Amplification of several genes that map to a region of chromosome 17p11.2, including COPS3, was observed in high-grade osteosarcoma. These genes were also shown to be overexpressed and may be involved in osteosarcoma tumorigenesis. COPS3 encodes a subunit of the COP9 signalosome implicated in the ubiquitination and ultimately degradation of the P53 tumor suppressor. To determine the relation between COPS3 amplification, P53 mutation, and patient outcome in osteosarcoma, tumors from a large cohort of patients with high-grade osteosarcoma and long-term clinical follow-up were examined.
Quantitative real-time polymerase chain reaction (PCR) was performed to detect copy number changes for COPS3, as well as additional genes (NCOR1, TOM1L2, and PMP22) from the 17p11.2 amplicon, in 155 osteosarcomas from a prospective collection of tumors with corresponding clinical data. Univariate and multivariate analyses were performed to assess differences in survival between groups.
Amplification of COPS3, detected in 31% of the osteosarcomas, was strongly associated with large tumor size (P = .0009), but was not associated with age at diagnosis, site, sex, and tumor necrosis. COPS3 amplification was significantly correlated with a shorter time to metastasis with an estimated hazard ratio (HR) of 1.61 (95% confidence interval [CI], 1.02–2.55) in univariate analysis (log-rank test, P = .042). However, in an a priori multivariate Cox model including the other clinical parameters, the HR for COPS3 amplification decreased to 1.32 (95% CI, 0.82–2.13, P = .25), mainly due to the strong correlation with tumor size. COPS3 amplification and P53 mutation frequently occurred in the same tumors, suggesting that these are not mutually exclusive events in osteosarcoma. Although not statistically significant, patients whose tumors exhibited both molecular alterations tended to be more likely to develop metastasis compared with patients with either COPS3 amplification or P53 mutation alone.
Cytogenetic studies reveal that the majority of osteosarcomas are characterized by complex chromosomal abnormalities.1 Loss of chromosomes 9, 10, 13, and 17 have been reported most commonly, with chromosomal regions 13q14 and 17p13 harboring deletions of the retinoblastoma (RB1) gene and P53 tumor suppressor gene, respectively. In addition, frequent chromosomal gains have been observed to involve chromosomal regions 1q, 3q, 6p, 8q, 12q, 14q, 17p, Xp, and Xq.1, 2 Between 13% and 47% of high-grade osteosarcomas have been found to contain amplification of several genes that map to a region of chromosome 17p11.2, including COP9 constitutive photomorphogenic homolog subunit 3 (COPS3), nuclear receptor corepressor (NCOR1), target of myb1-like 2 (TOM1L2), and peripheral myelin protein 22 (PMP22), and may be involved in osteosarcoma tumorigenesis.3–8
COPS3 encodes 1 subunit of a highly conserved complex, the COP9 signalosome (CSN), which has been implicated in the ubiquitination and ultimately degradation of the p53 tumor suppressor.9–11 Another protein involved in loss of the p53 tumor suppressor function is MDM2, whose amplification and/or overexpression has been observed in some osteosarcomas.12, 13 COP9 signalosome-specific phosphorylation has been shown to target p53 to MDM2-mediated degradation7, 11 and it has been postulated that amplification and overexpression of COPS3 in osteosarcoma may represent another route to p53 protein degradation equivalent to inactivation of P53 by mutation. An increased amount of the COP9 signalosome, induced by overexpression of COPS3, could function in diverse cellular processes, including regulation of the cell cycle, proliferation, differentiation, apoptosis, and signal transduction. It has also been suggested that COPS3 amplification and P53 mutation may be mutually exclusive events in osteosarcoma.7
We previously observed that P53 mutations are common in osteosarcoma, occurring in 20% of high-grade tumors.14 However, in a cohort of 196 patients with high-grade nonmetastatic osteosarcoma we found that P53 mutations were not associated with disease-free survival.15
We determined the amplification status of COPS3 and 3 other candidate genes, NCOR1, TOM1L2, and PMP22, that map to 17p11.2 (Fig. 1) in 155 high-grade osteosarcoma tumor samples and found that COPS3 was the most frequently amplified gene. To determine whether amplification of COPS3 might be of prognostic importance, we examined the association between amplification and outcome as well as other clinical characteristics in these 155 patients. We also investigated the relation between COPS3 amplification and P53 mutation because COPS3 amplification and overexpression might provide an alternative mechanism of p53 degradation.
MATERIALS AND METHODS
Clinical Data, Treatment, and Tumor Samples
One hundred fifty-five patients with high-grade osteosarcoma of the extremity, including 17 who presented with metastases at diagnosis, were evaluated in this study. Patients were treated at 1 of 6 tertiary care medical institutions: Mount Sinai Hospital, Toronto, Canada; Hospital for Sick Children, Toronto, Canada; Royal Orthopaedic Hospital, Birmingham, UK; Memorial Sloan Kettering Cancer Center, New York, NY; and Mayo Clinic, Rochester, Minn. Patients received pre- and postoperative chemotherapy or postoperative chemotherapy alone, which included doxorubicin. Surgical treatment of the primary tumor was by en bloc resection or amputation. Except for patients who presented with metastases at the time of diagnosis, all other patients were seen in regular follow-up for at least 4 years from the time of diagnosis or until systemic recurrence.
Patients were selected from a larger cohort14–16 on the basis of availability of tumor material. Each eligible patient provided a signed consent form before study entry, as approved by each participating institution's Research Ethics Board. Tumor specimens were chosen by a pathologist with the aid of frozen histologic analysis to ensure the presence of viable tumor without normal tissue contamination. Tumor samples were collected immediately after surgery, snap-frozen in liquid nitrogen, and stored at −70°C. Tumor genomic DNA was extracted by conventional techniques (Qiagen, Chatsworth, Calif). Human placenta DNA (Sigma, St. Louis, Mo) was used as an internal control.
Quantitative Real-Time Polymerase Chain Reaction (PCR) (TaqMan) Analysis of DNA Amplification
The DNA copy number was quantified for COPS3, NCOR1, TOM1L2, and PMP22 by real-time quantitative PCR using the ABI prism 7900HT system (Applied Biosystems, Foster City, Calif). Asparagine synthetase (AS) was used as an internal control gene. Primer sets and TaqMan probes were designed using the Primer Express Software v. 2.0 (Applied Biosystems; Table 1). The copy number of the gene of interest was determined using placenta DNA as a reference for standard curves. The AS copy number was used for normalization and the results are reported as copy number fold increase relative to AS.
Table 1. List of Primers and Probes of COPS3, NCOR1, TOM1L2, PMP22, and AS
5′-6-FAM- CAG GGC CGC CG TAA-3′
5′-6-FAM- ACG TCG TTC GAG GAA A-3′
Single-strand conformation polymorphism analysis (SSCP) and direct sequencing were used to analyze exons 4–10 of the P53 gene for alterations of each tumor as described previously.14, 15
Data Analysis and Statistical Tests
Tumors that exhibited a 2-fold or greater increase in gene copy number relative to control unamplified DNA from placenta were considered to have gene amplification. Tumor size is expressed as the largest dimension of the pathology specimen in cm. Differences in clinical variables based on amplification status were statistically assessed using the Fisher Exact Test (mid-P corrected) for categorical variables, and the exact Wilcoxon rank sum test (using Monte Carlo estimation based on 100,000 samples) for continuous variables. Disease-free survival was assessed using the time from diagnosis (based on the date of biopsy) to metastasis or last follow-up if continuously free of disease. Patients who presented with metastases at diagnosis were considered to have a disease-free period of 0.001 years for the purpose of the survival analysis. The log-rank test was used to assess differences in survival between groups. Cox proportional hazards regression (with ties handled using the exact conditional probability) was used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). Multivariate survival analysis was carried out using clinical parameters selected (without amplification status of any genes included in the model) from an a priori set of parameters based on an optimal Akaike information criterion (AIC) approach.17 Clinical parameters included age at diagnosis, site (distal or proximal), tumor size (untransformed, square root, and log-transformed values), sex, and necrosis (>90% vs ≤90% and missing). An optimized post-hoc model was also developed, in which necrosis was redefined to be a linear variable with a middle category for 50% to 90% necrosis, and tumor size was truncated at 12 cm and treated as a linear (untransformed) variable.
COPS3, NCOR1, TOM1L2, and PMP22 Amplification
TaqMan quantitative PCR was used to determine the DNA copy number of 4 candidate genes, COPS3, NCOR1, TOM1L2, and PMP22, in osteosarcoma specimens. We examined tumor DNAs from 155 patients for COPS3 copy number, 147 patients for PMP22 gene amplification status, and 133 patients for NCOR1 and TOM1L2 gene amplification status. COPS3 was found to be the most frequently amplified gene (48 of 155, 31%) in our patient set. Among the 48 COPS3 amplified tumors, amplification ratios were between 2.0–2.9 for 19 (40%), 3.0–3.9 for 9 tumors (19%), 4.0–4.9 for 4 tumors (8%), and greater than 5.5 for the remaining 16 (33%). The other 3 candidate genes, NCOR1, TOM1L2, and PMP22, were amplified in 22.6% (30 of 133), 20.3% (27 of 133), and 12.2% (18 of 147) of the cases, respectively. Of the 133 tumors in which all 4 genes were analyzed, 10 that were amplified for COPS3 did not exhibit amplification of the other 3 genes. Three tumor samples displayed gene amplification of NCOR1 alone and 1 tumor was amplified only for TOM1L2. No tumor specimens were found to be amplified for PMP22 alone. These results suggest that COPS3 is likely the target gene for this amplicon. For this reason, we focused on the clinical results for COPS3 below.
Patient and Tumor Characteristics
The patient and tumor characteristics for those with and without COPS3 amplification are summarized in Table 2. Of the clinicopathologic features studied, only tumor size varied with COPS3 status (P = .0009). Patients whose tumors exhibited COPS3 amplification were larger than those without gene amplification. Although not statistically significant at the conventional 5% level, patients with COPS3 amplified tumors were more likely to have metastases at diagnosis (odds ratio [OR] = 2.18, P = .13) and more likely to have proximal compared with distally located tumors (OR = 1.86, P = .12). There was no evidence of an association between COPS3 status and patient gender, age, or chemotherapy-induced tumor necrosis.
Table 2. Clinical and Tumor Characteristics According to COPS3 Amplification Status
Tumor site in the extremity refers to proximal or distal to the knee or elbow joint.
Chemotherapy-induced tumor necrosis data could not be obtained for 10 patients and were unavailable for 15 patients who received only postoperative chemotherapy.
Similar results were found for tumor size and amplification of PMP22 (P = .0006), NCOR1 (P = .0002), and TOM1L2 (P < .0001), which is not surprising given the strong correlation in the amplification status of the 4 genes. As well, patients with amplified TOM1L2 tended to be younger than those without amplification of this gene (median age of 15.1 years amplified, compared with 18.8 unamplified, P = .038).
Prognostic Significance of COPS3
We detected a statistically significant association between COPS3 status and the risk of metastasis, with an estimated HR of 1.61 (95% CI, 1.02–2.55) for those patients with COPS3 amplification (Fig. 2; log-rank test, P = .042). However, when included in an a priori multivariate Cox model including other clinical parameters (Table 3), the HR for COPS3 amplification decreased to 1.32 (95% CI: 0.82–2.13, P = .25), mainly due to the strong correlation with tumor size. The effect size was further reduced (HR 1.25, 95% CI, 0.78–2.02, P = .35) using an alternative model (not shown) with post-hoc definitions of tumor size and necrosis to improve model fit. As presented earlier, patients with metastatic disease at the time of diagnosis were roughly twice as likely to have COPS3 amplification. In order to determine the sensitivity of the analysis to this result, we reanalyzed the data using only the 138 patients without metastases at diagnosis, with similar results (eg, the HR for COPS3 in the univariate analysis changed from 1.61 to 1.50 in the reduced dataset, with a 95% CI of 0.88–2.54). Thus, although not ruling out the possibility of a clinically important independent effect (because the upper limits of the CIs are greater than 2), we have not observed an independent effect on disease-free survival of COPS3 amplification. In the multivariate analysis, tumor size, histologic necrosis, and patient sex all remained independent predictors of patient outcome (Table 3).
Table 3. Hazard Ratio Estimates (With P and 95% Confidence Intervals [CI]) From Multivariate Cox Regression Models With and Without COPS3 Amplification Status Included as a Covariate
Clinical variables only (n = 167)
COPS3 added to model (n = 155)
Log2 transformed, so hazard ratio estimates are for a doubling of the tumor size.
Patients with no necrosis value were grouped with the reference (≤90%) group.
In univariate analysis, neither age at diagnosis (P = .14) or tumor location (ie, proximal vs distal; P = .56) was significantly associated with the risk of metastasis. There was also a nonsignificant association between amplification and disease-free survival for the other genes, PMP22 (P = .10), NCOR1 (P = .09), and TOM1L2 (P = .06), perhaps due to limited sample size.
COPS3 Amplification and P53 Mutation Are Not Mutually Exclusive
In the tumors from our patient cohort, 141 specimens had been screened previously for mutations in the P53 gene by single-strand conformation polymorphism analysis, followed by direct sequencing.14P53 mutations were found in 27 of 141 tumors (19%) in exons 4–10 and included 17 missense and 10 nonsense alterations.
As shown in Table 2, 8 tumors exhibited both COPS3 amplification and P53 mutation, indicating that these 2 events can occur in the same tumor. There was no difference in the frequency of COPS3 amplification between those tumors with mutated P53 (8 of 27 specimens; 30%) and tumors that were wildtype for P53 (33 of 114 specimens; 29%) (OR = 1.03, P = .91). In contrast to COPS3, tumors with PMP22, NCOR1, and TOM1L2 amplification were less likely to contain P53 mutations relative to wildtype (OR = 0.28, 0.75, and 0.32 respectively), although this possible association only approached significance for TOM1L2 (P = .12).
Clinical Implication of Both COPS3 Amplification and P53 Mutation
There were 8 patients whose tumors had both COPS3 amplification and P53 mutation. We previously showed that P53 mutations do not have an effect on clinical outcome in osteosarcoma.15 However, it is interesting to note that those patients whose tumors had alterations of both P53 and COPS3 tended to have shorter disease-free survival compared with patients with COPS3 amplification alone (HR, 1.61; 95% CI, 0.63–4.12), although power was insufficient to detect an effect of this size (P = .32). The small numbers of patients whose tumors had amplification of PMP22, NCOR1, or TOM1L2 as well as a P53 mutation (1, 4, and 2, respectively) precluded meaningful statistical analysis of synergism between these genes.
Osteosarcoma is the most common nonhematologic primary malignant tumor of bone affecting children and young adults. Unfortunately, almost all osteosarcomas are high-grade and associated with a guarded prognosis. There are few clinical predictors of outcome for osteosarcoma.15, 16, 18 Patients who present with metastases at diagnosis have the worst prognosis and are rarely curable. In comparison, patients with localized tumors at presentation have a 25% to 40% risk of developing metastases after chemotherapy and radical surgery.2 Although this study confirmed that tumor size, histologic necrosis after preoperative chemotherapy, and patient sex are strong predictors of outcome, they are inaccurate for individual patients. A number of studies have been conducted to identify molecular abnormalities that might serve as therapeutic targets in osteosarcoma; however, none have had a clinical impact.1, 15, 16, 18
COPS3 was considered a potentially important molecular alteration because it targets the p53 protein for 26S proteasome degradation and increased degradation of p53 protein may result in a phenotype similar to P53 mutation.7, 11 In previous studies with small sample sizes, COPS3 had been found to be amplified in 32% to 63% of osteosarcoma specimens.3–5, 7 To determine whether amplification of COPS3 has clinical importance in osteosarcoma, we studied 155 high-grade osteosarcomas with corresponding clinical data and analyzed the association of COPS3 amplification, P53 mutation, and patient outcome.
In the present study COPS3 amplification was detected in 48 of 155 tumors (31%), similar to that observed in previous studies.4, 5, 7 Of the clinical and pathological characteristics that were analyzed only large tumor size strongly correlated with COPS3 amplification. More important, we found that COPS3 amplification was significantly correlated with poor patient outcome (P = .042). Although this association did not maintain independent significance in multivariate analysis (P = .25), the large confidence interval raises the possibility of an additional independent clinical effect of COPS3 amplification on survival. Osteosarcomas with COPS3 amplification and subsequent overexpression may have a growth advantage with a higher proliferative capacity leading to increased tumor size, and thereby acquiring additional alterations that together lead to poor patient survival. In comparison, the association of chemotherapy-induced tumor necrosis with clinical outcome is clearly independent of COPS3 amplification or other clinical factors.
Because the P53 status of the specimens examined in this study was known,15 we were able to make additional observations. We found that COPS3 amplification and P53 mutation can coexist in the same tumor, in contrast to a previous suggestion.7 There were 8 patients whose tumors had both COPS3 amplification and P53 mutation, indicating that COPS3 amplification and P53 mutation are not mutually exclusive events. In addition, those patients whose tumors had concomitant occurrence of COPS3 amplification and P53 mutation had an even shorter disease-free survival interval than patients whose tumors had either one or the other alteration alone. Multiple genetic alterations in the same tumor cell may indicate high chromosomal instability (CIN), which has been shown to correlate with invasive tumor cell behavior and poor patient outcome.19–21 The combination of both COPS3 amplification and P53 mutation could more effectively lead to CIN by driving abnormal centrosome amplification or other mechanisms.22, 23 However, the observed interaction between COPS3 and P53 was not statistically significant, likely because of the small number of patients whose tumors had both genetic alterations, and this hypothesis should be tested in a larger group.
Several cytogenetic and molecular genetic studies have reported amplification of 17p11.2 in osteosarcoma.3–5 In these studies some of the genes on the 17p11.2 amplicon were analyzed for amplification and subsequent expression. ADORA2B, DRG2, TOP3A, MAPK7 were excluded from the candidate genes list of 17p11.2 amplicon. However, COPS3, PMP22, and 3 expressed sequence tags (ESTs), AA918483, A126939, and R02360, were implicated as candidate oncogenes. AA918483 is the EST for nuclear receptor corepressor 1 (NCOR1), which mediates the transcriptional repression activity of some nuclear receptors. EST AA126939, tetratricopeptide repeat domain 19 (TTC19), encodes a protein of unknown function and R02360 is an EST from the gene target of myb1-like 2 (chicken) (TOM1L2) of unknown structure and function. These candidate genes with their relative locations on chromosome 17p11.2 are shown in Figure 1. PMP22 and NCOR1 flank the COPS3 gene, by 2 Mb and 1.1 Mb upstream, respectively, and TOM1L2 is 0.6 Mb downstream of COPS3. We also investigated the amplification status of PMP22, NCOR1, and TOM1L2 in osteosarcoma and excluded these genes as the likely driving force of this amplification unit.
In this study COPS3 was found to be the most frequently amplified gene and the amplification status of COPS3 was significantly correlated with large tumor size and poor patient outcome in osteosarcoma by univariate analysis. This is the first study to show the potential value of COPS3 amplification in high-grade osteosarcoma. Deregulation of COPS3 and P53 play important roles and together may define a subgroup of patients with a worse outcome. Thus, inhibitors of COPS3 may represent an attractive new therapeutic approach for high-grade osteosarcoma in future.
We thank our colleagues in the Department of Pathology and Laboratory Medicine, Mount Sinai Hospital, and the following physicians for osteosarcoma specimens: D. Malkin (Hospital for Sick Children, Toronto, Canada); R. Grimer (Royal Orthopaedic Hospital, Birmingham, UK); J. Healey (Memorial Sloan Kettering Cancer Center, New York, NY); and M. Rock (Mayo Clinic, Rochester, Minn).