, Perou CM, Alizadeh AA, Eisen MB, Pergamenschikov A, Williams CF et al. (1999). Nat Genet 23: 41–46.

  43. Pollack JR, Sorlie T, Perou CM, Rees CA, Jeffrey SS, Lonning PE et al. (2002). Proc Natl Acad Sci USA 99: 12963–12968.

  44. Raghavan M, Lillington DM, Skoulakis S, Debernardi S, Chaplin T, Foot NJ et al. (2005). Cancer Res 65: 375–378.

  45. Sano T, Kityama Y, Igarashi H, Suzuki M, Tanioka F, Chida K et al. (2006). Pathol Int 56: 117–125.

  46. Taguchi T, Jhanwar SC, Siegfried JM, Keller SM, Testa JR . (1993). Cancer Res 53: 4349–4355.

  47. Temple IK, James RS, Crolla JA, Sitch FL, Jacobs P, Howell WM et al. (1995). Nat Genet 9: 110–112.

  48. Thrash-Bingham CA, Salazar H, Freed JJ, Greenberg RE, Tartof KD . (1995). Cancer Res 55: 6189–6195.

  49. Upender MB, Habermann JK, McShane LM, Korn EL, Barrett JC, Difilippantonio MJ et al. (2004). Cancer Res 64: 6941–6949.

  50. Varrault A, Ciani E, Apiou F, Bilanges B, Hoffmann A, Pantaloni C et al. (1998). Proc Natl Acad Sci USA 95: 8835–8840.

  51. Wang DG, Fan JB, Siao CJ, Berno A, Young P, Sapolsky R et al. (1998). Science 280: 1077–1082.

  52. Weissenbach J, Gyapay G, Dib C, Vignal A, Morissette J, Millasseau P et al. (1992). Nature 359: 794–801.

  53. Zhao X, Li C, Paez JG, Chin K, Janne PA, Chen TH et al. (2004). Cancer Res 64: 3060–3071.

Download references


We thank Hiroko Meguro, Kunihiro Nishimura, Akira Watanabe, Kimihiro Yamashita, Megumi Ihara and Kiyoko Nagura for valuable technical assistance, and Panda Binaya for helpful discussion. This work was supported in part by a Grant-in-Aid for Scientific Research (S) 16101006 (HA) and (C) 17591378 (YM), Scientific Research on Priority Areas 17015008 (HA) and 17015017 (HS), and Special Coordination Fund for Science and Technology from The Ministry of Education, Science, Sports and Culture, and CREST, JST (HA), Mitsui Life Social Welfare Foundation (YM), Smoking Research Foundation (HS), the Program of Fundamental Studies in Health Sciences of the National Institute of Biomedical Innovation (NIBIO) and Focus 21 Project of New Energy and Industrial Technology Development Organization (NEDO).

Author information



Corresponding author

Correspondence to H Aburatani.

Additional information

Supplementary Information accompanies the paper on the Oncogene website (

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Midorikawa, Y., Yamamoto, S., Ishikawa, S. et al. Molecular karyotyping of human hepatocellular carcinoma using single-nucleotide polymorphism arrays. Oncogene 25, 5581–5590 (2006).

Download citation


  • liver cancer
  • oligonucleotide array
  • SNP
  • allelic imbalance
  • copy number

Further reading


Quick links