Screening of DNA copy-number aberrations in gastric cancer cell lines by array-based comparative genomic hybridization

Abstract

We performed genome-wide screening for deoxyribonucleic acid copy-number aberrations in 31 gastric cancer (GC) cell lines by using custom-made comparative genomic hybridization (CGH)-array. Copy-number gains were frequently detected at 1q, 3q, 5p, 7p, 7q, 8q, 11q, 17q, 20p, 20q, Xp and Xq, and losses at 3p, 4p, 4q, 8p, 9p, 18p and 18q. With respect to histological subtypes, copy-number gains at 1p, 16p, 20p, 20q and 22q, and losses at 8p, 10p, 10q and 18q were significantly frequent in cell lines derived from tumors of the well-differentiated type, whereas copy-number gains at 1q, 7p, 7q, Xp and Xq were frequent in the undifferentiated type. Homozygous deletions were seen at five loci, whereas high-level amplifications were detected in 15 of the 31 GC cell lines; these had occurred at 24 loci, including the segment containing CDK6 (7q21.2). Amplification of that gene had never been reported in GC before. Immunohistochemical studies showed increased levels of CDK6 protein in 54 of the 292 primary GC samples we examined (18.5%). Cytoplasmic localization of CDK6, as well as CDK6 over-expression, was more frequent in well-differentiated GC than in undifferentiated tumors. Nuclear expression of CDK6 was more frequent in early stage GC than in advanced tumors, suggesting that nuclear localization of CDK6 is likely to be a prognostic factor for GC. Taken together, our data indicate that CDK6 might be involved in the pathogenesis of GC and, more generally, that CGH-arrays have a powerful potential for identifying novel cancer-related genetic changes in a variety of tumors.

Figure 1

Normal‐versus‐normal control hybridizations. (a) Representative genomic profile obtained from one of the five control experiments. Clones are ordered from chromosomes 1–22, X and Y, and within each chromosome on the basis of the UCSC mapping position (http://genome.ucsc.edu/[version April, 2003]). Each dark spot (two spots/each clone) represents test over reference value after normalization and log2 transformation. Thresholds for copy‐number gain and loss were defined at log2ratios of 0.4 and − 0.4, respectively. None of the clones included in the final data set crossed these thresholds for the control experiment, (b) histogram of the ratios obtained for all five control hybridizations. Thresholds for copy‐number gain and loss were set at log2ratios of 0.4 and −0.4, respectively.

Figure 2

Genome‐wide frequencies of copy‐number gains (above 0, green) and losses (below 0, red) in 31 gastric cancer (GC) cell lines. Clones are ordered from chromosomes 1–22, X and Y, and within each chromosome on the basis of the UCSC mapping position (http://genome.ucsc.edu/[version April, 2003]). Green asterisks, clones with at least one high‐level amplification; red asterisks, clones with at least one homozygous deletion.

Figure 3

(a) Genetic changes observed on chromosome 1 of MKN45 cells. Copy‐number gain of MCL1 at 1q21.3 and loss of RNF28 at 1p36.11, neither of which had been detected by conventional comparative genomic hybridization (CGH) (upper left), were clearly revealed by CGH‐array analysis (upper right). A vertical line indicates the position of the centromere. These alterations were confirmed by fluorescence in situ hybridization  (FISH) analysis; one‐copy loss of RNF28 (red, arrow) was detected compared to MUC1 (blue, bottom left), whereas two‐copy gain of MCL1 (red, arrowheads) was detected compared to MUC1 (blue) in interphase (bottom middle) or prophase (bottom right) chromosome slides, (b) genetic changes observed on chromosome 13 of the HSC43 cell line. Homozygous deletion of RB1 (red arrowhead) at 13q14.2 was detected by CGH‐array (middle), although conventional CGH had failed to detect this alteration (left). FISH images specific for RB1 (red) confirmed the homozygous deletion, (c) genetic changes observed on chromosome 7 of the OKAJIMA cell line. CGH‐array analysis identified two independent high‐level amplifications, of CDK6 at 7q21.2 (middle, red arrowhead) and MET at 7q31.2 (green arrowhead), whereas conventional CGH had detected gain of almost the entire long‐arm of chromosome 7 (arrow, left). A vertical line in the middle panel indicates the position of the centromere. FISH analysis demonstrated that CDK6 (red) and MET (green) were independently amplified on different double minute chromosomes (right).

Figure 4

Genome‐wide frequency of copy‐number gains (above 0, green) and losses (below 0, red) in nine well‐ (a), versus 19 undifferentiated (b) types of gastric cancer (GC) cells. Clones are ordered from chromosomes 1–22, X, and Y, and within each chromosome on the basis of the UCSC mapping position (http://genome.ucsc.edu/[version April, 2003]). Green arrowheads or arrows, regions frequently gained; red arrowheads, regions frequently deleted in the well‐differentiated type.

Figure 5

Expression levels of CDK6 messenger ribonucleic acid (mRNA) in gastric cancer (GC) cell lines, compared with (a) copy‐number changes, and (b) histological type. The level of CDK6 mRNA in each sample was normalized on the basis of the respective GAPDH content and recorded as a relative expression level. We compared the expression of CDK6 between cells without copy‐number gain/amplification (n = 20) and those with copy‐number gain/amplification (n = 11), and between cells with well‐differentiated phenotype (n = 9) and those with undifferentiated phenotype (n = 19) by a non‐parametric Mann–Whitney U‐test. CDK6 expression levels in cell lines that had shown copy‐number gain/amplification in comparative genomic hybridization‐array analyses were significantly higher than in cell lines without gains (P = 0.0082). Differences between CDK6 expression levels and histological subtypes were less significant (P = 0.3132).

Figure 6

CDK6 expression in primary gastric tumors. (a–c) Representative staining patterns of CDK6 protein, from experiments using a tissue microarray (TMA) system to examine 292 cases of gastric cancer (GC). (a) 0, (b) cytoplasmic 2+, (c) nuclear 2+×200, (d,e) expression pattern of CDK6 versus overall survival in patients with GC. Nuclear CDK6 expression (d) was associated with better prognosis, although the difference was marginally significant (P = 0.1427), whereas cytoplasmic CDK6 expression (e) showed no correlation with prognosis (P = 0.6964).