Background
Medulloblastoma is the most common malignant brain tumor of childhood. Treatment with surgery, radiation, and chemotherapy successfully cures many patients, but survivors can suffer significant long-term toxicities affecting their neurocognitive and growth potential. Despite clinical advances, up to 30% of children with medulloblastoma experience tumor progression or recurrence, for which no curative therapy exists [
1]. The lack of more effective, less toxic therapies and the inability to stratify patients biologically result from imperfect understanding of the molecular processes that underlie medulloblastoma growth [
1‐
3].
The advent of genomic technologies has permitted a more global approach to tumor classification, diagnosis, and prognostication [
4‐
6]. Several medulloblastoma series have been analyzed for copy number aberrations (CNAs) with comparative genomic hybridization (CGH), widely considered a standard method for genome-wide screening [
7‐
12]. Using CGH, we have determined that gain of the long arm of chromosome 8 is associated with overall and progression-free survival of medulloblastoma patients.
Despite its limited resolution, chromosomal CGH can guide gene expression analysis, suggesting potential oncogenes or tumor suppressor genes involved in tumor growth. Our identification of 8q gain as clinically significant prompted the search for candidate genes mapped to that chromosomal region. Previous reports have identified MYC amplification and overexpression as clinically significant. However, we did not note such associations in our series. We investigated the possibility of other 8q-mapped candidate genes by exploiting the microarray analysis of an overlapping set of medulloblastoma specimens to complement the scale of our CGH dataset.
Of the 71 tumors we analyzed by CGH, a subset of 27 were among 64 medulloblastomas previously analyzed using gene expression microarrays, which permitted integrated analysis of both sets of data [
6]. We identified three 8q-mapped genes overexpressed in tumors with 8q gain and whose expression levels were significantly associated with overall survival in all 64 patients: eukaryotic translation elongation factor 1D (
EEF1D), ribosomal protein L30 (
RPL30), and ribosomal protein S20 (
RPS20). We corroborated the expression of each candidate gene in microarray analysis by quantitative real time-RTPCR (qRT-RTPCR). By analyzing expression microarray data and cytogenetic profiles, we have identified a group of clinically significant genes involved in translational regulation that display relatively small expression changes and previously eluded implication by single platform approaches.
Discussion
We report the results of our CGH analysis of a large series of medulloblastoma and the application of CNA-based outcome analysis of expression datasets to identify candidate genes. The relative frequencies of various CNAs are consistent with previous CGH reports [
7,
9‐
12]. The most common aberrations involved chromosome 17, including loss of 17p and gain of 17q, which were not significantly associated with outcome in our series. Several reports have studied the clinical significance of CNAs, in particular involving chromosome 17, but with discordant results. Loss of 17p correlated with survival in several series [
24‐
27]. Mendrzyk
et al [
28] and Pan
et al [
10] have described adverse outcomes associated with combined 17p loss and 17q gain, as in iso17q, respectively. Negative studies include those of Biegel
et al [
29], Emadian
et al [
30], Jung
et al [
31], and Nicholson
et al [
32]. In general, these studies evaluated fewer samples or follow-up was limited. In addition, multiple clinical factors probably contribute to discrepancies among different series.
Our CGH-based outcome analysis reveals a significant association between 8q gain (minimal common region 8q23-q24) and worse overall survival, confirmed by multivariate analysis by controlling for three known prognostic factors. We detected 8q gain in only seven cases of the 71 tumor specimens analyzed. The relatively low frequency of 8q gain probably prevented detection of its clinical significance in other series.
MYC amplification has been identified as a marker of poor prognosis, but previous studies did not further characterize cytogenetic changes involving 8q. In fact, excluding the only patient with tumor amplification at 8q23-q24 from the outcome analysis actually strengthens the association of 8q gain with overall survival. Although only a minority of specimens was available for confirmatory qRT-RTPCR quantitation, in those tumors tested,
MYC levels correlated well with microarray results.
MYC expression levels, however, were not associated with survival in our series, in contrast with previous reports [
33]. Overall, our results indicate that biological factors associated with 8q gain other than
MYC achieve clinical significance in our series of medulloblastoma patients. Due to the relatively low frequency of 8q gain, however, its prognostic significance requires confirmation in studies with larger sample size.
Our CGH and gene expression analysis indicate that other genes mapped to 8q may contribute to clinical outcome. We screened microarray profiles in an overlapping subset of tumors to identify differentially expressed candidate genes overexpressed in specimens with 8q gain. The expression of three candidate genes were associated with overall and progression-free survival: EEF1D, RPL30, and RPS20. By analyzing the larger group of 64 patients, we confirmed the association of these candidate genes with outcome. In fact, among the 64 patients, higher expression of EEF1D, RPL30, and RPS20 correlated with worse overall and progression-free survival regardless of cytogenetic profile. Since we also detected overexpression of these candidate genes in tumor samples without 8q gain, alternate mechanisms of induction may otherwise contribute to tumor phenotype and clinical outcome.
These candidate genes represent a novel group involved in translational regulation and have not been previously associated with outcome in medulloblastoma. Although we previously reported that
RPL30 expression was associated with classic histology, the smaller expression differences precluded prior detection and association with survival [
6]. In that study, we employed a two-class comparison of array profiles based on an unsupervised clustering algorithm (self-organizing map) to define a multi-gene predictor of adverse outcome. The multi-gene predictor included several other ribosomal genes, but did not include
EEF1D,
RPL30, or
RPS20. All three candidate genes are involved in ribosomal functions:
RPS20 encodes a component of the 60S ribosomal subunit, and
RPL30 part of the 40S subunit. The
EEF1D protein contributes to delivery of t-RNA to ribosomes. The overexpression of
EEF1D has been associated with advanced tumor stage in gastrointestinal carcinomas and
EEF1D reportedly displays oncogenic properties
in vitro [
34‐
36]. This supports the hypothesis that increased expression of ribosomal genes confers a growth advantage [
37].
Indeed, accumulating evidence indicates that aberrant regulation of ribosomes, their components, and their functions can be linked to cellular transformation. Several oncogenes and tumor suppressor genes, including
MYC and
MYCN, regulate the expression of rRNA and ribosomal proteins [
37]. Not surprisingly, cancer cells display increased metabolism and protein synthesis, which requires upregulated ribosomal proteins and rRNA [
37]. Since ribosomal proteins and translation factors directly regulate protein synthesis, influencing ribosomal biogenesis is one of the possible mechanisms by which cellular growth controls can be disrupted, resulting in increased proliferation. It remains unclear precisely how deregulation of rRNA and ribosomal functions are involved in tumor formation or progression. Nonetheless, our results suggest that better appreciation of the relative contributions of
EEF1D,
RPL30,
RPS20, and their associated regulatory mechanisms will impact our understanding of medulloblastoma biology. The identification of these three candidate genes indicates that specific mechanisms of ribosomal biosynthesis and translational regulation are certainly worthy of future study in medulloblastoma.
Conclusion
In summary, our CGH survey of 71 medulloblastomas indicates that the gain of 8q is adversely associated with overall and progression-free survival in patients with medulloblastoma. We applied our cytogenetic results to guide the expression analysis of a subset of twenty-seven tumors and subsequent outcome analysis in the larger group of 64 tumors. By exploiting these complementary genomic datasets, we have implicated the expression of three 8q-mapped candidate genes (EEF1D, RPL30, and RPS20) as adversely associated with overall and progression-free survival, independent of cytogenetic profile. Our results indicate the potential for hypothesis-generation by integrating these global approaches, so that more selective methods may now be applied in future studies of their contributions to tumor biology and clinical outcome.
Acknowledgements
We thank Linda L. Lin, Darlene G. Skapura, Jack M. Su, Meena Bhattacharjee, Murali Chintagumpala and Michael Sheldon for technical assistance; Cindy L. Nelson and Carolyn Pena for administrative assistance; and Angelo Rosolen for helpful discussions. This work was supported by funding from: Associazione Italiana contro le Leucemie (MDB); the Cancer Fighters of Houston, Inc. (CCL); the John S. Dunn Research Foundation; Hope Street Kids (JYHK); the Robert J. Kleberg, Jr. and Helen C. Kleberg Foundation; the Gillson Longenbaugh Foundation (CCL, JYHK); the National Children's Cancer Foundation/Children's Oncology Group Young Investigator Award (JYHK); NIH grants HD042977 (RCC), CA109467 and CA105607 (SLP), HD041648 and NS043517 (JYHK); and John and Carroll Goodman (JYHK).
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
MDB, XYL, JD, LMS, and PHR performed CGH analysis. MDB, RCC, JD, LMS, SLP, CCL, TM, and JYHK contributed to gene expression analysis using microarrays and qRT-RTPCR. MDB, RCC, AMA, LP, SLP, and JYHK analyzed tumor pathology, clinical characteristics and outcomes. MDB, JD, TM, PHR, and JYHK analyzed and integrated genomic data. All of the authors made significant contributions to data interpretation and drafting of the manuscript.