Purpose: Although many genomic alterations have been observed in lung cancer, their clinicopathologic significance has not been thoroughly investigated. This study screened the genomic aberrations across the whole genome of non–small cell lung cancer cells with high-resolution and investigated their clinicopathologic implications.

Experimental Design: One-megabase resolution array comparative genomic hybridization was applied to 29 squamous cell carcinomas and 21 adenocarcinomas of the lung. Tumor and normal tissues were microdissected and the extracted DNA was used directly for hybridization without genomic amplification. The recurrent genomic alterations were analyzed for their association with the clinicopathologic features of lung cancer.

Results: Overall, 36 amplicons, 3 homozygous deletions, and 17 minimally altered regions common to many lung cancers were identified. Among them, genomic changes on 13q21, 1p32, Xq, and Yp were found to be significantly associated with clinical features such as age, stage, and disease recurrence. Kaplan-Meier survival analysis revealed that genomic changes on 10p, 16q, 9p, 13q, 6p21, and 19q13 were associated with poor survival. Multivariate analysis showed that alterations on 6p21, 7p, 9q, and 9p remained as independent predictors of poor outcome. In addition, significant correlations were observed for three pairs of minimally altered regions (19q13 and 6p21, 19p13 and 19q13, and 8p12 and 8q11), which indicated their possible collaborative roles.

Conclusions: These results show that our approach is robust for high-resolution mapping of genomic alterations. The novel genomic alterations identified in this study, along with their clinicopathologic implications, would be useful to elucidate the molecular mechanisms of lung cancer and to identify reliable biomarkers for clinical application.

Lung cancer is the most common incident form of malignancy and is also the leading cause of cancer death worldwide (1, 2). A primary lung cancer is classified into four major histologic subtypes; squamous cell carcinomas, adenocarcinomas, large cell and small cell lung cancers. The former three classes, which are grouped as non–small cell lung cancers (NSCLC), make up almost 80% of all total lung cancer cases. Among the NSCLC, squamous cell carcinomas and adenocarcinomas are the two major subtypes. Histologically different subtypes have different clinical courses, and might require individual therapeutic approaches.

Some genomic aberrations in tumors have been suggested to be prognostic markers or can be used to identify the target genes for treatment or prevention (3, 4). Likewise, in other solid tumors, chromosomal aberrations are thought to be critical molecular events in the pathogenesis of lung cancer (5, 6). However, clinically applicable screening tools or prognostic markers are still underdeveloped. Because the lack of efficient screening methods and therapy accounts for the poor outcome of lung cancer, genome-wide assessment of aberrations could help in developing more accurate diagnostic and therapeutic strategies.

For this reason, previous cytogenetic studies using conventional comparative genomic hybridization (CGH) or fluorescence in situ hybridization have focused on identifying the chromosomal aberrations associated with NSCLC. Recurrent genomic alterations have been observed in NSCLC, including the gains of partial or whole chromosomal arms on 1q, 3q, 5p, and 8q along with the losses on 3p, 6q, 8p, 9p, 13q, and 17p (711). However, the ∼10 Mb resolution of conventional CGH is insufficient for the precise identification of submicroscopic changes (12). As accumulating evidence suggests that changes in the genomic dosage contribute to tumorigenesis by altering the expression levels of the cancer-related genes (13, 14), more detailed analyses with sufficient resolution are required.

For enhancing the resolution, array CGH using mapped bacterial or P1 artificial chromosomes (BAC/PAC) rather than metaphase chromosomes, has been recently developed (1517). This technique provides a high resolution that is directly related to the genomic density and insert size of the arrayed clones. Array CGH has emerged as a useful tool for detecting and mapping the genomic aberrations, which may contain putative oncogenes or tumor suppressor genes and for performing a molecular classification of tumors (18).

To see genomic alterations and their clinicopathologic implications in NSCLC, we applied genome-wide array CGH to the genomic DNA extracted from the microdissected tissues of 29 squamous cell carcinoma and 21 adenocarcinoma cases, on which the association study was done. Using this strategy, the genomic copy number changes specific to NSCLC including novel minimally altered regions (MAR) were identified. Those genomic alterations are likely to be related to tumorigenesis or the clinical outcomes of lung cancer.

Study materials. Frozen tissues were obtained from 50 NSCLC patients, who underwent surgical resection at Dankook University Hospital, Cheonan, Korea. Tissue collection and the full procedure of genetic analyses were done under the approval of Institutional Review Board of Kangnam St. Mary's Hospital, The Catholic University of Korea. The 50 NSCLC cases were histologically classified into squamous cell carcinomas (29 cases) and adenocarcinomas (21 cases). Tumor staging was done according to the standard tumor-node-metastasis classification in the American Joint Committee on Cancer guidelines. Of 50 patients whose mean age was 60 years, 88% (44 cases) were male. Other clinical information on the 50 patients is also available in Supplementary Table S1.

Tissue preparation. After surgical resection, tumor and adjacent normal tissues from the same patient were collected separately and snap-frozen in a deep freezer. Frozen sections were prepared of 10 μm thickness on a gelatin-coated slide using 2800 Frigocut (Reighert-Jung, Germany). After H&E staining, tumor cell–rich area (>60% of tumor cells) and histologically normal cell area were selected under the microscope and dissected manually. Microdissected tissues were transferred into the cell lysis buffer (1% proteinase-K in TE buffer) and DNA was extracted. DNA from normal tissue was used as reference DNA for array CGH. Extracted DNA was purified using a DNA purification Kit (Solgent, Daejeon, Korea) and used for dye labeling reactions.

Array comparative genomic hybridization and image analysis. We used human large insert clone arrays with 1 Mb resolution across the whole genome printed by the Sanger Institute Microarray Facility (19). DNA labeling, prehybridization, hybridization, and posthybridization processes were done as described previously (19, 20). Arrays were scanned using GenePix 4100A scanner (Axon Instruments, Union City, CA) and the image was processed using GenePix Pro 6.0.

Data processing, normalization, and mapping of BAC clones. Normalization and re-aligning raw array CGH data were done using the web-based array CGH analysis interface, ArrayCyGHt (http://genomics.catholic.ac.kr/arrayCGH/; ref. 21). Mapping of large insert clones was done according to the genomic location in the UCSC genome browser (May 2004 freeze). In total, 2,987 successfully mapped BAC clones out of initial 3,014 clones were processed subsequently. All the genomic coordinates such as cytogenetic bands or gene positions described in this study are based on the same version of the human genome available on the UCSC genome browser.

Data analysis for chromosomal alterations. To set the cutoff value for chromosomal alterations of individual large insert clones, we did a series of four independent normal hybridizations (three sex-matched and one male versus female hybridizations) as a control. The average SD value of the control batch was 0.081. Adopting the criteria of a previous study (22), the cutoff value for the copy number aberrations was set to be ±0.2 in log2 ratio in this study, >2-fold of control SD. The entire chromosome arm gain or loss was determined as previously described (23). Regional copy number change was defined as DNA copy number alteration limited to part of a chromosome. High-level amplification of clones was defined when their intensity ratios were >1.0 in log2 scale, and vice versa for homozygous deletion. The boundary of copy number change was assigned to be halfway between the two neighboring clones.

Definition of minimally altered regions. To define MARs of chromosomal gain or loss, we used CGH-Miner (http://www-stat.stanford.edu/∼wp57/CGH-Miner/) to smooth the raw intensity ratio and to identify the breakpoints of chromosomal alterations (24). A series of four normal hybridizations were combined as a control and the analysis was done with recommended program variables. The significant gains or losses reported by the program were directly used for subsequent aligning procedures. Minimal regions of chromosomal gains and losses were determined by altered segments recurring for at least seven samples.

Statistical analysis. The significance of the differences in chromosomal arm changes between squamous cell carcinomas and adenocarcinomas was tested by two-sided Fisher's exact test. The correlations between recurrent genetic changes on minimally altered regions were assessed using univariate pairwise Pearson's correlation. For multiple comparisons, the step-down Sidak method was used to adjust the overall level of significance. In this case, the pairs of genetic changes on the same chromosomal arm were excluded for the concordance analysis. The correlations between genetic alterations and clinical variables were analyzed by two-sided Fisher's exact test. All the MARs as well as chromosomal arm changes were included in the analysis. For comparison, four kinds of clinical variables were treated as categorical variables such as age (<60 versus ≥60 years), stage (stages I and II as early versus stages III and IV as advanced), lymph node status (negative versus positive), and the disease recurrence (presence versus absence). Kaplan-Meier method was used for survival analysis and the difference between survival curves was compared using the log-rank test in univariate model. To identify independent prognostic factors after adjusting clinical variables such as age, sex, stage, treatment, metastasis, and recurrence, Cox regression was done. In all statistical analyses, P < 0.05 was considered significant.

Comprehensive profiling of genomic alterations in non–small cell lung cancers. The overall genomic alterations observed in the 50 NSCLC cases (29 squamous cell carcinomas and 21 adenocarcinomas) are illustrated in Fig. 1A. The frequency plots of the chromosomal changes in the 50 NSCLC cases show that they are not randomly distributed but clustered in several hot regions across all the chromosomes (Fig. 1B). Eight chromosomal arms were frequently gained: 19q (40%, 20 of 50 cases), 20q (26%), 22q (24%), 3q (22%), 19p (22%), 1q (20%), 5p (20%), and 17q (20%). Also, six chromosomal arms were frequently lost: Yp (52%), Yq (46%), 9p (42%), 3p (26%), 17p (24%), and 4q (20%). Six chromosomal changes differentially distributed between squamous cell carcinomas and adenocarcinomas. Gains of 3q and 12p as well as losses of 3p, Yp, and Yq were found to be specific to squamous cell carcinomas, whereas a gain of 6p was found to be specific to adenocarcinomas (see Supplementary Table S2). The array CGH signal intensity ratio (in log2 scale) data of the 50 NSCLC can be downloaded from our web site (http://lib.cuk.ac.kr/micro/CGH/lung.htm).

Fig. 1.

Genome-wide copy number alterations in 50 cases of NSCLC. A, genomic profiles of 29 squamous cell carcinomas (top) and 21 adenocarcinomas (bottom). Fifty NSCLC cases are represented in individual lanes with corresponding sample numbers in two subtypes. Intensity ratios are schematically plotted in different color scales reflecting the extent of genomic gains (red) and losses (green) as indicated in the reference color bar. A total of 2,987 BAC clones were ordered (x-axis) according to the map positions and the chromosomal order from 1pter to Yqter. B, the genome-wide frequencies of all significant gains (>0.2 of intensity ratio, top plot) and losses (<−0.2 of intensity ratio, bottom plot) for each clone are shown for 29 cases of squamous cell carcinomas (black, above the x-axis) and 21 cases of adenocarcinomas (gray, below the x-axis), respectively. The boundaries of individual chromosome and the location of centromere are indicated by vertical bars and dotted lines below the plots, respectively.

Fig. 1.

Genome-wide copy number alterations in 50 cases of NSCLC. A, genomic profiles of 29 squamous cell carcinomas (top) and 21 adenocarcinomas (bottom). Fifty NSCLC cases are represented in individual lanes with corresponding sample numbers in two subtypes. Intensity ratios are schematically plotted in different color scales reflecting the extent of genomic gains (red) and losses (green) as indicated in the reference color bar. A total of 2,987 BAC clones were ordered (x-axis) according to the map positions and the chromosomal order from 1pter to Yqter. B, the genome-wide frequencies of all significant gains (>0.2 of intensity ratio, top plot) and losses (<−0.2 of intensity ratio, bottom plot) for each clone are shown for 29 cases of squamous cell carcinomas (black, above the x-axis) and 21 cases of adenocarcinomas (gray, below the x-axis), respectively. The boundaries of individual chromosome and the location of centromere are indicated by vertical bars and dotted lines below the plots, respectively.

Close modal

High-level amplification and homozygous deletion. In total, 98 large insert clones showed high-level amplifications at least in one case and they clustered in 36 different genomic segments. The genomic size of the amplicons ranged from 0.31 to 14.78 Mb. All the identified amplicons along with the putative cancer-related genes located in these amplicons are summarized in Table 1. Figure 2A shows an example of a high copy number gain observed recurrently around 3q26-q28. The first (3q26), which is as large as 14.78 Mb, harbors several putative oncogenes such as EVI1, SKIL, ECT2, and PIK3CA. The other (3q28), which is 10.31 Mb in size, contains the putative oncogenes, BCL6 and HES. In all 50 NSCLC cases, only three homozygous deletions were identified (Table 1). Among them, a homozygous deletion of RP11-765C10 (10q23.31) harbors the tumor suppressor gene PTEN (Fig. 2B).

Table 1.

Genomic segments representing high copy number changes in NSCLC

ChangeCloneCytobandMap position (Mb)Size (Mb)Observed cases*Putative cancer-related genes
Amplification RP11-45I3 1p36.13 15.94-16.84 0.9 SqCs4  
 RP11-184I16 1p34.1 43.42-44.18 0.76 SqCs15 PTPRF 
 RP5-881A21 1p12 118.51-118.96 0.45 SqCs17  
 RP4-790G17/RP11-172I6 1q21.2-q22 145.67-152.49 6.81 AdCs1, SqCs23 AF1Q, TPM3, CTSS 
 RP11-440P5/RP11-568N6 2p16.1-p14 59.90-63.96 4.06 AdCs16, SqCs9, SqCs14 REL 
 RP11-251C9 3q25.1 151.88-152.64 0.75 SqCs22  
 RP11-264D7/RP11-416O18 3q26.1-q26.33 168.03-182.82 14.78 SqCs2, SqCs8, SqCs9, SqCs12, SqCs15, SqCs16, SqCs20, SqCs22 EVI1, SKIL, ECT2, PIK3CA 
 RP11-110C15/RP11-506F8 3q27.2-3q29 185.80-196.12 10.31 SqCs2, SqCs7, SqCs9, SqCs15, SqCs23, SqCs24 BCL6, HES 
 CTD-2324F15 5p15.32 6.15-6.47 0.31 AdCs21  
 RP11-360O19 6p24.3 10.16-11.01 0.85 AdCs1  
 RP11-472M19 6p12.1 55.80-57.09 1.29 SqCs1  
 RP11-449P15/RP4-810E6 7p22.3-p22.1 0.69-5.85 5.16 AdCs1, AdCs13 NUDT1 
 RP11-449G3/RP11-339F13 7p11.2 53.47-55.02 1.55 SqCs11  
 RP5-1091E12/RP4-725G10 7p11.2 54.72-55.54 0.82 SqCs5 EGFR 
 RP5-905H7/RP11-340I6 7q11.21-q11.21 62.13-62.49 0.36 AdCs1  
 RP11-107L23 7q11.23 73.42-75.47 2.05 AdCs1  
 RP11-17I10 7q22.3 105.57-106.49 0.91 SqCs25 PIK3CG 
 RP11-115G12 8q12.3 65.01-66.36 1.35 SqCs29  
 RP11-399H11/RP11-83N9 9q34.3 134.47-135.81 1.33 AdCs13  
 RP11-554A11 11q13.3 68.38-68.93 0.55 SqCs9  
 RP11-21D20 11q13.4 69.78-70.34 0.56 SqCs25  
 RP11-45C5/RP11-21G19 11q22.1-q22.2 99.95-100.77 0.82 AdCs21  
 CTD-3245B9 11q23.3 117.67-118.56 0.88 AdCs21 MLL, DDX6 
 RP3-432E18/RP11-89H19 12q13.11 46.13-46.52 0.39 SqCs10  
 RP11-490O6/CTD-2504F3 16p13.13-p13.11 11.11-15.77 4.66 AdCs1  
 RP11-105C19/CTD-2515A14 16p12.1 22.31-24.18 1.87 AdCs1  
 RP5-906A24/RP11-94L15 17q12 33.91-35.02 1.1 AdCs1 MLLT6 
 RP11-769O8/RP11-291G24 18p11.32 0.52-1.33 0.8 SqCs10 YES1, TYMS 
 CTD-2547N9/CTC-444D3 19p13.2 8.06-8.78 0.71 AdCs13  
 CTC-260F20 19p13.11 18.59-20.01 1.41 SqCs1 JUND 
 CTD-2527I21/CTC-246B18 19q13.11-q13.2 39.22-44.14 4.92 SqCs9 HKR, SPINT2 
 RP11-158G19/CTD-2337J16 19q13.42 58.68-59.30 0.62 AdCs21  
 RP4-742J24/RP11-104O6 20p12.2-p12.1 11.17-12.28 1.11 SqCs25  
 RP3-324O17/RP5-857M17 20q11.21 28.92-29.65 0.73 SqCs1  
 CTA-433F6/RP11-50L23 22q11.21-q11.22 16.84-19.20 2.36 AdCs21  
 RP5-925J7/CTA-722E9 22q13.32-q13.33 47.47-47.94 0.46 SqCs1, SqCs3  
Homozygous deletion RP11-765C10 10q23.31 89.79-90.20 0.40 SqCs2 PTEN 
 RP11-122K13 10q26.3 134.36-135.11 0.75 SqCs25, SqCs26  
 CTD2547N9/CTD444D3 19p13.2 8.06-9.44 1.38 AdCs11  
ChangeCloneCytobandMap position (Mb)Size (Mb)Observed cases*Putative cancer-related genes
Amplification RP11-45I3 1p36.13 15.94-16.84 0.9 SqCs4  
 RP11-184I16 1p34.1 43.42-44.18 0.76 SqCs15 PTPRF 
 RP5-881A21 1p12 118.51-118.96 0.45 SqCs17  
 RP4-790G17/RP11-172I6 1q21.2-q22 145.67-152.49 6.81 AdCs1, SqCs23 AF1Q, TPM3, CTSS 
 RP11-440P5/RP11-568N6 2p16.1-p14 59.90-63.96 4.06 AdCs16, SqCs9, SqCs14 REL 
 RP11-251C9 3q25.1 151.88-152.64 0.75 SqCs22  
 RP11-264D7/RP11-416O18 3q26.1-q26.33 168.03-182.82 14.78 SqCs2, SqCs8, SqCs9, SqCs12, SqCs15, SqCs16, SqCs20, SqCs22 EVI1, SKIL, ECT2, PIK3CA 
 RP11-110C15/RP11-506F8 3q27.2-3q29 185.80-196.12 10.31 SqCs2, SqCs7, SqCs9, SqCs15, SqCs23, SqCs24 BCL6, HES 
 CTD-2324F15 5p15.32 6.15-6.47 0.31 AdCs21  
 RP11-360O19 6p24.3 10.16-11.01 0.85 AdCs1  
 RP11-472M19 6p12.1 55.80-57.09 1.29 SqCs1  
 RP11-449P15/RP4-810E6 7p22.3-p22.1 0.69-5.85 5.16 AdCs1, AdCs13 NUDT1 
 RP11-449G3/RP11-339F13 7p11.2 53.47-55.02 1.55 SqCs11  
 RP5-1091E12/RP4-725G10 7p11.2 54.72-55.54 0.82 SqCs5 EGFR 
 RP5-905H7/RP11-340I6 7q11.21-q11.21 62.13-62.49 0.36 AdCs1  
 RP11-107L23 7q11.23 73.42-75.47 2.05 AdCs1  
 RP11-17I10 7q22.3 105.57-106.49 0.91 SqCs25 PIK3CG 
 RP11-115G12 8q12.3 65.01-66.36 1.35 SqCs29  
 RP11-399H11/RP11-83N9 9q34.3 134.47-135.81 1.33 AdCs13  
 RP11-554A11 11q13.3 68.38-68.93 0.55 SqCs9  
 RP11-21D20 11q13.4 69.78-70.34 0.56 SqCs25  
 RP11-45C5/RP11-21G19 11q22.1-q22.2 99.95-100.77 0.82 AdCs21  
 CTD-3245B9 11q23.3 117.67-118.56 0.88 AdCs21 MLL, DDX6 
 RP3-432E18/RP11-89H19 12q13.11 46.13-46.52 0.39 SqCs10  
 RP11-490O6/CTD-2504F3 16p13.13-p13.11 11.11-15.77 4.66 AdCs1  
 RP11-105C19/CTD-2515A14 16p12.1 22.31-24.18 1.87 AdCs1  
 RP5-906A24/RP11-94L15 17q12 33.91-35.02 1.1 AdCs1 MLLT6 
 RP11-769O8/RP11-291G24 18p11.32 0.52-1.33 0.8 SqCs10 YES1, TYMS 
 CTD-2547N9/CTC-444D3 19p13.2 8.06-8.78 0.71 AdCs13  
 CTC-260F20 19p13.11 18.59-20.01 1.41 SqCs1 JUND 
 CTD-2527I21/CTC-246B18 19q13.11-q13.2 39.22-44.14 4.92 SqCs9 HKR, SPINT2 
 RP11-158G19/CTD-2337J16 19q13.42 58.68-59.30 0.62 AdCs21  
 RP4-742J24/RP11-104O6 20p12.2-p12.1 11.17-12.28 1.11 SqCs25  
 RP3-324O17/RP5-857M17 20q11.21 28.92-29.65 0.73 SqCs1  
 CTA-433F6/RP11-50L23 22q11.21-q11.22 16.84-19.20 2.36 AdCs21  
 RP5-925J7/CTA-722E9 22q13.32-q13.33 47.47-47.94 0.46 SqCs1, SqCs3  
Homozygous deletion RP11-765C10 10q23.31 89.79-90.20 0.40 SqCs2 PTEN 
 RP11-122K13 10q26.3 134.36-135.11 0.75 SqCs25, SqCs26  
 CTD2547N9/CTD444D3 19p13.2 8.06-9.44 1.38 AdCs11  

NOTE: The boundary of each high copy number of change is defined by the corresponding insert clone. Cytogenetic band and map position of clones are based on the public genome database (UCSC genome, May 2004 freeze).

*

In case of more than two observed cases, the boundary of high copy number change was defined as the most extended set of clones, so they were not necessarily overlapping.

Fig. 2.

Individual profiles of high copy number changes. A, high-level amplifications on 3q21-q29 for SqCs9 and SqCs22. B, a homozygous deletion on 10q23.31 for SqCs2. In the intensity ratio profiles, the x-axis represents the map position of corresponding clone according to the UCSC human genome (May 2004 freeze) and the intensity ratios were assigned to the y-axis. The schematic presentation of cytogenetic bands as well as a map position is shown below the plot.

Fig. 2.

Individual profiles of high copy number changes. A, high-level amplifications on 3q21-q29 for SqCs9 and SqCs22. B, a homozygous deletion on 10q23.31 for SqCs2. In the intensity ratio profiles, the x-axis represents the map position of corresponding clone according to the UCSC human genome (May 2004 freeze) and the intensity ratios were assigned to the y-axis. The schematic presentation of cytogenetic bands as well as a map position is shown below the plot.

Close modal

Minimal regions of recurrent genomic changes. High copy number changes were relatively rare among the 50 cases. However, single copy number changes were more common and widespread. In total, 13 MAR gains (MAR-G) and 4 MAR losses (MAR-L) were identified. Table 2 lists the map position, size, and cancer-related genes located in the 17 MARs. Examples of MAR-G and MAR-L are illustrated in Fig. 3. The MAR-G on 1p36-p34, which was observed in 12 cases, contains several putative cancer-related genes such as PAX7, FGR, LCK, and MYCL1. In addition, another MAR-G on 1p32.3, which was found in nine cases, contains the putative cancer-related gene, TTC4 (Fig. 3A). The MAR-L on 5q23.2-q31.1 in seven cases includes several putative tumor suppressor genes such as IRF1, CDKL3, and RAD50 (Fig. 3B).

Table 2.

Minimal regions of recurrent copy number changes

ChangeCloneCytobandMap positionSize (Mb)Frequency* (squamous cell carcinomas/adenocarcinomas)Putative cancer-related genes
Gain RP4-560M15/RP4-534D1 1p36.21-p34.1 14.94-45.92 30.97 12 (6/6) PAX7, FGR, LCK, MYCL1 
 RP5-1070D5 1p32.3 54.77-56.08 1.31 9 (4/5) TTC4 
 RP4-706A17/RP11-137A12 1q21.1-q23.3 142.95-158.14 15.19 9 (3/6) BCL9, AF1Q, TPM3, PRCC, NTRK1 
 RP11-260K8/RP11-335E8 2p16.1-p12 58.65-76.8 18.15 7 (4/3) REL, MEIS 
 RP11-498P15/RP11-525C11 3q26.1-q28 162.86-192.98 30.11 17 (17/0) EVI1, SKIL, ECT2, PIK3CA,BCL6 
 RP11-269G2/RP1-137K24 5p15.2-p15.1 12.75-15.75 2.99 9 (3/6)  
 RP3-349A12/RP11-501I18 6p21.31-p21.1 34.46-43.44 8.97 7 (2/5) PIM1, CCND3 
 RP11-350N15/RP11-44K6 8p12 37.82-40.24 2.42 8 (6/2) FGFR1 
 RP11-137L15/RP11-513O17 8q11.21-q12.1 48.33-58.91 10.58 7 (4/3) MOS, LYN 
 RP11-67N21/RP11-349C2 8q24.11-q24.3 117.89-145.77 27.88 9 (6/3) NOV, MYC, WISP1, PTK2 
 CTD-2547N9/CTD-3149D2 19p13.2-p13.11 8.06-18.59 10.53 9 (3/6) LYL1, BRD4, ICAM1, JAK3, TYK2 
 RP11-9B17 19q13.12 42.52-43.38 0.85 13 (5/8) HKR1, SPINT2 
 RP5-1107C24/RP13-152O15 20q13.33 59.71-62.16 2.44 7 (3/4) BIRC7, EEF1A2, PTK6, TNFRSF6B 
Loss RP11-434D11/CTB-28J9 5q23.2-q31.1 125.7-135.73 10.03 7 (5/2) CDKL3,RAD50, IRF1 
 RP11-478B20/RP11-516G5 13q21.1 54.6-56.16 1.55 7 (2/5)  
 RP11-480K16 13q34 111.9-112.49 0.58 8 (4/4)  
 RP4-715N11 20q13.2 49.82-51.19 1.37 9 (6/3)  
ChangeCloneCytobandMap positionSize (Mb)Frequency* (squamous cell carcinomas/adenocarcinomas)Putative cancer-related genes
Gain RP4-560M15/RP4-534D1 1p36.21-p34.1 14.94-45.92 30.97 12 (6/6) PAX7, FGR, LCK, MYCL1 
 RP5-1070D5 1p32.3 54.77-56.08 1.31 9 (4/5) TTC4 
 RP4-706A17/RP11-137A12 1q21.1-q23.3 142.95-158.14 15.19 9 (3/6) BCL9, AF1Q, TPM3, PRCC, NTRK1 
 RP11-260K8/RP11-335E8 2p16.1-p12 58.65-76.8 18.15 7 (4/3) REL, MEIS 
 RP11-498P15/RP11-525C11 3q26.1-q28 162.86-192.98 30.11 17 (17/0) EVI1, SKIL, ECT2, PIK3CA,BCL6 
 RP11-269G2/RP1-137K24 5p15.2-p15.1 12.75-15.75 2.99 9 (3/6)  
 RP3-349A12/RP11-501I18 6p21.31-p21.1 34.46-43.44 8.97 7 (2/5) PIM1, CCND3 
 RP11-350N15/RP11-44K6 8p12 37.82-40.24 2.42 8 (6/2) FGFR1 
 RP11-137L15/RP11-513O17 8q11.21-q12.1 48.33-58.91 10.58 7 (4/3) MOS, LYN 
 RP11-67N21/RP11-349C2 8q24.11-q24.3 117.89-145.77 27.88 9 (6/3) NOV, MYC, WISP1, PTK2 
 CTD-2547N9/CTD-3149D2 19p13.2-p13.11 8.06-18.59 10.53 9 (3/6) LYL1, BRD4, ICAM1, JAK3, TYK2 
 RP11-9B17 19q13.12 42.52-43.38 0.85 13 (5/8) HKR1, SPINT2 
 RP5-1107C24/RP13-152O15 20q13.33 59.71-62.16 2.44 7 (3/4) BIRC7, EEF1A2, PTK6, TNFRSF6B 
Loss RP11-434D11/CTB-28J9 5q23.2-q31.1 125.7-135.73 10.03 7 (5/2) CDKL3,RAD50, IRF1 
 RP11-478B20/RP11-516G5 13q21.1 54.6-56.16 1.55 7 (2/5)  
 RP11-480K16 13q34 111.9-112.49 0.58 8 (4/4)  
 RP4-715N11 20q13.2 49.82-51.19 1.37 9 (6/3)  

NOTE: Gain and loss in the first column represent MAR-G and MAR-L, respectively.

*

The frequency represents the number of samples with the corresponding genomic changes in two kinds of NSCLC subtypes.

Fig. 3.

Examples of minimal regions of genomic gain or loss. MAR was defined as a commonly altered segment recurring for at least seven cases. Each sample is represented as an individual lane. MARs are schematically shown as a colored box below the cytogenetic bands: red, copy number gain; green, copy number loss; black, no change. A, two minimally gained regions on chromosome 1p with different genomic sizes in 16 cases. B, minimal region of losses on chromosome 5q common to seven NSCLC samples.

Fig. 3.

Examples of minimal regions of genomic gain or loss. MAR was defined as a commonly altered segment recurring for at least seven cases. Each sample is represented as an individual lane. MARs are schematically shown as a colored box below the cytogenetic bands: red, copy number gain; green, copy number loss; black, no change. A, two minimally gained regions on chromosome 1p with different genomic sizes in 16 cases. B, minimal region of losses on chromosome 5q common to seven NSCLC samples.

Close modal

Correlation between minimally altered regions. Pairwise correlation analysis between the MARs was done to determine if such genomic changes appeared concordantly in a set of NSCLC cases. For comparison, all possible combinations between the 17 MARs were considered except for pairs on the same chromosomal arms. A significantly positive correlation was observed for three pairs of MARs (see Supplementary Table S3). The MAR-G on 19q13.1 correlated with the MAR-Gs on 6p21.3-p21.1 (r = 0.549; P = 0.0482) and 19p13.2-p13.1 (r = 0.672; P = 0.0016). Another significant association was found between two MAR-Gs on 8p12.2-p12.1 and 8q11.2-12.1 (r = 0.610; P = 0.0370).

Association between genomic aberrations and clinical characteristics. Four types of clinical variables (age, stage, lymph node, and recurrence) were analyzed for their association with the genomic alterations identified (see Supplementary Table S4). Significant associations were observed for the MAR-L on 13q21 with cancers from those aged <60 and an advanced stage (stages III and IV). The chromosomal gain of Xq was also found to be associated with an advanced stage and disease recurrence. The MAR-G on 1p32 and a chromosomal gain of Yp were associated with being lymph node–negative.

Survival analysis was done to assess the prognostic values of the genetic aberrations identified. Using Kaplan-Meier methods, we identified that six genetic aberrations associated with a relatively poor survival (Fig. 4); gain of 10p (P = 0.0091) and 16q (P = 0.0262), loss of 9p (P = 0.0082) and 13q (P = 0.0019), and the MAR-Gs on 6p21 and 19q13 (P = 0.0265 and 0.0295, respectively). Multivariate analyses using all the genetic alterations identified, as well as the clinical variables such as age, gender, stage, treatment, metastasis, and recurrence showed that four genetic alterations and three clinical variables remained independent factors to be significantly associated with a poor survival outcome (Table 3). One of the four genetic alterations was a MAR-G on 6p21 and the other three were a loss of 9p, and gains of 7p and 9p.

Fig. 4.

Kaplan-Meier survival curves. The survival curves for the cases with (thick line) or without (thin line) specific genomic changes are plotted using the Kaplan-Meier method. The chromosomal changes associated with relatively poor survival are presented with the significance level; gain of 10p (A) and 16q (B), loss of 9p (C) and 13q (D), and MAR-Gs on 6p21 (E) and 19q13 (F).

Fig. 4.

Kaplan-Meier survival curves. The survival curves for the cases with (thick line) or without (thin line) specific genomic changes are plotted using the Kaplan-Meier method. The chromosomal changes associated with relatively poor survival are presented with the significance level; gain of 10p (A) and 16q (B), loss of 9p (C) and 13q (D), and MAR-Gs on 6p21 (E) and 19q13 (F).

Close modal
Table 3.

Independent predictors of poor survival in 50 NSCLCs

VariableHazard ratio95% Confidence intervalP
MAR on 6p21 3.961 1.349-11.626 0.0122 
Loss of 9p 4.256 1.746-10.373 0.0014 
Gain of 7p 15.563 3.399-71.268 0.0004 
Gain of 9q 9.546 1.400-65.077 0.0212 
Sex (male) 9.528 1.360-66.733 0.0232 
Stage 3.916 1.212-12.659 0.0226 
Metastasis 4.428 1.763-11.121 0.0015 
VariableHazard ratio95% Confidence intervalP
MAR on 6p21 3.961 1.349-11.626 0.0122 
Loss of 9p 4.256 1.746-10.373 0.0014 
Gain of 7p 15.563 3.399-71.268 0.0004 
Gain of 9q 9.546 1.400-65.077 0.0212 
Sex (male) 9.528 1.360-66.733 0.0232 
Stage 3.916 1.212-12.659 0.0226 
Metastasis 4.428 1.763-11.121 0.0015 

NOTE: Cox proportional hazards regression after adjusting for age, treatment, and recurrence.

Using whole-genome array CGH strategy, we successfully identified novel chromosomal aberrations as well as previously identified ones in NSCLC. This study focused on the potentially meaningful genomic changes such as recurrent single copy changes as well as high-level amplifications or deletions. For this, microdissection was used to remove the nontumor tissues, and the microdissected DNA was hybridized to arrays without performing whole genome amplification, which reduced any possible bias due to a random amplification.

The frequent chromosomal changes in this study are largely consistent with previous cytogenetic analysis (711), including a loss of the Y chromosome in male patients (5, 6). It is notable that the copy number alterations on the small chromosomes such as 19, 20, and 22 were much more frequent in our study. This might be due to the differences in the analytic methods. However, it is more likely to reflect the potential of array CGH to improve the low resolution of conventional CGH as described elsewhere (25). The genomic size of the high copy number changes ranged from 0.31 to 14.78 Mb, and most of them were <5 Mb. Genomic changes <5 Mb are likely to be novel because they cannot be detected using conventional CGH (12).

To date, two array-based CGH analyses of lung cancer have been published (26, 27). Our results are in general agreement with the previous array CGH results. For example, the 30-Mb sized recurrently gained region around 3q26 reported by Massion et al. (26) overlapped with the MAR-G on 3q26-q28 in this study. Other metaphase CGH studies have also shown frequent amplifications around the same region in squamous cell carcinomas (911). However, some of the MARs identified were not consistent with them. Gain of 3q26 was the only recurrent alteration in Massion et al.'s study, whereas we identified not only 3q26 but also 16 more MARs across various chromosomes. The difference is thought to be due to the resolution. This study placed 2,987 large insert clones in 1 Mb intervals (average of 125 clones per chromosome), whereas they used 348 BAC clones (average of 15 clones per chromosome, at most 78 clones on chromosome 3). That might explain why more MARs could be defined. In Jiang et al.'s study (27), the recurrent genomic alterations were only partly compatible with this result, and there was no clear minimal recurrent gain on 3q. They used a cDNA microarray rather than a BAC/PAC array for CGH analysis. Despite its advantages, the sensitivity and reliability for detecting copy number changes, particularly single copy changes, is known to be limited (18).

Several interesting cancer-related genes are located in the genomic alteration regions identified in this study. For example, PIK3CA on 3q26-q28, which is one of the genes in the most common MAR-G in this study, is believed to contribute to the tumorigenesis of squamous cell carcinomas by involving in the phosphoinositide-3-OH kinase signaling pathway (26). The ECT2 oncogene, which is located in the same locus, is known to activate the Rho signaling pathway leading to a malignant transformation (28). In our unpublished data,6

6

M.S. Kwon, H.M. Kang, and Y.J. Chung, manuscript in preparation.

ECT2 is frequently overexpressed in primary lung cancers. Although, there has been no report of ECT2 overexpression in lung cancer, this suggests the involvement of ECT2 in malignant transformation or the progression of lung cancer. Most high-level amplifications were usually observed in one or two cases with the exceptions of the amplifications on 3q. They might reflect the individual nature of the genomic evolution for the respective NSCLC cases. Among the putative cancer-related genes in the high-level amplification region (Table 1), the expression of the AF1Q, TPM3, REL, SKIL, ECT2, BCL6, MLLT6, YES1, and HKR genes have not been reported in lung cancer.

A homozygous deletion on 10q23.31 observed in one squamous cell carcinoma case contains the well-known tumor suppressor gene, PTEN. PTEN is known to encode lipid phosphatase, which negatively controls the signaling proteins activated in the phosphoinositide-3-OH kinase pathway (29). This suggests a potential role of the phosphoinositide-3-OH kinase signaling pathway in NSCLC pathogenesis.

In contrast to the high copy number changes that were largely limited to a few samples, single copy changes were found in many more samples, which is indicative of a shared mechanism common to the earlier stage of NSCLC. Minimal recurrent gains and losses were successfully identified using high-resolution array CGH. Seventeen MARs of various sizes were defined. The MAR-Gs on 1p, 2p, 6p, 8p, 19p, and 20p along with MAR-Ls on 5q and 20q are believed to be novel features in lung cancer, which shows the advantage of genome-wide, high-resolution mapping of the genomic alterations. Interestingly, three pairs of MAR-Gs (19q13.1 and 6p21, 19p13 and 19q13.1, and 8p12 and 8q11-12) showed significant correlations among themselves, suggesting a possible collaborative role in the tumorigenesis of NSCLC. Further investigations will be needed to confirm the functional consequences of the associations between the MARs. Some of the MARs showed significant correlations with the clinical features. This suggests that the common single copy changes identified by high-resolution analysis can be useful biomarkers for the clinical characteristics of lung cancer.

Survival analysis revealed that six genetic alterations were associated with a poor survival outcome in the univariate model (Fig. 4). Among those six alterations, a loss of 9p was reported to be associated with a poor survival outcome (30). However, there has been no report about the association between the other five genomic alterations and survival outcomes in lung cancer. These genomic alterations might be a novel genetic indicator of the prognosis of NSCLC after the appropriate validation. In particular, two of these alterations are MARs, which appeared concordantly (P = 0.0482). These two MARs, MAR-Gs on 6p21 and 19q13, contain cancer-related genes such as PIM1, CCND3 (both in 6p21), and HKR1 (19q13). The high expression level of HKR1 after administering platinum drugs has been reported to be associated with the acquisition of resistance to chemotherapy (31). There is no report demonstrating an alteration of CCND3 and PIM1 proto-oncogene in lung cancer. However, both genes are well known to be involved in the tumorigenesis pathways of various tumors. Therefore, further investigations will be needed to evaluate their specific implications in lung cancer.

Subsequent Cox regression analysis identified seven factors, including four genomic alterations such as MAR-G on 6p21, 9p loss, 7q gain, and 9q gain, to be independent indicators of a poor survival outcome. This indicates that in addition to the clinical factors, precisely defined recurrent genetic alterations can be useful biomarkers for the prognosis of NSCLC. However, due to the limited number of samples in this study, further studies with a larger sample size will be needed to confirm the prognostic implication of these genomic alterations and to identify further reliable prognostic markers.

This study showed that a well-designed high-resolution array CGH could define more novel regions possibly associated with the tumorigenesis of lung cancer. Therefore, these results will give a clue for further studies to elucidate lung cancer pathogenesis or to develop biomarkers for predicting the prognosis or treatment response of lung cancer.

Grant support: Korea Health 21 R&D Project, Ministry of Health and Welfare, Republic of Korea (01-PJ3-PG6-01GN07-0004).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Note: T-M. Kim and S-H. Yim contributed equally to this paper.

Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).

We thank the Wellcome Trust Sanger Institute Microarray Facility for printing BAC array slides.

1
Boyle P, Ferlay J. Cancer incidence and mortality in Europe, 2004.
Ann Oncol
2005
;
16
:
481
–8.
2
Jemal A, Tiwari RC, Murray T, et al. Cancer statistics, 2004.
CA Cancer J Clin
2004
;
54
:
8
–29.
3
Liotta L, Petricoin E. Molecular profiling of human cancer.
Nat Rev Genet
2000
;
1
:
48
–56.
4
Albertson DG, Collins C, McCormick F, Gray JW. Chromosome aberrations in solid tumors.
Nat Genet
2003
;
34
:
369
–76.
5
Balsara BR, Testa JR. Chromosomal imbalances in human lung cancer.
Oncogene
2002
;
21
:
6877
–83.
6
Testa JR, Liu Z, Feder M, et al. Advances in the analysis of chromosome alterations in human lung carcinomas.
Cancer Genet Cytogenet
1997
;
95
:
20
–32.
7
Berrieman HK, Ashman JN, Cowen ME, Greenman J, Lind MJ, Cawkwell L. Chromosomal analysis of non-small-cell lung cancer by multicolour fluorescent in situ hybridisation.
Br J Cancer
2004
;
90
:
900
–5.
8
Luk C, Tsao MS, Bayani J, Shepherd F, Squire JA. Molecular cytogenetic analysis of non-small cell lung carcinoma by spectral karyotyping and comparative genomic hybridization.
Cancer Genet Cytogenet
2001
;
125
:
87
–99.
9
Petersen I, Bujard M, Petersen S, et al. Patterns of chromosomal imbalances in adenocarcinoma and squamous cell carcinoma of the lung.
Cancer Res
1997
;
57
:
2331
–5.
10
Pei J, Balsara BR, Li W, et al. Genomic imbalances in human lung adenocarcinomas and squamous cell carcinomas.
Genes Chromosomes Cancer
2001
;
31
:
282
–7.
11
Bjorkqvist AM, Husgafvel-Pursiainen K, Anttila S, et al. DNA gains in 3q occur frequently in squamous cell carcinoma of the lung, but not in adenocarcinoma.
Genes Chromosomes Cancer
1998
;
22
:
79
–82.
12
Kallioniemi A, Kallioniemi OP, Sudar D, et al. Comparative genomic hybridization for molecular cytogenetic analysis of solid tumors.
Science
1992
;
258
:
818
–21.
13
Tay ST, Leong SH, Yu K, et al. A combined comparative genomic hybridization and expression microarray analysis of gastric cancer reveals novel molecular subtypes.
Cancer Res
2003
;
63
:
3309
–16.
14
Mukasa A, Ueki K, Matsumoto S, et al. Distinction in gene expression profiles of oligodendrogliomas with and without allelic loss of 1p.
Oncogene
2002
;
21
:
3961
–8.
15
Pinkel D, Segraves R, Sudar D, et al. High resolution analysis of DNA copy number variation using comparative genomic hybridization to microarrays.
Nat Genet
1998
;
20
:
207
–11.
16
Albertson DG, Pinkel D. Genomic microarrays in human genetic disease and cancer.
Hum Mol Genet
2003
;
12
:
R145
–52.
17
Solinas-Toldo S, Lampel S, Stilgenbauer S, et al. Matrix-based comparative genomic hybridization: biochips to screen for genomic imbalances.
Genes Chromosomes Cancer
1997
;
20
:
399
–407.
18
Mantripragada KK, Buckley PG, de Stahl TD, Dumanski JP. Genomic microarrays in the spotlight.
Trends Genet
2004
;
20
:
87
–94.
19
Fiegler H, Carr P, Douglas EJ, et al. DNA microarrays for comparative genomic hybridization based on DOP-PCR amplification of BAC and PAC clones.
Genes Chromosomes Cancer
2003
;
36
:
361
–74.
20
Chung YJ, Jonkers J, Kitson H, et al. A whole-genome mouse BAC microarray with 1 Mb resolution for analysis of DNA copy number changes by array CGH.
Genome Res
2004
;
14
:
188
–96.
21
Kim SY, Nam SW, Lee SH, et al. ArrayCyGHt: a web application for analysis and visualization of array CGH data.
Bioinformatics
2005
;
21
:
2554
–5.
22
de Leeuw RJ, Davies JJ, Rosenwald A, et al. Comprehensive whole genome array CGH profiling of mantle cell lymphoma model genomes.
Hum Mol Genet
2004
;
13
:
1827
–37.
23
Hackett CS, Hodgson JG, Law ME, et al. Genome-wide array CGH analysis of murine neuroblastoma reveals distinct genomic aberrations which parallel those in human tumors.
Cancer Res
2003
;
63
:
5266
–73.
24
Wang P, Kim Y, Pollack J, Narasimhan B, Tibshirani R. A method for calling gains and losses in array CGH data.
Biostatistics
2005
;
6
:
45
–58.
25
Schraders M, Pfundt R, Straatman HM, et al. Novel chromosomal imbalances in mantle cell lymphoma detected by genome-wide array-based comparative genomic hybridization.
Blood
2005
;
105
:
1686
–93.
26
Massion PP, Kuo WL, Stokoe D, et al. Genomic copy number analysis of non-small cell lung cancer using array comparative genomic hybridization: implications of the phosphatidylinositol 3-kinase pathway.
Cancer Res
2002
;
62
:
3636
–40.
27
Jiang F, Yin Z, Caraway NP, Li R, Katz RL. Genomic profiles in stage I primary non–small cell lung cancer using comparative genomic hybridization analysis of cDNA microarrays.
Neoplasia
2004
;
6
:
623
–35.
28
Saito S, Liu XF, Kamijo K, et al. Deregulation and mislocalization of the cytokinesis regulator ECT2 activate the Rho signaling pathways leading to malignant transformation.
J Biol Chem
2004
;
279
:
7169
–79.
29
Cantley LC, Neel BG. New insights into tumor suppression: PTEN suppresses tumor formation by restraining the phosphoinositide 3-kinase/AKT pathway.
Proc Natl Acad Sci U S A
1999
;
96
:
4240
–5.
30
Tomizawa Y, Adachi J, Kohno T, et al. Prognostic significance of allelic imbalances on chromosome 9p in stage I non-small cell lung carcinoma.
Clin Cancer Res
1999
;
5
:
1139
–46.
31
Oguri T, Katoh O, Takahashi T, et al. The Kruppel-type zinc finger family gene, HKR1, is induced in lung cancer by exposure to platinum drugs.
Gene
1998
;
222
:
61
–7.

Supplementary data