Genome-wide array comparative genomic hybridization analysis reveals distinct amplifications in osteosarcoma

Osteosarcoma (OS) is a primary malignant tumor of bone arising from primitive bone-forming mesenchymal cells and it accounts for approximately 60% of malignant bone tumors in the first two decades of life [1]. These tumors typically arise in the metaphyseal regions of long bones, with the distal femur, proximal tibia and proximal humerus. A significant number of osteosarcomas are of conventional type which can be subdivided into three major categories based on their predominant differentiation of tumor cells: osteoblastic, chondroblastic, and fibroblastic. Currently, only the histological response (degree of necrosis) to therapy represent gold standard for predicting the outcome in a patient with non-metastatic osteosarcoma at the time of definitive surgery [2]. Patients with lower degree of necrosis have a higher risk of relapse and poor outcome even after chemotherapy and complete resection of the primary tumor. Therefore, a better understanding of the underlying molecular genetic events leading to tumor initiation and progression could result in the identification of potential diagnostic and therapeutic targets.

Chromosomal aberrations in osteosarcoma are highly complex and characterized by high frequency of amplifications. These amplifications may result in the overexpression of genes and contribute to the genomic instability in osteosarcoma. The identification of genes within the amplified sites is crucial for understanding the biology and clinical behavior of osteosarcoma. Until, recently gene amplification has been detected by PCR, southern blot analysis or FISH-based approach using gene specific probes. These techniques are inherently restricted to the previously known amplified genes in the genome. In contrast, genome-wide screening of amplified chromosomal regions with CGH has become an important tool for the detection of amplified regions in the tumor genome. So far published chromosomal CGH studies in osteosarcoma have identified several high-level chromosomal amplifications at 1p22, 1p31, 1q21, 1q23, 2q24, 3p25, 3q26, 6q24.3, 4q12, 5p14-p15, 5q33, 6p12-p21, 6q24.3, 7p21-p22, 8q12-q23, 10p21, 10q11.1, 10q22, 11q13, 11q23, 12p13, 12q12-q15, 17p11.2, 17q21, 18q22, 19p13.1 and 20p11.2 [3‐7]. However, conventional CGH has limited sensitivity and resolution (~10–15 megabases) because of its dependence on the morphology of metaphase chromosomes. In addition, extensive follow-up work is required to identify candidate genes after regions of gain or loss have been identified. Recently, novel method termed as array-based comparative genomic hybridization (array CGH) has been described, which enables high throughput quantitative measurement of high-resolution DNA copy number changes throughout the genome [8]. This method is based on hybridization of differentially labeled test and reference DNAs to an array of mapped human genomic DNA fragments (~100–200 kb) and has been recently applied to human and mouse tumors [9‐14]. To identify high-resolution copy number, we used array CGH to the panel of 48 tumors. The resolution of array CGH technology combined with human genome database not only allowed a precise identification of amplicons but also suggested the possible target genes within the amplicons.

To define the gains and losses in our experiments, we used invariant analysis for the first time to describe genomic changes by array CGH. In this method, we defined an invariant clone set that has low variance of log ratios among all the array experiments. After the mean and standard deviation of the log ratios in the invariant set of each experiment were calculated, clones that have higher or lower log ratios than the mean +/- 2SD of the invariant set (upper bound and lower bound) were used to define gains and losses. We chose to use this method because it addresses some of the shortcomings of the modeling method, such as using all information provided in a set of experiment to determine the unchanged population instead of using one experiment at a time. However, the variation of each experiment is accounted for because the thresholds are calculated using the invariant set from each experiment. It also does not require a separate reference set for comparison. Finally, it provides an adjustable cutoff to optimize the thresholds to the training data, if provided.

The amplified and homozygously deleted clones were defined to have at least 2 fold of the upper bound and 4-fold of lower bound, respectively. Figure 1 summarizes the high-resolution DNA copy number changes identified by array CGH in 48 osteosarcomas derived from 42 patients. Copy number changes were detected involving small genomic regions, whole chromosomes, and chromosomal arms showing homozygous deletions and high-level amplifications.

This study represents the first application of genome-wide copy number changes by array CGH in osteosarcoma. Recent studies in breast, renal and bladder cancer showed the potential assessment of this technology in detecting high-resolution copy number changes [9, 11, 14]. This approach will augment the identification of cancer causing genes by relating the clone information directly with sequence information from human genome database. In this study, we used array CGH to screen for high-resolution DNA copy number changes and precise identification of amplifications in a panel of 48 osteosarcomas.

Gene amplification is an important genetic mechanism in human cancers, as it clearly associated with tumor progression and has a prognostic significance and has even provided a target for therapeutics [17, 18]. These amplifications are often seen at the cytogenetic level as homozygously staining regions (hsrs) or double minute chromosomes (dms). However, cytogenetic recognition of amplifications doesn't contribute to the mapping and identification of amplified DNA sequences. The advent of CGH points an ever-increasing number of chromosomal amplifications in various tumors. These amplifications contribute to the genomic instability in tumors. We have recently shown that the mutation of p53 significantly correlates with genome-wide DNA instability and seems to represent a major genetic factor contributing to the extremely high levels of genomic instability found in high-grade osteosarcomas [19].

Our analysis have identified frequently amplified clones from 6p11.2-p21, 8q12.1, 8q24.13, 12q12, 12q13-q15, 12q21-q21.33, 16p13 and 17p11.2-p12. Amplification of clones from 6p12-p21 region was noted in 25% of the cases analyzed. This was consistent with the previously published results by CGH. By array CGH, we refined the 6p amplicon to 9.4 Mb with amplification peak for clone RP11-81F7. We recently demonstrated the origin of 6p amplicon as consequence of tandem duplication of clones RP11-81F7 and RP11-79F13 [7]. Based on combined array CGH and FISH analysis suggest CDC5L, HSPCB, and NFKBIE, and HGNC and MRPL14 are the target genes from 6p12-p21 amplicon. Of these genes, CDC5L may be an important gene in cancer because of its role as a positive cell cycle regulator for G2/M transition[20]. Consistent with our analysis, overexpression of HSPCB was shown recently by cDNA microarray studies on osteosarcoma [21]. This protein was shown to play an important role in assemble/disassembly of tubulin by inhibiting tubulin polymerization.

High-level amplification of clones on 12q revealed three distinct sites of amplifications – AMP1 (12q12), AMP2 (12q14.1) and AMP3 (12q21.33). Pervious studies have shown the amplification GLI, CHOP, SAS, HMGI-C, CDK4, HDM2, and PRIM1 from 12q13-q15 region in osteosarcoma [22, 23]. The present array CGH analysis identified a possible target gene IFNG from AMP2 (RP11-298M11; p = 0.0000001), which is physically mapped close to the HDM2 oncogene locus[24]. Previous studies demonstrated that T-cell production of IFNG strongly suppresses osteoclastogenesis by interfering with the RANKL-RANK signaling pathway. IFNG induces rapid degradation of the RANK adaptor protein, TRAF6, resulting in strong inhibition of the RANKL-induced activation of the transcription factor NFKB and JNK [25]. The AMP3, which was 23.3 Mb distal to AMP2 containing RP11-89F6. Our analysis from AMP3 revealed two interesting candidate genes: transcription factor ELK3 and PCTAIRE protein kinase 2 (PCTK2). ELK3 is a member of the ETS-domain transcription factor family and the protein is activated by signal-induced phosphorylation [26]. The protein encoded by PCTK2 belongs to the cdc2/cdkx subfamily of the ser/thr family of protein kinases and play an important role in the regulation of the mammalian cell cycle [27]. High-level amplification of three clones from 12p13 was noted in case 27 and the amplicon span 4.6 Mb with peak amplification for clone RP11-89D16. No candidate genes contained with in this BAC. Amplification 12p has been reported previously in 9/19 high-grade osteosarcomas by CGH. Recent FISH analysis has identified the amplification of CCND2, ETV6, and KRAS2 from 12p region [28].

Amplification of 17p11.2 was noted in 27% of the cases analyzed by array CGH. Our array CGH analysis has identified three clones with high-level amplifications that spans ~3.7 Mb region on 17p11.2. Several candidate genes were identified within these clones (TPP3A, SMCR5, DRG2, FL11, MYCD, SOX 17, ELAC2, and PMP22). Recent studies have shown the amplification of some of the genes identified in the present study (PMP22, and TOP3A) from 17p11.2-p12 in high-grade OS by semi-quantitative PCR and cDNA microarrays [29, 30].

The present array CGH analysis has identified seven recurrent clones exhibiting homozygous deletions from 1q25.1, 3p14.1, 13q12.2, 4p15.1, 6q12, 6q12 and 6q16.3. These chromosomal regions were consistent with previously reported studies by loss of heterozygosity (LOH) and CGH [3‐7, 31]. The clone, RP11-90M15 (13q12.2) contain possible candidate gene MTMR6, a protein-tyrosine phosphatase gene and shown to be present within a cloned region that encompasses a translocation breakpoint t(8;13) in an atypical myoproliferative disorder [32]. Homozygous deletions of two clones spanning approximately 2.6 Mb of 6q12 region containing candidate genes – nuclear fragile X mental retardation protein interacting protein 1 pseudogene (NUFIP1P) and BAI3 gene (brain-specific angiogenesis inhibitor gene), which is to homologous to BAI1 and shown to suppress glioblastoma [33].

In summary, high resolution array-based CGH revealed large number of chromosomal aberrations previously identified in osteosarcoma by chromosomal CGH and conventional cytogenetic methods. The present study allowed precise identification of smaller DNA copy number alterations, which suggest the presence of specific target genes in osteosarcoma. Although this study suggested several possible target genes from amplified regions from 6p, 8q, 12q and 17p, but these genes should be validated by other molecular and immunohistochemical approaches on well-defined large patient samples. Further, interaction or association studies between small genomic losses and gains will facilitate the identification of new genetic pathways in the pathogenesis of osteosarcoma.