2
Oncogene amplifications in non small cell lung cancers with supernumerary ring chromosomes: characterization by tiling resolution bacterial artificial chromosome microarrays
Maria Soller, MD, PhD1, Anna Karlsson, MSc2, Anna Collin, PhD1, Göran Jönsson, PhD2, Maria Planck, MD, PhD2*
Departments of Clinical Genetics 1 and Oncology 2 at Lund University Hospital, Sweden.
* Correspondence to: Maria Planck, Lund University, Clinical Sciences, Lund, Department of Oncology, Barngatan 2:1, SE-221 85 Lund, Sweden
Phone +46-46-177501, Fax +46-46-147327, E-mail
Abstract
A small fraction of lung cancers have a karyotype displaying supernumerary ring chromosomes. We obtained gene copy number profiles by array based comparative genomic hybridization, using whole-genome tiling resolution bacterial artificial chromosome microarrays, for five primary non small cell lung cancers (NSCLC) with a supernumerary ring chromosome as their sole clonal karyotypic aberration. Regions with about 2- to 3-fold amplifications were revealed in 4/5 tumors. Amplicons within the chromosomal regions 3q26.1-3q29, 17q11.1-17q21.2, and 19q12-19q13.2 were shared between the two squamous cell carcinomas whereas the two adenocarcinomas demonstrated a single high-level amplification on 14q13.2-14q21.2 and amplifications on 7p22.3-7p11.2 and 21q22.13-21q22.3, respectively. Several known oncogenes were identified as presumed targets of amplification, e.g. PIK3CA, ERBB2, EGFR and CCNE1. The rings and the candidate gene-containing amplifications may simply reflect independent, coexisting, alterations detected by different techniques. Alternatively, supernumerary ring chromosomes in NSCLC harbor amplified oncogenes and thus have a similar role in tumorigenesis as described for some mesenchymal tumors.
Keywords: Lung cancer, ring chromosome, array-based CGH, FISH, chromosomal imbalances, gene amplification
Introduction
Lung cancer, with 1.35 million cases diagnosed worldwide each year, is a common and increasing type of cancer, with the majority of cases attributable to cigarette smoking. The disease is the leading cause of cancer death in the world and, despite diagnostic and therapeutic improvements during the past two decades, the overall 5-year survival rate is still below 15% (1). Lung carcinomas are histopathologically divided into two major categories; small cell lungcancer (SCLC) and non small cell lung cancer (NSCLC). All major histological types of invasive lung carcinoma display multiple genetic alterations, generally believed to have accumulated during a multistep carcinogenesis. The tumors frequently exhibit multiple karyotypic changes with both numerical and structural aberrations that may involve any chromosome (2). In NSCLC, frequently reported karyotypic changes include losses on chromosomal arms 3p, 6q, 8p, 9p, 9q, 13q, 17p, 18q, 19p, 21q, and 22q. The chromosomal arms most frequently involved in gains revealed by karyotyping are 1q, 3q, 5p, 7p, 7q, 11q and 12q (3). Studies of recurrent alterations by comparative genomic hybridization (CGH) have made it possible to identify cancer-related genes through observations of lost or gained chromosome regions (4). CGH analysis has confirmed the multiple genomic imbalances in NSCLC, including gains of chromosome arms 1q, 3q, 5p and 8q, and loss of 3p, 8p, 9p, 13q, and 17p (3). Some imbalances seem to occur at different frequencies in the two major subtypes of NSCLC, squamous cell carcinoma (SCC) and adenocarcinoma (AC). Indeed, gain of 3q24-3q26 is twice as common in SCCs as in ACs, with PIK3CA, one of the most highly mutated oncogenes identified in human cancers, located at 3q26 as a presumed target of amplification (3, 5). On the other hand, gains in 1q22-32 are more frequent in ACs (3).
A small fraction of lung cancers demonstrate a karyotype displaying supernumerary ring chromosomes. These are unstable chromosomal structures that are generally rare in epithelial tumors but frequently observed in mesenchymal tumors (2, 6). Ring chromosomes are typically formed through breakage of both chromosomal arms and fusion of the proximal ends into a ring-shaped structure that thus may contain amplified chromosome material. The size and number of rings may vary greatly between cells within a tumor, perhaps due to frequent recombination based on breakage-fusion-bridge cycles (7). The flexible/unstable structure of ring chromosomes may constitute a useful mechanism for rapid change in oncogene copy number during tumor evolution (6). In mesenchymal tumors, ring chromosomes do typically contain amplified material from chromosome arm 12q that include known oncogenes such as MDM2, SAS and CDK4. The pathogenic effects of ring chromosomes in lung cancer, as in any epithelial tumor type, are, however, largely unknown. Cytogenetic analyses of a consecutive series of primary NSCLCs collected at our hospital revealed ring chromosomes as sole abnormality in 4/114 ACs (8), 1/26 large cell carcinomas (9) and in 2/111 SCCs (10). The origin of the supernumerary ring chromosomes could not be disclosed by conventional cytogenetic methods. In the present study, five ring-containing primary NSCLC were selected for characterization by array based CGH (aCGH), using whole-genome tiling resolution bacterial artificial chromosome (BAC) microarrays. This recently developed method allows for characterization of DNA copy number changes at a resolution only limited by the number of BAC clones used for the arrays (11, 12). We used arrays encompassing 32433 overlapping BAC clones covering the whole genome, i.e. the tumor DNA could be analyzed with an average resolution of 70 kbp.
Results
We obtained gene copy number profiles by aCGH for five primary NSCLCs with a supernumerary ring chromosome as their sole karyotypic aberration (Figure 1). The aCGH demonstrated a high frequency of DNA copy number changes (range 7-90 regions per tumor), most of which were at modest levels (about 1.5-fold) and with gains dominating over losses. A summary of all imbalances (log2 ratio > ± 0.2) is available as supplementary material. On 14q13.2-14q21.2 in tumor no. 5 (AC), on 3q26.1-3q29, 22q12.2 and 22q13.33 in tumor no. 2 (SCC), and on 17q11.1-17q21.2 and 19q12-19q13.2 in tumor no. 1 (SCC), we found high level amplifications (log2 ratio >1.5, i.e. at least 3-fold amplification), some of which contained known cancer-related genes (Figure 2, Table 2). Furthermore, the amplifications of the regions 3q26-3q29, 17q11-17q21, and 19q12-19q13 were seen in both SCC, although at different levels (see Table 2). Tumor no. 3 (AC) harbored distinct gains in the regions 7p22.3-7p11.2 and 21q22.13-21q22.3, at amplification levels that were approximately 2-fold (Table 2). Few (a total of 7), and modestly altered, regions of gains/losses were detected by aCGH in tumor no. 4 (AC).
Interphase FISH analysis was done using probes targeting the amplified genomic sequences mentioned above, including known oncogenes when possible. The results of the FISH analysis for the amplifications seen on the aCGH are shown in Table 2 and exemplified by Figure 3. The analysis confirmed the presence of these alterations, although the number of gene copies in the amplified region 14q13.2-14q21.2 (tumor no. 5) could not be examined, probably due to poor quality of the BAC clone selected as FISH probe.
Discussion
Ring chromosomes have been observed in various types of cancer but the knowledge on the tumor biology behind these changes is limited. Although common in a subgroup of mesenchymal tumors, ring-shaped chromosomal structures are rarely observed in epithelial tumors (2, 6). Rings were detected in only 7 of 251 primary NSCLCs that were cytogenetically analyzed at our hospital. The origin of the supernumerary ring chromosomes and their role in initiation and development of these tumors is unknown (8-10). All cases in which the ring constituted the sole karyotypic alteration (2 SCCs and 3 ACs) were selected for aCGH analysis.
The DNA copy number profiles generated by using the tiling resolution BAC arrays revealed regions with about 2- to 3-fold amplifications in 4/5 tumors. The loci for these imbalances, as well as known cancer related genes within the amplicons, are summarized in Table 2. Amplicons within the chromosomal regions 3q26.1-3q29 (Figure 2), 17q11.1-17q21.2, and 19q12-19q13.2 were shared between the two SCCs, a finding consistent with observations from studies on sarcomas harboring ring chromosomes, with each ring often containing material from multiple chromosomes (14). Similarly, tumor 3 (AC) contained amplifications of the regions 7p22.3-7p11.2 and 21q22.13-21q22.3. In contrast, in tumor 5 (AC), we demonstrated a single high-level amplification on 14q13.2-14q21.2 (Figure 2). A relatively low proportion of ring-containing cells in the tumors was observed by previous karyotyping (8, 10) and the remaining AC, tumor no. 4, may thus harbor too few amplicon-containing cells to show a sufficient copy number change in the aCGH analysis. Recurrent 14q13-14q21 amplification, or any other copy number change that could have explained the existence of the rings, can still exist undetected in this sample. The great similarities between the aCGH-patterns regarding presumed oncogenes in the two SCC with supernumerary rings as solitary cytogenetic abnormality may indicate that the amplifications reflect the presence of the ring chromosomes.
Probes targeting the regions that were amplified according to the aCGH analysis (log2 ratio >1.5) were used for subsequent FISH analysis that confirmed the amplifications at 3q11, 3q26, 3q29, 7p11, 17q12, 17q25, 19q12, 19p13, 21q22, and 22q12 (Table 2, Figure 3). The number and configuration of the corresponding extra gene copies demonstrated by interphase FISH in this study were well in line with the interphase structure of ring amplifications previously demonstrated in sarcomas (15).
In addition to the amplifications described above, the five tumors also displayed frequent low-level copy number alterations involving all chromosomes, despite having almost normal karyotypes according to previous cytogenetic analysis. All copy number changes are summarized in table and in frequency plot, available as supplementary material. Although of unknown importance, these imbalances illustrates the high resolution of the whole-genome tiling resolution 32k BAC arrays used, which, compared to conventional methods, enabled us to detect more and smaller changes (11, 12, 14). As expected from the karyotypes, gains dominated over losses.
In order to ensure the specificity of the high-level amplifications, we also used FISH probes targeting regions that were not gained as revealed by aCGH (log2 ratio < 0,2), i.e. presumed to be representative of the general DNA content of the tumors. By using these probes, we observed a polyploidy in all tumors, corresponding to approximately 4-5 copies per gene and thus at a level not comparable to the high number of copies demonstrated with the FISH probes specific for the amplified regions.
Ring chromosomes are frequently observed in bone and soft tissue tumors as well as in leukemias and lymphomas (2, 6, 14). The role of these chromosomal structures in other tumor types is largely unknown. There are, however, a few reported cases of acquired supernumerary ring chromosomes in cancers of the esophagus, ovary, urinary bladder, lung, brain, and kidney (10, 16-18). A number of difficulties must be addressed when studying the formation, structure and implications of somatically acquired ring chromosomes in tumors; There is an intratumoral variation of number and size of the rings and the karyotypic banding pattern is typically diffuse and the chromosomal origin hard to establish. Moreover, the highly unstable ring structures may be completely changed during tumor development, probably allowing changes up and down in copy numbers each cell division (6). The rings in our study may represent unimportant changes that simply coexist with the more obviously cancer-related amplifications observed by aCGH. Indeed, the rings may represent clones that simply had a growth advantage during cell culturing, a process that do not always allow the relevant abnormal tumor cells to grow. Given the low proportion of ring-containing clones in the tumor samples, the aCGH does not necessarily reflect the ring-containing karyotype but rather DNA copy number changes that were not possible to pin-point by cytogenetics or conventional CGH. Furthermore, the presumed importance of the oncogene-containing amplifications observed by aCGH for the development of the tumors in our study can also be argued against. Indeed, previous studies have demonstrated that far from all gene amplifications are overexpressed according to their gene expression profiles (14, 19). Thus, the exact role of ring chromosomes in the development of the lung cancers in our study remains to be elucidated. However, we believe that the specific amplicons, as revealed by aCGH, may constitute the origin of the rings. This hypothesis is strengthened by the interphase FISH analyses, which confirm the presence of the amplifications and display a pattern very similar to interphase FISH on ring-containing sarcomas (15), and the co-existence of similar amplifications in ring-containing tumors of the same histological type. The rings in the present study may thus harbor cancer-related genes in a mode consistent with the model suggested for MDM2 amplifications in ring chromosomes in malignant fibrous histiocytomas (6). Of the genes located within the amplified regions, Table 2 aims to summarize those with previously reported oncogenic function, involved in e.g. cell motility, transcription, or signal transduction. For example, the well characterized oncogene PIK3CA has been implicated in SCC tumor progression and the region 3q11.2-3q29 was accordingly amplified in both our SCC tumors (20). Furthermore, the amplified region on 17q11.1-17q21.2 harbors the oncogene ERBB2. Somatic mutation of ERBB2 or partial gain of chromosome 17 has been associated with a poor prognosis of NSCLC (21). The amplicon encompassing 7p22.3-7p11.2 contains, among other presumed oncogenes, the EGFR gene, with prognostic and therapeutic implications in NSCLC. In a previous study that used whole genome tiling BAC microarray for characterization of NSCLC cell lines, the regions 7p22.1-7p22.3 and 7p15.3-7p11.2 were amplified in more than 80% of the cases (22). We detected a high level amplification at 14q13.2-14q21.2 and part of this region (14q13.2-14q13.3) was highly amplified also in 6/28 NSCLC cell lines in the study by Garnis et al. (22). This region harbors for example the gene PAX9 which is frequently expressed in cancer and required for the growth and survival of cancer cells (23). In a study using single nucleotide polymorphism (SNP) arrays for determination of copy number alterations, this same 14q13.3 amplification was described as the most common (6%) focal event in lung adenocarcinomas, and the transcription factor NKX2-1 (essential for pneumocyte formation) was identified as the target gene and thus suggested as an important protooncogene in lung adenocarcinoma (Ref Weir et al. Nature 2007). Since the karyotypes have not been investigated in these studies, the occurrence of ring chromosomes as the mechanism of 14q amplification in these tumors is unknown (22 + Weir).
In summary, we identified several known oncogenes as targets of amplifications in primary NSCLC with supernumerary ring chromosomes. We observed these same amplifications as an interphase FISH pattern that is consistent with the behavior of interphase ring structures previously demonstrated in sarcomas. It is not finally proved whether ring chromosomes in NSCLC contribute to oncogenic change or constitute innocent cytogenetic alterations that exist in parallel with more significant amplifications of known oncogenes. Future aCGH studies on tumors available for metaphase FISH could hopefully give direct evidence for the presence of the aCGH amplifications within the rings.