Medulloblastoma outcome is adversely associated with overexpression of EEF1D, RPL30, and RPS20 on the long arm of chromosome 8

Seventy-one medulloblastoma samples were obtained following informed consents via Institutional Review Board (IRB)-approved protocols from patients (aged 7 months – 38 years) diagnosed between 1991 and 2004 at Children's Hospital (Boston, MA) and Texas Children's Hospital (Houston, TX). Of these, 27 tumors were also previously analyzed by expression profiling using oligonucleotide microarrays [6] [Figure 1]. All specimens were obtained at the time of diagnosis prior to radiation or chemotherapy and subjected to histopathologic review according to WHO criteria [3]. Tumor samples were snap-frozen and stored in liquid nitrogen.

All patients were initially treated with maximally feasible surgical resection. Subtotal or partial resection was achieved in eleven patients. Chemotherapy for most patients (n = 50) consisted of repeated cycles of cisplatin and vincristine, with combinations of carboplatin, etoposide, cyclophosphamide and/or lomustine [13‐16]. The other twenty-one patients received intensified chemotherapy cycles (vincristine, cisplatin, cyclophosphamide, and etoposide) with autologous stem cell support [17]. Patients greater than 36 months old received craniospinal irradiation 2,400 ± 360 centiGray (cGy) with a tumor dose of 5,300 ± 720 cGy. For the entire group, median follow up was 36 months (mean 39 months, range 4 – 109). All studies were performed with the approval of the Committee for Clinical Investigation and the IRB of Boston Children's Hospital, Harvard Medical School, and the IRB of Baylor College of Medicine for Texas Children's Hospital.

Tumor DNA was extracted using the DNeasy Tissue Protocol (Qiagen, Valencia, CA). In cases of limited tissue, DNA was extracted using TRIZOL according to the manufacturer's instructions (Invitrogen, Carlsbad, CA), or from paraffin-embedded tissues using standard methods with proteinase K digestion, phenol: chloroform extraction, and ethanol precipitation [18]. Total cellular RNA from a matching portion of frozen tissue was extracted using TRIZOL or RNeasy (Qiagen). The established medulloblastoma cell line Daoy (American Type Culture Collection, Manassas, VA) was maintained to provide control RNA, as previously described [19].

Normal metaphase spreads and chromosomal CGH were performed on 71 tumor specimens as previously described [20]. Briefly, DNA from normal female human placental controls and medulloblastoma samples were labeled by nick-translation with Texas red-5-dUTP and fluorescein (FITC)-12-dUTP, respectively (Dupont NEN, Boston, MA), and co-hybridized onto normal metaphase chromosomes. Images from at least ten to fifteen separate metaphases were captured by CCD camera (SenSys Photometrics, Tucson, AZ) mounted on a Nikon Eclipse 800 microscope. The fluorescence signal ratio (FITC:Texas red) along each chromosome was calculated with Quantitative Image Processing System (Applied Imaging, Santa Clara, CA). The average of fluorescence ratios and their 99% confidence intervals were determined for each tumor specimen. Threshold ratio values greater than 1.2 and lower than 0.8 defined tumor gains and losses, respectively. High-level amplification was defined as fluorescence ratios in excess of 2.0 at the chromosomal site.

For survival analysis of each CNA, patients were stratified into two groups based on the presence or absence of the specific cytogenetic lesion. Only those lesions detected in a minimum of five patients (7%) were analyzed. Clinical outcome measures, including overall and progression-free survival, were determined from the date of diagnosis using the method of Kaplan and Meier (SPSS v.11; SPSS, Inc, Chicago, IL). The significance of survival differences between patient groups was calculated by log-rank test. To test if 8q gain was an independent prognostic factor, we also performed multivariate analysis using Cox proportional hazard model for 8q gain with respect to three other clinical prognostic factors: extent of resection, age relative to 3 years and metastatic stage at diagnosis.

Twenty-seven of the 71 tumors characterized by CGH were previously analyzed for their gene expression profiles on Affymetrix HuGeneFL microarrays and were among the series of 64 reported by Pomeroy et al. [6]. We normalized the CEL files of 64 tumors and four normal cerebellar controls. The probe set intensities used for comparisons were calculated by model-based expression index using dCHIP software (v1.3, http://​www.​dCHIP.​org) [21]. Among the 44 tumors that were collected subsequently and not included in the previous report, seven were analyzed on the Affymetrix U133 Plus 2.0 microarrays and their resulting datasets were processed similarly for direct comparison with parallel qRT-RTPCR results.

Tumor RNA was analyzed by qRT-RTPCR performed in a Bio-Rad iQ4 Multicolor Real Time iCycler (Bio-Rad Laboratories, Hercules, CA). Total cellular RNA was reverse transcribed with M-MLV Reverse Transcriptase enzyme (Invitrogen) and oligo-(dT)12, according to the manufacturer's recommendations. PCR reactions containing cDNA, iQ Syber Green Supermix (Bio-Rad), and primers for EEF1D, RPL30, RPS20, MYC, or MYCN were performed in triplicate for 40 cycles (95°C 15 sec, 60°C 1 min). Amplification products were verified by melting curves, agarose gel electrophoresis, and sequencing. Copy numbers were internally normalized to GAPDH expression, quantitated relative to the Daoy cell line as a calibrator tissue control, and accounting for differences in primer efficiencies [22]. Results from at least three separate experiments were analyzed. Fisher's exact test was used to calculate the significance of the association between gene expression levels with tumor and clinical variables. We determined correlation coefficients (R²) between expression levels detected by microarray and by qRT-RTPCR using the Pearson method (relative to the Daoy cell line as a tissue control). Primer sequences are: EEF1D, sense 5'-CCCGCGTCCGCCGATTCCTC-3' and antisense 5'-CGCTGGCGCCGTTCTCCTG-3'; RPL30 sense, 5'-TGGTGGCTGCAAAGAAGAC-3' and antisense 5'-GCAGTTGTTAGCGAGAATGAC-3'; RPS20 sense, 5'-CACGCTCCTGCTCCTGACTC-3' and antisense 5'-GCGGCTTGTTAGGCTGATTCG-3'; MYC sense, 5'-TCTGGATCACCTTCTGCTGG-3' and antisense 5'-TGTTGCTGATCTGTCTCAGG-3'; MYCN sense, 5'-AGAGGACACCCTGAGCGATT-3' and antisense 5'-TCTTGGGACGCACAGTGATG-3';GAPDH sense, 5'-AAGGTGAAGGTCGGAGTCAA-3' and antisense, 5'-AATGAAGGGGTCATTGATGG-3'.

We analyzed overall and progression-free survival for patient groups stratified by clinical variables such as age, sex, metastatic (Chang) stage, histologic subtype, and treatment regimens. Not surprisingly, metastatic disease at diagnosis displayed a trend toward worse survival [Table 1]. For other reported risk factors, only age less than three years was weakly associated with adverse clinical outcome, which could be attributed to the lack of craniospinal radiation therapy in these patients. Neither sex nor tumor histology correlated with outcome. Furthermore, different treatment centers and regimens (standard dose chemotherapy plus radiation vs. intensified chemotherapy with autologous stem cell support) were not associated with differences in survival.

The 71 medulloblastomas analyzed by chromosomal CGH displayed a variety of CNAs [Figure 2; Table 2]. The most frequently observed CNAs involved chromosome 17 (i.e. loss of the short arm, 17q gain). The loss of 17p accompanied by concurrent gain of 17q, consistent with isochromosome 17q (i17q), was detected in 32% of tumors. Losses were more frequently observed than gains and predominated on chromosomes 16, 10, 8, and 11.

We also detected gain of 2p in 22% of tumors, including 5.6% with amplification at 2p23-p24, which was the most common amplification detected in our series (n = 4). Other recurrent amplifications were detected at 2q14, 7q34, and 12p13 in two specimens each and at 8q23-q24 in one tumor. Given the resolution of chromosomal CGH (approximately 10–20 Mb), the 2p23-p24 amplifications and the 8q23-q24 amplification probably include the MYCN (2p24.1) and MYC (8q24.12-q24.13) loci, respectively. Amplifications and low copy number gains at 2p23-p24 or 8q23-q24 were not associated with large cell or anaplastic histology.

To evaluate the clinical significance of common CNAs, we stratified 71 medulloblastoma patients according to the presence or absence of specific cytogenetic lesions in their tumors. Specific CNAs involving 58 chromosomal regions were observed in a minimum of 7% tumors and used to stratify patients for Kaplan-Meier survival analysis [Table 2]. The most common lesions, including loss of 17p and/or gain of 17q, were not significantly associated with outcome.

Another common CNA, gain of 2p (minimal region 2p23-p24) detected in 22.5% of tumors (n = 16) displayed a trend towards worse overall survival (p = 0.0848) [Table 3, Figure 3]. Amplification at 2p23-p24 seen in 5.6% of tumors (n = 4) was also associated with worse overall survival (p = 0.0152). However, gain of 2p without 2p23-p24 amplification in 17% of tumors (n = 12) was not associated with outcome.

Gain of 8q (minimal region 8q23-q24, in 10% of tumors) displayed the strongest association with significantly worse overall and progression-free survival (p = 0.0141 and 0.0004, respectively) [Table 3, Figure 4]. The gain of 8q was also associated with adverse outcome in the well described high-risk group of patients were younger than three years of age at diagnosis (n = 14) (p = 0.0023). One of seven tumors with 8q gain also displayed large cell/anaplastic histology with partial loss of 8p23-q22 and amplification at 8q23-q24, which includes the MYC locus (8q24.12-q24.13). When the single case of 8q amplification is excluded from the analysis, the other six patients with gain of 8q22-q24 displayed an even stronger association with overall survival (p = 0.0005).

We performed multivariate analysis to test the significance of 8q gain with regard to outcome. When controlled for widely accepted prognostic factors (i.e. extent of resection, age relative to 3 years and metastatic stage at diagnosis), 8q gain is still significantly associated with both PFS and OS (p = 0.003 and 0.013, respectively) [see Additional file 1].

To address the potential contribution of MYC or MYCN, we screened gene expression microarray data of 27 medulloblastoma specimens that were among 64 previously reported tumors [6]. We did not discern significant upregulation of MYCN or other 2p-mapped gene expression in the seven tumors with 2p gain or amplification relative to the twenty tumors without 2p gain [Table 4]. Nor did we detect an association between MYCN expression and overall survival in the larger group of 64 medulloblastomas. We corroborated the microarray data with qRT-RTPCR in 12 primary medulloblastomas and determined that MYCN was not significantly overexpressed in tumors with gain or amplifications of 2p, compared to others (p = 0.31) [Table 4].

We also examined the possible contribution of MYC expression to the clinical significance of 8q gain. Higher levels of MYC expression correlated with slightly worse outcome among the larger group of 64 medulloblastomas, although failed to achieve statistical significance (in agreement with previous results) [6]. In addition, MYC was not consistently overexpressed in microarray-analyzed tumors with 8q gain (n = 3) [Table 4]. We quantitated MYC in twelve available tumor specimens with qRT-RTPCR and found that MYC was not significantly upregulated in the two tumors with 8q gain (t = 0.49). Not surprisingly, the one tumor with 8q23-q24 amplification displayed the highest MYC level, while low copy number gains of 8q did not correlate with higher MYC levels in either microarray- or qRT-RTPCR-tested specimens. As tested by qRT-RTPCR, relative MYC expression correlated highly with microarray results (R² = 0.99). Overall, these data suggest that MYC expression, like 8q amplification, is not entirely responsible for the association of 8q gain with survival.

In the larger group of 64 microarray-profiled tumors, the expression levels of three of five 8q-mapped genes were adversely associated with overall and progression-free survival: eukaryotic translation elongation factor 1D (EEF1D;p values, 0.014 (OS) and 0.0085 (PFS), log-rank), ribosomal protein L30 (RPL30;p values, 0.014 (OS) and 0.0018 (PFS)), and ribosomal protein S20 (RPS20;p values, 0.0005 (OS) and 0.0213 (PFS)) [Table 4; Figures 5, 6, 7]. In fact, we detected elevated levels of these genes in many tumors without 8q gain. This suggests that not only do these candidate genes contribute to outcome in patients with 8q gain tumors, but also in those regardless of their cytogenetic profiles. The expression levels of EEF1D, RPL30, and RPS20 did not correlate with MYC expression in the 64 microarray-profiled tumors.

Because the 64 tumor specimens profiled on Affymetrix HuGeneFL, were unavailable for further analysis by qRT-RTPCR, we could not validate candidate gene expression directly. Alternatively, we assayed the expression of EEF1D, RPL30, and RPS20 with qRT-RTPCR in twelve available medulloblastoma specimens, including those analyzed by Affymetrix U133 Plus 2.0 microarrays [Table 4]. To corroborate the tumor profiles, we compared the expression of each candidate gene by qRT-RTPCR with the new set of microarray-generated data on Affymetrix U133 Plus 2.0. As expected, each candidate gene demonstrated a broader range of differential expression by qRT-RTPCR than by microarray, consistent with signal compression [23]. Our qRT-RTPCR results correlated well with the expression patterns noted in microarray data (R², 0.53–0.90) [Table 4], indicating that observed expression levels detected by microarray analysis likely reflect tumor differences.