Clinical implication of recurrent copy number alterations in hepatocellular carcinoma and putative oncogenes in recurrent gains on 1q

Abstract

To elucidate the pathogenesis of hepatocellular carcinoma (HCC) and develop useful prognosis predictors, it is necessary to identify biologically relevant genomic alterations in HCC. In our study, we defined recurrently altered regions (RARs) common to many cases of HCCs, which may contain tumor-related genes, using whole-genome array-CGH and explored their associations with the clinicopathologic features. Gene set enrichment analysis was performed to investigate functional implication of RARs. On an average, 23.1% of the total probes were altered per case. Mean numbers of altered probes are significantly higher in high-grade, bigger and microvascular invasion (MVI) positive tumors. In total, 32 RARs (14 gains and 18 losses) were defined and 4 most frequent RARs are gains in 1q21.1-q32.1 (64.5%), 1q32.1-q44 (59.2%), 8q11.21-q24.3 (48.7%) and a loss in 17p13.3-p12 (51.3%). Through focusing on RARs, we identified genes and functional pathways likely to be involved in hepatocarcinogenesis. Among genes in the recurrently gained regions on 1q, expression of KIF14 and TPM3 was significantly increased, suggesting their oncogenic potential in HCC. Some RARs showed the significant associations with the clinical features. Especially, the recurrent loss in 9p24.2-p21.1 and gain in 8q11.21-q24.3 are associated with the high tumor grade and MVI, respectively. Functional analysis showed that cytokine receptor binding and defense response to virus pathways are significantly enriched in high grade-related RARs. Taken together, our results and the strategy of analysis will help to elucidate pathogenesis of HCC and to develop biomarkers for predicting behaviors of HCC.

Hepatocellular carcinoma (HCC) is one of the most common human malignancies and responsible for ~5% of all cancer-related deaths in the world.¹ Given that the overall HCC incidence is still rising and prognosis of the disease remains poor, it is important to develop effective diagnostic and therapeutic modalities based on sound biological insights into hepatocarcinogenesis.^2,3

The copy number alterations observed in human solid tumors are known to contribute to the tumorigenesis by affecting the activities of cancer-related genes in the altered chromosomal regions.⁴ Thus, genome-wide mapping of copy number alterations in cancer can facilitate the identification of cancer-related genes, which will improve the understanding of tumorigenesis. Using conventional cytogenetic tools such as comparative genomic hybridization (CGH), copy number gains on 1q, 8q and 20q, along with losses on 1p, 4q, 8p, 13q, 16q and 17p have been previously identified in HCC.^5-7 However, the resolution of conventional cytogenetic analysis is insufficient to precisely identify submicroscopic changes. Recently introduced array-CGH, the combination of conventional CGH and microarray technology, enabled high-resolution screening of genome-wide copy number alterations containing potential cancer-related genes.^8,9 Through array-CGH analysis, novel oncogenes such as JAB1 or differentiation-specific regions have been identified in HCC.^5,10 Also etiology-dependent copy number alterations and genes relevant to hepatocarcinogenesis were suggested in HCC.¹¹ But, it is still difficult to identify the biologically relevant changes and their functional significance in a systematic manner due to the extensive and complex nature of chromosomal alterations.

We hypothesized that recurrent copy number changes common to many HCC cases may contain essential genes for hepatocarcinogenesis. Using this strategy, recurrently altered regions (RARs) were defined in 76 primary HCCs using whole-genome array CGH analysis, and the associations between RARs and clinicopathologic features were examined. Also, we functionally categorized the genes located in the RARs.

Material and methods

Study materials

Frozen tissues (tumor and adjacent normal tissue pairs) were obtained from 76 primary HCC patients (65 males and 11 females) who underwent surgical resection. This study was performed under the approval of the Institutional Review Board of the Catholic University Medical College of Korea. Tumor stage was determined according to the standard tumor-node-metastasis classification of AJCC guidelines (6th edition). Clinicopathologic information about the 76 cases is available in Supporting Table 1. Ten-micrometer-thick frozen sections were prepared, and tumor cell-rich areas (tumor cells comprise more than 80–90% of the selected area) without tumor cell necrosis were microdissected, from which genomic DNA was extracted as described previously.^12,13 Normal human genomic DNA was purchased from Promega (Madison, WI) and used as sex-matched reference DNA for array-CGH.

Array comparative genomic hybridization

A large insert clone array covering the entire human genome at 1 Mb resolution was used for profiling genomic alterations.¹⁴ Array-CGH was performed as described elsewhere using MAUI hybridization station (BioMicro Systems, Salt Lake city, UT).^12,13 Data processing, normalization and realigning of raw array-CGH data were performed using web-based array-CGH analysis software ArrayCyGHt (http://genomics.catholic.ac.kr/arrayCGH/).¹⁵ We used print-tip loess normalization method for analysis. Large insert clones (n = 2,958) and genomic coordinates such as cytogenetic bands or gene positions were mapped according to the UCSC genome browser (http://genome.ucsc.edu/; May 2004 freeze).

Identification of copy number alteration

To set the cutoff value for chromosomal alterations of individual clones based on standard deviation (SD), we performed 6 independent hybridizations using DNA from normal individuals (4 sex-matched and 2 male-to-female), from which control SD values of individual clones were obtained. Chromosomal gain or loss was assigned when the normalized log₂ intensity ratio of each data point exceeded or fell below ±3 SD derived from normal control hybridizations. Regional copy number change was defined as DNA copy number alteration stretching over 2 or more consecutive large insert clones, but not across an entire chromosomal arm. High-level amplification of clones was defined when their intensity ratios were higher than 1.0 in log₂ scale and vice versa for homozygous deletion. RAR was defined as regional copy number changes observed in 15 (20%) or more samples out of 76 HCC samples. The array-CGH data described in our study are available at website (http://www.systemsbiology.co.kr/micro/CGH/hepato.htm).

Real-time quantitative PCR assay

The first-strand cDNA was synthesized from total RNA of 20 HCCs using M-MLV reverse transcriptase (Invitrogen, Carlsbad, CA). cDNA from the nonneoplastic part of liver, which showed no septal fibrosis or liver cirrhosis, was used as normal control. Real-time quantitative PCR was performed using Mx3000P qPCR system and MxPro Version 3.00 software (Stratagene, CA). Twenty microliter of real-time qPCR mixture contained 10 ng of cDNA, 1×SYBR^® Green Tbr polymerase mixture (FINNZYMES, Finland), 1×ROX and 20 pmol of primers. GAPDH was used as an internal control in each procedure. Thermal cycling was done as follows: 10 min at 95°C followed by 40 cycles of 10 sec at 94°C, 30 sec at 53–58°C and 30 sec at 72°C. To verify specific amplification, melting curve analysis was performed (55–95°C, 0.5°C/sec). Relative quantification was performed by the ΔΔCT method.¹⁶ All the experiments were repeated twice and the mean value of intensity ratios with the SD was plotted for each case. Primer sequences for TPM3, RPS27, HAX1, PYGO2, CKS1B, ADAM15, CCT3, PRCC, KIF14, ELF3, TGFB2 and AKT3 are available in the Supporting Table 2.

Western blot analysis for KIF14 and TPM3

Tissue lysates (40 μg per lane for TPM3 and 100 μg per lane for KIF14) were electrophoresed in 10% sodium dodecyl sulfate (SDS)-polyacrylamide gel. Separated proteins were transferred to a PVDF membrane, which was blocked with 5% nonfat dried milk in TBST (20 mM Tris-HCl, 150 mM NaCl and 0.1% Tween 20, pH 7.5), and then incubated overnight with anti-KIF14 (Abcam, Cambridge, UK), anti-TPM3 (Abnova, Taipei, Taiwan) and anti β-actin antibody (Sigma, St. Louis, MO). After washing with TBST, the PVDF membrane was incubated with diluted horseradish peroxidase-conjugated anti-rabbit IgG or anti-mouse IgG for 2 hr at room temperature. The blots were detected using an enhanced chemiluminescence (ECL) system (Amersham-Pharmacia Biotech, Braunschweig, Germany).

Functional enrichment analysis

Functional enrichment analysis based on gene ontology was performed for the RARs that are significantly associated with the clinicopathologic characteristics. In brief, gene sets for enrichment analysis were prepared using 17,661 known genes with genomic coordinates downloaded from the UCSC genome browser (2004, May Freeze). Genes were grouped into specific sets using NetAffx Gene Ontology Mining Tools according to functions annotated in public gene databases such as GO (Gene Ontology), KEGG (Kyoto Encyclopedia of Genes and Genomes) and GenMAPP (Gene Map Annotator and Pathway Profiler).^17-20 A total of 1,632 gene sets were prepared for enrichment analysis. The functional enrichment analysis was performed using GEAR software (http://www.systemsbiology.co.kr/GEAR/).²¹ The significance of enrichment was calculated by hypergeometric distribution and p value less than 0.01 was considered significant.

Statistical analysis

To see the association between chromosomal changes such as RARs and total number of altered clones, and clinicopathologic phenotypes, 8 clinical parameters were treated as categorical variables; age (<50 vs. ≥50 years), sex, grade (Grade 1 and 2 as low vs. 3 and 4 as high), stage (Stage I and II as early vs. Stage III and IV as advanced), tumor size (<3.5 vs. ≥3.5 cm) along with the presence or absence of microvascular invasion (MVI), portal vein invasion and encapsulation. Significance of the different distribution of RARs in each category was tested by two-sided Fisher's exact test. The mean number of altered clones in 8 categories and expression levels of genes in different RAR groups were compared by independent t-test. Stata version 7.0 (Stata Corporation, Texas) was used and p value less than 0.05 was considered significant in all statistical analyses.

Results

Profiles of chromosomal alterations

The overall profile of chromosomal alterations identified in the 76 primary HCCs is illustrated in Figure 1a. Chromosomal alterations are not randomly distributed along the genome but clustered in certain chromosomal segments (Fig. 1b). The mean number of altered clones per case is 677.1 (338.7 clones gained and 338.4 clones lost) out of total 2,958 clones. In other words, 22.9% (11.5% gained and 11.4% lost) of the whole genome is altered per each case on an average. The mean numbers of altered clones are significantly higher in high-grade tumors than low-grade ones (769.9 vs. 594.6 clones; p = 0.002), in microinvasion positive cases than negative ones (764.1 vs. 598.9 clones; p = 0.016) and in bigger tumors than smaller ones (756.1 vs. 602.3 clones; p = 0.026). In the case of entire chromosomal arm changes, 5 chromosomal gains in 1q, 6p, 8q, 20p and 20q and 5 losses in 4q, 8p, 9p, 16q and 17p are repetitively observed in over 30% of the samples. The alteration frequency of chromosomal arms is summarized in Supporting Table 3.

Figure 1.

Whole-genome profiles and frequency plot of chromosomal alterations in hepatocellular carcinoma. (a) The genomic alterations in 76 primary human hepatocellular carcinomas are illustrated in individual lanes. A total of 2,958 large insert clones are mapped according to the UCSC genome browser (May, 2004 Freeze) and ordered by chromosomal position from 1pter to Yqter (X axis). Tumor vs. reference intensity ratios (in log₂ scale) for individual clones are plotted in different color scales reflecting the extent of copy number gains (red) and losses (green) as indicated in the reference color bar. The boundaries of individual chromosome are indicated by long vertical bars and the locations of centromeres by short vertical bars below the plot. (b) The frequency of chromosomal gains (>3 SD) and losses (<3 SD) for each clone is shown for 76 HCC samples.

Recurrently altered regions in HCC

Although entire chromosomal arm changes appeared occasionally, vast majority of copy number alterations in HCCs are localized regional changes. To delineate the frequently observed consensus regions, we defined the RARs, which are regional chromosomal alterations observed in at least 15 cases (20% of HCC cases) among the 76 HCCs. In total, 14 RAR gains (RAR-G) and 18 RAR losses (RAR-L) were detected (Table I). Five most frequent RARs are RAR-G1 (64.5%: 1q21.1-1q32.1), RAR-G2 (59.2%: 1q32.1-1q44), RAR-L17 (51.3%: 17p13.3-17p12), RARG9 (48.7%: 8q11.21-8q24.3) and RAR-L5/L8 (both 44.7%: 4q33-4q35.2/8p23.2-8p12). The average size of RAR-Gs (26.6 Mb; ranged 1–94 Mb) is larger than that of RAR-Ls (16.5 Mb; ranged 0.9–43 Mb). Well-known oncogenes such as MYC, FGF, EGFR and CCND3 along with tumor suppressor genes such as TP53, RB1, CDKN2A and CDKN2B are located in identified RARs. The other genes in the RARs are also thought to be potential cancer-related genes contributing to hepatocarcinogenesis directly or indirectly (Table I).

TABLE I.

GENERAL CHARACTERISTICS OF RECURRENTLY ALTERED REGIONS

RAR	Clone	Chromosome	Map position1	Size (Mb)	Cytoband	Number of cases	Cancer-related genes
G1	RP11-326G21-RP5-936P19	1	142391262-183515789	41	1q21.1-1q32.1	49	PDZK1, MCL1, ARNT, AF1Q, TPM3, ADAR, RPS27, HAX1, PYGO2, CKS1B, ADAM15, MUC1, HDGF, CCT3, PRCC, IFI16, AIM2, USF1, SELP, SELE, LAMC2, TPR, PTGS2,
G2	RP11-572A16-RP11-438H8	1	198883756-244440465	46	1q32.1-1q44	45	KIF14, ELF3, MDM4, ATF3, TGFB2, WNT3A, AKT3
G3	RP11-46C20-CTD-2291F22	5	27487154-35513594	8	5p14.1-5p13.2	19	AMACR
G4	CTB-55A14-CTB-73D21	5	167592449-174643724	7	5q34-5q35.2	16	FGF18
G5	RP1-136B1-RP11-524H19	6	2316508-54856430	52	6p25.2-6p12.1	23	DEK, ID4, E2F3, PRL, MICA, MICB, HMGA1, NOTCH4, MAPK14, PIM1, TFEB, CCND3, VEGF
G6	RP5-1091E12-RP11-339F13	7	54851795-55154846	1	7p11.2	16	EGFR
G7	RP5-1057M1-RP11-212B1	7	79345845-86861115	7	7q21.11	20	HGF, DMTF1, ABCB1
G8	RP5-1059M17-RP11-420H19	7	100783053-124638919	24	7q22.1-7q31.33	22	EPO, EPHB4, PIK3CG, CAV1&2, MET, WNT2
G9	RP11-167E7-RP11-65A5	8	48736257-141551817	94	8q11.21-8q24.3	37	PRKDC, MCM4, SNAI2, LYN, MOS, PLAG1, COPS5, TPD52, E2F5, MMP16, NBS1, EIF3S3, MYC, KCNK9, PTK2, EIF2C2, CCNE2
G10	RP11-472N13-RP11-505N10	10	31849328-33818450	2	10p11.22	15	MAP3K8, NRP1,
G11	RP11-95C14-RP11-265C7	13	91284427-112901415	21	13q31.3-13q34	16	FGF14, ERCC5,
G12	RP11-515O17-RP11-332H18	17	50654233-56847074	6	17q22-17q23.2	16	HLF, MPO, PPM1D, BCAS3, TBX2
G13	CTD-2043I16-CTC-416D1	19	33293135-36979856	3	19q12	19	CCNE1
G14	RP5-852M4-RP4-563E14	20	327036-61041280	61	20p13-20q13.33	26	CDC25B, JAG1, SSTR4, BCL2L1, PLAGL2, DNMT3B, E2F1, MMP24, SRC, TOP1, MYBL2, MMP9, NCOA3, PTPN1, ZNF217, STK6, BMP7
L1	RP1-37J18-RP11-338N10	1	4487199-7719107	3.3	1p36.32-1p36.23	18	CHD5, ICMT, CAMTA1
L2	RP11-285P3-RP4-560M15	1	14486429-15354152	0.9	1p36.21	24	PRDM2, RIZ, CASP9
L3	RP11-8J9-RP11-412F21	1	46823997-69303906	23	1p33-1p31.2	17	RAD54L, FAF1, S(T)IL, CDKN2C, TTC4, JUN, ARHI
L4	RP11-118B23- RP5-1108M17	1	84667459-103923047	19	1p22.3-1p21.1	16	BCL10, CLCA2, LMO4, GTF2B, TGFBR3, GFI1, EVI5
L5	RP11-213L8-RP11-553E4	4	172094998-190118103	18	4q33-4q35.2	34	CASP3, FAT
L6	RP1-273N12-RP11-347H8	6	99385861-102294657	3	6q16.2-6q16.3	17	CCNC, GRIK2
L7	RP1-84N20-RP3-470B24	6	125389372-168197568	43	6q22.31-6q27	16	CRSP3, PLAGL1, SASH1, LATS1, IGF2R, UNC93A, MLLT4
L8	RP11-336N16-RP11-75P13	8	2898583-34455078	31	8p23.2-8p12	34	CSMD1, DEFB1, NAT1, NAT2, PSD3, TNFRSF10A, TNFRSF10B, TNFRSF10C, RHOBTB2
L9	RP11-48M17-RP11-48L13	9	2136329-29639069	27	9p24.2-9p21.1	26	SMARCA2, MTAP, CDKN2B, CDKN2A, RECK, PAX5
L10	RP11-276H19-RP11-65B23	9	86827119-87654534	1	9q21.33	15	GAS1, DAPK1
L11	RP11-92C4-RP11-31J20	9	98644250-104754734	6	9q22.33-9q31.1	16	TGFBR1
L12	RP11-381K7-RP11-338L11	10	112963138-116971219	4	10q25.2-10q25.3	15	CASP7
L13	RP11-153M24-RP11-359P14	13	27414161-59888914	32	13q12.2-13q21.2	19	BRCA2, CCNA1, RB1, RFP2, DLEU1, DLEU2, DDX26
L14	RP11-353N19-RP11-68I8	14	91389738-99247779	8	14q32.12-14q32.2	15	BCL11B
L15	RP11-114I12-RP11-142A12	16	6846342-26727359	20	16p13.2-16p12.1	18	SOCS1, ERCC4
L16	RP11-325K4-RP11-514D23	16	55369591-84922042	29	16q13-16q24.1	28	CDH1, CDH3, BCAR1, WWOX, CDH13, WFDC1
L17	RP11-135N5-RP11-219A15	17	2312021-16718826	14	17p13.3-17p12	39	TP53
L18	RP1-270M7-RP5-1031P17	21	15134621-39788380	14	21q11.2-21q22.2	15	ADAMTS1

¹The mapping position refers to the UCSC genome browser (http://genome.ucsc.edu/; May 2004 freeze).

RARs and clinical characteristics

Eight types of clinical variables (age of onset, sex, grade, stage, size, portal vein invasion, MVI and encapsulation) were analyzed for their associations with the RARs (Table II). RARs and associated characteristics are as follows: RAR-L11 and -L16 with early onset of age, RAR-G5 and -G8 with male sex, RAR-L5, -L9, -L10, -L11 and -L13 with high tumor grade, especially RAR-L10 showing highly significant association, RAR-L4 with advanced tumor stage, RAR-G9, -G12 and -G13 with MVI, RAR-G13 and -L7 with portal vein invasion in negative direction and RAR-G4 with larger tumor size.

TABLE II.

SIGNIFICANT ASSOCIATIONS BETWEEN RARs AND CLINICAL FEATURES

		Early onset (<50)	Late onset (≥50)	Total	p value
RAR-L11	−	23	37	60	0.046
	+	11	5	16
RAR-L16	−	23	37	60	0.046
	+	11	5	16
		Female	Male	Total	p value
RAR-G5	−	11	42	53	0.028
	+	0	23	23
RAR-G8	−	11	43	54	0.028
	+	0	22	22
		Low grade (1, 2)	High grade (3, 4)	Total	p value
RAR-L5	−	30	12	42	0.020
	+	15	19	34
RAR-L9	−	36	14	42	0.003
	+	9	17	26
RAR-L10	−	42	19	51	0.001
	+	3	12	15
RAR-L11	−	40	20	60	0.020
	+	5	11	16
RAR-L13	−	38	19	57	0.031
	+	7	12	19
		Stage (I, II)	Stage (III, IV)	Total	p value
RAR-L4	−	34	26	60	0.047
	+	4	12	16
		Size (<3.5 cm)	Size (≥3.5 cm)	Total	p value
RAR-G4	−	35	25	60	0.024
	+	4	12	16
		MVI (−)	MVI (+)	Total	p value
RAR-G9	−	27	12	39	0.005
	+	13	24	37
RAR-G12	−	36	24	60	0.022
	+	4	12	16
RAR-G13	−	25	32	57	0.009
	+	15	4	19
		PVI (−)	PVI (+)	Total	p value
RAR-G13	−	43	14	57	0.016
	+	19	0	19
RAR-L7	−	46	14	60	0.033
	+	16	0	16

MVI, microvascular invasion; PVI, portal vein invasion.

Functional annotation of the clinically significant RARs

We investigated the functional categories of the genes enriched in the RARs, which showed significant associations with the clinical features. Table III lists the enriched functional pathways in tumor grade-associated RARs. Top 5 pathways have interferon-related gene families in common as member genes, because interferon-loci are included in tumor grade-related RARs. As all of these RARs are copy number losses, it can be assumed that the 5 interferon-related pathways are repressed. Cell cycle regulation and angiogenesis pathways are also found to be significantly associated with tumor grade-related RARs. Functional enrichment analysis results for other clinical feature-related RARs are available in the Supporting Table 4.

TABLE III.

- FUNCTIONAL PATHWAYS ENRICHED IN TUMOR GRADE-ASSOCIATED RARs

Functional annotations	Gene size1	Observed genes2	p-value3	Genes4
Hematopoietin/interferon-class cytokine receptor binding	20	14	5.25E-21	IFNA1, IFNA2, IFNA4, IFNA5, IFNA6, IFNA8, IFNA10, IFNA13, IFNA14, IFNA16, IFNA17, IFNA21, IFNB1, IFNW1
Interferon-alpha/beta receptor binding	9	9	1.18E-16	IFNA1, IFNA2, IFNA4, IFNA10, IFNA13, IFNA16, IFNA17, IFNB1, IFNW1
Response to virus	66	14	5.15E-12	IFN family5
Defense response	126	15	4.52E-09	IFN family, IFNE1
Cytokine activity	177	17	1.15E-08	IFN family, TNFSF11, CER1, IFNE1
Physiological process	18	4	0.0002	INSL4, RLN1, RLN2, INSL6
Regulation of cyclin dependent protein kinase activity	34	4	0.0027	CDKN2A, CDKN2B, CCNA1, RGC32
Condensed chromosome	6	2	0.0042	HMGB1, HMGB2
Angiogenesis	41	4	0.0053	COL15A1, FLT1, VEGFC, HAND2
Pregnancy	47	4	0.0086	FLT1, INSL4, RLN1, RLN2

¹Number of genes in the functionally annotated gene sets.

²Number of observed genes in tumor grade-related RARs.

³Significance level of enrichment was calculated using hypergeometric distribution and p < 0.01 was considered significant.

⁴Gene symbols for the observed genes.

⁵Includes 14 genes in hematopoietin/interferon-class cytokine receptor binding category. It is used to avoid repeating the gene symbols.

High-level copy number changes

In total, 33 amplifications and 10 homozygous deletions (HD) were identified (see Supporting Table 5). Most high-level copy number changes were observed in a single case, but some of them appeared recurrently. For example, the amplification on 8q11.1-8q24.3 containing MYC and EIF3S3 was observed in 10 cases and the amplification on 11q13.2-11q13.3 containing CCND1, FGF4, FGF3 and ORAOV1 in 5 cases. In addition, the amplifications on 1q31.1-1q43, 1q43-1q44, 13q31.1-13q34 and 17q12-17q25.3 were detected in 4 cases. All HDs were observed in a single case except for 9p21.3 containing CDKN2A and CDKN2B tumor suppressor.

Putative hepatocarcinogenesis related oncogenes in RARs on 1q

Many of the high-level copy number changes were found within RARs. Sixteen of 33 amplifications and 8 of 10 HDs overlapped with RAR-Gs and -Ls, respectively (see Supporting Table 5). We assumed high-level alterations overlapping with RARs might be more closely related to carcinogenesis. Thus, we examined the RNA profiles of the putative cancer-related genes as follows in RAR-G1 and -G2, which most extensively overlapped with amplifications; TPM3, RPS27, HAX1, PYGO2, CKS1B, ADAM15, CCT3, PRCC, KIF14, ELF3, TGFB2 and AKT3. Total RNA was available for 20 cases out of 76 HCCs, of which 13 cases are RAR-G1 positive and 12 cases are RAR-G2 positive. RNA expression levels of most candidate genes in the HCCs are higher than normal liver control RNA level (see Supporting Fig. 1). Especially, KIF14 (kinesin family member 14) and TPM3 (tropomyosin 3) are highly expressed in HCC and their RNA levels are significantly correlated with copy number status (Figs. 2a and 2b). The mean RNA expression ratio (tumor/normal) of KIF14 in RAR-G2 positive cases was 16.8, whereas 5.3 in RAR-G2 negative cases (p = 0.004). The mean RNA expression ratio of TPM3 was significantly higher in RAR-G1 positive cases than in RAR-G1 negative group (2.85 vs. 1.83, p = 0.026). CKS1B also showed the higher expression in RAR-G1 positive group than negatives (1.42 vs. 2.21), but the difference is not statistically significant. KIF14 and TPM3 protein expression levels (tumor/normal ratio) were also examined by immunoblotting. Both KIF14 and TPM3 were significantly overexpressed compared with normal liver tissue (Figs. 2c and 2d). The mean protein expression ratio (tumor/normal) was 3.63 for TPM3 (p < 0.001) and 1.99 for KIF14 (p = 0.005). When analysis was limited to HCCs, KIF14 protein expression (2.61 in RAR-G2 positive cases vs. 1.15 in RAR-G2 negative cases, p = 0.004) was significantly correlated with copy number status, whereas the correlation was not significant in TPM3 (Figs. 2a and 2b).

Figure 2.

Expression levels of KIF14 and TPM3 by copy number status. Mean RNA and protein expression levels (tumor/normal ratio) of KIF14 (a and c) and TPM3 (b and d) were calculated by real-time qRT-PCR and Western blotting in the presence of RAR. Human GAPDH gene and β-actin was used as an internal control for qRT-PCR and immunoblotting, respectively. KIF14 RNA level was significantly higher in samples with RAR-G2 (n = 12) than without (n = 8) (*p = 0.004), and KIF14 protein level was also significantly higher in samples with RAR-G2 (n = 8) than without (n = 6) (** p = 0.004). TPM3 RNA level was significantly higher in samples with RAR-G1 (n = 13) than without (n = 7) (*p = 0.026), but protein expression was not significantly different between RAR-G1 positives and negatives (n = 9 and 5, respectively). Tumor 1, 3, 5, 7, 8, 9, 10, 12 and 14: RAR-G1+; Tumor 2, 4, 6, 11 and 13: RAR-G1−. Tumor 1, 3, 7, 8, 9, 11, 12 and 14: RAR-G2+; Tumor 2, 4, 5, 6, 10 and 13: RAR-G2−. X axis represents RAR status and Y axis represents mean expression ratio. Error bars represent mean ± standard error of mean. N, normal.

Discussion

In our study, we analyzed the genome-wide copy number alteration profiles in HCC using BAC arrays with 1Mb resolution. Frequently observed entire arm chromosomal changes (>30% of samples) are largely consistent with previous observations in HCCs of Asians and Caucasians.^5,6,10,11 Some of the discrepancies might be caused by using the populations or criteria different from our study. The relatively frequent gains of 20p and 20q might have been observed due to the increased resolution of array platform we used.¹⁰

We delineated the RARs on the assumption that alterations shared by significant number of cases may contain genes essential to hepatocarcinogenesis. Among the 32 RARs identified in our study, 18 RARs (RAR-G1,-2,-5,-6,-9,-11,-12 and -14 along with RAR-L1,-2,-5,-7,-8,-9,-11,-13,-17 and -18) including the 5 most frequent RARs are consistent with those previously reported in HCC.^22-24 The other 14 RARs have not been identified in other studies. But, some of them such as RAR-G7 (7q21.11), RAR-G8 (7q22.1–7q31.33) and RAR-L16 (16q13–16q24.1) appeared in more than 20 cases in our study, which suggests that they are unlikely to be false positives. These potentially novel RARs may give a clue to understand hepatocarcinogenesis. For example, copy number gain of 7q21.11(RAR-G7), frequently observed in our study, could explain overexpression of HGF (hepatocyte growth factor) in HCCs. Higher expression of HGF, which is known to promote angiogenesis, has been frequently reported in HCC²⁵ and highlighted as a potential biomarker for the progression of HCC.^26,27 Another example is EPO located in 7q 22.1 (RAR-G8). EPO is also known to be involved in angiogenesis in HCC,²⁸ but copy number gain of EPO gene has been rarely reported in HCC.

In high-level copy number changes identified in our study, many well-known cancer related genes such as MYC, MDM4, EIF3S3, CCND1, FGF4, FGF3, CDKN2A and CDKN2B were found^11,29,30 along with other genes unexplored in HCCs. Twenty-four regions of amplifications or HDs were found to overlap with RARs across diverse chromosomes. It can be assumed that the genes located in this overlapped region may contribute to hepatocarcinogenesis more directly. Because this overlapping phenomenon was most apparent on chromosome 1q, we examined the RNA profile of the putative cancer-related genes on the 8 overlapped regions on 1q. Of the 12 candidate genes listed in the result, RNA levels of KIF14 and TPM3 were elevated in all tested cases except for 1 and showed the significant correlations with the copy number status. Similarly, both KIF14 and TPM3 proteins were significantly overexpressed in HCCs compared with normal liver tissue. Increased expression of TPM3 protein was observed in all the HCCs tested regardless of copy number status, which suggests the possibility of TPM3 as a HCC-biomarker. KIF14 expression was elevated in DNA dosage-dependent manner in both RNA and protein levels. Those results suggest that copy number gains on 1q (RAR-G1 and 2) induce overexpression of TPM3 and KIF14, subsequently influencing hepatocarcinogenesis. Although copy number gain on 1q31-q32 where KIF14 is located has been reported in diverse cancers,³¹ this has not been reported in HCC yet.

KIF14 is essential for cytokinesis and chromosome segregation.³² Its upregulation was previously observed in malignancies and suggested as a predictor for poor prognosis.³³ As TPM3 plays similar roles in cytokinesis,³⁴ it can be assumed that both molecules may be involved in carcinogenesis including the development of HCC. In our unpublished data, siRNA knockdown of KIF14 and TPM3 repressed the growth of a HCC cell line. However, in our study, we could not observe any significant association between their expression levels and invasiveness (MVI and PVI), tumor grade, size and survival (data not shown). It might be due to limited number of samples (20 for qRT-PCR and 14 for immunoblotting). Further large-scale studies will be required to validate this possibility in HCCs.

TPM3, located in 1q21.3, is known to be involved in hematopoietic tumorigenesis by forming TPM3-ALK fusion through (1;2) translocation.³⁵ TPM3-ALK fusion gene is also known to be involved in transformation, proliferation, invasion and metastasis.³⁶ When we examined the existence of TPM3-ALK fusion by real-time qRT-PCR for the HCCs showing TPM3 overexpression as previously described,³⁵ none of them showed fusion-specific signal (data not shown). But, the absence of the fusion protein in our samples cannot exclude the potential role of TPM3 in hepatocarcinogenesis. To our knowledge, this is the first report to suggest oncogenic potential of copy number gains related to overexpression of KIF14 and TPM3 in hepatocarcinogenesis.

Some RARs are significantly associated with clinical features. Especially, RAR-L9 and -L10 (9p21) and RAR-G9 (8q24) showed the most significant associations with the higher tumor grade and MVI, respectively. This result is consistent with previous reports showing that copy number losses on 9p21 and 8q24 amplification have been frequently associated with poor prognosis of HCC.^37,38 Our result indicates that these copy number alterations can affect prognosis of HCC.

To see functional implication of copy number alterations, we applied functional enrichment analysis on RARs associated with clinical features. Gene sets significantly enrichment in with phenotype-correlated RARs might represent functionally related genes, which are likely to work collaboratively in hepatocarcinogenesis. For example, cytokine receptor binding and defense response to virus pathways are enriched in recurrent copy number deletion regions (RAR-L5,-9,-10,-11 and -13), which are associated with the high tumor grade. This is comparable with previous studies reporting that the progression of liver disease and hepatocarcinogenesis is not only due to direct cytopathic effects of viruses but also due to repressed host immune response to viral proteins.^39,40 Given that more than 90% of HCC patients examined in our study are HBV or HCV positive, it seems reasonable to observe the significant correlation between the recurrent deletion of defense response to virus pathway and the high tumor grade. Also, cell cycle regulation and angiogenesis pathways were found to be enriched in the RAR-Ls associated with high grade. Disruption of cell cycle regulation in aggressive HCC has been consistently reported in HCC.⁴¹

Our study has some limitations. The resolution of arrays used is not high enough to detect microcopy number changes and to determine exact boundaries for altered regions. Further studies using higher resolution arrays and target region-focusing subarrays will be required to discover novel microalterations. Because of limited availability of cases for RNA and protein assays, potential roles of the RAR-associated genes in hepatocarcinogenesis could not be validated reliably. Further large scale study will be necessary to explore these genes. Lastly, because most of the study subjects are HBV or HCV positive, it is difficult to generalize our results to nonviral HCCs.

In conclusion, we successfully identified copy number changes including novel ones in 76 primary HCCs using whole-genome array-CGH and proposed KIF14 and TPM3 as potential oncogenes in HCC. We also demonstrated the relationship between RARs and tumor characteristics. Through functional enrichment analysis, we suggested several functional pathways likely to be involved in hepatocarcinogenesis. Taken together, our results and the strategy of analysis will help to elucidate pathogenesis of HCC and to develop biomarkers for predicting the behaviors of HCC.